diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/AndroidManifest.xml glmark2-2021.02/android/AndroidManifest.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/AndroidManifest.xml	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/AndroidManifest.xml	2021-04-25 01:47:22.000000000 +0800
@@ -1,8 +1,11 @@
 <?xml version="1.0" encoding="utf-8"?>
 <manifest xmlns:android="http://schemas.android.com/apk/res/android"
       android:versionCode="1"
-      android:versionName="2014.03" package="org.linaro.glmark2">
-    <application android:label="@string/app_name">
+      android:versionName="2021.02" package="org.linaro.glmark2">
+    <application android:icon="@mipmap/ic_launcher"
+                 android:label="@string/app_name"
+                 android:roundIcon="@mipmap/ic_launcher_round">
+
         <activity android:name="org.linaro.glmark2.MainActivity"
                   android:label="@string/title_activity_main">
             <intent-filter>
@@ -50,5 +53,6 @@
     <uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE" />
     <uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE" />
     <uses-feature android:glEsVersion="0x00020000"/>
-    <uses-sdk android:minSdkVersion="9"/>
+    <uses-sdk android:minSdkVersion="16"/>
+    <uses-sdk android:targetSdkVersion="23"/>
 </manifest>
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/meson.build glmark2-2021.02/android/assets/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/assets/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,14 @@
+install_subdir(
+    'shaders',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
+
+install_subdir(
+    'models',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
+
+install_subdir(
+    'textures',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/bump-height.frag glmark2-2021.02/android/assets/shaders/bump-height.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/bump-height.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/bump-height.frag	2021-04-25 01:47:22.000000000 +0800
@@ -4,7 +4,12 @@
 
 uniform sampler2D HeightMap;
 
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+varying highp vec2 TextureCoord;
+#else
 varying vec2 TextureCoord;
+#endif
+
 varying vec3 NormalEye;
 varying vec3 TangentEye;
 varying vec3 BitangentEye;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/conditionals.frag glmark2-2021.02/android/assets/shaders/conditionals.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/conditionals.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/conditionals.frag	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/function.frag glmark2-2021.02/android/assets/shaders/function.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/function.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/function.frag	2021-04-25 01:47:22.000000000 +0800
@@ -8,7 +8,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/ideas-lamp-lit.frag glmark2-2021.02/android/assets/shaders/ideas-lamp-lit.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/ideas-lamp-lit.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/ideas-lamp-lit.frag	2021-04-25 01:47:22.000000000 +0800
@@ -6,7 +6,6 @@
     vec4 position;
 };
 LightSourceParameters lightSource[3];
-uniform mat4 modelview;
 uniform vec4 light0Position;
 uniform vec4 light1Position;
 uniform vec4 light2Position;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/light-refract.frag glmark2-2021.02/android/assets/shaders/light-refract.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/light-refract.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/light-refract.frag	2021-04-25 01:47:22.000000000 +0800
@@ -23,7 +23,7 @@
     vec3 mc_perspective = (MapCoord.xyz / MapCoord.w) + front_refraction;
     vec2 dcoord = mc_perspective.st * point_five + point_five;
     vec4 distance_value = texture2D(DistanceMap, dcoord);
-    vec3 back_position = vertex_position.xyz + front_refraction * distance_value.z;
+    vec3 back_position = vertex_position.xyz + front_refraction * distance_value.x;
     // Use the exit point to index the map of back-side normals, and use the
     // back-side position and normal to find the transmitted vector out of the
     // object.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/loop.frag glmark2-2021.02/android/assets/shaders/loop.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/loop.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/loop.frag	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/shadow.frag glmark2-2021.02/android/assets/shaders/shadow.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/shadow.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/shadow.frag	2021-04-25 01:47:22.000000000 +0800
@@ -8,7 +8,7 @@
     vec4 sc_perspective = ShadowCoord / ShadowCoord.w;
     sc_perspective.z += 0.1505;
     vec4 shadow_value = texture2D(ShadowMap, sc_perspective.st);
-    float light_distance = shadow_value.z;
+    float light_distance = shadow_value.x;
     float shadow = 1.0;
     if (ShadowCoord.w > 0.0 && light_distance < sc_perspective.z) {
         shadow = 0.5;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain.frag glmark2-2021.02/android/assets/shaders/terrain.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/terrain.frag	2021-04-25 01:47:22.000000000 +0800
@@ -3,9 +3,6 @@
 uniform vec3 uSpecularColor;
 uniform float uShininess;
 uniform float uOpacity;
-uniform bool enableDiffuse1;
-uniform bool enableDiffuse2;
-uniform bool enableSpecular;
 uniform sampler2D tDiffuse1;
 uniform sampler2D tDiffuse2;
 uniform sampler2D tDetail;
@@ -70,7 +67,12 @@
     vec3 pointSpecular = vec3( 0.0 );
     for ( int i = 0; i < MAX_POINT_LIGHTS; i ++ ) {
         vec4 lPosition = viewMatrix * vec4( pointLightPosition[ i ], 1.0 );
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+        // should be highp for correct behaviour if mediump is implemented as fp16
+        highp vec3 lVector = lPosition.xyz + vViewPosition.xyz;
+#else
         vec3 lVector = lPosition.xyz + vViewPosition.xyz;
+#endif
         float lDistance = 1.0;
         if ( pointLightDistance[ i ] > 0.0 )
             lDistance = 1.0 - min( ( length( lVector ) / pointLightDistance[ i ] ), 1.0 );
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain-noise.frag glmark2-2021.02/android/assets/shaders/terrain-noise.frag
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain-noise.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/terrain-noise.frag	2021-04-25 01:47:22.000000000 +0800
@@ -17,7 +17,13 @@
 uniform MEDIUMP vec2 uvScale;
 varying vec2 vUv;
 
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+// x should be passed as highp since the intermediate multiplications can
+// overflow with mediump
+vec4 permute(highp vec4 x)
+#else
 vec4 permute(vec4 x)
+#endif
 {
     return mod(((x * 34.0) + 1.0) * x, 289.0);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain.vert glmark2-2021.02/android/assets/shaders/terrain.vert
--- glmark2-2014.03+git20150611.fa71af2d/android/assets/shaders/terrain.vert	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/assets/shaders/terrain.vert	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,6 @@
 uniform mat4 modelViewMatrix;
 uniform mat4 normalMatrix;
 uniform mat4 projectionMatrix;
-uniform mat4 viewMatrix;
 
 // Vertex attributes
 attribute vec3 position;
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/assets/textures/effect-2d.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/assets/textures/effect-2d.png differ
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/build.sh glmark2-2021.02/android/build.sh
--- glmark2-2014.03+git20150611.fa71af2d/android/build.sh	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/build.sh	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,72 @@
+#!/bin/bash
+
+# Copyright 2019 The Android Open Source Project
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+if [ -z "${ANDROID_SDK}" ];
+then echo "Please set ANDROID_SDK, exiting"; exit 1;
+else echo "ANDROID_SDK is ${ANDROID_SDK}";
+fi
+
+if [ -z "${ANDROID_NDK}" ];
+then echo "Please set ANDROID_NDK, exiting"; exit 1;
+else echo "ANDROID_NDK is ${ANDROID_NDK}";
+fi
+
+if [ -z "${JAVA_HOME}" ];
+then echo "Please set JAVA_HOME, exiting"; exit 1;
+else echo "JAVA_HOME is ${JAVA_HOME}";
+fi
+
+BUILD_TOOLS_VERSION=26.0.1
+BUILD_TOOLS=$ANDROID_SDK/build-tools/$BUILD_TOOLS_VERSION
+if [ ! -d "${BUILD_TOOLS}" ];
+then echo "Please download correct build-tools version: ${BUILD_TOOLS_VERSION}, exiting"; exit 1;
+else echo "BUILD_TOOLS is ${BUILD_TOOLS}";
+fi
+
+set -ev
+
+GLMARK2_BUILD_DIR=$PWD
+GLMARK2_BASE_DIR=$GLMARK2_BUILD_DIR/..
+echo GLMARK2_BASE_DIR="${GLMARK2_BASE_DIR}"
+echo GLMARK2_BUILD_DIR="${GLMARK2_BUILD_DIR}"
+
+# Android 16 is the minSdkVersion supported
+ANDROID_JAR=$ANDROID_SDK/platforms/android-16/android.jar
+
+function create_APK() {
+    mkdir -p bin/lib obj
+    cp -r $GLMARK2_BUILD_DIR/libs/* $GLMARK2_BUILD_DIR/bin/lib
+    cp -r $GLMARK2_BASE_DIR/data $GLMARK2_BUILD_DIR/bin/assets
+    $BUILD_TOOLS/aapt package -f -m -S res -J src -M AndroidManifest.xml -I $ANDROID_JAR
+    $JAVA_HOME/bin/javac -d ./obj -source 1.7 -target 1.7 -bootclasspath $JAVA_HOME/jre/lib/rt.jar -classpath $ANDROID_JAR:obj -sourcepath src src/org/linaro/glmark2/*.java
+    $BUILD_TOOLS/dx --dex --output=bin/classes.dex ./obj
+    $BUILD_TOOLS/aapt package -f -M AndroidManifest.xml -S res -I $ANDROID_JAR -F $1-unaligned.apk bin
+    $JAVA_HOME/bin/jarsigner -verbose -keystore ~/.android/debug.keystore -storepass android -keypass android  $1-unaligned.apk androiddebugkey
+    $BUILD_TOOLS/zipalign -f 4 $1-unaligned.apk $1.apk
+}
+
+#
+# build native libraries
+#
+$ANDROID_NDK/build/ndk-build -j $cores
+
+#
+# build glmark2 APK
+#
+create_APK glmark2
+
+echo Builds succeeded
+exit 0
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/Android.mk glmark2-2021.02/android/jni/Android.mk
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/Android.mk	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/Android.mk	2021-04-25 01:47:22.000000000 +0800
@@ -54,7 +54,7 @@
 LOCAL_MODULE_TAGS := optional
 LOCAL_MODULE := libglmark2-android
 LOCAL_STATIC_LIBRARIES := libglmark2-matrix libglmark2-png libglmark2-ideas libglmark2-jpeg
-LOCAL_CFLAGS := -DGLMARK_DATA_PATH="" -DGLMARK_VERSION="\"2014.03\"" \
+LOCAL_CFLAGS := -DGLMARK_DATA_PATH="" -DGLMARK_VERSION="\"2021.02\"" \
                 -DGLMARK2_USE_GLESv2 -Werror -Wall -Wextra -Wnon-virtual-dtor \
                 -Wno-error=unused-parameter
 LOCAL_SHARED_LIBRARIES := liblog libz libEGL libGLESv2 libandroid libdl libstlport
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/Android.ndk.mk glmark2-2021.02/android/jni/Android.ndk.mk
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/Android.ndk.mk	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/Android.ndk.mk	2021-04-25 01:47:22.000000000 +0800
@@ -6,7 +6,8 @@
 LOCAL_MODULE := libglmark2-matrix
 LOCAL_CFLAGS := -DGLMARK2_USE_GLESv2 -Werror -Wall -Wextra -Wnon-virtual-dtor \
                 -Wno-error=unused-parameter
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/src
+LOCAL_C_INCLUDES := $(LOCAL_PATH)/src \
+                    $(LOCAL_PATH)/src/glad/include
 LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libmatrix/*.cc))
 
 include $(BUILD_STATIC_LIBRARY)
@@ -24,10 +25,51 @@
 LOCAL_CFLAGS := -Werror -Wall -Wextra -Wno-error=attributes \
                 -Wno-error=unused-parameter -Wno-error=unused-function -Wno-error=unused-variable
 LOCAL_C_INCLUDES := $(LOCAL_PATH)/src/libjpeg-turbo/
-LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/simd/*.c)) \
-                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/simd/*.S)) \
-                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/*.c))
-
+LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcapimin.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcapistd.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jccoefct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jccolor.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcdctmgr.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jchuff.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcinit.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcmainct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcmarker.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcmaster.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcomapi.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcparam.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcphuff.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcprepct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jcsample.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdapimin.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdapistd.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdatadst.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdatasrc.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdcoefct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdcolor.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jddctmgr.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdhuff.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdinput.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdmainct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdmarker.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdmaster.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdmerge.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdphuff.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdpostct.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jdsample.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jerror.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jfdctflt.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jfdctfst.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jfdctint.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jidctflt.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jidctfst.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jidctint.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jidctred.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jmemmgr.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jmemnobs.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jquant1.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jquant2.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jutils.c)) \
+                   $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/libjpeg-turbo/jsimd_none.c))
 include $(BUILD_STATIC_LIBRARY)
 
 include $(CLEAR_VARS)
@@ -37,26 +79,40 @@
 LOCAL_CFLAGS := -DGLMARK_DATA_PATH="" -DGLMARK2_USE_GLESv2 -Werror -Wall -Wextra\
                 -Wnon-virtual-dtor -Wno-error=unused-parameter
 LOCAL_C_INCLUDES := $(LOCAL_PATH)/src \
-                    $(LOCAL_PATH)/src/libmatrix
+                    $(LOCAL_PATH)/src/libmatrix \
+                    $(LOCAL_PATH)/src/glad/include
 LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/scene-ideas/*.cc))
+include $(BUILD_STATIC_LIBRARY)
 
+include $(CLEAR_VARS)
+LOCAL_MODULE := libglad-egl
+LOCAL_CFLAGS := -Werror -Wall
+LOCAL_C_INCLUDES := $(LOCAL_PATH)/src/glad/include
+LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/glad/src/egl.c))
 include $(BUILD_STATIC_LIBRARY)
 
 include $(CLEAR_VARS)
+LOCAL_MODULE := libglad-glesv2
+LOCAL_CFLAGS := -Werror -Wall
+LOCAL_C_INCLUDES := $(LOCAL_PATH)/src/glad/include
+LOCAL_SRC_FILES := $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/glad/src/gles2.c))
+include $(BUILD_STATIC_LIBRARY)
 
+include $(CLEAR_VARS)
 LOCAL_MODULE_TAGS := optional
 LOCAL_MODULE := libglmark2-android
-LOCAL_STATIC_LIBRARIES := libglmark2-matrix libglmark2-png libglmark2-ideas libglmark2-jpeg
-LOCAL_CFLAGS := -DGLMARK_DATA_PATH="" -DGLMARK_VERSION="\"2014.03\"" \
+LOCAL_STATIC_LIBRARIES := libglmark2-matrix libglmark2-png libglmark2-ideas libglmark2-jpeg libglad-egl libglad-glesv2
+LOCAL_CFLAGS := -DGLMARK_DATA_PATH="" -DGLMARK_VERSION="\"2021.02\"" \
                 -DGLMARK2_USE_GLESv2 -Werror -Wall -Wextra -Wnon-virtual-dtor \
                 -Wno-error=unused-parameter
-LOCAL_LDLIBS := -landroid -llog -lGLESv2 -lEGL -lz
+LOCAL_LDLIBS := -landroid -llog -lz
 LOCAL_C_INCLUDES := $(LOCAL_PATH)/src \
                     $(LOCAL_PATH)/src/libmatrix \
                     $(LOCAL_PATH)/src/scene-ideas \
                     $(LOCAL_PATH)/src/scene-terrain \
                     $(LOCAL_PATH)/src/libjpeg-turbo \
-                    $(LOCAL_PATH)/src/libpng
+                    $(LOCAL_PATH)/src/libpng \
+                    $(LOCAL_PATH)/src/glad/include
 LOCAL_SRC_FILES := $(filter-out src/canvas% src/gl-state% src/native-state% src/main.cpp, \
                      $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/*.cpp))) \
                    $(subst $(LOCAL_PATH)/,,$(wildcard $(LOCAL_PATH)/src/scene-terrain/*.cpp)) \
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/Application.mk glmark2-2021.02/android/jni/Application.mk
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/Application.mk	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/Application.mk	2021-04-25 01:47:22.000000000 +0800
@@ -1,3 +1,4 @@
-APP_STL := stlport_static
-APP_PLATFORM := android-9
+APP_ABI := all
+APP_STL := c++_static
+APP_PLATFORM := android-16
 APP_BUILD_SCRIPT := jni/Android.ndk.mk
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-android.cpp glmark2-2021.02/android/jni/src/canvas-android.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-android.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/canvas-android.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,7 +24,6 @@
 #include "log.h"
 #include "options.h"
 #include "gl-headers.h"
-#include <EGL/egl.h>
 
 #include <fstream>
 #include <sstream>
@@ -41,8 +40,26 @@
         EGL_NONE
     };
 
+    /* Init the GLAD entrypoints */
+    if (!egl_lib_.open_from_alternatives({"libEGL.so", "libEGL.so.1" })) {
+        Log::error("Error loading EGL library\n");
+        return false;
+    }
+
+    if (gladLoadEGLUserPtr(EGL_NO_DISPLAY, load_proc, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
     /* Get the current EGL config */
     EGLDisplay egl_display(eglGetCurrentDisplay());
+
+    /* Reinitialize GLAD with a known display */
+    if (gladLoadEGLUserPtr(egl_display, load_proc, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points with display failed\n");
+        return false;
+    }
+
     EGLContext egl_context(eglGetCurrentContext());
     EGLConfig egl_config(0);
     EGLint num_configs;
@@ -51,6 +68,17 @@
 
     eglChooseConfig(egl_display, attribs, &egl_config, 1, &num_configs);
 
+    /* Before calling GLES functions, init GLAD GLES */
+    if (!gles_lib_.open("libGLESv2.so")) {
+        Log::error("Error loading GLES library\n");
+        return false;
+    }
+
+    if (!gladLoadGLES2UserPtr(load_proc, &gles_lib_)) {
+        Log::error("Loading GLESv2 entry points failed.");
+        return false;
+    }
+
     resize(width_, height_);
 
     if (!eglSwapInterval(egl_display, 0))
@@ -168,6 +196,20 @@
  * Private methods *
  *******************/
 
+GLADapiproc
+CanvasAndroid::load_proc(void *userdata, const char *name)
+{
+    if (eglGetProcAddress) {
+        GLADapiproc sym = reinterpret_cast<GLADapiproc>(eglGetProcAddress(name));
+        if (sym) {
+            return sym;
+        }
+    }
+
+    SharedLibrary* lib = reinterpret_cast<SharedLibrary*>(userdata);
+    return reinterpret_cast<GLADapiproc>(lib->load(name));
+}
+
 void
 CanvasAndroid::init_gl_extensions()
 {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-android.h glmark2-2021.02/android/jni/src/canvas-android.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-android.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/canvas-android.h	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,8 @@
 #define GLMARK2_CANVAS_ANDROID_H_
 
 #include "canvas.h"
+#include "shared-library.h"
+#include "glad/egl.h"
 
 /**
  * Canvas for rendering to Android surfaces.
@@ -49,6 +51,10 @@
     void resize(int width, int height);
 
 private:
+    SharedLibrary egl_lib_;
+    SharedLibrary gles_lib_;
+
+    static GLADapiproc load_proc(void *userdata, const char *name);
     void init_gl_extensions();
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-generic.cpp glmark2-2021.02/android/jni/src/canvas-generic.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-generic.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/canvas-generic.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -103,7 +103,7 @@
     Options::FrameEnd m = Options::frame_end;
 
     if (m == Options::FrameEndDefault) {
-        if (offscreen_)
+        if (offscreen_ || Options::validate)
             m = Options::FrameEndFinish;
         else
             m = Options::FrameEndSwap;
@@ -194,7 +194,8 @@
 bool
 CanvasGeneric::supports_gl2()
 {
-    std::string gl_version_str(reinterpret_cast<const char*>(glGetString(GL_VERSION)));
+    const char *version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
+    std::string gl_version_str(version ? version : "");
     int gl_major(0);
 
     size_t point_pos(gl_version_str.find('.'));
@@ -221,7 +222,7 @@
 {
     bool request_fullscreen = (width == -1 && height == -1);
 
-    int vid;
+    intptr_t vid;
     if (!gl_state_.gotNativeConfig(vid))
     {
         Log::error("Error: Couldn't get GL visual config!\n");
@@ -249,6 +250,7 @@
     }
 
     native_window_ = native_state_.window(cur_properties);
+    window_initialized_ = true;
 
     width_ = cur_properties.width;
     height_ = cur_properties.height;
@@ -274,6 +276,12 @@
 bool
 CanvasGeneric::do_make_current()
 {
+    if (!window_initialized_) {
+        Log::error("glwindow has never been initialized, check native-state code\n"
+                   "it should not return a valid window until create_window() is called\n");
+        return false;
+    }
+
     gl_state_.init_surface(native_window_);
 
     if (!gl_state_.valid()) {
@@ -281,7 +289,8 @@
         return false;
     }
 
-    gl_state_.init_gl_extensions();
+    if (!gl_state_.init_gl_extensions())
+        return false;
 
     if (offscreen_) {
         if (!ensure_fbo())
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-generic.h glmark2-2021.02/android/jni/src/canvas-generic.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/canvas-generic.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/canvas-generic.h	2021-04-25 01:47:22.000000000 +0800
@@ -37,9 +37,10 @@
     CanvasGeneric(NativeState& native_state, GLState& gl_state,
                   int width, int height)
         : Canvas(width, height),
-          native_state_(native_state), gl_state_(gl_state),
+          native_state_(native_state), gl_state_(gl_state), native_window_(0),
           gl_color_format_(0), gl_depth_format_(0),
-          color_renderbuffer_(0), depth_renderbuffer_(0), fbo_(0) {}
+          color_renderbuffer_(0), depth_renderbuffer_(0), fbo_(0),
+          window_initialized_(false) {}
 
     bool init();
     bool reset();
@@ -70,6 +71,7 @@
     GLuint color_renderbuffer_;
     GLuint depth_renderbuffer_;
     GLuint fbo_;
+    bool window_initialized_;
 };
 
 #endif /* GLMARK2_CANVAS_GENERIC_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/EGL/eglplatform.h glmark2-2021.02/android/jni/src/glad/include/EGL/eglplatform.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/EGL/eglplatform.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/EGL/eglplatform.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,176 @@
+#ifndef __eglplatform_h_
+#define __eglplatform_h_
+
+/*
+** Copyright (c) 2007-2016 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+/* Platform-specific types and definitions for egl.h
+ * $Revision: 30994 $ on $Date: 2015-04-30 13:36:48 -0700 (Thu, 30 Apr 2015) $
+ *
+ * Adopters may modify khrplatform.h and this file to suit their platform.
+ * You are encouraged to submit all modifications to the Khronos group so that
+ * they can be included in future versions of this file.  Please submit changes
+ * by sending them to the public Khronos Bugzilla (http://khronos.org/bugzilla)
+ * by filing a bug against product "EGL" component "Registry".
+ */
+
+#include <KHR/khrplatform.h>
+
+/* Macros used in EGL function prototype declarations.
+ *
+ * EGL functions should be prototyped as:
+ *
+ * EGLAPI return-type EGLAPIENTRY eglFunction(arguments);
+ * typedef return-type (EXPAPIENTRYP PFNEGLFUNCTIONPROC) (arguments);
+ *
+ * KHRONOS_APICALL and KHRONOS_APIENTRY are defined in KHR/khrplatform.h
+ */
+
+#ifndef EGLAPI
+#define EGLAPI KHRONOS_APICALL
+#endif
+
+#ifndef EGLAPIENTRY
+#define EGLAPIENTRY  KHRONOS_APIENTRY
+#endif
+#define EGLAPIENTRYP EGLAPIENTRY*
+
+/* The types NativeDisplayType, NativeWindowType, and NativePixmapType
+ * are aliases of window-system-dependent types, such as X Display * or
+ * Windows Device Context. They must be defined in platform-specific
+ * code below. The EGL-prefixed versions of Native*Type are the same
+ * types, renamed in EGL 1.3 so all types in the API start with "EGL".
+ *
+ * Khronos STRONGLY RECOMMENDS that you use the default definitions
+ * provided below, since these changes affect both binary and source
+ * portability of applications using EGL running on different EGL
+ * implementations.
+ */
+
+#if defined(_WIN32) || defined(__VC32__) && !defined(__CYGWIN__) && !defined(__SCITECH_SNAP__) /* Win32 and WinCE */
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN 1
+#endif
+#include <windows.h>
+
+typedef HDC     EGLNativeDisplayType;
+typedef HBITMAP EGLNativePixmapType;
+typedef HWND    EGLNativeWindowType;
+
+#elif defined(__EMSCRIPTEN__)
+
+typedef int EGLNativeDisplayType;
+typedef int EGLNativePixmapType;
+typedef int EGLNativeWindowType;
+
+#elif defined(__WINSCW__) || defined(__SYMBIAN32__)  /* Symbian */
+
+typedef int   EGLNativeDisplayType;
+typedef void *EGLNativePixmapType;
+typedef void *EGLNativeWindowType;
+
+#elif defined(WL_EGL_PLATFORM)
+
+typedef struct wl_display     *EGLNativeDisplayType;
+typedef struct wl_egl_pixmap  *EGLNativePixmapType;
+typedef struct wl_egl_window  *EGLNativeWindowType;
+
+#elif defined(__GBM__)
+
+typedef struct gbm_device  *EGLNativeDisplayType;
+typedef struct gbm_bo      *EGLNativePixmapType;
+typedef void               *EGLNativeWindowType;
+
+#elif defined(__ANDROID__) || defined(ANDROID)
+
+struct ANativeWindow;
+struct egl_native_pixmap_t;
+
+typedef void*                           EGLNativeDisplayType;
+typedef struct egl_native_pixmap_t*     EGLNativePixmapType;
+typedef struct ANativeWindow*           EGLNativeWindowType;
+
+#elif defined(USE_OZONE)
+
+typedef intptr_t EGLNativeDisplayType;
+typedef intptr_t EGLNativePixmapType;
+typedef intptr_t EGLNativeWindowType;
+
+#elif defined(__unix__) && defined(EGL_NO_X11)
+
+typedef void             *EGLNativeDisplayType;
+typedef khronos_uintptr_t EGLNativePixmapType;
+typedef khronos_uintptr_t EGLNativeWindowType;
+
+#elif defined(__unix__) || defined(USE_X11)
+
+/* X11 (tentative)  */
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+
+typedef Display *EGLNativeDisplayType;
+typedef Pixmap   EGLNativePixmapType;
+typedef Window   EGLNativeWindowType;
+
+#elif defined(__APPLE__)
+
+typedef int   EGLNativeDisplayType;
+typedef void *EGLNativePixmapType;
+typedef void *EGLNativeWindowType;
+
+#elif defined(__HAIKU__)
+
+#include <kernel/image.h>
+
+typedef void              *EGLNativeDisplayType;
+typedef khronos_uintptr_t  EGLNativePixmapType;
+typedef khronos_uintptr_t  EGLNativeWindowType;
+
+#else
+#error "Platform not recognized"
+#endif
+
+/* EGL 1.2 types, renamed for consistency in EGL 1.3 */
+typedef EGLNativeDisplayType NativeDisplayType;
+typedef EGLNativePixmapType  NativePixmapType;
+typedef EGLNativeWindowType  NativeWindowType;
+
+
+/* Define EGLint. This must be a signed integral type large enough to contain
+ * all legal attribute names and values passed into and out of EGL, whether
+ * their type is boolean, bitmask, enumerant (symbolic constant), integer,
+ * handle, or other.  While in general a 32-bit integer will suffice, if
+ * handles are 64 bit types, then EGLint should be defined as a signed 64-bit
+ * integer type.
+ */
+typedef khronos_int32_t EGLint;
+
+
+/* C++ / C typecast macros for special EGL handle values */
+#if defined(__cplusplus)
+#define EGL_CAST(type, value) (static_cast<type>(value))
+#else
+#define EGL_CAST(type, value) ((type) (value))
+#endif
+
+#endif /* __eglplatform_h */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/egl.h glmark2-2021.02/android/jni/src/glad/include/glad/egl.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/egl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/glad/egl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,554 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:15 2019
+ *
+ * Generator: C/C++
+ * Specification: egl
+ * Extensions: 4
+ *
+ * APIs:
+ *  - egl=1.5
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='egl=1.5' --extensions='EGL_EXT_platform_base,EGL_KHR_platform_gbm,EGL_KHR_platform_wayland,EGL_KHR_platform_x11' c
+ *
+ * Online:
+ *    http://glad.sh/#api=egl%3D1.5&extensions=EGL_EXT_platform_base%2CEGL_KHR_platform_gbm%2CEGL_KHR_platform_wayland%2CEGL_KHR_platform_x11&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_EGL_H_
+#define GLAD_EGL_H_
+
+
+#define GLAD_EGL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define EGL_ALPHA_FORMAT 0x3088
+#define EGL_ALPHA_FORMAT_NONPRE 0x308B
+#define EGL_ALPHA_FORMAT_PRE 0x308C
+#define EGL_ALPHA_MASK_SIZE 0x303E
+#define EGL_ALPHA_SIZE 0x3021
+#define EGL_BACK_BUFFER 0x3084
+#define EGL_BAD_ACCESS 0x3002
+#define EGL_BAD_ALLOC 0x3003
+#define EGL_BAD_ATTRIBUTE 0x3004
+#define EGL_BAD_CONFIG 0x3005
+#define EGL_BAD_CONTEXT 0x3006
+#define EGL_BAD_CURRENT_SURFACE 0x3007
+#define EGL_BAD_DISPLAY 0x3008
+#define EGL_BAD_MATCH 0x3009
+#define EGL_BAD_NATIVE_PIXMAP 0x300A
+#define EGL_BAD_NATIVE_WINDOW 0x300B
+#define EGL_BAD_PARAMETER 0x300C
+#define EGL_BAD_SURFACE 0x300D
+#define EGL_BIND_TO_TEXTURE_RGB 0x3039
+#define EGL_BIND_TO_TEXTURE_RGBA 0x303A
+#define EGL_BLUE_SIZE 0x3022
+#define EGL_BUFFER_DESTROYED 0x3095
+#define EGL_BUFFER_PRESERVED 0x3094
+#define EGL_BUFFER_SIZE 0x3020
+#define EGL_CLIENT_APIS 0x308D
+#define EGL_CL_EVENT_HANDLE 0x309C
+#define EGL_COLORSPACE 0x3087
+#define EGL_COLORSPACE_LINEAR 0x308A
+#define EGL_COLORSPACE_sRGB 0x3089
+#define EGL_COLOR_BUFFER_TYPE 0x303F
+#define EGL_CONDITION_SATISFIED 0x30F6
+#define EGL_CONFIG_CAVEAT 0x3027
+#define EGL_CONFIG_ID 0x3028
+#define EGL_CONFORMANT 0x3042
+#define EGL_CONTEXT_CLIENT_TYPE 0x3097
+#define EGL_CONTEXT_CLIENT_VERSION 0x3098
+#define EGL_CONTEXT_LOST 0x300E
+#define EGL_CONTEXT_MAJOR_VERSION 0x3098
+#define EGL_CONTEXT_MINOR_VERSION 0x30FB
+#define EGL_CONTEXT_OPENGL_COMPATIBILITY_PROFILE_BIT 0x00000002
+#define EGL_CONTEXT_OPENGL_CORE_PROFILE_BIT 0x00000001
+#define EGL_CONTEXT_OPENGL_DEBUG 0x31B0
+#define EGL_CONTEXT_OPENGL_FORWARD_COMPATIBLE 0x31B1
+#define EGL_CONTEXT_OPENGL_PROFILE_MASK 0x30FD
+#define EGL_CONTEXT_OPENGL_RESET_NOTIFICATION_STRATEGY 0x31BD
+#define EGL_CONTEXT_OPENGL_ROBUST_ACCESS 0x31B2
+#define EGL_CORE_NATIVE_ENGINE 0x305B
+#define EGL_DEFAULT_DISPLAY EGL_CAST(EGLNativeDisplayType,0)
+#define EGL_DEPTH_SIZE 0x3025
+#define EGL_DISPLAY_SCALING 10000
+#define EGL_DONT_CARE EGL_CAST(EGLint,-1)
+#define EGL_DRAW 0x3059
+#define EGL_EXTENSIONS 0x3055
+#define EGL_FALSE 0
+#define EGL_FOREVER 0xFFFFFFFFFFFFFFFF
+#define EGL_GL_COLORSPACE 0x309D
+#define EGL_GL_COLORSPACE_LINEAR 0x308A
+#define EGL_GL_COLORSPACE_SRGB 0x3089
+#define EGL_GL_RENDERBUFFER 0x30B9
+#define EGL_GL_TEXTURE_2D 0x30B1
+#define EGL_GL_TEXTURE_3D 0x30B2
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x30B4
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x30B6
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x30B8
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x30B3
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x30B5
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x30B7
+#define EGL_GL_TEXTURE_LEVEL 0x30BC
+#define EGL_GL_TEXTURE_ZOFFSET 0x30BD
+#define EGL_GREEN_SIZE 0x3023
+#define EGL_HEIGHT 0x3056
+#define EGL_HORIZONTAL_RESOLUTION 0x3090
+#define EGL_IMAGE_PRESERVED 0x30D2
+#define EGL_LARGEST_PBUFFER 0x3058
+#define EGL_LEVEL 0x3029
+#define EGL_LOSE_CONTEXT_ON_RESET 0x31BF
+#define EGL_LUMINANCE_BUFFER 0x308F
+#define EGL_LUMINANCE_SIZE 0x303D
+#define EGL_MATCH_NATIVE_PIXMAP 0x3041
+#define EGL_MAX_PBUFFER_HEIGHT 0x302A
+#define EGL_MAX_PBUFFER_PIXELS 0x302B
+#define EGL_MAX_PBUFFER_WIDTH 0x302C
+#define EGL_MAX_SWAP_INTERVAL 0x303C
+#define EGL_MIN_SWAP_INTERVAL 0x303B
+#define EGL_MIPMAP_LEVEL 0x3083
+#define EGL_MIPMAP_TEXTURE 0x3082
+#define EGL_MULTISAMPLE_RESOLVE 0x3099
+#define EGL_MULTISAMPLE_RESOLVE_BOX 0x309B
+#define EGL_MULTISAMPLE_RESOLVE_BOX_BIT 0x0200
+#define EGL_MULTISAMPLE_RESOLVE_DEFAULT 0x309A
+#define EGL_NATIVE_RENDERABLE 0x302D
+#define EGL_NATIVE_VISUAL_ID 0x302E
+#define EGL_NATIVE_VISUAL_TYPE 0x302F
+#define EGL_NONE 0x3038
+#define EGL_NON_CONFORMANT_CONFIG 0x3051
+#define EGL_NOT_INITIALIZED 0x3001
+#define EGL_NO_CONTEXT EGL_CAST(EGLContext,0)
+#define EGL_NO_DISPLAY EGL_CAST(EGLDisplay,0)
+#define EGL_NO_IMAGE EGL_CAST(EGLImage,0)
+#define EGL_NO_RESET_NOTIFICATION 0x31BE
+#define EGL_NO_SURFACE EGL_CAST(EGLSurface,0)
+#define EGL_NO_SYNC EGL_CAST(EGLSync,0)
+#define EGL_NO_TEXTURE 0x305C
+#define EGL_OPENGL_API 0x30A2
+#define EGL_OPENGL_BIT 0x0008
+#define EGL_OPENGL_ES2_BIT 0x0004
+#define EGL_OPENGL_ES3_BIT 0x00000040
+#define EGL_OPENGL_ES_API 0x30A0
+#define EGL_OPENGL_ES_BIT 0x0001
+#define EGL_OPENVG_API 0x30A1
+#define EGL_OPENVG_BIT 0x0002
+#define EGL_OPENVG_IMAGE 0x3096
+#define EGL_PBUFFER_BIT 0x0001
+#define EGL_PIXEL_ASPECT_RATIO 0x3092
+#define EGL_PIXMAP_BIT 0x0002
+#define EGL_PLATFORM_GBM_KHR 0x31D7
+#define EGL_PLATFORM_WAYLAND_KHR 0x31D8
+#define EGL_PLATFORM_X11_KHR 0x31D5
+#define EGL_PLATFORM_X11_SCREEN_KHR 0x31D6
+#define EGL_READ 0x305A
+#define EGL_RED_SIZE 0x3024
+#define EGL_RENDERABLE_TYPE 0x3040
+#define EGL_RENDER_BUFFER 0x3086
+#define EGL_RGB_BUFFER 0x308E
+#define EGL_SAMPLES 0x3031
+#define EGL_SAMPLE_BUFFERS 0x3032
+#define EGL_SIGNALED 0x30F2
+#define EGL_SINGLE_BUFFER 0x3085
+#define EGL_SLOW_CONFIG 0x3050
+#define EGL_STENCIL_SIZE 0x3026
+#define EGL_SUCCESS 0x3000
+#define EGL_SURFACE_TYPE 0x3033
+#define EGL_SWAP_BEHAVIOR 0x3093
+#define EGL_SWAP_BEHAVIOR_PRESERVED_BIT 0x0400
+#define EGL_SYNC_CL_EVENT 0x30FE
+#define EGL_SYNC_CL_EVENT_COMPLETE 0x30FF
+#define EGL_SYNC_CONDITION 0x30F8
+#define EGL_SYNC_FENCE 0x30F9
+#define EGL_SYNC_FLUSH_COMMANDS_BIT 0x0001
+#define EGL_SYNC_PRIOR_COMMANDS_COMPLETE 0x30F0
+#define EGL_SYNC_STATUS 0x30F1
+#define EGL_SYNC_TYPE 0x30F7
+#define EGL_TEXTURE_2D 0x305F
+#define EGL_TEXTURE_FORMAT 0x3080
+#define EGL_TEXTURE_RGB 0x305D
+#define EGL_TEXTURE_RGBA 0x305E
+#define EGL_TEXTURE_TARGET 0x3081
+#define EGL_TIMEOUT_EXPIRED 0x30F5
+#define EGL_TRANSPARENT_BLUE_VALUE 0x3035
+#define EGL_TRANSPARENT_GREEN_VALUE 0x3036
+#define EGL_TRANSPARENT_RED_VALUE 0x3037
+#define EGL_TRANSPARENT_RGB 0x3052
+#define EGL_TRANSPARENT_TYPE 0x3034
+#define EGL_TRUE 1
+#define EGL_UNKNOWN EGL_CAST(EGLint,-1)
+#define EGL_UNSIGNALED 0x30F3
+#define EGL_VENDOR 0x3053
+#define EGL_VERSION 0x3054
+#define EGL_VERTICAL_RESOLUTION 0x3091
+#define EGL_VG_ALPHA_FORMAT 0x3088
+#define EGL_VG_ALPHA_FORMAT_NONPRE 0x308B
+#define EGL_VG_ALPHA_FORMAT_PRE 0x308C
+#define EGL_VG_ALPHA_FORMAT_PRE_BIT 0x0040
+#define EGL_VG_COLORSPACE 0x3087
+#define EGL_VG_COLORSPACE_LINEAR 0x308A
+#define EGL_VG_COLORSPACE_LINEAR_BIT 0x0020
+#define EGL_VG_COLORSPACE_sRGB 0x3089
+#define EGL_WIDTH 0x3057
+#define EGL_WINDOW_BIT 0x0004
+
+
+#include <KHR/khrplatform.h>
+#include <EGL/eglplatform.h>
+
+
+
+
+
+
+
+
+
+
+
+struct AHardwareBuffer;
+
+typedef unsigned int EGLBoolean;
+typedef unsigned int EGLenum;
+typedef intptr_t EGLAttribKHR;
+typedef intptr_t EGLAttrib;
+typedef void *EGLClientBuffer;
+typedef void *EGLConfig;
+typedef void *EGLContext;
+typedef void *EGLDeviceEXT;
+typedef void *EGLDisplay;
+typedef void *EGLImage;
+typedef void *EGLImageKHR;
+typedef void *EGLLabelKHR;
+typedef void *EGLObjectKHR;
+typedef void *EGLOutputLayerEXT;
+typedef void *EGLOutputPortEXT;
+typedef void *EGLStreamKHR;
+typedef void *EGLSurface;
+typedef void *EGLSync;
+typedef void *EGLSyncKHR;
+typedef void *EGLSyncNV;
+typedef void (*__eglMustCastToProperFunctionPointerType)(void);
+typedef khronos_utime_nanoseconds_t EGLTimeKHR;
+typedef khronos_utime_nanoseconds_t EGLTime;
+typedef khronos_utime_nanoseconds_t EGLTimeNV;
+typedef khronos_utime_nanoseconds_t EGLuint64NV;
+typedef khronos_uint64_t EGLuint64KHR;
+typedef khronos_stime_nanoseconds_t EGLnsecsANDROID;
+typedef int EGLNativeFileDescriptorKHR;
+typedef khronos_ssize_t EGLsizeiANDROID;
+typedef void (*EGLSetBlobFuncANDROID) (const void *key, EGLsizeiANDROID keySize, const void *value, EGLsizeiANDROID valueSize);
+typedef EGLsizeiANDROID (*EGLGetBlobFuncANDROID) (const void *key, EGLsizeiANDROID keySize, void *value, EGLsizeiANDROID valueSize);
+struct EGLClientPixmapHI {
+    void  *pData;
+    EGLint iWidth;
+    EGLint iHeight;
+    EGLint iStride;
+};
+typedef void (GLAD_API_PTR *EGLDEBUGPROCKHR)(EGLenum error,const char *command,EGLint messageType,EGLLabelKHR threadLabel,EGLLabelKHR objectLabel,const char* message);
+
+
+#define EGL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_0;
+#define EGL_VERSION_1_1 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_1;
+#define EGL_VERSION_1_2 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_2;
+#define EGL_VERSION_1_3 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_3;
+#define EGL_VERSION_1_4 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_4;
+#define EGL_VERSION_1_5 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_5;
+#define EGL_EXT_platform_base 1
+GLAD_API_CALL int GLAD_EGL_EXT_platform_base;
+#define EGL_KHR_platform_gbm 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_gbm;
+#define EGL_KHR_platform_wayland 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_wayland;
+#define EGL_KHR_platform_x11 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_x11;
+
+
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLBINDAPIPROC)(EGLenum api);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLBINDTEXIMAGEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLCHOOSECONFIGPROC)(EGLDisplay dpy, const EGLint * attrib_list, EGLConfig * configs, EGLint config_size, EGLint * num_config);
+typedef EGLint (GLAD_API_PTR *PFNEGLCLIENTWAITSYNCPROC)(EGLDisplay dpy, EGLSync sync, EGLint flags, EGLTime timeout);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLCOPYBUFFERSPROC)(EGLDisplay dpy, EGLSurface surface, EGLNativePixmapType target);
+typedef EGLContext (GLAD_API_PTR *PFNEGLCREATECONTEXTPROC)(EGLDisplay dpy, EGLConfig config, EGLContext share_context, const EGLint * attrib_list);
+typedef EGLImage (GLAD_API_PTR *PFNEGLCREATEIMAGEPROC)(EGLDisplay dpy, EGLContext ctx, EGLenum target, EGLClientBuffer buffer, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC)(EGLDisplay dpy, EGLenum buftype, EGLClientBuffer buffer, EGLConfig config, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPBUFFERSURFACEPROC)(EGLDisplay dpy, EGLConfig config, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPIXMAPSURFACEPROC)(EGLDisplay dpy, EGLConfig config, EGLNativePixmapType pixmap, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC)(EGLDisplay dpy, EGLConfig config, void * native_pixmap, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC)(EGLDisplay dpy, EGLConfig config, void * native_pixmap, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMWINDOWSURFACEPROC)(EGLDisplay dpy, EGLConfig config, void * native_window, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC)(EGLDisplay dpy, EGLConfig config, void * native_window, const EGLint * attrib_list);
+typedef EGLSync (GLAD_API_PTR *PFNEGLCREATESYNCPROC)(EGLDisplay dpy, EGLenum type, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEWINDOWSURFACEPROC)(EGLDisplay dpy, EGLConfig config, EGLNativeWindowType win, const EGLint * attrib_list);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYCONTEXTPROC)(EGLDisplay dpy, EGLContext ctx);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYIMAGEPROC)(EGLDisplay dpy, EGLImage image);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYSURFACEPROC)(EGLDisplay dpy, EGLSurface surface);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYSYNCPROC)(EGLDisplay dpy, EGLSync sync);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETCONFIGATTRIBPROC)(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETCONFIGSPROC)(EGLDisplay dpy, EGLConfig * configs, EGLint config_size, EGLint * num_config);
+typedef EGLContext (GLAD_API_PTR *PFNEGLGETCURRENTCONTEXTPROC)(void);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETCURRENTDISPLAYPROC)(void);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLGETCURRENTSURFACEPROC)(EGLint readdraw);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETDISPLAYPROC)(EGLNativeDisplayType display_id);
+typedef EGLint (GLAD_API_PTR *PFNEGLGETERRORPROC)(void);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETPLATFORMDISPLAYPROC)(EGLenum platform, void * native_display, const EGLAttrib * attrib_list);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETPLATFORMDISPLAYEXTPROC)(EGLenum platform, void * native_display, const EGLint * attrib_list);
+typedef __eglMustCastToProperFunctionPointerType (GLAD_API_PTR *PFNEGLGETPROCADDRESSPROC)(const char * procname);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETSYNCATTRIBPROC)(EGLDisplay dpy, EGLSync sync, EGLint attribute, EGLAttrib * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLINITIALIZEPROC)(EGLDisplay dpy, EGLint * major, EGLint * minor);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLMAKECURRENTPROC)(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx);
+typedef EGLenum (GLAD_API_PTR *PFNEGLQUERYAPIPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLQUERYCONTEXTPROC)(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint * value);
+typedef const char * (GLAD_API_PTR *PFNEGLQUERYSTRINGPROC)(EGLDisplay dpy, EGLint name);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLQUERYSURFACEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLRELEASETEXIMAGEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLRELEASETHREADPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSURFACEATTRIBPROC)(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSWAPBUFFERSPROC)(EGLDisplay dpy, EGLSurface surface);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSWAPINTERVALPROC)(EGLDisplay dpy, EGLint interval);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLTERMINATEPROC)(EGLDisplay dpy);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITCLIENTPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITGLPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITNATIVEPROC)(EGLint engine);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITSYNCPROC)(EGLDisplay dpy, EGLSync sync, EGLint flags);
+
+GLAD_API_CALL PFNEGLBINDAPIPROC glad_eglBindAPI;
+#define eglBindAPI glad_eglBindAPI
+GLAD_API_CALL PFNEGLBINDTEXIMAGEPROC glad_eglBindTexImage;
+#define eglBindTexImage glad_eglBindTexImage
+GLAD_API_CALL PFNEGLCHOOSECONFIGPROC glad_eglChooseConfig;
+#define eglChooseConfig glad_eglChooseConfig
+GLAD_API_CALL PFNEGLCLIENTWAITSYNCPROC glad_eglClientWaitSync;
+#define eglClientWaitSync glad_eglClientWaitSync
+GLAD_API_CALL PFNEGLCOPYBUFFERSPROC glad_eglCopyBuffers;
+#define eglCopyBuffers glad_eglCopyBuffers
+GLAD_API_CALL PFNEGLCREATECONTEXTPROC glad_eglCreateContext;
+#define eglCreateContext glad_eglCreateContext
+GLAD_API_CALL PFNEGLCREATEIMAGEPROC glad_eglCreateImage;
+#define eglCreateImage glad_eglCreateImage
+GLAD_API_CALL PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC glad_eglCreatePbufferFromClientBuffer;
+#define eglCreatePbufferFromClientBuffer glad_eglCreatePbufferFromClientBuffer
+GLAD_API_CALL PFNEGLCREATEPBUFFERSURFACEPROC glad_eglCreatePbufferSurface;
+#define eglCreatePbufferSurface glad_eglCreatePbufferSurface
+GLAD_API_CALL PFNEGLCREATEPIXMAPSURFACEPROC glad_eglCreatePixmapSurface;
+#define eglCreatePixmapSurface glad_eglCreatePixmapSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC glad_eglCreatePlatformPixmapSurface;
+#define eglCreatePlatformPixmapSurface glad_eglCreatePlatformPixmapSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC glad_eglCreatePlatformPixmapSurfaceEXT;
+#define eglCreatePlatformPixmapSurfaceEXT glad_eglCreatePlatformPixmapSurfaceEXT
+GLAD_API_CALL PFNEGLCREATEPLATFORMWINDOWSURFACEPROC glad_eglCreatePlatformWindowSurface;
+#define eglCreatePlatformWindowSurface glad_eglCreatePlatformWindowSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC glad_eglCreatePlatformWindowSurfaceEXT;
+#define eglCreatePlatformWindowSurfaceEXT glad_eglCreatePlatformWindowSurfaceEXT
+GLAD_API_CALL PFNEGLCREATESYNCPROC glad_eglCreateSync;
+#define eglCreateSync glad_eglCreateSync
+GLAD_API_CALL PFNEGLCREATEWINDOWSURFACEPROC glad_eglCreateWindowSurface;
+#define eglCreateWindowSurface glad_eglCreateWindowSurface
+GLAD_API_CALL PFNEGLDESTROYCONTEXTPROC glad_eglDestroyContext;
+#define eglDestroyContext glad_eglDestroyContext
+GLAD_API_CALL PFNEGLDESTROYIMAGEPROC glad_eglDestroyImage;
+#define eglDestroyImage glad_eglDestroyImage
+GLAD_API_CALL PFNEGLDESTROYSURFACEPROC glad_eglDestroySurface;
+#define eglDestroySurface glad_eglDestroySurface
+GLAD_API_CALL PFNEGLDESTROYSYNCPROC glad_eglDestroySync;
+#define eglDestroySync glad_eglDestroySync
+GLAD_API_CALL PFNEGLGETCONFIGATTRIBPROC glad_eglGetConfigAttrib;
+#define eglGetConfigAttrib glad_eglGetConfigAttrib
+GLAD_API_CALL PFNEGLGETCONFIGSPROC glad_eglGetConfigs;
+#define eglGetConfigs glad_eglGetConfigs
+GLAD_API_CALL PFNEGLGETCURRENTCONTEXTPROC glad_eglGetCurrentContext;
+#define eglGetCurrentContext glad_eglGetCurrentContext
+GLAD_API_CALL PFNEGLGETCURRENTDISPLAYPROC glad_eglGetCurrentDisplay;
+#define eglGetCurrentDisplay glad_eglGetCurrentDisplay
+GLAD_API_CALL PFNEGLGETCURRENTSURFACEPROC glad_eglGetCurrentSurface;
+#define eglGetCurrentSurface glad_eglGetCurrentSurface
+GLAD_API_CALL PFNEGLGETDISPLAYPROC glad_eglGetDisplay;
+#define eglGetDisplay glad_eglGetDisplay
+GLAD_API_CALL PFNEGLGETERRORPROC glad_eglGetError;
+#define eglGetError glad_eglGetError
+GLAD_API_CALL PFNEGLGETPLATFORMDISPLAYPROC glad_eglGetPlatformDisplay;
+#define eglGetPlatformDisplay glad_eglGetPlatformDisplay
+GLAD_API_CALL PFNEGLGETPLATFORMDISPLAYEXTPROC glad_eglGetPlatformDisplayEXT;
+#define eglGetPlatformDisplayEXT glad_eglGetPlatformDisplayEXT
+GLAD_API_CALL PFNEGLGETPROCADDRESSPROC glad_eglGetProcAddress;
+#define eglGetProcAddress glad_eglGetProcAddress
+GLAD_API_CALL PFNEGLGETSYNCATTRIBPROC glad_eglGetSyncAttrib;
+#define eglGetSyncAttrib glad_eglGetSyncAttrib
+GLAD_API_CALL PFNEGLINITIALIZEPROC glad_eglInitialize;
+#define eglInitialize glad_eglInitialize
+GLAD_API_CALL PFNEGLMAKECURRENTPROC glad_eglMakeCurrent;
+#define eglMakeCurrent glad_eglMakeCurrent
+GLAD_API_CALL PFNEGLQUERYAPIPROC glad_eglQueryAPI;
+#define eglQueryAPI glad_eglQueryAPI
+GLAD_API_CALL PFNEGLQUERYCONTEXTPROC glad_eglQueryContext;
+#define eglQueryContext glad_eglQueryContext
+GLAD_API_CALL PFNEGLQUERYSTRINGPROC glad_eglQueryString;
+#define eglQueryString glad_eglQueryString
+GLAD_API_CALL PFNEGLQUERYSURFACEPROC glad_eglQuerySurface;
+#define eglQuerySurface glad_eglQuerySurface
+GLAD_API_CALL PFNEGLRELEASETEXIMAGEPROC glad_eglReleaseTexImage;
+#define eglReleaseTexImage glad_eglReleaseTexImage
+GLAD_API_CALL PFNEGLRELEASETHREADPROC glad_eglReleaseThread;
+#define eglReleaseThread glad_eglReleaseThread
+GLAD_API_CALL PFNEGLSURFACEATTRIBPROC glad_eglSurfaceAttrib;
+#define eglSurfaceAttrib glad_eglSurfaceAttrib
+GLAD_API_CALL PFNEGLSWAPBUFFERSPROC glad_eglSwapBuffers;
+#define eglSwapBuffers glad_eglSwapBuffers
+GLAD_API_CALL PFNEGLSWAPINTERVALPROC glad_eglSwapInterval;
+#define eglSwapInterval glad_eglSwapInterval
+GLAD_API_CALL PFNEGLTERMINATEPROC glad_eglTerminate;
+#define eglTerminate glad_eglTerminate
+GLAD_API_CALL PFNEGLWAITCLIENTPROC glad_eglWaitClient;
+#define eglWaitClient glad_eglWaitClient
+GLAD_API_CALL PFNEGLWAITGLPROC glad_eglWaitGL;
+#define eglWaitGL glad_eglWaitGL
+GLAD_API_CALL PFNEGLWAITNATIVEPROC glad_eglWaitNative;
+#define eglWaitNative glad_eglWaitNative
+GLAD_API_CALL PFNEGLWAITSYNCPROC glad_eglWaitSync;
+#define eglWaitSync glad_eglWaitSync
+
+
+
+
+
+GLAD_API_CALL int gladLoadEGLUserPtr(EGLDisplay display, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadEGL(EGLDisplay display, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/gles2.h glmark2-2021.02/android/jni/src/glad/include/glad/gles2.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/gles2.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/glad/gles2.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1015 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:14 2019
+ *
+ * Generator: C/C++
+ * Specification: gl
+ * Extensions: 2
+ *
+ * APIs:
+ *  - gles2=2.0
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='gles2=2.0' --extensions='GL_OES_mapbuffer,GL_OES_required_internalformat' c
+ *
+ * Online:
+ *    http://glad.sh/#api=gles2%3D2.0&extensions=GL_OES_mapbuffer%2CGL_OES_required_internalformat&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GLES2_H_
+#define GLAD_GLES2_H_
+
+#ifdef __gl2_h_
+    #error OpenGL ES 2 header already included (API: gles2), remove previous include!
+#endif
+#define __gl2_h_ 1
+
+#ifdef __gl3_h_
+    #error OpenGL ES 3 header already included (API: gles2), remove previous include!
+#endif
+#define __gl3_h_ 1
+
+
+#define GLAD_GLES2
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GL_ACTIVE_ATTRIBUTES 0x8B89
+#define GL_ACTIVE_ATTRIBUTE_MAX_LENGTH 0x8B8A
+#define GL_ACTIVE_TEXTURE 0x84E0
+#define GL_ACTIVE_UNIFORMS 0x8B86
+#define GL_ACTIVE_UNIFORM_MAX_LENGTH 0x8B87
+#define GL_ALIASED_LINE_WIDTH_RANGE 0x846E
+#define GL_ALIASED_POINT_SIZE_RANGE 0x846D
+#define GL_ALPHA 0x1906
+#define GL_ALPHA8_OES 0x803C
+#define GL_ALPHA_BITS 0x0D55
+#define GL_ALWAYS 0x0207
+#define GL_ARRAY_BUFFER 0x8892
+#define GL_ARRAY_BUFFER_BINDING 0x8894
+#define GL_ATTACHED_SHADERS 0x8B85
+#define GL_BACK 0x0405
+#define GL_BLEND 0x0BE2
+#define GL_BLEND_COLOR 0x8005
+#define GL_BLEND_DST_ALPHA 0x80CA
+#define GL_BLEND_DST_RGB 0x80C8
+#define GL_BLEND_EQUATION 0x8009
+#define GL_BLEND_EQUATION_ALPHA 0x883D
+#define GL_BLEND_EQUATION_RGB 0x8009
+#define GL_BLEND_SRC_ALPHA 0x80CB
+#define GL_BLEND_SRC_RGB 0x80C9
+#define GL_BLUE_BITS 0x0D54
+#define GL_BOOL 0x8B56
+#define GL_BOOL_VEC2 0x8B57
+#define GL_BOOL_VEC3 0x8B58
+#define GL_BOOL_VEC4 0x8B59
+#define GL_BUFFER_ACCESS_OES 0x88BB
+#define GL_BUFFER_MAPPED_OES 0x88BC
+#define GL_BUFFER_MAP_POINTER_OES 0x88BD
+#define GL_BUFFER_SIZE 0x8764
+#define GL_BUFFER_USAGE 0x8765
+#define GL_BYTE 0x1400
+#define GL_CCW 0x0901
+#define GL_CLAMP_TO_EDGE 0x812F
+#define GL_COLOR_ATTACHMENT0 0x8CE0
+#define GL_COLOR_BUFFER_BIT 0x00004000
+#define GL_COLOR_CLEAR_VALUE 0x0C22
+#define GL_COLOR_WRITEMASK 0x0C23
+#define GL_COMPILE_STATUS 0x8B81
+#define GL_COMPRESSED_TEXTURE_FORMATS 0x86A3
+#define GL_CONSTANT_ALPHA 0x8003
+#define GL_CONSTANT_COLOR 0x8001
+#define GL_CULL_FACE 0x0B44
+#define GL_CULL_FACE_MODE 0x0B45
+#define GL_CURRENT_PROGRAM 0x8B8D
+#define GL_CURRENT_VERTEX_ATTRIB 0x8626
+#define GL_CW 0x0900
+#define GL_DECR 0x1E03
+#define GL_DECR_WRAP 0x8508
+#define GL_DELETE_STATUS 0x8B80
+#define GL_DEPTH24_STENCIL8_OES 0x88F0
+#define GL_DEPTH_ATTACHMENT 0x8D00
+#define GL_DEPTH_BITS 0x0D56
+#define GL_DEPTH_BUFFER_BIT 0x00000100
+#define GL_DEPTH_CLEAR_VALUE 0x0B73
+#define GL_DEPTH_COMPONENT 0x1902
+#define GL_DEPTH_COMPONENT16 0x81A5
+#define GL_DEPTH_COMPONENT16_OES 0x81A5
+#define GL_DEPTH_COMPONENT24_OES 0x81A6
+#define GL_DEPTH_COMPONENT32_OES 0x81A7
+#define GL_DEPTH_FUNC 0x0B74
+#define GL_DEPTH_RANGE 0x0B70
+#define GL_DEPTH_TEST 0x0B71
+#define GL_DEPTH_WRITEMASK 0x0B72
+#define GL_DITHER 0x0BD0
+#define GL_DONT_CARE 0x1100
+#define GL_DST_ALPHA 0x0304
+#define GL_DST_COLOR 0x0306
+#define GL_DYNAMIC_DRAW 0x88E8
+#define GL_ELEMENT_ARRAY_BUFFER 0x8893
+#define GL_ELEMENT_ARRAY_BUFFER_BINDING 0x8895
+#define GL_EQUAL 0x0202
+#define GL_EXTENSIONS 0x1F03
+#define GL_FALSE 0
+#define GL_FASTEST 0x1101
+#define GL_FIXED 0x140C
+#define GL_FLOAT 0x1406
+#define GL_FLOAT_MAT2 0x8B5A
+#define GL_FLOAT_MAT3 0x8B5B
+#define GL_FLOAT_MAT4 0x8B5C
+#define GL_FLOAT_VEC2 0x8B50
+#define GL_FLOAT_VEC3 0x8B51
+#define GL_FLOAT_VEC4 0x8B52
+#define GL_FRAGMENT_SHADER 0x8B30
+#define GL_FRAMEBUFFER 0x8D40
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME 0x8CD1
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE 0x8CD0
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE 0x8CD3
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL 0x8CD2
+#define GL_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_FRAMEBUFFER_COMPLETE 0x8CD5
+#define GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT 0x8CD6
+#define GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS 0x8CD9
+#define GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT 0x8CD7
+#define GL_FRAMEBUFFER_UNSUPPORTED 0x8CDD
+#define GL_FRONT 0x0404
+#define GL_FRONT_AND_BACK 0x0408
+#define GL_FRONT_FACE 0x0B46
+#define GL_FUNC_ADD 0x8006
+#define GL_FUNC_REVERSE_SUBTRACT 0x800B
+#define GL_FUNC_SUBTRACT 0x800A
+#define GL_GENERATE_MIPMAP_HINT 0x8192
+#define GL_GEQUAL 0x0206
+#define GL_GREATER 0x0204
+#define GL_GREEN_BITS 0x0D53
+#define GL_HIGH_FLOAT 0x8DF2
+#define GL_HIGH_INT 0x8DF5
+#define GL_IMPLEMENTATION_COLOR_READ_FORMAT 0x8B9B
+#define GL_IMPLEMENTATION_COLOR_READ_TYPE 0x8B9A
+#define GL_INCR 0x1E02
+#define GL_INCR_WRAP 0x8507
+#define GL_INFO_LOG_LENGTH 0x8B84
+#define GL_INT 0x1404
+#define GL_INT_VEC2 0x8B53
+#define GL_INT_VEC3 0x8B54
+#define GL_INT_VEC4 0x8B55
+#define GL_INVALID_ENUM 0x0500
+#define GL_INVALID_FRAMEBUFFER_OPERATION 0x0506
+#define GL_INVALID_OPERATION 0x0502
+#define GL_INVALID_VALUE 0x0501
+#define GL_INVERT 0x150A
+#define GL_KEEP 0x1E00
+#define GL_LEQUAL 0x0203
+#define GL_LESS 0x0201
+#define GL_LINEAR 0x2601
+#define GL_LINEAR_MIPMAP_LINEAR 0x2703
+#define GL_LINEAR_MIPMAP_NEAREST 0x2701
+#define GL_LINES 0x0001
+#define GL_LINE_LOOP 0x0002
+#define GL_LINE_STRIP 0x0003
+#define GL_LINE_WIDTH 0x0B21
+#define GL_LINK_STATUS 0x8B82
+#define GL_LOW_FLOAT 0x8DF0
+#define GL_LOW_INT 0x8DF3
+#define GL_LUMINANCE 0x1909
+#define GL_LUMINANCE4_ALPHA4_OES 0x8043
+#define GL_LUMINANCE8_ALPHA8_OES 0x8045
+#define GL_LUMINANCE8_OES 0x8040
+#define GL_LUMINANCE_ALPHA 0x190A
+#define GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS 0x8B4D
+#define GL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C
+#define GL_MAX_FRAGMENT_UNIFORM_VECTORS 0x8DFD
+#define GL_MAX_RENDERBUFFER_SIZE 0x84E8
+#define GL_MAX_TEXTURE_IMAGE_UNITS 0x8872
+#define GL_MAX_TEXTURE_SIZE 0x0D33
+#define GL_MAX_VARYING_VECTORS 0x8DFC
+#define GL_MAX_VERTEX_ATTRIBS 0x8869
+#define GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS 0x8B4C
+#define GL_MAX_VERTEX_UNIFORM_VECTORS 0x8DFB
+#define GL_MAX_VIEWPORT_DIMS 0x0D3A
+#define GL_MEDIUM_FLOAT 0x8DF1
+#define GL_MEDIUM_INT 0x8DF4
+#define GL_MIRRORED_REPEAT 0x8370
+#define GL_NEAREST 0x2600
+#define GL_NEAREST_MIPMAP_LINEAR 0x2702
+#define GL_NEAREST_MIPMAP_NEAREST 0x2700
+#define GL_NEVER 0x0200
+#define GL_NICEST 0x1102
+#define GL_NONE 0
+#define GL_NOTEQUAL 0x0205
+#define GL_NO_ERROR 0
+#define GL_NUM_COMPRESSED_TEXTURE_FORMATS 0x86A2
+#define GL_NUM_SHADER_BINARY_FORMATS 0x8DF9
+#define GL_ONE 1
+#define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004
+#define GL_ONE_MINUS_CONSTANT_COLOR 0x8002
+#define GL_ONE_MINUS_DST_ALPHA 0x0305
+#define GL_ONE_MINUS_DST_COLOR 0x0307
+#define GL_ONE_MINUS_SRC_ALPHA 0x0303
+#define GL_ONE_MINUS_SRC_COLOR 0x0301
+#define GL_OUT_OF_MEMORY 0x0505
+#define GL_PACK_ALIGNMENT 0x0D05
+#define GL_POINTS 0x0000
+#define GL_POLYGON_OFFSET_FACTOR 0x8038
+#define GL_POLYGON_OFFSET_FILL 0x8037
+#define GL_POLYGON_OFFSET_UNITS 0x2A00
+#define GL_RED_BITS 0x0D52
+#define GL_RENDERBUFFER 0x8D41
+#define GL_RENDERBUFFER_ALPHA_SIZE 0x8D53
+#define GL_RENDERBUFFER_BINDING 0x8CA7
+#define GL_RENDERBUFFER_BLUE_SIZE 0x8D52
+#define GL_RENDERBUFFER_DEPTH_SIZE 0x8D54
+#define GL_RENDERBUFFER_GREEN_SIZE 0x8D51
+#define GL_RENDERBUFFER_HEIGHT 0x8D43
+#define GL_RENDERBUFFER_INTERNAL_FORMAT 0x8D44
+#define GL_RENDERBUFFER_RED_SIZE 0x8D50
+#define GL_RENDERBUFFER_STENCIL_SIZE 0x8D55
+#define GL_RENDERBUFFER_WIDTH 0x8D42
+#define GL_RENDERER 0x1F01
+#define GL_REPEAT 0x2901
+#define GL_REPLACE 0x1E01
+#define GL_RGB 0x1907
+#define GL_RGB10_A2_EXT 0x8059
+#define GL_RGB10_EXT 0x8052
+#define GL_RGB565 0x8D62
+#define GL_RGB565_OES 0x8D62
+#define GL_RGB5_A1 0x8057
+#define GL_RGB5_A1_OES 0x8057
+#define GL_RGB8_OES 0x8051
+#define GL_RGBA 0x1908
+#define GL_RGBA4 0x8056
+#define GL_RGBA4_OES 0x8056
+#define GL_RGBA8_OES 0x8058
+#define GL_SAMPLER_2D 0x8B5E
+#define GL_SAMPLER_CUBE 0x8B60
+#define GL_SAMPLES 0x80A9
+#define GL_SAMPLE_ALPHA_TO_COVERAGE 0x809E
+#define GL_SAMPLE_BUFFERS 0x80A8
+#define GL_SAMPLE_COVERAGE 0x80A0
+#define GL_SAMPLE_COVERAGE_INVERT 0x80AB
+#define GL_SAMPLE_COVERAGE_VALUE 0x80AA
+#define GL_SCISSOR_BOX 0x0C10
+#define GL_SCISSOR_TEST 0x0C11
+#define GL_SHADER_BINARY_FORMATS 0x8DF8
+#define GL_SHADER_COMPILER 0x8DFA
+#define GL_SHADER_SOURCE_LENGTH 0x8B88
+#define GL_SHADER_TYPE 0x8B4F
+#define GL_SHADING_LANGUAGE_VERSION 0x8B8C
+#define GL_SHORT 0x1402
+#define GL_SRC_ALPHA 0x0302
+#define GL_SRC_ALPHA_SATURATE 0x0308
+#define GL_SRC_COLOR 0x0300
+#define GL_STATIC_DRAW 0x88E4
+#define GL_STENCIL_ATTACHMENT 0x8D20
+#define GL_STENCIL_BACK_FAIL 0x8801
+#define GL_STENCIL_BACK_FUNC 0x8800
+#define GL_STENCIL_BACK_PASS_DEPTH_FAIL 0x8802
+#define GL_STENCIL_BACK_PASS_DEPTH_PASS 0x8803
+#define GL_STENCIL_BACK_REF 0x8CA3
+#define GL_STENCIL_BACK_VALUE_MASK 0x8CA4
+#define GL_STENCIL_BACK_WRITEMASK 0x8CA5
+#define GL_STENCIL_BITS 0x0D57
+#define GL_STENCIL_BUFFER_BIT 0x00000400
+#define GL_STENCIL_CLEAR_VALUE 0x0B91
+#define GL_STENCIL_FAIL 0x0B94
+#define GL_STENCIL_FUNC 0x0B92
+#define GL_STENCIL_INDEX8 0x8D48
+#define GL_STENCIL_PASS_DEPTH_FAIL 0x0B95
+#define GL_STENCIL_PASS_DEPTH_PASS 0x0B96
+#define GL_STENCIL_REF 0x0B97
+#define GL_STENCIL_TEST 0x0B90
+#define GL_STENCIL_VALUE_MASK 0x0B93
+#define GL_STENCIL_WRITEMASK 0x0B98
+#define GL_STREAM_DRAW 0x88E0
+#define GL_SUBPIXEL_BITS 0x0D50
+#define GL_TEXTURE 0x1702
+#define GL_TEXTURE0 0x84C0
+#define GL_TEXTURE1 0x84C1
+#define GL_TEXTURE10 0x84CA
+#define GL_TEXTURE11 0x84CB
+#define GL_TEXTURE12 0x84CC
+#define GL_TEXTURE13 0x84CD
+#define GL_TEXTURE14 0x84CE
+#define GL_TEXTURE15 0x84CF
+#define GL_TEXTURE16 0x84D0
+#define GL_TEXTURE17 0x84D1
+#define GL_TEXTURE18 0x84D2
+#define GL_TEXTURE19 0x84D3
+#define GL_TEXTURE2 0x84C2
+#define GL_TEXTURE20 0x84D4
+#define GL_TEXTURE21 0x84D5
+#define GL_TEXTURE22 0x84D6
+#define GL_TEXTURE23 0x84D7
+#define GL_TEXTURE24 0x84D8
+#define GL_TEXTURE25 0x84D9
+#define GL_TEXTURE26 0x84DA
+#define GL_TEXTURE27 0x84DB
+#define GL_TEXTURE28 0x84DC
+#define GL_TEXTURE29 0x84DD
+#define GL_TEXTURE3 0x84C3
+#define GL_TEXTURE30 0x84DE
+#define GL_TEXTURE31 0x84DF
+#define GL_TEXTURE4 0x84C4
+#define GL_TEXTURE5 0x84C5
+#define GL_TEXTURE6 0x84C6
+#define GL_TEXTURE7 0x84C7
+#define GL_TEXTURE8 0x84C8
+#define GL_TEXTURE9 0x84C9
+#define GL_TEXTURE_2D 0x0DE1
+#define GL_TEXTURE_BINDING_2D 0x8069
+#define GL_TEXTURE_BINDING_CUBE_MAP 0x8514
+#define GL_TEXTURE_CUBE_MAP 0x8513
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519
+#define GL_TEXTURE_MAG_FILTER 0x2800
+#define GL_TEXTURE_MIN_FILTER 0x2801
+#define GL_TEXTURE_WRAP_S 0x2802
+#define GL_TEXTURE_WRAP_T 0x2803
+#define GL_TRIANGLES 0x0004
+#define GL_TRIANGLE_FAN 0x0006
+#define GL_TRIANGLE_STRIP 0x0005
+#define GL_TRUE 1
+#define GL_UNPACK_ALIGNMENT 0x0CF5
+#define GL_UNSIGNED_BYTE 0x1401
+#define GL_UNSIGNED_INT 0x1405
+#define GL_UNSIGNED_SHORT 0x1403
+#define GL_UNSIGNED_SHORT_4_4_4_4 0x8033
+#define GL_UNSIGNED_SHORT_5_5_5_1 0x8034
+#define GL_UNSIGNED_SHORT_5_6_5 0x8363
+#define GL_VALIDATE_STATUS 0x8B83
+#define GL_VENDOR 0x1F00
+#define GL_VERSION 0x1F02
+#define GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING 0x889F
+#define GL_VERTEX_ATTRIB_ARRAY_ENABLED 0x8622
+#define GL_VERTEX_ATTRIB_ARRAY_NORMALIZED 0x886A
+#define GL_VERTEX_ATTRIB_ARRAY_POINTER 0x8645
+#define GL_VERTEX_ATTRIB_ARRAY_SIZE 0x8623
+#define GL_VERTEX_ATTRIB_ARRAY_STRIDE 0x8624
+#define GL_VERTEX_ATTRIB_ARRAY_TYPE 0x8625
+#define GL_VERTEX_SHADER 0x8B31
+#define GL_VIEWPORT 0x0BA2
+#define GL_WRITE_ONLY_OES 0x88B9
+#define GL_ZERO 0
+
+
+#include <KHR/khrplatform.h>
+typedef unsigned int GLenum;
+typedef unsigned char GLboolean;
+typedef unsigned int GLbitfield;
+typedef void GLvoid;
+typedef khronos_int8_t GLbyte;
+typedef khronos_uint8_t GLubyte;
+typedef khronos_int16_t GLshort;
+typedef khronos_uint16_t GLushort;
+typedef int GLint;
+typedef unsigned int GLuint;
+typedef khronos_int32_t GLclampx;
+typedef int GLsizei;
+typedef khronos_float_t GLfloat;
+typedef khronos_float_t GLclampf;
+typedef double GLdouble;
+typedef double GLclampd;
+typedef void *GLeglClientBufferEXT;
+typedef void *GLeglImageOES;
+typedef char GLchar;
+typedef char GLcharARB;
+#ifdef __APPLE__
+typedef void *GLhandleARB;
+#else
+typedef unsigned int GLhandleARB;
+#endif
+typedef khronos_uint16_t GLhalf;
+typedef khronos_uint16_t GLhalfARB;
+typedef khronos_int32_t GLfixed;
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptr;
+#else
+typedef khronos_intptr_t GLintptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptrARB;
+#else
+typedef khronos_intptr_t GLintptrARB;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptr;
+#else
+typedef khronos_ssize_t GLsizeiptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptrARB;
+#else
+typedef khronos_ssize_t GLsizeiptrARB;
+#endif
+typedef khronos_int64_t GLint64;
+typedef khronos_int64_t GLint64EXT;
+typedef khronos_uint64_t GLuint64;
+typedef khronos_uint64_t GLuint64EXT;
+typedef struct __GLsync *GLsync;
+struct _cl_context;
+struct _cl_event;
+typedef void (GLAD_API_PTR *GLDEBUGPROC)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCARB)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCKHR)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCAMD)(GLuint id,GLenum category,GLenum severity,GLsizei length,const GLchar *message,void *userParam);
+typedef unsigned short GLhalfNV;
+typedef GLintptr GLvdpauSurfaceNV;
+typedef void (GLAD_API_PTR *GLVULKANPROCNV)(void);
+
+
+#define GL_ES_VERSION_2_0 1
+GLAD_API_CALL int GLAD_GL_ES_VERSION_2_0;
+#define GL_OES_mapbuffer 1
+GLAD_API_CALL int GLAD_GL_OES_mapbuffer;
+#define GL_OES_required_internalformat 1
+GLAD_API_CALL int GLAD_GL_OES_required_internalformat;
+
+
+typedef void (GLAD_API_PTR *PFNGLACTIVETEXTUREPROC)(GLenum texture);
+typedef void (GLAD_API_PTR *PFNGLATTACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLBINDATTRIBLOCATIONPROC)(GLuint program, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERPROC)(GLenum target, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAMEBUFFERPROC)(GLenum target, GLuint framebuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDRENDERBUFFERPROC)(GLenum target, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDTEXTUREPROC)(GLenum target, GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLBLENDCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONSEPARATEPROC)(GLenum modeRGB, GLenum modeAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCPROC)(GLenum sfactor, GLenum dfactor);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCSEPARATEPROC)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+typedef void (GLAD_API_PTR *PFNGLBUFFERDATAPROC)(GLenum target, GLsizeiptr size, const void * data, GLenum usage);
+typedef void (GLAD_API_PTR *PFNGLBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, const void * data);
+typedef GLenum (GLAD_API_PTR *PFNGLCHECKFRAMEBUFFERSTATUSPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLCLEARPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLCLEARCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARDEPTHFPROC)(GLfloat d);
+typedef void (GLAD_API_PTR *PFNGLCLEARSTENCILPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKPROC)(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
+typedef void (GLAD_API_PTR *PFNGLCOMPILESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLsizei height, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATEPROGRAMPROC)(void);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATESHADERPROC)(GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCULLFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLDELETEBUFFERSPROC)(GLsizei n, const GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEFRAMEBUFFERSPROC)(GLsizei n, const GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLDELETERENDERBUFFERSPROC)(GLsizei n, const GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDELETETEXTURESPROC)(GLsizei n, const GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLDEPTHFUNCPROC)(GLenum func);
+typedef void (GLAD_API_PTR *PFNGLDEPTHMASKPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLDEPTHRANGEFPROC)(GLfloat n, GLfloat f);
+typedef void (GLAD_API_PTR *PFNGLDETACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDISABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLDISABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSPROC)(GLenum mode, GLint first, GLsizei count);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLENABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLENABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLFINISHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERRENDERBUFFERPROC)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE2DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRONTFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLGENBUFFERSPROC)(GLsizei n, GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLGENFRAMEBUFFERSPROC)(GLsizei n, GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLGENRENDERBUFFERSPROC)(GLsizei n, GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLGENTEXTURESPROC)(GLsizei n, GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLGENERATEMIPMAPPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEATTRIBPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETATTACHEDSHADERSPROC)(GLuint program, GLsizei maxCount, GLsizei * count, GLuint * shaders);
+typedef GLint (GLAD_API_PTR *PFNGLGETATTRIBLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANVPROC)(GLenum pname, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVPROC)(GLenum target, GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVOESPROC)(GLenum target, GLenum pname, void ** params);
+typedef GLenum (GLAD_API_PTR *PFNGLGETERRORPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLGETFLOATVPROC)(GLenum pname, GLfloat * data);
+typedef void (GLAD_API_PTR *PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC)(GLenum target, GLenum attachment, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERVPROC)(GLenum pname, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMINFOLOGPROC)(GLuint program, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMIVPROC)(GLuint program, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETRENDERBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERINFOLOGPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERPRECISIONFORMATPROC)(GLenum shadertype, GLenum precisiontype, GLint * range, GLint * precision);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERSOURCEPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * source);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERIVPROC)(GLuint shader, GLenum pname, GLint * params);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGPROC)(GLenum name);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef GLint (GLAD_API_PTR *PFNGLGETUNIFORMLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMFVPROC)(GLuint program, GLint location, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMIVPROC)(GLuint program, GLint location, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBPOINTERVPROC)(GLuint index, GLenum pname, void ** pointer);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBFVPROC)(GLuint index, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLHINTPROC)(GLenum target, GLenum mode);
+typedef GLboolean (GLAD_API_PTR *PFNGLISBUFFERPROC)(GLuint buffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDPROC)(GLenum cap);
+typedef GLboolean (GLAD_API_PTR *PFNGLISFRAMEBUFFERPROC)(GLuint framebuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISPROGRAMPROC)(GLuint program);
+typedef GLboolean (GLAD_API_PTR *PFNGLISRENDERBUFFERPROC)(GLuint renderbuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSHADERPROC)(GLuint shader);
+typedef GLboolean (GLAD_API_PTR *PFNGLISTEXTUREPROC)(GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLLINEWIDTHPROC)(GLfloat width);
+typedef void (GLAD_API_PTR *PFNGLLINKPROGRAMPROC)(GLuint program);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERPROC)(GLenum target, GLenum access);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFEROESPROC)(GLenum target, GLenum access);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONOFFSETPROC)(GLfloat factor, GLfloat units);
+typedef void (GLAD_API_PTR *PFNGLREADPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLRELEASESHADERCOMPILERPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEPROC)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSAMPLECOVERAGEPROC)(GLfloat value, GLboolean invert);
+typedef void (GLAD_API_PTR *PFNGLSCISSORPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSHADERBINARYPROC)(GLsizei count, const GLuint * shaders, GLenum binaryformat, const void * binary, GLsizei length);
+typedef void (GLAD_API_PTR *PFNGLSHADERSOURCEPROC)(GLuint shader, GLsizei count, const GLchar *const* string, const GLint * length);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCPROC)(GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCSEPARATEPROC)(GLenum face, GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKSEPARATEPROC)(GLenum face, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPPROC)(GLenum fail, GLenum zfail, GLenum zpass);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPSEPARATEPROC)(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FPROC)(GLint location, GLfloat v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IPROC)(GLint location, GLint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FPROC)(GLint location, GLfloat v0, GLfloat v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IPROC)(GLint location, GLint v0, GLint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IPROC)(GLint location, GLint v0, GLint v1, GLint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2, GLfloat v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IPROC)(GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFERPROC)(GLenum target);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFEROESPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLUSEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVALIDATEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FPROC)(GLuint index, GLfloat x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FPROC)(GLuint index, GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBPOINTERPROC)(GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVIEWPORTPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+
+GLAD_API_CALL PFNGLACTIVETEXTUREPROC glad_glActiveTexture;
+#define glActiveTexture glad_glActiveTexture
+GLAD_API_CALL PFNGLATTACHSHADERPROC glad_glAttachShader;
+#define glAttachShader glad_glAttachShader
+GLAD_API_CALL PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation;
+#define glBindAttribLocation glad_glBindAttribLocation
+GLAD_API_CALL PFNGLBINDBUFFERPROC glad_glBindBuffer;
+#define glBindBuffer glad_glBindBuffer
+GLAD_API_CALL PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer;
+#define glBindFramebuffer glad_glBindFramebuffer
+GLAD_API_CALL PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer;
+#define glBindRenderbuffer glad_glBindRenderbuffer
+GLAD_API_CALL PFNGLBINDTEXTUREPROC glad_glBindTexture;
+#define glBindTexture glad_glBindTexture
+GLAD_API_CALL PFNGLBLENDCOLORPROC glad_glBlendColor;
+#define glBlendColor glad_glBlendColor
+GLAD_API_CALL PFNGLBLENDEQUATIONPROC glad_glBlendEquation;
+#define glBlendEquation glad_glBlendEquation
+GLAD_API_CALL PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate;
+#define glBlendEquationSeparate glad_glBlendEquationSeparate
+GLAD_API_CALL PFNGLBLENDFUNCPROC glad_glBlendFunc;
+#define glBlendFunc glad_glBlendFunc
+GLAD_API_CALL PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate;
+#define glBlendFuncSeparate glad_glBlendFuncSeparate
+GLAD_API_CALL PFNGLBUFFERDATAPROC glad_glBufferData;
+#define glBufferData glad_glBufferData
+GLAD_API_CALL PFNGLBUFFERSUBDATAPROC glad_glBufferSubData;
+#define glBufferSubData glad_glBufferSubData
+GLAD_API_CALL PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus;
+#define glCheckFramebufferStatus glad_glCheckFramebufferStatus
+GLAD_API_CALL PFNGLCLEARPROC glad_glClear;
+#define glClear glad_glClear
+GLAD_API_CALL PFNGLCLEARCOLORPROC glad_glClearColor;
+#define glClearColor glad_glClearColor
+GLAD_API_CALL PFNGLCLEARDEPTHFPROC glad_glClearDepthf;
+#define glClearDepthf glad_glClearDepthf
+GLAD_API_CALL PFNGLCLEARSTENCILPROC glad_glClearStencil;
+#define glClearStencil glad_glClearStencil
+GLAD_API_CALL PFNGLCOLORMASKPROC glad_glColorMask;
+#define glColorMask glad_glColorMask
+GLAD_API_CALL PFNGLCOMPILESHADERPROC glad_glCompileShader;
+#define glCompileShader glad_glCompileShader
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D;
+#define glCompressedTexImage2D glad_glCompressedTexImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D;
+#define glCompressedTexSubImage2D glad_glCompressedTexSubImage2D
+GLAD_API_CALL PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D;
+#define glCopyTexImage2D glad_glCopyTexImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D;
+#define glCopyTexSubImage2D glad_glCopyTexSubImage2D
+GLAD_API_CALL PFNGLCREATEPROGRAMPROC glad_glCreateProgram;
+#define glCreateProgram glad_glCreateProgram
+GLAD_API_CALL PFNGLCREATESHADERPROC glad_glCreateShader;
+#define glCreateShader glad_glCreateShader
+GLAD_API_CALL PFNGLCULLFACEPROC glad_glCullFace;
+#define glCullFace glad_glCullFace
+GLAD_API_CALL PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers;
+#define glDeleteBuffers glad_glDeleteBuffers
+GLAD_API_CALL PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers;
+#define glDeleteFramebuffers glad_glDeleteFramebuffers
+GLAD_API_CALL PFNGLDELETEPROGRAMPROC glad_glDeleteProgram;
+#define glDeleteProgram glad_glDeleteProgram
+GLAD_API_CALL PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers;
+#define glDeleteRenderbuffers glad_glDeleteRenderbuffers
+GLAD_API_CALL PFNGLDELETESHADERPROC glad_glDeleteShader;
+#define glDeleteShader glad_glDeleteShader
+GLAD_API_CALL PFNGLDELETETEXTURESPROC glad_glDeleteTextures;
+#define glDeleteTextures glad_glDeleteTextures
+GLAD_API_CALL PFNGLDEPTHFUNCPROC glad_glDepthFunc;
+#define glDepthFunc glad_glDepthFunc
+GLAD_API_CALL PFNGLDEPTHMASKPROC glad_glDepthMask;
+#define glDepthMask glad_glDepthMask
+GLAD_API_CALL PFNGLDEPTHRANGEFPROC glad_glDepthRangef;
+#define glDepthRangef glad_glDepthRangef
+GLAD_API_CALL PFNGLDETACHSHADERPROC glad_glDetachShader;
+#define glDetachShader glad_glDetachShader
+GLAD_API_CALL PFNGLDISABLEPROC glad_glDisable;
+#define glDisable glad_glDisable
+GLAD_API_CALL PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray;
+#define glDisableVertexAttribArray glad_glDisableVertexAttribArray
+GLAD_API_CALL PFNGLDRAWARRAYSPROC glad_glDrawArrays;
+#define glDrawArrays glad_glDrawArrays
+GLAD_API_CALL PFNGLDRAWELEMENTSPROC glad_glDrawElements;
+#define glDrawElements glad_glDrawElements
+GLAD_API_CALL PFNGLENABLEPROC glad_glEnable;
+#define glEnable glad_glEnable
+GLAD_API_CALL PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray;
+#define glEnableVertexAttribArray glad_glEnableVertexAttribArray
+GLAD_API_CALL PFNGLFINISHPROC glad_glFinish;
+#define glFinish glad_glFinish
+GLAD_API_CALL PFNGLFLUSHPROC glad_glFlush;
+#define glFlush glad_glFlush
+GLAD_API_CALL PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer;
+#define glFramebufferRenderbuffer glad_glFramebufferRenderbuffer
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D;
+#define glFramebufferTexture2D glad_glFramebufferTexture2D
+GLAD_API_CALL PFNGLFRONTFACEPROC glad_glFrontFace;
+#define glFrontFace glad_glFrontFace
+GLAD_API_CALL PFNGLGENBUFFERSPROC glad_glGenBuffers;
+#define glGenBuffers glad_glGenBuffers
+GLAD_API_CALL PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers;
+#define glGenFramebuffers glad_glGenFramebuffers
+GLAD_API_CALL PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers;
+#define glGenRenderbuffers glad_glGenRenderbuffers
+GLAD_API_CALL PFNGLGENTEXTURESPROC glad_glGenTextures;
+#define glGenTextures glad_glGenTextures
+GLAD_API_CALL PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap;
+#define glGenerateMipmap glad_glGenerateMipmap
+GLAD_API_CALL PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib;
+#define glGetActiveAttrib glad_glGetActiveAttrib
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform;
+#define glGetActiveUniform glad_glGetActiveUniform
+GLAD_API_CALL PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders;
+#define glGetAttachedShaders glad_glGetAttachedShaders
+GLAD_API_CALL PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation;
+#define glGetAttribLocation glad_glGetAttribLocation
+GLAD_API_CALL PFNGLGETBOOLEANVPROC glad_glGetBooleanv;
+#define glGetBooleanv glad_glGetBooleanv
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv;
+#define glGetBufferParameteriv glad_glGetBufferParameteriv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv;
+#define glGetBufferPointerv glad_glGetBufferPointerv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVOESPROC glad_glGetBufferPointervOES;
+#define glGetBufferPointervOES glad_glGetBufferPointervOES
+GLAD_API_CALL PFNGLGETERRORPROC glad_glGetError;
+#define glGetError glad_glGetError
+GLAD_API_CALL PFNGLGETFLOATVPROC glad_glGetFloatv;
+#define glGetFloatv glad_glGetFloatv
+GLAD_API_CALL PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv;
+#define glGetFramebufferAttachmentParameteriv glad_glGetFramebufferAttachmentParameteriv
+GLAD_API_CALL PFNGLGETINTEGERVPROC glad_glGetIntegerv;
+#define glGetIntegerv glad_glGetIntegerv
+GLAD_API_CALL PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog;
+#define glGetProgramInfoLog glad_glGetProgramInfoLog
+GLAD_API_CALL PFNGLGETPROGRAMIVPROC glad_glGetProgramiv;
+#define glGetProgramiv glad_glGetProgramiv
+GLAD_API_CALL PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv;
+#define glGetRenderbufferParameteriv glad_glGetRenderbufferParameteriv
+GLAD_API_CALL PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog;
+#define glGetShaderInfoLog glad_glGetShaderInfoLog
+GLAD_API_CALL PFNGLGETSHADERPRECISIONFORMATPROC glad_glGetShaderPrecisionFormat;
+#define glGetShaderPrecisionFormat glad_glGetShaderPrecisionFormat
+GLAD_API_CALL PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource;
+#define glGetShaderSource glad_glGetShaderSource
+GLAD_API_CALL PFNGLGETSHADERIVPROC glad_glGetShaderiv;
+#define glGetShaderiv glad_glGetShaderiv
+GLAD_API_CALL PFNGLGETSTRINGPROC glad_glGetString;
+#define glGetString glad_glGetString
+GLAD_API_CALL PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv;
+#define glGetTexParameterfv glad_glGetTexParameterfv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv;
+#define glGetTexParameteriv glad_glGetTexParameteriv
+GLAD_API_CALL PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation;
+#define glGetUniformLocation glad_glGetUniformLocation
+GLAD_API_CALL PFNGLGETUNIFORMFVPROC glad_glGetUniformfv;
+#define glGetUniformfv glad_glGetUniformfv
+GLAD_API_CALL PFNGLGETUNIFORMIVPROC glad_glGetUniformiv;
+#define glGetUniformiv glad_glGetUniformiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv;
+#define glGetVertexAttribPointerv glad_glGetVertexAttribPointerv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv;
+#define glGetVertexAttribfv glad_glGetVertexAttribfv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv;
+#define glGetVertexAttribiv glad_glGetVertexAttribiv
+GLAD_API_CALL PFNGLHINTPROC glad_glHint;
+#define glHint glad_glHint
+GLAD_API_CALL PFNGLISBUFFERPROC glad_glIsBuffer;
+#define glIsBuffer glad_glIsBuffer
+GLAD_API_CALL PFNGLISENABLEDPROC glad_glIsEnabled;
+#define glIsEnabled glad_glIsEnabled
+GLAD_API_CALL PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer;
+#define glIsFramebuffer glad_glIsFramebuffer
+GLAD_API_CALL PFNGLISPROGRAMPROC glad_glIsProgram;
+#define glIsProgram glad_glIsProgram
+GLAD_API_CALL PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer;
+#define glIsRenderbuffer glad_glIsRenderbuffer
+GLAD_API_CALL PFNGLISSHADERPROC glad_glIsShader;
+#define glIsShader glad_glIsShader
+GLAD_API_CALL PFNGLISTEXTUREPROC glad_glIsTexture;
+#define glIsTexture glad_glIsTexture
+GLAD_API_CALL PFNGLLINEWIDTHPROC glad_glLineWidth;
+#define glLineWidth glad_glLineWidth
+GLAD_API_CALL PFNGLLINKPROGRAMPROC glad_glLinkProgram;
+#define glLinkProgram glad_glLinkProgram
+GLAD_API_CALL PFNGLMAPBUFFERPROC glad_glMapBuffer;
+#define glMapBuffer glad_glMapBuffer
+GLAD_API_CALL PFNGLMAPBUFFEROESPROC glad_glMapBufferOES;
+#define glMapBufferOES glad_glMapBufferOES
+GLAD_API_CALL PFNGLPIXELSTOREIPROC glad_glPixelStorei;
+#define glPixelStorei glad_glPixelStorei
+GLAD_API_CALL PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset;
+#define glPolygonOffset glad_glPolygonOffset
+GLAD_API_CALL PFNGLREADPIXELSPROC glad_glReadPixels;
+#define glReadPixels glad_glReadPixels
+GLAD_API_CALL PFNGLRELEASESHADERCOMPILERPROC glad_glReleaseShaderCompiler;
+#define glReleaseShaderCompiler glad_glReleaseShaderCompiler
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage;
+#define glRenderbufferStorage glad_glRenderbufferStorage
+GLAD_API_CALL PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage;
+#define glSampleCoverage glad_glSampleCoverage
+GLAD_API_CALL PFNGLSCISSORPROC glad_glScissor;
+#define glScissor glad_glScissor
+GLAD_API_CALL PFNGLSHADERBINARYPROC glad_glShaderBinary;
+#define glShaderBinary glad_glShaderBinary
+GLAD_API_CALL PFNGLSHADERSOURCEPROC glad_glShaderSource;
+#define glShaderSource glad_glShaderSource
+GLAD_API_CALL PFNGLSTENCILFUNCPROC glad_glStencilFunc;
+#define glStencilFunc glad_glStencilFunc
+GLAD_API_CALL PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate;
+#define glStencilFuncSeparate glad_glStencilFuncSeparate
+GLAD_API_CALL PFNGLSTENCILMASKPROC glad_glStencilMask;
+#define glStencilMask glad_glStencilMask
+GLAD_API_CALL PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate;
+#define glStencilMaskSeparate glad_glStencilMaskSeparate
+GLAD_API_CALL PFNGLSTENCILOPPROC glad_glStencilOp;
+#define glStencilOp glad_glStencilOp
+GLAD_API_CALL PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate;
+#define glStencilOpSeparate glad_glStencilOpSeparate
+GLAD_API_CALL PFNGLTEXIMAGE2DPROC glad_glTexImage2D;
+#define glTexImage2D glad_glTexImage2D
+GLAD_API_CALL PFNGLTEXPARAMETERFPROC glad_glTexParameterf;
+#define glTexParameterf glad_glTexParameterf
+GLAD_API_CALL PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv;
+#define glTexParameterfv glad_glTexParameterfv
+GLAD_API_CALL PFNGLTEXPARAMETERIPROC glad_glTexParameteri;
+#define glTexParameteri glad_glTexParameteri
+GLAD_API_CALL PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv;
+#define glTexParameteriv glad_glTexParameteriv
+GLAD_API_CALL PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D;
+#define glTexSubImage2D glad_glTexSubImage2D
+GLAD_API_CALL PFNGLUNIFORM1FPROC glad_glUniform1f;
+#define glUniform1f glad_glUniform1f
+GLAD_API_CALL PFNGLUNIFORM1FVPROC glad_glUniform1fv;
+#define glUniform1fv glad_glUniform1fv
+GLAD_API_CALL PFNGLUNIFORM1IPROC glad_glUniform1i;
+#define glUniform1i glad_glUniform1i
+GLAD_API_CALL PFNGLUNIFORM1IVPROC glad_glUniform1iv;
+#define glUniform1iv glad_glUniform1iv
+GLAD_API_CALL PFNGLUNIFORM2FPROC glad_glUniform2f;
+#define glUniform2f glad_glUniform2f
+GLAD_API_CALL PFNGLUNIFORM2FVPROC glad_glUniform2fv;
+#define glUniform2fv glad_glUniform2fv
+GLAD_API_CALL PFNGLUNIFORM2IPROC glad_glUniform2i;
+#define glUniform2i glad_glUniform2i
+GLAD_API_CALL PFNGLUNIFORM2IVPROC glad_glUniform2iv;
+#define glUniform2iv glad_glUniform2iv
+GLAD_API_CALL PFNGLUNIFORM3FPROC glad_glUniform3f;
+#define glUniform3f glad_glUniform3f
+GLAD_API_CALL PFNGLUNIFORM3FVPROC glad_glUniform3fv;
+#define glUniform3fv glad_glUniform3fv
+GLAD_API_CALL PFNGLUNIFORM3IPROC glad_glUniform3i;
+#define glUniform3i glad_glUniform3i
+GLAD_API_CALL PFNGLUNIFORM3IVPROC glad_glUniform3iv;
+#define glUniform3iv glad_glUniform3iv
+GLAD_API_CALL PFNGLUNIFORM4FPROC glad_glUniform4f;
+#define glUniform4f glad_glUniform4f
+GLAD_API_CALL PFNGLUNIFORM4FVPROC glad_glUniform4fv;
+#define glUniform4fv glad_glUniform4fv
+GLAD_API_CALL PFNGLUNIFORM4IPROC glad_glUniform4i;
+#define glUniform4i glad_glUniform4i
+GLAD_API_CALL PFNGLUNIFORM4IVPROC glad_glUniform4iv;
+#define glUniform4iv glad_glUniform4iv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv;
+#define glUniformMatrix2fv glad_glUniformMatrix2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv;
+#define glUniformMatrix3fv glad_glUniformMatrix3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv;
+#define glUniformMatrix4fv glad_glUniformMatrix4fv
+GLAD_API_CALL PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer;
+#define glUnmapBuffer glad_glUnmapBuffer
+GLAD_API_CALL PFNGLUNMAPBUFFEROESPROC glad_glUnmapBufferOES;
+#define glUnmapBufferOES glad_glUnmapBufferOES
+GLAD_API_CALL PFNGLUSEPROGRAMPROC glad_glUseProgram;
+#define glUseProgram glad_glUseProgram
+GLAD_API_CALL PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram;
+#define glValidateProgram glad_glValidateProgram
+GLAD_API_CALL PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f;
+#define glVertexAttrib1f glad_glVertexAttrib1f
+GLAD_API_CALL PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv;
+#define glVertexAttrib1fv glad_glVertexAttrib1fv
+GLAD_API_CALL PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f;
+#define glVertexAttrib2f glad_glVertexAttrib2f
+GLAD_API_CALL PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv;
+#define glVertexAttrib2fv glad_glVertexAttrib2fv
+GLAD_API_CALL PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f;
+#define glVertexAttrib3f glad_glVertexAttrib3f
+GLAD_API_CALL PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv;
+#define glVertexAttrib3fv glad_glVertexAttrib3fv
+GLAD_API_CALL PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f;
+#define glVertexAttrib4f glad_glVertexAttrib4f
+GLAD_API_CALL PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv;
+#define glVertexAttrib4fv glad_glVertexAttrib4fv
+GLAD_API_CALL PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer;
+#define glVertexAttribPointer glad_glVertexAttribPointer
+GLAD_API_CALL PFNGLVIEWPORTPROC glad_glViewport;
+#define glViewport glad_glViewport
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLES2UserPtr( GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGLES2( GLADloadfunc load);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/gl.h glmark2-2021.02/android/jni/src/glad/include/glad/gl.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/gl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/glad/gl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3686 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:14 2019
+ *
+ * Generator: C/C++
+ * Specification: gl
+ * Extensions: 0
+ *
+ * APIs:
+ *  - gl:compatibility=3.3
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='gl:compatibility=3.3' --extensions='' c
+ *
+ * Online:
+ *    http://glad.sh/#api=gl%3Acompatibility%3D3.3&extensions=&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GL_H_
+#define GLAD_GL_H_
+
+#ifdef __gl_h_
+    #error OpenGL header already included (API: gl), remove previous include!
+#endif
+#define __gl_h_ 1
+
+
+#define GLAD_GL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GL_2D 0x0600
+#define GL_2_BYTES 0x1407
+#define GL_3D 0x0601
+#define GL_3D_COLOR 0x0602
+#define GL_3D_COLOR_TEXTURE 0x0603
+#define GL_3_BYTES 0x1408
+#define GL_4D_COLOR_TEXTURE 0x0604
+#define GL_4_BYTES 0x1409
+#define GL_ACCUM 0x0100
+#define GL_ACCUM_ALPHA_BITS 0x0D5B
+#define GL_ACCUM_BLUE_BITS 0x0D5A
+#define GL_ACCUM_BUFFER_BIT 0x00000200
+#define GL_ACCUM_CLEAR_VALUE 0x0B80
+#define GL_ACCUM_GREEN_BITS 0x0D59
+#define GL_ACCUM_RED_BITS 0x0D58
+#define GL_ACTIVE_ATTRIBUTES 0x8B89
+#define GL_ACTIVE_ATTRIBUTE_MAX_LENGTH 0x8B8A
+#define GL_ACTIVE_TEXTURE 0x84E0
+#define GL_ACTIVE_UNIFORMS 0x8B86
+#define GL_ACTIVE_UNIFORM_BLOCKS 0x8A36
+#define GL_ACTIVE_UNIFORM_BLOCK_MAX_NAME_LENGTH 0x8A35
+#define GL_ACTIVE_UNIFORM_MAX_LENGTH 0x8B87
+#define GL_ADD 0x0104
+#define GL_ADD_SIGNED 0x8574
+#define GL_ALIASED_LINE_WIDTH_RANGE 0x846E
+#define GL_ALIASED_POINT_SIZE_RANGE 0x846D
+#define GL_ALL_ATTRIB_BITS 0xFFFFFFFF
+#define GL_ALPHA 0x1906
+#define GL_ALPHA12 0x803D
+#define GL_ALPHA16 0x803E
+#define GL_ALPHA4 0x803B
+#define GL_ALPHA8 0x803C
+#define GL_ALPHA_BIAS 0x0D1D
+#define GL_ALPHA_BITS 0x0D55
+#define GL_ALPHA_INTEGER 0x8D97
+#define GL_ALPHA_SCALE 0x0D1C
+#define GL_ALPHA_TEST 0x0BC0
+#define GL_ALPHA_TEST_FUNC 0x0BC1
+#define GL_ALPHA_TEST_REF 0x0BC2
+#define GL_ALREADY_SIGNALED 0x911A
+#define GL_ALWAYS 0x0207
+#define GL_AMBIENT 0x1200
+#define GL_AMBIENT_AND_DIFFUSE 0x1602
+#define GL_AND 0x1501
+#define GL_AND_INVERTED 0x1504
+#define GL_AND_REVERSE 0x1502
+#define GL_ANY_SAMPLES_PASSED 0x8C2F
+#define GL_ARRAY_BUFFER 0x8892
+#define GL_ARRAY_BUFFER_BINDING 0x8894
+#define GL_ATTACHED_SHADERS 0x8B85
+#define GL_ATTRIB_STACK_DEPTH 0x0BB0
+#define GL_AUTO_NORMAL 0x0D80
+#define GL_AUX0 0x0409
+#define GL_AUX1 0x040A
+#define GL_AUX2 0x040B
+#define GL_AUX3 0x040C
+#define GL_AUX_BUFFERS 0x0C00
+#define GL_BACK 0x0405
+#define GL_BACK_LEFT 0x0402
+#define GL_BACK_RIGHT 0x0403
+#define GL_BGR 0x80E0
+#define GL_BGRA 0x80E1
+#define GL_BGRA_INTEGER 0x8D9B
+#define GL_BGR_INTEGER 0x8D9A
+#define GL_BITMAP 0x1A00
+#define GL_BITMAP_TOKEN 0x0704
+#define GL_BLEND 0x0BE2
+#define GL_BLEND_COLOR 0x8005
+#define GL_BLEND_DST 0x0BE0
+#define GL_BLEND_DST_ALPHA 0x80CA
+#define GL_BLEND_DST_RGB 0x80C8
+#define GL_BLEND_EQUATION 0x8009
+#define GL_BLEND_EQUATION_ALPHA 0x883D
+#define GL_BLEND_EQUATION_RGB 0x8009
+#define GL_BLEND_SRC 0x0BE1
+#define GL_BLEND_SRC_ALPHA 0x80CB
+#define GL_BLEND_SRC_RGB 0x80C9
+#define GL_BLUE 0x1905
+#define GL_BLUE_BIAS 0x0D1B
+#define GL_BLUE_BITS 0x0D54
+#define GL_BLUE_INTEGER 0x8D96
+#define GL_BLUE_SCALE 0x0D1A
+#define GL_BOOL 0x8B56
+#define GL_BOOL_VEC2 0x8B57
+#define GL_BOOL_VEC3 0x8B58
+#define GL_BOOL_VEC4 0x8B59
+#define GL_BUFFER_ACCESS 0x88BB
+#define GL_BUFFER_ACCESS_FLAGS 0x911F
+#define GL_BUFFER_MAPPED 0x88BC
+#define GL_BUFFER_MAP_LENGTH 0x9120
+#define GL_BUFFER_MAP_OFFSET 0x9121
+#define GL_BUFFER_MAP_POINTER 0x88BD
+#define GL_BUFFER_SIZE 0x8764
+#define GL_BUFFER_USAGE 0x8765
+#define GL_BYTE 0x1400
+#define GL_C3F_V3F 0x2A24
+#define GL_C4F_N3F_V3F 0x2A26
+#define GL_C4UB_V2F 0x2A22
+#define GL_C4UB_V3F 0x2A23
+#define GL_CCW 0x0901
+#define GL_CLAMP 0x2900
+#define GL_CLAMP_FRAGMENT_COLOR 0x891B
+#define GL_CLAMP_READ_COLOR 0x891C
+#define GL_CLAMP_TO_BORDER 0x812D
+#define GL_CLAMP_TO_EDGE 0x812F
+#define GL_CLAMP_VERTEX_COLOR 0x891A
+#define GL_CLEAR 0x1500
+#define GL_CLIENT_ACTIVE_TEXTURE 0x84E1
+#define GL_CLIENT_ALL_ATTRIB_BITS 0xFFFFFFFF
+#define GL_CLIENT_ATTRIB_STACK_DEPTH 0x0BB1
+#define GL_CLIENT_PIXEL_STORE_BIT 0x00000001
+#define GL_CLIENT_VERTEX_ARRAY_BIT 0x00000002
+#define GL_CLIP_DISTANCE0 0x3000
+#define GL_CLIP_DISTANCE1 0x3001
+#define GL_CLIP_DISTANCE2 0x3002
+#define GL_CLIP_DISTANCE3 0x3003
+#define GL_CLIP_DISTANCE4 0x3004
+#define GL_CLIP_DISTANCE5 0x3005
+#define GL_CLIP_DISTANCE6 0x3006
+#define GL_CLIP_DISTANCE7 0x3007
+#define GL_CLIP_PLANE0 0x3000
+#define GL_CLIP_PLANE1 0x3001
+#define GL_CLIP_PLANE2 0x3002
+#define GL_CLIP_PLANE3 0x3003
+#define GL_CLIP_PLANE4 0x3004
+#define GL_CLIP_PLANE5 0x3005
+#define GL_COEFF 0x0A00
+#define GL_COLOR 0x1800
+#define GL_COLOR_ARRAY 0x8076
+#define GL_COLOR_ARRAY_BUFFER_BINDING 0x8898
+#define GL_COLOR_ARRAY_POINTER 0x8090
+#define GL_COLOR_ARRAY_SIZE 0x8081
+#define GL_COLOR_ARRAY_STRIDE 0x8083
+#define GL_COLOR_ARRAY_TYPE 0x8082
+#define GL_COLOR_ATTACHMENT0 0x8CE0
+#define GL_COLOR_ATTACHMENT1 0x8CE1
+#define GL_COLOR_ATTACHMENT10 0x8CEA
+#define GL_COLOR_ATTACHMENT11 0x8CEB
+#define GL_COLOR_ATTACHMENT12 0x8CEC
+#define GL_COLOR_ATTACHMENT13 0x8CED
+#define GL_COLOR_ATTACHMENT14 0x8CEE
+#define GL_COLOR_ATTACHMENT15 0x8CEF
+#define GL_COLOR_ATTACHMENT16 0x8CF0
+#define GL_COLOR_ATTACHMENT17 0x8CF1
+#define GL_COLOR_ATTACHMENT18 0x8CF2
+#define GL_COLOR_ATTACHMENT19 0x8CF3
+#define GL_COLOR_ATTACHMENT2 0x8CE2
+#define GL_COLOR_ATTACHMENT20 0x8CF4
+#define GL_COLOR_ATTACHMENT21 0x8CF5
+#define GL_COLOR_ATTACHMENT22 0x8CF6
+#define GL_COLOR_ATTACHMENT23 0x8CF7
+#define GL_COLOR_ATTACHMENT24 0x8CF8
+#define GL_COLOR_ATTACHMENT25 0x8CF9
+#define GL_COLOR_ATTACHMENT26 0x8CFA
+#define GL_COLOR_ATTACHMENT27 0x8CFB
+#define GL_COLOR_ATTACHMENT28 0x8CFC
+#define GL_COLOR_ATTACHMENT29 0x8CFD
+#define GL_COLOR_ATTACHMENT3 0x8CE3
+#define GL_COLOR_ATTACHMENT30 0x8CFE
+#define GL_COLOR_ATTACHMENT31 0x8CFF
+#define GL_COLOR_ATTACHMENT4 0x8CE4
+#define GL_COLOR_ATTACHMENT5 0x8CE5
+#define GL_COLOR_ATTACHMENT6 0x8CE6
+#define GL_COLOR_ATTACHMENT7 0x8CE7
+#define GL_COLOR_ATTACHMENT8 0x8CE8
+#define GL_COLOR_ATTACHMENT9 0x8CE9
+#define GL_COLOR_BUFFER_BIT 0x00004000
+#define GL_COLOR_CLEAR_VALUE 0x0C22
+#define GL_COLOR_INDEX 0x1900
+#define GL_COLOR_INDEXES 0x1603
+#define GL_COLOR_LOGIC_OP 0x0BF2
+#define GL_COLOR_MATERIAL 0x0B57
+#define GL_COLOR_MATERIAL_FACE 0x0B55
+#define GL_COLOR_MATERIAL_PARAMETER 0x0B56
+#define GL_COLOR_SUM 0x8458
+#define GL_COLOR_WRITEMASK 0x0C23
+#define GL_COMBINE 0x8570
+#define GL_COMBINE_ALPHA 0x8572
+#define GL_COMBINE_RGB 0x8571
+#define GL_COMPARE_REF_TO_TEXTURE 0x884E
+#define GL_COMPARE_R_TO_TEXTURE 0x884E
+#define GL_COMPILE 0x1300
+#define GL_COMPILE_AND_EXECUTE 0x1301
+#define GL_COMPILE_STATUS 0x8B81
+#define GL_COMPRESSED_ALPHA 0x84E9
+#define GL_COMPRESSED_INTENSITY 0x84EC
+#define GL_COMPRESSED_LUMINANCE 0x84EA
+#define GL_COMPRESSED_LUMINANCE_ALPHA 0x84EB
+#define GL_COMPRESSED_RED 0x8225
+#define GL_COMPRESSED_RED_RGTC1 0x8DBB
+#define GL_COMPRESSED_RG 0x8226
+#define GL_COMPRESSED_RGB 0x84ED
+#define GL_COMPRESSED_RGBA 0x84EE
+#define GL_COMPRESSED_RG_RGTC2 0x8DBD
+#define GL_COMPRESSED_SIGNED_RED_RGTC1 0x8DBC
+#define GL_COMPRESSED_SIGNED_RG_RGTC2 0x8DBE
+#define GL_COMPRESSED_SLUMINANCE 0x8C4A
+#define GL_COMPRESSED_SLUMINANCE_ALPHA 0x8C4B
+#define GL_COMPRESSED_SRGB 0x8C48
+#define GL_COMPRESSED_SRGB_ALPHA 0x8C49
+#define GL_COMPRESSED_TEXTURE_FORMATS 0x86A3
+#define GL_CONDITION_SATISFIED 0x911C
+#define GL_CONSTANT 0x8576
+#define GL_CONSTANT_ALPHA 0x8003
+#define GL_CONSTANT_ATTENUATION 0x1207
+#define GL_CONSTANT_COLOR 0x8001
+#define GL_CONTEXT_COMPATIBILITY_PROFILE_BIT 0x00000002
+#define GL_CONTEXT_CORE_PROFILE_BIT 0x00000001
+#define GL_CONTEXT_FLAGS 0x821E
+#define GL_CONTEXT_FLAG_FORWARD_COMPATIBLE_BIT 0x00000001
+#define GL_CONTEXT_PROFILE_MASK 0x9126
+#define GL_COORD_REPLACE 0x8862
+#define GL_COPY 0x1503
+#define GL_COPY_INVERTED 0x150C
+#define GL_COPY_PIXEL_TOKEN 0x0706
+#define GL_COPY_READ_BUFFER 0x8F36
+#define GL_COPY_WRITE_BUFFER 0x8F37
+#define GL_CULL_FACE 0x0B44
+#define GL_CULL_FACE_MODE 0x0B45
+#define GL_CURRENT_BIT 0x00000001
+#define GL_CURRENT_COLOR 0x0B00
+#define GL_CURRENT_FOG_COORD 0x8453
+#define GL_CURRENT_FOG_COORDINATE 0x8453
+#define GL_CURRENT_INDEX 0x0B01
+#define GL_CURRENT_NORMAL 0x0B02
+#define GL_CURRENT_PROGRAM 0x8B8D
+#define GL_CURRENT_QUERY 0x8865
+#define GL_CURRENT_RASTER_COLOR 0x0B04
+#define GL_CURRENT_RASTER_DISTANCE 0x0B09
+#define GL_CURRENT_RASTER_INDEX 0x0B05
+#define GL_CURRENT_RASTER_POSITION 0x0B07
+#define GL_CURRENT_RASTER_POSITION_VALID 0x0B08
+#define GL_CURRENT_RASTER_SECONDARY_COLOR 0x845F
+#define GL_CURRENT_RASTER_TEXTURE_COORDS 0x0B06
+#define GL_CURRENT_SECONDARY_COLOR 0x8459
+#define GL_CURRENT_TEXTURE_COORDS 0x0B03
+#define GL_CURRENT_VERTEX_ATTRIB 0x8626
+#define GL_CW 0x0900
+#define GL_DECAL 0x2101
+#define GL_DECR 0x1E03
+#define GL_DECR_WRAP 0x8508
+#define GL_DELETE_STATUS 0x8B80
+#define GL_DEPTH 0x1801
+#define GL_DEPTH24_STENCIL8 0x88F0
+#define GL_DEPTH32F_STENCIL8 0x8CAD
+#define GL_DEPTH_ATTACHMENT 0x8D00
+#define GL_DEPTH_BIAS 0x0D1F
+#define GL_DEPTH_BITS 0x0D56
+#define GL_DEPTH_BUFFER_BIT 0x00000100
+#define GL_DEPTH_CLAMP 0x864F
+#define GL_DEPTH_CLEAR_VALUE 0x0B73
+#define GL_DEPTH_COMPONENT 0x1902
+#define GL_DEPTH_COMPONENT16 0x81A5
+#define GL_DEPTH_COMPONENT24 0x81A6
+#define GL_DEPTH_COMPONENT32 0x81A7
+#define GL_DEPTH_COMPONENT32F 0x8CAC
+#define GL_DEPTH_FUNC 0x0B74
+#define GL_DEPTH_RANGE 0x0B70
+#define GL_DEPTH_SCALE 0x0D1E
+#define GL_DEPTH_STENCIL 0x84F9
+#define GL_DEPTH_STENCIL_ATTACHMENT 0x821A
+#define GL_DEPTH_TEST 0x0B71
+#define GL_DEPTH_TEXTURE_MODE 0x884B
+#define GL_DEPTH_WRITEMASK 0x0B72
+#define GL_DIFFUSE 0x1201
+#define GL_DITHER 0x0BD0
+#define GL_DOMAIN 0x0A02
+#define GL_DONT_CARE 0x1100
+#define GL_DOT3_RGB 0x86AE
+#define GL_DOT3_RGBA 0x86AF
+#define GL_DOUBLE 0x140A
+#define GL_DOUBLEBUFFER 0x0C32
+#define GL_DRAW_BUFFER 0x0C01
+#define GL_DRAW_BUFFER0 0x8825
+#define GL_DRAW_BUFFER1 0x8826
+#define GL_DRAW_BUFFER10 0x882F
+#define GL_DRAW_BUFFER11 0x8830
+#define GL_DRAW_BUFFER12 0x8831
+#define GL_DRAW_BUFFER13 0x8832
+#define GL_DRAW_BUFFER14 0x8833
+#define GL_DRAW_BUFFER15 0x8834
+#define GL_DRAW_BUFFER2 0x8827
+#define GL_DRAW_BUFFER3 0x8828
+#define GL_DRAW_BUFFER4 0x8829
+#define GL_DRAW_BUFFER5 0x882A
+#define GL_DRAW_BUFFER6 0x882B
+#define GL_DRAW_BUFFER7 0x882C
+#define GL_DRAW_BUFFER8 0x882D
+#define GL_DRAW_BUFFER9 0x882E
+#define GL_DRAW_FRAMEBUFFER 0x8CA9
+#define GL_DRAW_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_DRAW_PIXEL_TOKEN 0x0705
+#define GL_DST_ALPHA 0x0304
+#define GL_DST_COLOR 0x0306
+#define GL_DYNAMIC_COPY 0x88EA
+#define GL_DYNAMIC_DRAW 0x88E8
+#define GL_DYNAMIC_READ 0x88E9
+#define GL_EDGE_FLAG 0x0B43
+#define GL_EDGE_FLAG_ARRAY 0x8079
+#define GL_EDGE_FLAG_ARRAY_BUFFER_BINDING 0x889B
+#define GL_EDGE_FLAG_ARRAY_POINTER 0x8093
+#define GL_EDGE_FLAG_ARRAY_STRIDE 0x808C
+#define GL_ELEMENT_ARRAY_BUFFER 0x8893
+#define GL_ELEMENT_ARRAY_BUFFER_BINDING 0x8895
+#define GL_EMISSION 0x1600
+#define GL_ENABLE_BIT 0x00002000
+#define GL_EQUAL 0x0202
+#define GL_EQUIV 0x1509
+#define GL_EVAL_BIT 0x00010000
+#define GL_EXP 0x0800
+#define GL_EXP2 0x0801
+#define GL_EXTENSIONS 0x1F03
+#define GL_EYE_LINEAR 0x2400
+#define GL_EYE_PLANE 0x2502
+#define GL_FALSE 0
+#define GL_FASTEST 0x1101
+#define GL_FEEDBACK 0x1C01
+#define GL_FEEDBACK_BUFFER_POINTER 0x0DF0
+#define GL_FEEDBACK_BUFFER_SIZE 0x0DF1
+#define GL_FEEDBACK_BUFFER_TYPE 0x0DF2
+#define GL_FILL 0x1B02
+#define GL_FIRST_VERTEX_CONVENTION 0x8E4D
+#define GL_FIXED_ONLY 0x891D
+#define GL_FLAT 0x1D00
+#define GL_FLOAT 0x1406
+#define GL_FLOAT_32_UNSIGNED_INT_24_8_REV 0x8DAD
+#define GL_FLOAT_MAT2 0x8B5A
+#define GL_FLOAT_MAT2x3 0x8B65
+#define GL_FLOAT_MAT2x4 0x8B66
+#define GL_FLOAT_MAT3 0x8B5B
+#define GL_FLOAT_MAT3x2 0x8B67
+#define GL_FLOAT_MAT3x4 0x8B68
+#define GL_FLOAT_MAT4 0x8B5C
+#define GL_FLOAT_MAT4x2 0x8B69
+#define GL_FLOAT_MAT4x3 0x8B6A
+#define GL_FLOAT_VEC2 0x8B50
+#define GL_FLOAT_VEC3 0x8B51
+#define GL_FLOAT_VEC4 0x8B52
+#define GL_FOG 0x0B60
+#define GL_FOG_BIT 0x00000080
+#define GL_FOG_COLOR 0x0B66
+#define GL_FOG_COORD 0x8451
+#define GL_FOG_COORDINATE 0x8451
+#define GL_FOG_COORDINATE_ARRAY 0x8457
+#define GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING 0x889D
+#define GL_FOG_COORDINATE_ARRAY_POINTER 0x8456
+#define GL_FOG_COORDINATE_ARRAY_STRIDE 0x8455
+#define GL_FOG_COORDINATE_ARRAY_TYPE 0x8454
+#define GL_FOG_COORDINATE_SOURCE 0x8450
+#define GL_FOG_COORD_ARRAY 0x8457
+#define GL_FOG_COORD_ARRAY_BUFFER_BINDING 0x889D
+#define GL_FOG_COORD_ARRAY_POINTER 0x8456
+#define GL_FOG_COORD_ARRAY_STRIDE 0x8455
+#define GL_FOG_COORD_ARRAY_TYPE 0x8454
+#define GL_FOG_COORD_SRC 0x8450
+#define GL_FOG_DENSITY 0x0B62
+#define GL_FOG_END 0x0B64
+#define GL_FOG_HINT 0x0C54
+#define GL_FOG_INDEX 0x0B61
+#define GL_FOG_MODE 0x0B65
+#define GL_FOG_START 0x0B63
+#define GL_FRAGMENT_DEPTH 0x8452
+#define GL_FRAGMENT_SHADER 0x8B30
+#define GL_FRAGMENT_SHADER_DERIVATIVE_HINT 0x8B8B
+#define GL_FRAMEBUFFER 0x8D40
+#define GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE 0x8215
+#define GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE 0x8214
+#define GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING 0x8210
+#define GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE 0x8211
+#define GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE 0x8216
+#define GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE 0x8213
+#define GL_FRAMEBUFFER_ATTACHMENT_LAYERED 0x8DA7
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME 0x8CD1
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE 0x8CD0
+#define GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE 0x8212
+#define GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE 0x8217
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE 0x8CD3
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER 0x8CD4
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL 0x8CD2
+#define GL_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_FRAMEBUFFER_COMPLETE 0x8CD5
+#define GL_FRAMEBUFFER_DEFAULT 0x8218
+#define GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT 0x8CD6
+#define GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER 0x8CDB
+#define GL_FRAMEBUFFER_INCOMPLETE_LAYER_TARGETS 0x8DA8
+#define GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT 0x8CD7
+#define GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE 0x8D56
+#define GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER 0x8CDC
+#define GL_FRAMEBUFFER_SRGB 0x8DB9
+#define GL_FRAMEBUFFER_UNDEFINED 0x8219
+#define GL_FRAMEBUFFER_UNSUPPORTED 0x8CDD
+#define GL_FRONT 0x0404
+#define GL_FRONT_AND_BACK 0x0408
+#define GL_FRONT_FACE 0x0B46
+#define GL_FRONT_LEFT 0x0400
+#define GL_FRONT_RIGHT 0x0401
+#define GL_FUNC_ADD 0x8006
+#define GL_FUNC_REVERSE_SUBTRACT 0x800B
+#define GL_FUNC_SUBTRACT 0x800A
+#define GL_GENERATE_MIPMAP 0x8191
+#define GL_GENERATE_MIPMAP_HINT 0x8192
+#define GL_GEOMETRY_INPUT_TYPE 0x8917
+#define GL_GEOMETRY_OUTPUT_TYPE 0x8918
+#define GL_GEOMETRY_SHADER 0x8DD9
+#define GL_GEOMETRY_VERTICES_OUT 0x8916
+#define GL_GEQUAL 0x0206
+#define GL_GREATER 0x0204
+#define GL_GREEN 0x1904
+#define GL_GREEN_BIAS 0x0D19
+#define GL_GREEN_BITS 0x0D53
+#define GL_GREEN_INTEGER 0x8D95
+#define GL_GREEN_SCALE 0x0D18
+#define GL_HALF_FLOAT 0x140B
+#define GL_HINT_BIT 0x00008000
+#define GL_INCR 0x1E02
+#define GL_INCR_WRAP 0x8507
+#define GL_INDEX 0x8222
+#define GL_INDEX_ARRAY 0x8077
+#define GL_INDEX_ARRAY_BUFFER_BINDING 0x8899
+#define GL_INDEX_ARRAY_POINTER 0x8091
+#define GL_INDEX_ARRAY_STRIDE 0x8086
+#define GL_INDEX_ARRAY_TYPE 0x8085
+#define GL_INDEX_BITS 0x0D51
+#define GL_INDEX_CLEAR_VALUE 0x0C20
+#define GL_INDEX_LOGIC_OP 0x0BF1
+#define GL_INDEX_MODE 0x0C30
+#define GL_INDEX_OFFSET 0x0D13
+#define GL_INDEX_SHIFT 0x0D12
+#define GL_INDEX_WRITEMASK 0x0C21
+#define GL_INFO_LOG_LENGTH 0x8B84
+#define GL_INT 0x1404
+#define GL_INTENSITY 0x8049
+#define GL_INTENSITY12 0x804C
+#define GL_INTENSITY16 0x804D
+#define GL_INTENSITY4 0x804A
+#define GL_INTENSITY8 0x804B
+#define GL_INTERLEAVED_ATTRIBS 0x8C8C
+#define GL_INTERPOLATE 0x8575
+#define GL_INT_2_10_10_10_REV 0x8D9F
+#define GL_INT_SAMPLER_1D 0x8DC9
+#define GL_INT_SAMPLER_1D_ARRAY 0x8DCE
+#define GL_INT_SAMPLER_2D 0x8DCA
+#define GL_INT_SAMPLER_2D_ARRAY 0x8DCF
+#define GL_INT_SAMPLER_2D_MULTISAMPLE 0x9109
+#define GL_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910C
+#define GL_INT_SAMPLER_2D_RECT 0x8DCD
+#define GL_INT_SAMPLER_3D 0x8DCB
+#define GL_INT_SAMPLER_BUFFER 0x8DD0
+#define GL_INT_SAMPLER_CUBE 0x8DCC
+#define GL_INT_VEC2 0x8B53
+#define GL_INT_VEC3 0x8B54
+#define GL_INT_VEC4 0x8B55
+#define GL_INVALID_ENUM 0x0500
+#define GL_INVALID_FRAMEBUFFER_OPERATION 0x0506
+#define GL_INVALID_INDEX 0xFFFFFFFF
+#define GL_INVALID_OPERATION 0x0502
+#define GL_INVALID_VALUE 0x0501
+#define GL_INVERT 0x150A
+#define GL_KEEP 0x1E00
+#define GL_LAST_VERTEX_CONVENTION 0x8E4E
+#define GL_LEFT 0x0406
+#define GL_LEQUAL 0x0203
+#define GL_LESS 0x0201
+#define GL_LIGHT0 0x4000
+#define GL_LIGHT1 0x4001
+#define GL_LIGHT2 0x4002
+#define GL_LIGHT3 0x4003
+#define GL_LIGHT4 0x4004
+#define GL_LIGHT5 0x4005
+#define GL_LIGHT6 0x4006
+#define GL_LIGHT7 0x4007
+#define GL_LIGHTING 0x0B50
+#define GL_LIGHTING_BIT 0x00000040
+#define GL_LIGHT_MODEL_AMBIENT 0x0B53
+#define GL_LIGHT_MODEL_COLOR_CONTROL 0x81F8
+#define GL_LIGHT_MODEL_LOCAL_VIEWER 0x0B51
+#define GL_LIGHT_MODEL_TWO_SIDE 0x0B52
+#define GL_LINE 0x1B01
+#define GL_LINEAR 0x2601
+#define GL_LINEAR_ATTENUATION 0x1208
+#define GL_LINEAR_MIPMAP_LINEAR 0x2703
+#define GL_LINEAR_MIPMAP_NEAREST 0x2701
+#define GL_LINES 0x0001
+#define GL_LINES_ADJACENCY 0x000A
+#define GL_LINE_BIT 0x00000004
+#define GL_LINE_LOOP 0x0002
+#define GL_LINE_RESET_TOKEN 0x0707
+#define GL_LINE_SMOOTH 0x0B20
+#define GL_LINE_SMOOTH_HINT 0x0C52
+#define GL_LINE_STIPPLE 0x0B24
+#define GL_LINE_STIPPLE_PATTERN 0x0B25
+#define GL_LINE_STIPPLE_REPEAT 0x0B26
+#define GL_LINE_STRIP 0x0003
+#define GL_LINE_STRIP_ADJACENCY 0x000B
+#define GL_LINE_TOKEN 0x0702
+#define GL_LINE_WIDTH 0x0B21
+#define GL_LINE_WIDTH_GRANULARITY 0x0B23
+#define GL_LINE_WIDTH_RANGE 0x0B22
+#define GL_LINK_STATUS 0x8B82
+#define GL_LIST_BASE 0x0B32
+#define GL_LIST_BIT 0x00020000
+#define GL_LIST_INDEX 0x0B33
+#define GL_LIST_MODE 0x0B30
+#define GL_LOAD 0x0101
+#define GL_LOGIC_OP 0x0BF1
+#define GL_LOGIC_OP_MODE 0x0BF0
+#define GL_LOWER_LEFT 0x8CA1
+#define GL_LUMINANCE 0x1909
+#define GL_LUMINANCE12 0x8041
+#define GL_LUMINANCE12_ALPHA12 0x8047
+#define GL_LUMINANCE12_ALPHA4 0x8046
+#define GL_LUMINANCE16 0x8042
+#define GL_LUMINANCE16_ALPHA16 0x8048
+#define GL_LUMINANCE4 0x803F
+#define GL_LUMINANCE4_ALPHA4 0x8043
+#define GL_LUMINANCE6_ALPHA2 0x8044
+#define GL_LUMINANCE8 0x8040
+#define GL_LUMINANCE8_ALPHA8 0x8045
+#define GL_LUMINANCE_ALPHA 0x190A
+#define GL_MAJOR_VERSION 0x821B
+#define GL_MAP1_COLOR_4 0x0D90
+#define GL_MAP1_GRID_DOMAIN 0x0DD0
+#define GL_MAP1_GRID_SEGMENTS 0x0DD1
+#define GL_MAP1_INDEX 0x0D91
+#define GL_MAP1_NORMAL 0x0D92
+#define GL_MAP1_TEXTURE_COORD_1 0x0D93
+#define GL_MAP1_TEXTURE_COORD_2 0x0D94
+#define GL_MAP1_TEXTURE_COORD_3 0x0D95
+#define GL_MAP1_TEXTURE_COORD_4 0x0D96
+#define GL_MAP1_VERTEX_3 0x0D97
+#define GL_MAP1_VERTEX_4 0x0D98
+#define GL_MAP2_COLOR_4 0x0DB0
+#define GL_MAP2_GRID_DOMAIN 0x0DD2
+#define GL_MAP2_GRID_SEGMENTS 0x0DD3
+#define GL_MAP2_INDEX 0x0DB1
+#define GL_MAP2_NORMAL 0x0DB2
+#define GL_MAP2_TEXTURE_COORD_1 0x0DB3
+#define GL_MAP2_TEXTURE_COORD_2 0x0DB4
+#define GL_MAP2_TEXTURE_COORD_3 0x0DB5
+#define GL_MAP2_TEXTURE_COORD_4 0x0DB6
+#define GL_MAP2_VERTEX_3 0x0DB7
+#define GL_MAP2_VERTEX_4 0x0DB8
+#define GL_MAP_COLOR 0x0D10
+#define GL_MAP_FLUSH_EXPLICIT_BIT 0x0010
+#define GL_MAP_INVALIDATE_BUFFER_BIT 0x0008
+#define GL_MAP_INVALIDATE_RANGE_BIT 0x0004
+#define GL_MAP_READ_BIT 0x0001
+#define GL_MAP_STENCIL 0x0D11
+#define GL_MAP_UNSYNCHRONIZED_BIT 0x0020
+#define GL_MAP_WRITE_BIT 0x0002
+#define GL_MATRIX_MODE 0x0BA0
+#define GL_MAX 0x8008
+#define GL_MAX_3D_TEXTURE_SIZE 0x8073
+#define GL_MAX_ARRAY_TEXTURE_LAYERS 0x88FF
+#define GL_MAX_ATTRIB_STACK_DEPTH 0x0D35
+#define GL_MAX_CLIENT_ATTRIB_STACK_DEPTH 0x0D3B
+#define GL_MAX_CLIP_DISTANCES 0x0D32
+#define GL_MAX_CLIP_PLANES 0x0D32
+#define GL_MAX_COLOR_ATTACHMENTS 0x8CDF
+#define GL_MAX_COLOR_TEXTURE_SAMPLES 0x910E
+#define GL_MAX_COMBINED_FRAGMENT_UNIFORM_COMPONENTS 0x8A33
+#define GL_MAX_COMBINED_GEOMETRY_UNIFORM_COMPONENTS 0x8A32
+#define GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS 0x8B4D
+#define GL_MAX_COMBINED_UNIFORM_BLOCKS 0x8A2E
+#define GL_MAX_COMBINED_VERTEX_UNIFORM_COMPONENTS 0x8A31
+#define GL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C
+#define GL_MAX_DEPTH_TEXTURE_SAMPLES 0x910F
+#define GL_MAX_DRAW_BUFFERS 0x8824
+#define GL_MAX_DUAL_SOURCE_DRAW_BUFFERS 0x88FC
+#define GL_MAX_ELEMENTS_INDICES 0x80E9
+#define GL_MAX_ELEMENTS_VERTICES 0x80E8
+#define GL_MAX_EVAL_ORDER 0x0D30
+#define GL_MAX_FRAGMENT_INPUT_COMPONENTS 0x9125
+#define GL_MAX_FRAGMENT_UNIFORM_BLOCKS 0x8A2D
+#define GL_MAX_FRAGMENT_UNIFORM_COMPONENTS 0x8B49
+#define GL_MAX_GEOMETRY_INPUT_COMPONENTS 0x9123
+#define GL_MAX_GEOMETRY_OUTPUT_COMPONENTS 0x9124
+#define GL_MAX_GEOMETRY_OUTPUT_VERTICES 0x8DE0
+#define GL_MAX_GEOMETRY_TEXTURE_IMAGE_UNITS 0x8C29
+#define GL_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS 0x8DE1
+#define GL_MAX_GEOMETRY_UNIFORM_BLOCKS 0x8A2C
+#define GL_MAX_GEOMETRY_UNIFORM_COMPONENTS 0x8DDF
+#define GL_MAX_INTEGER_SAMPLES 0x9110
+#define GL_MAX_LIGHTS 0x0D31
+#define GL_MAX_LIST_NESTING 0x0B31
+#define GL_MAX_MODELVIEW_STACK_DEPTH 0x0D36
+#define GL_MAX_NAME_STACK_DEPTH 0x0D37
+#define GL_MAX_PIXEL_MAP_TABLE 0x0D34
+#define GL_MAX_PROGRAM_TEXEL_OFFSET 0x8905
+#define GL_MAX_PROJECTION_STACK_DEPTH 0x0D38
+#define GL_MAX_RECTANGLE_TEXTURE_SIZE 0x84F8
+#define GL_MAX_RENDERBUFFER_SIZE 0x84E8
+#define GL_MAX_SAMPLES 0x8D57
+#define GL_MAX_SAMPLE_MASK_WORDS 0x8E59
+#define GL_MAX_SERVER_WAIT_TIMEOUT 0x9111
+#define GL_MAX_TEXTURE_BUFFER_SIZE 0x8C2B
+#define GL_MAX_TEXTURE_COORDS 0x8871
+#define GL_MAX_TEXTURE_IMAGE_UNITS 0x8872
+#define GL_MAX_TEXTURE_LOD_BIAS 0x84FD
+#define GL_MAX_TEXTURE_SIZE 0x0D33
+#define GL_MAX_TEXTURE_STACK_DEPTH 0x0D39
+#define GL_MAX_TEXTURE_UNITS 0x84E2
+#define GL_MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS 0x8C8A
+#define GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_ATTRIBS 0x8C8B
+#define GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS 0x8C80
+#define GL_MAX_UNIFORM_BLOCK_SIZE 0x8A30
+#define GL_MAX_UNIFORM_BUFFER_BINDINGS 0x8A2F
+#define GL_MAX_VARYING_COMPONENTS 0x8B4B
+#define GL_MAX_VARYING_FLOATS 0x8B4B
+#define GL_MAX_VERTEX_ATTRIBS 0x8869
+#define GL_MAX_VERTEX_OUTPUT_COMPONENTS 0x9122
+#define GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS 0x8B4C
+#define GL_MAX_VERTEX_UNIFORM_BLOCKS 0x8A2B
+#define GL_MAX_VERTEX_UNIFORM_COMPONENTS 0x8B4A
+#define GL_MAX_VIEWPORT_DIMS 0x0D3A
+#define GL_MIN 0x8007
+#define GL_MINOR_VERSION 0x821C
+#define GL_MIN_PROGRAM_TEXEL_OFFSET 0x8904
+#define GL_MIRRORED_REPEAT 0x8370
+#define GL_MODELVIEW 0x1700
+#define GL_MODELVIEW_MATRIX 0x0BA6
+#define GL_MODELVIEW_STACK_DEPTH 0x0BA3
+#define GL_MODULATE 0x2100
+#define GL_MULT 0x0103
+#define GL_MULTISAMPLE 0x809D
+#define GL_MULTISAMPLE_BIT 0x20000000
+#define GL_N3F_V3F 0x2A25
+#define GL_NAME_STACK_DEPTH 0x0D70
+#define GL_NAND 0x150E
+#define GL_NEAREST 0x2600
+#define GL_NEAREST_MIPMAP_LINEAR 0x2702
+#define GL_NEAREST_MIPMAP_NEAREST 0x2700
+#define GL_NEVER 0x0200
+#define GL_NICEST 0x1102
+#define GL_NONE 0
+#define GL_NOOP 0x1505
+#define GL_NOR 0x1508
+#define GL_NORMALIZE 0x0BA1
+#define GL_NORMAL_ARRAY 0x8075
+#define GL_NORMAL_ARRAY_BUFFER_BINDING 0x8897
+#define GL_NORMAL_ARRAY_POINTER 0x808F
+#define GL_NORMAL_ARRAY_STRIDE 0x807F
+#define GL_NORMAL_ARRAY_TYPE 0x807E
+#define GL_NORMAL_MAP 0x8511
+#define GL_NOTEQUAL 0x0205
+#define GL_NO_ERROR 0
+#define GL_NUM_COMPRESSED_TEXTURE_FORMATS 0x86A2
+#define GL_NUM_EXTENSIONS 0x821D
+#define GL_OBJECT_LINEAR 0x2401
+#define GL_OBJECT_PLANE 0x2501
+#define GL_OBJECT_TYPE 0x9112
+#define GL_ONE 1
+#define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004
+#define GL_ONE_MINUS_CONSTANT_COLOR 0x8002
+#define GL_ONE_MINUS_DST_ALPHA 0x0305
+#define GL_ONE_MINUS_DST_COLOR 0x0307
+#define GL_ONE_MINUS_SRC1_ALPHA 0x88FB
+#define GL_ONE_MINUS_SRC1_COLOR 0x88FA
+#define GL_ONE_MINUS_SRC_ALPHA 0x0303
+#define GL_ONE_MINUS_SRC_COLOR 0x0301
+#define GL_OPERAND0_ALPHA 0x8598
+#define GL_OPERAND0_RGB 0x8590
+#define GL_OPERAND1_ALPHA 0x8599
+#define GL_OPERAND1_RGB 0x8591
+#define GL_OPERAND2_ALPHA 0x859A
+#define GL_OPERAND2_RGB 0x8592
+#define GL_OR 0x1507
+#define GL_ORDER 0x0A01
+#define GL_OR_INVERTED 0x150D
+#define GL_OR_REVERSE 0x150B
+#define GL_OUT_OF_MEMORY 0x0505
+#define GL_PACK_ALIGNMENT 0x0D05
+#define GL_PACK_IMAGE_HEIGHT 0x806C
+#define GL_PACK_LSB_FIRST 0x0D01
+#define GL_PACK_ROW_LENGTH 0x0D02
+#define GL_PACK_SKIP_IMAGES 0x806B
+#define GL_PACK_SKIP_PIXELS 0x0D04
+#define GL_PACK_SKIP_ROWS 0x0D03
+#define GL_PACK_SWAP_BYTES 0x0D00
+#define GL_PASS_THROUGH_TOKEN 0x0700
+#define GL_PERSPECTIVE_CORRECTION_HINT 0x0C50
+#define GL_PIXEL_MAP_A_TO_A 0x0C79
+#define GL_PIXEL_MAP_A_TO_A_SIZE 0x0CB9
+#define GL_PIXEL_MAP_B_TO_B 0x0C78
+#define GL_PIXEL_MAP_B_TO_B_SIZE 0x0CB8
+#define GL_PIXEL_MAP_G_TO_G 0x0C77
+#define GL_PIXEL_MAP_G_TO_G_SIZE 0x0CB7
+#define GL_PIXEL_MAP_I_TO_A 0x0C75
+#define GL_PIXEL_MAP_I_TO_A_SIZE 0x0CB5
+#define GL_PIXEL_MAP_I_TO_B 0x0C74
+#define GL_PIXEL_MAP_I_TO_B_SIZE 0x0CB4
+#define GL_PIXEL_MAP_I_TO_G 0x0C73
+#define GL_PIXEL_MAP_I_TO_G_SIZE 0x0CB3
+#define GL_PIXEL_MAP_I_TO_I 0x0C70
+#define GL_PIXEL_MAP_I_TO_I_SIZE 0x0CB0
+#define GL_PIXEL_MAP_I_TO_R 0x0C72
+#define GL_PIXEL_MAP_I_TO_R_SIZE 0x0CB2
+#define GL_PIXEL_MAP_R_TO_R 0x0C76
+#define GL_PIXEL_MAP_R_TO_R_SIZE 0x0CB6
+#define GL_PIXEL_MAP_S_TO_S 0x0C71
+#define GL_PIXEL_MAP_S_TO_S_SIZE 0x0CB1
+#define GL_PIXEL_MODE_BIT 0x00000020
+#define GL_PIXEL_PACK_BUFFER 0x88EB
+#define GL_PIXEL_PACK_BUFFER_BINDING 0x88ED
+#define GL_PIXEL_UNPACK_BUFFER 0x88EC
+#define GL_PIXEL_UNPACK_BUFFER_BINDING 0x88EF
+#define GL_POINT 0x1B00
+#define GL_POINTS 0x0000
+#define GL_POINT_BIT 0x00000002
+#define GL_POINT_DISTANCE_ATTENUATION 0x8129
+#define GL_POINT_FADE_THRESHOLD_SIZE 0x8128
+#define GL_POINT_SIZE 0x0B11
+#define GL_POINT_SIZE_GRANULARITY 0x0B13
+#define GL_POINT_SIZE_MAX 0x8127
+#define GL_POINT_SIZE_MIN 0x8126
+#define GL_POINT_SIZE_RANGE 0x0B12
+#define GL_POINT_SMOOTH 0x0B10
+#define GL_POINT_SMOOTH_HINT 0x0C51
+#define GL_POINT_SPRITE 0x8861
+#define GL_POINT_SPRITE_COORD_ORIGIN 0x8CA0
+#define GL_POINT_TOKEN 0x0701
+#define GL_POLYGON 0x0009
+#define GL_POLYGON_BIT 0x00000008
+#define GL_POLYGON_MODE 0x0B40
+#define GL_POLYGON_OFFSET_FACTOR 0x8038
+#define GL_POLYGON_OFFSET_FILL 0x8037
+#define GL_POLYGON_OFFSET_LINE 0x2A02
+#define GL_POLYGON_OFFSET_POINT 0x2A01
+#define GL_POLYGON_OFFSET_UNITS 0x2A00
+#define GL_POLYGON_SMOOTH 0x0B41
+#define GL_POLYGON_SMOOTH_HINT 0x0C53
+#define GL_POLYGON_STIPPLE 0x0B42
+#define GL_POLYGON_STIPPLE_BIT 0x00000010
+#define GL_POLYGON_TOKEN 0x0703
+#define GL_POSITION 0x1203
+#define GL_PREVIOUS 0x8578
+#define GL_PRIMARY_COLOR 0x8577
+#define GL_PRIMITIVES_GENERATED 0x8C87
+#define GL_PRIMITIVE_RESTART 0x8F9D
+#define GL_PRIMITIVE_RESTART_INDEX 0x8F9E
+#define GL_PROGRAM_POINT_SIZE 0x8642
+#define GL_PROJECTION 0x1701
+#define GL_PROJECTION_MATRIX 0x0BA7
+#define GL_PROJECTION_STACK_DEPTH 0x0BA4
+#define GL_PROVOKING_VERTEX 0x8E4F
+#define GL_PROXY_TEXTURE_1D 0x8063
+#define GL_PROXY_TEXTURE_1D_ARRAY 0x8C19
+#define GL_PROXY_TEXTURE_2D 0x8064
+#define GL_PROXY_TEXTURE_2D_ARRAY 0x8C1B
+#define GL_PROXY_TEXTURE_2D_MULTISAMPLE 0x9101
+#define GL_PROXY_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9103
+#define GL_PROXY_TEXTURE_3D 0x8070
+#define GL_PROXY_TEXTURE_CUBE_MAP 0x851B
+#define GL_PROXY_TEXTURE_RECTANGLE 0x84F7
+#define GL_Q 0x2003
+#define GL_QUADRATIC_ATTENUATION 0x1209
+#define GL_QUADS 0x0007
+#define GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION 0x8E4C
+#define GL_QUAD_STRIP 0x0008
+#define GL_QUERY_BY_REGION_NO_WAIT 0x8E16
+#define GL_QUERY_BY_REGION_WAIT 0x8E15
+#define GL_QUERY_COUNTER_BITS 0x8864
+#define GL_QUERY_NO_WAIT 0x8E14
+#define GL_QUERY_RESULT 0x8866
+#define GL_QUERY_RESULT_AVAILABLE 0x8867
+#define GL_QUERY_WAIT 0x8E13
+#define GL_R 0x2002
+#define GL_R11F_G11F_B10F 0x8C3A
+#define GL_R16 0x822A
+#define GL_R16F 0x822D
+#define GL_R16I 0x8233
+#define GL_R16UI 0x8234
+#define GL_R16_SNORM 0x8F98
+#define GL_R32F 0x822E
+#define GL_R32I 0x8235
+#define GL_R32UI 0x8236
+#define GL_R3_G3_B2 0x2A10
+#define GL_R8 0x8229
+#define GL_R8I 0x8231
+#define GL_R8UI 0x8232
+#define GL_R8_SNORM 0x8F94
+#define GL_RASTERIZER_DISCARD 0x8C89
+#define GL_READ_BUFFER 0x0C02
+#define GL_READ_FRAMEBUFFER 0x8CA8
+#define GL_READ_FRAMEBUFFER_BINDING 0x8CAA
+#define GL_READ_ONLY 0x88B8
+#define GL_READ_WRITE 0x88BA
+#define GL_RED 0x1903
+#define GL_RED_BIAS 0x0D15
+#define GL_RED_BITS 0x0D52
+#define GL_RED_INTEGER 0x8D94
+#define GL_RED_SCALE 0x0D14
+#define GL_REFLECTION_MAP 0x8512
+#define GL_RENDER 0x1C00
+#define GL_RENDERBUFFER 0x8D41
+#define GL_RENDERBUFFER_ALPHA_SIZE 0x8D53
+#define GL_RENDERBUFFER_BINDING 0x8CA7
+#define GL_RENDERBUFFER_BLUE_SIZE 0x8D52
+#define GL_RENDERBUFFER_DEPTH_SIZE 0x8D54
+#define GL_RENDERBUFFER_GREEN_SIZE 0x8D51
+#define GL_RENDERBUFFER_HEIGHT 0x8D43
+#define GL_RENDERBUFFER_INTERNAL_FORMAT 0x8D44
+#define GL_RENDERBUFFER_RED_SIZE 0x8D50
+#define GL_RENDERBUFFER_SAMPLES 0x8CAB
+#define GL_RENDERBUFFER_STENCIL_SIZE 0x8D55
+#define GL_RENDERBUFFER_WIDTH 0x8D42
+#define GL_RENDERER 0x1F01
+#define GL_RENDER_MODE 0x0C40
+#define GL_REPEAT 0x2901
+#define GL_REPLACE 0x1E01
+#define GL_RESCALE_NORMAL 0x803A
+#define GL_RETURN 0x0102
+#define GL_RG 0x8227
+#define GL_RG16 0x822C
+#define GL_RG16F 0x822F
+#define GL_RG16I 0x8239
+#define GL_RG16UI 0x823A
+#define GL_RG16_SNORM 0x8F99
+#define GL_RG32F 0x8230
+#define GL_RG32I 0x823B
+#define GL_RG32UI 0x823C
+#define GL_RG8 0x822B
+#define GL_RG8I 0x8237
+#define GL_RG8UI 0x8238
+#define GL_RG8_SNORM 0x8F95
+#define GL_RGB 0x1907
+#define GL_RGB10 0x8052
+#define GL_RGB10_A2 0x8059
+#define GL_RGB10_A2UI 0x906F
+#define GL_RGB12 0x8053
+#define GL_RGB16 0x8054
+#define GL_RGB16F 0x881B
+#define GL_RGB16I 0x8D89
+#define GL_RGB16UI 0x8D77
+#define GL_RGB16_SNORM 0x8F9A
+#define GL_RGB32F 0x8815
+#define GL_RGB32I 0x8D83
+#define GL_RGB32UI 0x8D71
+#define GL_RGB4 0x804F
+#define GL_RGB5 0x8050
+#define GL_RGB5_A1 0x8057
+#define GL_RGB8 0x8051
+#define GL_RGB8I 0x8D8F
+#define GL_RGB8UI 0x8D7D
+#define GL_RGB8_SNORM 0x8F96
+#define GL_RGB9_E5 0x8C3D
+#define GL_RGBA 0x1908
+#define GL_RGBA12 0x805A
+#define GL_RGBA16 0x805B
+#define GL_RGBA16F 0x881A
+#define GL_RGBA16I 0x8D88
+#define GL_RGBA16UI 0x8D76
+#define GL_RGBA16_SNORM 0x8F9B
+#define GL_RGBA2 0x8055
+#define GL_RGBA32F 0x8814
+#define GL_RGBA32I 0x8D82
+#define GL_RGBA32UI 0x8D70
+#define GL_RGBA4 0x8056
+#define GL_RGBA8 0x8058
+#define GL_RGBA8I 0x8D8E
+#define GL_RGBA8UI 0x8D7C
+#define GL_RGBA8_SNORM 0x8F97
+#define GL_RGBA_INTEGER 0x8D99
+#define GL_RGBA_MODE 0x0C31
+#define GL_RGB_INTEGER 0x8D98
+#define GL_RGB_SCALE 0x8573
+#define GL_RG_INTEGER 0x8228
+#define GL_RIGHT 0x0407
+#define GL_S 0x2000
+#define GL_SAMPLER_1D 0x8B5D
+#define GL_SAMPLER_1D_ARRAY 0x8DC0
+#define GL_SAMPLER_1D_ARRAY_SHADOW 0x8DC3
+#define GL_SAMPLER_1D_SHADOW 0x8B61
+#define GL_SAMPLER_2D 0x8B5E
+#define GL_SAMPLER_2D_ARRAY 0x8DC1
+#define GL_SAMPLER_2D_ARRAY_SHADOW 0x8DC4
+#define GL_SAMPLER_2D_MULTISAMPLE 0x9108
+#define GL_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910B
+#define GL_SAMPLER_2D_RECT 0x8B63
+#define GL_SAMPLER_2D_RECT_SHADOW 0x8B64
+#define GL_SAMPLER_2D_SHADOW 0x8B62
+#define GL_SAMPLER_3D 0x8B5F
+#define GL_SAMPLER_BINDING 0x8919
+#define GL_SAMPLER_BUFFER 0x8DC2
+#define GL_SAMPLER_CUBE 0x8B60
+#define GL_SAMPLER_CUBE_SHADOW 0x8DC5
+#define GL_SAMPLES 0x80A9
+#define GL_SAMPLES_PASSED 0x8914
+#define GL_SAMPLE_ALPHA_TO_COVERAGE 0x809E
+#define GL_SAMPLE_ALPHA_TO_ONE 0x809F
+#define GL_SAMPLE_BUFFERS 0x80A8
+#define GL_SAMPLE_COVERAGE 0x80A0
+#define GL_SAMPLE_COVERAGE_INVERT 0x80AB
+#define GL_SAMPLE_COVERAGE_VALUE 0x80AA
+#define GL_SAMPLE_MASK 0x8E51
+#define GL_SAMPLE_MASK_VALUE 0x8E52
+#define GL_SAMPLE_POSITION 0x8E50
+#define GL_SCISSOR_BIT 0x00080000
+#define GL_SCISSOR_BOX 0x0C10
+#define GL_SCISSOR_TEST 0x0C11
+#define GL_SECONDARY_COLOR_ARRAY 0x845E
+#define GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING 0x889C
+#define GL_SECONDARY_COLOR_ARRAY_POINTER 0x845D
+#define GL_SECONDARY_COLOR_ARRAY_SIZE 0x845A
+#define GL_SECONDARY_COLOR_ARRAY_STRIDE 0x845C
+#define GL_SECONDARY_COLOR_ARRAY_TYPE 0x845B
+#define GL_SELECT 0x1C02
+#define GL_SELECTION_BUFFER_POINTER 0x0DF3
+#define GL_SELECTION_BUFFER_SIZE 0x0DF4
+#define GL_SEPARATE_ATTRIBS 0x8C8D
+#define GL_SEPARATE_SPECULAR_COLOR 0x81FA
+#define GL_SET 0x150F
+#define GL_SHADER_SOURCE_LENGTH 0x8B88
+#define GL_SHADER_TYPE 0x8B4F
+#define GL_SHADE_MODEL 0x0B54
+#define GL_SHADING_LANGUAGE_VERSION 0x8B8C
+#define GL_SHININESS 0x1601
+#define GL_SHORT 0x1402
+#define GL_SIGNALED 0x9119
+#define GL_SIGNED_NORMALIZED 0x8F9C
+#define GL_SINGLE_COLOR 0x81F9
+#define GL_SLUMINANCE 0x8C46
+#define GL_SLUMINANCE8 0x8C47
+#define GL_SLUMINANCE8_ALPHA8 0x8C45
+#define GL_SLUMINANCE_ALPHA 0x8C44
+#define GL_SMOOTH 0x1D01
+#define GL_SMOOTH_LINE_WIDTH_GRANULARITY 0x0B23
+#define GL_SMOOTH_LINE_WIDTH_RANGE 0x0B22
+#define GL_SMOOTH_POINT_SIZE_GRANULARITY 0x0B13
+#define GL_SMOOTH_POINT_SIZE_RANGE 0x0B12
+#define GL_SOURCE0_ALPHA 0x8588
+#define GL_SOURCE0_RGB 0x8580
+#define GL_SOURCE1_ALPHA 0x8589
+#define GL_SOURCE1_RGB 0x8581
+#define GL_SOURCE2_ALPHA 0x858A
+#define GL_SOURCE2_RGB 0x8582
+#define GL_SPECULAR 0x1202
+#define GL_SPHERE_MAP 0x2402
+#define GL_SPOT_CUTOFF 0x1206
+#define GL_SPOT_DIRECTION 0x1204
+#define GL_SPOT_EXPONENT 0x1205
+#define GL_SRC0_ALPHA 0x8588
+#define GL_SRC0_RGB 0x8580
+#define GL_SRC1_ALPHA 0x8589
+#define GL_SRC1_COLOR 0x88F9
+#define GL_SRC1_RGB 0x8581
+#define GL_SRC2_ALPHA 0x858A
+#define GL_SRC2_RGB 0x8582
+#define GL_SRC_ALPHA 0x0302
+#define GL_SRC_ALPHA_SATURATE 0x0308
+#define GL_SRC_COLOR 0x0300
+#define GL_SRGB 0x8C40
+#define GL_SRGB8 0x8C41
+#define GL_SRGB8_ALPHA8 0x8C43
+#define GL_SRGB_ALPHA 0x8C42
+#define GL_STACK_OVERFLOW 0x0503
+#define GL_STACK_UNDERFLOW 0x0504
+#define GL_STATIC_COPY 0x88E6
+#define GL_STATIC_DRAW 0x88E4
+#define GL_STATIC_READ 0x88E5
+#define GL_STENCIL 0x1802
+#define GL_STENCIL_ATTACHMENT 0x8D20
+#define GL_STENCIL_BACK_FAIL 0x8801
+#define GL_STENCIL_BACK_FUNC 0x8800
+#define GL_STENCIL_BACK_PASS_DEPTH_FAIL 0x8802
+#define GL_STENCIL_BACK_PASS_DEPTH_PASS 0x8803
+#define GL_STENCIL_BACK_REF 0x8CA3
+#define GL_STENCIL_BACK_VALUE_MASK 0x8CA4
+#define GL_STENCIL_BACK_WRITEMASK 0x8CA5
+#define GL_STENCIL_BITS 0x0D57
+#define GL_STENCIL_BUFFER_BIT 0x00000400
+#define GL_STENCIL_CLEAR_VALUE 0x0B91
+#define GL_STENCIL_FAIL 0x0B94
+#define GL_STENCIL_FUNC 0x0B92
+#define GL_STENCIL_INDEX 0x1901
+#define GL_STENCIL_INDEX1 0x8D46
+#define GL_STENCIL_INDEX16 0x8D49
+#define GL_STENCIL_INDEX4 0x8D47
+#define GL_STENCIL_INDEX8 0x8D48
+#define GL_STENCIL_PASS_DEPTH_FAIL 0x0B95
+#define GL_STENCIL_PASS_DEPTH_PASS 0x0B96
+#define GL_STENCIL_REF 0x0B97
+#define GL_STENCIL_TEST 0x0B90
+#define GL_STENCIL_VALUE_MASK 0x0B93
+#define GL_STENCIL_WRITEMASK 0x0B98
+#define GL_STEREO 0x0C33
+#define GL_STREAM_COPY 0x88E2
+#define GL_STREAM_DRAW 0x88E0
+#define GL_STREAM_READ 0x88E1
+#define GL_SUBPIXEL_BITS 0x0D50
+#define GL_SUBTRACT 0x84E7
+#define GL_SYNC_CONDITION 0x9113
+#define GL_SYNC_FENCE 0x9116
+#define GL_SYNC_FLAGS 0x9115
+#define GL_SYNC_FLUSH_COMMANDS_BIT 0x00000001
+#define GL_SYNC_GPU_COMMANDS_COMPLETE 0x9117
+#define GL_SYNC_STATUS 0x9114
+#define GL_T 0x2001
+#define GL_T2F_C3F_V3F 0x2A2A
+#define GL_T2F_C4F_N3F_V3F 0x2A2C
+#define GL_T2F_C4UB_V3F 0x2A29
+#define GL_T2F_N3F_V3F 0x2A2B
+#define GL_T2F_V3F 0x2A27
+#define GL_T4F_C4F_N3F_V4F 0x2A2D
+#define GL_T4F_V4F 0x2A28
+#define GL_TEXTURE 0x1702
+#define GL_TEXTURE0 0x84C0
+#define GL_TEXTURE1 0x84C1
+#define GL_TEXTURE10 0x84CA
+#define GL_TEXTURE11 0x84CB
+#define GL_TEXTURE12 0x84CC
+#define GL_TEXTURE13 0x84CD
+#define GL_TEXTURE14 0x84CE
+#define GL_TEXTURE15 0x84CF
+#define GL_TEXTURE16 0x84D0
+#define GL_TEXTURE17 0x84D1
+#define GL_TEXTURE18 0x84D2
+#define GL_TEXTURE19 0x84D3
+#define GL_TEXTURE2 0x84C2
+#define GL_TEXTURE20 0x84D4
+#define GL_TEXTURE21 0x84D5
+#define GL_TEXTURE22 0x84D6
+#define GL_TEXTURE23 0x84D7
+#define GL_TEXTURE24 0x84D8
+#define GL_TEXTURE25 0x84D9
+#define GL_TEXTURE26 0x84DA
+#define GL_TEXTURE27 0x84DB
+#define GL_TEXTURE28 0x84DC
+#define GL_TEXTURE29 0x84DD
+#define GL_TEXTURE3 0x84C3
+#define GL_TEXTURE30 0x84DE
+#define GL_TEXTURE31 0x84DF
+#define GL_TEXTURE4 0x84C4
+#define GL_TEXTURE5 0x84C5
+#define GL_TEXTURE6 0x84C6
+#define GL_TEXTURE7 0x84C7
+#define GL_TEXTURE8 0x84C8
+#define GL_TEXTURE9 0x84C9
+#define GL_TEXTURE_1D 0x0DE0
+#define GL_TEXTURE_1D_ARRAY 0x8C18
+#define GL_TEXTURE_2D 0x0DE1
+#define GL_TEXTURE_2D_ARRAY 0x8C1A
+#define GL_TEXTURE_2D_MULTISAMPLE 0x9100
+#define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102
+#define GL_TEXTURE_3D 0x806F
+#define GL_TEXTURE_ALPHA_SIZE 0x805F
+#define GL_TEXTURE_ALPHA_TYPE 0x8C13
+#define GL_TEXTURE_BASE_LEVEL 0x813C
+#define GL_TEXTURE_BINDING_1D 0x8068
+#define GL_TEXTURE_BINDING_1D_ARRAY 0x8C1C
+#define GL_TEXTURE_BINDING_2D 0x8069
+#define GL_TEXTURE_BINDING_2D_ARRAY 0x8C1D
+#define GL_TEXTURE_BINDING_2D_MULTISAMPLE 0x9104
+#define GL_TEXTURE_BINDING_2D_MULTISAMPLE_ARRAY 0x9105
+#define GL_TEXTURE_BINDING_3D 0x806A
+#define GL_TEXTURE_BINDING_BUFFER 0x8C2C
+#define GL_TEXTURE_BINDING_CUBE_MAP 0x8514
+#define GL_TEXTURE_BINDING_RECTANGLE 0x84F6
+#define GL_TEXTURE_BIT 0x00040000
+#define GL_TEXTURE_BLUE_SIZE 0x805E
+#define GL_TEXTURE_BLUE_TYPE 0x8C12
+#define GL_TEXTURE_BORDER 0x1005
+#define GL_TEXTURE_BORDER_COLOR 0x1004
+#define GL_TEXTURE_BUFFER 0x8C2A
+#define GL_TEXTURE_BUFFER_DATA_STORE_BINDING 0x8C2D
+#define GL_TEXTURE_COMPARE_FUNC 0x884D
+#define GL_TEXTURE_COMPARE_MODE 0x884C
+#define GL_TEXTURE_COMPONENTS 0x1003
+#define GL_TEXTURE_COMPRESSED 0x86A1
+#define GL_TEXTURE_COMPRESSED_IMAGE_SIZE 0x86A0
+#define GL_TEXTURE_COMPRESSION_HINT 0x84EF
+#define GL_TEXTURE_COORD_ARRAY 0x8078
+#define GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING 0x889A
+#define GL_TEXTURE_COORD_ARRAY_POINTER 0x8092
+#define GL_TEXTURE_COORD_ARRAY_SIZE 0x8088
+#define GL_TEXTURE_COORD_ARRAY_STRIDE 0x808A
+#define GL_TEXTURE_COORD_ARRAY_TYPE 0x8089
+#define GL_TEXTURE_CUBE_MAP 0x8513
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519
+#define GL_TEXTURE_CUBE_MAP_SEAMLESS 0x884F
+#define GL_TEXTURE_DEPTH 0x8071
+#define GL_TEXTURE_DEPTH_SIZE 0x884A
+#define GL_TEXTURE_DEPTH_TYPE 0x8C16
+#define GL_TEXTURE_ENV 0x2300
+#define GL_TEXTURE_ENV_COLOR 0x2201
+#define GL_TEXTURE_ENV_MODE 0x2200
+#define GL_TEXTURE_FILTER_CONTROL 0x8500
+#define GL_TEXTURE_FIXED_SAMPLE_LOCATIONS 0x9107
+#define GL_TEXTURE_GEN_MODE 0x2500
+#define GL_TEXTURE_GEN_Q 0x0C63
+#define GL_TEXTURE_GEN_R 0x0C62
+#define GL_TEXTURE_GEN_S 0x0C60
+#define GL_TEXTURE_GEN_T 0x0C61
+#define GL_TEXTURE_GREEN_SIZE 0x805D
+#define GL_TEXTURE_GREEN_TYPE 0x8C11
+#define GL_TEXTURE_HEIGHT 0x1001
+#define GL_TEXTURE_INTENSITY_SIZE 0x8061
+#define GL_TEXTURE_INTENSITY_TYPE 0x8C15
+#define GL_TEXTURE_INTERNAL_FORMAT 0x1003
+#define GL_TEXTURE_LOD_BIAS 0x8501
+#define GL_TEXTURE_LUMINANCE_SIZE 0x8060
+#define GL_TEXTURE_LUMINANCE_TYPE 0x8C14
+#define GL_TEXTURE_MAG_FILTER 0x2800
+#define GL_TEXTURE_MATRIX 0x0BA8
+#define GL_TEXTURE_MAX_LEVEL 0x813D
+#define GL_TEXTURE_MAX_LOD 0x813B
+#define GL_TEXTURE_MIN_FILTER 0x2801
+#define GL_TEXTURE_MIN_LOD 0x813A
+#define GL_TEXTURE_PRIORITY 0x8066
+#define GL_TEXTURE_RECTANGLE 0x84F5
+#define GL_TEXTURE_RED_SIZE 0x805C
+#define GL_TEXTURE_RED_TYPE 0x8C10
+#define GL_TEXTURE_RESIDENT 0x8067
+#define GL_TEXTURE_SAMPLES 0x9106
+#define GL_TEXTURE_SHARED_SIZE 0x8C3F
+#define GL_TEXTURE_STACK_DEPTH 0x0BA5
+#define GL_TEXTURE_STENCIL_SIZE 0x88F1
+#define GL_TEXTURE_SWIZZLE_A 0x8E45
+#define GL_TEXTURE_SWIZZLE_B 0x8E44
+#define GL_TEXTURE_SWIZZLE_G 0x8E43
+#define GL_TEXTURE_SWIZZLE_R 0x8E42
+#define GL_TEXTURE_SWIZZLE_RGBA 0x8E46
+#define GL_TEXTURE_WIDTH 0x1000
+#define GL_TEXTURE_WRAP_R 0x8072
+#define GL_TEXTURE_WRAP_S 0x2802
+#define GL_TEXTURE_WRAP_T 0x2803
+#define GL_TIMEOUT_EXPIRED 0x911B
+#define GL_TIMEOUT_IGNORED 0xFFFFFFFFFFFFFFFF
+#define GL_TIMESTAMP 0x8E28
+#define GL_TIME_ELAPSED 0x88BF
+#define GL_TRANSFORM_BIT 0x00001000
+#define GL_TRANSFORM_FEEDBACK_BUFFER 0x8C8E
+#define GL_TRANSFORM_FEEDBACK_BUFFER_BINDING 0x8C8F
+#define GL_TRANSFORM_FEEDBACK_BUFFER_MODE 0x8C7F
+#define GL_TRANSFORM_FEEDBACK_BUFFER_SIZE 0x8C85
+#define GL_TRANSFORM_FEEDBACK_BUFFER_START 0x8C84
+#define GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN 0x8C88
+#define GL_TRANSFORM_FEEDBACK_VARYINGS 0x8C83
+#define GL_TRANSFORM_FEEDBACK_VARYING_MAX_LENGTH 0x8C76
+#define GL_TRANSPOSE_COLOR_MATRIX 0x84E6
+#define GL_TRANSPOSE_MODELVIEW_MATRIX 0x84E3
+#define GL_TRANSPOSE_PROJECTION_MATRIX 0x84E4
+#define GL_TRANSPOSE_TEXTURE_MATRIX 0x84E5
+#define GL_TRIANGLES 0x0004
+#define GL_TRIANGLES_ADJACENCY 0x000C
+#define GL_TRIANGLE_FAN 0x0006
+#define GL_TRIANGLE_STRIP 0x0005
+#define GL_TRIANGLE_STRIP_ADJACENCY 0x000D
+#define GL_TRUE 1
+#define GL_UNIFORM_ARRAY_STRIDE 0x8A3C
+#define GL_UNIFORM_BLOCK_ACTIVE_UNIFORMS 0x8A42
+#define GL_UNIFORM_BLOCK_ACTIVE_UNIFORM_INDICES 0x8A43
+#define GL_UNIFORM_BLOCK_BINDING 0x8A3F
+#define GL_UNIFORM_BLOCK_DATA_SIZE 0x8A40
+#define GL_UNIFORM_BLOCK_INDEX 0x8A3A
+#define GL_UNIFORM_BLOCK_NAME_LENGTH 0x8A41
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_FRAGMENT_SHADER 0x8A46
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_GEOMETRY_SHADER 0x8A45
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_VERTEX_SHADER 0x8A44
+#define GL_UNIFORM_BUFFER 0x8A11
+#define GL_UNIFORM_BUFFER_BINDING 0x8A28
+#define GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT 0x8A34
+#define GL_UNIFORM_BUFFER_SIZE 0x8A2A
+#define GL_UNIFORM_BUFFER_START 0x8A29
+#define GL_UNIFORM_IS_ROW_MAJOR 0x8A3E
+#define GL_UNIFORM_MATRIX_STRIDE 0x8A3D
+#define GL_UNIFORM_NAME_LENGTH 0x8A39
+#define GL_UNIFORM_OFFSET 0x8A3B
+#define GL_UNIFORM_SIZE 0x8A38
+#define GL_UNIFORM_TYPE 0x8A37
+#define GL_UNPACK_ALIGNMENT 0x0CF5
+#define GL_UNPACK_IMAGE_HEIGHT 0x806E
+#define GL_UNPACK_LSB_FIRST 0x0CF1
+#define GL_UNPACK_ROW_LENGTH 0x0CF2
+#define GL_UNPACK_SKIP_IMAGES 0x806D
+#define GL_UNPACK_SKIP_PIXELS 0x0CF4
+#define GL_UNPACK_SKIP_ROWS 0x0CF3
+#define GL_UNPACK_SWAP_BYTES 0x0CF0
+#define GL_UNSIGNALED 0x9118
+#define GL_UNSIGNED_BYTE 0x1401
+#define GL_UNSIGNED_BYTE_2_3_3_REV 0x8362
+#define GL_UNSIGNED_BYTE_3_3_2 0x8032
+#define GL_UNSIGNED_INT 0x1405
+#define GL_UNSIGNED_INT_10F_11F_11F_REV 0x8C3B
+#define GL_UNSIGNED_INT_10_10_10_2 0x8036
+#define GL_UNSIGNED_INT_24_8 0x84FA
+#define GL_UNSIGNED_INT_2_10_10_10_REV 0x8368
+#define GL_UNSIGNED_INT_5_9_9_9_REV 0x8C3E
+#define GL_UNSIGNED_INT_8_8_8_8 0x8035
+#define GL_UNSIGNED_INT_8_8_8_8_REV 0x8367
+#define GL_UNSIGNED_INT_SAMPLER_1D 0x8DD1
+#define GL_UNSIGNED_INT_SAMPLER_1D_ARRAY 0x8DD6
+#define GL_UNSIGNED_INT_SAMPLER_2D 0x8DD2
+#define GL_UNSIGNED_INT_SAMPLER_2D_ARRAY 0x8DD7
+#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE 0x910A
+#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910D
+#define GL_UNSIGNED_INT_SAMPLER_2D_RECT 0x8DD5
+#define GL_UNSIGNED_INT_SAMPLER_3D 0x8DD3
+#define GL_UNSIGNED_INT_SAMPLER_BUFFER 0x8DD8
+#define GL_UNSIGNED_INT_SAMPLER_CUBE 0x8DD4
+#define GL_UNSIGNED_INT_VEC2 0x8DC6
+#define GL_UNSIGNED_INT_VEC3 0x8DC7
+#define GL_UNSIGNED_INT_VEC4 0x8DC8
+#define GL_UNSIGNED_NORMALIZED 0x8C17
+#define GL_UNSIGNED_SHORT 0x1403
+#define GL_UNSIGNED_SHORT_1_5_5_5_REV 0x8366
+#define GL_UNSIGNED_SHORT_4_4_4_4 0x8033
+#define GL_UNSIGNED_SHORT_4_4_4_4_REV 0x8365
+#define GL_UNSIGNED_SHORT_5_5_5_1 0x8034
+#define GL_UNSIGNED_SHORT_5_6_5 0x8363
+#define GL_UNSIGNED_SHORT_5_6_5_REV 0x8364
+#define GL_UPPER_LEFT 0x8CA2
+#define GL_V2F 0x2A20
+#define GL_V3F 0x2A21
+#define GL_VALIDATE_STATUS 0x8B83
+#define GL_VENDOR 0x1F00
+#define GL_VERSION 0x1F02
+#define GL_VERTEX_ARRAY 0x8074
+#define GL_VERTEX_ARRAY_BINDING 0x85B5
+#define GL_VERTEX_ARRAY_BUFFER_BINDING 0x8896
+#define GL_VERTEX_ARRAY_POINTER 0x808E
+#define GL_VERTEX_ARRAY_SIZE 0x807A
+#define GL_VERTEX_ARRAY_STRIDE 0x807C
+#define GL_VERTEX_ARRAY_TYPE 0x807B
+#define GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING 0x889F
+#define GL_VERTEX_ATTRIB_ARRAY_DIVISOR 0x88FE
+#define GL_VERTEX_ATTRIB_ARRAY_ENABLED 0x8622
+#define GL_VERTEX_ATTRIB_ARRAY_INTEGER 0x88FD
+#define GL_VERTEX_ATTRIB_ARRAY_NORMALIZED 0x886A
+#define GL_VERTEX_ATTRIB_ARRAY_POINTER 0x8645
+#define GL_VERTEX_ATTRIB_ARRAY_SIZE 0x8623
+#define GL_VERTEX_ATTRIB_ARRAY_STRIDE 0x8624
+#define GL_VERTEX_ATTRIB_ARRAY_TYPE 0x8625
+#define GL_VERTEX_PROGRAM_POINT_SIZE 0x8642
+#define GL_VERTEX_PROGRAM_TWO_SIDE 0x8643
+#define GL_VERTEX_SHADER 0x8B31
+#define GL_VIEWPORT 0x0BA2
+#define GL_VIEWPORT_BIT 0x00000800
+#define GL_WAIT_FAILED 0x911D
+#define GL_WEIGHT_ARRAY_BUFFER_BINDING 0x889E
+#define GL_WRITE_ONLY 0x88B9
+#define GL_XOR 0x1506
+#define GL_ZERO 0
+#define GL_ZOOM_X 0x0D16
+#define GL_ZOOM_Y 0x0D17
+
+
+#include <KHR/khrplatform.h>
+typedef unsigned int GLenum;
+typedef unsigned char GLboolean;
+typedef unsigned int GLbitfield;
+typedef void GLvoid;
+typedef khronos_int8_t GLbyte;
+typedef khronos_uint8_t GLubyte;
+typedef khronos_int16_t GLshort;
+typedef khronos_uint16_t GLushort;
+typedef int GLint;
+typedef unsigned int GLuint;
+typedef khronos_int32_t GLclampx;
+typedef int GLsizei;
+typedef khronos_float_t GLfloat;
+typedef khronos_float_t GLclampf;
+typedef double GLdouble;
+typedef double GLclampd;
+typedef void *GLeglClientBufferEXT;
+typedef void *GLeglImageOES;
+typedef char GLchar;
+typedef char GLcharARB;
+#ifdef __APPLE__
+typedef void *GLhandleARB;
+#else
+typedef unsigned int GLhandleARB;
+#endif
+typedef khronos_uint16_t GLhalf;
+typedef khronos_uint16_t GLhalfARB;
+typedef khronos_int32_t GLfixed;
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptr;
+#else
+typedef khronos_intptr_t GLintptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptrARB;
+#else
+typedef khronos_intptr_t GLintptrARB;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptr;
+#else
+typedef khronos_ssize_t GLsizeiptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptrARB;
+#else
+typedef khronos_ssize_t GLsizeiptrARB;
+#endif
+typedef khronos_int64_t GLint64;
+typedef khronos_int64_t GLint64EXT;
+typedef khronos_uint64_t GLuint64;
+typedef khronos_uint64_t GLuint64EXT;
+typedef struct __GLsync *GLsync;
+struct _cl_context;
+struct _cl_event;
+typedef void (GLAD_API_PTR *GLDEBUGPROC)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCARB)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCKHR)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCAMD)(GLuint id,GLenum category,GLenum severity,GLsizei length,const GLchar *message,void *userParam);
+typedef unsigned short GLhalfNV;
+typedef GLintptr GLvdpauSurfaceNV;
+typedef void (GLAD_API_PTR *GLVULKANPROCNV)(void);
+
+
+#define GL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_0;
+#define GL_VERSION_1_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_1;
+#define GL_VERSION_1_2 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_2;
+#define GL_VERSION_1_3 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_3;
+#define GL_VERSION_1_4 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_4;
+#define GL_VERSION_1_5 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_5;
+#define GL_VERSION_2_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_2_0;
+#define GL_VERSION_2_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_2_1;
+#define GL_VERSION_3_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_0;
+#define GL_VERSION_3_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_1;
+#define GL_VERSION_3_2 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_2;
+#define GL_VERSION_3_3 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_3;
+
+
+typedef void (GLAD_API_PTR *PFNGLACCUMPROC)(GLenum op, GLfloat value);
+typedef void (GLAD_API_PTR *PFNGLACTIVETEXTUREPROC)(GLenum texture);
+typedef void (GLAD_API_PTR *PFNGLALPHAFUNCPROC)(GLenum func, GLfloat ref);
+typedef GLboolean (GLAD_API_PTR *PFNGLARETEXTURESRESIDENTPROC)(GLsizei n, const GLuint * textures, GLboolean * residences);
+typedef void (GLAD_API_PTR *PFNGLARRAYELEMENTPROC)(GLint i);
+typedef void (GLAD_API_PTR *PFNGLATTACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLBEGINPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBEGINCONDITIONALRENDERPROC)(GLuint id, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBEGINQUERYPROC)(GLenum target, GLuint id);
+typedef void (GLAD_API_PTR *PFNGLBEGINTRANSFORMFEEDBACKPROC)(GLenum primitiveMode);
+typedef void (GLAD_API_PTR *PFNGLBINDATTRIBLOCATIONPROC)(GLuint program, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERPROC)(GLenum target, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERBASEPROC)(GLenum target, GLuint index, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERRANGEPROC)(GLenum target, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAGDATALOCATIONPROC)(GLuint program, GLuint color, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAGDATALOCATIONINDEXEDPROC)(GLuint program, GLuint colorNumber, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAMEBUFFERPROC)(GLenum target, GLuint framebuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDRENDERBUFFERPROC)(GLenum target, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDSAMPLERPROC)(GLuint unit, GLuint sampler);
+typedef void (GLAD_API_PTR *PFNGLBINDTEXTUREPROC)(GLenum target, GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLBINDVERTEXARRAYPROC)(GLuint array);
+typedef void (GLAD_API_PTR *PFNGLBITMAPPROC)(GLsizei width, GLsizei height, GLfloat xorig, GLfloat yorig, GLfloat xmove, GLfloat ymove, const GLubyte * bitmap);
+typedef void (GLAD_API_PTR *PFNGLBLENDCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONSEPARATEPROC)(GLenum modeRGB, GLenum modeAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCPROC)(GLenum sfactor, GLenum dfactor);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCSEPARATEPROC)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLITFRAMEBUFFERPROC)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+typedef void (GLAD_API_PTR *PFNGLBUFFERDATAPROC)(GLenum target, GLsizeiptr size, const void * data, GLenum usage);
+typedef void (GLAD_API_PTR *PFNGLBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCALLLISTPROC)(GLuint list);
+typedef void (GLAD_API_PTR *PFNGLCALLLISTSPROC)(GLsizei n, GLenum type, const void * lists);
+typedef GLenum (GLAD_API_PTR *PFNGLCHECKFRAMEBUFFERSTATUSPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLCLAMPCOLORPROC)(GLenum target, GLenum clamp);
+typedef void (GLAD_API_PTR *PFNGLCLEARPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLCLEARACCUMPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERFIPROC)(GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERFVPROC)(GLenum buffer, GLint drawbuffer, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERIVPROC)(GLenum buffer, GLint drawbuffer, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERUIVPROC)(GLenum buffer, GLint drawbuffer, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARDEPTHPROC)(GLdouble depth);
+typedef void (GLAD_API_PTR *PFNGLCLEARINDEXPROC)(GLfloat c);
+typedef void (GLAD_API_PTR *PFNGLCLEARSTENCILPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLCLIENTACTIVETEXTUREPROC)(GLenum texture);
+typedef GLenum (GLAD_API_PTR *PFNGLCLIENTWAITSYNCPROC)(GLsync sync, GLbitfield flags, GLuint64 timeout);
+typedef void (GLAD_API_PTR *PFNGLCLIPPLANEPROC)(GLenum plane, const GLdouble * equation);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3BPROC)(GLbyte red, GLbyte green, GLbyte blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3DPROC)(GLdouble red, GLdouble green, GLdouble blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3FPROC)(GLfloat red, GLfloat green, GLfloat blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3IPROC)(GLint red, GLint green, GLint blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3SPROC)(GLshort red, GLshort green, GLshort blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UBPROC)(GLubyte red, GLubyte green, GLubyte blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UIPROC)(GLuint red, GLuint green, GLuint blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3USPROC)(GLushort red, GLushort green, GLushort blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4BPROC)(GLbyte red, GLbyte green, GLbyte blue, GLbyte alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4DPROC)(GLdouble red, GLdouble green, GLdouble blue, GLdouble alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4FPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4IPROC)(GLint red, GLint green, GLint blue, GLint alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4SPROC)(GLshort red, GLshort green, GLshort blue, GLshort alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UBPROC)(GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UIPROC)(GLuint red, GLuint green, GLuint blue, GLuint alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4USPROC)(GLushort red, GLushort green, GLushort blue, GLushort alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKPROC)(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKIPROC)(GLuint index, GLboolean r, GLboolean g, GLboolean b, GLboolean a);
+typedef void (GLAD_API_PTR *PFNGLCOLORMATERIALPROC)(GLenum face, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLCOLORP3UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP3UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP4UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP4UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLCOLORPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLCOMPILESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE1DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE3DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOPYBUFFERSUBDATAPROC)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size);
+typedef void (GLAD_API_PTR *PFNGLCOPYPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE1DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLsizei height, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATEPROGRAMPROC)(void);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATESHADERPROC)(GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCULLFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLDELETEBUFFERSPROC)(GLsizei n, const GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEFRAMEBUFFERSPROC)(GLsizei n, const GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETELISTSPROC)(GLuint list, GLsizei range);
+typedef void (GLAD_API_PTR *PFNGLDELETEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLDELETEQUERIESPROC)(GLsizei n, const GLuint * ids);
+typedef void (GLAD_API_PTR *PFNGLDELETERENDERBUFFERSPROC)(GLsizei n, const GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETESAMPLERSPROC)(GLsizei count, const GLuint * samplers);
+typedef void (GLAD_API_PTR *PFNGLDELETESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDELETESYNCPROC)(GLsync sync);
+typedef void (GLAD_API_PTR *PFNGLDELETETEXTURESPROC)(GLsizei n, const GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLDELETEVERTEXARRAYSPROC)(GLsizei n, const GLuint * arrays);
+typedef void (GLAD_API_PTR *PFNGLDEPTHFUNCPROC)(GLenum func);
+typedef void (GLAD_API_PTR *PFNGLDEPTHMASKPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLDEPTHRANGEPROC)(GLdouble n, GLdouble f);
+typedef void (GLAD_API_PTR *PFNGLDETACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDISABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLDISABLECLIENTSTATEPROC)(GLenum array);
+typedef void (GLAD_API_PTR *PFNGLDISABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDISABLEIPROC)(GLenum target, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSPROC)(GLenum mode, GLint first, GLsizei count);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSINSTANCEDPROC)(GLenum mode, GLint first, GLsizei count, GLsizei instancecount);
+typedef void (GLAD_API_PTR *PFNGLDRAWBUFFERPROC)(GLenum buf);
+typedef void (GLAD_API_PTR *PFNGLDRAWBUFFERSPROC)(GLsizei n, const GLenum * bufs);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSBASEVERTEXPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSINSTANCEDPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLsizei instancecount);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLsizei instancecount, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLDRAWPIXELSPROC)(GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLDRAWRANGEELEMENTSPROC)(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC)(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void * indices, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGPOINTERPROC)(GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGVPROC)(const GLboolean * flag);
+typedef void (GLAD_API_PTR *PFNGLENABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLENABLECLIENTSTATEPROC)(GLenum array);
+typedef void (GLAD_API_PTR *PFNGLENABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLENABLEIPROC)(GLenum target, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLENDPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDCONDITIONALRENDERPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDLISTPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDQUERYPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLENDTRANSFORMFEEDBACKPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1DPROC)(GLdouble u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1DVPROC)(const GLdouble * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1FPROC)(GLfloat u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1FVPROC)(const GLfloat * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2DPROC)(GLdouble u, GLdouble v);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2DVPROC)(const GLdouble * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2FPROC)(GLfloat u, GLfloat v);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2FVPROC)(const GLfloat * u);
+typedef void (GLAD_API_PTR *PFNGLEVALMESH1PROC)(GLenum mode, GLint i1, GLint i2);
+typedef void (GLAD_API_PTR *PFNGLEVALMESH2PROC)(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2);
+typedef void (GLAD_API_PTR *PFNGLEVALPOINT1PROC)(GLint i);
+typedef void (GLAD_API_PTR *PFNGLEVALPOINT2PROC)(GLint i, GLint j);
+typedef void (GLAD_API_PTR *PFNGLFEEDBACKBUFFERPROC)(GLsizei size, GLenum type, GLfloat * buffer);
+typedef GLsync (GLAD_API_PTR *PFNGLFENCESYNCPROC)(GLenum condition, GLbitfield flags);
+typedef void (GLAD_API_PTR *PFNGLFINISHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHMAPPEDBUFFERRANGEPROC)(GLenum target, GLintptr offset, GLsizeiptr length);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDDPROC)(GLdouble coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDDVPROC)(const GLdouble * coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDFPROC)(GLfloat coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDFVPROC)(const GLfloat * coord);
+typedef void (GLAD_API_PTR *PFNGLFOGFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLFOGFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLFOGIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLFOGIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERRENDERBUFFERPROC)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTUREPROC)(GLenum target, GLenum attachment, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE1DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE2DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE3DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURELAYERPROC)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer);
+typedef void (GLAD_API_PTR *PFNGLFRONTFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLFRUSTUMPROC)(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble zNear, GLdouble zFar);
+typedef void (GLAD_API_PTR *PFNGLGENBUFFERSPROC)(GLsizei n, GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLGENFRAMEBUFFERSPROC)(GLsizei n, GLuint * framebuffers);
+typedef GLuint (GLAD_API_PTR *PFNGLGENLISTSPROC)(GLsizei range);
+typedef void (GLAD_API_PTR *PFNGLGENQUERIESPROC)(GLsizei n, GLuint * ids);
+typedef void (GLAD_API_PTR *PFNGLGENRENDERBUFFERSPROC)(GLsizei n, GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLGENSAMPLERSPROC)(GLsizei count, GLuint * samplers);
+typedef void (GLAD_API_PTR *PFNGLGENTEXTURESPROC)(GLsizei n, GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLGENVERTEXARRAYSPROC)(GLsizei n, GLuint * arrays);
+typedef void (GLAD_API_PTR *PFNGLGENERATEMIPMAPPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEATTRIBPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC)(GLuint program, GLuint uniformBlockIndex, GLsizei bufSize, GLsizei * length, GLchar * uniformBlockName);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMBLOCKIVPROC)(GLuint program, GLuint uniformBlockIndex, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMNAMEPROC)(GLuint program, GLuint uniformIndex, GLsizei bufSize, GLsizei * length, GLchar * uniformName);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMSIVPROC)(GLuint program, GLsizei uniformCount, const GLuint * uniformIndices, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETATTACHEDSHADERSPROC)(GLuint program, GLsizei maxCount, GLsizei * count, GLuint * shaders);
+typedef GLint (GLAD_API_PTR *PFNGLGETATTRIBLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANI_VPROC)(GLenum target, GLuint index, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANVPROC)(GLenum pname, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERI64VPROC)(GLenum target, GLenum pname, GLint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVPROC)(GLenum target, GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, void * data);
+typedef void (GLAD_API_PTR *PFNGLGETCLIPPLANEPROC)(GLenum plane, GLdouble * equation);
+typedef void (GLAD_API_PTR *PFNGLGETCOMPRESSEDTEXIMAGEPROC)(GLenum target, GLint level, void * img);
+typedef void (GLAD_API_PTR *PFNGLGETDOUBLEVPROC)(GLenum pname, GLdouble * data);
+typedef GLenum (GLAD_API_PTR *PFNGLGETERRORPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLGETFLOATVPROC)(GLenum pname, GLfloat * data);
+typedef GLint (GLAD_API_PTR *PFNGLGETFRAGDATAINDEXPROC)(GLuint program, const GLchar * name);
+typedef GLint (GLAD_API_PTR *PFNGLGETFRAGDATALOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC)(GLenum target, GLenum attachment, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGER64I_VPROC)(GLenum target, GLuint index, GLint64 * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGER64VPROC)(GLenum pname, GLint64 * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERI_VPROC)(GLenum target, GLuint index, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERVPROC)(GLenum pname, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETLIGHTFVPROC)(GLenum light, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETLIGHTIVPROC)(GLenum light, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETMAPDVPROC)(GLenum target, GLenum query, GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLGETMAPFVPROC)(GLenum target, GLenum query, GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLGETMAPIVPROC)(GLenum target, GLenum query, GLint * v);
+typedef void (GLAD_API_PTR *PFNGLGETMATERIALFVPROC)(GLenum face, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETMATERIALIVPROC)(GLenum face, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETMULTISAMPLEFVPROC)(GLenum pname, GLuint index, GLfloat * val);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPFVPROC)(GLenum map, GLfloat * values);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPUIVPROC)(GLenum map, GLuint * values);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPUSVPROC)(GLenum map, GLushort * values);
+typedef void (GLAD_API_PTR *PFNGLGETPOINTERVPROC)(GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETPOLYGONSTIPPLEPROC)(GLubyte * mask);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMINFOLOGPROC)(GLuint program, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMIVPROC)(GLuint program, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTI64VPROC)(GLuint id, GLenum pname, GLint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTIVPROC)(GLuint id, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTUI64VPROC)(GLuint id, GLenum pname, GLuint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTUIVPROC)(GLuint id, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETRENDERBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIIVPROC)(GLuint sampler, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIUIVPROC)(GLuint sampler, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERFVPROC)(GLuint sampler, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIVPROC)(GLuint sampler, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERINFOLOGPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERSOURCEPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * source);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERIVPROC)(GLuint shader, GLenum pname, GLint * params);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGPROC)(GLenum name);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGIPROC)(GLenum name, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLGETSYNCIVPROC)(GLsync sync, GLenum pname, GLsizei count, GLsizei * length, GLint * values);
+typedef void (GLAD_API_PTR *PFNGLGETTEXENVFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXENVIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENDVPROC)(GLenum coord, GLenum pname, GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENFVPROC)(GLenum coord, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENIVPROC)(GLenum coord, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXIMAGEPROC)(GLenum target, GLint level, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLGETTEXLEVELPARAMETERFVPROC)(GLenum target, GLint level, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXLEVELPARAMETERIVPROC)(GLenum target, GLint level, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIUIVPROC)(GLenum target, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTRANSFORMFEEDBACKVARYINGPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLsizei * size, GLenum * type, GLchar * name);
+typedef GLuint (GLAD_API_PTR *PFNGLGETUNIFORMBLOCKINDEXPROC)(GLuint program, const GLchar * uniformBlockName);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMINDICESPROC)(GLuint program, GLsizei uniformCount, const GLchar *const* uniformNames, GLuint * uniformIndices);
+typedef GLint (GLAD_API_PTR *PFNGLGETUNIFORMLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMFVPROC)(GLuint program, GLint location, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMIVPROC)(GLuint program, GLint location, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMUIVPROC)(GLuint program, GLint location, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIUIVPROC)(GLuint index, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBPOINTERVPROC)(GLuint index, GLenum pname, void ** pointer);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBDVPROC)(GLuint index, GLenum pname, GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBFVPROC)(GLuint index, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLHINTPROC)(GLenum target, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLINDEXMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLINDEXPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLINDEXDPROC)(GLdouble c);
+typedef void (GLAD_API_PTR *PFNGLINDEXDVPROC)(const GLdouble * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXFPROC)(GLfloat c);
+typedef void (GLAD_API_PTR *PFNGLINDEXFVPROC)(const GLfloat * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXIPROC)(GLint c);
+typedef void (GLAD_API_PTR *PFNGLINDEXIVPROC)(const GLint * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXSPROC)(GLshort c);
+typedef void (GLAD_API_PTR *PFNGLINDEXSVPROC)(const GLshort * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXUBPROC)(GLubyte c);
+typedef void (GLAD_API_PTR *PFNGLINDEXUBVPROC)(const GLubyte * c);
+typedef void (GLAD_API_PTR *PFNGLINITNAMESPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLINTERLEAVEDARRAYSPROC)(GLenum format, GLsizei stride, const void * pointer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISBUFFERPROC)(GLuint buffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDPROC)(GLenum cap);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDIPROC)(GLenum target, GLuint index);
+typedef GLboolean (GLAD_API_PTR *PFNGLISFRAMEBUFFERPROC)(GLuint framebuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISLISTPROC)(GLuint list);
+typedef GLboolean (GLAD_API_PTR *PFNGLISPROGRAMPROC)(GLuint program);
+typedef GLboolean (GLAD_API_PTR *PFNGLISQUERYPROC)(GLuint id);
+typedef GLboolean (GLAD_API_PTR *PFNGLISRENDERBUFFERPROC)(GLuint renderbuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSAMPLERPROC)(GLuint sampler);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSHADERPROC)(GLuint shader);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSYNCPROC)(GLsync sync);
+typedef GLboolean (GLAD_API_PTR *PFNGLISTEXTUREPROC)(GLuint texture);
+typedef GLboolean (GLAD_API_PTR *PFNGLISVERTEXARRAYPROC)(GLuint array);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTFPROC)(GLenum light, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTFVPROC)(GLenum light, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTIPROC)(GLenum light, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTIVPROC)(GLenum light, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLLINESTIPPLEPROC)(GLint factor, GLushort pattern);
+typedef void (GLAD_API_PTR *PFNGLLINEWIDTHPROC)(GLfloat width);
+typedef void (GLAD_API_PTR *PFNGLLINKPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLLISTBASEPROC)(GLuint base);
+typedef void (GLAD_API_PTR *PFNGLLOADIDENTITYPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLLOADMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLLOADMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLLOADNAMEPROC)(GLuint name);
+typedef void (GLAD_API_PTR *PFNGLLOADTRANSPOSEMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLLOADTRANSPOSEMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLLOGICOPPROC)(GLenum opcode);
+typedef void (GLAD_API_PTR *PFNGLMAP1DPROC)(GLenum target, GLdouble u1, GLdouble u2, GLint stride, GLint order, const GLdouble * points);
+typedef void (GLAD_API_PTR *PFNGLMAP1FPROC)(GLenum target, GLfloat u1, GLfloat u2, GLint stride, GLint order, const GLfloat * points);
+typedef void (GLAD_API_PTR *PFNGLMAP2DPROC)(GLenum target, GLdouble u1, GLdouble u2, GLint ustride, GLint uorder, GLdouble v1, GLdouble v2, GLint vstride, GLint vorder, const GLdouble * points);
+typedef void (GLAD_API_PTR *PFNGLMAP2FPROC)(GLenum target, GLfloat u1, GLfloat u2, GLint ustride, GLint uorder, GLfloat v1, GLfloat v2, GLint vstride, GLint vorder, const GLfloat * points);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERPROC)(GLenum target, GLenum access);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERRANGEPROC)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID1DPROC)(GLint un, GLdouble u1, GLdouble u2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID1FPROC)(GLint un, GLfloat u1, GLfloat u2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID2DPROC)(GLint un, GLdouble u1, GLdouble u2, GLint vn, GLdouble v1, GLdouble v2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID2FPROC)(GLint un, GLfloat u1, GLfloat u2, GLint vn, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLMATERIALFPROC)(GLenum face, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLMATERIALFVPROC)(GLenum face, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLMATERIALIPROC)(GLenum face, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLMATERIALIVPROC)(GLenum face, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLMATRIXMODEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLMULTMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLMULTMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLMULTTRANSPOSEMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLMULTTRANSPOSEMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWARRAYSPROC)(GLenum mode, const GLint * first, const GLsizei * count, GLsizei drawcount);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWELEMENTSPROC)(GLenum mode, const GLsizei * count, GLenum type, const void *const* indices, GLsizei drawcount);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC)(GLenum mode, const GLsizei * count, GLenum type, const void *const* indices, GLsizei drawcount, const GLint * basevertex);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1DPROC)(GLenum target, GLdouble s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1FPROC)(GLenum target, GLfloat s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1IPROC)(GLenum target, GLint s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1SPROC)(GLenum target, GLshort s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2DPROC)(GLenum target, GLdouble s, GLdouble t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2FPROC)(GLenum target, GLfloat s, GLfloat t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2IPROC)(GLenum target, GLint s, GLint t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2SPROC)(GLenum target, GLshort s, GLshort t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3DPROC)(GLenum target, GLdouble s, GLdouble t, GLdouble r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3FPROC)(GLenum target, GLfloat s, GLfloat t, GLfloat r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3IPROC)(GLenum target, GLint s, GLint t, GLint r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3SPROC)(GLenum target, GLshort s, GLshort t, GLshort r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4DPROC)(GLenum target, GLdouble s, GLdouble t, GLdouble r, GLdouble q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4FPROC)(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4IPROC)(GLenum target, GLint s, GLint t, GLint r, GLint q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4SPROC)(GLenum target, GLshort s, GLshort t, GLshort r, GLshort q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP1UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP1UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP2UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP2UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP3UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP3UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP4UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP4UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLNEWLISTPROC)(GLuint list, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3BPROC)(GLbyte nx, GLbyte ny, GLbyte nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3DPROC)(GLdouble nx, GLdouble ny, GLdouble nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3FPROC)(GLfloat nx, GLfloat ny, GLfloat nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3IPROC)(GLint nx, GLint ny, GLint nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3SPROC)(GLshort nx, GLshort ny, GLshort nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLNORMALP3UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLNORMALP3UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLNORMALPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLORTHOPROC)(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble zNear, GLdouble zFar);
+typedef void (GLAD_API_PTR *PFNGLPASSTHROUGHPROC)(GLfloat token);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPFVPROC)(GLenum map, GLsizei mapsize, const GLfloat * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPUIVPROC)(GLenum map, GLsizei mapsize, const GLuint * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPUSVPROC)(GLenum map, GLsizei mapsize, const GLushort * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPIXELTRANSFERFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPIXELTRANSFERIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPIXELZOOMPROC)(GLfloat xfactor, GLfloat yfactor);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLPOINTSIZEPROC)(GLfloat size);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONMODEPROC)(GLenum face, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONOFFSETPROC)(GLfloat factor, GLfloat units);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONSTIPPLEPROC)(const GLubyte * mask);
+typedef void (GLAD_API_PTR *PFNGLPOPATTRIBPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPCLIENTATTRIBPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPMATRIXPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPNAMEPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPRIMITIVERESTARTINDEXPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLPRIORITIZETEXTURESPROC)(GLsizei n, const GLuint * textures, const GLfloat * priorities);
+typedef void (GLAD_API_PTR *PFNGLPROVOKINGVERTEXPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLPUSHATTRIBPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLPUSHCLIENTATTRIBPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLPUSHMATRIXPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPUSHNAMEPROC)(GLuint name);
+typedef void (GLAD_API_PTR *PFNGLQUERYCOUNTERPROC)(GLuint id, GLenum target);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4DPROC)(GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4FPROC)(GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4IPROC)(GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4SPROC)(GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLREADBUFFERPROC)(GLenum src);
+typedef void (GLAD_API_PTR *PFNGLREADPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLRECTDPROC)(GLdouble x1, GLdouble y1, GLdouble x2, GLdouble y2);
+typedef void (GLAD_API_PTR *PFNGLRECTDVPROC)(const GLdouble * v1, const GLdouble * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTFPROC)(GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2);
+typedef void (GLAD_API_PTR *PFNGLRECTFVPROC)(const GLfloat * v1, const GLfloat * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTIPROC)(GLint x1, GLint y1, GLint x2, GLint y2);
+typedef void (GLAD_API_PTR *PFNGLRECTIVPROC)(const GLint * v1, const GLint * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTSPROC)(GLshort x1, GLshort y1, GLshort x2, GLshort y2);
+typedef void (GLAD_API_PTR *PFNGLRECTSVPROC)(const GLshort * v1, const GLshort * v2);
+typedef GLint (GLAD_API_PTR *PFNGLRENDERMODEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEPROC)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLROTATEDPROC)(GLdouble angle, GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLROTATEFPROC)(GLfloat angle, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLSAMPLECOVERAGEPROC)(GLfloat value, GLboolean invert);
+typedef void (GLAD_API_PTR *PFNGLSAMPLEMASKIPROC)(GLuint maskNumber, GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIIVPROC)(GLuint sampler, GLenum pname, const GLint * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIUIVPROC)(GLuint sampler, GLenum pname, const GLuint * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERFPROC)(GLuint sampler, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERFVPROC)(GLuint sampler, GLenum pname, const GLfloat * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIPROC)(GLuint sampler, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIVPROC)(GLuint sampler, GLenum pname, const GLint * param);
+typedef void (GLAD_API_PTR *PFNGLSCALEDPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLSCALEFPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLSCISSORPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3BPROC)(GLbyte red, GLbyte green, GLbyte blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3DPROC)(GLdouble red, GLdouble green, GLdouble blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3FPROC)(GLfloat red, GLfloat green, GLfloat blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3IPROC)(GLint red, GLint green, GLint blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3SPROC)(GLshort red, GLshort green, GLshort blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UBPROC)(GLubyte red, GLubyte green, GLubyte blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UIPROC)(GLuint red, GLuint green, GLuint blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3USPROC)(GLushort red, GLushort green, GLushort blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORP3UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORP3UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLSELECTBUFFERPROC)(GLsizei size, GLuint * buffer);
+typedef void (GLAD_API_PTR *PFNGLSHADEMODELPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLSHADERSOURCEPROC)(GLuint shader, GLsizei count, const GLchar *const* string, const GLint * length);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCPROC)(GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCSEPARATEPROC)(GLenum face, GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKSEPARATEPROC)(GLenum face, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPPROC)(GLenum fail, GLenum zfail, GLenum zpass);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPSEPARATEPROC)(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass);
+typedef void (GLAD_API_PTR *PFNGLTEXBUFFERPROC)(GLenum target, GLenum internalformat, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1DPROC)(GLdouble s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1FPROC)(GLfloat s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1IPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1SPROC)(GLshort s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2DPROC)(GLdouble s, GLdouble t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2FPROC)(GLfloat s, GLfloat t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2IPROC)(GLint s, GLint t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2SPROC)(GLshort s, GLshort t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3DPROC)(GLdouble s, GLdouble t, GLdouble r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3FPROC)(GLfloat s, GLfloat t, GLfloat r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3IPROC)(GLint s, GLint t, GLint r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3SPROC)(GLshort s, GLshort t, GLshort r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4DPROC)(GLdouble s, GLdouble t, GLdouble r, GLdouble q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4FPROC)(GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4IPROC)(GLint s, GLint t, GLint r, GLint q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4SPROC)(GLshort s, GLshort t, GLshort r, GLshort q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP1UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP1UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP2UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP2UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP3UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP3UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP4UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP4UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLTEXENVFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXENVFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXENVIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXENVIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENDPROC)(GLenum coord, GLenum pname, GLdouble param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENDVPROC)(GLenum coord, GLenum pname, const GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENFPROC)(GLenum coord, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENFVPROC)(GLenum coord, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENIPROC)(GLenum coord, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENIVPROC)(GLenum coord, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE1DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLboolean fixedsamplelocations);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE3DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE3DMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLboolean fixedsamplelocations);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIUIVPROC)(GLenum target, GLenum pname, const GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTRANSFORMFEEDBACKVARYINGSPROC)(GLuint program, GLsizei count, const GLchar *const* varyings, GLenum bufferMode);
+typedef void (GLAD_API_PTR *PFNGLTRANSLATEDPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLTRANSLATEFPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FPROC)(GLint location, GLfloat v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IPROC)(GLint location, GLint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1UIPROC)(GLint location, GLuint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FPROC)(GLint location, GLfloat v0, GLfloat v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IPROC)(GLint location, GLint v0, GLint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2UIPROC)(GLint location, GLuint v0, GLuint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IPROC)(GLint location, GLint v0, GLint v1, GLint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3UIPROC)(GLint location, GLuint v0, GLuint v1, GLuint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2, GLfloat v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IPROC)(GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4UIPROC)(GLint location, GLuint v0, GLuint v1, GLuint v2, GLuint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMBLOCKBINDINGPROC)(GLuint program, GLuint uniformBlockIndex, GLuint uniformBlockBinding);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2X3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2X4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3X2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3X4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4X2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4X3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFERPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLUSEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVALIDATEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4DPROC)(GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4FPROC)(GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4IPROC)(GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4SPROC)(GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1DPROC)(GLuint index, GLdouble x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FPROC)(GLuint index, GLfloat x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1SPROC)(GLuint index, GLshort x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2DPROC)(GLuint index, GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FPROC)(GLuint index, GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2SPROC)(GLuint index, GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3DPROC)(GLuint index, GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3SPROC)(GLuint index, GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NBVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NIVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NSVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUBPROC)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUSVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4BVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4DPROC)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4SPROC)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4UBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4USVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBDIVISORPROC)(GLuint index, GLuint divisor);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1IPROC)(GLuint index, GLint x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1UIPROC)(GLuint index, GLuint x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2IPROC)(GLuint index, GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2UIPROC)(GLuint index, GLuint x, GLuint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3IPROC)(GLuint index, GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3UIPROC)(GLuint index, GLuint x, GLuint y, GLuint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4BVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4IPROC)(GLuint index, GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UIPROC)(GLuint index, GLuint x, GLuint y, GLuint z, GLuint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4USVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBIPOINTERPROC)(GLuint index, GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP1UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP1UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP2UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP2UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP3UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP3UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP4UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP4UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBPOINTERPROC)(GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP2UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP2UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP3UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP3UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP4UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP4UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVIEWPORTPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLWAITSYNCPROC)(GLsync sync, GLbitfield flags, GLuint64 timeout);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3SVPROC)(const GLshort * v);
+
+GLAD_API_CALL PFNGLACCUMPROC glad_glAccum;
+#define glAccum glad_glAccum
+GLAD_API_CALL PFNGLACTIVETEXTUREPROC glad_glActiveTexture;
+#define glActiveTexture glad_glActiveTexture
+GLAD_API_CALL PFNGLALPHAFUNCPROC glad_glAlphaFunc;
+#define glAlphaFunc glad_glAlphaFunc
+GLAD_API_CALL PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident;
+#define glAreTexturesResident glad_glAreTexturesResident
+GLAD_API_CALL PFNGLARRAYELEMENTPROC glad_glArrayElement;
+#define glArrayElement glad_glArrayElement
+GLAD_API_CALL PFNGLATTACHSHADERPROC glad_glAttachShader;
+#define glAttachShader glad_glAttachShader
+GLAD_API_CALL PFNGLBEGINPROC glad_glBegin;
+#define glBegin glad_glBegin
+GLAD_API_CALL PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender;
+#define glBeginConditionalRender glad_glBeginConditionalRender
+GLAD_API_CALL PFNGLBEGINQUERYPROC glad_glBeginQuery;
+#define glBeginQuery glad_glBeginQuery
+GLAD_API_CALL PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback;
+#define glBeginTransformFeedback glad_glBeginTransformFeedback
+GLAD_API_CALL PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation;
+#define glBindAttribLocation glad_glBindAttribLocation
+GLAD_API_CALL PFNGLBINDBUFFERPROC glad_glBindBuffer;
+#define glBindBuffer glad_glBindBuffer
+GLAD_API_CALL PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase;
+#define glBindBufferBase glad_glBindBufferBase
+GLAD_API_CALL PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange;
+#define glBindBufferRange glad_glBindBufferRange
+GLAD_API_CALL PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation;
+#define glBindFragDataLocation glad_glBindFragDataLocation
+GLAD_API_CALL PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed;
+#define glBindFragDataLocationIndexed glad_glBindFragDataLocationIndexed
+GLAD_API_CALL PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer;
+#define glBindFramebuffer glad_glBindFramebuffer
+GLAD_API_CALL PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer;
+#define glBindRenderbuffer glad_glBindRenderbuffer
+GLAD_API_CALL PFNGLBINDSAMPLERPROC glad_glBindSampler;
+#define glBindSampler glad_glBindSampler
+GLAD_API_CALL PFNGLBINDTEXTUREPROC glad_glBindTexture;
+#define glBindTexture glad_glBindTexture
+GLAD_API_CALL PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray;
+#define glBindVertexArray glad_glBindVertexArray
+GLAD_API_CALL PFNGLBITMAPPROC glad_glBitmap;
+#define glBitmap glad_glBitmap
+GLAD_API_CALL PFNGLBLENDCOLORPROC glad_glBlendColor;
+#define glBlendColor glad_glBlendColor
+GLAD_API_CALL PFNGLBLENDEQUATIONPROC glad_glBlendEquation;
+#define glBlendEquation glad_glBlendEquation
+GLAD_API_CALL PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate;
+#define glBlendEquationSeparate glad_glBlendEquationSeparate
+GLAD_API_CALL PFNGLBLENDFUNCPROC glad_glBlendFunc;
+#define glBlendFunc glad_glBlendFunc
+GLAD_API_CALL PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate;
+#define glBlendFuncSeparate glad_glBlendFuncSeparate
+GLAD_API_CALL PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer;
+#define glBlitFramebuffer glad_glBlitFramebuffer
+GLAD_API_CALL PFNGLBUFFERDATAPROC glad_glBufferData;
+#define glBufferData glad_glBufferData
+GLAD_API_CALL PFNGLBUFFERSUBDATAPROC glad_glBufferSubData;
+#define glBufferSubData glad_glBufferSubData
+GLAD_API_CALL PFNGLCALLLISTPROC glad_glCallList;
+#define glCallList glad_glCallList
+GLAD_API_CALL PFNGLCALLLISTSPROC glad_glCallLists;
+#define glCallLists glad_glCallLists
+GLAD_API_CALL PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus;
+#define glCheckFramebufferStatus glad_glCheckFramebufferStatus
+GLAD_API_CALL PFNGLCLAMPCOLORPROC glad_glClampColor;
+#define glClampColor glad_glClampColor
+GLAD_API_CALL PFNGLCLEARPROC glad_glClear;
+#define glClear glad_glClear
+GLAD_API_CALL PFNGLCLEARACCUMPROC glad_glClearAccum;
+#define glClearAccum glad_glClearAccum
+GLAD_API_CALL PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi;
+#define glClearBufferfi glad_glClearBufferfi
+GLAD_API_CALL PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv;
+#define glClearBufferfv glad_glClearBufferfv
+GLAD_API_CALL PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv;
+#define glClearBufferiv glad_glClearBufferiv
+GLAD_API_CALL PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv;
+#define glClearBufferuiv glad_glClearBufferuiv
+GLAD_API_CALL PFNGLCLEARCOLORPROC glad_glClearColor;
+#define glClearColor glad_glClearColor
+GLAD_API_CALL PFNGLCLEARDEPTHPROC glad_glClearDepth;
+#define glClearDepth glad_glClearDepth
+GLAD_API_CALL PFNGLCLEARINDEXPROC glad_glClearIndex;
+#define glClearIndex glad_glClearIndex
+GLAD_API_CALL PFNGLCLEARSTENCILPROC glad_glClearStencil;
+#define glClearStencil glad_glClearStencil
+GLAD_API_CALL PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture;
+#define glClientActiveTexture glad_glClientActiveTexture
+GLAD_API_CALL PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync;
+#define glClientWaitSync glad_glClientWaitSync
+GLAD_API_CALL PFNGLCLIPPLANEPROC glad_glClipPlane;
+#define glClipPlane glad_glClipPlane
+GLAD_API_CALL PFNGLCOLOR3BPROC glad_glColor3b;
+#define glColor3b glad_glColor3b
+GLAD_API_CALL PFNGLCOLOR3BVPROC glad_glColor3bv;
+#define glColor3bv glad_glColor3bv
+GLAD_API_CALL PFNGLCOLOR3DPROC glad_glColor3d;
+#define glColor3d glad_glColor3d
+GLAD_API_CALL PFNGLCOLOR3DVPROC glad_glColor3dv;
+#define glColor3dv glad_glColor3dv
+GLAD_API_CALL PFNGLCOLOR3FPROC glad_glColor3f;
+#define glColor3f glad_glColor3f
+GLAD_API_CALL PFNGLCOLOR3FVPROC glad_glColor3fv;
+#define glColor3fv glad_glColor3fv
+GLAD_API_CALL PFNGLCOLOR3IPROC glad_glColor3i;
+#define glColor3i glad_glColor3i
+GLAD_API_CALL PFNGLCOLOR3IVPROC glad_glColor3iv;
+#define glColor3iv glad_glColor3iv
+GLAD_API_CALL PFNGLCOLOR3SPROC glad_glColor3s;
+#define glColor3s glad_glColor3s
+GLAD_API_CALL PFNGLCOLOR3SVPROC glad_glColor3sv;
+#define glColor3sv glad_glColor3sv
+GLAD_API_CALL PFNGLCOLOR3UBPROC glad_glColor3ub;
+#define glColor3ub glad_glColor3ub
+GLAD_API_CALL PFNGLCOLOR3UBVPROC glad_glColor3ubv;
+#define glColor3ubv glad_glColor3ubv
+GLAD_API_CALL PFNGLCOLOR3UIPROC glad_glColor3ui;
+#define glColor3ui glad_glColor3ui
+GLAD_API_CALL PFNGLCOLOR3UIVPROC glad_glColor3uiv;
+#define glColor3uiv glad_glColor3uiv
+GLAD_API_CALL PFNGLCOLOR3USPROC glad_glColor3us;
+#define glColor3us glad_glColor3us
+GLAD_API_CALL PFNGLCOLOR3USVPROC glad_glColor3usv;
+#define glColor3usv glad_glColor3usv
+GLAD_API_CALL PFNGLCOLOR4BPROC glad_glColor4b;
+#define glColor4b glad_glColor4b
+GLAD_API_CALL PFNGLCOLOR4BVPROC glad_glColor4bv;
+#define glColor4bv glad_glColor4bv
+GLAD_API_CALL PFNGLCOLOR4DPROC glad_glColor4d;
+#define glColor4d glad_glColor4d
+GLAD_API_CALL PFNGLCOLOR4DVPROC glad_glColor4dv;
+#define glColor4dv glad_glColor4dv
+GLAD_API_CALL PFNGLCOLOR4FPROC glad_glColor4f;
+#define glColor4f glad_glColor4f
+GLAD_API_CALL PFNGLCOLOR4FVPROC glad_glColor4fv;
+#define glColor4fv glad_glColor4fv
+GLAD_API_CALL PFNGLCOLOR4IPROC glad_glColor4i;
+#define glColor4i glad_glColor4i
+GLAD_API_CALL PFNGLCOLOR4IVPROC glad_glColor4iv;
+#define glColor4iv glad_glColor4iv
+GLAD_API_CALL PFNGLCOLOR4SPROC glad_glColor4s;
+#define glColor4s glad_glColor4s
+GLAD_API_CALL PFNGLCOLOR4SVPROC glad_glColor4sv;
+#define glColor4sv glad_glColor4sv
+GLAD_API_CALL PFNGLCOLOR4UBPROC glad_glColor4ub;
+#define glColor4ub glad_glColor4ub
+GLAD_API_CALL PFNGLCOLOR4UBVPROC glad_glColor4ubv;
+#define glColor4ubv glad_glColor4ubv
+GLAD_API_CALL PFNGLCOLOR4UIPROC glad_glColor4ui;
+#define glColor4ui glad_glColor4ui
+GLAD_API_CALL PFNGLCOLOR4UIVPROC glad_glColor4uiv;
+#define glColor4uiv glad_glColor4uiv
+GLAD_API_CALL PFNGLCOLOR4USPROC glad_glColor4us;
+#define glColor4us glad_glColor4us
+GLAD_API_CALL PFNGLCOLOR4USVPROC glad_glColor4usv;
+#define glColor4usv glad_glColor4usv
+GLAD_API_CALL PFNGLCOLORMASKPROC glad_glColorMask;
+#define glColorMask glad_glColorMask
+GLAD_API_CALL PFNGLCOLORMASKIPROC glad_glColorMaski;
+#define glColorMaski glad_glColorMaski
+GLAD_API_CALL PFNGLCOLORMATERIALPROC glad_glColorMaterial;
+#define glColorMaterial glad_glColorMaterial
+GLAD_API_CALL PFNGLCOLORP3UIPROC glad_glColorP3ui;
+#define glColorP3ui glad_glColorP3ui
+GLAD_API_CALL PFNGLCOLORP3UIVPROC glad_glColorP3uiv;
+#define glColorP3uiv glad_glColorP3uiv
+GLAD_API_CALL PFNGLCOLORP4UIPROC glad_glColorP4ui;
+#define glColorP4ui glad_glColorP4ui
+GLAD_API_CALL PFNGLCOLORP4UIVPROC glad_glColorP4uiv;
+#define glColorP4uiv glad_glColorP4uiv
+GLAD_API_CALL PFNGLCOLORPOINTERPROC glad_glColorPointer;
+#define glColorPointer glad_glColorPointer
+GLAD_API_CALL PFNGLCOMPILESHADERPROC glad_glCompileShader;
+#define glCompileShader glad_glCompileShader
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D;
+#define glCompressedTexImage1D glad_glCompressedTexImage1D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D;
+#define glCompressedTexImage2D glad_glCompressedTexImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D;
+#define glCompressedTexImage3D glad_glCompressedTexImage3D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D;
+#define glCompressedTexSubImage1D glad_glCompressedTexSubImage1D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D;
+#define glCompressedTexSubImage2D glad_glCompressedTexSubImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D;
+#define glCompressedTexSubImage3D glad_glCompressedTexSubImage3D
+GLAD_API_CALL PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData;
+#define glCopyBufferSubData glad_glCopyBufferSubData
+GLAD_API_CALL PFNGLCOPYPIXELSPROC glad_glCopyPixels;
+#define glCopyPixels glad_glCopyPixels
+GLAD_API_CALL PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D;
+#define glCopyTexImage1D glad_glCopyTexImage1D
+GLAD_API_CALL PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D;
+#define glCopyTexImage2D glad_glCopyTexImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D;
+#define glCopyTexSubImage1D glad_glCopyTexSubImage1D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D;
+#define glCopyTexSubImage2D glad_glCopyTexSubImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D;
+#define glCopyTexSubImage3D glad_glCopyTexSubImage3D
+GLAD_API_CALL PFNGLCREATEPROGRAMPROC glad_glCreateProgram;
+#define glCreateProgram glad_glCreateProgram
+GLAD_API_CALL PFNGLCREATESHADERPROC glad_glCreateShader;
+#define glCreateShader glad_glCreateShader
+GLAD_API_CALL PFNGLCULLFACEPROC glad_glCullFace;
+#define glCullFace glad_glCullFace
+GLAD_API_CALL PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers;
+#define glDeleteBuffers glad_glDeleteBuffers
+GLAD_API_CALL PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers;
+#define glDeleteFramebuffers glad_glDeleteFramebuffers
+GLAD_API_CALL PFNGLDELETELISTSPROC glad_glDeleteLists;
+#define glDeleteLists glad_glDeleteLists
+GLAD_API_CALL PFNGLDELETEPROGRAMPROC glad_glDeleteProgram;
+#define glDeleteProgram glad_glDeleteProgram
+GLAD_API_CALL PFNGLDELETEQUERIESPROC glad_glDeleteQueries;
+#define glDeleteQueries glad_glDeleteQueries
+GLAD_API_CALL PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers;
+#define glDeleteRenderbuffers glad_glDeleteRenderbuffers
+GLAD_API_CALL PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers;
+#define glDeleteSamplers glad_glDeleteSamplers
+GLAD_API_CALL PFNGLDELETESHADERPROC glad_glDeleteShader;
+#define glDeleteShader glad_glDeleteShader
+GLAD_API_CALL PFNGLDELETESYNCPROC glad_glDeleteSync;
+#define glDeleteSync glad_glDeleteSync
+GLAD_API_CALL PFNGLDELETETEXTURESPROC glad_glDeleteTextures;
+#define glDeleteTextures glad_glDeleteTextures
+GLAD_API_CALL PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays;
+#define glDeleteVertexArrays glad_glDeleteVertexArrays
+GLAD_API_CALL PFNGLDEPTHFUNCPROC glad_glDepthFunc;
+#define glDepthFunc glad_glDepthFunc
+GLAD_API_CALL PFNGLDEPTHMASKPROC glad_glDepthMask;
+#define glDepthMask glad_glDepthMask
+GLAD_API_CALL PFNGLDEPTHRANGEPROC glad_glDepthRange;
+#define glDepthRange glad_glDepthRange
+GLAD_API_CALL PFNGLDETACHSHADERPROC glad_glDetachShader;
+#define glDetachShader glad_glDetachShader
+GLAD_API_CALL PFNGLDISABLEPROC glad_glDisable;
+#define glDisable glad_glDisable
+GLAD_API_CALL PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState;
+#define glDisableClientState glad_glDisableClientState
+GLAD_API_CALL PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray;
+#define glDisableVertexAttribArray glad_glDisableVertexAttribArray
+GLAD_API_CALL PFNGLDISABLEIPROC glad_glDisablei;
+#define glDisablei glad_glDisablei
+GLAD_API_CALL PFNGLDRAWARRAYSPROC glad_glDrawArrays;
+#define glDrawArrays glad_glDrawArrays
+GLAD_API_CALL PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced;
+#define glDrawArraysInstanced glad_glDrawArraysInstanced
+GLAD_API_CALL PFNGLDRAWBUFFERPROC glad_glDrawBuffer;
+#define glDrawBuffer glad_glDrawBuffer
+GLAD_API_CALL PFNGLDRAWBUFFERSPROC glad_glDrawBuffers;
+#define glDrawBuffers glad_glDrawBuffers
+GLAD_API_CALL PFNGLDRAWELEMENTSPROC glad_glDrawElements;
+#define glDrawElements glad_glDrawElements
+GLAD_API_CALL PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex;
+#define glDrawElementsBaseVertex glad_glDrawElementsBaseVertex
+GLAD_API_CALL PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced;
+#define glDrawElementsInstanced glad_glDrawElementsInstanced
+GLAD_API_CALL PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex;
+#define glDrawElementsInstancedBaseVertex glad_glDrawElementsInstancedBaseVertex
+GLAD_API_CALL PFNGLDRAWPIXELSPROC glad_glDrawPixels;
+#define glDrawPixels glad_glDrawPixels
+GLAD_API_CALL PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements;
+#define glDrawRangeElements glad_glDrawRangeElements
+GLAD_API_CALL PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex;
+#define glDrawRangeElementsBaseVertex glad_glDrawRangeElementsBaseVertex
+GLAD_API_CALL PFNGLEDGEFLAGPROC glad_glEdgeFlag;
+#define glEdgeFlag glad_glEdgeFlag
+GLAD_API_CALL PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer;
+#define glEdgeFlagPointer glad_glEdgeFlagPointer
+GLAD_API_CALL PFNGLEDGEFLAGVPROC glad_glEdgeFlagv;
+#define glEdgeFlagv glad_glEdgeFlagv
+GLAD_API_CALL PFNGLENABLEPROC glad_glEnable;
+#define glEnable glad_glEnable
+GLAD_API_CALL PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState;
+#define glEnableClientState glad_glEnableClientState
+GLAD_API_CALL PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray;
+#define glEnableVertexAttribArray glad_glEnableVertexAttribArray
+GLAD_API_CALL PFNGLENABLEIPROC glad_glEnablei;
+#define glEnablei glad_glEnablei
+GLAD_API_CALL PFNGLENDPROC glad_glEnd;
+#define glEnd glad_glEnd
+GLAD_API_CALL PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender;
+#define glEndConditionalRender glad_glEndConditionalRender
+GLAD_API_CALL PFNGLENDLISTPROC glad_glEndList;
+#define glEndList glad_glEndList
+GLAD_API_CALL PFNGLENDQUERYPROC glad_glEndQuery;
+#define glEndQuery glad_glEndQuery
+GLAD_API_CALL PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback;
+#define glEndTransformFeedback glad_glEndTransformFeedback
+GLAD_API_CALL PFNGLEVALCOORD1DPROC glad_glEvalCoord1d;
+#define glEvalCoord1d glad_glEvalCoord1d
+GLAD_API_CALL PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv;
+#define glEvalCoord1dv glad_glEvalCoord1dv
+GLAD_API_CALL PFNGLEVALCOORD1FPROC glad_glEvalCoord1f;
+#define glEvalCoord1f glad_glEvalCoord1f
+GLAD_API_CALL PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv;
+#define glEvalCoord1fv glad_glEvalCoord1fv
+GLAD_API_CALL PFNGLEVALCOORD2DPROC glad_glEvalCoord2d;
+#define glEvalCoord2d glad_glEvalCoord2d
+GLAD_API_CALL PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv;
+#define glEvalCoord2dv glad_glEvalCoord2dv
+GLAD_API_CALL PFNGLEVALCOORD2FPROC glad_glEvalCoord2f;
+#define glEvalCoord2f glad_glEvalCoord2f
+GLAD_API_CALL PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv;
+#define glEvalCoord2fv glad_glEvalCoord2fv
+GLAD_API_CALL PFNGLEVALMESH1PROC glad_glEvalMesh1;
+#define glEvalMesh1 glad_glEvalMesh1
+GLAD_API_CALL PFNGLEVALMESH2PROC glad_glEvalMesh2;
+#define glEvalMesh2 glad_glEvalMesh2
+GLAD_API_CALL PFNGLEVALPOINT1PROC glad_glEvalPoint1;
+#define glEvalPoint1 glad_glEvalPoint1
+GLAD_API_CALL PFNGLEVALPOINT2PROC glad_glEvalPoint2;
+#define glEvalPoint2 glad_glEvalPoint2
+GLAD_API_CALL PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer;
+#define glFeedbackBuffer glad_glFeedbackBuffer
+GLAD_API_CALL PFNGLFENCESYNCPROC glad_glFenceSync;
+#define glFenceSync glad_glFenceSync
+GLAD_API_CALL PFNGLFINISHPROC glad_glFinish;
+#define glFinish glad_glFinish
+GLAD_API_CALL PFNGLFLUSHPROC glad_glFlush;
+#define glFlush glad_glFlush
+GLAD_API_CALL PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange;
+#define glFlushMappedBufferRange glad_glFlushMappedBufferRange
+GLAD_API_CALL PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer;
+#define glFogCoordPointer glad_glFogCoordPointer
+GLAD_API_CALL PFNGLFOGCOORDDPROC glad_glFogCoordd;
+#define glFogCoordd glad_glFogCoordd
+GLAD_API_CALL PFNGLFOGCOORDDVPROC glad_glFogCoorddv;
+#define glFogCoorddv glad_glFogCoorddv
+GLAD_API_CALL PFNGLFOGCOORDFPROC glad_glFogCoordf;
+#define glFogCoordf glad_glFogCoordf
+GLAD_API_CALL PFNGLFOGCOORDFVPROC glad_glFogCoordfv;
+#define glFogCoordfv glad_glFogCoordfv
+GLAD_API_CALL PFNGLFOGFPROC glad_glFogf;
+#define glFogf glad_glFogf
+GLAD_API_CALL PFNGLFOGFVPROC glad_glFogfv;
+#define glFogfv glad_glFogfv
+GLAD_API_CALL PFNGLFOGIPROC glad_glFogi;
+#define glFogi glad_glFogi
+GLAD_API_CALL PFNGLFOGIVPROC glad_glFogiv;
+#define glFogiv glad_glFogiv
+GLAD_API_CALL PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer;
+#define glFramebufferRenderbuffer glad_glFramebufferRenderbuffer
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture;
+#define glFramebufferTexture glad_glFramebufferTexture
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D;
+#define glFramebufferTexture1D glad_glFramebufferTexture1D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D;
+#define glFramebufferTexture2D glad_glFramebufferTexture2D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D;
+#define glFramebufferTexture3D glad_glFramebufferTexture3D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer;
+#define glFramebufferTextureLayer glad_glFramebufferTextureLayer
+GLAD_API_CALL PFNGLFRONTFACEPROC glad_glFrontFace;
+#define glFrontFace glad_glFrontFace
+GLAD_API_CALL PFNGLFRUSTUMPROC glad_glFrustum;
+#define glFrustum glad_glFrustum
+GLAD_API_CALL PFNGLGENBUFFERSPROC glad_glGenBuffers;
+#define glGenBuffers glad_glGenBuffers
+GLAD_API_CALL PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers;
+#define glGenFramebuffers glad_glGenFramebuffers
+GLAD_API_CALL PFNGLGENLISTSPROC glad_glGenLists;
+#define glGenLists glad_glGenLists
+GLAD_API_CALL PFNGLGENQUERIESPROC glad_glGenQueries;
+#define glGenQueries glad_glGenQueries
+GLAD_API_CALL PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers;
+#define glGenRenderbuffers glad_glGenRenderbuffers
+GLAD_API_CALL PFNGLGENSAMPLERSPROC glad_glGenSamplers;
+#define glGenSamplers glad_glGenSamplers
+GLAD_API_CALL PFNGLGENTEXTURESPROC glad_glGenTextures;
+#define glGenTextures glad_glGenTextures
+GLAD_API_CALL PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays;
+#define glGenVertexArrays glad_glGenVertexArrays
+GLAD_API_CALL PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap;
+#define glGenerateMipmap glad_glGenerateMipmap
+GLAD_API_CALL PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib;
+#define glGetActiveAttrib glad_glGetActiveAttrib
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform;
+#define glGetActiveUniform glad_glGetActiveUniform
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName;
+#define glGetActiveUniformBlockName glad_glGetActiveUniformBlockName
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv;
+#define glGetActiveUniformBlockiv glad_glGetActiveUniformBlockiv
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName;
+#define glGetActiveUniformName glad_glGetActiveUniformName
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv;
+#define glGetActiveUniformsiv glad_glGetActiveUniformsiv
+GLAD_API_CALL PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders;
+#define glGetAttachedShaders glad_glGetAttachedShaders
+GLAD_API_CALL PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation;
+#define glGetAttribLocation glad_glGetAttribLocation
+GLAD_API_CALL PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v;
+#define glGetBooleani_v glad_glGetBooleani_v
+GLAD_API_CALL PFNGLGETBOOLEANVPROC glad_glGetBooleanv;
+#define glGetBooleanv glad_glGetBooleanv
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v;
+#define glGetBufferParameteri64v glad_glGetBufferParameteri64v
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv;
+#define glGetBufferParameteriv glad_glGetBufferParameteriv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv;
+#define glGetBufferPointerv glad_glGetBufferPointerv
+GLAD_API_CALL PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData;
+#define glGetBufferSubData glad_glGetBufferSubData
+GLAD_API_CALL PFNGLGETCLIPPLANEPROC glad_glGetClipPlane;
+#define glGetClipPlane glad_glGetClipPlane
+GLAD_API_CALL PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage;
+#define glGetCompressedTexImage glad_glGetCompressedTexImage
+GLAD_API_CALL PFNGLGETDOUBLEVPROC glad_glGetDoublev;
+#define glGetDoublev glad_glGetDoublev
+GLAD_API_CALL PFNGLGETERRORPROC glad_glGetError;
+#define glGetError glad_glGetError
+GLAD_API_CALL PFNGLGETFLOATVPROC glad_glGetFloatv;
+#define glGetFloatv glad_glGetFloatv
+GLAD_API_CALL PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex;
+#define glGetFragDataIndex glad_glGetFragDataIndex
+GLAD_API_CALL PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation;
+#define glGetFragDataLocation glad_glGetFragDataLocation
+GLAD_API_CALL PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv;
+#define glGetFramebufferAttachmentParameteriv glad_glGetFramebufferAttachmentParameteriv
+GLAD_API_CALL PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v;
+#define glGetInteger64i_v glad_glGetInteger64i_v
+GLAD_API_CALL PFNGLGETINTEGER64VPROC glad_glGetInteger64v;
+#define glGetInteger64v glad_glGetInteger64v
+GLAD_API_CALL PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v;
+#define glGetIntegeri_v glad_glGetIntegeri_v
+GLAD_API_CALL PFNGLGETINTEGERVPROC glad_glGetIntegerv;
+#define glGetIntegerv glad_glGetIntegerv
+GLAD_API_CALL PFNGLGETLIGHTFVPROC glad_glGetLightfv;
+#define glGetLightfv glad_glGetLightfv
+GLAD_API_CALL PFNGLGETLIGHTIVPROC glad_glGetLightiv;
+#define glGetLightiv glad_glGetLightiv
+GLAD_API_CALL PFNGLGETMAPDVPROC glad_glGetMapdv;
+#define glGetMapdv glad_glGetMapdv
+GLAD_API_CALL PFNGLGETMAPFVPROC glad_glGetMapfv;
+#define glGetMapfv glad_glGetMapfv
+GLAD_API_CALL PFNGLGETMAPIVPROC glad_glGetMapiv;
+#define glGetMapiv glad_glGetMapiv
+GLAD_API_CALL PFNGLGETMATERIALFVPROC glad_glGetMaterialfv;
+#define glGetMaterialfv glad_glGetMaterialfv
+GLAD_API_CALL PFNGLGETMATERIALIVPROC glad_glGetMaterialiv;
+#define glGetMaterialiv glad_glGetMaterialiv
+GLAD_API_CALL PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv;
+#define glGetMultisamplefv glad_glGetMultisamplefv
+GLAD_API_CALL PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv;
+#define glGetPixelMapfv glad_glGetPixelMapfv
+GLAD_API_CALL PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv;
+#define glGetPixelMapuiv glad_glGetPixelMapuiv
+GLAD_API_CALL PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv;
+#define glGetPixelMapusv glad_glGetPixelMapusv
+GLAD_API_CALL PFNGLGETPOINTERVPROC glad_glGetPointerv;
+#define glGetPointerv glad_glGetPointerv
+GLAD_API_CALL PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple;
+#define glGetPolygonStipple glad_glGetPolygonStipple
+GLAD_API_CALL PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog;
+#define glGetProgramInfoLog glad_glGetProgramInfoLog
+GLAD_API_CALL PFNGLGETPROGRAMIVPROC glad_glGetProgramiv;
+#define glGetProgramiv glad_glGetProgramiv
+GLAD_API_CALL PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v;
+#define glGetQueryObjecti64v glad_glGetQueryObjecti64v
+GLAD_API_CALL PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv;
+#define glGetQueryObjectiv glad_glGetQueryObjectiv
+GLAD_API_CALL PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v;
+#define glGetQueryObjectui64v glad_glGetQueryObjectui64v
+GLAD_API_CALL PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv;
+#define glGetQueryObjectuiv glad_glGetQueryObjectuiv
+GLAD_API_CALL PFNGLGETQUERYIVPROC glad_glGetQueryiv;
+#define glGetQueryiv glad_glGetQueryiv
+GLAD_API_CALL PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv;
+#define glGetRenderbufferParameteriv glad_glGetRenderbufferParameteriv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv;
+#define glGetSamplerParameterIiv glad_glGetSamplerParameterIiv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv;
+#define glGetSamplerParameterIuiv glad_glGetSamplerParameterIuiv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv;
+#define glGetSamplerParameterfv glad_glGetSamplerParameterfv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv;
+#define glGetSamplerParameteriv glad_glGetSamplerParameteriv
+GLAD_API_CALL PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog;
+#define glGetShaderInfoLog glad_glGetShaderInfoLog
+GLAD_API_CALL PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource;
+#define glGetShaderSource glad_glGetShaderSource
+GLAD_API_CALL PFNGLGETSHADERIVPROC glad_glGetShaderiv;
+#define glGetShaderiv glad_glGetShaderiv
+GLAD_API_CALL PFNGLGETSTRINGPROC glad_glGetString;
+#define glGetString glad_glGetString
+GLAD_API_CALL PFNGLGETSTRINGIPROC glad_glGetStringi;
+#define glGetStringi glad_glGetStringi
+GLAD_API_CALL PFNGLGETSYNCIVPROC glad_glGetSynciv;
+#define glGetSynciv glad_glGetSynciv
+GLAD_API_CALL PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv;
+#define glGetTexEnvfv glad_glGetTexEnvfv
+GLAD_API_CALL PFNGLGETTEXENVIVPROC glad_glGetTexEnviv;
+#define glGetTexEnviv glad_glGetTexEnviv
+GLAD_API_CALL PFNGLGETTEXGENDVPROC glad_glGetTexGendv;
+#define glGetTexGendv glad_glGetTexGendv
+GLAD_API_CALL PFNGLGETTEXGENFVPROC glad_glGetTexGenfv;
+#define glGetTexGenfv glad_glGetTexGenfv
+GLAD_API_CALL PFNGLGETTEXGENIVPROC glad_glGetTexGeniv;
+#define glGetTexGeniv glad_glGetTexGeniv
+GLAD_API_CALL PFNGLGETTEXIMAGEPROC glad_glGetTexImage;
+#define glGetTexImage glad_glGetTexImage
+GLAD_API_CALL PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv;
+#define glGetTexLevelParameterfv glad_glGetTexLevelParameterfv
+GLAD_API_CALL PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv;
+#define glGetTexLevelParameteriv glad_glGetTexLevelParameteriv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv;
+#define glGetTexParameterIiv glad_glGetTexParameterIiv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv;
+#define glGetTexParameterIuiv glad_glGetTexParameterIuiv
+GLAD_API_CALL PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv;
+#define glGetTexParameterfv glad_glGetTexParameterfv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv;
+#define glGetTexParameteriv glad_glGetTexParameteriv
+GLAD_API_CALL PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying;
+#define glGetTransformFeedbackVarying glad_glGetTransformFeedbackVarying
+GLAD_API_CALL PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex;
+#define glGetUniformBlockIndex glad_glGetUniformBlockIndex
+GLAD_API_CALL PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices;
+#define glGetUniformIndices glad_glGetUniformIndices
+GLAD_API_CALL PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation;
+#define glGetUniformLocation glad_glGetUniformLocation
+GLAD_API_CALL PFNGLGETUNIFORMFVPROC glad_glGetUniformfv;
+#define glGetUniformfv glad_glGetUniformfv
+GLAD_API_CALL PFNGLGETUNIFORMIVPROC glad_glGetUniformiv;
+#define glGetUniformiv glad_glGetUniformiv
+GLAD_API_CALL PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv;
+#define glGetUniformuiv glad_glGetUniformuiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv;
+#define glGetVertexAttribIiv glad_glGetVertexAttribIiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv;
+#define glGetVertexAttribIuiv glad_glGetVertexAttribIuiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv;
+#define glGetVertexAttribPointerv glad_glGetVertexAttribPointerv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv;
+#define glGetVertexAttribdv glad_glGetVertexAttribdv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv;
+#define glGetVertexAttribfv glad_glGetVertexAttribfv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv;
+#define glGetVertexAttribiv glad_glGetVertexAttribiv
+GLAD_API_CALL PFNGLHINTPROC glad_glHint;
+#define glHint glad_glHint
+GLAD_API_CALL PFNGLINDEXMASKPROC glad_glIndexMask;
+#define glIndexMask glad_glIndexMask
+GLAD_API_CALL PFNGLINDEXPOINTERPROC glad_glIndexPointer;
+#define glIndexPointer glad_glIndexPointer
+GLAD_API_CALL PFNGLINDEXDPROC glad_glIndexd;
+#define glIndexd glad_glIndexd
+GLAD_API_CALL PFNGLINDEXDVPROC glad_glIndexdv;
+#define glIndexdv glad_glIndexdv
+GLAD_API_CALL PFNGLINDEXFPROC glad_glIndexf;
+#define glIndexf glad_glIndexf
+GLAD_API_CALL PFNGLINDEXFVPROC glad_glIndexfv;
+#define glIndexfv glad_glIndexfv
+GLAD_API_CALL PFNGLINDEXIPROC glad_glIndexi;
+#define glIndexi glad_glIndexi
+GLAD_API_CALL PFNGLINDEXIVPROC glad_glIndexiv;
+#define glIndexiv glad_glIndexiv
+GLAD_API_CALL PFNGLINDEXSPROC glad_glIndexs;
+#define glIndexs glad_glIndexs
+GLAD_API_CALL PFNGLINDEXSVPROC glad_glIndexsv;
+#define glIndexsv glad_glIndexsv
+GLAD_API_CALL PFNGLINDEXUBPROC glad_glIndexub;
+#define glIndexub glad_glIndexub
+GLAD_API_CALL PFNGLINDEXUBVPROC glad_glIndexubv;
+#define glIndexubv glad_glIndexubv
+GLAD_API_CALL PFNGLINITNAMESPROC glad_glInitNames;
+#define glInitNames glad_glInitNames
+GLAD_API_CALL PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays;
+#define glInterleavedArrays glad_glInterleavedArrays
+GLAD_API_CALL PFNGLISBUFFERPROC glad_glIsBuffer;
+#define glIsBuffer glad_glIsBuffer
+GLAD_API_CALL PFNGLISENABLEDPROC glad_glIsEnabled;
+#define glIsEnabled glad_glIsEnabled
+GLAD_API_CALL PFNGLISENABLEDIPROC glad_glIsEnabledi;
+#define glIsEnabledi glad_glIsEnabledi
+GLAD_API_CALL PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer;
+#define glIsFramebuffer glad_glIsFramebuffer
+GLAD_API_CALL PFNGLISLISTPROC glad_glIsList;
+#define glIsList glad_glIsList
+GLAD_API_CALL PFNGLISPROGRAMPROC glad_glIsProgram;
+#define glIsProgram glad_glIsProgram
+GLAD_API_CALL PFNGLISQUERYPROC glad_glIsQuery;
+#define glIsQuery glad_glIsQuery
+GLAD_API_CALL PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer;
+#define glIsRenderbuffer glad_glIsRenderbuffer
+GLAD_API_CALL PFNGLISSAMPLERPROC glad_glIsSampler;
+#define glIsSampler glad_glIsSampler
+GLAD_API_CALL PFNGLISSHADERPROC glad_glIsShader;
+#define glIsShader glad_glIsShader
+GLAD_API_CALL PFNGLISSYNCPROC glad_glIsSync;
+#define glIsSync glad_glIsSync
+GLAD_API_CALL PFNGLISTEXTUREPROC glad_glIsTexture;
+#define glIsTexture glad_glIsTexture
+GLAD_API_CALL PFNGLISVERTEXARRAYPROC glad_glIsVertexArray;
+#define glIsVertexArray glad_glIsVertexArray
+GLAD_API_CALL PFNGLLIGHTMODELFPROC glad_glLightModelf;
+#define glLightModelf glad_glLightModelf
+GLAD_API_CALL PFNGLLIGHTMODELFVPROC glad_glLightModelfv;
+#define glLightModelfv glad_glLightModelfv
+GLAD_API_CALL PFNGLLIGHTMODELIPROC glad_glLightModeli;
+#define glLightModeli glad_glLightModeli
+GLAD_API_CALL PFNGLLIGHTMODELIVPROC glad_glLightModeliv;
+#define glLightModeliv glad_glLightModeliv
+GLAD_API_CALL PFNGLLIGHTFPROC glad_glLightf;
+#define glLightf glad_glLightf
+GLAD_API_CALL PFNGLLIGHTFVPROC glad_glLightfv;
+#define glLightfv glad_glLightfv
+GLAD_API_CALL PFNGLLIGHTIPROC glad_glLighti;
+#define glLighti glad_glLighti
+GLAD_API_CALL PFNGLLIGHTIVPROC glad_glLightiv;
+#define glLightiv glad_glLightiv
+GLAD_API_CALL PFNGLLINESTIPPLEPROC glad_glLineStipple;
+#define glLineStipple glad_glLineStipple
+GLAD_API_CALL PFNGLLINEWIDTHPROC glad_glLineWidth;
+#define glLineWidth glad_glLineWidth
+GLAD_API_CALL PFNGLLINKPROGRAMPROC glad_glLinkProgram;
+#define glLinkProgram glad_glLinkProgram
+GLAD_API_CALL PFNGLLISTBASEPROC glad_glListBase;
+#define glListBase glad_glListBase
+GLAD_API_CALL PFNGLLOADIDENTITYPROC glad_glLoadIdentity;
+#define glLoadIdentity glad_glLoadIdentity
+GLAD_API_CALL PFNGLLOADMATRIXDPROC glad_glLoadMatrixd;
+#define glLoadMatrixd glad_glLoadMatrixd
+GLAD_API_CALL PFNGLLOADMATRIXFPROC glad_glLoadMatrixf;
+#define glLoadMatrixf glad_glLoadMatrixf
+GLAD_API_CALL PFNGLLOADNAMEPROC glad_glLoadName;
+#define glLoadName glad_glLoadName
+GLAD_API_CALL PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd;
+#define glLoadTransposeMatrixd glad_glLoadTransposeMatrixd
+GLAD_API_CALL PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf;
+#define glLoadTransposeMatrixf glad_glLoadTransposeMatrixf
+GLAD_API_CALL PFNGLLOGICOPPROC glad_glLogicOp;
+#define glLogicOp glad_glLogicOp
+GLAD_API_CALL PFNGLMAP1DPROC glad_glMap1d;
+#define glMap1d glad_glMap1d
+GLAD_API_CALL PFNGLMAP1FPROC glad_glMap1f;
+#define glMap1f glad_glMap1f
+GLAD_API_CALL PFNGLMAP2DPROC glad_glMap2d;
+#define glMap2d glad_glMap2d
+GLAD_API_CALL PFNGLMAP2FPROC glad_glMap2f;
+#define glMap2f glad_glMap2f
+GLAD_API_CALL PFNGLMAPBUFFERPROC glad_glMapBuffer;
+#define glMapBuffer glad_glMapBuffer
+GLAD_API_CALL PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange;
+#define glMapBufferRange glad_glMapBufferRange
+GLAD_API_CALL PFNGLMAPGRID1DPROC glad_glMapGrid1d;
+#define glMapGrid1d glad_glMapGrid1d
+GLAD_API_CALL PFNGLMAPGRID1FPROC glad_glMapGrid1f;
+#define glMapGrid1f glad_glMapGrid1f
+GLAD_API_CALL PFNGLMAPGRID2DPROC glad_glMapGrid2d;
+#define glMapGrid2d glad_glMapGrid2d
+GLAD_API_CALL PFNGLMAPGRID2FPROC glad_glMapGrid2f;
+#define glMapGrid2f glad_glMapGrid2f
+GLAD_API_CALL PFNGLMATERIALFPROC glad_glMaterialf;
+#define glMaterialf glad_glMaterialf
+GLAD_API_CALL PFNGLMATERIALFVPROC glad_glMaterialfv;
+#define glMaterialfv glad_glMaterialfv
+GLAD_API_CALL PFNGLMATERIALIPROC glad_glMateriali;
+#define glMateriali glad_glMateriali
+GLAD_API_CALL PFNGLMATERIALIVPROC glad_glMaterialiv;
+#define glMaterialiv glad_glMaterialiv
+GLAD_API_CALL PFNGLMATRIXMODEPROC glad_glMatrixMode;
+#define glMatrixMode glad_glMatrixMode
+GLAD_API_CALL PFNGLMULTMATRIXDPROC glad_glMultMatrixd;
+#define glMultMatrixd glad_glMultMatrixd
+GLAD_API_CALL PFNGLMULTMATRIXFPROC glad_glMultMatrixf;
+#define glMultMatrixf glad_glMultMatrixf
+GLAD_API_CALL PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd;
+#define glMultTransposeMatrixd glad_glMultTransposeMatrixd
+GLAD_API_CALL PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf;
+#define glMultTransposeMatrixf glad_glMultTransposeMatrixf
+GLAD_API_CALL PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays;
+#define glMultiDrawArrays glad_glMultiDrawArrays
+GLAD_API_CALL PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements;
+#define glMultiDrawElements glad_glMultiDrawElements
+GLAD_API_CALL PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex;
+#define glMultiDrawElementsBaseVertex glad_glMultiDrawElementsBaseVertex
+GLAD_API_CALL PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d;
+#define glMultiTexCoord1d glad_glMultiTexCoord1d
+GLAD_API_CALL PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv;
+#define glMultiTexCoord1dv glad_glMultiTexCoord1dv
+GLAD_API_CALL PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f;
+#define glMultiTexCoord1f glad_glMultiTexCoord1f
+GLAD_API_CALL PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv;
+#define glMultiTexCoord1fv glad_glMultiTexCoord1fv
+GLAD_API_CALL PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i;
+#define glMultiTexCoord1i glad_glMultiTexCoord1i
+GLAD_API_CALL PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv;
+#define glMultiTexCoord1iv glad_glMultiTexCoord1iv
+GLAD_API_CALL PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s;
+#define glMultiTexCoord1s glad_glMultiTexCoord1s
+GLAD_API_CALL PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv;
+#define glMultiTexCoord1sv glad_glMultiTexCoord1sv
+GLAD_API_CALL PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d;
+#define glMultiTexCoord2d glad_glMultiTexCoord2d
+GLAD_API_CALL PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv;
+#define glMultiTexCoord2dv glad_glMultiTexCoord2dv
+GLAD_API_CALL PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f;
+#define glMultiTexCoord2f glad_glMultiTexCoord2f
+GLAD_API_CALL PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv;
+#define glMultiTexCoord2fv glad_glMultiTexCoord2fv
+GLAD_API_CALL PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i;
+#define glMultiTexCoord2i glad_glMultiTexCoord2i
+GLAD_API_CALL PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv;
+#define glMultiTexCoord2iv glad_glMultiTexCoord2iv
+GLAD_API_CALL PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s;
+#define glMultiTexCoord2s glad_glMultiTexCoord2s
+GLAD_API_CALL PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv;
+#define glMultiTexCoord2sv glad_glMultiTexCoord2sv
+GLAD_API_CALL PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d;
+#define glMultiTexCoord3d glad_glMultiTexCoord3d
+GLAD_API_CALL PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv;
+#define glMultiTexCoord3dv glad_glMultiTexCoord3dv
+GLAD_API_CALL PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f;
+#define glMultiTexCoord3f glad_glMultiTexCoord3f
+GLAD_API_CALL PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv;
+#define glMultiTexCoord3fv glad_glMultiTexCoord3fv
+GLAD_API_CALL PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i;
+#define glMultiTexCoord3i glad_glMultiTexCoord3i
+GLAD_API_CALL PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv;
+#define glMultiTexCoord3iv glad_glMultiTexCoord3iv
+GLAD_API_CALL PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s;
+#define glMultiTexCoord3s glad_glMultiTexCoord3s
+GLAD_API_CALL PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv;
+#define glMultiTexCoord3sv glad_glMultiTexCoord3sv
+GLAD_API_CALL PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d;
+#define glMultiTexCoord4d glad_glMultiTexCoord4d
+GLAD_API_CALL PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv;
+#define glMultiTexCoord4dv glad_glMultiTexCoord4dv
+GLAD_API_CALL PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f;
+#define glMultiTexCoord4f glad_glMultiTexCoord4f
+GLAD_API_CALL PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv;
+#define glMultiTexCoord4fv glad_glMultiTexCoord4fv
+GLAD_API_CALL PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i;
+#define glMultiTexCoord4i glad_glMultiTexCoord4i
+GLAD_API_CALL PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv;
+#define glMultiTexCoord4iv glad_glMultiTexCoord4iv
+GLAD_API_CALL PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s;
+#define glMultiTexCoord4s glad_glMultiTexCoord4s
+GLAD_API_CALL PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv;
+#define glMultiTexCoord4sv glad_glMultiTexCoord4sv
+GLAD_API_CALL PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui;
+#define glMultiTexCoordP1ui glad_glMultiTexCoordP1ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv;
+#define glMultiTexCoordP1uiv glad_glMultiTexCoordP1uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui;
+#define glMultiTexCoordP2ui glad_glMultiTexCoordP2ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv;
+#define glMultiTexCoordP2uiv glad_glMultiTexCoordP2uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui;
+#define glMultiTexCoordP3ui glad_glMultiTexCoordP3ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv;
+#define glMultiTexCoordP3uiv glad_glMultiTexCoordP3uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui;
+#define glMultiTexCoordP4ui glad_glMultiTexCoordP4ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv;
+#define glMultiTexCoordP4uiv glad_glMultiTexCoordP4uiv
+GLAD_API_CALL PFNGLNEWLISTPROC glad_glNewList;
+#define glNewList glad_glNewList
+GLAD_API_CALL PFNGLNORMAL3BPROC glad_glNormal3b;
+#define glNormal3b glad_glNormal3b
+GLAD_API_CALL PFNGLNORMAL3BVPROC glad_glNormal3bv;
+#define glNormal3bv glad_glNormal3bv
+GLAD_API_CALL PFNGLNORMAL3DPROC glad_glNormal3d;
+#define glNormal3d glad_glNormal3d
+GLAD_API_CALL PFNGLNORMAL3DVPROC glad_glNormal3dv;
+#define glNormal3dv glad_glNormal3dv
+GLAD_API_CALL PFNGLNORMAL3FPROC glad_glNormal3f;
+#define glNormal3f glad_glNormal3f
+GLAD_API_CALL PFNGLNORMAL3FVPROC glad_glNormal3fv;
+#define glNormal3fv glad_glNormal3fv
+GLAD_API_CALL PFNGLNORMAL3IPROC glad_glNormal3i;
+#define glNormal3i glad_glNormal3i
+GLAD_API_CALL PFNGLNORMAL3IVPROC glad_glNormal3iv;
+#define glNormal3iv glad_glNormal3iv
+GLAD_API_CALL PFNGLNORMAL3SPROC glad_glNormal3s;
+#define glNormal3s glad_glNormal3s
+GLAD_API_CALL PFNGLNORMAL3SVPROC glad_glNormal3sv;
+#define glNormal3sv glad_glNormal3sv
+GLAD_API_CALL PFNGLNORMALP3UIPROC glad_glNormalP3ui;
+#define glNormalP3ui glad_glNormalP3ui
+GLAD_API_CALL PFNGLNORMALP3UIVPROC glad_glNormalP3uiv;
+#define glNormalP3uiv glad_glNormalP3uiv
+GLAD_API_CALL PFNGLNORMALPOINTERPROC glad_glNormalPointer;
+#define glNormalPointer glad_glNormalPointer
+GLAD_API_CALL PFNGLORTHOPROC glad_glOrtho;
+#define glOrtho glad_glOrtho
+GLAD_API_CALL PFNGLPASSTHROUGHPROC glad_glPassThrough;
+#define glPassThrough glad_glPassThrough
+GLAD_API_CALL PFNGLPIXELMAPFVPROC glad_glPixelMapfv;
+#define glPixelMapfv glad_glPixelMapfv
+GLAD_API_CALL PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv;
+#define glPixelMapuiv glad_glPixelMapuiv
+GLAD_API_CALL PFNGLPIXELMAPUSVPROC glad_glPixelMapusv;
+#define glPixelMapusv glad_glPixelMapusv
+GLAD_API_CALL PFNGLPIXELSTOREFPROC glad_glPixelStoref;
+#define glPixelStoref glad_glPixelStoref
+GLAD_API_CALL PFNGLPIXELSTOREIPROC glad_glPixelStorei;
+#define glPixelStorei glad_glPixelStorei
+GLAD_API_CALL PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf;
+#define glPixelTransferf glad_glPixelTransferf
+GLAD_API_CALL PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi;
+#define glPixelTransferi glad_glPixelTransferi
+GLAD_API_CALL PFNGLPIXELZOOMPROC glad_glPixelZoom;
+#define glPixelZoom glad_glPixelZoom
+GLAD_API_CALL PFNGLPOINTPARAMETERFPROC glad_glPointParameterf;
+#define glPointParameterf glad_glPointParameterf
+GLAD_API_CALL PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv;
+#define glPointParameterfv glad_glPointParameterfv
+GLAD_API_CALL PFNGLPOINTPARAMETERIPROC glad_glPointParameteri;
+#define glPointParameteri glad_glPointParameteri
+GLAD_API_CALL PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv;
+#define glPointParameteriv glad_glPointParameteriv
+GLAD_API_CALL PFNGLPOINTSIZEPROC glad_glPointSize;
+#define glPointSize glad_glPointSize
+GLAD_API_CALL PFNGLPOLYGONMODEPROC glad_glPolygonMode;
+#define glPolygonMode glad_glPolygonMode
+GLAD_API_CALL PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset;
+#define glPolygonOffset glad_glPolygonOffset
+GLAD_API_CALL PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple;
+#define glPolygonStipple glad_glPolygonStipple
+GLAD_API_CALL PFNGLPOPATTRIBPROC glad_glPopAttrib;
+#define glPopAttrib glad_glPopAttrib
+GLAD_API_CALL PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib;
+#define glPopClientAttrib glad_glPopClientAttrib
+GLAD_API_CALL PFNGLPOPMATRIXPROC glad_glPopMatrix;
+#define glPopMatrix glad_glPopMatrix
+GLAD_API_CALL PFNGLPOPNAMEPROC glad_glPopName;
+#define glPopName glad_glPopName
+GLAD_API_CALL PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex;
+#define glPrimitiveRestartIndex glad_glPrimitiveRestartIndex
+GLAD_API_CALL PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures;
+#define glPrioritizeTextures glad_glPrioritizeTextures
+GLAD_API_CALL PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex;
+#define glProvokingVertex glad_glProvokingVertex
+GLAD_API_CALL PFNGLPUSHATTRIBPROC glad_glPushAttrib;
+#define glPushAttrib glad_glPushAttrib
+GLAD_API_CALL PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib;
+#define glPushClientAttrib glad_glPushClientAttrib
+GLAD_API_CALL PFNGLPUSHMATRIXPROC glad_glPushMatrix;
+#define glPushMatrix glad_glPushMatrix
+GLAD_API_CALL PFNGLPUSHNAMEPROC glad_glPushName;
+#define glPushName glad_glPushName
+GLAD_API_CALL PFNGLQUERYCOUNTERPROC glad_glQueryCounter;
+#define glQueryCounter glad_glQueryCounter
+GLAD_API_CALL PFNGLRASTERPOS2DPROC glad_glRasterPos2d;
+#define glRasterPos2d glad_glRasterPos2d
+GLAD_API_CALL PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv;
+#define glRasterPos2dv glad_glRasterPos2dv
+GLAD_API_CALL PFNGLRASTERPOS2FPROC glad_glRasterPos2f;
+#define glRasterPos2f glad_glRasterPos2f
+GLAD_API_CALL PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv;
+#define glRasterPos2fv glad_glRasterPos2fv
+GLAD_API_CALL PFNGLRASTERPOS2IPROC glad_glRasterPos2i;
+#define glRasterPos2i glad_glRasterPos2i
+GLAD_API_CALL PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv;
+#define glRasterPos2iv glad_glRasterPos2iv
+GLAD_API_CALL PFNGLRASTERPOS2SPROC glad_glRasterPos2s;
+#define glRasterPos2s glad_glRasterPos2s
+GLAD_API_CALL PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv;
+#define glRasterPos2sv glad_glRasterPos2sv
+GLAD_API_CALL PFNGLRASTERPOS3DPROC glad_glRasterPos3d;
+#define glRasterPos3d glad_glRasterPos3d
+GLAD_API_CALL PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv;
+#define glRasterPos3dv glad_glRasterPos3dv
+GLAD_API_CALL PFNGLRASTERPOS3FPROC glad_glRasterPos3f;
+#define glRasterPos3f glad_glRasterPos3f
+GLAD_API_CALL PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv;
+#define glRasterPos3fv glad_glRasterPos3fv
+GLAD_API_CALL PFNGLRASTERPOS3IPROC glad_glRasterPos3i;
+#define glRasterPos3i glad_glRasterPos3i
+GLAD_API_CALL PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv;
+#define glRasterPos3iv glad_glRasterPos3iv
+GLAD_API_CALL PFNGLRASTERPOS3SPROC glad_glRasterPos3s;
+#define glRasterPos3s glad_glRasterPos3s
+GLAD_API_CALL PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv;
+#define glRasterPos3sv glad_glRasterPos3sv
+GLAD_API_CALL PFNGLRASTERPOS4DPROC glad_glRasterPos4d;
+#define glRasterPos4d glad_glRasterPos4d
+GLAD_API_CALL PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv;
+#define glRasterPos4dv glad_glRasterPos4dv
+GLAD_API_CALL PFNGLRASTERPOS4FPROC glad_glRasterPos4f;
+#define glRasterPos4f glad_glRasterPos4f
+GLAD_API_CALL PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv;
+#define glRasterPos4fv glad_glRasterPos4fv
+GLAD_API_CALL PFNGLRASTERPOS4IPROC glad_glRasterPos4i;
+#define glRasterPos4i glad_glRasterPos4i
+GLAD_API_CALL PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv;
+#define glRasterPos4iv glad_glRasterPos4iv
+GLAD_API_CALL PFNGLRASTERPOS4SPROC glad_glRasterPos4s;
+#define glRasterPos4s glad_glRasterPos4s
+GLAD_API_CALL PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv;
+#define glRasterPos4sv glad_glRasterPos4sv
+GLAD_API_CALL PFNGLREADBUFFERPROC glad_glReadBuffer;
+#define glReadBuffer glad_glReadBuffer
+GLAD_API_CALL PFNGLREADPIXELSPROC glad_glReadPixels;
+#define glReadPixels glad_glReadPixels
+GLAD_API_CALL PFNGLRECTDPROC glad_glRectd;
+#define glRectd glad_glRectd
+GLAD_API_CALL PFNGLRECTDVPROC glad_glRectdv;
+#define glRectdv glad_glRectdv
+GLAD_API_CALL PFNGLRECTFPROC glad_glRectf;
+#define glRectf glad_glRectf
+GLAD_API_CALL PFNGLRECTFVPROC glad_glRectfv;
+#define glRectfv glad_glRectfv
+GLAD_API_CALL PFNGLRECTIPROC glad_glRecti;
+#define glRecti glad_glRecti
+GLAD_API_CALL PFNGLRECTIVPROC glad_glRectiv;
+#define glRectiv glad_glRectiv
+GLAD_API_CALL PFNGLRECTSPROC glad_glRects;
+#define glRects glad_glRects
+GLAD_API_CALL PFNGLRECTSVPROC glad_glRectsv;
+#define glRectsv glad_glRectsv
+GLAD_API_CALL PFNGLRENDERMODEPROC glad_glRenderMode;
+#define glRenderMode glad_glRenderMode
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage;
+#define glRenderbufferStorage glad_glRenderbufferStorage
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample;
+#define glRenderbufferStorageMultisample glad_glRenderbufferStorageMultisample
+GLAD_API_CALL PFNGLROTATEDPROC glad_glRotated;
+#define glRotated glad_glRotated
+GLAD_API_CALL PFNGLROTATEFPROC glad_glRotatef;
+#define glRotatef glad_glRotatef
+GLAD_API_CALL PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage;
+#define glSampleCoverage glad_glSampleCoverage
+GLAD_API_CALL PFNGLSAMPLEMASKIPROC glad_glSampleMaski;
+#define glSampleMaski glad_glSampleMaski
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv;
+#define glSamplerParameterIiv glad_glSamplerParameterIiv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv;
+#define glSamplerParameterIuiv glad_glSamplerParameterIuiv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf;
+#define glSamplerParameterf glad_glSamplerParameterf
+GLAD_API_CALL PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv;
+#define glSamplerParameterfv glad_glSamplerParameterfv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri;
+#define glSamplerParameteri glad_glSamplerParameteri
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv;
+#define glSamplerParameteriv glad_glSamplerParameteriv
+GLAD_API_CALL PFNGLSCALEDPROC glad_glScaled;
+#define glScaled glad_glScaled
+GLAD_API_CALL PFNGLSCALEFPROC glad_glScalef;
+#define glScalef glad_glScalef
+GLAD_API_CALL PFNGLSCISSORPROC glad_glScissor;
+#define glScissor glad_glScissor
+GLAD_API_CALL PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b;
+#define glSecondaryColor3b glad_glSecondaryColor3b
+GLAD_API_CALL PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv;
+#define glSecondaryColor3bv glad_glSecondaryColor3bv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d;
+#define glSecondaryColor3d glad_glSecondaryColor3d
+GLAD_API_CALL PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv;
+#define glSecondaryColor3dv glad_glSecondaryColor3dv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f;
+#define glSecondaryColor3f glad_glSecondaryColor3f
+GLAD_API_CALL PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv;
+#define glSecondaryColor3fv glad_glSecondaryColor3fv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i;
+#define glSecondaryColor3i glad_glSecondaryColor3i
+GLAD_API_CALL PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv;
+#define glSecondaryColor3iv glad_glSecondaryColor3iv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s;
+#define glSecondaryColor3s glad_glSecondaryColor3s
+GLAD_API_CALL PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv;
+#define glSecondaryColor3sv glad_glSecondaryColor3sv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub;
+#define glSecondaryColor3ub glad_glSecondaryColor3ub
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv;
+#define glSecondaryColor3ubv glad_glSecondaryColor3ubv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui;
+#define glSecondaryColor3ui glad_glSecondaryColor3ui
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv;
+#define glSecondaryColor3uiv glad_glSecondaryColor3uiv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us;
+#define glSecondaryColor3us glad_glSecondaryColor3us
+GLAD_API_CALL PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv;
+#define glSecondaryColor3usv glad_glSecondaryColor3usv
+GLAD_API_CALL PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui;
+#define glSecondaryColorP3ui glad_glSecondaryColorP3ui
+GLAD_API_CALL PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv;
+#define glSecondaryColorP3uiv glad_glSecondaryColorP3uiv
+GLAD_API_CALL PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer;
+#define glSecondaryColorPointer glad_glSecondaryColorPointer
+GLAD_API_CALL PFNGLSELECTBUFFERPROC glad_glSelectBuffer;
+#define glSelectBuffer glad_glSelectBuffer
+GLAD_API_CALL PFNGLSHADEMODELPROC glad_glShadeModel;
+#define glShadeModel glad_glShadeModel
+GLAD_API_CALL PFNGLSHADERSOURCEPROC glad_glShaderSource;
+#define glShaderSource glad_glShaderSource
+GLAD_API_CALL PFNGLSTENCILFUNCPROC glad_glStencilFunc;
+#define glStencilFunc glad_glStencilFunc
+GLAD_API_CALL PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate;
+#define glStencilFuncSeparate glad_glStencilFuncSeparate
+GLAD_API_CALL PFNGLSTENCILMASKPROC glad_glStencilMask;
+#define glStencilMask glad_glStencilMask
+GLAD_API_CALL PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate;
+#define glStencilMaskSeparate glad_glStencilMaskSeparate
+GLAD_API_CALL PFNGLSTENCILOPPROC glad_glStencilOp;
+#define glStencilOp glad_glStencilOp
+GLAD_API_CALL PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate;
+#define glStencilOpSeparate glad_glStencilOpSeparate
+GLAD_API_CALL PFNGLTEXBUFFERPROC glad_glTexBuffer;
+#define glTexBuffer glad_glTexBuffer
+GLAD_API_CALL PFNGLTEXCOORD1DPROC glad_glTexCoord1d;
+#define glTexCoord1d glad_glTexCoord1d
+GLAD_API_CALL PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv;
+#define glTexCoord1dv glad_glTexCoord1dv
+GLAD_API_CALL PFNGLTEXCOORD1FPROC glad_glTexCoord1f;
+#define glTexCoord1f glad_glTexCoord1f
+GLAD_API_CALL PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv;
+#define glTexCoord1fv glad_glTexCoord1fv
+GLAD_API_CALL PFNGLTEXCOORD1IPROC glad_glTexCoord1i;
+#define glTexCoord1i glad_glTexCoord1i
+GLAD_API_CALL PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv;
+#define glTexCoord1iv glad_glTexCoord1iv
+GLAD_API_CALL PFNGLTEXCOORD1SPROC glad_glTexCoord1s;
+#define glTexCoord1s glad_glTexCoord1s
+GLAD_API_CALL PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv;
+#define glTexCoord1sv glad_glTexCoord1sv
+GLAD_API_CALL PFNGLTEXCOORD2DPROC glad_glTexCoord2d;
+#define glTexCoord2d glad_glTexCoord2d
+GLAD_API_CALL PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv;
+#define glTexCoord2dv glad_glTexCoord2dv
+GLAD_API_CALL PFNGLTEXCOORD2FPROC glad_glTexCoord2f;
+#define glTexCoord2f glad_glTexCoord2f
+GLAD_API_CALL PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv;
+#define glTexCoord2fv glad_glTexCoord2fv
+GLAD_API_CALL PFNGLTEXCOORD2IPROC glad_glTexCoord2i;
+#define glTexCoord2i glad_glTexCoord2i
+GLAD_API_CALL PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv;
+#define glTexCoord2iv glad_glTexCoord2iv
+GLAD_API_CALL PFNGLTEXCOORD2SPROC glad_glTexCoord2s;
+#define glTexCoord2s glad_glTexCoord2s
+GLAD_API_CALL PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv;
+#define glTexCoord2sv glad_glTexCoord2sv
+GLAD_API_CALL PFNGLTEXCOORD3DPROC glad_glTexCoord3d;
+#define glTexCoord3d glad_glTexCoord3d
+GLAD_API_CALL PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv;
+#define glTexCoord3dv glad_glTexCoord3dv
+GLAD_API_CALL PFNGLTEXCOORD3FPROC glad_glTexCoord3f;
+#define glTexCoord3f glad_glTexCoord3f
+GLAD_API_CALL PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv;
+#define glTexCoord3fv glad_glTexCoord3fv
+GLAD_API_CALL PFNGLTEXCOORD3IPROC glad_glTexCoord3i;
+#define glTexCoord3i glad_glTexCoord3i
+GLAD_API_CALL PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv;
+#define glTexCoord3iv glad_glTexCoord3iv
+GLAD_API_CALL PFNGLTEXCOORD3SPROC glad_glTexCoord3s;
+#define glTexCoord3s glad_glTexCoord3s
+GLAD_API_CALL PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv;
+#define glTexCoord3sv glad_glTexCoord3sv
+GLAD_API_CALL PFNGLTEXCOORD4DPROC glad_glTexCoord4d;
+#define glTexCoord4d glad_glTexCoord4d
+GLAD_API_CALL PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv;
+#define glTexCoord4dv glad_glTexCoord4dv
+GLAD_API_CALL PFNGLTEXCOORD4FPROC glad_glTexCoord4f;
+#define glTexCoord4f glad_glTexCoord4f
+GLAD_API_CALL PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv;
+#define glTexCoord4fv glad_glTexCoord4fv
+GLAD_API_CALL PFNGLTEXCOORD4IPROC glad_glTexCoord4i;
+#define glTexCoord4i glad_glTexCoord4i
+GLAD_API_CALL PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv;
+#define glTexCoord4iv glad_glTexCoord4iv
+GLAD_API_CALL PFNGLTEXCOORD4SPROC glad_glTexCoord4s;
+#define glTexCoord4s glad_glTexCoord4s
+GLAD_API_CALL PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv;
+#define glTexCoord4sv glad_glTexCoord4sv
+GLAD_API_CALL PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui;
+#define glTexCoordP1ui glad_glTexCoordP1ui
+GLAD_API_CALL PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv;
+#define glTexCoordP1uiv glad_glTexCoordP1uiv
+GLAD_API_CALL PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui;
+#define glTexCoordP2ui glad_glTexCoordP2ui
+GLAD_API_CALL PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv;
+#define glTexCoordP2uiv glad_glTexCoordP2uiv
+GLAD_API_CALL PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui;
+#define glTexCoordP3ui glad_glTexCoordP3ui
+GLAD_API_CALL PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv;
+#define glTexCoordP3uiv glad_glTexCoordP3uiv
+GLAD_API_CALL PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui;
+#define glTexCoordP4ui glad_glTexCoordP4ui
+GLAD_API_CALL PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv;
+#define glTexCoordP4uiv glad_glTexCoordP4uiv
+GLAD_API_CALL PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer;
+#define glTexCoordPointer glad_glTexCoordPointer
+GLAD_API_CALL PFNGLTEXENVFPROC glad_glTexEnvf;
+#define glTexEnvf glad_glTexEnvf
+GLAD_API_CALL PFNGLTEXENVFVPROC glad_glTexEnvfv;
+#define glTexEnvfv glad_glTexEnvfv
+GLAD_API_CALL PFNGLTEXENVIPROC glad_glTexEnvi;
+#define glTexEnvi glad_glTexEnvi
+GLAD_API_CALL PFNGLTEXENVIVPROC glad_glTexEnviv;
+#define glTexEnviv glad_glTexEnviv
+GLAD_API_CALL PFNGLTEXGENDPROC glad_glTexGend;
+#define glTexGend glad_glTexGend
+GLAD_API_CALL PFNGLTEXGENDVPROC glad_glTexGendv;
+#define glTexGendv glad_glTexGendv
+GLAD_API_CALL PFNGLTEXGENFPROC glad_glTexGenf;
+#define glTexGenf glad_glTexGenf
+GLAD_API_CALL PFNGLTEXGENFVPROC glad_glTexGenfv;
+#define glTexGenfv glad_glTexGenfv
+GLAD_API_CALL PFNGLTEXGENIPROC glad_glTexGeni;
+#define glTexGeni glad_glTexGeni
+GLAD_API_CALL PFNGLTEXGENIVPROC glad_glTexGeniv;
+#define glTexGeniv glad_glTexGeniv
+GLAD_API_CALL PFNGLTEXIMAGE1DPROC glad_glTexImage1D;
+#define glTexImage1D glad_glTexImage1D
+GLAD_API_CALL PFNGLTEXIMAGE2DPROC glad_glTexImage2D;
+#define glTexImage2D glad_glTexImage2D
+GLAD_API_CALL PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample;
+#define glTexImage2DMultisample glad_glTexImage2DMultisample
+GLAD_API_CALL PFNGLTEXIMAGE3DPROC glad_glTexImage3D;
+#define glTexImage3D glad_glTexImage3D
+GLAD_API_CALL PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample;
+#define glTexImage3DMultisample glad_glTexImage3DMultisample
+GLAD_API_CALL PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv;
+#define glTexParameterIiv glad_glTexParameterIiv
+GLAD_API_CALL PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv;
+#define glTexParameterIuiv glad_glTexParameterIuiv
+GLAD_API_CALL PFNGLTEXPARAMETERFPROC glad_glTexParameterf;
+#define glTexParameterf glad_glTexParameterf
+GLAD_API_CALL PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv;
+#define glTexParameterfv glad_glTexParameterfv
+GLAD_API_CALL PFNGLTEXPARAMETERIPROC glad_glTexParameteri;
+#define glTexParameteri glad_glTexParameteri
+GLAD_API_CALL PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv;
+#define glTexParameteriv glad_glTexParameteriv
+GLAD_API_CALL PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D;
+#define glTexSubImage1D glad_glTexSubImage1D
+GLAD_API_CALL PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D;
+#define glTexSubImage2D glad_glTexSubImage2D
+GLAD_API_CALL PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D;
+#define glTexSubImage3D glad_glTexSubImage3D
+GLAD_API_CALL PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings;
+#define glTransformFeedbackVaryings glad_glTransformFeedbackVaryings
+GLAD_API_CALL PFNGLTRANSLATEDPROC glad_glTranslated;
+#define glTranslated glad_glTranslated
+GLAD_API_CALL PFNGLTRANSLATEFPROC glad_glTranslatef;
+#define glTranslatef glad_glTranslatef
+GLAD_API_CALL PFNGLUNIFORM1FPROC glad_glUniform1f;
+#define glUniform1f glad_glUniform1f
+GLAD_API_CALL PFNGLUNIFORM1FVPROC glad_glUniform1fv;
+#define glUniform1fv glad_glUniform1fv
+GLAD_API_CALL PFNGLUNIFORM1IPROC glad_glUniform1i;
+#define glUniform1i glad_glUniform1i
+GLAD_API_CALL PFNGLUNIFORM1IVPROC glad_glUniform1iv;
+#define glUniform1iv glad_glUniform1iv
+GLAD_API_CALL PFNGLUNIFORM1UIPROC glad_glUniform1ui;
+#define glUniform1ui glad_glUniform1ui
+GLAD_API_CALL PFNGLUNIFORM1UIVPROC glad_glUniform1uiv;
+#define glUniform1uiv glad_glUniform1uiv
+GLAD_API_CALL PFNGLUNIFORM2FPROC glad_glUniform2f;
+#define glUniform2f glad_glUniform2f
+GLAD_API_CALL PFNGLUNIFORM2FVPROC glad_glUniform2fv;
+#define glUniform2fv glad_glUniform2fv
+GLAD_API_CALL PFNGLUNIFORM2IPROC glad_glUniform2i;
+#define glUniform2i glad_glUniform2i
+GLAD_API_CALL PFNGLUNIFORM2IVPROC glad_glUniform2iv;
+#define glUniform2iv glad_glUniform2iv
+GLAD_API_CALL PFNGLUNIFORM2UIPROC glad_glUniform2ui;
+#define glUniform2ui glad_glUniform2ui
+GLAD_API_CALL PFNGLUNIFORM2UIVPROC glad_glUniform2uiv;
+#define glUniform2uiv glad_glUniform2uiv
+GLAD_API_CALL PFNGLUNIFORM3FPROC glad_glUniform3f;
+#define glUniform3f glad_glUniform3f
+GLAD_API_CALL PFNGLUNIFORM3FVPROC glad_glUniform3fv;
+#define glUniform3fv glad_glUniform3fv
+GLAD_API_CALL PFNGLUNIFORM3IPROC glad_glUniform3i;
+#define glUniform3i glad_glUniform3i
+GLAD_API_CALL PFNGLUNIFORM3IVPROC glad_glUniform3iv;
+#define glUniform3iv glad_glUniform3iv
+GLAD_API_CALL PFNGLUNIFORM3UIPROC glad_glUniform3ui;
+#define glUniform3ui glad_glUniform3ui
+GLAD_API_CALL PFNGLUNIFORM3UIVPROC glad_glUniform3uiv;
+#define glUniform3uiv glad_glUniform3uiv
+GLAD_API_CALL PFNGLUNIFORM4FPROC glad_glUniform4f;
+#define glUniform4f glad_glUniform4f
+GLAD_API_CALL PFNGLUNIFORM4FVPROC glad_glUniform4fv;
+#define glUniform4fv glad_glUniform4fv
+GLAD_API_CALL PFNGLUNIFORM4IPROC glad_glUniform4i;
+#define glUniform4i glad_glUniform4i
+GLAD_API_CALL PFNGLUNIFORM4IVPROC glad_glUniform4iv;
+#define glUniform4iv glad_glUniform4iv
+GLAD_API_CALL PFNGLUNIFORM4UIPROC glad_glUniform4ui;
+#define glUniform4ui glad_glUniform4ui
+GLAD_API_CALL PFNGLUNIFORM4UIVPROC glad_glUniform4uiv;
+#define glUniform4uiv glad_glUniform4uiv
+GLAD_API_CALL PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding;
+#define glUniformBlockBinding glad_glUniformBlockBinding
+GLAD_API_CALL PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv;
+#define glUniformMatrix2fv glad_glUniformMatrix2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv;
+#define glUniformMatrix2x3fv glad_glUniformMatrix2x3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv;
+#define glUniformMatrix2x4fv glad_glUniformMatrix2x4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv;
+#define glUniformMatrix3fv glad_glUniformMatrix3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv;
+#define glUniformMatrix3x2fv glad_glUniformMatrix3x2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv;
+#define glUniformMatrix3x4fv glad_glUniformMatrix3x4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv;
+#define glUniformMatrix4fv glad_glUniformMatrix4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv;
+#define glUniformMatrix4x2fv glad_glUniformMatrix4x2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv;
+#define glUniformMatrix4x3fv glad_glUniformMatrix4x3fv
+GLAD_API_CALL PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer;
+#define glUnmapBuffer glad_glUnmapBuffer
+GLAD_API_CALL PFNGLUSEPROGRAMPROC glad_glUseProgram;
+#define glUseProgram glad_glUseProgram
+GLAD_API_CALL PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram;
+#define glValidateProgram glad_glValidateProgram
+GLAD_API_CALL PFNGLVERTEX2DPROC glad_glVertex2d;
+#define glVertex2d glad_glVertex2d
+GLAD_API_CALL PFNGLVERTEX2DVPROC glad_glVertex2dv;
+#define glVertex2dv glad_glVertex2dv
+GLAD_API_CALL PFNGLVERTEX2FPROC glad_glVertex2f;
+#define glVertex2f glad_glVertex2f
+GLAD_API_CALL PFNGLVERTEX2FVPROC glad_glVertex2fv;
+#define glVertex2fv glad_glVertex2fv
+GLAD_API_CALL PFNGLVERTEX2IPROC glad_glVertex2i;
+#define glVertex2i glad_glVertex2i
+GLAD_API_CALL PFNGLVERTEX2IVPROC glad_glVertex2iv;
+#define glVertex2iv glad_glVertex2iv
+GLAD_API_CALL PFNGLVERTEX2SPROC glad_glVertex2s;
+#define glVertex2s glad_glVertex2s
+GLAD_API_CALL PFNGLVERTEX2SVPROC glad_glVertex2sv;
+#define glVertex2sv glad_glVertex2sv
+GLAD_API_CALL PFNGLVERTEX3DPROC glad_glVertex3d;
+#define glVertex3d glad_glVertex3d
+GLAD_API_CALL PFNGLVERTEX3DVPROC glad_glVertex3dv;
+#define glVertex3dv glad_glVertex3dv
+GLAD_API_CALL PFNGLVERTEX3FPROC glad_glVertex3f;
+#define glVertex3f glad_glVertex3f
+GLAD_API_CALL PFNGLVERTEX3FVPROC glad_glVertex3fv;
+#define glVertex3fv glad_glVertex3fv
+GLAD_API_CALL PFNGLVERTEX3IPROC glad_glVertex3i;
+#define glVertex3i glad_glVertex3i
+GLAD_API_CALL PFNGLVERTEX3IVPROC glad_glVertex3iv;
+#define glVertex3iv glad_glVertex3iv
+GLAD_API_CALL PFNGLVERTEX3SPROC glad_glVertex3s;
+#define glVertex3s glad_glVertex3s
+GLAD_API_CALL PFNGLVERTEX3SVPROC glad_glVertex3sv;
+#define glVertex3sv glad_glVertex3sv
+GLAD_API_CALL PFNGLVERTEX4DPROC glad_glVertex4d;
+#define glVertex4d glad_glVertex4d
+GLAD_API_CALL PFNGLVERTEX4DVPROC glad_glVertex4dv;
+#define glVertex4dv glad_glVertex4dv
+GLAD_API_CALL PFNGLVERTEX4FPROC glad_glVertex4f;
+#define glVertex4f glad_glVertex4f
+GLAD_API_CALL PFNGLVERTEX4FVPROC glad_glVertex4fv;
+#define glVertex4fv glad_glVertex4fv
+GLAD_API_CALL PFNGLVERTEX4IPROC glad_glVertex4i;
+#define glVertex4i glad_glVertex4i
+GLAD_API_CALL PFNGLVERTEX4IVPROC glad_glVertex4iv;
+#define glVertex4iv glad_glVertex4iv
+GLAD_API_CALL PFNGLVERTEX4SPROC glad_glVertex4s;
+#define glVertex4s glad_glVertex4s
+GLAD_API_CALL PFNGLVERTEX4SVPROC glad_glVertex4sv;
+#define glVertex4sv glad_glVertex4sv
+GLAD_API_CALL PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d;
+#define glVertexAttrib1d glad_glVertexAttrib1d
+GLAD_API_CALL PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv;
+#define glVertexAttrib1dv glad_glVertexAttrib1dv
+GLAD_API_CALL PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f;
+#define glVertexAttrib1f glad_glVertexAttrib1f
+GLAD_API_CALL PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv;
+#define glVertexAttrib1fv glad_glVertexAttrib1fv
+GLAD_API_CALL PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s;
+#define glVertexAttrib1s glad_glVertexAttrib1s
+GLAD_API_CALL PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv;
+#define glVertexAttrib1sv glad_glVertexAttrib1sv
+GLAD_API_CALL PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d;
+#define glVertexAttrib2d glad_glVertexAttrib2d
+GLAD_API_CALL PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv;
+#define glVertexAttrib2dv glad_glVertexAttrib2dv
+GLAD_API_CALL PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f;
+#define glVertexAttrib2f glad_glVertexAttrib2f
+GLAD_API_CALL PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv;
+#define glVertexAttrib2fv glad_glVertexAttrib2fv
+GLAD_API_CALL PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s;
+#define glVertexAttrib2s glad_glVertexAttrib2s
+GLAD_API_CALL PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv;
+#define glVertexAttrib2sv glad_glVertexAttrib2sv
+GLAD_API_CALL PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d;
+#define glVertexAttrib3d glad_glVertexAttrib3d
+GLAD_API_CALL PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv;
+#define glVertexAttrib3dv glad_glVertexAttrib3dv
+GLAD_API_CALL PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f;
+#define glVertexAttrib3f glad_glVertexAttrib3f
+GLAD_API_CALL PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv;
+#define glVertexAttrib3fv glad_glVertexAttrib3fv
+GLAD_API_CALL PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s;
+#define glVertexAttrib3s glad_glVertexAttrib3s
+GLAD_API_CALL PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv;
+#define glVertexAttrib3sv glad_glVertexAttrib3sv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv;
+#define glVertexAttrib4Nbv glad_glVertexAttrib4Nbv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv;
+#define glVertexAttrib4Niv glad_glVertexAttrib4Niv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv;
+#define glVertexAttrib4Nsv glad_glVertexAttrib4Nsv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub;
+#define glVertexAttrib4Nub glad_glVertexAttrib4Nub
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv;
+#define glVertexAttrib4Nubv glad_glVertexAttrib4Nubv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv;
+#define glVertexAttrib4Nuiv glad_glVertexAttrib4Nuiv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv;
+#define glVertexAttrib4Nusv glad_glVertexAttrib4Nusv
+GLAD_API_CALL PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv;
+#define glVertexAttrib4bv glad_glVertexAttrib4bv
+GLAD_API_CALL PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d;
+#define glVertexAttrib4d glad_glVertexAttrib4d
+GLAD_API_CALL PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv;
+#define glVertexAttrib4dv glad_glVertexAttrib4dv
+GLAD_API_CALL PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f;
+#define glVertexAttrib4f glad_glVertexAttrib4f
+GLAD_API_CALL PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv;
+#define glVertexAttrib4fv glad_glVertexAttrib4fv
+GLAD_API_CALL PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv;
+#define glVertexAttrib4iv glad_glVertexAttrib4iv
+GLAD_API_CALL PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s;
+#define glVertexAttrib4s glad_glVertexAttrib4s
+GLAD_API_CALL PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv;
+#define glVertexAttrib4sv glad_glVertexAttrib4sv
+GLAD_API_CALL PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv;
+#define glVertexAttrib4ubv glad_glVertexAttrib4ubv
+GLAD_API_CALL PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv;
+#define glVertexAttrib4uiv glad_glVertexAttrib4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv;
+#define glVertexAttrib4usv glad_glVertexAttrib4usv
+GLAD_API_CALL PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor;
+#define glVertexAttribDivisor glad_glVertexAttribDivisor
+GLAD_API_CALL PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i;
+#define glVertexAttribI1i glad_glVertexAttribI1i
+GLAD_API_CALL PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv;
+#define glVertexAttribI1iv glad_glVertexAttribI1iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui;
+#define glVertexAttribI1ui glad_glVertexAttribI1ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv;
+#define glVertexAttribI1uiv glad_glVertexAttribI1uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i;
+#define glVertexAttribI2i glad_glVertexAttribI2i
+GLAD_API_CALL PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv;
+#define glVertexAttribI2iv glad_glVertexAttribI2iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui;
+#define glVertexAttribI2ui glad_glVertexAttribI2ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv;
+#define glVertexAttribI2uiv glad_glVertexAttribI2uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i;
+#define glVertexAttribI3i glad_glVertexAttribI3i
+GLAD_API_CALL PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv;
+#define glVertexAttribI3iv glad_glVertexAttribI3iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui;
+#define glVertexAttribI3ui glad_glVertexAttribI3ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv;
+#define glVertexAttribI3uiv glad_glVertexAttribI3uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv;
+#define glVertexAttribI4bv glad_glVertexAttribI4bv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i;
+#define glVertexAttribI4i glad_glVertexAttribI4i
+GLAD_API_CALL PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv;
+#define glVertexAttribI4iv glad_glVertexAttribI4iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv;
+#define glVertexAttribI4sv glad_glVertexAttribI4sv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv;
+#define glVertexAttribI4ubv glad_glVertexAttribI4ubv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui;
+#define glVertexAttribI4ui glad_glVertexAttribI4ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv;
+#define glVertexAttribI4uiv glad_glVertexAttribI4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv;
+#define glVertexAttribI4usv glad_glVertexAttribI4usv
+GLAD_API_CALL PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer;
+#define glVertexAttribIPointer glad_glVertexAttribIPointer
+GLAD_API_CALL PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui;
+#define glVertexAttribP1ui glad_glVertexAttribP1ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv;
+#define glVertexAttribP1uiv glad_glVertexAttribP1uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui;
+#define glVertexAttribP2ui glad_glVertexAttribP2ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv;
+#define glVertexAttribP2uiv glad_glVertexAttribP2uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui;
+#define glVertexAttribP3ui glad_glVertexAttribP3ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv;
+#define glVertexAttribP3uiv glad_glVertexAttribP3uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui;
+#define glVertexAttribP4ui glad_glVertexAttribP4ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv;
+#define glVertexAttribP4uiv glad_glVertexAttribP4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer;
+#define glVertexAttribPointer glad_glVertexAttribPointer
+GLAD_API_CALL PFNGLVERTEXP2UIPROC glad_glVertexP2ui;
+#define glVertexP2ui glad_glVertexP2ui
+GLAD_API_CALL PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv;
+#define glVertexP2uiv glad_glVertexP2uiv
+GLAD_API_CALL PFNGLVERTEXP3UIPROC glad_glVertexP3ui;
+#define glVertexP3ui glad_glVertexP3ui
+GLAD_API_CALL PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv;
+#define glVertexP3uiv glad_glVertexP3uiv
+GLAD_API_CALL PFNGLVERTEXP4UIPROC glad_glVertexP4ui;
+#define glVertexP4ui glad_glVertexP4ui
+GLAD_API_CALL PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv;
+#define glVertexP4uiv glad_glVertexP4uiv
+GLAD_API_CALL PFNGLVERTEXPOINTERPROC glad_glVertexPointer;
+#define glVertexPointer glad_glVertexPointer
+GLAD_API_CALL PFNGLVIEWPORTPROC glad_glViewport;
+#define glViewport glad_glViewport
+GLAD_API_CALL PFNGLWAITSYNCPROC glad_glWaitSync;
+#define glWaitSync glad_glWaitSync
+GLAD_API_CALL PFNGLWINDOWPOS2DPROC glad_glWindowPos2d;
+#define glWindowPos2d glad_glWindowPos2d
+GLAD_API_CALL PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv;
+#define glWindowPos2dv glad_glWindowPos2dv
+GLAD_API_CALL PFNGLWINDOWPOS2FPROC glad_glWindowPos2f;
+#define glWindowPos2f glad_glWindowPos2f
+GLAD_API_CALL PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv;
+#define glWindowPos2fv glad_glWindowPos2fv
+GLAD_API_CALL PFNGLWINDOWPOS2IPROC glad_glWindowPos2i;
+#define glWindowPos2i glad_glWindowPos2i
+GLAD_API_CALL PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv;
+#define glWindowPos2iv glad_glWindowPos2iv
+GLAD_API_CALL PFNGLWINDOWPOS2SPROC glad_glWindowPos2s;
+#define glWindowPos2s glad_glWindowPos2s
+GLAD_API_CALL PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv;
+#define glWindowPos2sv glad_glWindowPos2sv
+GLAD_API_CALL PFNGLWINDOWPOS3DPROC glad_glWindowPos3d;
+#define glWindowPos3d glad_glWindowPos3d
+GLAD_API_CALL PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv;
+#define glWindowPos3dv glad_glWindowPos3dv
+GLAD_API_CALL PFNGLWINDOWPOS3FPROC glad_glWindowPos3f;
+#define glWindowPos3f glad_glWindowPos3f
+GLAD_API_CALL PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv;
+#define glWindowPos3fv glad_glWindowPos3fv
+GLAD_API_CALL PFNGLWINDOWPOS3IPROC glad_glWindowPos3i;
+#define glWindowPos3i glad_glWindowPos3i
+GLAD_API_CALL PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv;
+#define glWindowPos3iv glad_glWindowPos3iv
+GLAD_API_CALL PFNGLWINDOWPOS3SPROC glad_glWindowPos3s;
+#define glWindowPos3s glad_glWindowPos3s
+GLAD_API_CALL PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv;
+#define glWindowPos3sv glad_glWindowPos3sv
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLUserPtr( GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGL( GLADloadfunc load);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/glx.h glmark2-2021.02/android/jni/src/glad/include/glad/glx.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/glx.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/glad/glx.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,570 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:13 2019
+ *
+ * Generator: C/C++
+ * Specification: glx
+ * Extensions: 2
+ *
+ * APIs:
+ *  - glx=1.4
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='glx=1.4' --extensions='GLX_EXT_swap_control,GLX_MESA_swap_control' c
+ *
+ * Online:
+ *    http://glad.sh/#api=glx%3D1.4&extensions=GLX_EXT_swap_control%2CGLX_MESA_swap_control&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GLX_H_
+#define GLAD_GLX_H_
+
+#ifdef GLX_H
+    #error GLX header already included (API: glx), remove previous include!
+#endif
+#define GLX_H 1
+
+
+#include <X11/X.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+
+#include <glad/gl.h>
+
+#define GLAD_GLX
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GLX_ACCUM_ALPHA_SIZE 17
+#define GLX_ACCUM_BLUE_SIZE 16
+#define GLX_ACCUM_BUFFER_BIT 0x00000080
+#define GLX_ACCUM_GREEN_SIZE 15
+#define GLX_ACCUM_RED_SIZE 14
+#define GLX_ALPHA_SIZE 11
+#define GLX_AUX_BUFFERS 7
+#define GLX_AUX_BUFFERS_BIT 0x00000010
+#define GLX_BACK_LEFT_BUFFER_BIT 0x00000004
+#define GLX_BACK_RIGHT_BUFFER_BIT 0x00000008
+#define GLX_BAD_ATTRIBUTE 2
+#define GLX_BAD_CONTEXT 5
+#define GLX_BAD_ENUM 7
+#define GLX_BAD_SCREEN 1
+#define GLX_BAD_VALUE 6
+#define GLX_BAD_VISUAL 4
+#define GLX_BLUE_SIZE 10
+#define GLX_BUFFER_SIZE 2
+#define GLX_BufferSwapComplete 1
+#define GLX_COLOR_INDEX_BIT 0x00000002
+#define GLX_COLOR_INDEX_TYPE 0x8015
+#define GLX_CONFIG_CAVEAT 0x20
+#define GLX_DAMAGED 0x8020
+#define GLX_DEPTH_BUFFER_BIT 0x00000020
+#define GLX_DEPTH_SIZE 12
+#define GLX_DIRECT_COLOR 0x8003
+#define GLX_DONT_CARE 0xFFFFFFFF
+#define GLX_DOUBLEBUFFER 5
+#define GLX_DRAWABLE_TYPE 0x8010
+#define GLX_EVENT_MASK 0x801F
+#define GLX_EXTENSIONS 0x3
+#define GLX_EXTENSION_NAME "GLX"
+#define GLX_FBCONFIG_ID 0x8013
+#define GLX_FRONT_LEFT_BUFFER_BIT 0x00000001
+#define GLX_FRONT_RIGHT_BUFFER_BIT 0x00000002
+#define GLX_GRAY_SCALE 0x8006
+#define GLX_GREEN_SIZE 9
+#define GLX_HEIGHT 0x801E
+#define GLX_LARGEST_PBUFFER 0x801C
+#define GLX_LEVEL 3
+#define GLX_MAX_PBUFFER_HEIGHT 0x8017
+#define GLX_MAX_PBUFFER_PIXELS 0x8018
+#define GLX_MAX_PBUFFER_WIDTH 0x8016
+#define GLX_MAX_SWAP_INTERVAL_EXT 0x20F2
+#define GLX_NONE 0x8000
+#define GLX_NON_CONFORMANT_CONFIG 0x800D
+#define GLX_NO_EXTENSION 3
+#define GLX_PBUFFER 0x8023
+#define GLX_PBUFFER_BIT 0x00000004
+#define GLX_PBUFFER_CLOBBER_MASK 0x08000000
+#define GLX_PBUFFER_HEIGHT 0x8040
+#define GLX_PBUFFER_WIDTH 0x8041
+#define GLX_PIXMAP_BIT 0x00000002
+#define GLX_PRESERVED_CONTENTS 0x801B
+#define GLX_PSEUDO_COLOR 0x8004
+#define GLX_PbufferClobber 0
+#define GLX_RED_SIZE 8
+#define GLX_RENDER_TYPE 0x8011
+#define GLX_RGBA 4
+#define GLX_RGBA_BIT 0x00000001
+#define GLX_RGBA_TYPE 0x8014
+#define GLX_SAMPLES 100001
+#define GLX_SAMPLE_BUFFERS 100000
+#define GLX_SAVED 0x8021
+#define GLX_SCREEN 0x800C
+#define GLX_SLOW_CONFIG 0x8001
+#define GLX_STATIC_COLOR 0x8005
+#define GLX_STATIC_GRAY 0x8007
+#define GLX_STENCIL_BUFFER_BIT 0x00000040
+#define GLX_STENCIL_SIZE 13
+#define GLX_STEREO 6
+#define GLX_SWAP_INTERVAL_EXT 0x20F1
+#define GLX_TRANSPARENT_ALPHA_VALUE 0x28
+#define GLX_TRANSPARENT_BLUE_VALUE 0x27
+#define GLX_TRANSPARENT_GREEN_VALUE 0x26
+#define GLX_TRANSPARENT_INDEX 0x8009
+#define GLX_TRANSPARENT_INDEX_VALUE 0x24
+#define GLX_TRANSPARENT_RED_VALUE 0x25
+#define GLX_TRANSPARENT_RGB 0x8008
+#define GLX_TRANSPARENT_TYPE 0x23
+#define GLX_TRUE_COLOR 0x8002
+#define GLX_USE_GL 1
+#define GLX_VENDOR 0x1
+#define GLX_VERSION 0x2
+#define GLX_VISUAL_ID 0x800B
+#define GLX_WIDTH 0x801D
+#define GLX_WINDOW 0x8022
+#define GLX_WINDOW_BIT 0x00000001
+#define GLX_X_RENDERABLE 0x8012
+#define GLX_X_VISUAL_TYPE 0x22
+#define __GLX_NUMBER_EVENTS 17
+
+
+#ifndef GLEXT_64_TYPES_DEFINED
+/* This code block is duplicated in glext.h, so must be protected */
+#define GLEXT_64_TYPES_DEFINED
+/* Define int32_t, int64_t, and uint64_t types for UST/MSC */
+/* (as used in the GLX_OML_sync_control extension). */
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
+#include <inttypes.h>
+#elif defined(__sun__) || defined(__digital__)
+#include <inttypes.h>
+#if defined(__STDC__)
+#if defined(__arch64__) || defined(_LP64)
+typedef long int int64_t;
+typedef unsigned long int uint64_t;
+#else
+typedef long long int int64_t;
+typedef unsigned long long int uint64_t;
+#endif /* __arch64__ */
+#endif /* __STDC__ */
+#elif defined( __VMS ) || defined(__sgi)
+#include <inttypes.h>
+#elif defined(__SCO__) || defined(__USLC__)
+#include <stdint.h>
+#elif defined(__UNIXOS2__) || defined(__SOL64__)
+typedef long int int32_t;
+typedef long long int int64_t;
+typedef unsigned long long int uint64_t;
+#elif defined(_WIN32) && defined(__GNUC__)
+#include <stdint.h>
+#elif defined(_WIN32)
+typedef __int32 int32_t;
+typedef __int64 int64_t;
+typedef unsigned __int64 uint64_t;
+#else
+/* Fallback if nothing above works */
+#include <inttypes.h>
+#endif
+#endif
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+
+#else
+
+#endif
+
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+
+#else
+
+#endif
+
+
+
+
+
+
+
+typedef XID GLXFBConfigID;
+typedef struct __GLXFBConfigRec *GLXFBConfig;
+typedef XID GLXContextID;
+typedef struct __GLXcontextRec *GLXContext;
+typedef XID GLXPixmap;
+typedef XID GLXDrawable;
+typedef XID GLXWindow;
+typedef XID GLXPbuffer;
+typedef void (GLAD_API_PTR *__GLXextFuncPtr)(void);
+typedef XID GLXVideoCaptureDeviceNV;
+typedef unsigned int GLXVideoDeviceNV;
+typedef XID GLXVideoSourceSGIX;
+typedef XID GLXFBConfigIDSGIX;
+typedef struct __GLXFBConfigRec *GLXFBConfigSGIX;
+typedef XID GLXPbufferSGIX;
+typedef struct {
+    int event_type;             /* GLX_DAMAGED or GLX_SAVED */
+    int draw_type;              /* GLX_WINDOW or GLX_PBUFFER */
+    unsigned long serial;       /* # of last request processed by server */
+    Bool send_event;            /* true if this came for SendEvent request */
+    Display *display;           /* display the event was read from */
+    GLXDrawable drawable;       /* XID of Drawable */
+    unsigned int buffer_mask;   /* mask indicating which buffers are affected */
+    unsigned int aux_buffer;    /* which aux buffer was affected */
+    int x, y;
+    int width, height;
+    int count;                  /* if nonzero, at least this many more */
+} GLXPbufferClobberEvent;
+typedef struct {
+    int type;
+    unsigned long serial;       /* # of last request processed by server */
+    Bool send_event;            /* true if this came from a SendEvent request */
+    Display *display;           /* Display the event was read from */
+    GLXDrawable drawable;       /* drawable on which event was requested in event mask */
+    int event_type;
+    int64_t ust;
+    int64_t msc;
+    int64_t sbc;
+} GLXBufferSwapComplete;
+typedef union __GLXEvent {
+    GLXPbufferClobberEvent glxpbufferclobber;
+    GLXBufferSwapComplete glxbufferswapcomplete;
+    long pad[24];
+} GLXEvent;
+typedef struct {
+    int type;
+    unsigned long serial;
+    Bool send_event;
+    Display *display;
+    int extension;
+    int evtype;
+    GLXDrawable window;
+    Bool stereo_tree;
+} GLXStereoNotifyEventEXT;
+typedef struct {
+    int type;
+    unsigned long serial;   /* # of last request processed by server */
+    Bool send_event;        /* true if this came for SendEvent request */
+    Display *display;       /* display the event was read from */
+    GLXDrawable drawable;   /* i.d. of Drawable */
+    int event_type;         /* GLX_DAMAGED_SGIX or GLX_SAVED_SGIX */
+    int draw_type;          /* GLX_WINDOW_SGIX or GLX_PBUFFER_SGIX */
+    unsigned int mask;      /* mask indicating which buffers are affected*/
+    int x, y;
+    int width, height;
+    int count;              /* if nonzero, at least this many more */
+} GLXBufferClobberEventSGIX;
+typedef struct {
+    char    pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int     networkId;
+} GLXHyperpipeNetworkSGIX;
+typedef struct {
+    char    pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int     channel;
+    unsigned int participationType;
+    int     timeSlice;
+} GLXHyperpipeConfigSGIX;
+typedef struct {
+    char pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int srcXOrigin, srcYOrigin, srcWidth, srcHeight;
+    int destXOrigin, destYOrigin, destWidth, destHeight;
+} GLXPipeRect;
+typedef struct {
+    char pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int XOrigin, YOrigin, maxHeight, maxWidth;
+} GLXPipeRectLimits;
+
+
+#define GLX_VERSION_1_0 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_0;
+#define GLX_VERSION_1_1 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_1;
+#define GLX_VERSION_1_2 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_2;
+#define GLX_VERSION_1_3 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_3;
+#define GLX_VERSION_1_4 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_4;
+#define GLX_EXT_swap_control 1
+GLAD_API_CALL int GLAD_GLX_EXT_swap_control;
+#define GLX_MESA_swap_control 1
+GLAD_API_CALL int GLAD_GLX_MESA_swap_control;
+
+
+typedef GLXFBConfig * (GLAD_API_PTR *PFNGLXCHOOSEFBCONFIGPROC)(Display * dpy, int screen, const int * attrib_list, int * nelements);
+typedef XVisualInfo * (GLAD_API_PTR *PFNGLXCHOOSEVISUALPROC)(Display * dpy, int screen, int * attribList);
+typedef void (GLAD_API_PTR *PFNGLXCOPYCONTEXTPROC)(Display * dpy, GLXContext src, GLXContext dst, unsigned long mask);
+typedef GLXContext (GLAD_API_PTR *PFNGLXCREATECONTEXTPROC)(Display * dpy, XVisualInfo * vis, GLXContext shareList, Bool direct);
+typedef GLXPixmap (GLAD_API_PTR *PFNGLXCREATEGLXPIXMAPPROC)(Display * dpy, XVisualInfo * visual, Pixmap pixmap);
+typedef GLXContext (GLAD_API_PTR *PFNGLXCREATENEWCONTEXTPROC)(Display * dpy, GLXFBConfig config, int render_type, GLXContext share_list, Bool direct);
+typedef GLXPbuffer (GLAD_API_PTR *PFNGLXCREATEPBUFFERPROC)(Display * dpy, GLXFBConfig config, const int * attrib_list);
+typedef GLXPixmap (GLAD_API_PTR *PFNGLXCREATEPIXMAPPROC)(Display * dpy, GLXFBConfig config, Pixmap pixmap, const int * attrib_list);
+typedef GLXWindow (GLAD_API_PTR *PFNGLXCREATEWINDOWPROC)(Display * dpy, GLXFBConfig config, Window win, const int * attrib_list);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYCONTEXTPROC)(Display * dpy, GLXContext ctx);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYGLXPIXMAPPROC)(Display * dpy, GLXPixmap pixmap);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYPBUFFERPROC)(Display * dpy, GLXPbuffer pbuf);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYPIXMAPPROC)(Display * dpy, GLXPixmap pixmap);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYWINDOWPROC)(Display * dpy, GLXWindow win);
+typedef const char * (GLAD_API_PTR *PFNGLXGETCLIENTSTRINGPROC)(Display * dpy, int name);
+typedef int (GLAD_API_PTR *PFNGLXGETCONFIGPROC)(Display * dpy, XVisualInfo * visual, int attrib, int * value);
+typedef GLXContext (GLAD_API_PTR *PFNGLXGETCURRENTCONTEXTPROC)(void);
+typedef Display * (GLAD_API_PTR *PFNGLXGETCURRENTDISPLAYPROC)(void);
+typedef GLXDrawable (GLAD_API_PTR *PFNGLXGETCURRENTDRAWABLEPROC)(void);
+typedef GLXDrawable (GLAD_API_PTR *PFNGLXGETCURRENTREADDRAWABLEPROC)(void);
+typedef int (GLAD_API_PTR *PFNGLXGETFBCONFIGATTRIBPROC)(Display * dpy, GLXFBConfig config, int attribute, int * value);
+typedef GLXFBConfig * (GLAD_API_PTR *PFNGLXGETFBCONFIGSPROC)(Display * dpy, int screen, int * nelements);
+typedef __GLXextFuncPtr (GLAD_API_PTR *PFNGLXGETPROCADDRESSPROC)(const GLubyte * procName);
+typedef void (GLAD_API_PTR *PFNGLXGETSELECTEDEVENTPROC)(Display * dpy, GLXDrawable draw, unsigned long * event_mask);
+typedef int (GLAD_API_PTR *PFNGLXGETSWAPINTERVALMESAPROC)(void);
+typedef XVisualInfo * (GLAD_API_PTR *PFNGLXGETVISUALFROMFBCONFIGPROC)(Display * dpy, GLXFBConfig config);
+typedef Bool (GLAD_API_PTR *PFNGLXISDIRECTPROC)(Display * dpy, GLXContext ctx);
+typedef Bool (GLAD_API_PTR *PFNGLXMAKECONTEXTCURRENTPROC)(Display * dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx);
+typedef Bool (GLAD_API_PTR *PFNGLXMAKECURRENTPROC)(Display * dpy, GLXDrawable drawable, GLXContext ctx);
+typedef int (GLAD_API_PTR *PFNGLXQUERYCONTEXTPROC)(Display * dpy, GLXContext ctx, int attribute, int * value);
+typedef void (GLAD_API_PTR *PFNGLXQUERYDRAWABLEPROC)(Display * dpy, GLXDrawable draw, int attribute, unsigned int * value);
+typedef Bool (GLAD_API_PTR *PFNGLXQUERYEXTENSIONPROC)(Display * dpy, int * errorb, int * event);
+typedef const char * (GLAD_API_PTR *PFNGLXQUERYEXTENSIONSSTRINGPROC)(Display * dpy, int screen);
+typedef const char * (GLAD_API_PTR *PFNGLXQUERYSERVERSTRINGPROC)(Display * dpy, int screen, int name);
+typedef Bool (GLAD_API_PTR *PFNGLXQUERYVERSIONPROC)(Display * dpy, int * maj, int * min);
+typedef void (GLAD_API_PTR *PFNGLXSELECTEVENTPROC)(Display * dpy, GLXDrawable draw, unsigned long event_mask);
+typedef void (GLAD_API_PTR *PFNGLXSWAPBUFFERSPROC)(Display * dpy, GLXDrawable drawable);
+typedef void (GLAD_API_PTR *PFNGLXSWAPINTERVALEXTPROC)(Display * dpy, GLXDrawable drawable, int interval);
+typedef int (GLAD_API_PTR *PFNGLXSWAPINTERVALMESAPROC)(unsigned int interval);
+typedef void (GLAD_API_PTR *PFNGLXUSEXFONTPROC)(Font font, int first, int count, int list);
+typedef void (GLAD_API_PTR *PFNGLXWAITGLPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLXWAITXPROC)(void);
+
+GLAD_API_CALL PFNGLXCHOOSEFBCONFIGPROC glad_glXChooseFBConfig;
+#define glXChooseFBConfig glad_glXChooseFBConfig
+GLAD_API_CALL PFNGLXCHOOSEVISUALPROC glad_glXChooseVisual;
+#define glXChooseVisual glad_glXChooseVisual
+GLAD_API_CALL PFNGLXCOPYCONTEXTPROC glad_glXCopyContext;
+#define glXCopyContext glad_glXCopyContext
+GLAD_API_CALL PFNGLXCREATECONTEXTPROC glad_glXCreateContext;
+#define glXCreateContext glad_glXCreateContext
+GLAD_API_CALL PFNGLXCREATEGLXPIXMAPPROC glad_glXCreateGLXPixmap;
+#define glXCreateGLXPixmap glad_glXCreateGLXPixmap
+GLAD_API_CALL PFNGLXCREATENEWCONTEXTPROC glad_glXCreateNewContext;
+#define glXCreateNewContext glad_glXCreateNewContext
+GLAD_API_CALL PFNGLXCREATEPBUFFERPROC glad_glXCreatePbuffer;
+#define glXCreatePbuffer glad_glXCreatePbuffer
+GLAD_API_CALL PFNGLXCREATEPIXMAPPROC glad_glXCreatePixmap;
+#define glXCreatePixmap glad_glXCreatePixmap
+GLAD_API_CALL PFNGLXCREATEWINDOWPROC glad_glXCreateWindow;
+#define glXCreateWindow glad_glXCreateWindow
+GLAD_API_CALL PFNGLXDESTROYCONTEXTPROC glad_glXDestroyContext;
+#define glXDestroyContext glad_glXDestroyContext
+GLAD_API_CALL PFNGLXDESTROYGLXPIXMAPPROC glad_glXDestroyGLXPixmap;
+#define glXDestroyGLXPixmap glad_glXDestroyGLXPixmap
+GLAD_API_CALL PFNGLXDESTROYPBUFFERPROC glad_glXDestroyPbuffer;
+#define glXDestroyPbuffer glad_glXDestroyPbuffer
+GLAD_API_CALL PFNGLXDESTROYPIXMAPPROC glad_glXDestroyPixmap;
+#define glXDestroyPixmap glad_glXDestroyPixmap
+GLAD_API_CALL PFNGLXDESTROYWINDOWPROC glad_glXDestroyWindow;
+#define glXDestroyWindow glad_glXDestroyWindow
+GLAD_API_CALL PFNGLXGETCLIENTSTRINGPROC glad_glXGetClientString;
+#define glXGetClientString glad_glXGetClientString
+GLAD_API_CALL PFNGLXGETCONFIGPROC glad_glXGetConfig;
+#define glXGetConfig glad_glXGetConfig
+GLAD_API_CALL PFNGLXGETCURRENTCONTEXTPROC glad_glXGetCurrentContext;
+#define glXGetCurrentContext glad_glXGetCurrentContext
+GLAD_API_CALL PFNGLXGETCURRENTDISPLAYPROC glad_glXGetCurrentDisplay;
+#define glXGetCurrentDisplay glad_glXGetCurrentDisplay
+GLAD_API_CALL PFNGLXGETCURRENTDRAWABLEPROC glad_glXGetCurrentDrawable;
+#define glXGetCurrentDrawable glad_glXGetCurrentDrawable
+GLAD_API_CALL PFNGLXGETCURRENTREADDRAWABLEPROC glad_glXGetCurrentReadDrawable;
+#define glXGetCurrentReadDrawable glad_glXGetCurrentReadDrawable
+GLAD_API_CALL PFNGLXGETFBCONFIGATTRIBPROC glad_glXGetFBConfigAttrib;
+#define glXGetFBConfigAttrib glad_glXGetFBConfigAttrib
+GLAD_API_CALL PFNGLXGETFBCONFIGSPROC glad_glXGetFBConfigs;
+#define glXGetFBConfigs glad_glXGetFBConfigs
+GLAD_API_CALL PFNGLXGETPROCADDRESSPROC glad_glXGetProcAddress;
+#define glXGetProcAddress glad_glXGetProcAddress
+GLAD_API_CALL PFNGLXGETSELECTEDEVENTPROC glad_glXGetSelectedEvent;
+#define glXGetSelectedEvent glad_glXGetSelectedEvent
+GLAD_API_CALL PFNGLXGETSWAPINTERVALMESAPROC glad_glXGetSwapIntervalMESA;
+#define glXGetSwapIntervalMESA glad_glXGetSwapIntervalMESA
+GLAD_API_CALL PFNGLXGETVISUALFROMFBCONFIGPROC glad_glXGetVisualFromFBConfig;
+#define glXGetVisualFromFBConfig glad_glXGetVisualFromFBConfig
+GLAD_API_CALL PFNGLXISDIRECTPROC glad_glXIsDirect;
+#define glXIsDirect glad_glXIsDirect
+GLAD_API_CALL PFNGLXMAKECONTEXTCURRENTPROC glad_glXMakeContextCurrent;
+#define glXMakeContextCurrent glad_glXMakeContextCurrent
+GLAD_API_CALL PFNGLXMAKECURRENTPROC glad_glXMakeCurrent;
+#define glXMakeCurrent glad_glXMakeCurrent
+GLAD_API_CALL PFNGLXQUERYCONTEXTPROC glad_glXQueryContext;
+#define glXQueryContext glad_glXQueryContext
+GLAD_API_CALL PFNGLXQUERYDRAWABLEPROC glad_glXQueryDrawable;
+#define glXQueryDrawable glad_glXQueryDrawable
+GLAD_API_CALL PFNGLXQUERYEXTENSIONPROC glad_glXQueryExtension;
+#define glXQueryExtension glad_glXQueryExtension
+GLAD_API_CALL PFNGLXQUERYEXTENSIONSSTRINGPROC glad_glXQueryExtensionsString;
+#define glXQueryExtensionsString glad_glXQueryExtensionsString
+GLAD_API_CALL PFNGLXQUERYSERVERSTRINGPROC glad_glXQueryServerString;
+#define glXQueryServerString glad_glXQueryServerString
+GLAD_API_CALL PFNGLXQUERYVERSIONPROC glad_glXQueryVersion;
+#define glXQueryVersion glad_glXQueryVersion
+GLAD_API_CALL PFNGLXSELECTEVENTPROC glad_glXSelectEvent;
+#define glXSelectEvent glad_glXSelectEvent
+GLAD_API_CALL PFNGLXSWAPBUFFERSPROC glad_glXSwapBuffers;
+#define glXSwapBuffers glad_glXSwapBuffers
+GLAD_API_CALL PFNGLXSWAPINTERVALEXTPROC glad_glXSwapIntervalEXT;
+#define glXSwapIntervalEXT glad_glXSwapIntervalEXT
+GLAD_API_CALL PFNGLXSWAPINTERVALMESAPROC glad_glXSwapIntervalMESA;
+#define glXSwapIntervalMESA glad_glXSwapIntervalMESA
+GLAD_API_CALL PFNGLXUSEXFONTPROC glad_glXUseXFont;
+#define glXUseXFont glad_glXUseXFont
+GLAD_API_CALL PFNGLXWAITGLPROC glad_glXWaitGL;
+#define glXWaitGL glad_glXWaitGL
+GLAD_API_CALL PFNGLXWAITXPROC glad_glXWaitX;
+#define glXWaitX glad_glXWaitX
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLXUserPtr(Display *display, int screen, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGLX(Display *display, int screen, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/wgl.h glmark2-2021.02/android/jni/src/glad/include/glad/wgl.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/glad/wgl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/glad/wgl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,298 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:13 2019
+ *
+ * Generator: C/C++
+ * Specification: wgl
+ * Extensions: 3
+ *
+ * APIs:
+ *  - wgl=1.0
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='wgl=1.0' --extensions='WGL_ARB_extensions_string,WGL_EXT_extensions_string,WGL_EXT_swap_control' c
+ *
+ * Online:
+ *    http://glad.sh/#api=wgl%3D1.0&extensions=WGL_ARB_extensions_string%2CWGL_EXT_extensions_string%2CWGL_EXT_swap_control&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_WGL_H_
+#define GLAD_WGL_H_
+
+#include <windows.h>
+#include <glad/gl.h>
+
+#define GLAD_WGL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define WGL_FONT_LINES 0
+#define WGL_FONT_POLYGONS 1
+#define WGL_SWAP_MAIN_PLANE 0x00000001
+#define WGL_SWAP_OVERLAY1 0x00000002
+#define WGL_SWAP_OVERLAY10 0x00000400
+#define WGL_SWAP_OVERLAY11 0x00000800
+#define WGL_SWAP_OVERLAY12 0x00001000
+#define WGL_SWAP_OVERLAY13 0x00002000
+#define WGL_SWAP_OVERLAY14 0x00004000
+#define WGL_SWAP_OVERLAY15 0x00008000
+#define WGL_SWAP_OVERLAY2 0x00000004
+#define WGL_SWAP_OVERLAY3 0x00000008
+#define WGL_SWAP_OVERLAY4 0x00000010
+#define WGL_SWAP_OVERLAY5 0x00000020
+#define WGL_SWAP_OVERLAY6 0x00000040
+#define WGL_SWAP_OVERLAY7 0x00000080
+#define WGL_SWAP_OVERLAY8 0x00000100
+#define WGL_SWAP_OVERLAY9 0x00000200
+#define WGL_SWAP_UNDERLAY1 0x00010000
+#define WGL_SWAP_UNDERLAY10 0x02000000
+#define WGL_SWAP_UNDERLAY11 0x04000000
+#define WGL_SWAP_UNDERLAY12 0x08000000
+#define WGL_SWAP_UNDERLAY13 0x10000000
+#define WGL_SWAP_UNDERLAY14 0x20000000
+#define WGL_SWAP_UNDERLAY15 0x40000000
+#define WGL_SWAP_UNDERLAY2 0x00020000
+#define WGL_SWAP_UNDERLAY3 0x00040000
+#define WGL_SWAP_UNDERLAY4 0x00080000
+#define WGL_SWAP_UNDERLAY5 0x00100000
+#define WGL_SWAP_UNDERLAY6 0x00200000
+#define WGL_SWAP_UNDERLAY7 0x00400000
+#define WGL_SWAP_UNDERLAY8 0x00800000
+#define WGL_SWAP_UNDERLAY9 0x01000000
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+struct _GPU_DEVICE {
+    DWORD  cb;
+    CHAR   DeviceName[32];
+    CHAR   DeviceString[128];
+    DWORD  Flags;
+    RECT   rcVirtualScreen;
+};
+DECLARE_HANDLE(HPBUFFERARB);
+DECLARE_HANDLE(HPBUFFEREXT);
+DECLARE_HANDLE(HVIDEOOUTPUTDEVICENV);
+DECLARE_HANDLE(HPVIDEODEV);
+DECLARE_HANDLE(HPGPUNV);
+DECLARE_HANDLE(HGPUNV);
+DECLARE_HANDLE(HVIDEOINPUTDEVICENV);
+typedef struct _GPU_DEVICE GPU_DEVICE;
+typedef struct _GPU_DEVICE *PGPU_DEVICE;
+
+
+#define WGL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_WGL_VERSION_1_0;
+#define WGL_ARB_extensions_string 1
+GLAD_API_CALL int GLAD_WGL_ARB_extensions_string;
+#define WGL_EXT_extensions_string 1
+GLAD_API_CALL int GLAD_WGL_EXT_extensions_string;
+#define WGL_EXT_swap_control 1
+GLAD_API_CALL int GLAD_WGL_EXT_swap_control;
+
+
+typedef int (GLAD_API_PTR *PFNCHOOSEPIXELFORMATPROC)(HDC hDc, const PIXELFORMATDESCRIPTOR * pPfd);
+typedef int (GLAD_API_PTR *PFNDESCRIBEPIXELFORMATPROC)(HDC hdc, int ipfd, UINT cjpfd, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef UINT (GLAD_API_PTR *PFNGETENHMETAFILEPIXELFORMATPROC)(HENHMETAFILE hemf, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef int (GLAD_API_PTR *PFNGETPIXELFORMATPROC)(HDC hdc);
+typedef BOOL (GLAD_API_PTR *PFNSETPIXELFORMATPROC)(HDC hdc, int ipfd, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef BOOL (GLAD_API_PTR *PFNSWAPBUFFERSPROC)(HDC hdc);
+typedef BOOL (GLAD_API_PTR *PFNWGLCOPYCONTEXTPROC)(HGLRC hglrcSrc, HGLRC hglrcDst, UINT mask);
+typedef HGLRC (GLAD_API_PTR *PFNWGLCREATECONTEXTPROC)(HDC hDc);
+typedef HGLRC (GLAD_API_PTR *PFNWGLCREATELAYERCONTEXTPROC)(HDC hDc, int level);
+typedef BOOL (GLAD_API_PTR *PFNWGLDELETECONTEXTPROC)(HGLRC oldContext);
+typedef BOOL (GLAD_API_PTR *PFNWGLDESCRIBELAYERPLANEPROC)(HDC hDc, int pixelFormat, int layerPlane, UINT nBytes, const LAYERPLANEDESCRIPTOR * plpd);
+typedef HGLRC (GLAD_API_PTR *PFNWGLGETCURRENTCONTEXTPROC)(void);
+typedef HDC (GLAD_API_PTR *PFNWGLGETCURRENTDCPROC)(void);
+typedef const char * (GLAD_API_PTR *PFNWGLGETEXTENSIONSSTRINGARBPROC)(HDC hdc);
+typedef const char * (GLAD_API_PTR *PFNWGLGETEXTENSIONSSTRINGEXTPROC)(void);
+typedef int (GLAD_API_PTR *PFNWGLGETLAYERPALETTEENTRIESPROC)(HDC hdc, int iLayerPlane, int iStart, int cEntries, const COLORREF * pcr);
+typedef PROC (GLAD_API_PTR *PFNWGLGETPROCADDRESSPROC)(LPCSTR lpszProc);
+typedef int (GLAD_API_PTR *PFNWGLGETSWAPINTERVALEXTPROC)(void);
+typedef BOOL (GLAD_API_PTR *PFNWGLMAKECURRENTPROC)(HDC hDc, HGLRC newContext);
+typedef BOOL (GLAD_API_PTR *PFNWGLREALIZELAYERPALETTEPROC)(HDC hdc, int iLayerPlane, BOOL bRealize);
+typedef int (GLAD_API_PTR *PFNWGLSETLAYERPALETTEENTRIESPROC)(HDC hdc, int iLayerPlane, int iStart, int cEntries, const COLORREF * pcr);
+typedef BOOL (GLAD_API_PTR *PFNWGLSHARELISTSPROC)(HGLRC hrcSrvShare, HGLRC hrcSrvSource);
+typedef BOOL (GLAD_API_PTR *PFNWGLSWAPINTERVALEXTPROC)(int interval);
+typedef BOOL (GLAD_API_PTR *PFNWGLSWAPLAYERBUFFERSPROC)(HDC hdc, UINT fuFlags);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSAPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSWPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESAPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESWPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+
+GLAD_API_CALL PFNWGLGETEXTENSIONSSTRINGARBPROC glad_wglGetExtensionsStringARB;
+#define wglGetExtensionsStringARB glad_wglGetExtensionsStringARB
+GLAD_API_CALL PFNWGLGETEXTENSIONSSTRINGEXTPROC glad_wglGetExtensionsStringEXT;
+#define wglGetExtensionsStringEXT glad_wglGetExtensionsStringEXT
+GLAD_API_CALL PFNWGLGETSWAPINTERVALEXTPROC glad_wglGetSwapIntervalEXT;
+#define wglGetSwapIntervalEXT glad_wglGetSwapIntervalEXT
+GLAD_API_CALL PFNWGLSWAPINTERVALEXTPROC glad_wglSwapIntervalEXT;
+#define wglSwapIntervalEXT glad_wglSwapIntervalEXT
+
+
+
+
+
+GLAD_API_CALL int gladLoadWGLUserPtr(HDC hdc, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadWGL(HDC hdc, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/KHR/khrplatform.h glmark2-2021.02/android/jni/src/glad/include/KHR/khrplatform.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/include/KHR/khrplatform.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/include/KHR/khrplatform.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,290 @@
+#ifndef __khrplatform_h_
+#define __khrplatform_h_
+
+/*
+** Copyright (c) 2008-2018 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+/* Khronos platform-specific types and definitions.
+ *
+ * The master copy of khrplatform.h is maintained in the Khronos EGL
+ * Registry repository at https://github.com/KhronosGroup/EGL-Registry
+ * The last semantic modification to khrplatform.h was at commit ID:
+ *      67a3e0864c2d75ea5287b9f3d2eb74a745936692
+ *
+ * Adopters may modify this file to suit their platform. Adopters are
+ * encouraged to submit platform specific modifications to the Khronos
+ * group so that they can be included in future versions of this file.
+ * Please submit changes by filing pull requests or issues on
+ * the EGL Registry repository linked above.
+ *
+ *
+ * See the Implementer's Guidelines for information about where this file
+ * should be located on your system and for more details of its use:
+ *    http://www.khronos.org/registry/implementers_guide.pdf
+ *
+ * This file should be included as
+ *        #include <KHR/khrplatform.h>
+ * by Khronos client API header files that use its types and defines.
+ *
+ * The types in khrplatform.h should only be used to define API-specific types.
+ *
+ * Types defined in khrplatform.h:
+ *    khronos_int8_t              signed   8  bit
+ *    khronos_uint8_t             unsigned 8  bit
+ *    khronos_int16_t             signed   16 bit
+ *    khronos_uint16_t            unsigned 16 bit
+ *    khronos_int32_t             signed   32 bit
+ *    khronos_uint32_t            unsigned 32 bit
+ *    khronos_int64_t             signed   64 bit
+ *    khronos_uint64_t            unsigned 64 bit
+ *    khronos_intptr_t            signed   same number of bits as a pointer
+ *    khronos_uintptr_t           unsigned same number of bits as a pointer
+ *    khronos_ssize_t             signed   size
+ *    khronos_usize_t             unsigned size
+ *    khronos_float_t             signed   32 bit floating point
+ *    khronos_time_ns_t           unsigned 64 bit time in nanoseconds
+ *    khronos_utime_nanoseconds_t unsigned time interval or absolute time in
+ *                                         nanoseconds
+ *    khronos_stime_nanoseconds_t signed time interval in nanoseconds
+ *    khronos_boolean_enum_t      enumerated boolean type. This should
+ *      only be used as a base type when a client API's boolean type is
+ *      an enum. Client APIs which use an integer or other type for
+ *      booleans cannot use this as the base type for their boolean.
+ *
+ * Tokens defined in khrplatform.h:
+ *
+ *    KHRONOS_FALSE, KHRONOS_TRUE Enumerated boolean false/true values.
+ *
+ *    KHRONOS_SUPPORT_INT64 is 1 if 64 bit integers are supported; otherwise 0.
+ *    KHRONOS_SUPPORT_FLOAT is 1 if floats are supported; otherwise 0.
+ *
+ * Calling convention macros defined in this file:
+ *    KHRONOS_APICALL
+ *    KHRONOS_APIENTRY
+ *    KHRONOS_APIATTRIBUTES
+ *
+ * These may be used in function prototypes as:
+ *
+ *      KHRONOS_APICALL void KHRONOS_APIENTRY funcname(
+ *                                  int arg1,
+ *                                  int arg2) KHRONOS_APIATTRIBUTES;
+ */
+
+#if defined(__SCITECH_SNAP__) && !defined(KHRONOS_STATIC)
+#   define KHRONOS_STATIC 1
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APICALL
+ *-------------------------------------------------------------------------
+ * This precedes the return type of the function in the function prototype.
+ */
+#if defined(KHRONOS_STATIC)
+    /* If the preprocessor constant KHRONOS_STATIC is defined, make the
+     * header compatible with static linking. */
+#   define KHRONOS_APICALL
+#elif defined(_WIN32)
+#   define KHRONOS_APICALL __declspec(dllimport)
+#elif defined (__SYMBIAN32__)
+#   define KHRONOS_APICALL IMPORT_C
+#elif defined(__ANDROID__)
+#   define KHRONOS_APICALL __attribute__((visibility("default")))
+#else
+#   define KHRONOS_APICALL
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APIENTRY
+ *-------------------------------------------------------------------------
+ * This follows the return type of the function  and precedes the function
+ * name in the function prototype.
+ */
+#if defined(_WIN32) && !defined(_WIN32_WCE) && !defined(KHRONOS_STATIC)
+    /* Win32 but not WinCE */
+#   define KHRONOS_APIENTRY __stdcall
+#else
+#   define KHRONOS_APIENTRY
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APIATTRIBUTES
+ *-------------------------------------------------------------------------
+ * This follows the closing parenthesis of the function prototype arguments.
+ */
+#if defined (__ARMCC_2__)
+#define KHRONOS_APIATTRIBUTES __softfp
+#else
+#define KHRONOS_APIATTRIBUTES
+#endif
+
+/*-------------------------------------------------------------------------
+ * basic type definitions
+ *-----------------------------------------------------------------------*/
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(__GNUC__) || defined(__SCO__) || defined(__USLC__)
+
+
+/*
+ * Using <stdint.h>
+ */
+#include <stdint.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(__VMS ) || defined(__sgi)
+
+/*
+ * Using <inttypes.h>
+ */
+#include <inttypes.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(_WIN32) && !defined(__SCITECH_SNAP__)
+
+/*
+ * Win32
+ */
+typedef __int32                 khronos_int32_t;
+typedef unsigned __int32        khronos_uint32_t;
+typedef __int64                 khronos_int64_t;
+typedef unsigned __int64        khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(__sun__) || defined(__digital__)
+
+/*
+ * Sun or Digital
+ */
+typedef int                     khronos_int32_t;
+typedef unsigned int            khronos_uint32_t;
+#if defined(__arch64__) || defined(_LP64)
+typedef long int                khronos_int64_t;
+typedef unsigned long int       khronos_uint64_t;
+#else
+typedef long long int           khronos_int64_t;
+typedef unsigned long long int  khronos_uint64_t;
+#endif /* __arch64__ */
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif 0
+
+/*
+ * Hypothetical platform with no float or int64 support
+ */
+typedef int                     khronos_int32_t;
+typedef unsigned int            khronos_uint32_t;
+#define KHRONOS_SUPPORT_INT64   0
+#define KHRONOS_SUPPORT_FLOAT   0
+
+#else
+
+/*
+ * Generic fallback
+ */
+#include <stdint.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#endif
+
+
+/*
+ * Types that are (so far) the same on all platforms
+ */
+typedef signed   char          khronos_int8_t;
+typedef unsigned char          khronos_uint8_t;
+typedef signed   short int     khronos_int16_t;
+typedef unsigned short int     khronos_uint16_t;
+
+/*
+ * Types that differ between LLP64 and LP64 architectures - in LLP64,
+ * pointers are 64 bits, but 'long' is still 32 bits. Win64 appears
+ * to be the only LLP64 architecture in current use.
+ */
+#ifdef _WIN64
+typedef signed   long long int khronos_intptr_t;
+typedef unsigned long long int khronos_uintptr_t;
+typedef signed   long long int khronos_ssize_t;
+typedef unsigned long long int khronos_usize_t;
+#else
+typedef signed   long  int     khronos_intptr_t;
+typedef unsigned long  int     khronos_uintptr_t;
+typedef signed   long  int     khronos_ssize_t;
+typedef unsigned long  int     khronos_usize_t;
+#endif
+
+#if KHRONOS_SUPPORT_FLOAT
+/*
+ * Float type
+ */
+typedef          float         khronos_float_t;
+#endif
+
+#if KHRONOS_SUPPORT_INT64
+/* Time types
+ *
+ * These types can be used to represent a time interval in nanoseconds or
+ * an absolute Unadjusted System Time.  Unadjusted System Time is the number
+ * of nanoseconds since some arbitrary system event (e.g. since the last
+ * time the system booted).  The Unadjusted System Time is an unsigned
+ * 64 bit value that wraps back to 0 every 584 years.  Time intervals
+ * may be either signed or unsigned.
+ */
+typedef khronos_uint64_t       khronos_utime_nanoseconds_t;
+typedef khronos_int64_t        khronos_stime_nanoseconds_t;
+#endif
+
+/*
+ * Dummy value used to pad enum types to 32 bits.
+ */
+#ifndef KHRONOS_MAX_ENUM
+#define KHRONOS_MAX_ENUM 0x7FFFFFFF
+#endif
+
+/*
+ * Enumerated boolean type
+ *
+ * Values other than zero should be considered to be true.  Therefore
+ * comparisons should not be made against KHRONOS_TRUE.
+ */
+typedef enum {
+    KHRONOS_FALSE = 0,
+    KHRONOS_TRUE  = 1,
+    KHRONOS_BOOLEAN_ENUM_FORCE_SIZE = KHRONOS_MAX_ENUM
+} khronos_boolean_enum_t;
+
+#endif /* __khrplatform_h_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/README.md glmark2-2021.02/android/jni/src/glad/README.md
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/README.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/README.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,7 @@
+This directory contains a copy of the
+[glad2](https://github.com/Dav1dde/glad/tree/glad2) library.
+
+The contents of this directory can be recreated (at the latest available
+upstream version) with [this][permalink] permalink.
+
+[permalink]: http://gen.glad.sh/#profile=gl%3Dcompatibility%2Cgles1%3Dcommon&api=egl%3D1.5%2Cgl%3D3.3%2Cgles2%3D2.0%2Cglx%3D1.4%2Cwgl%3D1.0&extensions=EGL_EXT_platform_base%2CEGL_KHR_platform_gbm%2CEGL_KHR_platform_wayland%2CEGL_KHR_platform_x11%2CGLX_EXT_swap_control%2CGLX_MESA_swap_control%2CGL_OES_mapbuffer%2CGL_OES_required_internalformat%2CWGL_ARB_extensions_string%2CWGL_EXT_extensions_string%2CWGL_EXT_swap_control&generator=c
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/egl.c glmark2-2021.02/android/jni/src/glad/src/egl.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/egl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/src/egl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,258 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/egl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_EGL_VERSION_1_0 = 0;
+int GLAD_EGL_VERSION_1_1 = 0;
+int GLAD_EGL_VERSION_1_2 = 0;
+int GLAD_EGL_VERSION_1_3 = 0;
+int GLAD_EGL_VERSION_1_4 = 0;
+int GLAD_EGL_VERSION_1_5 = 0;
+int GLAD_EGL_EXT_platform_base = 0;
+int GLAD_EGL_KHR_platform_gbm = 0;
+int GLAD_EGL_KHR_platform_wayland = 0;
+int GLAD_EGL_KHR_platform_x11 = 0;
+
+
+
+PFNEGLBINDAPIPROC glad_eglBindAPI = NULL;
+PFNEGLBINDTEXIMAGEPROC glad_eglBindTexImage = NULL;
+PFNEGLCHOOSECONFIGPROC glad_eglChooseConfig = NULL;
+PFNEGLCLIENTWAITSYNCPROC glad_eglClientWaitSync = NULL;
+PFNEGLCOPYBUFFERSPROC glad_eglCopyBuffers = NULL;
+PFNEGLCREATECONTEXTPROC glad_eglCreateContext = NULL;
+PFNEGLCREATEIMAGEPROC glad_eglCreateImage = NULL;
+PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC glad_eglCreatePbufferFromClientBuffer = NULL;
+PFNEGLCREATEPBUFFERSURFACEPROC glad_eglCreatePbufferSurface = NULL;
+PFNEGLCREATEPIXMAPSURFACEPROC glad_eglCreatePixmapSurface = NULL;
+PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC glad_eglCreatePlatformPixmapSurface = NULL;
+PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC glad_eglCreatePlatformPixmapSurfaceEXT = NULL;
+PFNEGLCREATEPLATFORMWINDOWSURFACEPROC glad_eglCreatePlatformWindowSurface = NULL;
+PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC glad_eglCreatePlatformWindowSurfaceEXT = NULL;
+PFNEGLCREATESYNCPROC glad_eglCreateSync = NULL;
+PFNEGLCREATEWINDOWSURFACEPROC glad_eglCreateWindowSurface = NULL;
+PFNEGLDESTROYCONTEXTPROC glad_eglDestroyContext = NULL;
+PFNEGLDESTROYIMAGEPROC glad_eglDestroyImage = NULL;
+PFNEGLDESTROYSURFACEPROC glad_eglDestroySurface = NULL;
+PFNEGLDESTROYSYNCPROC glad_eglDestroySync = NULL;
+PFNEGLGETCONFIGATTRIBPROC glad_eglGetConfigAttrib = NULL;
+PFNEGLGETCONFIGSPROC glad_eglGetConfigs = NULL;
+PFNEGLGETCURRENTCONTEXTPROC glad_eglGetCurrentContext = NULL;
+PFNEGLGETCURRENTDISPLAYPROC glad_eglGetCurrentDisplay = NULL;
+PFNEGLGETCURRENTSURFACEPROC glad_eglGetCurrentSurface = NULL;
+PFNEGLGETDISPLAYPROC glad_eglGetDisplay = NULL;
+PFNEGLGETERRORPROC glad_eglGetError = NULL;
+PFNEGLGETPLATFORMDISPLAYPROC glad_eglGetPlatformDisplay = NULL;
+PFNEGLGETPLATFORMDISPLAYEXTPROC glad_eglGetPlatformDisplayEXT = NULL;
+PFNEGLGETPROCADDRESSPROC glad_eglGetProcAddress = NULL;
+PFNEGLGETSYNCATTRIBPROC glad_eglGetSyncAttrib = NULL;
+PFNEGLINITIALIZEPROC glad_eglInitialize = NULL;
+PFNEGLMAKECURRENTPROC glad_eglMakeCurrent = NULL;
+PFNEGLQUERYAPIPROC glad_eglQueryAPI = NULL;
+PFNEGLQUERYCONTEXTPROC glad_eglQueryContext = NULL;
+PFNEGLQUERYSTRINGPROC glad_eglQueryString = NULL;
+PFNEGLQUERYSURFACEPROC glad_eglQuerySurface = NULL;
+PFNEGLRELEASETEXIMAGEPROC glad_eglReleaseTexImage = NULL;
+PFNEGLRELEASETHREADPROC glad_eglReleaseThread = NULL;
+PFNEGLSURFACEATTRIBPROC glad_eglSurfaceAttrib = NULL;
+PFNEGLSWAPBUFFERSPROC glad_eglSwapBuffers = NULL;
+PFNEGLSWAPINTERVALPROC glad_eglSwapInterval = NULL;
+PFNEGLTERMINATEPROC glad_eglTerminate = NULL;
+PFNEGLWAITCLIENTPROC glad_eglWaitClient = NULL;
+PFNEGLWAITGLPROC glad_eglWaitGL = NULL;
+PFNEGLWAITNATIVEPROC glad_eglWaitNative = NULL;
+PFNEGLWAITSYNCPROC glad_eglWaitSync = NULL;
+
+
+static void glad_egl_load_EGL_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_0) return;
+    glad_eglChooseConfig = (PFNEGLCHOOSECONFIGPROC) load(userptr, "eglChooseConfig");
+    glad_eglCopyBuffers = (PFNEGLCOPYBUFFERSPROC) load(userptr, "eglCopyBuffers");
+    glad_eglCreateContext = (PFNEGLCREATECONTEXTPROC) load(userptr, "eglCreateContext");
+    glad_eglCreatePbufferSurface = (PFNEGLCREATEPBUFFERSURFACEPROC) load(userptr, "eglCreatePbufferSurface");
+    glad_eglCreatePixmapSurface = (PFNEGLCREATEPIXMAPSURFACEPROC) load(userptr, "eglCreatePixmapSurface");
+    glad_eglCreateWindowSurface = (PFNEGLCREATEWINDOWSURFACEPROC) load(userptr, "eglCreateWindowSurface");
+    glad_eglDestroyContext = (PFNEGLDESTROYCONTEXTPROC) load(userptr, "eglDestroyContext");
+    glad_eglDestroySurface = (PFNEGLDESTROYSURFACEPROC) load(userptr, "eglDestroySurface");
+    glad_eglGetConfigAttrib = (PFNEGLGETCONFIGATTRIBPROC) load(userptr, "eglGetConfigAttrib");
+    glad_eglGetConfigs = (PFNEGLGETCONFIGSPROC) load(userptr, "eglGetConfigs");
+    glad_eglGetCurrentDisplay = (PFNEGLGETCURRENTDISPLAYPROC) load(userptr, "eglGetCurrentDisplay");
+    glad_eglGetCurrentSurface = (PFNEGLGETCURRENTSURFACEPROC) load(userptr, "eglGetCurrentSurface");
+    glad_eglGetDisplay = (PFNEGLGETDISPLAYPROC) load(userptr, "eglGetDisplay");
+    glad_eglGetError = (PFNEGLGETERRORPROC) load(userptr, "eglGetError");
+    glad_eglGetProcAddress = (PFNEGLGETPROCADDRESSPROC) load(userptr, "eglGetProcAddress");
+    glad_eglInitialize = (PFNEGLINITIALIZEPROC) load(userptr, "eglInitialize");
+    glad_eglMakeCurrent = (PFNEGLMAKECURRENTPROC) load(userptr, "eglMakeCurrent");
+    glad_eglQueryContext = (PFNEGLQUERYCONTEXTPROC) load(userptr, "eglQueryContext");
+    glad_eglQueryString = (PFNEGLQUERYSTRINGPROC) load(userptr, "eglQueryString");
+    glad_eglQuerySurface = (PFNEGLQUERYSURFACEPROC) load(userptr, "eglQuerySurface");
+    glad_eglSwapBuffers = (PFNEGLSWAPBUFFERSPROC) load(userptr, "eglSwapBuffers");
+    glad_eglTerminate = (PFNEGLTERMINATEPROC) load(userptr, "eglTerminate");
+    glad_eglWaitGL = (PFNEGLWAITGLPROC) load(userptr, "eglWaitGL");
+    glad_eglWaitNative = (PFNEGLWAITNATIVEPROC) load(userptr, "eglWaitNative");
+}
+static void glad_egl_load_EGL_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_1) return;
+    glad_eglBindTexImage = (PFNEGLBINDTEXIMAGEPROC) load(userptr, "eglBindTexImage");
+    glad_eglReleaseTexImage = (PFNEGLRELEASETEXIMAGEPROC) load(userptr, "eglReleaseTexImage");
+    glad_eglSurfaceAttrib = (PFNEGLSURFACEATTRIBPROC) load(userptr, "eglSurfaceAttrib");
+    glad_eglSwapInterval = (PFNEGLSWAPINTERVALPROC) load(userptr, "eglSwapInterval");
+}
+static void glad_egl_load_EGL_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_2) return;
+    glad_eglBindAPI = (PFNEGLBINDAPIPROC) load(userptr, "eglBindAPI");
+    glad_eglCreatePbufferFromClientBuffer = (PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC) load(userptr, "eglCreatePbufferFromClientBuffer");
+    glad_eglQueryAPI = (PFNEGLQUERYAPIPROC) load(userptr, "eglQueryAPI");
+    glad_eglReleaseThread = (PFNEGLRELEASETHREADPROC) load(userptr, "eglReleaseThread");
+    glad_eglWaitClient = (PFNEGLWAITCLIENTPROC) load(userptr, "eglWaitClient");
+}
+static void glad_egl_load_EGL_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_4) return;
+    glad_eglGetCurrentContext = (PFNEGLGETCURRENTCONTEXTPROC) load(userptr, "eglGetCurrentContext");
+}
+static void glad_egl_load_EGL_VERSION_1_5( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_5) return;
+    glad_eglClientWaitSync = (PFNEGLCLIENTWAITSYNCPROC) load(userptr, "eglClientWaitSync");
+    glad_eglCreateImage = (PFNEGLCREATEIMAGEPROC) load(userptr, "eglCreateImage");
+    glad_eglCreatePlatformPixmapSurface = (PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC) load(userptr, "eglCreatePlatformPixmapSurface");
+    glad_eglCreatePlatformWindowSurface = (PFNEGLCREATEPLATFORMWINDOWSURFACEPROC) load(userptr, "eglCreatePlatformWindowSurface");
+    glad_eglCreateSync = (PFNEGLCREATESYNCPROC) load(userptr, "eglCreateSync");
+    glad_eglDestroyImage = (PFNEGLDESTROYIMAGEPROC) load(userptr, "eglDestroyImage");
+    glad_eglDestroySync = (PFNEGLDESTROYSYNCPROC) load(userptr, "eglDestroySync");
+    glad_eglGetPlatformDisplay = (PFNEGLGETPLATFORMDISPLAYPROC) load(userptr, "eglGetPlatformDisplay");
+    glad_eglGetSyncAttrib = (PFNEGLGETSYNCATTRIBPROC) load(userptr, "eglGetSyncAttrib");
+    glad_eglWaitSync = (PFNEGLWAITSYNCPROC) load(userptr, "eglWaitSync");
+}
+static void glad_egl_load_EGL_EXT_platform_base( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_EXT_platform_base) return;
+    glad_eglCreatePlatformPixmapSurfaceEXT = (PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC) load(userptr, "eglCreatePlatformPixmapSurfaceEXT");
+    glad_eglCreatePlatformWindowSurfaceEXT = (PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC) load(userptr, "eglCreatePlatformWindowSurfaceEXT");
+    glad_eglGetPlatformDisplayEXT = (PFNEGLGETPLATFORMDISPLAYEXTPROC) load(userptr, "eglGetPlatformDisplayEXT");
+}
+
+
+
+static int glad_egl_get_extensions(EGLDisplay display, const char **extensions) {
+    *extensions = eglQueryString(display, EGL_EXTENSIONS);
+
+    return extensions != NULL;
+}
+
+static int glad_egl_has_extension(const char *extensions, const char *ext) {
+    const char *loc;
+    const char *terminator;
+    if(extensions == NULL) {
+        return 0;
+    }
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL) {
+            return 0;
+        }
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0')) {
+            return 1;
+        }
+        extensions = terminator;
+    }
+}
+
+static GLADapiproc glad_egl_get_proc_from_userptr(void *userptr, const char *name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_egl_find_extensions_egl(EGLDisplay display) {
+    const char *extensions;
+    if (!glad_egl_get_extensions(display, &extensions)) return 0;
+
+    GLAD_EGL_EXT_platform_base = glad_egl_has_extension(extensions, "EGL_EXT_platform_base");
+    GLAD_EGL_KHR_platform_gbm = glad_egl_has_extension(extensions, "EGL_KHR_platform_gbm");
+    GLAD_EGL_KHR_platform_wayland = glad_egl_has_extension(extensions, "EGL_KHR_platform_wayland");
+    GLAD_EGL_KHR_platform_x11 = glad_egl_has_extension(extensions, "EGL_KHR_platform_x11");
+
+    return 1;
+}
+
+static int glad_egl_find_core_egl(EGLDisplay display) {
+    int major, minor;
+    const char *version;
+
+    if (display == NULL) {
+        display = EGL_NO_DISPLAY; /* this is usually NULL, better safe than sorry */
+    }
+    if (display == EGL_NO_DISPLAY) {
+        display = eglGetCurrentDisplay();
+    }
+#ifdef EGL_VERSION_1_4
+    if (display == EGL_NO_DISPLAY) {
+        display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
+    }
+#endif
+#ifndef EGL_VERSION_1_5
+    if (display == EGL_NO_DISPLAY) {
+        return 0;
+    }
+#endif
+
+    version = eglQueryString(display, EGL_VERSION);
+    (void) eglGetError();
+
+    if (version == NULL) {
+        major = 1;
+        minor = 0;
+    } else {
+        GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+    }
+
+    GLAD_EGL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_EGL_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_EGL_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_EGL_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_EGL_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    GLAD_EGL_VERSION_1_5 = (major == 1 && minor >= 5) || major > 1;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadEGLUserPtr(EGLDisplay display, GLADuserptrloadfunc load, void* userptr) {
+    int version;
+    eglGetDisplay = (PFNEGLGETDISPLAYPROC) load(userptr, "eglGetDisplay");
+    eglGetCurrentDisplay = (PFNEGLGETCURRENTDISPLAYPROC) load(userptr, "eglGetCurrentDisplay");
+    eglQueryString = (PFNEGLQUERYSTRINGPROC) load(userptr, "eglQueryString");
+    eglGetError = (PFNEGLGETERRORPROC) load(userptr, "eglGetError");
+    if (eglGetDisplay == NULL || eglGetCurrentDisplay == NULL || eglQueryString == NULL || eglGetError == NULL) return 0;
+
+    version = glad_egl_find_core_egl(display);
+    if (!version) return 0;
+    glad_egl_load_EGL_VERSION_1_0(load, userptr);
+    glad_egl_load_EGL_VERSION_1_1(load, userptr);
+    glad_egl_load_EGL_VERSION_1_2(load, userptr);
+    glad_egl_load_EGL_VERSION_1_4(load, userptr);
+    glad_egl_load_EGL_VERSION_1_5(load, userptr);
+
+    if (!glad_egl_find_extensions_egl(display)) return 0;
+    glad_egl_load_EGL_EXT_platform_base(load, userptr);
+
+    return version;
+}
+
+int gladLoadEGL(EGLDisplay display, GLADloadfunc load) {
+    return gladLoadEGLUserPtr(display, glad_egl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/gl.c glmark2-2021.02/android/jni/src/glad/src/gl.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/gl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/src/gl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1709 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/gl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GL_VERSION_1_0 = 0;
+int GLAD_GL_VERSION_1_1 = 0;
+int GLAD_GL_VERSION_1_2 = 0;
+int GLAD_GL_VERSION_1_3 = 0;
+int GLAD_GL_VERSION_1_4 = 0;
+int GLAD_GL_VERSION_1_5 = 0;
+int GLAD_GL_VERSION_2_0 = 0;
+int GLAD_GL_VERSION_2_1 = 0;
+int GLAD_GL_VERSION_3_0 = 0;
+int GLAD_GL_VERSION_3_1 = 0;
+int GLAD_GL_VERSION_3_2 = 0;
+int GLAD_GL_VERSION_3_3 = 0;
+
+
+
+PFNGLACCUMPROC glad_glAccum = NULL;
+PFNGLACTIVETEXTUREPROC glad_glActiveTexture = NULL;
+PFNGLALPHAFUNCPROC glad_glAlphaFunc = NULL;
+PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident = NULL;
+PFNGLARRAYELEMENTPROC glad_glArrayElement = NULL;
+PFNGLATTACHSHADERPROC glad_glAttachShader = NULL;
+PFNGLBEGINPROC glad_glBegin = NULL;
+PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender = NULL;
+PFNGLBEGINQUERYPROC glad_glBeginQuery = NULL;
+PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback = NULL;
+PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation = NULL;
+PFNGLBINDBUFFERPROC glad_glBindBuffer = NULL;
+PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase = NULL;
+PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange = NULL;
+PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation = NULL;
+PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed = NULL;
+PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer = NULL;
+PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer = NULL;
+PFNGLBINDSAMPLERPROC glad_glBindSampler = NULL;
+PFNGLBINDTEXTUREPROC glad_glBindTexture = NULL;
+PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray = NULL;
+PFNGLBITMAPPROC glad_glBitmap = NULL;
+PFNGLBLENDCOLORPROC glad_glBlendColor = NULL;
+PFNGLBLENDEQUATIONPROC glad_glBlendEquation = NULL;
+PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate = NULL;
+PFNGLBLENDFUNCPROC glad_glBlendFunc = NULL;
+PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate = NULL;
+PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer = NULL;
+PFNGLBUFFERDATAPROC glad_glBufferData = NULL;
+PFNGLBUFFERSUBDATAPROC glad_glBufferSubData = NULL;
+PFNGLCALLLISTPROC glad_glCallList = NULL;
+PFNGLCALLLISTSPROC glad_glCallLists = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus = NULL;
+PFNGLCLAMPCOLORPROC glad_glClampColor = NULL;
+PFNGLCLEARPROC glad_glClear = NULL;
+PFNGLCLEARACCUMPROC glad_glClearAccum = NULL;
+PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi = NULL;
+PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv = NULL;
+PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv = NULL;
+PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv = NULL;
+PFNGLCLEARCOLORPROC glad_glClearColor = NULL;
+PFNGLCLEARDEPTHPROC glad_glClearDepth = NULL;
+PFNGLCLEARINDEXPROC glad_glClearIndex = NULL;
+PFNGLCLEARSTENCILPROC glad_glClearStencil = NULL;
+PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture = NULL;
+PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync = NULL;
+PFNGLCLIPPLANEPROC glad_glClipPlane = NULL;
+PFNGLCOLOR3BPROC glad_glColor3b = NULL;
+PFNGLCOLOR3BVPROC glad_glColor3bv = NULL;
+PFNGLCOLOR3DPROC glad_glColor3d = NULL;
+PFNGLCOLOR3DVPROC glad_glColor3dv = NULL;
+PFNGLCOLOR3FPROC glad_glColor3f = NULL;
+PFNGLCOLOR3FVPROC glad_glColor3fv = NULL;
+PFNGLCOLOR3IPROC glad_glColor3i = NULL;
+PFNGLCOLOR3IVPROC glad_glColor3iv = NULL;
+PFNGLCOLOR3SPROC glad_glColor3s = NULL;
+PFNGLCOLOR3SVPROC glad_glColor3sv = NULL;
+PFNGLCOLOR3UBPROC glad_glColor3ub = NULL;
+PFNGLCOLOR3UBVPROC glad_glColor3ubv = NULL;
+PFNGLCOLOR3UIPROC glad_glColor3ui = NULL;
+PFNGLCOLOR3UIVPROC glad_glColor3uiv = NULL;
+PFNGLCOLOR3USPROC glad_glColor3us = NULL;
+PFNGLCOLOR3USVPROC glad_glColor3usv = NULL;
+PFNGLCOLOR4BPROC glad_glColor4b = NULL;
+PFNGLCOLOR4BVPROC glad_glColor4bv = NULL;
+PFNGLCOLOR4DPROC glad_glColor4d = NULL;
+PFNGLCOLOR4DVPROC glad_glColor4dv = NULL;
+PFNGLCOLOR4FPROC glad_glColor4f = NULL;
+PFNGLCOLOR4FVPROC glad_glColor4fv = NULL;
+PFNGLCOLOR4IPROC glad_glColor4i = NULL;
+PFNGLCOLOR4IVPROC glad_glColor4iv = NULL;
+PFNGLCOLOR4SPROC glad_glColor4s = NULL;
+PFNGLCOLOR4SVPROC glad_glColor4sv = NULL;
+PFNGLCOLOR4UBPROC glad_glColor4ub = NULL;
+PFNGLCOLOR4UBVPROC glad_glColor4ubv = NULL;
+PFNGLCOLOR4UIPROC glad_glColor4ui = NULL;
+PFNGLCOLOR4UIVPROC glad_glColor4uiv = NULL;
+PFNGLCOLOR4USPROC glad_glColor4us = NULL;
+PFNGLCOLOR4USVPROC glad_glColor4usv = NULL;
+PFNGLCOLORMASKPROC glad_glColorMask = NULL;
+PFNGLCOLORMASKIPROC glad_glColorMaski = NULL;
+PFNGLCOLORMATERIALPROC glad_glColorMaterial = NULL;
+PFNGLCOLORP3UIPROC glad_glColorP3ui = NULL;
+PFNGLCOLORP3UIVPROC glad_glColorP3uiv = NULL;
+PFNGLCOLORP4UIPROC glad_glColorP4ui = NULL;
+PFNGLCOLORP4UIVPROC glad_glColorP4uiv = NULL;
+PFNGLCOLORPOINTERPROC glad_glColorPointer = NULL;
+PFNGLCOMPILESHADERPROC glad_glCompileShader = NULL;
+PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D = NULL;
+PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData = NULL;
+PFNGLCOPYPIXELSPROC glad_glCopyPixels = NULL;
+PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D = NULL;
+PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D = NULL;
+PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D = NULL;
+PFNGLCREATEPROGRAMPROC glad_glCreateProgram = NULL;
+PFNGLCREATESHADERPROC glad_glCreateShader = NULL;
+PFNGLCULLFACEPROC glad_glCullFace = NULL;
+PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers = NULL;
+PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers = NULL;
+PFNGLDELETELISTSPROC glad_glDeleteLists = NULL;
+PFNGLDELETEPROGRAMPROC glad_glDeleteProgram = NULL;
+PFNGLDELETEQUERIESPROC glad_glDeleteQueries = NULL;
+PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers = NULL;
+PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers = NULL;
+PFNGLDELETESHADERPROC glad_glDeleteShader = NULL;
+PFNGLDELETESYNCPROC glad_glDeleteSync = NULL;
+PFNGLDELETETEXTURESPROC glad_glDeleteTextures = NULL;
+PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays = NULL;
+PFNGLDEPTHFUNCPROC glad_glDepthFunc = NULL;
+PFNGLDEPTHMASKPROC glad_glDepthMask = NULL;
+PFNGLDEPTHRANGEPROC glad_glDepthRange = NULL;
+PFNGLDETACHSHADERPROC glad_glDetachShader = NULL;
+PFNGLDISABLEPROC glad_glDisable = NULL;
+PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray = NULL;
+PFNGLDISABLEIPROC glad_glDisablei = NULL;
+PFNGLDRAWARRAYSPROC glad_glDrawArrays = NULL;
+PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced = NULL;
+PFNGLDRAWBUFFERPROC glad_glDrawBuffer = NULL;
+PFNGLDRAWBUFFERSPROC glad_glDrawBuffers = NULL;
+PFNGLDRAWELEMENTSPROC glad_glDrawElements = NULL;
+PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex = NULL;
+PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced = NULL;
+PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex = NULL;
+PFNGLDRAWPIXELSPROC glad_glDrawPixels = NULL;
+PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements = NULL;
+PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex = NULL;
+PFNGLEDGEFLAGPROC glad_glEdgeFlag = NULL;
+PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer = NULL;
+PFNGLEDGEFLAGVPROC glad_glEdgeFlagv = NULL;
+PFNGLENABLEPROC glad_glEnable = NULL;
+PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray = NULL;
+PFNGLENABLEIPROC glad_glEnablei = NULL;
+PFNGLENDPROC glad_glEnd = NULL;
+PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender = NULL;
+PFNGLENDLISTPROC glad_glEndList = NULL;
+PFNGLENDQUERYPROC glad_glEndQuery = NULL;
+PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback = NULL;
+PFNGLEVALCOORD1DPROC glad_glEvalCoord1d = NULL;
+PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv = NULL;
+PFNGLEVALCOORD1FPROC glad_glEvalCoord1f = NULL;
+PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv = NULL;
+PFNGLEVALCOORD2DPROC glad_glEvalCoord2d = NULL;
+PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv = NULL;
+PFNGLEVALCOORD2FPROC glad_glEvalCoord2f = NULL;
+PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv = NULL;
+PFNGLEVALMESH1PROC glad_glEvalMesh1 = NULL;
+PFNGLEVALMESH2PROC glad_glEvalMesh2 = NULL;
+PFNGLEVALPOINT1PROC glad_glEvalPoint1 = NULL;
+PFNGLEVALPOINT2PROC glad_glEvalPoint2 = NULL;
+PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer = NULL;
+PFNGLFENCESYNCPROC glad_glFenceSync = NULL;
+PFNGLFINISHPROC glad_glFinish = NULL;
+PFNGLFLUSHPROC glad_glFlush = NULL;
+PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange = NULL;
+PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer = NULL;
+PFNGLFOGCOORDDPROC glad_glFogCoordd = NULL;
+PFNGLFOGCOORDDVPROC glad_glFogCoorddv = NULL;
+PFNGLFOGCOORDFPROC glad_glFogCoordf = NULL;
+PFNGLFOGCOORDFVPROC glad_glFogCoordfv = NULL;
+PFNGLFOGFPROC glad_glFogf = NULL;
+PFNGLFOGFVPROC glad_glFogfv = NULL;
+PFNGLFOGIPROC glad_glFogi = NULL;
+PFNGLFOGIVPROC glad_glFogiv = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer = NULL;
+PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture = NULL;
+PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D = NULL;
+PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D = NULL;
+PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer = NULL;
+PFNGLFRONTFACEPROC glad_glFrontFace = NULL;
+PFNGLFRUSTUMPROC glad_glFrustum = NULL;
+PFNGLGENBUFFERSPROC glad_glGenBuffers = NULL;
+PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers = NULL;
+PFNGLGENLISTSPROC glad_glGenLists = NULL;
+PFNGLGENQUERIESPROC glad_glGenQueries = NULL;
+PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers = NULL;
+PFNGLGENSAMPLERSPROC glad_glGenSamplers = NULL;
+PFNGLGENTEXTURESPROC glad_glGenTextures = NULL;
+PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays = NULL;
+PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap = NULL;
+PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib = NULL;
+PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform = NULL;
+PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName = NULL;
+PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv = NULL;
+PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName = NULL;
+PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv = NULL;
+PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders = NULL;
+PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation = NULL;
+PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v = NULL;
+PFNGLGETBOOLEANVPROC glad_glGetBooleanv = NULL;
+PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v = NULL;
+PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv = NULL;
+PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv = NULL;
+PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData = NULL;
+PFNGLGETCLIPPLANEPROC glad_glGetClipPlane = NULL;
+PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage = NULL;
+PFNGLGETDOUBLEVPROC glad_glGetDoublev = NULL;
+PFNGLGETERRORPROC glad_glGetError = NULL;
+PFNGLGETFLOATVPROC glad_glGetFloatv = NULL;
+PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex = NULL;
+PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv = NULL;
+PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v = NULL;
+PFNGLGETINTEGER64VPROC glad_glGetInteger64v = NULL;
+PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v = NULL;
+PFNGLGETINTEGERVPROC glad_glGetIntegerv = NULL;
+PFNGLGETLIGHTFVPROC glad_glGetLightfv = NULL;
+PFNGLGETLIGHTIVPROC glad_glGetLightiv = NULL;
+PFNGLGETMAPDVPROC glad_glGetMapdv = NULL;
+PFNGLGETMAPFVPROC glad_glGetMapfv = NULL;
+PFNGLGETMAPIVPROC glad_glGetMapiv = NULL;
+PFNGLGETMATERIALFVPROC glad_glGetMaterialfv = NULL;
+PFNGLGETMATERIALIVPROC glad_glGetMaterialiv = NULL;
+PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv = NULL;
+PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv = NULL;
+PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv = NULL;
+PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv = NULL;
+PFNGLGETPOINTERVPROC glad_glGetPointerv = NULL;
+PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple = NULL;
+PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog = NULL;
+PFNGLGETPROGRAMIVPROC glad_glGetProgramiv = NULL;
+PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v = NULL;
+PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv = NULL;
+PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v = NULL;
+PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv = NULL;
+PFNGLGETQUERYIVPROC glad_glGetQueryiv = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv = NULL;
+PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv = NULL;
+PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv = NULL;
+PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv = NULL;
+PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv = NULL;
+PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog = NULL;
+PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource = NULL;
+PFNGLGETSHADERIVPROC glad_glGetShaderiv = NULL;
+PFNGLGETSTRINGPROC glad_glGetString = NULL;
+PFNGLGETSTRINGIPROC glad_glGetStringi = NULL;
+PFNGLGETSYNCIVPROC glad_glGetSynciv = NULL;
+PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv = NULL;
+PFNGLGETTEXENVIVPROC glad_glGetTexEnviv = NULL;
+PFNGLGETTEXGENDVPROC glad_glGetTexGendv = NULL;
+PFNGLGETTEXGENFVPROC glad_glGetTexGenfv = NULL;
+PFNGLGETTEXGENIVPROC glad_glGetTexGeniv = NULL;
+PFNGLGETTEXIMAGEPROC glad_glGetTexImage = NULL;
+PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv = NULL;
+PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv = NULL;
+PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv = NULL;
+PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv = NULL;
+PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv = NULL;
+PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv = NULL;
+PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying = NULL;
+PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex = NULL;
+PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices = NULL;
+PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation = NULL;
+PFNGLGETUNIFORMFVPROC glad_glGetUniformfv = NULL;
+PFNGLGETUNIFORMIVPROC glad_glGetUniformiv = NULL;
+PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv = NULL;
+PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv = NULL;
+PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv = NULL;
+PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv = NULL;
+PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv = NULL;
+PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv = NULL;
+PFNGLHINTPROC glad_glHint = NULL;
+PFNGLINDEXMASKPROC glad_glIndexMask = NULL;
+PFNGLINDEXPOINTERPROC glad_glIndexPointer = NULL;
+PFNGLINDEXDPROC glad_glIndexd = NULL;
+PFNGLINDEXDVPROC glad_glIndexdv = NULL;
+PFNGLINDEXFPROC glad_glIndexf = NULL;
+PFNGLINDEXFVPROC glad_glIndexfv = NULL;
+PFNGLINDEXIPROC glad_glIndexi = NULL;
+PFNGLINDEXIVPROC glad_glIndexiv = NULL;
+PFNGLINDEXSPROC glad_glIndexs = NULL;
+PFNGLINDEXSVPROC glad_glIndexsv = NULL;
+PFNGLINDEXUBPROC glad_glIndexub = NULL;
+PFNGLINDEXUBVPROC glad_glIndexubv = NULL;
+PFNGLINITNAMESPROC glad_glInitNames = NULL;
+PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays = NULL;
+PFNGLISBUFFERPROC glad_glIsBuffer = NULL;
+PFNGLISENABLEDPROC glad_glIsEnabled = NULL;
+PFNGLISENABLEDIPROC glad_glIsEnabledi = NULL;
+PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer = NULL;
+PFNGLISLISTPROC glad_glIsList = NULL;
+PFNGLISPROGRAMPROC glad_glIsProgram = NULL;
+PFNGLISQUERYPROC glad_glIsQuery = NULL;
+PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer = NULL;
+PFNGLISSAMPLERPROC glad_glIsSampler = NULL;
+PFNGLISSHADERPROC glad_glIsShader = NULL;
+PFNGLISSYNCPROC glad_glIsSync = NULL;
+PFNGLISTEXTUREPROC glad_glIsTexture = NULL;
+PFNGLISVERTEXARRAYPROC glad_glIsVertexArray = NULL;
+PFNGLLIGHTMODELFPROC glad_glLightModelf = NULL;
+PFNGLLIGHTMODELFVPROC glad_glLightModelfv = NULL;
+PFNGLLIGHTMODELIPROC glad_glLightModeli = NULL;
+PFNGLLIGHTMODELIVPROC glad_glLightModeliv = NULL;
+PFNGLLIGHTFPROC glad_glLightf = NULL;
+PFNGLLIGHTFVPROC glad_glLightfv = NULL;
+PFNGLLIGHTIPROC glad_glLighti = NULL;
+PFNGLLIGHTIVPROC glad_glLightiv = NULL;
+PFNGLLINESTIPPLEPROC glad_glLineStipple = NULL;
+PFNGLLINEWIDTHPROC glad_glLineWidth = NULL;
+PFNGLLINKPROGRAMPROC glad_glLinkProgram = NULL;
+PFNGLLISTBASEPROC glad_glListBase = NULL;
+PFNGLLOADIDENTITYPROC glad_glLoadIdentity = NULL;
+PFNGLLOADMATRIXDPROC glad_glLoadMatrixd = NULL;
+PFNGLLOADMATRIXFPROC glad_glLoadMatrixf = NULL;
+PFNGLLOADNAMEPROC glad_glLoadName = NULL;
+PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd = NULL;
+PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf = NULL;
+PFNGLLOGICOPPROC glad_glLogicOp = NULL;
+PFNGLMAP1DPROC glad_glMap1d = NULL;
+PFNGLMAP1FPROC glad_glMap1f = NULL;
+PFNGLMAP2DPROC glad_glMap2d = NULL;
+PFNGLMAP2FPROC glad_glMap2f = NULL;
+PFNGLMAPBUFFERPROC glad_glMapBuffer = NULL;
+PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange = NULL;
+PFNGLMAPGRID1DPROC glad_glMapGrid1d = NULL;
+PFNGLMAPGRID1FPROC glad_glMapGrid1f = NULL;
+PFNGLMAPGRID2DPROC glad_glMapGrid2d = NULL;
+PFNGLMAPGRID2FPROC glad_glMapGrid2f = NULL;
+PFNGLMATERIALFPROC glad_glMaterialf = NULL;
+PFNGLMATERIALFVPROC glad_glMaterialfv = NULL;
+PFNGLMATERIALIPROC glad_glMateriali = NULL;
+PFNGLMATERIALIVPROC glad_glMaterialiv = NULL;
+PFNGLMATRIXMODEPROC glad_glMatrixMode = NULL;
+PFNGLMULTMATRIXDPROC glad_glMultMatrixd = NULL;
+PFNGLMULTMATRIXFPROC glad_glMultMatrixf = NULL;
+PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd = NULL;
+PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf = NULL;
+PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays = NULL;
+PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements = NULL;
+PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex = NULL;
+PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d = NULL;
+PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv = NULL;
+PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f = NULL;
+PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv = NULL;
+PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i = NULL;
+PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv = NULL;
+PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s = NULL;
+PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv = NULL;
+PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d = NULL;
+PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv = NULL;
+PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f = NULL;
+PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv = NULL;
+PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i = NULL;
+PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv = NULL;
+PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s = NULL;
+PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv = NULL;
+PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d = NULL;
+PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv = NULL;
+PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f = NULL;
+PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv = NULL;
+PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i = NULL;
+PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv = NULL;
+PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s = NULL;
+PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv = NULL;
+PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d = NULL;
+PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv = NULL;
+PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f = NULL;
+PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv = NULL;
+PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i = NULL;
+PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv = NULL;
+PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s = NULL;
+PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv = NULL;
+PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui = NULL;
+PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv = NULL;
+PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui = NULL;
+PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv = NULL;
+PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui = NULL;
+PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv = NULL;
+PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui = NULL;
+PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv = NULL;
+PFNGLNEWLISTPROC glad_glNewList = NULL;
+PFNGLNORMAL3BPROC glad_glNormal3b = NULL;
+PFNGLNORMAL3BVPROC glad_glNormal3bv = NULL;
+PFNGLNORMAL3DPROC glad_glNormal3d = NULL;
+PFNGLNORMAL3DVPROC glad_glNormal3dv = NULL;
+PFNGLNORMAL3FPROC glad_glNormal3f = NULL;
+PFNGLNORMAL3FVPROC glad_glNormal3fv = NULL;
+PFNGLNORMAL3IPROC glad_glNormal3i = NULL;
+PFNGLNORMAL3IVPROC glad_glNormal3iv = NULL;
+PFNGLNORMAL3SPROC glad_glNormal3s = NULL;
+PFNGLNORMAL3SVPROC glad_glNormal3sv = NULL;
+PFNGLNORMALP3UIPROC glad_glNormalP3ui = NULL;
+PFNGLNORMALP3UIVPROC glad_glNormalP3uiv = NULL;
+PFNGLNORMALPOINTERPROC glad_glNormalPointer = NULL;
+PFNGLORTHOPROC glad_glOrtho = NULL;
+PFNGLPASSTHROUGHPROC glad_glPassThrough = NULL;
+PFNGLPIXELMAPFVPROC glad_glPixelMapfv = NULL;
+PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv = NULL;
+PFNGLPIXELMAPUSVPROC glad_glPixelMapusv = NULL;
+PFNGLPIXELSTOREFPROC glad_glPixelStoref = NULL;
+PFNGLPIXELSTOREIPROC glad_glPixelStorei = NULL;
+PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf = NULL;
+PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi = NULL;
+PFNGLPIXELZOOMPROC glad_glPixelZoom = NULL;
+PFNGLPOINTPARAMETERFPROC glad_glPointParameterf = NULL;
+PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv = NULL;
+PFNGLPOINTPARAMETERIPROC glad_glPointParameteri = NULL;
+PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv = NULL;
+PFNGLPOINTSIZEPROC glad_glPointSize = NULL;
+PFNGLPOLYGONMODEPROC glad_glPolygonMode = NULL;
+PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset = NULL;
+PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple = NULL;
+PFNGLPOPATTRIBPROC glad_glPopAttrib = NULL;
+PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib = NULL;
+PFNGLPOPMATRIXPROC glad_glPopMatrix = NULL;
+PFNGLPOPNAMEPROC glad_glPopName = NULL;
+PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex = NULL;
+PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures = NULL;
+PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex = NULL;
+PFNGLPUSHATTRIBPROC glad_glPushAttrib = NULL;
+PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib = NULL;
+PFNGLPUSHMATRIXPROC glad_glPushMatrix = NULL;
+PFNGLPUSHNAMEPROC glad_glPushName = NULL;
+PFNGLQUERYCOUNTERPROC glad_glQueryCounter = NULL;
+PFNGLRASTERPOS2DPROC glad_glRasterPos2d = NULL;
+PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv = NULL;
+PFNGLRASTERPOS2FPROC glad_glRasterPos2f = NULL;
+PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv = NULL;
+PFNGLRASTERPOS2IPROC glad_glRasterPos2i = NULL;
+PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv = NULL;
+PFNGLRASTERPOS2SPROC glad_glRasterPos2s = NULL;
+PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv = NULL;
+PFNGLRASTERPOS3DPROC glad_glRasterPos3d = NULL;
+PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv = NULL;
+PFNGLRASTERPOS3FPROC glad_glRasterPos3f = NULL;
+PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv = NULL;
+PFNGLRASTERPOS3IPROC glad_glRasterPos3i = NULL;
+PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv = NULL;
+PFNGLRASTERPOS3SPROC glad_glRasterPos3s = NULL;
+PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv = NULL;
+PFNGLRASTERPOS4DPROC glad_glRasterPos4d = NULL;
+PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv = NULL;
+PFNGLRASTERPOS4FPROC glad_glRasterPos4f = NULL;
+PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv = NULL;
+PFNGLRASTERPOS4IPROC glad_glRasterPos4i = NULL;
+PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv = NULL;
+PFNGLRASTERPOS4SPROC glad_glRasterPos4s = NULL;
+PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv = NULL;
+PFNGLREADBUFFERPROC glad_glReadBuffer = NULL;
+PFNGLREADPIXELSPROC glad_glReadPixels = NULL;
+PFNGLRECTDPROC glad_glRectd = NULL;
+PFNGLRECTDVPROC glad_glRectdv = NULL;
+PFNGLRECTFPROC glad_glRectf = NULL;
+PFNGLRECTFVPROC glad_glRectfv = NULL;
+PFNGLRECTIPROC glad_glRecti = NULL;
+PFNGLRECTIVPROC glad_glRectiv = NULL;
+PFNGLRECTSPROC glad_glRects = NULL;
+PFNGLRECTSVPROC glad_glRectsv = NULL;
+PFNGLRENDERMODEPROC glad_glRenderMode = NULL;
+PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage = NULL;
+PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample = NULL;
+PFNGLROTATEDPROC glad_glRotated = NULL;
+PFNGLROTATEFPROC glad_glRotatef = NULL;
+PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage = NULL;
+PFNGLSAMPLEMASKIPROC glad_glSampleMaski = NULL;
+PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv = NULL;
+PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv = NULL;
+PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf = NULL;
+PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv = NULL;
+PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri = NULL;
+PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv = NULL;
+PFNGLSCALEDPROC glad_glScaled = NULL;
+PFNGLSCALEFPROC glad_glScalef = NULL;
+PFNGLSCISSORPROC glad_glScissor = NULL;
+PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b = NULL;
+PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv = NULL;
+PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d = NULL;
+PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv = NULL;
+PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f = NULL;
+PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv = NULL;
+PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i = NULL;
+PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv = NULL;
+PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s = NULL;
+PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv = NULL;
+PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub = NULL;
+PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv = NULL;
+PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui = NULL;
+PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv = NULL;
+PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us = NULL;
+PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv = NULL;
+PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui = NULL;
+PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv = NULL;
+PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer = NULL;
+PFNGLSELECTBUFFERPROC glad_glSelectBuffer = NULL;
+PFNGLSHADEMODELPROC glad_glShadeModel = NULL;
+PFNGLSHADERSOURCEPROC glad_glShaderSource = NULL;
+PFNGLSTENCILFUNCPROC glad_glStencilFunc = NULL;
+PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate = NULL;
+PFNGLSTENCILMASKPROC glad_glStencilMask = NULL;
+PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate = NULL;
+PFNGLSTENCILOPPROC glad_glStencilOp = NULL;
+PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate = NULL;
+PFNGLTEXBUFFERPROC glad_glTexBuffer = NULL;
+PFNGLTEXCOORD1DPROC glad_glTexCoord1d = NULL;
+PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv = NULL;
+PFNGLTEXCOORD1FPROC glad_glTexCoord1f = NULL;
+PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv = NULL;
+PFNGLTEXCOORD1IPROC glad_glTexCoord1i = NULL;
+PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv = NULL;
+PFNGLTEXCOORD1SPROC glad_glTexCoord1s = NULL;
+PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv = NULL;
+PFNGLTEXCOORD2DPROC glad_glTexCoord2d = NULL;
+PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv = NULL;
+PFNGLTEXCOORD2FPROC glad_glTexCoord2f = NULL;
+PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv = NULL;
+PFNGLTEXCOORD2IPROC glad_glTexCoord2i = NULL;
+PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv = NULL;
+PFNGLTEXCOORD2SPROC glad_glTexCoord2s = NULL;
+PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv = NULL;
+PFNGLTEXCOORD3DPROC glad_glTexCoord3d = NULL;
+PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv = NULL;
+PFNGLTEXCOORD3FPROC glad_glTexCoord3f = NULL;
+PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv = NULL;
+PFNGLTEXCOORD3IPROC glad_glTexCoord3i = NULL;
+PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv = NULL;
+PFNGLTEXCOORD3SPROC glad_glTexCoord3s = NULL;
+PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv = NULL;
+PFNGLTEXCOORD4DPROC glad_glTexCoord4d = NULL;
+PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv = NULL;
+PFNGLTEXCOORD4FPROC glad_glTexCoord4f = NULL;
+PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv = NULL;
+PFNGLTEXCOORD4IPROC glad_glTexCoord4i = NULL;
+PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv = NULL;
+PFNGLTEXCOORD4SPROC glad_glTexCoord4s = NULL;
+PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv = NULL;
+PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui = NULL;
+PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv = NULL;
+PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui = NULL;
+PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv = NULL;
+PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui = NULL;
+PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv = NULL;
+PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui = NULL;
+PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv = NULL;
+PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer = NULL;
+PFNGLTEXENVFPROC glad_glTexEnvf = NULL;
+PFNGLTEXENVFVPROC glad_glTexEnvfv = NULL;
+PFNGLTEXENVIPROC glad_glTexEnvi = NULL;
+PFNGLTEXENVIVPROC glad_glTexEnviv = NULL;
+PFNGLTEXGENDPROC glad_glTexGend = NULL;
+PFNGLTEXGENDVPROC glad_glTexGendv = NULL;
+PFNGLTEXGENFPROC glad_glTexGenf = NULL;
+PFNGLTEXGENFVPROC glad_glTexGenfv = NULL;
+PFNGLTEXGENIPROC glad_glTexGeni = NULL;
+PFNGLTEXGENIVPROC glad_glTexGeniv = NULL;
+PFNGLTEXIMAGE1DPROC glad_glTexImage1D = NULL;
+PFNGLTEXIMAGE2DPROC glad_glTexImage2D = NULL;
+PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample = NULL;
+PFNGLTEXIMAGE3DPROC glad_glTexImage3D = NULL;
+PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample = NULL;
+PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv = NULL;
+PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv = NULL;
+PFNGLTEXPARAMETERFPROC glad_glTexParameterf = NULL;
+PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv = NULL;
+PFNGLTEXPARAMETERIPROC glad_glTexParameteri = NULL;
+PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv = NULL;
+PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D = NULL;
+PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D = NULL;
+PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D = NULL;
+PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings = NULL;
+PFNGLTRANSLATEDPROC glad_glTranslated = NULL;
+PFNGLTRANSLATEFPROC glad_glTranslatef = NULL;
+PFNGLUNIFORM1FPROC glad_glUniform1f = NULL;
+PFNGLUNIFORM1FVPROC glad_glUniform1fv = NULL;
+PFNGLUNIFORM1IPROC glad_glUniform1i = NULL;
+PFNGLUNIFORM1IVPROC glad_glUniform1iv = NULL;
+PFNGLUNIFORM1UIPROC glad_glUniform1ui = NULL;
+PFNGLUNIFORM1UIVPROC glad_glUniform1uiv = NULL;
+PFNGLUNIFORM2FPROC glad_glUniform2f = NULL;
+PFNGLUNIFORM2FVPROC glad_glUniform2fv = NULL;
+PFNGLUNIFORM2IPROC glad_glUniform2i = NULL;
+PFNGLUNIFORM2IVPROC glad_glUniform2iv = NULL;
+PFNGLUNIFORM2UIPROC glad_glUniform2ui = NULL;
+PFNGLUNIFORM2UIVPROC glad_glUniform2uiv = NULL;
+PFNGLUNIFORM3FPROC glad_glUniform3f = NULL;
+PFNGLUNIFORM3FVPROC glad_glUniform3fv = NULL;
+PFNGLUNIFORM3IPROC glad_glUniform3i = NULL;
+PFNGLUNIFORM3IVPROC glad_glUniform3iv = NULL;
+PFNGLUNIFORM3UIPROC glad_glUniform3ui = NULL;
+PFNGLUNIFORM3UIVPROC glad_glUniform3uiv = NULL;
+PFNGLUNIFORM4FPROC glad_glUniform4f = NULL;
+PFNGLUNIFORM4FVPROC glad_glUniform4fv = NULL;
+PFNGLUNIFORM4IPROC glad_glUniform4i = NULL;
+PFNGLUNIFORM4IVPROC glad_glUniform4iv = NULL;
+PFNGLUNIFORM4UIPROC glad_glUniform4ui = NULL;
+PFNGLUNIFORM4UIVPROC glad_glUniform4uiv = NULL;
+PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding = NULL;
+PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv = NULL;
+PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv = NULL;
+PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv = NULL;
+PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv = NULL;
+PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv = NULL;
+PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv = NULL;
+PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv = NULL;
+PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv = NULL;
+PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv = NULL;
+PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer = NULL;
+PFNGLUSEPROGRAMPROC glad_glUseProgram = NULL;
+PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram = NULL;
+PFNGLVERTEX2DPROC glad_glVertex2d = NULL;
+PFNGLVERTEX2DVPROC glad_glVertex2dv = NULL;
+PFNGLVERTEX2FPROC glad_glVertex2f = NULL;
+PFNGLVERTEX2FVPROC glad_glVertex2fv = NULL;
+PFNGLVERTEX2IPROC glad_glVertex2i = NULL;
+PFNGLVERTEX2IVPROC glad_glVertex2iv = NULL;
+PFNGLVERTEX2SPROC glad_glVertex2s = NULL;
+PFNGLVERTEX2SVPROC glad_glVertex2sv = NULL;
+PFNGLVERTEX3DPROC glad_glVertex3d = NULL;
+PFNGLVERTEX3DVPROC glad_glVertex3dv = NULL;
+PFNGLVERTEX3FPROC glad_glVertex3f = NULL;
+PFNGLVERTEX3FVPROC glad_glVertex3fv = NULL;
+PFNGLVERTEX3IPROC glad_glVertex3i = NULL;
+PFNGLVERTEX3IVPROC glad_glVertex3iv = NULL;
+PFNGLVERTEX3SPROC glad_glVertex3s = NULL;
+PFNGLVERTEX3SVPROC glad_glVertex3sv = NULL;
+PFNGLVERTEX4DPROC glad_glVertex4d = NULL;
+PFNGLVERTEX4DVPROC glad_glVertex4dv = NULL;
+PFNGLVERTEX4FPROC glad_glVertex4f = NULL;
+PFNGLVERTEX4FVPROC glad_glVertex4fv = NULL;
+PFNGLVERTEX4IPROC glad_glVertex4i = NULL;
+PFNGLVERTEX4IVPROC glad_glVertex4iv = NULL;
+PFNGLVERTEX4SPROC glad_glVertex4s = NULL;
+PFNGLVERTEX4SVPROC glad_glVertex4sv = NULL;
+PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d = NULL;
+PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv = NULL;
+PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f = NULL;
+PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv = NULL;
+PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s = NULL;
+PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv = NULL;
+PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d = NULL;
+PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv = NULL;
+PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f = NULL;
+PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv = NULL;
+PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s = NULL;
+PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv = NULL;
+PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d = NULL;
+PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv = NULL;
+PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f = NULL;
+PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv = NULL;
+PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s = NULL;
+PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv = NULL;
+PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv = NULL;
+PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv = NULL;
+PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv = NULL;
+PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub = NULL;
+PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv = NULL;
+PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv = NULL;
+PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv = NULL;
+PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv = NULL;
+PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d = NULL;
+PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv = NULL;
+PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f = NULL;
+PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv = NULL;
+PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv = NULL;
+PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s = NULL;
+PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv = NULL;
+PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv = NULL;
+PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv = NULL;
+PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv = NULL;
+PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor = NULL;
+PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i = NULL;
+PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv = NULL;
+PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui = NULL;
+PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv = NULL;
+PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i = NULL;
+PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv = NULL;
+PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui = NULL;
+PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv = NULL;
+PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i = NULL;
+PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv = NULL;
+PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui = NULL;
+PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv = NULL;
+PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv = NULL;
+PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i = NULL;
+PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv = NULL;
+PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv = NULL;
+PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv = NULL;
+PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui = NULL;
+PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv = NULL;
+PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv = NULL;
+PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer = NULL;
+PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui = NULL;
+PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv = NULL;
+PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui = NULL;
+PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv = NULL;
+PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui = NULL;
+PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv = NULL;
+PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui = NULL;
+PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv = NULL;
+PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer = NULL;
+PFNGLVERTEXP2UIPROC glad_glVertexP2ui = NULL;
+PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv = NULL;
+PFNGLVERTEXP3UIPROC glad_glVertexP3ui = NULL;
+PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv = NULL;
+PFNGLVERTEXP4UIPROC glad_glVertexP4ui = NULL;
+PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv = NULL;
+PFNGLVERTEXPOINTERPROC glad_glVertexPointer = NULL;
+PFNGLVIEWPORTPROC glad_glViewport = NULL;
+PFNGLWAITSYNCPROC glad_glWaitSync = NULL;
+PFNGLWINDOWPOS2DPROC glad_glWindowPos2d = NULL;
+PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv = NULL;
+PFNGLWINDOWPOS2FPROC glad_glWindowPos2f = NULL;
+PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv = NULL;
+PFNGLWINDOWPOS2IPROC glad_glWindowPos2i = NULL;
+PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv = NULL;
+PFNGLWINDOWPOS2SPROC glad_glWindowPos2s = NULL;
+PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv = NULL;
+PFNGLWINDOWPOS3DPROC glad_glWindowPos3d = NULL;
+PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv = NULL;
+PFNGLWINDOWPOS3FPROC glad_glWindowPos3f = NULL;
+PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv = NULL;
+PFNGLWINDOWPOS3IPROC glad_glWindowPos3i = NULL;
+PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv = NULL;
+PFNGLWINDOWPOS3SPROC glad_glWindowPos3s = NULL;
+PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv = NULL;
+
+
+static void glad_gl_load_GL_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_0) return;
+    glad_glAccum = (PFNGLACCUMPROC) load(userptr, "glAccum");
+    glad_glAlphaFunc = (PFNGLALPHAFUNCPROC) load(userptr, "glAlphaFunc");
+    glad_glBegin = (PFNGLBEGINPROC) load(userptr, "glBegin");
+    glad_glBitmap = (PFNGLBITMAPPROC) load(userptr, "glBitmap");
+    glad_glBlendFunc = (PFNGLBLENDFUNCPROC) load(userptr, "glBlendFunc");
+    glad_glCallList = (PFNGLCALLLISTPROC) load(userptr, "glCallList");
+    glad_glCallLists = (PFNGLCALLLISTSPROC) load(userptr, "glCallLists");
+    glad_glClear = (PFNGLCLEARPROC) load(userptr, "glClear");
+    glad_glClearAccum = (PFNGLCLEARACCUMPROC) load(userptr, "glClearAccum");
+    glad_glClearColor = (PFNGLCLEARCOLORPROC) load(userptr, "glClearColor");
+    glad_glClearDepth = (PFNGLCLEARDEPTHPROC) load(userptr, "glClearDepth");
+    glad_glClearIndex = (PFNGLCLEARINDEXPROC) load(userptr, "glClearIndex");
+    glad_glClearStencil = (PFNGLCLEARSTENCILPROC) load(userptr, "glClearStencil");
+    glad_glClipPlane = (PFNGLCLIPPLANEPROC) load(userptr, "glClipPlane");
+    glad_glColor3b = (PFNGLCOLOR3BPROC) load(userptr, "glColor3b");
+    glad_glColor3bv = (PFNGLCOLOR3BVPROC) load(userptr, "glColor3bv");
+    glad_glColor3d = (PFNGLCOLOR3DPROC) load(userptr, "glColor3d");
+    glad_glColor3dv = (PFNGLCOLOR3DVPROC) load(userptr, "glColor3dv");
+    glad_glColor3f = (PFNGLCOLOR3FPROC) load(userptr, "glColor3f");
+    glad_glColor3fv = (PFNGLCOLOR3FVPROC) load(userptr, "glColor3fv");
+    glad_glColor3i = (PFNGLCOLOR3IPROC) load(userptr, "glColor3i");
+    glad_glColor3iv = (PFNGLCOLOR3IVPROC) load(userptr, "glColor3iv");
+    glad_glColor3s = (PFNGLCOLOR3SPROC) load(userptr, "glColor3s");
+    glad_glColor3sv = (PFNGLCOLOR3SVPROC) load(userptr, "glColor3sv");
+    glad_glColor3ub = (PFNGLCOLOR3UBPROC) load(userptr, "glColor3ub");
+    glad_glColor3ubv = (PFNGLCOLOR3UBVPROC) load(userptr, "glColor3ubv");
+    glad_glColor3ui = (PFNGLCOLOR3UIPROC) load(userptr, "glColor3ui");
+    glad_glColor3uiv = (PFNGLCOLOR3UIVPROC) load(userptr, "glColor3uiv");
+    glad_glColor3us = (PFNGLCOLOR3USPROC) load(userptr, "glColor3us");
+    glad_glColor3usv = (PFNGLCOLOR3USVPROC) load(userptr, "glColor3usv");
+    glad_glColor4b = (PFNGLCOLOR4BPROC) load(userptr, "glColor4b");
+    glad_glColor4bv = (PFNGLCOLOR4BVPROC) load(userptr, "glColor4bv");
+    glad_glColor4d = (PFNGLCOLOR4DPROC) load(userptr, "glColor4d");
+    glad_glColor4dv = (PFNGLCOLOR4DVPROC) load(userptr, "glColor4dv");
+    glad_glColor4f = (PFNGLCOLOR4FPROC) load(userptr, "glColor4f");
+    glad_glColor4fv = (PFNGLCOLOR4FVPROC) load(userptr, "glColor4fv");
+    glad_glColor4i = (PFNGLCOLOR4IPROC) load(userptr, "glColor4i");
+    glad_glColor4iv = (PFNGLCOLOR4IVPROC) load(userptr, "glColor4iv");
+    glad_glColor4s = (PFNGLCOLOR4SPROC) load(userptr, "glColor4s");
+    glad_glColor4sv = (PFNGLCOLOR4SVPROC) load(userptr, "glColor4sv");
+    glad_glColor4ub = (PFNGLCOLOR4UBPROC) load(userptr, "glColor4ub");
+    glad_glColor4ubv = (PFNGLCOLOR4UBVPROC) load(userptr, "glColor4ubv");
+    glad_glColor4ui = (PFNGLCOLOR4UIPROC) load(userptr, "glColor4ui");
+    glad_glColor4uiv = (PFNGLCOLOR4UIVPROC) load(userptr, "glColor4uiv");
+    glad_glColor4us = (PFNGLCOLOR4USPROC) load(userptr, "glColor4us");
+    glad_glColor4usv = (PFNGLCOLOR4USVPROC) load(userptr, "glColor4usv");
+    glad_glColorMask = (PFNGLCOLORMASKPROC) load(userptr, "glColorMask");
+    glad_glColorMaterial = (PFNGLCOLORMATERIALPROC) load(userptr, "glColorMaterial");
+    glad_glCopyPixels = (PFNGLCOPYPIXELSPROC) load(userptr, "glCopyPixels");
+    glad_glCullFace = (PFNGLCULLFACEPROC) load(userptr, "glCullFace");
+    glad_glDeleteLists = (PFNGLDELETELISTSPROC) load(userptr, "glDeleteLists");
+    glad_glDepthFunc = (PFNGLDEPTHFUNCPROC) load(userptr, "glDepthFunc");
+    glad_glDepthMask = (PFNGLDEPTHMASKPROC) load(userptr, "glDepthMask");
+    glad_glDepthRange = (PFNGLDEPTHRANGEPROC) load(userptr, "glDepthRange");
+    glad_glDisable = (PFNGLDISABLEPROC) load(userptr, "glDisable");
+    glad_glDrawBuffer = (PFNGLDRAWBUFFERPROC) load(userptr, "glDrawBuffer");
+    glad_glDrawPixels = (PFNGLDRAWPIXELSPROC) load(userptr, "glDrawPixels");
+    glad_glEdgeFlag = (PFNGLEDGEFLAGPROC) load(userptr, "glEdgeFlag");
+    glad_glEdgeFlagv = (PFNGLEDGEFLAGVPROC) load(userptr, "glEdgeFlagv");
+    glad_glEnable = (PFNGLENABLEPROC) load(userptr, "glEnable");
+    glad_glEnd = (PFNGLENDPROC) load(userptr, "glEnd");
+    glad_glEndList = (PFNGLENDLISTPROC) load(userptr, "glEndList");
+    glad_glEvalCoord1d = (PFNGLEVALCOORD1DPROC) load(userptr, "glEvalCoord1d");
+    glad_glEvalCoord1dv = (PFNGLEVALCOORD1DVPROC) load(userptr, "glEvalCoord1dv");
+    glad_glEvalCoord1f = (PFNGLEVALCOORD1FPROC) load(userptr, "glEvalCoord1f");
+    glad_glEvalCoord1fv = (PFNGLEVALCOORD1FVPROC) load(userptr, "glEvalCoord1fv");
+    glad_glEvalCoord2d = (PFNGLEVALCOORD2DPROC) load(userptr, "glEvalCoord2d");
+    glad_glEvalCoord2dv = (PFNGLEVALCOORD2DVPROC) load(userptr, "glEvalCoord2dv");
+    glad_glEvalCoord2f = (PFNGLEVALCOORD2FPROC) load(userptr, "glEvalCoord2f");
+    glad_glEvalCoord2fv = (PFNGLEVALCOORD2FVPROC) load(userptr, "glEvalCoord2fv");
+    glad_glEvalMesh1 = (PFNGLEVALMESH1PROC) load(userptr, "glEvalMesh1");
+    glad_glEvalMesh2 = (PFNGLEVALMESH2PROC) load(userptr, "glEvalMesh2");
+    glad_glEvalPoint1 = (PFNGLEVALPOINT1PROC) load(userptr, "glEvalPoint1");
+    glad_glEvalPoint2 = (PFNGLEVALPOINT2PROC) load(userptr, "glEvalPoint2");
+    glad_glFeedbackBuffer = (PFNGLFEEDBACKBUFFERPROC) load(userptr, "glFeedbackBuffer");
+    glad_glFinish = (PFNGLFINISHPROC) load(userptr, "glFinish");
+    glad_glFlush = (PFNGLFLUSHPROC) load(userptr, "glFlush");
+    glad_glFogf = (PFNGLFOGFPROC) load(userptr, "glFogf");
+    glad_glFogfv = (PFNGLFOGFVPROC) load(userptr, "glFogfv");
+    glad_glFogi = (PFNGLFOGIPROC) load(userptr, "glFogi");
+    glad_glFogiv = (PFNGLFOGIVPROC) load(userptr, "glFogiv");
+    glad_glFrontFace = (PFNGLFRONTFACEPROC) load(userptr, "glFrontFace");
+    glad_glFrustum = (PFNGLFRUSTUMPROC) load(userptr, "glFrustum");
+    glad_glGenLists = (PFNGLGENLISTSPROC) load(userptr, "glGenLists");
+    glad_glGetBooleanv = (PFNGLGETBOOLEANVPROC) load(userptr, "glGetBooleanv");
+    glad_glGetClipPlane = (PFNGLGETCLIPPLANEPROC) load(userptr, "glGetClipPlane");
+    glad_glGetDoublev = (PFNGLGETDOUBLEVPROC) load(userptr, "glGetDoublev");
+    glad_glGetError = (PFNGLGETERRORPROC) load(userptr, "glGetError");
+    glad_glGetFloatv = (PFNGLGETFLOATVPROC) load(userptr, "glGetFloatv");
+    glad_glGetIntegerv = (PFNGLGETINTEGERVPROC) load(userptr, "glGetIntegerv");
+    glad_glGetLightfv = (PFNGLGETLIGHTFVPROC) load(userptr, "glGetLightfv");
+    glad_glGetLightiv = (PFNGLGETLIGHTIVPROC) load(userptr, "glGetLightiv");
+    glad_glGetMapdv = (PFNGLGETMAPDVPROC) load(userptr, "glGetMapdv");
+    glad_glGetMapfv = (PFNGLGETMAPFVPROC) load(userptr, "glGetMapfv");
+    glad_glGetMapiv = (PFNGLGETMAPIVPROC) load(userptr, "glGetMapiv");
+    glad_glGetMaterialfv = (PFNGLGETMATERIALFVPROC) load(userptr, "glGetMaterialfv");
+    glad_glGetMaterialiv = (PFNGLGETMATERIALIVPROC) load(userptr, "glGetMaterialiv");
+    glad_glGetPixelMapfv = (PFNGLGETPIXELMAPFVPROC) load(userptr, "glGetPixelMapfv");
+    glad_glGetPixelMapuiv = (PFNGLGETPIXELMAPUIVPROC) load(userptr, "glGetPixelMapuiv");
+    glad_glGetPixelMapusv = (PFNGLGETPIXELMAPUSVPROC) load(userptr, "glGetPixelMapusv");
+    glad_glGetPolygonStipple = (PFNGLGETPOLYGONSTIPPLEPROC) load(userptr, "glGetPolygonStipple");
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    glad_glGetTexEnvfv = (PFNGLGETTEXENVFVPROC) load(userptr, "glGetTexEnvfv");
+    glad_glGetTexEnviv = (PFNGLGETTEXENVIVPROC) load(userptr, "glGetTexEnviv");
+    glad_glGetTexGendv = (PFNGLGETTEXGENDVPROC) load(userptr, "glGetTexGendv");
+    glad_glGetTexGenfv = (PFNGLGETTEXGENFVPROC) load(userptr, "glGetTexGenfv");
+    glad_glGetTexGeniv = (PFNGLGETTEXGENIVPROC) load(userptr, "glGetTexGeniv");
+    glad_glGetTexImage = (PFNGLGETTEXIMAGEPROC) load(userptr, "glGetTexImage");
+    glad_glGetTexLevelParameterfv = (PFNGLGETTEXLEVELPARAMETERFVPROC) load(userptr, "glGetTexLevelParameterfv");
+    glad_glGetTexLevelParameteriv = (PFNGLGETTEXLEVELPARAMETERIVPROC) load(userptr, "glGetTexLevelParameteriv");
+    glad_glGetTexParameterfv = (PFNGLGETTEXPARAMETERFVPROC) load(userptr, "glGetTexParameterfv");
+    glad_glGetTexParameteriv = (PFNGLGETTEXPARAMETERIVPROC) load(userptr, "glGetTexParameteriv");
+    glad_glHint = (PFNGLHINTPROC) load(userptr, "glHint");
+    glad_glIndexMask = (PFNGLINDEXMASKPROC) load(userptr, "glIndexMask");
+    glad_glIndexd = (PFNGLINDEXDPROC) load(userptr, "glIndexd");
+    glad_glIndexdv = (PFNGLINDEXDVPROC) load(userptr, "glIndexdv");
+    glad_glIndexf = (PFNGLINDEXFPROC) load(userptr, "glIndexf");
+    glad_glIndexfv = (PFNGLINDEXFVPROC) load(userptr, "glIndexfv");
+    glad_glIndexi = (PFNGLINDEXIPROC) load(userptr, "glIndexi");
+    glad_glIndexiv = (PFNGLINDEXIVPROC) load(userptr, "glIndexiv");
+    glad_glIndexs = (PFNGLINDEXSPROC) load(userptr, "glIndexs");
+    glad_glIndexsv = (PFNGLINDEXSVPROC) load(userptr, "glIndexsv");
+    glad_glInitNames = (PFNGLINITNAMESPROC) load(userptr, "glInitNames");
+    glad_glIsEnabled = (PFNGLISENABLEDPROC) load(userptr, "glIsEnabled");
+    glad_glIsList = (PFNGLISLISTPROC) load(userptr, "glIsList");
+    glad_glLightModelf = (PFNGLLIGHTMODELFPROC) load(userptr, "glLightModelf");
+    glad_glLightModelfv = (PFNGLLIGHTMODELFVPROC) load(userptr, "glLightModelfv");
+    glad_glLightModeli = (PFNGLLIGHTMODELIPROC) load(userptr, "glLightModeli");
+    glad_glLightModeliv = (PFNGLLIGHTMODELIVPROC) load(userptr, "glLightModeliv");
+    glad_glLightf = (PFNGLLIGHTFPROC) load(userptr, "glLightf");
+    glad_glLightfv = (PFNGLLIGHTFVPROC) load(userptr, "glLightfv");
+    glad_glLighti = (PFNGLLIGHTIPROC) load(userptr, "glLighti");
+    glad_glLightiv = (PFNGLLIGHTIVPROC) load(userptr, "glLightiv");
+    glad_glLineStipple = (PFNGLLINESTIPPLEPROC) load(userptr, "glLineStipple");
+    glad_glLineWidth = (PFNGLLINEWIDTHPROC) load(userptr, "glLineWidth");
+    glad_glListBase = (PFNGLLISTBASEPROC) load(userptr, "glListBase");
+    glad_glLoadIdentity = (PFNGLLOADIDENTITYPROC) load(userptr, "glLoadIdentity");
+    glad_glLoadMatrixd = (PFNGLLOADMATRIXDPROC) load(userptr, "glLoadMatrixd");
+    glad_glLoadMatrixf = (PFNGLLOADMATRIXFPROC) load(userptr, "glLoadMatrixf");
+    glad_glLoadName = (PFNGLLOADNAMEPROC) load(userptr, "glLoadName");
+    glad_glLogicOp = (PFNGLLOGICOPPROC) load(userptr, "glLogicOp");
+    glad_glMap1d = (PFNGLMAP1DPROC) load(userptr, "glMap1d");
+    glad_glMap1f = (PFNGLMAP1FPROC) load(userptr, "glMap1f");
+    glad_glMap2d = (PFNGLMAP2DPROC) load(userptr, "glMap2d");
+    glad_glMap2f = (PFNGLMAP2FPROC) load(userptr, "glMap2f");
+    glad_glMapGrid1d = (PFNGLMAPGRID1DPROC) load(userptr, "glMapGrid1d");
+    glad_glMapGrid1f = (PFNGLMAPGRID1FPROC) load(userptr, "glMapGrid1f");
+    glad_glMapGrid2d = (PFNGLMAPGRID2DPROC) load(userptr, "glMapGrid2d");
+    glad_glMapGrid2f = (PFNGLMAPGRID2FPROC) load(userptr, "glMapGrid2f");
+    glad_glMaterialf = (PFNGLMATERIALFPROC) load(userptr, "glMaterialf");
+    glad_glMaterialfv = (PFNGLMATERIALFVPROC) load(userptr, "glMaterialfv");
+    glad_glMateriali = (PFNGLMATERIALIPROC) load(userptr, "glMateriali");
+    glad_glMaterialiv = (PFNGLMATERIALIVPROC) load(userptr, "glMaterialiv");
+    glad_glMatrixMode = (PFNGLMATRIXMODEPROC) load(userptr, "glMatrixMode");
+    glad_glMultMatrixd = (PFNGLMULTMATRIXDPROC) load(userptr, "glMultMatrixd");
+    glad_glMultMatrixf = (PFNGLMULTMATRIXFPROC) load(userptr, "glMultMatrixf");
+    glad_glNewList = (PFNGLNEWLISTPROC) load(userptr, "glNewList");
+    glad_glNormal3b = (PFNGLNORMAL3BPROC) load(userptr, "glNormal3b");
+    glad_glNormal3bv = (PFNGLNORMAL3BVPROC) load(userptr, "glNormal3bv");
+    glad_glNormal3d = (PFNGLNORMAL3DPROC) load(userptr, "glNormal3d");
+    glad_glNormal3dv = (PFNGLNORMAL3DVPROC) load(userptr, "glNormal3dv");
+    glad_glNormal3f = (PFNGLNORMAL3FPROC) load(userptr, "glNormal3f");
+    glad_glNormal3fv = (PFNGLNORMAL3FVPROC) load(userptr, "glNormal3fv");
+    glad_glNormal3i = (PFNGLNORMAL3IPROC) load(userptr, "glNormal3i");
+    glad_glNormal3iv = (PFNGLNORMAL3IVPROC) load(userptr, "glNormal3iv");
+    glad_glNormal3s = (PFNGLNORMAL3SPROC) load(userptr, "glNormal3s");
+    glad_glNormal3sv = (PFNGLNORMAL3SVPROC) load(userptr, "glNormal3sv");
+    glad_glOrtho = (PFNGLORTHOPROC) load(userptr, "glOrtho");
+    glad_glPassThrough = (PFNGLPASSTHROUGHPROC) load(userptr, "glPassThrough");
+    glad_glPixelMapfv = (PFNGLPIXELMAPFVPROC) load(userptr, "glPixelMapfv");
+    glad_glPixelMapuiv = (PFNGLPIXELMAPUIVPROC) load(userptr, "glPixelMapuiv");
+    glad_glPixelMapusv = (PFNGLPIXELMAPUSVPROC) load(userptr, "glPixelMapusv");
+    glad_glPixelStoref = (PFNGLPIXELSTOREFPROC) load(userptr, "glPixelStoref");
+    glad_glPixelStorei = (PFNGLPIXELSTOREIPROC) load(userptr, "glPixelStorei");
+    glad_glPixelTransferf = (PFNGLPIXELTRANSFERFPROC) load(userptr, "glPixelTransferf");
+    glad_glPixelTransferi = (PFNGLPIXELTRANSFERIPROC) load(userptr, "glPixelTransferi");
+    glad_glPixelZoom = (PFNGLPIXELZOOMPROC) load(userptr, "glPixelZoom");
+    glad_glPointSize = (PFNGLPOINTSIZEPROC) load(userptr, "glPointSize");
+    glad_glPolygonMode = (PFNGLPOLYGONMODEPROC) load(userptr, "glPolygonMode");
+    glad_glPolygonStipple = (PFNGLPOLYGONSTIPPLEPROC) load(userptr, "glPolygonStipple");
+    glad_glPopAttrib = (PFNGLPOPATTRIBPROC) load(userptr, "glPopAttrib");
+    glad_glPopMatrix = (PFNGLPOPMATRIXPROC) load(userptr, "glPopMatrix");
+    glad_glPopName = (PFNGLPOPNAMEPROC) load(userptr, "glPopName");
+    glad_glPushAttrib = (PFNGLPUSHATTRIBPROC) load(userptr, "glPushAttrib");
+    glad_glPushMatrix = (PFNGLPUSHMATRIXPROC) load(userptr, "glPushMatrix");
+    glad_glPushName = (PFNGLPUSHNAMEPROC) load(userptr, "glPushName");
+    glad_glRasterPos2d = (PFNGLRASTERPOS2DPROC) load(userptr, "glRasterPos2d");
+    glad_glRasterPos2dv = (PFNGLRASTERPOS2DVPROC) load(userptr, "glRasterPos2dv");
+    glad_glRasterPos2f = (PFNGLRASTERPOS2FPROC) load(userptr, "glRasterPos2f");
+    glad_glRasterPos2fv = (PFNGLRASTERPOS2FVPROC) load(userptr, "glRasterPos2fv");
+    glad_glRasterPos2i = (PFNGLRASTERPOS2IPROC) load(userptr, "glRasterPos2i");
+    glad_glRasterPos2iv = (PFNGLRASTERPOS2IVPROC) load(userptr, "glRasterPos2iv");
+    glad_glRasterPos2s = (PFNGLRASTERPOS2SPROC) load(userptr, "glRasterPos2s");
+    glad_glRasterPos2sv = (PFNGLRASTERPOS2SVPROC) load(userptr, "glRasterPos2sv");
+    glad_glRasterPos3d = (PFNGLRASTERPOS3DPROC) load(userptr, "glRasterPos3d");
+    glad_glRasterPos3dv = (PFNGLRASTERPOS3DVPROC) load(userptr, "glRasterPos3dv");
+    glad_glRasterPos3f = (PFNGLRASTERPOS3FPROC) load(userptr, "glRasterPos3f");
+    glad_glRasterPos3fv = (PFNGLRASTERPOS3FVPROC) load(userptr, "glRasterPos3fv");
+    glad_glRasterPos3i = (PFNGLRASTERPOS3IPROC) load(userptr, "glRasterPos3i");
+    glad_glRasterPos3iv = (PFNGLRASTERPOS3IVPROC) load(userptr, "glRasterPos3iv");
+    glad_glRasterPos3s = (PFNGLRASTERPOS3SPROC) load(userptr, "glRasterPos3s");
+    glad_glRasterPos3sv = (PFNGLRASTERPOS3SVPROC) load(userptr, "glRasterPos3sv");
+    glad_glRasterPos4d = (PFNGLRASTERPOS4DPROC) load(userptr, "glRasterPos4d");
+    glad_glRasterPos4dv = (PFNGLRASTERPOS4DVPROC) load(userptr, "glRasterPos4dv");
+    glad_glRasterPos4f = (PFNGLRASTERPOS4FPROC) load(userptr, "glRasterPos4f");
+    glad_glRasterPos4fv = (PFNGLRASTERPOS4FVPROC) load(userptr, "glRasterPos4fv");
+    glad_glRasterPos4i = (PFNGLRASTERPOS4IPROC) load(userptr, "glRasterPos4i");
+    glad_glRasterPos4iv = (PFNGLRASTERPOS4IVPROC) load(userptr, "glRasterPos4iv");
+    glad_glRasterPos4s = (PFNGLRASTERPOS4SPROC) load(userptr, "glRasterPos4s");
+    glad_glRasterPos4sv = (PFNGLRASTERPOS4SVPROC) load(userptr, "glRasterPos4sv");
+    glad_glReadBuffer = (PFNGLREADBUFFERPROC) load(userptr, "glReadBuffer");
+    glad_glReadPixels = (PFNGLREADPIXELSPROC) load(userptr, "glReadPixels");
+    glad_glRectd = (PFNGLRECTDPROC) load(userptr, "glRectd");
+    glad_glRectdv = (PFNGLRECTDVPROC) load(userptr, "glRectdv");
+    glad_glRectf = (PFNGLRECTFPROC) load(userptr, "glRectf");
+    glad_glRectfv = (PFNGLRECTFVPROC) load(userptr, "glRectfv");
+    glad_glRecti = (PFNGLRECTIPROC) load(userptr, "glRecti");
+    glad_glRectiv = (PFNGLRECTIVPROC) load(userptr, "glRectiv");
+    glad_glRects = (PFNGLRECTSPROC) load(userptr, "glRects");
+    glad_glRectsv = (PFNGLRECTSVPROC) load(userptr, "glRectsv");
+    glad_glRenderMode = (PFNGLRENDERMODEPROC) load(userptr, "glRenderMode");
+    glad_glRotated = (PFNGLROTATEDPROC) load(userptr, "glRotated");
+    glad_glRotatef = (PFNGLROTATEFPROC) load(userptr, "glRotatef");
+    glad_glScaled = (PFNGLSCALEDPROC) load(userptr, "glScaled");
+    glad_glScalef = (PFNGLSCALEFPROC) load(userptr, "glScalef");
+    glad_glScissor = (PFNGLSCISSORPROC) load(userptr, "glScissor");
+    glad_glSelectBuffer = (PFNGLSELECTBUFFERPROC) load(userptr, "glSelectBuffer");
+    glad_glShadeModel = (PFNGLSHADEMODELPROC) load(userptr, "glShadeModel");
+    glad_glStencilFunc = (PFNGLSTENCILFUNCPROC) load(userptr, "glStencilFunc");
+    glad_glStencilMask = (PFNGLSTENCILMASKPROC) load(userptr, "glStencilMask");
+    glad_glStencilOp = (PFNGLSTENCILOPPROC) load(userptr, "glStencilOp");
+    glad_glTexCoord1d = (PFNGLTEXCOORD1DPROC) load(userptr, "glTexCoord1d");
+    glad_glTexCoord1dv = (PFNGLTEXCOORD1DVPROC) load(userptr, "glTexCoord1dv");
+    glad_glTexCoord1f = (PFNGLTEXCOORD1FPROC) load(userptr, "glTexCoord1f");
+    glad_glTexCoord1fv = (PFNGLTEXCOORD1FVPROC) load(userptr, "glTexCoord1fv");
+    glad_glTexCoord1i = (PFNGLTEXCOORD1IPROC) load(userptr, "glTexCoord1i");
+    glad_glTexCoord1iv = (PFNGLTEXCOORD1IVPROC) load(userptr, "glTexCoord1iv");
+    glad_glTexCoord1s = (PFNGLTEXCOORD1SPROC) load(userptr, "glTexCoord1s");
+    glad_glTexCoord1sv = (PFNGLTEXCOORD1SVPROC) load(userptr, "glTexCoord1sv");
+    glad_glTexCoord2d = (PFNGLTEXCOORD2DPROC) load(userptr, "glTexCoord2d");
+    glad_glTexCoord2dv = (PFNGLTEXCOORD2DVPROC) load(userptr, "glTexCoord2dv");
+    glad_glTexCoord2f = (PFNGLTEXCOORD2FPROC) load(userptr, "glTexCoord2f");
+    glad_glTexCoord2fv = (PFNGLTEXCOORD2FVPROC) load(userptr, "glTexCoord2fv");
+    glad_glTexCoord2i = (PFNGLTEXCOORD2IPROC) load(userptr, "glTexCoord2i");
+    glad_glTexCoord2iv = (PFNGLTEXCOORD2IVPROC) load(userptr, "glTexCoord2iv");
+    glad_glTexCoord2s = (PFNGLTEXCOORD2SPROC) load(userptr, "glTexCoord2s");
+    glad_glTexCoord2sv = (PFNGLTEXCOORD2SVPROC) load(userptr, "glTexCoord2sv");
+    glad_glTexCoord3d = (PFNGLTEXCOORD3DPROC) load(userptr, "glTexCoord3d");
+    glad_glTexCoord3dv = (PFNGLTEXCOORD3DVPROC) load(userptr, "glTexCoord3dv");
+    glad_glTexCoord3f = (PFNGLTEXCOORD3FPROC) load(userptr, "glTexCoord3f");
+    glad_glTexCoord3fv = (PFNGLTEXCOORD3FVPROC) load(userptr, "glTexCoord3fv");
+    glad_glTexCoord3i = (PFNGLTEXCOORD3IPROC) load(userptr, "glTexCoord3i");
+    glad_glTexCoord3iv = (PFNGLTEXCOORD3IVPROC) load(userptr, "glTexCoord3iv");
+    glad_glTexCoord3s = (PFNGLTEXCOORD3SPROC) load(userptr, "glTexCoord3s");
+    glad_glTexCoord3sv = (PFNGLTEXCOORD3SVPROC) load(userptr, "glTexCoord3sv");
+    glad_glTexCoord4d = (PFNGLTEXCOORD4DPROC) load(userptr, "glTexCoord4d");
+    glad_glTexCoord4dv = (PFNGLTEXCOORD4DVPROC) load(userptr, "glTexCoord4dv");
+    glad_glTexCoord4f = (PFNGLTEXCOORD4FPROC) load(userptr, "glTexCoord4f");
+    glad_glTexCoord4fv = (PFNGLTEXCOORD4FVPROC) load(userptr, "glTexCoord4fv");
+    glad_glTexCoord4i = (PFNGLTEXCOORD4IPROC) load(userptr, "glTexCoord4i");
+    glad_glTexCoord4iv = (PFNGLTEXCOORD4IVPROC) load(userptr, "glTexCoord4iv");
+    glad_glTexCoord4s = (PFNGLTEXCOORD4SPROC) load(userptr, "glTexCoord4s");
+    glad_glTexCoord4sv = (PFNGLTEXCOORD4SVPROC) load(userptr, "glTexCoord4sv");
+    glad_glTexEnvf = (PFNGLTEXENVFPROC) load(userptr, "glTexEnvf");
+    glad_glTexEnvfv = (PFNGLTEXENVFVPROC) load(userptr, "glTexEnvfv");
+    glad_glTexEnvi = (PFNGLTEXENVIPROC) load(userptr, "glTexEnvi");
+    glad_glTexEnviv = (PFNGLTEXENVIVPROC) load(userptr, "glTexEnviv");
+    glad_glTexGend = (PFNGLTEXGENDPROC) load(userptr, "glTexGend");
+    glad_glTexGendv = (PFNGLTEXGENDVPROC) load(userptr, "glTexGendv");
+    glad_glTexGenf = (PFNGLTEXGENFPROC) load(userptr, "glTexGenf");
+    glad_glTexGenfv = (PFNGLTEXGENFVPROC) load(userptr, "glTexGenfv");
+    glad_glTexGeni = (PFNGLTEXGENIPROC) load(userptr, "glTexGeni");
+    glad_glTexGeniv = (PFNGLTEXGENIVPROC) load(userptr, "glTexGeniv");
+    glad_glTexImage1D = (PFNGLTEXIMAGE1DPROC) load(userptr, "glTexImage1D");
+    glad_glTexImage2D = (PFNGLTEXIMAGE2DPROC) load(userptr, "glTexImage2D");
+    glad_glTexParameterf = (PFNGLTEXPARAMETERFPROC) load(userptr, "glTexParameterf");
+    glad_glTexParameterfv = (PFNGLTEXPARAMETERFVPROC) load(userptr, "glTexParameterfv");
+    glad_glTexParameteri = (PFNGLTEXPARAMETERIPROC) load(userptr, "glTexParameteri");
+    glad_glTexParameteriv = (PFNGLTEXPARAMETERIVPROC) load(userptr, "glTexParameteriv");
+    glad_glTranslated = (PFNGLTRANSLATEDPROC) load(userptr, "glTranslated");
+    glad_glTranslatef = (PFNGLTRANSLATEFPROC) load(userptr, "glTranslatef");
+    glad_glVertex2d = (PFNGLVERTEX2DPROC) load(userptr, "glVertex2d");
+    glad_glVertex2dv = (PFNGLVERTEX2DVPROC) load(userptr, "glVertex2dv");
+    glad_glVertex2f = (PFNGLVERTEX2FPROC) load(userptr, "glVertex2f");
+    glad_glVertex2fv = (PFNGLVERTEX2FVPROC) load(userptr, "glVertex2fv");
+    glad_glVertex2i = (PFNGLVERTEX2IPROC) load(userptr, "glVertex2i");
+    glad_glVertex2iv = (PFNGLVERTEX2IVPROC) load(userptr, "glVertex2iv");
+    glad_glVertex2s = (PFNGLVERTEX2SPROC) load(userptr, "glVertex2s");
+    glad_glVertex2sv = (PFNGLVERTEX2SVPROC) load(userptr, "glVertex2sv");
+    glad_glVertex3d = (PFNGLVERTEX3DPROC) load(userptr, "glVertex3d");
+    glad_glVertex3dv = (PFNGLVERTEX3DVPROC) load(userptr, "glVertex3dv");
+    glad_glVertex3f = (PFNGLVERTEX3FPROC) load(userptr, "glVertex3f");
+    glad_glVertex3fv = (PFNGLVERTEX3FVPROC) load(userptr, "glVertex3fv");
+    glad_glVertex3i = (PFNGLVERTEX3IPROC) load(userptr, "glVertex3i");
+    glad_glVertex3iv = (PFNGLVERTEX3IVPROC) load(userptr, "glVertex3iv");
+    glad_glVertex3s = (PFNGLVERTEX3SPROC) load(userptr, "glVertex3s");
+    glad_glVertex3sv = (PFNGLVERTEX3SVPROC) load(userptr, "glVertex3sv");
+    glad_glVertex4d = (PFNGLVERTEX4DPROC) load(userptr, "glVertex4d");
+    glad_glVertex4dv = (PFNGLVERTEX4DVPROC) load(userptr, "glVertex4dv");
+    glad_glVertex4f = (PFNGLVERTEX4FPROC) load(userptr, "glVertex4f");
+    glad_glVertex4fv = (PFNGLVERTEX4FVPROC) load(userptr, "glVertex4fv");
+    glad_glVertex4i = (PFNGLVERTEX4IPROC) load(userptr, "glVertex4i");
+    glad_glVertex4iv = (PFNGLVERTEX4IVPROC) load(userptr, "glVertex4iv");
+    glad_glVertex4s = (PFNGLVERTEX4SPROC) load(userptr, "glVertex4s");
+    glad_glVertex4sv = (PFNGLVERTEX4SVPROC) load(userptr, "glVertex4sv");
+    glad_glViewport = (PFNGLVIEWPORTPROC) load(userptr, "glViewport");
+}
+static void glad_gl_load_GL_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_1) return;
+    glad_glAreTexturesResident = (PFNGLARETEXTURESRESIDENTPROC) load(userptr, "glAreTexturesResident");
+    glad_glArrayElement = (PFNGLARRAYELEMENTPROC) load(userptr, "glArrayElement");
+    glad_glBindTexture = (PFNGLBINDTEXTUREPROC) load(userptr, "glBindTexture");
+    glad_glColorPointer = (PFNGLCOLORPOINTERPROC) load(userptr, "glColorPointer");
+    glad_glCopyTexImage1D = (PFNGLCOPYTEXIMAGE1DPROC) load(userptr, "glCopyTexImage1D");
+    glad_glCopyTexImage2D = (PFNGLCOPYTEXIMAGE2DPROC) load(userptr, "glCopyTexImage2D");
+    glad_glCopyTexSubImage1D = (PFNGLCOPYTEXSUBIMAGE1DPROC) load(userptr, "glCopyTexSubImage1D");
+    glad_glCopyTexSubImage2D = (PFNGLCOPYTEXSUBIMAGE2DPROC) load(userptr, "glCopyTexSubImage2D");
+    glad_glDeleteTextures = (PFNGLDELETETEXTURESPROC) load(userptr, "glDeleteTextures");
+    glad_glDisableClientState = (PFNGLDISABLECLIENTSTATEPROC) load(userptr, "glDisableClientState");
+    glad_glDrawArrays = (PFNGLDRAWARRAYSPROC) load(userptr, "glDrawArrays");
+    glad_glDrawElements = (PFNGLDRAWELEMENTSPROC) load(userptr, "glDrawElements");
+    glad_glEdgeFlagPointer = (PFNGLEDGEFLAGPOINTERPROC) load(userptr, "glEdgeFlagPointer");
+    glad_glEnableClientState = (PFNGLENABLECLIENTSTATEPROC) load(userptr, "glEnableClientState");
+    glad_glGenTextures = (PFNGLGENTEXTURESPROC) load(userptr, "glGenTextures");
+    glad_glGetPointerv = (PFNGLGETPOINTERVPROC) load(userptr, "glGetPointerv");
+    glad_glIndexPointer = (PFNGLINDEXPOINTERPROC) load(userptr, "glIndexPointer");
+    glad_glIndexub = (PFNGLINDEXUBPROC) load(userptr, "glIndexub");
+    glad_glIndexubv = (PFNGLINDEXUBVPROC) load(userptr, "glIndexubv");
+    glad_glInterleavedArrays = (PFNGLINTERLEAVEDARRAYSPROC) load(userptr, "glInterleavedArrays");
+    glad_glIsTexture = (PFNGLISTEXTUREPROC) load(userptr, "glIsTexture");
+    glad_glNormalPointer = (PFNGLNORMALPOINTERPROC) load(userptr, "glNormalPointer");
+    glad_glPolygonOffset = (PFNGLPOLYGONOFFSETPROC) load(userptr, "glPolygonOffset");
+    glad_glPopClientAttrib = (PFNGLPOPCLIENTATTRIBPROC) load(userptr, "glPopClientAttrib");
+    glad_glPrioritizeTextures = (PFNGLPRIORITIZETEXTURESPROC) load(userptr, "glPrioritizeTextures");
+    glad_glPushClientAttrib = (PFNGLPUSHCLIENTATTRIBPROC) load(userptr, "glPushClientAttrib");
+    glad_glTexCoordPointer = (PFNGLTEXCOORDPOINTERPROC) load(userptr, "glTexCoordPointer");
+    glad_glTexSubImage1D = (PFNGLTEXSUBIMAGE1DPROC) load(userptr, "glTexSubImage1D");
+    glad_glTexSubImage2D = (PFNGLTEXSUBIMAGE2DPROC) load(userptr, "glTexSubImage2D");
+    glad_glVertexPointer = (PFNGLVERTEXPOINTERPROC) load(userptr, "glVertexPointer");
+}
+static void glad_gl_load_GL_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_2) return;
+    glad_glCopyTexSubImage3D = (PFNGLCOPYTEXSUBIMAGE3DPROC) load(userptr, "glCopyTexSubImage3D");
+    glad_glDrawRangeElements = (PFNGLDRAWRANGEELEMENTSPROC) load(userptr, "glDrawRangeElements");
+    glad_glTexImage3D = (PFNGLTEXIMAGE3DPROC) load(userptr, "glTexImage3D");
+    glad_glTexSubImage3D = (PFNGLTEXSUBIMAGE3DPROC) load(userptr, "glTexSubImage3D");
+}
+static void glad_gl_load_GL_VERSION_1_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_3) return;
+    glad_glActiveTexture = (PFNGLACTIVETEXTUREPROC) load(userptr, "glActiveTexture");
+    glad_glClientActiveTexture = (PFNGLCLIENTACTIVETEXTUREPROC) load(userptr, "glClientActiveTexture");
+    glad_glCompressedTexImage1D = (PFNGLCOMPRESSEDTEXIMAGE1DPROC) load(userptr, "glCompressedTexImage1D");
+    glad_glCompressedTexImage2D = (PFNGLCOMPRESSEDTEXIMAGE2DPROC) load(userptr, "glCompressedTexImage2D");
+    glad_glCompressedTexImage3D = (PFNGLCOMPRESSEDTEXIMAGE3DPROC) load(userptr, "glCompressedTexImage3D");
+    glad_glCompressedTexSubImage1D = (PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) load(userptr, "glCompressedTexSubImage1D");
+    glad_glCompressedTexSubImage2D = (PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) load(userptr, "glCompressedTexSubImage2D");
+    glad_glCompressedTexSubImage3D = (PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) load(userptr, "glCompressedTexSubImage3D");
+    glad_glGetCompressedTexImage = (PFNGLGETCOMPRESSEDTEXIMAGEPROC) load(userptr, "glGetCompressedTexImage");
+    glad_glLoadTransposeMatrixd = (PFNGLLOADTRANSPOSEMATRIXDPROC) load(userptr, "glLoadTransposeMatrixd");
+    glad_glLoadTransposeMatrixf = (PFNGLLOADTRANSPOSEMATRIXFPROC) load(userptr, "glLoadTransposeMatrixf");
+    glad_glMultTransposeMatrixd = (PFNGLMULTTRANSPOSEMATRIXDPROC) load(userptr, "glMultTransposeMatrixd");
+    glad_glMultTransposeMatrixf = (PFNGLMULTTRANSPOSEMATRIXFPROC) load(userptr, "glMultTransposeMatrixf");
+    glad_glMultiTexCoord1d = (PFNGLMULTITEXCOORD1DPROC) load(userptr, "glMultiTexCoord1d");
+    glad_glMultiTexCoord1dv = (PFNGLMULTITEXCOORD1DVPROC) load(userptr, "glMultiTexCoord1dv");
+    glad_glMultiTexCoord1f = (PFNGLMULTITEXCOORD1FPROC) load(userptr, "glMultiTexCoord1f");
+    glad_glMultiTexCoord1fv = (PFNGLMULTITEXCOORD1FVPROC) load(userptr, "glMultiTexCoord1fv");
+    glad_glMultiTexCoord1i = (PFNGLMULTITEXCOORD1IPROC) load(userptr, "glMultiTexCoord1i");
+    glad_glMultiTexCoord1iv = (PFNGLMULTITEXCOORD1IVPROC) load(userptr, "glMultiTexCoord1iv");
+    glad_glMultiTexCoord1s = (PFNGLMULTITEXCOORD1SPROC) load(userptr, "glMultiTexCoord1s");
+    glad_glMultiTexCoord1sv = (PFNGLMULTITEXCOORD1SVPROC) load(userptr, "glMultiTexCoord1sv");
+    glad_glMultiTexCoord2d = (PFNGLMULTITEXCOORD2DPROC) load(userptr, "glMultiTexCoord2d");
+    glad_glMultiTexCoord2dv = (PFNGLMULTITEXCOORD2DVPROC) load(userptr, "glMultiTexCoord2dv");
+    glad_glMultiTexCoord2f = (PFNGLMULTITEXCOORD2FPROC) load(userptr, "glMultiTexCoord2f");
+    glad_glMultiTexCoord2fv = (PFNGLMULTITEXCOORD2FVPROC) load(userptr, "glMultiTexCoord2fv");
+    glad_glMultiTexCoord2i = (PFNGLMULTITEXCOORD2IPROC) load(userptr, "glMultiTexCoord2i");
+    glad_glMultiTexCoord2iv = (PFNGLMULTITEXCOORD2IVPROC) load(userptr, "glMultiTexCoord2iv");
+    glad_glMultiTexCoord2s = (PFNGLMULTITEXCOORD2SPROC) load(userptr, "glMultiTexCoord2s");
+    glad_glMultiTexCoord2sv = (PFNGLMULTITEXCOORD2SVPROC) load(userptr, "glMultiTexCoord2sv");
+    glad_glMultiTexCoord3d = (PFNGLMULTITEXCOORD3DPROC) load(userptr, "glMultiTexCoord3d");
+    glad_glMultiTexCoord3dv = (PFNGLMULTITEXCOORD3DVPROC) load(userptr, "glMultiTexCoord3dv");
+    glad_glMultiTexCoord3f = (PFNGLMULTITEXCOORD3FPROC) load(userptr, "glMultiTexCoord3f");
+    glad_glMultiTexCoord3fv = (PFNGLMULTITEXCOORD3FVPROC) load(userptr, "glMultiTexCoord3fv");
+    glad_glMultiTexCoord3i = (PFNGLMULTITEXCOORD3IPROC) load(userptr, "glMultiTexCoord3i");
+    glad_glMultiTexCoord3iv = (PFNGLMULTITEXCOORD3IVPROC) load(userptr, "glMultiTexCoord3iv");
+    glad_glMultiTexCoord3s = (PFNGLMULTITEXCOORD3SPROC) load(userptr, "glMultiTexCoord3s");
+    glad_glMultiTexCoord3sv = (PFNGLMULTITEXCOORD3SVPROC) load(userptr, "glMultiTexCoord3sv");
+    glad_glMultiTexCoord4d = (PFNGLMULTITEXCOORD4DPROC) load(userptr, "glMultiTexCoord4d");
+    glad_glMultiTexCoord4dv = (PFNGLMULTITEXCOORD4DVPROC) load(userptr, "glMultiTexCoord4dv");
+    glad_glMultiTexCoord4f = (PFNGLMULTITEXCOORD4FPROC) load(userptr, "glMultiTexCoord4f");
+    glad_glMultiTexCoord4fv = (PFNGLMULTITEXCOORD4FVPROC) load(userptr, "glMultiTexCoord4fv");
+    glad_glMultiTexCoord4i = (PFNGLMULTITEXCOORD4IPROC) load(userptr, "glMultiTexCoord4i");
+    glad_glMultiTexCoord4iv = (PFNGLMULTITEXCOORD4IVPROC) load(userptr, "glMultiTexCoord4iv");
+    glad_glMultiTexCoord4s = (PFNGLMULTITEXCOORD4SPROC) load(userptr, "glMultiTexCoord4s");
+    glad_glMultiTexCoord4sv = (PFNGLMULTITEXCOORD4SVPROC) load(userptr, "glMultiTexCoord4sv");
+    glad_glSampleCoverage = (PFNGLSAMPLECOVERAGEPROC) load(userptr, "glSampleCoverage");
+}
+static void glad_gl_load_GL_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_4) return;
+    glad_glBlendColor = (PFNGLBLENDCOLORPROC) load(userptr, "glBlendColor");
+    glad_glBlendEquation = (PFNGLBLENDEQUATIONPROC) load(userptr, "glBlendEquation");
+    glad_glBlendFuncSeparate = (PFNGLBLENDFUNCSEPARATEPROC) load(userptr, "glBlendFuncSeparate");
+    glad_glFogCoordPointer = (PFNGLFOGCOORDPOINTERPROC) load(userptr, "glFogCoordPointer");
+    glad_glFogCoordd = (PFNGLFOGCOORDDPROC) load(userptr, "glFogCoordd");
+    glad_glFogCoorddv = (PFNGLFOGCOORDDVPROC) load(userptr, "glFogCoorddv");
+    glad_glFogCoordf = (PFNGLFOGCOORDFPROC) load(userptr, "glFogCoordf");
+    glad_glFogCoordfv = (PFNGLFOGCOORDFVPROC) load(userptr, "glFogCoordfv");
+    glad_glMultiDrawArrays = (PFNGLMULTIDRAWARRAYSPROC) load(userptr, "glMultiDrawArrays");
+    glad_glMultiDrawElements = (PFNGLMULTIDRAWELEMENTSPROC) load(userptr, "glMultiDrawElements");
+    glad_glPointParameterf = (PFNGLPOINTPARAMETERFPROC) load(userptr, "glPointParameterf");
+    glad_glPointParameterfv = (PFNGLPOINTPARAMETERFVPROC) load(userptr, "glPointParameterfv");
+    glad_glPointParameteri = (PFNGLPOINTPARAMETERIPROC) load(userptr, "glPointParameteri");
+    glad_glPointParameteriv = (PFNGLPOINTPARAMETERIVPROC) load(userptr, "glPointParameteriv");
+    glad_glSecondaryColor3b = (PFNGLSECONDARYCOLOR3BPROC) load(userptr, "glSecondaryColor3b");
+    glad_glSecondaryColor3bv = (PFNGLSECONDARYCOLOR3BVPROC) load(userptr, "glSecondaryColor3bv");
+    glad_glSecondaryColor3d = (PFNGLSECONDARYCOLOR3DPROC) load(userptr, "glSecondaryColor3d");
+    glad_glSecondaryColor3dv = (PFNGLSECONDARYCOLOR3DVPROC) load(userptr, "glSecondaryColor3dv");
+    glad_glSecondaryColor3f = (PFNGLSECONDARYCOLOR3FPROC) load(userptr, "glSecondaryColor3f");
+    glad_glSecondaryColor3fv = (PFNGLSECONDARYCOLOR3FVPROC) load(userptr, "glSecondaryColor3fv");
+    glad_glSecondaryColor3i = (PFNGLSECONDARYCOLOR3IPROC) load(userptr, "glSecondaryColor3i");
+    glad_glSecondaryColor3iv = (PFNGLSECONDARYCOLOR3IVPROC) load(userptr, "glSecondaryColor3iv");
+    glad_glSecondaryColor3s = (PFNGLSECONDARYCOLOR3SPROC) load(userptr, "glSecondaryColor3s");
+    glad_glSecondaryColor3sv = (PFNGLSECONDARYCOLOR3SVPROC) load(userptr, "glSecondaryColor3sv");
+    glad_glSecondaryColor3ub = (PFNGLSECONDARYCOLOR3UBPROC) load(userptr, "glSecondaryColor3ub");
+    glad_glSecondaryColor3ubv = (PFNGLSECONDARYCOLOR3UBVPROC) load(userptr, "glSecondaryColor3ubv");
+    glad_glSecondaryColor3ui = (PFNGLSECONDARYCOLOR3UIPROC) load(userptr, "glSecondaryColor3ui");
+    glad_glSecondaryColor3uiv = (PFNGLSECONDARYCOLOR3UIVPROC) load(userptr, "glSecondaryColor3uiv");
+    glad_glSecondaryColor3us = (PFNGLSECONDARYCOLOR3USPROC) load(userptr, "glSecondaryColor3us");
+    glad_glSecondaryColor3usv = (PFNGLSECONDARYCOLOR3USVPROC) load(userptr, "glSecondaryColor3usv");
+    glad_glSecondaryColorPointer = (PFNGLSECONDARYCOLORPOINTERPROC) load(userptr, "glSecondaryColorPointer");
+    glad_glWindowPos2d = (PFNGLWINDOWPOS2DPROC) load(userptr, "glWindowPos2d");
+    glad_glWindowPos2dv = (PFNGLWINDOWPOS2DVPROC) load(userptr, "glWindowPos2dv");
+    glad_glWindowPos2f = (PFNGLWINDOWPOS2FPROC) load(userptr, "glWindowPos2f");
+    glad_glWindowPos2fv = (PFNGLWINDOWPOS2FVPROC) load(userptr, "glWindowPos2fv");
+    glad_glWindowPos2i = (PFNGLWINDOWPOS2IPROC) load(userptr, "glWindowPos2i");
+    glad_glWindowPos2iv = (PFNGLWINDOWPOS2IVPROC) load(userptr, "glWindowPos2iv");
+    glad_glWindowPos2s = (PFNGLWINDOWPOS2SPROC) load(userptr, "glWindowPos2s");
+    glad_glWindowPos2sv = (PFNGLWINDOWPOS2SVPROC) load(userptr, "glWindowPos2sv");
+    glad_glWindowPos3d = (PFNGLWINDOWPOS3DPROC) load(userptr, "glWindowPos3d");
+    glad_glWindowPos3dv = (PFNGLWINDOWPOS3DVPROC) load(userptr, "glWindowPos3dv");
+    glad_glWindowPos3f = (PFNGLWINDOWPOS3FPROC) load(userptr, "glWindowPos3f");
+    glad_glWindowPos3fv = (PFNGLWINDOWPOS3FVPROC) load(userptr, "glWindowPos3fv");
+    glad_glWindowPos3i = (PFNGLWINDOWPOS3IPROC) load(userptr, "glWindowPos3i");
+    glad_glWindowPos3iv = (PFNGLWINDOWPOS3IVPROC) load(userptr, "glWindowPos3iv");
+    glad_glWindowPos3s = (PFNGLWINDOWPOS3SPROC) load(userptr, "glWindowPos3s");
+    glad_glWindowPos3sv = (PFNGLWINDOWPOS3SVPROC) load(userptr, "glWindowPos3sv");
+}
+static void glad_gl_load_GL_VERSION_1_5( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_5) return;
+    glad_glBeginQuery = (PFNGLBEGINQUERYPROC) load(userptr, "glBeginQuery");
+    glad_glBindBuffer = (PFNGLBINDBUFFERPROC) load(userptr, "glBindBuffer");
+    glad_glBufferData = (PFNGLBUFFERDATAPROC) load(userptr, "glBufferData");
+    glad_glBufferSubData = (PFNGLBUFFERSUBDATAPROC) load(userptr, "glBufferSubData");
+    glad_glDeleteBuffers = (PFNGLDELETEBUFFERSPROC) load(userptr, "glDeleteBuffers");
+    glad_glDeleteQueries = (PFNGLDELETEQUERIESPROC) load(userptr, "glDeleteQueries");
+    glad_glEndQuery = (PFNGLENDQUERYPROC) load(userptr, "glEndQuery");
+    glad_glGenBuffers = (PFNGLGENBUFFERSPROC) load(userptr, "glGenBuffers");
+    glad_glGenQueries = (PFNGLGENQUERIESPROC) load(userptr, "glGenQueries");
+    glad_glGetBufferParameteriv = (PFNGLGETBUFFERPARAMETERIVPROC) load(userptr, "glGetBufferParameteriv");
+    glad_glGetBufferPointerv = (PFNGLGETBUFFERPOINTERVPROC) load(userptr, "glGetBufferPointerv");
+    glad_glGetBufferSubData = (PFNGLGETBUFFERSUBDATAPROC) load(userptr, "glGetBufferSubData");
+    glad_glGetQueryObjectiv = (PFNGLGETQUERYOBJECTIVPROC) load(userptr, "glGetQueryObjectiv");
+    glad_glGetQueryObjectuiv = (PFNGLGETQUERYOBJECTUIVPROC) load(userptr, "glGetQueryObjectuiv");
+    glad_glGetQueryiv = (PFNGLGETQUERYIVPROC) load(userptr, "glGetQueryiv");
+    glad_glIsBuffer = (PFNGLISBUFFERPROC) load(userptr, "glIsBuffer");
+    glad_glIsQuery = (PFNGLISQUERYPROC) load(userptr, "glIsQuery");
+    glad_glMapBuffer = (PFNGLMAPBUFFERPROC) load(userptr, "glMapBuffer");
+    glad_glUnmapBuffer = (PFNGLUNMAPBUFFERPROC) load(userptr, "glUnmapBuffer");
+}
+static void glad_gl_load_GL_VERSION_2_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_2_0) return;
+    glad_glAttachShader = (PFNGLATTACHSHADERPROC) load(userptr, "glAttachShader");
+    glad_glBindAttribLocation = (PFNGLBINDATTRIBLOCATIONPROC) load(userptr, "glBindAttribLocation");
+    glad_glBlendEquationSeparate = (PFNGLBLENDEQUATIONSEPARATEPROC) load(userptr, "glBlendEquationSeparate");
+    glad_glCompileShader = (PFNGLCOMPILESHADERPROC) load(userptr, "glCompileShader");
+    glad_glCreateProgram = (PFNGLCREATEPROGRAMPROC) load(userptr, "glCreateProgram");
+    glad_glCreateShader = (PFNGLCREATESHADERPROC) load(userptr, "glCreateShader");
+    glad_glDeleteProgram = (PFNGLDELETEPROGRAMPROC) load(userptr, "glDeleteProgram");
+    glad_glDeleteShader = (PFNGLDELETESHADERPROC) load(userptr, "glDeleteShader");
+    glad_glDetachShader = (PFNGLDETACHSHADERPROC) load(userptr, "glDetachShader");
+    glad_glDisableVertexAttribArray = (PFNGLDISABLEVERTEXATTRIBARRAYPROC) load(userptr, "glDisableVertexAttribArray");
+    glad_glDrawBuffers = (PFNGLDRAWBUFFERSPROC) load(userptr, "glDrawBuffers");
+    glad_glEnableVertexAttribArray = (PFNGLENABLEVERTEXATTRIBARRAYPROC) load(userptr, "glEnableVertexAttribArray");
+    glad_glGetActiveAttrib = (PFNGLGETACTIVEATTRIBPROC) load(userptr, "glGetActiveAttrib");
+    glad_glGetActiveUniform = (PFNGLGETACTIVEUNIFORMPROC) load(userptr, "glGetActiveUniform");
+    glad_glGetAttachedShaders = (PFNGLGETATTACHEDSHADERSPROC) load(userptr, "glGetAttachedShaders");
+    glad_glGetAttribLocation = (PFNGLGETATTRIBLOCATIONPROC) load(userptr, "glGetAttribLocation");
+    glad_glGetProgramInfoLog = (PFNGLGETPROGRAMINFOLOGPROC) load(userptr, "glGetProgramInfoLog");
+    glad_glGetProgramiv = (PFNGLGETPROGRAMIVPROC) load(userptr, "glGetProgramiv");
+    glad_glGetShaderInfoLog = (PFNGLGETSHADERINFOLOGPROC) load(userptr, "glGetShaderInfoLog");
+    glad_glGetShaderSource = (PFNGLGETSHADERSOURCEPROC) load(userptr, "glGetShaderSource");
+    glad_glGetShaderiv = (PFNGLGETSHADERIVPROC) load(userptr, "glGetShaderiv");
+    glad_glGetUniformLocation = (PFNGLGETUNIFORMLOCATIONPROC) load(userptr, "glGetUniformLocation");
+    glad_glGetUniformfv = (PFNGLGETUNIFORMFVPROC) load(userptr, "glGetUniformfv");
+    glad_glGetUniformiv = (PFNGLGETUNIFORMIVPROC) load(userptr, "glGetUniformiv");
+    glad_glGetVertexAttribPointerv = (PFNGLGETVERTEXATTRIBPOINTERVPROC) load(userptr, "glGetVertexAttribPointerv");
+    glad_glGetVertexAttribdv = (PFNGLGETVERTEXATTRIBDVPROC) load(userptr, "glGetVertexAttribdv");
+    glad_glGetVertexAttribfv = (PFNGLGETVERTEXATTRIBFVPROC) load(userptr, "glGetVertexAttribfv");
+    glad_glGetVertexAttribiv = (PFNGLGETVERTEXATTRIBIVPROC) load(userptr, "glGetVertexAttribiv");
+    glad_glIsProgram = (PFNGLISPROGRAMPROC) load(userptr, "glIsProgram");
+    glad_glIsShader = (PFNGLISSHADERPROC) load(userptr, "glIsShader");
+    glad_glLinkProgram = (PFNGLLINKPROGRAMPROC) load(userptr, "glLinkProgram");
+    glad_glShaderSource = (PFNGLSHADERSOURCEPROC) load(userptr, "glShaderSource");
+    glad_glStencilFuncSeparate = (PFNGLSTENCILFUNCSEPARATEPROC) load(userptr, "glStencilFuncSeparate");
+    glad_glStencilMaskSeparate = (PFNGLSTENCILMASKSEPARATEPROC) load(userptr, "glStencilMaskSeparate");
+    glad_glStencilOpSeparate = (PFNGLSTENCILOPSEPARATEPROC) load(userptr, "glStencilOpSeparate");
+    glad_glUniform1f = (PFNGLUNIFORM1FPROC) load(userptr, "glUniform1f");
+    glad_glUniform1fv = (PFNGLUNIFORM1FVPROC) load(userptr, "glUniform1fv");
+    glad_glUniform1i = (PFNGLUNIFORM1IPROC) load(userptr, "glUniform1i");
+    glad_glUniform1iv = (PFNGLUNIFORM1IVPROC) load(userptr, "glUniform1iv");
+    glad_glUniform2f = (PFNGLUNIFORM2FPROC) load(userptr, "glUniform2f");
+    glad_glUniform2fv = (PFNGLUNIFORM2FVPROC) load(userptr, "glUniform2fv");
+    glad_glUniform2i = (PFNGLUNIFORM2IPROC) load(userptr, "glUniform2i");
+    glad_glUniform2iv = (PFNGLUNIFORM2IVPROC) load(userptr, "glUniform2iv");
+    glad_glUniform3f = (PFNGLUNIFORM3FPROC) load(userptr, "glUniform3f");
+    glad_glUniform3fv = (PFNGLUNIFORM3FVPROC) load(userptr, "glUniform3fv");
+    glad_glUniform3i = (PFNGLUNIFORM3IPROC) load(userptr, "glUniform3i");
+    glad_glUniform3iv = (PFNGLUNIFORM3IVPROC) load(userptr, "glUniform3iv");
+    glad_glUniform4f = (PFNGLUNIFORM4FPROC) load(userptr, "glUniform4f");
+    glad_glUniform4fv = (PFNGLUNIFORM4FVPROC) load(userptr, "glUniform4fv");
+    glad_glUniform4i = (PFNGLUNIFORM4IPROC) load(userptr, "glUniform4i");
+    glad_glUniform4iv = (PFNGLUNIFORM4IVPROC) load(userptr, "glUniform4iv");
+    glad_glUniformMatrix2fv = (PFNGLUNIFORMMATRIX2FVPROC) load(userptr, "glUniformMatrix2fv");
+    glad_glUniformMatrix3fv = (PFNGLUNIFORMMATRIX3FVPROC) load(userptr, "glUniformMatrix3fv");
+    glad_glUniformMatrix4fv = (PFNGLUNIFORMMATRIX4FVPROC) load(userptr, "glUniformMatrix4fv");
+    glad_glUseProgram = (PFNGLUSEPROGRAMPROC) load(userptr, "glUseProgram");
+    glad_glValidateProgram = (PFNGLVALIDATEPROGRAMPROC) load(userptr, "glValidateProgram");
+    glad_glVertexAttrib1d = (PFNGLVERTEXATTRIB1DPROC) load(userptr, "glVertexAttrib1d");
+    glad_glVertexAttrib1dv = (PFNGLVERTEXATTRIB1DVPROC) load(userptr, "glVertexAttrib1dv");
+    glad_glVertexAttrib1f = (PFNGLVERTEXATTRIB1FPROC) load(userptr, "glVertexAttrib1f");
+    glad_glVertexAttrib1fv = (PFNGLVERTEXATTRIB1FVPROC) load(userptr, "glVertexAttrib1fv");
+    glad_glVertexAttrib1s = (PFNGLVERTEXATTRIB1SPROC) load(userptr, "glVertexAttrib1s");
+    glad_glVertexAttrib1sv = (PFNGLVERTEXATTRIB1SVPROC) load(userptr, "glVertexAttrib1sv");
+    glad_glVertexAttrib2d = (PFNGLVERTEXATTRIB2DPROC) load(userptr, "glVertexAttrib2d");
+    glad_glVertexAttrib2dv = (PFNGLVERTEXATTRIB2DVPROC) load(userptr, "glVertexAttrib2dv");
+    glad_glVertexAttrib2f = (PFNGLVERTEXATTRIB2FPROC) load(userptr, "glVertexAttrib2f");
+    glad_glVertexAttrib2fv = (PFNGLVERTEXATTRIB2FVPROC) load(userptr, "glVertexAttrib2fv");
+    glad_glVertexAttrib2s = (PFNGLVERTEXATTRIB2SPROC) load(userptr, "glVertexAttrib2s");
+    glad_glVertexAttrib2sv = (PFNGLVERTEXATTRIB2SVPROC) load(userptr, "glVertexAttrib2sv");
+    glad_glVertexAttrib3d = (PFNGLVERTEXATTRIB3DPROC) load(userptr, "glVertexAttrib3d");
+    glad_glVertexAttrib3dv = (PFNGLVERTEXATTRIB3DVPROC) load(userptr, "glVertexAttrib3dv");
+    glad_glVertexAttrib3f = (PFNGLVERTEXATTRIB3FPROC) load(userptr, "glVertexAttrib3f");
+    glad_glVertexAttrib3fv = (PFNGLVERTEXATTRIB3FVPROC) load(userptr, "glVertexAttrib3fv");
+    glad_glVertexAttrib3s = (PFNGLVERTEXATTRIB3SPROC) load(userptr, "glVertexAttrib3s");
+    glad_glVertexAttrib3sv = (PFNGLVERTEXATTRIB3SVPROC) load(userptr, "glVertexAttrib3sv");
+    glad_glVertexAttrib4Nbv = (PFNGLVERTEXATTRIB4NBVPROC) load(userptr, "glVertexAttrib4Nbv");
+    glad_glVertexAttrib4Niv = (PFNGLVERTEXATTRIB4NIVPROC) load(userptr, "glVertexAttrib4Niv");
+    glad_glVertexAttrib4Nsv = (PFNGLVERTEXATTRIB4NSVPROC) load(userptr, "glVertexAttrib4Nsv");
+    glad_glVertexAttrib4Nub = (PFNGLVERTEXATTRIB4NUBPROC) load(userptr, "glVertexAttrib4Nub");
+    glad_glVertexAttrib4Nubv = (PFNGLVERTEXATTRIB4NUBVPROC) load(userptr, "glVertexAttrib4Nubv");
+    glad_glVertexAttrib4Nuiv = (PFNGLVERTEXATTRIB4NUIVPROC) load(userptr, "glVertexAttrib4Nuiv");
+    glad_glVertexAttrib4Nusv = (PFNGLVERTEXATTRIB4NUSVPROC) load(userptr, "glVertexAttrib4Nusv");
+    glad_glVertexAttrib4bv = (PFNGLVERTEXATTRIB4BVPROC) load(userptr, "glVertexAttrib4bv");
+    glad_glVertexAttrib4d = (PFNGLVERTEXATTRIB4DPROC) load(userptr, "glVertexAttrib4d");
+    glad_glVertexAttrib4dv = (PFNGLVERTEXATTRIB4DVPROC) load(userptr, "glVertexAttrib4dv");
+    glad_glVertexAttrib4f = (PFNGLVERTEXATTRIB4FPROC) load(userptr, "glVertexAttrib4f");
+    glad_glVertexAttrib4fv = (PFNGLVERTEXATTRIB4FVPROC) load(userptr, "glVertexAttrib4fv");
+    glad_glVertexAttrib4iv = (PFNGLVERTEXATTRIB4IVPROC) load(userptr, "glVertexAttrib4iv");
+    glad_glVertexAttrib4s = (PFNGLVERTEXATTRIB4SPROC) load(userptr, "glVertexAttrib4s");
+    glad_glVertexAttrib4sv = (PFNGLVERTEXATTRIB4SVPROC) load(userptr, "glVertexAttrib4sv");
+    glad_glVertexAttrib4ubv = (PFNGLVERTEXATTRIB4UBVPROC) load(userptr, "glVertexAttrib4ubv");
+    glad_glVertexAttrib4uiv = (PFNGLVERTEXATTRIB4UIVPROC) load(userptr, "glVertexAttrib4uiv");
+    glad_glVertexAttrib4usv = (PFNGLVERTEXATTRIB4USVPROC) load(userptr, "glVertexAttrib4usv");
+    glad_glVertexAttribPointer = (PFNGLVERTEXATTRIBPOINTERPROC) load(userptr, "glVertexAttribPointer");
+}
+static void glad_gl_load_GL_VERSION_2_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_2_1) return;
+    glad_glUniformMatrix2x3fv = (PFNGLUNIFORMMATRIX2X3FVPROC) load(userptr, "glUniformMatrix2x3fv");
+    glad_glUniformMatrix2x4fv = (PFNGLUNIFORMMATRIX2X4FVPROC) load(userptr, "glUniformMatrix2x4fv");
+    glad_glUniformMatrix3x2fv = (PFNGLUNIFORMMATRIX3X2FVPROC) load(userptr, "glUniformMatrix3x2fv");
+    glad_glUniformMatrix3x4fv = (PFNGLUNIFORMMATRIX3X4FVPROC) load(userptr, "glUniformMatrix3x4fv");
+    glad_glUniformMatrix4x2fv = (PFNGLUNIFORMMATRIX4X2FVPROC) load(userptr, "glUniformMatrix4x2fv");
+    glad_glUniformMatrix4x3fv = (PFNGLUNIFORMMATRIX4X3FVPROC) load(userptr, "glUniformMatrix4x3fv");
+}
+static void glad_gl_load_GL_VERSION_3_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_0) return;
+    glad_glBeginConditionalRender = (PFNGLBEGINCONDITIONALRENDERPROC) load(userptr, "glBeginConditionalRender");
+    glad_glBeginTransformFeedback = (PFNGLBEGINTRANSFORMFEEDBACKPROC) load(userptr, "glBeginTransformFeedback");
+    glad_glBindBufferBase = (PFNGLBINDBUFFERBASEPROC) load(userptr, "glBindBufferBase");
+    glad_glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC) load(userptr, "glBindBufferRange");
+    glad_glBindFragDataLocation = (PFNGLBINDFRAGDATALOCATIONPROC) load(userptr, "glBindFragDataLocation");
+    glad_glBindFramebuffer = (PFNGLBINDFRAMEBUFFERPROC) load(userptr, "glBindFramebuffer");
+    glad_glBindRenderbuffer = (PFNGLBINDRENDERBUFFERPROC) load(userptr, "glBindRenderbuffer");
+    glad_glBindVertexArray = (PFNGLBINDVERTEXARRAYPROC) load(userptr, "glBindVertexArray");
+    glad_glBlitFramebuffer = (PFNGLBLITFRAMEBUFFERPROC) load(userptr, "glBlitFramebuffer");
+    glad_glCheckFramebufferStatus = (PFNGLCHECKFRAMEBUFFERSTATUSPROC) load(userptr, "glCheckFramebufferStatus");
+    glad_glClampColor = (PFNGLCLAMPCOLORPROC) load(userptr, "glClampColor");
+    glad_glClearBufferfi = (PFNGLCLEARBUFFERFIPROC) load(userptr, "glClearBufferfi");
+    glad_glClearBufferfv = (PFNGLCLEARBUFFERFVPROC) load(userptr, "glClearBufferfv");
+    glad_glClearBufferiv = (PFNGLCLEARBUFFERIVPROC) load(userptr, "glClearBufferiv");
+    glad_glClearBufferuiv = (PFNGLCLEARBUFFERUIVPROC) load(userptr, "glClearBufferuiv");
+    glad_glColorMaski = (PFNGLCOLORMASKIPROC) load(userptr, "glColorMaski");
+    glad_glDeleteFramebuffers = (PFNGLDELETEFRAMEBUFFERSPROC) load(userptr, "glDeleteFramebuffers");
+    glad_glDeleteRenderbuffers = (PFNGLDELETERENDERBUFFERSPROC) load(userptr, "glDeleteRenderbuffers");
+    glad_glDeleteVertexArrays = (PFNGLDELETEVERTEXARRAYSPROC) load(userptr, "glDeleteVertexArrays");
+    glad_glDisablei = (PFNGLDISABLEIPROC) load(userptr, "glDisablei");
+    glad_glEnablei = (PFNGLENABLEIPROC) load(userptr, "glEnablei");
+    glad_glEndConditionalRender = (PFNGLENDCONDITIONALRENDERPROC) load(userptr, "glEndConditionalRender");
+    glad_glEndTransformFeedback = (PFNGLENDTRANSFORMFEEDBACKPROC) load(userptr, "glEndTransformFeedback");
+    glad_glFlushMappedBufferRange = (PFNGLFLUSHMAPPEDBUFFERRANGEPROC) load(userptr, "glFlushMappedBufferRange");
+    glad_glFramebufferRenderbuffer = (PFNGLFRAMEBUFFERRENDERBUFFERPROC) load(userptr, "glFramebufferRenderbuffer");
+    glad_glFramebufferTexture1D = (PFNGLFRAMEBUFFERTEXTURE1DPROC) load(userptr, "glFramebufferTexture1D");
+    glad_glFramebufferTexture2D = (PFNGLFRAMEBUFFERTEXTURE2DPROC) load(userptr, "glFramebufferTexture2D");
+    glad_glFramebufferTexture3D = (PFNGLFRAMEBUFFERTEXTURE3DPROC) load(userptr, "glFramebufferTexture3D");
+    glad_glFramebufferTextureLayer = (PFNGLFRAMEBUFFERTEXTURELAYERPROC) load(userptr, "glFramebufferTextureLayer");
+    glad_glGenFramebuffers = (PFNGLGENFRAMEBUFFERSPROC) load(userptr, "glGenFramebuffers");
+    glad_glGenRenderbuffers = (PFNGLGENRENDERBUFFERSPROC) load(userptr, "glGenRenderbuffers");
+    glad_glGenVertexArrays = (PFNGLGENVERTEXARRAYSPROC) load(userptr, "glGenVertexArrays");
+    glad_glGenerateMipmap = (PFNGLGENERATEMIPMAPPROC) load(userptr, "glGenerateMipmap");
+    glad_glGetBooleani_v = (PFNGLGETBOOLEANI_VPROC) load(userptr, "glGetBooleani_v");
+    glad_glGetFragDataLocation = (PFNGLGETFRAGDATALOCATIONPROC) load(userptr, "glGetFragDataLocation");
+    glad_glGetFramebufferAttachmentParameteriv = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC) load(userptr, "glGetFramebufferAttachmentParameteriv");
+    glad_glGetIntegeri_v = (PFNGLGETINTEGERI_VPROC) load(userptr, "glGetIntegeri_v");
+    glad_glGetRenderbufferParameteriv = (PFNGLGETRENDERBUFFERPARAMETERIVPROC) load(userptr, "glGetRenderbufferParameteriv");
+    glad_glGetStringi = (PFNGLGETSTRINGIPROC) load(userptr, "glGetStringi");
+    glad_glGetTexParameterIiv = (PFNGLGETTEXPARAMETERIIVPROC) load(userptr, "glGetTexParameterIiv");
+    glad_glGetTexParameterIuiv = (PFNGLGETTEXPARAMETERIUIVPROC) load(userptr, "glGetTexParameterIuiv");
+    glad_glGetTransformFeedbackVarying = (PFNGLGETTRANSFORMFEEDBACKVARYINGPROC) load(userptr, "glGetTransformFeedbackVarying");
+    glad_glGetUniformuiv = (PFNGLGETUNIFORMUIVPROC) load(userptr, "glGetUniformuiv");
+    glad_glGetVertexAttribIiv = (PFNGLGETVERTEXATTRIBIIVPROC) load(userptr, "glGetVertexAttribIiv");
+    glad_glGetVertexAttribIuiv = (PFNGLGETVERTEXATTRIBIUIVPROC) load(userptr, "glGetVertexAttribIuiv");
+    glad_glIsEnabledi = (PFNGLISENABLEDIPROC) load(userptr, "glIsEnabledi");
+    glad_glIsFramebuffer = (PFNGLISFRAMEBUFFERPROC) load(userptr, "glIsFramebuffer");
+    glad_glIsRenderbuffer = (PFNGLISRENDERBUFFERPROC) load(userptr, "glIsRenderbuffer");
+    glad_glIsVertexArray = (PFNGLISVERTEXARRAYPROC) load(userptr, "glIsVertexArray");
+    glad_glMapBufferRange = (PFNGLMAPBUFFERRANGEPROC) load(userptr, "glMapBufferRange");
+    glad_glRenderbufferStorage = (PFNGLRENDERBUFFERSTORAGEPROC) load(userptr, "glRenderbufferStorage");
+    glad_glRenderbufferStorageMultisample = (PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC) load(userptr, "glRenderbufferStorageMultisample");
+    glad_glTexParameterIiv = (PFNGLTEXPARAMETERIIVPROC) load(userptr, "glTexParameterIiv");
+    glad_glTexParameterIuiv = (PFNGLTEXPARAMETERIUIVPROC) load(userptr, "glTexParameterIuiv");
+    glad_glTransformFeedbackVaryings = (PFNGLTRANSFORMFEEDBACKVARYINGSPROC) load(userptr, "glTransformFeedbackVaryings");
+    glad_glUniform1ui = (PFNGLUNIFORM1UIPROC) load(userptr, "glUniform1ui");
+    glad_glUniform1uiv = (PFNGLUNIFORM1UIVPROC) load(userptr, "glUniform1uiv");
+    glad_glUniform2ui = (PFNGLUNIFORM2UIPROC) load(userptr, "glUniform2ui");
+    glad_glUniform2uiv = (PFNGLUNIFORM2UIVPROC) load(userptr, "glUniform2uiv");
+    glad_glUniform3ui = (PFNGLUNIFORM3UIPROC) load(userptr, "glUniform3ui");
+    glad_glUniform3uiv = (PFNGLUNIFORM3UIVPROC) load(userptr, "glUniform3uiv");
+    glad_glUniform4ui = (PFNGLUNIFORM4UIPROC) load(userptr, "glUniform4ui");
+    glad_glUniform4uiv = (PFNGLUNIFORM4UIVPROC) load(userptr, "glUniform4uiv");
+    glad_glVertexAttribI1i = (PFNGLVERTEXATTRIBI1IPROC) load(userptr, "glVertexAttribI1i");
+    glad_glVertexAttribI1iv = (PFNGLVERTEXATTRIBI1IVPROC) load(userptr, "glVertexAttribI1iv");
+    glad_glVertexAttribI1ui = (PFNGLVERTEXATTRIBI1UIPROC) load(userptr, "glVertexAttribI1ui");
+    glad_glVertexAttribI1uiv = (PFNGLVERTEXATTRIBI1UIVPROC) load(userptr, "glVertexAttribI1uiv");
+    glad_glVertexAttribI2i = (PFNGLVERTEXATTRIBI2IPROC) load(userptr, "glVertexAttribI2i");
+    glad_glVertexAttribI2iv = (PFNGLVERTEXATTRIBI2IVPROC) load(userptr, "glVertexAttribI2iv");
+    glad_glVertexAttribI2ui = (PFNGLVERTEXATTRIBI2UIPROC) load(userptr, "glVertexAttribI2ui");
+    glad_glVertexAttribI2uiv = (PFNGLVERTEXATTRIBI2UIVPROC) load(userptr, "glVertexAttribI2uiv");
+    glad_glVertexAttribI3i = (PFNGLVERTEXATTRIBI3IPROC) load(userptr, "glVertexAttribI3i");
+    glad_glVertexAttribI3iv = (PFNGLVERTEXATTRIBI3IVPROC) load(userptr, "glVertexAttribI3iv");
+    glad_glVertexAttribI3ui = (PFNGLVERTEXATTRIBI3UIPROC) load(userptr, "glVertexAttribI3ui");
+    glad_glVertexAttribI3uiv = (PFNGLVERTEXATTRIBI3UIVPROC) load(userptr, "glVertexAttribI3uiv");
+    glad_glVertexAttribI4bv = (PFNGLVERTEXATTRIBI4BVPROC) load(userptr, "glVertexAttribI4bv");
+    glad_glVertexAttribI4i = (PFNGLVERTEXATTRIBI4IPROC) load(userptr, "glVertexAttribI4i");
+    glad_glVertexAttribI4iv = (PFNGLVERTEXATTRIBI4IVPROC) load(userptr, "glVertexAttribI4iv");
+    glad_glVertexAttribI4sv = (PFNGLVERTEXATTRIBI4SVPROC) load(userptr, "glVertexAttribI4sv");
+    glad_glVertexAttribI4ubv = (PFNGLVERTEXATTRIBI4UBVPROC) load(userptr, "glVertexAttribI4ubv");
+    glad_glVertexAttribI4ui = (PFNGLVERTEXATTRIBI4UIPROC) load(userptr, "glVertexAttribI4ui");
+    glad_glVertexAttribI4uiv = (PFNGLVERTEXATTRIBI4UIVPROC) load(userptr, "glVertexAttribI4uiv");
+    glad_glVertexAttribI4usv = (PFNGLVERTEXATTRIBI4USVPROC) load(userptr, "glVertexAttribI4usv");
+    glad_glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC) load(userptr, "glVertexAttribIPointer");
+}
+static void glad_gl_load_GL_VERSION_3_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_1) return;
+    glad_glBindBufferBase = (PFNGLBINDBUFFERBASEPROC) load(userptr, "glBindBufferBase");
+    glad_glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC) load(userptr, "glBindBufferRange");
+    glad_glCopyBufferSubData = (PFNGLCOPYBUFFERSUBDATAPROC) load(userptr, "glCopyBufferSubData");
+    glad_glDrawArraysInstanced = (PFNGLDRAWARRAYSINSTANCEDPROC) load(userptr, "glDrawArraysInstanced");
+    glad_glDrawElementsInstanced = (PFNGLDRAWELEMENTSINSTANCEDPROC) load(userptr, "glDrawElementsInstanced");
+    glad_glGetActiveUniformBlockName = (PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC) load(userptr, "glGetActiveUniformBlockName");
+    glad_glGetActiveUniformBlockiv = (PFNGLGETACTIVEUNIFORMBLOCKIVPROC) load(userptr, "glGetActiveUniformBlockiv");
+    glad_glGetActiveUniformName = (PFNGLGETACTIVEUNIFORMNAMEPROC) load(userptr, "glGetActiveUniformName");
+    glad_glGetActiveUniformsiv = (PFNGLGETACTIVEUNIFORMSIVPROC) load(userptr, "glGetActiveUniformsiv");
+    glad_glGetIntegeri_v = (PFNGLGETINTEGERI_VPROC) load(userptr, "glGetIntegeri_v");
+    glad_glGetUniformBlockIndex = (PFNGLGETUNIFORMBLOCKINDEXPROC) load(userptr, "glGetUniformBlockIndex");
+    glad_glGetUniformIndices = (PFNGLGETUNIFORMINDICESPROC) load(userptr, "glGetUniformIndices");
+    glad_glPrimitiveRestartIndex = (PFNGLPRIMITIVERESTARTINDEXPROC) load(userptr, "glPrimitiveRestartIndex");
+    glad_glTexBuffer = (PFNGLTEXBUFFERPROC) load(userptr, "glTexBuffer");
+    glad_glUniformBlockBinding = (PFNGLUNIFORMBLOCKBINDINGPROC) load(userptr, "glUniformBlockBinding");
+}
+static void glad_gl_load_GL_VERSION_3_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_2) return;
+    glad_glClientWaitSync = (PFNGLCLIENTWAITSYNCPROC) load(userptr, "glClientWaitSync");
+    glad_glDeleteSync = (PFNGLDELETESYNCPROC) load(userptr, "glDeleteSync");
+    glad_glDrawElementsBaseVertex = (PFNGLDRAWELEMENTSBASEVERTEXPROC) load(userptr, "glDrawElementsBaseVertex");
+    glad_glDrawElementsInstancedBaseVertex = (PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC) load(userptr, "glDrawElementsInstancedBaseVertex");
+    glad_glDrawRangeElementsBaseVertex = (PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC) load(userptr, "glDrawRangeElementsBaseVertex");
+    glad_glFenceSync = (PFNGLFENCESYNCPROC) load(userptr, "glFenceSync");
+    glad_glFramebufferTexture = (PFNGLFRAMEBUFFERTEXTUREPROC) load(userptr, "glFramebufferTexture");
+    glad_glGetBufferParameteri64v = (PFNGLGETBUFFERPARAMETERI64VPROC) load(userptr, "glGetBufferParameteri64v");
+    glad_glGetInteger64i_v = (PFNGLGETINTEGER64I_VPROC) load(userptr, "glGetInteger64i_v");
+    glad_glGetInteger64v = (PFNGLGETINTEGER64VPROC) load(userptr, "glGetInteger64v");
+    glad_glGetMultisamplefv = (PFNGLGETMULTISAMPLEFVPROC) load(userptr, "glGetMultisamplefv");
+    glad_glGetSynciv = (PFNGLGETSYNCIVPROC) load(userptr, "glGetSynciv");
+    glad_glIsSync = (PFNGLISSYNCPROC) load(userptr, "glIsSync");
+    glad_glMultiDrawElementsBaseVertex = (PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC) load(userptr, "glMultiDrawElementsBaseVertex");
+    glad_glProvokingVertex = (PFNGLPROVOKINGVERTEXPROC) load(userptr, "glProvokingVertex");
+    glad_glSampleMaski = (PFNGLSAMPLEMASKIPROC) load(userptr, "glSampleMaski");
+    glad_glTexImage2DMultisample = (PFNGLTEXIMAGE2DMULTISAMPLEPROC) load(userptr, "glTexImage2DMultisample");
+    glad_glTexImage3DMultisample = (PFNGLTEXIMAGE3DMULTISAMPLEPROC) load(userptr, "glTexImage3DMultisample");
+    glad_glWaitSync = (PFNGLWAITSYNCPROC) load(userptr, "glWaitSync");
+}
+static void glad_gl_load_GL_VERSION_3_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_3) return;
+    glad_glBindFragDataLocationIndexed = (PFNGLBINDFRAGDATALOCATIONINDEXEDPROC) load(userptr, "glBindFragDataLocationIndexed");
+    glad_glBindSampler = (PFNGLBINDSAMPLERPROC) load(userptr, "glBindSampler");
+    glad_glColorP3ui = (PFNGLCOLORP3UIPROC) load(userptr, "glColorP3ui");
+    glad_glColorP3uiv = (PFNGLCOLORP3UIVPROC) load(userptr, "glColorP3uiv");
+    glad_glColorP4ui = (PFNGLCOLORP4UIPROC) load(userptr, "glColorP4ui");
+    glad_glColorP4uiv = (PFNGLCOLORP4UIVPROC) load(userptr, "glColorP4uiv");
+    glad_glDeleteSamplers = (PFNGLDELETESAMPLERSPROC) load(userptr, "glDeleteSamplers");
+    glad_glGenSamplers = (PFNGLGENSAMPLERSPROC) load(userptr, "glGenSamplers");
+    glad_glGetFragDataIndex = (PFNGLGETFRAGDATAINDEXPROC) load(userptr, "glGetFragDataIndex");
+    glad_glGetQueryObjecti64v = (PFNGLGETQUERYOBJECTI64VPROC) load(userptr, "glGetQueryObjecti64v");
+    glad_glGetQueryObjectui64v = (PFNGLGETQUERYOBJECTUI64VPROC) load(userptr, "glGetQueryObjectui64v");
+    glad_glGetSamplerParameterIiv = (PFNGLGETSAMPLERPARAMETERIIVPROC) load(userptr, "glGetSamplerParameterIiv");
+    glad_glGetSamplerParameterIuiv = (PFNGLGETSAMPLERPARAMETERIUIVPROC) load(userptr, "glGetSamplerParameterIuiv");
+    glad_glGetSamplerParameterfv = (PFNGLGETSAMPLERPARAMETERFVPROC) load(userptr, "glGetSamplerParameterfv");
+    glad_glGetSamplerParameteriv = (PFNGLGETSAMPLERPARAMETERIVPROC) load(userptr, "glGetSamplerParameteriv");
+    glad_glIsSampler = (PFNGLISSAMPLERPROC) load(userptr, "glIsSampler");
+    glad_glMultiTexCoordP1ui = (PFNGLMULTITEXCOORDP1UIPROC) load(userptr, "glMultiTexCoordP1ui");
+    glad_glMultiTexCoordP1uiv = (PFNGLMULTITEXCOORDP1UIVPROC) load(userptr, "glMultiTexCoordP1uiv");
+    glad_glMultiTexCoordP2ui = (PFNGLMULTITEXCOORDP2UIPROC) load(userptr, "glMultiTexCoordP2ui");
+    glad_glMultiTexCoordP2uiv = (PFNGLMULTITEXCOORDP2UIVPROC) load(userptr, "glMultiTexCoordP2uiv");
+    glad_glMultiTexCoordP3ui = (PFNGLMULTITEXCOORDP3UIPROC) load(userptr, "glMultiTexCoordP3ui");
+    glad_glMultiTexCoordP3uiv = (PFNGLMULTITEXCOORDP3UIVPROC) load(userptr, "glMultiTexCoordP3uiv");
+    glad_glMultiTexCoordP4ui = (PFNGLMULTITEXCOORDP4UIPROC) load(userptr, "glMultiTexCoordP4ui");
+    glad_glMultiTexCoordP4uiv = (PFNGLMULTITEXCOORDP4UIVPROC) load(userptr, "glMultiTexCoordP4uiv");
+    glad_glNormalP3ui = (PFNGLNORMALP3UIPROC) load(userptr, "glNormalP3ui");
+    glad_glNormalP3uiv = (PFNGLNORMALP3UIVPROC) load(userptr, "glNormalP3uiv");
+    glad_glQueryCounter = (PFNGLQUERYCOUNTERPROC) load(userptr, "glQueryCounter");
+    glad_glSamplerParameterIiv = (PFNGLSAMPLERPARAMETERIIVPROC) load(userptr, "glSamplerParameterIiv");
+    glad_glSamplerParameterIuiv = (PFNGLSAMPLERPARAMETERIUIVPROC) load(userptr, "glSamplerParameterIuiv");
+    glad_glSamplerParameterf = (PFNGLSAMPLERPARAMETERFPROC) load(userptr, "glSamplerParameterf");
+    glad_glSamplerParameterfv = (PFNGLSAMPLERPARAMETERFVPROC) load(userptr, "glSamplerParameterfv");
+    glad_glSamplerParameteri = (PFNGLSAMPLERPARAMETERIPROC) load(userptr, "glSamplerParameteri");
+    glad_glSamplerParameteriv = (PFNGLSAMPLERPARAMETERIVPROC) load(userptr, "glSamplerParameteriv");
+    glad_glSecondaryColorP3ui = (PFNGLSECONDARYCOLORP3UIPROC) load(userptr, "glSecondaryColorP3ui");
+    glad_glSecondaryColorP3uiv = (PFNGLSECONDARYCOLORP3UIVPROC) load(userptr, "glSecondaryColorP3uiv");
+    glad_glTexCoordP1ui = (PFNGLTEXCOORDP1UIPROC) load(userptr, "glTexCoordP1ui");
+    glad_glTexCoordP1uiv = (PFNGLTEXCOORDP1UIVPROC) load(userptr, "glTexCoordP1uiv");
+    glad_glTexCoordP2ui = (PFNGLTEXCOORDP2UIPROC) load(userptr, "glTexCoordP2ui");
+    glad_glTexCoordP2uiv = (PFNGLTEXCOORDP2UIVPROC) load(userptr, "glTexCoordP2uiv");
+    glad_glTexCoordP3ui = (PFNGLTEXCOORDP3UIPROC) load(userptr, "glTexCoordP3ui");
+    glad_glTexCoordP3uiv = (PFNGLTEXCOORDP3UIVPROC) load(userptr, "glTexCoordP3uiv");
+    glad_glTexCoordP4ui = (PFNGLTEXCOORDP4UIPROC) load(userptr, "glTexCoordP4ui");
+    glad_glTexCoordP4uiv = (PFNGLTEXCOORDP4UIVPROC) load(userptr, "glTexCoordP4uiv");
+    glad_glVertexAttribDivisor = (PFNGLVERTEXATTRIBDIVISORPROC) load(userptr, "glVertexAttribDivisor");
+    glad_glVertexAttribP1ui = (PFNGLVERTEXATTRIBP1UIPROC) load(userptr, "glVertexAttribP1ui");
+    glad_glVertexAttribP1uiv = (PFNGLVERTEXATTRIBP1UIVPROC) load(userptr, "glVertexAttribP1uiv");
+    glad_glVertexAttribP2ui = (PFNGLVERTEXATTRIBP2UIPROC) load(userptr, "glVertexAttribP2ui");
+    glad_glVertexAttribP2uiv = (PFNGLVERTEXATTRIBP2UIVPROC) load(userptr, "glVertexAttribP2uiv");
+    glad_glVertexAttribP3ui = (PFNGLVERTEXATTRIBP3UIPROC) load(userptr, "glVertexAttribP3ui");
+    glad_glVertexAttribP3uiv = (PFNGLVERTEXATTRIBP3UIVPROC) load(userptr, "glVertexAttribP3uiv");
+    glad_glVertexAttribP4ui = (PFNGLVERTEXATTRIBP4UIPROC) load(userptr, "glVertexAttribP4ui");
+    glad_glVertexAttribP4uiv = (PFNGLVERTEXATTRIBP4UIVPROC) load(userptr, "glVertexAttribP4uiv");
+    glad_glVertexP2ui = (PFNGLVERTEXP2UIPROC) load(userptr, "glVertexP2ui");
+    glad_glVertexP2uiv = (PFNGLVERTEXP2UIVPROC) load(userptr, "glVertexP2uiv");
+    glad_glVertexP3ui = (PFNGLVERTEXP3UIPROC) load(userptr, "glVertexP3ui");
+    glad_glVertexP3uiv = (PFNGLVERTEXP3UIVPROC) load(userptr, "glVertexP3uiv");
+    glad_glVertexP4ui = (PFNGLVERTEXP4UIPROC) load(userptr, "glVertexP4ui");
+    glad_glVertexP4uiv = (PFNGLVERTEXP4UIVPROC) load(userptr, "glVertexP4uiv");
+}
+
+
+
+#if defined(GL_ES_VERSION_3_0) || defined(GL_VERSION_3_0)
+#define GLAD_GL_IS_SOME_NEW_VERSION 1
+#else
+#define GLAD_GL_IS_SOME_NEW_VERSION 0
+#endif
+
+static int glad_gl_get_extensions( int version, const char **out_exts, unsigned int *out_num_exts_i, char ***out_exts_i) {
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    if(GLAD_VERSION_MAJOR(version) < 3) {
+#else
+    (void) version;
+    (void) out_num_exts_i;
+    (void) out_exts_i;
+#endif
+        if (glad_glGetString == NULL) {
+            return 0;
+        }
+        *out_exts = (const char *)glad_glGetString(GL_EXTENSIONS);
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    } else {
+        unsigned int index = 0;
+        unsigned int num_exts_i = 0;
+        char **exts_i = NULL;
+        if (glad_glGetStringi == NULL || glad_glGetIntegerv == NULL) {
+            return 0;
+        }
+        glad_glGetIntegerv(GL_NUM_EXTENSIONS, (int*) &num_exts_i);
+        if (num_exts_i > 0) {
+            exts_i = (char **) malloc(num_exts_i * (sizeof *exts_i));
+        }
+        if (exts_i == NULL) {
+            return 0;
+        }
+        for(index = 0; index < num_exts_i; index++) {
+            const char *gl_str_tmp = (const char*) glad_glGetStringi(GL_EXTENSIONS, index);
+            size_t len = strlen(gl_str_tmp) + 1;
+
+            char *local_str = (char*) malloc(len * sizeof(char));
+            if(local_str != NULL) {
+                memcpy(local_str, gl_str_tmp, len * sizeof(char));
+            }
+
+            exts_i[index] = local_str;
+        }
+
+        *out_num_exts_i = num_exts_i;
+        *out_exts_i = exts_i;
+    }
+#endif
+    return 1;
+}
+static void glad_gl_free_extensions(char **exts_i, unsigned int num_exts_i) {
+    if (exts_i != NULL) {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            free((void *) (exts_i[index]));
+        }
+        free((void *)exts_i);
+        exts_i = NULL;
+    }
+}
+static int glad_gl_has_extension(int version, const char *exts, unsigned int num_exts_i, char **exts_i, const char *ext) {
+    if(GLAD_VERSION_MAJOR(version) < 3 || !GLAD_GL_IS_SOME_NEW_VERSION) {
+        const char *extensions;
+        const char *loc;
+        const char *terminator;
+        extensions = exts;
+        if(extensions == NULL || ext == NULL) {
+            return 0;
+        }
+        while(1) {
+            loc = strstr(extensions, ext);
+            if(loc == NULL) {
+                return 0;
+            }
+            terminator = loc + strlen(ext);
+            if((loc == extensions || *(loc - 1) == ' ') &&
+                (*terminator == ' ' || *terminator == '\0')) {
+                return 1;
+            }
+            extensions = terminator;
+        }
+    } else {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            const char *e = exts_i[index];
+            if(strcmp(e, ext) == 0) {
+                return 1;
+            }
+        }
+    }
+    return 0;
+}
+
+static GLADapiproc glad_gl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_gl_find_extensions_gl( int version) {
+    const char *exts = NULL;
+    unsigned int num_exts_i = 0;
+    char **exts_i = NULL;
+    if (!glad_gl_get_extensions(version, &exts, &num_exts_i, &exts_i)) return 0;
+
+    (void) glad_gl_has_extension;
+
+    glad_gl_free_extensions(exts_i, num_exts_i);
+
+    return 1;
+}
+
+static int glad_gl_find_core_gl(void) {
+    int i, major, minor;
+    const char* version;
+    const char* prefixes[] = {
+        "OpenGL ES-CM ",
+        "OpenGL ES-CL ",
+        "OpenGL ES ",
+        NULL
+    };
+    version = (const char*) glad_glGetString(GL_VERSION);
+    if (!version) return 0;
+    for (i = 0;  prefixes[i];  i++) {
+        const size_t length = strlen(prefixes[i]);
+        if (strncmp(version, prefixes[i], length) == 0) {
+            version += length;
+            break;
+        }
+    }
+
+    GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+
+    GLAD_GL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_GL_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_GL_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_GL_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_GL_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    GLAD_GL_VERSION_1_5 = (major == 1 && minor >= 5) || major > 1;
+    GLAD_GL_VERSION_2_0 = (major == 2 && minor >= 0) || major > 2;
+    GLAD_GL_VERSION_2_1 = (major == 2 && minor >= 1) || major > 2;
+    GLAD_GL_VERSION_3_0 = (major == 3 && minor >= 0) || major > 3;
+    GLAD_GL_VERSION_3_1 = (major == 3 && minor >= 1) || major > 3;
+    GLAD_GL_VERSION_3_2 = (major == 3 && minor >= 2) || major > 3;
+    GLAD_GL_VERSION_3_3 = (major == 3 && minor >= 3) || major > 3;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLUserPtr( GLADuserptrloadfunc load, void *userptr) {
+    int version;
+
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    if(glad_glGetString == NULL) return 0;
+    if(glad_glGetString(GL_VERSION) == NULL) return 0;
+    version = glad_gl_find_core_gl();
+
+    glad_gl_load_GL_VERSION_1_0(load, userptr);
+    glad_gl_load_GL_VERSION_1_1(load, userptr);
+    glad_gl_load_GL_VERSION_1_2(load, userptr);
+    glad_gl_load_GL_VERSION_1_3(load, userptr);
+    glad_gl_load_GL_VERSION_1_4(load, userptr);
+    glad_gl_load_GL_VERSION_1_5(load, userptr);
+    glad_gl_load_GL_VERSION_2_0(load, userptr);
+    glad_gl_load_GL_VERSION_2_1(load, userptr);
+    glad_gl_load_GL_VERSION_3_0(load, userptr);
+    glad_gl_load_GL_VERSION_3_1(load, userptr);
+    glad_gl_load_GL_VERSION_3_2(load, userptr);
+    glad_gl_load_GL_VERSION_3_3(load, userptr);
+
+    if (!glad_gl_find_extensions_gl(version)) return 0;
+
+
+
+    return version;
+}
+
+
+int gladLoadGL( GLADloadfunc load) {
+    return gladLoadGLUserPtr( glad_gl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/gles2.c glmark2-2021.02/android/jni/src/glad/src/gles2.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/gles2.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/src/gles2.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,495 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/gles2.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GL_ES_VERSION_2_0 = 0;
+int GLAD_GL_OES_mapbuffer = 0;
+int GLAD_GL_OES_required_internalformat = 0;
+
+
+
+PFNGLACTIVETEXTUREPROC glad_glActiveTexture = NULL;
+PFNGLATTACHSHADERPROC glad_glAttachShader = NULL;
+PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation = NULL;
+PFNGLBINDBUFFERPROC glad_glBindBuffer = NULL;
+PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer = NULL;
+PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer = NULL;
+PFNGLBINDTEXTUREPROC glad_glBindTexture = NULL;
+PFNGLBLENDCOLORPROC glad_glBlendColor = NULL;
+PFNGLBLENDEQUATIONPROC glad_glBlendEquation = NULL;
+PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate = NULL;
+PFNGLBLENDFUNCPROC glad_glBlendFunc = NULL;
+PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate = NULL;
+PFNGLBUFFERDATAPROC glad_glBufferData = NULL;
+PFNGLBUFFERSUBDATAPROC glad_glBufferSubData = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus = NULL;
+PFNGLCLEARPROC glad_glClear = NULL;
+PFNGLCLEARCOLORPROC glad_glClearColor = NULL;
+PFNGLCLEARDEPTHFPROC glad_glClearDepthf = NULL;
+PFNGLCLEARSTENCILPROC glad_glClearStencil = NULL;
+PFNGLCOLORMASKPROC glad_glColorMask = NULL;
+PFNGLCOMPILESHADERPROC glad_glCompileShader = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D = NULL;
+PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D = NULL;
+PFNGLCREATEPROGRAMPROC glad_glCreateProgram = NULL;
+PFNGLCREATESHADERPROC glad_glCreateShader = NULL;
+PFNGLCULLFACEPROC glad_glCullFace = NULL;
+PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers = NULL;
+PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers = NULL;
+PFNGLDELETEPROGRAMPROC glad_glDeleteProgram = NULL;
+PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers = NULL;
+PFNGLDELETESHADERPROC glad_glDeleteShader = NULL;
+PFNGLDELETETEXTURESPROC glad_glDeleteTextures = NULL;
+PFNGLDEPTHFUNCPROC glad_glDepthFunc = NULL;
+PFNGLDEPTHMASKPROC glad_glDepthMask = NULL;
+PFNGLDEPTHRANGEFPROC glad_glDepthRangef = NULL;
+PFNGLDETACHSHADERPROC glad_glDetachShader = NULL;
+PFNGLDISABLEPROC glad_glDisable = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray = NULL;
+PFNGLDRAWARRAYSPROC glad_glDrawArrays = NULL;
+PFNGLDRAWELEMENTSPROC glad_glDrawElements = NULL;
+PFNGLENABLEPROC glad_glEnable = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray = NULL;
+PFNGLFINISHPROC glad_glFinish = NULL;
+PFNGLFLUSHPROC glad_glFlush = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D = NULL;
+PFNGLFRONTFACEPROC glad_glFrontFace = NULL;
+PFNGLGENBUFFERSPROC glad_glGenBuffers = NULL;
+PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers = NULL;
+PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers = NULL;
+PFNGLGENTEXTURESPROC glad_glGenTextures = NULL;
+PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap = NULL;
+PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib = NULL;
+PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform = NULL;
+PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders = NULL;
+PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation = NULL;
+PFNGLGETBOOLEANVPROC glad_glGetBooleanv = NULL;
+PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv = NULL;
+PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv = NULL;
+PFNGLGETBUFFERPOINTERVOESPROC glad_glGetBufferPointervOES = NULL;
+PFNGLGETERRORPROC glad_glGetError = NULL;
+PFNGLGETFLOATVPROC glad_glGetFloatv = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv = NULL;
+PFNGLGETINTEGERVPROC glad_glGetIntegerv = NULL;
+PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog = NULL;
+PFNGLGETPROGRAMIVPROC glad_glGetProgramiv = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv = NULL;
+PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog = NULL;
+PFNGLGETSHADERPRECISIONFORMATPROC glad_glGetShaderPrecisionFormat = NULL;
+PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource = NULL;
+PFNGLGETSHADERIVPROC glad_glGetShaderiv = NULL;
+PFNGLGETSTRINGPROC glad_glGetString = NULL;
+PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv = NULL;
+PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv = NULL;
+PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation = NULL;
+PFNGLGETUNIFORMFVPROC glad_glGetUniformfv = NULL;
+PFNGLGETUNIFORMIVPROC glad_glGetUniformiv = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv = NULL;
+PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv = NULL;
+PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv = NULL;
+PFNGLHINTPROC glad_glHint = NULL;
+PFNGLISBUFFERPROC glad_glIsBuffer = NULL;
+PFNGLISENABLEDPROC glad_glIsEnabled = NULL;
+PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer = NULL;
+PFNGLISPROGRAMPROC glad_glIsProgram = NULL;
+PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer = NULL;
+PFNGLISSHADERPROC glad_glIsShader = NULL;
+PFNGLISTEXTUREPROC glad_glIsTexture = NULL;
+PFNGLLINEWIDTHPROC glad_glLineWidth = NULL;
+PFNGLLINKPROGRAMPROC glad_glLinkProgram = NULL;
+PFNGLMAPBUFFERPROC glad_glMapBuffer = NULL;
+PFNGLMAPBUFFEROESPROC glad_glMapBufferOES = NULL;
+PFNGLPIXELSTOREIPROC glad_glPixelStorei = NULL;
+PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset = NULL;
+PFNGLREADPIXELSPROC glad_glReadPixels = NULL;
+PFNGLRELEASESHADERCOMPILERPROC glad_glReleaseShaderCompiler = NULL;
+PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage = NULL;
+PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage = NULL;
+PFNGLSCISSORPROC glad_glScissor = NULL;
+PFNGLSHADERBINARYPROC glad_glShaderBinary = NULL;
+PFNGLSHADERSOURCEPROC glad_glShaderSource = NULL;
+PFNGLSTENCILFUNCPROC glad_glStencilFunc = NULL;
+PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate = NULL;
+PFNGLSTENCILMASKPROC glad_glStencilMask = NULL;
+PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate = NULL;
+PFNGLSTENCILOPPROC glad_glStencilOp = NULL;
+PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate = NULL;
+PFNGLTEXIMAGE2DPROC glad_glTexImage2D = NULL;
+PFNGLTEXPARAMETERFPROC glad_glTexParameterf = NULL;
+PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv = NULL;
+PFNGLTEXPARAMETERIPROC glad_glTexParameteri = NULL;
+PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv = NULL;
+PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D = NULL;
+PFNGLUNIFORM1FPROC glad_glUniform1f = NULL;
+PFNGLUNIFORM1FVPROC glad_glUniform1fv = NULL;
+PFNGLUNIFORM1IPROC glad_glUniform1i = NULL;
+PFNGLUNIFORM1IVPROC glad_glUniform1iv = NULL;
+PFNGLUNIFORM2FPROC glad_glUniform2f = NULL;
+PFNGLUNIFORM2FVPROC glad_glUniform2fv = NULL;
+PFNGLUNIFORM2IPROC glad_glUniform2i = NULL;
+PFNGLUNIFORM2IVPROC glad_glUniform2iv = NULL;
+PFNGLUNIFORM3FPROC glad_glUniform3f = NULL;
+PFNGLUNIFORM3FVPROC glad_glUniform3fv = NULL;
+PFNGLUNIFORM3IPROC glad_glUniform3i = NULL;
+PFNGLUNIFORM3IVPROC glad_glUniform3iv = NULL;
+PFNGLUNIFORM4FPROC glad_glUniform4f = NULL;
+PFNGLUNIFORM4FVPROC glad_glUniform4fv = NULL;
+PFNGLUNIFORM4IPROC glad_glUniform4i = NULL;
+PFNGLUNIFORM4IVPROC glad_glUniform4iv = NULL;
+PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv = NULL;
+PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv = NULL;
+PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv = NULL;
+PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer = NULL;
+PFNGLUNMAPBUFFEROESPROC glad_glUnmapBufferOES = NULL;
+PFNGLUSEPROGRAMPROC glad_glUseProgram = NULL;
+PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram = NULL;
+PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f = NULL;
+PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv = NULL;
+PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f = NULL;
+PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv = NULL;
+PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f = NULL;
+PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv = NULL;
+PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f = NULL;
+PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv = NULL;
+PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer = NULL;
+PFNGLVIEWPORTPROC glad_glViewport = NULL;
+
+
+static void glad_gl_load_GL_ES_VERSION_2_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_ES_VERSION_2_0) return;
+    glad_glActiveTexture = (PFNGLACTIVETEXTUREPROC) load(userptr, "glActiveTexture");
+    glad_glAttachShader = (PFNGLATTACHSHADERPROC) load(userptr, "glAttachShader");
+    glad_glBindAttribLocation = (PFNGLBINDATTRIBLOCATIONPROC) load(userptr, "glBindAttribLocation");
+    glad_glBindBuffer = (PFNGLBINDBUFFERPROC) load(userptr, "glBindBuffer");
+    glad_glBindFramebuffer = (PFNGLBINDFRAMEBUFFERPROC) load(userptr, "glBindFramebuffer");
+    glad_glBindRenderbuffer = (PFNGLBINDRENDERBUFFERPROC) load(userptr, "glBindRenderbuffer");
+    glad_glBindTexture = (PFNGLBINDTEXTUREPROC) load(userptr, "glBindTexture");
+    glad_glBlendColor = (PFNGLBLENDCOLORPROC) load(userptr, "glBlendColor");
+    glad_glBlendEquation = (PFNGLBLENDEQUATIONPROC) load(userptr, "glBlendEquation");
+    glad_glBlendEquationSeparate = (PFNGLBLENDEQUATIONSEPARATEPROC) load(userptr, "glBlendEquationSeparate");
+    glad_glBlendFunc = (PFNGLBLENDFUNCPROC) load(userptr, "glBlendFunc");
+    glad_glBlendFuncSeparate = (PFNGLBLENDFUNCSEPARATEPROC) load(userptr, "glBlendFuncSeparate");
+    glad_glBufferData = (PFNGLBUFFERDATAPROC) load(userptr, "glBufferData");
+    glad_glBufferSubData = (PFNGLBUFFERSUBDATAPROC) load(userptr, "glBufferSubData");
+    glad_glCheckFramebufferStatus = (PFNGLCHECKFRAMEBUFFERSTATUSPROC) load(userptr, "glCheckFramebufferStatus");
+    glad_glClear = (PFNGLCLEARPROC) load(userptr, "glClear");
+    glad_glClearColor = (PFNGLCLEARCOLORPROC) load(userptr, "glClearColor");
+    glad_glClearDepthf = (PFNGLCLEARDEPTHFPROC) load(userptr, "glClearDepthf");
+    glad_glClearStencil = (PFNGLCLEARSTENCILPROC) load(userptr, "glClearStencil");
+    glad_glColorMask = (PFNGLCOLORMASKPROC) load(userptr, "glColorMask");
+    glad_glCompileShader = (PFNGLCOMPILESHADERPROC) load(userptr, "glCompileShader");
+    glad_glCompressedTexImage2D = (PFNGLCOMPRESSEDTEXIMAGE2DPROC) load(userptr, "glCompressedTexImage2D");
+    glad_glCompressedTexSubImage2D = (PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) load(userptr, "glCompressedTexSubImage2D");
+    glad_glCopyTexImage2D = (PFNGLCOPYTEXIMAGE2DPROC) load(userptr, "glCopyTexImage2D");
+    glad_glCopyTexSubImage2D = (PFNGLCOPYTEXSUBIMAGE2DPROC) load(userptr, "glCopyTexSubImage2D");
+    glad_glCreateProgram = (PFNGLCREATEPROGRAMPROC) load(userptr, "glCreateProgram");
+    glad_glCreateShader = (PFNGLCREATESHADERPROC) load(userptr, "glCreateShader");
+    glad_glCullFace = (PFNGLCULLFACEPROC) load(userptr, "glCullFace");
+    glad_glDeleteBuffers = (PFNGLDELETEBUFFERSPROC) load(userptr, "glDeleteBuffers");
+    glad_glDeleteFramebuffers = (PFNGLDELETEFRAMEBUFFERSPROC) load(userptr, "glDeleteFramebuffers");
+    glad_glDeleteProgram = (PFNGLDELETEPROGRAMPROC) load(userptr, "glDeleteProgram");
+    glad_glDeleteRenderbuffers = (PFNGLDELETERENDERBUFFERSPROC) load(userptr, "glDeleteRenderbuffers");
+    glad_glDeleteShader = (PFNGLDELETESHADERPROC) load(userptr, "glDeleteShader");
+    glad_glDeleteTextures = (PFNGLDELETETEXTURESPROC) load(userptr, "glDeleteTextures");
+    glad_glDepthFunc = (PFNGLDEPTHFUNCPROC) load(userptr, "glDepthFunc");
+    glad_glDepthMask = (PFNGLDEPTHMASKPROC) load(userptr, "glDepthMask");
+    glad_glDepthRangef = (PFNGLDEPTHRANGEFPROC) load(userptr, "glDepthRangef");
+    glad_glDetachShader = (PFNGLDETACHSHADERPROC) load(userptr, "glDetachShader");
+    glad_glDisable = (PFNGLDISABLEPROC) load(userptr, "glDisable");
+    glad_glDisableVertexAttribArray = (PFNGLDISABLEVERTEXATTRIBARRAYPROC) load(userptr, "glDisableVertexAttribArray");
+    glad_glDrawArrays = (PFNGLDRAWARRAYSPROC) load(userptr, "glDrawArrays");
+    glad_glDrawElements = (PFNGLDRAWELEMENTSPROC) load(userptr, "glDrawElements");
+    glad_glEnable = (PFNGLENABLEPROC) load(userptr, "glEnable");
+    glad_glEnableVertexAttribArray = (PFNGLENABLEVERTEXATTRIBARRAYPROC) load(userptr, "glEnableVertexAttribArray");
+    glad_glFinish = (PFNGLFINISHPROC) load(userptr, "glFinish");
+    glad_glFlush = (PFNGLFLUSHPROC) load(userptr, "glFlush");
+    glad_glFramebufferRenderbuffer = (PFNGLFRAMEBUFFERRENDERBUFFERPROC) load(userptr, "glFramebufferRenderbuffer");
+    glad_glFramebufferTexture2D = (PFNGLFRAMEBUFFERTEXTURE2DPROC) load(userptr, "glFramebufferTexture2D");
+    glad_glFrontFace = (PFNGLFRONTFACEPROC) load(userptr, "glFrontFace");
+    glad_glGenBuffers = (PFNGLGENBUFFERSPROC) load(userptr, "glGenBuffers");
+    glad_glGenFramebuffers = (PFNGLGENFRAMEBUFFERSPROC) load(userptr, "glGenFramebuffers");
+    glad_glGenRenderbuffers = (PFNGLGENRENDERBUFFERSPROC) load(userptr, "glGenRenderbuffers");
+    glad_glGenTextures = (PFNGLGENTEXTURESPROC) load(userptr, "glGenTextures");
+    glad_glGenerateMipmap = (PFNGLGENERATEMIPMAPPROC) load(userptr, "glGenerateMipmap");
+    glad_glGetActiveAttrib = (PFNGLGETACTIVEATTRIBPROC) load(userptr, "glGetActiveAttrib");
+    glad_glGetActiveUniform = (PFNGLGETACTIVEUNIFORMPROC) load(userptr, "glGetActiveUniform");
+    glad_glGetAttachedShaders = (PFNGLGETATTACHEDSHADERSPROC) load(userptr, "glGetAttachedShaders");
+    glad_glGetAttribLocation = (PFNGLGETATTRIBLOCATIONPROC) load(userptr, "glGetAttribLocation");
+    glad_glGetBooleanv = (PFNGLGETBOOLEANVPROC) load(userptr, "glGetBooleanv");
+    glad_glGetBufferParameteriv = (PFNGLGETBUFFERPARAMETERIVPROC) load(userptr, "glGetBufferParameteriv");
+    glad_glGetError = (PFNGLGETERRORPROC) load(userptr, "glGetError");
+    glad_glGetFloatv = (PFNGLGETFLOATVPROC) load(userptr, "glGetFloatv");
+    glad_glGetFramebufferAttachmentParameteriv = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC) load(userptr, "glGetFramebufferAttachmentParameteriv");
+    glad_glGetIntegerv = (PFNGLGETINTEGERVPROC) load(userptr, "glGetIntegerv");
+    glad_glGetProgramInfoLog = (PFNGLGETPROGRAMINFOLOGPROC) load(userptr, "glGetProgramInfoLog");
+    glad_glGetProgramiv = (PFNGLGETPROGRAMIVPROC) load(userptr, "glGetProgramiv");
+    glad_glGetRenderbufferParameteriv = (PFNGLGETRENDERBUFFERPARAMETERIVPROC) load(userptr, "glGetRenderbufferParameteriv");
+    glad_glGetShaderInfoLog = (PFNGLGETSHADERINFOLOGPROC) load(userptr, "glGetShaderInfoLog");
+    glad_glGetShaderPrecisionFormat = (PFNGLGETSHADERPRECISIONFORMATPROC) load(userptr, "glGetShaderPrecisionFormat");
+    glad_glGetShaderSource = (PFNGLGETSHADERSOURCEPROC) load(userptr, "glGetShaderSource");
+    glad_glGetShaderiv = (PFNGLGETSHADERIVPROC) load(userptr, "glGetShaderiv");
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    glad_glGetTexParameterfv = (PFNGLGETTEXPARAMETERFVPROC) load(userptr, "glGetTexParameterfv");
+    glad_glGetTexParameteriv = (PFNGLGETTEXPARAMETERIVPROC) load(userptr, "glGetTexParameteriv");
+    glad_glGetUniformLocation = (PFNGLGETUNIFORMLOCATIONPROC) load(userptr, "glGetUniformLocation");
+    glad_glGetUniformfv = (PFNGLGETUNIFORMFVPROC) load(userptr, "glGetUniformfv");
+    glad_glGetUniformiv = (PFNGLGETUNIFORMIVPROC) load(userptr, "glGetUniformiv");
+    glad_glGetVertexAttribPointerv = (PFNGLGETVERTEXATTRIBPOINTERVPROC) load(userptr, "glGetVertexAttribPointerv");
+    glad_glGetVertexAttribfv = (PFNGLGETVERTEXATTRIBFVPROC) load(userptr, "glGetVertexAttribfv");
+    glad_glGetVertexAttribiv = (PFNGLGETVERTEXATTRIBIVPROC) load(userptr, "glGetVertexAttribiv");
+    glad_glHint = (PFNGLHINTPROC) load(userptr, "glHint");
+    glad_glIsBuffer = (PFNGLISBUFFERPROC) load(userptr, "glIsBuffer");
+    glad_glIsEnabled = (PFNGLISENABLEDPROC) load(userptr, "glIsEnabled");
+    glad_glIsFramebuffer = (PFNGLISFRAMEBUFFERPROC) load(userptr, "glIsFramebuffer");
+    glad_glIsProgram = (PFNGLISPROGRAMPROC) load(userptr, "glIsProgram");
+    glad_glIsRenderbuffer = (PFNGLISRENDERBUFFERPROC) load(userptr, "glIsRenderbuffer");
+    glad_glIsShader = (PFNGLISSHADERPROC) load(userptr, "glIsShader");
+    glad_glIsTexture = (PFNGLISTEXTUREPROC) load(userptr, "glIsTexture");
+    glad_glLineWidth = (PFNGLLINEWIDTHPROC) load(userptr, "glLineWidth");
+    glad_glLinkProgram = (PFNGLLINKPROGRAMPROC) load(userptr, "glLinkProgram");
+    glad_glPixelStorei = (PFNGLPIXELSTOREIPROC) load(userptr, "glPixelStorei");
+    glad_glPolygonOffset = (PFNGLPOLYGONOFFSETPROC) load(userptr, "glPolygonOffset");
+    glad_glReadPixels = (PFNGLREADPIXELSPROC) load(userptr, "glReadPixels");
+    glad_glReleaseShaderCompiler = (PFNGLRELEASESHADERCOMPILERPROC) load(userptr, "glReleaseShaderCompiler");
+    glad_glRenderbufferStorage = (PFNGLRENDERBUFFERSTORAGEPROC) load(userptr, "glRenderbufferStorage");
+    glad_glSampleCoverage = (PFNGLSAMPLECOVERAGEPROC) load(userptr, "glSampleCoverage");
+    glad_glScissor = (PFNGLSCISSORPROC) load(userptr, "glScissor");
+    glad_glShaderBinary = (PFNGLSHADERBINARYPROC) load(userptr, "glShaderBinary");
+    glad_glShaderSource = (PFNGLSHADERSOURCEPROC) load(userptr, "glShaderSource");
+    glad_glStencilFunc = (PFNGLSTENCILFUNCPROC) load(userptr, "glStencilFunc");
+    glad_glStencilFuncSeparate = (PFNGLSTENCILFUNCSEPARATEPROC) load(userptr, "glStencilFuncSeparate");
+    glad_glStencilMask = (PFNGLSTENCILMASKPROC) load(userptr, "glStencilMask");
+    glad_glStencilMaskSeparate = (PFNGLSTENCILMASKSEPARATEPROC) load(userptr, "glStencilMaskSeparate");
+    glad_glStencilOp = (PFNGLSTENCILOPPROC) load(userptr, "glStencilOp");
+    glad_glStencilOpSeparate = (PFNGLSTENCILOPSEPARATEPROC) load(userptr, "glStencilOpSeparate");
+    glad_glTexImage2D = (PFNGLTEXIMAGE2DPROC) load(userptr, "glTexImage2D");
+    glad_glTexParameterf = (PFNGLTEXPARAMETERFPROC) load(userptr, "glTexParameterf");
+    glad_glTexParameterfv = (PFNGLTEXPARAMETERFVPROC) load(userptr, "glTexParameterfv");
+    glad_glTexParameteri = (PFNGLTEXPARAMETERIPROC) load(userptr, "glTexParameteri");
+    glad_glTexParameteriv = (PFNGLTEXPARAMETERIVPROC) load(userptr, "glTexParameteriv");
+    glad_glTexSubImage2D = (PFNGLTEXSUBIMAGE2DPROC) load(userptr, "glTexSubImage2D");
+    glad_glUniform1f = (PFNGLUNIFORM1FPROC) load(userptr, "glUniform1f");
+    glad_glUniform1fv = (PFNGLUNIFORM1FVPROC) load(userptr, "glUniform1fv");
+    glad_glUniform1i = (PFNGLUNIFORM1IPROC) load(userptr, "glUniform1i");
+    glad_glUniform1iv = (PFNGLUNIFORM1IVPROC) load(userptr, "glUniform1iv");
+    glad_glUniform2f = (PFNGLUNIFORM2FPROC) load(userptr, "glUniform2f");
+    glad_glUniform2fv = (PFNGLUNIFORM2FVPROC) load(userptr, "glUniform2fv");
+    glad_glUniform2i = (PFNGLUNIFORM2IPROC) load(userptr, "glUniform2i");
+    glad_glUniform2iv = (PFNGLUNIFORM2IVPROC) load(userptr, "glUniform2iv");
+    glad_glUniform3f = (PFNGLUNIFORM3FPROC) load(userptr, "glUniform3f");
+    glad_glUniform3fv = (PFNGLUNIFORM3FVPROC) load(userptr, "glUniform3fv");
+    glad_glUniform3i = (PFNGLUNIFORM3IPROC) load(userptr, "glUniform3i");
+    glad_glUniform3iv = (PFNGLUNIFORM3IVPROC) load(userptr, "glUniform3iv");
+    glad_glUniform4f = (PFNGLUNIFORM4FPROC) load(userptr, "glUniform4f");
+    glad_glUniform4fv = (PFNGLUNIFORM4FVPROC) load(userptr, "glUniform4fv");
+    glad_glUniform4i = (PFNGLUNIFORM4IPROC) load(userptr, "glUniform4i");
+    glad_glUniform4iv = (PFNGLUNIFORM4IVPROC) load(userptr, "glUniform4iv");
+    glad_glUniformMatrix2fv = (PFNGLUNIFORMMATRIX2FVPROC) load(userptr, "glUniformMatrix2fv");
+    glad_glUniformMatrix3fv = (PFNGLUNIFORMMATRIX3FVPROC) load(userptr, "glUniformMatrix3fv");
+    glad_glUniformMatrix4fv = (PFNGLUNIFORMMATRIX4FVPROC) load(userptr, "glUniformMatrix4fv");
+    glad_glUseProgram = (PFNGLUSEPROGRAMPROC) load(userptr, "glUseProgram");
+    glad_glValidateProgram = (PFNGLVALIDATEPROGRAMPROC) load(userptr, "glValidateProgram");
+    glad_glVertexAttrib1f = (PFNGLVERTEXATTRIB1FPROC) load(userptr, "glVertexAttrib1f");
+    glad_glVertexAttrib1fv = (PFNGLVERTEXATTRIB1FVPROC) load(userptr, "glVertexAttrib1fv");
+    glad_glVertexAttrib2f = (PFNGLVERTEXATTRIB2FPROC) load(userptr, "glVertexAttrib2f");
+    glad_glVertexAttrib2fv = (PFNGLVERTEXATTRIB2FVPROC) load(userptr, "glVertexAttrib2fv");
+    glad_glVertexAttrib3f = (PFNGLVERTEXATTRIB3FPROC) load(userptr, "glVertexAttrib3f");
+    glad_glVertexAttrib3fv = (PFNGLVERTEXATTRIB3FVPROC) load(userptr, "glVertexAttrib3fv");
+    glad_glVertexAttrib4f = (PFNGLVERTEXATTRIB4FPROC) load(userptr, "glVertexAttrib4f");
+    glad_glVertexAttrib4fv = (PFNGLVERTEXATTRIB4FVPROC) load(userptr, "glVertexAttrib4fv");
+    glad_glVertexAttribPointer = (PFNGLVERTEXATTRIBPOINTERPROC) load(userptr, "glVertexAttribPointer");
+    glad_glViewport = (PFNGLVIEWPORTPROC) load(userptr, "glViewport");
+}
+static void glad_gl_load_GL_OES_mapbuffer( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_OES_mapbuffer) return;
+    glad_glGetBufferPointerv = (PFNGLGETBUFFERPOINTERVPROC) load(userptr, "glGetBufferPointerv");
+    glad_glGetBufferPointervOES = (PFNGLGETBUFFERPOINTERVOESPROC) load(userptr, "glGetBufferPointervOES");
+    glad_glMapBuffer = (PFNGLMAPBUFFERPROC) load(userptr, "glMapBuffer");
+    glad_glMapBufferOES = (PFNGLMAPBUFFEROESPROC) load(userptr, "glMapBufferOES");
+    glad_glUnmapBuffer = (PFNGLUNMAPBUFFERPROC) load(userptr, "glUnmapBuffer");
+    glad_glUnmapBufferOES = (PFNGLUNMAPBUFFEROESPROC) load(userptr, "glUnmapBufferOES");
+}
+
+
+
+#if defined(GL_ES_VERSION_3_0) || defined(GL_VERSION_3_0)
+#define GLAD_GL_IS_SOME_NEW_VERSION 1
+#else
+#define GLAD_GL_IS_SOME_NEW_VERSION 0
+#endif
+
+static int glad_gl_get_extensions( int version, const char **out_exts, unsigned int *out_num_exts_i, char ***out_exts_i) {
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    if(GLAD_VERSION_MAJOR(version) < 3) {
+#else
+    (void) version;
+    (void) out_num_exts_i;
+    (void) out_exts_i;
+#endif
+        if (glad_glGetString == NULL) {
+            return 0;
+        }
+        *out_exts = (const char *)glad_glGetString(GL_EXTENSIONS);
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    } else {
+        unsigned int index = 0;
+        unsigned int num_exts_i = 0;
+        char **exts_i = NULL;
+        if (glad_glGetStringi == NULL || glad_glGetIntegerv == NULL) {
+            return 0;
+        }
+        glad_glGetIntegerv(GL_NUM_EXTENSIONS, (int*) &num_exts_i);
+        if (num_exts_i > 0) {
+            exts_i = (char **) malloc(num_exts_i * (sizeof *exts_i));
+        }
+        if (exts_i == NULL) {
+            return 0;
+        }
+        for(index = 0; index < num_exts_i; index++) {
+            const char *gl_str_tmp = (const char*) glad_glGetStringi(GL_EXTENSIONS, index);
+            size_t len = strlen(gl_str_tmp) + 1;
+
+            char *local_str = (char*) malloc(len * sizeof(char));
+            if(local_str != NULL) {
+                memcpy(local_str, gl_str_tmp, len * sizeof(char));
+            }
+
+            exts_i[index] = local_str;
+        }
+
+        *out_num_exts_i = num_exts_i;
+        *out_exts_i = exts_i;
+    }
+#endif
+    return 1;
+}
+static void glad_gl_free_extensions(char **exts_i, unsigned int num_exts_i) {
+    if (exts_i != NULL) {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            free((void *) (exts_i[index]));
+        }
+        free((void *)exts_i);
+        exts_i = NULL;
+    }
+}
+static int glad_gl_has_extension(int version, const char *exts, unsigned int num_exts_i, char **exts_i, const char *ext) {
+    if(GLAD_VERSION_MAJOR(version) < 3 || !GLAD_GL_IS_SOME_NEW_VERSION) {
+        const char *extensions;
+        const char *loc;
+        const char *terminator;
+        extensions = exts;
+        if(extensions == NULL || ext == NULL) {
+            return 0;
+        }
+        while(1) {
+            loc = strstr(extensions, ext);
+            if(loc == NULL) {
+                return 0;
+            }
+            terminator = loc + strlen(ext);
+            if((loc == extensions || *(loc - 1) == ' ') &&
+                (*terminator == ' ' || *terminator == '\0')) {
+                return 1;
+            }
+            extensions = terminator;
+        }
+    } else {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            const char *e = exts_i[index];
+            if(strcmp(e, ext) == 0) {
+                return 1;
+            }
+        }
+    }
+    return 0;
+}
+
+static GLADapiproc glad_gl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_gl_find_extensions_gles2( int version) {
+    const char *exts = NULL;
+    unsigned int num_exts_i = 0;
+    char **exts_i = NULL;
+    if (!glad_gl_get_extensions(version, &exts, &num_exts_i, &exts_i)) return 0;
+
+    GLAD_GL_OES_mapbuffer = glad_gl_has_extension(version, exts, num_exts_i, exts_i, "GL_OES_mapbuffer");
+    GLAD_GL_OES_required_internalformat = glad_gl_has_extension(version, exts, num_exts_i, exts_i, "GL_OES_required_internalformat");
+
+    glad_gl_free_extensions(exts_i, num_exts_i);
+
+    return 1;
+}
+
+static int glad_gl_find_core_gles2(void) {
+    int i, major, minor;
+    const char* version;
+    const char* prefixes[] = {
+        "OpenGL ES-CM ",
+        "OpenGL ES-CL ",
+        "OpenGL ES ",
+        NULL
+    };
+    version = (const char*) glad_glGetString(GL_VERSION);
+    if (!version) return 0;
+    for (i = 0;  prefixes[i];  i++) {
+        const size_t length = strlen(prefixes[i]);
+        if (strncmp(version, prefixes[i], length) == 0) {
+            version += length;
+            break;
+        }
+    }
+
+    GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+
+    GLAD_GL_ES_VERSION_2_0 = (major == 2 && minor >= 0) || major > 2;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLES2UserPtr( GLADuserptrloadfunc load, void *userptr) {
+    int version;
+
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    if(glad_glGetString == NULL) return 0;
+    if(glad_glGetString(GL_VERSION) == NULL) return 0;
+    version = glad_gl_find_core_gles2();
+
+    glad_gl_load_GL_ES_VERSION_2_0(load, userptr);
+
+    if (!glad_gl_find_extensions_gles2(version)) return 0;
+    glad_gl_load_GL_OES_mapbuffer(load, userptr);
+
+
+
+    return version;
+}
+
+
+int gladLoadGLES2( GLADloadfunc load) {
+    return gladLoadGLES2UserPtr( glad_gl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/glx.c glmark2-2021.02/android/jni/src/glad/src/glx.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/glx.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/src/glx.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,232 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/glx.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GLX_VERSION_1_0 = 0;
+int GLAD_GLX_VERSION_1_1 = 0;
+int GLAD_GLX_VERSION_1_2 = 0;
+int GLAD_GLX_VERSION_1_3 = 0;
+int GLAD_GLX_VERSION_1_4 = 0;
+int GLAD_GLX_EXT_swap_control = 0;
+int GLAD_GLX_MESA_swap_control = 0;
+
+
+
+PFNGLXCHOOSEFBCONFIGPROC glad_glXChooseFBConfig = NULL;
+PFNGLXCHOOSEVISUALPROC glad_glXChooseVisual = NULL;
+PFNGLXCOPYCONTEXTPROC glad_glXCopyContext = NULL;
+PFNGLXCREATECONTEXTPROC glad_glXCreateContext = NULL;
+PFNGLXCREATEGLXPIXMAPPROC glad_glXCreateGLXPixmap = NULL;
+PFNGLXCREATENEWCONTEXTPROC glad_glXCreateNewContext = NULL;
+PFNGLXCREATEPBUFFERPROC glad_glXCreatePbuffer = NULL;
+PFNGLXCREATEPIXMAPPROC glad_glXCreatePixmap = NULL;
+PFNGLXCREATEWINDOWPROC glad_glXCreateWindow = NULL;
+PFNGLXDESTROYCONTEXTPROC glad_glXDestroyContext = NULL;
+PFNGLXDESTROYGLXPIXMAPPROC glad_glXDestroyGLXPixmap = NULL;
+PFNGLXDESTROYPBUFFERPROC glad_glXDestroyPbuffer = NULL;
+PFNGLXDESTROYPIXMAPPROC glad_glXDestroyPixmap = NULL;
+PFNGLXDESTROYWINDOWPROC glad_glXDestroyWindow = NULL;
+PFNGLXGETCLIENTSTRINGPROC glad_glXGetClientString = NULL;
+PFNGLXGETCONFIGPROC glad_glXGetConfig = NULL;
+PFNGLXGETCURRENTCONTEXTPROC glad_glXGetCurrentContext = NULL;
+PFNGLXGETCURRENTDISPLAYPROC glad_glXGetCurrentDisplay = NULL;
+PFNGLXGETCURRENTDRAWABLEPROC glad_glXGetCurrentDrawable = NULL;
+PFNGLXGETCURRENTREADDRAWABLEPROC glad_glXGetCurrentReadDrawable = NULL;
+PFNGLXGETFBCONFIGATTRIBPROC glad_glXGetFBConfigAttrib = NULL;
+PFNGLXGETFBCONFIGSPROC glad_glXGetFBConfigs = NULL;
+PFNGLXGETPROCADDRESSPROC glad_glXGetProcAddress = NULL;
+PFNGLXGETSELECTEDEVENTPROC glad_glXGetSelectedEvent = NULL;
+PFNGLXGETSWAPINTERVALMESAPROC glad_glXGetSwapIntervalMESA = NULL;
+PFNGLXGETVISUALFROMFBCONFIGPROC glad_glXGetVisualFromFBConfig = NULL;
+PFNGLXISDIRECTPROC glad_glXIsDirect = NULL;
+PFNGLXMAKECONTEXTCURRENTPROC glad_glXMakeContextCurrent = NULL;
+PFNGLXMAKECURRENTPROC glad_glXMakeCurrent = NULL;
+PFNGLXQUERYCONTEXTPROC glad_glXQueryContext = NULL;
+PFNGLXQUERYDRAWABLEPROC glad_glXQueryDrawable = NULL;
+PFNGLXQUERYEXTENSIONPROC glad_glXQueryExtension = NULL;
+PFNGLXQUERYEXTENSIONSSTRINGPROC glad_glXQueryExtensionsString = NULL;
+PFNGLXQUERYSERVERSTRINGPROC glad_glXQueryServerString = NULL;
+PFNGLXQUERYVERSIONPROC glad_glXQueryVersion = NULL;
+PFNGLXSELECTEVENTPROC glad_glXSelectEvent = NULL;
+PFNGLXSWAPBUFFERSPROC glad_glXSwapBuffers = NULL;
+PFNGLXSWAPINTERVALEXTPROC glad_glXSwapIntervalEXT = NULL;
+PFNGLXSWAPINTERVALMESAPROC glad_glXSwapIntervalMESA = NULL;
+PFNGLXUSEXFONTPROC glad_glXUseXFont = NULL;
+PFNGLXWAITGLPROC glad_glXWaitGL = NULL;
+PFNGLXWAITXPROC glad_glXWaitX = NULL;
+
+
+static void glad_glx_load_GLX_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_0) return;
+    glad_glXChooseVisual = (PFNGLXCHOOSEVISUALPROC) load(userptr, "glXChooseVisual");
+    glad_glXCopyContext = (PFNGLXCOPYCONTEXTPROC) load(userptr, "glXCopyContext");
+    glad_glXCreateContext = (PFNGLXCREATECONTEXTPROC) load(userptr, "glXCreateContext");
+    glad_glXCreateGLXPixmap = (PFNGLXCREATEGLXPIXMAPPROC) load(userptr, "glXCreateGLXPixmap");
+    glad_glXDestroyContext = (PFNGLXDESTROYCONTEXTPROC) load(userptr, "glXDestroyContext");
+    glad_glXDestroyGLXPixmap = (PFNGLXDESTROYGLXPIXMAPPROC) load(userptr, "glXDestroyGLXPixmap");
+    glad_glXGetConfig = (PFNGLXGETCONFIGPROC) load(userptr, "glXGetConfig");
+    glad_glXGetCurrentContext = (PFNGLXGETCURRENTCONTEXTPROC) load(userptr, "glXGetCurrentContext");
+    glad_glXGetCurrentDrawable = (PFNGLXGETCURRENTDRAWABLEPROC) load(userptr, "glXGetCurrentDrawable");
+    glad_glXIsDirect = (PFNGLXISDIRECTPROC) load(userptr, "glXIsDirect");
+    glad_glXMakeCurrent = (PFNGLXMAKECURRENTPROC) load(userptr, "glXMakeCurrent");
+    glad_glXQueryExtension = (PFNGLXQUERYEXTENSIONPROC) load(userptr, "glXQueryExtension");
+    glad_glXQueryVersion = (PFNGLXQUERYVERSIONPROC) load(userptr, "glXQueryVersion");
+    glad_glXSwapBuffers = (PFNGLXSWAPBUFFERSPROC) load(userptr, "glXSwapBuffers");
+    glad_glXUseXFont = (PFNGLXUSEXFONTPROC) load(userptr, "glXUseXFont");
+    glad_glXWaitGL = (PFNGLXWAITGLPROC) load(userptr, "glXWaitGL");
+    glad_glXWaitX = (PFNGLXWAITXPROC) load(userptr, "glXWaitX");
+}
+static void glad_glx_load_GLX_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_1) return;
+    glad_glXGetClientString = (PFNGLXGETCLIENTSTRINGPROC) load(userptr, "glXGetClientString");
+    glad_glXQueryExtensionsString = (PFNGLXQUERYEXTENSIONSSTRINGPROC) load(userptr, "glXQueryExtensionsString");
+    glad_glXQueryServerString = (PFNGLXQUERYSERVERSTRINGPROC) load(userptr, "glXQueryServerString");
+}
+static void glad_glx_load_GLX_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_2) return;
+    glad_glXGetCurrentDisplay = (PFNGLXGETCURRENTDISPLAYPROC) load(userptr, "glXGetCurrentDisplay");
+}
+static void glad_glx_load_GLX_VERSION_1_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_3) return;
+    glad_glXChooseFBConfig = (PFNGLXCHOOSEFBCONFIGPROC) load(userptr, "glXChooseFBConfig");
+    glad_glXCreateNewContext = (PFNGLXCREATENEWCONTEXTPROC) load(userptr, "glXCreateNewContext");
+    glad_glXCreatePbuffer = (PFNGLXCREATEPBUFFERPROC) load(userptr, "glXCreatePbuffer");
+    glad_glXCreatePixmap = (PFNGLXCREATEPIXMAPPROC) load(userptr, "glXCreatePixmap");
+    glad_glXCreateWindow = (PFNGLXCREATEWINDOWPROC) load(userptr, "glXCreateWindow");
+    glad_glXDestroyPbuffer = (PFNGLXDESTROYPBUFFERPROC) load(userptr, "glXDestroyPbuffer");
+    glad_glXDestroyPixmap = (PFNGLXDESTROYPIXMAPPROC) load(userptr, "glXDestroyPixmap");
+    glad_glXDestroyWindow = (PFNGLXDESTROYWINDOWPROC) load(userptr, "glXDestroyWindow");
+    glad_glXGetCurrentReadDrawable = (PFNGLXGETCURRENTREADDRAWABLEPROC) load(userptr, "glXGetCurrentReadDrawable");
+    glad_glXGetFBConfigAttrib = (PFNGLXGETFBCONFIGATTRIBPROC) load(userptr, "glXGetFBConfigAttrib");
+    glad_glXGetFBConfigs = (PFNGLXGETFBCONFIGSPROC) load(userptr, "glXGetFBConfigs");
+    glad_glXGetSelectedEvent = (PFNGLXGETSELECTEDEVENTPROC) load(userptr, "glXGetSelectedEvent");
+    glad_glXGetVisualFromFBConfig = (PFNGLXGETVISUALFROMFBCONFIGPROC) load(userptr, "glXGetVisualFromFBConfig");
+    glad_glXMakeContextCurrent = (PFNGLXMAKECONTEXTCURRENTPROC) load(userptr, "glXMakeContextCurrent");
+    glad_glXQueryContext = (PFNGLXQUERYCONTEXTPROC) load(userptr, "glXQueryContext");
+    glad_glXQueryDrawable = (PFNGLXQUERYDRAWABLEPROC) load(userptr, "glXQueryDrawable");
+    glad_glXSelectEvent = (PFNGLXSELECTEVENTPROC) load(userptr, "glXSelectEvent");
+}
+static void glad_glx_load_GLX_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_4) return;
+    glad_glXGetProcAddress = (PFNGLXGETPROCADDRESSPROC) load(userptr, "glXGetProcAddress");
+}
+static void glad_glx_load_GLX_EXT_swap_control( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_EXT_swap_control) return;
+    glad_glXSwapIntervalEXT = (PFNGLXSWAPINTERVALEXTPROC) load(userptr, "glXSwapIntervalEXT");
+}
+static void glad_glx_load_GLX_MESA_swap_control( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_MESA_swap_control) return;
+    glad_glXGetSwapIntervalMESA = (PFNGLXGETSWAPINTERVALMESAPROC) load(userptr, "glXGetSwapIntervalMESA");
+    glad_glXSwapIntervalMESA = (PFNGLXSWAPINTERVALMESAPROC) load(userptr, "glXSwapIntervalMESA");
+}
+
+
+
+static int glad_glx_has_extension(Display *display, int screen, const char *ext) {
+#ifndef GLX_VERSION_1_1
+    (void) display;
+    (void) screen;
+    (void) ext;
+#else
+    const char *terminator;
+    const char *loc;
+    const char *extensions;
+
+    if (glXQueryExtensionsString == NULL) {
+        return 0;
+    }
+
+    extensions = glXQueryExtensionsString(display, screen);
+
+    if(extensions == NULL || ext == NULL) {
+        return 0;
+    }
+
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL)
+            break;
+
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0')) {
+            return 1;
+        }
+        extensions = terminator;
+    }
+#endif
+
+    return 0;
+}
+
+static GLADapiproc glad_glx_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_glx_find_extensions(Display *display, int screen) {
+    GLAD_GLX_EXT_swap_control = glad_glx_has_extension(display, screen, "GLX_EXT_swap_control");
+    GLAD_GLX_MESA_swap_control = glad_glx_has_extension(display, screen, "GLX_MESA_swap_control");
+    return 1;
+}
+
+static int glad_glx_find_core_glx(Display **display, int *screen) {
+    int major = 0, minor = 0;
+    if(*display == NULL) {
+#ifdef GLAD_GLX_NO_X11
+        (void) screen;
+        return 0;
+#else
+        *display = XOpenDisplay(0);
+        if (*display == NULL) {
+            return 0;
+        }
+        *screen = XScreenNumberOfScreen(XDefaultScreenOfDisplay(*display));
+#endif
+    }
+    glXQueryVersion(*display, &major, &minor);
+    GLAD_GLX_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_GLX_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_GLX_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_GLX_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_GLX_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLXUserPtr(Display *display, int screen, GLADuserptrloadfunc load, void *userptr) {
+    int version;
+    glXQueryVersion = (PFNGLXQUERYVERSIONPROC) load(userptr, "glXQueryVersion");
+    if(glXQueryVersion == NULL) return 0;
+    version = glad_glx_find_core_glx(&display, &screen);
+
+    glad_glx_load_GLX_VERSION_1_0(load, userptr);
+    glad_glx_load_GLX_VERSION_1_1(load, userptr);
+    glad_glx_load_GLX_VERSION_1_2(load, userptr);
+    glad_glx_load_GLX_VERSION_1_3(load, userptr);
+    glad_glx_load_GLX_VERSION_1_4(load, userptr);
+
+    if (!glad_glx_find_extensions(display, screen)) return 0;
+    glad_glx_load_GLX_EXT_swap_control(load, userptr);
+    glad_glx_load_GLX_MESA_swap_control(load, userptr);
+
+    return version;
+}
+
+int gladLoadGLX(Display *display, int screen, GLADloadfunc load) {
+    return gladLoadGLXUserPtr(display, screen, glad_glx_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/wgl.c glmark2-2021.02/android/jni/src/glad/src/wgl.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/glad/src/wgl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/glad/src/wgl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,117 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/wgl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_WGL_VERSION_1_0 = 0;
+int GLAD_WGL_ARB_extensions_string = 0;
+int GLAD_WGL_EXT_extensions_string = 0;
+int GLAD_WGL_EXT_swap_control = 0;
+
+
+
+PFNWGLGETEXTENSIONSSTRINGARBPROC glad_wglGetExtensionsStringARB = NULL;
+PFNWGLGETEXTENSIONSSTRINGEXTPROC glad_wglGetExtensionsStringEXT = NULL;
+PFNWGLGETSWAPINTERVALEXTPROC glad_wglGetSwapIntervalEXT = NULL;
+PFNWGLSWAPINTERVALEXTPROC glad_wglSwapIntervalEXT = NULL;
+
+
+static void glad_wgl_load_WGL_ARB_extensions_string(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_ARB_extensions_string) return;
+    glad_wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC) load(userptr, "wglGetExtensionsStringARB");
+}
+static void glad_wgl_load_WGL_EXT_extensions_string(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_EXT_extensions_string) return;
+    glad_wglGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC) load(userptr, "wglGetExtensionsStringEXT");
+}
+static void glad_wgl_load_WGL_EXT_swap_control(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_EXT_swap_control) return;
+    glad_wglGetSwapIntervalEXT = (PFNWGLGETSWAPINTERVALEXTPROC) load(userptr, "wglGetSwapIntervalEXT");
+    glad_wglSwapIntervalEXT = (PFNWGLSWAPINTERVALEXTPROC) load(userptr, "wglSwapIntervalEXT");
+}
+
+
+
+static int glad_wgl_has_extension(HDC hdc, const char *ext) {
+    const char *terminator;
+    const char *loc;
+    const char *extensions;
+
+    if(wglGetExtensionsStringEXT == NULL && wglGetExtensionsStringARB == NULL)
+        return 0;
+
+    if(wglGetExtensionsStringARB == NULL || hdc == INVALID_HANDLE_VALUE)
+        extensions = wglGetExtensionsStringEXT();
+    else
+        extensions = wglGetExtensionsStringARB(hdc);
+
+    if(extensions == NULL || ext == NULL)
+        return 0;
+
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL)
+            break;
+
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0'))
+        {
+            return 1;
+        }
+        extensions = terminator;
+    }
+
+    return 0;
+}
+
+static GLADapiproc glad_wgl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_wgl_find_extensions_wgl(HDC hdc) {
+    GLAD_WGL_ARB_extensions_string = glad_wgl_has_extension(hdc, "WGL_ARB_extensions_string");
+    GLAD_WGL_EXT_extensions_string = glad_wgl_has_extension(hdc, "WGL_EXT_extensions_string");
+    GLAD_WGL_EXT_swap_control = glad_wgl_has_extension(hdc, "WGL_EXT_swap_control");
+    return 1;
+}
+
+static int glad_wgl_find_core_wgl(void) {
+    int major = 1, minor = 0;
+    GLAD_WGL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadWGLUserPtr(HDC hdc, GLADuserptrloadfunc load, void *userptr) {
+    int version;
+    wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC) load(userptr, "wglGetExtensionsStringARB");
+    wglGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC) load(userptr, "wglGetExtensionsStringEXT");
+    if(wglGetExtensionsStringARB == NULL && wglGetExtensionsStringEXT == NULL) return 0;
+    version = glad_wgl_find_core_wgl();
+
+
+    if (!glad_wgl_find_extensions_wgl(hdc)) return 0;
+    glad_wgl_load_WGL_ARB_extensions_string(load, userptr);
+    glad_wgl_load_WGL_EXT_extensions_string(load, userptr);
+    glad_wgl_load_WGL_EXT_swap_control(load, userptr);
+
+    return version;
+}
+
+int gladLoadWGL(HDC hdc, GLADloadfunc load) {
+    return gladLoadWGLUserPtr(hdc, glad_wgl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-headers.cpp glmark2-2021.02/android/jni/src/gl-headers.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-headers.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-headers.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,8 +21,8 @@
  */
 #include "gl-headers.h"
 
-void* (*GLExtensions::MapBuffer) (GLenum target, GLenum access) = 0;
-GLboolean (*GLExtensions::UnmapBuffer) (GLenum target) = 0;
+void* (GLAD_API_PTR *GLExtensions::MapBuffer) (GLenum target, GLenum access) = 0;
+GLboolean (GLAD_API_PTR *GLExtensions::UnmapBuffer) (GLenum target) = 0;
 
 bool
 GLExtensions::support(const std::string &ext)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-headers.h glmark2-2021.02/android/jni/src/gl-headers.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-headers.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-headers.h	2021-04-25 01:47:22.000000000 +0800
@@ -25,14 +25,12 @@
 #define GL_GLEXT_PROTOTYPES
 
 #if GLMARK2_USE_GL
-#include <GL/gl.h>
-#include <GL/glext.h>
+#include <glad/gl.h>
 #ifndef GL_RGB565
 #define GL_RGB565 0x8D62
 #endif
 #elif GLMARK2_USE_GLESv2
-#include <GLES2/gl2.h>
-#include <GLES2/gl2ext.h>
+#include <glad/gles2.h>
 #ifndef GL_WRITE_ONLY
 #define GL_WRITE_ONLY GL_WRITE_ONLY_OES
 #endif
@@ -64,8 +62,8 @@
      */
     static bool support(const std::string &ext);
 
-    static void* (*MapBuffer) (GLenum target, GLenum access);
-    static GLboolean (*UnmapBuffer) (GLenum target);
+    static void* (GLAD_API_PTR *MapBuffer) (GLenum target, GLenum access);
+    static GLboolean (GLAD_API_PTR *UnmapBuffer) (GLenum target);
 };
 
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-egl.cpp glmark2-2021.02/android/jni/src/gl-state-egl.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-egl.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-egl.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -27,10 +27,17 @@
 #include "gl-headers.h"
 #include <iomanip>
 #include <sstream>
+#include <cstring>
 
 using std::vector;
 using std::string;
 
+GLADapiproc load_egl_func(void *userdata, const char *name)
+{
+    SharedLibrary *lib = reinterpret_cast<SharedLibrary *>(userdata);
+    return reinterpret_cast<GLADapiproc>(lib->load(name));
+}
+
 /****************************
  * EGLConfig public methods *
  ****************************/
@@ -287,9 +294,34 @@
  * GLStateEGL public methods *
  ****************************/
 
+GLStateEGL::~GLStateEGL()
+{
+    if(egl_display_ != nullptr){
+        if(!eglTerminate(egl_display_))
+            Log::error("eglTerminate failed\n");
+    }
+
+    if(eglReleaseThread && !eglReleaseThread())
+       Log::error("eglReleaseThread failed\n");
+}
+
 bool
 GLStateEGL::init_display(void* native_display, GLVisualConfig& visual_config)
 {
+#if defined(WIN32)
+    if (!egl_lib_.open("libEGL.dll")) {
+#else
+    if (!egl_lib_.open_from_alternatives({"libEGL.so", "libEGL.so.1" })) {
+#endif
+        Log::error("Error loading EGL library\n");
+        return false;
+    }
+
+    if (gladLoadEGLUserPtr(EGL_NO_DISPLAY, load_egl_func, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
     native_display_ = reinterpret_cast<EGLNativeDisplayType>(native_display);
     requested_visual_config_ = visual_config;
 
@@ -304,20 +336,28 @@
     return gotValidSurface();
 }
 
-void
+bool
 GLStateEGL::init_gl_extensions()
 {
 #if GLMARK2_USE_GLESv2
+    if (!gladLoadGLES2UserPtr(load_proc, this)) {
+        Log::error("Loading GLESv2 entry points failed.");
+        return false;
+    }
+
     if (GLExtensions::support("GL_OES_mapbuffer")) {
-        GLExtensions::MapBuffer =
-            reinterpret_cast<PFNGLMAPBUFFEROESPROC>(eglGetProcAddress("glMapBufferOES"));
-        GLExtensions::UnmapBuffer =
-            reinterpret_cast<PFNGLUNMAPBUFFEROESPROC>(eglGetProcAddress("glUnmapBufferOES"));
+        GLExtensions::MapBuffer = glMapBufferOES;
+        GLExtensions::UnmapBuffer = glUnmapBufferOES;
     }
 #elif GLMARK2_USE_GL
+    if (!gladLoadGLUserPtr(load_proc, this)) {
+        Log::error("Loading GL entry points failed.");
+        return false;
+    }
     GLExtensions::MapBuffer = glMapBuffer;
     GLExtensions::UnmapBuffer = glUnmapBuffer;
 #endif
+    return true;
 }
 
 bool
@@ -347,7 +387,9 @@
         Log::info("** Failed to set swap interval. Results may be bounded above by refresh rate.\n");
     }
 
-    init_gl_extensions();
+    if (!init_gl_extensions()) {
+        return false;
+    }
 
     return true;
 }
@@ -379,7 +421,7 @@
 }
 
 bool
-GLStateEGL::gotNativeConfig(int& vid)
+GLStateEGL::gotNativeConfig(intptr_t& vid)
 {
     if (!gotValidConfig())
         return false;
@@ -409,17 +451,64 @@
  * GLStateEGL private methods *
  *****************************/
 
+#ifdef GLMARK2_USE_X11
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_X11_KHR
+#elif  GLMARK2_USE_WAYLAND
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_WAYLAND_KHR
+#elif  GLMARK2_USE_DRM
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_GBM_KHR
+#elif  GLMARK2_USE_MIR
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_MIR_KHR
+#else
+// Platforms not in the above platform enums fall back to eglGetDisplay.
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM 0
+#endif
+
 bool
 GLStateEGL::gotValidDisplay()
 {
     if (egl_display_)
         return true;
 
-    egl_display_ = eglGetDisplay(native_display_);
+    char const * __restrict const supported_extensions =
+        eglQueryString(EGL_NO_DISPLAY, EGL_EXTENSIONS);
+
+    if (GLMARK2_NATIVE_EGL_DISPLAY_ENUM != 0 && supported_extensions
+        && strstr(supported_extensions, "EGL_EXT_platform_base"))
+    {
+        Log::debug("Using eglGetPlatformDisplayEXT()\n");
+        PFNEGLGETPLATFORMDISPLAYEXTPROC get_platform_display =
+            reinterpret_cast<PFNEGLGETPLATFORMDISPLAYEXTPROC>(
+                eglGetProcAddress("eglGetPlatformDisplayEXT"));
+
+        if (get_platform_display != nullptr) {
+            egl_display_ = get_platform_display(
+                GLMARK2_NATIVE_EGL_DISPLAY_ENUM,
+                reinterpret_cast<void*>(native_display_),
+                nullptr);
+        }
+
+        if (!egl_display_) {
+            Log::debug("eglGetPlatformDisplayEXT() failed with error: 0x%x\n",
+                       eglGetError());
+        }
+    }
+    else
+    {
+        Log::debug("eglGetPlatformDisplayEXT() seems unsupported\n");
+    }
+
+    /* Just in case get_platform_display failed... */
+    if (!egl_display_) {
+        Log::debug("Falling back to eglGetDisplay()\n");
+        egl_display_ = eglGetDisplay(native_display_);
+    }
+
     if (!egl_display_) {
         Log::error("eglGetDisplay() failed with error: 0x%x\n", eglGetError());
         return false;
     }
+
     int egl_major(-1);
     int egl_minor(-1);
     if (!eglInitialize(egl_display_, &egl_major, &egl_minor)) {
@@ -428,16 +517,33 @@
         return false;
     }
 
+    /* Reinitialize GLAD with a known display */
+    if (gladLoadEGLUserPtr(egl_display_, load_egl_func, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
 #if GLMARK2_USE_GLESv2
     EGLenum apiType(EGL_OPENGL_ES_API);
+#if defined(WIN32)
+    std::initializer_list<const char *> libNames = { "libGLESv2.dll" };
+#else
+    std::initializer_list<const char *> libNames = { "libGLESv2.so", "libGLESv2.so.2" };
+#endif
 #elif GLMARK2_USE_GL
     EGLenum apiType(EGL_OPENGL_API);
+    std::initializer_list<const char *> libNames = { "libGL.so", "libGL.so.1" };
 #endif
     if (!eglBindAPI(apiType)) {
         Log::error("Failed to bind api EGL_OPENGL_ES_API\n");
         return false;
     }
 
+    if (!gl_lib_.open_from_alternatives(libNames)) {
+        Log::error("Error loading GL library\n");
+        return false;
+    }
+
     return true;
 }
 
@@ -617,3 +723,16 @@
     return true;
 }
 
+GLADapiproc
+GLStateEGL::load_proc(void *userptr, const char* name)
+{
+    if (eglGetProcAddress) {
+        GLADapiproc sym = reinterpret_cast<GLADapiproc>(eglGetProcAddress(name));
+        if (sym) {
+            return sym;
+        }
+    }
+
+    GLStateEGL* state = reinterpret_cast<GLStateEGL*>(userptr);
+    return reinterpret_cast<GLADapiproc>(state->gl_lib_.load(name));
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-egl.h glmark2-2021.02/android/jni/src/gl-state-egl.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-egl.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-egl.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -23,9 +23,10 @@
 #define GLMARK2_GL_STATE_EGL_H_
 
 #include <vector>
-#include <EGL/egl.h>
+#include <glad/egl.h>
 #include "gl-state.h"
 #include "gl-visual-config.h"
+#include "shared-library.h"
 
 class EglConfig
 {
@@ -124,12 +125,17 @@
     EGLSurface egl_surface_;
     GLVisualConfig requested_visual_config_;
     EglConfig best_config_;
+    SharedLibrary egl_lib_;
+    SharedLibrary gl_lib_;
     bool gotValidDisplay();
     bool gotValidConfig();
     bool gotValidSurface();
     bool gotValidContext();
     void get_glvisualconfig(EGLConfig config, GLVisualConfig& visual_config);
     EGLConfig select_best_config(std::vector<EGLConfig>& configs);
+
+    static GLADapiproc load_proc(void *userptr, const char* name);
+
 public:
     GLStateEGL() :
         native_display_(0),
@@ -138,15 +144,16 @@
         egl_config_(0),
         egl_context_(0),
         egl_surface_(0) {}
+    ~GLStateEGL();
 
     bool init_display(void* native_display, GLVisualConfig& config_pref);
     bool init_surface(void* native_window);
-    void init_gl_extensions();
+    bool init_gl_extensions();
     bool valid();
     bool reset();
     void swap();
     // Performs a config search, returning a native visual ID on success
-    bool gotNativeConfig(int& vid);
+    bool gotNativeConfig(intptr_t& vid);
     void getVisualConfig(GLVisualConfig& vc);
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-glx.cpp glmark2-2021.02/android/jni/src/gl-state-glx.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-glx.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-glx.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,13 +26,6 @@
 
 #include <climits>
 
-namespace
-{
-PFNGLXSWAPINTERVALEXTPROC glXSwapIntervalEXT_;
-PFNGLXSWAPINTERVALMESAPROC glXSwapIntervalMESA_;
-PFNGLXGETSWAPINTERVALMESAPROC glXGetSwapIntervalMESA_;
-}
-
 /******************
  * Public methods *
  ******************/
@@ -43,6 +36,12 @@
     xdpy_ = reinterpret_cast<Display*>(native_display);
     requested_visual_config_ = visual_config;
 
+    if (!lib_.open_from_alternatives({"libGL.so", "libGL.so.1"})) {
+        Log::error("Failed to load libGL\n");
+        return false;
+    }
+
+    gladLoadGLXUserPtr(xdpy_, DefaultScreen(xdpy_), load_proc, this);
     return (xdpy_ != 0);
 }
 
@@ -54,11 +53,12 @@
     return (xwin_ != 0);
 }
 
-void
+bool
 GLStateGLX::init_gl_extensions()
 {
     GLExtensions::MapBuffer = glMapBuffer;
     GLExtensions::UnmapBuffer = glUnmapBuffer;
+    return true;
 }
 
 bool
@@ -80,20 +80,26 @@
         return false;
     }
 
-    init_gl_extensions();
+    if (gladLoadGLUserPtr(load_proc, this) == 0) {
+        Log::error("Failed to load GL entry points\n");
+        return false;
+    }
+
+    if (!init_gl_extensions())
+        return false;
 
     unsigned int desired_swap(0);
     unsigned int actual_swap(-1);
-    if (glXSwapIntervalEXT_) {
-        glXSwapIntervalEXT_(xdpy_, xwin_, desired_swap);
+    if (glXSwapIntervalEXT) {
+        glXSwapIntervalEXT(xdpy_, xwin_, desired_swap);
         glXQueryDrawable(xdpy_, xwin_, GLX_SWAP_INTERVAL_EXT, &actual_swap);
         if (actual_swap == desired_swap)
             return true;
     }
 
-    if (glXSwapIntervalMESA_) {
-        glXSwapIntervalMESA_(desired_swap);
-        actual_swap = glXGetSwapIntervalMESA_();
+    if (glXSwapIntervalMESA) {
+        glXSwapIntervalMESA(desired_swap);
+        actual_swap = glXGetSwapIntervalMESA();
         if (actual_swap == desired_swap)
             return true;
     }
@@ -123,7 +129,7 @@
 }
 
 bool
-GLStateGLX::gotNativeConfig(int& vid)
+GLStateGLX::gotNativeConfig(intptr_t& vid)
 {
     if (!ensure_glx_fbconfig())
         return false;
@@ -186,31 +192,7 @@
      * GLX_SGI_swap_control is not enough because it doesn't allow 0 as a valid
      * value (i.e. you can't turn off VSync).
      */
-    if (extString.find("GLX_EXT_swap_control") != std::string::npos) {
-        glXSwapIntervalEXT_ =
-            reinterpret_cast<PFNGLXSWAPINTERVALEXTPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXSwapIntervalEXT")
-                )
-            );
-    }
-    else if (extString.find("GLX_MESA_swap_control") != std::string::npos) {
-        glXSwapIntervalMESA_ =
-            reinterpret_cast<PFNGLXSWAPINTERVALMESAPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXSwapIntervalMESA")
-                )
-            );
-        glXGetSwapIntervalMESA_ =
-            reinterpret_cast<PFNGLXGETSWAPINTERVALMESAPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXGetSwapIntervalMESA")
-                )
-            );
-    }
-
-
-    if (!glXSwapIntervalEXT_ && !glXSwapIntervalMESA_) {
+    if (!glXSwapIntervalEXT && !glXSwapIntervalMESA) {
         Log::info("** GLX does not support GLX_EXT_swap_control or GLX_MESA_swap_control!\n");
     }
 }
@@ -346,3 +328,18 @@
 
     return best_config;
 }
+
+GLADapiproc
+GLStateGLX::load_proc(void *userptr, const char* name)
+{
+    if (glXGetProcAddress) {
+        const GLubyte* bytes = reinterpret_cast<const GLubyte*>(name);
+        GLADapiproc sym = glXGetProcAddress(bytes);
+        if (sym) {
+            return sym;
+        }
+    }
+
+    GLStateGLX* state = reinterpret_cast<GLStateGLX*>(userptr);
+    return reinterpret_cast<GLADapiproc>(state->lib_.load(name));
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-glx.h glmark2-2021.02/android/jni/src/gl-state-glx.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-glx.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-glx.h	2021-04-25 01:47:22.000000000 +0800
@@ -26,13 +26,11 @@
 #include "gl-state.h"
 #include "gl-visual-config.h"
 #include "gl-headers.h"
+#include "shared-library.h"
 
 #include <vector>
 
-#include <X11/Xlib.h>
-#define GLX_GLXEXT_PROTOTYPES
-#include <GL/glx.h>
-#include <GL/glxext.h>
+#include <glad/glx.h>
 
 class GLStateGLX : public GLState
 {
@@ -43,10 +41,10 @@
     bool valid();
     bool init_display(void* native_display, GLVisualConfig& config_pref);
     bool init_surface(void* native_window);
-    void init_gl_extensions();
+    bool init_gl_extensions();
     bool reset();
     void swap();
-    bool gotNativeConfig(int& vid);
+    bool gotNativeConfig(intptr_t& vid);
     void getVisualConfig(GLVisualConfig& vc);
 
 private:
@@ -57,11 +55,14 @@
     void get_glvisualconfig_glx(GLXFBConfig config, GLVisualConfig &visual_config);
     GLXFBConfig select_best_config(std::vector<GLXFBConfig> configs);
 
+    static GLADapiproc load_proc(void* userptr, const char* name);
+
     Display* xdpy_;
     Window xwin_;
     GLXFBConfig glx_fbconfig_;
     GLXContext glx_context_;
     GLVisualConfig requested_visual_config_;
+    SharedLibrary lib_;
 };
 
 #endif /* GLMARK2_GL_STATE_GLX_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state.h glmark2-2021.02/android/jni/src/gl-state.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,6 +22,8 @@
 #ifndef GLMARK2_GL_STATE_H_
 #define GLMARK2_GL_STATE_H_
 
+#include <stdint.h>
+
 class GLVisualConfig;
 
 class GLState
@@ -31,11 +33,11 @@
 
     virtual bool init_display(void *native_display, GLVisualConfig& config_pref) = 0;
     virtual bool init_surface(void *native_window) = 0;
-    virtual void init_gl_extensions() = 0;
+    virtual bool init_gl_extensions() = 0;
     virtual bool valid() = 0;
     virtual bool reset() = 0;
     virtual void swap() = 0;
-    virtual bool gotNativeConfig(int& vid) = 0;
+    virtual bool gotNativeConfig(intptr_t& vid) = 0;
     virtual void getVisualConfig(GLVisualConfig& vc) = 0;
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-wgl.cpp glmark2-2021.02/android/jni/src/gl-state-wgl.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-wgl.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-wgl.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,178 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+#include "gl-state-wgl.h"
+
+#include <glad/wgl.h>
+#include "gl-headers.h"
+#include "log.h"
+#include "options.h"
+
+/******************
+ * Public methods *
+ ******************/
+
+GLStateWGL::GLStateWGL() : hdc_(nullptr), wgl_context_(nullptr), get_proc_addr_(nullptr) {}
+
+GLStateWGL::~GLStateWGL()
+{
+    if (wgl_context_) {
+        wglDeleteContext(wgl_context_);
+        wgl_context_ = nullptr;
+    }
+}
+
+bool
+GLStateWGL::init_display(void* native_display, GLVisualConfig& /*visual_config*/)
+{
+    if (!wgl_library_.open("opengl32.dll")) {
+        Log::error("Failed to load opengl32.dll\n");
+        return false;
+    }
+
+    hdc_ = reinterpret_cast<HDC>(native_display);
+
+    PIXELFORMATDESCRIPTOR pixel_format_desc = {};
+    pixel_format_desc.nSize = sizeof(pixel_format_desc);
+    pixel_format_desc.nVersion = 1;
+    pixel_format_desc.dwFlags =
+        PFD_DRAW_TO_WINDOW | PFD_GENERIC_ACCELERATED | PFD_SUPPORT_OPENGL | PFD_DOUBLEBUFFER;
+    pixel_format_desc.iPixelType = PFD_TYPE_RGBA;
+    pixel_format_desc.cColorBits = 24;
+    pixel_format_desc.cAlphaBits = 8;
+    pixel_format_desc.cDepthBits = 24;
+    pixel_format_desc.cStencilBits = 8;
+    pixel_format_desc.iLayerType = PFD_MAIN_PLANE;
+
+    int pixelFormat = ChoosePixelFormat(hdc_, &pixel_format_desc);
+    if (pixelFormat == 0) {
+        Log::error("Could not find a compatible pixel format\n");
+        return false;
+    }
+
+    if (SetPixelFormat(hdc_, pixelFormat, &pixel_format_desc) != TRUE) {
+        Log::error("Failed to set the pixel format\n");
+        return false;
+    }
+
+    wgl_context_ = wglCreateContext(hdc_);
+    if (!wgl_context_) {
+        Log::error("Failed to create a WGL context\n");
+        return false;
+    }
+
+    if (wglMakeCurrent(hdc_, wgl_context_) == FALSE) {
+        Log::error("Error during wglMakeCurrent\n");
+        return false;
+    }
+
+    get_proc_addr_ = wgl_library_.load("wglGetProcAddress");
+    if (!get_proc_addr_) {
+        Log::error("Failed to find wglGetProcAddress\n");
+        return false;
+    }
+
+    if (gladLoadGLUserPtr(load_proc, this) == 0) {
+        Log::error("Failed to load GL entry points\n");
+        return false;
+    }
+
+    if (gladLoadWGLUserPtr(hdc_, load_proc, this) == 0) {
+        Log::error("Failed to load WGL entry points\n");
+        return false;
+    }
+
+    return true;
+}
+
+bool
+GLStateWGL::init_surface(void* /*native_window*/)
+{
+    return true;
+}
+
+bool
+GLStateWGL::init_gl_extensions()
+{
+    GLExtensions::MapBuffer = glMapBuffer;
+    GLExtensions::UnmapBuffer = glUnmapBuffer;
+    return true;
+}
+
+bool
+GLStateWGL::valid()
+{
+    if (!wglSwapIntervalEXT || wglSwapIntervalEXT(0) == FALSE) {
+        Log::info("** Failed to set swap interval. Results may be bounded above by refresh rate.\n");
+    }
+
+    return wgl_context_ != nullptr;
+}
+
+
+bool
+GLStateWGL::reset()
+{
+    return true;
+}
+
+void
+GLStateWGL::swap()
+{
+    if (SwapBuffers(hdc_) == FALSE) {
+        Log::error("Error during SwapBuffers\n");
+    }
+}
+
+bool
+GLStateWGL::gotNativeConfig(intptr_t& vid)
+{
+    vid = 1;
+    return true;
+}
+
+void
+GLStateWGL::getVisualConfig(GLVisualConfig& vc)
+{
+    vc.buffer = 32;
+    vc.red = 8;
+    vc.green = 8;
+    vc.blue = 8;
+    vc.alpha = 8;
+    vc.depth = 24;
+    vc.stencil = 8;
+}
+
+/*******************
+ * Private methods *
+ *******************/
+
+GLStateWGL::api_proc GLStateWGL::load_proc(void *userptr, const char *name)
+{
+    GLStateWGL* state = reinterpret_cast<GLStateWGL*>(userptr);
+    auto get_proc_addr = reinterpret_cast<decltype(&wglGetProcAddress)>(state->get_proc_addr_);
+    auto proc = reinterpret_cast<GLStateWGL::api_proc>(get_proc_addr(name));
+    if (proc) {
+        return proc;
+    }
+    return reinterpret_cast<GLStateWGL::api_proc>(state->wgl_library_.load(name));
+}
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-wgl.h glmark2-2021.02/android/jni/src/gl-state-wgl.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-state-wgl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-state-wgl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,54 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+#ifndef GLMARK2_GL_STATE_WGL_H_
+#define GLMARK2_GL_STATE_WGL_H_
+
+#include "gl-state.h"
+#include "shared-library.h"
+#include <windows.h>
+
+class GLStateWGL : public GLState
+{
+public:
+    GLStateWGL();
+    ~GLStateWGL();
+
+    bool init_display(void* native_display, GLVisualConfig& config_pref);
+    bool init_surface(void* native_window);
+    bool init_gl_extensions();
+    bool valid();
+    bool reset();
+    void swap();
+    bool gotNativeConfig(intptr_t& vid);
+    void getVisualConfig(GLVisualConfig& vc);
+
+private:
+    using api_proc = void (*)();
+    static api_proc load_proc(void *userptr, const char *name);
+
+    SharedLibrary wgl_library_;
+    HDC hdc_;
+    HGLRC wgl_context_;
+    void* get_proc_addr_;
+};
+
+#endif // GLMARK2_GL_STATE_WGL_H_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-visual-config.cpp glmark2-2021.02/android/jni/src/gl-visual-config.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/gl-visual-config.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/gl-visual-config.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -108,6 +108,14 @@
         /* Reward exact matches with the maximum per component score */
         score = MAXIMUM_COMPONENT_SCORE;
     }
+    else if (component > 8 && target <= 8 && scale > 1)
+    {
+        /* Penalize RGBA component widths larger than 8, since they are
+         * unlikely to be what the users want or properly supported for
+         * display. Such widths can still be used, but only if explicitly
+         * requested. */
+        score = UNACCEPTABLE_COMPONENT_PENALTY;
+    }
     else
     {
         /* 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/image-reader.cpp glmark2-2021.02/android/jni/src/image-reader.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/image-reader.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/image-reader.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -120,7 +120,7 @@
 
     Log::debug("Reading PNG file %s\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> is_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> is_ptr(Util::get_resource(filename));
     if (!(*is_ptr)) {
         Log::error("Cannot open file %s!\n", filename.c_str());
         return false;
@@ -268,7 +268,7 @@
             size_t n = static_cast<size_t>(num_bytes);
             while (n > src->pub.bytes_in_buffer) {
                 n -= src->pub.bytes_in_buffer;
-                (*src->fill_input_buffer)(cinfo);
+                src->fill_input_buffer(cinfo);
             }
             src->pub.next_input_byte += n;
             src->pub.bytes_in_buffer -= n;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/dirent.h glmark2-2021.02/android/jni/src/include/dirent.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/dirent.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/include/dirent.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1224 @@
+/*
+ * Dirent interface for Microsoft Visual Studio
+ *
+ * Copyright (C) 2006-2012 Toni Ronkko
+ * This file is part of dirent.  Dirent may be freely distributed
+ * under the MIT license.  For all details and documentation, see
+ * https://github.com/tronkko/dirent
+ */
+#ifndef DIRENT_H
+#define DIRENT_H
+ 
+/*
+ * Include windows.h without Windows Sockets 1.1 to prevent conflicts with
+ * Windows Sockets 2.0.
+ */
+#ifndef WIN32_LEAN_AND_MEAN
+#   define WIN32_LEAN_AND_MEAN
+#endif
+#include <windows.h>
+
+#include <errno.h>
+#include <malloc.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <wchar.h>
+
+/* Indicates that d_type field is available in dirent structure */
+#define _DIRENT_HAVE_D_TYPE
+
+/* Indicates that d_namlen field is available in dirent structure */
+#define _DIRENT_HAVE_D_NAMLEN
+
+/* Entries missing from MSVC 6.0 */
+#if !defined(FILE_ATTRIBUTE_DEVICE)
+#   define FILE_ATTRIBUTE_DEVICE 0x40
+#endif
+
+/* File type and permission flags for stat(), general mask */
+#if !defined(S_IFMT)
+#   define S_IFMT _S_IFMT
+#endif
+
+/* Directory bit */
+#if !defined(S_IFDIR)
+#   define S_IFDIR _S_IFDIR
+#endif
+
+/* Character device bit */
+#if !defined(S_IFCHR)
+#   define S_IFCHR _S_IFCHR
+#endif
+
+/* Pipe bit */
+#if !defined(S_IFFIFO)
+#   define S_IFFIFO _S_IFFIFO
+#endif
+
+/* Regular file bit */
+#if !defined(S_IFREG)
+#   define S_IFREG _S_IFREG
+#endif
+
+/* Read permission */
+#if !defined(S_IREAD)
+#   define S_IREAD _S_IREAD
+#endif
+
+/* Write permission */
+#if !defined(S_IWRITE)
+#   define S_IWRITE _S_IWRITE
+#endif
+
+/* Execute permission */
+#if !defined(S_IEXEC)
+#   define S_IEXEC _S_IEXEC
+#endif
+
+/* Pipe */
+#if !defined(S_IFIFO)
+#   define S_IFIFO _S_IFIFO
+#endif
+
+/* Block device */
+#if !defined(S_IFBLK)
+#   define S_IFBLK 0
+#endif
+
+/* Link */
+#if !defined(S_IFLNK)
+#   define S_IFLNK 0
+#endif
+
+/* Socket */
+#if !defined(S_IFSOCK)
+#   define S_IFSOCK 0
+#endif
+
+/* Read user permission */
+#if !defined(S_IRUSR)
+#   define S_IRUSR S_IREAD
+#endif
+
+/* Write user permission */
+#if !defined(S_IWUSR)
+#   define S_IWUSR S_IWRITE
+#endif
+
+/* Execute user permission */
+#if !defined(S_IXUSR)
+#   define S_IXUSR 0
+#endif
+
+/* Read group permission */
+#if !defined(S_IRGRP)
+#   define S_IRGRP 0
+#endif
+
+/* Write group permission */
+#if !defined(S_IWGRP)
+#   define S_IWGRP 0
+#endif
+
+/* Execute group permission */
+#if !defined(S_IXGRP)
+#   define S_IXGRP 0
+#endif
+
+/* Read others permission */
+#if !defined(S_IROTH)
+#   define S_IROTH 0
+#endif
+
+/* Write others permission */
+#if !defined(S_IWOTH)
+#   define S_IWOTH 0
+#endif
+
+/* Execute others permission */
+#if !defined(S_IXOTH)
+#   define S_IXOTH 0
+#endif
+
+/* Maximum length of file name */
+#if !defined(PATH_MAX)
+#   define PATH_MAX MAX_PATH
+#endif
+#if !defined(FILENAME_MAX)
+#   define FILENAME_MAX MAX_PATH
+#endif
+#if !defined(NAME_MAX)
+#   define NAME_MAX FILENAME_MAX
+#endif
+
+/* File type flags for d_type */
+#define DT_UNKNOWN 0
+#define DT_REG S_IFREG
+#define DT_DIR S_IFDIR
+#define DT_FIFO S_IFIFO
+#define DT_SOCK S_IFSOCK
+#define DT_CHR S_IFCHR
+#define DT_BLK S_IFBLK
+#define DT_LNK S_IFLNK
+
+/* Macros for converting between st_mode and d_type */
+#define IFTODT(mode) ((mode) & S_IFMT)
+#define DTTOIF(type) (type)
+
+/*
+ * File type macros.  Note that block devices, sockets and links cannot be
+ * distinguished on Windows and the macros S_ISBLK, S_ISSOCK and S_ISLNK are
+ * only defined for compatibility.  These macros should always return false
+ * on Windows.
+ */
+#if !defined(S_ISFIFO)
+#   define S_ISFIFO(mode) (((mode) & S_IFMT) == S_IFIFO)
+#endif
+#if !defined(S_ISDIR)
+#   define S_ISDIR(mode) (((mode) & S_IFMT) == S_IFDIR)
+#endif
+#if !defined(S_ISREG)
+#   define S_ISREG(mode) (((mode) & S_IFMT) == S_IFREG)
+#endif
+#if !defined(S_ISLNK)
+#   define S_ISLNK(mode) (((mode) & S_IFMT) == S_IFLNK)
+#endif
+#if !defined(S_ISSOCK)
+#   define S_ISSOCK(mode) (((mode) & S_IFMT) == S_IFSOCK)
+#endif
+#if !defined(S_ISCHR)
+#   define S_ISCHR(mode) (((mode) & S_IFMT) == S_IFCHR)
+#endif
+#if !defined(S_ISBLK)
+#   define S_ISBLK(mode) (((mode) & S_IFMT) == S_IFBLK)
+#endif
+
+/* Return the exact length of the file name without zero terminator */
+#define _D_EXACT_NAMLEN(p) ((p)->d_namlen)
+
+/* Return the maximum size of a file name */
+#define _D_ALLOC_NAMLEN(p) ((PATH_MAX)+1)
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/* Wide-character version */
+struct _wdirent {
+    /* Always zero */
+    long d_ino;
+
+    /* File position within stream */
+    long d_off;
+
+    /* Structure size */
+    unsigned short d_reclen;
+
+    /* Length of name without \0 */
+    size_t d_namlen;
+
+    /* File type */
+    int d_type;
+
+    /* File name */
+    wchar_t d_name[PATH_MAX+1];
+};
+typedef struct _wdirent _wdirent;
+
+struct _WDIR {
+    /* Current directory entry */
+    struct _wdirent ent;
+
+    /* Private file data */
+    WIN32_FIND_DATAW data;
+
+    /* True if data is valid */
+    int cached;
+
+    /* Win32 search handle */
+    HANDLE handle;
+
+    /* Initial directory name */
+    wchar_t *patt;
+};
+typedef struct _WDIR _WDIR;
+
+/* Multi-byte character version */
+struct dirent {
+    /* Always zero */
+    long d_ino;
+
+    /* File position within stream */
+    long d_off;
+
+    /* Structure size */
+    unsigned short d_reclen;
+
+    /* Length of name without \0 */
+    size_t d_namlen;
+
+    /* File type */
+    int d_type;
+
+    /* File name */
+    char d_name[PATH_MAX+1];
+};
+typedef struct dirent dirent;
+
+struct DIR {
+    struct dirent ent;
+    struct _WDIR *wdirp;
+};
+typedef struct DIR DIR;
+
+
+/* Dirent functions */
+static DIR *opendir (const char *dirname);
+static _WDIR *_wopendir (const wchar_t *dirname);
+
+static struct dirent *readdir (DIR *dirp);
+static struct _wdirent *_wreaddir (_WDIR *dirp);
+
+static int readdir_r(
+    DIR *dirp, struct dirent *entry, struct dirent **result);
+static int _wreaddir_r(
+    _WDIR *dirp, struct _wdirent *entry, struct _wdirent **result);
+
+static int closedir (DIR *dirp);
+static int _wclosedir (_WDIR *dirp);
+
+static void rewinddir (DIR* dirp);
+static void _wrewinddir (_WDIR* dirp);
+
+static int scandir (const char *dirname, struct dirent ***namelist,
+    int (*filter)(const struct dirent*),
+    int (*compare)(const struct dirent**, const struct dirent**));
+
+static int alphasort (const struct dirent **a, const struct dirent **b);
+
+static int versionsort (const struct dirent **a, const struct dirent **b);
+
+
+/* For compatibility with Symbian */
+#define wdirent _wdirent
+#define WDIR _WDIR
+#define wopendir _wopendir
+#define wreaddir _wreaddir
+#define wclosedir _wclosedir
+#define wrewinddir _wrewinddir
+
+
+/* Internal utility functions */
+static WIN32_FIND_DATAW *dirent_first (_WDIR *dirp);
+static WIN32_FIND_DATAW *dirent_next (_WDIR *dirp);
+
+static int dirent_mbstowcs_s(
+    size_t *pReturnValue,
+    wchar_t *wcstr,
+    size_t sizeInWords,
+    const char *mbstr,
+    size_t count);
+
+static int dirent_wcstombs_s(
+    size_t *pReturnValue,
+    char *mbstr,
+    size_t sizeInBytes,
+    const wchar_t *wcstr,
+    size_t count);
+
+static void dirent_set_errno (int error);
+
+
+/*
+ * Open directory stream DIRNAME for read and return a pointer to the
+ * internal working area that is used to retrieve individual directory
+ * entries.
+ */
+static _WDIR*
+_wopendir(
+    const wchar_t *dirname)
+{
+    _WDIR *dirp = NULL;
+    int error;
+
+    /* Must have directory name */
+    if (dirname == NULL  ||  dirname[0] == '\0') {
+        dirent_set_errno (ENOENT);
+        return NULL;
+    }
+
+    /* Allocate new _WDIR structure */
+    dirp = (_WDIR*) malloc (sizeof (struct _WDIR));
+    if (dirp != NULL) {
+        DWORD n;
+
+        /* Reset _WDIR structure */
+        dirp->handle = INVALID_HANDLE_VALUE;
+        dirp->patt = NULL;
+        dirp->cached = 0;
+
+        /* Compute the length of full path plus zero terminator
+         *
+         * Note that on WinRT there's no way to convert relative paths
+         * into absolute paths, so just assume it is an absolute path.
+         */
+#       if defined(WINAPI_FAMILY) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP)
+            n = wcslen(dirname);
+#       else
+            n = GetFullPathNameW (dirname, 0, NULL, NULL);
+#       endif
+
+        /* Allocate room for absolute directory name and search pattern */
+        dirp->patt = (wchar_t*) malloc (sizeof (wchar_t) * n + 16);
+        if (dirp->patt) {
+
+            /*
+             * Convert relative directory name to an absolute one.  This
+             * allows rewinddir() to function correctly even when current
+             * working directory is changed between opendir() and rewinddir().
+             *
+             * Note that on WinRT there's no way to convert relative paths
+             * into absolute paths, so just assume it is an absolute path.
+             */
+#           if defined(WINAPI_FAMILY) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP)
+                wcsncpy_s(dirp->patt, n+1, dirname, n);
+#           else
+                n = GetFullPathNameW (dirname, n, dirp->patt, NULL);
+#           endif
+            if (n > 0) {
+                wchar_t *p;
+
+                /* Append search pattern \* to the directory name */
+                p = dirp->patt + n;
+                if (dirp->patt < p) {
+                    switch (p[-1]) {
+                    case '\\':
+                    case '/':
+                    case ':':
+                        /* Directory ends in path separator, e.g. c:\temp\ */
+                        /*NOP*/;
+                        break;
+
+                    default:
+                        /* Directory name doesn't end in path separator */
+                        *p++ = '\\';
+                    }
+                }
+                *p++ = '*';
+                *p = '\0';
+
+                /* Open directory stream and retrieve the first entry */
+                if (dirent_first (dirp)) {
+                    /* Directory stream opened successfully */
+                    error = 0;
+                } else {
+                    /* Cannot retrieve first entry */
+                    error = 1;
+                    dirent_set_errno (ENOENT);
+                }
+
+            } else {
+                /* Cannot retrieve full path name */
+                dirent_set_errno (ENOENT);
+                error = 1;
+            }
+
+        } else {
+            /* Cannot allocate memory for search pattern */
+            error = 1;
+        }
+
+    } else {
+        /* Cannot allocate _WDIR structure */
+        error = 1;
+    }
+
+    /* Clean up in case of error */
+    if (error  &&  dirp) {
+        _wclosedir (dirp);
+        dirp = NULL;
+    }
+
+    return dirp;
+}
+
+/*
+ * Read next directory entry.
+ *
+ * Returns pointer to static directory entry which may be overwritten by
+ * subsequent calls to _wreaddir().
+ */
+static struct _wdirent*
+_wreaddir(
+    _WDIR *dirp)
+{
+    struct _wdirent *entry;
+
+    /*
+     * Read directory entry to buffer.  We can safely ignore the return value
+     * as entry will be set to NULL in case of error.
+     */
+    (void) _wreaddir_r (dirp, &dirp->ent, &entry);
+
+    /* Return pointer to statically allocated directory entry */
+    return entry;
+}
+
+/*
+ * Read next directory entry.
+ *
+ * Returns zero on success.  If end of directory stream is reached, then sets
+ * result to NULL and returns zero.
+ */
+static int
+_wreaddir_r(
+    _WDIR *dirp,
+    struct _wdirent *entry,
+    struct _wdirent **result)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Read next directory entry */
+    datap = dirent_next (dirp);
+    if (datap) {
+        size_t n;
+        DWORD attr;
+
+        /*
+         * Copy file name as wide-character string.  If the file name is too
+         * long to fit in to the destination buffer, then truncate file name
+         * to PATH_MAX characters and zero-terminate the buffer.
+         */
+        n = 0;
+        while (n < PATH_MAX  &&  datap->cFileName[n] != 0) {
+            entry->d_name[n] = datap->cFileName[n];
+            n++;
+        }
+        entry->d_name[n] = 0;
+
+        /* Length of file name excluding zero terminator */
+        entry->d_namlen = n;
+
+        /* File type */
+        attr = datap->dwFileAttributes;
+        if ((attr & FILE_ATTRIBUTE_DEVICE) != 0) {
+            entry->d_type = DT_CHR;
+        } else if ((attr & FILE_ATTRIBUTE_DIRECTORY) != 0) {
+            entry->d_type = DT_DIR;
+        } else {
+            entry->d_type = DT_REG;
+        }
+
+        /* Reset dummy fields */
+        entry->d_ino = 0;
+        entry->d_off = 0;
+        entry->d_reclen = sizeof (struct _wdirent);
+
+        /* Set result address */
+        *result = entry;
+
+    } else {
+
+        /* Return NULL to indicate end of directory */
+        *result = NULL;
+
+    }
+
+    return /*OK*/0;
+}
+
+/*
+ * Close directory stream opened by opendir() function.  This invalidates the
+ * DIR structure as well as any directory entry read previously by
+ * _wreaddir().
+ */
+static int
+_wclosedir(
+    _WDIR *dirp)
+{
+    int ok;
+    if (dirp) {
+
+        /* Release search handle */
+        if (dirp->handle != INVALID_HANDLE_VALUE) {
+            FindClose (dirp->handle);
+            dirp->handle = INVALID_HANDLE_VALUE;
+        }
+
+        /* Release search pattern */
+        if (dirp->patt) {
+            free (dirp->patt);
+            dirp->patt = NULL;
+        }
+
+        /* Release directory structure */
+        free (dirp);
+        ok = /*success*/0;
+
+    } else {
+
+        /* Invalid directory stream */
+        dirent_set_errno (EBADF);
+        ok = /*failure*/-1;
+
+    }
+    return ok;
+}
+
+/*
+ * Rewind directory stream such that _wreaddir() returns the very first
+ * file name again.
+ */
+static void
+_wrewinddir(
+    _WDIR* dirp)
+{
+    if (dirp) {
+        /* Release existing search handle */
+        if (dirp->handle != INVALID_HANDLE_VALUE) {
+            FindClose (dirp->handle);
+        }
+
+        /* Open new search handle */
+        dirent_first (dirp);
+    }
+}
+
+/* Get first directory entry (internal) */
+static WIN32_FIND_DATAW*
+dirent_first(
+    _WDIR *dirp)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Open directory and retrieve the first entry */
+    dirp->handle = FindFirstFileExW(
+        dirp->patt, FindExInfoStandard, &dirp->data,
+        FindExSearchNameMatch, NULL, 0);
+    if (dirp->handle != INVALID_HANDLE_VALUE) {
+
+        /* a directory entry is now waiting in memory */
+        datap = &dirp->data;
+        dirp->cached = 1;
+
+    } else {
+
+        /* Failed to re-open directory: no directory entry in memory */
+        dirp->cached = 0;
+        datap = NULL;
+
+    }
+    return datap;
+}
+
+/*
+ * Get next directory entry (internal).
+ *
+ * Returns
+ */
+static WIN32_FIND_DATAW*
+dirent_next(
+    _WDIR *dirp)
+{
+    WIN32_FIND_DATAW *p;
+
+    /* Get next directory entry */
+    if (dirp->cached != 0) {
+
+        /* A valid directory entry already in memory */
+        p = &dirp->data;
+        dirp->cached = 0;
+
+    } else if (dirp->handle != INVALID_HANDLE_VALUE) {
+
+        /* Get the next directory entry from stream */
+        if (FindNextFileW (dirp->handle, &dirp->data) != FALSE) {
+            /* Got a file */
+            p = &dirp->data;
+        } else {
+            /* The very last entry has been processed or an error occurred */
+            FindClose (dirp->handle);
+            dirp->handle = INVALID_HANDLE_VALUE;
+            p = NULL;
+        }
+
+    } else {
+
+        /* End of directory stream reached */
+        p = NULL;
+
+    }
+
+    return p;
+}
+
+/*
+ * Open directory stream using plain old C-string.
+ */
+static DIR*
+opendir(
+    const char *dirname)
+{
+    struct DIR *dirp;
+    int error;
+
+    /* Must have directory name */
+    if (dirname == NULL  ||  dirname[0] == '\0') {
+        dirent_set_errno (ENOENT);
+        return NULL;
+    }
+
+    /* Allocate memory for DIR structure */
+    dirp = (DIR*) malloc (sizeof (struct DIR));
+    if (dirp) {
+        wchar_t wname[PATH_MAX + 1];
+        size_t n;
+
+        /* Convert directory name to wide-character string */
+        error = dirent_mbstowcs_s(
+            &n, wname, PATH_MAX + 1, dirname, PATH_MAX + 1);
+        if (!error) {
+
+            /* Open directory stream using wide-character name */
+            dirp->wdirp = _wopendir (wname);
+            if (dirp->wdirp) {
+                /* Directory stream opened */
+                error = 0;
+            } else {
+                /* Failed to open directory stream */
+                error = 1;
+            }
+
+        } else {
+            /*
+             * Cannot convert file name to wide-character string.  This
+             * occurs if the string contains invalid multi-byte sequences or
+             * the output buffer is too small to contain the resulting
+             * string.
+             */
+            error = 1;
+        }
+
+    } else {
+        /* Cannot allocate DIR structure */
+        error = 1;
+    }
+
+    /* Clean up in case of error */
+    if (error  &&  dirp) {
+        free (dirp);
+        dirp = NULL;
+    }
+
+    return dirp;
+}
+
+/*
+ * Read next directory entry.
+ */
+static struct dirent*
+readdir(
+    DIR *dirp)
+{
+    struct dirent *entry;
+
+    /*
+     * Read directory entry to buffer.  We can safely ignore the return value
+     * as entry will be set to NULL in case of error.
+     */
+    (void) readdir_r (dirp, &dirp->ent, &entry);
+
+    /* Return pointer to statically allocated directory entry */
+    return entry;
+}
+
+/*
+ * Read next directory entry into called-allocated buffer.
+ *
+ * Returns zero on success.  If the end of directory stream is reached, then
+ * sets result to NULL and returns zero.
+ */
+static int
+readdir_r(
+    DIR *dirp,
+    struct dirent *entry,
+    struct dirent **result)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Read next directory entry */
+    datap = dirent_next (dirp->wdirp);
+    if (datap) {
+        size_t n;
+        int error;
+
+        /* Attempt to convert file name to multi-byte string */
+        error = dirent_wcstombs_s(
+            &n, entry->d_name, PATH_MAX + 1, datap->cFileName, PATH_MAX + 1);
+
+        /*
+         * If the file name cannot be represented by a multi-byte string,
+         * then attempt to use old 8+3 file name.  This allows traditional
+         * Unix-code to access some file names despite of unicode
+         * characters, although file names may seem unfamiliar to the user.
+         *
+         * Be ware that the code below cannot come up with a short file
+         * name unless the file system provides one.  At least
+         * VirtualBox shared folders fail to do this.
+         */
+        if (error  &&  datap->cAlternateFileName[0] != '\0') {
+            error = dirent_wcstombs_s(
+                &n, entry->d_name, PATH_MAX + 1,
+                datap->cAlternateFileName, PATH_MAX + 1);
+        }
+
+        if (!error) {
+            DWORD attr;
+
+            /* Length of file name excluding zero terminator */
+            entry->d_namlen = n - 1;
+
+            /* File attributes */
+            attr = datap->dwFileAttributes;
+            if ((attr & FILE_ATTRIBUTE_DEVICE) != 0) {
+                entry->d_type = DT_CHR;
+            } else if ((attr & FILE_ATTRIBUTE_DIRECTORY) != 0) {
+                entry->d_type = DT_DIR;
+            } else {
+                entry->d_type = DT_REG;
+            }
+
+            /* Reset dummy fields */
+            entry->d_ino = 0;
+            entry->d_off = 0;
+            entry->d_reclen = sizeof (struct dirent);
+
+        } else {
+
+            /*
+             * Cannot convert file name to multi-byte string so construct
+             * an erroneous directory entry and return that.  Note that
+             * we cannot return NULL as that would stop the processing
+             * of directory entries completely.
+             */
+            entry->d_name[0] = '?';
+            entry->d_name[1] = '\0';
+            entry->d_namlen = 1;
+            entry->d_type = DT_UNKNOWN;
+            entry->d_ino = 0;
+            entry->d_off = -1;
+            entry->d_reclen = 0;
+
+        }
+
+        /* Return pointer to directory entry */
+        *result = entry;
+
+    } else {
+
+        /* No more directory entries */
+        *result = NULL;
+
+    }
+
+    return /*OK*/0;
+}
+
+/*
+ * Close directory stream.
+ */
+static int
+closedir(
+    DIR *dirp)
+{
+    int ok;
+    if (dirp) {
+
+        /* Close wide-character directory stream */
+        ok = _wclosedir (dirp->wdirp);
+        dirp->wdirp = NULL;
+
+        /* Release multi-byte character version */
+        free (dirp);
+
+    } else {
+
+        /* Invalid directory stream */
+        dirent_set_errno (EBADF);
+        ok = /*failure*/-1;
+
+    }
+    return ok;
+}
+
+/*
+ * Rewind directory stream to beginning.
+ */
+static void
+rewinddir(
+    DIR* dirp)
+{
+    /* Rewind wide-character string directory stream */
+    _wrewinddir (dirp->wdirp);
+}
+
+/*
+ * Scan directory for entries.
+ */
+static int
+scandir(
+    const char *dirname,
+    struct dirent ***namelist,
+    int (*filter)(const struct dirent*),
+    int (*compare)(const struct dirent**, const struct dirent**))
+{
+    struct dirent **files = NULL;
+    size_t size = 0;
+    size_t allocated = 0;
+    const size_t init_size = 1;
+    DIR *dir = NULL;
+    struct dirent *entry;
+    struct dirent *tmp = NULL;
+    size_t i;
+    int result = 0;
+
+    /* Open directory stream */
+    dir = opendir (dirname);
+    if (dir) {
+
+        /* Read directory entries to memory */
+        while (1) {
+
+            /* Enlarge pointer table to make room for another pointer */
+            if (size >= allocated) {
+                void *p;
+                size_t num_entries;
+
+                /* Compute number of entries in the enlarged pointer table */
+                if (size < init_size) {
+                    /* Allocate initial pointer table */
+                    num_entries = init_size;
+                } else {
+                    /* Double the size */
+                    num_entries = size * 2;
+                }
+
+                /* Allocate first pointer table or enlarge existing table */
+                p = realloc (files, sizeof (void*) * num_entries);
+                if (p != NULL) {
+                    /* Got the memory */
+                    files = (dirent**) p;
+                    allocated = num_entries;
+                } else {
+                    /* Out of memory */
+                    result = -1;
+                    break;
+                }
+
+            }
+
+            /* Allocate room for temporary directory entry */
+            if (tmp == NULL) {
+                tmp = (struct dirent*) malloc (sizeof (struct dirent));
+                if (tmp == NULL) {
+                    /* Cannot allocate temporary directory entry */
+                    result = -1;
+                    break;
+                }
+            }
+
+            /* Read directory entry to temporary area */
+            if (readdir_r (dir, tmp, &entry) == /*OK*/0) {
+
+                /* Did we get an entry? */
+                if (entry != NULL) {
+                    int pass;
+
+                    /* Determine whether to include the entry in result */
+                    if (filter) {
+                        /* Let the filter function decide */
+                        pass = filter (tmp);
+                    } else {
+                        /* No filter function, include everything */
+                        pass = 1;
+                    }
+
+                    if (pass) {
+                        /* Store the temporary entry to pointer table */
+                        files[size++] = tmp;
+                        tmp = NULL;
+
+                        /* Keep up with the number of files */
+                        result++;
+                    }
+
+                } else {
+
+                    /*
+                     * End of directory stream reached => sort entries and
+                     * exit.
+                     */
+                    qsort (files, size, sizeof (void*),
+                        (int (*) (const void*, const void*)) compare);
+                    break;
+
+                }
+
+            } else {
+                /* Error reading directory entry */
+                result = /*Error*/ -1;
+                break;
+            }
+
+        }
+
+    } else {
+        /* Cannot open directory */
+        result = /*Error*/ -1;
+    }
+
+    /* Release temporary directory entry */
+    if (tmp) {
+        free (tmp);
+    }
+
+    /* Release allocated memory on error */
+    if (result < 0) {
+        for (i = 0; i < size; i++) {
+            free (files[i]);
+        }
+        free (files);
+        files = NULL;
+    }
+
+    /* Close directory stream */
+    if (dir) {
+        closedir (dir);
+    }
+
+    /* Pass pointer table to caller */
+    if (namelist) {
+        *namelist = files;
+    }
+    return result;
+}
+
+/* Alphabetical sorting */
+static int
+alphasort(
+    const struct dirent **a, const struct dirent **b)
+{
+    return strcoll ((*a)->d_name, (*b)->d_name);
+}
+
+/* Sort versions */
+static int
+versionsort(
+    const struct dirent **a, const struct dirent **b)
+{
+    /* FIXME: implement strverscmp and use that */
+    return alphasort (a, b);
+}
+
+/* Convert multi-byte string to wide character string */
+static int
+dirent_mbstowcs_s(
+    size_t *pReturnValue,
+    wchar_t *wcstr,
+    size_t sizeInWords,
+    const char *mbstr,
+    size_t count)
+{
+    int error;
+    int n;
+    size_t len;
+    UINT cp;
+    DWORD flags;
+
+    /* Determine code page for multi-byte string */
+    if (AreFileApisANSI ()) {
+        /* Default ANSI code page */
+        cp = GetACP ();
+    } else {
+        /* Default OEM code page */
+        cp = GetOEMCP ();
+    }
+
+    /*
+     * Determine flags based on the character set.  For more information,
+     * please see https://docs.microsoft.com/fi-fi/windows/desktop/api/stringapiset/nf-stringapiset-multibytetowidechar
+     */
+    switch (cp) {
+    case 42:
+    case 50220:
+    case 50221:
+    case 50222:
+    case 50225:
+    case 50227:
+    case 50229:
+    case 57002:
+    case 57003:
+    case 57004:
+    case 57005:
+    case 57006:
+    case 57007:
+    case 57008:
+    case 57009:
+    case 57010:
+    case 57011:
+    case 65000:
+        /* MultiByteToWideChar does not support MB_ERR_INVALID_CHARS */
+        flags = 0;
+        break;
+
+    default:
+        /*
+         * Ask MultiByteToWideChar to return an error if a multi-byte
+         * character cannot be converted to a wide-character.
+         */
+        flags = MB_ERR_INVALID_CHARS;
+    }
+
+    /* Compute the length of input string without zero-terminator */
+    len = 0;
+    while (mbstr[len] != '\0'  &&  len < count) {
+        len++;
+    }
+
+    /* Convert to wide-character string */
+    n = MultiByteToWideChar(
+        /* Source code page */ cp,
+        /* Flags */ flags,
+        /* Pointer to string to convert */ mbstr,
+        /* Size of multi-byte string */ (int) len,
+        /* Pointer to output buffer */ wcstr,
+        /* Size of output buffer */ (int)sizeInWords - 1
+    );
+
+    /* Ensure that output buffer is zero-terminated */
+    wcstr[n] = '\0';
+
+    /* Return length of wide-character string with zero-terminator */
+    *pReturnValue = (size_t) (n + 1);
+
+    /* Return zero if conversion succeeded */
+    if (n > 0) {
+        error = 0;
+    } else {
+        error = 1;
+    }
+    return error;
+}
+
+/* Convert wide-character string to multi-byte string */
+static int
+dirent_wcstombs_s(
+    size_t *pReturnValue,
+    char *mbstr,
+    size_t sizeInBytes, /* max size of mbstr */
+    const wchar_t *wcstr,
+    size_t count)
+{
+    int n;
+    int error;
+    UINT cp;
+    size_t len;
+    BOOL flag = 0;
+    LPBOOL pflag;
+
+    /* Determine code page for multi-byte string */
+    if (AreFileApisANSI ()) {
+        /* Default ANSI code page */
+        cp = GetACP ();
+    } else {
+        /* Default OEM code page */
+        cp = GetOEMCP ();
+    }
+
+    /* Compute the length of input string without zero-terminator */
+    len = 0;
+    while (wcstr[len] != '\0'  &&  len < count) {
+        len++;
+    }
+
+    /*
+     * Determine if we can ask WideCharToMultiByte to return information on
+     * broken characters.  For more information, please see
+     * https://docs.microsoft.com/en-us/windows/desktop/api/stringapiset/nf-stringapiset-widechartomultibyte
+     */
+    switch (cp) {
+    case CP_UTF7:
+    case CP_UTF8:
+        /*
+         * WideCharToMultiByte fails if we request information on default
+         * characters.  This is just a nuisance but doesn't affect the
+         * conversion: if Windows is configured to use UTF-8, then the default
+         * character should not be needed anyway.
+         */
+        pflag = NULL;
+        break;
+
+    default:
+        /*
+         * Request that WideCharToMultiByte sets the flag if it uses the
+         * default character.
+         */
+        pflag = &flag;
+    }
+
+    /* Convert wide-character string to multi-byte character string */
+    n = WideCharToMultiByte(
+        /* Target code page */ cp,
+        /* Flags */ 0,
+        /* Pointer to unicode string */ wcstr,
+        /* Length of unicode string */ (int) len,
+        /* Pointer to output buffer */ mbstr,
+        /* Size of output buffer */ (int)sizeInBytes - 1,
+        /* Default character */ NULL,
+        /* Whether default character was used or not */ pflag
+    );
+
+    /* Ensure that output buffer is zero-terminated */
+    mbstr[n] = '\0';
+
+    /* Return length of multi-byte string with zero-terminator */
+    *pReturnValue = (size_t) (n + 1);
+
+    /* Return zero if conversion succeeded without using default characters */
+    if (n > 0  &&  flag == 0) {
+        error = 0;
+    } else {
+        error = 1;
+    }
+    return error;
+}
+
+/* Set errno variable */
+static void
+dirent_set_errno(
+    int error)
+{
+#if defined(_MSC_VER)  &&  _MSC_VER >= 1400
+
+    /* Microsoft Visual Studio 2005 and later */
+    _set_errno (error);
+
+#else
+
+    /* Non-Microsoft compiler or older Microsoft compiler */
+    errno = error;
+
+#endif
+}
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif /*DIRENT_H*/
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/getopt.h glmark2-2021.02/android/jni/src/include/getopt.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/getopt.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/include/getopt.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,653 @@
+#ifndef __GETOPT_H__
+/**
+ * DISCLAIMER
+ * This file is part of the mingw-w64 runtime package.
+ *
+ * The mingw-w64 runtime package and its code is distributed in the hope that it 
+ * will be useful but WITHOUT ANY WARRANTY.  ALL WARRANTIES, EXPRESSED OR 
+ * IMPLIED ARE HEREBY DISCLAIMED.  This includes but is not limited to 
+ * warranties of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ */
+ /*
+ * Copyright (c) 2002 Todd C. Miller <Todd.Miller@courtesan.com>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * Sponsored in part by the Defense Advanced Research Projects
+ * Agency (DARPA) and Air Force Research Laboratory, Air Force
+ * Materiel Command, USAF, under agreement number F39502-99-1-0512.
+ */
+/*-
+ * Copyright (c) 2000 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Dieter Baron and Thomas Klausner.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#pragma warning(disable:4996)
+
+#define __GETOPT_H__
+
+/* All the headers include this file. */
+#include <crtdefs.h>
+#include <errno.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <windows.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define	REPLACE_GETOPT		/* use this getopt as the system getopt(3) */
+
+#ifdef REPLACE_GETOPT
+int	opterr = 1;		/* if error message should be printed */
+int	optind = 1;		/* index into parent argv vector */
+int	optopt = '?';		/* character checked for validity */
+#undef	optreset		/* see getopt.h */
+#define	optreset		__mingw_optreset
+int	optreset;		/* reset getopt */
+char    *optarg;		/* argument associated with option */
+#endif
+
+//extern int optind;		/* index of first non-option in argv      */
+//extern int optopt;		/* single option character, as parsed     */
+//extern int opterr;		/* flag to enable built-in diagnostics... */
+//				/* (user may set to zero, to suppress)    */
+//
+//extern char *optarg;		/* pointer to argument of current option  */
+
+#define PRINT_ERROR	((opterr) && (*options != ':'))
+
+#define FLAG_PERMUTE	0x01	/* permute non-options to the end of argv */
+#define FLAG_ALLARGS	0x02	/* treat non-options as args to option "-1" */
+#define FLAG_LONGONLY	0x04	/* operate as getopt_long_only */
+
+/* return values */
+#define	BADCH		(int)'?'
+#define	BADARG		((*options == ':') ? (int)':' : (int)'?')
+#define	INORDER 	(int)1
+
+#ifndef __CYGWIN__
+#define __progname __argv[0]
+#else
+extern char __declspec(dllimport) *__progname;
+#endif
+
+#ifdef __CYGWIN__
+static char EMSG[] = "";
+#else
+#define	EMSG		""
+#endif
+
+static int getopt_internal(int, char * const *, const char *,
+			   const struct option *, int *, int);
+static int parse_long_options(char * const *, const char *,
+			      const struct option *, int *, int);
+static int gcd(int, int);
+static void permute_args(int, int, int, char * const *);
+
+static char *place = EMSG; /* option letter processing */
+
+/* XXX: set optreset to 1 rather than these two */
+static int nonopt_start = -1; /* first non option argument (for permute) */
+static int nonopt_end = -1;   /* first option after non options (for permute) */
+
+/* Error messages */
+static const char recargchar[] = "option requires an argument -- %c";
+static const char recargstring[] = "option requires an argument -- %s";
+static const char ambig[] = "ambiguous option -- %.*s";
+static const char noarg[] = "option doesn't take an argument -- %.*s";
+static const char illoptchar[] = "unknown option -- %c";
+static const char illoptstring[] = "unknown option -- %s";
+
+static void
+_vwarnx(const char *fmt,va_list ap)
+{
+  (void)fprintf(stderr,"%s: ",__progname);
+  if (fmt != NULL)
+    (void)vfprintf(stderr,fmt,ap);
+  (void)fprintf(stderr,"\n");
+}
+
+static void
+warnx(const char *fmt,...)
+{
+  va_list ap;
+  va_start(ap,fmt);
+  _vwarnx(fmt,ap);
+  va_end(ap);
+}
+
+/*
+ * Compute the greatest common divisor of a and b.
+ */
+static int
+gcd(int a, int b)
+{
+	int c;
+
+	c = a % b;
+	while (c != 0) {
+		a = b;
+		b = c;
+		c = a % b;
+	}
+
+	return (b);
+}
+
+/*
+ * Exchange the block from nonopt_start to nonopt_end with the block
+ * from nonopt_end to opt_end (keeping the same order of arguments
+ * in each block).
+ */
+static void
+permute_args(int panonopt_start, int panonopt_end, int opt_end,
+	char * const *nargv)
+{
+	int cstart, cyclelen, i, j, ncycle, nnonopts, nopts, pos;
+	char *swap;
+
+	/*
+	 * compute lengths of blocks and number and size of cycles
+	 */
+	nnonopts = panonopt_end - panonopt_start;
+	nopts = opt_end - panonopt_end;
+	ncycle = gcd(nnonopts, nopts);
+	cyclelen = (opt_end - panonopt_start) / ncycle;
+
+	for (i = 0; i < ncycle; i++) {
+		cstart = panonopt_end+i;
+		pos = cstart;
+		for (j = 0; j < cyclelen; j++) {
+			if (pos >= panonopt_end)
+				pos -= nnonopts;
+			else
+				pos += nopts;
+			swap = nargv[pos];
+			/* LINTED const cast */
+			((char **) nargv)[pos] = nargv[cstart];
+			/* LINTED const cast */
+			((char **)nargv)[cstart] = swap;
+		}
+	}
+}
+
+#ifdef REPLACE_GETOPT
+/*
+ * getopt --
+ *	Parse argc/argv argument vector.
+ *
+ * [eventually this will replace the BSD getopt]
+ */
+int
+getopt(int nargc, char * const *nargv, const char *options)
+{
+
+	/*
+	 * We don't pass FLAG_PERMUTE to getopt_internal() since
+	 * the BSD getopt(3) (unlike GNU) has never done this.
+	 *
+	 * Furthermore, since many privileged programs call getopt()
+	 * before dropping privileges it makes sense to keep things
+	 * as simple (and bug-free) as possible.
+	 */
+	return (getopt_internal(nargc, nargv, options, NULL, NULL, 0));
+}
+#endif /* REPLACE_GETOPT */
+
+//extern int getopt(int nargc, char * const *nargv, const char *options);
+
+#ifdef _BSD_SOURCE
+/*
+ * BSD adds the non-standard `optreset' feature, for reinitialisation
+ * of `getopt' parsing.  We support this feature, for applications which
+ * proclaim their BSD heritage, before including this header; however,
+ * to maintain portability, developers are advised to avoid it.
+ */
+# define optreset  __mingw_optreset
+extern int optreset;
+#endif
+#ifdef __cplusplus
+}
+#endif
+/*
+ * POSIX requires the `getopt' API to be specified in `unistd.h';
+ * thus, `unistd.h' includes this header.  However, we do not want
+ * to expose the `getopt_long' or `getopt_long_only' APIs, when
+ * included in this manner.  Thus, close the standard __GETOPT_H__
+ * declarations block, and open an additional __GETOPT_LONG_H__
+ * specific block, only when *not* __UNISTD_H_SOURCED__, in which
+ * to declare the extended API.
+ */
+#endif /* !defined(__GETOPT_H__) */
+
+#if !defined(__UNISTD_H_SOURCED__) && !defined(__GETOPT_LONG_H__)
+#define __GETOPT_LONG_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct option		/* specification for a long form option...	*/
+{
+  const char *name;		/* option name, without leading hyphens */
+  int         has_arg;		/* does it take an argument?		*/
+  int        *flag;		/* where to save its status, or NULL	*/
+  int         val;		/* its associated status value		*/
+};
+
+enum    		/* permitted values for its `has_arg' field...	*/
+{
+  no_argument = 0,      	/* option never takes an argument	*/
+  required_argument,		/* option always requires an argument	*/
+  optional_argument		/* option may take an argument		*/
+};
+
+/*
+ * parse_long_options --
+ *	Parse long options in argc/argv argument vector.
+ * Returns -1 if short_too is set and the option does not match long_options.
+ */
+static int
+parse_long_options(char * const *nargv, const char *options,
+	const struct option *long_options, int *idx, int short_too)
+{
+	char *current_argv, *has_equal;
+	size_t current_argv_len;
+	int i, ambiguous, match;
+
+#define IDENTICAL_INTERPRETATION(_x, _y)                                \
+	(long_options[(_x)].has_arg == long_options[(_y)].has_arg &&    \
+	 long_options[(_x)].flag == long_options[(_y)].flag &&          \
+	 long_options[(_x)].val == long_options[(_y)].val)
+
+	current_argv = place;
+	match = -1;
+	ambiguous = 0;
+
+	optind++;
+
+	if ((has_equal = strchr(current_argv, '=')) != NULL) {
+		/* argument found (--option=arg) */
+		current_argv_len = has_equal - current_argv;
+		has_equal++;
+	} else
+		current_argv_len = strlen(current_argv);
+
+	for (i = 0; long_options[i].name; i++) {
+		/* find matching long option */
+		if (strncmp(current_argv, long_options[i].name,
+		    current_argv_len))
+			continue;
+
+		if (strlen(long_options[i].name) == current_argv_len) {
+			/* exact match */
+			match = i;
+			ambiguous = 0;
+			break;
+		}
+		/*
+		 * If this is a known short option, don't allow
+		 * a partial match of a single character.
+		 */
+		if (short_too && current_argv_len == 1)
+			continue;
+
+		if (match == -1)	/* partial match */
+			match = i;
+		else if (!IDENTICAL_INTERPRETATION(i, match))
+			ambiguous = 1;
+	}
+	if (ambiguous) {
+		/* ambiguous abbreviation */
+		if (PRINT_ERROR)
+			warnx(ambig, (int)current_argv_len,
+			     current_argv);
+		optopt = 0;
+		return (BADCH);
+	}
+	if (match != -1) {		/* option found */
+		if (long_options[match].has_arg == no_argument
+		    && has_equal) {
+			if (PRINT_ERROR)
+				warnx(noarg, (int)current_argv_len,
+				     current_argv);
+			/*
+			 * XXX: GNU sets optopt to val regardless of flag
+			 */
+			if (long_options[match].flag == NULL)
+				optopt = long_options[match].val;
+			else
+				optopt = 0;
+			return (BADARG);
+		}
+		if (long_options[match].has_arg == required_argument ||
+		    long_options[match].has_arg == optional_argument) {
+			if (has_equal)
+				optarg = has_equal;
+			else if (long_options[match].has_arg ==
+			    required_argument) {
+				/*
+				 * optional argument doesn't use next nargv
+				 */
+				optarg = nargv[optind++];
+			}
+		}
+		if ((long_options[match].has_arg == required_argument)
+		    && (optarg == NULL)) {
+			/*
+			 * Missing argument; leading ':' indicates no error
+			 * should be generated.
+			 */
+			if (PRINT_ERROR)
+				warnx(recargstring,
+				    current_argv);
+			/*
+			 * XXX: GNU sets optopt to val regardless of flag
+			 */
+			if (long_options[match].flag == NULL)
+				optopt = long_options[match].val;
+			else
+				optopt = 0;
+			--optind;
+			return (BADARG);
+		}
+	} else {			/* unknown option */
+		if (short_too) {
+			--optind;
+			return (-1);
+		}
+		if (PRINT_ERROR)
+			warnx(illoptstring, current_argv);
+		optopt = 0;
+		return (BADCH);
+	}
+	if (idx)
+		*idx = match;
+	if (long_options[match].flag) {
+		*long_options[match].flag = long_options[match].val;
+		return (0);
+	} else
+		return (long_options[match].val);
+#undef IDENTICAL_INTERPRETATION
+}
+
+/*
+ * getopt_internal --
+ *	Parse argc/argv argument vector.  Called by user level routines.
+ */
+static int
+getopt_internal(int nargc, char * const *nargv, const char *options,
+	const struct option *long_options, int *idx, int flags)
+{
+	char *oli;				/* option letter list index */
+	int optchar, short_too;
+	static int posixly_correct = -1;
+
+	if (options == NULL)
+		return (-1);
+
+	/*
+	 * XXX Some GNU programs (like cvs) set optind to 0 instead of
+	 * XXX using optreset.  Work around this braindamage.
+	 */
+	if (optind == 0)
+		optind = optreset = 1;
+
+	/*
+	 * Disable GNU extensions if POSIXLY_CORRECT is set or options
+	 * string begins with a '+'.
+	 *
+	 * CV, 2009-12-14: Check POSIXLY_CORRECT anew if optind == 0 or
+	 *                 optreset != 0 for GNU compatibility.
+	 */
+	if (posixly_correct == -1 || optreset != 0)
+		posixly_correct = (getenv("POSIXLY_CORRECT") != NULL);
+	if (*options == '-')
+		flags |= FLAG_ALLARGS;
+	else if (posixly_correct || *options == '+')
+		flags &= ~FLAG_PERMUTE;
+	if (*options == '+' || *options == '-')
+		options++;
+
+	optarg = NULL;
+	if (optreset)
+		nonopt_start = nonopt_end = -1;
+start:
+	if (optreset || !*place) {		/* update scanning pointer */
+		optreset = 0;
+		if (optind >= nargc) {          /* end of argument vector */
+			place = EMSG;
+			if (nonopt_end != -1) {
+				/* do permutation, if we have to */
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				optind -= nonopt_end - nonopt_start;
+			}
+			else if (nonopt_start != -1) {
+				/*
+				 * If we skipped non-options, set optind
+				 * to the first of them.
+				 */
+				optind = nonopt_start;
+			}
+			nonopt_start = nonopt_end = -1;
+			return (-1);
+		}
+		if (*(place = nargv[optind]) != '-' ||
+		    (place[1] == '\0' && strchr(options, '-') == NULL)) {
+			place = EMSG;		/* found non-option */
+			if (flags & FLAG_ALLARGS) {
+				/*
+				 * GNU extension:
+				 * return non-option as argument to option 1
+				 */
+				optarg = nargv[optind++];
+				return (INORDER);
+			}
+			if (!(flags & FLAG_PERMUTE)) {
+				/*
+				 * If no permutation wanted, stop parsing
+				 * at first non-option.
+				 */
+				return (-1);
+			}
+			/* do permutation */
+			if (nonopt_start == -1)
+				nonopt_start = optind;
+			else if (nonopt_end != -1) {
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				nonopt_start = optind -
+				    (nonopt_end - nonopt_start);
+				nonopt_end = -1;
+			}
+			optind++;
+			/* process next argument */
+			goto start;
+		}
+		if (nonopt_start != -1 && nonopt_end == -1)
+			nonopt_end = optind;
+
+		/*
+		 * If we have "-" do nothing, if "--" we are done.
+		 */
+		if (place[1] != '\0' && *++place == '-' && place[1] == '\0') {
+			optind++;
+			place = EMSG;
+			/*
+			 * We found an option (--), so if we skipped
+			 * non-options, we have to permute.
+			 */
+			if (nonopt_end != -1) {
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				optind -= nonopt_end - nonopt_start;
+			}
+			nonopt_start = nonopt_end = -1;
+			return (-1);
+		}
+	}
+
+	/*
+	 * Check long options if:
+	 *  1) we were passed some
+	 *  2) the arg is not just "-"
+	 *  3) either the arg starts with -- we are getopt_long_only()
+	 */
+	if (long_options != NULL && place != nargv[optind] &&
+	    (*place == '-' || (flags & FLAG_LONGONLY))) {
+		short_too = 0;
+		if (*place == '-')
+			place++;		/* --foo long option */
+		else if (*place != ':' && strchr(options, *place) != NULL)
+			short_too = 1;		/* could be short option too */
+
+		optchar = parse_long_options(nargv, options, long_options,
+		    idx, short_too);
+		if (optchar != -1) {
+			place = EMSG;
+			return (optchar);
+		}
+	}
+
+	if ((optchar = (int)*place++) == (int)':' ||
+	    (optchar == (int)'-' && *place != '\0') ||
+	    (oli = (char*)strchr(options, optchar)) == NULL) {
+		/*
+		 * If the user specified "-" and  '-' isn't listed in
+		 * options, return -1 (non-option) as per POSIX.
+		 * Otherwise, it is an unknown option character (or ':').
+		 */
+		if (optchar == (int)'-' && *place == '\0')
+			return (-1);
+		if (!*place)
+			++optind;
+		if (PRINT_ERROR)
+			warnx(illoptchar, optchar);
+		optopt = optchar;
+		return (BADCH);
+	}
+	if (long_options != NULL && optchar == 'W' && oli[1] == ';') {
+		/* -W long-option */
+		if (*place)			/* no space */
+			/* NOTHING */;
+		else if (++optind >= nargc) {	/* no arg */
+			place = EMSG;
+			if (PRINT_ERROR)
+				warnx(recargchar, optchar);
+			optopt = optchar;
+			return (BADARG);
+		} else				/* white space */
+			place = nargv[optind];
+		optchar = parse_long_options(nargv, options, long_options,
+		    idx, 0);
+		place = EMSG;
+		return (optchar);
+	}
+	if (*++oli != ':') {			/* doesn't take argument */
+		if (!*place)
+			++optind;
+	} else {				/* takes (optional) argument */
+		optarg = NULL;
+		if (*place)			/* no white space */
+			optarg = place;
+		else if (oli[1] != ':') {	/* arg not optional */
+			if (++optind >= nargc) {	/* no arg */
+				place = EMSG;
+				if (PRINT_ERROR)
+					warnx(recargchar, optchar);
+				optopt = optchar;
+				return (BADARG);
+			} else
+				optarg = nargv[optind];
+		}
+		place = EMSG;
+		++optind;
+	}
+	/* dump back option letter */
+	return (optchar);
+}
+
+/*
+ * getopt_long --
+ *	Parse argc/argv argument vector.
+ */
+int
+getopt_long(int nargc, char * const *nargv, const char *options,
+    const struct option *long_options, int *idx)
+{
+
+	return (getopt_internal(nargc, nargv, options, long_options, idx,
+	    FLAG_PERMUTE));
+}
+
+/*
+ * getopt_long_only --
+ *	Parse argc/argv argument vector.
+ */
+int
+getopt_long_only(int nargc, char * const *nargv, const char *options,
+    const struct option *long_options, int *idx)
+{
+
+	return (getopt_internal(nargc, nargv, options, long_options, idx,
+	    FLAG_PERMUTE|FLAG_LONGONLY));
+}
+
+//extern int getopt_long(int nargc, char * const *nargv, const char *options,
+//    const struct option *long_options, int *idx);
+//extern int getopt_long_only(int nargc, char * const *nargv, const char *options,
+//    const struct option *long_options, int *idx);
+/*
+ * Previous MinGW implementation had...
+ */
+#ifndef HAVE_DECL_GETOPT
+/*
+ * ...for the long form API only; keep this for compatibility.
+ */
+# define HAVE_DECL_GETOPT	1
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* !defined(__UNISTD_H_SOURCED__) && !defined(__GETOPT_LONG_H__) */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/sys/time.h glmark2-2021.02/android/jni/src/include/sys/time.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/include/sys/time.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/include/sys/time.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,45 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+// From Stack Overflow under Create Commons license:
+// https://stackoverflow.com/a/31335254
+
+#ifndef SYS_TIME_H_FOR_WINDOWS_
+#define SYS_TIME_H_FOR_WINDOWS_
+
+#ifdef _WIN32
+#include <time.h>
+#include <windows.h>
+#define CLOCK_MONOTONIC 0
+inline int clock_gettime(int, struct timespec *spec)
+{
+    __int64 wintime;
+    GetSystemTimeAsFileTime((FILETIME*)&wintime);
+    wintime      -= 116444736000000000i64;           // 1601 to 1970
+    spec->tv_sec  = wintime / 10000000i64;           // seconds
+    spec->tv_nsec = wintime % 10000000i64 *100;      // nanoseconds
+    return 0;
+}
+#else
+#include <sys/time.h>
+#endif
+
+#endif  // SYS_TIME_H_FOR_WINDOWS_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/acinclude.m4 glmark2-2021.02/android/jni/src/libjpeg-turbo/acinclude.m4
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/acinclude.m4	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/acinclude.m4	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,254 @@
+# AC_PROG_NASM
+# --------------------------
+# Check that NASM exists and determine flags
+AC_DEFUN([AC_PROG_NASM],[
+
+AC_ARG_VAR(NASM, [NASM command (used to build the x86/x86-64 SIMD code)])
+if test "x$NASM" = "x"; then
+  AC_CHECK_PROGS(NASM, [nasm nasmw yasm])
+  test -z "$NASM" && AC_MSG_ERROR([no nasm (Netwide Assembler) found])
+fi
+
+AC_MSG_CHECKING([for object file format of host system])
+case "$host_os" in
+  cygwin* | mingw* | pw32* | interix*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='Win64-COFF'
+        ;;
+      *)
+        objfmt='Win32-COFF'
+        ;;
+    esac
+  ;;
+  msdosdjgpp* | go32*)
+    objfmt='COFF'
+  ;;
+  os2-emx*)			# not tested
+    objfmt='MSOMF'		# obj
+  ;;
+  linux*coff* | linux*oldld*)
+    objfmt='COFF'		# ???
+  ;;
+  linux*aout*)
+    objfmt='a.out'
+  ;;
+  linux*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='ELF64'
+        ;;
+      *)
+        objfmt='ELF'
+        ;;
+    esac
+  ;;
+  kfreebsd* | freebsd* | netbsd* | openbsd*)
+    if echo __ELF__ | $CC -E - | grep __ELF__ > /dev/null; then
+      objfmt='BSD-a.out'
+    else
+      case "$host_cpu" in
+        x86_64 | amd64)
+          objfmt='ELF64'
+          ;;
+        *)
+          objfmt='ELF'
+          ;;
+      esac
+    fi
+  ;;
+  solaris* | sunos* | sysv* | sco*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='ELF64'
+        ;;
+      *)
+        objfmt='ELF'
+        ;;
+    esac
+  ;;
+  darwin* | rhapsody* | nextstep* | openstep* | macos*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='Mach-O64'
+        ;;
+      *)
+        objfmt='Mach-O'
+        ;;
+    esac
+  ;;
+  *)
+    objfmt='ELF ?'
+  ;;
+esac
+
+AC_MSG_RESULT([$objfmt])
+if test "$objfmt" = 'ELF ?'; then
+  objfmt='ELF'
+  AC_MSG_WARN([unexpected host system. assumed that the format is $objfmt.])
+fi
+
+AC_MSG_CHECKING([for object file format specifier (NAFLAGS) ])
+case "$objfmt" in
+  MSOMF)      NAFLAGS='-fobj -DOBJ32';;
+  Win32-COFF) NAFLAGS='-fwin32 -DWIN32';;
+  Win64-COFF) NAFLAGS='-fwin64 -DWIN64 -D__x86_64__';;
+  COFF)       NAFLAGS='-fcoff -DCOFF';;
+  a.out)      NAFLAGS='-faout -DAOUT';;
+  BSD-a.out)  NAFLAGS='-faoutb -DAOUT';;
+  ELF)        NAFLAGS='-felf -DELF';;
+  ELF64)      NAFLAGS='-felf64 -DELF -D__x86_64__';;
+  RDF)        NAFLAGS='-frdf -DRDF';;
+  Mach-O)     NAFLAGS='-fmacho -DMACHO';;
+  Mach-O64)   NAFLAGS='-fmacho64 -DMACHO -D__x86_64__';;
+esac
+AC_MSG_RESULT([$NAFLAGS])
+AC_SUBST([NAFLAGS])
+
+AC_MSG_CHECKING([whether the assembler ($NASM $NAFLAGS) works])
+cat > conftest.asm <<EOF
+[%line __oline__ "configure"
+        section .text
+        global  _main,main
+_main:
+main:   xor     eax,eax
+        ret
+]EOF
+try_nasm='$NASM $NAFLAGS -o conftest.o conftest.asm'
+if AC_TRY_EVAL(try_nasm) && test -s conftest.o; then
+  AC_MSG_RESULT(yes)
+else
+  echo "configure: failed program was:" >&AC_FD_CC
+  cat conftest.asm >&AC_FD_CC
+  rm -rf conftest*
+  AC_MSG_RESULT(no)
+  AC_MSG_ERROR([installation or configuration problem: assembler cannot create object files.])
+fi
+
+AC_MSG_CHECKING([whether the linker accepts assembler output])
+try_nasm='${CC-cc} -o conftest${ac_exeext} $LDFLAGS conftest.o $LIBS 1>&AC_FD_CC'
+if AC_TRY_EVAL(try_nasm) && test -s conftest${ac_exeext}; then
+  rm -rf conftest*
+  AC_MSG_RESULT(yes)
+else
+  rm -rf conftest*
+  AC_MSG_RESULT(no)
+  AC_MSG_ERROR([configuration problem: maybe object file format mismatch.])
+fi
+
+])
+
+# AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE
+# --------------------------
+# Test whether the assembler is suitable and supports NEON instructions
+AC_DEFUN([AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE],[
+  ac_good_gnu_arm_assembler=no
+  ac_save_CC="$CC"
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -x assembler-with-cpp"
+  CC="$CCAS"
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+    .text
+    .fpu neon
+    .arch armv7a
+    .object_arch armv4
+    .arm
+    pld [r0]
+    vmovn.u16 d0, q0]])], ac_good_gnu_arm_assembler=yes)
+
+  ac_use_gas_preprocessor=no
+  if test "x$ac_good_gnu_arm_assembler" = "xno" ; then
+    CC="gas-preprocessor.pl $CCAS"
+    AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+      .text
+      .fpu neon
+      .arch armv7a
+      .object_arch armv4
+      .arm
+      pld [r0]
+      vmovn.u16 d0, q0]])], ac_use_gas_preprocessor=yes)
+  fi
+  CFLAGS="$ac_save_CFLAGS"
+  CC="$ac_save_CC"
+
+  if test "x$ac_use_gas_preprocessor" = "xyes" ; then
+    CCAS="gas-preprocessor.pl $CCAS"
+    AC_SUBST([CCAS])
+    ac_good_gnu_arm_assembler=yes
+  fi
+
+  if test "x$ac_good_gnu_arm_assembler" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
+
+# AC_CHECK_COMPATIBLE_MIPSEL_ASSEMBLER_IFELSE
+# --------------------------
+# Test whether the assembler is suitable and supports MIPS instructions
+AC_DEFUN([AC_CHECK_COMPATIBLE_MIPS_ASSEMBLER_IFELSE],[
+  have_mips_dspr2=no
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -mdspr2"
+
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+
+  int main ()
+  {
+    int c = 0, a = 0, b = 0;
+    __asm__ __volatile__ (
+        "precr.qb.ph %[c], %[a], %[b]          \n\t"
+        : [c] "=r" (c)
+        : [a] "r" (a), [b] "r" (b)
+    );
+    return c;
+  }
+  ]])], have_mips_dspr2=yes)
+  CFLAGS=$ac_save_CFLAGS
+
+  if test "x$have_mips_dspr2" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
+
+AC_DEFUN([AC_CHECK_COMPATIBLE_ARM64_ASSEMBLER_IFELSE],[
+  ac_good_gnu_arm_assembler=no
+  ac_save_CC="$CC"
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -x assembler-with-cpp"
+  CC="$CCAS"
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+    .text
+    MYVAR .req x0
+    movi v0.16b, #100
+    mov MYVAR, #100
+    .unreq MYVAR]])], ac_good_gnu_arm_assembler=yes)
+
+  ac_use_gas_preprocessor=no
+  if test "x$ac_good_gnu_arm_assembler" = "xno" ; then
+    CC="gas-preprocessor.pl $CCAS"
+    AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+      .text
+      MYVAR .req x0
+      movi v0.16b, #100
+      mov MYVAR, #100
+      .unreq MYVAR]])], ac_use_gas_preprocessor=yes)
+  fi
+  CFLAGS="$ac_save_CFLAGS"
+  CC="$ac_save_CC"
+
+  if test "x$ac_use_gas_preprocessor" = "xyes" ; then
+    CCAS="gas-preprocessor.pl $CCAS"
+    AC_SUBST([CCAS])
+    ac_good_gnu_arm_assembler=yes
+  fi
+
+  if test "x$ac_good_gnu_arm_assembler" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/bmp.c glmark2-2021.02/android/jni/src/libjpeg-turbo/bmp.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/bmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/bmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,334 @@
+/*
+ * Copyright (C)2011, 2015 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include <setjmp.h>
+#include <errno.h>
+#include "cdjpeg.h"
+#include <jpeglib.h>
+#include <jpegint.h>
+#include "tjutil.h"
+#include "bmp.h"
+
+
+/* This duplicates the functionality of the VirtualGL bitmap library using
+   the components from cjpeg and djpeg */
+
+
+/* Error handling (based on example in example.c) */
+
+static char errStr[JMSG_LENGTH_MAX]="No error";
+
+struct my_error_mgr
+{
+	struct jpeg_error_mgr pub;
+	jmp_buf setjmp_buffer;
+};
+typedef struct my_error_mgr *my_error_ptr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	(*cinfo->err->output_message)(cinfo);
+	longjmp(myerr->setjmp_buffer, 1);
+}
+
+/* Based on output_message() in jerror.c */
+
+static void my_output_message(j_common_ptr cinfo)
+{
+	(*cinfo->err->format_message)(cinfo, errStr);
+}
+
+#define _throw(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s", m);  \
+	retval=-1;  goto bailout;}
+#define _throwunix(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s\n%s", m,  \
+	strerror(errno));  retval=-1;  goto bailout;}
+
+
+static void pixelconvert(unsigned char *srcbuf, int srcpf, int srcbottomup,
+	unsigned char *dstbuf, int dstpf, int dstbottomup, int w, int h)
+{
+	unsigned char *srcrowptr=srcbuf, *srccolptr;
+	int srcps=tjPixelSize[srcpf];
+	int srcstride=srcbottomup? -w*srcps:w*srcps;
+	unsigned char *dstrowptr=dstbuf, *dstcolptr;
+	int dstps=tjPixelSize[dstpf];
+	int dststride=dstbottomup? -w*dstps:w*dstps;
+	int row, col;
+
+	if(srcbottomup) srcrowptr=&srcbuf[w*srcps*(h-1)];
+	if(dstbottomup) dstrowptr=&dstbuf[w*dstps*(h-1)];
+
+	/* NOTE: These quick & dirty CMYK<->RGB conversion routines are for testing
+	   purposes only.  Properly converting between CMYK and RGB requires a color
+	   management system. */
+
+	if(dstpf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, srccolptr+=srcps)
+			{
+				double c=1.0-((double)(srccolptr[tjRedOffset[srcpf]])/255.);
+				double m=1.0-((double)(srccolptr[tjGreenOffset[srcpf]])/255.);
+				double y=1.0-((double)(srccolptr[tjBlueOffset[srcpf]])/255.);
+				double k=min(min(c,m),min(y,1.0));
+				if(k==1.0) c=m=y=0.0;
+				else
+				{
+					c=(c-k)/(1.0-k);
+					m=(m-k)/(1.0-k);
+					y=(y-k)/(1.0-k);
+				}
+				if(c>1.0) c=1.0;  if(c<0.) c=0.;
+				if(m>1.0) m=1.0;  if(m<0.) m=0.;
+				if(y>1.0) y=1.0;  if(y<0.) y=0.;
+				if(k>1.0) k=1.0;  if(k<0.) k=0.;
+				*dstcolptr++=(unsigned char)(255.0-c*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-m*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-y*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-k*255.0+0.5);
+			}
+		}
+	}
+	else if(srcpf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, dstcolptr+=dstps)
+			{
+				double c=(double)(*srccolptr++);
+				double m=(double)(*srccolptr++);
+				double y=(double)(*srccolptr++);
+				double k=(double)(*srccolptr++);
+				double r=c*k/255.;
+				double g=m*k/255.;
+				double b=y*k/255.;
+				if(r>255.0) r=255.0;  if(r<0.) r=0.;
+				if(g>255.0) g=255.0;  if(g<0.) g=0.;
+				if(b>255.0) b=255.0;  if(b<0.) b=0.;
+				dstcolptr[tjRedOffset[dstpf]]=(unsigned char)(r+0.5);
+				dstcolptr[tjGreenOffset[dstpf]]=(unsigned char)(g+0.5);
+				dstcolptr[tjBlueOffset[dstpf]]=(unsigned char)(b+0.5);
+			}
+		}
+	}
+	else
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, srccolptr+=srcps, dstcolptr+=dstps)
+			{
+				dstcolptr[tjRedOffset[dstpf]]=srccolptr[tjRedOffset[srcpf]];
+				dstcolptr[tjGreenOffset[dstpf]]=srccolptr[tjGreenOffset[srcpf]];
+				dstcolptr[tjBlueOffset[dstpf]]=srccolptr[tjBlueOffset[srcpf]];
+			}
+		}
+	}
+}
+
+
+int loadbmp(char *filename, unsigned char **buf, int *w, int *h,
+	int dstpf, int bottomup)
+{
+	int retval=0, dstps, srcpf, tempc;
+	struct jpeg_compress_struct cinfo;
+	struct my_error_mgr jerr;
+	cjpeg_source_ptr src;
+	FILE *file=NULL;
+
+	memset(&cinfo, 0, sizeof(struct jpeg_compress_struct));
+
+	if(!filename || !buf || !w || !h || dstpf<0 || dstpf>=TJ_NUMPF)
+		_throw("loadbmp(): Invalid argument");
+
+	if((file=fopen(filename, "rb"))==NULL)
+		_throwunix("loadbmp(): Cannot open input file");
+
+	cinfo.err=jpeg_std_error(&jerr.pub);
+	jerr.pub.error_exit=my_error_exit;
+	jerr.pub.output_message=my_output_message;
+
+	if(setjmp(jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;  goto bailout;
+	}
+
+	jpeg_create_compress(&cinfo);
+	if((tempc=getc(file))<0 || ungetc(tempc, file)==EOF)
+		_throwunix("loadbmp(): Could not read input file")
+	else if(tempc==EOF) _throw("loadbmp(): Input file contains no data");
+
+	if(tempc=='B')
+	{
+		if((src=jinit_read_bmp(&cinfo))==NULL)
+			_throw("loadbmp(): Could not initialize bitmap loader");
+	}
+	else if(tempc=='P')
+	{
+		if((src=jinit_read_ppm(&cinfo))==NULL)
+			_throw("loadbmp(): Could not initialize bitmap loader");
+	}
+	else _throw("loadbmp(): Unsupported file type");
+
+	src->input_file=file;
+	(*src->start_input)(&cinfo, src);
+	(*cinfo.mem->realize_virt_arrays)((j_common_ptr)&cinfo);
+
+	*w=cinfo.image_width;  *h=cinfo.image_height;
+
+	if(cinfo.input_components==1 && cinfo.in_color_space==JCS_RGB)
+		srcpf=TJPF_GRAY;
+	else srcpf=TJPF_RGB;
+
+	dstps=tjPixelSize[dstpf];
+	if((*buf=(unsigned char *)malloc((*w)*(*h)*dstps))==NULL)
+		_throw("loadbmp(): Memory allocation failure");
+
+	while(cinfo.next_scanline<cinfo.image_height)
+	{
+		int i, nlines=(*src->get_pixel_rows)(&cinfo, src);
+		for(i=0; i<nlines; i++)
+		{
+			unsigned char *outbuf;  int row;
+			row=cinfo.next_scanline+i;
+			if(bottomup) outbuf=&(*buf)[((*h)-row-1)*(*w)*dstps];
+			else outbuf=&(*buf)[row*(*w)*dstps];
+			pixelconvert(src->buffer[i], srcpf, 0, outbuf, dstpf, bottomup, *w,
+				nlines);
+		}
+		cinfo.next_scanline+=nlines;
+	}
+
+	(*src->finish_input)(&cinfo, src);
+
+	bailout:
+	jpeg_destroy_compress(&cinfo);
+	if(file) fclose(file);
+	if(retval<0 && buf && *buf) {free(*buf);  *buf=NULL;}
+	return retval;
+}
+
+
+int savebmp(char *filename, unsigned char *buf, int w, int h, int srcpf,
+	int bottomup)
+{
+	int retval=0, srcps, dstpf;
+	struct jpeg_decompress_struct dinfo;
+	struct my_error_mgr jerr;
+	djpeg_dest_ptr dst;
+	FILE *file=NULL;
+	char *ptr=NULL;
+
+	memset(&dinfo, 0, sizeof(struct jpeg_decompress_struct));
+
+	if(!filename || !buf || w<1 || h<1 || srcpf<0 || srcpf>=TJ_NUMPF)
+		_throw("savebmp(): Invalid argument");
+
+	if((file=fopen(filename, "wb"))==NULL)
+		_throwunix("savebmp(): Cannot open output file");
+
+	dinfo.err=jpeg_std_error(&jerr.pub);
+	jerr.pub.error_exit=my_error_exit;
+	jerr.pub.output_message=my_output_message;
+
+	if(setjmp(jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;  goto bailout;
+	}
+
+	jpeg_create_decompress(&dinfo);
+	if(srcpf==TJPF_GRAY)
+	{
+		dinfo.out_color_components=dinfo.output_components=1;
+		dinfo.out_color_space=JCS_GRAYSCALE;
+	}
+	else
+	{
+		dinfo.out_color_components=dinfo.output_components=3;
+		dinfo.out_color_space=JCS_RGB;
+	}
+	dinfo.image_width=w;  dinfo.image_height=h;
+	dinfo.global_state=DSTATE_READY;
+	dinfo.scale_num=dinfo.scale_denom=1;
+
+	ptr=strrchr(filename, '.');
+	if(ptr && !strcasecmp(ptr, ".bmp"))
+	{
+		if((dst=jinit_write_bmp(&dinfo, 0))==NULL)
+			_throw("savebmp(): Could not initialize bitmap writer");
+	}
+	else
+	{
+		if((dst=jinit_write_ppm(&dinfo))==NULL)
+			_throw("savebmp(): Could not initialize PPM writer");
+	}
+
+	dst->output_file=file;
+	(*dst->start_output)(&dinfo, dst);
+	(*dinfo.mem->realize_virt_arrays)((j_common_ptr)&dinfo);
+
+	if(srcpf==TJPF_GRAY) dstpf=srcpf;
+	else dstpf=TJPF_RGB;
+	srcps=tjPixelSize[srcpf];
+
+	while(dinfo.output_scanline<dinfo.output_height)
+	{
+		int i, nlines=dst->buffer_height;
+		for(i=0; i<nlines; i++)
+		{
+			unsigned char *inbuf;  int row;
+			row=dinfo.output_scanline+i;
+			if(bottomup) inbuf=&buf[(h-row-1)*w*srcps];
+			else inbuf=&buf[row*w*srcps];
+			pixelconvert(inbuf, srcpf, bottomup, dst->buffer[i], dstpf, 0, w,
+				nlines);
+		}
+		(*dst->put_pixel_rows)(&dinfo, dst, nlines);
+		dinfo.output_scanline+=nlines;
+	}
+
+	(*dst->finish_output)(&dinfo, dst);
+
+	bailout:
+	jpeg_destroy_decompress(&dinfo);
+	if(file) fclose(file);
+	return retval;
+}
+
+const char *bmpgeterr(void)
+{
+	return errStr;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/bmp.h glmark2-2021.02/android/jni/src/libjpeg-turbo/bmp.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/bmp.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/bmp.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,42 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __BMP_H__
+#define __BMP_H__
+
+#include "./turbojpeg.h"
+
+int loadbmp(char *filename, unsigned char **buf, int *w, int *h, int pf,
+	int bottomup);
+
+int savebmp(char *filename, unsigned char *buf, int w, int h, int pf,
+	int bottomup);
+
+const char *bmpgeterr(void);
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cderror.h glmark2-2021.02/android/jni/src/libjpeg-turbo/cderror.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cderror.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/cderror.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,136 @@
+/*
+ * cderror.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009 by Guido Vollbeding.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file defines the error and message codes for the cjpeg/djpeg
+ * applications.  These strings are not needed as part of the JPEG library
+ * proper.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef CDERROR_H
+#define CDERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code,string)
+#endif /* CDERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code,string)   code ,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_FIRSTADDONCODE=1000, NULL) /* Must be first entry! */
+
+#ifdef BMP_SUPPORTED
+JMESSAGE(JERR_BMP_BADCMAP, "Unsupported BMP colormap format")
+JMESSAGE(JERR_BMP_BADDEPTH, "Only 8- and 24-bit BMP files are supported")
+JMESSAGE(JERR_BMP_BADHEADER, "Invalid BMP file: bad header length")
+JMESSAGE(JERR_BMP_BADPLANES, "Invalid BMP file: biPlanes not equal to 1")
+JMESSAGE(JERR_BMP_COLORSPACE, "BMP output must be grayscale or RGB")
+JMESSAGE(JERR_BMP_COMPRESSED, "Sorry, compressed BMPs not yet supported")
+JMESSAGE(JERR_BMP_EMPTY, "Empty BMP image")
+JMESSAGE(JERR_BMP_NOT, "Not a BMP file - does not start with BM")
+JMESSAGE(JTRC_BMP, "%ux%u 24-bit BMP image")
+JMESSAGE(JTRC_BMP_MAPPED, "%ux%u 8-bit colormapped BMP image")
+JMESSAGE(JTRC_BMP_OS2, "%ux%u 24-bit OS2 BMP image")
+JMESSAGE(JTRC_BMP_OS2_MAPPED, "%ux%u 8-bit colormapped OS2 BMP image")
+#endif /* BMP_SUPPORTED */
+
+#ifdef GIF_SUPPORTED
+JMESSAGE(JERR_GIF_BUG, "GIF output got confused")
+JMESSAGE(JERR_GIF_CODESIZE, "Bogus GIF codesize %d")
+JMESSAGE(JERR_GIF_COLORSPACE, "GIF output must be grayscale or RGB")
+JMESSAGE(JERR_GIF_IMAGENOTFOUND, "Too few images in GIF file")
+JMESSAGE(JERR_GIF_NOT, "Not a GIF file")
+JMESSAGE(JTRC_GIF, "%ux%ux%d GIF image")
+JMESSAGE(JTRC_GIF_BADVERSION,
+         "Warning: unexpected GIF version number '%c%c%c'")
+JMESSAGE(JTRC_GIF_EXTENSION, "Ignoring GIF extension block of type 0x%02x")
+JMESSAGE(JTRC_GIF_NONSQUARE, "Caution: nonsquare pixels in input")
+JMESSAGE(JWRN_GIF_BADDATA, "Corrupt data in GIF file")
+JMESSAGE(JWRN_GIF_CHAR, "Bogus char 0x%02x in GIF file, ignoring")
+JMESSAGE(JWRN_GIF_ENDCODE, "Premature end of GIF image")
+JMESSAGE(JWRN_GIF_NOMOREDATA, "Ran out of GIF bits")
+#endif /* GIF_SUPPORTED */
+
+#ifdef PPM_SUPPORTED
+JMESSAGE(JERR_PPM_COLORSPACE, "PPM output must be grayscale or RGB")
+JMESSAGE(JERR_PPM_NONNUMERIC, "Nonnumeric data in PPM file")
+JMESSAGE(JERR_PPM_TOOLARGE, "Integer value too large in PPM file")
+JMESSAGE(JERR_PPM_NOT, "Not a PPM/PGM file")
+JMESSAGE(JTRC_PGM, "%ux%u PGM image")
+JMESSAGE(JTRC_PGM_TEXT, "%ux%u text PGM image")
+JMESSAGE(JTRC_PPM, "%ux%u PPM image")
+JMESSAGE(JTRC_PPM_TEXT, "%ux%u text PPM image")
+#endif /* PPM_SUPPORTED */
+
+#ifdef RLE_SUPPORTED
+JMESSAGE(JERR_RLE_BADERROR, "Bogus error code from RLE library")
+JMESSAGE(JERR_RLE_COLORSPACE, "RLE output must be grayscale or RGB")
+JMESSAGE(JERR_RLE_DIMENSIONS, "Image dimensions (%ux%u) too large for RLE")
+JMESSAGE(JERR_RLE_EMPTY, "Empty RLE file")
+JMESSAGE(JERR_RLE_EOF, "Premature EOF in RLE header")
+JMESSAGE(JERR_RLE_MEM, "Insufficient memory for RLE header")
+JMESSAGE(JERR_RLE_NOT, "Not an RLE file")
+JMESSAGE(JERR_RLE_TOOMANYCHANNELS, "Cannot handle %d output channels for RLE")
+JMESSAGE(JERR_RLE_UNSUPPORTED, "Cannot handle this RLE setup")
+JMESSAGE(JTRC_RLE, "%ux%u full-color RLE file")
+JMESSAGE(JTRC_RLE_FULLMAP, "%ux%u full-color RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_GRAY, "%ux%u grayscale RLE file")
+JMESSAGE(JTRC_RLE_MAPGRAY, "%ux%u grayscale RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_MAPPED, "%ux%u colormapped RLE file with map of length %d")
+#endif /* RLE_SUPPORTED */
+
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_TGA_BADCMAP, "Unsupported Targa colormap format")
+JMESSAGE(JERR_TGA_BADPARMS, "Invalid or unsupported Targa file")
+JMESSAGE(JERR_TGA_COLORSPACE, "Targa output must be grayscale or RGB")
+JMESSAGE(JTRC_TGA, "%ux%u RGB Targa image")
+JMESSAGE(JTRC_TGA_GRAY, "%ux%u grayscale Targa image")
+JMESSAGE(JTRC_TGA_MAPPED, "%ux%u colormapped Targa image")
+#else
+JMESSAGE(JERR_TGA_NOTCOMP, "Targa support was not compiled")
+#endif /* TARGA_SUPPORTED */
+
+JMESSAGE(JERR_BAD_CMAP_FILE,
+         "Color map file is invalid or of unsupported format")
+JMESSAGE(JERR_TOO_MANY_COLORS,
+         "Output file format cannot handle %d colormap entries")
+JMESSAGE(JERR_UNGETC_FAILED, "ungetc failed")
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_UNKNOWN_FORMAT,
+         "Unrecognized input file format --- perhaps you need -targa")
+#else
+JMESSAGE(JERR_UNKNOWN_FORMAT, "Unrecognized input file format")
+#endif
+JMESSAGE(JERR_UNSUPPORTED_FORMAT, "Unsupported output file format")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTADDONCODE
+} ADDON_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cdjpeg.c glmark2-2021.02/android/jni/src/libjpeg-turbo/cdjpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cdjpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/cdjpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,144 @@
+/*
+ * cdjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common support routines used by the IJG application
+ * programs (cjpeg, djpeg, jpegtran).
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+
+/*
+ * Optional progress monitor: display a percent-done figure on stderr.
+ */
+
+#ifdef PROGRESS_REPORT
+
+METHODDEF(void)
+progress_monitor (j_common_ptr cinfo)
+{
+  cd_progress_ptr prog = (cd_progress_ptr) cinfo->progress;
+  int total_passes = prog->pub.total_passes + prog->total_extra_passes;
+  int percent_done = (int) (prog->pub.pass_counter*100L/prog->pub.pass_limit);
+
+  if (percent_done != prog->percent_done) {
+    prog->percent_done = percent_done;
+    if (total_passes > 1) {
+      fprintf(stderr, "\rPass %d/%d: %3d%% ",
+              prog->pub.completed_passes + prog->completed_extra_passes + 1,
+              total_passes, percent_done);
+    } else {
+      fprintf(stderr, "\r %3d%% ", percent_done);
+    }
+    fflush(stderr);
+  }
+}
+
+
+GLOBAL(void)
+start_progress_monitor (j_common_ptr cinfo, cd_progress_ptr progress)
+{
+  /* Enable progress display, unless trace output is on */
+  if (cinfo->err->trace_level == 0) {
+    progress->pub.progress_monitor = progress_monitor;
+    progress->completed_extra_passes = 0;
+    progress->total_extra_passes = 0;
+    progress->percent_done = -1;
+    cinfo->progress = &progress->pub;
+  }
+}
+
+
+GLOBAL(void)
+end_progress_monitor (j_common_ptr cinfo)
+{
+  /* Clear away progress display */
+  if (cinfo->err->trace_level == 0) {
+    fprintf(stderr, "\r                \r");
+    fflush(stderr);
+  }
+}
+
+#endif
+
+
+/*
+ * Case-insensitive matching of possibly-abbreviated keyword switches.
+ * keyword is the constant keyword (must be lower case already),
+ * minchars is length of minimum legal abbreviation.
+ */
+
+GLOBAL(boolean)
+keymatch (char *arg, const char *keyword, int minchars)
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return FALSE;             /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return FALSE;             /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return FALSE;
+  return TRUE;                  /* A-OK */
+}
+
+
+/*
+ * Routines to establish binary I/O mode for stdin and stdout.
+ * Non-Unix systems often require some hacking to get out of text mode.
+ */
+
+GLOBAL(FILE *)
+read_stdin (void)
+{
+  FILE * input_file = stdin;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((input_file = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdin\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return input_file;
+}
+
+
+GLOBAL(FILE *)
+write_stdout (void)
+{
+  FILE * output_file = stdout;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((output_file = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdout\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return output_file;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cdjpeg.h glmark2-2021.02/android/jni/src/libjpeg-turbo/cdjpeg.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cdjpeg.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/cdjpeg.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,145 @@
+/*
+ * cdjpeg.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common declarations for the sample applications
+ * cjpeg and djpeg.  It is NOT used by the core JPEG library.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* define proper options in jconfig.h */
+#define JPEG_INTERNAL_OPTIONS   /* cjpeg.c,djpeg.c need to see xxx_SUPPORTED */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"             /* get library error codes too */
+#include "cderror.h"            /* get application-specific error codes */
+
+
+/*
+ * Object interface for cjpeg's source file decoding modules
+ */
+
+typedef struct cjpeg_source_struct *cjpeg_source_ptr;
+
+struct cjpeg_source_struct {
+  void (*start_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  void (*finish_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+
+  FILE *input_file;
+
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * Object interface for djpeg's output file encoding modules
+ */
+
+typedef struct djpeg_dest_struct *djpeg_dest_ptr;
+
+struct djpeg_dest_struct {
+  /* start_output is called after jpeg_start_decompress finishes.
+   * The color map will be ready at this time, if one is needed.
+   */
+  void (*start_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+  /* Emit the specified number of pixel rows from the buffer. */
+  void (*put_pixel_rows) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                          JDIMENSION rows_supplied);
+  /* Finish up at the end of the image. */
+  void (*finish_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+
+  /* Target file spec; filled in by djpeg.c after object is created. */
+  FILE *output_file;
+
+  /* Output pixel-row buffer.  Created by module init or start_output.
+   * Width is cinfo->output_width * cinfo->output_components;
+   * height is buffer_height.
+   */
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * cjpeg/djpeg may need to perform extra passes to convert to or from
+ * the source/destination file format.  The JPEG library does not know
+ * about these passes, but we'd like them to be counted by the progress
+ * monitor.  We use an expanded progress monitor object to hold the
+ * additional pass count.
+ */
+
+struct cdjpeg_progress_mgr {
+  struct jpeg_progress_mgr pub; /* fields known to JPEG library */
+  int completed_extra_passes;   /* extra passes completed */
+  int total_extra_passes;       /* total extra */
+  /* last printed percentage stored here to avoid multiple printouts */
+  int percent_done;
+};
+
+typedef struct cdjpeg_progress_mgr *cd_progress_ptr;
+
+
+/* Module selection routines for I/O modules. */
+
+EXTERN(cjpeg_source_ptr) jinit_read_bmp (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_bmp (j_decompress_ptr cinfo,
+                                        boolean is_os2);
+EXTERN(cjpeg_source_ptr) jinit_read_gif (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_gif (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_ppm (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_ppm (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_rle (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_rle (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_targa (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_targa (j_decompress_ptr cinfo);
+
+/* cjpeg support routines (in rdswitch.c) */
+
+EXTERN(boolean) read_quant_tables (j_compress_ptr cinfo, char *filename,
+                                   boolean force_baseline);
+EXTERN(boolean) read_scan_script (j_compress_ptr cinfo, char *filename);
+EXTERN(boolean) set_quality_ratings (j_compress_ptr cinfo, char *arg,
+                                     boolean force_baseline);
+EXTERN(boolean) set_quant_slots (j_compress_ptr cinfo, char *arg);
+EXTERN(boolean) set_sample_factors (j_compress_ptr cinfo, char *arg);
+
+/* djpeg support routines (in rdcolmap.c) */
+
+EXTERN(void) read_color_map (j_decompress_ptr cinfo, FILE *infile);
+
+/* common support routines (in cdjpeg.c) */
+
+EXTERN(void) enable_signal_catcher (j_common_ptr cinfo);
+EXTERN(void) start_progress_monitor (j_common_ptr cinfo,
+                                     cd_progress_ptr progress);
+EXTERN(void) end_progress_monitor (j_common_ptr cinfo);
+EXTERN(boolean) keymatch (char *arg, const char *keyword, int minchars);
+EXTERN(FILE *) read_stdin (void);
+EXTERN(FILE *) write_stdout (void);
+
+/* miscellaneous useful macros */
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+#ifndef EXIT_WARNING
+#define EXIT_WARNING  2
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/change.log glmark2-2021.02/android/jni/src/libjpeg-turbo/change.log
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/change.log	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/change.log	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,313 @@
+NOTE:  This file was modified by The libjpeg-turbo Project to include only
+information relevant to libjpeg-turbo.
+
+CHANGE LOG for Independent JPEG Group's JPEG software
+
+
+Version 9b  17-Jan-2016
+-----------------------
+
+Document 'f' specifier for jpegtran -crop specification.
+Thank to Michele Martone for suggestion.
+
+
+Version 9  13-Jan-2013
+----------------------
+
+Add remark for jpeg_mem_dest() in jdatadst.c.
+Thank to Elie-Gregoire Khoury for the hint.
+
+Correct argument type in format string, avoid compiler warnings.
+Thank to Vincent Torri for hint.
+
+
+Version 8d  15-Jan-2012
+-----------------------
+
+Add cjpeg -rgb option to create RGB JPEG files.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+Thank to Michael Koch for the initial suggestion.
+
+Add option to disable the region adjustment in the transupp crop code.
+Thank to Jeffrey Friedl for the suggestion.
+
+
+Version 8b  16-May-2010
+-----------------------
+
+Repair problem in new memory source manager with corrupt JPEG data.
+Thank to Ted Campbell and Samuel Chun for the report.
+
+
+Version 8a  28-Feb-2010
+-----------------------
+
+Writing tables-only datastreams via jpeg_write_tables works again.
+
+Support 32-bit BMPs (RGB image with Alpha channel) for read in cjpeg.
+Thank to Brett Blackham for the suggestion.
+
+
+Version 8  10-Jan-2010
+----------------------
+
+Add sanity check in BMP reader module to avoid cjpeg crash for empty input
+image (thank to Isaev Ildar of ISP RAS, Moscow, RU for reporting this error).
+
+Add data source and destination managers for read from and write to
+memory buffers.  New API functions jpeg_mem_src and jpeg_mem_dest.
+Thank to Roberto Boni from Italy for the suggestion.
+
+
+Version 7  27-Jun-2009
+----------------------
+
+New scaled DCTs implemented.
+djpeg now supports scalings N/8 with all N from 1 to 16.
+
+cjpeg -quality option has been extended for support of separate quality
+settings for luminance and chrominance (or in general, for every provided
+quantization table slot).
+New API function jpeg_default_qtables() and q_scale_factor array in library.
+
+Support arithmetic entropy encoding and decoding.
+Added files jaricom.c, jcarith.c, jdarith.c.
+
+jpegtran has a new "lossless" cropping feature.
+
+Implement -perfect option in jpegtran, new API function
+jtransform_perfect_transform() in transupp. (DP 204_perfect.dpatch)
+
+Better error messages for jpegtran fopen failure.
+(DP 203_jpegtran_errmsg.dpatch)
+
+Fix byte order issue with 16bit PPM/PGM files in rdppm.c/wrppm.c:
+according to Netpbm, the de facto standard implementation of the PNM formats,
+the most significant byte is first. (DP 203_rdppm.dpatch)
+
+Add -raw option to rdjpgcom not to mangle the output.
+(DP 205_rdjpgcom_raw.dpatch)
+
+Make rdjpgcom locale aware. (DP 201_rdjpgcom_locale.dpatch)
+
+Add extern "C" to jpeglib.h.
+This avoids the need to put extern "C" { ... } around #include "jpeglib.h"
+in your C++ application.  Defining the symbol DONT_USE_EXTERN_C in the
+configuration prevents this. (DP 202_jpeglib.h_c++.dpatch)
+
+
+Version 6b  27-Mar-1998
+-----------------------
+
+jpegtran has new features for lossless image transformations (rotation
+and flipping) as well as "lossless" reduction to grayscale.
+
+jpegtran now copies comments by default; it has a -copy switch to enable
+copying all APPn blocks as well, or to suppress comments.  (Formerly it
+always suppressed comments and APPn blocks.)  jpegtran now also preserves
+JFIF version and resolution information.
+
+New decompressor library feature: COM and APPn markers found in the input
+file can be saved in memory for later use by the application.  (Before,
+you had to code this up yourself with a custom marker processor.)
+
+There is an unused field "void * client_data" now in compress and decompress
+parameter structs; this may be useful in some applications.
+
+JFIF version number information is now saved by the decoder and accepted by
+the encoder.  jpegtran uses this to copy the source file's version number,
+to ensure "jpegtran -copy all" won't create bogus files that contain JFXX
+extensions but claim to be version 1.01.  Applications that generate their
+own JFXX extension markers also (finally) have a supported way to cause the
+encoder to emit JFIF version number 1.02.
+
+djpeg's trace mode reports JFIF 1.02 thumbnail images as such, rather
+than as unknown APP0 markers.
+
+In -verbose mode, djpeg and rdjpgcom will try to print the contents of
+APP12 markers as text.  Some digital cameras store useful text information
+in APP12 markers.
+
+Handling of truncated data streams is more robust: blocks beyond the one in
+which the error occurs will be output as uniform gray, or left unchanged
+if decoding a progressive JPEG.  The appearance no longer depends on the
+Huffman tables being used.
+
+Huffman tables are checked for validity much more carefully than before.
+
+To avoid the Unisys LZW patent, djpeg's GIF output capability has been
+changed to produce "uncompressed GIFs", and cjpeg's GIF input capability
+has been removed altogether.  We're not happy about it either, but there
+seems to be no good alternative.
+
+The configure script now supports building libjpeg as a shared library
+on many flavors of Unix (all the ones that GNU libtool knows how to
+build shared libraries for).  Use "./configure --enable-shared" to
+try this out.
+
+New jconfig file and makefiles for Microsoft Visual C++ and Developer Studio.
+Also, a jconfig file and a build script for Metrowerks CodeWarrior
+on Apple Macintosh.  makefile.dj has been updated for DJGPP v2, and there
+are miscellaneous other minor improvements in the makefiles.
+
+jmemmac.c now knows how to create temporary files following Mac System 7
+conventions.
+
+djpeg's -map switch is now able to read raw-format PPM files reliably.
+
+cjpeg -progressive -restart no longer generates any unnecessary DRI markers.
+
+Multiple calls to jpeg_simple_progression for a single JPEG object
+no longer leak memory.
+
+
+Version 6a  7-Feb-96
+--------------------
+
+Library initialization sequence modified to detect version mismatches
+and struct field packing mismatches between library and calling application.
+This change requires applications to be recompiled, but does not require
+any application source code change.
+
+All routine declarations changed to the style "GLOBAL(type) name ...",
+that is, GLOBAL, LOCAL, METHODDEF, EXTERN are now macros taking the
+routine's return type as an argument.  This makes it possible to add
+Microsoft-style linkage keywords to all the routines by changing just
+these macros.  Note that any application code that was using these macros
+will have to be changed.
+
+DCT coefficient quantization tables are now stored in normal array order
+rather than zigzag order.  Application code that calls jpeg_add_quant_table,
+or otherwise manipulates quantization tables directly, will need to be
+changed.  If you need to make such code work with either older or newer
+versions of the library, a test like "#if JPEG_LIB_VERSION >= 61" is
+recommended.
+
+djpeg's trace capability now dumps DQT tables in natural order, not zigzag
+order.  This allows the trace output to be made into a "-qtables" file
+more easily.
+
+New system-dependent memory manager module for use on Apple Macintosh.
+
+Fix bug in cjpeg's -smooth option: last one or two scanlines would be
+duplicates of the prior line unless the image height mod 16 was 1 or 2.
+
+Repair minor problems in VMS, BCC, MC6 makefiles.
+
+New configure script based on latest GNU Autoconf.
+
+Correct the list of include files needed by MetroWerks C for ccommand().
+
+Numerous small documentation updates.
+
+
+Version 6  2-Aug-95
+-------------------
+
+Progressive JPEG support: library can read and write full progressive JPEG
+files.  A "buffered image" mode supports incremental decoding for on-the-fly
+display of progressive images.  Simply recompiling an existing IJG-v5-based
+decoder with v6 should allow it to read progressive files, though of course
+without any special progressive display.
+
+New "jpegtran" application performs lossless transcoding between different
+JPEG formats; primarily, it can be used to convert baseline to progressive
+JPEG and vice versa.  In support of jpegtran, the library now allows lossless
+reading and writing of JPEG files as DCT coefficient arrays.  This ability
+may be of use in other applications.
+
+Notes for programmers:
+* We changed jpeg_start_decompress() to be able to suspend; this makes all
+decoding modes available to suspending-input applications.  However,
+existing applications that use suspending input will need to be changed
+to check the return value from jpeg_start_decompress().  You don't need to
+do anything if you don't use a suspending data source.
+* We changed the interface to the virtual array routines: access_virt_array
+routines now take a count of the number of rows to access this time.  The
+last parameter to request_virt_array routines is now interpreted as the
+maximum number of rows that may be accessed at once, but not necessarily
+the height of every access.
+
+
+Version 5b  15-Mar-95
+---------------------
+
+Correct bugs with grayscale images having v_samp_factor > 1.
+
+jpeg_write_raw_data() now supports output suspension.
+
+Correct bugs in "configure" script for case of compiling in
+a directory other than the one containing the source files.
+
+Repair bug in jquant1.c: sometimes didn't use as many colors as it could.
+
+Borland C makefile and jconfig file work under either MS-DOS or OS/2.
+
+Miscellaneous improvements to documentation.
+
+
+Version 5a  7-Dec-94
+--------------------
+
+Changed color conversion roundoff behavior so that grayscale values are
+represented exactly.  (This causes test image files to change.)
+
+Make ordered dither use 16x16 instead of 4x4 pattern for a small quality
+improvement.
+
+New configure script based on latest GNU Autoconf.
+Fix configure script to handle CFLAGS correctly.
+Rename *.auto files to *.cfg, so that configure script still works if
+file names have been truncated for DOS.
+
+Fix bug in rdbmp.c: didn't allow for extra data between header and image.
+
+Modify rdppm.c/wrppm.c to handle 2-byte raw PPM/PGM formats for 12-bit data.
+
+Fix several bugs in rdrle.c.
+
+NEED_SHORT_EXTERNAL_NAMES option was broken.
+
+Revise jerror.h/jerror.c for more flexibility in message table.
+
+Repair oversight in jmemname.c NO_MKTEMP case: file could be there
+but unreadable.
+
+
+Version 5  24-Sep-94
+--------------------
+
+Version 5 represents a nearly complete redesign and rewrite of the IJG
+software.  Major user-visible changes include:
+  * Automatic configuration simplifies installation for most Unix systems.
+  * A range of speed vs. image quality tradeoffs are supported.
+    This includes resizing of an image during decompression: scaling down
+    by a factor of 1/2, 1/4, or 1/8 is handled very efficiently.
+  * New programs rdjpgcom and wrjpgcom allow insertion and extraction
+    of text comments in a JPEG file.
+
+The application programmer's interface to the library has changed completely.
+Notable improvements include:
+  * We have eliminated the use of callback routines for handling the
+    uncompressed image data.  The application now sees the library as a
+    set of routines that it calls to read or write image data on a
+    scanline-by-scanline basis.
+  * The application image data is represented in a conventional interleaved-
+    pixel format, rather than as a separate array for each color channel.
+    This can save a copying step in many programs.
+  * The handling of compressed data has been cleaned up: the application can
+    supply routines to source or sink the compressed data.  It is possible to
+    suspend processing on source/sink buffer overrun, although this is not
+    supported in all operating modes.
+  * All static state has been eliminated from the library, so that multiple
+    instances of compression or decompression can be active concurrently.
+  * JPEG abbreviated datastream formats are supported, ie, quantization and
+    Huffman tables can be stored separately from the image data.
+  * And not only that, but the documentation of the library has improved
+    considerably!
+
+
+The last widely used release before the version 5 rewrite was version 4A of
+18-Feb-93.  Change logs before that point have been discarded, since they
+are not of much interest after the rewrite.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/ChangeLog.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/ChangeLog.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/ChangeLog.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/ChangeLog.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,861 @@
+1.5 beta1
+=========
+
+[1] Added full SIMD acceleration for PowerPC platforms using AltiVec VMX
+(128-bit SIMD) instructions.  Although the performance of libjpeg-turbo on
+PowerPC was already good, due to the increased number of registers available
+to the compiler vs. x86, it was still possible to speed up compression by about
+3-4x and decompression by about 2-2.5x (relative to libjpeg v6b) through the
+use of AltiVec instructions.
+
+[2] Added two new libjpeg API functions (jpeg_skip_scanlines() and
+jpeg_crop_scanline()) that can be used to partially decode a JPEG image.  See
+libjpeg.txt for more details.
+
+[3] The TJCompressor and TJDecompressor classes in the TurboJPEG Java API now
+implement the Closeable interface, so those classes can be used with a
+try-with-resources statement.
+
+[4] The TurboJPEG Java classes now throw unchecked idiomatic exceptions
+(IllegalArgumentException, IllegalStateException) for unrecoverable errors
+caused by incorrect API usage, and those classes throw a new checked exception
+type (TJException) for errors that are passed through from the C library.
+
+[5] Source buffers for the TurboJPEG C API functions, as well as the
+jpeg_mem_src() function in the libjpeg API, are now declared as const pointers.
+This facilitates passing read-only buffers to those functions and ensures the
+caller that the source buffer will not be modified.  This should not create any
+backward API or ABI incompatibilities with prior libjpeg-turbo releases.
+
+[6] The MIPS DSPr2 SIMD code can now be compiled to support either FR=0 or FR=1
+FPUs.
+
+[7] Fixed additional negative left shifts and other issues reported by the GCC
+and Clang undefined behavior sanitizers.  Most of these issues affected only
+32-bit code, and none of them was known to pose a security threat, but removing
+the warnings makes it easier to detect actual security issues, should they
+arise in the future.
+
+[8] Removed the unnecessary .arch directive from the ARM64 NEON SIMD code.
+This directive was preventing the code from assembling using the clang
+integrated assembler.
+
+[9] Fixed a regression caused by 1.4.1[6] that prevented 32-bit and 64-bit
+libjpeg-turbo RPMs from being installed simultaneously on recent Red Hat/Fedora
+distributions.  This was due to the addition of a macro in jconfig.h that
+allows the Huffman codec to determine the word size at compile time.  Since
+that macro differs between 32-bit and 64-bit builds, this caused a conflict
+between the i386 and x86_64 RPMs (any differing files, other than executables,
+are not allowed when 32-bit and 64-bit RPMs are installed simultaneously.)
+Since the macro is used only internally, it has been moved into jconfigint.h.
+
+[10] The x86-64 SIMD code can now be disabled at run time by setting the
+JSIMD_FORCENONE environment variable to 1 (the other SIMD implementations
+already had this capability.)
+
+[11] Added a new command-line argument to TJBench (-nowrite) that prevents the
+benchmark from outputting any images.  This removes any potential operating
+system overhead that might be caused by lazy writes to disk and thus improves
+the consistency of the performance measurements.
+
+[12] Added SIMD acceleration for Huffman encoding on SSE2-capable x86 and
+x86-64 platforms.  This speeds up the compression of full-color JPEGs by about
+10-15% on average (relative to libjpeg-turbo 1.4.x) when using modern Intel and
+AMD CPUs.  Additionally, this works around an issue in the clang optimizer that
+prevents it (as of this writing) from achieving the same performance as GCC
+when compiling the C version of the Huffman encoder
+(https://llvm.org/bugs/show_bug.cgi?id=16035). For the purposes of benchmarking
+or regression testing, SIMD-accelerated Huffman encoding can be disabled by
+setting the JSIMD_NOHUFFENC environment variable to 1.
+
+[13] Added ARM 64-bit (ARMv8) NEON SIMD implementations of the commonly-used
+compression algorithms (including the slow integer forward DCT and h2v2 & h2v1
+downsampling algorithms, which are not accelerated in the 32-bit NEON
+implementation.)  This speeds up the compression of full-color JPEGs by about
+75% on average on a Cavium ThunderX processor and by about 2-2.5x on average on
+Cortex-A53 and Cortex-A57 cores.
+
+[14] Added SIMD acceleration for Huffman encoding on NEON-capable ARM 32-bit
+and 64-bit platforms.
+
+For 32-bit code, this speeds up the compression of full-color JPEGs by about
+30% on average on a typical iOS device (iPhone 4S, Cortex-A9) and by about 6-7%
+on average on a typical Android device (Nexus 5X, Cortex-A53 and Cortex-A57),
+relative to libjpeg-turbo 1.4.x.  Note that the larger speedup under iOS is due
+to the fact that iOS builds use LLVM, which does not optimize the C Huffman
+encoder as well as GCC does.
+
+For 64-bit code, NEON-accelerated Huffman encoding speeds up the compression of
+full-color JPEGs by about 40% on average on a typical iOS device (iPhone 5S,
+Apple A7) and by about 7-8% on average on a typical Android device (Nexus 5X,
+Cortex-A53 and Cortex-A57), in addition to the speedup described in [13] above.
+
+For the purposes of benchmarking or regression testing, SIMD-accelerated
+Huffman encoding can be disabled by setting the JSIMD_NOHUFFENC environment
+variable to 1.
+
+[15] pkg-config (.pc) scripts are now included for both the libjpeg and
+TurboJPEG API libraries on Un*x systems.  Note that if a project's build system
+relies on these scripts, then it will not be possible to build that project
+with libjpeg or with a prior version of libjpeg-turbo.
+
+[16] Optimized the ARM 64-bit (ARMv8) NEON SIMD decompression routines to
+improve performance on CPUs with in-order pipelines.  This speeds up the
+decompression of full-color JPEGs by nearly 2x on average on a Cavium ThunderX
+processor and by about 15% on average on a Cortex-A53 core.
+
+[17] Fixed an issue in the accelerated Huffman decoder that could have caused
+the decoder to read past the end of the input buffer when a malformed,
+specially-crafted JPEG image was being decompressed.  In prior versions of
+libjpeg-turbo, the accelerated Huffman decoder was invoked (in most cases) only
+if there were > 128 bytes of data in the input buffer.  However, it is possible
+to construct a JPEG image in which a single Huffman block is over 430 bytes
+long, so this version of libjpeg-turbo activates the accelerated Huffman
+decoder only if there are > 512 bytes of data in the input buffer.
+
+[18] Fixed a memory leak in tjunittest encountered when running the program
+with the -yuv option.
+
+
+1.4.2
+=====
+
+[1] Fixed an issue whereby cjpeg would segfault if a Windows bitmap with a
+negative width or height was used as an input image (Windows bitmaps can have
+a negative height if they are stored in top-down order, but such files are
+rare and not supported by libjpeg-turbo.)
+
+[2] Fixed an issue whereby, under certain circumstances, libjpeg-turbo would
+incorrectly encode certain JPEG images when quality=100 and the fast integer
+forward DCT were used.  This was known to cause 'make test' to fail when the
+library was built with '-march=haswell' on x86 systems.
+
+[3] Fixed an issue whereby libjpeg-turbo would crash when built with the latest
+& greatest development version of the Clang/LLVM compiler.  This was caused by
+an x86-64 ABI conformance issue in some of libjpeg-turbo's 64-bit SSE2 SIMD
+routines.  Those routines were incorrectly using a 64-bit mov instruction to
+transfer a 32-bit JDIMENSION argument, whereas the x86-64 ABI allows the upper
+(unused) 32 bits of a 32-bit argument's register to be undefined.  The new
+Clang/LLVM optimizer uses load combining to transfer multiple adjacent 32-bit
+structure members into a single 64-bit register, and this exposed the ABI
+conformance issue.
+
+[4] Fixed a bug in the MIPS DSPr2 4:2:0 "plain" (non-fancy and non-merged)
+upsampling routine that caused a buffer overflow (and subsequent segfault) when
+decompressing a 4:2:0 JPEG image whose scaled output width was less than 16
+pixels.  The "plain" upsampling routines are normally only used when
+decompressing a non-YCbCr JPEG image, but they are also used when decompressing
+a JPEG image whose scaled output height is 1.
+
+[5] Fixed various negative left shifts and other issues reported by the GCC and
+Clang undefined behavior sanitizers.  None of these was known to pose a
+security threat, but removing the warnings makes it easier to detect actual
+security issues, should they arise in the future.
+
+
+1.4.1
+=====
+
+[1] tjbench now properly handles CMYK/YCCK JPEG files.  Passing an argument of
+-cmyk (instead of, for instance, -rgb) will cause tjbench to internally convert
+the source bitmap to CMYK prior to compression, to generate YCCK JPEG files,
+and to internally convert the decompressed CMYK pixels back to RGB after
+decompression (the latter is done automatically if a CMYK or YCCK JPEG is
+passed to tjbench as a source image.)  The CMYK<->RGB conversion operation is
+not benchmarked.  NOTE: The quick & dirty CMYK<->RGB conversions that tjbench
+uses are suitable for testing only.  Proper conversion between CMYK and RGB
+requires a color management system.
+
+[2] 'make test' now performs additional bitwise regression tests using tjbench,
+mainly for the purpose of testing compression from/decompression to a subregion
+of a larger image buffer.
+
+[3] 'make test' no longer tests the regression of the floating point DCT/IDCT
+by default, since the results of those tests can vary if the algorithms in
+question are not implemented using SIMD instructions on a particular platform.
+See the comments in Makefile.am for information on how to re-enable the tests
+and to specify an expected result for them based on the particulars of your
+platform.
+
+[4] The NULL color conversion routines have been significantly optimized,
+which speeds up the compression of RGB and CMYK JPEGs by 5-20% when using
+64-bit code and 0-3% when using 32-bit code, and the decompression of those
+images by 10-30% when using 64-bit code and 3-12% when using 32-bit code.
+
+[5] Fixed an "illegal instruction" error that occurred when djpeg from a
+SIMD-enabled libjpeg-turbo MIPS build was executed with the -nosmooth option on
+a MIPS machine that lacked DSPr2 support.  The MIPS SIMD routines for h2v1 and
+h2v2 merged upsampling were not properly checking for the existence of DSPr2.
+
+[6] Performance has been improved significantly on 64-bit non-Linux and
+non-Windows platforms (generally 10-20% faster compression and 5-10% faster
+decompression.)  Due to an oversight, the 64-bit version of the accelerated
+Huffman codec was not being compiled in when libjpeg-turbo was built on
+platforms other than Windows or Linux.  Oops.
+
+[7] Fixed an extremely rare bug in the Huffman encoder that caused 64-bit
+builds of libjpeg-turbo to incorrectly encode a few specific test images when
+quality=98, an optimized Huffman table, and the slow integer forward DCT were
+used.
+
+[8] The Windows (CMake) build system now supports building only static or only
+shared libraries.  This is accomplished by adding either -DENABLE_STATIC=0 or
+-DENABLE_SHARED=0 to the CMake command line.
+
+[9] TurboJPEG API functions will now return an error code if a warning is
+triggered in the underlying libjpeg API.  For instance, if a JPEG file is
+corrupt, the TurboJPEG decompression functions will attempt to decompress
+as much of the image as possible, but those functions will now return -1 to
+indicate that the decompression was not entirely successful.
+
+[10] Fixed a bug in the MIPS DSPr2 4:2:2 fancy upsampling routine that caused a
+buffer overflow (and subsequent segfault) when decompressing a 4:2:2 JPEG image
+in which the right-most MCU was 5 or 6 pixels wide.
+
+
+1.4.0
+=====
+
+[1] Fixed a build issue on OS X PowerPC platforms (md5cmp failed to build
+because OS X does not provide the le32toh() and htole32() functions.)
+
+[2] The non-SIMD RGB565 color conversion code did not work correctly on big
+endian machines.  This has been fixed.
+
+[3] Fixed an issue in tjPlaneSizeYUV() whereby it would erroneously return 1
+instead of -1 if componentID was > 0 and subsamp was TJSAMP_GRAY.
+
+[3] Fixed an issue in tjBufSizeYUV2() whereby it would erroneously return 0
+instead of -1 if width was < 1.
+
+[5] The Huffman encoder now uses clz and bsr instructions for bit counting on
+ARM64 platforms (see 1.4 beta1 [5].)
+
+[6] The close() method in the TJCompressor and TJDecompressor Java classes is
+now idempotent.  Previously, that method would call the native tjDestroy()
+function even if the TurboJPEG instance had already been destroyed.  This
+caused an exception to be thrown during finalization, if the close() method had
+already been called.  The exception was caught, but it was still an expensive
+operation.
+
+[7] The TurboJPEG API previously generated an error ("Could not determine
+subsampling type for JPEG image") when attempting to decompress grayscale JPEG
+images that were compressed with a sampling factor other than 1 (for instance,
+with 'cjpeg -grayscale -sample 2x2').  Subsampling technically has no meaning
+with grayscale JPEGs, and thus the horizontal and vertical sampling factors
+for such images are ignored by the decompressor.  However, the TurboJPEG API
+was being too rigid and was expecting the sampling factors to be equal to 1
+before it treated the image as a grayscale JPEG.
+
+[8] cjpeg, djpeg, and jpegtran now accept an argument of -version, which will
+print the library version and exit.
+
+[9] Referring to 1.4 beta1 [15], another extremely rare circumstance was
+discovered under which the Huffman encoder's local buffer can be overrun
+when a buffered destination manager is being used and an
+extremely-high-frequency block (basically junk image data) is being encoded.
+Even though the Huffman local buffer was increased from 128 bytes to 136 bytes
+to address the previous issue, the new issue caused even the larger buffer to
+be overrun.  Further analysis reveals that, in the absolute worst case (such as
+setting alternating AC coefficients to 32767 and -32768 in the JPEG scanning
+order), the Huffman encoder can produce encoded blocks that approach double the
+size of the unencoded blocks.  Thus, the Huffman local buffer was increased to
+256 bytes, which should prevent any such issue from re-occurring in the future.
+
+[10] The new tjPlaneSizeYUV(), tjPlaneWidth(), and tjPlaneHeight() functions
+were not actually usable on any platform except OS X and Windows, because
+those functions were not included in the libturbojpeg mapfile.  This has been
+fixed.
+
+[11] Restored the JPP(), JMETHOD(), and FAR macros in the libjpeg-turbo header
+files.  The JPP() and JMETHOD() macros were originally implemented in libjpeg
+as a way of supporting non-ANSI compilers that lacked support for prototype
+parameters.  libjpeg-turbo has never supported such compilers, but some
+software packages still use the macros to define their own prototypes.
+Similarly, libjpeg-turbo has never supported MS-DOS and other platforms that
+have far symbols, but some software packages still use the FAR macro.  A pretty
+good argument can be made that this is a bad practice on the part of the
+software in question, but since this affects more than one package, it's just
+easier to fix it here.
+
+[12] Fixed issues that were preventing the ARM 64-bit SIMD code from compiling
+for iOS, and included an ARMv8 architecture in all of the binaries installed by
+the "official" libjpeg-turbo SDK for OS X.
+
+
+1.3.90 (1.4 beta1)
+==================
+
+[1] New features in the TurboJPEG API:
+-- YUV planar images can now be generated with an arbitrary line padding
+(previously only 4-byte padding, which was compatible with X Video, was
+supported.)
+-- The decompress-to-YUV function has been extended to support image scaling.
+-- JPEG images can now be compressed from YUV planar source images.
+-- YUV planar images can now be decoded into RGB or grayscale images.
+-- 4:1:1 subsampling is now supported.  This is mainly included for
+compatibility, since 4:1:1 is not fully accelerated in libjpeg-turbo and has no
+significant advantages relative to 4:2:0.
+-- CMYK images are now supported.  This feature allows CMYK source images to be
+compressed to YCCK JPEGs and YCCK or CMYK JPEGs to be decompressed to CMYK
+destination images.  Conversion between CMYK/YCCK and RGB or YUV images is not
+supported.  Such conversion requires a color management system and is thus out
+of scope for a codec library.
+-- The handling of YUV images in the Java API has been significantly refactored
+and should now be much more intuitive.
+-- The Java API now supports encoding a YUV image from an arbitrary position in
+a large image buffer.
+-- All of the YUV functions now have a corresponding function that operates on
+separate image planes instead of a unified image buffer.  This allows for
+compressing/decoding from or decompressing/encoding to a subregion of a larger
+YUV image.  It also allows for handling YUV formats that swap the order of the
+U and V planes.
+
+[2] Added SIMD acceleration for DSPr2-capable MIPS platforms.  This speeds up
+the compression of full-color JPEGs by 70-80% on such platforms and
+decompression by 25-35%.
+
+[3] If an application attempts to decompress a Huffman-coded JPEG image whose
+header does not contain Huffman tables, libjpeg-turbo will now insert the
+default Huffman tables.  In order to save space, many motion JPEG video frames
+are encoded without the default Huffman tables, so these frames can now be
+successfully decompressed by libjpeg-turbo without additional work on the part
+of the application.  An application can still override the Huffman tables, for
+instance to re-use tables from a previous frame of the same video.
+
+[4] The Mac packaging system now uses pkgbuild and productbuild rather than
+PackageMaker (which is obsolete and no longer supported.)  This means that
+OS X 10.6 "Snow Leopard" or later must be used when packaging libjpeg-turbo,
+although the packages produced can be installed on OS X 10.5 "Leopard" or
+later.  OS X 10.4 "Tiger" is no longer supported.
+
+[5] The Huffman encoder now uses clz and bsr instructions for bit counting on
+ARM platforms rather than a lookup table.  This reduces the memory footprint
+by 64k, which may be important for some mobile applications.  Out of four
+Android devices that were tested, two demonstrated a small overall performance
+loss (~3-4% on average) with ARMv6 code and a small gain (also ~3-4%) with
+ARMv7 code when enabling this new feature, but the other two devices
+demonstrated a significant overall performance gain with both ARMv6 and ARMv7
+code (~10-20%) when enabling the feature.  Actual mileage may vary.
+
+[6] Worked around an issue with Visual C++ 2010 and later that caused incorrect
+pixels to be generated when decompressing a JPEG image to a 256-color bitmap,
+if compiler optimization was enabled when libjpeg-turbo was built.  This caused
+the regression tests to fail when doing a release build under Visual C++ 2010
+and later.
+
+[7] Improved the accuracy and performance of the non-SIMD implementation of the
+floating point inverse DCT (using code borrowed from libjpeg v8a and later.)
+The accuracy of this implementation now matches the accuracy of the SSE/SSE2
+implementation.  Note, however, that the floating point DCT/IDCT algorithms are
+mainly a legacy feature.  They generally do not produce significantly better
+accuracy than the slow integer DCT/IDCT algorithms, and they are quite a bit
+slower.
+
+[8] Added a new output colorspace (JCS_RGB565) to the libjpeg API that allows
+for decompressing JPEG images into RGB565 (16-bit) pixels.  If dithering is not
+used, then this code path is SIMD-accelerated on ARM platforms.
+
+[9] Numerous obsolete features, such as support for non-ANSI compilers and
+support for the MS-DOS memory model, were removed from the libjpeg code,
+greatly improving its readability and making it easier to maintain and extend.
+
+[10] Fixed a segfault that occurred when calling output_message() with msg_code
+set to JMSG_COPYRIGHT.
+
+[11] Fixed an issue whereby wrjpgcom was allowing comments longer than 65k
+characters to be passed on the command line, which was causing it to generate
+incorrect JPEG files.
+
+[12] Fixed a bug in the build system that was causing the Windows version of
+wrjpgcom to be built using the rdjpgcom source code.
+
+[13] Restored 12-bit-per-component JPEG support.  A 12-bit version of
+libjpeg-turbo can now be built by passing an argument of --with-12bit to
+configure (Unix) or -DWITH_12BIT=1 to cmake (Windows.)  12-bit JPEG support is
+included only for convenience.  Enabling this feature disables all of the
+performance features in libjpeg-turbo, as well as arithmetic coding and the
+TurboJPEG API.  The resulting library still contains the other libjpeg-turbo
+features (such as the colorspace extensions), but in general, it performs no
+faster than libjpeg v6b.
+
+[14] Added ARM 64-bit SIMD acceleration for the YCC-to-RGB color conversion
+and IDCT algorithms (both are used during JPEG decompression.)  For unknown
+reasons (probably related to clang), this code cannot currently be compiled for
+iOS.
+
+[15] Fixed an extremely rare bug that could cause the Huffman encoder's local
+buffer to overrun when a very high-frequency MCU is compressed using quality
+100 and no subsampling, and when the JPEG output buffer is being dynamically
+resized by the destination manager.  This issue was so rare that, even with a
+test program specifically designed to make the bug occur (by injecting random
+high-frequency YUV data into the compressor), it was reproducible only once in
+about every 25 million iterations.
+
+[16] Fixed an oversight in the TurboJPEG C wrapper:  if any of the JPEG
+compression functions was called repeatedly with the same
+automatically-allocated destination buffer, then TurboJPEG would erroneously
+assume that the jpegSize parameter was equal to the size of the buffer, when in
+fact that parameter was probably equal to the size of the most recently
+compressed JPEG image.  If the size of the previous JPEG image was not as large
+as the current JPEG image, then TurboJPEG would unnecessarily reallocate the
+destination buffer.
+
+
+1.3.1
+=====
+
+[1] On Un*x systems, 'make install' now installs the libjpeg-turbo libraries
+into /opt/libjpeg-turbo/lib32 by default on any 32-bit system, not just x86,
+and into /opt/libjpeg-turbo/lib64 by default on any 64-bit system, not just
+x86-64.  You can override this by overriding either the 'prefix' or 'libdir'
+configure variables.
+
+[2] The Windows installer now places a copy of the TurboJPEG DLLs in the same
+directory as the rest of the libjpeg-turbo binaries.  This was mainly done
+to support TurboVNC 1.3, which bundles the DLLs in its Windows installation.
+When using a 32-bit version of CMake on 64-bit Windows, it is impossible to
+access the c:\WINDOWS\system32 directory, which made it impossible for the
+TurboVNC build scripts to bundle the 64-bit TurboJPEG DLL.
+
+[3] Fixed a bug whereby attempting to encode a progressive JPEG with arithmetic
+entropy coding (by passing arguments of -progressive -arithmetic to cjpeg or
+jpegtran, for instance) would result in an error, "Requested feature was
+omitted at compile time".
+
+[4] Fixed a couple of issues whereby malformed JPEG images would cause
+libjpeg-turbo to use uninitialized memory during decompression.
+
+[5] Fixed an error ("Buffer passed to JPEG library is too small") that occurred
+when calling the TurboJPEG YUV encoding function with a very small (< 5x5)
+source image, and added a unit test to check for this error.
+
+[6] The Java classes should now build properly under Visual Studio 2010 and
+later.
+
+[7] Fixed an issue that prevented SRPMs generated using the in-tree packaging
+tools from being rebuilt on certain newer Linux distributions.
+
+[8] Numerous minor fixes to eliminate compilation and build/packaging system
+warnings, fix cosmetic issues, improve documentation clarity, and other general
+source cleanup.
+
+
+1.3.0
+=====
+
+[1] 'make test' now works properly on FreeBSD, and it no longer requires the
+md5sum executable to be present on other Un*x platforms.
+
+[2] Overhauled the packaging system:
+-- To avoid conflict with vendor-supplied libjpeg-turbo packages, the
+official RPMs and DEBs for libjpeg-turbo have been renamed to
+"libjpeg-turbo-official".
+-- The TurboJPEG libraries are now located under /opt/libjpeg-turbo in the
+official Linux and Mac packages, to avoid conflict with vendor-supplied
+packages and also to streamline the packaging system.
+-- Release packages are now created with the directory structure defined
+by the configure variables "prefix", "bindir", "libdir", etc. (Un*x) or by the
+CMAKE_INSTALL_PREFIX variable (Windows.)  The exception is that the docs are
+always located under the system default documentation directory on Un*x and Mac
+systems, and on Windows, the TurboJPEG DLL is always located in the Windows
+system directory.
+-- To avoid confusion, official libjpeg-turbo packages on Linux/Unix platforms
+(except for Mac) will always install the 32-bit libraries in
+/opt/libjpeg-turbo/lib32 and the 64-bit libraries in /opt/libjpeg-turbo/lib64.
+-- Fixed an issue whereby, in some cases, the libjpeg-turbo executables on Un*x
+systems were not properly linking with the shared libraries installed by the
+same package.
+-- Fixed an issue whereby building the "installer" target on Windows when
+WITH_JAVA=1 would fail if the TurboJPEG JAR had not been previously built.
+-- Building the "install" target on Windows now installs files into the same
+places that the installer does.
+
+[3] Fixed a Huffman encoder bug that prevented I/O suspension from working
+properly.
+
+
+1.2.90 (1.3 beta1)
+==================
+
+[1] Added support for additional scaling factors (3/8, 5/8, 3/4, 7/8, 9/8, 5/4,
+11/8, 3/2, 13/8, 7/4, 15/8, and 2) when decompressing.  Note that the IDCT will
+not be SIMD-accelerated when using any of these new scaling factors.
+
+[2] The TurboJPEG dynamic library is now versioned.  It was not strictly
+necessary to do so, because TurboJPEG uses versioned symbols, and if a function
+changes in an ABI-incompatible way, that function is renamed and a legacy
+function is provided to maintain backward compatibility.  However, certain
+Linux distro maintainers have a policy against accepting any library that isn't
+versioned.
+
+[3] Extended the TurboJPEG Java API so that it can be used to compress a JPEG
+image from and decompress a JPEG image to an arbitrary position in a large
+image buffer.
+
+[4] The tjDecompressToYUV() function now supports the TJFLAG_FASTDCT flag.
+
+[5] The 32-bit supplementary package for amd64 Debian systems now provides
+symlinks in /usr/lib/i386-linux-gnu for the TurboJPEG libraries in /usr/lib32.
+This allows those libraries to be used on MultiArch-compatible systems (such as
+Ubuntu 11 and later) without setting the linker path.
+
+[6] The TurboJPEG Java wrapper should now find the JNI library on Mac systems
+without having to pass -Djava.library.path=/usr/lib to java.
+
+[7] TJBench has been ported to Java to provide a convenient way of validating
+the performance of the TurboJPEG Java API.  It can be run with
+'java -cp turbojpeg.jar TJBench'.
+
+[8] cjpeg can now be used to generate JPEG files with the RGB colorspace
+(feature ported from jpeg-8d.)
+
+[9] The width and height in the -crop argument passed to jpegtran can now be
+suffixed with "f" to indicate that, when the upper left corner of the cropping
+region is automatically moved to the nearest iMCU boundary, the bottom right
+corner should be moved by the same amount.  In other words, this feature causes
+jpegtran to strictly honor the specified width/height rather than the specified
+bottom right corner (feature ported from jpeg-8d.)
+
+[10] JPEG files using the RGB colorspace can now be decompressed into grayscale
+images (feature ported from jpeg-8d.)
+
+[11] Fixed a regression caused by 1.2.1[7] whereby the build would fail with
+multiple "Mismatch in operand sizes" errors when attempting to build the x86
+SIMD code with NASM 0.98.
+
+[12] The in-memory source/destination managers (jpeg_mem_src() and
+jpeg_mem_dest()) are now included by default when building libjpeg-turbo with
+libjpeg v6b or v7 emulation, so that programs can take advantage of these
+functions without requiring the use of the backward-incompatible libjpeg v8
+ABI.  The "age number" of the libjpeg-turbo library on Un*x systems has been
+incremented by 1 to reflect this.  You can disable this feature with a
+configure/CMake switch in order to retain strict API/ABI compatibility with the
+libjpeg v6b or v7 API/ABI (or with previous versions of libjpeg-turbo.)  See
+README.md for more details.
+
+[13] Added ARMv7s architecture to libjpeg.a and libturbojpeg.a in the official
+libjpeg-turbo binary package for OS X, so that those libraries can be used to
+build applications that leverage the faster CPUs in the iPhone 5 and iPad 4.
+
+
+1.2.1
+=====
+
+[1] Creating or decoding a JPEG file that uses the RGB colorspace should now
+properly work when the input or output colorspace is one of the libjpeg-turbo
+colorspace extensions.
+
+[2] When libjpeg-turbo was built without SIMD support and merged (non-fancy)
+upsampling was used along with an alpha-enabled colorspace during
+decompression, the unused byte of the decompressed pixels was not being set to
+0xFF.  This has been fixed.  TJUnitTest has also been extended to test for the
+correct behavior of the colorspace extensions when merged upsampling is used.
+
+[3] Fixed a bug whereby the libjpeg-turbo SSE2 SIMD code would not preserve the
+upper 64 bits of xmm6 and xmm7 on Win64 platforms, which violated the Win64
+calling conventions.
+
+[4] Fixed a regression caused by 1.2.0[6] whereby decompressing corrupt JPEG
+images (specifically, images in which the component count was erroneously set
+to a large value) would cause libjpeg-turbo to segfault.
+
+[5] Worked around a severe performance issue with "Bobcat" (AMD Embedded APU)
+processors.  The MASKMOVDQU instruction, which was used by the libjpeg-turbo
+SSE2 SIMD code, is apparently implemented in microcode on AMD processors, and
+it is painfully slow on Bobcat processors in particular.  Eliminating the use
+of this instruction improved performance by an order of magnitude on Bobcat
+processors and by a small amount (typically 5%) on AMD desktop processors.
+
+[6] Added SIMD acceleration for performing 4:2:2 upsampling on NEON-capable ARM
+platforms.  This speeds up the decompression of 4:2:2 JPEGs by 20-25% on such
+platforms.
+
+[7] Fixed a regression caused by 1.2.0[2] whereby, on Linux/x86 platforms
+running the 32-bit SSE2 SIMD code in libjpeg-turbo, decompressing a 4:2:0 or
+4:2:2 JPEG image into a 32-bit (RGBX, BGRX, etc.) buffer without using fancy
+upsampling would produce several incorrect columns of pixels at the right-hand
+side of the output image if each row in the output image was not evenly
+divisible by 16 bytes.
+
+[8] Fixed an issue whereby attempting to build the SIMD extensions with Xcode
+4.3 on OS X platforms would cause NASM to return numerous errors of the form
+"'%define' expects a macro identifier".
+
+[9] Added flags to the TurboJPEG API that allow the caller to force the use of
+either the fast or the accurate DCT/IDCT algorithms in the underlying codec.
+
+
+1.2.0
+=====
+
+[1] Fixed build issue with YASM on Unix systems (the libjpeg-turbo build system
+was not adding the current directory to the assembler include path, so YASM
+was not able to find jsimdcfg.inc.)
+
+[2] Fixed out-of-bounds read in SSE2 SIMD code that occurred when decompressing
+a JPEG image to a bitmap buffer whose size was not a multiple of 16 bytes.
+This was more of an annoyance than an actual bug, since it did not cause any
+actual run-time problems, but the issue showed up when running libjpeg-turbo in
+valgrind.  See http://crbug.com/72399 for more information.
+
+[3] Added a compile-time macro (LIBJPEG_TURBO_VERSION) that can be used to
+check the version of libjpeg-turbo against which an application was compiled.
+
+[4] Added new RGBA/BGRA/ABGR/ARGB colorspace extension constants (libjpeg API)
+and pixel formats (TurboJPEG API), which allow applications to specify that,
+when decompressing to a 4-component RGB buffer, the unused byte should be set
+to 0xFF so that it can be interpreted as an opaque alpha channel.
+
+[5] Fixed regression issue whereby DevIL failed to build against libjpeg-turbo
+because libjpeg-turbo's distributed version of jconfig.h contained an INLINE
+macro, which conflicted with a similar macro in DevIL.  This macro is used only
+internally when building libjpeg-turbo, so it was moved into config.h.
+
+[6] libjpeg-turbo will now correctly decompress erroneous CMYK/YCCK JPEGs whose
+K component is assigned a component ID of 1 instead of 4.  Although these files
+are in violation of the spec, other JPEG implementations handle them
+correctly.
+
+[7] Added ARMv6 and ARMv7 architectures to libjpeg.a and libturbojpeg.a in
+the official libjpeg-turbo binary package for OS X, so that those libraries can
+be used to build both OS X and iOS applications.
+
+
+1.1.90 (1.2 beta1)
+==================
+
+[1] Added a Java wrapper for the TurboJPEG API.  See java/README for more
+details.
+
+[2] The TurboJPEG API can now be used to scale down images during
+decompression.
+
+[3] Added SIMD routines for RGB-to-grayscale color conversion, which
+significantly improves the performance of grayscale JPEG compression from an
+RGB source image.
+
+[4] Improved the performance of the C color conversion routines, which are used
+on platforms for which SIMD acceleration is not available.
+
+[5] Added a function to the TurboJPEG API that performs lossless transforms.
+This function is implemented using the same back end as jpegtran, but it
+performs transcoding entirely in memory and allows multiple transforms and/or
+crop operations to be batched together, so the source coefficients only need to
+be read once.  This is useful when generating image tiles from a single source
+JPEG.
+
+[6] Added tests for the new TurboJPEG scaled decompression and lossless
+transform features to tjbench (the TurboJPEG benchmark, formerly called
+"jpgtest".)
+
+[7] Added support for 4:4:0 (transposed 4:2:2) subsampling in TurboJPEG, which
+was necessary in order for it to read 4:2:2 JPEG files that had been losslessly
+transposed or rotated 90 degrees.
+
+[8] All legacy VirtualGL code has been re-factored, and this has allowed
+libjpeg-turbo, in its entirety, to be re-licensed under a BSD-style license.
+
+[9] libjpeg-turbo can now be built with YASM.
+
+[10] Added SIMD acceleration for ARM Linux and iOS platforms that support
+NEON instructions.
+
+[11] Refactored the TurboJPEG C API and documented it using Doxygen.  The
+TurboJPEG 1.2 API uses pixel formats to define the size and component order of
+the uncompressed source/destination images, and it includes a more efficient
+version of TJBUFSIZE() that computes a worst-case JPEG size based on the level
+of chrominance subsampling.  The refactored implementation of the TurboJPEG API
+now uses the libjpeg memory source and destination managers, which allows the
+TurboJPEG compressor to grow the JPEG buffer as necessary.
+
+[12] Eliminated errors in the output of jpegtran on Windows that occurred when
+the application was invoked using I/O redirection
+(jpegtran <input.jpg >output.jpg).
+
+[13] The inclusion of libjpeg v7 and v8 emulation as well as arithmetic coding
+support in libjpeg-turbo v1.1.0 introduced several new error constants in
+jerror.h, and these were mistakenly enabled for all emulation modes, causing
+the error enum in libjpeg-turbo to sometimes have different values than the
+same enum in libjpeg.  This represents an ABI incompatibility, and it caused
+problems with rare applications that took specific action based on a particular
+error value.  The fix was to include the new error constants conditionally
+based on whether libjpeg v7 or v8 emulation was enabled.
+
+[14] Fixed an issue whereby Windows applications that used libjpeg-turbo would
+fail to compile if the Windows system headers were included before jpeglib.h.
+This issue was caused by a conflict in the definition of the INT32 type.
+
+[15] Fixed 32-bit supplementary package for amd64 Debian systems, which was
+broken by enhancements to the packaging system in 1.1.
+
+[16] When decompressing a JPEG image using an output colorspace of
+JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR, or JCS_EXT_XRGB, libjpeg-turbo will
+now set the unused byte to 0xFF, which allows applications to interpret that
+byte as an alpha channel (0xFF = opaque).
+
+
+1.1.1
+=====
+
+[1] Fixed a 1-pixel error in row 0, column 21 of the luminance plane generated
+by tjEncodeYUV().
+
+[2] libjpeg-turbo's accelerated Huffman decoder previously ignored unexpected
+markers found in the middle of the JPEG data stream during decompression.  It
+will now hand off decoding of a particular block to the unaccelerated Huffman
+decoder if an unexpected marker is found, so that the unaccelerated Huffman
+decoder can generate an appropriate warning.
+
+[3] Older versions of MinGW64 prefixed symbol names with underscores by
+default, which differed from the behavior of 64-bit Visual C++.  MinGW64 1.0
+has adopted the behavior of 64-bit Visual C++ as the default, so to accommodate
+this, the libjpeg-turbo SIMD function names are no longer prefixed with an
+underscore when building with MinGW64.  This means that, when building
+libjpeg-turbo with older versions of MinGW64, you will now have to add
+-fno-leading-underscore to the CFLAGS.
+
+[4] Fixed a regression bug in the NSIS script that caused the Windows installer
+build to fail when using the Visual Studio IDE.
+
+[5] Fixed a bug in jpeg_read_coefficients() whereby it would not initialize
+cinfo->image_width and cinfo->image_height if libjpeg v7 or v8 emulation was
+enabled.  This specifically caused the jpegoptim program to fail if it was
+linked against a version of libjpeg-turbo that was built with libjpeg v7 or v8
+emulation.
+
+[6] Eliminated excessive I/O overhead that occurred when reading BMP files in
+cjpeg.
+
+[7] Eliminated errors in the output of cjpeg on Windows that occurred when the
+application was invoked using I/O redirection (cjpeg <inputfile >output.jpg).
+
+
+1.1.0
+=====
+
+[1] The algorithm used by the SIMD quantization function cannot produce correct
+results when the JPEG quality is >= 98 and the fast integer forward DCT is
+used.  Thus, the non-SIMD quantization function is now used for those cases,
+and libjpeg-turbo should now produce identical output to libjpeg v6b in all
+cases.
+
+[2] Despite the above, the fast integer forward DCT still degrades somewhat for
+JPEG qualities greater than 95, so the TurboJPEG wrapper will now automatically
+use the slow integer forward DCT when generating JPEG images of quality 96 or
+greater.  This reduces compression performance by as much as 15% for these
+high-quality images but is necessary to ensure that the images are perceptually
+lossless.  It also ensures that the library can avoid the performance pitfall
+created by [1].
+
+[3] Ported jpgtest.cxx to pure C to avoid the need for a C++ compiler.
+
+[4] Fixed visual artifacts in grayscale JPEG compression caused by a typo in
+the RGB-to-luminance lookup tables.
+
+[5] The Windows distribution packages now include the libjpeg run-time programs
+(cjpeg, etc.)
+
+[6] All packages now include jpgtest.
+
+[7] The TurboJPEG dynamic library now uses versioned symbols.
+
+[8] Added two new TurboJPEG API functions, tjEncodeYUV() and
+tjDecompressToYUV(), to replace the somewhat hackish TJ_YUV flag.
+
+
+1.0.90 (1.1 beta1)
+==================
+
+[1] Added emulation of the libjpeg v7 and v8 APIs and ABIs.  See
+README.md for more details.  This feature was sponsored by CamTrace SAS.
+
+[2] Created a new CMake-based build system for the Visual C++ and MinGW builds.
+
+[3] Grayscale bitmaps can now be compressed from/decompressed to using the
+TurboJPEG API.
+
+[4] jpgtest can now be used to test decompression performance with existing
+JPEG images.
+
+[5] If the default install prefix (/opt/libjpeg-turbo) is used, then
+'make install' now creates /opt/libjpeg-turbo/lib32 and
+/opt/libjpeg-turbo/lib64 sym links to duplicate the behavior of the binary
+packages.
+
+[6] All symbols in the libjpeg-turbo dynamic library are now versioned, even
+when the library is built with libjpeg v6b emulation.
+
+[7] Added arithmetic encoding and decoding support (can be disabled with
+configure or CMake options)
+
+[8] Added a TJ_YUV flag to the TurboJPEG API, which causes both the compressor
+and decompressor to output planar YUV images.
+
+[9] Added an extended version of tjDecompressHeader() to the TurboJPEG API,
+which allows the caller to determine the type of subsampling used in a JPEG
+image.
+
+[10] Added further protections against invalid Huffman codes.
+
+
+1.0.1
+=====
+
+[1] The Huffman decoder will now handle erroneous Huffman codes (for instance,
+from a corrupt JPEG image.)  Previously, these would cause libjpeg-turbo to
+crash under certain circumstances.
+
+[2] Fixed typo in SIMD dispatch routines that was causing 4:2:2 upsampling to
+be used instead of 4:2:0 when decompressing JPEG images using SSE2 code.
+
+[3] configure script will now automatically determine whether the
+INCOMPLETE_TYPES_BROKEN macro should be defined.
+
+
+1.0.0
+=====
+
+[1] 2983700: Further FreeBSD build tweaks (no longer necessary to specify
+--host when configuring on a 64-bit system)
+
+[2] Created symlinks in the Unix/Linux packages so that the TurboJPEG
+include file can always be found in /opt/libjpeg-turbo/include, the 32-bit
+static libraries can always be found in /opt/libjpeg-turbo/lib32, and the
+64-bit static libraries can always be found in /opt/libjpeg-turbo/lib64.
+
+[3] The Unix/Linux distribution packages now include the libjpeg run-time
+programs (cjpeg, etc.) and man pages.
+
+[4] Created a 32-bit supplementary package for amd64 Debian systems, which
+contains just the 32-bit libjpeg-turbo libraries.
+
+[5] Moved the libraries from */lib32 to */lib in the i386 Debian package.
+
+[6] Include distribution package for Cygwin
+
+[7] No longer necessary to specify --without-simd on non-x86 architectures, and
+unit tests now work on those architectures.
+
+
+0.0.93
+======
+
+[1] 2982659, Fixed x86-64 build on FreeBSD systems
+
+[2] 2988188: Added support for Windows 64-bit systems
+
+
+0.0.91
+======
+
+[1] Added documentation to .deb packages
+
+[2] 2968313: Fixed data corruption issues when decompressing large JPEG images
+and/or using buffered I/O with the libjpeg-turbo decompressor
+
+
+0.0.90
+======
+
+Initial release
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cjpeg.1 glmark2-2021.02/android/jni/src/libjpeg-turbo/cjpeg.1
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cjpeg.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/cjpeg.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,351 @@
+.TH CJPEG 1 "17 February 2016"
+.SH NAME
+cjpeg \- compress an image file to a JPEG file
+.SH SYNOPSIS
+.B cjpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B cjpeg
+compresses the named image file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+The currently supported input file formats are: PPM (PBMPLUS color
+format), PGM (PBMPLUS grayscale format), BMP, Targa, and RLE (Utah Raster
+Toolkit format).  (RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-quality " N[,...]"
+Scale quantization tables to adjust image quality.  Quality is 0 (worst) to
+100 (best); default is 75.  (See below for more info.)
+.TP
+.B \-grayscale
+Create monochrome JPEG file from color input.  Be sure to use this switch when
+compressing a grayscale BMP file, because
+.B cjpeg
+isn't bright enough to notice whether a BMP file uses only shades of gray.
+By saying
+.BR \-grayscale ,
+you'll get a smaller JPEG file that takes less time to process.
+.TP
+.B \-rgb
+Create RGB JPEG file.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.  Without this, default
+encoding parameters are used.
+.B \-optimize
+usually makes the JPEG file a little smaller, but
+.B cjpeg
+runs somewhat slower and needs much more memory.  Image quality and speed of
+decompression are unaffected by
+.BR \-optimize .
+.TP
+.B \-progressive
+Create progressive JPEG file (see below).
+.TP
+.B \-targa
+Input file is Targa format.  Targa files that contain an "identification"
+field will not be automatically recognized by
+.BR cjpeg ;
+for such files you must specify
+.B \-targa
+to make
+.B cjpeg
+treat the input as Targa format.
+For most Targa files, you won't need this switch.
+.PP
+The
+.B \-quality
+switch lets you trade off compressed file size against quality of the
+reconstructed image: the higher the quality setting, the larger the JPEG file,
+and the closer the output image will be to the original input.  Normally you
+want to use the lowest quality setting (smallest file) that decompresses into
+something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at
+.B \-quality
+75, then go up 5 or 10 counts at a time until you are happy with the output
+image.  (The optimal setting will vary from one image to another.)
+.PP
+.B \-quality
+100 will generate a quantization table of all 1's, minimizing loss in the
+quantization step (but there is still information loss in subsampling, as well
+as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+.PP
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try
+.B \-quality
+2 (or so) for some amusing Cubist effects.  (Note: quality
+values below about 25 generate 2-byte quantization tables, which are
+considered optional in the JPEG standard.
+.B cjpeg
+emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use
+.B \-baseline
+if you need to ensure compatibility at low quality values.)
+.PP
+The \fB-quality\fR option has been extended in this version of \fBcjpeg\fR to
+support separate quality settings for luminance and chrominance (or, in
+general, separate settings for every quantization table slot.)  The principle
+is the same as chrominance subsampling:  since the human eye is more sensitive
+to spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+.PP
+The \fB-quality\fR option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.
+More (or customized) quantization tables can be set with the \fB-qtables\fR
+option and assigned to components with the \fB-qslots\fR option (see the
+"wizard" switches below.)
+.PP
+JPEG files generated with separate luminance and chrominance quality are fully
+compliant with standard JPEG decoders.
+.PP
+.BR CAUTION:
+For this setting to be useful, be sure to pass an argument of \fB-sample 1x1\fR
+to \fBcjpeg\fR to disable chrominance subsampling.  Otherwise, the default
+subsampling level (2x2, AKA "4:2:0") will be used.
+.PP
+The
+.B \-progressive
+switch creates a "progressive JPEG" file.  In this type of JPEG file, the data
+is stored in multiple scans of increasing quality.  If the file is being
+transmitted over a slow communications link, the decoder can use the first
+scan to display a low-quality image very quickly, and can then improve the
+display with each subsequent scan.  The final image is exactly equivalent to a
+standard JPEG file of the same quality setting, and the total file size is
+about the same --- often a little smaller.
+.PP
+Switches for advanced users:
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.B Caution:
+arithmetic coded JPEG is not yet widely implemented, so many decoders will be
+unable to view an arithmetic coded JPEG file at all.
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+In libjpeg-turbo, the fast method is generally about 5-15% faster than the int
+method when using the x86/x86-64 SIMD extensions (results may vary with other
+SIMD implementations, or when using libjpeg-turbo without SIMD extensions.)
+For quality levels of 90 and below, there should be little or no perceptible
+difference between the two algorithms.  For quality levels above 90, however,
+the difference between the fast and the int methods becomes more pronounced.
+With quality=97, for instance, the fast method incurs generally about a 1-3 dB
+loss (in PSNR) relative to the int method, but this can be larger for some
+images.  Do not use the fast method with quality levels above 97.  The
+algorithm often degenerates at quality=98 and above and can actually produce a
+more lossy image than if lower quality levels had been used.  Also, in
+libjpeg-turbo, the fast method is not fully accelerated for quality levels
+above 97, so it will be slower than the int method.
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is mainly a legacy feature.  It does not produce significantly
+more accurate results than the int method, and it is much slower.  The float
+method may also give different results on different machines due to varying
+roundoff behavior, whereas the integer methods should give the same results on
+all machines.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.B \-restart 0
+(the default) means no restart markers.
+.TP
+.BI \-smooth " N"
+Smooth the input image to eliminate dithering noise.  N, ranging from 1 to
+100, indicates the strength of smoothing.  0 (the default) means no smoothing.
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memdst
+Compress to memory instead of a file.  This feature was implemented mainly as a
+way of testing the in-memory destination manager (jpeg_mem_dest()), but it is
+also useful for benchmarking, since it reduces the I/O overhead.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.PP
+The
+.B \-restart
+option inserts extra markers that allow a JPEG decoder to resynchronize after
+a transmission error.  Without restart markers, any damage to a compressed
+file will usually ruin the image from the point of the error to the end of the
+image; with restart markers, the damage is usually confined to the portion of
+the image up to the next restart marker.  Of course, the restart markers
+occupy extra space.  We recommend
+.B \-restart 1
+for images that will be transmitted across unreliable networks such as Usenet.
+.PP
+The
+.B \-smooth
+option filters the input to eliminate fine-scale noise.  This is often useful
+when converting dithered images to JPEG: a moderate smoothing factor of 10 to
+50 gets rid of dithering patterns in the input file, resulting in a smaller
+JPEG file and a better-looking image.  Too large a smoothing factor will
+visibly blur the image, however.
+.PP
+Switches for wizards:
+.TP
+.B \-baseline
+Force baseline-compatible quantization tables to be generated.  This clamps
+quantization values to 8 bits even at low quality settings.  (This switch is
+poorly named, since it does not ensure that the output is actually baseline
+JPEG.  For example, you can use
+.B \-baseline
+and
+.B \-progressive
+together.)
+.TP
+.BI \-qtables " file"
+Use the quantization tables given in the specified text file.
+.TP
+.BI \-qslots " N[,...]"
+Select which quantization table to use for each color component.
+.TP
+.BI \-sample " HxV[,...]"
+Set JPEG sampling factors for each color component.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, \fBdon't use them\fR.  These switches are
+documented further in the file wizard.txt.
+.SH EXAMPLES
+.LP
+This example compresses the PPM file foo.ppm with a quality factor of
+60 and saves the output as foo.jpg:
+.IP
+.B cjpeg \-quality
+.I 60 foo.ppm
+.B >
+.I foo.jpg
+.SH HINTS
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with
+.BR cjpeg 's
+.B \-quality
+and
+.B \-smooth
+options to get a satisfactory conversion.
+.B \-smooth 10
+or so is often helpful.
+.PP
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+.PP
+The
+.B \-optimize
+option to
+.B cjpeg
+is worth using when you are making a "final" version for posting or archiving.
+It's also a win when you are using low quality settings to make very small
+JPEG files; the percentage improvement is often a lot more than it is on
+larger files.  (At present,
+.B \-optimize
+mode is always selected when generating progressive JPEG files.)
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.
+.SH ISSUES
+Support for GIF input files was removed in cjpeg v6b due to concerns over
+the Unisys LZW patent.  Although this patent expired in 2006, cjpeg still
+lacks GIF support, for these historical reasons.  (Conversion of GIF files to
+JPEG is usually a bad idea anyway, since GIF is a 256-color format.)
+.PP
+Not all variants of BMP and Targa file formats are supported.
+.PP
+The
+.B \-targa
+switch is not a bug, it's a feature.  (It would be a bug if the Targa format
+designers had not been clueless.)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cjpeg.c glmark2-2021.02/android/jni/src/libjpeg-turbo/cjpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/cjpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/cjpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,644 @@
+/*
+ * cjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2013-2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG compressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      cjpeg [options]  inputfile outputfile
+ *      cjpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      cjpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)   string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This routine determines what format the input file is,
+ * and selects the appropriate input-reading module.
+ *
+ * To determine which family of input formats the file belongs to,
+ * we may look only at the first byte of the file, since C does not
+ * guarantee that more than one character can be pushed back with ungetc.
+ * Looking at additional bytes would require one of these approaches:
+ *     1) assume we can fseek() the input file (fails for piped input);
+ *     2) assume we can push back more than one character (works in
+ *        some C implementations, but unportable);
+ *     3) provide our own buffering (breaks input readers that want to use
+ *        stdio directly, such as the RLE library);
+ * or  4) don't put back the data, and modify the input_init methods to assume
+ *        they start reading after the start of file (also breaks RLE library).
+ * #1 is attractive for MS-DOS but is untenable on Unix.
+ *
+ * The most portable solution for file types that can't be identified by their
+ * first byte is to make the user tell us what they are.  This is also the
+ * only approach for "raw" file types that contain only arbitrary values.
+ * We presently apply this method for Targa files.  Most of the time Targa
+ * files start with 0x00, so we recognize that case.  Potentially, however,
+ * a Targa file could start with any byte value (byte 0 is the length of the
+ * seldom-used ID field), so we provide a switch to force Targa input mode.
+ */
+
+static boolean is_targa;        /* records user -targa switch */
+
+
+LOCAL(cjpeg_source_ptr)
+select_file_type (j_compress_ptr cinfo, FILE *infile)
+{
+  int c;
+
+  if (is_targa) {
+#ifdef TARGA_SUPPORTED
+    return jinit_read_targa(cinfo);
+#else
+    ERREXIT(cinfo, JERR_TGA_NOTCOMP);
+#endif
+  }
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+  if (ungetc(c, infile) == EOF)
+    ERREXIT(cinfo, JERR_UNGETC_FAILED);
+
+  switch (c) {
+#ifdef BMP_SUPPORTED
+  case 'B':
+    return jinit_read_bmp(cinfo);
+#endif
+#ifdef GIF_SUPPORTED
+  case 'G':
+    return jinit_read_gif(cinfo);
+#endif
+#ifdef PPM_SUPPORTED
+  case 'P':
+    return jinit_read_ppm(cinfo);
+#endif
+#ifdef RLE_SUPPORTED
+  case 'R':
+    return jinit_read_rle(cinfo);
+#endif
+#ifdef TARGA_SUPPORTED
+  case 0x00:
+    return jinit_read_targa(cinfo);
+#endif
+  default:
+    ERREXIT(cinfo, JERR_UNKNOWN_FORMAT);
+    break;
+  }
+
+  return NULL;                  /* suppress compiler warnings */
+}
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+boolean memdst;                 /* for -memdst switch */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -quality N[,...]   Compression quality (0..100; 5-95 is most useful range,\n");
+  fprintf(stderr, "                     default is 75)\n");
+  fprintf(stderr, "  -grayscale     Create monochrome JPEG file\n");
+  fprintf(stderr, "  -rgb           Create RGB JPEG file\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Input file is Targa format (usually not needed)\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  fprintf(stderr, "  -smooth N      Smooth dithered input (N=1..100 is strength)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memdst        Compress to memory instead of file (useful for benchmarking)\n");
+#endif
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+  fprintf(stderr, "  -baseline      Force baseline quantization tables\n");
+  fprintf(stderr, "  -qtables file  Use quantization tables given in file\n");
+  fprintf(stderr, "  -qslots N[,...]    Set component quantization tables\n");
+  fprintf(stderr, "  -sample HxV[,...]  Set component sampling factors\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean force_baseline;
+  boolean simple_progressive;
+  char *qualityarg = NULL;      /* saves -quality parm if any */
+  char *qtablefile = NULL;      /* saves -qtables filename if any */
+  char *qslotsarg = NULL;       /* saves -qslots parm if any */
+  char *samplearg = NULL;       /* saves -sample parm if any */
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+
+  force_baseline = FALSE;       /* by default, allow 16-bit quantizers */
+  simple_progressive = FALSE;
+  is_targa = FALSE;
+  outfilename = NULL;
+  memdst = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "baseline", 1)) {
+      /* Force baseline-compatible output (8-bit quantizer values). */
+      force_baseline = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select DCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force a monochrome JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+
+    } else if (keymatch(arg, "rgb", 3)) {
+      /* Force an RGB JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_RGB);
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "progressive", 1)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "memdst", 2)) {
+      /* Use in-memory destination manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memdst = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory destination manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "quality", 1)) {
+      /* Quality ratings (quantization table scaling factors). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qualityarg = argv[argn];
+
+    } else if (keymatch(arg, "qslots", 2)) {
+      /* Quantization table slot numbers. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qslotsarg = argv[argn];
+      /* Must delay setting qslots until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default quant table numbers.
+       */
+
+    } else if (keymatch(arg, "qtables", 2)) {
+      /* Quantization tables fetched from file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qtablefile = argv[argn];
+      /* We postpone actually reading the file in case -quality comes later. */
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int) lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int) lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "sample", 2)) {
+      /* Set sampling factors. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      samplearg = argv[argn];
+      /* Must delay setting sample factors until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default sampling factors.
+       */
+
+    } else if (keymatch(arg, "scans", 4)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "smooth", 2)) {
+      /* Set input smoothing factor. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+        usage();
+      if (val < 0 || val > 100)
+        usage();
+      cinfo->smoothing_factor = val;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Input file is Targa format. */
+      is_targa = TRUE;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+    /* Set quantization tables for selected quality. */
+    /* Some or all may be overridden if -qtables is present. */
+    if (qualityarg != NULL)     /* process -quality if it was present */
+      if (! set_quality_ratings(cinfo, qualityarg, force_baseline))
+        usage();
+
+    if (qtablefile != NULL)     /* process -qtables if it was present */
+      if (! read_quant_tables(cinfo, qtablefile, force_baseline))
+        usage();
+
+    if (qslotsarg != NULL)      /* process -qslots if it was present */
+      if (! set_quant_slots(cinfo, qslotsarg))
+        usage();
+
+    if (samplearg != NULL)      /* process -sample if it was present */
+      if (! set_sample_factors(cinfo, samplearg))
+        usage();
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_compress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  cjpeg_source_ptr src_mgr;
+  FILE *input_file;
+  FILE *output_file = NULL;
+  unsigned char *outbuffer = NULL;
+  unsigned long outsize = 0;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "cjpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG compression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_compress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Initialize JPEG parameters.
+   * Much of this may be overridden later.
+   * In particular, we don't yet know the input file's color space,
+   * but we need to provide some value for jpeg_set_defaults() to work.
+   */
+
+  cinfo.in_color_space = JCS_RGB; /* arbitrary guess */
+  jpeg_set_defaults(&cinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  if (!memdst) {
+    /* Must have either -outfile switch or explicit output file name */
+    if (outfilename == NULL) {
+      if (file_index != argc-2) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+      outfilename = argv[file_index+1];
+    } else {
+      if (file_index != argc-1) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else if (!memdst) {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Figure out the input file format, and set up to read it. */
+  src_mgr = select_file_type(&cinfo, input_file);
+  src_mgr->input_file = input_file;
+
+  /* Read the input file header to obtain file size & colorspace. */
+  (*src_mgr->start_input) (&cinfo, src_mgr);
+
+  /* Now that we know input colorspace, fix colorspace-dependent defaults */
+  jpeg_default_colorspace(&cinfo);
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memdst)
+    jpeg_mem_dest(&cinfo, &outbuffer, &outsize);
+  else
+#endif
+    jpeg_stdio_dest(&cinfo, output_file);
+
+  /* Start compressor */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Process data */
+  while (cinfo.next_scanline < cinfo.image_height) {
+    num_scanlines = (*src_mgr->get_pixel_rows) (&cinfo, src_mgr);
+    (void) jpeg_write_scanlines(&cinfo, src_mgr->buffer, num_scanlines);
+  }
+
+  /* Finish compression and release memory */
+  (*src_mgr->finish_input) (&cinfo, src_mgr);
+  jpeg_finish_compress(&cinfo);
+  jpeg_destroy_compress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout && output_file != NULL)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  if (memdst) {
+    fprintf(stderr, "Compressed size:  %lu bytes\n", outsize);
+    if (outbuffer != NULL)
+      free(outbuffer);
+  }
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/coderules.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/coderules.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/coderules.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/coderules.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,78 @@
+IJG JPEG LIBRARY:  CODING RULES
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-1996, Thomas G. Lane.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+Since numerous people will be contributing code and bug fixes, it's important
+to establish a common coding style.  The goal of using similar coding styles
+is much more important than the details of just what that style is.
+
+In general we follow the recommendations of "Recommended C Style and Coding
+Standards" revision 6.1 (Cannon et al. as modified by Spencer, Keppel and
+Brader).  This document is available in the IJG FTP archive (see
+jpeg/doc/cstyle.ms.tbl.Z, or cstyle.txt.Z for those without nroff/tbl).
+
+Block comments should be laid out thusly:
+
+/*
+ *  Block comments in this style.
+ */
+
+We indent statements in K&R style, e.g.,
+        if (test) {
+          then-part;
+        } else {
+          else-part;
+        }
+with two spaces per indentation level.  (This indentation convention is
+handled automatically by GNU Emacs and many other text editors.)
+
+Multi-word names should be written in lower case with underscores, e.g.,
+multi_word_name (not multiWordName).  Preprocessor symbols and enum constants
+are similar but upper case (MULTI_WORD_NAME).  Names should be unique within
+the first fifteen characters.
+
+Note that each function definition must begin with GLOBAL(type), LOCAL(type),
+or METHODDEF(type).  These macros expand to "static type" or just "type" as
+appropriate.  They provide a readable indication of the routine's usage and
+can readily be changed for special needs.  (For instance, special linkage
+keywords can be inserted for use in Windows DLLs.)
+
+A similar solution is used for external function declarations (see the EXTERN
+macro.)
+
+
+The JPEG library is intended to be used within larger programs.  Furthermore,
+we want it to be reentrant so that it can be used by applications that process
+multiple images concurrently.  The following rules support these requirements:
+
+1. Avoid direct use of file I/O, "malloc", error report printouts, etc;
+pass these through the common routines provided.
+
+2. Minimize global namespace pollution.  Functions should be declared static
+wherever possible.  (Note that our method-based calling conventions help this
+a lot: in many modules only the initialization function will ever need to be
+called directly, so only that function need be externally visible.)  All
+global function names should begin with "jpeg_".
+
+3. Don't use global variables; anything that must be used in another module
+should be in the common data structures.
+
+4. Don't use static variables except for read-only constant tables.  Variables
+that should be private to a module can be placed into private structures (see
+the system architecture document, structure.txt).
+
+5. Source file names should begin with "j" for files that are part of the
+library proper; source files that are not part of the library, such as cjpeg.c
+and djpeg.c, do not begin with "j".  Keep compression and decompression code in
+separate source files --- some applications may want only one half of the
+library.
+
+Note: these rules (particularly #4) are not followed religiously in the
+modules that are used in cjpeg/djpeg but are not part of the JPEG library
+proper.  Those modules are not really intended to be used in other
+applications.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/config.h glmark2-2021.02/android/jni/src/libjpeg-turbo/config.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/config.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/config.h	1970-01-01 08:00:00.000000000 +0800
@@ -1,137 +0,0 @@
-/* config.h.  Generated from config.h.in by configure.  */
-/* config.h.in.  Generated from configure.ac by autoheader.  */
-
-/* Build number */
-#define BUILD "20120626"
-
-/* Support arithmetic encoding */
-#define C_ARITH_CODING_SUPPORTED 1
-
-/* Support arithmetic decoding */
-#define D_ARITH_CODING_SUPPORTED 1
-
-/* Define to 1 if you have the <dlfcn.h> header file. */
-#define HAVE_DLFCN_H 1
-
-/* Define to 1 if you have the <inttypes.h> header file. */
-#define HAVE_INTTYPES_H 1
-
-/* Define to 1 if you have the <jni.h> header file. */
-/* #undef HAVE_JNI_H */
-
-/* Define to 1 if you have the `memcpy' function. */
-#define HAVE_MEMCPY 1
-
-/* Define to 1 if you have the <memory.h> header file. */
-#define HAVE_MEMORY_H 1
-
-/* Define to 1 if you have the `memset' function. */
-#define HAVE_MEMSET 1
-
-/* Define if your compiler supports prototypes */
-#define HAVE_PROTOTYPES 1
-
-/* Define to 1 if you have the <stddef.h> header file. */
-#define HAVE_STDDEF_H 1
-
-/* Define to 1 if you have the <stdint.h> header file. */
-#define HAVE_STDINT_H 1
-
-/* Define to 1 if you have the <stdlib.h> header file. */
-#define HAVE_STDLIB_H 1
-
-/* Define to 1 if you have the <strings.h> header file. */
-#define HAVE_STRINGS_H 1
-
-/* Define to 1 if you have the <string.h> header file. */
-#define HAVE_STRING_H 1
-
-/* Define to 1 if you have the <sys/stat.h> header file. */
-#define HAVE_SYS_STAT_H 1
-
-/* Define to 1 if you have the <sys/types.h> header file. */
-#define HAVE_SYS_TYPES_H 1
-
-/* Define to 1 if you have the <unistd.h> header file. */
-#define HAVE_UNISTD_H 1
-
-/* Define to 1 if the system has the type `unsigned char'. */
-#define HAVE_UNSIGNED_CHAR 1
-
-/* Define to 1 if the system has the type `unsigned short'. */
-#define HAVE_UNSIGNED_SHORT 1
-
-/* Compiler does not support pointers to undefined structures. */
-/* #undef INCOMPLETE_TYPES_BROKEN */
-
-/* How to obtain function inlining. */
-#define INLINE __attribute__((always_inline))
-
-/* libjpeg API version */
-#define JPEG_LIB_VERSION 62
-
-/* libjpeg-turbo version */
-#define LIBJPEG_TURBO_VERSION 1.2.0
-
-/* Define to the sub-directory in which libtool stores uninstalled libraries.
-   */
-#define LT_OBJDIR ".libs/"
-
-/* Define if you have BSD-like bzero and bcopy */
-/* #undef NEED_BSD_STRINGS */
-
-/* Define if you need short function names */
-/* #undef NEED_SHORT_EXTERNAL_NAMES */
-
-/* Define if you have sys/types.h */
-#define NEED_SYS_TYPES_H 1
-
-/* Name of package */
-#define PACKAGE "libjpeg-turbo"
-
-/* Define to the address where bug reports for this package should be sent. */
-#define PACKAGE_BUGREPORT ""
-
-/* Define to the full name of this package. */
-#define PACKAGE_NAME "libjpeg-turbo"
-
-/* Define to the full name and version of this package. */
-#define PACKAGE_STRING "libjpeg-turbo 1.2.0"
-
-/* Define to the one symbol short name of this package. */
-#define PACKAGE_TARNAME "libjpeg-turbo"
-
-/* Define to the home page for this package. */
-#define PACKAGE_URL ""
-
-/* Define to the version of this package. */
-#define PACKAGE_VERSION "1.2.0"
-
-/* Define if shift is unsigned */
-/* #undef RIGHT_SHIFT_IS_UNSIGNED */
-
-/* Define to 1 if you have the ANSI C header files. */
-#define STDC_HEADERS 1
-
-/* Version number of package */
-#define VERSION "1.2.0"
-
-/* Use accelerated SIMD routines. */
-#define WITH_SIMD 1
-
-/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
-#ifndef __CHAR_UNSIGNED__
-/* # undef __CHAR_UNSIGNED__ */
-#endif
-
-/* Define to empty if `const' does not conform to ANSI C. */
-/* #undef const */
-
-/* Define to `__inline__' or `__inline' if that's what the C compiler
-   calls it, or to nothing if 'inline' is not supported under any name.  */
-#ifndef __cplusplus
-/* #undef inline */
-#endif
-
-/* Define to `unsigned int' if <sys/types.h> does not define. */
-/* #undef size_t */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/configure.ac glmark2-2021.02/android/jni/src/libjpeg-turbo/configure.ac
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/configure.ac	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/configure.ac	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,595 @@
+#                                               -*- Autoconf -*-
+# Process this file with autoconf to produce a configure script.
+
+AC_PREREQ([2.56])
+AC_INIT([libjpeg-turbo], [1.4.90])
+
+AM_INIT_AUTOMAKE([-Wall foreign dist-bzip2])
+AC_PREFIX_DEFAULT(/opt/libjpeg-turbo)
+
+m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
+
+# Checks for programs.
+SAVED_CFLAGS=${CFLAGS}
+SAVED_CPPFLAGS=${CPPFLAGS}
+AC_PROG_CPP
+AC_PROG_CC
+m4_ifdef([AM_PROG_AR], [AM_PROG_AR])
+AM_PROG_AS
+AM_PROG_CC_C_O
+AC_PROG_INSTALL
+AC_PROG_LIBTOOL
+AC_PROG_LN_S
+
+AC_ARG_WITH([build-date], [Use custom build string to enable reproducible builds (default: YYMMDD)],
+  [BUILD="$with_build_date"],
+  [BUILD=`date +%Y%m%d`])
+
+PKG_PROG_PKG_CONFIG
+
+# When the prefix is /opt/libjpeg-turbo, we assume that an "official" binary is
+# being created, and thus we install things into specific locations.
+
+old_prefix=${prefix}
+if test "x$prefix" = "xNONE" -a "x$ac_default_prefix" != "x"; then
+  prefix=$ac_default_prefix
+fi
+DATADIR=`eval echo ${datadir}`
+DATADIR=`eval echo $DATADIR`
+if test "$DATADIR" = "/opt/libjpeg-turbo/share"; then
+  datadir='${prefix}'
+fi
+DATADIR=`eval echo ${datarootdir}`
+DATADIR=`eval echo $DATADIR`
+if test "$DATADIR" = "/opt/libjpeg-turbo/share"; then
+  datarootdir='${prefix}'
+fi
+DOCDIR=`eval echo ${docdir}`
+DOCDIR=`eval echo $DOCDIR`
+if test "$DOCDIR" = "/opt/libjpeg-turbo/doc/libjpeg-turbo"; then
+  docdir='${datadir}/doc'
+fi
+
+old_exec_prefix=${exec_prefix}
+if test "x$exec_prefix" = "xNONE"; then
+  exec_prefix=${prefix}
+fi
+
+AC_CHECK_SIZEOF(size_t)
+
+if test "x${libdir}" = 'x${exec_prefix}/lib' -o "x${libdir}" = 'x${prefix}/lib'; then
+  LIBDIR=`eval echo ${libdir}`
+  LIBDIR=`eval echo $LIBDIR`
+  if test "$LIBDIR" = "/opt/libjpeg-turbo/lib"; then
+    case $host_os in
+      darwin*)
+        ;;
+      *)
+        if test "${ac_cv_sizeof_size_t}" = "8"; then
+          libdir='${exec_prefix}/lib64'
+        elif test "${ac_cv_sizeof_size_t}" = "4"; then
+          libdir='${exec_prefix}/lib32'
+        fi
+        ;;
+    esac
+  fi
+fi
+exec_prefix=${old_exec_prefix}
+prefix=${old_prefix}
+
+# Check whether compiler supports pointers to undefined structures
+AC_MSG_CHECKING(whether compiler supports pointers to undefined structures)
+AC_TRY_COMPILE([ typedef struct undefined_structure *undef_struct_ptr; ], ,
+  AC_MSG_RESULT(yes),
+  [AC_MSG_RESULT(no)
+   AC_DEFINE([INCOMPLETE_TYPES_BROKEN], [1],
+     [Compiler does not support pointers to undefined structures.])])
+
+if test "x${GCC}" = "xyes"; then
+  if test "x${SAVED_CFLAGS}" = "x"; then
+    CFLAGS=-O3
+  fi
+  if test "x${SAVED_CPPFLAGS}" = "x"; then
+    CPPFLAGS=-Wall
+  fi
+fi
+
+AC_CHECK_DECL([__SUNPRO_C], [SUNCC="yes"], [SUNCC="no"])
+if test "x${SUNCC}" = "xyes"; then
+  if test "x${SAVED_CFLAGS}" = "x"; then
+    CFLAGS=-xO5
+  fi
+fi
+
+# Checks for libraries.
+
+# Checks for header files.
+AC_HEADER_STDC
+AC_CHECK_HEADERS([stddef.h stdlib.h locale.h string.h])
+AC_CHECK_HEADER([sys/types.h],
+  AC_DEFINE([NEED_SYS_TYPES_H], 1, [Define if you need to include <sys/types.h> to get size_t.]))
+
+# Checks for typedefs, structures, and compiler characteristics.
+AC_C_CONST
+AC_C_CHAR_UNSIGNED
+AC_C_INLINE
+AC_TYPE_SIZE_T
+AC_CHECK_TYPES([unsigned char, unsigned short])
+
+AC_MSG_CHECKING([if right shift is signed])
+AC_TRY_RUN(
+  [#include <stdio.h>
+   int is_shifting_signed (long arg) {
+     long res = arg >> 4;
+
+     if (res == -0x7F7E80CL)
+       return 1; /* right shift is signed */
+
+     /* see if unsigned-shift hack will fix it. */
+     /* we can't just test exact value since it depends on width of long... */
+     res |= (~0L) << (32-4);
+     if (res == -0x7F7E80CL)
+       return 0; /* right shift is unsigned */
+
+     printf("Right shift isn't acting as I expect it to.\n");
+     printf("I fear the JPEG software will not work at all.\n\n");
+     return 0; /* try it with unsigned anyway */
+   }
+   int main (void) {
+     exit(is_shifting_signed(-0x7F7E80B1L));
+   }],
+  [AC_MSG_RESULT(no)
+   AC_DEFINE([RIGHT_SHIFT_IS_UNSIGNED], 1,
+     [Define if your (broken) compiler shifts signed values as if they were unsigned.])],
+  [AC_MSG_RESULT(yes)],
+  [AC_MSG_RESULT(Assuming that right shift is signed on target machine.)])
+
+# Checks for library functions.
+AC_CHECK_FUNCS([memset memcpy], [],
+  [AC_DEFINE([NEED_BSD_STRINGS], 1,
+     [Define if you have BSD-like bzero and bcopy in <strings.h> rather than memset/memcpy in <string.h>.])])
+
+AC_MSG_CHECKING([libjpeg API version])
+AC_ARG_VAR(JPEG_LIB_VERSION, [libjpeg API version (62, 70, or 80)])
+if test "x$JPEG_LIB_VERSION" = "x"; then
+  AC_ARG_WITH([jpeg7],
+    AC_HELP_STRING([--with-jpeg7],
+      [Emulate libjpeg v7 API/ABI (this makes libjpeg-turbo backward incompatible with libjpeg v6b.)]))
+  AC_ARG_WITH([jpeg8],
+    AC_HELP_STRING([--with-jpeg8],
+      [Emulate libjpeg v8 API/ABI (this makes libjpeg-turbo backward incompatible with libjpeg v6b.)]))
+  if test "x${with_jpeg8}" = "xyes"; then
+    JPEG_LIB_VERSION=80
+  else
+    if test "x${with_jpeg7}" = "xyes"; then
+      JPEG_LIB_VERSION=70
+    else
+      JPEG_LIB_VERSION=62
+    fi
+  fi
+fi
+JPEG_LIB_VERSION_DECIMAL=`expr $JPEG_LIB_VERSION / 10`.`expr $JPEG_LIB_VERSION % 10`
+AC_SUBST(JPEG_LIB_VERSION_DECIMAL)
+AC_MSG_RESULT([$JPEG_LIB_VERSION_DECIMAL])
+AC_DEFINE_UNQUOTED(JPEG_LIB_VERSION, [$JPEG_LIB_VERSION],
+  [libjpeg API version])
+
+AC_ARG_VAR(SO_MAJOR_VERSION,
+  [Major version of the libjpeg-turbo shared library (default is determined by the API version)])
+AC_ARG_VAR(SO_MINOR_VERSION,
+  [Minor version of the libjpeg-turbo shared library (default is determined by the API version)])
+if test "x$SO_MAJOR_VERSION" = "x"; then
+  case "$JPEG_LIB_VERSION" in
+    62)  SO_MAJOR_VERSION=$JPEG_LIB_VERSION ;;
+    *)   SO_MAJOR_VERSION=`expr $JPEG_LIB_VERSION / 10` ;;
+  esac
+fi
+if test "x$SO_MINOR_VERSION" = "x"; then
+  case "$JPEG_LIB_VERSION" in
+    80)  SO_MINOR_VERSION=2 ;;
+    *)   SO_MINOR_VERSION=0 ;;
+  esac
+fi
+
+RPM_CONFIG_ARGS=
+
+# Memory source/destination managers
+SO_AGE=1
+MEM_SRCDST_FUNCTIONS=
+if test "x${with_jpeg8}" != "xyes"; then
+  AC_MSG_CHECKING([whether to include in-memory source/destination managers])
+  AC_ARG_WITH([mem-srcdst],
+    AC_HELP_STRING([--without-mem-srcdst],
+      [Do not include in-memory source/destination manager functions when emulating the libjpeg v6b or v7 API/ABI]))
+  if test "x$with_mem_srcdst" != "xno"; then
+    AC_MSG_RESULT(yes)
+    AC_DEFINE([MEM_SRCDST_SUPPORTED], [1],
+      [Support in-memory source/destination managers])
+    SO_AGE=2
+    MEM_SRCDST_FUNCTIONS="global:  jpeg_mem_dest;  jpeg_mem_src;";
+  else
+    AC_MSG_RESULT(no)
+    RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-mem-srcdst"
+  fi
+fi
+
+AC_MSG_CHECKING([libjpeg shared library version])
+AC_MSG_RESULT([$SO_MAJOR_VERSION.$SO_AGE.$SO_MINOR_VERSION])
+LIBTOOL_CURRENT=`expr $SO_MAJOR_VERSION + $SO_AGE`
+AC_SUBST(LIBTOOL_CURRENT)
+AC_SUBST(SO_MAJOR_VERSION)
+AC_SUBST(SO_MINOR_VERSION)
+AC_SUBST(SO_AGE)
+AC_SUBST(MEM_SRCDST_FUNCTIONS)
+
+AC_DEFINE_UNQUOTED(LIBJPEG_TURBO_VERSION, [$VERSION], [libjpeg-turbo version])
+
+VERSION_SCRIPT=yes
+AC_ARG_ENABLE([ld-version-script],
+  AS_HELP_STRING([--disable-ld-version-script],
+    [Disable linker version script for libjpeg-turbo (default is to use linker version script if the linker supports it)]),
+  [VERSION_SCRIPT=$enableval], [])
+
+AC_MSG_CHECKING([whether the linker supports version scripts])
+SAVED_LDFLAGS="$LDFLAGS"
+LDFLAGS="$LDFLAGS -Wl,--version-script,conftest.map"
+cat > conftest.map <<EOF
+VERS_1 {
+  global: *;
+};
+EOF
+AC_LINK_IFELSE([AC_LANG_PROGRAM([], [])],
+  [VERSION_SCRIPT_FLAG=-Wl,--version-script,;
+   AC_MSG_RESULT([yes (GNU style)])],
+  [])
+if test "x$VERSION_SCRIPT_FLAG" = "x"; then
+  LDFLAGS="$SAVED_LDFLAGS -Wl,-M,conftest.map"
+  AC_LINK_IFELSE([AC_LANG_PROGRAM([], [])],
+    [VERSION_SCRIPT_FLAG=-Wl,-M,;
+     AC_MSG_RESULT([yes (Sun style)])],
+    [])
+fi
+if test "x$VERSION_SCRIPT_FLAG" = "x"; then
+  VERSION_SCRIPT=no
+  AC_MSG_RESULT(no)
+fi
+LDFLAGS="$SAVED_LDFLAGS"
+
+AC_MSG_CHECKING([whether to use version script when building libjpeg-turbo])
+AC_MSG_RESULT($VERSION_SCRIPT)
+
+AM_CONDITIONAL(VERSION_SCRIPT, test "x$VERSION_SCRIPT" = "xyes")
+AC_SUBST(VERSION_SCRIPT_FLAG)
+
+# Check for non-broken inline under various spellings
+AC_MSG_CHECKING(for inline)
+ljt_cv_inline=""
+AC_TRY_COMPILE(, [} inline __attribute__((always_inline)) int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="inline __attribute__((always_inline))",
+AC_TRY_COMPILE(, [} __inline__ int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline__",
+AC_TRY_COMPILE(, [} __inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline",
+AC_TRY_COMPILE(, [} inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="inline"))))
+AC_MSG_RESULT($ljt_cv_inline)
+AC_DEFINE_UNQUOTED([INLINE],[$ljt_cv_inline],[How to obtain function inlining.])
+
+# Arithmetic coding support
+AC_MSG_CHECKING([whether to include arithmetic encoding support])
+AC_ARG_WITH([arith-enc],
+  AC_HELP_STRING([--without-arith-enc],
+    [Do not include arithmetic encoding support]))
+if test "x$with_12bit" = "xyes"; then
+  with_arith_enc=no
+fi
+if test "x$with_arith_enc" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-arith-enc"
+else
+  AC_DEFINE([C_ARITH_CODING_SUPPORTED], [1], [Support arithmetic encoding])
+  AC_MSG_RESULT(yes)
+fi
+AM_CONDITIONAL([WITH_ARITH_ENC], [test "x$with_arith_enc" != "xno"])
+
+AC_MSG_CHECKING([whether to include arithmetic decoding support])
+AC_ARG_WITH([arith-dec],
+  AC_HELP_STRING([--without-arith-dec],
+    [Do not include arithmetic decoding support]))
+if test "x$with_12bit" = "xyes"; then
+  with_arith_dec=no
+fi
+if test "x$with_arith_dec" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-arith-dec"
+else
+  AC_DEFINE([D_ARITH_CODING_SUPPORTED], [1], [Support arithmetic decoding])
+  AC_MSG_RESULT(yes)
+fi
+AM_CONDITIONAL([WITH_ARITH_DEC], [test "x$with_arith_dec" != "xno"])
+
+AM_CONDITIONAL([WITH_ARITH],
+  [test "x$with_arith_dec" != "xno" -o "x$with_arith_enc" != "xno"])
+
+# 12-bit component support
+AC_MSG_CHECKING([whether to use 12-bit samples])
+AC_ARG_WITH([12bit],
+  AC_HELP_STRING([--with-12bit], [Encode/decode JPEG images with 12-bit samples (implies --without-simd --without-turbojpeg --without-arith-dec --without-arith-enc)]))
+if test "x$with_12bit" = "xyes"; then
+  AC_DEFINE([BITS_IN_JSAMPLE], [12], [use 8 or 12])
+  AC_MSG_RESULT(yes)
+else
+  AC_MSG_RESULT(no)
+fi
+AM_CONDITIONAL([WITH_12BIT], [test "x$with_12bit" = "xyes"])
+
+# TurboJPEG support
+AC_MSG_CHECKING([whether to build TurboJPEG C wrapper])
+AC_ARG_WITH([turbojpeg],
+  AC_HELP_STRING([--without-turbojpeg],
+    [Do not include the TurboJPEG wrapper library and associated test programs]))
+if test "x$with_12bit" = "xyes"; then
+  with_turbojpeg=no
+fi
+if test "x$with_turbojpeg" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-turbojpeg"
+else
+  AC_MSG_RESULT(yes)
+fi
+
+# Java support
+AC_ARG_VAR(JAVAC, [Java compiler command (default: javac)])
+if test "x$JAVAC" = "x"; then
+  JAVAC=javac
+fi
+AC_SUBST(JAVAC)
+AC_ARG_VAR(JAVACFLAGS, [Java compiler flags])
+AC_SUBST(JAVACFLAGS)
+AC_ARG_VAR(JAR, [Java archive command (default: jar)])
+if test "x$JAR" = "x"; then
+  JAR=jar
+fi
+AC_SUBST(JAR)
+AC_ARG_VAR(JAVA, [Java runtime command (default: java)])
+if test "x$JAVA" = "x"; then
+  JAVA=java
+fi
+AC_SUBST(JAVA)
+AC_ARG_VAR(JNI_CFLAGS,
+  [C compiler flags needed to include jni.h (default: -I/System/Library/Frameworks/JavaVM.framework/Headers on OS X, '-I/usr/java/include -I/usr/java/include/solaris' on Solaris, and '-I/usr/java/default/include -I/usr/java/default/include/linux' on Linux)])
+
+AC_MSG_CHECKING([whether to build TurboJPEG Java wrapper])
+AC_ARG_WITH([java],
+  AC_HELP_STRING([--with-java], [Build Java wrapper for the TurboJPEG library]))
+if test "x$with_12bit" = "xyes" -o "x$with_turbojpeg" = "xno"; then
+  with_java=no
+fi
+
+WITH_JAVA=0
+if test "x$with_java" = "xyes"; then
+  AC_MSG_RESULT(yes)
+
+  case $host_os in
+    darwin*)
+      DEFAULT_JNI_CFLAGS=-I/System/Library/Frameworks/JavaVM.framework/Headers
+      ;;
+    solaris*)
+      DEFAULT_JNI_CFLAGS='-I/usr/java/include -I/usr/java/include/solaris'
+      ;;
+    linux*)
+      DEFAULT_JNI_CFLAGS='-I/usr/java/default/include -I/usr/java/default/include/linux'
+      ;;
+  esac
+  if test "x$JNI_CFLAGS" = "x"; then
+    JNI_CFLAGS=$DEFAULT_JNI_CFLAGS
+  fi
+
+  SAVE_CPPFLAGS=${CPPFLAGS}
+  CPPFLAGS="${CPPFLAGS} ${JNI_CFLAGS}"
+  AC_CHECK_HEADERS([jni.h], [DUMMY=1],
+    [AC_MSG_ERROR([Could not find JNI header file])])
+  CPPFLAGS=${SAVE_CPPFLAGS}
+  AC_SUBST(JNI_CFLAGS)
+
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --with-java"
+  JAVA_RPM_CONTENTS_1='%dir %{_datadir}/classes'
+  JAVA_RPM_CONTENTS_2=%{_datadir}/classes/turbojpeg.jar
+  WITH_JAVA=1
+else
+  AC_MSG_RESULT(no)
+fi
+AM_CONDITIONAL([WITH_JAVA], [test "x$with_java" = "xyes"])
+AC_SUBST(WITH_JAVA)
+AC_SUBST(JAVA_RPM_CONTENTS_1)
+AC_SUBST(JAVA_RPM_CONTENTS_2)
+
+# optionally force using gas-preprocessor.pl for compatibility testing
+AC_ARG_WITH([gas-preprocessor],
+  AC_HELP_STRING([--with-gas-preprocessor],
+    [Force using gas-preprocessor.pl on ARM.]))
+if test "x${with_gas_preprocessor}" = "xyes"; then
+  case $host_os in
+    darwin*)
+      CCAS="gas-preprocessor.pl -fix-unreq $CC"
+      ;;
+    *)
+      CCAS="gas-preprocessor.pl -no-fix-unreq $CC"
+      ;;
+  esac
+  AC_SUBST([CCAS])
+fi
+
+# SIMD is optional
+AC_ARG_WITH([simd],
+  AC_HELP_STRING([--without-simd], [Do not include SIMD extensions]))
+if test "x$with_12bit" = "xyes"; then
+  with_simd=no
+fi
+if test "x${with_simd}" != "xno"; then
+  require_simd=no
+  if test "x${with_simd}" = "xyes"; then
+    require_simd=yes
+  fi
+  # Check if we're on a supported CPU
+  AC_MSG_CHECKING([if we have SIMD optimisations for cpu type])
+  case "$host_cpu" in
+    x86_64 | amd64)
+      AC_MSG_RESULT([yes (x86_64)])
+      AC_PROG_NASM
+      simd_arch=x86_64
+      ;;
+    i*86 | x86 | ia32)
+      AC_MSG_RESULT([yes (i386)])
+      AC_PROG_NASM
+      simd_arch=i386
+      ;;
+    arm*)
+      AC_MSG_RESULT([yes (arm)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE(
+        [if test "x$ac_use_gas_preprocessor" = "xyes"; then
+           AC_MSG_RESULT([yes (with gas-preprocessor)])
+         else
+           AC_MSG_RESULT([yes])
+         fi
+         simd_arch=arm],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    aarch64*)
+      AC_MSG_RESULT([yes (arm64)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_ARM64_ASSEMBLER_IFELSE(
+        [if test "x$ac_use_gas_preprocessor" = "xyes"; then
+           AC_MSG_RESULT([yes (with gas-preprocessor)])
+         else
+           AC_MSG_RESULT([yes])
+         fi
+         simd_arch=aarch64],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    mips*)
+      AC_MSG_RESULT([yes (mips)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_MIPS_ASSEMBLER_IFELSE(
+        [AC_MSG_RESULT([yes])
+         simd_arch=mips],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    powerpc*)
+      AC_MSG_RESULT([yes (powerpc)])
+      simd_arch=powerpc
+      ;;
+    *)
+      AC_MSG_RESULT([no ("$host_cpu")])
+      with_simd=no;
+      if test "x${require_simd}" = "xyes"; then
+        AC_MSG_ERROR([SIMD support not available for this CPU.])
+      else
+        AC_MSG_WARN([SIMD support not available for this CPU.  Performance will suffer.])
+      fi
+      ;;
+  esac
+
+  if test "x${with_simd}" != "xno"; then
+    AC_DEFINE([WITH_SIMD], [1], [Use accelerated SIMD routines.])
+  fi
+else
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-simd"
+fi
+
+AM_CONDITIONAL([WITH_SIMD], [test "x$with_simd" != "xno"])
+AM_CONDITIONAL([WITH_SSE_FLOAT_DCT], [test "x$simd_arch" = "xx86_64" -o "x$simd_arch" = "xi386"])
+AM_CONDITIONAL([SIMD_I386], [test "x$simd_arch" = "xi386"])
+AM_CONDITIONAL([SIMD_X86_64], [test "x$simd_arch" = "xx86_64"])
+AM_CONDITIONAL([SIMD_ARM], [test "x$simd_arch" = "xarm"])
+AM_CONDITIONAL([SIMD_ARM_64], [test "x$simd_arch" = "xaarch64"])
+AM_CONDITIONAL([SIMD_MIPS], [test "x$simd_arch" = "xmips"])
+AM_CONDITIONAL([SIMD_POWERPC], [test "x$simd_arch" = "xpowerpc"])
+AM_CONDITIONAL([X86_64], [test "x$host_cpu" = "xx86_64" -o "x$host_cpu" = "xamd64"])
+AM_CONDITIONAL([WITH_TURBOJPEG], [test "x$with_turbojpeg" != "xno"])
+AM_CONDITIONAL([CROSS_COMPILING], [test "x$cross_compiling" = "xyes"])
+
+AC_ARG_VAR(PKGNAME, [distribution package name (default: libjpeg-turbo)])
+if test "x$PKGNAME" = "x"; then
+  PKGNAME=$PACKAGE_NAME
+fi
+AC_SUBST(PKGNAME)
+
+case "$host_cpu" in
+  x86_64)
+    RPMARCH=x86_64
+    DEBARCH=amd64
+    ;;
+  i*86 | x86 | ia32)
+    RPMARCH=i386
+    DEBARCH=i386
+    ;;
+  *)
+    RPMARCH=`uname -m`
+    DEBARCH=$RPMARCH
+    ;;
+esac
+
+if test "${docdir}" = ""; then
+  docdir=${datadir}/doc
+  AC_SUBST(docdir)
+fi
+
+AC_SUBST(RPMARCH)
+AC_SUBST(RPM_CONFIG_ARGS)
+AC_SUBST(DEBARCH)
+AC_SUBST(BUILD)
+AC_DEFINE_UNQUOTED([BUILD], "$BUILD", [libjpeg-turbo build number])
+
+# NOTE: autoheader automatically modifies the input file of the first
+# invocation of AC_CONFIG_HEADERS, so we put config.h first to prevent
+# jconfig.h.in from being clobbered.  config.h is used only internally, whereas
+# jconfig.h contains macros that are relevant to external programs (macros that
+# specify which features were built into the library.)
+AC_CONFIG_HEADERS([config.h])
+AC_CONFIG_HEADERS([jconfig.h])
+AC_CONFIG_HEADERS([jconfigint.h])
+AC_CONFIG_FILES([pkgscripts/libjpeg-turbo.spec.tmpl:release/libjpeg-turbo.spec.in])
+AC_CONFIG_FILES([pkgscripts/makecygwinpkg.tmpl:release/makecygwinpkg.in])
+AC_CONFIG_FILES([pkgscripts/makedpkg.tmpl:release/makedpkg.in])
+AC_CONFIG_FILES([pkgscripts/makemacpkg.tmpl:release/makemacpkg.in])
+AC_CONFIG_FILES([pkgscripts/uninstall.tmpl:release/uninstall.in])
+AC_CONFIG_FILES([pkgscripts/libjpeg.pc:release/libjpeg.pc.in])
+AC_CONFIG_FILES([pkgscripts/libturbojpeg.pc:release/libturbojpeg.pc.in])
+if test "x$with_turbojpeg" != "xno"; then
+  AC_CONFIG_FILES([tjbenchtest])
+fi
+if test "x$with_java" = "xyes"; then
+  AC_CONFIG_FILES([tjbenchtest.java])
+  AC_CONFIG_FILES([tjexampletest])
+fi
+AC_CONFIG_FILES([libjpeg.map])
+AC_CONFIG_FILES([Makefile simd/Makefile])
+AC_CONFIG_FILES([java/Makefile])
+AC_CONFIG_FILES([md5/Makefile])
+AC_OUTPUT
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/djpeg.1 glmark2-2021.02/android/jni/src/libjpeg-turbo/djpeg.1
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/djpeg.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/djpeg.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,292 @@
+.TH DJPEG 1 "18 February 2016"
+.SH NAME
+djpeg \- decompress a JPEG file to an image file
+.SH SYNOPSIS
+.B djpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B djpeg
+decompresses the named JPEG file, or the standard input if no file is named,
+and produces an image file on the standard output.  PBMPLUS (PPM/PGM), BMP,
+GIF, Targa, or RLE (Utah Raster Toolkit) output format can be selected.
+(RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-colors " N"
+Reduce image to at most N colors.  This reduces the number of colors used in
+the output image, so that it can be displayed on a colormapped display or
+stored in a colormapped file format.  For example, if you have an 8-bit
+display, you'd need to reduce to 256 or fewer colors.
+.TP
+.BI \-quantize " N"
+Same as
+.BR \-colors .
+.B \-colors
+is the recommended name,
+.B \-quantize
+is provided only for backwards compatibility.
+.TP
+.B \-fast
+Select recommended processing options for fast, low quality output.  (The
+default options are chosen for highest quality output.)  Currently, this is
+equivalent to \fB\-dct fast \-nosmooth \-onepass \-dither ordered\fR.
+.TP
+.B \-grayscale
+Force grayscale output even if JPEG file is color.  Useful for viewing on
+monochrome displays; also,
+.B djpeg
+runs noticeably faster in this mode.
+.TP
+.B \-rgb
+Force RGB output even if JPEG file is grayscale.
+.TP
+.BI \-scale " M/N"
+Scale the output image by a factor M/N.  Currently the scale factor must be
+M/8, where M is an integer between 1 and 16 inclusive, or any reduced fraction
+thereof (such as 1/2, 3/4, etc.)  Scaling is handy if the image is larger than
+your screen; also,
+.B djpeg
+runs much faster when scaling down the output.
+.TP
+.B \-bmp
+Select BMP output format (Windows flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-gif
+Select GIF output format.  Since GIF does not support more than 256 colors,
+.B \-colors 256
+is assumed (unless you specify a smaller number of colors).
+.TP
+.B \-os2
+Select BMP output format (OS/2 1.x flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-pnm
+Select PBMPLUS (PPM/PGM) output format (this is the default format).
+PGM is emitted if the JPEG file is grayscale or if
+.B \-grayscale
+is specified; otherwise PPM is emitted.
+.TP
+.B \-rle
+Select RLE output format.  (Requires URT library.)
+.TP
+.B \-targa
+Select Targa output format.  Grayscale format is emitted if the JPEG file is
+grayscale or if
+.B \-grayscale
+is specified; otherwise, colormapped format is emitted if
+.B \-colors
+is specified; otherwise, 24-bit full-color format is emitted.
+.PP
+Switches for advanced users:
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+In libjpeg-turbo, the fast method is generally about 5-15% faster than the int
+method when using the x86/x86-64 SIMD extensions (results may vary with other
+SIMD implementations, or when using libjpeg-turbo without SIMD extensions.)  If
+the JPEG image was compressed using a quality level of 85 or below, then there
+should be little or no perceptible difference between the two algorithms.  When
+decompressing images that were compressed using quality levels above 85,
+however, the difference between the fast and int methods becomes more
+pronounced.  With images compressed using quality=97, for instance, the fast
+method incurs generally about a 4-6 dB loss (in PSNR) relative to the int
+method, but this can be larger for some images.  If you can avoid it, do not
+use the fast method when decompressing images that were compressed using
+quality levels above 97.  The algorithm often degenerates for such images and
+can actually produce a more lossy output image than if the JPEG image had been
+compressed using lower quality levels.
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is mainly a legacy feature.  It does not produce significantly
+more accurate results than the int method, and it is much slower.  The float
+method may also give different results on different machines due to varying
+roundoff behavior, whereas the integer methods should give the same results on
+all machines.
+.TP
+.B \-dither fs
+Use Floyd-Steinberg dithering in color quantization.
+.TP
+.B \-dither ordered
+Use ordered dithering in color quantization.
+.TP
+.B \-dither none
+Do not use dithering in color quantization.
+By default, Floyd-Steinberg dithering is applied when quantizing colors; this
+is slow but usually produces the best results.  Ordered dither is a compromise
+between speed and quality; no dithering is fast but usually looks awful.  Note
+that these switches have no effect unless color quantization is being done.
+Ordered dither is only available in
+.B \-onepass
+mode.
+.TP
+.BI \-map " file"
+Quantize to the colors used in the specified image file.  This is useful for
+producing multiple files with identical color maps, or for forcing a
+predefined set of colors to be used.  The
+.I file
+must be a GIF or PPM file. This option overrides
+.B \-colors
+and
+.BR \-onepass .
+.TP
+.B \-nosmooth
+Use a faster, lower-quality upsampling routine.
+.TP
+.B \-onepass
+Use one-pass instead of two-pass color quantization.  The one-pass method is
+faster and needs less memory, but it produces a lower-quality image.
+.B \-onepass
+is ignored unless you also say
+.B \-colors
+.IR N .
+Also, the one-pass method is always used for grayscale output (the two-pass
+method is no improvement then).
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memsrc
+Load input file into memory before decompressing.  This feature was implemented
+mainly as a way of testing the in-memory source manager (jpeg_mem_src().)
+.TP
+.BI \-skip " Y0,Y1"
+Decompress all rows of the JPEG image except those between Y0 and Y1
+(inclusive.)  Note that if decompression scaling is being used, then Y0 and Y1
+are relative to the scaled image dimensions.
+.TP
+.BI \-crop " WxH+X+Y"
+Decompress only a rectangular subregion of the image, starting at point X,Y
+with width W and height H.  If necessary, X will be shifted left to the nearest
+iMCU boundary, and the width will be increased accordingly.  Note that if
+decompression scaling is being used, then X, Y, W, and H are relative to the
+scaled image dimensions.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example decompresses the JPEG file foo.jpg, quantizes it to
+256 colors, and saves the output in 8-bit BMP format in foo.bmp:
+.IP
+.B djpeg \-colors 256 \-bmp
+.I foo.jpg
+.B >
+.I foo.bmp
+.SH HINTS
+To get a quick preview of an image, use the
+.B \-grayscale
+and/or
+.B \-scale
+switches.
+.B \-grayscale \-scale 1/8
+is the fastest case.
+.PP
+Several options are available that trade off image quality to gain speed.
+.B \-fast
+turns on the recommended settings.
+.PP
+.B \-dct fast
+and/or
+.B \-nosmooth
+gain speed at a small sacrifice in quality.
+When producing a color-quantized image,
+.B \-onepass \-dither ordered
+is fast but much lower quality than the default behavior.
+.B \-dither none
+may give acceptable results in two-pass mode, but is seldom tolerable in
+one-pass mode.
+.PP
+If you are fortunate enough to have very fast floating point hardware,
+\fB\-dct float\fR may be even faster than \fB\-dct fast\fR.  But on most
+machines \fB\-dct float\fR is slower than \fB\-dct int\fR; in this case it is
+not worth using, because its theoretical accuracy advantage is too small to be
+significant in practice.
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.
+.SH ISSUES
+Support for compressed GIF output files was removed in djpeg v6b due to
+concerns over the Unisys LZW patent.  Although this patent expired in 2006,
+djpeg still lacks compressed GIF support, for these historical reasons.
+(Conversion of JPEG files to GIF is usually a bad idea anyway, since GIF is a
+256-color format.)  The uncompressed GIF files that djpeg generates are larger
+than they should be, but they are readable by standard GIF decoders.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/djpeg.c glmark2-2021.02/android/jni/src/libjpeg-turbo/djpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/djpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/djpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,781 @@
+/*
+ * djpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2013 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2013-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG decompressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      djpeg [options]  inputfile outputfile
+ *      djpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      djpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+#include "wrppm.h"
+
+#include <ctype.h>              /* to declare isprint() */
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)   string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This list defines the known output image formats
+ * (not all of which need be supported by a given version).
+ * You can change the default output format by defining DEFAULT_FMT;
+ * indeed, you had better do so if you undefine PPM_SUPPORTED.
+ */
+
+typedef enum {
+        FMT_BMP,                /* BMP format (Windows flavor) */
+        FMT_GIF,                /* GIF format */
+        FMT_OS2,                /* BMP format (OS/2 flavor) */
+        FMT_PPM,                /* PPM/PGM (PBMPLUS formats) */
+        FMT_RLE,                /* RLE format */
+        FMT_TARGA,              /* Targa format */
+        FMT_TIFF                /* TIFF format */
+} IMAGE_FORMATS;
+
+#ifndef DEFAULT_FMT             /* so can override from CFLAGS in Makefile */
+#define DEFAULT_FMT     FMT_PPM
+#endif
+
+static IMAGE_FORMATS requested_fmt;
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+boolean memsrc;                 /* for -memsrc switch */
+boolean skip, crop;
+JDIMENSION skip_start, skip_end;
+JDIMENSION crop_x, crop_y, crop_width, crop_height;
+#define INPUT_BUF_SIZE  4096
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -colors N      Reduce image to no more than N colors\n");
+  fprintf(stderr, "  -fast          Fast, low-quality processing\n");
+  fprintf(stderr, "  -grayscale     Force grayscale output\n");
+  fprintf(stderr, "  -rgb           Force RGB output\n");
+  fprintf(stderr, "  -rgb565        Force RGB565 output\n");
+#ifdef IDCT_SCALING_SUPPORTED
+  fprintf(stderr, "  -scale M/N     Scale output image by fraction M/N, eg, 1/8\n");
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -bmp           Select BMP output format (Windows style)%s\n",
+          (DEFAULT_FMT == FMT_BMP ? " (default)" : ""));
+#endif
+#ifdef GIF_SUPPORTED
+  fprintf(stderr, "  -gif           Select GIF output format%s\n",
+          (DEFAULT_FMT == FMT_GIF ? " (default)" : ""));
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -os2           Select BMP output format (OS/2 style)%s\n",
+          (DEFAULT_FMT == FMT_OS2 ? " (default)" : ""));
+#endif
+#ifdef PPM_SUPPORTED
+  fprintf(stderr, "  -pnm           Select PBMPLUS (PPM/PGM) output format%s\n",
+          (DEFAULT_FMT == FMT_PPM ? " (default)" : ""));
+#endif
+#ifdef RLE_SUPPORTED
+  fprintf(stderr, "  -rle           Select Utah RLE output format%s\n",
+          (DEFAULT_FMT == FMT_RLE ? " (default)" : ""));
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Select Targa output format%s\n",
+          (DEFAULT_FMT == FMT_TARGA ? " (default)" : ""));
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -dither fs     Use F-S dithering (default)\n");
+  fprintf(stderr, "  -dither none   Don't use dithering in quantization\n");
+  fprintf(stderr, "  -dither ordered  Use ordered dither (medium speed, quality)\n");
+#ifdef QUANT_2PASS_SUPPORTED
+  fprintf(stderr, "  -map FILE      Map to colors used in named image file\n");
+#endif
+  fprintf(stderr, "  -nosmooth      Don't use high-quality upsampling\n");
+#ifdef QUANT_1PASS_SUPPORTED
+  fprintf(stderr, "  -onepass       Use 1-pass quantization (fast, low quality)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memsrc        Load input file into memory before decompressing\n");
+#endif
+
+  fprintf(stderr, "  -skip Y0,Y1    Decompress all rows except those between Y0 and Y1 (inclusive)\n");
+  fprintf(stderr, "  -crop WxH+X+Y  Decompress only a rectangular subregion of the image\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_decompress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+
+  /* Set up default JPEG parameters. */
+  requested_fmt = DEFAULT_FMT;  /* set default output file format */
+  outfilename = NULL;
+  memsrc = FALSE;
+  skip = FALSE;
+  crop = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "bmp", 1)) {
+      /* BMP output format. */
+      requested_fmt = FMT_BMP;
+
+    } else if (keymatch(arg, "colors", 1) || keymatch(arg, "colours", 1) ||
+               keymatch(arg, "quantize", 1) || keymatch(arg, "quantise", 1)) {
+      /* Do color quantization. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+        usage();
+      cinfo->desired_number_of_colors = val;
+      cinfo->quantize_colors = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select IDCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "dither", 2)) {
+      /* Select dithering algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "fs", 2)) {
+        cinfo->dither_mode = JDITHER_FS;
+      } else if (keymatch(argv[argn], "none", 2)) {
+        cinfo->dither_mode = JDITHER_NONE;
+      } else if (keymatch(argv[argn], "ordered", 2)) {
+        cinfo->dither_mode = JDITHER_ORDERED;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "fast", 1)) {
+      /* Select recommended processing options for quick-and-dirty output. */
+      cinfo->two_pass_quantize = FALSE;
+      cinfo->dither_mode = JDITHER_ORDERED;
+      if (! cinfo->quantize_colors) /* don't override an earlier -colors */
+        cinfo->desired_number_of_colors = 216;
+      cinfo->dct_method = JDCT_FASTEST;
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "gif", 1)) {
+      /* GIF output format. */
+      requested_fmt = FMT_GIF;
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force monochrome output. */
+      cinfo->out_color_space = JCS_GRAYSCALE;
+
+    } else if (keymatch(arg, "rgb", 2)) {
+      /* Force RGB output. */
+      cinfo->out_color_space = JCS_RGB;
+
+    } else if (keymatch(arg, "rgb565", 2)) {
+      /* Force RGB565 output. */
+      cinfo->out_color_space = JCS_RGB565;
+
+    } else if (keymatch(arg, "map", 3)) {
+      /* Quantize to a color map taken from an input file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (for_real) {           /* too expensive to do twice! */
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+        FILE *mapfile;
+
+        if ((mapfile = fopen(argv[argn], READ_BINARY)) == NULL) {
+          fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+          exit(EXIT_FAILURE);
+        }
+        read_color_map(cinfo, mapfile);
+        fclose(mapfile);
+        cinfo->quantize_colors = TRUE;
+#else
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+      }
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "nosmooth", 3)) {
+      /* Suppress fancy upsampling */
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "onepass", 3)) {
+      /* Use fast one-pass quantization. */
+      cinfo->two_pass_quantize = FALSE;
+
+    } else if (keymatch(arg, "os2", 3)) {
+      /* BMP output format (OS/2 flavor). */
+      requested_fmt = FMT_OS2;
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "memsrc", 2)) {
+      /* Use in-memory source manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memsrc = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory source manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "pnm", 1) || keymatch(arg, "ppm", 1)) {
+      /* PPM/PGM output format. */
+      requested_fmt = FMT_PPM;
+
+    } else if (keymatch(arg, "rle", 1)) {
+      /* RLE output format. */
+      requested_fmt = FMT_RLE;
+
+    } else if (keymatch(arg, "scale", 2)) {
+      /* Scale the output image by a fraction M/N. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%u/%u",
+                 &cinfo->scale_num, &cinfo->scale_denom) != 2)
+        usage();
+
+    } else if (keymatch(arg, "skip", 2)) {
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argv[argn], "%u,%u", &skip_start, &skip_end) != 2 ||
+          skip_start > skip_end)
+        usage();
+      skip = TRUE;
+
+    } else if (keymatch(arg, "crop", 2)) {
+      char c;
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argv[argn], "%u%c%u+%u+%u", &crop_width, &c, &crop_height,
+                 &crop_x, &crop_y) != 5 ||
+          (c != 'X' && c != 'x') || crop_width < 1 || crop_height < 1)
+        usage();
+      crop = TRUE;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Targa output format. */
+      requested_fmt = FMT_TARGA;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * Marker processor for COM and interesting APPn markers.
+ * This replaces the library's built-in processor, which just skips the marker.
+ * We want to print out the marker as text, to the extent possible.
+ * Note this code relies on a non-suspending data source.
+ */
+
+LOCAL(unsigned int)
+jpeg_getc (j_decompress_ptr cinfo)
+/* Read next byte */
+{
+  struct jpeg_source_mgr *datasrc = cinfo->src;
+
+  if (datasrc->bytes_in_buffer == 0) {
+    if (! (*datasrc->fill_input_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+  datasrc->bytes_in_buffer--;
+  return GETJOCTET(*datasrc->next_input_byte++);
+}
+
+
+METHODDEF(boolean)
+print_text_marker (j_decompress_ptr cinfo)
+{
+  boolean traceit = (cinfo->err->trace_level >= 1);
+  long length;
+  unsigned int ch;
+  unsigned int lastch = 0;
+
+  length = jpeg_getc(cinfo) << 8;
+  length += jpeg_getc(cinfo);
+  length -= 2;                  /* discount the length word itself */
+
+  if (traceit) {
+    if (cinfo->unread_marker == JPEG_COM)
+      fprintf(stderr, "Comment, length %ld:\n", (long) length);
+    else                        /* assume it is an APPn otherwise */
+      fprintf(stderr, "APP%d, length %ld:\n",
+              cinfo->unread_marker - JPEG_APP0, (long) length);
+  }
+
+  while (--length >= 0) {
+    ch = jpeg_getc(cinfo);
+    if (traceit) {
+      /* Emit the character in a readable form.
+       * Nonprintables are converted to \nnn form,
+       * while \ is converted to \\.
+       * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+       */
+      if (ch == '\r') {
+        fprintf(stderr, "\n");
+      } else if (ch == '\n') {
+        if (lastch != '\r')
+          fprintf(stderr, "\n");
+      } else if (ch == '\\') {
+        fprintf(stderr, "\\\\");
+      } else if (isprint(ch)) {
+        putc(ch, stderr);
+      } else {
+        fprintf(stderr, "\\%03o", ch);
+      }
+      lastch = ch;
+    }
+  }
+
+  if (traceit)
+    fprintf(stderr, "\n");
+
+  return TRUE;
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  djpeg_dest_ptr dest_mgr = NULL;
+  FILE *input_file;
+  FILE *output_file;
+  unsigned char *inbuffer = NULL;
+  unsigned long insize = 0;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "djpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_decompress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Insert custom marker processor for COM and APP12.
+   * APP12 is used by some digital camera makers for textual info,
+   * so we provide the ability to display it as text.
+   * If you like, additional APPn marker types can be selected for display,
+   * but don't try to override APP0 or APP14 this way (see libjpeg.txt).
+   */
+  jpeg_set_marker_processor(&cinfo, JPEG_COM, print_text_marker);
+  jpeg_set_marker_processor(&cinfo, JPEG_APP0+12, print_text_marker);
+
+  /* Scan command line to find file names. */
+  /* It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   * (Exception: tracing level set here controls verbosity for COM markers
+   * found during jpeg_read_header...)
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memsrc) {
+    size_t nbytes;
+    do {
+      inbuffer = (unsigned char *)realloc(inbuffer, insize + INPUT_BUF_SIZE);
+      if (inbuffer == NULL) {
+        fprintf(stderr, "%s: memory allocation failure\n", progname);
+        exit(EXIT_FAILURE);
+      }
+      nbytes = JFREAD(input_file, &inbuffer[insize], INPUT_BUF_SIZE);
+      if (nbytes < INPUT_BUF_SIZE && ferror(input_file)) {
+        if (file_index < argc)
+          fprintf(stderr, "%s: can't read from %s\n", progname,
+                  argv[file_index]);
+        else
+          fprintf(stderr, "%s: can't read from stdin\n", progname);
+      }
+      insize += (unsigned long)nbytes;
+    } while (nbytes == INPUT_BUF_SIZE);
+    fprintf(stderr, "Compressed size:  %lu bytes\n", insize);
+    jpeg_mem_src(&cinfo, inbuffer, insize);
+  } else
+#endif
+    jpeg_stdio_src(&cinfo, input_file);
+
+  /* Read file header, set default decompression parameters */
+  (void) jpeg_read_header(&cinfo, TRUE);
+
+  /* Adjust default decompression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Initialize the output module now to let it override any crucial
+   * option settings (for instance, GIF wants to force color quantization).
+   */
+  switch (requested_fmt) {
+#ifdef BMP_SUPPORTED
+  case FMT_BMP:
+    dest_mgr = jinit_write_bmp(&cinfo, FALSE);
+    break;
+  case FMT_OS2:
+    dest_mgr = jinit_write_bmp(&cinfo, TRUE);
+    break;
+#endif
+#ifdef GIF_SUPPORTED
+  case FMT_GIF:
+    dest_mgr = jinit_write_gif(&cinfo);
+    break;
+#endif
+#ifdef PPM_SUPPORTED
+  case FMT_PPM:
+    dest_mgr = jinit_write_ppm(&cinfo);
+    break;
+#endif
+#ifdef RLE_SUPPORTED
+  case FMT_RLE:
+    dest_mgr = jinit_write_rle(&cinfo);
+    break;
+#endif
+#ifdef TARGA_SUPPORTED
+  case FMT_TARGA:
+    dest_mgr = jinit_write_targa(&cinfo);
+    break;
+#endif
+  default:
+    ERREXIT(&cinfo, JERR_UNSUPPORTED_FORMAT);
+    break;
+  }
+  dest_mgr->output_file = output_file;
+
+  /* Start decompressor */
+  (void) jpeg_start_decompress(&cinfo);
+
+  /* Skip rows */
+  if (skip) {
+    JDIMENSION tmp;
+
+    /* Check for valid skip_end.  We cannot check this value until after
+     * jpeg_start_decompress() is called.  Note that we have already verified
+     * that skip_start <= skip_end.
+     */
+    if (skip_end > cinfo.output_height - 1) {
+      fprintf(stderr, "%s: skip region exceeds image height %d\n", progname,
+              cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height -= (skip_end - skip_start + 1);
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    while (cinfo.output_scanline < skip_start) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    jpeg_skip_scanlines(&cinfo, skip_end - skip_start + 1);
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+
+  /* Decompress a subregion */
+  } else if (crop) {
+    JDIMENSION tmp;
+
+    /* Check for valid crop dimensions.  We cannot check these values until
+     * after jpeg_start_decompress() is called.
+     */
+    if (crop_x + crop_width > cinfo.output_width ||
+        crop_y + crop_height > cinfo.output_height) {
+      fprintf(stderr, "%s: crop dimensions exceed image dimensions %d x %d\n",
+              progname, cinfo.output_width, cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    jpeg_crop_scanline(&cinfo, &crop_x, &crop_width);
+    ((ppm_dest_ptr) dest_mgr)->buffer_width = cinfo.output_width *
+                                              cinfo.out_color_components *
+                                              sizeof(JSAMPLE);
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height = crop_height;
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    jpeg_skip_scanlines(&cinfo, crop_y);
+    while (cinfo.output_scanline < crop_y + crop_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    jpeg_skip_scanlines(&cinfo, cinfo.output_height - crop_y - crop_height);
+
+  /* Normal full-image decompress */
+  } else {
+    /* Write output file header */
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+
+    /* Process data */
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  /* Hack: count final pass as done in case finish_output does an extra pass.
+   * The library won't have updated completed_passes.
+   */
+  progress.pub.completed_passes = progress.pub.total_passes;
+#endif
+
+  /* Finish decompression and release memory.
+   * I must do it in this order because output module has allocated memory
+   * of lifespan JPOOL_IMAGE; it needs to finish before releasing memory.
+   */
+  (*dest_mgr->finish_output) (&cinfo, dest_mgr);
+  (void) jpeg_finish_decompress(&cinfo);
+  jpeg_destroy_decompress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  if (memsrc && inbuffer != NULL)
+    free(inbuffer);
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/doxygen.config glmark2-2021.02/android/jni/src/libjpeg-turbo/doxygen.config
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/doxygen.config	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/doxygen.config	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,16 @@
+PROJECT_NAME = TurboJPEG
+PROJECT_NUMBER = 1.5
+OUTPUT_DIRECTORY = doc/
+USE_WINDOWS_ENCODING = NO
+OPTIMIZE_OUTPUT_FOR_C = YES
+WARN_NO_PARAMDOC = YES
+GENERATE_LATEX = NO
+FILE_PATTERNS = turbojpeg.h
+HIDE_UNDOC_MEMBERS = YES
+VERBATIM_HEADERS = NO
+EXTRACT_STATIC = YES
+JAVADOC_AUTOBRIEF = YES
+MAX_INITIALIZER_LINES = 0
+ALWAYS_DETAILED_SEC = YES
+HTML_TIMESTAMP = NO
+HTML_EXTRA_STYLESHEET = doxygen-extra.css
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/doxygen-extra.css glmark2-2021.02/android/jni/src/libjpeg-turbo/doxygen-extra.css
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/doxygen-extra.css	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/doxygen-extra.css	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3 @@
+code {
+	color: #4665A2; 
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/example.c glmark2-2021.02/android/jni/src/libjpeg-turbo/example.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/example.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/example.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,433 @@
+/*
+ * example.c
+ *
+ * This file illustrates how to use the IJG code as a subroutine library
+ * to read or write JPEG image files.  You should look at this code in
+ * conjunction with the documentation file libjpeg.txt.
+ *
+ * This code will not do anything useful as-is, but it may be helpful as a
+ * skeleton for constructing routines that call the JPEG library.
+ *
+ * We present these routines in the same coding style used in the JPEG code
+ * (ANSI function definitions, etc); but you are of course free to code your
+ * routines in a different style if you prefer.
+ */
+
+#include <stdio.h>
+
+/*
+ * Include file for users of JPEG library.
+ * You will need to have included system headers that define at least
+ * the typedefs FILE and size_t before you can include jpeglib.h.
+ * (stdio.h is sufficient on ANSI-conforming systems.)
+ * You may also wish to include "jerror.h".
+ */
+
+#include "jpeglib.h"
+
+/*
+ * <setjmp.h> is used for the optional error recovery mechanism shown in
+ * the second part of the example.
+ */
+
+#include <setjmp.h>
+
+
+
+/******************** JPEG COMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to feed data into the JPEG compressor.
+ * We present a minimal version that does not worry about refinements such
+ * as error recovery (the JPEG code will just exit() if it gets an error).
+ */
+
+
+/*
+ * IMAGE DATA FORMATS:
+ *
+ * The standard input image format is a rectangular array of pixels, with
+ * each pixel having the same number of "component" values (color channels).
+ * Each pixel row is an array of JSAMPLEs (which typically are unsigned chars).
+ * If you are working with color data, then the color values for each pixel
+ * must be adjacent in the row; for example, R,G,B,R,G,B,R,G,B,... for 24-bit
+ * RGB color.
+ *
+ * For this example, we'll assume that this data structure matches the way
+ * our application has stored the image in memory, so we can just pass a
+ * pointer to our image buffer.  In particular, let's say that the image is
+ * RGB color and is described by:
+ */
+
+extern JSAMPLE *image_buffer;   /* Points to large array of R,G,B-order data */
+extern int image_height;        /* Number of rows in image */
+extern int image_width;         /* Number of columns in image */
+
+
+/*
+ * Sample routine for JPEG compression.  We assume that the target file name
+ * and a compression quality factor are passed in.
+ */
+
+GLOBAL(void)
+write_JPEG_file (char *filename, int quality)
+{
+  /* This struct contains the JPEG compression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   * It is possible to have several such structures, representing multiple
+   * compression/decompression processes, in existence at once.  We refer
+   * to any one struct (and its associated working data) as a "JPEG object".
+   */
+  struct jpeg_compress_struct cinfo;
+  /* This struct represents a JPEG error handler.  It is declared separately
+   * because applications often want to supply a specialized error handler
+   * (see the second half of this file for an example).  But here we just
+   * take the easy way out and use the standard error handler, which will
+   * print a message on stderr and call exit() if compression fails.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct jpeg_error_mgr jerr;
+  /* More stuff */
+  FILE *outfile;                /* target file */
+  JSAMPROW row_pointer[1];      /* pointer to JSAMPLE row[s] */
+  int row_stride;               /* physical row width in image buffer */
+
+  /* Step 1: allocate and initialize JPEG compression object */
+
+  /* We have to set up the error handler first, in case the initialization
+   * step fails.  (Unlikely, but it could happen if you are out of memory.)
+   * This routine fills in the contents of struct jerr, and returns jerr's
+   * address which we place into the link field in cinfo.
+   */
+  cinfo.err = jpeg_std_error(&jerr);
+  /* Now we can initialize the JPEG compression object. */
+  jpeg_create_compress(&cinfo);
+
+  /* Step 2: specify data destination (eg, a file) */
+  /* Note: steps 2 and 3 can be done in either order. */
+
+  /* Here we use the library-supplied code to send compressed data to a
+   * stdio stream.  You can also write your own code to do something else.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to write binary files.
+   */
+  if ((outfile = fopen(filename, "wb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    exit(1);
+  }
+  jpeg_stdio_dest(&cinfo, outfile);
+
+  /* Step 3: set parameters for compression */
+
+  /* First we supply a description of the input image.
+   * Four fields of the cinfo struct must be filled in:
+   */
+  cinfo.image_width = image_width;      /* image width and height, in pixels */
+  cinfo.image_height = image_height;
+  cinfo.input_components = 3;           /* # of color components per pixel */
+  cinfo.in_color_space = JCS_RGB;       /* colorspace of input image */
+  /* Now use the library's routine to set default compression parameters.
+   * (You must set at least cinfo.in_color_space before calling this,
+   * since the defaults depend on the source color space.)
+   */
+  jpeg_set_defaults(&cinfo);
+  /* Now you can set any non-default parameters you wish to.
+   * Here we just illustrate the use of quality (quantization table) scaling:
+   */
+  jpeg_set_quality(&cinfo, quality, TRUE /* limit to baseline-JPEG values */);
+
+  /* Step 4: Start compressor */
+
+  /* TRUE ensures that we will write a complete interchange-JPEG file.
+   * Pass TRUE unless you are very sure of what you're doing.
+   */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Step 5: while (scan lines remain to be written) */
+  /*           jpeg_write_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.next_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   * To keep things simple, we pass one scanline per call; you can pass
+   * more if you wish, though.
+   */
+  row_stride = image_width * 3; /* JSAMPLEs per row in image_buffer */
+
+  while (cinfo.next_scanline < cinfo.image_height) {
+    /* jpeg_write_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could pass
+     * more than one scanline at a time if that's more convenient.
+     */
+    row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+    (void) jpeg_write_scanlines(&cinfo, row_pointer, 1);
+  }
+
+  /* Step 6: Finish compression */
+
+  jpeg_finish_compress(&cinfo);
+  /* After finish_compress, we can close the output file. */
+  fclose(outfile);
+
+  /* Step 7: release JPEG compression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_compress(&cinfo);
+
+  /* And we're done! */
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above loop, we ignored the return value of jpeg_write_scanlines,
+ * which is the number of scanlines actually written.  We could get away
+ * with this because we were only relying on the value of cinfo.next_scanline,
+ * which will be incremented correctly.  If you maintain additional loop
+ * variables then you should be careful to increment them properly.
+ * Actually, for output to a stdio stream you needn't worry, because
+ * then jpeg_write_scanlines will write all the lines passed (or else exit
+ * with a fatal error).  Partial writes can only occur if you use a data
+ * destination module that can demand suspension of the compressor.
+ * (If you don't know what that's for, you don't need it.)
+ *
+ * If the compressor requires full-image buffers (for entropy-coding
+ * optimization or a multi-scan JPEG file), it will create temporary
+ * files for anything that doesn't fit within the maximum-memory setting.
+ * (Note that temp files are NOT needed if you use the default parameters.)
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ *
+ * Scanlines MUST be supplied in top-to-bottom order if you want your JPEG
+ * files to be compatible with everyone else's.  If you cannot readily read
+ * your data in that order, you'll need an intermediate array to hold the
+ * image.  See rdtarga.c or rdbmp.c for examples of handling bottom-to-top
+ * source data using the JPEG code's internal virtual-array mechanisms.
+ */
+
+
+
+/******************** JPEG DECOMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to read data from the JPEG decompressor.
+ * It's a bit more refined than the above, in that we show:
+ *   (a) how to modify the JPEG library's standard error-reporting behavior;
+ *   (b) how to allocate workspace using the library's memory manager.
+ *
+ * Just to make this example a little different from the first one, we'll
+ * assume that we do not intend to put the whole image into an in-memory
+ * buffer, but to send it line-by-line someplace else.  We need a one-
+ * scanline-high JSAMPLE array as a work buffer, and we will let the JPEG
+ * memory manager allocate it for us.  This approach is actually quite useful
+ * because we don't need to remember to deallocate the buffer separately: it
+ * will go away automatically when the JPEG object is cleaned up.
+ */
+
+
+/*
+ * ERROR HANDLING:
+ *
+ * The JPEG library's standard error handler (jerror.c) is divided into
+ * several "methods" which you can override individually.  This lets you
+ * adjust the behavior without duplicating a lot of code, which you might
+ * have to update with each future release.
+ *
+ * Our example here shows how to override the "error_exit" method so that
+ * control is returned to the library's caller when a fatal error occurs,
+ * rather than calling exit() as the standard error_exit method does.
+ *
+ * We use C's setjmp/longjmp facility to return control.  This means that the
+ * routine which calls the JPEG library must first execute a setjmp() call to
+ * establish the return point.  We want the replacement error_exit to do a
+ * longjmp().  But we need to make the setjmp buffer accessible to the
+ * error_exit routine.  To do this, we make a private extension of the
+ * standard JPEG error handler object.  (If we were using C++, we'd say we
+ * were making a subclass of the regular error handler.)
+ *
+ * Here's the extended error handler struct:
+ */
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;    /* "public" fields */
+
+  jmp_buf setjmp_buffer;        /* for return to caller */
+};
+
+typedef struct my_error_mgr *my_error_ptr;
+
+/*
+ * Here's the routine that will replace the standard error_exit method:
+ */
+
+METHODDEF(void)
+my_error_exit (j_common_ptr cinfo)
+{
+  /* cinfo->err really points to a my_error_mgr struct, so coerce pointer */
+  my_error_ptr myerr = (my_error_ptr) cinfo->err;
+
+  /* Always display the message. */
+  /* We could postpone this until after returning, if we chose. */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Return control to the setjmp point */
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+
+/*
+ * Sample routine for JPEG decompression.  We assume that the source file name
+ * is passed in.  We want to return 1 on success, 0 on error.
+ */
+
+
+GLOBAL(int)
+read_JPEG_file (char *filename)
+{
+  /* This struct contains the JPEG decompression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   */
+  struct jpeg_decompress_struct cinfo;
+  /* We use our private extension JPEG error handler.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct my_error_mgr jerr;
+  /* More stuff */
+  FILE *infile;                 /* source file */
+  JSAMPARRAY buffer;            /* Output row buffer */
+  int row_stride;               /* physical row width in output buffer */
+
+  /* In this example we want to open the input file before doing anything else,
+   * so that the setjmp() error recovery below can assume the file is open.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to read binary files.
+   */
+
+  if ((infile = fopen(filename, "rb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    return 0;
+  }
+
+  /* Step 1: allocate and initialize JPEG decompression object */
+
+  /* We set up the normal JPEG error routines, then override error_exit. */
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  /* Establish the setjmp return context for my_error_exit to use. */
+  if (setjmp(jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error.
+     * We need to clean up the JPEG object, close the input file, and return.
+     */
+    jpeg_destroy_decompress(&cinfo);
+    fclose(infile);
+    return 0;
+  }
+  /* Now we can initialize the JPEG decompression object. */
+  jpeg_create_decompress(&cinfo);
+
+  /* Step 2: specify data source (eg, a file) */
+
+  jpeg_stdio_src(&cinfo, infile);
+
+  /* Step 3: read file parameters with jpeg_read_header() */
+
+  (void) jpeg_read_header(&cinfo, TRUE);
+  /* We can ignore the return value from jpeg_read_header since
+   *   (a) suspension is not possible with the stdio data source, and
+   *   (b) we passed TRUE to reject a tables-only JPEG file as an error.
+   * See libjpeg.txt for more info.
+   */
+
+  /* Step 4: set parameters for decompression */
+
+  /* In this example, we don't need to change any of the defaults set by
+   * jpeg_read_header(), so we do nothing here.
+   */
+
+  /* Step 5: Start decompressor */
+
+  (void) jpeg_start_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* We may need to do some setup of our own at this point before reading
+   * the data.  After jpeg_start_decompress() we have the correct scaled
+   * output image dimensions available, as well as the output colormap
+   * if we asked for color quantization.
+   * In this example, we need to make an output work buffer of the right size.
+   */
+  /* JSAMPLEs per row in output buffer */
+  row_stride = cinfo.output_width * cinfo.output_components;
+  /* Make a one-row-high sample array that will go away when done with image */
+  buffer = (*cinfo.mem->alloc_sarray)
+                ((j_common_ptr) &cinfo, JPOOL_IMAGE, row_stride, 1);
+
+  /* Step 6: while (scan lines remain to be read) */
+  /*           jpeg_read_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.output_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   */
+  while (cinfo.output_scanline < cinfo.output_height) {
+    /* jpeg_read_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could ask for
+     * more than one scanline at a time if that's more convenient.
+     */
+    (void) jpeg_read_scanlines(&cinfo, buffer, 1);
+    /* Assume put_scanline_someplace wants a pointer and sample count. */
+    put_scanline_someplace(buffer[0], row_stride);
+  }
+
+  /* Step 7: Finish decompression */
+
+  (void) jpeg_finish_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* Step 8: Release JPEG decompression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_decompress(&cinfo);
+
+  /* After finish_decompress, we can close the input file.
+   * Here we postpone it until after no more JPEG errors are possible,
+   * so as to simplify the setjmp error logic above.  (Actually, I don't
+   * think that jpeg_destroy can do an error exit, but why assume anything...)
+   */
+  fclose(infile);
+
+  /* At this point you may want to check to see whether any corrupt-data
+   * warnings occurred (test whether jerr.pub.num_warnings is nonzero).
+   */
+
+  /* And we're done! */
+  return 1;
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above code, we ignored the return value of jpeg_read_scanlines,
+ * which is the number of scanlines actually read.  We could get away with
+ * this because we asked for only one line at a time and we weren't using
+ * a suspending data source.  See libjpeg.txt for more info.
+ *
+ * We cheated a bit by calling alloc_sarray() after jpeg_start_decompress();
+ * we should have done it beforehand to ensure that the space would be
+ * counted against the JPEG max_memory setting.  In some systems the above
+ * code would risk an out-of-memory error.  However, in general we don't
+ * know the output image dimensions before jpeg_start_decompress(), unless we
+ * call jpeg_calc_output_dimensions().  See libjpeg.txt for more about this.
+ *
+ * Scanlines are returned in the same order as they appear in the JPEG file,
+ * which is standardly top-to-bottom.  If you must emit data bottom-to-top,
+ * you can use one of the virtual arrays provided by the JPEG memory manager
+ * to invert the data.  See wrbmp.c for an example.
+ *
+ * As with compression, some operating modes may require temporary files.
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jaricom.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jaricom.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jaricom.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jaricom.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jaricom.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains probability estimation tables for common use in
  * arithmetic entropy encoding and decoding routines.
@@ -18,7 +21,7 @@
 #include "jpeglib.h"
 
 /* The following #define specifies the packing of the four components
- * into the compact INT32 representation.
+ * into the compact JLONG representation.
  * Note that this formula must match the actual arithmetic encoder
  * and decoder implementation.  The implementation has to be changed
  * if this formula is changed.
@@ -26,9 +29,9 @@
  * implementation (jbig_tab.c).
  */
 
-#define V(i,a,b,c,d) (((INT32)a << 16) | ((INT32)c << 8) | ((INT32)d << 7) | b)
+#define V(i,a,b,c,d) (((JLONG)a << 16) | ((JLONG)c << 8) | ((JLONG)d << 7) | b)
 
-const INT32 jpeg_aritab[113+1] = {
+const JLONG jpeg_aritab[113+1] = {
 /*
  * Index, Qe_Value, Next_Index_LPS, Next_Index_MPS, Switch_MPS
  */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcapimin.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcapimin.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcapimin.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcapimin.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jcapimin.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
  * Modified 2003-2010 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the compression half
  * of the JPEG library.  These are the "minimum" API routines that may be
@@ -33,12 +36,12 @@
   int i;
 
   /* Guard against version mismatches between library and caller. */
-  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
   if (version != JPEG_LIB_VERSION)
     ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
-  if (structsize != SIZEOF(struct jpeg_compress_struct))
-    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
-	     (int) SIZEOF(struct jpeg_compress_struct), (int) structsize);
+  if (structsize != sizeof(struct jpeg_compress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int) sizeof(struct jpeg_compress_struct), (int) structsize);
 
   /* For debugging purposes, we zero the whole master structure.
    * But the application has already set the err pointer, and may have set
@@ -47,9 +50,9 @@
    * complain here.
    */
   {
-    struct jpeg_error_mgr * err = cinfo->err;
-    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
-    MEMZERO(cinfo, SIZEOF(struct jpeg_compress_struct));
+    struct jpeg_error_mgr *err = cinfo->err;
+    void *client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, sizeof(struct jpeg_compress_struct));
     cinfo->err = err;
     cinfo->client_data = client_data;
   }
@@ -85,7 +88,7 @@
 
   cinfo->script_space = NULL;
 
-  cinfo->input_gamma = 1.0;	/* in case application forgets */
+  cinfo->input_gamma = 1.0;     /* in case application forgets */
 
   /* OK, I'm ready */
   cinfo->global_state = CSTATE_START;
@@ -131,8 +134,8 @@
 jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress)
 {
   int i;
-  JQUANT_TBL * qtbl;
-  JHUFF_TBL * htbl;
+  JQUANT_TBL *qtbl;
+  JHUFF_TBL *htbl;
 
   for (i = 0; i < NUM_QUANT_TBLS; i++) {
     if ((qtbl = cinfo->quant_tbl_ptrs[i]) != NULL)
@@ -173,15 +176,15 @@
     (*cinfo->master->prepare_for_pass) (cinfo);
     for (iMCU_row = 0; iMCU_row < cinfo->total_iMCU_rows; iMCU_row++) {
       if (cinfo->progress != NULL) {
-	cinfo->progress->pass_counter = (long) iMCU_row;
-	cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        cinfo->progress->pass_counter = (long) iMCU_row;
+        cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       }
       /* We bypass the main controller and invoke coef controller directly;
        * all work is being done from the coefficient buffer.
        */
       if (! (*cinfo->coef->compress_data) (cinfo, (JSAMPIMAGE) NULL))
-	ERREXIT(cinfo, JERR_CANT_SUSPEND);
+        ERREXIT(cinfo, JERR_CANT_SUSPEND);
     }
     (*cinfo->master->finish_pass) (cinfo);
   }
@@ -202,9 +205,9 @@
 
 GLOBAL(void)
 jpeg_write_marker (j_compress_ptr cinfo, int marker,
-		   const JOCTET *dataptr, unsigned int datalen)
+                   const JOCTET *dataptr, unsigned int datalen)
 {
-  JMETHOD(void, write_marker_byte, (j_compress_ptr info, int val));
+  void (*write_marker_byte) (j_compress_ptr info, int val);
 
   if (cinfo->next_scanline != 0 ||
       (cinfo->global_state != CSTATE_SCANNING &&
@@ -213,7 +216,7 @@
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
   (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
-  write_marker_byte = cinfo->marker->write_marker_byte;	/* copy for speed */
+  write_marker_byte = cinfo->marker->write_marker_byte; /* copy for speed */
   while (datalen--) {
     (*write_marker_byte) (cinfo, *dataptr);
     dataptr++;
@@ -248,14 +251,14 @@
  * To produce a pair of files containing abbreviated tables and abbreviated
  * image data, one would proceed as follows:
  *
- *		initialize JPEG object
- *		set JPEG parameters
- *		set destination to table file
- *		jpeg_write_tables(cinfo);
- *		set destination to image file
- *		jpeg_start_compress(cinfo, FALSE);
- *		write data...
- *		jpeg_finish_compress(cinfo);
+ *              initialize JPEG object
+ *              set JPEG parameters
+ *              set destination to table file
+ *              jpeg_write_tables(cinfo);
+ *              set destination to image file
+ *              jpeg_start_compress(cinfo, FALSE);
+ *              write data...
+ *              jpeg_finish_compress(cinfo);
  *
  * jpeg_write_tables has the side effect of marking all tables written
  * (same as jpeg_suppress_tables(..., TRUE)).  Thus a subsequent start_compress
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcapistd.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcapistd.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcapistd.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcapistd.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1994-1996, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the compression half
  * of the JPEG library.  These are the "standard" API routines that are
@@ -41,7 +42,7 @@
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
   if (write_all_tables)
-    jpeg_suppress_tables(cinfo, FALSE);	/* mark all tables to be written */
+    jpeg_suppress_tables(cinfo, FALSE); /* mark all tables to be written */
 
   /* (Re)initialize error mgr and destination modules */
   (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
@@ -75,7 +76,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_write_scanlines (j_compress_ptr cinfo, JSAMPARRAY scanlines,
-		      JDIMENSION num_lines)
+                      JDIMENSION num_lines)
 {
   JDIMENSION row_ctr, rows_left;
 
@@ -118,7 +119,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
-		     JDIMENSION num_lines)
+                     JDIMENSION num_lines)
 {
   JDIMENSION lines_per_iMCU_row;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcarith.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcarith.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcarith.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcarith.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcarith.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains portable arithmetic entropy encoding routines for JPEG
  * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
@@ -23,10 +26,10 @@
 typedef struct {
   struct jpeg_entropy_encoder pub; /* public fields */
 
-  INT32 c; /* C register, base of coding interval, layout as in sec. D.1.3 */
-  INT32 a;               /* A register, normalized size of coding interval */
-  INT32 sc;        /* counter for stacked 0xFF values which might overflow */
-  INT32 zc;          /* counter for pending 0x00 output values which might *
+  JLONG c; /* C register, base of coding interval, layout as in sec. D.1.3 */
+  JLONG a;               /* A register, normalized size of coding interval */
+  JLONG sc;        /* counter for stacked 0xFF values which might overflow */
+  JLONG zc;          /* counter for pending 0x00 output values which might *
                           * be discarded at the end ("Pacman" termination) */
   int ct;  /* bit shift counter, determines when next byte will be written */
   int buffer;                /* buffer for most recent output byte != 0xFF */
@@ -34,18 +37,18 @@
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
   int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to statistics areas (these workspaces have image lifespan) */
-  unsigned char * dc_stats[NUM_ARITH_TBLS];
-  unsigned char * ac_stats[NUM_ARITH_TBLS];
+  unsigned char *dc_stats[NUM_ARITH_TBLS];
+  unsigned char *ac_stats[NUM_ARITH_TBLS];
 
   /* Statistics bin for coding with fixed probability 0.5 */
   unsigned char fixed_bin[4];
 } arith_entropy_encoder;
 
-typedef arith_entropy_encoder * arith_entropy_ptr;
+typedef arith_entropy_encoder *arith_entropy_ptr;
 
 /* The following two definitions specify the allocation chunk size
  * for the statistics area.
@@ -95,20 +98,20 @@
 #define CALCULATE_SPECTRAL_CONDITIONING
  */
 
-/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
- * We assume that int right shift is unsigned if INT32 right shift is,
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
  * which should be safe.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	int ishift_temp;
+#define ISHIFT_TEMPS    int ishift_temp;
 #define IRIGHT_SHIFT(x,shft)  \
-	((ishift_temp = (x)) < 0 ? \
-	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
-	 (ishift_temp >> (shft)))
+        ((ishift_temp = (x)) < 0 ? \
+         (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+         (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 
@@ -116,7 +119,7 @@
 emit_byte (int val, j_compress_ptr cinfo)
 /* Write next output byte; we do not support suspension in this module. */
 {
-  struct jpeg_destination_mgr * dest = cinfo->dest;
+  struct jpeg_destination_mgr *dest = cinfo->dest;
 
   *dest->next_output_byte++ = (JOCTET) val;
   if (--dest->free_in_buffer == 0)
@@ -133,7 +136,7 @@
 finish_pass (j_compress_ptr cinfo)
 {
   arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
-  INT32 temp;
+  JLONG temp;
 
   /* Section D.1.8: Termination of encoding */
 
@@ -149,11 +152,11 @@
     /* One final overflow has to be handled */
     if (e->buffer >= 0) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       emit_byte(e->buffer + 1, cinfo);
       if (e->buffer + 1 == 0xFF)
-	emit_byte(0x00, cinfo);
+        emit_byte(0x00, cinfo);
     }
     e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
     e->sc = 0;
@@ -162,17 +165,17 @@
       ++e->zc;
     else if (e->buffer >= 0) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       emit_byte(e->buffer, cinfo);
     }
     if (e->sc) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       do {
-	emit_byte(0xFF, cinfo);
-	emit_byte(0x00, cinfo);
+        emit_byte(0xFF, cinfo);
+        emit_byte(0x00, cinfo);
       } while (--e->sc);
     }
   }
@@ -187,7 +190,7 @@
     if (e->c & 0x7F800L) {
       emit_byte((e->c >> 11) & 0xFF, cinfo);
       if (((e->c >> 11) & 0xFF) == 0xFF)
-	emit_byte(0x00, cinfo);
+        emit_byte(0x00, cinfo);
     }
   }
 }
@@ -216,20 +219,20 @@
  */
 
 LOCAL(void)
-arith_encode (j_compress_ptr cinfo, unsigned char *st, int val) 
+arith_encode (j_compress_ptr cinfo, unsigned char *st, int val)
 {
   register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
   register unsigned char nl, nm;
-  register INT32 qe, temp;
+  register JLONG qe, temp;
   register int sv;
 
   /* Fetch values from our compact representation of Table D.2:
    * Qe values and probability estimation state machine
    */
   sv = *st;
-  qe = jpeg_aritab[sv & 0x7F];	/* => Qe_Value */
-  nl = qe & 0xFF; qe >>= 8;	/* Next_Index_LPS + Switch_MPS */
-  nm = qe & 0xFF; qe >>= 8;	/* Next_Index_MPS */
+  qe = jpeg_aritab[sv & 0x7F];  /* => Qe_Value */
+  nl = qe & 0xFF; qe >>= 8;     /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF; qe >>= 8;     /* Next_Index_MPS */
 
   /* Encode & estimation procedures per sections D.1.4 & D.1.5 */
   e->a -= qe;
@@ -243,7 +246,7 @@
       e->c += e->a;
       e->a = qe;
     }
-    *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
+    *st = (sv & 0x80) ^ nl;     /* Estimate_after_LPS */
   } else {
     /* Encode the more probable symbol */
     if (e->a >= 0x8000L)
@@ -255,7 +258,7 @@
       e->c += e->a;
       e->a = qe;
     }
-    *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+    *st = (sv & 0x80) ^ nm;     /* Estimate_after_MPS */
   }
 
   /* Renormalization & data output per section D.1.6 */
@@ -266,43 +269,43 @@
       /* Another byte is ready for output */
       temp = e->c >> 19;
       if (temp > 0xFF) {
-	/* Handle overflow over all stacked 0xFF bytes */
-	if (e->buffer >= 0) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  emit_byte(e->buffer + 1, cinfo);
-	  if (e->buffer + 1 == 0xFF)
-	    emit_byte(0x00, cinfo);
-	}
-	e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
-	e->sc = 0;
-	/* Note: The 3 spacer bits in the C register guarantee
-	 * that the new buffer byte can't be 0xFF here
-	 * (see page 160 in the P&M JPEG book). */
-	e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
+        /* Handle overflow over all stacked 0xFF bytes */
+        if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer + 1, cinfo);
+          if (e->buffer + 1 == 0xFF)
+            emit_byte(0x00, cinfo);
+        }
+        e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
+        e->sc = 0;
+        /* Note: The 3 spacer bits in the C register guarantee
+         * that the new buffer byte can't be 0xFF here
+         * (see page 160 in the P&M JPEG book). */
+        e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
       } else if (temp == 0xFF) {
-	++e->sc;  /* stack 0xFF byte (which might overflow later) */
+        ++e->sc;  /* stack 0xFF byte (which might overflow later) */
       } else {
-	/* Output all stacked 0xFF bytes, they will not overflow any more */
-	if (e->buffer == 0)
-	  ++e->zc;
-	else if (e->buffer >= 0) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  emit_byte(e->buffer, cinfo);
-	}
-	if (e->sc) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  do {
-	    emit_byte(0xFF, cinfo);
-	    emit_byte(0x00, cinfo);
-	  } while (--e->sc);
-	}
-	e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
+        /* Output all stacked 0xFF bytes, they will not overflow any more */
+        if (e->buffer == 0)
+          ++e->zc;
+        else if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer, cinfo);
+        }
+        if (e->sc) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          do {
+            emit_byte(0xFF, cinfo);
+            emit_byte(0x00, cinfo);
+          } while (--e->sc);
+        }
+        e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
       }
       e->c &= 0x7FFFFL;
       e->ct += 8;
@@ -320,7 +323,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   finish_pass(cinfo);
 
@@ -398,45 +401,45 @@
     /* Figure F.4: Encode_DC_DIFF */
     if ((v = m - entropy->last_dc_val[ci]) == 0) {
       arith_encode(cinfo, st, 0);
-      entropy->dc_context[ci] = 0;	/* zero diff category */
+      entropy->dc_context[ci] = 0;      /* zero diff category */
     } else {
       entropy->last_dc_val[ci] = m;
       arith_encode(cinfo, st, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, st + 1, 0);	/* Table F.4: SS = S0 + 1 */
-	st += 2;			/* Table F.4: SP = S0 + 2 */
-	entropy->dc_context[ci] = 4;	/* small positive diff category */
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_context[ci] = 4;    /* small positive diff category */
       } else {
-	v = -v;
-	arith_encode(cinfo, st + 1, 1);	/* Table F.4: SS = S0 + 1 */
-	st += 3;			/* Table F.4: SN = S0 + 3 */
-	entropy->dc_context[ci] = 8;	/* small negative diff category */
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_context[ci] = 8;    /* small negative diff category */
       }
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;	/* zero diff category */
+        entropy->dc_context[ci] = 0;    /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] += 8;	/* large diff category */
+        entropy->dc_context[ci] += 8;   /* large diff category */
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
   }
 
@@ -491,21 +494,21 @@
   /* Figure F.5: Encode_AC_Coefficients */
   for (k = cinfo->Ss; k <= ke; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
-    arith_encode(cinfo, st, 0);		/* EOB decision */
+    arith_encode(cinfo, st, 0);         /* EOB decision */
     for (;;) {
       if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
-	if (v >>= cinfo->Al) {
-	  arith_encode(cinfo, st + 1, 1);
-	  arith_encode(cinfo, entropy->fixed_bin, 0);
-	  break;
-	}
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 0);
+          break;
+        }
       } else {
-	v = -v;
-	if (v >>= cinfo->Al) {
-	  arith_encode(cinfo, st + 1, 1);
-	  arith_encode(cinfo, entropy->fixed_bin, 1);
-	  break;
-	}
+        v = -v;
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 1);
+          break;
+        }
       }
       arith_encode(cinfo, st + 1, 0); st += 3; k++;
     }
@@ -517,15 +520,15 @@
       m = 1;
       v2 = v;
       if (v2 >>= 1) {
-	arith_encode(cinfo, st, 1);
-	m <<= 1;
-	st = entropy->ac_stats[tbl] +
-	     (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m <<= 1;
+        st = entropy->ac_stats[tbl] +
+             (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
     }
     arith_encode(cinfo, st, 0);
@@ -566,7 +569,7 @@
     entropy->restarts_to_go--;
   }
 
-  st = entropy->fixed_bin;	/* use fixed probability estimation */
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
   Al = cinfo->Al;
 
   /* Encode the MCU data blocks */
@@ -635,29 +638,29 @@
   for (k = cinfo->Ss; k <= ke; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
     if (k > kex)
-      arith_encode(cinfo, st, 0);	/* EOB decision */
+      arith_encode(cinfo, st, 0);       /* EOB decision */
     for (;;) {
       if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
-	if (v >>= cinfo->Al) {
-	  if (v >> 1)			/* previously nonzero coef */
-	    arith_encode(cinfo, st + 2, (v & 1));
-	  else {			/* newly nonzero coef */
-	    arith_encode(cinfo, st + 1, 1);
-	    arith_encode(cinfo, entropy->fixed_bin, 0);
-	  }
-	  break;
-	}
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 0);
+          }
+          break;
+        }
       } else {
-	v = -v;
-	if (v >>= cinfo->Al) {
-	  if (v >> 1)			/* previously nonzero coef */
-	    arith_encode(cinfo, st + 2, (v & 1));
-	  else {			/* newly nonzero coef */
-	    arith_encode(cinfo, st + 1, 1);
-	    arith_encode(cinfo, entropy->fixed_bin, 1);
-	  }
-	  break;
-	}
+        v = -v;
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 1);
+          }
+          break;
+        }
       }
       arith_encode(cinfo, st + 1, 0); st += 3; k++;
     }
@@ -680,7 +683,7 @@
 encode_mcu (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JBLOCKROW block;
   unsigned char *st;
   int blkn, ci, tbl, k, ke;
@@ -713,45 +716,45 @@
     /* Figure F.4: Encode_DC_DIFF */
     if ((v = (*block)[0] - entropy->last_dc_val[ci]) == 0) {
       arith_encode(cinfo, st, 0);
-      entropy->dc_context[ci] = 0;	/* zero diff category */
+      entropy->dc_context[ci] = 0;      /* zero diff category */
     } else {
       entropy->last_dc_val[ci] = (*block)[0];
       arith_encode(cinfo, st, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, st + 1, 0);	/* Table F.4: SS = S0 + 1 */
-	st += 2;			/* Table F.4: SP = S0 + 2 */
-	entropy->dc_context[ci] = 4;	/* small positive diff category */
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_context[ci] = 4;    /* small positive diff category */
       } else {
-	v = -v;
-	arith_encode(cinfo, st + 1, 1);	/* Table F.4: SS = S0 + 1 */
-	st += 3;			/* Table F.4: SN = S0 + 3 */
-	entropy->dc_context[ci] = 8;	/* small negative diff category */
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_context[ci] = 8;    /* small negative diff category */
       }
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;	/* zero diff category */
+        entropy->dc_context[ci] = 0;    /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] += 8;	/* large diff category */
+        entropy->dc_context[ci] += 8;   /* large diff category */
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
 
     /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
@@ -765,43 +768,43 @@
     /* Figure F.5: Encode_AC_Coefficients */
     for (k = 1; k <= ke; k++) {
       st = entropy->ac_stats[tbl] + 3 * (k - 1);
-      arith_encode(cinfo, st, 0);	/* EOB decision */
+      arith_encode(cinfo, st, 0);       /* EOB decision */
       while ((v = (*block)[jpeg_natural_order[k]]) == 0) {
-	arith_encode(cinfo, st + 1, 0); st += 3; k++;
+        arith_encode(cinfo, st + 1, 0); st += 3; k++;
       }
       arith_encode(cinfo, st + 1, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, entropy->fixed_bin, 0);
+        arith_encode(cinfo, entropy->fixed_bin, 0);
       } else {
-	v = -v;
-	arith_encode(cinfo, entropy->fixed_bin, 1);
+        v = -v;
+        arith_encode(cinfo, entropy->fixed_bin, 1);
       }
       st += 2;
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	if (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st = entropy->ac_stats[tbl] +
-	       (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	  while (v2 >>= 1) {
-	    arith_encode(cinfo, st, 1);
-	    m <<= 1;
-	    st += 1;
-	  }
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        if (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st = entropy->ac_stats[tbl] +
+               (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+          while (v2 >>= 1) {
+            arith_encode(cinfo, st, 1);
+            m <<= 1;
+            st += 1;
+          }
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
     /* Encode EOB decision only if k <= DCTSIZE2 - 1 */
     if (k <= DCTSIZE2 - 1) {
@@ -823,7 +826,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (gather_statistics)
     /* Make sure to avoid that in the master control logic!
@@ -838,14 +841,14 @@
   if (cinfo->progressive_mode) {
     if (cinfo->Ah == 0) {
       if (cinfo->Ss == 0)
-	entropy->pub.encode_mcu = encode_mcu_DC_first;
+        entropy->pub.encode_mcu = encode_mcu_DC_first;
       else
-	entropy->pub.encode_mcu = encode_mcu_AC_first;
+        entropy->pub.encode_mcu = encode_mcu_AC_first;
     } else {
       if (cinfo->Ss == 0)
-	entropy->pub.encode_mcu = encode_mcu_DC_refine;
+        entropy->pub.encode_mcu = encode_mcu_DC_refine;
       else
-	entropy->pub.encode_mcu = encode_mcu_AC_refine;
+        entropy->pub.encode_mcu = encode_mcu_AC_refine;
     }
   } else
     entropy->pub.encode_mcu = encode_mcu;
@@ -857,10 +860,10 @@
     if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       tbl = compptr->dc_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->dc_stats[tbl] == NULL)
-	entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+        entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
       MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
       /* Initialize DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
@@ -870,15 +873,15 @@
     if (cinfo->progressive_mode == 0 || cinfo->Se) {
       tbl = compptr->ac_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->ac_stats[tbl] == NULL)
-	entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+        entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
       MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
 #ifdef CALCULATE_SPECTRAL_CONDITIONING
       if (cinfo->progressive_mode)
-	/* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
-	cinfo->arith_ac_K[tbl] = cinfo->Ss + ((8 + cinfo->Se - cinfo->Ss) >> 4);
+        /* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
+        cinfo->arith_ac_K[tbl] = cinfo->Ss + ((8 + cinfo->Se - cinfo->Ss) >> 4);
 #endif
     }
   }
@@ -909,7 +912,7 @@
 
   entropy = (arith_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(arith_entropy_encoder));
+                                sizeof(arith_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass;
   entropy->pub.finish_pass = finish_pass;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccoefct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jccoefct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccoefct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jccoefct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jccoefct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the coefficient buffer controller for compression.
  * This controller is the top level of the JPEG compressor proper.
@@ -34,19 +37,16 @@
 typedef struct {
   struct jpeg_c_coef_controller pub; /* public fields */
 
-  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
-  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
 
   /* For single-pass compression, it's sufficient to buffer just one MCU
    * (although this may prove a bit slow in practice).  We allocate a
    * workspace of C_MAX_BLOCKS_IN_MCU coefficient blocks, and reuse it for each
-   * MCU constructed and sent.  (On 80x86, the workspace is FAR even though
-   * it's not really very big; this is to keep the module interfaces unchanged
-   * when a large coefficient buffer is necessary.)
-   * In multi-pass modes, this array points to the current MCU's blocks
-   * within the virtual arrays.
+   * MCU constructed and sent.  In multi-pass modes, this array points to the
+   * current MCU's blocks within the virtual arrays.
    */
   JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
 
@@ -54,17 +54,17 @@
   jvirt_barray_ptr whole_image[MAX_COMPONENTS];
 } my_coef_controller;
 
-typedef my_coef_controller * my_coef_ptr;
+typedef my_coef_controller *my_coef_ptr;
 
 
 /* Forward declarations */
 METHODDEF(boolean) compress_data
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 #ifdef FULL_COEF_BUFFER_SUPPORTED
 METHODDEF(boolean) compress_first_pass
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 METHODDEF(boolean) compress_output
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 #endif
 
 
@@ -143,7 +143,7 @@
 compress_data (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, bi, ci, yindex, yoffset, blockcnt;
@@ -154,7 +154,7 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num <= last_MCU_col;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Determine where data comes from in input_buf and do the DCT thing.
        * Each call on forward_DCT processes a horizontal row of DCT blocks
        * as wide as an MCU; we rely on having allocated the MCU_buffer[] blocks
@@ -166,46 +166,46 @@
        */
       blkn = 0;
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						: compptr->last_col_width;
-	xpos = MCU_col_num * compptr->MCU_sample_width;
-	ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (coef->iMCU_row_num < last_iMCU_row ||
-	      yoffset+yindex < compptr->last_row_height) {
-	    (*cinfo->fdct->forward_DCT) (cinfo, compptr,
-					 input_buf[compptr->component_index],
-					 coef->MCU_buffer[blkn],
-					 ypos, xpos, (JDIMENSION) blockcnt);
-	    if (blockcnt < compptr->MCU_width) {
-	      /* Create some dummy blocks at the right edge of the image. */
-	      jzero_far((void FAR *) coef->MCU_buffer[blkn + blockcnt],
-			(compptr->MCU_width - blockcnt) * SIZEOF(JBLOCK));
-	      for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
-		coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
-	      }
-	    }
-	  } else {
-	    /* Create a row of dummy blocks at the bottom of the image. */
-	    jzero_far((void FAR *) coef->MCU_buffer[blkn],
-		      compptr->MCU_width * SIZEOF(JBLOCK));
-	    for (bi = 0; bi < compptr->MCU_width; bi++) {
-	      coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
-	    }
-	  }
-	  blkn += compptr->MCU_width;
-	  ypos += DCTSIZE;
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                : compptr->last_col_width;
+        xpos = MCU_col_num * compptr->MCU_sample_width;
+        ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yoffset+yindex < compptr->last_row_height) {
+            (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+                                         input_buf[compptr->component_index],
+                                         coef->MCU_buffer[blkn],
+                                         ypos, xpos, (JDIMENSION) blockcnt);
+            if (blockcnt < compptr->MCU_width) {
+              /* Create some dummy blocks at the right edge of the image. */
+              jzero_far((void *) coef->MCU_buffer[blkn + blockcnt],
+                        (compptr->MCU_width - blockcnt) * sizeof(JBLOCK));
+              for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
+                coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
+              }
+            }
+          } else {
+            /* Create a row of dummy blocks at the bottom of the image. */
+            jzero_far((void *) coef->MCU_buffer[blkn],
+                      compptr->MCU_width * sizeof(JBLOCK));
+            for (bi = 0; bi < compptr->MCU_width; bi++) {
+              coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
+            }
+          }
+          blkn += compptr->MCU_width;
+          ypos += DCTSIZE;
+        }
       }
       /* Try to write the MCU.  In event of a suspension failure, we will
        * re-DCT the MCU on restart (a bit inefficient, could be fixed...)
        */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -280,17 +280,17 @@
     for (block_row = 0; block_row < block_rows; block_row++) {
       thisblockrow = buffer[block_row];
       (*cinfo->fdct->forward_DCT) (cinfo, compptr,
-				   input_buf[ci], thisblockrow,
-				   (JDIMENSION) (block_row * DCTSIZE),
-				   (JDIMENSION) 0, blocks_across);
+                                   input_buf[ci], thisblockrow,
+                                   (JDIMENSION) (block_row * DCTSIZE),
+                                   (JDIMENSION) 0, blocks_across);
       if (ndummy > 0) {
-	/* Create dummy blocks at the right edge of the image. */
-	thisblockrow += blocks_across; /* => first dummy block */
-	jzero_far((void FAR *) thisblockrow, ndummy * SIZEOF(JBLOCK));
-	lastDC = thisblockrow[-1][0];
-	for (bi = 0; bi < ndummy; bi++) {
-	  thisblockrow[bi][0] = lastDC;
-	}
+        /* Create dummy blocks at the right edge of the image. */
+        thisblockrow += blocks_across; /* => first dummy block */
+        jzero_far((void *) thisblockrow, ndummy * sizeof(JBLOCK));
+        lastDC = thisblockrow[-1][0];
+        for (bi = 0; bi < ndummy; bi++) {
+          thisblockrow[bi][0] = lastDC;
+        }
       }
     }
     /* If at end of image, create dummy block rows as needed.
@@ -299,22 +299,22 @@
      * This squeezes a few more bytes out of the resulting file...
      */
     if (coef->iMCU_row_num == last_iMCU_row) {
-      blocks_across += ndummy;	/* include lower right corner */
+      blocks_across += ndummy;  /* include lower right corner */
       MCUs_across = blocks_across / h_samp_factor;
       for (block_row = block_rows; block_row < compptr->v_samp_factor;
-	   block_row++) {
-	thisblockrow = buffer[block_row];
-	lastblockrow = buffer[block_row-1];
-	jzero_far((void FAR *) thisblockrow,
-		  (size_t) (blocks_across * SIZEOF(JBLOCK)));
-	for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
-	  lastDC = lastblockrow[h_samp_factor-1][0];
-	  for (bi = 0; bi < h_samp_factor; bi++) {
-	    thisblockrow[bi][0] = lastDC;
-	  }
-	  thisblockrow += h_samp_factor; /* advance to next MCU in row */
-	  lastblockrow += h_samp_factor;
-	}
+           block_row++) {
+        thisblockrow = buffer[block_row];
+        lastblockrow = buffer[block_row-1];
+        jzero_far((void *) thisblockrow,
+                  (size_t) (blocks_across * sizeof(JBLOCK)));
+        for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
+          lastDC = lastblockrow[h_samp_factor-1][0];
+          for (bi = 0; bi < h_samp_factor; bi++) {
+            thisblockrow[bi][0] = lastDC;
+          }
+          thisblockrow += h_samp_factor; /* advance to next MCU in row */
+          lastblockrow += h_samp_factor;
+        }
       }
     }
   }
@@ -341,7 +341,7 @@
 compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   int blkn, ci, xindex, yindex, yoffset;
   JDIMENSION start_col;
   JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
@@ -364,25 +364,25 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
-	    coef->MCU_buffer[blkn++] = buffer_ptr++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+          }
+        }
       }
       /* Try to write the MCU. */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -408,7 +408,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_c_coef_controller *) coef;
   coef->pub.start_pass = start_pass_coef;
 
@@ -421,14 +421,14 @@
     jpeg_component_info *compptr;
 
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
-				(long) compptr->h_samp_factor),
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor),
-	 (JDIMENSION) compptr->v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+         (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+                                (long) compptr->h_samp_factor),
+         (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+                                (long) compptr->v_samp_factor),
+         (JDIMENSION) compptr->v_samp_factor);
     }
 #else
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
@@ -440,7 +440,7 @@
 
     buffer = (JBLOCKROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                  C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
     for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
       coef->MCU_buffer[i] = buffer + i;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolext.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolext.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,148 @@
+/*
+ * jccolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+
+/* This file is included by jccolor.c */
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+rgb_ycc_convert_internal (j_compress_ptr cinfo,
+                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                          JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+                ((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+                 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+                ((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+INLINE
+LOCAL(void)
+rgb_gray_convert_internal (j_compress_ptr cinfo,
+                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                           JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptr[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles extended RGB->plain RGB conversion
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal (j_compress_ptr cinfo,
+                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                          JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptr0[col] = GETJSAMPLE(inptr[RGB_RED]);
+      outptr1[col] = GETJSAMPLE(inptr[RGB_GREEN]);
+      outptr2[col] = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolext.c.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,114 +0,0 @@
-/*
- * jccolext.c
- *
- * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains input colorspace conversion routines.
- */
-
-
-/* This file is included by jccolor.c */
-
-
-/*
- * Convert some rows of samples to the JPEG colorspace.
- *
- * Note that we change from the application's interleaved-pixel format
- * to our internal noninterleaved, one-plane-per-component format.
- * The input buffer is therefore three times as wide as the output buffer.
- *
- * A starting row offset is provided only for the output buffer.  The caller
- * can easily adjust the passed input_buf value to accommodate any row
- * offset required on that side.
- */
-
-INLINE
-LOCAL(void)
-rgb_ycc_convert_internal (j_compress_ptr cinfo,
-                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-                          JDIMENSION output_row, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr0, outptr1, outptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr0 = output_buf[0][output_row];
-    outptr1 = output_buf[1][output_row];
-    outptr2 = output_buf[2][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
-      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
-       * must be too; we do not need an explicit range-limiting operation.
-       * Hence the value being shifted is never negative, and we don't
-       * need the general RIGHT_SHIFT macro.
-       */
-      /* Y */
-      outptr0[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-      /* Cb */
-      outptr1[col] = (JSAMPLE)
-		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
-		 >> SCALEBITS);
-      /* Cr */
-      outptr2[col] = (JSAMPLE)
-		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
-		 >> SCALEBITS);
-    }
-  }
-}
-
-
-/**************** Cases other than RGB -> YCbCr **************/
-
-
-/*
- * Convert some rows of samples to the JPEG colorspace.
- * This version handles RGB->grayscale conversion, which is the same
- * as the RGB->Y portion of RGB->YCbCr.
- * We assume rgb_ycc_start has been called (we only use the Y tables).
- */
-
-INLINE
-LOCAL(void)
-rgb_gray_convert_internal (j_compress_ptr cinfo,
-                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-                           JDIMENSION output_row, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr = output_buf[0][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
-      /* Y */
-      outptr[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-    }
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolor.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolor.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jccolor.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jccolor.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jccolor.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009-2012, 2015 D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains input colorspace conversion routines.
  */
@@ -14,7 +17,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 
 /* Private subobject */
@@ -23,10 +26,10 @@
   struct jpeg_color_converter pub; /* public fields */
 
   /* Private state for RGB->YCC conversion */
-  INT32 * rgb_ycc_tab;		/* => table for RGB to YCbCr conversion */
+  JLONG *rgb_ycc_tab;           /* => table for RGB to YCbCr conversion */
 } my_color_converter;
 
-typedef my_color_converter * my_cconvert_ptr;
+typedef my_color_converter *my_cconvert_ptr;
 
 
 /**************** RGB -> YCbCr conversion: most common case **************/
@@ -35,9 +38,9 @@
  * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
  * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
  * The conversion equations to be implemented are therefore
- *	Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
- *	Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
- *	Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
+ *      Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *      Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
+ *      Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
  * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
  * Note: older versions of the IJG code used a zero offset of MAXJSAMPLE/2,
  * rather than CENTERJSAMPLE, for Cb and Cr.  This gave equal positive and
@@ -59,10 +62,10 @@
  * in the tables to save adding them separately in the inner loop.
  */
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define CBCR_OFFSET	((INT32) CENTERJSAMPLE << SCALEBITS)
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define CBCR_OFFSET     ((JLONG) CENTERJSAMPLE << SCALEBITS)
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
 
 /* We allocate one big table and divide it up into eight parts, instead of
  * doing eight alloc_small requests.  This lets us use a single table base
@@ -70,21 +73,21 @@
  * machines (more than can hold all eight addresses, anyway).
  */
 
-#define R_Y_OFF		0			/* offset to R => Y section */
-#define G_Y_OFF		(1*(MAXJSAMPLE+1))	/* offset to G => Y section */
-#define B_Y_OFF		(2*(MAXJSAMPLE+1))	/* etc. */
-#define R_CB_OFF	(3*(MAXJSAMPLE+1))
-#define G_CB_OFF	(4*(MAXJSAMPLE+1))
-#define B_CB_OFF	(5*(MAXJSAMPLE+1))
-#define R_CR_OFF	B_CB_OFF		/* B=>Cb, R=>Cr are the same */
-#define G_CR_OFF	(6*(MAXJSAMPLE+1))
-#define B_CR_OFF	(7*(MAXJSAMPLE+1))
-#define TABLE_SIZE	(8*(MAXJSAMPLE+1))
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1*(MAXJSAMPLE+1))      /* offset to G => Y section */
+#define B_Y_OFF         (2*(MAXJSAMPLE+1))      /* etc. */
+#define R_CB_OFF        (3*(MAXJSAMPLE+1))
+#define G_CB_OFF        (4*(MAXJSAMPLE+1))
+#define B_CB_OFF        (5*(MAXJSAMPLE+1))
+#define R_CR_OFF        B_CB_OFF                /* B=>Cb, R=>Cr are the same */
+#define G_CR_OFF        (6*(MAXJSAMPLE+1))
+#define B_CR_OFF        (7*(MAXJSAMPLE+1))
+#define TABLE_SIZE      (8*(MAXJSAMPLE+1))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jccolext.c.inc"
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -96,13 +99,15 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define rgb_ycc_convert_internal extrgb_ycc_convert_internal
 #define rgb_gray_convert_internal extrgb_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extrgb_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
@@ -110,13 +115,15 @@
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define rgb_ycc_convert_internal extrgbx_ycc_convert_internal
 #define rgb_gray_convert_internal extrgbx_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extrgbx_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGR_RED
 #define RGB_GREEN EXT_BGR_GREEN
@@ -124,13 +131,15 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define rgb_ycc_convert_internal extbgr_ycc_convert_internal
 #define rgb_gray_convert_internal extbgr_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extbgr_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
@@ -138,13 +147,15 @@
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define rgb_ycc_convert_internal extbgrx_ycc_convert_internal
 #define rgb_gray_convert_internal extbgrx_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extbgrx_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
@@ -152,13 +163,15 @@
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define rgb_ycc_convert_internal extxbgr_ycc_convert_internal
 #define rgb_gray_convert_internal extxbgr_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extxbgr_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
@@ -166,13 +179,15 @@
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define rgb_ycc_convert_internal extxrgb_ycc_convert_internal
 #define rgb_gray_convert_internal extxrgb_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extxrgb_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 
 /*
@@ -183,13 +198,13 @@
 rgb_ycc_start (j_compress_ptr cinfo)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  INT32 * rgb_ycc_tab;
-  INT32 i;
+  JLONG *rgb_ycc_tab;
+  JLONG i;
 
   /* Allocate and fill in the conversion tables. */
-  cconvert->rgb_ycc_tab = rgb_ycc_tab = (INT32 *)
+  cconvert->rgb_ycc_tab = rgb_ycc_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(TABLE_SIZE * SIZEOF(INT32)));
+                                (TABLE_SIZE * sizeof(JLONG)));
 
   for (i = 0; i <= MAXJSAMPLE; i++) {
     rgb_ycc_tab[i+R_Y_OFF] = FIX(0.29900) * i;
@@ -217,8 +232,8 @@
 
 METHODDEF(void)
 rgb_ycc_convert (j_compress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		 JDIMENSION output_row, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                 JDIMENSION output_row, int num_rows)
 {
   switch (cinfo->in_color_space) {
     case JCS_EXT_RGB:
@@ -266,8 +281,8 @@
 
 METHODDEF(void)
 rgb_gray_convert (j_compress_ptr cinfo,
-		  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		  JDIMENSION output_row, int num_rows)
+                  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                  JDIMENSION output_row, int num_rows)
 {
   switch (cinfo->in_color_space) {
     case JCS_EXT_RGB:
@@ -307,6 +322,52 @@
 
 
 /*
+ * Extended RGB to plain RGB conversion
+ */
+
+METHODDEF(void)
+rgb_rgb_convert (j_compress_ptr cinfo,
+                  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                  JDIMENSION output_row, int num_rows)
+{
+  switch (cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      extrgbx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      extbgrx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      extxbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      extxrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    default:
+      rgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                               num_rows);
+      break;
+  }
+}
+
+
+/*
  * Convert some rows of samples to the JPEG colorspace.
  * This version handles Adobe-style CMYK->YCCK conversion,
  * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
@@ -316,12 +377,12 @@
 
 METHODDEF(void)
 cmyk_ycck_convert (j_compress_ptr cinfo,
-		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		   JDIMENSION output_row, int num_rows)
+                   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                   JDIMENSION output_row, int num_rows)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JLONG *ctab = cconvert->rgb_ycc_tab;
   register JSAMPROW inptr;
   register JSAMPROW outptr0, outptr1, outptr2, outptr3;
   register JDIMENSION col;
@@ -339,7 +400,7 @@
       g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
       b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
       /* K passes through as-is */
-      outptr3[col] = inptr[3];	/* don't need GETJSAMPLE here */
+      outptr3[col] = inptr[3];  /* don't need GETJSAMPLE here */
       inptr += 4;
       /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
        * must be too; we do not need an explicit range-limiting operation.
@@ -348,16 +409,16 @@
        */
       /* Y */
       outptr0[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
       /* Cb */
       outptr1[col] = (JSAMPLE)
-		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+                 >> SCALEBITS);
       /* Cr */
       outptr2[col] = (JSAMPLE)
-		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+                 >> SCALEBITS);
     }
   }
 }
@@ -371,8 +432,8 @@
 
 METHODDEF(void)
 grayscale_convert (j_compress_ptr cinfo,
-		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		   JDIMENSION output_row, int num_rows)
+                   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                   JDIMENSION output_row, int num_rows)
 {
   register JSAMPROW inptr;
   register JSAMPROW outptr;
@@ -385,7 +446,7 @@
     outptr = output_buf[0][output_row];
     output_row++;
     for (col = 0; col < num_cols; col++) {
-      outptr[col] = inptr[0];	/* don't need GETJSAMPLE() here */
+      outptr[col] = inptr[0];   /* don't need GETJSAMPLE() here */
       inptr += instride;
     }
   }
@@ -400,28 +461,58 @@
 
 METHODDEF(void)
 null_convert (j_compress_ptr cinfo,
-	      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-	      JDIMENSION output_row, int num_rows)
+              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+              JDIMENSION output_row, int num_rows)
 {
   register JSAMPROW inptr;
-  register JSAMPROW outptr;
+  register JSAMPROW outptr, outptr0, outptr1, outptr2, outptr3;
   register JDIMENSION col;
   register int ci;
   int nc = cinfo->num_components;
   JDIMENSION num_cols = cinfo->image_width;
 
-  while (--num_rows >= 0) {
-    /* It seems fastest to make a separate pass for each component. */
-    for (ci = 0; ci < nc; ci++) {
-      inptr = *input_buf;
-      outptr = output_buf[ci][output_row];
+  if (nc == 3) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf[0][output_row];
+      outptr1 = output_buf[1][output_row];
+      outptr2 = output_buf[2][output_row];
+      output_row++;
       for (col = 0; col < num_cols; col++) {
-	outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
-	inptr += nc;
+        outptr0[col] = *inptr++;
+        outptr1[col] = *inptr++;
+        outptr2[col] = *inptr++;
       }
     }
-    input_buf++;
-    output_row++;
+  } else if (nc == 4) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf[0][output_row];
+      outptr1 = output_buf[1][output_row];
+      outptr2 = output_buf[2][output_row];
+      outptr3 = output_buf[3][output_row];
+      output_row++;
+      for (col = 0; col < num_cols; col++) {
+        outptr0[col] = *inptr++;
+        outptr1[col] = *inptr++;
+        outptr2[col] = *inptr++;
+        outptr3[col] = *inptr++;
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      /* It seems fastest to make a separate pass for each component. */
+      for (ci = 0; ci < nc; ci++) {
+        inptr = *input_buf;
+        outptr = output_buf[ci][output_row];
+        for (col = 0; col < num_cols; col++) {
+          outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
+          inptr += nc;
+        }
+      }
+      input_buf++;
+      output_row++;
+    }
   }
 }
 
@@ -448,7 +539,7 @@
 
   cconvert = (my_cconvert_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_color_converter));
+                                sizeof(my_color_converter));
   cinfo->cconvert = (struct jpeg_color_converter *) cconvert;
   /* set start_pass to null method until we find out differently */
   cconvert->pub.start_pass = null_method;
@@ -486,7 +577,7 @@
       ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
     break;
 
-  default:			/* JCS_UNKNOWN can be anything */
+  default:                      /* JCS_UNKNOWN can be anything */
     if (cinfo->input_components < 1)
       ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
     break;
@@ -523,21 +614,30 @@
     break;
 
   case JCS_RGB:
-  case JCS_EXT_RGB:
-  case JCS_EXT_RGBX:
-  case JCS_EXT_BGR:
-  case JCS_EXT_BGRX:
-  case JCS_EXT_XBGR:
-  case JCS_EXT_XRGB:
-  case JCS_EXT_RGBA:
-  case JCS_EXT_BGRA:
-  case JCS_EXT_ABGR:
-  case JCS_EXT_ARGB:
     if (cinfo->num_components != 3)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space == cinfo->jpeg_color_space &&
-      rgb_pixelsize[cinfo->in_color_space] == 3)
-      cconvert->pub.color_convert = null_convert;
+    if (rgb_red[cinfo->in_color_space] == 0 &&
+        rgb_green[cinfo->in_color_space] == 1 &&
+        rgb_blue[cinfo->in_color_space] == 2 &&
+        rgb_pixelsize[cinfo->in_color_space] == 3) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else if (cinfo->in_color_space == JCS_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGBX ||
+               cinfo->in_color_space == JCS_EXT_BGR ||
+               cinfo->in_color_space == JCS_EXT_BGRX ||
+               cinfo->in_color_space == JCS_EXT_XBGR ||
+               cinfo->in_color_space == JCS_EXT_XRGB ||
+               cinfo->in_color_space == JCS_EXT_RGBA ||
+               cinfo->in_color_space == JCS_EXT_BGRA ||
+               cinfo->in_color_space == JCS_EXT_ABGR ||
+               cinfo->in_color_space == JCS_EXT_ARGB)
+      cconvert->pub.color_convert = rgb_rgb_convert;
     else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
@@ -562,18 +662,28 @@
         cconvert->pub.start_pass = rgb_ycc_start;
         cconvert->pub.color_convert = rgb_ycc_convert;
       }
-    } else if (cinfo->in_color_space == JCS_YCbCr)
-      cconvert->pub.color_convert = null_convert;
-    else
+    } else if (cinfo->in_color_space == JCS_YCbCr) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
   case JCS_CMYK:
     if (cinfo->num_components != 4)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space == JCS_CMYK)
-      cconvert->pub.color_convert = null_convert;
-    else
+    if (cinfo->in_color_space == JCS_CMYK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
@@ -583,17 +693,27 @@
     if (cinfo->in_color_space == JCS_CMYK) {
       cconvert->pub.start_pass = rgb_ycc_start;
       cconvert->pub.color_convert = cmyk_ycck_convert;
-    } else if (cinfo->in_color_space == JCS_YCCK)
-      cconvert->pub.color_convert = null_convert;
-    else
+    } else if (cinfo->in_color_space == JCS_YCCK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
-  default:			/* allow null conversion of JCS_UNKNOWN */
+  default:                      /* allow null conversion of JCS_UNKNOWN */
     if (cinfo->jpeg_color_space != cinfo->in_color_space ||
-	cinfo->num_components != cinfo->input_components)
+        cinfo->num_components != cinfo->input_components)
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    cconvert->pub.color_convert = null_convert;
+#if defined(__mips__)
+    if (jsimd_c_can_null_convert())
+      cconvert->pub.color_convert = jsimd_c_null_convert;
+    else
+#endif
+      cconvert->pub.color_convert = null_convert;
     break;
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcdctmgr.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcdctmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcdctmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcdctmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,14 @@
 /*
  * jcdctmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2011 D. R. Commander
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2011, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the forward-DCT management logic.
  * This code selects a particular DCT implementation to be used,
@@ -17,33 +19,32 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 #include "jsimddct.h"
 
 
 /* Private subobject for this module */
 
-typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data));
-typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
+typedef void (*forward_DCT_method_ptr) (DCTELEM *data);
+typedef void (*float_DCT_method_ptr) (FAST_FLOAT *data);
 
-typedef JMETHOD(void, convsamp_method_ptr,
-                (JSAMPARRAY sample_data, JDIMENSION start_col,
-                 DCTELEM * workspace));
-typedef JMETHOD(void, float_convsamp_method_ptr,
-                (JSAMPARRAY sample_data, JDIMENSION start_col,
-                 FAST_FLOAT *workspace));
-
-typedef JMETHOD(void, quantize_method_ptr,
-                (JCOEFPTR coef_block, DCTELEM * divisors,
-                 DCTELEM * workspace));
-typedef JMETHOD(void, float_quantize_method_ptr,
-                (JCOEFPTR coef_block, FAST_FLOAT * divisors,
-                 FAST_FLOAT * workspace));
+typedef void (*convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                     JDIMENSION start_col,
+                                     DCTELEM *workspace);
+typedef void (*float_convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                           JDIMENSION start_col,
+                                           FAST_FLOAT *workspace);
+
+typedef void (*quantize_method_ptr) (JCOEFPTR coef_block, DCTELEM *divisors,
+                                     DCTELEM *workspace);
+typedef void (*float_quantize_method_ptr) (JCOEFPTR coef_block,
+                                           FAST_FLOAT *divisors,
+                                           FAST_FLOAT *workspace);
 
 METHODDEF(void) quantize (JCOEFPTR, DCTELEM *, DCTELEM *);
 
 typedef struct {
-  struct jpeg_forward_dct pub;	/* public fields */
+  struct jpeg_forward_dct pub;  /* public fields */
 
   /* Pointer to the DCT routine actually in use */
   forward_DCT_method_ptr dct;
@@ -54,27 +55,30 @@
    * entries, because of scaling (especially for an unnormalized DCT).
    * Each table is given in normal array order.
    */
-  DCTELEM * divisors[NUM_QUANT_TBLS];
+  DCTELEM *divisors[NUM_QUANT_TBLS];
 
   /* work area for FDCT subroutine */
-  DCTELEM * workspace;
+  DCTELEM *workspace;
 
 #ifdef DCT_FLOAT_SUPPORTED
   /* Same as above for the floating-point case. */
   float_DCT_method_ptr float_dct;
   float_convsamp_method_ptr float_convsamp;
   float_quantize_method_ptr float_quantize;
-  FAST_FLOAT * float_divisors[NUM_QUANT_TBLS];
-  FAST_FLOAT * float_workspace;
+  FAST_FLOAT *float_divisors[NUM_QUANT_TBLS];
+  FAST_FLOAT *float_workspace;
 #endif
 } my_fdct_controller;
 
-typedef my_fdct_controller * my_fdct_ptr;
+typedef my_fdct_controller *my_fdct_ptr;
 
 
+#if BITS_IN_JSAMPLE == 8
+
 /*
  * Find the highest bit in an integer through binary search.
  */
+
 LOCAL(int)
 flss (UINT16 val)
 {
@@ -105,6 +109,7 @@
   return bit;
 }
 
+
 /*
  * Compute values to do a division using reciprocal.
  *
@@ -146,7 +151,7 @@
  *
  * In order to allow SIMD implementations we also tweak the values to
  * allow the same calculation to be made at all times:
- * 
+ *
  *   dctbl[0] = f rounded to nearest integer
  *   dctbl[1] = divisor / 2 (+ 1 if fractional part of f < 0.5)
  *   dctbl[2] = 1 << ((word size) * 2 - r)
@@ -163,13 +168,27 @@
  * of in a consecutive manner, yet again in order to allow SIMD
  * routines.
  */
+
 LOCAL(int)
-compute_reciprocal (UINT16 divisor, DCTELEM * dtbl)
+compute_reciprocal (UINT16 divisor, DCTELEM *dtbl)
 {
   UDCTELEM2 fq, fr;
   UDCTELEM c;
   int b, r;
 
+  if (divisor == 1) {
+    /* divisor == 1 means unquantized, so these reciprocal/correction/shift
+     * values will cause the C quantization algorithm to act like the
+     * identity function.  Since only the C quantization algorithm is used in
+     * these cases, the scale value is irrelevant.
+     */
+    dtbl[DCTSIZE2 * 0] = (DCTELEM) 1;                       /* reciprocal */
+    dtbl[DCTSIZE2 * 1] = (DCTELEM) 0;                       /* correction */
+    dtbl[DCTSIZE2 * 2] = (DCTELEM) 1;                       /* scale */
+    dtbl[DCTSIZE2 * 3] = -(DCTELEM) (sizeof(DCTELEM) * 8);  /* shift */
+    return 0;
+  }
+
   b = flss(divisor) - 1;
   r  = sizeof(DCTELEM) * 8 + b;
 
@@ -190,13 +209,20 @@
 
   dtbl[DCTSIZE2 * 0] = (DCTELEM) fq;      /* reciprocal */
   dtbl[DCTSIZE2 * 1] = (DCTELEM) c;       /* correction + roundfactor */
+#ifdef WITH_SIMD
   dtbl[DCTSIZE2 * 2] = (DCTELEM) (1 << (sizeof(DCTELEM)*8*2 - r));  /* scale */
+#else
+  dtbl[DCTSIZE2 * 2] = 1;
+#endif
   dtbl[DCTSIZE2 * 3] = (DCTELEM) r - sizeof(DCTELEM)*8; /* shift */
 
   if(r <= 16) return 0;
   else return 1;
 }
 
+#endif
+
+
 /*
  * Initialize for a processing pass.
  * Verify that all referenced Q-tables are present, and set up
@@ -212,15 +238,15 @@
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
   int ci, qtblno, i;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtbl;
-  DCTELEM * dtbl;
+  JQUANT_TBL *qtbl;
+  DCTELEM *dtbl;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     qtblno = compptr->quant_tbl_no;
     /* Make sure specified quantization table is present */
     if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
-	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+        cinfo->quant_tbl_ptrs[qtblno] == NULL)
       ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
     qtbl = cinfo->quant_tbl_ptrs[qtblno];
     /* Compute divisors for this quant table */
@@ -232,91 +258,102 @@
        * coefficients multiplied by 8 (to counteract scaling).
        */
       if (fdct->divisors[qtblno] == NULL) {
-	fdct->divisors[qtblno] = (DCTELEM *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      (DCTSIZE2 * 4) * SIZEOF(DCTELEM));
+        fdct->divisors[qtblno] = (DCTELEM *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      (DCTSIZE2 * 4) * sizeof(DCTELEM));
       }
       dtbl = fdct->divisors[qtblno];
       for (i = 0; i < DCTSIZE2; i++) {
-	if(!compute_reciprocal(qtbl->quantval[i] << 3, &dtbl[i])
-	  && fdct->quantize == jsimd_quantize)
-	  fdct->quantize = quantize;
+#if BITS_IN_JSAMPLE == 8
+        if (!compute_reciprocal(qtbl->quantval[i] << 3, &dtbl[i]) &&
+            fdct->quantize == jsimd_quantize)
+          fdct->quantize = quantize;
+#else
+        dtbl[i] = ((DCTELEM) qtbl->quantval[i]) << 3;
+#endif
       }
       break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
     case JDCT_IFAST:
       {
-	/* For AA&N IDCT method, divisors are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * We apply a further scale factor of 8.
-	 */
+        /* For AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         */
 #define CONST_BITS 14
-	static const INT16 aanscales[DCTSIZE2] = {
-	  /* precomputed values scaled up by 14 bits */
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
-	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
-	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
-	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
-	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
-	};
-	SHIFT_TEMPS
-
-	if (fdct->divisors[qtblno] == NULL) {
-	  fdct->divisors[qtblno] = (DCTELEM *)
-	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-					(DCTSIZE2 * 4) * SIZEOF(DCTELEM));
-	}
-	dtbl = fdct->divisors[qtblno];
-	for (i = 0; i < DCTSIZE2; i++) {
-	  if(!compute_reciprocal(
-	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
-				  (INT32) aanscales[i]),
-		    CONST_BITS-3), &dtbl[i])
-	    && fdct->quantize == jsimd_quantize)
-	    fdct->quantize = quantize;
-	}
+        static const INT16 aanscales[DCTSIZE2] = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        if (fdct->divisors[qtblno] == NULL) {
+          fdct->divisors[qtblno] = (DCTELEM *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                        (DCTSIZE2 * 4) * sizeof(DCTELEM));
+        }
+        dtbl = fdct->divisors[qtblno];
+        for (i = 0; i < DCTSIZE2; i++) {
+#if BITS_IN_JSAMPLE == 8
+          if (!compute_reciprocal(
+                DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                      (JLONG) aanscales[i]),
+                        CONST_BITS-3), &dtbl[i]) &&
+              fdct->quantize == jsimd_quantize)
+            fdct->quantize = quantize;
+#else
+           dtbl[i] = (DCTELEM)
+             DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                   (JLONG) aanscales[i]),
+                     CONST_BITS-3);
+#endif
+        }
       }
       break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
     case JDCT_FLOAT:
       {
-	/* For float AA&N IDCT method, divisors are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * We apply a further scale factor of 8.
-	 * What's actually stored is 1/divisor so that the inner loop can
-	 * use a multiplication rather than a division.
-	 */
-	FAST_FLOAT * fdtbl;
-	int row, col;
-	static const double aanscalefactor[DCTSIZE] = {
-	  1.0, 1.387039845, 1.306562965, 1.175875602,
-	  1.0, 0.785694958, 0.541196100, 0.275899379
-	};
-
-	if (fdct->float_divisors[qtblno] == NULL) {
-	  fdct->float_divisors[qtblno] = (FAST_FLOAT *)
-	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-					DCTSIZE2 * SIZEOF(FAST_FLOAT));
-	}
-	fdtbl = fdct->float_divisors[qtblno];
-	i = 0;
-	for (row = 0; row < DCTSIZE; row++) {
-	  for (col = 0; col < DCTSIZE; col++) {
-	    fdtbl[i] = (FAST_FLOAT)
-	      (1.0 / (((double) qtbl->quantval[i] *
-		       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
-	    i++;
-	  }
-	}
+        /* For float AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         * What's actually stored is 1/divisor so that the inner loop can
+         * use a multiplication rather than a division.
+         */
+        FAST_FLOAT *fdtbl;
+        int row, col;
+        static const double aanscalefactor[DCTSIZE] = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        if (fdct->float_divisors[qtblno] == NULL) {
+          fdct->float_divisors[qtblno] = (FAST_FLOAT *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                        DCTSIZE2 * sizeof(FAST_FLOAT));
+        }
+        fdtbl = fdct->float_divisors[qtblno];
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fdtbl[i] = (FAST_FLOAT)
+              (1.0 / (((double) qtbl->quantval[i] *
+                       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
+            i++;
+          }
+        }
       }
       break;
 #endif
@@ -333,7 +370,7 @@
  */
 
 METHODDEF(void)
-convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace)
+convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace)
 {
   register DCTELEM *workspaceptr;
   register JSAMPROW elemptr;
@@ -343,7 +380,7 @@
   for (elemr = 0; elemr < DCTSIZE; elemr++) {
     elemptr = sample_data[elemr] + start_col;
 
-#if DCTSIZE == 8		/* unroll the inner loop */
+#if DCTSIZE == 8                /* unroll the inner loop */
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
@@ -368,14 +405,18 @@
  */
 
 METHODDEF(void)
-quantize (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace)
+quantize (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
 {
   int i;
   DCTELEM temp;
-  UDCTELEM recip, corr, shift;
-  UDCTELEM2 product;
   JCOEFPTR output_ptr = coef_block;
 
+#if BITS_IN_JSAMPLE == 8
+
+  UDCTELEM recip, corr;
+  int shift;
+  UDCTELEM2 product;
+
   for (i = 0; i < DCTSIZE2; i++) {
     temp = workspace[i];
     recip = divisors[i + DCTSIZE2 * 0];
@@ -386,16 +427,54 @@
       temp = -temp;
       product = (UDCTELEM2)(temp + corr) * recip;
       product >>= shift + sizeof(DCTELEM)*8;
-      temp = product;
+      temp = (DCTELEM)product;
       temp = -temp;
     } else {
       product = (UDCTELEM2)(temp + corr) * recip;
       product >>= shift + sizeof(DCTELEM)*8;
-      temp = product;
+      temp = (DCTELEM)product;
     }
+    output_ptr[i] = (JCOEF) temp;
+  }
+
+#else
+
+  register DCTELEM qval;
 
+  for (i = 0; i < DCTSIZE2; i++) {
+    qval = divisors[i];
+    temp = workspace[i];
+    /* Divide the coefficient value by qval, ensuring proper rounding.
+     * Since C does not specify the direction of rounding for negative
+     * quotients, we have to force the dividend positive for portability.
+     *
+     * In most files, at least half of the output values will be zero
+     * (at default quantization settings, more like three-quarters...)
+     * so we should ensure that this case is fast.  On many machines,
+     * a comparison is enough cheaper than a divide to make a special test
+     * a win.  Since both inputs will be nonnegative, we need only test
+     * for a < b to discover whether a/b is 0.
+     * If your machine's division is fast enough, define FAST_DIVIDE.
+     */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a,b)  a /= b
+#else
+#define DIVIDE_BY(a,b)  if (a >= b) a /= b; else a = 0
+#endif
+    if (temp < 0) {
+      temp = -temp;
+      temp += qval>>1;  /* for rounding */
+      DIVIDE_BY(temp, qval);
+      temp = -temp;
+    } else {
+      temp += qval>>1;  /* for rounding */
+      DIVIDE_BY(temp, qval);
+    }
     output_ptr[i] = (JCOEF) temp;
   }
+
+#endif
+
 }
 
 
@@ -408,16 +487,16 @@
  */
 
 METHODDEF(void)
-forward_DCT (j_compress_ptr cinfo, jpeg_component_info * compptr,
-	     JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-	     JDIMENSION start_row, JDIMENSION start_col,
-	     JDIMENSION num_blocks)
+forward_DCT (j_compress_ptr cinfo, jpeg_component_info *compptr,
+             JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+             JDIMENSION start_row, JDIMENSION start_col,
+             JDIMENSION num_blocks)
 /* This version is used for integer DCT implementations. */
 {
   /* This routine is heavily used, so it's worth coding it tightly. */
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
-  DCTELEM * divisors = fdct->divisors[compptr->quant_tbl_no];
-  DCTELEM * workspace;
+  DCTELEM *divisors = fdct->divisors[compptr->quant_tbl_no];
+  DCTELEM *workspace;
   JDIMENSION bi;
 
   /* Make sure the compiler doesn't look up these every pass */
@@ -426,7 +505,7 @@
   quantize_method_ptr do_quantize = fdct->quantize;
   workspace = fdct->workspace;
 
-  sample_data += start_row;	/* fold in the vertical offset once */
+  sample_data += start_row;     /* fold in the vertical offset once */
 
   for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
     /* Load data into workspace, applying unsigned->signed conversion */
@@ -445,7 +524,7 @@
 
 
 METHODDEF(void)
-convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace)
+convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace)
 {
   register FAST_FLOAT *workspaceptr;
   register JSAMPROW elemptr;
@@ -454,7 +533,7 @@
   workspaceptr = workspace;
   for (elemr = 0; elemr < DCTSIZE; elemr++) {
     elemptr = sample_data[elemr] + start_col;
-#if DCTSIZE == 8		/* unroll the inner loop */
+#if DCTSIZE == 8                /* unroll the inner loop */
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
@@ -476,7 +555,7 @@
 
 
 METHODDEF(void)
-quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace)
+quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace)
 {
   register FAST_FLOAT temp;
   register int i;
@@ -498,16 +577,16 @@
 
 
 METHODDEF(void)
-forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-		   JDIMENSION start_row, JDIMENSION start_col,
-		   JDIMENSION num_blocks)
+forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                   JDIMENSION start_row, JDIMENSION start_col,
+                   JDIMENSION num_blocks)
 /* This version is used for floating-point DCT implementations. */
 {
   /* This routine is heavily used, so it's worth coding it tightly. */
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
-  FAST_FLOAT * divisors = fdct->float_divisors[compptr->quant_tbl_no];
-  FAST_FLOAT * workspace;
+  FAST_FLOAT *divisors = fdct->float_divisors[compptr->quant_tbl_no];
+  FAST_FLOAT *workspace;
   JDIMENSION bi;
 
 
@@ -517,7 +596,7 @@
   float_quantize_method_ptr do_quantize = fdct->float_quantize;
   workspace = fdct->float_workspace;
 
-  sample_data += start_row;	/* fold in the vertical offset once */
+  sample_data += start_row;     /* fold in the vertical offset once */
 
   for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
     /* Load data into workspace, applying unsigned->signed conversion */
@@ -546,7 +625,7 @@
 
   fdct = (my_fdct_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_fdct_controller));
+                                sizeof(my_fdct_controller));
   cinfo->fdct = (struct jpeg_forward_dct *) fdct;
   fdct->pub.start_pass = start_pass_fdctmgr;
 
@@ -625,12 +704,12 @@
   if (cinfo->dct_method == JDCT_FLOAT)
     fdct->float_workspace = (FAST_FLOAT *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(FAST_FLOAT) * DCTSIZE2);
+                                  sizeof(FAST_FLOAT) * DCTSIZE2);
   else
 #endif
     fdct->workspace = (DCTELEM *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(DCTELEM) * DCTSIZE2);
+                                  sizeof(DCTELEM) * DCTSIZE2);
 
   /* Mark divisor tables unallocated */
   for (i = 0; i < NUM_QUANT_TBLS; i++) {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jchuff.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jchuff.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jchuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jchuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jchuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2014-2016 D. R. Commander.
+ * Copyright (C) 2015 Matthieu Darbois.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy encoding routines.
  *
@@ -18,11 +21,39 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jchuff.h"		/* Declarations shared with jcphuff.c */
+#include "jsimd.h"
+#include "jconfigint.h"
 #include <limits.h>
 
-static unsigned char jpeg_nbits_table[65536];
-static int jpeg_nbits_table_init = 0;
+/*
+ * NOTE: If USE_CLZ_INTRINSIC is defined, then clz/bsr instructions will be
+ * used for bit counting rather than the lookup table.  This will reduce the
+ * memory footprint by 64k, which is important for some mobile applications
+ * that create many isolated instances of libjpeg-turbo (web browsers, for
+ * instance.)  This may improve performance on some mobile platforms as well.
+ * This feature is enabled by default only on ARM processors, because some x86
+ * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
+ * shown to have a significant performance impact even on the x86 chips that
+ * have a fast implementation of it.  When building for ARMv6, you can
+ * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
+ * flags (this defines __thumb__).
+ */
+
+/* NOTE: Both GCC and Clang define __GNUC__ */
+#if defined __GNUC__ && (defined __arm__ || defined __aarch64__)
+#if !defined __thumb__ || defined __thumb2__
+#define USE_CLZ_INTRINSIC
+#endif
+#endif
+
+#ifdef USE_CLZ_INTRINSIC
+#define JPEG_NBITS_NONZERO(x) (32 - __builtin_clz(x))
+#define JPEG_NBITS(x) (x ? JPEG_NBITS_NONZERO(x) : 0)
+#else
+#include "jpeg_nbits_table.h"
+#define JPEG_NBITS(x) (jpeg_nbits_table[x])
+#define JPEG_NBITS_NONZERO(x) JPEG_NBITS(x)
+#endif
 
 #ifndef min
  #define min(a,b) ((a)<(b)?(a):(b))
@@ -36,8 +67,8 @@
  */
 
 typedef struct {
-  size_t put_buffer;		/* current bit-accumulation buffer */
-  int put_bits;			/* # of bits now in it */
+  size_t put_buffer;            /* current bit-accumulation buffer */
+  int put_bits;                 /* # of bits now in it */
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 } savable_state;
 
@@ -51,12 +82,12 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).put_buffer = (src).put_buffer, \
-	 (dest).put_bits = (src).put_bits, \
-	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).put_buffer = (src).put_buffer, \
+         (dest).put_bits = (src).put_bits, \
+         (dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -64,44 +95,45 @@
 typedef struct {
   struct jpeg_entropy_encoder pub; /* public fields */
 
-  savable_state saved;		/* Bit buffer & DC state at start of MCU */
+  savable_state saved;          /* Bit buffer & DC state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  c_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
-  c_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *dc_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *ac_derived_tbls[NUM_HUFF_TBLS];
 
-#ifdef ENTROPY_OPT_SUPPORTED	/* Statistics tables for optimization */
-  long * dc_count_ptrs[NUM_HUFF_TBLS];
-  long * ac_count_ptrs[NUM_HUFF_TBLS];
+#ifdef ENTROPY_OPT_SUPPORTED    /* Statistics tables for optimization */
+  long *dc_count_ptrs[NUM_HUFF_TBLS];
+  long *ac_count_ptrs[NUM_HUFF_TBLS];
 #endif
+
+  int simd;
 } huff_entropy_encoder;
 
-typedef huff_entropy_encoder * huff_entropy_ptr;
+typedef huff_entropy_encoder *huff_entropy_ptr;
 
 /* Working state while writing an MCU.
  * This struct contains all the fields that are needed by subroutines.
  */
 
 typedef struct {
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
-  savable_state cur;		/* Current bit buffer & DC state */
-  j_compress_ptr cinfo;		/* dump_buffer needs access to this */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  savable_state cur;            /* Current bit buffer & DC state */
+  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
 } working_state;
 
 
 /* Forward declarations */
-METHODDEF(boolean) encode_mcu_huff JPP((j_compress_ptr cinfo,
-					JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_huff JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_huff (j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_huff (j_compress_ptr cinfo);
 #ifdef ENTROPY_OPT_SUPPORTED
-METHODDEF(boolean) encode_mcu_gather JPP((j_compress_ptr cinfo,
-					  JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_gather JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_gather (j_compress_ptr cinfo,
+                                      JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_gather (j_compress_ptr cinfo);
 #endif
 
 
@@ -116,7 +148,7 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, dctbl, actbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (gather_statistics) {
 #ifdef ENTROPY_OPT_SUPPORTED
@@ -130,6 +162,8 @@
     entropy->pub.finish_pass = finish_pass_huff;
   }
 
+  entropy->simd = jsimd_can_huff_encode_one_block();
+
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
     dctbl = compptr->dc_tbl_no;
@@ -139,29 +173,29 @@
       /* Check for invalid table indexes */
       /* (make_c_derived_tbl does this in the other path) */
       if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
-	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
       if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
-	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
       /* Allocate and zero the statistics tables */
       /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
       if (entropy->dc_count_ptrs[dctbl] == NULL)
-	entropy->dc_count_ptrs[dctbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * SIZEOF(long));
+        entropy->dc_count_ptrs[dctbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * sizeof(long));
       if (entropy->ac_count_ptrs[actbl] == NULL)
-	entropy->ac_count_ptrs[actbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * SIZEOF(long));
+        entropy->ac_count_ptrs[actbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * sizeof(long));
 #endif
     } else {
       /* Compute derived values for Huffman tables */
       /* We may do this more than once for a table, but it's not expensive */
       jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
-			      & entropy->dc_derived_tbls[dctbl]);
+                              & entropy->dc_derived_tbls[dctbl]);
       jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
-			      & entropy->ac_derived_tbls[actbl]);
+                              & entropy->ac_derived_tbls[actbl]);
     }
     /* Initialize DC predictions to 0 */
     entropy->saved.last_dc_val[ci] = 0;
@@ -186,7 +220,7 @@
 
 GLOBAL(void)
 jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
-			 c_derived_tbl ** pdtbl)
+                         c_derived_tbl **pdtbl)
 {
   JHUFF_TBL *htbl;
   c_derived_tbl *dtbl;
@@ -211,22 +245,22 @@
   if (*pdtbl == NULL)
     *pdtbl = (c_derived_tbl *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(c_derived_tbl));
+                                  sizeof(c_derived_tbl));
   dtbl = *pdtbl;
-  
+
   /* Figure C.1: make table of Huffman code length for each symbol */
 
   p = 0;
   for (l = 1; l <= 16; l++) {
     i = (int) htbl->bits[l];
-    if (i < 0 || p + i > 256)	/* protect against table overrun */
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     while (i--)
       huffsize[p++] = (char) l;
   }
   huffsize[p] = 0;
   lastp = p;
-  
+
   /* Figure C.2: generate the codes themselves */
   /* We also validate that the counts represent a legal Huffman code tree. */
 
@@ -241,12 +275,12 @@
     /* code is now 1 more than the last code used for codelength si; but
      * it must still fit in si bits, since no code is allowed to be all ones.
      */
-    if (((INT32) code) >= (((INT32) 1) << si))
+    if (((JLONG) code) >= (((JLONG) 1) << si))
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     code <<= 1;
     si++;
   }
-  
+
   /* Figure C.3: generate encoding tables */
   /* These are code and size indexed by symbol value */
 
@@ -254,7 +288,7 @@
    * this lets us detect duplicate VAL entries here, and later
    * allows emit_bits to detect any attempt to emit such symbols.
    */
-  MEMZERO(dtbl->ehufsi, SIZEOF(dtbl->ehufsi));
+  MEMZERO(dtbl->ehufsi, sizeof(dtbl->ehufsi));
 
   /* This is also a convenient place to check for out-of-range
    * and duplicated VAL entries.  We allow 0..255 for AC symbols
@@ -270,15 +304,6 @@
     dtbl->ehufco[i] = huffcode[p];
     dtbl->ehufsi[i] = huffsize[p];
   }
-
-  if(!jpeg_nbits_table_init) {
-    for(i = 0; i < 65536; i++) {
-      int nbits = 0, temp = i;
-      while (temp) {temp >>= 1;  nbits++;}
-      jpeg_nbits_table[i] = nbits;
-    }
-    jpeg_nbits_table_init = 1;
-  }
 }
 
 
@@ -286,19 +311,17 @@
 
 /* Emit a byte, taking 'action' if must suspend. */
 #define emit_byte(state,val,action)  \
-	{ *(state)->next_output_byte++ = (JOCTET) (val);  \
-	  if (--(state)->free_in_buffer == 0)  \
-	    if (! dump_buffer(state))  \
-	      { action; } }
+        { *(state)->next_output_byte++ = (JOCTET) (val);  \
+          if (--(state)->free_in_buffer == 0)  \
+            if (! dump_buffer(state))  \
+              { action; } }
 
 
 LOCAL(boolean)
-dump_buffer (working_state * state)
+dump_buffer (working_state *state)
 /* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
 {
-  struct jpeg_destination_mgr * dest = state->cinfo->dest;
-
-  dest->free_in_buffer = state->free_in_buffer;
+  struct jpeg_destination_mgr *dest = state->cinfo->dest;
 
   if (! (*dest->empty_output_buffer) (state->cinfo))
     return FALSE;
@@ -360,7 +383,11 @@
   } \
 }
 
-#if __WORDSIZE==64 || defined(_WIN64)
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
 #define EMIT_BITS(code, size) { \
   CHECKBUF47() \
@@ -368,7 +395,7 @@
 }
 
 #define EMIT_CODE(code, size) { \
-  temp2 &= (((INT32) 1)<<nbits) - 1; \
+  temp2 &= (((JLONG) 1)<<nbits) - 1; \
   CHECKBUF31() \
   PUT_BITS(code, size) \
   PUT_BITS(temp2, nbits) \
@@ -382,7 +409,7 @@
 }
 
 #define EMIT_CODE(code, size) { \
-  temp2 &= (((INT32) 1)<<nbits) - 1; \
+  temp2 &= (((JLONG) 1)<<nbits) - 1; \
   PUT_BITS(code, size) \
   CHECKBUF15() \
   PUT_BITS(temp2, nbits) \
@@ -392,7 +419,16 @@
 #endif
 
 
-#define BUFSIZE (DCTSIZE2 * 2)
+/* Although it is exceedingly rare, it is possible for a Huffman-encoded
+ * coefficient block to be larger than the 128-byte unencoded block.  For each
+ * of the 64 coefficients, PUT_BITS is invoked twice, and each invocation can
+ * theoretically store 16 bits (for a maximum of 2048 bits or 256 bytes per
+ * encoded block.)  If, for instance, one artificially sets the AC
+ * coefficients to alternating values of 32767 and -32768 (using the JPEG
+ * scanning order-- 1, 8, 16, etc.), then this will produce an encoded block
+ * larger than 200 bytes.
+ */
+#define BUFSIZE (DCTSIZE2 * 4)
 
 #define LOAD_BUFFER() { \
   if (state->free_in_buffer < BUFSIZE) { \
@@ -425,7 +461,7 @@
 
 
 LOCAL(boolean)
-flush_bits (working_state * state)
+flush_bits (working_state *state)
 {
   JOCTET _buffer[BUFSIZE], *buffer;
   size_t put_buffer;  int put_bits;
@@ -439,7 +475,7 @@
   PUT_BITS(0x7F, 7)
   while (put_bits >= 8) EMIT_BYTE()
 
-  state->cur.put_buffer = 0;	/* and reset bit-buffer to empty */
+  state->cur.put_buffer = 0;    /* and reset bit-buffer to empty */
   state->cur.put_bits = 0;
   STORE_BUFFER()
 
@@ -450,8 +486,25 @@
 /* Encode a single block's worth of coefficients */
 
 LOCAL(boolean)
-encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
-		  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+encode_one_block_simd (working_state *state, JCOEFPTR block, int last_dc_val,
+                       c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  JOCTET _buffer[BUFSIZE], *buffer;
+  size_t bytes, bytestocopy;  int localbuf = 0;
+
+  LOAD_BUFFER()
+
+  buffer = jsimd_huff_encode_one_block(state, buffer, block, last_dc_val,
+                                       dctbl, actbl);
+
+  STORE_BUFFER()
+
+  return TRUE;
+}
+
+LOCAL(boolean)
+encode_one_block (working_state *state, JCOEFPTR block, int last_dc_val,
+                  c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
   int temp, temp2, temp3;
   int nbits;
@@ -466,7 +519,7 @@
   LOAD_BUFFER()
 
   /* Encode the DC coefficient difference per section F.1.2.1 */
-  
+
   temp = temp2 = block[0] - last_dc_val;
 
  /* This is a well-known technique for obtaining the absolute value without a
@@ -483,25 +536,23 @@
   temp2 += temp3;
 
   /* Find the number of bits needed for the magnitude of the coefficient */
-  nbits = jpeg_nbits_table[temp];
+  nbits = JPEG_NBITS(temp);
 
   /* Emit the Huffman-coded symbol for the number of bits */
   code = dctbl->ehufco[nbits];
   size = dctbl->ehufsi[nbits];
-  PUT_BITS(code, size)
-  CHECKBUF15()
+  EMIT_BITS(code, size)
 
   /* Mask off any extra bits in code */
-  temp2 &= (((INT32) 1)<<nbits) - 1;
+  temp2 &= (((JLONG) 1)<<nbits) - 1;
 
   /* Emit that number of bits of the value, if positive, */
   /* or the complement of its magnitude, if negative. */
-  PUT_BITS(temp2, nbits)
-  CHECKBUF15()
+  EMIT_BITS(temp2, nbits)
 
   /* Encode the AC coefficients per section F.1.2.2 */
-  
-  r = 0;			/* r = run length of zeros */
+
+  r = 0;                        /* r = run length of zeros */
 
 /* Manually unroll the k loop to eliminate the counter variable.  This
  * improves performance greatly on systems with a limited number of
@@ -517,7 +568,7 @@
     temp ^= temp3; \
     temp -= temp3; \
     temp2 += temp3; \
-    nbits = jpeg_nbits_table[temp]; \
+    nbits = JPEG_NBITS_NONZERO(temp); \
     /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
     while (r > 15) { \
       EMIT_BITS(code_0xf0, size_0xf0) \
@@ -565,7 +616,7 @@
  */
 
 LOCAL(boolean)
-emit_restart (working_state * state, int restart_num)
+emit_restart (working_state *state, int restart_num)
 {
   int ci;
 
@@ -595,7 +646,7 @@
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   working_state state;
   int blkn, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Load up working state */
   state.next_output_byte = cinfo->dest->next_output_byte;
@@ -607,20 +658,34 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! emit_restart(&state, entropy->next_restart_num))
-	return FALSE;
+        return FALSE;
   }
 
   /* Encode the MCU data blocks */
-  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    ci = cinfo->MCU_membership[blkn];
-    compptr = cinfo->cur_comp_info[ci];
-    if (! encode_one_block(&state,
-			   MCU_data[blkn][0], state.cur.last_dc_val[ci],
-			   entropy->dc_derived_tbls[compptr->dc_tbl_no],
-			   entropy->ac_derived_tbls[compptr->ac_tbl_no]))
-      return FALSE;
-    /* Update last_dc_val */
-    state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+  if (entropy->simd) {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      if (! encode_one_block_simd(&state,
+                                  MCU_data[blkn][0], state.cur.last_dc_val[ci],
+                                  entropy->dc_derived_tbls[compptr->dc_tbl_no],
+                                  entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+    }
+  } else {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      if (! encode_one_block(&state,
+                             MCU_data[blkn][0], state.cur.last_dc_val[ci],
+                             entropy->dc_derived_tbls[compptr->dc_tbl_no],
+                             entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+    }
   }
 
   /* Completed MCU, so update state */
@@ -687,18 +752,18 @@
 
 LOCAL(void)
 htest_one_block (j_compress_ptr cinfo, JCOEFPTR block, int last_dc_val,
-		 long dc_counts[], long ac_counts[])
+                 long dc_counts[], long ac_counts[])
 {
   register int temp;
   register int nbits;
   register int k, r;
-  
+
   /* Encode the DC coefficient difference per section F.1.2.1 */
-  
+
   temp = block[0] - last_dc_val;
   if (temp < 0)
     temp = -temp;
-  
+
   /* Find the number of bits needed for the magnitude of the coefficient */
   nbits = 0;
   while (temp) {
@@ -713,36 +778,36 @@
 
   /* Count the Huffman symbol for the number of bits */
   dc_counts[nbits]++;
-  
+
   /* Encode the AC coefficients per section F.1.2.2 */
-  
-  r = 0;			/* r = run length of zeros */
-  
+
+  r = 0;                        /* r = run length of zeros */
+
   for (k = 1; k < DCTSIZE2; k++) {
     if ((temp = block[jpeg_natural_order[k]]) == 0) {
       r++;
     } else {
       /* if run length > 15, must emit special run-length-16 codes (0xF0) */
       while (r > 15) {
-	ac_counts[0xF0]++;
-	r -= 16;
+        ac_counts[0xF0]++;
+        r -= 16;
       }
-      
+
       /* Find the number of bits needed for the magnitude of the coefficient */
       if (temp < 0)
-	temp = -temp;
-      
+        temp = -temp;
+
       /* Find the number of bits needed for the magnitude of the coefficient */
-      nbits = 1;		/* there must be at least one 1 bit */
+      nbits = 1;                /* there must be at least one 1 bit */
       while ((temp >>= 1))
-	nbits++;
+        nbits++;
       /* Check for out-of-range coefficient values */
       if (nbits > MAX_COEF_BITS)
-	ERREXIT(cinfo, JERR_BAD_DCT_COEF);
-      
+        ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
       /* Count Huffman symbol for run length / number of bits */
       ac_counts[(r << 4) + nbits]++;
-      
+
       r = 0;
     }
   }
@@ -763,14 +828,14 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int blkn, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Take care of restart intervals if needed */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0) {
       /* Re-initialize DC predictions to 0 */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++)
-	entropy->saved.last_dc_val[ci] = 0;
+        entropy->saved.last_dc_val[ci] = 0;
       /* Update restart state */
       entropy->restarts_to_go = cinfo->restart_interval;
     }
@@ -781,8 +846,8 @@
     ci = cinfo->MCU_membership[blkn];
     compptr = cinfo->cur_comp_info[ci];
     htest_one_block(cinfo, MCU_data[blkn][0], entropy->saved.last_dc_val[ci],
-		    entropy->dc_count_ptrs[compptr->dc_tbl_no],
-		    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
+                    entropy->dc_count_ptrs[compptr->dc_tbl_no],
+                    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
     entropy->saved.last_dc_val[ci] = MCU_data[blkn][0][0];
   }
 
@@ -819,24 +884,24 @@
  */
 
 GLOBAL(void)
-jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
+jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL *htbl, long freq[])
 {
-#define MAX_CLEN 32		/* assumed maximum initial code length */
-  UINT8 bits[MAX_CLEN+1];	/* bits[k] = # of symbols with code length k */
-  int codesize[257];		/* codesize[k] = code length of symbol k */
-  int others[257];		/* next symbol in current branch of tree */
+#define MAX_CLEN 32             /* assumed maximum initial code length */
+  UINT8 bits[MAX_CLEN+1];       /* bits[k] = # of symbols with code length k */
+  int codesize[257];            /* codesize[k] = code length of symbol k */
+  int others[257];              /* next symbol in current branch of tree */
   int c1, c2;
   int p, i, j;
   long v;
 
   /* This algorithm is explained in section K.2 of the JPEG standard */
 
-  MEMZERO(bits, SIZEOF(bits));
-  MEMZERO(codesize, SIZEOF(codesize));
+  MEMZERO(bits, sizeof(bits));
+  MEMZERO(codesize, sizeof(codesize));
   for (i = 0; i < 257; i++)
-    others[i] = -1;		/* init links to empty */
-  
-  freq[256] = 1;		/* make sure 256 has a nonzero count */
+    others[i] = -1;             /* init links to empty */
+
+  freq[256] = 1;                /* make sure 256 has a nonzero count */
   /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
    * that no real symbol is given code-value of all ones, because 256
    * will be placed last in the largest codeword category.
@@ -851,8 +916,8 @@
     v = 1000000000L;
     for (i = 0; i <= 256; i++) {
       if (freq[i] && freq[i] <= v) {
-	v = freq[i];
-	c1 = i;
+        v = freq[i];
+        c1 = i;
       }
     }
 
@@ -862,15 +927,15 @@
     v = 1000000000L;
     for (i = 0; i <= 256; i++) {
       if (freq[i] && freq[i] <= v && i != c1) {
-	v = freq[i];
-	c2 = i;
+        v = freq[i];
+        c2 = i;
       }
     }
 
     /* Done if we've merged everything into one frequency */
     if (c2 < 0)
       break;
-    
+
     /* Else merge the two counts/trees */
     freq[c1] += freq[c2];
     freq[c2] = 0;
@@ -881,9 +946,9 @@
       c1 = others[c1];
       codesize[c1]++;
     }
-    
-    others[c1] = c2;		/* chain c2 onto c1's tree branch */
-    
+
+    others[c1] = c2;            /* chain c2 onto c1's tree branch */
+
     /* Increment the codesize of everything in c2's tree branch */
     codesize[c2]++;
     while (others[c2] >= 0) {
@@ -898,7 +963,7 @@
       /* The JPEG standard seems to think that this can't happen, */
       /* but I'm paranoid... */
       if (codesize[i] > MAX_CLEN)
-	ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
+        ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
 
       bits[codesize[i]]++;
     }
@@ -914,28 +979,28 @@
    * shortest nonzero BITS entry is converted into a prefix for two code words
    * one bit longer.
    */
-  
+
   for (i = MAX_CLEN; i > 16; i--) {
     while (bits[i] > 0) {
-      j = i - 2;		/* find length of new prefix to be used */
+      j = i - 2;                /* find length of new prefix to be used */
       while (bits[j] == 0)
-	j--;
-      
-      bits[i] -= 2;		/* remove two symbols */
-      bits[i-1]++;		/* one goes in this length */
-      bits[j+1] += 2;		/* two new symbols in this length */
-      bits[j]--;		/* symbol of this length is now a prefix */
+        j--;
+
+      bits[i] -= 2;             /* remove two symbols */
+      bits[i-1]++;              /* one goes in this length */
+      bits[j+1] += 2;           /* two new symbols in this length */
+      bits[j]--;                /* symbol of this length is now a prefix */
     }
   }
 
   /* Remove the count for the pseudo-symbol 256 from the largest codelength */
-  while (bits[i] == 0)		/* find largest codelength still in use */
+  while (bits[i] == 0)          /* find largest codelength still in use */
     i--;
   bits[i]--;
-  
+
   /* Return final symbol counts (only for lengths 0..16) */
-  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
-  
+  MEMCOPY(htbl->bits, bits, sizeof(htbl->bits));
+
   /* Return a list of the symbols sorted by code length */
   /* It's not real clear to me why we don't need to consider the codelength
    * changes made above, but the JPEG spec seems to think this works.
@@ -944,8 +1009,8 @@
   for (i = 1; i <= MAX_CLEN; i++) {
     for (j = 0; j <= 255; j++) {
       if (codesize[j] == i) {
-	htbl->huffval[p] = (UINT8) j;
-	p++;
+        htbl->huffval[p] = (UINT8) j;
+        p++;
       }
     }
   }
@@ -964,7 +1029,7 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, dctbl, actbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JHUFF_TBL **htblptr;
   boolean did_dc[NUM_HUFF_TBLS];
   boolean did_ac[NUM_HUFF_TBLS];
@@ -972,8 +1037,8 @@
   /* It's important not to apply jpeg_gen_optimal_table more than once
    * per table, because it clobbers the input frequency counts!
    */
-  MEMZERO(did_dc, SIZEOF(did_dc));
-  MEMZERO(did_ac, SIZEOF(did_ac));
+  MEMZERO(did_dc, sizeof(did_dc));
+  MEMZERO(did_ac, sizeof(did_ac));
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
@@ -982,14 +1047,14 @@
     if (! did_dc[dctbl]) {
       htblptr = & cinfo->dc_huff_tbl_ptrs[dctbl];
       if (*htblptr == NULL)
-	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
       jpeg_gen_optimal_table(cinfo, *htblptr, entropy->dc_count_ptrs[dctbl]);
       did_dc[dctbl] = TRUE;
     }
     if (! did_ac[actbl]) {
       htblptr = & cinfo->ac_huff_tbl_ptrs[actbl];
       if (*htblptr == NULL)
-	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
       jpeg_gen_optimal_table(cinfo, *htblptr, entropy->ac_count_ptrs[actbl]);
       did_ac[actbl] = TRUE;
     }
@@ -1012,7 +1077,7 @@
 
   entropy = (huff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(huff_entropy_encoder));
+                                sizeof(huff_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass_huff;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jchuff.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jchuff.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jchuff.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jchuff.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jchuff.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains declarations for Huffman entropy encoding routines
  * that are shared between the sequential encoder (jchuff.c) and the
@@ -25,23 +28,16 @@
 /* Derived data constructed for each Huffman table */
 
 typedef struct {
-  unsigned int ehufco[256];	/* code for each symbol */
-  char ehufsi[256];		/* length of code for each symbol */
+  unsigned int ehufco[256];     /* code for each symbol */
+  char ehufsi[256];             /* length of code for each symbol */
   /* If no code has been allocated for a symbol S, ehufsi[S] contains 0 */
 } c_derived_tbl;
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_make_c_derived_tbl	jMkCDerived
-#define jpeg_gen_optimal_table	jGenOptTbl
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
 /* Expand a Huffman table definition into the derived format */
 EXTERN(void) jpeg_make_c_derived_tbl
-	JPP((j_compress_ptr cinfo, boolean isDC, int tblno,
-	     c_derived_tbl ** pdtbl));
+        (j_compress_ptr cinfo, boolean isDC, int tblno,
+         c_derived_tbl ** pdtbl);
 
 /* Generate an optimal table definition given the specified counts */
 EXTERN(void) jpeg_gen_optimal_table
-	JPP((j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[]));
+        (j_compress_ptr cinfo, JHUFF_TBL *htbl, long freq[]);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcinit.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcinit.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcinit.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcinit.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains initialization logic for the JPEG compressor.
  * This routine is in charge of selecting the modules to be executed and
@@ -60,7 +61,7 @@
 
   /* Need a full-image coefficient buffer in any multi-pass mode. */
   jinit_c_coef_controller(cinfo,
-		(boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
+                (boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
   jinit_c_main_controller(cinfo, FALSE /* never need full buffer here */);
 
   jinit_marker_writer(cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmainct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmainct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmainct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmainct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcmainct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the main buffer controller for compression.
  * The main buffer lies between the pre-processor and the JPEG
@@ -15,50 +18,30 @@
 #include "jpeglib.h"
 
 
-/* Note: currently, there is no operating mode in which a full-image buffer
- * is needed at this step.  If there were, that mode could not be used with
- * "raw data" input, since this module is bypassed in that case.  However,
- * we've left the code here for possible use in special applications.
- */
-#undef FULL_MAIN_BUFFER_SUPPORTED
-
-
 /* Private buffer controller object */
 
 typedef struct {
   struct jpeg_c_main_controller pub; /* public fields */
 
-  JDIMENSION cur_iMCU_row;	/* number of current iMCU row */
-  JDIMENSION rowgroup_ctr;	/* counts row groups received in iMCU row */
-  boolean suspended;		/* remember if we suspended output */
-  J_BUF_MODE pass_mode;		/* current operating mode */
+  JDIMENSION cur_iMCU_row;      /* number of current iMCU row */
+  JDIMENSION rowgroup_ctr;      /* counts row groups received in iMCU row */
+  boolean suspended;            /* remember if we suspended output */
+  J_BUF_MODE pass_mode;         /* current operating mode */
 
   /* If using just a strip buffer, this points to the entire set of buffers
    * (we allocate one for each component).  In the full-image case, this
    * points to the currently accessible strips of the virtual arrays.
    */
   JSAMPARRAY buffer[MAX_COMPONENTS];
-
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-  /* If using full-image storage, this array holds pointers to virtual-array
-   * control blocks for each component.  Unused if not full-image storage.
-   */
-  jvirt_sarray_ptr whole_image[MAX_COMPONENTS];
-#endif
 } my_main_controller;
 
-typedef my_main_controller * my_main_ptr;
+typedef my_main_controller *my_main_ptr;
 
 
 /* Forward declarations */
 METHODDEF(void) process_data_simple_main
-	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
-	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-METHODDEF(void) process_data_buffer_main
-	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
-	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
-#endif
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+         JDIMENSION in_rows_avail);
 
 
 /*
@@ -74,32 +57,14 @@
   if (cinfo->raw_data_in)
     return;
 
-  main_ptr->cur_iMCU_row = 0;	/* initialize counters */
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  main_ptr->cur_iMCU_row = 0;   /* initialize counters */
   main_ptr->rowgroup_ctr = 0;
   main_ptr->suspended = FALSE;
-  main_ptr->pass_mode = pass_mode;	/* save mode for use by process_data */
-
-  switch (pass_mode) {
-  case JBUF_PASS_THRU:
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    if (main_ptr->whole_image[0] != NULL)
-      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-#endif
-    main_ptr->pub.process_data = process_data_simple_main;
-    break;
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-  case JBUF_SAVE_SOURCE:
-  case JBUF_CRANK_DEST:
-  case JBUF_SAVE_AND_PASS:
-    if (main_ptr->whole_image[0] == NULL)
-      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-    main_ptr->pub.process_data = process_data_buffer_main;
-    break;
-#endif
-  default:
-    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-    break;
-  }
+  main_ptr->pass_mode = pass_mode;      /* save mode for use by process_data */
+  main_ptr->pub.process_data = process_data_simple_main;
 }
 
 
@@ -111,8 +76,8 @@
 
 METHODDEF(void)
 process_data_simple_main (j_compress_ptr cinfo,
-			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			  JDIMENSION in_rows_avail)
+                          JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                          JDIMENSION in_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
 
@@ -120,9 +85,9 @@
     /* Read input data if we haven't filled the main buffer yet */
     if (main_ptr->rowgroup_ctr < DCTSIZE)
       (*cinfo->prep->pre_process_data) (cinfo,
-					input_buf, in_row_ctr, in_rows_avail,
-					main_ptr->buffer, &main_ptr->rowgroup_ctr,
-					(JDIMENSION) DCTSIZE);
+                                        input_buf, in_row_ctr, in_rows_avail,
+                                        main_ptr->buffer, &main_ptr->rowgroup_ctr,
+                                        (JDIMENSION) DCTSIZE);
 
     /* If we don't have a full iMCU row buffered, return to application for
      * more data.  Note that preprocessor will always pad to fill the iMCU row
@@ -140,8 +105,8 @@
        * think we were done.
        */
       if (! main_ptr->suspended) {
-	(*in_row_ctr)--;
-	main_ptr->suspended = TRUE;
+        (*in_row_ctr)--;
+        main_ptr->suspended = TRUE;
       }
       return;
     }
@@ -158,85 +123,6 @@
 }
 
 
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-
-/*
- * Process some data.
- * This routine handles all of the modes that use a full-size buffer.
- */
-
-METHODDEF(void)
-process_data_buffer_main (j_compress_ptr cinfo,
-			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			  JDIMENSION in_rows_avail)
-{
-  my_main_ptr main = (my_main_ptr) cinfo->main;
-  int ci;
-  jpeg_component_info *compptr;
-  boolean writing = (main_ptr->pass_mode != JBUF_CRANK_DEST);
-
-  while (main_ptr->cur_iMCU_row < cinfo->total_iMCU_rows) {
-    /* Realign the virtual buffers if at the start of an iMCU row. */
-    if (main_ptr->rowgroup_ctr == 0) {
-      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	   ci++, compptr++) {
-	main_ptr->buffer[ci] = (*cinfo->mem->access_virt_sarray)
-	  ((j_common_ptr) cinfo, main_ptr->whole_image[ci],
-	   main_ptr->cur_iMCU_row * (compptr->v_samp_factor * DCTSIZE),
-	   (JDIMENSION) (compptr->v_samp_factor * DCTSIZE), writing);
-      }
-      /* In a read pass, pretend we just read some source data. */
-      if (! writing) {
-	*in_row_ctr += cinfo->max_v_samp_factor * DCTSIZE;
-	main_ptr->rowgroup_ctr = DCTSIZE;
-      }
-    }
-
-    /* If a write pass, read input data until the current iMCU row is full. */
-    /* Note: preprocessor will pad if necessary to fill the last iMCU row. */
-    if (writing) {
-      (*cinfo->prep->pre_process_data) (cinfo,
-					input_buf, in_row_ctr, in_rows_avail,
-					main_ptr->buffer, &main_ptr->rowgroup_ctr,
-					(JDIMENSION) DCTSIZE);
-      /* Return to application if we need more data to fill the iMCU row. */
-      if (main_ptr->rowgroup_ctr < DCTSIZE)
-	return;
-    }
-
-    /* Emit data, unless this is a sink-only pass. */
-    if (main_ptr->pass_mode != JBUF_SAVE_SOURCE) {
-      if (! (*cinfo->coef->compress_data) (cinfo, main_ptr->buffer)) {
-	/* If compressor did not consume the whole row, then we must need to
-	 * suspend processing and return to the application.  In this situation
-	 * we pretend we didn't yet consume the last input row; otherwise, if
-	 * it happened to be the last row of the image, the application would
-	 * think we were done.
-	 */
-	if (! main_ptr->suspended) {
-	  (*in_row_ctr)--;
-	  main_ptr->suspended = TRUE;
-	}
-	return;
-      }
-      /* We did finish the row.  Undo our little suspension hack if a previous
-       * call suspended; then mark the main buffer empty.
-       */
-      if (main_ptr->suspended) {
-	(*in_row_ctr)++;
-	main_ptr->suspended = FALSE;
-      }
-    }
-
-    /* If get here, we are done with this iMCU row.  Mark buffer empty. */
-    main_ptr->rowgroup_ctr = 0;
-    main_ptr->cur_iMCU_row++;
-  }
-}
-
-#endif /* FULL_MAIN_BUFFER_SUPPORTED */
-
-
 /*
  * Initialize main buffer controller.
  */
@@ -250,7 +136,7 @@
 
   main_ptr = (my_main_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_main_controller));
+                                sizeof(my_main_controller));
   cinfo->main = (struct jpeg_c_main_controller *) main_ptr;
   main_ptr->pub.start_pass = start_pass_main;
 
@@ -262,32 +148,15 @@
    * may be of a different size.
    */
   if (need_full_buffer) {
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    /* Allocate a full-image virtual array for each component */
-    /* Note we pad the bottom to a multiple of the iMCU height */
-    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
-      main_ptr->whole_image[ci] = (*cinfo->mem->request_virt_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 compptr->width_in_blocks * DCTSIZE,
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor) * DCTSIZE,
-	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
-    }
-#else
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-#endif
   } else {
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    main_ptr->whole_image[0] = NULL; /* flag for no virtual arrays */
-#endif
     /* Allocate a strip buffer for each component */
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 compptr->width_in_blocks * DCTSIZE,
-	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         compptr->width_in_blocks * DCTSIZE,
+         (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmarker.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmarker.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmarker.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmarker.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jcmarker.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains routines to write JPEG datastream markers.
  */
@@ -15,29 +18,29 @@
 #include "jpegcomp.h"
 
 
-typedef enum {			/* JPEG marker codes */
+typedef enum {                  /* JPEG marker codes */
   M_SOF0  = 0xc0,
   M_SOF1  = 0xc1,
   M_SOF2  = 0xc2,
   M_SOF3  = 0xc3,
-  
+
   M_SOF5  = 0xc5,
   M_SOF6  = 0xc6,
   M_SOF7  = 0xc7,
-  
+
   M_JPG   = 0xc8,
   M_SOF9  = 0xc9,
   M_SOF10 = 0xca,
   M_SOF11 = 0xcb,
-  
+
   M_SOF13 = 0xcd,
   M_SOF14 = 0xce,
   M_SOF15 = 0xcf,
-  
+
   M_DHT   = 0xc4,
-  
+
   M_DAC   = 0xcc,
-  
+
   M_RST0  = 0xd0,
   M_RST1  = 0xd1,
   M_RST2  = 0xd2,
@@ -46,7 +49,7 @@
   M_RST5  = 0xd5,
   M_RST6  = 0xd6,
   M_RST7  = 0xd7,
-  
+
   M_SOI   = 0xd8,
   M_EOI   = 0xd9,
   M_SOS   = 0xda,
@@ -55,7 +58,7 @@
   M_DRI   = 0xdd,
   M_DHP   = 0xde,
   M_EXP   = 0xdf,
-  
+
   M_APP0  = 0xe0,
   M_APP1  = 0xe1,
   M_APP2  = 0xe2,
@@ -72,13 +75,13 @@
   M_APP13 = 0xed,
   M_APP14 = 0xee,
   M_APP15 = 0xef,
-  
+
   M_JPG0  = 0xf0,
   M_JPG13 = 0xfd,
   M_COM   = 0xfe,
-  
+
   M_TEM   = 0x01,
-  
+
   M_ERROR = 0x100
 } JPEG_MARKER;
 
@@ -91,7 +94,7 @@
   unsigned int last_restart_interval; /* last DRI value emitted; 0 after SOI */
 } my_marker_writer;
 
-typedef my_marker_writer * my_marker_ptr;
+typedef my_marker_writer *my_marker_ptr;
 
 
 /*
@@ -110,7 +113,7 @@
 emit_byte (j_compress_ptr cinfo, int val)
 /* Emit a byte */
 {
-  struct jpeg_destination_mgr * dest = cinfo->dest;
+  struct jpeg_destination_mgr *dest = cinfo->dest;
 
   *(dest->next_output_byte)++ = (JOCTET) val;
   if (--dest->free_in_buffer == 0) {
@@ -147,7 +150,7 @@
 /* Emit a DQT marker */
 /* Returns the precision used (0 = 8bits, 1 = 16bits) for baseline checking */
 {
-  JQUANT_TBL * qtbl = cinfo->quant_tbl_ptrs[index];
+  JQUANT_TBL *qtbl = cinfo->quant_tbl_ptrs[index];
   int prec;
   int i;
 
@@ -171,7 +174,7 @@
       /* The table entries must be emitted in zigzag order. */
       unsigned int qval = qtbl->quantval[jpeg_natural_order[i]];
       if (prec)
-	emit_byte(cinfo, (int) (qval >> 8));
+        emit_byte(cinfo, (int) (qval >> 8));
       emit_byte(cinfo, (int) (qval & 0xFF));
     }
 
@@ -186,35 +189,35 @@
 emit_dht (j_compress_ptr cinfo, int index, boolean is_ac)
 /* Emit a DHT marker */
 {
-  JHUFF_TBL * htbl;
+  JHUFF_TBL *htbl;
   int length, i;
-  
+
   if (is_ac) {
     htbl = cinfo->ac_huff_tbl_ptrs[index];
-    index += 0x10;		/* output index has AC bit set */
+    index += 0x10;              /* output index has AC bit set */
   } else {
     htbl = cinfo->dc_huff_tbl_ptrs[index];
   }
 
   if (htbl == NULL)
     ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, index);
-  
+
   if (! htbl->sent_table) {
     emit_marker(cinfo, M_DHT);
-    
+
     length = 0;
     for (i = 1; i <= 16; i++)
       length += htbl->bits[i];
-    
+
     emit_2bytes(cinfo, length + 2 + 1 + 16);
     emit_byte(cinfo, index);
-    
+
     for (i = 1; i <= 16; i++)
       emit_byte(cinfo, htbl->bits[i]);
-    
+
     for (i = 0; i < length; i++)
       emit_byte(cinfo, htbl->huffval[i]);
-    
+
     htbl->sent_table = TRUE;
   }
 }
@@ -231,32 +234,38 @@
   char ac_in_use[NUM_ARITH_TBLS];
   int length, i;
   jpeg_component_info *compptr;
-  
+
   for (i = 0; i < NUM_ARITH_TBLS; i++)
     dc_in_use[i] = ac_in_use[i] = 0;
-  
+
   for (i = 0; i < cinfo->comps_in_scan; i++) {
     compptr = cinfo->cur_comp_info[i];
-    dc_in_use[compptr->dc_tbl_no] = 1;
-    ac_in_use[compptr->ac_tbl_no] = 1;
+    /* DC needs no table for refinement scan */
+    if (cinfo->Ss == 0 && cinfo->Ah == 0)
+      dc_in_use[compptr->dc_tbl_no] = 1;
+    /* AC needs no table when not present */
+    if (cinfo->Se)
+      ac_in_use[compptr->ac_tbl_no] = 1;
   }
-  
+
   length = 0;
   for (i = 0; i < NUM_ARITH_TBLS; i++)
     length += dc_in_use[i] + ac_in_use[i];
-  
-  emit_marker(cinfo, M_DAC);
-  
-  emit_2bytes(cinfo, length*2 + 2);
-  
-  for (i = 0; i < NUM_ARITH_TBLS; i++) {
-    if (dc_in_use[i]) {
-      emit_byte(cinfo, i);
-      emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
-    }
-    if (ac_in_use[i]) {
-      emit_byte(cinfo, i + 0x10);
-      emit_byte(cinfo, cinfo->arith_ac_K[i]);
+
+  if (length) {
+    emit_marker(cinfo, M_DAC);
+
+    emit_2bytes(cinfo, length*2 + 2);
+
+    for (i = 0; i < NUM_ARITH_TBLS; i++) {
+      if (dc_in_use[i]) {
+        emit_byte(cinfo, i);
+        emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
+      }
+      if (ac_in_use[i]) {
+        emit_byte(cinfo, i + 0x10);
+        emit_byte(cinfo, cinfo->arith_ac_K[i]);
+      }
     }
   }
 #endif /* C_ARITH_CODING_SUPPORTED */
@@ -268,8 +277,8 @@
 /* Emit a DRI marker */
 {
   emit_marker(cinfo, M_DRI);
-  
-  emit_2bytes(cinfo, 4);	/* fixed length */
+
+  emit_2bytes(cinfo, 4);        /* fixed length */
 
   emit_2bytes(cinfo, (int) cinfo->restart_interval);
 }
@@ -281,9 +290,9 @@
 {
   int ci;
   jpeg_component_info *compptr;
-  
+
   emit_marker(cinfo, code);
-  
+
   emit_2bytes(cinfo, 3 * cinfo->num_components + 2 + 5 + 1); /* length */
 
   /* Make sure image isn't bigger than SOF field can handle */
@@ -312,32 +321,26 @@
 {
   int i, td, ta;
   jpeg_component_info *compptr;
-  
+
   emit_marker(cinfo, M_SOS);
-  
+
   emit_2bytes(cinfo, 2 * cinfo->comps_in_scan + 2 + 1 + 3); /* length */
-  
+
   emit_byte(cinfo, cinfo->comps_in_scan);
-  
+
   for (i = 0; i < cinfo->comps_in_scan; i++) {
     compptr = cinfo->cur_comp_info[i];
     emit_byte(cinfo, compptr->component_id);
-    td = compptr->dc_tbl_no;
-    ta = compptr->ac_tbl_no;
-    if (cinfo->progressive_mode) {
-      /* Progressive mode: only DC or only AC tables are used in one scan;
-       * furthermore, Huffman coding of DC refinement uses no table at all.
-       * We emit 0 for unused field(s); this is recommended by the P&M text
-       * but does not seem to be specified in the standard.
-       */
-      if (cinfo->Ss == 0) {
-	ta = 0;			/* DC scan */
-	if (cinfo->Ah != 0 && !cinfo->arith_code)
-	  td = 0;		/* no DC table either */
-      } else {
-	td = 0;			/* AC scan */
-      }
-    }
+
+    /* We emit 0 for unused field(s); this is recommended by the P&M text
+     * but does not seem to be specified in the standard.
+     */
+
+    /* DC needs no table for refinement scan */
+    td = cinfo->Ss == 0 && cinfo->Ah == 0 ? compptr->dc_tbl_no : 0;
+    /* AC needs no table when not present */
+    ta = cinfo->Se ? compptr->ac_tbl_no : 0;
+
     emit_byte(cinfo, (td << 4) + ta);
   }
 
@@ -352,22 +355,22 @@
 /* Emit a JFIF-compliant APP0 marker */
 {
   /*
-   * Length of APP0 block	(2 bytes)
-   * Block ID			(4 bytes - ASCII "JFIF")
-   * Zero byte			(1 byte to terminate the ID string)
-   * Version Major, Minor	(2 bytes - major first)
-   * Units			(1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
-   * Xdpu			(2 bytes - dots per unit horizontal)
-   * Ydpu			(2 bytes - dots per unit vertical)
-   * Thumbnail X size		(1 byte)
-   * Thumbnail Y size		(1 byte)
+   * Length of APP0 block       (2 bytes)
+   * Block ID                   (4 bytes - ASCII "JFIF")
+   * Zero byte                  (1 byte to terminate the ID string)
+   * Version Major, Minor       (2 bytes - major first)
+   * Units                      (1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
+   * Xdpu                       (2 bytes - dots per unit horizontal)
+   * Ydpu                       (2 bytes - dots per unit vertical)
+   * Thumbnail X size           (1 byte)
+   * Thumbnail Y size           (1 byte)
    */
-  
+
   emit_marker(cinfo, M_APP0);
-  
+
   emit_2bytes(cinfo, 2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); /* length */
 
-  emit_byte(cinfo, 0x4A);	/* Identifier: ASCII "JFIF" */
+  emit_byte(cinfo, 0x4A);       /* Identifier: ASCII "JFIF" */
   emit_byte(cinfo, 0x46);
   emit_byte(cinfo, 0x49);
   emit_byte(cinfo, 0x46);
@@ -377,7 +380,7 @@
   emit_byte(cinfo, cinfo->density_unit); /* Pixel size information */
   emit_2bytes(cinfo, (int) cinfo->X_density);
   emit_2bytes(cinfo, (int) cinfo->Y_density);
-  emit_byte(cinfo, 0);		/* No thumbnail image */
+  emit_byte(cinfo, 0);          /* No thumbnail image */
   emit_byte(cinfo, 0);
 }
 
@@ -387,12 +390,12 @@
 /* Emit an Adobe APP14 marker */
 {
   /*
-   * Length of APP14 block	(2 bytes)
-   * Block ID			(5 bytes - ASCII "Adobe")
-   * Version Number		(2 bytes - currently 100)
-   * Flags0			(2 bytes - currently 0)
-   * Flags1			(2 bytes - currently 0)
-   * Color transform		(1 byte)
+   * Length of APP14 block      (2 bytes)
+   * Block ID                   (5 bytes - ASCII "Adobe")
+   * Version Number             (2 bytes - currently 100)
+   * Flags0                     (2 bytes - currently 0)
+   * Flags1                     (2 bytes - currently 0)
+   * Color transform            (1 byte)
    *
    * Although Adobe TN 5116 mentions Version = 101, all the Adobe files
    * now in circulation seem to use Version = 100, so that's what we write.
@@ -401,28 +404,28 @@
    * YCbCr, 2 if it's YCCK, 0 otherwise.  Adobe's definition has to do with
    * whether the encoder performed a transformation, which is pretty useless.
    */
-  
+
   emit_marker(cinfo, M_APP14);
-  
+
   emit_2bytes(cinfo, 2 + 5 + 2 + 2 + 2 + 1); /* length */
 
-  emit_byte(cinfo, 0x41);	/* Identifier: ASCII "Adobe" */
+  emit_byte(cinfo, 0x41);       /* Identifier: ASCII "Adobe" */
   emit_byte(cinfo, 0x64);
   emit_byte(cinfo, 0x6F);
   emit_byte(cinfo, 0x62);
   emit_byte(cinfo, 0x65);
-  emit_2bytes(cinfo, 100);	/* Version */
-  emit_2bytes(cinfo, 0);	/* Flags0 */
-  emit_2bytes(cinfo, 0);	/* Flags1 */
+  emit_2bytes(cinfo, 100);      /* Version */
+  emit_2bytes(cinfo, 0);        /* Flags0 */
+  emit_2bytes(cinfo, 0);        /* Flags1 */
   switch (cinfo->jpeg_color_space) {
   case JCS_YCbCr:
-    emit_byte(cinfo, 1);	/* Color transform = 1 */
+    emit_byte(cinfo, 1);        /* Color transform = 1 */
     break;
   case JCS_YCCK:
-    emit_byte(cinfo, 2);	/* Color transform = 2 */
+    emit_byte(cinfo, 2);        /* Color transform = 2 */
     break;
   default:
-    emit_byte(cinfo, 0);	/* Color transform = 0 */
+    emit_byte(cinfo, 0);        /* Color transform = 0 */
     break;
   }
 }
@@ -440,12 +443,12 @@
 write_marker_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
 /* Emit an arbitrary marker header */
 {
-  if (datalen > (unsigned int) 65533)		/* safety check */
+  if (datalen > (unsigned int) 65533)           /* safety check */
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
   emit_marker(cinfo, (JPEG_MARKER) marker);
 
-  emit_2bytes(cinfo, (int) (datalen + 2));	/* total length */
+  emit_2bytes(cinfo, (int) (datalen + 2));      /* total length */
 }
 
 METHODDEF(void)
@@ -472,12 +475,12 @@
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
-  emit_marker(cinfo, M_SOI);	/* first the SOI */
+  emit_marker(cinfo, M_SOI);    /* first the SOI */
 
   /* SOI is defined to reset restart interval to 0 */
   marker->last_restart_interval = 0;
 
-  if (cinfo->write_JFIF_header)	/* next an optional JFIF APP0 */
+  if (cinfo->write_JFIF_header) /* next an optional JFIF APP0 */
     emit_jfif_app0(cinfo);
   if (cinfo->write_Adobe_marker) /* next an optional Adobe APP14 */
     emit_adobe_app14(cinfo);
@@ -498,7 +501,7 @@
   int ci, prec;
   boolean is_baseline;
   jpeg_component_info *compptr;
-  
+
   /* Emit DQT for each quantization table.
    * Note that emit_dqt() suppresses any duplicate tables.
    */
@@ -518,9 +521,9 @@
   } else {
     is_baseline = TRUE;
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       if (compptr->dc_tbl_no > 1 || compptr->ac_tbl_no > 1)
-	is_baseline = FALSE;
+        is_baseline = FALSE;
     }
     if (prec && is_baseline) {
       is_baseline = FALSE;
@@ -531,14 +534,17 @@
 
   /* Emit the proper SOF marker */
   if (cinfo->arith_code) {
-    emit_sof(cinfo, M_SOF9);	/* SOF code for arithmetic coding */
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF10); /* SOF code for progressive arithmetic */
+    else
+      emit_sof(cinfo, M_SOF9);  /* SOF code for sequential arithmetic */
   } else {
     if (cinfo->progressive_mode)
-      emit_sof(cinfo, M_SOF2);	/* SOF code for progressive Huffman */
+      emit_sof(cinfo, M_SOF2);  /* SOF code for progressive Huffman */
     else if (is_baseline)
-      emit_sof(cinfo, M_SOF0);	/* SOF code for baseline implementation */
+      emit_sof(cinfo, M_SOF0);  /* SOF code for baseline implementation */
     else
-      emit_sof(cinfo, M_SOF1);	/* SOF code for non-baseline Huffman file */
+      emit_sof(cinfo, M_SOF1);  /* SOF code for non-baseline Huffman file */
   }
 }
 
@@ -568,19 +574,12 @@
      */
     for (i = 0; i < cinfo->comps_in_scan; i++) {
       compptr = cinfo->cur_comp_info[i];
-      if (cinfo->progressive_mode) {
-	/* Progressive mode: only DC or only AC tables are used in one scan */
-	if (cinfo->Ss == 0) {
-	  if (cinfo->Ah == 0)	/* DC needs no table for refinement scan */
-	    emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
-	} else {
-	  emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
-	}
-      } else {
-	/* Sequential mode: need both DC and AC tables */
-	emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
-	emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
-      }
+      /* DC needs no table for refinement scan */
+      if (cinfo->Ss == 0 && cinfo->Ah == 0)
+        emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+      /* AC needs no table when not present */
+      if (cinfo->Se)
+        emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
     }
   }
 
@@ -629,9 +628,9 @@
   if (! cinfo->arith_code) {
     for (i = 0; i < NUM_HUFF_TBLS; i++) {
       if (cinfo->dc_huff_tbl_ptrs[i] != NULL)
-	emit_dht(cinfo, i, FALSE);
+        emit_dht(cinfo, i, FALSE);
       if (cinfo->ac_huff_tbl_ptrs[i] != NULL)
-	emit_dht(cinfo, i, TRUE);
+        emit_dht(cinfo, i, TRUE);
     }
   }
 
@@ -651,7 +650,7 @@
   /* Create the subobject */
   marker = (my_marker_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_marker_writer));
+                                sizeof(my_marker_writer));
   cinfo->marker = (struct jpeg_marker_writer *) marker;
   /* Initialize method pointers */
   marker->pub.write_file_header = write_file_header;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmaster.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmaster.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcmaster.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcmaster.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,17 @@
 /*
  * jcmaster.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains master control logic for the JPEG compressor.
  * These routines are concerned with parameter validation, initial setup,
- * and inter-pass control (determining the number of passes and the work 
+ * and inter-pass control (determining the number of passes and the work
  * to be done in each pass).
  */
 
@@ -22,23 +24,23 @@
 /* Private state */
 
 typedef enum {
-	main_pass,		/* input data, also do first output step */
-	huff_opt_pass,		/* Huffman code optimization pass */
-	output_pass		/* data output pass */
+        main_pass,              /* input data, also do first output step */
+        huff_opt_pass,          /* Huffman code optimization pass */
+        output_pass             /* data output pass */
 } c_pass_type;
 
 typedef struct {
-  struct jpeg_comp_master pub;	/* public fields */
+  struct jpeg_comp_master pub;  /* public fields */
 
-  c_pass_type pass_type;	/* the type of the current pass */
+  c_pass_type pass_type;        /* the type of the current pass */
 
-  int pass_number;		/* # of passes completed */
-  int total_passes;		/* total # of passes needed */
+  int pass_number;              /* # of passes completed */
+  int total_passes;             /* total # of passes needed */
 
-  int scan_number;		/* current index in scan_info[] */
+  int scan_number;              /* current index in scan_info[] */
 } my_comp_master;
 
-typedef my_comp_master * my_master_ptr;
+typedef my_comp_master *my_master_ptr;
 
 
 /*
@@ -104,7 +106,7 @@
   /* Check that number of components won't exceed internal array sizes */
   if (cinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
 
   /* Compute maximum sampling factors; check factor validity */
   cinfo->max_h_samp_factor = 1;
@@ -112,12 +114,12 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
-	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+        compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
       ERREXIT(cinfo, JERR_BAD_SAMPLING);
     cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
-				   compptr->h_samp_factor);
+                                   compptr->h_samp_factor);
     cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
-				   compptr->v_samp_factor);
+                                   compptr->v_samp_factor);
   }
 
   /* Compute dimensions of components */
@@ -134,17 +136,17 @@
     /* Size in DCT blocks */
     compptr->width_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width * (long) compptr->h_samp_factor,
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->height_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height * (long) compptr->v_samp_factor,
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
     /* Size in samples */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width * (long) compptr->h_samp_factor,
-		    (long) cinfo->max_h_samp_factor);
+                    (long) cinfo->max_h_samp_factor);
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height * (long) compptr->v_samp_factor,
-		    (long) cinfo->max_v_samp_factor);
+                    (long) cinfo->max_v_samp_factor);
     /* Mark component needed (this flag isn't actually used for compression) */
     compptr->component_needed = TRUE;
   }
@@ -154,7 +156,7 @@
    */
   cinfo->total_iMCU_rows = (JDIMENSION)
     jdiv_round_up((long) cinfo->_jpeg_height,
-		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+                  (long) (cinfo->max_v_samp_factor*DCTSIZE));
 }
 
 
@@ -166,12 +168,12 @@
  * determine whether it uses progressive JPEG, and set cinfo->progressive_mode.
  */
 {
-  const jpeg_scan_info * scanptr;
+  const jpeg_scan_info *scanptr;
   int scanno, ncomps, ci, coefi, thisi;
   int Ss, Se, Ah, Al;
   boolean component_sent[MAX_COMPONENTS];
 #ifdef C_PROGRESSIVE_SUPPORTED
-  int * last_bitpos_ptr;
+  int *last_bitpos_ptr;
   int last_bitpos[MAX_COMPONENTS][DCTSIZE2];
   /* -1 until that coefficient has been seen; then last Al for it */
 #endif
@@ -187,15 +189,15 @@
 #ifdef C_PROGRESSIVE_SUPPORTED
     cinfo->progressive_mode = TRUE;
     last_bitpos_ptr = & last_bitpos[0][0];
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       for (coefi = 0; coefi < DCTSIZE2; coefi++)
-	*last_bitpos_ptr++ = -1;
+        *last_bitpos_ptr++ = -1;
 #else
     ERREXIT(cinfo, JERR_NOT_COMPILED);
 #endif
   } else {
     cinfo->progressive_mode = FALSE;
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       component_sent[ci] = FALSE;
   }
 
@@ -207,10 +209,10 @@
     for (ci = 0; ci < ncomps; ci++) {
       thisi = scanptr->component_index[ci];
       if (thisi < 0 || thisi >= cinfo->num_components)
-	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
       /* Components must appear in SOF order within each scan */
       if (ci > 0 && thisi <= scanptr->component_index[ci-1])
-	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
     }
     /* Validate progression parameters */
     Ss = scanptr->Ss;
@@ -232,43 +234,43 @@
 #define MAX_AH_AL 13
 #endif
       if (Ss < 0 || Ss >= DCTSIZE2 || Se < Ss || Se >= DCTSIZE2 ||
-	  Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
-	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       if (Ss == 0) {
-	if (Se != 0)		/* DC and AC together not OK */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        if (Se != 0)            /* DC and AC together not OK */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       } else {
-	if (ncomps != 1)	/* AC scans must be for only one component */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        if (ncomps != 1)        /* AC scans must be for only one component */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       }
       for (ci = 0; ci < ncomps; ci++) {
-	last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
-	if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	for (coefi = Ss; coefi <= Se; coefi++) {
-	  if (last_bitpos_ptr[coefi] < 0) {
-	    /* first scan of this coefficient */
-	    if (Ah != 0)
-	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	  } else {
-	    /* not first scan */
-	    if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
-	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	  }
-	  last_bitpos_ptr[coefi] = Al;
-	}
+        last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
+        if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        for (coefi = Ss; coefi <= Se; coefi++) {
+          if (last_bitpos_ptr[coefi] < 0) {
+            /* first scan of this coefficient */
+            if (Ah != 0)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          } else {
+            /* not first scan */
+            if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          }
+          last_bitpos_ptr[coefi] = Al;
+        }
       }
 #endif
     } else {
       /* For sequential JPEG, all progression parameters must be these: */
       if (Ss != 0 || Se != DCTSIZE2-1 || Ah != 0 || Al != 0)
-	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       /* Make sure components are not sent twice */
       for (ci = 0; ci < ncomps; ci++) {
-	thisi = scanptr->component_index[ci];
-	if (component_sent[thisi])
-	  ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
-	component_sent[thisi] = TRUE;
+        thisi = scanptr->component_index[ci];
+        if (component_sent[thisi])
+          ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        component_sent[thisi] = TRUE;
       }
     }
   }
@@ -283,13 +285,13 @@
      */
     for (ci = 0; ci < cinfo->num_components; ci++) {
       if (last_bitpos[ci][0] < 0)
-	ERREXIT(cinfo, JERR_MISSING_DATA);
+        ERREXIT(cinfo, JERR_MISSING_DATA);
     }
 #endif
   } else {
     for (ci = 0; ci < cinfo->num_components; ci++) {
       if (! component_sent[ci])
-	ERREXIT(cinfo, JERR_MISSING_DATA);
+        ERREXIT(cinfo, JERR_MISSING_DATA);
     }
   }
 }
@@ -307,12 +309,12 @@
   if (cinfo->scan_info != NULL) {
     /* Prepare for current scan --- the script is already validated */
     my_master_ptr master = (my_master_ptr) cinfo->master;
-    const jpeg_scan_info * scanptr = cinfo->scan_info + master->scan_number;
+    const jpeg_scan_info *scanptr = cinfo->scan_info + master->scan_number;
 
     cinfo->comps_in_scan = scanptr->comps_in_scan;
     for (ci = 0; ci < scanptr->comps_in_scan; ci++) {
       cinfo->cur_comp_info[ci] =
-	&cinfo->comp_info[scanptr->component_index[ci]];
+        &cinfo->comp_info[scanptr->component_index[ci]];
     }
     cinfo->Ss = scanptr->Ss;
     cinfo->Se = scanptr->Se;
@@ -325,7 +327,7 @@
     /* Prepare for single sequential-JPEG scan containing all components */
     if (cinfo->num_components > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	       MAX_COMPS_IN_SCAN);
+               MAX_COMPS_IN_SCAN);
     cinfo->comps_in_scan = cinfo->num_components;
     for (ci = 0; ci < cinfo->num_components; ci++) {
       cinfo->cur_comp_info[ci] = &cinfo->comp_info[ci];
@@ -345,16 +347,16 @@
 {
   int ci, mcublks, tmp;
   jpeg_component_info *compptr;
-  
+
   if (cinfo->comps_in_scan == 1) {
-    
+
     /* Noninterleaved (single-component) scan */
     compptr = cinfo->cur_comp_info[0];
-    
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = compptr->width_in_blocks;
     cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
-    
+
     /* For noninterleaved scan, always one block per MCU */
     compptr->MCU_width = 1;
     compptr->MCU_height = 1;
@@ -367,28 +369,28 @@
     tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
     if (tmp == 0) tmp = compptr->v_samp_factor;
     compptr->last_row_height = tmp;
-    
+
     /* Prepare array describing MCU composition */
     cinfo->blocks_in_MCU = 1;
     cinfo->MCU_membership[0] = 0;
-    
+
   } else {
-    
+
     /* Interleaved (multi-component) scan */
     if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
-	       MAX_COMPS_IN_SCAN);
-    
+               MAX_COMPS_IN_SCAN);
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width,
-		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor*DCTSIZE));
     cinfo->MCU_rows_in_scan = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height,
-		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
-    
+                    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
     cinfo->blocks_in_MCU = 0;
-    
+
     for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
       compptr = cinfo->cur_comp_info[ci];
       /* Sampling factors give # of blocks of component in each MCU */
@@ -406,12 +408,12 @@
       /* Prepare array describing MCU composition */
       mcublks = compptr->MCU_blocks;
       if (cinfo->blocks_in_MCU + mcublks > C_MAX_BLOCKS_IN_MCU)
-	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
       while (mcublks-- > 0) {
-	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+        cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
       }
     }
-    
+
   }
 
   /* Convert restart specified in rows to actual MCU count. */
@@ -451,8 +453,8 @@
     (*cinfo->fdct->start_pass) (cinfo);
     (*cinfo->entropy->start_pass) (cinfo, cinfo->optimize_coding);
     (*cinfo->coef->start_pass) (cinfo,
-				(master->total_passes > 1 ?
-				 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+                                (master->total_passes > 1 ?
+                                 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
     (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
     if (cinfo->optimize_coding) {
       /* No immediate data output; postpone writing frame/scan headers */
@@ -580,7 +582,7 @@
 
   master = (my_master_ptr)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(my_comp_master));
+                                  sizeof(my_comp_master));
   cinfo->master = (struct jpeg_comp_master *) master;
   master->pub.prepare_for_pass = prepare_for_pass;
   master->pub.pass_startup = pass_startup;
@@ -601,7 +603,7 @@
     cinfo->num_scans = 1;
   }
 
-  if (cinfo->progressive_mode)	/*  TEMPORARY HACK ??? */
+  if (cinfo->progressive_mode && !cinfo->arith_code)  /*  TEMPORARY HACK ??? */
     cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
 
   /* Initialize my private state */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcomapi.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcomapi.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcomapi.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcomapi.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcomapi.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface routines that are used for both
  * compression and decompression.
@@ -72,8 +75,8 @@
   /* NB: mem pointer is NULL if memory mgr failed to initialize. */
   if (cinfo->mem != NULL)
     (*cinfo->mem->self_destruct) (cinfo);
-  cinfo->mem = NULL;		/* be safe if jpeg_destroy is called twice */
-  cinfo->global_state = 0;	/* mark it destroyed */
+  cinfo->mem = NULL;            /* be safe if jpeg_destroy is called twice */
+  cinfo->global_state = 0;      /* mark it destroyed */
 }
 
 
@@ -88,8 +91,8 @@
   JQUANT_TBL *tbl;
 
   tbl = (JQUANT_TBL *)
-    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JQUANT_TBL));
-  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JQUANT_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
   return tbl;
 }
 
@@ -100,7 +103,7 @@
   JHUFF_TBL *tbl;
 
   tbl = (JHUFF_TBL *)
-    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JHUFF_TBL));
-  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JHUFF_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
   return tbl;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.h	2021-04-25 01:47:22.000000000 +0800
@@ -5,16 +5,27 @@
 #define JPEG_LIB_VERSION 62
 
 /* libjpeg-turbo version */
-#define LIBJPEG_TURBO_VERSION 1.2.0
+#define LIBJPEG_TURBO_VERSION 1.4.2
 
 /* Support arithmetic encoding */
-#define C_ARITH_CODING_SUPPORTED 1
+/* #define C_ARITH_CODING_SUPPORTED 1 */
 
 /* Support arithmetic decoding */
-#define D_ARITH_CODING_SUPPORTED 1
+/* #define D_ARITH_CODING_SUPPORTED 1 */
 
-/* Compiler supports function prototypes. */
-#define HAVE_PROTOTYPES 1
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8      /* use 8 or 12 */
+
+/* Define to 1 if you have the <locale.h> header file. */
+#define HAVE_LOCALE_H 1
 
 /* Define to 1 if you have the <stddef.h> header file. */
 #define HAVE_STDDEF_H 1
@@ -22,25 +33,27 @@
 /* Define to 1 if you have the <stdlib.h> header file. */
 #define HAVE_STDLIB_H 1
 
-/* Compiler supports 'unsigned char'. */
+/* Define to 1 if the system has the type `unsigned char'. */
 #define HAVE_UNSIGNED_CHAR 1
 
-/* Compiler supports 'unsigned short'. */
+/* Define to 1 if the system has the type `unsigned short'. */
 #define HAVE_UNSIGNED_SHORT 1
 
-/* Compiler does not support pointers to unspecified structures. */
+/* Compiler does not support pointers to undefined structures. */
 /* #undef INCOMPLETE_TYPES_BROKEN */
 
-/* Compiler has <strings.h> rather than standard <string.h>. */
-/* #undef NEED_BSD_STRINGS */
+/* Support in-memory source/destination managers */
+/* #undef MEM_SRCDST_SUPPORTED */
 
-/* Linker requires that global names be unique in first 15 characters. */
-/* #undef NEED_SHORT_EXTERNAL_NAMES */
+/* Define if you have BSD-like bzero and bcopy in <strings.h> rather than
+   memset/memcpy in <string.h>. */
+/* #undef NEED_BSD_STRINGS */
 
-/* Need to include <sys/types.h> in order to obtain size_t. */
-#define NEED_SYS_TYPES_H 1
+/* Define if you need to include <sys/types.h> to get size_t. */
+/* #undef NEED_SYS_TYPES_H 1 */
 
-/* Broken compiler shifts signed values as an unsigned shift. */
+/* Define if your (broken) compiler shifts signed values as if they were
+   unsigned. */
 /* #undef RIGHT_SHIFT_IS_UNSIGNED */
 
 /* Use accelerated SIMD routines. */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.h.in glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.h.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.h.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.h.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,70 @@
+/* Version ID for the JPEG library.
+ * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60".
+ */
+#define JPEG_LIB_VERSION  62	/* Version 6b */
+
+/* libjpeg-turbo version */
+#define LIBJPEG_TURBO_VERSION 0
+
+/* Support arithmetic encoding */
+#undef C_ARITH_CODING_SUPPORTED
+
+/* Support arithmetic decoding */
+#undef D_ARITH_CODING_SUPPORTED
+
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8      /* use 8 or 12 */
+
+/* Define to 1 if you have the <locale.h> header file. */
+#undef HAVE_LOCALE_H
+
+/* Define to 1 if you have the <stddef.h> header file. */
+#undef HAVE_STDDEF_H
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#undef HAVE_STDLIB_H
+
+/* Define to 1 if the system has the type `unsigned char'. */
+#undef HAVE_UNSIGNED_CHAR
+
+/* Define to 1 if the system has the type `unsigned short'. */
+#undef HAVE_UNSIGNED_SHORT
+
+/* Compiler does not support pointers to undefined structures. */
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Support in-memory source/destination managers */
+#undef MEM_SRCDST_SUPPORTED
+
+/* Define if you have BSD-like bzero and bcopy in <strings.h> rather than
+   memset/memcpy in <string.h>. */
+#undef NEED_BSD_STRINGS
+
+/* Define if you need to include <sys/types.h> to get size_t. */
+#undef NEED_SYS_TYPES_H
+
+/* Define if your (broken) compiler shifts signed values as if they were
+   unsigned. */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+/* Use accelerated SIMD routines. */
+#undef WITH_SIMD
+
+/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
+#ifndef __CHAR_UNSIGNED__
+# undef __CHAR_UNSIGNED__
+#endif
+
+/* Define to empty if `const' does not conform to ANSI C. */
+#undef const
+
+/* Define to `unsigned int' if <sys/types.h> does not define. */
+#undef size_t
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfigint.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfigint.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfigint.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfigint.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,27 @@
+/* jconfigint.h.  Generated from jconfigint.h.in by configure.  */
+/* libjpeg-turbo build number */
+#define BUILD ""
+
+/* How to obtain function inlining. */
+#ifndef INLINE
+#if defined(__GNUC__)
+#define INLINE inline __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#define INLINE __forceinline
+#else
+#define INLINE
+#endif
+#endif
+
+/* Define to the full name of this package. */
+#define PACKAGE_NAME "libjpeg-turbo"
+
+/* Version number of package */
+#define VERSION "1.4.90"
+
+/* The size of `size_t', as computed by sizeof. */
+#if __WORDSIZE==64 || defined(_WIN64)
+#define SIZEOF_SIZE_T 8
+#else
+#define SIZEOF_SIZE_T 4
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfigint.h.in glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfigint.h.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfigint.h.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfigint.h.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,14 @@
+/* libjpeg-turbo build number */
+#undef BUILD
+
+/* How to obtain function inlining. */
+#undef INLINE
+
+/* Define to the full name of this package. */
+#undef PACKAGE_NAME
+
+/* Version number of package */
+#undef VERSION
+
+/* The size of `size_t', as computed by sizeof. */
+#undef SIZEOF_SIZE_T
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jconfig.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jconfig.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,143 @@
+/*
+ * jconfig.txt
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if an ordinary "char" type is unsigned.
+ * If you're not sure, leaving it undefined will work at some cost in speed.
+ * If you defined HAVE_UNSIGNED_CHAR then the speed difference is minimal.
+ */
+#undef __CHAR_UNSIGNED__
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Define "boolean" as unsigned char, not int, on Windows systems.
+ */
+#ifdef _WIN32
+#ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#endif
+#define HAVE_BOOLEAN            /* prevent jmorecfg.h from redefining it */
+#endif
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED           /* BMP image file format */
+#define GIF_SUPPORTED           /* GIF image file format */
+#define PPM_SUPPORTED           /* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED            /* Utah RLE image file format */
+#define TARGA_SUPPORTED         /* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* By default, we open image files with fopen(...,"rb") or fopen(...,"wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcparam.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcparam.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcparam.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcparam.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,13 @@
 /*
  * jcparam.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
  * Modified 2003-2008 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains optional default-setting code for the JPEG compressor.
  * Applications do not have to use this file, but those that don't use it
@@ -15,6 +17,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
+#include "jstdhuff.c"
 
 
 /*
@@ -23,15 +26,15 @@
 
 GLOBAL(void)
 jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
-		      const unsigned int *basic_table,
-		      int scale_factor, boolean force_baseline)
+                      const unsigned int *basic_table,
+                      int scale_factor, boolean force_baseline)
 /* Define a quantization table equal to the basic_table times
  * a scale factor (given as a percentage).
  * If force_baseline is TRUE, the computed quantization table entries
  * are limited to 1..255 for JPEG baseline compatibility.
  */
 {
-  JQUANT_TBL ** qtblptr;
+  JQUANT_TBL **qtblptr;
   int i;
   long temp;
 
@@ -53,7 +56,7 @@
     if (temp <= 0L) temp = 1L;
     if (temp > 32767L) temp = 32767L; /* max quantizer needed for 12 bits */
     if (force_baseline && temp > 255L)
-      temp = 255L;		/* limit to baseline range if requested */
+      temp = 255L;              /* limit to baseline range if requested */
     (*qtblptr)->quantval[i] = (UINT16) temp;
   }
 
@@ -98,16 +101,16 @@
 {
   /* Set up two quantization tables using the specified scaling */
   jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
-		       cinfo->q_scale_factor[0], force_baseline);
+                       cinfo->q_scale_factor[0], force_baseline);
   jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
-		       cinfo->q_scale_factor[1], force_baseline);
+                       cinfo->q_scale_factor[1], force_baseline);
 }
 #endif
 
 
 GLOBAL(void)
 jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
-			 boolean force_baseline)
+                         boolean force_baseline)
 /* Set or change the 'quality' (quantization) setting, using default tables
  * and a straight percentage-scaling quality scale.  In most cases it's better
  * to use jpeg_set_quality (below); this entry point is provided for
@@ -116,9 +119,9 @@
 {
   /* Set up two quantization tables using the specified scaling */
   jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
-		       scale_factor, force_baseline);
+                       scale_factor, force_baseline);
   jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
-		       scale_factor, force_baseline);
+                       scale_factor, force_baseline);
 }
 
 
@@ -165,116 +168,6 @@
 
 
 /*
- * Huffman table setup routines
- */
-
-LOCAL(void)
-add_huff_table (j_compress_ptr cinfo,
-		JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
-/* Define a Huffman table */
-{
-  int nsymbols, len;
-
-  if (*htblptr == NULL)
-    *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
-
-  /* Copy the number-of-symbols-of-each-code-length counts */
-  MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
-
-  /* Validate the counts.  We do this here mainly so we can copy the right
-   * number of symbols from the val[] array, without risking marching off
-   * the end of memory.  jchuff.c will do a more thorough test later.
-   */
-  nsymbols = 0;
-  for (len = 1; len <= 16; len++)
-    nsymbols += bits[len];
-  if (nsymbols < 1 || nsymbols > 256)
-    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
-
-  MEMCOPY((*htblptr)->huffval, val, nsymbols * SIZEOF(UINT8));
-
-  /* Initialize sent_table FALSE so table will be written to JPEG file. */
-  (*htblptr)->sent_table = FALSE;
-}
-
-
-LOCAL(void)
-std_huff_tables (j_compress_ptr cinfo)
-/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
-/* IMPORTANT: these are only valid for 8-bit data precision! */
-{
-  static const UINT8 bits_dc_luminance[17] =
-    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_luminance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-  
-  static const UINT8 bits_dc_chrominance[17] =
-    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_chrominance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-  
-  static const UINT8 bits_ac_luminance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
-  static const UINT8 val_ac_luminance[] =
-    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
-      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
-      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
-      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
-      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
-      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
-      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
-      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
-      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
-      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
-      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
-      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
-      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
-      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
-      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
-      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
-      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
-      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
-      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
-      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-  
-  static const UINT8 bits_ac_chrominance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
-  static const UINT8 val_ac_chrominance[] =
-    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
-      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
-      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
-      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
-      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
-      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
-      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
-      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
-      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
-      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
-      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
-      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
-      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
-      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
-      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
-      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
-      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
-      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
-      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
-      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-  
-  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[0],
-		 bits_dc_luminance, val_dc_luminance);
-  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[0],
-		 bits_ac_luminance, val_ac_luminance);
-  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[1],
-		 bits_dc_chrominance, val_dc_chrominance);
-  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[1],
-		 bits_ac_chrominance, val_ac_chrominance);
-}
-
-
-/*
  * Default parameter setup for compression.
  *
  * Applications that don't choose to use this routine must do their
@@ -300,19 +193,19 @@
   if (cinfo->comp_info == NULL)
     cinfo->comp_info = (jpeg_component_info *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  MAX_COMPONENTS * SIZEOF(jpeg_component_info));
+                                  MAX_COMPONENTS * sizeof(jpeg_component_info));
 
   /* Initialize everything not dependent on the color space */
 
 #if JPEG_LIB_VERSION >= 70
-  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_num = 1;         /* 1:1 scaling */
   cinfo->scale_denom = 1;
 #endif
   cinfo->data_precision = BITS_IN_JSAMPLE;
   /* Set up two quantization tables using default quality of 75 */
   jpeg_set_quality(cinfo, 75, TRUE);
   /* Set up two Huffman tables */
-  std_huff_tables(cinfo);
+  std_huff_tables((j_common_ptr) cinfo);
 
   /* Initialize default arithmetic coding conditioning */
   for (i = 0; i < NUM_ARITH_TBLS; i++) {
@@ -370,8 +263,8 @@
    */
   cinfo->JFIF_major_version = 1; /* Default JFIF version = 1.01 */
   cinfo->JFIF_minor_version = 1;
-  cinfo->density_unit = 0;	/* Pixel size is unknown by default */
-  cinfo->X_density = 1;		/* Pixel aspect ratio is square by default */
+  cinfo->density_unit = 0;      /* Pixel size is unknown by default */
+  cinfo->X_density = 1;         /* Pixel aspect ratio is square by default */
   cinfo->Y_density = 1;
 
   /* Choose JPEG colorspace based on input space, set defaults accordingly */
@@ -429,7 +322,7 @@
 GLOBAL(void)
 jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
 {
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   int ci;
 
 #define SET_COMP(index,id,hsamp,vsamp,quant,dctbl,actbl)  \
@@ -497,7 +390,7 @@
     cinfo->num_components = cinfo->input_components;
     if (cinfo->num_components < 1 || cinfo->num_components > MAX_COMPONENTS)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	       MAX_COMPONENTS);
+               MAX_COMPONENTS);
     for (ci = 0; ci < cinfo->num_components; ci++) {
       SET_COMP(ci, ci, 1,1, 0, 0,0);
     }
@@ -511,8 +404,8 @@
 #ifdef C_PROGRESSIVE_SUPPORTED
 
 LOCAL(jpeg_scan_info *)
-fill_a_scan (jpeg_scan_info * scanptr, int ci,
-	     int Ss, int Se, int Ah, int Al)
+fill_a_scan (jpeg_scan_info *scanptr, int ci,
+             int Ss, int Se, int Ah, int Al)
 /* Support routine: generate one scan for specified component */
 {
   scanptr->comps_in_scan = 1;
@@ -526,8 +419,8 @@
 }
 
 LOCAL(jpeg_scan_info *)
-fill_scans (jpeg_scan_info * scanptr, int ncomps,
-	    int Ss, int Se, int Ah, int Al)
+fill_scans (jpeg_scan_info *scanptr, int ncomps,
+            int Ss, int Se, int Ah, int Al)
 /* Support routine: generate one scan for each component */
 {
   int ci;
@@ -545,7 +438,7 @@
 }
 
 LOCAL(jpeg_scan_info *)
-fill_dc_scans (jpeg_scan_info * scanptr, int ncomps, int Ah, int Al)
+fill_dc_scans (jpeg_scan_info *scanptr, int ncomps, int Ah, int Al)
 /* Support routine: generate interleaved DC scan if possible, else N scans */
 {
   int ci;
@@ -577,7 +470,7 @@
 {
   int ncomps = cinfo->num_components;
   int nscans;
-  jpeg_scan_info * scanptr;
+  jpeg_scan_info *scanptr;
 
   /* Safety check to ensure start_compress not called yet. */
   if (cinfo->global_state != CSTATE_START)
@@ -590,9 +483,9 @@
   } else {
     /* All-purpose script for other color spaces. */
     if (ncomps > MAX_COMPS_IN_SCAN)
-      nscans = 6 * ncomps;	/* 2 DC + 4 AC scans per component */
+      nscans = 6 * ncomps;      /* 2 DC + 4 AC scans per component */
     else
-      nscans = 2 + 4 * ncomps;	/* 2 DC scans; 4 AC scans per component */
+      nscans = 2 + 4 * ncomps;  /* 2 DC scans; 4 AC scans per component */
   }
 
   /* Allocate space for script.
@@ -606,7 +499,7 @@
     cinfo->script_space_size = MAX(nscans, 10);
     cinfo->script_space = (jpeg_scan_info *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-			cinfo->script_space_size * SIZEOF(jpeg_scan_info));
+                        cinfo->script_space_size * sizeof(jpeg_scan_info));
   }
   scanptr = cinfo->script_space;
   cinfo->scan_info = scanptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcphuff.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcphuff.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcphuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcphuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcphuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy encoding routines for progressive JPEG.
  *
@@ -15,7 +18,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jchuff.h"		/* Declarations shared with jchuff.c */
+#include "jchuff.h"             /* Declarations shared with jchuff.c */
 
 #ifdef C_PROGRESSIVE_SUPPORTED
 
@@ -30,36 +33,36 @@
   /* Bit-level coding status.
    * next_output_byte/free_in_buffer are local copies of cinfo->dest fields.
    */
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
-  INT32 put_buffer;		/* current bit-accumulation buffer */
-  int put_bits;			/* # of bits now in it */
-  j_compress_ptr cinfo;		/* link to cinfo (needed for dump_buffer) */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  size_t put_buffer;            /* current bit-accumulation buffer */
+  int put_bits;                 /* # of bits now in it */
+  j_compress_ptr cinfo;         /* link to cinfo (needed for dump_buffer) */
 
   /* Coding status for DC components */
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 
   /* Coding status for AC components */
-  int ac_tbl_no;		/* the table number of the single component */
-  unsigned int EOBRUN;		/* run length of EOBs */
-  unsigned int BE;		/* # of buffered correction bits before MCU */
-  char * bit_buffer;		/* buffer for correction bits (1 per char) */
+  int ac_tbl_no;                /* the table number of the single component */
+  unsigned int EOBRUN;          /* run length of EOBs */
+  unsigned int BE;              /* # of buffered correction bits before MCU */
+  char *bit_buffer;             /* buffer for correction bits (1 per char) */
   /* packing correction bits tightly would save some space but cost time... */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to derived tables (these workspaces have image lifespan).
    * Since any one scan codes only DC or only AC, we only need one set
    * of tables, not one for DC and one for AC.
    */
-  c_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *derived_tbls[NUM_HUFF_TBLS];
 
   /* Statistics tables for optimization; again, one set is enough */
-  long * count_ptrs[NUM_HUFF_TBLS];
+  long *count_ptrs[NUM_HUFF_TBLS];
 } phuff_entropy_encoder;
 
-typedef phuff_entropy_encoder * phuff_entropy_ptr;
+typedef phuff_entropy_encoder *phuff_entropy_ptr;
 
 /* MAX_CORR_BITS is the number of bits the AC refinement correction-bit
  * buffer can hold.  Larger sizes may slightly improve compression, but
@@ -67,35 +70,35 @@
  * The minimum safe size is 64 bits.
  */
 
-#define MAX_CORR_BITS  1000	/* Max # of correction bits I can buffer */
+#define MAX_CORR_BITS  1000     /* Max # of correction bits I can buffer */
 
-/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
- * We assume that int right shift is unsigned if INT32 right shift is,
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
  * which should be safe.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	int ishift_temp;
+#define ISHIFT_TEMPS    int ishift_temp;
 #define IRIGHT_SHIFT(x,shft)  \
-	((ishift_temp = (x)) < 0 ? \
-	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
-	 (ishift_temp >> (shft)))
+        ((ishift_temp = (x)) < 0 ? \
+         (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+         (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 /* Forward declarations */
-METHODDEF(boolean) encode_mcu_DC_first JPP((j_compress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_AC_first JPP((j_compress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_DC_refine JPP((j_compress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_AC_refine JPP((j_compress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_phuff JPP((j_compress_ptr cinfo));
-METHODDEF(void) finish_pass_gather_phuff JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_DC_first (j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_AC_first (j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_DC_refine (j_compress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_AC_refine (j_compress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_phuff (j_compress_ptr cinfo);
+METHODDEF(void) finish_pass_gather_phuff (j_compress_ptr cinfo);
 
 
 /*
@@ -104,11 +107,11 @@
 
 METHODDEF(void)
 start_pass_phuff (j_compress_ptr cinfo, boolean gather_statistics)
-{  
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   entropy->cinfo = cinfo;
   entropy->gather_statistics = gather_statistics;
@@ -130,9 +133,9 @@
       entropy->pub.encode_mcu = encode_mcu_AC_refine;
       /* AC refinement needs a correction bit buffer */
       if (entropy->bit_buffer == NULL)
-	entropy->bit_buffer = (char *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      MAX_CORR_BITS * SIZEOF(char));
+        entropy->bit_buffer = (char *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      MAX_CORR_BITS * sizeof(char));
     }
   }
   if (gather_statistics)
@@ -149,8 +152,8 @@
     entropy->last_dc_val[ci] = 0;
     /* Get table index */
     if (is_DC_band) {
-      if (cinfo->Ah != 0)	/* DC refinement needs no table */
-	continue;
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
       tbl = compptr->dc_tbl_no;
     } else {
       entropy->ac_tbl_no = tbl = compptr->ac_tbl_no;
@@ -163,15 +166,15 @@
       /* Allocate and zero the statistics tables */
       /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
       if (entropy->count_ptrs[tbl] == NULL)
-	entropy->count_ptrs[tbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->count_ptrs[tbl], 257 * SIZEOF(long));
+        entropy->count_ptrs[tbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->count_ptrs[tbl], 257 * sizeof(long));
     } else {
       /* Compute derived values for Huffman table */
       /* We may do this more than once for a table, but it's not expensive */
       jpeg_make_c_derived_tbl(cinfo, is_DC_band, tbl,
-			      & entropy->derived_tbls[tbl]);
+                              & entropy->derived_tbls[tbl]);
     }
   }
 
@@ -196,16 +199,16 @@
 
 /* Emit a byte */
 #define emit_byte(entropy,val)  \
-	{ *(entropy)->next_output_byte++ = (JOCTET) (val);  \
-	  if (--(entropy)->free_in_buffer == 0)  \
-	    dump_buffer(entropy); }
+        { *(entropy)->next_output_byte++ = (JOCTET) (val);  \
+          if (--(entropy)->free_in_buffer == 0)  \
+            dump_buffer(entropy); }
 
 
 LOCAL(void)
 dump_buffer (phuff_entropy_ptr entropy)
 /* Empty the output buffer; we do not support suspension in this module. */
 {
-  struct jpeg_destination_mgr * dest = entropy->cinfo->dest;
+  struct jpeg_destination_mgr *dest = entropy->cinfo->dest;
 
   if (! (*dest->empty_output_buffer) (entropy->cinfo))
     ERREXIT(entropy->cinfo, JERR_CANT_SUSPEND);
@@ -228,7 +231,7 @@
 /* Emit some bits, unless we are in gather mode */
 {
   /* This routine is heavily used, so it's worth coding tightly. */
-  register INT32 put_buffer = (INT32) code;
+  register size_t put_buffer = (size_t) code;
   register int put_bits = entropy->put_bits;
 
   /* if size is 0, caller used an invalid Huffman table entry */
@@ -236,21 +239,21 @@
     ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
 
   if (entropy->gather_statistics)
-    return;			/* do nothing if we're only getting stats */
+    return;                     /* do nothing if we're only getting stats */
+
+  put_buffer &= (((size_t) 1)<<size) - 1; /* mask off any extra bits in code */
+
+  put_bits += size;             /* new number of bits in buffer */
 
-  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
-  
-  put_bits += size;		/* new number of bits in buffer */
-  
   put_buffer <<= 24 - put_bits; /* align incoming bits */
 
   put_buffer |= entropy->put_buffer; /* and merge with old buffer contents */
 
   while (put_bits >= 8) {
     int c = (int) ((put_buffer >> 16) & 0xFF);
-    
+
     emit_byte(entropy, c);
-    if (c == 0xFF) {		/* need to stuff a zero byte? */
+    if (c == 0xFF) {            /* need to stuff a zero byte? */
       emit_byte(entropy, 0);
     }
     put_buffer <<= 8;
@@ -281,7 +284,7 @@
   if (entropy->gather_statistics)
     entropy->count_ptrs[tbl_no][symbol]++;
   else {
-    c_derived_tbl * tbl = entropy->derived_tbls[tbl_no];
+    c_derived_tbl *tbl = entropy->derived_tbls[tbl_no];
     emit_bits(entropy, tbl->ehufco[symbol], tbl->ehufsi[symbol]);
   }
 }
@@ -292,11 +295,11 @@
  */
 
 LOCAL(void)
-emit_buffered_bits (phuff_entropy_ptr entropy, char * bufstart,
-		    unsigned int nbits)
+emit_buffered_bits (phuff_entropy_ptr entropy, char *bufstart,
+                    unsigned int nbits)
 {
   if (entropy->gather_statistics)
-    return;			/* no real work */
+    return;                     /* no real work */
 
   while (nbits > 0) {
     emit_bits(entropy, (unsigned int) (*bufstart), 1);
@@ -315,7 +318,7 @@
 {
   register int temp, nbits;
 
-  if (entropy->EOBRUN > 0) {	/* if there is any pending EOBRUN */
+  if (entropy->EOBRUN > 0) {    /* if there is any pending EOBRUN */
     temp = entropy->EOBRUN;
     nbits = 0;
     while ((temp >>= 1))
@@ -380,7 +383,7 @@
   int blkn, ci;
   int Al = cinfo->Al;
   JBLOCKROW block;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   ISHIFT_TEMPS
 
   entropy->next_output_byte = cinfo->dest->next_output_byte;
@@ -409,12 +412,12 @@
     /* Encode the DC coefficient difference per section G.1.2.1 */
     temp2 = temp;
     if (temp < 0) {
-      temp = -temp;		/* temp is abs value of input */
+      temp = -temp;             /* temp is abs value of input */
       /* For a negative input, want temp2 = bitwise complement of abs(input) */
       /* This code assumes we are on a two's complement machine */
       temp2--;
     }
-    
+
     /* Find the number of bits needed for the magnitude of the coefficient */
     nbits = 0;
     while (temp) {
@@ -426,13 +429,13 @@
      */
     if (nbits > MAX_COEF_BITS+1)
       ERREXIT(cinfo, JERR_BAD_DCT_COEF);
-    
+
     /* Count/emit the Huffman-coded symbol for the number of bits */
     emit_symbol(entropy, compptr->dc_tbl_no, nbits);
-    
+
     /* Emit that number of bits of the value, if positive, */
     /* or the complement of its magnitude, if negative. */
-    if (nbits)			/* emit_bits rejects calls with size 0 */
+    if (nbits)                  /* emit_bits rejects calls with size 0 */
       emit_bits(entropy, (unsigned int) temp2, nbits);
   }
 
@@ -481,9 +484,9 @@
   block = MCU_data[0];
 
   /* Encode the AC coefficients per section G.1.2.2, fig. G.3 */
-  
-  r = 0;			/* r = run length of zeros */
-   
+
+  r = 0;                        /* r = run length of zeros */
+
   for (k = cinfo->Ss; k <= Se; k++) {
     if ((temp = (*block)[jpeg_natural_order[k]]) == 0) {
       r++;
@@ -495,12 +498,12 @@
      * interwoven with finding the abs value (temp) and output bits (temp2).
      */
     if (temp < 0) {
-      temp = -temp;		/* temp is abs value of input */
-      temp >>= Al;		/* apply the point transform */
+      temp = -temp;             /* temp is abs value of input */
+      temp >>= Al;              /* apply the point transform */
       /* For a negative coef, want temp2 = bitwise complement of abs(coef) */
       temp2 = ~temp;
     } else {
-      temp >>= Al;		/* apply the point transform */
+      temp >>= Al;              /* apply the point transform */
       temp2 = temp;
     }
     /* Watch out for case that nonzero coef is zero after point transform */
@@ -519,7 +522,7 @@
     }
 
     /* Find the number of bits needed for the magnitude of the coefficient */
-    nbits = 1;			/* there must be at least one 1 bit */
+    nbits = 1;                  /* there must be at least one 1 bit */
     while ((temp >>= 1))
       nbits++;
     /* Check for out-of-range coefficient values */
@@ -533,13 +536,13 @@
     /* or the complement of its magnitude, if negative. */
     emit_bits(entropy, (unsigned int) temp2, nbits);
 
-    r = 0;			/* reset zero run length */
+    r = 0;                      /* reset zero run length */
   }
 
-  if (r > 0) {			/* If there are trailing zeroes, */
-    entropy->EOBRUN++;		/* count an EOB */
+  if (r > 0) {                  /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
     if (entropy->EOBRUN == 0x7FFF)
-      emit_eobrun(entropy);	/* force it out to avoid overflow */
+      emit_eobrun(entropy);     /* force it out to avoid overflow */
   }
 
   cinfo->dest->next_output_byte = entropy->next_output_byte;
@@ -648,17 +651,17 @@
      * in C, we shift after obtaining the absolute value.
      */
     if (temp < 0)
-      temp = -temp;		/* temp is abs value of input */
-    temp >>= Al;		/* apply the point transform */
-    absvalues[k] = temp;	/* save abs value for main pass */
+      temp = -temp;             /* temp is abs value of input */
+    temp >>= Al;                /* apply the point transform */
+    absvalues[k] = temp;        /* save abs value for main pass */
     if (temp == 1)
-      EOB = k;			/* EOB = index of last newly-nonzero coef */
+      EOB = k;                  /* EOB = index of last newly-nonzero coef */
   }
 
   /* Encode the AC coefficients per section G.1.2.3, fig. G.7 */
-  
-  r = 0;			/* r = run length of zeros */
-  BR = 0;			/* BR = count of buffered bits added now */
+
+  r = 0;                        /* r = run length of zeros */
+  BR = 0;                       /* BR = count of buffered bits added now */
   BR_buffer = entropy->bit_buffer + entropy->BE; /* Append bits to buffer */
 
   for (k = cinfo->Ss; k <= Se; k++) {
@@ -705,12 +708,12 @@
     emit_buffered_bits(entropy, BR_buffer, BR);
     BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
     BR = 0;
-    r = 0;			/* reset zero run length */
+    r = 0;                      /* reset zero run length */
   }
 
-  if (r > 0 || BR > 0) {	/* If there are trailing zeroes, */
-    entropy->EOBRUN++;		/* count an EOB */
-    entropy->BE += BR;		/* concat my correction bits to older ones */
+  if (r > 0 || BR > 0) {        /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
+    entropy->BE += BR;          /* concat my correction bits to older ones */
     /* We force out the EOB if we risk either:
      * 1. overflow of the EOB counter;
      * 2. overflow of the correction bit buffer during the next MCU.
@@ -742,7 +745,7 @@
 
 METHODDEF(void)
 finish_pass_phuff (j_compress_ptr cinfo)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
 
   entropy->next_output_byte = cinfo->dest->next_output_byte;
@@ -767,7 +770,7 @@
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JHUFF_TBL **htblptr;
   boolean did[NUM_HUFF_TBLS];
 
@@ -779,13 +782,13 @@
   /* It's important not to apply jpeg_gen_optimal_table more than once
    * per table, because it clobbers the input frequency counts!
    */
-  MEMZERO(did, SIZEOF(did));
+  MEMZERO(did, sizeof(did));
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
     if (is_DC_band) {
-      if (cinfo->Ah != 0)	/* DC refinement needs no table */
-	continue;
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
       tbl = compptr->dc_tbl_no;
     } else {
       tbl = compptr->ac_tbl_no;
@@ -816,7 +819,7 @@
 
   entropy = (phuff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(phuff_entropy_encoder));
+                                sizeof(phuff_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass_phuff;
 
@@ -825,7 +828,7 @@
     entropy->derived_tbls[i] = NULL;
     entropy->count_ptrs[i] = NULL;
   }
-  entropy->bit_buffer = NULL;	/* needed only in AC refinement scan */
+  entropy->bit_buffer = NULL;   /* needed only in AC refinement scan */
 }
 
 #endif /* C_PROGRESSIVE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcprepct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcprepct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcprepct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcprepct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcprepct.c
  *
+ * This file is part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the compression preprocessing controller.
  * This controller manages the color conversion, downsampling,
@@ -58,16 +61,16 @@
    */
   JSAMPARRAY color_buf[MAX_COMPONENTS];
 
-  JDIMENSION rows_to_go;	/* counts rows remaining in source image */
-  int next_buf_row;		/* index of next row to store in color_buf */
+  JDIMENSION rows_to_go;        /* counts rows remaining in source image */
+  int next_buf_row;             /* index of next row to store in color_buf */
 
-#ifdef CONTEXT_ROWS_SUPPORTED	/* only needed for context case */
-  int this_row_group;		/* starting row index of group to process */
-  int next_buf_stop;		/* downsample when we reach this index */
+#ifdef CONTEXT_ROWS_SUPPORTED   /* only needed for context case */
+  int this_row_group;           /* starting row index of group to process */
+  int next_buf_stop;            /* downsample when we reach this index */
 #endif
 } my_prep_controller;
 
-typedef my_prep_controller * my_prep_ptr;
+typedef my_prep_controller *my_prep_ptr;
 
 
 /*
@@ -104,13 +107,13 @@
 
 LOCAL(void)
 expand_bottom_edge (JSAMPARRAY image_data, JDIMENSION num_cols,
-		    int input_rows, int output_rows)
+                    int input_rows, int output_rows)
 {
   register int row;
 
   for (row = input_rows; row < output_rows; row++) {
     jcopy_sample_rows(image_data, input_rows-1, image_data, row,
-		      1, num_cols);
+                      1, num_cols);
   }
 }
 
@@ -126,43 +129,43 @@
 
 METHODDEF(void)
 pre_process_data (j_compress_ptr cinfo,
-		  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-		  JDIMENSION in_rows_avail,
-		  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
-		  JDIMENSION out_row_groups_avail)
+                  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                  JDIMENSION in_rows_avail,
+                  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                  JDIMENSION out_row_groups_avail)
 {
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int numrows, ci;
   JDIMENSION inrows;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   while (*in_row_ctr < in_rows_avail &&
-	 *out_row_group_ctr < out_row_groups_avail) {
+         *out_row_group_ctr < out_row_groups_avail) {
     /* Do color conversion to fill the conversion buffer. */
     inrows = in_rows_avail - *in_row_ctr;
     numrows = cinfo->max_v_samp_factor - prep->next_buf_row;
     numrows = (int) MIN((JDIMENSION) numrows, inrows);
     (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
-				       prep->color_buf,
-				       (JDIMENSION) prep->next_buf_row,
-				       numrows);
+                                       prep->color_buf,
+                                       (JDIMENSION) prep->next_buf_row,
+                                       numrows);
     *in_row_ctr += numrows;
     prep->next_buf_row += numrows;
     prep->rows_to_go -= numrows;
     /* If at bottom of image, pad to fill the conversion buffer. */
     if (prep->rows_to_go == 0 &&
-	prep->next_buf_row < cinfo->max_v_samp_factor) {
+        prep->next_buf_row < cinfo->max_v_samp_factor) {
       for (ci = 0; ci < cinfo->num_components; ci++) {
-	expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
-			   prep->next_buf_row, cinfo->max_v_samp_factor);
+        expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+                           prep->next_buf_row, cinfo->max_v_samp_factor);
       }
       prep->next_buf_row = cinfo->max_v_samp_factor;
     }
     /* If we've filled the conversion buffer, empty it. */
     if (prep->next_buf_row == cinfo->max_v_samp_factor) {
       (*cinfo->downsample->downsample) (cinfo,
-					prep->color_buf, (JDIMENSION) 0,
-					output_buf, *out_row_group_ctr);
+                                        prep->color_buf, (JDIMENSION) 0,
+                                        output_buf, *out_row_group_ctr);
       prep->next_buf_row = 0;
       (*out_row_group_ctr)++;
     }
@@ -170,16 +173,16 @@
      * Note we assume the caller is providing a one-iMCU-height output buffer!
      */
     if (prep->rows_to_go == 0 &&
-	*out_row_group_ctr < out_row_groups_avail) {
+        *out_row_group_ctr < out_row_groups_avail) {
       for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	   ci++, compptr++) {
-	expand_bottom_edge(output_buf[ci],
-			   compptr->width_in_blocks * DCTSIZE,
-			   (int) (*out_row_group_ctr * compptr->v_samp_factor),
-			   (int) (out_row_groups_avail * compptr->v_samp_factor));
+           ci++, compptr++) {
+        expand_bottom_edge(output_buf[ci],
+                           compptr->width_in_blocks * DCTSIZE,
+                           (int) (*out_row_group_ctr * compptr->v_samp_factor),
+                           (int) (out_row_groups_avail * compptr->v_samp_factor));
       }
       *out_row_group_ctr = out_row_groups_avail;
-      break;			/* can exit outer loop without test */
+      break;                    /* can exit outer loop without test */
     }
   }
 }
@@ -193,10 +196,10 @@
 
 METHODDEF(void)
 pre_process_context (j_compress_ptr cinfo,
-		     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-		     JDIMENSION in_rows_avail,
-		     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
-		     JDIMENSION out_row_groups_avail)
+                     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                     JDIMENSION in_rows_avail,
+                     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                     JDIMENSION out_row_groups_avail)
 {
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int numrows, ci;
@@ -210,19 +213,19 @@
       numrows = prep->next_buf_stop - prep->next_buf_row;
       numrows = (int) MIN((JDIMENSION) numrows, inrows);
       (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
-					 prep->color_buf,
-					 (JDIMENSION) prep->next_buf_row,
-					 numrows);
+                                         prep->color_buf,
+                                         (JDIMENSION) prep->next_buf_row,
+                                         numrows);
       /* Pad at top of image, if first time through */
       if (prep->rows_to_go == cinfo->image_height) {
-	for (ci = 0; ci < cinfo->num_components; ci++) {
-	  int row;
-	  for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
-	    jcopy_sample_rows(prep->color_buf[ci], 0,
-			      prep->color_buf[ci], -row,
-			      1, cinfo->image_width);
-	  }
-	}
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          int row;
+          for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
+            jcopy_sample_rows(prep->color_buf[ci], 0,
+                              prep->color_buf[ci], -row,
+                              1, cinfo->image_width);
+          }
+        }
       }
       *in_row_ctr += numrows;
       prep->next_buf_row += numrows;
@@ -230,29 +233,29 @@
     } else {
       /* Return for more data, unless we are at the bottom of the image. */
       if (prep->rows_to_go != 0)
-	break;
+        break;
       /* When at bottom of image, pad to fill the conversion buffer. */
       if (prep->next_buf_row < prep->next_buf_stop) {
-	for (ci = 0; ci < cinfo->num_components; ci++) {
-	  expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
-			     prep->next_buf_row, prep->next_buf_stop);
-	}
-	prep->next_buf_row = prep->next_buf_stop;
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+                             prep->next_buf_row, prep->next_buf_stop);
+        }
+        prep->next_buf_row = prep->next_buf_stop;
       }
     }
     /* If we've gotten enough data, downsample a row group. */
     if (prep->next_buf_row == prep->next_buf_stop) {
       (*cinfo->downsample->downsample) (cinfo,
-					prep->color_buf,
-					(JDIMENSION) prep->this_row_group,
-					output_buf, *out_row_group_ctr);
+                                        prep->color_buf,
+                                        (JDIMENSION) prep->this_row_group,
+                                        output_buf, *out_row_group_ctr);
       (*out_row_group_ctr)++;
       /* Advance pointers with wraparound as necessary. */
       prep->this_row_group += cinfo->max_v_samp_factor;
       if (prep->this_row_group >= buf_height)
-	prep->this_row_group = 0;
+        prep->this_row_group = 0;
       if (prep->next_buf_row >= buf_height)
-	prep->next_buf_row = 0;
+        prep->next_buf_row = 0;
       prep->next_buf_stop = prep->next_buf_row + cinfo->max_v_samp_factor;
     }
   }
@@ -269,7 +272,7 @@
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int rgroup_height = cinfo->max_v_samp_factor;
   int ci, i;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JSAMPARRAY true_buffer, fake_buffer;
 
   /* Grab enough space for fake row pointers for all the components;
@@ -277,8 +280,8 @@
    */
   fake_buffer = (JSAMPARRAY)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(cinfo->num_components * 5 * rgroup_height) *
-				SIZEOF(JSAMPROW));
+                                (cinfo->num_components * 5 * rgroup_height) *
+                                sizeof(JSAMPROW));
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
@@ -289,11 +292,11 @@
     true_buffer = (*cinfo->mem->alloc_sarray)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
        (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
-		      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+                      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
        (JDIMENSION) (3 * rgroup_height));
     /* Copy true buffer row pointers into the middle of the fake row array */
     MEMCOPY(fake_buffer + rgroup_height, true_buffer,
-	    3 * rgroup_height * SIZEOF(JSAMPROW));
+            3 * rgroup_height * sizeof(JSAMPROW));
     /* Fill in the above and below wraparound pointers */
     for (i = 0; i < rgroup_height; i++) {
       fake_buffer[i] = true_buffer[2 * rgroup_height + i];
@@ -316,14 +319,14 @@
 {
   my_prep_ptr prep;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
-  if (need_full_buffer)		/* safety check */
+  if (need_full_buffer)         /* safety check */
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 
   prep = (my_prep_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_prep_controller));
+                                sizeof(my_prep_controller));
   cinfo->prep = (struct jpeg_c_prep_controller *) prep;
   prep->pub.start_pass = start_pass_prep;
 
@@ -343,12 +346,12 @@
     /* No context, just make it tall enough for one row group */
     prep->pub.pre_process_data = pre_process_data;
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       prep->color_buf[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
-			cinfo->max_h_samp_factor) / compptr->h_samp_factor),
-	 (JDIMENSION) cinfo->max_v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+                        cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+         (JDIMENSION) cinfo->max_v_samp_factor);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcsample.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcsample.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcsample.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcsample.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jcsample.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains downsampling routines.
  *
@@ -53,20 +57,21 @@
 
 
 /* Pointer to routine to downsample a single component */
-typedef JMETHOD(void, downsample1_ptr,
-		(j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data));
+typedef void (*downsample1_ptr) (j_compress_ptr cinfo,
+                                 jpeg_component_info *compptr,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY output_data);
 
 /* Private subobject */
 
 typedef struct {
-  struct jpeg_downsampler pub;	/* public fields */
+  struct jpeg_downsampler pub;  /* public fields */
 
   /* Downsampling method pointers, one per component */
   downsample1_ptr methods[MAX_COMPONENTS];
 } my_downsampler;
 
-typedef my_downsampler * my_downsample_ptr;
+typedef my_downsampler *my_downsample_ptr;
 
 
 /*
@@ -87,7 +92,7 @@
 
 LOCAL(void)
 expand_right_edge (JSAMPARRAY image_data, int num_rows,
-		   JDIMENSION input_cols, JDIMENSION output_cols)
+                   JDIMENSION input_cols, JDIMENSION output_cols)
 {
   register JSAMPROW ptr;
   register JSAMPLE pixval;
@@ -98,9 +103,9 @@
   if (numcols > 0) {
     for (row = 0; row < num_rows; row++) {
       ptr = image_data[row] + input_cols;
-      pixval = ptr[-1];		/* don't need GETJSAMPLE() here */
+      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
       for (count = numcols; count > 0; count--)
-	*ptr++ = pixval;
+        *ptr++ = pixval;
     }
   }
 }
@@ -114,12 +119,12 @@
 
 METHODDEF(void)
 sep_downsample (j_compress_ptr cinfo,
-		JSAMPIMAGE input_buf, JDIMENSION in_row_index,
-		JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
+                JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+                JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
 {
   my_downsample_ptr downsample = (my_downsample_ptr) cinfo->downsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JSAMPARRAY in_ptr, out_ptr;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -139,14 +144,14 @@
  */
 
 METHODDEF(void)
-int_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		JSAMPARRAY input_data, JSAMPARRAY output_data)
+int_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow, h_expand, v_expand, numpix, numpix2, h, v;
-  JDIMENSION outcol, outcol_h;	/* outcol_h == outcol*h_expand */
+  JDIMENSION outcol, outcol_h;  /* outcol_h == outcol*h_expand */
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   JSAMPROW inptr, outptr;
-  INT32 outvalue;
+  JLONG outvalue;
 
   h_expand = cinfo->max_h_samp_factor / compptr->h_samp_factor;
   v_expand = cinfo->max_v_samp_factor / compptr->v_samp_factor;
@@ -158,19 +163,19 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * h_expand);
+                    cinfo->image_width, output_cols * h_expand);
 
   inrow = 0;
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     for (outcol = 0, outcol_h = 0; outcol < output_cols;
-	 outcol++, outcol_h += h_expand) {
+         outcol++, outcol_h += h_expand) {
       outvalue = 0;
       for (v = 0; v < v_expand; v++) {
-	inptr = input_data[inrow+v] + outcol_h;
-	for (h = 0; h < h_expand; h++) {
-	  outvalue += (INT32) GETJSAMPLE(*inptr++);
-	}
+        inptr = input_data[inrow+v] + outcol_h;
+        for (h = 0; h < h_expand; h++) {
+          outvalue += (JLONG) GETJSAMPLE(*inptr++);
+        }
       }
       *outptr++ = (JSAMPLE) ((outvalue + numpix2) / numpix);
     }
@@ -186,15 +191,15 @@
  */
 
 METHODDEF(void)
-fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY output_data)
+fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   /* Copy the data */
   jcopy_sample_rows(input_data, 0, output_data, 0,
-		    cinfo->max_v_samp_factor, cinfo->image_width);
+                    cinfo->max_v_samp_factor, cinfo->image_width);
   /* Edge-expand */
   expand_right_edge(output_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
+                    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
 }
 
 
@@ -211,8 +216,8 @@
  */
 
 METHODDEF(void)
-h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                 JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int outrow;
   JDIMENSION outcol;
@@ -225,16 +230,16 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     inptr = input_data[outrow];
-    bias = 0;			/* bias = 0,1,0,1,... for successive samples */
+    bias = 0;                   /* bias = 0,1,0,1,... for successive samples */
     for (outcol = 0; outcol < output_cols; outcol++) {
       *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1])
-			      + bias) >> 1);
-      bias ^= 1;		/* 0=>1, 1=>0 */
+                              + bias) >> 1);
+      bias ^= 1;                /* 0=>1, 1=>0 */
       inptr += 2;
     }
   }
@@ -248,8 +253,8 @@
  */
 
 METHODDEF(void)
-h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                 JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow;
   JDIMENSION outcol;
@@ -262,19 +267,19 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   inrow = 0;
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     inptr0 = input_data[inrow];
     inptr1 = input_data[inrow+1];
-    bias = 1;			/* bias = 1,2,1,2,... for successive samples */
+    bias = 1;                   /* bias = 1,2,1,2,... for successive samples */
     for (outcol = 0; outcol < output_cols; outcol++) {
       *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-			      GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
-			      + bias) >> 2);
-      bias ^= 3;		/* 1=>2, 2=>1 */
+                              GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
+                              + bias) >> 2);
+      bias ^= 3;                /* 1=>2, 2=>1 */
       inptr0 += 2; inptr1 += 2;
     }
     inrow += 2;
@@ -291,21 +296,21 @@
  */
 
 METHODDEF(void)
-h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-			JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow;
   JDIMENSION colctr;
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   register JSAMPROW inptr0, inptr1, above_ptr, below_ptr, outptr;
-  INT32 membersum, neighsum, memberscale, neighscale;
+  JLONG membersum, neighsum, memberscale, neighscale;
 
   /* Expand input data enough to let all the output samples be generated
    * by the standard loop.  Special-casing padded output would be more
    * efficient.
    */
   expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   /* We don't bother to form the individual "smoothed" input pixel values;
    * we can directly compute the output which is the average of the four
@@ -333,14 +338,14 @@
 
     /* Special case for first column: pretend column -1 is same as column 0 */
     membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
     neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-	       GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
-	       GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
+               GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
     neighsum += neighsum;
     neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
-		GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+                GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
     inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
@@ -348,17 +353,17 @@
     for (colctr = output_cols - 2; colctr > 0; colctr--) {
       /* sum of pixels directly mapped to this output element */
       membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
       /* sum of edge-neighbor pixels */
       neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-		 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-		 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
-		 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
+                 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+                 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
+                 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
       /* The edge-neighbors count twice as much as corner-neighbors */
       neighsum += neighsum;
       /* Add in the corner-neighbors */
       neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
-		  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+                  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
       /* form final output scaled up by 2^16 */
       membersum = membersum * memberscale + neighsum * neighscale;
       /* round, descale and output it */
@@ -368,14 +373,14 @@
 
     /* Special case for last column */
     membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
     neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-	       GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
-	       GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
+               GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
     neighsum += neighsum;
     neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
-		GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
+                GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
 
@@ -392,13 +397,13 @@
 
 METHODDEF(void)
 fullsize_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
-			    JSAMPARRAY input_data, JSAMPARRAY output_data)
+                            JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int outrow;
   JDIMENSION colctr;
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   register JSAMPROW inptr, above_ptr, below_ptr, outptr;
-  INT32 membersum, neighsum, memberscale, neighscale;
+  JLONG membersum, neighsum, memberscale, neighscale;
   int colsum, lastcolsum, nextcolsum;
 
   /* Expand input data enough to let all the output samples be generated
@@ -406,7 +411,7 @@
    * efficient.
    */
   expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
-		    cinfo->image_width, output_cols);
+                    cinfo->image_width, output_cols);
 
   /* Each of the eight neighbor pixels contributes a fraction SF to the
    * smoothed pixel, while the main pixel contributes (1-8*SF).  In order
@@ -425,10 +430,10 @@
 
     /* Special case for first column */
     colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
-	     GETJSAMPLE(*inptr);
+             GETJSAMPLE(*inptr);
     membersum = GETJSAMPLE(*inptr++);
     nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-		 GETJSAMPLE(*inptr);
+                 GETJSAMPLE(*inptr);
     neighsum = colsum + (colsum - membersum) + nextcolsum;
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
@@ -438,7 +443,7 @@
       membersum = GETJSAMPLE(*inptr++);
       above_ptr++; below_ptr++;
       nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-		   GETJSAMPLE(*inptr);
+                   GETJSAMPLE(*inptr);
       neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
       membersum = membersum * memberscale + neighsum * neighscale;
       *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
@@ -467,12 +472,12 @@
 {
   my_downsample_ptr downsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   boolean smoothok = TRUE;
 
   downsample = (my_downsample_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_downsampler));
+                                sizeof(my_downsampler));
   cinfo->downsample = (struct jpeg_downsampler *) downsample;
   downsample->pub.start_pass = start_pass_downsample;
   downsample->pub.downsample = sep_downsample;
@@ -485,35 +490,42 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor == cinfo->max_h_samp_factor &&
-	compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+        compptr->v_samp_factor == cinfo->max_v_samp_factor) {
 #ifdef INPUT_SMOOTHING_SUPPORTED
       if (cinfo->smoothing_factor) {
-	downsample->methods[ci] = fullsize_smooth_downsample;
-	downsample->pub.need_context_rows = TRUE;
+        downsample->methods[ci] = fullsize_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
       } else
 #endif
-	downsample->methods[ci] = fullsize_downsample;
+        downsample->methods[ci] = fullsize_downsample;
     } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
-	       compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+               compptr->v_samp_factor == cinfo->max_v_samp_factor) {
       smoothok = FALSE;
       if (jsimd_can_h2v1_downsample())
         downsample->methods[ci] = jsimd_h2v1_downsample;
       else
         downsample->methods[ci] = h2v1_downsample;
     } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
-	       compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
+               compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
 #ifdef INPUT_SMOOTHING_SUPPORTED
       if (cinfo->smoothing_factor) {
-	downsample->methods[ci] = h2v2_smooth_downsample;
-	downsample->pub.need_context_rows = TRUE;
+#if defined(__mips__)
+        if (jsimd_can_h2v2_smooth_downsample())
+          downsample->methods[ci] = jsimd_h2v2_smooth_downsample;
+        else
+#endif
+          downsample->methods[ci] = h2v2_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
       } else
 #endif
-	if (jsimd_can_h2v2_downsample())
-	  downsample->methods[ci] = jsimd_h2v2_downsample;
-	else
-	  downsample->methods[ci] = h2v2_downsample;
+      {
+        if (jsimd_can_h2v2_downsample())
+          downsample->methods[ci] = jsimd_h2v2_downsample;
+        else
+          downsample->methods[ci] = h2v2_downsample;
+      }
     } else if ((cinfo->max_h_samp_factor % compptr->h_samp_factor) == 0 &&
-	       (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
+               (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
       smoothok = FALSE;
       downsample->methods[ci] = int_downsample;
     } else
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcstest.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jcstest.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jcstest.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jcstest.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* This program demonstrates how to check for the colorspace extension
+   capabilities of libjpeg-turbo at both compile time and run time. */
+
+#include <stdio.h>
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+
+#ifndef JCS_EXTENSIONS
+#define JCS_EXT_RGB 6
+#endif
+#if !defined(JCS_EXTENSIONS) || !defined(JCS_ALPHA_EXTENSIONS)
+#define JCS_EXT_RGBA 12
+#endif
+
+static char lasterror[JMSG_LENGTH_MAX] = "No error";
+
+typedef struct _error_mgr {
+  struct jpeg_error_mgr pub;
+  jmp_buf jb;
+} error_mgr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+  error_mgr *myerr = (error_mgr *)cinfo->err;
+  (*cinfo->err->output_message)(cinfo);
+  longjmp(myerr->jb, 1);
+}
+
+static void my_output_message(j_common_ptr cinfo)
+{
+  (*cinfo->err->format_message)(cinfo, lasterror);
+}
+
+int main(void)
+{
+  int jcs_valid = -1, jcs_alpha_valid = -1;
+  struct jpeg_compress_struct cinfo;
+  error_mgr jerr;
+
+  printf("libjpeg-turbo colorspace extensions:\n");
+  #if JCS_EXTENSIONS
+  printf("  Present at compile time\n");
+  #else
+  printf("  Not present at compile time\n");
+  #endif
+
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  jerr.pub.output_message = my_output_message;
+
+  if(setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_valid = 0;
+    goto done;
+  }
+
+  jpeg_create_compress(&cinfo);
+  cinfo.input_components = 3;
+  jpeg_set_defaults(&cinfo);
+  cinfo.in_color_space = JCS_EXT_RGB;
+  jpeg_default_colorspace(&cinfo);
+  jcs_valid = 1;
+
+  done:
+  if (jcs_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  printf("libjpeg-turbo alpha colorspace extensions:\n");
+  #if JCS_ALPHA_EXTENSIONS
+  printf("  Present at compile time\n");
+  #else
+  printf("  Not present at compile time\n");
+  #endif
+
+  if(setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_alpha_valid = 0;
+    goto done2;
+  }
+
+  cinfo.in_color_space = JCS_EXT_RGBA;
+  jpeg_default_colorspace(&cinfo);
+  jcs_alpha_valid = 1;
+
+  done2:
+  if (jcs_alpha_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  jpeg_destroy_compress(&cinfo);
+  return 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jctrans.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jctrans.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jctrans.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jctrans.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jctrans.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1998, Thomas G. Lane.
  * Modified 2000-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains library routines for transcoding compression,
  * that is, writing raw DCT coefficient arrays to an output JPEG file.
@@ -18,9 +21,9 @@
 
 /* Forward declarations */
 LOCAL(void) transencode_master_selection
-	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+        (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays);
 LOCAL(void) transencode_coef_controller
-	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+        (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays);
 
 
 /*
@@ -36,7 +39,7 @@
  */
 
 GLOBAL(void)
-jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays)
+jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays)
 {
   if (cinfo->global_state != CSTATE_START)
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
@@ -48,7 +51,7 @@
   /* Perform master selection of active modules */
   transencode_master_selection(cinfo, coef_arrays);
   /* Wait for jpeg_finish_compress() call */
-  cinfo->next_scanline = 0;	/* so jpeg_write_marker works */
+  cinfo->next_scanline = 0;     /* so jpeg_write_marker works */
   cinfo->global_state = CSTATE_WRCOEFS;
 }
 
@@ -62,9 +65,9 @@
 
 GLOBAL(void)
 jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
-			       j_compress_ptr dstinfo)
+                               j_compress_ptr dstinfo)
 {
-  JQUANT_TBL ** qtblptr;
+  JQUANT_TBL **qtblptr;
   jpeg_component_info *incomp, *outcomp;
   JQUANT_TBL *c_quant, *slot_quant;
   int tblno, ci, coefi;
@@ -96,10 +99,10 @@
     if (srcinfo->quant_tbl_ptrs[tblno] != NULL) {
       qtblptr = & dstinfo->quant_tbl_ptrs[tblno];
       if (*qtblptr == NULL)
-	*qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
+        *qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
       MEMCOPY((*qtblptr)->quantval,
-	      srcinfo->quant_tbl_ptrs[tblno]->quantval,
-	      SIZEOF((*qtblptr)->quantval));
+              srcinfo->quant_tbl_ptrs[tblno]->quantval,
+              sizeof((*qtblptr)->quantval));
       (*qtblptr)->sent_table = FALSE;
     }
   }
@@ -109,7 +112,7 @@
   dstinfo->num_components = srcinfo->num_components;
   if (dstinfo->num_components < 1 || dstinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(dstinfo, JERR_COMPONENT_COUNT, dstinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
   for (ci = 0, incomp = srcinfo->comp_info, outcomp = dstinfo->comp_info;
        ci < dstinfo->num_components; ci++, incomp++, outcomp++) {
     outcomp->component_id = incomp->component_id;
@@ -122,14 +125,14 @@
      */
     tblno = outcomp->quant_tbl_no;
     if (tblno < 0 || tblno >= NUM_QUANT_TBLS ||
-	srcinfo->quant_tbl_ptrs[tblno] == NULL)
+        srcinfo->quant_tbl_ptrs[tblno] == NULL)
       ERREXIT1(dstinfo, JERR_NO_QUANT_TABLE, tblno);
     slot_quant = srcinfo->quant_tbl_ptrs[tblno];
     c_quant = incomp->quant_table;
     if (c_quant != NULL) {
       for (coefi = 0; coefi < DCTSIZE2; coefi++) {
-	if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
-	  ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
+        if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
+          ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
       }
     }
     /* Note: we do not copy the source's Huffman table assignments;
@@ -163,7 +166,7 @@
 
 LOCAL(void)
 transencode_master_selection (j_compress_ptr cinfo,
-			      jvirt_barray_ptr * coef_arrays)
+                              jvirt_barray_ptr *coef_arrays)
 {
   /* Although we don't actually use input_components for transcoding,
    * jcmaster.c's initial_setup will complain if input_components is 0.
@@ -219,19 +222,19 @@
 typedef struct {
   struct jpeg_c_coef_controller pub; /* public fields */
 
-  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
-  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
 
   /* Virtual block array for each component. */
-  jvirt_barray_ptr * whole_image;
+  jvirt_barray_ptr *whole_image;
 
   /* Workspace for constructing dummy blocks at right/bottom edges. */
   JBLOCKROW dummy_buffer[C_MAX_BLOCKS_IN_MCU];
 } my_coef_controller;
 
-typedef my_coef_controller * my_coef_ptr;
+typedef my_coef_controller *my_coef_ptr;
 
 
 LOCAL(void)
@@ -289,7 +292,7 @@
 compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, ci, xindex, yindex, yoffset, blockcnt;
@@ -312,44 +315,44 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						: compptr->last_col_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (coef->iMCU_row_num < last_iMCU_row ||
-	      yindex+yoffset < compptr->last_row_height) {
-	    /* Fill in pointers to real blocks in this row */
-	    buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	    for (xindex = 0; xindex < blockcnt; xindex++)
-	      MCU_buffer[blkn++] = buffer_ptr++;
-	  } else {
-	    /* At bottom of image, need a whole row of dummy blocks */
-	    xindex = 0;
-	  }
-	  /* Fill in any dummy blocks needed in this row.
-	   * Dummy blocks are filled in the same way as in jccoefct.c:
-	   * all zeroes in the AC entries, DC entries equal to previous
-	   * block's DC value.  The init routine has already zeroed the
-	   * AC entries, so we need only set the DC entries correctly.
-	   */
-	  for (; xindex < compptr->MCU_width; xindex++) {
-	    MCU_buffer[blkn] = coef->dummy_buffer[blkn];
-	    MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
-	    blkn++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                : compptr->last_col_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yindex+yoffset < compptr->last_row_height) {
+            /* Fill in pointers to real blocks in this row */
+            buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+            for (xindex = 0; xindex < blockcnt; xindex++)
+              MCU_buffer[blkn++] = buffer_ptr++;
+          } else {
+            /* At bottom of image, need a whole row of dummy blocks */
+            xindex = 0;
+          }
+          /* Fill in any dummy blocks needed in this row.
+           * Dummy blocks are filled in the same way as in jccoefct.c:
+           * all zeroes in the AC entries, DC entries equal to previous
+           * block's DC value.  The init routine has already zeroed the
+           * AC entries, so we need only set the DC entries correctly.
+           */
+          for (; xindex < compptr->MCU_width; xindex++) {
+            MCU_buffer[blkn] = coef->dummy_buffer[blkn];
+            MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
+            blkn++;
+          }
+        }
       }
       /* Try to write the MCU. */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -372,7 +375,7 @@
 
 LOCAL(void)
 transencode_coef_controller (j_compress_ptr cinfo,
-			     jvirt_barray_ptr * coef_arrays)
+                             jvirt_barray_ptr *coef_arrays)
 {
   my_coef_ptr coef;
   JBLOCKROW buffer;
@@ -380,7 +383,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_c_coef_controller *) coef;
   coef->pub.start_pass = start_pass_coef;
   coef->pub.compress_data = compress_output;
@@ -391,8 +394,8 @@
   /* Allocate and pre-zero space for dummy DCT blocks. */
   buffer = (JBLOCKROW)
     (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
-  jzero_far((void FAR *) buffer, C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+  jzero_far((void *) buffer, C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
   for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
     coef->dummy_buffer[i] = buffer + i;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdapimin.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdapimin.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdapimin.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdapimin.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdapimin.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the decompression half
  * of the JPEG library.  These are the "minimum" API routines that may be
@@ -19,6 +22,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
+#include "jdmaster.h"
 
 
 /*
@@ -32,12 +36,12 @@
   int i;
 
   /* Guard against version mismatches between library and caller. */
-  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
   if (version != JPEG_LIB_VERSION)
     ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
-  if (structsize != SIZEOF(struct jpeg_decompress_struct))
-    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
-	     (int) SIZEOF(struct jpeg_decompress_struct), (int) structsize);
+  if (structsize != sizeof(struct jpeg_decompress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int) sizeof(struct jpeg_decompress_struct), (int) structsize);
 
   /* For debugging purposes, we zero the whole master structure.
    * But the application has already set the err pointer, and may have set
@@ -48,7 +52,7 @@
   {
     struct jpeg_error_mgr * err = cinfo->err;
     void * client_data = cinfo->client_data; /* ignore Purify complaint here */
-    MEMZERO(cinfo, SIZEOF(struct jpeg_decompress_struct));
+    MEMZERO(cinfo, sizeof(struct jpeg_decompress_struct));
     cinfo->err = err;
     cinfo->client_data = client_data;
   }
@@ -80,6 +84,14 @@
 
   /* OK, I'm ready */
   cinfo->global_state = DSTATE_START;
+
+  /* The master struct is used to store extension parameters, so we allocate it
+   * here.
+   */
+  cinfo->master = (struct jpeg_decomp_master *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_decomp_master));
+  MEMZERO(cinfo->master, sizeof(my_decomp_master));
 }
 
 
@@ -121,22 +133,22 @@
     cinfo->jpeg_color_space = JCS_GRAYSCALE;
     cinfo->out_color_space = JCS_GRAYSCALE;
     break;
-    
+
   case 3:
     if (cinfo->saw_JFIF_marker) {
       cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */
     } else if (cinfo->saw_Adobe_marker) {
       switch (cinfo->Adobe_transform) {
       case 0:
-	cinfo->jpeg_color_space = JCS_RGB;
-	break;
+        cinfo->jpeg_color_space = JCS_RGB;
+        break;
       case 1:
-	cinfo->jpeg_color_space = JCS_YCbCr;
-	break;
+        cinfo->jpeg_color_space = JCS_YCbCr;
+        break;
       default:
-	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
-	break;
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+        break;
       }
     } else {
       /* Saw no special markers, try to guess from the component IDs */
@@ -145,31 +157,31 @@
       int cid2 = cinfo->comp_info[2].component_id;
 
       if (cid0 == 1 && cid1 == 2 && cid2 == 3)
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
       else if (cid0 == 82 && cid1 == 71 && cid2 == 66)
-	cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
+        cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
       else {
-	TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+        TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
       }
     }
     /* Always guess RGB is proper output colorspace. */
     cinfo->out_color_space = JCS_RGB;
     break;
-    
+
   case 4:
     if (cinfo->saw_Adobe_marker) {
       switch (cinfo->Adobe_transform) {
       case 0:
-	cinfo->jpeg_color_space = JCS_CMYK;
-	break;
+        cinfo->jpeg_color_space = JCS_CMYK;
+        break;
       case 2:
-	cinfo->jpeg_color_space = JCS_YCCK;
-	break;
+        cinfo->jpeg_color_space = JCS_YCCK;
+        break;
       default:
-	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
-	cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
-	break;
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
+        break;
       }
     } else {
       /* No special markers, assume straight CMYK. */
@@ -177,7 +189,7 @@
     }
     cinfo->out_color_space = JCS_CMYK;
     break;
-    
+
   default:
     cinfo->jpeg_color_space = JCS_UNKNOWN;
     cinfo->out_color_space = JCS_UNKNOWN;
@@ -185,7 +197,7 @@
   }
 
   /* Set defaults for other decompression parameters. */
-  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_num = 1;         /* 1:1 scaling */
   cinfo->scale_denom = 1;
   cinfo->output_gamma = 1.0;
   cinfo->buffered_image = FALSE;
@@ -253,7 +265,7 @@
     retcode = JPEG_HEADER_OK;
     break;
   case JPEG_REACHED_EOI:
-    if (require_image)		/* Complain if application wanted an image */
+    if (require_image)          /* Complain if application wanted an image */
       ERREXIT(cinfo, JERR_NO_IMAGE);
     /* Reset to start state; it would be safer to require the application to
      * call jpeg_abort, but we can't change it now for compatibility reasons.
@@ -385,7 +397,7 @@
   /* Read until EOI */
   while (! cinfo->inputctl->eoi_reached) {
     if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
-      return FALSE;		/* Suspend, come back later */
+      return FALSE;             /* Suspend, come back later */
   }
   /* Do final cleanup */
   (*cinfo->src->term_source) (cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdapistd.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdapistd.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdapistd.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdapistd.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jdapistd.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the decompression half
  * of the JPEG library.  These are the "standard" API routines that are
@@ -15,14 +18,14 @@
  * whole decompression library into a transcoder.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
-#include "jpegcomp.h"
-
+#include "jdmainct.h"
+#include "jdcoefct.h"
+#include "jdsample.h"
+#include "jmemsys.h"
 
 /* Forward declarations */
-LOCAL(boolean) output_pass_setup JPP((j_decompress_ptr cinfo));
+LOCAL(boolean) output_pass_setup (j_decompress_ptr cinfo);
 
 
 /*
@@ -54,24 +57,24 @@
     if (cinfo->inputctl->has_multiple_scans) {
 #ifdef D_MULTISCAN_FILES_SUPPORTED
       for (;;) {
-	int retcode;
-	/* Call progress monitor hook if present */
-	if (cinfo->progress != NULL)
-	  (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
-	/* Absorb some more input */
-	retcode = (*cinfo->inputctl->consume_input) (cinfo);
-	if (retcode == JPEG_SUSPENDED)
-	  return FALSE;
-	if (retcode == JPEG_REACHED_EOI)
-	  break;
-	/* Advance progress counter if appropriate */
-	if (cinfo->progress != NULL &&
-	    (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
-	  if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
-	    /* jdmaster underestimated number of scans; ratchet up one scan */
-	    cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
-	  }
-	}
+        int retcode;
+        /* Call progress monitor hook if present */
+        if (cinfo->progress != NULL)
+          (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        /* Absorb some more input */
+        retcode = (*cinfo->inputctl->consume_input) (cinfo);
+        if (retcode == JPEG_SUSPENDED)
+          return FALSE;
+        if (retcode == JPEG_REACHED_EOI)
+          break;
+        /* Advance progress counter if appropriate */
+        if (cinfo->progress != NULL &&
+            (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+          if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+            /* jdmaster underestimated number of scans; ratchet up one scan */
+            cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+          }
+        }
       }
 #else
       ERREXIT(cinfo, JERR_NOT_COMPILED);
@@ -110,16 +113,16 @@
       JDIMENSION last_scanline;
       /* Call progress monitor hook if present */
       if (cinfo->progress != NULL) {
-	cinfo->progress->pass_counter = (long) cinfo->output_scanline;
-	cinfo->progress->pass_limit = (long) cinfo->output_height;
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+        cinfo->progress->pass_limit = (long) cinfo->output_height;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       }
       /* Process some data */
       last_scanline = cinfo->output_scanline;
       (*cinfo->main->process_data) (cinfo, (JSAMPARRAY) NULL,
-				    &cinfo->output_scanline, (JDIMENSION) 0);
+                                    &cinfo->output_scanline, (JDIMENSION) 0);
       if (cinfo->output_scanline == last_scanline)
-	return FALSE;		/* No progress made, must suspend */
+        return FALSE;           /* No progress made, must suspend */
     }
     /* Finish up dummy pass, and set up for another one */
     (*cinfo->master->finish_output_pass) (cinfo);
@@ -138,6 +141,110 @@
 
 
 /*
+ * Enable partial scanline decompression
+ *
+ * Must be called after jpeg_start_decompress() and before any calls to
+ * jpeg_read_scanlines() or jpeg_skip_scanlines().
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(void)
+jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                    JDIMENSION *width)
+{
+  int ci, align, orig_downsampled_width;
+  JDIMENSION input_xoffset;
+  boolean reinit_upsampler = FALSE;
+  jpeg_component_info *compptr;
+
+  if (cinfo->global_state != DSTATE_SCANNING || cinfo->output_scanline != 0)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (!xoffset || !width)
+    ERREXIT(cinfo, JERR_BAD_CROP_SPEC);
+
+  /* xoffset and width must fall within the output image dimensions. */
+  if (*width == 0 || *xoffset + *width > cinfo->output_width)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* No need to do anything if the caller wants the entire width. */
+  if (*width == cinfo->output_width)
+    return;
+
+  /* Ensuring the proper alignment of xoffset is tricky.  At minimum, it
+   * must align with an MCU boundary, because:
+   *
+   *   (1) The IDCT is performed in blocks, and it is not feasible to modify
+   *       the algorithm so that it can transform partial blocks.
+   *   (2) Because of the SIMD extensions, any input buffer passed to the
+   *       upsampling and color conversion routines must be aligned to the
+   *       SIMD word size (for instance, 128-bit in the case of SSE2.)  The
+   *       easiest way to accomplish this without copying data is to ensure
+   *       that upsampling and color conversion begin at the start of the
+   *       first MCU column that will be inverse transformed.
+   *
+   * In practice, we actually impose a stricter alignment requirement.  We
+   * require that xoffset be a multiple of the maximum MCU column width of all
+   * of the components (the "iMCU column width.")  This is to simplify the
+   * single-pass decompression case, allowing us to use the same MCU column
+   * width for all of the components.
+   */
+  align = cinfo->_min_DCT_scaled_size * cinfo->max_h_samp_factor;
+
+  /* Adjust xoffset to the nearest iMCU boundary <= the requested value */
+  input_xoffset = *xoffset;
+  *xoffset = (input_xoffset / align) * align;
+
+  /* Adjust the width so that the right edge of the output image is as
+   * requested (only the left edge is altered.)  It is important that calling
+   * programs check this value after this function returns, so that they can
+   * allocate an output buffer with the appropriate size.
+   */
+  *width = *width + input_xoffset - *xoffset;
+  cinfo->output_width = *width;
+
+  /* Set the first and last iMCU columns that we must decompress.  These values
+   * will be used in single-scan decompressions.
+   */
+  cinfo->master->first_iMCU_col =
+    (JDIMENSION) (long) (*xoffset) / (long) align;
+  cinfo->master->last_iMCU_col =
+    (JDIMENSION) jdiv_round_up((long) (*xoffset + cinfo->output_width),
+                               (long) align) - 1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Set downsampled_width to the new output width. */
+    orig_downsampled_width = compptr->downsampled_width;
+    compptr->downsampled_width =
+      (JDIMENSION) jdiv_round_up((long) (cinfo->output_width *
+                                         compptr->h_samp_factor),
+                                 (long) cinfo->max_h_samp_factor);
+    if (compptr->downsampled_width < 2 && orig_downsampled_width >= 2)
+      reinit_upsampler = TRUE;
+
+    /* Set the first and last iMCU columns that we must decompress.  These
+     * values will be used in multi-scan decompressions.
+     */
+    cinfo->master->first_MCU_col[ci] =
+      (JDIMENSION) (long) (*xoffset * compptr->h_samp_factor) /
+                   (long) align;
+    cinfo->master->last_MCU_col[ci] =
+      (JDIMENSION) jdiv_round_up((long) ((*xoffset + cinfo->output_width) *
+                                         compptr->h_samp_factor),
+                                 (long) align) - 1;
+  }
+
+  if (reinit_upsampler) {
+    cinfo->master->jinit_upsampler_no_alloc = TRUE;
+    jinit_upsampler(cinfo);
+    cinfo->master->jinit_upsampler_no_alloc = FALSE;
+  }
+}
+
+
+/*
  * Read some scanlines of data from the JPEG decompressor.
  *
  * The return value will be the number of lines actually read.
@@ -152,7 +259,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_read_scanlines (j_decompress_ptr cinfo, JSAMPARRAY scanlines,
-		     JDIMENSION max_lines)
+                     JDIMENSION max_lines)
 {
   JDIMENSION row_ctr;
 
@@ -178,6 +285,236 @@
 }
 
 
+/* Dummy color convert function used by jpeg_skip_scanlines() */
+LOCAL(void)
+noop_convert (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+              JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+
+/*
+ * In some cases, it is best to call jpeg_read_scanlines() and discard the
+ * output, rather than skipping the scanlines, because this allows us to
+ * maintain the internal state of the context-based upsampler.  In these cases,
+ * we set up and tear down a dummy color converter in order to avoid valgrind
+ * errors and to achieve the best possible performance.
+ */
+
+LOCAL(void)
+read_and_discard_scanlines (j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  JDIMENSION n;
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows);
+
+  color_convert = cinfo->cconvert->color_convert;
+  cinfo->cconvert->color_convert = noop_convert;
+
+  for (n = 0; n < num_lines; n++)
+    jpeg_read_scanlines(cinfo, NULL, 1);
+
+  cinfo->cconvert->color_convert = color_convert;
+}
+
+
+/*
+ * Called by jpeg_skip_scanlines().  This partially skips a decompress block by
+ * incrementing the rowgroup counter.
+ */
+
+LOCAL(void)
+increment_simple_rowgroup_ctr (j_decompress_ptr cinfo, JDIMENSION rows)
+{
+  JDIMENSION rows_left;
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  /* Increment the counter to the next row group after the skipped rows. */
+  main_ptr->rowgroup_ctr += rows / cinfo->max_v_samp_factor;
+
+  /* Partially skipping a row group would involve modifying the internal state
+   * of the upsampler, so read the remaining rows into a dummy buffer instead.
+   */
+  rows_left = rows % cinfo->max_v_samp_factor;
+  cinfo->output_scanline += rows - rows_left;
+
+  read_and_discard_scanlines(cinfo, rows_left);
+}
+
+/*
+ * Skips some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually skipped.  If skipping
+ * num_lines would move beyond the end of the image, then the actual number of
+ * lines remaining in the image is returned.  Otherwise, the return value will
+ * be equal to num_lines.
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_skip_scanlines (j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JDIMENSION i, x;
+  int y;
+  JDIMENSION lines_per_iMCU_row, lines_left_in_iMCU_row, lines_after_iMCU_row;
+  JDIMENSION lines_to_skip, lines_to_read;
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Do not skip past the bottom of the image. */
+  if (cinfo->output_scanline + num_lines >= cinfo->output_height) {
+    cinfo->output_scanline = cinfo->output_height;
+    return cinfo->output_height - cinfo->output_scanline;
+  }
+
+  if (num_lines == 0)
+    return 0;
+
+  lines_per_iMCU_row = cinfo->_min_DCT_scaled_size * cinfo->max_v_samp_factor;
+  lines_left_in_iMCU_row =
+    (lines_per_iMCU_row - (cinfo->output_scanline % lines_per_iMCU_row)) %
+    lines_per_iMCU_row;
+  lines_after_iMCU_row = num_lines - lines_left_in_iMCU_row;
+
+  /* Skip the lines remaining in the current iMCU row.  When upsampling
+   * requires context rows, we need the previous and next rows in order to read
+   * the current row.  This adds some complexity.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    /* If the skipped lines would not move us past the current iMCU row, we
+     * read the lines and ignore them.  There might be a faster way of doing
+     * this, but we are facing increasing complexity for diminishing returns.
+     * The increasing complexity would be a by-product of meddling with the
+     * state machine used to skip context rows.  Near the end of an iMCU row,
+     * the next iMCU row may have already been entropy-decoded.  In this unique
+     * case, we will read the next iMCU row if we cannot skip past it as well.
+     */
+    if ((num_lines < lines_left_in_iMCU_row + 1) ||
+        (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full &&
+         lines_after_iMCU_row < lines_per_iMCU_row + 1)) {
+      read_and_discard_scanlines(cinfo, num_lines);
+      return num_lines;
+    }
+
+    /* If the next iMCU row has already been entropy-decoded, make sure that
+     * we do not skip too far.
+     */
+    if (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full) {
+      cinfo->output_scanline += lines_left_in_iMCU_row + lines_per_iMCU_row;
+      lines_after_iMCU_row -= lines_per_iMCU_row;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+    }
+
+    /* If we have just completed the first block, adjust the buffer pointers */
+    if (main_ptr->iMCU_row_ctr == 0 ||
+        (main_ptr->iMCU_row_ctr == 1 && lines_left_in_iMCU_row > 2))
+      set_wraparound_pointers(cinfo);
+    main_ptr->buffer_full = FALSE;
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+    upsample->next_row_out = cinfo->max_v_samp_factor;
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+  }
+
+  /* Skipping is much simpler when context rows are not required. */
+  else {
+    if (num_lines < lines_left_in_iMCU_row) {
+      increment_simple_rowgroup_ctr(cinfo, num_lines);
+      return num_lines;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+      main_ptr->buffer_full = FALSE;
+      main_ptr->rowgroup_ctr = 0;
+      upsample->next_row_out = cinfo->max_v_samp_factor;
+      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    }
+  }
+
+  /* Calculate how many full iMCU rows we can skip. */
+  if (cinfo->upsample->need_context_rows)
+    lines_to_skip = ((lines_after_iMCU_row - 1) / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  else
+    lines_to_skip = (lines_after_iMCU_row / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  /* Calculate the number of lines that remain to be skipped after skipping all
+   * of the full iMCU rows that we can.  We will not read these lines unless we
+   * have to.
+   */
+  lines_to_read = lines_after_iMCU_row - lines_to_skip;
+
+  /* For images requiring multiple scans (progressive, non-interleaved, etc.),
+   * all of the entropy decoding occurs in jpeg_start_decompress(), assuming
+   * that the input data source is non-suspending.  This makes skipping easy.
+   */
+  if (cinfo->inputctl->has_multiple_scans) {
+    if (cinfo->upsample->need_context_rows) {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      main_ptr->iMCU_row_ctr += lines_after_iMCU_row / lines_per_iMCU_row;
+      /* It is complex to properly move to the middle of a context block, so
+       * read the remaining lines instead of skipping them.
+       */
+      read_and_discard_scanlines(cinfo, lines_to_read);
+    } else {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+    }
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    return num_lines;
+  }
+
+  /* Skip the iMCU rows that we can safely skip. */
+  for (i = 0; i < lines_to_skip; i += lines_per_iMCU_row) {
+    for (y = 0; y < coef->MCU_rows_per_iMCU_row; y++) {
+      for (x = 0; x < cinfo->MCUs_per_row; x++) {
+        /* Calling decode_mcu() with a NULL pointer causes it to discard the
+         * decoded coefficients.  This is ~5% faster for large subsets, but
+         * it's tough to tell a difference for smaller images.
+         */
+        (*cinfo->entropy->decode_mcu) (cinfo, NULL);
+      }
+    }
+    cinfo->input_iMCU_row++;
+    cinfo->output_iMCU_row++;
+    if (cinfo->input_iMCU_row < cinfo->total_iMCU_rows)
+      start_iMCU_row(cinfo);
+    else
+      (*cinfo->inputctl->finish_input_pass) (cinfo);
+  }
+  cinfo->output_scanline += lines_to_skip;
+
+  if (cinfo->upsample->need_context_rows) {
+    /* Context-based upsampling keeps track of iMCU rows. */
+    main_ptr->iMCU_row_ctr += lines_to_skip / lines_per_iMCU_row;
+
+    /* It is complex to properly move to the middle of a context block, so
+     * read the remaining lines instead of skipping them.
+     */
+    read_and_discard_scanlines(cinfo, lines_to_read);
+  } else {
+    increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+  }
+
+  /* Since skipping lines involves skipping the upsampling step, the value of
+   * "rows_to_go" will become invalid unless we set it here.  NOTE: This is a
+   * bit odd, since "rows_to_go" seems to be redundantly keeping track of
+   * output_scanline.
+   */
+  upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+
+  /* Always skip the requested number of lines. */
+  return num_lines;
+}
+
 /*
  * Alternate entry point to read raw data.
  * Processes exactly one iMCU row per call, unless suspended.
@@ -185,7 +522,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
-		    JDIMENSION max_lines)
+                    JDIMENSION max_lines)
 {
   JDIMENSION lines_per_iMCU_row;
 
@@ -210,7 +547,7 @@
 
   /* Decompress directly into user's buffer. */
   if (! (*cinfo->coef->decompress_data) (cinfo, data))
-    return 0;			/* suspension forced, can do nothing more */
+    return 0;                   /* suspension forced, can do nothing more */
 
   /* OK, we processed one iMCU row. */
   cinfo->output_scanline += lines_per_iMCU_row;
@@ -266,9 +603,9 @@
   }
   /* Read markers looking for SOS or EOI */
   while (cinfo->input_scan_number <= cinfo->output_scan_number &&
-	 ! cinfo->inputctl->eoi_reached) {
+         ! cinfo->inputctl->eoi_reached) {
     if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
-      return FALSE;		/* Suspend, come back later */
+      return FALSE;             /* Suspend, come back later */
   }
   cinfo->global_state = DSTATE_BUFIMAGE;
   return TRUE;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdarith.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdarith.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdarith.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdarith.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdarith.c
  *
- * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * This file was part of the Independent JPEG Group's software:
+ * Developed 1997-2015 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains portable arithmetic entropy decoding routines for JPEG
  * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
@@ -23,8 +26,8 @@
 typedef struct {
   struct jpeg_entropy_decoder pub; /* public fields */
 
-  INT32 c;       /* C register, base of coding interval + input bit buffer */
-  INT32 a;               /* A register, normalized size of coding interval */
+  JLONG c;       /* C register, base of coding interval + input bit buffer */
+  JLONG a;               /* A register, normalized size of coding interval */
   int ct;     /* bit shift counter, # of bits left in bit buffer part of C */
                                                          /* init: ct = -16 */
                                                          /* run: ct = 0..7 */
@@ -32,17 +35,17 @@
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
   int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to statistics areas (these workspaces have image lifespan) */
-  unsigned char * dc_stats[NUM_ARITH_TBLS];
-  unsigned char * ac_stats[NUM_ARITH_TBLS];
+  unsigned char *dc_stats[NUM_ARITH_TBLS];
+  unsigned char *ac_stats[NUM_ARITH_TBLS];
 
   /* Statistics bin for coding with fixed probability 0.5 */
   unsigned char fixed_bin[4];
 } arith_entropy_decoder;
 
-typedef arith_entropy_decoder * arith_entropy_ptr;
+typedef arith_entropy_decoder *arith_entropy_ptr;
 
 /* The following two definitions specify the allocation chunk size
  * for the statistics area.
@@ -65,7 +68,7 @@
 get_byte (j_decompress_ptr cinfo)
 /* Read next input byte; we do not support suspension in this module. */
 {
-  struct jpeg_source_mgr * src = cinfo->src;
+  struct jpeg_source_mgr *src = cinfo->src;
 
   if (src->bytes_in_buffer == 0)
     if (! (*src->fill_input_buffer) (cinfo))
@@ -94,7 +97,7 @@
  * (instead of fixed) with the bit shift counter CT.
  * Thus, we also need only one (variable instead of
  * fixed size) shift for the LPS/MPS decision, and
- * we can get away with any renormalization update
+ * we can do away with any renormalization update
  * of C (except for new data insertion, of course).
  *
  * I've also introduced a new scheme for accessing
@@ -107,7 +110,7 @@
 {
   register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
   register unsigned char nl, nm;
-  register INT32 qe, temp;
+  register JLONG qe, temp;
   register int sv, data;
 
   /* Renormalization & data input per section D.2.6 */
@@ -115,32 +118,32 @@
     if (--e->ct < 0) {
       /* Need to fetch next data byte */
       if (cinfo->unread_marker)
-	data = 0;		/* stuff zero data */
+        data = 0;               /* stuff zero data */
       else {
-	data = get_byte(cinfo);	/* read next input byte */
-	if (data == 0xFF) {	/* zero stuff or marker code */
-	  do data = get_byte(cinfo);
-	  while (data == 0xFF);	/* swallow extra 0xFF bytes */
-	  if (data == 0)
-	    data = 0xFF;	/* discard stuffed zero byte */
-	  else {
-	    /* Note: Different from the Huffman decoder, hitting
-	     * a marker while processing the compressed data
-	     * segment is legal in arithmetic coding.
-	     * The convention is to supply zero data
-	     * then until decoding is complete.
-	     */
-	    cinfo->unread_marker = data;
-	    data = 0;
-	  }
-	}
+        data = get_byte(cinfo); /* read next input byte */
+        if (data == 0xFF) {     /* zero stuff or marker code */
+          do data = get_byte(cinfo);
+          while (data == 0xFF); /* swallow extra 0xFF bytes */
+          if (data == 0)
+            data = 0xFF;        /* discard stuffed zero byte */
+          else {
+            /* Note: Different from the Huffman decoder, hitting
+             * a marker while processing the compressed data
+             * segment is legal in arithmetic coding.
+             * The convention is to supply zero data
+             * then until decoding is complete.
+             */
+            cinfo->unread_marker = data;
+            data = 0;
+          }
+        }
       }
       e->c = (e->c << 8) | data; /* insert data into C register */
-      if ((e->ct += 8) < 0)	 /* update bit shift counter */
-	/* Need more initial bytes */
-	if (++e->ct == 0)
-	  /* Got 2 initial bytes -> re-init A and exit loop */
-	  e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
+      if ((e->ct += 8) < 0)      /* update bit shift counter */
+        /* Need more initial bytes */
+        if (++e->ct == 0)
+          /* Got 2 initial bytes -> re-init A and exit loop */
+          e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
     }
     e->a <<= 1;
   }
@@ -149,9 +152,9 @@
    * Qe values and probability estimation state machine
    */
   sv = *st;
-  qe = jpeg_aritab[sv & 0x7F];	/* => Qe_Value */
-  nl = qe & 0xFF; qe >>= 8;	/* Next_Index_LPS + Switch_MPS */
-  nm = qe & 0xFF; qe >>= 8;	/* Next_Index_MPS */
+  qe = jpeg_aritab[sv & 0x7F];  /* => Qe_Value */
+  nl = qe & 0xFF; qe >>= 8;     /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF; qe >>= 8;     /* Next_Index_MPS */
 
   /* Decode & estimation procedures per sections D.2.4 & D.2.5 */
   temp = e->a - qe;
@@ -162,19 +165,19 @@
     /* Conditional LPS (less probable symbol) exchange */
     if (e->a < qe) {
       e->a = qe;
-      *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
     } else {
       e->a = qe;
-      *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
-      sv ^= 0x80;		/* Exchange LPS/MPS */
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
     }
   } else if (e->a < 0x8000L) {
     /* Conditional MPS (more probable symbol) exchange */
     if (e->a < qe) {
-      *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
-      sv ^= 0x80;		/* Exchange LPS/MPS */
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
     } else {
-      *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
     }
   }
 
@@ -191,7 +194,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Advance past the RSTn marker */
   if (! (*cinfo->marker->read_restart_marker) (cinfo))
@@ -200,13 +203,13 @@
   /* Re-initialize statistics areas */
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
-    if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       MEMZERO(entropy->dc_stats[compptr->dc_tbl_no], DC_STAT_BINS);
       /* Reset DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
       entropy->dc_context[ci] = 0;
     }
-    if (! cinfo->progressive_mode || cinfo->Ss) {
+    if (!cinfo->progressive_mode || cinfo->Ss) {
       MEMZERO(entropy->ac_stats[compptr->ac_tbl_no], AC_STAT_BINS);
     }
   }
@@ -214,7 +217,7 @@
   /* Reset arithmetic decoding variables */
   entropy->c = 0;
   entropy->a = 0;
-  entropy->ct = -16;	/* force reading 2 initial bytes to fill C */
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
 
   /* Reset restart counter */
   entropy->restarts_to_go = cinfo->restart_interval;
@@ -253,7 +256,7 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* Outer loop handles each block in the MCU */
 
@@ -277,34 +280,34 @@
       st += 2; st += sign;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        st = entropy->dc_stats[tbl] + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;		   /* zero diff category */
+        entropy->dc_context[ci] = 0;               /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+        entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
       else
-	entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
+        entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
       entropy->last_dc_val[ci] += v;
     }
 
     /* Scale and output the DC coefficient (assumes jpeg_natural_order[0]=0) */
-    (*block)[0] = (JCOEF) (entropy->last_dc_val[ci] << cinfo->Al);
+    (*block)[0] = (JCOEF) LEFT_SHIFT(entropy->last_dc_val[ci], cinfo->Al);
   }
 
   return TRUE;
@@ -332,7 +335,7 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* There is always only one block per MCU */
   block = MCU_data[0];
@@ -343,13 +346,13 @@
   /* Figure F.20: Decode_AC_coefficients */
   for (k = cinfo->Ss; k <= cinfo->Se; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
-    if (arith_decode(cinfo, st)) break;		/* EOB flag */
+    if (arith_decode(cinfo, st)) break;         /* EOB flag */
     while (arith_decode(cinfo, st + 1) == 0) {
       st += 3; k++;
       if (k > cinfo->Se) {
-	WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	entropy->ct = -1;			/* spectral overflow */
-	return TRUE;
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
       }
     }
     /* Figure F.21: Decoding nonzero value v */
@@ -359,17 +362,17 @@
     /* Figure F.23: Decoding the magnitude category of v */
     if ((m = arith_decode(cinfo, st)) != 0) {
       if (arith_decode(cinfo, st)) {
-	m <<= 1;
-	st = entropy->ac_stats[tbl] +
-	     (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        m <<= 1;
+        st = entropy->ac_stats[tbl] +
+             (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
     }
     v = m;
@@ -404,8 +407,8 @@
     entropy->restarts_to_go--;
   }
 
-  st = entropy->fixed_bin;	/* use fixed probability estimation */
-  p1 = 1 << cinfo->Al;		/* 1 in the bit position being coded */
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
 
   /* Outer loop handles each block in the MCU */
 
@@ -440,14 +443,14 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* There is always only one block per MCU */
   block = MCU_data[0];
   tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
 
-  p1 = 1 << cinfo->Al;		/* 1 in the bit position being coded */
-  m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
+  m1 = (-1) << cinfo->Al;       /* -1 in the bit position being coded */
 
   /* Establish EOBx (previous stage end-of-block) index */
   for (kex = cinfo->Se; kex > 0; kex--)
@@ -456,30 +459,30 @@
   for (k = cinfo->Ss; k <= cinfo->Se; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
     if (k > kex)
-      if (arith_decode(cinfo, st)) break;	/* EOB flag */
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
     for (;;) {
       thiscoef = *block + jpeg_natural_order[k];
-      if (*thiscoef) {				/* previously nonzero coef */
-	if (arith_decode(cinfo, st + 2)) {
-	  if (*thiscoef < 0)
-	    *thiscoef += m1;
-	  else
-	    *thiscoef += p1;
-	}
-	break;
-      }
-      if (arith_decode(cinfo, st + 1)) {	/* newly nonzero coef */
-	if (arith_decode(cinfo, entropy->fixed_bin))
-	  *thiscoef = m1;
-	else
-	  *thiscoef = p1;
-	break;
+      if (*thiscoef) {                          /* previously nonzero coef */
+        if (arith_decode(cinfo, st + 2)) {
+          if (*thiscoef < 0)
+            *thiscoef += m1;
+          else
+            *thiscoef += p1;
+        }
+        break;
+      }
+      if (arith_decode(cinfo, st + 1)) {        /* newly nonzero coef */
+        if (arith_decode(cinfo, entropy->fixed_bin))
+          *thiscoef = m1;
+        else
+          *thiscoef = p1;
+        break;
       }
       st += 3; k++;
       if (k > cinfo->Se) {
-	WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	entropy->ct = -1;			/* spectral overflow */
-	return TRUE;
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
       }
     }
   }
@@ -496,7 +499,7 @@
 decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JBLOCKROW block;
   unsigned char *st;
   int blkn, ci, tbl, sign, k;
@@ -509,12 +512,12 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* Outer loop handles each block in the MCU */
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    block = MCU_data[blkn];
+    block = MCU_data ? MCU_data[blkn] : NULL;
     ci = cinfo->MCU_membership[blkn];
     compptr = cinfo->cur_comp_info[ci];
 
@@ -535,33 +538,34 @@
       st += 2; st += sign;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        st = entropy->dc_stats[tbl] + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;		   /* zero diff category */
+        entropy->dc_context[ci] = 0;               /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+        entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
       else
-	entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
+        entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
       entropy->last_dc_val[ci] += v;
     }
 
-    (*block)[0] = (JCOEF) entropy->last_dc_val[ci];
+    if (block)
+      (*block)[0] = (JCOEF) entropy->last_dc_val[ci];
 
     /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
 
@@ -570,14 +574,14 @@
     /* Figure F.20: Decode_AC_coefficients */
     for (k = 1; k <= DCTSIZE2 - 1; k++) {
       st = entropy->ac_stats[tbl] + 3 * (k - 1);
-      if (arith_decode(cinfo, st)) break;	/* EOB flag */
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
       while (arith_decode(cinfo, st + 1) == 0) {
-	st += 3; k++;
-	if (k > DCTSIZE2 - 1) {
-	  WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	  entropy->ct = -1;			/* spectral overflow */
-	  return TRUE;
-	}
+        st += 3; k++;
+        if (k > DCTSIZE2 - 1) {
+          WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+          entropy->ct = -1;                     /* spectral overflow */
+          return TRUE;
+        }
       }
       /* Figure F.21: Decoding nonzero value v */
       /* Figure F.22: Decoding the sign of v */
@@ -585,27 +589,28 @@
       st += 2;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	if (arith_decode(cinfo, st)) {
-	  m <<= 1;
-	  st = entropy->ac_stats[tbl] +
-	       (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	  while (arith_decode(cinfo, st)) {
-	    if ((m <<= 1) == 0x8000) {
-	      WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	      entropy->ct = -1;			/* magnitude overflow */
-	      return TRUE;
-	    }
-	    st += 1;
-	  }
-	}
+        if (arith_decode(cinfo, st)) {
+          m <<= 1;
+          st = entropy->ac_stats[tbl] +
+               (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+          while (arith_decode(cinfo, st)) {
+            if ((m <<= 1) == 0x8000) {
+              WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+              entropy->ct = -1;                 /* magnitude overflow */
+              return TRUE;
+            }
+            st += 1;
+          }
+        }
       }
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
-      (*block)[jpeg_natural_order[k]] = (JCOEF) v;
+      if (block)
+        (*block)[jpeg_natural_order[k]] = (JCOEF) v;
     }
   }
 
@@ -622,30 +627,30 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (cinfo->progressive_mode) {
     /* Validate progressive scan parameters */
     if (cinfo->Ss == 0) {
       if (cinfo->Se != 0)
-	goto bad;
+        goto bad;
     } else {
       /* need not check Ss/Se < 0 since they came from unsigned bytes */
       if (cinfo->Se < cinfo->Ss || cinfo->Se > DCTSIZE2 - 1)
-	goto bad;
+        goto bad;
       /* AC scans may have only one component */
       if (cinfo->comps_in_scan != 1)
-	goto bad;
+        goto bad;
     }
     if (cinfo->Ah != 0) {
       /* Successive approximation refinement scan: must have Al = Ah-1. */
       if (cinfo->Ah-1 != cinfo->Al)
-	goto bad;
+        goto bad;
     }
-    if (cinfo->Al > 13) {	/* need not check for < 0 */
+    if (cinfo->Al > 13) {       /* need not check for < 0 */
       bad:
       ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
-	       cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+               cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
     }
     /* Update progression status, and verify that scan order is legal.
      * Note that inter-scan inconsistencies are treated as warnings
@@ -655,32 +660,32 @@
       int coefi, cindex = cinfo->cur_comp_info[ci]->component_index;
       int *coef_bit_ptr = & cinfo->coef_bits[cindex][0];
       if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
-	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
       for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
-	int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
-	if (cinfo->Ah != expected)
-	  WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
-	coef_bit_ptr[coefi] = cinfo->Al;
+        int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
+        if (cinfo->Ah != expected)
+          WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+        coef_bit_ptr[coefi] = cinfo->Al;
       }
     }
     /* Select MCU decoding routine */
     if (cinfo->Ah == 0) {
       if (cinfo->Ss == 0)
-	entropy->pub.decode_mcu = decode_mcu_DC_first;
+        entropy->pub.decode_mcu = decode_mcu_DC_first;
       else
-	entropy->pub.decode_mcu = decode_mcu_AC_first;
+        entropy->pub.decode_mcu = decode_mcu_AC_first;
     } else {
       if (cinfo->Ss == 0)
-	entropy->pub.decode_mcu = decode_mcu_DC_refine;
+        entropy->pub.decode_mcu = decode_mcu_DC_refine;
       else
-	entropy->pub.decode_mcu = decode_mcu_AC_refine;
+        entropy->pub.decode_mcu = decode_mcu_AC_refine;
     }
   } else {
     /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
      * This ought to be an error condition, but we make it a warning.
      */
     if (cinfo->Ss != 0 || cinfo->Ah != 0 || cinfo->Al != 0 ||
-	(cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1))
+        (cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1))
       WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
     /* Select MCU decoding routine */
     entropy->pub.decode_mcu = decode_mcu;
@@ -689,25 +694,25 @@
   /* Allocate & initialize requested statistics areas */
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
-    if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       tbl = compptr->dc_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->dc_stats[tbl] == NULL)
-	entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+        entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
       MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
       /* Initialize DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
       entropy->dc_context[ci] = 0;
     }
-    if (! cinfo->progressive_mode || cinfo->Ss) {
+    if (!cinfo->progressive_mode || cinfo->Ss) {
       tbl = compptr->ac_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->ac_stats[tbl] == NULL)
-	entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+        entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
       MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
     }
   }
@@ -715,7 +720,7 @@
   /* Initialize arithmetic decoding variables */
   entropy->c = 0;
   entropy->a = 0;
-  entropy->ct = -16;	/* force reading 2 initial bytes to fill C */
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
 
   /* Initialize restart counter */
   entropy->restarts_to_go = cinfo->restart_interval;
@@ -734,7 +739,7 @@
 
   entropy = (arith_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(arith_entropy_decoder));
+                                sizeof(arith_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass;
 
@@ -752,10 +757,10 @@
     int *coef_bit_ptr, ci;
     cinfo->coef_bits = (int (*)[DCTSIZE2])
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  cinfo->num_components*DCTSIZE2*SIZEOF(int));
+                                  cinfo->num_components*DCTSIZE2*sizeof(int));
     coef_bit_ptr = & cinfo->coef_bits[0][0];
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       for (i = 0; i < DCTSIZE2; i++)
-	*coef_bit_ptr++ = -1;
+        *coef_bit_ptr++ = -1;
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatadst.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatadst.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatadst.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatadst.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,283 @@
+/*
+ * jdatadst.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to memory or to a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
+#endif
+
+
+/* Expanded data destination object for stdio output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  FILE *outfile;                /* target stream */
+  JOCTET *buffer;               /* start of buffer */
+} my_destination_mgr;
+
+typedef my_destination_mgr *my_dest_ptr;
+
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/* Expanded data destination object for memory output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
+  size_t bufsize;
+} my_mem_destination_mgr;
+
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
+#endif
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  /* Allocate the output buffer --- it will be released when done with image */
+  dest->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  OUTPUT_BUF_SIZE * sizeof(JOCTET));
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_destination (j_compress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_output_buffer (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) !=
+      (size_t) OUTPUT_BUF_SIZE)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+empty_mem_output_buffer (j_compress_ptr cinfo)
+{
+  size_t nextsize;
+  JOCTET *nextbuffer;
+  my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
+
+  /* Try to allocate new buffer with double size */
+  nextsize = dest->bufsize * 2;
+  nextbuffer = (JOCTET *) malloc(nextsize);
+
+  if (nextbuffer == NULL)
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+
+  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);
+
+  if (dest->newbuffer != NULL)
+    free(dest->newbuffer);
+
+  dest->newbuffer = nextbuffer;
+
+  dest->pub.next_output_byte = nextbuffer + dest->bufsize;
+  dest->pub.free_in_buffer = dest->bufsize;
+
+  dest->buffer = nextbuffer;
+  dest->bufsize = nextsize;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+  size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer;
+
+  /* Write any data remaining in the buffer */
+  if (datacount > 0) {
+    if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount)
+      ERREXIT(cinfo, JERR_FILE_WRITE);
+  }
+  fflush(dest->outfile);
+  /* Make sure we wrote the output file OK */
+  if (ferror(dest->outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+term_mem_destination (j_compress_ptr cinfo)
+{
+  my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
+
+  *dest->outbuffer = dest->buffer;
+  *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer);
+}
+#endif
+
+
+/*
+ * Prepare for output to a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing compression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_dest (j_compress_ptr cinfo, FILE *outfile)
+{
+  my_dest_ptr dest;
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same file without re-executing jpeg_stdio_dest.
+   * This makes it dangerous to use this manager and a different destination
+   * manager serially with the same JPEG object, because their private object
+   * sizes may be different.  Caveat programmer.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_destination_mgr));
+  }
+
+  dest = (my_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_destination;
+  dest->pub.empty_output_buffer = empty_output_buffer;
+  dest->pub.term_destination = term_destination;
+  dest->outfile = outfile;
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for output to a memory buffer.
+ * The caller may supply an own initial buffer with appropriate size.
+ * Otherwise, or when the actual data output exceeds the given size,
+ * the library adapts the buffer size as necessary.
+ * The standard library functions malloc/free are used for allocating
+ * larger memory, so the buffer is available to the application after
+ * finishing compression, and then the application is responsible for
+ * freeing the requested memory.
+ * Note:  An initial buffer supplied by the caller is expected to be
+ * managed by the application.  The library does not free such buffer
+ * when allocating a larger buffer.
+ */
+
+GLOBAL(void)
+jpeg_mem_dest (j_compress_ptr cinfo,
+               unsigned char **outbuffer, unsigned long *outsize)
+{
+  my_mem_dest_ptr dest;
+
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same buffer without re-executing jpeg_mem_dest.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_mem_destination_mgr));
+  }
+
+  dest = (my_mem_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_mem_destination;
+  dest->pub.empty_output_buffer = empty_mem_output_buffer;
+  dest->pub.term_destination = term_mem_destination;
+  dest->outbuffer = outbuffer;
+  dest->outsize = outsize;
+  dest->newbuffer = NULL;
+
+  if (*outbuffer == NULL || *outsize == 0) {
+    /* Allocate initial buffer */
+    dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE);
+    if (dest->newbuffer == NULL)
+      ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+    *outsize = OUTPUT_BUF_SIZE;
+  }
+
+  dest->pub.next_output_byte = dest->buffer = *outbuffer;
+  dest->pub.free_in_buffer = dest->bufsize = *outsize;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatadst-tj.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatadst-tj.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatadst-tj.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatadst-tj.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
- * jdatadst.c
+ * jdatadst-tj.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2014 D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains compression data destination routines for the case of
  * emitting JPEG data to memory or to a file (or any stdio stream).
@@ -20,13 +23,13 @@
 #include "jpeglib.h"
 #include "jerror.h"
 
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
-extern void * malloc JPP((size_t size));
-extern void free JPP((void *ptr));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
 #endif
 
 
-#define OUTPUT_BUF_SIZE  4096	/* choose an efficiently fwrite'able size */
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
 
 
 /* Expanded data destination object for memory output */
@@ -34,15 +37,15 @@
 typedef struct {
   struct jpeg_destination_mgr pub; /* public fields */
 
-  unsigned char ** outbuffer;	/* target buffer */
-  unsigned long * outsize;
-  unsigned char * newbuffer;	/* newly allocated buffer */
-  JOCTET * buffer;		/* start of buffer */
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
   size_t bufsize;
   boolean alloc;
 } my_mem_destination_mgr;
 
-typedef my_mem_destination_mgr * my_mem_dest_ptr;
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
 
 
 /*
@@ -84,14 +87,14 @@
 empty_mem_output_buffer (j_compress_ptr cinfo)
 {
   size_t nextsize;
-  JOCTET * nextbuffer;
+  JOCTET *nextbuffer;
   my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
 
   if (!dest->alloc) ERREXIT(cinfo, JERR_BUFFER_SIZE);
 
   /* Try to allocate new buffer with double size */
   nextsize = dest->bufsize * 2;
-  nextbuffer = malloc(nextsize);
+  nextbuffer = (JOCTET *) malloc(nextsize);
 
   if (nextbuffer == NULL)
     ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
@@ -145,29 +148,33 @@
 
 GLOBAL(void)
 jpeg_mem_dest_tj (j_compress_ptr cinfo,
-	       unsigned char ** outbuffer, unsigned long * outsize,
-	       boolean alloc)
+               unsigned char **outbuffer, unsigned long *outsize,
+               boolean alloc)
 {
+  boolean reused = FALSE;
   my_mem_dest_ptr dest;
 
-  if (outbuffer == NULL || outsize == NULL)	/* sanity check */
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
     ERREXIT(cinfo, JERR_BUFFER_SIZE);
 
   /* The destination object is made permanent so that multiple JPEG images
    * can be written to the same buffer without re-executing jpeg_mem_dest.
    */
-  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
     cinfo->dest = (struct jpeg_destination_mgr *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  SIZEOF(my_mem_destination_mgr));
+                                  sizeof(my_mem_destination_mgr));
     dest = (my_mem_dest_ptr) cinfo->dest;
     dest->newbuffer = NULL;
+    dest->buffer = NULL;
   }
 
   dest = (my_mem_dest_ptr) cinfo->dest;
   dest->pub.init_destination = init_mem_destination;
   dest->pub.empty_output_buffer = empty_mem_output_buffer;
   dest->pub.term_destination = term_mem_destination;
+  if (dest->buffer == *outbuffer && *outbuffer != NULL && alloc)
+    reused = TRUE;
   dest->outbuffer = outbuffer;
   dest->outsize = outsize;
   dest->alloc = alloc;
@@ -175,7 +182,7 @@
   if (*outbuffer == NULL || *outsize == 0) {
     if (alloc) {
       /* Allocate initial buffer */
-      dest->newbuffer = *outbuffer = malloc(OUTPUT_BUF_SIZE);
+      dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE);
       if (dest->newbuffer == NULL)
         ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
       *outsize = OUTPUT_BUF_SIZE;
@@ -184,5 +191,7 @@
   }
 
   dest->pub.next_output_byte = dest->buffer = *outbuffer;
-  dest->pub.free_in_buffer = dest->bufsize = *outsize;
+  if (!reused)
+    dest->bufsize = *outsize;
+  dest->pub.free_in_buffer = dest->bufsize;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatasrc.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatasrc.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatasrc.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatasrc.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,284 @@
+/*
+ * jdatasrc.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from memory or from a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data source object for stdio input */
+
+typedef struct {
+  struct jpeg_source_mgr pub;   /* public fields */
+
+  FILE *infile;                 /* source stream */
+  JOCTET *buffer;               /* start of buffer */
+  boolean start_of_file;        /* have we gotten any data yet? */
+} my_source_mgr;
+
+typedef my_source_mgr *my_src_ptr;
+
+#define INPUT_BUF_SIZE  4096    /* choose an efficiently fread'able size */
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_source (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* We reset the empty-input-file flag for each image,
+   * but we don't clear the input buffer.
+   * This is correct behavior for reading a series of images from one source.
+   */
+  src->start_of_file = TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_input_buffer (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+  size_t nbytes;
+
+  nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE);
+
+  if (nbytes <= 0) {
+    if (src->start_of_file)     /* Treat empty input file as fatal error */
+      ERREXIT(cinfo, JERR_INPUT_EMPTY);
+    WARNMS(cinfo, JWRN_JPEG_EOF);
+    /* Insert a fake EOI marker */
+    src->buffer[0] = (JOCTET) 0xFF;
+    src->buffer[1] = (JOCTET) JPEG_EOI;
+    nbytes = 2;
+  }
+
+  src->pub.next_input_byte = src->buffer;
+  src->pub.bytes_in_buffer = nbytes;
+  src->start_of_file = FALSE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+fill_mem_input_buffer (j_decompress_ptr cinfo)
+{
+  static const JOCTET mybuffer[4] = {
+    (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0
+  };
+
+  /* The whole JPEG data is expected to reside in the supplied memory
+   * buffer, so any request for more data beyond the given buffer size
+   * is treated as an error.
+   */
+  WARNMS(cinfo, JWRN_JPEG_EOF);
+
+  /* Insert a fake EOI marker */
+
+  cinfo->src->next_input_byte = mybuffer;
+  cinfo->src->bytes_in_buffer = 2;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+{
+  struct jpeg_source_mgr *src = cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long) src->bytes_in_buffer) {
+      num_bytes -= (long) src->bytes_in_buffer;
+      (void) (*src->fill_input_buffer) (cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->next_input_byte += (size_t) num_bytes;
+    src->bytes_in_buffer -= (size_t) num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing decompression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_src (j_decompress_ptr cinfo, FILE *infile)
+{
+  my_src_ptr src;
+
+  /* The source object and input buffer are made permanent so that a series
+   * of JPEG images can be read from the same file by calling jpeg_stdio_src
+   * only before the first one.  (If we discarded the buffer at the end of
+   * one image, we'd likely lose the start of the next one.)
+   * This makes it unsafe to use this manager and a different source
+   * manager serially with the same JPEG object.  Caveat programmer.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_source_mgr));
+    src = (my_src_ptr) cinfo->src;
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  INPUT_BUF_SIZE * sizeof(JOCTET));
+  }
+
+  src = (my_src_ptr) cinfo->src;
+  src->pub.init_source = init_source;
+  src->pub.fill_input_buffer = fill_input_buffer;
+  src->pub.skip_input_data = skip_input_data;
+  src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->pub.term_source = term_source;
+  src->infile = infile;
+  src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */
+  src->pub.next_input_byte = NULL; /* until buffer loaded */
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for input from a supplied memory buffer.
+ * The buffer must contain the whole JPEG data.
+ */
+
+GLOBAL(void)
+jpeg_mem_src (j_decompress_ptr cinfo,
+              const unsigned char *inbuffer, unsigned long insize)
+{
+  struct jpeg_source_mgr *src;
+
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+
+  /* The source object is made permanent so that a series of JPEG images
+   * can be read from the same buffer by calling jpeg_mem_src only before
+   * the first one.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(struct jpeg_source_mgr));
+  }
+
+  src = cinfo->src;
+  src->init_source = init_mem_source;
+  src->fill_input_buffer = fill_mem_input_buffer;
+  src->skip_input_data = skip_input_data;
+  src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->term_source = term_source;
+  src->bytes_in_buffer = (size_t) insize;
+  src->next_input_byte = (const JOCTET *) inbuffer;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatasrc-tj.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatasrc-tj.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdatasrc-tj.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdatasrc-tj.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
- * jdatasrc.c
+ * jdatasrc-tj.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009-2010 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains decompression data source routines for the case of
  * reading JPEG data from memory or from a file (or any stdio stream).
@@ -69,16 +72,17 @@
 METHODDEF(boolean)
 fill_mem_input_buffer (j_decompress_ptr cinfo)
 {
-  static JOCTET mybuffer[4];
+  static const JOCTET mybuffer[4] = {
+    (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0
+  };
 
   /* The whole JPEG data is expected to reside in the supplied memory
    * buffer, so any request for more data beyond the given buffer size
    * is treated as an error.
    */
   WARNMS(cinfo, JWRN_JPEG_EOF);
+
   /* Insert a fake EOI marker */
-  mybuffer[0] = (JOCTET) 0xFF;
-  mybuffer[1] = (JOCTET) JPEG_EOI;
 
   cinfo->src->next_input_byte = mybuffer;
   cinfo->src->bytes_in_buffer = 2;
@@ -102,7 +106,7 @@
 METHODDEF(void)
 skip_input_data (j_decompress_ptr cinfo, long num_bytes)
 {
-  struct jpeg_source_mgr * src = cinfo->src;
+  struct jpeg_source_mgr *src = cinfo->src;
 
   /* Just a dumb implementation for now.  Could use fseek() except
    * it doesn't work on pipes.  Not clear that being smart is worth
@@ -154,21 +158,21 @@
 
 GLOBAL(void)
 jpeg_mem_src_tj (j_decompress_ptr cinfo,
-	      unsigned char * inbuffer, unsigned long insize)
+                 const unsigned char *inbuffer, unsigned long insize)
 {
-  struct jpeg_source_mgr * src;
+  struct jpeg_source_mgr *src;
 
-  if (inbuffer == NULL || insize == 0)	/* Treat empty input as fatal error */
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
     ERREXIT(cinfo, JERR_INPUT_EMPTY);
 
   /* The source object is made permanent so that a series of JPEG images
    * can be read from the same buffer by calling jpeg_mem_src only before
    * the first one.
    */
-  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
     cinfo->src = (struct jpeg_source_mgr *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  SIZEOF(struct jpeg_source_mgr));
+                                  sizeof(struct jpeg_source_mgr));
   }
 
   src = cinfo->src;
@@ -178,5 +182,5 @@
   src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
   src->term_source = term_source;
   src->bytes_in_buffer = (size_t) insize;
-  src->next_input_byte = (JOCTET *) inbuffer;
+  src->next_input_byte = (const JOCTET *) inbuffer;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcoefct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcoefct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcoefct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcoefct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jdcoefct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the coefficient buffer controller for decompression.
  * This controller is the top level of the JPEG decompressor proper.
@@ -15,95 +19,25 @@
  * Also, the input side (only) is used when reading a file for transcoding.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
+#include "jdcoefct.h"
 #include "jpegcomp.h"
 
-/* Block smoothing is only applicable for progressive JPEG, so: */
-#ifndef D_PROGRESSIVE_SUPPORTED
-#undef BLOCK_SMOOTHING_SUPPORTED
-#endif
-
-/* Private buffer controller object */
-
-typedef struct {
-  struct jpeg_d_coef_controller pub; /* public fields */
-
-  /* These variables keep track of the current location of the input side. */
-  /* cinfo->input_iMCU_row is also used for this. */
-  JDIMENSION MCU_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
-
-  /* The output side's location is represented by cinfo->output_iMCU_row. */
-
-  /* In single-pass modes, it's sufficient to buffer just one MCU.
-   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
-   * and let the entropy decoder write into that workspace each time.
-   * (On 80x86, the workspace is FAR even though it's not really very big;
-   * this is to keep the module interfaces unchanged when a large coefficient
-   * buffer is necessary.)
-   * In multi-pass modes, this array points to the current MCU's blocks
-   * within the virtual arrays; it is used only by the input side.
-   */
-  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
-
-  /* Temporary workspace for one MCU */
-  JCOEF * workspace;
-
-#ifdef D_MULTISCAN_FILES_SUPPORTED
-  /* In multi-pass modes, we need a virtual block array for each component. */
-  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
-#endif
-
-#ifdef BLOCK_SMOOTHING_SUPPORTED
-  /* When doing block smoothing, we latch coefficient Al values here */
-  int * coef_bits_latch;
-#define SAVED_COEFS  6		/* we save coef_bits[0..5] */
-#endif
-} my_coef_controller;
-
-typedef my_coef_controller * my_coef_ptr;
 
 /* Forward declarations */
 METHODDEF(int) decompress_onepass
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #ifdef D_MULTISCAN_FILES_SUPPORTED
 METHODDEF(int) decompress_data
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #endif
 #ifdef BLOCK_SMOOTHING_SUPPORTED
-LOCAL(boolean) smoothing_ok JPP((j_decompress_ptr cinfo));
+LOCAL(boolean) smoothing_ok (j_decompress_ptr cinfo);
 METHODDEF(int) decompress_smooth_data
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #endif
 
 
-LOCAL(void)
-start_iMCU_row (j_decompress_ptr cinfo)
-/* Reset within-iMCU-row counters for a new row (input side) */
-{
-  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-
-  /* In an interleaved scan, an MCU row is the same as an iMCU row.
-   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
-   * But at the bottom of the image, process only what's left.
-   */
-  if (cinfo->comps_in_scan > 1) {
-    coef->MCU_rows_per_iMCU_row = 1;
-  } else {
-    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
-      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
-    else
-      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
-  }
-
-  coef->MCU_ctr = 0;
-  coef->MCU_vert_offset = 0;
-}
-
-
 /*
  * Initialize for an input processing pass.
  */
@@ -152,7 +86,7 @@
 decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, ci, xindex, yindex, yoffset, useful_width;
@@ -165,49 +99,57 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
-      jzero_far((void FAR *) coef->MCU_buffer[0],
-		(size_t) (cinfo->blocks_in_MCU * SIZEOF(JBLOCK)));
+      jzero_far((void *) coef->MCU_buffer[0],
+                (size_t) (cinfo->blocks_in_MCU * sizeof(JBLOCK)));
       if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->MCU_ctr = MCU_col_num;
-	return JPEG_SUSPENDED;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
       }
-      /* Determine where data should go in output_buf and do the IDCT thing.
-       * We skip dummy blocks at the right and bottom edges (but blkn gets
-       * incremented past them!).  Note the inner loop relies on having
-       * allocated the MCU_buffer[] blocks sequentially.
+
+      /* Only perform the IDCT on blocks that are contained within the desired
+       * cropping region.
        */
-      blkn = 0;			/* index of current DCT block within MCU */
-      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	/* Don't bother to IDCT an uninteresting component. */
-	if (! compptr->component_needed) {
-	  blkn += compptr->MCU_blocks;
-	  continue;
-	}
-	inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
-	useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						    : compptr->last_col_width;
-	output_ptr = output_buf[compptr->component_index] +
-	  yoffset * compptr->_DCT_scaled_size;
-	start_col = MCU_col_num * compptr->MCU_sample_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (cinfo->input_iMCU_row < last_iMCU_row ||
-	      yoffset+yindex < compptr->last_row_height) {
-	    output_col = start_col;
-	    for (xindex = 0; xindex < useful_width; xindex++) {
-	      (*inverse_DCT) (cinfo, compptr,
-			      (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
-			      output_ptr, output_col);
-	      output_col += compptr->_DCT_scaled_size;
-	    }
-	  }
-	  blkn += compptr->MCU_width;
-	  output_ptr += compptr->_DCT_scaled_size;
-	}
+      if (MCU_col_num >= cinfo->master->first_iMCU_col &&
+          MCU_col_num <= cinfo->master->last_iMCU_col) {
+        /* Determine where data should go in output_buf and do the IDCT thing.
+         * We skip dummy blocks at the right and bottom edges (but blkn gets
+         * incremented past them!).  Note the inner loop relies on having
+         * allocated the MCU_buffer[] blocks sequentially.
+         */
+        blkn = 0;                 /* index of current DCT block within MCU */
+        for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+          compptr = cinfo->cur_comp_info[ci];
+          /* Don't bother to IDCT an uninteresting component. */
+          if (! compptr->component_needed) {
+            blkn += compptr->MCU_blocks;
+            continue;
+          }
+          inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
+          useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                      : compptr->last_col_width;
+          output_ptr = output_buf[compptr->component_index] +
+            yoffset * compptr->_DCT_scaled_size;
+          start_col = (MCU_col_num - cinfo->master->first_iMCU_col) *
+              compptr->MCU_sample_width;
+          for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+            if (cinfo->input_iMCU_row < last_iMCU_row ||
+                yoffset+yindex < compptr->last_row_height) {
+              output_col = start_col;
+              for (xindex = 0; xindex < useful_width; xindex++) {
+                (*inverse_DCT) (cinfo, compptr,
+                                (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
+                                output_ptr, output_col);
+                output_col += compptr->_DCT_scaled_size;
+              }
+            }
+            blkn += compptr->MCU_width;
+            output_ptr += compptr->_DCT_scaled_size;
+          }
+        }
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -232,7 +174,7 @@
 METHODDEF(int)
 dummy_consume_data (j_decompress_ptr cinfo)
 {
-  return JPEG_SUSPENDED;	/* Always indicate nothing was done */
+  return JPEG_SUSPENDED;        /* Always indicate nothing was done */
 }
 
 
@@ -249,7 +191,7 @@
 consume_data (j_decompress_ptr cinfo)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   int blkn, ci, xindex, yindex, yoffset;
   JDIMENSION start_col;
   JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
@@ -273,25 +215,25 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
-	    coef->MCU_buffer[blkn++] = buffer_ptr++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+          }
+        }
       }
       /* Try to fetch the MCU. */
       if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->MCU_ctr = MCU_col_num;
-	return JPEG_SUSPENDED;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -332,8 +274,8 @@
 
   /* Force some input to be done if we are getting ahead of the input. */
   while (cinfo->input_scan_number < cinfo->output_scan_number ||
-	 (cinfo->input_scan_number == cinfo->output_scan_number &&
-	  cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
+         (cinfo->input_scan_number == cinfo->output_scan_number &&
+          cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
     if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
       return JPEG_SUSPENDED;
   }
@@ -361,13 +303,14 @@
     output_ptr = output_buf[ci];
     /* Loop over all DCT blocks to be processed. */
     for (block_row = 0; block_row < block_rows; block_row++) {
-      buffer_ptr = buffer[block_row];
+      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
       output_col = 0;
-      for (block_num = 0; block_num < compptr->width_in_blocks; block_num++) {
-	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
-			output_ptr, output_col);
-	buffer_ptr++;
-	output_col += compptr->_DCT_scaled_size;
+      for (block_num = cinfo->master->first_MCU_col[ci];
+           block_num <= cinfo->master->last_MCU_col[ci]; block_num++) {
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
+                        output_ptr, output_col);
+        buffer_ptr++;
+        output_col += compptr->_DCT_scaled_size;
       }
       output_ptr += compptr->_DCT_scaled_size;
     }
@@ -413,9 +356,9 @@
   boolean smoothing_useful = FALSE;
   int ci, coefi;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtable;
-  int * coef_bits;
-  int * coef_bits_latch;
+  JQUANT_TBL *qtable;
+  int *coef_bits;
+  int *coef_bits_latch;
 
   if (! cinfo->progressive_mode || cinfo->coef_bits == NULL)
     return FALSE;
@@ -424,8 +367,8 @@
   if (coef->coef_bits_latch == NULL)
     coef->coef_bits_latch = (int *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  cinfo->num_components *
-				  (SAVED_COEFS * SIZEOF(int)));
+                                  cinfo->num_components *
+                                  (SAVED_COEFS * sizeof(int)));
   coef_bits_latch = coef->coef_bits_latch;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -435,11 +378,11 @@
       return FALSE;
     /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
     if (qtable->quantval[0] == 0 ||
-	qtable->quantval[Q01_POS] == 0 ||
-	qtable->quantval[Q10_POS] == 0 ||
-	qtable->quantval[Q20_POS] == 0 ||
-	qtable->quantval[Q11_POS] == 0 ||
-	qtable->quantval[Q02_POS] == 0)
+        qtable->quantval[Q01_POS] == 0 ||
+        qtable->quantval[Q10_POS] == 0 ||
+        qtable->quantval[Q20_POS] == 0 ||
+        qtable->quantval[Q11_POS] == 0 ||
+        qtable->quantval[Q02_POS] == 0)
       return FALSE;
     /* DC values must be at least partly known for all components. */
     coef_bits = cinfo->coef_bits[ci];
@@ -449,7 +392,7 @@
     for (coefi = 1; coefi <= 5; coefi++) {
       coef_bits_latch[coefi] = coef_bits[coefi];
       if (coef_bits[coefi] != 0)
-	smoothing_useful = TRUE;
+        smoothing_useful = TRUE;
     }
     coef_bits_latch += SAVED_COEFS;
   }
@@ -476,10 +419,10 @@
   jpeg_component_info *compptr;
   inverse_DCT_method_ptr inverse_DCT;
   boolean first_row, last_row;
-  JCOEF * workspace;
+  JCOEF *workspace;
   int *coef_bits;
   JQUANT_TBL *quanttbl;
-  INT32 Q00,Q01,Q02,Q10,Q11,Q20, num;
+  JLONG Q00,Q01,Q02,Q10,Q11,Q20, num;
   int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9;
   int Al, pred;
 
@@ -488,7 +431,7 @@
 
   /* Force some input to be done if we are getting ahead of the input. */
   while (cinfo->input_scan_number <= cinfo->output_scan_number &&
-	 ! cinfo->inputctl->eoi_reached) {
+         ! cinfo->inputctl->eoi_reached) {
     if (cinfo->input_scan_number == cinfo->output_scan_number) {
       /* If input is working on current scan, we ordinarily want it to
        * have completed the current row.  But if input scan is DC,
@@ -497,7 +440,7 @@
        */
       JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
       if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta)
-	break;
+        break;
     }
     if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
       return JPEG_SUSPENDED;
@@ -525,15 +468,15 @@
     if (cinfo->output_iMCU_row > 0) {
       access_rows += compptr->v_samp_factor; /* prior iMCU row too */
       buffer = (*cinfo->mem->access_virt_barray)
-	((j_common_ptr) cinfo, coef->whole_image[ci],
-	 (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
-	 (JDIMENSION) access_rows, FALSE);
-      buffer += compptr->v_samp_factor;	/* point to current iMCU row */
+        ((j_common_ptr) cinfo, coef->whole_image[ci],
+         (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
+         (JDIMENSION) access_rows, FALSE);
+      buffer += compptr->v_samp_factor; /* point to current iMCU row */
       first_row = FALSE;
     } else {
       buffer = (*cinfo->mem->access_virt_barray)
-	((j_common_ptr) cinfo, coef->whole_image[ci],
-	 (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
+        ((j_common_ptr) cinfo, coef->whole_image[ci],
+         (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
       first_row = TRUE;
     }
     /* Fetch component-dependent info */
@@ -549,15 +492,15 @@
     output_ptr = output_buf[ci];
     /* Loop over all DCT blocks to be processed. */
     for (block_row = 0; block_row < block_rows; block_row++) {
-      buffer_ptr = buffer[block_row];
+      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
       if (first_row && block_row == 0)
-	prev_block_row = buffer_ptr;
+        prev_block_row = buffer_ptr;
       else
-	prev_block_row = buffer[block_row-1];
+        prev_block_row = buffer[block_row-1];
       if (last_row && block_row == block_rows-1)
-	next_block_row = buffer_ptr;
+        next_block_row = buffer_ptr;
       else
-	next_block_row = buffer[block_row+1];
+        next_block_row = buffer[block_row+1];
       /* We fetch the surrounding DC values using a sliding-register approach.
        * Initialize all nine here so as to do the right thing on narrow pics.
        */
@@ -566,103 +509,104 @@
       DC7 = DC8 = DC9 = (int) next_block_row[0][0];
       output_col = 0;
       last_block_column = compptr->width_in_blocks - 1;
-      for (block_num = 0; block_num <= last_block_column; block_num++) {
-	/* Fetch current DCT block into workspace so we can modify it. */
-	jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
-	/* Update DC values */
-	if (block_num < last_block_column) {
-	  DC3 = (int) prev_block_row[1][0];
-	  DC6 = (int) buffer_ptr[1][0];
-	  DC9 = (int) next_block_row[1][0];
-	}
-	/* Compute coefficient estimates per K.8.
-	 * An estimate is applied only if coefficient is still zero,
-	 * and is not known to be fully accurate.
-	 */
-	/* AC01 */
-	if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
-	  num = 36 * Q00 * (DC4 - DC6);
-	  if (num >= 0) {
-	    pred = (int) (((Q01<<7) + num) / (Q01<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q01<<7) - num) / (Q01<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[1] = (JCOEF) pred;
-	}
-	/* AC10 */
-	if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
-	  num = 36 * Q00 * (DC2 - DC8);
-	  if (num >= 0) {
-	    pred = (int) (((Q10<<7) + num) / (Q10<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q10<<7) - num) / (Q10<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[8] = (JCOEF) pred;
-	}
-	/* AC20 */
-	if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
-	  num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
-	  if (num >= 0) {
-	    pred = (int) (((Q20<<7) + num) / (Q20<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q20<<7) - num) / (Q20<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[16] = (JCOEF) pred;
-	}
-	/* AC11 */
-	if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
-	  num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
-	  if (num >= 0) {
-	    pred = (int) (((Q11<<7) + num) / (Q11<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q11<<7) - num) / (Q11<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[9] = (JCOEF) pred;
-	}
-	/* AC02 */
-	if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
-	  num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
-	  if (num >= 0) {
-	    pred = (int) (((Q02<<7) + num) / (Q02<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q02<<7) - num) / (Q02<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[2] = (JCOEF) pred;
-	}
-	/* OK, do the IDCT */
-	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
-			output_ptr, output_col);
-	/* Advance for next column */
-	DC1 = DC2; DC2 = DC3;
-	DC4 = DC5; DC5 = DC6;
-	DC7 = DC8; DC8 = DC9;
-	buffer_ptr++, prev_block_row++, next_block_row++;
-	output_col += compptr->_DCT_scaled_size;
+      for (block_num = cinfo->master->first_MCU_col[ci];
+           block_num <= cinfo->master->last_MCU_col[ci]; block_num++) {
+        /* Fetch current DCT block into workspace so we can modify it. */
+        jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
+        /* Update DC values */
+        if (block_num < last_block_column) {
+          DC3 = (int) prev_block_row[1][0];
+          DC6 = (int) buffer_ptr[1][0];
+          DC9 = (int) next_block_row[1][0];
+        }
+        /* Compute coefficient estimates per K.8.
+         * An estimate is applied only if coefficient is still zero,
+         * and is not known to be fully accurate.
+         */
+        /* AC01 */
+        if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
+          num = 36 * Q00 * (DC4 - DC6);
+          if (num >= 0) {
+            pred = (int) (((Q01<<7) + num) / (Q01<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q01<<7) - num) / (Q01<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[1] = (JCOEF) pred;
+        }
+        /* AC10 */
+        if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
+          num = 36 * Q00 * (DC2 - DC8);
+          if (num >= 0) {
+            pred = (int) (((Q10<<7) + num) / (Q10<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q10<<7) - num) / (Q10<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[8] = (JCOEF) pred;
+        }
+        /* AC20 */
+        if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
+          num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
+          if (num >= 0) {
+            pred = (int) (((Q20<<7) + num) / (Q20<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q20<<7) - num) / (Q20<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[16] = (JCOEF) pred;
+        }
+        /* AC11 */
+        if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
+          num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+          if (num >= 0) {
+            pred = (int) (((Q11<<7) + num) / (Q11<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q11<<7) - num) / (Q11<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[9] = (JCOEF) pred;
+        }
+        /* AC02 */
+        if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
+          num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
+          if (num >= 0) {
+            pred = (int) (((Q02<<7) + num) / (Q02<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q02<<7) - num) / (Q02<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[2] = (JCOEF) pred;
+        }
+        /* OK, do the IDCT */
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
+                        output_ptr, output_col);
+        /* Advance for next column */
+        DC1 = DC2; DC2 = DC3;
+        DC4 = DC5; DC5 = DC6;
+        DC7 = DC8; DC8 = DC9;
+        buffer_ptr++, prev_block_row++, next_block_row++;
+        output_col += compptr->_DCT_scaled_size;
       }
       output_ptr += compptr->_DCT_scaled_size;
     }
@@ -687,7 +631,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_d_coef_controller *) coef;
   coef->pub.start_input_pass = start_input_pass;
   coef->pub.start_output_pass = start_output_pass;
@@ -705,20 +649,20 @@
     jpeg_component_info *compptr;
 
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       access_rows = compptr->v_samp_factor;
 #ifdef BLOCK_SMOOTHING_SUPPORTED
       /* If block smoothing could be used, need a bigger window */
       if (cinfo->progressive_mode)
-	access_rows *= 3;
+        access_rows *= 3;
 #endif
       coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
-	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
-				(long) compptr->h_samp_factor),
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor),
-	 (JDIMENSION) access_rows);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
+         (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+                                (long) compptr->h_samp_factor),
+         (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+                                (long) compptr->v_samp_factor),
+         (JDIMENSION) access_rows);
     }
     coef->pub.consume_data = consume_data;
     coef->pub.decompress_data = decompress_data;
@@ -733,7 +677,7 @@
 
     buffer = (JBLOCKROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                  D_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
     for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
       coef->MCU_buffer[i] = buffer + i;
     }
@@ -745,5 +689,5 @@
   /* Allocate the workspace buffer */
   coef->workspace = (JCOEF *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-                                SIZEOF(JCOEF) * DCTSIZE2);
+                                sizeof(JCOEF) * DCTSIZE2);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcoefct.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcoefct.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcoefct.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcoefct.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,82 @@
+/*
+ * jdcoefct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Block smoothing is only applicable for progressive JPEG, so: */
+#ifndef D_PROGRESSIVE_SUPPORTED
+#undef BLOCK_SMOOTHING_SUPPORTED
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_coef_controller pub; /* public fields */
+
+  /* These variables keep track of the current location of the input side. */
+  /* cinfo->input_iMCU_row is also used for this. */
+  JDIMENSION MCU_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
+
+  /* The output side's location is represented by cinfo->output_iMCU_row. */
+
+  /* In single-pass modes, it's sufficient to buffer just one MCU.
+   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
+   * and let the entropy decoder write into that workspace each time.
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays; it is used only by the input side.
+   */
+  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
+
+  /* Temporary workspace for one MCU */
+  JCOEF *workspace;
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+#endif
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  /* When doing block smoothing, we latch coefficient Al values here */
+  int *coef_bits_latch;
+#define SAVED_COEFS  6          /* we save coef_bits[0..5] */
+#endif
+} my_coef_controller;
+
+typedef my_coef_controller *my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row (j_decompress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row (input side) */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->MCU_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcol565.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcol565.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcol565.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcol565.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,384 @@
+/*
+ * jdcol565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+/* This file is included by jdcolor.c */
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565_convert_internal (j_decompress_ptr cinfo,
+                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+                             JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565_convert_internal (j_decompress_ptr cinfo,
+                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+                             JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = GETJSAMPLE(*inptr0);
+      g = GETJSAMPLE(*inptr1);
+      b = GETJSAMPLE(*inptr2);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      g = *inptr++;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf, JDIMENSION input_row,
+                               JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      d0 = DITHER_ROTATE(d0);
+
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      d0 = DITHER_ROTATE(d0);
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolext.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolext.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,143 @@
+/*
+ * jdcolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+
+/* This file is included by jdcolor.c */
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+ycc_rgb_convert_internal (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+INLINE
+LOCAL(void)
+gray_rgb_convert_internal (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION input_row,
+                           JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert RGB to extended RGB: just swap the order of source pixels
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = inptr0[col];
+      outptr[RGB_GREEN] = inptr1[col];
+      outptr[RGB_BLUE] = inptr2[col];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolext.c.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,104 +0,0 @@
-/*
- * jdcolext.c
- *
- * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains output colorspace conversion routines.
- */
-
-
-/* This file is included by jdcolor.c */
-
-
-/*
- * Convert some rows of samples to the output colorspace.
- *
- * Note that we change from noninterleaved, one-plane-per-component format
- * to interleaved-pixel format.  The output buffer is therefore three times
- * as wide as the input buffer.
- * A starting row offset is provided only for the input buffer.  The caller
- * can easily adjust the passed output_buf value to accommodate any row
- * offset required on that side.
- */
-
-INLINE
-LOCAL(void)
-ycc_rgb_convert_internal (j_decompress_ptr cinfo,
-                          JSAMPIMAGE input_buf, JDIMENSION input_row,
-                          JSAMPARRAY output_buf, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int y, cb, cr;
-  register JSAMPROW outptr;
-  register JSAMPROW inptr0, inptr1, inptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->output_width;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  register int * Crrtab = cconvert->Cr_r_tab;
-  register int * Cbbtab = cconvert->Cb_b_tab;
-  register INT32 * Crgtab = cconvert->Cr_g_tab;
-  register INT32 * Cbgtab = cconvert->Cb_g_tab;
-  SHIFT_TEMPS
-
-  while (--num_rows >= 0) {
-    inptr0 = input_buf[0][input_row];
-    inptr1 = input_buf[1][input_row];
-    inptr2 = input_buf[2][input_row];
-    input_row++;
-    outptr = *output_buf++;
-    for (col = 0; col < num_cols; col++) {
-      y  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
-      /* Range-limiting is essential due to noise introduced by DCT losses. */
-      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
-      outptr[RGB_GREEN] = range_limit[y +
-			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
-						 SCALEBITS))];
-      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
-      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
-      /* alpha channel value */
-#ifdef RGB_ALPHA
-      outptr[RGB_ALPHA] = 0xFF;
-#endif
-      outptr += RGB_PIXELSIZE;
-    }
-  }
-}
-
-
-/*
- * Convert grayscale to RGB: just duplicate the graylevel three times.
- * This is provided to support applications that don't want to cope
- * with grayscale as a separate case.
- */
-
-INLINE
-LOCAL(void)
-gray_rgb_convert_internal (j_decompress_ptr cinfo,
-                           JSAMPIMAGE input_buf, JDIMENSION input_row,
-                           JSAMPARRAY output_buf, int num_rows)
-{
-  register JSAMPROW inptr, outptr;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->output_width;
-
-  while (--num_rows >= 0) {
-    inptr = input_buf[0][input_row++];
-    outptr = *output_buf++;
-    for (col = 0; col < num_cols; col++) {
-      /* We can dispense with GETJSAMPLE() here */
-      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
-      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
-      /* alpha channel value */
-#ifdef RGB_ALPHA
-      outptr[RGB_ALPHA] = 0xFF;
-#endif
-      outptr += RGB_PIXELSIZE;
-    }
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolor.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolor.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdcolor.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdcolor.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jdcolor.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009, 2011-2012, 2014-2015, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains output colorspace conversion routines.
  */
@@ -14,7 +18,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 
 /* Private subobject */
@@ -23,24 +27,32 @@
   struct jpeg_color_deconverter pub; /* public fields */
 
   /* Private state for YCC->RGB conversion */
-  int * Cr_r_tab;		/* => table for Cr to R conversion */
-  int * Cb_b_tab;		/* => table for Cb to B conversion */
-  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
-  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
+
+  /* Private state for RGB->Y conversion */
+  JLONG *rgb_y_tab;             /* => table for RGB to Y conversion */
 } my_color_deconverter;
 
-typedef my_color_deconverter * my_cconvert_ptr;
+typedef my_color_deconverter *my_cconvert_ptr;
 
 
 /**************** YCbCr -> RGB conversion: most common case **************/
+/****************   RGB -> Y   conversion: less common case **************/
 
 /*
  * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
  * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
  * The conversion equations to be implemented are therefore
- *	R = Y                + 1.40200 * Cr
- *	G = Y - 0.34414 * Cb - 0.71414 * Cr
- *	B = Y + 1.77200 * Cb
+ *
+ *      R = Y                + 1.40200 * Cr
+ *      G = Y - 0.34414 * Cb - 0.71414 * Cr
+ *      B = Y + 1.77200 * Cb
+ *
+ *      Y = 0.29900 * R + 0.58700 * G + 0.11400 * B
+ *
  * where Cb and Cr represent the incoming values less CENTERJSAMPLE.
  * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
  *
@@ -61,14 +73,26 @@
  * together before rounding.
  */
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
+
+/* We allocate one big table for RGB->Y conversion and divide it up into
+ * three parts, instead of doing three alloc_small requests.  This lets us
+ * use a single table base address, which can be held in a register in the
+ * inner loops on many machines (more than can hold all three addresses,
+ * anyway).
+ */
+
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1*(MAXJSAMPLE+1))      /* offset to G => Y section */
+#define B_Y_OFF         (2*(MAXJSAMPLE+1))      /* etc. */
+#define TABLE_SIZE      (3*(MAXJSAMPLE+1))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jdcolext.c.inc"
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -80,13 +104,15 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extrgb_convert_internal
 #define gray_rgb_convert_internal gray_extrgb_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extrgb_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
@@ -95,7 +121,8 @@
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extrgbx_convert_internal
 #define gray_rgb_convert_internal gray_extrgbx_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extrgbx_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -103,6 +130,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGR_RED
 #define RGB_GREEN EXT_BGR_GREEN
@@ -110,13 +138,15 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extbgr_convert_internal
 #define gray_rgb_convert_internal gray_extbgr_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extbgr_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
@@ -125,7 +155,8 @@
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extbgrx_convert_internal
 #define gray_rgb_convert_internal gray_extbgrx_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extbgrx_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -133,6 +164,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
@@ -141,7 +173,8 @@
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extxbgr_convert_internal
 #define gray_rgb_convert_internal gray_extxbgr_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extxbgr_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -149,6 +182,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
@@ -157,7 +191,8 @@
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extxrgb_convert_internal
 #define gray_rgb_convert_internal gray_extxrgb_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extxrgb_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -165,6 +200,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 
 /*
@@ -176,31 +212,31 @@
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   int i;
-  INT32 x;
+  JLONG x;
   SHIFT_TEMPS
 
   cconvert->Cr_r_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
+                                (MAXJSAMPLE+1) * sizeof(int));
   cconvert->Cb_b_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cr_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(int));
+  cconvert->Cr_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
-  cconvert->Cb_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
+  cconvert->Cb_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
 
   for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
     /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
     /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
     /* Cr=>R value is nearest int to 1.40200 * x */
     cconvert->Cr_r_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
     /* Cb=>B value is nearest int to 1.77200 * x */
     cconvert->Cb_b_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
     /* Cr=>G value is scaled-up -0.71414 * x */
     cconvert->Cr_g_tab[i] = (- FIX(0.71414)) * x;
     /* Cb=>G value is scaled-up -0.34414 * x */
@@ -216,8 +252,8 @@
 
 METHODDEF(void)
 ycc_rgb_convert (j_decompress_ptr cinfo,
-		 JSAMPIMAGE input_buf, JDIMENSION input_row,
-		 JSAMPARRAY output_buf, int num_rows)
+                 JSAMPIMAGE input_buf, JDIMENSION input_row,
+                 JSAMPARRAY output_buf, int num_rows)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -260,32 +296,122 @@
 
 
 /*
+ * Initialize for RGB->grayscale colorspace conversion.
+ */
+
+LOCAL(void)
+build_rgb_y_table (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  JLONG *rgb_y_tab;
+  JLONG i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_y_tab = rgb_y_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (TABLE_SIZE * sizeof(JLONG)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_y_tab[i+R_Y_OFF] = FIX(0.29900) * i;
+    rgb_y_tab[i+G_Y_OFF] = FIX(0.58700) * i;
+    rgb_y_tab[i+B_Y_OFF] = FIX(0.11400) * i + ONE_HALF;
+  }
+}
+
+
+/*
+ * Convert RGB to grayscale.
+ */
+
+METHODDEF(void)
+rgb_gray_convert (j_decompress_ptr cinfo,
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_y_tab;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr0[col]);
+      g = GETJSAMPLE(inptr1[col]);
+      b = GETJSAMPLE(inptr2[col]);
+      /* Y */
+      outptr[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
  * Color conversion for no colorspace change: just copy the data,
  * converting from separate-planes to interleaved representation.
  */
 
 METHODDEF(void)
 null_convert (j_decompress_ptr cinfo,
-	      JSAMPIMAGE input_buf, JDIMENSION input_row,
-	      JSAMPARRAY output_buf, int num_rows)
+              JSAMPIMAGE input_buf, JDIMENSION input_row,
+              JSAMPARRAY output_buf, int num_rows)
 {
-  register JSAMPROW inptr, outptr;
-  register JDIMENSION count;
+  register JSAMPROW inptr, inptr0, inptr1, inptr2, inptr3, outptr;
+  register JDIMENSION col;
   register int num_components = cinfo->num_components;
   JDIMENSION num_cols = cinfo->output_width;
   int ci;
 
-  while (--num_rows >= 0) {
-    for (ci = 0; ci < num_components; ci++) {
-      inptr = input_buf[ci][input_row];
-      outptr = output_buf[0] + ci;
-      for (count = num_cols; count > 0; count--) {
-	*outptr = *inptr++;	/* needn't bother with GETJSAMPLE() here */
-	outptr += num_components;
+  if (num_components == 3) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf[0][input_row];
+      inptr1 = input_buf[1][input_row];
+      inptr2 = input_buf[2][input_row];
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0[col];
+        *outptr++ = inptr1[col];
+        *outptr++ = inptr2[col];
       }
     }
-    input_row++;
-    output_buf++;
+  } else if (num_components == 4) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf[0][input_row];
+      inptr1 = input_buf[1][input_row];
+      inptr2 = input_buf[2][input_row];
+      inptr3 = input_buf[3][input_row];
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0[col];
+        *outptr++ = inptr1[col];
+        *outptr++ = inptr2[col];
+        *outptr++ = inptr3[col];
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      for (ci = 0; ci < num_components; ci++) {
+        inptr = input_buf[ci][input_row];
+        outptr = *output_buf;
+        for (col = 0; col < num_cols; col++) {
+          outptr[ci] = inptr[col];
+          outptr += num_components;
+        }
+      }
+      output_buf++;
+      input_row++;
+    }
   }
 }
 
@@ -298,11 +424,11 @@
 
 METHODDEF(void)
 grayscale_convert (j_decompress_ptr cinfo,
-		   JSAMPIMAGE input_buf, JDIMENSION input_row,
-		   JSAMPARRAY output_buf, int num_rows)
+                   JSAMPIMAGE input_buf, JDIMENSION input_row,
+                   JSAMPARRAY output_buf, int num_rows)
 {
   jcopy_sample_rows(input_buf[0], (int) input_row, output_buf, 0,
-		    num_rows, cinfo->output_width);
+                    num_rows, cinfo->output_width);
 }
 
 
@@ -312,8 +438,8 @@
 
 METHODDEF(void)
 gray_rgb_convert (j_decompress_ptr cinfo,
-		  JSAMPIMAGE input_buf, JDIMENSION input_row,
-		  JSAMPARRAY output_buf, int num_rows)
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -353,6 +479,52 @@
 
 
 /*
+ * Convert plain RGB to extended RGB
+ */
+
+METHODDEF(void)
+rgb_rgb_convert (j_decompress_ptr cinfo,
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
+{
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      rgb_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      rgb_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      rgb_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      rgb_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      rgb_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      rgb_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    default:
+      rgb_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                               num_rows);
+      break;
+  }
+}
+
+
+/*
  * Adobe-style YCCK->CMYK conversion.
  * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
  * conversion as above, while passing K (black) unchanged.
@@ -361,8 +533,8 @@
 
 METHODDEF(void)
 ycck_cmyk_convert (j_decompress_ptr cinfo,
-		   JSAMPIMAGE input_buf, JDIMENSION input_row,
-		   JSAMPARRAY output_buf, int num_rows)
+                   JSAMPIMAGE input_buf, JDIMENSION input_row,
+                   JSAMPARRAY output_buf, int num_rows)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   register int y, cb, cr;
@@ -371,11 +543,11 @@
   register JDIMENSION col;
   JDIMENSION num_cols = cinfo->output_width;
   /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  register int * Crrtab = cconvert->Cr_r_tab;
-  register int * Cbbtab = cconvert->Cb_b_tab;
-  register INT32 * Crgtab = cconvert->Cr_g_tab;
-  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
   SHIFT_TEMPS
 
   while (--num_rows >= 0) {
@@ -390,13 +562,13 @@
       cb = GETJSAMPLE(inptr1[col]);
       cr = GETJSAMPLE(inptr2[col]);
       /* Range-limiting is essential due to noise introduced by DCT losses. */
-      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];	/* red */
-      outptr[1] = range_limit[MAXJSAMPLE - (y +			/* green */
-			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
-						 SCALEBITS)))];
-      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];	/* blue */
+      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];   /* red */
+      outptr[1] = range_limit[MAXJSAMPLE - (y +                 /* green */
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS)))];
+      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];   /* blue */
       /* K passes through unchanged */
-      outptr[3] = inptr3[col];	/* don't need GETJSAMPLE here */
+      outptr[3] = inptr3[col];  /* don't need GETJSAMPLE here */
       outptr += 4;
     }
   }
@@ -404,6 +576,164 @@
 
 
 /*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b)   ((((r) << 8) & 0xF800) |  \
+                                      (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)   (((r) & 0xF8) | ((g) >> 5) |  \
+                                      (((g) << 11) & 0xE000) |  \
+                                      (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)     ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)     ((l << 16) | r)
+
+#define PACK_NEED_ALIGNMENT(ptr)     (((size_t)(ptr)) & 3)
+
+#define WRITE_TWO_ALIGNED_PIXELS(addr, pixels)  ((*(int *)(addr)) = pixels)
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficent for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565 PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE
+#define ycc_rgb565_convert_internal ycc_rgb565_convert_le
+#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_le
+#define rgb_rgb565_convert_internal rgb_rgb565_convert_le
+#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_le
+#define gray_rgb565_convert_internal gray_rgb565_convert_le
+#define gray_rgb565D_convert_internal gray_rgb565D_convert_le
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+#define PACK_SHORT_565 PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE
+#define ycc_rgb565_convert_internal ycc_rgb565_convert_be
+#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_be
+#define rgb_rgb565_convert_internal rgb_rgb565_convert_be
+#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_be
+#define gray_rgb565_convert_internal gray_rgb565_convert_be
+#define gray_rgb565D_convert_internal gray_rgb565D_convert_be
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+
+METHODDEF(void)
+ycc_rgb565_convert (j_decompress_ptr cinfo,
+                    JSAMPIMAGE input_buf, JDIMENSION input_row,
+                    JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+ycc_rgb565D_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565_convert (j_decompress_ptr cinfo,
+                    JSAMPIMAGE input_buf, JDIMENSION input_row,
+                    JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565D_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565D_convert (j_decompress_ptr cinfo,
+                      JSAMPIMAGE input_buf, JDIMENSION input_row,
+                      JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+/*
  * Empty method for start_pass.
  */
 
@@ -426,7 +756,7 @@
 
   cconvert = (my_cconvert_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_color_deconverter));
+                                sizeof(my_color_deconverter));
   cinfo->cconvert = (struct jpeg_color_deconverter *) cconvert;
   cconvert->pub.start_pass = start_pass_dcolor;
 
@@ -449,7 +779,7 @@
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
     break;
 
-  default:			/* JCS_UNKNOWN can be anything */
+  default:                      /* JCS_UNKNOWN can be anything */
     if (cinfo->num_components < 1)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
     break;
@@ -464,11 +794,14 @@
   case JCS_GRAYSCALE:
     cinfo->out_color_components = 1;
     if (cinfo->jpeg_color_space == JCS_GRAYSCALE ||
-	cinfo->jpeg_color_space == JCS_YCbCr) {
+        cinfo->jpeg_color_space == JCS_YCbCr) {
       cconvert->pub.color_convert = grayscale_convert;
       /* For color->grayscale conversion, only the Y (0) component is needed */
       for (ci = 1; ci < cinfo->num_components; ci++)
-	cinfo->comp_info[ci].component_needed = FALSE;
+        cinfo->comp_info[ci].component_needed = FALSE;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      cconvert->pub.color_convert = rgb_gray_convert;
+      build_rgb_y_table(cinfo);
     } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
@@ -494,13 +827,48 @@
       }
     } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
       cconvert->pub.color_convert = gray_rgb_convert;
-    } else if (cinfo->jpeg_color_space == cinfo->out_color_space &&
-      rgb_pixelsize[cinfo->out_color_space] == 3) {
-      cconvert->pub.color_convert = null_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      if (rgb_red[cinfo->out_color_space] == 0 &&
+          rgb_green[cinfo->out_color_space] == 1 &&
+          rgb_blue[cinfo->out_color_space] == 2 &&
+          rgb_pixelsize[cinfo->out_color_space] == 3)
+        cconvert->pub.color_convert = null_convert;
+      else
+        cconvert->pub.color_convert = rgb_rgb_convert;
     } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
+  case JCS_RGB565:
+    cinfo->out_color_components = 3;
+    if (cinfo->dither_mode == JDITHER_NONE) {
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+         if (jsimd_can_ycc_rgb565())
+           cconvert->pub.color_convert = jsimd_ycc_rgb565_convert;
+         else {
+           cconvert->pub.color_convert = ycc_rgb565_convert;
+           build_ycc_rgb_table(cinfo);
+        }
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    } else {
+      /* only ordered dithering is supported */
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+        cconvert->pub.color_convert = ycc_rgb565D_convert;
+        build_ycc_rgb_table(cinfo);
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565D_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565D_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    }
+    break;
+
   case JCS_CMYK:
     cinfo->out_color_components = 4;
     if (cinfo->jpeg_color_space == JCS_YCCK) {
@@ -517,7 +885,7 @@
     if (cinfo->out_color_space == cinfo->jpeg_color_space) {
       cinfo->out_color_components = cinfo->num_components;
       cconvert->pub.color_convert = null_convert;
-    } else			/* unsupported non-null conversion */
+    } else                      /* unsupported non-null conversion */
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdct.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdct.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdct.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdct.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,14 +1,17 @@
 /*
  * jdct.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This include file contains common declarations for the forward and
  * inverse DCT modules.  These declarations are private to the DCT managers
  * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
- * The individual DCT algorithms are kept in separate files to ease 
+ * The individual DCT algorithms are kept in separate files to ease
  * machine-dependent tuning (e.g., assembly coding).
  */
 
@@ -16,7 +19,7 @@
 /*
  * A forward DCT routine is given a pointer to a work area of type DCTELEM[];
  * the DCT is to be performed in-place in that buffer.  Type DCTELEM is int
- * for 8-bit samples, INT32 for 12-bit samples.  (NOTE: Floating-point DCT
+ * for 8-bit samples, JLONG for 12-bit samples.  (NOTE: Floating-point DCT
  * implementations use an array of type FAST_FLOAT, instead.)
  * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
  * The DCT outputs are returned scaled up by a factor of 8; they therefore
@@ -29,7 +32,7 @@
 
 #if BITS_IN_JSAMPLE == 8
 #ifndef WITH_SIMD
-typedef int DCTELEM;		/* 16 or 32 bits is fine */
+typedef int DCTELEM;            /* 16 or 32 bits is fine */
 typedef unsigned int UDCTELEM;
 typedef unsigned long long UDCTELEM2;
 #else
@@ -38,8 +41,7 @@
 typedef unsigned int UDCTELEM2;
 #endif
 #else
-typedef INT32 DCTELEM;		/* must have 32 bits */
-typedef UINT32 UDCTELEM;
+typedef JLONG DCTELEM;          /* must have 32 bits */
 typedef unsigned long long UDCTELEM2;
 #endif
 
@@ -64,10 +66,10 @@
 typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
 #if BITS_IN_JSAMPLE == 8
 typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
-#define IFAST_SCALE_BITS  2	/* fractional bits in scale factors */
+#define IFAST_SCALE_BITS  2     /* fractional bits in scale factors */
 #else
-typedef INT32 IFAST_MULT_TYPE;	/* need 32 bits for scaled quantizers */
-#define IFAST_SCALE_BITS  13	/* fractional bits in scale factors */
+typedef JLONG IFAST_MULT_TYPE;  /* need 32 bits for scaled quantizers */
+#define IFAST_SCALE_BITS  13    /* fractional bits in scale factors */
 #endif
 typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
 
@@ -86,57 +88,79 @@
 #define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_fdct_islow		jFDislow
-#define jpeg_fdct_ifast		jFDifast
-#define jpeg_fdct_float		jFDfloat
-#define jpeg_idct_islow		jRDislow
-#define jpeg_idct_ifast		jRDifast
-#define jpeg_idct_float		jRDfloat
-#define jpeg_idct_4x4		jRD4x4
-#define jpeg_idct_2x2		jRD2x2
-#define jpeg_idct_1x1		jRD1x1
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
 /* Extern declarations for the forward and inverse DCT routines. */
 
-EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
-EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
-EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
+EXTERN(void) jpeg_fdct_islow (DCTELEM *data);
+EXTERN(void) jpeg_fdct_ifast (DCTELEM *data);
+EXTERN(void) jpeg_fdct_float (FAST_FLOAT *data);
 
 EXTERN(void) jpeg_idct_islow
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_ifast
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_float
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_7x7
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_6x6
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_5x5
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_4x4
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_3x3
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_2x2
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_1x1
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_9x9
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_10x10
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_11x11
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_12x12
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_13x13
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_14x14
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_15x15
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_16x16
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 
 
 /*
  * Macros for handling fixed-point arithmetic; these are used by many
  * but not all of the DCT/IDCT modules.
  *
- * All values are expected to be of type INT32.
+ * All values are expected to be of type JLONG.
  * Fractional constants are scaled left by CONST_BITS bits.
  * CONST_BITS is defined within each module using these macros,
  * and may differ from one module to the next.
  */
 
-#define ONE	((INT32) 1)
+#define ONE     ((JLONG) 1)
 #define CONST_SCALE (ONE << CONST_BITS)
 
 /* Convert a positive real constant to an integer scaled by CONST_SCALE.
@@ -144,16 +168,16 @@
  * thus causing a lot of useless floating-point operations at run time.
  */
 
-#define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5))
+#define FIX(x)  ((JLONG) ((x) * CONST_SCALE + 0.5))
 
-/* Descale and correctly round an INT32 value that's scaled by N bits.
+/* Descale and correctly round a JLONG value that's scaled by N bits.
  * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
  * the fudge factor is correct for either sign of X.
  */
 
 #define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
  * This macro is used only when the two inputs will actually be no more than
  * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
  * full 32x32 multiply.  This provides a useful speedup on many machines.
@@ -162,23 +186,23 @@
  * correct combination of casts.
  */
 
-#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
 #define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT16) (const)))
 #endif
-#ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
-#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT32) (const)))
+#ifdef SHORTxLCONST_32          /* known to work with Microsoft C 6.0 */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((JLONG) (const)))
 #endif
 
-#ifndef MULTIPLY16C16		/* default definition */
+#ifndef MULTIPLY16C16           /* default definition */
 #define MULTIPLY16C16(var,const)  ((var) * (const))
 #endif
 
 /* Same except both inputs are variables. */
 
-#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
 #define MULTIPLY16V16(var1,var2)  (((INT16) (var1)) * ((INT16) (var2)))
 #endif
 
-#ifndef MULTIPLY16V16		/* default definition */
+#ifndef MULTIPLY16V16           /* default definition */
 #define MULTIPLY16V16(var1,var2)  ((var1) * (var2))
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jddctmgr.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jddctmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jddctmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jddctmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jddctmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2002-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2010, 2015, D. R. Commander.
+ * Copyright (C) 2013, MIPS Technologies, Inc., California
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the inverse-DCT management logic.
  * This code selects a particular IDCT implementation to be used,
@@ -20,7 +24,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 #include "jsimddct.h"
 #include "jpegcomp.h"
 
@@ -45,7 +49,7 @@
 /* Private subobject for this module */
 
 typedef struct {
-  struct jpeg_inverse_dct pub;	/* public fields */
+  struct jpeg_inverse_dct pub;  /* public fields */
 
   /* This array contains the IDCT method code that each multiplier table
    * is currently set up for, or -1 if it's not yet set up.
@@ -55,7 +59,7 @@
   int cur_method[MAX_COMPONENTS];
 } my_idct_controller;
 
-typedef my_idct_controller * my_idct_ptr;
+typedef my_idct_controller *my_idct_ptr;
 
 
 /* Allocated multiplier tables: big enough for any supported variant */
@@ -97,7 +101,7 @@
   jpeg_component_info *compptr;
   int method = 0;
   inverse_DCT_method_ptr method_ptr = NULL;
-  JQUANT_TBL * qtbl;
+  JQUANT_TBL *qtbl;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
@@ -106,57 +110,117 @@
 #ifdef IDCT_SCALING_SUPPORTED
     case 1:
       method_ptr = jpeg_idct_1x1;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
       break;
     case 2:
       if (jsimd_can_idct_2x2())
         method_ptr = jsimd_idct_2x2;
       else
         method_ptr = jpeg_idct_2x2;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 3:
+      method_ptr = jpeg_idct_3x3;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
       break;
     case 4:
       if (jsimd_can_idct_4x4())
         method_ptr = jsimd_idct_4x4;
       else
         method_ptr = jpeg_idct_4x4;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 5:
+      method_ptr = jpeg_idct_5x5;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 6:
+#if defined(__mips__)
+      if (jsimd_can_idct_6x6())
+        method_ptr = jsimd_idct_6x6;
+      else
+#endif
+      method_ptr = jpeg_idct_6x6;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 7:
+      method_ptr = jpeg_idct_7x7;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
       break;
 #endif
     case DCTSIZE:
       switch (cinfo->dct_method) {
 #ifdef DCT_ISLOW_SUPPORTED
       case JDCT_ISLOW:
-	if (jsimd_can_idct_islow())
-	  method_ptr = jsimd_idct_islow;
-	else
-	  method_ptr = jpeg_idct_islow;
-	method = JDCT_ISLOW;
-	break;
+        if (jsimd_can_idct_islow())
+          method_ptr = jsimd_idct_islow;
+        else
+          method_ptr = jpeg_idct_islow;
+        method = JDCT_ISLOW;
+        break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
       case JDCT_IFAST:
-	if (jsimd_can_idct_ifast())
-	  method_ptr = jsimd_idct_ifast;
-	else
-	  method_ptr = jpeg_idct_ifast;
-	method = JDCT_IFAST;
-	break;
+        if (jsimd_can_idct_ifast())
+          method_ptr = jsimd_idct_ifast;
+        else
+          method_ptr = jpeg_idct_ifast;
+        method = JDCT_IFAST;
+        break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
       case JDCT_FLOAT:
-	if (jsimd_can_idct_float())
-	  method_ptr = jsimd_idct_float;
-	else
-	  method_ptr = jpeg_idct_float;
-	method = JDCT_FLOAT;
-	break;
+        if (jsimd_can_idct_float())
+          method_ptr = jsimd_idct_float;
+        else
+          method_ptr = jpeg_idct_float;
+        method = JDCT_FLOAT;
+        break;
 #endif
       default:
-	ERREXIT(cinfo, JERR_NOT_COMPILED);
-	break;
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+        break;
       }
       break;
+#ifdef IDCT_SCALING_SUPPORTED
+    case 9:
+      method_ptr = jpeg_idct_9x9;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 10:
+      method_ptr = jpeg_idct_10x10;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 11:
+      method_ptr = jpeg_idct_11x11;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 12:
+#if defined(__mips__)
+      if (jsimd_can_idct_12x12())
+        method_ptr = jsimd_idct_12x12;
+      else
+#endif
+      method_ptr = jpeg_idct_12x12;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 13:
+      method_ptr = jpeg_idct_13x13;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 14:
+      method_ptr = jpeg_idct_14x14;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 15:
+      method_ptr = jpeg_idct_15x15;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 16:
+      method_ptr = jpeg_idct_16x16;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+#endif
     default:
       ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->_DCT_scaled_size);
       break;
@@ -172,81 +236,81 @@
     if (! compptr->component_needed || idct->cur_method[ci] == method)
       continue;
     qtbl = compptr->quant_table;
-    if (qtbl == NULL)		/* happens if no data yet for component */
+    if (qtbl == NULL)           /* happens if no data yet for component */
       continue;
     idct->cur_method[ci] = method;
     switch (method) {
 #ifdef PROVIDE_ISLOW_TABLES
     case JDCT_ISLOW:
       {
-	/* For LL&M IDCT method, multipliers are equal to raw quantization
-	 * coefficients, but are stored as ints to ensure access efficiency.
-	 */
-	ISLOW_MULT_TYPE * ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
-	for (i = 0; i < DCTSIZE2; i++) {
-	  ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
-	}
+        /* For LL&M IDCT method, multipliers are equal to raw quantization
+         * coefficients, but are stored as ints to ensure access efficiency.
+         */
+        ISLOW_MULT_TYPE *ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
+        for (i = 0; i < DCTSIZE2; i++) {
+          ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
+        }
       }
       break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
     case JDCT_IFAST:
       {
-	/* For AA&N IDCT method, multipliers are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * For integer operation, the multiplier table is to be scaled by
-	 * IFAST_SCALE_BITS.
-	 */
-	IFAST_MULT_TYPE * ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
+        /* For AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * For integer operation, the multiplier table is to be scaled by
+         * IFAST_SCALE_BITS.
+         */
+        IFAST_MULT_TYPE *ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
 #define CONST_BITS 14
-	static const INT16 aanscales[DCTSIZE2] = {
-	  /* precomputed values scaled up by 14 bits */
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
-	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
-	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
-	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
-	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
-	};
-	SHIFT_TEMPS
-
-	for (i = 0; i < DCTSIZE2; i++) {
-	  ifmtbl[i] = (IFAST_MULT_TYPE)
-	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
-				  (INT32) aanscales[i]),
-		    CONST_BITS-IFAST_SCALE_BITS);
-	}
+        static const INT16 aanscales[DCTSIZE2] = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        for (i = 0; i < DCTSIZE2; i++) {
+          ifmtbl[i] = (IFAST_MULT_TYPE)
+            DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                  (JLONG) aanscales[i]),
+                    CONST_BITS-IFAST_SCALE_BITS);
+        }
       }
       break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
     case JDCT_FLOAT:
       {
-	/* For float AA&N IDCT method, multipliers are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 */
-	FLOAT_MULT_TYPE * fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
-	int row, col;
-	static const double aanscalefactor[DCTSIZE] = {
-	  1.0, 1.387039845, 1.306562965, 1.175875602,
-	  1.0, 0.785694958, 0.541196100, 0.275899379
-	};
-
-	i = 0;
-	for (row = 0; row < DCTSIZE; row++) {
-	  for (col = 0; col < DCTSIZE; col++) {
-	    fmtbl[i] = (FLOAT_MULT_TYPE)
-	      ((double) qtbl->quantval[i] *
-	       aanscalefactor[row] * aanscalefactor[col]);
-	    i++;
-	  }
-	}
+        /* For float AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         */
+        FLOAT_MULT_TYPE *fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
+        int row, col;
+        static const double aanscalefactor[DCTSIZE] = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fmtbl[i] = (FLOAT_MULT_TYPE)
+              ((double) qtbl->quantval[i] *
+               aanscalefactor[row] * aanscalefactor[col]);
+            i++;
+          }
+        }
       }
       break;
 #endif
@@ -271,7 +335,7 @@
 
   idct = (my_idct_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_idct_controller));
+                                sizeof(my_idct_controller));
   cinfo->idct = (struct jpeg_inverse_dct *) idct;
   idct->pub.start_pass = start_pass;
 
@@ -280,8 +344,8 @@
     /* Allocate and pre-zero a multiplier table for each component */
     compptr->dct_table =
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(multiplier_table));
-    MEMZERO(compptr->dct_table, SIZEOF(multiplier_table));
+                                  sizeof(multiplier_table));
+    MEMZERO(compptr->dct_table, sizeof(multiplier_table));
     /* Mark multiplier table not yet set up for any method */
     idct->cur_method[ci] = -1;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdhuff.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdhuff.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdhuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdhuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdhuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy decoding routines.
  *
@@ -18,8 +20,9 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdhuff.h"		/* Declarations shared with jdphuff.c */
+#include "jdhuff.h"             /* Declarations shared with jdphuff.c */
 #include "jpegcomp.h"
+#include "jstdhuff.c"
 
 
 /*
@@ -43,10 +46,10 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -57,27 +60,27 @@
   /* These fields are loaded into local variables at start of each MCU.
    * In case of suspension, we exit WITHOUT updating them.
    */
-  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
-  savable_state saved;		/* Other state at start of MCU */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  d_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
-  d_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *dc_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *ac_derived_tbls[NUM_HUFF_TBLS];
 
   /* Precalculated info set up by start_pass for use in decode_mcu: */
 
   /* Pointers to derived tables to be used for each block within an MCU */
-  d_derived_tbl * dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
-  d_derived_tbl * ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl *dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl *ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
   /* Whether we care about the DC and AC coefficient values for each block */
   boolean dc_needed[D_MAX_BLOCKS_IN_MCU];
   boolean ac_needed[D_MAX_BLOCKS_IN_MCU];
 } huff_entropy_decoder;
 
-typedef huff_entropy_decoder * huff_entropy_ptr;
+typedef huff_entropy_decoder *huff_entropy_ptr;
 
 
 /*
@@ -89,7 +92,8 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, blkn, dctbl, actbl;
-  jpeg_component_info * compptr;
+  d_derived_tbl **pdtbl;
+  jpeg_component_info *compptr;
 
   /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
    * This ought to be an error condition, but we make it a warning because
@@ -105,10 +109,10 @@
     actbl = compptr->ac_tbl_no;
     /* Compute derived values for Huffman tables */
     /* We may do this more than once for a table, but it's not expensive */
-    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl,
-			    & entropy->dc_derived_tbls[dctbl]);
-    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl,
-			    & entropy->ac_derived_tbls[actbl]);
+    pdtbl = entropy->dc_derived_tbls + dctbl;
+    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl, pdtbl);
+    pdtbl = entropy->ac_derived_tbls + actbl;
+    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl, pdtbl);
     /* Initialize DC predictions to 0 */
     entropy->saved.last_dc_val[ci] = 0;
   }
@@ -149,7 +153,7 @@
 
 GLOBAL(void)
 jpeg_make_d_derived_tbl (j_decompress_ptr cinfo, boolean isDC, int tblno,
-			 d_derived_tbl ** pdtbl)
+                         d_derived_tbl **pdtbl)
 {
   JHUFF_TBL *htbl;
   d_derived_tbl *dtbl;
@@ -175,26 +179,26 @@
   if (*pdtbl == NULL)
     *pdtbl = (d_derived_tbl *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(d_derived_tbl));
+                                  sizeof(d_derived_tbl));
   dtbl = *pdtbl;
-  dtbl->pub = htbl;		/* fill in back link */
-  
+  dtbl->pub = htbl;             /* fill in back link */
+
   /* Figure C.1: make table of Huffman code length for each symbol */
 
   p = 0;
   for (l = 1; l <= 16; l++) {
     i = (int) htbl->bits[l];
-    if (i < 0 || p + i > 256)	/* protect against table overrun */
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     while (i--)
       huffsize[p++] = (char) l;
   }
   huffsize[p] = 0;
   numsymbols = p;
-  
+
   /* Figure C.2: generate the codes themselves */
   /* We also validate that the counts represent a legal Huffman code tree. */
-  
+
   code = 0;
   si = huffsize[0];
   p = 0;
@@ -206,7 +210,7 @@
     /* code is now 1 more than the last code used for codelength si; but
      * it must still fit in si bits, since no code is allowed to be all ones.
      */
-    if (((INT32) code) >= (((INT32) 1) << si))
+    if (((JLONG) code) >= (((JLONG) 1) << si))
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     code <<= 1;
     si++;
@@ -220,11 +224,11 @@
       /* valoffset[l] = huffval[] index of 1st symbol of code length l,
        * minus the minimum code of length l
        */
-      dtbl->valoffset[l] = (INT32) p - (INT32) huffcode[p];
+      dtbl->valoffset[l] = (JLONG) p - (JLONG) huffcode[p];
       p += htbl->bits[l];
       dtbl->maxcode[l] = huffcode[p-1]; /* maximum code of length l */
     } else {
-      dtbl->maxcode[l] = -1;	/* -1 if no codes of this length */
+      dtbl->maxcode[l] = -1;    /* -1 if no codes of this length */
     }
   }
   dtbl->valoffset[17] = 0;
@@ -247,8 +251,8 @@
       /* Generate left-justified code followed by all possible bit sequences */
       lookbits = huffcode[p] << (HUFF_LOOKAHEAD-l);
       for (ctr = 1 << (HUFF_LOOKAHEAD-l); ctr > 0; ctr--) {
-	dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p];
-	lookbits++;
+        dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p];
+        lookbits++;
       }
     }
   }
@@ -263,7 +267,7 @@
     for (i = 0; i < numsymbols; i++) {
       int sym = htbl->huffval[i];
       if (sym < 0 || sym > 15)
-	ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+        ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     }
   }
 }
@@ -285,20 +289,20 @@
  */
 
 #ifdef SLOW_SHIFT_32
-#define MIN_GET_BITS  15	/* minimum allowable value */
+#define MIN_GET_BITS  15        /* minimum allowable value */
 #else
 #define MIN_GET_BITS  (BIT_BUF_SIZE-7)
 #endif
 
 
 GLOBAL(boolean)
-jpeg_fill_bit_buffer (bitread_working_state * state,
-		      register bit_buf_type get_buffer, register int bits_left,
-		      int nbits)
+jpeg_fill_bit_buffer (bitread_working_state *state,
+                      register bit_buf_type get_buffer, register int bits_left,
+                      int nbits)
 /* Load up the bit buffer to a depth of at least nbits */
 {
   /* Copy heavily used state fields into locals (hopefully registers) */
-  register const JOCTET * next_input_byte = state->next_input_byte;
+  register const JOCTET *next_input_byte = state->next_input_byte;
   register size_t bytes_in_buffer = state->bytes_in_buffer;
   j_decompress_ptr cinfo = state->cinfo;
 
@@ -306,54 +310,54 @@
   /* (It is assumed that no request will be for more than that many bits.) */
   /* We fail to do so only if we hit a marker or are forced to suspend. */
 
-  if (cinfo->unread_marker == 0) {	/* cannot advance past a marker */
+  if (cinfo->unread_marker == 0) {      /* cannot advance past a marker */
     while (bits_left < MIN_GET_BITS) {
       register int c;
 
       /* Attempt to read a byte */
       if (bytes_in_buffer == 0) {
-	if (! (*cinfo->src->fill_input_buffer) (cinfo))
-	  return FALSE;
-	next_input_byte = cinfo->src->next_input_byte;
-	bytes_in_buffer = cinfo->src->bytes_in_buffer;
+        if (! (*cinfo->src->fill_input_buffer) (cinfo))
+          return FALSE;
+        next_input_byte = cinfo->src->next_input_byte;
+        bytes_in_buffer = cinfo->src->bytes_in_buffer;
       }
       bytes_in_buffer--;
       c = GETJOCTET(*next_input_byte++);
 
       /* If it's 0xFF, check and discard stuffed zero byte */
       if (c == 0xFF) {
-	/* Loop here to discard any padding FF's on terminating marker,
-	 * so that we can save a valid unread_marker value.  NOTE: we will
-	 * accept multiple FF's followed by a 0 as meaning a single FF data
-	 * byte.  This data pattern is not valid according to the standard.
-	 */
-	do {
-	  if (bytes_in_buffer == 0) {
-	    if (! (*cinfo->src->fill_input_buffer) (cinfo))
-	      return FALSE;
-	    next_input_byte = cinfo->src->next_input_byte;
-	    bytes_in_buffer = cinfo->src->bytes_in_buffer;
-	  }
-	  bytes_in_buffer--;
-	  c = GETJOCTET(*next_input_byte++);
-	} while (c == 0xFF);
-
-	if (c == 0) {
-	  /* Found FF/00, which represents an FF data byte */
-	  c = 0xFF;
-	} else {
-	  /* Oops, it's actually a marker indicating end of compressed data.
-	   * Save the marker code for later use.
-	   * Fine point: it might appear that we should save the marker into
-	   * bitread working state, not straight into permanent state.  But
-	   * once we have hit a marker, we cannot need to suspend within the
-	   * current MCU, because we will read no more bytes from the data
-	   * source.  So it is OK to update permanent state right away.
-	   */
-	  cinfo->unread_marker = c;
-	  /* See if we need to insert some fake zero bits. */
-	  goto no_more_bytes;
-	}
+        /* Loop here to discard any padding FF's on terminating marker,
+         * so that we can save a valid unread_marker value.  NOTE: we will
+         * accept multiple FF's followed by a 0 as meaning a single FF data
+         * byte.  This data pattern is not valid according to the standard.
+         */
+        do {
+          if (bytes_in_buffer == 0) {
+            if (! (*cinfo->src->fill_input_buffer) (cinfo))
+              return FALSE;
+            next_input_byte = cinfo->src->next_input_byte;
+            bytes_in_buffer = cinfo->src->bytes_in_buffer;
+          }
+          bytes_in_buffer--;
+          c = GETJOCTET(*next_input_byte++);
+        } while (c == 0xFF);
+
+        if (c == 0) {
+          /* Found FF/00, which represents an FF data byte */
+          c = 0xFF;
+        } else {
+          /* Oops, it's actually a marker indicating end of compressed data.
+           * Save the marker code for later use.
+           * Fine point: it might appear that we should save the marker into
+           * bitread working state, not straight into permanent state.  But
+           * once we have hit a marker, we cannot need to suspend within the
+           * current MCU, because we will read no more bytes from the data
+           * source.  So it is OK to update permanent state right away.
+           */
+          cinfo->unread_marker = c;
+          /* See if we need to insert some fake zero bits. */
+          goto no_more_bytes;
+        }
       }
 
       /* OK, load c into get_buffer */
@@ -373,8 +377,8 @@
        * appears per data segment.
        */
       if (! cinfo->entropy->insufficient_data) {
-	WARNMS(cinfo, JWRN_HIT_MARKER);
-	cinfo->entropy->insufficient_data = TRUE;
+        WARNMS(cinfo, JWRN_HIT_MARKER);
+        cinfo->entropy->insufficient_data = TRUE;
       }
       /* Fill the buffer with zero bits */
       get_buffer <<= MIN_GET_BITS - bits_left;
@@ -417,11 +421,11 @@
   } \
 }
 
-#if __WORDSIZE == 64 || defined(_WIN64)
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
 /* Pre-fetch 48 bytes, because the holding register is 64-bit */
 #define FILL_BIT_BUFFER_FAST \
-  if (bits_left < 16) { \
+  if (bits_left <= 16) { \
     GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE \
   }
 
@@ -429,7 +433,7 @@
 
 /* Pre-fetch 16 bytes, because the holding register is 32-bit */
 #define FILL_BIT_BUFFER_FAST \
-  if (bits_left < 16) { \
+  if (bits_left <= 16) { \
     GET_BYTE GET_BYTE \
   }
 
@@ -442,12 +446,12 @@
  */
 
 GLOBAL(int)
-jpeg_huff_decode (bitread_working_state * state,
-		  register bit_buf_type get_buffer, register int bits_left,
-		  d_derived_tbl * htbl, int min_bits)
+jpeg_huff_decode (bitread_working_state *state,
+                  register bit_buf_type get_buffer, register int bits_left,
+                  d_derived_tbl *htbl, int min_bits)
 {
   register int l = min_bits;
-  register INT32 code;
+  register JLONG code;
 
   /* HUFF_DECODE has determined that the code is at least min_bits */
   /* bits long, so fetch that many bits in one swoop. */
@@ -473,7 +477,7 @@
 
   if (l > 16) {
     WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
-    return 0;			/* fake a zero as the safest result */
+    return 0;                   /* fake a zero as the safest result */
   }
 
   return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ];
@@ -488,7 +492,8 @@
 #define AVOID_TABLES
 #ifdef AVOID_TABLES
 
-#define HUFF_EXTEND(x,s)  ((x) + ((((x) - (1<<((s)-1))) >> 31) & (((-1)<<(s)) + 1)))
+#define NEG_1 ((unsigned int)-1)
+#define HUFF_EXTEND(x,s)  ((x) + ((((x) - (1<<((s)-1))) >> 31) & (((NEG_1)<<(s)) + 1)))
 
 #else
 
@@ -560,9 +565,9 @@
   ASSIGN_STATE(state, entropy->saved);
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    JBLOCKROW block = MCU_data[blkn];
-    d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
-    d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
+    d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
     register int s, k, r;
 
     /* Decode a single block's worth of coefficients */
@@ -580,11 +585,13 @@
       int ci = cinfo->MCU_membership[blkn];
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
-      /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
-      (*block)[0] = (JCOEF) s;
+      if (block) {
+        /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
+        (*block)[0] = (JCOEF) s;
+      }
     }
 
-    if (entropy->ac_needed[blkn]) {
+    if (entropy->ac_needed[blkn] && block) {
 
       /* Section F.2.2.2: decode the AC coefficients */
       /* Since zeroes are skipped, output area must be cleared beforehand */
@@ -593,7 +600,7 @@
 
         r = s >> 4;
         s &= 15;
-      
+
         if (s) {
           k += r;
           CHECK_BIT_BUFFER(br_state, s, return FALSE);
@@ -657,12 +664,12 @@
   ASSIGN_STATE(state, entropy->saved);
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    JBLOCKROW block = MCU_data[blkn];
-    d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
-    d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
+    d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
     register int s, k, r, l;
 
-    HUFF_DECODE_FAST(s, l, dctbl);
+    HUFF_DECODE_FAST(s, l, dctbl, slow_decode_mcu);
     if (s) {
       FILL_BIT_BUFFER_FAST
       r = GET_BITS(s);
@@ -673,16 +680,17 @@
       int ci = cinfo->MCU_membership[blkn];
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
-      (*block)[0] = (JCOEF) s;
+      if (block)
+        (*block)[0] = (JCOEF) s;
     }
 
-    if (entropy->ac_needed[blkn]) {
+    if (entropy->ac_needed[blkn] && block) {
 
       for (k = 1; k < DCTSIZE2; k++) {
-        HUFF_DECODE_FAST(s, l, actbl);
+        HUFF_DECODE_FAST(s, l, actbl, slow_decode_mcu);
         r = s >> 4;
         s &= 15;
-      
+
         if (s) {
           k += r;
           FILL_BIT_BUFFER_FAST
@@ -698,7 +706,7 @@
     } else {
 
       for (k = 1; k < DCTSIZE2; k++) {
-        HUFF_DECODE_FAST(s, l, actbl);
+        HUFF_DECODE_FAST(s, l, actbl, slow_decode_mcu);
         r = s >> 4;
         s &= 15;
 
@@ -715,6 +723,7 @@
   }
 
   if (cinfo->unread_marker != 0) {
+slow_decode_mcu:
     cinfo->unread_marker = 0;
     return FALSE;
   }
@@ -742,7 +751,7 @@
  * this module, since we'll just re-assign them on the next call.)
  */
 
-#define BUFSIZE (DCTSIZE2 * 2)
+#define BUFSIZE (DCTSIZE2 * 8)
 
 METHODDEF(boolean)
 decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
@@ -754,7 +763,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
     usefast = 0;
   }
 
@@ -794,9 +803,15 @@
   huff_entropy_ptr entropy;
   int i;
 
+  /* Motion JPEG frames typically do not include the Huffman tables if they
+     are the default tables.  Thus, if the tables are not set by the time
+     the Huffman decoder is initialized (usually within the body of
+     jpeg_start_decompress()), we set them to default values. */
+  std_huff_tables((j_common_ptr) cinfo);
+
   entropy = (huff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(huff_entropy_decoder));
+                                sizeof(huff_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass_huff_decoder;
   entropy->pub.decode_mcu = decode_mcu;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdhuff.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdhuff.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdhuff.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdhuff.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,34 +1,30 @@
 /*
  * jdhuff.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2010-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2015-2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains declarations for Huffman entropy decoding routines
  * that are shared between the sequential decoder (jdhuff.c) and the
  * progressive decoder (jdphuff.c).  No other modules need to see these.
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_make_d_derived_tbl	jMkDDerived
-#define jpeg_fill_bit_buffer	jFilBitBuf
-#define jpeg_huff_decode	jHufDecode
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#include "jconfigint.h"
 
 
 /* Derived data constructed for each Huffman table */
 
-#define HUFF_LOOKAHEAD	8	/* # of bits of lookahead */
+#define HUFF_LOOKAHEAD  8       /* # of bits of lookahead */
 
 typedef struct {
   /* Basic tables: (element [0] of each array is unused) */
-  INT32 maxcode[18];		/* largest code of length k (-1 if none) */
+  JLONG maxcode[18];            /* largest code of length k (-1 if none) */
   /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */
-  INT32 valoffset[18];		/* huffval[] offset for codes of length k */
+  JLONG valoffset[18];          /* huffval[] offset for codes of length k */
   /* valoffset[k] = huffval[] index of 1st symbol of code length k, less
    * the smallest code of length k; so given a code of length k, the
    * corresponding symbol is huffval[code + valoffset[k]]
@@ -52,8 +48,8 @@
 
 /* Expand a Huffman table definition into the derived format */
 EXTERN(void) jpeg_make_d_derived_tbl
-	JPP((j_decompress_ptr cinfo, boolean isDC, int tblno,
-	     d_derived_tbl ** pdtbl));
+        (j_decompress_ptr cinfo, boolean isDC, int tblno,
+         d_derived_tbl ** pdtbl);
 
 
 /*
@@ -74,15 +70,19 @@
  * necessary.
  */
 
-#if __WORDSIZE == 64 || defined(_WIN64)
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
-typedef size_t bit_buf_type;	/* type of bit-extraction buffer */
-#define BIT_BUF_SIZE  64		/* size of buffer in bits */
+typedef size_t bit_buf_type;            /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  64                /* size of buffer in bits */
 
 #else
 
-typedef INT32 bit_buf_type;	/* type of bit-extraction buffer */
-#define BIT_BUF_SIZE  32		/* size of buffer in bits */
+typedef unsigned long bit_buf_type;     /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  32                /* size of buffer in bits */
 
 #endif
 
@@ -93,43 +93,43 @@
  * because not all machines measure sizeof in 8-bit bytes.
  */
 
-typedef struct {		/* Bitreading state saved across MCUs */
-  bit_buf_type get_buffer;	/* current bit-extraction buffer */
-  int bits_left;		/* # of unused bits in it */
+typedef struct {                /* Bitreading state saved across MCUs */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
 } bitread_perm_state;
 
-typedef struct {		/* Bitreading working state within an MCU */
+typedef struct {                /* Bitreading working state within an MCU */
   /* Current data source location */
   /* We need a copy, rather than munging the original, in case of suspension */
-  const JOCTET * next_input_byte; /* => next byte to read from source */
-  size_t bytes_in_buffer;	/* # of bytes remaining in source buffer */
+  const JOCTET *next_input_byte; /* => next byte to read from source */
+  size_t bytes_in_buffer;       /* # of bytes remaining in source buffer */
   /* Bit input buffer --- note these values are kept in register variables,
    * not in this struct, inside the inner loops.
    */
-  bit_buf_type get_buffer;	/* current bit-extraction buffer */
-  int bits_left;		/* # of unused bits in it */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
   /* Pointer needed by jpeg_fill_bit_buffer. */
-  j_decompress_ptr cinfo;	/* back link to decompress master record */
+  j_decompress_ptr cinfo;       /* back link to decompress master record */
 } bitread_working_state;
 
 /* Macros to declare and load/save bitread local variables. */
 #define BITREAD_STATE_VARS  \
-	register bit_buf_type get_buffer;  \
-	register int bits_left;  \
-	bitread_working_state br_state
+        register bit_buf_type get_buffer;  \
+        register int bits_left;  \
+        bitread_working_state br_state
 
 #define BITREAD_LOAD_STATE(cinfop,permstate)  \
-	br_state.cinfo = cinfop; \
-	br_state.next_input_byte = cinfop->src->next_input_byte; \
-	br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
-	get_buffer = permstate.get_buffer; \
-	bits_left = permstate.bits_left;
+        br_state.cinfo = cinfop; \
+        br_state.next_input_byte = cinfop->src->next_input_byte; \
+        br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
+        get_buffer = permstate.get_buffer; \
+        bits_left = permstate.bits_left;
 
 #define BITREAD_SAVE_STATE(cinfop,permstate)  \
-	cinfop->src->next_input_byte = br_state.next_input_byte; \
-	cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
-	permstate.get_buffer = get_buffer; \
-	permstate.bits_left = bits_left
+        cinfop->src->next_input_byte = br_state.next_input_byte; \
+        cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
+        permstate.get_buffer = get_buffer; \
+        permstate.bits_left = bits_left
 
 /*
  * These macros provide the in-line portion of bit fetching.
@@ -137,37 +137,37 @@
  * before using GET_BITS, PEEK_BITS, or DROP_BITS.
  * The variables get_buffer and bits_left are assumed to be locals,
  * but the state struct might not be (jpeg_huff_decode needs this).
- *	CHECK_BIT_BUFFER(state,n,action);
- *		Ensure there are N bits in get_buffer; if suspend, take action.
+ *      CHECK_BIT_BUFFER(state,n,action);
+ *              Ensure there are N bits in get_buffer; if suspend, take action.
  *      val = GET_BITS(n);
- *		Fetch next N bits.
+ *              Fetch next N bits.
  *      val = PEEK_BITS(n);
- *		Fetch next N bits without removing them from the buffer.
- *	DROP_BITS(n);
- *		Discard next N bits.
+ *              Fetch next N bits without removing them from the buffer.
+ *      DROP_BITS(n);
+ *              Discard next N bits.
  * The value N should be a simple variable, not an expression, because it
  * is evaluated multiple times.
  */
 
 #define CHECK_BIT_BUFFER(state,nbits,action) \
-	{ if (bits_left < (nbits)) {  \
-	    if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
-	      { action; }  \
-	    get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
+        { if (bits_left < (nbits)) {  \
+            if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
+              { action; }  \
+            get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
 
 #define GET_BITS(nbits) \
-	(((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
+        (((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
 
 #define PEEK_BITS(nbits) \
-	(((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
+        (((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
 
 #define DROP_BITS(nbits) \
-	(bits_left -= (nbits))
+        (bits_left -= (nbits))
 
 /* Load up the bit buffer to a depth of at least nbits */
 EXTERN(boolean) jpeg_fill_bit_buffer
-	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
-	     register int bits_left, int nbits));
+        (bitread_working_state *state, register bit_buf_type get_buffer,
+         register int bits_left, int nbits);
 
 
 /*
@@ -203,12 +203,12 @@
   } else { \
 slowlabel: \
     if ((result=jpeg_huff_decode(&state,get_buffer,bits_left,htbl,nb)) < 0) \
-	{ failaction; } \
+        { failaction; } \
     get_buffer = state.get_buffer; bits_left = state.bits_left; \
   } \
 }
 
-#define HUFF_DECODE_FAST(s,nb,htbl) \
+#define HUFF_DECODE_FAST(s,nb,htbl,slowlabel) \
   FILL_BIT_BUFFER_FAST; \
   s = PEEK_BITS(HUFF_LOOKAHEAD); \
   s = htbl->lookup[s]; \
@@ -225,10 +225,12 @@
       s |= GET_BITS(1); \
       nb++; \
     } \
-    s = htbl->pub->huffval[ (int) (s + htbl->valoffset[nb]) & 0xFF ]; \
+    if (nb > 16) \
+      goto slowlabel; \
+    s = htbl->pub->huffval[ (int) (s + htbl->valoffset[nb]) ]; \
   }
 
 /* Out-of-line case for Huffman code fetching */
 EXTERN(int) jpeg_huff_decode
-	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
-	     register int bits_left, d_derived_tbl * htbl, int min_bits));
+        (bitread_working_state *state, register bit_buf_type get_buffer,
+         register int bits_left, d_derived_tbl *htbl, int min_bits);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdinput.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdinput.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdinput.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdinput.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,13 @@
 /*
  * jdinput.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Modified 2002-2009 by Guido Vollbeding.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains input control logic for the JPEG decompressor.
  * These routines are concerned with controlling the decompressor's input
@@ -24,93 +26,20 @@
 typedef struct {
   struct jpeg_input_controller pub; /* public fields */
 
-  boolean inheaders;		/* TRUE until first SOS is reached */
+  boolean inheaders;            /* TRUE until first SOS is reached */
 } my_input_controller;
 
-typedef my_input_controller * my_inputctl_ptr;
+typedef my_input_controller *my_inputctl_ptr;
 
 
 /* Forward declarations */
-METHODDEF(int) consume_markers JPP((j_decompress_ptr cinfo));
+METHODDEF(int) consume_markers (j_decompress_ptr cinfo);
 
 
 /*
  * Routines to calculate various quantities related to the size of the image.
  */
 
-
-#if JPEG_LIB_VERSION >= 80
-/*
- * Compute output image dimensions and related values.
- * NOTE: this is exported for possible use by application.
- * Hence it mustn't do anything that can't be done twice.
- */
-
-GLOBAL(void)
-jpeg_core_output_dimensions (j_decompress_ptr cinfo)
-/* Do computations that are needed before master selection phase.
- * This function is used for transcoding and full decompression.
- */
-{
-#ifdef IDCT_SCALING_SUPPORTED
-  int ci;
-  jpeg_component_info *compptr;
-
-  /* Compute actual output image dimensions and DCT scaling choices. */
-  if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom) {
-    /* Provide 1/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 1;
-    cinfo->min_DCT_v_scaled_size = 1;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 2) {
-    /* Provide 2/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 2L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 2L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 2;
-    cinfo->min_DCT_v_scaled_size = 2;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 4) {
-    /* Provide 4/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 4L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 4L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 4;
-    cinfo->min_DCT_v_scaled_size = 4;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 8) {
-    /* Provide 8/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 8L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 8L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 8;
-    cinfo->min_DCT_v_scaled_size = 8;
-  }
-  /* Recompute dimensions of components */
-  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-       ci++, compptr++) {
-    compptr->DCT_h_scaled_size = cinfo->min_DCT_h_scaled_size;
-    compptr->DCT_v_scaled_size = cinfo->min_DCT_v_scaled_size;
-  }
-
-#else /* !IDCT_SCALING_SUPPORTED */
-
-  /* Hardwire it to "no scaling" */
-  cinfo->output_width = cinfo->image_width;
-  cinfo->output_height = cinfo->image_height;
-  /* jdinput.c has already initialized DCT_scaled_size,
-   * and has computed unscaled downsampled_width and downsampled_height.
-   */
-
-#endif /* IDCT_SCALING_SUPPORTED */
-}
-#endif
-
-
 LOCAL(void)
 initial_setup (j_decompress_ptr cinfo)
 /* Called once, when first SOS marker is reached */
@@ -130,7 +59,7 @@
   /* Check that number of components won't exceed internal array sizes */
   if (cinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
 
   /* Compute maximum sampling factors; check factor validity */
   cinfo->max_h_samp_factor = 1;
@@ -138,12 +67,12 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
-	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+        compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
       ERREXIT(cinfo, JERR_BAD_SAMPLING);
     cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
-				   compptr->h_samp_factor);
+                                   compptr->h_samp_factor);
     cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
-				   compptr->v_samp_factor);
+                                   compptr->v_samp_factor);
   }
 
 #if JPEG_LIB_VERSION >=80
@@ -173,10 +102,15 @@
     /* Size in DCT blocks */
     compptr->width_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->height_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* Set the first and last MCU columns to decompress from multi-scan images.
+     * By default, decompress all of the MCU columns.
+     */
+    cinfo->master->first_MCU_col[ci] = 0;
+    cinfo->master->last_MCU_col[ci] = compptr->width_in_blocks - 1;
     /* downsampled_width and downsampled_height will also be overridden by
      * jdmaster.c if we are doing full decompression.  The transcoder library
      * doesn't use these values, but the calling application might.
@@ -184,10 +118,10 @@
     /* Size in samples */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
-		    (long) cinfo->max_h_samp_factor);
+                    (long) cinfo->max_h_samp_factor);
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
-		    (long) cinfo->max_v_samp_factor);
+                    (long) cinfo->max_v_samp_factor);
     /* Mark component needed, until color conversion says otherwise */
     compptr->component_needed = TRUE;
     /* Mark no quantization table yet saved for component */
@@ -197,7 +131,7 @@
   /* Compute number of fully interleaved MCU rows. */
   cinfo->total_iMCU_rows = (JDIMENSION)
     jdiv_round_up((long) cinfo->image_height,
-		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+                  (long) (cinfo->max_v_samp_factor*DCTSIZE));
 
   /* Decide whether file contains multiple scans */
   if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode)
@@ -214,16 +148,16 @@
 {
   int ci, mcublks, tmp;
   jpeg_component_info *compptr;
-  
+
   if (cinfo->comps_in_scan == 1) {
-    
+
     /* Noninterleaved (single-component) scan */
     compptr = cinfo->cur_comp_info[0];
-    
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = compptr->width_in_blocks;
     cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
-    
+
     /* For noninterleaved scan, always one block per MCU */
     compptr->MCU_width = 1;
     compptr->MCU_height = 1;
@@ -236,28 +170,28 @@
     tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
     if (tmp == 0) tmp = compptr->v_samp_factor;
     compptr->last_row_height = tmp;
-    
+
     /* Prepare array describing MCU composition */
     cinfo->blocks_in_MCU = 1;
     cinfo->MCU_membership[0] = 0;
-    
+
   } else {
-    
+
     /* Interleaved (multi-component) scan */
     if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
-	       MAX_COMPS_IN_SCAN);
-    
+               MAX_COMPS_IN_SCAN);
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width,
-		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor*DCTSIZE));
     cinfo->MCU_rows_in_scan = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height,
-		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
-    
+                    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
     cinfo->blocks_in_MCU = 0;
-    
+
     for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
       compptr = cinfo->cur_comp_info[ci];
       /* Sampling factors give # of blocks of component in each MCU */
@@ -275,12 +209,12 @@
       /* Prepare array describing MCU composition */
       mcublks = compptr->MCU_blocks;
       if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU)
-	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
       while (mcublks-- > 0) {
-	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+        cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
       }
     }
-    
+
   }
 }
 
@@ -311,7 +245,7 @@
 {
   int ci, qtblno;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtbl;
+  JQUANT_TBL *qtbl;
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
@@ -321,13 +255,13 @@
     /* Make sure specified quantization table is present */
     qtblno = compptr->quant_tbl_no;
     if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
-	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+        cinfo->quant_tbl_ptrs[qtblno] == NULL)
       ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
     /* OK, save away the quantization table */
     qtbl = (JQUANT_TBL *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(JQUANT_TBL));
-    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], SIZEOF(JQUANT_TBL));
+                                  sizeof(JQUANT_TBL));
+    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], sizeof(JQUANT_TBL));
     compptr->quant_table = qtbl;
   }
 }
@@ -386,31 +320,31 @@
   val = (*cinfo->marker->read_markers) (cinfo);
 
   switch (val) {
-  case JPEG_REACHED_SOS:	/* Found SOS */
-    if (inputctl->inheaders) {	/* 1st SOS */
+  case JPEG_REACHED_SOS:        /* Found SOS */
+    if (inputctl->inheaders) {  /* 1st SOS */
       initial_setup(cinfo);
       inputctl->inheaders = FALSE;
       /* Note: start_input_pass must be called by jdmaster.c
        * before any more input can be consumed.  jdapimin.c is
        * responsible for enforcing this sequencing.
        */
-    } else {			/* 2nd or later SOS marker */
+    } else {                    /* 2nd or later SOS marker */
       if (! inputctl->pub.has_multiple_scans)
-	ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
+        ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
       start_input_pass(cinfo);
     }
     break;
-  case JPEG_REACHED_EOI:	/* Found EOI */
+  case JPEG_REACHED_EOI:        /* Found EOI */
     inputctl->pub.eoi_reached = TRUE;
-    if (inputctl->inheaders) {	/* Tables-only datastream, apparently */
+    if (inputctl->inheaders) {  /* Tables-only datastream, apparently */
       if (cinfo->marker->saw_SOF)
-	ERREXIT(cinfo, JERR_SOF_NO_SOS);
+        ERREXIT(cinfo, JERR_SOF_NO_SOS);
     } else {
       /* Prevent infinite loop in coef ctlr's decompress_data routine
        * if user set output_scan_number larger than number of scans.
        */
       if (cinfo->output_scan_number > cinfo->input_scan_number)
-	cinfo->output_scan_number = cinfo->input_scan_number;
+        cinfo->output_scan_number = cinfo->input_scan_number;
     }
     break;
   case JPEG_SUSPENDED:
@@ -455,7 +389,7 @@
   /* Create subobject in permanent pool */
   inputctl = (my_inputctl_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				SIZEOF(my_input_controller));
+                                sizeof(my_input_controller));
   cinfo->inputctl = (struct jpeg_input_controller *) inputctl;
   /* Initialize method pointers */
   inputctl->pub.consume_input = consume_markers;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmainct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmainct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmainct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmainct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdmainct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the main buffer controller for decompression.
  * The main buffer lies between the JPEG decompressor proper and the
@@ -14,10 +16,8 @@
  * supplies the equivalent of the main buffer in that case.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
-#include "jpegcomp.h"
+#include "jdmainct.h"
 
 
 /*
@@ -111,47 +111,17 @@
  */
 
 
-/* Private buffer controller object */
-
-typedef struct {
-  struct jpeg_d_main_controller pub; /* public fields */
-
-  /* Pointer to allocated workspace (M or M+2 row groups). */
-  JSAMPARRAY buffer[MAX_COMPONENTS];
-
-  boolean buffer_full;		/* Have we gotten an iMCU row from decoder? */
-  JDIMENSION rowgroup_ctr;	/* counts row groups output to postprocessor */
-
-  /* Remaining fields are only used in the context case. */
-
-  /* These are the master pointers to the funny-order pointer lists. */
-  JSAMPIMAGE xbuffer[2];	/* pointers to weird pointer lists */
-
-  int whichptr;			/* indicates which pointer set is now in use */
-  int context_state;		/* process_data state machine status */
-  JDIMENSION rowgroups_avail;	/* row groups available to postprocessor */
-  JDIMENSION iMCU_row_ctr;	/* counts iMCU rows to detect image top/bot */
-} my_main_controller;
-
-typedef my_main_controller * my_main_ptr;
-
-/* context_state values: */
-#define CTX_PREPARE_FOR_IMCU	0	/* need to prepare for MCU row */
-#define CTX_PROCESS_IMCU	1	/* feeding iMCU to postprocessor */
-#define CTX_POSTPONED_ROW	2	/* feeding postponed row group */
-
-
 /* Forward declarations */
 METHODDEF(void) process_data_simple_main
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 METHODDEF(void) process_data_context_main
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 #ifdef QUANT_2PASS_SUPPORTED
 METHODDEF(void) process_data_crank_post
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 #endif
 
 
@@ -172,7 +142,7 @@
    */
   main_ptr->xbuffer[0] = (JSAMPIMAGE)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
+                                cinfo->num_components * 2 * sizeof(JSAMPARRAY));
   main_ptr->xbuffer[1] = main_ptr->xbuffer[0] + cinfo->num_components;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -184,8 +154,8 @@
      */
     xbuf = (JSAMPARRAY)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
-    xbuf += rgroup;		/* want one row group at negative offsets */
+                                  2 * (rgroup * (M + 4)) * sizeof(JSAMPROW));
+    xbuf += rgroup;             /* want one row group at negative offsets */
     main_ptr->xbuffer[0][ci] = xbuf;
     xbuf += rgroup * (M + 4);
     main_ptr->xbuffer[1][ci] = xbuf;
@@ -237,34 +207,6 @@
 
 
 LOCAL(void)
-set_wraparound_pointers (j_decompress_ptr cinfo)
-/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
- * This changes the pointer list state from top-of-image to the normal state.
- */
-{
-  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
-  int ci, i, rgroup;
-  int M = cinfo->_min_DCT_scaled_size;
-  jpeg_component_info *compptr;
-  JSAMPARRAY xbuf0, xbuf1;
-
-  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-       ci++, compptr++) {
-    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
-      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
-    xbuf0 = main_ptr->xbuffer[0][ci];
-    xbuf1 = main_ptr->xbuffer[1][ci];
-    for (i = 0; i < rgroup; i++) {
-      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
-      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
-      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
-      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
-    }
-  }
-}
-
-
-LOCAL(void)
 set_bottom_pointers (j_decompress_ptr cinfo)
 /* Change the pointer lists to duplicate the last sample row at the bottom
  * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
@@ -315,14 +257,14 @@
     if (cinfo->upsample->need_context_rows) {
       main_ptr->pub.process_data = process_data_context_main;
       make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
-      main_ptr->whichptr = 0;	/* Read first iMCU row into xbuffer[0] */
+      main_ptr->whichptr = 0;   /* Read first iMCU row into xbuffer[0] */
       main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
       main_ptr->iMCU_row_ctr = 0;
     } else {
       /* Simple case with no context needed */
       main_ptr->pub.process_data = process_data_simple_main;
     }
-    main_ptr->buffer_full = FALSE;	/* Mark buffer empty */
+    main_ptr->buffer_full = FALSE;      /* Mark buffer empty */
     main_ptr->rowgroup_ctr = 0;
     break;
 #ifdef QUANT_2PASS_SUPPORTED
@@ -345,8 +287,8 @@
 
 METHODDEF(void)
 process_data_simple_main (j_decompress_ptr cinfo,
-			  JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			  JDIMENSION out_rows_avail)
+                          JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                          JDIMENSION out_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
   JDIMENSION rowgroups_avail;
@@ -354,8 +296,8 @@
   /* Read input data if we haven't filled the main buffer yet */
   if (! main_ptr->buffer_full) {
     if (! (*cinfo->coef->decompress_data) (cinfo, main_ptr->buffer))
-      return;			/* suspension forced, can do nothing more */
-    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
   }
 
   /* There are always min_DCT_scaled_size row groups in an iMCU row. */
@@ -367,8 +309,8 @@
 
   /* Feed the postprocessor */
   (*cinfo->post->post_process_data) (cinfo, main_ptr->buffer,
-				     &main_ptr->rowgroup_ctr, rowgroups_avail,
-				     output_buf, out_row_ctr, out_rows_avail);
+                                     &main_ptr->rowgroup_ctr, rowgroups_avail,
+                                     output_buf, out_row_ctr, out_rows_avail);
 
   /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
   if (main_ptr->rowgroup_ctr >= rowgroups_avail) {
@@ -385,18 +327,18 @@
 
 METHODDEF(void)
 process_data_context_main (j_decompress_ptr cinfo,
-			   JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			   JDIMENSION out_rows_avail)
+                           JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                           JDIMENSION out_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
 
   /* Read input data if we haven't filled the main buffer yet */
   if (! main_ptr->buffer_full) {
     if (! (*cinfo->coef->decompress_data) (cinfo,
-					   main_ptr->xbuffer[main_ptr->whichptr]))
-      return;			/* suspension forced, can do nothing more */
-    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
-    main_ptr->iMCU_row_ctr++;	/* count rows received */
+                                           main_ptr->xbuffer[main_ptr->whichptr]))
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
+    main_ptr->iMCU_row_ctr++;   /* count rows received */
   }
 
   /* Postprocessor typically will not swallow all the input data it is handed
@@ -408,13 +350,13 @@
   case CTX_POSTPONED_ROW:
     /* Call postprocessor using previously set pointers for postponed row */
     (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
-			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
-			output_buf, out_row_ctr, out_rows_avail);
+                        &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+                        output_buf, out_row_ctr, out_rows_avail);
     if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
-      return;			/* Need to suspend */
+      return;                   /* Need to suspend */
     main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
     if (*out_row_ctr >= out_rows_avail)
-      return;			/* Postprocessor exactly filled output buf */
+      return;                   /* Postprocessor exactly filled output buf */
     /*FALLTHROUGH*/
   case CTX_PREPARE_FOR_IMCU:
     /* Prepare to process first M-1 row groups of this iMCU row */
@@ -430,15 +372,15 @@
   case CTX_PROCESS_IMCU:
     /* Call postprocessor using previously set pointers */
     (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
-			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
-			output_buf, out_row_ctr, out_rows_avail);
+                        &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+                        output_buf, out_row_ctr, out_rows_avail);
     if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
-      return;			/* Need to suspend */
+      return;                   /* Need to suspend */
     /* After the first iMCU, change wraparound pointers to normal state */
     if (main_ptr->iMCU_row_ctr == 1)
       set_wraparound_pointers(cinfo);
     /* Prepare to load new iMCU row using other xbuffer list */
-    main_ptr->whichptr ^= 1;	/* 0=>1 or 1=>0 */
+    main_ptr->whichptr ^= 1;    /* 0=>1 or 1=>0 */
     main_ptr->buffer_full = FALSE;
     /* Still need to process last row group of this iMCU row, */
     /* which is saved at index M+1 of the other xbuffer */
@@ -459,12 +401,12 @@
 
 METHODDEF(void)
 process_data_crank_post (j_decompress_ptr cinfo,
-			 JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			 JDIMENSION out_rows_avail)
+                         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                         JDIMENSION out_rows_avail)
 {
   (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
-				     (JDIMENSION *) NULL, (JDIMENSION) 0,
-				     output_buf, out_row_ctr, out_rows_avail);
+                                     (JDIMENSION *) NULL, (JDIMENSION) 0,
+                                     output_buf, out_row_ctr, out_rows_avail);
 }
 
 #endif /* QUANT_2PASS_SUPPORTED */
@@ -483,11 +425,11 @@
 
   main_ptr = (my_main_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_main_controller));
+                                sizeof(my_main_controller));
   cinfo->main = (struct jpeg_d_main_controller *) main_ptr;
   main_ptr->pub.start_pass = start_pass_main;
 
-  if (need_full_buffer)		/* shouldn't happen */
+  if (need_full_buffer)         /* shouldn't happen */
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 
   /* Allocate the workspace.
@@ -507,8 +449,8 @@
     rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
       cinfo->_min_DCT_scaled_size; /* height of a row group of component */
     main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
-			((j_common_ptr) cinfo, JPOOL_IMAGE,
-			 compptr->width_in_blocks * compptr->_DCT_scaled_size,
-			 (JDIMENSION) (rgroup * ngroups));
+                        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                         compptr->width_in_blocks * compptr->_DCT_scaled_size,
+                         (JDIMENSION) (rgroup * ngroups));
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmainct.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmainct.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmainct.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmainct.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,71 @@
+/*
+ * jdmainct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_main_controller pub; /* public fields */
+
+  /* Pointer to allocated workspace (M or M+2 row groups). */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+  boolean buffer_full;          /* Have we gotten an iMCU row from decoder? */
+  JDIMENSION rowgroup_ctr;      /* counts row groups output to postprocessor */
+
+  /* Remaining fields are only used in the context case. */
+
+  /* These are the master pointers to the funny-order pointer lists. */
+  JSAMPIMAGE xbuffer[2];        /* pointers to weird pointer lists */
+
+  int whichptr;                 /* indicates which pointer set is now in use */
+  int context_state;            /* process_data state machine status */
+  JDIMENSION rowgroups_avail;   /* row groups available to postprocessor */
+  JDIMENSION iMCU_row_ctr;      /* counts iMCU rows to detect image top/bot */
+} my_main_controller;
+
+typedef my_main_controller *my_main_ptr;
+
+
+/* context_state values: */
+#define CTX_PREPARE_FOR_IMCU    0       /* need to prepare for MCU row */
+#define CTX_PROCESS_IMCU        1       /* feeding iMCU to postprocessor */
+#define CTX_POSTPONED_ROW       2       /* feeding postponed row group */
+
+
+LOCAL(void)
+set_wraparound_pointers (j_decompress_ptr cinfo)
+/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
+ * This changes the pointer list state from top-of-image to the normal state.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->_min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer[0][ci];
+    xbuf1 = main_ptr->xbuffer[1][ci];
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
+      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
+      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
+      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmarker.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmarker.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmarker.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmarker.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdmarker.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * Copyright (C) 2012, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains routines to decode JPEG datastream markers.
  * Most of the complexity arises from our desire to support input
@@ -18,29 +20,29 @@
 #include "jpeglib.h"
 
 
-typedef enum {			/* JPEG marker codes */
+typedef enum {                  /* JPEG marker codes */
   M_SOF0  = 0xc0,
   M_SOF1  = 0xc1,
   M_SOF2  = 0xc2,
   M_SOF3  = 0xc3,
-  
+
   M_SOF5  = 0xc5,
   M_SOF6  = 0xc6,
   M_SOF7  = 0xc7,
-  
+
   M_JPG   = 0xc8,
   M_SOF9  = 0xc9,
   M_SOF10 = 0xca,
   M_SOF11 = 0xcb,
-  
+
   M_SOF13 = 0xcd,
   M_SOF14 = 0xce,
   M_SOF15 = 0xcf,
-  
+
   M_DHT   = 0xc4,
-  
+
   M_DAC   = 0xcc,
-  
+
   M_RST0  = 0xd0,
   M_RST1  = 0xd1,
   M_RST2  = 0xd2,
@@ -49,7 +51,7 @@
   M_RST5  = 0xd5,
   M_RST6  = 0xd6,
   M_RST7  = 0xd7,
-  
+
   M_SOI   = 0xd8,
   M_EOI   = 0xd9,
   M_SOS   = 0xda,
@@ -58,7 +60,7 @@
   M_DRI   = 0xdd,
   M_DHP   = 0xde,
   M_EXP   = 0xdf,
-  
+
   M_APP0  = 0xe0,
   M_APP1  = 0xe1,
   M_APP2  = 0xe2,
@@ -75,13 +77,13 @@
   M_APP13 = 0xed,
   M_APP14 = 0xee,
   M_APP15 = 0xef,
-  
+
   M_JPG0  = 0xf0,
   M_JPG13 = 0xfd,
   M_COM   = 0xfe,
-  
+
   M_TEM   = 0x01,
-  
+
   M_ERROR = 0x100
 } JPEG_MARKER;
 
@@ -100,12 +102,12 @@
   unsigned int length_limit_APPn[16];
 
   /* Status of COM/APPn marker saving */
-  jpeg_saved_marker_ptr cur_marker;	/* NULL if not processing a marker */
-  unsigned int bytes_read;		/* data bytes read so far in marker */
+  jpeg_saved_marker_ptr cur_marker;     /* NULL if not processing a marker */
+  unsigned int bytes_read;              /* data bytes read so far in marker */
   /* Note: cur_marker is not linked into marker_list until it's all read. */
 } my_marker_reader;
 
-typedef my_marker_reader * my_marker_ptr;
+typedef my_marker_reader *my_marker_ptr;
 
 
 /*
@@ -118,49 +120,49 @@
 
 /* Declare and initialize local copies of input pointer/count */
 #define INPUT_VARS(cinfo)  \
-	struct jpeg_source_mgr * datasrc = (cinfo)->src;  \
-	const JOCTET * next_input_byte = datasrc->next_input_byte;  \
-	size_t bytes_in_buffer = datasrc->bytes_in_buffer
+        struct jpeg_source_mgr *datasrc = (cinfo)->src;  \
+        const JOCTET *next_input_byte = datasrc->next_input_byte;  \
+        size_t bytes_in_buffer = datasrc->bytes_in_buffer
 
 /* Unload the local copies --- do this only at a restart boundary */
 #define INPUT_SYNC(cinfo)  \
-	( datasrc->next_input_byte = next_input_byte,  \
-	  datasrc->bytes_in_buffer = bytes_in_buffer )
+        ( datasrc->next_input_byte = next_input_byte,  \
+          datasrc->bytes_in_buffer = bytes_in_buffer )
 
 /* Reload the local copies --- used only in MAKE_BYTE_AVAIL */
 #define INPUT_RELOAD(cinfo)  \
-	( next_input_byte = datasrc->next_input_byte,  \
-	  bytes_in_buffer = datasrc->bytes_in_buffer )
+        ( next_input_byte = datasrc->next_input_byte,  \
+          bytes_in_buffer = datasrc->bytes_in_buffer )
 
 /* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available.
  * Note we do *not* do INPUT_SYNC before calling fill_input_buffer,
  * but we must reload the local copies after a successful fill.
  */
 #define MAKE_BYTE_AVAIL(cinfo,action)  \
-	if (bytes_in_buffer == 0) {  \
-	  if (! (*datasrc->fill_input_buffer) (cinfo))  \
-	    { action; }  \
-	  INPUT_RELOAD(cinfo);  \
-	}
+        if (bytes_in_buffer == 0) {  \
+          if (! (*datasrc->fill_input_buffer) (cinfo))  \
+            { action; }  \
+          INPUT_RELOAD(cinfo);  \
+        }
 
 /* Read a byte into variable V.
  * If must suspend, take the specified action (typically "return FALSE").
  */
 #define INPUT_BYTE(cinfo,V,action)  \
-	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V = GETJOCTET(*next_input_byte++); )
+        MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V = GETJOCTET(*next_input_byte++); )
 
 /* As above, but read two bytes interpreted as an unsigned 16-bit integer.
- * V should be declared unsigned int or perhaps INT32.
+ * V should be declared unsigned int or perhaps JLONG.
  */
 #define INPUT_2BYTES(cinfo,V,action)  \
-	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
-		  MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V += GETJOCTET(*next_input_byte++); )
+        MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
+                  MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V += GETJOCTET(*next_input_byte++); )
 
 
 /*
@@ -199,7 +201,7 @@
 /* Process an SOI marker */
 {
   int i;
-  
+
   TRACEMS(cinfo, 1, JTRC_SOI);
 
   if (cinfo->marker->saw_SOI)
@@ -238,9 +240,9 @@
 get_sof (j_decompress_ptr cinfo, boolean is_prog, boolean is_arith)
 /* Process a SOFn marker */
 {
-  INT32 length;
+  JLONG length;
   int c, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   INPUT_VARS(cinfo);
 
   cinfo->progressive_mode = is_prog;
@@ -256,8 +258,8 @@
   length -= 8;
 
   TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker,
-	   (int) cinfo->image_width, (int) cinfo->image_height,
-	   cinfo->num_components);
+           (int) cinfo->image_width, (int) cinfo->image_height,
+           cinfo->num_components);
 
   if (cinfo->marker->saw_SOF)
     ERREXIT(cinfo, JERR_SOF_DUPLICATE);
@@ -272,11 +274,11 @@
   if (length != (cinfo->num_components * 3))
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
-  if (cinfo->comp_info == NULL)	/* do only once, even if suspend */
+  if (cinfo->comp_info == NULL) /* do only once, even if suspend */
     cinfo->comp_info = (jpeg_component_info *) (*cinfo->mem->alloc_small)
-			((j_common_ptr) cinfo, JPOOL_IMAGE,
-			 cinfo->num_components * SIZEOF(jpeg_component_info));
-  
+                        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                         cinfo->num_components * sizeof(jpeg_component_info));
+
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     compptr->component_index = ci;
@@ -287,8 +289,8 @@
     INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE);
 
     TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT,
-	     compptr->component_id, compptr->h_samp_factor,
-	     compptr->v_samp_factor, compptr->quant_tbl_no);
+             compptr->component_id, compptr->h_samp_factor,
+             compptr->v_samp_factor, compptr->quant_tbl_no);
   }
 
   cinfo->marker->saw_SOF = TRUE;
@@ -302,9 +304,9 @@
 get_sos (j_decompress_ptr cinfo)
 /* Process a SOS marker */
 {
-  INT32 length;
-  int i, ci, n, c, cc;
-  jpeg_component_info * compptr;
+  JLONG length;
+  int i, ci, n, c, cc, pi;
+  jpeg_component_info *compptr;
   INPUT_VARS(cinfo);
 
   if (! cinfo->marker->saw_SOF)
@@ -323,17 +325,18 @@
 
   /* Collect the component-spec parameters */
 
-  for (i = 0; i < cinfo->num_components; i++)
+  for (i = 0; i < MAX_COMPS_IN_SCAN; i++)
     cinfo->cur_comp_info[i] = NULL;
 
   for (i = 0; i < n; i++) {
     INPUT_BYTE(cinfo, cc, return FALSE);
     INPUT_BYTE(cinfo, c, return FALSE);
-    
-    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+
+    for (ci = 0, compptr = cinfo->comp_info;
+         ci < cinfo->num_components && ci < MAX_COMPS_IN_SCAN;
+         ci++, compptr++) {
       if (cc == compptr->component_id && !cinfo->cur_comp_info[ci])
-	goto id_found;
+        goto id_found;
     }
 
     ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
@@ -343,9 +346,16 @@
     cinfo->cur_comp_info[i] = compptr;
     compptr->dc_tbl_no = (c >> 4) & 15;
     compptr->ac_tbl_no = (c     ) & 15;
-    
+
     TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc,
-	     compptr->dc_tbl_no, compptr->ac_tbl_no);
+             compptr->dc_tbl_no, compptr->ac_tbl_no);
+
+    /* This CSi (cc) should differ from the previous CSi */
+    for (pi = 0; pi < i; pi++) {
+      if (cinfo->cur_comp_info[pi] == compptr) {
+        ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+      }
+    }
   }
 
   /* Collect the additional scan parameters Ss, Se, Ah/Al. */
@@ -358,7 +368,7 @@
   cinfo->Al = (c     ) & 15;
 
   TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se,
-	   cinfo->Ah, cinfo->Al);
+           cinfo->Ah, cinfo->Al);
 
   /* Prepare to scan data & restart markers */
   cinfo->marker->next_restart_num = 0;
@@ -377,13 +387,13 @@
 get_dac (j_decompress_ptr cinfo)
 /* Process a DAC marker */
 {
-  INT32 length;
+  JLONG length;
   int index, val;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   while (length > 0) {
     INPUT_BYTE(cinfo, index, return FALSE);
     INPUT_BYTE(cinfo, val, return FALSE);
@@ -397,11 +407,11 @@
 
     if (index >= NUM_ARITH_TBLS) { /* define AC table */
       cinfo->arith_ac_K[index-NUM_ARITH_TBLS] = (UINT8) val;
-    } else {			/* define DC table */
+    } else {                    /* define DC table */
       cinfo->arith_dc_L[index] = (UINT8) (val & 0x0F);
       cinfo->arith_dc_U[index] = (UINT8) (val >> 4);
       if (cinfo->arith_dc_L[index] > cinfo->arith_dc_U[index])
-	ERREXIT1(cinfo, JERR_DAC_VALUE, val);
+        ERREXIT1(cinfo, JERR_DAC_VALUE, val);
     }
   }
 
@@ -423,7 +433,7 @@
 get_dht (j_decompress_ptr cinfo)
 /* Process a DHT marker */
 {
-  INT32 length;
+  JLONG length;
   UINT8 bits[17];
   UINT8 huffval[256];
   int i, index, count;
@@ -432,12 +442,12 @@
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   while (length > 16) {
     INPUT_BYTE(cinfo, index, return FALSE);
 
     TRACEMS1(cinfo, 1, JTRC_DHT, index);
-      
+
     bits[0] = 0;
     count = 0;
     for (i = 1; i <= 16; i++) {
@@ -448,38 +458,41 @@
     length -= 1 + 16;
 
     TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
-	     bits[1], bits[2], bits[3], bits[4],
-	     bits[5], bits[6], bits[7], bits[8]);
+             bits[1], bits[2], bits[3], bits[4],
+             bits[5], bits[6], bits[7], bits[8]);
     TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
-	     bits[9], bits[10], bits[11], bits[12],
-	     bits[13], bits[14], bits[15], bits[16]);
+             bits[9], bits[10], bits[11], bits[12],
+             bits[13], bits[14], bits[15], bits[16]);
 
     /* Here we just do minimal validation of the counts to avoid walking
      * off the end of our table space.  jdhuff.c will check more carefully.
      */
-    if (count > 256 || ((INT32) count) > length)
+    if (count > 256 || ((JLONG) count) > length)
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
 
     for (i = 0; i < count; i++)
       INPUT_BYTE(cinfo, huffval[i], return FALSE);
 
+    MEMZERO(&huffval[count], (256 - count) * sizeof(UINT8));
+
     length -= count;
 
-    if (index & 0x10) {		/* AC table definition */
+    if (index & 0x10) {         /* AC table definition */
       index -= 0x10;
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
       htblptr = &cinfo->ac_huff_tbl_ptrs[index];
-    } else {			/* DC table definition */
+    } else {                    /* DC table definition */
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
       htblptr = &cinfo->dc_huff_tbl_ptrs[index];
     }
 
-    if (index < 0 || index >= NUM_HUFF_TBLS)
-      ERREXIT1(cinfo, JERR_DHT_INDEX, index);
-
     if (*htblptr == NULL)
       *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
-  
-    MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
-    MEMCOPY((*htblptr)->huffval, huffval, SIZEOF((*htblptr)->huffval));
+
+    MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+    MEMCOPY((*htblptr)->huffval, huffval, sizeof((*htblptr)->huffval));
   }
 
   if (length != 0)
@@ -494,7 +507,7 @@
 get_dqt (j_decompress_ptr cinfo)
 /* Process a DQT marker */
 {
-  INT32 length;
+  JLONG length;
   int n, i, prec;
   unsigned int tmp;
   JQUANT_TBL *quant_ptr;
@@ -512,27 +525,27 @@
 
     if (n >= NUM_QUANT_TBLS)
       ERREXIT1(cinfo, JERR_DQT_INDEX, n);
-      
+
     if (cinfo->quant_tbl_ptrs[n] == NULL)
       cinfo->quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) cinfo);
     quant_ptr = cinfo->quant_tbl_ptrs[n];
 
     for (i = 0; i < DCTSIZE2; i++) {
       if (prec)
-	INPUT_2BYTES(cinfo, tmp, return FALSE);
+        INPUT_2BYTES(cinfo, tmp, return FALSE);
       else
-	INPUT_BYTE(cinfo, tmp, return FALSE);
+        INPUT_BYTE(cinfo, tmp, return FALSE);
       /* We convert the zigzag-order table to natural array order. */
       quant_ptr->quantval[jpeg_natural_order[i]] = (UINT16) tmp;
     }
 
     if (cinfo->err->trace_level >= 2) {
       for (i = 0; i < DCTSIZE2; i += 8) {
-	TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
-		 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
-		 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
-		 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
-		 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
+        TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
+                 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
+                 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
+                 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
+                 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
       }
     }
 
@@ -552,12 +565,12 @@
 get_dri (j_decompress_ptr cinfo)
 /* Process a DRI marker */
 {
-  INT32 length;
+  JLONG length;
   unsigned int tmp;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
-  
+
   if (length != 4)
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
@@ -579,20 +592,20 @@
  * JFIF and Adobe markers, respectively.
  */
 
-#define APP0_DATA_LEN	14	/* Length of interesting data in APP0 */
-#define APP14_DATA_LEN	12	/* Length of interesting data in APP14 */
-#define APPN_DATA_LEN	14	/* Must be the largest of the above!! */
+#define APP0_DATA_LEN   14      /* Length of interesting data in APP0 */
+#define APP14_DATA_LEN  12      /* Length of interesting data in APP14 */
+#define APPN_DATA_LEN   14      /* Must be the largest of the above!! */
 
 
 LOCAL(void)
-examine_app0 (j_decompress_ptr cinfo, JOCTET FAR * data,
-	      unsigned int datalen, INT32 remaining)
+examine_app0 (j_decompress_ptr cinfo, JOCTET *data,
+              unsigned int datalen, JLONG remaining)
 /* Examine first few bytes from an APP0.
  * Take appropriate action if it is a JFIF marker.
  * datalen is # of bytes at data[], remaining is length of rest of marker data.
  */
 {
-  INT32 totallen = (INT32) datalen + remaining;
+  JLONG totallen = (JLONG) datalen + remaining;
 
   if (datalen >= APP0_DATA_LEN &&
       GETJOCTET(data[0]) == 0x4A &&
@@ -615,18 +628,18 @@
      */
     if (cinfo->JFIF_major_version != 1)
       WARNMS2(cinfo, JWRN_JFIF_MAJOR,
-	      cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
+              cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
     /* Generate trace messages */
     TRACEMS5(cinfo, 1, JTRC_JFIF,
-	     cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
-	     cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
+             cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
+             cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
     /* Validate thumbnail dimensions and issue appropriate messages */
     if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
       TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
-	       GETJOCTET(data[12]), GETJOCTET(data[13]));
+               GETJOCTET(data[12]), GETJOCTET(data[13]));
     totallen -= APP0_DATA_LEN;
     if (totallen !=
-	((INT32)GETJOCTET(data[12]) * (INT32)GETJOCTET(data[13]) * (INT32) 3))
+        ((JLONG)GETJOCTET(data[12]) * (JLONG)GETJOCTET(data[13]) * (JLONG) 3))
       TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int) totallen);
   } else if (datalen >= 6 &&
       GETJOCTET(data[0]) == 0x4A &&
@@ -650,7 +663,7 @@
       break;
     default:
       TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
-	       GETJOCTET(data[5]), (int) totallen);
+               GETJOCTET(data[5]), (int) totallen);
       break;
     }
   } else {
@@ -661,8 +674,8 @@
 
 
 LOCAL(void)
-examine_app14 (j_decompress_ptr cinfo, JOCTET FAR * data,
-	       unsigned int datalen, INT32 remaining)
+examine_app14 (j_decompress_ptr cinfo, JOCTET *data,
+               unsigned int datalen, JLONG remaining)
 /* Examine first few bytes from an APP14.
  * Take appropriate action if it is an Adobe marker.
  * datalen is # of bytes at data[], remaining is length of rest of marker data.
@@ -695,7 +708,7 @@
 get_interesting_appn (j_decompress_ptr cinfo)
 /* Process an APP0 or APP14 marker without saving it */
 {
-  INT32 length;
+  JLONG length;
   JOCTET b[APPN_DATA_LEN];
   unsigned int i, numtoread;
   INPUT_VARS(cinfo);
@@ -717,10 +730,10 @@
   /* process it */
   switch (cinfo->unread_marker) {
   case M_APP0:
-    examine_app0(cinfo, (JOCTET FAR *) b, numtoread, length);
+    examine_app0(cinfo, (JOCTET *) b, numtoread, length);
     break;
   case M_APP14:
-    examine_app14(cinfo, (JOCTET FAR *) b, numtoread, length);
+    examine_app14(cinfo, (JOCTET *) b, numtoread, length);
     break;
   default:
     /* can't get here unless jpeg_save_markers chooses wrong processor */
@@ -746,33 +759,33 @@
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
   jpeg_saved_marker_ptr cur_marker = marker->cur_marker;
   unsigned int bytes_read, data_length;
-  JOCTET FAR * data;
-  INT32 length = 0;
+  JOCTET *data;
+  JLONG length = 0;
   INPUT_VARS(cinfo);
 
   if (cur_marker == NULL) {
     /* begin reading a marker */
     INPUT_2BYTES(cinfo, length, return FALSE);
     length -= 2;
-    if (length >= 0) {		/* watch out for bogus length word */
+    if (length >= 0) {          /* watch out for bogus length word */
       /* figure out how much we want to save */
       unsigned int limit;
       if (cinfo->unread_marker == (int) M_COM)
-	limit = marker->length_limit_COM;
+        limit = marker->length_limit_COM;
       else
-	limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
+        limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
       if ((unsigned int) length < limit)
-	limit = (unsigned int) length;
+        limit = (unsigned int) length;
       /* allocate and initialize the marker item */
       cur_marker = (jpeg_saved_marker_ptr)
-	(*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				    SIZEOF(struct jpeg_marker_struct) + limit);
+        (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                    sizeof(struct jpeg_marker_struct) + limit);
       cur_marker->next = NULL;
       cur_marker->marker = (UINT8) cinfo->unread_marker;
       cur_marker->original_length = (unsigned int) length;
       cur_marker->data_length = limit;
       /* data area is just beyond the jpeg_marker_struct */
-      data = cur_marker->data = (JOCTET FAR *) (cur_marker + 1);
+      data = cur_marker->data = (JOCTET *) (cur_marker + 1);
       marker->cur_marker = cur_marker;
       marker->bytes_read = 0;
       bytes_read = 0;
@@ -790,7 +803,7 @@
   }
 
   while (bytes_read < data_length) {
-    INPUT_SYNC(cinfo);		/* move the restart point to here */
+    INPUT_SYNC(cinfo);          /* move the restart point to here */
     marker->bytes_read = bytes_read;
     /* If there's not at least one byte in buffer, suspend */
     MAKE_BYTE_AVAIL(cinfo, return FALSE);
@@ -803,14 +816,14 @@
   }
 
   /* Done reading what we want to read */
-  if (cur_marker != NULL) {	/* will be NULL if bogus length word */
+  if (cur_marker != NULL) {     /* will be NULL if bogus length word */
     /* Add new marker to end of list */
     if (cinfo->marker_list == NULL) {
       cinfo->marker_list = cur_marker;
     } else {
       jpeg_saved_marker_ptr prev = cinfo->marker_list;
       while (prev->next != NULL)
-	prev = prev->next;
+        prev = prev->next;
       prev->next = cur_marker;
     }
     /* Reset pointer & calc remaining data length */
@@ -830,12 +843,12 @@
     break;
   default:
     TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker,
-	     (int) (data_length + length));
+             (int) (data_length + length));
     break;
   }
 
   /* skip any remaining data -- could be lots */
-  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
   if (length > 0)
     (*cinfo->src->skip_input_data) (cinfo, (long) length);
 
@@ -849,15 +862,15 @@
 skip_variable (j_decompress_ptr cinfo)
 /* Skip over an unknown or uninteresting variable-length marker */
 {
-  INT32 length;
+  JLONG length;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int) length);
 
-  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
   if (length > 0)
     (*cinfo->src->skip_input_data) (cinfo, (long) length);
 
@@ -901,7 +914,7 @@
       INPUT_BYTE(cinfo, c, return FALSE);
     } while (c == 0xFF);
     if (c != 0)
-      break;			/* found a valid marker, exit loop */
+      break;                    /* found a valid marker, exit loop */
     /* Reach here if we found a stuffed-zero data sequence (FF/00).
      * Discard it and loop back to try again.
      */
@@ -961,11 +974,11 @@
     /* NB: first_marker() enforces the requirement that SOI appear first. */
     if (cinfo->unread_marker == 0) {
       if (! cinfo->marker->saw_SOI) {
-	if (! first_marker(cinfo))
-	  return JPEG_SUSPENDED;
+        if (! first_marker(cinfo))
+          return JPEG_SUSPENDED;
       } else {
-	if (! next_marker(cinfo))
-	  return JPEG_SUSPENDED;
+        if (! next_marker(cinfo))
+          return JPEG_SUSPENDED;
       }
     }
     /* At this point cinfo->unread_marker contains the marker code and the
@@ -975,74 +988,74 @@
     switch (cinfo->unread_marker) {
     case M_SOI:
       if (! get_soi(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF0:		/* Baseline */
-    case M_SOF1:		/* Extended sequential, Huffman */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
       if (! get_sof(cinfo, FALSE, FALSE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF2:		/* Progressive, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
       if (! get_sof(cinfo, TRUE, FALSE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF9:		/* Extended sequential, arithmetic */
+    case M_SOF9:                /* Extended sequential, arithmetic */
       if (! get_sof(cinfo, FALSE, TRUE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF10:		/* Progressive, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
       if (! get_sof(cinfo, TRUE, TRUE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
     /* Currently unsupported SOFn types */
-    case M_SOF3:		/* Lossless, Huffman */
-    case M_SOF5:		/* Differential sequential, Huffman */
-    case M_SOF6:		/* Differential progressive, Huffman */
-    case M_SOF7:		/* Differential lossless, Huffman */
-    case M_JPG:			/* Reserved for JPEG extensions */
-    case M_SOF11:		/* Lossless, arithmetic */
-    case M_SOF13:		/* Differential sequential, arithmetic */
-    case M_SOF14:		/* Differential progressive, arithmetic */
-    case M_SOF15:		/* Differential lossless, arithmetic */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_JPG:                 /* Reserved for JPEG extensions */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
       ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker);
       break;
 
     case M_SOS:
       if (! get_sos(cinfo))
-	return JPEG_SUSPENDED;
-      cinfo->unread_marker = 0;	/* processed the marker */
+        return JPEG_SUSPENDED;
+      cinfo->unread_marker = 0; /* processed the marker */
       return JPEG_REACHED_SOS;
-    
+
     case M_EOI:
       TRACEMS(cinfo, 1, JTRC_EOI);
-      cinfo->unread_marker = 0;	/* processed the marker */
+      cinfo->unread_marker = 0; /* processed the marker */
       return JPEG_REACHED_EOI;
-      
+
     case M_DAC:
       if (! get_dac(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DHT:
       if (! get_dht(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DQT:
       if (! get_dqt(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DRI:
       if (! get_dri(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_APP0:
     case M_APP1:
     case M_APP2:
@@ -1060,16 +1073,16 @@
     case M_APP14:
     case M_APP15:
       if (! (*((my_marker_ptr) cinfo->marker)->process_APPn[
-		cinfo->unread_marker - (int) M_APP0]) (cinfo))
-	return JPEG_SUSPENDED;
+                cinfo->unread_marker - (int) M_APP0]) (cinfo))
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_COM:
       if (! (*((my_marker_ptr) cinfo->marker)->process_COM) (cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_RST0:		/* these are all parameterless */
+    case M_RST0:                /* these are all parameterless */
     case M_RST1:
     case M_RST2:
     case M_RST3:
@@ -1081,12 +1094,12 @@
       TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker);
       break;
 
-    case M_DNL:			/* Ignore DNL ... perhaps the wrong thing */
+    case M_DNL:                 /* Ignore DNL ... perhaps the wrong thing */
       if (! skip_variable(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    default:			/* must be DHP, EXP, JPGn, or RESn */
+    default:                    /* must be DHP, EXP, JPGn, or RESn */
       /* For now, we treat the reserved markers as fatal errors since they are
        * likely to be used to signal incompatible JPEG Part 3 extensions.
        * Once the JPEG 3 version-number marker is well defined, this code
@@ -1132,7 +1145,7 @@
     /* Uh-oh, the restart markers have been messed up. */
     /* Let the data source manager determine how to resync. */
     if (! (*cinfo->src->resync_to_restart) (cinfo,
-					    cinfo->marker->next_restart_num))
+                                            cinfo->marker->next_restart_num))
       return FALSE;
   }
 
@@ -1197,25 +1210,25 @@
 {
   int marker = cinfo->unread_marker;
   int action = 1;
-  
+
   /* Always put up a warning. */
   WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired);
-  
+
   /* Outer loop handles repeated decision after scanning forward. */
   for (;;) {
     if (marker < (int) M_SOF0)
-      action = 2;		/* invalid marker */
+      action = 2;               /* invalid marker */
     else if (marker < (int) M_RST0 || marker > (int) M_RST7)
-      action = 3;		/* valid non-restart marker */
+      action = 3;               /* valid non-restart marker */
     else {
       if (marker == ((int) M_RST0 + ((desired+1) & 7)) ||
-	  marker == ((int) M_RST0 + ((desired+2) & 7)))
-	action = 3;		/* one of the next two expected restarts */
+          marker == ((int) M_RST0 + ((desired+2) & 7)))
+        action = 3;             /* one of the next two expected restarts */
       else if (marker == ((int) M_RST0 + ((desired-1) & 7)) ||
-	       marker == ((int) M_RST0 + ((desired-2) & 7)))
-	action = 2;		/* a prior restart, so advance */
+               marker == ((int) M_RST0 + ((desired-2) & 7)))
+        action = 2;             /* a prior restart, so advance */
       else
-	action = 1;		/* desired restart or too far away */
+        action = 1;             /* desired restart or too far away */
     }
     TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action);
     switch (action) {
@@ -1226,7 +1239,7 @@
     case 2:
       /* Scan to the next marker, and repeat the decision loop. */
       if (! next_marker(cinfo))
-	return FALSE;
+        return FALSE;
       marker = cinfo->unread_marker;
       break;
     case 3:
@@ -1247,10 +1260,10 @@
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
-  cinfo->comp_info = NULL;		/* until allocated by get_sof */
-  cinfo->input_scan_number = 0;		/* no SOS seen yet */
-  cinfo->unread_marker = 0;		/* no pending marker */
-  marker->pub.saw_SOI = FALSE;		/* set internal state too */
+  cinfo->comp_info = NULL;              /* until allocated by get_sof */
+  cinfo->input_scan_number = 0;         /* no SOS seen yet */
+  cinfo->unread_marker = 0;             /* no pending marker */
+  marker->pub.saw_SOI = FALSE;          /* set internal state too */
   marker->pub.saw_SOF = FALSE;
   marker->pub.discarded_bytes = 0;
   marker->cur_marker = NULL;
@@ -1271,7 +1284,7 @@
   /* Create subobject in permanent pool */
   marker = (my_marker_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				SIZEOF(my_marker_reader));
+                                sizeof(my_marker_reader));
   cinfo->marker = (struct jpeg_marker_reader *) marker;
   /* Initialize public method pointers */
   marker->pub.reset_marker_reader = reset_marker_reader;
@@ -1302,7 +1315,7 @@
 
 GLOBAL(void)
 jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
-		   unsigned int length_limit)
+                   unsigned int length_limit)
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
   long maxlength;
@@ -1311,7 +1324,7 @@
   /* Length limit mustn't be larger than what we can allocate
    * (should only be a concern in a 16-bit environment).
    */
-  maxlength = cinfo->mem->max_alloc_chunk - SIZEOF(struct jpeg_marker_struct);
+  maxlength = cinfo->mem->max_alloc_chunk - sizeof(struct jpeg_marker_struct);
   if (((long) length_limit) > maxlength)
     length_limit = (unsigned int) maxlength;
 
@@ -1351,7 +1364,7 @@
 
 GLOBAL(void)
 jpeg_set_marker_processor (j_decompress_ptr cinfo, int marker_code,
-			   jpeg_marker_parser_method routine)
+                           jpeg_marker_parser_method routine)
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmaster.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmaster.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmaster.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmaster.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,15 @@
 /*
  * jdmaster.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains master control logic for the JPEG decompressor.
  * These routines are concerned with selecting the modules to be executed
@@ -16,25 +21,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jpegcomp.h"
-
-
-/* Private state */
-
-typedef struct {
-  struct jpeg_decomp_master pub; /* public fields */
-
-  int pass_number;		/* # of passes completed */
-
-  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
-
-  /* Saved references to initialized quantizer modules,
-   * in case we need to switch modes.
-   */
-  struct jpeg_color_quantizer * quantizer_1pass;
-  struct jpeg_color_quantizer * quantizer_2pass;
-} my_decomp_master;
-
-typedef my_decomp_master * my_master_ptr;
+#include "jdmaster.h"
 
 
 /*
@@ -49,9 +36,10 @@
   /* Merging is the equivalent of plain box-filter upsampling */
   if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling)
     return FALSE;
-  /* jdmerge.c only supports YCC=>RGB color conversion */
+  /* jdmerge.c only supports YCC=>RGB and YCC=>RGB565 color conversion */
   if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 ||
       (cinfo->out_color_space != JCS_RGB &&
+      cinfo->out_color_space != JCS_RGB565 &&
       cinfo->out_color_space != JCS_EXT_RGB &&
       cinfo->out_color_space != JCS_EXT_RGBX &&
       cinfo->out_color_space != JCS_EXT_BGR &&
@@ -61,8 +49,12 @@
       cinfo->out_color_space != JCS_EXT_RGBA &&
       cinfo->out_color_space != JCS_EXT_BGRA &&
       cinfo->out_color_space != JCS_EXT_ABGR &&
-      cinfo->out_color_space != JCS_EXT_ARGB) ||
-      cinfo->out_color_components != rgb_pixelsize[cinfo->out_color_space])
+      cinfo->out_color_space != JCS_EXT_ARGB))
+    return FALSE;
+  if ((cinfo->out_color_space == JCS_RGB565 &&
+      cinfo->out_color_components != 3) ||
+      (cinfo->out_color_space != JCS_RGB565 &&
+      cinfo->out_color_components != rgb_pixelsize[cinfo->out_color_space]))
     return FALSE;
   /* and it only handles 2h1v or 2h2v sampling ratios */
   if (cinfo->comp_info[0].h_samp_factor != 2 ||
@@ -78,7 +70,7 @@
       cinfo->comp_info[2]._DCT_scaled_size != cinfo->_min_DCT_scaled_size)
     return FALSE;
   /* ??? also need to test for upsample-time rescaling, when & if supported */
-  return TRUE;			/* by golly, it'll work... */
+  return TRUE;                  /* by golly, it'll work... */
 #else
   return FALSE;
 #endif
@@ -89,6 +81,177 @@
  * Compute output image dimensions and related values.
  * NOTE: this is exported for possible use by application.
  * Hence it mustn't do anything that can't be done twice.
+ */
+
+#if JPEG_LIB_VERSION >= 80
+GLOBAL(void)
+#else
+LOCAL(void)
+#endif
+jpeg_core_output_dimensions (j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase.
+ * This function is used for transcoding and full decompression.
+ */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Compute actual output image dimensions and DCT scaling choices. */
+  if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom) {
+    /* Provide 1/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 1;
+    cinfo->_min_DCT_v_scaled_size = 1;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 2) {
+    /* Provide 2/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 2L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 2L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 2;
+    cinfo->_min_DCT_v_scaled_size = 2;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 3) {
+    /* Provide 3/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 3L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 3L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 3;
+    cinfo->_min_DCT_v_scaled_size = 3;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 4) {
+    /* Provide 4/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 4L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 4L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 4;
+    cinfo->_min_DCT_v_scaled_size = 4;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 5) {
+    /* Provide 5/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 5L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 5L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 5;
+    cinfo->_min_DCT_v_scaled_size = 5;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 6) {
+    /* Provide 6/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 6L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 6L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 6;
+    cinfo->_min_DCT_v_scaled_size = 6;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 7) {
+    /* Provide 7/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 7L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 7L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 7;
+    cinfo->_min_DCT_v_scaled_size = 7;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 8) {
+    /* Provide 8/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 8L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 8L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 8;
+    cinfo->_min_DCT_v_scaled_size = 8;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 9) {
+    /* Provide 9/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 9L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 9L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 9;
+    cinfo->_min_DCT_v_scaled_size = 9;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 10) {
+    /* Provide 10/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 10L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 10L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 10;
+    cinfo->_min_DCT_v_scaled_size = 10;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 11) {
+    /* Provide 11/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 11L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 11L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 11;
+    cinfo->_min_DCT_v_scaled_size = 11;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 12) {
+    /* Provide 12/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 12L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 12L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 12;
+    cinfo->_min_DCT_v_scaled_size = 12;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 13) {
+    /* Provide 13/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 13L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 13L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 13;
+    cinfo->_min_DCT_v_scaled_size = 13;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 14) {
+    /* Provide 14/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 14L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 14L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 14;
+    cinfo->_min_DCT_v_scaled_size = 14;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 15) {
+    /* Provide 15/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 15L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 15L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 15;
+    cinfo->_min_DCT_v_scaled_size = 15;
+  } else {
+    /* Provide 16/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 16L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 16L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 16;
+    cinfo->_min_DCT_v_scaled_size = 16;
+  }
+
+  /* Recompute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->_DCT_h_scaled_size = cinfo->_min_DCT_h_scaled_size;
+    compptr->_DCT_v_scaled_size = cinfo->_min_DCT_v_scaled_size;
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
  * Also note that it may be called before the master module is initialized!
  */
 
@@ -105,65 +268,24 @@
   if (cinfo->global_state != DSTATE_READY)
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
+  /* Compute core output image dimensions and DCT scaling choices. */
+  jpeg_core_output_dimensions(cinfo);
+
 #ifdef IDCT_SCALING_SUPPORTED
 
-  /* Compute actual output image dimensions and DCT scaling choices. */
-  if (cinfo->scale_num * 8 <= cinfo->scale_denom) {
-    /* Provide 1/8 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 8L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 8L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 1;
-#else
-    cinfo->min_DCT_scaled_size = 1;
-#endif
-  } else if (cinfo->scale_num * 4 <= cinfo->scale_denom) {
-    /* Provide 1/4 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 4L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 4L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 2;
-#else
-    cinfo->min_DCT_scaled_size = 2;
-#endif
-  } else if (cinfo->scale_num * 2 <= cinfo->scale_denom) {
-    /* Provide 1/2 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 2L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 2L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 4;
-#else
-    cinfo->min_DCT_scaled_size = 4;
-#endif
-  } else {
-    /* Provide 1/1 scaling */
-    cinfo->output_width = cinfo->image_width;
-    cinfo->output_height = cinfo->image_height;
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = DCTSIZE;
-#else
-    cinfo->min_DCT_scaled_size = DCTSIZE;
-#endif
-  }
   /* In selecting the actual DCT scaling for each component, we try to
    * scale up the chroma components via IDCT scaling rather than upsampling.
    * This saves time if the upsampler gets to use 1:1 scaling.
-   * Note this code assumes that the supported DCT scalings are powers of 2.
+   * Note this code adapts subsampling ratios which are powers of 2.
    */
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     int ssize = cinfo->_min_DCT_scaled_size;
     while (ssize < DCTSIZE &&
-	   (compptr->h_samp_factor * ssize * 2 <=
-	    cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) &&
-	   (compptr->v_samp_factor * ssize * 2 <=
-	    cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size)) {
+           ((cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->h_samp_factor * ssize * 2) == 0) &&
+           ((cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->v_samp_factor * ssize * 2) == 0)) {
       ssize = ssize * 2;
     }
 #if JPEG_LIB_VERSION >= 70
@@ -181,12 +303,12 @@
     /* Size in samples, after IDCT scaling */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width *
-		    (long) (compptr->h_samp_factor * compptr->_DCT_scaled_size),
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (compptr->h_samp_factor * compptr->_DCT_scaled_size),
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height *
-		    (long) (compptr->v_samp_factor * compptr->_DCT_scaled_size),
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (compptr->v_samp_factor * compptr->_DCT_scaled_size),
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
   }
 
 #else /* !IDCT_SCALING_SUPPORTED */
@@ -220,18 +342,19 @@
     cinfo->out_color_components = rgb_pixelsize[cinfo->out_color_space];
     break;
   case JCS_YCbCr:
+  case JCS_RGB565:
     cinfo->out_color_components = 3;
     break;
   case JCS_CMYK:
   case JCS_YCCK:
     cinfo->out_color_components = 4;
     break;
-  default:			/* else must be same colorspace as in file */
+  default:                      /* else must be same colorspace as in file */
     cinfo->out_color_components = cinfo->num_components;
     break;
   }
   cinfo->output_components = (cinfo->quantize_colors ? 1 :
-			      cinfo->out_color_components);
+                              cinfo->out_color_components);
 
   /* See if upsampler will want to emit more than one row at a time */
   if (use_merged_upsample(cinfo))
@@ -248,20 +371,20 @@
  * processes are inner loops and need to be as fast as possible.  On most
  * machines, particularly CPUs with pipelines or instruction prefetch,
  * a (subscript-check-less) C table lookup
- *		x = sample_range_limit[x];
+ *              x = sample_range_limit[x];
  * is faster than explicit tests
- *		if (x < 0)  x = 0;
- *		else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
+ *              if (x < 0)  x = 0;
+ *              else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
  * These processes all use a common table prepared by the routine below.
  *
  * For most steps we can mathematically guarantee that the initial value
  * of x is within MAXJSAMPLE+1 of the legal range, so a table running from
  * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient.  But for the initial
- * limiting step (just after the IDCT), a wildly out-of-range value is 
+ * limiting step (just after the IDCT), a wildly out-of-range value is
  * possible if the input data is corrupt.  To avoid any chance of indexing
  * off the end of memory and getting a bad-pointer trap, we perform the
  * post-IDCT limiting thus:
- *		x = range_limit[x & MASK];
+ *              x = range_limit[x & MASK];
  * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit
  * samples.  Under normal circumstances this is more than enough range and
  * a correct output will be generated; with bogus input data the mask will
@@ -279,37 +402,34 @@
  * We can save some space by overlapping the start of the post-IDCT table
  * with the simpler range limiting table.  The post-IDCT table begins at
  * sample_range_limit + CENTERJSAMPLE.
- *
- * Note that the table is allocated in near data space on PCs; it's small
- * enough and used often enough to justify this.
  */
 
 LOCAL(void)
 prepare_range_limit_table (j_decompress_ptr cinfo)
 /* Allocate and fill in the sample_range_limit table */
 {
-  JSAMPLE * table;
+  JSAMPLE *table;
   int i;
 
   table = (JSAMPLE *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-		(5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * SIZEOF(JSAMPLE));
-  table += (MAXJSAMPLE+1);	/* allow negative subscripts of simple table */
+                (5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * sizeof(JSAMPLE));
+  table += (MAXJSAMPLE+1);      /* allow negative subscripts of simple table */
   cinfo->sample_range_limit = table;
   /* First segment of "simple" table: limit[x] = 0 for x < 0 */
-  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * SIZEOF(JSAMPLE));
+  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * sizeof(JSAMPLE));
   /* Main part of "simple" table: limit[x] = x */
   for (i = 0; i <= MAXJSAMPLE; i++)
     table[i] = (JSAMPLE) i;
-  table += CENTERJSAMPLE;	/* Point to where post-IDCT table starts */
+  table += CENTERJSAMPLE;       /* Point to where post-IDCT table starts */
   /* End of simple table, rest of first half of post-IDCT table */
   for (i = CENTERJSAMPLE; i < 2*(MAXJSAMPLE+1); i++)
     table[i] = MAXJSAMPLE;
   /* Second half of post-IDCT table */
   MEMZERO(table + (2 * (MAXJSAMPLE+1)),
-	  (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+          (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * sizeof(JSAMPLE));
   MEMCOPY(table + (4 * (MAXJSAMPLE+1) - CENTERJSAMPLE),
-	  cinfo->sample_range_limit, CENTERJSAMPLE * SIZEOF(JSAMPLE));
+          cinfo->sample_range_limit, CENTERJSAMPLE * sizeof(JSAMPLE));
 }
 
 
@@ -442,6 +562,12 @@
   /* Initialize input side of decompressor to consume first scan. */
   (*cinfo->inputctl->start_input_pass) (cinfo);
 
+  /* Set the first and last iMCU columns to decompress from single-scan images.
+   * By default, decompress all of the iMCU columns.
+   */
+  cinfo->master->first_iMCU_col = 0;
+  cinfo->master->last_iMCU_col = cinfo->MCUs_per_row - 1;
+
 #ifdef D_MULTISCAN_FILES_SUPPORTED
   /* If jpeg_start_decompress will read the whole file, initialize
    * progress monitoring appropriately.  The input step is counted
@@ -497,24 +623,24 @@
     if (cinfo->quantize_colors && cinfo->colormap == NULL) {
       /* Select new quantization method */
       if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) {
-	cinfo->cquantize = master->quantizer_2pass;
-	master->pub.is_dummy_pass = TRUE;
+        cinfo->cquantize = master->quantizer_2pass;
+        master->pub.is_dummy_pass = TRUE;
       } else if (cinfo->enable_1pass_quant) {
-	cinfo->cquantize = master->quantizer_1pass;
+        cinfo->cquantize = master->quantizer_1pass;
       } else {
-	ERREXIT(cinfo, JERR_MODE_CHANGE);
+        ERREXIT(cinfo, JERR_MODE_CHANGE);
       }
     }
     (*cinfo->idct->start_pass) (cinfo);
     (*cinfo->coef->start_output_pass) (cinfo);
     if (! cinfo->raw_data_out) {
       if (! master->using_merged_upsample)
-	(*cinfo->cconvert->start_pass) (cinfo);
+        (*cinfo->cconvert->start_pass) (cinfo);
       (*cinfo->upsample->start_pass) (cinfo);
       if (cinfo->quantize_colors)
-	(*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
+        (*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
       (*cinfo->post->start_pass) (cinfo,
-	    (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+            (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
       (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
     }
   }
@@ -523,7 +649,7 @@
   if (cinfo->progress != NULL) {
     cinfo->progress->completed_passes = master->pass_number;
     cinfo->progress->total_passes = master->pass_number +
-				    (master->pub.is_dummy_pass ? 2 : 1);
+                                    (master->pub.is_dummy_pass ? 2 : 1);
     /* In buffered-image mode, we assume one more output pass if EOI not
      * yet reached, but no more passes if EOI has been reached.
      */
@@ -586,16 +712,13 @@
 GLOBAL(void)
 jinit_master_decompress (j_decompress_ptr cinfo)
 {
-  my_master_ptr master;
+  my_master_ptr master = (my_master_ptr) cinfo->master;
 
-  master = (my_master_ptr)
-      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(my_decomp_master));
-  cinfo->master = (struct jpeg_decomp_master *) master;
   master->pub.prepare_for_output_pass = prepare_for_output_pass;
   master->pub.finish_output_pass = finish_output_pass;
 
   master->pub.is_dummy_pass = FALSE;
+  master->pub.jinit_upsampler_no_alloc = FALSE;
 
   master_selection(cinfo);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmaster.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmaster.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmaster.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmaster.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+/*
+ * jdmaster.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1995, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the master control structure for the JPEG decompressor.
+ */
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_decomp_master pub; /* public fields */
+
+  int pass_number;              /* # of passes completed */
+
+  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
+
+  /* Saved references to initialized quantizer modules,
+   * in case we need to switch modes.
+   */
+  struct jpeg_color_quantizer *quantizer_1pass;
+  struct jpeg_color_quantizer *quantizer_2pass;
+} my_decomp_master;
+
+typedef my_decomp_master *my_master_ptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmerge.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmerge.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmerge.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmerge.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jdmerge.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009, 2011, 2014-2015, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains code for merged upsampling/color conversion.
  *
@@ -16,19 +19,19 @@
  * (ie, box filtering), we can save some work in color conversion by
  * calculating all the output pixels corresponding to a pair of chroma
  * samples at one time.  In the conversion equations
- *	R = Y           + K1 * Cr
- *	G = Y + K2 * Cb + K3 * Cr
- *	B = Y + K4 * Cb
+ *      R = Y           + K1 * Cr
+ *      G = Y + K2 * Cb + K3 * Cr
+ *      B = Y + K4 * Cb
  * only the Y term varies among the group of pixels corresponding to a pair
  * of chroma samples, so the rest of the terms can be calculated just once.
  * At typical sampling ratios, this eliminates half or three-quarters of the
  * multiplications needed for color conversion.
  *
  * This file currently provides implementations for the following cases:
- *	YCbCr => RGB color conversion only.
- *	Sampling ratios of 2h1v or 2h2v.
- *	No scaling needed at upsample time.
- *	Corner-aligned (non-CCIR601) sampling alignment.
+ *      YCbCr => RGB color conversion only.
+ *      Sampling ratios of 2h1v or 2h2v.
+ *      No scaling needed at upsample time.
+ *      Corner-aligned (non-CCIR601) sampling alignment.
  * Other special cases could be added, but in most applications these are
  * the only common cases.  (For uncommon cases we fall back on the more
  * general code in jdsample.c and jdcolor.c.)
@@ -38,7 +41,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 #ifdef UPSAMPLE_MERGING_SUPPORTED
 
@@ -46,18 +49,17 @@
 /* Private subobject */
 
 typedef struct {
-  struct jpeg_upsampler pub;	/* public fields */
+  struct jpeg_upsampler pub;    /* public fields */
 
   /* Pointer to routine to do actual upsampling/conversion of one row group */
-  JMETHOD(void, upmethod, (j_decompress_ptr cinfo,
-			   JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-			   JSAMPARRAY output_buf));
+  void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
   /* Private state for YCC->RGB conversion */
-  int * Cr_r_tab;		/* => table for Cr to R conversion */
-  int * Cb_b_tab;		/* => table for Cb to B conversion */
-  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
-  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
 
   /* For 2:1 vertical sampling, we produce two output rows at a time.
    * We need a "spare" row buffer to hold the second output row if the
@@ -65,22 +67,22 @@
    * to discard the dummy last row if the image height is odd.
    */
   JSAMPROW spare_row;
-  boolean spare_full;		/* T if spare buffer is occupied */
+  boolean spare_full;           /* T if spare buffer is occupied */
 
-  JDIMENSION out_row_width;	/* samples per output row */
-  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+  JDIMENSION out_row_width;     /* samples per output row */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
 } my_upsampler;
 
-typedef my_upsampler * my_upsample_ptr;
+typedef my_upsampler *my_upsample_ptr;
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -92,7 +94,7 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define h2v1_merged_upsample_internal extrgb_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extrgb_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -103,13 +105,15 @@
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
 #define RGB_BLUE EXT_RGBX_BLUE
+#define RGB_ALPHA 3
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define h2v1_merged_upsample_internal extrgbx_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extrgbx_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -120,7 +124,7 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define h2v1_merged_upsample_internal extbgr_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extbgr_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -131,13 +135,15 @@
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
 #define RGB_BLUE EXT_BGRX_BLUE
+#define RGB_ALPHA 3
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define h2v1_merged_upsample_internal extbgrx_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extbgrx_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -145,13 +151,15 @@
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
 #define RGB_BLUE EXT_XBGR_BLUE
+#define RGB_ALPHA 0
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define h2v1_merged_upsample_internal extxbgr_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extxbgr_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -159,13 +167,15 @@
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
 #define RGB_BLUE EXT_XRGB_BLUE
+#define RGB_ALPHA 0
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define h2v1_merged_upsample_internal extxrgb_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extxrgb_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -181,31 +191,31 @@
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   int i;
-  INT32 x;
+  JLONG x;
   SHIFT_TEMPS
 
   upsample->Cr_r_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
+                                (MAXJSAMPLE+1) * sizeof(int));
   upsample->Cb_b_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
-  upsample->Cr_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(int));
+  upsample->Cr_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
-  upsample->Cb_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
+  upsample->Cb_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
 
   for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
     /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
     /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
     /* Cr=>R value is nearest int to 1.40200 * x */
     upsample->Cr_r_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
     /* Cb=>B value is nearest int to 1.77200 * x */
     upsample->Cb_b_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
     /* Cr=>G value is scaled-up -0.71414 * x */
     upsample->Cr_g_tab[i] = (- FIX(0.71414)) * x;
     /* Cb=>G value is scaled-up -0.34414 * x */
@@ -239,20 +249,23 @@
 
 METHODDEF(void)
 merged_2v_upsample (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 /* 2:1 vertical sampling case: may need a spare row. */
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   JSAMPROW work_ptrs[2];
-  JDIMENSION num_rows;		/* number of rows returned to caller */
+  JDIMENSION num_rows;          /* number of rows returned to caller */
 
   if (upsample->spare_full) {
     /* If we have a spare row saved from a previous cycle, just return it. */
+    JDIMENSION size = upsample->out_row_width;
+    if (cinfo->out_color_space == JCS_RGB565)
+      size = cinfo->output_width * 2;
     jcopy_sample_rows(& upsample->spare_row, 0, output_buf + *out_row_ctr, 0,
-		      1, upsample->out_row_width);
+                      1, size);
     num_rows = 1;
     upsample->spare_full = FALSE;
   } else {
@@ -288,17 +301,17 @@
 
 METHODDEF(void)
 merged_1v_upsample (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 /* 1:1 vertical sampling case: much easier, never need a spare row. */
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
 
   /* Just do the upsampling. */
   (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
-			 output_buf + *out_row_ctr);
+                         output_buf + *out_row_ctr);
   /* Adjust counts */
   (*out_row_ctr)++;
   (*in_row_group_ctr)++;
@@ -321,8 +334,8 @@
 
 METHODDEF(void)
 h2v1_merged_upsample (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-		      JSAMPARRAY output_buf)
+                      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                      JSAMPARRAY output_buf)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -367,8 +380,8 @@
 
 METHODDEF(void)
 h2v2_merged_upsample (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-		      JSAMPARRAY output_buf)
+                      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                      JSAMPARRAY output_buf)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -408,6 +421,151 @@
 
 
 /*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b)   ((((r) << 8) & 0xF800) |  \
+                                      (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)   (((r) & 0xF8) | ((g) >> 5) |  \
+                                      (((g) << 11) & 0xE000) |  \
+                                      (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)     ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)     ((l << 16) | r)
+
+#define PACK_NEED_ALIGNMENT(ptr)  (((size_t)(ptr)) & 3)
+
+#define WRITE_TWO_PIXELS_LE(addr, pixels) {  \
+  ((INT16*)(addr))[0] = (INT16)(pixels);  \
+  ((INT16*)(addr))[1] = (INT16)((pixels) >> 16);  \
+}
+#define WRITE_TWO_PIXELS_BE(addr, pixels) {  \
+  ((INT16*)(addr))[1] = (INT16)(pixels);  \
+  ((INT16*)(addr))[0] = (INT16)((pixels) >> 16);  \
+}
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficent for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565 PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE
+#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_LE
+#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_le
+#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_le
+#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_le
+#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_le
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+#define PACK_SHORT_565 PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE
+#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_BE
+#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_be
+#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_be
+#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_be
+#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_be
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565 (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                          JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v1_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+ }
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565D (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                           JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v1_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565 (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                          JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v2_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565D (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                           JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v2_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+/*
  * Module initialization routine for merged upsampling/color conversion.
  *
  * NB: this is called under the conditions determined by use_merged_upsample()
@@ -422,7 +580,7 @@
 
   upsample = (my_upsample_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_upsampler));
+                                sizeof(my_upsampler));
   cinfo->upsample = (struct jpeg_upsampler *) upsample;
   upsample->pub.start_pass = start_pass_merged_upsample;
   upsample->pub.need_context_rows = FALSE;
@@ -435,16 +593,30 @@
       upsample->upmethod = jsimd_h2v2_merged_upsample;
     else
       upsample->upmethod = h2v2_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v2_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v2_merged_upsample_565;
+      }
+    }
     /* Allocate a spare row buffer */
     upsample->spare_row = (JSAMPROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-		(size_t) (upsample->out_row_width * SIZEOF(JSAMPLE)));
+                (size_t) (upsample->out_row_width * sizeof(JSAMPLE)));
   } else {
     upsample->pub.upsample = merged_1v_upsample;
     if (jsimd_can_h2v1_merged_upsample())
       upsample->upmethod = jsimd_h2v1_merged_upsample;
     else
       upsample->upmethod = h2v1_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v1_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v1_merged_upsample_565;
+      }
+    }
     /* No spare row needed */
     upsample->spare_row = NULL;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrg565.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrg565.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrg565.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrg565.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,356 @@
+/*
+ * jdmrg565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565_internal (j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr = (INT16)rgb;
+   }
+ }
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565D_internal (j_decompress_ptr cinfo,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565_internal (j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr * 2];
+  inptr01 = input_buf[0][in_row_group_ctr * 2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr0 = (INT16)rgb;
+
+    y  = GETJSAMPLE(*inptr01);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr1 = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565D_internal (j_decompress_ptr cinfo,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  JLONG d1 = dither_matrix[(cinfo->output_scanline+1) & DITHER_MASK];
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr0 = (INT16)rgb;
+
+    y  = GETJSAMPLE(*inptr01);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr1 = (INT16)rgb;
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrgext.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrgext.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrgext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrgext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,186 @@
+/*
+ * jdmrgext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+/* This file is included by jdmerge.c */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    y  = GETJSAMPLE(*inptr01);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrgext.c.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrgext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdmrgext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdmrgext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,156 +0,0 @@
-/*
- * jdmrgext.c
- *
- * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains code for merged upsampling/color conversion.
- */
-
-
-/* This file is included by jdmerge.c */
-
-
-/*
- * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
- */
-
-INLINE
-LOCAL(void)
-h2v1_merged_upsample_internal (j_decompress_ptr cinfo,
-                               JSAMPIMAGE input_buf,
-                               JDIMENSION in_row_group_ctr,
-                               JSAMPARRAY output_buf)
-{
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr;
-  JSAMPROW inptr0, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr0 = input_buf[0][in_row_group_ctr];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr = output_buf[0];
-  /* Loop for each pair of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-    outptr += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-    outptr += RGB_PIXELSIZE;
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-  }
-}
-
-
-/*
- * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
- */
-
-INLINE
-LOCAL(void)
-h2v2_merged_upsample_internal (j_decompress_ptr cinfo,
-                               JSAMPIMAGE input_buf,
-                               JDIMENSION in_row_group_ctr,
-                               JSAMPARRAY output_buf)
-{
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr0, outptr1;
-  JSAMPROW inptr00, inptr01, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr00 = input_buf[0][in_row_group_ctr*2];
-  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr0 = output_buf[0];
-  outptr1 = output_buf[1];
-  /* Loop for each group of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr00);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    y  = GETJSAMPLE(*inptr01);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdphuff.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdphuff.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdphuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdphuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdphuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy decoding routines for progressive JPEG.
  *
@@ -17,7 +20,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdhuff.h"		/* Declarations shared with jdhuff.c */
+#include "jdhuff.h"             /* Declarations shared with jdhuff.c */
 
 
 #ifdef D_PROGRESSIVE_SUPPORTED
@@ -30,8 +33,8 @@
  */
 
 typedef struct {
-  unsigned int EOBRUN;			/* remaining EOBs in EOBRUN */
-  int last_dc_val[MAX_COMPS_IN_SCAN];	/* last DC coef for each component */
+  unsigned int EOBRUN;                  /* remaining EOBs in EOBRUN */
+  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;
 
 /* This macro is to work around compilers with missing or broken
@@ -44,11 +47,11 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).EOBRUN = (src).EOBRUN, \
-	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).EOBRUN = (src).EOBRUN, \
+         (dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -59,29 +62,29 @@
   /* These fields are loaded into local variables at start of each MCU.
    * In case of suspension, we exit WITHOUT updating them.
    */
-  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
-  savable_state saved;		/* Other state at start of MCU */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  d_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *derived_tbls[NUM_HUFF_TBLS];
 
-  d_derived_tbl * ac_derived_tbl; /* active table during an AC scan */
+  d_derived_tbl *ac_derived_tbl; /* active table during an AC scan */
 } phuff_entropy_decoder;
 
-typedef phuff_entropy_decoder * phuff_entropy_ptr;
+typedef phuff_entropy_decoder *phuff_entropy_ptr;
 
 /* Forward declarations */
-METHODDEF(boolean) decode_mcu_DC_first JPP((j_decompress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_AC_first JPP((j_decompress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_DC_refine JPP((j_decompress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_AC_refine JPP((j_decompress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_DC_first (j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_first (j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_DC_refine (j_decompress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_refine (j_decompress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
 
 
 /*
@@ -94,8 +97,9 @@
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band, bad;
   int ci, coefi, tbl;
+  d_derived_tbl **pdtbl;
   int *coef_bit_ptr;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   is_DC_band = (cinfo->Ss == 0);
 
@@ -117,7 +121,7 @@
     if (cinfo->Al != cinfo->Ah-1)
       bad = TRUE;
   }
-  if (cinfo->Al > 13)		/* need not check for < 0 */
+  if (cinfo->Al > 13)           /* need not check for < 0 */
     bad = TRUE;
   /* Arguably the maximum Al value should be less than 13 for 8-bit precision,
    * but the spec doesn't say so, and we try to be liberal about what we
@@ -127,7 +131,7 @@
    */
   if (bad)
     ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
-	     cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+             cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
   /* Update progression status, and verify that scan order is legal.
    * Note that inter-scan inconsistencies are treated as warnings
    * not fatal errors ... not clear if this is right way to behave.
@@ -140,7 +144,7 @@
     for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
       int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
       if (cinfo->Ah != expected)
-	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
       coef_bit_ptr[coefi] = cinfo->Al;
     }
   }
@@ -164,15 +168,15 @@
      * We may build same derived table more than once, but it's not expensive.
      */
     if (is_DC_band) {
-      if (cinfo->Ah == 0) {	/* DC refinement needs no table */
-	tbl = compptr->dc_tbl_no;
-	jpeg_make_d_derived_tbl(cinfo, TRUE, tbl,
-				& entropy->derived_tbls[tbl]);
+      if (cinfo->Ah == 0) {     /* DC refinement needs no table */
+        tbl = compptr->dc_tbl_no;
+        pdtbl = entropy->derived_tbls + tbl;
+        jpeg_make_d_derived_tbl(cinfo, TRUE, tbl, pdtbl);
       }
     } else {
       tbl = compptr->ac_tbl_no;
-      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl,
-			      & entropy->derived_tbls[tbl]);
+      pdtbl = entropy->derived_tbls + tbl;
+      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl, pdtbl);
       /* remember the single active table */
       entropy->ac_derived_tbl = entropy->derived_tbls[tbl];
     }
@@ -198,9 +202,11 @@
  * On some machines, a shift and add will be faster than a table lookup.
  */
 
+#define AVOID_TABLES
 #ifdef AVOID_TABLES
 
-#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+#define NEG_1 ((unsigned)-1)
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((NEG_1)<<(s)) + 1) : (x))
 
 #else
 
@@ -263,7 +269,7 @@
 /*
  * Huffman MCU decoding.
  * Each of these routines decodes and returns one MCU's worth of
- * Huffman-compressed coefficients. 
+ * Huffman-compressed coefficients.
  * The coefficients are reordered from zigzag order into natural array order,
  * but are not dequantized.
  *
@@ -284,7 +290,7 @@
 
 METHODDEF(boolean)
 decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Al = cinfo->Al;
   register int s, r;
@@ -292,14 +298,14 @@
   JBLOCKROW block;
   BITREAD_STATE_VARS;
   savable_state state;
-  d_derived_tbl * tbl;
-  jpeg_component_info * compptr;
+  d_derived_tbl *tbl;
+  jpeg_component_info *compptr;
 
   /* Process restart marker if needed; may have to suspend */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, just leave the MCU set to zeroes.
@@ -324,16 +330,16 @@
       /* Section F.2.2.1: decode the DC coefficient difference */
       HUFF_DECODE(s, br_state, tbl, return FALSE, label1);
       if (s) {
-	CHECK_BIT_BUFFER(br_state, s, return FALSE);
-	r = GET_BITS(s);
-	s = HUFF_EXTEND(r, s);
+        CHECK_BIT_BUFFER(br_state, s, return FALSE);
+        r = GET_BITS(s);
+        s = HUFF_EXTEND(r, s);
       }
 
       /* Convert DC difference to actual value, update last_dc_val */
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
       /* Scale and output the coefficient (assumes jpeg_natural_order[0]=0) */
-      (*block)[0] = (JCOEF) (s << Al);
+      (*block)[0] = (JCOEF) LEFT_SHIFT(s, Al);
     }
 
     /* Completed MCU, so update state */
@@ -355,7 +361,7 @@
 
 METHODDEF(boolean)
 decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Se = cinfo->Se;
   int Al = cinfo->Al;
@@ -363,13 +369,13 @@
   unsigned int EOBRUN;
   JBLOCKROW block;
   BITREAD_STATE_VARS;
-  d_derived_tbl * tbl;
+  d_derived_tbl *tbl;
 
   /* Process restart marker if needed; may have to suspend */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, just leave the MCU set to zeroes.
@@ -380,49 +386,49 @@
     /* Load up working state.
      * We can avoid loading/saving bitread state if in an EOB run.
      */
-    EOBRUN = entropy->saved.EOBRUN;	/* only part of saved state we need */
+    EOBRUN = entropy->saved.EOBRUN;     /* only part of saved state we need */
 
     /* There is always only one block per MCU */
 
-    if (EOBRUN > 0)		/* if it's a band of zeroes... */
-      EOBRUN--;			/* ...process it now (we do nothing) */
+    if (EOBRUN > 0)             /* if it's a band of zeroes... */
+      EOBRUN--;                 /* ...process it now (we do nothing) */
     else {
       BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
       block = MCU_data[0];
       tbl = entropy->ac_derived_tbl;
 
       for (k = cinfo->Ss; k <= Se; k++) {
-	HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
-	r = s >> 4;
-	s &= 15;
-	if (s) {
-	  k += r;
-	  CHECK_BIT_BUFFER(br_state, s, return FALSE);
-	  r = GET_BITS(s);
-	  s = HUFF_EXTEND(r, s);
-	  /* Scale and output coefficient in natural (dezigzagged) order */
-	  (*block)[jpeg_natural_order[k]] = (JCOEF) (s << Al);
-	} else {
-	  if (r == 15) {	/* ZRL */
-	    k += 15;		/* skip 15 zeroes in band */
-	  } else {		/* EOBr, run length is 2^r + appended bits */
-	    EOBRUN = 1 << r;
-	    if (r) {		/* EOBr, r > 0 */
-	      CHECK_BIT_BUFFER(br_state, r, return FALSE);
-	      r = GET_BITS(r);
-	      EOBRUN += r;
-	    }
-	    EOBRUN--;		/* this band is processed at this moment */
-	    break;		/* force end-of-band */
-	  }
-	}
+        HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          r = GET_BITS(s);
+          s = HUFF_EXTEND(r, s);
+          /* Scale and output coefficient in natural (dezigzagged) order */
+          (*block)[jpeg_natural_order[k]] = (JCOEF) LEFT_SHIFT(s, Al);
+        } else {
+          if (r == 15) {        /* ZRL */
+            k += 15;            /* skip 15 zeroes in band */
+          } else {              /* EOBr, run length is 2^r + appended bits */
+            EOBRUN = 1 << r;
+            if (r) {            /* EOBr, r > 0 */
+              CHECK_BIT_BUFFER(br_state, r, return FALSE);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            EOBRUN--;           /* this band is processed at this moment */
+            break;              /* force end-of-band */
+          }
+        }
       }
 
       BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
     }
 
     /* Completed MCU, so update state */
-    entropy->saved.EOBRUN = EOBRUN;	/* only part of saved state we need */
+    entropy->saved.EOBRUN = EOBRUN;     /* only part of saved state we need */
   }
 
   /* Account for restart interval (no-op if not using restarts) */
@@ -440,9 +446,9 @@
 
 METHODDEF(boolean)
 decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
-  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int p1 = 1 << cinfo->Al;      /* 1 in the bit position being coded */
   int blkn;
   JBLOCKROW block;
   BITREAD_STATE_VARS;
@@ -451,7 +457,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* Not worth the cycles to check insufficient_data here,
@@ -489,17 +495,17 @@
 
 METHODDEF(boolean)
 decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Se = cinfo->Se;
-  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
-  int m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  int p1 = 1 << cinfo->Al;        /* 1 in the bit position being coded */
+  int m1 = (NEG_1) << cinfo->Al;  /* -1 in the bit position being coded */
   register int s, k, r;
   unsigned int EOBRUN;
   JBLOCKROW block;
   JCOEFPTR thiscoef;
   BITREAD_STATE_VARS;
-  d_derived_tbl * tbl;
+  d_derived_tbl *tbl;
   int num_newnz;
   int newnz_pos[DCTSIZE2];
 
@@ -507,7 +513,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, don't modify the MCU.
@@ -535,58 +541,58 @@
 
     if (EOBRUN == 0) {
       for (; k <= Se; k++) {
-	HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
-	r = s >> 4;
-	s &= 15;
-	if (s) {
-	  if (s != 1)		/* size of new coef should always be 1 */
-	    WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
-	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	  if (GET_BITS(1))
-	    s = p1;		/* newly nonzero coef is positive */
-	  else
-	    s = m1;		/* newly nonzero coef is negative */
-	} else {
-	  if (r != 15) {
-	    EOBRUN = 1 << r;	/* EOBr, run length is 2^r + appended bits */
-	    if (r) {
-	      CHECK_BIT_BUFFER(br_state, r, goto undoit);
-	      r = GET_BITS(r);
-	      EOBRUN += r;
-	    }
-	    break;		/* rest of block is handled by EOB logic */
-	  }
-	  /* note s = 0 for processing ZRL */
-	}
-	/* Advance over already-nonzero coefs and r still-zero coefs,
-	 * appending correction bits to the nonzeroes.  A correction bit is 1
-	 * if the absolute value of the coefficient must be increased.
-	 */
-	do {
-	  thiscoef = *block + jpeg_natural_order[k];
-	  if (*thiscoef != 0) {
-	    CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	    if (GET_BITS(1)) {
-	      if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
-		if (*thiscoef >= 0)
-		  *thiscoef += p1;
-		else
-		  *thiscoef += m1;
-	      }
-	    }
-	  } else {
-	    if (--r < 0)
-	      break;		/* reached target zero coefficient */
-	  }
-	  k++;
-	} while (k <= Se);
-	if (s) {
-	  int pos = jpeg_natural_order[k];
-	  /* Output newly nonzero coefficient */
-	  (*block)[pos] = (JCOEF) s;
-	  /* Remember its position in case we have to suspend */
-	  newnz_pos[num_newnz++] = pos;
-	}
+        HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          if (s != 1)           /* size of new coef should always be 1 */
+            WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1))
+            s = p1;             /* newly nonzero coef is positive */
+          else
+            s = m1;             /* newly nonzero coef is negative */
+        } else {
+          if (r != 15) {
+            EOBRUN = 1 << r;    /* EOBr, run length is 2^r + appended bits */
+            if (r) {
+              CHECK_BIT_BUFFER(br_state, r, goto undoit);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            break;              /* rest of block is handled by EOB logic */
+          }
+          /* note s = 0 for processing ZRL */
+        }
+        /* Advance over already-nonzero coefs and r still-zero coefs,
+         * appending correction bits to the nonzeroes.  A correction bit is 1
+         * if the absolute value of the coefficient must be increased.
+         */
+        do {
+          thiscoef = *block + jpeg_natural_order[k];
+          if (*thiscoef != 0) {
+            CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+            if (GET_BITS(1)) {
+              if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
+                if (*thiscoef >= 0)
+                  *thiscoef += p1;
+                else
+                  *thiscoef += m1;
+              }
+            }
+          } else {
+            if (--r < 0)
+              break;            /* reached target zero coefficient */
+          }
+          k++;
+        } while (k <= Se);
+        if (s) {
+          int pos = jpeg_natural_order[k];
+          /* Output newly nonzero coefficient */
+          (*block)[pos] = (JCOEF) s;
+          /* Remember its position in case we have to suspend */
+          newnz_pos[num_newnz++] = pos;
+        }
       }
     }
 
@@ -597,18 +603,18 @@
        * if the absolute value of the coefficient must be increased.
        */
       for (; k <= Se; k++) {
-	thiscoef = *block + jpeg_natural_order[k];
-	if (*thiscoef != 0) {
-	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	  if (GET_BITS(1)) {
-	    if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
-	      if (*thiscoef >= 0)
-		*thiscoef += p1;
-	      else
-		*thiscoef += m1;
-	    }
-	  }
-	}
+        thiscoef = *block + jpeg_natural_order[k];
+        if (*thiscoef != 0) {
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1)) {
+            if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
+              if (*thiscoef >= 0)
+                *thiscoef += p1;
+              else
+                *thiscoef += m1;
+            }
+          }
+        }
       }
       /* Count one block completed in EOB run */
       EOBRUN--;
@@ -646,7 +652,7 @@
 
   entropy = (phuff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(phuff_entropy_decoder));
+                                sizeof(phuff_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass_phuff_decoder;
 
@@ -658,9 +664,9 @@
   /* Create progression status table */
   cinfo->coef_bits = (int (*)[DCTSIZE2])
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				cinfo->num_components*DCTSIZE2*SIZEOF(int));
+                                cinfo->num_components*DCTSIZE2*sizeof(int));
   coef_bit_ptr = & cinfo->coef_bits[0][0];
-  for (ci = 0; ci < cinfo->num_components; ci++) 
+  for (ci = 0; ci < cinfo->num_components; ci++)
     for (i = 0; i < DCTSIZE2; i++)
       *coef_bit_ptr++ = -1;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdpostct.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdpostct.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdpostct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdpostct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdpostct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the decompression postprocessing controller.
  * This controller manages the upsampling, color conversion, and color
@@ -31,37 +34,34 @@
    * For two-pass color quantization, we need a full-image buffer;
    * for one-pass operation, a strip buffer is sufficient.
    */
-  jvirt_sarray_ptr whole_image;	/* virtual array, or NULL if one-pass */
-  JSAMPARRAY buffer;		/* strip buffer, or current strip of virtual */
-  JDIMENSION strip_height;	/* buffer size in rows */
+  jvirt_sarray_ptr whole_image; /* virtual array, or NULL if one-pass */
+  JSAMPARRAY buffer;            /* strip buffer, or current strip of virtual */
+  JDIMENSION strip_height;      /* buffer size in rows */
   /* for two-pass mode only: */
-  JDIMENSION starting_row;	/* row # of first row in current strip */
-  JDIMENSION next_row;		/* index of next row to fill/empty in strip */
+  JDIMENSION starting_row;      /* row # of first row in current strip */
+  JDIMENSION next_row;          /* index of next row to fill/empty in strip */
 } my_post_controller;
 
-typedef my_post_controller * my_post_ptr;
+typedef my_post_controller *my_post_ptr;
 
 
 /* Forward declarations */
 METHODDEF(void) post_process_1pass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 #ifdef QUANT_2PASS_SUPPORTED
 METHODDEF(void) post_process_prepass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 METHODDEF(void) post_process_2pass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 #endif
 
 
@@ -84,9 +84,9 @@
        * allocate a strip buffer.  Use the virtual-array buffer as workspace.
        */
       if (post->buffer == NULL) {
-	post->buffer = (*cinfo->mem->access_virt_sarray)
-	  ((j_common_ptr) cinfo, post->whole_image,
-	   (JDIMENSION) 0, post->strip_height, TRUE);
+        post->buffer = (*cinfo->mem->access_virt_sarray)
+          ((j_common_ptr) cinfo, post->whole_image,
+           (JDIMENSION) 0, post->strip_height, TRUE);
       }
     } else {
       /* For single-pass processing without color quantization,
@@ -124,10 +124,10 @@
 
 METHODDEF(void)
 post_process_1pass (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION num_rows, max_rows;
@@ -139,11 +139,11 @@
     max_rows = post->strip_height;
   num_rows = 0;
   (*cinfo->upsample->upsample) (cinfo,
-		input_buf, in_row_group_ctr, in_row_groups_avail,
-		post->buffer, &num_rows, max_rows);
+                input_buf, in_row_group_ctr, in_row_groups_avail,
+                post->buffer, &num_rows, max_rows);
   /* Quantize and emit data. */
   (*cinfo->cquantize->color_quantize) (cinfo,
-		post->buffer, output_buf + *out_row_ctr, (int) num_rows);
+                post->buffer, output_buf + *out_row_ctr, (int) num_rows);
   *out_row_ctr += num_rows;
 }
 
@@ -156,10 +156,10 @@
 
 METHODDEF(void)
 post_process_prepass (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		      JDIMENSION in_row_groups_avail,
-		      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		      JDIMENSION out_rows_avail)
+                      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                      JDIMENSION in_row_groups_avail,
+                      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                      JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION old_next_row, num_rows;
@@ -167,22 +167,22 @@
   /* Reposition virtual buffer if at start of strip. */
   if (post->next_row == 0) {
     post->buffer = (*cinfo->mem->access_virt_sarray)
-	((j_common_ptr) cinfo, post->whole_image,
-	 post->starting_row, post->strip_height, TRUE);
+        ((j_common_ptr) cinfo, post->whole_image,
+         post->starting_row, post->strip_height, TRUE);
   }
 
   /* Upsample some data (up to a strip height's worth). */
   old_next_row = post->next_row;
   (*cinfo->upsample->upsample) (cinfo,
-		input_buf, in_row_group_ctr, in_row_groups_avail,
-		post->buffer, &post->next_row, post->strip_height);
+                input_buf, in_row_group_ctr, in_row_groups_avail,
+                post->buffer, &post->next_row, post->strip_height);
 
   /* Allow quantizer to scan new data.  No data is emitted, */
   /* but we advance out_row_ctr so outer loop can tell when we're done. */
   if (post->next_row > old_next_row) {
     num_rows = post->next_row - old_next_row;
     (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row,
-					 (JSAMPARRAY) NULL, (int) num_rows);
+                                         (JSAMPARRAY) NULL, (int) num_rows);
     *out_row_ctr += num_rows;
   }
 
@@ -200,10 +200,10 @@
 
 METHODDEF(void)
 post_process_2pass (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION num_rows, max_rows;
@@ -211,8 +211,8 @@
   /* Reposition virtual buffer if at start of strip. */
   if (post->next_row == 0) {
     post->buffer = (*cinfo->mem->access_virt_sarray)
-	((j_common_ptr) cinfo, post->whole_image,
-	 post->starting_row, post->strip_height, FALSE);
+        ((j_common_ptr) cinfo, post->whole_image,
+         post->starting_row, post->strip_height, FALSE);
   }
 
   /* Determine number of rows to emit. */
@@ -227,8 +227,8 @@
 
   /* Quantize and emit data. */
   (*cinfo->cquantize->color_quantize) (cinfo,
-		post->buffer + post->next_row, output_buf + *out_row_ctr,
-		(int) num_rows);
+                post->buffer + post->next_row, output_buf + *out_row_ctr,
+                (int) num_rows);
   *out_row_ctr += num_rows;
 
   /* Advance if we filled the strip. */
@@ -253,11 +253,11 @@
 
   post = (my_post_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_post_controller));
+                                sizeof(my_post_controller));
   cinfo->post = (struct jpeg_d_post_controller *) post;
   post->pub.start_pass = start_pass_dpost;
-  post->whole_image = NULL;	/* flag for no virtual arrays */
-  post->buffer = NULL;		/* flag for no strip buffer */
+  post->whole_image = NULL;     /* flag for no virtual arrays */
+  post->buffer = NULL;          /* flag for no strip buffer */
 
   /* Create the quantization buffer, if needed */
   if (cinfo->quantize_colors) {
@@ -271,20 +271,20 @@
       /* We round up the number of rows to a multiple of the strip height. */
 #ifdef QUANT_2PASS_SUPPORTED
       post->whole_image = (*cinfo->mem->request_virt_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 cinfo->output_width * cinfo->out_color_components,
-	 (JDIMENSION) jround_up((long) cinfo->output_height,
-				(long) post->strip_height),
-	 post->strip_height);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+         cinfo->output_width * cinfo->out_color_components,
+         (JDIMENSION) jround_up((long) cinfo->output_height,
+                                (long) post->strip_height),
+         post->strip_height);
 #else
       ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 #endif /* QUANT_2PASS_SUPPORTED */
     } else {
       /* One-pass color quantization: just make a strip buffer. */
       post->buffer = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 cinfo->output_width * cinfo->out_color_components,
-	 post->strip_height);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         cinfo->output_width * cinfo->out_color_components,
+         post->strip_height);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdsample.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdsample.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdsample.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdsample.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jdsample.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains upsampling routines.
  *
@@ -20,50 +24,12 @@
  *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
+#include "jdsample.h"
 #include "jsimd.h"
 #include "jpegcomp.h"
 
 
-/* Pointer to routine to upsample a single component */
-typedef JMETHOD(void, upsample1_ptr,
-		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-
-/* Private subobject */
-
-typedef struct {
-  struct jpeg_upsampler pub;	/* public fields */
-
-  /* Color conversion buffer.  When using separate upsampling and color
-   * conversion steps, this buffer holds one upsampled row group until it
-   * has been color converted and output.
-   * Note: we do not allocate any storage for component(s) which are full-size,
-   * ie do not need rescaling.  The corresponding entry of color_buf[] is
-   * simply set to point to the input data array, thereby avoiding copying.
-   */
-  JSAMPARRAY color_buf[MAX_COMPONENTS];
-
-  /* Per-component upsampling method pointers */
-  upsample1_ptr methods[MAX_COMPONENTS];
-
-  int next_row_out;		/* counts rows emitted from color_buf */
-  JDIMENSION rows_to_go;	/* counts rows remaining in image */
-
-  /* Height of an input row group for each component. */
-  int rowgroup_height[MAX_COMPONENTS];
-
-  /* These arrays save pixel expansion factors so that int_expand need not
-   * recompute them each time.  They are unused for other upsampling methods.
-   */
-  UINT8 h_expand[MAX_COMPONENTS];
-  UINT8 v_expand[MAX_COMPONENTS];
-} my_upsampler;
-
-typedef my_upsampler * my_upsample_ptr;
-
 
 /*
  * Initialize for an upsampling pass.
@@ -91,26 +57,26 @@
 
 METHODDEF(void)
 sep_upsample (j_decompress_ptr cinfo,
-	      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	      JDIMENSION in_row_groups_avail,
-	      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	      JDIMENSION out_rows_avail)
+              JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+              JDIMENSION in_row_groups_avail,
+              JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+              JDIMENSION out_rows_avail)
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JDIMENSION num_rows;
 
   /* Fill the conversion buffer, if it's empty */
   if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       /* Invoke per-component upsample method.  Notice we pass a POINTER
        * to color_buf[ci], so that fullsize_upsample can change it.
        */
       (*upsample->methods[ci]) (cinfo, compptr,
-	input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
-	upsample->color_buf + ci);
+        input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
+        upsample->color_buf + ci);
     }
     upsample->next_row_out = 0;
   }
@@ -122,7 +88,7 @@
   /* Not more than the distance to the end of the image.  Need this test
    * in case the image height is not a multiple of max_v_samp_factor:
    */
-  if (num_rows > upsample->rows_to_go) 
+  if (num_rows > upsample->rows_to_go)
     num_rows = upsample->rows_to_go;
   /* And not more than what the client can accept: */
   out_rows_avail -= *out_row_ctr;
@@ -130,9 +96,9 @@
     num_rows = out_rows_avail;
 
   (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
-				     (JDIMENSION) upsample->next_row_out,
-				     output_buf + *out_row_ctr,
-				     (int) num_rows);
+                                     (JDIMENSION) upsample->next_row_out,
+                                     output_buf + *out_row_ctr,
+                                     (int) num_rows);
 
   /* Adjust counts */
   *out_row_ctr += num_rows;
@@ -158,8 +124,8 @@
  */
 
 METHODDEF(void)
-fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		   JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   *output_data_ptr = input_data;
 }
@@ -171,10 +137,10 @@
  */
 
 METHODDEF(void)
-noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+noop_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
-  *output_data_ptr = NULL;	/* safety check */
+  *output_data_ptr = NULL;      /* safety check */
 }
 
 
@@ -190,8 +156,8 @@
  */
 
 METHODDEF(void)
-int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	      JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   JSAMPARRAY output_data = *output_data_ptr;
@@ -212,15 +178,15 @@
     outptr = output_data[outrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       for (h = h_expand; h > 0; h--) {
-	*outptr++ = invalue;
+        *outptr++ = invalue;
       }
     }
     /* Generate any additional output rows by duplicating the first one */
     if (v_expand > 1) {
       jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
-			v_expand-1, cinfo->output_width);
+                        v_expand-1, cinfo->output_width);
     }
     inrow++;
     outrow += v_expand;
@@ -234,8 +200,8 @@
  */
 
 METHODDEF(void)
-h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -248,7 +214,7 @@
     outptr = output_data[inrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       *outptr++ = invalue;
       *outptr++ = invalue;
     }
@@ -262,8 +228,8 @@
  */
 
 METHODDEF(void)
-h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -277,12 +243,12 @@
     outptr = output_data[outrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       *outptr++ = invalue;
       *outptr++ = invalue;
     }
     jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
-		      1, cinfo->output_width);
+                      1, cinfo->output_width);
     inrow++;
     outrow += 2;
   }
@@ -305,8 +271,8 @@
  */
 
 METHODDEF(void)
-h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -346,15 +312,15 @@
  */
 
 METHODDEF(void)
-h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr0, inptr1, outptr;
 #if BITS_IN_JSAMPLE == 8
   register int thiscolsum, lastcolsum, nextcolsum;
 #else
-  register INT32 thiscolsum, lastcolsum, nextcolsum;
+  register JLONG thiscolsum, lastcolsum, nextcolsum;
 #endif
   register JDIMENSION colctr;
   int inrow, outrow, v;
@@ -364,10 +330,10 @@
     for (v = 0; v < 2; v++) {
       /* inptr0 points to nearest input row, inptr1 points to next nearest */
       inptr0 = input_data[inrow];
-      if (v == 0)		/* next nearest is row above */
-	inptr1 = input_data[inrow-1];
-      else			/* next nearest is row below */
-	inptr1 = input_data[inrow+1];
+      if (v == 0)               /* next nearest is row above */
+        inptr1 = input_data[inrow-1];
+      else                      /* next nearest is row below */
+        inptr1 = input_data[inrow+1];
       outptr = output_data[outrow++];
 
       /* Special case for first column */
@@ -378,12 +344,12 @@
       lastcolsum = thiscolsum; thiscolsum = nextcolsum;
 
       for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
-	/* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
-	/* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
-	nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
-	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
-	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
-	lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+        /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
+        /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
+        nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+        *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+        lastcolsum = thiscolsum; thiscolsum = nextcolsum;
       }
 
       /* Special case for last column */
@@ -404,19 +370,22 @@
 {
   my_upsample_ptr upsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   boolean need_buffer, do_fancy;
   int h_in_group, v_in_group, h_out_group, v_out_group;
 
-  upsample = (my_upsample_ptr)
-    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_upsampler));
-  cinfo->upsample = (struct jpeg_upsampler *) upsample;
-  upsample->pub.start_pass = start_pass_upsample;
-  upsample->pub.upsample = sep_upsample;
-  upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+  if (!cinfo->master->jinit_upsampler_no_alloc) {
+    upsample = (my_upsample_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(my_upsampler));
+    cinfo->upsample = (struct jpeg_upsampler *) upsample;
+    upsample->pub.start_pass = start_pass_upsample;
+    upsample->pub.upsample = sep_upsample;
+    upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+  } else
+    upsample = (my_upsample_ptr) cinfo->upsample;
 
-  if (cinfo->CCIR601_sampling)	/* this isn't supported */
+  if (cinfo->CCIR601_sampling)  /* this isn't supported */
     ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
 
   /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
@@ -433,9 +402,9 @@
      * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
      */
     h_in_group = (compptr->h_samp_factor * compptr->_DCT_scaled_size) /
-		 cinfo->_min_DCT_scaled_size;
+                 cinfo->_min_DCT_scaled_size;
     v_in_group = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
-		 cinfo->_min_DCT_scaled_size;
+                 cinfo->_min_DCT_scaled_size;
     h_out_group = cinfo->max_h_samp_factor;
     v_out_group = cinfo->max_v_samp_factor;
     upsample->rowgroup_height[ci] = v_in_group; /* save for use later */
@@ -449,48 +418,53 @@
       upsample->methods[ci] = fullsize_upsample;
       need_buffer = FALSE;
     } else if (h_in_group * 2 == h_out_group &&
-	       v_in_group == v_out_group) {
+               v_in_group == v_out_group) {
       /* Special cases for 2h1v upsampling */
       if (do_fancy && compptr->downsampled_width > 2) {
-	if (jsimd_can_h2v1_fancy_upsample())
-	  upsample->methods[ci] = jsimd_h2v1_fancy_upsample;
-	else
-	  upsample->methods[ci] = h2v1_fancy_upsample;
+        if (jsimd_can_h2v1_fancy_upsample())
+          upsample->methods[ci] = jsimd_h2v1_fancy_upsample;
+        else
+          upsample->methods[ci] = h2v1_fancy_upsample;
       } else {
-	if (jsimd_can_h2v1_upsample())
-	  upsample->methods[ci] = jsimd_h2v1_upsample;
-	else
-	  upsample->methods[ci] = h2v1_upsample;
+        if (jsimd_can_h2v1_upsample())
+          upsample->methods[ci] = jsimd_h2v1_upsample;
+        else
+          upsample->methods[ci] = h2v1_upsample;
       }
     } else if (h_in_group * 2 == h_out_group &&
-	       v_in_group * 2 == v_out_group) {
+               v_in_group * 2 == v_out_group) {
       /* Special cases for 2h2v upsampling */
       if (do_fancy && compptr->downsampled_width > 2) {
-	if (jsimd_can_h2v2_fancy_upsample())
-	  upsample->methods[ci] = jsimd_h2v2_fancy_upsample;
-	else
-	  upsample->methods[ci] = h2v2_fancy_upsample;
-	upsample->pub.need_context_rows = TRUE;
+        if (jsimd_can_h2v2_fancy_upsample())
+          upsample->methods[ci] = jsimd_h2v2_fancy_upsample;
+        else
+          upsample->methods[ci] = h2v2_fancy_upsample;
+        upsample->pub.need_context_rows = TRUE;
       } else {
-	if (jsimd_can_h2v2_upsample())
-	  upsample->methods[ci] = jsimd_h2v2_upsample;
-	else
-	  upsample->methods[ci] = h2v2_upsample;
+        if (jsimd_can_h2v2_upsample())
+          upsample->methods[ci] = jsimd_h2v2_upsample;
+        else
+          upsample->methods[ci] = h2v2_upsample;
       }
     } else if ((h_out_group % h_in_group) == 0 &&
-	       (v_out_group % v_in_group) == 0) {
+               (v_out_group % v_in_group) == 0) {
       /* Generic integral-factors upsampling method */
-      upsample->methods[ci] = int_upsample;
+#if defined(__mips__)
+      if (jsimd_can_int_upsample())
+        upsample->methods[ci] = jsimd_int_upsample;
+      else
+#endif
+        upsample->methods[ci] = int_upsample;
       upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group);
       upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group);
     } else
       ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
-    if (need_buffer) {
+    if (need_buffer && !cinfo->master->jinit_upsampler_no_alloc) {
       upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 (JDIMENSION) jround_up((long) cinfo->output_width,
-				(long) cinfo->max_h_samp_factor),
-	 (JDIMENSION) cinfo->max_v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         (JDIMENSION) jround_up((long) cinfo->output_width,
+                                (long) cinfo->max_h_samp_factor),
+         (JDIMENSION) cinfo->max_v_samp_factor);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdsample.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jdsample.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdsample.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdsample.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,50 @@
+/*
+ * jdsample.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Pointer to routine to upsample a single component */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;    /* public fields */
+
+  /* Color conversion buffer.  When using separate upsampling and color
+   * conversion steps, this buffer holds one upsampled row group until it
+   * has been color converted and output.
+   * Note: we do not allocate any storage for component(s) which are full-size,
+   * ie do not need rescaling.  The corresponding entry of color_buf[] is
+   * simply set to point to the input data array, thereby avoiding copying.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  /* Per-component upsampling method pointers */
+  upsample1_ptr methods[MAX_COMPONENTS];
+
+  int next_row_out;             /* counts rows emitted from color_buf */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
+
+  /* Height of an input row group for each component. */
+  int rowgroup_height[MAX_COMPONENTS];
+
+  /* These arrays save pixel expansion factors so that int_expand need not
+   * recompute them each time.  They are unused for other upsampling methods.
+   */
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdtrans.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jdtrans.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jdtrans.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jdtrans.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdtrans.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains library routines for transcoding decompression,
  * that is, reading raw DCT coefficient arrays from an input JPEG file.
@@ -16,7 +19,7 @@
 
 
 /* Forward declarations */
-LOCAL(void) transdecode_master_selection JPP((j_decompress_ptr cinfo));
+LOCAL(void) transdecode_master_selection (j_decompress_ptr cinfo);
 
 
 /*
@@ -55,20 +58,20 @@
       int retcode;
       /* Call progress monitor hook if present */
       if (cinfo->progress != NULL)
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       /* Absorb some more input */
       retcode = (*cinfo->inputctl->consume_input) (cinfo);
       if (retcode == JPEG_SUSPENDED)
-	return NULL;
+        return NULL;
       if (retcode == JPEG_REACHED_EOI)
-	break;
+        break;
       /* Advance progress counter if appropriate */
       if (cinfo->progress != NULL &&
-	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
-	if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
-	  /* startup underestimated number of scans; ratchet up one scan */
-	  cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
-	}
+          (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+          /* startup underestimated number of scans; ratchet up one scan */
+          cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+        }
       }
     }
     /* Set state so that jpeg_finish_decompress does the right thing */
@@ -84,7 +87,7 @@
   }
   /* Oops, improper usage */
   ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
-  return NULL;			/* keep compiler happy */
+  return NULL;                  /* keep compiler happy */
 }
 
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jerror.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jerror.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jerror.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jerror.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jerror.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains simple error-reporting and trace-message routines.
  * These are suitable for Unix-like systems and others where writing to
@@ -28,7 +31,7 @@
 #include <windows.h>
 #endif
 
-#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
 #define EXIT_FAILURE  1
 #endif
 
@@ -41,11 +44,7 @@
  * want to refer to it directly.
  */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_std_message_table	jMsgTable
-#endif
-
-#define JMESSAGE(code,string)	string ,
+#define JMESSAGE(code,string)   string ,
 
 const char * const jpeg_std_message_table[] = {
 #include "jerror.h"
@@ -105,7 +104,7 @@
 #ifdef USE_WINDOWS_MESSAGEBOX
   /* Display it in a message dialog box */
   MessageBox(GetActiveWindow(), buffer, "JPEG Library Error",
-	     MB_OK | MB_ICONERROR);
+             MB_OK | MB_ICONERROR);
 #else
   /* Send it to stderr, adding a newline */
   fprintf(stderr, "%s\n", buffer);
@@ -127,7 +126,7 @@
 METHODDEF(void)
 emit_message (j_common_ptr cinfo, int msg_level)
 {
-  struct jpeg_error_mgr * err = cinfo->err;
+  struct jpeg_error_mgr *err = cinfo->err;
 
   if (msg_level < 0) {
     /* It's a warning message.  Since corrupt files may generate many warnings,
@@ -154,12 +153,12 @@
  */
 
 METHODDEF(void)
-format_message (j_common_ptr cinfo, char * buffer)
+format_message (j_common_ptr cinfo, char *buffer)
 {
-  struct jpeg_error_mgr * err = cinfo->err;
+  struct jpeg_error_mgr *err = cinfo->err;
   int msg_code = err->msg_code;
-  const char * msgtext = NULL;
-  const char * msgptr;
+  const char *msgtext = NULL;
+  const char *msgptr;
   char ch;
   boolean isstring;
 
@@ -167,8 +166,8 @@
   if (msg_code > 0 && msg_code <= err->last_jpeg_message) {
     msgtext = err->jpeg_message_table[msg_code];
   } else if (err->addon_message_table != NULL &&
-	     msg_code >= err->first_addon_message &&
-	     msg_code <= err->last_addon_message) {
+             msg_code >= err->first_addon_message &&
+             msg_code <= err->last_addon_message) {
     msgtext = err->addon_message_table[msg_code - err->first_addon_message];
   }
 
@@ -193,10 +192,10 @@
     sprintf(buffer, msgtext, err->msg_parm.s);
   else
     sprintf(buffer, msgtext,
-	    err->msg_parm.i[0], err->msg_parm.i[1],
-	    err->msg_parm.i[2], err->msg_parm.i[3],
-	    err->msg_parm.i[4], err->msg_parm.i[5],
-	    err->msg_parm.i[6], err->msg_parm.i[7]);
+            err->msg_parm.i[0], err->msg_parm.i[1],
+            err->msg_parm.i[2], err->msg_parm.i[3],
+            err->msg_parm.i[4], err->msg_parm.i[5],
+            err->msg_parm.i[6], err->msg_parm.i[7]);
 }
 
 
@@ -213,22 +212,22 @@
 {
   cinfo->err->num_warnings = 0;
   /* trace_level is not reset since it is an application-supplied parameter */
-  cinfo->err->msg_code = 0;	/* may be useful as a flag for "no error" */
+  cinfo->err->msg_code = 0;     /* may be useful as a flag for "no error" */
 }
 
 
 /*
  * Fill in the standard error-handling methods in a jpeg_error_mgr object.
  * Typical call is:
- *	struct jpeg_compress_struct cinfo;
- *	struct jpeg_error_mgr err;
+ *      struct jpeg_compress_struct cinfo;
+ *      struct jpeg_error_mgr err;
  *
- *	cinfo.err = jpeg_std_error(&err);
+ *      cinfo.err = jpeg_std_error(&err);
  * after which the application may override some of the methods.
  */
 
 GLOBAL(struct jpeg_error_mgr *)
-jpeg_std_error (struct jpeg_error_mgr * err)
+jpeg_std_error (struct jpeg_error_mgr *err)
 {
   err->error_exit = error_exit;
   err->emit_message = emit_message;
@@ -236,16 +235,16 @@
   err->format_message = format_message;
   err->reset_error_mgr = reset_error_mgr;
 
-  err->trace_level = 0;		/* default = no tracing */
-  err->num_warnings = 0;	/* no warnings emitted yet */
-  err->msg_code = 0;		/* may be useful as a flag for "no error" */
+  err->trace_level = 0;         /* default = no tracing */
+  err->num_warnings = 0;        /* no warnings emitted yet */
+  err->msg_code = 0;            /* may be useful as a flag for "no error" */
 
   /* Initialize message table pointers */
   err->jpeg_message_table = jpeg_std_message_table;
   err->last_jpeg_message = (int) JMSG_LASTMSGCODE - 1;
 
   err->addon_message_table = NULL;
-  err->first_addon_message = 0;	/* for safety */
+  err->first_addon_message = 0; /* for safety */
   err->last_addon_message = 0;
 
   return err;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jerror.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jerror.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jerror.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jerror.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jerror.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
  * Modified 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file defines the error and message codes for the JPEG library.
  * Edit this file to add new codes, or to translate the message strings to
@@ -33,7 +36,7 @@
 
 typedef enum {
 
-#define JMESSAGE(code,string)	code ,
+#define JMESSAGE(code,string)   code ,
 
 #endif /* JMAKE_ENUM_LIST */
 
@@ -42,7 +45,7 @@
 /* For maintenance convenience, list is alphabetical by message code name */
 #if JPEG_LIB_VERSION < 70
 JMESSAGE(JERR_ARITH_NOTIMPL,
-	 "Sorry, arithmetic coding is not implemented")
+         "Sorry, arithmetic coding is not implemented")
 #endif
 JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix")
 JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix")
@@ -55,26 +58,26 @@
 JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported")
 #if JPEG_LIB_VERSION >= 70
 JMESSAGE(JERR_BAD_DROP_SAMPLING,
-	 "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
+         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
 #endif
 JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition")
 JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace")
 JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace")
 JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length")
 JMESSAGE(JERR_BAD_LIB_VERSION,
-	 "Wrong JPEG library version: library is %d, caller expects %d")
+         "Wrong JPEG library version: library is %d, caller expects %d")
 JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan")
 JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d")
 JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d")
 JMESSAGE(JERR_BAD_PROGRESSION,
-	 "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
+         "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
 JMESSAGE(JERR_BAD_PROG_SCRIPT,
-	 "Invalid progressive parameters at scan script entry %d")
+         "Invalid progressive parameters at scan script entry %d")
 JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors")
 JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d")
 JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d")
 JMESSAGE(JERR_BAD_STRUCT_SIZE,
-	 "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
+         "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
 JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access")
 JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small")
 JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here")
@@ -98,7 +101,7 @@
 JMESSAGE(JERR_INPUT_EMPTY, "Empty input file")
 JMESSAGE(JERR_INPUT_EOF, "Premature end of input file")
 JMESSAGE(JERR_MISMATCHED_QUANT_TABLE,
-	 "Cannot transcode due to multiple use of quantization table %d")
+         "Cannot transcode due to multiple use of quantization table %d")
 JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data")
 JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
 JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
@@ -113,7 +116,7 @@
 JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x")
 JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)")
 JMESSAGE(JERR_QUANT_COMPONENTS,
-	 "Cannot quantize more than %d color components")
+         "Cannot quantize more than %d color components")
 JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors")
 JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors")
 JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers")
@@ -125,19 +128,19 @@
 JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file")
 JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file")
 JMESSAGE(JERR_TFILE_WRITE,
-	 "Write failed on temporary file --- out of disk space?")
+         "Write failed on temporary file --- out of disk space?")
 JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines")
 JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x")
 JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up")
 JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation")
 JMESSAGE(JERR_XMS_READ, "Read from XMS failed")
 JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed")
-JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT)
+JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT_SHORT)
 JMESSAGE(JMSG_VERSION, JVERSION)
 JMESSAGE(JTRC_16BIT_TABLES,
-	 "Caution: quantization tables are too coarse for baseline JPEG")
+         "Caution: quantization tables are too coarse for baseline JPEG")
 JMESSAGE(JTRC_ADOBE,
-	 "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
+         "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
 JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u")
 JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u")
 JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x")
@@ -150,9 +153,9 @@
 JMESSAGE(JTRC_HUFFBITS, "        %3d %3d %3d %3d %3d %3d %3d %3d")
 JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d  %d")
 JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE,
-	 "Warning: thumbnail image size does not match data length %u")
+         "Warning: thumbnail image size does not match data length %u")
 JMESSAGE(JTRC_JFIF_EXTENSION,
-	 "JFIF extension marker: type 0x%02x, length %u")
+         "JFIF extension marker: type 0x%02x, length %u")
 JMESSAGE(JTRC_JFIF_THUMBNAIL, "    with %d x %d thumbnail image")
 JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u")
 JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x")
@@ -163,7 +166,7 @@
 JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d")
 JMESSAGE(JTRC_RST, "RST%d")
 JMESSAGE(JTRC_SMOOTH_NOTIMPL,
-	 "Smoothing not supported with nonstandard sampling ratios")
+         "Smoothing not supported with nonstandard sampling ratios")
 JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d")
 JMESSAGE(JTRC_SOF_COMPONENT, "    Component %d: %dhx%dv q=%d")
 JMESSAGE(JTRC_SOI, "Start of Image")
@@ -173,13 +176,13 @@
 JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s")
 JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s")
 JMESSAGE(JTRC_THUMB_JPEG,
-	 "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
+         "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
 JMESSAGE(JTRC_THUMB_PALETTE,
-	 "JFIF extension marker: palette thumbnail image, length %u")
+         "JFIF extension marker: palette thumbnail image, length %u")
 JMESSAGE(JTRC_THUMB_RGB,
-	 "JFIF extension marker: RGB thumbnail image, length %u")
+         "JFIF extension marker: RGB thumbnail image, length %u")
 JMESSAGE(JTRC_UNKNOWN_IDS,
-	 "Unrecognized component IDs %d %d %d, assuming YCbCr")
+         "Unrecognized component IDs %d %d %d, assuming YCbCr")
 JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
 JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
 JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
@@ -187,15 +190,15 @@
 JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
 #endif
 JMESSAGE(JWRN_BOGUS_PROGRESSION,
-	 "Inconsistent progression sequence for component %d coefficient %d")
+         "Inconsistent progression sequence for component %d coefficient %d")
 JMESSAGE(JWRN_EXTRANEOUS_DATA,
-	 "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
+         "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
 JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment")
 JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code")
 JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d")
 JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file")
 JMESSAGE(JWRN_MUST_RESYNC,
-	 "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
+         "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
 JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG")
 JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines")
 #if JPEG_LIB_VERSION < 70
@@ -255,7 +258,7 @@
    strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
    (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
 
-#define MAKESTMT(stuff)		do { stuff } while (0)
+#define MAKESTMT(stuff)         do { stuff } while (0)
 
 /* Nonfatal errors (we can keep going, but the data is probably corrupt) */
 #define WARNMS(cinfo,code)  \
@@ -286,26 +289,26 @@
    (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
 #define TRACEMS3(cinfo,lvl,code,p1,p2,p3)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS4(cinfo,lvl,code,p1,p2,p3,p4)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS5(cinfo,lvl,code,p1,p2,p3,p4,p5)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   _mp[4] = (p5); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           _mp[4] = (p5); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS8(cinfo,lvl,code,p1,p2,p3,p4,p5,p6,p7,p8)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMSS(cinfo,lvl,code,str)  \
   ((cinfo)->err->msg_code = (code), \
    strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctflt.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctflt.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctflt.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctflt.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1994-1996, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a floating-point implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -20,8 +21,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -37,7 +38,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_FLOAT_SUPPORTED
 
@@ -56,7 +57,7 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_float (FAST_FLOAT * data)
+jpeg_fdct_float (FAST_FLOAT *data)
 {
   FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
@@ -76,24 +77,24 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[0] = tmp10 + tmp11; /* phase 3 */
     dataptr[4] = tmp10 - tmp11;
-    
+
     z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
-    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[2] = tmp13 + z1;    /* phase 5 */
     dataptr[6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -103,15 +104,15 @@
     z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
     z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
-    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[5] = z13 + z2;      /* phase 6 */
     dataptr[3] = z13 - z2;
     dataptr[1] = z11 + z4;
     dataptr[7] = z11 - z4;
 
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns. */
@@ -126,24 +127,24 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
     dataptr[DCTSIZE*4] = tmp10 - tmp11;
-    
+
     z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
     dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
     dataptr[DCTSIZE*6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -153,7 +154,7 @@
     z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
     z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
     dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
@@ -161,7 +162,7 @@
     dataptr[DCTSIZE*1] = z11 + z4;
     dataptr[DCTSIZE*7] = z11 - z4;
 
-    dataptr++;			/* advance pointer to next column */
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctfst.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctfst.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctfst.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctfst.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jfdctfst.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a fast, not so accurate integer implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -15,8 +18,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -33,7 +36,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_IFAST_SUPPORTED
 
@@ -76,10 +79,10 @@
  */
 
 #if CONST_BITS == 8
-#define FIX_0_382683433  ((INT32)   98)		/* FIX(0.382683433) */
-#define FIX_0_541196100  ((INT32)  139)		/* FIX(0.541196100) */
-#define FIX_0_707106781  ((INT32)  181)		/* FIX(0.707106781) */
-#define FIX_1_306562965  ((INT32)  334)		/* FIX(1.306562965) */
+#define FIX_0_382683433  ((JLONG)   98)         /* FIX(0.382683433) */
+#define FIX_0_541196100  ((JLONG)  139)         /* FIX(0.541196100) */
+#define FIX_0_707106781  ((JLONG)  181)         /* FIX(0.707106781) */
+#define FIX_1_306562965  ((JLONG)  334)         /* FIX(1.306562965) */
 #else
 #define FIX_0_382683433  FIX(0.382683433)
 #define FIX_0_541196100  FIX(0.541196100)
@@ -99,7 +102,7 @@
 #endif
 
 
-/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
  * descale to yield a DCTELEM result.
  */
 
@@ -111,7 +114,7 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_ifast (DCTELEM * data)
+jpeg_fdct_ifast (DCTELEM *data)
 {
   DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   DCTELEM tmp10, tmp11, tmp12, tmp13;
@@ -132,24 +135,24 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[0] = tmp10 + tmp11; /* phase 3 */
     dataptr[4] = tmp10 - tmp11;
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
-    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[2] = tmp13 + z1;    /* phase 5 */
     dataptr[6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -159,15 +162,15 @@
     z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
     z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
-    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[5] = z13 + z2;      /* phase 6 */
     dataptr[3] = z13 - z2;
     dataptr[1] = z11 + z4;
     dataptr[7] = z11 - z4;
 
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns. */
@@ -182,24 +185,24 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
     dataptr[DCTSIZE*4] = tmp10 - tmp11;
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
     dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
     dataptr[DCTSIZE*6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -209,7 +212,7 @@
     z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
     z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
     dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
@@ -217,7 +220,7 @@
     dataptr[DCTSIZE*1] = z11 + z4;
     dataptr[DCTSIZE*7] = z11 - z4;
 
-    dataptr++;			/* advance pointer to next column */
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctint.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctint.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jfdctint.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jfdctint.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jfdctint.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a slow-but-accurate integer implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -26,7 +29,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_ISLOW_SUPPORTED
 
@@ -67,7 +70,7 @@
  * they are represented to better-than-integral precision.  These outputs
  * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
  * with the recommended scaling.  (For 12-bit sample data, the intermediate
- * array is INT32 anyway.)
+ * array is JLONG anyway.)
  *
  * To avoid overflow of the 32-bit intermediate results in pass 2, we must
  * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
@@ -79,7 +82,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -90,18 +93,18 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
-#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
-#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
-#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
-#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#define FIX_0_298631336  ((JLONG)  2446)        /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)  3196)        /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)  4433)        /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)  9633)        /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)  12299)       /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)  16069)       /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)  16819)       /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)  25172)       /* FIX(3.072711026) */
 #else
 #define FIX_0_298631336  FIX(0.298631336)
 #define FIX_0_390180644  FIX(0.390180644)
@@ -118,7 +121,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -137,11 +140,11 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_islow (DCTELEM * data)
+jpeg_fdct_islow (DCTELEM *data)
 {
-  INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  INT32 tmp10, tmp11, tmp12, tmp13;
-  INT32 z1, z2, z3, z4, z5;
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
@@ -160,36 +163,36 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part per LL&M figure 1 --- note that published figure is faulty;
      * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
      */
-    
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
-    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
-    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
-    
+
+    dataptr[0] = (DCTELEM) LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS);
+    dataptr[4] = (DCTELEM) LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS);
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
-				   CONST_BITS-PASS1_BITS);
+                                   CONST_BITS-PASS1_BITS);
     dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
-				   CONST_BITS-PASS1_BITS);
-    
+                                   CONST_BITS-PASS1_BITS);
+
     /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
      * cK represents cos(K*pi/16).
      * i0..i3 in the paper are tmp4..tmp7 here.
      */
-    
+
     z1 = tmp4 + tmp7;
     z2 = tmp5 + tmp6;
     z3 = tmp4 + tmp6;
     z4 = tmp5 + tmp7;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -198,16 +201,16 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
     dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
     dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
     dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
-    
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns.
@@ -225,36 +228,36 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part per LL&M figure 1 --- note that published figure is faulty;
      * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
      */
-    
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
     dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
-					   CONST_BITS+PASS1_BITS);
-    
+                                           CONST_BITS+PASS1_BITS);
+
     /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
      * cK represents cos(K*pi/16).
      * i0..i3 in the paper are tmp4..tmp7 here.
      */
-    
+
     z1 = tmp4 + tmp7;
     z2 = tmp5 + tmp6;
     z3 = tmp4 + tmp6;
     z4 = tmp5 + tmp7;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -263,20 +266,20 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
-					   CONST_BITS+PASS1_BITS);
-    
-    dataptr++;			/* advance pointer to next column */
+                                           CONST_BITS+PASS1_BITS);
+
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctflt.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctflt.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctflt.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctflt.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
  * jidctflt.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a floating-point implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -22,8 +26,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -39,7 +43,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_FLOAT_SUPPORTED
 
@@ -65,21 +69,21 @@
  */
 
 GLOBAL(void)
-jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
   FAST_FLOAT z5, z10, z11, z12, z13;
   JCOEFPTR inptr;
-  FLOAT_MULT_TYPE * quantptr;
-  FAST_FLOAT * wsptr;
+  FLOAT_MULT_TYPE *quantptr;
+  FAST_FLOAT *wsptr;
   JSAMPROW outptr;
-  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
   int ctr;
   FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */
-  SHIFT_TEMPS
+  #define _0_125 ((FLOAT_MULT_TYPE)0.125)
 
   /* Pass 1: process columns from input, store into work array. */
 
@@ -95,14 +99,15 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
-      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-      
+      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0],
+                                    quantptr[DCTSIZE*0] * _0_125);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
@@ -111,53 +116,53 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part */
 
-    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
-    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
-    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0] * _0_125);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2] * _0_125);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4] * _0_125);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6] * _0_125);
 
-    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
     tmp11 = tmp0 - tmp2;
 
-    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
     tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */
 
-    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
     tmp3 = tmp10 - tmp13;
     tmp1 = tmp11 + tmp12;
     tmp2 = tmp11 - tmp12;
-    
+
     /* Odd part */
 
-    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
-    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
-    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1] * _0_125);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3] * _0_125);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5] * _0_125);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7] * _0_125);
 
-    z13 = tmp6 + tmp5;		/* phase 6 */
+    z13 = tmp6 + tmp5;          /* phase 6 */
     z10 = tmp6 - tmp5;
     z11 = tmp4 + tmp7;
     z12 = tmp4 - tmp7;
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */
 
     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
-    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
-    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+    tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
-    tmp4 = tmp10 + tmp5;
+    tmp4 = tmp10 - tmp5;
 
     wsptr[DCTSIZE*0] = tmp0 + tmp7;
     wsptr[DCTSIZE*7] = tmp0 - tmp7;
@@ -165,16 +170,15 @@
     wsptr[DCTSIZE*6] = tmp1 - tmp6;
     wsptr[DCTSIZE*2] = tmp2 + tmp5;
     wsptr[DCTSIZE*5] = tmp2 - tmp5;
-    wsptr[DCTSIZE*4] = tmp3 + tmp4;
-    wsptr[DCTSIZE*3] = tmp3 - tmp4;
+    wsptr[DCTSIZE*3] = tmp3 + tmp4;
+    wsptr[DCTSIZE*4] = tmp3 - tmp4;
 
-    inptr++;			/* advance pointers to next column */
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
-  /* Note that we must descale the results by a factor of 8 == 2**3. */
 
   wsptr = workspace;
   for (ctr = 0; ctr < DCTSIZE; ctr++) {
@@ -184,11 +188,13 @@
      * the simplification applies less often (typically 5% to 10% of the time).
      * And testing floats for zero is relatively expensive, so we don't bother.
      */
-    
+
     /* Even part */
 
-    tmp10 = wsptr[0] + wsptr[4];
-    tmp11 = wsptr[0] - wsptr[4];
+    /* Apply signed->unsigned and prepare float->int conversion */
+    z5 = wsptr[0] + ((FAST_FLOAT) CENTERJSAMPLE + (FAST_FLOAT) 0.5);
+    tmp10 = z5 + wsptr[4];
+    tmp11 = z5 - wsptr[4];
 
     tmp13 = wsptr[2] + wsptr[6];
     tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13;
@@ -209,33 +215,25 @@
     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562);
 
     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
-    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
-    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+    tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */
 
     tmp6 = tmp12 - tmp7;
     tmp5 = tmp11 - tmp6;
-    tmp4 = tmp10 + tmp5;
+    tmp4 = tmp10 - tmp5;
+
+    /* Final output stage: float->int conversion and range-limit */
 
-    /* Final output stage: scale down by a factor of 8 and range-limit */
+    outptr[0] = range_limit[((int) (tmp0 + tmp7)) & RANGE_MASK];
+    outptr[7] = range_limit[((int) (tmp0 - tmp7)) & RANGE_MASK];
+    outptr[1] = range_limit[((int) (tmp1 + tmp6)) & RANGE_MASK];
+    outptr[6] = range_limit[((int) (tmp1 - tmp6)) & RANGE_MASK];
+    outptr[2] = range_limit[((int) (tmp2 + tmp5)) & RANGE_MASK];
+    outptr[5] = range_limit[((int) (tmp2 - tmp5)) & RANGE_MASK];
+    outptr[3] = range_limit[((int) (tmp3 + tmp4)) & RANGE_MASK];
+    outptr[4] = range_limit[((int) (tmp3 - tmp4)) & RANGE_MASK];
 
-    outptr[0] = range_limit[(int) DESCALE((INT32) (tmp0 + tmp7), 3)
-			    & RANGE_MASK];
-    outptr[7] = range_limit[(int) DESCALE((INT32) (tmp0 - tmp7), 3)
-			    & RANGE_MASK];
-    outptr[1] = range_limit[(int) DESCALE((INT32) (tmp1 + tmp6), 3)
-			    & RANGE_MASK];
-    outptr[6] = range_limit[(int) DESCALE((INT32) (tmp1 - tmp6), 3)
-			    & RANGE_MASK];
-    outptr[2] = range_limit[(int) DESCALE((INT32) (tmp2 + tmp5), 3)
-			    & RANGE_MASK];
-    outptr[5] = range_limit[(int) DESCALE((INT32) (tmp2 - tmp5), 3)
-			    & RANGE_MASK];
-    outptr[4] = range_limit[(int) DESCALE((INT32) (tmp3 + tmp4), 3)
-			    & RANGE_MASK];
-    outptr[3] = range_limit[(int) DESCALE((INT32) (tmp3 - tmp4), 3)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctfst.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctfst.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctfst.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctfst.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jidctfst.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a fast, not so accurate integer implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -17,8 +20,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -35,7 +38,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_IFAST_SUPPORTED
 
@@ -78,7 +81,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  8
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -89,10 +92,10 @@
  */
 
 #if CONST_BITS == 8
-#define FIX_1_082392200  ((INT32)  277)		/* FIX(1.082392200) */
-#define FIX_1_414213562  ((INT32)  362)		/* FIX(1.414213562) */
-#define FIX_1_847759065  ((INT32)  473)		/* FIX(1.847759065) */
-#define FIX_2_613125930  ((INT32)  669)		/* FIX(2.613125930) */
+#define FIX_1_082392200  ((JLONG)  277)         /* FIX(1.082392200) */
+#define FIX_1_414213562  ((JLONG)  362)         /* FIX(1.414213562) */
+#define FIX_1_847759065  ((JLONG)  473)         /* FIX(1.847759065) */
+#define FIX_2_613125930  ((JLONG)  669)         /* FIX(2.613125930) */
 #else
 #define FIX_1_082392200  FIX(1.082392200)
 #define FIX_1_414213562  FIX(1.414213562)
@@ -112,7 +115,7 @@
 #endif
 
 
-/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
  * descale to yield a DCTELEM result.
  */
 
@@ -122,27 +125,27 @@
 /* Dequantize a coefficient by multiplying it by the multiplier-table
  * entry; produce a DCTELEM result.  For 8-bit data a 16x16->16
  * multiplication will do.  For 12-bit data, the multiplier table is
- * declared INT32, so a 32-bit multiply will be used.
+ * declared JLONG, so a 32-bit multiply will be used.
  */
 
 #if BITS_IN_JSAMPLE == 8
 #define DEQUANTIZE(coef,quantval)  (((IFAST_MULT_TYPE) (coef)) * (quantval))
 #else
 #define DEQUANTIZE(coef,quantval)  \
-	DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
+        DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
 #endif
 
 
 /* Like DESCALE, but applies to a DCTELEM and produces an int.
- * We assume that int right shift is unsigned if INT32 right shift is.
+ * We assume that int right shift is unsigned if JLONG right shift is.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	DCTELEM ishift_temp;
+#define ISHIFT_TEMPS    DCTELEM ishift_temp;
 #if BITS_IN_JSAMPLE == 8
-#define DCTELEMBITS  16		/* DCTELEM may be 16 or 32 bits */
+#define DCTELEMBITS  16         /* DCTELEM may be 16 or 32 bits */
 #else
-#define DCTELEMBITS  32		/* DCTELEM must be 32 bits */
+#define DCTELEMBITS  32         /* DCTELEM must be 32 bits */
 #endif
 #define IRIGHT_SHIFT(x,shft)  \
     ((ishift_temp = (x)) < 0 ? \
@@ -150,7 +153,7 @@
      (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 #ifdef USE_ACCURATE_ROUNDING
@@ -165,22 +168,22 @@
  */
 
 GLOBAL(void)
-jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   DCTELEM tmp10, tmp11, tmp12, tmp13;
   DCTELEM z5, z10, z11, z12, z13;
   JCOEFPTR inptr;
-  IFAST_MULT_TYPE * quantptr;
-  int * wsptr;
+  IFAST_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE2];	/* buffers data between passes */
-  SHIFT_TEMPS			/* for DESCALE */
-  ISHIFT_TEMPS			/* for IDESCALE */
+  int workspace[DCTSIZE2];      /* buffers data between passes */
+  SHIFT_TEMPS                   /* for DESCALE */
+  ISHIFT_TEMPS                  /* for IDESCALE */
 
   /* Pass 1: process columns from input, store into work array. */
 
@@ -196,11 +199,11 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
       int dcval = (int) DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
 
@@ -212,13 +215,13 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part */
 
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
@@ -226,17 +229,17 @@
     tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
     tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
 
-    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
     tmp11 = tmp0 - tmp2;
 
-    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
     tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */
 
-    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
     tmp3 = tmp10 - tmp13;
     tmp1 = tmp11 + tmp12;
     tmp2 = tmp11 - tmp12;
-    
+
     /* Odd part */
 
     tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
@@ -244,19 +247,19 @@
     tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
 
-    z13 = tmp6 + tmp5;		/* phase 6 */
+    z13 = tmp6 + tmp5;          /* phase 6 */
     z10 = tmp6 - tmp5;
     z11 = tmp4 + tmp7;
     z12 = tmp4 - tmp7;
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
 
     z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
     tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
     tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
     tmp4 = tmp10 + tmp5;
 
@@ -269,11 +272,11 @@
     wsptr[DCTSIZE*4] = (int) (tmp3 + tmp4);
     wsptr[DCTSIZE*3] = (int) (tmp3 - tmp4);
 
-    inptr++;			/* advance pointers to next column */
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
   /* Note that we must descale the results by a factor of 8 == 2**3, */
   /* and also undo the PASS1_BITS scaling. */
@@ -288,14 +291,14 @@
      * test takes more time than it's worth.  In that case this section
      * may be commented out.
      */
-    
+
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
       JSAMPLE dcval = range_limit[IDESCALE(wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
@@ -305,11 +308,11 @@
       outptr[6] = dcval;
       outptr[7] = dcval;
 
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
 
     tmp10 = ((DCTELEM) wsptr[0] + (DCTELEM) wsptr[4]);
@@ -317,7 +320,7 @@
 
     tmp13 = ((DCTELEM) wsptr[2] + (DCTELEM) wsptr[6]);
     tmp12 = MULTIPLY((DCTELEM) wsptr[2] - (DCTELEM) wsptr[6], FIX_1_414213562)
-	    - tmp13;
+            - tmp13;
 
     tmp0 = tmp10 + tmp13;
     tmp3 = tmp10 - tmp13;
@@ -331,37 +334,37 @@
     z11 = (DCTELEM) wsptr[1] + (DCTELEM) wsptr[7];
     z12 = (DCTELEM) wsptr[1] - (DCTELEM) wsptr[7];
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
 
     z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
     tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
     tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
     tmp4 = tmp10 + tmp5;
 
     /* Final output stage: scale down by a factor of 8 and range-limit */
 
     outptr[0] = range_limit[IDESCALE(tmp0 + tmp7, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[7] = range_limit[IDESCALE(tmp0 - tmp7, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[1] = range_limit[IDESCALE(tmp1 + tmp6, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[6] = range_limit[IDESCALE(tmp1 - tmp6, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[2] = range_limit[IDESCALE(tmp2 + tmp5, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[5] = range_limit[IDESCALE(tmp2 - tmp5, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[4] = range_limit[IDESCALE(tmp3 + tmp4, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[3] = range_limit[IDESCALE(tmp3 - tmp4, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
 
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctint.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctint.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctint.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctint.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
  * jidctint.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modification developed 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a slow-but-accurate integer implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -23,12 +27,33 @@
  * The advantage of this method is that no data path contains more than one
  * multiplication; this allows a very simple and accurate implementation in
  * scaled fixed-point arithmetic, with a minimal number of shifts.
+ *
+ * We also provide IDCT routines with various output sample block sizes for
+ * direct resolution reduction or enlargement without additional resampling:
+ * NxN (N=1...16) pixels for one 8x8 input DCT block.
+ *
+ * For N<8 we simply take the corresponding low-frequency coefficients of
+ * the 8x8 input DCT block and apply an NxN point IDCT on the sub-block
+ * to yield the downscaled outputs.
+ * This can be seen as direct low-pass downsampling from the DCT domain
+ * point of view rather than the usual spatial domain point of view,
+ * yielding significant computational savings and results at least
+ * as good as common bilinear (averaging) spatial downsampling.
+ *
+ * For N>8 we apply a partial NxN IDCT on the 8 input coefficients as
+ * lower frequencies and higher frequencies assumed to be zero.
+ * It turns out that the computational effort is similar to the 8x8 IDCT
+ * regarding the output size.
+ * Furthermore, the scaling and descaling is the same for all IDCT sizes.
+ *
+ * CAUTION: We rely on the FIX() macro except for the N=1,2,4,8 cases
+ * since there would be too many additional constants to pre-calculate.
  */
 
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_ISLOW_SUPPORTED
 
@@ -38,7 +63,7 @@
  */
 
 #if DCTSIZE != 8
-  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+  Sorry, this code only copes with 8x8 DCT blocks. /* deliberate syntax err */
 #endif
 
 
@@ -67,7 +92,7 @@
  * they are represented to better-than-integral precision.  These outputs
  * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
  * with the recommended scaling.  (To scale up 12-bit sample data further, an
- * intermediate INT32 array would be needed.)
+ * intermediate JLONG array would be needed.)
  *
  * To avoid overflow of the 32-bit intermediate results in pass 2, we must
  * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
@@ -79,7 +104,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -90,18 +115,18 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
-#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
-#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
-#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
-#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#define FIX_0_298631336  ((JLONG)  2446)        /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)  3196)        /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)  4433)        /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)  9633)        /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)  12299)       /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)  16069)       /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)  16819)       /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)  25172)       /* FIX(3.072711026) */
 #else
 #define FIX_0_298631336  FIX(0.298631336)
 #define FIX_0_390180644  FIX(0.390180644)
@@ -118,7 +143,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -145,20 +170,20 @@
  */
 
 GLOBAL(void)
-jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp1, tmp2, tmp3;
-  INT32 tmp10, tmp11, tmp12, tmp13;
-  INT32 z1, z2, z3, z4, z5;
+  JLONG tmp0, tmp1, tmp2, tmp3;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE2];	/* buffers data between passes */
+  int workspace[DCTSIZE2];      /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -177,14 +202,15 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
@@ -193,49 +219,49 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part: reverse the even part of the forward DCT. */
     /* The rotator is sqrt(2)*c(-6). */
-    
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
-    
+
     z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
     tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
     tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
-    
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
 
-    tmp0 = (z2 + z3) << CONST_BITS;
-    tmp1 = (z2 - z3) << CONST_BITS;
-    
+    tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS);
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     /* Odd part per figure 8; the matrix is unitary and hence its
      * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
      */
-    
+
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
     tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
     tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    
+
     z1 = tmp0 + tmp3;
     z2 = tmp1 + tmp2;
     z3 = tmp0 + tmp2;
     z4 = tmp1 + tmp3;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -244,17 +270,17 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     tmp0 += z1 + z3;
     tmp1 += z2 + z4;
     tmp2 += z2 + z3;
     tmp3 += z1 + z4;
-    
+
     /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
@@ -263,12 +289,12 @@
     wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
-    
-    inptr++;			/* advance pointers to next column */
+
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
   /* Note that we must descale the results by a factor of 8 == 2**3, */
   /* and also undo the PASS1_BITS scaling. */
@@ -283,14 +309,14 @@
      * test takes more time than it's worth.  In that case this section
      * may be commented out.
      */
-    
+
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
@@ -300,44 +326,44 @@
       outptr[6] = dcval;
       outptr[7] = dcval;
 
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part: reverse the even part of the forward DCT. */
     /* The rotator is sqrt(2)*c(-6). */
-    
-    z2 = (INT32) wsptr[2];
-    z3 = (INT32) wsptr[6];
-    
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[6];
+
     z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
     tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
     tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
-    
-    tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS;
-    tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS;
-    
+
+    tmp0 = LEFT_SHIFT((JLONG) wsptr[0] + (JLONG) wsptr[4], CONST_BITS);
+    tmp1 = LEFT_SHIFT((JLONG) wsptr[0] - (JLONG) wsptr[4], CONST_BITS);
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     /* Odd part per figure 8; the matrix is unitary and hence its
      * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
      */
-    
-    tmp0 = (INT32) wsptr[7];
-    tmp1 = (INT32) wsptr[5];
-    tmp2 = (INT32) wsptr[3];
-    tmp3 = (INT32) wsptr[1];
-    
+
+    tmp0 = (JLONG) wsptr[7];
+    tmp1 = (JLONG) wsptr[5];
+    tmp2 = (JLONG) wsptr[3];
+    tmp3 = (JLONG) wsptr[1];
+
     z1 = tmp0 + tmp3;
     z2 = tmp1 + tmp2;
     z3 = tmp0 + tmp2;
     z4 = tmp1 + tmp3;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -346,44 +372,2256 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     tmp0 += z1 + z3;
     tmp1 += z2 + z4;
     tmp2 += z2 + z3;
     tmp3 += z1 + z4;
-    
+
     /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 7x7 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/14).
+ */
+
+GLOBAL(void)
+jpeg_idct_7x7 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[7*7];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp13 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp13 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276));    /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    wsptr[7*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[7*6] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[7*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[7*5] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[7*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[7*4] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[7*3] = (int) RIGHT_SHIFT(tmp13, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 7 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp13 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276));    /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 7;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 6x6 output block.
+ *
+ * Optimized algorithm with 3 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/12).
+ */
+
+GLOBAL(void)
+jpeg_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[6*6];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = RIGHT_SHIFT(tmp0 - tmp10 - tmp10, CONST_BITS-PASS1_BITS);
+    tmp10 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr[6*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[6*5] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[6*1] = (int) (tmp11 + tmp1);
+    wsptr[6*4] = (int) (tmp11 - tmp1);
+    wsptr[6*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[6*3] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 6 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG) wsptr[4];
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = tmp0 - tmp10 - tmp10;
+    tmp10 = (JLONG) wsptr[2];
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, CONST_BITS);
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 6;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 5x5 output block.
+ *
+ * Optimized algorithm with 5 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/10).
+ */
+
+GLOBAL(void)
+jpeg_idct_5x5 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[5*5];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp12 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp12 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    wsptr[5*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[5*4] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[5*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[5*3] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[5*2] = (int) RIGHT_SHIFT(tmp12, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 5 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp12 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    tmp0 = (JLONG) wsptr[2];
+    tmp1 = (JLONG) wsptr[4];
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = (JLONG) wsptr[1];
+    z3 = (JLONG) wsptr[3];
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 5;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 3x3 output block.
+ *
+ * Optimized algorithm with 2 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/6).
+ */
+
+GLOBAL(void)
+jpeg_idct_3x3 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[3*3];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    wsptr[3*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[3*2] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[3*1] = (int) RIGHT_SHIFT(tmp2, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 3 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG) wsptr[2];
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = (JLONG) wsptr[1];
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 3;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 9x9 output block.
+ *
+ * Optimized algorithm with 10 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/18).
+ */
+
+GLOBAL(void)
+jpeg_idct_9x9 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*9];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    wsptr[8*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[8*8] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[8*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[8*7] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[8*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[8*6] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[8*3] = (int) RIGHT_SHIFT(tmp13 + tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[8*5] = (int) RIGHT_SHIFT(tmp13 - tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[8*4] = (int) RIGHT_SHIFT(tmp14, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 9 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 9; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13 + tmp3,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp13 - tmp3,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 10x10 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/20).
+ */
+
+GLOBAL(void)
+jpeg_idct_10x10 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24;
+  JLONG z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*10];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS-PASS1_BITS-1);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = RIGHT_SHIFT(z3 - LEFT_SHIFT(z1 - z2, 1),
+                        CONST_BITS-PASS1_BITS);  /* c0 = (c4-c8)*2 */
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+    z5 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z5 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z5 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13 - z3, PASS1_BITS);
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*9] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*8] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2] = (int) (tmp22 + tmp12);
+    wsptr[8*7] = (int) (tmp22 - tmp12);
+    wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*6] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 10 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 10; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG) wsptr[4];
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = z3 - LEFT_SHIFT(z1 - z2, 1);         /* c0 = (c4-c8)*2 */
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[6];
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z3 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z3 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13, CONST_BITS) - z3;
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 11x11 output block.
+ *
+ * Optimized algorithm with 24 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/22).
+ */
+
+GLOBAL(void)
+jpeg_idct_11x11 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*11];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp10 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp10 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, - FIX(1.155664402));        /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, - FIX(1.798248910));       /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, - FIX(1.467221301)) +          /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 11 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 11; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp10 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, - FIX(1.155664402));        /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, - FIX(1.798248910));       /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, - FIX(1.467221301)) +          /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 12x12 output block.
+ *
+ * Optimized algorithm with 15 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/24).
+ */
+
+GLOBAL(void)
+jpeg_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*12];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, - FIX_0_541196100);                 /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580));           /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 12 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 12; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z4 = (JLONG) wsptr[4];
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = (JLONG) wsptr[2];
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = (JLONG) wsptr[6];
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, - FIX_0_541196100);                 /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580));           /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 13x13 output block.
+ *
+ * Optimized algorithm with 29 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/26).
+ */
+
+GLOBAL(void)
+jpeg_idct_13x13 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*13];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */
+    tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458));   /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945));   /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813));   /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 13 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 13; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[4];
+    z4 = (JLONG) wsptr[6];
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */
+    tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458));   /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945));   /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813));   /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 14x14 output block.
+ *
+ * Optimized algorithm with 20 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/28).
+ */
+
+GLOBAL(void)
+jpeg_idct_14x14 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*14];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = RIGHT_SHIFT(z1 - LEFT_SHIFT(z2 + z3 - z4, 1),
+                        CONST_BITS-PASS1_BITS);  /* c0 = (c4+c12-c8)*2 */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp13 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + tmp13 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - tmp13;        /* c11 */
+    tmp16 += tmp15;
+    z1    += z4;
+    z4    = MULTIPLY(z2 + z3, - FIX(0.158341681)) - tmp13; /* -c13 */
+    tmp11 += z4 - MULTIPLY(z2, FIX(0.424103948));          /* c3-c9-c13 */
+    tmp12 += z4 - MULTIPLY(z3, FIX(2.373959773));          /* c3+c5-c13 */
+    z4    = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += z4 + tmp13 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += z4 + MULTIPLY(z2, FIX(0.674957567));          /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) (tmp23 + tmp13);
+    wsptr[8*10] = (int) (tmp23 - tmp13);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 14 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 14; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z4 = (JLONG) wsptr[4];
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = z1 - LEFT_SHIFT(z2 + z3 - z4, 1);    /* c0 = (c4+c12-c8)*2 */
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[6];
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+    z4 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + z4 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - z4;           /* c11 */
+    tmp16 += tmp15;
+    tmp13 = MULTIPLY(z2 + z3, - FIX(0.158341681)) - z4;    /* -c13 */
+    tmp11 += tmp13 - MULTIPLY(z2, FIX(0.424103948));       /* c3-c9-c13 */
+    tmp12 += tmp13 - MULTIPLY(z3, FIX(2.373959773));       /* c3+c5-c13 */
+    tmp13 = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += tmp13 + z4 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += tmp13 + MULTIPLY(z2, FIX(0.674957567));       /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, CONST_BITS) + z4;
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 15x15 output block.
+ *
+ * Optimized algorithm with 22 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/30).
+ */
+
+GLOBAL(void)
+jpeg_idct_15x15 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*15];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, - FIX(0.831253876));               /* -c9 */
+    tmp15 = MULTIPLY(z2, - FIX(1.344997024));               /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*14] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp27, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 15 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 15; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[4];
+    z4 = (JLONG) wsptr[6];
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z4 = (JLONG) wsptr[5];
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = (JLONG) wsptr[7];
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, - FIX(0.831253876));               /* -c9 */
+    tmp15 = MULTIPLY(z2, - FIX(1.344997024));               /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp27,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 16x16 output block.
+ *
+ * Optimized algorithm with 28 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/32).
+ */
+
+GLOBAL(void)
+jpeg_idct_16x16 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*16];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += 1 << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c12[16] = c6[8] */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c14[16] = c7[8] */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c2[16] = c1[8] */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)[16] = (c3+c1)[8] */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)[16] = (c3-c7)[8] */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, - FIX(0.666655658));      /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, - FIX(1.247225013));      /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp0,  CONST_BITS-PASS1_BITS);
+    wsptr[8*15] = (int) RIGHT_SHIFT(tmp20 - tmp0,  CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp1,  CONST_BITS-PASS1_BITS);
+    wsptr[8*14] = (int) RIGHT_SHIFT(tmp21 - tmp1,  CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp2,  CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp22 - tmp2,  CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp3,  CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp23 - tmp3,  CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp24 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp25 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp26 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp27 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp27 - tmp13, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 16 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 16; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG) wsptr[4];
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c12[16] = c6[8] */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[6];
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c14[16] = c7[8] */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c2[16] = c1[8] */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)[16] = (c3+c1)[8] */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)[16] = (c3-c7)[8] */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, - FIX(0.666655658));      /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, - FIX(1.247225013));      /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp0,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[15] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp0,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp1,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp1,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp2,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp2,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp3,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp3,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp27 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp27 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
   }
 }
 
+#endif /* IDCT_SCALING_SUPPORTED */
 #endif /* DCT_ISLOW_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctred.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctred.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jidctred.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jidctred.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jidctred.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains inverse-DCT routines that produce reduced-size output:
  * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block.
@@ -23,7 +26,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef IDCT_SCALING_SUPPORTED
 
@@ -44,7 +47,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -55,20 +58,20 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_211164243  ((INT32)  1730)	/* FIX(0.211164243) */
-#define FIX_0_509795579  ((INT32)  4176)	/* FIX(0.509795579) */
-#define FIX_0_601344887  ((INT32)  4926)	/* FIX(0.601344887) */
-#define FIX_0_720959822  ((INT32)  5906)	/* FIX(0.720959822) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_850430095  ((INT32)  6967)	/* FIX(0.850430095) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_061594337  ((INT32)  8697)	/* FIX(1.061594337) */
-#define FIX_1_272758580  ((INT32)  10426)	/* FIX(1.272758580) */
-#define FIX_1_451774981  ((INT32)  11893)	/* FIX(1.451774981) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_2_172734803  ((INT32)  17799)	/* FIX(2.172734803) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_624509785  ((INT32)  29692)	/* FIX(3.624509785) */
+#define FIX_0_211164243  ((JLONG)  1730)        /* FIX(0.211164243) */
+#define FIX_0_509795579  ((JLONG)  4176)        /* FIX(0.509795579) */
+#define FIX_0_601344887  ((JLONG)  4926)        /* FIX(0.601344887) */
+#define FIX_0_720959822  ((JLONG)  5906)        /* FIX(0.720959822) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_850430095  ((JLONG)  6967)        /* FIX(0.850430095) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_061594337  ((JLONG)  8697)        /* FIX(1.061594337) */
+#define FIX_1_272758580  ((JLONG)  10426)       /* FIX(1.272758580) */
+#define FIX_1_451774981  ((JLONG)  11893)       /* FIX(1.451774981) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_2_172734803  ((JLONG)  17799)       /* FIX(2.172734803) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_624509785  ((JLONG)  29692)       /* FIX(3.624509785) */
 #else
 #define FIX_0_211164243  FIX(0.211164243)
 #define FIX_0_509795579  FIX(0.509795579)
@@ -87,7 +90,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -115,19 +118,19 @@
  */
 
 GLOBAL(void)
-jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp2, tmp10, tmp12;
-  INT32 z1, z2, z3, z4;
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JLONG z1, z2, z3, z4;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE*4];	/* buffers data between passes */
+  int workspace[DCTSIZE*4];     /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -140,57 +143,58 @@
     if (ctr == DCTSIZE-4)
       continue;
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
-	inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
+        inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero; we need not examine term 4 for 4x4 output */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
       wsptr[DCTSIZE*3] = dcval;
-      
+
       continue;
     }
-    
+
     /* Even part */
-    
+
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp0 <<= (CONST_BITS+1);
-    
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS+1);
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
 
     tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, - FIX_0_765366865);
-    
+
     tmp10 = tmp0 + tmp2;
     tmp12 = tmp0 - tmp2;
-    
+
     /* Odd part */
-    
+
     z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
     z2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
     z4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    
+
     tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
-	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
-	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
-	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
-    
+         + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+         + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+         + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+
     tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
-	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
-	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
-	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+         + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+         + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+         + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
 
     /* Final output stage */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp2, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*3] = (int) DESCALE(tmp10 - tmp2, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp12 + tmp0, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 - tmp0, CONST_BITS-PASS1_BITS+1);
   }
-  
+
   /* Pass 2: process 4 rows from work array, store into output array. */
 
   wsptr = workspace;
@@ -200,64 +204,64 @@
 
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
       outptr[3] = dcval;
-      
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
-    
-    tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1);
-    
-    tmp2 = MULTIPLY((INT32) wsptr[2], FIX_1_847759065)
-	 + MULTIPLY((INT32) wsptr[6], - FIX_0_765366865);
-    
+
+    tmp0 = LEFT_SHIFT((JLONG) wsptr[0], CONST_BITS+1);
+
+    tmp2 = MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+         + MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865);
+
     tmp10 = tmp0 + tmp2;
     tmp12 = tmp0 - tmp2;
-    
+
     /* Odd part */
-    
-    z1 = (INT32) wsptr[7];
-    z2 = (INT32) wsptr[5];
-    z3 = (INT32) wsptr[3];
-    z4 = (INT32) wsptr[1];
-    
+
+    z1 = (JLONG) wsptr[7];
+    z2 = (JLONG) wsptr[5];
+    z3 = (JLONG) wsptr[3];
+    z4 = (JLONG) wsptr[1];
+
     tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
-	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
-	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
-	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
-    
+         + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+         + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+         + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+
     tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
-	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
-	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
-	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+         + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+         + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+         + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
 
     /* Final output stage */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp2,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[3] = range_limit[(int) DESCALE(tmp10 - tmp2,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp12 + tmp0,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[2] = range_limit[(int) DESCALE(tmp12 - tmp0,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
@@ -268,18 +272,18 @@
  */
 
 GLOBAL(void)
-jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp10, z1;
+  JLONG tmp0, tmp10, z1;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE*2];	/* buffers data between passes */
+  int workspace[DCTSIZE*2];     /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -292,21 +296,22 @@
     if (ctr == DCTSIZE-2 || ctr == DCTSIZE-4 || ctr == DCTSIZE-6)
       continue;
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*3] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
-      
+
       continue;
     }
-    
+
     /* Even part */
-    
+
     z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp10 = z1 << (CONST_BITS+2);
-    
+    tmp10 = LEFT_SHIFT(z1, CONST_BITS+2);
+
     /* Odd part */
 
     z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
@@ -319,11 +324,11 @@
     tmp0 += MULTIPLY(z1, FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
 
     /* Final output stage */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp0, CONST_BITS-PASS1_BITS+2);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp10 - tmp0, CONST_BITS-PASS1_BITS+2);
   }
-  
+
   /* Pass 2: process 2 rows from work array, store into output array. */
 
   wsptr = workspace;
@@ -334,38 +339,38 @@
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[3] == 0 && wsptr[5] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
-      
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
-    
-    tmp10 = ((INT32) wsptr[0]) << (CONST_BITS+2);
-    
+
+    tmp10 = LEFT_SHIFT((JLONG) wsptr[0], CONST_BITS+2);
+
     /* Odd part */
 
-    tmp0 = MULTIPLY((INT32) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
-	 + MULTIPLY((INT32) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
-	 + MULTIPLY((INT32) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
-	 + MULTIPLY((INT32) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+    tmp0 = MULTIPLY((JLONG) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
+         + MULTIPLY((JLONG) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
+         + MULTIPLY((JLONG) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
+         + MULTIPLY((JLONG) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
 
     /* Final output stage */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp0,
-					  CONST_BITS+PASS1_BITS+3+2)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+2)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp10 - tmp0,
-					  CONST_BITS+PASS1_BITS+3+2)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3+2)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
@@ -376,12 +381,12 @@
  */
 
 GLOBAL(void)
-jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   int dcval;
-  ISLOW_MULT_TYPE * quantptr;
+  ISLOW_MULT_TYPE *quantptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   SHIFT_TEMPS
 
@@ -390,7 +395,7 @@
    */
   quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
   dcval = DEQUANTIZE(coef_block[0], quantptr[0]);
-  dcval = (int) DESCALE((INT32) dcval, 3);
+  dcval = (int) DESCALE((JLONG) dcval, 3);
 
   output_buf[0][output_col] = range_limit[dcval & RANGE_MASK];
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jinclude.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jinclude.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jinclude.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jinclude.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jinclude.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1994, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file exists to provide a single place to fix any problems with
  * including the wrong system include files.  (Common problems are taken
@@ -17,8 +20,8 @@
 
 /* Include auto-config file to find out which system include files we need. */
 
-#include "jconfig.h"		/* auto configuration options */
-#define JCONFIG_INCLUDED	/* so that jpeglib.h doesn't do it again */
+#include "jconfig.h"            /* auto configuration options */
+#define JCONFIG_INCLUDED        /* so that jpeglib.h doesn't do it again */
 
 /*
  * We need the NULL macro and size_t typedef.
@@ -58,28 +61,18 @@
 #ifdef NEED_BSD_STRINGS
 
 #include <strings.h>
-#define MEMZERO(target,size)	bzero((void *)(target), (size_t)(size))
-#define MEMCOPY(dest,src,size)	bcopy((const void *)(src), (void *)(dest), (size_t)(size))
+#define MEMZERO(target,size)    bzero((void *)(target), (size_t)(size))
+#define MEMCOPY(dest,src,size)  bcopy((const void *)(src), (void *)(dest), (size_t)(size))
 
 #else /* not BSD, assume ANSI/SysV string lib */
 
 #include <string.h>
-#define MEMZERO(target,size)	memset((void *)(target), 0, (size_t)(size))
-#define MEMCOPY(dest,src,size)	memcpy((void *)(dest), (const void *)(src), (size_t)(size))
+#define MEMZERO(target,size)    memset((void *)(target), 0, (size_t)(size))
+#define MEMCOPY(dest,src,size)  memcpy((void *)(dest), (const void *)(src), (size_t)(size))
 
 #endif
 
 /*
- * In ANSI C, and indeed any rational implementation, size_t is also the
- * type returned by sizeof().  However, it seems there are some irrational
- * implementations out there, in which sizeof() returns an int even though
- * size_t is defined as long or unsigned long.  To ensure consistent results
- * we always use this SIZEOF() macro in place of using sizeof() directly.
- */
-
-#define SIZEOF(object)	((size_t) sizeof(object))
-
-/*
  * The modules that use fread() and fwrite() always invoke them through
  * these macros.  On some systems you may need to twiddle the argument casts.
  * CAUTION: argument order is different from underlying functions!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemmgr.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the JPEG system-independent memory management
  * routines.  This code is usable across a wide variety of machines; most
@@ -25,14 +28,14 @@
  */
 
 #define JPEG_INTERNALS
-#define AM_MEMORY_MANAGER	/* we define jvirt_Xarray_control structs */
+#define AM_MEMORY_MANAGER       /* we define jvirt_Xarray_control structs */
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jmemsys.h"		/* import the system-dependent declarations */
+#include "jmemsys.h"            /* import the system-dependent declarations */
 
 #ifndef NO_GETENV
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare getenv() */
-extern char * getenv JPP((const char * name));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare getenv() */
+extern char *getenv (const char *name);
 #endif
 #endif
 
@@ -67,9 +70,9 @@
  * There isn't any really portable way to determine the worst-case alignment
  * requirement.  This module assumes that the alignment requirement is
  * multiples of ALIGN_SIZE.
- * By default, we define ALIGN_SIZE as sizeof(double).  This is necessary on some
- * workstations (where doubles really do need 8-byte alignment) and will work
- * fine on nearly everything.  If your machine has lesser alignment needs,
+ * By default, we define ALIGN_SIZE as sizeof(double).  This is necessary on
+ * some workstations (where doubles really do need 8-byte alignment) and will
+ * work fine on nearly everything.  If your machine has lesser alignment needs,
  * you can save a few bytes by making ALIGN_SIZE smaller.
  * The only place I know of where this will NOT work is certain Macintosh
  * 680x0 compilers that define double as a 10-byte IEEE extended float.
@@ -78,9 +81,9 @@
  * such a compiler.
  */
 
-#ifndef ALIGN_SIZE		/* so can override from jconfig.h */
+#ifndef ALIGN_SIZE              /* so can override from jconfig.h */
 #ifndef WITH_SIMD
-#define ALIGN_SIZE  SIZEOF(double)
+#define ALIGN_SIZE  sizeof(double)
 #else
 #define ALIGN_SIZE  16 /* Most SIMD implementations require this */
 #endif
@@ -91,24 +94,23 @@
  * request to jpeg_get_small() or jpeg_get_large().  There is no per-object
  * overhead within a pool, except for alignment padding.  Each pool has a
  * header with a link to the next pool of the same class.
- * Small and large pool headers are identical except that the latter's
- * link pointer must be FAR on 80x86 machines.
+ * Small and large pool headers are identical.
  */
 
-typedef struct small_pool_struct * small_pool_ptr;
+typedef struct small_pool_struct *small_pool_ptr;
 
 typedef struct small_pool_struct {
-  small_pool_ptr next;	/* next in list of pools */
-  size_t bytes_used;		/* how many bytes already used within pool */
-  size_t bytes_left;		/* bytes still available in this pool */
+  small_pool_ptr next;  /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
 } small_pool_hdr;
 
-typedef struct large_pool_struct FAR * large_pool_ptr;
+typedef struct large_pool_struct *large_pool_ptr;
 
 typedef struct large_pool_struct {
-  large_pool_ptr next;	/* next in list of pools */
-  size_t bytes_used;		/* how many bytes already used within pool */
-  size_t bytes_left;		/* bytes still available in this pool */
+  large_pool_ptr next;  /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
 } large_pool_hdr;
 
 /*
@@ -116,7 +118,7 @@
  */
 
 typedef struct {
-  struct jpeg_memory_mgr pub;	/* public fields */
+  struct jpeg_memory_mgr pub;   /* public fields */
 
   /* Each pool identifier (lifetime class) names a linked list of pools. */
   small_pool_ptr small_list[JPOOL_NUMPOOLS];
@@ -136,10 +138,10 @@
   /* alloc_sarray and alloc_barray set this value for use by virtual
    * array routines.
    */
-  JDIMENSION last_rowsperchunk;	/* from most recent alloc_sarray/barray */
+  JDIMENSION last_rowsperchunk; /* from most recent alloc_sarray/barray */
 } my_memory_mgr;
 
-typedef my_memory_mgr * my_mem_ptr;
+typedef my_memory_mgr *my_mem_ptr;
 
 
 /*
@@ -150,39 +152,39 @@
  */
 
 struct jvirt_sarray_control {
-  JSAMPARRAY mem_buffer;	/* => the in-memory buffer */
-  JDIMENSION rows_in_array;	/* total virtual array height */
-  JDIMENSION samplesperrow;	/* width of array (and of memory buffer) */
-  JDIMENSION maxaccess;		/* max rows accessed by access_virt_sarray */
-  JDIMENSION rows_in_mem;	/* height of memory buffer */
-  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
-  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
-  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
-  boolean pre_zero;		/* pre-zero mode requested? */
-  boolean dirty;		/* do current buffer contents need written? */
-  boolean b_s_open;		/* is backing-store data valid? */
-  jvirt_sarray_ptr next;	/* link to next virtual sarray control block */
-  backing_store_info b_s_info;	/* System-dependent control info */
+  JSAMPARRAY mem_buffer;        /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION samplesperrow;     /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_sarray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_sarray_ptr next;        /* link to next virtual sarray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
 };
 
 struct jvirt_barray_control {
-  JBLOCKARRAY mem_buffer;	/* => the in-memory buffer */
-  JDIMENSION rows_in_array;	/* total virtual array height */
-  JDIMENSION blocksperrow;	/* width of array (and of memory buffer) */
-  JDIMENSION maxaccess;		/* max rows accessed by access_virt_barray */
-  JDIMENSION rows_in_mem;	/* height of memory buffer */
-  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
-  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
-  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
-  boolean pre_zero;		/* pre-zero mode requested? */
-  boolean dirty;		/* do current buffer contents need written? */
-  boolean b_s_open;		/* is backing-store data valid? */
-  jvirt_barray_ptr next;	/* link to next virtual barray control block */
-  backing_store_info b_s_info;	/* System-dependent control info */
+  JBLOCKARRAY mem_buffer;       /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION blocksperrow;      /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_barray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_barray_ptr next;        /* link to next virtual barray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
 };
 
 
-#ifdef MEM_STATS		/* optional extra stuff for statistics */
+#ifdef MEM_STATS                /* optional extra stuff for statistics */
 
 LOCAL(void)
 print_mem_stats (j_common_ptr cinfo, int pool_id)
@@ -196,19 +198,19 @@
    * This is helpful because message parm array can't handle longs.
    */
   fprintf(stderr, "Freeing pool %d, total space = %ld\n",
-	  pool_id, mem->total_space_allocated);
+          pool_id, mem->total_space_allocated);
 
   for (lhdr_ptr = mem->large_list[pool_id]; lhdr_ptr != NULL;
        lhdr_ptr = lhdr_ptr->next) {
     fprintf(stderr, "  Large chunk used %ld\n",
-	    (long) lhdr_ptr->bytes_used);
+            (long) lhdr_ptr->bytes_used);
   }
 
   for (shdr_ptr = mem->small_list[pool_id]; shdr_ptr != NULL;
        shdr_ptr = shdr_ptr->next) {
     fprintf(stderr, "  Small chunk used %ld free %ld\n",
-	    (long) shdr_ptr->bytes_used,
-	    (long) shdr_ptr->bytes_left);
+            (long) shdr_ptr->bytes_used,
+            (long) shdr_ptr->bytes_left);
   }
 }
 
@@ -221,7 +223,7 @@
 /* If we compiled MEM_STATS support, report alloc requests before dying */
 {
 #ifdef MEM_STATS
-  cinfo->err->trace_level = 2;	/* force self_destruct to report stats */
+  cinfo->err->trace_level = 2;  /* force self_destruct to report stats */
 #endif
   ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which);
 }
@@ -244,19 +246,19 @@
  * adjustment.
  */
 
-static const size_t first_pool_slop[JPOOL_NUMPOOLS] = 
+static const size_t first_pool_slop[JPOOL_NUMPOOLS] =
 {
-	1600,			/* first PERMANENT pool */
-	16000			/* first IMAGE pool */
+        1600,                   /* first PERMANENT pool */
+        16000                   /* first IMAGE pool */
 };
 
-static const size_t extra_pool_slop[JPOOL_NUMPOOLS] = 
+static const size_t extra_pool_slop[JPOOL_NUMPOOLS] =
 {
-	0,			/* additional PERMANENT pools */
-	5000			/* additional IMAGE pools */
+        0,                      /* additional PERMANENT pools */
+        5000                    /* additional IMAGE pools */
 };
 
-#define MIN_SLOP  50		/* greater than 0 to avoid futile looping */
+#define MIN_SLOP  50            /* greater than 0 to avoid futile looping */
 
 
 METHODDEF(void *)
@@ -265,7 +267,7 @@
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
   small_pool_ptr hdr_ptr, prev_hdr_ptr;
-  char * data_ptr;
+  char *data_ptr;
   size_t min_request, slop;
 
   /*
@@ -274,20 +276,26 @@
    * and so that algorithms can straddle outside the proper area up
    * to the next alignment.
    */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 7);
+  }
   sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
 
   /* Check for unsatisfiable request (do now to ensure no overflow below) */
-  if ((SIZEOF(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK)
-    out_of_memory(cinfo, 1);	/* request exceeds malloc's ability */
+  if ((sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 1);    /* request exceeds malloc's ability */
 
   /* See if space is available in any existing pool */
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
   prev_hdr_ptr = NULL;
   hdr_ptr = mem->small_list[pool_id];
   while (hdr_ptr != NULL) {
     if (hdr_ptr->bytes_left >= sizeofobject)
-      break;			/* found pool with enough space */
+      break;                    /* found pool with enough space */
     prev_hdr_ptr = hdr_ptr;
     hdr_ptr = hdr_ptr->next;
   }
@@ -295,8 +303,8 @@
   /* Time to make a new pool? */
   if (hdr_ptr == NULL) {
     /* min_request is what we need now, slop is what will be leftover */
-    min_request = SIZEOF(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1;
-    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+    min_request = sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1;
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
       slop = first_pool_slop[pool_id];
     else
       slop = extra_pool_slop[pool_id];
@@ -307,17 +315,17 @@
     for (;;) {
       hdr_ptr = (small_pool_ptr) jpeg_get_small(cinfo, min_request + slop);
       if (hdr_ptr != NULL)
-	break;
+        break;
       slop /= 2;
-      if (slop < MIN_SLOP)	/* give up when it gets real small */
-	out_of_memory(cinfo, 2); /* jpeg_get_small failed */
+      if (slop < MIN_SLOP)      /* give up when it gets real small */
+        out_of_memory(cinfo, 2); /* jpeg_get_small failed */
     }
     mem->total_space_allocated += min_request + slop;
     /* Success, initialize the new pool header and add to end of list */
     hdr_ptr->next = NULL;
     hdr_ptr->bytes_used = 0;
     hdr_ptr->bytes_left = sizeofobject + slop;
-    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
       mem->small_list[pool_id] = hdr_ptr;
     else
       prev_hdr_ptr->next = hdr_ptr;
@@ -325,7 +333,7 @@
 
   /* OK, allocate the object from the current pool */
   data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */
-  data_ptr += SIZEOF(small_pool_hdr); /* ...by skipping the header... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
   if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
     data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
   data_ptr += hdr_ptr->bytes_used; /* point to place for object */
@@ -339,9 +347,8 @@
 /*
  * Allocation of "large" objects.
  *
- * The external semantics of these are the same as "small" objects,
- * except that FAR pointers are used on 80x86.  However the pool
- * management heuristics are quite different.  We assume that each
+ * The external semantics of these are the same as "small" objects.  However,
+ * the pool management heuristics are quite different.  We assume that each
  * request is large enough that it may as well be passed directly to
  * jpeg_get_large; the pool management just links everything together
  * so that we can free it all on demand.
@@ -350,35 +357,42 @@
  * deliberately bunch rows together to ensure a large request size.
  */
 
-METHODDEF(void FAR *)
+METHODDEF(void *)
 alloc_large (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
 /* Allocate a "large" object */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
   large_pool_ptr hdr_ptr;
-  char FAR * data_ptr;
+  char *data_ptr;
 
   /*
    * Round up the requested size to a multiple of ALIGN_SIZE so that
    * algorithms can straddle outside the proper area up to the next
    * alignment.
    */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 8);
+  }
   sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
 
   /* Check for unsatisfiable request (do now to ensure no overflow below) */
-  if ((SIZEOF(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK)
-    out_of_memory(cinfo, 3);	/* request exceeds malloc's ability */
+  if ((sizeof(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 3);    /* request exceeds malloc's ability */
 
   /* Always make a new pool */
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   hdr_ptr = (large_pool_ptr) jpeg_get_large(cinfo, sizeofobject +
-					    SIZEOF(large_pool_hdr) +
-					    ALIGN_SIZE - 1);
+                                            sizeof(large_pool_hdr) +
+                                            ALIGN_SIZE - 1);
   if (hdr_ptr == NULL)
-    out_of_memory(cinfo, 4);	/* jpeg_get_large failed */
-  mem->total_space_allocated += sizeofobject + SIZEOF(large_pool_hdr) + ALIGN_SIZE - 1;
+    out_of_memory(cinfo, 4);    /* jpeg_get_large failed */
+  mem->total_space_allocated += sizeofobject + sizeof(large_pool_hdr) +
+                                ALIGN_SIZE - 1;
 
   /* Success, initialize the new pool header and add to list */
   hdr_ptr->next = mem->large_list[pool_id];
@@ -390,17 +404,16 @@
   mem->large_list[pool_id] = hdr_ptr;
 
   data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */
-  data_ptr += SIZEOF(small_pool_hdr); /* ...by skipping the header... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
   if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
     data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
 
-  return (void FAR *) data_ptr;
+  return (void *) data_ptr;
 }
 
 
 /*
  * Creation of 2-D sample arrays.
- * The pointers are in near heap, the samples themselves in FAR heap.
  *
  * To minimize allocation overhead and to allow I/O of large contiguous
  * blocks, we allocate the sample rows in groups of as many rows as possible
@@ -417,7 +430,7 @@
 
 METHODDEF(JSAMPARRAY)
 alloc_sarray (j_common_ptr cinfo, int pool_id,
-	      JDIMENSION samplesperrow, JDIMENSION numrows)
+              JDIMENSION samplesperrow, JDIMENSION numrows)
 /* Allocate a 2-D sample array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -427,13 +440,20 @@
   long ltemp;
 
   /* Make sure each row is properly aligned */
-  if ((ALIGN_SIZE % SIZEOF(JSAMPLE)) != 0)
-    out_of_memory(cinfo, 5);	/* safety check */
-  samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) / SIZEOF(JSAMPLE));
+  if ((ALIGN_SIZE % sizeof(JSAMPLE)) != 0)
+    out_of_memory(cinfo, 5);    /* safety check */
+
+  if (samplesperrow > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 9);
+  }
+  samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) /
+                                                           sizeof(JSAMPLE));
 
   /* Calculate max # of rows allowed in one allocation chunk */
-  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
-	  ((long) samplesperrow * SIZEOF(JSAMPLE));
+  ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) /
+          ((long) samplesperrow * sizeof(JSAMPLE));
   if (ltemp <= 0)
     ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
   if (ltemp < (long) numrows)
@@ -444,15 +464,15 @@
 
   /* Get space for row pointers (small object) */
   result = (JSAMPARRAY) alloc_small(cinfo, pool_id,
-				    (size_t) (numrows * SIZEOF(JSAMPROW)));
+                                    (size_t) (numrows * sizeof(JSAMPROW)));
 
   /* Get the rows themselves (large objects) */
   currow = 0;
   while (currow < numrows) {
     rowsperchunk = MIN(rowsperchunk, numrows - currow);
     workspace = (JSAMPROW) alloc_large(cinfo, pool_id,
-	(size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
-		  * SIZEOF(JSAMPLE)));
+        (size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
+                  * sizeof(JSAMPLE)));
     for (i = rowsperchunk; i > 0; i--) {
       result[currow++] = workspace;
       workspace += samplesperrow;
@@ -470,7 +490,7 @@
 
 METHODDEF(JBLOCKARRAY)
 alloc_barray (j_common_ptr cinfo, int pool_id,
-	      JDIMENSION blocksperrow, JDIMENSION numrows)
+              JDIMENSION blocksperrow, JDIMENSION numrows)
 /* Allocate a 2-D coefficient-block array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -480,12 +500,12 @@
   long ltemp;
 
   /* Make sure each row is properly aligned */
-  if ((SIZEOF(JBLOCK) % ALIGN_SIZE) != 0)
-    out_of_memory(cinfo, 6);	/* safety check */
+  if ((sizeof(JBLOCK) % ALIGN_SIZE) != 0)
+    out_of_memory(cinfo, 6);    /* safety check */
 
   /* Calculate max # of rows allowed in one allocation chunk */
-  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
-	  ((long) blocksperrow * SIZEOF(JBLOCK));
+  ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) /
+          ((long) blocksperrow * sizeof(JBLOCK));
   if (ltemp <= 0)
     ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
   if (ltemp < (long) numrows)
@@ -496,15 +516,15 @@
 
   /* Get space for row pointers (small object) */
   result = (JBLOCKARRAY) alloc_small(cinfo, pool_id,
-				     (size_t) (numrows * SIZEOF(JBLOCKROW)));
+                                     (size_t) (numrows * sizeof(JBLOCKROW)));
 
   /* Get the rows themselves (large objects) */
   currow = 0;
   while (currow < numrows) {
     rowsperchunk = MIN(rowsperchunk, numrows - currow);
     workspace = (JBLOCKROW) alloc_large(cinfo, pool_id,
-	(size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
-		  * SIZEOF(JBLOCK)));
+        (size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
+                  * sizeof(JBLOCK)));
     for (i = rowsperchunk; i > 0; i--) {
       result[currow++] = workspace;
       workspace += blocksperrow;
@@ -554,8 +574,8 @@
 
 METHODDEF(jvirt_sarray_ptr)
 request_virt_sarray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
-		     JDIMENSION samplesperrow, JDIMENSION numrows,
-		     JDIMENSION maxaccess)
+                     JDIMENSION samplesperrow, JDIMENSION numrows,
+                     JDIMENSION maxaccess)
 /* Request a virtual 2-D sample array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -563,18 +583,18 @@
 
   /* Only IMAGE-lifetime virtual arrays are currently supported */
   if (pool_id != JPOOL_IMAGE)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   /* get control block */
   result = (jvirt_sarray_ptr) alloc_small(cinfo, pool_id,
-					  SIZEOF(struct jvirt_sarray_control));
+                                          sizeof(struct jvirt_sarray_control));
 
-  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->mem_buffer = NULL;    /* marks array not yet realized */
   result->rows_in_array = numrows;
   result->samplesperrow = samplesperrow;
   result->maxaccess = maxaccess;
   result->pre_zero = pre_zero;
-  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->b_s_open = FALSE;     /* no associated backing-store object */
   result->next = mem->virt_sarray_list; /* add to list of virtual arrays */
   mem->virt_sarray_list = result;
 
@@ -584,8 +604,8 @@
 
 METHODDEF(jvirt_barray_ptr)
 request_virt_barray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
-		     JDIMENSION blocksperrow, JDIMENSION numrows,
-		     JDIMENSION maxaccess)
+                     JDIMENSION blocksperrow, JDIMENSION numrows,
+                     JDIMENSION maxaccess)
 /* Request a virtual 2-D coefficient-block array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -593,18 +613,18 @@
 
   /* Only IMAGE-lifetime virtual arrays are currently supported */
   if (pool_id != JPOOL_IMAGE)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   /* get control block */
   result = (jvirt_barray_ptr) alloc_small(cinfo, pool_id,
-					  SIZEOF(struct jvirt_barray_control));
+                                          sizeof(struct jvirt_barray_control));
 
-  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->mem_buffer = NULL;    /* marks array not yet realized */
   result->rows_in_array = numrows;
   result->blocksperrow = blocksperrow;
   result->maxaccess = maxaccess;
   result->pre_zero = pre_zero;
-  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->b_s_open = FALSE;     /* no associated backing-store object */
   result->next = mem->virt_barray_list; /* add to list of virtual arrays */
   mem->virt_barray_list = result;
 
@@ -631,26 +651,26 @@
   for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
     if (sptr->mem_buffer == NULL) { /* if not realized yet */
       space_per_minheight += (long) sptr->maxaccess *
-			     (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+                             (long) sptr->samplesperrow * sizeof(JSAMPLE);
       maximum_space += (long) sptr->rows_in_array *
-		       (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+                       (long) sptr->samplesperrow * sizeof(JSAMPLE);
     }
   }
   for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
     if (bptr->mem_buffer == NULL) { /* if not realized yet */
       space_per_minheight += (long) bptr->maxaccess *
-			     (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+                             (long) bptr->blocksperrow * sizeof(JBLOCK);
       maximum_space += (long) bptr->rows_in_array *
-		       (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+                       (long) bptr->blocksperrow * sizeof(JBLOCK);
     }
   }
 
   if (space_per_minheight <= 0)
-    return;			/* no unrealized arrays, no work */
+    return;                     /* no unrealized arrays, no work */
 
   /* Determine amount of memory to actually use; this is system-dependent. */
   avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space,
-				 mem->total_space_allocated);
+                                 mem->total_space_allocated);
 
   /* If the maximum space needed is available, make all the buffers full
    * height; otherwise parcel it out with the same number of minheights
@@ -673,19 +693,19 @@
     if (sptr->mem_buffer == NULL) { /* if not realized yet */
       minheights = ((long) sptr->rows_in_array - 1L) / sptr->maxaccess + 1L;
       if (minheights <= max_minheights) {
-	/* This buffer fits in memory */
-	sptr->rows_in_mem = sptr->rows_in_array;
+        /* This buffer fits in memory */
+        sptr->rows_in_mem = sptr->rows_in_array;
       } else {
-	/* It doesn't fit in memory, create backing store. */
-	sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
-	jpeg_open_backing_store(cinfo, & sptr->b_s_info,
-				(long) sptr->rows_in_array *
-				(long) sptr->samplesperrow *
-				(long) SIZEOF(JSAMPLE));
-	sptr->b_s_open = TRUE;
+        /* It doesn't fit in memory, create backing store. */
+        sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
+        jpeg_open_backing_store(cinfo, & sptr->b_s_info,
+                                (long) sptr->rows_in_array *
+                                (long) sptr->samplesperrow *
+                                (long) sizeof(JSAMPLE));
+        sptr->b_s_open = TRUE;
       }
       sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE,
-				      sptr->samplesperrow, sptr->rows_in_mem);
+                                      sptr->samplesperrow, sptr->rows_in_mem);
       sptr->rowsperchunk = mem->last_rowsperchunk;
       sptr->cur_start_row = 0;
       sptr->first_undef_row = 0;
@@ -697,19 +717,19 @@
     if (bptr->mem_buffer == NULL) { /* if not realized yet */
       minheights = ((long) bptr->rows_in_array - 1L) / bptr->maxaccess + 1L;
       if (minheights <= max_minheights) {
-	/* This buffer fits in memory */
-	bptr->rows_in_mem = bptr->rows_in_array;
+        /* This buffer fits in memory */
+        bptr->rows_in_mem = bptr->rows_in_array;
       } else {
-	/* It doesn't fit in memory, create backing store. */
-	bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
-	jpeg_open_backing_store(cinfo, & bptr->b_s_info,
-				(long) bptr->rows_in_array *
-				(long) bptr->blocksperrow *
-				(long) SIZEOF(JBLOCK));
-	bptr->b_s_open = TRUE;
+        /* It doesn't fit in memory, create backing store. */
+        bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
+        jpeg_open_backing_store(cinfo, & bptr->b_s_info,
+                                (long) bptr->rows_in_array *
+                                (long) bptr->blocksperrow *
+                                (long) sizeof(JBLOCK));
+        bptr->b_s_open = TRUE;
       }
       bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE,
-				      bptr->blocksperrow, bptr->rows_in_mem);
+                                      bptr->blocksperrow, bptr->rows_in_mem);
       bptr->rowsperchunk = mem->last_rowsperchunk;
       bptr->cur_start_row = 0;
       bptr->first_undef_row = 0;
@@ -725,7 +745,7 @@
 {
   long bytesperrow, file_offset, byte_count, rows, thisrow, i;
 
-  bytesperrow = (long) ptr->samplesperrow * SIZEOF(JSAMPLE);
+  bytesperrow = (long) ptr->samplesperrow * sizeof(JSAMPLE);
   file_offset = ptr->cur_start_row * bytesperrow;
   /* Loop to read or write each allocation chunk in mem_buffer */
   for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
@@ -736,17 +756,17 @@
     rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
     /* Transfer no more than fits in file */
     rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
-    if (rows <= 0)		/* this chunk might be past end of file! */
+    if (rows <= 0)              /* this chunk might be past end of file! */
       break;
     byte_count = rows * bytesperrow;
     if (writing)
       (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
-					    (void FAR *) ptr->mem_buffer[i],
-					    file_offset, byte_count);
+                                            (void *) ptr->mem_buffer[i],
+                                            file_offset, byte_count);
     else
       (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
-					   (void FAR *) ptr->mem_buffer[i],
-					   file_offset, byte_count);
+                                           (void *) ptr->mem_buffer[i],
+                                           file_offset, byte_count);
     file_offset += byte_count;
   }
 }
@@ -758,7 +778,7 @@
 {
   long bytesperrow, file_offset, byte_count, rows, thisrow, i;
 
-  bytesperrow = (long) ptr->blocksperrow * SIZEOF(JBLOCK);
+  bytesperrow = (long) ptr->blocksperrow * sizeof(JBLOCK);
   file_offset = ptr->cur_start_row * bytesperrow;
   /* Loop to read or write each allocation chunk in mem_buffer */
   for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
@@ -769,17 +789,17 @@
     rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
     /* Transfer no more than fits in file */
     rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
-    if (rows <= 0)		/* this chunk might be past end of file! */
+    if (rows <= 0)              /* this chunk might be past end of file! */
       break;
     byte_count = rows * bytesperrow;
     if (writing)
       (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
-					    (void FAR *) ptr->mem_buffer[i],
-					    file_offset, byte_count);
+                                            (void *) ptr->mem_buffer[i],
+                                            file_offset, byte_count);
     else
       (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
-					   (void FAR *) ptr->mem_buffer[i],
-					   file_offset, byte_count);
+                                           (void *) ptr->mem_buffer[i],
+                                           file_offset, byte_count);
     file_offset += byte_count;
   }
 }
@@ -787,8 +807,8 @@
 
 METHODDEF(JSAMPARRAY)
 access_virt_sarray (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
-		    JDIMENSION start_row, JDIMENSION num_rows,
-		    boolean writable)
+                    JDIMENSION start_row, JDIMENSION num_rows,
+                    boolean writable)
 /* Access the part of a virtual sample array starting at start_row */
 /* and extending for num_rows rows.  writable is true if  */
 /* caller intends to modify the accessed area. */
@@ -826,7 +846,7 @@
 
       ltemp = (long) end_row - (long) ptr->rows_in_mem;
       if (ltemp < 0)
-	ltemp = 0;		/* don't fall off front end of file */
+        ltemp = 0;              /* don't fall off front end of file */
       ptr->cur_start_row = (JDIMENSION) ltemp;
     }
     /* Read in the selected part of the array.
@@ -841,25 +861,25 @@
    */
   if (ptr->first_undef_row < end_row) {
     if (ptr->first_undef_row < start_row) {
-      if (writable)		/* writer skipped over a section of array */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
-      undef_row = start_row;	/* but reader is allowed to read ahead */
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
     } else {
       undef_row = ptr->first_undef_row;
     }
     if (writable)
       ptr->first_undef_row = end_row;
     if (ptr->pre_zero) {
-      size_t bytesperrow = (size_t) ptr->samplesperrow * SIZEOF(JSAMPLE);
+      size_t bytesperrow = (size_t) ptr->samplesperrow * sizeof(JSAMPLE);
       undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
       end_row -= ptr->cur_start_row;
       while (undef_row < end_row) {
-	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
-	undef_row++;
+        jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow);
+        undef_row++;
       }
     } else {
-      if (! writable)		/* reader looking at undefined data */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      if (! writable)           /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
     }
   }
   /* Flag the buffer dirty if caller will write in it */
@@ -872,8 +892,8 @@
 
 METHODDEF(JBLOCKARRAY)
 access_virt_barray (j_common_ptr cinfo, jvirt_barray_ptr ptr,
-		    JDIMENSION start_row, JDIMENSION num_rows,
-		    boolean writable)
+                    JDIMENSION start_row, JDIMENSION num_rows,
+                    boolean writable)
 /* Access the part of a virtual block array starting at start_row */
 /* and extending for num_rows rows.  writable is true if  */
 /* caller intends to modify the accessed area. */
@@ -911,7 +931,7 @@
 
       ltemp = (long) end_row - (long) ptr->rows_in_mem;
       if (ltemp < 0)
-	ltemp = 0;		/* don't fall off front end of file */
+        ltemp = 0;              /* don't fall off front end of file */
       ptr->cur_start_row = (JDIMENSION) ltemp;
     }
     /* Read in the selected part of the array.
@@ -926,25 +946,25 @@
    */
   if (ptr->first_undef_row < end_row) {
     if (ptr->first_undef_row < start_row) {
-      if (writable)		/* writer skipped over a section of array */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
-      undef_row = start_row;	/* but reader is allowed to read ahead */
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
     } else {
       undef_row = ptr->first_undef_row;
     }
     if (writable)
       ptr->first_undef_row = end_row;
     if (ptr->pre_zero) {
-      size_t bytesperrow = (size_t) ptr->blocksperrow * SIZEOF(JBLOCK);
+      size_t bytesperrow = (size_t) ptr->blocksperrow * sizeof(JBLOCK);
       undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
       end_row -= ptr->cur_start_row;
       while (undef_row < end_row) {
-	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
-	undef_row++;
+        jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow);
+        undef_row++;
       }
     } else {
-      if (! writable)		/* reader looking at undefined data */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      if (! writable)           /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
     }
   }
   /* Flag the buffer dirty if caller will write in it */
@@ -968,7 +988,7 @@
   size_t space_freed;
 
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
 #ifdef MEM_STATS
   if (cinfo->err->trace_level > 1)
@@ -981,16 +1001,16 @@
     jvirt_barray_ptr bptr;
 
     for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
-      if (sptr->b_s_open) {	/* there may be no backing store */
-	sptr->b_s_open = FALSE;	/* prevent recursive close if error */
-	(*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
+      if (sptr->b_s_open) {     /* there may be no backing store */
+        sptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
       }
     }
     mem->virt_sarray_list = NULL;
     for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
-      if (bptr->b_s_open) {	/* there may be no backing store */
-	bptr->b_s_open = FALSE;	/* prevent recursive close if error */
-	(*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
+      if (bptr->b_s_open) {     /* there may be no backing store */
+        bptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
       }
     }
     mem->virt_barray_list = NULL;
@@ -1003,9 +1023,9 @@
   while (lhdr_ptr != NULL) {
     large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
     space_freed = lhdr_ptr->bytes_used +
-		  lhdr_ptr->bytes_left +
-		  SIZEOF(large_pool_hdr);
-    jpeg_free_large(cinfo, (void FAR *) lhdr_ptr, space_freed);
+                  lhdr_ptr->bytes_left +
+                  sizeof(large_pool_hdr);
+    jpeg_free_large(cinfo, (void *) lhdr_ptr, space_freed);
     mem->total_space_allocated -= space_freed;
     lhdr_ptr = next_lhdr_ptr;
   }
@@ -1017,8 +1037,8 @@
   while (shdr_ptr != NULL) {
     small_pool_ptr next_shdr_ptr = shdr_ptr->next;
     space_freed = shdr_ptr->bytes_used +
-		  shdr_ptr->bytes_left +
-		  SIZEOF(small_pool_hdr);
+                  shdr_ptr->bytes_left +
+                  sizeof(small_pool_hdr);
     jpeg_free_small(cinfo, (void *) shdr_ptr, space_freed);
     mem->total_space_allocated -= space_freed;
     shdr_ptr = next_shdr_ptr;
@@ -1045,10 +1065,10 @@
   }
 
   /* Release the memory manager control block too. */
-  jpeg_free_small(cinfo, (void *) cinfo->mem, SIZEOF(my_memory_mgr));
-  cinfo->mem = NULL;		/* ensures I will be called only once */
+  jpeg_free_small(cinfo, (void *) cinfo->mem, sizeof(my_memory_mgr));
+  cinfo->mem = NULL;            /* ensures I will be called only once */
 
-  jpeg_mem_term(cinfo);		/* system-dependent cleanup */
+  jpeg_mem_term(cinfo);         /* system-dependent cleanup */
 }
 
 
@@ -1065,10 +1085,10 @@
   int pool;
   size_t test_mac;
 
-  cinfo->mem = NULL;		/* for safety if init fails */
+  cinfo->mem = NULL;            /* for safety if init fails */
 
   /* Check for configuration errors.
-   * SIZEOF(ALIGN_TYPE) should be a power of 2; otherwise, it probably
+   * sizeof(ALIGN_TYPE) should be a power of 2; otherwise, it probably
    * doesn't reflect any real hardware alignment requirement.
    * The test is a little tricky: for X>0, X and X-1 have no one-bits
    * in common if and only if X is a power of 2, ie has only one one-bit.
@@ -1089,10 +1109,10 @@
   max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */
 
   /* Attempt to allocate memory manager's control block */
-  mem = (my_mem_ptr) jpeg_get_small(cinfo, SIZEOF(my_memory_mgr));
+  mem = (my_mem_ptr) jpeg_get_small(cinfo, sizeof(my_memory_mgr));
 
   if (mem == NULL) {
-    jpeg_mem_term(cinfo);	/* system-dependent cleanup */
+    jpeg_mem_term(cinfo);       /* system-dependent cleanup */
     ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0);
   }
 
@@ -1122,7 +1142,7 @@
   mem->virt_sarray_list = NULL;
   mem->virt_barray_list = NULL;
 
-  mem->total_space_allocated = SIZEOF(my_memory_mgr);
+  mem->total_space_allocated = sizeof(my_memory_mgr);
 
   /* Declare ourselves open for business */
   cinfo->mem = & mem->pub;
@@ -1134,15 +1154,15 @@
    * this feature.
    */
 #ifndef NO_GETENV
-  { char * memenv;
+  { char *memenv;
 
     if ((memenv = getenv("JPEGMEM")) != NULL) {
       char ch = 'x';
 
       if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) {
-	if (ch == 'm' || ch == 'M')
-	  max_to_use *= 1000L;
-	mem->pub.max_memory_to_use = max_to_use * 1000L;
+        if (ch == 'm' || ch == 'M')
+          max_to_use *= 1000L;
+        mem->pub.max_memory_to_use = max_to_use * 1000L;
       }
     }
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemnobs.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemnobs.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemnobs.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemnobs.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemnobs.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1992-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file provides a really simple implementation of the system-
  * dependent portion of the JPEG memory manager.  This implementation
@@ -18,11 +21,11 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jmemsys.h"		/* import the system-dependent declarations */
+#include "jmemsys.h"            /* import the system-dependent declarations */
 
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
-extern void * malloc JPP((size_t size));
-extern void free JPP((void *ptr));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
 #endif
 
 
@@ -38,7 +41,7 @@
 }
 
 GLOBAL(void)
-jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+jpeg_free_small (j_common_ptr cinfo, void *object, size_t sizeofobject)
 {
   free(object);
 }
@@ -46,19 +49,16 @@
 
 /*
  * "Large" objects are treated the same as "small" ones.
- * NB: although we include FAR keywords in the routine declarations,
- * this file won't actually work in 80x86 small/medium model; at least,
- * you probably won't be able to process useful-size images in only 64KB.
  */
 
-GLOBAL(void FAR *)
+GLOBAL(void *)
 jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
 {
-  return (void FAR *) malloc(sizeofobject);
+  return (void *) malloc(sizeofobject);
 }
 
 GLOBAL(void)
-jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+jpeg_free_large (j_common_ptr cinfo, void *object, size_t sizeofobject)
 {
   free(object);
 }
@@ -71,7 +71,7 @@
 
 GLOBAL(size_t)
 jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed,
-		    size_t max_bytes_needed, size_t already_allocated)
+                    size_t max_bytes_needed, size_t already_allocated)
 {
   return max_bytes_needed;
 }
@@ -85,7 +85,7 @@
 
 GLOBAL(void)
 jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
-			 long total_bytes_needed)
+                         long total_bytes_needed)
 {
   ERREXIT(cinfo, JERR_NO_BACKING_STORE);
 }
@@ -99,7 +99,7 @@
 GLOBAL(long)
 jpeg_mem_init (j_common_ptr cinfo)
 {
-  return 0;			/* just set max_memory_to_use to 0 */
+  return 0;                     /* just set max_memory_to_use to 0 */
 }
 
 GLOBAL(void)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemsys.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemsys.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmemsys.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jmemsys.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemsys.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1992-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This include file defines the interface between the system-independent
  * and system-dependent portions of the JPEG memory manager.  No other
@@ -14,25 +17,10 @@
  * in the IJG distribution.  You may need to modify it if you write a
  * custom memory manager.  If system-dependent changes are needed in
  * this file, the best method is to #ifdef them based on a configuration
- * symbol supplied in jconfig.h, as we have done with USE_MSDOS_MEMMGR
- * and USE_MAC_MEMMGR.
+ * symbol supplied in jconfig.h.
  */
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_get_small		jGetSmall
-#define jpeg_free_small		jFreeSmall
-#define jpeg_get_large		jGetLarge
-#define jpeg_free_large		jFreeLarge
-#define jpeg_mem_available	jMemAvail
-#define jpeg_open_backing_store	jOpenBackStore
-#define jpeg_mem_init		jMemInit
-#define jpeg_mem_term		jMemTerm
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-
 /*
  * These two functions are used to allocate and release small chunks of
  * memory.  (Typically the total amount requested through jpeg_get_small is
@@ -41,40 +29,36 @@
  * and free; in particular, jpeg_get_small must return NULL on failure.
  * On most systems, these ARE malloc and free.  jpeg_free_small is passed the
  * size of the object being freed, just in case it's needed.
- * On an 80x86 machine using small-data memory model, these manage near heap.
  */
 
-EXTERN(void *) jpeg_get_small JPP((j_common_ptr cinfo, size_t sizeofobject));
-EXTERN(void) jpeg_free_small JPP((j_common_ptr cinfo, void * object,
-				  size_t sizeofobject));
+EXTERN(void *) jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_small (j_common_ptr cinfo, void *object,
+                              size_t sizeofobject);
 
 /*
  * These two functions are used to allocate and release large chunks of
  * memory (up to the total free space designated by jpeg_mem_available).
- * The interface is the same as above, except that on an 80x86 machine,
- * far pointers are used.  On most other machines these are identical to
- * the jpeg_get/free_small routines; but we keep them separate anyway,
- * in case a different allocation strategy is desirable for large chunks.
+ * These are identical to the jpeg_get/free_small routines; but we keep them
+ * separate anyway, in case a different allocation strategy is desirable for
+ * large chunks.
  */
 
-EXTERN(void FAR *) jpeg_get_large JPP((j_common_ptr cinfo,
-				       size_t sizeofobject));
-EXTERN(void) jpeg_free_large JPP((j_common_ptr cinfo, void FAR * object,
-				  size_t sizeofobject));
+EXTERN(void *) jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_large (j_common_ptr cinfo, void *object,
+                              size_t sizeofobject);
 
 /*
  * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may
  * be requested in a single call to jpeg_get_large (and jpeg_get_small for that
- * matter, but that case should never come into play).  This macro is needed
+ * matter, but that case should never come into play).  This macro was needed
  * to model the 64Kb-segment-size limit of far addressing on 80x86 machines.
- * On those machines, we expect that jconfig.h will provide a proper value.
- * On machines with 32-bit flat address spaces, any large constant may be used.
+ * On machines with flat address spaces, any large constant may be used.
  *
  * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type
  * size_t and will be a multiple of sizeof(align_type).
  */
 
-#ifndef MAX_ALLOC_CHUNK		/* may be overridden in jconfig.h */
+#ifndef MAX_ALLOC_CHUNK         /* may be overridden in jconfig.h */
 #define MAX_ALLOC_CHUNK  1000000000L
 #endif
 
@@ -100,10 +84,9 @@
  * Conversely, zero may be returned to always use the minimum amount of memory.
  */
 
-EXTERN(size_t) jpeg_mem_available JPP((j_common_ptr cinfo,
-				     size_t min_bytes_needed,
-				     size_t max_bytes_needed,
-				     size_t already_allocated));
+EXTERN(size_t) jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed,
+                                   size_t max_bytes_needed,
+                                   size_t already_allocated);
 
 
 /*
@@ -113,56 +96,53 @@
  * are private to the system-dependent backing store routines.
  */
 
-#define TEMP_NAME_LENGTH   64	/* max length of a temporary file's name */
+#define TEMP_NAME_LENGTH   64   /* max length of a temporary file's name */
 
 
-#ifdef USE_MSDOS_MEMMGR		/* DOS-specific junk */
+#ifdef USE_MSDOS_MEMMGR         /* DOS-specific junk */
 
-typedef unsigned short XMSH;	/* type of extended-memory handles */
-typedef unsigned short EMSH;	/* type of expanded-memory handles */
+typedef unsigned short XMSH;    /* type of extended-memory handles */
+typedef unsigned short EMSH;    /* type of expanded-memory handles */
 
 typedef union {
-  short file_handle;		/* DOS file handle if it's a temp file */
-  XMSH xms_handle;		/* handle if it's a chunk of XMS */
-  EMSH ems_handle;		/* handle if it's a chunk of EMS */
+  short file_handle;            /* DOS file handle if it's a temp file */
+  XMSH xms_handle;              /* handle if it's a chunk of XMS */
+  EMSH ems_handle;              /* handle if it's a chunk of EMS */
 } handle_union;
 
 #endif /* USE_MSDOS_MEMMGR */
 
-#ifdef USE_MAC_MEMMGR		/* Mac-specific junk */
+#ifdef USE_MAC_MEMMGR           /* Mac-specific junk */
 #include <Files.h>
 #endif /* USE_MAC_MEMMGR */
 
 
-typedef struct backing_store_struct * backing_store_ptr;
+typedef struct backing_store_struct *backing_store_ptr;
 
 typedef struct backing_store_struct {
   /* Methods for reading/writing/closing this backing-store object */
-  JMETHOD(void, read_backing_store, (j_common_ptr cinfo,
-				     backing_store_ptr info,
-				     void FAR * buffer_address,
-				     long file_offset, long byte_count));
-  JMETHOD(void, write_backing_store, (j_common_ptr cinfo,
-				      backing_store_ptr info,
-				      void FAR * buffer_address,
-				      long file_offset, long byte_count));
-  JMETHOD(void, close_backing_store, (j_common_ptr cinfo,
-				      backing_store_ptr info));
+  void (*read_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                              void *buffer_address, long file_offset,
+                              long byte_count);
+  void (*write_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                               void *buffer_address, long file_offset,
+                               long byte_count);
+  void (*close_backing_store) (j_common_ptr cinfo, backing_store_ptr info);
 
   /* Private fields for system-dependent backing-store management */
 #ifdef USE_MSDOS_MEMMGR
   /* For the MS-DOS manager (jmemdos.c), we need: */
-  handle_union handle;		/* reference to backing-store storage object */
+  handle_union handle;          /* reference to backing-store storage object */
   char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
 #else
 #ifdef USE_MAC_MEMMGR
   /* For the Mac manager (jmemmac.c), we need: */
-  short temp_file;		/* file reference number to temp file */
-  FSSpec tempSpec;		/* the FSSpec for the temp file */
+  short temp_file;              /* file reference number to temp file */
+  FSSpec tempSpec;              /* the FSSpec for the temp file */
   char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
 #else
   /* For a typical implementation with temp files, we need: */
-  FILE * temp_file;		/* stdio reference to temp file */
+  FILE *temp_file;              /* stdio reference to temp file */
   char temp_name[TEMP_NAME_LENGTH]; /* name of temp file */
 #endif
 #endif
@@ -177,9 +157,9 @@
  * just take an error exit.)
  */
 
-EXTERN(void) jpeg_open_backing_store JPP((j_common_ptr cinfo,
-					  backing_store_ptr info,
-					  long total_bytes_needed));
+EXTERN(void) jpeg_open_backing_store (j_common_ptr cinfo,
+                                      backing_store_ptr info,
+                                      long total_bytes_needed);
 
 
 /*
@@ -194,5 +174,5 @@
  * all opened backing-store objects have been closed.
  */
 
-EXTERN(long) jpeg_mem_init JPP((j_common_ptr cinfo));
-EXTERN(void) jpeg_mem_term JPP((j_common_ptr cinfo));
+EXTERN(long) jpeg_mem_init (j_common_ptr cinfo);
+EXTERN(void) jpeg_mem_term (j_common_ptr cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmorecfg.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jmorecfg.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jmorecfg.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jmorecfg.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jmorecfg.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains additional configuration options that customize the
  * JPEG software for special applications or support machine-dependent
@@ -13,18 +16,6 @@
 
 
 /*
- * Define BITS_IN_JSAMPLE as either
- *   8   for 8-bit sample values (the usual setting)
- *   12  for 12-bit sample values
- * Only 8 and 12 are legal data precisions for lossy JPEG according to the
- * JPEG standard, and the IJG code does not support anything else!
- * We do not support run-time selection of data precision, sorry.
- */
-
-#define BITS_IN_JSAMPLE  8	/* use 8 or 12 */
-
-
-/*
  * Maximum number of components (color channels) allowed in JPEG image.
  * To meet the letter of the JPEG spec, set this to 255.  However, darn
  * few applications need more than 4 channels (maybe 5 for CMYK + alpha
@@ -33,7 +24,7 @@
  * bytes of storage, whether actually used in an image or not.)
  */
 
-#define MAX_COMPONENTS  10	/* maximum number of image components */
+#define MAX_COMPONENTS  10      /* maximum number of image components */
 
 
 /*
@@ -71,8 +62,8 @@
 
 #endif /* HAVE_UNSIGNED_CHAR */
 
-#define MAXJSAMPLE	255
-#define CENTERJSAMPLE	128
+#define MAXJSAMPLE      255
+#define CENTERJSAMPLE   128
 
 #endif /* BITS_IN_JSAMPLE == 8 */
 
@@ -85,8 +76,8 @@
 typedef short JSAMPLE;
 #define GETJSAMPLE(value)  ((int) (value))
 
-#define MAXJSAMPLE	4095
-#define CENTERJSAMPLE	2048
+#define MAXJSAMPLE      4095
+#define CENTERJSAMPLE   2048
 
 #endif /* BITS_IN_JSAMPLE == 12 */
 
@@ -152,21 +143,52 @@
 
 /* INT16 must hold at least the values -32768..32767. */
 
-#ifndef XMD_H			/* X11/xmd.h correctly defines INT16 */
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT16 */
 typedef short INT16;
 #endif
 
-/* INT32 must hold at least signed 32-bit values. */
+/* INT32 must hold at least signed 32-bit values.
+ *
+ * NOTE: The INT32 typedef dates back to libjpeg v5 (1994.)  Integers were
+ * sometimes 16-bit back then (MS-DOS), which is why INT32 is typedef'd to
+ * long.  It also wasn't common (or at least as common) in 1994 for INT32 to be
+ * defined by platform headers.  Since then, however, INT32 is defined in
+ * several other common places:
+ *
+ * Xmd.h (X11 header) typedefs INT32 to int on 64-bit platforms and long on
+ * 32-bit platforms (i.e always a 32-bit signed type.)
+ *
+ * basetsd.h (Win32 header) typedefs INT32 to int (always a 32-bit signed type
+ * on modern platforms.)
+ *
+ * qglobal.h (Qt header) typedefs INT32 to int (always a 32-bit signed type on
+ * modern platforms.)
+ *
+ * This is a recipe for conflict, since "long" and "int" aren't always
+ * compatible types.  Since the definition of INT32 has technically been part
+ * of the libjpeg API for more than 20 years, we can't remove it, but we do not
+ * use it internally any longer.  We instead define a separate type (JLONG)
+ * for internal use, which ensures that internal behavior will always be the
+ * same regardless of any external headers that may be included.
+ */
 
-#ifndef XMD_H			/* X11/xmd.h correctly defines INT32 */
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT32 */
+#ifndef _BASETSD_H_		/* Microsoft defines it in basetsd.h */
+#ifndef _BASETSD_H		/* MinGW is slightly different */
+#ifndef QGLOBAL_H		/* Qt defines it in qglobal.h */
 typedef long INT32;
 #endif
+#endif
+#endif
+#endif
 
 /* Datatype used for image dimensions.  The JPEG standard only supports
  * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
  * "unsigned int" is sufficient on all machines.  However, if you need to
  * handle larger images and you don't mind deviating from the spec, you
- * can change this datatype.
+ * can change this datatype.  (Note that changing this datatype will
+ * potentially require modifying the SIMD code.  The x86-64 SIMD extensions,
+ * in particular, assume a 32-bit JDIMENSION.)
  */
 
 typedef unsigned int JDIMENSION;
@@ -182,39 +204,31 @@
  */
 
 /* a function called through method pointers: */
-#define METHODDEF(type)		static type
+#define METHODDEF(type)         static type
 /* a function used only in its module: */
-#define LOCAL(type)		static type
+#define LOCAL(type)             static type
 /* a function referenced thru EXTERNs: */
-#define GLOBAL(type)		type
+#define GLOBAL(type)            type
 /* a reference to a GLOBAL function: */
-#define EXTERN(type)		extern type
+#define EXTERN(type)            extern type
 
 
-/* This macro is used to declare a "method", that is, a function pointer.
- * We want to supply prototype parameters if the compiler can cope.
- * Note that the arglist parameter must be parenthesized!
- * Again, you can customize this if you need special linkage keywords.
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
  */
 
-#ifdef HAVE_PROTOTYPES
 #define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
-#else
-#define JMETHOD(type,methodname,arglist)  type (*methodname) ()
-#endif
 
 
-/* Here is the pseudo-keyword for declaring pointers that must be "far"
- * on 80x86 machines.  Most of the specialized coding for 80x86 is handled
- * by just saying "FAR *" where such a pointer is needed.  In a few places
- * explicit coding is needed; see uses of the NEED_FAR_POINTERS symbol.
+/* libjpeg-turbo no longer supports platforms that have far symbols (MS-DOS),
+ * but again, some software relies on this macro.
  */
 
-#ifdef NEED_FAR_POINTERS
-#define FAR  far
-#else
+#undef FAR
 #define FAR
-#endif
 
 
 /*
@@ -227,11 +241,11 @@
 #ifndef HAVE_BOOLEAN
 typedef int boolean;
 #endif
-#ifndef FALSE			/* in case these macros already exist */
-#define FALSE	0		/* values of boolean */
+#ifndef FALSE                   /* in case these macros already exist */
+#define FALSE   0               /* values of boolean */
 #endif
 #ifndef TRUE
-#define TRUE	1
+#define TRUE    1
 #endif
 
 
@@ -259,15 +273,15 @@
 
 /* Capability options common to encoder and decoder: */
 
-#define DCT_ISLOW_SUPPORTED	/* slow but accurate integer algorithm */
-#define DCT_IFAST_SUPPORTED	/* faster, less accurate integer method */
-#define DCT_FLOAT_SUPPORTED	/* floating-point: accurate, fast on fast HW */
+#define DCT_ISLOW_SUPPORTED     /* slow but accurate integer algorithm */
+#define DCT_IFAST_SUPPORTED     /* faster, less accurate integer method */
+#define DCT_FLOAT_SUPPORTED     /* floating-point: accurate, fast on fast HW */
 
 /* Encoder capability options: */
 
 #define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
-#define C_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
-#define ENTROPY_OPT_SUPPORTED	    /* Optimization of entropy coding parms? */
+#define C_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define ENTROPY_OPT_SUPPORTED       /* Optimization of entropy coding parms? */
 /* Note: if you selected 12-bit data precision, it is dangerous to turn off
  * ENTROPY_OPT_SUPPORTED.  The standard Huffman tables are only good for 8-bit
  * precision, so jchuff.c normally uses entropy optimization to compute
@@ -281,39 +295,45 @@
 /* Decoder capability options: */
 
 #define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
-#define D_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
-#define SAVE_MARKERS_SUPPORTED	    /* jpeg_save_markers() needed? */
+#define D_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define SAVE_MARKERS_SUPPORTED      /* jpeg_save_markers() needed? */
 #define BLOCK_SMOOTHING_SUPPORTED   /* Block smoothing? (Progressive only) */
-#define IDCT_SCALING_SUPPORTED	    /* Output rescaling via IDCT? */
+#define IDCT_SCALING_SUPPORTED      /* Output rescaling via IDCT? */
 #undef  UPSAMPLE_SCALING_SUPPORTED  /* Output rescaling at upsample stage? */
 #define UPSAMPLE_MERGING_SUPPORTED  /* Fast path for sloppy upsampling? */
-#define QUANT_1PASS_SUPPORTED	    /* 1-pass color quantization? */
-#define QUANT_2PASS_SUPPORTED	    /* 2-pass color quantization? */
+#define QUANT_1PASS_SUPPORTED       /* 1-pass color quantization? */
+#define QUANT_2PASS_SUPPORTED       /* 2-pass color quantization? */
 
 /* more capability options later, no doubt */
 
 
 /*
- * Ordering of RGB data in scanlines passed to or from the application.
- * If your application wants to deal with data in the order B,G,R, just
- * change these macros.  You can also deal with formats such as R,G,B,X
- * (one extra byte per pixel) by changing RGB_PIXELSIZE.  Note that changing
- * the offsets will also change the order in which colormap data is organized.
- * RESTRICTIONS:
- * 1. The sample applications cjpeg,djpeg do NOT support modified RGB formats.
- * 2. These macros only affect RGB<=>YCbCr color conversion, so they are not
- *    useful if you are using JPEG color spaces other than YCbCr or grayscale.
- * 3. The color quantizer modules will not behave desirably if RGB_PIXELSIZE
- *    is not 3 (they don't understand about dummy color components!).  So you
- *    can't use color quantization if you change that value.
- */
-
-#define RGB_RED		0	/* Offset of Red in an RGB scanline element */
-#define RGB_GREEN	1	/* Offset of Green */
-#define RGB_BLUE	2	/* Offset of Blue */
-#define RGB_PIXELSIZE	3	/* JSAMPLEs per RGB scanline element */
+ * The RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros are a vestigial
+ * feature of libjpeg.  The idea was that, if an application developer needed
+ * to compress from/decompress to a BGR/BGRX/RGBX/XBGR/XRGB buffer, they could
+ * change these macros, rebuild libjpeg, and link their application statically
+ * with it.  In reality, few people ever did this, because there were some
+ * severe restrictions involved (cjpeg and djpeg no longer worked properly,
+ * compressing/decompressing RGB JPEGs no longer worked properly, and the color
+ * quantizer wouldn't work with pixel sizes other than 3.)  Further, since all
+ * of the O/S-supplied versions of libjpeg were built with the default values
+ * of RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE, many applications have
+ * come to regard these values as immutable.
+ *
+ * The libjpeg-turbo colorspace extensions provide a much cleaner way of
+ * compressing from/decompressing to buffers with arbitrary component orders
+ * and pixel sizes.  Thus, we do not support changing the values of RGB_RED,
+ * RGB_GREEN, RGB_BLUE, or RGB_PIXELSIZE.  In addition to the restrictions
+ * listed above, changing these values will also break the SIMD extensions and
+ * the regression tests.
+ */
+
+#define RGB_RED         0       /* Offset of Red in an RGB scanline element */
+#define RGB_GREEN       1       /* Offset of Green */
+#define RGB_BLUE        2       /* Offset of Blue */
+#define RGB_PIXELSIZE   3       /* JSAMPLEs per RGB scanline element */
 
-#define JPEG_NUMCS 16
+#define JPEG_NUMCS 17
 
 #define EXT_RGB_RED        0
 #define EXT_RGB_GREEN      1
@@ -348,25 +368,29 @@
 static const int rgb_red[JPEG_NUMCS] = {
   -1, -1, RGB_RED, -1, -1, -1, EXT_RGB_RED, EXT_RGBX_RED,
   EXT_BGR_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
-  EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED
+  EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
+  -1
 };
 
 static const int rgb_green[JPEG_NUMCS] = {
   -1, -1, RGB_GREEN, -1, -1, -1, EXT_RGB_GREEN, EXT_RGBX_GREEN,
   EXT_BGR_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
-  EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN
+  EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
+  -1
 };
 
 static const int rgb_blue[JPEG_NUMCS] = {
   -1, -1, RGB_BLUE, -1, -1, -1, EXT_RGB_BLUE, EXT_RGBX_BLUE,
   EXT_BGR_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
-  EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE
+  EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
+  -1
 };
 
 static const int rgb_pixelsize[JPEG_NUMCS] = {
   -1, -1, RGB_PIXELSIZE, -1, -1, -1, EXT_RGB_PIXELSIZE, EXT_RGBX_PIXELSIZE,
   EXT_BGR_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
-  EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE
+  EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
+  -1
 };
 
 /* Definitions for speed-related optimizations. */
@@ -378,7 +402,7 @@
 
 #ifndef MULTIPLIER
 #ifndef WITH_SIMD
-#define MULTIPLIER  int		/* type for fastest integer multiply */
+#define MULTIPLIER  int         /* type for fastest integer multiply */
 #else
 #define MULTIPLIER short  /* prefer 16-bit with SIMD for parellelism */
 #endif
@@ -388,17 +412,10 @@
 /* FAST_FLOAT should be either float or double, whichever is done faster
  * by your compiler.  (Note that this type is only used in the floating point
  * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.)
- * Typically, float is faster in ANSI C compilers, while double is faster in
- * pre-ANSI compilers (because they insist on converting to double anyway).
- * The code below therefore chooses float if we have ANSI-style prototypes.
  */
 
 #ifndef FAST_FLOAT
-#ifdef HAVE_PROTOTYPES
 #define FAST_FLOAT  float
-#else
-#define FAST_FLOAT  double
-#endif
 #endif
 
 #endif /* JPEG_INTERNAL_OPTIONS */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegcomp.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegcomp.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegcomp.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegcomp.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,8 @@
  * jpegcomp.h
  *
  * Copyright (C) 2010, D. R. Commander
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * JPEG compatibility macros
  * These declarations are considered internal to the JPEG library; most
@@ -11,6 +12,8 @@
 
 #if JPEG_LIB_VERSION >= 70
 #define _DCT_scaled_size DCT_h_scaled_size
+#define _DCT_h_scaled_size DCT_h_scaled_size
+#define _DCT_v_scaled_size DCT_v_scaled_size
 #define _min_DCT_scaled_size min_DCT_h_scaled_size
 #define _min_DCT_h_scaled_size min_DCT_h_scaled_size
 #define _min_DCT_v_scaled_size min_DCT_v_scaled_size
@@ -18,6 +21,8 @@
 #define _jpeg_height jpeg_height
 #else
 #define _DCT_scaled_size DCT_scaled_size
+#define _DCT_h_scaled_size DCT_scaled_size
+#define _DCT_v_scaled_size DCT_scaled_size
 #define _min_DCT_scaled_size min_DCT_scaled_size
 #define _min_DCT_h_scaled_size min_DCT_scaled_size
 #define _min_DCT_v_scaled_size min_DCT_scaled_size
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegint.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegint.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegint.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegint.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jpegint.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * Modified 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2016, D. R. Commander
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file provides common declarations for the various JPEG modules.
  * These declarations are considered internal to the JPEG library; most
@@ -14,121 +18,127 @@
 
 /* Declarations for both compression & decompression */
 
-typedef enum {			/* Operating modes for buffer controllers */
-	JBUF_PASS_THRU,		/* Plain stripwise operation */
-	/* Remaining modes require a full-image buffer to have been created */
-	JBUF_SAVE_SOURCE,	/* Run source subobject only, save output */
-	JBUF_CRANK_DEST,	/* Run dest subobject only, using saved data */
-	JBUF_SAVE_AND_PASS	/* Run both subobjects, save output */
+typedef enum {            /* Operating modes for buffer controllers */
+  JBUF_PASS_THRU,         /* Plain stripwise operation */
+  /* Remaining modes require a full-image buffer to have been created */
+  JBUF_SAVE_SOURCE,       /* Run source subobject only, save output */
+  JBUF_CRANK_DEST,        /* Run dest subobject only, using saved data */
+  JBUF_SAVE_AND_PASS      /* Run both subobjects, save output */
 } J_BUF_MODE;
 
 /* Values of global_state field (jdapi.c has some dependencies on ordering!) */
-#define CSTATE_START	100	/* after create_compress */
-#define CSTATE_SCANNING	101	/* start_compress done, write_scanlines OK */
-#define CSTATE_RAW_OK	102	/* start_compress done, write_raw_data OK */
-#define CSTATE_WRCOEFS	103	/* jpeg_write_coefficients done */
-#define DSTATE_START	200	/* after create_decompress */
-#define DSTATE_INHEADER	201	/* reading header markers, no SOS yet */
-#define DSTATE_READY	202	/* found SOS, ready for start_decompress */
-#define DSTATE_PRELOAD	203	/* reading multiscan file in start_decompress*/
-#define DSTATE_PRESCAN	204	/* performing dummy pass for 2-pass quant */
-#define DSTATE_SCANNING	205	/* start_decompress done, read_scanlines OK */
-#define DSTATE_RAW_OK	206	/* start_decompress done, read_raw_data OK */
-#define DSTATE_BUFIMAGE	207	/* expecting jpeg_start_output */
-#define DSTATE_BUFPOST	208	/* looking for SOS/EOI in jpeg_finish_output */
-#define DSTATE_RDCOEFS	209	/* reading file in jpeg_read_coefficients */
-#define DSTATE_STOPPING	210	/* looking for EOI in jpeg_finish_decompress */
+#define CSTATE_START    100     /* after create_compress */
+#define CSTATE_SCANNING 101     /* start_compress done, write_scanlines OK */
+#define CSTATE_RAW_OK   102     /* start_compress done, write_raw_data OK */
+#define CSTATE_WRCOEFS  103     /* jpeg_write_coefficients done */
+#define DSTATE_START    200     /* after create_decompress */
+#define DSTATE_INHEADER 201     /* reading header markers, no SOS yet */
+#define DSTATE_READY    202     /* found SOS, ready for start_decompress */
+#define DSTATE_PRELOAD  203     /* reading multiscan file in start_decompress*/
+#define DSTATE_PRESCAN  204     /* performing dummy pass for 2-pass quant */
+#define DSTATE_SCANNING 205     /* start_decompress done, read_scanlines OK */
+#define DSTATE_RAW_OK   206     /* start_decompress done, read_raw_data OK */
+#define DSTATE_BUFIMAGE 207     /* expecting jpeg_start_output */
+#define DSTATE_BUFPOST  208     /* looking for SOS/EOI in jpeg_finish_output */
+#define DSTATE_RDCOEFS  209     /* reading file in jpeg_read_coefficients */
+#define DSTATE_STOPPING 210     /* looking for EOI in jpeg_finish_decompress */
+
+
+/* JLONG must hold at least signed 32-bit values. */
+typedef long JLONG;
+
+
+/*
+ * Left shift macro that handles a negative operand without causing any
+ * sanitizer warnings
+ */
+
+#define LEFT_SHIFT(a, b) ((JLONG)((unsigned long)(a) << (b)))
 
 
 /* Declarations for compression modules */
 
 /* Master control module */
 struct jpeg_comp_master {
-  JMETHOD(void, prepare_for_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, pass_startup, (j_compress_ptr cinfo));
-  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+  void (*prepare_for_pass) (j_compress_ptr cinfo);
+  void (*pass_startup) (j_compress_ptr cinfo);
+  void (*finish_pass) (j_compress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean call_pass_startup;	/* True if pass_startup must be called */
-  boolean is_last_pass;		/* True during last pass */
+  boolean call_pass_startup;    /* True if pass_startup must be called */
+  boolean is_last_pass;         /* True during last pass */
 };
 
 /* Main buffer control (downsampled-data buffer) */
 struct jpeg_c_main_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, process_data, (j_compress_ptr cinfo,
-			       JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			       JDIMENSION in_rows_avail));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                        JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail);
 };
 
 /* Compression preprocessing (downsampling input buffer control) */
 struct jpeg_c_prep_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, pre_process_data, (j_compress_ptr cinfo,
-				   JSAMPARRAY input_buf,
-				   JDIMENSION *in_row_ctr,
-				   JDIMENSION in_rows_avail,
-				   JSAMPIMAGE output_buf,
-				   JDIMENSION *out_row_group_ctr,
-				   JDIMENSION out_row_groups_avail));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*pre_process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                            JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail,
+                            JSAMPIMAGE output_buf,
+                            JDIMENSION *out_row_group_ctr,
+                            JDIMENSION out_row_groups_avail);
 };
 
 /* Coefficient buffer control */
 struct jpeg_c_coef_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(boolean, compress_data, (j_compress_ptr cinfo,
-				   JSAMPIMAGE input_buf));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  boolean (*compress_data) (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 };
 
 /* Colorspace conversion */
 struct jpeg_color_converter {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, color_convert, (j_compress_ptr cinfo,
-				JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-				JDIMENSION output_row, int num_rows));
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*color_convert) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JSAMPIMAGE output_buf, JDIMENSION output_row,
+                         int num_rows);
 };
 
 /* Downsampling */
 struct jpeg_downsampler {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, downsample, (j_compress_ptr cinfo,
-			     JSAMPIMAGE input_buf, JDIMENSION in_row_index,
-			     JSAMPIMAGE output_buf,
-			     JDIMENSION out_row_group_index));
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*downsample) (j_compress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION in_row_index, JSAMPIMAGE output_buf,
+                      JDIMENSION out_row_group_index);
 
-  boolean need_context_rows;	/* TRUE if need rows above & below */
+  boolean need_context_rows;    /* TRUE if need rows above & below */
 };
 
 /* Forward DCT (also controls coefficient quantization) */
 struct jpeg_forward_dct {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  void (*start_pass) (j_compress_ptr cinfo);
   /* perhaps this should be an array??? */
-  JMETHOD(void, forward_DCT, (j_compress_ptr cinfo,
-			      jpeg_component_info * compptr,
-			      JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-			      JDIMENSION start_row, JDIMENSION start_col,
-			      JDIMENSION num_blocks));
+  void (*forward_DCT) (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                       JDIMENSION start_row, JDIMENSION start_col,
+                       JDIMENSION num_blocks);
 };
 
 /* Entropy encoding */
 struct jpeg_entropy_encoder {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, boolean gather_statistics));
-  JMETHOD(boolean, encode_mcu, (j_compress_ptr cinfo, JBLOCKROW *MCU_data));
-  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+  void (*start_pass) (j_compress_ptr cinfo, boolean gather_statistics);
+  boolean (*encode_mcu) (j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+  void (*finish_pass) (j_compress_ptr cinfo);
 };
 
 /* Marker writing */
 struct jpeg_marker_writer {
-  JMETHOD(void, write_file_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_frame_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_scan_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_file_trailer, (j_compress_ptr cinfo));
-  JMETHOD(void, write_tables_only, (j_compress_ptr cinfo));
+  void (*write_file_header) (j_compress_ptr cinfo);
+  void (*write_frame_header) (j_compress_ptr cinfo);
+  void (*write_scan_header) (j_compress_ptr cinfo);
+  void (*write_file_trailer) (j_compress_ptr cinfo);
+  void (*write_tables_only) (j_compress_ptr cinfo);
   /* These routines are exported to allow insertion of extra markers */
   /* Probably only COM and APPn markers should be written this way */
-  JMETHOD(void, write_marker_header, (j_compress_ptr cinfo, int marker,
-				      unsigned int datalen));
-  JMETHOD(void, write_marker_byte, (j_compress_ptr cinfo, int val));
+  void (*write_marker_header) (j_compress_ptr cinfo, int marker,
+                               unsigned int datalen);
+  void (*write_marker_byte) (j_compress_ptr cinfo, int val);
 };
 
 
@@ -136,138 +146,137 @@
 
 /* Master control module */
 struct jpeg_decomp_master {
-  JMETHOD(void, prepare_for_output_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, finish_output_pass, (j_decompress_ptr cinfo));
+  void (*prepare_for_output_pass) (j_decompress_ptr cinfo);
+  void (*finish_output_pass) (j_decompress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean is_dummy_pass;	/* True during 1st pass for 2-pass quant */
+  boolean is_dummy_pass;        /* True during 1st pass for 2-pass quant */
+
+  /* Partial decompression variables */
+  JDIMENSION first_iMCU_col;
+  JDIMENSION last_iMCU_col;
+  JDIMENSION first_MCU_col[MAX_COMPS_IN_SCAN];
+  JDIMENSION last_MCU_col[MAX_COMPS_IN_SCAN];
+  boolean jinit_upsampler_no_alloc;
 };
 
 /* Input control module */
 struct jpeg_input_controller {
-  JMETHOD(int, consume_input, (j_decompress_ptr cinfo));
-  JMETHOD(void, reset_input_controller, (j_decompress_ptr cinfo));
-  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, finish_input_pass, (j_decompress_ptr cinfo));
+  int (*consume_input) (j_decompress_ptr cinfo);
+  void (*reset_input_controller) (j_decompress_ptr cinfo);
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  void (*finish_input_pass) (j_decompress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean has_multiple_scans;	/* True if file has multiple scans */
-  boolean eoi_reached;		/* True when EOI has been consumed */
+  boolean has_multiple_scans;   /* True if file has multiple scans */
+  boolean eoi_reached;          /* True when EOI has been consumed */
 };
 
 /* Main buffer control (downsampled-data buffer) */
 struct jpeg_d_main_controller {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, process_data, (j_decompress_ptr cinfo,
-			       JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			       JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                        JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 };
 
 /* Coefficient buffer control */
 struct jpeg_d_coef_controller {
-  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
-  JMETHOD(int, consume_data, (j_decompress_ptr cinfo));
-  JMETHOD(void, start_output_pass, (j_decompress_ptr cinfo));
-  JMETHOD(int, decompress_data, (j_decompress_ptr cinfo,
-				 JSAMPIMAGE output_buf));
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  int (*consume_data) (j_decompress_ptr cinfo);
+  void (*start_output_pass) (j_decompress_ptr cinfo);
+  int (*decompress_data) (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
   /* Pointer to array of coefficient virtual arrays, or NULL if none */
   jvirt_barray_ptr *coef_arrays;
 };
 
 /* Decompression postprocessing (color quantization buffer control) */
 struct jpeg_d_post_controller {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, post_process_data, (j_decompress_ptr cinfo,
-				    JSAMPIMAGE input_buf,
-				    JDIMENSION *in_row_group_ctr,
-				    JDIMENSION in_row_groups_avail,
-				    JSAMPARRAY output_buf,
-				    JDIMENSION *out_row_ctr,
-				    JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*post_process_data) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION *in_row_group_ctr,
+                             JDIMENSION in_row_groups_avail,
+                             JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                             JDIMENSION out_rows_avail);
 };
 
 /* Marker reading & parsing */
 struct jpeg_marker_reader {
-  JMETHOD(void, reset_marker_reader, (j_decompress_ptr cinfo));
+  void (*reset_marker_reader) (j_decompress_ptr cinfo);
   /* Read markers until SOS or EOI.
    * Returns same codes as are defined for jpeg_consume_input:
    * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
    */
-  JMETHOD(int, read_markers, (j_decompress_ptr cinfo));
+  int (*read_markers) (j_decompress_ptr cinfo);
   /* Read a restart marker --- exported for use by entropy decoder only */
   jpeg_marker_parser_method read_restart_marker;
 
   /* State of marker reader --- nominally internal, but applications
    * supplying COM or APPn handlers might like to know the state.
    */
-  boolean saw_SOI;		/* found SOI? */
-  boolean saw_SOF;		/* found SOF? */
-  int next_restart_num;		/* next restart number expected (0-7) */
-  unsigned int discarded_bytes;	/* # of bytes skipped looking for a marker */
+  boolean saw_SOI;              /* found SOI? */
+  boolean saw_SOF;              /* found SOF? */
+  int next_restart_num;         /* next restart number expected (0-7) */
+  unsigned int discarded_bytes; /* # of bytes skipped looking for a marker */
 };
 
 /* Entropy decoding */
 struct jpeg_entropy_decoder {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(boolean, decode_mcu, (j_decompress_ptr cinfo,
-				JBLOCKROW *MCU_data));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  boolean (*decode_mcu) (j_decompress_ptr cinfo, JBLOCKROW *MCU_data);
 
   /* This is here to share code between baseline and progressive decoders; */
   /* other modules probably should not use it */
-  boolean insufficient_data;	/* set TRUE after emitting warning */
+  boolean insufficient_data;    /* set TRUE after emitting warning */
 };
 
 /* Inverse DCT (also performs dequantization) */
-typedef JMETHOD(void, inverse_DCT_method_ptr,
-		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col));
+typedef void (*inverse_DCT_method_ptr) (j_decompress_ptr cinfo,
+                                        jpeg_component_info *compptr,
+                                        JCOEFPTR coef_block,
+                                        JSAMPARRAY output_buf,
+                                        JDIMENSION output_col);
 
 struct jpeg_inverse_dct {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  void (*start_pass) (j_decompress_ptr cinfo);
   /* It is useful to allow each component to have a separate IDCT method. */
   inverse_DCT_method_ptr inverse_DCT[MAX_COMPONENTS];
 };
 
 /* Upsampling (note that upsampler must also call color converter) */
 struct jpeg_upsampler {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, upsample, (j_decompress_ptr cinfo,
-			   JSAMPIMAGE input_buf,
-			   JDIMENSION *in_row_group_ctr,
-			   JDIMENSION in_row_groups_avail,
-			   JSAMPARRAY output_buf,
-			   JDIMENSION *out_row_ctr,
-			   JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*upsample) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                    JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 
-  boolean need_context_rows;	/* TRUE if need rows above & below */
+  boolean need_context_rows;    /* TRUE if need rows above & below */
 };
 
 /* Colorspace conversion */
 struct jpeg_color_deconverter {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, color_convert, (j_decompress_ptr cinfo,
-				JSAMPIMAGE input_buf, JDIMENSION input_row,
-				JSAMPARRAY output_buf, int num_rows));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows);
 };
 
 /* Color quantization or color precision reduction */
 struct jpeg_color_quantizer {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, boolean is_pre_scan));
-  JMETHOD(void, color_quantize, (j_decompress_ptr cinfo,
-				 JSAMPARRAY input_buf, JSAMPARRAY output_buf,
-				 int num_rows));
-  JMETHOD(void, finish_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, new_color_map, (j_decompress_ptr cinfo));
+  void (*start_pass) (j_decompress_ptr cinfo, boolean is_pre_scan);
+  void (*color_quantize) (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                          JSAMPARRAY output_buf, int num_rows);
+  void (*finish_pass) (j_decompress_ptr cinfo);
+  void (*new_color_map) (j_decompress_ptr cinfo);
 };
 
 
 /* Miscellaneous useful macros */
 
 #undef MAX
-#define MAX(a,b)	((a) > (b) ? (a) : (b))
+#define MAX(a,b)        ((a) > (b) ? (a) : (b))
 #undef MIN
-#define MIN(a,b)	((a) < (b) ? (a) : (b))
+#define MIN(a,b)        ((a) < (b) ? (a) : (b))
 
 
 /* We assume that right shift corresponds to signed division by 2 with
@@ -275,126 +284,84 @@
  * shift" instructions that shift in copies of the sign bit.  But some
  * C compilers implement >> with an unsigned shift.  For these machines you
  * must define RIGHT_SHIFT_IS_UNSIGNED.
- * RIGHT_SHIFT provides a proper signed right shift of an INT32 quantity.
+ * RIGHT_SHIFT provides a proper signed right shift of a JLONG quantity.
  * It is only applied with constant shift counts.  SHIFT_TEMPS must be
  * included in the variables of any routine using RIGHT_SHIFT.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define SHIFT_TEMPS	INT32 shift_temp;
+#define SHIFT_TEMPS     JLONG shift_temp;
 #define RIGHT_SHIFT(x,shft)  \
-	((shift_temp = (x)) < 0 ? \
-	 (shift_temp >> (shft)) | ((~((INT32) 0)) << (32-(shft))) : \
-	 (shift_temp >> (shft)))
+        ((shift_temp = (x)) < 0 ? \
+         (shift_temp >> (shft)) | ((~((JLONG) 0)) << (32-(shft))) : \
+         (shift_temp >> (shft)))
 #else
 #define SHIFT_TEMPS
-#define RIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define RIGHT_SHIFT(x,shft)     ((x) >> (shft))
 #endif
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jinit_compress_master	jICompress
-#define jinit_c_master_control	jICMaster
-#define jinit_c_main_controller	jICMainC
-#define jinit_c_prep_controller	jICPrepC
-#define jinit_c_coef_controller	jICCoefC
-#define jinit_color_converter	jICColor
-#define jinit_downsampler	jIDownsampler
-#define jinit_forward_dct	jIFDCT
-#define jinit_huff_encoder	jIHEncoder
-#define jinit_phuff_encoder	jIPHEncoder
-#define jinit_arith_encoder	jIAEncoder
-#define jinit_marker_writer	jIMWriter
-#define jinit_master_decompress	jIDMaster
-#define jinit_d_main_controller	jIDMainC
-#define jinit_d_coef_controller	jIDCoefC
-#define jinit_d_post_controller	jIDPostC
-#define jinit_input_controller	jIInCtlr
-#define jinit_marker_reader	jIMReader
-#define jinit_huff_decoder	jIHDecoder
-#define jinit_phuff_decoder	jIPHDecoder
-#define jinit_arith_decoder	jIADecoder
-#define jinit_inverse_dct	jIIDCT
-#define jinit_upsampler		jIUpsampler
-#define jinit_color_deconverter	jIDColor
-#define jinit_1pass_quantizer	jI1Quant
-#define jinit_2pass_quantizer	jI2Quant
-#define jinit_merged_upsampler	jIMUpsampler
-#define jinit_memory_mgr	jIMemMgr
-#define jdiv_round_up		jDivRound
-#define jround_up		jRound
-#define jcopy_sample_rows	jCopySamples
-#define jcopy_block_row		jCopyBlocks
-#define jzero_far		jZeroFar
-#define jpeg_zigzag_order	jZIGTable
-#define jpeg_natural_order	jZAGTable
-#define jpeg_aritab		jAriTab
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-
 /* Compression module initialization routines */
-EXTERN(void) jinit_compress_master JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_c_master_control JPP((j_compress_ptr cinfo,
-					 boolean transcode_only));
-EXTERN(void) jinit_c_main_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_c_prep_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_c_coef_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_color_converter JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_downsampler JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_forward_dct JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_huff_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_phuff_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_arith_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_marker_writer JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_compress_master (j_compress_ptr cinfo);
+EXTERN(void) jinit_c_master_control (j_compress_ptr cinfo,
+                                     boolean transcode_only);
+EXTERN(void) jinit_c_main_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_c_prep_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_c_coef_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_color_converter (j_compress_ptr cinfo);
+EXTERN(void) jinit_downsampler (j_compress_ptr cinfo);
+EXTERN(void) jinit_forward_dct (j_compress_ptr cinfo);
+EXTERN(void) jinit_huff_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_phuff_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_arith_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_marker_writer (j_compress_ptr cinfo);
 /* Decompression module initialization routines */
-EXTERN(void) jinit_master_decompress JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_d_main_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_d_coef_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_d_post_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_input_controller JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_marker_reader JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_huff_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_phuff_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_arith_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_inverse_dct JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_upsampler JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_color_deconverter JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_1pass_quantizer JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_2pass_quantizer JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_merged_upsampler JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_master_decompress (j_decompress_ptr cinfo);
+EXTERN(void) jinit_d_main_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_d_coef_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_d_post_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_input_controller (j_decompress_ptr cinfo);
+EXTERN(void) jinit_marker_reader (j_decompress_ptr cinfo);
+EXTERN(void) jinit_huff_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_phuff_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_arith_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_inverse_dct (j_decompress_ptr cinfo);
+EXTERN(void) jinit_upsampler (j_decompress_ptr cinfo);
+EXTERN(void) jinit_color_deconverter (j_decompress_ptr cinfo);
+EXTERN(void) jinit_1pass_quantizer (j_decompress_ptr cinfo);
+EXTERN(void) jinit_2pass_quantizer (j_decompress_ptr cinfo);
+EXTERN(void) jinit_merged_upsampler (j_decompress_ptr cinfo);
 /* Memory manager initialization */
-EXTERN(void) jinit_memory_mgr JPP((j_common_ptr cinfo));
+EXTERN(void) jinit_memory_mgr (j_common_ptr cinfo);
 
 /* Utility routines in jutils.c */
-EXTERN(long) jdiv_round_up JPP((long a, long b));
-EXTERN(long) jround_up JPP((long a, long b));
-EXTERN(void) jcopy_sample_rows JPP((JSAMPARRAY input_array, int source_row,
-				    JSAMPARRAY output_array, int dest_row,
-				    int num_rows, JDIMENSION num_cols));
-EXTERN(void) jcopy_block_row JPP((JBLOCKROW input_row, JBLOCKROW output_row,
-				  JDIMENSION num_blocks));
-EXTERN(void) jzero_far JPP((void FAR * target, size_t bytestozero));
+EXTERN(long) jdiv_round_up (long a, long b);
+EXTERN(long) jround_up (long a, long b);
+EXTERN(void) jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
+                                JSAMPARRAY output_array, int dest_row,
+                                int num_rows, JDIMENSION num_cols);
+EXTERN(void) jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
+                              JDIMENSION num_blocks);
+EXTERN(void) jzero_far (void *target, size_t bytestozero);
 /* Constant tables in jutils.c */
-#if 0				/* This table is not actually needed in v6a */
+#if 0                           /* This table is not actually needed in v6a */
 extern const int jpeg_zigzag_order[]; /* natural coef order to zigzag order */
 #endif
 extern const int jpeg_natural_order[]; /* zigzag coef order to natural order */
 
 /* Arithmetic coding probability estimation tables in jaricom.c */
-extern const INT32 jpeg_aritab[];
+extern const JLONG jpeg_aritab[];
 
 /* Suppress undefined-structure complaints if necessary. */
 
 #ifdef INCOMPLETE_TYPES_BROKEN
-#ifndef AM_MEMORY_MANAGER	/* only jmemmgr.c defines these */
+#ifndef AM_MEMORY_MANAGER       /* only jmemmgr.c defines these */
 struct jvirt_sarray_control { long dummy; };
 struct jvirt_barray_control { long dummy; };
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeglib.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeglib.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeglib.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeglib.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jpeglib.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
  * Modified 2002-2009 by Guido Vollbeding.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2013-2014, 2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file defines the application interface for the JPEG library.
  * Most applications using the library need only include this file,
@@ -15,6 +18,10 @@
 #ifndef JPEGLIB_H
 #define JPEGLIB_H
 
+/* Begin chromium edits */
+#include "jpeglibmangler.h"
+/* End chromium edits */
+
 /*
  * First we include the configuration files that record how this
  * installation of the JPEG library is set up.  jconfig.h can be
@@ -22,10 +29,10 @@
  * manual configuration options that most people need not worry about.
  */
 
-#ifndef JCONFIG_INCLUDED	/* in case jinclude.h already did */
-#include "jconfig.h"		/* widely used configuration options */
+#ifndef JCONFIG_INCLUDED        /* in case jinclude.h already did */
+#include "jconfig.h"            /* widely used configuration options */
 #endif
-#include "jmorecfg.h"		/* seldom changed options */
+#include "jmorecfg.h"           /* seldom changed options */
 
 
 #ifdef __cplusplus
@@ -40,13 +47,13 @@
  * if you want to be compatible.
  */
 
-#define DCTSIZE		    8	/* The basic DCT block is 8x8 samples */
-#define DCTSIZE2	    64	/* DCTSIZE squared; # of elements in a block */
-#define NUM_QUANT_TBLS      4	/* Quantization tables are numbered 0..3 */
-#define NUM_HUFF_TBLS       4	/* Huffman tables are numbered 0..3 */
-#define NUM_ARITH_TBLS      16	/* Arith-coding tables are numbered 0..15 */
-#define MAX_COMPS_IN_SCAN   4	/* JPEG limit on # of components in one scan */
-#define MAX_SAMP_FACTOR     4	/* JPEG limit on sampling factors */
+#define DCTSIZE             8   /* The basic DCT block is 8x8 samples */
+#define DCTSIZE2            64  /* DCTSIZE squared; # of elements in a block */
+#define NUM_QUANT_TBLS      4   /* Quantization tables are numbered 0..3 */
+#define NUM_HUFF_TBLS       4   /* Huffman tables are numbered 0..3 */
+#define NUM_ARITH_TBLS      16  /* Arith-coding tables are numbered 0..15 */
+#define MAX_COMPS_IN_SCAN   4   /* JPEG limit on # of components in one scan */
+#define MAX_SAMP_FACTOR     4   /* JPEG limit on sampling factors */
 /* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard;
  * the PostScript DCT filter can emit files with many more than 10 blocks/MCU.
  * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU
@@ -61,20 +68,18 @@
 
 
 /* Data structures for images (arrays of samples and of DCT coefficients).
- * On 80x86 machines, the image arrays are too big for near pointers,
- * but the pointer arrays can fit in near memory.
  */
 
-typedef JSAMPLE FAR *JSAMPROW;	/* ptr to one image row of pixel samples. */
-typedef JSAMPROW *JSAMPARRAY;	/* ptr to some rows (a 2-D sample array) */
-typedef JSAMPARRAY *JSAMPIMAGE;	/* a 3-D sample array: top index is color */
-
-typedef JCOEF JBLOCK[DCTSIZE2];	/* one block of coefficients */
-typedef JBLOCK FAR *JBLOCKROW;	/* pointer to one row of coefficient blocks */
-typedef JBLOCKROW *JBLOCKARRAY;		/* a 2-D array of coefficient blocks */
-typedef JBLOCKARRAY *JBLOCKIMAGE;	/* a 3-D array of coefficient blocks */
+typedef JSAMPLE *JSAMPROW;      /* ptr to one image row of pixel samples. */
+typedef JSAMPROW *JSAMPARRAY;   /* ptr to some rows (a 2-D sample array) */
+typedef JSAMPARRAY *JSAMPIMAGE; /* a 3-D sample array: top index is color */
+
+typedef JCOEF JBLOCK[DCTSIZE2]; /* one block of coefficients */
+typedef JBLOCK *JBLOCKROW;      /* pointer to one row of coefficient blocks */
+typedef JBLOCKROW *JBLOCKARRAY;         /* a 2-D array of coefficient blocks */
+typedef JBLOCKARRAY *JBLOCKIMAGE;       /* a 3-D array of coefficient blocks */
 
-typedef JCOEF FAR *JCOEFPTR;	/* useful in a couple of places */
+typedef JCOEF *JCOEFPTR;        /* useful in a couple of places */
 
 
 /* Types for JPEG compression parameters and working tables. */
@@ -87,13 +92,13 @@
    * (not the zigzag order in which they are stored in a JPEG DQT marker).
    * CAUTION: IJG versions prior to v6a kept this array in zigzag order.
    */
-  UINT16 quantval[DCTSIZE2];	/* quantization step for each coefficient */
+  UINT16 quantval[DCTSIZE2];    /* quantization step for each coefficient */
   /* This field is used only during compression.  It's initialized FALSE when
    * the table is created, and set TRUE when it's been output to the file.
    * You could suppress output of a table by setting this to TRUE.
    * (See jpeg_suppress_tables for an example.)
    */
-  boolean sent_table;		/* TRUE when table has been output */
+  boolean sent_table;           /* TRUE when table has been output */
 } JQUANT_TBL;
 
 
@@ -101,15 +106,15 @@
 
 typedef struct {
   /* These two fields directly represent the contents of a JPEG DHT marker */
-  UINT8 bits[17];		/* bits[k] = # of symbols with codes of */
-				/* length k bits; bits[0] is unused */
-  UINT8 huffval[256];		/* The symbols, in order of incr code length */
+  UINT8 bits[17];               /* bits[k] = # of symbols with codes of */
+                                /* length k bits; bits[0] is unused */
+  UINT8 huffval[256];           /* The symbols, in order of incr code length */
   /* This field is used only during compression.  It's initialized FALSE when
    * the table is created, and set TRUE when it's been output to the file.
    * You could suppress output of a table by setting this to TRUE.
    * (See jpeg_suppress_tables for an example.)
    */
-  boolean sent_table;		/* TRUE when table has been output */
+  boolean sent_table;           /* TRUE when table has been output */
 } JHUFF_TBL;
 
 
@@ -119,20 +124,20 @@
   /* These values are fixed over the whole image. */
   /* For compression, they must be supplied by parameter setup; */
   /* for decompression, they are read from the SOF marker. */
-  int component_id;		/* identifier for this component (0..255) */
-  int component_index;		/* its index in SOF or cinfo->comp_info[] */
-  int h_samp_factor;		/* horizontal sampling factor (1..4) */
-  int v_samp_factor;		/* vertical sampling factor (1..4) */
-  int quant_tbl_no;		/* quantization table selector (0..3) */
+  int component_id;             /* identifier for this component (0..255) */
+  int component_index;          /* its index in SOF or cinfo->comp_info[] */
+  int h_samp_factor;            /* horizontal sampling factor (1..4) */
+  int v_samp_factor;            /* vertical sampling factor (1..4) */
+  int quant_tbl_no;             /* quantization table selector (0..3) */
   /* These values may vary between scans. */
   /* For compression, they must be supplied by parameter setup; */
   /* for decompression, they are read from the SOS marker. */
   /* The decompressor output side may not use these variables. */
-  int dc_tbl_no;		/* DC entropy table selector (0..3) */
-  int ac_tbl_no;		/* AC entropy table selector (0..3) */
-  
+  int dc_tbl_no;                /* DC entropy table selector (0..3) */
+  int ac_tbl_no;                /* AC entropy table selector (0..3) */
+
   /* Remaining fields should be treated as private by applications. */
-  
+
   /* These values are computed during compression or decompression startup: */
   /* Component's size in DCT blocks.
    * Any dummy blocks added to complete an MCU are not counted; therefore
@@ -143,8 +148,8 @@
   /* Size of a DCT block in samples.  Always DCTSIZE for compression.
    * For decompression this is the size of the output from one DCT block,
    * reflecting any scaling we choose to apply during the IDCT step.
-   * Values of 1,2,4,8 are likely to be supported.  Note that different
-   * components may receive different IDCT scalings.
+   * Values from 1 to 16 are supported.
+   * Note that different components may receive different IDCT scalings.
    */
 #if JPEG_LIB_VERSION >= 70
   int DCT_h_scaled_size;
@@ -158,53 +163,53 @@
    * and similarly for height.  For decompression, IDCT scaling is included, so
    * downsampled_width = ceil(image_width * Hi/Hmax * DCT_[h_]scaled_size/DCTSIZE)
    */
-  JDIMENSION downsampled_width;	 /* actual width in samples */
+  JDIMENSION downsampled_width;  /* actual width in samples */
   JDIMENSION downsampled_height; /* actual height in samples */
   /* This flag is used only for decompression.  In cases where some of the
    * components will be ignored (eg grayscale output from YCbCr image),
    * we can skip most computations for the unused components.
    */
-  boolean component_needed;	/* do we need the value of this component? */
+  boolean component_needed;     /* do we need the value of this component? */
 
   /* These values are computed before starting a scan of the component. */
   /* The decompressor output side may not use these variables. */
-  int MCU_width;		/* number of blocks per MCU, horizontally */
-  int MCU_height;		/* number of blocks per MCU, vertically */
-  int MCU_blocks;		/* MCU_width * MCU_height */
-  int MCU_sample_width;		/* MCU width in samples, MCU_width*DCT_[h_]scaled_size */
-  int last_col_width;		/* # of non-dummy blocks across in last MCU */
-  int last_row_height;		/* # of non-dummy blocks down in last MCU */
+  int MCU_width;                /* number of blocks per MCU, horizontally */
+  int MCU_height;               /* number of blocks per MCU, vertically */
+  int MCU_blocks;               /* MCU_width * MCU_height */
+  int MCU_sample_width;         /* MCU width in samples, MCU_width*DCT_[h_]scaled_size */
+  int last_col_width;           /* # of non-dummy blocks across in last MCU */
+  int last_row_height;          /* # of non-dummy blocks down in last MCU */
 
   /* Saved quantization table for component; NULL if none yet saved.
    * See jdinput.c comments about the need for this information.
    * This field is currently used only for decompression.
    */
-  JQUANT_TBL * quant_table;
+  JQUANT_TBL *quant_table;
 
   /* Private per-component storage for DCT or IDCT subsystem. */
-  void * dct_table;
+  void *dct_table;
 } jpeg_component_info;
 
 
 /* The script for encoding a multiple-scan file is an array of these: */
 
 typedef struct {
-  int comps_in_scan;		/* number of components encoded in this scan */
+  int comps_in_scan;            /* number of components encoded in this scan */
   int component_index[MAX_COMPS_IN_SCAN]; /* their SOF/comp_info[] indexes */
-  int Ss, Se;			/* progressive JPEG spectral selection parms */
-  int Ah, Al;			/* progressive JPEG successive approx. parms */
+  int Ss, Se;                   /* progressive JPEG spectral selection parms */
+  int Ah, Al;                   /* progressive JPEG successive approx. parms */
 } jpeg_scan_info;
 
 /* The decompressor can save APPn and COM markers in a list of these: */
 
-typedef struct jpeg_marker_struct FAR * jpeg_saved_marker_ptr;
+typedef struct jpeg_marker_struct *jpeg_saved_marker_ptr;
 
 struct jpeg_marker_struct {
-  jpeg_saved_marker_ptr next;	/* next in list, or NULL */
-  UINT8 marker;			/* marker code: JPEG_COM, or JPEG_APP0+n */
-  unsigned int original_length;	/* # bytes of data in the file */
-  unsigned int data_length;	/* # bytes of data saved at data[] */
-  JOCTET FAR * data;		/* the data contained in the marker */
+  jpeg_saved_marker_ptr next;   /* next in list, or NULL */
+  UINT8 marker;                 /* marker code: JPEG_COM, or JPEG_APP0+n */
+  unsigned int original_length; /* # bytes of data in the file */
+  unsigned int data_length;     /* # bytes of data saved at data[] */
+  JOCTET *data;                 /* the data contained in the marker */
   /* the marker length word is not counted in data_length or original_length */
 };
 
@@ -214,102 +219,102 @@
 #define JCS_ALPHA_EXTENSIONS 1
 
 typedef enum {
-	JCS_UNKNOWN,		/* error/unspecified */
-	JCS_GRAYSCALE,		/* monochrome */
-	JCS_RGB,		/* red/green/blue as specified by the RGB_RED, RGB_GREEN,
-				   RGB_BLUE, and RGB_PIXELSIZE macros */
-	JCS_YCbCr,		/* Y/Cb/Cr (also known as YUV) */
-	JCS_CMYK,		/* C/M/Y/K */
-	JCS_YCCK,		/* Y/Cb/Cr/K */
-	JCS_EXT_RGB,		/* red/green/blue */
-	JCS_EXT_RGBX,		/* red/green/blue/x */
-	JCS_EXT_BGR,		/* blue/green/red */
-	JCS_EXT_BGRX,		/* blue/green/red/x */
-	JCS_EXT_XBGR,		/* x/blue/green/red */
-	JCS_EXT_XRGB,		/* x/red/green/blue */
-	/* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX,
-	   JCS_EXT_XBGR, or JCS_EXT_XRGB during decompression, the X byte is
-	   undefined, and in order to ensure the best performance,
-	   libjpeg-turbo can set that byte to whatever value it wishes.  Use
-	   the following colorspace constants to ensure that the X byte is set
-	   to 0xFF, so that it can be interpreted as an opaque alpha
-	   channel. */
-	JCS_EXT_RGBA,		/* red/green/blue/alpha */
-	JCS_EXT_BGRA,		/* blue/green/red/alpha */
-	JCS_EXT_ABGR,		/* alpha/blue/green/red */
-	JCS_EXT_ARGB		/* alpha/red/green/blue */
+  JCS_UNKNOWN,            /* error/unspecified */
+  JCS_GRAYSCALE,          /* monochrome */
+  JCS_RGB,                /* red/green/blue as specified by the RGB_RED,
+                             RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros */
+  JCS_YCbCr,              /* Y/Cb/Cr (also known as YUV) */
+  JCS_CMYK,               /* C/M/Y/K */
+  JCS_YCCK,               /* Y/Cb/Cr/K */
+  JCS_EXT_RGB,            /* red/green/blue */
+  JCS_EXT_RGBX,           /* red/green/blue/x */
+  JCS_EXT_BGR,            /* blue/green/red */
+  JCS_EXT_BGRX,           /* blue/green/red/x */
+  JCS_EXT_XBGR,           /* x/blue/green/red */
+  JCS_EXT_XRGB,           /* x/red/green/blue */
+  /* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR,
+     or JCS_EXT_XRGB during decompression, the X byte is undefined, and in
+     order to ensure the best performance, libjpeg-turbo can set that byte to
+     whatever value it wishes.  Use the following colorspace constants to
+     ensure that the X byte is set to 0xFF, so that it can be interpreted as an
+     opaque alpha channel. */
+  JCS_EXT_RGBA,           /* red/green/blue/alpha */
+  JCS_EXT_BGRA,           /* blue/green/red/alpha */
+  JCS_EXT_ABGR,           /* alpha/blue/green/red */
+  JCS_EXT_ARGB,           /* alpha/red/green/blue */
+  JCS_RGB565              /* 5-bit red/6-bit green/5-bit blue */
 } J_COLOR_SPACE;
 
 /* DCT/IDCT algorithm options. */
 
 typedef enum {
-	JDCT_ISLOW,		/* slow but accurate integer algorithm */
-	JDCT_IFAST,		/* faster, less accurate integer method */
-	JDCT_FLOAT		/* floating-point: accurate, fast on fast HW */
+  JDCT_ISLOW,             /* slow but accurate integer algorithm */
+  JDCT_IFAST,             /* faster, less accurate integer method */
+  JDCT_FLOAT              /* floating-point: accurate, fast on fast HW */
 } J_DCT_METHOD;
 
-#ifndef JDCT_DEFAULT		/* may be overridden in jconfig.h */
+#ifndef JDCT_DEFAULT            /* may be overridden in jconfig.h */
 #define JDCT_DEFAULT  JDCT_ISLOW
 #endif
-#ifndef JDCT_FASTEST		/* may be overridden in jconfig.h */
+#ifndef JDCT_FASTEST            /* may be overridden in jconfig.h */
 #define JDCT_FASTEST  JDCT_IFAST
 #endif
 
 /* Dithering options for decompression. */
 
 typedef enum {
-	JDITHER_NONE,		/* no dithering */
-	JDITHER_ORDERED,	/* simple ordered dither */
-	JDITHER_FS		/* Floyd-Steinberg error diffusion dither */
+  JDITHER_NONE,           /* no dithering */
+  JDITHER_ORDERED,        /* simple ordered dither */
+  JDITHER_FS              /* Floyd-Steinberg error diffusion dither */
 } J_DITHER_MODE;
 
 
 /* Common fields between JPEG compression and decompression master structs. */
 
 #define jpeg_common_fields \
-  struct jpeg_error_mgr * err;	/* Error handler module */\
-  struct jpeg_memory_mgr * mem;	/* Memory manager module */\
-  struct jpeg_progress_mgr * progress; /* Progress monitor, or NULL if none */\
-  void * client_data;		/* Available for use by application */\
-  boolean is_decompressor;	/* So common code can tell which is which */\
-  int global_state		/* For checking call sequence validity */
+  struct jpeg_error_mgr *err;   /* Error handler module */\
+  struct jpeg_memory_mgr *mem;  /* Memory manager module */\
+  struct jpeg_progress_mgr *progress; /* Progress monitor, or NULL if none */\
+  void *client_data;            /* Available for use by application */\
+  boolean is_decompressor;      /* So common code can tell which is which */\
+  int global_state              /* For checking call sequence validity */
 
 /* Routines that are to be used by both halves of the library are declared
  * to receive a pointer to this structure.  There are no actual instances of
  * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct.
  */
 struct jpeg_common_struct {
-  jpeg_common_fields;		/* Fields common to both master struct types */
+  jpeg_common_fields;           /* Fields common to both master struct types */
   /* Additional fields follow in an actual jpeg_compress_struct or
    * jpeg_decompress_struct.  All three structs must agree on these
    * initial fields!  (This would be a lot cleaner in C++.)
    */
 };
 
-typedef struct jpeg_common_struct * j_common_ptr;
-typedef struct jpeg_compress_struct * j_compress_ptr;
-typedef struct jpeg_decompress_struct * j_decompress_ptr;
+typedef struct jpeg_common_struct *j_common_ptr;
+typedef struct jpeg_compress_struct *j_compress_ptr;
+typedef struct jpeg_decompress_struct *j_decompress_ptr;
 
 
 /* Master record for a compression instance */
 
 struct jpeg_compress_struct {
-  jpeg_common_fields;		/* Fields shared with jpeg_decompress_struct */
+  jpeg_common_fields;           /* Fields shared with jpeg_decompress_struct */
 
   /* Destination for compressed data */
-  struct jpeg_destination_mgr * dest;
+  struct jpeg_destination_mgr *dest;
 
   /* Description of source image --- these fields must be filled in by
    * outer application before starting compression.  in_color_space must
    * be correct before you can even call jpeg_set_defaults().
    */
 
-  JDIMENSION image_width;	/* input image width */
-  JDIMENSION image_height;	/* input image height */
-  int input_components;		/* # of color components in input image */
-  J_COLOR_SPACE in_color_space;	/* colorspace of input image */
+  JDIMENSION image_width;       /* input image width */
+  JDIMENSION image_height;      /* input image height */
+  int input_components;         /* # of color components in input image */
+  J_COLOR_SPACE in_color_space; /* colorspace of input image */
 
-  double input_gamma;		/* image gamma of input image */
+  double input_gamma;           /* image gamma of input image */
 
   /* Compression parameters --- these fields must be set before calling
    * jpeg_start_compress().  We recommend calling jpeg_set_defaults() to
@@ -322,8 +327,8 @@
 #if JPEG_LIB_VERSION >= 70
   unsigned int scale_num, scale_denom; /* fraction by which to scale image */
 
-  JDIMENSION jpeg_width;	/* scaled JPEG image width */
-  JDIMENSION jpeg_height;	/* scaled JPEG image height */
+  JDIMENSION jpeg_width;        /* scaled JPEG image width */
+  JDIMENSION jpeg_height;       /* scaled JPEG image height */
   /* Dimensions of actual JPEG image that will be written to file,
    * derived from input dimensions by scaling factors above.
    * These fields are computed by jpeg_start_compress().
@@ -332,15 +337,15 @@
    */
 #endif
 
-  int data_precision;		/* bits of precision in image data */
+  int data_precision;           /* bits of precision in image data */
 
-  int num_components;		/* # of color components in JPEG image */
+  int num_components;           /* # of color components in JPEG image */
   J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
 
-  jpeg_component_info * comp_info;
+  jpeg_component_info *comp_info;
   /* comp_info[i] describes component that appears i'th in SOF */
 
-  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS];
 #if JPEG_LIB_VERSION >= 70
   int q_scale_factor[NUM_QUANT_TBLS];
 #endif
@@ -348,30 +353,30 @@
    * and corresponding scale factors (percentage, initialized 100).
    */
 
-  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
-  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
   /* ptrs to Huffman coding tables, or NULL if not defined */
 
   UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
   UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
   UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
 
-  int num_scans;		/* # of entries in scan_info array */
-  const jpeg_scan_info * scan_info; /* script for multi-scan file, or NULL */
+  int num_scans;                /* # of entries in scan_info array */
+  const jpeg_scan_info *scan_info; /* script for multi-scan file, or NULL */
   /* The default value of scan_info is NULL, which causes a single-scan
    * sequential JPEG file to be emitted.  To create a multi-scan file,
    * set num_scans and scan_info to point to an array of scan definitions.
    */
 
-  boolean raw_data_in;		/* TRUE=caller supplies downsampled data */
-  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
-  boolean optimize_coding;	/* TRUE=optimize entropy encoding parms */
-  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  boolean raw_data_in;          /* TRUE=caller supplies downsampled data */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean optimize_coding;      /* TRUE=optimize entropy encoding parms */
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
 #if JPEG_LIB_VERSION >= 70
   boolean do_fancy_downsampling; /* TRUE=apply fancy downsampling */
 #endif
-  int smoothing_factor;		/* 1..100, or 0 for no input smoothing */
-  J_DCT_METHOD dct_method;	/* DCT algorithm selector */
+  int smoothing_factor;         /* 1..100, or 0 for no input smoothing */
+  J_DCT_METHOD dct_method;      /* DCT algorithm selector */
 
   /* The restart interval can be specified in absolute MCUs by setting
    * restart_interval, or in MCU rows by setting restart_in_rows
@@ -379,28 +384,28 @@
    * for each scan).
    */
   unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */
-  int restart_in_rows;		/* if > 0, MCU rows per restart interval */
+  int restart_in_rows;          /* if > 0, MCU rows per restart interval */
 
   /* Parameters controlling emission of special markers. */
 
-  boolean write_JFIF_header;	/* should a JFIF marker be written? */
-  UINT8 JFIF_major_version;	/* What to write for the JFIF version number */
+  boolean write_JFIF_header;    /* should a JFIF marker be written? */
+  UINT8 JFIF_major_version;     /* What to write for the JFIF version number */
   UINT8 JFIF_minor_version;
   /* These three values are not used by the JPEG code, merely copied */
   /* into the JFIF APP0 marker.  density_unit can be 0 for unknown, */
   /* 1 for dots/inch, or 2 for dots/cm.  Note that the pixel aspect */
   /* ratio is defined by X_density/Y_density even when density_unit=0. */
-  UINT8 density_unit;		/* JFIF code for pixel size units */
-  UINT16 X_density;		/* Horizontal pixel density */
-  UINT16 Y_density;		/* Vertical pixel density */
-  boolean write_Adobe_marker;	/* should an Adobe marker be written? */
-  
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean write_Adobe_marker;   /* should an Adobe marker be written? */
+
   /* State variable: index of next scanline to be written to
    * jpeg_write_scanlines().  Application may use this to control its
    * processing loop, e.g., "while (next_scanline < image_height)".
    */
 
-  JDIMENSION next_scanline;	/* 0 .. image_height-1  */
+  JDIMENSION next_scanline;     /* 0 .. image_height-1  */
 
   /* Remaining fields are known throughout compressor, but generally
    * should not be touched by a surrounding application.
@@ -409,59 +414,59 @@
   /*
    * These fields are computed during compression startup
    */
-  boolean progressive_mode;	/* TRUE if scan script uses progressive mode */
-  int max_h_samp_factor;	/* largest h_samp_factor */
-  int max_v_samp_factor;	/* largest v_samp_factor */
+  boolean progressive_mode;     /* TRUE if scan script uses progressive mode */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
 
 #if JPEG_LIB_VERSION >= 70
-  int min_DCT_h_scaled_size;	/* smallest DCT_h_scaled_size of any component */
-  int min_DCT_v_scaled_size;	/* smallest DCT_v_scaled_size of any component */
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
 #endif
 
-  JDIMENSION total_iMCU_rows;	/* # of iMCU rows to be input to coef ctlr */
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows to be input to coef ctlr */
   /* The coefficient controller receives data in units of MCU rows as defined
    * for fully interleaved scans (whether the JPEG file is interleaved or not).
    * There are v_samp_factor * DCTSIZE sample rows of each component in an
    * "iMCU" (interleaved MCU) row.
    */
-  
+
   /*
    * These fields are valid during any one scan.
    * They describe the components and MCUs actually appearing in the scan.
    */
-  int comps_in_scan;		/* # of JPEG components in this scan */
-  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_info[MAX_COMPS_IN_SCAN];
   /* *cur_comp_info[i] describes component that appears i'th in SOS */
-  
-  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
-  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
-  
-  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
+
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
   int MCU_membership[C_MAX_BLOCKS_IN_MCU];
   /* MCU_membership[i] is index in cur_comp_info of component owning */
   /* i'th block in an MCU */
 
-  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
 
 #if JPEG_LIB_VERSION >= 80
-  int block_size;		/* the basic DCT block size: 1..16 */
-  const int * natural_order;	/* natural-order position array */
-  int lim_Se;			/* min( Se, DCTSIZE2-1 ) */
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order;     /* natural-order position array */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) */
 #endif
 
   /*
    * Links to compression subobjects (methods and private variables of modules)
    */
-  struct jpeg_comp_master * master;
-  struct jpeg_c_main_controller * main;
-  struct jpeg_c_prep_controller * prep;
-  struct jpeg_c_coef_controller * coef;
-  struct jpeg_marker_writer * marker;
-  struct jpeg_color_converter * cconvert;
-  struct jpeg_downsampler * downsample;
-  struct jpeg_forward_dct * fdct;
-  struct jpeg_entropy_encoder * entropy;
-  jpeg_scan_info * script_space; /* workspace for jpeg_simple_progression */
+  struct jpeg_comp_master *master;
+  struct jpeg_c_main_controller *main;
+  struct jpeg_c_prep_controller *prep;
+  struct jpeg_c_coef_controller *coef;
+  struct jpeg_marker_writer *marker;
+  struct jpeg_color_converter *cconvert;
+  struct jpeg_downsampler *downsample;
+  struct jpeg_forward_dct *fdct;
+  struct jpeg_entropy_encoder *entropy;
+  jpeg_scan_info *script_space; /* workspace for jpeg_simple_progression */
   int script_space_size;
 };
 
@@ -469,17 +474,17 @@
 /* Master record for a decompression instance */
 
 struct jpeg_decompress_struct {
-  jpeg_common_fields;		/* Fields shared with jpeg_compress_struct */
+  jpeg_common_fields;           /* Fields shared with jpeg_compress_struct */
 
   /* Source of compressed data */
-  struct jpeg_source_mgr * src;
+  struct jpeg_source_mgr *src;
 
   /* Basic description of image --- filled in by jpeg_read_header(). */
   /* Application may inspect these values to decide how to process image. */
 
-  JDIMENSION image_width;	/* nominal image width (from SOF marker) */
-  JDIMENSION image_height;	/* nominal image height */
-  int num_components;		/* # of color components in JPEG image */
+  JDIMENSION image_width;       /* nominal image width (from SOF marker) */
+  JDIMENSION image_height;      /* nominal image height */
+  int num_components;           /* # of color components in JPEG image */
   J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
 
   /* Decompression processing parameters --- these fields must be set before
@@ -491,24 +496,24 @@
 
   unsigned int scale_num, scale_denom; /* fraction by which to scale image */
 
-  double output_gamma;		/* image gamma wanted in output */
+  double output_gamma;          /* image gamma wanted in output */
 
-  boolean buffered_image;	/* TRUE=multiple output passes */
-  boolean raw_data_out;		/* TRUE=downsampled data wanted */
+  boolean buffered_image;       /* TRUE=multiple output passes */
+  boolean raw_data_out;         /* TRUE=downsampled data wanted */
 
-  J_DCT_METHOD dct_method;	/* IDCT algorithm selector */
-  boolean do_fancy_upsampling;	/* TRUE=apply fancy upsampling */
-  boolean do_block_smoothing;	/* TRUE=apply interblock smoothing */
+  J_DCT_METHOD dct_method;      /* IDCT algorithm selector */
+  boolean do_fancy_upsampling;  /* TRUE=apply fancy upsampling */
+  boolean do_block_smoothing;   /* TRUE=apply interblock smoothing */
 
-  boolean quantize_colors;	/* TRUE=colormapped output wanted */
+  boolean quantize_colors;      /* TRUE=colormapped output wanted */
   /* the following are ignored if not quantize_colors: */
-  J_DITHER_MODE dither_mode;	/* type of color dithering to use */
-  boolean two_pass_quantize;	/* TRUE=use two-pass color quantization */
-  int desired_number_of_colors;	/* max # colors to use in created colormap */
+  J_DITHER_MODE dither_mode;    /* type of color dithering to use */
+  boolean two_pass_quantize;    /* TRUE=use two-pass color quantization */
+  int desired_number_of_colors; /* max # colors to use in created colormap */
   /* these are significant only in buffered-image mode: */
-  boolean enable_1pass_quant;	/* enable future use of 1-pass quantizer */
+  boolean enable_1pass_quant;   /* enable future use of 1-pass quantizer */
   boolean enable_external_quant;/* enable future use of external colormap */
-  boolean enable_2pass_quant;	/* enable future use of 2-pass quantizer */
+  boolean enable_2pass_quant;   /* enable future use of 2-pass quantizer */
 
   /* Description of actual output image that will be returned to application.
    * These fields are computed by jpeg_start_decompress().
@@ -516,14 +521,14 @@
    * in advance of calling jpeg_start_decompress().
    */
 
-  JDIMENSION output_width;	/* scaled image width */
-  JDIMENSION output_height;	/* scaled image height */
-  int out_color_components;	/* # of color components in out_color_space */
-  int output_components;	/* # of color components returned */
+  JDIMENSION output_width;      /* scaled image width */
+  JDIMENSION output_height;     /* scaled image height */
+  int out_color_components;     /* # of color components in out_color_space */
+  int output_components;        /* # of color components returned */
   /* output_components is 1 (a colormap index) when quantizing colors;
    * otherwise it equals out_color_components.
    */
-  int rec_outbuf_height;	/* min recommended height of scanline buffer */
+  int rec_outbuf_height;        /* min recommended height of scanline buffer */
   /* If the buffer passed to jpeg_read_scanlines() is less than this many rows
    * high, space and time will be wasted due to unnecessary data copying.
    * Usually rec_outbuf_height will be 1 or 2, at most 4.
@@ -535,8 +540,8 @@
    * jpeg_start_decompress or jpeg_start_output.
    * The map has out_color_components rows and actual_number_of_colors columns.
    */
-  int actual_number_of_colors;	/* number of entries in use */
-  JSAMPARRAY colormap;		/* The color map as a 2-D pixel array */
+  int actual_number_of_colors;  /* number of entries in use */
+  JSAMPARRAY colormap;          /* The color map as a 2-D pixel array */
 
   /* State variables: these variables indicate the progress of decompression.
    * The application may examine these but must not modify them.
@@ -546,20 +551,20 @@
    * Application may use this to control its processing loop, e.g.,
    * "while (output_scanline < output_height)".
    */
-  JDIMENSION output_scanline;	/* 0 .. output_height-1  */
+  JDIMENSION output_scanline;   /* 0 .. output_height-1  */
 
   /* Current input scan number and number of iMCU rows completed in scan.
    * These indicate the progress of the decompressor input side.
    */
-  int input_scan_number;	/* Number of SOS markers seen so far */
-  JDIMENSION input_iMCU_row;	/* Number of iMCU rows completed */
+  int input_scan_number;        /* Number of SOS markers seen so far */
+  JDIMENSION input_iMCU_row;    /* Number of iMCU rows completed */
 
   /* The "output scan number" is the notional scan being displayed by the
    * output side.  The decompressor will not allow output scan/row number
    * to get ahead of input scan/row, but it can fall arbitrarily far behind.
    */
-  int output_scan_number;	/* Nominal scan number being displayed */
-  JDIMENSION output_iMCU_row;	/* Number of iMCU rows read */
+  int output_scan_number;       /* Nominal scan number being displayed */
+  JDIMENSION output_iMCU_row;   /* Number of iMCU rows read */
 
   /* Current progression status.  coef_bits[c][i] indicates the precision
    * with which component c's DCT coefficient i (in zigzag order) is known.
@@ -568,7 +573,7 @@
    * (thus, 0 at completion of the progression).
    * This pointer is NULL when reading a non-progressive file.
    */
-  int (*coef_bits)[DCTSIZE2];	/* -1 or current Al value for each coef */
+  int (*coef_bits)[DCTSIZE2];   /* -1 or current Al value for each coef */
 
   /* Internal JPEG parameters --- the application usually need not look at
    * these fields.  Note that the decompressor output side may not use
@@ -579,27 +584,27 @@
    * datastreams when processing abbreviated JPEG datastreams.
    */
 
-  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS];
   /* ptrs to coefficient quantization tables, or NULL if not defined */
 
-  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
-  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
   /* ptrs to Huffman coding tables, or NULL if not defined */
 
   /* These parameters are never carried across datastreams, since they
    * are given in SOF/SOS markers or defined to be reset by SOI.
    */
 
-  int data_precision;		/* bits of precision in image data */
+  int data_precision;           /* bits of precision in image data */
 
-  jpeg_component_info * comp_info;
+  jpeg_component_info *comp_info;
   /* comp_info[i] describes component that appears i'th in SOF */
 
 #if JPEG_LIB_VERSION >= 80
-  boolean is_baseline;		/* TRUE if Baseline SOF0 encountered */
+  boolean is_baseline;          /* TRUE if Baseline SOF0 encountered */
 #endif
-  boolean progressive_mode;	/* TRUE if SOFn specifies progressive mode */
-  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean progressive_mode;     /* TRUE if SOFn specifies progressive mode */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
 
   UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
   UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
@@ -610,17 +615,17 @@
   /* These fields record data obtained from optional markers recognized by
    * the JPEG library.
    */
-  boolean saw_JFIF_marker;	/* TRUE iff a JFIF APP0 marker was found */
+  boolean saw_JFIF_marker;      /* TRUE iff a JFIF APP0 marker was found */
   /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */
-  UINT8 JFIF_major_version;	/* JFIF version number */
+  UINT8 JFIF_major_version;     /* JFIF version number */
   UINT8 JFIF_minor_version;
-  UINT8 density_unit;		/* JFIF code for pixel size units */
-  UINT16 X_density;		/* Horizontal pixel density */
-  UINT16 Y_density;		/* Vertical pixel density */
-  boolean saw_Adobe_marker;	/* TRUE iff an Adobe APP14 marker was found */
-  UINT8 Adobe_transform;	/* Color transform code from Adobe marker */
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean saw_Adobe_marker;     /* TRUE iff an Adobe APP14 marker was found */
+  UINT8 Adobe_transform;        /* Color transform code from Adobe marker */
 
-  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
 
   /* Aside from the specific data retained from APPn markers known to the
    * library, the uninterpreted contents of any or all APPn and COM markers
@@ -635,17 +640,17 @@
   /*
    * These fields are computed during decompression startup
    */
-  int max_h_samp_factor;	/* largest h_samp_factor */
-  int max_v_samp_factor;	/* largest v_samp_factor */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
 
 #if JPEG_LIB_VERSION >= 70
-  int min_DCT_h_scaled_size;	/* smallest DCT_h_scaled_size of any component */
-  int min_DCT_v_scaled_size;	/* smallest DCT_v_scaled_size of any component */
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
 #else
-  int min_DCT_scaled_size;	/* smallest DCT_scaled_size of any component */
+  int min_DCT_scaled_size;      /* smallest DCT_scaled_size of any component */
 #endif
 
-  JDIMENSION total_iMCU_rows;	/* # of iMCU rows in image */
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows in image */
   /* The coefficient controller's input and output progress is measured in
    * units of "iMCU" (interleaved MCU) rows.  These are the same as MCU rows
    * in fully interleaved JPEG scans, but are used whether the scan is
@@ -654,33 +659,33 @@
    * v_samp_factor*DCT_[v_]scaled_size sample rows of a component per iMCU row.
    */
 
-  JSAMPLE * sample_range_limit; /* table for fast range-limiting */
+  JSAMPLE *sample_range_limit;  /* table for fast range-limiting */
 
   /*
    * These fields are valid during any one scan.
    * They describe the components and MCUs actually appearing in the scan.
    * Note that the decompressor output side must not use these fields.
    */
-  int comps_in_scan;		/* # of JPEG components in this scan */
-  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_info[MAX_COMPS_IN_SCAN];
   /* *cur_comp_info[i] describes component that appears i'th in SOS */
 
-  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
-  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
 
-  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
   int MCU_membership[D_MAX_BLOCKS_IN_MCU];
   /* MCU_membership[i] is index in cur_comp_info of component owning */
   /* i'th block in an MCU */
 
-  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
 
 #if JPEG_LIB_VERSION >= 80
   /* These fields are derived from Se of first SOS marker.
    */
-  int block_size;		/* the basic DCT block size: 1..16 */
-  const int * natural_order; /* natural-order position array for entropy decode */
-  int lim_Se;			/* min( Se, DCTSIZE2-1 ) for entropy decode */
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order; /* natural-order position array for entropy decode */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) for entropy decode */
 #endif
 
   /* This field is shared between entropy decoder and marker parser.
@@ -692,17 +697,17 @@
   /*
    * Links to decompression subobjects (methods, private variables of modules)
    */
-  struct jpeg_decomp_master * master;
-  struct jpeg_d_main_controller * main;
-  struct jpeg_d_coef_controller * coef;
-  struct jpeg_d_post_controller * post;
-  struct jpeg_input_controller * inputctl;
-  struct jpeg_marker_reader * marker;
-  struct jpeg_entropy_decoder * entropy;
-  struct jpeg_inverse_dct * idct;
-  struct jpeg_upsampler * upsample;
-  struct jpeg_color_deconverter * cconvert;
-  struct jpeg_color_quantizer * cquantize;
+  struct jpeg_decomp_master *master;
+  struct jpeg_d_main_controller *main;
+  struct jpeg_d_coef_controller *coef;
+  struct jpeg_d_post_controller *post;
+  struct jpeg_input_controller *inputctl;
+  struct jpeg_marker_reader *marker;
+  struct jpeg_entropy_decoder *entropy;
+  struct jpeg_inverse_dct *idct;
+  struct jpeg_upsampler *upsample;
+  struct jpeg_color_deconverter *cconvert;
+  struct jpeg_color_quantizer *cquantize;
 };
 
 
@@ -718,17 +723,17 @@
 
 struct jpeg_error_mgr {
   /* Error exit handler: does not return to caller */
-  JMETHOD(void, error_exit, (j_common_ptr cinfo));
+  void (*error_exit) (j_common_ptr cinfo);
   /* Conditionally emit a trace or warning message */
-  JMETHOD(void, emit_message, (j_common_ptr cinfo, int msg_level));
+  void (*emit_message) (j_common_ptr cinfo, int msg_level);
   /* Routine that actually outputs a trace or error message */
-  JMETHOD(void, output_message, (j_common_ptr cinfo));
+  void (*output_message) (j_common_ptr cinfo);
   /* Format a message string for the most recent JPEG error or message */
-  JMETHOD(void, format_message, (j_common_ptr cinfo, char * buffer));
-#define JMSG_LENGTH_MAX  200	/* recommended size of format_message buffer */
+  void (*format_message) (j_common_ptr cinfo, char *buffer);
+#define JMSG_LENGTH_MAX  200    /* recommended size of format_message buffer */
   /* Reset error state variables at start of a new image */
-  JMETHOD(void, reset_error_mgr, (j_common_ptr cinfo));
-  
+  void (*reset_error_mgr) (j_common_ptr cinfo);
+
   /* The message ID code and any parameters are saved here.
    * A message can have one string parameter or up to 8 int parameters.
    */
@@ -738,18 +743,18 @@
     int i[8];
     char s[JMSG_STR_PARM_MAX];
   } msg_parm;
-  
+
   /* Standard state variables for error facility */
-  
-  int trace_level;		/* max msg_level that will be displayed */
-  
+
+  int trace_level;              /* max msg_level that will be displayed */
+
   /* For recoverable corrupt-data errors, we emit a warning message,
    * but keep going unless emit_message chooses to abort.  emit_message
    * should count warnings in num_warnings.  The surrounding application
    * can check for bad data by seeing if num_warnings is nonzero at the
    * end of processing.
    */
-  long num_warnings;		/* number of corrupt-data warnings */
+  long num_warnings;            /* number of corrupt-data warnings */
 
   /* These fields point to the table(s) of error message strings.
    * An application can change the table pointer to switch to a different
@@ -761,52 +766,52 @@
    * First table includes all errors generated by JPEG library itself.
    * Error code 0 is reserved for a "no such error string" message.
    */
-  const char * const * jpeg_message_table; /* Library errors */
+  const char * const *jpeg_message_table; /* Library errors */
   int last_jpeg_message;    /* Table contains strings 0..last_jpeg_message */
   /* Second table can be added by application (see cjpeg/djpeg for example).
    * It contains strings numbered first_addon_message..last_addon_message.
    */
-  const char * const * addon_message_table; /* Non-library errors */
-  int first_addon_message;	/* code for first string in addon table */
-  int last_addon_message;	/* code for last string in addon table */
+  const char * const *addon_message_table; /* Non-library errors */
+  int first_addon_message;      /* code for first string in addon table */
+  int last_addon_message;       /* code for last string in addon table */
 };
 
 
 /* Progress monitor object */
 
 struct jpeg_progress_mgr {
-  JMETHOD(void, progress_monitor, (j_common_ptr cinfo));
+  void (*progress_monitor) (j_common_ptr cinfo);
 
-  long pass_counter;		/* work units completed in this pass */
-  long pass_limit;		/* total number of work units in this pass */
-  int completed_passes;		/* passes completed so far */
-  int total_passes;		/* total number of passes expected */
+  long pass_counter;            /* work units completed in this pass */
+  long pass_limit;              /* total number of work units in this pass */
+  int completed_passes;         /* passes completed so far */
+  int total_passes;             /* total number of passes expected */
 };
 
 
 /* Data destination object for compression */
 
 struct jpeg_destination_mgr {
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
 
-  JMETHOD(void, init_destination, (j_compress_ptr cinfo));
-  JMETHOD(boolean, empty_output_buffer, (j_compress_ptr cinfo));
-  JMETHOD(void, term_destination, (j_compress_ptr cinfo));
+  void (*init_destination) (j_compress_ptr cinfo);
+  boolean (*empty_output_buffer) (j_compress_ptr cinfo);
+  void (*term_destination) (j_compress_ptr cinfo);
 };
 
 
 /* Data source object for decompression */
 
 struct jpeg_source_mgr {
-  const JOCTET * next_input_byte; /* => next byte to read from buffer */
-  size_t bytes_in_buffer;	/* # of bytes remaining in buffer */
+  const JOCTET *next_input_byte; /* => next byte to read from buffer */
+  size_t bytes_in_buffer;       /* # of bytes remaining in buffer */
 
-  JMETHOD(void, init_source, (j_decompress_ptr cinfo));
-  JMETHOD(boolean, fill_input_buffer, (j_decompress_ptr cinfo));
-  JMETHOD(void, skip_input_data, (j_decompress_ptr cinfo, long num_bytes));
-  JMETHOD(boolean, resync_to_restart, (j_decompress_ptr cinfo, int desired));
-  JMETHOD(void, term_source, (j_decompress_ptr cinfo));
+  void (*init_source) (j_decompress_ptr cinfo);
+  boolean (*fill_input_buffer) (j_decompress_ptr cinfo);
+  void (*skip_input_data) (j_decompress_ptr cinfo, long num_bytes);
+  boolean (*resync_to_restart) (j_decompress_ptr cinfo, int desired);
+  void (*term_source) (j_decompress_ptr cinfo);
 };
 
 
@@ -821,51 +826,42 @@
  * successful.
  */
 
-#define JPOOL_PERMANENT	0	/* lasts until master record is destroyed */
-#define JPOOL_IMAGE	1	/* lasts until done with image/datastream */
-#define JPOOL_NUMPOOLS	2
+#define JPOOL_PERMANENT 0       /* lasts until master record is destroyed */
+#define JPOOL_IMAGE     1       /* lasts until done with image/datastream */
+#define JPOOL_NUMPOOLS  2
 
-typedef struct jvirt_sarray_control * jvirt_sarray_ptr;
-typedef struct jvirt_barray_control * jvirt_barray_ptr;
+typedef struct jvirt_sarray_control *jvirt_sarray_ptr;
+typedef struct jvirt_barray_control *jvirt_barray_ptr;
 
 
 struct jpeg_memory_mgr {
   /* Method pointers */
-  JMETHOD(void *, alloc_small, (j_common_ptr cinfo, int pool_id,
-				size_t sizeofobject));
-  JMETHOD(void FAR *, alloc_large, (j_common_ptr cinfo, int pool_id,
-				     size_t sizeofobject));
-  JMETHOD(JSAMPARRAY, alloc_sarray, (j_common_ptr cinfo, int pool_id,
-				     JDIMENSION samplesperrow,
-				     JDIMENSION numrows));
-  JMETHOD(JBLOCKARRAY, alloc_barray, (j_common_ptr cinfo, int pool_id,
-				      JDIMENSION blocksperrow,
-				      JDIMENSION numrows));
-  JMETHOD(jvirt_sarray_ptr, request_virt_sarray, (j_common_ptr cinfo,
-						  int pool_id,
-						  boolean pre_zero,
-						  JDIMENSION samplesperrow,
-						  JDIMENSION numrows,
-						  JDIMENSION maxaccess));
-  JMETHOD(jvirt_barray_ptr, request_virt_barray, (j_common_ptr cinfo,
-						  int pool_id,
-						  boolean pre_zero,
-						  JDIMENSION blocksperrow,
-						  JDIMENSION numrows,
-						  JDIMENSION maxaccess));
-  JMETHOD(void, realize_virt_arrays, (j_common_ptr cinfo));
-  JMETHOD(JSAMPARRAY, access_virt_sarray, (j_common_ptr cinfo,
-					   jvirt_sarray_ptr ptr,
-					   JDIMENSION start_row,
-					   JDIMENSION num_rows,
-					   boolean writable));
-  JMETHOD(JBLOCKARRAY, access_virt_barray, (j_common_ptr cinfo,
-					    jvirt_barray_ptr ptr,
-					    JDIMENSION start_row,
-					    JDIMENSION num_rows,
-					    boolean writable));
-  JMETHOD(void, free_pool, (j_common_ptr cinfo, int pool_id));
-  JMETHOD(void, self_destruct, (j_common_ptr cinfo));
+  void *(*alloc_small) (j_common_ptr cinfo, int pool_id, size_t sizeofobject);
+  void *(*alloc_large) (j_common_ptr cinfo, int pool_id,
+                        size_t sizeofobject);
+  JSAMPARRAY (*alloc_sarray) (j_common_ptr cinfo, int pool_id,
+                              JDIMENSION samplesperrow, JDIMENSION numrows);
+  JBLOCKARRAY (*alloc_barray) (j_common_ptr cinfo, int pool_id,
+                               JDIMENSION blocksperrow, JDIMENSION numrows);
+  jvirt_sarray_ptr (*request_virt_sarray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION samplesperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  jvirt_barray_ptr (*request_virt_barray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION blocksperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  void (*realize_virt_arrays) (j_common_ptr cinfo);
+  JSAMPARRAY (*access_virt_sarray) (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+                                    JDIMENSION start_row, JDIMENSION num_rows,
+                                    boolean writable);
+  JBLOCKARRAY (*access_virt_barray) (j_common_ptr cinfo, jvirt_barray_ptr ptr,
+                                     JDIMENSION start_row, JDIMENSION num_rows,
+                                     boolean writable);
+  void (*free_pool) (j_common_ptr cinfo, int pool_id);
+  void (*self_destruct) (j_common_ptr cinfo);
 
   /* Limit on memory allocation for this JPEG object.  (Note that this is
    * merely advisory, not a guaranteed maximum; it only affects the space
@@ -882,96 +878,21 @@
 /* Routine signature for application-supplied marker processing methods.
  * Need not pass marker code since it is stored in cinfo->unread_marker.
  */
-typedef JMETHOD(boolean, jpeg_marker_parser_method, (j_decompress_ptr cinfo));
-
-
-/* Declarations for routines called by application.
- * The JPP macro hides prototype parameters from compilers that can't cope.
- * Note JPP requires double parentheses.
- */
-
-#ifdef HAVE_PROTOTYPES
-#define JPP(arglist)	arglist
-#else
-#define JPP(arglist)	()
-#endif
+typedef boolean (*jpeg_marker_parser_method) (j_decompress_ptr cinfo);
 
 
-/* Short forms of external names for systems with brain-damaged linkers.
- * We shorten external names to be unique in the first six letters, which
- * is good enough for all known systems.
- * (If your compiler itself needs names to be unique in less than 15 
- * characters, you are out of luck.  Get a better compiler.)
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
  */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_std_error		jStdError
-#define jpeg_CreateCompress	jCreaCompress
-#define jpeg_CreateDecompress	jCreaDecompress
-#define jpeg_destroy_compress	jDestCompress
-#define jpeg_destroy_decompress	jDestDecompress
-#define jpeg_stdio_dest		jStdDest
-#define jpeg_stdio_src		jStdSrc
-#if JPEG_LIB_VERSION >= 80
-#define jpeg_mem_dest		jMemDest
-#define jpeg_mem_src		jMemSrc
-#endif
-#define jpeg_set_defaults	jSetDefaults
-#define jpeg_set_colorspace	jSetColorspace
-#define jpeg_default_colorspace	jDefColorspace
-#define jpeg_set_quality	jSetQuality
-#define jpeg_set_linear_quality	jSetLQuality
-#if JPEG_LIB_VERSION >= 70
-#define jpeg_default_qtables	jDefQTables
-#endif
-#define jpeg_add_quant_table	jAddQuantTable
-#define jpeg_quality_scaling	jQualityScaling
-#define jpeg_simple_progression	jSimProgress
-#define jpeg_suppress_tables	jSuppressTables
-#define jpeg_alloc_quant_table	jAlcQTable
-#define jpeg_alloc_huff_table	jAlcHTable
-#define jpeg_start_compress	jStrtCompress
-#define jpeg_write_scanlines	jWrtScanlines
-#define jpeg_finish_compress	jFinCompress
-#if JPEG_LIB_VERSION >= 70
-#define jpeg_calc_jpeg_dimensions	jCjpegDimensions
-#endif
-#define jpeg_write_raw_data	jWrtRawData
-#define jpeg_write_marker	jWrtMarker
-#define jpeg_write_m_header	jWrtMHeader
-#define jpeg_write_m_byte	jWrtMByte
-#define jpeg_write_tables	jWrtTables
-#define jpeg_read_header	jReadHeader
-#define jpeg_start_decompress	jStrtDecompress
-#define jpeg_read_scanlines	jReadScanlines
-#define jpeg_finish_decompress	jFinDecompress
-#define jpeg_read_raw_data	jReadRawData
-#define jpeg_has_multiple_scans	jHasMultScn
-#define jpeg_start_output	jStrtOutput
-#define jpeg_finish_output	jFinOutput
-#define jpeg_input_complete	jInComplete
-#define jpeg_new_colormap	jNewCMap
-#define jpeg_consume_input	jConsumeInput
-#if JPEG_LIB_VERSION >= 80
-#define jpeg_core_output_dimensions	jCoreDimensions
-#endif
-#define jpeg_calc_output_dimensions	jCalcDimensions
-#define jpeg_save_markers	jSaveMarkers
-#define jpeg_set_marker_processor	jSetMarker
-#define jpeg_read_coefficients	jReadCoefs
-#define jpeg_write_coefficients	jWrtCoefs
-#define jpeg_copy_critical_parameters	jCopyCrit
-#define jpeg_abort_compress	jAbrtCompress
-#define jpeg_abort_decompress	jAbrtDecompress
-#define jpeg_abort		jAbort
-#define jpeg_destroy		jDestroy
-#define jpeg_resync_to_restart	jResyncRestart
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#define JPP(arglist)    arglist
 
 
 /* Default error-management setup */
-EXTERN(struct jpeg_error_mgr *) jpeg_std_error
-	JPP((struct jpeg_error_mgr * err));
+EXTERN(struct jpeg_error_mgr *) jpeg_std_error (struct jpeg_error_mgr *err);
 
 /* Initialization of JPEG compression objects.
  * jpeg_create_compress() and jpeg_create_decompress() are the exported
@@ -982,97 +903,89 @@
  */
 #define jpeg_create_compress(cinfo) \
     jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \
-			(size_t) sizeof(struct jpeg_compress_struct))
+                        (size_t) sizeof(struct jpeg_compress_struct))
 #define jpeg_create_decompress(cinfo) \
     jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \
-			  (size_t) sizeof(struct jpeg_decompress_struct))
-EXTERN(void) jpeg_CreateCompress JPP((j_compress_ptr cinfo,
-				      int version, size_t structsize));
-EXTERN(void) jpeg_CreateDecompress JPP((j_decompress_ptr cinfo,
-					int version, size_t structsize));
+                          (size_t) sizeof(struct jpeg_decompress_struct))
+EXTERN(void) jpeg_CreateCompress (j_compress_ptr cinfo, int version,
+                                  size_t structsize);
+EXTERN(void) jpeg_CreateDecompress (j_decompress_ptr cinfo, int version,
+                                    size_t structsize);
 /* Destruction of JPEG compression objects */
-EXTERN(void) jpeg_destroy_compress JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_destroy_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_destroy_compress (j_compress_ptr cinfo);
+EXTERN(void) jpeg_destroy_decompress (j_decompress_ptr cinfo);
 
 /* Standard data source and destination managers: stdio streams. */
 /* Caller is responsible for opening the file before and closing after. */
-EXTERN(void) jpeg_stdio_dest JPP((j_compress_ptr cinfo, FILE * outfile));
-EXTERN(void) jpeg_stdio_src JPP((j_decompress_ptr cinfo, FILE * infile));
+EXTERN(void) jpeg_stdio_dest (j_compress_ptr cinfo, FILE *outfile);
+EXTERN(void) jpeg_stdio_src (j_decompress_ptr cinfo, FILE *infile);
 
-#if JPEG_LIB_VERSION >= 80
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
 /* Data source and destination managers: memory buffers. */
-EXTERN(void) jpeg_mem_dest JPP((j_compress_ptr cinfo,
-			       unsigned char ** outbuffer,
-			       unsigned long * outsize));
-EXTERN(void) jpeg_mem_src JPP((j_decompress_ptr cinfo,
-			      unsigned char * inbuffer,
-			      unsigned long insize));
+EXTERN(void) jpeg_mem_dest (j_compress_ptr cinfo, unsigned char **outbuffer,
+                            unsigned long *outsize);
+EXTERN(void) jpeg_mem_src (j_decompress_ptr cinfo,
+                           const unsigned char *inbuffer,
+                           unsigned long insize);
 #endif
 
 /* Default parameter setup for compression */
-EXTERN(void) jpeg_set_defaults JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_set_defaults (j_compress_ptr cinfo);
 /* Compression parameter setup aids */
-EXTERN(void) jpeg_set_colorspace JPP((j_compress_ptr cinfo,
-				      J_COLOR_SPACE colorspace));
-EXTERN(void) jpeg_default_colorspace JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_set_quality JPP((j_compress_ptr cinfo, int quality,
-				   boolean force_baseline));
-EXTERN(void) jpeg_set_linear_quality JPP((j_compress_ptr cinfo,
-					  int scale_factor,
-					  boolean force_baseline));
+EXTERN(void) jpeg_set_colorspace (j_compress_ptr cinfo,
+                                  J_COLOR_SPACE colorspace);
+EXTERN(void) jpeg_default_colorspace (j_compress_ptr cinfo);
+EXTERN(void) jpeg_set_quality (j_compress_ptr cinfo, int quality,
+                               boolean force_baseline);
+EXTERN(void) jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+                                      boolean force_baseline);
 #if JPEG_LIB_VERSION >= 70
-EXTERN(void) jpeg_default_qtables JPP((j_compress_ptr cinfo,
-				       boolean force_baseline));
+EXTERN(void) jpeg_default_qtables (j_compress_ptr cinfo,
+                                   boolean force_baseline);
 #endif
-EXTERN(void) jpeg_add_quant_table JPP((j_compress_ptr cinfo, int which_tbl,
-				       const unsigned int *basic_table,
-				       int scale_factor,
-				       boolean force_baseline));
-EXTERN(int) jpeg_quality_scaling JPP((int quality));
-EXTERN(void) jpeg_simple_progression JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_suppress_tables JPP((j_compress_ptr cinfo,
-				       boolean suppress));
-EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table JPP((j_common_ptr cinfo));
-EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+                                   const unsigned int *basic_table,
+                                   int scale_factor, boolean force_baseline);
+EXTERN(int) jpeg_quality_scaling (int quality);
+EXTERN(void) jpeg_simple_progression (j_compress_ptr cinfo);
+EXTERN(void) jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress);
+EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table (j_common_ptr cinfo);
+EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table (j_common_ptr cinfo);
 
 /* Main entry points for compression */
-EXTERN(void) jpeg_start_compress JPP((j_compress_ptr cinfo,
-				      boolean write_all_tables));
-EXTERN(JDIMENSION) jpeg_write_scanlines JPP((j_compress_ptr cinfo,
-					     JSAMPARRAY scanlines,
-					     JDIMENSION num_lines));
-EXTERN(void) jpeg_finish_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_start_compress (j_compress_ptr cinfo,
+                                  boolean write_all_tables);
+EXTERN(JDIMENSION) jpeg_write_scanlines (j_compress_ptr cinfo,
+                                         JSAMPARRAY scanlines,
+                                         JDIMENSION num_lines);
+EXTERN(void) jpeg_finish_compress (j_compress_ptr cinfo);
 
 #if JPEG_LIB_VERSION >= 70
 /* Precalculate JPEG dimensions for current compression parameters. */
-EXTERN(void) jpeg_calc_jpeg_dimensions JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_calc_jpeg_dimensions (j_compress_ptr cinfo);
 #endif
 
 /* Replaces jpeg_write_scanlines when writing raw downsampled data. */
-EXTERN(JDIMENSION) jpeg_write_raw_data JPP((j_compress_ptr cinfo,
-					    JSAMPIMAGE data,
-					    JDIMENSION num_lines));
+EXTERN(JDIMENSION) jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
+                                        JDIMENSION num_lines);
 
 /* Write a special marker.  See libjpeg.txt concerning safe usage. */
-EXTERN(void) jpeg_write_marker
-	JPP((j_compress_ptr cinfo, int marker,
-	     const JOCTET * dataptr, unsigned int datalen));
+EXTERN(void) jpeg_write_marker (j_compress_ptr cinfo, int marker,
+                                const JOCTET *dataptr, unsigned int datalen);
 /* Same, but piecemeal. */
-EXTERN(void) jpeg_write_m_header
-	JPP((j_compress_ptr cinfo, int marker, unsigned int datalen));
-EXTERN(void) jpeg_write_m_byte
-	JPP((j_compress_ptr cinfo, int val));
+EXTERN(void) jpeg_write_m_header (j_compress_ptr cinfo, int marker,
+                                  unsigned int datalen);
+EXTERN(void) jpeg_write_m_byte (j_compress_ptr cinfo, int val);
 
 /* Alternate compression function: just write an abbreviated table file */
-EXTERN(void) jpeg_write_tables JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_write_tables (j_compress_ptr cinfo);
 
 /* Decompression startup: read start of JPEG datastream to see what's there */
-EXTERN(int) jpeg_read_header JPP((j_decompress_ptr cinfo,
-				  boolean require_image));
+EXTERN(int) jpeg_read_header (j_decompress_ptr cinfo, boolean require_image);
 /* Return value is one of: */
-#define JPEG_SUSPENDED		0 /* Suspended due to lack of input data */
-#define JPEG_HEADER_OK		1 /* Found valid image datastream */
-#define JPEG_HEADER_TABLES_ONLY	2 /* Found valid table-specs-only datastream */
+#define JPEG_SUSPENDED          0 /* Suspended due to lack of input data */
+#define JPEG_HEADER_OK          1 /* Found valid image datastream */
+#define JPEG_HEADER_TABLES_ONLY 2 /* Found valid table-specs-only datastream */
 /* If you pass require_image = TRUE (normal case), you need not check for
  * a TABLES_ONLY return code; an abbreviated file will cause an error exit.
  * JPEG_SUSPENDED is only possible if you use a data source module that can
@@ -1080,54 +993,55 @@
  */
 
 /* Main entry points for decompression */
-EXTERN(boolean) jpeg_start_decompress JPP((j_decompress_ptr cinfo));
-EXTERN(JDIMENSION) jpeg_read_scanlines JPP((j_decompress_ptr cinfo,
-					    JSAMPARRAY scanlines,
-					    JDIMENSION max_lines));
-EXTERN(boolean) jpeg_finish_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_start_decompress (j_decompress_ptr cinfo);
+EXTERN(JDIMENSION) jpeg_read_scanlines (j_decompress_ptr cinfo,
+                                        JSAMPARRAY scanlines,
+                                        JDIMENSION max_lines);
+EXTERN(JDIMENSION) jpeg_skip_scanlines (j_decompress_ptr cinfo,
+                                        JDIMENSION num_lines);
+EXTERN(void) jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                                 JDIMENSION *width);
+EXTERN(boolean) jpeg_finish_decompress (j_decompress_ptr cinfo);
 
 /* Replaces jpeg_read_scanlines when reading raw downsampled data. */
-EXTERN(JDIMENSION) jpeg_read_raw_data JPP((j_decompress_ptr cinfo,
-					   JSAMPIMAGE data,
-					   JDIMENSION max_lines));
+EXTERN(JDIMENSION) jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
+                                       JDIMENSION max_lines);
 
 /* Additional entry points for buffered-image mode. */
-EXTERN(boolean) jpeg_has_multiple_scans JPP((j_decompress_ptr cinfo));
-EXTERN(boolean) jpeg_start_output JPP((j_decompress_ptr cinfo,
-				       int scan_number));
-EXTERN(boolean) jpeg_finish_output JPP((j_decompress_ptr cinfo));
-EXTERN(boolean) jpeg_input_complete JPP((j_decompress_ptr cinfo));
-EXTERN(void) jpeg_new_colormap JPP((j_decompress_ptr cinfo));
-EXTERN(int) jpeg_consume_input JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_has_multiple_scans (j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_start_output (j_decompress_ptr cinfo, int scan_number);
+EXTERN(boolean) jpeg_finish_output (j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_input_complete (j_decompress_ptr cinfo);
+EXTERN(void) jpeg_new_colormap (j_decompress_ptr cinfo);
+EXTERN(int) jpeg_consume_input (j_decompress_ptr cinfo);
 /* Return value is one of: */
-/* #define JPEG_SUSPENDED	0    Suspended due to lack of input data */
-#define JPEG_REACHED_SOS	1 /* Reached start of new scan */
-#define JPEG_REACHED_EOI	2 /* Reached end of image */
-#define JPEG_ROW_COMPLETED	3 /* Completed one iMCU row */
-#define JPEG_SCAN_COMPLETED	4 /* Completed last iMCU row of a scan */
+/* #define JPEG_SUSPENDED       0    Suspended due to lack of input data */
+#define JPEG_REACHED_SOS        1 /* Reached start of new scan */
+#define JPEG_REACHED_EOI        2 /* Reached end of image */
+#define JPEG_ROW_COMPLETED      3 /* Completed one iMCU row */
+#define JPEG_SCAN_COMPLETED     4 /* Completed last iMCU row of a scan */
 
 /* Precalculate output dimensions for current decompression parameters. */
 #if JPEG_LIB_VERSION >= 80
-EXTERN(void) jpeg_core_output_dimensions JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_core_output_dimensions (j_decompress_ptr cinfo);
 #endif
-EXTERN(void) jpeg_calc_output_dimensions JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_calc_output_dimensions (j_decompress_ptr cinfo);
 
 /* Control saving of COM and APPn markers into marker_list. */
-EXTERN(void) jpeg_save_markers
-	JPP((j_decompress_ptr cinfo, int marker_code,
-	     unsigned int length_limit));
+EXTERN(void) jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
+                                unsigned int length_limit);
 
 /* Install a special processing method for COM or APPn markers. */
-EXTERN(void) jpeg_set_marker_processor
-	JPP((j_decompress_ptr cinfo, int marker_code,
-	     jpeg_marker_parser_method routine));
+EXTERN(void) jpeg_set_marker_processor (j_decompress_ptr cinfo,
+                                        int marker_code,
+                                        jpeg_marker_parser_method routine);
 
 /* Read or write raw DCT coefficients --- useful for lossless transcoding. */
-EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients JPP((j_decompress_ptr cinfo));
-EXTERN(void) jpeg_write_coefficients JPP((j_compress_ptr cinfo,
-					  jvirt_barray_ptr * coef_arrays));
-EXTERN(void) jpeg_copy_critical_parameters JPP((j_decompress_ptr srcinfo,
-						j_compress_ptr dstinfo));
+EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients (j_decompress_ptr cinfo);
+EXTERN(void) jpeg_write_coefficients (j_compress_ptr cinfo,
+                                      jvirt_barray_ptr *coef_arrays);
+EXTERN(void) jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
+                                            j_compress_ptr dstinfo);
 
 /* If you choose to abort compression or decompression before completing
  * jpeg_finish_(de)compress, then you need to clean up to release memory,
@@ -1135,28 +1049,27 @@
  * if you're done with the JPEG object, but if you want to clean it up and
  * reuse it, call this:
  */
-EXTERN(void) jpeg_abort_compress JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_abort_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_abort_compress (j_compress_ptr cinfo);
+EXTERN(void) jpeg_abort_decompress (j_decompress_ptr cinfo);
 
 /* Generic versions of jpeg_abort and jpeg_destroy that work on either
  * flavor of JPEG object.  These may be more convenient in some places.
  */
-EXTERN(void) jpeg_abort JPP((j_common_ptr cinfo));
-EXTERN(void) jpeg_destroy JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_abort (j_common_ptr cinfo);
+EXTERN(void) jpeg_destroy (j_common_ptr cinfo);
 
 /* Default restart-marker-resync procedure for use by data source modules */
-EXTERN(boolean) jpeg_resync_to_restart JPP((j_decompress_ptr cinfo,
-					    int desired));
+EXTERN(boolean) jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired);
 
 
 /* These marker codes are exported since applications and data source modules
  * are likely to want to use them.
  */
 
-#define JPEG_RST0	0xD0	/* RST0 marker code */
-#define JPEG_EOI	0xD9	/* EOI marker code */
-#define JPEG_APP0	0xE0	/* APP0 marker code */
-#define JPEG_COM	0xFE	/* COM marker code */
+#define JPEG_RST0       0xD0    /* RST0 marker code */
+#define JPEG_EOI        0xD9    /* EOI marker code */
+#define JPEG_APP0       0xE0    /* APP0 marker code */
+#define JPEG_COM        0xFE    /* COM marker code */
 
 
 /* If we have a brain-damaged compiler that emits warnings (or worse, errors)
@@ -1165,7 +1078,7 @@
  */
 
 #ifdef INCOMPLETE_TYPES_BROKEN
-#ifndef JPEG_INTERNALS		/* will be defined in jpegint.h */
+#ifndef JPEG_INTERNALS          /* will be defined in jpegint.h */
 struct jvirt_sarray_control { long dummy; };
 struct jvirt_barray_control { long dummy; };
 struct jpeg_comp_master { long dummy; };
@@ -1200,8 +1113,8 @@
  */
 
 #ifdef JPEG_INTERNALS
-#include "jpegint.h"		/* fetch private declarations */
-#include "jerror.h"		/* fetch error codes too */
+#include "jpegint.h"            /* fetch private declarations */
+#include "jerror.h"             /* fetch error codes too */
 #endif
 
 #ifdef __cplusplus
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeglibmangler.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeglibmangler.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeglibmangler.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeglibmangler.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,117 @@
+// Copyright (c) 2009 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+#define THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+
+// Mangle all externally visible function names so we can build our own libjpeg
+// without system libraries trying to use it.
+
+#define jpeg_make_c_derived_tbl chromium_jpeg_make_c_derived_tbl
+#define jpeg_gen_optimal_table chromium_jpeg_gen_optimal_table
+#define jpeg_make_d_derived_tbl chromium_jpeg_make_d_derived_tbl
+#define jpeg_fill_bit_buffer chromium_jpeg_fill_bit_buffer
+#define jpeg_huff_decode chromium_jpeg_huff_decode
+#define jpeg_fdct_islow chromium_jpeg_fdct_islow
+#define jpeg_fdct_ifast chromium_jpeg_fdct_ifast
+#define jpeg_fdct_float chromium_jpeg_fdct_float
+#define jpeg_idct_islow chromium_jpeg_idct_islow
+#define jpeg_idct_ifast chromium_jpeg_idct_ifast
+#define jpeg_idct_float chromium_jpeg_idct_float
+#define jpeg_idct_4x4 chromium_jpeg_idct_4x4
+#define jpeg_idct_2x2 chromium_jpeg_idct_2x2
+#define jpeg_idct_1x1 chromium_jpeg_idct_1x1
+#define jinit_compress_master chromium_jinit_compress_master
+#define jinit_c_master_control chromium_jinit_c_master_control
+#define jinit_c_main_controller chromium_jinit_c_main_controller
+#define jinit_c_prep_controller chromium_jinit_c_prep_controller
+#define jinit_c_coef_controller chromium_jinit_c_coef_controller
+#define jinit_color_converter chromium_jinit_color_converter
+#define jinit_downsampler chromium_jinit_downsampler
+#define jinit_forward_dct chromium_jinit_forward_dct
+#define jinit_huff_encoder chromium_jinit_huff_encoder
+#define jinit_phuff_encoder chromium_jinit_phuff_encoder
+#define jinit_marker_writer chromium_jinit_marker_writer
+#define jinit_master_decompress chromium_jinit_master_decompress
+#define jinit_d_main_controller chromium_jinit_d_main_controller
+#define jinit_d_coef_controller chromium_jinit_d_coef_controller
+#define jinit_d_post_controller chromium_jinit_d_post_controller
+#define jinit_input_controller chromium_jinit_input_controller
+#define jinit_marker_reader chromium_jinit_marker_reader
+#define jinit_huff_decoder chromium_jinit_huff_decoder
+#define jinit_phuff_decoder chromium_jinit_phuff_decoder
+#define jinit_inverse_dct chromium_jinit_inverse_dct
+#define jinit_upsampler chromium_jinit_upsampler
+#define jinit_color_deconverter chromium_jinit_color_deconverter
+#define jinit_1pass_quantizer chromium_jinit_1pass_quantizer
+#define jinit_2pass_quantizer chromium_jinit_2pass_quantizer
+#define jinit_merged_upsampler chromium_jinit_merged_upsampler
+#define jinit_memory_mgr chromium_jinit_memory_mgr
+#define jdiv_round_up chromium_jdiv_round_up
+#define jround_up chromium_jround_up
+#define jcopy_sample_rows chromium_jcopy_sample_rows
+#define jcopy_block_row chromium_jcopy_block_row
+#define jzero_far chromium_jzero_far
+#define jpeg_std_error chromium_jpeg_std_error
+#define jpeg_CreateCompress chromium_jpeg_CreateCompress
+#define jpeg_CreateDecompress chromium_jpeg_CreateDecompress
+#define jpeg_destroy_compress chromium_jpeg_destroy_compress
+#define jpeg_destroy_decompress chromium_jpeg_destroy_decompress
+#define jpeg_stdio_dest chromium_jpeg_stdio_dest
+#define jpeg_stdio_src chromium_jpeg_stdio_src
+#define jpeg_set_defaults chromium_jpeg_set_defaults
+#define jpeg_set_colorspace chromium_jpeg_set_colorspace
+#define jpeg_default_colorspace chromium_jpeg_default_colorspace
+#define jpeg_set_quality chromium_jpeg_set_quality
+#define jpeg_set_linear_quality chromium_jpeg_set_linear_quality
+#define jpeg_add_quant_table chromium_jpeg_add_quant_table
+#define jpeg_quality_scaling chromium_jpeg_quality_scaling
+#define jpeg_simple_progression chromium_jpeg_simple_progression
+#define jpeg_suppress_tables chromium_jpeg_suppress_tables
+#define jpeg_alloc_quant_table chromium_jpeg_alloc_quant_table
+#define jpeg_alloc_huff_table chromium_jpeg_alloc_huff_table
+#define jpeg_start_compress chromium_jpeg_start_compress
+#define jpeg_write_scanlines chromium_jpeg_write_scanlines
+#define jpeg_finish_compress chromium_jpeg_finish_compress
+#define jpeg_write_raw_data chromium_jpeg_write_raw_data
+#define jpeg_write_marker chromium_jpeg_write_marker
+#define jpeg_write_m_header chromium_jpeg_write_m_header
+#define jpeg_write_m_byte chromium_jpeg_write_m_byte
+#define jpeg_write_tables chromium_jpeg_write_tables
+#define jpeg_read_header chromium_jpeg_read_header
+#define jpeg_start_decompress chromium_jpeg_start_decompress
+#define jpeg_read_scanlines chromium_jpeg_read_scanlines
+#define jpeg_skip_scanlines chromium_jpeg_skip_scanlines
+#define jpeg_crop_scanline chromium_jpeg_crop_scanline
+#define jpeg_finish_decompress chromium_jpeg_finish_decompress
+#define jpeg_read_raw_data chromium_jpeg_read_raw_data
+#define jpeg_has_multiple_scans chromium_jpeg_has_multiple_scans
+#define jpeg_start_output chromium_jpeg_start_output
+#define jpeg_finish_output chromium_jpeg_finish_output
+#define jpeg_input_complete chromium_jpeg_input_complete
+#define jpeg_new_colormap chromium_jpeg_new_colormap
+#define jpeg_consume_input chromium_jpeg_consume_input
+#define jpeg_calc_output_dimensions chromium_jpeg_calc_output_dimensions
+#define jpeg_save_markers chromium_jpeg_save_markers
+#define jpeg_set_marker_processor chromium_jpeg_set_marker_processor
+#define jpeg_read_coefficients chromium_jpeg_read_coefficients
+#define jpeg_write_coefficients chromium_jpeg_write_coefficients
+#define jpeg_copy_critical_parameters chromium_jpeg_copy_critical_parameters
+#define jpeg_abort_compress chromium_jpeg_abort_compress
+#define jpeg_abort_decompress chromium_jpeg_abort_decompress
+#define jpeg_abort chromium_jpeg_abort
+#define jpeg_destroy chromium_jpeg_destroy
+#define jpeg_resync_to_restart chromium_jpeg_resync_to_restart
+#define jpeg_get_small chromium_jpeg_get_small
+#define jpeg_free_small chromium_jpeg_free_small
+#define jpeg_get_large chromium_jpeg_get_large
+#define jpeg_free_large chromium_jpeg_free_large
+#define jpeg_mem_available chromium_jpeg_mem_available
+#define jpeg_open_backing_store chromium_jpeg_open_backing_store
+#define jpeg_mem_init chromium_jpeg_mem_init
+#define jpeg_mem_term chromium_jpeg_mem_term
+#define jpeg_std_message_table chromium_jpeg_std_message_table
+#define jpeg_natural_order chromium_jpeg_natural_order
+
+#endif  // THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeg_nbits_table.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeg_nbits_table.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpeg_nbits_table.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpeg_nbits_table.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4098 @@
+static const unsigned char jpeg_nbits_table[65536] = {
+   0,  1,  2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,
+   5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegtran.1 glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegtran.1
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegtran.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegtran.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,290 @@
+.TH JPEGTRAN 1 "18 February 2016"
+.SH NAME
+jpegtran \- lossless transformation of JPEG files
+.SH SYNOPSIS
+.B jpegtran
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B jpegtran
+performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+.PP
+For EXIF files and JPEG files containing Exif data, you may prefer to use
+.B exiftran
+instead.
+.PP
+.B jpegtran
+works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+.B djpeg
+followed by
+.B cjpeg
+to accomplish the same conversion.  But by the same token,
+.B jpegtran
+cannot perform lossy operations such as changing the image quality.  However,
+while the image data is losslessly transformed, metadata can be removed.  See
+the
+.B \-copy
+option for specifics.
+.PP
+.B jpegtran
+reads the named JPEG/JFIF file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-optimize
+may be written
+.B \-opt
+or
+.BR \-o .
+Upper and lower case are equivalent.
+British spellings are also accepted (e.g.,
+.BR \-optimise ),
+though for brevity these are not mentioned below.
+.PP
+To specify the coded JPEG representation used in the output file,
+.B jpegtran
+accepts a subset of the switches recognized by
+.BR cjpeg :
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.
+.TP
+.B \-progressive
+Create progressive JPEG file.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+See
+.BR cjpeg (1)
+for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+.PP
+The image can be losslessly transformed by giving one of these switches:
+.TP
+.B \-flip horizontal
+Mirror image horizontally (left-right).
+.TP
+.B \-flip vertical
+Mirror image vertically (top-bottom).
+.TP
+.B \-rotate 90
+Rotate image 90 degrees clockwise.
+.TP
+.B \-rotate 180
+Rotate image 180 degrees.
+.TP
+.B \-rotate 270
+Rotate image 270 degrees clockwise (or 90 ccw).
+.TP
+.B \-transpose
+Transpose image (across UL-to-LR axis).
+.TP
+.B \-transverse
+Transverse transpose (across UR-to-LL axis).
+.PP
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+.PP
+.BR jpegtran 's
+default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+.PP
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the
+.B \-trim
+switch:
+.TP
+.B \-trim
+Drop non-transformable edge blocks.
+.IP
+Obviously, a transformation with
+.B \-trim
+is not reversible, so strictly speaking
+.B jpegtran
+with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+.B \-rot 270 -trim
+trims only the bottom edge, but
+.B \-rot 90 -trim
+followed by
+.B \-rot 180 -trim
+trims both edges.
+.TP
+.B \-perfect
+If you are only interested in perfect transformations, add the
+.B \-perfect
+switch.  This causes
+.B jpegtran
+to fail with an error if the transformation is not perfect.
+.IP
+For example, you may want to do
+.IP
+.B (jpegtran \-rot 90 -perfect
+.I foo.jpg
+.B || djpeg
+.I foo.jpg
+.B | pnmflip \-r90 | cjpeg)
+.IP
+to do a perfect rotation, if available, or an approximated one if not.
+.PP
+This version of \fBjpegtran\fR also offers a lossless crop option, which
+discards data outside of a given image region but losslessly preserves what is
+inside. Like the rotate and flip transforms, lossless crop is restricted by the
+current JPEG format; the upper left corner of the selected region must fall on
+an iMCU boundary.  If it doesn't, then it is silently moved up and/or left to
+the nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the
+output image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching
+an 'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+.TP
+.B \-crop WxH+X+Y
+Crop the image to a rectangular region of width W and height H, starting at
+point X,Y.  The lossless crop feature discards data outside of a given image
+region but losslessly preserves what is inside.  Like the rotate and flip
+transforms, lossless crop is restricted by the current JPEG format; the upper
+left corner of the selected region must fall on an iMCU boundary.  If it
+doesn't, then it is silently moved up and/or left to the nearest iMCU boundary
+(the lower right corner is unchanged.)
+.PP
+Other not-strictly-lossless transformation switches are:
+.TP
+.B \-grayscale
+Force grayscale output.
+.IP
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+.PP
+.B jpegtran
+also recognizes these switches that control what to do with "extra" markers,
+such as comment blocks:
+.TP
+.B \-copy none
+Copy no extra markers from source file.  This setting suppresses all
+comments and other metadata in the source file.
+.TP
+.B \-copy comments
+Copy only comment markers.  This setting copies comments from the source file
+but discards any other metadata.
+.TP
+.B \-copy all
+Copy all extra markers.  This setting preserves miscellaneous markers
+found in the source file, such as JFIF thumbnails, Exif data, and Photoshop
+settings.  In some files, these extra markers can be sizable.  Note that this
+option will copy thumbnails as-is; they will not be transformed.
+.PP
+The default behavior is \fB-copy comments\fR.  (Note: in IJG releases v6 and
+v6a, \fBjpegtran\fR always did the equivalent of \fB-copy none\fR.)
+.PP
+Additional switches recognized by jpegtran are:
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example converts a baseline JPEG file to progressive form:
+.IP
+.B jpegtran \-progressive
+.I foo.jpg
+.B >
+.I fooprog.jpg
+.PP
+This example rotates an image 90 degrees clockwise, discarding any
+unrotatable edge pixels:
+.IP
+.B jpegtran \-rot 90 -trim
+.I foo.jpg
+.B >
+.I foo90.jpg
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo and to wordsmith certain sections.
+.SH BUGS
+The transform options can't transform odd-size images perfectly.  Use
+.B \-trim
+or
+.B \-perfect
+if you don't like the results.
+.PP
+The entire image is read into memory and then written out again, even in
+cases where this isn't really necessary.  Expect swapping on large images,
+especially when using the more complex transform options.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegtran.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegtran.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jpegtran.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jpegtran.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,551 @@
+/*
+ * jpegtran.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-2010, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for JPEG transcoding.
+ * It is very similar to cjpeg.c, and partly to djpeg.c, but provides
+ * lossless transcoding between different JPEG file formats.  It also
+ * provides some lossless and sort-of-lossless transformations of JPEG data.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "transupp.h"           /* Support routines for jpegtran */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+static JCOPY_OPTION copyoption; /* -copy switch */
+static jpeg_transform_info transformoption; /* image transformation options */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -copy none     Copy no extra markers from source file\n");
+  fprintf(stderr, "  -copy comments Copy only comment markers (default)\n");
+  fprintf(stderr, "  -copy all      Copy all extra markers\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+  fprintf(stderr, "Switches for modifying the image:\n");
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -crop WxH+X+Y  Crop to a rectangular subarea\n");
+  fprintf(stderr, "  -grayscale     Reduce to grayscale (omit color data)\n");
+  fprintf(stderr, "  -flip [horizontal|vertical]  Mirror image (left-right or top-bottom)\n");
+  fprintf(stderr, "  -perfect       Fail if there is non-transformable edge blocks\n");
+  fprintf(stderr, "  -rotate [90|180|270]         Rotate image (degrees clockwise)\n");
+#endif
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -transpose     Transpose image\n");
+  fprintf(stderr, "  -transverse    Transverse transpose image\n");
+  fprintf(stderr, "  -trim          Drop non-transformable edge blocks\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(void)
+select_transform (JXFORM_CODE transform)
+/* Silly little routine to detect multiple transform options,
+ * which we can't handle.
+ */
+{
+#if TRANSFORMS_SUPPORTED
+  if (transformoption.transform == JXFORM_NONE ||
+      transformoption.transform == transform) {
+    transformoption.transform = transform;
+  } else {
+    fprintf(stderr, "%s: can only do one image transformation at a time\n",
+            progname);
+    usage();
+  }
+#else
+  fprintf(stderr, "%s: sorry, image transformation was not compiled\n",
+          progname);
+  exit(EXIT_FAILURE);
+#endif
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean simple_progressive;
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+  simple_progressive = FALSE;
+  outfilename = NULL;
+  copyoption = JCOPYOPT_DEFAULT;
+  transformoption.transform = JXFORM_NONE;
+  transformoption.perfect = FALSE;
+  transformoption.trim = FALSE;
+  transformoption.force_grayscale = FALSE;
+  transformoption.crop = FALSE;
+  transformoption.slow_hflip = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "copy", 2)) {
+      /* Select which extra markers to copy. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "none", 1)) {
+        copyoption = JCOPYOPT_NONE;
+      } else if (keymatch(argv[argn], "comments", 1)) {
+        copyoption = JCOPYOPT_COMMENTS;
+      } else if (keymatch(argv[argn], "all", 1)) {
+        copyoption = JCOPYOPT_ALL;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "crop", 2)) {
+      /* Perform lossless cropping. */
+#if TRANSFORMS_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (! jtransform_parse_crop_spec(&transformoption, argv[argn])) {
+        fprintf(stderr, "%s: bogus -crop argument '%s'\n",
+                progname, argv[argn]);
+        exit(EXIT_FAILURE);
+      }
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "flip", 1)) {
+      /* Mirror left-right or top-bottom. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "horizontal", 1))
+        select_transform(JXFORM_FLIP_H);
+      else if (keymatch(argv[argn], "vertical", 1))
+        select_transform(JXFORM_FLIP_V);
+      else
+        usage();
+
+    } else if (keymatch(arg, "grayscale", 1) || keymatch(arg, "greyscale",1)) {
+      /* Force to grayscale. */
+#if TRANSFORMS_SUPPORTED
+      transformoption.force_grayscale = TRUE;
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "perfect", 2)) {
+      /* Fail if there is any partial edge MCUs that the transform can't
+       * handle. */
+      transformoption.perfect = TRUE;
+
+    } else if (keymatch(arg, "progressive", 2)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int) lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int) lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "rotate", 2)) {
+      /* Rotate 90, 180, or 270 degrees (measured clockwise). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "90", 2))
+        select_transform(JXFORM_ROT_90);
+      else if (keymatch(argv[argn], "180", 3))
+        select_transform(JXFORM_ROT_180);
+      else if (keymatch(argv[argn], "270", 3))
+        select_transform(JXFORM_ROT_270);
+      else
+        usage();
+
+    } else if (keymatch(arg, "scans", 1)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "transpose", 1)) {
+      /* Transpose (across UL-to-LR axis). */
+      select_transform(JXFORM_TRANSPOSE);
+
+    } else if (keymatch(arg, "transverse", 6)) {
+      /* Transverse transpose (across UR-to-LL axis). */
+      select_transform(JXFORM_TRANSVERSE);
+
+    } else if (keymatch(arg, "trim", 3)) {
+      /* Trim off any partial edge MCUs that the transform can't handle. */
+      transformoption.trim = TRUE;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct srcinfo;
+  struct jpeg_compress_struct dstinfo;
+  struct jpeg_error_mgr jsrcerr, jdsterr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  jvirt_barray_ptr *src_coef_arrays;
+  jvirt_barray_ptr *dst_coef_arrays;
+  int file_index;
+  /* We assume all-in-memory processing and can therefore use only a
+   * single file pointer for sequential input and output operation.
+   */
+  FILE *fp;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "jpegtran";      /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  srcinfo.err = jpeg_std_error(&jsrcerr);
+  jpeg_create_decompress(&srcinfo);
+  /* Initialize the JPEG compression object with default error handling. */
+  dstinfo.err = jpeg_std_error(&jdsterr);
+  jpeg_create_compress(&dstinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are mostly ignored; we will rescan the switches after
+   * opening the input file.  Also note that most of the switches affect the
+   * destination JPEG object, so we parse into that and then copy over what
+   * needs to affects the source too.
+   */
+
+  file_index = parse_switches(&dstinfo, argc, argv, 0, FALSE);
+  jsrcerr.trace_level = jdsterr.trace_level;
+  srcinfo.mem->max_memory_to_use = dstinfo.mem->max_memory_to_use;
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((fp = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for reading\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    fp = read_stdin();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &dstinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+  jpeg_stdio_src(&srcinfo, fp);
+
+  /* Enable saving of extra markers that we want to copy */
+  jcopy_markers_setup(&srcinfo, copyoption);
+
+  /* Read file header */
+  (void) jpeg_read_header(&srcinfo, TRUE);
+
+  /* Any space needed by a transform option must be requested before
+   * jpeg_read_coefficients so that memory allocation will be done right.
+   */
+#if TRANSFORMS_SUPPORTED
+  /* Fail right away if -perfect is given and transformation is not perfect.
+   */
+  if (!jtransform_request_workspace(&srcinfo, &transformoption)) {
+    fprintf(stderr, "%s: transformation is not perfect\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#endif
+
+  /* Read source file as DCT coefficients */
+  src_coef_arrays = jpeg_read_coefficients(&srcinfo);
+
+  /* Initialize destination compression parameters from source values */
+  jpeg_copy_critical_parameters(&srcinfo, &dstinfo);
+
+  /* Adjust destination parameters if required by transform options;
+   * also find out which set of coefficient arrays will hold the output.
+   */
+#if TRANSFORMS_SUPPORTED
+  dst_coef_arrays = jtransform_adjust_parameters(&srcinfo, &dstinfo,
+                                                 src_coef_arrays,
+                                                 &transformoption);
+#else
+  dst_coef_arrays = src_coef_arrays;
+#endif
+
+  /* Close input file, if we opened it.
+   * Note: we assume that jpeg_read_coefficients consumed all input
+   * until JPEG_REACHED_EOI, and that jpeg_finish_decompress will
+   * only consume more while (! cinfo->inputctl->eoi_reached).
+   * We cannot call jpeg_finish_decompress here since we still need the
+   * virtual arrays allocated from the source object for processing.
+   */
+  if (fp != stdin)
+    fclose(fp);
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((fp = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for writing\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    fp = write_stdout();
+  }
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&dstinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+  jpeg_stdio_dest(&dstinfo, fp);
+
+  /* Start compressor (note no image data is actually written here) */
+  jpeg_write_coefficients(&dstinfo, dst_coef_arrays);
+
+  /* Copy to the output file any extra markers that we want to preserve */
+  jcopy_markers_execute(&srcinfo, &dstinfo, copyoption);
+
+  /* Execute image transformation, if any */
+#if TRANSFORMS_SUPPORTED
+  jtransform_execute_transformation(&srcinfo, &dstinfo,
+                                    src_coef_arrays,
+                                    &transformoption);
+#endif
+
+  /* Finish compression and release memory */
+  jpeg_finish_compress(&dstinfo);
+  jpeg_destroy_compress(&dstinfo);
+  (void) jpeg_finish_decompress(&srcinfo);
+  jpeg_destroy_decompress(&srcinfo);
+
+  /* Close output file, if we opened it */
+  if (fp != stdout)
+    fclose(fp);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &dstinfo);
+#endif
+
+  /* All done. */
+  exit(jsrcerr.num_warnings + jdsterr.num_warnings ?EXIT_WARNING:EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jquant1.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jquant1.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jquant1.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jquant1.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jquant1.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009, D. R. Commander
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains 1-pass color quantization (color mapping) routines.
  * These routines provide mapping to a fixed color map using equally spaced
@@ -69,9 +71,9 @@
  * table in both directions.
  */
 
-#define ODITHER_SIZE  16	/* dimension of dither matrix */
+#define ODITHER_SIZE  16        /* dimension of dither matrix */
 /* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */
-#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)	/* # cells in matrix */
+#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)       /* # cells in matrix */
 #define ODITHER_MASK  (ODITHER_SIZE-1) /* mask for wrapping around counters */
 
 typedef int ODITHER_MATRIX[ODITHER_SIZE][ODITHER_SIZE];
@@ -106,8 +108,8 @@
  * Errors are accumulated into the array fserrors[], at a resolution of
  * 1/16th of a pixel count.  The error at a given pixel is propagated
  * to its not-yet-processed neighbors using the standard F-S fractions,
- *		...	(here)	7/16
- *		3/16	5/16	1/16
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
  * We work left-to-right on even rows, right-to-left on odd rows.
  *
  * We can get away with a single array (holding one row's worth of errors)
@@ -120,52 +122,49 @@
  * The fserrors[] array is indexed [component#][position].
  * We provide (#columns + 2) entries per component; the extra entry at each
  * end saves us from special-casing the first and last pixels.
- *
- * Note: on a wide image, we might not have enough room in a PC's near data
- * segment to hold the error array; so it is allocated with alloc_large.
  */
 
 #if BITS_IN_JSAMPLE == 8
-typedef INT16 FSERROR;		/* 16 bits should be enough */
-typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
 #else
-typedef INT32 FSERROR;		/* may need more than 16 bits */
-typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
 #endif
 
-typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+typedef FSERROR *FSERRPTR;  /* pointer to error array */
 
 
 /* Private subobject */
 
-#define MAX_Q_COMPS 4		/* max components I can handle */
+#define MAX_Q_COMPS 4           /* max components I can handle */
 
 typedef struct {
   struct jpeg_color_quantizer pub; /* public fields */
 
   /* Initially allocated colormap is saved here */
-  JSAMPARRAY sv_colormap;	/* The color map as a 2-D pixel array */
-  int sv_actual;		/* number of entries in use */
+  JSAMPARRAY sv_colormap;       /* The color map as a 2-D pixel array */
+  int sv_actual;                /* number of entries in use */
 
-  JSAMPARRAY colorindex;	/* Precomputed mapping for speed */
+  JSAMPARRAY colorindex;        /* Precomputed mapping for speed */
   /* colorindex[i][j] = index of color closest to pixel value j in component i,
    * premultiplied as described above.  Since colormap indexes must fit into
    * JSAMPLEs, the entries of this array will too.
    */
-  boolean is_padded;		/* is the colorindex padded for odither? */
+  boolean is_padded;            /* is the colorindex padded for odither? */
 
-  int Ncolors[MAX_Q_COMPS];	/* # of values alloced to each component */
+  int Ncolors[MAX_Q_COMPS];     /* # of values alloced to each component */
 
   /* Variables for ordered dithering */
-  int row_index;		/* cur row's vertical index in dither matrix */
+  int row_index;                /* cur row's vertical index in dither matrix */
   ODITHER_MATRIX_PTR odither[MAX_Q_COMPS]; /* one dither array per component */
 
   /* Variables for Floyd-Steinberg dithering */
   FSERRPTR fserrors[MAX_Q_COMPS]; /* accumulated errors */
-  boolean on_odd_row;		/* flag to remember which row we are on */
+  boolean on_odd_row;           /* flag to remember which row we are on */
 } my_cquantizer;
 
-typedef my_cquantizer * my_cquantize_ptr;
+typedef my_cquantizer *my_cquantize_ptr;
 
 
 /*
@@ -204,11 +203,11 @@
   iroot = 1;
   do {
     iroot++;
-    temp = iroot;		/* set temp = iroot ** nc */
+    temp = iroot;               /* set temp = iroot ** nc */
     for (i = 1; i < nc; i++)
       temp *= iroot;
   } while (temp <= (long) max_colors); /* repeat till iroot exceeds root */
-  iroot--;			/* now iroot = floor(root) */
+  iroot--;                      /* now iroot = floor(root) */
 
   /* Must have at least 2 color values per component */
   if (iroot < 2)
@@ -232,10 +231,10 @@
       j = (cinfo->out_color_space == JCS_RGB ? RGB_order[i] : i);
       /* calculate new total_colors if Ncolors[j] is incremented */
       temp = total_colors / Ncolors[j];
-      temp *= Ncolors[j]+1;	/* done in long arith to avoid oflo */
+      temp *= Ncolors[j]+1;     /* done in long arith to avoid oflo */
       if (temp > (long) max_colors)
-	break;			/* won't fit, done with this pass */
-      Ncolors[j]++;		/* OK, apply the increment */
+        break;                  /* won't fit, done with this pass */
+      Ncolors[j]++;             /* OK, apply the increment */
       total_colors = (int) temp;
       changed = TRUE;
     }
@@ -255,7 +254,7 @@
    * (Forcing the upper and lower values to the limits ensures that
    * dithering can't produce a color outside the selected gamut.)
    */
-  return (int) (((INT32) j * MAXJSAMPLE + maxj/2) / maxj);
+  return (int) (((JLONG) j * MAXJSAMPLE + maxj/2) / maxj);
 }
 
 
@@ -265,7 +264,7 @@
 /* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */
 {
   /* Breakpoints are halfway between values returned by output_value */
-  return (int) (((INT32) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
+  return (int) (((JLONG) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
 }
 
 
@@ -277,8 +276,8 @@
 create_colormap (j_decompress_ptr cinfo)
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  JSAMPARRAY colormap;		/* Created colormap */
-  int total_colors;		/* Number of distinct output colors */
+  JSAMPARRAY colormap;          /* Created colormap */
+  int total_colors;             /* Number of distinct output colors */
   int i,j,k, nci, blksize, blkdist, ptr, val;
 
   /* Select number of colors for each component */
@@ -287,8 +286,8 @@
   /* Report selected color counts */
   if (cinfo->out_color_components == 3)
     TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS,
-	     total_colors, cquantize->Ncolors[0],
-	     cquantize->Ncolors[1], cquantize->Ncolors[2]);
+             total_colors, cquantize->Ncolors[0],
+             cquantize->Ncolors[1], cquantize->Ncolors[2]);
   else
     TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors);
 
@@ -313,12 +312,12 @@
       val = output_value(cinfo, i, j, nci-1);
       /* Fill in all colormap entries that have this value of this component */
       for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) {
-	/* fill in blksize entries beginning at ptr */
-	for (k = 0; k < blksize; k++)
-	  colormap[i][ptr+k] = (JSAMPLE) val;
+        /* fill in blksize entries beginning at ptr */
+        for (k = 0; k < blksize; k++)
+          colormap[i][ptr+k] = (JSAMPLE) val;
       }
     }
-    blkdist = blksize;		/* blksize of this color is blkdist of next */
+    blkdist = blksize;          /* blksize of this color is blkdist of next */
   }
 
   /* Save the colormap in private storage,
@@ -376,16 +375,16 @@
     val = 0;
     k = largest_input_value(cinfo, i, 0, nci-1);
     for (j = 0; j <= MAXJSAMPLE; j++) {
-      while (j > k)		/* advance val if past boundary */
-	k = largest_input_value(cinfo, i, ++val, nci-1);
+      while (j > k)             /* advance val if past boundary */
+        k = largest_input_value(cinfo, i, ++val, nci-1);
       /* premultiply so that no multiplication needed in main processing */
       indexptr[j] = (JSAMPLE) (val * blksize);
     }
     /* Pad at both ends if necessary */
     if (pad)
       for (j = 1; j <= MAXJSAMPLE; j++) {
-	indexptr[-j] = indexptr[0];
-	indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
+        indexptr[-j] = indexptr[0];
+        indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
       }
   }
 }
@@ -401,21 +400,21 @@
 {
   ODITHER_MATRIX_PTR odither;
   int j,k;
-  INT32 num,den;
+  JLONG num,den;
 
   odither = (ODITHER_MATRIX_PTR)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(ODITHER_MATRIX));
+                                sizeof(ODITHER_MATRIX));
   /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1).
    * Hence the dither value for the matrix cell with fill order f
    * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1).
    * On 16-bit-int machine, be careful to avoid overflow.
    */
-  den = 2 * ODITHER_CELLS * ((INT32) (ncolors - 1));
+  den = 2 * ODITHER_CELLS * ((JLONG) (ncolors - 1));
   for (j = 0; j < ODITHER_SIZE; j++) {
     for (k = 0; k < ODITHER_SIZE; k++) {
-      num = ((INT32) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
-	    * MAXJSAMPLE;
+      num = ((JLONG) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
+            * MAXJSAMPLE;
       /* Ensure round towards zero despite C's lack of consistency
        * about rounding negative values in integer division...
        */
@@ -428,7 +427,7 @@
 
 /*
  * Create the ordered-dither tables.
- * Components having the same number of representative colors may 
+ * Components having the same number of representative colors may
  * share a dither table.
  */
 
@@ -441,14 +440,14 @@
 
   for (i = 0; i < cinfo->out_color_components; i++) {
     nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
-    odither = NULL;		/* search for matching prior component */
+    odither = NULL;             /* search for matching prior component */
     for (j = 0; j < i; j++) {
       if (nci == cquantize->Ncolors[j]) {
-	odither = cquantize->odither[j];
-	break;
+        odither = cquantize->odither[j];
+        break;
       }
     }
-    if (odither == NULL)	/* need a new table? */
+    if (odither == NULL)        /* need a new table? */
       odither = make_odither_array(cinfo, nci);
     cquantize->odither[i] = odither;
   }
@@ -461,7 +460,7 @@
 
 METHODDEF(void)
 color_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		JSAMPARRAY output_buf, int num_rows)
+                JSAMPARRAY output_buf, int num_rows)
 /* General case, no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -479,7 +478,7 @@
     for (col = width; col > 0; col--) {
       pixcode = 0;
       for (ci = 0; ci < nc; ci++) {
-	pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+        pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
       }
       *ptrout++ = (JSAMPLE) pixcode;
     }
@@ -489,7 +488,7 @@
 
 METHODDEF(void)
 color_quantize3 (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		 JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY output_buf, int num_rows)
 /* Fast path for out_color_components==3, no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -517,15 +516,15 @@
 
 METHODDEF(void)
 quantize_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		     JSAMPARRAY output_buf, int num_rows)
+                     JSAMPARRAY output_buf, int num_rows)
 /* General case, with ordered dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   register JSAMPROW input_ptr;
   register JSAMPROW output_ptr;
   JSAMPROW colorindex_ci;
-  int * dither;			/* points to active row of dither matrix */
-  int row_index, col_index;	/* current indexes into dither matrix */
+  int *dither;                  /* points to active row of dither matrix */
+  int row_index, col_index;     /* current indexes into dither matrix */
   int nc = cinfo->out_color_components;
   int ci;
   int row;
@@ -534,8 +533,7 @@
 
   for (row = 0; row < num_rows; row++) {
     /* Initialize output values to 0 so can process components separately */
-    jzero_far((void FAR *) output_buf[row],
-	      (size_t) (width * SIZEOF(JSAMPLE)));
+    jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE)));
     row_index = cquantize->row_index;
     for (ci = 0; ci < nc; ci++) {
       input_ptr = input_buf[row] + ci;
@@ -545,17 +543,17 @@
       col_index = 0;
 
       for (col = width; col > 0; col--) {
-	/* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
-	 * select output value, accumulate into output code for this pixel.
-	 * Range-limiting need not be done explicitly, as we have extended
-	 * the colorindex table to produce the right answers for out-of-range
-	 * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
-	 * required amount of padding.
-	 */
-	*output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
-	input_ptr += nc;
-	output_ptr++;
-	col_index = (col_index + 1) & ODITHER_MASK;
+        /* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
+         * select output value, accumulate into output code for this pixel.
+         * Range-limiting need not be done explicitly, as we have extended
+         * the colorindex table to produce the right answers for out-of-range
+         * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
+         * required amount of padding.
+         */
+        *output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
+        input_ptr += nc;
+        output_ptr++;
+        col_index = (col_index + 1) & ODITHER_MASK;
       }
     }
     /* Advance row index for next row */
@@ -567,7 +565,7 @@
 
 METHODDEF(void)
 quantize3_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		      JSAMPARRAY output_buf, int num_rows)
+                      JSAMPARRAY output_buf, int num_rows)
 /* Fast path for out_color_components==3, with ordered dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -577,10 +575,10 @@
   JSAMPROW colorindex0 = cquantize->colorindex[0];
   JSAMPROW colorindex1 = cquantize->colorindex[1];
   JSAMPROW colorindex2 = cquantize->colorindex[2];
-  int * dither0;		/* points to active row of dither matrix */
-  int * dither1;
-  int * dither2;
-  int row_index, col_index;	/* current indexes into dither matrix */
+  int *dither0;                 /* points to active row of dither matrix */
+  int *dither1;
+  int *dither2;
+  int row_index, col_index;     /* current indexes into dither matrix */
   int row;
   JDIMENSION col;
   JDIMENSION width = cinfo->output_width;
@@ -596,11 +594,11 @@
 
     for (col = width; col > 0; col--) {
       pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) +
-					dither0[col_index]]);
+                                        dither0[col_index]]);
       pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) +
-					dither1[col_index]]);
+                                        dither1[col_index]]);
       pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) +
-					dither2[col_index]]);
+                                        dither2[col_index]]);
       *output_ptr++ = (JSAMPLE) pixcode;
       col_index = (col_index + 1) & ODITHER_MASK;
     }
@@ -612,24 +610,24 @@
 
 METHODDEF(void)
 quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		    JSAMPARRAY output_buf, int num_rows)
+                    JSAMPARRAY output_buf, int num_rows)
 /* General case, with Floyd-Steinberg dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  register LOCFSERROR cur;	/* current error or pixel value */
-  LOCFSERROR belowerr;		/* error for pixel below cur */
-  LOCFSERROR bpreverr;		/* error for below/prev col */
-  LOCFSERROR bnexterr;		/* error for below/next col */
+  register LOCFSERROR cur;      /* current error or pixel value */
+  LOCFSERROR belowerr;          /* error for pixel below cur */
+  LOCFSERROR bpreverr;          /* error for below/prev col */
+  LOCFSERROR bnexterr;          /* error for below/next col */
   LOCFSERROR delta;
-  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  register FSERRPTR errorptr;   /* => fserrors[] at column before current */
   register JSAMPROW input_ptr;
   register JSAMPROW output_ptr;
   JSAMPROW colorindex_ci;
   JSAMPROW colormap_ci;
   int pixcode;
   int nc = cinfo->out_color_components;
-  int dir;			/* 1 for left-to-right, -1 for right-to-left */
-  int dirnc;			/* dir * nc */
+  int dir;                      /* 1 for left-to-right, -1 for right-to-left */
+  int dirnc;                    /* dir * nc */
   int ci;
   int row;
   JDIMENSION col;
@@ -639,23 +637,22 @@
 
   for (row = 0; row < num_rows; row++) {
     /* Initialize output values to 0 so can process components separately */
-    jzero_far((void FAR *) output_buf[row],
-	      (size_t) (width * SIZEOF(JSAMPLE)));
+    jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE)));
     for (ci = 0; ci < nc; ci++) {
       input_ptr = input_buf[row] + ci;
       output_ptr = output_buf[row];
       if (cquantize->on_odd_row) {
-	/* work right to left in this row */
-	input_ptr += (width-1) * nc; /* so point to rightmost pixel */
-	output_ptr += width-1;
-	dir = -1;
-	dirnc = -nc;
-	errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
+        /* work right to left in this row */
+        input_ptr += (width-1) * nc; /* so point to rightmost pixel */
+        output_ptr += width-1;
+        dir = -1;
+        dirnc = -nc;
+        errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
       } else {
-	/* work left to right in this row */
-	dir = 1;
-	dirnc = nc;
-	errorptr = cquantize->fserrors[ci]; /* => entry before first column */
+        /* work left to right in this row */
+        dir = 1;
+        dirnc = nc;
+        errorptr = cquantize->fserrors[ci]; /* => entry before first column */
       }
       colorindex_ci = cquantize->colorindex[ci];
       colormap_ci = cquantize->sv_colormap[ci];
@@ -665,47 +662,47 @@
       belowerr = bpreverr = 0;
 
       for (col = width; col > 0; col--) {
-	/* cur holds the error propagated from the previous pixel on the
-	 * current line.  Add the error propagated from the previous line
-	 * to form the complete error correction term for this pixel, and
-	 * round the error term (which is expressed * 16) to an integer.
-	 * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
-	 * for either sign of the error value.
-	 * Note: errorptr points to *previous* column's array entry.
-	 */
-	cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
-	/* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
-	 * The maximum error is +- MAXJSAMPLE; this sets the required size
-	 * of the range_limit array.
-	 */
-	cur += GETJSAMPLE(*input_ptr);
-	cur = GETJSAMPLE(range_limit[cur]);
-	/* Select output value, accumulate into output code for this pixel */
-	pixcode = GETJSAMPLE(colorindex_ci[cur]);
-	*output_ptr += (JSAMPLE) pixcode;
-	/* Compute actual representation error at this pixel */
-	/* Note: we can do this even though we don't have the final */
-	/* pixel code, because the colormap is orthogonal. */
-	cur -= GETJSAMPLE(colormap_ci[pixcode]);
-	/* Compute error fractions to be propagated to adjacent pixels.
-	 * Add these into the running sums, and simultaneously shift the
-	 * next-line error sums left by 1 column.
-	 */
-	bnexterr = cur;
-	delta = cur * 2;
-	cur += delta;		/* form error * 3 */
-	errorptr[0] = (FSERROR) (bpreverr + cur);
-	cur += delta;		/* form error * 5 */
-	bpreverr = belowerr + cur;
-	belowerr = bnexterr;
-	cur += delta;		/* form error * 7 */
-	/* At this point cur contains the 7/16 error value to be propagated
-	 * to the next pixel on the current line, and all the errors for the
-	 * next line have been shifted over. We are therefore ready to move on.
-	 */
-	input_ptr += dirnc;	/* advance input ptr to next column */
-	output_ptr += dir;	/* advance output ptr to next column */
-	errorptr += dir;	/* advance errorptr to current column */
+        /* cur holds the error propagated from the previous pixel on the
+         * current line.  Add the error propagated from the previous line
+         * to form the complete error correction term for this pixel, and
+         * round the error term (which is expressed * 16) to an integer.
+         * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+         * for either sign of the error value.
+         * Note: errorptr points to *previous* column's array entry.
+         */
+        cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
+        /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+         * The maximum error is +- MAXJSAMPLE; this sets the required size
+         * of the range_limit array.
+         */
+        cur += GETJSAMPLE(*input_ptr);
+        cur = GETJSAMPLE(range_limit[cur]);
+        /* Select output value, accumulate into output code for this pixel */
+        pixcode = GETJSAMPLE(colorindex_ci[cur]);
+        *output_ptr += (JSAMPLE) pixcode;
+        /* Compute actual representation error at this pixel */
+        /* Note: we can do this even though we don't have the final */
+        /* pixel code, because the colormap is orthogonal. */
+        cur -= GETJSAMPLE(colormap_ci[pixcode]);
+        /* Compute error fractions to be propagated to adjacent pixels.
+         * Add these into the running sums, and simultaneously shift the
+         * next-line error sums left by 1 column.
+         */
+        bnexterr = cur;
+        delta = cur * 2;
+        cur += delta;           /* form error * 3 */
+        errorptr[0] = (FSERROR) (bpreverr + cur);
+        cur += delta;           /* form error * 5 */
+        bpreverr = belowerr + cur;
+        belowerr = bnexterr;
+        cur += delta;           /* form error * 7 */
+        /* At this point cur contains the 7/16 error value to be propagated
+         * to the next pixel on the current line, and all the errors for the
+         * next line have been shifted over. We are therefore ready to move on.
+         */
+        input_ptr += dirnc;     /* advance input ptr to next column */
+        output_ptr += dir;      /* advance output ptr to next column */
+        errorptr += dir;        /* advance errorptr to current column */
       }
       /* Post-loop cleanup: we must unload the final error value into the
        * final fserrors[] entry.  Note we need not unload belowerr because
@@ -729,7 +726,7 @@
   size_t arraysize;
   int i;
 
-  arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+  arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR));
   for (i = 0; i < cinfo->out_color_components; i++) {
     cquantize->fserrors[i] = (FSERRPTR)
       (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
@@ -765,7 +762,7 @@
       cquantize->pub.color_quantize = quantize3_ord_dither;
     else
       cquantize->pub.color_quantize = quantize_ord_dither;
-    cquantize->row_index = 0;	/* initialize state for ordered dither */
+    cquantize->row_index = 0;   /* initialize state for ordered dither */
     /* If user changed to ordered dither from another mode,
      * we must recreate the color index table with padding.
      * This will cost extra space, but probably isn't very likely.
@@ -783,9 +780,9 @@
     if (cquantize->fserrors[0] == NULL)
       alloc_fs_workspace(cinfo);
     /* Initialize the propagated errors to zero. */
-    arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+    arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR));
     for (i = 0; i < cinfo->out_color_components; i++)
-      jzero_far((void FAR *) cquantize->fserrors[i], arraysize);
+      jzero_far((void *) cquantize->fserrors[i], arraysize);
     break;
   default:
     ERREXIT(cinfo, JERR_NOT_COMPILED);
@@ -828,13 +825,13 @@
 
   cquantize = (my_cquantize_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_cquantizer));
+                                sizeof(my_cquantizer));
   cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
   cquantize->pub.start_pass = start_pass_1_quant;
   cquantize->pub.finish_pass = finish_pass_1_quant;
   cquantize->pub.new_color_map = new_color_map_1_quant;
   cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */
-  cquantize->odither[0] = NULL;	/* Also flag odither arrays not allocated */
+  cquantize->odither[0] = NULL; /* Also flag odither arrays not allocated */
 
   /* Make sure my internal arrays won't overflow */
   if (cinfo->out_color_components > MAX_Q_COMPS)
@@ -848,10 +845,10 @@
   create_colorindex(cinfo);
 
   /* Allocate Floyd-Steinberg workspace now if requested.
-   * We do this now since it is FAR storage and may affect the memory
-   * manager's space calculations.  If the user changes to FS dither
-   * mode in a later pass, we will allocate the space then, and will
-   * possibly overrun the max_memory_to_use setting.
+   * We do this now since it may affect the memory manager's space
+   * calculations.  If the user changes to FS dither mode in a later pass, we
+   * will allocate the space then, and will possibly overrun the
+   * max_memory_to_use setting.
    */
   if (cinfo->dither_mode == JDITHER_FS)
     alloc_fs_workspace(cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jquant2.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jquant2.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jquant2.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jquant2.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jquant2.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains 2-pass color quantization (color mapping) routines.
  * These routines provide selection of a custom color map for an image,
@@ -42,7 +44,7 @@
  * color space, and repeatedly splits the "largest" remaining box until we
  * have as many boxes as desired colors.  Then the mean color in each
  * remaining box becomes one of the possible output colors.
- * 
+ *
  * The second pass over the image maps each input pixel to the closest output
  * color (optionally after applying a Floyd-Steinberg dithering correction).
  * This mapping is logically trivial, but making it go fast enough requires
@@ -71,9 +73,9 @@
  * probably need to change these scale factors.
  */
 
-#define R_SCALE 2		/* scale R distances by this much */
-#define G_SCALE 3		/* scale G distances by this much */
-#define B_SCALE 1		/* and B by this much */
+#define R_SCALE 2               /* scale R distances by this much */
+#define G_SCALE 3               /* scale G distances by this much */
+#define B_SCALE 1               /* and B by this much */
 
 static const int c_scales[3]={R_SCALE, G_SCALE, B_SCALE};
 #define C0_SCALE c_scales[rgb_red[cinfo->out_color_space]]
@@ -101,9 +103,7 @@
  * machines, we can't just allocate the histogram in one chunk.  Instead
  * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
  * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
- * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.  Note that
- * on 80x86 machines, the pointer row is in near memory but the actual
- * arrays are in far memory (same arrangement as we use for image arrays).
+ * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.
  */
 
 #define MAXNUMCOLORS  (MAXJSAMPLE+1) /* maximum size of colormap */
@@ -111,9 +111,9 @@
 /* These will do the right thing for either R,G,B or B,G,R color order,
  * but you may not like the results for other color orders.
  */
-#define HIST_C0_BITS  5		/* bits of precision in R/B histogram */
-#define HIST_C1_BITS  6		/* bits of precision in G histogram */
-#define HIST_C2_BITS  5		/* bits of precision in B/R histogram */
+#define HIST_C0_BITS  5         /* bits of precision in R/B histogram */
+#define HIST_C1_BITS  6         /* bits of precision in G histogram */
+#define HIST_C2_BITS  5         /* bits of precision in B/R histogram */
 
 /* Number of elements along histogram axes. */
 #define HIST_C0_ELEMS  (1<<HIST_C0_BITS)
@@ -126,13 +126,13 @@
 #define C2_SHIFT  (BITS_IN_JSAMPLE-HIST_C2_BITS)
 
 
-typedef UINT16 histcell;	/* histogram cell; prefer an unsigned type */
+typedef UINT16 histcell;        /* histogram cell; prefer an unsigned type */
 
-typedef histcell FAR * histptr;	/* for pointers to histogram cells */
+typedef histcell *histptr; /* for pointers to histogram cells */
 
 typedef histcell hist1d[HIST_C2_ELEMS]; /* typedefs for the array */
-typedef hist1d FAR * hist2d;	/* type for the 2nd-level pointers */
-typedef hist2d * hist3d;	/* type for top-level pointer */
+typedef hist1d *hist2d;         /* type for the 2nd-level pointers */
+typedef hist2d *hist3d;         /* type for top-level pointer */
 
 
 /* Declarations for Floyd-Steinberg dithering.
@@ -140,8 +140,8 @@
  * Errors are accumulated into the array fserrors[], at a resolution of
  * 1/16th of a pixel count.  The error at a given pixel is propagated
  * to its not-yet-processed neighbors using the standard F-S fractions,
- *		...	(here)	7/16
- *		3/16	5/16	1/16
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
  * We work left-to-right on even rows, right-to-left on odd rows.
  *
  * We can get away with a single array (holding one row's worth of errors)
@@ -154,20 +154,17 @@
  * The fserrors[] array has (#columns + 2) entries; the extra entry at
  * each end saves us from special-casing the first and last pixels.
  * Each entry is three values long, one value for each color component.
- *
- * Note: on a wide image, we might not have enough room in a PC's near data
- * segment to hold the error array; so it is allocated with alloc_large.
  */
 
 #if BITS_IN_JSAMPLE == 8
-typedef INT16 FSERROR;		/* 16 bits should be enough */
-typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
 #else
-typedef INT32 FSERROR;		/* may need more than 16 bits */
-typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
 #endif
 
-typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+typedef FSERROR *FSERRPTR;      /* pointer to error array */
 
 
 /* Private subobject */
@@ -176,21 +173,21 @@
   struct jpeg_color_quantizer pub; /* public fields */
 
   /* Space for the eventually created colormap is stashed here */
-  JSAMPARRAY sv_colormap;	/* colormap allocated at init time */
-  int desired;			/* desired # of colors = size of colormap */
+  JSAMPARRAY sv_colormap;       /* colormap allocated at init time */
+  int desired;                  /* desired # of colors = size of colormap */
 
   /* Variables for accumulating image statistics */
-  hist3d histogram;		/* pointer to the histogram */
+  hist3d histogram;             /* pointer to the histogram */
 
-  boolean needs_zeroed;		/* TRUE if next pass must zero histogram */
+  boolean needs_zeroed;         /* TRUE if next pass must zero histogram */
 
   /* Variables for Floyd-Steinberg dithering */
-  FSERRPTR fserrors;		/* accumulated errors */
-  boolean on_odd_row;		/* flag to remember which row we are on */
-  int * error_limiter;		/* table for clamping the applied error */
+  FSERRPTR fserrors;            /* accumulated errors */
+  boolean on_odd_row;           /* flag to remember which row we are on */
+  int *error_limiter;           /* table for clamping the applied error */
 } my_cquantizer;
 
-typedef my_cquantizer * my_cquantize_ptr;
+typedef my_cquantizer *my_cquantize_ptr;
 
 
 /*
@@ -204,7 +201,7 @@
 
 METHODDEF(void)
 prescan_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		  JSAMPARRAY output_buf, int num_rows)
+                  JSAMPARRAY output_buf, int num_rows)
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   register JSAMPROW ptr;
@@ -219,11 +216,11 @@
     for (col = width; col > 0; col--) {
       /* get pixel value and index into the histogram */
       histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
-			 [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
-			 [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+                         [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
+                         [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
       /* increment, check for overflow and undo increment if so. */
       if (++(*histp) <= 0)
-	(*histp)--;
+        (*histp)--;
       ptr += 3;
     }
   }
@@ -243,12 +240,12 @@
   int c1min, c1max;
   int c2min, c2max;
   /* The volume (actually 2-norm) of the box */
-  INT32 volume;
+  JLONG volume;
   /* The number of nonzero histogram cells within this box */
   long colorcount;
 } box;
 
-typedef box * boxptr;
+typedef box *boxptr;
 
 
 LOCAL(boxptr)
@@ -260,7 +257,7 @@
   register int i;
   register long maxc = 0;
   boxptr which = NULL;
-  
+
   for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
     if (boxp->colorcount > maxc && boxp->volume > 0) {
       which = boxp;
@@ -278,9 +275,9 @@
 {
   register boxptr boxp;
   register int i;
-  register INT32 maxv = 0;
+  register JLONG maxv = 0;
   boxptr which = NULL;
-  
+
   for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
     if (boxp->volume > maxv) {
       which = boxp;
@@ -301,77 +298,77 @@
   histptr histp;
   int c0,c1,c2;
   int c0min,c0max,c1min,c1max,c2min,c2max;
-  INT32 dist0,dist1,dist2;
+  JLONG dist0,dist1,dist2;
   long ccount;
-  
+
   c0min = boxp->c0min;  c0max = boxp->c0max;
   c1min = boxp->c1min;  c1max = boxp->c1max;
   c2min = boxp->c2min;  c2max = boxp->c2max;
-  
+
   if (c0max > c0min)
     for (c0 = c0min; c0 <= c0max; c0++)
       for (c1 = c1min; c1 <= c1max; c1++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c0min = c0min = c0;
-	    goto have_c0min;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0min = c0min = c0;
+            goto have_c0min;
+          }
       }
  have_c0min:
   if (c0max > c0min)
     for (c0 = c0max; c0 >= c0min; c0--)
       for (c1 = c1min; c1 <= c1max; c1++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c0max = c0max = c0;
-	    goto have_c0max;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0max = c0max = c0;
+            goto have_c0max;
+          }
       }
  have_c0max:
   if (c1max > c1min)
     for (c1 = c1min; c1 <= c1max; c1++)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c1min = c1min = c1;
-	    goto have_c1min;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1min = c1min = c1;
+            goto have_c1min;
+          }
       }
  have_c1min:
   if (c1max > c1min)
     for (c1 = c1max; c1 >= c1min; c1--)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c1max = c1max = c1;
-	    goto have_c1max;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1max = c1max = c1;
+            goto have_c1max;
+          }
       }
  have_c1max:
   if (c2max > c2min)
     for (c2 = c2min; c2 <= c2max; c2++)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1min][c2];
-	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
-	  if (*histp != 0) {
-	    boxp->c2min = c2min = c2;
-	    goto have_c2min;
-	  }
+        histp = & histogram[c0][c1min][c2];
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2min = c2min = c2;
+            goto have_c2min;
+          }
       }
  have_c2min:
   if (c2max > c2min)
     for (c2 = c2max; c2 >= c2min; c2--)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1min][c2];
-	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
-	  if (*histp != 0) {
-	    boxp->c2max = c2max = c2;
-	    goto have_c2max;
-	  }
+        histp = & histogram[c0][c1min][c2];
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2max = c2max = c2;
+            goto have_c2max;
+          }
       }
  have_c2max:
 
@@ -387,16 +384,16 @@
   dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
   dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
   boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2;
-  
+
   /* Now scan remaining volume of box and compute population */
   ccount = 0;
   for (c0 = c0min; c0 <= c0max; c0++)
     for (c1 = c1min; c1 <= c1max; c1++) {
       histp = & histogram[c0][c1][c2min];
       for (c2 = c2min; c2 <= c2max; c2++, histp++)
-	if (*histp != 0) {
-	  ccount++;
-	}
+        if (*histp != 0) {
+          ccount++;
+        }
     }
   boxp->colorcount = ccount;
 }
@@ -404,7 +401,7 @@
 
 LOCAL(int)
 median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
-	    int desired_colors)
+            int desired_colors)
 /* Repeatedly select and split the largest box until we have enough boxes */
 {
   int n,lb;
@@ -420,9 +417,9 @@
     } else {
       b1 = find_biggest_volume(boxlist, numboxes);
     }
-    if (b1 == NULL)		/* no splittable boxes left! */
+    if (b1 == NULL)             /* no splittable boxes left! */
       break;
-    b2 = &boxlist[numboxes];	/* where new box will go */
+    b2 = &boxlist[numboxes];    /* where new box will go */
     /* Copy the color bounds to the new box. */
     b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max;
     b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min;
@@ -494,24 +491,24 @@
   long c0total = 0;
   long c1total = 0;
   long c2total = 0;
-  
+
   c0min = boxp->c0min;  c0max = boxp->c0max;
   c1min = boxp->c1min;  c1max = boxp->c1max;
   c2min = boxp->c2min;  c2max = boxp->c2max;
-  
+
   for (c0 = c0min; c0 <= c0max; c0++)
     for (c1 = c1min; c1 <= c1max; c1++) {
       histp = & histogram[c0][c1][c2min];
       for (c2 = c2min; c2 <= c2max; c2++) {
-	if ((count = *histp++) != 0) {
-	  total += count;
-	  c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
-	  c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
-	  c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
-	}
+        if ((count = *histp++) != 0) {
+          total += count;
+          c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
+          c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
+          c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
+        }
       }
     }
-  
+
   cinfo->colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total);
   cinfo->colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total);
   cinfo->colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total);
@@ -528,7 +525,7 @@
 
   /* Allocate workspace for box list */
   boxlist = (boxptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * SIZEOF(box));
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * sizeof(box));
   /* Initialize one box containing whole space */
   numboxes = 1;
   boxlist[0].c0min = 0;
@@ -575,7 +572,7 @@
  * distance from every colormap entry to every histogram cell.  Unfortunately,
  * it needs a work array to hold the best-distance-so-far for each histogram
  * cell (because the inner loop has to be over cells, not colormap entries).
- * The work array elements have to be INT32s, so the work array would need
+ * The work array elements have to be JLONGs, so the work array would need
  * 256Kb at our recommended precision.  This is not feasible in DOS machines.
  *
  * To get around these problems, we apply Thomas' method to compute the
@@ -627,7 +624,7 @@
 
 LOCAL(int)
 find_nearby_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
-		    JSAMPLE colorlist[])
+                    JSAMPLE colorlist[])
 /* Locate the colormap entries close enough to an update box to be candidates
  * for the nearest entry to some cell(s) in the update box.  The update box
  * is specified by the center coordinates of its first cell.  The number of
@@ -641,8 +638,8 @@
   int maxc0, maxc1, maxc2;
   int centerc0, centerc1, centerc2;
   int i, x, ncolors;
-  INT32 minmaxdist, min_dist, max_dist, tdist;
-  INT32 mindist[MAXNUMCOLORS];	/* min distance to colormap entry i */
+  JLONG minmaxdist, min_dist, max_dist, tdist;
+  JLONG mindist[MAXNUMCOLORS];  /* min distance to colormap entry i */
 
   /* Compute true coordinates of update box's upper corner and center.
    * Actually we compute the coordinates of the center of the upper-corner
@@ -684,11 +681,11 @@
       /* within cell range so no contribution to min_dist */
       min_dist = 0;
       if (x <= centerc0) {
-	tdist = (x - maxc0) * C0_SCALE;
-	max_dist = tdist*tdist;
+        tdist = (x - maxc0) * C0_SCALE;
+        max_dist = tdist*tdist;
       } else {
-	tdist = (x - minc0) * C0_SCALE;
-	max_dist = tdist*tdist;
+        tdist = (x - minc0) * C0_SCALE;
+        max_dist = tdist*tdist;
       }
     }
 
@@ -706,11 +703,11 @@
     } else {
       /* within cell range so no contribution to min_dist */
       if (x <= centerc1) {
-	tdist = (x - maxc1) * C1_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - maxc1) * C1_SCALE;
+        max_dist += tdist*tdist;
       } else {
-	tdist = (x - minc1) * C1_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - minc1) * C1_SCALE;
+        max_dist += tdist*tdist;
       }
     }
 
@@ -728,15 +725,15 @@
     } else {
       /* within cell range so no contribution to min_dist */
       if (x <= centerc2) {
-	tdist = (x - maxc2) * C2_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - maxc2) * C2_SCALE;
+        max_dist += tdist*tdist;
       } else {
-	tdist = (x - minc2) * C2_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - minc2) * C2_SCALE;
+        max_dist += tdist*tdist;
       }
     }
 
-    mindist[i] = min_dist;	/* save away the results */
+    mindist[i] = min_dist;      /* save away the results */
     if (max_dist < minmaxdist)
       minmaxdist = max_dist;
   }
@@ -756,7 +753,7 @@
 
 LOCAL(void)
 find_best_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
-		  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
+                  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
 /* Find the closest colormap entry for each cell in the update box,
  * given the list of candidate colors prepared by find_nearby_colors.
  * Return the indexes of the closest entries in the bestcolor[] array.
@@ -766,31 +763,31 @@
 {
   int ic0, ic1, ic2;
   int i, icolor;
-  register INT32 * bptr;	/* pointer into bestdist[] array */
-  JSAMPLE * cptr;		/* pointer into bestcolor[] array */
-  INT32 dist0, dist1;		/* initial distance values */
-  register INT32 dist2;		/* current distance in inner loop */
-  INT32 xx0, xx1;		/* distance increments */
-  register INT32 xx2;
-  INT32 inc0, inc1, inc2;	/* initial values for increments */
+  register JLONG *bptr;         /* pointer into bestdist[] array */
+  JSAMPLE *cptr;                /* pointer into bestcolor[] array */
+  JLONG dist0, dist1;           /* initial distance values */
+  register JLONG dist2;         /* current distance in inner loop */
+  JLONG xx0, xx1;               /* distance increments */
+  register JLONG xx2;
+  JLONG inc0, inc1, inc2;       /* initial values for increments */
   /* This array holds the distance to the nearest-so-far color for each cell */
-  INT32 bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+  JLONG bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
 
   /* Initialize best-distance for each cell of the update box */
   bptr = bestdist;
   for (i = BOX_C0_ELEMS*BOX_C1_ELEMS*BOX_C2_ELEMS-1; i >= 0; i--)
     *bptr++ = 0x7FFFFFFFL;
-  
+
   /* For each color selected by find_nearby_colors,
    * compute its distance to the center of each cell in the box.
    * If that's less than best-so-far, update best distance and color number.
    */
-  
+
   /* Nominal steps between cell centers ("x" in Thomas article) */
 #define STEP_C0  ((1 << C0_SHIFT) * C0_SCALE)
 #define STEP_C1  ((1 << C1_SHIFT) * C1_SCALE)
 #define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)
-  
+
   for (i = 0; i < numcolors; i++) {
     icolor = GETJSAMPLE(colorlist[i]);
     /* Compute (square of) distance from minc0/c1/c2 to this color */
@@ -812,20 +809,20 @@
       dist1 = dist0;
       xx1 = inc1;
       for (ic1 = BOX_C1_ELEMS-1; ic1 >= 0; ic1--) {
-	dist2 = dist1;
-	xx2 = inc2;
-	for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
-	  if (dist2 < *bptr) {
-	    *bptr = dist2;
-	    *cptr = (JSAMPLE) icolor;
-	  }
-	  dist2 += xx2;
-	  xx2 += 2 * STEP_C2 * STEP_C2;
-	  bptr++;
-	  cptr++;
-	}
-	dist1 += xx1;
-	xx1 += 2 * STEP_C1 * STEP_C1;
+        dist2 = dist1;
+        xx2 = inc2;
+        for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
+          if (dist2 < *bptr) {
+            *bptr = dist2;
+            *cptr = (JSAMPLE) icolor;
+          }
+          dist2 += xx2;
+          xx2 += 2 * STEP_C2 * STEP_C2;
+          bptr++;
+          cptr++;
+        }
+        dist1 += xx1;
+        xx1 += 2 * STEP_C1 * STEP_C1;
       }
       dist0 += xx0;
       xx0 += 2 * STEP_C0 * STEP_C0;
@@ -842,13 +839,13 @@
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   hist3d histogram = cquantize->histogram;
-  int minc0, minc1, minc2;	/* lower left corner of update box */
+  int minc0, minc1, minc2;      /* lower left corner of update box */
   int ic0, ic1, ic2;
-  register JSAMPLE * cptr;	/* pointer into bestcolor[] array */
-  register histptr cachep;	/* pointer into main cache array */
+  register JSAMPLE *cptr;       /* pointer into bestcolor[] array */
+  register histptr cachep;      /* pointer into main cache array */
   /* This array lists the candidate colormap indexes. */
   JSAMPLE colorlist[MAXNUMCOLORS];
-  int numcolors;		/* number of candidate colors */
+  int numcolors;                /* number of candidate colors */
   /* This array holds the actually closest colormap index for each cell. */
   JSAMPLE bestcolor[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
 
@@ -864,7 +861,7 @@
   minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1);
   minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1);
   minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1);
-  
+
   /* Determine which colormap entries are close enough to be candidates
    * for the nearest entry to some cell in the update box.
    */
@@ -872,10 +869,10 @@
 
   /* Determine the actually nearest colors. */
   find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist,
-		   bestcolor);
+                   bestcolor);
 
   /* Save the best color numbers (plus 1) in the main cache array */
-  c0 <<= BOX_C0_LOG;		/* convert ID back to base cell indexes */
+  c0 <<= BOX_C0_LOG;            /* convert ID back to base cell indexes */
   c1 <<= BOX_C1_LOG;
   c2 <<= BOX_C2_LOG;
   cptr = bestcolor;
@@ -883,7 +880,7 @@
     for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
       cachep = & histogram[c0+ic0][c1+ic1][c2];
       for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
-	*cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
+        *cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
       }
     }
   }
@@ -896,7 +893,7 @@
 
 METHODDEF(void)
 pass2_no_dither (j_decompress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
 /* This version performs no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -920,7 +917,7 @@
       /* If we have not seen this color before, find nearest colormap entry */
       /* and update the cache */
       if (*cachep == 0)
-	fill_inverse_cmap(cinfo, c0,c1,c2);
+        fill_inverse_cmap(cinfo, c0,c1,c2);
       /* Now emit the colormap index for this cell */
       *outptr++ = (JSAMPLE) (*cachep - 1);
     }
@@ -930,20 +927,20 @@
 
 METHODDEF(void)
 pass2_fs_dither (j_decompress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
 /* This version performs Floyd-Steinberg dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   hist3d histogram = cquantize->histogram;
-  register LOCFSERROR cur0, cur1, cur2;	/* current error or pixel value */
+  register LOCFSERROR cur0, cur1, cur2; /* current error or pixel value */
   LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */
   LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */
-  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
-  JSAMPROW inptr;		/* => current input pixel */
-  JSAMPROW outptr;		/* => current output pixel */
+  register FSERRPTR errorptr;   /* => fserrors[] at column before current */
+  JSAMPROW inptr;               /* => current input pixel */
+  JSAMPROW outptr;              /* => current output pixel */
   histptr cachep;
-  int dir;			/* +1 or -1 depending on direction */
-  int dir3;			/* 3*dir, for advancing inptr & errorptr */
+  int dir;                      /* +1 or -1 depending on direction */
+  int dir3;                     /* 3*dir, for advancing inptr & errorptr */
   int row;
   JDIMENSION col;
   JDIMENSION width = cinfo->output_width;
@@ -959,7 +956,7 @@
     outptr = output_buf[row];
     if (cquantize->on_odd_row) {
       /* work right to left in this row */
-      inptr += (width-1) * 3;	/* so point to rightmost pixel */
+      inptr += (width-1) * 3;   /* so point to rightmost pixel */
       outptr += width-1;
       dir = -1;
       dir3 = -3;
@@ -1011,53 +1008,44 @@
       /* If we have not seen this color before, find nearest colormap */
       /* entry and update the cache */
       if (*cachep == 0)
-	fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
+        fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
       /* Now emit the colormap index for this cell */
       { register int pixcode = *cachep - 1;
-	*outptr = (JSAMPLE) pixcode;
-	/* Compute representation error for this pixel */
-	cur0 -= GETJSAMPLE(colormap0[pixcode]);
-	cur1 -= GETJSAMPLE(colormap1[pixcode]);
-	cur2 -= GETJSAMPLE(colormap2[pixcode]);
+        *outptr = (JSAMPLE) pixcode;
+        /* Compute representation error for this pixel */
+        cur0 -= GETJSAMPLE(colormap0[pixcode]);
+        cur1 -= GETJSAMPLE(colormap1[pixcode]);
+        cur2 -= GETJSAMPLE(colormap2[pixcode]);
       }
       /* Compute error fractions to be propagated to adjacent pixels.
        * Add these into the running sums, and simultaneously shift the
        * next-line error sums left by 1 column.
        */
-      { register LOCFSERROR bnexterr, delta;
+      { register LOCFSERROR bnexterr;
 
-	bnexterr = cur0;	/* Process component 0 */
-	delta = cur0 * 2;
-	cur0 += delta;		/* form error * 3 */
-	errorptr[0] = (FSERROR) (bpreverr0 + cur0);
-	cur0 += delta;		/* form error * 5 */
-	bpreverr0 = belowerr0 + cur0;
-	belowerr0 = bnexterr;
-	cur0 += delta;		/* form error * 7 */
-	bnexterr = cur1;	/* Process component 1 */
-	delta = cur1 * 2;
-	cur1 += delta;		/* form error * 3 */
-	errorptr[1] = (FSERROR) (bpreverr1 + cur1);
-	cur1 += delta;		/* form error * 5 */
-	bpreverr1 = belowerr1 + cur1;
-	belowerr1 = bnexterr;
-	cur1 += delta;		/* form error * 7 */
-	bnexterr = cur2;	/* Process component 2 */
-	delta = cur2 * 2;
-	cur2 += delta;		/* form error * 3 */
-	errorptr[2] = (FSERROR) (bpreverr2 + cur2);
-	cur2 += delta;		/* form error * 5 */
-	bpreverr2 = belowerr2 + cur2;
-	belowerr2 = bnexterr;
-	cur2 += delta;		/* form error * 7 */
+        bnexterr = cur0;        /* Process component 0 */
+        errorptr[0] = (FSERROR) (bpreverr0 + cur0 * 3);
+        bpreverr0 = belowerr0 + cur0 * 5;
+        belowerr0 = bnexterr;
+        cur0 *= 7;
+        bnexterr = cur1;        /* Process component 1 */
+        errorptr[1] = (FSERROR) (bpreverr1 + cur1 * 3);
+        bpreverr1 = belowerr1 + cur1 * 5;
+        belowerr1 = bnexterr;
+        cur1 *= 7;
+        bnexterr = cur2;        /* Process component 2 */
+        errorptr[2] = (FSERROR) (bpreverr2 + cur2 * 3);
+        bpreverr2 = belowerr2 + cur2 * 5;
+        belowerr2 = bnexterr;
+        cur2 *= 7;
       }
       /* At this point curN contains the 7/16 error value to be propagated
        * to the next pixel on the current line, and all the errors for the
        * next line have been shifted over.  We are therefore ready to move on.
        */
-      inptr += dir3;		/* Advance pixel pointers to next column */
+      inptr += dir3;            /* Advance pixel pointers to next column */
       outptr += dir;
-      errorptr += dir3;		/* advance errorptr to current column */
+      errorptr += dir3;         /* advance errorptr to current column */
     }
     /* Post-loop cleanup: we must unload the final error values into the
      * final fserrors[] entry.  Note we need not unload belowerrN because
@@ -1092,12 +1080,12 @@
 /* Allocate and fill in the error_limiter table */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  int * table;
+  int *table;
   int in, out;
 
   table = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * SIZEOF(int));
-  table += MAXJSAMPLE;		/* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * sizeof(int));
+  table += MAXJSAMPLE;          /* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
   cquantize->error_limiter = table;
 
 #define STEPSIZE ((MAXJSAMPLE+1)/16)
@@ -1180,16 +1168,16 @@
 
     if (cinfo->dither_mode == JDITHER_FS) {
       size_t arraysize = (size_t) ((cinfo->output_width + 2) *
-				   (3 * SIZEOF(FSERROR)));
+                                   (3 * sizeof(FSERROR)));
       /* Allocate Floyd-Steinberg workspace if we didn't already. */
       if (cquantize->fserrors == NULL)
-	cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+        cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
       /* Initialize the propagated errors to zero. */
-      jzero_far((void FAR *) cquantize->fserrors, arraysize);
+      jzero_far((void *) cquantize->fserrors, arraysize);
       /* Make the error-limit table if we didn't already. */
       if (cquantize->error_limiter == NULL)
-	init_error_limit(cinfo);
+        init_error_limit(cinfo);
       cquantize->on_odd_row = FALSE;
     }
 
@@ -1197,8 +1185,8 @@
   /* Zero the histogram or inverse color map, if necessary */
   if (cquantize->needs_zeroed) {
     for (i = 0; i < HIST_C0_ELEMS; i++) {
-      jzero_far((void FAR *) histogram[i],
-		HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+      jzero_far((void *) histogram[i],
+                HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell));
     }
     cquantize->needs_zeroed = FALSE;
   }
@@ -1231,11 +1219,11 @@
 
   cquantize = (my_cquantize_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_cquantizer));
+                                sizeof(my_cquantizer));
   cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
   cquantize->pub.start_pass = start_pass_2_quant;
   cquantize->pub.new_color_map = new_color_map_2_quant;
-  cquantize->fserrors = NULL;	/* flag optional arrays not allocated */
+  cquantize->fserrors = NULL;   /* flag optional arrays not allocated */
   cquantize->error_limiter = NULL;
 
   /* Make sure jdmaster didn't give me a case I can't handle */
@@ -1244,17 +1232,17 @@
 
   /* Allocate the histogram/inverse colormap storage */
   cquantize->histogram = (hist3d) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * SIZEOF(hist2d));
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * sizeof(hist2d));
   for (i = 0; i < HIST_C0_ELEMS; i++) {
     cquantize->histogram[i] = (hist2d) (*cinfo->mem->alloc_large)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
-       HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+       HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell));
   }
   cquantize->needs_zeroed = TRUE; /* histogram is garbage now */
 
   /* Allocate storage for the completed colormap, if required.
-   * We do this now since it is FAR storage and may affect
-   * the memory manager's space calculations.
+   * We do this now since it may affect the memory manager's space
+   * calculations.
    */
   if (cinfo->enable_2pass_quant) {
     /* Make sure color count is acceptable */
@@ -1277,14 +1265,15 @@
     cinfo->dither_mode = JDITHER_FS;
 
   /* Allocate Floyd-Steinberg workspace if necessary.
-   * This isn't really needed until pass 2, but again it is FAR storage.
-   * Although we will cope with a later change in dither_mode,
-   * we do not promise to honor max_memory_to_use if dither_mode changes.
+   * This isn't really needed until pass 2, but again it may affect the memory
+   * manager's space calculations.  Although we will cope with a later change
+   * in dither_mode, we do not promise to honor max_memory_to_use if
+   * dither_mode changes.
    */
   if (cinfo->dither_mode == JDITHER_FS) {
     cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
-       (size_t) ((cinfo->output_width + 2) * (3 * SIZEOF(FSERROR))));
+       (size_t) ((cinfo->output_width + 2) * (3 * sizeof(FSERROR))));
     /* Might as well create the error-limiting table too. */
     init_error_limit(cinfo);
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimddct.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimddct.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimddct.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimddct.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,101 +2,73 @@
  * jsimddct.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * 
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jsimd_can_convsamp                jSCanConv
-#define jsimd_can_convsamp_float          jSCanConvF
-#define jsimd_convsamp                    jSConv
-#define jsimd_convsamp_float              jSConvF
-#define jsimd_can_fdct_islow              jSCanFDCTIS
-#define jsimd_can_fdct_ifast              jSCanFDCTIF
-#define jsimd_can_fdct_float              jSCanFDCTFl
-#define jsimd_fdct_islow                  jSFDCTIS
-#define jsimd_fdct_ifast                  jSFDCTIF
-#define jsimd_fdct_float                  jSFDCTFl
-#define jsimd_can_quantize                jSCanQuant
-#define jsimd_can_quantize_float          jSCanQuantF
-#define jsimd_quantize                    jSQuant
-#define jsimd_quantize_float              jSQuantF
-#define jsimd_can_idct_2x2                jSCanIDCT22
-#define jsimd_can_idct_4x4                jSCanIDCT44
-#define jsimd_idct_2x2                    jSIDCT22
-#define jsimd_idct_4x4                    jSIDCT44
-#define jsimd_can_idct_islow              jSCanIDCTIS
-#define jsimd_can_idct_ifast              jSCanIDCTIF
-#define jsimd_can_idct_float              jSCanIDCTFl
-#define jsimd_idct_islow                  jSIDCTIS
-#define jsimd_idct_ifast                  jSIDCTIF
-#define jsimd_idct_float                  jSIDCTFl
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-EXTERN(int) jsimd_can_convsamp JPP((void));
-EXTERN(int) jsimd_can_convsamp_float JPP((void));
-
-EXTERN(void) jsimd_convsamp JPP((JSAMPARRAY sample_data,
-                                 JDIMENSION start_col,
-                                 DCTELEM * workspace));
-EXTERN(void) jsimd_convsamp_float JPP((JSAMPARRAY sample_data,
-                                       JDIMENSION start_col,
-                                       FAST_FLOAT * workspace));
-
-EXTERN(int) jsimd_can_fdct_islow JPP((void));
-EXTERN(int) jsimd_can_fdct_ifast JPP((void));
-EXTERN(int) jsimd_can_fdct_float JPP((void));
-
-EXTERN(void) jsimd_fdct_islow JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_ifast JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_float JPP((FAST_FLOAT * data));
-
-EXTERN(int) jsimd_can_quantize JPP((void));
-EXTERN(int) jsimd_can_quantize_float JPP((void));
-
-EXTERN(void) jsimd_quantize JPP((JCOEFPTR coef_block,
-                                 DCTELEM * divisors,
-                                 DCTELEM * workspace));
-EXTERN(void) jsimd_quantize_float JPP((JCOEFPTR coef_block,
-                                       FAST_FLOAT * divisors,
-                                       FAST_FLOAT * workspace));
-
-EXTERN(int) jsimd_can_idct_2x2 JPP((void));
-EXTERN(int) jsimd_can_idct_4x4 JPP((void));
-
-EXTERN(void) jsimd_idct_2x2 JPP((j_decompress_ptr cinfo,
-                                 jpeg_component_info * compptr,
-                                 JCOEFPTR coef_block,
-                                 JSAMPARRAY output_buf,
-                                 JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4 JPP((j_decompress_ptr cinfo,
-                                 jpeg_component_info * compptr,
-                                 JCOEFPTR coef_block,
-                                 JSAMPARRAY output_buf,
-                                 JDIMENSION output_col));
-
-EXTERN(int) jsimd_can_idct_islow JPP((void));
-EXTERN(int) jsimd_can_idct_ifast JPP((void));
-EXTERN(int) jsimd_can_idct_float JPP((void));
-
-EXTERN(void) jsimd_idct_islow JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
-EXTERN(void) jsimd_idct_float JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
+EXTERN(int) jsimd_can_convsamp (void);
+EXTERN(int) jsimd_can_convsamp_float (void);
 
+EXTERN(void) jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                             DCTELEM *workspace);
+EXTERN(void) jsimd_convsamp_float (JSAMPARRAY sample_data,
+                                   JDIMENSION start_col,
+                                   FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_fdct_islow (void);
+EXTERN(int) jsimd_can_fdct_ifast (void);
+EXTERN(int) jsimd_can_fdct_float (void);
+
+EXTERN(void) jsimd_fdct_islow (DCTELEM *data);
+EXTERN(void) jsimd_fdct_ifast (DCTELEM *data);
+EXTERN(void) jsimd_fdct_float (FAST_FLOAT *data);
+
+EXTERN(int) jsimd_can_quantize (void);
+EXTERN(int) jsimd_can_quantize_float (void);
+
+EXTERN(void) jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                             DCTELEM *workspace);
+EXTERN(void) jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                                   FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_idct_2x2 (void);
+EXTERN(int) jsimd_can_idct_4x4 (void);
+EXTERN(int) jsimd_can_idct_6x6 (void);
+EXTERN(int) jsimd_can_idct_12x12 (void);
+
+EXTERN(void) jsimd_idct_2x2 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_6x6 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12 (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+
+EXTERN(int) jsimd_can_idct_islow (void);
+EXTERN(int) jsimd_can_idct_ifast (void);
+EXTERN(int) jsimd_can_idct_float (void);
+
+EXTERN(void) jsimd_idct_islow (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+EXTERN(void) jsimd_idct_ifast (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+EXTERN(void) jsimd_idct_float (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimd.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimd.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimd.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimd.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,97 +2,92 @@
  * jsimd.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2011 D. R. Commander
- * 
+ * Copyright 2011, 2014 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
+#include "jchuff.h"             /* Declarations shared with jcphuff.c */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jsimd_can_rgb_ycc                 jSCanRgbYcc
-#define jsimd_can_rgb_gray                jSCanRgbGry
-#define jsimd_can_ycc_rgb                 jSCanYccRgb
-#define jsimd_rgb_ycc_convert             jSRgbYccConv
-#define jsimd_rgb_gray_convert            jSRgbGryConv
-#define jsimd_ycc_rgb_convert             jSYccRgbConv
-#define jsimd_can_h2v2_downsample         jSCanH2V2Down
-#define jsimd_can_h2v1_downsample         jSCanH2V1Down
-#define jsimd_h2v2_downsample             jSH2V2Down
-#define jsimd_h2v1_downsample             jSH2V1Down
-#define jsimd_can_h2v2_upsample           jSCanH2V2Up
-#define jsimd_can_h2v1_upsample           jSCanH2V1Up
-#define jsimd_h2v2_upsample               jSH2V2Up
-#define jsimd_h2v1_upsample               jSH2V1Up
-#define jsimd_can_h2v2_fancy_upsample     jSCanH2V2FUp
-#define jsimd_can_h2v1_fancy_upsample     jSCanH2V1FUp
-#define jsimd_h2v2_fancy_upsample         jSH2V2FUp
-#define jsimd_h2v1_fancy_upsample         jSH2V1FUp
-#define jsimd_can_h2v2_merged_upsample    jSCanH2V2MUp
-#define jsimd_can_h2v1_merged_upsample    jSCanH2V1MUp
-#define jsimd_h2v2_merged_upsample        jSH2V2MUp
-#define jsimd_h2v1_merged_upsample        jSH2V1MUp
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-EXTERN(int) jsimd_can_rgb_ycc JPP((void));
-EXTERN(int) jsimd_can_rgb_gray JPP((void));
-EXTERN(int) jsimd_can_ycc_rgb JPP((void));
+EXTERN(int) jsimd_can_rgb_ycc (void);
+EXTERN(int) jsimd_can_rgb_gray (void);
+EXTERN(int) jsimd_can_ycc_rgb (void);
+EXTERN(int) jsimd_can_ycc_rgb565 (void);
+EXTERN(int) jsimd_c_can_null_convert (void);
 
 EXTERN(void) jsimd_rgb_ycc_convert
-        JPP((j_compress_ptr cinfo,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_rgb_gray_convert
-        JPP((j_compress_ptr cinfo,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_ycc_rgb_convert
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_c_null_convert
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
-EXTERN(int) jsimd_can_h2v2_downsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_downsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_downsample (void);
+EXTERN(int) jsimd_can_h2v1_downsample (void);
 
 EXTERN(void) jsimd_h2v2_downsample
-        JPP((j_compress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(int) jsimd_can_h2v2_smooth_downsample (void);
+
+EXTERN(void) jsimd_h2v2_smooth_downsample
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
 EXTERN(void) jsimd_h2v1_downsample
-        JPP((j_compress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+        JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-EXTERN(int) jsimd_can_h2v2_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_upsample (void);
+EXTERN(int) jsimd_can_h2v1_upsample (void);
+EXTERN(int) jsimd_can_int_upsample (void);
 
 EXTERN(void) jsimd_h2v2_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 EXTERN(void) jsimd_h2v1_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_int_upsample
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(int) jsimd_can_h2v2_fancy_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_fancy_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_fancy_upsample (void);
+EXTERN(int) jsimd_can_h2v1_fancy_upsample (void);
 
 EXTERN(void) jsimd_h2v2_fancy_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 EXTERN(void) jsimd_h2v1_fancy_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(int) jsimd_can_h2v2_merged_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_merged_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_merged_upsample (void);
+EXTERN(int) jsimd_can_h2v1_merged_upsample (void);
 
 EXTERN(void) jsimd_h2v2_merged_upsample
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-             JSAMPARRAY output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_merged_upsample
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-             JSAMPARRAY output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(int) jsimd_can_huff_encode_one_block (void);
 
+EXTERN(JOCTET*) jsimd_huff_encode_one_block
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimd_none.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimd_none.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jsimd_none.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jsimd_none.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,404 @@
+/*
+ * jsimd_none.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains stubs for when there is no SIMD support available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jdct.h"
+#include "jsimddct.h"
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert (j_compress_ptr cinfo,
+                      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                      JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jstdhuff.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jstdhuff.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jstdhuff.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jstdhuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,135 @@
+/*
+ * jstdhuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to set the default Huffman tables, if they are
+ * not already set.
+ */
+
+/*
+ * Huffman table setup routines
+ */
+
+LOCAL(void)
+add_huff_table (j_common_ptr cinfo,
+                JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
+/* Define a Huffman table */
+{
+  int nsymbols, len;
+
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table(cinfo);
+  else
+    return;
+
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val[] array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bits[len];
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * sizeof(UINT8));
+  MEMZERO(&((*htblptr)->huffval[nsymbols]), (256 - nsymbols) * sizeof(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables (j_common_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  JHUFF_TBL **dc_huff_tbl_ptrs, **ac_huff_tbl_ptrs;
+
+  static const UINT8 bits_dc_luminance[17] =
+    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_luminance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+  static const UINT8 bits_dc_chrominance[17] =
+    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_chrominance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+  static const UINT8 bits_ac_luminance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+  static const UINT8 val_ac_luminance[] =
+    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+
+  static const UINT8 bits_ac_chrominance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+  static const UINT8 val_ac_chrominance[] =
+    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+
+  if (cinfo->is_decompressor) {
+    dc_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  } else {
+    dc_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  }
+
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs[0], bits_dc_luminance,
+                 val_dc_luminance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs[0], bits_ac_luminance,
+                 val_ac_luminance);
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs[1], bits_dc_chrominance,
+                 val_dc_chrominance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs[1], bits_ac_chrominance,
+                 val_ac_chrominance);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jutils.c glmark2-2021.02/android/jni/src/libjpeg-turbo/jutils.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jutils.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jutils.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jutils.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code
+ * relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains tables and miscellaneous utility routines needed
  * for both compression and decompression.
@@ -21,7 +24,7 @@
  * of a DCT block read in natural order (left to right, top to bottom).
  */
 
-#if 0				/* This table is not actually needed in v6a */
+#if 0                           /* This table is not actually needed in v6a */
 
 const int jpeg_zigzag_order[DCTSIZE2] = {
    0,  1,  5,  6, 14, 15, 27, 28,
@@ -87,30 +90,10 @@
 }
 
 
-/* On normal machines we can apply MEMCOPY() and MEMZERO() to sample arrays
- * and coefficient-block arrays.  This won't work on 80x86 because the arrays
- * are FAR and we're assuming a small-pointer memory model.  However, some
- * DOS compilers provide far-pointer versions of memcpy() and memset() even
- * in the small-model libraries.  These will be used if USE_FMEM is defined.
- * Otherwise, the routines below do it the hard way.  (The performance cost
- * is not all that great, because these routines aren't very heavily used.)
- */
-
-#ifndef NEED_FAR_POINTERS	/* normal case, same as regular macros */
-#define FMEMCOPY(dest,src,size)	MEMCOPY(dest,src,size)
-#define FMEMZERO(target,size)	MEMZERO(target,size)
-#else				/* 80x86 case, define if we can */
-#ifdef USE_FMEM
-#define FMEMCOPY(dest,src,size)	_fmemcpy((void FAR *)(dest), (const void FAR *)(src), (size_t)(size))
-#define FMEMZERO(target,size)	_fmemset((void FAR *)(target), 0, (size_t)(size))
-#endif
-#endif
-
-
 GLOBAL(void)
 jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
-		   JSAMPARRAY output_array, int dest_row,
-		   int num_rows, JDIMENSION num_cols)
+                   JSAMPARRAY output_array, int dest_row,
+                   int num_rows, JDIMENSION num_cols)
 /* Copy some rows of samples from one place to another.
  * num_rows rows are copied from input_array[source_row++]
  * to output_array[dest_row++]; these areas may overlap for duplication.
@@ -118,11 +101,7 @@
  */
 {
   register JSAMPROW inptr, outptr;
-#ifdef FMEMCOPY
-  register size_t count = (size_t) (num_cols * SIZEOF(JSAMPLE));
-#else
-  register JDIMENSION count;
-#endif
+  register size_t count = (size_t) (num_cols * sizeof(JSAMPLE));
   register int row;
 
   input_array += source_row;
@@ -131,49 +110,24 @@
   for (row = num_rows; row > 0; row--) {
     inptr = *input_array++;
     outptr = *output_array++;
-#ifdef FMEMCOPY
-    FMEMCOPY(outptr, inptr, count);
-#else
-    for (count = num_cols; count > 0; count--)
-      *outptr++ = *inptr++;	/* needn't bother with GETJSAMPLE() here */
-#endif
+    MEMCOPY(outptr, inptr, count);
   }
 }
 
 
 GLOBAL(void)
 jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
-		 JDIMENSION num_blocks)
+                 JDIMENSION num_blocks)
 /* Copy a row of coefficient blocks from one place to another. */
 {
-#ifdef FMEMCOPY
-  FMEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * SIZEOF(JCOEF)));
-#else
-  register JCOEFPTR inptr, outptr;
-  register long count;
-
-  inptr = (JCOEFPTR) input_row;
-  outptr = (JCOEFPTR) output_row;
-  for (count = (long) num_blocks * DCTSIZE2; count > 0; count--) {
-    *outptr++ = *inptr++;
-  }
-#endif
+  MEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * sizeof(JCOEF)));
 }
 
 
 GLOBAL(void)
-jzero_far (void FAR * target, size_t bytestozero)
-/* Zero out a chunk of FAR memory. */
+jzero_far (void *target, size_t bytestozero)
+/* Zero out a chunk of memory. */
 /* This might be sample-array data, block-array data, or alloc_large data. */
 {
-#ifdef FMEMZERO
-  FMEMZERO(target, bytestozero);
-#else
-  register char FAR * ptr = (char FAR *) target;
-  register size_t count;
-
-  for (count = bytestozero; count > 0; count--) {
-    *ptr++ = 0;
-  }
-#endif
+  MEMZERO(target, bytestozero);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jversion.h glmark2-2021.02/android/jni/src/libjpeg-turbo/jversion.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/jversion.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/jversion.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jversion.h
  *
- * Copyright (C) 1991-2010, Thomas G. Lane, Guido Vollbeding.
- * Copyright (C) 2010, 2012, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2012-2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains software version identification.
  */
@@ -12,7 +14,7 @@
 
 #if JPEG_LIB_VERSION >= 80
 
-#define JVERSION	"8b  16-May-2010"
+#define JVERSION        "8d  15-Jan-2012"
 
 #elif JPEG_LIB_VERSION >= 70
 
@@ -20,12 +22,28 @@
 
 #else
 
-#define JVERSION	"6b  27-Mar-1998"
+#define JVERSION        "6b  27-Mar-1998"
 
 #endif
 
-#define JCOPYRIGHT	"Copyright (C) 1991-2010 Thomas G. Lane, Guido Vollbeding\n" \
-			"Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
-			"Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \
-			"Copyright (C) 2009-2012 D. R. Commander\n" \
-			"Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)"
+/*
+ * NOTE: It is our convention to place the authors in the following order:
+ * - libjpeg-turbo authors (2009-) in descending order of the date of their
+ *   most recent contribution to the project, then in ascending order of the
+ *   date of their first contribution to the project
+ * - Upstream authors in descending order of the date of the first inclusion of
+ *   their code
+ */
+
+#define JCOPYRIGHT      "Copyright (C) 2009-2016 D. R. Commander\n" \
+                        "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
+                        "Copyright (C) 2015-2016 Matthieu Darbois\n" \
+                        "Copyright (C) 2015 Google, Inc.\n" \
+                        "Copyright (C) 2013-2014 MIPS Technologies, Inc.\n" \
+                        "Copyright (C) 2013 Linaro Limited\n" \
+                        "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
+                        "Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \
+                        "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
+                        "Copyright (C) 1991-2016 Thomas G. Lane, Guido Vollbeding" \
+
+#define JCOPYRIGHT_SHORT "Copyright (C) 1991-2016 The libjpeg-turbo Project and many others"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/libjpeg.map.in glmark2-2021.02/android/jni/src/libjpeg-turbo/libjpeg.map.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/libjpeg.map.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/libjpeg.map.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+LIBJPEGTURBO_@JPEG_LIB_VERSION_DECIMAL@ {
+  @MEM_SRCDST_FUNCTIONS@
+  local:
+    jsimd_*;
+    jconst_*;
+};
+
+LIBJPEG_@JPEG_LIB_VERSION_DECIMAL@ {
+  global:
+    *;
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/libjpeg.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/libjpeg.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/libjpeg.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/libjpeg.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3105 @@
+USING THE IJG JPEG LIBRARY
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1994-2013, Thomas G. Lane, Guido Vollbeding.
+libjpeg-turbo Modifications:
+Copyright (C) 2010, 2014-2016, D. R. Commander.
+Copyright (C) 2015, Google, Inc.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file describes how to use the IJG JPEG library within an application
+program.  Read it if you want to write a program that uses the library.
+
+The file example.c provides heavily commented skeleton code for calling the
+JPEG library.  Also see jpeglib.h (the include file to be used by application
+programs) for full details about data structures and function parameter lists.
+The library source code, of course, is the ultimate reference.
+
+Note that there have been *major* changes from the application interface
+presented by IJG version 4 and earlier versions.  The old design had several
+inherent limitations, and it had accumulated a lot of cruft as we added
+features while trying to minimize application-interface changes.  We have
+sacrificed backward compatibility in the version 5 rewrite, but we think the
+improvements justify this.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Overview:
+        Functions provided by the library
+        Outline of typical usage
+Basic library usage:
+        Data formats
+        Compression details
+        Decompression details
+        Mechanics of usage: include files, linking, etc
+Advanced features:
+        Compression parameter selection
+        Decompression parameter selection
+        Special color spaces
+        Error handling
+        Compressed data handling (source and destination managers)
+        I/O suspension
+        Progressive JPEG support
+        Buffered-image mode
+        Abbreviated datastreams and multiple images
+        Special markers
+        Raw (downsampled) image data
+        Really raw data: DCT coefficients
+        Progress monitoring
+        Memory management
+        Memory usage
+        Library compile-time options
+        Portability considerations
+
+You should read at least the overview and basic usage sections before trying
+to program with the library.  The sections on advanced features can be read
+if and when you need them.
+
+
+OVERVIEW
+========
+
+Functions provided by the library
+---------------------------------
+
+The IJG JPEG library provides C code to read and write JPEG-compressed image
+files.  The surrounding application program receives or supplies image data a
+scanline at a time, using a straightforward uncompressed image format.  All
+details of color conversion and other preprocessing/postprocessing can be
+handled by the library.
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  The application indirectly selects use of this code
+by specifying the format in which it wishes to supply or receive image data.
+For example, if colormapped output is requested, then the decompression
+library automatically invokes color quantization.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+A word about functions *not* provided by the library.  We handle a subset of
+the ISO JPEG standard; most baseline, extended-sequential, and progressive
+JPEG processes are supported.  (Our subset includes all features now in common
+use.)  Unsupported ISO options include:
+        * Hierarchical storage
+        * Lossless JPEG
+        * DNL marker
+        * Nonintegral subsampling ratios
+We support both 8- and 12-bit data precision, but this is a compile-time
+choice rather than a run-time choice; hence it is difficult to use both
+precisions in a single application.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, this library is
+used by the free LIBTIFF library to support JPEG compression in TIFF.)
+
+
+Outline of typical usage
+------------------------
+
+The rough outline of a JPEG compression operation is:
+
+        Allocate and initialize a JPEG compression object
+        Specify the destination for the compressed data (eg, a file)
+        Set parameters for compression, including image size & colorspace
+        jpeg_start_compress(...);
+        while (scan lines remain to be written)
+                jpeg_write_scanlines(...);
+        jpeg_finish_compress(...);
+        Release the JPEG compression object
+
+A JPEG compression object holds parameters and working state for the JPEG
+library.  We make creation/destruction of the object separate from starting
+or finishing compression of an image; the same object can be re-used for a
+series of image compression operations.  This makes it easy to re-use the
+same parameter settings for a sequence of images.  Re-use of a JPEG object
+also has important implications for processing abbreviated JPEG datastreams,
+as discussed later.
+
+The image data to be compressed is supplied to jpeg_write_scanlines() from
+in-memory buffers.  If the application is doing file-to-file compression,
+reading image data from the source file is the application's responsibility.
+The library emits compressed data by calling a "data destination manager",
+which typically will write the data into a file; but the application can
+provide its own destination manager to do something else.
+
+Similarly, the rough outline of a JPEG decompression operation is:
+
+        Allocate and initialize a JPEG decompression object
+        Specify the source of the compressed data (eg, a file)
+        Call jpeg_read_header() to obtain image info
+        Set parameters for decompression
+        jpeg_start_decompress(...);
+        while (scan lines remain to be read)
+                jpeg_read_scanlines(...);
+        jpeg_finish_decompress(...);
+        Release the JPEG decompression object
+
+This is comparable to the compression outline except that reading the
+datastream header is a separate step.  This is helpful because information
+about the image's size, colorspace, etc is available when the application
+selects decompression parameters.  For example, the application can choose an
+output scaling ratio that will fit the image into the available screen size.
+
+The decompression library obtains compressed data by calling a data source
+manager, which typically will read the data from a file; but other behaviors
+can be obtained with a custom source manager.  Decompressed data is delivered
+into in-memory buffers passed to jpeg_read_scanlines().
+
+It is possible to abort an incomplete compression or decompression operation
+by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
+simply release it by calling jpeg_destroy().
+
+JPEG compression and decompression objects are two separate struct types.
+However, they share some common fields, and certain routines such as
+jpeg_destroy() can work on either type of object.
+
+The JPEG library has no static variables: all state is in the compression
+or decompression object.  Therefore it is possible to process multiple
+compression and decompression operations concurrently, using multiple JPEG
+objects.
+
+Both compression and decompression can be done in an incremental memory-to-
+memory fashion, if suitable source/destination managers are used.  See the
+section on "I/O suspension" for more details.
+
+
+BASIC LIBRARY USAGE
+===================
+
+Data formats
+------------
+
+Before diving into procedural details, it is helpful to understand the
+image data format that the JPEG library expects or returns.
+
+The standard input image format is a rectangular array of pixels, with each
+pixel having the same number of "component" or "sample" values (color
+channels).  You must specify how many components there are and the colorspace
+interpretation of the components.  Most applications will use RGB data
+(three components per pixel) or grayscale data (one component per pixel).
+PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
+A remarkable number of people manage to miss this, only to find that their
+programs don't work with grayscale JPEG files.
+
+There is no provision for colormapped input.  JPEG files are always full-color
+or full grayscale (or sometimes another colorspace such as CMYK).  You can
+feed in a colormapped image by expanding it to full-color format.  However
+JPEG often doesn't work very well with source data that has been colormapped,
+because of dithering noise.  This is discussed in more detail in the JPEG FAQ
+and the other references mentioned in the README.ijg file.
+
+Pixels are stored by scanlines, with each scanline running from left to
+right.  The component values for each pixel are adjacent in the row; for
+example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color.  Each scanline is an
+array of data type JSAMPLE --- which is typically "unsigned char", unless
+you've changed jmorecfg.h.  (You can also change the RGB pixel layout, say
+to B,G,R order, by modifying jmorecfg.h.  But see the restrictions listed in
+that file before doing so.)
+
+A 2-D array of pixels is formed by making a list of pointers to the starts of
+scanlines; so the scanlines need not be physically adjacent in memory.  Even
+if you process just one scanline at a time, you must make a one-element
+pointer array to conform to this structure.  Pointers to JSAMPLE rows are of
+type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
+
+The library accepts or supplies one or more complete scanlines per call.
+It is not possible to process part of a row at a time.  Scanlines are always
+processed top-to-bottom.  You can process an entire image in one call if you
+have it all in memory, but usually it's simplest to process one scanline at
+a time.
+
+For best results, source data values should have the precision specified by
+BITS_IN_JSAMPLE (normally 8 bits).  For instance, if you choose to compress
+data that's only 6 bits/channel, you should left-justify each value in a
+byte before passing it to the compressor.  If you need to compress data
+that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
+(See "Library compile-time options", later.)
+
+
+The data format returned by the decompressor is the same in all details,
+except that colormapped output is supported.  (Again, a JPEG file is never
+colormapped.  But you can ask the decompressor to perform on-the-fly color
+quantization to deliver colormapped output.)  If you request colormapped
+output then the returned data array contains a single JSAMPLE per pixel;
+its value is an index into a color map.  The color map is represented as
+a 2-D JSAMPARRAY in which each row holds the values of one color component,
+that is, colormap[i][j] is the value of the i'th color component for pixel
+value (map index) j.  Note that since the colormap indexes are stored in
+JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
+(ie, at most 256 colors for an 8-bit JPEG library).
+
+
+Compression details
+-------------------
+
+Here we revisit the JPEG compression outline given in the overview.
+
+1. Allocate and initialize a JPEG compression object.
+
+A JPEG compression object is a "struct jpeg_compress_struct".  (It also has
+a bunch of subsidiary structures which are allocated via malloc(), but the
+application doesn't control those directly.)  This struct can be just a local
+variable in the calling routine, if a single routine is going to execute the
+whole JPEG compression sequence.  Otherwise it can be static or allocated
+from malloc().
+
+You will also need a structure representing a JPEG error handler.  The part
+of this that the library cares about is a "struct jpeg_error_mgr".  If you
+are providing your own error handler, you'll typically want to embed the
+jpeg_error_mgr struct in a larger structure; this is discussed later under
+"Error handling".  For now we'll assume you are just using the default error
+handler.  The default error handler will print JPEG error/warning messages
+on stderr, and it will call exit() if a fatal error occurs.
+
+You must initialize the error handler structure, store a pointer to it into
+the JPEG object's "err" field, and then call jpeg_create_compress() to
+initialize the rest of the JPEG object.
+
+Typical code for this step, if you are using the default error handler, is
+
+        struct jpeg_compress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_compress(&cinfo);
+
+jpeg_create_compress allocates a small amount of memory, so it could fail
+if you are out of memory.  In that case it will exit via the error handler;
+that's why the error handler must be initialized first.
+
+
+2. Specify the destination for the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library delivers compressed data to a
+"data destination" module.  The library includes one data destination
+module which knows how to write to a stdio stream.  You can use your own
+destination module if you want to do something else, as discussed later.
+
+If you use the standard destination module, you must open the target stdio
+stream beforehand.  Typical code for this step looks like:
+
+        FILE *outfile;
+        ...
+        if ((outfile = fopen(filename, "wb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_dest(&cinfo, outfile);
+
+where the last line invokes the standard destination module.
+
+WARNING: it is critical that the binary compressed data be delivered to the
+output file unchanged.  On non-Unix systems the stdio library may perform
+newline translation or otherwise corrupt binary data.  To suppress this
+behavior, you may need to use a "b" option to fopen (as shown above), or use
+setmode() or another routine to put the stdio stream in binary mode.  See
+cjpeg.c and djpeg.c for code that has been found to work on many systems.
+
+You can select the data destination after setting other parameters (step 3),
+if that's more convenient.  You may not change the destination between
+calling jpeg_start_compress() and jpeg_finish_compress().
+
+
+3. Set parameters for compression, including image size & colorspace.
+
+You must supply information about the source image by setting the following
+fields in the JPEG object (cinfo structure):
+
+        image_width             Width of image, in pixels
+        image_height            Height of image, in pixels
+        input_components        Number of color channels (samples per pixel)
+        in_color_space          Color space of source image
+
+The image dimensions are, hopefully, obvious.  JPEG supports image dimensions
+of 1 to 64K pixels in either direction.  The input color space is typically
+RGB or grayscale, and input_components is 3 or 1 accordingly.  (See "Special
+color spaces", later, for more info.)  The in_color_space field must be
+assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
+JCS_GRAYSCALE.
+
+JPEG has a large number of compression parameters that determine how the
+image is encoded.  Most applications don't need or want to know about all
+these parameters.  You can set all the parameters to reasonable defaults by
+calling jpeg_set_defaults(); then, if there are particular values you want
+to change, you can do so after that.  The "Compression parameter selection"
+section tells about all the parameters.
+
+You must set in_color_space correctly before calling jpeg_set_defaults(),
+because the defaults depend on the source image colorspace.  However the
+other three source image parameters need not be valid until you call
+jpeg_start_compress().  There's no harm in calling jpeg_set_defaults() more
+than once, if that happens to be convenient.
+
+Typical code for a 24-bit RGB source image is
+
+        cinfo.image_width = Width;      /* image width and height, in pixels */
+        cinfo.image_height = Height;
+        cinfo.input_components = 3;     /* # of color components per pixel */
+        cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
+
+        jpeg_set_defaults(&cinfo);
+        /* Make optional parameter settings here */
+
+
+4. jpeg_start_compress(...);
+
+After you have established the data destination and set all the necessary
+source image info and other parameters, call jpeg_start_compress() to begin
+a compression cycle.  This will initialize internal state, allocate working
+storage, and emit the first few bytes of the JPEG datastream header.
+
+Typical code:
+
+        jpeg_start_compress(&cinfo, TRUE);
+
+The "TRUE" parameter ensures that a complete JPEG interchange datastream
+will be written.  This is appropriate in most cases.  If you think you might
+want to use an abbreviated datastream, read the section on abbreviated
+datastreams, below.
+
+Once you have called jpeg_start_compress(), you may not alter any JPEG
+parameters or other fields of the JPEG object until you have completed
+the compression cycle.
+
+
+5. while (scan lines remain to be written)
+        jpeg_write_scanlines(...);
+
+Now write all the required image data by calling jpeg_write_scanlines()
+one or more times.  You can pass one or more scanlines in each call, up
+to the total image height.  In most applications it is convenient to pass
+just one or a few scanlines at a time.  The expected format for the passed
+data is discussed under "Data formats", above.
+
+Image data should be written in top-to-bottom scanline order.  The JPEG spec
+contains some weasel wording about how top and bottom are application-defined
+terms (a curious interpretation of the English language...) but if you want
+your files to be compatible with everyone else's, you WILL use top-to-bottom
+order.  If the source data must be read in bottom-to-top order, you can use
+the JPEG library's virtual array mechanism to invert the data efficiently.
+Examples of this can be found in the sample application cjpeg.
+
+The library maintains a count of the number of scanlines written so far
+in the next_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.next_scanline < cinfo.image_height)".
+
+Code for this step depends heavily on the way that you store the source data.
+example.c shows the following code for the case of a full-size 2-D source
+array containing 3-byte RGB pixels:
+
+        JSAMPROW row_pointer[1];        /* pointer to a single row */
+        int row_stride;                 /* physical row width in buffer */
+
+        row_stride = image_width * 3;   /* JSAMPLEs per row in image_buffer */
+
+        while (cinfo.next_scanline < cinfo.image_height) {
+            row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+            jpeg_write_scanlines(&cinfo, row_pointer, 1);
+        }
+
+jpeg_write_scanlines() returns the number of scanlines actually written.
+This will normally be equal to the number passed in, so you can usually
+ignore the return value.  It is different in just two cases:
+  * If you try to write more scanlines than the declared image height,
+    the additional scanlines are ignored.
+  * If you use a suspending data destination manager, output buffer overrun
+    will cause the compressor to return before accepting all the passed lines.
+    This feature is discussed under "I/O suspension", below.  The normal
+    stdio destination manager will NOT cause this to happen.
+In any case, the return value is the same as the change in the value of
+next_scanline.
+
+
+6. jpeg_finish_compress(...);
+
+After all the image data has been written, call jpeg_finish_compress() to
+complete the compression cycle.  This step is ESSENTIAL to ensure that the
+last bufferload of data is written to the data destination.
+jpeg_finish_compress() also releases working memory associated with the JPEG
+object.
+
+Typical code:
+
+        jpeg_finish_compress(&cinfo);
+
+If using the stdio destination manager, don't forget to close the output
+stdio stream (if necessary) afterwards.
+
+If you have requested a multi-pass operating mode, such as Huffman code
+optimization, jpeg_finish_compress() will perform the additional passes using
+data buffered by the first pass.  In this case jpeg_finish_compress() may take
+quite a while to complete.  With the default compression parameters, this will
+not happen.
+
+It is an error to call jpeg_finish_compress() before writing the necessary
+total number of scanlines.  If you wish to abort compression, call
+jpeg_abort() as discussed below.
+
+After completing a compression cycle, you may dispose of the JPEG object
+as discussed next, or you may use it to compress another image.  In that case
+return to step 2, 3, or 4 as appropriate.  If you do not change the
+destination manager, the new datastream will be written to the same target.
+If you do not change any JPEG parameters, the new datastream will be written
+with the same parameters as before.  Note that you can change the input image
+dimensions freely between cycles, but if you change the input colorspace, you
+should call jpeg_set_defaults() to adjust for the new colorspace; and then
+you'll need to repeat all of step 3.
+
+
+7. Release the JPEG compression object.
+
+When you are done with a JPEG compression object, destroy it by calling
+jpeg_destroy_compress().  This will free all subsidiary memory (regardless of
+the previous state of the object).  Or you can call jpeg_destroy(), which
+works for either compression or decompression objects --- this may be more
+convenient if you are sharing code between compression and decompression
+cases.  (Actually, these routines are equivalent except for the declared type
+of the passed pointer.  To avoid gripes from ANSI C compilers, jpeg_destroy()
+should be passed a j_common_ptr.)
+
+If you allocated the jpeg_compress_struct structure from malloc(), freeing
+it is your responsibility --- jpeg_destroy() won't.  Ditto for the error
+handler structure.
+
+Typical code:
+
+        jpeg_destroy_compress(&cinfo);
+
+
+8. Aborting.
+
+If you decide to abort a compression cycle before finishing, you can clean up
+in either of two ways:
+
+* If you don't need the JPEG object any more, just call
+  jpeg_destroy_compress() or jpeg_destroy() to release memory.  This is
+  legitimate at any point after calling jpeg_create_compress() --- in fact,
+  it's safe even if jpeg_create_compress() fails.
+
+* If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
+  jpeg_abort() which works on both compression and decompression objects.
+  This will return the object to an idle state, releasing any working memory.
+  jpeg_abort() is allowed at any time after successful object creation.
+
+Note that cleaning up the data destination, if required, is your
+responsibility; neither of these routines will call term_destination().
+(See "Compressed data handling", below, for more about that.)
+
+jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
+object that has reported an error by calling error_exit (see "Error handling"
+for more info).  The internal state of such an object is likely to be out of
+whack.  Either of these two routines will return the object to a known state.
+
+
+Decompression details
+---------------------
+
+Here we revisit the JPEG decompression outline given in the overview.
+
+1. Allocate and initialize a JPEG decompression object.
+
+This is just like initialization for compression, as discussed above,
+except that the object is a "struct jpeg_decompress_struct" and you
+call jpeg_create_decompress().  Error handling is exactly the same.
+
+Typical code:
+
+        struct jpeg_decompress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_decompress(&cinfo);
+
+(Both here and in the IJG code, we usually use variable name "cinfo" for
+both compression and decompression objects.)
+
+
+2. Specify the source of the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library reads compressed data from a "data
+source" module.  The library includes one data source module which knows how
+to read from a stdio stream.  You can use your own source module if you want
+to do something else, as discussed later.
+
+If you use the standard source module, you must open the source stdio stream
+beforehand.  Typical code for this step looks like:
+
+        FILE *infile;
+        ...
+        if ((infile = fopen(filename, "rb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_src(&cinfo, infile);
+
+where the last line invokes the standard source module.
+
+WARNING: it is critical that the binary compressed data be read unchanged.
+On non-Unix systems the stdio library may perform newline translation or
+otherwise corrupt binary data.  To suppress this behavior, you may need to use
+a "b" option to fopen (as shown above), or use setmode() or another routine to
+put the stdio stream in binary mode.  See cjpeg.c and djpeg.c for code that
+has been found to work on many systems.
+
+You may not change the data source between calling jpeg_read_header() and
+jpeg_finish_decompress().  If you wish to read a series of JPEG images from
+a single source file, you should repeat the jpeg_read_header() to
+jpeg_finish_decompress() sequence without reinitializing either the JPEG
+object or the data source module; this prevents buffered input data from
+being discarded.
+
+
+3. Call jpeg_read_header() to obtain image info.
+
+Typical code for this step is just
+
+        jpeg_read_header(&cinfo, TRUE);
+
+This will read the source datastream header markers, up to the beginning
+of the compressed data proper.  On return, the image dimensions and other
+info have been stored in the JPEG object.  The application may wish to
+consult this information before selecting decompression parameters.
+
+More complex code is necessary if
+  * A suspending data source is used --- in that case jpeg_read_header()
+    may return before it has read all the header data.  See "I/O suspension",
+    below.  The normal stdio source manager will NOT cause this to happen.
+  * Abbreviated JPEG files are to be processed --- see the section on
+    abbreviated datastreams.  Standard applications that deal only in
+    interchange JPEG files need not be concerned with this case either.
+
+It is permissible to stop at this point if you just wanted to find out the
+image dimensions and other header info for a JPEG file.  In that case,
+call jpeg_destroy() when you are done with the JPEG object, or call
+jpeg_abort() to return it to an idle state before selecting a new data
+source and reading another header.
+
+
+4. Set parameters for decompression.
+
+jpeg_read_header() sets appropriate default decompression parameters based on
+the properties of the image (in particular, its colorspace).  However, you
+may well want to alter these defaults before beginning the decompression.
+For example, the default is to produce full color output from a color file.
+If you want colormapped output you must ask for it.  Other options allow the
+returned image to be scaled and allow various speed/quality tradeoffs to be
+selected.  "Decompression parameter selection", below, gives details.
+
+If the defaults are appropriate, nothing need be done at this step.
+
+Note that all default values are set by each call to jpeg_read_header().
+If you reuse a decompression object, you cannot expect your parameter
+settings to be preserved across cycles, as you can for compression.
+You must set desired parameter values each time.
+
+
+5. jpeg_start_decompress(...);
+
+Once the parameter values are satisfactory, call jpeg_start_decompress() to
+begin decompression.  This will initialize internal state, allocate working
+memory, and prepare for returning data.
+
+Typical code is just
+
+        jpeg_start_decompress(&cinfo);
+
+If you have requested a multi-pass operating mode, such as 2-pass color
+quantization, jpeg_start_decompress() will do everything needed before data
+output can begin.  In this case jpeg_start_decompress() may take quite a while
+to complete.  With a single-scan (non progressive) JPEG file and default
+decompression parameters, this will not happen; jpeg_start_decompress() will
+return quickly.
+
+After this call, the final output image dimensions, including any requested
+scaling, are available in the JPEG object; so is the selected colormap, if
+colormapped output has been requested.  Useful fields include
+
+        output_width            image width and height, as scaled
+        output_height
+        out_color_components    # of color components in out_color_space
+        output_components       # of color components returned per pixel
+        colormap                the selected colormap, if any
+        actual_number_of_colors         number of entries in colormap
+
+output_components is 1 (a colormap index) when quantizing colors; otherwise it
+equals out_color_components.  It is the number of JSAMPLE values that will be
+emitted per pixel in the output arrays.
+
+Typically you will need to allocate data buffers to hold the incoming image.
+You will need output_width * output_components JSAMPLEs per scanline in your
+output buffer, and a total of output_height scanlines will be returned.
+
+Note: if you are using the JPEG library's internal memory manager to allocate
+data buffers (as djpeg does), then the manager's protocol requires that you
+request large buffers *before* calling jpeg_start_decompress().  This is a
+little tricky since the output_XXX fields are not normally valid then.  You
+can make them valid by calling jpeg_calc_output_dimensions() after setting the
+relevant parameters (scaling, output color space, and quantization flag).
+
+
+6. while (scan lines remain to be read)
+        jpeg_read_scanlines(...);
+
+Now you can read the decompressed image data by calling jpeg_read_scanlines()
+one or more times.  At each call, you pass in the maximum number of scanlines
+to be read (ie, the height of your working buffer); jpeg_read_scanlines()
+will return up to that many lines.  The return value is the number of lines
+actually read.  The format of the returned data is discussed under "Data
+formats", above.  Don't forget that grayscale and color JPEGs will return
+different data formats!
+
+Image data is returned in top-to-bottom scanline order.  If you must write
+out the image in bottom-to-top order, you can use the JPEG library's virtual
+array mechanism to invert the data efficiently.  Examples of this can be
+found in the sample application djpeg.
+
+The library maintains a count of the number of scanlines returned so far
+in the output_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.output_scanline < cinfo.output_height)".  (Note that the test
+should NOT be against image_height, unless you never use scaling.  The
+image_height field is the height of the original unscaled image.)
+The return value always equals the change in the value of output_scanline.
+
+If you don't use a suspending data source, it is safe to assume that
+jpeg_read_scanlines() reads at least one scanline per call, until the
+bottom of the image has been reached.
+
+If you use a buffer larger than one scanline, it is NOT safe to assume that
+jpeg_read_scanlines() fills it.  (The current implementation returns only a
+few scanlines per call, no matter how large a buffer you pass.)  So you must
+always provide a loop that calls jpeg_read_scanlines() repeatedly until the
+whole image has been read.
+
+
+7. jpeg_finish_decompress(...);
+
+After all the image data has been read, call jpeg_finish_decompress() to
+complete the decompression cycle.  This causes working memory associated
+with the JPEG object to be released.
+
+Typical code:
+
+        jpeg_finish_decompress(&cinfo);
+
+If using the stdio source manager, don't forget to close the source stdio
+stream if necessary.
+
+It is an error to call jpeg_finish_decompress() before reading the correct
+total number of scanlines.  If you wish to abort decompression, call
+jpeg_abort() as discussed below.
+
+After completing a decompression cycle, you may dispose of the JPEG object as
+discussed next, or you may use it to decompress another image.  In that case
+return to step 2 or 3 as appropriate.  If you do not change the source
+manager, the next image will be read from the same source.
+
+
+8. Release the JPEG decompression object.
+
+When you are done with a JPEG decompression object, destroy it by calling
+jpeg_destroy_decompress() or jpeg_destroy().  The previous discussion of
+destroying compression objects applies here too.
+
+Typical code:
+
+        jpeg_destroy_decompress(&cinfo);
+
+
+9. Aborting.
+
+You can abort a decompression cycle by calling jpeg_destroy_decompress() or
+jpeg_destroy() if you don't need the JPEG object any more, or
+jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
+The previous discussion of aborting compression cycles applies here too.
+
+
+Partial image decompression
+---------------------------
+
+Partial image decompression is convenient for performance-critical applications
+that wish to view only a portion of a large JPEG image without decompressing
+the whole thing.  It it also useful in memory-constrained environments (such as
+on mobile devices.)  This library provides the following functions to support
+partial image decompression:
+
+1. Skipping rows when decompressing
+
+        jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines);
+
+This function provides application programmers with the ability to skip over
+multiple rows in the JPEG image.
+
+Suspending data sources are not supported by this function.  Calling
+jpeg_skip_scanlines() with a suspending data source will result in undefined
+behavior.
+
+jpeg_skip_scanlines() will not allow skipping past the bottom of the image.  If
+the value of num_lines is large enough to skip past the bottom of the image,
+then the function will skip to the end of the image instead.
+
+If the value of num_lines is valid, then jpeg_skip_scanlines() will always
+skip all of the input rows requested.  There is no need to inspect the return
+value of the function in that case.
+
+Best results will be achieved by calling jpeg_skip_scanlines() for large chunks
+of rows.  The function should be viewed as a way to quickly jump to a
+particular vertical offset in the JPEG image in order to decode a subset of the
+image.  Used in this manner, it will provide significant performance
+improvements.
+
+Calling jpeg_skip_scanlines() for small values of num_lines has several
+potential drawbacks:
+    1) JPEG decompression occurs in blocks, so if jpeg_skip_scanlines() is
+       called from the middle of a decompression block, then it is likely that
+       much of the decompression work has already been done for the first
+       couple of rows that need to be skipped.
+    2) When this function returns, it must leave the decompressor in a state
+       such that it is ready to read the next line.  This may involve
+       decompressing a block that must be partially skipped.
+These issues are especially tricky for cases in which upsampling requires
+context rows.  In the worst case, jpeg_skip_scanlines() will perform similarly
+to jpeg_read_scanlines() (since it will actually call jpeg_read_scanlines().)
+
+2. Decompressing partial scanlines
+
+        jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                            JDIMENSION *width)
+
+This function provides application programmers with the ability to decompress
+only a portion of each row in the JPEG image.  It must be called after
+jpeg_start_decompress() and before any calls to jpeg_read_scanlines() or
+jpeg_skip_scanlines().
+
+If xoffset and width do not form a valid subset of the image row, then this
+function will generate an error.  Note that if the output image is scaled, then
+xoffset and width are relative to the scaled image dimensions.
+
+xoffset and width are passed by reference because xoffset must fall on an iMCU
+boundary.  If it doesn't, then it will be moved left to the nearest iMCU
+boundary, and width will be increased accordingly.  If the calling program does
+not like the adjusted values of xoffset and width, then it can call
+jpeg_crop_scanline() again with new values (for instance, if it wants to move
+xoffset to the nearest iMCU boundary to the right instead of to the left.)
+
+After calling this function, cinfo->output_width will be set to the adjusted
+width.  This value should be used when allocating an output buffer to pass to
+jpeg_read_scanlines().
+
+The output image from a partial-width decompression will be identical to the
+corresponding image region from a full decode, with one exception:  The "fancy"
+(smooth) h2v2 (4:2:0) and h2v1 (4:2:2) upsampling algorithms fill in the
+missing chroma components by averaging the chroma components from neighboring
+pixels, except on the right and left edges of the image (where there are no
+neighboring pixels.)  When performing a partial-width decompression, these
+"fancy" upsampling algorithms may treat the left and right edges of the partial
+image region as if they are the left and right edges of the image, meaning that
+the upsampling algorithm may be simplified.  The result is that the pixels on
+the left or right edge of the partial image may not be exactly identical to the
+corresponding pixels in the original image.
+
+
+Mechanics of usage: include files, linking, etc
+-----------------------------------------------
+
+Applications using the JPEG library should include the header file jpeglib.h
+to obtain declarations of data types and routines.  Before including
+jpeglib.h, include system headers that define at least the typedefs FILE and
+size_t.  On ANSI-conforming systems, including <stdio.h> is sufficient; on
+older Unix systems, you may need <sys/types.h> to define size_t.
+
+If the application needs to refer to individual JPEG library error codes, also
+include jerror.h to define those symbols.
+
+jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h.  If you are
+installing the JPEG header files in a system directory, you will want to
+install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
+
+The most convenient way to include the JPEG code into your executable program
+is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
+machines) and reference it at your link step.  If you use only half of the
+library (only compression or only decompression), only that much code will be
+included from the library, unless your linker is hopelessly brain-damaged.
+The supplied makefiles build libjpeg.a automatically (see install.txt).
+
+While you can build the JPEG library as a shared library if the whim strikes
+you, we don't really recommend it.  The trouble with shared libraries is that
+at some point you'll probably try to substitute a new version of the library
+without recompiling the calling applications.  That generally doesn't work
+because the parameter struct declarations usually change with each new
+version.  In other words, the library's API is *not* guaranteed binary
+compatible across versions; we only try to ensure source-code compatibility.
+(In hindsight, it might have been smarter to hide the parameter structs from
+applications and introduce a ton of access functions instead.  Too late now,
+however.)
+
+It may be worth pointing out that the core JPEG library does not actually
+require the stdio library: only the default source/destination managers and
+error handler need it.  You can use the library in a stdio-less environment
+if you replace those modules and use jmemnobs.c (or another memory manager of
+your own devising).  More info about the minimum system library requirements
+may be found in jinclude.h.
+
+
+ADVANCED FEATURES
+=================
+
+Compression parameter selection
+-------------------------------
+
+This section describes all the optional parameters you can set for JPEG
+compression, as well as the "helper" routines provided to assist in this
+task.  Proper setting of some parameters requires detailed understanding
+of the JPEG standard; if you don't know what a parameter is for, it's best
+not to mess with it!  See REFERENCES in the README.ijg file for pointers to
+more info about JPEG.
+
+It's a good idea to call jpeg_set_defaults() first, even if you plan to set
+all the parameters; that way your code is more likely to work with future JPEG
+libraries that have additional parameters.  For the same reason, we recommend
+you use a helper routine where one is provided, in preference to twiddling
+cinfo fields directly.
+
+The helper routines are:
+
+jpeg_set_defaults (j_compress_ptr cinfo)
+        This routine sets all JPEG parameters to reasonable defaults, using
+        only the input image's color space (field in_color_space, which must
+        already be set in cinfo).  Many applications will only need to use
+        this routine and perhaps jpeg_set_quality().
+
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+        Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
+        and sets other color-space-dependent parameters appropriately.  See
+        "Special color spaces", below, before using this.  A large number of
+        parameters, including all per-component parameters, are set by this
+        routine; if you want to twiddle individual parameters you should call
+        jpeg_set_colorspace() before rather than after.
+
+jpeg_default_colorspace (j_compress_ptr cinfo)
+        Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
+        and calls jpeg_set_colorspace().  This is actually a subroutine of
+        jpeg_set_defaults().  It's broken out in case you want to change
+        just the colorspace-dependent JPEG parameters.
+
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+        Constructs JPEG quantization tables appropriate for the indicated
+        quality setting.  The quality value is expressed on the 0..100 scale
+        recommended by IJG (cjpeg's "-quality" switch uses this routine).
+        Note that the exact mapping from quality values to tables may change
+        in future IJG releases as more is learned about DCT quantization.
+        If the force_baseline parameter is TRUE, then the quantization table
+        entries are constrained to the range 1..255 for full JPEG baseline
+        compatibility.  In the current implementation, this only makes a
+        difference for quality settings below 25, and it effectively prevents
+        very small/low quality files from being generated.  The IJG decoder
+        is capable of reading the non-baseline files generated at low quality
+        settings when force_baseline is FALSE, but other decoders may not be.
+
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+                         boolean force_baseline)
+        Same as jpeg_set_quality() except that the generated tables are the
+        sample tables given in the JPEC spec section K.1, multiplied by the
+        specified scale factor (which is expressed as a percentage; thus
+        scale_factor = 100 reproduces the spec's tables).  Note that larger
+        scale factors give lower quality.  This entry point is useful for
+        conforming to the Adobe PostScript DCT conventions, but we do not
+        recommend linear scaling as a user-visible quality scale otherwise.
+        force_baseline again constrains the computed table entries to 1..255.
+
+int jpeg_quality_scaling (int quality)
+        Converts a value on the IJG-recommended quality scale to a linear
+        scaling percentage.  Note that this routine may change or go away
+        in future releases --- IJG may choose to adopt a scaling method that
+        can't be expressed as a simple scalar multiplier, in which case the
+        premise of this routine collapses.  Caveat user.
+
+jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
+        [libjpeg v7+ API/ABI emulation only]
+        Set default quantization tables with linear q_scale_factor[] values
+        (see below).
+
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+                      const unsigned int *basic_table,
+                      int scale_factor, boolean force_baseline)
+        Allows an arbitrary quantization table to be created.  which_tbl
+        indicates which table slot to fill.  basic_table points to an array
+        of 64 unsigned ints given in normal array order.  These values are
+        multiplied by scale_factor/100 and then clamped to the range 1..65535
+        (or to 1..255 if force_baseline is TRUE).
+        CAUTION: prior to library version 6a, jpeg_add_quant_table expected
+        the basic table to be given in JPEG zigzag order.  If you need to
+        write code that works with either older or newer versions of this
+        routine, you must check the library version number.  Something like
+        "#if JPEG_LIB_VERSION >= 61" is the right test.
+
+jpeg_simple_progression (j_compress_ptr cinfo)
+        Generates a default scan script for writing a progressive-JPEG file.
+        This is the recommended method of creating a progressive file,
+        unless you want to make a custom scan sequence.  You must ensure that
+        the JPEG color space is set correctly before calling this routine.
+
+
+Compression parameters (cinfo fields) include:
+
+boolean arith_code
+	If TRUE, use arithmetic coding.
+	If FALSE, use Huffman coding.
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: slow but accurate integer algorithm
+                JDCT_IFAST: faster, less accurate integer method
+                JDCT_FLOAT: floating-point method
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        In libjpeg-turbo, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW when using the x86/x86-64 SIMD extensions (results may vary
+        with other SIMD implementations, or when using libjpeg-turbo without
+        SIMD extensions.)  For quality levels of 90 and below, there should be
+        little or no perceptible difference between the two algorithms.  For
+        quality levels above 90, however, the difference between JDCT_IFAST and
+        JDCT_ISLOW becomes more pronounced.  With quality=97, for instance,
+        JDCT_IFAST incurs generally about a 1-3 dB loss (in PSNR) relative to
+        JDCT_ISLOW, but this can be larger for some images.  Do not use
+        JDCT_IFAST with quality levels above 97.  The algorithm often
+        degenerates at quality=98 and above and can actually produce a more
+        lossy image than if lower quality levels had been used.  Also, in
+        libjpeg-turbo, JDCT_IFAST is not fully accelerated for quality levels
+        above 97, so it will be slower than JDCT_ISLOW.  JDCT_FLOAT is mainly a
+        legacy feature.  It does not produce significantly more accurate
+        results than the ISLOW method, and it is much slower.  The FLOAT method
+        may also give different results on different machines due to varying
+        roundoff behavior, whereas the integer methods should give the same
+        results on all machines.
+
+J_COLOR_SPACE jpeg_color_space
+int num_components
+        The JPEG color space and corresponding number of components; see
+        "Special color spaces", below, for more info.  We recommend using
+        jpeg_set_color_space() if you want to change these.
+
+boolean optimize_coding
+        TRUE causes the compressor to compute optimal Huffman coding tables
+        for the image.  This requires an extra pass over the data and
+        therefore costs a good deal of space and time.  The default is
+        FALSE, which tells the compressor to use the supplied or default
+        Huffman tables.  In most cases optimal tables save only a few percent
+        of file size compared to the default tables.  Note that when this is
+        TRUE, you need not supply Huffman tables at all, and any you do
+        supply will be overwritten.
+
+unsigned int restart_interval
+int restart_in_rows
+        To emit restart markers in the JPEG file, set one of these nonzero.
+        Set restart_interval to specify the exact interval in MCU blocks.
+        Set restart_in_rows to specify the interval in MCU rows.  (If
+        restart_in_rows is not 0, then restart_interval is set after the
+        image width in MCUs is computed.)  Defaults are zero (no restarts).
+        One restart marker per MCU row is often a good choice.
+        NOTE: the overhead of restart markers is higher in grayscale JPEG
+        files than in color files, and MUCH higher in progressive JPEGs.
+        If you use restarts, you may want to use larger intervals in those
+        cases.
+
+const jpeg_scan_info *scan_info
+int num_scans
+        By default, scan_info is NULL; this causes the compressor to write a
+        single-scan sequential JPEG file.  If not NULL, scan_info points to
+        an array of scan definition records of length num_scans.  The
+        compressor will then write a JPEG file having one scan for each scan
+        definition record.  This is used to generate noninterleaved or
+        progressive JPEG files.  The library checks that the scan array
+        defines a valid JPEG scan sequence.  (jpeg_simple_progression creates
+        a suitable scan definition array for progressive JPEG.)  This is
+        discussed further under "Progressive JPEG support".
+
+int smoothing_factor
+        If non-zero, the input image is smoothed; the value should be 1 for
+        minimal smoothing to 100 for maximum smoothing.  Consult jcsample.c
+        for details of the smoothing algorithm.  The default is zero.
+
+boolean write_JFIF_header
+        If TRUE, a JFIF APP0 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
+        (ie, YCbCr or grayscale) is selected, otherwise FALSE.
+
+UINT8 JFIF_major_version
+UINT8 JFIF_minor_version
+        The version number to be written into the JFIF marker.
+        jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
+        You should set it to 1.02 (major=1, minor=2) if you plan to write
+        any JFIF 1.02 extension markers.
+
+UINT8 density_unit
+UINT16 X_density
+UINT16 Y_density
+        The resolution information to be written into the JFIF marker;
+        not used otherwise.  density_unit may be 0 for unknown,
+        1 for dots/inch, or 2 for dots/cm.  The default values are 0,1,1
+        indicating square pixels of unknown size.
+
+boolean write_Adobe_marker
+        If TRUE, an Adobe APP14 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
+        or YCCK is selected, otherwise FALSE.  It is generally a bad idea
+        to set both write_JFIF_header and write_Adobe_marker.  In fact,
+        you probably shouldn't change the default settings at all --- the
+        default behavior ensures that the JPEG file's color space can be
+        recognized by the decoder.
+
+JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS]
+        Pointers to coefficient quantization tables, one per table slot,
+        or NULL if no table is defined for a slot.  Usually these should
+        be set via one of the above helper routines; jpeg_add_quant_table()
+        is general enough to define any quantization table.  The other
+        routines will set up table slot 0 for luminance quality and table
+        slot 1 for chrominance.
+
+int q_scale_factor[NUM_QUANT_TBLS]
+        [libjpeg v7+ API/ABI emulation only]
+        Linear quantization scaling factors (0-100, default 100)
+        for use with jpeg_default_qtables().
+        See rdswitch.c and cjpeg.c for an example of usage.
+        Note that the q_scale_factor[] values use "linear" scales, so JPEG
+        quality levels chosen by the user must be converted to these scales
+        using jpeg_quality_scaling().  Here is an example that corresponds to
+        cjpeg -quality 90,70:
+
+                jpeg_set_defaults(cinfo);
+
+                /* Set luminance quality 90. */
+                cinfo->q_scale_factor[0] = jpeg_quality_scaling(90);
+                /* Set chrominance quality 70. */
+                cinfo->q_scale_factor[1] = jpeg_quality_scaling(70);
+
+                jpeg_default_qtables(cinfo, force_baseline);
+
+        CAUTION: Setting separate quality levels for chrominance and luminance
+        is mainly only useful if chrominance subsampling is disabled.  2x2
+        chrominance subsampling (AKA "4:2:0") is the default, but you can
+        explicitly disable subsampling as follows:
+
+                cinfo->comp_info[0].v_samp_factor = 1;
+                cinfo->comp_info[0].h_samp_factor = 1;
+
+JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
+JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
+        Pointers to Huffman coding tables, one per table slot, or NULL if
+        no table is defined for a slot.  Slots 0 and 1 are filled with the
+        JPEG sample tables by jpeg_set_defaults().  If you need to allocate
+        more table structures, jpeg_alloc_huff_table() may be used.
+        Note that optimal Huffman tables can be computed for an image
+        by setting optimize_coding, as discussed above; there's seldom
+        any need to mess with providing your own Huffman tables.
+
+
+[libjpeg v7+ API/ABI emulation only]
+The actual dimensions of the JPEG image that will be written to the file are
+given by the following fields.  These are computed from the input image
+dimensions and the compression parameters by jpeg_start_compress().  You can
+also call jpeg_calc_jpeg_dimensions() to obtain the values that will result
+from the current parameter settings.  This can be useful if you are trying
+to pick a scaling ratio that will get close to a desired target size.
+
+JDIMENSION jpeg_width           Actual dimensions of output image.
+JDIMENSION jpeg_height
+
+
+Per-component parameters are stored in the struct cinfo.comp_info[i] for
+component number i.  Note that components here refer to components of the
+JPEG color space, *not* the source image color space.  A suitably large
+comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
+to use that routine, it's up to you to allocate the array.
+
+int component_id
+        The one-byte identifier code to be recorded in the JPEG file for
+        this component.  For the standard color spaces, we recommend you
+        leave the default values alone.
+
+int h_samp_factor
+int v_samp_factor
+        Horizontal and vertical sampling factors for the component; must
+        be 1..4 according to the JPEG standard.  Note that larger sampling
+        factors indicate a higher-resolution component; many people find
+        this behavior quite unintuitive.  The default values are 2,2 for
+        luminance components and 1,1 for chrominance components, except
+        for grayscale where 1,1 is used.
+
+int quant_tbl_no
+        Quantization table number for component.  The default value is
+        0 for luminance components and 1 for chrominance components.
+
+int dc_tbl_no
+int ac_tbl_no
+        DC and AC entropy coding table numbers.  The default values are
+        0 for luminance components and 1 for chrominance components.
+
+int component_index
+        Must equal the component's index in comp_info[].  (Beginning in
+        release v6, the compressor library will fill this in automatically;
+        you don't have to.)
+
+
+Decompression parameter selection
+---------------------------------
+
+Decompression parameter selection is somewhat simpler than compression
+parameter selection, since all of the JPEG internal parameters are
+recorded in the source file and need not be supplied by the application.
+(Unless you are working with abbreviated files, in which case see
+"Abbreviated datastreams", below.)  Decompression parameters control
+the postprocessing done on the image to deliver it in a format suitable
+for the application's use.  Many of the parameters control speed/quality
+tradeoffs, in which faster decompression may be obtained at the price of
+a poorer-quality image.  The defaults select the highest quality (slowest)
+processing.
+
+The following fields in the JPEG object are set by jpeg_read_header() and
+may be useful to the application in choosing decompression parameters:
+
+JDIMENSION image_width                  Width and height of image
+JDIMENSION image_height
+int num_components                      Number of color components
+J_COLOR_SPACE jpeg_color_space          Colorspace of image
+boolean saw_JFIF_marker                 TRUE if a JFIF APP0 marker was seen
+  UINT8 JFIF_major_version              Version information from JFIF marker
+  UINT8 JFIF_minor_version
+  UINT8 density_unit                    Resolution data from JFIF marker
+  UINT16 X_density
+  UINT16 Y_density
+boolean saw_Adobe_marker                TRUE if an Adobe APP14 marker was seen
+  UINT8 Adobe_transform                 Color transform code from Adobe marker
+
+The JPEG color space, unfortunately, is something of a guess since the JPEG
+standard proper does not provide a way to record it.  In practice most files
+adhere to the JFIF or Adobe conventions, and the decoder will recognize these
+correctly.  See "Special color spaces", below, for more info.
+
+
+The decompression parameters that determine the basic properties of the
+returned image are:
+
+J_COLOR_SPACE out_color_space
+        Output color space.  jpeg_read_header() sets an appropriate default
+        based on jpeg_color_space; typically it will be RGB or grayscale.
+        The application can change this field to request output in a different
+        colorspace.  For example, set it to JCS_GRAYSCALE to get grayscale
+        output from a color file.  (This is useful for previewing: grayscale
+        output is faster than full color since the color components need not
+        be processed.)  Note that not all possible color space transforms are
+        currently implemented; you may need to extend jdcolor.c if you want an
+        unusual conversion.
+
+unsigned int scale_num, scale_denom
+        Scale the image by the fraction scale_num/scale_denom.  Default is
+        1/1, or no scaling.  Currently, the only supported scaling ratios
+        are M/8 with all M from 1 to 16, or any reduced fraction thereof (such
+        as 1/2, 3/4, etc.)  (The library design allows for arbitrary
+        scaling ratios but this is not likely to be implemented any time soon.)
+        Smaller scaling ratios permit significantly faster decoding since
+        fewer pixels need be processed and a simpler IDCT method can be used.
+
+boolean quantize_colors
+        If set TRUE, colormapped output will be delivered.  Default is FALSE,
+        meaning that full-color output will be delivered.
+
+The next three parameters are relevant only if quantize_colors is TRUE.
+
+int desired_number_of_colors
+        Maximum number of colors to use in generating a library-supplied color
+        map (the actual number of colors is returned in a different field).
+        Default 256.  Ignored when the application supplies its own color map.
+
+boolean two_pass_quantize
+        If TRUE, an extra pass over the image is made to select a custom color
+        map for the image.  This usually looks a lot better than the one-size-
+        fits-all colormap that is used otherwise.  Default is TRUE.  Ignored
+        when the application supplies its own color map.
+
+J_DITHER_MODE dither_mode
+        Selects color dithering method.  Supported values are:
+                JDITHER_NONE    no dithering: fast, very low quality
+                JDITHER_ORDERED ordered dither: moderate speed and quality
+                JDITHER_FS      Floyd-Steinberg dither: slow, high quality
+        Default is JDITHER_FS.  (At present, ordered dither is implemented
+        only in the single-pass, standard-colormap case.  If you ask for
+        ordered dither when two_pass_quantize is TRUE or when you supply
+        an external color map, you'll get F-S dithering.)
+
+When quantize_colors is TRUE, the target color map is described by the next
+two fields.  colormap is set to NULL by jpeg_read_header().  The application
+can supply a color map by setting colormap non-NULL and setting
+actual_number_of_colors to the map size.  Otherwise, jpeg_start_decompress()
+selects a suitable color map and sets these two fields itself.
+[Implementation restriction: at present, an externally supplied colormap is
+only accepted for 3-component output color spaces.]
+
+JSAMPARRAY colormap
+        The color map, represented as a 2-D pixel array of out_color_components
+        rows and actual_number_of_colors columns.  Ignored if not quantizing.
+        CAUTION: if the JPEG library creates its own colormap, the storage
+        pointed to by this field is released by jpeg_finish_decompress().
+        Copy the colormap somewhere else first, if you want to save it.
+
+int actual_number_of_colors
+        The number of colors in the color map.
+
+Additional decompression parameters that the application may set include:
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: slow but accurate integer algorithm
+                JDCT_IFAST: faster, less accurate integer method
+                JDCT_FLOAT: floating-point method
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        In libjpeg-turbo, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW when using the x86/x86-64 SIMD extensions (results may vary
+        with other SIMD implementations, or when using libjpeg-turbo without
+        SIMD extensions.)  If the JPEG image was compressed using a quality
+        level of 85 or below, then there should be little or no perceptible
+        difference between the two algorithms.  When decompressing images that
+        were compressed using quality levels above 85, however, the difference
+        between JDCT_IFAST and JDCT_ISLOW becomes more pronounced.  With images
+        compressed using quality=97, for instance, JDCT_IFAST incurs generally
+        about a 4-6 dB loss (in PSNR) relative to JDCT_ISLOW, but this can be
+        larger for some images.  If you can avoid it, do not use JDCT_IFAST
+        when decompressing images that were compressed using quality levels
+        above 97.  The algorithm often degenerates for such images and can
+        actually produce a more lossy output image than if the JPEG image had
+        been compressed using lower quality levels.  JDCT_FLOAT is mainly a
+        legacy feature.  It does not produce significantly more accurate
+        results than the ISLOW method, and it is much slower.  The FLOAT method
+        may also give different results on different machines due to varying
+        roundoff behavior, whereas the integer methods should give the same
+        results on all machines.
+
+boolean do_fancy_upsampling
+        If TRUE, do careful upsampling of chroma components.  If FALSE,
+        a faster but sloppier method is used.  Default is TRUE.  The visual
+        impact of the sloppier method is often very small.
+
+boolean do_block_smoothing
+        If TRUE, interblock smoothing is applied in early stages of decoding
+        progressive JPEG files; if FALSE, not.  Default is TRUE.  Early
+        progression stages look "fuzzy" with smoothing, "blocky" without.
+        In any case, block smoothing ceases to be applied after the first few
+        AC coefficients are known to full accuracy, so it is relevant only
+        when using buffered-image mode for progressive images.
+
+boolean enable_1pass_quant
+boolean enable_external_quant
+boolean enable_2pass_quant
+        These are significant only in buffered-image mode, which is
+        described in its own section below.
+
+
+The output image dimensions are given by the following fields.  These are
+computed from the source image dimensions and the decompression parameters
+by jpeg_start_decompress().  You can also call jpeg_calc_output_dimensions()
+to obtain the values that will result from the current parameter settings.
+This can be useful if you are trying to pick a scaling ratio that will get
+close to a desired target size.  It's also important if you are using the
+JPEG library's memory manager to allocate output buffer space, because you
+are supposed to request such buffers *before* jpeg_start_decompress().
+
+JDIMENSION output_width         Actual dimensions of output image.
+JDIMENSION output_height
+int out_color_components        Number of color components in out_color_space.
+int output_components           Number of color components returned.
+int rec_outbuf_height           Recommended height of scanline buffer.
+
+When quantizing colors, output_components is 1, indicating a single color map
+index per pixel.  Otherwise it equals out_color_components.  The output arrays
+are required to be output_width * output_components JSAMPLEs wide.
+
+rec_outbuf_height is the recommended minimum height (in scanlines) of the
+buffer passed to jpeg_read_scanlines().  If the buffer is smaller, the
+library will still work, but time will be wasted due to unnecessary data
+copying.  In high-quality modes, rec_outbuf_height is always 1, but some
+faster, lower-quality modes set it to larger values (typically 2 to 4).
+If you are going to ask for a high-speed processing mode, you may as well
+go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
+(An output buffer larger than rec_outbuf_height lines is OK, but won't
+provide any material speed improvement over that height.)
+
+
+Special color spaces
+--------------------
+
+The JPEG standard itself is "color blind" and doesn't specify any particular
+color space.  It is customary to convert color data to a luminance/chrominance
+color space before compressing, since this permits greater compression.  The
+existing de-facto JPEG file format standards specify YCbCr or grayscale data
+(JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe).  For special
+applications such as multispectral images, other color spaces can be used,
+but it must be understood that such files will be unportable.
+
+The JPEG library can handle the most common colorspace conversions (namely
+RGB <=> YCbCr and CMYK <=> YCCK).  It can also deal with data of an unknown
+color space, passing it through without conversion.  If you deal extensively
+with an unusual color space, you can easily extend the library to understand
+additional color spaces and perform appropriate conversions.
+
+For compression, the source data's color space is specified by field
+in_color_space.  This is transformed to the JPEG file's color space given
+by jpeg_color_space.  jpeg_set_defaults() chooses a reasonable JPEG color
+space depending on in_color_space, but you can override this by calling
+jpeg_set_colorspace().  Of course you must select a supported transformation.
+jccolor.c currently supports the following transformations:
+        RGB => YCbCr
+        RGB => GRAYSCALE
+        YCbCr => GRAYSCALE
+        CMYK => YCCK
+plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
+YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
+
+The de-facto file format standards (JFIF and Adobe) specify APPn markers that
+indicate the color space of the JPEG file.  It is important to ensure that
+these are written correctly, or omitted if the JPEG file's color space is not
+one of the ones supported by the de-facto standards.  jpeg_set_colorspace()
+will set the compression parameters to include or omit the APPn markers
+properly, so long as it is told the truth about the JPEG color space.
+For example, if you are writing some random 3-component color space without
+conversion, don't try to fake out the library by setting in_color_space and
+jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN.  You may want to write an
+APPn marker of your own devising to identify the colorspace --- see "Special
+markers", below.
+
+When told that the color space is UNKNOWN, the library will default to using
+luminance-quality compression parameters for all color components.  You may
+well want to change these parameters.  See the source code for
+jpeg_set_colorspace(), in jcparam.c, for details.
+
+For decompression, the JPEG file's color space is given in jpeg_color_space,
+and this is transformed to the output color space out_color_space.
+jpeg_read_header's setting of jpeg_color_space can be relied on if the file
+conforms to JFIF or Adobe conventions, but otherwise it is no better than a
+guess.  If you know the JPEG file's color space for certain, you can override
+jpeg_read_header's guess by setting jpeg_color_space.  jpeg_read_header also
+selects a default output color space based on (its guess of) jpeg_color_space;
+set out_color_space to override this.  Again, you must select a supported
+transformation.  jdcolor.c currently supports
+        YCbCr => RGB
+        YCbCr => GRAYSCALE
+        RGB => GRAYSCALE
+        GRAYSCALE => RGB
+        YCCK => CMYK
+as well as the null transforms.  (Since GRAYSCALE=>RGB is provided, an
+application can force grayscale JPEGs to look like color JPEGs if it only
+wants to handle one case.)
+
+The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
+(it weights distances appropriately for RGB colors).  You'll need to modify
+the code if you want to use it for non-RGB output color spaces.  Note that
+jquant2.c is used to map to an application-supplied colormap as well as for
+the normal two-pass colormap selection process.
+
+CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
+files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
+This is arguably a bug in Photoshop, but if you need to work with Photoshop
+CMYK files, you will have to deal with it in your application.  We cannot
+"fix" this in the library by inverting the data during the CMYK<=>YCCK
+transform, because that would break other applications, notably Ghostscript.
+Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
+data in the same inverted-YCCK representation used in bare JPEG files, but
+the surrounding PostScript code performs an inversion using the PS image
+operator.  I am told that Photoshop 3.0 will write uninverted YCCK in
+EPS/JPEG files, and will omit the PS-level inversion.  (But the data
+polarity used in bare JPEG files will not change in 3.0.)  In either case,
+the JPEG library must not invert the data itself, or else Ghostscript would
+read these EPS files incorrectly.
+
+
+Error handling
+--------------
+
+When the default error handler is used, any error detected inside the JPEG
+routines will cause a message to be printed on stderr, followed by exit().
+You can supply your own error handling routines to override this behavior
+and to control the treatment of nonfatal warnings and trace/debug messages.
+The file example.c illustrates the most common case, which is to have the
+application regain control after an error rather than exiting.
+
+The JPEG library never writes any message directly; it always goes through
+the error handling routines.  Three classes of messages are recognized:
+  * Fatal errors: the library cannot continue.
+  * Warnings: the library can continue, but the data is corrupt, and a
+    damaged output image is likely to result.
+  * Trace/informational messages.  These come with a trace level indicating
+    the importance of the message; you can control the verbosity of the
+    program by adjusting the maximum trace level that will be displayed.
+
+You may, if you wish, simply replace the entire JPEG error handling module
+(jerror.c) with your own code.  However, you can avoid code duplication by
+only replacing some of the routines depending on the behavior you need.
+This is accomplished by calling jpeg_std_error() as usual, but then overriding
+some of the method pointers in the jpeg_error_mgr struct, as illustrated by
+example.c.
+
+All of the error handling routines will receive a pointer to the JPEG object
+(a j_common_ptr which points to either a jpeg_compress_struct or a
+jpeg_decompress_struct; if you need to tell which, test the is_decompressor
+field).  This struct includes a pointer to the error manager struct in its
+"err" field.  Frequently, custom error handler routines will need to access
+additional data which is not known to the JPEG library or the standard error
+handler.  The most convenient way to do this is to embed either the JPEG
+object or the jpeg_error_mgr struct in a larger structure that contains
+additional fields; then casting the passed pointer provides access to the
+additional fields.  Again, see example.c for one way to do it.  (Beginning
+with IJG version 6b, there is also a void pointer "client_data" in each
+JPEG object, which the application can also use to find related data.
+The library does not touch client_data at all.)
+
+The individual methods that you might wish to override are:
+
+error_exit (j_common_ptr cinfo)
+        Receives control for a fatal error.  Information sufficient to
+        generate the error message has been stored in cinfo->err; call
+        output_message to display it.  Control must NOT return to the caller;
+        generally this routine will exit() or longjmp() somewhere.
+        Typically you would override this routine to get rid of the exit()
+        default behavior.  Note that if you continue processing, you should
+        clean up the JPEG object with jpeg_abort() or jpeg_destroy().
+
+output_message (j_common_ptr cinfo)
+        Actual output of any JPEG message.  Override this to send messages
+        somewhere other than stderr.  Note that this method does not know
+        how to generate a message, only where to send it.
+
+format_message (j_common_ptr cinfo, char *buffer)
+        Constructs a readable error message string based on the error info
+        stored in cinfo->err.  This method is called by output_message.  Few
+        applications should need to override this method.  One possible
+        reason for doing so is to implement dynamic switching of error message
+        language.
+
+emit_message (j_common_ptr cinfo, int msg_level)
+        Decide whether or not to emit a warning or trace message; if so,
+        calls output_message.  The main reason for overriding this method
+        would be to abort on warnings.  msg_level is -1 for warnings,
+        0 and up for trace messages.
+
+Only error_exit() and emit_message() are called from the rest of the JPEG
+library; the other two are internal to the error handler.
+
+The actual message texts are stored in an array of strings which is pointed to
+by the field err->jpeg_message_table.  The messages are numbered from 0 to
+err->last_jpeg_message, and it is these code numbers that are used in the
+JPEG library code.  You could replace the message texts (for instance, with
+messages in French or German) by changing the message table pointer.  See
+jerror.h for the default texts.  CAUTION: this table will almost certainly
+change or grow from one library version to the next.
+
+It may be useful for an application to add its own message texts that are
+handled by the same mechanism.  The error handler supports a second "add-on"
+message table for this purpose.  To define an addon table, set the pointer
+err->addon_message_table and the message numbers err->first_addon_message and
+err->last_addon_message.  If you number the addon messages beginning at 1000
+or so, you won't have to worry about conflicts with the library's built-in
+messages.  See the sample applications cjpeg/djpeg for an example of using
+addon messages (the addon messages are defined in cderror.h).
+
+Actual invocation of the error handler is done via macros defined in jerror.h:
+        ERREXITn(...)   for fatal errors
+        WARNMSn(...)    for corrupt-data warnings
+        TRACEMSn(...)   for trace and informational messages.
+These macros store the message code and any additional parameters into the
+error handler struct, then invoke the error_exit() or emit_message() method.
+The variants of each macro are for varying numbers of additional parameters.
+The additional parameters are inserted into the generated message using
+standard printf() format codes.
+
+See jerror.h and jerror.c for further details.
+
+
+Compressed data handling (source and destination managers)
+----------------------------------------------------------
+
+The JPEG compression library sends its compressed data to a "destination
+manager" module.  The default destination manager just writes the data to a
+memory buffer or to a stdio stream, but you can provide your own manager to
+do something else.  Similarly, the decompression library calls a "source
+manager" to obtain the compressed data; you can provide your own source
+manager if you want the data to come from somewhere other than a memory
+buffer or a stdio stream.
+
+In both cases, compressed data is processed a bufferload at a time: the
+destination or source manager provides a work buffer, and the library invokes
+the manager only when the buffer is filled or emptied.  (You could define a
+one-character buffer to force the manager to be invoked for each byte, but
+that would be rather inefficient.)  The buffer's size and location are
+controlled by the manager, not by the library.  For example, the memory
+source manager just makes the buffer pointer and length point to the original
+data in memory.  In this case the buffer-reload procedure will be invoked
+only if the decompressor ran off the end of the datastream, which would
+indicate an erroneous datastream.
+
+The work buffer is defined as an array of datatype JOCTET, which is generally
+"char" or "unsigned char".  On a machine where char is not exactly 8 bits
+wide, you must define JOCTET as a wider data type and then modify the data
+source and destination modules to transcribe the work arrays into 8-bit units
+on external storage.
+
+A data destination manager struct contains a pointer and count defining the
+next byte to write in the work buffer and the remaining free space:
+
+        JOCTET *next_output_byte;   /* => next byte to write in buffer */
+        size_t free_in_buffer;      /* # of byte spaces remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is filled.  The manager's empty_output_buffer method must reset the pointer
+and count.  The manager is expected to remember the buffer's starting address
+and total size in private fields not visible to the library.
+
+A data destination manager provides three methods:
+
+init_destination (j_compress_ptr cinfo)
+        Initialize destination.  This is called by jpeg_start_compress()
+        before any data is actually written.  It must initialize
+        next_output_byte and free_in_buffer.  free_in_buffer must be
+        initialized to a positive value.
+
+empty_output_buffer (j_compress_ptr cinfo)
+        This is called whenever the buffer has filled (free_in_buffer
+        reaches zero).  In typical applications, it should write out the
+        *entire* buffer (use the saved start address and buffer length;
+        ignore the current state of next_output_byte and free_in_buffer).
+        Then reset the pointer & count to the start of the buffer, and
+        return TRUE indicating that the buffer has been dumped.
+        free_in_buffer must be set to a positive value when TRUE is
+        returned.  A FALSE return should only be used when I/O suspension is
+        desired (this operating mode is discussed in the next section).
+
+term_destination (j_compress_ptr cinfo)
+        Terminate destination --- called by jpeg_finish_compress() after all
+        data has been written.  In most applications, this must flush any
+        data remaining in the buffer.  Use either next_output_byte or
+        free_in_buffer to determine how much data is in the buffer.
+
+term_destination() is NOT called by jpeg_abort() or jpeg_destroy().  If you
+want the destination manager to be cleaned up during an abort, you must do it
+yourself.
+
+You will also need code to create a jpeg_destination_mgr struct, fill in its
+method pointers, and insert a pointer to the struct into the "dest" field of
+the JPEG compression object.  This can be done in-line in your setup code if
+you like, but it's probably cleaner to provide a separate routine similar to
+the jpeg_stdio_dest() or jpeg_mem_dest() routines of the supplied destination
+managers.
+
+Decompression source managers follow a parallel design, but with some
+additional frammishes.  The source manager struct contains a pointer and count
+defining the next byte to read from the work buffer and the number of bytes
+remaining:
+
+        const JOCTET *next_input_byte;  /* => next byte to read from buffer */
+        size_t bytes_in_buffer;         /* # of bytes remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is emptied.  The manager's fill_input_buffer method must reset the pointer and
+count.  In most applications, the manager must remember the buffer's starting
+address and total size in private fields not visible to the library.
+
+A data source manager provides five methods:
+
+init_source (j_decompress_ptr cinfo)
+        Initialize source.  This is called by jpeg_read_header() before any
+        data is actually read.  Unlike init_destination(), it may leave
+        bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
+        will occur immediately).
+
+fill_input_buffer (j_decompress_ptr cinfo)
+        This is called whenever bytes_in_buffer has reached zero and more
+        data is wanted.  In typical applications, it should read fresh data
+        into the buffer (ignoring the current state of next_input_byte and
+        bytes_in_buffer), reset the pointer & count to the start of the
+        buffer, and return TRUE indicating that the buffer has been reloaded.
+        It is not necessary to fill the buffer entirely, only to obtain at
+        least one more byte.  bytes_in_buffer MUST be set to a positive value
+        if TRUE is returned.  A FALSE return should only be used when I/O
+        suspension is desired (this mode is discussed in the next section).
+
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+        Skip num_bytes worth of data.  The buffer pointer and count should
+        be advanced over num_bytes input bytes, refilling the buffer as
+        needed.  This is used to skip over a potentially large amount of
+        uninteresting data (such as an APPn marker).  In some applications
+        it may be possible to optimize away the reading of the skipped data,
+        but it's not clear that being smart is worth much trouble; large
+        skips are uncommon.  bytes_in_buffer may be zero on return.
+        A zero or negative skip count should be treated as a no-op.
+
+resync_to_restart (j_decompress_ptr cinfo, int desired)
+        This routine is called only when the decompressor has failed to find
+        a restart (RSTn) marker where one is expected.  Its mission is to
+        find a suitable point for resuming decompression.  For most
+        applications, we recommend that you just use the default resync
+        procedure, jpeg_resync_to_restart().  However, if you are able to back
+        up in the input data stream, or if you have a-priori knowledge about
+        the likely location of restart markers, you may be able to do better.
+        Read the read_restart_marker() and jpeg_resync_to_restart() routines
+        in jdmarker.c if you think you'd like to implement your own resync
+        procedure.
+
+term_source (j_decompress_ptr cinfo)
+        Terminate source --- called by jpeg_finish_decompress() after all
+        data has been read.  Often a no-op.
+
+For both fill_input_buffer() and skip_input_data(), there is no such thing
+as an EOF return.  If the end of the file has been reached, the routine has
+a choice of exiting via ERREXIT() or inserting fake data into the buffer.
+In most cases, generating a warning message and inserting a fake EOI marker
+is the best course of action --- this will allow the decompressor to output
+however much of the image is there.  In pathological cases, the decompressor
+may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
+jdatasrc.c illustrates the recommended error recovery behavior.
+
+term_source() is NOT called by jpeg_abort() or jpeg_destroy().  If you want
+the source manager to be cleaned up during an abort, you must do it yourself.
+
+You will also need code to create a jpeg_source_mgr struct, fill in its method
+pointers, and insert a pointer to the struct into the "src" field of the JPEG
+decompression object.  This can be done in-line in your setup code if you
+like, but it's probably cleaner to provide a separate routine similar to the
+jpeg_stdio_src() or jpeg_mem_src() routines of the supplied source managers.
+
+For more information, consult the memory and stdio source and destination
+managers in jdatasrc.c and jdatadst.c.
+
+
+I/O suspension
+--------------
+
+Some applications need to use the JPEG library as an incremental memory-to-
+memory filter: when the compressed data buffer is filled or emptied, they want
+control to return to the outer loop, rather than expecting that the buffer can
+be emptied or reloaded within the data source/destination manager subroutine.
+The library supports this need by providing an "I/O suspension" mode, which we
+describe in this section.
+
+The I/O suspension mode is not a panacea: nothing is guaranteed about the
+maximum amount of time spent in any one call to the library, so it will not
+eliminate response-time problems in single-threaded applications.  If you
+need guaranteed response time, we suggest you "bite the bullet" and implement
+a real multi-tasking capability.
+
+To use I/O suspension, cooperation is needed between the calling application
+and the data source or destination manager; you will always need a custom
+source/destination manager.  (Please read the previous section if you haven't
+already.)  The basic idea is that the empty_output_buffer() or
+fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
+that it has done nothing.  Upon seeing this, the JPEG library suspends
+operation and returns to its caller.  The surrounding application is
+responsible for emptying or refilling the work buffer before calling the
+JPEG library again.
+
+Compression suspension:
+
+For compression suspension, use an empty_output_buffer() routine that returns
+FALSE; typically it will not do anything else.  This will cause the
+compressor to return to the caller of jpeg_write_scanlines(), with the return
+value indicating that not all the supplied scanlines have been accepted.
+The application must make more room in the output buffer, adjust the output
+buffer pointer/count appropriately, and then call jpeg_write_scanlines()
+again, pointing to the first unconsumed scanline.
+
+When forced to suspend, the compressor will backtrack to a convenient stopping
+point (usually the start of the current MCU); it will regenerate some output
+data when restarted.  Therefore, although empty_output_buffer() is only
+called when the buffer is filled, you should NOT write out the entire buffer
+after a suspension.  Write only the data up to the current position of
+next_output_byte/free_in_buffer.  The data beyond that point will be
+regenerated after resumption.
+
+Because of the backtracking behavior, a good-size output buffer is essential
+for efficiency; you don't want the compressor to suspend often.  (In fact, an
+overly small buffer could lead to infinite looping, if a single MCU required
+more data than would fit in the buffer.)  We recommend a buffer of at least
+several Kbytes.  You may want to insert explicit code to ensure that you don't
+call jpeg_write_scanlines() unless there is a reasonable amount of space in
+the output buffer; in other words, flush the buffer before trying to compress
+more data.
+
+The compressor does not allow suspension while it is trying to write JPEG
+markers at the beginning and end of the file.  This means that:
+  * At the beginning of a compression operation, there must be enough free
+    space in the output buffer to hold the header markers (typically 600 or
+    so bytes).  The recommended buffer size is bigger than this anyway, so
+    this is not a problem as long as you start with an empty buffer.  However,
+    this restriction might catch you if you insert large special markers, such
+    as a JFIF thumbnail image, without flushing the buffer afterwards.
+  * When you call jpeg_finish_compress(), there must be enough space in the
+    output buffer to emit any buffered data and the final EOI marker.  In the
+    current implementation, half a dozen bytes should suffice for this, but
+    for safety's sake we recommend ensuring that at least 100 bytes are free
+    before calling jpeg_finish_compress().
+
+A more significant restriction is that jpeg_finish_compress() cannot suspend.
+This means you cannot use suspension with multi-pass operating modes, namely
+Huffman code optimization and multiple-scan output.  Those modes write the
+whole file during jpeg_finish_compress(), which will certainly result in
+buffer overrun.  (Note that this restriction applies only to compression,
+not decompression.  The decompressor supports input suspension in all of its
+operating modes.)
+
+Decompression suspension:
+
+For decompression suspension, use a fill_input_buffer() routine that simply
+returns FALSE (except perhaps during error recovery, as discussed below).
+This will cause the decompressor to return to its caller with an indication
+that suspension has occurred.  This can happen at four places:
+  * jpeg_read_header(): will return JPEG_SUSPENDED.
+  * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
+  * jpeg_read_scanlines(): will return the number of scanlines already
+        completed (possibly 0).
+  * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
+The surrounding application must recognize these cases, load more data into
+the input buffer, and repeat the call.  In the case of jpeg_read_scanlines(),
+increment the passed pointers past any scanlines successfully read.
+
+Just as with compression, the decompressor will typically backtrack to a
+convenient restart point before suspending.  When fill_input_buffer() is
+called, next_input_byte/bytes_in_buffer point to the current restart point,
+which is where the decompressor will backtrack to if FALSE is returned.
+The data beyond that position must NOT be discarded if you suspend; it needs
+to be re-read upon resumption.  In most implementations, you'll need to shift
+this data down to the start of your work buffer and then load more data after
+it.  Again, this behavior means that a several-Kbyte work buffer is essential
+for decent performance; furthermore, you should load a reasonable amount of
+new data before resuming decompression.  (If you loaded, say, only one new
+byte each time around, you could waste a LOT of cycles.)
+
+The skip_input_data() source manager routine requires special care in a
+suspension scenario.  This routine is NOT granted the ability to suspend the
+decompressor; it can decrement bytes_in_buffer to zero, but no more.  If the
+requested skip distance exceeds the amount of data currently in the input
+buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
+additional skip distance somewhere else.  The decompressor will immediately
+call fill_input_buffer(), which should return FALSE, which will cause a
+suspension return.  The surrounding application must then arrange to discard
+the recorded number of bytes before it resumes loading the input buffer.
+(Yes, this design is rather baroque, but it avoids complexity in the far more
+common case where a non-suspending source manager is used.)
+
+If the input data has been exhausted, we recommend that you emit a warning
+and insert dummy EOI markers just as a non-suspending data source manager
+would do.  This can be handled either in the surrounding application logic or
+within fill_input_buffer(); the latter is probably more efficient.  If
+fill_input_buffer() knows that no more data is available, it can set the
+pointer/count to point to a dummy EOI marker and then return TRUE just as
+though it had read more data in a non-suspending situation.
+
+The decompressor does not attempt to suspend within standard JPEG markers;
+instead it will backtrack to the start of the marker and reprocess the whole
+marker next time.  Hence the input buffer must be large enough to hold the
+longest standard marker in the file.  Standard JPEG markers should normally
+not exceed a few hundred bytes each (DHT tables are typically the longest).
+We recommend at least a 2K buffer for performance reasons, which is much
+larger than any correct marker is likely to be.  For robustness against
+damaged marker length counts, you may wish to insert a test in your
+application for the case that the input buffer is completely full and yet
+the decoder has suspended without consuming any data --- otherwise, if this
+situation did occur, it would lead to an endless loop.  (The library can't
+provide this test since it has no idea whether "the buffer is full", or
+even whether there is a fixed-size input buffer.)
+
+The input buffer would need to be 64K to allow for arbitrary COM or APPn
+markers, but these are handled specially: they are either saved into allocated
+memory, or skipped over by calling skip_input_data().  In the former case,
+suspension is handled correctly, and in the latter case, the problem of
+buffer overrun is placed on skip_input_data's shoulders, as explained above.
+Note that if you provide your own marker handling routine for large markers,
+you should consider how to deal with buffer overflow.
+
+Multiple-buffer management:
+
+In some applications it is desirable to store the compressed data in a linked
+list of buffer areas, so as to avoid data copying.  This can be handled by
+having empty_output_buffer() or fill_input_buffer() set the pointer and count
+to reference the next available buffer; FALSE is returned only if no more
+buffers are available.  Although seemingly straightforward, there is a
+pitfall in this approach: the backtrack that occurs when FALSE is returned
+could back up into an earlier buffer.  For example, when fill_input_buffer()
+is called, the current pointer & count indicate the backtrack restart point.
+Since fill_input_buffer() will set the pointer and count to refer to a new
+buffer, the restart position must be saved somewhere else.  Suppose a second
+call to fill_input_buffer() occurs in the same library call, and no
+additional input data is available, so fill_input_buffer must return FALSE.
+If the JPEG library has not moved the pointer/count forward in the current
+buffer, then *the correct restart point is the saved position in the prior
+buffer*.  Prior buffers may be discarded only after the library establishes
+a restart point within a later buffer.  Similar remarks apply for output into
+a chain of buffers.
+
+The library will never attempt to backtrack over a skip_input_data() call,
+so any skipped data can be permanently discarded.  You still have to deal
+with the case of skipping not-yet-received data, however.
+
+It's much simpler to use only a single buffer; when fill_input_buffer() is
+called, move any unconsumed data (beyond the current pointer/count) down to
+the beginning of this buffer and then load new data into the remaining buffer
+space.  This approach requires a little more data copying but is far easier
+to get right.
+
+
+Progressive JPEG support
+------------------------
+
+Progressive JPEG rearranges the stored data into a series of scans of
+increasing quality.  In situations where a JPEG file is transmitted across a
+slow communications link, a decoder can generate a low-quality image very
+quickly from the first scan, then gradually improve the displayed quality as
+more scans are received.  The final image after all scans are complete is
+identical to that of a regular (sequential) JPEG file of the same quality
+setting.  Progressive JPEG files are often slightly smaller than equivalent
+sequential JPEG files, but the possibility of incremental display is the main
+reason for using progressive JPEG.
+
+The IJG encoder library generates progressive JPEG files when given a
+suitable "scan script" defining how to divide the data into scans.
+Creation of progressive JPEG files is otherwise transparent to the encoder.
+Progressive JPEG files can also be read transparently by the decoder library.
+If the decoding application simply uses the library as defined above, it
+will receive a final decoded image without any indication that the file was
+progressive.  Of course, this approach does not allow incremental display.
+To perform incremental display, an application needs to use the decoder
+library's "buffered-image" mode, in which it receives a decoded image
+multiple times.
+
+Each displayed scan requires about as much work to decode as a full JPEG
+image of the same size, so the decoder must be fairly fast in relation to the
+data transmission rate in order to make incremental display useful.  However,
+it is possible to skip displaying the image and simply add the incoming bits
+to the decoder's coefficient buffer.  This is fast because only Huffman
+decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
+The IJG decoder library allows the application to switch dynamically between
+displaying the image and simply absorbing the incoming bits.  A properly
+coded application can automatically adapt the number of display passes to
+suit the time available as the image is received.  Also, a final
+higher-quality display cycle can be performed from the buffered data after
+the end of the file is reached.
+
+Progressive compression:
+
+To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
+set the scan_info cinfo field to point to an array of scan descriptors, and
+perform compression as usual.  Instead of constructing your own scan list,
+you can call the jpeg_simple_progression() helper routine to create a
+recommended progression sequence; this method should be used by all
+applications that don't want to get involved in the nitty-gritty of
+progressive scan sequence design.  (If you want to provide user control of
+scan sequences, you may wish to borrow the scan script reading code found
+in rdswitch.c, so that you can read scan script files just like cjpeg's.)
+When scan_info is not NULL, the compression library will store DCT'd data
+into a buffer array as jpeg_write_scanlines() is called, and will emit all
+the requested scans during jpeg_finish_compress().  This implies that
+multiple-scan output cannot be created with a suspending data destination
+manager, since jpeg_finish_compress() does not support suspension.  We
+should also note that the compressor currently forces Huffman optimization
+mode when creating a progressive JPEG file, because the default Huffman
+tables are unsuitable for progressive files.
+
+Progressive decompression:
+
+When buffered-image mode is not used, the decoder library will read all of
+a multi-scan file during jpeg_start_decompress(), so that it can provide a
+final decoded image.  (Here "multi-scan" means either progressive or
+multi-scan sequential.)  This makes multi-scan files transparent to the
+decoding application.  However, existing applications that used suspending
+input with version 5 of the IJG library will need to be modified to check
+for a suspension return from jpeg_start_decompress().
+
+To perform incremental display, an application must use the library's
+buffered-image mode.  This is described in the next section.
+
+
+Buffered-image mode
+-------------------
+
+In buffered-image mode, the library stores the partially decoded image in a
+coefficient buffer, from which it can be read out as many times as desired.
+This mode is typically used for incremental display of progressive JPEG files,
+but it can be used with any JPEG file.  Each scan of a progressive JPEG file
+adds more data (more detail) to the buffered image.  The application can
+display in lockstep with the source file (one display pass per input scan),
+or it can allow input processing to outrun display processing.  By making
+input and display processing run independently, it is possible for the
+application to adapt progressive display to a wide range of data transmission
+rates.
+
+The basic control flow for buffered-image decoding is
+
+        jpeg_create_decompress()
+        set data source
+        jpeg_read_header()
+        set overall decompression parameters
+        cinfo.buffered_image = TRUE;    /* select buffered-image mode */
+        jpeg_start_decompress()
+        for (each output pass) {
+            adjust output decompression parameters if required
+            jpeg_start_output()         /* start a new output pass */
+            for (all scanlines in image) {
+                jpeg_read_scanlines()
+                display scanlines
+            }
+            jpeg_finish_output()        /* terminate output pass */
+        }
+        jpeg_finish_decompress()
+        jpeg_destroy_decompress()
+
+This differs from ordinary unbuffered decoding in that there is an additional
+level of looping.  The application can choose how many output passes to make
+and how to display each pass.
+
+The simplest approach to displaying progressive images is to do one display
+pass for each scan appearing in the input file.  In this case the outer loop
+condition is typically
+        while (! jpeg_input_complete(&cinfo))
+and the start-output call should read
+        jpeg_start_output(&cinfo, cinfo.input_scan_number);
+The second parameter to jpeg_start_output() indicates which scan of the input
+file is to be displayed; the scans are numbered starting at 1 for this
+purpose.  (You can use a loop counter starting at 1 if you like, but using
+the library's input scan counter is easier.)  The library automatically reads
+data as necessary to complete each requested scan, and jpeg_finish_output()
+advances to the next scan or end-of-image marker (hence input_scan_number
+will be incremented by the time control arrives back at jpeg_start_output()).
+With this technique, data is read from the input file only as needed, and
+input and output processing run in lockstep.
+
+After reading the final scan and reaching the end of the input file, the
+buffered image remains available; it can be read additional times by
+repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
+sequence.  For example, a useful technique is to use fast one-pass color
+quantization for display passes made while the image is arriving, followed by
+a final display pass using two-pass quantization for highest quality.  This
+is done by changing the library parameters before the final output pass.
+Changing parameters between passes is discussed in detail below.
+
+In general the last scan of a progressive file cannot be recognized as such
+until after it is read, so a post-input display pass is the best approach if
+you want special processing in the final pass.
+
+When done with the image, be sure to call jpeg_finish_decompress() to release
+the buffered image (or just use jpeg_destroy_decompress()).
+
+If input data arrives faster than it can be displayed, the application can
+cause the library to decode input data in advance of what's needed to produce
+output.  This is done by calling the routine jpeg_consume_input().
+The return value is one of the following:
+        JPEG_REACHED_SOS:    reached an SOS marker (the start of a new scan)
+        JPEG_REACHED_EOI:    reached the EOI marker (end of image)
+        JPEG_ROW_COMPLETED:  completed reading one MCU row of compressed data
+        JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
+        JPEG_SUSPENDED:      suspended before completing any of the above
+(JPEG_SUSPENDED can occur only if a suspending data source is used.)  This
+routine can be called at any time after initializing the JPEG object.  It
+reads some additional data and returns when one of the indicated significant
+events occurs.  (If called after the EOI marker is reached, it will
+immediately return JPEG_REACHED_EOI without attempting to read more data.)
+
+The library's output processing will automatically call jpeg_consume_input()
+whenever the output processing overtakes the input; thus, simple lockstep
+display requires no direct calls to jpeg_consume_input().  But by adding
+calls to jpeg_consume_input(), you can absorb data in advance of what is
+being displayed.  This has two benefits:
+  * You can limit buildup of unprocessed data in your input buffer.
+  * You can eliminate extra display passes by paying attention to the
+    state of the library's input processing.
+
+The first of these benefits only requires interspersing calls to
+jpeg_consume_input() with your display operations and any other processing
+you may be doing.  To avoid wasting cycles due to backtracking, it's best to
+call jpeg_consume_input() only after a hundred or so new bytes have arrived.
+This is discussed further under "I/O suspension", above.  (Note: the JPEG
+library currently is not thread-safe.  You must not call jpeg_consume_input()
+from one thread of control if a different library routine is working on the
+same JPEG object in another thread.)
+
+When input arrives fast enough that more than one new scan is available
+before you start a new output pass, you may as well skip the output pass
+corresponding to the completed scan.  This occurs for free if you pass
+cinfo.input_scan_number as the target scan number to jpeg_start_output().
+The input_scan_number field is simply the index of the scan currently being
+consumed by the input processor.  You can ensure that this is up-to-date by
+emptying the input buffer just before calling jpeg_start_output(): call
+jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
+JPEG_REACHED_EOI.
+
+The target scan number passed to jpeg_start_output() is saved in the
+cinfo.output_scan_number field.  The library's output processing calls
+jpeg_consume_input() whenever the current input scan number and row within
+that scan is less than or equal to the current output scan number and row.
+Thus, input processing can "get ahead" of the output processing but is not
+allowed to "fall behind".  You can achieve several different effects by
+manipulating this interlock rule.  For example, if you pass a target scan
+number greater than the current input scan number, the output processor will
+wait until that scan starts to arrive before producing any output.  (To avoid
+an infinite loop, the target scan number is automatically reset to the last
+scan number when the end of image is reached.  Thus, if you specify a large
+target scan number, the library will just absorb the entire input file and
+then perform an output pass.  This is effectively the same as what
+jpeg_start_decompress() does when you don't select buffered-image mode.)
+When you pass a target scan number equal to the current input scan number,
+the image is displayed no faster than the current input scan arrives.  The
+final possibility is to pass a target scan number less than the current input
+scan number; this disables the input/output interlock and causes the output
+processor to simply display whatever it finds in the image buffer, without
+waiting for input.  (However, the library will not accept a target scan
+number less than one, so you can't avoid waiting for the first scan.)
+
+When data is arriving faster than the output display processing can advance
+through the image, jpeg_consume_input() will store data into the buffered
+image beyond the point at which the output processing is reading data out
+again.  If the input arrives fast enough, it may "wrap around" the buffer to
+the point where the input is more than one whole scan ahead of the output.
+If the output processing simply proceeds through its display pass without
+paying attention to the input, the effect seen on-screen is that the lower
+part of the image is one or more scans better in quality than the upper part.
+Then, when the next output scan is started, you have a choice of what target
+scan number to use.  The recommended choice is to use the current input scan
+number at that time, which implies that you've skipped the output scans
+corresponding to the input scans that were completed while you processed the
+previous output scan.  In this way, the decoder automatically adapts its
+speed to the arriving data, by skipping output scans as necessary to keep up
+with the arriving data.
+
+When using this strategy, you'll want to be sure that you perform a final
+output pass after receiving all the data; otherwise your last display may not
+be full quality across the whole screen.  So the right outer loop logic is
+something like this:
+        do {
+            absorb any waiting input by calling jpeg_consume_input()
+            final_pass = jpeg_input_complete(&cinfo);
+            adjust output decompression parameters if required
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+        } while (! final_pass);
+rather than quitting as soon as jpeg_input_complete() returns TRUE.  This
+arrangement makes it simple to use higher-quality decoding parameters
+for the final pass.  But if you don't want to use special parameters for
+the final pass, the right loop logic is like this:
+        for (;;) {
+            absorb any waiting input by calling jpeg_consume_input()
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+            if (jpeg_input_complete(&cinfo) &&
+                cinfo.input_scan_number == cinfo.output_scan_number)
+              break;
+        }
+In this case you don't need to know in advance whether an output pass is to
+be the last one, so it's not necessary to have reached EOF before starting
+the final output pass; rather, what you want to test is whether the output
+pass was performed in sync with the final input scan.  This form of the loop
+will avoid an extra output pass whenever the decoder is able (or nearly able)
+to keep up with the incoming data.
+
+When the data transmission speed is high, you might begin a display pass,
+then find that much or all of the file has arrived before you can complete
+the pass.  (You can detect this by noting the JPEG_REACHED_EOI return code
+from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
+In this situation you may wish to abort the current display pass and start a
+new one using the newly arrived information.  To do so, just call
+jpeg_finish_output() and then start a new pass with jpeg_start_output().
+
+A variant strategy is to abort and restart display if more than one complete
+scan arrives during an output pass; this can be detected by noting
+JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number.  This
+idea should be employed with caution, however, since the display process
+might never get to the bottom of the image before being aborted, resulting
+in the lower part of the screen being several passes worse than the upper.
+In most cases it's probably best to abort an output pass only if the whole
+file has arrived and you want to begin the final output pass immediately.
+
+When receiving data across a communication link, we recommend always using
+the current input scan number for the output target scan number; if a
+higher-quality final pass is to be done, it should be started (aborting any
+incomplete output pass) as soon as the end of file is received.  However,
+many other strategies are possible.  For example, the application can examine
+the parameters of the current input scan and decide whether to display it or
+not.  If the scan contains only chroma data, one might choose not to use it
+as the target scan, expecting that the scan will be small and will arrive
+quickly.  To skip to the next scan, call jpeg_consume_input() until it
+returns JPEG_REACHED_SOS or JPEG_REACHED_EOI.  Or just use the next higher
+number as the target scan for jpeg_start_output(); but that method doesn't
+let you inspect the next scan's parameters before deciding to display it.
+
+
+In buffered-image mode, jpeg_start_decompress() never performs input and
+thus never suspends.  An application that uses input suspension with
+buffered-image mode must be prepared for suspension returns from these
+routines:
+* jpeg_start_output() performs input only if you request 2-pass quantization
+  and the target scan isn't fully read yet.  (This is discussed below.)
+* jpeg_read_scanlines(), as always, returns the number of scanlines that it
+  was able to produce before suspending.
+* jpeg_finish_output() will read any markers following the target scan,
+  up to the end of the file or the SOS marker that begins another scan.
+  (But it reads no input if jpeg_consume_input() has already reached the
+  end of the file or a SOS marker beyond the target output scan.)
+* jpeg_finish_decompress() will read until the end of file, and thus can
+  suspend if the end hasn't already been reached (as can be tested by
+  calling jpeg_input_complete()).
+jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
+all return TRUE if they completed their tasks, FALSE if they had to suspend.
+In the event of a FALSE return, the application must load more input data
+and repeat the call.  Applications that use non-suspending data sources need
+not check the return values of these three routines.
+
+
+It is possible to change decoding parameters between output passes in the
+buffered-image mode.  The decoder library currently supports only very
+limited changes of parameters.  ONLY THE FOLLOWING parameter changes are
+allowed after jpeg_start_decompress() is called:
+* dct_method can be changed before each call to jpeg_start_output().
+  For example, one could use a fast DCT method for early scans, changing
+  to a higher quality method for the final scan.
+* dither_mode can be changed before each call to jpeg_start_output();
+  of course this has no impact if not using color quantization.  Typically
+  one would use ordered dither for initial passes, then switch to
+  Floyd-Steinberg dither for the final pass.  Caution: changing dither mode
+  can cause more memory to be allocated by the library.  Although the amount
+  of memory involved is not large (a scanline or so), it may cause the
+  initial max_memory_to_use specification to be exceeded, which in the worst
+  case would result in an out-of-memory failure.
+* do_block_smoothing can be changed before each call to jpeg_start_output().
+  This setting is relevant only when decoding a progressive JPEG image.
+  During the first DC-only scan, block smoothing provides a very "fuzzy" look
+  instead of the very "blocky" look seen without it; which is better seems a
+  matter of personal taste.  But block smoothing is nearly always a win
+  during later stages, especially when decoding a successive-approximation
+  image: smoothing helps to hide the slight blockiness that otherwise shows
+  up on smooth gradients until the lowest coefficient bits are sent.
+* Color quantization mode can be changed under the rules described below.
+  You *cannot* change between full-color and quantized output (because that
+  would alter the required I/O buffer sizes), but you can change which
+  quantization method is used.
+
+When generating color-quantized output, changing quantization method is a
+very useful way of switching between high-speed and high-quality display.
+The library allows you to change among its three quantization methods:
+1. Single-pass quantization to a fixed color cube.
+   Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
+2. Single-pass quantization to an application-supplied colormap.
+   Selected by setting cinfo.colormap to point to the colormap (the value of
+   two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
+3. Two-pass quantization to a colormap chosen specifically for the image.
+   Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
+   (This is the default setting selected by jpeg_read_header, but it is
+   probably NOT what you want for the first pass of progressive display!)
+These methods offer successively better quality and lesser speed.  However,
+only the first method is available for quantizing in non-RGB color spaces.
+
+IMPORTANT: because the different quantizer methods have very different
+working-storage requirements, the library requires you to indicate which
+one(s) you intend to use before you call jpeg_start_decompress().  (If we did
+not require this, the max_memory_to_use setting would be a complete fiction.)
+You do this by setting one or more of these three cinfo fields to TRUE:
+        enable_1pass_quant              Fixed color cube colormap
+        enable_external_quant           Externally-supplied colormap
+        enable_2pass_quant              Two-pass custom colormap
+All three are initialized FALSE by jpeg_read_header().  But
+jpeg_start_decompress() automatically sets TRUE the one selected by the
+current two_pass_quantize and colormap settings, so you only need to set the
+enable flags for any other quantization methods you plan to change to later.
+
+After setting the enable flags correctly at jpeg_start_decompress() time, you
+can change to any enabled quantization method by setting two_pass_quantize
+and colormap properly just before calling jpeg_start_output().  The following
+special rules apply:
+1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
+   or 2-pass mode from a different mode, or when you want the 2-pass
+   quantizer to be re-run to generate a new colormap.
+2. To switch to an external colormap, or to change to a different external
+   colormap than was used on the prior pass, you must call
+   jpeg_new_colormap() after setting cinfo.colormap.
+NOTE: if you want to use the same colormap as was used in the prior pass,
+you should not do either of these things.  This will save some nontrivial
+switchover costs.
+(These requirements exist because cinfo.colormap will always be non-NULL
+after completing a prior output pass, since both the 1-pass and 2-pass
+quantizers set it to point to their output colormaps.  Thus you have to
+do one of these two things to notify the library that something has changed.
+Yup, it's a bit klugy, but it's necessary to do it this way for backwards
+compatibility.)
+
+Note that in buffered-image mode, the library generates any requested colormap
+during jpeg_start_output(), not during jpeg_start_decompress().
+
+When using two-pass quantization, jpeg_start_output() makes a pass over the
+buffered image to determine the optimum color map; it therefore may take a
+significant amount of time, whereas ordinarily it does little work.  The
+progress monitor hook is called during this pass, if defined.  It is also
+important to realize that if the specified target scan number is greater than
+or equal to the current input scan number, jpeg_start_output() will attempt
+to consume input as it makes this pass.  If you use a suspending data source,
+you need to check for a FALSE return from jpeg_start_output() under these
+conditions.  The combination of 2-pass quantization and a not-yet-fully-read
+target scan is the only case in which jpeg_start_output() will consume input.
+
+
+Application authors who support buffered-image mode may be tempted to use it
+for all JPEG images, even single-scan ones.  This will work, but it is
+inefficient: there is no need to create an image-sized coefficient buffer for
+single-scan images.  Requesting buffered-image mode for such an image wastes
+memory.  Worse, it can cost time on large images, since the buffered data has
+to be swapped out or written to a temporary file.  If you are concerned about
+maximum performance on baseline JPEG files, you should use buffered-image
+mode only when the incoming file actually has multiple scans.  This can be
+tested by calling jpeg_has_multiple_scans(), which will return a correct
+result at any time after jpeg_read_header() completes.
+
+It is also worth noting that when you use jpeg_consume_input() to let input
+processing get ahead of output processing, the resulting pattern of access to
+the coefficient buffer is quite nonsequential.  It's best to use the memory
+manager jmemnobs.c if you can (ie, if you have enough real or virtual main
+memory).  If not, at least make sure that max_memory_to_use is set as high as
+possible.  If the JPEG memory manager has to use a temporary file, you will
+probably see a lot of disk traffic and poor performance.  (This could be
+improved with additional work on the memory manager, but we haven't gotten
+around to it yet.)
+
+In some applications it may be convenient to use jpeg_consume_input() for all
+input processing, including reading the initial markers; that is, you may
+wish to call jpeg_consume_input() instead of jpeg_read_header() during
+startup.  This works, but note that you must check for JPEG_REACHED_SOS and
+JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
+Once the first SOS marker has been reached, you must call
+jpeg_start_decompress() before jpeg_consume_input() will consume more input;
+it'll just keep returning JPEG_REACHED_SOS until you do.  If you read a
+tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
+without ever returning JPEG_REACHED_SOS; be sure to check for this case.
+If this happens, the decompressor will not read any more input until you call
+jpeg_abort() to reset it.  It is OK to call jpeg_consume_input() even when not
+using buffered-image mode, but in that case it's basically a no-op after the
+initial markers have been read: it will just return JPEG_SUSPENDED.
+
+
+Abbreviated datastreams and multiple images
+-------------------------------------------
+
+A JPEG compression or decompression object can be reused to process multiple
+images.  This saves a small amount of time per image by eliminating the
+"create" and "destroy" operations, but that isn't the real purpose of the
+feature.  Rather, reuse of an object provides support for abbreviated JPEG
+datastreams.  Object reuse can also simplify processing a series of images in
+a single input or output file.  This section explains these features.
+
+A JPEG file normally contains several hundred bytes worth of quantization
+and Huffman tables.  In a situation where many images will be stored or
+transmitted with identical tables, this may represent an annoying overhead.
+The JPEG standard therefore permits tables to be omitted.  The standard
+defines three classes of JPEG datastreams:
+  * "Interchange" datastreams contain an image and all tables needed to decode
+     the image.  These are the usual kind of JPEG file.
+  * "Abbreviated image" datastreams contain an image, but are missing some or
+    all of the tables needed to decode that image.
+  * "Abbreviated table specification" (henceforth "tables-only") datastreams
+    contain only table specifications.
+To decode an abbreviated image, it is necessary to load the missing table(s)
+into the decoder beforehand.  This can be accomplished by reading a separate
+tables-only file.  A variant scheme uses a series of images in which the first
+image is an interchange (complete) datastream, while subsequent ones are
+abbreviated and rely on the tables loaded by the first image.  It is assumed
+that once the decoder has read a table, it will remember that table until a
+new definition for the same table number is encountered.
+
+It is the application designer's responsibility to figure out how to associate
+the correct tables with an abbreviated image.  While abbreviated datastreams
+can be useful in a closed environment, their use is strongly discouraged in
+any situation where data exchange with other applications might be needed.
+Caveat designer.
+
+The JPEG library provides support for reading and writing any combination of
+tables-only datastreams and abbreviated images.  In both compression and
+decompression objects, a quantization or Huffman table will be retained for
+the lifetime of the object, unless it is overwritten by a new table definition.
+
+
+To create abbreviated image datastreams, it is only necessary to tell the
+compressor not to emit some or all of the tables it is using.  Each
+quantization and Huffman table struct contains a boolean field "sent_table",
+which normally is initialized to FALSE.  For each table used by the image, the
+header-writing process emits the table and sets sent_table = TRUE unless it is
+already TRUE.  (In normal usage, this prevents outputting the same table
+definition multiple times, as would otherwise occur because the chroma
+components typically share tables.)  Thus, setting this field to TRUE before
+calling jpeg_start_compress() will prevent the table from being written at
+all.
+
+If you want to create a "pure" abbreviated image file containing no tables,
+just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
+tables.  If you want to emit some but not all tables, you'll need to set the
+individual sent_table fields directly.
+
+To create an abbreviated image, you must also call jpeg_start_compress()
+with a second parameter of FALSE, not TRUE.  Otherwise jpeg_start_compress()
+will force all the sent_table fields to FALSE.  (This is a safety feature to
+prevent abbreviated images from being created accidentally.)
+
+To create a tables-only file, perform the same parameter setup that you
+normally would, but instead of calling jpeg_start_compress() and so on, call
+jpeg_write_tables(&cinfo).  This will write an abbreviated datastream
+containing only SOI, DQT and/or DHT markers, and EOI.  All the quantization
+and Huffman tables that are currently defined in the compression object will
+be emitted unless their sent_tables flag is already TRUE, and then all the
+sent_tables flags will be set TRUE.
+
+A sure-fire way to create matching tables-only and abbreviated image files
+is to proceed as follows:
+
+        create JPEG compression object
+        set JPEG parameters
+        set destination to tables-only file
+        jpeg_write_tables(&cinfo);
+        set destination to image file
+        jpeg_start_compress(&cinfo, FALSE);
+        write data...
+        jpeg_finish_compress(&cinfo);
+
+Since the JPEG parameters are not altered between writing the table file and
+the abbreviated image file, the same tables are sure to be used.  Of course,
+you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
+many times to produce many abbreviated image files matching the table file.
+
+You cannot suppress output of the computed Huffman tables when Huffman
+optimization is selected.  (If you could, there'd be no way to decode the
+image...)  Generally, you don't want to set optimize_coding = TRUE when
+you are trying to produce abbreviated files.
+
+In some cases you might want to compress an image using tables which are
+not stored in the application, but are defined in an interchange or
+tables-only file readable by the application.  This can be done by setting up
+a JPEG decompression object to read the specification file, then copying the
+tables into your compression object.  See jpeg_copy_critical_parameters()
+for an example of copying quantization tables.
+
+
+To read abbreviated image files, you simply need to load the proper tables
+into the decompression object before trying to read the abbreviated image.
+If the proper tables are stored in the application program, you can just
+allocate the table structs and fill in their contents directly.  For example,
+to load a fixed quantization table into table slot "n":
+
+    if (cinfo.quant_tbl_ptrs[n] == NULL)
+      cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
+    quant_ptr = cinfo.quant_tbl_ptrs[n];        /* quant_ptr is JQUANT_TBL* */
+    for (i = 0; i < 64; i++) {
+      /* Qtable[] is desired quantization table, in natural array order */
+      quant_ptr->quantval[i] = Qtable[i];
+    }
+
+Code to load a fixed Huffman table is typically (for AC table "n"):
+
+    if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
+      cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
+    huff_ptr = cinfo.ac_huff_tbl_ptrs[n];       /* huff_ptr is JHUFF_TBL* */
+    for (i = 1; i <= 16; i++) {
+      /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
+      huff_ptr->bits[i] = counts[i];
+    }
+    for (i = 0; i < 256; i++) {
+      /* symbols[] is the list of Huffman symbols, in code-length order */
+      huff_ptr->huffval[i] = symbols[i];
+    }
+
+(Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
+constant JQUANT_TBL object is not safe.  If the incoming file happened to
+contain a quantization table definition, your master table would get
+overwritten!  Instead allocate a working table copy and copy the master table
+into it, as illustrated above.  Ditto for Huffman tables, of course.)
+
+You might want to read the tables from a tables-only file, rather than
+hard-wiring them into your application.  The jpeg_read_header() call is
+sufficient to read a tables-only file.  You must pass a second parameter of
+FALSE to indicate that you do not require an image to be present.  Thus, the
+typical scenario is
+
+        create JPEG decompression object
+        set source to tables-only file
+        jpeg_read_header(&cinfo, FALSE);
+        set source to abbreviated image file
+        jpeg_read_header(&cinfo, TRUE);
+        set decompression parameters
+        jpeg_start_decompress(&cinfo);
+        read data...
+        jpeg_finish_decompress(&cinfo);
+
+In some cases, you may want to read a file without knowing whether it contains
+an image or just tables.  In that case, pass FALSE and check the return value
+from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
+JPEG_HEADER_TABLES_ONLY if only tables were found.  (A third return value,
+JPEG_SUSPENDED, is possible when using a suspending data source manager.)
+Note that jpeg_read_header() will not complain if you read an abbreviated
+image for which you haven't loaded the missing tables; the missing-table check
+occurs later, in jpeg_start_decompress().
+
+
+It is possible to read a series of images from a single source file by
+repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
+without releasing/recreating the JPEG object or the data source module.
+(If you did reinitialize, any partial bufferload left in the data source
+buffer at the end of one image would be discarded, causing you to lose the
+start of the next image.)  When you use this method, stored tables are
+automatically carried forward, so some of the images can be abbreviated images
+that depend on tables from earlier images.
+
+If you intend to write a series of images into a single destination file,
+you might want to make a specialized data destination module that doesn't
+flush the output buffer at term_destination() time.  This would speed things
+up by some trifling amount.  Of course, you'd need to remember to flush the
+buffer after the last image.  You can make the later images be abbreviated
+ones by passing FALSE to jpeg_start_compress().
+
+
+Special markers
+---------------
+
+Some applications may need to insert or extract special data in the JPEG
+datastream.  The JPEG standard provides marker types "COM" (comment) and
+"APP0" through "APP15" (application) to hold application-specific data.
+Unfortunately, the use of these markers is not specified by the standard.
+COM markers are fairly widely used to hold user-supplied text.  The JFIF file
+format spec uses APP0 markers with specified initial strings to hold certain
+data.  Adobe applications use APP14 markers beginning with the string "Adobe"
+for miscellaneous data.  Other APPn markers are rarely seen, but might
+contain almost anything.
+
+If you wish to store user-supplied text, we recommend you use COM markers
+and place readable 7-bit ASCII text in them.  Newline conventions are not
+standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
+(Mac style).  A robust COM reader should be able to cope with random binary
+garbage, including nulls, since some applications generate COM markers
+containing non-ASCII junk.  (But yours should not be one of them.)
+
+For program-supplied data, use an APPn marker, and be sure to begin it with an
+identifying string so that you can tell whether the marker is actually yours.
+It's probably best to avoid using APP0 or APP14 for any private markers.
+(NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
+not use APP8 markers for any private purposes, either.)
+
+Keep in mind that at most 65533 bytes can be put into one marker, but you
+can have as many markers as you like.
+
+By default, the IJG compression library will write a JFIF APP0 marker if the
+selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
+the selected colorspace is RGB, CMYK, or YCCK.  You can disable this, but
+we don't recommend it.  The decompression library will recognize JFIF and
+Adobe markers and will set the JPEG colorspace properly when one is found.
+
+
+You can write special markers immediately following the datastream header by
+calling jpeg_write_marker() after jpeg_start_compress() and before the first
+call to jpeg_write_scanlines().  When you do this, the markers appear after
+the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
+all else.  Specify the marker type parameter as "JPEG_COM" for COM or
+"JPEG_APP0 + n" for APPn.  (Actually, jpeg_write_marker will let you write
+any marker type, but we don't recommend writing any other kinds of marker.)
+For example, to write a user comment string pointed to by comment_text:
+        jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
+
+If it's not convenient to store all the marker data in memory at once,
+you can instead call jpeg_write_m_header() followed by multiple calls to
+jpeg_write_m_byte().  If you do it this way, it's your responsibility to
+call jpeg_write_m_byte() exactly the number of times given in the length
+parameter to jpeg_write_m_header().  (This method lets you empty the
+output buffer partway through a marker, which might be important when
+using a suspending data destination module.  In any case, if you are using
+a suspending destination, you should flush its buffer after inserting
+any special markers.  See "I/O suspension".)
+
+Or, if you prefer to synthesize the marker byte sequence yourself,
+you can just cram it straight into the data destination module.
+
+If you are writing JFIF 1.02 extension markers (thumbnail images), don't
+forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
+correct JFIF version number in the JFIF header marker.  The library's default
+is to write version 1.01, but that's wrong if you insert any 1.02 extension
+markers.  (We could probably get away with just defaulting to 1.02, but there
+used to be broken decoders that would complain about unknown minor version
+numbers.  To reduce compatibility risks it's safest not to write 1.02 unless
+you are actually using 1.02 extensions.)
+
+
+When reading, two methods of handling special markers are available:
+1. You can ask the library to save the contents of COM and/or APPn markers
+into memory, and then examine them at your leisure afterwards.
+2. You can supply your own routine to process COM and/or APPn markers
+on-the-fly as they are read.
+The first method is simpler to use, especially if you are using a suspending
+data source; writing a marker processor that copes with input suspension is
+not easy (consider what happens if the marker is longer than your available
+input buffer).  However, the second method conserves memory since the marker
+data need not be kept around after it's been processed.
+
+For either method, you'd normally set up marker handling after creating a
+decompression object and before calling jpeg_read_header(), because the
+markers of interest will typically be near the head of the file and so will
+be scanned by jpeg_read_header.  Once you've established a marker handling
+method, it will be used for the life of that decompression object
+(potentially many datastreams), unless you change it.  Marker handling is
+determined separately for COM markers and for each APPn marker code.
+
+
+To save the contents of special markers in memory, call
+        jpeg_save_markers(cinfo, marker_code, length_limit)
+where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
+(To arrange to save all the special marker types, you need to call this
+routine 17 times, for COM and APP0-APP15.)  If the incoming marker is longer
+than length_limit data bytes, only length_limit bytes will be saved; this
+parameter allows you to avoid chewing up memory when you only need to see the
+first few bytes of a potentially large marker.  If you want to save all the
+data, set length_limit to 0xFFFF; that is enough since marker lengths are only
+16 bits.  As a special case, setting length_limit to 0 prevents that marker
+type from being saved at all.  (That is the default behavior, in fact.)
+
+After jpeg_read_header() completes, you can examine the special markers by
+following the cinfo->marker_list pointer chain.  All the special markers in
+the file appear in this list, in order of their occurrence in the file (but
+omitting any markers of types you didn't ask for).  Both the original data
+length and the saved data length are recorded for each list entry; the latter
+will not exceed length_limit for the particular marker type.  Note that these
+lengths exclude the marker length word, whereas the stored representation
+within the JPEG file includes it.  (Hence the maximum data length is really
+only 65533.)
+
+It is possible that additional special markers appear in the file beyond the
+SOS marker at which jpeg_read_header stops; if so, the marker list will be
+extended during reading of the rest of the file.  This is not expected to be
+common, however.  If you are short on memory you may want to reset the length
+limit to zero for all marker types after finishing jpeg_read_header, to
+ensure that the max_memory_to_use setting cannot be exceeded due to addition
+of later markers.
+
+The marker list remains stored until you call jpeg_finish_decompress or
+jpeg_abort, at which point the memory is freed and the list is set to empty.
+(jpeg_destroy also releases the storage, of course.)
+
+Note that the library is internally interested in APP0 and APP14 markers;
+if you try to set a small nonzero length limit on these types, the library
+will silently force the length up to the minimum it wants.  (But you can set
+a zero length limit to prevent them from being saved at all.)  Also, in a
+16-bit environment, the maximum length limit may be constrained to less than
+65533 by malloc() limitations.  It is therefore best not to assume that the
+effective length limit is exactly what you set it to be.
+
+
+If you want to supply your own marker-reading routine, you do it by calling
+jpeg_set_marker_processor().  A marker processor routine must have the
+signature
+        boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
+Although the marker code is not explicitly passed, the routine can find it
+in cinfo->unread_marker.  At the time of call, the marker proper has been
+read from the data source module.  The processor routine is responsible for
+reading the marker length word and the remaining parameter bytes, if any.
+Return TRUE to indicate success.  (FALSE should be returned only if you are
+using a suspending data source and it tells you to suspend.  See the standard
+marker processors in jdmarker.c for appropriate coding methods if you need to
+use a suspending data source.)
+
+If you override the default APP0 or APP14 processors, it is up to you to
+recognize JFIF and Adobe markers if you want colorspace recognition to occur
+properly.  We recommend copying and extending the default processors if you
+want to do that.  (A better idea is to save these marker types for later
+examination by calling jpeg_save_markers(); that method doesn't interfere
+with the library's own processing of these markers.)
+
+jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
+--- if you call one it overrides any previous call to the other, for the
+particular marker type specified.
+
+A simple example of an external COM processor can be found in djpeg.c.
+Also, see jpegtran.c for an example of using jpeg_save_markers.
+
+
+Raw (downsampled) image data
+----------------------------
+
+Some applications need to supply already-downsampled image data to the JPEG
+compressor, or to receive raw downsampled data from the decompressor.  The
+library supports this requirement by allowing the application to write or
+read raw data, bypassing the normal preprocessing or postprocessing steps.
+The interface is different from the standard one and is somewhat harder to
+use.  If your interest is merely in bypassing color conversion, we recommend
+that you use the standard interface and simply set jpeg_color_space =
+in_color_space (or jpeg_color_space = out_color_space for decompression).
+The mechanism described in this section is necessary only to supply or
+receive downsampled image data, in which not all components have the same
+dimensions.
+
+
+To compress raw data, you must supply the data in the colorspace to be used
+in the JPEG file (please read the earlier section on Special color spaces)
+and downsampled to the sampling factors specified in the JPEG parameters.
+You must supply the data in the format used internally by the JPEG library,
+namely a JSAMPIMAGE array.  This is an array of pointers to two-dimensional
+arrays, each of type JSAMPARRAY.  Each 2-D array holds the values for one
+color component.  This structure is necessary since the components are of
+different sizes.  If the image dimensions are not a multiple of the MCU size,
+you must also pad the data correctly (usually, this is done by replicating
+the last column and/or row).  The data must be padded to a multiple of a DCT
+block in each component: that is, each downsampled row must contain a
+multiple of 8 valid samples, and there must be a multiple of 8 sample rows
+for each component.  (For applications such as conversion of digital TV
+images, the standard image size is usually a multiple of the DCT block size,
+so that no padding need actually be done.)
+
+The procedure for compression of raw data is basically the same as normal
+compression, except that you call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  Before calling jpeg_start_compress(), you must do
+the following:
+  * Set cinfo->raw_data_in to TRUE.  (It is set FALSE by jpeg_set_defaults().)
+    This notifies the library that you will be supplying raw data.
+  * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
+    call is a good idea.  Note that since color conversion is bypassed,
+    in_color_space is ignored, except that jpeg_set_defaults() uses it to
+    choose the default jpeg_color_space setting.
+  * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
+    cinfo->comp_info[i].v_samp_factor, are correct.  Since these indicate the
+    dimensions of the data you are supplying, it's wise to set them
+    explicitly, rather than assuming the library's defaults are what you want.
+
+To pass raw data to the library, call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  The two routines work similarly except that
+jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
+The scanlines count passed to and returned from jpeg_write_raw_data is
+measured in terms of the component with the largest v_samp_factor.
+
+jpeg_write_raw_data() processes one MCU row per call, which is to say
+v_samp_factor*DCTSIZE sample rows of each component.  The passed num_lines
+value must be at least max_v_samp_factor*DCTSIZE, and the return value will
+be exactly that amount (or possibly some multiple of that amount, in future
+library versions).  This is true even on the last call at the bottom of the
+image; don't forget to pad your data as necessary.
+
+The required dimensions of the supplied data can be computed for each
+component as
+        cinfo->comp_info[i].width_in_blocks*DCTSIZE  samples per row
+        cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
+after jpeg_start_compress() has initialized those fields.  If the valid data
+is smaller than this, it must be padded appropriately.  For some sampling
+factors and image sizes, additional dummy DCT blocks are inserted to make
+the image a multiple of the MCU dimensions.  The library creates such dummy
+blocks itself; it does not read them from your supplied data.  Therefore you
+need never pad by more than DCTSIZE samples.  An example may help here.
+Assume 2h2v downsampling of YCbCr data, that is
+        cinfo->comp_info[0].h_samp_factor = 2           for Y
+        cinfo->comp_info[0].v_samp_factor = 2
+        cinfo->comp_info[1].h_samp_factor = 1           for Cb
+        cinfo->comp_info[1].v_samp_factor = 1
+        cinfo->comp_info[2].h_samp_factor = 1           for Cr
+        cinfo->comp_info[2].v_samp_factor = 1
+and suppose that the nominal image dimensions (cinfo->image_width and
+cinfo->image_height) are 101x101 pixels.  Then jpeg_start_compress() will
+compute downsampled_width = 101 and width_in_blocks = 13 for Y,
+downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
+for the height fields).  You must pad the Y data to at least 13*8 = 104
+columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows.  The
+MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
+scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
+sample rows of Y and 8 each of Cb and Cr.  A total of 7 MCU rows are needed,
+so you must pass a total of 7*16 = 112 "scanlines".  The last DCT block row
+of Y data is dummy, so it doesn't matter what you pass for it in the data
+arrays, but the scanlines count must total up to 112 so that all of the Cb
+and Cr data gets passed.
+
+Output suspension is supported with raw-data compression: if the data
+destination module suspends, jpeg_write_raw_data() will return 0.
+In this case the same data rows must be passed again on the next call.
+
+
+Decompression with raw data output implies bypassing all postprocessing:
+you cannot ask for rescaling or color quantization, for instance.  More
+seriously, you must deal with the color space and sampling factors present in
+the incoming file.  If your application only handles, say, 2h1v YCbCr data,
+you must check for and fail on other color spaces or other sampling factors.
+The library will not convert to a different color space for you.
+
+To obtain raw data output, set cinfo->raw_data_out = TRUE before
+jpeg_start_decompress() (it is set FALSE by jpeg_read_header()).  Be sure to
+verify that the color space and sampling factors are ones you can handle.
+Then call jpeg_read_raw_data() in place of jpeg_read_scanlines().  The
+decompression process is otherwise the same as usual.
+
+jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
+buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
+the same as for raw-data compression).  The buffer you pass must be large
+enough to hold the actual data plus padding to DCT-block boundaries.  As with
+compression, any entirely dummy DCT blocks are not processed so you need not
+allocate space for them, but the total scanline count includes them.  The
+above example of computing buffer dimensions for raw-data compression is
+equally valid for decompression.
+
+Input suspension is supported with raw-data decompression: if the data source
+module suspends, jpeg_read_raw_data() will return 0.  You can also use
+buffered-image mode to read raw data in multiple passes.
+
+
+Really raw data: DCT coefficients
+---------------------------------
+
+It is possible to read or write the contents of a JPEG file as raw DCT
+coefficients.  This facility is mainly intended for use in lossless
+transcoding between different JPEG file formats.  Other possible applications
+include lossless cropping of a JPEG image, lossless reassembly of a
+multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
+
+To read the contents of a JPEG file as DCT coefficients, open the file and do
+jpeg_read_header() as usual.  But instead of calling jpeg_start_decompress()
+and jpeg_read_scanlines(), call jpeg_read_coefficients().  This will read the
+entire image into a set of virtual coefficient-block arrays, one array per
+component.  The return value is a pointer to an array of virtual-array
+descriptors.  Each virtual array can be accessed directly using the JPEG
+memory manager's access_virt_barray method (see Memory management, below,
+and also read structure.txt's discussion of virtual array handling).  Or,
+for simple transcoding to a different JPEG file format, the array list can
+just be handed directly to jpeg_write_coefficients().
+
+Each block in the block arrays contains quantized coefficient values in
+normal array order (not JPEG zigzag order).  The block arrays contain only
+DCT blocks containing real data; any entirely-dummy blocks added to fill out
+interleaved MCUs at the right or bottom edges of the image are discarded
+during reading and are not stored in the block arrays.  (The size of each
+block array can be determined from the width_in_blocks and height_in_blocks
+fields of the component's comp_info entry.)  This is also the data format
+expected by jpeg_write_coefficients().
+
+When you are done using the virtual arrays, call jpeg_finish_decompress()
+to release the array storage and return the decompression object to an idle
+state; or just call jpeg_destroy() if you don't need to reuse the object.
+
+If you use a suspending data source, jpeg_read_coefficients() will return
+NULL if it is forced to suspend; a non-NULL return value indicates successful
+completion.  You need not test for a NULL return value when using a
+non-suspending data source.
+
+It is also possible to call jpeg_read_coefficients() to obtain access to the
+decoder's coefficient arrays during a normal decode cycle in buffered-image
+mode.  This frammish might be useful for progressively displaying an incoming
+image and then re-encoding it without loss.  To do this, decode in buffered-
+image mode as discussed previously, then call jpeg_read_coefficients() after
+the last jpeg_finish_output() call.  The arrays will be available for your use
+until you call jpeg_finish_decompress().
+
+
+To write the contents of a JPEG file as DCT coefficients, you must provide
+the DCT coefficients stored in virtual block arrays.  You can either pass
+block arrays read from an input JPEG file by jpeg_read_coefficients(), or
+allocate virtual arrays from the JPEG compression object and fill them
+yourself.  In either case, jpeg_write_coefficients() is substituted for
+jpeg_start_compress() and jpeg_write_scanlines().  Thus the sequence is
+  * Create compression object
+  * Set all compression parameters as necessary
+  * Request virtual arrays if needed
+  * jpeg_write_coefficients()
+  * jpeg_finish_compress()
+  * Destroy or re-use compression object
+jpeg_write_coefficients() is passed a pointer to an array of virtual block
+array descriptors; the number of arrays is equal to cinfo.num_components.
+
+The virtual arrays need only have been requested, not realized, before
+jpeg_write_coefficients() is called.  A side-effect of
+jpeg_write_coefficients() is to realize any virtual arrays that have been
+requested from the compression object's memory manager.  Thus, when obtaining
+the virtual arrays from the compression object, you should fill the arrays
+after calling jpeg_write_coefficients().  The data is actually written out
+when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
+the file header.
+
+When writing raw DCT coefficients, it is crucial that the JPEG quantization
+tables and sampling factors match the way the data was encoded, or the
+resulting file will be invalid.  For transcoding from an existing JPEG file,
+we recommend using jpeg_copy_critical_parameters().  This routine initializes
+all the compression parameters to default values (like jpeg_set_defaults()),
+then copies the critical information from a source decompression object.
+The decompression object should have just been used to read the entire
+JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
+
+jpeg_write_coefficients() marks all tables stored in the compression object
+as needing to be written to the output file (thus, it acts like
+jpeg_start_compress(cinfo, TRUE)).  This is for safety's sake, to avoid
+emitting abbreviated JPEG files by accident.  If you really want to emit an
+abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
+individual sent_table flags, between calling jpeg_write_coefficients() and
+jpeg_finish_compress().
+
+
+Progress monitoring
+-------------------
+
+Some applications may need to regain control from the JPEG library every so
+often.  The typical use of this feature is to produce a percent-done bar or
+other progress display.  (For a simple example, see cjpeg.c or djpeg.c.)
+Although you do get control back frequently during the data-transferring pass
+(the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
+will occur inside jpeg_finish_compress or jpeg_start_decompress; those
+routines may take a long time to execute, and you don't get control back
+until they are done.
+
+You can define a progress-monitor routine which will be called periodically
+by the library.  No guarantees are made about how often this call will occur,
+so we don't recommend you use it for mouse tracking or anything like that.
+At present, a call will occur once per MCU row, scanline, or sample row
+group, whichever unit is convenient for the current processing mode; so the
+wider the image, the longer the time between calls.  During the data
+transferring pass, only one call occurs per call of jpeg_read_scanlines or
+jpeg_write_scanlines, so don't pass a large number of scanlines at once if
+you want fine resolution in the progress count.  (If you really need to use
+the callback mechanism for time-critical tasks like mouse tracking, you could
+insert additional calls inside some of the library's inner loops.)
+
+To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
+fill in its progress_monitor field with a pointer to your callback routine,
+and set cinfo->progress to point to the struct.  The callback will be called
+whenever cinfo->progress is non-NULL.  (This pointer is set to NULL by
+jpeg_create_compress or jpeg_create_decompress; the library will not change
+it thereafter.  So if you allocate dynamic storage for the progress struct,
+make sure it will live as long as the JPEG object does.  Allocating from the
+JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.)  You
+can use the same callback routine for both compression and decompression.
+
+The jpeg_progress_mgr struct contains four fields which are set by the library:
+        long pass_counter;      /* work units completed in this pass */
+        long pass_limit;        /* total number of work units in this pass */
+        int completed_passes;   /* passes completed so far */
+        int total_passes;       /* total number of passes expected */
+During any one pass, pass_counter increases from 0 up to (not including)
+pass_limit; the step size is usually but not necessarily 1.  The pass_limit
+value may change from one pass to another.  The expected total number of
+passes is in total_passes, and the number of passes already completed is in
+completed_passes.  Thus the fraction of work completed may be estimated as
+                completed_passes + (pass_counter/pass_limit)
+                --------------------------------------------
+                                total_passes
+ignoring the fact that the passes may not be equal amounts of work.
+
+When decompressing, pass_limit can even change within a pass, because it
+depends on the number of scans in the JPEG file, which isn't always known in
+advance.  The computed fraction-of-work-done may jump suddenly (if the library
+discovers it has overestimated the number of scans) or even decrease (in the
+opposite case).  It is not wise to put great faith in the work estimate.
+
+When using the decompressor's buffered-image mode, the progress monitor work
+estimate is likely to be completely unhelpful, because the library has no way
+to know how many output passes will be demanded of it.  Currently, the library
+sets total_passes based on the assumption that there will be one more output
+pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
+TRUE), but no more output passes if the file end has been reached when the
+output pass is started.  This means that total_passes will rise as additional
+output passes are requested.  If you have a way of determining the input file
+size, estimating progress based on the fraction of the file that's been read
+will probably be more useful than using the library's value.
+
+
+Memory management
+-----------------
+
+This section covers some key facts about the JPEG library's built-in memory
+manager.  For more info, please read structure.txt's section about the memory
+manager, and consult the source code if necessary.
+
+All memory and temporary file allocation within the library is done via the
+memory manager.  If necessary, you can replace the "back end" of the memory
+manager to control allocation yourself (for example, if you don't want the
+library to use malloc() and free() for some reason).
+
+Some data is allocated "permanently" and will not be freed until the JPEG
+object is destroyed.  Most data is allocated "per image" and is freed by
+jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort.  You can call the
+memory manager yourself to allocate structures that will automatically be
+freed at these times.  Typical code for this is
+  ptr = (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, size);
+Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
+Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
+There are also alloc_sarray and alloc_barray routines that automatically
+build 2-D sample or block arrays.
+
+The library's minimum space requirements to process an image depend on the
+image's width, but not on its height, because the library ordinarily works
+with "strip" buffers that are as wide as the image but just a few rows high.
+Some operating modes (eg, two-pass color quantization) require full-image
+buffers.  Such buffers are treated as "virtual arrays": only the current strip
+need be in memory, and the rest can be swapped out to a temporary file.
+
+If you use the simplest memory manager back end (jmemnobs.c), then no
+temporary files are used; virtual arrays are simply malloc()'d.  Images bigger
+than memory can be processed only if your system supports virtual memory.
+The other memory manager back ends support temporary files of various flavors
+and thus work in machines without virtual memory.  They may also be useful on
+Unix machines if you need to process images that exceed available swap space.
+
+When using temporary files, the library will make the in-memory buffers for
+its virtual arrays just big enough to stay within a "maximum memory" setting.
+Your application can set this limit by setting cinfo->mem->max_memory_to_use
+after creating the JPEG object.  (Of course, there is still a minimum size for
+the buffers, so the max-memory setting is effective only if it is bigger than
+the minimum space needed.)  If you allocate any large structures yourself, you
+must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
+order to have them counted against the max memory limit.  Also keep in mind
+that space allocated with alloc_small() is ignored, on the assumption that
+it's too small to be worth worrying about; so a reasonable safety margin
+should be left when setting max_memory_to_use.
+
+
+Memory usage
+------------
+
+Working memory requirements while performing compression or decompression
+depend on image dimensions, image characteristics (such as colorspace and
+JPEG process), and operating mode (application-selected options).
+
+As of v6b, the decompressor requires:
+ 1. About 24K in more-or-less-fixed-size data.  This varies a bit depending
+    on operating mode and image characteristics (particularly color vs.
+    grayscale), but it doesn't depend on image dimensions.
+ 2. Strip buffers (of size proportional to the image width) for IDCT and
+    upsampling results.  The worst case for commonly used sampling factors
+    is about 34 bytes * width in pixels for a color image.  A grayscale image
+    only needs about 8 bytes per pixel column.
+ 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
+    file (including progressive JPEGs), or whenever you select buffered-image
+    mode.  This takes 2 bytes/coefficient.  At typical 2x2 sampling, that's
+    3 bytes per pixel for a color image.  Worst case (1x1 sampling) requires
+    6 bytes/pixel.  For grayscale, figure 2 bytes/pixel.
+ 4. To perform 2-pass color quantization, the decompressor also needs a
+    128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
+This does not count any memory allocated by the application, such as a
+buffer to hold the final output image.
+
+The above figures are valid for 8-bit JPEG data precision and a machine with
+32-bit ints.  For 12-bit JPEG data, double the size of the strip buffers and
+quantization pixel buffer.  The "fixed-size" data will be somewhat smaller
+with 16-bit ints, larger with 64-bit ints.  Also, CMYK or other unusual
+color spaces will require different amounts of space.
+
+The full-image coefficient and pixel buffers, if needed at all, do not
+have to be fully RAM resident; you can have the library use temporary
+files instead when the total memory usage would exceed a limit you set.
+(But if your OS supports virtual memory, it's probably better to just use
+jmemnobs and let the OS do the swapping.)
+
+The compressor's memory requirements are similar, except that it has no need
+for color quantization.  Also, it needs a full-image DCT coefficient buffer
+if Huffman-table optimization is asked for, even if progressive mode is not
+requested.
+
+If you need more detailed information about memory usage in a particular
+situation, you can enable the MEM_STATS code in jmemmgr.c.
+
+
+Library compile-time options
+----------------------------
+
+A number of compile-time options are available by modifying jmorecfg.h.
+
+The JPEG standard provides for both the baseline 8-bit DCT process and
+a 12-bit DCT process.  The IJG code supports 12-bit lossy JPEG if you define
+BITS_IN_JSAMPLE as 12 rather than 8.  Note that this causes JSAMPLE to be
+larger than a char, so it affects the surrounding application's image data.
+The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
+and GIF file formats; you must disable the other file formats to compile a
+12-bit cjpeg or djpeg.  (install.txt has more information about that.)
+At present, a 12-bit library can handle *only* 12-bit images, not both
+precisions.
+
+Note that a 12-bit library always compresses in Huffman optimization mode,
+in order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.  If you need to output 12-bit
+files in one pass, you'll have to supply suitable default Huffman tables.
+You may also want to supply your own DCT quantization tables; the existing
+quality-scaling code has been developed for 8-bit use, and probably doesn't
+generate especially good tables for 12-bit.
+
+The maximum number of components (color channels) in the image is determined
+by MAX_COMPONENTS.  The JPEG standard allows up to 255 components, but we
+expect that few applications will need more than four or so.
+
+On machines with unusual data type sizes, you may be able to improve
+performance or reduce memory space by tweaking the various typedefs in
+jmorecfg.h.  In particular, on some RISC CPUs, access to arrays of "short"s
+is quite slow; consider trading memory for speed by making JCOEF, INT16, and
+UINT16 be "int" or "unsigned int".  UINT8 is also a candidate to become int.
+You probably don't want to make JSAMPLE be int unless you have lots of memory
+to burn.
+
+You can reduce the size of the library by compiling out various optional
+functions.  To do this, undefine xxx_SUPPORTED symbols as necessary.
+
+You can also save a few K by not having text error messages in the library;
+the standard error message table occupies about 5Kb.  This is particularly
+reasonable for embedded applications where there's no good way to display
+a message anyway.  To do this, remove the creation of the message table
+(jpeg_std_message_table[]) from jerror.c, and alter format_message to do
+something reasonable without it.  You could output the numeric value of the
+message code number, for example.  If you do this, you can also save a couple
+more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
+you don't need trace capability anyway, right?
+
+
+Portability considerations
+--------------------------
+
+The JPEG library has been written to be extremely portable; the sample
+applications cjpeg and djpeg are slightly less so.  This section summarizes
+the design goals in this area.  (If you encounter any bugs that cause the
+library to be less portable than is claimed here, we'd appreciate hearing
+about them.)
+
+The code works fine on ANSI C and C++ compilers, using any of the popular
+system include file setups, and some not-so-popular ones too.
+
+The code is not dependent on the exact sizes of the C data types.  As
+distributed, we make the assumptions that
+        char    is at least 8 bits wide
+        short   is at least 16 bits wide
+        int     is at least 16 bits wide
+        long    is at least 32 bits wide
+(These are the minimum requirements of the ANSI C standard.)  Wider types will
+work fine, although memory may be used inefficiently if char is much larger
+than 8 bits or short is much bigger than 16 bits.  The code should work
+equally well with 16- or 32-bit ints.
+
+In a system where these assumptions are not met, you may be able to make the
+code work by modifying the typedefs in jmorecfg.h.  However, you will probably
+have difficulty if int is less than 16 bits wide, since references to plain
+int abound in the code.
+
+char can be either signed or unsigned, although the code runs faster if an
+unsigned char type is available.  If char is wider than 8 bits, you will need
+to redefine JOCTET and/or provide custom data source/destination managers so
+that JOCTET represents exactly 8 bits of data on external storage.
+
+The JPEG library proper does not assume ASCII representation of characters.
+But some of the image file I/O modules in cjpeg/djpeg do have ASCII
+dependencies in file-header manipulation; so does cjpeg's select_file_type()
+routine.
+
+The JPEG library does not rely heavily on the C library.  In particular, C
+stdio is used only by the data source/destination modules and the error
+handler, all of which are application-replaceable.  (cjpeg/djpeg are more
+heavily dependent on stdio.)  malloc and free are called only from the memory
+manager "back end" module, so you can use a different memory allocator by
+replacing that one file.
+
+More info about porting the code may be gleaned by reading jconfig.txt,
+jmorecfg.h, and jinclude.h.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/LICENSE.md glmark2-2021.02/android/jni/src/libjpeg-turbo/LICENSE.md
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/LICENSE.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/LICENSE.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,88 @@
+libjpeg-turbo Licenses
+======================
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses:
+
+- The IJG (Independent JPEG Group) License, which is listed in
+  [README.ijg](README.ijg)
+
+  This license applies to the libjpeg API library and associated programs
+  (any code inherited from libjpeg, and any modifications to that code.)
+
+- The Modified (3-clause) BSD License, which is listed in
+  [turbojpeg.c](turbojpeg.c)
+
+  This license covers the TurboJPEG API library and associated programs.
+
+- The zlib License, which is listed in [simd/jsimdext.inc](simd/jsimdext.inc)
+
+  This license is a subset of the other two, and it covers the libjpeg-turbo
+  SIMD extensions.
+
+
+Complying with the libjpeg-turbo Licenses
+=========================================
+
+This section provides a roll-up of the libjpeg-turbo licensing terms, to the
+best of our understanding.
+
+1.  If you are distributing a modified version of the libjpeg-turbo source,
+    then:
+
+    1.  You cannot alter or remove any existing copyright or license notices
+        from the source.
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 1 of the Modified BSD License
+        - Clauses 1 and 3 of the zlib License
+
+    2.  You must add your own copyright notice to the header of each source
+        file you modified, so others can tell that you modified that file (if
+        there is not an existing copyright header in that file, then you can
+        simply add a notice stating that you modified the file.)
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 2 of the zlib License
+
+    3.  You must include the IJG README file, and you must not alter any of the
+        copyright or license text in that file.
+
+        **Origin**
+        - Clause 1 of the IJG License
+
+2.  If you are distributing only libjpeg-turbo binaries without the source, or
+    if you are distributing an application that statically links with
+    libjpeg-turbo, then:
+
+    1.  Your product documentation must include a message stating:
+
+        This software is based in part on the work of the Independent JPEG
+        Group.
+
+        **Origin**
+        - Clause 2 of the IJG license
+
+    2.  If your binary distribution includes or uses the TurboJPEG API, then
+        your product documentation must include the text of the Modified BSD
+        License.
+
+        **Origin**
+        - Clause 2 of the Modified BSD License
+
+3.  You cannot use the name of the IJG or The libjpeg-turbo Project or the
+    contributors thereof in advertising, publicity, etc.
+
+    **Origin**
+    - IJG License
+    - Clause 3 of the Modified BSD License
+
+4.  The IJG and The libjpeg-turbo Project do not warrant libjpeg-turbo to be
+    free of defects, nor do we accept any liability for undesirable
+    consequences resulting from your use of the software.
+
+    **Origin**
+    - IJG License
+    - Modified BSD License
+    - zlib License
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/Makefile.am glmark2-2021.02/android/jni/src/libjpeg-turbo/Makefile.am
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/Makefile.am	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/Makefile.am	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,778 @@
+lib_LTLIBRARIES = libjpeg.la
+libjpeg_la_LDFLAGS = -version-info ${LIBTOOL_CURRENT}:${SO_MINOR_VERSION}:${SO_AGE} -no-undefined
+include_HEADERS = jerror.h jmorecfg.h jpeglib.h
+
+if WITH_TURBOJPEG
+lib_LTLIBRARIES += libturbojpeg.la
+libturbojpeg_la_LDFLAGS = -version-info 1:0:1 -no-undefined
+include_HEADERS += turbojpeg.h
+endif
+
+nodist_include_HEADERS = jconfig.h
+
+pkgconfigdir = $(libdir)/pkgconfig
+pkgconfig_DATA = pkgscripts/libjpeg.pc pkgscripts/libturbojpeg.pc
+
+HDRS = jchuff.h jdct.h jdhuff.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+	jpegint.h jpeglib.h jversion.h jsimd.h jsimddct.h jpegcomp.h \
+	jpeg_nbits_table.h
+
+libjpeg_la_SOURCES = $(HDRS) jcapimin.c jcapistd.c jccoefct.c jccolor.c \
+	jcdctmgr.c jchuff.c jcinit.c jcmainct.c jcmarker.c jcmaster.c \
+	jcomapi.c jcparam.c jcphuff.c jcprepct.c jcsample.c jctrans.c \
+	jdapimin.c jdapistd.c jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c \
+	jddctmgr.c jdhuff.c jdinput.c jdmainct.c jdmarker.c jdmaster.c \
+	jdmerge.c jdphuff.c jdpostct.c jdsample.c jdtrans.c jerror.c \
+	jfdctflt.c jfdctfst.c jfdctint.c jidctflt.c jidctfst.c jidctint.c \
+	jidctred.c jquant1.c jquant2.c jutils.c jmemmgr.c jmemnobs.c
+
+if WITH_ARITH
+libjpeg_la_SOURCES += jaricom.c
+endif
+
+if WITH_ARITH_ENC
+libjpeg_la_SOURCES += jcarith.c
+endif
+
+if WITH_ARITH_DEC
+libjpeg_la_SOURCES += jdarith.c
+endif
+
+
+SUBDIRS = java
+
+
+if WITH_TURBOJPEG
+
+libturbojpeg_la_SOURCES = $(libjpeg_la_SOURCES) turbojpeg.c turbojpeg.h \
+	transupp.c transupp.h jdatadst-tj.c jdatasrc-tj.c
+
+if WITH_JAVA
+
+libturbojpeg_la_SOURCES += turbojpeg-jni.c
+libturbojpeg_la_CFLAGS = ${JNI_CFLAGS}
+TJMAPFILE = turbojpeg-mapfile.jni
+
+else
+
+TJMAPFILE = turbojpeg-mapfile
+
+endif
+
+libturbojpeg_la_SOURCES += $(TJMAPFILE)
+
+if VERSION_SCRIPT
+libturbojpeg_la_LDFLAGS += $(VERSION_SCRIPT_FLAG)$(srcdir)/$(TJMAPFILE)
+endif
+
+endif
+
+
+if VERSION_SCRIPT
+libjpeg_la_LDFLAGS += $(VERSION_SCRIPT_FLAG)libjpeg.map
+endif
+
+
+if WITH_SIMD
+
+SUBDIRS += simd
+libjpeg_la_LIBADD = simd/libsimd.la
+libturbojpeg_la_LIBADD = simd/libsimd.la
+
+else
+
+libjpeg_la_SOURCES += jsimd_none.c
+
+endif
+
+
+bin_PROGRAMS = cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+noinst_PROGRAMS = jcstest
+
+
+if WITH_TURBOJPEG
+
+bin_PROGRAMS += tjbench
+
+noinst_PROGRAMS += tjunittest
+
+tjbench_SOURCES = tjbench.c bmp.h bmp.c tjutil.h tjutil.c rdbmp.c rdppm.c \
+	wrbmp.c wrppm.c
+
+tjbench_LDADD = libturbojpeg.la libjpeg.la -lm
+
+tjbench_CFLAGS = -DBMP_SUPPORTED -DPPM_SUPPORTED
+
+tjunittest_SOURCES = tjunittest.c tjutil.h tjutil.c
+
+tjunittest_LDADD = libturbojpeg.la
+
+endif
+
+
+cjpeg_SOURCES = cdjpeg.h cderror.h cdjpeg.c cjpeg.c rdgif.c rdppm.c rdswitch.c
+if WITH_12BIT
+else
+cjpeg_SOURCES += rdbmp.c rdtarga.c
+endif
+
+cjpeg_LDADD = libjpeg.la
+
+cjpeg_CFLAGS = -DGIF_SUPPORTED -DPPM_SUPPORTED
+if WITH_12BIT
+else
+cjpeg_CFLAGS += -DBMP_SUPPORTED -DTARGA_SUPPORTED
+endif
+
+djpeg_SOURCES = cdjpeg.h cderror.h cdjpeg.c djpeg.c rdcolmap.c rdswitch.c \
+	wrgif.c wrppm.c
+if WITH_12BIT
+else
+djpeg_SOURCES += wrbmp.c wrtarga.c
+endif
+
+djpeg_LDADD = libjpeg.la
+
+djpeg_CFLAGS = -DGIF_SUPPORTED -DPPM_SUPPORTED
+if WITH_12BIT
+else
+djpeg_CFLAGS += -DBMP_SUPPORTED -DTARGA_SUPPORTED
+endif
+
+jpegtran_SOURCES = jpegtran.c rdswitch.c cdjpeg.c transupp.c transupp.h
+
+jpegtran_LDADD = libjpeg.la
+
+rdjpgcom_SOURCES = rdjpgcom.c
+
+rdjpgcom_LDADD = libjpeg.la
+
+wrjpgcom_SOURCES = wrjpgcom.c
+
+wrjpgcom_LDADD = libjpeg.la
+
+jcstest_SOURCES = jcstest.c
+
+jcstest_LDADD = libjpeg.la
+
+dist_man1_MANS = cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 wrjpgcom.1
+
+DOCS= coderules.txt jconfig.txt change.log rdrle.c wrrle.c BUILDING.md \
+	ChangeLog.txt
+
+dist_doc_DATA = README.ijg README.md libjpeg.txt structure.txt usage.txt \
+	wizard.txt LICENSE.md
+
+exampledir = $(docdir)
+dist_example_DATA = example.c
+
+
+EXTRA_DIST = win release $(DOCS) testimages CMakeLists.txt \
+	sharedlib/CMakeLists.txt cmakescripts libjpeg.map.in doc doxygen.config \
+	doxygen-extra.css jccolext.c jdcolext.c jdcol565.c jdmrgext.c jdmrg565.c \
+	jstdhuff.c jdcoefct.h jdmainct.h jdmaster.h jdsample.h wrppm.h \
+	md5/CMakeLists.txt
+
+dist-hook:
+	rm -rf `find $(distdir) -name .svn`
+
+
+SUBDIRS += md5
+
+if WITH_12BIT
+
+TESTORIG = testorig12.jpg
+MD5_JPEG_RGB_ISLOW = 9620f424569594bb9242b48498ad801f
+MD5_PPM_RGB_ISLOW = f3301d2219783b8b3d942b7239fa50c0
+MD5_JPEG_422_IFAST_OPT = 7322e3bd2f127f7de4b40d4480ce60e4
+MD5_PPM_422_IFAST = 79807fa552899e66a04708f533e16950
+MD5_PPM_422M_IFAST = 07737bfe8a7c1c87aaa393a0098d16b0
+MD5_JPEG_420_IFAST_Q100_PROG = a1da220b5604081863a504297ed59e55
+MD5_PPM_420_Q100_IFAST = 1b3730122709f53d007255e8dfd3305e
+MD5_PPM_420M_Q100_IFAST = 980a1a3c5bf9510022869d30b7d26566
+MD5_JPEG_GRAY_ISLOW = 235c90707b16e2e069f37c888b2636d9
+MD5_PPM_GRAY_ISLOW = 7213c10af507ad467da5578ca5ee1fca
+MD5_PPM_GRAY_ISLOW_RGB = e96ee81c30a6ed422d466338bd3de65d
+MD5_JPEG_420S_IFAST_OPT = 7af8e60be4d9c227ec63ac9b6630855e
+MD5_JPEG_3x2_FLOAT_PROG_SSE = a8c17daf77b457725ec929e215b603f8
+MD5_PPM_3x2_FLOAT_SSE = 42876ab9e5c2f76a87d08db5fbd57956
+MD5_JPEG_3x2_FLOAT_PROG_32BIT = a8c17daf77b457725ec929e215b603f8
+MD5_PPM_3x2_FLOAT_32BIT = 42876ab9e5c2f76a87d08db5fbd57956
+MD5_PPM_3x2_FLOAT_64BIT = d6fbc71153b3d8ded484dbc17c7b9cf4
+MD5_JPEG_3x2_IFAST_PROG = 1396cc2b7185cfe943d408c9d305339e
+MD5_PPM_3x2_IFAST = 3975985ef6eeb0a2cdc58daa651ccc00
+MD5_PPM_420M_ISLOW_2_1 = 4ca6be2a6f326ff9eaab63e70a8259c0
+MD5_PPM_420M_ISLOW_15_8 = 12aa9f9534c1b3d7ba047322226365eb
+MD5_PPM_420M_ISLOW_13_8 = f7e22817c7b25e1393e4ec101e9d4e96
+MD5_PPM_420M_ISLOW_11_8 = 800a16f9f4dc9b293197bfe11be10a82
+MD5_PPM_420M_ISLOW_9_8 = 06b7a92a9bc69f4dc36ec40f1937d55c
+MD5_PPM_420M_ISLOW_7_8 = 3ec444a14a4ab4eab88ffc49c48eca43
+MD5_PPM_420M_ISLOW_3_4 = 3e726b7ea872445b19437d1c1d4f0d93
+MD5_PPM_420M_ISLOW_5_8 = a8a771abdc94301d20ffac119b2caccd
+MD5_PPM_420M_ISLOW_1_2 = b419124dd5568b085787234866102866
+MD5_PPM_420M_ISLOW_3_8 = 343d19015531b7bbe746124127244fa8
+MD5_PPM_420M_ISLOW_1_4 = 35fd59d866e44659edfa3c18db2a3edb
+MD5_PPM_420M_ISLOW_1_8 = ccaed48ac0aedefda5d4abe4013f4ad7
+MD5_PPM_420_ISLOW_SKIP15_31 = 86664cd9dc956536409e44e244d20a97
+MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 = 452a21656115a163029cfba5c04fa76a
+MD5_PPM_444_ISLOW_SKIP1_6 = ef63901f71ef7a75cd78253fc0914f84
+MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 = 15b173fb5872d9575572fbcc1b05956f
+MD5_JPEG_CROP = cdb35ff4b4519392690ea040c56ea99c
+
+else
+
+TESTORIG = testorig.jpg
+MD5_JPEG_RGB_ISLOW = 768e970dd57b340ff1b83c9d3d47c77b
+MD5_PPM_RGB_ISLOW = 00a257f5393fef8821f2b88ac7421291
+MD5_BMP_RGB_ISLOW_565 = f07d2e75073e4bb10f6c6f4d36e2e3be
+MD5_BMP_RGB_ISLOW_565D = 4cfa0928ef3e6bb626d7728c924cfda4
+MD5_JPEG_422_IFAST_OPT = 2540287b79d913f91665e660303ab2c8
+MD5_PPM_422_IFAST = 35bd6b3f833bad23de82acea847129fa
+MD5_PPM_422M_IFAST = 8dbc65323d62cca7c91ba02dd1cfa81d
+MD5_BMP_422M_IFAST_565 = 3294bd4d9a1f2b3d08ea6020d0db7065
+MD5_BMP_422M_IFAST_565D = da98c9c7b6039511be4a79a878a9abc1
+MD5_JPEG_420_IFAST_Q100_PROG = 990cbe0329c882420a2094da7e5adade
+MD5_PPM_420_Q100_IFAST = 5a732542015c278ff43635e473a8a294
+MD5_PPM_420M_Q100_IFAST = ff692ee9323a3b424894862557c092f1
+MD5_JPEG_GRAY_ISLOW = 72b51f894b8f4a10b3ee3066770aa38d
+MD5_PPM_GRAY_ISLOW = 8d3596c56eace32f205deccc229aa5ed
+MD5_PPM_GRAY_ISLOW_RGB = 116424ac07b79e5e801f00508eab48ec
+MD5_BMP_GRAY_ISLOW_565 = 12f78118e56a2f48b966f792fedf23cc
+MD5_BMP_GRAY_ISLOW_565D = bdbbd616441a24354c98553df5dc82db
+MD5_JPEG_420S_IFAST_OPT = 388708217ac46273ca33086b22827ed8
+# See README.md for more details on why this next bit is necessary.
+MD5_JPEG_3x2_FLOAT_PROG_SSE = 343e3f8caf8af5986ebaf0bdc13b5c71
+MD5_PPM_3x2_FLOAT_SSE = 1a75f36e5904d6fc3a85a43da9ad89bb
+MD5_JPEG_3x2_FLOAT_PROG_32BIT = 9bca803d2042bd1eb03819e2bf92b3e5
+MD5_PPM_3x2_FLOAT_32BIT = f6bfab038438ed8f5522fbd33595dcdc
+MD5_PPM_3x2_FLOAT_64BIT = 0e917a34193ef976b679a6b069b1be26
+MD5_JPEG_3x2_IFAST_PROG = 1ee5d2c1a77f2da495f993c8c7cceca5
+MD5_PPM_3x2_IFAST = fd283664b3b49127984af0a7f118fccd
+MD5_JPEG_420_ISLOW_ARI = e986fb0a637a8d833d96e8a6d6d84ea1
+MD5_JPEG_444_ISLOW_PROGARI = 0a8f1c8f66e113c3cf635df0a475a617
+MD5_PPM_420M_IFAST_ARI = 72b59a99bcf1de24c5b27d151bde2437
+MD5_JPEG_420_ISLOW = 9a68f56bc76e466aa7e52f415d0f4a5f
+MD5_PPM_420M_ISLOW_2_1 = 9f9de8c0612f8d06869b960b05abf9c9
+MD5_PPM_420M_ISLOW_15_8 = b6875bc070720b899566cc06459b63b7
+MD5_PPM_420M_ISLOW_13_8 = bc3452573c8152f6ae552939ee19f82f
+MD5_PPM_420M_ISLOW_11_8 = d8cc73c0aaacd4556569b59437ba00a5
+MD5_PPM_420M_ISLOW_9_8 = d25e61bc7eac0002f5b393aa223747b6
+MD5_PPM_420M_ISLOW_7_8 = ddb564b7c74a09494016d6cd7502a946
+MD5_PPM_420M_ISLOW_3_4 = 8ed8e68808c3fbc4ea764fc9d2968646
+MD5_PPM_420M_ISLOW_5_8 = a3363274999da2366a024efae6d16c9b
+MD5_PPM_420M_ISLOW_1_2 = e692a315cea26b988c8e8b29a5dbcd81
+MD5_PPM_420M_ISLOW_3_8 = 79eca9175652ced755155c90e785a996
+MD5_PPM_420M_ISLOW_1_4 = 79cd778f8bf1a117690052cacdd54eca
+MD5_PPM_420M_ISLOW_1_8 = 391b3d4aca640c8567d6f8745eb2142f
+MD5_BMP_420_ISLOW_256 = 4980185e3776e89bd931736e1cddeee6
+MD5_BMP_420_ISLOW_565 = bf9d13e16c4923b92e1faa604d7922cb
+MD5_BMP_420_ISLOW_565D = 6bde71526acc44bcff76f696df8638d2
+MD5_BMP_420M_ISLOW_565 = 8dc0185245353cfa32ad97027342216f
+MD5_BMP_420M_ISLOW_565D =d1be3a3339166255e76fa50a0d70d73e
+MD5_PPM_420_ISLOW_SKIP15_31 = c4c65c1e43d7275cd50328a61e6534f0
+MD5_PPM_420_ISLOW_ARI_SKIP16_139 = 087c6b123db16ac00cb88c5b590bb74a
+MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 = 26eb36ccc7d1f0cb80cdabb0ac8b5d99
+MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4 = 886c6775af22370257122f8b16207e6d
+MD5_PPM_444_ISLOW_SKIP1_6 = 5606f86874cf26b8fcee1117a0a436a6
+MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 = db87dc7ce26bcdc7a6b56239ce2b9d6c
+MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0 = cb57b32bd6d03e35432362f7bf184b6d
+MD5_JPEG_CROP = b4197f377e621c4e9b1d20471432610d
+
+endif
+
+.PHONY: test
+test: tjquicktest tjbittest bittest
+
+if CROSS_COMPILING
+tjquicktest: testclean
+else
+tjquicktest: testclean all
+endif
+
+if WITH_TURBOJPEG
+if WITH_JAVA
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -bi
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -noyuvpad
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -bi
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -bi -noyuvpad
+endif
+	./tjunittest
+	./tjunittest -alloc
+	./tjunittest -yuv
+	./tjunittest -yuv -alloc
+	./tjunittest -yuv -noyuvpad
+endif
+	echo GREAT SUCCESS!
+
+if CROSS_COMPILING
+tjbittest: testclean
+else
+tjbittest: testclean all
+endif
+
+if WITH_TURBOJPEG
+
+MD5_PPM_GRAY_TILE = 89d3ca21213d9d864b50b4e4e7de4ca6
+MD5_PPM_420_8x8_TILE = 847fceab15c5b7b911cb986cf0f71de3
+MD5_PPM_420_16x16_TILE = ca45552a93687e078f7137cc4126a7b0
+MD5_PPM_420_32x32_TILE = d8676f1d6b68df358353bba9844f4a00
+MD5_PPM_420_64x64_TILE = 4e4c1a3d7ea4bace4f868bcbe83b7050
+MD5_PPM_420_128x128_TILE = f24c3429c52265832beab9df72a0ceae
+MD5_PPM_420M_8x8_TILE = bc25320e1f4c31ce2e610e43e9fd173c
+MD5_PPM_420M_TILE = 75ffdf14602258c5c189522af57fa605
+MD5_PPM_422_8x8_TILE = d83dacd9fc73b0a6f10c09acad64eb1e
+MD5_PPM_422_16x16_TILE = 35077fb610d72dd743b1eb0cbcfe10fb
+MD5_PPM_422_32x32_TILE = e6902ed8a449ecc0f0d6f2bf945f65f7
+MD5_PPM_422_64x64_TILE = 2b4502a8f316cedbde1da7bce3d2231e
+MD5_PPM_422_128x128_TILE = f0b5617d578f5e13c8eee215d64d4877
+MD5_PPM_422M_8x8_TILE = 828941d7f41cd6283abd6beffb7fd51d
+MD5_PPM_422M_TILE = e877ae1324c4a280b95376f7f018172f
+MD5_PPM_444_TILE = 7964e41e67cfb8d0a587c0aa4798f9c3
+
+# Test compressing from/decompressing to an arbitrary subregion of a larger
+# image buffer
+	cp $(srcdir)/testimages/testorig.ppm testout_tile.ppm
+	./tjbench testout_tile.ppm 95 -rgb -quiet -tile -benchtime 0.01 >/dev/null 2>&1
+	for i in 8 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_GRAY_TILE) testout_tile_GRAY_Q95_$$i\x$$i.ppm; \
+	done
+	md5/md5cmp $(MD5_PPM_420_8x8_TILE) testout_tile_420_Q95_8x8.ppm
+	md5/md5cmp $(MD5_PPM_420_16x16_TILE) testout_tile_420_Q95_16x16.ppm
+	md5/md5cmp $(MD5_PPM_420_32x32_TILE) testout_tile_420_Q95_32x32.ppm
+	md5/md5cmp $(MD5_PPM_420_64x64_TILE) testout_tile_420_Q95_64x64.ppm
+	md5/md5cmp $(MD5_PPM_420_128x128_TILE) testout_tile_420_Q95_128x128.ppm
+	md5/md5cmp $(MD5_PPM_422_8x8_TILE) testout_tile_422_Q95_8x8.ppm
+	md5/md5cmp $(MD5_PPM_422_16x16_TILE) testout_tile_422_Q95_16x16.ppm
+	md5/md5cmp $(MD5_PPM_422_32x32_TILE) testout_tile_422_Q95_32x32.ppm
+	md5/md5cmp $(MD5_PPM_422_64x64_TILE) testout_tile_422_Q95_64x64.ppm
+	md5/md5cmp $(MD5_PPM_422_128x128_TILE) testout_tile_422_Q95_128x128.ppm
+	for i in 8 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_444_TILE) testout_tile_444_Q95_$$i\x$$i.ppm; \
+	done
+	rm -f testout_tile_GRAY_* testout_tile_420_* testout_tile_422_* testout_tile_444_*
+
+	./tjbench testout_tile.ppm 95 -rgb -fastupsample -quiet -tile -benchtime 0.01 >/dev/null 2>&1
+	md5/md5cmp $(MD5_PPM_420M_8x8_TILE) testout_tile_420_Q95_8x8.ppm
+	for i in 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_420M_TILE) testout_tile_420_Q95_$$i\x$$i.ppm; \
+	done
+	md5/md5cmp $(MD5_PPM_422M_8x8_TILE) testout_tile_422_Q95_8x8.ppm
+	for i in 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_422M_TILE) testout_tile_422_Q95_$$i\x$$i.ppm; \
+	done
+	rm -f testout_tile_GRAY_* testout_tile_420_* testout_tile_422_* testout_tile_444_* testout_tile.ppm
+	echo GREAT SUCCESS!
+
+endif
+
+if CROSS_COMPILING
+bittest: testclean
+else
+bittest: testclean all
+endif
+
+# These tests are carefully crafted to provide full coverage of as many of the
+# underlying algorithms as possible (including all of the SIMD-accelerated
+# ones.)
+
+# CC: null  SAMP: fullsize  FDCT: islow  ENT: huff
+	./cjpeg -rgb -dct int -outfile testout_rgb_islow.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_RGB_ISLOW) testout_rgb_islow.jpg
+# CC: null  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -ppm -outfile testout_rgb_islow.ppm testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_PPM_RGB_ISLOW) testout_rgb_islow.ppm
+	rm -f testout_rgb_islow.ppm
+if WITH_12BIT
+	rm -f testout_rgb_islow.jpg
+else
+# CC: RGB->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_rgb_islow_565.bmp testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_BMP_RGB_ISLOW_565) testout_rgb_islow_565.bmp
+	rm -f testout_rgb_islow_565.bmp
+# CC: RGB->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_rgb_islow_565D.bmp testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_BMP_RGB_ISLOW_565D) testout_rgb_islow_565D.bmp
+	rm -f testout_rgb_islow_565D.bmp testout_rgb_islow.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize/h2v1  FDCT: ifast  ENT: 2-pass huff
+	./cjpeg -sample 2x1 -dct fast -opt -outfile testout_422_ifast_opt.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_422_IFAST_OPT) testout_422_ifast_opt.jpg
+# CC: YCC->RGB  SAMP: fullsize/h2v1 fancy  IDCT: ifast  ENT: huff
+	./djpeg -dct fast -outfile testout_422_ifast.ppm testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_PPM_422_IFAST) testout_422_ifast.ppm
+	rm -f testout_422_ifast.ppm
+# CC: YCC->RGB  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct fast -nosmooth -outfile testout_422m_ifast.ppm testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_PPM_422M_IFAST) testout_422m_ifast.ppm
+	rm -f testout_422m_ifast.ppm
+if WITH_12BIT
+	rm -f testout_422_ifast_opt.jpg
+else
+# CC: YCC->RGB565  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -dither none -bmp -outfile testout_422m_ifast_565.bmp testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_BMP_422M_IFAST_565) testout_422m_ifast_565.bmp
+	rm -f testout_422m_ifast_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -bmp -outfile testout_422m_ifast_565D.bmp testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_BMP_422M_IFAST_565D) testout_422m_ifast_565D.bmp
+	rm -f testout_422m_ifast_565D.bmp testout_422_ifast_opt.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize/h2v2  FDCT: ifast  ENT: prog huff
+	./cjpeg -sample 2x2 -quality 100 -dct fast -prog -outfile testout_420_q100_ifast_prog.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420_IFAST_Q100_PROG) testout_420_q100_ifast_prog.jpg
+# CC: YCC->RGB  SAMP: fullsize/h2v2 fancy  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -outfile testout_420_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_420_Q100_IFAST) testout_420_q100_ifast.ppm
+	rm -f testout_420_q100_ifast.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -nosmooth -outfile testout_420m_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_420M_Q100_IFAST) testout_420m_q100_ifast.ppm
+	rm -f testout_420m_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+
+# CC: RGB->Gray  SAMP: fullsize  FDCT: islow  ENT: huff
+	./cjpeg -gray -dct int -outfile testout_gray_islow.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_GRAY_ISLOW) testout_gray_islow.jpg
+# CC: Gray->Gray  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -outfile testout_gray_islow.ppm testout_gray_islow.jpg
+	md5/md5cmp $(MD5_PPM_GRAY_ISLOW) testout_gray_islow.ppm
+	rm -f testout_gray_islow.ppm
+# CC: Gray->RGB  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb -outfile testout_gray_islow_rgb.ppm testout_gray_islow.jpg
+	md5/md5cmp $(MD5_PPM_GRAY_ISLOW_RGB) testout_gray_islow_rgb.ppm
+	rm -f testout_gray_islow_rgb.ppm
+if WITH_12BIT
+	rm -f testout_gray_islow.jpg
+else
+# CC: Gray->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_gray_islow_565.bmp testout_gray_islow.jpg
+	md5/md5cmp $(MD5_BMP_GRAY_ISLOW_565) testout_gray_islow_565.bmp
+	rm -f testout_gray_islow_565.bmp
+# CC: Gray->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_gray_islow_565D.bmp testout_gray_islow.jpg
+	md5/md5cmp $(MD5_BMP_GRAY_ISLOW_565D) testout_gray_islow_565D.bmp
+	rm -f testout_gray_islow_565D.bmp testout_gray_islow.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize smooth/h2v2 smooth  FDCT: islow
+# ENT: 2-pass huff
+	./cjpeg -sample 2x2 -smooth 1 -dct int -opt -outfile testout_420s_ifast_opt.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420S_IFAST_OPT) testout_420s_ifast_opt.jpg
+	rm -f testout_420s_ifast_opt.jpg
+
+# The output of the floating point tests is not validated by default, because
+# the output differs depending on the type of floating point math used, and
+# this is only deterministic if the DCT/IDCT are implemented using SIMD
+# instructions on a particular platform.  Pass one of the following on the make
+# command line to validate the floating point tests against one of the expected
+# results:
+#
+# FLOATTEST=sse  validate against the expected results from the libjpeg-turbo
+#                SSE SIMD extensions
+# FLOATTEST=32bit  validate against the expected results from the C code
+#                  when running on a 32-bit FPU (or when SSE is being used for
+#                  floating point math, which is generally the default with
+#                  x86-64 compilers)
+# FLOATTEST=64bit  validate against the exepected results from the C code
+#                  when running on a 64-bit FPU
+
+# CC: RGB->YCC  SAMP: fullsize/int  FDCT: float  ENT: prog huff
+	./cjpeg -sample 3x2 -dct float -prog -outfile testout_3x2_float_prog.jpg $(srcdir)/testimages/testorig.ppm
+	if [ "${FLOATTEST}" = "sse" ]; then \
+		md5/md5cmp $(MD5_JPEG_3x2_FLOAT_PROG_SSE) testout_3x2_float_prog.jpg; \
+	elif [ "${FLOATTEST}" = "32bit" -o "${FLOATTEST}" = "64bit" ]; then \
+		md5/md5cmp $(MD5_JPEG_3x2_FLOAT_PROG_32BIT) testout_3x2_float_prog.jpg; \
+	fi
+# CC: YCC->RGB  SAMP: fullsize/int  IDCT: float  ENT: prog huff
+	./djpeg -dct float -outfile testout_3x2_float.ppm testout_3x2_float_prog.jpg
+	if [ "${FLOATTEST}" = "sse" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_SSE) testout_3x2_float.ppm; \
+	elif [ "${FLOATTEST}" = "32bit" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_32BIT) testout_3x2_float.ppm; \
+	elif [ "${FLOATTEST}" = "64bit" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_64BIT) testout_3x2_float.ppm; \
+	fi
+	rm -f testout_3x2_float.ppm testout_3x2_float_prog.jpg
+
+# CC: RGB->YCC  SAMP: fullsize/int  FDCT: ifast  ENT: prog huff
+	./cjpeg -sample 3x2 -dct fast -prog -outfile testout_3x2_ifast_prog.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_3x2_IFAST_PROG) testout_3x2_ifast_prog.jpg
+# CC: YCC->RGB  SAMP: fullsize/int  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -outfile testout_3x2_ifast.ppm testout_3x2_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_3x2_IFAST) testout_3x2_ifast.ppm
+	rm -f testout_3x2_ifast.ppm testout_3x2_ifast_prog.jpg
+
+if WITH_ARITH_ENC
+# CC: YCC->RGB  SAMP: fullsize/h2v2  FDCT: islow  ENT: arith
+	./cjpeg -dct int -arithmetic -outfile testout_420_islow_ari.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420_ISLOW_ARI) testout_420_islow_ari.jpg
+	rm -f testout_420_islow_ari.jpg
+	./jpegtran -arithmetic -outfile testout_420_islow_ari.jpg $(srcdir)/testimages/testimgint.jpg
+	md5/md5cmp $(MD5_JPEG_420_ISLOW_ARI) testout_420_islow_ari.jpg
+	rm -f testout_420_islow_ari.jpg
+# CC: YCC->RGB  SAMP: fullsize  FDCT: islow  ENT: prog arith
+	./cjpeg -sample 1x1 -dct int -prog -arithmetic -outfile testout_444_islow_progari.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_444_ISLOW_PROGARI) testout_444_islow_progari.jpg
+	rm -f testout_444_islow_progari.jpg
+endif
+if WITH_ARITH_DEC
+# CC: RGB->YCC  SAMP: h2v2 merged  IDCT: ifast  ENT: arith
+	./djpeg -fast -ppm -outfile testout_420m_ifast_ari.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420M_IFAST_ARI) testout_420m_ifast_ari.ppm
+	rm -f testout_420m_ifast_ari.ppm
+	./jpegtran -outfile testout_420_islow.jpg $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_JPEG_420_ISLOW) testout_420_islow.jpg
+	rm -f testout_420_islow.jpg
+endif
+
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 16x16 islow  ENT: huff
+	./djpeg -dct int -scale 2/1 -nosmooth -ppm -outfile testout_420m_islow_2_1.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_2_1) testout_420m_islow_2_1.ppm
+	rm -f testout_420m_islow_2_1.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 15x15 islow  ENT: huff
+	./djpeg -dct int -scale 15/8 -nosmooth -ppm -outfile testout_420m_islow_15_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_15_8) testout_420m_islow_15_8.ppm
+	rm -f testout_420m_islow_15_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 13x13 islow  ENT: huff
+	./djpeg -dct int -scale 13/8 -nosmooth -ppm -outfile testout_420m_islow_13_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_13_8) testout_420m_islow_13_8.ppm
+	rm -f testout_420m_islow_13_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 11x11 islow  ENT: huff
+	./djpeg -dct int -scale 11/8 -nosmooth -ppm -outfile testout_420m_islow_11_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_11_8) testout_420m_islow_11_8.ppm
+	rm -f testout_420m_islow_11_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 9x9 islow  ENT: huff
+	./djpeg -dct int -scale 9/8 -nosmooth -ppm -outfile testout_420m_islow_9_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_9_8) testout_420m_islow_9_8.ppm
+	rm -f testout_420m_islow_9_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 7x7 islow/14x14 islow  ENT: huff
+	./djpeg -dct int -scale 7/8 -nosmooth -ppm -outfile testout_420m_islow_7_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_7_8) testout_420m_islow_7_8.ppm
+	rm -f testout_420m_islow_7_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 6x6 islow/12x12 islow  ENT: huff
+	./djpeg -dct int -scale 3/4 -nosmooth -ppm -outfile testout_420m_islow_3_4.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_3_4) testout_420m_islow_3_4.ppm
+	rm -f testout_420m_islow_3_4.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 5x5 islow/10x10 islow  ENT: huff
+	./djpeg -dct int -scale 5/8 -nosmooth -ppm -outfile testout_420m_islow_5_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_5_8) testout_420m_islow_5_8.ppm
+	rm -f testout_420m_islow_5_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 4x4 islow/8x8 islow  ENT: huff
+	./djpeg -dct int -scale 1/2 -nosmooth -ppm -outfile testout_420m_islow_1_2.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_2) testout_420m_islow_1_2.ppm
+	rm -f testout_420m_islow_1_2.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 3x3 islow/6x6 islow  ENT: huff
+	./djpeg -dct int -scale 3/8 -nosmooth -ppm -outfile testout_420m_islow_3_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_3_8) testout_420m_islow_3_8.ppm
+	rm -f testout_420m_islow_3_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 2x2 islow/4x4 islow  ENT: huff
+	./djpeg -dct int -scale 1/4 -nosmooth -ppm -outfile testout_420m_islow_1_4.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_4) testout_420m_islow_1_4.ppm
+	rm -f testout_420m_islow_1_4.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 1x1 islow/2x2 islow  ENT: huff
+	./djpeg -dct int -scale 1/8 -nosmooth -ppm -outfile testout_420m_islow_1_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_8) testout_420m_islow_1_8.ppm
+	rm -f testout_420m_islow_1_8.ppm
+if WITH_12BIT
+else
+# CC: YCC->RGB (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -colors 256 -bmp -outfile testout_420_islow_256.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_256) testout_420_islow_256.bmp
+	rm -f testout_420_islow_256.bmp
+# CC: YCC->RGB565  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_420_islow_565.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_565) testout_420_islow_565.bmp
+	rm -f testout_420_islow_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_420_islow_565D.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_565D) testout_420_islow_565D.bmp
+	rm -f testout_420_islow_565D.bmp
+# CC: YCC->RGB565  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -dither none -bmp -outfile testout_420m_islow_565.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420M_ISLOW_565) testout_420m_islow_565.bmp
+	rm -f testout_420m_islow_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -bmp -outfile testout_420m_islow_565D.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420M_ISLOW_565D) testout_420m_islow_565D.bmp
+	rm -f testout_420m_islow_565D.bmp
+endif
+
+# Partial decode tests.  These tests are designed to cover all of the possible
+# code paths in jpeg_skip_scanlines().
+
+# Context rows: Yes  Intra-iMCU row: Yes  iMCU row prefetch: No   ENT: huff
+	./djpeg -dct int -skip 15,31 -ppm -outfile testout_420_islow_skip15,31.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420_ISLOW_SKIP15_31) testout_420_islow_skip15,31.ppm
+	rm -f testout_420_islow_skip15,31.ppm
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: Yes  ENT: arith
+if WITH_ARITH_DEC
+	./djpeg -dct int -skip 16,139 -ppm -outfile testout_420_islow_ari_skip16,139.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_ARI_SKIP16_139) testout_420_islow_ari_skip16,139.ppm
+	rm -f testout_420_islow_ari_skip16,139.ppm
+endif
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: prog huff
+	./cjpeg -dct int -prog -outfile testout_420_islow_prog.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -crop 62x62+71+71 -ppm -outfile testout_420_islow_prog_crop62x62,71,71.ppm testout_420_islow_prog.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71) testout_420_islow_prog_crop62x62,71,71.ppm
+	rm -f testout_420_islow_prog_crop62x62,71,71.ppm testout_420_islow_prog.jpg
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: arith
+if WITH_ARITH_DEC
+	./djpeg -dct int -crop 53x53+4+4 -ppm -outfile testout_420_islow_ari_crop53x53,4,4.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4) testout_420_islow_ari_crop53x53,4,4.ppm
+	rm -f testout_420_islow_ari_crop53x53,4,4.ppm
+endif
+# Context rows: No   Intra-iMCU row: Yes  ENT: huff
+	./cjpeg -dct int -sample 1x1 -outfile testout_444_islow.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -skip 1,6 -ppm -outfile testout_444_islow_skip1,6.ppm testout_444_islow.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_SKIP1_6) testout_444_islow_skip1,6.ppm
+	rm -f testout_444_islow_skip1,6.ppm testout_444_islow.jpg
+# Context rows: No   Intra-iMCU row: No   ENT: prog huff
+	./cjpeg -dct int -prog -sample 1x1 -outfile testout_444_islow_prog.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -crop 98x98+13+13 -ppm -outfile testout_444_islow_prog_crop98x98,13,13.ppm testout_444_islow_prog.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13) testout_444_islow_prog_crop98x98,13,13.ppm
+	rm -f testout_444_islow_prog_crop98x98,13,13.ppm testout_444_islow_prog.jpg
+# Context rows: No   Intra-iMCU row: No   ENT: arith
+if WITH_ARITH_ENC
+	./cjpeg -dct int -arithmetic -sample 1x1 -outfile testout_444_islow_ari.jpg $(srcdir)/testimages/testorig.ppm
+if WITH_ARITH_DEC
+	./djpeg -dct int -crop 37x37+0+0 -ppm -outfile testout_444_islow_ari_crop37x37,0,0.ppm testout_444_islow_ari.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0) testout_444_islow_ari_crop37x37,0,0.ppm
+	rm -f testout_444_islow_ari_crop37x37,0,0.ppm
+endif
+	rm -f testout_444_islow_ari.jpg
+endif
+
+	./jpegtran -crop 120x90+20+50 -transpose -perfect -outfile testout_crop.jpg $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_JPEG_CROP) testout_crop.jpg
+	rm -f testout_crop.jpg
+	echo GREAT SUCCESS!
+
+
+testclean:
+	rm -f testout*
+	rm -f *_GRAY_*.bmp
+	rm -f *_GRAY_*.png
+	rm -f *_GRAY_*.ppm
+	rm -f *_GRAY_*.jpg
+	rm -f *_GRAY.yuv
+	rm -f *_420_*.bmp
+	rm -f *_420_*.png
+	rm -f *_420_*.ppm
+	rm -f *_420_*.jpg
+	rm -f *_420.yuv
+	rm -f *_422_*.bmp
+	rm -f *_422_*.png
+	rm -f *_422_*.ppm
+	rm -f *_422_*.jpg
+	rm -f *_422.yuv
+	rm -f *_444_*.bmp
+	rm -f *_444_*.png
+	rm -f *_444_*.ppm
+	rm -f *_444_*.jpg
+	rm -f *_444.yuv
+	rm -f *_440_*.bmp
+	rm -f *_440_*.png
+	rm -f *_440_*.ppm
+	rm -f *_440_*.jpg
+	rm -f *_440.yuv
+	rm -f *_411_*.bmp
+	rm -f *_411_*.png
+	rm -f *_411_*.ppm
+	rm -f *_411_*.jpg
+	rm -f *_411.yuv
+
+
+tjtest:
+	sh ./tjbenchtest
+	sh ./tjbenchtest -alloc
+	sh ./tjbenchtest -yuv
+	sh ./tjbenchtest -yuv -alloc
+if WITH_JAVA
+	sh ./tjbenchtest.java
+	sh ./tjbenchtest.java -yuv
+endif
+
+
+pkgscripts/libjpeg-turbo.spec: pkgscripts/libjpeg-turbo.spec.tmpl
+	cat pkgscripts/libjpeg-turbo.spec.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__datadir}@$(datadir)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__includedir}@$(includedir)@g | \
+		sed s@%{__libdir}@$(libdir)@g | sed s@%{__mandir}@$(mandir)@g \
+		> pkgscripts/libjpeg-turbo.spec
+
+rpm: all pkgscripts/libjpeg-turbo.spec
+	TMPDIR=`mktemp -d /tmp/${PACKAGE_NAME}-build.XXXXXX`; \
+	mkdir -p $$TMPDIR/RPMS; \
+	ln -fs `pwd` $$TMPDIR/BUILD; \
+	rm -f ${PKGNAME}-${VERSION}.${RPMARCH}.rpm; \
+	rpmbuild -bb --define "_blddir $$TMPDIR/buildroot"  \
+		--define "_topdir $$TMPDIR" \
+		--target ${RPMARCH} pkgscripts/libjpeg-turbo.spec; \
+	cp $$TMPDIR/RPMS/${RPMARCH}/${PKGNAME}-${VERSION}-${BUILD}.${RPMARCH}.rpm \
+		${PKGNAME}-${VERSION}.${RPMARCH}.rpm; \
+	rm -rf $$TMPDIR
+
+srpm: dist-gzip pkgscripts/libjpeg-turbo.spec
+	TMPDIR=`mktemp -d /tmp/${PACKAGE_NAME}-build.XXXXXX`; \
+	mkdir -p $$TMPDIR/RPMS; \
+	mkdir -p $$TMPDIR/SRPMS; \
+	mkdir -p $$TMPDIR/BUILD; \
+	mkdir -p $$TMPDIR/SOURCES; \
+	mkdir -p $$TMPDIR/SPECS; \
+	rm -f ${PKGNAME}-${VERSION}.src.rpm; \
+	cp ${PACKAGE_NAME}-${VERSION}.tar.gz $$TMPDIR/SOURCES; \
+	cat pkgscripts/libjpeg-turbo.spec | sed s/%{_blddir}/%{_tmppath}/g \
+		| sed s/#--\>//g \
+		> $$TMPDIR/SPECS/libjpeg-turbo.spec; \
+	rpmbuild -bs --define "_topdir $$TMPDIR" $$TMPDIR/SPECS/libjpeg-turbo.spec; \
+	cp $$TMPDIR/SRPMS/${PKGNAME}-${VERSION}-${BUILD}.src.rpm \
+		${PKGNAME}-${VERSION}.src.rpm; \
+	rm -rf $$TMPDIR
+
+pkgscripts/makedpkg: pkgscripts/makedpkg.tmpl
+	cat pkgscripts/makedpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__libdir}@$(libdir)@g \
+		> pkgscripts/makedpkg
+
+deb: all pkgscripts/makedpkg
+	sh pkgscripts/makedpkg
+
+pkgscripts/uninstall: pkgscripts/uninstall.tmpl
+	cat pkgscripts/uninstall.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__datadir}@$(datadir)@g | \
+		sed s@%{__includedir}@$(includedir)@g | sed s@%{__libdir}@$(libdir)@g | \
+		sed s@%{__mandir}@$(mandir)@g > pkgscripts/uninstall
+
+pkgscripts/makemacpkg: pkgscripts/makemacpkg.tmpl
+	cat pkgscripts/makemacpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__docdir}@$(docdir)@g | \
+		sed s@%{__libdir}@$(libdir)@g > pkgscripts/makemacpkg
+
+if X86_64
+
+udmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -build32 ${BUILDDIR32}
+
+iosdmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -build32 ${BUILDDIR32} -buildarmv6 ${BUILDDIRARMV6} -buildarmv7 ${BUILDDIRARMV7} -buildarmv7s ${BUILDDIRARMV7S} -buildarmv8 ${BUILDDIRARMV8} -lipo "${LIPO}"
+
+else
+
+iosdmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -buildarmv6 ${BUILDDIRARMV6} -buildarmv7 ${BUILDDIRARMV7} -buildarmv7s ${BUILDDIRARMV7S} -buildarmv8 ${BUILDDIRARMV8} -lipo "${LIPO}"
+
+endif
+
+dmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg
+
+pkgscripts/makecygwinpkg: pkgscripts/makecygwinpkg.tmpl
+	cat pkgscripts/makecygwinpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__libdir}@$(libdir)@g \
+		> pkgscripts/makecygwinpkg
+
+cygwinpkg: all pkgscripts/makecygwinpkg
+	sh pkgscripts/makecygwinpkg
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdbmp.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdbmp.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdbmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdbmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,483 @@
+/*
+ * rdbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Modified 2011 by Siarhei Siamashka.
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Microsoft "BMP"
+ * format (MS Windows 3.x, OS/2 1.x, and OS/2 2.x flavors).
+ * Currently, only 8-bit and 24-bit images are supported, not 1-bit or
+ * 4-bit (feeding such low-depth images into JPEG would be silly anyway).
+ * Also, we don't support RLE-compressed files.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed BMP format).
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef BMP_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _bmp_source_struct *bmp_source_ptr;
+
+typedef struct _bmp_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* BMP colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed to reverse row order */
+  JDIMENSION source_row;        /* Current source row number */
+  JDIMENSION row_width;         /* Physical width of scanlines in file */
+
+  int bits_per_pixel;           /* remembers 8- or 24-bit format */
+} bmp_source_struct;
+
+
+LOCAL(int)
+read_byte (bmp_source_ptr sinfo)
+/* Read next byte from BMP file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (bmp_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a BMP file */
+{
+  int i;
+
+  switch (mapentrysize) {
+  case 3:
+    /* BGR format (occurs in OS/2 files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+    }
+    break;
+  case 4:
+    /* BGR0 format (occurs in MS Windows files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+      (void) read_byte(sinfo);
+    }
+    break;
+  default:
+    ERREXIT(sinfo->cinfo, JERR_BMP_BADCMAP);
+    break;
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ * The image has been read into the whole_image array, but is otherwise
+ * unprocessed.  We must read it out in top-to-bottom row order, and if
+ * it is an 8-bit image, we must expand colormapped pixels to 24bit format.
+ */
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register JSAMPARRAY colormap = source->colormap;
+  JSAMPARRAY image_ptr;
+  register int t;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Expand the colormap indexes to real data */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    t = GETJSAMPLE(*inptr++);
+    *outptr++ = colormap[0][t]; /* can omit GETJSAMPLE() safely */
+    *outptr++ = colormap[1][t];
+    *outptr++ = colormap[2][t];
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    outptr += 3;
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_32bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 32-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    inptr++;                    /* skip the 4th byte (Alpha channel) */
+    outptr += 3;
+  }
+
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_8bit_row, get_24bit_row, or get_32bit_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register FILE *infile = source->pub.input_file;
+  register JSAMPROW out_ptr;
+  JSAMPARRAY image_ptr;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image,
+       row, (JDIMENSION) 1, TRUE);
+    out_ptr = image_ptr[0];
+    if (fread(out_ptr, 1, source->row_width, infile) != source->row_width) {
+      if (feof(infile))
+        ERREXIT(cinfo, JERR_INPUT_EOF);
+      else
+        ERREXIT(cinfo, JERR_FILE_READ);
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in top-to-bottom order */
+  switch (source->bits_per_pixel) {
+  case 8:
+    source->pub.get_pixel_rows = get_8bit_row;
+    break;
+  case 24:
+    source->pub.get_pixel_rows = get_24bit_row;
+    break;
+  case 32:
+    source->pub.get_pixel_rows = get_32bit_row;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+  }
+  source->source_row = cinfo->image_height;
+
+  /* And read the first row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  U_CHAR bmpfileheader[14];
+  U_CHAR bmpinfoheader[64];
+#define GET_2B(array,offset)  ((unsigned short) UCH(array[offset]) + \
+                               (((unsigned short) UCH(array[offset+1])) << 8))
+#define GET_4B(array,offset)  ((unsigned int) UCH(array[offset]) + \
+                               (((unsigned int) UCH(array[offset+1])) << 8) + \
+                               (((unsigned int) UCH(array[offset+2])) << 16) + \
+                               (((unsigned int) UCH(array[offset+3])) << 24))
+  unsigned int bfOffBits;
+  unsigned int headerSize;
+  int biWidth;
+  int biHeight;
+  unsigned short biPlanes;
+  unsigned int biCompression;
+  int biXPelsPerMeter,biYPelsPerMeter;
+  unsigned int biClrUsed = 0;
+  int mapentrysize = 0;         /* 0 indicates no colormap */
+  int bPad;
+  JDIMENSION row_width;
+
+  /* Read and verify the bitmap file header */
+  if (! ReadOK(source->pub.input_file, bmpfileheader, 14))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  if (GET_2B(bmpfileheader,0) != 0x4D42) /* 'BM' */
+    ERREXIT(cinfo, JERR_BMP_NOT);
+  bfOffBits = GET_4B(bmpfileheader,10);
+  /* We ignore the remaining fileheader fields */
+
+  /* The infoheader might be 12 bytes (OS/2 1.x), 40 bytes (Windows),
+   * or 64 bytes (OS/2 2.x).  Check the first 4 bytes to find out which.
+   */
+  if (! ReadOK(source->pub.input_file, bmpinfoheader, 4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  headerSize = GET_4B(bmpinfoheader,0);
+  if (headerSize < 12 || headerSize > 64)
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  if (! ReadOK(source->pub.input_file, bmpinfoheader+4, headerSize-4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  switch (headerSize) {
+  case 12:
+    /* Decode OS/2 1.x header (Microsoft calls this a BITMAPCOREHEADER) */
+    biWidth = (int) GET_2B(bmpinfoheader,4);
+    biHeight = (int) GET_2B(bmpinfoheader,6);
+    biPlanes = GET_2B(bmpinfoheader,8);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,10);
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 3;         /* OS/2 uses RGBTRIPLE colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2, biWidth, biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    break;
+  case 40:
+  case 64:
+    /* Decode Windows 3.x header (Microsoft calls this a BITMAPINFOHEADER) */
+    /* or OS/2 2.x header, which has additional fields that we ignore */
+    biWidth = (int) GET_4B(bmpinfoheader,4);
+    biHeight = (int) GET_4B(bmpinfoheader,8);
+    biPlanes = GET_2B(bmpinfoheader,12);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,14);
+    biCompression = GET_4B(bmpinfoheader,16);
+    biXPelsPerMeter = (int) GET_4B(bmpinfoheader,24);
+    biYPelsPerMeter = (int) GET_4B(bmpinfoheader,28);
+    biClrUsed = GET_4B(bmpinfoheader,32);
+    /* biSizeImage, biClrImportant fields are ignored */
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 4;         /* Windows uses RGBQUAD colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP, biWidth, biHeight);
+      break;
+    case 32:                    /* RGB image + Alpha channel */
+      TRACEMS2(cinfo, 1, JTRC_BMP, biWidth, biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    if (biCompression != 0)
+      ERREXIT(cinfo, JERR_BMP_COMPRESSED);
+
+    if (biXPelsPerMeter > 0 && biYPelsPerMeter > 0) {
+      /* Set JFIF density parameters from the BMP data */
+      cinfo->X_density = (UINT16) (biXPelsPerMeter/100); /* 100 cm per meter */
+      cinfo->Y_density = (UINT16) (biYPelsPerMeter/100);
+      cinfo->density_unit = 2;  /* dots/cm */
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+    return;
+  }
+
+  if (biWidth <= 0 || biHeight <= 0)
+    ERREXIT(cinfo, JERR_BMP_EMPTY);
+  if (biPlanes != 1)
+    ERREXIT(cinfo, JERR_BMP_BADPLANES);
+
+  /* Compute distance to bitmap data --- will adjust for colormap below */
+  bPad = bfOffBits - (headerSize + 14);
+
+  /* Read the colormap, if any */
+  if (mapentrysize > 0) {
+    if (biClrUsed <= 0)
+      biClrUsed = 256;          /* assume it's 256 */
+    else if (biClrUsed > 256)
+      ERREXIT(cinfo, JERR_BMP_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) biClrUsed, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) biClrUsed, mapentrysize);
+    /* account for size of colormap */
+    bPad -= biClrUsed * mapentrysize;
+  }
+
+  /* Skip any remaining pad bytes */
+  if (bPad < 0)                 /* incorrect bfOffBits value? */
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  while (--bPad >= 0) {
+    (void) read_byte(source);
+  }
+
+  /* Compute row width in file, including padding to 4-byte boundary */
+  if (source->bits_per_pixel == 24)
+    row_width = (JDIMENSION) (biWidth * 3);
+  else if (source->bits_per_pixel == 32)
+    row_width = (JDIMENSION) (biWidth * 4);
+  else
+    row_width = (JDIMENSION) biWidth;
+  while ((row_width & 3) != 0) row_width++;
+  source->row_width = row_width;
+
+  /* Allocate space for inversion array, prepare for preload pass */
+  source->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     row_width, (JDIMENSION) biHeight, (JDIMENSION) 1);
+  source->pub.get_pixel_rows = preload_image;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Allocate one-row buffer for returned data */
+  source->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (biWidth * 3), (JDIMENSION) 1);
+  source->pub.buffer_height = 1;
+
+  cinfo->in_color_space = JCS_RGB;
+  cinfo->input_components = 3;
+  cinfo->data_precision = 8;
+  cinfo->image_width = (JDIMENSION) biWidth;
+  cinfo->image_height = (JDIMENSION) biHeight;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for BMP format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_bmp (j_compress_ptr cinfo)
+{
+  bmp_source_ptr source;
+
+  /* Create module interface object */
+  source = (bmp_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(bmp_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_bmp;
+  source->pub.finish_input = finish_input_bmp;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* BMP_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdcolmap.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdcolmap.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdcolmap.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdcolmap.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,254 @@
+/*
+ * rdcolmap.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file implements djpeg's "-map file" switch.  It reads a source image
+ * and constructs a colormap to be supplied to the JPEG decompressor.
+ *
+ * Currently, these file formats are supported for the map file:
+ *   GIF: the contents of the GIF's global colormap are used.
+ *   PPM (either text or raw flavor): the entire file is read and
+ *      each unique pixel value is entered in the map.
+ * Note that reading a large PPM file will be horrendously slow.
+ * Typically, a PPM-format map file should contain just one pixel
+ * of each desired color.  Such a file can be extracted from an
+ * ordinary image PPM file with ppmtomap(1).
+ *
+ * Rescaling a PPM that has a maxval unequal to MAXJSAMPLE is not
+ * currently implemented.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/*
+ * Add a (potentially) new color to the color map.
+ */
+
+LOCAL(void)
+add_map_entry (j_decompress_ptr cinfo, int R, int G, int B)
+{
+  JSAMPROW colormap0 = cinfo->colormap[0];
+  JSAMPROW colormap1 = cinfo->colormap[1];
+  JSAMPROW colormap2 = cinfo->colormap[2];
+  int ncolors = cinfo->actual_number_of_colors;
+  int index;
+
+  /* Check for duplicate color. */
+  for (index = 0; index < ncolors; index++) {
+    if (GETJSAMPLE(colormap0[index]) == R &&
+        GETJSAMPLE(colormap1[index]) == G &&
+        GETJSAMPLE(colormap2[index]) == B)
+      return;                   /* color is already in map */
+  }
+
+  /* Check for map overflow. */
+  if (ncolors >= (MAXJSAMPLE+1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, (MAXJSAMPLE+1));
+
+  /* OK, add color to map. */
+  colormap0[ncolors] = (JSAMPLE) R;
+  colormap1[ncolors] = (JSAMPLE) G;
+  colormap2[ncolors] = (JSAMPLE) B;
+  cinfo->actual_number_of_colors++;
+}
+
+
+/*
+ * Extract color map from a GIF file.
+ */
+
+LOCAL(void)
+read_gif_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  int header[13];
+  int i, colormaplen;
+  int R, G, B;
+
+  /* Initial 'G' has already been read by read_color_map */
+  /* Read the rest of the GIF header and logical screen descriptor */
+  for (i = 1; i < 13; i++) {
+    if ((header[i] = getc(infile)) == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  }
+
+  /* Verify GIF Header */
+  if (header[1] != 'I' || header[2] != 'F')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* There must be a global color map. */
+  if ((header[10] & 0x80) == 0)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* OK, fetch it. */
+  colormaplen = 2 << (header[10] & 0x07);
+
+  for (i = 0; i < colormaplen; i++) {
+    R = getc(infile);
+    G = getc(infile);
+    B = getc(infile);
+    if (R == EOF || G == EOF || B == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    add_map_entry(cinfo,
+                  R << (BITS_IN_JSAMPLE-8),
+                  G << (BITS_IN_JSAMPLE-8),
+                  B << (BITS_IN_JSAMPLE-8));
+  }
+}
+
+
+/* Support routines for reading PPM */
+
+
+LOCAL(int)
+pbm_getc (FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_decompress_ptr cinfo, FILE *infile)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+  return val;
+}
+
+
+/*
+ * Extract color map from a PPM file.
+ */
+
+LOCAL(void)
+read_ppm_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  int c;
+  unsigned int w, h, maxval, row, col;
+  int R, G, B;
+
+  /* Initial 'P' has already been read by read_color_map */
+  c = getc(infile);             /* save format discriminator for a sec */
+
+  /* while we fetch the remaining header info */
+  w = read_pbm_integer(cinfo, infile);
+  h = read_pbm_integer(cinfo, infile);
+  maxval = read_pbm_integer(cinfo, infile);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* For now, we don't support rescaling from an unusual maxval. */
+  if (maxval != (unsigned int) MAXJSAMPLE)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  switch (c) {
+  case '3':                     /* it's a text-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = read_pbm_integer(cinfo, infile);
+        G = read_pbm_integer(cinfo, infile);
+        B = read_pbm_integer(cinfo, infile);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = getc(infile);
+        G = getc(infile);
+        B = getc(infile);
+        if (R == EOF || G == EOF || B == EOF)
+          ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+
+/*
+ * Main entry point from djpeg.c.
+ *  Input: opened input file (from file name argument on command line).
+ *  Output: colormap and actual_number_of_colors fields are set in cinfo.
+ */
+
+GLOBAL(void)
+read_color_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  /* Allocate space for a color map of maximum supported size. */
+  cinfo->colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (MAXJSAMPLE+1), (JDIMENSION) 3);
+  cinfo->actual_number_of_colors = 0; /* initialize map to empty */
+
+  /* Read first byte to determine file format */
+  switch (getc(infile)) {
+  case 'G':
+    read_gif_map(cinfo, infile);
+    break;
+  case 'P':
+    read_ppm_map(cinfo, infile);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdgif.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdgif.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdgif.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdgif.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,39 @@
+/*
+ * rdgif.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in GIF format.
+ *
+ *****************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,      *
+ * the ability to read GIF files has been removed from the IJG distribution. *
+ * Sorry about that.                                                         *
+ *****************************************************************************
+ *
+ * We are required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+/*
+ * The module selection routine for GIF format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_gif (j_compress_ptr cinfo)
+{
+  fprintf(stderr, "GIF input is unsupported for legal reasons.  Sorry.\n");
+  exit(EXIT_FAILURE);
+  return NULL;                  /* keep compiler happy */
+}
+
+#endif /* GIF_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdjpgcom.1 glmark2-2021.02/android/jni/src/libjpeg-turbo/rdjpgcom.1
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdjpgcom.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdjpgcom.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,63 @@
+.TH RDJPGCOM 1 "02 April 2009"
+.SH NAME
+rdjpgcom \- display text comments from a JPEG file
+.SH SYNOPSIS
+.B rdjpgcom
+[
+.B \-raw
+]
+[
+.B \-verbose
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B rdjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and prints any text comments found in the file on the standard output.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.SH OPTIONS
+.TP
+.B \-raw
+Normally
+.B rdjpgcom
+escapes non-printable characters in comments, for security reasons.
+This option avoids that.
+.PP
+.B \-verbose
+Causes
+.B rdjpgcom
+to also display the JPEG image dimensions.
+.PP
+Switch names may be abbreviated, and are not case sensitive.
+.SH HINTS
+.B rdjpgcom
+does not depend on the IJG JPEG library.  Its source code is intended as an
+illustration of the minimum amount of code required to parse a JPEG file
+header correctly.
+.PP
+In
+.B \-verbose
+mode,
+.B rdjpgcom
+will also attempt to print the contents of any "APP12" markers as text.
+Some digital cameras produce APP12 markers containing useful textual
+information.  If you like, you can modify the source code to print
+other APPn marker types as well.
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR wrjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdjpgcom.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdjpgcom.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdjpgcom.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdjpgcom.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,510 @@
+/*
+ * rdjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that displays
+ * the text in COM (comment) markers in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifdef HAVE_LOCALE_H
+#include <locale.h>             /* Bill Allombert: use locale for isprint */
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#else
+#define READ_BINARY     "rb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+
+/*
+ * These macros are used to read the input file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0            /* Start Of Frame N */
+#define M_SOF1  0xC1            /* N indicates which compression process */
+#define M_SOF2  0xC2            /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5            /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8            /* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9            /* End Of Image (end of datastream) */
+#define M_SOS   0xDA            /* Start Of Scan (begins compressed data) */
+#define M_APP0  0xE0            /* Application-specific marker, type N */
+#define M_APP12 0xEC            /* (we don't bother to list all 16 APPn's) */
+#define M_COM   0xFE            /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Process a COM marker.
+ * We want to print out the marker contents as legible text;
+ * we must guard against non-text junk and varying newline representations.
+ */
+
+static void
+process_COM (int raw)
+{
+  unsigned int length;
+  int ch;
+  int lastch = 0;
+
+  /* Bill Allombert: set locale properly for isprint */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "");
+#endif
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+
+  while (length > 0) {
+    ch = read_1_byte();
+    if (raw) {
+      putc(ch, stdout);
+    /* Emit the character in a readable form.
+     * Nonprintables are converted to \nnn form,
+     * while \ is converted to \\.
+     * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+     */
+    } else if (ch == '\r') {
+      printf("\n");
+    } else if (ch == '\n') {
+      if (lastch != '\r')
+        printf("\n");
+    } else if (ch == '\\') {
+      printf("\\\\");
+    } else if (isprint(ch)) {
+      putc(ch, stdout);
+    } else {
+      printf("\\%03o", ch);
+    }
+    lastch = ch;
+    length--;
+  }
+  printf("\n");
+
+  /* Bill Allombert: revert to C locale */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "C");
+#endif
+}
+
+
+/*
+ * Process a SOFn marker.
+ * This code is only needed if you want to know the image dimensions...
+ */
+
+static void
+process_SOFn (int marker)
+{
+  unsigned int length;
+  unsigned int image_height, image_width;
+  int data_precision, num_components;
+  const char *process;
+  int ci;
+
+  length = read_2_bytes();      /* usual parameter length count */
+
+  data_precision = read_1_byte();
+  image_height = read_2_bytes();
+  image_width = read_2_bytes();
+  num_components = read_1_byte();
+
+  switch (marker) {
+  case M_SOF0:  process = "Baseline";  break;
+  case M_SOF1:  process = "Extended sequential";  break;
+  case M_SOF2:  process = "Progressive";  break;
+  case M_SOF3:  process = "Lossless";  break;
+  case M_SOF5:  process = "Differential sequential";  break;
+  case M_SOF6:  process = "Differential progressive";  break;
+  case M_SOF7:  process = "Differential lossless";  break;
+  case M_SOF9:  process = "Extended sequential, arithmetic coding";  break;
+  case M_SOF10: process = "Progressive, arithmetic coding";  break;
+  case M_SOF11: process = "Lossless, arithmetic coding";  break;
+  case M_SOF13: process = "Differential sequential, arithmetic coding";  break;
+  case M_SOF14: process = "Differential progressive, arithmetic coding"; break;
+  case M_SOF15: process = "Differential lossless, arithmetic coding";  break;
+  default:      process = "Unknown";  break;
+  }
+
+  printf("JPEG image is %uw * %uh, %d color components, %d bits per sample\n",
+         image_width, image_height, num_components, data_precision);
+  printf("JPEG process: %s\n", process);
+
+  if (length != (unsigned int) (8 + num_components * 3))
+    ERREXIT("Bogus SOF marker length");
+
+  for (ci = 0; ci < num_components; ci++) {
+    (void) read_1_byte();       /* Component ID code */
+    (void) read_1_byte();       /* H, V sampling factors */
+    (void) read_1_byte();       /* Quantization table number */
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOS or EOI is seen;
+ * display any COM markers.
+ * While the companion program wrjpgcom will always insert COM markers before
+ * SOFn, other implementations might not, so we scan to SOS before stopping.
+ * If we were only interested in the image dimensions, we would stop at SOFn.
+ * (Conversely, if we only cared about COM markers, there would be no need
+ * for special code to handle SOFn; we could treat it like other markers.)
+ */
+
+static int
+scan_JPEG_header (int verbose, int raw)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+
+  /* Scan miscellaneous markers until we reach SOS. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      if (verbose)
+        process_SOFn(marker);
+      else
+        skip_variable();
+      break;
+
+    case M_SOS:                 /* stop before hitting compressed data */
+      return marker;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:
+      process_COM(raw);
+      break;
+
+    case M_APP12:
+      /* Some digital camera makers put useful textual information into
+       * APP12 markers, so we print those out too when in -verbose mode.
+       */
+      if (verbose) {
+        printf("APP12 contains:\n");
+        process_COM(raw);
+      } else
+        skip_variable();
+      break;
+
+    default:                    /* Anything else just gets skipped */
+      skip_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "rdjpgcom displays any textual comments in a JPEG file.\n");
+
+  fprintf(stderr, "Usage: %s [switches] [inputfile]\n", progname);
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -raw        Display non-printable characters in comments (unsafe)\n");
+  fprintf(stderr, "  -verbose    Also display dimensions of JPEG image\n");
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int verbose = 0, raw = 0;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "rdjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "verbose", 1)) {
+      verbose++;
+    } else if (keymatch(arg, "raw", 1)) {
+      raw = 1;
+    } else
+      usage();
+  }
+
+  /* Open the input file. */
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Scan the JPEG headers. */
+  (void) scan_JPEG_header(verbose, raw);
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdppm.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdppm.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdppm.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdppm.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,461 @@
+/*
+ * rdppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed PPM format).
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  U_CHAR *iobuffer;             /* fread's I/O buffer */
+  JSAMPROW pixrow;              /* compressor input buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JSAMPLE *rescale;             /* => maxval-remapping array, or NULL */
+  int maxval;
+} ppm_source_struct;
+
+typedef ppm_source_struct *ppm_source_ptr;
+
+
+LOCAL(int)
+pbm_getc (FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_compress_ptr cinfo, FILE *infile, unsigned int maxval)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_PPM_NONNUMERIC);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+
+  if (val > maxval)
+    ERREXIT(cinfo, JERR_PPM_TOOLARGE);
+
+  return val;
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ * In all cases, input is scaled to the size of JSAMPLE.
+ *
+ * A really fast path is provided for reading byte/sample raw files with
+ * maxval = MAXJSAMPLE, which is the normal case for 8-bit data.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_text_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_text_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_raw_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format files with maxval = MAXJSAMPLE.
+ * In this case we just read right into the JSAMPLE buffer!
+ * Note that same code works for PPM and PGM files.
+ */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  int c;
+  unsigned int w, h, maxval;
+  boolean need_iobuffer, use_raw_buffer, need_rescale;
+
+  if (getc(source->pub.input_file) != 'P')
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  c = getc(source->pub.input_file); /* subformat discriminator character */
+
+  /* detect unsupported variants (ie, PBM) before trying to read header */
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+  case '3':                     /* it's a text-format PPM file */
+  case '5':                     /* it's a raw-format PGM file */
+  case '6':                     /* it's a raw-format PPM file */
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_NOT);
+    break;
+  }
+
+  /* fetch the remaining header info */
+  w = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  h = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  maxval = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  cinfo->data_precision = BITS_IN_JSAMPLE; /* we always rescale data to this */
+  cinfo->image_width = (JDIMENSION) w;
+  cinfo->image_height = (JDIMENSION) h;
+  source->maxval = maxval;
+
+  /* initialize flags to most common settings */
+  need_iobuffer = TRUE;         /* do we need an I/O buffer? */
+  use_raw_buffer = FALSE;       /* do we map input buffer onto I/O buffer? */
+  need_rescale = TRUE;          /* do we need a rescale array? */
+
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_gray_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '3':                     /* it's a text-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_rgb_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '5':                     /* it's a raw-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_gray_row;
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_gray_row;
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_rgb_row;
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_rgb_row;
+    }
+    break;
+  }
+
+  /* Allocate space for I/O buffer: 1 or 3 bytes or words/pixel. */
+  if (need_iobuffer) {
+    source->buffer_width = (size_t) w * cinfo->input_components *
+      ((maxval <= 255) ? sizeof(U_CHAR) : (2 * sizeof(U_CHAR)));
+    source->iobuffer = (U_CHAR *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  source->buffer_width);
+  }
+
+  /* Create compressor input buffer. */
+  if (use_raw_buffer) {
+    /* For unscaled raw-input case, we can just map it onto the I/O buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    source->pixrow = (JSAMPROW) source->iobuffer;
+    source->pub.buffer = & source->pixrow;
+    source->pub.buffer_height = 1;
+  } else {
+    /* Need to translate anyway, so make a separate sample buffer. */
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) w * cinfo->input_components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+  }
+
+  /* Compute the rescaling array if required. */
+  if (need_rescale) {
+    long val, half_maxval;
+
+    /* On 16-bit-int machines we have to be careful of maxval = 65535 */
+    source->rescale = (JSAMPLE *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  (size_t) (((long) maxval + 1L) *
+                                            sizeof(JSAMPLE)));
+    half_maxval = maxval / 2;
+    for (val = 0; val <= (long) maxval; val++) {
+      /* The multiplication here must be done in 32 bits to avoid overflow */
+      source->rescale[val] = (JSAMPLE) ((val * MAXJSAMPLE + half_maxval) /
+                                        maxval);
+    }
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for PPM format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_ppm (j_compress_ptr cinfo)
+{
+  ppm_source_ptr source;
+
+  /* Create module interface object */
+  source = (ppm_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(ppm_source_struct));
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_ppm;
+  source->pub.finish_input = finish_input_ppm;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* PPM_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdrle.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdrle.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdrle.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdrle.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,389 @@
+/*
+ * rdrle.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Utah RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed RLE format).
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * We support the following types of RLE files:
+ *
+ *   GRAYSCALE   - 8 bits, no colormap
+ *   MAPPEDGRAY  - 8 bits, 1 channel colomap
+ *   PSEUDOCOLOR - 8 bits, 3 channel colormap
+ *   TRUECOLOR   - 24 bits, 3 channel colormap
+ *   DIRECTCOLOR - 24 bits, no colormap
+ *
+ * For now, we ignore any alpha channel in the image.
+ */
+
+typedef enum
+  { GRAYSCALE, MAPPEDGRAY, PSEUDOCOLOR, TRUECOLOR, DIRECTCOLOR } rle_kind;
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * to conform to JPEG's top-to-bottom order.  To do this, we read the
+ * incoming image into a virtual array on the first get_pixel_rows call,
+ * then fetch the required row from the virtual array on subsequent calls.
+ */
+
+typedef struct _rle_source_struct *rle_source_ptr;
+
+typedef struct _rle_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  rle_kind visual;              /* actual type of input file */
+  jvirt_sarray_ptr image;       /* virtual array to hold the image */
+  JDIMENSION row;               /* current row # in the virtual array */
+  rle_hdr header;               /* Input file information */
+  rle_pixel **rle_row;          /* holds a row returned by rle_getrow() */
+
+} rle_source_struct;
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION width, height;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Use RLE library routine to get the header info */
+  source->header = *rle_hdr_init(NULL);
+  source->header.rle_file = source->pub.input_file;
+  switch (rle_get_setup(&(source->header))) {
+  case RLE_SUCCESS:
+    /* A-OK */
+    break;
+  case RLE_NOT_RLE:
+    ERREXIT(cinfo, JERR_RLE_NOT);
+    break;
+  case RLE_NO_SPACE:
+    ERREXIT(cinfo, JERR_RLE_MEM);
+    break;
+  case RLE_EMPTY:
+    ERREXIT(cinfo, JERR_RLE_EMPTY);
+    break;
+  case RLE_EOF:
+    ERREXIT(cinfo, JERR_RLE_EOF);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_RLE_BADERROR);
+    break;
+  }
+
+  /* Figure out what we have, set private vars and return values accordingly */
+
+  width  = source->header.xmax - source->header.xmin + 1;
+  height = source->header.ymax - source->header.ymin + 1;
+  source->header.xmin = 0;              /* realign horizontally */
+  source->header.xmax = width-1;
+
+  cinfo->image_width      = width;
+  cinfo->image_height     = height;
+  cinfo->data_precision   = 8;  /* we can only handle 8 bit data */
+
+  if (source->header.ncolors == 1 && source->header.ncmap == 0) {
+    source->visual     = GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_RLE_GRAY, width, height);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 1) {
+    source->visual     = MAPPEDGRAY;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPGRAY, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 3) {
+    source->visual     = PSEUDOCOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPPED, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 3) {
+    source->visual     = TRUECOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_FULLMAP, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 0) {
+    source->visual     = DIRECTCOLOR;
+    TRACEMS2(cinfo, 1, JTRC_RLE, width, height);
+  } else
+    ERREXIT(cinfo, JERR_RLE_UNSUPPORTED);
+
+  if (source->visual == GRAYSCALE || source->visual == MAPPEDGRAY) {
+    cinfo->in_color_space   = JCS_GRAYSCALE;
+    cinfo->input_components = 1;
+  } else {
+    cinfo->in_color_space   = JCS_RGB;
+    cinfo->input_components = 3;
+  }
+
+  /*
+   * A place to hold each scanline while it's converted.
+   * (GRAYSCALE scanlines don't need converting)
+   */
+  if (source->visual != GRAYSCALE) {
+    source->rle_row = (rle_pixel**) (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width, (JDIMENSION) cinfo->input_components);
+  }
+
+  /* request a virtual array to hold the image */
+  source->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (width * source->header.ncolors),
+     (JDIMENSION) height, (JDIMENSION) 1);
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    /* count file input as separate pass */
+    progress->total_extra_passes++;
+  }
+#endif
+
+  source->pub.buffer_height = 1;
+}
+
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for GRAYSCALE, MAPPEDGRAY, TRUECOLOR, and DIRECTCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_rle_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+
+  source->row--;
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  return 1;
+}
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for PSEUDOCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_pseudocolor_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JSAMPROW src_row, dest_row;
+  JDIMENSION col;
+  rle_map *colormap;
+  int val;
+
+  colormap = source->header.cmap;
+  dest_row = source->pub.buffer[0];
+  source->row--;
+  src_row = *(*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  for (col = cinfo->image_width; col > 0; col--) {
+    val = GETJSAMPLE(*src_row++);
+    *dest_row++ = (JSAMPLE) (colormap[val      ] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 256] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 512] >> 8);
+  }
+
+  return 1;
+}
+
+
+/*
+ * Load the image into a virtual array.  We have to do this because RLE
+ * files start at the lower left while the JPEG standard has them starting
+ * in the upper left.  This is called the first time we want to get a row
+ * of input.  What we do is load the RLE data into the array and then call
+ * the appropriate routine to read one row from the array.  Before returning,
+ * we set source->pub.get_pixel_rows so that subsequent calls go straight to
+ * the appropriate row-reading routine.
+ */
+
+METHODDEF(JDIMENSION)
+load_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION row, col;
+  JSAMPROW  scanline, red_ptr, green_ptr, blue_ptr;
+  rle_pixel **rle_row;
+  rle_map *colormap;
+  char channel;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  colormap = source->header.cmap;
+  rle_row = source->rle_row;
+
+  /* Read the RLE data into our virtual array.
+   * We assume here that rle_pixel is represented the same as JSAMPLE.
+   */
+  RLE_CLR_BIT(source->header, RLE_ALPHA); /* don't read the alpha channel */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->image_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  switch (source->visual) {
+
+  case GRAYSCALE:
+  case PSEUDOCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+         ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case MAPPEDGRAY:
+  case TRUECOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_row = source->rle_row;
+      rle_getrow(&source->header, rle_row);
+
+      for (col = 0; col < cinfo->image_width; col++) {
+        for (channel = 0; channel < source->header.ncolors; channel++) {
+          *scanline++ = (JSAMPLE)
+            (colormap[GETJSAMPLE(rle_row[channel][col]) + 256 * channel] >> 8);
+        }
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case DIRECTCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+
+      red_ptr   = rle_row[0];
+      green_ptr = rle_row[1];
+      blue_ptr  = rle_row[2];
+
+      for (col = cinfo->image_width; col > 0; col--) {
+        *scanline++ = *red_ptr++;
+        *scanline++ = *green_ptr++;
+        *scanline++ = *blue_ptr++;
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Set up to call proper row-extraction routine in future */
+  if (source->visual == PSEUDOCOLOR) {
+    source->pub.buffer = source->rle_row;
+    source->pub.get_pixel_rows = get_pseudocolor_row;
+  } else {
+    source->pub.get_pixel_rows = get_rle_row;
+  }
+  source->row = cinfo->image_height;
+
+  /* And fetch the topmost (bottommost) row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for RLE format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_rle (j_compress_ptr cinfo)
+{
+  rle_source_ptr source;
+
+  /* Create module interface object */
+  source = (rle_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(rle_source_struct));
+  /* Fill in method ptrs */
+  source->pub.start_input = start_input_rle;
+  source->pub.finish_input = finish_input_rle;
+  source->pub.get_pixel_rows = load_image;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* RLE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdswitch.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdswitch.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdswitch.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdswitch.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,424 @@
+/*
+ * rdswitch.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to process some of cjpeg's more complicated
+ * command-line switches.  Switches processed here are:
+ *      -qtables file           Read quantization tables from text file
+ *      -scans file             Read scan script from text file
+ *      -quality N[,N,...]      Set quality ratings
+ *      -qslots N[,N,...]       Set component quantization table selectors
+ *      -sample HxV[,HxV,...]   Set component sampling factors
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isdigit(), isspace() */
+
+
+LOCAL(int)
+text_getc (FILE *file)
+/* Read next char, skipping over any comments (# to end of line) */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(file);
+  if (ch == '#') {
+    do {
+      ch = getc(file);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(boolean)
+read_text_integer (FILE *file, long *result, int *termchar)
+/* Read an unsigned decimal integer from a file, store it in result */
+/* Reads one trailing character after the integer; returns it in termchar */
+{
+  register int ch;
+  register long val;
+
+  /* Skip any leading whitespace, detect EOF */
+  do {
+    ch = text_getc(file);
+    if (ch == EOF) {
+      *termchar = ch;
+      return FALSE;
+    }
+  } while (isspace(ch));
+
+  if (! isdigit(ch)) {
+    *termchar = ch;
+    return FALSE;
+  }
+
+  val = ch - '0';
+  while ((ch = text_getc(file)) != EOF) {
+    if (! isdigit(ch))
+      break;
+    val *= 10;
+    val += ch - '0';
+  }
+  *result = val;
+  *termchar = ch;
+  return TRUE;
+}
+
+
+#if JPEG_LIB_VERSION < 70
+static int q_scale_factor[NUM_QUANT_TBLS] = {100, 100, 100, 100};
+#endif
+
+GLOBAL(boolean)
+read_quant_tables (j_compress_ptr cinfo, char *filename,
+                   boolean force_baseline)
+/* Read a set of quantization tables from the specified file.
+ * The file is plain ASCII text: decimal numbers with whitespace between.
+ * Comments preceded by '#' may be included in the file.
+ * There may be one to NUM_QUANT_TBLS tables in the file, each of 64 values.
+ * The tables are implicitly numbered 0,1,etc.
+ * NOTE: does not affect the qslots mapping, which will default to selecting
+ * table 0 for luminance (or primary) components, 1 for chrominance components.
+ * You must use -qslots if you want a different component->table mapping.
+ */
+{
+  FILE *fp;
+  int tblno, i, termchar;
+  long val;
+  unsigned int table[DCTSIZE2];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open table file %s\n", filename);
+    return FALSE;
+  }
+  tblno = 0;
+
+  while (read_text_integer(fp, &val, &termchar)) { /* read 1st element of table */
+    if (tblno >= NUM_QUANT_TBLS) {
+      fprintf(stderr, "Too many tables in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    table[0] = (unsigned int) val;
+    for (i = 1; i < DCTSIZE2; i++) {
+      if (! read_text_integer(fp, &val, &termchar)) {
+        fprintf(stderr, "Invalid table data in file %s\n", filename);
+        fclose(fp);
+        return FALSE;
+      }
+      table[i] = (unsigned int) val;
+    }
+#if JPEG_LIB_VERSION >= 70
+    jpeg_add_quant_table(cinfo, tblno, table, cinfo->q_scale_factor[tblno],
+                         force_baseline);
+#else
+    jpeg_add_quant_table(cinfo, tblno, table, q_scale_factor[tblno],
+                         force_baseline);
+#endif
+    tblno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(boolean)
+read_scan_integer (FILE *file, long *result, int *termchar)
+/* Variant of read_text_integer that always looks for a non-space termchar;
+ * this simplifies parsing of punctuation in scan scripts.
+ */
+{
+  register int ch;
+
+  if (! read_text_integer(file, result, termchar))
+    return FALSE;
+  ch = *termchar;
+  while (ch != EOF && isspace(ch))
+    ch = text_getc(file);
+  if (isdigit(ch)) {            /* oops, put it back */
+    if (ungetc(ch, file) == EOF)
+      return FALSE;
+    ch = ' ';
+  } else {
+    /* Any separators other than ';' and ':' are ignored;
+     * this allows user to insert commas, etc, if desired.
+     */
+    if (ch != EOF && ch != ';' && ch != ':')
+      ch = ' ';
+  }
+  *termchar = ch;
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+read_scan_script (j_compress_ptr cinfo, char *filename)
+/* Read a scan script from the specified text file.
+ * Each entry in the file defines one scan to be emitted.
+ * Entries are separated by semicolons ';'.
+ * An entry contains one to four component indexes,
+ * optionally followed by a colon ':' and four progressive-JPEG parameters.
+ * The component indexes denote which component(s) are to be transmitted
+ * in the current scan.  The first component has index 0.
+ * Sequential JPEG is used if the progressive-JPEG parameters are omitted.
+ * The file is free format text: any whitespace may appear between numbers
+ * and the ':' and ';' punctuation marks.  Also, other punctuation (such
+ * as commas or dashes) can be placed between numbers if desired.
+ * Comments preceded by '#' may be included in the file.
+ * Note: we do very little validity checking here;
+ * jcmaster.c will validate the script parameters.
+ */
+{
+  FILE *fp;
+  int scanno, ncomps, termchar;
+  long val;
+  jpeg_scan_info *scanptr;
+#define MAX_SCANS  100          /* quite arbitrary limit */
+  jpeg_scan_info scans[MAX_SCANS];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open scan definition file %s\n", filename);
+    return FALSE;
+  }
+  scanptr = scans;
+  scanno = 0;
+
+  while (read_scan_integer(fp, &val, &termchar)) {
+    if (scanno >= MAX_SCANS) {
+      fprintf(stderr, "Too many scans defined in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr->component_index[0] = (int) val;
+    ncomps = 1;
+    while (termchar == ' ') {
+      if (ncomps >= MAX_COMPS_IN_SCAN) {
+        fprintf(stderr, "Too many components in one scan in file %s\n",
+                filename);
+        fclose(fp);
+        return FALSE;
+      }
+      if (! read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->component_index[ncomps] = (int) val;
+      ncomps++;
+    }
+    scanptr->comps_in_scan = ncomps;
+    if (termchar == ':') {
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ss = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Se = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ah = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->Al = (int) val;
+    } else {
+      /* set non-progressive parameters */
+      scanptr->Ss = 0;
+      scanptr->Se = DCTSIZE2-1;
+      scanptr->Ah = 0;
+      scanptr->Al = 0;
+    }
+    if (termchar != ';' && termchar != EOF) {
+bogus:
+      fprintf(stderr, "Invalid scan entry format in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr++, scanno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  if (scanno > 0) {
+    /* Stash completed scan list in cinfo structure.
+     * NOTE: for cjpeg's use, JPOOL_IMAGE is the right lifetime for this data,
+     * but if you want to compress multiple images you'd want JPOOL_PERMANENT.
+     */
+    scanptr = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  scanno * sizeof(jpeg_scan_info));
+    MEMCOPY(scanptr, scans, scanno * sizeof(jpeg_scan_info));
+    cinfo->scan_info = scanptr;
+    cinfo->num_scans = scanno;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+#if JPEG_LIB_VERSION < 70
+/* These are the sample quantization tables given in JPEG spec section K.1.
+ * The spec says that the values given produce "good" quality, and
+ * when divided by 2, "very good" quality.
+ */
+static const unsigned int std_luminance_quant_tbl[DCTSIZE2] = {
+  16,  11,  10,  16,  24,  40,  51,  61,
+  12,  12,  14,  19,  26,  58,  60,  55,
+  14,  13,  16,  24,  40,  57,  69,  56,
+  14,  17,  22,  29,  51,  87,  80,  62,
+  18,  22,  37,  56,  68, 109, 103,  77,
+  24,  35,  55,  64,  81, 104, 113,  92,
+  49,  64,  78,  87, 103, 121, 120, 101,
+  72,  92,  95,  98, 112, 100, 103,  99
+};
+static const unsigned int std_chrominance_quant_tbl[DCTSIZE2] = {
+  17,  18,  24,  47,  99,  99,  99,  99,
+  18,  21,  26,  66,  99,  99,  99,  99,
+  24,  26,  56,  99,  99,  99,  99,  99,
+  47,  66,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99
+};
+
+
+LOCAL(void)
+jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
+{
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+                       q_scale_factor[0], force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+                       q_scale_factor[1], force_baseline);
+}
+#endif
+
+
+GLOBAL(boolean)
+set_quality_ratings (j_compress_ptr cinfo, char *arg, boolean force_baseline)
+/* Process a quality-ratings parameter string, of the form
+ *     N[,N,...]
+ * If there are more q-table slots than parameters, the last value is replicated.
+ */
+{
+  int val = 75;                 /* default value */
+  int tblno;
+  char ch;
+
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      /* Convert user 0-100 rating to percentage scaling */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#else
+      q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#endif
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining factors to last value */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#else
+      q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#endif
+    }
+  }
+  jpeg_default_qtables(cinfo, force_baseline);
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_quant_slots (j_compress_ptr cinfo, char *arg)
+/* Process a quantization-table-selectors parameter string, of the form
+ *     N[,N,...]
+ * If there are more components than parameters, the last value is replicated.
+ */
+{
+  int val = 0;                  /* default table # */
+  int ci;
+  char ch;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      if (val < 0 || val >= NUM_QUANT_TBLS) {
+        fprintf(stderr, "JPEG quantization tables are numbered 0..%d\n",
+                NUM_QUANT_TBLS-1);
+        return FALSE;
+      }
+      cinfo->comp_info[ci].quant_tbl_no = val;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining components to last table */
+      cinfo->comp_info[ci].quant_tbl_no = val;
+    }
+  }
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_sample_factors (j_compress_ptr cinfo, char *arg)
+/* Process a sample-factors parameter string, of the form
+ *     HxV[,HxV,...]
+ * If there are more components than parameters, "1x1" is assumed for the rest.
+ */
+{
+  int ci, val1, val2;
+  char ch1, ch2;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch2 = ',';                /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c%d%c", &val1, &ch1, &val2, &ch2) < 3)
+        return FALSE;
+      if ((ch1 != 'x' && ch1 != 'X') || ch2 != ',') /* syntax check */
+        return FALSE;
+      if (val1 <= 0 || val1 > 4 || val2 <= 0 || val2 > 4) {
+        fprintf(stderr, "JPEG sampling factors must be 1..4\n");
+        return FALSE;
+      }
+      cinfo->comp_info[ci].h_samp_factor = val1;
+      cinfo->comp_info[ci].v_samp_factor = val2;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining components to 1x1 sampling */
+      cinfo->comp_info[ci].h_samp_factor = 1;
+      cinfo->comp_info[ci].v_samp_factor = 1;
+    }
+  }
+  return TRUE;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdtarga.c glmark2-2021.02/android/jni/src/libjpeg-turbo/rdtarga.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/rdtarga.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/rdtarga.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,503 @@
+/*
+ * rdtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed Targa format).
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _tga_source_struct *tga_source_ptr;
+
+typedef struct _tga_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* Targa colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed if funny input row order */
+  JDIMENSION current_row;       /* Current logical row number to read */
+
+  /* Pointer to routine to extract next Targa pixel from input file */
+  void (*read_pixel) (tga_source_ptr sinfo);
+
+  /* Result of read_pixel is delivered here: */
+  U_CHAR tga_pixel[4];
+
+  int pixel_size;               /* Bytes per Targa pixel (1 to 4) */
+
+  /* State info for reading RLE-coded pixels; both counts must be init to 0 */
+  int block_count;              /* # of pixels remaining in RLE block */
+  int dup_pixel_count;          /* # of times to duplicate previous pixel */
+
+  /* This saves the correct pixel-row-expansion method for preload_image */
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+} tga_source_struct;
+
+
+/* For expanding 5-bit pixel values to 8-bit with best rounding */
+
+static const UINT8 c5to8bits[32] = {
+    0,   8,  16,  25,  33,  41,  49,  58,
+   66,  74,  82,  90,  99, 107, 115, 123,
+  132, 140, 148, 156, 165, 173, 181, 189,
+  197, 206, 214, 222, 230, 239, 247, 255
+};
+
+
+
+LOCAL(int)
+read_byte (tga_source_ptr sinfo)
+/* Read next byte from Targa file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (tga_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a Targa file */
+{
+  int i;
+
+  /* Presently only handles 24-bit BGR format */
+  if (mapentrysize != 24)
+    ERREXIT(sinfo->cinfo, JERR_TGA_BADCMAP);
+
+  for (i = 0; i < cmaplen; i++) {
+    sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+  }
+}
+
+
+/*
+ * read_pixel methods: get a single pixel from Targa file into tga_pixel[]
+ */
+
+METHODDEF(void)
+read_non_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file; no RLE expansion */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+METHODDEF(void)
+read_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file, expanding RLE data as needed */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  /* Duplicate previously read pixel? */
+  if (sinfo->dup_pixel_count > 0) {
+    sinfo->dup_pixel_count--;
+    return;
+  }
+
+  /* Time to read RLE block header? */
+  if (--sinfo->block_count < 0) { /* decrement pixels remaining in block */
+    i = read_byte(sinfo);
+    if (i & 0x80) {             /* Start of duplicate-pixel block? */
+      sinfo->dup_pixel_count = i & 0x7F; /* number of dups after this one */
+      sinfo->block_count = 0;   /* then read new block header */
+    } else {
+      sinfo->block_count = i & 0x7F; /* number of pixels after this one */
+    }
+  }
+
+  /* Read next pixel */
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_8bit_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit grayscale pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    *ptr++ = colormap[0][t];
+    *ptr++ = colormap[1][t];
+    *ptr++ = colormap[2][t];
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_16bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 16-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    t += UCH(source->tga_pixel[1]) << 8;
+    /* We expand 5 bit data to 8 bit sample width.
+     * The format of the 16-bit (LSB first) input word is
+     *     xRRRRRGGGGGBBBBB
+     */
+    ptr[2] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[1] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[0] = (JSAMPLE) c5to8bits[t & 0x1F];
+    ptr += 3;
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[2]); /* change BGR to RGB order */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[1]);
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+/*
+ * Targa also defines a 32-bit pixel format with order B,G,R,A.
+ * We presently ignore the attribute byte, so the code for reading
+ * these pixels is identical to the 24-bit routine above.
+ * This works because the actual pixel length is only known to read_pixel.
+ */
+
+#define get_32bit_row  get_24bit_row
+
+
+/*
+ * This method is for re-reading the input data in standard top-down
+ * row order.  The entire image has already been read into whole_image
+ * with proper conversion of pixel format, but it's in a funny row order.
+ */
+
+METHODDEF(JDIMENSION)
+get_memory_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION source_row;
+
+  /* Compute row of source that maps to current_row of normal order */
+  /* For now, assume image is bottom-up and not interlaced. */
+  /* NEEDS WORK to support interlaced images! */
+  source_row = cinfo->image_height - source->current_row - 1;
+
+  /* Fetch that row from virtual array */
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source_row, (JDIMENSION) 1, FALSE);
+
+  source->current_row++;
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_memory_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image, row, (JDIMENSION) 1, TRUE);
+    (*source->get_pixel_rows) (cinfo, sinfo);
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in unscrambled order */
+  source->pub.get_pixel_rows = get_memory_row;
+  source->current_row = 0;
+  /* And read the first row */
+  return get_memory_row(cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  U_CHAR targaheader[18];
+  int idlen, cmaptype, subtype, flags, interlace_type, components;
+  unsigned int width, height, maplen;
+  boolean is_bottom_up;
+
+#define GET_2B(offset)  ((unsigned int) UCH(targaheader[offset]) + \
+                         (((unsigned int) UCH(targaheader[offset+1])) << 8))
+
+  if (! ReadOK(source->pub.input_file, targaheader, 18))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  /* Pretend "15-bit" pixels are 16-bit --- we ignore attribute bit anyway */
+  if (targaheader[16] == 15)
+    targaheader[16] = 16;
+
+  idlen = UCH(targaheader[0]);
+  cmaptype = UCH(targaheader[1]);
+  subtype = UCH(targaheader[2]);
+  maplen = GET_2B(5);
+  width = GET_2B(12);
+  height = GET_2B(14);
+  source->pixel_size = UCH(targaheader[16]) >> 3;
+  flags = UCH(targaheader[17]); /* Image Descriptor byte */
+
+  is_bottom_up = ((flags & 0x20) == 0); /* bit 5 set => top-down */
+  interlace_type = flags >> 6;  /* bits 6/7 are interlace code */
+
+  if (cmaptype > 1 ||           /* cmaptype must be 0 or 1 */
+      source->pixel_size < 1 || source->pixel_size > 4 ||
+      (UCH(targaheader[16]) & 7) != 0 || /* bits/pixel must be multiple of 8 */
+      interlace_type != 0 ||      /* currently don't allow interlaced image */
+      width == 0 || height == 0)  /* image width/height must be non-zero */
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+
+  if (subtype > 8) {
+    /* It's an RLE-coded file */
+    source->read_pixel = read_rle_pixel;
+    source->block_count = source->dup_pixel_count = 0;
+    subtype -= 8;
+  } else {
+    /* Non-RLE file */
+    source->read_pixel = read_non_rle_pixel;
+  }
+
+  /* Now should have subtype 1, 2, or 3 */
+  components = 3;               /* until proven different */
+  cinfo->in_color_space = JCS_RGB;
+
+  switch (subtype) {
+  case 1:                       /* Colormapped image */
+    if (source->pixel_size == 1 && cmaptype == 1)
+      source->get_pixel_rows = get_8bit_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_MAPPED, width, height);
+    break;
+  case 2:                       /* RGB image */
+    switch (source->pixel_size) {
+    case 2:
+      source->get_pixel_rows = get_16bit_row;
+      break;
+    case 3:
+      source->get_pixel_rows = get_24bit_row;
+      break;
+    case 4:
+      source->get_pixel_rows = get_32bit_row;
+      break;
+    default:
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+      break;
+    }
+    TRACEMS2(cinfo, 1, JTRC_TGA, width, height);
+    break;
+  case 3:                       /* Grayscale image */
+    components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    if (source->pixel_size == 1)
+      source->get_pixel_rows = get_8bit_gray_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_GRAY, width, height);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    break;
+  }
+
+  if (is_bottom_up) {
+    /* Create a virtual array to buffer the upside-down image. */
+    source->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+       (JDIMENSION) width * components, (JDIMENSION) height, (JDIMENSION) 1);
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+    /* source->pub.buffer will point to the virtual array. */
+    source->pub.buffer_height = 1; /* in case anyone looks at it */
+    source->pub.get_pixel_rows = preload_image;
+  } else {
+    /* Don't need a virtual array, but do need a one-row input buffer. */
+    source->whole_image = NULL;
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width * components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+    source->pub.get_pixel_rows = source->get_pixel_rows;
+  }
+
+  while (idlen--)               /* Throw away ID field */
+    (void) read_byte(source);
+
+  if (maplen > 0) {
+    if (maplen > 256 || GET_2B(3) != 0)
+      ERREXIT(cinfo, JERR_TGA_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, (JDIMENSION) maplen, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) maplen, UCH(targaheader[7]));
+  } else {
+    if (cmaptype)               /* but you promised a cmap! */
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    source->colormap = NULL;
+  }
+
+  cinfo->input_components = components;
+  cinfo->data_precision = 8;
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for Targa format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_targa (j_compress_ptr cinfo)
+{
+  tga_source_ptr source;
+
+  /* Create module interface object */
+  source = (tga_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(tga_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_tga;
+  source->pub.finish_input = finish_input_tga;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* TARGA_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README glmark2-2021.02/android/jni/src/libjpeg-turbo/README
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/README	1970-01-01 08:00:00.000000000 +0800
@@ -1,290 +0,0 @@
-libjpeg-turbo note:  This file contains portions of the libjpeg v6b and v8
-README files, with additional wordsmithing by The libjpeg-turbo Project.
-It is included only for reference, as some parts of it may not apply to
-libjpeg-turbo.  Please see README-turbo.txt for information specific to
-libjpeg-turbo.
-
-
-The Independent JPEG Group's JPEG software
-==========================================
-
-This distribution contains a release of the Independent JPEG Group's free JPEG
-software.  You are welcome to redistribute this software and to use it for any
-purpose, subject to the conditions under LEGAL ISSUES, below.
-
-This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
-Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
-Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
-and other members of the Independent JPEG Group.
-
-IJG is not affiliated with the official ISO JPEG standards committee.
-
-
-DOCUMENTATION ROADMAP
-=====================
-
-This file contains the following sections:
-
-OVERVIEW            General description of JPEG and the IJG software.
-LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
-REFERENCES          Where to learn more about JPEG.
-ARCHIVE LOCATIONS   Where to find newer versions of this software.
-FILE FORMAT WARS    Software *not* to get.
-TO DO               Plans for future IJG releases.
-
-Other documentation files in the distribution are:
-
-User documentation:
-  install.txt       How to configure and install the IJG software.
-  usage.txt         Usage instructions for cjpeg, djpeg, jpegtran,
-                    rdjpgcom, and wrjpgcom.
-  *.1               Unix-style man pages for programs (same info as usage.txt).
-  wizard.txt        Advanced usage instructions for JPEG wizards only.
-  change.log        Version-to-version change highlights.
-Programmer and internal documentation:
-  libjpeg.txt       How to use the JPEG library in your own programs.
-  example.c         Sample code for calling the JPEG library.
-  structure.txt     Overview of the JPEG library's internal structure.
-  filelist.txt      Road map of IJG files.
-  coderules.txt     Coding style rules --- please read if you contribute code.
-
-Please read at least the files install.txt and usage.txt.  Some information
-can also be found in the JPEG FAQ (Frequently Asked Questions) article.  See
-ARCHIVE LOCATIONS below to find out where to obtain the FAQ article.
-
-If you want to understand how the JPEG code works, we suggest reading one or
-more of the REFERENCES, then looking at the documentation files (in roughly
-the order listed) before diving into the code.
-
-
-OVERVIEW
-========
-
-This package contains C software to implement JPEG image encoding, decoding,
-and transcoding.  JPEG (pronounced "jay-peg") is a standardized compression
-method for full-color and gray-scale images.  JPEG's strong suit is compressing
-photographic images or other types of images that have smooth color and
-brightness transitions between neighboring pixels.  Images with sharp lines or
-other abrupt features may not compress well with JPEG, and a higher JPEG
-quality may have to be used to avoid visible compression artifacts with such
-images.
-
-JPEG is lossy, meaning that the output pixels are not necessarily identical to
-the input pixels.  However, on photographic content and other "smooth" images,
-very good compression ratios can be obtained with no visible compression
-artifacts, and extremely high compression ratios are possible if you are
-willing to sacrifice image quality (by reducing the "quality" setting in the
-compressor.)
-
-This software implements JPEG baseline, extended-sequential, and progressive
-compression processes.  Provision is made for supporting all variants of these
-processes, although some uncommon parameter settings aren't implemented yet.
-We have made no provision for supporting the hierarchical or lossless
-processes defined in the standard.
-
-We provide a set of library routines for reading and writing JPEG image files,
-plus two sample applications "cjpeg" and "djpeg", which use the library to
-perform conversion between JPEG and some other popular image file formats.
-The library is intended to be reused in other applications.
-
-In order to support file conversion and viewing software, we have included
-considerable functionality beyond the bare JPEG coding/decoding capability;
-for example, the color quantization modules are not strictly part of JPEG
-decoding, but they are essential for output to colormapped file formats or
-colormapped displays.  These extra functions can be compiled out of the
-library if not required for a particular application.
-
-We have also included "jpegtran", a utility for lossless transcoding between
-different JPEG processes, and "rdjpgcom" and "wrjpgcom", two simple
-applications for inserting and extracting textual comments in JFIF files.
-
-The emphasis in designing this software has been on achieving portability and
-flexibility, while also making it fast enough to be useful.  In particular,
-the software is not intended to be read as a tutorial on JPEG.  (See the
-REFERENCES section for introductory material.)  Rather, it is intended to
-be reliable, portable, industrial-strength code.  We do not claim to have
-achieved that goal in every aspect of the software, but we strive for it.
-
-We welcome the use of this software as a component of commercial products.
-No royalty is required, but we do ask for an acknowledgement in product
-documentation, as described under LEGAL ISSUES.
-
-
-LEGAL ISSUES
-============
-
-In plain English:
-
-1. We don't promise that this software works.  (But if you find any bugs,
-   please let us know!)
-2. You can use this software for whatever you want.  You don't have to pay us.
-3. You may not pretend that you wrote this software.  If you use it in a
-   program, you must acknowledge somewhere in your documentation that
-   you've used the IJG code.
-
-In legalese:
-
-The authors make NO WARRANTY or representation, either express or implied,
-with respect to this software, its quality, accuracy, merchantability, or
-fitness for a particular purpose.  This software is provided "AS IS", and you,
-its user, assume the entire risk as to its quality and accuracy.
-
-This software is copyright (C) 1991-2010, Thomas G. Lane, Guido Vollbeding.
-All Rights Reserved except as specified below.
-
-Permission is hereby granted to use, copy, modify, and distribute this
-software (or portions thereof) for any purpose, without fee, subject to these
-conditions:
-(1) If any part of the source code for this software is distributed, then this
-README file must be included, with this copyright and no-warranty notice
-unaltered; and any additions, deletions, or changes to the original files
-must be clearly indicated in accompanying documentation.
-(2) If only executable code is distributed, then the accompanying
-documentation must state that "this software is based in part on the work of
-the Independent JPEG Group".
-(3) Permission for use of this software is granted only if the user accepts
-full responsibility for any undesirable consequences; the authors accept
-NO LIABILITY for damages of any kind.
-
-These conditions apply to any software derived from or based on the IJG code,
-not just to the unmodified library.  If you use our work, you ought to
-acknowledge us.
-
-Permission is NOT granted for the use of any IJG author's name or company name
-in advertising or publicity relating to this software or products derived from
-it.  This software may be referred to only as "the Independent JPEG Group's
-software".
-
-We specifically permit and encourage the use of this software as the basis of
-commercial products, provided that all warranty or liability claims are
-assumed by the product vendor.
-
-
-ansi2knr.c is included in this distribution by permission of L. Peter Deutsch,
-sole proprietor of its copyright holder, Aladdin Enterprises of Menlo Park, CA.
-ansi2knr.c is NOT covered by the above copyright and conditions, but instead
-by the usual distribution terms of the Free Software Foundation; principally,
-that you must include source code if you redistribute it.  (See the file
-ansi2knr.c for full details.)  However, since ansi2knr.c is not needed as part
-of any program generated from the IJG code, this does not limit you more than
-the foregoing paragraphs do.
-
-The Unix configuration script "configure" was produced with GNU Autoconf.
-It is copyright by the Free Software Foundation but is freely distributable.
-The same holds for its supporting scripts (config.guess, config.sub,
-ltmain.sh).  Another support script, install-sh, is copyright by X Consortium
-but is also freely distributable.
-
-The IJG distribution formerly included code to read and write GIF files.
-To avoid entanglement with the Unisys LZW patent, GIF reading support has
-been removed altogether, and the GIF writer has been simplified to produce
-"uncompressed GIFs".  This technique does not use the LZW algorithm; the
-resulting GIF files are larger than usual, but are readable by all standard
-GIF decoders.
-
-We are required to state that
-    "The Graphics Interchange Format(c) is the Copyright property of
-    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
-    CompuServe Incorporated."
-
-
-REFERENCES
-==========
-
-We recommend reading one or more of these references before trying to
-understand the innards of the JPEG software.
-
-The best short technical introduction to the JPEG compression algorithm is
-	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
-	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
-(Adjacent articles in that issue discuss MPEG motion picture compression,
-applications of JPEG, and related topics.)  If you don't have the CACM issue
-handy, a PostScript file containing a revised version of Wallace's article is
-available at http://www.ijg.org/files/wallace.ps.gz.  The file (actually
-a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
-omits the sample images that appeared in CACM, but it includes corrections
-and some added material.  Note: the Wallace article is copyright ACM and IEEE,
-and it may not be used for commercial purposes.
-
-A somewhat less technical, more leisurely introduction to JPEG can be found in
-"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
-M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
-good explanations and example C code for a multitude of compression methods
-including JPEG.  It is an excellent source if you are comfortable reading C
-code but don't know much about data compression in general.  The book's JPEG
-sample code is far from industrial-strength, but when you are ready to look
-at a full implementation, you've got one here...
-
-The best currently available description of JPEG is the textbook "JPEG Still
-Image Data Compression Standard" by William B. Pennebaker and Joan L.
-Mitchell, published by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
-Price US$59.95, 638 pp.  The book includes the complete text of the ISO JPEG
-standards (DIS 10918-1 and draft DIS 10918-2).
-
-The original JPEG standard is divided into two parts, Part 1 being the actual
-specification, while Part 2 covers compliance testing methods.  Part 1 is
-titled "Digital Compression and Coding of Continuous-tone Still Images,
-Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
-10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
-Continuous-tone Still Images, Part 2: Compliance testing" and has document
-numbers ISO/IEC IS 10918-2, ITU-T T.83.
-
-The JPEG standard does not specify all details of an interchangeable file
-format.  For the omitted details we follow the "JFIF" conventions, revision
-1.02.  JFIF 1.02 has been adopted as an Ecma International Technical Report
-and thus received a formal publication status.  It is available as a free
-download in PDF format from
-http://www.ecma-international.org/publications/techreports/E-TR-098.htm.
-A PostScript version of the JFIF document is available at
-http://www.ijg.org/files/jfif.ps.gz.  There is also a plain text version at
-http://www.ijg.org/files/jfif.txt.gz, but it is missing the figures.
-
-The TIFF 6.0 file format specification can be obtained by FTP from
-ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
-found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
-IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
-Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
-(Compression tag 7).  Copies of this Note can be obtained from
-http://www.ijg.org/files/.  It is expected that the next revision
-of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
-Although IJG's own code does not support TIFF/JPEG, the free libtiff library
-uses our library to implement TIFF/JPEG per the Note.
-
-
-ARCHIVE LOCATIONS
-=================
-
-The "official" archive site for this software is www.ijg.org.
-The most recent released version can always be found there in
-directory "files".  This particular version will be archived as
-http://www.ijg.org/files/jpegsrc.v8d.tar.gz, and in Windows-compatible
-"zip" archive format as http://www.ijg.org/files/jpegsr8d.zip.
-
-The JPEG FAQ (Frequently Asked Questions) article is a source of some
-general information about JPEG.
-It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
-and other news.answers archive sites, including the official news.answers
-archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
-If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
-with body
-	send usenet/news.answers/jpeg-faq/part1
-	send usenet/news.answers/jpeg-faq/part2
-
-
-FILE FORMAT WARS
-================
-
-The ISO JPEG standards committee actually promotes different formats like
-"JPEG 2000" or "JPEG XR", which are incompatible with original DCT-based
-JPEG.  IJG therefore does not support these formats (see REFERENCES).  Indeed,
-one of the original reasons for developing this free software was to help
-force convergence on common, interoperable format standards for JPEG files.
-Don't use an incompatible file format!
-(In any case, our decoder will remain capable of reading existing JPEG
-image files indefinitely.)
-
-
-TO DO
-=====
-
-Please send bug reports, offers of help, etc. to jpeg-info@jpegclub.org.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README.ijg glmark2-2021.02/android/jni/src/libjpeg-turbo/README.ijg
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README.ijg	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/README.ijg	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,279 @@
+libjpeg-turbo note:  This file has been modified by The libjpeg-turbo Project
+to include only information relevant to libjpeg-turbo, to wordsmith certain
+sections, and to remove impolitic language that existed in the libjpeg v8
+README.  It is included only for reference.  Please see README.md for
+information specific to libjpeg-turbo.
+
+
+The Independent JPEG Group's JPEG software
+==========================================
+
+This distribution contains a release of the Independent JPEG Group's free JPEG
+software.  You are welcome to redistribute this software and to use it for any
+purpose, subject to the conditions under LEGAL ISSUES, below.
+
+This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
+Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
+Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
+and other members of the Independent JPEG Group.
+
+IJG is not affiliated with the ISO/IEC JTC1/SC29/WG1 standards committee
+(also known as JPEG, together with ITU-T SG16).
+
+
+DOCUMENTATION ROADMAP
+=====================
+
+This file contains the following sections:
+
+OVERVIEW            General description of JPEG and the IJG software.
+LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
+REFERENCES          Where to learn more about JPEG.
+ARCHIVE LOCATIONS   Where to find newer versions of this software.
+FILE FORMAT WARS    Software *not* to get.
+TO DO               Plans for future IJG releases.
+
+Other documentation files in the distribution are:
+
+User documentation:
+  usage.txt         Usage instructions for cjpeg, djpeg, jpegtran,
+                    rdjpgcom, and wrjpgcom.
+  *.1               Unix-style man pages for programs (same info as usage.txt).
+  wizard.txt        Advanced usage instructions for JPEG wizards only.
+  change.log        Version-to-version change highlights.
+Programmer and internal documentation:
+  libjpeg.txt       How to use the JPEG library in your own programs.
+  example.c         Sample code for calling the JPEG library.
+  structure.txt     Overview of the JPEG library's internal structure.
+  coderules.txt     Coding style rules --- please read if you contribute code.
+
+Please read at least usage.txt.  Some information can also be found in the JPEG
+FAQ (Frequently Asked Questions) article.  See ARCHIVE LOCATIONS below to find
+out where to obtain the FAQ article.
+
+If you want to understand how the JPEG code works, we suggest reading one or
+more of the REFERENCES, then looking at the documentation files (in roughly
+the order listed) before diving into the code.
+
+
+OVERVIEW
+========
+
+This package contains C software to implement JPEG image encoding, decoding,
+and transcoding.  JPEG (pronounced "jay-peg") is a standardized compression
+method for full-color and grayscale images.  JPEG's strong suit is compressing
+photographic images or other types of images that have smooth color and
+brightness transitions between neighboring pixels.  Images with sharp lines or
+other abrupt features may not compress well with JPEG, and a higher JPEG
+quality may have to be used to avoid visible compression artifacts with such
+images.
+
+JPEG is lossy, meaning that the output pixels are not necessarily identical to
+the input pixels.  However, on photographic content and other "smooth" images,
+very good compression ratios can be obtained with no visible compression
+artifacts, and extremely high compression ratios are possible if you are
+willing to sacrifice image quality (by reducing the "quality" setting in the
+compressor.)
+
+This software implements JPEG baseline, extended-sequential, and progressive
+compression processes.  Provision is made for supporting all variants of these
+processes, although some uncommon parameter settings aren't implemented yet.
+We have made no provision for supporting the hierarchical or lossless
+processes defined in the standard.
+
+We provide a set of library routines for reading and writing JPEG image files,
+plus two sample applications "cjpeg" and "djpeg", which use the library to
+perform conversion between JPEG and some other popular image file formats.
+The library is intended to be reused in other applications.
+
+In order to support file conversion and viewing software, we have included
+considerable functionality beyond the bare JPEG coding/decoding capability;
+for example, the color quantization modules are not strictly part of JPEG
+decoding, but they are essential for output to colormapped file formats or
+colormapped displays.  These extra functions can be compiled out of the
+library if not required for a particular application.
+
+We have also included "jpegtran", a utility for lossless transcoding between
+different JPEG processes, and "rdjpgcom" and "wrjpgcom", two simple
+applications for inserting and extracting textual comments in JFIF files.
+
+The emphasis in designing this software has been on achieving portability and
+flexibility, while also making it fast enough to be useful.  In particular,
+the software is not intended to be read as a tutorial on JPEG.  (See the
+REFERENCES section for introductory material.)  Rather, it is intended to
+be reliable, portable, industrial-strength code.  We do not claim to have
+achieved that goal in every aspect of the software, but we strive for it.
+
+We welcome the use of this software as a component of commercial products.
+No royalty is required, but we do ask for an acknowledgement in product
+documentation, as described under LEGAL ISSUES.
+
+
+LEGAL ISSUES
+============
+
+In plain English:
+
+1. We don't promise that this software works.  (But if you find any bugs,
+   please let us know!)
+2. You can use this software for whatever you want.  You don't have to pay us.
+3. You may not pretend that you wrote this software.  If you use it in a
+   program, you must acknowledge somewhere in your documentation that
+   you've used the IJG code.
+
+In legalese:
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-2016, Thomas G. Lane, Guido Vollbeding.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
+
+
+The Unix configuration script "configure" was produced with GNU Autoconf.
+It is copyright by the Free Software Foundation but is freely distributable.
+The same holds for its supporting scripts (config.guess, config.sub,
+ltmain.sh).  Another support script, install-sh, is copyright by X Consortium
+but is also freely distributable.
+
+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent (now expired), GIF reading
+support has been removed altogether, and the GIF writer has been simplified
+to produce "uncompressed GIFs".  This technique does not use the LZW
+algorithm; the resulting GIF files are larger than usual, but are readable
+by all standard GIF decoders.
+
+We are required to state that
+    "The Graphics Interchange Format(c) is the Copyright property of
+    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
+    CompuServe Incorporated."
+
+
+REFERENCES
+==========
+
+We recommend reading one or more of these references before trying to
+understand the innards of the JPEG software.
+
+The best short technical introduction to the JPEG compression algorithm is
+	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
+(Adjacent articles in that issue discuss MPEG motion picture compression,
+applications of JPEG, and related topics.)  If you don't have the CACM issue
+handy, a PDF file containing a revised version of Wallace's article is
+available at http://www.ijg.org/files/Wallace.JPEG.pdf.  The file (actually
+a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
+omits the sample images that appeared in CACM, but it includes corrections
+and some added material.  Note: the Wallace article is copyright ACM and IEEE,
+and it may not be used for commercial purposes.
+
+A somewhat less technical, more leisurely introduction to JPEG can be found in
+"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
+M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
+good explanations and example C code for a multitude of compression methods
+including JPEG.  It is an excellent source if you are comfortable reading C
+code but don't know much about data compression in general.  The book's JPEG
+sample code is far from industrial-strength, but when you are ready to look
+at a full implementation, you've got one here...
+
+The best currently available description of JPEG is the textbook "JPEG Still
+Image Data Compression Standard" by William B. Pennebaker and Joan L.
+Mitchell, published by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
+Price US$59.95, 638 pp.  The book includes the complete text of the ISO JPEG
+standards (DIS 10918-1 and draft DIS 10918-2).
+
+The original JPEG standard is divided into two parts, Part 1 being the actual
+specification, while Part 2 covers compliance testing methods.  Part 1 is
+titled "Digital Compression and Coding of Continuous-tone Still Images,
+Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
+10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
+Continuous-tone Still Images, Part 2: Compliance testing" and has document
+numbers ISO/IEC IS 10918-2, ITU-T T.83.
+
+The JPEG standard does not specify all details of an interchangeable file
+format.  For the omitted details we follow the "JFIF" conventions, revision
+1.02.  JFIF 1.02 has been adopted as an Ecma International Technical Report
+and thus received a formal publication status.  It is available as a free
+download in PDF format from
+http://www.ecma-international.org/publications/techreports/E-TR-098.htm.
+A PostScript version of the JFIF document is available at
+http://www.ijg.org/files/jfif.ps.gz.  There is also a plain text version at
+http://www.ijg.org/files/jfif.txt.gz, but it is missing the figures.
+
+The TIFF 6.0 file format specification can be obtained by FTP from
+ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
+found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
+IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
+Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
+(Compression tag 7).  Copies of this Note can be obtained from
+http://www.ijg.org/files/.  It is expected that the next revision
+of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
+Although IJG's own code does not support TIFF/JPEG, the free libtiff library
+uses our library to implement TIFF/JPEG per the Note.
+
+
+ARCHIVE LOCATIONS
+=================
+
+The "official" archive site for this software is www.ijg.org.
+The most recent released version can always be found there in
+directory "files".
+
+The JPEG FAQ (Frequently Asked Questions) article is a source of some
+general information about JPEG.
+It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
+and other news.answers archive sites, including the official news.answers
+archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
+If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
+with body
+	send usenet/news.answers/jpeg-faq/part1
+	send usenet/news.answers/jpeg-faq/part2
+
+
+FILE FORMAT WARS
+================
+
+The ISO/IEC JTC1/SC29/WG1 standards committee (also known as JPEG, together
+with ITU-T SG16) currently promotes different formats containing the name
+"JPEG" which are incompatible with original DCT-based JPEG.  IJG therefore does
+not support these formats (see REFERENCES).  Indeed, one of the original
+reasons for developing this free software was to help force convergence on
+common, interoperable format standards for JPEG files.
+Don't use an incompatible file format!
+(In any case, our decoder will remain capable of reading existing JPEG
+image files indefinitely.)
+
+
+TO DO
+=====
+
+Please send bug reports, offers of help, etc. to jpeg-info@jpegclub.org.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README.md glmark2-2021.02/android/jni/src/libjpeg-turbo/README.md
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/README.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,335 @@
+Background
+==========
+
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions (MMX, SSE2,
+NEON, AltiVec) to accelerate baseline JPEG compression and decompression on
+x86, x86-64, ARM, and PowerPC systems.  On such systems, libjpeg-turbo is
+generally 2-6x as fast as libjpeg, all else being equal.  On other types of
+systems, libjpeg-turbo can still outperform libjpeg by a significant amount, by
+virtue of its highly-optimized Huffman coding routines.  In many cases, the
+performance of libjpeg-turbo rivals that of proprietary high-speed JPEG codecs.
+
+libjpeg-turbo implements both the traditional libjpeg API as well as the less
+powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features
+colorspace extensions that allow it to compress from/decompress to 32-bit and
+big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java
+interface.
+
+libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated
+derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and
+VirtualGL projects made numerous enhancements to the codec in 2009, and in
+early 2010, libjpeg-turbo spun off into an independent project, with the goal
+of making high-speed JPEG compression/decompression technology available to a
+broader range of users and developers.
+
+
+License
+=======
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses.
+Refer to [LICENSE.md](LICENSE.md) for a roll-up of license terms.
+
+
+Using libjpeg-turbo
+===================
+
+libjpeg-turbo includes two APIs that can be used to compress and decompress
+JPEG images:
+
+- **TurboJPEG API**
+  This API provides an easy-to-use interface for compressing and decompressing
+  JPEG images in memory.  It also provides some functionality that would not be
+  straightforward to achieve using the underlying libjpeg API, such as
+  generating planar YUV images and performing multiple simultaneous lossless
+  transforms on an image.  The Java interface for libjpeg-turbo is written on
+  top of the TurboJPEG API.
+
+- **libjpeg API**
+  This is the de facto industry-standard API for compressing and decompressing
+  JPEG images.  It is more difficult to use than the TurboJPEG API but also
+  more powerful.  The libjpeg API implementation in libjpeg-turbo is both
+  API/ABI-compatible and mathematically compatible with libjpeg v6b.  It can
+  also optionally be configured to be API/ABI-compatible with libjpeg v7 and v8
+  (see below.)
+
+There is no significant performance advantage to either API when both are used
+to perform similar operations.
+
+Colorspace Extensions
+---------------------
+
+libjpeg-turbo includes extensions that allow JPEG images to be compressed
+directly from (and decompressed directly to) buffers that use BGR, BGRX,
+RGBX, XBGR, and XRGB pixel ordering.  This is implemented with ten new
+colorspace constants:
+
+    JCS_EXT_RGB   /* red/green/blue */
+    JCS_EXT_RGBX  /* red/green/blue/x */
+    JCS_EXT_BGR   /* blue/green/red */
+    JCS_EXT_BGRX  /* blue/green/red/x */
+    JCS_EXT_XBGR  /* x/blue/green/red */
+    JCS_EXT_XRGB  /* x/red/green/blue */
+    JCS_EXT_RGBA  /* red/green/blue/alpha */
+    JCS_EXT_BGRA  /* blue/green/red/alpha */
+    JCS_EXT_ABGR  /* alpha/blue/green/red */
+    JCS_EXT_ARGB  /* alpha/red/green/blue */
+
+Setting `cinfo.in_color_space` (compression) or `cinfo.out_color_space`
+(decompression) to one of these values will cause libjpeg-turbo to read the
+red, green, and blue values from (or write them to) the appropriate position in
+the pixel when compressing from/decompressing to an RGB buffer.
+
+Your application can check for the existence of these extensions at compile
+time with:
+
+    #ifdef JCS_EXTENSIONS
+
+At run time, attempting to use these extensions with a libjpeg implementation
+that does not support them will result in a "Bogus input colorspace" error.
+Applications can trap this error in order to test whether run-time support is
+available for the colorspace extensions.
+
+When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
+X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
+can set that byte to whatever value it wishes.  If an application expects the X
+byte to be used as an alpha channel, then it should specify `JCS_EXT_RGBA`,
+`JCS_EXT_BGRA`, `JCS_EXT_ABGR`, or `JCS_EXT_ARGB`.  When these colorspace
+constants are used, the X byte is guaranteed to be 0xFF, which is interpreted
+as opaque.
+
+Your application can check for the existence of the alpha channel colorspace
+extensions at compile time with:
+
+    #ifdef JCS_ALPHA_EXTENSIONS
+
+[jcstest.c](jcstest.c), located in the libjpeg-turbo source tree, demonstrates
+how to check for the existence of the colorspace extensions at compile time and
+run time.
+
+libjpeg v7 and v8 API/ABI Emulation
+-----------------------------------
+
+With libjpeg v7 and v8, new features were added that necessitated extending the
+compression and decompression structures.  Unfortunately, due to the exposed
+nature of those structures, extending them also necessitated breaking backward
+ABI compatibility with previous libjpeg releases.  Thus, programs that were
+built to use libjpeg v7 or v8 did not work with libjpeg-turbo, since it is
+based on the libjpeg v6b code base.  Although libjpeg v7 and v8 are not
+as widely used as v6b, enough programs (including a few Linux distros) made
+the switch that there was a demand to emulate the libjpeg v7 and v8 ABIs
+in libjpeg-turbo.  It should be noted, however, that this feature was added
+primarily so that applications that had already been compiled to use libjpeg
+v7+ could take advantage of accelerated baseline JPEG encoding/decoding
+without recompiling.  libjpeg-turbo does not claim to support all of the
+libjpeg v7+ features, nor to produce identical output to libjpeg v7+ in all
+cases (see below.)
+
+By passing an argument of `--with-jpeg7` or `--with-jpeg8` to `configure`, or
+an argument of `-DWITH_JPEG7=1` or `-DWITH_JPEG8=1` to `cmake`, you can build a
+version of libjpeg-turbo that emulates the libjpeg v7 or v8 ABI, so that
+programs that are built against libjpeg v7 or v8 can be run with libjpeg-turbo.
+The following section describes which libjpeg v7+ features are supported and
+which aren't.
+
+### Support for libjpeg v7 and v8 Features
+
+#### Fully supported
+
+- **libjpeg: IDCT scaling extensions in decompressor**
+  libjpeg-turbo supports IDCT scaling with scaling factors of 1/8, 1/4, 3/8,
+  1/2, 5/8, 3/4, 7/8, 9/8, 5/4, 11/8, 3/2, 13/8, 7/4, 15/8, and 2/1 (only 1/4
+  and 1/2 are SIMD-accelerated.)
+
+- **libjpeg: Arithmetic coding**
+
+- **libjpeg: In-memory source and destination managers**
+  See notes below.
+
+- **cjpeg: Separate quality settings for luminance and chrominance**
+  Note that the libpjeg v7+ API was extended to accommodate this feature only
+  for convenience purposes.  It has always been possible to implement this
+  feature with libjpeg v6b (see rdswitch.c for an example.)
+
+- **cjpeg: 32-bit BMP support**
+
+- **cjpeg: `-rgb` option**
+
+- **jpegtran: Lossless cropping**
+
+- **jpegtran: `-perfect` option**
+
+- **jpegtran: Forcing width/height when performing lossless crop**
+
+- **rdjpgcom: `-raw` option**
+
+- **rdjpgcom: Locale awareness**
+
+
+#### Not supported
+
+NOTE:  As of this writing, extensive research has been conducted into the
+usefulness of DCT scaling as a means of data reduction and SmartScale as a
+means of quality improvement.  The reader is invited to peruse the research at
+http://www.libjpeg-turbo.org/About/SmartScale and draw his/her own conclusions,
+but it is the general belief of our project that these features have not
+demonstrated sufficient usefulness to justify inclusion in libjpeg-turbo.
+
+- **libjpeg: DCT scaling in compressor**
+  `cinfo.scale_num` and `cinfo.scale_denom` are silently ignored.
+  There is no technical reason why DCT scaling could not be supported when
+  emulating the libjpeg v7+ API/ABI, but without the SmartScale extension (see
+  below), only scaling factors of 1/2, 8/15, 4/7, 8/13, 2/3, 8/11, 4/5, and
+  8/9 would be available, which is of limited usefulness.
+
+- **libjpeg: SmartScale**
+  `cinfo.block_size` is silently ignored.
+  SmartScale is an extension to the JPEG format that allows for DCT block
+  sizes other than 8x8.  Providing support for this new format would be
+  feasible (particularly without full acceleration.)  However, until/unless
+  the format becomes either an official industry standard or, at minimum, an
+  accepted solution in the community, we are hesitant to implement it, as
+  there is no sense of whether or how it might change in the future.  It is
+  our belief that SmartScale has not demonstrated sufficient usefulness as a
+  lossless format nor as a means of quality enhancement, and thus our primary
+  interest in providing this feature would be as a means of supporting
+  additional DCT scaling factors.
+
+- **libjpeg: Fancy downsampling in compressor**
+  `cinfo.do_fancy_downsampling` is silently ignored.
+  This requires the DCT scaling feature, which is not supported.
+
+- **jpegtran: Scaling**
+  This requires both the DCT scaling and SmartScale features, which are not
+  supported.
+
+- **Lossless RGB JPEG files**
+  This requires the SmartScale feature, which is not supported.
+
+### What About libjpeg v9?
+
+libjpeg v9 introduced yet another field to the JPEG compression structure
+(`color_transform`), thus making the ABI backward incompatible with that of
+libjpeg v8.  This new field was introduced solely for the purpose of supporting
+lossless SmartScale encoding.  Furthermore, there was actually no reason to
+extend the API in this manner, as the color transform could have just as easily
+been activated by way of a new JPEG colorspace constant, thus preserving
+backward ABI compatibility.
+
+Our research (see link above) has shown that lossless SmartScale does not
+generally accomplish anything that can't already be accomplished better with
+existing, standard lossless formats.  Therefore, at this time it is our belief
+that there is not sufficient technical justification for software projects to
+upgrade from libjpeg v8 to libjpeg v9, and thus there is not sufficient
+echnical justification for us to emulate the libjpeg v9 ABI.
+
+In-Memory Source/Destination Managers
+-------------------------------------
+
+By default, libjpeg-turbo 1.3 and later includes the `jpeg_mem_src()` and
+`jpeg_mem_dest()` functions, even when not emulating the libjpeg v8 API/ABI.
+Previously, it was necessary to build libjpeg-turbo from source with libjpeg v8
+API/ABI emulation in order to use the in-memory source/destination managers,
+but several projects requested that those functions be included when emulating
+the libjpeg v6b API/ABI as well.  This allows the use of those functions by
+programs that need them, without breaking ABI compatibility for programs that
+don't, and it allows those functions to be provided in the "official"
+libjpeg-turbo binaries.
+
+Those who are concerned about maintaining strict conformance with the libjpeg
+v6b or v7 API can pass an argument of `--without-mem-srcdst` to `configure` or
+an argument of `-DWITH_MEM_SRCDST=0` to `cmake` prior to building
+libjpeg-turbo.  This will restore the pre-1.3 behavior, in which
+`jpeg_mem_src()` and `jpeg_mem_dest()` are only included when emulating the
+libjpeg v8 API/ABI.
+
+On Un*x systems, including the in-memory source/destination managers changes
+the dynamic library version from 62.0.0 to 62.1.0 if using libjpeg v6b API/ABI
+emulation and from 7.0.0 to 7.1.0 if using libjpeg v7 API/ABI emulation.
+
+Note that, on most Un*x systems, the dynamic linker will not look for a
+function in a library until that function is actually used.  Thus, if a program
+is built against libjpeg-turbo 1.3+ and uses `jpeg_mem_src()` or
+`jpeg_mem_dest()`, that program will not fail if run against an older version
+of libjpeg-turbo or against libjpeg v7- until the program actually tries to
+call `jpeg_mem_src()` or `jpeg_mem_dest()`.  Such is not the case on Windows.
+If a program is built against the libjpeg-turbo 1.3+ DLL and uses
+`jpeg_mem_src()` or `jpeg_mem_dest()`, then it must use the libjpeg-turbo 1.3+
+DLL at run time.
+
+Both cjpeg and djpeg have been extended to allow testing the in-memory
+source/destination manager functions.  See their respective man pages for more
+details.
+
+
+Mathematical Compatibility
+==========================
+
+For the most part, libjpeg-turbo should produce identical output to libjpeg
+v6b.  The one exception to this is when using the floating point DCT/IDCT, in
+which case the outputs of libjpeg v6b and libjpeg-turbo can differ for the
+following reasons:
+
+- The SSE/SSE2 floating point DCT implementation in libjpeg-turbo is ever so
+  slightly more accurate than the implementation in libjpeg v6b, but not by
+  any amount perceptible to human vision (generally in the range of 0.01 to
+  0.08 dB gain in PNSR.)
+
+- When not using the SIMD extensions, libjpeg-turbo uses the more accurate
+  (and slightly faster) floating point IDCT algorithm introduced in libjpeg
+  v8a as opposed to the algorithm used in libjpeg v6b.  It should be noted,
+  however, that this algorithm basically brings the accuracy of the floating
+  point IDCT in line with the accuracy of the slow integer IDCT.  The floating
+  point DCT/IDCT algorithms are mainly a legacy feature, and they do not
+  produce significantly more accuracy than the slow integer algorithms (to put
+  numbers on this, the typical difference in PNSR between the two algorithms
+  is less than 0.10 dB, whereas changing the quality level by 1 in the upper
+  range of the quality scale is typically more like a 1.0 dB difference.)
+
+- If the floating point algorithms in libjpeg-turbo are not implemented using
+  SIMD instructions on a particular platform, then the accuracy of the
+  floating point DCT/IDCT can depend on the compiler settings.
+
+While libjpeg-turbo does emulate the libjpeg v8 API/ABI, under the hood it is
+still using the same algorithms as libjpeg v6b, so there are several specific
+cases in which libjpeg-turbo cannot be expected to produce the same output as
+libjpeg v8:
+
+- When decompressing using scaling factors of 1/2 and 1/4, because libjpeg v8
+  implements those scaling algorithms differently than libjpeg v6b does, and
+  libjpeg-turbo's SIMD extensions are based on the libjpeg v6b behavior.
+
+- When using chrominance subsampling, because libjpeg v8 implements this
+  with its DCT/IDCT scaling algorithms rather than with a separate
+  downsampling/upsampling algorithm.  In our testing, the subsampled/upsampled
+  output of libjpeg v8 is less accurate than that of libjpeg v6b for this
+  reason.
+
+- When decompressing using a scaling factor > 1 and merged (AKA "non-fancy" or
+  "non-smooth") chrominance upsampling, because libjpeg v8 does not support
+  merged upsampling with scaling factors > 1.
+
+
+Performance Pitfalls
+====================
+
+Restart Markers
+---------------
+
+The optimized Huffman decoder in libjpeg-turbo does not handle restart markers
+in a way that makes the rest of the libjpeg infrastructure happy, so it is
+necessary to use the slow Huffman decoder when decompressing a JPEG image that
+has restart markers.  This can cause the decompression performance to drop by
+as much as 20%, but the performance will still be much greater than that of
+libjpeg.  Many consumer packages, such as PhotoShop, use restart markers when
+generating JPEG images, so images generated by those programs will experience
+this issue.
+
+Fast Integer Forward DCT at High Quality Levels
+-----------------------------------------------
+
+The algorithm used by the SIMD-accelerated quantization function cannot produce
+correct results whenever the fast integer forward DCT is used along with a JPEG
+quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
+function in those cases.  This causes performance to drop by as much as 40%.
+It is therefore strongly advised that you use the slow integer forward DCT
+whenever encoding images with a JPEG quality of 98 or higher.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README-turbo.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/README-turbo.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/README-turbo.txt	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/README-turbo.txt	1970-01-01 08:00:00.000000000 +0800
@@ -1,361 +0,0 @@
-*******************************************************************************
-**     Background
-*******************************************************************************
-
-libjpeg-turbo is a derivative of libjpeg that uses SIMD instructions (MMX,
-SSE2, NEON) to accelerate baseline JPEG compression and decompression on x86,
-x86-64, and ARM systems.  On such systems, libjpeg-turbo is generally 2-4x as
-fast as the unmodified version of libjpeg, all else being equal.
-
-libjpeg-turbo was originally based on libjpeg/SIMD by Miyasaka Masaru, but
-the TigerVNC and VirtualGL projects made numerous enhancements to the codec in
-2009, including improved support for Mac OS X, 64-bit support, support for
-32-bit and big-endian pixel formats (RGBX, XBGR, etc.), accelerated Huffman
-encoding/decoding, and various bug fixes.  The goal was to produce a fully
-open-source codec that could replace the partially closed-source TurboJPEG/IPP
-codec used by VirtualGL and TurboVNC.  libjpeg-turbo generally achieves 80-120%
-of the performance of TurboJPEG/IPP.  It is faster in some areas but slower in
-others.
-
-In early 2010, libjpeg-turbo spun off into its own independent project, with
-the goal of making high-speed JPEG compression/decompression technology
-available to a broader range of users and developers.
-
-
-*******************************************************************************
-**     License
-*******************************************************************************
-
-Most of libjpeg-turbo inherits the non-restrictive, BSD-style license used by
-libjpeg (see README.)  The TurboJPEG/OSS wrapper (both C and Java versions) and
-associated test programs bear a similar license, which is reproduced below:
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
-- Redistributions of source code must retain the above copyright notice,
-  this list of conditions and the following disclaimer.
-- Redistributions in binary form must reproduce the above copyright notice,
-  this list of conditions and the following disclaimer in the documentation
-  and/or other materials provided with the distribution.
-- Neither the name of the libjpeg-turbo Project nor the names of its
-  contributors may be used to endorse or promote products derived from this
-  software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
-AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
-ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
-LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
-CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
-SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
-INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
-CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
-ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
-POSSIBILITY OF SUCH DAMAGE.
-
-
-*******************************************************************************
-**     Using libjpeg-turbo
-*******************************************************************************
-
-libjpeg-turbo includes two APIs that can be used to compress and decompress
-JPEG images:
-
-  TurboJPEG API:  This API provides an easy-to-use interface for compressing
-  and decompressing JPEG images in memory.  It also provides some functionality
-  that would not be straightforward to achieve using the underlying libjpeg
-  API, such as generating planar YUV images and performing multiple
-  simultaneous lossless transforms on an image.  The Java interface for
-  libjpeg-turbo is written on top of the TurboJPEG API.
-
-  libjpeg API:  This is the de facto industry-standard API for compressing and
-  decompressing JPEG images.  It is more difficult to use than the TurboJPEG
-  API but also more powerful.  libjpeg-turbo is both API/ABI-compatible and
-  mathematically compatible with libjpeg v6b.  It can also optionally be
-  configured to be API/ABI-compatible with libjpeg v7 and v8 (see below.)
-
-
-=============================
-Replacing libjpeg at Run Time
-=============================
-
-If a Unix application is dynamically linked with libjpeg, then you can replace
-libjpeg with libjpeg-turbo at run time by manipulating LD_LIBRARY_PATH.
-For instance:
-
-  [Using libjpeg]
-  > time cjpeg <vgl_5674_0098.ppm >vgl_5674_0098.jpg
-  real  0m0.392s
-  user  0m0.074s
-  sys   0m0.020s
-
-  [Using libjpeg-turbo]
-  > export LD_LIBRARY_PATH=/opt/libjpeg-turbo/{lib}:$LD_LIBRARY_PATH
-  > time cjpeg <vgl_5674_0098.ppm >vgl_5674_0098.jpg
-  real  0m0.109s
-  user  0m0.029s
-  sys   0m0.010s
-
-NOTE: {lib} can be lib, lib32, lib64, or lib/64, depending on the O/S and
-architecture.
-
-System administrators can also replace the libjpeg sym links in /usr/{lib} with
-links to the libjpeg-turbo dynamic library located in /opt/libjpeg-turbo/{lib}.
-This will effectively accelerate every application that uses the libjpeg
-dynamic library on the system.
-
-The libjpeg-turbo SDK for Visual C++ installs the libjpeg-turbo DLL
-(jpeg62.dll, jpeg7.dll, or jpeg8.dll, depending on whether it was built with
-libjpeg v6b, v7, or v8 emulation) into c:\libjpeg-turbo[64]\bin, and the PATH
-environment variable can be modified such that this directory is searched
-before any others that might contain a libjpeg DLL.  However, if a libjpeg
-DLL exists in an application's install directory, then Windows will load this
-DLL first whenever the application is launched.  Thus, if an application ships
-with jpeg62.dll, jpeg7.dll, or jpeg8.dll, then back up the application's
-version of this DLL and copy c:\libjpeg-turbo[64]\bin\jpeg*.dll into the
-application's install directory to accelerate it.
-
-The version of the libjpeg-turbo DLL distributed in the libjpeg-turbo SDK for
-Visual C++ requires the Visual C++ 2008 C run-time DLL (msvcr90.dll).
-msvcr90.dll ships with more recent versions of Windows, but users of older
-Windows releases can obtain it from the Visual C++ 2008 Redistributable
-Package, which is available as a free download from Microsoft's web site.
-
-NOTE:  Features of libjpeg that require passing a C run-time structure, such
-as a file handle, from an application to libjpeg will probably not work with
-the version of the libjpeg-turbo DLL distributed in the libjpeg-turbo SDK for
-Visual C++, unless the application is also built to use the Visual C++ 2008 C
-run-time DLL.  In particular, this affects jpeg_stdio_dest() and
-jpeg_stdio_src().
-
-Mac applications typically embed their own copies of the libjpeg dylib inside
-the (hidden) application bundle, so it is not possible to globally replace
-libjpeg on OS X systems.  If an application uses a shared library version of
-libjpeg, then it may be possible to replace the application's version of it.
-This would generally involve copying libjpeg.*.dylib from libjpeg-turbo into
-the appropriate place in the application bundle and using install_name_tool to
-repoint the dylib to the new directory.  This requires an advanced knowledge of
-OS X and would not survive an upgrade or a re-install of the application.
-Thus, it is not recommended for most users.
-
-=======================
-Replacing TurboJPEG/IPP
-=======================
-
-libjpeg-turbo is a drop-in replacement for the TurboJPEG/IPP SDK used by
-VirtualGL 2.1.x and TurboVNC 0.6 (and prior.)  libjpeg-turbo contains a wrapper
-library (TurboJPEG/OSS) that emulates the TurboJPEG API using libjpeg-turbo
-instead of the closed-source Intel Performance Primitives.  You can replace the
-TurboJPEG/IPP package on Linux systems with the libjpeg-turbo package in order
-to make existing releases of VirtualGL 2.1.x and TurboVNC 0.x use the new codec
-at run time.  Note that the 64-bit libjpeg-turbo packages contain only 64-bit
-binaries, whereas the TurboJPEG/IPP 64-bit packages contained both 64-bit and
-32-bit binaries.  Thus, to replace a TurboJPEG/IPP 64-bit package, install
-both the 64-bit and 32-bit versions of libjpeg-turbo.
-
-You can also build the VirtualGL 2.1.x and TurboVNC 0.6 source code with
-the libjpeg-turbo SDK instead of TurboJPEG/IPP.  It should work identically.
-libjpeg-turbo also includes static library versions of TurboJPEG/OSS, which
-are used to build VirtualGL 2.2 and TurboVNC 1.0 and later.
-
-========================================
-Using libjpeg-turbo in Your Own Programs
-========================================
-
-For the most part, libjpeg-turbo should work identically to libjpeg, so in
-most cases, an application can be built against libjpeg and then run against
-libjpeg-turbo.  On Unix systems (including Cygwin), you can build against
-libjpeg-turbo instead of libjpeg by setting
-
-  CPATH=/opt/libjpeg-turbo/include
-  and
-  LIBRARY_PATH=/opt/libjpeg-turbo/{lib}
-
-({lib} = lib32 or lib64, depending on whether you are building a 32-bit or a
-64-bit application.)
-
-If using MinGW, then set
-
-  CPATH=/c/libjpeg-turbo-gcc[64]/include
-  and
-  LIBRARY_PATH=/c/libjpeg-turbo-gcc[64]/lib
-
-Building against libjpeg-turbo is useful, for instance, if you want to build an
-application that leverages the libjpeg-turbo colorspace extensions (see below.)
-On Linux and Solaris systems, you would still need to manipulate
-LD_LIBRARY_PATH or create appropriate sym links to use libjpeg-turbo at run
-time.  On such systems, you can pass -R /opt/libjpeg-turbo/{lib} to the linker
-to force the use of libjpeg-turbo at run time rather than libjpeg (also useful
-if you want to leverage the colorspace extensions), or you can link against the
-libjpeg-turbo static library.
-
-To force a Linux, Solaris, or MinGW application to link against the static
-version of libjpeg-turbo, you can use the following linker options:
-
-  -Wl,-Bstatic -ljpeg -Wl,-Bdynamic
-
-On OS X, simply add /opt/libjpeg-turbo/lib/libjpeg.a to the linker command
-line (this also works on Linux and Solaris.)
-
-To build Visual C++ applications using libjpeg-turbo, add
-c:\libjpeg-turbo[64]\include to the system or user INCLUDE environment
-variable and c:\libjpeg-turbo[64]\lib to the system or user LIB environment
-variable, and then link against either jpeg.lib (to use the DLL version of
-libjpeg-turbo) or jpeg-static.lib (to use the static version of libjpeg-turbo.)
-
-=====================
-Colorspace Extensions
-=====================
-
-libjpeg-turbo includes extensions that allow JPEG images to be compressed
-directly from (and decompressed directly to) buffers that use BGR, BGRX,
-RGBX, XBGR, and XRGB pixel ordering.  This is implemented with ten new
-colorspace constants:
-
-  JCS_EXT_RGB   /* red/green/blue */
-  JCS_EXT_RGBX  /* red/green/blue/x */
-  JCS_EXT_BGR   /* blue/green/red */
-  JCS_EXT_BGRX  /* blue/green/red/x */
-  JCS_EXT_XBGR  /* x/blue/green/red */
-  JCS_EXT_XRGB  /* x/red/green/blue */
-  JCS_EXT_RGBA  /* red/green/blue/alpha */
-  JCS_EXT_BGRA  /* blue/green/red/alpha */
-  JCS_EXT_ABGR  /* alpha/blue/green/red */
-  JCS_EXT_ARGB  /* alpha/red/green/blue */
-
-Setting cinfo.in_color_space (compression) or cinfo.out_color_space
-(decompression) to one of these values will cause libjpeg-turbo to read the
-red, green, and blue values from (or write them to) the appropriate position in
-the pixel when compressing from/decompressing to an RGB buffer.
-
-Your application can check for the existence of these extensions at compile
-time with:
-
-  #ifdef JCS_EXTENSIONS
-
-At run time, attempting to use these extensions with a version of libjpeg
-that doesn't support them will result in a "Bogus input colorspace" error.
-
-When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
-X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
-can set that byte to whatever value it wishes.  If an application expects the X
-byte to be used as an alpha channel, then it should specify JCS_EXT_RGBA,
-JCS_EXT_BGRA, JCS_EXT_ABGR, or JCS_EXT_ARGB.  When these colorspace constants
-are used, the X byte is guaranteed to be 0xFF, which is interpreted as opaque.
-
-Your application can check for the existence of the alpha channel colorspace
-extensions at compile time with:
-
-  #ifdef JCS_ALPHA_EXTENSIONS
-
-jcstest.c, located in the libjpeg-turbo source tree, demonstrates how to check
-for the existence of the colorspace extensions at compile time and run time.
-
-=================================
-libjpeg v7 and v8 API/ABI support
-=================================
-
-With libjpeg v7 and v8, new features were added that necessitated extending the
-compression and decompression structures.  Unfortunately, due to the exposed
-nature of those structures, extending them also necessitated breaking backward
-ABI compatibility with previous libjpeg releases.  Thus, programs that are
-built to use libjpeg v7 or v8 did not work with libjpeg-turbo, since it is
-based on the libjpeg v6b code base.  Although libjpeg v7 and v8 are still not
-as widely used as v6b, enough programs (including a few Linux distros) have
-made the switch that it was desirable to provide support for the libjpeg v7/v8
-API/ABI in libjpeg-turbo.  Although libjpeg-turbo can now be configured as a
-drop-in replacement for libjpeg v7 or v8, it should be noted that not all of
-the features in libjpeg v7 and v8 are supported (see below.)
-
-By passing an argument of --with-jpeg7 or --with-jpeg8 to configure, or an
-argument of -DWITH_JPEG7=1 or -DWITH_JPEG8=1 to cmake, you can build a version
-of libjpeg-turbo that emulates the libjpeg v7 or v8 API/ABI, so that programs
-that are built against libjpeg v7 or v8 can be run with libjpeg-turbo.  The
-following section describes which libjpeg v7+ features are supported and which
-aren't.
-
-libjpeg v7 and v8 Features:
----------------------------
-
-Fully supported:
-
--- cjpeg: Separate quality settings for luminance and chrominance
-   Note that the libpjeg v7+ API was extended to accommodate this feature only
-   for convenience purposes.  It has always been possible to implement this
-   feature with libjpeg v6b (see rdswitch.c for an example.)
-
--- cjpeg: 32-bit BMP support
-
--- jpegtran: lossless cropping
-
--- jpegtran: -perfect option
-
--- rdjpgcom: -raw option
-
--- rdjpgcom: locale awareness
-
-
-Fully supported when using libjpeg v7/v8 emulation:
-
--- libjpeg: In-memory source and destination managers
-
-
-Not supported:
-
--- libjpeg: DCT scaling in compressor
-   cinfo.scale_num and cinfo.scale_denom are silently ignored.
-   There is no technical reason why DCT scaling cannot be supported, but
-   without the SmartScale extension (see below), it would only be able to
-   down-scale using ratios of 1/2, 8/15, 4/7, 8/13, 2/3, 8/11, 4/5, and 8/9,
-   which is of limited usefulness.
-
--- libjpeg: SmartScale
-   cinfo.block_size is silently ignored.
-   SmartScale is an extension to the JPEG format that allows for DCT block
-   sizes other than 8x8.  It would be difficult to support this feature while
-   retaining backward compatibility with libjpeg v6b.
-
--- libjpeg: IDCT scaling extensions in decompressor
-   libjpeg-turbo still supports IDCT scaling with scaling factors of 1/2, 1/4,
-   and 1/8 (same as libjpeg v6b.)
-
--- libjpeg: Fancy downsampling in compressor
-   cinfo.do_fancy_downsampling is silently ignored.
-   This requires the DCT scaling feature, which is not supported.
-
--- jpegtran: Scaling
-   This requires both the DCT scaling and SmartScale features, which are not
-   supported.
-
--- Lossless RGB JPEG files
-   This requires the SmartScale feature, which is not supported.
-
-
-*******************************************************************************
-**     Performance pitfalls
-*******************************************************************************
-
-===============
-Restart Markers
-===============
-
-The optimized Huffman decoder in libjpeg-turbo does not handle restart markers
-in a way that makes the rest of the libjpeg infrastructure happy, so it is
-necessary to use the slow Huffman decoder when decompressing a JPEG image that
-has restart markers.  This can cause the decompression performance to drop by
-as much as 20%, but the performance will still be much greater than that of
-libjpeg.  Many consumer packages, such as PhotoShop, use restart markers when
-generating JPEG images, so images generated by those programs will experience
-this issue.
-
-===============================================
-Fast Integer Forward DCT at High Quality Levels
-===============================================
-
-The algorithm used by the SIMD-accelerated quantization function cannot produce
-correct results whenever the fast integer forward DCT is used along with a JPEG
-quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
-function in those cases.  This causes performance to drop by as much as 40%.
-It is therefore strongly advised that you use the slow integer forward DCT
-whenever encoding images with a JPEG quality of 98 or higher.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/CMakeLists.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/CMakeLists.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/CMakeLists.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/CMakeLists.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,80 @@
+if(NOT DEFINED NASM)
+  set(NASM nasm CACHE FILEPATH "Path to NASM/YASM executable")
+endif()
+
+if(SIMD_X86_64)
+  set(NAFLAGS -fwin64 -DWIN64 -D__x86_64__)
+else()
+  if(BORLAND)
+    set(NAFLAGS -fobj -DOBJ32)
+  else()
+    set(NAFLAGS -fwin32 -DWIN32)
+  endif()
+endif()
+set(NAFLAGS ${NAFLAGS} -I${CMAKE_SOURCE_DIR}/win/ -I${CMAKE_CURRENT_SOURCE_DIR}/)
+
+# This only works if building from the command line.  There is currently no way
+# to set a variable's value based on the build type when using the MSVC IDE.
+if(CMAKE_BUILD_TYPE STREQUAL "Debug"
+  OR CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
+  set(NAFLAGS ${NAFLAGS} -g)
+endif()
+
+if(SIMD_X86_64)
+  set(SIMD_BASENAMES jfdctflt-sse-64 jccolor-sse2-64 jcgray-sse2-64
+    jchuff-sse2-64 jcsample-sse2-64 jdcolor-sse2-64 jdmerge-sse2-64
+    jdsample-sse2-64 jfdctfst-sse2-64 jfdctint-sse2-64 jidctflt-sse2-64
+    jidctfst-sse2-64 jidctint-sse2-64 jidctred-sse2-64 jquantf-sse2-64
+    jquanti-sse2-64)
+  message(STATUS "Building x86_64 SIMD extensions")
+else()
+  set(SIMD_BASENAMES jsimdcpu jfdctflt-3dn jidctflt-3dn jquant-3dn jccolor-mmx
+    jcgray-mmx jcsample-mmx jdcolor-mmx jdmerge-mmx jdsample-mmx jfdctfst-mmx
+    jfdctint-mmx jidctfst-mmx jidctint-mmx jidctred-mmx jquant-mmx jfdctflt-sse
+    jidctflt-sse jquant-sse jccolor-sse2 jcgray-sse2 jchuff-sse2 jcsample-sse2
+    jdcolor-sse2 jdmerge-sse2 jdsample-sse2 jfdctfst-sse2 jfdctint-sse2
+    jidctflt-sse2 jidctfst-sse2 jidctint-sse2 jidctred-sse2 jquantf-sse2
+    jquanti-sse2)
+  message(STATUS "Building i386 SIMD extensions")
+endif()
+
+if(MSVC_IDE)
+  set(OBJDIR "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}")
+else()
+  set(OBJDIR ${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
+file(GLOB INC_FILES *.inc)
+
+foreach(file ${SIMD_BASENAMES})
+  set(DEPFILE "")
+  set(SIMD_SRC ${CMAKE_CURRENT_SOURCE_DIR}/${file}.asm)
+  if(${file} MATCHES jccolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jccolor" "jccolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jcgray)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jcgray" "jcgryext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdcolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdcolor" "jdcolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdmerge)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdmerge" "jdmrgext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  set(SIMD_OBJ ${OBJDIR}/${file}.obj)
+  add_custom_command(OUTPUT ${SIMD_OBJ}
+    DEPENDS ${SIMD_SRC} ${DEPFILE} ${INC_FILES}
+    COMMAND ${NASM} ${NAFLAGS} ${SIMD_SRC} -o${SIMD_OBJ})
+  set(SIMD_OBJS ${SIMD_OBJS} ${SIMD_OBJ})
+endforeach()
+
+set(SIMD_OBJS ${SIMD_OBJS} PARENT_SCOPE)
+add_custom_target(simd DEPENDS ${SIMD_OBJS})
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,267 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * Copyright (C) 2014, Jay Foad.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-altivec.c */
+
+
+void jsimd_rgb_ycc_convert_altivec (JDIMENSION img_width, JSAMPARRAY input_buf,
+                                    JSAMPIMAGE output_buf,
+                                    JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr0, outptr1, outptr2;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1 = {0}, rgb2 = {0},
+    rgbg0, rgbg1, rgbg2, rgbg3, y, cb, cr;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = {0};
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = {0};
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh, crl, crh, cbl, cbh;
+  __vector int y0, y1, y2, y3, cr0, cr1, cr2, cr3, cb0, cb1, cb2, cb3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) },
+    pw_mf016_mf033 = { __4X2(-F_0_168, -F_0_331) },
+    pw_mf008_mf041 = { __4X2(-F_0_081, -F_0_418) };
+  __vector unsigned short pw_f050_f000 = { __4X2(F_0_500, 0) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) },
+    pd_onehalfm1_cj = { __4X(ONE_HALF - 1 + (CENTERJSAMPLE << SCALEBITS)) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr0 += 16, outptr1 += 16, outptr2 += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = VEC_LD(0, tmpbuf);
+          rgb1 = VEC_LD(16, tmpbuf);
+          rgb2 = VEC_LD(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = VEC_LD(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = VEC_LD(0, inptr);
+          if (num_cols > 16)
+            rgb1 = VEC_LD(16, inptr);
+          if (num_cols > 32)
+            rgb2 = VEC_LD(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = VEC_LD(48, inptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr0);
+
+      /* Calculate Cb values */
+      cb0 = vec_msums(rg0, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb1 = vec_msums(rg1, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb2 = vec_msums(rg2, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb3 = vec_msums(rg3, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb0 = (__vector int)vec_msum((__vector unsigned short)bg0, pw_f050_f000,
+                                   (__vector unsigned int)cb0);
+      cb1 = (__vector int)vec_msum((__vector unsigned short)bg1, pw_f050_f000,
+                                   (__vector unsigned int)cb1);
+      cb2 = (__vector int)vec_msum((__vector unsigned short)bg2, pw_f050_f000,
+                                   (__vector unsigned int)cb2);
+      cb3 = (__vector int)vec_msum((__vector unsigned short)bg3, pw_f050_f000,
+                                   (__vector unsigned int)cb3);
+      cbl = vec_perm((__vector unsigned short)cb0,
+                     (__vector unsigned short)cb1, shift_pack_index);
+      cbh = vec_perm((__vector unsigned short)cb2,
+                     (__vector unsigned short)cb3, shift_pack_index);
+      cb = vec_pack(cbl, cbh);
+      vec_st(cb, 0, outptr1);
+
+      /* Calculate Cr values */
+      cr0 = vec_msums(bg0, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr1 = vec_msums(bg1, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr2 = vec_msums(bg2, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr3 = vec_msums(bg3, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr0 = (__vector int)vec_msum((__vector unsigned short)rg0, pw_f050_f000,
+                                   (__vector unsigned int)cr0);
+      cr1 = (__vector int)vec_msum((__vector unsigned short)rg1, pw_f050_f000,
+                                   (__vector unsigned int)cr1);
+      cr2 = (__vector int)vec_msum((__vector unsigned short)rg2, pw_f050_f000,
+                                   (__vector unsigned int)cr2);
+      cr3 = (__vector int)vec_msum((__vector unsigned short)rg3, pw_f050_f000,
+                                   (__vector unsigned int)cr3);
+      crl = vec_perm((__vector unsigned short)cr0,
+                     (__vector unsigned short)cr1, shift_pack_index);
+      crh = vec_perm((__vector unsigned short)cr2,
+                     (__vector unsigned short)cr3, shift_pack_index);
+      cr = vec_pack(crl, crh);
+      vec_st(cr, 0, outptr2);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,477 @@
+;
+; jccolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_mmx (JDIMENSION img_width,
+;                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                           JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          8
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_rgb_ycc_convert_mmx) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edx
+        push    ebx
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        mov     ebx, JSAMPROW [ebx]     ; outptr1
+        mov     edx, JSAMPROW [edx]     ; outptr2
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        xor     eax,eax
+        mov     al, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        xor     edx,edx
+        mov     dx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    mmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    mmG, DWORD [esi+ecx]
+        psllq   mmA, DWORD_BIT
+        por     mmA,mmG
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        movq    mmG,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        mov     ecx, SIZEOF_MMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld16:
+        test    cl, 2*SIZEOF_MMWORD
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_ycc_cnv
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+2*SIZEOF_MMWORD]
+
+.rgb_ycc_cnv:
+        ; mmA=(00 10 20 01 11 21 02 12)
+        ; mmG=(22 03 13 23 04 14 24 05)
+        ; mmF=(15 25 06 16 26 07 17 27)
+
+        movq      mmD,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 10 20 01)
+        psrlq     mmD,4*BYTE_BIT        ; mmD=(11 21 02 12 -- -- -- --)
+
+        punpckhbw mmA,mmG               ; mmA=(00 04 10 14 20 24 01 05)
+        psllq     mmG,4*BYTE_BIT        ; mmG=(-- -- -- -- 22 03 13 23)
+
+        punpcklbw mmD,mmF               ; mmD=(11 15 21 25 02 06 12 16)
+        punpckhbw mmG,mmF               ; mmG=(22 26 03 07 13 17 23 27)
+
+        movq      mmE,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 04 10 14)
+        psrlq     mmE,4*BYTE_BIT        ; mmE=(20 24 01 05 -- -- -- --)
+
+        punpckhbw mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        psllq     mmD,4*BYTE_BIT        ; mmD=(-- -- -- -- 11 15 21 25)
+
+        punpcklbw mmE,mmG               ; mmE=(20 22 24 26 01 03 05 07)
+        punpckhbw mmD,mmG               ; mmD=(11 13 15 17 21 23 25 27)
+
+        pxor      mmH,mmH
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmH               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmH               ; mmC=(10 12 14 16)
+
+        movq      mmB,mmE
+        punpcklbw mmE,mmH               ; mmE=(20 22 24 26)
+        punpckhbw mmB,mmH               ; mmB=(01 03 05 07)
+
+        movq      mmF,mmD
+        punpcklbw mmD,mmH               ; mmD=(11 13 15 17)
+        punpckhbw mmF,mmH               ; mmF=(21 23 25 27)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_MMWORD/8
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_MMWORD/8
+        movd    mmA, DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_MMWORD/4
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_MMWORD/4
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld4:
+        test    cl, SIZEOF_MMWORD/2
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_ycc_cnv
+        movq    mmD,mmA
+        movq    mmC,mmF
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmD, MMWORD [esi+2*SIZEOF_MMWORD]
+        movq    mmC, MMWORD [esi+3*SIZEOF_MMWORD]
+
+.rgb_ycc_cnv:
+        ; mmA=(00 10 20 30 01 11 21 31)
+        ; mmF=(02 12 22 32 03 13 23 33)
+        ; mmD=(04 14 24 34 05 15 25 35)
+        ; mmC=(06 16 26 36 07 17 27 37)
+
+        movq      mmB,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 10 12 20 22 30 32)
+        punpckhbw mmB,mmF               ; mmB=(01 03 11 13 21 23 31 33)
+
+        movq      mmG,mmD
+        punpcklbw mmD,mmC               ; mmD=(04 06 14 16 24 26 34 36)
+        punpckhbw mmG,mmC               ; mmG=(05 07 15 17 25 27 35 37)
+
+        movq      mmE,mmA
+        punpcklwd mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        punpckhwd mmE,mmD               ; mmE=(20 22 24 26 30 32 34 36)
+
+        movq      mmH,mmB
+        punpcklwd mmB,mmG               ; mmB=(01 03 05 07 11 13 15 17)
+        punpckhwd mmH,mmG               ; mmH=(21 23 25 27 31 33 35 37)
+
+        pxor      mmF,mmF
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmF               ; mmC=(10 12 14 16)
+
+        movq      mmD,mmB
+        punpcklbw mmB,mmF               ; mmB=(01 03 05 07)
+        punpckhbw mmD,mmF               ; mmD=(11 13 15 17)
+
+        movq      mmG,mmE
+        punpcklbw mmE,mmF               ; mmE=(20 22 24 26)
+        punpckhbw mmG,mmF               ; mmG=(30 32 34 36)
+
+        punpcklbw mmF,mmH
+        punpckhbw mmH,mmH
+        psrlw     mmF,BYTE_BIT          ; mmF=(21 23 25 27)
+        psrlw     mmH,BYTE_BIT          ; mmH=(31 33 35 37)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+        ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movq      MMWORD [wk(0)], mm0   ; wk(0)=RE
+        movq      MMWORD [wk(1)], mm1   ; wk(1)=RO
+        movq      MMWORD [wk(2)], mm4   ; wk(2)=BE
+        movq      MMWORD [wk(3)], mm5   ; wk(3)=BO
+
+        movq      mm6,mm1
+        punpcklwd mm1,mm3
+        punpckhwd mm6,mm3
+        movq      mm7,mm1
+        movq      mm4,mm6
+        pmaddwd   mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF016_MF033)] ; mm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movq      MMWORD [wk(4)], mm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movq      MMWORD [wk(5)], mm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      mm1,mm1
+        pxor      mm6,mm6
+        punpcklwd mm1,mm5               ; mm1=BOL
+        punpckhwd mm6,mm5               ; mm6=BOH
+        psrld     mm1,1                 ; mm1=BOL*FIX(0.500)
+        psrld     mm6,1                 ; mm6=BOH*FIX(0.500)
+
+        movq      mm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm5=[PD_ONEHALFM1_CJ]
+
+        paddd     mm7,mm1
+        paddd     mm4,mm6
+        paddd     mm7,mm5
+        paddd     mm4,mm5
+        psrld     mm7,SCALEBITS         ; mm7=CbOL
+        psrld     mm4,SCALEBITS         ; mm4=CbOH
+        packssdw  mm7,mm4               ; mm7=CbO
+
+        movq      mm1, MMWORD [wk(2)]   ; mm1=BE
+
+        movq      mm6,mm0
+        punpcklwd mm0,mm2
+        punpckhwd mm6,mm2
+        movq      mm5,mm0
+        movq      mm4,mm6
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF016_MF033)] ; mm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movq      MMWORD [wk(6)], mm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movq      MMWORD [wk(7)], mm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      mm0,mm0
+        pxor      mm6,mm6
+        punpcklwd mm0,mm1               ; mm0=BEL
+        punpckhwd mm6,mm1               ; mm6=BEH
+        psrld     mm0,1                 ; mm0=BEL*FIX(0.500)
+        psrld     mm6,1                 ; mm6=BEH*FIX(0.500)
+
+        movq      mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ]
+
+        paddd     mm5,mm0
+        paddd     mm4,mm6
+        paddd     mm5,mm1
+        paddd     mm4,mm1
+        psrld     mm5,SCALEBITS         ; mm5=CbEL
+        psrld     mm4,SCALEBITS         ; mm4=CbEH
+        packssdw  mm5,mm4               ; mm5=CbE
+
+        psllw     mm7,BYTE_BIT
+        por       mm5,mm7               ; mm5=Cb
+        movq      MMWORD [ebx], mm5     ; Save Cb
+
+        movq      mm0, MMWORD [wk(3)]   ; mm0=BO
+        movq      mm6, MMWORD [wk(2)]   ; mm6=BE
+        movq      mm1, MMWORD [wk(1)]   ; mm1=RO
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm3
+        punpckhwd mm4,mm3
+        movq      mm7,mm0
+        movq      mm5,mm4
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF008_MF041)] ; mm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movq      mm3,[GOTOFF(eax,PD_ONEHALF)]  ; mm3=[PD_ONEHALF]
+
+        paddd     mm0, MMWORD [wk(4)]
+        paddd     mm4, MMWORD [wk(5)]
+        paddd     mm0,mm3
+        paddd     mm4,mm3
+        psrld     mm0,SCALEBITS         ; mm0=YOL
+        psrld     mm4,SCALEBITS         ; mm4=YOH
+        packssdw  mm0,mm4               ; mm0=YO
+
+        pxor      mm3,mm3
+        pxor      mm4,mm4
+        punpcklwd mm3,mm1               ; mm3=ROL
+        punpckhwd mm4,mm1               ; mm4=ROH
+        psrld     mm3,1                 ; mm3=ROL*FIX(0.500)
+        psrld     mm4,1                 ; mm4=ROH*FIX(0.500)
+
+        movq      mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ]
+
+        paddd     mm7,mm3
+        paddd     mm5,mm4
+        paddd     mm7,mm1
+        paddd     mm5,mm1
+        psrld     mm7,SCALEBITS         ; mm7=CrOL
+        psrld     mm5,SCALEBITS         ; mm5=CrOH
+        packssdw  mm7,mm5               ; mm7=CrO
+
+        movq      mm3, MMWORD [wk(0)]   ; mm3=RE
+
+        movq      mm4,mm6
+        punpcklwd mm6,mm2
+        punpckhwd mm4,mm2
+        movq      mm1,mm6
+        movq      mm5,mm4
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   mm1,[GOTOFF(eax,PW_MF008_MF041)] ; mm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movq      mm2,[GOTOFF(eax,PD_ONEHALF)]  ; mm2=[PD_ONEHALF]
+
+        paddd     mm6, MMWORD [wk(6)]
+        paddd     mm4, MMWORD [wk(7)]
+        paddd     mm6,mm2
+        paddd     mm4,mm2
+        psrld     mm6,SCALEBITS         ; mm6=YEL
+        psrld     mm4,SCALEBITS         ; mm4=YEH
+        packssdw  mm6,mm4               ; mm6=YE
+
+        psllw     mm0,BYTE_BIT
+        por       mm6,mm0               ; mm6=Y
+        movq      MMWORD [edi], mm6     ; Save Y
+
+        pxor      mm2,mm2
+        pxor      mm4,mm4
+        punpcklwd mm2,mm3               ; mm2=REL
+        punpckhwd mm4,mm3               ; mm4=REH
+        psrld     mm2,1                 ; mm2=REL*FIX(0.500)
+        psrld     mm4,1                 ; mm4=REH*FIX(0.500)
+
+        movq      mm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm0=[PD_ONEHALFM1_CJ]
+
+        paddd     mm1,mm2
+        paddd     mm5,mm4
+        paddd     mm1,mm0
+        paddd     mm5,mm0
+        psrld     mm1,SCALEBITS         ; mm1=CrEL
+        psrld     mm5,SCALEBITS         ; mm5=CrEH
+        packssdw  mm1,mm5               ; mm1=CrE
+
+        psllw     mm7,BYTE_BIT
+        por       mm1,mm7               ; mm1=Cr
+        movq      MMWORD [edx], mm1     ; Save Cr
+
+        sub     ecx, byte SIZEOF_MMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; inptr
+        add     edi, byte SIZEOF_MMWORD                 ; outptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; outptr1
+        add     edx, byte SIZEOF_MMWORD                 ; outptr2
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        pop     ebx
+        pop     edx
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,485 @@
+;
+; jccolext.asm - colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+; r10 = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13 = JDIMENSION output_row
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          8
+
+        align   16
+
+        global  EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov rsi, r12
+        mov ecx, r13d
+        mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rsi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rsi+2*SIZEOF_JSAMPARRAY]
+        lea     rdi, [rdi+rcx*SIZEOF_JSAMPROW]
+        lea     rbx, [rbx+rcx*SIZEOF_JSAMPROW]
+        lea     rdx, [rdx+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov rsi, r11
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rdx
+        push    rbx
+        push    rdi
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr0
+        mov     rbx, JSAMPROW [rbx]     ; outptr1
+        mov     rdx, JSAMPROW [rdx]     ; outptr2
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    rax
+        push    rdx
+        lea     rcx,[rcx+rcx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_BYTE
+        movzx   rax, BYTE [rsi+rcx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_WORD
+        movzx   rdx, WORD [rsi+rcx]
+        shl     rax, WORD_BIT
+        or      rax,rdx
+.column_ld4:
+        movd    xmmA,eax
+        pop     rdx
+        pop     rax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     rcx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=RE
+        movdqa    XMMWORD [wk(1)], xmm1 ; wk(1)=RO
+        movdqa    XMMWORD [wk(2)], xmm4 ; wk(2)=BE
+        movdqa    XMMWORD [wk(3)], xmm5 ; wk(3)=BO
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        movdqa    xmm7,xmm1
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   xmm7,[rel PW_MF016_MF033] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   xmm4,[rel PW_MF016_MF033] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movdqa    XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      xmm1,xmm1
+        pxor      xmm6,xmm6
+        punpcklwd xmm1,xmm5             ; xmm1=BOL
+        punpckhwd xmm6,xmm5             ; xmm6=BOH
+        psrld     xmm1,1                ; xmm1=BOL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BOH*FIX(0.500)
+
+        movdqa    xmm5,[rel PD_ONEHALFM1_CJ] ; xmm5=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm1
+        paddd     xmm4,xmm6
+        paddd     xmm7,xmm5
+        paddd     xmm4,xmm5
+        psrld     xmm7,SCALEBITS        ; xmm7=CbOL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbOH
+        packssdw  xmm7,xmm4             ; xmm7=CbO
+
+        movdqa    xmm1, XMMWORD [wk(2)] ; xmm1=BE
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   xmm5,[rel PW_MF016_MF033] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   xmm4,[rel PW_MF016_MF033] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      xmm0,xmm0
+        pxor      xmm6,xmm6
+        punpcklwd xmm0,xmm1             ; xmm0=BEL
+        punpckhwd xmm6,xmm1             ; xmm6=BEH
+        psrld     xmm0,1                ; xmm0=BEL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BEH*FIX(0.500)
+
+        movdqa    xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm5,xmm0
+        paddd     xmm4,xmm6
+        paddd     xmm5,xmm1
+        paddd     xmm4,xmm1
+        psrld     xmm5,SCALEBITS        ; xmm5=CbEL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbEH
+        packssdw  xmm5,xmm4             ; xmm5=CbE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm5,xmm7             ; xmm5=Cb
+        movdqa    XMMWORD [rbx], xmm5   ; Save Cb
+
+        movdqa    xmm0, XMMWORD [wk(3)] ; xmm0=BO
+        movdqa    xmm6, XMMWORD [wk(2)] ; xmm6=BE
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=RO
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        movdqa    xmm7,xmm0
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   xmm7,[rel PW_MF008_MF041] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   xmm5,[rel PW_MF008_MF041] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movdqa    xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, XMMWORD [wk(4)]
+        paddd     xmm4, XMMWORD [wk(5)]
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        pxor      xmm3,xmm3
+        pxor      xmm4,xmm4
+        punpcklwd xmm3,xmm1             ; xmm3=ROL
+        punpckhwd xmm4,xmm1             ; xmm4=ROH
+        psrld     xmm3,1                ; xmm3=ROL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=ROH*FIX(0.500)
+
+        movdqa    xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm3
+        paddd     xmm5,xmm4
+        paddd     xmm7,xmm1
+        paddd     xmm5,xmm1
+        psrld     xmm7,SCALEBITS        ; xmm7=CrOL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrOH
+        packssdw  xmm7,xmm5             ; xmm7=CrO
+
+        movdqa    xmm3, XMMWORD [wk(0)] ; xmm3=RE
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   xmm1,[rel PW_MF008_MF041] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   xmm5,[rel PW_MF008_MF041] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movdqa    xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(6)]
+        paddd     xmm4, XMMWORD [wk(7)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [rdi], xmm6   ; Save Y
+
+        pxor      xmm2,xmm2
+        pxor      xmm4,xmm4
+        punpcklwd xmm2,xmm3             ; xmm2=REL
+        punpckhwd xmm4,xmm3             ; xmm4=REH
+        psrld     xmm2,1                ; xmm2=REL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=REH*FIX(0.500)
+
+        movdqa    xmm0,[rel PD_ONEHALFM1_CJ] ; xmm0=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm1,xmm2
+        paddd     xmm5,xmm4
+        paddd     xmm1,xmm0
+        paddd     xmm5,xmm0
+        psrld     xmm1,SCALEBITS        ; xmm1=CrEL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrEH
+        packssdw  xmm1,xmm5             ; xmm1=CrE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm1,xmm7             ; xmm1=Cr
+        movdqa    XMMWORD [rdx], xmm1   ; Save Cr
+
+        sub     rcx, byte SIZEOF_XMMWORD
+        add     rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     rdi, byte SIZEOF_XMMWORD                ; outptr0
+        add     rbx, byte SIZEOF_XMMWORD                ; outptr1
+        add     rdx, byte SIZEOF_XMMWORD                ; outptr2
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .column_ld1
+
+        pop     rcx                     ; col
+        pop     rsi
+        pop     rdi
+        pop     rbx
+        pop     rdx
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     rdi, byte SIZEOF_JSAMPROW
+        add     rbx, byte SIZEOF_JSAMPROW
+        add     rdx, byte SIZEOF_JSAMPROW
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,503 @@
+;
+; jccolext.asm - colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          8
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+
+        global  EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edx
+        push    ebx
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        mov     ebx, JSAMPROW [ebx]     ; outptr1
+        mov     edx, JSAMPROW [edx]     ; outptr2
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        movzx   eax, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        movzx   edx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    xmmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     ecx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=RE
+        movdqa    XMMWORD [wk(1)], xmm1 ; wk(1)=RO
+        movdqa    XMMWORD [wk(2)], xmm4 ; wk(2)=BE
+        movdqa    XMMWORD [wk(3)], xmm5 ; wk(3)=BO
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        movdqa    xmm7,xmm1
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF016_MF033)] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movdqa    XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      xmm1,xmm1
+        pxor      xmm6,xmm6
+        punpcklwd xmm1,xmm5             ; xmm1=BOL
+        punpckhwd xmm6,xmm5             ; xmm6=BOH
+        psrld     xmm1,1                ; xmm1=BOL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BOH*FIX(0.500)
+
+        movdqa    xmm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm5=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm1
+        paddd     xmm4,xmm6
+        paddd     xmm7,xmm5
+        paddd     xmm4,xmm5
+        psrld     xmm7,SCALEBITS        ; xmm7=CbOL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbOH
+        packssdw  xmm7,xmm4             ; xmm7=CbO
+
+        movdqa    xmm1, XMMWORD [wk(2)] ; xmm1=BE
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF016_MF033)] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      xmm0,xmm0
+        pxor      xmm6,xmm6
+        punpcklwd xmm0,xmm1             ; xmm0=BEL
+        punpckhwd xmm6,xmm1             ; xmm6=BEH
+        psrld     xmm0,1                ; xmm0=BEL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BEH*FIX(0.500)
+
+        movdqa    xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm5,xmm0
+        paddd     xmm4,xmm6
+        paddd     xmm5,xmm1
+        paddd     xmm4,xmm1
+        psrld     xmm5,SCALEBITS        ; xmm5=CbEL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbEH
+        packssdw  xmm5,xmm4             ; xmm5=CbE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm5,xmm7             ; xmm5=Cb
+        movdqa    XMMWORD [ebx], xmm5   ; Save Cb
+
+        movdqa    xmm0, XMMWORD [wk(3)] ; xmm0=BO
+        movdqa    xmm6, XMMWORD [wk(2)] ; xmm6=BE
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=RO
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        movdqa    xmm7,xmm0
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF008_MF041)] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movdqa    xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, XMMWORD [wk(4)]
+        paddd     xmm4, XMMWORD [wk(5)]
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        pxor      xmm3,xmm3
+        pxor      xmm4,xmm4
+        punpcklwd xmm3,xmm1             ; xmm3=ROL
+        punpckhwd xmm4,xmm1             ; xmm4=ROH
+        psrld     xmm3,1                ; xmm3=ROL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=ROH*FIX(0.500)
+
+        movdqa    xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm3
+        paddd     xmm5,xmm4
+        paddd     xmm7,xmm1
+        paddd     xmm5,xmm1
+        psrld     xmm7,SCALEBITS        ; xmm7=CrOL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrOH
+        packssdw  xmm7,xmm5             ; xmm7=CrO
+
+        movdqa    xmm3, XMMWORD [wk(0)] ; xmm3=RE
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   xmm1,[GOTOFF(eax,PW_MF008_MF041)] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movdqa    xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(6)]
+        paddd     xmm4, XMMWORD [wk(7)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [edi], xmm6   ; Save Y
+
+        pxor      xmm2,xmm2
+        pxor      xmm4,xmm4
+        punpcklwd xmm2,xmm3             ; xmm2=REL
+        punpckhwd xmm4,xmm3             ; xmm4=REH
+        psrld     xmm2,1                ; xmm2=REL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=REH*FIX(0.500)
+
+        movdqa    xmm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm0=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm1,xmm2
+        paddd     xmm5,xmm4
+        paddd     xmm1,xmm0
+        paddd     xmm5,xmm0
+        psrld     xmm1,SCALEBITS        ; xmm1=CrEL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrEH
+        packssdw  xmm1,xmm5             ; xmm1=CrE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm1,xmm7             ; xmm1=Cr
+        movdqa    XMMWORD [edx], xmm1   ; Save Cr
+
+        sub     ecx, byte SIZEOF_XMMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     edi, byte SIZEOF_XMMWORD                ; outptr0
+        add     ebx, byte SIZEOF_XMMWORD                ; outptr1
+        add     edx, byte SIZEOF_XMMWORD                ; outptr2
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        pop     ebx
+        pop     edx
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,104 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> YCC CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_081 5329                 /* FIX(0.08131) */
+#define F_0_114 7471                 /* FIX(0.11400) */
+#define F_0_168 11059                /* FIX(0.16874) */
+#define F_0_250 16384                /* FIX(0.25000) */
+#define F_0_299 19595                /* FIX(0.29900) */
+#define F_0_331 21709                /* FIX(0.33126) */
+#define F_0_418 27439                /* FIX(0.41869) */
+#define F_0_500 32768                /* FIX(0.50000) */
+#define F_0_587 38470                /* FIX(0.58700) */
+#define F_0_337 (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 {0,1,3,4,6,7,9,10,2,1,5,4,8,7,11,10}
+#define RGBG_INDEX1 {12,13,15,16,18,19,21,22,14,13,17,16,20,19,23,22}
+#define RGBG_INDEX2 {8,9,11,12,14,15,17,18,10,9,13,12,16,15,19,18}
+#define RGBG_INDEX3 {4,5,7,8,10,11,13,14,6,5,9,8,12,11,15,14}
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_altivec jsimd_extrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX {0,1,4,5,8,9,12,13,2,1,6,5,10,9,14,13}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extrgbx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 {2,1,5,4,8,7,11,10,0,1,3,4,6,7,9,10}
+#define RGBG_INDEX1 {14,13,17,16,20,19,23,22,12,13,15,16,18,19,21,22}
+#define RGBG_INDEX2 {10,9,13,12,16,15,19,18,8,9,11,12,14,15,17,18}
+#define RGBG_INDEX3 {6,5,9,8,12,11,15,14,4,5,7,8,10,11,13,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX {2,1,6,5,10,9,14,13,0,1,4,5,8,9,12,13}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extbgrx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX {3,2,7,6,11,10,15,14,1,2,5,6,9,10,13,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extxbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX {1,2,5,6,9,10,13,14,3,2,7,6,11,10,15,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extxrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,123 @@
+;
+; jccolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_mmx) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_mmx):
+
+PW_F0299_F0337  times 2 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 2 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 2 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 2 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 2 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 2 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extrgbx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extbgrx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extxbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extxrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jccolor-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jccolor-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,99 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> GRAYSCALE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_114 7471                 /* FIX(0.11400) */
+#define F_0_250 16384                /* FIX(0.25000) */
+#define F_0_299 19595                /* FIX(0.29900) */
+#define F_0_587 38470                /* FIX(0.58700) */
+#define F_0_337 (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 {0,1,3,4,6,7,9,10,2,1,5,4,8,7,11,10}
+#define RGBG_INDEX1 {12,13,15,16,18,19,21,22,14,13,17,16,20,19,23,22}
+#define RGBG_INDEX2 {8,9,11,12,14,15,17,18,10,9,13,12,16,15,19,18}
+#define RGBG_INDEX3 {4,5,7,8,10,11,13,14,6,5,9,8,12,11,15,14}
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_altivec jsimd_extrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX {0,1,4,5,8,9,12,13,2,1,6,5,10,9,14,13}
+#define jsimd_rgb_gray_convert_altivec jsimd_extrgbx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 {2,1,5,4,8,7,11,10,0,1,3,4,6,7,9,10}
+#define RGBG_INDEX1 {14,13,17,16,20,19,23,22,12,13,15,16,18,19,21,22}
+#define RGBG_INDEX2 {10,9,13,12,16,15,19,18,8,9,11,12,14,15,17,18}
+#define RGBG_INDEX3 {6,5,9,8,12,11,15,14,4,5,7,8,10,11,13,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX {2,1,6,5,10,9,14,13,0,1,4,5,8,9,12,13}
+#define jsimd_rgb_gray_convert_altivec jsimd_extbgrx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX {3,2,7,6,11,10,15,14,1,2,5,6,9,10,13,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extxbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX {1,2,5,6,9,10,13,14,3,2,7,6,11,10,15,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extxrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,116 @@
+;
+; jcgray.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2011 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_mmx) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_mmx):
+
+PW_F0299_F0337  times 2 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 2 dw  F_0_114, F_0_250
+PD_ONEHALF      times 2 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extrgbx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extbgrx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extxbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extxrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgray-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgray-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,227 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * Copyright (C) 2014, Jay Foad.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jcgray-altivec.c */
+
+
+void jsimd_rgb_gray_convert_altivec (JDIMENSION img_width,
+                                     JSAMPARRAY input_buf,
+                                     JSAMPIMAGE output_buf,
+                                     JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+#endif
+
+  __vector unsigned char rgb0, rgb1 = {0}, rgb2 = {0},
+    rgbg0, rgbg1, rgbg2, rgbg3, y;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = {0};
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = {0};
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh;
+  __vector int y0, y1, y2, y3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (num_cols > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (num_cols > 32)
+            rgb2 = vec_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = vec_ld(48, inptr);
+#endif
+        }
+      }
+#else
+      /* Little endian */
+      rgb0 = vec_vsx_ld(0, inptr);
+      if (num_cols > 16)
+        rgb1 = vec_vsx_ld(16, inptr);
+      if (num_cols > 32)
+        rgb2 = vec_vsx_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+      if (num_cols > 48)
+        rgb3 = vec_vsx_ld(48, inptr);
+#endif
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,357 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2011 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_mmx (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_rgb_gray_convert_mmx) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        xor     eax,eax
+        mov     al, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        xor     edx,edx
+        mov     dx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    mmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    mmG, DWORD [esi+ecx]
+        psllq   mmA, DWORD_BIT
+        por     mmA,mmG
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        movq    mmG,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        mov     ecx, SIZEOF_MMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld16:
+        test    cl, 2*SIZEOF_MMWORD
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_gray_cnv
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+2*SIZEOF_MMWORD]
+
+.rgb_gray_cnv:
+        ; mmA=(00 10 20 01 11 21 02 12)
+        ; mmG=(22 03 13 23 04 14 24 05)
+        ; mmF=(15 25 06 16 26 07 17 27)
+
+        movq      mmD,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 10 20 01)
+        psrlq     mmD,4*BYTE_BIT        ; mmD=(11 21 02 12 -- -- -- --)
+
+        punpckhbw mmA,mmG               ; mmA=(00 04 10 14 20 24 01 05)
+        psllq     mmG,4*BYTE_BIT        ; mmG=(-- -- -- -- 22 03 13 23)
+
+        punpcklbw mmD,mmF               ; mmD=(11 15 21 25 02 06 12 16)
+        punpckhbw mmG,mmF               ; mmG=(22 26 03 07 13 17 23 27)
+
+        movq      mmE,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 04 10 14)
+        psrlq     mmE,4*BYTE_BIT        ; mmE=(20 24 01 05 -- -- -- --)
+
+        punpckhbw mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        psllq     mmD,4*BYTE_BIT        ; mmD=(-- -- -- -- 11 15 21 25)
+
+        punpcklbw mmE,mmG               ; mmE=(20 22 24 26 01 03 05 07)
+        punpckhbw mmD,mmG               ; mmD=(11 13 15 17 21 23 25 27)
+
+        pxor      mmH,mmH
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmH               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmH               ; mmC=(10 12 14 16)
+
+        movq      mmB,mmE
+        punpcklbw mmE,mmH               ; mmE=(20 22 24 26)
+        punpckhbw mmB,mmH               ; mmB=(01 03 05 07)
+
+        movq      mmF,mmD
+        punpcklbw mmD,mmH               ; mmD=(11 13 15 17)
+        punpckhbw mmF,mmH               ; mmF=(21 23 25 27)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_MMWORD/8
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_MMWORD/8
+        movd    mmA, DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_MMWORD/4
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_MMWORD/4
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld4:
+        test    cl, SIZEOF_MMWORD/2
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_gray_cnv
+        movq    mmD,mmA
+        movq    mmC,mmF
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmD, MMWORD [esi+2*SIZEOF_MMWORD]
+        movq    mmC, MMWORD [esi+3*SIZEOF_MMWORD]
+
+.rgb_gray_cnv:
+        ; mmA=(00 10 20 30 01 11 21 31)
+        ; mmF=(02 12 22 32 03 13 23 33)
+        ; mmD=(04 14 24 34 05 15 25 35)
+        ; mmC=(06 16 26 36 07 17 27 37)
+
+        movq      mmB,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 10 12 20 22 30 32)
+        punpckhbw mmB,mmF               ; mmB=(01 03 11 13 21 23 31 33)
+
+        movq      mmG,mmD
+        punpcklbw mmD,mmC               ; mmD=(04 06 14 16 24 26 34 36)
+        punpckhbw mmG,mmC               ; mmG=(05 07 15 17 25 27 35 37)
+
+        movq      mmE,mmA
+        punpcklwd mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        punpckhwd mmE,mmD               ; mmE=(20 22 24 26 30 32 34 36)
+
+        movq      mmH,mmB
+        punpcklwd mmB,mmG               ; mmB=(01 03 05 07 11 13 15 17)
+        punpckhwd mmH,mmG               ; mmH=(21 23 25 27 31 33 35 37)
+
+        pxor      mmF,mmF
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmF               ; mmC=(10 12 14 16)
+
+        movq      mmD,mmB
+        punpcklbw mmB,mmF               ; mmB=(01 03 05 07)
+        punpckhbw mmD,mmF               ; mmD=(11 13 15 17)
+
+        movq      mmG,mmE
+        punpcklbw mmE,mmF               ; mmE=(20 22 24 26)
+        punpckhbw mmG,mmF               ; mmG=(30 32 34 36)
+
+        punpcklbw mmF,mmH
+        punpckhbw mmH,mmH
+        psrlw     mmF,BYTE_BIT          ; mmF=(21 23 25 27)
+        psrlw     mmH,BYTE_BIT          ; mmH=(31 33 35 37)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+        ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movq      mm6,mm1
+        punpcklwd mm1,mm3
+        punpckhwd mm6,mm3
+        pmaddwd   mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movq      mm7, mm6      ; mm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movq      mm6,mm0
+        punpcklwd mm0,mm2
+        punpckhwd mm6,mm2
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movq      MMWORD [wk(0)], mm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movq      MMWORD [wk(1)], mm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movq      mm0, mm5      ; mm0=BO
+        movq      mm6, mm4      ; mm6=BE
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm3
+        punpckhwd mm4,mm3
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movq      mm3,[GOTOFF(eax,PD_ONEHALF)]  ; mm3=[PD_ONEHALF]
+
+        paddd     mm0, mm1
+        paddd     mm4, mm7
+        paddd     mm0,mm3
+        paddd     mm4,mm3
+        psrld     mm0,SCALEBITS         ; mm0=YOL
+        psrld     mm4,SCALEBITS         ; mm4=YOH
+        packssdw  mm0,mm4               ; mm0=YO
+
+        movq      mm4,mm6
+        punpcklwd mm6,mm2
+        punpckhwd mm4,mm2
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movq      mm2,[GOTOFF(eax,PD_ONEHALF)]  ; mm2=[PD_ONEHALF]
+
+        paddd     mm6, MMWORD [wk(0)]
+        paddd     mm4, MMWORD [wk(1)]
+        paddd     mm6,mm2
+        paddd     mm4,mm2
+        psrld     mm6,SCALEBITS         ; mm6=YEL
+        psrld     mm4,SCALEBITS         ; mm4=YEH
+        packssdw  mm6,mm4               ; mm6=YE
+
+        psllw     mm0,BYTE_BIT
+        por       mm6,mm0               ; mm6=Y
+        movq      MMWORD [edi], mm6     ; Save Y
+
+        sub     ecx, byte SIZEOF_MMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; inptr
+        add     edi, byte SIZEOF_MMWORD                 ; outptr0
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,364 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width,
+;                              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                              JDIMENSION output_row, int num_rows);
+;
+
+; r10 = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13 = JDIMENSION output_row
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+
+        global  EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov rsi, r12
+        mov ecx, r13d
+        mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
+        lea     rdi, [rdi+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov rsi, r11
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rdi
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr0
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    rax
+        push    rdx
+        lea     rcx,[rcx+rcx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_BYTE
+        movzx   rax, BYTE [rsi+rcx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_WORD
+        movzx   rdx, WORD [rsi+rcx]
+        shl     rax, WORD_BIT
+        or      rax,rdx
+.column_ld4:
+        movd    xmmA,eax
+        pop     rdx
+        pop     rax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     rcx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        pmaddwd   xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm7, xmm6    ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        pmaddwd   xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    xmm0, xmm5    ; xmm0=BO
+        movdqa    xmm6, xmm4    ; xmm6=BE
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        pmaddwd   xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movdqa    xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, xmm1
+        paddd     xmm4, xmm7
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        pmaddwd   xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movdqa    xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(0)]
+        paddd     xmm4, XMMWORD [wk(1)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [rdi], xmm6   ; Save Y
+
+        sub     rcx, byte SIZEOF_XMMWORD
+        add     rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     rdi, byte SIZEOF_XMMWORD                ; outptr0
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .column_ld1
+
+        pop     rcx                     ; col
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     rdi, byte SIZEOF_JSAMPROW
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcgryext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,383 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width,
+;                              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                              JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+
+        global  EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        movzx   eax, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        movzx   edx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    xmmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     ecx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        pmaddwd   xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm7, xmm6    ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    xmm0, xmm5    ; xmm0=BO
+        movdqa    xmm6, xmm4    ; xmm6=BE
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movdqa    xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, xmm1
+        paddd     xmm4, xmm7
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movdqa    xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(0)]
+        paddd     xmm4, XMMWORD [wk(1)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [edi], xmm6   ; Save Y
+
+        sub     ecx, byte SIZEOF_XMMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     edi, byte SIZEOF_XMMWORD                ; outptr0
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jchuff-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jchuff-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jchuff-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jchuff-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,361 @@
+;
+; jchuff-sse2-64.asm - Huffman entropy encoding (64-bit SSE2)
+;
+; Copyright 2009-2011, 2014-2016 D. R. Commander.
+; Copyright 2015 Matthieu Darbois
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based directly on jchuff.c; see jchuff.c for more
+; details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_huff_encode_one_block) PRIVATE
+
+EXTN(jconst_huff_encode_one_block):
+
+%include "jpeg_nbits_table.inc"
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+; These macros perform the same task as the emit_bits() function in the
+; original libjpeg code.  In addition to reducing overhead by explicitly
+; inlining the code, additional performance is achieved by taking into
+; account the size of the bit buffer and waiting until it is almost full
+; before emptying it.  This mostly benefits 64-bit platforms, since 6
+; bytes can be stored in a 64-bit bit buffer before it has to be emptied.
+
+%macro EMIT_BYTE 0
+        sub put_bits, 8  ; put_bits -= 8;
+        mov rdx, put_buffer
+        mov ecx, put_bits
+        shr rdx, cl  ; c = (JOCTET)GETJOCTET(put_buffer >> put_bits);
+        mov byte [buffer], dl  ; *buffer++ = c;
+        add buffer, 1
+        cmp dl, 0xFF  ; need to stuff a zero byte?
+        jne %%.EMIT_BYTE_END
+        mov byte [buffer], 0  ; *buffer++ = 0;
+        add buffer, 1
+%%.EMIT_BYTE_END:
+%endmacro
+
+%macro PUT_BITS 1
+        add put_bits, ecx  ; put_bits += size;
+        shl put_buffer, cl  ; put_buffer = (put_buffer << size);
+        or  put_buffer, %1
+%endmacro
+
+%macro CHECKBUF31 0
+        cmp put_bits, 32  ; if (put_bits > 31) {
+        jl %%.CHECKBUF31_END
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+%%.CHECKBUF31_END:
+%endmacro
+
+%macro CHECKBUF47 0
+        cmp put_bits, 48  ; if (put_bits > 47) {
+        jl %%.CHECKBUF47_END
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+%%.CHECKBUF47_END:
+%endmacro
+
+%macro EMIT_BITS 2
+        CHECKBUF47
+        mov ecx, %2
+        PUT_BITS %1
+%endmacro
+
+%macro kloop_prepare 37  ;(ko, jno0, ..., jno31, xmm0, xmm1, xmm2, xmm3)
+    pxor xmm8, xmm8  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm9, xmm9  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm10, xmm10  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm11, xmm11  ; __m128i neg = _mm_setzero_si128();
+    pinsrw %34, word [r12 + %2  * SIZEOF_WORD], 0  ; xmm_shadow[0] = block[jno0];
+    pinsrw %35, word [r12 + %10 * SIZEOF_WORD], 0  ; xmm_shadow[8] = block[jno8];
+    pinsrw %36, word [r12 + %18 * SIZEOF_WORD], 0  ; xmm_shadow[16] = block[jno16];
+    pinsrw %37, word [r12 + %26 * SIZEOF_WORD], 0  ; xmm_shadow[24] = block[jno24];
+    pinsrw %34, word [r12 + %3  * SIZEOF_WORD], 1  ; xmm_shadow[1] = block[jno1];
+    pinsrw %35, word [r12 + %11 * SIZEOF_WORD], 1  ; xmm_shadow[9] = block[jno9];
+    pinsrw %36, word [r12 + %19 * SIZEOF_WORD], 1  ; xmm_shadow[17] = block[jno17];
+    pinsrw %37, word [r12 + %27 * SIZEOF_WORD], 1  ; xmm_shadow[25] = block[jno25];
+    pinsrw %34, word [r12 + %4  * SIZEOF_WORD], 2  ; xmm_shadow[2] = block[jno2];
+    pinsrw %35, word [r12 + %12 * SIZEOF_WORD], 2  ; xmm_shadow[10] = block[jno10];
+    pinsrw %36, word [r12 + %20 * SIZEOF_WORD], 2  ; xmm_shadow[18] = block[jno18];
+    pinsrw %37, word [r12 + %28 * SIZEOF_WORD], 2  ; xmm_shadow[26] = block[jno26];
+    pinsrw %34, word [r12 + %5  * SIZEOF_WORD], 3  ; xmm_shadow[3] = block[jno3];
+    pinsrw %35, word [r12 + %13 * SIZEOF_WORD], 3  ; xmm_shadow[11] = block[jno11];
+    pinsrw %36, word [r12 + %21 * SIZEOF_WORD], 3  ; xmm_shadow[19] = block[jno19];
+    pinsrw %37, word [r12 + %29 * SIZEOF_WORD], 3  ; xmm_shadow[27] = block[jno27];
+    pinsrw %34, word [r12 + %6  * SIZEOF_WORD], 4  ; xmm_shadow[4] = block[jno4];
+    pinsrw %35, word [r12 + %14 * SIZEOF_WORD], 4  ; xmm_shadow[12] = block[jno12];
+    pinsrw %36, word [r12 + %22 * SIZEOF_WORD], 4  ; xmm_shadow[20] = block[jno20];
+    pinsrw %37, word [r12 + %30 * SIZEOF_WORD], 4  ; xmm_shadow[28] = block[jno28];
+    pinsrw %34, word [r12 + %7  * SIZEOF_WORD], 5  ; xmm_shadow[5] = block[jno5];
+    pinsrw %35, word [r12 + %15 * SIZEOF_WORD], 5  ; xmm_shadow[13] = block[jno13];
+    pinsrw %36, word [r12 + %23 * SIZEOF_WORD], 5  ; xmm_shadow[21] = block[jno21];
+    pinsrw %37, word [r12 + %31 * SIZEOF_WORD], 5  ; xmm_shadow[29] = block[jno29];
+    pinsrw %34, word [r12 + %8  * SIZEOF_WORD], 6  ; xmm_shadow[6] = block[jno6];
+    pinsrw %35, word [r12 + %16 * SIZEOF_WORD], 6  ; xmm_shadow[14] = block[jno14];
+    pinsrw %36, word [r12 + %24 * SIZEOF_WORD], 6  ; xmm_shadow[22] = block[jno22];
+    pinsrw %37, word [r12 + %32 * SIZEOF_WORD], 6  ; xmm_shadow[30] = block[jno30];
+    pinsrw %34, word [r12 + %9  * SIZEOF_WORD], 7  ; xmm_shadow[7] = block[jno7];
+    pinsrw %35, word [r12 + %17 * SIZEOF_WORD], 7  ; xmm_shadow[15] = block[jno15];
+    pinsrw %36, word [r12 + %25 * SIZEOF_WORD], 7  ; xmm_shadow[23] = block[jno23];
+%if %1 != 32
+    pinsrw %37, word [r12 + %33 * SIZEOF_WORD], 7  ; xmm_shadow[31] = block[jno31];
+%else
+    pinsrw %37, ebx, 7  ; xmm_shadow[31] = block[jno31];
+%endif
+    pcmpgtw xmm8, %34  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm9, %35  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm10, %36  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm11, %37  ; neg = _mm_cmpgt_epi16(neg, x1);
+    paddw %34, xmm8   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %35, xmm9   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %36, xmm10  ; x1 = _mm_add_epi16(x1, neg);
+    paddw %37, xmm11  ; x1 = _mm_add_epi16(x1, neg);
+    pxor %34, xmm8    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %35, xmm9    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %36, xmm10   ; x1 = _mm_xor_si128(x1, neg);
+    pxor %37, xmm11   ; x1 = _mm_xor_si128(x1, neg);
+    pxor xmm8, %34    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm9, %35    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm10, %36   ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm11, %37   ; neg = _mm_xor_si128(neg, x1);
+    movdqa XMMWORD [t1 + %1 * SIZEOF_WORD], %34  ; _mm_storeu_si128((__m128i *)(t1 + ko), x1);
+    movdqa XMMWORD [t1 + (%1 + 8) * SIZEOF_WORD], %35  ; _mm_storeu_si128((__m128i *)(t1 + ko + 8), x1);
+    movdqa XMMWORD [t1 + (%1 + 16) * SIZEOF_WORD], %36  ; _mm_storeu_si128((__m128i *)(t1 + ko + 16), x1);
+    movdqa XMMWORD [t1 + (%1 + 24) * SIZEOF_WORD], %37  ; _mm_storeu_si128((__m128i *)(t1 + ko + 24), x1);
+    movdqa XMMWORD [t2 + %1 * SIZEOF_WORD], xmm8  ; _mm_storeu_si128((__m128i *)(t2 + ko), neg);
+    movdqa XMMWORD [t2 + (%1 + 8) * SIZEOF_WORD], xmm9  ; _mm_storeu_si128((__m128i *)(t2 + ko + 8), neg);
+    movdqa XMMWORD [t2 + (%1 + 16) * SIZEOF_WORD], xmm10  ; _mm_storeu_si128((__m128i *)(t2 + ko + 16), neg);
+    movdqa XMMWORD [t2 + (%1 + 24) * SIZEOF_WORD], xmm11  ; _mm_storeu_si128((__m128i *)(t2 + ko + 24), neg);
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET*)
+; jsimd_huff_encode_one_block_sse2 (working_state *state, JOCTET *buffer,
+;                                   JCOEFPTR block, int last_dc_val,
+;                                   c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+
+; r10 = working_state *state
+; r11 = JOCTET *buffer
+; r12 = JCOEFPTR block
+; r13 = int last_dc_val
+; r14 = c_derived_tbl *dctbl
+; r15 = c_derived_tbl *actbl
+
+%define t1              rbp-(DCTSIZE2*SIZEOF_WORD)
+%define t2              t1-(DCTSIZE2*SIZEOF_WORD)
+%define put_buffer      r8
+%define put_bits        r9d
+%define buffer          rax
+
+        align   16
+        global  EXTN(jsimd_huff_encode_one_block_sse2) PRIVATE
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [t2]
+        collect_args
+%ifdef WIN64
+        movaps  XMMWORD [rsp-1*SIZEOF_XMMWORD], xmm8
+        movaps  XMMWORD [rsp-2*SIZEOF_XMMWORD], xmm9
+        movaps  XMMWORD [rsp-3*SIZEOF_XMMWORD], xmm10
+        movaps  XMMWORD [rsp-4*SIZEOF_XMMWORD], xmm11
+        sub     rsp, 4*SIZEOF_XMMWORD
+%endif
+        push rbx
+
+        mov buffer, r11  ; r11 is now sratch
+
+        mov put_buffer, MMWORD [r10+16]  ; put_buffer = state->cur.put_buffer;
+        mov put_bits,    DWORD [r10+24]  ; put_bits = state->cur.put_bits;
+        push r10  ; r10 is now scratch
+
+        ; Encode the DC coefficient difference per section F.1.2.1
+        movsx edi, word [r12]  ; temp = temp2 = block[0] - last_dc_val;
+        sub   edi, r13d  ; r13 is not used anymore
+        mov   ebx, edi
+
+        ; This is a well-known technique for obtaining the absolute value
+        ; without a branch.  It is derived from an assembly language technique
+        ; presented in "How to Optimize for the Pentium Processors",
+        ; Copyright (c) 1996, 1997 by Agner Fog.
+        mov esi, edi
+        sar esi, 31   ; temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+        xor edi, esi  ; temp ^= temp3;
+        sub edi, esi  ; temp -= temp3;
+
+        ; For a negative input, want temp2 = bitwise complement of abs(input)
+        ; This code assumes we are on a two's complement machine
+        add ebx, esi  ; temp2 += temp3;
+
+        ; Find the number of bits needed for the magnitude of the coefficient
+        lea   r11, [rel jpeg_nbits_table]
+        movzx rdi, byte [r11 + rdi]  ; nbits = JPEG_NBITS(temp);
+        ; Emit the Huffman-coded symbol for the number of bits
+        mov   r11d,  INT [r14 + rdi * 4]  ; code = dctbl->ehufco[nbits];
+        movzx  esi, byte [r14 + rdi + 1024]  ; size = dctbl->ehufsi[nbits];
+        EMIT_BITS r11, esi  ; EMIT_BITS(code, size)
+
+        ; Mask off any extra bits in code
+        mov esi, 1
+        mov ecx, edi
+        shl esi, cl
+        dec esi
+        and ebx, esi  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+
+        ; Emit that number of bits of the value, if positive,
+        ; or the complement of its magnitude, if negative.
+        EMIT_BITS rbx, edi  ; EMIT_BITS(temp2, nbits)
+
+        ; Prepare data
+        xor ebx, ebx
+        kloop_prepare  0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, \
+                       18, 11, 4,  5,  12, 19, 26, 33, 40, 48, 41, 34, \
+                       27, 20, 13, 6,  7,  14, 21, 28, 35, \
+                       xmm0, xmm1, xmm2, xmm3
+        kloop_prepare  32, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, \
+                       30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, \
+                       53, 60, 61, 54, 47, 55, 62, 63, 63, \
+                       xmm4, xmm5, xmm6, xmm7
+
+        pxor xmm8, xmm8
+        pcmpeqw xmm0, xmm8  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm8  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm8  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm8  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        pcmpeqw xmm4, xmm8  ; tmp4 = _mm_cmpeq_epi16(tmp4, zero);
+        pcmpeqw xmm5, xmm8  ; tmp5 = _mm_cmpeq_epi16(tmp5, zero);
+        pcmpeqw xmm6, xmm8  ; tmp6 = _mm_cmpeq_epi16(tmp6, zero);
+        pcmpeqw xmm7, xmm8  ; tmp7 = _mm_cmpeq_epi16(tmp7, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        packsswb xmm4, xmm5  ; tmp4 = _mm_packs_epi16(tmp4, tmp5);
+        packsswb xmm6, xmm7  ; tmp6 = _mm_packs_epi16(tmp6, tmp7);
+        pmovmskb r11d, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb r12d, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        pmovmskb r13d, xmm4  ; index  = ((uint64_t)_mm_movemask_epi8(tmp4)) << 32;
+        pmovmskb r14d, xmm6  ; index  = ((uint64_t)_mm_movemask_epi8(tmp6)) << 48;
+        shl r12, 16
+        shl r14, 16
+        or  r11, r12
+        or  r13, r14
+        shl r13, 32
+        or  r11, r13
+        not r11  ; index = ~index;
+
+        ;mov MMWORD [ t1 + DCTSIZE2 * SIZEOF_WORD ], r11
+        ;jmp .EFN
+
+        mov   r13d,  INT [r15 + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx r14d, byte [r15 + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        lea rsi, [t1]
+.BLOOP:
+        bsf r12, r11  ; r = __builtin_ctzl(index);
+        jz .ELOOP
+        mov rcx, r12
+        lea rsi, [rsi+r12*2]  ; k += r;
+        shr r11, cl  ; index >>= r;
+        movzx rdi, word [rsi]  ; temp = t1[k];
+        lea   rbx, [rel jpeg_nbits_table]
+        movzx rdi, byte [rbx + rdi]  ; nbits = JPEG_NBITS(temp);
+.BRLOOP:
+        cmp r12, 16  ; while (r > 15) {
+        jl .ERLOOP
+        EMIT_BITS r13, r14d  ; EMIT_BITS(code_0xf0, size_0xf0)
+        sub r12, 16  ; r -= 16;
+        jmp .BRLOOP
+.ERLOOP:
+        ; Emit Huffman symbol for run length / number of bits
+        CHECKBUF31  ; uses rcx, rdx
+
+        shl r12, 4  ; temp3 = (r << 4) + nbits;
+        add r12, rdi
+        mov   ebx,  INT [r15 + r12 * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [r15 + r12 + 1024]  ; size = actbl->ehufsi[temp3];
+        PUT_BITS rbx
+
+        ;EMIT_CODE(code, size)
+
+        movsx ebx, word [rsi-DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov rcx, rdi
+        mov rdx, 1
+        shl rdx, cl
+        dec rdx
+        and rbx, rdx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        PUT_BITS rbx  ; PUT_BITS(temp2, nbits)
+
+        shr r11, 1  ; index >>= 1;
+        add rsi, 2  ; ++k;
+        jmp .BLOOP
+.ELOOP:
+        ; If the last coef(s) were zero, emit an end-of-block code
+        lea rdi, [t1 + (DCTSIZE2-1) * 2]  ; r = DCTSIZE2-1-k;
+        cmp rdi, rsi  ; if (r > 0) {
+        je .EFN
+        mov   ebx,  INT [r15]  ; code = actbl->ehufco[0];
+        movzx r12d, byte [r15 + 1024]  ; size = actbl->ehufsi[0];
+        EMIT_BITS rbx, r12d
+.EFN:
+        pop r10
+        ; Save put_buffer & put_bits
+        mov MMWORD [r10+16], put_buffer  ; state->cur.put_buffer = put_buffer;
+        mov DWORD  [r10+24], put_bits  ; state->cur.put_bits = put_bits;
+
+        pop rbx
+%ifdef WIN64
+        movaps  xmm11, XMMWORD [rsp+0*SIZEOF_XMMWORD]
+        movaps  xmm10, XMMWORD [rsp+1*SIZEOF_XMMWORD]
+        movaps  xmm9, XMMWORD [rsp+2*SIZEOF_XMMWORD]
+        movaps  xmm8, XMMWORD [rsp+3*SIZEOF_XMMWORD]
+        add     rsp, 4*SIZEOF_XMMWORD
+%endif
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jchuff-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jchuff-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jchuff-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jchuff-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,427 @@
+;
+; jchuff-sse2.asm - Huffman entropy encoding (SSE2)
+;
+; Copyright 2009-2011, 2014-2016 D. R. Commander.
+; Copyright 2015 Matthieu Darbois
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based directly on jchuff.c; see jchuff.c for more
+; details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_huff_encode_one_block) PRIVATE
+
+EXTN(jconst_huff_encode_one_block):
+
+%include "jpeg_nbits_table.inc"
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+; These macros perform the same task as the emit_bits() function in the
+; original libjpeg code.  In addition to reducing overhead by explicitly
+; inlining the code, additional performance is achieved by taking into
+; account the size of the bit buffer and waiting until it is almost full
+; before emptying it.  This mostly benefits 64-bit platforms, since 6
+; bytes can be stored in a 64-bit bit buffer before it has to be emptied.
+
+%macro EMIT_BYTE 0
+        sub put_bits, 8  ; put_bits -= 8;
+        mov edx, put_buffer
+        mov ecx, put_bits
+        shr edx, cl  ; c = (JOCTET)GETJOCTET(put_buffer >> put_bits);
+        mov byte [eax], dl  ; *buffer++ = c;
+        add eax, 1
+        cmp dl, 0xFF  ; need to stuff a zero byte?
+        jne %%.EMIT_BYTE_END
+        mov byte [eax], 0  ; *buffer++ = 0;
+        add eax, 1
+%%.EMIT_BYTE_END:
+%endmacro
+
+%macro PUT_BITS 1
+        add put_bits, ecx  ; put_bits += size;
+        shl put_buffer, cl  ; put_buffer = (put_buffer << size);
+        or  put_buffer, %1
+%endmacro
+
+%macro CHECKBUF15 0
+        cmp put_bits, 16  ; if (put_bits > 31) {
+        jl %%.CHECKBUF15_END
+        mov eax, POINTER [esp+buffer]
+        EMIT_BYTE
+        EMIT_BYTE
+        mov POINTER [esp+buffer], eax
+%%.CHECKBUF15_END:
+%endmacro
+
+%macro EMIT_BITS 1
+        PUT_BITS %1
+        CHECKBUF15
+%endmacro
+
+%macro kloop_prepare 37  ;(ko, jno0, ..., jno31, xmm0, xmm1, xmm2, xmm3)
+    pxor xmm4, xmm4  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm5, xmm5  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm6, xmm6  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm7, xmm7  ; __m128i neg = _mm_setzero_si128();
+    pinsrw %34, word [esi + %2  * SIZEOF_WORD], 0  ; xmm_shadow[0] = block[jno0];
+    pinsrw %35, word [esi + %10 * SIZEOF_WORD], 0  ; xmm_shadow[8] = block[jno8];
+    pinsrw %36, word [esi + %18 * SIZEOF_WORD], 0  ; xmm_shadow[16] = block[jno16];
+    pinsrw %37, word [esi + %26 * SIZEOF_WORD], 0  ; xmm_shadow[24] = block[jno24];
+    pinsrw %34, word [esi + %3  * SIZEOF_WORD], 1  ; xmm_shadow[1] = block[jno1];
+    pinsrw %35, word [esi + %11 * SIZEOF_WORD], 1  ; xmm_shadow[9] = block[jno9];
+    pinsrw %36, word [esi + %19 * SIZEOF_WORD], 1  ; xmm_shadow[17] = block[jno17];
+    pinsrw %37, word [esi + %27 * SIZEOF_WORD], 1  ; xmm_shadow[25] = block[jno25];
+    pinsrw %34, word [esi + %4  * SIZEOF_WORD], 2  ; xmm_shadow[2] = block[jno2];
+    pinsrw %35, word [esi + %12 * SIZEOF_WORD], 2  ; xmm_shadow[10] = block[jno10];
+    pinsrw %36, word [esi + %20 * SIZEOF_WORD], 2  ; xmm_shadow[18] = block[jno18];
+    pinsrw %37, word [esi + %28 * SIZEOF_WORD], 2  ; xmm_shadow[26] = block[jno26];
+    pinsrw %34, word [esi + %5  * SIZEOF_WORD], 3  ; xmm_shadow[3] = block[jno3];
+    pinsrw %35, word [esi + %13 * SIZEOF_WORD], 3  ; xmm_shadow[11] = block[jno11];
+    pinsrw %36, word [esi + %21 * SIZEOF_WORD], 3  ; xmm_shadow[19] = block[jno19];
+    pinsrw %37, word [esi + %29 * SIZEOF_WORD], 3  ; xmm_shadow[27] = block[jno27];
+    pinsrw %34, word [esi + %6  * SIZEOF_WORD], 4  ; xmm_shadow[4] = block[jno4];
+    pinsrw %35, word [esi + %14 * SIZEOF_WORD], 4  ; xmm_shadow[12] = block[jno12];
+    pinsrw %36, word [esi + %22 * SIZEOF_WORD], 4  ; xmm_shadow[20] = block[jno20];
+    pinsrw %37, word [esi + %30 * SIZEOF_WORD], 4  ; xmm_shadow[28] = block[jno28];
+    pinsrw %34, word [esi + %7  * SIZEOF_WORD], 5  ; xmm_shadow[5] = block[jno5];
+    pinsrw %35, word [esi + %15 * SIZEOF_WORD], 5  ; xmm_shadow[13] = block[jno13];
+    pinsrw %36, word [esi + %23 * SIZEOF_WORD], 5  ; xmm_shadow[21] = block[jno21];
+    pinsrw %37, word [esi + %31 * SIZEOF_WORD], 5  ; xmm_shadow[29] = block[jno29];
+    pinsrw %34, word [esi + %8  * SIZEOF_WORD], 6  ; xmm_shadow[6] = block[jno6];
+    pinsrw %35, word [esi + %16 * SIZEOF_WORD], 6  ; xmm_shadow[14] = block[jno14];
+    pinsrw %36, word [esi + %24 * SIZEOF_WORD], 6  ; xmm_shadow[22] = block[jno22];
+    pinsrw %37, word [esi + %32 * SIZEOF_WORD], 6  ; xmm_shadow[30] = block[jno30];
+    pinsrw %34, word [esi + %9  * SIZEOF_WORD], 7  ; xmm_shadow[7] = block[jno7];
+    pinsrw %35, word [esi + %17 * SIZEOF_WORD], 7  ; xmm_shadow[15] = block[jno15];
+    pinsrw %36, word [esi + %25 * SIZEOF_WORD], 7  ; xmm_shadow[23] = block[jno23];
+%if %1 != 32
+    pinsrw %37, word [esi + %33 * SIZEOF_WORD], 7  ; xmm_shadow[31] = block[jno31];
+%else
+    pinsrw %37, ecx, 7  ; xmm_shadow[31] = block[jno31];
+%endif
+    pcmpgtw xmm4, %34  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm5, %35  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm6, %36  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm7, %37  ; neg = _mm_cmpgt_epi16(neg, x1);
+    paddw %34, xmm4   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %35, xmm5   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %36, xmm6  ; x1 = _mm_add_epi16(x1, neg);
+    paddw %37, xmm7  ; x1 = _mm_add_epi16(x1, neg);
+    pxor %34, xmm4    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %35, xmm5    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %36, xmm6   ; x1 = _mm_xor_si128(x1, neg);
+    pxor %37, xmm7   ; x1 = _mm_xor_si128(x1, neg);
+    pxor xmm4, %34    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm5, %35    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm6, %36   ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm7, %37   ; neg = _mm_xor_si128(neg, x1);
+    movdqa XMMWORD [esp + t1 + %1 * SIZEOF_WORD], %34  ; _mm_storeu_si128((__m128i *)(t1 + ko), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 8) * SIZEOF_WORD], %35  ; _mm_storeu_si128((__m128i *)(t1 + ko + 8), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 16) * SIZEOF_WORD], %36  ; _mm_storeu_si128((__m128i *)(t1 + ko + 16), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 24) * SIZEOF_WORD], %37  ; _mm_storeu_si128((__m128i *)(t1 + ko + 24), x1);
+    movdqa XMMWORD [esp + t2 + %1 * SIZEOF_WORD], xmm4  ; _mm_storeu_si128((__m128i *)(t2 + ko), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 8) * SIZEOF_WORD], xmm5  ; _mm_storeu_si128((__m128i *)(t2 + ko + 8), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 16) * SIZEOF_WORD], xmm6  ; _mm_storeu_si128((__m128i *)(t2 + ko + 16), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 24) * SIZEOF_WORD], xmm7  ; _mm_storeu_si128((__m128i *)(t2 + ko + 24), neg);
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET*)
+; jsimd_huff_encode_one_block_sse2 (working_state *state, JOCTET *buffer,
+;                                   JCOEFPTR block, int last_dc_val,
+;                                   c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+
+; eax + 8 = working_state *state
+; eax + 12 = JOCTET *buffer
+; eax + 16 = JCOEFPTR block
+; eax + 20 = int last_dc_val
+; eax + 24 = c_derived_tbl *dctbl
+; eax + 28 = c_derived_tbl *actbl
+
+%define pad             6*SIZEOF_DWORD  ; Align to 16 bytes
+%define t1              pad
+%define t2              t1+(DCTSIZE2*SIZEOF_WORD)
+%define block           t2+(DCTSIZE2*SIZEOF_WORD)
+%define actbl           block+SIZEOF_DWORD
+%define buffer          actbl+SIZEOF_DWORD
+%define temp            buffer+SIZEOF_DWORD
+%define temp2           temp+SIZEOF_DWORD
+%define temp3           temp2+SIZEOF_DWORD
+%define temp4           temp3+SIZEOF_DWORD
+%define temp5           temp4+SIZEOF_DWORD
+%define gotptr          temp5+SIZEOF_DWORD  ; void *gotptr
+%define put_buffer      ebx
+%define put_bits        edi
+
+        align   16
+        global  EXTN(jsimd_huff_encode_one_block_sse2) PRIVATE
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        sub     esp, temp5+9*SIZEOF_DWORD-pad
+        push    ebx
+        push    ecx
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+        push    ebp
+
+        mov esi, POINTER [eax+8]        ; (working_state *state)
+        mov put_buffer,  DWORD [esi+8]  ; put_buffer = state->cur.put_buffer;
+        mov put_bits,    DWORD [esi+12]  ; put_bits = state->cur.put_bits;
+        push esi  ; esi is now scratch
+
+        get_GOT edx                       ; get GOT address
+        movpic POINTER [esp+gotptr], edx  ; save GOT address
+
+        mov ecx, POINTER [eax+28]
+        mov edx, POINTER [eax+16]
+        mov esi, POINTER [eax+12]
+        mov POINTER [esp+actbl],  ecx
+        mov POINTER [esp+block],  edx
+        mov POINTER [esp+buffer], esi
+
+        ; Encode the DC coefficient difference per section F.1.2.1
+        mov esi, POINTER [esp+block]        ; block
+        movsx ecx, word [esi]  ; temp = temp2 = block[0] - last_dc_val;
+        sub   ecx, DWORD [eax+20]
+        mov   esi, ecx
+
+        ; This is a well-known technique for obtaining the absolute value
+        ; without a branch.  It is derived from an assembly language technique
+        ; presented in "How to Optimize for the Pentium Processors",
+        ; Copyright (c) 1996, 1997 by Agner Fog.
+        mov edx, ecx
+        sar edx, 31   ; temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+        xor ecx, edx ; temp ^= temp3;
+        sub ecx, edx ; temp -= temp3;
+
+        ; For a negative input, want temp2 = bitwise complement of abs(input)
+        ; This code assumes we are on a two's complement machine
+        add esi, edx  ; temp2 += temp3;
+        mov DWORD [esp+temp], esi  ; backup temp2 in temp
+
+        ; Find the number of bits needed for the magnitude of the coefficient
+        movpic ebp, POINTER [esp+gotptr]   ; load GOT address (ebp)
+        movzx edx, byte [GOTOFF(ebp, jpeg_nbits_table + ecx)]  ; nbits = JPEG_NBITS(temp);
+        mov DWORD [esp+temp2], edx  ; backup nbits in temp2
+
+        ; Emit the Huffman-coded symbol for the number of bits
+        mov    ebp, POINTER [eax+24]  ; After this point, arguments are not accessible anymore
+        mov    eax,  INT [ebp + edx * 4]  ; code = dctbl->ehufco[nbits];
+        movzx  ecx, byte [ebp + edx + 1024]  ; size = dctbl->ehufsi[nbits];
+        EMIT_BITS eax  ; EMIT_BITS(code, size)
+
+        mov ecx, DWORD [esp+temp2]  ; restore nbits
+
+        ; Mask off any extra bits in code
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, DWORD [esp+temp]  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+
+        ; Emit that number of bits of the value, if positive,
+        ; or the complement of its magnitude, if negative.
+        EMIT_BITS eax  ; EMIT_BITS(temp2, nbits)
+
+        ; Prepare data
+        xor ecx, ecx
+        mov esi, POINTER [esp+block]
+        kloop_prepare  0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, \
+                       18, 11, 4,  5,  12, 19, 26, 33, 40, 48, 41, 34, \
+                       27, 20, 13, 6,  7,  14, 21, 28, 35, \
+                       xmm0, xmm1, xmm2, xmm3
+        kloop_prepare  32, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, \
+                       30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, \
+                       53, 60, 61, 54, 47, 55, 62, 63, 63, \
+                       xmm0, xmm1, xmm2, xmm3
+
+        pxor xmm7, xmm7
+        movdqa xmm0, XMMWORD [esp + t1 + 0 * SIZEOF_WORD]   ; __m128i tmp0 = _mm_loadu_si128((__m128i *)(t1 + 0));
+        movdqa xmm1, XMMWORD [esp + t1 + 8 * SIZEOF_WORD]   ; __m128i tmp1 = _mm_loadu_si128((__m128i *)(t1 + 8));
+        movdqa xmm2, XMMWORD [esp + t1 + 16 * SIZEOF_WORD]  ; __m128i tmp2 = _mm_loadu_si128((__m128i *)(t1 + 16));
+        movdqa xmm3, XMMWORD [esp + t1 + 24 * SIZEOF_WORD]  ; __m128i tmp3 = _mm_loadu_si128((__m128i *)(t1 + 24));
+        pcmpeqw xmm0, xmm7  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm7  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm7  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm7  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        pmovmskb edx, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb ecx, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        shl ecx, 16
+        or  edx, ecx
+        not edx  ; index = ~index;
+
+        lea esi, [esp+t1]
+        mov ebp, POINTER [esp+actbl]  ; ebp = actbl
+
+.BLOOP:
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP
+        lea esi, [esi+ecx*2]  ; k += r;
+        shr edx, cl  ; index >>= r;
+        mov DWORD [esp+temp3], edx
+.BRLOOP:
+        cmp ecx, 16  ; while (r > 15) {
+        jl .ERLOOP
+        sub ecx, 16 ; r -= 16;
+        mov DWORD [esp+temp], ecx
+        mov   eax, INT [ebp + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx ecx, byte [ebp + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        EMIT_BITS eax  ; EMIT_BITS(code_0xf0, size_0xf0)
+        mov ecx, DWORD [esp+temp]
+        jmp .BRLOOP
+.ERLOOP:
+        movsx eax, word [esi]  ; temp = t1[k];
+        movpic edx, POINTER [esp+gotptr]   ; load GOT address (edx)
+        movzx eax, byte [GOTOFF(edx, jpeg_nbits_table + eax)]  ; nbits = JPEG_NBITS(temp);
+        mov DWORD [esp+temp2], eax
+        ; Emit Huffman symbol for run length / number of bits
+        shl ecx, 4  ; temp3 = (r << 4) + nbits;
+        add ecx, eax
+        mov   eax,  INT [ebp + ecx * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [ebp + ecx + 1024]  ; size = actbl->ehufsi[temp3];
+        EMIT_BITS eax
+
+        movsx edx, word [esi+DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov ecx, DWORD [esp+temp2]
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, edx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        EMIT_BITS eax  ; PUT_BITS(temp2, nbits)
+        mov edx, DWORD [esp+temp3]
+        add esi, 2  ; ++k;
+        shr edx, 1  ; index >>= 1;
+
+        jmp .BLOOP
+.ELOOP:
+        movdqa xmm0, XMMWORD [esp + t1 + 32 * SIZEOF_WORD]  ; __m128i tmp0 = _mm_loadu_si128((__m128i *)(t1 + 0));
+        movdqa xmm1, XMMWORD [esp + t1 + 40 * SIZEOF_WORD]  ; __m128i tmp1 = _mm_loadu_si128((__m128i *)(t1 + 8));
+        movdqa xmm2, XMMWORD [esp + t1 + 48 * SIZEOF_WORD]  ; __m128i tmp2 = _mm_loadu_si128((__m128i *)(t1 + 16));
+        movdqa xmm3, XMMWORD [esp + t1 + 56 * SIZEOF_WORD]  ; __m128i tmp3 = _mm_loadu_si128((__m128i *)(t1 + 24));
+        pcmpeqw xmm0, xmm7  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm7  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm7  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm7  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        pmovmskb edx, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb ecx, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        shl ecx, 16
+        or  edx, ecx
+        not edx  ; index = ~index;
+
+        lea eax, [esp + t1 + (DCTSIZE2/2) * 2]
+        sub eax, esi
+        shr eax, 1
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP2
+        shr edx, cl  ; index >>= r;
+        add ecx, eax
+        lea esi, [esi+ecx*2]  ; k += r;
+        mov DWORD [esp+temp3], edx
+        jmp .BRLOOP2
+.BLOOP2:
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP2
+        lea esi, [esi+ecx*2]  ; k += r;
+        shr edx, cl  ; index >>= r;
+        mov DWORD [esp+temp3], edx
+.BRLOOP2:
+        cmp ecx, 16  ; while (r > 15) {
+        jl .ERLOOP2
+        sub ecx, 16  ; r -= 16;
+        mov DWORD [esp+temp], ecx
+        mov   eax, INT [ebp + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx ecx, byte [ebp + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        EMIT_BITS eax  ; EMIT_BITS(code_0xf0, size_0xf0)
+        mov ecx, DWORD [esp+temp]
+        jmp .BRLOOP2
+.ERLOOP2:
+        movsx eax, word [esi]  ; temp = t1[k];
+        bsr eax, eax  ; nbits = 32 - __builtin_clz(temp);
+        inc eax
+        mov DWORD [esp+temp2], eax
+        ; Emit Huffman symbol for run length / number of bits
+        shl ecx, 4  ; temp3 = (r << 4) + nbits;
+        add ecx, eax
+        mov   eax,  INT [ebp + ecx * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [ebp + ecx + 1024]  ; size = actbl->ehufsi[temp3];
+        EMIT_BITS eax
+
+        movsx edx, word [esi+DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov ecx, DWORD [esp+temp2]
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, edx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        EMIT_BITS eax  ; PUT_BITS(temp2, nbits)
+        mov edx, DWORD [esp+temp3]
+        add esi, 2  ; ++k;
+        shr edx, 1  ; index >>= 1;
+
+        jmp .BLOOP2
+.ELOOP2:
+        ; If the last coef(s) were zero, emit an end-of-block code
+        lea edx, [esp + t1 + (DCTSIZE2-1) * 2]  ; r = DCTSIZE2-1-k;
+        cmp edx, esi  ; if (r > 0) {
+        je .EFN
+        mov   eax,  INT [ebp]  ; code = actbl->ehufco[0];
+        movzx ecx, byte [ebp + 1024]  ; size = actbl->ehufsi[0];
+        EMIT_BITS eax
+.EFN:
+        mov eax, [esp+buffer]
+        pop esi
+        ; Save put_buffer & put_bits
+        mov DWORD [esi+8], put_buffer  ; state->cur.put_buffer = put_buffer;
+        mov DWORD [esi+12], put_bits  ; state->cur.put_bits = put_bits;
+
+        pop     ebp
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+        pop     ecx
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcolsamp.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcolsamp.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcolsamp.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcolsamp.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,105 @@
+;
+; jcolsamp.inc - private declarations for color conversion & up/downsampling
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; [TAB8]
+
+; --------------------------------------------------------------------------
+
+; pseudo-resisters to make ordering of RGB configurable
+;
+%if RGB_RED == 0
+%define  mmA  mm0
+%define  mmB  mm1
+%define xmmA xmm0
+%define xmmB xmm1
+%elif RGB_GREEN == 0
+%define  mmA  mm2
+%define  mmB  mm3
+%define xmmA xmm2
+%define xmmB xmm3
+%elif RGB_BLUE == 0
+%define  mmA  mm4
+%define  mmB  mm5
+%define xmmA xmm4
+%define xmmB xmm5
+%else
+%define  mmA  mm6
+%define  mmB  mm7
+%define xmmA xmm6
+%define xmmB xmm7
+%endif
+
+%if RGB_RED == 1
+%define  mmC  mm0
+%define  mmD  mm1
+%define xmmC xmm0
+%define xmmD xmm1
+%elif RGB_GREEN == 1
+%define  mmC  mm2
+%define  mmD  mm3
+%define xmmC xmm2
+%define xmmD xmm3
+%elif RGB_BLUE == 1
+%define  mmC  mm4
+%define  mmD  mm5
+%define xmmC xmm4
+%define xmmD xmm5
+%else
+%define  mmC  mm6
+%define  mmD  mm7
+%define xmmC xmm6
+%define xmmD xmm7
+%endif
+
+%if RGB_RED == 2
+%define  mmE  mm0
+%define  mmF  mm1
+%define xmmE xmm0
+%define xmmF xmm1
+%elif RGB_GREEN == 2
+%define  mmE  mm2
+%define  mmF  mm3
+%define xmmE xmm2
+%define xmmF xmm3
+%elif RGB_BLUE == 2
+%define  mmE  mm4
+%define  mmF  mm5
+%define xmmE xmm4
+%define xmmF xmm5
+%else
+%define  mmE  mm6
+%define  mmF  mm7
+%define xmmE xmm6
+%define xmmF xmm7
+%endif
+
+%if RGB_RED == 3
+%define  mmG  mm0
+%define  mmH  mm1
+%define xmmG xmm0
+%define xmmH xmm1
+%elif RGB_GREEN == 3
+%define  mmG  mm2
+%define  mmH  mm3
+%define xmmG xmm2
+%define xmmH xmm3
+%elif RGB_BLUE == 3
+%define  mmG  mm4
+%define  mmH  mm5
+%define xmmG xmm4
+%define xmmH xmm5
+%else
+%define  mmG  mm6
+%define  mmH  mm7
+%define xmmG xmm6
+%define xmmH xmm7
+%endif
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,158 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA DOWNSAMPLING */
+
+#include "jsimd_altivec.h"
+#include "jcsample.h"
+
+
+void
+jsimd_h2v1_downsample_altivec (JDIMENSION image_width, int max_v_samp_factor,
+                               JDIMENSION v_samp_factor,
+                               JDIMENSION width_blocks,
+                               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow, outcol;
+  JDIMENSION output_cols = width_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+
+  __vector unsigned char this0, next0, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(0, 1) },
+    pw_one = { __8X(1) };
+  __vector unsigned char even_odd_index =
+    {0,2,4,6,8,10,12,14,1,3,5,7,9,11,13,15},
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (outrow = 0; outrow < v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      outl = vec_add(this0e, this0o);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_one);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        outh = vec_add(next0e, next0o);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_one);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_downsample_altivec (JDIMENSION image_width, int max_v_samp_factor,
+                               JDIMENSION v_samp_factor,
+                               JDIMENSION width_blocks,
+                               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, outcol;
+  JDIMENSION output_cols = width_blocks * DCTSIZE;
+  JSAMPROW inptr0, inptr1, outptr;
+
+  __vector unsigned char this0, next0, this1, next1, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, this1e, this1o,
+    next1e, next1o, out0l, out0h, out1l, out1h, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(1, 2) },
+    pw_two = { __8X(2) };
+  __vector unsigned char even_odd_index =
+    { 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15 },
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (inrow = 0, outrow = 0; outrow < v_samp_factor;
+       inrow += 2, outrow++) {
+
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow + 1];
+    outptr = output_data[outrow];
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr0 += 32, inptr1 += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr0);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      out0l = vec_add(this0e, this0o);
+
+      this1 = vec_ld(0, inptr1);
+      this1 = vec_perm(this1, this1, even_odd_index);
+      this1e = (__vector unsigned short)VEC_UNPACKHU(this1);
+      this1o = (__vector unsigned short)VEC_UNPACKLU(this1);
+      out1l = vec_add(this1e, this1o);
+
+      outl = vec_add(out0l, out1l);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_two);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr0);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        out0h = vec_add(next0e, next0o);
+
+        next1 = vec_ld(16, inptr1);
+        next1 = vec_perm(next1, next1, even_odd_index);
+        next1e = (__vector unsigned short)VEC_UNPACKHU(next1);
+        next1o = (__vector unsigned short)VEC_UNPACKLU(next1);
+        out1h = vec_add(next1e, next1o);
+
+        outh = vec_add(out0h, out1h);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_two);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample.h glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+/*
+ * jcsample.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+LOCAL(void)
+expand_right_edge (JSAMPARRAY image_data, int num_rows,
+                   JDIMENSION input_cols, JDIMENSION output_cols)
+{
+  register JSAMPROW ptr;
+  register JSAMPLE pixval;
+  register int count;
+  int row;
+  int numcols = (int) (output_cols - input_cols);
+
+  if (numcols > 0) {
+    for (row = 0; row < num_rows; row++) {
+      ptr = image_data[row] + input_cols;
+      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
+      for (count = numcols; count > 0; count--)
+        *ptr++ = pixval;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,324 @@
+;
+; jcsample.asm - downsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                            JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov       edx, 0x00010000       ; bias pattern
+        movd      mm7,edx
+        pcmpeqw   mm6,mm6
+        punpckldq mm7,mm7               ; mm7={0, 1, 0, 1}
+        psrlw     mm6,BYTE_BIT          ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm1, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mm2,mm0
+        movq    mm3,mm1
+
+        pand    mm0,mm6
+        psrlw   mm2,BYTE_BIT
+        pand    mm1,mm6
+        psrlw   mm3,BYTE_BIT
+
+        paddw   mm0,mm2
+        paddw   mm1,mm3
+        paddw   mm0,mm7
+        paddw   mm1,mm7
+        psrlw   mm0,1
+        psrlw   mm1,1
+
+        packuswb mm0,mm1
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 1*SIZEOF_MMWORD       ; outptr
+        sub     ecx, byte SIZEOF_MMWORD         ; outcol
+        jnz     short .columnloop
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     eax                             ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                            JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov       edx, 0x00020001       ; bias pattern
+        movd      mm7,edx
+        pcmpeqw   mm6,mm6
+        punpckldq mm7,mm7               ; mm7={1, 2, 1, 2}
+        psrlw     mm6,BYTE_BIT          ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     edi, JSAMPROW [edi]                     ; outptr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [edx+0*SIZEOF_MMWORD]
+        movq    mm1, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm2, MMWORD [edx+1*SIZEOF_MMWORD]
+        movq    mm3, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pand    mm0,mm6
+        psrlw   mm4,BYTE_BIT
+        pand    mm1,mm6
+        psrlw   mm5,BYTE_BIT
+        paddw   mm0,mm4
+        paddw   mm1,mm5
+
+        movq    mm4,mm2
+        movq    mm5,mm3
+        pand    mm2,mm6
+        psrlw   mm4,BYTE_BIT
+        pand    mm3,mm6
+        psrlw   mm5,BYTE_BIT
+        paddw   mm2,mm4
+        paddw   mm3,mm5
+
+        paddw   mm0,mm1
+        paddw   mm2,mm3
+        paddw   mm0,mm7
+        paddw   mm2,mm7
+        psrlw   mm0,2
+        psrlw   mm2,2
+
+        packuswb mm0,mm2
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+
+        add     edx, byte 2*SIZEOF_MMWORD       ; inptr0
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr1
+        add     edi, byte 1*SIZEOF_MMWORD       ; outptr
+        sub     ecx, byte SIZEOF_MMWORD         ; outcol
+        jnz     near .columnloop
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,330 @@
+;
+; jcsample.asm - downsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+; r10 = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12 = JDIMENSION v_samp_factor
+; r13 = JDIMENSION width_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov ecx, r13d
+        shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
+        jz      near .return
+
+        mov edx, r10d
+
+        ; -- expand_right_edge
+
+        push    rcx
+        shl     rcx,1                           ; output_cols * 2
+        sub     rcx,rdx
+        jle     short .expand_end
+
+        mov     rax, r11
+        test    rax,rax
+        jle     short .expand_end
+
+        cld
+        mov     rsi, r14        ; input_data
+.expandloop:
+        push    rax
+        push    rcx
+
+        mov     rdi, JSAMPROW [rsi]
+        add     rdi,rdx
+        mov     al, JSAMPLE [rdi-1]
+
+        rep stosb
+
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        dec     rax
+        jg      short .expandloop
+
+.expand_end:
+        pop     rcx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, r12d        ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     rdx, 0x00010000         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     rsi, r14        ; input_data
+        mov     rdi, r15        ; output_data
+.rowloop:
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]             ; inptr
+        mov rdi, JSAMPROW [rdi]         ; outptr
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        pxor    xmm1,xmm1
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .downsample
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm2,xmm0
+        movdqa  xmm3,xmm1
+
+        pand    xmm0,xmm6
+        psrlw   xmm2,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm3,BYTE_BIT
+
+        paddw   xmm0,xmm2
+        paddw   xmm1,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+        psrlw   xmm0,1
+        psrlw   xmm1,1
+
+        packuswb xmm0,xmm1
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     rcx, byte SIZEOF_XMMWORD        ; outcol
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        test    rcx,rcx
+        jnz     short .columnloop_r8
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+; r10 = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12 = JDIMENSION v_samp_factor
+; r13 = JDIMENSION width_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     ecx, r13d
+        shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
+        jz      near .return
+
+        mov     edx, r10d
+
+        ; -- expand_right_edge
+
+        push    rcx
+        shl     rcx,1                           ; output_cols * 2
+        sub     rcx,rdx
+        jle     short .expand_end
+
+        mov     rax, r11
+        test    rax,rax
+        jle     short .expand_end
+
+        cld
+        mov     rsi, r14        ; input_data
+.expandloop:
+        push    rax
+        push    rcx
+
+        mov     rdi, JSAMPROW [rsi]
+        add     rdi,rdx
+        mov     al, JSAMPLE [rdi-1]
+
+        rep stosb
+
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        dec     rax
+        jg      short .expandloop
+
+.expand_end:
+        pop     rcx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, r12d        ; rowctr
+        test    rax,rax
+        jle     near .return
+
+        mov     rdx, 0x00020001         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     rsi, r14        ; input_data
+        mov     rdi, r15        ; output_data
+.rowloop:
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rdx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     rdi, JSAMPROW [rdi]                     ; outptr
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        pxor    xmm2,xmm2
+        pxor    xmm3,xmm3
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .downsample
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        pand    xmm0,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm0,xmm4
+        paddw   xmm1,xmm5
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm5,xmm3
+        pand    xmm2,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm3,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm2,xmm4
+        paddw   xmm3,xmm5
+
+        paddw   xmm0,xmm1
+        paddw   xmm2,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm7
+        psrlw   xmm0,2
+        psrlw   xmm2,2
+
+        packuswb xmm0,xmm2
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     rcx, byte SIZEOF_XMMWORD        ; outcol
+        add     rdx, byte 2*SIZEOF_XMMWORD      ; inptr0
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr1
+        add     rdi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .columnloop_r8
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+
+        add     rsi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     rax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jcsample-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jcsample-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,351 @@
+;
+; jcsample.asm - downsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     edx, 0x00010000         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        pxor    xmm1,xmm1
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .downsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm2,xmm0
+        movdqa  xmm3,xmm1
+
+        pand    xmm0,xmm6
+        psrlw   xmm2,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm3,BYTE_BIT
+
+        paddw   xmm0,xmm2
+        paddw   xmm1,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+        psrlw   xmm0,1
+        psrlw   xmm1,1
+
+        packuswb xmm0,xmm1
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     ecx, byte SIZEOF_XMMWORD        ; outcol
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        test    ecx,ecx
+        jnz     short .columnloop_r8
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     edx, 0x00020001         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     edi, JSAMPROW [edi]                     ; outptr
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        pxor    xmm2,xmm2
+        pxor    xmm3,xmm3
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .downsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2, XMMWORD [edx+1*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        pand    xmm0,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm0,xmm4
+        paddw   xmm1,xmm5
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm5,xmm3
+        pand    xmm2,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm3,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm2,xmm4
+        paddw   xmm3,xmm5
+
+        paddw   xmm0,xmm1
+        paddw   xmm2,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm7
+        psrlw   xmm0,2
+        psrlw   xmm2,2
+
+        packuswb xmm0,xmm2
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     ecx, byte SIZEOF_XMMWORD        ; outcol
+        add     edx, byte 2*SIZEOF_XMMWORD      ; inptr0
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr1
+        add     edi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .columnloop_r8
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,274 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdcolor-altivec.c */
+
+
+void jsimd_ycc_rgb_convert_altivec (JDIMENSION out_width, JSAMPIMAGE input_buf,
+                                    JDIMENSION input_row,
+                                    JSAMPARRAY output_buf, int num_rows)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = out_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, yl, yh, cbl, cbh,
+    crl, crh, rl, rh, gl, gh, bl, bh, g0w, g1w, g2w, g3w;
+  __vector int g0, g1, g2, g3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, outptr += RGB_PIXELSIZE * 16,
+         inptr0 += 16, inptr1 += 16, inptr2 += 16) {
+
+      y = vec_ld(0, inptr0);
+      /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      yl = (__vector signed short)VEC_UNPACKHU(y);
+      yh = (__vector signed short)VEC_UNPACKLU(y);
+
+      cb = vec_ld(0, inptr1);
+      cbl = (__vector signed short)VEC_UNPACKHU(cb);
+      cbh = (__vector signed short)VEC_UNPACKLU(cb);
+      cbl = vec_sub(cbl, pw_cj);
+      cbh = vec_sub(cbh, pw_cj);
+
+      cr = vec_ld(0, inptr2);
+      crl = (__vector signed short)VEC_UNPACKHU(cr);
+      crh = (__vector signed short)VEC_UNPACKLU(cr);
+      crl = vec_sub(crl, pw_cj);
+      crh = vec_sub(crh, pw_cj);
+
+      /* (Original)
+       * R = Y                + 1.40200 * Cr
+       * G = Y - 0.34414 * Cb - 0.71414 * Cr
+       * B = Y + 1.77200 * Cb
+       *
+       * (This implementation)
+       * R = Y                + 0.40200 * Cr + Cr
+       * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+       * B = Y - 0.22800 * Cb + Cb + Cb
+       */
+      bl = vec_add(cbl, cbl);
+      bh = vec_add(cbh, cbh);
+      bl = vec_madds(bl, pw_mf0228, pw_one);
+      bh = vec_madds(bh, pw_mf0228, pw_one);
+      bl = vec_sra(bl, (__vector unsigned short)pw_one);
+      bh = vec_sra(bh, (__vector unsigned short)pw_one);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, yl);
+      bh = vec_add(bh, yh);
+
+      rl = vec_add(crl, crl);
+      rh = vec_add(crh, crh);
+      rl = vec_madds(rl, pw_f0402, pw_one);
+      rh = vec_madds(rh, pw_f0402, pw_one);
+      rl = vec_sra(rl, (__vector unsigned short)pw_one);
+      rh = vec_sra(rh, (__vector unsigned short)pw_one);
+      rl = vec_add(rl, crl);
+      rh = vec_add(rh, crh);
+      rl = vec_add(rl, yl);
+      rh = vec_add(rh, yh);
+
+      g0w = vec_mergeh(cbl, crl);
+      g1w = vec_mergel(cbl, crl);
+      g0 = vec_msums(g0w, pw_mf0344_f0285, pd_onehalf);
+      g1 = vec_msums(g1w, pw_mf0344_f0285, pd_onehalf);
+      g2w = vec_mergeh(cbh, crh);
+      g3w = vec_mergel(cbh, crh);
+      g2 = vec_msums(g2w, pw_mf0344_f0285, pd_onehalf);
+      g3 = vec_msums(g3w, pw_mf0344_f0285, pd_onehalf);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      gl = vec_perm((__vector short)g0, (__vector short)g1, shift_pack_index);
+      gh = vec_perm((__vector short)g2, (__vector short)g3, shift_pack_index);
+      gl = vec_sub(gl, crl);
+      gh = vec_sub(gh, crh);
+      gl = vec_add(gl, yl);
+      gh = vec_add(gh, yh);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,405 @@
+;
+; jdcolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_mmx (JDIMENSION out_width,
+;                            JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                            JSAMPARRAY output_buf, int num_rows)
+;
+
+%define out_width(b)    (b)+8           ; JDIMENSION out_width
+%define input_buf(b)    (b)+12          ; JSAMPIMAGE input_buf
+%define input_row(b)    (b)+16          ; JDIMENSION input_row
+%define output_buf(b)   (b)+20          ; JSAMPARRAY output_buf
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_mmx) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [out_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [input_row(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        push    eax
+        push    edi
+        push    edx
+        push    ebx
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr0
+        mov     ebx, JSAMPROW [ebx]     ; inptr1
+        mov     edx, JSAMPROW [edx]     ; inptr2
+        mov     edi, JSAMPROW [edi]     ; outptr
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+        alignx  16,7
+.columnloop:
+
+        movq    mm5, MMWORD [ebx]       ; mm5=Cb(01234567)
+        movq    mm1, MMWORD [edx]       ; mm1=Cr(01234567)
+
+        pcmpeqw mm4,mm4
+        pcmpeqw mm7,mm7
+        psrlw   mm4,BYTE_BIT
+        psllw   mm7,7                   ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+        movq    mm0,mm4                 ; mm0=mm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    mm4,mm5                 ; mm4=Cb(0246)=CbE
+        psrlw   mm5,BYTE_BIT            ; mm5=Cb(1357)=CbO
+        pand    mm0,mm1                 ; mm0=Cr(0246)=CrE
+        psrlw   mm1,BYTE_BIT            ; mm1=Cr(1357)=CrO
+
+        paddw   mm4,mm7
+        paddw   mm5,mm7
+        paddw   mm0,mm7
+        paddw   mm1,mm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movq    mm2,mm4                 ; mm2=CbE
+        movq    mm3,mm5                 ; mm3=CbO
+        paddw   mm4,mm4                 ; mm4=2*CbE
+        paddw   mm5,mm5                 ; mm5=2*CbO
+        movq    mm6,mm0                 ; mm6=CrE
+        movq    mm7,mm1                 ; mm7=CrO
+        paddw   mm0,mm0                 ; mm0=2*CrE
+        paddw   mm1,mm1                 ; mm1=2*CrO
+
+        pmulhw  mm4,[GOTOFF(eax,PW_MF0228)]     ; mm4=(2*CbE * -FIX(0.22800))
+        pmulhw  mm5,[GOTOFF(eax,PW_MF0228)]     ; mm5=(2*CbO * -FIX(0.22800))
+        pmulhw  mm0,[GOTOFF(eax,PW_F0402)]      ; mm0=(2*CrE * FIX(0.40200))
+        pmulhw  mm1,[GOTOFF(eax,PW_F0402)]      ; mm1=(2*CrO * FIX(0.40200))
+
+        paddw   mm4,[GOTOFF(eax,PW_ONE)]
+        paddw   mm5,[GOTOFF(eax,PW_ONE)]
+        psraw   mm4,1                   ; mm4=(CbE * -FIX(0.22800))
+        psraw   mm5,1                   ; mm5=(CbO * -FIX(0.22800))
+        paddw   mm0,[GOTOFF(eax,PW_ONE)]
+        paddw   mm1,[GOTOFF(eax,PW_ONE)]
+        psraw   mm0,1                   ; mm0=(CrE * FIX(0.40200))
+        psraw   mm1,1                   ; mm1=(CrO * FIX(0.40200))
+
+        paddw   mm4,mm2
+        paddw   mm5,mm3
+        paddw   mm4,mm2                 ; mm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   mm5,mm3                 ; mm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   mm0,mm6                 ; mm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   mm1,mm7                 ; mm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(B-Y)E
+        movq    MMWORD [wk(1)], mm5     ; wk(1)=(B-Y)O
+
+        movq      mm4,mm2
+        movq      mm5,mm3
+        punpcklwd mm2,mm6
+        punpckhwd mm4,mm6
+        pmaddwd   mm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd mm3,mm7
+        punpckhwd mm5,mm7
+        pmaddwd   mm3,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     mm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm4,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm2,SCALEBITS
+        psrad     mm4,SCALEBITS
+        paddd     mm3,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm5,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm3,SCALEBITS
+        psrad     mm5,SCALEBITS
+
+        packssdw  mm2,mm4       ; mm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  mm3,mm5       ; mm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     mm2,mm6       ; mm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     mm3,mm7       ; mm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movq      mm5, MMWORD [esi]     ; mm5=Y(01234567)
+
+        pcmpeqw   mm4,mm4
+        psrlw     mm4,BYTE_BIT          ; mm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      mm4,mm5               ; mm4=Y(0246)=YE
+        psrlw     mm5,BYTE_BIT          ; mm5=Y(1357)=YO
+
+        paddw     mm0,mm4               ; mm0=((R-Y)E+YE)=RE=(R0 R2 R4 R6)
+        paddw     mm1,mm5               ; mm1=((R-Y)O+YO)=RO=(R1 R3 R5 R7)
+        packuswb  mm0,mm0               ; mm0=(R0 R2 R4 R6 ** ** ** **)
+        packuswb  mm1,mm1               ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+        paddw     mm2,mm4               ; mm2=((G-Y)E+YE)=GE=(G0 G2 G4 G6)
+        paddw     mm3,mm5               ; mm3=((G-Y)O+YO)=GO=(G1 G3 G5 G7)
+        packuswb  mm2,mm2               ; mm2=(G0 G2 G4 G6 ** ** ** **)
+        packuswb  mm3,mm3               ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+        paddw     mm4, MMWORD [wk(0)]   ; mm4=(YE+(B-Y)E)=BE=(B0 B2 B4 B6)
+        paddw     mm5, MMWORD [wk(1)]   ; mm5=(YO+(B-Y)O)=BO=(B1 B3 B5 B7)
+        packuswb  mm4,mm4               ; mm4=(B0 B2 B4 B6 ** ** ** **)
+        packuswb  mm5,mm5               ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmB               ; mmE=(20 01 22 03 24 05 26 07)
+        punpcklbw mmD,mmF               ; mmD=(11 21 13 23 15 25 17 27)
+
+        movq      mmG,mmA
+        movq      mmH,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 01 02 12 22 03)
+        punpckhwd mmG,mmE               ; mmG=(04 14 24 05 06 16 26 07)
+
+        psrlq     mmH,2*BYTE_BIT        ; mmH=(02 12 04 14 06 16 -- --)
+        psrlq     mmE,2*BYTE_BIT        ; mmE=(22 03 24 05 26 07 -- --)
+
+        movq      mmC,mmD
+        movq      mmB,mmD
+        punpcklwd mmD,mmH               ; mmD=(11 21 02 12 13 23 04 14)
+        punpckhwd mmC,mmH               ; mmC=(15 25 06 16 17 27 -- --)
+
+        psrlq     mmB,2*BYTE_BIT        ; mmB=(13 23 15 25 17 27 -- --)
+
+        movq      mmF,mmE
+        punpcklwd mmE,mmB               ; mmE=(22 03 13 23 24 05 15 25)
+        punpckhwd mmF,mmB               ; mmF=(26 07 17 27 -- -- -- --)
+
+        punpckldq mmA,mmD               ; mmA=(00 10 20 01 11 21 02 12)
+        punpckldq mmE,mmG               ; mmE=(22 03 13 23 04 14 24 05)
+        punpckldq mmC,mmF               ; mmC=(15 25 06 16 26 07 17 27)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        lea     ecx, [ecx+ecx*2]        ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_MMWORD
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    mmA,mmC
+        sub     ecx, byte 2*SIZEOF_MMWORD
+        add     edi, byte 2*SIZEOF_MMWORD
+        jmp     short .column_st4
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmE
+        sub     ecx, byte SIZEOF_MMWORD
+        add     edi, byte SIZEOF_MMWORD
+.column_st4:
+        movd    eax,mmA
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st2
+        mov     DWORD [edi+0*SIZEOF_DWORD], eax
+        psrlq   mmA,DWORD_BIT
+        movd    eax,mmA
+        sub     ecx, byte SIZEOF_DWORD
+        add     edi, byte SIZEOF_DWORD
+.column_st2:
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi+0*SIZEOF_WORD], ax
+        shr     eax,WORD_BIT
+        sub     ecx, byte SIZEOF_WORD
+        add     edi, byte SIZEOF_WORD
+.column_st1:
+        cmp     ecx, byte SIZEOF_BYTE
+        jb      short .nextrow
+        mov     BYTE [edi+0*SIZEOF_BYTE], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pcmpeqb   mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+        pxor      mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pxor      mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmG               ; mmE=(20 30 22 32 24 34 26 36)
+        punpcklbw mmB,mmD               ; mmB=(01 11 03 13 05 15 07 17)
+        punpcklbw mmF,mmH               ; mmF=(21 31 23 33 25 35 27 37)
+
+        movq      mmC,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 30 02 12 22 32)
+        punpckhwd mmC,mmE               ; mmC=(04 14 24 34 06 16 26 36)
+        movq      mmG,mmB
+        punpcklwd mmB,mmF               ; mmB=(01 11 21 31 03 13 23 33)
+        punpckhwd mmG,mmF               ; mmG=(05 15 25 35 07 17 27 37)
+
+        movq      mmD,mmA
+        punpckldq mmA,mmB               ; mmA=(00 10 20 30 01 11 21 31)
+        punpckhdq mmD,mmB               ; mmD=(02 12 22 32 03 13 23 33)
+        movq      mmH,mmC
+        punpckldq mmC,mmG               ; mmC=(04 14 24 34 05 15 25 35)
+        punpckhdq mmH,mmG               ; mmH=(06 16 26 36 07 17 27 37)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mmH
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        cmp     ecx, byte SIZEOF_MMWORD/2
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    mmA,mmC
+        movq    mmD,mmH
+        sub     ecx, byte SIZEOF_MMWORD/2
+        add     edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD/4
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmD
+        sub     ecx, byte SIZEOF_MMWORD/4
+        add     edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+        cmp     ecx, byte SIZEOF_MMWORD/8
+        jb      short .nextrow
+        movd    DWORD [edi+0*SIZEOF_DWORD], mmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        alignx  16,7
+
+.nextrow:
+        pop     ecx
+        pop     esi
+        pop     ebx
+        pop     edx
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        add     edi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,441 @@
+;
+; jdcolext.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009, 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width,
+;                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                             JSAMPARRAY output_buf, int num_rows)
+;
+
+; r10 = JDIMENSION out_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION input_row
+; r13 = JSAMPARRAY output_buf
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d        ; num_cols
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        lea     rsi, [rsi+rcx*SIZEOF_JSAMPROW]
+        lea     rbx, [rbx+rcx*SIZEOF_JSAMPROW]
+        lea     rdx, [rdx+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov     rdi, r13
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rax
+        push    rdi
+        push    rdx
+        push    rbx
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr0
+        mov     rbx, JSAMPROW [rbx]     ; inptr1
+        mov     rdx, JSAMPROW [rdx]     ; inptr2
+        mov     rdi, JSAMPROW [rdi]     ; outptr
+.columnloop:
+
+        movdqa  xmm5, XMMWORD [rbx]     ; xmm5=Cb(0123456789ABCDEF)
+        movdqa  xmm1, XMMWORD [rdx]     ; xmm1=Cr(0123456789ABCDEF)
+
+        pcmpeqw xmm4,xmm4
+        pcmpeqw xmm7,xmm7
+        psrlw   xmm4,BYTE_BIT
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+        movdqa  xmm0,xmm4               ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    xmm4,xmm5               ; xmm4=Cb(02468ACE)=CbE
+        psrlw   xmm5,BYTE_BIT           ; xmm5=Cb(13579BDF)=CbO
+        pand    xmm0,xmm1               ; xmm0=Cr(02468ACE)=CrE
+        psrlw   xmm1,BYTE_BIT           ; xmm1=Cr(13579BDF)=CrO
+
+        paddw   xmm4,xmm7
+        paddw   xmm5,xmm7
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm2,xmm4               ; xmm2=CbE
+        movdqa  xmm3,xmm5               ; xmm3=CbO
+        paddw   xmm4,xmm4               ; xmm4=2*CbE
+        paddw   xmm5,xmm5               ; xmm5=2*CbO
+        movdqa  xmm6,xmm0               ; xmm6=CrE
+        movdqa  xmm7,xmm1               ; xmm7=CrO
+        paddw   xmm0,xmm0               ; xmm0=2*CrE
+        paddw   xmm1,xmm1               ; xmm1=2*CrO
+
+        pmulhw  xmm4,[rel PW_MF0228]    ; xmm4=(2*CbE * -FIX(0.22800))
+        pmulhw  xmm5,[rel PW_MF0228]    ; xmm5=(2*CbO * -FIX(0.22800))
+        pmulhw  xmm0,[rel PW_F0402]     ; xmm0=(2*CrE * FIX(0.40200))
+        pmulhw  xmm1,[rel PW_F0402]     ; xmm1=(2*CrO * FIX(0.40200))
+
+        paddw   xmm4,[rel PW_ONE]
+        paddw   xmm5,[rel PW_ONE]
+        psraw   xmm4,1                  ; xmm4=(CbE * -FIX(0.22800))
+        psraw   xmm5,1                  ; xmm5=(CbO * -FIX(0.22800))
+        paddw   xmm0,[rel PW_ONE]
+        paddw   xmm1,[rel PW_ONE]
+        psraw   xmm0,1                  ; xmm0=(CrE * FIX(0.40200))
+        psraw   xmm1,1                  ; xmm1=(CrO * FIX(0.40200))
+
+        paddw   xmm4,xmm2
+        paddw   xmm5,xmm3
+        paddw   xmm4,xmm2               ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   xmm5,xmm3               ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   xmm0,xmm6               ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   xmm1,xmm7               ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=(B-Y)E
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(B-Y)O
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm5,xmm3
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm4,xmm6
+        pmaddwd   xmm2,[rel PW_MF0344_F0285]
+        pmaddwd   xmm4,[rel PW_MF0344_F0285]
+        punpcklwd xmm3,xmm7
+        punpckhwd xmm5,xmm7
+        pmaddwd   xmm3,[rel PW_MF0344_F0285]
+        pmaddwd   xmm5,[rel PW_MF0344_F0285]
+
+        paddd     xmm2,[rel PD_ONEHALF]
+        paddd     xmm4,[rel PD_ONEHALF]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm4,SCALEBITS
+        paddd     xmm3,[rel PD_ONEHALF]
+        paddd     xmm5,[rel PD_ONEHALF]
+        psrad     xmm3,SCALEBITS
+        psrad     xmm5,SCALEBITS
+
+        packssdw  xmm2,xmm4     ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  xmm3,xmm5     ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     xmm2,xmm6     ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     xmm3,xmm7     ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movdqa    xmm5, XMMWORD [rsi]   ; xmm5=Y(0123456789ABCDEF)
+
+        pcmpeqw   xmm4,xmm4
+        psrlw     xmm4,BYTE_BIT         ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      xmm4,xmm5             ; xmm4=Y(02468ACE)=YE
+        psrlw     xmm5,BYTE_BIT         ; xmm5=Y(13579BDF)=YO
+
+        paddw     xmm0,xmm4             ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm5             ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm4             ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm5             ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+        paddw     xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        lea     rcx, [rcx+rcx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     rcx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     rcx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_MMWORD
+        sub     rcx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_DWORD
+        sub     rcx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of rax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     rcx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [rdi], ax
+        add     rdi, byte SIZEOF_WORD
+        sub     rcx, byte SIZEOF_WORD
+        shr     rax, 16
+.column_st1:
+        ; Store the lower 1 byte of rax to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .nextrow
+        mov     BYTE [rdi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        cmp     rcx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_XMMWORD/8*4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .nextrow
+        movd    XMM_DWORD [rdi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.nextrow:
+        pop     rcx
+        pop     rsi
+        pop     rbx
+        pop     rdx
+        pop     rdi
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        add     rbx, byte SIZEOF_JSAMPROW
+        add     rdx, byte SIZEOF_JSAMPROW
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+        sfence          ; flush the write buffer
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,460 @@
+;
+; jdcolext.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width,
+;                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                             JSAMPARRAY output_buf, int num_rows)
+;
+
+%define out_width(b)    (b)+8           ; JDIMENSION out_width
+%define input_buf(b)    (b)+12          ; JSAMPIMAGE input_buf
+%define input_row(b)    (b)+16          ; JDIMENSION input_row
+%define output_buf(b)   (b)+20          ; JSAMPARRAY output_buf
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [out_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [input_row(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        push    eax
+        push    edi
+        push    edx
+        push    ebx
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr0
+        mov     ebx, JSAMPROW [ebx]     ; inptr1
+        mov     edx, JSAMPROW [edx]     ; inptr2
+        mov     edi, JSAMPROW [edi]     ; outptr
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm5, XMMWORD [ebx]     ; xmm5=Cb(0123456789ABCDEF)
+        movdqa  xmm1, XMMWORD [edx]     ; xmm1=Cr(0123456789ABCDEF)
+
+        pcmpeqw xmm4,xmm4
+        pcmpeqw xmm7,xmm7
+        psrlw   xmm4,BYTE_BIT
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+        movdqa  xmm0,xmm4               ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    xmm4,xmm5               ; xmm4=Cb(02468ACE)=CbE
+        psrlw   xmm5,BYTE_BIT           ; xmm5=Cb(13579BDF)=CbO
+        pand    xmm0,xmm1               ; xmm0=Cr(02468ACE)=CrE
+        psrlw   xmm1,BYTE_BIT           ; xmm1=Cr(13579BDF)=CrO
+
+        paddw   xmm4,xmm7
+        paddw   xmm5,xmm7
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm2,xmm4               ; xmm2=CbE
+        movdqa  xmm3,xmm5               ; xmm3=CbO
+        paddw   xmm4,xmm4               ; xmm4=2*CbE
+        paddw   xmm5,xmm5               ; xmm5=2*CbO
+        movdqa  xmm6,xmm0               ; xmm6=CrE
+        movdqa  xmm7,xmm1               ; xmm7=CrO
+        paddw   xmm0,xmm0               ; xmm0=2*CrE
+        paddw   xmm1,xmm1               ; xmm1=2*CrO
+
+        pmulhw  xmm4,[GOTOFF(eax,PW_MF0228)]    ; xmm4=(2*CbE * -FIX(0.22800))
+        pmulhw  xmm5,[GOTOFF(eax,PW_MF0228)]    ; xmm5=(2*CbO * -FIX(0.22800))
+        pmulhw  xmm0,[GOTOFF(eax,PW_F0402)]     ; xmm0=(2*CrE * FIX(0.40200))
+        pmulhw  xmm1,[GOTOFF(eax,PW_F0402)]     ; xmm1=(2*CrO * FIX(0.40200))
+
+        paddw   xmm4,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm5,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm4,1                  ; xmm4=(CbE * -FIX(0.22800))
+        psraw   xmm5,1                  ; xmm5=(CbO * -FIX(0.22800))
+        paddw   xmm0,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm1,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm0,1                  ; xmm0=(CrE * FIX(0.40200))
+        psraw   xmm1,1                  ; xmm1=(CrO * FIX(0.40200))
+
+        paddw   xmm4,xmm2
+        paddw   xmm5,xmm3
+        paddw   xmm4,xmm2               ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   xmm5,xmm3               ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   xmm0,xmm6               ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   xmm1,xmm7               ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=(B-Y)E
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(B-Y)O
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm5,xmm3
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm4,xmm6
+        pmaddwd   xmm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd xmm3,xmm7
+        punpckhwd xmm5,xmm7
+        pmaddwd   xmm3,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     xmm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm4,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm4,SCALEBITS
+        paddd     xmm3,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm5,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm3,SCALEBITS
+        psrad     xmm5,SCALEBITS
+
+        packssdw  xmm2,xmm4     ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  xmm3,xmm5     ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     xmm2,xmm6     ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     xmm3,xmm7     ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movdqa    xmm5, XMMWORD [esi]   ; xmm5=Y(0123456789ABCDEF)
+
+        pcmpeqw   xmm4,xmm4
+        psrlw     xmm4,BYTE_BIT         ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      xmm4,xmm5             ; xmm4=Y(02468ACE)=YE
+        psrlw     xmm5,BYTE_BIT         ; xmm5=Y(13579BDF)=YO
+
+        paddw     xmm0,xmm4             ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm5             ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm4             ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm5             ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+        paddw     xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        lea     ecx, [ecx+ecx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     ecx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_MMWORD
+        sub     ecx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [edi], xmmA
+        add     edi, byte SIZEOF_DWORD
+        sub     ecx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of eax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi], ax
+        add     edi, byte SIZEOF_WORD
+        sub     ecx, byte SIZEOF_WORD
+        shr     eax, 16
+.column_st1:
+        ; Store the lower 1 byte of eax to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .nextrow
+        mov     BYTE [edi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        cmp     ecx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_XMMWORD/8*4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .nextrow
+        movd    XMM_DWORD [edi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        alignx  16,7
+
+.nextrow:
+        pop     ecx
+        pop     esi
+        pop     ebx
+        pop     edx
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        add     edi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        sfence          ; flush the write buffer
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,96 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* YCC --> RGB CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344 22554              /* FIX(0.34414) */
+#define F_0_714 46802              /* FIX(0.71414) */
+#define F_1_402 91881              /* FIX(1.40200) */
+#define F_1_772 116130             /* FIX(1.77200) */
+#define F_0_402 (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285 (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228 (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 {0,1,8,2,3,10,4,5,12,6,7,14,16,17,24,18}
+#define RGB_INDEX1 {3,10,4,5,12,6,7,14,16,17,24,18,19,26,20,21}
+#define RGB_INDEX2 {12,6,7,14,16,17,24,18,19,26,20,21,28,22,23,30}
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGB_INDEX {0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extrgbx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 {8,1,0,10,3,2,12,5,4,14,7,6,24,17,16,26}
+#define RGB_INDEX1 {3,2,12,5,4,14,7,6,24,17,16,26,19,18,28,21}
+#define RGB_INDEX2 {4,14,7,6,24,17,16,26,19,18,28,21,20,30,23,22}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGB_INDEX {8,1,0,9,10,3,2,11,12,5,4,13,14,7,6,15}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extbgrx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGB_INDEX {9,8,1,0,11,10,3,2,13,12,5,4,15,14,7,6}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extxbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGB_INDEX {9,0,1,8,11,2,3,10,13,4,5,12,15,6,7,14}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extxrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_mmx) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgb_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgbx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgrx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxrgb_convert_mmx
+%include "jdcolext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdcolor-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdct.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdct.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdct.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdct.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+;
+; jdct.inc - private declarations for forward & reverse DCT subsystems
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; [TAB8]
+
+; Each IDCT routine is responsible for range-limiting its results and
+; converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+; be quite far out of range if the input data is corrupt, so a bulletproof
+; range-limiting step is required.  We use a mask-and-table-lookup method
+; to do the combined operations quickly.
+;
+%define RANGE_MASK  (MAXJSAMPLE * 4 + 3)  ; 2 bits wider than legal samples
+
+%define ROW(n,b,s)              ((b)+(n)*(s))
+%define COL(n,b,s)              ((b)+(n)*(s)*DCTSIZE)
+
+%define DWBLOCK(m,n,b,s)        ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_DWORD)
+%define MMBLOCK(m,n,b,s)        ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_MMWORD)
+%define XMMBLOCK(m,n,b,s)       ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_XMMWORD)
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,108 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* MERGED YCC --> RGB CONVERSION AND UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344 22554              /* FIX(0.34414) */
+#define F_0_714 46802              /* FIX(0.71414) */
+#define F_1_402 91881              /* FIX(1.40200) */
+#define F_1_772 116130             /* FIX(1.77200) */
+#define F_0_402 (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285 (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228 (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 {0,1,8,2,3,10,4,5,12,6,7,14,16,17,24,18}
+#define RGB_INDEX1 {3,10,4,5,12,6,7,14,16,17,24,18,19,26,20,21}
+#define RGB_INDEX2 {12,6,7,14,16,17,24,18,19,26,20,21,28,22,23,30}
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGB_INDEX {0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extrgbx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extrgbx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 {8,1,0,10,3,2,12,5,4,14,7,6,24,17,16,26}
+#define RGB_INDEX1 {3,2,12,5,4,14,7,6,24,17,16,26,19,18,28,21}
+#define RGB_INDEX2 {4,14,7,6,24,17,16,26,19,18,28,21,20,30,23,22}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGB_INDEX {8,1,0,9,10,3,2,11,12,5,4,13,14,7,6,15}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extbgrx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extbgrx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGB_INDEX {9,8,1,0,11,10,3,2,13,12,5,4,15,14,7,6}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extxbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extxbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGB_INDEX {9,0,1,8,11,2,3,10,13,4,5,12,15,6,7,14}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extxrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extxrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_mmx) PRIVATE
+
+EXTN(jconst_merged_upsample_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgbx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgbx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgrx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgrx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_sse2) PRIVATE
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmerge-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_sse2) PRIVATE
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,323 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdmerge-altivec.c */
+
+
+void jsimd_h2v1_merged_upsample_altivec (JDIMENSION output_width,
+                                         JSAMPIMAGE input_buf,
+                                         JDIMENSION in_row_group_ctr,
+                                         JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = output_width * RGB_PIXELSIZE, num_cols, yloop;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, ye, yo, cbl, cbh,
+    crl, crh, r_yl, r_yh, g_yl, g_yh, b_yl, b_yh, g_y0w, g_y1w, g_y2w, g_y3w,
+    rl, rh, gl, gh, bl, bh, re, ro, ge, go, be, bo;
+  __vector int g_y0, g_y1, g_y2, g_y3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29},
+    even_index = {0,16,0,18,0,20,0,22,0,24,0,26,0,28,0,30},
+    odd_index = {0,17,0,19,0,21,0,23,0,25,0,27,0,29,0,31};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31},
+    even_index = {16,0,18,0,20,0,22,0,24,0,26,0,28,0,30,0},
+    odd_index = {17,0,19,0,21,0,23,0,25,0,27,0,29,0,31,0};
+#endif
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  for (num_cols = pitch; num_cols > 0; inptr1 += 16, inptr2 += 16) {
+
+    cb = vec_ld(0, inptr1);
+    /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+     * support unsigned vectors.
+     */
+    cbl = (__vector signed short)VEC_UNPACKHU(cb);
+    cbh = (__vector signed short)VEC_UNPACKLU(cb);
+    cbl = vec_sub(cbl, pw_cj);
+    cbh = vec_sub(cbh, pw_cj);
+
+    cr = vec_ld(0, inptr2);
+    crl = (__vector signed short)VEC_UNPACKHU(cr);
+    crh = (__vector signed short)VEC_UNPACKLU(cr);
+    crl = vec_sub(crl, pw_cj);
+    crh = vec_sub(crh, pw_cj);
+
+    /* (Original)
+     * R = Y                + 1.40200 * Cr
+     * G = Y - 0.34414 * Cb - 0.71414 * Cr
+     * B = Y + 1.77200 * Cb
+     *
+     * (This implementation)
+     * R = Y                + 0.40200 * Cr + Cr
+     * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+     * B = Y - 0.22800 * Cb + Cb + Cb
+     */
+    b_yl = vec_add(cbl, cbl);
+    b_yh = vec_add(cbh, cbh);
+    b_yl = vec_madds(b_yl, pw_mf0228, pw_one);
+    b_yh = vec_madds(b_yh, pw_mf0228, pw_one);
+    b_yl = vec_sra(b_yl, (__vector unsigned short)pw_one);
+    b_yh = vec_sra(b_yh, (__vector unsigned short)pw_one);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+
+    r_yl = vec_add(crl, crl);
+    r_yh = vec_add(crh, crh);
+    r_yl = vec_madds(r_yl, pw_f0402, pw_one);
+    r_yh = vec_madds(r_yh, pw_f0402, pw_one);
+    r_yl = vec_sra(r_yl, (__vector unsigned short)pw_one);
+    r_yh = vec_sra(r_yh, (__vector unsigned short)pw_one);
+    r_yl = vec_add(r_yl, crl);
+    r_yh = vec_add(r_yh, crh);
+
+    g_y0w = vec_mergeh(cbl, crl);
+    g_y1w = vec_mergel(cbl, crl);
+    g_y0 = vec_msums(g_y0w, pw_mf0344_f0285, pd_onehalf);
+    g_y1 = vec_msums(g_y1w, pw_mf0344_f0285, pd_onehalf);
+    g_y2w = vec_mergeh(cbh, crh);
+    g_y3w = vec_mergel(cbh, crh);
+    g_y2 = vec_msums(g_y2w, pw_mf0344_f0285, pd_onehalf);
+    g_y3 = vec_msums(g_y3w, pw_mf0344_f0285, pd_onehalf);
+    /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+     * each dword into a new 16-bit vector, which is the equivalent of
+     * descaling the 32-bit results (right-shifting by 16 bits) and then
+     * packing them.
+     */
+    g_yl = vec_perm((__vector short)g_y0, (__vector short)g_y1,
+                    shift_pack_index);
+    g_yh = vec_perm((__vector short)g_y2, (__vector short)g_y3,
+                    shift_pack_index);
+    g_yl = vec_sub(g_yl, crl);
+    g_yh = vec_sub(g_yh, crh);
+
+    for (yloop = 0; yloop < 2 && num_cols > 0; yloop++,
+         num_cols -= RGB_PIXELSIZE * 16,
+         outptr += RGB_PIXELSIZE * 16, inptr0 += 16) {
+
+      y = vec_ld(0, inptr0);
+      ye = (__vector signed short)vec_perm(pb_zero, y, even_index);
+      yo = (__vector signed short)vec_perm(pb_zero, y, odd_index);
+
+      if (yloop == 0) {
+        be = vec_add(b_yl, ye);
+        bo = vec_add(b_yl, yo);
+        re = vec_add(r_yl, ye);
+        ro = vec_add(r_yl, yo);
+        ge = vec_add(g_yl, ye);
+        go = vec_add(g_yl, yo);
+      } else {
+        be = vec_add(b_yh, ye);
+        bo = vec_add(b_yh, yo);
+        re = vec_add(r_yh, ye);
+        ro = vec_add(r_yh, yo);
+        ge = vec_add(g_yh, ye);
+        go = vec_add(g_yh, yo);
+      }
+
+      rl = vec_mergeh(re, ro);
+      rh = vec_mergel(re, ro);
+      gl = vec_mergeh(ge, go);
+      gh = vec_mergel(ge, go);
+      bl = vec_mergeh(be, bo);
+      bh = vec_mergel(be, bo);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}
+
+
+void jsimd_h2v2_merged_upsample_altivec (JDIMENSION output_width,
+                                         JSAMPIMAGE input_buf,
+                                         JDIMENSION in_row_group_ctr,
+                                         JSAMPARRAY output_buf)
+{
+  JSAMPROW inptr, outptr;
+
+  inptr = input_buf[0][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  input_buf[0][in_row_group_ctr] = input_buf[0][in_row_group_ctr * 2];
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf[0][in_row_group_ctr] = input_buf[0][in_row_group_ctr * 2 + 1];
+  output_buf[0] = output_buf[1];
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf[0][in_row_group_ctr] = inptr;
+  output_buf[0] = outptr;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,464 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_mmx (JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          3
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [output_width(eax)]     ; col
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     edi, JSAMPROW [edi]                             ; outptr
+
+        pop     ecx                     ; col
+
+        alignx  16,7
+.columnloop:
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        movq      mm6, MMWORD [ebx]     ; mm6=Cb(01234567)
+        movq      mm7, MMWORD [edx]     ; mm7=Cr(01234567)
+
+        pxor      mm1,mm1               ; mm1=(all 0's)
+        pcmpeqw   mm3,mm3
+        psllw     mm3,7                 ; mm3={0xFF80 0xFF80 0xFF80 0xFF80}
+
+        movq      mm4,mm6
+        punpckhbw mm6,mm1               ; mm6=Cb(4567)=CbH
+        punpcklbw mm4,mm1               ; mm4=Cb(0123)=CbL
+        movq      mm0,mm7
+        punpckhbw mm7,mm1               ; mm7=Cr(4567)=CrH
+        punpcklbw mm0,mm1               ; mm0=Cr(0123)=CrL
+
+        paddw     mm6,mm3
+        paddw     mm4,mm3
+        paddw     mm7,mm3
+        paddw     mm0,mm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movq    mm5,mm6                 ; mm5=CbH
+        movq    mm2,mm4                 ; mm2=CbL
+        paddw   mm6,mm6                 ; mm6=2*CbH
+        paddw   mm4,mm4                 ; mm4=2*CbL
+        movq    mm1,mm7                 ; mm1=CrH
+        movq    mm3,mm0                 ; mm3=CrL
+        paddw   mm7,mm7                 ; mm7=2*CrH
+        paddw   mm0,mm0                 ; mm0=2*CrL
+
+        pmulhw  mm6,[GOTOFF(eax,PW_MF0228)]     ; mm6=(2*CbH * -FIX(0.22800))
+        pmulhw  mm4,[GOTOFF(eax,PW_MF0228)]     ; mm4=(2*CbL * -FIX(0.22800))
+        pmulhw  mm7,[GOTOFF(eax,PW_F0402)]      ; mm7=(2*CrH * FIX(0.40200))
+        pmulhw  mm0,[GOTOFF(eax,PW_F0402)]      ; mm0=(2*CrL * FIX(0.40200))
+
+        paddw   mm6,[GOTOFF(eax,PW_ONE)]
+        paddw   mm4,[GOTOFF(eax,PW_ONE)]
+        psraw   mm6,1                   ; mm6=(CbH * -FIX(0.22800))
+        psraw   mm4,1                   ; mm4=(CbL * -FIX(0.22800))
+        paddw   mm7,[GOTOFF(eax,PW_ONE)]
+        paddw   mm0,[GOTOFF(eax,PW_ONE)]
+        psraw   mm7,1                   ; mm7=(CrH * FIX(0.40200))
+        psraw   mm0,1                   ; mm0=(CrL * FIX(0.40200))
+
+        paddw   mm6,mm5
+        paddw   mm4,mm2
+        paddw   mm6,mm5                 ; mm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   mm4,mm2                 ; mm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   mm7,mm1                 ; mm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   mm0,mm3                 ; mm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movq    MMWORD [wk(0)], mm6     ; wk(0)=(B-Y)H
+        movq    MMWORD [wk(1)], mm7     ; wk(1)=(R-Y)H
+
+        movq      mm6,mm5
+        movq      mm7,mm2
+        punpcklwd mm5,mm1
+        punpckhwd mm6,mm1
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm6,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd mm2,mm3
+        punpckhwd mm7,mm3
+        pmaddwd   mm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     mm5,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm6,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm5,SCALEBITS
+        psrad     mm6,SCALEBITS
+        paddd     mm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm7,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm2,SCALEBITS
+        psrad     mm7,SCALEBITS
+
+        packssdw  mm5,mm6       ; mm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  mm2,mm7       ; mm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     mm5,mm1       ; mm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     mm2,mm3       ; mm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movq    MMWORD [wk(2)], mm5     ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+        alignx  16,7
+
+.Yloop_2nd:
+        movq    mm0, MMWORD [wk(1)]     ; mm0=(R-Y)H
+        movq    mm2, MMWORD [wk(2)]     ; mm2=(G-Y)H
+        movq    mm4, MMWORD [wk(0)]     ; mm4=(B-Y)H
+        alignx  16,7
+
+.Yloop_1st:
+        movq    mm7, MMWORD [esi]       ; mm7=Y(01234567)
+
+        pcmpeqw mm6,mm6
+        psrlw   mm6,BYTE_BIT            ; mm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    mm6,mm7                 ; mm6=Y(0246)=YE
+        psrlw   mm7,BYTE_BIT            ; mm7=Y(1357)=YO
+
+        movq    mm1,mm0                 ; mm1=mm0=(R-Y)(L/H)
+        movq    mm3,mm2                 ; mm3=mm2=(G-Y)(L/H)
+        movq    mm5,mm4                 ; mm5=mm4=(B-Y)(L/H)
+
+        paddw     mm0,mm6               ; mm0=((R-Y)+YE)=RE=(R0 R2 R4 R6)
+        paddw     mm1,mm7               ; mm1=((R-Y)+YO)=RO=(R1 R3 R5 R7)
+        packuswb  mm0,mm0               ; mm0=(R0 R2 R4 R6 ** ** ** **)
+        packuswb  mm1,mm1               ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+        paddw     mm2,mm6               ; mm2=((G-Y)+YE)=GE=(G0 G2 G4 G6)
+        paddw     mm3,mm7               ; mm3=((G-Y)+YO)=GO=(G1 G3 G5 G7)
+        packuswb  mm2,mm2               ; mm2=(G0 G2 G4 G6 ** ** ** **)
+        packuswb  mm3,mm3               ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+        paddw     mm4,mm6               ; mm4=((B-Y)+YE)=BE=(B0 B2 B4 B6)
+        paddw     mm5,mm7               ; mm5=((B-Y)+YO)=BO=(B1 B3 B5 B7)
+        packuswb  mm4,mm4               ; mm4=(B0 B2 B4 B6 ** ** ** **)
+        packuswb  mm5,mm5               ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmB               ; mmE=(20 01 22 03 24 05 26 07)
+        punpcklbw mmD,mmF               ; mmD=(11 21 13 23 15 25 17 27)
+
+        movq      mmG,mmA
+        movq      mmH,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 01 02 12 22 03)
+        punpckhwd mmG,mmE               ; mmG=(04 14 24 05 06 16 26 07)
+
+        psrlq     mmH,2*BYTE_BIT        ; mmH=(02 12 04 14 06 16 -- --)
+        psrlq     mmE,2*BYTE_BIT        ; mmE=(22 03 24 05 26 07 -- --)
+
+        movq      mmC,mmD
+        movq      mmB,mmD
+        punpcklwd mmD,mmH               ; mmD=(11 21 02 12 13 23 04 14)
+        punpckhwd mmC,mmH               ; mmC=(15 25 06 16 17 27 -- --)
+
+        psrlq     mmB,2*BYTE_BIT        ; mmB=(13 23 15 25 17 27 -- --)
+
+        movq      mmF,mmE
+        punpcklwd mmE,mmB               ; mmE=(22 03 13 23 24 05 15 25)
+        punpckhwd mmF,mmB               ; mmF=(26 07 17 27 -- -- -- --)
+
+        punpckldq mmA,mmD               ; mmA=(00 10 20 01 11 21 02 12)
+        punpckldq mmE,mmG               ; mmE=(22 03 13 23 04 14 24 05)
+        punpckldq mmC,mmF               ; mmC=(15 25 06 16 26 07 17 27)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      near .endcolumn
+
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        lea     ecx, [ecx+ecx*2]        ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_MMWORD
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    mmA,mmC
+        sub     ecx, byte 2*SIZEOF_MMWORD
+        add     edi, byte 2*SIZEOF_MMWORD
+        jmp     short .column_st4
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmE
+        sub     ecx, byte SIZEOF_MMWORD
+        add     edi, byte SIZEOF_MMWORD
+.column_st4:
+        movd    eax,mmA
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st2
+        mov     DWORD [edi+0*SIZEOF_DWORD], eax
+        psrlq   mmA,DWORD_BIT
+        movd    eax,mmA
+        sub     ecx, byte SIZEOF_DWORD
+        add     edi, byte SIZEOF_DWORD
+.column_st2:
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi+0*SIZEOF_WORD], ax
+        shr     eax,WORD_BIT
+        sub     ecx, byte SIZEOF_WORD
+        add     edi, byte SIZEOF_WORD
+.column_st1:
+        cmp     ecx, byte SIZEOF_BYTE
+        jb      short .endcolumn
+        mov     BYTE [edi+0*SIZEOF_BYTE], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pcmpeqb   mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+        pxor      mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pxor      mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmG               ; mmE=(20 30 22 32 24 34 26 36)
+        punpcklbw mmB,mmD               ; mmB=(01 11 03 13 05 15 07 17)
+        punpcklbw mmF,mmH               ; mmF=(21 31 23 33 25 35 27 37)
+
+        movq      mmC,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 30 02 12 22 32)
+        punpckhwd mmC,mmE               ; mmC=(04 14 24 34 06 16 26 36)
+        movq      mmG,mmB
+        punpcklwd mmB,mmF               ; mmB=(01 11 21 31 03 13 23 33)
+        punpckhwd mmG,mmF               ; mmG=(05 15 25 35 07 17 27 37)
+
+        movq      mmD,mmA
+        punpckldq mmA,mmB               ; mmA=(00 10 20 30 01 11 21 31)
+        punpckhdq mmD,mmB               ; mmD=(02 12 22 32 03 13 23 33)
+        movq      mmH,mmC
+        punpckldq mmC,mmG               ; mmC=(04 14 24 34 05 15 25 35)
+        punpckhdq mmH,mmG               ; mmH=(06 16 26 36 07 17 27 37)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mmH
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .endcolumn
+
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        cmp     ecx, byte SIZEOF_MMWORD/2
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    mmA,mmC
+        movq    mmD,mmH
+        sub     ecx, byte SIZEOF_MMWORD/2
+        add     edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD/4
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmD
+        sub     ecx, byte SIZEOF_MMWORD/4
+        add     edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+        cmp     ecx, byte SIZEOF_MMWORD/8
+        jb      short .endcolumn
+        movd    DWORD [edi+0*SIZEOF_DWORD], mmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_mmx (JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     eax, JDIMENSION [output_width(ebp)]
+
+        mov     edi, JSAMPIMAGE [input_buf(ebp)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(ebp)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(ebp)]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+
+        push    edx                     ; inptr2
+        push    ebx                     ; inptr1
+        push    esi                     ; inptr00
+        mov     ebx,esp
+
+        push    edi                     ; output_buf (outptr0)
+        push    ecx                     ; in_row_group_ctr
+        push    ebx                     ; input_buf
+        push    eax                     ; output_width
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+        add     esi, byte SIZEOF_JSAMPROW       ; inptr01
+        add     edi, byte SIZEOF_JSAMPROW       ; outptr1
+        mov     POINTER [ebx+0*SIZEOF_POINTER], esi
+        mov     POINTER [ebx-1*SIZEOF_POINTER], edi
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+        add     esp, byte 7*SIZEOF_DWORD
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,538 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009, 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+; r10 = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          3
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d        ; col
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        mov     rdi, r13
+        mov     rsi, JSAMPROW [rsi+rcx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     rbx, JSAMPROW [rbx+rcx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     rdx, JSAMPROW [rdx+rcx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     rdi, JSAMPROW [rdi]                             ; outptr
+
+        pop     rcx                     ; col
+
+.columnloop:
+
+        movdqa    xmm6, XMMWORD [rbx]   ; xmm6=Cb(0123456789ABCDEF)
+        movdqa    xmm7, XMMWORD [rdx]   ; xmm7=Cr(0123456789ABCDEF)
+
+        pxor      xmm1,xmm1             ; xmm1=(all 0's)
+        pcmpeqw   xmm3,xmm3
+        psllw     xmm3,7                ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        movdqa    xmm4,xmm6
+        punpckhbw xmm6,xmm1             ; xmm6=Cb(89ABCDEF)=CbH
+        punpcklbw xmm4,xmm1             ; xmm4=Cb(01234567)=CbL
+        movdqa    xmm0,xmm7
+        punpckhbw xmm7,xmm1             ; xmm7=Cr(89ABCDEF)=CrH
+        punpcklbw xmm0,xmm1             ; xmm0=Cr(01234567)=CrL
+
+        paddw     xmm6,xmm3
+        paddw     xmm4,xmm3
+        paddw     xmm7,xmm3
+        paddw     xmm0,xmm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm5,xmm6               ; xmm5=CbH
+        movdqa  xmm2,xmm4               ; xmm2=CbL
+        paddw   xmm6,xmm6               ; xmm6=2*CbH
+        paddw   xmm4,xmm4               ; xmm4=2*CbL
+        movdqa  xmm1,xmm7               ; xmm1=CrH
+        movdqa  xmm3,xmm0               ; xmm3=CrL
+        paddw   xmm7,xmm7               ; xmm7=2*CrH
+        paddw   xmm0,xmm0               ; xmm0=2*CrL
+
+        pmulhw  xmm6,[rel PW_MF0228]    ; xmm6=(2*CbH * -FIX(0.22800))
+        pmulhw  xmm4,[rel PW_MF0228]    ; xmm4=(2*CbL * -FIX(0.22800))
+        pmulhw  xmm7,[rel PW_F0402]     ; xmm7=(2*CrH * FIX(0.40200))
+        pmulhw  xmm0,[rel PW_F0402]     ; xmm0=(2*CrL * FIX(0.40200))
+
+        paddw   xmm6,[rel PW_ONE]
+        paddw   xmm4,[rel PW_ONE]
+        psraw   xmm6,1                  ; xmm6=(CbH * -FIX(0.22800))
+        psraw   xmm4,1                  ; xmm4=(CbL * -FIX(0.22800))
+        paddw   xmm7,[rel PW_ONE]
+        paddw   xmm0,[rel PW_ONE]
+        psraw   xmm7,1                  ; xmm7=(CrH * FIX(0.40200))
+        psraw   xmm0,1                  ; xmm0=(CrL * FIX(0.40200))
+
+        paddw   xmm6,xmm5
+        paddw   xmm4,xmm2
+        paddw   xmm6,xmm5               ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   xmm4,xmm2               ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   xmm7,xmm1               ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   xmm0,xmm3               ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=(B-Y)H
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(R-Y)H
+
+        movdqa    xmm6,xmm5
+        movdqa    xmm7,xmm2
+        punpcklwd xmm5,xmm1
+        punpckhwd xmm6,xmm1
+        pmaddwd   xmm5,[rel PW_MF0344_F0285]
+        pmaddwd   xmm6,[rel PW_MF0344_F0285]
+        punpcklwd xmm2,xmm3
+        punpckhwd xmm7,xmm3
+        pmaddwd   xmm2,[rel PW_MF0344_F0285]
+        pmaddwd   xmm7,[rel PW_MF0344_F0285]
+
+        paddd     xmm5,[rel PD_ONEHALF]
+        paddd     xmm6,[rel PD_ONEHALF]
+        psrad     xmm5,SCALEBITS
+        psrad     xmm6,SCALEBITS
+        paddd     xmm2,[rel PD_ONEHALF]
+        paddd     xmm7,[rel PD_ONEHALF]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm7,SCALEBITS
+
+        packssdw  xmm5,xmm6     ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  xmm2,xmm7     ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     xmm5,xmm1     ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     xmm2,xmm3     ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movdqa  XMMWORD [wk(2)], xmm5   ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+
+.Yloop_2nd:
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(R-Y)H
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(G-Y)H
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(B-Y)H
+
+.Yloop_1st:
+        movdqa  xmm7, XMMWORD [rsi]     ; xmm7=Y(0123456789ABCDEF)
+
+        pcmpeqw xmm6,xmm6
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    xmm6,xmm7               ; xmm6=Y(02468ACE)=YE
+        psrlw   xmm7,BYTE_BIT           ; xmm7=Y(13579BDF)=YO
+
+        movdqa  xmm1,xmm0               ; xmm1=xmm0=(R-Y)(L/H)
+        movdqa  xmm3,xmm2               ; xmm3=xmm2=(G-Y)(L/H)
+        movdqa  xmm5,xmm4               ; xmm5=xmm4=(B-Y)(L/H)
+
+        paddw     xmm0,xmm6             ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm7             ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm6             ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm7             ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4,xmm6             ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+        paddw     xmm5,xmm7             ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        lea     rcx, [rcx+rcx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     rcx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     rcx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_MMWORD
+        sub     rcx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_DWORD
+        sub     rcx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of rax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     rcx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [rdi], ax
+        add     rdi, byte SIZEOF_WORD
+        sub     rcx, byte SIZEOF_WORD
+        shr     rax, 16
+.column_st1:
+        ; Store the lower 1 byte of rax to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .endcolumn
+        mov     BYTE [rdi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        cmp     rcx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_XMMWORD/8*4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .endcolumn
+        movd    XMM_DWORD [rdi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        sfence          ; flush the write buffer
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+; r10 = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        mov     eax, r10d
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        mov     rdi, r13
+        lea     rsi, [rsi+rcx*SIZEOF_JSAMPROW]
+
+        push    rdx                     ; inptr2
+        push    rbx                     ; inptr1
+        push    rsi                     ; inptr00
+        mov     rbx,rsp
+
+        push    rdi
+        push    rcx
+        push    rax
+
+        %ifdef WIN64
+        mov r8, rcx
+        mov r9, rdi
+        mov rcx, rax
+        mov rdx, rbx
+        %else
+        mov rdx, rcx
+        mov rcx, rdi
+        mov     rdi, rax
+        mov rsi, rbx
+        %endif
+
+        call    EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        pop rax
+        pop rcx
+        pop rdi
+        pop rsi
+        pop rbx
+        pop rdx
+
+        add     rdi, byte SIZEOF_JSAMPROW       ; outptr1
+        add     rsi, byte SIZEOF_JSAMPROW       ; inptr01
+
+        push    rdx                     ; inptr2
+        push    rbx                     ; inptr1
+        push    rsi                     ; inptr00
+        mov     rbx,rsp
+
+        push    rdi
+        push    rcx
+        push    rax
+
+        %ifdef WIN64
+        mov r8, rcx
+        mov r9, rdi
+        mov rcx, rax
+        mov rdx, rbx
+        %else
+        mov rdx, rcx
+        mov rcx, rdi
+        mov     rdi, rax
+        mov rsi, rbx
+        %endif
+
+        call    EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        pop rax
+        pop rcx
+        pop rdi
+        pop rsi
+        pop rbx
+        pop rdx
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdmrgext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,519 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          3
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [output_width(eax)]     ; col
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     edi, JSAMPROW [edi]                             ; outptr
+
+        pop     ecx                     ; col
+
+        alignx  16,7
+.columnloop:
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        movdqa    xmm6, XMMWORD [ebx]   ; xmm6=Cb(0123456789ABCDEF)
+        movdqa    xmm7, XMMWORD [edx]   ; xmm7=Cr(0123456789ABCDEF)
+
+        pxor      xmm1,xmm1             ; xmm1=(all 0's)
+        pcmpeqw   xmm3,xmm3
+        psllw     xmm3,7                ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        movdqa    xmm4,xmm6
+        punpckhbw xmm6,xmm1             ; xmm6=Cb(89ABCDEF)=CbH
+        punpcklbw xmm4,xmm1             ; xmm4=Cb(01234567)=CbL
+        movdqa    xmm0,xmm7
+        punpckhbw xmm7,xmm1             ; xmm7=Cr(89ABCDEF)=CrH
+        punpcklbw xmm0,xmm1             ; xmm0=Cr(01234567)=CrL
+
+        paddw     xmm6,xmm3
+        paddw     xmm4,xmm3
+        paddw     xmm7,xmm3
+        paddw     xmm0,xmm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm5,xmm6               ; xmm5=CbH
+        movdqa  xmm2,xmm4               ; xmm2=CbL
+        paddw   xmm6,xmm6               ; xmm6=2*CbH
+        paddw   xmm4,xmm4               ; xmm4=2*CbL
+        movdqa  xmm1,xmm7               ; xmm1=CrH
+        movdqa  xmm3,xmm0               ; xmm3=CrL
+        paddw   xmm7,xmm7               ; xmm7=2*CrH
+        paddw   xmm0,xmm0               ; xmm0=2*CrL
+
+        pmulhw  xmm6,[GOTOFF(eax,PW_MF0228)]    ; xmm6=(2*CbH * -FIX(0.22800))
+        pmulhw  xmm4,[GOTOFF(eax,PW_MF0228)]    ; xmm4=(2*CbL * -FIX(0.22800))
+        pmulhw  xmm7,[GOTOFF(eax,PW_F0402)]     ; xmm7=(2*CrH * FIX(0.40200))
+        pmulhw  xmm0,[GOTOFF(eax,PW_F0402)]     ; xmm0=(2*CrL * FIX(0.40200))
+
+        paddw   xmm6,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm4,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm6,1                  ; xmm6=(CbH * -FIX(0.22800))
+        psraw   xmm4,1                  ; xmm4=(CbL * -FIX(0.22800))
+        paddw   xmm7,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm0,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm7,1                  ; xmm7=(CrH * FIX(0.40200))
+        psraw   xmm0,1                  ; xmm0=(CrL * FIX(0.40200))
+
+        paddw   xmm6,xmm5
+        paddw   xmm4,xmm2
+        paddw   xmm6,xmm5               ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   xmm4,xmm2               ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   xmm7,xmm1               ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   xmm0,xmm3               ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=(B-Y)H
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(R-Y)H
+
+        movdqa    xmm6,xmm5
+        movdqa    xmm7,xmm2
+        punpcklwd xmm5,xmm1
+        punpckhwd xmm6,xmm1
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm6,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd xmm2,xmm3
+        punpckhwd xmm7,xmm3
+        pmaddwd   xmm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     xmm5,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm6,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm5,SCALEBITS
+        psrad     xmm6,SCALEBITS
+        paddd     xmm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm7,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm7,SCALEBITS
+
+        packssdw  xmm5,xmm6     ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  xmm2,xmm7     ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     xmm5,xmm1     ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     xmm2,xmm3     ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movdqa  XMMWORD [wk(2)], xmm5   ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+        alignx  16,7
+
+.Yloop_2nd:
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(R-Y)H
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(G-Y)H
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(B-Y)H
+        alignx  16,7
+
+.Yloop_1st:
+        movdqa  xmm7, XMMWORD [esi]     ; xmm7=Y(0123456789ABCDEF)
+
+        pcmpeqw xmm6,xmm6
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    xmm6,xmm7               ; xmm6=Y(02468ACE)=YE
+        psrlw   xmm7,BYTE_BIT           ; xmm7=Y(13579BDF)=YO
+
+        movdqa  xmm1,xmm0               ; xmm1=xmm0=(R-Y)(L/H)
+        movdqa  xmm3,xmm2               ; xmm3=xmm2=(G-Y)(L/H)
+        movdqa  xmm5,xmm4               ; xmm5=xmm4=(B-Y)(L/H)
+
+        paddw     xmm0,xmm6             ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm7             ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm6             ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm7             ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4,xmm6             ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+        paddw     xmm5,xmm7             ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        lea     ecx, [ecx+ecx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     ecx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_MMWORD
+        sub     ecx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [edi], xmmA
+        add     edi, byte SIZEOF_DWORD
+        sub     ecx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of eax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi], ax
+        add     edi, byte SIZEOF_WORD
+        sub     ecx, byte SIZEOF_WORD
+        shr     eax, 16
+.column_st1:
+        ; Store the lower 1 byte of eax to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .endcolumn
+        mov     BYTE [edi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        cmp     ecx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_XMMWORD/8*4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .endcolumn
+        movd    XMM_DWORD [edi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        sfence          ; flush the write buffer
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     eax, POINTER [output_width(ebp)]
+
+        mov     edi, JSAMPIMAGE [input_buf(ebp)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(ebp)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(ebp)]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+
+        push    edx                     ; inptr2
+        push    ebx                     ; inptr1
+        push    esi                     ; inptr00
+        mov     ebx,esp
+
+        push    edi                     ; output_buf (outptr0)
+        push    ecx                     ; in_row_group_ctr
+        push    ebx                     ; input_buf
+        push    eax                     ; output_width
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        add     esi, byte SIZEOF_JSAMPROW       ; inptr01
+        add     edi, byte SIZEOF_JSAMPROW       ; outptr1
+        mov     POINTER [ebx+0*SIZEOF_POINTER], esi
+        mov     POINTER [ebx-1*SIZEOF_POINTER], edi
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        add     esp, byte 7*SIZEOF_DWORD
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,392 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+void
+jsimd_h2v1_fancy_upsample_altivec (int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char this0, last0, p_last0, next0 = {0}, p_next0,
+    out;
+  __vector short this0e, this0o, this0l, this0h, last0l, last0h,
+    next0l, next0h, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) }, pb_three = { __16X(3) },
+    last_index_col0 = {0,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14},
+    last_index = {15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30},
+    next_index = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16},
+    next_index_lastcol = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,15},
+#if __BIG_ENDIAN__
+    merge_pack_index = {1,17,3,19,5,21,7,23,9,25,11,27,13,29,15,31};
+#else
+    merge_pack_index = {0,16,2,18,4,20,6,22,8,24,10,26,12,28,14,30};
+#endif
+  __vector short pw_one = { __8X(1) }, pw_two = { __8X(2) };
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+
+    if (downsampled_width & 15)
+      inptr[downsampled_width] = inptr[downsampled_width - 1];
+
+    this0 = vec_ld(0, inptr);
+    p_last0 = vec_perm(this0, this0, last_index_col0);
+    last0 = this0;
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr += 16, outptr += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_last0 = vec_perm(last0, this0, last_index);
+        last0 = this0;
+      }
+
+      if (incol <= 16)
+        p_next0 = vec_perm(this0, this0, next_index_lastcol);
+      else {
+        next0 = vec_ld(16, inptr);
+        p_next0 = vec_perm(this0, next0, next_index);
+      }
+
+      this0e = (__vector short)vec_mule(this0, pb_three);
+      this0o = (__vector short)vec_mulo(this0, pb_three);
+      this0l = vec_mergeh(this0e, this0o);
+      this0h = vec_mergel(this0e, this0o);
+
+      last0l = (__vector short)VEC_UNPACKHU(p_last0);
+      last0h = (__vector short)VEC_UNPACKLU(p_last0);
+      last0l = vec_add(last0l, pw_one);
+
+      next0l = (__vector short)VEC_UNPACKHU(p_next0);
+      next0h = (__vector short)VEC_UNPACKLU(p_next0);
+      next0l = vec_add(next0l, pw_two);
+
+      outle = vec_add(this0l, last0l);
+      outlo = vec_add(this0l, next0l);
+      outle = vec_sr(outle, (__vector unsigned short)pw_two);
+      outlo = vec_sr(outlo, (__vector unsigned short)pw_two);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr);
+
+      if (incol > 8) {
+        last0h = vec_add(last0h, pw_one);
+        next0h = vec_add(next0h, pw_two);
+
+        outhe = vec_add(this0h, last0h);
+        outho = vec_add(this0h, next0h);
+        outhe = vec_sr(outhe, (__vector unsigned short)pw_two);
+        outho = vec_sr(outho, (__vector unsigned short)pw_two);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr);
+      }
+
+      this0 = next0;
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_fancy_upsample_altivec (int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr_1, inptr0, inptr1, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char this_1, this0, this1, out;
+  __vector short this_1l, this_1h, this0l, this0h, this1l, this1h,
+    lastcolsum_1h, lastcolsum1h,
+    p_lastcolsum_1l, p_lastcolsum_1h, p_lastcolsum1l, p_lastcolsum1h,
+    thiscolsum_1l, thiscolsum_1h, thiscolsum1l, thiscolsum1h,
+    nextcolsum_1l = {0}, nextcolsum_1h = {0},
+    nextcolsum1l = {0}, nextcolsum1h = {0},
+    p_nextcolsum_1l, p_nextcolsum_1h, p_nextcolsum1l, p_nextcolsum1h,
+    tmpl, tmph, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) },
+    last_index_col0 = {0,1,0,1,2,3,4,5,6,7,8,9,10,11,12,13},
+    last_index={14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29},
+    next_index = {2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17},
+    next_index_lastcol = {2,3,4,5,6,7,8,9,10,11,12,13,14,15,14,15},
+#if __BIG_ENDIAN__
+    merge_pack_index = {1,17,3,19,5,21,7,23,9,25,11,27,13,29,15,31};
+#else
+    merge_pack_index = {0,16,2,18,4,20,6,22,8,24,10,26,12,28,14,30};
+#endif
+  __vector short pw_zero = { __8X(0) }, pw_three = { __8X(3) },
+    pw_seven = { __8X(7) }, pw_eight = { __8X(8) };
+  __vector unsigned short pw_four = { __8X(4) };
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr_1 = input_data[inrow - 1];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow + 1];
+    outptr0 = output_data[outrow++];
+    outptr1 = output_data[outrow++];
+
+    if (downsampled_width & 15) {
+      inptr_1[downsampled_width] = inptr_1[downsampled_width - 1];
+      inptr0[downsampled_width] = inptr0[downsampled_width - 1];
+      inptr1[downsampled_width] = inptr1[downsampled_width - 1];
+    }
+
+    this0 = vec_ld(0, inptr0);
+    this0l = (__vector short)VEC_UNPACKHU(this0);
+    this0h = (__vector short)VEC_UNPACKLU(this0);
+    this0l = vec_mladd(this0l, pw_three, pw_zero);
+    this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+    this_1 = vec_ld(0, inptr_1);
+    this_1l = (__vector short)VEC_UNPACKHU(this_1);
+    this_1h = (__vector short)VEC_UNPACKLU(this_1);
+    thiscolsum_1l = vec_add(this0l, this_1l);
+    thiscolsum_1h = vec_add(this0h, this_1h);
+    lastcolsum_1h = thiscolsum_1h;
+    p_lastcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1l, last_index_col0);
+    p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+
+    this1 = vec_ld(0, inptr1);
+    this1l = (__vector short)VEC_UNPACKHU(this1);
+    this1h = (__vector short)VEC_UNPACKLU(this1);
+    thiscolsum1l = vec_add(this0l, this1l);
+    thiscolsum1h = vec_add(this0h, this1h);
+    lastcolsum1h = thiscolsum1h;
+    p_lastcolsum1l = vec_perm(thiscolsum1l, thiscolsum1l, last_index_col0);
+    p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr_1 += 16, inptr0 += 16, inptr1 += 16,
+         outptr0 += 32, outptr1 += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_lastcolsum_1l = vec_perm(lastcolsum_1h, thiscolsum_1l, last_index);
+        p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+        p_lastcolsum1l = vec_perm(lastcolsum1h, thiscolsum1l, last_index);
+        p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+        lastcolsum_1h = thiscolsum_1h;  lastcolsum1h = thiscolsum1h;
+      }
+
+      if (incol <= 16) {
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, thiscolsum_1h,
+                                   next_index_lastcol);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, thiscolsum1h,
+                                  next_index_lastcol);
+      } else {
+        this0 = vec_ld(16, inptr0);
+        this0l = (__vector short)VEC_UNPACKHU(this0);
+        this0h = (__vector short)VEC_UNPACKLU(this0);
+        this0l = vec_mladd(this0l, pw_three, pw_zero);
+        this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+        this_1 = vec_ld(16, inptr_1);
+        this_1l = (__vector short)VEC_UNPACKHU(this_1);
+        this_1h = (__vector short)VEC_UNPACKLU(this_1);
+        nextcolsum_1l = vec_add(this0l, this_1l);
+        nextcolsum_1h = vec_add(this0h, this_1h);
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, nextcolsum_1l, next_index);
+
+        this1 = vec_ld(16, inptr1);
+        this1l = (__vector short)VEC_UNPACKHU(this1);
+        this1h = (__vector short)VEC_UNPACKLU(this1);
+        nextcolsum1l = vec_add(this0l, this1l);
+        nextcolsum1h = vec_add(this0h, this1h);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, nextcolsum1l, next_index);
+      }
+
+      /* Process the upper row */
+
+      tmpl = vec_mladd(thiscolsum_1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum_1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum_1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr0);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum_1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum_1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum_1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr0);
+      }
+
+      /* Process the lower row */
+
+      tmpl = vec_mladd(thiscolsum1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr1);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr1);
+      }
+
+      thiscolsum_1l = nextcolsum_1l;  thiscolsum_1h = nextcolsum_1h;
+      thiscolsum1l = nextcolsum1l;  thiscolsum1h = nextcolsum1h;
+    }
+  }
+}
+
+
+/* These are rarely used (mainly just for decompressing YCCK images) */
+
+void
+jsimd_h2v1_upsample_altivec (int max_v_samp_factor,
+                             JDIMENSION output_width,
+                             JSAMPARRAY input_data,
+                             JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr);
+      vec_st(inh, 16, outptr);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr);
+        vec_st(inh, 48, outptr);
+      }
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_upsample_altivec (int max_v_samp_factor,
+                             JDIMENSION output_width,
+                             JSAMPARRAY input_data,
+                             JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr = input_data[inrow];
+    outptr0 = output_data[outrow++];
+    outptr1 = output_data[outrow++];
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr0 += 64, outptr1 += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr0);
+      vec_st(inl, 0, outptr1);
+
+      vec_st(inh, 16, outptr0);
+      vec_st(inh, 16, outptr1);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr0);
+        vec_st(inl, 32, outptr1);
+
+        vec_st(inh, 48, outptr0);
+        vec_st(inh, 48, outptr1);
+      }
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,737 @@
+;
+; jdsample.asm - upsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_mmx) PRIVATE
+
+EXTN(jconst_fancy_upsample_mmx):
+
+PW_ONE          times 4 dw  1
+PW_TWO          times 4 dw  2
+PW_THREE        times 4 dw  3
+PW_SEVEN        times 4 dw  7
+PW_EIGHT        times 4 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_mmx (int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        mov     eax, JDIMENSION [downsamp_width(ebp)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                     ; colctr
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr
+
+        test    eax, SIZEOF_MMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    mm0,mm0                 ; mm0=(all 0's)
+        pcmpeqb mm7,mm7
+        psrlq   mm7,(SIZEOF_MMWORD-1)*BYTE_BIT
+        pand    mm7, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        add     eax, byte SIZEOF_MMWORD-1
+        and     eax, byte -SIZEOF_MMWORD
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        pcmpeqb mm6,mm6
+        psllq   mm6,(SIZEOF_MMWORD-1)*BYTE_BIT
+        pand    mm6, MMWORD [esi+0*SIZEOF_MMWORD]
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        movq    mm6, MMWORD [esi+1*SIZEOF_MMWORD]
+        psllq   mm6,(SIZEOF_MMWORD-1)*BYTE_BIT
+
+.upsample:
+        movq    mm1, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm2,mm1
+        movq    mm3,mm1                 ; mm1=( 0 1 2 3 4 5 6 7)
+        psllq   mm2,BYTE_BIT            ; mm2=( - 0 1 2 3 4 5 6)
+        psrlq   mm3,BYTE_BIT            ; mm3=( 1 2 3 4 5 6 7 -)
+
+        por     mm2,mm7                 ; mm2=(-1 0 1 2 3 4 5 6)
+        por     mm3,mm6                 ; mm3=( 1 2 3 4 5 6 7 8)
+
+        movq    mm7,mm1
+        psrlq   mm7,(SIZEOF_MMWORD-1)*BYTE_BIT  ; mm7=( 7 - - - - - - -)
+
+        movq      mm4,mm1
+        punpcklbw mm1,mm0               ; mm1=( 0 1 2 3)
+        punpckhbw mm4,mm0               ; mm4=( 4 5 6 7)
+        movq      mm5,mm2
+        punpcklbw mm2,mm0               ; mm2=(-1 0 1 2)
+        punpckhbw mm5,mm0               ; mm5=( 3 4 5 6)
+        movq      mm6,mm3
+        punpcklbw mm3,mm0               ; mm3=( 1 2 3 4)
+        punpckhbw mm6,mm0               ; mm6=( 5 6 7 8)
+
+        pmullw  mm1,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm2,[GOTOFF(ebx,PW_ONE)]
+        paddw   mm5,[GOTOFF(ebx,PW_ONE)]
+        paddw   mm3,[GOTOFF(ebx,PW_TWO)]
+        paddw   mm6,[GOTOFF(ebx,PW_TWO)]
+
+        paddw   mm2,mm1
+        paddw   mm5,mm4
+        psrlw   mm2,2                   ; mm2=OutLE=( 0  2  4  6)
+        psrlw   mm5,2                   ; mm5=OutHE=( 8 10 12 14)
+        paddw   mm3,mm1
+        paddw   mm6,mm4
+        psrlw   mm3,2                   ; mm3=OutLO=( 1  3  5  7)
+        psrlw   mm6,2                   ; mm6=OutHO=( 9 11 13 15)
+
+        psllw   mm3,BYTE_BIT
+        psllw   mm6,BYTE_BIT
+        por     mm2,mm3                 ; mm2=OutL=( 0  1  2  3  4  5  6  7)
+        por     mm5,mm6                 ; mm5=OutH=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm5
+
+        sub     eax, byte SIZEOF_MMWORD
+        add     esi, byte 1*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 2*SIZEOF_MMWORD       ; outptr
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_mmx (int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          4
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void *gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     edx,eax                         ; edx = original ebp
+        mov     eax, JDIMENSION [downsamp_width(edx)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(edx)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(edx)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(edx)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                                     ; colctr
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    eax, SIZEOF_MMWORD-1
+        jz      short .skip
+        push    edx
+        mov     dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     edx
+.skip:
+        ; -- process the first column block
+
+        movq    mm0, MMWORD [ebx+0*SIZEOF_MMWORD]       ; mm0=row[ 0][0]
+        movq    mm1, MMWORD [ecx+0*SIZEOF_MMWORD]       ; mm1=row[-1][0]
+        movq    mm2, MMWORD [esi+0*SIZEOF_MMWORD]       ; mm2=row[+1][0]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      mm3,mm3               ; mm3=(all 0's)
+        movq      mm4,mm0
+        punpcklbw mm0,mm3               ; mm0=row[ 0][0]( 0 1 2 3)
+        punpckhbw mm4,mm3               ; mm4=row[ 0][0]( 4 5 6 7)
+        movq      mm5,mm1
+        punpcklbw mm1,mm3               ; mm1=row[-1][0]( 0 1 2 3)
+        punpckhbw mm5,mm3               ; mm5=row[-1][0]( 4 5 6 7)
+        movq      mm6,mm2
+        punpcklbw mm2,mm3               ; mm2=row[+1][0]( 0 1 2 3)
+        punpckhbw mm6,mm3               ; mm6=row[+1][0]( 4 5 6 7)
+
+        pmullw  mm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+
+        pcmpeqb mm7,mm7
+        psrlq   mm7,(SIZEOF_MMWORD-2)*BYTE_BIT
+
+        paddw   mm1,mm0                 ; mm1=Int0L=( 0 1 2 3)
+        paddw   mm5,mm4                 ; mm5=Int0H=( 4 5 6 7)
+        paddw   mm2,mm0                 ; mm2=Int1L=( 0 1 2 3)
+        paddw   mm6,mm4                 ; mm6=Int1H=( 4 5 6 7)
+
+        movq    MMWORD [edx+0*SIZEOF_MMWORD], mm1       ; temporarily save
+        movq    MMWORD [edx+1*SIZEOF_MMWORD], mm5       ; the intermediate data
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm6
+
+        pand    mm1,mm7                 ; mm1=( 0 - - -)
+        pand    mm2,mm7                 ; mm2=( 0 - - -)
+
+        movq    MMWORD [wk(0)], mm1
+        movq    MMWORD [wk(1)], mm2
+
+        poppic  ebx
+
+        add     eax, byte SIZEOF_MMWORD-1
+        and     eax, byte -SIZEOF_MMWORD
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pcmpeqb mm1,mm1
+        psllq   mm1,(SIZEOF_MMWORD-2)*BYTE_BIT
+        movq    mm2,mm1
+
+        pand    mm1, MMWORD [edx+1*SIZEOF_MMWORD]       ; mm1=( - - - 7)
+        pand    mm2, MMWORD [edi+1*SIZEOF_MMWORD]       ; mm2=( - - - 7)
+
+        movq    MMWORD [wk(2)], mm1
+        movq    MMWORD [wk(3)], mm2
+
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        ; -- process the next column block
+
+        movq    mm0, MMWORD [ebx+1*SIZEOF_MMWORD]       ; mm0=row[ 0][1]
+        movq    mm1, MMWORD [ecx+1*SIZEOF_MMWORD]       ; mm1=row[-1][1]
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]       ; mm2=row[+1][1]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      mm3,mm3               ; mm3=(all 0's)
+        movq      mm4,mm0
+        punpcklbw mm0,mm3               ; mm0=row[ 0][1]( 0 1 2 3)
+        punpckhbw mm4,mm3               ; mm4=row[ 0][1]( 4 5 6 7)
+        movq      mm5,mm1
+        punpcklbw mm1,mm3               ; mm1=row[-1][1]( 0 1 2 3)
+        punpckhbw mm5,mm3               ; mm5=row[-1][1]( 4 5 6 7)
+        movq      mm6,mm2
+        punpcklbw mm2,mm3               ; mm2=row[+1][1]( 0 1 2 3)
+        punpckhbw mm6,mm3               ; mm6=row[+1][1]( 4 5 6 7)
+
+        pmullw  mm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+
+        paddw   mm1,mm0                 ; mm1=Int0L=( 0 1 2 3)
+        paddw   mm5,mm4                 ; mm5=Int0H=( 4 5 6 7)
+        paddw   mm2,mm0                 ; mm2=Int1L=( 0 1 2 3)
+        paddw   mm6,mm4                 ; mm6=Int1H=( 4 5 6 7)
+
+        movq    MMWORD [edx+2*SIZEOF_MMWORD], mm1       ; temporarily save
+        movq    MMWORD [edx+3*SIZEOF_MMWORD], mm5       ; the intermediate data
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm6
+
+        psllq   mm1,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm1=( - - - 0)
+        psllq   mm2,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm2=( - - - 0)
+
+        movq    MMWORD [wk(2)], mm1
+        movq    MMWORD [wk(3)], mm2
+
+.upsample:
+        ; -- process the upper row
+
+        movq    mm7, MMWORD [edx+0*SIZEOF_MMWORD]       ; mm7=Int0L=( 0 1 2 3)
+        movq    mm3, MMWORD [edx+1*SIZEOF_MMWORD]       ; mm3=Int0H=( 4 5 6 7)
+
+        movq    mm0,mm7
+        movq    mm4,mm3
+        psrlq   mm0,2*BYTE_BIT                  ; mm0=( 1 2 3 -)
+        psllq   mm4,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( - - - 4)
+        movq    mm5,mm7
+        movq    mm6,mm3
+        psrlq   mm5,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm5=( 3 - - -)
+        psllq   mm6,2*BYTE_BIT                  ; mm6=( - 4 5 6)
+
+        por     mm0,mm4                         ; mm0=( 1 2 3 4)
+        por     mm5,mm6                         ; mm5=( 3 4 5 6)
+
+        movq    mm1,mm7
+        movq    mm2,mm3
+        psllq   mm1,2*BYTE_BIT                  ; mm1=( - 0 1 2)
+        psrlq   mm2,2*BYTE_BIT                  ; mm2=( 5 6 7 -)
+        movq    mm4,mm3
+        psrlq   mm4,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( 7 - - -)
+
+        por     mm1, MMWORD [wk(0)]             ; mm1=(-1 0 1 2)
+        por     mm2, MMWORD [wk(2)]             ; mm2=( 5 6 7 8)
+
+        movq    MMWORD [wk(0)], mm4
+
+        pmullw  mm7,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm3,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm5,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm0,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   mm2,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   mm1,mm7
+        paddw   mm5,mm3
+        psrlw   mm1,4                   ; mm1=Out0LE=( 0  2  4  6)
+        psrlw   mm5,4                   ; mm5=Out0HE=( 8 10 12 14)
+        paddw   mm0,mm7
+        paddw   mm2,mm3
+        psrlw   mm0,4                   ; mm0=Out0LO=( 1  3  5  7)
+        psrlw   mm2,4                   ; mm2=Out0HO=( 9 11 13 15)
+
+        psllw   mm0,BYTE_BIT
+        psllw   mm2,BYTE_BIT
+        por     mm1,mm0                 ; mm1=Out0L=( 0  1  2  3  4  5  6  7)
+        por     mm5,mm2                 ; mm5=Out0H=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edx+0*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edx+1*SIZEOF_MMWORD], mm5
+
+        ; -- process the lower row
+
+        movq    mm6, MMWORD [edi+0*SIZEOF_MMWORD]       ; mm6=Int1L=( 0 1 2 3)
+        movq    mm4, MMWORD [edi+1*SIZEOF_MMWORD]       ; mm4=Int1H=( 4 5 6 7)
+
+        movq    mm7,mm6
+        movq    mm3,mm4
+        psrlq   mm7,2*BYTE_BIT                  ; mm7=( 1 2 3 -)
+        psllq   mm3,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( - - - 4)
+        movq    mm0,mm6
+        movq    mm2,mm4
+        psrlq   mm0,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm0=( 3 - - -)
+        psllq   mm2,2*BYTE_BIT                  ; mm2=( - 4 5 6)
+
+        por     mm7,mm3                         ; mm7=( 1 2 3 4)
+        por     mm0,mm2                         ; mm0=( 3 4 5 6)
+
+        movq    mm1,mm6
+        movq    mm5,mm4
+        psllq   mm1,2*BYTE_BIT                  ; mm1=( - 0 1 2)
+        psrlq   mm5,2*BYTE_BIT                  ; mm5=( 5 6 7 -)
+        movq    mm3,mm4
+        psrlq   mm3,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( 7 - - -)
+
+        por     mm1, MMWORD [wk(1)]             ; mm1=(-1 0 1 2)
+        por     mm5, MMWORD [wk(3)]             ; mm5=( 5 6 7 8)
+
+        movq    MMWORD [wk(1)], mm3
+
+        pmullw  mm6,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm0,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm7,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   mm5,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   mm1,mm6
+        paddw   mm0,mm4
+        psrlw   mm1,4                   ; mm1=Out1LE=( 0  2  4  6)
+        psrlw   mm0,4                   ; mm0=Out1HE=( 8 10 12 14)
+        paddw   mm7,mm6
+        paddw   mm5,mm4
+        psrlw   mm7,4                   ; mm7=Out1LO=( 1  3  5  7)
+        psrlw   mm5,4                   ; mm5=Out1HO=( 9 11 13 15)
+
+        psllw   mm7,BYTE_BIT
+        psllw   mm5,BYTE_BIT
+        por     mm1,mm7                 ; mm1=Out1L=( 0  1  2  3  4  5  6  7)
+        por     mm0,mm5                 ; mm0=Out1H=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm0
+
+        poppic  ebx
+
+        sub     eax, byte SIZEOF_MMWORD
+        add     ecx, byte 1*SIZEOF_MMWORD       ; inptr1(above)
+        add     ebx, byte 1*SIZEOF_MMWORD       ; inptr0
+        add     esi, byte 1*SIZEOF_MMWORD       ; inptr1(below)
+        add     edx, byte 2*SIZEOF_MMWORD       ; outptr0
+        add     edi, byte 2*SIZEOF_MMWORD       ; outptr1
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     ecx
+        pop     eax
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_mmx (int max_v_samp_factor,
+;                          JDIMENSION output_width,
+;                          JSAMPARRAY input_data,
+;                          JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_MMWORD)-1
+        and     edx, byte -(2*SIZEOF_MMWORD)
+        jz      short .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        mov     eax,edx                         ; colctr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        movq      mm1,mm0
+        punpcklbw mm0,mm0
+        punpckhbw mm1,mm1
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm1
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq      mm3,mm2
+        punpcklbw mm2,mm2
+        punpckhbw mm3,mm3
+
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm3
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 4*SIZEOF_MMWORD       ; outptr
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_mmx (int max_v_samp_factor,
+;                          JDIMENSION output_width,
+;                          JSAMPARRAY input_data,
+;                          JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_MMWORD)-1
+        and     edx, byte -(2*SIZEOF_MMWORD)
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]                     ; inptr
+        mov     ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     eax,edx                                 ; colctr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        movq      mm1,mm0
+        punpcklbw mm0,mm0
+        punpckhbw mm1,mm1
+
+        movq    MMWORD [ebx+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [ebx+1*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm1
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq      mm3,mm2
+        punpcklbw mm2,mm2
+        punpckhbw mm3,mm3
+
+        movq    MMWORD [ebx+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [ebx+3*SIZEOF_MMWORD], mm3
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm3
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     ebx, byte 4*SIZEOF_MMWORD       ; outptr0
+        add     edi, byte 4*SIZEOF_MMWORD       ; outptr1
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,671 @@
+;
+; jdsample.asm - upsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_sse2) PRIVATE
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE          times 8 dw  1
+PW_TWO          times 8 dw  2
+PW_THREE        times 8 dw  3
+PW_SEVEN        times 8 dw  7
+PW_EIGHT        times 8 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     eax, r11d  ; colctr
+        test    rax,rax
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rax                     ; colctr
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr
+
+        test    rax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    xmm0,xmm0               ; xmm0=(all 0's)
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-1)
+        pand    xmm7, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        add     rax, byte SIZEOF_XMMWORD-1
+        and     rax, byte -SIZEOF_XMMWORD
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+
+.columnloop_last:
+        pcmpeqb xmm6,xmm6
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+        pand    xmm6, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        jmp     short .upsample
+
+.columnloop:
+        movdqa  xmm6, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+
+.upsample:
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2,xmm1
+        movdqa  xmm3,xmm1               ; xmm1=( 0  1  2 ... 13 14 15)
+        pslldq  xmm2,1                  ; xmm2=(--  0  1 ... 12 13 14)
+        psrldq  xmm3,1                  ; xmm3=( 1  2  3 ... 14 15 --)
+
+        por     xmm2,xmm7               ; xmm2=(-1  0  1 ... 12 13 14)
+        por     xmm3,xmm6               ; xmm3=( 1  2  3 ... 14 15 16)
+
+        movdqa  xmm7,xmm1
+        psrldq  xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --)
+
+        movdqa    xmm4,xmm1
+        punpcklbw xmm1,xmm0             ; xmm1=( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm0             ; xmm4=( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm2
+        punpcklbw xmm2,xmm0             ; xmm2=(-1  0  1  2  3  4  5  6)
+        punpckhbw xmm5,xmm0             ; xmm5=( 7  8  9 10 11 12 13 14)
+        movdqa    xmm6,xmm3
+        punpcklbw xmm3,xmm0             ; xmm3=( 1  2  3  4  5  6  7  8)
+        punpckhbw xmm6,xmm0             ; xmm6=( 9 10 11 12 13 14 15 16)
+
+        pmullw  xmm1,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+        paddw   xmm2,[rel PW_ONE]
+        paddw   xmm5,[rel PW_ONE]
+        paddw   xmm3,[rel PW_TWO]
+        paddw   xmm6,[rel PW_TWO]
+
+        paddw   xmm2,xmm1
+        paddw   xmm5,xmm4
+        psrlw   xmm2,2                  ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,2                  ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+        paddw   xmm3,xmm1
+        paddw   xmm6,xmm4
+        psrlw   xmm3,2                  ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm6,2                  ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm3,BYTE_BIT
+        psllw   xmm6,BYTE_BIT
+        por     xmm2,xmm3               ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm6               ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm5
+
+        sub     rax, byte SIZEOF_XMMWORD
+        add     rsi, byte 1*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     rsi
+        pop     rdi
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rcx                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          4
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     eax, r11d  ; colctr
+        test    rax,rax
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rax                                     ; colctr
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rcx, JSAMPROW [rsi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    rax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        push    rdx
+        mov     dl, JSAMPLE [rcx+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rcx+rax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [rbx+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rbx+rax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     rdx
+.skip:
+        ; -- process the first column block
+
+        movdqa  xmm0, XMMWORD [rbx+0*SIZEOF_XMMWORD]    ; xmm0=row[ 0][0]
+        movdqa  xmm1, XMMWORD [rcx+0*SIZEOF_XMMWORD]    ; xmm1=row[-1][0]
+        movdqa  xmm2, XMMWORD [rsi+0*SIZEOF_XMMWORD]    ; xmm2=row[+1][0]
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-2)
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm6
+
+        pand    xmm1,xmm7               ; xmm1=( 0 -- -- -- -- -- -- --)
+        pand    xmm2,xmm7               ; xmm2=( 0 -- -- -- -- -- -- --)
+
+        movdqa  XMMWORD [wk(0)], xmm1
+        movdqa  XMMWORD [wk(1)], xmm2
+
+        add     rax, byte SIZEOF_XMMWORD-1
+        and     rax, byte -SIZEOF_XMMWORD
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pcmpeqb xmm1,xmm1
+        pslldq  xmm1,(SIZEOF_XMMWORD-2)
+        movdqa  xmm2,xmm1
+
+        pand    xmm1, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+        pand    xmm2, XMMWORD [rdi+1*SIZEOF_XMMWORD]
+
+        movdqa  XMMWORD [wk(2)], xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+        movdqa  XMMWORD [wk(3)], xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+        jmp     near .upsample
+
+.columnloop:
+        ; -- process the next column block
+
+        movdqa  xmm0, XMMWORD [rbx+1*SIZEOF_XMMWORD]    ; xmm0=row[ 0][1]
+        movdqa  xmm1, XMMWORD [rcx+1*SIZEOF_XMMWORD]    ; xmm1=row[-1][1]
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]    ; xmm2=row[+1][1]
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [rdx+2*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [rdx+3*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm6
+
+        pslldq  xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- --  0)
+        pslldq  xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- --  0)
+
+        movdqa  XMMWORD [wk(2)], xmm1
+        movdqa  XMMWORD [wk(3)], xmm2
+
+.upsample:
+        ; -- process the upper row
+
+        movdqa  xmm7, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm0,xmm7               ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm4,xmm3               ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm0,2                  ; xmm0=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- --  8)
+        movdqa  xmm5,xmm7
+        movdqa  xmm6,xmm3
+        psrldq  xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm6,2                  ; xmm6=(--  8  9 10 11 12 13 14)
+
+        por     xmm0,xmm4               ; xmm0=( 1  2  3  4  5  6  7  8)
+        por     xmm5,xmm6               ; xmm5=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm2,xmm3
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm2,2                  ; xmm2=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm4,xmm3
+        psrldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(0)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm2, XMMWORD [wk(2)]   ; xmm2=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(0)], xmm4
+
+        pmullw  xmm7,[rel PW_THREE]
+        pmullw  xmm3,[rel PW_THREE]
+        paddw   xmm1,[rel PW_EIGHT]
+        paddw   xmm5,[rel PW_EIGHT]
+        paddw   xmm0,[rel PW_SEVEN]
+        paddw   xmm2,[rel PW_SEVEN]
+
+        paddw   xmm1,xmm7
+        paddw   xmm5,xmm3
+        psrlw   xmm1,4                  ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,4                  ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm3
+        psrlw   xmm0,4                  ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm2,4                  ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm0,BYTE_BIT
+        psllw   xmm2,BYTE_BIT
+        por     xmm1,xmm0               ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm2               ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5
+
+        ; -- process the lower row
+
+        movdqa  xmm6, XMMWORD [rdi+0*SIZEOF_XMMWORD]
+        movdqa  xmm4, XMMWORD [rdi+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm7,xmm6               ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm3,xmm4               ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm7,2                  ; xmm7=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- --  8)
+        movdqa  xmm0,xmm6
+        movdqa  xmm2,xmm4
+        psrldq  xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm2,2                  ; xmm2=(--  8  9 10 11 12 13 14)
+
+        por     xmm7,xmm3               ; xmm7=( 1  2  3  4  5  6  7  8)
+        por     xmm0,xmm2               ; xmm0=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm4
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm5,2                  ; xmm5=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm3,xmm4
+        psrldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(1)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm5, XMMWORD [wk(3)]   ; xmm5=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(1)], xmm3
+
+        pmullw  xmm6,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+        paddw   xmm1,[rel PW_EIGHT]
+        paddw   xmm0,[rel PW_EIGHT]
+        paddw   xmm7,[rel PW_SEVEN]
+        paddw   xmm5,[rel PW_SEVEN]
+
+        paddw   xmm1,xmm6
+        paddw   xmm0,xmm4
+        psrlw   xmm1,4                  ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm0,4                  ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm7,xmm6
+        paddw   xmm5,xmm4
+        psrlw   xmm7,4                  ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm5,4                  ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm7,BYTE_BIT
+        psllw   xmm5,BYTE_BIT
+        por     xmm1,xmm7               ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+        por     xmm0,xmm5               ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm0
+
+        sub     rax, byte SIZEOF_XMMWORD
+        add     rcx, byte 1*SIZEOF_XMMWORD      ; inptr1(above)
+        add     rbx, byte 1*SIZEOF_XMMWORD      ; inptr0
+        add     rsi, byte 1*SIZEOF_XMMWORD      ; inptr1(below)
+        add     rdx, byte 2*SIZEOF_XMMWORD      ; outptr0
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr1
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    rax,rax
+        jnz     near .columnloop_last
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     rcx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     edx, r11d
+        add     rdx, byte (2*SIZEOF_XMMWORD)-1
+        and     rdx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      short .return
+
+        mov     rsi, r12 ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]             ; inptr
+        mov     rdi, JSAMPROW [rdi]             ; outptr
+        mov     rax,rdx                         ; colctr
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 4*SIZEOF_XMMWORD      ; outptr
+        jmp     short .columnloop
+
+.nextrow:
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rcx                             ; rowctr
+        jg      short .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        mov     edx, r11d
+        add     rdx, byte (2*SIZEOF_XMMWORD)-1
+        and     rdx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]                     ; inptr
+        mov     rbx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     rax,rdx                                 ; colctr
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [rbx+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rbx+1*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [rbx+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rbx+3*SIZEOF_XMMWORD], xmm3
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rbx, byte 4*SIZEOF_XMMWORD      ; outptr0
+        add     rdi, byte 4*SIZEOF_XMMWORD      ; outptr1
+        jmp     short .columnloop
+
+.nextrow:
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     rcx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jdsample-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jdsample-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,729 @@
+;
+; jdsample.asm - upsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_sse2) PRIVATE
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE          times 8 dw  1
+PW_TWO          times 8 dw  2
+PW_THREE        times 8 dw  3
+PW_SEVEN        times 8 dw  7
+PW_EIGHT        times 8 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        mov     eax, JDIMENSION [downsamp_width(ebp)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                     ; colctr
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr
+
+        test    eax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    xmm0,xmm0               ; xmm0=(all 0's)
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-1)
+        pand    xmm7, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        add     eax, byte SIZEOF_XMMWORD-1
+        and     eax, byte -SIZEOF_XMMWORD
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        pcmpeqb xmm6,xmm6
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+        pand    xmm6, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm6, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+
+.upsample:
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2,xmm1
+        movdqa  xmm3,xmm1               ; xmm1=( 0  1  2 ... 13 14 15)
+        pslldq  xmm2,1                  ; xmm2=(--  0  1 ... 12 13 14)
+        psrldq  xmm3,1                  ; xmm3=( 1  2  3 ... 14 15 --)
+
+        por     xmm2,xmm7               ; xmm2=(-1  0  1 ... 12 13 14)
+        por     xmm3,xmm6               ; xmm3=( 1  2  3 ... 14 15 16)
+
+        movdqa  xmm7,xmm1
+        psrldq  xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --)
+
+        movdqa    xmm4,xmm1
+        punpcklbw xmm1,xmm0             ; xmm1=( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm0             ; xmm4=( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm2
+        punpcklbw xmm2,xmm0             ; xmm2=(-1  0  1  2  3  4  5  6)
+        punpckhbw xmm5,xmm0             ; xmm5=( 7  8  9 10 11 12 13 14)
+        movdqa    xmm6,xmm3
+        punpcklbw xmm3,xmm0             ; xmm3=( 1  2  3  4  5  6  7  8)
+        punpckhbw xmm6,xmm0             ; xmm6=( 9 10 11 12 13 14 15 16)
+
+        pmullw  xmm1,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm2,[GOTOFF(ebx,PW_ONE)]
+        paddw   xmm5,[GOTOFF(ebx,PW_ONE)]
+        paddw   xmm3,[GOTOFF(ebx,PW_TWO)]
+        paddw   xmm6,[GOTOFF(ebx,PW_TWO)]
+
+        paddw   xmm2,xmm1
+        paddw   xmm5,xmm4
+        psrlw   xmm2,2                  ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,2                  ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+        paddw   xmm3,xmm1
+        paddw   xmm6,xmm4
+        psrlw   xmm3,2                  ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm6,2                  ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm3,BYTE_BIT
+        psllw   xmm6,BYTE_BIT
+        por     xmm2,xmm3               ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm6               ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm5
+
+        sub     eax, byte SIZEOF_XMMWORD
+        add     esi, byte 1*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          4
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void *gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     edx,eax                         ; edx = original ebp
+        mov     eax, JDIMENSION [downsamp_width(edx)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(edx)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(edx)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(edx)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                                     ; colctr
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    eax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        push    edx
+        mov     dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     edx
+.skip:
+        ; -- process the first column block
+
+        movdqa  xmm0, XMMWORD [ebx+0*SIZEOF_XMMWORD]    ; xmm0=row[ 0][0]
+        movdqa  xmm1, XMMWORD [ecx+0*SIZEOF_XMMWORD]    ; xmm1=row[-1][0]
+        movdqa  xmm2, XMMWORD [esi+0*SIZEOF_XMMWORD]    ; xmm2=row[+1][0]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-2)
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm6
+
+        pand    xmm1,xmm7               ; xmm1=( 0 -- -- -- -- -- -- --)
+        pand    xmm2,xmm7               ; xmm2=( 0 -- -- -- -- -- -- --)
+
+        movdqa  XMMWORD [wk(0)], xmm1
+        movdqa  XMMWORD [wk(1)], xmm2
+
+        poppic  ebx
+
+        add     eax, byte SIZEOF_XMMWORD-1
+        and     eax, byte -SIZEOF_XMMWORD
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pcmpeqb xmm1,xmm1
+        pslldq  xmm1,(SIZEOF_XMMWORD-2)
+        movdqa  xmm2,xmm1
+
+        pand    xmm1, XMMWORD [edx+1*SIZEOF_XMMWORD]
+        pand    xmm2, XMMWORD [edi+1*SIZEOF_XMMWORD]
+
+        movdqa  XMMWORD [wk(2)], xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+        movdqa  XMMWORD [wk(3)], xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+        jmp     near .upsample
+        alignx  16,7
+
+.columnloop:
+        ; -- process the next column block
+
+        movdqa  xmm0, XMMWORD [ebx+1*SIZEOF_XMMWORD]    ; xmm0=row[ 0][1]
+        movdqa  xmm1, XMMWORD [ecx+1*SIZEOF_XMMWORD]    ; xmm1=row[-1][1]
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]    ; xmm2=row[+1][1]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [edx+2*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [edx+3*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm6
+
+        pslldq  xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- --  0)
+        pslldq  xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- --  0)
+
+        movdqa  XMMWORD [wk(2)], xmm1
+        movdqa  XMMWORD [wk(3)], xmm2
+
+.upsample:
+        ; -- process the upper row
+
+        movdqa  xmm7, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [edx+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm0,xmm7               ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm4,xmm3               ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm0,2                  ; xmm0=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- --  8)
+        movdqa  xmm5,xmm7
+        movdqa  xmm6,xmm3
+        psrldq  xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm6,2                  ; xmm6=(--  8  9 10 11 12 13 14)
+
+        por     xmm0,xmm4               ; xmm0=( 1  2  3  4  5  6  7  8)
+        por     xmm5,xmm6               ; xmm5=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm2,xmm3
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm2,2                  ; xmm2=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm4,xmm3
+        psrldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(0)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm2, XMMWORD [wk(2)]   ; xmm2=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(0)], xmm4
+
+        pmullw  xmm7,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm3,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm5,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm0,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   xmm2,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   xmm1,xmm7
+        paddw   xmm5,xmm3
+        psrlw   xmm1,4                  ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,4                  ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm3
+        psrlw   xmm0,4                  ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm2,4                  ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm0,BYTE_BIT
+        psllw   xmm2,BYTE_BIT
+        por     xmm1,xmm0               ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm2               ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5
+
+        ; -- process the lower row
+
+        movdqa  xmm6, XMMWORD [edi+0*SIZEOF_XMMWORD]
+        movdqa  xmm4, XMMWORD [edi+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm7,xmm6               ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm3,xmm4               ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm7,2                  ; xmm7=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- --  8)
+        movdqa  xmm0,xmm6
+        movdqa  xmm2,xmm4
+        psrldq  xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm2,2                  ; xmm2=(--  8  9 10 11 12 13 14)
+
+        por     xmm7,xmm3               ; xmm7=( 1  2  3  4  5  6  7  8)
+        por     xmm0,xmm2               ; xmm0=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm4
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm5,2                  ; xmm5=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm3,xmm4
+        psrldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(1)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm5, XMMWORD [wk(3)]   ; xmm5=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(1)], xmm3
+
+        pmullw  xmm6,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm0,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm7,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   xmm5,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   xmm1,xmm6
+        paddw   xmm0,xmm4
+        psrlw   xmm1,4                  ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm0,4                  ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm7,xmm6
+        paddw   xmm5,xmm4
+        psrlw   xmm7,4                  ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm5,4                  ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm7,BYTE_BIT
+        psllw   xmm5,BYTE_BIT
+        por     xmm1,xmm7               ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+        por     xmm0,xmm5               ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm0
+
+        poppic  ebx
+
+        sub     eax, byte SIZEOF_XMMWORD
+        add     ecx, byte 1*SIZEOF_XMMWORD      ; inptr1(above)
+        add     ebx, byte 1*SIZEOF_XMMWORD      ; inptr0
+        add     esi, byte 1*SIZEOF_XMMWORD      ; inptr1(below)
+        add     edx, byte 2*SIZEOF_XMMWORD      ; outptr0
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr1
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     ecx
+        pop     eax
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_XMMWORD)-1
+        and     edx, byte -(2*SIZEOF_XMMWORD)
+        jz      short .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        mov     eax,edx                         ; colctr
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 4*SIZEOF_XMMWORD      ; outptr
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      short .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_XMMWORD)-1
+        and     edx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]                     ; inptr
+        mov     ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     eax,edx                                 ; colctr
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [ebx+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [ebx+1*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [ebx+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [ebx+3*SIZEOF_XMMWORD], xmm3
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     ebx, byte 4*SIZEOF_XMMWORD      ; outptr0
+        add     edi, byte 4*SIZEOF_XMMWORD      ; outptr1
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      short .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-3dn.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,320 @@
+;
+; jfdctflt.asm - floating-point FDCT (3DNow!)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_3dnow) PRIVATE
+
+EXTN(jconst_fdct_float_3dnow):
+
+PD_0_382        times 2 dd  0.382683432365089771728460
+PD_0_707        times 2 dd  0.707106781186547524400844
+PD_0_541        times 2 dd  0.541196100146196984399723
+PD_1_306        times 2 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_3dnow (FAST_FLOAT *data)
+;
+
+%define data(b)         (b)+8           ; FAST_FLOAT *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_3dnow) PRIVATE
+
+EXTN(jsimd_fdct_float_3dnow):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm0=(00 01), mm1=(10 11), mm2=(06 07), mm3=(16 17)
+
+        movq      mm4,mm0               ; transpose coefficients
+        punpckldq mm0,mm1               ; mm0=(00 10)=data0
+        punpckhdq mm4,mm1               ; mm4=(01 11)=data1
+        movq      mm5,mm2               ; transpose coefficients
+        punpckldq mm2,mm3               ; mm2=(06 16)=data6
+        punpckhdq mm5,mm3               ; mm5=(07 17)=data7
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm2                 ; mm4=data1-data6=tmp6
+        pfsub   mm0,mm5                 ; mm0=data0-data7=tmp7
+        pfadd   mm6,mm2                 ; mm6=data1+data6=tmp1
+        pfadd   mm7,mm5                 ; mm7=data0+data7=tmp0
+
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm1=(02 03), mm3=(12 13), mm2=(04 05), mm5=(14 15)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=tmp7
+
+        movq      mm4,mm1               ; transpose coefficients
+        punpckldq mm1,mm3               ; mm1=(02 12)=data2
+        punpckhdq mm4,mm3               ; mm4=(03 13)=data3
+        movq      mm0,mm2               ; transpose coefficients
+        punpckldq mm2,mm5               ; mm2=(04 14)=data4
+        punpckhdq mm0,mm5               ; mm0=(05 15)=data5
+
+        movq    mm3,mm4
+        movq    mm5,mm1
+        pfadd   mm4,mm2                 ; mm4=data3+data4=tmp3
+        pfadd   mm1,mm0                 ; mm1=data2+data5=tmp2
+        pfsub   mm3,mm2                 ; mm3=data3-data4=tmp4
+        pfsub   mm5,mm0                 ; mm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm2,mm7
+        movq    mm0,mm6
+        pfsub   mm7,mm4                 ; mm7=tmp13
+        pfsub   mm6,mm1                 ; mm6=tmp12
+        pfadd   mm2,mm4                 ; mm2=tmp10
+        pfadd   mm0,mm1                 ; mm0=tmp11
+
+        pfadd   mm6,mm7
+        pfmul   mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1
+
+        movq    mm4,mm2
+        movq    mm1,mm7
+        pfsub   mm2,mm0                 ; mm2=data4
+        pfsub   mm7,mm6                 ; mm7=data6
+        pfadd   mm4,mm0                 ; mm4=data0
+        pfadd   mm1,mm6                 ; mm1=data2
+
+        movq    MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [wk(0)]     ; mm0=tmp6
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp7
+
+        pfadd   mm3,mm5                 ; mm3=tmp10
+        pfadd   mm5,mm0                 ; mm5=tmp11
+        pfadd   mm0,mm6                 ; mm0=tmp12, mm6=tmp7
+
+        pfmul   mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3
+
+        movq    mm2,mm3                 ; mm2=tmp10
+        pfsub   mm3,mm0
+        pfmul   mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5
+        pfmul   mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+        pfmul   mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+        pfadd   mm2,mm3                 ; mm2=z2
+        pfadd   mm0,mm3                 ; mm0=z4
+
+        movq    mm7,mm6
+        pfsub   mm6,mm5                 ; mm6=z13
+        pfadd   mm7,mm5                 ; mm7=z11
+
+        movq    mm4,mm6
+        movq    mm1,mm7
+        pfsub   mm6,mm2                 ; mm6=data3
+        pfsub   mm7,mm0                 ; mm7=data7
+        pfadd   mm4,mm2                 ; mm4=data5
+        pfadd   mm1,mm0                 ; mm1=data1
+
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        add     edx, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm0=(00 10), mm1=(01 11), mm2=(60 70), mm3=(61 71)
+
+        movq      mm4,mm0               ; transpose coefficients
+        punpckldq mm0,mm1               ; mm0=(00 01)=data0
+        punpckhdq mm4,mm1               ; mm4=(10 11)=data1
+        movq      mm5,mm2               ; transpose coefficients
+        punpckldq mm2,mm3               ; mm2=(60 61)=data6
+        punpckhdq mm5,mm3               ; mm5=(70 71)=data7
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm2                 ; mm4=data1-data6=tmp6
+        pfsub   mm0,mm5                 ; mm0=data0-data7=tmp7
+        pfadd   mm6,mm2                 ; mm6=data1+data6=tmp1
+        pfadd   mm7,mm5                 ; mm7=data0+data7=tmp0
+
+        movq    mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm1=(20 30), mm3=(21 31), mm2=(40 50), mm5=(41 51)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=tmp7
+
+        movq      mm4,mm1               ; transpose coefficients
+        punpckldq mm1,mm3               ; mm1=(20 21)=data2
+        punpckhdq mm4,mm3               ; mm4=(30 31)=data3
+        movq      mm0,mm2               ; transpose coefficients
+        punpckldq mm2,mm5               ; mm2=(40 41)=data4
+        punpckhdq mm0,mm5               ; mm0=(50 51)=data5
+
+        movq    mm3,mm4
+        movq    mm5,mm1
+        pfadd   mm4,mm2                 ; mm4=data3+data4=tmp3
+        pfadd   mm1,mm0                 ; mm1=data2+data5=tmp2
+        pfsub   mm3,mm2                 ; mm3=data3-data4=tmp4
+        pfsub   mm5,mm0                 ; mm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm2,mm7
+        movq    mm0,mm6
+        pfsub   mm7,mm4                 ; mm7=tmp13
+        pfsub   mm6,mm1                 ; mm6=tmp12
+        pfadd   mm2,mm4                 ; mm2=tmp10
+        pfadd   mm0,mm1                 ; mm0=tmp11
+
+        pfadd   mm6,mm7
+        pfmul   mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1
+
+        movq    mm4,mm2
+        movq    mm1,mm7
+        pfsub   mm2,mm0                 ; mm2=data4
+        pfsub   mm7,mm6                 ; mm7=data6
+        pfadd   mm4,mm0                 ; mm4=data0
+        pfadd   mm1,mm6                 ; mm1=data2
+
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [wk(0)]     ; mm0=tmp6
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp7
+
+        pfadd   mm3,mm5                 ; mm3=tmp10
+        pfadd   mm5,mm0                 ; mm5=tmp11
+        pfadd   mm0,mm6                 ; mm0=tmp12, mm6=tmp7
+
+        pfmul   mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3
+
+        movq    mm2,mm3                 ; mm2=tmp10
+        pfsub   mm3,mm0
+        pfmul   mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5
+        pfmul   mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+        pfmul   mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+        pfadd   mm2,mm3                 ; mm2=z2
+        pfadd   mm0,mm3                 ; mm0=z4
+
+        movq    mm7,mm6
+        pfsub   mm6,mm5                 ; mm6=z13
+        pfadd   mm7,mm5                 ; mm7=z11
+
+        movq    mm4,mm6
+        movq    mm1,mm7
+        pfsub   mm6,mm2                 ; mm6=data3
+        pfsub   mm7,mm0                 ; mm7=data7
+        pfadd   mm4,mm2                 ; mm4=data5
+        pfadd   mm1,mm0                 ; mm1=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        add     edx, byte 2*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .columnloop
+
+        femms           ; empty MMX/3DNow! state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,358 @@
+;
+; jfdctflt.asm - floating-point FDCT (64-bit SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_sse) PRIVATE
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382        times 4 dd  0.382683432365089771728460
+PD_0_707        times 4 dd  0.707106781186547524400844
+PD_0_541        times 4 dd  0.541196100146196984399723
+PD_1_306        times 4 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse (FAST_FLOAT *data)
+;
+
+; r10 = FAST_FLOAT *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_sse) PRIVATE
+
+EXTN(jsimd_fdct_float_sse):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (FAST_FLOAT *)
+        mov     rcx, DCTSIZE/4
+.rowloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+        ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(20 30 21 31)
+        unpckhps xmm4,xmm1              ; xmm4=(22 32 23 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(24 34 25 35)
+        unpckhps xmm5,xmm3              ; xmm5=(26 36 27 37)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+        ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 32 23 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(24 34 25 35)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 10 01 11)
+        unpckhps xmm4,xmm7              ; xmm4=(02 12 03 13)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(04 14 05 15)
+        unpckhps xmm2,xmm3              ; xmm2=(06 16 07 17)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 10 20 30)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(01 11 21 31)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(06 16 26 36)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(07 17 27 37)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 32 23 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(24 34 25 35)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(02 12 22 32)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(03 13 23 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(04 14 24 34)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(05 15 25 35)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[rel PD_0_707] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[rel PD_0_707] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[rel PD_0_382] ; xmm2=z5
+        mulps   xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     rdx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     rdx, r10        ; (FAST_FLOAT *)
+        mov     rcx, DCTSIZE/4
+.columnloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+        ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(02 03 12 13)
+        unpckhps xmm4,xmm1              ; xmm4=(22 23 32 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(42 43 52 53)
+        unpckhps xmm5,xmm3              ; xmm5=(62 63 72 73)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+        ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 23 32 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(42 43 52 53)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 01 10 11)
+        unpckhps xmm4,xmm7              ; xmm4=(20 21 30 31)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(40 41 50 51)
+        unpckhps xmm2,xmm3              ; xmm2=(60 61 70 71)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 01 02 03)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(10 11 12 13)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(60 61 62 63)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(70 71 72 73)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 23 32 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(20 21 22 23)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(30 31 32 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(40 41 42 43)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(50 51 52 53)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[rel PD_0_707] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[rel PD_0_707] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[rel PD_0_382] ; xmm2=z5
+        mulps   xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     rdx, byte 4*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     near .columnloop
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctflt-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,370 @@
+;
+; jfdctflt.asm - floating-point FDCT (SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_sse) PRIVATE
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382        times 4 dd  0.382683432365089771728460
+PD_0_707        times 4 dd  0.707106781186547524400844
+PD_0_541        times 4 dd  0.541196100146196984399723
+PD_1_306        times 4 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse (FAST_FLOAT *data)
+;
+
+%define data(b)         (b)+8           ; FAST_FLOAT *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_sse) PRIVATE
+
+EXTN(jsimd_fdct_float_sse):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+        ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(20 30 21 31)
+        unpckhps xmm4,xmm1              ; xmm4=(22 32 23 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(24 34 25 35)
+        unpckhps xmm5,xmm3              ; xmm5=(26 36 27 37)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+        ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 32 23 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(24 34 25 35)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 10 01 11)
+        unpckhps xmm4,xmm7              ; xmm4=(02 12 03 13)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(04 14 05 15)
+        unpckhps xmm2,xmm3              ; xmm2=(06 16 07 17)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 10 20 30)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(01 11 21 31)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(06 16 26 36)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(07 17 27 37)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 32 23 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(24 34 25 35)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(02 12 22 32)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(03 13 23 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(04 14 24 34)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(05 15 25 35)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5
+        mulps   xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     edx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+        ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(02 03 12 13)
+        unpckhps xmm4,xmm1              ; xmm4=(22 23 32 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(42 43 52 53)
+        unpckhps xmm5,xmm3              ; xmm5=(62 63 72 73)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+        ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 23 32 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(42 43 52 53)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 01 10 11)
+        unpckhps xmm4,xmm7              ; xmm4=(20 21 30 31)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(40 41 50 51)
+        unpckhps xmm2,xmm3              ; xmm2=(60 61 70 71)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 01 02 03)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(10 11 12 13)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(60 61 62 63)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(70 71 72 73)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 23 32 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(20 21 22 23)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(30 31 32 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(40 41 42 43)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(50 51 52 53)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5
+        mulps   xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     edx, byte 4*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .columnloop
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,156 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER FORWARD DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_382 98   /* FIX(0.382683433) */
+#define F_0_541 139  /* FIX(0.541196100) */
+#define F_0_707 181  /* FIX(0.707106781) */
+#define F_1_306 334  /* FIX(1.306562965) */
+
+#define CONST_BITS 8
+#define PRE_MULTIPLY_SCALE_BITS 2
+#define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_FDCT()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  \
+  z1 = vec_add(tmp12, tmp13);  \
+  z1 = vec_sl(z1, pre_multiply_scale_bits);  \
+  z1 = vec_madds(z1, pw_0707, pw_zero);  \
+  \
+  out2 = vec_add(tmp13, z1);  \
+  out6 = vec_sub(tmp13, z1);  \
+  \
+  /* Odd part */  \
+  \
+  tmp10 = vec_add(tmp4, tmp5);  \
+  tmp11 = vec_add(tmp5, tmp6);  \
+  tmp12 = vec_add(tmp6, tmp7);  \
+  \
+  tmp10 = vec_sl(tmp10, pre_multiply_scale_bits);  \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits);  \
+  z5 = vec_sub(tmp10, tmp12);  \
+  z5 = vec_madds(z5, pw_0382, pw_zero);  \
+  \
+  z2 = vec_madds(tmp10, pw_0541, z5);  \
+  z4 = vec_madds(tmp12, pw_1306, z5);  \
+  \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits);  \
+  z3 = vec_madds(tmp11, pw_0707, pw_zero);  \
+  \
+  z11 = vec_add(tmp7, z3);  \
+  z13 = vec_sub(tmp7, z3);  \
+  \
+  out5 = vec_add(z13, z2);  \
+  out3 = vec_sub(z13, z2);  \
+  out1 = vec_add(z11, z4);  \
+  out7 = vec_sub(z11, z4);  \
+}
+
+
+void
+jsimd_fdct_ifast_altivec (DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z1, z2, z3, z4, z5, z11, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_0382 = { __8X(F_0_382 << CONST_SHIFT) },
+    pw_0541 = { __8X(F_0_541 << CONST_SHIFT) },
+    pw_0707 = { __8X(F_0_707 << CONST_SHIFT) },
+    pw_1306 = { __8X(F_1_306 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,397 @@
+;
+; jfdctfst.asm - fast integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_mmx) PRIVATE
+
+EXTN(jconst_fdct_ifast_mmx):
+
+PW_F0707        times 4 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 4 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 4 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 4 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_mmx (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_mmx) PRIVATE
+
+EXTN(jsimd_fdct_ifast_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+        ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(24 34 25 35)
+        punpckhwd mm5,mm3               ; mm5=(26 36 27 37)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+        ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 32 23 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(24 34 25 35)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm4,mm7               ; mm4=(02 12 03 13)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(04 14 05 15)
+        punpckhwd mm2,mm3               ; mm2=(06 16 07 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 10 20 30)=data0
+        punpckhdq mm7,mm0               ; mm7=(01 11 21 31)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(06 16 26 36)=data6
+        punpckhdq mm3,mm5               ; mm3=(07 17 27 37)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 32 23 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(24 34 25 35)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(02 12 22 32)=data2
+        punpckhdq mm7,mm2               ; mm7=(03 13 23 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(04 14 24 34)=data4
+        punpckhdq mm6,mm3               ; mm6=(05 15 25 35)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        psubw   mm5,mm7                 ; mm5=tmp13
+        psubw   mm0,mm4                 ; mm0=tmp12
+        paddw   mm1,mm7                 ; mm1=tmp10
+        paddw   mm6,mm4                 ; mm6=tmp11
+
+        paddw   mm0,mm5
+        psllw   mm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1
+
+        movq    mm7,mm1
+        movq    mm4,mm5
+        psubw   mm1,mm6                 ; mm1=data4
+        psubw   mm5,mm0                 ; mm5=data6
+        paddw   mm7,mm6                 ; mm7=data0
+        paddw   mm4,mm0                 ; mm4=data2
+
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+
+        ; -- Odd part
+
+        movq    mm6, MMWORD [wk(0)]     ; mm6=tmp6
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp7
+
+        paddw   mm2,mm3                 ; mm2=tmp10
+        paddw   mm3,mm6                 ; mm3=tmp11
+        paddw   mm6,mm0                 ; mm6=tmp12, mm0=tmp7
+
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm6,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   mm3,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3
+
+        movq    mm1,mm2                 ; mm1=tmp10
+        psubw   mm2,mm6
+        pmulhw  mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5
+        pmulhw  mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+        pmulhw  mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+        paddw   mm1,mm2                 ; mm1=z2
+        paddw   mm6,mm2                 ; mm6=z4
+
+        movq    mm5,mm0
+        psubw   mm0,mm3                 ; mm0=z13
+        paddw   mm5,mm3                 ; mm5=z11
+
+        movq    mm7,mm0
+        movq    mm4,mm5
+        psubw   mm0,mm1                 ; mm0=data3
+        psubw   mm5,mm6                 ; mm5=data7
+        paddw   mm7,mm1                 ; mm7=data5
+        paddw   mm4,mm6                 ; mm4=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4
+
+        add     edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+        ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(02 03 12 13)
+        punpckhwd mm4,mm1               ; mm4=(22 23 32 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(42 43 52 53)
+        punpckhwd mm5,mm3               ; mm5=(62 63 72 73)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+        ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 23 32 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(42 43 52 53)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 01 10 11)
+        punpckhwd mm4,mm7               ; mm4=(20 21 30 31)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(40 41 50 51)
+        punpckhwd mm2,mm3               ; mm2=(60 61 70 71)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03)=data0
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(60 61 62 63)=data6
+        punpckhdq mm3,mm5               ; mm3=(70 71 72 73)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 23 32 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(42 43 52 53)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(20 21 22 23)=data2
+        punpckhdq mm7,mm2               ; mm7=(30 31 32 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(40 41 42 43)=data4
+        punpckhdq mm6,mm3               ; mm6=(50 51 52 53)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        psubw   mm5,mm7                 ; mm5=tmp13
+        psubw   mm0,mm4                 ; mm0=tmp12
+        paddw   mm1,mm7                 ; mm1=tmp10
+        paddw   mm6,mm4                 ; mm6=tmp11
+
+        paddw   mm0,mm5
+        psllw   mm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1
+
+        movq    mm7,mm1
+        movq    mm4,mm5
+        psubw   mm1,mm6                 ; mm1=data4
+        psubw   mm5,mm0                 ; mm5=data6
+        paddw   mm7,mm6                 ; mm7=data0
+        paddw   mm4,mm0                 ; mm4=data2
+
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+
+        ; -- Odd part
+
+        movq    mm6, MMWORD [wk(0)]     ; mm6=tmp6
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp7
+
+        paddw   mm2,mm3                 ; mm2=tmp10
+        paddw   mm3,mm6                 ; mm3=tmp11
+        paddw   mm6,mm0                 ; mm6=tmp12, mm0=tmp7
+
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm6,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   mm3,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3
+
+        movq    mm1,mm2                 ; mm1=tmp10
+        psubw   mm2,mm6
+        pmulhw  mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5
+        pmulhw  mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+        pmulhw  mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+        paddw   mm1,mm2                 ; mm1=z2
+        paddw   mm6,mm2                 ; mm6=z4
+
+        movq    mm5,mm0
+        psubw   mm0,mm3                 ; mm0=z13
+        paddw   mm5,mm3                 ; mm5=z11
+
+        movq    mm7,mm0
+        movq    mm4,mm5
+        psubw   mm0,mm1                 ; mm0=data3
+        psubw   mm5,mm6                 ; mm5=data7
+        paddw   mm7,mm1                 ; mm7=data5
+        paddw   mm4,mm6                 ; mm4=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4
+
+        add     edx, byte 4*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .columnloop
+
+        emms            ; empty MMX state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,392 @@
+;
+; jfdctfst.asm - fast integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_sse2) PRIVATE
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707        times 8 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 8 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 8 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 8 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2 (DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_fdct_ifast_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(0)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        psubw   xmm3,xmm1               ; xmm3=tmp13
+        psubw   xmm6,xmm7               ; xmm6=tmp12
+        paddw   xmm4,xmm1               ; xmm4=tmp10
+        paddw   xmm0,xmm7               ; xmm0=tmp11
+
+        paddw   xmm6,xmm3
+        psllw   xmm6,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm6,[rel PW_F0707] ; xmm6=z1
+
+        movdqa  xmm1,xmm4
+        movdqa  xmm7,xmm3
+        psubw   xmm4,xmm0               ; xmm4=data4
+        psubw   xmm3,xmm6               ; xmm3=data6
+        paddw   xmm1,xmm0               ; xmm1=data0
+        paddw   xmm7,xmm6               ; xmm7=data2
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=tmp6
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp7
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=data4
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=data6
+
+        ; -- Odd part
+
+        paddw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm5,xmm0               ; xmm5=tmp11
+        paddw   xmm0,xmm6               ; xmm0=tmp12, xmm6=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F0707] ; xmm5=z3
+
+        movdqa  xmm4,xmm2               ; xmm4=tmp10
+        psubw   xmm2,xmm0
+        pmulhw  xmm2,[rel PW_F0382] ; xmm2=z5
+        pmulhw  xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm0,[rel PW_F1306] ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm2               ; xmm4=z2
+        paddw   xmm0,xmm2               ; xmm0=z4
+
+        movdqa  xmm3,xmm6
+        psubw   xmm6,xmm5               ; xmm6=z13
+        paddw   xmm3,xmm5               ; xmm3=z11
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm5,xmm3
+        psubw   xmm6,xmm4               ; xmm6=data3
+        psubw   xmm3,xmm0               ; xmm3=data7
+        paddw   xmm2,xmm4               ; xmm2=data5
+        paddw   xmm5,xmm0               ; xmm5=data1
+
+        ; ---- Pass 2: process columns.
+
+        ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+        ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm4,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm4,xmm5             ; xmm4=(40 41 50 51 60 61 70 71)
+        movdqa    xmm0,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm6             ; xmm7=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm0,xmm6             ; xmm0=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=col4
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=col6
+
+        ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+        ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm7,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm2             ; xmm5=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm7,xmm2             ; xmm7=(44 45 54 55 64 65 74 75)
+        movdqa    xmm0,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm3             ; xmm6=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm0,xmm3             ; xmm0=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm2,xmm5             ; transpose coefficients(phase 2)
+        punpckldq xmm5,xmm6             ; xmm5=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm2,xmm6             ; xmm2=(24 25 26 27 34 35 36 37)
+        movdqa    xmm3,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm0             ; xmm7=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm3,xmm0             ; xmm3=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=(02 03 12 13 22 23 32 33)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm6             ; xmm1=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm2,xmm6             ; xmm2=(20 21 22 23 30 31 32 33)
+        movdqa    xmm7,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm7,xmm0             ; xmm7=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm5            ; xmm1=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm6,xmm5            ; xmm6=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm0,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm3            ; xmm7=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm0,xmm3            ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm3,xmm1
+        psubw   xmm6,xmm7               ; xmm6=data1-data6=tmp6
+        psubw   xmm1,xmm0               ; xmm1=data0-data7=tmp7
+        paddw   xmm5,xmm7               ; xmm5=data1+data6=tmp1
+        paddw   xmm3,xmm0               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(24 25 26 27 34 35 36 37)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=tmp7
+
+        movdqa     xmm6,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm7            ; xmm2=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm6,xmm7            ; xmm6=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm1,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm1,xmm0            ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm7,xmm6
+        movdqa  xmm0,xmm2
+        paddw   xmm6,xmm4               ; xmm6=data3+data4=tmp3
+        paddw   xmm2,xmm1               ; xmm2=data2+data5=tmp2
+        psubw   xmm7,xmm4               ; xmm7=data3-data4=tmp4
+        psubw   xmm0,xmm1               ; xmm0=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm1,xmm5
+        psubw   xmm3,xmm6               ; xmm3=tmp13
+        psubw   xmm5,xmm2               ; xmm5=tmp12
+        paddw   xmm4,xmm6               ; xmm4=tmp10
+        paddw   xmm1,xmm2               ; xmm1=tmp11
+
+        paddw   xmm5,xmm3
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F0707] ; xmm5=z1
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm2,xmm3
+        psubw   xmm4,xmm1               ; xmm4=data4
+        psubw   xmm3,xmm5               ; xmm3=data6
+        paddw   xmm6,xmm1               ; xmm6=data0
+        paddw   xmm2,xmm5               ; xmm2=data2
+
+        movdqa  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm2
+
+        ; -- Odd part
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        paddw   xmm7,xmm0               ; xmm7=tmp10
+        paddw   xmm0,xmm1               ; xmm0=tmp11
+        paddw   xmm1,xmm5               ; xmm1=tmp12, xmm5=tmp7
+
+        psllw   xmm7,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm0,[rel PW_F0707] ; xmm0=z3
+
+        movdqa  xmm4,xmm7               ; xmm4=tmp10
+        psubw   xmm7,xmm1
+        pmulhw  xmm7,[rel PW_F0382] ; xmm7=z5
+        pmulhw  xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm1,[rel PW_F1306] ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm7               ; xmm4=z2
+        paddw   xmm1,xmm7               ; xmm1=z4
+
+        movdqa  xmm3,xmm5
+        psubw   xmm5,xmm0               ; xmm5=z13
+        paddw   xmm3,xmm0               ; xmm3=z11
+
+        movdqa  xmm6,xmm5
+        movdqa  xmm2,xmm3
+        psubw   xmm5,xmm4               ; xmm5=data3
+        psubw   xmm3,xmm1               ; xmm3=data7
+        paddw   xmm6,xmm4               ; xmm6=data5
+        paddw   xmm2,xmm1               ; xmm2=data1
+
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm5
+        movdqa  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm2
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctfst-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,404 @@
+;
+; jfdctfst.asm - fast integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_sse2) PRIVATE
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707        times 8 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 8 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 8 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 8 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2 (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_fdct_ifast_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(0)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        psubw   xmm3,xmm1               ; xmm3=tmp13
+        psubw   xmm6,xmm7               ; xmm6=tmp12
+        paddw   xmm4,xmm1               ; xmm4=tmp10
+        paddw   xmm0,xmm7               ; xmm0=tmp11
+
+        paddw   xmm6,xmm3
+        psllw   xmm6,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm6,[GOTOFF(ebx,PW_F0707)] ; xmm6=z1
+
+        movdqa  xmm1,xmm4
+        movdqa  xmm7,xmm3
+        psubw   xmm4,xmm0               ; xmm4=data4
+        psubw   xmm3,xmm6               ; xmm3=data6
+        paddw   xmm1,xmm0               ; xmm1=data0
+        paddw   xmm7,xmm6               ; xmm7=data2
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=tmp6
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp7
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=data4
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=data6
+
+        ; -- Odd part
+
+        paddw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm5,xmm0               ; xmm5=tmp11
+        paddw   xmm0,xmm6               ; xmm0=tmp12, xmm6=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z3
+
+        movdqa  xmm4,xmm2               ; xmm4=tmp10
+        psubw   xmm2,xmm0
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F0382)] ; xmm2=z5
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F1306)] ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm2               ; xmm4=z2
+        paddw   xmm0,xmm2               ; xmm0=z4
+
+        movdqa  xmm3,xmm6
+        psubw   xmm6,xmm5               ; xmm6=z13
+        paddw   xmm3,xmm5               ; xmm3=z11
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm5,xmm3
+        psubw   xmm6,xmm4               ; xmm6=data3
+        psubw   xmm3,xmm0               ; xmm3=data7
+        paddw   xmm2,xmm4               ; xmm2=data5
+        paddw   xmm5,xmm0               ; xmm5=data1
+
+        ; ---- Pass 2: process columns.
+
+;       mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+        ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm4,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm4,xmm5             ; xmm4=(40 41 50 51 60 61 70 71)
+        movdqa    xmm0,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm6             ; xmm7=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm0,xmm6             ; xmm0=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=col4
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=col6
+
+        ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+        ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm7,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm2             ; xmm5=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm7,xmm2             ; xmm7=(44 45 54 55 64 65 74 75)
+        movdqa    xmm0,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm3             ; xmm6=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm0,xmm3             ; xmm0=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm2,xmm5             ; transpose coefficients(phase 2)
+        punpckldq xmm5,xmm6             ; xmm5=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm2,xmm6             ; xmm2=(24 25 26 27 34 35 36 37)
+        movdqa    xmm3,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm0             ; xmm7=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm3,xmm0             ; xmm3=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=(02 03 12 13 22 23 32 33)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm6             ; xmm1=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm2,xmm6             ; xmm2=(20 21 22 23 30 31 32 33)
+        movdqa    xmm7,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm7,xmm0             ; xmm7=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm5            ; xmm1=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm6,xmm5            ; xmm6=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm0,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm3            ; xmm7=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm0,xmm3            ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm3,xmm1
+        psubw   xmm6,xmm7               ; xmm6=data1-data6=tmp6
+        psubw   xmm1,xmm0               ; xmm1=data0-data7=tmp7
+        paddw   xmm5,xmm7               ; xmm5=data1+data6=tmp1
+        paddw   xmm3,xmm0               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(24 25 26 27 34 35 36 37)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=tmp7
+
+        movdqa     xmm6,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm7            ; xmm2=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm6,xmm7            ; xmm6=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm1,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm1,xmm0            ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm7,xmm6
+        movdqa  xmm0,xmm2
+        paddw   xmm6,xmm4               ; xmm6=data3+data4=tmp3
+        paddw   xmm2,xmm1               ; xmm2=data2+data5=tmp2
+        psubw   xmm7,xmm4               ; xmm7=data3-data4=tmp4
+        psubw   xmm0,xmm1               ; xmm0=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm1,xmm5
+        psubw   xmm3,xmm6               ; xmm3=tmp13
+        psubw   xmm5,xmm2               ; xmm5=tmp12
+        paddw   xmm4,xmm6               ; xmm4=tmp10
+        paddw   xmm1,xmm2               ; xmm1=tmp11
+
+        paddw   xmm5,xmm3
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z1
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm2,xmm3
+        psubw   xmm4,xmm1               ; xmm4=data4
+        psubw   xmm3,xmm5               ; xmm3=data6
+        paddw   xmm6,xmm1               ; xmm6=data0
+        paddw   xmm2,xmm5               ; xmm2=data2
+
+        movdqa  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm2
+
+        ; -- Odd part
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        paddw   xmm7,xmm0               ; xmm7=tmp10
+        paddw   xmm0,xmm1               ; xmm0=tmp11
+        paddw   xmm1,xmm5               ; xmm1=tmp12, xmm5=tmp7
+
+        psllw   xmm7,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F0707)] ; xmm0=z3
+
+        movdqa  xmm4,xmm7               ; xmm4=tmp10
+        psubw   xmm7,xmm1
+        pmulhw  xmm7,[GOTOFF(ebx,PW_F0382)] ; xmm7=z5
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm1,[GOTOFF(ebx,PW_F1306)] ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm7               ; xmm4=z2
+        paddw   xmm1,xmm7               ; xmm1=z4
+
+        movdqa  xmm3,xmm5
+        psubw   xmm5,xmm0               ; xmm5=z13
+        paddw   xmm3,xmm0               ; xmm3=z11
+
+        movdqa  xmm6,xmm5
+        movdqa  xmm2,xmm3
+        psubw   xmm5,xmm4               ; xmm5=data3
+        psubw   xmm3,xmm1               ; xmm3=data7
+        paddw   xmm6,xmm4               ; xmm6=data5
+        paddw   xmm2,xmm1               ; xmm2=data1
+
+        movdqa  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm5
+        movdqa  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm2
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,262 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* SLOW INTEGER FORWARD DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298 2446   /* FIX(0.298631336) */
+#define F_0_390 3196   /* FIX(0.390180644) */
+#define F_0_541 4433   /* FIX(0.541196100) */
+#define F_0_765 6270   /* FIX(0.765366865) */
+#define F_0_899 7373   /* FIX(0.899976223) */
+#define F_1_175 9633   /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+#define DESCALE_P1 (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS + PASS1_BITS)
+
+
+#define DO_FDCT_COMMON(PASS)  \
+{  \
+  /* (Original)  \
+   * z1 = (tmp12 + tmp13) * 0.541196100;  \
+   * data2 = z1 + tmp13 * 0.765366865;  \
+   * data6 = z1 + tmp12 * -1.847759065;  \
+   *  \
+   * (This implementation)  \
+   * data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;  \
+   * data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);  \
+   */  \
+  \
+  tmp1312l = vec_mergeh(tmp13, tmp12);  \
+  tmp1312h = vec_mergel(tmp13, tmp12);  \
+  \
+  out2l = vec_msums(tmp1312l, pw_f130_f054, pd_descale_p##PASS);  \
+  out2h = vec_msums(tmp1312h, pw_f130_f054, pd_descale_p##PASS);  \
+  out6l = vec_msums(tmp1312l, pw_f054_mf130, pd_descale_p##PASS);  \
+  out6h = vec_msums(tmp1312h, pw_f054_mf130, pd_descale_p##PASS);  \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS);  \
+  out2h = vec_sra(out2h, descale_p##PASS);  \
+  out6l = vec_sra(out6l, descale_p##PASS);  \
+  out6h = vec_sra(out6h, descale_p##PASS);  \
+  \
+  out2 = vec_pack(out2l, out2h);  \
+  out6 = vec_pack(out6l, out6h);  \
+  \
+  /* Odd part */  \
+  \
+  z3 = vec_add(tmp4, tmp6);  \
+  z4 = vec_add(tmp5, tmp7);  \
+  \
+  /* (Original)  \
+   * z5 = (z3 + z4) * 1.175875602;  \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;  \
+   * z3 += z5;  z4 += z5;  \
+   *  \
+   * (This implementation)  \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;  \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);  \
+   */  \
+  \
+  z34l = vec_mergeh(z3, z4);  \
+  z34h = vec_mergel(z3, z4);  \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_descale_p##PASS);  \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_descale_p##PASS);  \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_descale_p##PASS);  \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_descale_p##PASS);  \
+  \
+  /* (Original)  \
+   * z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;  \
+   * tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;  \
+   * tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;  \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;  \
+   * data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;  \
+   * data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;  \
+   *  \
+   * (This implementation)  \
+   * tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;  \
+   * tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;  \
+   * tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);  \
+   * tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);  \
+   * data7 = tmp4 + z3;  data5 = tmp5 + z4;  \
+   * data3 = tmp6 + z3;  data1 = tmp7 + z4;  \
+   */  \
+  \
+  tmp47l = vec_mergeh(tmp4, tmp7);  \
+  tmp47h = vec_mergel(tmp4, tmp7);  \
+  \
+  out7l = vec_msums(tmp47l, pw_mf060_mf089, z3l);  \
+  out7h = vec_msums(tmp47h, pw_mf060_mf089, z3h);  \
+  out1l = vec_msums(tmp47l, pw_mf089_f060, z4l);  \
+  out1h = vec_msums(tmp47h, pw_mf089_f060, z4h);  \
+  \
+  out7l = vec_sra(out7l, descale_p##PASS);  \
+  out7h = vec_sra(out7h, descale_p##PASS);  \
+  out1l = vec_sra(out1l, descale_p##PASS);  \
+  out1h = vec_sra(out1h, descale_p##PASS);  \
+  \
+  out7 = vec_pack(out7l, out7h);  \
+  out1 = vec_pack(out1l, out1h);  \
+  \
+  tmp56l = vec_mergeh(tmp5, tmp6);  \
+  tmp56h = vec_mergel(tmp5, tmp6);  \
+  \
+  out5l = vec_msums(tmp56l, pw_mf050_mf256, z4l);  \
+  out5h = vec_msums(tmp56h, pw_mf050_mf256, z4h);  \
+  out3l = vec_msums(tmp56l, pw_mf256_f050, z3l);  \
+  out3h = vec_msums(tmp56h, pw_mf256_f050, z3h);  \
+  \
+  out5l = vec_sra(out5l, descale_p##PASS);  \
+  out5h = vec_sra(out5h, descale_p##PASS);  \
+  out3l = vec_sra(out3l, descale_p##PASS);  \
+  out3h = vec_sra(out3h, descale_p##PASS);  \
+  \
+  out5 = vec_pack(out5l, out5h);  \
+  out3 = vec_pack(out3l, out3h);  \
+}
+
+#define DO_FDCT_PASS1()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out0  = vec_sl(out0, pass1_bits);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  out4  = vec_sl(out4, pass1_bits);  \
+  \
+  DO_FDCT_COMMON(1);  \
+}
+
+#define DO_FDCT_PASS2()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out0  = vec_add(out0, pw_descale_p2x);  \
+  out0  = vec_sra(out0, pass1_bits);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  out4  = vec_add(out4, pw_descale_p2x);  \
+  out4  = vec_sra(out4, pass1_bits);  \
+  \
+  DO_FDCT_COMMON(2);  \
+}
+
+
+void
+jsimd_fdct_islow_altivec (DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    tmp47l, tmp47h, tmp56l, tmp56h, tmp1312l, tmp1312h,
+    z3, z4, z34l, z34h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int z3l, z3h, z4l, z4h,
+    out1l, out1h, out2l, out2h, out3l, out3h, out5l, out5h, out6l, out6h,
+    out7l, out7h;
+
+  /* Constants */
+  __vector short
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) },
+    pw_descale_p2x = { __8X(1 << (PASS1_BITS - 1)) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT_PASS1();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT_PASS2();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,622 @@
+;
+; jfdctint.asm - accurate integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_mmx) PRIVATE
+
+EXTN(jconst_fdct_islow_mmx):
+
+PW_F130_F054    times 2 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 2 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 2 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 2 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 2 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 2 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 2 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 2 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 2 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 2 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 4 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_mmx (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_mmx) PRIVATE
+
+EXTN(jsimd_fdct_islow_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+        ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(24 34 25 35)
+        punpckhwd mm5,mm3               ; mm5=(26 36 27 37)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+        ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 32 23 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(24 34 25 35)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm4,mm7               ; mm4=(02 12 03 13)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(04 14 05 15)
+        punpckhwd mm2,mm3               ; mm2=(06 16 07 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 10 20 30)=data0
+        punpckhdq mm7,mm0               ; mm7=(01 11 21 31)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(06 16 26 36)=data6
+        punpckhdq mm3,mm5               ; mm3=(07 17 27 37)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 32 23 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(24 34 25 35)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(02 12 22 32)=data2
+        punpckhdq mm7,mm2               ; mm7=(03 13 23 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(04 14 24 34)=data4
+        punpckhdq mm6,mm3               ; mm6=(05 15 25 35)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        paddw   mm5,mm7                 ; mm5=tmp10
+        paddw   mm0,mm4                 ; mm0=tmp11
+        psubw   mm1,mm7                 ; mm1=tmp13
+        psubw   mm6,mm4                 ; mm6=tmp12
+
+        movq    mm7,mm5
+        paddw   mm5,mm0                 ; mm5=tmp10+tmp11
+        psubw   mm7,mm0                 ; mm7=tmp10-tmp11
+
+        psllw   mm5,PASS1_BITS          ; mm5=data0
+        psllw   mm7,PASS1_BITS          ; mm7=data4
+
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm7
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movq      mm4,mm1               ; mm1=tmp13
+        movq      mm0,mm1
+        punpcklwd mm4,mm6               ; mm6=tmp12
+        punpckhwd mm0,mm6
+        movq      mm1,mm4
+        movq      mm6,mm0
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=data2L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F130_F054)]        ; mm0=data2H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=data6L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F054_MF130)]       ; mm6=data6H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm0,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm4,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm6,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm1,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm4,mm0               ; mm4=data2
+        packssdw  mm1,mm6               ; mm1=data6
+
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm1
+
+        ; -- Odd part
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp6
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp7
+
+        movq    mm0,mm2                 ; mm2=tmp4
+        movq    mm6,mm3                 ; mm3=tmp5
+        paddw   mm0,mm5                 ; mm0=z3
+        paddw   mm6,mm7                 ; mm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm4,mm0
+        movq      mm1,mm0
+        punpcklwd mm4,mm6
+        punpckhwd mm1,mm6
+        movq      mm0,mm4
+        movq      mm6,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF078_F117)]       ; mm4=z3L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF078_F117)]       ; mm1=z3H
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F117_F078)]        ; mm0=z4L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F117_F078)]        ; mm6=z4H
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=z3L
+        movq    MMWORD [wk(1)], mm1     ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movq      mm4,mm2
+        movq      mm1,mm2
+        punpcklwd mm4,mm7
+        punpckhwd mm1,mm7
+        movq      mm2,mm4
+        movq      mm7,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm4=tmp4L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm1=tmp4H
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF089_F060)]       ; mm2=tmp7L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF089_F060)]       ; mm7=tmp7H
+
+        paddd   mm4, MMWORD [wk(0)]     ; mm4=data7L
+        paddd   mm1, MMWORD [wk(1)]     ; mm1=data7H
+        paddd   mm2,mm0                 ; mm2=data1L
+        paddd   mm7,mm6                 ; mm7=data1H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm4,DESCALE_P1
+        psrad   mm1,DESCALE_P1
+        paddd   mm2,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm2,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+
+        packssdw  mm4,mm1               ; mm4=data7
+        packssdw  mm2,mm7               ; mm2=data1
+
+        movq    MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2
+
+        movq      mm1,mm3
+        movq      mm7,mm3
+        punpcklwd mm1,mm5
+        punpckhwd mm7,mm5
+        movq      mm3,mm1
+        movq      mm5,mm7
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm1=tmp5L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm7=tmp5H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF256_F050)]       ; mm3=tmp6L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_MF256_F050)]       ; mm5=tmp6H
+
+        paddd   mm1,mm0                 ; mm1=data5L
+        paddd   mm7,mm6                 ; mm7=data5H
+        paddd   mm3, MMWORD [wk(0)]     ; mm3=data3L
+        paddd   mm5, MMWORD [wk(1)]     ; mm5=data3H
+
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm1,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+        paddd   mm3,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm5,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm3,DESCALE_P1
+        psrad   mm5,DESCALE_P1
+
+        packssdw  mm1,mm7               ; mm1=data5
+        packssdw  mm3,mm5               ; mm3=data3
+
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3
+
+        add     edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+        ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(02 03 12 13)
+        punpckhwd mm4,mm1               ; mm4=(22 23 32 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(42 43 52 53)
+        punpckhwd mm5,mm3               ; mm5=(62 63 72 73)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+        ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 23 32 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(42 43 52 53)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 01 10 11)
+        punpckhwd mm4,mm7               ; mm4=(20 21 30 31)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(40 41 50 51)
+        punpckhwd mm2,mm3               ; mm2=(60 61 70 71)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03)=data0
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(60 61 62 63)=data6
+        punpckhdq mm3,mm5               ; mm3=(70 71 72 73)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 23 32 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(42 43 52 53)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(20 21 22 23)=data2
+        punpckhdq mm7,mm2               ; mm7=(30 31 32 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(40 41 42 43)=data4
+        punpckhdq mm6,mm3               ; mm6=(50 51 52 53)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        paddw   mm5,mm7                 ; mm5=tmp10
+        paddw   mm0,mm4                 ; mm0=tmp11
+        psubw   mm1,mm7                 ; mm1=tmp13
+        psubw   mm6,mm4                 ; mm6=tmp12
+
+        movq    mm7,mm5
+        paddw   mm5,mm0                 ; mm5=tmp10+tmp11
+        psubw   mm7,mm0                 ; mm7=tmp10-tmp11
+
+        paddw   mm5,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        paddw   mm7,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        psraw   mm5,PASS1_BITS          ; mm5=data0
+        psraw   mm7,PASS1_BITS          ; mm7=data4
+
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm7
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movq      mm4,mm1               ; mm1=tmp13
+        movq      mm0,mm1
+        punpcklwd mm4,mm6               ; mm6=tmp12
+        punpckhwd mm0,mm6
+        movq      mm1,mm4
+        movq      mm6,mm0
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=data2L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F130_F054)]        ; mm0=data2H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=data6L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F054_MF130)]       ; mm6=data6H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm0,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm4,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm6,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm1,DESCALE_P2
+        psrad   mm6,DESCALE_P2
+
+        packssdw  mm4,mm0               ; mm4=data2
+        packssdw  mm1,mm6               ; mm1=data6
+
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm1
+
+        ; -- Odd part
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp6
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp7
+
+        movq    mm0,mm2                 ; mm2=tmp4
+        movq    mm6,mm3                 ; mm3=tmp5
+        paddw   mm0,mm5                 ; mm0=z3
+        paddw   mm6,mm7                 ; mm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm4,mm0
+        movq      mm1,mm0
+        punpcklwd mm4,mm6
+        punpckhwd mm1,mm6
+        movq      mm0,mm4
+        movq      mm6,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF078_F117)]       ; mm4=z3L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF078_F117)]       ; mm1=z3H
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F117_F078)]        ; mm0=z4L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F117_F078)]        ; mm6=z4H
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=z3L
+        movq    MMWORD [wk(1)], mm1     ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movq      mm4,mm2
+        movq      mm1,mm2
+        punpcklwd mm4,mm7
+        punpckhwd mm1,mm7
+        movq      mm2,mm4
+        movq      mm7,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm4=tmp4L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm1=tmp4H
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF089_F060)]       ; mm2=tmp7L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF089_F060)]       ; mm7=tmp7H
+
+        paddd   mm4, MMWORD [wk(0)]     ; mm4=data7L
+        paddd   mm1, MMWORD [wk(1)]     ; mm1=data7H
+        paddd   mm2,mm0                 ; mm2=data1L
+        paddd   mm7,mm6                 ; mm7=data1H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm4,DESCALE_P2
+        psrad   mm1,DESCALE_P2
+        paddd   mm2,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm2,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+
+        packssdw  mm4,mm1               ; mm4=data7
+        packssdw  mm2,mm7               ; mm2=data1
+
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2
+
+        movq      mm1,mm3
+        movq      mm7,mm3
+        punpcklwd mm1,mm5
+        punpckhwd mm7,mm5
+        movq      mm3,mm1
+        movq      mm5,mm7
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm1=tmp5L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm7=tmp5H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF256_F050)]       ; mm3=tmp6L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_MF256_F050)]       ; mm5=tmp6H
+
+        paddd   mm1,mm0                 ; mm1=data5L
+        paddd   mm7,mm6                 ; mm7=data5H
+        paddd   mm3, MMWORD [wk(0)]     ; mm3=data3L
+        paddd   mm5, MMWORD [wk(1)]     ; mm5=data3H
+
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm1,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+        paddd   mm3,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm3,DESCALE_P2
+        psrad   mm5,DESCALE_P2
+
+        packssdw  mm1,mm7               ; mm1=data5
+        packssdw  mm3,mm5               ; mm3=data3
+
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3
+
+        add     edx, byte 4*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .columnloop
+
+        emms            ; empty MMX state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,622 @@
+;
+; jfdctint.asm - accurate integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_sse2) PRIVATE
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 8 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2 (DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          6
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_sse2) PRIVATE
+
+EXTN(jsimd_fdct_islow_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(3)], xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(3)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        paddw   xmm3,xmm1               ; xmm3=tmp10
+        paddw   xmm6,xmm7               ; xmm6=tmp11
+        psubw   xmm4,xmm1               ; xmm4=tmp13
+        psubw   xmm0,xmm7               ; xmm0=tmp12
+
+        movdqa  xmm1,xmm3
+        paddw   xmm3,xmm6               ; xmm3=tmp10+tmp11
+        psubw   xmm1,xmm6               ; xmm1=tmp10-tmp11
+
+        psllw   xmm3,PASS1_BITS         ; xmm3=data0
+        psllw   xmm1,PASS1_BITS         ; xmm1=data4
+
+        movdqa  XMMWORD [wk(2)], xmm3   ; wk(2)=data0
+        movdqa  XMMWORD [wk(3)], xmm1   ; wk(3)=data4
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm7,xmm4             ; xmm4=tmp13
+        movdqa    xmm6,xmm4
+        punpcklwd xmm7,xmm0             ; xmm0=tmp12
+        punpckhwd xmm6,xmm0
+        movdqa    xmm4,xmm7
+        movdqa    xmm0,xmm6
+        pmaddwd   xmm7,[rel PW_F130_F054]       ; xmm7=data2L
+        pmaddwd   xmm6,[rel PW_F130_F054]       ; xmm6=data2H
+        pmaddwd   xmm4,[rel PW_F054_MF130]      ; xmm4=data6L
+        pmaddwd   xmm0,[rel PW_F054_MF130]      ; xmm0=data6H
+
+        paddd   xmm7,[rel PD_DESCALE_P1]
+        paddd   xmm6,[rel PD_DESCALE_P1]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        paddd   xmm0,[rel PD_DESCALE_P1]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm7,xmm6             ; xmm7=data2
+        packssdw  xmm4,xmm0             ; xmm4=data6
+
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=data2
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=data6
+
+        ; -- Odd part
+
+        movdqa  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp6
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp7
+
+        movdqa  xmm6,xmm2               ; xmm2=tmp4
+        movdqa  xmm0,xmm5               ; xmm5=tmp5
+        paddw   xmm6,xmm3               ; xmm6=z3
+        paddw   xmm0,xmm1               ; xmm0=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm7,xmm6
+        movdqa    xmm4,xmm6
+        punpcklwd xmm7,xmm0
+        punpckhwd xmm4,xmm0
+        movdqa    xmm6,xmm7
+        movdqa    xmm0,xmm4
+        pmaddwd   xmm7,[rel PW_MF078_F117]      ; xmm7=z3L
+        pmaddwd   xmm4,[rel PW_MF078_F117]      ; xmm4=z3H
+        pmaddwd   xmm6,[rel PW_F117_F078]       ; xmm6=z4L
+        pmaddwd   xmm0,[rel PW_F117_F078]       ; xmm0=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm7,xmm2
+        movdqa    xmm4,xmm2
+        punpcklwd xmm7,xmm1
+        punpckhwd xmm4,xmm1
+        movdqa    xmm2,xmm7
+        movdqa    xmm1,xmm4
+        pmaddwd   xmm7,[rel PW_MF060_MF089]     ; xmm7=tmp4L
+        pmaddwd   xmm4,[rel PW_MF060_MF089]     ; xmm4=tmp4H
+        pmaddwd   xmm2,[rel PW_MF089_F060]      ; xmm2=tmp7L
+        pmaddwd   xmm1,[rel PW_MF089_F060]      ; xmm1=tmp7H
+
+        paddd   xmm7, XMMWORD [wk(0)]   ; xmm7=data7L
+        paddd   xmm4, XMMWORD [wk(1)]   ; xmm4=data7H
+        paddd   xmm2,xmm6               ; xmm2=data1L
+        paddd   xmm1,xmm0               ; xmm1=data1H
+
+        paddd   xmm7,[rel PD_DESCALE_P1]
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm2,[rel PD_DESCALE_P1]
+        paddd   xmm1,[rel PD_DESCALE_P1]
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+
+        packssdw  xmm7,xmm4             ; xmm7=data7
+        packssdw  xmm2,xmm1             ; xmm2=data1
+
+        movdqa    xmm4,xmm5
+        movdqa    xmm1,xmm5
+        punpcklwd xmm4,xmm3
+        punpckhwd xmm1,xmm3
+        movdqa    xmm5,xmm4
+        movdqa    xmm3,xmm1
+        pmaddwd   xmm4,[rel PW_MF050_MF256]     ; xmm4=tmp5L
+        pmaddwd   xmm1,[rel PW_MF050_MF256]     ; xmm1=tmp5H
+        pmaddwd   xmm5,[rel PW_MF256_F050]      ; xmm5=tmp6L
+        pmaddwd   xmm3,[rel PW_MF256_F050]      ; xmm3=tmp6H
+
+        paddd   xmm4,xmm6               ; xmm4=data5L
+        paddd   xmm1,xmm0               ; xmm1=data5H
+        paddd   xmm5, XMMWORD [wk(0)]   ; xmm5=data3L
+        paddd   xmm3, XMMWORD [wk(1)]   ; xmm3=data3H
+
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        paddd   xmm1,[rel PD_DESCALE_P1]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+        paddd   xmm5,[rel PD_DESCALE_P1]
+        paddd   xmm3,[rel PD_DESCALE_P1]
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+
+        packssdw  xmm4,xmm1             ; xmm4=data5
+        packssdw  xmm5,xmm3             ; xmm5=data3
+
+        ; ---- Pass 2: process columns.
+
+        movdqa  xmm6, XMMWORD [wk(2)]   ; xmm6=col0
+        movdqa  xmm0, XMMWORD [wk(4)]   ; xmm0=col2
+
+        ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+        ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm1,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm2             ; xmm6=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm1,xmm2             ; xmm1=(40 41 50 51 60 61 70 71)
+        movdqa    xmm3,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm3,xmm5             ; xmm3=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm2, XMMWORD [wk(3)]   ; xmm2=col4
+        movdqa  xmm5, XMMWORD [wk(5)]   ; xmm5=col6
+
+        ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+        ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm4             ; xmm2=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm0,xmm4             ; xmm0=(44 45 54 55 64 65 74 75)
+        movdqa    xmm3,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm7             ; xmm5=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm3,xmm7             ; xmm3=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm4,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm4,xmm5             ; xmm4=(24 25 26 27 34 35 36 37)
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm7,xmm3             ; xmm7=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=(02 03 12 13 22 23 32 33)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(2)], xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(3)], xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm4,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm5             ; xmm6=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm4,xmm5             ; xmm4=(20 21 22 23 30 31 32 33)
+        movdqa    xmm0,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm3             ; xmm1=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm0,xmm3             ; xmm0=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm5,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm2            ; xmm6=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm5,xmm2            ; xmm5=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm7            ; xmm0=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm3,xmm7            ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm7,xmm6
+        psubw   xmm5,xmm0               ; xmm5=data1-data6=tmp6
+        psubw   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        paddw   xmm2,xmm0               ; xmm2=data1+data6=tmp1
+        paddw   xmm7,xmm3               ; xmm7=data0+data7=tmp0
+
+        movdqa  xmm0, XMMWORD [wk(2)]   ; xmm0=(24 25 26 27 34 35 36 37)
+        movdqa  xmm3, XMMWORD [wk(3)]   ; xmm3=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movdqa     xmm5,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm5,xmm0            ; xmm5=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm3            ; xmm1=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm6,xmm3            ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm3,xmm4
+        paddw   xmm5,xmm1               ; xmm5=data3+data4=tmp3
+        paddw   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        psubw   xmm0,xmm1               ; xmm0=data3-data4=tmp4
+        psubw   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm6,xmm2
+        paddw   xmm7,xmm5               ; xmm7=tmp10
+        paddw   xmm2,xmm4               ; xmm2=tmp11
+        psubw   xmm1,xmm5               ; xmm1=tmp13
+        psubw   xmm6,xmm4               ; xmm6=tmp12
+
+        movdqa  xmm5,xmm7
+        paddw   xmm7,xmm2               ; xmm7=tmp10+tmp11
+        psubw   xmm5,xmm2               ; xmm5=tmp10-tmp11
+
+        paddw   xmm7,[rel PW_DESCALE_P2X]
+        paddw   xmm5,[rel PW_DESCALE_P2X]
+        psraw   xmm7,PASS1_BITS         ; xmm7=data0
+        psraw   xmm5,PASS1_BITS         ; xmm5=data4
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm7
+        movdqa  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm5
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm4,xmm1             ; xmm1=tmp13
+        movdqa    xmm2,xmm1
+        punpcklwd xmm4,xmm6             ; xmm6=tmp12
+        punpckhwd xmm2,xmm6
+        movdqa    xmm1,xmm4
+        movdqa    xmm6,xmm2
+        pmaddwd   xmm4,[rel PW_F130_F054]       ; xmm4=data2L
+        pmaddwd   xmm2,[rel PW_F130_F054]       ; xmm2=data2H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=data6L
+        pmaddwd   xmm6,[rel PW_F054_MF130]      ; xmm6=data6H
+
+        paddd   xmm4,[rel PD_DESCALE_P2]
+        paddd   xmm2,[rel PD_DESCALE_P2]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        paddd   xmm6,[rel PD_DESCALE_P2]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm6,DESCALE_P2
+
+        packssdw  xmm4,xmm2             ; xmm4=data2
+        packssdw  xmm1,xmm6             ; xmm1=data6
+
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm1
+
+        ; -- Odd part
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        movdqa  xmm2,xmm0               ; xmm0=tmp4
+        movdqa  xmm6,xmm3               ; xmm3=tmp5
+        paddw   xmm2,xmm7               ; xmm2=z3
+        paddw   xmm6,xmm5               ; xmm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm1,xmm2
+        punpcklwd xmm4,xmm6
+        punpckhwd xmm1,xmm6
+        movdqa    xmm2,xmm4
+        movdqa    xmm6,xmm1
+        pmaddwd   xmm4,[rel PW_MF078_F117]      ; xmm4=z3L
+        pmaddwd   xmm1,[rel PW_MF078_F117]      ; xmm1=z3H
+        pmaddwd   xmm2,[rel PW_F117_F078]       ; xmm2=z4L
+        pmaddwd   xmm6,[rel PW_F117_F078]       ; xmm6=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm1,xmm0
+        punpcklwd xmm4,xmm5
+        punpckhwd xmm1,xmm5
+        movdqa    xmm0,xmm4
+        movdqa    xmm5,xmm1
+        pmaddwd   xmm4,[rel PW_MF060_MF089]     ; xmm4=tmp4L
+        pmaddwd   xmm1,[rel PW_MF060_MF089]     ; xmm1=tmp4H
+        pmaddwd   xmm0,[rel PW_MF089_F060]      ; xmm0=tmp7L
+        pmaddwd   xmm5,[rel PW_MF089_F060]      ; xmm5=tmp7H
+
+        paddd   xmm4, XMMWORD [wk(0)]   ; xmm4=data7L
+        paddd   xmm1, XMMWORD [wk(1)]   ; xmm1=data7H
+        paddd   xmm0,xmm2               ; xmm0=data1L
+        paddd   xmm5,xmm6               ; xmm5=data1H
+
+        paddd   xmm4,[rel PD_DESCALE_P2]
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm0,[rel PD_DESCALE_P2]
+        paddd   xmm5,[rel PD_DESCALE_P2]
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+
+        packssdw  xmm4,xmm1             ; xmm4=data7
+        packssdw  xmm0,xmm5             ; xmm0=data1
+
+        movdqa  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm0
+
+        movdqa    xmm1,xmm3
+        movdqa    xmm5,xmm3
+        punpcklwd xmm1,xmm7
+        punpckhwd xmm5,xmm7
+        movdqa    xmm3,xmm1
+        movdqa    xmm7,xmm5
+        pmaddwd   xmm1,[rel PW_MF050_MF256]     ; xmm1=tmp5L
+        pmaddwd   xmm5,[rel PW_MF050_MF256]     ; xmm5=tmp5H
+        pmaddwd   xmm3,[rel PW_MF256_F050]      ; xmm3=tmp6L
+        pmaddwd   xmm7,[rel PW_MF256_F050]      ; xmm7=tmp6H
+
+        paddd   xmm1,xmm2               ; xmm1=data5L
+        paddd   xmm5,xmm6               ; xmm5=data5H
+        paddd   xmm3, XMMWORD [wk(0)]   ; xmm3=data3L
+        paddd   xmm7, XMMWORD [wk(1)]   ; xmm7=data3H
+
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        paddd   xmm5,[rel PD_DESCALE_P2]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+        paddd   xmm3,[rel PD_DESCALE_P2]
+        paddd   xmm7,[rel PD_DESCALE_P2]
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm1,xmm5             ; xmm1=data5
+        packssdw  xmm3,xmm7             ; xmm3=data3
+
+        movdqa  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm3
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jfdctint-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,634 @@
+;
+; jfdctint.asm - accurate integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_sse2) PRIVATE
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 8 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2 (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          6
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_sse2) PRIVATE
+
+EXTN(jsimd_fdct_islow_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(3)], xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(3)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        paddw   xmm3,xmm1               ; xmm3=tmp10
+        paddw   xmm6,xmm7               ; xmm6=tmp11
+        psubw   xmm4,xmm1               ; xmm4=tmp13
+        psubw   xmm0,xmm7               ; xmm0=tmp12
+
+        movdqa  xmm1,xmm3
+        paddw   xmm3,xmm6               ; xmm3=tmp10+tmp11
+        psubw   xmm1,xmm6               ; xmm1=tmp10-tmp11
+
+        psllw   xmm3,PASS1_BITS         ; xmm3=data0
+        psllw   xmm1,PASS1_BITS         ; xmm1=data4
+
+        movdqa  XMMWORD [wk(2)], xmm3   ; wk(2)=data0
+        movdqa  XMMWORD [wk(3)], xmm1   ; wk(3)=data4
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm7,xmm4             ; xmm4=tmp13
+        movdqa    xmm6,xmm4
+        punpcklwd xmm7,xmm0             ; xmm0=tmp12
+        punpckhwd xmm6,xmm0
+        movdqa    xmm4,xmm7
+        movdqa    xmm0,xmm6
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F130_F054)]       ; xmm7=data2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F130_F054)]       ; xmm6=data2H
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm4=data6L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm0=data6H
+
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm6,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm0,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm7,xmm6             ; xmm7=data2
+        packssdw  xmm4,xmm0             ; xmm4=data6
+
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=data2
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=data6
+
+        ; -- Odd part
+
+        movdqa  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp6
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp7
+
+        movdqa  xmm6,xmm2               ; xmm2=tmp4
+        movdqa  xmm0,xmm5               ; xmm5=tmp5
+        paddw   xmm6,xmm3               ; xmm6=z3
+        paddw   xmm0,xmm1               ; xmm0=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm7,xmm6
+        movdqa    xmm4,xmm6
+        punpcklwd xmm7,xmm0
+        punpckhwd xmm4,xmm0
+        movdqa    xmm6,xmm7
+        movdqa    xmm0,xmm4
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm7=z3L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm4=z3H
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F117_F078)]       ; xmm6=z4L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F117_F078)]       ; xmm0=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm7,xmm2
+        movdqa    xmm4,xmm2
+        punpcklwd xmm7,xmm1
+        punpckhwd xmm4,xmm1
+        movdqa    xmm2,xmm7
+        movdqa    xmm1,xmm4
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm7=tmp4L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm4=tmp4H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm2=tmp7L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm1=tmp7H
+
+        paddd   xmm7, XMMWORD [wk(0)]   ; xmm7=data7L
+        paddd   xmm4, XMMWORD [wk(1)]   ; xmm4=data7H
+        paddd   xmm2,xmm6               ; xmm2=data1L
+        paddd   xmm1,xmm0               ; xmm1=data1H
+
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm2,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+
+        packssdw  xmm7,xmm4             ; xmm7=data7
+        packssdw  xmm2,xmm1             ; xmm2=data1
+
+        movdqa    xmm4,xmm5
+        movdqa    xmm1,xmm5
+        punpcklwd xmm4,xmm3
+        punpckhwd xmm1,xmm3
+        movdqa    xmm5,xmm4
+        movdqa    xmm3,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm4=tmp5L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm1=tmp5H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm5=tmp6L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm3=tmp6H
+
+        paddd   xmm4,xmm6               ; xmm4=data5L
+        paddd   xmm1,xmm0               ; xmm1=data5H
+        paddd   xmm5, XMMWORD [wk(0)]   ; xmm5=data3L
+        paddd   xmm3, XMMWORD [wk(1)]   ; xmm3=data3H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm3,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+
+        packssdw  xmm4,xmm1             ; xmm4=data5
+        packssdw  xmm5,xmm3             ; xmm5=data3
+
+        ; ---- Pass 2: process columns.
+
+;       mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm6, XMMWORD [wk(2)]   ; xmm6=col0
+        movdqa  xmm0, XMMWORD [wk(4)]   ; xmm0=col2
+
+        ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+        ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm1,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm2             ; xmm6=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm1,xmm2             ; xmm1=(40 41 50 51 60 61 70 71)
+        movdqa    xmm3,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm3,xmm5             ; xmm3=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm2, XMMWORD [wk(3)]   ; xmm2=col4
+        movdqa  xmm5, XMMWORD [wk(5)]   ; xmm5=col6
+
+        ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+        ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm4             ; xmm2=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm0,xmm4             ; xmm0=(44 45 54 55 64 65 74 75)
+        movdqa    xmm3,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm7             ; xmm5=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm3,xmm7             ; xmm3=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm4,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm4,xmm5             ; xmm4=(24 25 26 27 34 35 36 37)
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm7,xmm3             ; xmm7=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=(02 03 12 13 22 23 32 33)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(2)], xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(3)], xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm4,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm5             ; xmm6=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm4,xmm5             ; xmm4=(20 21 22 23 30 31 32 33)
+        movdqa    xmm0,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm3             ; xmm1=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm0,xmm3             ; xmm0=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm5,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm2            ; xmm6=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm5,xmm2            ; xmm5=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm7            ; xmm0=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm3,xmm7            ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm7,xmm6
+        psubw   xmm5,xmm0               ; xmm5=data1-data6=tmp6
+        psubw   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        paddw   xmm2,xmm0               ; xmm2=data1+data6=tmp1
+        paddw   xmm7,xmm3               ; xmm7=data0+data7=tmp0
+
+        movdqa  xmm0, XMMWORD [wk(2)]   ; xmm0=(24 25 26 27 34 35 36 37)
+        movdqa  xmm3, XMMWORD [wk(3)]   ; xmm3=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movdqa     xmm5,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm5,xmm0            ; xmm5=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm3            ; xmm1=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm6,xmm3            ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm3,xmm4
+        paddw   xmm5,xmm1               ; xmm5=data3+data4=tmp3
+        paddw   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        psubw   xmm0,xmm1               ; xmm0=data3-data4=tmp4
+        psubw   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm6,xmm2
+        paddw   xmm7,xmm5               ; xmm7=tmp10
+        paddw   xmm2,xmm4               ; xmm2=tmp11
+        psubw   xmm1,xmm5               ; xmm1=tmp13
+        psubw   xmm6,xmm4               ; xmm6=tmp12
+
+        movdqa  xmm5,xmm7
+        paddw   xmm7,xmm2               ; xmm7=tmp10+tmp11
+        psubw   xmm5,xmm2               ; xmm5=tmp10-tmp11
+
+        paddw   xmm7,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        paddw   xmm5,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        psraw   xmm7,PASS1_BITS         ; xmm7=data0
+        psraw   xmm5,PASS1_BITS         ; xmm5=data4
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm7
+        movdqa  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm5
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm4,xmm1             ; xmm1=tmp13
+        movdqa    xmm2,xmm1
+        punpcklwd xmm4,xmm6             ; xmm6=tmp12
+        punpckhwd xmm2,xmm6
+        movdqa    xmm1,xmm4
+        movdqa    xmm6,xmm2
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F130_F054)]       ; xmm4=data2L
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F130_F054)]       ; xmm2=data2H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=data6L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm6=data6H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm2,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm6,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm6,DESCALE_P2
+
+        packssdw  xmm4,xmm2             ; xmm4=data2
+        packssdw  xmm1,xmm6             ; xmm1=data6
+
+        movdqa  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm1
+
+        ; -- Odd part
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        movdqa  xmm2,xmm0               ; xmm0=tmp4
+        movdqa  xmm6,xmm3               ; xmm3=tmp5
+        paddw   xmm2,xmm7               ; xmm2=z3
+        paddw   xmm6,xmm5               ; xmm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm1,xmm2
+        punpcklwd xmm4,xmm6
+        punpckhwd xmm1,xmm6
+        movdqa    xmm2,xmm4
+        movdqa    xmm6,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm4=z3L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm1=z3H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F117_F078)]       ; xmm2=z4L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F117_F078)]       ; xmm6=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm1,xmm0
+        punpcklwd xmm4,xmm5
+        punpckhwd xmm1,xmm5
+        movdqa    xmm0,xmm4
+        movdqa    xmm5,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm4=tmp4L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm1=tmp4H
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm0=tmp7L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm5=tmp7H
+
+        paddd   xmm4, XMMWORD [wk(0)]   ; xmm4=data7L
+        paddd   xmm1, XMMWORD [wk(1)]   ; xmm1=data7H
+        paddd   xmm0,xmm2               ; xmm0=data1L
+        paddd   xmm5,xmm6               ; xmm5=data1H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm0,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+
+        packssdw  xmm4,xmm1             ; xmm4=data7
+        packssdw  xmm0,xmm5             ; xmm0=data1
+
+        movdqa  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm0
+
+        movdqa    xmm1,xmm3
+        movdqa    xmm5,xmm3
+        punpcklwd xmm1,xmm7
+        punpckhwd xmm5,xmm7
+        movdqa    xmm3,xmm1
+        movdqa    xmm7,xmm5
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm1=tmp5L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm5=tmp5H
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm3=tmp6L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm7=tmp6H
+
+        paddd   xmm1,xmm2               ; xmm1=data5L
+        paddd   xmm5,xmm6               ; xmm5=data5H
+        paddd   xmm3, XMMWORD [wk(0)]   ; xmm3=data3L
+        paddd   xmm7, XMMWORD [wk(1)]   ; xmm7=data3H
+
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+        paddd   xmm3,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm1,xmm5             ; xmm1=data5
+        packssdw  xmm3,xmm7             ; xmm3=data3
+
+        movdqa  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm3
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-3dn.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,452 @@
+;
+; jidctflt.asm - floating-point IDCT (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_3dnow) PRIVATE
+
+EXTN(jconst_idct_float_3dnow):
+
+PD_1_414        times 2 dd  1.414213562373095048801689
+PD_1_847        times 2 dd  1.847759065022573512256366
+PD_1_082        times 2 dd  1.082392200292393968799446
+PD_2_613        times 2 dd  2.613125929752753055713286
+PD_RNDINT_MAGIC times 2 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_3dnow (void *dct_table, JCOEFPTR coef_block,
+;                         JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_3dnow) PRIVATE
+
+EXTN(jsimd_idct_float_3dnow):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/2                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_3DNOW
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        pushpic ebx             ; save GOT address
+        mov     ebx, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        mov     eax, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        or      ebx, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        or      ebx, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        or      eax,ebx
+        poppic  ebx             ; restore GOT address
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movd      mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm0,mm0
+        psrad     mm0,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm0,mm0
+
+        pfmul     mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0
+        punpckhdq mm1,mm1
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movd      mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movd      mm1, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movd      mm2, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movd      mm3, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm0,mm0
+        punpcklwd mm1,mm1
+        psrad     mm0,(DWORD_BIT-WORD_BIT)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm0,mm0
+        pi2fd     mm1,mm1
+
+        pfmul     mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        punpcklwd mm2,mm2
+        punpcklwd mm3,mm3
+        psrad     mm2,(DWORD_BIT-WORD_BIT)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm2,mm2
+        pi2fd     mm3,mm3
+
+        pfmul     mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pfsub   mm0,mm2                 ; mm0=tmp11
+        pfsub   mm1,mm3
+        pfadd   mm4,mm2                 ; mm4=tmp10
+        pfadd   mm5,mm3                 ; mm5=tmp13
+
+        pfmul   mm1,[GOTOFF(ebx,PD_1_414)]
+        pfsub   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm5                 ; mm4=tmp3
+        pfsub   mm0,mm1                 ; mm0=tmp2
+        pfadd   mm6,mm5                 ; mm6=tmp0
+        pfadd   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; tmp3
+        movq    MMWORD [wk(0)], mm0     ; tmp2
+
+        ; -- Odd part
+
+        movd      mm2, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movd      mm3, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movd      mm5, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movd      mm1, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm2,mm2
+        punpcklwd mm3,mm3
+        psrad     mm2,(DWORD_BIT-WORD_BIT)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm2,mm2
+        pi2fd     mm3,mm3
+
+        pfmul     mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        punpcklwd mm5,mm5
+        punpcklwd mm1,mm1
+        psrad     mm5,(DWORD_BIT-WORD_BIT)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm5,mm5
+        pi2fd     mm1,mm1
+
+        pfmul     mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        pfadd   mm2,mm1                 ; mm2=z11
+        pfadd   mm5,mm3                 ; mm5=z13
+        pfsub   mm4,mm1                 ; mm4=z12
+        pfsub   mm0,mm3                 ; mm0=z10
+
+        movq    mm1,mm2
+        pfsub   mm2,mm5
+        pfadd   mm1,mm5                 ; mm1=tmp7
+
+        pfmul   mm2,[GOTOFF(ebx,PD_1_414)]      ; mm2=tmp11
+
+        movq    mm3,mm0
+        pfadd   mm0,mm4
+        pfmul   mm0,[GOTOFF(ebx,PD_1_847)]      ; mm0=z5
+        pfmul   mm3,[GOTOFF(ebx,PD_2_613)]      ; mm3=(z10 * 2.613125930)
+        pfmul   mm4,[GOTOFF(ebx,PD_1_082)]      ; mm4=(z12 * 1.082392200)
+        pfsubr  mm3,mm0                 ; mm3=tmp12
+        pfsub   mm4,mm0                 ; mm4=tmp10
+
+        ; -- Final output stage
+
+        pfsub   mm3,mm1                 ; mm3=tmp6
+        movq    mm5,mm6
+        movq    mm0,mm7
+        pfadd   mm6,mm1                 ; mm6=data0=(00 01)
+        pfadd   mm7,mm3                 ; mm7=data1=(10 11)
+        pfsub   mm5,mm1                 ; mm5=data7=(70 71)
+        pfsub   mm0,mm3                 ; mm0=data6=(60 61)
+        pfsub   mm2,mm3                 ; mm2=tmp5
+
+        movq      mm1,mm6               ; transpose coefficients
+        punpckldq mm6,mm7               ; mm6=(00 10)
+        punpckhdq mm1,mm7               ; mm1=(01 11)
+        movq      mm3,mm0               ; transpose coefficients
+        punpckldq mm0,mm5               ; mm0=(60 70)
+        punpckhdq mm3,mm5               ; mm3=(61 71)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm3
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp2
+        movq    mm5, MMWORD [wk(1)]     ; mm5=tmp3
+
+        pfadd   mm4,mm2                 ; mm4=tmp4
+        movq    mm6,mm7
+        movq    mm1,mm5
+        pfadd   mm7,mm2                 ; mm7=data2=(20 21)
+        pfadd   mm5,mm4                 ; mm5=data4=(40 41)
+        pfsub   mm6,mm2                 ; mm6=data5=(50 51)
+        pfsub   mm1,mm4                 ; mm1=data3=(30 31)
+
+        movq      mm0,mm7               ; transpose coefficients
+        punpckldq mm7,mm1               ; mm7=(20 30)
+        punpckhdq mm0,mm1               ; mm0=(21 31)
+        movq      mm3,mm5               ; transpose coefficients
+        punpckldq mm5,mm6               ; mm5=(40 50)
+        punpckhdq mm3,mm6               ; mm3=(41 51)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm3
+
+.nextcolumn:
+        add     esi, byte 2*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 2*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/2                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pfsub   mm0,mm2                 ; mm0=tmp11
+        pfsub   mm1,mm3
+        pfadd   mm4,mm2                 ; mm4=tmp10
+        pfadd   mm5,mm3                 ; mm5=tmp13
+
+        pfmul   mm1,[GOTOFF(ebx,PD_1_414)]
+        pfsub   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm5                 ; mm4=tmp3
+        pfsub   mm0,mm1                 ; mm0=tmp2
+        pfadd   mm6,mm5                 ; mm6=tmp0
+        pfadd   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; tmp3
+        movq    MMWORD [wk(0)], mm0     ; tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        pfadd   mm2,mm1                 ; mm2=z11
+        pfadd   mm5,mm3                 ; mm5=z13
+        pfsub   mm4,mm1                 ; mm4=z12
+        pfsub   mm0,mm3                 ; mm0=z10
+
+        movq    mm1,mm2
+        pfsub   mm2,mm5
+        pfadd   mm1,mm5                 ; mm1=tmp7
+
+        pfmul   mm2,[GOTOFF(ebx,PD_1_414)]      ; mm2=tmp11
+
+        movq    mm3,mm0
+        pfadd   mm0,mm4
+        pfmul   mm0,[GOTOFF(ebx,PD_1_847)]      ; mm0=z5
+        pfmul   mm3,[GOTOFF(ebx,PD_2_613)]      ; mm3=(z10 * 2.613125930)
+        pfmul   mm4,[GOTOFF(ebx,PD_1_082)]      ; mm4=(z12 * 1.082392200)
+        pfsubr  mm3,mm0                 ; mm3=tmp12
+        pfsub   mm4,mm0                 ; mm4=tmp10
+
+        ; -- Final output stage
+
+        pfsub   mm3,mm1                 ; mm3=tmp6
+        movq    mm5,mm6
+        movq    mm0,mm7
+        pfadd   mm6,mm1                 ; mm6=data0=(00 10)
+        pfadd   mm7,mm3                 ; mm7=data1=(01 11)
+        pfsub   mm5,mm1                 ; mm5=data7=(07 17)
+        pfsub   mm0,mm3                 ; mm0=data6=(06 16)
+        pfsub   mm2,mm3                 ; mm2=tmp5
+
+        movq    mm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)]       ; mm1=[PD_RNDINT_MAGIC]
+        pcmpeqd mm3,mm3
+        psrld   mm3,WORD_BIT            ; mm3={0xFFFF 0x0000 0xFFFF 0x0000}
+
+        pfadd   mm6,mm1                 ; mm6=roundint(data0/8)=(00 ** 10 **)
+        pfadd   mm7,mm1                 ; mm7=roundint(data1/8)=(01 ** 11 **)
+        pfadd   mm0,mm1                 ; mm0=roundint(data6/8)=(06 ** 16 **)
+        pfadd   mm5,mm1                 ; mm5=roundint(data7/8)=(07 ** 17 **)
+
+        pand    mm6,mm3                 ; mm6=(00 -- 10 --)
+        pslld   mm7,WORD_BIT            ; mm7=(-- 01 -- 11)
+        pand    mm0,mm3                 ; mm0=(06 -- 16 --)
+        pslld   mm5,WORD_BIT            ; mm5=(-- 07 -- 17)
+        por     mm6,mm7                 ; mm6=(00 01 10 11)
+        por     mm0,mm5                 ; mm0=(06 07 16 17)
+
+        movq    mm1, MMWORD [wk(0)]     ; mm1=tmp2
+        movq    mm3, MMWORD [wk(1)]     ; mm3=tmp3
+
+        pfadd   mm4,mm2                 ; mm4=tmp4
+        movq    mm7,mm1
+        movq    mm5,mm3
+        pfadd   mm1,mm2                 ; mm1=data2=(02 12)
+        pfadd   mm3,mm4                 ; mm3=data4=(04 14)
+        pfsub   mm7,mm2                 ; mm7=data5=(05 15)
+        pfsub   mm5,mm4                 ; mm5=data3=(03 13)
+
+        movq    mm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)]       ; mm2=[PD_RNDINT_MAGIC]
+        pcmpeqd mm4,mm4
+        psrld   mm4,WORD_BIT            ; mm4={0xFFFF 0x0000 0xFFFF 0x0000}
+
+        pfadd   mm3,mm2                 ; mm3=roundint(data4/8)=(04 ** 14 **)
+        pfadd   mm7,mm2                 ; mm7=roundint(data5/8)=(05 ** 15 **)
+        pfadd   mm1,mm2                 ; mm1=roundint(data2/8)=(02 ** 12 **)
+        pfadd   mm5,mm2                 ; mm5=roundint(data3/8)=(03 ** 13 **)
+
+        pand    mm3,mm4                 ; mm3=(04 -- 14 --)
+        pslld   mm7,WORD_BIT            ; mm7=(-- 05 -- 15)
+        pand    mm1,mm4                 ; mm1=(02 -- 12 --)
+        pslld   mm5,WORD_BIT            ; mm5=(-- 03 -- 13)
+        por     mm3,mm7                 ; mm3=(04 05 14 15)
+        por     mm1,mm5                 ; mm1=(02 03 12 13)
+
+        movq      mm2,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm2=[PB_CENTERJSAMP]
+
+        packsswb  mm6,mm3               ; mm6=(00 01 10 11 04 05 14 15)
+        packsswb  mm1,mm0               ; mm1=(02 03 12 13 06 07 16 17)
+        paddb     mm6,mm2
+        paddb     mm1,mm2
+
+        movq      mm4,mm6               ; transpose coefficients(phase 2)
+        punpcklwd mm6,mm1               ; mm6=(00 01 02 03 10 11 12 13)
+        punpckhwd mm4,mm1               ; mm4=(04 05 06 07 14 15 16 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 3)
+        punpckldq mm6,mm4               ; mm6=(00 01 02 03 04 05 06 07)
+        punpckhdq mm7,mm4               ; mm7=(10 11 12 13 14 15 16 17)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 2*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 2*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,483 @@
+;
+; jidctflt.asm - floating-point IDCT (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse2) PRIVATE
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse2) PRIVATE
+
+EXTN(jsimd_idct_float_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [workspace]
+        collect_args
+        push    rbx
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+        lea     rdi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     rcx, DCTSIZE/4                          ; ctr
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    xmm1, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm2, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm4, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm6, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm7, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm2
+        por     xmm3,xmm4
+        por     xmm5,xmm6
+        por     xmm1,xmm3
+        por     xmm5,xmm7
+        por     xmm1,xmm5
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm2, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpcklwd xmm1,xmm1             ; xmm1=(20 20 21 21 22 22 23 23)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in2=(20 21 22 23)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in2=(20 21 22 23)
+
+        punpcklwd xmm2,xmm2             ; xmm2=(40 40 41 41 42 42 43 43)
+        punpcklwd xmm3,xmm3             ; xmm3=(60 60 61 61 62 62 63 63)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in4=(40 41 42 43)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in6=(60 61 62 63)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in4=(40 41 42 43)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in6=(60 61 62 63)
+
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[rel PD_1_414]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      xmm2, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm2,xmm2             ; xmm2=(10 10 11 11 12 12 13 13)
+        punpcklwd xmm3,xmm3             ; xmm3=(30 30 31 31 32 32 33 33)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in1=(10 11 12 13)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in3=(30 31 32 33)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in1=(10 11 12 13)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in3=(30 31 32 33)
+
+        punpcklwd xmm5,xmm5             ; xmm5=(50 50 51 51 52 52 53 53)
+        punpcklwd xmm1,xmm1             ; xmm1=(70 70 71 71 72 72 73 73)
+        psrad     xmm5,(DWORD_BIT-WORD_BIT)     ; xmm5=in5=(50 51 52 53)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in7=(70 71 72 73)
+        cvtdq2ps  xmm5,xmm5                     ; xmm5=in5=(50 51 52 53)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in7=(70 71 72 73)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[rel PD_1_414]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[rel PD_1_847]     ; xmm0=z5
+        mulps   xmm3,[rel PD_M2_613]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[rel PD_1_082]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     rsi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     rdx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     rdi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     rcx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        lea     rsi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+        mov     rcx, DCTSIZE/4                          ; ctr
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[rel PD_1_414]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[rel PD_1_414]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[rel PD_1_847]     ; xmm0=z5
+        mulps   xmm3,[rel PD_M2_613]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[rel PD_1_082]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[rel PD_RNDINT_MAGIC]      ; xmm1=[rel PD_RNDINT_MAGIC]
+        pcmpeqd xmm3,xmm3
+        psrld   xmm3,WORD_BIT           ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm6,xmm1       ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+        addps   xmm7,xmm1       ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+        addps   xmm0,xmm1       ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+        addps   xmm5,xmm1       ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+        pand    xmm6,xmm3               ; xmm6=(00 -- 10 -- 20 -- 30 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+        pand    xmm0,xmm3               ; xmm0=(06 -- 16 -- 26 -- 36 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+        por     xmm6,xmm7               ; xmm6=(00 01 10 11 20 21 30 31)
+        por     xmm0,xmm5               ; xmm0=(06 07 16 17 26 27 36 37)
+
+        movaps  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp2
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=tmp3
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm7,xmm1
+        movaps  xmm5,xmm3
+        addps   xmm1,xmm2               ; xmm1=data2=(02 12 22 32)
+        addps   xmm3,xmm4               ; xmm3=data4=(04 14 24 34)
+        subps   xmm7,xmm2               ; xmm7=data5=(05 15 25 35)
+        subps   xmm5,xmm4               ; xmm5=data3=(03 13 23 33)
+
+        movaps  xmm2,[rel PD_RNDINT_MAGIC]      ; xmm2=[rel PD_RNDINT_MAGIC]
+        pcmpeqd xmm4,xmm4
+        psrld   xmm4,WORD_BIT           ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm3,xmm2       ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+        addps   xmm7,xmm2       ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+        addps   xmm1,xmm2       ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+        addps   xmm5,xmm2       ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+        pand    xmm3,xmm4               ; xmm3=(04 -- 14 -- 24 -- 34 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+        pand    xmm1,xmm4               ; xmm1=(02 -- 12 -- 22 -- 32 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+        por     xmm3,xmm7               ; xmm3=(04 05 14 15 24 25 34 35)
+        por     xmm1,xmm5               ; xmm1=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm2,[rel PB_CENTERJSAMP]     ; xmm2=[rel PB_CENTERJSAMP]
+
+        packsswb  xmm6,xmm3     ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+        packsswb  xmm1,xmm0     ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+        paddb     xmm6,xmm2
+        paddb     xmm1,xmm2
+
+        movdqa    xmm4,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm1     ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm1     ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+        movdqa    xmm7,xmm6     ; transpose coefficients(phase 3)
+        punpckldq xmm6,xmm4     ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm7,xmm4     ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+        pshufd  xmm5,xmm6,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm3,xmm7,0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rbx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm7
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rbx, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm3
+
+        add     rsi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     rdi, byte 4*SIZEOF_JSAMPROW
+        dec     rcx                             ; ctr
+        jnz     near .rowloop
+
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,498 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse2) PRIVATE
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse2) PRIVATE
+
+EXTN(jsimd_idct_float_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    xmm1, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    xmm2, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    xmm4, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    xmm6, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        movq    xmm7, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm2
+        por     xmm3,xmm4
+        por     xmm5,xmm6
+        por     xmm1,xmm3
+        por     xmm5,xmm7
+        por     xmm1,xmm5
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq      xmm2, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpcklwd xmm1,xmm1             ; xmm1=(20 20 21 21 22 22 23 23)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in2=(20 21 22 23)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in2=(20 21 22 23)
+
+        punpcklwd xmm2,xmm2             ; xmm2=(40 40 41 41 42 42 43 43)
+        punpcklwd xmm3,xmm3             ; xmm3=(60 60 61 61 62 62 63 63)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in4=(40 41 42 43)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in6=(60 61 62 63)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in4=(40 41 42 43)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in6=(60 61 62 63)
+
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      xmm2, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq      xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm2,xmm2             ; xmm2=(10 10 11 11 12 12 13 13)
+        punpcklwd xmm3,xmm3             ; xmm3=(30 30 31 31 32 32 33 33)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in1=(10 11 12 13)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in3=(30 31 32 33)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in1=(10 11 12 13)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in3=(30 31 32 33)
+
+        punpcklwd xmm5,xmm5             ; xmm5=(50 50 51 51 52 52 53 53)
+        punpcklwd xmm1,xmm1             ; xmm1=(70 70 71 71 72 72 73 73)
+        psrad     xmm5,(DWORD_BIT-WORD_BIT)     ; xmm5=in5=(50 51 52 53)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in7=(70 71 72 73)
+        cvtdq2ps  xmm5,xmm5                     ; xmm5=in5=(50 51 52 53)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in7=(70 71 72 73)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)]      ; xmm1=[PD_RNDINT_MAGIC]
+        pcmpeqd xmm3,xmm3
+        psrld   xmm3,WORD_BIT           ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm6,xmm1       ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+        addps   xmm7,xmm1       ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+        addps   xmm0,xmm1       ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+        addps   xmm5,xmm1       ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+        pand    xmm6,xmm3               ; xmm6=(00 -- 10 -- 20 -- 30 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+        pand    xmm0,xmm3               ; xmm0=(06 -- 16 -- 26 -- 36 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+        por     xmm6,xmm7               ; xmm6=(00 01 10 11 20 21 30 31)
+        por     xmm0,xmm5               ; xmm0=(06 07 16 17 26 27 36 37)
+
+        movaps  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp2
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=tmp3
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm7,xmm1
+        movaps  xmm5,xmm3
+        addps   xmm1,xmm2               ; xmm1=data2=(02 12 22 32)
+        addps   xmm3,xmm4               ; xmm3=data4=(04 14 24 34)
+        subps   xmm7,xmm2               ; xmm7=data5=(05 15 25 35)
+        subps   xmm5,xmm4               ; xmm5=data3=(03 13 23 33)
+
+        movaps  xmm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)]      ; xmm2=[PD_RNDINT_MAGIC]
+        pcmpeqd xmm4,xmm4
+        psrld   xmm4,WORD_BIT           ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm3,xmm2       ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+        addps   xmm7,xmm2       ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+        addps   xmm1,xmm2       ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+        addps   xmm5,xmm2       ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+        pand    xmm3,xmm4               ; xmm3=(04 -- 14 -- 24 -- 34 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+        pand    xmm1,xmm4               ; xmm1=(02 -- 12 -- 22 -- 32 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+        por     xmm3,xmm7               ; xmm3=(04 05 14 15 24 25 34 35)
+        por     xmm1,xmm5               ; xmm1=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm2=[PB_CENTERJSAMP]
+
+        packsswb  xmm6,xmm3     ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+        packsswb  xmm1,xmm0     ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+        paddb     xmm6,xmm2
+        paddb     xmm1,xmm2
+
+        movdqa    xmm4,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm1     ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm1     ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+        movdqa    xmm7,xmm6     ; transpose coefficients(phase 3)
+        punpckldq xmm6,xmm4     ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm7,xmm4     ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+        pshufd  xmm5,xmm6,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm3,xmm7,0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm7
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm3
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctflt-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctflt-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,572 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse) PRIVATE
+
+EXTN(jconst_idct_float_sse):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_0_125        times 4 dd  0.125       ; 1/8
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse) PRIVATE
+
+EXTN(jsimd_idct_float_sse):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm1,mm0                       ; mm1=(** 02 ** 03)
+        punpcklwd mm0,mm0                       ; mm0=(00 00 01 01)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in0H=(02 03)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in0L=(00 01)
+        cvtpi2ps  xmm3,mm1                      ; xmm3=(02 03 ** **)
+        cvtpi2ps  xmm0,mm0                      ; xmm0=(00 01 ** **)
+        movlhps   xmm0,xmm3                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq      mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq      mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq      mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm4,mm0                       ; mm4=(** 02 ** 03)
+        punpcklwd mm0,mm0                       ; mm0=(00 00 01 01)
+        punpckhwd mm5,mm1                       ; mm5=(** 22 ** 23)
+        punpcklwd mm1,mm1                       ; mm1=(20 20 21 21)
+
+        psrad     mm4,(DWORD_BIT-WORD_BIT)      ; mm4=in0H=(02 03)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in0L=(00 01)
+        cvtpi2ps  xmm4,mm4                      ; xmm4=(02 03 ** **)
+        cvtpi2ps  xmm0,mm0                      ; xmm0=(00 01 ** **)
+        psrad     mm5,(DWORD_BIT-WORD_BIT)      ; mm5=in2H=(22 23)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in2L=(20 21)
+        cvtpi2ps  xmm5,mm5                      ; xmm5=(22 23 ** **)
+        cvtpi2ps  xmm1,mm1                      ; xmm1=(20 21 ** **)
+
+        punpckhwd mm6,mm2                       ; mm6=(** 42 ** 43)
+        punpcklwd mm2,mm2                       ; mm2=(40 40 41 41)
+        punpckhwd mm7,mm3                       ; mm7=(** 62 ** 63)
+        punpcklwd mm3,mm3                       ; mm3=(60 60 61 61)
+
+        psrad     mm6,(DWORD_BIT-WORD_BIT)      ; mm6=in4H=(42 43)
+        psrad     mm2,(DWORD_BIT-WORD_BIT)      ; mm2=in4L=(40 41)
+        cvtpi2ps  xmm6,mm6                      ; xmm6=(42 43 ** **)
+        cvtpi2ps  xmm2,mm2                      ; xmm2=(40 41 ** **)
+        psrad     mm7,(DWORD_BIT-WORD_BIT)      ; mm7=in6H=(62 63)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)      ; mm3=in6L=(60 61)
+        cvtpi2ps  xmm7,mm7                      ; xmm7=(62 63 ** **)
+        cvtpi2ps  xmm3,mm3                      ; xmm3=(60 61 ** **)
+
+        movlhps   xmm0,xmm4                     ; xmm0=in0=(00 01 02 03)
+        movlhps   xmm1,xmm5                     ; xmm1=in2=(20 21 22 23)
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movlhps   xmm2,xmm6                     ; xmm2=in4=(40 41 42 43)
+        movlhps   xmm3,xmm7                     ; xmm3=in6=(60 61 62 63)
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq      mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq      mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq      mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm6,mm4                       ; mm6=(** 12 ** 13)
+        punpcklwd mm4,mm4                       ; mm4=(10 10 11 11)
+        punpckhwd mm2,mm0                       ; mm2=(** 32 ** 33)
+        punpcklwd mm0,mm0                       ; mm0=(30 30 31 31)
+
+        psrad     mm6,(DWORD_BIT-WORD_BIT)      ; mm6=in1H=(12 13)
+        psrad     mm4,(DWORD_BIT-WORD_BIT)      ; mm4=in1L=(10 11)
+        cvtpi2ps  xmm4,mm6                      ; xmm4=(12 13 ** **)
+        cvtpi2ps  xmm2,mm4                      ; xmm2=(10 11 ** **)
+        psrad     mm2,(DWORD_BIT-WORD_BIT)      ; mm2=in3H=(32 33)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in3L=(30 31)
+        cvtpi2ps  xmm0,mm2                      ; xmm0=(32 33 ** **)
+        cvtpi2ps  xmm3,mm0                      ; xmm3=(30 31 ** **)
+
+        punpckhwd mm7,mm5                       ; mm7=(** 52 ** 53)
+        punpcklwd mm5,mm5                       ; mm5=(50 50 51 51)
+        punpckhwd mm3,mm1                       ; mm3=(** 72 ** 73)
+        punpcklwd mm1,mm1                       ; mm1=(70 70 71 71)
+
+        movlhps   xmm2,xmm4                     ; xmm2=in1=(10 11 12 13)
+        movlhps   xmm3,xmm0                     ; xmm3=in3=(30 31 32 33)
+
+        psrad     mm7,(DWORD_BIT-WORD_BIT)      ; mm7=in5H=(52 53)
+        psrad     mm5,(DWORD_BIT-WORD_BIT)      ; mm5=in5L=(50 51)
+        cvtpi2ps  xmm4,mm7                      ; xmm4=(52 53 ** **)
+        cvtpi2ps  xmm5,mm5                      ; xmm5=(50 51 ** **)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)      ; mm3=in7H=(72 73)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in7L=(70 71)
+        cvtpi2ps  xmm0,mm3                      ; xmm0=(72 73 ** **)
+        cvtpi2ps  xmm1,mm1                      ; xmm1=(70 71 ** **)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movlhps   xmm5,xmm4                     ; xmm5=in5=(50 51 52 53)
+        movlhps   xmm1,xmm0                     ; xmm1=in7=(70 71 72 73)
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[GOTOFF(ebx,PD_0_125)]     ; xmm1=[PD_0_125]
+
+        mulps   xmm6,xmm1               ; descale(1/8)
+        mulps   xmm7,xmm1               ; descale(1/8)
+        mulps   xmm5,xmm1               ; descale(1/8)
+        mulps   xmm0,xmm1               ; descale(1/8)
+
+        movhlps   xmm3,xmm6
+        movhlps   xmm1,xmm7
+        cvtps2pi  mm0,xmm6              ; round to int32, mm0=data0L=(00 10)
+        cvtps2pi  mm1,xmm7              ; round to int32, mm1=data1L=(01 11)
+        cvtps2pi  mm2,xmm3              ; round to int32, mm2=data0H=(20 30)
+        cvtps2pi  mm3,xmm1              ; round to int32, mm3=data1H=(21 31)
+        packssdw  mm0,mm2               ; mm0=data0=(00 10 20 30)
+        packssdw  mm1,mm3               ; mm1=data1=(01 11 21 31)
+
+        movhlps   xmm6,xmm5
+        movhlps   xmm7,xmm0
+        cvtps2pi  mm4,xmm5              ; round to int32, mm4=data7L=(07 17)
+        cvtps2pi  mm5,xmm0              ; round to int32, mm5=data6L=(06 16)
+        cvtps2pi  mm6,xmm6              ; round to int32, mm6=data7H=(27 37)
+        cvtps2pi  mm7,xmm7              ; round to int32, mm7=data6H=(26 36)
+        packssdw  mm4,mm6               ; mm4=data7=(07 17 27 37)
+        packssdw  mm5,mm7               ; mm5=data6=(06 16 26 36)
+
+        packsswb  mm0,mm5               ; mm0=(00 10 20 30 06 16 26 36)
+        packsswb  mm1,mm4               ; mm1=(01 11 21 31 07 17 27 37)
+
+        movaps  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp2
+        movaps  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movaps  xmm6,[GOTOFF(ebx,PD_0_125)]     ; xmm6=[PD_0_125]
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm5,xmm3
+        movaps  xmm0,xmm1
+        addps   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+        addps   xmm1,xmm4               ; xmm1=data4=(04 14 24 34)
+        subps   xmm5,xmm2               ; xmm5=data5=(05 15 25 35)
+        subps   xmm0,xmm4               ; xmm0=data3=(03 13 23 33)
+
+        mulps   xmm3,xmm6               ; descale(1/8)
+        mulps   xmm1,xmm6               ; descale(1/8)
+        mulps   xmm5,xmm6               ; descale(1/8)
+        mulps   xmm0,xmm6               ; descale(1/8)
+
+        movhlps   xmm7,xmm3
+        movhlps   xmm2,xmm1
+        cvtps2pi  mm2,xmm3              ; round to int32, mm2=data2L=(02 12)
+        cvtps2pi  mm3,xmm1              ; round to int32, mm3=data4L=(04 14)
+        cvtps2pi  mm6,xmm7              ; round to int32, mm6=data2H=(22 32)
+        cvtps2pi  mm7,xmm2              ; round to int32, mm7=data4H=(24 34)
+        packssdw  mm2,mm6               ; mm2=data2=(02 12 22 32)
+        packssdw  mm3,mm7               ; mm3=data4=(04 14 24 34)
+
+        movhlps   xmm4,xmm5
+        movhlps   xmm6,xmm0
+        cvtps2pi  mm5,xmm5              ; round to int32, mm5=data5L=(05 15)
+        cvtps2pi  mm4,xmm0              ; round to int32, mm4=data3L=(03 13)
+        cvtps2pi  mm6,xmm4              ; round to int32, mm6=data5H=(25 35)
+        cvtps2pi  mm7,xmm6              ; round to int32, mm7=data3H=(23 33)
+        packssdw  mm5,mm6               ; mm5=data5=(05 15 25 35)
+        packssdw  mm4,mm7               ; mm4=data3=(03 13 23 33)
+
+        movq      mm6,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm6=[PB_CENTERJSAMP]
+
+        packsswb  mm2,mm3               ; mm2=(02 12 22 32 04 14 24 34)
+        packsswb  mm4,mm5               ; mm4=(03 13 23 33 05 15 25 35)
+
+        paddb     mm0,mm6
+        paddb     mm1,mm6
+        paddb     mm2,mm6
+        paddb     mm4,mm6
+
+        movq      mm7,mm0               ; transpose coefficients(phase 1)
+        punpcklbw mm0,mm1               ; mm0=(00 01 10 11 20 21 30 31)
+        punpckhbw mm7,mm1               ; mm7=(06 07 16 17 26 27 36 37)
+        movq      mm3,mm2               ; transpose coefficients(phase 1)
+        punpcklbw mm2,mm4               ; mm2=(02 03 12 13 22 23 32 33)
+        punpckhbw mm3,mm4               ; mm3=(04 05 14 15 24 25 34 35)
+
+        movq      mm5,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm2               ; mm0=(00 01 02 03 10 11 12 13)
+        punpckhwd mm5,mm2               ; mm5=(20 21 22 23 30 31 32 33)
+        movq      mm6,mm3               ; transpose coefficients(phase 2)
+        punpcklwd mm3,mm7               ; mm3=(04 05 06 07 14 15 16 17)
+        punpckhwd mm6,mm7               ; mm6=(24 25 26 27 34 35 36 37)
+
+        movq      mm1,mm0               ; transpose coefficients(phase 3)
+        punpckldq mm0,mm3               ; mm0=(00 01 02 03 04 05 06 07)
+        punpckhdq mm1,mm3               ; mm1=(10 11 12 13 14 15 16 17)
+        movq      mm4,mm5               ; transpose coefficients(phase 3)
+        punpckldq mm5,mm6               ; mm5=(20 21 22 23 24 25 26 27)
+        punpckhdq mm4,mm6               ; mm4=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,257 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER INVERSE DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_1_082 277              /* FIX(1.082392200) */
+#define F_1_414 362              /* FIX(1.414213562) */
+#define F_1_847 473              /* FIX(1.847759065) */
+#define F_2_613 669              /* FIX(2.613125930) */
+#define F_1_613 (F_2_613 - 256)  /* FIX(2.613125930) - FIX(1) */
+
+#define CONST_BITS 8
+#define PASS1_BITS 2
+#define PRE_MULTIPLY_SCALE_BITS 2
+#define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_IDCT(in)  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(in##0, in##4);  \
+  tmp11 = vec_sub(in##0, in##4);  \
+  tmp13 = vec_add(in##2, in##6);  \
+  \
+  tmp12 = vec_sub(in##2, in##6);  \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits);  \
+  tmp12 = vec_madds(tmp12, pw_F1414, pw_zero);  \
+  tmp12 = vec_sub(tmp12, tmp13);  \
+  \
+  tmp0 = vec_add(tmp10, tmp13);  \
+  tmp3 = vec_sub(tmp10, tmp13);  \
+  tmp1 = vec_add(tmp11, tmp12);  \
+  tmp2 = vec_sub(tmp11, tmp12);  \
+  \
+  /* Odd part */  \
+  \
+  z13 = vec_add(in##5, in##3);  \
+  z10 = vec_sub(in##5, in##3);  \
+  z10s = vec_sl(z10, pre_multiply_scale_bits);  \
+  z11 = vec_add(in##1, in##7);  \
+  z12s = vec_sub(in##1, in##7);  \
+  z12s = vec_sl(z12s, pre_multiply_scale_bits);  \
+  \
+  tmp11 = vec_sub(z11, z13);  \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits);  \
+  tmp11 = vec_madds(tmp11, pw_F1414, pw_zero);  \
+  \
+  tmp7 = vec_add(z11, z13);  \
+  \
+  /* To avoid overflow...  \
+   *  \
+   * (Original)  \
+   * tmp12 = -2.613125930 * z10 + z5;  \
+   *  \
+   * (This implementation)  \
+   * tmp12 = (-1.613125930 - 1) * z10 + z5;  \
+   *       = -1.613125930 * z10 - z10 + z5;  \
+   */  \
+  \
+  z5 = vec_add(z10s, z12s);  \
+  z5 = vec_madds(z5, pw_F1847, pw_zero);  \
+  \
+  tmp10 = vec_madds(z12s, pw_F1082, pw_zero);  \
+  tmp10 = vec_sub(tmp10, z5);  \
+  tmp12 = vec_madds(z10s, pw_MF1613, z5);  \
+  tmp12 = vec_sub(tmp12, z10);  \
+  \
+  tmp6 = vec_sub(tmp12, tmp7);  \
+  tmp5 = vec_sub(tmp11, tmp6);  \
+  tmp4 = vec_add(tmp10, tmp5);  \
+  \
+  out0 = vec_add(tmp0, tmp7);  \
+  out1 = vec_add(tmp1, tmp6);  \
+  out2 = vec_add(tmp2, tmp5);  \
+  out3 = vec_sub(tmp3, tmp4);  \
+  out4 = vec_add(tmp3, tmp4);  \
+  out5 = vec_sub(tmp2, tmp5);  \
+  out6 = vec_sub(tmp1, tmp6);  \
+  out7 = vec_sub(tmp0, tmp7);  \
+}
+
+
+void
+jsimd_idct_ifast_altivec (void *dct_table_, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z5, z10, z10s, z11, z12s, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_F1414 = { __8X(F_1_414 << CONST_SHIFT) },
+    pw_F1847 = { __8X(F_1_847 << CONST_SHIFT) },
+    pw_MF1613 = { __8X(-F_1_613 << CONST_SHIFT) },
+    pw_F1082 = { __8X(F_1_082 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) },
+    pass1_bits3 = { __8X(PASS1_BITS + 3) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row);
+
+  out0 = vec_sra(out0, pass1_bits3);
+  out1 = vec_sra(out1, pass1_bits3);
+  out2 = vec_sra(out2, pass1_bits3);
+  out3 = vec_sra(out3, pass1_bits3);
+  out4 = vec_sra(out4, pass1_bits3);
+  out5 = vec_sra(out5, pass1_bits3);
+  out6 = vec_sra(out6, pass1_bits3);
+  out7 = vec_sra(out7, pass1_bits3);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[0] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[1] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[2] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[3] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[4] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[5] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[6] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[7] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,500 @@
+;
+; jidctfst.asm - fast integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_mmx) PRIVATE
+
+EXTN(jconst_idct_ifast_mmx):
+
+PW_F1414        times 4 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 4 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 4 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 4 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_mmx (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_mmx) PRIVATE
+
+EXTN(jsimd_idct_ifast_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        psubw   mm0,mm2                 ; mm0=tmp11
+        psubw   mm1,mm3
+        paddw   mm4,mm2                 ; mm4=tmp10
+        paddw   mm5,mm3                 ; mm5=tmp13
+
+        psllw   mm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        psubw   mm4,mm5                 ; mm4=tmp3
+        psubw   mm0,mm1                 ; mm0=tmp2
+        paddw   mm6,mm5                 ; mm6=tmp0
+        paddw   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=tmp3
+        movq    MMWORD [wk(0)], mm0     ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        psubw   mm2,mm1                 ; mm2=z12
+        psubw   mm5,mm3                 ; mm5=z10
+        paddw   mm4,mm1                 ; mm4=z11
+        paddw   mm0,mm3                 ; mm0=z13
+
+        movq    mm1,mm5                 ; mm1=z10(unscaled)
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm5,PRE_MULTIPLY_SCALE_BITS
+
+        movq    mm3,mm4
+        psubw   mm4,mm0
+        paddw   mm3,mm0                 ; mm3=tmp7
+
+        psllw   mm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm4,[GOTOFF(ebx,PW_F1414)]      ; mm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movq    mm0,mm5
+        paddw   mm5,mm2
+        pmulhw  mm5,[GOTOFF(ebx,PW_F1847)]      ; mm5=z5
+        pmulhw  mm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  mm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   mm0,mm1
+        psubw   mm2,mm5                 ; mm2=tmp10
+        paddw   mm0,mm5                 ; mm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   mm0,mm3                 ; mm0=tmp6
+        movq    mm1,mm6
+        movq    mm5,mm7
+        paddw   mm6,mm3                 ; mm6=data0=(00 01 02 03)
+        paddw   mm7,mm0                 ; mm7=data1=(10 11 12 13)
+        psubw   mm1,mm3                 ; mm1=data7=(70 71 72 73)
+        psubw   mm5,mm0                 ; mm5=data6=(60 61 62 63)
+        psubw   mm4,mm0                 ; mm4=tmp5
+
+        movq      mm3,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm3,mm7               ; mm3=(02 12 03 13)
+        movq      mm0,mm5               ; transpose coefficients(phase 1)
+        punpcklwd mm5,mm1               ; mm5=(60 70 61 71)
+        punpckhwd mm0,mm1               ; mm0=(62 72 63 73)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp2
+        movq    mm1, MMWORD [wk(1)]     ; mm1=tmp3
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(60 70 61 71)
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=(62 72 63 73)
+
+        paddw   mm2,mm4                 ; mm2=tmp4
+        movq    mm5,mm7
+        movq    mm0,mm1
+        paddw   mm7,mm4                 ; mm7=data2=(20 21 22 23)
+        paddw   mm1,mm2                 ; mm1=data4=(40 41 42 43)
+        psubw   mm5,mm4                 ; mm5=data5=(50 51 52 53)
+        psubw   mm0,mm2                 ; mm0=data3=(30 31 32 33)
+
+        movq      mm4,mm7               ; transpose coefficients(phase 1)
+        punpcklwd mm7,mm0               ; mm7=(20 30 21 31)
+        punpckhwd mm4,mm0               ; mm4=(22 32 23 33)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm5               ; mm1=(40 50 41 51)
+        punpckhwd mm2,mm5               ; mm2=(42 52 43 53)
+
+        movq      mm0,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm7               ; mm6=(00 10 20 30)
+        punpckhdq mm0,mm7               ; mm0=(01 11 21 31)
+        movq      mm5,mm3               ; transpose coefficients(phase 2)
+        punpckldq mm3,mm4               ; mm3=(02 12 22 32)
+        punpckhdq mm5,mm4               ; mm5=(03 13 23 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=(60 70 61 71)
+        movq    mm4, MMWORD [wk(1)]     ; mm4=(62 72 63 73)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5
+
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm7               ; mm1=(40 50 60 70)
+        punpckhdq mm6,mm7               ; mm6=(41 51 61 71)
+        movq      mm0,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm4               ; mm2=(42 52 62 72)
+        punpckhdq mm0,mm4               ; mm0=(43 53 63 73)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm0
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_IFAST_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        psubw   mm0,mm2                 ; mm0=tmp11
+        psubw   mm1,mm3
+        paddw   mm4,mm2                 ; mm4=tmp10
+        paddw   mm5,mm3                 ; mm5=tmp13
+
+        psllw   mm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        psubw   mm4,mm5                 ; mm4=tmp3
+        psubw   mm0,mm1                 ; mm0=tmp2
+        paddw   mm6,mm5                 ; mm6=tmp0
+        paddw   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=tmp3
+        movq    MMWORD [wk(0)], mm0     ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        psubw   mm2,mm1                 ; mm2=z12
+        psubw   mm5,mm3                 ; mm5=z10
+        paddw   mm4,mm1                 ; mm4=z11
+        paddw   mm0,mm3                 ; mm0=z13
+
+        movq    mm1,mm5                 ; mm1=z10(unscaled)
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm5,PRE_MULTIPLY_SCALE_BITS
+
+        movq    mm3,mm4
+        psubw   mm4,mm0
+        paddw   mm3,mm0                 ; mm3=tmp7
+
+        psllw   mm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm4,[GOTOFF(ebx,PW_F1414)]      ; mm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movq    mm0,mm5
+        paddw   mm5,mm2
+        pmulhw  mm5,[GOTOFF(ebx,PW_F1847)]      ; mm5=z5
+        pmulhw  mm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  mm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   mm0,mm1
+        psubw   mm2,mm5                 ; mm2=tmp10
+        paddw   mm0,mm5                 ; mm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   mm0,mm3                 ; mm0=tmp6
+        movq    mm1,mm6
+        movq    mm5,mm7
+        paddw   mm6,mm3                 ; mm6=data0=(00 10 20 30)
+        paddw   mm7,mm0                 ; mm7=data1=(01 11 21 31)
+        psraw   mm6,(PASS1_BITS+3)      ; descale
+        psraw   mm7,(PASS1_BITS+3)      ; descale
+        psubw   mm1,mm3                 ; mm1=data7=(07 17 27 37)
+        psubw   mm5,mm0                 ; mm5=data6=(06 16 26 36)
+        psraw   mm1,(PASS1_BITS+3)      ; descale
+        psraw   mm5,(PASS1_BITS+3)      ; descale
+        psubw   mm4,mm0                 ; mm4=tmp5
+
+        packsswb  mm6,mm5               ; mm6=(00 10 20 30 06 16 26 36)
+        packsswb  mm7,mm1               ; mm7=(01 11 21 31 07 17 27 37)
+
+        movq    mm3, MMWORD [wk(0)]     ; mm3=tmp2
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp3
+
+        paddw   mm2,mm4                 ; mm2=tmp4
+        movq    mm5,mm3
+        movq    mm1,mm0
+        paddw   mm3,mm4                 ; mm3=data2=(02 12 22 32)
+        paddw   mm0,mm2                 ; mm0=data4=(04 14 24 34)
+        psraw   mm3,(PASS1_BITS+3)      ; descale
+        psraw   mm0,(PASS1_BITS+3)      ; descale
+        psubw   mm5,mm4                 ; mm5=data5=(05 15 25 35)
+        psubw   mm1,mm2                 ; mm1=data3=(03 13 23 33)
+        psraw   mm5,(PASS1_BITS+3)      ; descale
+        psraw   mm1,(PASS1_BITS+3)      ; descale
+
+        movq      mm4,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm4=[PB_CENTERJSAMP]
+
+        packsswb  mm3,mm0               ; mm3=(02 12 22 32 04 14 24 34)
+        packsswb  mm1,mm5               ; mm1=(03 13 23 33 05 15 25 35)
+
+        paddb     mm6,mm4
+        paddb     mm7,mm4
+        paddb     mm3,mm4
+        paddb     mm1,mm4
+
+        movq      mm2,mm6               ; transpose coefficients(phase 1)
+        punpcklbw mm6,mm7               ; mm6=(00 01 10 11 20 21 30 31)
+        punpckhbw mm2,mm7               ; mm2=(06 07 16 17 26 27 36 37)
+        movq      mm0,mm3               ; transpose coefficients(phase 1)
+        punpcklbw mm3,mm1               ; mm3=(02 03 12 13 22 23 32 33)
+        punpckhbw mm0,mm1               ; mm0=(04 05 14 15 24 25 34 35)
+
+        movq      mm5,mm6               ; transpose coefficients(phase 2)
+        punpcklwd mm6,mm3               ; mm6=(00 01 02 03 10 11 12 13)
+        punpckhwd mm5,mm3               ; mm5=(20 21 22 23 30 31 32 33)
+        movq      mm4,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm2               ; mm0=(04 05 06 07 14 15 16 17)
+        punpckhwd mm4,mm2               ; mm4=(24 25 26 27 34 35 36 37)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 3)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03 04 05 06 07)
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13 14 15 16 17)
+        movq      mm1,mm5               ; transpose coefficients(phase 3)
+        punpckldq mm5,mm4               ; mm5=(20 21 22 23 24 25 26 27)
+        punpckhdq mm1,mm4               ; mm1=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_JCOEF        ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,492 @@
+;
+; jidctfst.asm - fast integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_sse2) PRIVATE
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414        times 8 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 8 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 8 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 8 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_idct_ifast_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm7,xmm0             ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm7,xmm7             ; xmm7=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm6,xmm0,0x00          ; xmm6=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm2,xmm0,0x55          ; xmm2=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm5,xmm0,0xAA          ; xmm5=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm0,xmm0,0xFF          ; xmm0=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm1,xmm7,0x00          ; xmm1=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm4,xmm7,0x55          ; xmm4=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm3,xmm7,0xAA          ; xmm3=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm7,xmm7,0xFF          ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=col3
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        psubw   xmm0,xmm2               ; xmm0=tmp11
+        psubw   xmm1,xmm3
+        paddw   xmm4,xmm2               ; xmm4=tmp10
+        paddw   xmm5,xmm3               ; xmm5=tmp13
+
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm1,[rel PW_F1414]
+        psubw   xmm1,xmm5               ; xmm1=tmp12
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm7,xmm0
+        psubw   xmm4,xmm5               ; xmm4=tmp3
+        psubw   xmm0,xmm1               ; xmm0=tmp2
+        paddw   xmm6,xmm5               ; xmm6=tmp0
+        paddw   xmm7,xmm1               ; xmm7=tmp1
+
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=tmp3
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movdqa  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm0,xmm5
+        psubw   xmm2,xmm1               ; xmm2=z12
+        psubw   xmm5,xmm3               ; xmm5=z10
+        paddw   xmm4,xmm1               ; xmm4=z11
+        paddw   xmm0,xmm3               ; xmm0=z13
+
+        movdqa  xmm1,xmm5               ; xmm1=z10(unscaled)
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm3,xmm4
+        psubw   xmm4,xmm0
+        paddw   xmm3,xmm0               ; xmm3=tmp7
+
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm4,[rel PW_F1414]     ; xmm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm0,xmm5
+        paddw   xmm5,xmm2
+        pmulhw  xmm5,[rel PW_F1847]     ; xmm5=z5
+        pmulhw  xmm0,[rel PW_MF1613]
+        pmulhw  xmm2,[rel PW_F1082]
+        psubw   xmm0,xmm1
+        psubw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm0,xmm5               ; xmm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm0,xmm3               ; xmm0=tmp6
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm7
+        paddw   xmm6,xmm3               ; xmm6=data0=(00 01 02 03 04 05 06 07)
+        paddw   xmm7,xmm0               ; xmm7=data1=(10 11 12 13 14 15 16 17)
+        psubw   xmm1,xmm3               ; xmm1=data7=(70 71 72 73 74 75 76 77)
+        psubw   xmm5,xmm0               ; xmm5=data6=(60 61 62 63 64 65 66 67)
+        psubw   xmm4,xmm0               ; xmm4=tmp5
+
+        movdqa    xmm3,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm3,xmm7             ; xmm3=(04 14 05 15 06 16 07 17)
+        movdqa    xmm0,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm1             ; xmm5=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm0,xmm1             ; xmm0=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+        paddw   xmm2,xmm4               ; xmm2=tmp4
+        movdqa  xmm5,xmm7
+        movdqa  xmm0,xmm1
+        paddw   xmm7,xmm4               ; xmm7=data2=(20 21 22 23 24 25 26 27)
+        paddw   xmm1,xmm2               ; xmm1=data4=(40 41 42 43 44 45 46 47)
+        psubw   xmm5,xmm4               ; xmm5=data5=(50 51 52 53 54 55 56 57)
+        psubw   xmm0,xmm2               ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm0             ; xmm7=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm0             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm5             ; xmm2=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm0,xmm3             ; transpose coefficients(phase 2)
+        punpckldq xmm3,xmm4             ; xmm3=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm0,xmm4             ; xmm0=(06 16 26 36 07 17 27 37)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7             ; xmm6=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm5,xmm7             ; xmm5=(02 12 22 32 03 13 23 33)
+
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(60 70 61 71 62 72 63 73)
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm3,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm4             ; xmm1=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm3,xmm4             ; xmm3=(42 52 62 72 43 53 63 73)
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm7             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm0,xmm7             ; xmm0=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm4,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm1            ; xmm6=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm4,xmm1            ; xmm4=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm7,xmm5            ; transpose coefficients(phase 3)
+        punpcklqdq xmm5,xmm3            ; xmm5=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm7,xmm3            ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(04 14 24 34 05 15 25 35)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=col3
+
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm4,xmm2            ; xmm4=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm7,xmm3            ; transpose coefficients(phase 3)
+        punpcklqdq xmm3,xmm0            ; xmm3=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm7,xmm0            ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm0,xmm5
+        psubw   xmm6,xmm1               ; xmm6=tmp11
+        psubw   xmm5,xmm3
+        paddw   xmm2,xmm1               ; xmm2=tmp10
+        paddw   xmm0,xmm3               ; xmm0=tmp13
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F1414]
+        psubw   xmm5,xmm0               ; xmm5=tmp12
+
+        movdqa  xmm1,xmm2
+        movdqa  xmm3,xmm6
+        psubw   xmm2,xmm0               ; xmm2=tmp3
+        psubw   xmm6,xmm5               ; xmm6=tmp2
+        paddw   xmm1,xmm0               ; xmm1=tmp0
+        paddw   xmm3,xmm5               ; xmm3=tmp1
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=col1
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=col3
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp3
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp2
+
+        ; -- Odd part
+
+        ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+        movdqa  xmm2,xmm0
+        movdqa  xmm6,xmm4
+        psubw   xmm0,xmm7               ; xmm0=z12
+        psubw   xmm4,xmm5               ; xmm4=z10
+        paddw   xmm2,xmm7               ; xmm2=z11
+        paddw   xmm6,xmm5               ; xmm6=z13
+
+        movdqa  xmm7,xmm4               ; xmm7=z10(unscaled)
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm5,xmm2
+        psubw   xmm2,xmm6
+        paddw   xmm5,xmm6               ; xmm5=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm2,[rel PW_F1414]     ; xmm2=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm6,xmm4
+        paddw   xmm4,xmm0
+        pmulhw  xmm4,[rel PW_F1847]     ; xmm4=z5
+        pmulhw  xmm6,[rel PW_MF1613]
+        pmulhw  xmm0,[rel PW_F1082]
+        psubw   xmm6,xmm7
+        psubw   xmm0,xmm4               ; xmm0=tmp10
+        paddw   xmm6,xmm4               ; xmm6=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm6,xmm5               ; xmm6=tmp6
+        movdqa  xmm7,xmm1
+        movdqa  xmm4,xmm3
+        paddw   xmm1,xmm5               ; xmm1=data0=(00 10 20 30 40 50 60 70)
+        paddw   xmm3,xmm6               ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        psraw   xmm1,(PASS1_BITS+3)     ; descale
+        psraw   xmm3,(PASS1_BITS+3)     ; descale
+        psubw   xmm7,xmm5               ; xmm7=data7=(07 17 27 37 47 57 67 77)
+        psubw   xmm4,xmm6               ; xmm4=data6=(06 16 26 36 46 56 66 76)
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psubw   xmm2,xmm6               ; xmm2=tmp5
+
+        packsswb  xmm1,xmm4     ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm7     ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp2
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp3
+
+        paddw   xmm0,xmm2               ; xmm0=tmp4
+        movdqa  xmm4,xmm5
+        movdqa  xmm7,xmm6
+        paddw   xmm5,xmm2               ; xmm5=data2=(02 12 22 32 42 52 62 72)
+        paddw   xmm6,xmm0               ; xmm6=data4=(04 14 24 34 44 54 64 74)
+        psraw   xmm5,(PASS1_BITS+3)     ; descale
+        psraw   xmm6,(PASS1_BITS+3)     ; descale
+        psubw   xmm4,xmm2               ; xmm4=data5=(05 15 25 35 45 55 65 75)
+        psubw   xmm7,xmm0               ; xmm7=data3=(03 13 23 33 43 53 63 73)
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+
+        movdqa    xmm2,[rel PB_CENTERJSAMP]     ; xmm2=[rel PB_CENTERJSAMP]
+
+        packsswb  xmm5,xmm6     ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm7,xmm4     ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm1,xmm2
+        paddb     xmm3,xmm2
+        paddb     xmm5,xmm2
+        paddb     xmm7,xmm2
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 1)
+        punpcklbw xmm1,xmm3     ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm3     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm6,xmm5     ; transpose coefficients(phase 1)
+        punpcklbw xmm5,xmm7     ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm6,xmm7     ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm1     ; transpose coefficients(phase 2)
+        punpcklwd xmm1,xmm5     ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm5     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm0     ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm2,xmm0     ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm3,xmm1     ; transpose coefficients(phase 3)
+        punpckldq xmm1,xmm6     ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm3,xmm6     ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm7,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm2     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm7,xmm2     ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm5,xmm1,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm3,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm6,xmm4,0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm2,xmm7,0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+        mov     rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm7
+
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0
+        mov     rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctfst-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,502 @@
+;
+; jidctfst.asm - fast integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_sse2) PRIVATE
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414        times 8 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 8 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 8 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 8 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_idct_ifast_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm7,xmm0             ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm7,xmm7             ; xmm7=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm6,xmm0,0x00          ; xmm6=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm2,xmm0,0x55          ; xmm2=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm5,xmm0,0xAA          ; xmm5=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm0,xmm0,0xFF          ; xmm0=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm1,xmm7,0x00          ; xmm1=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm4,xmm7,0x55          ; xmm4=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm3,xmm7,0xAA          ; xmm3=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm7,xmm7,0xFF          ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=col3
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        psubw   xmm0,xmm2               ; xmm0=tmp11
+        psubw   xmm1,xmm3
+        paddw   xmm4,xmm2               ; xmm4=tmp10
+        paddw   xmm5,xmm3               ; xmm5=tmp13
+
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   xmm1,xmm5               ; xmm1=tmp12
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm7,xmm0
+        psubw   xmm4,xmm5               ; xmm4=tmp3
+        psubw   xmm0,xmm1               ; xmm0=tmp2
+        paddw   xmm6,xmm5               ; xmm6=tmp0
+        paddw   xmm7,xmm1               ; xmm7=tmp1
+
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=tmp3
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movdqa  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm0,xmm5
+        psubw   xmm2,xmm1               ; xmm2=z12
+        psubw   xmm5,xmm3               ; xmm5=z10
+        paddw   xmm4,xmm1               ; xmm4=z11
+        paddw   xmm0,xmm3               ; xmm0=z13
+
+        movdqa  xmm1,xmm5               ; xmm1=z10(unscaled)
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm3,xmm4
+        psubw   xmm4,xmm0
+        paddw   xmm3,xmm0               ; xmm3=tmp7
+
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F1414)]     ; xmm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm0,xmm5
+        paddw   xmm5,xmm2
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F1847)]     ; xmm5=z5
+        pmulhw  xmm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   xmm0,xmm1
+        psubw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm0,xmm5               ; xmm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm0,xmm3               ; xmm0=tmp6
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm7
+        paddw   xmm6,xmm3               ; xmm6=data0=(00 01 02 03 04 05 06 07)
+        paddw   xmm7,xmm0               ; xmm7=data1=(10 11 12 13 14 15 16 17)
+        psubw   xmm1,xmm3               ; xmm1=data7=(70 71 72 73 74 75 76 77)
+        psubw   xmm5,xmm0               ; xmm5=data6=(60 61 62 63 64 65 66 67)
+        psubw   xmm4,xmm0               ; xmm4=tmp5
+
+        movdqa    xmm3,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm3,xmm7             ; xmm3=(04 14 05 15 06 16 07 17)
+        movdqa    xmm0,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm1             ; xmm5=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm0,xmm1             ; xmm0=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+        paddw   xmm2,xmm4               ; xmm2=tmp4
+        movdqa  xmm5,xmm7
+        movdqa  xmm0,xmm1
+        paddw   xmm7,xmm4               ; xmm7=data2=(20 21 22 23 24 25 26 27)
+        paddw   xmm1,xmm2               ; xmm1=data4=(40 41 42 43 44 45 46 47)
+        psubw   xmm5,xmm4               ; xmm5=data5=(50 51 52 53 54 55 56 57)
+        psubw   xmm0,xmm2               ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm0             ; xmm7=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm0             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm5             ; xmm2=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm0,xmm3             ; transpose coefficients(phase 2)
+        punpckldq xmm3,xmm4             ; xmm3=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm0,xmm4             ; xmm0=(06 16 26 36 07 17 27 37)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7             ; xmm6=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm5,xmm7             ; xmm5=(02 12 22 32 03 13 23 33)
+
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(60 70 61 71 62 72 63 73)
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm3,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm4             ; xmm1=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm3,xmm4             ; xmm3=(42 52 62 72 43 53 63 73)
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm7             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm0,xmm7             ; xmm0=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm4,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm1            ; xmm6=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm4,xmm1            ; xmm4=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm7,xmm5            ; transpose coefficients(phase 3)
+        punpcklqdq xmm5,xmm3            ; xmm5=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm7,xmm3            ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(04 14 24 34 05 15 25 35)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=col3
+
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm4,xmm2            ; xmm4=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm7,xmm3            ; transpose coefficients(phase 3)
+        punpcklqdq xmm3,xmm0            ; xmm3=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm7,xmm0            ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm0,xmm5
+        psubw   xmm6,xmm1               ; xmm6=tmp11
+        psubw   xmm5,xmm3
+        paddw   xmm2,xmm1               ; xmm2=tmp10
+        paddw   xmm0,xmm3               ; xmm0=tmp13
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F1414)]
+        psubw   xmm5,xmm0               ; xmm5=tmp12
+
+        movdqa  xmm1,xmm2
+        movdqa  xmm3,xmm6
+        psubw   xmm2,xmm0               ; xmm2=tmp3
+        psubw   xmm6,xmm5               ; xmm6=tmp2
+        paddw   xmm1,xmm0               ; xmm1=tmp0
+        paddw   xmm3,xmm5               ; xmm3=tmp1
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=col1
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=col3
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp3
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp2
+
+        ; -- Odd part
+
+        ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+        movdqa  xmm2,xmm0
+        movdqa  xmm6,xmm4
+        psubw   xmm0,xmm7               ; xmm0=z12
+        psubw   xmm4,xmm5               ; xmm4=z10
+        paddw   xmm2,xmm7               ; xmm2=z11
+        paddw   xmm6,xmm5               ; xmm6=z13
+
+        movdqa  xmm7,xmm4               ; xmm7=z10(unscaled)
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm5,xmm2
+        psubw   xmm2,xmm6
+        paddw   xmm5,xmm6               ; xmm5=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F1414)]     ; xmm2=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm6,xmm4
+        paddw   xmm4,xmm0
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F1847)]     ; xmm4=z5
+        pmulhw  xmm6,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F1082)]
+        psubw   xmm6,xmm7
+        psubw   xmm0,xmm4               ; xmm0=tmp10
+        paddw   xmm6,xmm4               ; xmm6=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm6,xmm5               ; xmm6=tmp6
+        movdqa  xmm7,xmm1
+        movdqa  xmm4,xmm3
+        paddw   xmm1,xmm5               ; xmm1=data0=(00 10 20 30 40 50 60 70)
+        paddw   xmm3,xmm6               ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        psraw   xmm1,(PASS1_BITS+3)     ; descale
+        psraw   xmm3,(PASS1_BITS+3)     ; descale
+        psubw   xmm7,xmm5               ; xmm7=data7=(07 17 27 37 47 57 67 77)
+        psubw   xmm4,xmm6               ; xmm4=data6=(06 16 26 36 46 56 66 76)
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psubw   xmm2,xmm6               ; xmm2=tmp5
+
+        packsswb  xmm1,xmm4     ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm7     ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp2
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp3
+
+        paddw   xmm0,xmm2               ; xmm0=tmp4
+        movdqa  xmm4,xmm5
+        movdqa  xmm7,xmm6
+        paddw   xmm5,xmm2               ; xmm5=data2=(02 12 22 32 42 52 62 72)
+        paddw   xmm6,xmm0               ; xmm6=data4=(04 14 24 34 44 54 64 74)
+        psraw   xmm5,(PASS1_BITS+3)     ; descale
+        psraw   xmm6,(PASS1_BITS+3)     ; descale
+        psubw   xmm4,xmm2               ; xmm4=data5=(05 15 25 35 45 55 65 75)
+        psubw   xmm7,xmm0               ; xmm7=data3=(03 13 23 33 43 53 63 73)
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+
+        movdqa    xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm2=[PB_CENTERJSAMP]
+
+        packsswb  xmm5,xmm6     ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm7,xmm4     ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm1,xmm2
+        paddb     xmm3,xmm2
+        paddb     xmm5,xmm2
+        paddb     xmm7,xmm2
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 1)
+        punpcklbw xmm1,xmm3     ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm3     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm6,xmm5     ; transpose coefficients(phase 1)
+        punpcklbw xmm5,xmm7     ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm6,xmm7     ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm1     ; transpose coefficients(phase 2)
+        punpcklwd xmm1,xmm5     ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm5     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm0     ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm2,xmm0     ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm3,xmm1     ; transpose coefficients(phase 3)
+        punpckldq xmm1,xmm6     ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm3,xmm6     ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm7,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm2     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm7,xmm2     ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm5,xmm1,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm3,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm6,xmm4,0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm2,xmm7,0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm1
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+        mov     edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm7
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0
+        mov     edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm2
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,359 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* SLOW INTEGER INVERSE DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298 2446   /* FIX(0.298631336) */
+#define F_0_390 3196   /* FIX(0.390180644) */
+#define F_0_541 4433   /* FIX(0.541196100) */
+#define F_0_765 6270   /* FIX(0.765366865) */
+#define F_0_899 7373   /* FIX(0.899976223) */
+#define F_1_175 9633   /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+#define DESCALE_P1 (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS + PASS1_BITS + 3)
+
+
+#define DO_IDCT(in, PASS)  \
+{  \
+  /* Even part  \
+   *  \
+   * (Original)  \
+   * z1 = (z2 + z3) * 0.541196100;  \
+   * tmp2 = z1 + z3 * -1.847759065;  \
+   * tmp3 = z1 + z2 * 0.765366865;  \
+   *  \
+   * (This implementation)  \
+   * tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);  \
+   * tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;  \
+   */  \
+  \
+  in##26l = vec_mergeh(in##2, in##6);  \
+  in##26h = vec_mergel(in##2, in##6);  \
+  \
+  tmp3l = vec_msums(in##26l, pw_f130_f054, pd_zero);  \
+  tmp3h = vec_msums(in##26h, pw_f130_f054, pd_zero);  \
+  tmp2l = vec_msums(in##26l, pw_f054_mf130, pd_zero);  \
+  tmp2h = vec_msums(in##26h, pw_f054_mf130, pd_zero);  \
+  \
+  tmp0 = vec_add(in##0, in##4);  \
+  tmp1 = vec_sub(in##0, in##4);  \
+  \
+  tmp0l = vec_unpackh(tmp0);  \
+  tmp0h = vec_unpackl(tmp0);  \
+  tmp0l = vec_sl(tmp0l, const_bits);  \
+  tmp0h = vec_sl(tmp0h, const_bits);  \
+  tmp0l = vec_add(tmp0l, pd_descale_p##PASS);  \
+  tmp0h = vec_add(tmp0h, pd_descale_p##PASS);  \
+  \
+  tmp10l = vec_add(tmp0l, tmp3l);  \
+  tmp10h = vec_add(tmp0h, tmp3h);  \
+  tmp13l = vec_sub(tmp0l, tmp3l);  \
+  tmp13h = vec_sub(tmp0h, tmp3h);  \
+  \
+  tmp1l = vec_unpackh(tmp1);  \
+  tmp1h = vec_unpackl(tmp1);  \
+  tmp1l = vec_sl(tmp1l, const_bits);  \
+  tmp1h = vec_sl(tmp1h, const_bits);  \
+  tmp1l = vec_add(tmp1l, pd_descale_p##PASS);  \
+  tmp1h = vec_add(tmp1h, pd_descale_p##PASS);  \
+  \
+  tmp11l = vec_add(tmp1l, tmp2l);  \
+  tmp11h = vec_add(tmp1h, tmp2h);  \
+  tmp12l = vec_sub(tmp1l, tmp2l);  \
+  tmp12h = vec_sub(tmp1h, tmp2h);  \
+  \
+  /* Odd part */  \
+  \
+  z3 = vec_add(in##3, in##7);  \
+  z4 = vec_add(in##1, in##5);  \
+  \
+  /* (Original)  \
+   * z5 = (z3 + z4) * 1.175875602;  \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;  \
+   * z3 += z5;  z4 += z5;  \
+   *  \
+   * (This implementation)  \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;  \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);  \
+   */  \
+  \
+  z34l = vec_mergeh(z3, z4);  \
+  z34h = vec_mergel(z3, z4);  \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_zero);  \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_zero);  \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_zero);  \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_zero);  \
+  \
+  /* (Original)  \
+   * z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;  \
+   * tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;  \
+   * tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;  \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;  \
+   * tmp0 += z1 + z3;  tmp1 += z2 + z4;  \
+   * tmp2 += z2 + z3;  tmp3 += z1 + z4;  \
+   *  \
+   * (This implementation)  \
+   * tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;  \
+   * tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;  \
+   * tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);  \
+   * tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);  \
+   * tmp0 += z3;  tmp1 += z4;  \
+   * tmp2 += z3;  tmp3 += z4;  \
+   */  \
+  \
+  in##71l = vec_mergeh(in##7, in##1);  \
+  in##71h = vec_mergel(in##7, in##1);  \
+  \
+  tmp0l = vec_msums(in##71l, pw_mf060_mf089, z3l);  \
+  tmp0h = vec_msums(in##71h, pw_mf060_mf089, z3h);  \
+  tmp3l = vec_msums(in##71l, pw_mf089_f060, z4l);  \
+  tmp3h = vec_msums(in##71h, pw_mf089_f060, z4h);  \
+  \
+  in##53l = vec_mergeh(in##5, in##3);  \
+  in##53h = vec_mergel(in##5, in##3);  \
+  \
+  tmp1l = vec_msums(in##53l, pw_mf050_mf256, z4l);  \
+  tmp1h = vec_msums(in##53h, pw_mf050_mf256, z4h);  \
+  tmp2l = vec_msums(in##53l, pw_mf256_f050, z3l);  \
+  tmp2h = vec_msums(in##53h, pw_mf256_f050, z3h);  \
+  \
+  /* Final output stage */  \
+  \
+  out0l = vec_add(tmp10l, tmp3l);  \
+  out0h = vec_add(tmp10h, tmp3h);  \
+  out7l = vec_sub(tmp10l, tmp3l);  \
+  out7h = vec_sub(tmp10h, tmp3h);  \
+  \
+  out0l = vec_sra(out0l, descale_p##PASS);  \
+  out0h = vec_sra(out0h, descale_p##PASS);  \
+  out7l = vec_sra(out7l, descale_p##PASS);  \
+  out7h = vec_sra(out7h, descale_p##PASS);  \
+  \
+  out0 = vec_pack(out0l, out0h);  \
+  out7 = vec_pack(out7l, out7h);  \
+  \
+  out1l = vec_add(tmp11l, tmp2l);  \
+  out1h = vec_add(tmp11h, tmp2h);  \
+  out6l = vec_sub(tmp11l, tmp2l);  \
+  out6h = vec_sub(tmp11h, tmp2h);  \
+  \
+  out1l = vec_sra(out1l, descale_p##PASS);  \
+  out1h = vec_sra(out1h, descale_p##PASS);  \
+  out6l = vec_sra(out6l, descale_p##PASS);  \
+  out6h = vec_sra(out6h, descale_p##PASS);  \
+  \
+  out1 = vec_pack(out1l, out1h);  \
+  out6 = vec_pack(out6l, out6h);  \
+  \
+  out2l = vec_add(tmp12l, tmp1l);  \
+  out2h = vec_add(tmp12h, tmp1h);  \
+  out5l = vec_sub(tmp12l, tmp1l);  \
+  out5h = vec_sub(tmp12h, tmp1h);  \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS);  \
+  out2h = vec_sra(out2h, descale_p##PASS);  \
+  out5l = vec_sra(out5l, descale_p##PASS);  \
+  out5h = vec_sra(out5h, descale_p##PASS);  \
+  \
+  out2 = vec_pack(out2l, out2h);  \
+  out5 = vec_pack(out5l, out5h);  \
+  \
+  out3l = vec_add(tmp13l, tmp0l);  \
+  out3h = vec_add(tmp13h, tmp0h);  \
+  out4l = vec_sub(tmp13l, tmp0l);  \
+  out4h = vec_sub(tmp13h, tmp0h);  \
+  \
+  out3l = vec_sra(out3l, descale_p##PASS);  \
+  out3h = vec_sra(out3h, descale_p##PASS);  \
+  out4l = vec_sra(out4l, descale_p##PASS);  \
+  out4h = vec_sra(out4h, descale_p##PASS);  \
+  \
+  out3 = vec_pack(out3l, out3h);  \
+  out4 = vec_pack(out4l, out4h);  \
+}
+
+
+void
+jsimd_idct_islow_altivec (void *dct_table_, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, z3, z4,
+    z34l, z34h, col71l, col71h, col26l, col26h, col53l, col53h,
+    row71l, row71h, row26l, row26h, row53l, row53h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h,
+    tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h,
+    z3l, z3h, z4l, z4h,
+    out0l, out0h, out1l, out1h, out2l, out2h, out3l, out3h, out4l, out4h,
+    out5l, out5h, out6l, out6h, out7l, out7h;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_zero = { __4X(0) },
+    pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) },
+    const_bits = { __4X(CONST_BITS) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    col0 = vec_sl(col0, pass1_bits);
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col, 1);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row, 2);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[0] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[1] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[2] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[3] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[4] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[5] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[6] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[7] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,852 @@
+;
+; jidctint.asm - accurate integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_mmx) PRIVATE
+
+EXTN(jconst_idct_islow_mmx):
+
+PW_F130_F054    times 2 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 2 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 2 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 2 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 2 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 2 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 2 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 2 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 2 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 2 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_mmx (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          12
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_islow_mmx) PRIVATE
+
+EXTN(jsimd_idct_islow_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   mm0,PASS1_BITS
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movq      mm4,mm1               ; mm1=in2=z2
+        movq      mm5,mm1
+        punpcklwd mm4,mm3               ; mm3=in6=z3
+        punpckhwd mm5,mm3
+        movq      mm1,mm4
+        movq      mm3,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=tmp3L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F130_F054)]        ; mm5=tmp3H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F054_MF130)]       ; mm3=tmp2H
+
+        movq      mm6,mm0
+        paddw     mm0,mm2               ; mm0=in0+in4
+        psubw     mm6,mm2               ; mm6=in0-in4
+
+        pxor      mm7,mm7
+        pxor      mm2,mm2
+        punpcklwd mm7,mm0               ; mm7=tmp0L
+        punpckhwd mm2,mm0               ; mm2=tmp0H
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+        psrad     mm2,(16-CONST_BITS)   ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+        movq    mm0,mm7
+        paddd   mm7,mm4                 ; mm7=tmp10L
+        psubd   mm0,mm4                 ; mm0=tmp13L
+        movq    mm4,mm2
+        paddd   mm2,mm5                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp13H
+
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp10L
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=tmp10H
+        movq    MMWORD [wk(2)], mm0     ; wk(2)=tmp13L
+        movq    MMWORD [wk(3)], mm4     ; wk(3)=tmp13H
+
+        pxor      mm5,mm5
+        pxor      mm7,mm7
+        punpcklwd mm5,mm6               ; mm5=tmp1L
+        punpckhwd mm7,mm6               ; mm7=tmp1H
+        psrad     mm5,(16-CONST_BITS)   ; psrad mm5,16 & pslld mm5,CONST_BITS
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+        movq    mm2,mm5
+        paddd   mm5,mm1                 ; mm5=tmp11L
+        psubd   mm2,mm1                 ; mm2=tmp12L
+        movq    mm0,mm7
+        paddd   mm7,mm3                 ; mm7=tmp11H
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        movq    MMWORD [wk(4)], mm5     ; wk(4)=tmp11L
+        movq    MMWORD [wk(5)], mm7     ; wk(5)=tmp11H
+        movq    MMWORD [wk(6)], mm2     ; wk(6)=tmp12L
+        movq    MMWORD [wk(7)], mm0     ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movq    mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm4, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm6, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm1, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movq    mm5,mm6
+        movq    mm7,mm4
+        paddw   mm5,mm3                 ; mm5=z3
+        paddw   mm7,mm1                 ; mm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm2,mm5
+        movq      mm0,mm5
+        punpcklwd mm2,mm7
+        punpckhwd mm0,mm7
+        movq      mm5,mm2
+        movq      mm7,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF078_F117)]       ; mm2=z3L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF078_F117)]       ; mm0=z3H
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F117_F078)]        ; mm5=z4L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_F117_F078)]        ; mm7=z4H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=z3L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movq      mm2,mm3
+        movq      mm0,mm3
+        punpcklwd mm2,mm4
+        punpckhwd mm0,mm4
+        movq      mm3,mm2
+        movq      mm4,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm2=tmp0L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm0=tmp0H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF089_F060)]       ; mm3=tmp3L
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF089_F060)]       ; mm4=tmp3H
+
+        paddd   mm2, MMWORD [wk(10)]    ; mm2=tmp0L
+        paddd   mm0, MMWORD [wk(11)]    ; mm0=tmp0H
+        paddd   mm3,mm5                 ; mm3=tmp3L
+        paddd   mm4,mm7                 ; mm4=tmp3H
+
+        movq    MMWORD [wk(8)], mm2     ; wk(8)=tmp0L
+        movq    MMWORD [wk(9)], mm0     ; wk(9)=tmp0H
+
+        movq      mm2,mm1
+        movq      mm0,mm1
+        punpcklwd mm2,mm6
+        punpckhwd mm0,mm6
+        movq      mm1,mm2
+        movq      mm6,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm2=tmp1L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm0=tmp1H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF256_F050)]       ; mm1=tmp2L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF256_F050)]       ; mm6=tmp2H
+
+        paddd   mm2,mm5                 ; mm2=tmp1L
+        paddd   mm0,mm7                 ; mm0=tmp1H
+        paddd   mm1, MMWORD [wk(10)]    ; mm1=tmp2L
+        paddd   mm6, MMWORD [wk(11)]    ; mm6=tmp2H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=tmp1L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp10L
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp10H
+
+        movq    mm2,mm5
+        movq    mm0,mm7
+        paddd   mm5,mm3                 ; mm5=data0L
+        paddd   mm7,mm4                 ; mm7=data0H
+        psubd   mm2,mm3                 ; mm2=data7L
+        psubd   mm0,mm4                 ; mm0=data7H
+
+        movq    mm3,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm3=[PD_DESCALE_P1]
+
+        paddd   mm5,mm3
+        paddd   mm7,mm3
+        psrad   mm5,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+        paddd   mm2,mm3
+        paddd   mm0,mm3
+        psrad   mm2,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+
+        packssdw  mm5,mm7               ; mm5=data0=(00 01 02 03)
+        packssdw  mm2,mm0               ; mm2=data7=(70 71 72 73)
+
+        movq    mm4, MMWORD [wk(4)]     ; mm4=tmp11L
+        movq    mm3, MMWORD [wk(5)]     ; mm3=tmp11H
+
+        movq    mm7,mm4
+        movq    mm0,mm3
+        paddd   mm4,mm1                 ; mm4=data1L
+        paddd   mm3,mm6                 ; mm3=data1H
+        psubd   mm7,mm1                 ; mm7=data6L
+        psubd   mm0,mm6                 ; mm0=data6H
+
+        movq    mm1,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm1=[PD_DESCALE_P1]
+
+        paddd   mm4,mm1
+        paddd   mm3,mm1
+        psrad   mm4,DESCALE_P1
+        psrad   mm3,DESCALE_P1
+        paddd   mm7,mm1
+        paddd   mm0,mm1
+        psrad   mm7,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+
+        packssdw  mm4,mm3               ; mm4=data1=(10 11 12 13)
+        packssdw  mm7,mm0               ; mm7=data6=(60 61 62 63)
+
+        movq      mm6,mm5               ; transpose coefficients(phase 1)
+        punpcklwd mm5,mm4               ; mm5=(00 10 01 11)
+        punpckhwd mm6,mm4               ; mm6=(02 12 03 13)
+        movq      mm1,mm7               ; transpose coefficients(phase 1)
+        punpcklwd mm7,mm2               ; mm7=(60 70 61 71)
+        punpckhwd mm1,mm2               ; mm1=(62 72 63 73)
+
+        movq    mm3, MMWORD [wk(6)]     ; mm3=tmp12L
+        movq    mm0, MMWORD [wk(7)]     ; mm0=tmp12H
+        movq    mm4, MMWORD [wk(10)]    ; mm4=tmp1L
+        movq    mm2, MMWORD [wk(11)]    ; mm2=tmp1H
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(00 10 01 11)
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=(02 12 03 13)
+        movq    MMWORD [wk(4)], mm7     ; wk(4)=(60 70 61 71)
+        movq    MMWORD [wk(5)], mm1     ; wk(5)=(62 72 63 73)
+
+        movq    mm5,mm3
+        movq    mm6,mm0
+        paddd   mm3,mm4                 ; mm3=data2L
+        paddd   mm0,mm2                 ; mm0=data2H
+        psubd   mm5,mm4                 ; mm5=data5L
+        psubd   mm6,mm2                 ; mm6=data5H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm7=[PD_DESCALE_P1]
+
+        paddd   mm3,mm7
+        paddd   mm0,mm7
+        psrad   mm3,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+        paddd   mm5,mm7
+        paddd   mm6,mm7
+        psrad   mm5,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm3,mm0               ; mm3=data2=(20 21 22 23)
+        packssdw  mm5,mm6               ; mm5=data5=(50 51 52 53)
+
+        movq    mm1, MMWORD [wk(2)]     ; mm1=tmp13L
+        movq    mm4, MMWORD [wk(3)]     ; mm4=tmp13H
+        movq    mm2, MMWORD [wk(8)]     ; mm2=tmp0L
+        movq    mm7, MMWORD [wk(9)]     ; mm7=tmp0H
+
+        movq    mm0,mm1
+        movq    mm6,mm4
+        paddd   mm1,mm2                 ; mm1=data3L
+        paddd   mm4,mm7                 ; mm4=data3H
+        psubd   mm0,mm2                 ; mm0=data4L
+        psubd   mm6,mm7                 ; mm6=data4H
+
+        movq    mm2,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm2=[PD_DESCALE_P1]
+
+        paddd   mm1,mm2
+        paddd   mm4,mm2
+        psrad   mm1,DESCALE_P1
+        psrad   mm4,DESCALE_P1
+        paddd   mm0,mm2
+        paddd   mm6,mm2
+        psrad   mm0,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm1,mm4               ; mm1=data3=(30 31 32 33)
+        packssdw  mm0,mm6               ; mm0=data4=(40 41 42 43)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=(00 10 01 11)
+        movq    mm2, MMWORD [wk(1)]     ; mm2=(02 12 03 13)
+
+        movq      mm4,mm3               ; transpose coefficients(phase 1)
+        punpcklwd mm3,mm1               ; mm3=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm6,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm5               ; mm0=(40 50 41 51)
+        punpckhwd mm6,mm5               ; mm6=(42 52 43 53)
+
+        movq      mm1,mm7               ; transpose coefficients(phase 2)
+        punpckldq mm7,mm3               ; mm7=(00 10 20 30)
+        punpckhdq mm1,mm3               ; mm1=(01 11 21 31)
+        movq      mm5,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm4               ; mm2=(02 12 22 32)
+        punpckhdq mm5,mm4               ; mm5=(03 13 23 33)
+
+        movq    mm3, MMWORD [wk(4)]     ; mm3=(60 70 61 71)
+        movq    mm4, MMWORD [wk(5)]     ; mm4=(62 72 63 73)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5
+
+        movq      mm7,mm0               ; transpose coefficients(phase 2)
+        punpckldq mm0,mm3               ; mm0=(40 50 60 70)
+        punpckhdq mm7,mm3               ; mm7=(41 51 61 71)
+        movq      mm1,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm4               ; mm6=(42 52 62 72)
+        punpckhdq mm1,mm4               ; mm1=(43 53 63 73)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm7
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm1
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_ISLOW_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movq      mm4,mm1               ; mm1=in2=z2
+        movq      mm5,mm1
+        punpcklwd mm4,mm3               ; mm3=in6=z3
+        punpckhwd mm5,mm3
+        movq      mm1,mm4
+        movq      mm3,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=tmp3L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F130_F054)]        ; mm5=tmp3H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F054_MF130)]       ; mm3=tmp2H
+
+        movq      mm6,mm0
+        paddw     mm0,mm2               ; mm0=in0+in4
+        psubw     mm6,mm2               ; mm6=in0-in4
+
+        pxor      mm7,mm7
+        pxor      mm2,mm2
+        punpcklwd mm7,mm0               ; mm7=tmp0L
+        punpckhwd mm2,mm0               ; mm2=tmp0H
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+        psrad     mm2,(16-CONST_BITS)   ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+        movq    mm0,mm7
+        paddd   mm7,mm4                 ; mm7=tmp10L
+        psubd   mm0,mm4                 ; mm0=tmp13L
+        movq    mm4,mm2
+        paddd   mm2,mm5                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp13H
+
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp10L
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=tmp10H
+        movq    MMWORD [wk(2)], mm0     ; wk(2)=tmp13L
+        movq    MMWORD [wk(3)], mm4     ; wk(3)=tmp13H
+
+        pxor      mm5,mm5
+        pxor      mm7,mm7
+        punpcklwd mm5,mm6               ; mm5=tmp1L
+        punpckhwd mm7,mm6               ; mm7=tmp1H
+        psrad     mm5,(16-CONST_BITS)   ; psrad mm5,16 & pslld mm5,CONST_BITS
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+        movq    mm2,mm5
+        paddd   mm5,mm1                 ; mm5=tmp11L
+        psubd   mm2,mm1                 ; mm2=tmp12L
+        movq    mm0,mm7
+        paddd   mm7,mm3                 ; mm7=tmp11H
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        movq    MMWORD [wk(4)], mm5     ; wk(4)=tmp11L
+        movq    MMWORD [wk(5)], mm7     ; wk(5)=tmp11H
+        movq    MMWORD [wk(6)], mm2     ; wk(6)=tmp12L
+        movq    MMWORD [wk(7)], mm0     ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movq    mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm5,mm6
+        movq    mm7,mm4
+        paddw   mm5,mm3                 ; mm5=z3
+        paddw   mm7,mm1                 ; mm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm2,mm5
+        movq      mm0,mm5
+        punpcklwd mm2,mm7
+        punpckhwd mm0,mm7
+        movq      mm5,mm2
+        movq      mm7,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF078_F117)]       ; mm2=z3L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF078_F117)]       ; mm0=z3H
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F117_F078)]        ; mm5=z4L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_F117_F078)]        ; mm7=z4H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=z3L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movq      mm2,mm3
+        movq      mm0,mm3
+        punpcklwd mm2,mm4
+        punpckhwd mm0,mm4
+        movq      mm3,mm2
+        movq      mm4,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm2=tmp0L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm0=tmp0H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF089_F060)]       ; mm3=tmp3L
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF089_F060)]       ; mm4=tmp3H
+
+        paddd   mm2, MMWORD [wk(10)]    ; mm2=tmp0L
+        paddd   mm0, MMWORD [wk(11)]    ; mm0=tmp0H
+        paddd   mm3,mm5                 ; mm3=tmp3L
+        paddd   mm4,mm7                 ; mm4=tmp3H
+
+        movq    MMWORD [wk(8)], mm2     ; wk(8)=tmp0L
+        movq    MMWORD [wk(9)], mm0     ; wk(9)=tmp0H
+
+        movq      mm2,mm1
+        movq      mm0,mm1
+        punpcklwd mm2,mm6
+        punpckhwd mm0,mm6
+        movq      mm1,mm2
+        movq      mm6,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm2=tmp1L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm0=tmp1H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF256_F050)]       ; mm1=tmp2L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF256_F050)]       ; mm6=tmp2H
+
+        paddd   mm2,mm5                 ; mm2=tmp1L
+        paddd   mm0,mm7                 ; mm0=tmp1H
+        paddd   mm1, MMWORD [wk(10)]    ; mm1=tmp2L
+        paddd   mm6, MMWORD [wk(11)]    ; mm6=tmp2H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=tmp1L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp10L
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp10H
+
+        movq    mm2,mm5
+        movq    mm0,mm7
+        paddd   mm5,mm3                 ; mm5=data0L
+        paddd   mm7,mm4                 ; mm7=data0H
+        psubd   mm2,mm3                 ; mm2=data7L
+        psubd   mm0,mm4                 ; mm0=data7H
+
+        movq    mm3,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm3=[PD_DESCALE_P2]
+
+        paddd   mm5,mm3
+        paddd   mm7,mm3
+        psrad   mm5,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+        paddd   mm2,mm3
+        paddd   mm0,mm3
+        psrad   mm2,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+
+        packssdw  mm5,mm7               ; mm5=data0=(00 10 20 30)
+        packssdw  mm2,mm0               ; mm2=data7=(07 17 27 37)
+
+        movq    mm4, MMWORD [wk(4)]     ; mm4=tmp11L
+        movq    mm3, MMWORD [wk(5)]     ; mm3=tmp11H
+
+        movq    mm7,mm4
+        movq    mm0,mm3
+        paddd   mm4,mm1                 ; mm4=data1L
+        paddd   mm3,mm6                 ; mm3=data1H
+        psubd   mm7,mm1                 ; mm7=data6L
+        psubd   mm0,mm6                 ; mm0=data6H
+
+        movq    mm1,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm1=[PD_DESCALE_P2]
+
+        paddd   mm4,mm1
+        paddd   mm3,mm1
+        psrad   mm4,DESCALE_P2
+        psrad   mm3,DESCALE_P2
+        paddd   mm7,mm1
+        paddd   mm0,mm1
+        psrad   mm7,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+
+        packssdw  mm4,mm3               ; mm4=data1=(01 11 21 31)
+        packssdw  mm7,mm0               ; mm7=data6=(06 16 26 36)
+
+        packsswb  mm5,mm7               ; mm5=(00 10 20 30 06 16 26 36)
+        packsswb  mm4,mm2               ; mm4=(01 11 21 31 07 17 27 37)
+
+        movq    mm6, MMWORD [wk(6)]     ; mm6=tmp12L
+        movq    mm1, MMWORD [wk(7)]     ; mm1=tmp12H
+        movq    mm3, MMWORD [wk(10)]    ; mm3=tmp1L
+        movq    mm0, MMWORD [wk(11)]    ; mm0=tmp1H
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(00 10 20 30 06 16 26 36)
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=(01 11 21 31 07 17 27 37)
+
+        movq    mm7,mm6
+        movq    mm2,mm1
+        paddd   mm6,mm3                 ; mm6=data2L
+        paddd   mm1,mm0                 ; mm1=data2H
+        psubd   mm7,mm3                 ; mm7=data5L
+        psubd   mm2,mm0                 ; mm2=data5H
+
+        movq    mm5,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm5=[PD_DESCALE_P2]
+
+        paddd   mm6,mm5
+        paddd   mm1,mm5
+        psrad   mm6,DESCALE_P2
+        psrad   mm1,DESCALE_P2
+        paddd   mm7,mm5
+        paddd   mm2,mm5
+        psrad   mm7,DESCALE_P2
+        psrad   mm2,DESCALE_P2
+
+        packssdw  mm6,mm1               ; mm6=data2=(02 12 22 32)
+        packssdw  mm7,mm2               ; mm7=data5=(05 15 25 35)
+
+        movq    mm4, MMWORD [wk(2)]     ; mm4=tmp13L
+        movq    mm3, MMWORD [wk(3)]     ; mm3=tmp13H
+        movq    mm0, MMWORD [wk(8)]     ; mm0=tmp0L
+        movq    mm5, MMWORD [wk(9)]     ; mm5=tmp0H
+
+        movq    mm1,mm4
+        movq    mm2,mm3
+        paddd   mm4,mm0                 ; mm4=data3L
+        paddd   mm3,mm5                 ; mm3=data3H
+        psubd   mm1,mm0                 ; mm1=data4L
+        psubd   mm2,mm5                 ; mm2=data4H
+
+        movq    mm0,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm0=[PD_DESCALE_P2]
+
+        paddd   mm4,mm0
+        paddd   mm3,mm0
+        psrad   mm4,DESCALE_P2
+        psrad   mm3,DESCALE_P2
+        paddd   mm1,mm0
+        paddd   mm2,mm0
+        psrad   mm1,DESCALE_P2
+        psrad   mm2,DESCALE_P2
+
+        movq      mm5,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm5=[PB_CENTERJSAMP]
+
+        packssdw  mm4,mm3               ; mm4=data3=(03 13 23 33)
+        packssdw  mm1,mm2               ; mm1=data4=(04 14 24 34)
+
+        movq      mm0, MMWORD [wk(0)]   ; mm0=(00 10 20 30 06 16 26 36)
+        movq      mm3, MMWORD [wk(1)]   ; mm3=(01 11 21 31 07 17 27 37)
+
+        packsswb  mm6,mm1               ; mm6=(02 12 22 32 04 14 24 34)
+        packsswb  mm4,mm7               ; mm4=(03 13 23 33 05 15 25 35)
+
+        paddb     mm0,mm5
+        paddb     mm3,mm5
+        paddb     mm6,mm5
+        paddb     mm4,mm5
+
+        movq      mm2,mm0               ; transpose coefficients(phase 1)
+        punpcklbw mm0,mm3               ; mm0=(00 01 10 11 20 21 30 31)
+        punpckhbw mm2,mm3               ; mm2=(06 07 16 17 26 27 36 37)
+        movq      mm1,mm6               ; transpose coefficients(phase 1)
+        punpcklbw mm6,mm4               ; mm6=(02 03 12 13 22 23 32 33)
+        punpckhbw mm1,mm4               ; mm1=(04 05 14 15 24 25 34 35)
+
+        movq      mm7,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm6               ; mm0=(00 01 02 03 10 11 12 13)
+        punpckhwd mm7,mm6               ; mm7=(20 21 22 23 30 31 32 33)
+        movq      mm5,mm1               ; transpose coefficients(phase 2)
+        punpcklwd mm1,mm2               ; mm1=(04 05 06 07 14 15 16 17)
+        punpckhwd mm5,mm2               ; mm5=(24 25 26 27 34 35 36 37)
+
+        movq      mm3,mm0               ; transpose coefficients(phase 3)
+        punpckldq mm0,mm1               ; mm0=(00 01 02 03 04 05 06 07)
+        punpckhdq mm3,mm1               ; mm3=(10 11 12 13 14 15 16 17)
+        movq      mm4,mm7               ; transpose coefficients(phase 3)
+        punpckldq mm7,mm5               ; mm7=(20 21 22 23 24 25 26 27)
+        punpckhdq mm4,mm5               ; mm4=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm3
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm7
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_JCOEF        ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,848 @@
+;
+; jidctint.asm - accurate integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_sse2) PRIVATE
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          12
+
+        align   16
+        global  EXTN(jsimd_idct_islow_sse2) PRIVATE
+
+EXTN(jsimd_idct_islow_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm5, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm5,PASS1_BITS
+
+        movdqa    xmm4,xmm5             ; xmm5=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm5,xmm5             ; xmm5=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm4,xmm4             ; xmm4=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm7,xmm5,0x00          ; xmm7=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm6,xmm5,0x55          ; xmm6=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm1,xmm5,0xAA          ; xmm1=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm5,xmm5,0xFF          ; xmm5=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm0,xmm4,0x00          ; xmm0=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm3,xmm4,0x55          ; xmm3=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm2,xmm4,0xAA          ; xmm2=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm4,xmm4,0xFF          ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(8)], xmm6   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm5   ; wk(9)=col3
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm4,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm4,xmm3             ; xmm3=in6=z3
+        punpckhwd xmm5,xmm3
+        movdqa    xmm1,xmm4
+        movdqa    xmm3,xmm5
+        pmaddwd   xmm4,[rel PW_F130_F054]       ; xmm4=tmp3L
+        pmaddwd   xmm5,[rel PW_F130_F054]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=tmp2L
+        pmaddwd   xmm3,[rel PW_F054_MF130]      ; xmm3=tmp2H
+
+        movdqa    xmm6,xmm0
+        paddw     xmm0,xmm2             ; xmm0=in0+in4
+        psubw     xmm6,xmm2             ; xmm6=in0-in4
+
+        pxor      xmm7,xmm7
+        pxor      xmm2,xmm2
+        punpcklwd xmm7,xmm0             ; xmm7=tmp0L
+        punpckhwd xmm2,xmm0             ; xmm2=tmp0H
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+        psrad     xmm2,(16-CONST_BITS)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm4               ; xmm7=tmp10L
+        psubd   xmm0,xmm4               ; xmm0=tmp13L
+        movdqa  xmm4,xmm2
+        paddd   xmm2,xmm5               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm0   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm4   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm7,xmm7
+        punpcklwd xmm5,xmm6             ; xmm5=tmp1L
+        punpckhwd xmm7,xmm6             ; xmm7=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+        movdqa  xmm2,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm2,xmm1               ; xmm2=tmp12L
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm3               ; xmm7=tmp11H
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm7   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm0   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm7,xmm4
+        paddw   xmm5,xmm3               ; xmm5=z3
+        paddw   xmm7,xmm1               ; xmm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm2,xmm5
+        movdqa    xmm0,xmm5
+        punpcklwd xmm2,xmm7
+        punpckhwd xmm0,xmm7
+        movdqa    xmm5,xmm2
+        movdqa    xmm7,xmm0
+        pmaddwd   xmm2,[rel PW_MF078_F117]      ; xmm2=z3L
+        pmaddwd   xmm0,[rel PW_MF078_F117]      ; xmm0=z3H
+        pmaddwd   xmm5,[rel PW_F117_F078]       ; xmm5=z4L
+        pmaddwd   xmm7,[rel PW_F117_F078]       ; xmm7=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm2,xmm3
+        movdqa    xmm0,xmm3
+        punpcklwd xmm2,xmm4
+        punpckhwd xmm0,xmm4
+        movdqa    xmm3,xmm2
+        movdqa    xmm4,xmm0
+        pmaddwd   xmm2,[rel PW_MF060_MF089]     ; xmm2=tmp0L
+        pmaddwd   xmm0,[rel PW_MF060_MF089]     ; xmm0=tmp0H
+        pmaddwd   xmm3,[rel PW_MF089_F060]      ; xmm3=tmp3L
+        pmaddwd   xmm4,[rel PW_MF089_F060]      ; xmm4=tmp3H
+
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp0L
+        paddd   xmm0, XMMWORD [wk(11)]  ; xmm0=tmp0H
+        paddd   xmm3,xmm5               ; xmm3=tmp3L
+        paddd   xmm4,xmm7               ; xmm4=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm2   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm0   ; wk(9)=tmp0H
+
+        movdqa    xmm2,xmm1
+        movdqa    xmm0,xmm1
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm0,xmm6
+        movdqa    xmm1,xmm2
+        movdqa    xmm6,xmm0
+        pmaddwd   xmm2,[rel PW_MF050_MF256]     ; xmm2=tmp1L
+        pmaddwd   xmm0,[rel PW_MF050_MF256]     ; xmm0=tmp1H
+        pmaddwd   xmm1,[rel PW_MF256_F050]      ; xmm1=tmp2L
+        pmaddwd   xmm6,[rel PW_MF256_F050]      ; xmm6=tmp2H
+
+        paddd   xmm2,xmm5               ; xmm2=tmp1L
+        paddd   xmm0,xmm7               ; xmm0=tmp1H
+        paddd   xmm1, XMMWORD [wk(10)]  ; xmm1=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=tmp10H
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm0,xmm7
+        paddd   xmm5,xmm3               ; xmm5=data0L
+        paddd   xmm7,xmm4               ; xmm7=data0H
+        psubd   xmm2,xmm3               ; xmm2=data7L
+        psubd   xmm0,xmm4               ; xmm0=data7H
+
+        movdqa  xmm3,[rel PD_DESCALE_P1]        ; xmm3=[rel PD_DESCALE_P1]
+
+        paddd   xmm5,xmm3
+        paddd   xmm7,xmm3
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm7,DESCALE_P1
+        paddd   xmm2,xmm3
+        paddd   xmm0,xmm3
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm5,xmm7             ; xmm5=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm2,xmm0             ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+        movdqa  xmm4, XMMWORD [wk(4)]   ; xmm4=tmp11L
+        movdqa  xmm3, XMMWORD [wk(5)]   ; xmm3=tmp11H
+
+        movdqa  xmm7,xmm4
+        movdqa  xmm0,xmm3
+        paddd   xmm4,xmm1               ; xmm4=data1L
+        paddd   xmm3,xmm6               ; xmm3=data1H
+        psubd   xmm7,xmm1               ; xmm7=data6L
+        psubd   xmm0,xmm6               ; xmm0=data6H
+
+        movdqa  xmm1,[rel PD_DESCALE_P1]        ; xmm1=[rel PD_DESCALE_P1]
+
+        paddd   xmm4,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+        paddd   xmm7,xmm1
+        paddd   xmm0,xmm1
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm4,xmm3             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm7,xmm0             ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+        movdqa    xmm6,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm4             ; xmm5=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4             ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm2             ; xmm7=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm1,xmm2             ; xmm1=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(6)]   ; xmm3=tmp12L
+        movdqa  xmm0, XMMWORD [wk(7)]   ; xmm0=tmp12H
+        movdqa  xmm4, XMMWORD [wk(10)]  ; xmm4=tmp1L
+        movdqa  xmm2, XMMWORD [wk(11)]  ; xmm2=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(5)], xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm5,xmm3
+        movdqa  xmm6,xmm0
+        paddd   xmm3,xmm4               ; xmm3=data2L
+        paddd   xmm0,xmm2               ; xmm0=data2H
+        psubd   xmm5,xmm4               ; xmm5=data5L
+        psubd   xmm6,xmm2               ; xmm6=data5H
+
+        movdqa  xmm7,[rel PD_DESCALE_P1]        ; xmm7=[rel PD_DESCALE_P1]
+
+        paddd   xmm3,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm3,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+        paddd   xmm5,xmm7
+        paddd   xmm6,xmm7
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm3,xmm0             ; xmm3=data2=(20 21 22 23 24 25 26 27)
+        packssdw  xmm5,xmm6             ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+        movdqa  xmm1, XMMWORD [wk(2)]   ; xmm1=tmp13L
+        movdqa  xmm4, XMMWORD [wk(3)]   ; xmm4=tmp13H
+        movdqa  xmm2, XMMWORD [wk(8)]   ; xmm2=tmp0L
+        movdqa  xmm7, XMMWORD [wk(9)]   ; xmm7=tmp0H
+
+        movdqa  xmm0,xmm1
+        movdqa  xmm6,xmm4
+        paddd   xmm1,xmm2               ; xmm1=data3L
+        paddd   xmm4,xmm7               ; xmm4=data3H
+        psubd   xmm0,xmm2               ; xmm0=data4L
+        psubd   xmm6,xmm7               ; xmm6=data4H
+
+        movdqa  xmm2,[rel PD_DESCALE_P1]        ; xmm2=[rel PD_DESCALE_P1]
+
+        paddd   xmm1,xmm2
+        paddd   xmm4,xmm2
+        psrad   xmm1,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm0,xmm2
+        paddd   xmm6,xmm2
+        psrad   xmm0,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm1,xmm4             ; xmm1=data3=(30 31 32 33 34 35 36 37)
+        packssdw  xmm0,xmm6             ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(00 10 01 11 02 12 03 13)
+        movdqa  xmm2, XMMWORD [wk(1)]   ; xmm2=(04 14 05 15 06 16 07 17)
+
+        movdqa    xmm4,xmm3             ; transpose coefficients(phase 1)
+        punpcklwd xmm3,xmm1             ; xmm3=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm1             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm6,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm6,xmm5             ; xmm6=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm3             ; xmm7=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm1,xmm3             ; xmm1=(02 12 22 32 03 13 23 33)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm4             ; xmm2=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm5,xmm4             ; xmm5=(06 16 26 36 07 17 27 37)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=(60 70 61 71 62 72 63 73)
+        movdqa  xmm4, XMMWORD [wk(5)]   ; xmm4=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(7)], xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm2,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm2,xmm3             ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm4             ; xmm6=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm5,xmm4             ; xmm5=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm3,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm0            ; xmm7=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm3,xmm0            ; xmm3=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm4,xmm2            ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm0, XMMWORD [wk(6)]   ; xmm0=(04 14 24 34 05 15 25 35)
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(8)], xmm3   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm4   ; wk(9)=col3
+
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm3,xmm6            ; xmm3=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm4,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm5            ; xmm2=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm4,xmm5            ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm6,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm6,xmm2             ; xmm2=in6=z3
+        punpckhwd xmm5,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm2,xmm5
+        pmaddwd   xmm6,[rel PW_F130_F054]       ; xmm6=tmp3L
+        pmaddwd   xmm5,[rel PW_F130_F054]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=tmp2L
+        pmaddwd   xmm2,[rel PW_F054_MF130]      ; xmm2=tmp2H
+
+        movdqa    xmm3,xmm7
+        paddw     xmm7,xmm0             ; xmm7=in0+in4
+        psubw     xmm3,xmm0             ; xmm3=in0-in4
+
+        pxor      xmm4,xmm4
+        pxor      xmm0,xmm0
+        punpcklwd xmm4,xmm7             ; xmm4=tmp0L
+        punpckhwd xmm0,xmm7             ; xmm0=tmp0H
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+        psrad     xmm0,(16-CONST_BITS)  ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm6               ; xmm4=tmp10L
+        psubd   xmm7,xmm6               ; xmm7=tmp13L
+        movdqa  xmm6,xmm0
+        paddd   xmm0,xmm5               ; xmm0=tmp10H
+        psubd   xmm6,xmm5               ; xmm6=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm6   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm4,xmm4
+        punpcklwd xmm5,xmm3             ; xmm5=tmp1L
+        punpckhwd xmm4,xmm3             ; xmm4=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+        movdqa  xmm0,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm0,xmm1               ; xmm0=tmp12L
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm2               ; xmm4=tmp11H
+        psubd   xmm7,xmm2               ; xmm7=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm0   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm7   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm6, XMMWORD [wk(9)]   ; xmm6=col3
+        movdqa  xmm3, XMMWORD [wk(8)]   ; xmm3=col1
+        movdqa  xmm1, XMMWORD [wk(11)]  ; xmm1=col7
+        movdqa  xmm2, XMMWORD [wk(10)]  ; xmm2=col5
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm4,xmm3
+        paddw   xmm5,xmm1               ; xmm5=z3
+        paddw   xmm4,xmm2               ; xmm4=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm0,xmm5
+        movdqa    xmm7,xmm5
+        punpcklwd xmm0,xmm4
+        punpckhwd xmm7,xmm4
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm7
+        pmaddwd   xmm0,[rel PW_MF078_F117]      ; xmm0=z3L
+        pmaddwd   xmm7,[rel PW_MF078_F117]      ; xmm7=z3H
+        pmaddwd   xmm5,[rel PW_F117_F078]       ; xmm5=z4L
+        pmaddwd   xmm4,[rel PW_F117_F078]       ; xmm4=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm0,xmm1
+        movdqa    xmm7,xmm1
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm1,xmm0
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm0,[rel PW_MF060_MF089]     ; xmm0=tmp0L
+        pmaddwd   xmm7,[rel PW_MF060_MF089]     ; xmm7=tmp0H
+        pmaddwd   xmm1,[rel PW_MF089_F060]      ; xmm1=tmp3L
+        pmaddwd   xmm3,[rel PW_MF089_F060]      ; xmm3=tmp3H
+
+        paddd   xmm0, XMMWORD [wk(10)]  ; xmm0=tmp0L
+        paddd   xmm7, XMMWORD [wk(11)]  ; xmm7=tmp0H
+        paddd   xmm1,xmm5               ; xmm1=tmp3L
+        paddd   xmm3,xmm4               ; xmm3=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm0   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm7   ; wk(9)=tmp0H
+
+        movdqa    xmm0,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm0,xmm6
+        punpckhwd xmm7,xmm6
+        movdqa    xmm2,xmm0
+        movdqa    xmm6,xmm7
+        pmaddwd   xmm0,[rel PW_MF050_MF256]     ; xmm0=tmp1L
+        pmaddwd   xmm7,[rel PW_MF050_MF256]     ; xmm7=tmp1H
+        pmaddwd   xmm2,[rel PW_MF256_F050]      ; xmm2=tmp2L
+        pmaddwd   xmm6,[rel PW_MF256_F050]      ; xmm6=tmp2H
+
+        paddd   xmm0,xmm5               ; xmm0=tmp1L
+        paddd   xmm7,xmm4               ; xmm7=tmp1H
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm4, XMMWORD [wk(1)]   ; xmm4=tmp10H
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm7,xmm4
+        paddd   xmm5,xmm1               ; xmm5=data0L
+        paddd   xmm4,xmm3               ; xmm4=data0H
+        psubd   xmm0,xmm1               ; xmm0=data7L
+        psubd   xmm7,xmm3               ; xmm7=data7H
+
+        movdqa  xmm1,[rel PD_DESCALE_P2]        ; xmm1=[rel PD_DESCALE_P2]
+
+        paddd   xmm5,xmm1
+        paddd   xmm4,xmm1
+        psrad   xmm5,DESCALE_P2
+        psrad   xmm4,DESCALE_P2
+        paddd   xmm0,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm5,xmm4             ; xmm5=data0=(00 10 20 30 40 50 60 70)
+        packssdw  xmm0,xmm7             ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=tmp11L
+        movdqa  xmm1, XMMWORD [wk(5)]   ; xmm1=tmp11H
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm7,xmm1
+        paddd   xmm3,xmm2               ; xmm3=data1L
+        paddd   xmm1,xmm6               ; xmm1=data1H
+        psubd   xmm4,xmm2               ; xmm4=data6L
+        psubd   xmm7,xmm6               ; xmm7=data6H
+
+        movdqa  xmm2,[rel PD_DESCALE_P2]        ; xmm2=[rel PD_DESCALE_P2]
+
+        paddd   xmm3,xmm2
+        paddd   xmm1,xmm2
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm4,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm3,xmm1             ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        packssdw  xmm4,xmm7             ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+        packsswb  xmm5,xmm4             ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm0             ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm6, XMMWORD [wk(6)]   ; xmm6=tmp12L
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=tmp12H
+        movdqa  xmm1, XMMWORD [wk(10)]  ; xmm1=tmp1L
+        movdqa  xmm7, XMMWORD [wk(11)]  ; xmm7=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm4,xmm6
+        movdqa  xmm0,xmm2
+        paddd   xmm6,xmm1               ; xmm6=data2L
+        paddd   xmm2,xmm7               ; xmm2=data2H
+        psubd   xmm4,xmm1               ; xmm4=data5L
+        psubd   xmm0,xmm7               ; xmm0=data5H
+
+        movdqa  xmm5,[rel PD_DESCALE_P2]        ; xmm5=[rel PD_DESCALE_P2]
+
+        paddd   xmm6,xmm5
+        paddd   xmm2,xmm5
+        psrad   xmm6,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm4,xmm5
+        paddd   xmm0,xmm5
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        packssdw  xmm6,xmm2             ; xmm6=data2=(02 12 22 32 42 52 62 72)
+        packssdw  xmm4,xmm0             ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+        movdqa  xmm3, XMMWORD [wk(2)]   ; xmm3=tmp13L
+        movdqa  xmm1, XMMWORD [wk(3)]   ; xmm1=tmp13H
+        movdqa  xmm7, XMMWORD [wk(8)]   ; xmm7=tmp0L
+        movdqa  xmm5, XMMWORD [wk(9)]   ; xmm5=tmp0H
+
+        movdqa  xmm2,xmm3
+        movdqa  xmm0,xmm1
+        paddd   xmm3,xmm7               ; xmm3=data3L
+        paddd   xmm1,xmm5               ; xmm1=data3H
+        psubd   xmm2,xmm7               ; xmm2=data4L
+        psubd   xmm0,xmm5               ; xmm0=data4H
+
+        movdqa  xmm7,[rel PD_DESCALE_P2]        ; xmm7=[rel PD_DESCALE_P2]
+
+        paddd   xmm3,xmm7
+        paddd   xmm1,xmm7
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm2,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm2,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        movdqa    xmm5,[rel PB_CENTERJSAMP]     ; xmm5=[rel PB_CENTERJSAMP]
+
+        packssdw  xmm3,xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+        packssdw  xmm2,xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+        movdqa    xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        packsswb  xmm6,xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm3,xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm7,xmm5
+        paddb     xmm1,xmm5
+        paddb     xmm6,xmm5
+        paddb     xmm3,xmm5
+
+        movdqa    xmm0,xmm7     ; transpose coefficients(phase 1)
+        punpcklbw xmm7,xmm1     ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm1     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 1)
+        punpcklbw xmm6,xmm3     ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm2,xmm3     ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm7     ; transpose coefficients(phase 2)
+        punpcklwd xmm7,xmm6     ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm6     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm5,xmm2     ; transpose coefficients(phase 2)
+        punpcklwd xmm2,xmm0     ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm5,xmm0     ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm1,xmm7     ; transpose coefficients(phase 3)
+        punpckldq xmm7,xmm2     ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm1,xmm2     ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm3,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm5     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm3,xmm5     ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm6,xmm7,0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm1,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm2,xmm4,0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm5,xmm3,0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm7
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm1
+        mov     rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0
+        mov     rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm2
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm5
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctint-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctint-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,859 @@
+;
+; jidctint.asm - accurate integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_sse2) PRIVATE
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          12
+
+        align   16
+        global  EXTN(jsimd_idct_islow_sse2) PRIVATE
+
+EXTN(jsimd_idct_islow_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm5, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm5,PASS1_BITS
+
+        movdqa    xmm4,xmm5             ; xmm5=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm5,xmm5             ; xmm5=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm4,xmm4             ; xmm4=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm7,xmm5,0x00          ; xmm7=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm6,xmm5,0x55          ; xmm6=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm1,xmm5,0xAA          ; xmm1=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm5,xmm5,0xFF          ; xmm5=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm0,xmm4,0x00          ; xmm0=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm3,xmm4,0x55          ; xmm3=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm2,xmm4,0xAA          ; xmm2=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm4,xmm4,0xFF          ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(8)], xmm6   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm5   ; wk(9)=col3
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm4,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm4,xmm3             ; xmm3=in6=z3
+        punpckhwd xmm5,xmm3
+        movdqa    xmm1,xmm4
+        movdqa    xmm3,xmm5
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F130_F054)]       ; xmm4=tmp3L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F130_F054)]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=tmp2L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm3=tmp2H
+
+        movdqa    xmm6,xmm0
+        paddw     xmm0,xmm2             ; xmm0=in0+in4
+        psubw     xmm6,xmm2             ; xmm6=in0-in4
+
+        pxor      xmm7,xmm7
+        pxor      xmm2,xmm2
+        punpcklwd xmm7,xmm0             ; xmm7=tmp0L
+        punpckhwd xmm2,xmm0             ; xmm2=tmp0H
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+        psrad     xmm2,(16-CONST_BITS)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm4               ; xmm7=tmp10L
+        psubd   xmm0,xmm4               ; xmm0=tmp13L
+        movdqa  xmm4,xmm2
+        paddd   xmm2,xmm5               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm0   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm4   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm7,xmm7
+        punpcklwd xmm5,xmm6             ; xmm5=tmp1L
+        punpckhwd xmm7,xmm6             ; xmm7=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+        movdqa  xmm2,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm2,xmm1               ; xmm2=tmp12L
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm3               ; xmm7=tmp11H
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm7   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm0   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm7,xmm4
+        paddw   xmm5,xmm3               ; xmm5=z3
+        paddw   xmm7,xmm1               ; xmm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm2,xmm5
+        movdqa    xmm0,xmm5
+        punpcklwd xmm2,xmm7
+        punpckhwd xmm0,xmm7
+        movdqa    xmm5,xmm2
+        movdqa    xmm7,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm2=z3L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm0=z3H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F117_F078)]       ; xmm5=z4L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F117_F078)]       ; xmm7=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm2,xmm3
+        movdqa    xmm0,xmm3
+        punpcklwd xmm2,xmm4
+        punpckhwd xmm0,xmm4
+        movdqa    xmm3,xmm2
+        movdqa    xmm4,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm2=tmp0L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm0=tmp0H
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm3=tmp3L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm4=tmp3H
+
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp0L
+        paddd   xmm0, XMMWORD [wk(11)]  ; xmm0=tmp0H
+        paddd   xmm3,xmm5               ; xmm3=tmp3L
+        paddd   xmm4,xmm7               ; xmm4=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm2   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm0   ; wk(9)=tmp0H
+
+        movdqa    xmm2,xmm1
+        movdqa    xmm0,xmm1
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm0,xmm6
+        movdqa    xmm1,xmm2
+        movdqa    xmm6,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm2=tmp1L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm0=tmp1H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm1=tmp2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm6=tmp2H
+
+        paddd   xmm2,xmm5               ; xmm2=tmp1L
+        paddd   xmm0,xmm7               ; xmm0=tmp1H
+        paddd   xmm1, XMMWORD [wk(10)]  ; xmm1=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=tmp10H
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm0,xmm7
+        paddd   xmm5,xmm3               ; xmm5=data0L
+        paddd   xmm7,xmm4               ; xmm7=data0H
+        psubd   xmm2,xmm3               ; xmm2=data7L
+        psubd   xmm0,xmm4               ; xmm0=data7H
+
+        movdqa  xmm3,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm3=[PD_DESCALE_P1]
+
+        paddd   xmm5,xmm3
+        paddd   xmm7,xmm3
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm7,DESCALE_P1
+        paddd   xmm2,xmm3
+        paddd   xmm0,xmm3
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm5,xmm7             ; xmm5=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm2,xmm0             ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+        movdqa  xmm4, XMMWORD [wk(4)]   ; xmm4=tmp11L
+        movdqa  xmm3, XMMWORD [wk(5)]   ; xmm3=tmp11H
+
+        movdqa  xmm7,xmm4
+        movdqa  xmm0,xmm3
+        paddd   xmm4,xmm1               ; xmm4=data1L
+        paddd   xmm3,xmm6               ; xmm3=data1H
+        psubd   xmm7,xmm1               ; xmm7=data6L
+        psubd   xmm0,xmm6               ; xmm0=data6H
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm1=[PD_DESCALE_P1]
+
+        paddd   xmm4,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+        paddd   xmm7,xmm1
+        paddd   xmm0,xmm1
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm4,xmm3             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm7,xmm0             ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+        movdqa    xmm6,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm4             ; xmm5=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4             ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm2             ; xmm7=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm1,xmm2             ; xmm1=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(6)]   ; xmm3=tmp12L
+        movdqa  xmm0, XMMWORD [wk(7)]   ; xmm0=tmp12H
+        movdqa  xmm4, XMMWORD [wk(10)]  ; xmm4=tmp1L
+        movdqa  xmm2, XMMWORD [wk(11)]  ; xmm2=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(5)], xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm5,xmm3
+        movdqa  xmm6,xmm0
+        paddd   xmm3,xmm4               ; xmm3=data2L
+        paddd   xmm0,xmm2               ; xmm0=data2H
+        psubd   xmm5,xmm4               ; xmm5=data5L
+        psubd   xmm6,xmm2               ; xmm6=data5H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm7=[PD_DESCALE_P1]
+
+        paddd   xmm3,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm3,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+        paddd   xmm5,xmm7
+        paddd   xmm6,xmm7
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm3,xmm0             ; xmm3=data2=(20 21 22 23 24 25 26 27)
+        packssdw  xmm5,xmm6             ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+        movdqa  xmm1, XMMWORD [wk(2)]   ; xmm1=tmp13L
+        movdqa  xmm4, XMMWORD [wk(3)]   ; xmm4=tmp13H
+        movdqa  xmm2, XMMWORD [wk(8)]   ; xmm2=tmp0L
+        movdqa  xmm7, XMMWORD [wk(9)]   ; xmm7=tmp0H
+
+        movdqa  xmm0,xmm1
+        movdqa  xmm6,xmm4
+        paddd   xmm1,xmm2               ; xmm1=data3L
+        paddd   xmm4,xmm7               ; xmm4=data3H
+        psubd   xmm0,xmm2               ; xmm0=data4L
+        psubd   xmm6,xmm7               ; xmm6=data4H
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm2=[PD_DESCALE_P1]
+
+        paddd   xmm1,xmm2
+        paddd   xmm4,xmm2
+        psrad   xmm1,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm0,xmm2
+        paddd   xmm6,xmm2
+        psrad   xmm0,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm1,xmm4             ; xmm1=data3=(30 31 32 33 34 35 36 37)
+        packssdw  xmm0,xmm6             ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(00 10 01 11 02 12 03 13)
+        movdqa  xmm2, XMMWORD [wk(1)]   ; xmm2=(04 14 05 15 06 16 07 17)
+
+        movdqa    xmm4,xmm3             ; transpose coefficients(phase 1)
+        punpcklwd xmm3,xmm1             ; xmm3=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm1             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm6,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm6,xmm5             ; xmm6=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm3             ; xmm7=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm1,xmm3             ; xmm1=(02 12 22 32 03 13 23 33)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm4             ; xmm2=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm5,xmm4             ; xmm5=(06 16 26 36 07 17 27 37)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=(60 70 61 71 62 72 63 73)
+        movdqa  xmm4, XMMWORD [wk(5)]   ; xmm4=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(7)], xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm2,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm2,xmm3             ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm4             ; xmm6=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm5,xmm4             ; xmm5=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm3,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm0            ; xmm7=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm3,xmm0            ; xmm3=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm4,xmm2            ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm0, XMMWORD [wk(6)]   ; xmm0=(04 14 24 34 05 15 25 35)
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(8)], xmm3   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm4   ; wk(9)=col3
+
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm3,xmm6            ; xmm3=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm4,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm5            ; xmm2=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm4,xmm5            ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm6,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm6,xmm2             ; xmm2=in6=z3
+        punpckhwd xmm5,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm2,xmm5
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F130_F054)]       ; xmm6=tmp3L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F130_F054)]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=tmp2L
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm2=tmp2H
+
+        movdqa    xmm3,xmm7
+        paddw     xmm7,xmm0             ; xmm7=in0+in4
+        psubw     xmm3,xmm0             ; xmm3=in0-in4
+
+        pxor      xmm4,xmm4
+        pxor      xmm0,xmm0
+        punpcklwd xmm4,xmm7             ; xmm4=tmp0L
+        punpckhwd xmm0,xmm7             ; xmm0=tmp0H
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+        psrad     xmm0,(16-CONST_BITS)  ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm6               ; xmm4=tmp10L
+        psubd   xmm7,xmm6               ; xmm7=tmp13L
+        movdqa  xmm6,xmm0
+        paddd   xmm0,xmm5               ; xmm0=tmp10H
+        psubd   xmm6,xmm5               ; xmm6=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm6   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm4,xmm4
+        punpcklwd xmm5,xmm3             ; xmm5=tmp1L
+        punpckhwd xmm4,xmm3             ; xmm4=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+        movdqa  xmm0,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm0,xmm1               ; xmm0=tmp12L
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm2               ; xmm4=tmp11H
+        psubd   xmm7,xmm2               ; xmm7=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm0   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm7   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm6, XMMWORD [wk(9)]   ; xmm6=col3
+        movdqa  xmm3, XMMWORD [wk(8)]   ; xmm3=col1
+        movdqa  xmm1, XMMWORD [wk(11)]  ; xmm1=col7
+        movdqa  xmm2, XMMWORD [wk(10)]  ; xmm2=col5
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm4,xmm3
+        paddw   xmm5,xmm1               ; xmm5=z3
+        paddw   xmm4,xmm2               ; xmm4=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm0,xmm5
+        movdqa    xmm7,xmm5
+        punpcklwd xmm0,xmm4
+        punpckhwd xmm7,xmm4
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm0=z3L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm7=z3H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F117_F078)]       ; xmm5=z4L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F117_F078)]       ; xmm4=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm0,xmm1
+        movdqa    xmm7,xmm1
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm1,xmm0
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm0=tmp0L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm7=tmp0H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm1=tmp3L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm3=tmp3H
+
+        paddd   xmm0, XMMWORD [wk(10)]  ; xmm0=tmp0L
+        paddd   xmm7, XMMWORD [wk(11)]  ; xmm7=tmp0H
+        paddd   xmm1,xmm5               ; xmm1=tmp3L
+        paddd   xmm3,xmm4               ; xmm3=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm0   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm7   ; wk(9)=tmp0H
+
+        movdqa    xmm0,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm0,xmm6
+        punpckhwd xmm7,xmm6
+        movdqa    xmm2,xmm0
+        movdqa    xmm6,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm0=tmp1L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm7=tmp1H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm2=tmp2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm6=tmp2H
+
+        paddd   xmm0,xmm5               ; xmm0=tmp1L
+        paddd   xmm7,xmm4               ; xmm7=tmp1H
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm4, XMMWORD [wk(1)]   ; xmm4=tmp10H
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm7,xmm4
+        paddd   xmm5,xmm1               ; xmm5=data0L
+        paddd   xmm4,xmm3               ; xmm4=data0H
+        psubd   xmm0,xmm1               ; xmm0=data7L
+        psubd   xmm7,xmm3               ; xmm7=data7H
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm1=[PD_DESCALE_P2]
+
+        paddd   xmm5,xmm1
+        paddd   xmm4,xmm1
+        psrad   xmm5,DESCALE_P2
+        psrad   xmm4,DESCALE_P2
+        paddd   xmm0,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm5,xmm4             ; xmm5=data0=(00 10 20 30 40 50 60 70)
+        packssdw  xmm0,xmm7             ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=tmp11L
+        movdqa  xmm1, XMMWORD [wk(5)]   ; xmm1=tmp11H
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm7,xmm1
+        paddd   xmm3,xmm2               ; xmm3=data1L
+        paddd   xmm1,xmm6               ; xmm1=data1H
+        psubd   xmm4,xmm2               ; xmm4=data6L
+        psubd   xmm7,xmm6               ; xmm7=data6H
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm2=[PD_DESCALE_P2]
+
+        paddd   xmm3,xmm2
+        paddd   xmm1,xmm2
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm4,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm3,xmm1             ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        packssdw  xmm4,xmm7             ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+        packsswb  xmm5,xmm4             ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm0             ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm6, XMMWORD [wk(6)]   ; xmm6=tmp12L
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=tmp12H
+        movdqa  xmm1, XMMWORD [wk(10)]  ; xmm1=tmp1L
+        movdqa  xmm7, XMMWORD [wk(11)]  ; xmm7=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm4,xmm6
+        movdqa  xmm0,xmm2
+        paddd   xmm6,xmm1               ; xmm6=data2L
+        paddd   xmm2,xmm7               ; xmm2=data2H
+        psubd   xmm4,xmm1               ; xmm4=data5L
+        psubd   xmm0,xmm7               ; xmm0=data5H
+
+        movdqa  xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm5=[PD_DESCALE_P2]
+
+        paddd   xmm6,xmm5
+        paddd   xmm2,xmm5
+        psrad   xmm6,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm4,xmm5
+        paddd   xmm0,xmm5
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        packssdw  xmm6,xmm2             ; xmm6=data2=(02 12 22 32 42 52 62 72)
+        packssdw  xmm4,xmm0             ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+        movdqa  xmm3, XMMWORD [wk(2)]   ; xmm3=tmp13L
+        movdqa  xmm1, XMMWORD [wk(3)]   ; xmm1=tmp13H
+        movdqa  xmm7, XMMWORD [wk(8)]   ; xmm7=tmp0L
+        movdqa  xmm5, XMMWORD [wk(9)]   ; xmm5=tmp0H
+
+        movdqa  xmm2,xmm3
+        movdqa  xmm0,xmm1
+        paddd   xmm3,xmm7               ; xmm3=data3L
+        paddd   xmm1,xmm5               ; xmm1=data3H
+        psubd   xmm2,xmm7               ; xmm2=data4L
+        psubd   xmm0,xmm5               ; xmm0=data4H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm7=[PD_DESCALE_P2]
+
+        paddd   xmm3,xmm7
+        paddd   xmm1,xmm7
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm2,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm2,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        movdqa    xmm5,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm5=[PB_CENTERJSAMP]
+
+        packssdw  xmm3,xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+        packssdw  xmm2,xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+        movdqa    xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        packsswb  xmm6,xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm3,xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm7,xmm5
+        paddb     xmm1,xmm5
+        paddb     xmm6,xmm5
+        paddb     xmm3,xmm5
+
+        movdqa    xmm0,xmm7     ; transpose coefficients(phase 1)
+        punpcklbw xmm7,xmm1     ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm1     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 1)
+        punpcklbw xmm6,xmm3     ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm2,xmm3     ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm7     ; transpose coefficients(phase 2)
+        punpcklwd xmm7,xmm6     ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm6     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm5,xmm2     ; transpose coefficients(phase 2)
+        punpcklwd xmm2,xmm0     ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm5,xmm0     ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm1,xmm7     ; transpose coefficients(phase 3)
+        punpckldq xmm7,xmm2     ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm1,xmm2     ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm3,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm5     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm3,xmm5     ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm6,xmm7,0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm1,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm2,xmm4,0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm5,xmm3,0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm7
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm1
+        mov     edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0
+        mov     edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm2
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm5
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,706 @@
+;
+; jidctred.asm - reduced-size IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_mmx) PRIVATE
+
+EXTN(jconst_idct_red_mmx):
+
+PW_F184_MF076   times 2 dw  F_1_847,-F_0_765
+PW_F256_F089    times 2 dw  F_2_562, F_0_899
+PW_F106_MF217   times 2 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 2 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 2 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 2 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 2 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 2 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 2 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 2 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 2 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_mmx (void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_mmx) PRIVATE
+
+EXTN(jsimd_idct_4x4_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_4X4_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm0,mm1
+        packsswb mm0,mm0
+        movd    eax,mm0
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   mm0,PASS1_BITS
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movq      mm4,mm0
+        movq      mm5,mm0
+        punpcklwd mm4,mm1
+        punpckhwd mm5,mm1
+        movq      mm0,mm4
+        movq      mm1,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F256_F089)]        ; mm4=(tmp2L)
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F256_F089)]        ; mm5=(tmp2H)
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F106_MF217)]       ; mm0=(tmp0L)
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F106_MF217)]       ; mm1=(tmp0H)
+
+        movq      mm6,mm2
+        movq      mm7,mm2
+        punpcklwd mm6,mm3
+        punpckhwd mm7,mm3
+        movq      mm2,mm6
+        movq      mm3,mm7
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm6=(tmp2L)
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm7=(tmp2H)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F145_MF021)]       ; mm2=(tmp0L)
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F145_MF021)]       ; mm3=(tmp0H)
+
+        paddd   mm6,mm4                 ; mm6=tmp2L
+        paddd   mm7,mm5                 ; mm7=tmp2H
+        paddd   mm2,mm0                 ; mm2=tmp0L
+        paddd   mm3,mm1                 ; mm3=tmp0H
+
+        movq    MMWORD [wk(0)], mm2     ; wk(0)=tmp0L
+        movq    MMWORD [wk(1)], mm3     ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movq    mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm4, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm0, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      mm1,mm1
+        pxor      mm2,mm2
+        punpcklwd mm1,mm4               ; mm1=tmp0L
+        punpckhwd mm2,mm4               ; mm2=tmp0H
+        psrad     mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+        psrad     mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+        movq      mm3,mm5               ; mm5=in2=z2
+        punpcklwd mm5,mm0               ; mm0=in6=z3
+        punpckhwd mm3,mm0
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F184_MF076)]       ; mm5=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F184_MF076)]       ; mm3=tmp2H
+
+        movq    mm4,mm1
+        movq    mm0,mm2
+        paddd   mm1,mm5                 ; mm1=tmp10L
+        paddd   mm2,mm3                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp12L
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        ; -- Final output stage
+
+        movq    mm5,mm1
+        movq    mm3,mm2
+        paddd   mm1,mm6                 ; mm1=data0L
+        paddd   mm2,mm7                 ; mm2=data0H
+        psubd   mm5,mm6                 ; mm5=data3L
+        psubd   mm3,mm7                 ; mm3=data3H
+
+        movq    mm6,[GOTOFF(ebx,PD_DESCALE_P1_4)]       ; mm6=[PD_DESCALE_P1_4]
+
+        paddd   mm1,mm6
+        paddd   mm2,mm6
+        psrad   mm1,DESCALE_P1_4
+        psrad   mm2,DESCALE_P1_4
+        paddd   mm5,mm6
+        paddd   mm3,mm6
+        psrad   mm5,DESCALE_P1_4
+        psrad   mm3,DESCALE_P1_4
+
+        packssdw  mm1,mm2               ; mm1=data0=(00 01 02 03)
+        packssdw  mm5,mm3               ; mm5=data3=(30 31 32 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp0L
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp0H
+
+        movq    mm2,mm4
+        movq    mm3,mm0
+        paddd   mm4,mm7                 ; mm4=data1L
+        paddd   mm0,mm6                 ; mm0=data1H
+        psubd   mm2,mm7                 ; mm2=data2L
+        psubd   mm3,mm6                 ; mm3=data2H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1_4)]       ; mm7=[PD_DESCALE_P1_4]
+
+        paddd   mm4,mm7
+        paddd   mm0,mm7
+        psrad   mm4,DESCALE_P1_4
+        psrad   mm0,DESCALE_P1_4
+        paddd   mm2,mm7
+        paddd   mm3,mm7
+        psrad   mm2,DESCALE_P1_4
+        psrad   mm3,DESCALE_P1_4
+
+        packssdw  mm4,mm0               ; mm4=data1=(10 11 12 13)
+        packssdw  mm2,mm3               ; mm2=data2=(20 21 22 23)
+
+        movq      mm6,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm4               ; mm1=(00 10 01 11)
+        punpckhwd mm6,mm4               ; mm6=(02 12 03 13)
+        movq      mm7,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm5               ; mm2=(20 30 21 31)
+        punpckhwd mm7,mm5               ; mm7=(22 32 23 33)
+
+        movq      mm0,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm2               ; mm1=(00 10 20 30)
+        punpckhdq mm0,mm2               ; mm0=(01 11 21 31)
+        movq      mm3,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm7               ; mm6=(02 12 22 32)
+        punpckhdq mm3,mm7               ; mm3=(03 13 23 33)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_ISLOW_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq      mm4,mm0
+        movq      mm5,mm0
+        punpcklwd mm4,mm1
+        punpckhwd mm5,mm1
+        movq      mm0,mm4
+        movq      mm1,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F256_F089)]        ; mm4=(tmp2L)
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F256_F089)]        ; mm5=(tmp2H)
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F106_MF217)]       ; mm0=(tmp0L)
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F106_MF217)]       ; mm1=(tmp0H)
+
+        movq      mm6,mm2
+        movq      mm7,mm2
+        punpcklwd mm6,mm3
+        punpckhwd mm7,mm3
+        movq      mm2,mm6
+        movq      mm3,mm7
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm6=(tmp2L)
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm7=(tmp2H)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F145_MF021)]       ; mm2=(tmp0L)
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F145_MF021)]       ; mm3=(tmp0H)
+
+        paddd   mm6,mm4                 ; mm6=tmp2L
+        paddd   mm7,mm5                 ; mm7=tmp2H
+        paddd   mm2,mm0                 ; mm2=tmp0L
+        paddd   mm3,mm1                 ; mm3=tmp0H
+
+        movq    MMWORD [wk(0)], mm2     ; wk(0)=tmp0L
+        movq    MMWORD [wk(1)], mm3     ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movq    mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        pxor      mm1,mm1
+        pxor      mm2,mm2
+        punpcklwd mm1,mm4               ; mm1=tmp0L
+        punpckhwd mm2,mm4               ; mm2=tmp0H
+        psrad     mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+        psrad     mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+        movq      mm3,mm5               ; mm5=in2=z2
+        punpcklwd mm5,mm0               ; mm0=in6=z3
+        punpckhwd mm3,mm0
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F184_MF076)]       ; mm5=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F184_MF076)]       ; mm3=tmp2H
+
+        movq    mm4,mm1
+        movq    mm0,mm2
+        paddd   mm1,mm5                 ; mm1=tmp10L
+        paddd   mm2,mm3                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp12L
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        ; -- Final output stage
+
+        movq    mm5,mm1
+        movq    mm3,mm2
+        paddd   mm1,mm6                 ; mm1=data0L
+        paddd   mm2,mm7                 ; mm2=data0H
+        psubd   mm5,mm6                 ; mm5=data3L
+        psubd   mm3,mm7                 ; mm3=data3H
+
+        movq    mm6,[GOTOFF(ebx,PD_DESCALE_P2_4)]       ; mm6=[PD_DESCALE_P2_4]
+
+        paddd   mm1,mm6
+        paddd   mm2,mm6
+        psrad   mm1,DESCALE_P2_4
+        psrad   mm2,DESCALE_P2_4
+        paddd   mm5,mm6
+        paddd   mm3,mm6
+        psrad   mm5,DESCALE_P2_4
+        psrad   mm3,DESCALE_P2_4
+
+        packssdw  mm1,mm2               ; mm1=data0=(00 10 20 30)
+        packssdw  mm5,mm3               ; mm5=data3=(03 13 23 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp0L
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp0H
+
+        movq    mm2,mm4
+        movq    mm3,mm0
+        paddd   mm4,mm7                 ; mm4=data1L
+        paddd   mm0,mm6                 ; mm0=data1H
+        psubd   mm2,mm7                 ; mm2=data2L
+        psubd   mm3,mm6                 ; mm3=data2H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P2_4)]       ; mm7=[PD_DESCALE_P2_4]
+
+        paddd   mm4,mm7
+        paddd   mm0,mm7
+        psrad   mm4,DESCALE_P2_4
+        psrad   mm0,DESCALE_P2_4
+        paddd   mm2,mm7
+        paddd   mm3,mm7
+        psrad   mm2,DESCALE_P2_4
+        psrad   mm3,DESCALE_P2_4
+
+        packssdw  mm4,mm0               ; mm4=data1=(01 11 21 31)
+        packssdw  mm2,mm3               ; mm2=data2=(02 12 22 32)
+
+        movq      mm6,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm6=[PB_CENTERJSAMP]
+
+        packsswb  mm1,mm2               ; mm1=(00 10 20 30 02 12 22 32)
+        packsswb  mm4,mm5               ; mm4=(01 11 21 31 03 13 23 33)
+        paddb     mm1,mm6
+        paddb     mm4,mm6
+
+        movq      mm7,mm1               ; transpose coefficients(phase 1)
+        punpcklbw mm1,mm4               ; mm1=(00 01 10 11 20 21 30 31)
+        punpckhbw mm7,mm4               ; mm7=(02 03 12 13 22 23 32 33)
+
+        movq      mm0,mm1               ; transpose coefficients(phase 2)
+        punpcklwd mm1,mm7               ; mm1=(00 01 02 03 10 11 12 13)
+        punpckhwd mm0,mm7               ; mm0=(20 21 22 23 30 31 32 33)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movd    DWORD [edx+eax*SIZEOF_JSAMPLE], mm1
+        movd    DWORD [esi+eax*SIZEOF_JSAMPLE], mm0
+
+        psrlq   mm1,4*BYTE_BIT
+        psrlq   mm0,4*BYTE_BIT
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movd    DWORD [edx+eax*SIZEOF_JSAMPLE], mm1
+        movd    DWORD [esi+eax*SIZEOF_JSAMPLE], mm0
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_mmx (void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_mmx) PRIVATE
+
+EXTN(jsimd_idct_2x2_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     edx, POINTER [dct_table(ebp)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(ebp)]         ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm0=(10 11 ** 13), mm1=(30 31 ** 33)
+        ; mm2=(50 51 ** 53), mm3=(70 71 ** 73)
+
+        pcmpeqd   mm7,mm7
+        pslld     mm7,WORD_BIT          ; mm7={0x0000 0xFFFF 0x0000 0xFFFF}
+
+        movq      mm4,mm0               ; mm4=(10 11 ** 13)
+        movq      mm5,mm2               ; mm5=(50 51 ** 53)
+        punpcklwd mm4,mm1               ; mm4=(10 30 11 31)
+        punpcklwd mm5,mm3               ; mm5=(50 70 51 71)
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        psrld   mm0,WORD_BIT            ; mm0=(11 -- 13 --)
+        pand    mm1,mm7                 ; mm1=(-- 31 -- 33)
+        psrld   mm2,WORD_BIT            ; mm2=(51 -- 53 --)
+        pand    mm3,mm7                 ; mm3=(-- 71 -- 73)
+        por     mm0,mm1                 ; mm0=(11 31 13 33)
+        por     mm2,mm3                 ; mm2=(51 71 53 73)
+        pmaddwd mm0,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd mm2,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   mm4,mm5                 ; mm4=tmp0[col0 col1]
+
+        movq    mm6, MMWORD [MMBLOCK(1,1,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm6, MMWORD [MMBLOCK(1,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm3, MMWORD [MMBLOCK(5,1,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(7,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm3, MMWORD [MMBLOCK(5,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(7,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm6=(** 15 ** 17), mm1=(** 35 ** 37)
+        ; mm3=(** 55 ** 57), mm5=(** 75 ** 77)
+
+        psrld   mm6,WORD_BIT            ; mm6=(15 -- 17 --)
+        pand    mm1,mm7                 ; mm1=(-- 35 -- 37)
+        psrld   mm3,WORD_BIT            ; mm3=(55 -- 57 --)
+        pand    mm5,mm7                 ; mm5=(-- 75 -- 77)
+        por     mm6,mm1                 ; mm6=(15 35 17 37)
+        por     mm3,mm5                 ; mm3=(55 75 57 77)
+        pmaddwd mm6,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd mm3,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   mm0,mm2                 ; mm0=tmp0[col1 col3]
+        paddd   mm6,mm3                 ; mm6=tmp0[col5 col7]
+
+        ; -- Even part
+
+        movq    mm1, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(0,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm1, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(0,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm1=(00 01 ** 03), mm5=(** 05 ** 07)
+
+        movq    mm2,mm1                         ; mm2=(00 01 ** 03)
+        pslld   mm1,WORD_BIT                    ; mm1=(-- 00 -- **)
+        psrad   mm1,(WORD_BIT-CONST_BITS-2)     ; mm1=tmp10[col0 ****]
+
+        pand    mm2,mm7                         ; mm2=(-- 01 -- 03)
+        pand    mm5,mm7                         ; mm5=(-- 05 -- 07)
+        psrad   mm2,(WORD_BIT-CONST_BITS-2)     ; mm2=tmp10[col1 col3]
+        psrad   mm5,(WORD_BIT-CONST_BITS-2)     ; mm5=tmp10[col5 col7]
+
+        ; -- Final output stage
+
+        movq      mm3,mm1
+        paddd     mm1,mm4               ; mm1=data0[col0 ****]=(A0 **)
+        psubd     mm3,mm4               ; mm3=data1[col0 ****]=(B0 **)
+        punpckldq mm1,mm3               ; mm1=(A0 B0)
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1_2)]       ; mm7=[PD_DESCALE_P1_2]
+
+        movq    mm4,mm2
+        movq    mm3,mm5
+        paddd   mm2,mm0                 ; mm2=data0[col1 col3]=(A1 A3)
+        paddd   mm5,mm6                 ; mm5=data0[col5 col7]=(A5 A7)
+        psubd   mm4,mm0                 ; mm4=data1[col1 col3]=(B1 B3)
+        psubd   mm3,mm6                 ; mm3=data1[col5 col7]=(B5 B7)
+
+        paddd   mm1,mm7
+        psrad   mm1,DESCALE_P1_2
+
+        paddd   mm2,mm7
+        paddd   mm5,mm7
+        psrad   mm2,DESCALE_P1_2
+        psrad   mm5,DESCALE_P1_2
+        paddd   mm4,mm7
+        paddd   mm3,mm7
+        psrad   mm4,DESCALE_P1_2
+        psrad   mm3,DESCALE_P1_2
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     edi, JSAMPARRAY [output_buf(ebp)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(ebp)]
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  mm2,mm4               ; mm2=(A1 A3 B1 B3)
+        packssdw  mm5,mm3               ; mm5=(A5 A7 B5 B7)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd     mm2,mm5               ; mm2=tmp0[row0 row1]
+
+        ; -- Even part
+
+        pslld     mm1,(CONST_BITS+2)    ; mm1=tmp10[row0 row1]
+
+        ; -- Final output stage
+
+        movq      mm0,[GOTOFF(ebx,PD_DESCALE_P2_2)]     ; mm0=[PD_DESCALE_P2_2]
+
+        movq      mm6,mm1
+        paddd     mm1,mm2               ; mm1=data0[row0 row1]=(C0 C1)
+        psubd     mm6,mm2               ; mm6=data1[row0 row1]=(D0 D1)
+
+        paddd     mm1,mm0
+        paddd     mm6,mm0
+        psrad     mm1,DESCALE_P2_2
+        psrad     mm6,DESCALE_P2_2
+
+        movq      mm7,mm1               ; transpose coefficients
+        punpckldq mm1,mm6               ; mm1=(C0 D0)
+        punpckhdq mm7,mm6               ; mm7=(C1 D1)
+
+        packssdw  mm1,mm7               ; mm1=(C0 D0 C1 D1)
+        packsswb  mm1,mm1               ; mm1=(C0 D0 C1 D1 C0 D0 C1 D1)
+        paddb     mm1,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        movd    ecx,mm1
+        movd    ebx,mm1                 ; ebx=(C0 D0 C1 D1)
+        shr     ecx,2*BYTE_BIT          ; ecx=(C1 D1 -- --)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     WORD [edx+eax*SIZEOF_JSAMPLE], bx
+        mov     WORD [esi+eax*SIZEOF_JSAMPLE], cx
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,576 @@
+;
+; jidctred.asm - reduced-size IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_sse2) PRIVATE
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4 dw  F_1_847,-F_0_765
+PW_F256_F089    times 4 dw  F_2_562, F_0_899
+PW_F106_MF217   times 4 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 4 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 4 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 4 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 4 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 4 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 4 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 4 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 4 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_sse2) PRIVATE
+
+EXTN(jsimd_idct_4x4_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0,xmm1
+        packsswb xmm0,xmm0
+        packsswb xmm0,xmm0
+        movd    eax,xmm0
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm0,PASS1_BITS
+
+        movdqa    xmm3,xmm0     ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0     ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm3,xmm3     ; xmm3=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm1,xmm0,0x50  ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01)
+        pshufd  xmm0,xmm0,0xFA  ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03)
+        pshufd  xmm6,xmm3,0x50  ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05)
+        pshufd  xmm3,xmm3,0xFA  ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07)
+
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm5,xmm0
+        punpcklwd xmm4,xmm1
+        punpckhwd xmm5,xmm1
+        movdqa    xmm0,xmm4
+        movdqa    xmm1,xmm5
+        pmaddwd   xmm4,[rel PW_F256_F089]       ; xmm4=(tmp2L)
+        pmaddwd   xmm5,[rel PW_F256_F089]       ; xmm5=(tmp2H)
+        pmaddwd   xmm0,[rel PW_F106_MF217]      ; xmm0=(tmp0L)
+        pmaddwd   xmm1,[rel PW_F106_MF217]      ; xmm1=(tmp0H)
+
+        movdqa    xmm6,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm6,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm2,xmm6
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm6,[rel PW_MF060_MF050]     ; xmm6=(tmp2L)
+        pmaddwd   xmm7,[rel PW_MF060_MF050]     ; xmm7=(tmp2H)
+        pmaddwd   xmm2,[rel PW_F145_MF021]      ; xmm2=(tmp0L)
+        pmaddwd   xmm3,[rel PW_F145_MF021]      ; xmm3=(tmp0H)
+
+        paddd   xmm6,xmm4               ; xmm6=tmp2L
+        paddd   xmm7,xmm5               ; xmm7=tmp2H
+        paddd   xmm2,xmm0               ; xmm2=tmp0L
+        paddd   xmm3,xmm1               ; xmm3=tmp0H
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp0L
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      xmm1,xmm1
+        pxor      xmm2,xmm2
+        punpcklwd xmm1,xmm4             ; xmm1=tmp0L
+        punpckhwd xmm2,xmm4             ; xmm2=tmp0H
+        psrad     xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+        psrad     xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+        movdqa    xmm3,xmm5             ; xmm5=in2=z2
+        punpcklwd xmm5,xmm0             ; xmm0=in6=z3
+        punpckhwd xmm3,xmm0
+        pmaddwd   xmm5,[rel PW_F184_MF076]      ; xmm5=tmp2L
+        pmaddwd   xmm3,[rel PW_F184_MF076]      ; xmm3=tmp2H
+
+        movdqa  xmm4,xmm1
+        movdqa  xmm0,xmm2
+        paddd   xmm1,xmm5               ; xmm1=tmp10L
+        paddd   xmm2,xmm3               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp12L
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        ; -- Final output stage
+
+        movdqa  xmm5,xmm1
+        movdqa  xmm3,xmm2
+        paddd   xmm1,xmm6               ; xmm1=data0L
+        paddd   xmm2,xmm7               ; xmm2=data0H
+        psubd   xmm5,xmm6               ; xmm5=data3L
+        psubd   xmm3,xmm7               ; xmm3=data3H
+
+        movdqa  xmm6,[rel PD_DESCALE_P1_4]      ; xmm6=[rel PD_DESCALE_P1_4]
+
+        paddd   xmm1,xmm6
+        paddd   xmm2,xmm6
+        psrad   xmm1,DESCALE_P1_4
+        psrad   xmm2,DESCALE_P1_4
+        paddd   xmm5,xmm6
+        paddd   xmm3,xmm6
+        psrad   xmm5,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm1,xmm2             ; xmm1=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm5,xmm3             ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp0L
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp0H
+
+        movdqa  xmm2,xmm4
+        movdqa  xmm3,xmm0
+        paddd   xmm4,xmm7               ; xmm4=data1L
+        paddd   xmm0,xmm6               ; xmm0=data1H
+        psubd   xmm2,xmm7               ; xmm2=data2L
+        psubd   xmm3,xmm6               ; xmm3=data2H
+
+        movdqa  xmm7,[rel PD_DESCALE_P1_4]      ; xmm7=[rel PD_DESCALE_P1_4]
+
+        paddd   xmm4,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm4,DESCALE_P1_4
+        psrad   xmm0,DESCALE_P1_4
+        paddd   xmm2,xmm7
+        paddd   xmm3,xmm7
+        psrad   xmm2,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm4,xmm0             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm2,xmm3             ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+        movdqa    xmm6,xmm1     ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm4     ; xmm1=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4     ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm7,xmm2     ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm5     ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm7,xmm5     ; xmm7=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm2     ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm0,xmm2     ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33)
+        movdqa    xmm3,xmm6     ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7     ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm3,xmm7     ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        pxor      xmm4,xmm4
+        punpcklwd xmm4,xmm1             ; xmm4=tmp0
+        psrad     xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+        ; -- Odd part
+
+        punpckhwd xmm1,xmm0
+        punpckhwd xmm6,xmm3
+        movdqa    xmm5,xmm1
+        movdqa    xmm2,xmm6
+        pmaddwd   xmm1,[rel PW_F256_F089]       ; xmm1=(tmp2)
+        pmaddwd   xmm6,[rel PW_MF060_MF050]     ; xmm6=(tmp2)
+        pmaddwd   xmm5,[rel PW_F106_MF217]      ; xmm5=(tmp0)
+        pmaddwd   xmm2,[rel PW_F145_MF021]      ; xmm2=(tmp0)
+
+        paddd     xmm6,xmm1             ; xmm6=tmp2
+        paddd     xmm2,xmm5             ; xmm2=tmp0
+
+        ; -- Even part
+
+        punpcklwd xmm0,xmm3
+        pmaddwd   xmm0,[rel PW_F184_MF076]      ; xmm0=tmp2
+
+        movdqa    xmm7,xmm4
+        paddd     xmm4,xmm0             ; xmm4=tmp10
+        psubd     xmm7,xmm0             ; xmm7=tmp12
+
+        ; -- Final output stage
+
+        movdqa  xmm1,[rel PD_DESCALE_P2_4]      ; xmm1=[rel PD_DESCALE_P2_4]
+
+        movdqa  xmm5,xmm4
+        movdqa  xmm3,xmm7
+        paddd   xmm4,xmm6               ; xmm4=data0=(00 10 20 30)
+        paddd   xmm7,xmm2               ; xmm7=data1=(01 11 21 31)
+        psubd   xmm5,xmm6               ; xmm5=data3=(03 13 23 33)
+        psubd   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+
+        paddd   xmm4,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm4,DESCALE_P2_4
+        psrad   xmm7,DESCALE_P2_4
+        paddd   xmm5,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm5,DESCALE_P2_4
+        psrad   xmm3,DESCALE_P2_4
+
+        packssdw  xmm4,xmm3             ; xmm4=(00 10 20 30 02 12 22 32)
+        packssdw  xmm7,xmm5             ; xmm7=(01 11 21 31 03 13 23 33)
+
+        movdqa    xmm0,xmm4             ; transpose coefficients(phase 1)
+        punpcklwd xmm4,xmm7             ; xmm4=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm0,xmm7             ; xmm0=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm6,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm6,xmm0             ; xmm6=(20 21 22 23 30 31 32 33)
+
+        packsswb  xmm4,xmm6             ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+        paddb     xmm4,[rel PB_CENTERJSAMP]
+
+        pshufd    xmm2,xmm4,0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+        pshufd    xmm1,xmm4,0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+        pshufd    xmm3,xmm4,0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movd    XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2
+        mov     rdx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1
+        movd    XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_sse2) PRIVATE
+
+EXTN(jsimd_idct_2x2_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+        ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+        pcmpeqd   xmm7,xmm7
+        pslld     xmm7,WORD_BIT         ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+        movdqa    xmm4,xmm0             ; xmm4=(10 11 ** 13 ** 15 ** 17)
+        movdqa    xmm5,xmm2             ; xmm5=(50 51 ** 53 ** 55 ** 57)
+        punpcklwd xmm4,xmm1             ; xmm4=(10 30 11 31 ** ** 13 33)
+        punpcklwd xmm5,xmm3             ; xmm5=(50 70 51 71 ** ** 53 73)
+        pmaddwd   xmm4,[rel PW_F362_MF127]
+        pmaddwd   xmm5,[rel PW_F085_MF072]
+
+        psrld   xmm0,WORD_BIT           ; xmm0=(11 -- 13 -- 15 -- 17 --)
+        pand    xmm1,xmm7               ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+        psrld   xmm2,WORD_BIT           ; xmm2=(51 -- 53 -- 55 -- 57 --)
+        pand    xmm3,xmm7               ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+        por     xmm0,xmm1               ; xmm0=(11 31 13 33 15 35 17 37)
+        por     xmm2,xmm3               ; xmm2=(51 71 53 73 55 75 57 77)
+        pmaddwd xmm0,[rel PW_F362_MF127]
+        pmaddwd xmm2,[rel PW_F085_MF072]
+
+        paddd   xmm4,xmm5               ; xmm4=tmp0[col0 col1 **** col3]
+        paddd   xmm0,xmm2               ; xmm0=tmp0[col1 col3 col5 col7]
+
+        ; -- Even part
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+        movdqa  xmm1,xmm6               ; xmm1=(00 01 ** 03 ** 05 ** 07)
+        pslld   xmm6,WORD_BIT           ; xmm6=(-- 00 -- ** -- ** -- **)
+        pand    xmm1,xmm7               ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+        psrad   xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****]
+        psrad   xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7]
+
+        ; -- Final output stage
+
+        movdqa  xmm3,xmm6
+        movdqa  xmm5,xmm1
+        paddd   xmm6,xmm4       ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **)
+        paddd   xmm1,xmm0       ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7)
+        psubd   xmm3,xmm4       ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **)
+        psubd   xmm5,xmm0       ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7)
+
+        movdqa  xmm2,[rel PD_DESCALE_P1_2]      ; xmm2=[rel PD_DESCALE_P1_2]
+
+        punpckldq  xmm6,xmm3            ; xmm6=(A0 B0 ** **)
+
+        movdqa     xmm7,xmm1
+        punpcklqdq xmm1,xmm5            ; xmm1=(A1 A3 B1 B3)
+        punpckhqdq xmm7,xmm5            ; xmm7=(A5 A7 B5 B7)
+
+        paddd   xmm6,xmm2
+        psrad   xmm6,DESCALE_P1_2
+
+        paddd   xmm1,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm1,DESCALE_P1_2
+        psrad   xmm7,DESCALE_P1_2
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  xmm1,xmm1             ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+        packssdw  xmm7,xmm7             ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+        pmaddwd   xmm1,[rel PW_F362_MF127]
+        pmaddwd   xmm7,[rel PW_F085_MF072]
+
+        paddd     xmm1,xmm7             ; xmm1=tmp0[row0 row1 row0 row1]
+
+        ; -- Even part
+
+        pslld     xmm6,(CONST_BITS+2)   ; xmm6=tmp10[row0 row1 **** ****]
+
+        ; -- Final output stage
+
+        movdqa    xmm4,xmm6
+        paddd     xmm6,xmm1     ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **)
+        psubd     xmm4,xmm1     ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **)
+
+        punpckldq xmm6,xmm4     ; xmm6=(C0 D0 C1 D1)
+
+        paddd     xmm6,[rel PD_DESCALE_P2_2]
+        psrad     xmm6,DESCALE_P2_2
+
+        packssdw  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+        packsswb  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+        paddb     xmm6,[rel PB_CENTERJSAMP]
+
+        pextrw  ebx,xmm6,0x00           ; ebx=(C0 D0 -- --)
+        pextrw  ecx,xmm6,0x01           ; ecx=(C1 D1 -- --)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     WORD [rdx+rax*SIZEOF_JSAMPLE], bx
+        mov     WORD [rsi+rax*SIZEOF_JSAMPLE], cx
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jidctred-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jidctred-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,594 @@
+;
+; jidctred.asm - reduced-size IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_sse2) PRIVATE
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4 dw  F_1_847,-F_0_765
+PW_F256_F089    times 4 dw  F_2_562, F_0_899
+PW_F106_MF217   times 4 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 4 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 4 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 4 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 4 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 4 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 4 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 4 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 4 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_sse2) PRIVATE
+
+EXTN(jsimd_idct_4x4_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm0,xmm1
+        packsswb xmm0,xmm0
+        packsswb xmm0,xmm0
+        movd    eax,xmm0
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm0,PASS1_BITS
+
+        movdqa    xmm3,xmm0     ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0     ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm3,xmm3     ; xmm3=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm1,xmm0,0x50  ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01)
+        pshufd  xmm0,xmm0,0xFA  ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03)
+        pshufd  xmm6,xmm3,0x50  ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05)
+        pshufd  xmm3,xmm3,0xFA  ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07)
+
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm5,xmm0
+        punpcklwd xmm4,xmm1
+        punpckhwd xmm5,xmm1
+        movdqa    xmm0,xmm4
+        movdqa    xmm1,xmm5
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F256_F089)]       ; xmm4=(tmp2L)
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F256_F089)]       ; xmm5=(tmp2H)
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm0=(tmp0L)
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm1=(tmp0H)
+
+        movdqa    xmm6,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm6,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm2,xmm6
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm6=(tmp2L)
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm7=(tmp2H)
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm2=(tmp0L)
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm3=(tmp0H)
+
+        paddd   xmm6,xmm4               ; xmm6=tmp2L
+        paddd   xmm7,xmm5               ; xmm7=tmp2H
+        paddd   xmm2,xmm0               ; xmm2=tmp0L
+        paddd   xmm3,xmm1               ; xmm3=tmp0H
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp0L
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      xmm1,xmm1
+        pxor      xmm2,xmm2
+        punpcklwd xmm1,xmm4             ; xmm1=tmp0L
+        punpckhwd xmm2,xmm4             ; xmm2=tmp0H
+        psrad     xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+        psrad     xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+        movdqa    xmm3,xmm5             ; xmm5=in2=z2
+        punpcklwd xmm5,xmm0             ; xmm0=in6=z3
+        punpckhwd xmm3,xmm0
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm5=tmp2L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm3=tmp2H
+
+        movdqa  xmm4,xmm1
+        movdqa  xmm0,xmm2
+        paddd   xmm1,xmm5               ; xmm1=tmp10L
+        paddd   xmm2,xmm3               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp12L
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        ; -- Final output stage
+
+        movdqa  xmm5,xmm1
+        movdqa  xmm3,xmm2
+        paddd   xmm1,xmm6               ; xmm1=data0L
+        paddd   xmm2,xmm7               ; xmm2=data0H
+        psubd   xmm5,xmm6               ; xmm5=data3L
+        psubd   xmm3,xmm7               ; xmm3=data3H
+
+        movdqa  xmm6,[GOTOFF(ebx,PD_DESCALE_P1_4)]      ; xmm6=[PD_DESCALE_P1_4]
+
+        paddd   xmm1,xmm6
+        paddd   xmm2,xmm6
+        psrad   xmm1,DESCALE_P1_4
+        psrad   xmm2,DESCALE_P1_4
+        paddd   xmm5,xmm6
+        paddd   xmm3,xmm6
+        psrad   xmm5,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm1,xmm2             ; xmm1=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm5,xmm3             ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp0L
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp0H
+
+        movdqa  xmm2,xmm4
+        movdqa  xmm3,xmm0
+        paddd   xmm4,xmm7               ; xmm4=data1L
+        paddd   xmm0,xmm6               ; xmm0=data1H
+        psubd   xmm2,xmm7               ; xmm2=data2L
+        psubd   xmm3,xmm6               ; xmm3=data2H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P1_4)]      ; xmm7=[PD_DESCALE_P1_4]
+
+        paddd   xmm4,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm4,DESCALE_P1_4
+        psrad   xmm0,DESCALE_P1_4
+        paddd   xmm2,xmm7
+        paddd   xmm3,xmm7
+        psrad   xmm2,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm4,xmm0             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm2,xmm3             ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+        movdqa    xmm6,xmm1     ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm4     ; xmm1=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4     ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm7,xmm2     ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm5     ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm7,xmm5     ; xmm7=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm2     ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm0,xmm2     ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33)
+        movdqa    xmm3,xmm6     ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7     ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm3,xmm7     ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        pxor      xmm4,xmm4
+        punpcklwd xmm4,xmm1             ; xmm4=tmp0
+        psrad     xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+        ; -- Odd part
+
+        punpckhwd xmm1,xmm0
+        punpckhwd xmm6,xmm3
+        movdqa    xmm5,xmm1
+        movdqa    xmm2,xmm6
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F256_F089)]       ; xmm1=(tmp2)
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm6=(tmp2)
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm5=(tmp0)
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm2=(tmp0)
+
+        paddd     xmm6,xmm1             ; xmm6=tmp2
+        paddd     xmm2,xmm5             ; xmm2=tmp0
+
+        ; -- Even part
+
+        punpcklwd xmm0,xmm3
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm0=tmp2
+
+        movdqa    xmm7,xmm4
+        paddd     xmm4,xmm0             ; xmm4=tmp10
+        psubd     xmm7,xmm0             ; xmm7=tmp12
+
+        ; -- Final output stage
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P2_4)]      ; xmm1=[PD_DESCALE_P2_4]
+
+        movdqa  xmm5,xmm4
+        movdqa  xmm3,xmm7
+        paddd   xmm4,xmm6               ; xmm4=data0=(00 10 20 30)
+        paddd   xmm7,xmm2               ; xmm7=data1=(01 11 21 31)
+        psubd   xmm5,xmm6               ; xmm5=data3=(03 13 23 33)
+        psubd   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+
+        paddd   xmm4,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm4,DESCALE_P2_4
+        psrad   xmm7,DESCALE_P2_4
+        paddd   xmm5,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm5,DESCALE_P2_4
+        psrad   xmm3,DESCALE_P2_4
+
+        packssdw  xmm4,xmm3             ; xmm4=(00 10 20 30 02 12 22 32)
+        packssdw  xmm7,xmm5             ; xmm7=(01 11 21 31 03 13 23 33)
+
+        movdqa    xmm0,xmm4             ; transpose coefficients(phase 1)
+        punpcklwd xmm4,xmm7             ; xmm4=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm0,xmm7             ; xmm0=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm6,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm6,xmm0             ; xmm6=(20 21 22 23 30 31 32 33)
+
+        packsswb  xmm4,xmm6             ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+        paddb     xmm4,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        pshufd    xmm2,xmm4,0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+        pshufd    xmm1,xmm4,0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+        pshufd    xmm3,xmm4,0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movd    XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm2
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm1
+        movd    XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_sse2) PRIVATE
+
+EXTN(jsimd_idct_2x2_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     edx, POINTER [dct_table(ebp)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(ebp)]         ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+        ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+        pcmpeqd   xmm7,xmm7
+        pslld     xmm7,WORD_BIT         ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+        movdqa    xmm4,xmm0             ; xmm4=(10 11 ** 13 ** 15 ** 17)
+        movdqa    xmm5,xmm2             ; xmm5=(50 51 ** 53 ** 55 ** 57)
+        punpcklwd xmm4,xmm1             ; xmm4=(10 30 11 31 ** ** 13 33)
+        punpcklwd xmm5,xmm3             ; xmm5=(50 70 51 71 ** ** 53 73)
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        psrld   xmm0,WORD_BIT           ; xmm0=(11 -- 13 -- 15 -- 17 --)
+        pand    xmm1,xmm7               ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+        psrld   xmm2,WORD_BIT           ; xmm2=(51 -- 53 -- 55 -- 57 --)
+        pand    xmm3,xmm7               ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+        por     xmm0,xmm1               ; xmm0=(11 31 13 33 15 35 17 37)
+        por     xmm2,xmm3               ; xmm2=(51 71 53 73 55 75 57 77)
+        pmaddwd xmm0,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd xmm2,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   xmm4,xmm5               ; xmm4=tmp0[col0 col1 **** col3]
+        paddd   xmm0,xmm2               ; xmm0=tmp0[col1 col3 col5 col7]
+
+        ; -- Even part
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+        movdqa  xmm1,xmm6               ; xmm1=(00 01 ** 03 ** 05 ** 07)
+        pslld   xmm6,WORD_BIT           ; xmm6=(-- 00 -- ** -- ** -- **)
+        pand    xmm1,xmm7               ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+        psrad   xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****]
+        psrad   xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7]
+
+        ; -- Final output stage
+
+        movdqa  xmm3,xmm6
+        movdqa  xmm5,xmm1
+        paddd   xmm6,xmm4       ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **)
+        paddd   xmm1,xmm0       ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7)
+        psubd   xmm3,xmm4       ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **)
+        psubd   xmm5,xmm0       ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7)
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P1_2)]      ; xmm2=[PD_DESCALE_P1_2]
+
+        punpckldq  xmm6,xmm3            ; xmm6=(A0 B0 ** **)
+
+        movdqa     xmm7,xmm1
+        punpcklqdq xmm1,xmm5            ; xmm1=(A1 A3 B1 B3)
+        punpckhqdq xmm7,xmm5            ; xmm7=(A5 A7 B5 B7)
+
+        paddd   xmm6,xmm2
+        psrad   xmm6,DESCALE_P1_2
+
+        paddd   xmm1,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm1,DESCALE_P1_2
+        psrad   xmm7,DESCALE_P1_2
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     edi, JSAMPARRAY [output_buf(ebp)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(ebp)]
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  xmm1,xmm1             ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+        packssdw  xmm7,xmm7             ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd     xmm1,xmm7             ; xmm1=tmp0[row0 row1 row0 row1]
+
+        ; -- Even part
+
+        pslld     xmm6,(CONST_BITS+2)   ; xmm6=tmp10[row0 row1 **** ****]
+
+        ; -- Final output stage
+
+        movdqa    xmm4,xmm6
+        paddd     xmm6,xmm1     ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **)
+        psubd     xmm4,xmm1     ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **)
+
+        punpckldq xmm6,xmm4     ; xmm6=(C0 D0 C1 D1)
+
+        paddd     xmm6,[GOTOFF(ebx,PD_DESCALE_P2_2)]
+        psrad     xmm6,DESCALE_P2_2
+
+        packssdw  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+        packsswb  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+        paddb     xmm6,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        pextrw  ebx,xmm6,0x00           ; ebx=(C0 D0 -- --)
+        pextrw  ecx,xmm6,0x01           ; ecx=(C1 D1 -- --)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     WORD [edx+eax*SIZEOF_JSAMPLE], bx
+        mov     WORD [esi+eax*SIZEOF_JSAMPLE], cx
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jpeg_nbits_table.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jpeg_nbits_table.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jpeg_nbits_table.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jpeg_nbits_table.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4097 @@
+jpeg_nbits_table db  \
+   0,  1,  2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,  \
+   5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  \
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  \
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-3dn.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,233 @@
+;
+; jquant.asm - sample data conversion and quantization (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_3dnow (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                             FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_3dnow) PRIVATE
+
+EXTN(jsimd_convsamp_float_3dnow):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  mm7,mm7
+        psllw    mm7,7
+        packsswb mm7,mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   mm0,mm7                         ; mm0=(01234567)
+        psubb   mm1,mm7                         ; mm1=(89ABCDEF)
+
+        punpcklbw mm2,mm0                       ; mm2=(*0*1*2*3)
+        punpckhbw mm0,mm0                       ; mm0=(*4*5*6*7)
+        punpcklbw mm3,mm1                       ; mm3=(*8*9*A*B)
+        punpckhbw mm1,mm1                       ; mm1=(*C*D*E*F)
+
+        punpcklwd mm4,mm2                       ; mm4=(***0***1)
+        punpckhwd mm2,mm2                       ; mm2=(***2***3)
+        punpcklwd mm5,mm0                       ; mm5=(***4***5)
+        punpckhwd mm0,mm0                       ; mm0=(***6***7)
+
+        psrad   mm4,(DWORD_BIT-BYTE_BIT)        ; mm4=(01)
+        psrad   mm2,(DWORD_BIT-BYTE_BIT)        ; mm2=(23)
+        pi2fd   mm4,mm4
+        pi2fd   mm2,mm2
+        psrad   mm5,(DWORD_BIT-BYTE_BIT)        ; mm5=(45)
+        psrad   mm0,(DWORD_BIT-BYTE_BIT)        ; mm0=(67)
+        pi2fd   mm5,mm5
+        pi2fd   mm0,mm0
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+
+        punpcklwd mm6,mm3                       ; mm6=(***8***9)
+        punpckhwd mm3,mm3                       ; mm3=(***A***B)
+        punpcklwd mm4,mm1                       ; mm4=(***C***D)
+        punpckhwd mm1,mm1                       ; mm1=(***E***F)
+
+        psrad   mm6,(DWORD_BIT-BYTE_BIT)        ; mm6=(89)
+        psrad   mm3,(DWORD_BIT-BYTE_BIT)        ; mm3=(AB)
+        pi2fd   mm6,mm6
+        pi2fd   mm3,mm3
+        psrad   mm4,(DWORD_BIT-BYTE_BIT)        ; mm4=(CD)
+        psrad   mm1,(DWORD_BIT-BYTE_BIT)        ; mm1=(EF)
+        pi2fd   mm4,mm4
+        pi2fd   mm1,mm1
+
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm3
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .convloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_3dnow (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                             FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_3dnow) PRIVATE
+
+EXTN(jsimd_quantize_float_3dnow):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov       eax, 0x4B400000       ; (float)0x00C00000 (rndint_magic)
+        movd      mm7,eax
+        punpckldq mm7,mm7               ; mm7={12582912.0F 12582912.0F}
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,2,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(0,3,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm3, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)]
+
+        pfadd   mm0,mm7                 ; mm0=(00 ** 01 **)
+        pfadd   mm1,mm7                 ; mm1=(02 ** 03 **)
+        pfadd   mm2,mm7                 ; mm0=(04 ** 05 **)
+        pfadd   mm3,mm7                 ; mm1=(06 ** 07 **)
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm1               ; mm0=(00 02 ** **)
+        punpckhwd mm4,mm1               ; mm4=(01 03 ** **)
+        movq      mm5,mm2
+        punpcklwd mm2,mm3               ; mm2=(04 06 ** **)
+        punpckhwd mm5,mm3               ; mm5=(05 07 ** **)
+
+        punpcklwd mm0,mm4               ; mm0=(00 01 02 03)
+        punpcklwd mm2,mm5               ; mm2=(04 05 06 07)
+
+        movq    mm6, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm6, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm1, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,2,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm4, MMWORD [MMBLOCK(1,3,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm3, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm4, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)]
+
+        pfadd   mm6,mm7                 ; mm0=(10 ** 11 **)
+        pfadd   mm1,mm7                 ; mm4=(12 ** 13 **)
+        pfadd   mm3,mm7                 ; mm0=(14 ** 15 **)
+        pfadd   mm4,mm7                 ; mm4=(16 ** 17 **)
+
+        movq      mm5,mm6
+        punpcklwd mm6,mm1               ; mm6=(10 12 ** **)
+        punpckhwd mm5,mm1               ; mm5=(11 13 ** **)
+        movq      mm1,mm3
+        punpcklwd mm3,mm4               ; mm3=(14 16 ** **)
+        punpckhwd mm1,mm4               ; mm1=(15 17 ** **)
+
+        punpcklwd mm6,mm5               ; mm6=(10 11 12 13)
+        punpcklwd mm3,mm1               ; mm3=(14 15 16 17)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     near .quantloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquantf-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquantf-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquantf-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquantf-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,158 @@
+;
+; jquantf.asm - sample data conversion and quantization (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                            FAST_FLOAT *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11 = JDIMENSION start_col
+; r12 = FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        pcmpeqw  xmm7,xmm7
+        psllw    xmm7,7
+        packsswb xmm7,xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov rsi, r10
+        mov     eax, r11d
+        mov rdi, r12
+        mov     rcx, DCTSIZE/2
+.convloop:
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]       ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]
+        movq    xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]
+
+        psubb   xmm0,xmm7                       ; xmm0=(01234567)
+        psubb   xmm1,xmm7                       ; xmm1=(89ABCDEF)
+
+        punpcklbw xmm0,xmm0                     ; xmm0=(*0*1*2*3*4*5*6*7)
+        punpcklbw xmm1,xmm1                     ; xmm1=(*8*9*A*B*C*D*E*F)
+
+        punpcklwd xmm2,xmm0                     ; xmm2=(***0***1***2***3)
+        punpckhwd xmm0,xmm0                     ; xmm0=(***4***5***6***7)
+        punpcklwd xmm3,xmm1                     ; xmm3=(***8***9***A***B)
+        punpckhwd xmm1,xmm1                     ; xmm1=(***C***D***E***F)
+
+        psrad     xmm2,(DWORD_BIT-BYTE_BIT)     ; xmm2=(0123)
+        psrad     xmm0,(DWORD_BIT-BYTE_BIT)     ; xmm0=(4567)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=(0123)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=(4567)
+        psrad     xmm3,(DWORD_BIT-BYTE_BIT)     ; xmm3=(89AB)
+        psrad     xmm1,(DWORD_BIT-BYTE_BIT)     ; xmm1=(CDEF)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=(89AB)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1
+
+        add     rsi, byte 2*SIZEOF_JSAMPROW
+        add     rdi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     short .convloop
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                         FAST_FLOAT *workspace);
+;
+
+; r10 = JCOEFPTR coef_block
+; r11 = FAST_FLOAT *divisors
+; r12 = FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse2) PRIVATE
+
+EXTN(jsimd_quantize_float_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov rsi, r12
+        mov rdx, r11
+        mov rdi, r10
+        mov     rax, DCTSIZE2/16
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,rsi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,rsi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        cvtps2dq xmm0,xmm0
+        cvtps2dq xmm1,xmm1
+        cvtps2dq xmm2,xmm2
+        cvtps2dq xmm3,xmm3
+
+        packssdw xmm0,xmm1
+        packssdw xmm2,xmm3
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_JCOEF)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_JCOEF)], xmm2
+
+        add     rsi, byte 16*SIZEOF_FAST_FLOAT
+        add     rdx, byte 16*SIZEOF_FAST_FLOAT
+        add     rdi, byte 16*SIZEOF_JCOEF
+        dec     rax
+        jnz     short .quantloop
+
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquantf-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquantf-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquantf-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquantf-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,171 @@
+;
+; jquantf.asm - sample data conversion and quantization (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                            FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  xmm7,xmm7
+        psllw    xmm7,7
+        packsswb xmm7,xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   xmm0,xmm7                       ; xmm0=(01234567)
+        psubb   xmm1,xmm7                       ; xmm1=(89ABCDEF)
+
+        punpcklbw xmm0,xmm0                     ; xmm0=(*0*1*2*3*4*5*6*7)
+        punpcklbw xmm1,xmm1                     ; xmm1=(*8*9*A*B*C*D*E*F)
+
+        punpcklwd xmm2,xmm0                     ; xmm2=(***0***1***2***3)
+        punpckhwd xmm0,xmm0                     ; xmm0=(***4***5***6***7)
+        punpcklwd xmm3,xmm1                     ; xmm3=(***8***9***A***B)
+        punpckhwd xmm1,xmm1                     ; xmm1=(***C***D***E***F)
+
+        psrad     xmm2,(DWORD_BIT-BYTE_BIT)     ; xmm2=(0123)
+        psrad     xmm0,(DWORD_BIT-BYTE_BIT)     ; xmm0=(4567)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=(0123)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=(4567)
+        psrad     xmm3,(DWORD_BIT-BYTE_BIT)     ; xmm3=(89AB)
+        psrad     xmm1,(DWORD_BIT-BYTE_BIT)     ; xmm1=(CDEF)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=(89AB)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     short .convloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                            FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse2) PRIVATE
+
+EXTN(jsimd_quantize_float_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        cvtps2dq xmm0,xmm0
+        cvtps2dq xmm1,xmm1
+        cvtps2dq xmm2,xmm2
+        cvtps2dq xmm3,xmm3
+
+        packssdw xmm0,xmm1
+        packssdw xmm2,xmm3
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_JCOEF)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_JCOEF)], xmm2
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     short .quantloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-altivec.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,252 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* INTEGER QUANTIZATION AND SAMPLE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+/* NOTE: The address will either be aligned or offset by 8 bytes, so we can
+ * always get the data we want by using a single vector load (although we may
+ * have to permute the result.)
+ */
+#if __BIG_ENDIAN__
+
+#define LOAD_ROW(row) {  \
+  elemptr = sample_data[row] + start_col;  \
+  in##row = vec_ld(0, elemptr);  \
+  if ((size_t)elemptr & 15)  \
+    in##row = vec_perm(in##row, in##row, vec_lvsl(0, elemptr));  \
+}
+
+#else
+
+#define LOAD_ROW(row) {  \
+  elemptr = sample_data[row] + start_col;  \
+  in##row = vec_vsx_ld(0, elemptr);  \
+}
+
+#endif
+
+
+void
+jsimd_convsamp_altivec (JSAMPARRAY sample_data, JDIMENSION start_col,
+                        DCTELEM *workspace)
+{
+  JSAMPROW elemptr;
+
+  __vector unsigned char in0, in1, in2, in3, in4, in5, in6, in7;
+  __vector short out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_centerjsamp = { __8X(CENTERJSAMPLE) };
+  __vector unsigned char pb_zero = { __16X(0) };
+
+  LOAD_ROW(0);
+  LOAD_ROW(1);
+  LOAD_ROW(2);
+  LOAD_ROW(3);
+  LOAD_ROW(4);
+  LOAD_ROW(5);
+  LOAD_ROW(6);
+  LOAD_ROW(7);
+
+  out0 = (__vector short)VEC_UNPACKHU(in0);
+  out1 = (__vector short)VEC_UNPACKHU(in1);
+  out2 = (__vector short)VEC_UNPACKHU(in2);
+  out3 = (__vector short)VEC_UNPACKHU(in3);
+  out4 = (__vector short)VEC_UNPACKHU(in4);
+  out5 = (__vector short)VEC_UNPACKHU(in5);
+  out6 = (__vector short)VEC_UNPACKHU(in6);
+  out7 = (__vector short)VEC_UNPACKHU(in7);
+
+  out0 = vec_sub(out0, pw_centerjsamp);
+  out1 = vec_sub(out1, pw_centerjsamp);
+  out2 = vec_sub(out2, pw_centerjsamp);
+  out3 = vec_sub(out3, pw_centerjsamp);
+  out4 = vec_sub(out4, pw_centerjsamp);
+  out5 = vec_sub(out5, pw_centerjsamp);
+  out6 = vec_sub(out6, pw_centerjsamp);
+  out7 = vec_sub(out7, pw_centerjsamp);
+
+  vec_st(out0, 0, workspace);
+  vec_st(out1, 16, workspace);
+  vec_st(out2, 32, workspace);
+  vec_st(out3, 48, workspace);
+  vec_st(out4, 64, workspace);
+  vec_st(out5, 80, workspace);
+  vec_st(out6, 96, workspace);
+  vec_st(out7, 112, workspace);
+}
+
+
+#define WORD_BIT 16
+
+/* There is no AltiVec 16-bit unsigned multiply instruction, hence this.
+   We basically need an unsigned equivalent of vec_madds(). */
+
+#define MULTIPLY(vs0, vs1, out) {  \
+  tmpe = vec_mule((__vector unsigned short)vs0,  \
+                  (__vector unsigned short)vs1);  \
+  tmpo = vec_mulo((__vector unsigned short)vs0,  \
+                  (__vector unsigned short)vs1);  \
+  out = (__vector short)vec_perm((__vector unsigned short)tmpe,  \
+                                 (__vector unsigned short)tmpo,  \
+                                 shift_pack_index);  \
+}
+
+void
+jsimd_quantize_altivec (JCOEFPTR coef_block, DCTELEM *divisors,
+                        DCTELEM *workspace)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    row0s, row1s, row2s, row3s, row4s, row5s, row6s, row7s,
+    corr0, corr1, corr2, corr3, corr4, corr5, corr6, corr7,
+    recip0, recip1, recip2, recip3, recip4, recip5, recip6, recip7,
+    scale0, scale1, scale2, scale3, scale4, scale5, scale6, scale7;
+  __vector unsigned int tmpe, tmpo;
+
+  /* Constants */
+  __vector unsigned short pw_word_bit_m1 = { __8X(WORD_BIT - 1) };
+#if __BIG_ENDIAN__
+  __vector unsigned char shift_pack_index =
+    {0,1,16,17,4,5,20,21,8,9,24,25,12,13,28,29};
+#else
+  __vector unsigned char shift_pack_index =
+    {2,3,18,19,6,7,22,23,10,11,26,27,14,15,30,31};
+#endif
+
+  row0 = vec_ld(0, workspace);
+  row1 = vec_ld(16, workspace);
+  row2 = vec_ld(32, workspace);
+  row3 = vec_ld(48, workspace);
+  row4 = vec_ld(64, workspace);
+  row5 = vec_ld(80, workspace);
+  row6 = vec_ld(96, workspace);
+  row7 = vec_ld(112, workspace);
+
+  /* Branch-less absolute value */
+  row0s = vec_sra(row0, pw_word_bit_m1);
+  row1s = vec_sra(row1, pw_word_bit_m1);
+  row2s = vec_sra(row2, pw_word_bit_m1);
+  row3s = vec_sra(row3, pw_word_bit_m1);
+  row4s = vec_sra(row4, pw_word_bit_m1);
+  row5s = vec_sra(row5, pw_word_bit_m1);
+  row6s = vec_sra(row6, pw_word_bit_m1);
+  row7s = vec_sra(row7, pw_word_bit_m1);
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  corr0 = vec_ld(DCTSIZE2 * 2, divisors);
+  corr1 = vec_ld(DCTSIZE2 * 2 + 16, divisors);
+  corr2 = vec_ld(DCTSIZE2 * 2 + 32, divisors);
+  corr3 = vec_ld(DCTSIZE2 * 2 + 48, divisors);
+  corr4 = vec_ld(DCTSIZE2 * 2 + 64, divisors);
+  corr5 = vec_ld(DCTSIZE2 * 2 + 80, divisors);
+  corr6 = vec_ld(DCTSIZE2 * 2 + 96, divisors);
+  corr7 = vec_ld(DCTSIZE2 * 2 + 112, divisors);
+
+  row0 = vec_add(row0, corr0);
+  row1 = vec_add(row1, corr1);
+  row2 = vec_add(row2, corr2);
+  row3 = vec_add(row3, corr3);
+  row4 = vec_add(row4, corr4);
+  row5 = vec_add(row5, corr5);
+  row6 = vec_add(row6, corr6);
+  row7 = vec_add(row7, corr7);
+
+  recip0 = vec_ld(0, divisors);
+  recip1 = vec_ld(16, divisors);
+  recip2 = vec_ld(32, divisors);
+  recip3 = vec_ld(48, divisors);
+  recip4 = vec_ld(64, divisors);
+  recip5 = vec_ld(80, divisors);
+  recip6 = vec_ld(96, divisors);
+  recip7 = vec_ld(112, divisors);
+
+  MULTIPLY(row0, recip0, row0);
+  MULTIPLY(row1, recip1, row1);
+  MULTIPLY(row2, recip2, row2);
+  MULTIPLY(row3, recip3, row3);
+  MULTIPLY(row4, recip4, row4);
+  MULTIPLY(row5, recip5, row5);
+  MULTIPLY(row6, recip6, row6);
+  MULTIPLY(row7, recip7, row7);
+
+  scale0 = vec_ld(DCTSIZE2 * 4, divisors);
+  scale1 = vec_ld(DCTSIZE2 * 4 + 16, divisors);
+  scale2 = vec_ld(DCTSIZE2 * 4 + 32, divisors);
+  scale3 = vec_ld(DCTSIZE2 * 4 + 48, divisors);
+  scale4 = vec_ld(DCTSIZE2 * 4 + 64, divisors);
+  scale5 = vec_ld(DCTSIZE2 * 4 + 80, divisors);
+  scale6 = vec_ld(DCTSIZE2 * 4 + 96, divisors);
+  scale7 = vec_ld(DCTSIZE2 * 4 + 112, divisors);
+
+  MULTIPLY(row0, scale0, row0);
+  MULTIPLY(row1, scale1, row1);
+  MULTIPLY(row2, scale2, row2);
+  MULTIPLY(row3, scale3, row3);
+  MULTIPLY(row4, scale4, row4);
+  MULTIPLY(row5, scale5, row5);
+  MULTIPLY(row6, scale6, row6);
+  MULTIPLY(row7, scale7, row7);
+
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  vec_st(row0, 0, coef_block);
+  vec_st(row1, 16, coef_block);
+  vec_st(row2, 32, coef_block);
+  vec_st(row3, 48, coef_block);
+  vec_st(row4, 64, coef_block);
+  vec_st(row5, 80, coef_block);
+  vec_st(row6, 96, coef_block);
+  vec_st(row7, 112, coef_block);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-sse2-64.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,187 @@
+;
+; jquanti.asm - sample data conversion and quantization (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                      DCTELEM *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11 = JDIMENSION start_col
+; r12 = DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        pxor    xmm6,xmm6               ; xmm6=(all 0's)
+        pcmpeqw xmm7,xmm7
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        mov rsi, r10
+        mov eax, r11d
+        mov rdi, r12
+        mov     rcx, DCTSIZE/4
+.convloop:
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]       ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]       ; xmm0=(01234567)
+        movq    xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]       ; xmm1=(89ABCDEF)
+
+        mov     rbx, JSAMPROW [rsi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     rdx, JSAMPROW [rsi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm2, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]       ; xmm2=(GHIJKLMN)
+        movq    xmm3, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]       ; xmm3=(OPQRSTUV)
+
+        punpcklbw xmm0,xmm6             ; xmm0=(01234567)
+        punpcklbw xmm1,xmm6             ; xmm1=(89ABCDEF)
+        paddw     xmm0,xmm7
+        paddw     xmm1,xmm7
+        punpcklbw xmm2,xmm6             ; xmm2=(GHIJKLMN)
+        punpcklbw xmm3,xmm6             ; xmm3=(OPQRSTUV)
+        paddw     xmm2,xmm7
+        paddw     xmm3,xmm7
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3
+
+        add     rsi, byte 4*SIZEOF_JSAMPROW
+        add     rdi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     rcx
+        jnz     short .convloop
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM *divisors,
+;                      DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+
+; r10 = JCOEFPTR coef_block
+; r11 = DCTELEM *divisors
+; r12 = DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_sse2) PRIVATE
+
+EXTN(jsimd_quantize_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov rsi, r12
+        mov rdx, r11
+        mov rdi, r10
+        mov     rax, DCTSIZE2/32
+.quantloop:
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm0,xmm4
+        movdqa  xmm1,xmm5
+        movdqa  xmm2,xmm6
+        movdqa  xmm3,xmm7
+        psraw   xmm4,(WORD_BIT-1)
+        psraw   xmm5,(WORD_BIT-1)
+        psraw   xmm6,(WORD_BIT-1)
+        psraw   xmm7,(WORD_BIT-1)
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4               ; if (xmm0 < 0) xmm0 = -xmm0;
+        psubw   xmm1,xmm5               ; if (xmm1 < 0) xmm1 = -xmm1;
+        psubw   xmm2,xmm6               ; if (xmm2 < 0) xmm2 = -xmm2;
+        psubw   xmm3,xmm7               ; if (xmm3 < 0) xmm3 = -xmm3;
+
+        paddw   xmm0, XMMWORD [CORRECTION(0,0,rdx)]  ; correction + roundfactor
+        paddw   xmm1, XMMWORD [CORRECTION(1,0,rdx)]
+        paddw   xmm2, XMMWORD [CORRECTION(2,0,rdx)]
+        paddw   xmm3, XMMWORD [CORRECTION(3,0,rdx)]
+        pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,rdx)]  ; reciprocal
+        pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,rdx)]
+        pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,rdx)]
+        pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,rdx)]
+        pmulhuw xmm0, XMMWORD [SCALE(0,0,rdx)]  ; scale
+        pmulhuw xmm1, XMMWORD [SCALE(1,0,rdx)]
+        pmulhuw xmm2, XMMWORD [SCALE(2,0,rdx)]
+        pmulhuw xmm3, XMMWORD [SCALE(3,0,rdx)]
+
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4
+        psubw   xmm1,xmm5
+        psubw   xmm2,xmm6
+        psubw   xmm3,xmm7
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3
+
+        add     rsi, byte 32*SIZEOF_DCTELEM
+        add     rdx, byte 32*SIZEOF_DCTELEM
+        add     rdi, byte 32*SIZEOF_JCOEF
+        dec     rax
+        jnz     near .quantloop
+
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-sse2.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquanti-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquanti-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,200 @@
+;
+; jquanti.asm - sample data conversion and quantization (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                      DCTELEM *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pxor    xmm6,xmm6               ; xmm6=(all 0's)
+        pcmpeqw xmm7,xmm7
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]       ; xmm0=(01234567)
+        movq    xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]       ; xmm1=(89ABCDEF)
+
+        mov     ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm2, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]       ; xmm2=(GHIJKLMN)
+        movq    xmm3, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]       ; xmm3=(OPQRSTUV)
+
+        punpcklbw xmm0,xmm6             ; xmm0=(01234567)
+        punpcklbw xmm1,xmm6             ; xmm1=(89ABCDEF)
+        paddw     xmm0,xmm7
+        paddw     xmm1,xmm7
+        punpcklbw xmm2,xmm6             ; xmm2=(GHIJKLMN)
+        punpcklbw xmm3,xmm6             ; xmm3=(OPQRSTUV)
+        paddw     xmm2,xmm7
+        paddw     xmm3,xmm7
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3
+
+        add     esi, byte 4*SIZEOF_JSAMPROW
+        add     edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     short .convloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM *divisors,
+;                      DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; DCTELEM *divisors
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_sse2) PRIVATE
+
+EXTN(jsimd_quantize_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/32
+        alignx  16,7
+.quantloop:
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm0,xmm4
+        movdqa  xmm1,xmm5
+        movdqa  xmm2,xmm6
+        movdqa  xmm3,xmm7
+        psraw   xmm4,(WORD_BIT-1)
+        psraw   xmm5,(WORD_BIT-1)
+        psraw   xmm6,(WORD_BIT-1)
+        psraw   xmm7,(WORD_BIT-1)
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4               ; if (xmm0 < 0) xmm0 = -xmm0;
+        psubw   xmm1,xmm5               ; if (xmm1 < 0) xmm1 = -xmm1;
+        psubw   xmm2,xmm6               ; if (xmm2 < 0) xmm2 = -xmm2;
+        psubw   xmm3,xmm7               ; if (xmm3 < 0) xmm3 = -xmm3;
+
+        paddw   xmm0, XMMWORD [CORRECTION(0,0,edx)]  ; correction + roundfactor
+        paddw   xmm1, XMMWORD [CORRECTION(1,0,edx)]
+        paddw   xmm2, XMMWORD [CORRECTION(2,0,edx)]
+        paddw   xmm3, XMMWORD [CORRECTION(3,0,edx)]
+        pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,edx)]  ; reciprocal
+        pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,edx)]
+        pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,edx)]
+        pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,edx)]
+        pmulhuw xmm0, XMMWORD [SCALE(0,0,edx)]  ; scale
+        pmulhuw xmm1, XMMWORD [SCALE(1,0,edx)]
+        pmulhuw xmm2, XMMWORD [SCALE(2,0,edx)]
+        pmulhuw xmm3, XMMWORD [SCALE(3,0,edx)]
+
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4
+        psubw   xmm1,xmm5
+        psubw   xmm2,xmm6
+        psubw   xmm3,xmm7
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3
+
+        add     esi, byte 32*SIZEOF_DCTELEM
+        add     edx, byte 32*SIZEOF_DCTELEM
+        add     edi, byte 32*SIZEOF_JCOEF
+        dec     eax
+        jnz     near .quantloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-mmx.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,274 @@
+;
+; jquant.asm - sample data conversion and quantization (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_mmx (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_mmx) PRIVATE
+
+EXTN(jsimd_convsamp_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pxor    mm6,mm6                 ; mm6=(all 0's)
+        pcmpeqw mm7,mm7
+        psllw   mm7,7                   ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]    ; mm0=(01234567)
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]    ; mm1=(89ABCDEF)
+
+        mov     ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm2, MMWORD [ebx+eax*SIZEOF_JSAMPLE]    ; mm2=(GHIJKLMN)
+        movq    mm3, MMWORD [edx+eax*SIZEOF_JSAMPLE]    ; mm3=(OPQRSTUV)
+
+        movq      mm4,mm0
+        punpcklbw mm0,mm6               ; mm0=(0123)
+        punpckhbw mm4,mm6               ; mm4=(4567)
+        movq      mm5,mm1
+        punpcklbw mm1,mm6               ; mm1=(89AB)
+        punpckhbw mm5,mm6               ; mm5=(CDEF)
+
+        paddw   mm0,mm7
+        paddw   mm4,mm7
+        paddw   mm1,mm7
+        paddw   mm5,mm7
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_DCTELEM)], mm5
+
+        movq      mm0,mm2
+        punpcklbw mm2,mm6               ; mm2=(GHIJ)
+        punpckhbw mm0,mm6               ; mm0=(KLMN)
+        movq      mm4,mm3
+        punpcklbw mm3,mm6               ; mm3=(OPQR)
+        punpckhbw mm4,mm6               ; mm4=(STUV)
+
+        paddw   mm2,mm7
+        paddw   mm0,mm7
+        paddw   mm3,mm7
+        paddw   mm4,mm7
+
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_DCTELEM)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_DCTELEM)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_DCTELEM)], mm4
+
+        add     esi, byte 4*SIZEOF_JSAMPROW
+        add     edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     short .convloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_mmx (JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) MMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) MMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      MMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+%define SHIFT(m,n,b)      MMBLOCK(DCTSIZE*3+(m),(n),(b),SIZEOF_DCTELEM)
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; DCTELEM *divisors
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_mmx) PRIVATE
+
+EXTN(jsimd_quantize_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     ah, 2
+        alignx  16,7
+.quantloop1:
+        mov     al, DCTSIZE2/8/2
+        alignx  16,7
+.quantloop2:
+        movq    mm2, MMWORD [MMBLOCK(0,0,esi,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(0,1,esi,SIZEOF_DCTELEM)]
+
+        movq    mm0,mm2
+        movq    mm1,mm3
+
+        psraw   mm2,(WORD_BIT-1)  ; -1 if value < 0, 0 otherwise
+        psraw   mm3,(WORD_BIT-1)
+
+        pxor    mm0,mm2   ; val = -val
+        pxor    mm1,mm3
+        psubw   mm0,mm2
+        psubw   mm1,mm3
+
+        ;
+        ; MMX is an annoyingly crappy instruction set. It has two
+        ; misfeatures that are causing problems here:
+        ;
+        ; - All multiplications are signed.
+        ;
+        ; - The second operand for the shifts is not treated as packed.
+        ;
+        ;
+        ; We work around the first problem by implementing this algorithm:
+        ;
+        ; unsigned long unsigned_multiply(unsigned short x, unsigned short y)
+        ; {
+        ;   enum { SHORT_BIT = 16 };
+        ;   signed short sx = (signed short) x;
+        ;   signed short sy = (signed short) y;
+        ;   signed long sz;
+        ;
+        ;   sz = (long) sx * (long) sy;     /* signed multiply */
+        ;
+        ;   if (sx < 0) sz += (long) sy << SHORT_BIT;
+        ;   if (sy < 0) sz += (long) sx << SHORT_BIT;
+        ;
+        ;   return (unsigned long) sz;
+        ; }
+        ;
+        ; (note that a negative sx adds _sy_ and vice versa)
+        ;
+        ; For the second problem, we replace the shift by a multiplication.
+        ; Unfortunately that means we have to deal with the signed issue again.
+        ;
+
+        paddw   mm0, MMWORD [CORRECTION(0,0,edx)]   ; correction + roundfactor
+        paddw   mm1, MMWORD [CORRECTION(0,1,edx)]
+
+        movq    mm4,mm0   ; store current value for later
+        movq    mm5,mm1
+        pmulhw  mm0, MMWORD [RECIPROCAL(0,0,edx)]   ; reciprocal
+        pmulhw  mm1, MMWORD [RECIPROCAL(0,1,edx)]
+        paddw   mm0,mm4         ; reciprocal is always negative (MSB=1),
+        paddw   mm1,mm5   ; so we always need to add the initial value
+                        ; (input value is never negative as we
+                        ; inverted it at the start of this routine)
+
+        ; here it gets a bit tricky as both scale
+        ; and mm0/mm1 can be negative
+        movq    mm6, MMWORD [SCALE(0,0,edx)]    ; scale
+        movq    mm7, MMWORD [SCALE(0,1,edx)]
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pmulhw  mm0,mm6
+        pmulhw  mm1,mm7
+
+        psraw   mm6,(WORD_BIT-1)    ; determine if scale is negative
+        psraw   mm7,(WORD_BIT-1)
+
+        pand    mm6,mm4             ; and add input if it is
+        pand    mm7,mm5
+        paddw   mm0,mm6
+        paddw   mm1,mm7
+
+        psraw   mm4,(WORD_BIT-1)    ; then check if negative input
+        psraw   mm5,(WORD_BIT-1)
+
+        pand    mm4, MMWORD [SCALE(0,0,edx)]    ; and add scale if it is
+        pand    mm5, MMWORD [SCALE(0,1,edx)]
+        paddw   mm0,mm4
+        paddw   mm1,mm5
+
+        pxor    mm0,mm2   ; val = -val
+        pxor    mm1,mm3
+        psubw   mm0,mm2
+        psubw   mm1,mm3
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm1
+
+        add     esi, byte 8*SIZEOF_DCTELEM
+        add     edx, byte 8*SIZEOF_DCTELEM
+        add     edi, byte 8*SIZEOF_JCOEF
+        dec     al
+        jnz     near .quantloop2
+        dec     ah
+        jnz     near .quantloop1        ; to avoid branch misprediction
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-sse.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jquant-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jquant-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,211 @@
+;
+; jquant.asm - sample data conversion and quantization (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                           FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  mm7,mm7
+        psllw    mm7,7
+        packsswb mm7,mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   mm0,mm7                         ; mm0=(01234567)
+        psubb   mm1,mm7                         ; mm1=(89ABCDEF)
+
+        punpcklbw mm2,mm0                       ; mm2=(*0*1*2*3)
+        punpckhbw mm0,mm0                       ; mm0=(*4*5*6*7)
+        punpcklbw mm3,mm1                       ; mm3=(*8*9*A*B)
+        punpckhbw mm1,mm1                       ; mm1=(*C*D*E*F)
+
+        punpcklwd mm4,mm2                       ; mm4=(***0***1)
+        punpckhwd mm2,mm2                       ; mm2=(***2***3)
+        punpcklwd mm5,mm0                       ; mm5=(***4***5)
+        punpckhwd mm0,mm0                       ; mm0=(***6***7)
+
+        psrad     mm4,(DWORD_BIT-BYTE_BIT)      ; mm4=(01)
+        psrad     mm2,(DWORD_BIT-BYTE_BIT)      ; mm2=(23)
+        cvtpi2ps  xmm0,mm4                      ; xmm0=(01**)
+        cvtpi2ps  xmm1,mm2                      ; xmm1=(23**)
+        psrad     mm5,(DWORD_BIT-BYTE_BIT)      ; mm5=(45)
+        psrad     mm0,(DWORD_BIT-BYTE_BIT)      ; mm0=(67)
+        cvtpi2ps  xmm2,mm5                      ; xmm2=(45**)
+        cvtpi2ps  xmm3,mm0                      ; xmm3=(67**)
+
+        punpcklwd mm6,mm3                       ; mm6=(***8***9)
+        punpckhwd mm3,mm3                       ; mm3=(***A***B)
+        punpcklwd mm4,mm1                       ; mm4=(***C***D)
+        punpckhwd mm1,mm1                       ; mm1=(***E***F)
+
+        psrad     mm6,(DWORD_BIT-BYTE_BIT)      ; mm6=(89)
+        psrad     mm3,(DWORD_BIT-BYTE_BIT)      ; mm3=(AB)
+        cvtpi2ps  xmm4,mm6                      ; xmm4=(89**)
+        cvtpi2ps  xmm5,mm3                      ; xmm5=(AB**)
+        psrad     mm4,(DWORD_BIT-BYTE_BIT)      ; mm4=(CD)
+        psrad     mm1,(DWORD_BIT-BYTE_BIT)      ; mm1=(EF)
+        cvtpi2ps  xmm6,mm4                      ; xmm6=(CD**)
+        cvtpi2ps  xmm7,mm1                      ; xmm7=(EF**)
+
+        movlhps   xmm0,xmm1                     ; xmm0=(0123)
+        movlhps   xmm2,xmm3                     ; xmm2=(4567)
+        movlhps   xmm4,xmm5                     ; xmm4=(89AB)
+        movlhps   xmm6,xmm7                     ; xmm6=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .convloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                           FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse) PRIVATE
+
+EXTN(jsimd_quantize_float_sse):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        movhlps  xmm4,xmm0
+        movhlps  xmm5,xmm1
+
+        cvtps2pi mm0,xmm0
+        cvtps2pi mm1,xmm1
+        cvtps2pi mm4,xmm4
+        cvtps2pi mm5,xmm5
+
+        movhlps  xmm6,xmm2
+        movhlps  xmm7,xmm3
+
+        cvtps2pi mm2,xmm2
+        cvtps2pi mm3,xmm3
+        cvtps2pi mm6,xmm6
+        cvtps2pi mm7,xmm7
+
+        packssdw mm0,mm4
+        packssdw mm1,mm5
+        packssdw mm2,mm6
+        packssdw mm3,mm7
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     short .quantloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_altivec.h glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_altivec.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_altivec.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_altivec.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,99 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+#include <altivec.h>
+
+
+/* Common code */
+
+#define __4X(a) a, a, a, a
+#define __4X2(a, b) a, b, a, b, a, b, a, b
+#define __8X(a) __4X(a), __4X(a)
+#define __16X(a) __8X(a), __8X(a)
+
+#define TRANSPOSE(row, col)  \
+{  \
+  __vector short row04l, row04h, row15l, row15h,  \
+                 row26l, row26h, row37l, row37h;  \
+  __vector short col01e, col01o, col23e, col23o,  \
+                 col45e, col45o, col67e, col67o;  \
+  \
+                                       /* transpose coefficients (phase 1) */ \
+  row04l = vec_mergeh(row##0, row##4); /* row04l=(00 40 01 41 02 42 03 43) */ \
+  row04h = vec_mergel(row##0, row##4); /* row04h=(04 44 05 45 06 46 07 47) */ \
+  row15l = vec_mergeh(row##1, row##5); /* row15l=(10 50 11 51 12 52 13 53) */ \
+  row15h = vec_mergel(row##1, row##5); /* row15h=(14 54 15 55 16 56 17 57) */ \
+  row26l = vec_mergeh(row##2, row##6); /* row26l=(20 60 21 61 22 62 23 63) */ \
+  row26h = vec_mergel(row##2, row##6); /* row26h=(24 64 25 65 26 66 27 67) */ \
+  row37l = vec_mergeh(row##3, row##7); /* row37l=(30 70 31 71 32 72 33 73) */ \
+  row37h = vec_mergel(row##3, row##7); /* row37h=(34 74 35 75 36 76 37 77) */ \
+  \
+                                       /* transpose coefficients (phase 2) */ \
+  col01e = vec_mergeh(row04l, row26l); /* col01e=(00 20 40 60 01 21 41 61) */ \
+  col23e = vec_mergel(row04l, row26l); /* col23e=(02 22 42 62 03 23 43 63) */ \
+  col45e = vec_mergeh(row04h, row26h); /* col45e=(04 24 44 64 05 25 45 65) */ \
+  col67e = vec_mergel(row04h, row26h); /* col67e=(06 26 46 66 07 27 47 67) */ \
+  col01o = vec_mergeh(row15l, row37l); /* col01o=(10 30 50 70 11 31 51 71) */ \
+  col23o = vec_mergel(row15l, row37l); /* col23o=(12 32 52 72 13 33 53 73) */ \
+  col45o = vec_mergeh(row15h, row37h); /* col45o=(14 34 54 74 15 35 55 75) */ \
+  col67o = vec_mergel(row15h, row37h); /* col67o=(16 36 56 76 17 37 57 77) */ \
+  \
+                                       /* transpose coefficients (phase 3) */ \
+  col##0 = vec_mergeh(col01e, col01o); /* col0=(00 10 20 30 40 50 60 70) */   \
+  col##1 = vec_mergel(col01e, col01o); /* col1=(01 11 21 31 41 51 61 71) */   \
+  col##2 = vec_mergeh(col23e, col23o); /* col2=(02 12 22 32 42 52 62 72) */   \
+  col##3 = vec_mergel(col23e, col23o); /* col3=(03 13 23 33 43 53 63 73) */   \
+  col##4 = vec_mergeh(col45e, col45o); /* col4=(04 14 24 34 44 54 64 74) */   \
+  col##5 = vec_mergel(col45e, col45o); /* col5=(05 15 25 35 45 55 65 75) */   \
+  col##6 = vec_mergeh(col67e, col67o); /* col6=(06 16 26 36 46 56 66 76) */   \
+  col##7 = vec_mergel(col67e, col67o); /* col7=(07 17 27 37 47 57 67 77) */   \
+}
+
+#ifndef min
+#define min(a,b) ((a) < (b) ? (a) : (b))
+#endif
+
+
+/* Macros to abstract big/little endian bit twiddling */
+
+#if __BIG_ENDIAN__
+
+#define VEC_LD(a, b) vec_ld(a, b)
+#define VEC_ST(a, b, c) vec_st(a, b, c)
+#define VEC_UNPACKHU(a) vec_mergeh(pb_zero, a)
+#define VEC_UNPACKLU(a) vec_mergel(pb_zero, a)
+
+#else
+
+#define VEC_LD(a, b) vec_vsx_ld(a, b)
+#define VEC_ST(a, b, c) vec_vsx_st(a, b, c)
+#define VEC_UNPACKHU(a) vec_mergeh(a, pb_zero)
+#define VEC_UNPACKLU(a) vec_mergel(a, pb_zero)
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm64.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm64.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm64.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm64.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,802 @@
+/*
+ * jsimd_arm64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015-2016 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit ARM architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+#define JSIMD_FASTLD3 1
+#define JSIMD_FASTST3 2
+#define JSIMD_FASTTBL 4
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+static unsigned int simd_features = JSIMD_FASTLD3 | JSIMD_FASTST3 |
+                                    JSIMD_FASTTBL;
+
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
+
+LOCAL(int)
+check_cpuinfo (char *buffer, const char *field, char *value)
+{
+  char *p;
+  if (*value == 0)
+    return 0;
+  if (strncmp(buffer, field, strlen(field)) != 0)
+    return 0;
+  buffer += strlen(field);
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'value' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, value))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(value);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo (int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_cpuinfo(buffer, "CPU part", "0xd03") ||
+          check_cpuinfo(buffer, "CPU part", "0xd07"))
+        /* The Cortex-A53 has a slow tbl implementation.  We can gain a few
+           percent speedup by disabling the use of that instruction.  The
+           speedup on Cortex-A57 is more subtle but still measurable. */
+        simd_features &= ~JSIMD_FASTTBL;
+      else if (check_cpuinfo(buffer, "CPU part", "0x0a1"))
+        /* The SIMD version of Huffman encoding is slower than the C version on
+           Cavium ThunderX.  Also, ld3 and st3 are abyssmally slow on that
+           CPU. */
+        simd_huffman = simd_features = 0;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+
+/*
+ * ARMv8 architectures support NEON extensions by default.
+ * It is no longer optional as it was with ARMv7.
+ */
+
+
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+  simd_support |= JSIMD_ARM_NEON;
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#endif
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENEON");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_ARM_NEON;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+  env = getenv("JSIMD_FASTLD3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTLD3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTLD3;
+  env = getenv("JSIMD_FASTST3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTST3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTST3;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extrgb_ycc_convert_neon;
+      else
+        neonfct=jsimd_extrgb_ycc_convert_neon_slowld3;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct=jsimd_extrgbx_ycc_convert_neon;
+      break;
+    case JCS_EXT_BGR:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extbgr_ycc_convert_neon;
+      else
+        neonfct=jsimd_extbgr_ycc_convert_neon_slowld3;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct=jsimd_extbgrx_ycc_convert_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct=jsimd_extxbgr_ycc_convert_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct=jsimd_extxrgb_ycc_convert_neon;
+      break;
+    default:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extrgb_ycc_convert_neon;
+      else
+        neonfct=jsimd_extrgb_ycc_convert_neon_slowld3;
+      break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extrgb_convert_neon;
+      else
+        neonfct=jsimd_ycc_extrgb_convert_neon_slowst3;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct=jsimd_ycc_extrgbx_convert_neon;
+      break;
+    case JCS_EXT_BGR:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extbgr_convert_neon;
+      else
+        neonfct=jsimd_ycc_extbgr_convert_neon_slowst3;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct=jsimd_ycc_extbgrx_convert_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct=jsimd_ycc_extxbgr_convert_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct=jsimd_ycc_extxrgb_convert_neon;
+      break;
+    default:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extrgb_convert_neon;
+      else
+        neonfct=jsimd_ycc_extrgb_convert_neon_slowst3;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_neon(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  if (simd_features & JSIMD_FASTTBL)
+    return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                            dctbl, actbl);
+  else
+    return jsimd_huff_encode_one_block_neon_slowtbl(state, buffer, block,
+                                                    last_dc_val, dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm64_neon.S glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm64_neon.S
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm64_neon.S	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm64_neon.S	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3453 @@
+/*
+ * ARMv8 NEON optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies).
+ * All rights reserved.
+ * Author: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2013-2014, Linaro Limited
+ * Author: Ragesh Radhakrishnan <ragesh.r@linaro.org>
+ * Copyright (C) 2014-2016, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016, Matthieu Darbois.  All Rights Reserved.
+ * Copyright (C) 2016, Siarhei Siamashka.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
+#endif
+
+.text
+
+
+#define RESPECT_STRICT_ALIGNMENT 1
+
+
+/*****************************************************************************/
+
+/* Supplementary macro for setting function attributes */
+.macro asm_function fname
+#ifdef __APPLE__
+    .globl _\fname
+_\fname:
+#else
+    .global \fname
+#ifdef __ELF__
+    .hidden \fname
+    .type \fname, %function
+#endif
+\fname:
+#endif
+.endm
+
+/* Transpose elements of single 128 bit registers */
+.macro transpose_single x0, x1, xi, xilen, literal
+    ins             \xi\xilen[0], \x0\xilen[0]
+    ins             \x1\xilen[0], \x0\xilen[1]
+    trn1            \x0\literal, \x0\literal, \x1\literal
+    trn2            \x1\literal, \xi\literal, \x1\literal
+.endm
+
+/* Transpose elements of 2 differnet registers */
+.macro transpose x0, x1, xi, xilen, literal
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\literal, \x0\literal, \x1\literal
+    trn2            \x1\literal, \xi\literal, \x1\literal
+.endm
+
+/* Transpose a block of 4x4 coefficients in four 64-bit registers */
+.macro transpose_4x4_32 x0, x0len, x1, x1len, x2, x2len, x3, x3len, xi, xilen
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\x0len, \x0\x0len, \x2\x2len
+    trn2            \x2\x2len, \xi\x0len, \x2\x2len
+    mov             \xi\xilen, \x1\xilen
+    trn1            \x1\x1len, \x1\x1len, \x3\x3len
+    trn2            \x3\x3len, \xi\x1len, \x3\x3len
+.endm
+
+.macro transpose_4x4_16 x0, x0len, x1, x1len, x2, x2len, x3, x3len, xi, xilen
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\x0len, \x0\x0len, \x1\x1len
+    trn2            \x1\x2len, \xi\x0len, \x1\x2len
+    mov             \xi\xilen, \x2\xilen
+    trn1            \x2\x2len, \x2\x2len, \x3\x3len
+    trn2            \x3\x2len, \xi\x1len, \x3\x3len
+.endm
+
+.macro transpose_4x4 x0, x1, x2, x3, x5
+    transpose_4x4_16 \x0, .4h, \x1, .4h, \x2, .4h, \x3, .4h, \x5, .16b
+    transpose_4x4_32 \x0, .2s, \x1, .2s, \x2, .2s, \x3, .2s, \x5, .16b
+.endm
+
+.macro transpose_8x8 l0, l1, l2, l3, l4, l5, l6, l7, t0, t1, t2, t3
+    trn1            \t0\().8h, \l0\().8h, \l1\().8h
+    trn1            \t1\().8h, \l2\().8h, \l3\().8h
+    trn1            \t2\().8h, \l4\().8h, \l5\().8h
+    trn1            \t3\().8h, \l6\().8h, \l7\().8h
+    trn2            \l1\().8h, \l0\().8h, \l1\().8h
+    trn2            \l3\().8h, \l2\().8h, \l3\().8h
+    trn2            \l5\().8h, \l4\().8h, \l5\().8h
+    trn2            \l7\().8h, \l6\().8h, \l7\().8h
+
+    trn1            \l4\().4s, \t2\().4s, \t3\().4s
+    trn2            \t3\().4s, \t2\().4s, \t3\().4s
+    trn1            \t2\().4s, \t0\().4s, \t1\().4s
+    trn2            \l2\().4s, \t0\().4s, \t1\().4s
+    trn1            \t0\().4s, \l1\().4s, \l3\().4s
+    trn2            \l3\().4s, \l1\().4s, \l3\().4s
+    trn2            \t1\().4s, \l5\().4s, \l7\().4s
+    trn1            \l5\().4s, \l5\().4s, \l7\().4s
+
+    trn2            \l6\().2d, \l2\().2d, \t3\().2d
+    trn1            \l0\().2d, \t2\().2d, \l4\().2d
+    trn1            \l1\().2d, \t0\().2d, \l5\().2d
+    trn2            \l7\().2d, \l3\().2d, \t1\().2d
+    trn1            \l2\().2d, \l2\().2d, \t3\().2d
+    trn2            \l4\().2d, \t2\().2d, \l4\().2d
+    trn1            \l3\().2d, \l3\().2d, \t1\().2d
+    trn2            \l5\().2d, \t0\().2d, \l5\().2d
+.endm
+
+
+#define CENTERJSAMPLE 128
+
+/*****************************************************************************/
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ *
+ * GLOBAL(void)
+ * jsimd_idct_islow_neon (void *dct_table, JCOEFPTR coef_block,
+ *                        JSAMPARRAY output_buf, JDIMENSION output_col)
+ */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+
+#define F_0_298  2446  /* FIX(0.298631336) */
+#define F_0_390  3196  /* FIX(0.390180644) */
+#define F_0_541  4433  /* FIX(0.541196100) */
+#define F_0_765  6270  /* FIX(0.765366865) */
+#define F_0_899  7373  /* FIX(0.899976223) */
+#define F_1_175  9633  /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_idct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+
+#define XFIX_P_0_298 v0.h[0]
+#define XFIX_N_0_390 v0.h[1]
+#define XFIX_P_0_541 v0.h[2]
+#define XFIX_P_0_765 v0.h[3]
+#define XFIX_N_0_899 v0.h[4]
+#define XFIX_P_1_175 v0.h[5]
+#define XFIX_P_1_501 v0.h[6]
+#define XFIX_N_1_847 v0.h[7]
+#define XFIX_N_1_961 v1.h[0]
+#define XFIX_P_2_053 v1.h[1]
+#define XFIX_N_2_562 v1.h[2]
+#define XFIX_P_3_072 v1.h[3]
+
+asm_function jsimd_idct_islow_neon
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x9
+    TMP4            .req x10
+    TMP5            .req x11
+    TMP6            .req x12
+    TMP7            .req x13
+    TMP8            .req x14
+
+    sub             sp, sp, #64
+    adr             x15, Ljsimd_idct_islow_neon_consts
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], #32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], #32
+    ld1             {v0.8h, v1.8h}, [x15]
+    ld1             {v2.8h, v3.8h, v4.8h, v5.8h}, [COEF_BLOCK], #64
+    ld1             {v18.8h, v19.8h, v20.8h, v21.8h}, [DCT_TABLE], #64
+    ld1             {v6.8h, v7.8h, v8.8h, v9.8h}, [COEF_BLOCK], #64
+    ld1             {v22.8h, v23.8h, v24.8h, v25.8h}, [DCT_TABLE], #64
+
+    cmeq            v16.8h, v3.8h, #0
+    cmeq            v26.8h, v4.8h, #0
+    cmeq            v27.8h, v5.8h, #0
+    cmeq            v28.8h, v6.8h, #0
+    cmeq            v29.8h, v7.8h, #0
+    cmeq            v30.8h, v8.8h, #0
+    cmeq            v31.8h, v9.8h, #0
+
+    and             v10.16b, v16.16b, v26.16b
+    and             v11.16b, v27.16b, v28.16b
+    and             v12.16b, v29.16b, v30.16b
+    and             v13.16b, v31.16b, v10.16b
+    and             v14.16b, v11.16b, v12.16b
+    mul             v2.8h, v2.8h, v18.8h
+    and             v15.16b, v13.16b, v14.16b
+    shl             v10.8h, v2.8h, #(PASS1_BITS)
+    sqxtn           v16.8b, v15.8h
+    mov             TMP1, v16.d[0]
+    sub             sp, sp, #64
+    mvn             TMP2, TMP1
+
+    cbnz            TMP2, 2f
+    /* case all AC coeffs are zeros */
+    dup             v2.2d, v10.d[0]
+    dup             v6.2d, v10.d[1]
+    mov             v3.16b, v2.16b
+    mov             v7.16b, v6.16b
+    mov             v4.16b, v2.16b
+    mov             v8.16b, v6.16b
+    mov             v5.16b, v2.16b
+    mov             v9.16b, v6.16b
+1:
+    /* for this transpose, we should organise data like this:
+     * 00, 01, 02, 03, 40, 41, 42, 43
+     * 10, 11, 12, 13, 50, 51, 52, 53
+     * 20, 21, 22, 23, 60, 61, 62, 63
+     * 30, 31, 32, 33, 70, 71, 72, 73
+     * 04, 05, 06, 07, 44, 45, 46, 47
+     * 14, 15, 16, 17, 54, 55, 56, 57
+     * 24, 25, 26, 27, 64, 65, 66, 67
+     * 34, 35, 36, 37, 74, 75, 76, 77
+     */
+    trn1            v28.8h, v2.8h, v3.8h
+    trn1            v29.8h, v4.8h, v5.8h
+    trn1            v30.8h, v6.8h, v7.8h
+    trn1            v31.8h, v8.8h, v9.8h
+    trn2            v16.8h, v2.8h, v3.8h
+    trn2            v17.8h, v4.8h, v5.8h
+    trn2            v18.8h, v6.8h, v7.8h
+    trn2            v19.8h, v8.8h, v9.8h
+    trn1            v2.4s, v28.4s, v29.4s
+    trn1            v6.4s, v30.4s, v31.4s
+    trn1            v3.4s, v16.4s, v17.4s
+    trn1            v7.4s, v18.4s, v19.4s
+    trn2            v4.4s, v28.4s, v29.4s
+    trn2            v8.4s, v30.4s, v31.4s
+    trn2            v5.4s, v16.4s, v17.4s
+    trn2            v9.4s, v18.4s, v19.4s
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    shrn            v2.4h, v18.4s, #16  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v9.4h, v20.4s, #16  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v3.4h, v22.4s, #16  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v8.4h, v24.4s, #16  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v4.4h, v26.4s, #16  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v7.4h, v28.4s, #16  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v5.4h, v14.4s, #16  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn            v6.4h, v16.4s, #16  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v2.8h, v19.4s, #16  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v9.8h, v21.4s, #16  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v3.8h, v23.4s, #16  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v8.8h, v25.4s, #16  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v4.8h, v27.4s, #16  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v7.8h, v29.4s, #16  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v5.8h, v15.4s, #16  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v6.8h, v17.4s, #16  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    movi            v0.16b, #(CENTERJSAMPLE)
+    /* Prepare pointers (dual-issue with NEON instructions) */
+      ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    sqrshrn         v28.8b, v2.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP3, TMP4, [OUTPUT_BUF], 16
+    sqrshrn         v29.8b, v3.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP1, TMP1, OUTPUT_COL
+    sqrshrn         v30.8b, v4.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP2, TMP2, OUTPUT_COL
+    sqrshrn         v31.8b, v5.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP3, TMP3, OUTPUT_COL
+    sqrshrn2        v28.16b, v6.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP4, TMP4, OUTPUT_COL
+    sqrshrn2        v29.16b, v7.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP5, TMP6, [OUTPUT_BUF], 16
+    sqrshrn2        v30.16b, v8.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP7, TMP8, [OUTPUT_BUF], 16
+    sqrshrn2        v31.16b, v9.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP5, TMP5, OUTPUT_COL
+    add             v16.16b, v28.16b, v0.16b
+      add             TMP6, TMP6, OUTPUT_COL
+    add             v18.16b, v29.16b, v0.16b
+      add             TMP7, TMP7, OUTPUT_COL
+    add             v20.16b, v30.16b, v0.16b
+      add             TMP8, TMP8, OUTPUT_COL
+    add             v22.16b, v31.16b, v0.16b
+
+    /* Transpose the final 8-bit samples */
+    trn1            v28.16b, v16.16b, v18.16b
+    trn1            v30.16b, v20.16b, v22.16b
+    trn2            v29.16b, v16.16b, v18.16b
+    trn2            v31.16b, v20.16b, v22.16b
+
+    trn1            v16.8h, v28.8h, v30.8h
+    trn2            v18.8h, v28.8h, v30.8h
+    trn1            v20.8h, v29.8h, v31.8h
+    trn2            v22.8h, v29.8h, v31.8h
+
+    uzp1            v28.4s, v16.4s, v18.4s
+    uzp2            v30.4s, v16.4s, v18.4s
+    uzp1            v29.4s, v20.4s, v22.4s
+    uzp2            v31.4s, v20.4s, v22.4s
+
+    /* Store results to the output buffer */
+    st1             {v28.d}[0], [TMP1]
+    st1             {v29.d}[0], [TMP2]
+    st1             {v28.d}[1], [TMP3]
+    st1             {v29.d}[1], [TMP4]
+    st1             {v30.d}[0], [TMP5]
+    st1             {v31.d}[0], [TMP6]
+    st1             {v30.d}[1], [TMP7]
+    st1             {v31.d}[1], [TMP8]
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], #32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], #32
+    blr             x30
+
+.balign 16
+2:
+    mul             v3.8h, v3.8h, v19.8h
+    mul             v4.8h, v4.8h, v20.8h
+    mul             v5.8h, v5.8h, v21.8h
+    add             TMP4, xzr, TMP2, LSL #32
+    mul             v6.8h, v6.8h, v22.8h
+    mul             v7.8h, v7.8h, v23.8h
+    adds            TMP3, xzr, TMP2, LSR #32
+    mul             v8.8h, v8.8h, v24.8h
+    mul             v9.8h, v9.8h, v25.8h
+    b.ne            3f
+    /* Right AC coef is zero */
+    dup             v15.2d, v10.d[1]
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.4h, v4.4h, v8.4h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.4h, v2.4h, v6.4h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    sub             v26.4h, v2.4h, v6.4h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.4h, v9.4h, v5.4h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.4h, v7.4h, v3.4h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.4h, v9.4h, v3.4h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.4h, v7.4h, v5.4h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.4h, v22.4h, v24.4h  /* z5 = z3 + z4 */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    mov             v6.16b, v15.16b
+    mov             v7.16b, v15.16b
+    mov             v8.16b, v15.16b
+    mov             v9.16b, v15.16b
+    b               1b
+
+.balign 16
+3:
+    cbnz            TMP4, 4f
+    /* Left AC coef is zero */
+    dup             v14.2d, v10.d[0]
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    mov             v2.16b, v14.16b
+    mov             v3.16b, v14.16b
+    mov             v4.16b, v14.16b
+    mov             v5.16b, v14.16b
+    rshrn           v6.4h, v19.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+.balign 16
+4:
+    /* "No" AC coef is zero */
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn           v6.4h, v19.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+
+#undef CENTERJSAMPLE
+#undef CONST_BITS
+#undef PASS1_BITS
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_ifast_neon
+ *
+ * This function contains a fast, not so accurate integer implementation of
+ * the inverse DCT (Discrete Cosine Transform). It uses the same calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_ifast'
+ * function from jidctfst.c
+ *
+ * Normally 1-D AAN DCT needs 5 multiplications and 29 additions.
+ * But in ARM NEON case some extra additions are required because VQDMULH
+ * instruction can't handle the constants larger than 1. So the expressions
+ * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x",
+ * which introduces an extra addition. Overall, there are 6 extra additions
+ * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions.
+ */
+
+#define XFIX_1_082392200 v0.h[0]
+#define XFIX_1_414213562 v0.h[1]
+#define XFIX_1_847759065 v0.h[2]
+#define XFIX_2_613125930 v0.h[3]
+
+.balign 16
+Ljsimd_idct_ifast_neon_consts:
+  .short (277 * 128 - 256 * 128)  /* XFIX_1_082392200 */
+  .short (362 * 128 - 256 * 128)  /* XFIX_1_414213562 */
+  .short (473 * 128 - 256 * 128)  /* XFIX_1_847759065 */
+  .short (669 * 128 - 512 * 128)  /* XFIX_2_613125930 */
+
+asm_function jsimd_idct_ifast_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x9
+    TMP4            .req x10
+    TMP5            .req x11
+    TMP6            .req x12
+    TMP7            .req x13
+    TMP8            .req x14
+
+    /* Load and dequantize coefficients into NEON registers
+     * with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17     ( v16.8h )
+     *   1 | d18     | d19     ( v17.8h )
+     *   2 | d20     | d21     ( v18.8h )
+     *   3 | d22     | d23     ( v19.8h )
+     *   4 | d24     | d25     ( v20.8h )
+     *   5 | d26     | d27     ( v21.8h )
+     *   6 | d28     | d29     ( v22.8h )
+     *   7 | d30     | d31     ( v23.8h )
+     */
+    /* Save NEON registers used in fast IDCT */
+    adr             TMP5, Ljsimd_idct_ifast_neon_consts
+    ld1             {v16.8h, v17.8h}, [COEF_BLOCK], 32
+    ld1             {v0.8h, v1.8h}, [DCT_TABLE], 32
+    ld1             {v18.8h, v19.8h}, [COEF_BLOCK], 32
+    mul             v16.8h, v16.8h, v0.8h
+    ld1             {v2.8h, v3.8h}, [DCT_TABLE], 32
+    mul             v17.8h, v17.8h, v1.8h
+    ld1             {v20.8h, v21.8h}, [COEF_BLOCK], 32
+    mul             v18.8h, v18.8h, v2.8h
+    ld1             {v0.8h, v1.8h}, [DCT_TABLE], 32
+    mul             v19.8h, v19.8h, v3.8h
+    ld1             {v22.8h, v23.8h}, [COEF_BLOCK], 32
+    mul             v20.8h, v20.8h, v0.8h
+    ld1             {v2.8h, v3.8h}, [DCT_TABLE], 32
+    mul             v22.8h, v22.8h, v2.8h
+    mul             v21.8h, v21.8h, v1.8h
+    ld1             {v0.4h}, [TMP5]        /* load constants */
+    mul             v23.8h, v23.8h, v3.8h
+
+    /* 1-D IDCT, pass 1 */
+    sub             v2.8h, v18.8h, v22.8h
+    add             v22.8h, v18.8h, v22.8h
+    sub             v1.8h, v19.8h, v21.8h
+    add             v21.8h, v19.8h, v21.8h
+    sub             v5.8h, v17.8h, v23.8h
+    add             v23.8h, v17.8h, v23.8h
+    sqdmulh         v4.8h, v2.8h, XFIX_1_414213562
+    sqdmulh         v6.8h, v1.8h, XFIX_2_613125930
+    add             v3.8h, v1.8h, v1.8h
+    sub             v1.8h, v5.8h, v1.8h
+    add             v18.8h, v2.8h, v4.8h
+    sqdmulh         v4.8h, v1.8h, XFIX_1_847759065
+    sub             v2.8h, v23.8h, v21.8h
+    add             v3.8h, v3.8h, v6.8h
+    sqdmulh         v6.8h, v2.8h, XFIX_1_414213562
+    add             v1.8h, v1.8h, v4.8h
+    sqdmulh         v4.8h, v5.8h, XFIX_1_082392200
+    sub             v18.8h, v18.8h, v22.8h
+    add             v2.8h, v2.8h, v6.8h
+    sub             v6.8h, v16.8h, v20.8h
+    add             v20.8h, v16.8h, v20.8h
+    add             v17.8h, v5.8h, v4.8h
+    add             v5.8h, v6.8h, v18.8h
+    sub             v18.8h, v6.8h, v18.8h
+    add             v6.8h, v23.8h, v21.8h
+    add             v16.8h, v20.8h, v22.8h
+    sub             v3.8h, v6.8h, v3.8h
+    sub             v20.8h, v20.8h, v22.8h
+    sub             v3.8h, v3.8h, v1.8h
+    sub             v1.8h, v17.8h, v1.8h
+    add             v2.8h, v3.8h, v2.8h
+    sub             v23.8h, v16.8h, v6.8h
+    add             v1.8h, v1.8h, v2.8h
+    add             v16.8h, v16.8h, v6.8h
+    add             v22.8h, v5.8h, v3.8h
+    sub             v17.8h, v5.8h, v3.8h
+    sub             v21.8h, v18.8h, v2.8h
+    add             v18.8h, v18.8h, v2.8h
+    sub             v19.8h, v20.8h, v1.8h
+    add             v20.8h, v20.8h, v1.8h
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v28, v29, v30, v31
+    /* 1-D IDCT, pass 2 */
+    sub             v2.8h, v18.8h, v22.8h
+    add             v22.8h, v18.8h, v22.8h
+    sub             v1.8h, v19.8h, v21.8h
+    add             v21.8h, v19.8h, v21.8h
+    sub             v5.8h, v17.8h, v23.8h
+    add             v23.8h, v17.8h, v23.8h
+    sqdmulh         v4.8h, v2.8h, XFIX_1_414213562
+    sqdmulh         v6.8h, v1.8h, XFIX_2_613125930
+    add             v3.8h, v1.8h, v1.8h
+    sub             v1.8h, v5.8h, v1.8h
+    add             v18.8h, v2.8h, v4.8h
+    sqdmulh         v4.8h, v1.8h, XFIX_1_847759065
+    sub             v2.8h, v23.8h, v21.8h
+    add             v3.8h, v3.8h, v6.8h
+    sqdmulh         v6.8h, v2.8h, XFIX_1_414213562
+    add             v1.8h, v1.8h, v4.8h
+    sqdmulh         v4.8h, v5.8h, XFIX_1_082392200
+    sub             v18.8h, v18.8h, v22.8h
+    add             v2.8h, v2.8h, v6.8h
+    sub             v6.8h, v16.8h, v20.8h
+    add             v20.8h, v16.8h, v20.8h
+    add             v17.8h, v5.8h, v4.8h
+    add             v5.8h, v6.8h, v18.8h
+    sub             v18.8h, v6.8h, v18.8h
+    add             v6.8h, v23.8h, v21.8h
+    add             v16.8h, v20.8h, v22.8h
+    sub             v3.8h, v6.8h, v3.8h
+    sub             v20.8h, v20.8h, v22.8h
+    sub             v3.8h, v3.8h, v1.8h
+    sub             v1.8h, v17.8h, v1.8h
+    add             v2.8h, v3.8h, v2.8h
+    sub             v23.8h, v16.8h, v6.8h
+    add             v1.8h, v1.8h, v2.8h
+    add             v16.8h, v16.8h, v6.8h
+    add             v22.8h, v5.8h, v3.8h
+    sub             v17.8h, v5.8h, v3.8h
+    sub             v21.8h, v18.8h, v2.8h
+    add             v18.8h, v18.8h, v2.8h
+    sub             v19.8h, v20.8h, v1.8h
+    add             v20.8h, v20.8h, v1.8h
+    /* Descale to 8-bit and range limit */
+    movi            v0.16b, #0x80
+      /* Prepare pointers (dual-issue with NEON instructions) */
+      ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    sqshrn          v28.8b, v16.8h, #5
+      ldp             TMP3, TMP4, [OUTPUT_BUF], 16
+    sqshrn          v29.8b, v17.8h, #5
+      add             TMP1, TMP1, OUTPUT_COL
+    sqshrn          v30.8b, v18.8h, #5
+      add             TMP2, TMP2, OUTPUT_COL
+    sqshrn          v31.8b, v19.8h, #5
+      add             TMP3, TMP3, OUTPUT_COL
+    sqshrn2         v28.16b, v20.8h, #5
+      add             TMP4, TMP4, OUTPUT_COL
+    sqshrn2         v29.16b, v21.8h, #5
+      ldp             TMP5, TMP6, [OUTPUT_BUF], 16
+    sqshrn2         v30.16b, v22.8h, #5
+      ldp             TMP7, TMP8, [OUTPUT_BUF], 16
+    sqshrn2         v31.16b, v23.8h, #5
+      add             TMP5, TMP5, OUTPUT_COL
+    add             v16.16b, v28.16b, v0.16b
+      add             TMP6, TMP6, OUTPUT_COL
+    add             v18.16b, v29.16b, v0.16b
+      add             TMP7, TMP7, OUTPUT_COL
+    add             v20.16b, v30.16b, v0.16b
+      add             TMP8, TMP8, OUTPUT_COL
+    add             v22.16b, v31.16b, v0.16b
+
+    /* Transpose the final 8-bit samples */
+    trn1            v28.16b, v16.16b, v18.16b
+    trn1            v30.16b, v20.16b, v22.16b
+    trn2            v29.16b, v16.16b, v18.16b
+    trn2            v31.16b, v20.16b, v22.16b
+
+    trn1            v16.8h, v28.8h, v30.8h
+    trn2            v18.8h, v28.8h, v30.8h
+    trn1            v20.8h, v29.8h, v31.8h
+    trn2            v22.8h, v29.8h, v31.8h
+
+    uzp1            v28.4s, v16.4s, v18.4s
+    uzp2            v30.4s, v16.4s, v18.4s
+    uzp1            v29.4s, v20.4s, v22.4s
+    uzp2            v31.4s, v20.4s, v22.4s
+
+    /* Store results to the output buffer */
+    st1             {v28.d}[0], [TMP1]
+    st1             {v29.d}[0], [TMP2]
+    st1             {v28.d}[1], [TMP3]
+    st1             {v29.d}[1], [TMP4]
+    st1             {v30.d}[0], [TMP5]
+    st1             {v31.d}[0], [TMP6]
+    st1             {v30.d}[1], [TMP7]
+    st1             {v31.d}[1], [TMP8]
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_4x4_neon
+ *
+ * This function contains inverse-DCT code for getting reduced-size
+ * 4x4 pixels output from an 8x8 DCT block. It uses the same  calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_4x4'
+ * function from jpeg-6b (jidctred.c).
+ *
+ * NOTE: jpeg-8 has an improved implementation of 4x4 inverse-DCT, which
+ *       requires much less arithmetic operations and hence should be faster.
+ *       The primary purpose of this particular NEON optimized function is
+ *       bit exact compatibility with jpeg-6b.
+ *
+ * TODO: a bit better instructions scheduling can be achieved by expanding
+ *       idct_helper/transpose_4x4 macros and reordering instructions,
+ *       but readability will suffer somewhat.
+ */
+
+#define CONST_BITS  13
+
+#define FIX_0_211164243 (1730)   /* FIX(0.211164243) */
+#define FIX_0_509795579 (4176)   /* FIX(0.509795579) */
+#define FIX_0_601344887 (4926)   /* FIX(0.601344887) */
+#define FIX_0_720959822 (5906)   /* FIX(0.720959822) */
+#define FIX_0_765366865 (6270)   /* FIX(0.765366865) */
+#define FIX_0_850430095 (6967)   /* FIX(0.850430095) */
+#define FIX_0_899976223 (7373)   /* FIX(0.899976223) */
+#define FIX_1_061594337 (8697)   /* FIX(1.061594337) */
+#define FIX_1_272758580 (10426)  /* FIX(1.272758580) */
+#define FIX_1_451774981 (11893)  /* FIX(1.451774981) */
+#define FIX_1_847759065 (15137)  /* FIX(1.847759065) */
+#define FIX_2_172734803 (17799)  /* FIX(2.172734803) */
+#define FIX_2_562915447 (20995)  /* FIX(2.562915447) */
+#define FIX_3_624509785 (29692)  /* FIX(3.624509785) */
+
+.balign 16
+Ljsimd_idct_4x4_neon_consts:
+  .short FIX_1_847759065      /* v0.h[0] */
+  .short -FIX_0_765366865     /* v0.h[1] */
+  .short -FIX_0_211164243     /* v0.h[2] */
+  .short FIX_1_451774981      /* v0.h[3] */
+  .short -FIX_2_172734803     /* d1[0] */
+  .short FIX_1_061594337      /* d1[1] */
+  .short -FIX_0_509795579     /* d1[2] */
+  .short -FIX_0_601344887     /* d1[3] */
+  .short FIX_0_899976223      /* v2.h[0] */
+  .short FIX_2_562915447      /* v2.h[1] */
+  .short 1 << (CONST_BITS+1)  /* v2.h[2] */
+  .short 0                    /* v2.h[3] */
+
+.macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29
+    smull           v28.4s, \x4, v2.h[2]
+    smlal           v28.4s, \x8, v0.h[0]
+    smlal           v28.4s, \x14, v0.h[1]
+
+    smull           v26.4s, \x16, v1.h[2]
+    smlal           v26.4s, \x12, v1.h[3]
+    smlal           v26.4s, \x10, v2.h[0]
+    smlal           v26.4s, \x6, v2.h[1]
+
+    smull           v30.4s, \x4, v2.h[2]
+    smlsl           v30.4s, \x8, v0.h[0]
+    smlsl           v30.4s, \x14, v0.h[1]
+
+    smull           v24.4s, \x16, v0.h[2]
+    smlal           v24.4s, \x12, v0.h[3]
+    smlal           v24.4s, \x10, v1.h[0]
+    smlal           v24.4s, \x6, v1.h[1]
+
+    add             v20.4s, v28.4s, v26.4s
+    sub             v28.4s, v28.4s, v26.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v28.4s, v28.4s, #\shift
+    xtn             \y26, v20.4s
+    xtn             \y29, v28.4s
+  .else
+    rshrn           \y26, v20.4s, #\shift
+    rshrn           \y29, v28.4s, #\shift
+  .endif
+
+    add             v20.4s, v30.4s, v24.4s
+    sub             v30.4s, v30.4s, v24.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v30.4s, v30.4s, #\shift
+    xtn             \y27, v20.4s
+    xtn             \y28, v30.4s
+  .else
+    rshrn           \y27, v20.4s, #\shift
+    rshrn           \y28, v30.4s, #\shift
+  .endif
+.endm
+
+asm_function jsimd_idct_4x4_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x2
+    TMP4            .req x15
+
+    /* Save all used NEON registers */
+    sub             sp, sp, 272
+    str             x15, [sp], 16
+    /* Load constants (v3.4h is just used for padding) */
+    adr             TMP4, Ljsimd_idct_4x4_neon_consts
+    st1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    ld1             {v0.4h, v1.4h, v2.4h, v3.4h}, [TMP4]
+
+    /* Load all COEF_BLOCK into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | v4.4h   | v5.4h
+     *   1 | v6.4h   | v7.4h
+     *   2 | v8.4h   | v9.4h
+     *   3 | v10.4h  | v11.4h
+     *   4 | -       | -
+     *   5 | v12.4h  | v13.4h
+     *   6 | v14.4h  | v15.4h
+     *   7 | v16.4h  | v17.4h
+     */
+    ld1             {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32
+    ld1             {v8.4h, v9.4h, v10.4h, v11.4h}, [COEF_BLOCK], 32
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v12.4h, v13.4h, v14.4h, v15.4h}, [COEF_BLOCK], 32
+    ld1             {v16.4h, v17.4h}, [COEF_BLOCK], 16
+    /* dequantize */
+    ld1             {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32
+    mul             v4.4h, v4.4h, v18.4h
+    mul             v5.4h, v5.4h, v19.4h
+    ins             v4.d[1], v5.d[0]              /* 128 bit q4 */
+    ld1             {v22.4h, v23.4h, v24.4h, v25.4h}, [DCT_TABLE], 32
+    mul             v6.4h, v6.4h, v20.4h
+    mul             v7.4h, v7.4h, v21.4h
+    ins             v6.d[1], v7.d[0]              /* 128 bit q6 */
+    mul             v8.4h, v8.4h, v22.4h
+    mul             v9.4h, v9.4h, v23.4h
+    ins             v8.d[1], v9.d[0]              /* 128 bit q8 */
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v26.4h, v27.4h, v28.4h, v29.4h}, [DCT_TABLE], 32
+    mul             v10.4h, v10.4h, v24.4h
+    mul             v11.4h, v11.4h, v25.4h
+    ins             v10.d[1], v11.d[0]            /* 128 bit q10 */
+    mul             v12.4h, v12.4h, v26.4h
+    mul             v13.4h, v13.4h, v27.4h
+    ins             v12.d[1], v13.d[0]            /* 128 bit q12 */
+    ld1             {v30.4h, v31.4h}, [DCT_TABLE], 16
+    mul             v14.4h, v14.4h, v28.4h
+    mul             v15.4h, v15.4h, v29.4h
+    ins             v14.d[1], v15.d[0]            /* 128 bit q14 */
+    mul             v16.4h, v16.4h, v30.4h
+    mul             v17.4h, v17.4h, v31.4h
+    ins             v16.d[1], v17.d[0]            /* 128 bit q16 */
+
+    /* Pass 1 */
+    idct_helper     v4.4h, v6.4h, v8.4h, v10.4h, v12.4h, v14.4h, v16.4h, 12, \
+                    v4.4h, v6.4h, v8.4h, v10.4h
+    transpose_4x4   v4, v6, v8, v10, v3
+    ins             v10.d[1], v11.d[0]
+    idct_helper     v5.4h, v7.4h, v9.4h, v11.4h, v13.4h, v15.4h, v17.4h, 12, \
+                    v5.4h, v7.4h, v9.4h, v11.4h
+    transpose_4x4   v5, v7, v9, v11, v3
+    ins             v10.d[1], v11.d[0]
+
+    /* Pass 2 */
+    idct_helper     v4.4h, v6.4h, v8.4h, v10.4h, v7.4h, v9.4h, v11.4h, 19, \
+                    v26.4h, v27.4h, v28.4h, v29.4h
+    transpose_4x4   v26, v27, v28, v29, v3
+
+    /* Range limit */
+    movi            v30.8h, #0x80
+    ins             v26.d[1], v27.d[0]
+    ins             v28.d[1], v29.d[0]
+    add             v26.8h, v26.8h, v30.8h
+    add             v28.8h, v28.8h, v30.8h
+    sqxtun          v26.8b, v26.8h
+    sqxtun          v27.8b, v28.8h
+
+    /* Store results to the output buffer */
+    ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    ldp             TMP3, TMP4, [OUTPUT_BUF]
+    add             TMP1, TMP1, OUTPUT_COL
+    add             TMP2, TMP2, OUTPUT_COL
+    add             TMP3, TMP3, OUTPUT_COL
+    add             TMP4, TMP4, OUTPUT_COL
+
+#if defined(__ARMEL__) && !RESPECT_STRICT_ALIGNMENT
+    /* We can use much less instructions on little endian systems if the
+     * OS kernel is not configured to trap unaligned memory accesses
+     */
+    st1             {v26.s}[0], [TMP1], 4
+    st1             {v27.s}[0], [TMP3], 4
+    st1             {v26.s}[1], [TMP2], 4
+    st1             {v27.s}[1], [TMP4], 4
+#else
+    st1             {v26.b}[0], [TMP1], 1
+    st1             {v27.b}[0], [TMP3], 1
+    st1             {v26.b}[1], [TMP1], 1
+    st1             {v27.b}[1], [TMP3], 1
+    st1             {v26.b}[2], [TMP1], 1
+    st1             {v27.b}[2], [TMP3], 1
+    st1             {v26.b}[3], [TMP1], 1
+    st1             {v27.b}[3], [TMP3], 1
+
+    st1             {v26.b}[4], [TMP2], 1
+    st1             {v27.b}[4], [TMP4], 1
+    st1             {v26.b}[5], [TMP2], 1
+    st1             {v27.b}[5], [TMP4], 1
+    st1             {v26.b}[6], [TMP2], 1
+    st1             {v27.b}[6], [TMP4], 1
+    st1             {v26.b}[7], [TMP2], 1
+    st1             {v27.b}[7], [TMP4], 1
+#endif
+
+    /* vpop            {v8.4h - v15.4h}    ;not available */
+    sub             sp, sp, #272
+    ldr             x15, [sp], 16
+    ld1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+
+.purgem idct_helper
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_2x2_neon
+ *
+ * This function contains inverse-DCT code for getting reduced-size
+ * 2x2 pixels output from an 8x8 DCT block. It uses the same  calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_2x2'
+ * function from jpeg-6b (jidctred.c).
+ *
+ * NOTE: jpeg-8 has an improved implementation of 2x2 inverse-DCT, which
+ *       requires much less arithmetic operations and hence should be faster.
+ *       The primary purpose of this particular NEON optimized function is
+ *       bit exact compatibility with jpeg-6b.
+ */
+
+.balign 8
+Ljsimd_idct_2x2_neon_consts:
+  .short -FIX_0_720959822  /* v14[0] */
+  .short FIX_0_850430095   /* v14[1] */
+  .short -FIX_1_272758580  /* v14[2] */
+  .short FIX_3_624509785   /* v14[3] */
+
+.macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27
+    sshll           v15.4s, \x4, #15
+    smull           v26.4s, \x6, v14.h[3]
+    smlal           v26.4s, \x10, v14.h[2]
+    smlal           v26.4s, \x12, v14.h[1]
+    smlal           v26.4s, \x16, v14.h[0]
+
+    add             v20.4s, v15.4s, v26.4s
+    sub             v15.4s, v15.4s, v26.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v15.4s, v15.4s, #\shift
+    xtn             \y26, v20.4s
+    xtn             \y27, v15.4s
+  .else
+    rshrn           \y26, v20.4s, #\shift
+    rshrn           \y27, v15.4s, #\shift
+  .endif
+.endm
+
+asm_function jsimd_idct_2x2_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x15
+
+    /* vpush           {v8.4h - v15.4h}            ; not available */
+    sub             sp, sp, 208
+    str             x15, [sp], 16
+
+    /* Load constants */
+    adr             TMP2, Ljsimd_idct_2x2_neon_consts
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v21.8b, v22.8b}, [sp], 16
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v30.8b, v31.8b}, [sp], 16
+    ld1             {v14.4h}, [TMP2]
+
+    /* Load all COEF_BLOCK into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | v4.4h   | v5.4h
+     *   1 | v6.4h   | v7.4h
+     *   2 | -       | -
+     *   3 | v10.4h  | v11.4h
+     *   4 | -       | -
+     *   5 | v12.4h  | v13.4h
+     *   6 | -       | -
+     *   7 | v16.4h  | v17.4h
+     */
+    ld1             {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v10.4h, v11.4h}, [COEF_BLOCK], 16
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v12.4h, v13.4h}, [COEF_BLOCK], 16
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v16.4h, v17.4h}, [COEF_BLOCK], 16
+    /* Dequantize */
+    ld1             {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32
+    mul             v4.4h, v4.4h, v18.4h
+    mul             v5.4h, v5.4h, v19.4h
+    ins             v4.d[1], v5.d[0]
+    mul             v6.4h, v6.4h, v20.4h
+    mul             v7.4h, v7.4h, v21.4h
+    ins             v6.d[1], v7.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v24.4h, v25.4h}, [DCT_TABLE], 16
+    mul             v10.4h, v10.4h, v24.4h
+    mul             v11.4h, v11.4h, v25.4h
+    ins             v10.d[1], v11.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v26.4h, v27.4h}, [DCT_TABLE], 16
+    mul             v12.4h, v12.4h, v26.4h
+    mul             v13.4h, v13.4h, v27.4h
+    ins             v12.d[1], v13.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v30.4h, v31.4h}, [DCT_TABLE], 16
+    mul             v16.4h, v16.4h, v30.4h
+    mul             v17.4h, v17.4h, v31.4h
+    ins             v16.d[1], v17.d[0]
+
+    /* Pass 1 */
+#if 0
+    idct_helper     v4.4h, v6.4h, v10.4h, v12.4h, v16.4h, 13, v4.4h, v6.4h
+    transpose_4x4   v4.4h, v6.4h, v8.4h, v10.4h
+    idct_helper     v5.4h, v7.4h, v11.4h, v13.4h, v17.4h, 13, v5.4h, v7.4h
+    transpose_4x4   v5.4h, v7.4h, v9.4h, v11.4h
+#else
+    smull           v26.4s, v6.4h, v14.h[3]
+    smlal           v26.4s, v10.4h, v14.h[2]
+    smlal           v26.4s, v12.4h, v14.h[1]
+    smlal           v26.4s, v16.4h, v14.h[0]
+    smull           v24.4s, v7.4h, v14.h[3]
+    smlal           v24.4s, v11.4h, v14.h[2]
+    smlal           v24.4s, v13.4h, v14.h[1]
+    smlal           v24.4s, v17.4h, v14.h[0]
+    sshll           v15.4s, v4.4h, #15
+    sshll           v30.4s, v5.4h, #15
+    add             v20.4s, v15.4s, v26.4s
+    sub             v15.4s, v15.4s, v26.4s
+    rshrn           v4.4h, v20.4s, #13
+    rshrn           v6.4h, v15.4s, #13
+    add             v20.4s, v30.4s, v24.4s
+    sub             v15.4s, v30.4s, v24.4s
+    rshrn           v5.4h, v20.4s, #13
+    rshrn           v7.4h, v15.4s, #13
+    ins             v4.d[1], v5.d[0]
+    ins             v6.d[1], v7.d[0]
+    transpose       v4, v6, v3, .16b, .8h
+    transpose       v6, v10, v3, .16b, .4s
+    ins             v11.d[0], v10.d[1]
+    ins             v7.d[0], v6.d[1]
+#endif
+
+    /* Pass 2 */
+    idct_helper     v4.4h, v6.4h, v10.4h, v7.4h, v11.4h, 20, v26.4h, v27.4h
+
+    /* Range limit */
+    movi            v30.8h, #0x80
+    ins             v26.d[1], v27.d[0]
+    add             v26.8h, v26.8h, v30.8h
+    sqxtun          v30.8b, v26.8h
+    ins             v26.d[0], v30.d[0]
+    sqxtun          v27.8b, v26.8h
+
+    /* Store results to the output buffer */
+    ldp             TMP1, TMP2, [OUTPUT_BUF]
+    add             TMP1, TMP1, OUTPUT_COL
+    add             TMP2, TMP2, OUTPUT_COL
+
+    st1             {v26.b}[0], [TMP1], 1
+    st1             {v27.b}[4], [TMP1], 1
+    st1             {v26.b}[1], [TMP2], 1
+    st1             {v27.b}[5], [TMP2], 1
+
+    sub             sp, sp, #208
+    ldr             x15, [sp], 16
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v21.8b, v22.8b}, [sp], 16
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v30.8b, v31.8b}, [sp], 16
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+
+.purgem idct_helper
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_ycc_extrgb_convert_neon
+ * jsimd_ycc_extbgr_convert_neon
+ * jsimd_ycc_extrgbx_convert_neon
+ * jsimd_ycc_extbgrx_convert_neon
+ * jsimd_ycc_extxbgr_convert_neon
+ * jsimd_ycc_extxrgb_convert_neon
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro do_load size
+  .if \size == 8
+    ld1             {v4.8b}, [U], 8
+    ld1             {v5.8b}, [V], 8
+    ld1             {v0.8b}, [Y], 8
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+  .elseif \size == 4
+    ld1             {v4.b}[0], [U], 1
+    ld1             {v4.b}[1], [U], 1
+    ld1             {v4.b}[2], [U], 1
+    ld1             {v4.b}[3], [U], 1
+    ld1             {v5.b}[0], [V], 1
+    ld1             {v5.b}[1], [V], 1
+    ld1             {v5.b}[2], [V], 1
+    ld1             {v5.b}[3], [V], 1
+    ld1             {v0.b}[0], [Y], 1
+    ld1             {v0.b}[1], [Y], 1
+    ld1             {v0.b}[2], [Y], 1
+    ld1             {v0.b}[3], [Y], 1
+  .elseif \size == 2
+    ld1             {v4.b}[4], [U], 1
+    ld1             {v4.b}[5], [U], 1
+    ld1             {v5.b}[4], [V], 1
+    ld1             {v5.b}[5], [V], 1
+    ld1             {v0.b}[4], [Y], 1
+    ld1             {v0.b}[5], [Y], 1
+  .elseif \size == 1
+    ld1             {v4.b}[6], [U], 1
+    ld1             {v5.b}[6], [V], 1
+    ld1             {v0.b}[6], [Y], 1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_store bpp, size, fast_st3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_st3 == 1
+        st3         {v10.8b, v11.8b, v12.8b}, [RGB], 24
+      .else
+        st1         {v10.b}[0], [RGB], #1
+        st1         {v11.b}[0], [RGB], #1
+        st1         {v12.b}[0], [RGB], #1
+
+        st1         {v10.b}[1], [RGB], #1
+        st1         {v11.b}[1], [RGB], #1
+        st1         {v12.b}[1], [RGB], #1
+
+        st1         {v10.b}[2], [RGB], #1
+        st1         {v11.b}[2], [RGB], #1
+        st1         {v12.b}[2], [RGB], #1
+
+        st1         {v10.b}[3], [RGB], #1
+        st1         {v11.b}[3], [RGB], #1
+        st1         {v12.b}[3], [RGB], #1
+
+        st1         {v10.b}[4], [RGB], #1
+        st1         {v11.b}[4], [RGB], #1
+        st1         {v12.b}[4], [RGB], #1
+
+        st1         {v10.b}[5], [RGB], #1
+        st1         {v11.b}[5], [RGB], #1
+        st1         {v12.b}[5], [RGB], #1
+
+        st1         {v10.b}[6], [RGB], #1
+        st1         {v11.b}[6], [RGB], #1
+        st1         {v12.b}[6], [RGB], #1
+
+        st1         {v10.b}[7], [RGB], #1
+        st1         {v11.b}[7], [RGB], #1
+        st1         {v12.b}[7], [RGB], #1
+      .endif
+    .elseif \size == 4
+      st3           {v10.b, v11.b, v12.b}[0], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[1], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[2], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[3], [RGB], 3
+    .elseif \size == 2
+      st3           {v10.b, v11.b, v12.b}[4], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[5], [RGB], 3
+    .elseif \size == 1
+      st3           {v10.b, v11.b, v12.b}[6], [RGB], 3
+    .else
+     .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      st4           {v10.8b, v11.8b, v12.8b, v13.8b}, [RGB], 32
+    .elseif \size == 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[0], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[1], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[2], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[3], [RGB], 4
+    .elseif \size == 2
+      st4           {v10.b, v11.b, v12.b, v13.b}[4], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[5], [RGB], 4
+    .elseif \size == 1
+      st4           {v10.b, v11.b, v12.b, v13.b}[6], [RGB], 4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp==16
+    .if \size == 8
+      st1           {v25.8h}, [RGB], 16
+    .elseif \size == 4
+      st1           {v25.4h}, [RGB], 8
+    .elseif \size == 2
+      st1           {v25.h}[4], [RGB], 2
+      st1           {v25.h}[5], [RGB], 2
+    .elseif \size == 1
+      st1           {v25.h}[6], [RGB], 2
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, rsize, \
+                                           g_offs, gsize, b_offs, bsize, \
+                                           defsize, fast_st3
+
+/*
+ * 2-stage pipelined YCbCr->RGB conversion
+ */
+
+.macro do_yuv_to_rgb_stage1
+    uaddw           v6.8h, v2.8h, v4.8b     /* q3 = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+    smull           v28.4s, v6.4h, v1.h[3]  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h[3]  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb_stage2
+    rshrn           v20.4h, v20.4s, #15
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16
+    sqxtun          v1\g_offs\defsize, v20.8h
+    sqxtun          v1\r_offs\defsize, v24.8h
+    sqxtun          v1\b_offs\defsize, v28.8h
+  .else
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    sri             v25.8h, v21.8h, #5
+    sri             v25.8h, v29.8h, #11
+  .endif
+.endm
+
+.macro do_yuv_to_rgb_stage2_store_load_stage1 fast_st3
+    rshrn           v20.4h, v20.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    ld1             {v4.8b}, [U], 8
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    ld1             {v5.8b}, [V], 8
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16  /**************** rgb24/rgb32 ******************************/
+    sqxtun          v1\g_offs\defsize, v20.8h
+    ld1             {v0.8b}, [Y], 8
+    sqxtun          v1\r_offs\defsize, v24.8h
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+    sqxtun          v1\b_offs\defsize, v28.8h
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+  .else  /**************************** rgb565 ********************************/
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    ld1             {v0.8b}, [Y], 8
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    sri             v25.8h, v21.8h, #5
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+    sri             v25.8h, v29.8h, #11
+  .endif
+    do_store        \bpp, 8, \fast_st3
+    smull           v28.4s, v6.4h, v1.h[3]  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h[3]  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb
+    do_yuv_to_rgb_stage1
+    do_yuv_to_rgb_stage2
+.endm
+
+/* Apple gas crashes on adrl, work around that by using adr.
+ * But this requires a copy of these constants for each function.
+ */
+
+.balign 16
+.if \fast_st3 == 1
+Ljsimd_ycc_\colorid\()_neon_consts:
+.else
+Ljsimd_ycc_\colorid\()_neon_slowst3_consts:
+.endif
+  .short 0,      0,     0,      0
+  .short 22971, -11277, -23401, 29033
+  .short -128,  -128,   -128,   -128
+  .short -128,  -128,   -128,   -128
+
+.if \fast_st3 == 1
+asm_function jsimd_ycc_\colorid\()_convert_neon
+.else
+asm_function jsimd_ycc_\colorid\()_convert_neon_slowst3
+.endif
+    OUTPUT_WIDTH    .req x0
+    INPUT_BUF       .req x1
+    INPUT_ROW       .req x2
+    OUTPUT_BUF      .req x3
+    NUM_ROWS        .req x4
+
+    INPUT_BUF0      .req x5
+    INPUT_BUF1      .req x6
+    INPUT_BUF2      .req x1
+
+    RGB             .req x7
+    Y               .req x8
+    U               .req x9
+    V               .req x10
+    N               .req x15
+
+    sub             sp, sp, 336
+    str             x15, [sp], 16
+
+    /* Load constants to d1, d2, d3 (v0.4h is just used for padding) */
+    .if \fast_st3 == 1
+      adr           x15, Ljsimd_ycc_\colorid\()_neon_consts
+    .else
+      adr           x15, Ljsimd_ycc_\colorid\()_neon_slowst3_consts
+    .endif
+
+    /* Save NEON registers */
+    st1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    ld1             {v0.4h, v1.4h}, [x15], 16
+    ld1             {v2.8h}, [x15]
+
+    /* Save ARM registers and handle input arguments */
+    /* push            {x4, x5, x6, x7, x8, x9, x10, x30} */
+    stp             x4, x5, [sp], 16
+    stp             x6, x7, [sp], 16
+    stp             x8, x9, [sp], 16
+    stp             x10, x30, [sp], 16
+    ldr             INPUT_BUF0, [INPUT_BUF]
+    ldr             INPUT_BUF1, [INPUT_BUF, #8]
+    ldr             INPUT_BUF2, [INPUT_BUF, #16]
+    .unreq          INPUT_BUF
+
+    /* Initially set v10, v11.4h, v12.8b, d13 to 0xFF */
+    movi            v10.16b, #255
+    movi            v13.16b, #255
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    lsl             x16, INPUT_ROW, #3
+    ldr             Y, [INPUT_BUF0, x16]
+    ldr             U, [INPUT_BUF1, x16]
+    mov             N, OUTPUT_WIDTH
+    ldr             V, [INPUT_BUF2, x16]
+    add             INPUT_ROW, INPUT_ROW, #1
+    ldr             RGB, [OUTPUT_BUF], #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         8
+    do_yuv_to_rgb_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_yuv_to_rgb_stage2_store_load_stage1 \fast_st3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_yuv_to_rgb_stage2
+    do_store        \bpp, 8, \fast_st3
+    tst             N, #7
+    b.eq            8f
+3:
+    tst             N, #4
+    b.eq            3f
+    do_load         4
+3:
+    tst             N, #2
+    b.eq            4f
+    do_load         2
+4:
+    tst             N, #1
+    b.eq            5f
+    do_load         1
+5:
+    do_yuv_to_rgb
+    tst             N, #4
+    b.eq            6f
+    do_store        \bpp, 4, \fast_st3
+6:
+    tst             N, #2
+    b.eq            7f
+    do_store        \bpp, 2, \fast_st3
+7:
+    tst             N, #1
+    b.eq            8f
+    do_store        \bpp, 1, \fast_st3
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    sub             sp, sp, #336
+    ldr             x15, [sp], 16
+    ld1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    /* pop             {r4, r5, r6, r7, r8, r9, r10, pc} */
+    ldp             x4, x5, [sp], 16
+    ldp             x6, x7, [sp], 16
+    ldp             x8, x9, [sp], 16
+    ldp             x10, x30, [sp], 16
+    br              x30
+    .unreq          OUTPUT_WIDTH
+    .unreq          INPUT_ROW
+    .unreq          OUTPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          INPUT_BUF0
+    .unreq          INPUT_BUF1
+    .unreq          INPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_yuv_to_rgb
+.purgem do_yuv_to_rgb_stage1
+.purgem do_yuv_to_rgb_stage2
+.purgem do_yuv_to_rgb_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  rsize G  gsize B  bsize defsize fast_st3*/
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extrgbx, 32, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, .4h,  2, .4h,  1, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, .4h,  2, .4h,  3, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon rgb565,  16, 0, .4h,  0, .4h,  0, .4h,  .8b,    1
+
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    0
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_extrgb_ycc_convert_neon
+ * jsimd_extbgr_ycc_convert_neon
+ * jsimd_extrgbx_ycc_convert_neon
+ * jsimd_extbgrx_ycc_convert_neon
+ * jsimd_extxbgr_ycc_convert_neon
+ * jsimd_extxrgb_ycc_convert_neon
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro do_store size
+  .if \size == 8
+    st1             {v20.8b}, [Y], #8
+    st1             {v21.8b}, [U], #8
+    st1             {v22.8b}, [V], #8
+  .elseif \size == 4
+    st1             {v20.b}[0], [Y], #1
+    st1             {v20.b}[1], [Y], #1
+    st1             {v20.b}[2], [Y], #1
+    st1             {v20.b}[3], [Y], #1
+    st1             {v21.b}[0], [U], #1
+    st1             {v21.b}[1], [U], #1
+    st1             {v21.b}[2], [U], #1
+    st1             {v21.b}[3], [U], #1
+    st1             {v22.b}[0], [V], #1
+    st1             {v22.b}[1], [V], #1
+    st1             {v22.b}[2], [V], #1
+    st1             {v22.b}[3], [V], #1
+  .elseif \size == 2
+    st1             {v20.b}[4], [Y], #1
+    st1             {v20.b}[5], [Y], #1
+    st1             {v21.b}[4], [U], #1
+    st1             {v21.b}[5], [U], #1
+    st1             {v22.b}[4], [V], #1
+    st1             {v22.b}[5], [V], #1
+  .elseif \size == 1
+    st1             {v20.b}[6], [Y], #1
+    st1             {v21.b}[6], [U], #1
+    st1             {v22.b}[6], [V], #1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size, fast_ld3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_ld3 == 1
+        ld3         {v10.8b, v11.8b, v12.8b}, [RGB], #24
+      .else
+        ld1         {v10.b}[0], [RGB], #1
+        ld1         {v11.b}[0], [RGB], #1
+        ld1         {v12.b}[0], [RGB], #1
+
+        ld1         {v10.b}[1], [RGB], #1
+        ld1         {v11.b}[1], [RGB], #1
+        ld1         {v12.b}[1], [RGB], #1
+
+        ld1         {v10.b}[2], [RGB], #1
+        ld1         {v11.b}[2], [RGB], #1
+        ld1         {v12.b}[2], [RGB], #1
+
+        ld1         {v10.b}[3], [RGB], #1
+        ld1         {v11.b}[3], [RGB], #1
+        ld1         {v12.b}[3], [RGB], #1
+
+        ld1         {v10.b}[4], [RGB], #1
+        ld1         {v11.b}[4], [RGB], #1
+        ld1         {v12.b}[4], [RGB], #1
+
+        ld1         {v10.b}[5], [RGB], #1
+        ld1         {v11.b}[5], [RGB], #1
+        ld1         {v12.b}[5], [RGB], #1
+
+        ld1         {v10.b}[6], [RGB], #1
+        ld1         {v11.b}[6], [RGB], #1
+        ld1         {v12.b}[6], [RGB], #1
+
+        ld1         {v10.b}[7], [RGB], #1
+        ld1         {v11.b}[7], [RGB], #1
+        ld1         {v12.b}[7], [RGB], #1
+      .endif
+      prfm          pldl1keep, [RGB, #128]
+    .elseif \size == 4
+      ld3           {v10.b, v11.b, v12.b}[0], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[1], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[2], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[3], [RGB], #3
+    .elseif \size == 2
+      ld3           {v10.b, v11.b, v12.b}[4], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[5], [RGB], #3
+    .elseif \size == 1
+      ld3           {v10.b, v11.b, v12.b}[6], [RGB], #3
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      ld4           {v10.8b, v11.8b, v12.8b, v13.8b}, [RGB], #32
+      prfm          pldl1keep, [RGB, #128]
+    .elseif \size == 4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[0], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[1], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[2], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[3], [RGB], #4
+    .elseif \size == 2
+      ld4           {v10.b, v11.b, v12.b, v13.b}[4], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[5], [RGB], #4
+    .elseif \size == 1
+      ld4           {v10.b, v11.b, v12.b, v13.b}[6], [RGB], #4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, \
+                                           b_offs, fast_ld3
+
+/*
+ * 2-stage pipelined RGB->YCbCr conversion
+ */
+
+.macro do_rgb_to_yuv_stage1
+    ushll           v4.8h, v1\r_offs\().8b, #0  /* r = v4 */
+    ushll           v6.8h, v1\g_offs\().8b, #0  /* g = v6 */
+    ushll           v8.8h, v1\b_offs\().8b, #0  /* b = v8 */
+    rev64           v18.4s, v1.4s
+    rev64           v26.4s, v1.4s
+    rev64           v28.4s, v1.4s
+    rev64           v30.4s, v1.4s
+    umull           v14.4s, v4.4h, v0.h[0]
+    umull2          v16.4s, v4.8h, v0.h[0]
+    umlsl           v18.4s, v4.4h, v0.h[3]
+    umlsl2          v26.4s, v4.8h, v0.h[3]
+    umlal           v28.4s, v4.4h, v0.h[5]
+    umlal2          v30.4s, v4.8h, v0.h[5]
+    umlal           v14.4s, v6.4h, v0.h[1]
+    umlal2          v16.4s, v6.8h, v0.h[1]
+    umlsl           v18.4s, v6.4h, v0.h[4]
+    umlsl2          v26.4s, v6.8h, v0.h[4]
+    umlsl           v28.4s, v6.4h, v0.h[6]
+    umlsl2          v30.4s, v6.8h, v0.h[6]
+    umlal           v14.4s, v8.4h, v0.h[2]
+    umlal2          v16.4s, v8.8h, v0.h[2]
+    umlal           v18.4s, v8.4h, v0.h[5]
+    umlal2          v26.4s, v8.8h, v0.h[5]
+    umlsl           v28.4s, v8.4h, v0.h[7]
+    umlsl2          v30.4s, v8.8h, v0.h[7]
+.endm
+
+.macro do_rgb_to_yuv_stage2
+    rshrn           v20.4h, v14.4s, #16
+    shrn            v22.4h, v18.4s, #16
+    shrn            v24.4h, v28.4s, #16
+    rshrn2          v20.8h, v16.4s, #16
+    shrn2           v22.8h, v26.4s, #16
+    shrn2           v24.8h, v30.4s, #16
+    xtn             v20.8b, v20.8h       /* v20 = y */
+    xtn             v21.8b, v22.8h       /* v21 = u */
+    xtn             v22.8b, v24.8h       /* v22 = v */
+.endm
+
+.macro do_rgb_to_yuv
+    do_rgb_to_yuv_stage1
+    do_rgb_to_yuv_stage2
+.endm
+
+/* TODO: expand macros and interleave instructions if some in-order
+ *       ARM64 processor actually can dual-issue LOAD/STORE with ALU */
+.macro do_rgb_to_yuv_stage2_store_load_stage1 fast_ld3
+    do_rgb_to_yuv_stage2
+    do_load         \bpp, 8, \fast_ld3
+    st1             {v20.8b}, [Y], #8
+    st1             {v21.8b}, [U], #8
+    st1             {v22.8b}, [V], #8
+    do_rgb_to_yuv_stage1
+.endm
+
+.balign 16
+.if \fast_ld3 == 1
+Ljsimd_\colorid\()_ycc_neon_consts:
+.else
+Ljsimd_\colorid\()_ycc_neon_slowld3_consts:
+.endif
+  .short 19595, 38470, 7471, 11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128, 32767, 128
+  .short 32767, 128, 32767, 128
+
+.if \fast_ld3 == 1
+asm_function jsimd_\colorid\()_ycc_convert_neon
+.else
+asm_function jsimd_\colorid\()_ycc_convert_neon_slowld3
+.endif
+    OUTPUT_WIDTH    .req w0
+    INPUT_BUF       .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_ROW      .req x3
+    NUM_ROWS        .req x4
+
+    OUTPUT_BUF0     .req x5
+    OUTPUT_BUF1     .req x6
+    OUTPUT_BUF2     .req x2  /* OUTPUT_BUF */
+
+    RGB             .req x7
+    Y               .req x9
+    U               .req x10
+    V               .req x11
+    N               .req w12
+
+    /* Load constants to d0, d1, d2, d3 */
+    .if \fast_ld3 == 1
+      adr           x13, Ljsimd_\colorid\()_ycc_neon_consts
+    .else
+      adr           x13, Ljsimd_\colorid\()_ycc_neon_slowld3_consts
+    .endif
+    ld1             {v0.8h, v1.8h}, [x13]
+
+    ldr             OUTPUT_BUF0, [OUTPUT_BUF]
+    ldr             OUTPUT_BUF1, [OUTPUT_BUF, #8]
+    ldr             OUTPUT_BUF2, [OUTPUT_BUF, #16]
+    .unreq          OUTPUT_BUF
+
+    /* Save NEON registers */
+    sub             sp, sp, #64
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    ldr             Y, [OUTPUT_BUF0, OUTPUT_ROW, lsl #3]
+    ldr             U, [OUTPUT_BUF1, OUTPUT_ROW, lsl #3]
+    mov             N, OUTPUT_WIDTH
+    ldr             V, [OUTPUT_BUF2, OUTPUT_ROW, lsl #3]
+    add             OUTPUT_ROW, OUTPUT_ROW, #1
+    ldr             RGB, [INPUT_BUF], #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         \bpp, 8, \fast_ld3
+    do_rgb_to_yuv_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_rgb_to_yuv_stage2_store_load_stage1 \fast_ld3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_rgb_to_yuv_stage2
+    do_store        8
+    tst             N, #7
+    b.eq            8f
+3:
+    tbz             N, #2, 3f
+    do_load         \bpp, 4, \fast_ld3
+3:
+    tbz             N, #1, 4f
+    do_load         \bpp, 2, \fast_ld3
+4:
+    tbz             N, #0, 5f
+    do_load         \bpp, 1, \fast_ld3
+5:
+    do_rgb_to_yuv
+    tbz             N, #2, 6f
+    do_store        4
+6:
+    tbz             N, #1, 7f
+    do_store        2
+7:
+    tbz             N, #0, 8f
+    do_store        1
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    sub             sp, sp, #64
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    br              x30
+
+    .unreq          OUTPUT_WIDTH
+    .unreq          OUTPUT_ROW
+    .unreq          INPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          OUTPUT_BUF0
+    .unreq          OUTPUT_BUF1
+    .unreq          OUTPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_rgb_to_yuv
+.purgem do_rgb_to_yuv_stage1
+.purgem do_rgb_to_yuv_stage2
+.purgem do_rgb_to_yuv_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  G  B  Fast LD3 */
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extrgbx, 32, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgrx, 32, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extxbgr, 32, 3, 2, 1, 1
+generate_jsimd_rgb_ycc_convert_neon extxrgb, 32, 1, 2, 3, 1
+
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 0
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * Load data into workspace, applying unsigned->signed conversion
+ *
+ * TODO: can be combined with 'jsimd_fdct_ifast_neon' to get
+ *       rid of VST1.16 instructions
+ */
+
+asm_function jsimd_convsamp_neon
+    SAMPLE_DATA     .req x0
+    START_COL       .req x1
+    WORKSPACE       .req x2
+    TMP1            .req x9
+    TMP2            .req x10
+    TMP3            .req x11
+    TMP4            .req x12
+    TMP5            .req x13
+    TMP6            .req x14
+    TMP7            .req x15
+    TMP8            .req x4
+    TMPDUP          .req w3
+
+    mov             TMPDUP, #128
+    ldp             TMP1, TMP2, [SAMPLE_DATA], 16
+    ldp             TMP3, TMP4, [SAMPLE_DATA], 16
+    dup             v0.8b, TMPDUP
+    add             TMP1, TMP1, START_COL
+    add             TMP2, TMP2, START_COL
+    ldp             TMP5, TMP6, [SAMPLE_DATA], 16
+    add             TMP3, TMP3, START_COL
+    add             TMP4, TMP4, START_COL
+    ldp             TMP7, TMP8, [SAMPLE_DATA], 16
+    add             TMP5, TMP5, START_COL
+    add             TMP6, TMP6, START_COL
+    ld1             {v16.8b}, [TMP1]
+    add             TMP7, TMP7, START_COL
+    add             TMP8, TMP8, START_COL
+    ld1             {v17.8b}, [TMP2]
+    usubl           v16.8h, v16.8b, v0.8b
+    ld1             {v18.8b}, [TMP3]
+    usubl           v17.8h, v17.8b, v0.8b
+    ld1             {v19.8b}, [TMP4]
+    usubl           v18.8h, v18.8b, v0.8b
+    ld1             {v20.8b}, [TMP5]
+    usubl           v19.8h, v19.8b, v0.8b
+    ld1             {v21.8b}, [TMP6]
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [WORKSPACE], 64
+    usubl           v20.8h, v20.8b, v0.8b
+    ld1             {v22.8b}, [TMP7]
+    usubl           v21.8h, v21.8b, v0.8b
+    ld1             {v23.8b}, [TMP8]
+    usubl           v22.8h, v22.8b, v0.8b
+    usubl           v23.8h, v23.8b, v0.8b
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [WORKSPACE], 64
+
+    br              x30
+
+    .unreq          SAMPLE_DATA
+    .unreq          START_COL
+    .unreq          WORKSPACE
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+    .unreq          TMPDUP
+
+/*****************************************************************************/
+
+/*
+ * jsimd_fdct_islow_neon
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform). The following code is based
+ * directly on the IJG''s original jfdctint.c; see the jfdctint.c for
+ * more details.
+ *
+ * TODO: can be combined with 'jsimd_convsamp_neon' to get
+ *       rid of a bunch of VLD1.16 instructions
+ */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+
+#define DESCALE_P1 (CONST_BITS-PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS+PASS1_BITS)
+
+#define F_0_298  2446  /* FIX(0.298631336) */
+#define F_0_390  3196  /* FIX(0.390180644) */
+#define F_0_541  4433  /* FIX(0.541196100) */
+#define F_0_765  6270  /* FIX(0.765366865) */
+#define F_0_899  7373  /* FIX(0.899976223) */
+#define F_1_175  9633  /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_fdct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+#define XFIX_P_0_298 v0.h[0]
+#define XFIX_N_0_390 v0.h[1]
+#define XFIX_P_0_541 v0.h[2]
+#define XFIX_P_0_765 v0.h[3]
+#define XFIX_N_0_899 v0.h[4]
+#define XFIX_P_1_175 v0.h[5]
+#define XFIX_P_1_501 v0.h[6]
+#define XFIX_N_1_847 v0.h[7]
+#define XFIX_N_1_961 v1.h[0]
+#define XFIX_P_2_053 v1.h[1]
+#define XFIX_N_2_562 v1.h[2]
+#define XFIX_P_3_072 v1.h[3]
+
+asm_function jsimd_fdct_islow_neon
+
+    DATA            .req x0
+    TMP             .req x9
+
+    /* Load constants */
+    adr             TMP, Ljsimd_fdct_islow_neon_consts
+    ld1             {v0.8h, v1.8h}, [TMP]
+
+    /* Save NEON registers */
+    sub             sp, sp, #64
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    /* Load all DATA into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17    | v16.8h
+     *   1 | d18     | d19    | v17.8h
+     *   2 | d20     | d21    | v18.8h
+     *   3 | d22     | d23    | v19.8h
+     *   4 | d24     | d25    | v20.8h
+     *   5 | d26     | d27    | v21.8h
+     *   6 | d28     | d29    | v22.8h
+     *   7 | d30     | d31    | v23.8h
+     */
+
+    ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+    sub             DATA, DATA, #64
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr[0] + dataptr[7]; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr[0] - dataptr[7]; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr[1] + dataptr[6]; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr[1] - dataptr[6]; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr[2] + dataptr[5]; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr[2] - dataptr[5]; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr[3] + dataptr[4]; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr[3] - dataptr[4]; */
+
+    /* even part */
+
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    shl             v16.8h, v16.8h, #PASS1_BITS  /* dataptr[0] = (DCTELEM) LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS); */
+    shl             v20.8h, v20.8h, #PASS1_BITS  /* dataptr[4] = (DCTELEM) LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P1
+    rshrn           v22.4h, v22.4s, #DESCALE_P1
+    rshrn2          v18.8h, v24.4s, #DESCALE_P1  /* dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P1  /* dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+
+    add             v8.8h, v28.8h, v31.8h        /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h        /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h       /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h       /* z4 = tmp5 + tmp7; */
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, - FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, - FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s  /* z3 += z5 */
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s  /* z4 += z5 */
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P1
+    rshrn           v21.4h, v29.4s, #DESCALE_P1
+    rshrn           v19.4h, v30.4s, #DESCALE_P1
+    rshrn           v17.4h, v31.4s, #DESCALE_P1
+    rshrn2          v23.8h, v24.4s, #DESCALE_P1  /* dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P1  /* dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P1  /* dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P1  /* dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr[0] + dataptr[7]; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr[0] - dataptr[7]; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr[1] + dataptr[6]; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr[1] - dataptr[6]; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr[2] + dataptr[5]; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr[2] - dataptr[5]; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr[3] + dataptr[4]; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr[3] - dataptr[4]; */
+
+    /* even part */
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    srshr           v16.8h, v16.8h, #PASS1_BITS  /* dataptr[0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS); */
+    srshr           v20.8h, v20.8h, #PASS1_BITS  /* dataptr[4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P2
+    rshrn           v22.4h, v22.4s, #DESCALE_P2
+    rshrn2          v18.8h, v24.4s, #DESCALE_P2  /* dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P2  /* dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+    add             v8.8h, v28.8h, v31.8h   /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h   /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h  /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h  /* z4 = tmp5 + tmp7; */
+
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, - FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, - FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P2
+    rshrn           v21.4h, v29.4s, #DESCALE_P2
+    rshrn           v19.4h, v30.4s, #DESCALE_P2
+    rshrn           v17.4h, v31.4s, #DESCALE_P2
+    rshrn2          v23.8h, v24.4s, #DESCALE_P2  /* dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P2  /* dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P2  /* dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P2  /* dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* store results */
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+
+    /* Restore NEON registers */
+    sub             sp, sp, #64
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    br              x30
+
+    .unreq          DATA
+    .unreq          TMP
+
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_fdct_ifast_neon
+ *
+ * This function contains a fast, not so accurate integer implementation of
+ * the forward DCT (Discrete Cosine Transform). It uses the same calculations
+ * and produces exactly the same output as IJG's original 'jpeg_fdct_ifast'
+ * function from jfdctfst.c
+ *
+ * TODO: can be combined with 'jsimd_convsamp_neon' to get
+ *       rid of a bunch of VLD1.16 instructions
+ */
+
+#undef XFIX_0_541196100
+#define XFIX_0_382683433 v0.h[0]
+#define XFIX_0_541196100 v0.h[1]
+#define XFIX_0_707106781 v0.h[2]
+#define XFIX_1_306562965 v0.h[3]
+
+.balign 16
+Ljsimd_fdct_ifast_neon_consts:
+  .short (98 * 128)               /* XFIX_0_382683433 */
+  .short (139 * 128)              /* XFIX_0_541196100 */
+  .short (181 * 128)              /* XFIX_0_707106781 */
+  .short (334 * 128 - 256 * 128)  /* XFIX_1_306562965 */
+
+asm_function jsimd_fdct_ifast_neon
+
+    DATA            .req x0
+    TMP             .req x9
+
+    /* Load constants */
+    adr             TMP, Ljsimd_fdct_ifast_neon_consts
+    ld1             {v0.4h}, [TMP]
+
+    /* Load all DATA into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17    | v0.8h
+     *   1 | d18     | d19    | q9
+     *   2 | d20     | d21    | q10
+     *   3 | d22     | d23    | q11
+     *   4 | d24     | d25    | q12
+     *   5 | d26     | d27    | q13
+     *   6 | d28     | d29    | q14
+     *   7 | d30     | d31    | q15
+     */
+
+    ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+    mov             TMP, #2
+    sub             DATA, DATA, #64
+1:
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v1, v2, v3, v4
+    subs            TMP, TMP, #1
+    /* 1-D FDCT */
+    add             v4.8h, v19.8h, v20.8h
+    sub             v20.8h, v19.8h, v20.8h
+    sub             v28.8h, v18.8h, v21.8h
+    add             v18.8h, v18.8h, v21.8h
+    sub             v29.8h, v17.8h, v22.8h
+    add             v17.8h, v17.8h, v22.8h
+    sub             v21.8h, v16.8h, v23.8h
+    add             v16.8h, v16.8h, v23.8h
+    sub             v6.8h, v17.8h, v18.8h
+    sub             v7.8h, v16.8h, v4.8h
+    add             v5.8h, v17.8h, v18.8h
+    add             v6.8h, v6.8h, v7.8h
+    add             v4.8h, v16.8h, v4.8h
+    sqdmulh         v6.8h, v6.8h, XFIX_0_707106781
+    add             v19.8h, v20.8h, v28.8h
+    add             v16.8h, v4.8h, v5.8h
+    sub             v20.8h, v4.8h, v5.8h
+    add             v5.8h, v28.8h, v29.8h
+    add             v29.8h, v29.8h, v21.8h
+    sqdmulh         v5.8h, v5.8h, XFIX_0_707106781
+    sub             v28.8h, v19.8h, v29.8h
+    add             v18.8h, v7.8h, v6.8h
+    sqdmulh         v28.8h, v28.8h, XFIX_0_382683433
+    sub             v22.8h, v7.8h, v6.8h
+    sqdmulh         v19.8h, v19.8h, XFIX_0_541196100
+    sqdmulh         v7.8h, v29.8h, XFIX_1_306562965
+    add             v6.8h, v21.8h, v5.8h
+    sub             v5.8h, v21.8h, v5.8h
+    add             v29.8h, v29.8h, v28.8h
+    add             v19.8h, v19.8h, v28.8h
+    add             v29.8h, v29.8h, v7.8h
+    add             v21.8h, v5.8h, v19.8h
+    sub             v19.8h, v5.8h, v19.8h
+    add             v17.8h, v6.8h, v29.8h
+    sub             v23.8h, v6.8h, v29.8h
+
+    b.ne            1b
+
+    /* store results */
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+
+    br              x30
+
+    .unreq          DATA
+    .unreq          TMP
+#undef XFIX_0_382683433
+#undef XFIX_0_541196100
+#undef XFIX_0_707106781
+#undef XFIX_1_306562965
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(void)
+ * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM *divisors,
+ *                      DCTELEM *workspace);
+ *
+ */
+asm_function jsimd_quantize_neon
+
+    COEF_BLOCK      .req x0
+    DIVISORS        .req x1
+    WORKSPACE       .req x2
+
+    RECIPROCAL      .req DIVISORS
+    CORRECTION      .req x9
+    SHIFT           .req x10
+    LOOP_COUNT      .req x11
+
+    mov             LOOP_COUNT, #2
+    add             CORRECTION, DIVISORS, #(64 * 2)
+    add             SHIFT, DIVISORS, #(64 * 6)
+1:
+    subs            LOOP_COUNT, LOOP_COUNT, #1
+    ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [WORKSPACE], 64
+    ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [CORRECTION], 64
+    abs             v20.8h, v0.8h
+    abs             v21.8h, v1.8h
+    abs             v22.8h, v2.8h
+    abs             v23.8h, v3.8h
+    ld1             {v28.8h, v29.8h, v30.8h, v31.8h}, [RECIPROCAL], 64
+    add             v20.8h, v20.8h, v4.8h  /* add correction */
+    add             v21.8h, v21.8h, v5.8h
+    add             v22.8h, v22.8h, v6.8h
+    add             v23.8h, v23.8h, v7.8h
+    umull           v4.4s, v20.4h, v28.4h  /* multiply by reciprocal */
+    umull2          v16.4s, v20.8h, v28.8h
+    umull           v5.4s, v21.4h, v29.4h
+    umull2          v17.4s, v21.8h, v29.8h
+    umull           v6.4s, v22.4h, v30.4h  /* multiply by reciprocal */
+    umull2          v18.4s, v22.8h, v30.8h
+    umull           v7.4s, v23.4h, v31.4h
+    umull2          v19.4s, v23.8h, v31.8h
+    ld1             {v24.8h, v25.8h, v26.8h, v27.8h}, [SHIFT], 64
+    shrn            v4.4h, v4.4s, #16
+    shrn            v5.4h, v5.4s, #16
+    shrn            v6.4h, v6.4s, #16
+    shrn            v7.4h, v7.4s, #16
+    shrn2           v4.8h, v16.4s, #16
+    shrn2           v5.8h, v17.4s, #16
+    shrn2           v6.8h, v18.4s, #16
+    shrn2           v7.8h, v19.4s, #16
+    neg             v24.8h, v24.8h
+    neg             v25.8h, v25.8h
+    neg             v26.8h, v26.8h
+    neg             v27.8h, v27.8h
+    sshr            v0.8h, v0.8h, #15  /* extract sign */
+    sshr            v1.8h, v1.8h, #15
+    sshr            v2.8h, v2.8h, #15
+    sshr            v3.8h, v3.8h, #15
+    ushl            v4.8h, v4.8h, v24.8h  /* shift */
+    ushl            v5.8h, v5.8h, v25.8h
+    ushl            v6.8h, v6.8h, v26.8h
+    ushl            v7.8h, v7.8h, v27.8h
+
+    eor             v4.16b, v4.16b, v0.16b  /* restore sign */
+    eor             v5.16b, v5.16b, v1.16b
+    eor             v6.16b, v6.16b, v2.16b
+    eor             v7.16b, v7.16b, v3.16b
+    sub             v4.8h, v4.8h, v0.8h
+    sub             v5.8h, v5.8h, v1.8h
+    sub             v6.8h, v6.8h, v2.8h
+    sub             v7.8h, v7.8h, v3.8h
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [COEF_BLOCK], 64
+
+    b.ne            1b
+
+    br              x30  /* return */
+
+    .unreq          COEF_BLOCK
+    .unreq          DIVISORS
+    .unreq          WORKSPACE
+    .unreq          RECIPROCAL
+    .unreq          CORRECTION
+    .unreq          SHIFT
+    .unreq          LOOP_COUNT
+
+
+/*****************************************************************************/
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ *
+ * GLOBAL(void)
+ * jsimd_h2v1_downsample_neon (JDIMENSION image_width, int max_v_samp_factor,
+ *                             JDIMENSION v_samp_factor,
+ *                             JDIMENSION width_blocks, JSAMPARRAY input_data,
+ *                             JSAMPARRAY output_data);
+ */
+
+.balign 16
+Ljsimd_h2_downsample_neon_consts:
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F  /* diff 0 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0E  /* diff 1 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0D, 0x0D  /* diff 2 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0C, 0x0C, 0x0C  /* diff 3 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0B, 0x0B, 0x0B, 0x0B  /* diff 4 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A  /* diff 5 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09  /* diff 6 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08  /* diff 7 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07  /* diff 8 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x06, \
+        0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06  /* diff 9 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x05, 0x05, \
+        0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05  /* diff 10 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x04, 0x04, 0x04, \
+        0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04  /* diff 11 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x03, 0x03, 0x03, 0x03, \
+        0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03  /* diff 12 */
+  .byte 0x00, 0x01, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, \
+        0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02  /* diff 13 */
+  .byte 0x00, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, \
+        0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01  /* diff 14 */
+  .byte 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
+        0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00  /* diff 15 */
+
+asm_function jsimd_h2v1_downsample_neon
+    IMAGE_WIDTH     .req x0
+    MAX_V_SAMP      .req x1
+    V_SAMP          .req x2
+    BLOCK_WIDTH     .req x3
+    INPUT_DATA      .req x4
+    OUTPUT_DATA     .req x5
+    OUTPTR          .req x9
+    INPTR           .req x10
+    TMP1            .req x11
+    TMP2            .req x12
+    TMP3            .req x13
+    TMPDUP          .req w15
+
+    mov             TMPDUP, #0x10000
+    lsl             TMP2, BLOCK_WIDTH, #4
+    sub             TMP2, TMP2, IMAGE_WIDTH
+    adr             TMP3, Ljsimd_h2_downsample_neon_consts
+    add             TMP3, TMP3, TMP2, lsl #4
+    dup             v16.4s, TMPDUP
+    ld1             {v18.16b}, [TMP3]
+
+1:  /* row loop */
+    ldr             INPTR, [INPUT_DATA], #8
+    ldr             OUTPTR, [OUTPUT_DATA], #8
+    subs            TMP1, BLOCK_WIDTH, #1
+    b.eq            3f
+2:  /* columns */
+    ld1             {v0.16b}, [INPTR], #16
+    mov             v4.16b, v16.16b
+    subs            TMP1, TMP1, #1
+    uadalp          v4.8h, v0.16b
+    shrn            v6.8b, v4.8h, #1
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            2b
+3:  /* last columns */
+    ld1             {v0.16b}, [INPTR]
+    mov             v4.16b, v16.16b
+    subs            V_SAMP, V_SAMP, #1
+    /* expand right */
+    tbl             v2.16b, {v0.16b}, v18.16b
+    uadalp          v4.8h, v2.16b
+    shrn            v6.8b, v4.8h, #1
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            1b
+
+    br              x30
+
+    .unreq          IMAGE_WIDTH
+    .unreq          MAX_V_SAMP
+    .unreq          V_SAMP
+    .unreq          BLOCK_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA
+    .unreq          OUTPTR
+    .unreq          INPTR
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMPDUP
+
+
+/*****************************************************************************/
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ *
+ * GLOBAL(void)
+ * jsimd_h2v2_downsample_neon (JDIMENSION image_width, int max_v_samp_factor,
+ *                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+ *                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+ */
+
+.balign 16
+asm_function jsimd_h2v2_downsample_neon
+    IMAGE_WIDTH     .req x0
+    MAX_V_SAMP      .req x1
+    V_SAMP          .req x2
+    BLOCK_WIDTH     .req x3
+    INPUT_DATA      .req x4
+    OUTPUT_DATA     .req x5
+    OUTPTR          .req x9
+    INPTR0          .req x10
+    INPTR1          .req x14
+    TMP1            .req x11
+    TMP2            .req x12
+    TMP3            .req x13
+    TMPDUP          .req w15
+
+    mov             TMPDUP, #1
+    lsl             TMP2, BLOCK_WIDTH, #4
+    lsl             TMPDUP, TMPDUP, #17
+    sub             TMP2, TMP2, IMAGE_WIDTH
+    adr             TMP3, Ljsimd_h2_downsample_neon_consts
+    orr             TMPDUP, TMPDUP, #1
+    add             TMP3, TMP3, TMP2, lsl #4
+    dup             v16.4s, TMPDUP
+    ld1             {v18.16b}, [TMP3]
+
+1:  /* row loop */
+    ldr             INPTR0, [INPUT_DATA], #8
+    ldr             OUTPTR, [OUTPUT_DATA], #8
+    ldr             INPTR1, [INPUT_DATA], #8
+    subs            TMP1, BLOCK_WIDTH, #1
+    b.eq            3f
+2:  /* columns */
+    ld1             {v0.16b}, [INPTR0], #16
+    ld1             {v1.16b}, [INPTR1], #16
+    mov             v4.16b, v16.16b
+    subs            TMP1, TMP1, #1
+    uadalp          v4.8h, v0.16b
+    uadalp          v4.8h, v1.16b
+    shrn            v6.8b, v4.8h, #2
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            2b
+3:  /* last columns */
+    ld1             {v0.16b}, [INPTR0], #16
+    ld1             {v1.16b}, [INPTR1], #16
+    mov             v4.16b, v16.16b
+    subs            V_SAMP, V_SAMP, #1
+    /* expand right */
+    tbl             v2.16b, {v0.16b}, v18.16b
+    tbl             v3.16b, {v1.16b}, v18.16b
+    uadalp          v4.8h, v2.16b
+    uadalp          v4.8h, v3.16b
+    shrn            v6.8b, v4.8h, #2
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            1b
+
+    br              x30
+
+    .unreq          IMAGE_WIDTH
+    .unreq          MAX_V_SAMP
+    .unreq          V_SAMP
+    .unreq          BLOCK_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA
+    .unreq          OUTPTR
+    .unreq          INPTR0
+    .unreq          INPTR1
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMPDUP
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(JOCTET*)
+ * jsimd_huff_encode_one_block (working_state *state, JOCTET *buffer,
+ *                              JCOEFPTR block, int last_dc_val,
+ *                              c_derived_tbl *dctbl, c_derived_tbl *actbl)
+ *
+ */
+
+    BUFFER          .req x1
+    PUT_BUFFER      .req x6
+    PUT_BITS        .req x7
+    PUT_BITSw       .req w7
+
+.macro emit_byte
+    sub             PUT_BITS, PUT_BITS, #0x8
+    lsr             x19, PUT_BUFFER, PUT_BITS
+    uxtb            w19, w19
+    strb            w19, [BUFFER, #1]!
+    cmp             w19, #0xff
+    b.ne            14f
+    strb            wzr, [BUFFER, #1]!
+14:
+.endm
+.macro put_bits CODE, SIZE
+    lsl             PUT_BUFFER, PUT_BUFFER, \SIZE
+    add             PUT_BITS, PUT_BITS, \SIZE
+    orr             PUT_BUFFER, PUT_BUFFER, \CODE
+.endm
+.macro checkbuf31
+    cmp             PUT_BITS, #0x20
+    b.lt            31f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+31:
+.endm
+.macro checkbuf47
+    cmp             PUT_BITS, #0x30
+    b.lt            47f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+47:
+.endm
+
+.macro generate_jsimd_huff_encode_one_block fast_tbl
+
+.balign 16
+.if \fast_tbl == 1
+Ljsimd_huff_encode_one_block_neon_consts:
+.else
+Ljsimd_huff_encode_one_block_neon_slowtbl_consts:
+.endif
+    .byte 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, \
+          0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80
+.if \fast_tbl == 1
+    .byte    0,   1,   2,   3,  16,  17,  32,  33, \
+            18,  19,   4,   5,   6,   7,  20,  21  /* L0 => L3 : 4 lines OK */
+    .byte   34,  35,  48,  49, 255, 255,  50,  51, \
+            36,  37,  22,  23,   8,   9,  10,  11  /* L0 => L3 : 4 lines OK */
+    .byte    8,   9,  22,  23,  36,  37,  50,  51, \
+           255, 255, 255, 255, 255, 255,  52,  53  /* L1 => L4 : 4 lines OK */
+    .byte   54,  55,  40,  41,  26,  27,  12,  13, \
+            14,  15,  28,  29,  42,  43,  56,  57  /* L0 => L3 : 4 lines OK */
+    .byte    6,   7,  20,  21,  34,  35,  48,  49, \
+            50,  51,  36,  37,  22,  23,   8,   9  /* L4 => L7 : 4 lines OK */
+    .byte   42,  43,  28,  29,  14,  15,  30,  31, \
+            44,  45,  58,  59, 255, 255, 255, 255  /* L1 => L4 : 4 lines OK */
+    .byte  255, 255, 255, 255,  56,  57,  42,  43, \
+            28,  29,  14,  15,  30,  31,  44,  45  /* L3 => L6 : 4 lines OK */
+    .byte   26,  27,  40,  41,  42,  43,  28,  29, \
+            14,  15,  30,  31,  44,  45,  46,  47  /* L5 => L7 : 3 lines OK */
+    .byte  255, 255, 255, 255,   0,   1, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L4 : 1 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+             0,   1,  16,  17,   2,   3, 255, 255  /* L5 => L6 : 2 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+           255, 255, 255, 255,   8,   9,  22,  23  /* L5 => L6 : 2 lines OK */
+    .byte    4,   5,   6,   7, 255, 255, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L7 : 1 line OK */
+.endif
+
+.if \fast_tbl == 1
+asm_function jsimd_huff_encode_one_block_neon
+.else
+asm_function jsimd_huff_encode_one_block_neon_slowtbl
+.endif
+    sub             sp, sp, 272
+    sub             BUFFER, BUFFER, #0x1    /* BUFFER=buffer-- */
+    /* Save ARM registers */
+    stp             x19, x20, [sp], 16
+.if \fast_tbl == 1
+    adr             x15, Ljsimd_huff_encode_one_block_neon_consts
+.else
+    adr             x15, Ljsimd_huff_encode_one_block_neon_slowtbl_consts
+.endif
+    ldr             PUT_BUFFER, [x0, #0x10]
+    ldr             PUT_BITSw, [x0, #0x18]
+    ldrsh           w12, [x2]               /* load DC coeff in w12 */
+    /* prepare data */
+.if \fast_tbl == 1
+    ld1             {v23.16b}, [x15], #16
+    ld1             {v0.16b, v1.16b, v2.16b, v3.16b}, [x15], #64
+    ld1             {v4.16b, v5.16b, v6.16b, v7.16b}, [x15], #64
+    ld1             {v16.16b, v17.16b, v18.16b, v19.16b}, [x15], #64
+    ld1             {v24.16b, v25.16b, v26.16b, v27.16b}, [x2], #64
+    ld1             {v28.16b, v29.16b, v30.16b, v31.16b}, [x2], #64
+    sub             w12, w12, w3      /* last_dc_val, not used afterwards */
+    /* ZigZag 8x8 */
+    tbl             v0.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v0.16b
+    tbl             v1.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v1.16b
+    tbl             v2.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v2.16b
+    tbl             v3.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v3.16b
+    tbl             v4.16b, {v28.16b, v29.16b, v30.16b, v31.16b}, v4.16b
+    tbl             v5.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v5.16b
+    tbl             v6.16b, {v27.16b, v28.16b, v29.16b, v30.16b}, v6.16b
+    tbl             v7.16b, {v29.16b, v30.16b, v31.16b}, v7.16b
+    ins             v0.h[0], w12
+    tbx             v1.16b, {v28.16b}, v16.16b
+    tbx             v2.16b, {v29.16b, v30.16b}, v17.16b
+    tbx             v5.16b, {v29.16b, v30.16b}, v18.16b
+    tbx             v6.16b, {v31.16b}, v19.16b
+.else
+      add             x13, x2, #0x22
+      sub             w12, w12, w3    /* last_dc_val, not used afterwards */
+    ld1             {v23.16b}, [x15]
+      add             x14, x2, #0x18
+      add             x3, x2, #0x36
+    ins             v0.h[0], w12
+      add             x9, x2, #0x2
+    ld1             {v1.h}[0], [x13]
+      add             x15, x2, #0x30
+    ld1             {v2.h}[0], [x14]
+      add             x19, x2, #0x26
+    ld1             {v3.h}[0], [x3]
+      add             x20, x2, #0x28
+    ld1             {v0.h}[1], [x9]
+      add             x12, x2, #0x10
+    ld1             {v1.h}[1], [x15]
+      add             x13, x2, #0x40
+    ld1             {v2.h}[1], [x19]
+      add             x14, x2, #0x34
+    ld1             {v3.h}[1], [x20]
+      add             x3, x2, #0x1a
+    ld1             {v0.h}[2], [x12]
+      add             x9, x2, #0x20
+    ld1             {v1.h}[2], [x13]
+      add             x15, x2, #0x32
+    ld1             {v2.h}[2], [x14]
+      add             x19, x2, #0x42
+    ld1             {v3.h}[2], [x3]
+      add             x20, x2, #0xc
+    ld1             {v0.h}[3], [x9]
+      add             x12, x2, #0x12
+    ld1             {v1.h}[3], [x15]
+      add             x13, x2, #0x24
+    ld1             {v2.h}[3], [x19]
+      add             x14, x2, #0x50
+    ld1             {v3.h}[3], [x20]
+      add             x3, x2, #0xe
+    ld1             {v0.h}[4], [x12]
+      add             x9, x2, #0x4
+    ld1             {v1.h}[4], [x13]
+      add             x15, x2, #0x16
+    ld1             {v2.h}[4], [x14]
+      add             x19, x2, #0x60
+    ld1             {v3.h}[4], [x3]
+      add             x20, x2, #0x1c
+    ld1             {v0.h}[5], [x9]
+      add             x12, x2, #0x6
+    ld1             {v1.h}[5], [x15]
+      add             x13, x2, #0x8
+    ld1             {v2.h}[5], [x19]
+      add             x14, x2, #0x52
+    ld1             {v3.h}[5], [x20]
+      add             x3, x2, #0x2a
+    ld1             {v0.h}[6], [x12]
+      add             x9, x2, #0x14
+    ld1             {v1.h}[6], [x13]
+      add             x15, x2, #0xa
+    ld1             {v2.h}[6], [x14]
+      add             x19, x2, #0x44
+    ld1             {v3.h}[6], [x3]
+      add             x20, x2, #0x38
+    ld1             {v0.h}[7], [x9]
+      add             x12, x2, #0x46
+    ld1             {v1.h}[7], [x15]
+      add             x13, x2, #0x3a
+    ld1             {v2.h}[7], [x19]
+      add             x14, x2, #0x74
+    ld1             {v3.h}[7], [x20]
+      add             x3, x2, #0x6a
+    ld1             {v4.h}[0], [x12]
+      add             x9, x2, #0x54
+    ld1             {v5.h}[0], [x13]
+      add             x15, x2, #0x2c
+    ld1             {v6.h}[0], [x14]
+      add             x19, x2, #0x76
+    ld1             {v7.h}[0], [x3]
+      add             x20, x2, #0x78
+    ld1             {v4.h}[1], [x9]
+      add             x12, x2, #0x62
+    ld1             {v5.h}[1], [x15]
+      add             x13, x2, #0x1e
+    ld1             {v6.h}[1], [x19]
+      add             x14, x2, #0x68
+    ld1             {v7.h}[1], [x20]
+      add             x3, x2, #0x7a
+    ld1             {v4.h}[2], [x12]
+      add             x9, x2, #0x70
+    ld1             {v5.h}[2], [x13]
+      add             x15, x2, #0x2e
+    ld1             {v6.h}[2], [x14]
+      add             x19, x2, #0x5a
+    ld1             {v7.h}[2], [x3]
+      add             x20, x2, #0x6c
+    ld1             {v4.h}[3], [x9]
+      add             x12, x2, #0x72
+    ld1             {v5.h}[3], [x15]
+      add             x13, x2, #0x3c
+    ld1             {v6.h}[3], [x19]
+      add             x14, x2, #0x4c
+    ld1             {v7.h}[3], [x20]
+      add             x3, x2, #0x5e
+    ld1             {v4.h}[4], [x12]
+      add             x9, x2, #0x64
+    ld1             {v5.h}[4], [x13]
+      add             x15, x2, #0x4a
+    ld1             {v6.h}[4], [x14]
+      add             x19, x2, #0x3e
+    ld1             {v7.h}[4], [x3]
+      add             x20, x2, #0x6e
+    ld1             {v4.h}[5], [x9]
+      add             x12, x2, #0x56
+    ld1             {v5.h}[5], [x15]
+      add             x13, x2, #0x58
+    ld1             {v6.h}[5], [x19]
+      add             x14, x2, #0x4e
+    ld1             {v7.h}[5], [x20]
+      add             x3, x2, #0x7c
+    ld1             {v4.h}[6], [x12]
+      add             x9, x2, #0x48
+    ld1             {v5.h}[6], [x13]
+      add             x15, x2, #0x66
+    ld1             {v6.h}[6], [x14]
+      add             x19, x2, #0x5c
+    ld1             {v7.h}[6], [x3]
+      add             x20, x2, #0x7e
+    ld1             {v4.h}[7], [x9]
+    ld1             {v5.h}[7], [x15]
+    ld1             {v6.h}[7], [x19]
+    ld1             {v7.h}[7], [x20]
+.endif
+    cmlt            v24.8h, v0.8h, #0
+    cmlt            v25.8h, v1.8h, #0
+    cmlt            v26.8h, v2.8h, #0
+    cmlt            v27.8h, v3.8h, #0
+    cmlt            v28.8h, v4.8h, #0
+    cmlt            v29.8h, v5.8h, #0
+    cmlt            v30.8h, v6.8h, #0
+    cmlt            v31.8h, v7.8h, #0
+    abs             v0.8h, v0.8h
+    abs             v1.8h, v1.8h
+    abs             v2.8h, v2.8h
+    abs             v3.8h, v3.8h
+    abs             v4.8h, v4.8h
+    abs             v5.8h, v5.8h
+    abs             v6.8h, v6.8h
+    abs             v7.8h, v7.8h
+    eor             v24.16b, v24.16b, v0.16b
+    eor             v25.16b, v25.16b, v1.16b
+    eor             v26.16b, v26.16b, v2.16b
+    eor             v27.16b, v27.16b, v3.16b
+    eor             v28.16b, v28.16b, v4.16b
+    eor             v29.16b, v29.16b, v5.16b
+    eor             v30.16b, v30.16b, v6.16b
+    eor             v31.16b, v31.16b, v7.16b
+    cmeq            v16.8h, v0.8h, #0
+    cmeq            v17.8h, v1.8h, #0
+    cmeq            v18.8h, v2.8h, #0
+    cmeq            v19.8h, v3.8h, #0
+    cmeq            v20.8h, v4.8h, #0
+    cmeq            v21.8h, v5.8h, #0
+    cmeq            v22.8h, v6.8h, #0
+    xtn             v16.8b, v16.8h
+    xtn             v18.8b, v18.8h
+    xtn             v20.8b, v20.8h
+    xtn             v22.8b, v22.8h
+      umov            w14, v0.h[0]
+    xtn2            v16.16b, v17.8h
+      umov            w13, v24.h[0]
+    xtn2            v18.16b, v19.8h
+      clz             w14, w14
+    xtn2            v20.16b, v21.8h
+      lsl             w13, w13, w14
+    cmeq            v17.8h, v7.8h, #0
+      sub             w12, w14, #32
+    xtn2            v22.16b, v17.8h
+      lsr             w13, w13, w14
+    and             v16.16b, v16.16b, v23.16b
+      neg             w12, w12
+    and             v18.16b, v18.16b, v23.16b
+      add             x3, x4, #0x400           /* r1 = dctbl->ehufsi */
+    and             v20.16b, v20.16b, v23.16b
+      add             x15, sp, #0x80           /* x15 = t2 */
+    and             v22.16b, v22.16b, v23.16b
+      ldr             w10, [x4, x12, lsl #2]
+    addp            v16.16b, v16.16b, v18.16b
+      ldrb            w11, [x3, x12]
+    addp            v20.16b, v20.16b, v22.16b
+      checkbuf47
+    addp            v16.16b, v16.16b, v20.16b
+      put_bits        x10, x11
+    addp            v16.16b, v16.16b, v18.16b
+      checkbuf47
+    umov            x9,v16.D[0]
+      put_bits        x13, x12
+    cnt             v17.8b, v16.8b
+      mvn             x9, x9
+    addv            B18, v17.8b
+      add             x4, x5, #0x400   /* x4 = actbl->ehufsi */
+    umov            w12, v18.b[0]
+      lsr             x9, x9, #0x1     /* clear AC coeff */
+    ldr             w13, [x5, #0x3c0]  /* x13 = actbl->ehufco[0xf0] */
+    rbit            x9, x9             /* x9 = index0 */
+    ldrb            w14, [x4, #0xf0]   /* x14 = actbl->ehufsi[0xf0] */
+    cmp             w12, #(64-8)
+    mov             x11, sp
+    b.lt            4f
+    cbz             x9, 6f
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x11], #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x11], #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, [x11], #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, [x11], #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w20, [x15, #-126]
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    clz             w20, w20
+    ldrh            w3, [x15, #2]!
+    sub             w11, w20, #32
+    lsl             w3, w3, w20
+    neg             w11, w11
+    lsr             w3, w3, w20
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, [x5, x2, lsl #2]
+    ldrb            w10, [x4, x2]
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+    b               6f
+4:
+    movi            v21.8h, #0x0010
+    clz             v0.8h, v0.8h
+    clz             v1.8h, v1.8h
+    clz             v2.8h, v2.8h
+    clz             v3.8h, v3.8h
+    clz             v4.8h, v4.8h
+    clz             v5.8h, v5.8h
+    clz             v6.8h, v6.8h
+    clz             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    neg             v0.8h, v0.8h
+    neg             v1.8h, v1.8h
+    neg             v2.8h, v2.8h
+    neg             v3.8h, v3.8h
+    neg             v4.8h, v4.8h
+    neg             v5.8h, v5.8h
+    neg             v6.8h, v6.8h
+    neg             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    add             v0.8h, v21.8h, v0.8h
+    add             v1.8h, v21.8h, v1.8h
+    add             v2.8h, v21.8h, v2.8h
+    add             v3.8h, v21.8h, v3.8h
+    add             v4.8h, v21.8h, v4.8h
+    add             v5.8h, v21.8h, v5.8h
+    add             v6.8h, v21.8h, v6.8h
+    add             v7.8h, v21.8h, v7.8h
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x11], #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x11], #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, [x11], #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, [x11], #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w11, [x15, #-126]
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    ldrh            w3, [x15, #2]!
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, [x5, x2, lsl #2]
+    ldrb            w10, [x4, x2]
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+6:
+    add             x13, sp, #0xfe
+    cmp             x15, x13
+    b.hs            1f
+    ldr             w12, [x5]
+    ldrb            w14, [x4]
+    checkbuf47
+    put_bits        x12, x14
+1:
+    sub             sp, sp, 16
+    str             PUT_BUFFER, [x0, #0x10]
+    str             PUT_BITSw, [x0, #0x18]
+    ldp             x19, x20, [sp], 16
+    add             x0, BUFFER, #0x1
+    add             sp, sp, 256
+    br              x30
+
+.endm
+
+generate_jsimd_huff_encode_one_block 1
+generate_jsimd_huff_encode_one_block 0
+
+    .unreq          BUFFER
+    .unreq          PUT_BUFFER
+    .unreq          PUT_BITS
+    .unreq          PUT_BITSw
+
+.purgem emit_byte
+.purgem put_bits
+.purgem checkbuf31
+.purgem checkbuf47
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm.c	2021-04-25 01:47:22.000000000 +0800
@@ -2,17 +2,16 @@
  * jsimd_arm.c
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009-2011 D. R. Commander
- * 
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015-2016 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  * This file contains the interface between the "normal" portions
- * of the library and the SIMD implementations when running on
- * ARM architecture.
- *
- * Based on the stubs from 'jsimd_none.c'
+ * of the library and the SIMD implementations when running on a
+ * 32-bit ARM architecture.
  */
 
 #define JPEG_INTERNALS
@@ -28,8 +27,9 @@
 #include <ctype.h>
 
 static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
 
-#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
 
 #define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
 
@@ -100,7 +100,7 @@
 init_simd (void)
 {
   char *env = NULL;
-#if !defined(__ARM_NEON__) && defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
   int bufsize = 1024; /* an initial guess for the line buffer size limit */
 #endif
 
@@ -123,12 +123,15 @@
 #endif
 
   /* Force different settings through environment variables */
-  env = getenv("JSIMD_FORCE_ARM_NEON");
+  env = getenv("JSIMD_FORCENEON");
   if ((env != NULL) && (strcmp(env, "1") == 0))
     simd_support &= JSIMD_ARM_NEON;
-  env = getenv("JSIMD_FORCE_NO_SIMD");
+  env = getenv("JSIMD_FORCENONE");
   if ((env != NULL) && (strcmp(env, "1") == 0))
     simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
 }
 
 GLOBAL(int)
@@ -170,6 +173,24 @@
     return 0;
   if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
     return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
@@ -183,8 +204,7 @@
 {
   void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
 
-  switch(cinfo->in_color_space)
-  {
+  switch(cinfo->in_color_space) {
     case JCS_EXT_RGB:
       neonfct=jsimd_extrgb_ycc_convert_neon;
       break;
@@ -212,9 +232,7 @@
       break;
   }
 
-  if (simd_support & JSIMD_ARM_NEON)
-    neonfct(cinfo->image_width, input_buf,
-        output_buf, output_row, num_rows);
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
 }
 
 GLOBAL(void)
@@ -231,8 +249,7 @@
 {
   void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
 
-  switch(cinfo->out_color_space)
-  {
+  switch(cinfo->out_color_space) {
     case JCS_EXT_RGB:
       neonfct=jsimd_ycc_extrgb_convert_neon;
       break;
@@ -255,14 +272,21 @@
     case JCS_EXT_ARGB:
       neonfct=jsimd_ycc_extxrgb_convert_neon;
       break;
-  default:
+    default:
       neonfct=jsimd_ycc_extrgb_convert_neon;
       break;
   }
 
-  if (simd_support & JSIMD_ARM_NEON)
-    neonfct(cinfo->output_width, input_buf,
-        input_row, output_buf, num_rows);
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
 }
 
 GLOBAL(int)
@@ -282,13 +306,13 @@
 }
 
 GLOBAL(void)
-jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
 }
 
 GLOBAL(void)
-jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
 }
@@ -311,17 +335,17 @@
 
 GLOBAL(void)
 jsimd_h2v2_upsample (j_decompress_ptr cinfo,
-                     jpeg_component_info * compptr, 
+                     jpeg_component_info *compptr,
                      JSAMPARRAY input_data,
-                     JSAMPARRAY * output_data_ptr)
+                     JSAMPARRAY *output_data_ptr)
 {
 }
 
 GLOBAL(void)
 jsimd_h2v1_upsample (j_decompress_ptr cinfo,
-                     jpeg_component_info * compptr, 
+                     jpeg_component_info *compptr,
                      JSAMPARRAY input_data,
-                     JSAMPARRAY * output_data_ptr)
+                     JSAMPARRAY *output_data_ptr)
 {
 }
 
@@ -338,23 +362,35 @@
 {
   init_simd();
 
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
   return 0;
 }
 
 GLOBAL(void)
 jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
-                           jpeg_component_info * compptr, 
+                           jpeg_component_info *compptr,
                            JSAMPARRAY input_data,
-                           JSAMPARRAY * output_data_ptr)
+                           JSAMPARRAY *output_data_ptr)
 {
 }
 
 GLOBAL(void)
 jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
-                           jpeg_component_info * compptr, 
+                           jpeg_component_info *compptr,
                            JSAMPARRAY input_data,
-                           JSAMPARRAY * output_data_ptr)
+                           JSAMPARRAY *output_data_ptr)
 {
+  jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
 }
 
 GLOBAL(int)
@@ -420,15 +456,14 @@
 
 GLOBAL(void)
 jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
-                DCTELEM * workspace)
+                DCTELEM *workspace)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_convsamp_neon(sample_data, start_col, workspace);
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
 }
 
 GLOBAL(void)
 jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
-                      FAST_FLOAT * workspace)
+                      FAST_FLOAT *workspace)
 {
 }
 
@@ -466,19 +501,18 @@
 }
 
 GLOBAL(void)
-jsimd_fdct_islow (DCTELEM * data)
+jsimd_fdct_islow (DCTELEM *data)
 {
 }
 
 GLOBAL(void)
-jsimd_fdct_ifast (DCTELEM * data)
+jsimd_fdct_ifast (DCTELEM *data)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_fdct_ifast_neon(data);
+  jsimd_fdct_ifast_neon(data);
 }
 
 GLOBAL(void)
-jsimd_fdct_float (FAST_FLOAT * data)
+jsimd_fdct_float (FAST_FLOAT *data)
 {
 }
 
@@ -510,16 +544,15 @@
 }
 
 GLOBAL(void)
-jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors,
-                DCTELEM * workspace)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_quantize_neon(coef_block, divisors, workspace);
+  jsimd_quantize_neon(coef_block, divisors, workspace);
 }
 
 GLOBAL(void)
-jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors,
-                      FAST_FLOAT * workspace)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
 {
 }
 
@@ -540,7 +573,7 @@
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
@@ -563,28 +596,28 @@
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
 }
 
 GLOBAL(void)
-jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
                 JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
                 JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
 }
 
 GLOBAL(int)
@@ -629,7 +662,7 @@
   if (IFAST_SCALE_BITS != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
@@ -644,27 +677,51 @@
 }
 
 GLOBAL(void)
-jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
 }
 
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm_neon.S glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm_neon.S
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_arm_neon.S	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_arm_neon.S	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
- * ARM NEON optimizations for libjpeg-turbo
+ * ARMv7 NEON optimizations for libjpeg-turbo
  *
  * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies).
  * All rights reserved.
  * Author: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2014 Siarhei Siamashka.  All Rights Reserved.
+ * Copyright (C) 2014 Linaro Limited.  All Rights Reserved.
+ * Copyright (C) 2015 D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016 Matthieu Darbois.  All Rights Reserved.
  *
  * This software is provided 'as-is', without any express or implied
  * warranty.  In no event will the authors be held liable for any damages
@@ -23,7 +27,7 @@
  */
 
 #if defined(__linux__) && defined(__ELF__)
-.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
 #endif
 
 .text
@@ -31,20 +35,20 @@
 .arch armv7a
 .object_arch armv4
 .arm
+.syntax unified
 
 
 #define RESPECT_STRICT_ALIGNMENT 1
 
+
 /*****************************************************************************/
 
 /* Supplementary macro for setting function attributes */
 .macro asm_function fname
 #ifdef __APPLE__
-    .func _\fname
     .globl _\fname
 _\fname:
 #else
-    .func \fname
     .global \fname
 #ifdef __ELF__
     .hidden \fname
@@ -56,12 +60,13 @@
 
 /* Transpose a block of 4x4 coefficients in four 64-bit registers */
 .macro transpose_4x4 x0, x1, x2, x3
-    vtrn.16 \x0, \x1
-    vtrn.16 \x2, \x3
-    vtrn.32 \x0, \x2
-    vtrn.32 \x1, \x3
+    vtrn.16         \x0, \x1
+    vtrn.16         \x2, \x3
+    vtrn.32         \x0, \x2
+    vtrn.32         \x1, \x3
 .endm
 
+
 #define CENTERJSAMPLE 128
 
 /*****************************************************************************/
@@ -70,22 +75,22 @@
  * Perform dequantization and inverse DCT on one block of coefficients.
  *
  * GLOBAL(void)
- * jsimd_idct_islow_neon (void * dct_table, JCOEFPTR coef_block,
+ * jsimd_idct_islow_neon (void *dct_table, JCOEFPTR coef_block,
  *                        JSAMPARRAY output_buf, JDIMENSION output_col)
  */
 
-#define FIX_0_298631336  (2446)
-#define FIX_0_390180644  (3196)
-#define FIX_0_541196100  (4433)
-#define FIX_0_765366865  (6270)
-#define FIX_0_899976223  (7373)
-#define FIX_1_175875602  (9633)
-#define FIX_1_501321110  (12299)
-#define FIX_1_847759065  (15137)
-#define FIX_1_961570560  (16069)
-#define FIX_2_053119869  (16819)
-#define FIX_2_562915447  (20995)
-#define FIX_3_072711026  (25172)
+#define FIX_0_298631336 (2446)
+#define FIX_0_390180644 (3196)
+#define FIX_0_541196100 (4433)
+#define FIX_0_765366865 (6270)
+#define FIX_0_899976223 (7373)
+#define FIX_1_175875602 (9633)
+#define FIX_1_501321110 (12299)
+#define FIX_1_847759065 (15137)
+#define FIX_1_961570560 (16069)
+#define FIX_2_053119869 (16819)
+#define FIX_2_562915447 (20995)
+#define FIX_3_072711026 (25172)
 
 #define FIX_1_175875602_MINUS_1_961570560 (FIX_1_175875602 - FIX_1_961570560)
 #define FIX_1_175875602_MINUS_0_390180644 (FIX_1_175875602 - FIX_0_390180644)
@@ -103,8 +108,8 @@
 #define REF_1D_IDCT(xrow0, xrow1, xrow2, xrow3, xrow4, xrow5, xrow6, xrow7)   \
 {                                                                             \
     DCTELEM row0, row1, row2, row3, row4, row5, row6, row7;                   \
-    INT32   q1, q2, q3, q4, q5, q6, q7;                                       \
-    INT32   tmp11_plus_tmp2, tmp11_minus_tmp2;                                \
+    JLONG   q1, q2, q3, q4, q5, q6, q7;                                       \
+    JLONG   tmp11_plus_tmp2, tmp11_minus_tmp2;                                \
                                                                               \
     /* 1-D iDCT input data */                                                 \
     row0 = xrow0;                                                             \
@@ -125,7 +130,7 @@
     q2 = MULTIPLY(row2, FIX_0_541196100) +                                    \
          MULTIPLY(row6, FIX_0_541196100_MINUS_1_847759065);                   \
     q4 = q6;                                                                  \
-    q3 = ((INT32) row0 - (INT32) row4) << 13;                                 \
+    q3 = ((JLONG) row0 - (JLONG) row4) << 13;                                 \
     q6 += MULTIPLY(row5, -FIX_2_562915447) +                                  \
           MULTIPLY(row3, FIX_3_072711026_MINUS_2_562915447);                  \
     /* now we can use q1 (reloadable constants have been used up) */          \
@@ -152,7 +157,7 @@
     /* (tmp11 - tmp2) has been calculated (out_row6 before descale) */        \
     tmp11_minus_tmp2 = q1;                                                    \
                                                                               \
-    q1 = ((INT32) row0 + (INT32) row4) << 13;                                 \
+    q1 = ((JLONG) row0 + (JLONG) row4) << 13;                                 \
     q2 = q1 + q6;                                                             \
     q1 = q1 - q6;                                                             \
                                                                               \
@@ -167,34 +172,34 @@
     tmp13 = q1;                                                               \
 }
 
-#define XFIX_0_899976223                    d0[0]
-#define XFIX_0_541196100                    d0[1]
-#define XFIX_2_562915447                    d0[2]
-#define XFIX_0_298631336_MINUS_0_899976223  d0[3]
-#define XFIX_1_501321110_MINUS_0_899976223  d1[0]
-#define XFIX_2_053119869_MINUS_2_562915447  d1[1]
-#define XFIX_0_541196100_PLUS_0_765366865   d1[2]
-#define XFIX_1_175875602                    d1[3]
-#define XFIX_1_175875602_MINUS_0_390180644  d2[0]
-#define XFIX_0_541196100_MINUS_1_847759065  d2[1]
-#define XFIX_3_072711026_MINUS_2_562915447  d2[2]
-#define XFIX_1_175875602_MINUS_1_961570560  d2[3]
+#define XFIX_0_899976223                   d0[0]
+#define XFIX_0_541196100                   d0[1]
+#define XFIX_2_562915447                   d0[2]
+#define XFIX_0_298631336_MINUS_0_899976223 d0[3]
+#define XFIX_1_501321110_MINUS_0_899976223 d1[0]
+#define XFIX_2_053119869_MINUS_2_562915447 d1[1]
+#define XFIX_0_541196100_PLUS_0_765366865  d1[2]
+#define XFIX_1_175875602                   d1[3]
+#define XFIX_1_175875602_MINUS_0_390180644 d2[0]
+#define XFIX_0_541196100_MINUS_1_847759065 d2[1]
+#define XFIX_3_072711026_MINUS_2_562915447 d2[2]
+#define XFIX_1_175875602_MINUS_1_961570560 d2[3]
 
 .balign 16
 jsimd_idct_islow_neon_consts:
-    .short FIX_0_899976223                    /* d0[0] */
-    .short FIX_0_541196100                    /* d0[1] */
-    .short FIX_2_562915447                    /* d0[2] */
-    .short FIX_0_298631336_MINUS_0_899976223  /* d0[3] */
-    .short FIX_1_501321110_MINUS_0_899976223  /* d1[0] */
-    .short FIX_2_053119869_MINUS_2_562915447  /* d1[1] */
-    .short FIX_0_541196100_PLUS_0_765366865   /* d1[2] */
-    .short FIX_1_175875602                    /* d1[3] */
-    /* reloadable constants */
-    .short FIX_1_175875602_MINUS_0_390180644  /* d2[0] */
-    .short FIX_0_541196100_MINUS_1_847759065  /* d2[1] */
-    .short FIX_3_072711026_MINUS_2_562915447  /* d2[2] */
-    .short FIX_1_175875602_MINUS_1_961570560  /* d2[3] */
+  .short FIX_0_899976223                    /* d0[0] */
+  .short FIX_0_541196100                    /* d0[1] */
+  .short FIX_2_562915447                    /* d0[2] */
+  .short FIX_0_298631336_MINUS_0_899976223  /* d0[3] */
+  .short FIX_1_501321110_MINUS_0_899976223  /* d1[0] */
+  .short FIX_2_053119869_MINUS_2_562915447  /* d1[1] */
+  .short FIX_0_541196100_PLUS_0_765366865   /* d1[2] */
+  .short FIX_1_175875602                    /* d1[3] */
+  /* reloadable constants */
+  .short FIX_1_175875602_MINUS_0_390180644  /* d2[0] */
+  .short FIX_0_541196100_MINUS_1_847759065  /* d2[1] */
+  .short FIX_3_072711026_MINUS_2_562915447  /* d2[2] */
+  .short FIX_1_175875602_MINUS_1_961570560  /* d2[3] */
 
 asm_function jsimd_idct_islow_neon
 
@@ -253,140 +258,141 @@
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
     vmul.s16        q14, q14, q2
     vmul.s16        q13, q13, q1
-    vld1.16         {d0, d1, d2, d3}, [ip, :128] /* load constants */
+    vld1.16         {d0, d1, d2, d3}, [ip, :128]  /* load constants */
     add             ip, ip, #16
     vmul.s16        q15, q15, q3
-    vpush           {d8-d15} /* save NEON registers */
+    vpush           {d8-d15}                      /* save NEON registers */
     /* 1-D IDCT, pass 1, left 4x8 half */
-    vadd.s16        d4,    ROW7L, ROW3L
-    vadd.s16        d5,    ROW5L, ROW1L
-    vmull.s16       q6,    d4,    XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    d5,    XFIX_1_175875602
-    vmull.s16       q7,    d4,    XFIX_1_175875602
+    vadd.s16        d4, ROW7L, ROW3L
+    vadd.s16        d5, ROW5L, ROW1L
+    vmull.s16       q6, d4, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d5, XFIX_1_175875602
+    vmull.s16       q7, d4, XFIX_1_175875602
       /* Check for the zero coefficients in the right 4x8 half */
       push            {r4, r5}
-    vmlal.s16       q7,    d5,    XFIX_1_175875602_MINUS_0_390180644
-    vsubl.s16       q3,    ROW0L, ROW4L
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))]
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vmlal.s16       q2,    ROW6L, XFIX_0_541196100_MINUS_1_847759065
-      orr             r0,    r4,    r5
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5L, XFIX_2_562915447
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))]
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vshl.s32        q3,    q3,    #13
-      orr             r0,    r0,    r4
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-      orr             r0,    r0,    r5
-    vadd.s32        q1,    q3,    q2
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))]
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-      orr             r0,    r0,    r4
-    vmlsl.s16       q7,    ROW7L, XFIX_0_899976223
-      orr             r0,    r0,    r5
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vrshrn.s32      ROW1L, q1,    #11
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))]
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5L, XFIX_2_053119869_MINUS_2_562915447
-      orr             r0,    r0,    r4
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-      orr             r0,    r0,    r5
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))]
-    vmlal.s16       q6,    ROW6L, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-      orr             r0,    r0,    r4
-    vrshrn.s32      ROW6L, q1,    #11
-      orr             r0,    r0,    r5
-    vadd.s32        q1,    q3,    q5
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))]
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0L, ROW4L
-      orr             r0,    r0,    r4
-    vrshrn.s32      ROW2L, q1,    #11
-      orr             r0,    r0,    r5
-    vrshrn.s32      ROW5L, q3,    #11
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))]
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7L, XFIX_0_298631336_MINUS_0_899976223
-      orr             r0,    r0,    r4
-    vadd.s32        q2,    q5,    q6
-      orrs            r0,    r0,    r5
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))]
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-      orr             r0,    r4,    r5
-    vsub.s32        q3,    q1,    q4
+    vmlal.s16       q7, d5, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW4L
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))]
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW6L, XFIX_0_541196100_MINUS_1_847759065
+      orr             r0, r4, r5
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5L, XFIX_2_562915447
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))]
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+      orr             r0, r0, r4
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q2
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))]
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+      orr             r0, r0, r4
+    vmlsl.s16       q7, ROW7L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1L, q1, #11
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))]
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5L, XFIX_2_053119869_MINUS_2_562915447
+      orr             r0, r0, r4
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+      orr             r0, r0, r5
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))]
+    vmlal.s16       q6, ROW6L, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+      orr             r0, r0, r4
+    vrshrn.s32      ROW6L, q1, #11
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q5
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))]
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW4L
+      orr             r0, r0, r4
+    vrshrn.s32      ROW2L, q1, #11
+      orr             r0, r0, r5
+    vrshrn.s32      ROW5L, q3, #11
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))]
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7L, XFIX_0_298631336_MINUS_0_899976223
+      orr             r0, r0, r4
+    vadd.s32        q2, q5, q6
+      orrs            r0, r0, r5
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))]
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+      orr             r0, r4, r5
+    vsub.s32        q3, q1, q4
       pop             {r4, r5}
-    vrshrn.s32      ROW7L, q2,    #11
-    vrshrn.s32      ROW3L, q5,    #11
-    vrshrn.s32      ROW0L, q6,    #11
-    vrshrn.s32      ROW4L, q3,    #11
+    vrshrn.s32      ROW7L, q2, #11
+    vrshrn.s32      ROW3L, q5, #11
+    vrshrn.s32      ROW0L, q6, #11
+    vrshrn.s32      ROW4L, q3, #11
 
-      beq             3f /* Go to do some special handling for the sparse right 4x8 half */
+      beq             3f  /* Go to do some special handling for the sparse
+                             right 4x8 half */
 
     /* 1-D IDCT, pass 1, right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vadd.s16        d10,   ROW7R, ROW3R
-    vadd.s16        d8,    ROW5R, ROW1R
+    vld1.s16        {d2}, [ip, :64]  /* reload constants */
+    vadd.s16        d10, ROW7R, ROW3R
+    vadd.s16        d8, ROW5R, ROW1R
       /* Transpose left 4x8 half */
       vtrn.16         ROW6L, ROW7L
-    vmull.s16       q6,    d10,   XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    d8,    XFIX_1_175875602
+    vmull.s16       q6, d10, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d8, XFIX_1_175875602
       vtrn.16         ROW2L, ROW3L
-    vmull.s16       q7,    d10,   XFIX_1_175875602
-    vmlal.s16       q7,    d8,    XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q7, d10, XFIX_1_175875602
+    vmlal.s16       q7, d8, XFIX_1_175875602_MINUS_0_390180644
       vtrn.16         ROW0L, ROW1L
-    vsubl.s16       q3,    ROW0R, ROW4R
-    vmull.s16       q2,    ROW2R, XFIX_0_541196100
-    vmlal.s16       q2,    ROW6R, XFIX_0_541196100_MINUS_1_847759065
+    vsubl.s16       q3, ROW0R, ROW4R
+    vmull.s16       q2, ROW2R, XFIX_0_541196100
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
       vtrn.16         ROW4L, ROW5L
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5R, XFIX_2_562915447
-    vmlal.s16       q6,    ROW3R, XFIX_3_072711026_MINUS_2_562915447
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW3R, XFIX_3_072711026_MINUS_2_562915447
       vtrn.32         ROW1L, ROW3L
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW1R, XFIX_0_899976223
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1R, XFIX_0_899976223
       vtrn.32         ROW4L, ROW6L
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
       vtrn.32         ROW0L, ROW2L
-    vmlsl.s16       q7,    ROW7R, XFIX_0_899976223
-    vmlal.s16       q7,    ROW1R, XFIX_1_501321110_MINUS_0_899976223
-    vrshrn.s32      ROW1R, q1,    #11
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW1R, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1R, q1, #11
       vtrn.32         ROW5L, ROW7L
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5R, XFIX_2_053119869_MINUS_2_562915447
-    vmlsl.s16       q5,    ROW3R, XFIX_2_562915447
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2R, XFIX_0_541196100_PLUS_0_765366865
-    vmlal.s16       q6,    ROW6R, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-    vrshrn.s32      ROW6R, q1,    #11
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0R, ROW4R
-    vrshrn.s32      ROW2R, q1,    #11
-    vrshrn.s32      ROW5R, q3,    #11
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7R, XFIX_0_298631336_MINUS_0_899976223
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vrshrn.s32      ROW7R, q2,    #11
-    vrshrn.s32      ROW3R, q5,    #11
-    vrshrn.s32      ROW0R, q6,    #11
-    vrshrn.s32      ROW4R, q3,    #11
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW3R, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2R, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vrshrn.s32      ROW6R, q1, #11
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0R, ROW4R
+    vrshrn.s32      ROW2R, q1, #11
+    vrshrn.s32      ROW5R, q3, #11
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vrshrn.s32      ROW7R, q2, #11
+    vrshrn.s32      ROW3R, q5, #11
+    vrshrn.s32      ROW0R, q6, #11
+    vrshrn.s32      ROW4R, q3, #11
     /* Transpose right 4x8 half */
     vtrn.16         ROW6R, ROW7R
     vtrn.16         ROW2R, ROW3R
@@ -398,122 +404,122 @@
     vtrn.32         ROW5R, ROW7R
 
 1:  /* 1-D IDCT, pass 2 (normal variant), left 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW1R, XFIX_1_175875602 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW1L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW3R, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */
-    vmlal.s16       q6,    ROW3L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW3R, XFIX_1_175875602 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW3L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW1R, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */
-    vmlal.s16       q7,    ROW1L, XFIX_1_175875602_MINUS_0_390180644
-    vsubl.s16       q3,    ROW0L, ROW0R /* ROW4L <-> ROW0R */
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vmlal.s16       q2,    ROW2R, XFIX_0_541196100_MINUS_1_847759065 /* ROW6L <-> ROW2R */
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW1R, XFIX_2_562915447 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-    vmlsl.s16       q7,    ROW3R, XFIX_0_899976223 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vshrn.s32       ROW1L, q1,    #16
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW1R, XFIX_2_053119869_MINUS_2_562915447 /* ROW5L <-> ROW1R */
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-    vmlal.s16       q6,    ROW2R, XFIX_0_541196100 /* ROW6L <-> ROW2R */
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW2R, q1,    #16 /* ROW6L <-> ROW2R */
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0L, ROW0R /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW2L, q1,    #16
-    vshrn.s32       ROW1R, q3,    #16 /* ROW5L <-> ROW1R */
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW3R, XFIX_0_298631336_MINUS_0_899976223 /* ROW7L <-> ROW3R */
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW3R, q2,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW3L, q5,    #16
-    vshrn.s32       ROW0L, q6,    #16
-    vshrn.s32       ROW0R, q3,    #16 /* ROW4L <-> ROW0R */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW1R, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3R, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3R, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1R, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW2R, XFIX_0_541196100_MINUS_1_847759065  /* ROW6L <-> ROW2R */
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW1R, XFIX_2_562915447   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW3R, XFIX_0_899976223   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW1R, XFIX_2_053119869_MINUS_2_562915447  /* ROW5L <-> ROW1R */
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW2R, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW3R, XFIX_0_298631336_MINUS_0_899976223  /* ROW7L <-> ROW3R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
     /* 1-D IDCT, pass 2, right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW5R, XFIX_1_175875602
-    vmlal.s16       q6,    ROW5L, XFIX_1_175875602 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW7R, XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    ROW7L, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */
-    vmull.s16       q7,    ROW7R, XFIX_1_175875602
-    vmlal.s16       q7,    ROW7L, XFIX_1_175875602 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW5R, XFIX_1_175875602_MINUS_0_390180644
-    vmlal.s16       q7,    ROW5L, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */
-    vsubl.s16       q3,    ROW4L, ROW4R /* ROW4L <-> ROW0R */
-    vmull.s16       q2,    ROW6L, XFIX_0_541196100 /* ROW6L <-> ROW2R */
-    vmlal.s16       q2,    ROW6R, XFIX_0_541196100_MINUS_1_847759065
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5R, XFIX_2_562915447
-    vmlal.s16       q6,    ROW7L, XFIX_3_072711026_MINUS_2_562915447 /* ROW7L <-> ROW3R */
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW5L, XFIX_0_899976223 /* ROW5L <-> ROW1R */
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-    vmlsl.s16       q7,    ROW7R, XFIX_0_899976223
-    vmlal.s16       q7,    ROW5L, XFIX_1_501321110_MINUS_0_899976223 /* ROW5L <-> ROW1R */
-    vshrn.s32       ROW5L, q1,    #16 /* ROW5L <-> ROW1R */
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5R, XFIX_2_053119869_MINUS_2_562915447
-    vmlsl.s16       q5,    ROW7L, XFIX_2_562915447 /* ROW7L <-> ROW3R */
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW6L, XFIX_0_541196100_PLUS_0_765366865 /* ROW6L <-> ROW2R */
-    vmlal.s16       q6,    ROW6R, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW6R, q1,    #16
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW4L, ROW4R /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW6L, q1,    #16 /* ROW6L <-> ROW2R */
-    vshrn.s32       ROW5R, q3,    #16
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7R, XFIX_0_298631336_MINUS_0_899976223
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW7R, q2,    #16
-    vshrn.s32       ROW7L, q5,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW4L, q6,    #16 /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW4R, q3,    #16
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW5R, XFIX_1_175875602
+    vmlal.s16       q6, ROW5L, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW7R, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmull.s16       q7, ROW7R, XFIX_1_175875602
+    vmlal.s16       q7, ROW7L, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW5R, XFIX_1_175875602_MINUS_0_390180644
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vsubl.s16       q3, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW6L, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447  /* ROW7L <-> ROW3R */
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223   /* ROW5L <-> ROW1R */
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223  /* ROW5L <-> ROW1R */
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447   /* ROW7L <-> ROW3R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865  /* ROW6L <-> ROW2R */
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
 
 2:  /* Descale to 8-bit and range limit */
-    vqrshrn.s16     d16,   q8,    #2
-    vqrshrn.s16     d17,   q9,    #2
-    vqrshrn.s16     d18,   q10,   #2
-    vqrshrn.s16     d19,   q11,   #2
-    vpop            {d8-d15} /* restore NEON registers */
-    vqrshrn.s16     d20,   q12,   #2
+    vqrshrn.s16     d16, q8, #2
+    vqrshrn.s16     d17, q9, #2
+    vqrshrn.s16     d18, q10, #2
+    vqrshrn.s16     d19, q11, #2
+    vpop            {d8-d15}                      /* restore NEON registers */
+    vqrshrn.s16     d20, q12, #2
       /* Transpose the final 8-bit samples and do signed->unsigned conversion */
-      vtrn.16         q8,    q9
-    vqrshrn.s16     d21,   q13,   #2
-    vqrshrn.s16     d22,   q14,   #2
-      vmov.u8         q0,    #(CENTERJSAMPLE)
-    vqrshrn.s16     d23,   q15,   #2
-      vtrn.8          d16,   d17
-      vtrn.8          d18,   d19
-      vadd.u8         q8,    q8,    q0
-      vadd.u8         q9,    q9,    q0
-      vtrn.16         q10,   q11
+      vtrn.16         q8, q9
+    vqrshrn.s16     d21, q13, #2
+    vqrshrn.s16     d22, q14, #2
+      vmov.u8         q0, #(CENTERJSAMPLE)
+    vqrshrn.s16     d23, q15, #2
+      vtrn.8          d16, d17
+      vtrn.8          d18, d19
+      vadd.u8         q8, q8, q0
+      vadd.u8         q9, q9, q0
+      vtrn.16         q10, q11
         /* Store results to the output buffer */
         ldmia           OUTPUT_BUF!, {TMP1, TMP2}
         add             TMP1, TMP1, OUTPUT_COL
@@ -525,7 +531,7 @@
         add             TMP1, TMP1, OUTPUT_COL
         add             TMP2, TMP2, OUTPUT_COL
         vst1.8          {d18}, [TMP1]
-      vadd.u8         q10,   q10,   q0
+      vadd.u8         q10, q10, q0
         vst1.8          {d19}, [TMP2]
         ldmia           OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4}
         add             TMP1, TMP1, OUTPUT_COL
@@ -534,7 +540,7 @@
         add             TMP4, TMP4, OUTPUT_COL
       vtrn.8          d22, d23
         vst1.8          {d20}, [TMP1]
-      vadd.u8         q11,   q11,   q0
+      vadd.u8         q11, q11, q0
         vst1.8          {d21}, [TMP2]
         vst1.8          {d22}, [TMP3]
         vst1.8          {d23}, [TMP4]
@@ -547,14 +553,15 @@
     vtrn.16         ROW2L, ROW3L
     vtrn.16         ROW0L, ROW1L
     vtrn.16         ROW4L, ROW5L
-    vshl.s16        ROW0R, ROW0R, #2 /* PASS1_BITS */
+    vshl.s16        ROW0R, ROW0R, #2  /* PASS1_BITS */
     vtrn.32         ROW1L, ROW3L
     vtrn.32         ROW4L, ROW6L
     vtrn.32         ROW0L, ROW2L
     vtrn.32         ROW5L, ROW7L
 
     cmp             r0, #0
-    beq             4f /* Right 4x8 half has all zeros, go to 'sparse' second pass */
+    beq             4f  /* Right 4x8 half has all zeros, go to 'sparse' second
+                           pass */
 
     /* Only row 0 is non-zero for the right 4x8 half  */
     vdup.s16        ROW1R, ROW0R[1]
@@ -565,83 +572,83 @@
     vdup.s16        ROW6R, ROW0R[2]
     vdup.s16        ROW7R, ROW0R[3]
     vdup.s16        ROW0R, ROW0R[0]
-    b               1b /* Go to 'normal' second pass */
+    b               1b  /* Go to 'normal' second pass */
 
 4:  /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), left 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW1L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW3L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW3L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW1L, XFIX_1_175875602_MINUS_0_390180644
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vshll.s16       q3,    ROW0L, #13
-    vmov            q4,    q6
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vadd.s32        q1,    q1,    q6
-    vadd.s32        q6,    q6,    q6
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-    vshrn.s32       ROW1L, q1,    #16
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW2R, q1,    #16 /* ROW6L <-> ROW2R */
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vshll.s16       q5,    ROW0L, #13
-    vshrn.s32       ROW2L, q1,    #16
-    vshrn.s32       ROW1R, q3,    #16 /* ROW5L <-> ROW1R */
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW3R, q2,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW3L, q5,    #16
-    vshrn.s32       ROW0L, q6,    #16
-    vshrn.s32       ROW0R, q3,    #16 /* ROW4L <-> ROW0R */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vshll.s16       q3, ROW0L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW0L, #13
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
     /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW5L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW7L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW7L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW5L, XFIX_1_175875602_MINUS_0_390180644
-    vmull.s16       q2,    ROW6L, XFIX_0_541196100
-    vshll.s16       q3,    ROW4L, #13
-    vmov            q4,    q6
-    vmlal.s16       q6,    ROW7L, XFIX_3_072711026_MINUS_2_562915447
-    vmlsl.s16       q4,    ROW5L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vmlal.s16       q7,    ROW5L, XFIX_1_501321110_MINUS_0_899976223
-    vadd.s32        q1,    q1,    q6
-    vadd.s32        q6,    q6,    q6
-    vmlsl.s16       q5,    ROW7L, XFIX_2_562915447
-    vshrn.s32       ROW5L, q1,    #16 /* ROW5L <-> ROW1R */
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW6L, XFIX_0_541196100_PLUS_0_765366865
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW6R, q1,    #16
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vshll.s16       q5,    ROW4L, #13
-    vshrn.s32       ROW6L, q1,    #16 /* ROW6L <-> ROW2R */
-    vshrn.s32       ROW5R, q3,    #16
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW7R, q2,    #16
-    vshrn.s32       ROW7L, q5,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW4L, q6,    #16 /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW4R, q3,    #16
-    b               2b /* Go to epilogue */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW5L, XFIX_1_175875602
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW7L, XFIX_1_175875602
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW6L, XFIX_0_541196100
+    vshll.s16       q3, ROW4L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW4L, #13
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
+    b               2b                            /* Go to epilogue */
 
     .unreq          DCT_TABLE
     .unreq          COEF_BLOCK
@@ -668,7 +675,7 @@
     .unreq          ROW6R
     .unreq          ROW7L
     .unreq          ROW7R
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -695,10 +702,10 @@
 
 .balign 16
 jsimd_idct_ifast_neon_consts:
-    .short (277 * 128 - 256 * 128) /* XFIX_1_082392200 */
-    .short (362 * 128 - 256 * 128) /* XFIX_1_414213562 */
-    .short (473 * 128 - 256 * 128) /* XFIX_1_847759065 */
-    .short (669 * 128 - 512 * 128) /* XFIX_2_613125930 */
+  .short (277 * 128 - 256 * 128)  /* XFIX_1_082392200 */
+  .short (362 * 128 - 256 * 128)  /* XFIX_1_414213562 */
+  .short (473 * 128 - 256 * 128)  /* XFIX_1_847759065 */
+  .short (669 * 128 - 512 * 128)  /* XFIX_2_613125930 */
 
 asm_function jsimd_idct_ifast_neon
 
@@ -728,9 +735,9 @@
     vld1.16         {d16, d17, d18, d19}, [COEF_BLOCK, :128]!
     vld1.16         {d0, d1, d2, d3}, [DCT_TABLE, :128]!
     vld1.16         {d20, d21, d22, d23}, [COEF_BLOCK, :128]!
-    vmul.s16        q8,  q8,  q0
+    vmul.s16        q8, q8, q0
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
-    vmul.s16        q9,  q9,  q1
+    vmul.s16        q9, q9, q1
     vld1.16         {d24, d25, d26, d27}, [COEF_BLOCK, :128]!
     vmul.s16        q10, q10, q2
     vld1.16         {d0, d1, d2, d3}, [DCT_TABLE, :128]!
@@ -740,124 +747,124 @@
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
     vmul.s16        q14, q14, q2
     vmul.s16        q13, q13, q1
-    vld1.16         {d0}, [ip, :64] /* load constants */
+    vld1.16         {d0}, [ip, :64]  /* load constants */
     vmul.s16        q15, q15, q3
-    vpush           {d8-d13}        /* save NEON registers */
+    vpush           {d8-d13}         /* save NEON registers */
     /* 1-D IDCT, pass 1 */
-    vsub.s16        q2,  q10, q14
+    vsub.s16        q2, q10, q14
     vadd.s16        q14, q10, q14
-    vsub.s16        q1,  q11, q13
+    vsub.s16        q1, q11, q13
     vadd.s16        q13, q11, q13
-    vsub.s16        q5,  q9,  q15
-    vadd.s16        q15, q9,  q15
-    vqdmulh.s16     q4,  q2,  XFIX_1_414213562
-    vqdmulh.s16     q6,  q1,  XFIX_2_613125930
-    vadd.s16        q3,  q1,  q1
-    vsub.s16        q1,  q5,  q1
-    vadd.s16        q10, q2,  q4
-    vqdmulh.s16     q4,  q1,  XFIX_1_847759065
-    vsub.s16        q2,  q15, q13
-    vadd.s16        q3,  q3,  q6
-    vqdmulh.s16     q6,  q2,  XFIX_1_414213562
-    vadd.s16        q1,  q1,  q4
-    vqdmulh.s16     q4,  q5,  XFIX_1_082392200
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
     vsub.s16        q10, q10, q14
-    vadd.s16        q2,  q2,  q6
-    vsub.s16        q6,  q8,  q12
-    vadd.s16        q12, q8,  q12
-    vadd.s16        q9,  q5,  q4
-    vadd.s16        q5,  q6,  q10
-    vsub.s16        q10, q6,  q10
-    vadd.s16        q6,  q15, q13
-    vadd.s16        q8,  q12, q14
-    vsub.s16        q3,  q6,  q3
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
     vsub.s16        q12, q12, q14
-    vsub.s16        q3,  q3,  q1
-    vsub.s16        q1,  q9,  q1
-    vadd.s16        q2,  q3,  q2
-    vsub.s16        q15, q8,  q6
-    vadd.s16        q1,  q1,  q2
-    vadd.s16        q8,  q8,  q6
-    vadd.s16        q14, q5,  q3
-    vsub.s16        q9,  q5,  q3
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
     vsub.s16        q13, q10, q2
     vadd.s16        q10, q10, q2
       /* Transpose */
-      vtrn.16         q8,  q9
+      vtrn.16         q8, q9
     vsub.s16        q11, q12, q1
       vtrn.16         q14, q15
     vadd.s16        q12, q12, q1
       vtrn.16         q10, q11
       vtrn.16         q12, q13
-      vtrn.32         q9,  q11
+      vtrn.32         q9, q11
       vtrn.32         q12, q14
-      vtrn.32         q8,  q10
+      vtrn.32         q8, q10
       vtrn.32         q13, q15
       vswp            d28, d21
       vswp            d26, d19
     /* 1-D IDCT, pass 2 */
-    vsub.s16        q2,  q10, q14
+    vsub.s16        q2, q10, q14
       vswp            d30, d23
     vadd.s16        q14, q10, q14
       vswp            d24, d17
-    vsub.s16        q1,  q11, q13
+    vsub.s16        q1, q11, q13
     vadd.s16        q13, q11, q13
-    vsub.s16        q5,  q9,  q15
-    vadd.s16        q15, q9,  q15
-    vqdmulh.s16     q4,  q2,  XFIX_1_414213562
-    vqdmulh.s16     q6,  q1,  XFIX_2_613125930
-    vadd.s16        q3,  q1,  q1
-    vsub.s16        q1,  q5,  q1
-    vadd.s16        q10, q2,  q4
-    vqdmulh.s16     q4,  q1,  XFIX_1_847759065
-    vsub.s16        q2,  q15, q13
-    vadd.s16        q3,  q3,  q6
-    vqdmulh.s16     q6,  q2,  XFIX_1_414213562
-    vadd.s16        q1,  q1,  q4
-    vqdmulh.s16     q4,  q5,  XFIX_1_082392200
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
     vsub.s16        q10, q10, q14
-    vadd.s16        q2,  q2,  q6
-    vsub.s16        q6,  q8,  q12
-    vadd.s16        q12, q8,  q12
-    vadd.s16        q9,  q5,  q4
-    vadd.s16        q5,  q6,  q10
-    vsub.s16        q10, q6,  q10
-    vadd.s16        q6,  q15, q13
-    vadd.s16        q8,  q12, q14
-    vsub.s16        q3,  q6,  q3
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
     vsub.s16        q12, q12, q14
-    vsub.s16        q3,  q3,  q1
-    vsub.s16        q1,  q9,  q1
-    vadd.s16        q2,  q3,  q2
-    vsub.s16        q15, q8,  q6
-    vadd.s16        q1,  q1,  q2
-    vadd.s16        q8,  q8,  q6
-    vadd.s16        q14, q5,  q3
-    vsub.s16        q9,  q5,  q3
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
     vsub.s16        q13, q10, q2
-    vpop            {d8-d13}        /* restore NEON registers */
+    vpop            {d8-d13}      /* restore NEON registers */
     vadd.s16        q10, q10, q2
     vsub.s16        q11, q12, q1
     vadd.s16        q12, q12, q1
     /* Descale to 8-bit and range limit */
-    vmov.u8         q0,  #0x80
-    vqshrn.s16      d16, q8,  #5
-    vqshrn.s16      d17, q9,  #5
+    vmov.u8         q0, #0x80
+    vqshrn.s16      d16, q8, #5
+    vqshrn.s16      d17, q9, #5
     vqshrn.s16      d18, q10, #5
     vqshrn.s16      d19, q11, #5
     vqshrn.s16      d20, q12, #5
     vqshrn.s16      d21, q13, #5
     vqshrn.s16      d22, q14, #5
     vqshrn.s16      d23, q15, #5
-    vadd.u8         q8,  q8,  q0
-    vadd.u8         q9,  q9,  q0
+    vadd.u8         q8, q8, q0
+    vadd.u8         q9, q9, q0
     vadd.u8         q10, q10, q0
     vadd.u8         q11, q11, q0
     /* Transpose the final 8-bit samples */
-    vtrn.16         q8,  q9
+    vtrn.16         q8, q9
     vtrn.16         q10, q11
-    vtrn.32         q8,  q10
-    vtrn.32         q9,  q11
+    vtrn.32         q8, q10
+    vtrn.32         q9, q11
     vtrn.8          d16, d17
     vtrn.8          d18, d19
       /* Store results to the output buffer */
@@ -892,7 +899,7 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -916,81 +923,80 @@
 
 #define CONST_BITS  13
 
-#define FIX_0_211164243  (1730)  /* FIX(0.211164243) */
-#define FIX_0_509795579  (4176)  /* FIX(0.509795579) */
-#define FIX_0_601344887  (4926)  /* FIX(0.601344887) */
-#define FIX_0_720959822  (5906)  /* FIX(0.720959822) */
-#define FIX_0_765366865  (6270)  /* FIX(0.765366865) */
-#define FIX_0_850430095  (6967)  /* FIX(0.850430095) */
-#define FIX_0_899976223  (7373)  /* FIX(0.899976223) */
-#define FIX_1_061594337  (8697)  /* FIX(1.061594337) */
-#define FIX_1_272758580  (10426) /* FIX(1.272758580) */
-#define FIX_1_451774981  (11893) /* FIX(1.451774981) */
-#define FIX_1_847759065  (15137) /* FIX(1.847759065) */
-#define FIX_2_172734803  (17799) /* FIX(2.172734803) */
-#define FIX_2_562915447  (20995) /* FIX(2.562915447) */
-#define FIX_3_624509785  (29692) /* FIX(3.624509785) */
+#define FIX_0_211164243 (1730)   /* FIX(0.211164243) */
+#define FIX_0_509795579 (4176)   /* FIX(0.509795579) */
+#define FIX_0_601344887 (4926)   /* FIX(0.601344887) */
+#define FIX_0_720959822 (5906)   /* FIX(0.720959822) */
+#define FIX_0_765366865 (6270)   /* FIX(0.765366865) */
+#define FIX_0_850430095 (6967)   /* FIX(0.850430095) */
+#define FIX_0_899976223 (7373)   /* FIX(0.899976223) */
+#define FIX_1_061594337 (8697)   /* FIX(1.061594337) */
+#define FIX_1_272758580 (10426)  /* FIX(1.272758580) */
+#define FIX_1_451774981 (11893)  /* FIX(1.451774981) */
+#define FIX_1_847759065 (15137)  /* FIX(1.847759065) */
+#define FIX_2_172734803 (17799)  /* FIX(2.172734803) */
+#define FIX_2_562915447 (20995)  /* FIX(2.562915447) */
+#define FIX_3_624509785 (29692)  /* FIX(3.624509785) */
 
 .balign 16
 jsimd_idct_4x4_neon_consts:
-    .short     FIX_1_847759065     /* d0[0] */
-    .short     -FIX_0_765366865    /* d0[1] */
-    .short     -FIX_0_211164243    /* d0[2] */
-    .short     FIX_1_451774981     /* d0[3] */
-    .short     -FIX_2_172734803    /* d1[0] */
-    .short     FIX_1_061594337     /* d1[1] */
-    .short     -FIX_0_509795579    /* d1[2] */
-    .short     -FIX_0_601344887    /* d1[3] */
-    .short     FIX_0_899976223     /* d2[0] */
-    .short     FIX_2_562915447     /* d2[1] */
-    .short     1 << (CONST_BITS+1) /* d2[2] */
-    .short     0                   /* d2[3] */
+  .short FIX_1_847759065      /* d0[0] */
+  .short -FIX_0_765366865     /* d0[1] */
+  .short -FIX_0_211164243     /* d0[2] */
+  .short FIX_1_451774981      /* d0[3] */
+  .short -FIX_2_172734803     /* d1[0] */
+  .short FIX_1_061594337      /* d1[1] */
+  .short -FIX_0_509795579     /* d1[2] */
+  .short -FIX_0_601344887     /* d1[3] */
+  .short FIX_0_899976223      /* d2[0] */
+  .short FIX_2_562915447      /* d2[1] */
+  .short 1 << (CONST_BITS+1)  /* d2[2] */
+  .short 0                    /* d2[3] */
 
 .macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29
-    vmull.s16       q14, \x4,  d2[2]
-    vmlal.s16       q14, \x8,  d0[0]
+    vmull.s16       q14, \x4, d2[2]
+    vmlal.s16       q14, \x8, d0[0]
     vmlal.s16       q14, \x14, d0[1]
 
     vmull.s16       q13, \x16, d1[2]
     vmlal.s16       q13, \x12, d1[3]
     vmlal.s16       q13, \x10, d2[0]
-    vmlal.s16       q13, \x6,  d2[1]
+    vmlal.s16       q13, \x6, d2[1]
 
-    vmull.s16       q15, \x4,  d2[2]
-    vmlsl.s16       q15, \x8,  d0[0]
+    vmull.s16       q15, \x4, d2[2]
+    vmlsl.s16       q15, \x8, d0[0]
     vmlsl.s16       q15, \x14, d0[1]
 
     vmull.s16       q12, \x16, d0[2]
     vmlal.s16       q12, \x12, d0[3]
     vmlal.s16       q12, \x10, d1[0]
-    vmlal.s16       q12, \x6,  d1[1]
+    vmlal.s16       q12, \x6, d1[1]
 
     vadd.s32        q10, q14, q13
     vsub.s32        q14, q14, q13
 
-.if \shift > 16
-    vrshr.s32       q10,  q10, #\shift
-    vrshr.s32       q14,  q14, #\shift
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q14, q14, #\shift
     vmovn.s32       \y26, q10
     vmovn.s32       \y29, q14
-.else
+  .else
     vrshrn.s32      \y26, q10, #\shift
     vrshrn.s32      \y29, q14, #\shift
-.endif
+  .endif
 
     vadd.s32        q10, q15, q12
     vsub.s32        q15, q15, q12
 
-.if \shift > 16
-    vrshr.s32       q10,  q10, #\shift
-    vrshr.s32       q15,  q15, #\shift
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q15, q15, #\shift
     vmovn.s32       \y27, q10
     vmovn.s32       \y28, q15
-.else
+  .else
     vrshrn.s32      \y27, q10, #\shift
     vrshrn.s32      \y28, q15, #\shift
-.endif
-
+  .endif
 .endm
 
 asm_function jsimd_idct_4x4_neon
@@ -1104,10 +1110,10 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
 
 .purgem idct_helper
 
+
 /*****************************************************************************/
 
 /*
@@ -1126,31 +1132,30 @@
 
 .balign 8
 jsimd_idct_2x2_neon_consts:
-    .short     -FIX_0_720959822    /* d0[0] */
-    .short     FIX_0_850430095     /* d0[1] */
-    .short     -FIX_1_272758580    /* d0[2] */
-    .short     FIX_3_624509785     /* d0[3] */
+  .short -FIX_0_720959822  /* d0[0] */
+  .short FIX_0_850430095   /* d0[1] */
+  .short -FIX_1_272758580  /* d0[2] */
+  .short FIX_3_624509785   /* d0[3] */
 
 .macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27
-    vshll.s16  q14,  \x4,  #15
-    vmull.s16  q13,  \x6,  d0[3]
-    vmlal.s16  q13,  \x10, d0[2]
-    vmlal.s16  q13,  \x12, d0[1]
-    vmlal.s16  q13,  \x16, d0[0]
-
-    vadd.s32   q10,  q14,  q13
-    vsub.s32   q14,  q14,  q13
-
-.if \shift > 16
-    vrshr.s32  q10,  q10,  #\shift
-    vrshr.s32  q14,  q14,  #\shift
-    vmovn.s32  \y26, q10
-    vmovn.s32  \y27, q14
-.else
-    vrshrn.s32 \y26, q10,  #\shift
-    vrshrn.s32 \y27, q14,  #\shift
-.endif
+    vshll.s16       q14, \x4, #15
+    vmull.s16       q13, \x6, d0[3]
+    vmlal.s16       q13, \x10, d0[2]
+    vmlal.s16       q13, \x12, d0[1]
+    vmlal.s16       q13, \x16, d0[0]
 
+    vadd.s32        q10, q14, q13
+    vsub.s32        q14, q14, q13
+
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q14, q14, #\shift
+    vmovn.s32       \y26, q10
+    vmovn.s32       \y27, q14
+  .else
+    vrshrn.s32      \y26, q10, #\shift
+    vrshrn.s32      \y27, q14, #\shift
+  .endif
 .endm
 
 asm_function jsimd_idct_2x2_neon
@@ -1204,30 +1209,30 @@
     /* Pass 1 */
 #if 0
     idct_helper     d4, d6, d10, d12, d16, 13, d4, d6
-    transpose_4x4   d4, d6, d8,  d10
+    transpose_4x4   d4, d6, d8, d10
     idct_helper     d5, d7, d11, d13, d17, 13, d5, d7
-    transpose_4x4   d5, d7, d9,  d11
+    transpose_4x4   d5, d7, d9, d11
 #else
-    vmull.s16       q13, d6,  d0[3]
+    vmull.s16       q13, d6, d0[3]
     vmlal.s16       q13, d10, d0[2]
     vmlal.s16       q13, d12, d0[1]
     vmlal.s16       q13, d16, d0[0]
-    vmull.s16       q12, d7,  d0[3]
+    vmull.s16       q12, d7, d0[3]
     vmlal.s16       q12, d11, d0[2]
     vmlal.s16       q12, d13, d0[1]
     vmlal.s16       q12, d17, d0[0]
-    vshll.s16       q14, d4,  #15
-    vshll.s16       q15, d5,  #15
+    vshll.s16       q14, d4, #15
+    vshll.s16       q15, d5, #15
     vadd.s32        q10, q14, q13
     vsub.s32        q14, q14, q13
-    vrshrn.s32      d4,  q10, #13
-    vrshrn.s32      d6,  q14, #13
+    vrshrn.s32      d4, q10, #13
+    vrshrn.s32      d6, q14, #13
     vadd.s32        q10, q15, q12
     vsub.s32        q14, q15, q12
-    vrshrn.s32      d5,  q10, #13
-    vrshrn.s32      d7,  q14, #13
-    vtrn.16         q2,  q3
-    vtrn.32         q3,  q5
+    vrshrn.s32      d5, q10, #13
+    vrshrn.s32      d7, q14, #13
+    vtrn.16         q2, q3
+    vtrn.32         q3, q5
 #endif
 
     /* Pass 2 */
@@ -1258,10 +1263,10 @@
     .unreq          OUTPUT_COL
     .unreq          TMP1
     .unreq          TMP2
-.endfunc
 
 .purgem idct_helper
 
+
 /*****************************************************************************/
 
 /*
@@ -1277,97 +1282,110 @@
 
 
 .macro do_load size
+  .if \size == 8
+    vld1.8          {d4}, [U, :64]!
+    vld1.8          {d5}, [V, :64]!
+    vld1.8          {d0}, [Y, :64]!
+    pld             [U, #64]
+    pld             [V, #64]
+    pld             [Y, #64]
+  .elseif \size == 4
+    vld1.8          {d4[0]}, [U]!
+    vld1.8          {d4[1]}, [U]!
+    vld1.8          {d4[2]}, [U]!
+    vld1.8          {d4[3]}, [U]!
+    vld1.8          {d5[0]}, [V]!
+    vld1.8          {d5[1]}, [V]!
+    vld1.8          {d5[2]}, [V]!
+    vld1.8          {d5[3]}, [V]!
+    vld1.8          {d0[0]}, [Y]!
+    vld1.8          {d0[1]}, [Y]!
+    vld1.8          {d0[2]}, [Y]!
+    vld1.8          {d0[3]}, [Y]!
+  .elseif \size == 2
+    vld1.8          {d4[4]}, [U]!
+    vld1.8          {d4[5]}, [U]!
+    vld1.8          {d5[4]}, [V]!
+    vld1.8          {d5[5]}, [V]!
+    vld1.8          {d0[4]}, [Y]!
+    vld1.8          {d0[5]}, [Y]!
+  .elseif \size == 1
+    vld1.8          {d4[6]}, [U]!
+    vld1.8          {d5[6]}, [V]!
+    vld1.8          {d0[6]}, [Y]!
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_store bpp, size
+  .if \bpp == 24
     .if \size == 8
-        vld1.8  {d4}, [U, :64]!
-        vld1.8  {d5}, [V, :64]!
-        vld1.8  {d0}, [Y, :64]!
-        pld     [U, #64]
-        pld     [V, #64]
-        pld     [Y, #64]
+      vst3.8        {d10, d11, d12}, [RGB]!
     .elseif \size == 4
-        vld1.8  {d4[0]}, [U]!
-        vld1.8  {d4[1]}, [U]!
-        vld1.8  {d4[2]}, [U]!
-        vld1.8  {d4[3]}, [U]!
-        vld1.8  {d5[0]}, [V]!
-        vld1.8  {d5[1]}, [V]!
-        vld1.8  {d5[2]}, [V]!
-        vld1.8  {d5[3]}, [V]!
-        vld1.8  {d0[0]}, [Y]!
-        vld1.8  {d0[1]}, [Y]!
-        vld1.8  {d0[2]}, [Y]!
-        vld1.8  {d0[3]}, [Y]!
+      vst3.8        {d10[0], d11[0], d12[0]}, [RGB]!
+      vst3.8        {d10[1], d11[1], d12[1]}, [RGB]!
+      vst3.8        {d10[2], d11[2], d12[2]}, [RGB]!
+      vst3.8        {d10[3], d11[3], d12[3]}, [RGB]!
     .elseif \size == 2
-        vld1.8  {d4[4]}, [U]!
-        vld1.8  {d4[5]}, [U]!
-        vld1.8  {d5[4]}, [V]!
-        vld1.8  {d5[5]}, [V]!
-        vld1.8  {d0[4]}, [Y]!
-        vld1.8  {d0[5]}, [Y]!
+      vst3.8        {d10[4], d11[4], d12[4]}, [RGB]!
+      vst3.8        {d10[5], d11[5], d12[5]}, [RGB]!
     .elseif \size == 1
-        vld1.8  {d4[6]}, [U]!
-        vld1.8  {d5[6]}, [V]!
-        vld1.8  {d0[6]}, [Y]!
+      vst3.8        {d10[6], d11[6], d12[6]}, [RGB]!
     .else
-        .error unsupported macroblock size
+      .error unsupported macroblock size
     .endif
-.endm
-
-.macro do_store bpp, size
-    .if \bpp == 24
-        .if \size == 8
-            vst3.8  {d10, d11, d12}, [RGB]!
-        .elseif \size == 4
-            vst3.8  {d10[0], d11[0], d12[0]}, [RGB]!
-            vst3.8  {d10[1], d11[1], d12[1]}, [RGB]!
-            vst3.8  {d10[2], d11[2], d12[2]}, [RGB]!
-            vst3.8  {d10[3], d11[3], d12[3]}, [RGB]!
-        .elseif \size == 2
-            vst3.8  {d10[4], d11[4], d12[4]}, [RGB]!
-            vst3.8  {d10[5], d11[5], d12[5]}, [RGB]!
-        .elseif \size == 1
-            vst3.8  {d10[6], d11[6], d12[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
-    .elseif \bpp == 32
-        .if \size == 8
-            vst4.8  {d10, d11, d12, d13}, [RGB]!
-        .elseif \size == 4
-            vst4.8  {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
-            vst4.8  {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
-            vst4.8  {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
-            vst4.8  {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
-        .elseif \size == 2
-            vst4.8  {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
-            vst4.8  {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
-        .elseif \size == 1
-            vst4.8  {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      vst4.8        {d10, d11, d12, d13}, [RGB]!
+    .elseif \size == 4
+      vst4.8        {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
+      vst4.8        {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
+      vst4.8        {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
+      vst4.8        {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
+    .elseif \size == 2
+      vst4.8        {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
+      vst4.8        {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
+    .elseif \size == 1
+      vst4.8        {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
     .else
-        .error unsupported bpp
+      .error unsupported macroblock size
     .endif
+  .elseif \bpp == 16
+    .if \size == 8
+      vst1.16       {q15}, [RGB]!
+    .elseif \size == 4
+      vst1.16       {d30}, [RGB]!
+    .elseif \size == 2
+      vst1.16       {d31[0]}, [RGB]!
+      vst1.16       {d31[1]}, [RGB]!
+    .elseif \size == 1
+      vst1.16       {d31[2]}, [RGB]!
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
 .endm
 
 .macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, g_offs, b_offs
 
 /*
- * 2 stage pipelined YCbCr->RGB conversion
+ * 2-stage pipelined YCbCr->RGB conversion
  */
 
 .macro do_yuv_to_rgb_stage1
-    vaddw.u8        q3, q1, d4     /* q3 = u - 128 */
-    vaddw.u8        q4, q1, d5     /* q2 = v - 128 */
-    vmull.s16       q10, d6, d1[1] /* multiply by -11277 */
-    vmlal.s16       q10, d8, d1[2] /* multiply by -23401 */
-    vmull.s16       q11, d7, d1[1] /* multiply by -11277 */
-    vmlal.s16       q11, d9, d1[2] /* multiply by -23401 */
-    vmull.s16       q12, d8, d1[0] /* multiply by 22971 */
-    vmull.s16       q13, d9, d1[0] /* multiply by 22971 */
-    vmull.s16       q14, d6, d1[3] /* multiply by 29033 */
-    vmull.s16       q15, d7, d1[3] /* multiply by 29033 */
+    vaddw.u8        q3, q1, d4      /* q3 = u - 128 */
+    vaddw.u8        q4, q1, d5      /* q2 = v - 128 */
+    vmull.s16       q10, d6, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q10, d8, d1[2]  /* multiply by -23401 */
+    vmull.s16       q11, d7, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q11, d9, d1[2]  /* multiply by -23401 */
+    vmull.s16       q12, d8, d1[0]  /* multiply by 22971 */
+    vmull.s16       q13, d9, d1[0]  /* multiply by 22971 */
+    vmull.s16       q14, d6, d1[3]  /* multiply by 29033 */
+    vmull.s16       q15, d7, d1[3]  /* multiply by 29033 */
 .endm
 
 .macro do_yuv_to_rgb_stage2
@@ -1377,44 +1395,71 @@
     vrshrn.s32      d25, q13, #14
     vrshrn.s32      d28, q14, #14
     vrshrn.s32      d29, q15, #14
-    vaddw.u8        q10, q10, d0
+    vaddw.u8        q11, q10, d0
     vaddw.u8        q12, q12, d0
     vaddw.u8        q14, q14, d0
-    vqmovun.s16     d1\g_offs, q10
+  .if \bpp != 16
+    vqmovun.s16     d1\g_offs, q11
     vqmovun.s16     d1\r_offs, q12
     vqmovun.s16     d1\b_offs, q14
+  .else  /* rgb565 */
+    vqshlu.s16      q13, q11, #8
+    vqshlu.s16      q15, q12, #8
+    vqshlu.s16      q14, q14, #8
+    vsri.u16        q15, q13, #5
+    vsri.u16        q15, q14, #11
+  .endif
 .endm
 
 .macro do_yuv_to_rgb_stage2_store_load_stage1
+                                       /* "do_yuv_to_rgb_stage2" and "store" */
+                                       vrshrn.s32      d20, q10, #15
+    /* "load" and "do_yuv_to_rgb_stage1" */
+    pld             [U, #64]
+                                       vrshrn.s32      d21, q11, #15
+    pld             [V, #64]
+                                       vrshrn.s32      d24, q12, #14
+                                       vrshrn.s32      d25, q13, #14
     vld1.8          {d4}, [U, :64]!
-      vrshrn.s32      d20, q10, #15
-      vrshrn.s32      d21, q11, #15
-      vrshrn.s32      d24, q12, #14
-      vrshrn.s32      d25, q13, #14
-      vrshrn.s32      d28, q14, #14
+                                       vrshrn.s32      d28, q14, #14
     vld1.8          {d5}, [V, :64]!
-      vrshrn.s32      d29, q15, #14
-      vaddw.u8        q10, q10, d0
-      vaddw.u8        q12, q12, d0
-      vaddw.u8        q14, q14, d0
-      vqmovun.s16     d1\g_offs, q10
+                                       vrshrn.s32      d29, q15, #14
+    vaddw.u8        q3, q1, d4      /* q3 = u - 128 */
+    vaddw.u8        q4, q1, d5      /* q2 = v - 128 */
+                                       vaddw.u8        q11, q10, d0
+    vmull.s16       q10, d6, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q10, d8, d1[2]  /* multiply by -23401 */
+                                       vaddw.u8        q12, q12, d0
+                                       vaddw.u8        q14, q14, d0
+  .if \bpp != 16  /**************** rgb24/rgb32 ******************************/
+                                       vqmovun.s16     d1\g_offs, q11
+    pld             [Y, #64]
+                                       vqmovun.s16     d1\r_offs, q12
     vld1.8          {d0}, [Y, :64]!
-      vqmovun.s16     d1\r_offs, q12
-    pld             [U, #64]
-    pld             [V, #64]
+                                       vqmovun.s16     d1\b_offs, q14
+    vmull.s16       q11, d7, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q11, d9, d1[2]  /* multiply by -23401 */
+                                       do_store        \bpp, 8
+    vmull.s16       q12, d8, d1[0]  /* multiply by 22971 */
+    vmull.s16       q13, d9, d1[0]  /* multiply by 22971 */
+    vmull.s16       q14, d6, d1[3]  /* multiply by 29033 */
+    vmull.s16       q15, d7, d1[3]  /* multiply by 29033 */
+  .else  /**************************** rgb565 ********************************/
+                                       vqshlu.s16      q13, q11, #8
     pld             [Y, #64]
-      vqmovun.s16     d1\b_offs, q14
-    vaddw.u8        q3, q1, d4     /* q3 = u - 128 */
-    vaddw.u8        q4, q1, d5     /* q2 = v - 128 */
-      do_store        \bpp, 8
-    vmull.s16       q10, d6, d1[1] /* multiply by -11277 */
-    vmlal.s16       q10, d8, d1[2] /* multiply by -23401 */
-    vmull.s16       q11, d7, d1[1] /* multiply by -11277 */
-    vmlal.s16       q11, d9, d1[2] /* multiply by -23401 */
-    vmull.s16       q12, d8, d1[0] /* multiply by 22971 */
-    vmull.s16       q13, d9, d1[0] /* multiply by 22971 */
-    vmull.s16       q14, d6, d1[3] /* multiply by 29033 */
-    vmull.s16       q15, d7, d1[3] /* multiply by 29033 */
+                                       vqshlu.s16      q15, q12, #8
+                                       vqshlu.s16      q14, q14, #8
+    vld1.8          {d0}, [Y, :64]!
+    vmull.s16       q11, d7, d1[1]
+    vmlal.s16       q11, d9, d1[2]
+                                       vsri.u16        q15, q13, #5
+    vmull.s16       q12, d8, d1[0]
+                                       vsri.u16        q15, q14, #11
+    vmull.s16       q13, d9, d1[0]
+    vmull.s16       q14, d6, d1[3]
+                                       do_store        \bpp, 8
+    vmull.s16       q15, d7, d1[3]
+  .endif
 .endm
 
 .macro do_yuv_to_rgb
@@ -1428,10 +1473,10 @@
 
 .balign 16
 jsimd_ycc_\colorid\()_neon_consts:
-    .short          0,      0,     0,      0
-    .short          22971, -11277, -23401, 29033
-    .short          -128,  -128,   -128,   -128
-    .short          -128,  -128,   -128,   -128
+  .short 0,      0,     0,      0
+  .short 22971, -11277, -23401, 29033
+  .short -128,  -128,   -128,   -128
+  .short -128,  -128,   -128,   -128
 
 asm_function jsimd_ycc_\colorid\()_convert_neon
     OUTPUT_WIDTH    .req r0
@@ -1541,7 +1586,6 @@
     .unreq          U
     .unreq          V
     .unreq          N
-.endfunc
 
 .purgem do_yuv_to_rgb
 .purgem do_yuv_to_rgb_stage1
@@ -1557,10 +1601,12 @@
 generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, 1, 0
 generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, 2, 1
 generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, 2, 3
+generate_jsimd_ycc_rgb_convert_neon rgb565,  16, 0, 0, 0
 
 .purgem do_load
 .purgem do_store
 
+
 /*****************************************************************************/
 
 /*
@@ -1575,123 +1621,123 @@
  */
 
 .macro do_store size
+  .if \size == 8
+    vst1.8          {d20}, [Y]!
+    vst1.8          {d21}, [U]!
+    vst1.8          {d22}, [V]!
+  .elseif \size == 4
+    vst1.8          {d20[0]}, [Y]!
+    vst1.8          {d20[1]}, [Y]!
+    vst1.8          {d20[2]}, [Y]!
+    vst1.8          {d20[3]}, [Y]!
+    vst1.8          {d21[0]}, [U]!
+    vst1.8          {d21[1]}, [U]!
+    vst1.8          {d21[2]}, [U]!
+    vst1.8          {d21[3]}, [U]!
+    vst1.8          {d22[0]}, [V]!
+    vst1.8          {d22[1]}, [V]!
+    vst1.8          {d22[2]}, [V]!
+    vst1.8          {d22[3]}, [V]!
+  .elseif \size == 2
+    vst1.8          {d20[4]}, [Y]!
+    vst1.8          {d20[5]}, [Y]!
+    vst1.8          {d21[4]}, [U]!
+    vst1.8          {d21[5]}, [U]!
+    vst1.8          {d22[4]}, [V]!
+    vst1.8          {d22[5]}, [V]!
+  .elseif \size == 1
+    vst1.8          {d20[6]}, [Y]!
+    vst1.8          {d21[6]}, [U]!
+    vst1.8          {d22[6]}, [V]!
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size
+  .if \bpp == 24
     .if \size == 8
-        vst1.8  {d20}, [Y]!
-        vst1.8  {d21}, [U]!
-        vst1.8  {d22}, [V]!
+      vld3.8        {d10, d11, d12}, [RGB]!
+      pld           [RGB, #128]
     .elseif \size == 4
-        vst1.8  {d20[0]}, [Y]!
-        vst1.8  {d20[1]}, [Y]!
-        vst1.8  {d20[2]}, [Y]!
-        vst1.8  {d20[3]}, [Y]!
-        vst1.8  {d21[0]}, [U]!
-        vst1.8  {d21[1]}, [U]!
-        vst1.8  {d21[2]}, [U]!
-        vst1.8  {d21[3]}, [U]!
-        vst1.8  {d22[0]}, [V]!
-        vst1.8  {d22[1]}, [V]!
-        vst1.8  {d22[2]}, [V]!
-        vst1.8  {d22[3]}, [V]!
+      vld3.8        {d10[0], d11[0], d12[0]}, [RGB]!
+      vld3.8        {d10[1], d11[1], d12[1]}, [RGB]!
+      vld3.8        {d10[2], d11[2], d12[2]}, [RGB]!
+      vld3.8        {d10[3], d11[3], d12[3]}, [RGB]!
     .elseif \size == 2
-        vst1.8  {d20[4]}, [Y]!
-        vst1.8  {d20[5]}, [Y]!
-        vst1.8  {d21[4]}, [U]!
-        vst1.8  {d21[5]}, [U]!
-        vst1.8  {d22[4]}, [V]!
-        vst1.8  {d22[5]}, [V]!
+      vld3.8        {d10[4], d11[4], d12[4]}, [RGB]!
+      vld3.8        {d10[5], d11[5], d12[5]}, [RGB]!
     .elseif \size == 1
-        vst1.8  {d20[6]}, [Y]!
-        vst1.8  {d21[6]}, [U]!
-        vst1.8  {d22[6]}, [V]!
+      vld3.8        {d10[6], d11[6], d12[6]}, [RGB]!
     .else
-        .error unsupported macroblock size
+      .error unsupported macroblock size
     .endif
-.endm
-
-.macro do_load bpp, size
-    .if \bpp == 24
-        .if \size == 8
-            vld3.8  {d10, d11, d12}, [RGB]!
-            pld     [RGB, #128]
-        .elseif \size == 4
-            vld3.8  {d10[0], d11[0], d12[0]}, [RGB]!
-            vld3.8  {d10[1], d11[1], d12[1]}, [RGB]!
-            vld3.8  {d10[2], d11[2], d12[2]}, [RGB]!
-            vld3.8  {d10[3], d11[3], d12[3]}, [RGB]!
-        .elseif \size == 2
-            vld3.8  {d10[4], d11[4], d12[4]}, [RGB]!
-            vld3.8  {d10[5], d11[5], d12[5]}, [RGB]!
-        .elseif \size == 1
-            vld3.8  {d10[6], d11[6], d12[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
-    .elseif \bpp == 32
-        .if \size == 8
-            vld4.8  {d10, d11, d12, d13}, [RGB]!
-            pld     [RGB, #128]
-        .elseif \size == 4
-            vld4.8  {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
-            vld4.8  {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
-            vld4.8  {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
-            vld4.8  {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
-        .elseif \size == 2
-            vld4.8  {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
-            vld4.8  {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
-        .elseif \size == 1
-            vld4.8  {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      vld4.8        {d10, d11, d12, d13}, [RGB]!
+      pld           [RGB, #128]
+    .elseif \size == 4
+      vld4.8        {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
+      vld4.8        {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
+      vld4.8        {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
+      vld4.8        {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
+    .elseif \size == 2
+      vld4.8        {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
+      vld4.8        {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
+    .elseif \size == 1
+      vld4.8        {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
     .else
-        .error unsupported bpp
+      .error unsupported macroblock size
     .endif
+  .else
+    .error unsupported bpp
+  .endif
 .endm
 
 .macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, b_offs
 
 /*
- * 2 stage pipelined RGB->YCbCr conversion
+ * 2-stage pipelined RGB->YCbCr conversion
  */
 
 .macro do_rgb_to_yuv_stage1
-    vmovl.u8    q2, d1\r_offs /* r = { d4, d5 } */
-    vmovl.u8    q3, d1\g_offs /* g = { d6, d7 } */
-    vmovl.u8    q4, d1\b_offs /* b = { d8, d9 } */
-    vmull.u16   q7, d4, d0[0]
-    vmlal.u16   q7, d6, d0[1]
-    vmlal.u16   q7, d8, d0[2]
-    vmull.u16   q8, d5, d0[0]
-    vmlal.u16   q8, d7, d0[1]
-    vmlal.u16   q8, d9, d0[2]
-    vrev64.32   q9,  q1
-    vrev64.32   q13, q1
-    vmlsl.u16   q9,  d4, d0[3]
-    vmlsl.u16   q9,  d6, d1[0]
-    vmlal.u16   q9,  d8, d1[1]
-    vmlsl.u16   q13, d5, d0[3]
-    vmlsl.u16   q13, d7, d1[0]
-    vmlal.u16   q13, d9, d1[1]
-    vrev64.32   q14, q1
-    vrev64.32   q15, q1
-    vmlal.u16   q14, d4, d1[1]
-    vmlsl.u16   q14, d6, d1[2]
-    vmlsl.u16   q14, d8, d1[3]
-    vmlal.u16   q15, d5, d1[1]
-    vmlsl.u16   q15, d7, d1[2]
-    vmlsl.u16   q15, d9, d1[3]
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d0[0]
+    vmlal.u16       q7, d6, d0[1]
+    vmlal.u16       q7, d8, d0[2]
+    vmull.u16       q8, d5, d0[0]
+    vmlal.u16       q8, d7, d0[1]
+    vmlal.u16       q8, d9, d0[2]
+    vrev64.32       q9, q1
+    vrev64.32       q13, q1
+    vmlsl.u16       q9, d4, d0[3]
+    vmlsl.u16       q9, d6, d1[0]
+    vmlal.u16       q9, d8, d1[1]
+    vmlsl.u16       q13, d5, d0[3]
+    vmlsl.u16       q13, d7, d1[0]
+    vmlal.u16       q13, d9, d1[1]
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d1[1]
+    vmlsl.u16       q14, d6, d1[2]
+    vmlsl.u16       q14, d8, d1[3]
+    vmlal.u16       q15, d5, d1[1]
+    vmlsl.u16       q15, d7, d1[2]
+    vmlsl.u16       q15, d9, d1[3]
 .endm
 
 .macro do_rgb_to_yuv_stage2
-    vrshrn.u32  d20, q7,  #16
-    vrshrn.u32  d21, q8,  #16
-    vshrn.u32   d22, q9,  #16
-    vshrn.u32   d23, q13, #16
-    vshrn.u32   d24, q14, #16
-    vshrn.u32   d25, q15, #16
-    vmovn.u16   d20, q10      /* d20 = y */
-    vmovn.u16   d21, q11      /* d21 = u */
-    vmovn.u16   d22, q12      /* d22 = v */
+    vrshrn.u32      d20, q7, #16
+    vrshrn.u32      d21, q8, #16
+    vshrn.u32       d22, q9, #16
+    vshrn.u32       d23, q13, #16
+    vshrn.u32       d24, q14, #16
+    vshrn.u32       d25, q15, #16
+    vmovn.u16       d20, q10       /* d20 = y */
+    vmovn.u16       d21, q11       /* d21 = u */
+    vmovn.u16       d22, q12       /* d22 = v */
 .endm
 
 .macro do_rgb_to_yuv
@@ -1700,52 +1746,52 @@
 .endm
 
 .macro do_rgb_to_yuv_stage2_store_load_stage1
-      vrshrn.u32  d20, q7,  #16
-      vrshrn.u32  d21, q8,  #16
-      vshrn.u32   d22, q9,  #16
-    vrev64.32   q9,  q1
-      vshrn.u32   d23, q13, #16
-    vrev64.32   q13, q1
-      vshrn.u32   d24, q14, #16
-      vshrn.u32   d25, q15, #16
-    do_load     \bpp, 8
-      vmovn.u16   d20, q10      /* d20 = y */
-    vmovl.u8    q2, d1\r_offs   /* r = { d4, d5 } */
-      vmovn.u16   d21, q11      /* d21 = u */
-    vmovl.u8    q3, d1\g_offs   /* g = { d6, d7 } */
-      vmovn.u16   d22, q12      /* d22 = v */
-    vmovl.u8    q4, d1\b_offs   /* b = { d8, d9 } */
-    vmull.u16   q7, d4, d0[0]
-    vmlal.u16   q7, d6, d0[1]
-    vmlal.u16   q7, d8, d0[2]
-      vst1.8      {d20}, [Y]!
-    vmull.u16   q8, d5, d0[0]
-    vmlal.u16   q8, d7, d0[1]
-    vmlal.u16   q8, d9, d0[2]
-    vmlsl.u16   q9,  d4, d0[3]
-    vmlsl.u16   q9,  d6, d1[0]
-    vmlal.u16   q9,  d8, d1[1]
-      vst1.8      {d21}, [U]!
-    vmlsl.u16   q13, d5, d0[3]
-    vmlsl.u16   q13, d7, d1[0]
-    vmlal.u16   q13, d9, d1[1]
-    vrev64.32   q14, q1
-    vrev64.32   q15, q1
-    vmlal.u16   q14, d4, d1[1]
-    vmlsl.u16   q14, d6, d1[2]
-    vmlsl.u16   q14, d8, d1[3]
-      vst1.8      {d22}, [V]!
-    vmlal.u16   q15, d5, d1[1]
-    vmlsl.u16   q15, d7, d1[2]
-    vmlsl.u16   q15, d9, d1[3]
+      vrshrn.u32      d20, q7, #16
+      vrshrn.u32      d21, q8, #16
+      vshrn.u32       d22, q9, #16
+    vrev64.32       q9, q1
+      vshrn.u32       d23, q13, #16
+    vrev64.32       q13, q1
+      vshrn.u32       d24, q14, #16
+      vshrn.u32       d25, q15, #16
+    do_load         \bpp, 8
+      vmovn.u16       d20, q10     /* d20 = y */
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+      vmovn.u16       d21, q11     /* d21 = u */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+      vmovn.u16       d22, q12     /* d22 = v */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d0[0]
+    vmlal.u16       q7, d6, d0[1]
+    vmlal.u16       q7, d8, d0[2]
+      vst1.8          {d20}, [Y]!
+    vmull.u16       q8, d5, d0[0]
+    vmlal.u16       q8, d7, d0[1]
+    vmlal.u16       q8, d9, d0[2]
+    vmlsl.u16       q9, d4, d0[3]
+    vmlsl.u16       q9, d6, d1[0]
+    vmlal.u16       q9, d8, d1[1]
+      vst1.8          {d21}, [U]!
+    vmlsl.u16       q13, d5, d0[3]
+    vmlsl.u16       q13, d7, d1[0]
+    vmlal.u16       q13, d9, d1[1]
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d1[1]
+    vmlsl.u16       q14, d6, d1[2]
+    vmlsl.u16       q14, d8, d1[3]
+      vst1.8          {d22}, [V]!
+    vmlal.u16       q15, d5, d1[1]
+    vmlsl.u16       q15, d7, d1[2]
+    vmlsl.u16       q15, d9, d1[3]
 .endm
 
 .balign 16
 jsimd_\colorid\()_ycc_neon_consts:
-    .short          19595, 38470, 7471,  11059
-    .short          21709, 32768, 27439, 5329
-    .short          32767, 128,   32767, 128
-    .short          32767, 128,   32767, 128
+  .short 19595, 38470, 7471,  11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128,   32767, 128
+  .short 32767, 128,   32767, 128
 
 asm_function jsimd_\colorid\()_ycc_convert_neon
     OUTPUT_WIDTH    .req r0
@@ -1851,7 +1897,6 @@
     .unreq          U
     .unreq          V
     .unreq          N
-.endfunc
 
 .purgem do_rgb_to_yuv
 .purgem do_rgb_to_yuv_stage1
@@ -1871,6 +1916,7 @@
 .purgem do_load
 .purgem do_store
 
+
 /*****************************************************************************/
 
 /*
@@ -1932,7 +1978,7 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -1955,10 +2001,10 @@
 
 .balign 16
 jsimd_fdct_ifast_neon_consts:
-    .short (98 * 128)              /* XFIX_0_382683433 */
-    .short (139 * 128)             /* XFIX_0_541196100 */
-    .short (181 * 128)             /* XFIX_0_707106781 */
-    .short (334 * 128 - 256 * 128) /* XFIX_1_306562965 */
+  .short (98 * 128)               /* XFIX_0_382683433 */
+  .short (139 * 128)              /* XFIX_0_541196100 */
+  .short (181 * 128)              /* XFIX_0_707106781 */
+  .short (334 * 128 - 256 * 128)  /* XFIX_1_306562965 */
 
 asm_function jsimd_fdct_ifast_neon
 
@@ -1995,52 +2041,52 @@
     /* Transpose */
     vtrn.16         q12, q13
     vtrn.16         q10, q11
-    vtrn.16         q8,  q9
+    vtrn.16         q8, q9
     vtrn.16         q14, q15
-    vtrn.32         q9,  q11
+    vtrn.32         q9, q11
     vtrn.32         q13, q15
-    vtrn.32         q8,  q10
+    vtrn.32         q8, q10
     vtrn.32         q12, q14
     vswp            d30, d23
     vswp            d24, d17
     vswp            d26, d19
       /* 1-D FDCT */
-      vadd.s16        q2,  q11, q12
+      vadd.s16        q2, q11, q12
     vswp            d28, d21
       vsub.s16        q12, q11, q12
-      vsub.s16        q6,  q10, q13
+      vsub.s16        q6, q10, q13
       vadd.s16        q10, q10, q13
-      vsub.s16        q7,  q9,  q14
-      vadd.s16        q9,  q9,  q14
-      vsub.s16        q1,  q8,  q15
-      vadd.s16        q8,  q8,  q15
-      vsub.s16        q4,  q9,  q10
-      vsub.s16        q5,  q8,  q2
-      vadd.s16        q3,  q9,  q10
-      vadd.s16        q4,  q4,  q5
-      vadd.s16        q2,  q8,  q2
-      vqdmulh.s16     q4,  q4,  XFIX_0_707106781
+      vsub.s16        q7, q9, q14
+      vadd.s16        q9, q9, q14
+      vsub.s16        q1, q8, q15
+      vadd.s16        q8, q8, q15
+      vsub.s16        q4, q9, q10
+      vsub.s16        q5, q8, q2
+      vadd.s16        q3, q9, q10
+      vadd.s16        q4, q4, q5
+      vadd.s16        q2, q8, q2
+      vqdmulh.s16     q4, q4, XFIX_0_707106781
       vadd.s16        q11, q12, q6
-      vadd.s16        q8,  q2,  q3
-      vsub.s16        q12, q2,  q3
-      vadd.s16        q3,  q6,  q7
-      vadd.s16        q7,  q7,  q1
-      vqdmulh.s16     q3,  q3,  XFIX_0_707106781
-      vsub.s16        q6,  q11, q7
-      vadd.s16        q10, q5,  q4
-      vqdmulh.s16     q6,  q6,  XFIX_0_382683433
-      vsub.s16        q14, q5,  q4
+      vadd.s16        q8, q2, q3
+      vsub.s16        q12, q2, q3
+      vadd.s16        q3, q6, q7
+      vadd.s16        q7, q7, q1
+      vqdmulh.s16     q3, q3, XFIX_0_707106781
+      vsub.s16        q6, q11, q7
+      vadd.s16        q10, q5, q4
+      vqdmulh.s16     q6, q6, XFIX_0_382683433
+      vsub.s16        q14, q5, q4
       vqdmulh.s16     q11, q11, XFIX_0_541196100
-      vqdmulh.s16     q5,  q7,  XFIX_1_306562965
-      vadd.s16        q4,  q1,  q3
-      vsub.s16        q3,  q1,  q3
-      vadd.s16        q7,  q7,  q6
+      vqdmulh.s16     q5, q7, XFIX_1_306562965
+      vadd.s16        q4, q1, q3
+      vsub.s16        q3, q1, q3
+      vadd.s16        q7, q7, q6
       vadd.s16        q11, q11, q6
-      vadd.s16        q7,  q7,  q5
-      vadd.s16        q13, q3,  q11
-      vsub.s16        q11, q3,  q11
-      vadd.s16        q9,  q4,  q7
-      vsub.s16        q15, q4,  q7
+      vadd.s16        q7, q7, q5
+      vadd.s16        q13, q3, q11
+      vsub.s16        q11, q3, q11
+      vadd.s16        q9, q4, q7
+      vsub.s16        q15, q4, q7
     subs            TMP, TMP, #1
     bne             1b
 
@@ -2055,14 +2101,14 @@
 
     .unreq          DATA
     .unreq          TMP
-.endfunc
+
 
 /*****************************************************************************/
 
 /*
  * GLOBAL(void)
- * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM * divisors,
- *                      DCTELEM * workspace);
+ * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM *divisors,
+ *                      DCTELEM *workspace);
  *
  * Note: the code uses 2 stage pipelining in order to improve instructions
  *       scheduling and eliminate stalls (this provides ~15% better
@@ -2089,22 +2135,22 @@
     vld1.16         {d20, d21, d22, d23}, [CORRECTION, :128]!
     vabs.s16        q13, q1
     vld1.16         {d16, d17, d18, d19}, [RECIPROCAL, :128]!
-    vadd.u16        q12, q12, q10 /* add correction */
+    vadd.u16        q12, q12, q10  /* add correction */
     vadd.u16        q13, q13, q11
-    vmull.u16       q10, d24, d16 /* multiply by reciprocal */
+    vmull.u16       q10, d24, d16  /* multiply by reciprocal */
     vmull.u16       q11, d25, d17
-    vmull.u16       q8,  d26, d18
-    vmull.u16       q9,  d27, d19
+    vmull.u16       q8, d26, d18
+    vmull.u16       q9, d27, d19
     vld1.16         {d24, d25, d26, d27}, [SHIFT, :128]!
     vshrn.u32       d20, q10, #16
     vshrn.u32       d21, q11, #16
-    vshrn.u32       d22, q8,  #16
-    vshrn.u32       d23, q9,  #16
+    vshrn.u32       d22, q8, #16
+    vshrn.u32       d23, q9, #16
     vneg.s16        q12, q12
     vneg.s16        q13, q13
-    vshr.s16        q2,  q0,  #15 /* extract sign */
-    vshr.s16        q3,  q1,  #15
-    vshl.u16        q14, q10, q12 /* shift */
+    vshr.s16        q2, q0, #15    /* extract sign */
+    vshr.s16        q3, q1, #15
+    vshl.u16        q14, q10, q12  /* shift */
     vshl.u16        q15, q11, q13
 
     push            {r4, r5}
@@ -2117,25 +2163,25 @@
     vabs.s16        q13, q1
       veor.u16        q15, q15, q3
     vld1.16         {d16, d17, d18, d19}, [RECIPROCAL, :128]!
-    vadd.u16        q12, q12, q10 /* add correction */
+    vadd.u16        q12, q12, q10  /* add correction */
     vadd.u16        q13, q13, q11
-    vmull.u16       q10, d24, d16 /* multiply by reciprocal */
+    vmull.u16       q10, d24, d16  /* multiply by reciprocal */
     vmull.u16       q11, d25, d17
-    vmull.u16       q8,  d26, d18
-    vmull.u16       q9,  d27, d19
+    vmull.u16       q8, d26, d18
+    vmull.u16       q9, d27, d19
       vsub.u16        q14, q14, q2
     vld1.16         {d24, d25, d26, d27}, [SHIFT, :128]!
       vsub.u16        q15, q15, q3
     vshrn.u32       d20, q10, #16
     vshrn.u32       d21, q11, #16
       vst1.16         {d28, d29, d30, d31}, [COEF_BLOCK, :128]!
-    vshrn.u32       d22, q8,  #16
-    vshrn.u32       d23, q9,  #16
+    vshrn.u32       d22, q8, #16
+    vshrn.u32       d23, q9, #16
     vneg.s16        q12, q12
     vneg.s16        q13, q13
-    vshr.s16        q2,  q0,  #15 /* extract sign */
-    vshr.s16        q3,  q1,  #15
-    vshl.u16        q14, q10, q12 /* shift */
+    vshr.s16        q2, q0, #15    /* extract sign */
+    vshr.s16        q3, q1, #15
+    vshl.u16        q14, q10, q12  /* shift */
     vshl.u16        q15, q11, q13
     subs            LOOP_COUNT, LOOP_COUNT, #1
     bne             1b
@@ -2147,7 +2193,7 @@
       vsub.u16        q15, q15, q3
       vst1.16         {d28, d29, d30, d31}, [COEF_BLOCK, :128]!
 
-    bx              lr /* return */
+    bx              lr  /* return */
 
     .unreq          COEF_BLOCK
     .unreq          DIVISORS
@@ -2156,4 +2202,677 @@
     .unreq          CORRECTION
     .unreq          SHIFT
     .unreq          LOOP_COUNT
-.endfunc
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(void)
+ * jsimd_h2v1_fancy_upsample_neon (int max_v_samp_factor,
+ *                                 JDIMENSION downsampled_width,
+ *                                 JSAMPARRAY input_data,
+ *                                 JSAMPARRAY *output_data_ptr);
+ *
+ * Note: the use of unaligned writes is the main remaining bottleneck in
+ *       this code, which can be potentially solved to get up to tens
+ *       of percents performance improvement on Cortex-A8/Cortex-A9.
+ */
+
+/*
+ * Upsample 16 source pixels to 32 destination pixels. The new 16 source
+ * pixels are loaded to q0. The previous 16 source pixels are in q1. The
+ * shifted-by-one source pixels are constructed in q2 by using q0 and q1.
+ * Register d28 is used for multiplication by 3. Register q15 is used
+ * for adding +1 bias.
+ */
+.macro upsample16 OUTPTR, INPTR
+    vld1.8          {q0}, [\INPTR]!
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+    vmov            q1, q0        /* backup source pixels to q1 */
+    vrshrn.u16      d6, q8, #2
+    vrshrn.u16      d7, q9, #2
+    vshrn.u16       d8, q10, #2
+    vshrn.u16       d9, q11, #2
+    vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+.endm
+
+/*
+ * Upsample 32 source pixels to 64 destination pixels. Compared to 'usample16'
+ * macro, the roles of q0 and q1 registers are reversed for even and odd
+ * groups of 16 pixels, that's why "vmov q1, q0" instructions are not needed.
+ * Also this unrolling allows to reorder loads and stores to compensate
+ * multiplication latency and reduce stalls.
+ */
+.macro upsample32 OUTPTR, INPTR
+    /* even 16 pixels group */
+    vld1.8          {q0}, [\INPTR]!
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+      /* odd 16 pixels group */
+      vld1.8          {q1}, [\INPTR]!
+    vrshrn.u16      d6, q8, #2
+    vrshrn.u16      d7, q9, #2
+    vshrn.u16       d8, q10, #2
+    vshrn.u16       d9, q11, #2
+      vmovl.u8        q8, d2
+      vext.8          q2, q0, q1, #15
+      vmovl.u8        q9, d3
+      vaddw.u8        q10, q15, d4
+      vaddw.u8        q11, q15, d5
+      vmlal.u8        q8, d4, d28
+      vmlal.u8        q9, d5, d28
+      vmlal.u8        q10, d2, d28
+      vmlal.u8        q11, d3, d28
+    vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+      vrshrn.u16      d6, q8, #2
+      vrshrn.u16      d7, q9, #2
+      vshrn.u16       d8, q10, #2
+      vshrn.u16       d9, q11, #2
+      vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+.endm
+
+/*
+ * Upsample a row of WIDTH pixels from INPTR to OUTPTR.
+ */
+.macro upsample_row OUTPTR, INPTR, WIDTH, TMP1
+    /* special case for the first and last pixels */
+    sub             \WIDTH, \WIDTH, #1
+    add             \OUTPTR, \OUTPTR, #1
+    ldrb            \TMP1, [\INPTR, \WIDTH]
+    strb            \TMP1, [\OUTPTR, \WIDTH, asl #1]
+    ldrb            \TMP1, [\INPTR], #1
+    strb            \TMP1, [\OUTPTR, #-1]
+    vmov.8          d3[7], \TMP1
+
+    subs            \WIDTH, \WIDTH, #32
+    blt             5f
+0:  /* process 32 pixels per iteration */
+    upsample32      \OUTPTR, \INPTR
+    subs            \WIDTH, \WIDTH, #32
+    bge             0b
+5:
+    adds            \WIDTH, \WIDTH, #16
+    blt             1f
+0:  /* process 16 pixels if needed */
+    upsample16      \OUTPTR, \INPTR
+    subs            \WIDTH, \WIDTH, #16
+1:
+    adds            \WIDTH, \WIDTH, #16
+    beq             9f
+
+    /* load the remaining 1-15 pixels */
+    add             \INPTR, \INPTR, \WIDTH
+    tst             \WIDTH, #1
+    beq             2f
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #2
+    beq             2f
+    vext.8          d0, d0, d0, #6
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[1]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #4
+    beq             2f
+    vrev64.32       d0, d0
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[3]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[2]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[1]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #8
+    beq             2f
+    vmov            d1, d0
+    sub             \INPTR, \INPTR, #8
+    vld1.8          {d0}, [\INPTR]
+2:  /* upsample the remaining pixels */
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+    vrshrn.u16      d10, q8, #2
+    vrshrn.u16      d12, q9, #2
+    vshrn.u16       d11, q10, #2
+    vshrn.u16       d13, q11, #2
+    vzip.8          d10, d11
+    vzip.8          d12, d13
+    /* store the remaining pixels */
+    tst             \WIDTH, #8
+    beq             2f
+    vst1.8          {d10, d11}, [\OUTPTR]!
+    vmov            q5, q6
+2:
+    tst             \WIDTH, #4
+    beq             2f
+    vst1.8          {d10}, [\OUTPTR]!
+    vmov            d10, d11
+2:
+    tst             \WIDTH, #2
+    beq             2f
+    vst1.8          {d10[0]}, [\OUTPTR]!
+    vst1.8          {d10[1]}, [\OUTPTR]!
+    vst1.8          {d10[2]}, [\OUTPTR]!
+    vst1.8          {d10[3]}, [\OUTPTR]!
+    vext.8          d10, d10, d10, #4
+2:
+    tst             \WIDTH, #1
+    beq             2f
+    vst1.8          {d10[0]}, [\OUTPTR]!
+    vst1.8          {d10[1]}, [\OUTPTR]!
+2:
+9:
+.endm
+
+asm_function jsimd_h2v1_fancy_upsample_neon
+
+    MAX_V_SAMP_FACTOR .req r0
+    DOWNSAMPLED_WIDTH .req r1
+    INPUT_DATA        .req r2
+    OUTPUT_DATA_PTR   .req r3
+    OUTPUT_DATA       .req OUTPUT_DATA_PTR
+
+    OUTPTR            .req r4
+    INPTR             .req r5
+    WIDTH             .req ip
+    TMP               .req lr
+
+    push            {r4, r5, r6, lr}
+    vpush           {d8-d15}
+
+    ldr             OUTPUT_DATA, [OUTPUT_DATA_PTR]
+    cmp             MAX_V_SAMP_FACTOR, #0
+    ble             99f
+
+    /* initialize constants */
+    vmov.u8         d28, #3
+    vmov.u16        q15, #1
+11:
+    ldr             INPTR, [INPUT_DATA], #4
+    ldr             OUTPTR, [OUTPUT_DATA], #4
+    mov             WIDTH, DOWNSAMPLED_WIDTH
+    upsample_row    OUTPTR, INPTR, WIDTH, TMP
+    subs            MAX_V_SAMP_FACTOR, MAX_V_SAMP_FACTOR, #1
+    bgt             11b
+
+99:
+    vpop            {d8-d15}
+    pop             {r4, r5, r6, pc}
+
+    .unreq          MAX_V_SAMP_FACTOR
+    .unreq          DOWNSAMPLED_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA_PTR
+    .unreq          OUTPUT_DATA
+
+    .unreq          OUTPTR
+    .unreq          INPTR
+    .unreq          WIDTH
+    .unreq          TMP
+
+.purgem upsample16
+.purgem upsample32
+.purgem upsample_row
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(JOCTET*)
+ * jsimd_huff_encode_one_block (working_state *state, JOCTET *buffer,
+ *                              JCOEFPTR block, int last_dc_val,
+ *                              c_derived_tbl *dctbl, c_derived_tbl *actbl)
+ *
+ */
+
+.macro emit_byte BUFFER, PUT_BUFFER, PUT_BITS, ZERO, TMP
+    sub             \PUT_BITS, \PUT_BITS, #0x8
+    lsr             \TMP, \PUT_BUFFER, \PUT_BITS
+    uxtb            \TMP, \TMP
+    strb            \TMP, [\BUFFER, #1]!
+    cmp             \TMP, #0xff
+    /*it eq*/
+    strbeq          \ZERO, [\BUFFER, #1]!
+.endm
+
+.macro put_bits PUT_BUFFER, PUT_BITS, CODE, SIZE
+    /*lsl             \PUT_BUFFER, \PUT_BUFFER, \SIZE*/
+    add             \PUT_BITS, \SIZE
+    /*orr             \PUT_BUFFER, \PUT_BUFFER, \CODE*/
+    orr             \PUT_BUFFER, \CODE, \PUT_BUFFER, lsl \SIZE
+.endm
+
+.macro checkbuf15 BUFFER, PUT_BUFFER, PUT_BITS, ZERO, TMP
+  cmp               \PUT_BITS, #0x10
+  blt               15f
+    eor               \ZERO, \ZERO, \ZERO
+    emit_byte         \BUFFER, \PUT_BUFFER, \PUT_BITS, \ZERO, \TMP
+    emit_byte         \BUFFER, \PUT_BUFFER, \PUT_BITS, \ZERO, \TMP
+15:
+.endm
+
+.balign 16
+jsimd_huff_encode_one_block_neon_consts:
+  .byte 0x01
+  .byte 0x02
+  .byte 0x04
+  .byte 0x08
+  .byte 0x10
+  .byte 0x20
+  .byte 0x40
+  .byte 0x80
+
+asm_function jsimd_huff_encode_one_block_neon
+    push            {r4, r5, r6, r7, r8, r9, r10, r11, lr}
+    add             r7, sp, #0x1c
+    sub             r4, sp, #0x40
+    bfc             r4, #0, #5
+    mov             sp, r4           /* align sp on 32 bytes */
+    vst1.64         {d8, d9, d10, d11}, [r4, :128]!
+    vst1.64         {d12, d13, d14, d15}, [r4, :128]
+    sub             sp, #0x140       /* reserve 320 bytes */
+    str             r0, [sp, #0x18]  /* working state > sp + Ox18 */
+    add             r4, sp, #0x20    /* r4 = t1 */
+    ldr             lr, [r7, #0x8]   /* lr = dctbl */
+    sub             r10, r1, #0x1    /* r10=buffer-- */
+    ldrsh           r1, [r2]
+    mov             r9, #0x10
+    mov             r8, #0x1
+    adr             r5, jsimd_huff_encode_one_block_neon_consts
+    /* prepare data */
+    vld1.8          {d26}, [r5, :64]
+    veor            q8, q8, q8
+    veor            q9, q9, q9
+    vdup.16         q14, r9
+    vdup.16         q15, r8
+    veor            q10, q10, q10
+    veor            q11, q11, q11
+    sub             r1, r1, r3
+    add             r9, r2, #0x22
+    add             r8, r2, #0x18
+    add             r3, r2, #0x36
+    vmov.16         d0[0], r1
+    vld1.16         {d2[0]}, [r9, :16]
+    vld1.16         {d4[0]}, [r8, :16]
+    vld1.16         {d6[0]}, [r3, :16]
+    add             r1, r2, #0x2
+    add             r9, r2, #0x30
+    add             r8, r2, #0x26
+    add             r3, r2, #0x28
+    vld1.16         {d0[1]}, [r1, :16]
+    vld1.16         {d2[1]}, [r9, :16]
+    vld1.16         {d4[1]}, [r8, :16]
+    vld1.16         {d6[1]}, [r3, :16]
+    add             r1, r2, #0x10
+    add             r9, r2, #0x40
+    add             r8, r2, #0x34
+    add             r3, r2, #0x1a
+    vld1.16         {d0[2]}, [r1, :16]
+    vld1.16         {d2[2]}, [r9, :16]
+    vld1.16         {d4[2]}, [r8, :16]
+    vld1.16         {d6[2]}, [r3, :16]
+    add             r1, r2, #0x20
+    add             r9, r2, #0x32
+    add             r8, r2, #0x42
+    add             r3, r2, #0xc
+    vld1.16         {d0[3]}, [r1, :16]
+    vld1.16         {d2[3]}, [r9, :16]
+    vld1.16         {d4[3]}, [r8, :16]
+    vld1.16         {d6[3]}, [r3, :16]
+    add             r1, r2, #0x12
+    add             r9, r2, #0x24
+    add             r8, r2, #0x50
+    add             r3, r2, #0xe
+    vld1.16         {d1[0]}, [r1, :16]
+    vld1.16         {d3[0]}, [r9, :16]
+    vld1.16         {d5[0]}, [r8, :16]
+    vld1.16         {d7[0]}, [r3, :16]
+    add             r1, r2, #0x4
+    add             r9, r2, #0x16
+    add             r8, r2, #0x60
+    add             r3, r2, #0x1c
+    vld1.16         {d1[1]}, [r1, :16]
+    vld1.16         {d3[1]}, [r9, :16]
+    vld1.16         {d5[1]}, [r8, :16]
+    vld1.16         {d7[1]}, [r3, :16]
+    add             r1, r2, #0x6
+    add             r9, r2, #0x8
+    add             r8, r2, #0x52
+    add             r3, r2, #0x2a
+    vld1.16         {d1[2]}, [r1, :16]
+    vld1.16         {d3[2]}, [r9, :16]
+    vld1.16         {d5[2]}, [r8, :16]
+    vld1.16         {d7[2]}, [r3, :16]
+    add             r1, r2, #0x14
+    add             r9, r2, #0xa
+    add             r8, r2, #0x44
+    add             r3, r2, #0x38
+    vld1.16         {d1[3]}, [r1, :16]
+    vld1.16         {d3[3]}, [r9, :16]
+    vld1.16         {d5[3]}, [r8, :16]
+    vld1.16         {d7[3]}, [r3, :16]
+    vcgt.s16        q8, q8, q0
+    vcgt.s16        q9, q9, q1
+    vcgt.s16        q10, q10, q2
+    vcgt.s16        q11, q11, q3
+    vabs.s16        q0, q0
+    vabs.s16        q1, q1
+    vabs.s16        q2, q2
+    vabs.s16        q3, q3
+    veor            q8, q8, q0
+    veor            q9, q9, q1
+    veor            q10, q10, q2
+    veor            q11, q11, q3
+    add             r9, r4, #0x20
+    add             r8, r4, #0x80
+    add             r3, r4, #0xa0
+    vclz.i16        q0, q0
+    vclz.i16        q1, q1
+    vclz.i16        q2, q2
+    vclz.i16        q3, q3
+    vsub.i16        q0, q14, q0
+    vsub.i16        q1, q14, q1
+    vsub.i16        q2, q14, q2
+    vsub.i16        q3, q14, q3
+    vst1.16         {d0, d1, d2, d3}, [r4, :256]
+    vst1.16         {d4, d5, d6, d7}, [r9, :256]
+    vshl.s16        q0, q15, q0
+    vshl.s16        q1, q15, q1
+    vshl.s16        q2, q15, q2
+    vshl.s16        q3, q15, q3
+    vsub.i16        q0, q0, q15
+    vsub.i16        q1, q1, q15
+    vsub.i16        q2, q2, q15
+    vsub.i16        q3, q3, q15
+    vand            q8, q8, q0
+    vand            q9, q9, q1
+    vand            q10, q10, q2
+    vand            q11, q11, q3
+    vst1.16         {d16, d17, d18, d19}, [r8, :256]
+    vst1.16         {d20, d21, d22, d23}, [r3, :256]
+    add             r1, r2, #0x46
+    add             r9, r2, #0x3a
+    add             r8, r2, #0x74
+    add             r3, r2, #0x6a
+    vld1.16         {d8[0]}, [r1, :16]
+    vld1.16         {d10[0]}, [r9, :16]
+    vld1.16         {d12[0]}, [r8, :16]
+    vld1.16         {d14[0]}, [r3, :16]
+    veor            q8, q8, q8
+    veor            q9, q9, q9
+    veor            q10, q10, q10
+    veor            q11, q11, q11
+    add             r1, r2, #0x54
+    add             r9, r2, #0x2c
+    add             r8, r2, #0x76
+    add             r3, r2, #0x78
+    vld1.16         {d8[1]}, [r1, :16]
+    vld1.16         {d10[1]}, [r9, :16]
+    vld1.16         {d12[1]}, [r8, :16]
+    vld1.16         {d14[1]}, [r3, :16]
+    add             r1, r2, #0x62
+    add             r9, r2, #0x1e
+    add             r8, r2, #0x68
+    add             r3, r2, #0x7a
+    vld1.16         {d8[2]}, [r1, :16]
+    vld1.16         {d10[2]}, [r9, :16]
+    vld1.16         {d12[2]}, [r8, :16]
+    vld1.16         {d14[2]}, [r3, :16]
+    add             r1, r2, #0x70
+    add             r9, r2, #0x2e
+    add             r8, r2, #0x5a
+    add             r3, r2, #0x6c
+    vld1.16         {d8[3]}, [r1, :16]
+    vld1.16         {d10[3]}, [r9, :16]
+    vld1.16         {d12[3]}, [r8, :16]
+    vld1.16         {d14[3]}, [r3, :16]
+    add             r1, r2, #0x72
+    add             r9, r2, #0x3c
+    add             r8, r2, #0x4c
+    add             r3, r2, #0x5e
+    vld1.16         {d9[0]}, [r1, :16]
+    vld1.16         {d11[0]}, [r9, :16]
+    vld1.16         {d13[0]}, [r8, :16]
+    vld1.16         {d15[0]}, [r3, :16]
+    add             r1, r2, #0x64
+    add             r9, r2, #0x4a
+    add             r8, r2, #0x3e
+    add             r3, r2, #0x6e
+    vld1.16         {d9[1]}, [r1, :16]
+    vld1.16         {d11[1]}, [r9, :16]
+    vld1.16         {d13[1]}, [r8, :16]
+    vld1.16         {d15[1]}, [r3, :16]
+    add             r1, r2, #0x56
+    add             r9, r2, #0x58
+    add             r8, r2, #0x4e
+    add             r3, r2, #0x7c
+    vld1.16         {d9[2]}, [r1, :16]
+    vld1.16         {d11[2]}, [r9, :16]
+    vld1.16         {d13[2]}, [r8, :16]
+    vld1.16         {d15[2]}, [r3, :16]
+    add             r1, r2, #0x48
+    add             r9, r2, #0x66
+    add             r8, r2, #0x5c
+    add             r3, r2, #0x7e
+    vld1.16         {d9[3]}, [r1, :16]
+    vld1.16         {d11[3]}, [r9, :16]
+    vld1.16         {d13[3]}, [r8, :16]
+    vld1.16         {d15[3]}, [r3, :16]
+    vcgt.s16        q8, q8, q4
+    vcgt.s16        q9, q9, q5
+    vcgt.s16        q10, q10, q6
+    vcgt.s16        q11, q11, q7
+    vabs.s16        q4, q4
+    vabs.s16        q5, q5
+    vabs.s16        q6, q6
+    vabs.s16        q7, q7
+    veor            q8, q8, q4
+    veor            q9, q9, q5
+    veor            q10, q10, q6
+    veor            q11, q11, q7
+    add             r1, r4, #0x40
+    add             r9, r4, #0x60
+    add             r8, r4, #0xc0
+    add             r3, r4, #0xe0
+    vclz.i16        q4, q4
+    vclz.i16        q5, q5
+    vclz.i16        q6, q6
+    vclz.i16        q7, q7
+    vsub.i16        q4, q14, q4
+    vsub.i16        q5, q14, q5
+    vsub.i16        q6, q14, q6
+    vsub.i16        q7, q14, q7
+    vst1.16         {d8, d9, d10, d11}, [r1, :256]
+    vst1.16         {d12, d13, d14, d15}, [r9, :256]
+    vshl.s16        q4, q15, q4
+    vshl.s16        q5, q15, q5
+    vshl.s16        q6, q15, q6
+    vshl.s16        q7, q15, q7
+    vsub.i16        q4, q4, q15
+    vsub.i16        q5, q5, q15
+    vsub.i16        q6, q6, q15
+    vsub.i16        q7, q7, q15
+    vand            q8, q8, q4
+    vand            q9, q9, q5
+    vand            q10, q10, q6
+    vand            q11, q11, q7
+    vst1.16         {d16, d17, d18, d19}, [r8, :256]
+    vst1.16         {d20, d21, d22, d23}, [r3, :256]
+    ldr             r12, [r7, #0xc]       /* r12 = actbl */
+    add             r1, lr, #0x400        /* r1 = dctbl->ehufsi */
+    mov             r9, r12               /* r9 = actbl */
+    add             r6, r4, #0x80         /* r6 = t2 */
+    ldr             r11, [r0, #0x8]       /* r11 = put_buffer */
+    ldr             r4, [r0, #0xc]        /* r4  = put_bits */
+    ldrh            r2, [r6, #-128]       /* r2  = nbits */
+    ldrh            r3, [r6]              /* r3  = temp2 & (((JLONG) 1)<<nbits) - 1; */
+    ldr             r0, [lr, r2, lsl #2]
+    ldrb            r5, [r1, r2]
+    put_bits        r11, r4, r0, r5
+    checkbuf15      r10, r11, r4, r5, r0
+    put_bits        r11, r4, r3, r2
+    checkbuf15      r10, r11, r4, r5, r0
+    mov             lr, r6                /* lr = t2 */
+    add             r5, r9, #0x400        /* r5 = actbl->ehufsi */
+    ldrsb           r6, [r5, #0xf0]       /* r6 = actbl->ehufsi[0xf0] */
+    veor            q8, q8, q8
+    vceq.i16        q0, q0, q8
+    vceq.i16        q1, q1, q8
+    vceq.i16        q2, q2, q8
+    vceq.i16        q3, q3, q8
+    vceq.i16        q4, q4, q8
+    vceq.i16        q5, q5, q8
+    vceq.i16        q6, q6, q8
+    vceq.i16        q7, q7, q8
+    vmovn.i16       d0, q0
+    vmovn.i16       d2, q1
+    vmovn.i16       d4, q2
+    vmovn.i16       d6, q3
+    vmovn.i16       d8, q4
+    vmovn.i16       d10, q5
+    vmovn.i16       d12, q6
+    vmovn.i16       d14, q7
+    vand            d0, d0, d26
+    vand            d2, d2, d26
+    vand            d4, d4, d26
+    vand            d6, d6, d26
+    vand            d8, d8, d26
+    vand            d10, d10, d26
+    vand            d12, d12, d26
+    vand            d14, d14, d26
+    vpadd.i8        d0, d0, d2
+    vpadd.i8        d4, d4, d6
+    vpadd.i8        d8, d8, d10
+    vpadd.i8        d12, d12, d14
+    vpadd.i8        d0, d0, d4
+    vpadd.i8        d8, d8, d12
+    vpadd.i8        d0, d0, d8
+    vmov.32         r1, d0[1]
+    vmov.32         r8, d0[0]
+    mvn             r1, r1
+    mvn             r8, r8
+    lsrs            r1, r1, #0x1
+    rrx             r8, r8            /* shift in last r1 bit while shifting out DC bit */
+    rbit            r1, r1            /* r1 = index1 */
+    rbit            r8, r8            /* r8 = index0 */
+    ldr             r0, [r9, #0x3c0]  /* r0 = actbl->ehufco[0xf0] */
+    str             r1, [sp, #0x14]   /* index1 > sp + 0x14 */
+    cmp             r8, #0x0
+    beq             6f
+1:
+    clz             r2, r8
+    add             lr, lr, r2, lsl #1
+    lsl             r8, r8, r2
+    ldrh            r1, [lr, #-126]
+2:
+    cmp             r2, #0x10
+    blt             3f
+    sub             r2, r2, #0x10
+    put_bits        r11, r4, r0, r6
+    cmp             r4, #0x10
+    blt             2b
+    eor             r3, r3, r3
+    emit_byte       r10, r11, r4, r3, r12
+    emit_byte       r10, r11, r4, r3, r12
+    b               2b
+3:
+    add             r2, r1, r2, lsl #4
+    ldrh            r3, [lr, #2]!
+    ldr             r12, [r9, r2, lsl #2]
+    ldrb            r2, [r5, r2]
+    put_bits        r11, r4, r12, r2
+    checkbuf15      r10, r11, r4, r2, r12
+    put_bits        r11, r4, r3, r1
+    checkbuf15      r10, r11, r4, r2, r12
+    lsls            r8, r8, #0x1
+    bne             1b
+6:
+    add             r12, sp, #0x20   /* r12 = t1 */
+    ldr             r8, [sp, #0x14]  /* r8 = index1 */
+    adds            r12, #0xc0       /* r12 = t2 + (DCTSIZE2/2) */
+    cmp             r8, #0x0
+    beq             6f
+    clz             r2, r8
+    sub             r12, r12, lr
+    lsl             r8, r8, r2
+    add             r2, r2, r12, lsr #1
+    add             lr, lr, r2, lsl #1
+    b               7f
+1:
+    clz             r2, r8
+    add             lr, lr, r2, lsl #1
+    lsl             r8, r8, r2
+7:
+    ldrh            r1, [lr, #-126]
+2:
+    cmp             r2, #0x10
+    blt             3f
+    sub             r2, r2, #0x10
+    put_bits        r11, r4, r0, r6
+    cmp             r4, #0x10
+    blt             2b
+    eor             r3, r3, r3
+    emit_byte       r10, r11, r4, r3, r12
+    emit_byte       r10, r11, r4, r3, r12
+    b               2b
+3:
+    add             r2, r1, r2, lsl #4
+    ldrh            r3, [lr, #2]!
+    ldr             r12, [r9, r2, lsl #2]
+    ldrb            r2, [r5, r2]
+    put_bits        r11, r4, r12, r2
+    checkbuf15      r10, r11, r4, r2, r12
+    put_bits        r11, r4, r3, r1
+    checkbuf15      r10, r11, r4, r2, r12
+    lsls            r8, r8, #0x1
+    bne             1b
+6:
+    add             r0, sp, #0x20
+    add             r0, #0xfe
+    cmp             lr, r0
+    bhs             1f
+    ldr             r1, [r9]
+    ldrb            r0, [r5]
+    put_bits        r11, r4, r1, r0
+    checkbuf15      r10, r11, r4, r0, r1
+1:
+    ldr             r12, [sp, #0x18]
+    str             r11, [r12, #0x8]
+    str             r4, [r12, #0xc]
+    add             r0, r10, #0x1
+    add             r4, sp, #0x140
+    vld1.64         {d8, d9, d10, d11}, [r4, :128]!
+    vld1.64         {d12, d13, d14, d15}, [r4, :128]
+    sub             r4, r7, #0x1c
+    mov             sp, r4
+    pop             {r4, r5, r6, r7, r8, r9, r10, r11, pc}
+
+.purgem emit_byte
+.purgem put_bits
+.purgem checkbuf15
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,92 @@
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+;
+; -- jpeglib.h
+;
+%define DCTSIZE 8
+%define DCTSIZE2 64
+;
+; -- jmorecfg.h
+;
+%define RGB_RED 0
+%define RGB_GREEN 1
+%define RGB_BLUE 2
+%define RGB_PIXELSIZE 3
+%define EXT_RGB_RED 0
+%define EXT_RGB_GREEN 1
+%define EXT_RGB_BLUE 2
+%define EXT_RGB_PIXELSIZE 3
+%define EXT_RGBX_RED 0
+%define EXT_RGBX_GREEN 1
+%define EXT_RGBX_BLUE 2
+%define EXT_RGBX_PIXELSIZE 4
+%define EXT_BGR_RED 2
+%define EXT_BGR_GREEN 1
+%define EXT_BGR_BLUE 0
+%define EXT_BGR_PIXELSIZE 3
+%define EXT_BGRX_RED 2
+%define EXT_BGRX_GREEN 1
+%define EXT_BGRX_BLUE 0
+%define EXT_BGRX_PIXELSIZE 4
+%define EXT_XBGR_RED 3
+%define EXT_XBGR_GREEN 2
+%define EXT_XBGR_BLUE 1
+%define EXT_XBGR_PIXELSIZE 4
+%define EXT_XRGB_RED 1
+%define EXT_XRGB_GREEN 2
+%define EXT_XRGB_BLUE 3
+%define EXT_XRGB_PIXELSIZE 4
+%define RGBX_FILLER_0XFF 1
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+%define JSAMPLE byte ; unsigned char
+%define SIZEOF_JSAMPLE SIZEOF_BYTE ; sizeof(JSAMPLE)
+%define CENTERJSAMPLE 128
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF word ; short
+%define SIZEOF_JCOEF SIZEOF_WORD ; sizeof(JCOEF)
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION dword ; unsigned int
+%define SIZEOF_JDIMENSION SIZEOF_DWORD ; sizeof(JDIMENSION)
+%define JSAMPROW POINTER ; JSAMPLE * (jpeglib.h)
+%define JSAMPARRAY POINTER ; JSAMPROW * (jpeglib.h)
+%define JSAMPIMAGE POINTER ; JSAMPARRAY * (jpeglib.h)
+%define JCOEFPTR POINTER ; JCOEF * (jpeglib.h)
+%define SIZEOF_JSAMPROW SIZEOF_POINTER ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY SIZEOF_POINTER ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE SIZEOF_POINTER ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR SIZEOF_POINTER ; sizeof(JCOEFPTR)
+;
+; -- jdct.h
+;
+; A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM word ; short
+%define SIZEOF_DCTELEM SIZEOF_WORD ; sizeof(DCTELEM)
+%define float FP32 ; float
+%define SIZEOF_FAST_FLOAT SIZEOF_FP32 ; sizeof(float)
+; To maximize parallelism, Type short is changed to short.
+;
+%define ISLOW_MULT_TYPE word ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE SIZEOF_WORD ; sizeof(ISLOW_MULT_TYPE)
+%define IFAST_MULT_TYPE word ; must be short
+%define SIZEOF_IFAST_MULT_TYPE SIZEOF_WORD ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS 2 ; fractional bits in scale factors
+%define FLOAT_MULT_TYPE FP32 ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE SIZEOF_FP32 ; sizeof(FLOAT_MULT_TYPE)
+;
+; -- jsimd.h
+;
+%define JSIMD_NONE 0x00
+%define JSIMD_MMX 0x01
+%define JSIMD_3DNOW 0x02
+%define JSIMD_SSE 0x04
+%define JSIMD_SSE2 0x08
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc.h glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcfg.inc.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,130 @@
+// This file generates the include file for the assembly
+// implementations by abusing the C preprocessor.
+//
+// Note: Some things are manually defined as they need to
+// be mapped to NASM types.
+
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+
+#define JPEG_INTERNALS
+
+#include "../jpeglib.h"
+#include "../jconfig.h"
+#include "../jmorecfg.h"
+#include "jsimd.h"
+
+;
+; -- jpeglib.h
+;
+
+%define _cpp_protection_DCTSIZE DCTSIZE
+%define _cpp_protection_DCTSIZE2 DCTSIZE2
+
+;
+; -- jmorecfg.h
+;
+
+%define _cpp_protection_RGB_RED RGB_RED
+%define _cpp_protection_RGB_GREEN RGB_GREEN
+%define _cpp_protection_RGB_BLUE RGB_BLUE
+%define _cpp_protection_RGB_PIXELSIZE RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGB_RED EXT_RGB_RED
+%define _cpp_protection_EXT_RGB_GREEN EXT_RGB_GREEN
+%define _cpp_protection_EXT_RGB_BLUE EXT_RGB_BLUE
+%define _cpp_protection_EXT_RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGBX_RED EXT_RGBX_RED
+%define _cpp_protection_EXT_RGBX_GREEN EXT_RGBX_GREEN
+%define _cpp_protection_EXT_RGBX_BLUE EXT_RGBX_BLUE
+%define _cpp_protection_EXT_RGBX_PIXELSIZE EXT_RGBX_PIXELSIZE
+
+%define _cpp_protection_EXT_BGR_RED EXT_BGR_RED
+%define _cpp_protection_EXT_BGR_GREEN EXT_BGR_GREEN
+%define _cpp_protection_EXT_BGR_BLUE EXT_BGR_BLUE
+%define _cpp_protection_EXT_BGR_PIXELSIZE EXT_BGR_PIXELSIZE
+
+%define _cpp_protection_EXT_BGRX_RED EXT_BGRX_RED
+%define _cpp_protection_EXT_BGRX_GREEN EXT_BGRX_GREEN
+%define _cpp_protection_EXT_BGRX_BLUE EXT_BGRX_BLUE
+%define _cpp_protection_EXT_BGRX_PIXELSIZE EXT_BGRX_PIXELSIZE
+
+%define _cpp_protection_EXT_XBGR_RED EXT_XBGR_RED
+%define _cpp_protection_EXT_XBGR_GREEN EXT_XBGR_GREEN
+%define _cpp_protection_EXT_XBGR_BLUE EXT_XBGR_BLUE
+%define _cpp_protection_EXT_XBGR_PIXELSIZE EXT_XBGR_PIXELSIZE
+
+%define _cpp_protection_EXT_XRGB_RED EXT_XRGB_RED
+%define _cpp_protection_EXT_XRGB_GREEN EXT_XRGB_GREEN
+%define _cpp_protection_EXT_XRGB_BLUE EXT_XRGB_BLUE
+%define _cpp_protection_EXT_XRGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+
+%define RGBX_FILLER_0XFF        1
+
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+
+%define JSAMPLE                 byte          ; unsigned char
+%define SIZEOF_JSAMPLE          SIZEOF_BYTE   ; sizeof(JSAMPLE)
+
+%define _cpp_protection_CENTERJSAMPLE CENTERJSAMPLE
+
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF                   word          ; short
+%define SIZEOF_JCOEF            SIZEOF_WORD   ; sizeof(JCOEF)
+
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION              dword         ; unsigned int
+%define SIZEOF_JDIMENSION       SIZEOF_DWORD  ; sizeof(JDIMENSION)
+
+%define JSAMPROW                POINTER       ; JSAMPLE *     (jpeglib.h)
+%define JSAMPARRAY              POINTER       ; JSAMPROW *    (jpeglib.h)
+%define JSAMPIMAGE              POINTER       ; JSAMPARRAY *  (jpeglib.h)
+%define JCOEFPTR                POINTER       ; JCOEF *       (jpeglib.h)
+%define SIZEOF_JSAMPROW         SIZEOF_POINTER  ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY       SIZEOF_POINTER  ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE       SIZEOF_POINTER  ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR         SIZEOF_POINTER  ; sizeof(JCOEFPTR)
+
+;
+; -- jdct.h
+;
+
+; A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM                 word          ; short
+%define SIZEOF_DCTELEM          SIZEOF_WORD   ; sizeof(DCTELEM)
+
+%define FAST_FLOAT              FP32            ; float
+%define SIZEOF_FAST_FLOAT       SIZEOF_FP32     ; sizeof(FAST_FLOAT)
+
+; To maximize parallelism, Type MULTIPLIER is changed to short.
+;
+%define ISLOW_MULT_TYPE         word          ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE  SIZEOF_WORD   ; sizeof(ISLOW_MULT_TYPE)
+
+%define IFAST_MULT_TYPE         word          ; must be short
+%define SIZEOF_IFAST_MULT_TYPE  SIZEOF_WORD   ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS        2             ; fractional bits in scale factors
+
+%define FLOAT_MULT_TYPE         FP32          ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE  SIZEOF_FP32   ; sizeof(FLOAT_MULT_TYPE)
+
+;
+; -- jsimd.h
+;
+
+%define _cpp_protection_JSIMD_NONE JSIMD_NONE
+%define _cpp_protection_JSIMD_MMX JSIMD_MMX
+%define _cpp_protection_JSIMD_3DNOW JSIMD_3DNOW
+%define _cpp_protection_JSIMD_SSE JSIMD_SSE
+%define _cpp_protection_JSIMD_SSE2 JSIMD_SSE2
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcpu.asm glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcpu.asm
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdcpu.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdcpu.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,105 @@
+;
+; jsimdcpu.asm - SIMD instruction support check
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Check if the CPU supports SIMD instructions
+;
+; GLOBAL(unsigned int)
+; jpeg_simd_cpu_support (void)
+;
+
+        align   16
+        global  EXTN(jpeg_simd_cpu_support) PRIVATE
+
+EXTN(jpeg_simd_cpu_support):
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+        push    edi
+
+        xor     edi,edi                 ; simd support flag
+
+        pushfd
+        pop     eax
+        mov     edx,eax
+        xor     eax, 1<<21              ; flip ID bit in EFLAGS
+        push    eax
+        popfd
+        pushfd
+        pop     eax
+        xor     eax,edx
+        jz      short .return           ; CPUID is not supported
+
+        ; Check for MMX instruction support
+        xor     eax,eax
+        cpuid
+        test    eax,eax
+        jz      short .return
+
+        xor     eax,eax
+        inc     eax
+        cpuid
+        mov     eax,edx                 ; eax = Standard feature flags
+
+        test    eax, 1<<23              ; bit23:MMX
+        jz      short .no_mmx
+        or      edi, byte JSIMD_MMX
+.no_mmx:
+        test    eax, 1<<25              ; bit25:SSE
+        jz      short .no_sse
+        or      edi, byte JSIMD_SSE
+.no_sse:
+        test    eax, 1<<26              ; bit26:SSE2
+        jz      short .no_sse2
+        or      edi, byte JSIMD_SSE2
+.no_sse2:
+
+        ; Check for 3DNow! instruction support
+        mov     eax, 0x80000000
+        cpuid
+        cmp     eax, 0x80000000
+        jbe     short .return
+
+        mov     eax, 0x80000001
+        cpuid
+        mov     eax,edx                 ; eax = Extended feature flags
+
+        test    eax, 1<<31              ; bit31:3DNow!(vendor independent)
+        jz      short .no_3dnow
+        or      edi, byte JSIMD_3DNOW
+.no_3dnow:
+
+.return:
+        mov     eax,edi
+
+        pop     edi
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdext.inc glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdext.inc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimdext.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimdext.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,386 @@
+;
+; jsimdext.inc - common declarations
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2010 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library - version 1.02
+;
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+;
+; This software is provided 'as-is', without any express or implied
+; warranty.  In no event will the authors be held liable for any damages
+; arising from the use of this software.
+;
+; Permission is granted to anyone to use this software for any purpose,
+; including commercial applications, and to alter it and redistribute it
+; freely, subject to the following restrictions:
+;
+; 1. The origin of this software must not be misrepresented; you must not
+;    claim that you wrote the original software. If you use this software
+;    in a product, an acknowledgment in the product documentation would be
+;    appreciated but is not required.
+; 2. Altered source versions must be plainly marked as such, and must not be
+;    misrepresented as being the original software.
+; 3. This notice may not be removed or altered from any source distribution.
+;
+; [TAB8]
+
+; ==========================================================================
+;  System-dependent configurations
+
+%ifdef WIN32    ; ----(nasm -fwin32 -DWIN32 ...)--------
+; * Microsoft Visual C++
+; * MinGW (Minimalist GNU for Windows)
+; * CygWin
+; * LCC-Win32
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
+%define SEG_TEXT    .text  align=16 public use32 class=CODE
+%define SEG_CONST   .rdata align=16 public use32 class=CONST
+%endif
+
+%elifdef WIN64  ; ----(nasm -fwin64 -DWIN64 ...)--------
+; * Microsoft Visual C++
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
+%define SEG_TEXT    .text  align=16 public use64 class=CODE
+%define SEG_CONST   .rdata align=16 public use64 class=CONST
+%endif
+%define EXTN(name)  name                        ; foo() -> foo
+
+%elifdef OBJ32  ; ----(nasm -fobj -DOBJ32 ...)----------
+; * Borland C++ (Win32)
+
+; -- segment definition --
+;
+%define SEG_TEXT    _text  align=16 public use32 class=CODE
+%define SEG_CONST   _data  align=16 public use32 class=DATA
+
+%elifdef ELF    ; ----(nasm -felf[64] -DELF ...)------------
+; * Linux
+; * *BSD family Unix using elf format
+; * Unix System V, including Solaris x86, UnixWare and SCO Unix
+
+; mark stack as non-executable
+section .note.GNU-stack noalloc noexec nowrite progbits
+
+; -- segment definition --
+;
+%ifdef __x86_64__
+%define SEG_TEXT    .text   progbits align=16
+%define SEG_CONST   .rodata progbits align=16
+%else
+%define SEG_TEXT    .text   progbits alloc exec   nowrite align=16
+%define SEG_CONST   .rodata progbits alloc noexec nowrite align=16
+%endif
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  _GLOBAL_OFFSET_TABLE_       ; ELF supports PIC
+%define EXTN(name)  name                        ; foo() -> foo
+
+%elifdef AOUT   ; ----(nasm -faoutb/aout -DAOUT ...)----
+; * Older Linux using a.out format  (nasm -f aout -DAOUT ...)
+; * *BSD family Unix using a.out format  (nasm -f aoutb -DAOUT ...)
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text
+%define SEG_CONST   .data
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  __GLOBAL_OFFSET_TABLE_      ; BSD-style a.out supports PIC
+
+%elifdef MACHO  ; ----(nasm -fmacho -DMACHO ...)--------
+; * NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (Mach-O format)
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text  ;align=16    ; nasm doesn't accept align=16. why?
+%define SEG_CONST   .rodata align=16
+
+; The generation of position-independent code (PIC) is the default on Darwin.
+;
+%define PIC
+%define GOT_SYMBOL  _MACHO_PIC_         ; Mach-O style code-relative addressing
+
+%else           ; ----(Other case)----------------------
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text
+%define SEG_CONST   .data
+
+%endif  ; ----------------------------------------------
+
+; ==========================================================================
+
+; --------------------------------------------------------------------------
+;  Common types
+;
+%ifdef __x86_64__
+%define POINTER                 qword           ; general pointer type
+%define SIZEOF_POINTER          SIZEOF_QWORD    ; sizeof(POINTER)
+%define POINTER_BIT             QWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%else
+%define POINTER                 dword           ; general pointer type
+%define SIZEOF_POINTER          SIZEOF_DWORD    ; sizeof(POINTER)
+%define POINTER_BIT             DWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%endif
+
+%define INT                     dword           ; signed integer type
+%define SIZEOF_INT              SIZEOF_DWORD    ; sizeof(INT)
+%define INT_BIT                 DWORD_BIT       ; sizeof(INT)*BYTE_BIT
+
+%define FP32                    dword           ; IEEE754 single
+%define SIZEOF_FP32             SIZEOF_DWORD    ; sizeof(FP32)
+%define FP32_BIT                DWORD_BIT       ; sizeof(FP32)*BYTE_BIT
+
+%define MMWORD                  qword           ; int64  (MMX register)
+%define SIZEOF_MMWORD           SIZEOF_QWORD    ; sizeof(MMWORD)
+%define MMWORD_BIT              QWORD_BIT       ; sizeof(MMWORD)*BYTE_BIT
+
+; NASM is buggy and doesn't properly handle operand sizes for SSE
+; instructions, so for now we have to define XMMWORD as blank.
+%define XMMWORD                                 ; int128 (SSE register)
+%define SIZEOF_XMMWORD          SIZEOF_OWORD    ; sizeof(XMMWORD)
+%define XMMWORD_BIT             OWORD_BIT       ; sizeof(XMMWORD)*BYTE_BIT
+
+; Similar hacks for when we load a dword or MMWORD into an xmm# register
+%define XMM_DWORD
+%define XMM_MMWORD
+
+%define SIZEOF_BYTE             1               ; sizeof(BYTE)
+%define SIZEOF_WORD             2               ; sizeof(WORD)
+%define SIZEOF_DWORD            4               ; sizeof(DWORD)
+%define SIZEOF_QWORD            8               ; sizeof(QWORD)
+%define SIZEOF_OWORD            16              ; sizeof(OWORD)
+
+%define BYTE_BIT                8               ; CHAR_BIT in C
+%define WORD_BIT                16              ; sizeof(WORD)*BYTE_BIT
+%define DWORD_BIT               32              ; sizeof(DWORD)*BYTE_BIT
+%define QWORD_BIT               64              ; sizeof(QWORD)*BYTE_BIT
+%define OWORD_BIT               128             ; sizeof(OWORD)*BYTE_BIT
+
+; --------------------------------------------------------------------------
+;  External Symbol Name
+;
+%ifndef EXTN
+%define EXTN(name)   _ %+ name          ; foo() -> _foo
+%endif
+
+; --------------------------------------------------------------------------
+;  Macros for position-independent code (PIC) support
+;
+%ifndef GOT_SYMBOL
+%undef PIC
+%endif
+
+%ifdef PIC ; -------------------------------------------
+
+%ifidn GOT_SYMBOL,_MACHO_PIC_ ; --------------------
+
+; At present, nasm doesn't seem to support PIC generation for Mach-O.
+; The PIC support code below is a little tricky.
+
+        SECTION SEG_CONST
+const_base:
+
+%define GOTOFF(got,sym) (got) + (sym) - const_base
+
+%imacro get_GOT 1
+        ; NOTE: this macro destroys ecx resister.
+        call    %%geteip
+        add     ecx, byte (%%ref - $)
+        jmp     short %%adjust
+%%geteip:
+        mov     ecx, POINTER [esp]
+        ret
+%%adjust:
+        push    ebp
+        xor     ebp,ebp         ; ebp = 0
+%ifidni %1,ebx  ; (%1 == ebx)
+        ; db 0x8D,0x9C + jmp near const_base =
+        ;   lea ebx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,9C,E9,(offset32)
+        db      0x8D,0x9C               ; 8D,9C
+        jmp     near const_base         ; E9,(const_base-%%ref)
+%%ref:
+%else  ; (%1 != ebx)
+        ; db 0x8D,0x8C + jmp near const_base =
+        ;   lea ecx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,8C,E9,(offset32)
+        db      0x8D,0x8C               ; 8D,8C
+        jmp     near const_base         ; E9,(const_base-%%ref)
+%%ref:  mov     %1, ecx
+%endif ; (%1 == ebx)
+        pop     ebp
+%endmacro
+
+%else   ; GOT_SYMBOL != _MACHO_PIC_ ----------------
+
+%define GOTOFF(got,sym) (got) + (sym) wrt ..gotoff
+
+%imacro get_GOT 1
+        extern  GOT_SYMBOL
+        call    %%geteip
+        add     %1, GOT_SYMBOL + $$ - $ wrt ..gotpc
+        jmp     short %%done
+%%geteip:
+        mov     %1, POINTER [esp]
+        ret
+%%done:
+%endmacro
+
+%endif  ; GOT_SYMBOL == _MACHO_PIC_ ----------------
+
+%imacro pushpic 1.nolist
+        push    %1
+%endmacro
+%imacro poppic  1.nolist
+        pop     %1
+%endmacro
+%imacro movpic  2.nolist
+        mov     %1,%2
+%endmacro
+
+%else   ; !PIC -----------------------------------------
+
+%define GOTOFF(got,sym) (sym)
+
+%imacro get_GOT 1.nolist
+%endmacro
+%imacro pushpic 1.nolist
+%endmacro
+%imacro poppic  1.nolist
+%endmacro
+%imacro movpic  2.nolist
+%endmacro
+
+%endif  ;  PIC -----------------------------------------
+
+; --------------------------------------------------------------------------
+;  Align the next instruction on {2,4,8,16,..}-byte boundary.
+;  ".balign n,,m" in GNU as
+;
+%define MSKLE(x,y)  (~(((y) & 0xFFFF) - ((x) & 0xFFFF)) >> 16)
+%define FILLB(b,n)  (($$-(b)) & ((n)-1))
+
+%imacro alignx 1-2.nolist 0xFFFF
+%%bs:   times MSKLE(FILLB(%%bs,%1),%2) & MSKLE(16,FILLB($,%1)) & FILLB($,%1) \
+               db 0x90                               ; nop
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/9 \
+               db 0x8D,0x9C,0x23,0x00,0x00,0x00,0x00 ; lea ebx,[ebx+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/7 \
+               db 0x8D,0xAC,0x25,0x00,0x00,0x00,0x00 ; lea ebp,[ebp+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/6 \
+               db 0x8D,0xAD,0x00,0x00,0x00,0x00      ; lea ebp,[ebp+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/4 \
+               db 0x8D,0x6C,0x25,0x00                ; lea ebp,[ebp+0x00]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/3 \
+               db 0x8D,0x6D,0x00                     ; lea ebp,[ebp+0x00]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/2 \
+               db 0x8B,0xED                          ; mov ebp,ebp
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/1 \
+               db 0x90                               ; nop
+%endmacro
+
+; Align the next data on {2,4,8,16,..}-byte boundary.
+;
+%imacro alignz 1.nolist
+        align %1, db 0          ; filling zeros
+%endmacro
+
+%ifdef __x86_64__
+
+%ifdef WIN64
+
+%imacro collect_args 0
+        push r12
+        push r13
+        push r14
+        push r15
+        mov r10, rcx
+        mov r11, rdx
+        mov r12, r8
+        mov r13, r9
+        mov r14, [rax+48]
+        mov r15, [rax+56]
+        push rsi
+        push rdi
+        sub     rsp, SIZEOF_XMMWORD
+        movaps  XMMWORD [rsp], xmm6
+        sub     rsp, SIZEOF_XMMWORD
+        movaps  XMMWORD [rsp], xmm7
+%endmacro
+
+%imacro uncollect_args 0
+        movaps  xmm7, XMMWORD [rsp]
+        add     rsp, SIZEOF_XMMWORD
+        movaps  xmm6, XMMWORD [rsp]
+        add     rsp, SIZEOF_XMMWORD
+        pop rdi
+        pop rsi
+        pop r15
+        pop r14
+        pop r13
+        pop r12
+%endmacro
+
+%else
+
+%imacro collect_args 0
+        push r10
+        push r11
+        push r12
+        push r13
+        push r14
+        push r15
+        mov r10, rdi
+        mov r11, rsi
+        mov r12, rdx
+        mov r13, rcx
+        mov r14, r8
+        mov r15, r9
+%endmacro
+
+%imacro uncollect_args 0
+        pop r15
+        pop r14
+        pop r13
+        pop r12
+        pop r11
+        pop r10
+%endmacro
+
+%endif
+
+%endif
+
+; Begin chromium edits
+%ifdef MACHO ; ----(nasm -fmacho -DMACHO ...)--------
+%define PRIVATE :private_extern
+%elifdef ELF ; ----(nasm -felf[64] -DELF ...)------------
+%define PRIVATE :hidden
+%else
+%define PRIVATE
+%endif
+; End chromium edits
+
+; --------------------------------------------------------------------------
+;  Defines picked up from the C headers
+;
+%include "jsimdcfg.inc"
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd.h glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,8 +2,11 @@
  * simd/jsimd.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2011 D. R. Commander
- * 
+ * Copyright (C) 2011, 2014-2016 D. R. Commander
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2014 Linaro Limited
+ * Copyright (C) 2015-2016 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
@@ -18,649 +21,851 @@
 #define JSIMD_SSE        0x04
 #define JSIMD_SSE2       0x08
 #define JSIMD_ARM_NEON   0x10
-
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_simd_cpu_support                 jSiCpuSupport
-#define jsimd_rgb_ycc_convert_mmx             jSRGBYCCM
-#define jsimd_extrgb_ycc_convert_mmx          jSEXTRGBYCCM
-#define jsimd_extrgbx_ycc_convert_mmx         jSEXTRGBXYCCM
-#define jsimd_extbgr_ycc_convert_mmx          jSEXTBGRYCCM
-#define jsimd_extbgrx_ycc_convert_mmx         jSEXTBGRXYCCM
-#define jsimd_extxbgr_ycc_convert_mmx         jSEXTXBGRYCCM
-#define jsimd_extxrgb_ycc_convert_mmx         jSEXTXRGBYCCM
-#define jsimd_rgb_gray_convert_mmx            jSRGBGRYM
-#define jsimd_extrgb_gray_convert_mmx         jSEXTRGBGRYM
-#define jsimd_extrgbx_gray_convert_mmx        jSEXTRGBXGRYM
-#define jsimd_extbgr_gray_convert_mmx         jSEXTBGRGRYM
-#define jsimd_extbgrx_gray_convert_mmx        jSEXTBGRXGRYM
-#define jsimd_extxbgr_gray_convert_mmx        jSEXTXBGRGRYM
-#define jsimd_extxrgb_gray_convert_mmx        jSEXTXRGBGRYM
-#define jsimd_ycc_rgb_convert_mmx             jSYCCRGBM
-#define jsimd_ycc_extrgb_convert_mmx          jSYCCEXTRGBM
-#define jsimd_ycc_extrgbx_convert_mmx         jSYCCEXTRGBXM
-#define jsimd_ycc_extbgr_convert_mmx          jSYCCEXTBGRM
-#define jsimd_ycc_extbgrx_convert_mmx         jSYCCEXTBGRXM
-#define jsimd_ycc_extxbgr_convert_mmx         jSYCCEXTXBGRM
-#define jsimd_ycc_extxrgb_convert_mmx         jSYCCEXTXRGBM
-#define jconst_rgb_ycc_convert_sse2           jSCRGBYCCS2
-#define jsimd_rgb_ycc_convert_sse2            jSRGBYCCS2
-#define jsimd_extrgb_ycc_convert_sse2         jSEXTRGBYCCS2
-#define jsimd_extrgbx_ycc_convert_sse2        jSEXTRGBXYCCS2
-#define jsimd_extbgr_ycc_convert_sse2         jSEXTBGRYCCS2
-#define jsimd_extbgrx_ycc_convert_sse2        jSEXTBGRXYCCS2
-#define jsimd_extxbgr_ycc_convert_sse2        jSEXTXBGRYCCS2
-#define jsimd_extxrgb_ycc_convert_sse2        jSEXTXRGBYCCS2
-#define jconst_rgb_gray_convert_sse2          jSCRGBGRYS2
-#define jsimd_rgb_gray_convert_sse2           jSRGBGRYS2
-#define jsimd_extrgb_gray_convert_sse2        jSEXTRGBGRYS2
-#define jsimd_extrgbx_gray_convert_sse2       jSEXTRGBXGRYS2
-#define jsimd_extbgr_gray_convert_sse2        jSEXTBGRGRYS2
-#define jsimd_extbgrx_gray_convert_sse2       jSEXTBGRXGRYS2
-#define jsimd_extxbgr_gray_convert_sse2       jSEXTXBGRGRYS2
-#define jsimd_extxrgb_gray_convert_sse2       jSEXTXRGBGRYS2
-#define jconst_ycc_rgb_convert_sse2           jSCYCCRGBS2
-#define jsimd_ycc_rgb_convert_sse2            jSYCCRGBS2
-#define jsimd_ycc_extrgb_convert_sse2         jSYCCEXTRGBS2
-#define jsimd_ycc_extrgbx_convert_sse2        jSYCCEXTRGBXS2
-#define jsimd_ycc_extbgr_convert_sse2         jSYCCEXTBGRS2
-#define jsimd_ycc_extbgrx_convert_sse2        jSYCCEXTBGRXS2
-#define jsimd_ycc_extxbgr_convert_sse2        jSYCCEXTXBGRS2
-#define jsimd_ycc_extxrgb_convert_sse2        jSYCCEXTXRGBS2
-#define jsimd_h2v2_downsample_mmx             jSDnH2V2M
-#define jsimd_h2v1_downsample_mmx             jSDnH2V1M
-#define jsimd_h2v2_downsample_sse2            jSDnH2V2S2
-#define jsimd_h2v1_downsample_sse2            jSDnH2V1S2
-#define jsimd_h2v2_upsample_mmx               jSUpH2V2M
-#define jsimd_h2v1_upsample_mmx               jSUpH2V1M
-#define jsimd_h2v2_fancy_upsample_mmx         jSFUpH2V2M
-#define jsimd_h2v1_fancy_upsample_mmx         jSFUpH2V1M
-#define jsimd_h2v2_merged_upsample_mmx        jSMUpH2V2M
-#define jsimd_h2v2_extrgb_merged_upsample_mmx jSMUpH2V2EXTRGBM
-#define jsimd_h2v2_extrgbx_merged_upsample_mmx jSMUpH2V2EXTRGBXM
-#define jsimd_h2v2_extbgr_merged_upsample_mmx jSMUpH2V2EXTBGRM
-#define jsimd_h2v2_extbgrx_merged_upsample_mmx jSMUpH2V2EXTBGRXM
-#define jsimd_h2v2_extxbgr_merged_upsample_mmx jSMUpH2V2EXTXBGRM
-#define jsimd_h2v2_extxrgb_merged_upsample_mmx jSMUpH2V2EXTXRGBM
-#define jsimd_h2v1_merged_upsample_mmx        jSMUpH2V1M
-#define jsimd_h2v1_extrgb_merged_upsample_mmx jSMUpH2V1EXTRGBM
-#define jsimd_h2v1_extrgbx_merged_upsample_mmx jSMUpH2V1EXTRGBXM
-#define jsimd_h2v1_extbgr_merged_upsample_mmx jSMUpH2V1EXTBGRM
-#define jsimd_h2v1_extbgrx_merged_upsample_mmx jSMUpH2V1EXTBGRXM
-#define jsimd_h2v1_extxbgr_merged_upsample_mmx jSMUpH2V1EXTXBGRM
-#define jsimd_h2v1_extxrgb_merged_upsample_mmx jSMUpH2V1EXTXRGBM
-#define jsimd_h2v2_upsample_sse2              jSUpH2V2S2
-#define jsimd_h2v1_upsample_sse2              jSUpH2V1S2
-#define jconst_fancy_upsample_sse2            jSCFUpS2
-#define jsimd_h2v2_fancy_upsample_sse2        jSFUpH2V2S2
-#define jsimd_h2v1_fancy_upsample_sse2        jSFUpH2V1S2
-#define jconst_merged_upsample_sse2           jSCMUpS2
-#define jsimd_h2v2_merged_upsample_sse2       jSMUpH2V2S2
-#define jsimd_h2v2_extrgb_merged_upsample_sse2 jSMUpH2V2EXTRGBS2
-#define jsimd_h2v2_extrgbx_merged_upsample_sse2 jSMUpH2V2EXTRGBXS2
-#define jsimd_h2v2_extbgr_merged_upsample_sse2 jSMUpH2V2EXTBGRS2
-#define jsimd_h2v2_extbgrx_merged_upsample_sse2 jSMUpH2V2EXTBGRXS2
-#define jsimd_h2v2_extxbgr_merged_upsample_sse2 jSMUpH2V2EXTXBGRS2
-#define jsimd_h2v2_extxrgb_merged_upsample_sse2 jSMUpH2V2EXTXRGBS2
-#define jsimd_h2v1_merged_upsample_sse2       jSMUpH2V1S2
-#define jsimd_h2v1_extrgb_merged_upsample_sse2 jSMUpH2V1EXTRGBS2
-#define jsimd_h2v1_extrgbx_merged_upsample_sse2 jSMUpH2V1EXTRGBXS2
-#define jsimd_h2v1_extbgr_merged_upsample_sse2 jSMUpH2V1EXTBGRS2
-#define jsimd_h2v1_extbgrx_merged_upsample_sse2 jSMUpH2V1EXTBGRXS2
-#define jsimd_h2v1_extxbgr_merged_upsample_sse2 jSMUpH2V1EXTXBGRS2
-#define jsimd_h2v1_extxrgb_merged_upsample_sse2 jSMUpH2V1EXTXRGBS2
-#define jsimd_convsamp_mmx                    jSConvM
-#define jsimd_convsamp_sse2                   jSConvS2
-#define jsimd_convsamp_float_3dnow            jSConvF3D
-#define jsimd_convsamp_float_sse              jSConvFS
-#define jsimd_convsamp_float_sse2             jSConvFS2
-#define jsimd_fdct_islow_mmx                  jSFDMIS
-#define jsimd_fdct_ifast_mmx                  jSFDMIF
-#define jconst_fdct_islow_sse2                jSCFDS2IS
-#define jsimd_fdct_islow_sse2                 jSFDS2IS
-#define jconst_fdct_ifast_sse2                jSCFDS2IF
-#define jsimd_fdct_ifast_sse2                 jSFDS2IF
-#define jsimd_fdct_float_3dnow                jSFD3DF
-#define jconst_fdct_float_sse                 jSCFDSF
-#define jsimd_fdct_float_sse                  jSFDSF
-#define jsimd_quantize_mmx                    jSQuantM
-#define jsimd_quantize_sse2                   jSQuantS2
-#define jsimd_quantize_float_3dnow            jSQuantF3D
-#define jsimd_quantize_float_sse              jSQuantFS
-#define jsimd_quantize_float_sse2             jSQuantFS2
-#define jsimd_idct_2x2_mmx                    jSIDM22
-#define jsimd_idct_4x4_mmx                    jSIDM44
-#define jconst_idct_red_sse2                  jSCIDS2R
-#define jsimd_idct_2x2_sse2                   jSIDS222
-#define jsimd_idct_4x4_sse2                   jSIDS244
-#define jsimd_idct_islow_mmx                  jSIDMIS
-#define jsimd_idct_ifast_mmx                  jSIDMIF
-#define jconst_idct_islow_sse2                jSCIDS2IS
-#define jsimd_idct_islow_sse2                 jSIDS2IS
-#define jconst_idct_ifast_sse2                jSCIDS2IF
-#define jsimd_idct_ifast_sse2                 jSIDS2IF
-#define jsimd_idct_float_3dnow                jSID3DF
-#define jconst_fdct_float_sse                 jSCIDSF
-#define jsimd_idct_float_sse                  jSIDSF
-#define jconst_fdct_float_sse2                jSCIDS2F
-#define jsimd_idct_float_sse2                 jSIDS2F
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#define JSIMD_MIPS_DSPR2 0x20
+#define JSIMD_ALTIVEC    0x40
 
 /* SIMD Ext: retrieve SIMD/CPU information */
-EXTERN(unsigned int) jpeg_simd_cpu_support JPP((void));
+EXTERN(unsigned int) jpeg_simd_cpu_support (void);
 
-/* SIMD Color Space Conversion */
+/* RGB & extended RGB --> YCC Colorspace Conversion */
 EXTERN(void) jsimd_rgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-
-EXTERN(void) jsimd_rgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgbx_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgr_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgrx_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxbgr_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxrgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-
-EXTERN(void) jsimd_ycc_rgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extrgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extrgbx_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extbgr_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extbgrx_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extxbgr_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extxrgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
 extern const int jconst_rgb_ycc_convert_sse2[];
 EXTERN(void) jsimd_rgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_extrgb_ycc_convert_neon_slowld3
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon_slowld3
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+/* RGB & extended RGB --> Grayscale Colorspace Conversion */
+EXTERN(void) jsimd_rgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
 extern const int jconst_rgb_gray_convert_sse2[];
 EXTERN(void) jsimd_rgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+/* YCC --> RGB & extended RGB Colorspace Conversion */
+EXTERN(void) jsimd_ycc_rgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 
 extern const int jconst_ycc_rgb_convert_sse2[];
 EXTERN(void) jsimd_ycc_rgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgbx_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgr_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgrx_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxbgr_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxrgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-
-EXTERN(void) jsimd_rgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgbx_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgr_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgrx_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxbgr_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxrgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 
 EXTERN(void) jsimd_ycc_rgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgbx_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgr_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgrx_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxbgr_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxrgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert_neon
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_extrgb_convert_neon_slowst3
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_neon_slowst3
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+/* NULL Colorspace Conversion */
+EXTERN(void) jsimd_c_null_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows, int num_components);
 
-/* SIMD Downsample */
-EXTERN(void) jsimd_h2v2_downsample_mmx
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+/* h2v1 Downsampling */
 EXTERN(void) jsimd_h2v1_downsample_mmx
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-EXTERN(void) jsimd_h2v2_downsample_sse2
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
 EXTERN(void) jsimd_h2v1_downsample_sse2
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_neon
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_mips_dspr2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_altivec
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-/* SIMD Upsample */
-EXTERN(void) jsimd_h2v2_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+/* h2v2 Downsampling */
+EXTERN(void) jsimd_h2v2_downsample_mmx
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_sse2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_neon
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_mips_dspr2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_altivec
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+/* h2v2 Smooth Downsampling */
+EXTERN(void) jsimd_h2v2_smooth_downsample_mips_dspr2
+        (JSAMPARRAY input_data, JSAMPARRAY output_data,
+         JDIMENSION v_samp_factor, int max_v_samp_factor,
+         int smoothing_factor, JDIMENSION width_blocks,
+         JDIMENSION image_width);
+
+
+/* Upsampling */
 EXTERN(void) jsimd_h2v1_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_mmx
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
 
-EXTERN(void) jsimd_h2v2_fancy_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+EXTERN(void) jsimd_h2v1_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_int_upsample_mips_dspr2
+        (UINT8 h_expand, UINT8 v_expand, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr, JDIMENSION output_width,
+         int max_v_samp_factor);
+
+EXTERN(void) jsimd_h2v1_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+/* Fancy Upsampling */
 EXTERN(void) jsimd_h2v1_fancy_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mmx
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(void) jsimd_h2v2_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+extern const int jconst_fancy_upsample_sse2[];
+EXTERN(void) jsimd_h2v1_fancy_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_neon
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+/* Merged Upsampling */
 EXTERN(void) jsimd_h2v1_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-
-EXTERN(void) jsimd_h2v2_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-EXTERN(void) jsimd_h2v1_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
-extern const int jconst_fancy_upsample_sse2[];
-EXTERN(void) jsimd_h2v2_fancy_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-EXTERN(void) jsimd_h2v1_fancy_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+EXTERN(void) jsimd_h2v2_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
 extern const int jconst_merged_upsample_sse2[];
-EXTERN(void) jsimd_h2v2_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
 EXTERN(void) jsimd_h2v1_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
-/* SIMD Sample Conversion */
-EXTERN(void) jsimd_convsamp_mmx JPP((JSAMPARRAY sample_data,
-                                     JDIMENSION start_col,
-                                     DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_sse2 JPP((JSAMPARRAY sample_data,
-                                      JDIMENSION start_col,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_neon JPP((JSAMPARRAY sample_data,
-                                      JDIMENSION start_col,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_float_3dnow JPP((JSAMPARRAY sample_data,
-                                             JDIMENSION start_col,
-                                             FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_convsamp_float_sse JPP((JSAMPARRAY sample_data,
-                                           JDIMENSION start_col,
-                                           FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_convsamp_float_sse2 JPP((JSAMPARRAY sample_data,
-                                            JDIMENSION start_col,
-                                            FAST_FLOAT * workspace));
-
-/* SIMD Forward DCT */
-EXTERN(void) jsimd_fdct_islow_mmx JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_ifast_mmx JPP((DCTELEM * data));
+EXTERN(void) jsimd_h2v1_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+/* Sample Conversion */
+EXTERN(void) jsimd_convsamp_mmx
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_sse2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_neon
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_mips_dspr2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_altivec
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+/* Floating Point Sample Conversion */
+EXTERN(void) jsimd_convsamp_float_3dnow
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_mips_dspr2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+/* Slow Integer Forward DCT */
+EXTERN(void) jsimd_fdct_islow_mmx (DCTELEM *data);
 
-extern const int jconst_fdct_ifast_sse2[];
-EXTERN(void) jsimd_fdct_islow_sse2 JPP((DCTELEM * data));
 extern const int jconst_fdct_islow_sse2[];
-EXTERN(void) jsimd_fdct_ifast_sse2 JPP((DCTELEM * data));
+EXTERN(void) jsimd_fdct_islow_sse2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_neon (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_mips_dspr2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_altivec (DCTELEM *data);
 
-EXTERN(void) jsimd_fdct_ifast_neon JPP((DCTELEM * data));
+/* Fast Integer Forward DCT */
+EXTERN(void) jsimd_fdct_ifast_mmx (DCTELEM *data);
 
-EXTERN(void) jsimd_fdct_float_3dnow JPP((FAST_FLOAT * data));
+extern const int jconst_fdct_ifast_sse2[];
+EXTERN(void) jsimd_fdct_ifast_sse2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_neon (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_mips_dspr2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_altivec (DCTELEM *data);
+
+/* Floating Point Forward DCT */
+EXTERN(void) jsimd_fdct_float_3dnow (FAST_FLOAT *data);
 
 extern const int jconst_fdct_float_sse[];
-EXTERN(void) jsimd_fdct_float_sse JPP((FAST_FLOAT * data));
+EXTERN(void) jsimd_fdct_float_sse (FAST_FLOAT *data);
+
+/* Quantization */
+EXTERN(void) jsimd_quantize_mmx
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_sse2
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_neon
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
 
-/* SIMD Quantization */
-EXTERN(void) jsimd_quantize_mmx JPP((JCOEFPTR coef_block,
-                                     DCTELEM * divisors,
-                                     DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_sse2 JPP((JCOEFPTR coef_block,
-                                      DCTELEM * divisors,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_neon JPP((JCOEFPTR coef_block,
-                                      DCTELEM * divisors,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_float_3dnow JPP((JCOEFPTR coef_block,
-                                             FAST_FLOAT * divisors,
-                                             FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_quantize_float_sse JPP((JCOEFPTR coef_block,
-                                           FAST_FLOAT * divisors,
-                                           FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_quantize_float_sse2 JPP((JCOEFPTR coef_block,
-                                            FAST_FLOAT * divisors,
-                                            FAST_FLOAT * workspace));
-
-/* SIMD Reduced Inverse DCT */
-EXTERN(void) jsimd_idct_2x2_mmx JPP((void * dct_table,
-                                     JCOEFPTR coef_block,
-                                     JSAMPARRAY output_buf,
-                                     JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_mmx JPP((void * dct_table,
-                                     JCOEFPTR coef_block,
-                                     JSAMPARRAY output_buf,
-                                     JDIMENSION output_col));
+EXTERN(void) jsimd_quantize_mips_dspr2
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_altivec
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+/* Floating Point Quantization */
+EXTERN(void) jsimd_quantize_float_3dnow
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse2
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_mips_dspr2
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+/* Scaled Inverse DCT */
+EXTERN(void) jsimd_idct_2x2_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_red_sse2[];
-EXTERN(void) jsimd_idct_2x2_sse2 JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_sse2 JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_2x2_neon JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_neon JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-
-/* SIMD Inverse DCT */
-EXTERN(void) jsimd_idct_islow_mmx JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast_mmx JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
+EXTERN(void) jsimd_idct_2x2_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col, int *workspace);
+EXTERN(void) jsimd_idct_6x6_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12_pass1_mips_dspr2
+        (JCOEFPTR coef_block, void *dct_table, int *workspace);
+EXTERN(void) jsimd_idct_12x12_pass2_mips_dspr2
+        (int *workspace, int *output);
+
+/* Slow Integer Inverse DCT */
+EXTERN(void) jsimd_idct_islow_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_islow_sse2[];
-EXTERN(void) jsimd_idct_islow_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
+EXTERN(void) jsimd_idct_islow_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, int *output_buf,
+         JSAMPLE *output_col);
+
+EXTERN(void) jsimd_idct_islow_altivec
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Fast Integer Inverse DCT */
+EXTERN(void) jsimd_idct_ifast_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
 extern const int jconst_idct_ifast_sse2[];
-EXTERN(void) jsimd_idct_ifast_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_islow_neon JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast_neon JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_float_3dnow JPP((void * dct_table,
-                                         JCOEFPTR coef_block,
-                                         JSAMPARRAY output_buf,
-                                         JDIMENSION output_col));
+EXTERN(void) jsimd_idct_ifast_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_cols_mips_dspr2
+        (JCOEF *inptr, IFAST_MULT_TYPE *quantptr, DCTELEM *wsptr,
+         const int *idct_coefs);
+EXTERN(void) jsimd_idct_ifast_rows_mips_dspr2
+        (DCTELEM *wsptr, JSAMPARRAY output_buf, JDIMENSION output_col,
+         const int *idct_coefs);
+
+EXTERN(void) jsimd_idct_ifast_altivec
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Floating Point Inverse DCT */
+EXTERN(void) jsimd_idct_float_3dnow
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_float_sse[];
-EXTERN(void) jsimd_idct_float_sse JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
+EXTERN(void) jsimd_idct_float_sse
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_float_sse2[];
-EXTERN(void) jsimd_idct_float_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
+EXTERN(void) jsimd_idct_float_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Huffman coding */
+extern const int jconst_huff_encode_one_block[];
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_sse2
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_neon
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_neon_slowtbl
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_i386.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_i386.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_i386.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_i386.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1091 @@
+/*
+ * jsimd_i386.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 32-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order) (((unsigned)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = jpeg_simd_cpu_support();
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEMMX");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_MMX;
+  env = getenv("JSIMD_FORCE3DNOW");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_3DNOW|JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE|JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE2;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_ycc_convert_sse2;
+      mmxfct=jsimd_extrgb_ycc_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_ycc_convert_sse2;
+      mmxfct=jsimd_extrgbx_ycc_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_ycc_convert_sse2;
+      mmxfct=jsimd_extbgr_ycc_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_ycc_convert_sse2;
+      mmxfct=jsimd_extbgrx_ycc_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_ycc_convert_sse2;
+      mmxfct=jsimd_extxbgr_ycc_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_ycc_convert_sse2;
+      mmxfct=jsimd_extxrgb_ycc_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_rgb_ycc_convert_sse2;
+      mmxfct=jsimd_rgb_ycc_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      mmxfct=jsimd_extrgb_gray_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      mmxfct=jsimd_extrgbx_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      mmxfct=jsimd_extbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      mmxfct=jsimd_extbgrx_gray_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      mmxfct=jsimd_extxbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      mmxfct=jsimd_extxrgb_gray_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      mmxfct=jsimd_rgb_gray_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_ycc_extrgb_convert_sse2;
+      mmxfct=jsimd_ycc_extrgb_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_ycc_extrgbx_convert_sse2;
+      mmxfct=jsimd_ycc_extrgbx_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_ycc_extbgr_convert_sse2;
+      mmxfct=jsimd_ycc_extbgr_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_ycc_extbgrx_convert_sse2;
+      mmxfct=jsimd_ycc_extbgrx_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_ycc_extxbgr_convert_sse2;
+      mmxfct=jsimd_ycc_extxbgr_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_ycc_extxrgb_convert_sse2;
+      mmxfct=jsimd_ycc_extxrgb_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_ycc_rgb_convert_sse2;
+      mmxfct=jsimd_ycc_rgb_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extrgb_merged_upsample_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extrgbx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extbgrx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extxbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extxrgb_merged_upsample_mmx;
+      break;
+    default:
+      sse2fct=jsimd_h2v2_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_merged_upsample_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extrgb_merged_upsample_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extrgbx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extbgrx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extxbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extxrgb_merged_upsample_mmx;
+      break;
+    default:
+      sse2fct=jsimd_h2v1_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_merged_upsample_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_sse2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_convsamp_mmx(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_convsamp_float_sse(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_convsamp_float_3dnow(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    jsimd_fdct_islow_sse2(data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_fdct_islow_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    jsimd_fdct_ifast_sse2(data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_fdct_ifast_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    jsimd_fdct_float_sse(data);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_fdct_float_3dnow(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_sse2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_quantize_mmx(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_quantize_float_sse(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_quantize_float_3dnow(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_2x2_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_4x4_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_islow_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_ifast_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    jsimd_idct_float_sse(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_idct_float_3dnow(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1130 @@
+/*
+ * jsimd_mips.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014 D. R. Commander
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * MIPS architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+static unsigned int simd_support = ~0;
+
+#if defined(__linux__)
+
+LOCAL(int)
+parse_proc_cpuinfo(const char* search_string)
+{
+  const char* file_name = "/proc/cpuinfo";
+  char cpuinfo_line[256];
+  FILE* f = NULL;
+  simd_support = 0;
+
+  if ((f = fopen(file_name, "r")) != NULL) {
+    while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) {
+      if (strstr(cpuinfo_line, search_string) != NULL) {
+        fclose(f);
+        simd_support |= JSIMD_MIPS_DSPR2;
+        return 1;
+      }
+    }
+    fclose(f);
+  }
+  /* Did not find string in the proc file, or not Linux ELF. */
+  return 0;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__MIPSEL__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2)
+  simd_support |= JSIMD_MIPS_DSPR2;
+#elif defined(__linux__)
+  /* We still have a chance to use MIPS DSPR2 regardless of globally used
+   * -mdspr2 options passed to gcc by performing runtime detection via
+   * /proc/cpuinfo parsing on linux */
+  if (!parse_proc_cpuinfo("MIPS 74K"))
+    return;
+#endif
+}
+
+static const int mips_idct_ifast_coefs[4] = {
+  0x45404540,           // FIX( 1.082392200 / 2) =  17734 = 0x4546
+  0x5A805A80,           // FIX( 1.414213562 / 2) =  23170 = 0x5A82
+  0x76407640,           // FIX( 1.847759065 / 2) =  30274 = 0x7642
+  0xAC60AC60            // FIX(-2.613125930 / 4) = -21407 = 0xAC61
+};
+
+/* The following struct is borrowed from jdsample.c */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+
+typedef struct {
+  struct jpeg_upsampler pub;
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+  upsample1_ptr methods[MAX_COMPONENTS];
+  int next_row_out;
+  JDIMENSION rows_to_go;
+  int rowgroup_height[MAX_COMPONENTS];
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_extrgbx_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_extbgr_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_extbgrx_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_extxbgr_ycc_convert_mips_dspr2;
+
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_extxrgb_ycc_convert_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_extrgbx_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_extbgr_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_extbgrx_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_extxbgr_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_extxrgb_gray_convert_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_ycc_extrgbx_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_ycc_extbgr_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_ycc_extbgrx_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_ycc_extxbgr_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_ycc_extxrgb_convert_mips_dspr2;
+      break;
+  default:
+      mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->output_width, input_buf, input_row, output_buf,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert (j_compress_ptr cinfo,
+                      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                      JDIMENSION output_row, int num_rows)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_c_null_convert_mips_dspr2(cinfo->image_width, input_buf,
+                                    output_buf, output_row, num_rows,
+                                    cinfo->num_components);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if(DCTSIZE != 8)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_downsample_mips_dspr2(cinfo->image_width,
+                                     cinfo->max_v_samp_factor,
+                                     compptr->v_samp_factor,
+                                     compptr->width_in_blocks, input_data,
+                                     output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_smooth_downsample_mips_dspr2(input_data, output_data,
+                                          compptr->v_samp_factor,
+                                          cinfo->max_v_samp_factor,
+                                          cinfo->smoothing_factor,
+                                          compptr->width_in_blocks,
+                                          cinfo->image_width);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_downsample_mips_dspr2(cinfo->image_width,
+                                     cinfo->max_v_samp_factor,
+                                     compptr->v_samp_factor,
+                                     compptr->width_in_blocks,
+                                     input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                   cinfo->output_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                   cinfo->output_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  jsimd_int_upsample_mips_dspr2(upsample->h_expand[compptr->component_index],
+                                upsample->v_expand[compptr->component_index],
+                                input_data, output_data_ptr,
+                                cinfo->output_width,
+                                cinfo->max_v_samp_factor);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                         compptr->downsampled_width,
+                                         input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                         compptr->downsampled_width,
+                                         input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY,
+                       JSAMPLE *);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_h2v2_extbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2;
+      break;
+  }
+
+  mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+               cinfo->sample_range_limit);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY,
+                       JSAMPLE *);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_h2v1_extbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2;
+      break;
+  }
+
+  mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+               cinfo->sample_range_limit);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_convsamp_mips_dspr2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  if ((simd_support & JSIMD_MIPS_DSPR2))
+    jsimd_convsamp_float_mips_dspr2(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_fdct_islow_mips_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_fdct_ifast_mips_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_quantize_mips_dspr2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_quantize_float_mips_dspr2(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12 (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_idct_2x2_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                              output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int workspace[DCTSIZE*4];  /* buffers data between passes */
+    jsimd_idct_4x4_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                              output_col, workspace);
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+           JCOEFPTR coef_block, JSAMPARRAY output_buf,
+           JDIMENSION output_col)
+{
+    if (simd_support & JSIMD_MIPS_DSPR2)
+      jsimd_idct_6x6_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                                output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block,
+                  JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int workspace[96];
+    int output[12] = {
+      (int)(output_buf[0] + output_col),
+      (int)(output_buf[1] + output_col),
+      (int)(output_buf[2] + output_col),
+      (int)(output_buf[3] + output_col),
+      (int)(output_buf[4] + output_col),
+      (int)(output_buf[5] + output_col),
+      (int)(output_buf[6] + output_col),
+      (int)(output_buf[7] + output_col),
+      (int)(output_buf[8] + output_col),
+      (int)(output_buf[9] + output_col),
+      (int)(output_buf[10] + output_col),
+      (int)(output_buf[11] + output_col),
+    };
+    jsimd_idct_12x12_pass1_mips_dspr2(coef_block, compptr->dct_table,
+                                      workspace);
+    jsimd_idct_12x12_pass2_mips_dspr2(workspace, output);
+  }
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int output[8] = {
+      (int)(output_buf[0] + output_col),
+      (int)(output_buf[1] + output_col),
+      (int)(output_buf[2] + output_col),
+      (int)(output_buf[3] + output_col),
+      (int)(output_buf[4] + output_col),
+      (int)(output_buf[5] + output_col),
+      (int)(output_buf[6] + output_col),
+      (int)(output_buf[7] + output_col),
+    };
+
+    jsimd_idct_islow_mips_dspr2(coef_block, compptr->dct_table,
+                                output, IDCT_range_limit(cinfo));
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    JCOEFPTR inptr;
+    IFAST_MULT_TYPE *quantptr;
+    DCTELEM workspace[DCTSIZE2];  /* buffers data between passes */
+
+    /* Pass 1: process columns from input, store into work array. */
+
+    inptr = coef_block;
+    quantptr = (IFAST_MULT_TYPE *) compptr->dct_table;
+
+    jsimd_idct_ifast_cols_mips_dspr2(inptr, quantptr,
+                                     workspace, mips_idct_ifast_coefs);
+
+    /* Pass 2: process rows from work array, store into output array. */
+    /* Note that we must descale the results by a factor of 8 == 2**3, */
+    /* and also undo the PASS1_BITS scaling. */
+
+    jsimd_idct_ifast_rows_mips_dspr2(workspace, output_buf,
+                                     output_col, mips_idct_ifast_coefs);
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,285 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013, MIPS Technologies, Inc., California.
+ * All rights reserved.
+ * Authors:  Teodora Novkovic (teodora.novkovic@imgtec.com)
+ *           Darko Laus       (darko.laus@imgtec.com)
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define zero $0
+#define AT   $1
+#define v0   $2
+#define v1   $3
+#define a0   $4
+#define a1   $5
+#define a2   $6
+#define a3   $7
+#define t0   $8
+#define t1   $9
+#define t2   $10
+#define t3   $11
+#define t4   $12
+#define t5   $13
+#define t6   $14
+#define t7   $15
+#define s0   $16
+#define s1   $17
+#define s2   $18
+#define s3   $19
+#define s4   $20
+#define s5   $21
+#define s6   $22
+#define s7   $23
+#define t8   $24
+#define t9   $25
+#define k0   $26
+#define k1   $27
+#define gp   $28
+#define sp   $29
+#define fp   $30
+#define s8   $30
+#define ra   $31
+
+#define f0   $f0
+#define f1   $f1
+#define f2   $f2
+#define f3   $f3
+#define f4   $f4
+#define f5   $f5
+#define f6   $f6
+#define f7   $f7
+#define f8   $f8
+#define f9   $f9
+#define f10  $f10
+#define f11  $f11
+#define f12  $f12
+#define f13  $f13
+#define f14  $f14
+#define f15  $f15
+#define f16  $f16
+#define f17  $f17
+#define f18  $f18
+#define f19  $f19
+#define f20  $f20
+#define f21  $f21
+#define f22  $f22
+#define f23  $f23
+#define f24  $f24
+#define f25  $f25
+#define f26  $f26
+#define f27  $f27
+#define f28  $f28
+#define f29  $f29
+#define f30  $f30
+#define f31  $f31
+
+/*
+ * LEAF_MIPS32R2 - declare leaf routine for MIPS32r2
+ */
+#define LEAF_MIPS32R2(symbol)                           \
+                .globl  symbol;                         \
+                .align  2;                              \
+                .type   symbol, @function;              \
+                .ent    symbol, 0;                      \
+symbol:         .frame  sp, 0, ra;                      \
+                .set    push;                           \
+                .set    arch=mips32r2;                  \
+                .set    noreorder;                      \
+                .set    noat;
+
+/*
+ * LEAF_MIPS_DSPR2 - declare leaf routine for MIPS DSPr2
+ */
+#define LEAF_MIPS_DSPR2(symbol)                         \
+LEAF_MIPS32R2(symbol)                                   \
+                .set    dspr2;
+
+/*
+ * END - mark end of function
+ */
+#define END(function)                                   \
+                .set    pop;                            \
+                .end    function;                       \
+                .size   function,.-function
+
+/*
+ * Checks if stack offset is big enough for storing/restoring regs_num
+ * number of register to/from stack. Stack offset must be greater than
+ * or equal to the number of bytes needed for storing registers (regs_num*4).
+ * Since MIPS ABI allows usage of first 16 bytes of stack frame (this is
+ * preserved for input arguments of the functions, already stored in a0-a3),
+ * stack size can be further optimized by utilizing this space.
+ */
+.macro CHECK_STACK_OFFSET regs_num, stack_offset
+.if \stack_offset < \regs_num * 4 - 16
+.error "Stack offset too small."
+.endif
+.endm
+
+/*
+ * Saves set of registers on stack. Maximum number of registers that
+ * can be saved on stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * before registers are pushed in order to provide enough space on stack
+ * (offset must be multiple of 4, and must be big enough, as described by
+ * CHECK_STACK_OFFSET macro). This macro is intended to be used in
+ * combination with RESTORE_REGS_FROM_STACK macro. Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro SAVE_REGS_ON_STACK stack_offset = 0, r1, \
+                          r2  = 0, r3  = 0, r4  = 0, \
+                          r5  = 0, r6  = 0, r7  = 0, \
+                          r8  = 0, r9  = 0, r10 = 0, \
+                          r11 = 0, r12 = 0, r13 = 0, \
+                          r14 = 0
+    .if (\stack_offset < 0) || (\stack_offset - (\stack_offset / 4) * 4)
+    .error "Stack offset must be pozitive and multiple of 4."
+    .endif
+    .if \stack_offset != 0
+    addiu           sp, sp, -\stack_offset
+    .endif
+    sw              \r1, 0(sp)
+    .if \r2 != 0
+    sw              \r2, 4(sp)
+    .endif
+    .if \r3 != 0
+    sw              \r3, 8(sp)
+    .endif
+    .if \r4 != 0
+    sw              \r4, 12(sp)
+    .endif
+    .if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    sw              \r5, 16(sp)
+    .endif
+    .if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    sw              \r6, 20(sp)
+    .endif
+    .if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    sw              \r7, 24(sp)
+    .endif
+    .if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    sw              \r8, 28(sp)
+    .endif
+    .if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    sw              \r9, 32(sp)
+    .endif
+    .if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    sw              \r10, 36(sp)
+    .endif
+    .if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    sw              \r11, 40(sp)
+    .endif
+    .if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    sw              \r12, 44(sp)
+    .endif
+    .if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    sw              \r13, 48(sp)
+    .endif
+    .if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    sw              \r14, 52(sp)
+    .endif
+.endm
+
+/*
+ * Restores set of registers from stack. Maximum number of registers that
+ * can be restored from stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * after registers are restored (offset must be multiple of 4, and must
+ * be big enough, as described by CHECK_STACK_OFFSET macro). This macro is
+ * intended to be used in combination with RESTORE_REGS_FROM_STACK macro.
+ * Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro RESTORE_REGS_FROM_STACK stack_offset = 0, r1, \
+                               r2  = 0, r3  = 0, r4  = 0, \
+                               r5  = 0, r6  = 0, r7  = 0, \
+                               r8  = 0, r9  = 0, r10 = 0, \
+                               r11 = 0, r12 = 0, r13 = 0, \
+                               r14 = 0
+    .if (\stack_offset < 0) || (\stack_offset - (\stack_offset/4)*4)
+    .error "Stack offset must be pozitive and multiple of 4."
+    .endif
+    lw              \r1, 0(sp)
+    .if \r2 != 0
+    lw              \r2, 4(sp)
+    .endif
+    .if \r3 != 0
+    lw              \r3, 8(sp)
+    .endif
+    .if \r4 != 0
+    lw              \r4, 12(sp)
+    .endif
+    .if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    lw              \r5, 16(sp)
+    .endif
+    .if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    lw              \r6, 20(sp)
+    .endif
+    .if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    lw              \r7, 24(sp)
+    .endif
+    .if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    lw              \r8, 28(sp)
+    .endif
+    .if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    lw              \r9, 32(sp)
+    .endif
+    .if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    lw              \r10, 36(sp)
+    .endif
+    .if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    lw              \r11, 40(sp)
+    .endif
+    .if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    lw              \r12, 44(sp)
+    .endif
+    .if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    lw              \r13, 48(sp)
+    .endif
+    .if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    lw              \r14, 52(sp)
+    .endif
+    .if \stack_offset != 0
+    addiu           sp, sp, \stack_offset
+    .endif
+.endm
+
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4487 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ * All rights reserved.
+ * Authors:  Teodora Novkovic (teodora.novkovic@imgtec.com)
+ *           Darko Laus       (darko.laus@imgtec.com)
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#include "jsimd_mips_dspr2_asm.h"
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_c_null_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ * 20(sp) - cinfo->num_components
+ *
+ * Null conversion for compression
+ */
+
+    SAVE_REGS_ON_STACK 8, s0, s1
+
+    lw        t9, 24(sp)   // t9 = num_rows
+    lw        s0, 28(sp)   // s0 = cinfo->num_components
+    andi      t0, a0, 3    // t0 = cinfo->image_width & 3
+    beqz      t0, 4f       // no residual
+     nop
+0:
+    addiu     t9, t9, -1
+    bltz      t9, 7f
+     li       t1, 0
+1:
+    sll       t3, t1, 2
+    lwx       t5, t3(a2)   // t5 = outptr = output_buf[ci]
+    lw        t2, 0(a1)    // t2 = inptr = *input_buf
+    sll       t4, a3, 2
+    lwx       t5, t4(t5)   // t5 = outptr = output_buf[ci][output_row]
+    addu      t2, t2, t1
+    addu      s1, t5, a0
+    addu      t6, t5, t0
+2:
+    lbu       t3, 0(t2)
+    addiu     t5, t5, 1
+    sb        t3, -1(t5)
+    bne       t6, t5, 2b
+     addu     t2, t2, s0
+3:
+    lbu       t3, 0(t2)
+    addu      t4, t2, s0
+    addu      t7, t4, s0
+    addu      t8, t7, s0
+    addu      t2, t8, s0
+    lbu       t4, 0(t4)
+    lbu       t7, 0(t7)
+    lbu       t8, 0(t8)
+    addiu     t5, t5, 4
+    sb        t3, -4(t5)
+    sb        t4, -3(t5)
+    sb        t7, -2(t5)
+    bne       s1, t5, 3b
+     sb       t8, -1(t5)
+    addiu     t1, t1, 1
+    bne       t1, s0, 1b
+     nop
+    addiu     a1, a1, 4
+    bgez      t9, 0b
+     addiu    a3, a3, 1
+    b         7f
+     nop
+4:
+    addiu     t9, t9, -1
+    bltz      t9, 7f
+     li       t1, 0
+5:
+    sll       t3, t1, 2
+    lwx       t5, t3(a2)   // t5 = outptr = output_buf[ci]
+    lw        t2, 0(a1)    // t2 = inptr = *input_buf
+    sll       t4, a3, 2
+    lwx       t5, t4(t5)   // t5 = outptr = output_buf[ci][output_row]
+    addu      t2, t2, t1
+    addu      s1, t5, a0
+    addu      t6, t5, t0
+6:
+    lbu       t3, 0(t2)
+    addu      t4, t2, s0
+    addu      t7, t4, s0
+    addu      t8, t7, s0
+    addu      t2, t8, s0
+    lbu       t4, 0(t4)
+    lbu       t7, 0(t7)
+    lbu       t8, 0(t8)
+    addiu     t5, t5, 4
+    sb        t3, -4(t5)
+    sb        t4, -3(t5)
+    sb        t7, -2(t5)
+    bne       s1, t5, 6b
+     sb       t8, -1(t5)
+    addiu     t1, t1, 1
+    bne       t1, s0, 5b
+     nop
+    addiu     a1, a1, 4
+    bgez      t9, 4b
+     addiu    a3, a3, 1
+7:
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j         ra
+     nop
+
+END(jsimd_c_null_convert_mips_dspr2)
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_ycc_convert_mips_dspr2
+ * jsimd_extbgr_ycc_convert_mips_dspr2
+ * jsimd_extrgbx_ycc_convert_mips_dspr2
+ * jsimd_extbgrx_ycc_convert_mips_dspr2
+ * jsimd_extxbgr_ycc_convert_mips_dspr2
+ * jsimd_extxrgb_ycc_convert_mips_dspr2
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_YCC r,    \
+                     g,    \
+                     b,    \
+                     inptr
+    lbu     \r, \r_offs(\inptr)
+    lbu     \g, \g_offs(\inptr)
+    lbu     \b, \b_offs(\inptr)
+    addiu   \inptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_\colorid\()_ycc_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw      t7, 48(sp)        // t7 = num_rows
+    li      s0, 0x4c8b        // FIX(0.29900)
+    li      s1, 0x9646        // FIX(0.58700)
+    li      s2, 0x1d2f        // FIX(0.11400)
+    li      s3, 0xffffd4cd    // -FIX(0.16874)
+    li      s4, 0xffffab33    // -FIX(0.33126)
+    li      s5, 0x8000        // FIX(0.50000)
+    li      s6, 0xffff94d1    // -FIX(0.41869)
+    li      s7, 0xffffeb2f    // -FIX(0.08131)
+    li      t8, 0x807fff      // CBCR_OFFSET + ONE_HALF-1
+
+0:
+    addiu   t7, -1            // --num_rows
+    lw      t6, 0(a1)         // t6 = input_buf[0]
+    lw      t0, 0(a2)
+    lw      t1, 4(a2)
+    lw      t2, 8(a2)
+    sll     t3, a3, 2
+    lwx     t0, t3(t0)        // t0 = output_buf[0][output_row]
+    lwx     t1, t3(t1)        // t1 = output_buf[1][output_row]
+    lwx     t2, t3(t2)        // t2 = output_buf[2][output_row]
+
+    addu    t9, t2, a0        // t9 = end address
+    addiu   a3, 1
+
+1:
+    DO_RGB_TO_YCC t3, t4, t5, t6
+
+    mtlo    s5, $ac0
+    mtlo    t8, $ac1
+    mtlo    t8, $ac2
+    maddu   $ac0, s2, t5
+    maddu   $ac1, s5, t5
+    maddu   $ac2, s5, t3
+    maddu   $ac0, s0, t3
+    maddu   $ac1, s3, t3
+    maddu   $ac2, s6, t4
+    maddu   $ac0, s1, t4
+    maddu   $ac1, s4, t4
+    maddu   $ac2, s7, t5
+    extr.w  t3, $ac0, 16
+    extr.w  t4, $ac1, 16
+    extr.w  t5, $ac2, 16
+    sb      t3, 0(t0)
+    sb      t4, 0(t1)
+    sb      t5, 0(t2)
+    addiu   t0, 1
+    addiu   t2, 1
+    bne     t2, t9, 1b
+     addiu  t1, 1
+    bgtz    t7, 0b
+     addiu  a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_\colorid\()_ycc_convert_mips_dspr2)
+
+.purgem DO_RGB_TO_YCC
+
+.endm
+
+/*------------------------------------------id -- pix R  G  B */
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3
+
+/*****************************************************************************/
+/*
+ * jsimd_ycc_extrgb_convert_mips_dspr2
+ * jsimd_ycc_extbgr_convert_mips_dspr2
+ * jsimd_ycc_extrgbx_convert_mips_dspr2
+ * jsimd_ycc_extbgrx_convert_mips_dspr2
+ * jsimd_ycc_extxbgr_convert_mips_dspr2
+ * jsimd_ycc_extxrgb_convert_mips_dspr2
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs, a_offs
+
+.macro STORE_YCC_TO_RGB  scratch0 \
+                         scratch1 \
+                         scratch2 \
+                         outptr
+    sb       \scratch0, \r_offs(\outptr)
+    sb       \scratch1, \g_offs(\outptr)
+    sb       \scratch2, \b_offs(\outptr)
+.if (\pixel_size == 4)
+    li       t0, 0xFF
+    sb       t0, \a_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_ycc_\colorid\()_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - input_row
+ * a3     - output_buf
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw         s1, 48(sp)
+    li         t3, 0x8000
+    li         t4, 0x166e9     // FIX(1.40200)
+    li         t5, 0x1c5a2     // FIX(1.77200)
+    li         t6, 0xffff492e  // -FIX(0.71414)
+    li         t7, 0xffffa7e6  // -FIX(0.34414)
+    repl.ph    t8, 128
+
+0:
+    lw         s0, 0(a3)
+    lw         t0, 0(a1)
+    lw         t1, 4(a1)
+    lw         t2, 8(a1)
+    sll        s5, a2, 2
+    addiu      s1, -1
+    lwx        s2, s5(t0)
+    lwx        s3, s5(t1)
+    lwx        s4, s5(t2)
+    addu       t9, s2, a0
+    addiu      a2, 1
+
+1:
+    lbu        s7, 0(s4)       // cr
+    lbu        s6, 0(s3)       // cb
+    lbu        s5, 0(s2)       // y
+    addiu      s2, 1
+    addiu      s4, 1
+    addiu      s7, -128
+    addiu      s6, -128
+    mul        t2, t7, s6
+    mul        t0, t6, s7      // Crgtab[cr]
+    sll        s7, 15
+    mulq_rs.w  t1, t4, s7      // Crrtab[cr]
+    sll        s6, 15
+    addu       t2, t3          // Cbgtab[cb]
+    addu       t2, t0
+
+    mulq_rs.w  t0, t5, s6      // Cbbtab[cb]
+    sra        t2, 16
+    addu       t1, s5
+    addu       t2, s5          // add y
+    ins        t2, t1, 16, 16
+    subu.ph    t2, t2, t8
+    addu       t0, s5
+    shll_s.ph  t2, t2, 8
+    subu       t0, 128
+    shra.ph    t2, t2, 8
+    shll_s.w   t0, t0, 24
+    addu.ph    t2, t2, t8      // clip & store
+    sra        t0, t0, 24
+    sra        t1, t2, 16
+    addiu      t0, 128
+
+    STORE_YCC_TO_RGB t1, t2, t0, s0
+
+    bne        s2, t9, 1b
+     addiu     s3, 1
+    bgtz       s1, 0b
+     addiu     a3, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_ycc_\colorid\()_convert_mips_dspr2)
+
+.purgem STORE_YCC_TO_RGB
+
+.endm
+
+/*------------------------------------------id -- pix R  G  B  A */
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1, 0
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3, 0
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_gray_convert_mips_dspr2
+ * jsimd_extbgr_gray_convert_mips_dspr2
+ * jsimd_extrgbx_gray_convert_mips_dspr2
+ * jsimd_extbgrx_gray_convert_mips_dspr2
+ * jsimd_extxbgr_gray_convert_mips_dspr2
+ * jsimd_extxrgb_gray_convert_mips_dspr2
+ *
+ * Colorspace conversion RGB -> GRAY
+ */
+
+.macro GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_GRAY r,    \
+                      g,    \
+                      b,    \
+                      inptr
+    lbu     \r, \r_offs(\inptr)
+    lbu     \g, \g_offs(\inptr)
+    lbu     \b, \b_offs(\inptr)
+    addiu   \inptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_\colorid\()_gray_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    li      s0, 0x4c8b             // s0 = FIX(0.29900)
+    li      s1, 0x9646             // s1 = FIX(0.58700)
+    li      s2, 0x1d2f             // s2 = FIX(0.11400)
+    li      s7, 0x8000             // s7 = FIX(0.50000)
+    lw      s6, 48(sp)
+    andi    t7, a0, 3
+
+0:
+    addiu   s6, -1                 // s6 = num_rows
+    lw      t0, 0(a1)
+    lw      t1, 0(a2)
+    sll     t3, a3, 2
+    lwx     t1, t3(t1)
+    addiu   a3, 1
+    addu    t9, t1, a0
+    subu    t8, t9, t7
+    beq     t1, t8, 2f
+     nop
+
+1:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    maddu   $ac0, s0, t3
+    mtlo    s7, $ac1
+    maddu   $ac1, s2, s5
+    maddu   $ac1, s1, s4
+    maddu   $ac1, s0, s3
+    extr.w  t6, $ac0, 16
+
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    extr.w  t2, $ac1, 16
+    maddu   $ac0, s0, t3
+    mtlo    s7, $ac1
+    maddu   $ac1, s2, s5
+    maddu   $ac1, s1, s4
+    maddu   $ac1, s0, s3
+    extr.w  t5, $ac0, 16
+    sb      t6, 0(t1)
+    sb      t2, 1(t1)
+    extr.w  t3, $ac1, 16
+    addiu   t1, 4
+    sb      t5, -2(t1)
+    sb      t3, -1(t1)
+    bne     t1, t8, 1b
+     nop
+
+2:
+    beqz    t7, 4f
+     nop
+
+3:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    maddu   $ac0, s0, t3
+    extr.w  t6, $ac0, 16
+    sb      t6, 0(t1)
+    addiu   t1, 1
+    bne     t1, t9, 3b
+     nop
+
+4:
+    bgtz    s6, 0b
+     addiu  a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_\colorid\()_gray_convert_mips_dspr2)
+
+.purgem DO_RGB_TO_GRAY
+
+.endm
+
+/*------------------------------------------id --  pix R  G  B */
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extrgb_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2
+ *
+ * Merged h2v2 upsample routines
+ */
+.macro GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 colorid,    \
+                                                pixel_size, \
+                                                r1_offs,    \
+                                                g1_offs,    \
+                                                b1_offs,    \
+                                                a1_offs,    \
+                                                r2_offs,    \
+                                                g2_offs,    \
+                                                b2_offs,    \
+                                                a2_offs
+
+.macro STORE_H2V2_2_PIXELS  scratch0 \
+                            scratch1 \
+                            scratch2 \
+                            scratch3 \
+                            scratch4 \
+                            scratch5 \
+                            outptr
+    sb       \scratch0, \r1_offs(\outptr)
+    sb       \scratch1, \g1_offs(\outptr)
+    sb       \scratch2, \b1_offs(\outptr)
+    sb       \scratch3, \r2_offs(\outptr)
+    sb       \scratch4, \g2_offs(\outptr)
+    sb       \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li       \scratch0, 0xFF
+    sb       \scratch0, \a1_offs(\outptr)
+    sb       \scratch0, \a2_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V2_1_PIXEL  scratch0 \
+                           scratch1 \
+                           scratch2 \
+                           outptr
+    sb    \scratch0, \r1_offs(\outptr)
+    sb    \scratch1, \g1_offs(\outptr)
+    sb    \scratch2, \b1_offs(\outptr)
+
+.if (\pixel_size == 8)
+    li    t0, 0xFF
+    sb    t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->output_width
+ * a1     - input_buf
+ * a2     - in_row_group_ctr
+ * a3     - output_buf
+ * 16(sp) - cinfo->sample_range_limit
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    lw           t9, 56(sp)        // cinfo->sample_range_limit
+    lw           v0, 0(a1)
+    lw           v1, 4(a1)
+    lw           t0, 8(a1)
+    sll          t1, a2, 3
+    addiu        t2, t1, 4
+    sll          t3, a2, 2
+    lw           t4, 0(a3)         // t4 = output_buf[0]
+    lwx          t1, t1(v0)        // t1 = input_buf[0][in_row_group_ctr*2]
+    lwx          t2, t2(v0)        // t2 = input_buf[0][in_row_group_ctr*2 + 1]
+    lwx          t5, t3(v1)        // t5 = input_buf[1][in_row_group_ctr]
+    lwx          t6, t3(t0)        // t6 = input_buf[2][in_row_group_ctr]
+    lw           t7, 4(a3)         // t7 = output_buf[1]
+    li           s1, 0xe6ea
+    addiu        t8, s1, 0x7fff    // t8 = 0x166e9 [FIX(1.40200)]
+    addiu        s0, t8, 0x5eb9    // s0 = 0x1c5a2 [FIX(1.77200)]
+    addiu        s1, zero, 0xa7e6  // s4 = 0xffffa7e6 [-FIX(0.34414)]
+    xori         s2, s1, 0xeec8    // s3 = 0xffff492e [-FIX(0.71414)]
+    srl          t3, a0, 1
+    blez         t3, 2f
+     addu        t0, t5, t3        // t0 = end address
+ 1:
+    lbu          t3, 0(t5)
+    lbu          s3, 0(t6)
+    addiu        t5, t5, 1
+    addiu        t3, t3, -128      // (cb - 128)
+    addiu        s3, s3, -128      // (cr - 128)
+    mult         $ac1, s1, t3
+    madd         $ac1, s2, s3
+    sll          s3, s3, 15
+    sll          t3, t3, 15
+    mulq_rs.w    s4, t8, s3        // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS
+    extr_r.w     s5, $ac1, 16
+    mulq_rs.w    s6, s0, t3        // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS
+    lbu          v0, 0(t1)
+    addiu        t6, t6, 1
+    addiu        t1, t1, 2
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          AT, 0(t3)
+    lbu          s7, 0(s3)
+    lbu          ra, 0(v1)
+    lbu          v0, -1(t1)
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+    lbu          v0, 0(t2)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t4
+
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          AT, 0(t3)
+    lbu          s7, 0(s3)
+    lbu          ra, 0(v1)
+    lbu          v0, 1(t2)
+    addiu        t2, t2, 2
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t7
+
+    bne          t0, t5, 1b
+     nop
+2:
+    andi         t0, a0, 1
+    beqz         t0, 4f
+     lbu          t3, 0(t5)
+    lbu          s3, 0(t6)
+    addiu        t3, t3, -128      // (cb - 128)
+    addiu        s3, s3, -128      // (cr - 128)
+    mult         $ac1, s1, t3
+    madd         $ac1, s2, s3
+    sll          s3, s3, 15
+    sll          t3, t3, 15
+    lbu          v0, 0(t1)
+    extr_r.w     s5, $ac1, 16
+    mulq_rs.w    s4, t8, s3        // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS
+    mulq_rs.w    s6, s0, t3        // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+    lbu          v0, 0(t2)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t4
+
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2)
+
+.purgem STORE_H2V2_1_PIXEL
+.purgem STORE_H2V2_2_PIXELS
+.endm
+
+/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+/*****************************************************************************/
+/*
+ * jsimd_h2v1_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extrgb_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2
+ *
+ * Merged h2v1 upsample routines
+ */
+
+.macro GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 colorid,    \
+                                                pixel_size, \
+                                                r1_offs,    \
+                                                g1_offs,    \
+                                                b1_offs,    \
+                                                a1_offs,    \
+                                                r2_offs,    \
+                                                g2_offs,    \
+                                                b2_offs,    \
+                                                a2_offs
+
+.macro STORE_H2V1_2_PIXELS  scratch0 \
+                            scratch1 \
+                            scratch2 \
+                            scratch3 \
+                            scratch4 \
+                            scratch5 \
+                            outptr
+    sb       \scratch0, \r1_offs(\outptr)
+    sb       \scratch1, \g1_offs(\outptr)
+    sb       \scratch2, \b1_offs(\outptr)
+    sb       \scratch3, \r2_offs(\outptr)
+    sb       \scratch4, \g2_offs(\outptr)
+    sb       \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li       t0, 0xFF
+    sb       t0, \a1_offs(\outptr)
+    sb       t0, \a2_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V1_1_PIXEL  scratch0 \
+                           scratch1 \
+                           scratch2 \
+                           outptr
+    sb    \scratch0, \r1_offs(\outptr)
+    sb    \scratch1, \g1_offs(\outptr)
+    sb    \scratch2, \b1_offs(\outptr)
+.if (\pixel_size == 8)
+    li    t0, 0xFF
+    sb    t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->output_width
+ * a1     - input_buf
+ * a2     - in_row_group_ctr
+ * a3     - output_buf
+ * 16(sp) - range_limit
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    li           t0, 0xe6ea
+    lw           t1, 0(a1)         // t1 = input_buf[0]
+    lw           t2, 4(a1)         // t2 = input_buf[1]
+    lw           t3, 8(a1)         // t3 = input_buf[2]
+    lw           t8, 56(sp)        // t8 = range_limit
+    addiu        s1, t0, 0x7fff    // s1 = 0x166e9 [FIX(1.40200)]
+    addiu        s2, s1, 0x5eb9    // s2 = 0x1c5a2 [FIX(1.77200)]
+    addiu        s0, t0, 0x9916    // s0 = 0x8000
+    addiu        s4, zero, 0xa7e6  // s4 = 0xffffa7e6 [-FIX(0.34414)]
+    xori         s3, s4, 0xeec8    // s3 = 0xffff492e [-FIX(0.71414)]
+    srl          t0, a0, 1
+    sll          t4, a2, 2
+    lwx          s5, t4(t1)        // s5 = inptr0
+    lwx          s6, t4(t2)        // s6 = inptr1
+    lwx          s7, t4(t3)        // s7 = inptr2
+    lw           t7, 0(a3)         // t7 = outptr
+    blez         t0, 2f
+     addu        t9, s6, t0        // t9 = end address
+1:
+    lbu          t2, 0(s6)         // t2 = cb
+    lbu          t0, 0(s7)         // t0 = cr
+    lbu          t1, 0(s5)         // t1 = y
+    addiu        t2, t2, -128      // t2 = cb - 128
+    addiu        t0, t0, -128      // t0 = cr - 128
+    mult         $ac1, s4, t2
+    madd         $ac1, s3, t0
+    sll          t0, t0, 15
+    sll          t2, t2, 15
+    mulq_rs.w    t0, s1, t0        // t0 = (C1*cr + ONE_HALF)>> SCALEBITS
+    extr_r.w     t5, $ac1, 16
+    mulq_rs.w    t6, s2, t2        // t6 = (C2*cb + ONE_HALF)>> SCALEBITS
+    addiu        s7, s7, 1
+    addiu        s6, s6, 1
+    addu         t2, t1, t0        // t2 = y + cred
+    addu         t3, t1, t5        // t3 = y + cgreen
+    addu         t4, t1, t6        // t4 = y + cblue
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t1, 1(s5)
+    lbu          v0, 0(t2)
+    lbu          v1, 0(t3)
+    lbu          ra, 0(t4)
+    addu         t2, t1, t0
+    addu         t3, t1, t5
+    addu         t4, t1, t6
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t2, 0(t2)
+    lbu          t3, 0(t3)
+    lbu          t4, 0(t4)
+
+    STORE_H2V1_2_PIXELS v0, v1, ra, t2, t3, t4, t7
+
+    bne          t9, s6, 1b
+     addiu       s5, s5, 2
+2:
+    andi         t0, a0, 1
+    beqz         t0, 4f
+     nop
+3:
+    lbu          t2, 0(s6)
+    lbu          t0, 0(s7)
+    lbu          t1, 0(s5)
+    addiu        t2, t2, -128      //(cb - 128)
+    addiu        t0, t0, -128      //(cr - 128)
+    mul          t3, s4, t2
+    mul          t4, s3, t0
+    sll          t0, t0, 15
+    sll          t2, t2, 15
+    mulq_rs.w    t0, s1, t0       // (C1*cr + ONE_HALF)>> SCALEBITS
+    mulq_rs.w    t6, s2, t2       // (C2*cb + ONE_HALF)>> SCALEBITS
+    addu         t3, t3, s0
+    addu         t3, t4, t3
+    sra          t5, t3, 16       // (C4*cb + ONE_HALF + C3*cr)>> SCALEBITS
+    addu         t2, t1, t0       // y + cred
+    addu         t3, t1, t5       // y + cgreen
+    addu         t4, t1, t6       // y + cblue
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t2, 0(t2)
+    lbu          t3, 0(t3)
+    lbu          t4, 0(t4)
+
+    STORE_H2V1_1_PIXEL t2, t3, t4, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j            ra
+     nop
+
+END(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2)
+
+.purgem STORE_H2V1_1_PIXEL
+.purgem STORE_H2V1_2_PIXELS
+.endm
+
+/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_fancy_upsample_mips_dspr2
+ *
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ */
+LEAF_MIPS_DSPR2(jsimd_h2v2_fancy_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - downsampled_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    li             s4, 0
+    lw             s2, 0(a3)       // s2 = *output_data_ptr
+0:
+    li             t9, 2
+    lw             s1, -4(a2)      // s1 = inptr1
+
+1:
+    lw             s0, 0(a2)       // s0 = inptr0
+    lwx            s3, s4(s2)
+    addiu          s5, a1, -2      // s5 = downsampled_width - 2
+    srl            t4, s5, 1
+    sll            t4, t4, 1
+    lbu            t0, 0(s0)
+    lbu            t1, 1(s0)
+    lbu            t2, 0(s1)
+    lbu            t3, 1(s1)
+    addiu          s0, 2
+    addiu          s1, 2
+    addu           t8, s0, t4      // t8 = end address
+    andi           s5, s5, 1       // s5 = residual
+    sll            t4, t0, 1
+    sll            t6, t1, 1
+    addu           t0, t0, t4      // t0 = (*inptr0++) * 3
+    addu           t1, t1, t6      // t1 = (*inptr0++) * 3
+    addu           t7, t0, t2      // t7 = thiscolsum
+    addu           t6, t1, t3      // t5 = nextcolsum
+    sll            t0, t7, 2       // t0 = thiscolsum * 4
+    subu           t1, t0, t7      // t1 = thiscolsum * 3
+    shra_r.w       t0, t0, 4
+    addiu          t1, 7
+    addu           t1, t1, t6
+    srl            t1, t1, 4
+    sb             t0, 0(s3)
+    sb             t1, 1(s3)
+    beq            t8, s0, 22f     // skip to final iteration if width == 3
+     addiu          s3, 2
+2:
+    lh             t0, 0(s0)       // t0 = A3|A2
+    lh             t2, 0(s1)       // t2 = B3|B2
+    addiu          s0, 2
+    addiu          s1, 2
+    preceu.ph.qbr  t0, t0          // t0 = 0|A3|0|A2
+    preceu.ph.qbr  t2, t2          // t2 = 0|B3|0|B2
+    shll.ph        t1, t0, 1
+    sll            t3, t6, 1
+    addu.ph        t0, t1, t0      // t0 = A3*3|A2*3
+    addu           t3, t3, t6      // t3 = this * 3
+    addu.ph        t0, t0, t2      // t0 = next2|next1
+    addu           t1, t3, t7
+    andi           t7, t0, 0xFFFF  // t7 = next1
+    sll            t2, t7, 1
+    addu           t2, t7, t2      // t2 = next1*3
+    addu           t4, t2, t6
+    srl            t6, t0, 16      // t6 = next2
+    shra_r.w       t1, t1, 4       // t1 = (this*3 + last + 8) >> 4
+    addu           t0, t3, t7
+    addiu          t0, 7
+    srl            t0, t0, 4       // t0 = (this*3 + next1 + 7) >> 4
+    shra_r.w       t4, t4, 4       // t3 = (next1*3 + this + 8) >> 4
+    addu           t2, t2, t6
+    addiu          t2, 7
+    srl            t2, t2, 4       // t2 = (next1*3 + next2 + 7) >> 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    sb             t4, 2(s3)
+    sb             t2, 3(s3)
+    bne            t8, s0, 2b
+     addiu         s3, 4
+22:
+    beqz           s5, 4f
+     addu          t8, s0, s5
+3:
+    lbu            t0, 0(s0)
+    lbu            t2, 0(s1)
+    addiu          s0, 1
+    addiu          s1, 1
+    sll            t3, t6, 1
+    sll            t1, t0, 1
+    addu           t1, t0, t1      // t1 = inptr0 * 3
+    addu           t3, t3, t6      // t3 = thiscolsum * 3
+    addu           t5, t1, t2
+    addu           t1, t3, t7
+    shra_r.w       t1, t1, 4
+    addu           t0, t3, t5
+    addiu          t0, 7
+    srl            t0, t0, 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    addiu          s3, 2
+    move           t7, t6
+    bne            t8, s0, 3b
+     move          t6, t5
+4:
+    sll            t0, t6, 2       // t0 = thiscolsum * 4
+    subu           t1, t0, t6      // t1 = thiscolsum * 3
+    addu           t1, t1, t7
+    addiu          s4, 4
+    shra_r.w       t1, t1, 4
+    addiu          t0, 7
+    srl            t0, t0, 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    addiu          t9, -1
+    addiu          s3, 2
+    bnez           t9, 1b
+     lw            s1, 4(a2)
+    srl            t0, s4, 2
+    subu           t0, a0, t0
+    bgtz           t0, 0b
+     addiu         a2, 4
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j ra
+     nop
+END(jsimd_h2v2_fancy_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_fancy_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - downsampled_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    .set at
+
+    beqz           a0, 3f
+     sll           t0, a0, 2
+    lw             s1, 0(a3)
+    li             s3, 0x10001
+    addu           s0, s1, t0
+0:
+    addiu          t8, a1, -2
+    srl            t9, t8, 2
+    lw             t7, 0(a2)
+    lw             s2, 0(s1)
+    lbu            t0, 0(t7)
+    lbu            t1, 1(t7)   // t1 = inptr[1]
+    sll            t2, t0, 1
+    addu           t2, t2, t0  // t2 = invalue*3
+    addu           t2, t2, t1
+    shra_r.w       t2, t2, 2
+    sb             t0, 0(s2)
+    sb             t2, 1(s2)
+    beqz           t9, 11f
+     addiu         s2, 2
+1:
+    ulw            t0, 0(t7)   // t0 = |P3|P2|P1|P0|
+    ulw            t1, 1(t7)
+    ulh            t2, 4(t7)   // t2 = |0|0|P5|P4|
+    preceu.ph.qbl  t3, t0      // t3 = |0|P3|0|P2|
+    preceu.ph.qbr  t0, t0      // t0 = |0|P1|0|P0|
+    preceu.ph.qbr  t2, t2      // t2 = |0|P5|0|P4|
+    preceu.ph.qbl  t4, t1      // t4 = |0|P4|0|P3|
+    preceu.ph.qbr  t1, t1      // t1 = |0|P2|0|P1|
+    shll.ph        t5, t4, 1
+    shll.ph        t6, t1, 1
+    addu.ph        t5, t5, t4  // t5 = |P4*3|P3*3|
+    addu.ph        t6, t6, t1  // t6 = |P2*3|P1*3|
+    addu.ph        t4, t3, s3
+    addu.ph        t0, t0, s3
+    addu.ph        t4, t4, t5
+    addu.ph        t0, t0, t6
+    shrl.ph        t4, t4, 2   // t4 = |0|P3|0|P2|
+    shrl.ph        t0, t0, 2   // t0 = |0|P1|0|P0|
+    addu.ph        t2, t2, t5
+    addu.ph        t3, t3, t6
+    shra_r.ph      t2, t2, 2   // t2 = |0|P5|0|P4|
+    shra_r.ph      t3, t3, 2   // t3 = |0|P3|0|P2|
+    shll.ph        t2, t2, 8
+    shll.ph        t3, t3, 8
+    or             t2, t4, t2
+    or             t3, t3, t0
+    addiu          t9, -1
+    usw            t3, 0(s2)
+    usw            t2, 4(s2)
+    addiu          s2, 8
+    bgtz           t9, 1b
+     addiu         t7, 4
+11:
+    andi           t8, 3
+    beqz           t8, 22f
+     addiu         t7, 1
+
+2:
+    lbu            t0, 0(t7)
+    addiu          t7, 1
+    sll            t1, t0, 1
+    addu           t2, t0, t1  // t2 = invalue
+    lbu            t3, -2(t7)
+    lbu            t4, 0(t7)
+    addiu          t3, 1
+    addiu          t4, 2
+    addu           t3, t3, t2
+    addu           t4, t4, t2
+    srl            t3, 2
+    srl            t4, 2
+    sb             t3, 0(s2)
+    sb             t4, 1(s2)
+    addiu          t8, -1
+    bgtz           t8, 2b
+     addiu         s2, 2
+
+22:
+    lbu            t0, 0(t7)
+    lbu            t2, -1(t7)
+    sll            t1, t0, 1
+    addu           t1, t1, t0 // t1 = invalue * 3
+    addu           t1, t1, t2
+    addiu          t1, 1
+    srl            t1, t1, 2
+    sb             t1, 0(s2)
+    sb             t0, 1(s2)
+    addiu          s1, 4
+    bne            s1, s0, 0b
+     addiu         a2, 4
+3:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j              ra
+     nop
+END(jsimd_h2v1_fancy_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_downsample_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - cinfo->max_v_samp_factor
+ * a2     - compptr->v_samp_factor
+ * a3     - compptr->width_in_blocks
+ * 16(sp) - input_data
+ * 20(sp) - output_data
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4
+
+    beqz        a2, 7f
+     lw         s1, 44(sp)  // s1 = output_data
+    lw          s0, 40(sp)  // s0 = input_data
+    srl         s2, a0, 2
+    andi        t9, a0, 2
+    srl         t7, t9, 1
+    addu        s2, t7, s2
+    sll         t0, a3, 3   // t0 = width_in_blocks*DCT
+    srl         t7, t0, 1
+    subu        s2, t7, s2
+0:
+    andi        t6, a0, 1   // t6 = temp_index
+    addiu       t6, -1
+    lw          t4, 0(s1)   // t4 = outptr
+    lw          t5, 0(s0)   // t5 = inptr0
+    li          s3, 0       // s3 = bias
+    srl         t7, a0, 1   // t7 = image_width1
+    srl         s4, t7, 2
+    andi        t8, t7, 3
+1:
+    ulhu        t0, 0(t5)
+    ulhu        t1, 2(t5)
+    ulhu        t2, 4(t5)
+    ulhu        t3, 6(t5)
+    raddu.w.qb  t0, t0
+    raddu.w.qb  t1, t1
+    raddu.w.qb  t2, t2
+    raddu.w.qb  t3, t3
+    shra.ph     t0, t0, 1
+    shra_r.ph   t1, t1, 1
+    shra.ph     t2, t2, 1
+    shra_r.ph   t3, t3, 1
+    sb          t0, 0(t4)
+    sb          t1, 1(t4)
+    sb          t2, 2(t4)
+    sb          t3, 3(t4)
+    addiu       s4, -1
+    addiu       t4, 4
+    bgtz        s4, 1b
+     addiu      t5, 8
+    beqz        t8, 3f
+     addu       s4, t4, t8
+2:
+    ulhu        t0, 0(t5)
+    raddu.w.qb  t0, t0
+    addqh.w     t0, t0, s3
+    xori        s3, s3, 1
+    sb          t0, 0(t4)
+    addiu       t4, 1
+    bne         t4, s4, 2b
+     addiu      t5, 2
+3:
+    lbux        t1, t6(t5)
+    sll         t1, 1
+    addqh.w     t2, t1, s3  // t2 = pixval1
+    xori        s3, s3, 1
+    addqh.w     t3, t1, s3  // t3 = pixval2
+    blez        s2, 5f
+     append     t3, t2,  8
+    addu        t5, t4, s2  // t5 = loop_end2
+4:
+    ush         t3, 0(t4)
+    addiu       s2, -1
+    bgtz        s2, 4b
+     addiu      t4,  2
+5:
+    beqz        t9, 6f
+     nop
+    sb          t2, 0(t4)
+6:
+    addiu       s1, 4
+    addiu       a2, -1
+    bnez        a2, 0b
+     addiu      s0, 4
+7:
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4
+
+    j           ra
+    nop
+END(jsimd_h2v1_downsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_downsample_mips_dspr2)
+
+/*
+ * a0     - cinfo->image_width
+ * a1     - cinfo->max_v_samp_factor
+ * a2     - compptr->v_samp_factor
+ * a3     - compptr->width_in_blocks
+ * 16(sp) - input_data
+ * 20(sp) - output_data
+ */
+    .set at
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    beqz         a2, 8f
+     lw          s1, 52(sp)      // s1 = output_data
+    lw           s0, 48(sp)      // s0 = input_data
+
+    andi         t6, a0, 1       // t6 = temp_index
+    addiu        t6, -1
+    srl          t7, a0, 1       // t7 = image_width1
+    srl          s4, t7, 2
+    andi         t8, t7, 3
+    andi         t9, a0, 2
+    srl          s2, a0, 2
+    srl          t7, t9, 1
+    addu         s2, t7, s2
+    sll          t0, a3, 3       // s2 = width_in_blocks*DCT
+    srl          t7, t0, 1
+    subu         s2, t7, s2
+0:
+    lw           t4, 0(s1)       // t4 = outptr
+    lw           t5, 0(s0)       // t5 = inptr0
+    lw           s7, 4(s0)       // s7 = inptr1
+    li           s6, 1           // s6 = bias
+2:
+    ulw          t0, 0(t5)       // t0 = |P3|P2|P1|P0|
+    ulw          t1, 0(s7)       // t1 = |Q3|Q2|Q1|Q0|
+    ulw          t2, 4(t5)
+    ulw          t3, 4(s7)
+    precrq.ph.w  t7, t0, t1      // t2 = |P3|P2|Q3|Q2|
+    ins          t0, t1, 16, 16  // t0 = |Q1|Q0|P1|P0|
+    raddu.w.qb   t1, t7
+    raddu.w.qb   t0, t0
+    shra_r.w     t1, t1, 2
+    addiu        t0, 1
+    srl          t0, 2
+    precrq.ph.w  t7, t2, t3
+    ins          t2, t3, 16, 16
+    raddu.w.qb   t7, t7
+    raddu.w.qb   t2, t2
+    shra_r.w     t7, t7, 2
+    addiu        t2, 1
+    srl          t2, 2
+    sb           t0, 0(t4)
+    sb           t1, 1(t4)
+    sb           t2, 2(t4)
+    sb           t7, 3(t4)
+    addiu        t4, 4
+    addiu        t5, 8
+    addiu        s4, s4, -1
+    bgtz         s4, 2b
+     addiu       s7, 8
+    beqz         t8, 4f
+     addu        t8, t4, t8
+3:
+    ulhu         t0, 0(t5)
+    ulhu         t1, 0(s7)
+    ins          t0, t1, 16, 16
+    raddu.w.qb   t0, t0
+    addu         t0, t0, s6
+    srl          t0, 2
+    xori         s6, s6, 3
+    sb           t0, 0(t4)
+    addiu        t5, 2
+    addiu        t4, 1
+    bne          t8, t4, 3b
+     addiu       s7, 2
+4:
+    lbux         t1, t6(t5)
+    sll          t1, 1
+    lbux         t0, t6(s7)
+    sll          t0, 1
+    addu         t1, t1, t0
+    addu         t3, t1, s6
+    srl          t0, t3, 2       // t2 = pixval1
+    xori         s6, s6, 3
+    addu         t2, t1, s6
+    srl          t1, t2, 2       // t3 = pixval2
+    blez         s2, 6f
+     append      t1, t0, 8
+5:
+    ush          t1, 0(t4)
+    addiu        s2, -1
+    bgtz         s2, 5b
+     addiu       t4, 2
+6:
+    beqz         t9, 7f
+     nop
+    sb           t0, 0(t4)
+7:
+    addiu        s1, 4
+    addiu        a2, -1
+    bnez         a2, 0b
+     addiu       s0, 8
+8:
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j            ra
+     nop
+END(jsimd_h2v2_downsample_mips_dspr2)
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_smooth_downsample_mips_dspr2)
+/*
+ * a0     - input_data
+ * a1     - output_data
+ * a2     - compptr->v_samp_factor
+ * a3     - cinfo->max_v_samp_factor
+ * 16(sp) - cinfo->smoothing_factor
+ * 20(sp) - compptr->width_in_blocks
+ * 24(sp) - cinfo->image_width
+ */
+
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          s7, 52(sp)      // compptr->width_in_blocks
+    lw          s0, 56(sp)      // cinfo->image_width
+    lw          s6, 48(sp)      // cinfo->smoothing_factor
+    sll         s7, 3           // output_cols = width_in_blocks * DCTSIZE
+    sll         v0, s7, 1
+    subu        v0, v0, s0
+    blez        v0, 2f
+    move        v1, zero
+    addiu       t0, a3, 2       // t0 = cinfo->max_v_samp_factor + 2
+0:
+    addiu       t1, a0, -4
+    sll         t2, v1, 2
+    lwx         t1, t2(t1)
+    move        t3, v0
+    addu        t1, t1, s0
+    lbu         t2, -1(t1)
+1:
+    addiu       t3, t3, -1
+    sb          t2, 0(t1)
+    bgtz        t3, 1b
+    addiu       t1, t1, 1
+    addiu       v1, v1, 1
+    bne         v1, t0, 0b
+    nop
+2:
+    li          v0, 80
+    mul         v0, s6, v0
+    li          v1, 16384
+    move        t4, zero
+    move        t5, zero
+    subu        t6, v1, v0      // t6 = 16384 - tmp_smoot_f * 80
+    sll         t7, s6, 4       // t7 = tmp_smoot_f * 16
+3:
+/* Special case for first column: pretend column -1 is same as column 0 */
+    sll         v0, t4, 2
+    lwx         t8, v0(a1)      //  outptr = output_data[outrow]
+    sll         v1, t5, 2
+    addiu       t9, v1, 4
+    addiu       s0, v1, -4
+    addiu       s1, v1, 8
+    lwx         s2, v1(a0)      // inptr0 = input_data[inrow]
+    lwx         t9, t9(a0)      // inptr1 = input_data[inrow+1]
+    lwx         s0, s0(a0)      // above_ptr = input_data[inrow-1]
+    lwx         s1, s1(a0)      // below_ptr = input_data[inrow+2]
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, 0(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, 0(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1,t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, 0(s0)
+    lbu         t0, 0(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1,s3, t7
+    extr_r.w    v0, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    addiu       s1, s1, 2
+    sb          v0, -1(t8)
+    addiu       s4, s7, -2
+    and         s4, s4, 3
+    addu        s5, s4, t8      //end adress
+4:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    sb          t2, -1(t8)
+    bne         s5, t8, 4b
+    addiu       s1, s1, 2
+    addiu       s5, s7, -2
+    subu        s5, s5, s4
+    addu        s5, s5, t8      //end adress
+5:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    lh          v1, 2(t9)
+    addu        t0, t0, v0
+    lh          v0, 2(s2)
+    addu        s3, t0, s3
+    lh          t0, 2(s0)
+    lh          t1, 2(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 4(s2)
+    lbu         t0, 1(t9)
+    lbu         t1, 4(t9)
+    sb          t2, 0(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 1(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 4(s0)
+    addu        t0, t0, v0
+    lbu         v0, 1(s0)
+    addu        s3, t0, s3
+    lbu         t0, 1(s1)
+    lbu         t3, 4(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 4(t9)
+    addu        t0, t0, v0
+    lh          v0, 4(s2)
+    addu        s3, t0, s3
+    lh          t0, 4(s0)
+    lh          t1, 4(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 6(s2)
+    lbu         t0, 3(t9)
+    lbu         t1, 6(t9)
+    sb          t2, 1(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 3(s2)
+    addu        t0, t0,t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 6(s0)
+    addu        t0, t0, v0
+    lbu         v0, 3(s0)
+    addu        s3, t0, s3
+    lbu         t0, 3(s1)
+    lbu         t3, 6(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 6(t9)
+    addu        t0, t0, v0
+    lh          v0, 6(s2)
+    addu        s3, t0, s3
+    lh          t0, 6(s0)
+    lh          t1, 6(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t3, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 8(s2)
+    lbu         t0, 5(t9)
+    lbu         t1, 8(t9)
+    sb          t3, 2(t8)
+    raddu.w.qb  t2, v0
+    lbu         v0, 5(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t2, t6
+    addu        v0, v0, v1
+    lbu         t2, 8(s0)
+    addu        t0, t0, v0
+    lbu         v0, 5(s0)
+    addu        s3, t0, s3
+    lbu         t0, 5(s1)
+    lbu         t3, 8(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    addiu       t8, t8, 4
+    addu        t0, t0, v0
+    addiu       s2, s2, 8
+    addu        s3, t0, s3
+    addiu       t9, t9, 8
+    madd        $ac1, s3, t7
+    extr_r.w    t1, $ac1, 16
+    addiu       s0, s0, 8
+    addiu       s1, s1, 8
+    bne         s5, t8, 5b
+    sb          t1, -1(t8)
+/* Special case for last column */
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 1(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 1(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 1(s0)
+    addu        t0, t0, v0
+    lbu         t3, 1(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t0, $ac1, 16
+    addiu       t5, t5, 2
+    sb          t0, 0(t8)
+    addiu       t4, t4, 1
+    bne         t4, a2, 3b
+    addiu       t5, t5, 2
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_smooth_downsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_int_upsample_mips_dspr2)
+/*
+ * a0     - upsample->h_expand[compptr->component_index]
+ * a1     - upsample->v_expand[compptr->component_index]
+ * a2     - input_data
+ * a3     - output_data_ptr
+ * 16(sp) - cinfo->output_width
+ * 20(sp) - cinfo->max_v_samp_factor
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    lw      s0, 0(a3)    // s0 = output_data
+    lw      s1, 32(sp)   // s1 = cinfo->output_width
+    lw      s2, 36(sp)   // s2 = cinfo->max_v_samp_factor
+    li      t6, 0        // t6 = inrow
+    beqz    s2, 10f
+     li     s3, 0        // s3 = outrow
+0:
+    addu    t0, a2, t6
+    addu    t7, s0, s3
+    lw      t3, 0(t0)    // t3 = inptr
+    lw      t8, 0(t7)    // t8 = outptr
+    beqz    s1, 4f
+     addu   t5, t8, s1   // t5 = outend
+1:
+    lb      t2, 0(t3)    // t2 = invalue = *inptr++
+    addiu   t3, 1
+    beqz    a0, 3f
+     move   t0, a0       // t0 = h_expand
+2:
+    sb      t2, 0(t8)
+    addiu   t0, -1
+    bgtz    t0, 2b
+     addiu  t8, 1
+3:
+    bgt     t5, t8, 1b
+     nop
+4:
+    addiu   t9, a1, -1   // t9 = v_expand - 1
+    blez    t9, 9f
+     nop
+5:
+    lw      t3, 0(s0)
+    lw      t4, 4(s0)
+    subu    t0, s1, 0xF
+    blez    t0, 7f
+     addu   t5, t3, s1   // t5 = end address
+    andi    t7, s1, 0xF  // t7 = residual
+    subu    t8, t5, t7
+6:
+    ulw     t0, 0(t3)
+    ulw     t1, 4(t3)
+    ulw     t2, 8(t3)
+    usw     t0, 0(t4)
+    ulw     t0, 12(t3)
+    usw     t1, 4(t4)
+    usw     t2, 8(t4)
+    usw     t0, 12(t4)
+    addiu   t3, 16
+    bne     t3, t8, 6b
+     addiu  t4, 16
+    beqz    t7, 8f
+     nop
+7:
+    lbu     t0, 0(t3)
+    sb      t0, 0(t4)
+    addiu   t3, 1
+    bne     t3, t5, 7b
+     addiu  t4, 1
+8:
+    addiu   t9, -1
+    bgtz    t9, 5b
+     addiu  s0, 8
+9:
+    addu    s3, s3, a1
+    bne     s3, s2, 0b
+     addiu  t6, 1
+10:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j       ra
+     nop
+END(jsimd_int_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - cinfo->output_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+    lw      t7, 0(a3)       // t7 = output_data
+    andi    t8, a1, 0xf     // t8 = residual
+    sll     t0, a0, 2
+    blez    a0, 4f
+     addu   t9, t7, t0      // t9 = output_data end address
+0:
+    lw      t5, 0(t7)       // t5 = outptr
+    lw      t6, 0(a2)       // t6 = inptr
+    addu    t3, t5, a1      // t3 = outptr + output_width (end address)
+    subu    t3, t8          // t3 = end address - residual
+    beq     t5, t3, 2f
+     move   t4, t8
+1:
+    ulw     t0, 0(t6)       // t0 = |P3|P2|P1|P0|
+    ulw     t2, 4(t6)       // t2 = |P7|P6|P5|P4|
+    srl     t1, t0, 16      // t1 = |X|X|P3|P2|
+    ins     t0, t0, 16, 16  // t0 = |P1|P0|P1|P0|
+    ins     t1, t1, 16, 16  // t1 = |P3|P2|P3|P2|
+    ins     t0, t0, 8, 16   // t0 = |P1|P1|P0|P0|
+    ins     t1, t1, 8, 16   // t1 = |P3|P3|P2|P2|
+    usw     t0, 0(t5)
+    usw     t1, 4(t5)
+    srl     t0, t2, 16      // t0 = |X|X|P7|P6|
+    ins     t2, t2, 16, 16  // t2 = |P5|P4|P5|P4|
+    ins     t0, t0, 16, 16  // t0 = |P7|P6|P7|P6|
+    ins     t2, t2, 8, 16   // t2 = |P5|P5|P4|P4|
+    ins     t0, t0, 8, 16   // t0 = |P7|P7|P6|P6|
+    usw     t2, 8(t5)
+    usw     t0, 12(t5)
+    addiu   t5, 16
+    bne     t5, t3, 1b
+     addiu  t6, 8
+    beqz    t8, 3f
+     move   t4, t8
+2:
+    lbu     t1, 0(t6)
+    sb      t1, 0(t5)
+    sb      t1, 1(t5)
+    addiu   t4, -2
+    addiu   t6, 1
+    bgtz    t4, 2b
+     addiu  t5, 2
+3:
+    addiu   t7, 4
+    bne     t9, t7, 0b
+     addiu  a2, 4
+4:
+    j       ra
+     nop
+END(jsimd_h2v1_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - cinfo->output_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+    lw      t7, 0(a3)
+    blez    a0, 7f
+     andi   t9, a1, 0xf     // t9 = residual
+0:
+    lw      t6, 0(a2)       // t6 = inptr
+    lw      t5, 0(t7)       // t5 = outptr
+    addu    t8, t5, a1      // t8 = outptr end address
+    subu    t8, t9          // t8 = end address - residual
+    beq     t5, t8, 2f
+     move   t4, t9
+1:
+    ulw     t0, 0(t6)
+    srl     t1, t0, 16
+    ins     t0, t0, 16, 16
+    ins     t0, t0, 8, 16
+    ins     t1, t1, 16, 16
+    ins     t1, t1, 8, 16
+    ulw     t2, 4(t6)
+    usw     t0, 0(t5)
+    usw     t1, 4(t5)
+    srl     t3, t2, 16
+    ins     t2, t2, 16, 16
+    ins     t2, t2, 8, 16
+    ins     t3, t3, 16, 16
+    ins     t3, t3, 8, 16
+    usw     t2, 8(t5)
+    usw     t3, 12(t5)
+    addiu   t5, 16
+    bne     t5, t8, 1b
+     addiu  t6, 8
+    beqz    t9, 3f
+     move   t4, t9
+2:
+    lbu     t0, 0(t6)
+    sb      t0, 0(t5)
+    sb      t0, 1(t5)
+    addiu   t4, -2
+    addiu   t6, 1
+    bgtz    t4, 2b
+     addiu  t5, 2
+3:
+    lw      t6, 0(t7)       // t6 = outptr[0]
+    lw      t5, 4(t7)       // t5 = outptr[1]
+    addu    t4, t6, a1      // t4 = new end address
+    beq     a1, t9, 5f
+     subu   t8, t4, t9
+4:
+    ulw     t0, 0(t6)
+    ulw     t1, 4(t6)
+    ulw     t2, 8(t6)
+    usw     t0, 0(t5)
+    ulw     t0, 12(t6)
+    usw     t1, 4(t5)
+    usw     t2, 8(t5)
+    usw     t0, 12(t5)
+    addiu   t6, 16
+    bne     t6, t8, 4b
+     addiu  t5, 16
+    beqz    t9, 6f
+     nop
+5:
+    lbu     t0, 0(t6)
+    sb      t0, 0(t5)
+    addiu   t6, 1
+    bne     t6, t4, 5b
+     addiu  t5, 1
+6:
+    addiu   t7, 8
+    addiu   a0, -2
+    bgtz    a0, 0b
+     addiu  a2, 4
+7:
+    j       ra
+     nop
+END(jsimd_h2v2_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_islow_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - compptr->dcttable
+ * a2     - output
+ * a3     - range_limit
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu     sp, sp, -256
+    move      v0, sp
+    addiu     v1, zero, 8      // v1 = DCTSIZE = 8
+1:
+    lh        s4, 32(a0)       // s4 = inptr[16]
+    lh        s5, 64(a0)       // s5 = inptr[32]
+    lh        s6, 96(a0)       // s6 = inptr[48]
+    lh        t1, 112(a0)      // t1 = inptr[56]
+    lh        t7, 16(a0)       // t7 = inptr[8]
+    lh        t5, 80(a0)       // t5 = inptr[40]
+    lh        t3, 48(a0)       // t3 = inptr[24]
+    or        s4, s4, t1
+    or        s4, s4, t3
+    or        s4, s4, t5
+    or        s4, s4, t7
+    or        s4, s4, s5
+    or        s4, s4, s6
+    bnez      s4, 2f
+     addiu    v1, v1, -1
+    lh        s5, 0(a1)        // quantptr[DCTSIZE*0]
+    lh        s6, 0(a0)        // inptr[DCTSIZE*0]
+    mul       s5, s5, s6       // DEQUANTIZE(inptr[0], quantptr[0])
+    sll       s5, s5, 2
+    sw        s5, 0(v0)
+    sw        s5, 32(v0)
+    sw        s5, 64(v0)
+    sw        s5, 96(v0)
+    sw        s5, 128(v0)
+    sw        s5, 160(v0)
+    sw        s5, 192(v0)
+    b         3f
+     sw       s5, 224(v0)
+2:
+    lh        t0, 112(a1)
+    lh        t2, 48(a1)
+    lh        t4, 80(a1)
+    lh        t6, 16(a1)
+    mul       t0, t0, t1       // DEQUANTIZE(inptr[DCTSIZE*7],quant[DCTSIZE*7])
+    mul       t1, t2, t3       // DEQUANTIZE(inptr[DCTSIZE*3],quant[DCTSIZE*3])
+    mul       t2, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*5],quant[DCTSIZE*5])
+    mul       t3, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*1],quant[DCTSIZE*1])
+    lh        t4, 32(a1)
+    lh        t5, 32(a0)
+    lh        t6, 96(a1)
+    lh        t7, 96(a0)
+    addu      s0, t0, t1       // z3 = tmp0 + tmp2
+    addu      s1, t1, t2       // z2 = tmp1 + tmp2
+    addu      s2, t2, t3       // z4 = tmp1 + tmp3
+    addu      s3, s0, s2       // z3 + z4
+    addiu     t9, zero, 9633   // FIX_1_175875602
+    mul       s3, s3, t9       // z5 = MULTIPLY(z3 + z4, FIX_1_175875602)
+    addu      t8, t0, t3       // z1 = tmp0 + tmp3
+    addiu     t9, zero, 2446   // FIX_0_298631336
+    mul       t0, t0, t9       // tmp0 = MULTIPLY(tmp0, FIX_0_298631336)
+    addiu     t9, zero, 16819  // FIX_2_053119869
+    mul       t2, t2, t9       // tmp1 = MULTIPLY(tmp1, FIX_2_053119869)
+    addiu     t9, zero, 25172  // FIX_3_072711026
+    mul       t1, t1, t9       // tmp2 = MULTIPLY(tmp2, FIX_3_072711026)
+    addiu     t9, zero, 12299  // FIX_1_501321110
+    mul       t3, t3, t9       // tmp3 = MULTIPLY(tmp3, FIX_1_501321110)
+    addiu     t9, zero, 16069  // FIX_1_961570560
+    mul       s0, s0, t9       // -z3 = MULTIPLY(z3, FIX_1_961570560)
+    addiu     t9, zero, 3196   // FIX_0_390180644
+    mul       s2, s2, t9       // -z4 = MULTIPLY(z4, FIX_0_390180644)
+    addiu     t9, zero, 7373   // FIX_0_899976223
+    mul       t8, t8, t9       // -z1 = MULTIPLY(z1, FIX_0_899976223)
+    addiu     t9, zero, 20995  // FIX_2_562915447
+    mul       s1, s1, t9       // -z2 = MULTIPLY(z2, FIX_2_562915447)
+    subu      s0, s3, s0       // z3 += z5
+    addu      t0, t0, s0       // tmp0 += z3
+    addu      t1, t1, s0       // tmp2 += z3
+    subu      s2, s3, s2       // z4 += z5
+    addu      t2, t2, s2       // tmp1 += z4
+    addu      t3, t3, s2       // tmp3 += z4
+    subu      t0, t0, t8       // tmp0 += z1
+    subu      t1, t1, s1       // tmp2 += z2
+    subu      t2, t2, s1       // tmp1 += z2
+    subu      t3, t3, t8       // tmp3 += z1
+    mul       s0, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*2],quant[DCTSIZE*2])
+    addiu     t9, zero, 6270   // FIX_0_765366865
+    mul       s1, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*6],quant[DCTSIZE*6])
+    lh        t4, 0(a1)
+    lh        t5, 0(a0)
+    lh        t6, 64(a1)
+    lh        t7, 64(a0)
+    mul       s2, t9, s0       // MULTIPLY(z2, FIX_0_765366865)
+    mul       t5, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*0],quant[DCTSIZE*0])
+    mul       t6, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*4],quant[DCTSIZE*4])
+    addiu     t9, zero, 4433   // FIX_0_541196100
+    addu      s3, s0, s1       // z2 + z3
+    mul       s3, s3, t9       // z1 = MULTIPLY(z2 + z3, FIX_0_541196100)
+    addiu     t9, zero, 15137  // FIX_1_847759065
+    mul       t8, s1, t9       // MULTIPLY(z3, FIX_1_847759065)
+    addu      t4, t5, t6
+    subu      t5, t5, t6
+    sll       t4, t4, 13       // tmp0 = (z2 + z3) << CONST_BITS
+    sll       t5, t5, 13       // tmp1 = (z2 - z3) << CONST_BITS
+    addu      t7, s3, s2       // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865)
+    subu      t6, s3, t8       // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065)
+    addu      s0, t4, t7
+    subu      s1, t4, t7
+    addu      s2, t5, t6
+    subu      s3, t5, t6
+    addu      t4, s0, t3
+    subu      s0, s0, t3
+    addu      t3, s2, t1
+    subu      s2, s2, t1
+    addu      t1, s3, t2
+    subu      s3, s3, t2
+    addu      t2, s1, t0
+    subu      s1, s1, t0
+    shra_r.w  t4, t4, 11
+    shra_r.w  t3, t3, 11
+    shra_r.w  t1, t1, 11
+    shra_r.w  t2, t2, 11
+    shra_r.w  s1, s1, 11
+    shra_r.w  s3, s3, 11
+    shra_r.w  s2, s2, 11
+    shra_r.w  s0, s0, 11
+    sw        t4, 0(v0)
+    sw        t3, 32(v0)
+    sw        t1, 64(v0)
+    sw        t2, 96(v0)
+    sw        s1, 128(v0)
+    sw        s3, 160(v0)
+    sw        s2, 192(v0)
+    sw        s0, 224(v0)
+3:
+    addiu     a1, a1, 2
+    addiu     a0, a0, 2
+    bgtz      v1, 1b
+     addiu    v0, v0, 4
+    move      v0, sp
+    addiu     v1, zero, 8
+4:
+    lw        t0, 8(v0)        // z2 = (JLONG) wsptr[2]
+    lw        t1, 24(v0)       // z3 = (JLONG) wsptr[6]
+    lw        t2, 0(v0)        // (JLONG) wsptr[0]
+    lw        t3, 16(v0)       // (JLONG) wsptr[4]
+    lw        s4, 4(v0)        // (JLONG) wsptr[1]
+    lw        s5, 12(v0)       // (JLONG) wsptr[3]
+    lw        s6, 20(v0)       // (JLONG) wsptr[5]
+    lw        s7, 28(v0)       // (JLONG) wsptr[7]
+    or        s4, s4, t0
+    or        s4, s4, t1
+    or        s4, s4, t3
+    or        s4, s4, s7
+    or        s4, s4, s5
+    or        s4, s4, s6
+    bnez      s4, 5f
+     addiu    v1, v1, -1
+    shra_r.w  s5, t2, 5
+    andi      s5, s5, 0x3ff
+    lbux      s5, s5(a3)
+    lw        s1, 0(a2)
+    replv.qb  s5, s5
+    usw       s5, 0(s1)
+    usw       s5, 4(s1)
+    b         6f
+     nop
+5:
+    addu      t4, t0, t1       // z2 + z3
+    addiu     t8, zero, 4433   // FIX_0_541196100
+    mul       t5, t4, t8       // z1 = MULTIPLY(z2 + z3, FIX_0_541196100)
+    addiu     t8, zero, 15137  // FIX_1_847759065
+    mul       t1, t1, t8       // MULTIPLY(z3, FIX_1_847759065)
+    addiu     t8, zero, 6270   // FIX_0_765366865
+    mul       t0, t0, t8       // MULTIPLY(z2, FIX_0_765366865)
+    addu      t4, t2, t3       // (JLONG) wsptr[0] + (JLONG) wsptr[4]
+    subu      t2, t2, t3       // (JLONG) wsptr[0] - (JLONG) wsptr[4]
+    sll       t4, t4, 13       // tmp0 = ((wsptr[0] + wsptr[4]) << CONST_BITS
+    sll       t2, t2, 13       // tmp1 = ((wsptr[0] - wsptr[4]) << CONST_BITS
+    subu      t1, t5, t1       // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065)
+    subu      t3, t2, t1       // tmp12 = tmp1 - tmp2
+    addu      t2, t2, t1       // tmp11 = tmp1 + tmp2
+    addu      t5, t5, t0       // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865)
+    subu      t1, t4, t5       // tmp13 = tmp0 - tmp3
+    addu      t0, t4, t5       // tmp10 = tmp0 + tmp3
+    lw        t4, 28(v0)       // tmp0 = (JLONG) wsptr[7]
+    lw        t6, 12(v0)       // tmp2 = (JLONG) wsptr[3]
+    lw        t5, 20(v0)       // tmp1 = (JLONG) wsptr[5]
+    lw        t7, 4(v0)        // tmp3 = (JLONG) wsptr[1]
+    addu      s0, t4, t6       // z3 = tmp0 + tmp2
+    addiu     t8, zero, 9633   // FIX_1_175875602
+    addu      s1, t5, t7       // z4 = tmp1 + tmp3
+    addu      s2, s0, s1       // z3 + z4
+    mul       s2, s2, t8       // z5 = MULTIPLY(z3 + z4, FIX_1_175875602)
+    addu      s3, t4, t7       // z1 = tmp0 + tmp3
+    addu      t9, t5, t6       // z2 = tmp1 + tmp2
+    addiu     t8, zero, 16069  // FIX_1_961570560
+    mul       s0, s0, t8       // -z3 = MULTIPLY(z3, FIX_1_961570560)
+    addiu     t8, zero, 3196   // FIX_0_390180644
+    mul       s1, s1, t8       // -z4 = MULTIPLY(z4, FIX_0_390180644)
+    addiu     t8, zero, 2446   // FIX_0_298631336
+    mul       t4, t4, t8       // tmp0 = MULTIPLY(tmp0, FIX_0_298631336)
+    addiu     t8, zero, 7373   // FIX_0_899976223
+    mul       s3, s3, t8       // -z1 = MULTIPLY(z1, FIX_0_899976223)
+    addiu     t8, zero, 16819  // FIX_2_053119869
+    mul       t5, t5, t8       // tmp1 = MULTIPLY(tmp1, FIX_2_053119869)
+    addiu     t8, zero, 20995  // FIX_2_562915447
+    mul       t9, t9, t8       // -z2 = MULTIPLY(z2, FIX_2_562915447)
+    addiu     t8, zero, 25172  // FIX_3_072711026
+    mul       t6, t6, t8       // tmp2 = MULTIPLY(tmp2, FIX_3_072711026)
+    addiu     t8, zero, 12299  // FIX_1_501321110
+    mul       t7, t7, t8       // tmp3 = MULTIPLY(tmp3, FIX_1_501321110)
+    subu      s0, s2, s0       // z3 += z5
+    subu      s1, s2, s1       // z4 += z5
+    addu      t4, t4, s0
+    subu      t4, t4, s3       // tmp0
+    addu      t5, t5, s1
+    subu      t5, t5, t9       // tmp1
+    addu      t6, t6, s0
+    subu      t6, t6, t9       // tmp2
+    addu      t7, t7, s1
+    subu      t7, t7, s3       // tmp3
+    addu      s0, t0, t7
+    subu      t0, t0, t7
+    addu      t7, t2, t6
+    subu      t2, t2, t6
+    addu      t6, t3, t5
+    subu      t3, t3, t5
+    addu      t5, t1, t4
+    subu      t1, t1, t4
+    shra_r.w  s0, s0, 18
+    shra_r.w  t7, t7, 18
+    shra_r.w  t6, t6, 18
+    shra_r.w  t5, t5, 18
+    shra_r.w  t1, t1, 18
+    shra_r.w  t3, t3, 18
+    shra_r.w  t2, t2, 18
+    shra_r.w  t0, t0, 18
+    andi      s0, s0, 0x3ff
+    andi      t7, t7, 0x3ff
+    andi      t6, t6, 0x3ff
+    andi      t5, t5, 0x3ff
+    andi      t1, t1, 0x3ff
+    andi      t3, t3, 0x3ff
+    andi      t2, t2, 0x3ff
+    andi      t0, t0, 0x3ff
+    lw        s1, 0(a2)
+    lbux      s0, s0(a3)
+    lbux      t7, t7(a3)
+    lbux      t6, t6(a3)
+    lbux      t5, t5(a3)
+    lbux      t1, t1(a3)
+    lbux      t3, t3(a3)
+    lbux      t2, t2(a3)
+    lbux      t0, t0(a3)
+    sb        s0, 0(s1)
+    sb        t7, 1(s1)
+    sb        t6, 2(s1)
+    sb        t5, 3(s1)
+    sb        t1, 4(s1)
+    sb        t3, 5(s1)
+    sb        t2, 6(s1)
+    sb        t0, 7(s1)
+6:
+    addiu     v0, v0, 32
+    bgtz      v1, 4b
+     addiu    a2, a2, 4
+    addiu     sp, sp, 256
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+
+END(jsimd_idct_islow_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_ifast_cols_mips_dspr2)
+/*
+ * a0     - inptr
+ * a1     - quantptr
+ * a2     - wsptr
+ * a3     - mips_idct_ifast_coefs
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu          t9, a0, 16            // end address
+    or             AT, a3, zero
+
+0:
+    lw             s0, 0(a1)             // quantptr[DCTSIZE*0]
+    lw             t0, 0(a0)             // inptr[DCTSIZE*0]
+    lw             t1, 16(a0)            // inptr[DCTSIZE*1]
+    muleq_s.w.phl  v0, t0, s0            // tmp0 ...
+    lw             t2, 32(a0)            // inptr[DCTSIZE*2]
+    lw             t3, 48(a0)            // inptr[DCTSIZE*3]
+    lw             t4, 64(a0)            // inptr[DCTSIZE*4]
+    lw             t5, 80(a0)            // inptr[DCTSIZE*5]
+    muleq_s.w.phr  t0, t0, s0            // ... tmp0 ...
+    lw             t6, 96(a0)            // inptr[DCTSIZE*6]
+    lw             t7, 112(a0)           // inptr[DCTSIZE*7]
+    or             s4, t1, t2
+    or             s5, t3, t4
+    bnez           s4, 1f
+     ins           t0, v0, 16, 16        // ... tmp0
+    bnez           s5, 1f
+     or            s6, t5, t6
+    or             s6, s6, t7
+    bnez           s6, 1f
+     sw            t0, 0(a2)             // wsptr[DCTSIZE*0]
+    sw             t0, 16(a2)            // wsptr[DCTSIZE*1]
+    sw             t0, 32(a2)            // wsptr[DCTSIZE*2]
+    sw             t0, 48(a2)            // wsptr[DCTSIZE*3]
+    sw             t0, 64(a2)            // wsptr[DCTSIZE*4]
+    sw             t0, 80(a2)            // wsptr[DCTSIZE*5]
+    sw             t0, 96(a2)            // wsptr[DCTSIZE*6]
+    sw             t0, 112(a2)           // wsptr[DCTSIZE*7]
+    addiu          a0, a0, 4
+    b              2f
+     addiu         a1, a1, 4
+
+1:
+    lw             s1, 32(a1)            // quantptr[DCTSIZE*2]
+    lw             s2, 64(a1)            // quantptr[DCTSIZE*4]
+    muleq_s.w.phl  v0, t2, s1            // tmp1 ...
+    muleq_s.w.phr  t2, t2, s1            // ... tmp1 ...
+    lw             s0, 16(a1)            // quantptr[DCTSIZE*1]
+    lw             s1, 48(a1)            // quantptr[DCTSIZE*3]
+    lw             s3, 96(a1)            // quantptr[DCTSIZE*6]
+    muleq_s.w.phl  v1, t4, s2            // tmp2 ...
+    muleq_s.w.phr  t4, t4, s2            // ... tmp2 ...
+    lw             s2, 80(a1)            // quantptr[DCTSIZE*5]
+    lw             t8, 4(AT)             // FIX(1.414213562)
+    ins            t2, v0, 16, 16        // ... tmp1
+    muleq_s.w.phl  v0, t6, s3            // tmp3 ...
+    muleq_s.w.phr  t6, t6, s3            // ... tmp3 ...
+    ins            t4, v1, 16, 16        // ... tmp2
+    addq.ph        s4, t0, t4            // tmp10
+    subq.ph        s5, t0, t4            // tmp11
+    ins            t6, v0, 16, 16        // ... tmp3
+    subq.ph        s6, t2, t6            // tmp12 ...
+    addq.ph        s7, t2, t6            // tmp13
+    mulq_s.ph      s6, s6, t8            // ... tmp12 ...
+    addq.ph        t0, s4, s7            // tmp0
+    subq.ph        t6, s4, s7            // tmp3
+    muleq_s.w.phl  v0, t1, s0            // tmp4 ...
+    muleq_s.w.phr  t1, t1, s0            // ... tmp4 ...
+    shll_s.ph      s6, s6, 1             // x2
+    lw             s3, 112(a1)           // quantptr[DCTSIZE*7]
+    subq.ph        s6, s6, s7            // ... tmp12
+    muleq_s.w.phl  v1, t7, s3            // tmp7 ...
+    muleq_s.w.phr  t7, t7, s3            // ... tmp7 ...
+    ins            t1, v0, 16, 16        // ... tmp4
+    addq.ph        t2, s5, s6            // tmp1
+    subq.ph        t4, s5, s6            // tmp2
+    muleq_s.w.phl  v0, t5, s2            // tmp6 ...
+    muleq_s.w.phr  t5, t5, s2            // ... tmp6 ...
+    ins            t7, v1, 16, 16        // ... tmp7
+    addq.ph        s5, t1, t7            // z11
+    subq.ph        s6, t1, t7            // z12
+    muleq_s.w.phl  v1, t3, s1            // tmp5 ...
+    muleq_s.w.phr  t3, t3, s1            // ... tmp5 ...
+    ins            t5, v0, 16, 16        // ... tmp6
+    ins            t3, v1, 16, 16        // ... tmp5
+    addq.ph        s7, t5, t3            // z13
+    subq.ph        v0, t5, t3            // z10
+    addq.ph        t7, s5, s7            // tmp7
+    subq.ph        s5, s5, s7            // tmp11 ...
+    addq.ph        v1, v0, s6            // z5 ...
+    mulq_s.ph      s5, s5, t8            // ... tmp11
+    lw             t8, 8(AT)             // FIX(1.847759065)
+    lw             s4, 0(AT)             // FIX(1.082392200)
+    addq.ph        s0, t0, t7
+    subq.ph        s1, t0, t7
+    mulq_s.ph      v1, v1, t8            // ... z5
+    shll_s.ph      s5, s5, 1             // x2
+    lw             t8, 12(AT)            // FIX(-2.613125930)
+    sw             s0, 0(a2)             // wsptr[DCTSIZE*0]
+    shll_s.ph      v0, v0, 1             // x4
+    mulq_s.ph      v0, v0, t8            // tmp12 ...
+    mulq_s.ph      s4, s6, s4            // tmp10 ...
+    shll_s.ph      v1, v1, 1             // x2
+    addiu          a0, a0, 4
+    addiu          a1, a1, 4
+    sw             s1, 112(a2)           // wsptr[DCTSIZE*7]
+    shll_s.ph      s6, v0, 1             // x4
+    shll_s.ph      s4, s4, 1             // x2
+    addq.ph        s6, s6, v1            // ... tmp12
+    subq.ph        t5, s6, t7            // tmp6
+    subq.ph        s4, s4, v1            // ... tmp10
+    subq.ph        t3, s5, t5            // tmp5
+    addq.ph        s2, t2, t5
+    addq.ph        t1, s4, t3            // tmp4
+    subq.ph        s3, t2, t5
+    sw             s2, 16(a2)            // wsptr[DCTSIZE*1]
+    sw             s3, 96(a2)            // wsptr[DCTSIZE*6]
+    addq.ph        v0, t4, t3
+    subq.ph        v1, t4, t3
+    sw             v0, 32(a2)            // wsptr[DCTSIZE*2]
+    sw             v1, 80(a2)            // wsptr[DCTSIZE*5]
+    addq.ph        v0, t6, t1
+    subq.ph        v1, t6, t1
+    sw             v0, 64(a2)            // wsptr[DCTSIZE*4]
+    sw             v1, 48(a2)            // wsptr[DCTSIZE*3]
+
+2:
+    bne            a0, t9, 0b
+     addiu         a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j              ra
+     nop
+
+END(jsimd_idct_ifast_cols_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_ifast_rows_mips_dspr2)
+/*
+ * a0     - wsptr
+ * a1     - output_buf
+ * a2     - output_col
+ * a3     - mips_idct_ifast_coefs
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    addiu         t9, a0, 128        // end address
+    lui           s8, 0x8080
+    ori           s8, s8, 0x8080
+
+0:
+    lw            AT, 36(sp)         // restore $a3 (mips_idct_ifast_coefs)
+    lw            t0, 0(a0)          // wsptr[DCTSIZE*0+0/1]  b a
+    lw            s0, 16(a0)         // wsptr[DCTSIZE*1+0/1]  B A
+    lw            t2, 4(a0)          // wsptr[DCTSIZE*0+2/3]  d c
+    lw            s2, 20(a0)         // wsptr[DCTSIZE*1+2/3]  D C
+    lw            t4, 8(a0)          // wsptr[DCTSIZE*0+4/5]  f e
+    lw            s4, 24(a0)         // wsptr[DCTSIZE*1+4/5]  F E
+    lw            t6, 12(a0)         // wsptr[DCTSIZE*0+6/7]  h g
+    lw            s6, 28(a0)         // wsptr[DCTSIZE*1+6/7]  H G
+    precrq.ph.w   t1, s0, t0         // B b
+    ins           t0, s0, 16, 16     // A a
+    bnez          t1, 1f
+     or           s0, t2, s2
+    bnez          s0, 1f
+     or           s0, t4, s4
+    bnez          s0, 1f
+     or           s0, t6, s6
+    bnez          s0, 1f
+     shll_s.ph    s0, t0, 2          // A a
+    lw            a3, 0(a1)
+    lw            AT, 4(a1)
+    precrq.ph.w   t0, s0, s0         // A A
+    ins           s0, s0, 16, 16     // a a
+    addu          a3, a3, a2
+    addu          AT, AT, a2
+    precrq.qb.ph  t0, t0, t0         // A A A A
+    precrq.qb.ph  s0, s0, s0         // a a a a
+    addu.qb       s0, s0, s8
+    addu.qb       t0, t0, s8
+    sw            s0, 0(a3)
+    sw            s0, 4(a3)
+    sw            t0, 0(AT)
+    sw            t0, 4(AT)
+    addiu         a0, a0, 32
+    bne           a0, t9, 0b
+     addiu        a1, a1, 8
+    b             2f
+     nop
+
+1:
+    precrq.ph.w   t3, s2, t2
+    ins           t2, s2, 16, 16
+    precrq.ph.w   t5, s4, t4
+    ins           t4, s4, 16, 16
+    precrq.ph.w   t7, s6, t6
+    ins           t6, s6, 16, 16
+    lw            t8, 4(AT)          // FIX(1.414213562)
+    addq.ph       s4, t0, t4         // tmp10
+    subq.ph       s5, t0, t4         // tmp11
+    subq.ph       s6, t2, t6         // tmp12 ...
+    addq.ph       s7, t2, t6         // tmp13
+    mulq_s.ph     s6, s6, t8         // ... tmp12 ...
+    addq.ph       t0, s4, s7         // tmp0
+    subq.ph       t6, s4, s7         // tmp3
+    shll_s.ph     s6, s6, 1          // x2
+    subq.ph       s6, s6, s7         // ... tmp12
+    addq.ph       t2, s5, s6         // tmp1
+    subq.ph       t4, s5, s6         // tmp2
+    addq.ph       s5, t1, t7         // z11
+    subq.ph       s6, t1, t7         // z12
+    addq.ph       s7, t5, t3         // z13
+    subq.ph       v0, t5, t3         // z10
+    addq.ph       t7, s5, s7         // tmp7
+    subq.ph       s5, s5, s7         // tmp11 ...
+    addq.ph       v1, v0, s6         // z5 ...
+    mulq_s.ph     s5, s5, t8         // ... tmp11
+    lw            t8, 8(AT)          // FIX(1.847759065)
+    lw            s4, 0(AT)          // FIX(1.082392200)
+    addq.ph       s0, t0, t7         // tmp0 + tmp7
+    subq.ph       s7, t0, t7         // tmp0 - tmp7
+    mulq_s.ph     v1, v1, t8         // ... z5
+    lw            a3, 0(a1)
+    lw            t8, 12(AT)         // FIX(-2.613125930)
+    shll_s.ph     s5, s5, 1          // x2
+    addu          a3, a3, a2
+    shll_s.ph     v0, v0, 1          // x4
+    mulq_s.ph     v0, v0, t8         // tmp12 ...
+    mulq_s.ph     s4, s6, s4         // tmp10 ...
+    shll_s.ph     v1, v1, 1          // x2
+    addiu         a0, a0, 32
+    addiu         a1, a1, 8
+    shll_s.ph     s6, v0, 1          // x4
+    shll_s.ph     s4, s4, 1          // x2
+    addq.ph       s6, s6, v1         // ... tmp12
+    shll_s.ph     s0, s0, 2
+    subq.ph       t5, s6, t7         // tmp6
+    subq.ph       s4, s4, v1         // ... tmp10
+    subq.ph       t3, s5, t5         // tmp5
+    shll_s.ph     s7, s7, 2
+    addq.ph       t1, s4, t3         // tmp4
+    addq.ph       s1, t2, t5         // tmp1 + tmp6
+    subq.ph       s6, t2, t5         // tmp1 - tmp6
+    addq.ph       s2, t4, t3         // tmp2 + tmp5
+    subq.ph       s5, t4, t3         // tmp2 - tmp5
+    addq.ph       s4, t6, t1         // tmp3 + tmp4
+    subq.ph       s3, t6, t1         // tmp3 - tmp4
+    shll_s.ph     s1, s1, 2
+    shll_s.ph     s2, s2, 2
+    shll_s.ph     s3, s3, 2
+    shll_s.ph     s4, s4, 2
+    shll_s.ph     s5, s5, 2
+    shll_s.ph     s6, s6, 2
+    precrq.ph.w   t0, s1, s0         // B A
+    ins           s0, s1, 16, 16     // b a
+    precrq.ph.w   t2, s3, s2         // D C
+    ins           s2, s3, 16, 16     // d c
+    precrq.ph.w   t4, s5, s4         // F E
+    ins           s4, s5, 16, 16     // f e
+    precrq.ph.w   t6, s7, s6         // H G
+    ins           s6, s7, 16, 16     // h g
+    precrq.qb.ph  t0, t2, t0         // D C B A
+    precrq.qb.ph  s0, s2, s0         // d c b a
+    precrq.qb.ph  t4, t6, t4         // H G F E
+    precrq.qb.ph  s4, s6, s4         // h g f e
+    addu.qb       s0, s0, s8
+    addu.qb       s4, s4, s8
+    sw            s0, 0(a3)          // outptr[0/1/2/3]       d c b a
+    sw            s4, 4(a3)          // outptr[4/5/6/7]       h g f e
+    lw            a3, -4(a1)
+    addu.qb       t0, t0, s8
+    addu          a3, a3, a2
+    addu.qb       t4, t4, s8
+    sw            t0, 0(a3)          // outptr[0/1/2/3]       D C B A
+    bne           a0, t9, 0b
+     sw           t4, 4(a3)          // outptr[4/5/6/7]       H G F E
+
+2:
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    j             ra
+     nop
+
+END(jsimd_idct_ifast_rows_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_fdct_islow_mips_dspr2)
+/*
+ * a0     - data
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    lui       t0, 6437
+    ori       t0, 2260
+    lui       t1, 9633
+    ori       t1, 11363
+    lui       t2, 0xd39e
+    ori       t2, 0xe6dc
+    lui       t3, 0xf72d
+    ori       t3, 9633
+    lui       t4, 2261
+    ori       t4, 9633
+    lui       t5, 0xd39e
+    ori       t5, 6437
+    lui       t6, 9633
+    ori       t6, 0xd39d
+    lui       t7, 0xe6dc
+    ori       t7, 2260
+    lui       t8, 4433
+    ori       t8, 10703
+    lui       t9, 0xd630
+    ori       t9, 4433
+    li        s8, 8
+    move      a1, a0
+1:
+    lw        s0, 0(a1)     // tmp0 = 1|0
+    lw        s1, 4(a1)     // tmp1 = 3|2
+    lw        s2, 8(a1)     // tmp2 = 5|4
+    lw        s3, 12(a1)    // tmp3 = 7|6
+    packrl.ph s1, s1, s1    // tmp1 = 2|3
+    packrl.ph s3, s3, s3    // tmp3 = 6|7
+    subq.ph   s7, s1, s2    // tmp7 = 2-5|3-4 = t5|t4
+    subq.ph   s5, s0, s3    // tmp5 = 1-6|0-7 = t6|t7
+    mult      $0, $0        // ac0  = 0
+    dpa.w.ph  $ac0, s7, t0  // ac0 += t5*  6437 + t4*  2260
+    dpa.w.ph  $ac0, s5, t1  // ac0 += t6*  9633 + t7* 11363
+    mult      $ac1, $0, $0  // ac1  = 0
+    dpa.w.ph  $ac1, s7, t2  // ac1 += t5*-11362 + t4* -6436
+    dpa.w.ph  $ac1, s5, t3  // ac1 += t6* -2259 + t7*  9633
+    mult      $ac2, $0, $0  // ac2  = 0
+    dpa.w.ph  $ac2, s7, t4  // ac2 += t5*  2261 + t4*  9633
+    dpa.w.ph  $ac2, s5, t5  // ac2 += t6*-11362 + t7*  6437
+    mult      $ac3, $0, $0  // ac3  = 0
+    dpa.w.ph  $ac3, s7, t6  // ac3 += t5*  9633 + t4*-11363
+    dpa.w.ph  $ac3, s5, t7  // ac3 += t6* -6436 + t7*  2260
+    addq.ph   s6, s1, s2    // tmp6 = 2+5|3+4 = t2|t3
+    addq.ph   s4, s0, s3    // tmp4 = 1+6|0+7 = t1|t0
+    extr_r.w  s0, $ac0, 11  // tmp0 = (ac0 + 1024) >> 11
+    extr_r.w  s1, $ac1, 11  // tmp1 = (ac1 + 1024) >> 11
+    extr_r.w  s2, $ac2, 11  // tmp2 = (ac2 + 1024) >> 11
+    extr_r.w  s3, $ac3, 11  // tmp3 = (ac3 + 1024) >> 11
+    addq.ph   s5, s4, s6    // tmp5 = t1+t2|t0+t3 = t11|t10
+    subq.ph   s7, s4, s6    // tmp7 = t1-t2|t0-t3 = t12|t13
+    sh        s0, 2(a1)
+    sh        s1, 6(a1)
+    sh        s2, 10(a1)
+    sh        s3, 14(a1)
+    mult      $0, $0        // ac0  = 0
+    dpa.w.ph  $ac0, s7, t8  // ac0 += t12*  4433 + t13* 10703
+    mult      $ac1, $0, $0  // ac1  = 0
+    dpa.w.ph  $ac1, s7, t9  // ac1 += t12*-10704 + t13*  4433
+    sra       s4, s5, 16    // tmp4 = t11
+    addiu     a1, a1, 16
+    addiu     s8, s8, -1
+    extr_r.w  s0, $ac0, 11  // tmp0 = (ac0 + 1024) >> 11
+    extr_r.w  s1, $ac1, 11  // tmp1 = (ac1 + 1024) >> 11
+    addu      s2, s5, s4    // tmp2 = t10 + t11
+    subu      s3, s5, s4    // tmp3 = t10 - t11
+    sll       s2, s2, 2     // tmp2 = (t10 + t11) << 2
+    sll       s3, s3, 2     // tmp3 = (t10 - t11) << 2
+    sh        s2, -16(a1)
+    sh        s3, -8(a1)
+    sh        s0, -12(a1)
+    bgtz      s8, 1b
+     sh       s1, -4(a1)
+    li        t0, 2260
+    li        t1, 11363
+    li        t2, 9633
+    li        t3, 6436
+    li        t4, 6437
+    li        t5, 2261
+    li        t6, 11362
+    li        t7, 2259
+    li        t8, 4433
+    li        t9, 10703
+    li        a1, 10704
+    li        s8, 8
+
+2:
+    lh        a2, 0(a0)     // 0
+    lh        a3, 16(a0)    // 8
+    lh        v0, 32(a0)    // 16
+    lh        v1, 48(a0)    // 24
+    lh        s4, 64(a0)    // 32
+    lh        s5, 80(a0)    // 40
+    lh        s6, 96(a0)    // 48
+    lh        s7, 112(a0)   // 56
+    addu      s2, v0, s5    // tmp2 = 16 + 40
+    subu      s5, v0, s5    // tmp5 = 16 - 40
+    addu      s3, v1, s4    // tmp3 = 24 + 32
+    subu      s4, v1, s4    // tmp4 = 24 - 32
+    addu      s0, a2, s7    // tmp0 =  0 + 56
+    subu      s7, a2, s7    // tmp7 =  0 - 56
+    addu      s1, a3, s6    // tmp1 =  8 + 48
+    subu      s6, a3, s6    // tmp6 =  8 - 48
+    addu      a2, s0, s3    // tmp10 = tmp0 + tmp3
+    subu      v1, s0, s3    // tmp13 = tmp0 - tmp3
+    addu      a3, s1, s2    // tmp11 = tmp1 + tmp2
+    subu      v0, s1, s2    // tmp12 = tmp1 - tmp2
+    mult      s7, t1        // ac0  = tmp7 * c1
+    madd      s4, t0        // ac0 += tmp4 * c0
+    madd      s5, t4        // ac0 += tmp5 * c4
+    madd      s6, t2        // ac0 += tmp6 * c2
+    mult      $ac1, s7, t2  // ac1  = tmp7 * c2
+    msub      $ac1, s4, t3  // ac1 -= tmp4 * c3
+    msub      $ac1, s5, t6  // ac1 -= tmp5 * c6
+    msub      $ac1, s6, t7  // ac1 -= tmp6 * c7
+    mult      $ac2, s7, t4  // ac2  = tmp7 * c4
+    madd      $ac2, s4, t2  // ac2 += tmp4 * c2
+    madd      $ac2, s5, t5  // ac2 += tmp5 * c5
+    msub      $ac2, s6, t6  // ac2 -= tmp6 * c6
+    mult      $ac3, s7, t0  // ac3  = tmp7 * c0
+    msub      $ac3, s4, t1  // ac3 -= tmp4 * c1
+    madd      $ac3, s5, t2  // ac3 += tmp5 * c2
+    msub      $ac3, s6, t3  // ac3 -= tmp6 * c3
+    extr_r.w  s0, $ac0, 15  // tmp0 = (ac0 + 16384) >> 15
+    extr_r.w  s1, $ac1, 15  // tmp1 = (ac1 + 16384) >> 15
+    extr_r.w  s2, $ac2, 15  // tmp2 = (ac2 + 16384) >> 15
+    extr_r.w  s3, $ac3, 15  // tmp3 = (ac3 + 16384) >> 15
+    addiu     s8, s8, -1
+    addu      s4, a2, a3    // tmp4 = tmp10 + tmp11
+    subu      s5, a2, a3    // tmp5 = tmp10 - tmp11
+    sh        s0, 16(a0)
+    sh        s1, 48(a0)
+    sh        s2, 80(a0)
+    sh        s3, 112(a0)
+    mult      v0, t8        // ac0  = tmp12 * c8
+    madd      v1, t9        // ac0 += tmp13 * c9
+    mult      $ac1, v1, t8  // ac1  = tmp13 * c8
+    msub      $ac1, v0, a1  // ac1 -= tmp12 * c10
+    addiu     a0, a0, 2
+    extr_r.w  s6, $ac0, 15  // tmp6 = (ac0 + 16384) >> 15
+    extr_r.w  s7, $ac1, 15  // tmp7 = (ac1 + 16384) >> 15
+    shra_r.w  s4, s4, 2     // tmp4 = (tmp4 + 2) >> 2
+    shra_r.w  s5, s5, 2     // tmp5 = (tmp5 + 2) >> 2
+    sh        s4, -2(a0)
+    sh        s5, 62(a0)
+    sh        s6, 30(a0)
+    bgtz      s8, 2b
+     sh       s7, 94(a0)
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    jr       ra
+     nop
+
+END(jsimd_fdct_islow_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_fdct_ifast_mips_dspr2)
+/*
+ * a0     - data
+ */
+    .set at
+    SAVE_REGS_ON_STACK 8, s0, s1
+    li           a1, 0x014e014e  // FIX_1_306562965 (334 << 16)|(334 & 0xffff)
+    li           a2, 0x008b008b  // FIX_0_541196100 (139 << 16)|(139 & 0xffff)
+    li           a3, 0x00620062  // FIX_0_382683433 (98 << 16) |(98 & 0xffff)
+    li           s1, 0x00b500b5  // FIX_0_707106781 (181 << 16)|(181 & 0xffff)
+
+    move         v0, a0
+    addiu        v1, v0, 128     // end address
+
+0:
+    lw           t0, 0(v0)       // tmp0 = 1|0
+    lw           t1, 4(v0)       // tmp1 = 3|2
+    lw           t2, 8(v0)       // tmp2 = 5|4
+    lw           t3, 12(v0)      // tmp3 = 7|6
+    packrl.ph    t1, t1, t1      // tmp1 = 2|3
+    packrl.ph    t3, t3, t3      // tmp3 = 6|7
+    subq.ph      t7, t1, t2      // tmp7 = 2-5|3-4 = t5|t4
+    subq.ph      t5, t0, t3      // tmp5 = 1-6|0-7 = t6|t7
+    addq.ph      t6, t1, t2      // tmp6 = 2+5|3+4 = t2|t3
+    addq.ph      t4, t0, t3      // tmp4 = 1+6|0+7 = t1|t0
+    addq.ph      t8, t4, t6      // tmp5 = t1+t2|t0+t3 = t11|t10
+    subq.ph      t9, t4, t6      // tmp7 = t1-t2|t0-t3 = t12|t13
+    sra          t4, t8, 16      // tmp4 = t11
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t9, s1
+    mult         $ac1, $0, $0    // ac1  = 0
+    dpa.w.ph     $ac1, t7, a3    // ac1 += t4*98 + t5*98
+    dpsx.w.ph    $ac1, t5, a3    // ac1 += t6*98 + t7*98
+    mult         $ac2, $0, $0    // ac2  = 0
+    dpa.w.ph     $ac2, t7, a2    // ac2 += t4*139 + t5*139
+    mult         $ac3, $0, $0    // ac3  = 0
+    dpa.w.ph     $ac3, t5, a1    // ac3 += t6*334 + t7*334
+    precrq.ph.w  t0, t5, t7      // t0 = t5|t6
+    addq.ph      t2, t8, t4      // tmp2 = t10 + t11
+    subq.ph      t3, t8, t4      // tmp3 = t10 - t11
+    extr.w       t4, $ac0, 8
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t0, s1    // ac0 += t5*181 + t6*181
+    extr.w       t0, $ac1, 8     // t0 = z5
+    extr.w       t1, $ac2, 8     // t1 = MULTIPLY(tmp10, 139)
+    extr.w       t7, $ac3, 8     // t2 = MULTIPLY(tmp12, 334)
+    extr.w       t8, $ac0, 8     // t8 = z3 = MULTIPLY(tmp11, 181)
+    add          t6, t1, t0      // t6 = z2
+    add          t7, t7, t0      // t7 = z4
+    subq.ph      t0, t5, t8      // t0 = z13 = tmp7 - z3
+    addq.ph      t8, t5, t8      // t9 = z11 = tmp7 + z3
+    addq.ph      t1, t0, t6      // t1 = z13 + z2
+    subq.ph      t6, t0, t6      // t6 = z13 - z2
+    addq.ph      t0, t8, t7      // t0 = z11 + z4
+    subq.ph      t7, t8, t7      // t7 = z11 - z4
+    addq.ph      t5, t4, t9
+    subq.ph      t4, t9, t4
+    sh           t2, 0(v0)
+    sh           t5, 4(v0)
+    sh           t3, 8(v0)
+    sh           t4, 12(v0)
+    sh           t1, 10(v0)
+    sh           t6, 6(v0)
+    sh           t0, 2(v0)
+    sh           t7, 14(v0)
+    addiu        v0, 16
+    bne          v1, v0, 0b
+     nop
+    move         v0, a0
+    addiu        v1, v0, 16
+
+1:
+    lh           t0, 0(v0)       // 0
+    lh           t1, 16(v0)      // 8
+    lh           t2, 32(v0)      // 16
+    lh           t3, 48(v0)      // 24
+    lh           t4, 64(v0)      // 32
+    lh           t5, 80(v0)      // 40
+    lh           t6, 96(v0)      // 48
+    lh           t7, 112(v0)     // 56
+    add          t8, t0, t7      // t8 = tmp0
+    sub          t7, t0, t7      // t7 = tmp7
+    add          t0, t1, t6      // t0 = tmp1
+    sub          t1, t1, t6      // t1 = tmp6
+    add          t6, t2, t5      // t6 = tmp2
+    sub          t5, t2, t5      // t5 = tmp5
+    add          t2, t3, t4      // t2 = tmp3
+    sub          t3, t3, t4      // t3 = tmp4
+    add          t4, t8, t2      // t4 = tmp10 = tmp0 + tmp3
+    sub          t8, t8, t2      // t8 = tmp13 = tmp0 - tmp3
+    sub          s0, t0, t6      // s0 = tmp12 = tmp1 - tmp2
+    ins          t8, s0, 16, 16  // t8 = tmp12|tmp13
+    add          t2, t0, t6      // t2 = tmp11 = tmp1 + tmp2
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t8, s1    // ac0 += t12*181 + t13*181
+    add          s0, t4, t2      // t8 = tmp10+tmp11
+    sub          t4, t4, t2      // t4 = tmp10-tmp11
+    sh           s0, 0(v0)
+    sh           t4, 64(v0)
+    extr.w       t2, $ac0, 8     // z1 = MULTIPLY(tmp12+tmp13,FIX_0_707106781)
+    addq.ph      t4, t8, t2      // t9 = tmp13 + z1
+    subq.ph      t8, t8, t2      // t2 = tmp13 - z1
+    sh           t4, 32(v0)
+    sh           t8, 96(v0)
+    add          t3, t3, t5      // t3 = tmp10 = tmp4 + tmp5
+    add          t0, t5, t1      // t0 = tmp11 = tmp5 + tmp6
+    add          t1, t1, t7      // t1 = tmp12 = tmp6 + tmp7
+    andi         t4, a1, 0xffff
+    mul          s0, t1, t4
+    sra          s0, s0, 8       // s0 = z4 = MULTIPLY(tmp12, FIX_1_306562965)
+    ins          t1, t3, 16, 16  // t1 = tmp10|tmp12
+    mult         $0, $0          // ac0  = 0
+    mulsa.w.ph   $ac0, t1, a3    // ac0 += t10*98 - t12*98
+    extr.w       t8, $ac0, 8     // z5 = MULTIPLY(tmp10-tmp12,FIX_0_382683433)
+    add          t2, t7, t8      // t2 = tmp7 + z5
+    sub          t7, t7, t8      // t7 = tmp7 - z5
+    andi         t4, a2, 0xffff
+    mul          t8, t3, t4
+    sra          t8, t8, 8       // t8 = z2 = MULTIPLY(tmp10, FIX_0_541196100)
+    andi         t4, s1, 0xffff
+    mul          t6, t0, t4
+    sra          t6, t6, 8       // t6 = z3 = MULTIPLY(tmp11, FIX_0_707106781)
+    add          t0, t6, t8      // t0 = z3 + z2
+    sub          t1, t6, t8      // t1 = z3 - z2
+    add          t3, t6, s0      // t3 = z3 + z4
+    sub          t4, t6, s0      // t4 = z3 - z4
+    sub          t5, t2, t1      // t5 = dataptr[5]
+    sub          t6, t7, t0      // t6 = dataptr[3]
+    add          t3, t2, t3      // t3 = dataptr[1]
+    add          t4, t7, t4      // t4 = dataptr[7]
+    sh           t5, 80(v0)
+    sh           t6, 48(v0)
+    sh           t3, 16(v0)
+    sh           t4, 112(v0)
+    addiu        v0, 2
+    bne          v0, v1, 1b
+     nop
+
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j            ra
+     nop
+END(jsimd_fdct_ifast_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_quantize_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - divisors
+ * a2     - workspace
+ */
+
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2
+
+    addiu   v0, a2, 124  // v0 = workspace_end
+    lh      t0, 0(a2)
+    lh      t1, 0(a1)
+    lh      t2, 128(a1)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0, t3
+    lh      t4, 384(a1)
+    lh      t5, 130(a1)
+    lh      t6, 2(a2)
+    lh      t7, 2(a1)
+    lh      t8, 386(a1)
+
+1:
+    andi    t1, 0xffff
+    add     t9, t0, t2
+    andi    t9, 0xffff
+    mul     v1, t9, t1
+    sra     s0, t6, 15
+    sll     s0, s0, 1
+    addiu   s0, s0, 1
+    addiu   t9, t4, 16
+    srav    v1, v1, t9
+    mul     v1, v1, t3
+    mul     t6, t6, s0
+    andi    t7, 0xffff
+    addiu   a2, a2, 4
+    addiu   a1, a1, 4
+    add     s1, t6, t5
+    andi    s1, 0xffff
+    sh      v1, 0(a0)
+
+    mul     s2, s1, t7
+    addiu   s1, t8, 16
+    srav    s2, s2, s1
+    mul     s2,s2, s0
+    lh      t0, 0(a2)
+    lh      t1, 0(a1)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0, t3
+    lh      t2, 128(a1)
+    lh      t4, 384(a1)
+    lh      t5, 130(a1)
+    lh      t8, 386(a1)
+    lh      t6, 2(a2)
+    lh      t7, 2(a1)
+    sh      s2, 2(a0)
+    lh      t0, 0(a2)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0,t3
+    bne     a2, v0, 1b
+     addiu  a0, a0, 4
+
+    andi    t1, 0xffff
+    add     t9, t0, t2
+    andi    t9, 0xffff
+    mul     v1, t9, t1
+    sra     s0, t6, 15
+    sll     s0, s0, 1
+    addiu   s0, s0, 1
+    addiu   t9, t4, 16
+    srav    v1, v1, t9
+    mul     v1, v1, t3
+    mul     t6, t6, s0
+    andi    t7, 0xffff
+    sh      v1, 0(a0)
+    add     s1, t6, t5
+    andi    s1, 0xffff
+    mul     s2, s1, t7
+    addiu   s1, t8, 16
+    addiu   a2, a2, 4
+    addiu   a1, a1, 4
+    srav    s2, s2, s1
+    mul     s2, s2, s0
+    sh      s2, 2(a0)
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2
+
+    j       ra
+     nop
+
+END(jsimd_quantize_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_quantize_float_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - divisors
+ * a2     - workspace
+ */
+
+    .set at
+
+    li         t1, 0x46800100     //integer representation 16384.5
+    mtc1       t1, f0
+    li         t0, 63
+0:
+    lwc1       f2, 0(a2)
+    lwc1       f10, 0(a1)
+    lwc1       f4, 4(a2)
+    lwc1       f12, 4(a1)
+    lwc1       f6, 8(a2)
+    lwc1       f14, 8(a1)
+    lwc1       f8, 12(a2)
+    lwc1       f16, 12(a1)
+    madd.s     f2, f0, f2, f10
+    madd.s     f4, f0, f4, f12
+    madd.s     f6, f0, f6, f14
+    madd.s     f8, f0, f8, f16
+    lwc1       f10, 16(a1)
+    lwc1       f12, 20(a1)
+    trunc.w.s  f2, f2
+    trunc.w.s  f4, f4
+    trunc.w.s  f6, f6
+    trunc.w.s  f8, f8
+    lwc1       f14, 24(a1)
+    lwc1       f16, 28(a1)
+    mfc1       t1, f2
+    mfc1       t2, f4
+    mfc1       t3, f6
+    mfc1       t4, f8
+    lwc1       f2, 16(a2)
+    lwc1       f4, 20(a2)
+    lwc1       f6, 24(a2)
+    lwc1       f8, 28(a2)
+    madd.s     f2, f0, f2, f10
+    madd.s     f4, f0, f4, f12
+    madd.s     f6, f0, f6, f14
+    madd.s     f8, f0, f8, f16
+    addiu      t1, t1, -16384
+    addiu      t2, t2, -16384
+    addiu      t3, t3, -16384
+    addiu      t4, t4, -16384
+    trunc.w.s  f2, f2
+    trunc.w.s  f4, f4
+    trunc.w.s  f6, f6
+    trunc.w.s  f8, f8
+    sh         t1, 0(a0)
+    sh         t2, 2(a0)
+    sh         t3, 4(a0)
+    sh         t4, 6(a0)
+    mfc1       t1, f2
+    mfc1       t2, f4
+    mfc1       t3, f6
+    mfc1       t4, f8
+    addiu      t0, t0, -8
+    addiu      a2, a2, 32
+    addiu      a1, a1, 32
+    addiu      t1, t1, -16384
+    addiu      t2, t2, -16384
+    addiu      t3, t3, -16384
+    addiu      t4, t4, -16384
+    sh         t1, 8(a0)
+    sh         t2, 10(a0)
+    sh         t3, 12(a0)
+    sh         t4, 14(a0)
+    bgez       t0, 0b
+     addiu     a0, a0, 16
+
+    j          ra
+     nop
+
+END(jsimd_quantize_float_mips_dspr2)
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_2x2_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    addiu     sp, sp, -40
+    move      v0, sp
+    addiu     s2, zero, 29692
+    addiu     s3, zero, -10426
+    addiu     s4, zero, 6967
+    addiu     s5, zero, -5906
+    lh        t0, 0(a1)         // t0 = inptr[DCTSIZE*0]
+    lh        t5, 0(a0)         // t5 = quantptr[DCTSIZE*0]
+    lh        t1, 48(a1)        // t1 = inptr[DCTSIZE*3]
+    lh        t6, 48(a0)        // t6 = quantptr[DCTSIZE*3]
+    mul       t4, t5, t0
+    lh        t0, 16(a1)        // t0 = inptr[DCTSIZE*1]
+    lh        t5, 16(a0)        // t5 = quantptr[DCTSIZE*1]
+    mul       t6, t6, t1
+    mul       t5, t5, t0
+    lh        t2, 80(a1)        // t2 = inptr[DCTSIZE*5]
+    lh        t7, 80(a0)        // t7 = quantptr[DCTSIZE*5]
+    lh        t3, 112(a1)       // t3 = inptr[DCTSIZE*7]
+    lh        t8, 112(a0)       // t8 = quantptr[DCTSIZE*7]
+    mul       t7, t7, t2
+    mult      zero, zero
+    mul       t8, t8, t3
+    li        s0, 0x73FCD746    // s0 = (29692 << 16) | (-10426 & 0xffff)
+    li        s1, 0x1B37E8EE    // s1 = (6967 << 16) | (-5906 & 0xffff)
+    ins       t6, t5, 16, 16    // t6 = t5|t6
+    sll       t4, t4, 15
+    dpa.w.ph  $ac0, t6, s0
+    lh        t1, 2(a1)
+    lh        t6, 2(a0)
+    ins       t8, t7, 16, 16    // t8 = t7|t8
+    dpa.w.ph  $ac0, t8, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 18(a1)
+    lh        t6, 18(a0)
+    lh        t2, 50(a1)
+    lh        t7, 50(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 82(a1)
+    lh        t2, 82(a0)
+    lh        t3, 114(a1)
+    lh        t4, 114(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 0(v0)
+    sw        t8, 20(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 6(a1)
+    lh        t6, 6(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 22(a1)
+    lh        t6, 22(a0)
+    lh        t2, 54(a1)
+    lh        t7, 54(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 86(a1)
+    lh        t2, 86(a0)
+    lh        t3, 118(a1)
+    lh        t4, 118(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 4(v0)
+    sw        t8, 24(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 10(a1)
+    lh        t6, 10(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 26(a1)
+    lh        t6, 26(a0)
+    lh        t2, 58(a1)
+    lh        t7, 58(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 90(a1)
+    lh        t2, 90(a0)
+    lh        t3, 122(a1)
+    lh        t4, 122(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 8(v0)
+    sw        t8, 28(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 14(a1)
+    lh        t6, 14(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 30(a1)
+    lh        t6, 30(a0)
+    lh        t2, 62(a1)
+    lh        t7, 62(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 94(a1)
+    lh        t2, 94(a0)
+    lh        t3, 126(a1)
+    lh        t4, 126(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 12(v0)
+    sw        t8, 32(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    lw        t9, 0(a2)
+    lw        t3, 0(v0)
+    lw        t7, 4(v0)
+    lw        t1, 8(v0)
+    addu      t9, t9, a3
+    sll       t3, t3, 15
+    subu      t8, t4, t0
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    shra_r.w  t8, t8, 13
+    sw        t0, 16(v0)
+    sw        t8, 36(v0)
+    lw        t5, 12(v0)
+    lw        t6, 16(v0)
+    mult      t7, s2
+    madd      t1, s3
+    madd      t5, s4
+    madd      t6, s5
+    lw        t5, 24(v0)
+    lw        t7, 28(v0)
+    mflo      t0, $ac0
+    lw        t8, 32(v0)
+    lw        t2, 36(v0)
+    mult      $ac1, t5, s2
+    madd      $ac1, t7, s3
+    madd      $ac1, t8, s4
+    madd      $ac1, t2, s5
+    addu      t1, t3, t0
+    subu      t6, t3, t0
+    shra_r.w  t1, t1, 20
+    shra_r.w  t6, t6, 20
+    mflo      t4, $ac1
+    shll_s.w  t1, t1, 24
+    shll_s.w  t6, t6, 24
+    sra       t1, t1, 24
+    sra       t6, t6, 24
+    addiu     t1, t1, 128
+    addiu     t6, t6, 128
+    lw        t0, 20(v0)
+    sb        t1, 0(t9)
+    sb        t6, 1(t9)
+    sll       t0, t0, 15
+    lw        t9, 4(a2)
+    addu      t1, t0, t4
+    subu      t6, t0, t4
+    addu      t9, t9, a3
+    shra_r.w  t1, t1, 20
+    shra_r.w  t6, t6, 20
+    shll_s.w  t1, t1, 24
+    shll_s.w  t6, t6, 24
+    sra       t1, t1, 24
+    sra       t6, t6, 24
+    addiu     t1, t1, 128
+    addiu     t6, t6, 128
+    sb        t1, 0(t9)
+    sb        t6, 1(t9)
+    addiu     sp, sp, 40
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j         ra
+     nop
+
+END(jsimd_idct_2x2_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_4x4_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ * 16(sp) - workspace[DCTSIZE*4];  // buffers data between passes
+ */
+
+    .set at
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw        v1, 48(sp)
+    move      t0, a1
+    move      t1, v1
+    li        t9, 4
+    li        s0, 0x2e75f93e
+    li        s1, 0x21f9ba79
+    li        s2, 0xecc2efb0
+    li        s3, 0x52031ccd
+
+0:
+    lh        s6, 32(t0)        // inptr[DCTSIZE*2]
+    lh        t6, 32(a0)        // quantptr[DCTSIZE*2]
+    lh        s7, 96(t0)        // inptr[DCTSIZE*6]
+    lh        t7, 96(a0)        // quantptr[DCTSIZE*6]
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        s4, 0(t0)         // inptr[DCTSIZE*0]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s5, 0(a0)         // quantptr[0]
+    li        s6, 15137
+    li        s7, 6270
+    mul       t2, s4, s5        // tmp0 = (inptr[0] * quantptr[0])
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        t5, 112(t0)       // inptr[DCTSIZE*7]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s4, 112(a0)       // quantptr[DCTSIZE*7]
+    lh        v0, 80(t0)        // inptr[DCTSIZE*5]
+    lh        s5, 80(a0)        // quantptr[DCTSIZE*5]
+    lh        s6, 48(a0)        // quantptr[DCTSIZE*3]
+    sll       t2, t2, 14        // tmp0 <<= (CONST_BITS+1)
+    lh        s7, 16(a0)        // quantptr[DCTSIZE*1]
+    lh        t8, 16(t0)        // inptr[DCTSIZE*1]
+    subu      t6, t6, t7        // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6)
+    lh        t7, 48(t0)        // inptr[DCTSIZE*3]
+    mul       t5, s4, t5        // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7])
+    mul       v0, s5, v0        // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5])
+    mul       t7, s6, t7        // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3])
+    mul       t8, s7, t8        // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1])
+    addu      t3, t2, t6        // tmp10 = tmp0 + z2
+    subu      t4, t2, t6        // tmp10 = tmp0 - z2
+    mult      $ac0, zero, zero
+    mult      $ac1, zero, zero
+    ins       t5, v0, 16, 16
+    ins       t7, t8, 16, 16
+    addiu     t9, t9, -1
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    mflo      s4, $ac0
+    mflo      s5, $ac1
+    addiu     a0, a0, 2
+    addiu     t1, t1, 4
+    addiu     t0, t0, 2
+    addu      t6, t4, s4
+    subu      t5, t4, s4
+    addu      s6, t3, s5
+    subu      s7, t3, s5
+    shra_r.w  t6, t6, 12        // DESCALE(tmp12 + temp1, 12)
+    shra_r.w  t5, t5, 12        // DESCALE(tmp12 - temp1, 12)
+    shra_r.w  s6, s6, 12        // DESCALE(tmp10 + temp2, 12)
+    shra_r.w  s7, s7, 12        // DESCALE(tmp10 - temp2, 12)
+    sw        t6, 28(t1)
+    sw        t5, 60(t1)
+    sw        s6, -4(t1)
+    bgtz      t9, 0b
+     sw       s7, 92(t1)
+    // second loop three pass
+    li        t9, 3
+1:
+    lh        s6, 34(t0)        // inptr[DCTSIZE*2]
+    lh        t6, 34(a0)        // quantptr[DCTSIZE*2]
+    lh        s7, 98(t0)        // inptr[DCTSIZE*6]
+    lh        t7, 98(a0)        // quantptr[DCTSIZE*6]
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        s4, 2(t0)         // inptr[DCTSIZE*0]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s5, 2(a0)         // quantptr[DCTSIZE*0]
+    li        s6, 15137
+    li        s7, 6270
+    mul       t2, s4, s5        // tmp0 = (inptr[0] * quantptr[0])
+    mul       v0, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        t5, 114(t0)       // inptr[DCTSIZE*7]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s4, 114(a0)       // quantptr[DCTSIZE*7]
+    lh        s5, 82(a0)        // quantptr[DCTSIZE*5]
+    lh        t6, 82(t0)        // inptr[DCTSIZE*5]
+    sll       t2, t2, 14        // tmp0 <<= (CONST_BITS+1)
+    lh        s6, 50(a0)        // quantptr[DCTSIZE*3]
+    lh        t8, 18(t0)        // inptr[DCTSIZE*1]
+    subu      v0, v0, t7        // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6)
+    lh        t7, 50(t0)        // inptr[DCTSIZE*3]
+    lh        s7, 18(a0)        // quantptr[DCTSIZE*1]
+    mul       t5, s4, t5        // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7])
+    mul       t6, s5, t6        // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5])
+    mul       t7, s6, t7        // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3])
+    mul       t8, s7, t8        // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1])
+    addu      t3, t2, v0        // tmp10 = tmp0 + z2
+    subu      t4, t2, v0        // tmp10 = tmp0 - z2
+    mult      $ac0, zero, zero
+    mult      $ac1, zero, zero
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    mflo      t5, $ac0
+    mflo      t6, $ac1
+    addiu     t9, t9, -1
+    addiu     t0, t0, 2
+    addiu     a0, a0, 2
+    addiu     t1, t1, 4
+    addu      s5, t4, t5
+    subu      s4, t4, t5
+    addu      s6, t3, t6
+    subu      s7, t3, t6
+    shra_r.w  s5, s5, 12        // DESCALE(tmp12 + temp1, 12)
+    shra_r.w  s4, s4, 12        // DESCALE(tmp12 - temp1, 12)
+    shra_r.w  s6, s6, 12        // DESCALE(tmp10 + temp2, 12)
+    shra_r.w  s7, s7, 12        // DESCALE(tmp10 - temp2, 12)
+    sw        s5, 32(t1)
+    sw        s4, 64(t1)
+    sw        s6, 0(t1)
+    bgtz      t9, 1b
+     sw       s7, 96(t1)
+    move      t1, v1
+    li        s4, 15137
+    lw        s6, 8(t1)         // wsptr[2]
+    li        s5, 6270
+    lw        s7, 24(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 0(t1)         // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 28(t1)        // wsptr[7]
+    lh        t6, 20(t1)        // wsptr[5]
+    lh        t7, 12(t1)        // wsptr[3]
+    lh        t8, 4(t1)         // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 0(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    // 2
+    li        s4, 15137
+    lw        s6, 40(t1)        // wsptr[2]
+    li        s5, 6270
+    lw        s7, 56(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 32(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 60(t1)        // wsptr[7]
+    lh        t6, 52(t1)        // wsptr[5]
+    lh        t7, 44(t1)        // wsptr[3]
+    lh        t8, 36(t1)        // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, CONST_BITS-PASS1_BITS+1)
+    sll       s4, t9, 2
+    lw        v0, 4(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    // 3
+    li        s4, 15137
+    lw        s6, 72(t1)        // wsptr[2]
+    li        s5, 6270
+    lw        s7, 88(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 64(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 92(t1)        // wsptr[7]
+    lh        t6, 84(t1)        // wsptr[5]
+    lh        t7, 76(t1)        // wsptr[3]
+    lh        t8, 68(t1)        // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 8(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    li        s4, 15137
+    lw        s6, 104(t1)       // wsptr[2]
+    li        s5, 6270
+    lw        s7, 120(t1)       // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 96(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], -FIX_0_765366865)
+    lh        t5, 124(t1)       // wsptr[7]
+    lh        t6, 116(t1)       // wsptr[5]
+    lh        t7, 108(t1)       // wsptr[3]
+    lh        t8, 100(t1)       // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2;
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2;
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 12(a2)        // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+END(jsimd_idct_4x4_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_6x6_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu     sp, sp, -144
+    move      v0, sp
+    addiu     v1, v0, 24
+    addiu     t9, zero, 5793
+    addiu     s0, zero, 10033
+    addiu     s1, zero, 2998
+
+1:
+    lh        s2, 0(a0)   // q0 = quantptr[ 0]
+    lh        s3, 32(a0)  // q1 = quantptr[16]
+    lh        s4, 64(a0)  // q2 = quantptr[32]
+    lh        t2, 64(a1)  // tmp2 = inptr[32]
+    lh        t1, 32(a1)  // tmp1 = inptr[16]
+    lh        t0, 0(a1)   // tmp0 = inptr[ 0]
+    mul       t2, t2, s4  // tmp2 = tmp2 * q2
+    mul       t1, t1, s3  // tmp1 = tmp1 * q1
+    mul       t0, t0, s2  // tmp0 = tmp0 * q0
+    lh        t6, 16(a1)  // z1 = inptr[ 8]
+    lh        t8, 80(a1)  // z3 = inptr[40]
+    lh        t7, 48(a1)  // z2 = inptr[24]
+    lh        s2, 16(a0)  // q0 = quantptr[ 8]
+    lh        s4, 80(a0)  // q2 = quantptr[40]
+    lh        s3, 48(a0)  // q1 = quantptr[24]
+    mul       t2, t2, t9  // tmp2 = tmp2 * 5793
+    mul       t1, t1, s0  // tmp1 = tmp1 * 10033
+    sll       t0, t0, 13  // tmp0 = tmp0 << 13
+    mul       t6, t6, s2  // z1 = z1 * q0
+    mul       t8, t8, s4  // z3 = z3 * q2
+    mul       t7, t7, s3  // z2 = z2 * q1
+    addu      t3, t0, t2  // tmp10 = tmp0 + tmp2
+    sll       t2, t2, 1   // tmp2 = tmp2 << 2
+    subu      t4, t0, t2  // tmp11 = tmp0 - tmp2;
+    subu      t5, t3, t1  // tmp12 = tmp10 - tmp1
+    addu      t3, t3, t1  // tmp10 = tmp10 + tmp1
+    addu      t1, t6, t8  // tmp1 = z1 + z3
+    mul       t1, t1, s1  // tmp1 = tmp1 * 2998
+    shra_r.w  t4, t4, 11  // tmp11 = (tmp11 + 1024) >> 11
+    subu      t2, t6, t8  // tmp2 = z1 - z3
+    subu      t2, t2, t7  // tmp2 = tmp2 - z2
+    sll       t2, t2, 2   // tmp2 = tmp2 << 2
+    addu      t0, t6, t7  // tmp0 = z1 + z2
+    sll       t0, t0, 13  // tmp0 = tmp0 << 13
+    subu      s2, t8, t7  // q0 = z3 - z2
+    sll       s2, s2, 13  // q0 = q0 << 13
+    addu      t0, t0, t1  // tmp0 = tmp0 + tmp1
+    addu      t1, s2, t1  // tmp1 = q0 + tmp1
+    addu      s2, t4, t2  // q0 = tmp11 + tmp2
+    subu      s3, t4, t2  // q1 = tmp11 - tmp2
+    addu      t6, t3, t0  // z1 = tmp10 + tmp0
+    subu      t7, t3, t0  // z2 = tmp10 - tmp0
+    addu      t4, t5, t1  // tmp11 = tmp12 + tmp1
+    subu      t5, t5, t1  // tmp12 = tmp12 - tmp1
+    shra_r.w  t6, t6, 11  // z1 = (z1 + 1024) >> 11
+    shra_r.w  t7, t7, 11  // z2 = (z2 + 1024) >> 11
+    shra_r.w  t4, t4, 11  // tmp11 = (tmp11 + 1024) >> 11
+    shra_r.w  t5, t5, 11  // tmp12 = (tmp12 + 1024) >> 11
+    sw        s2, 24(v0)
+    sw        s3, 96(v0)
+    sw        t6, 0(v0)
+    sw        t7, 120(v0)
+    sw        t4, 48(v0)
+    sw        t5, 72(v0)
+    addiu     v0, v0, 4
+    addiu     a1, a1, 2
+    bne       v0, v1, 1b
+     addiu    a0, a0, 2
+
+    /* Pass 2: process 6 rows from work array, store into output array. */
+    move      v0, sp
+    addiu     v1, v0, 144
+
+2:
+    lw        t0, 0(v0)
+    lw        t2, 16(v0)
+    lw        s5, 0(a2)
+    addiu     t0, t0, 16
+    sll       t0, t0, 13
+    mul       t3, t2, t9
+    lw        t6, 4(v0)
+    lw        t8, 20(v0)
+    lw        t7, 12(v0)
+    addu      s5, s5, a3
+    addu      s6, t6, t8
+    mul       s6, s6, s1
+    addu      t1, t0, t3
+    subu      t4, t0, t3
+    subu      t4, t4, t3
+    lw        t3, 8(v0)
+    mul       t0, t3, s0
+    addu      s7, t6, t7
+    sll       s7, s7, 13
+    addu      s7, s6, s7
+    subu      t2, t8, t7
+    sll       t2, t2, 13
+    addu      t2, s6, t2
+    subu      s6, t6, t7
+    subu      s6, s6, t8
+    sll       s6, s6, 13
+    addu      t3, t1, t0
+    subu      t5, t1, t0
+    addu      t6, t3, s7
+    subu      t3, t3, s7
+    addu      t7, t4, s6
+    subu      t4, t4, s6
+    addu      t8, t5, t2
+    subu      t5, t5, t2
+    shll_s.w  t6, t6, 6
+    shll_s.w  t3, t3, 6
+    shll_s.w  t7, t7, 6
+    shll_s.w  t4, t4, 6
+    shll_s.w  t8, t8, 6
+    shll_s.w  t5, t5, 6
+    sra       t6, t6, 24
+    addiu     t6, t6, 128
+    sra       t3, t3, 24
+    addiu     t3, t3, 128
+    sb        t6, 0(s5)
+    sra       t7, t7, 24
+    addiu     t7, t7, 128
+    sb        t3, 5(s5)
+    sra       t4, t4, 24
+    addiu     t4, t4, 128
+    sb        t7, 1(s5)
+    sra       t8, t8, 24
+    addiu     t8, t8, 128
+    sb        t4, 4(s5)
+    addiu     v0, v0, 24
+    sra       t5, t5, 24
+    addiu     t5, t5, 128
+    sb        t8, 2(s5)
+    addiu     a2, a2,  4
+    bne       v0, v1, 2b
+     sb       t5, 3(s5)
+
+    addiu     sp, sp, 144
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+
+END(jsimd_idct_6x6_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass1_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - workspace
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li         a3, 8
+
+1:
+    // odd part
+    lh         t0, 48(a1)
+    lh         t1, 48(a0)
+    lh         t2, 16(a1)
+    lh         t3, 16(a0)
+    lh         t4, 80(a1)
+    lh         t5, 80(a0)
+    lh         t6, 112(a1)
+    lh         t7, 112(a0)
+    mul        t0, t0, t1    // z2
+    mul        t1, t2, t3    // z1
+    mul        t2, t4, t5    // z3
+    mul        t3, t6, t7    // z4
+    li         t4, 10703     // FIX(1.306562965)
+    li         t5, 4433      // FIX_0_541196100
+    li         t6, 7053      // FIX(0.860918669)
+    mul        t4, t0,t4     // tmp11
+    mul        t5, t0,t5     // -tmp14
+    addu       t7, t1,t2     // tmp10
+    addu       t8, t7,t3     // tmp10 + z4
+    mul        t6, t6, t8    // tmp15
+    li         t8, 2139      // FIX(0.261052384)
+    mul        t8, t7, t8    // MULTIPLY(tmp10, FIX(0.261052384))
+    li         t7, 2295      // FIX(0.280143716)
+    mul        t7, t1, t7    // MULTIPLY(z1, FIX(0.280143716))
+    addu       t9, t2, t3    // z3 + z4
+    li         s0, 8565      // FIX(1.045510580)
+    mul        t9, t9, s0    // -tmp13
+    li         s0, 12112     // FIX(1.478575242)
+    mul        s0, t2, s0    // MULTIPLY(z3, FIX(1.478575242)
+    li         s1, 12998     // FIX(1.586706681)
+    mul        s1, t3, s1    // MULTIPLY(z4, FIX(1.586706681))
+    li         s2, 5540      // FIX(0.676326758)
+    mul        s2, t1, s2    // MULTIPLY(z1, FIX(0.676326758))
+    li         s3, 16244     // FIX(1.982889723)
+    mul        s3, t3, s3    // MULTIPLY(z4, FIX(1.982889723))
+    subu       t1, t1, t3    // z1-=z4
+    subu       t0, t0, t2    // z2-=z3
+    addu       t2, t0, t1    // z1+z2
+    li         t3, 4433      // FIX_0_541196100
+    mul        t2, t2, t3    // z3
+    li         t3, 6270      // FIX_0_765366865
+    mul        t1, t1, t3    // MULTIPLY(z1, FIX_0_765366865)
+    li         t3, 15137     // FIX_0_765366865
+    mul        t0, t0, t3    // MULTIPLY(z2, FIX_1_847759065)
+    addu       t8, t6, t8    // tmp12
+    addu       t3, t8, t4    // tmp12 + tmp11
+    addu       t3, t3, t7    // tmp10
+    subu       t8, t8, t9    // tmp12 + tmp13
+    addu       s0, t5, s0
+    subu       t8, t8, s0    // tmp12
+    subu       t9, t6, t9
+    subu       s1, s1, t4
+    addu       t9, t9, s1    // tmp13
+    subu       t6, t6, t5
+    subu       t6, t6, s2
+    subu       t6, t6, s3    // tmp15
+    // even part start
+    lh         t4, 64(a1)
+    lh         t5, 64(a0)
+    lh         t7, 32(a1)
+    lh         s0, 32(a0)
+    lh         s1, 0(a1)
+    lh         s2, 0(a0)
+    lh         s3, 96(a1)
+    lh         v0, 96(a0)
+    mul        t4, t4, t5    // DEQUANTIZE(inptr[DCTSIZE*4],quantptr[DCTSIZE*4])
+    mul        t5, t7, s0    // DEQUANTIZE(inptr[DCTSIZE*2],quantptr[DCTSIZE*2])
+    mul        t7, s1, s2    // DEQUANTIZE(inptr[DCTSIZE*0],quantptr[DCTSIZE*0])
+    mul        s0, s3, v0    // DEQUANTIZE(inptr[DCTSIZE*6],quantptr[DCTSIZE*6])
+    // odd part end
+    addu       t1, t2, t1    // tmp11
+    subu       t0, t2, t0    // tmp14
+    // update counter and pointers
+    addiu      a3, a3, -1
+    addiu      a0, a0, 2
+    addiu      a1, a1, 2
+    // even part rest
+    li         s1, 10033
+    li         s2, 11190
+    mul        t4, t4, s1    // z4
+    mul        s1, t5, s2    // z4
+    sll        t5, t5, 13    // z1
+    sll        t7, t7, 13
+    addiu      t7, t7, 1024  // z3
+    sll        s0, s0, 13    // z2
+    addu       s2, t7, t4    // tmp10
+    subu       t4, t7, t4    // tmp11
+    subu       s3, t5, s0    // tmp12
+    addu       t2, t7, s3    // tmp21
+    subu       s3, t7, s3    // tmp24
+    addu       t7, s1, s0    // tmp12
+    addu       v0, s2, t7    // tmp20
+    subu       s2, s2, t7    // tmp25
+    subu       s1, s1, t5    // z4 - z1
+    subu       s1, s1, s0    // tmp12
+    addu       s0, t4, s1    // tmp22
+    subu       t4, t4, s1    // tmp23
+    // final output stage
+    addu       t5, v0, t3
+    subu       v0, v0, t3
+    addu       t3, t2, t1
+    subu       t2, t2, t1
+    addu       t1, s0, t8
+    subu       s0, s0, t8
+    addu       t8, t4, t9
+    subu       t4, t4, t9
+    addu       t9, s3, t0
+    subu       s3, s3, t0
+    addu       t0, s2, t6
+    subu       s2, s2, t6
+    sra        t5, t5, 11
+    sra        t3, t3, 11
+    sra        t1, t1, 11
+    sra        t8, t8, 11
+    sra        t9, t9, 11
+    sra        t0, t0, 11
+    sra        s2, s2, 11
+    sra        s3, s3, 11
+    sra        t4, t4, 11
+    sra        s0, s0, 11
+    sra        t2, t2, 11
+    sra        v0, v0, 11
+    sw         t5, 0(a2)
+    sw         t3, 32(a2)
+    sw         t1, 64(a2)
+    sw         t8, 96(a2)
+    sw         t9, 128(a2)
+    sw         t0, 160(a2)
+    sw         s2, 192(a2)
+    sw         s3, 224(a2)
+    sw         t4, 256(a2)
+    sw         s0, 288(a2)
+    sw         t2, 320(a2)
+    sw         v0, 352(a2)
+    bgtz       a3, 1b
+     addiu     a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j          ra
+     nop
+
+END(jsimd_idct_12x12_pass1_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass2_mips_dspr2)
+/*
+ * a0     - workspace
+ * a1     - output
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li        a3, 12
+
+1:
+    // Odd part
+    lw        t0, 12(a0)
+    lw        t1, 4(a0)
+    lw        t2, 20(a0)
+    lw        t3, 28(a0)
+    li        t4, 10703     // FIX(1.306562965)
+    li        t5, 4433      // FIX_0_541196100
+    mul       t4, t0, t4    // tmp11
+    mul       t5, t0, t5    // -tmp14
+    addu      t6, t1, t2    // tmp10
+    li        t7, 2139      // FIX(0.261052384)
+    mul       t7, t6, t7    // MULTIPLY(tmp10, FIX(0.261052384))
+    addu      t6, t6, t3    // tmp10 + z4
+    li        t8, 7053      // FIX(0.860918669)
+    mul       t6, t6, t8    // tmp15
+    li        t8, 2295      // FIX(0.280143716)
+    mul       t8, t1, t8    // MULTIPLY(z1, FIX(0.280143716))
+    addu      t9, t2, t3    // z3 + z4
+    li        s0, 8565      // FIX(1.045510580)
+    mul       t9, t9, s0    // -tmp13
+    li        s0, 12112     // FIX(1.478575242)
+    mul       s0, t2, s0    // MULTIPLY(z3, FIX(1.478575242))
+    li        s1, 12998     // FIX(1.586706681)
+    mul       s1, t3, s1    // MULTIPLY(z4, FIX(1.586706681))
+    li        s2, 5540      // FIX(0.676326758)
+    mul       s2, t1, s2    // MULTIPLY(z1, FIX(0.676326758))
+    li        s3, 16244     // FIX(1.982889723)
+    mul       s3, t3, s3    // MULTIPLY(z4, FIX(1.982889723))
+    subu      t1, t1, t3    // z1 -= z4
+    subu      t0, t0, t2    // z2 -= z3
+    addu      t2, t1, t0    // z1 + z2
+    li        t3, 4433      // FIX_0_541196100
+    mul       t2, t2, t3    // z3
+    li        t3, 6270      // FIX_0_765366865
+    mul       t1, t1, t3    // MULTIPLY(z1, FIX_0_765366865)
+    li        t3, 15137     // FIX_1_847759065
+    mul       t0, t0, t3    // MULTIPLY(z2, FIX_1_847759065)
+    addu      t3, t6, t7    // tmp12
+    addu      t7, t3, t4
+    addu      t7, t7, t8    // tmp10
+    subu      t3, t3, t9
+    subu      t3, t3, t5
+    subu      t3, t3, s0    // tmp12
+    subu      t9, t6, t9
+    subu      t9, t9, t4
+    addu      t9, t9, s1    // tmp13
+    subu      t6, t6, t5
+    subu      t6, t6, s2
+    subu      t6, t6, s3    // tmp15
+    addu      t1, t2, t1    // tmp11
+    subu      t0, t2, t0    // tmp14
+    // even part
+    lw        t2, 16(a0)    // z4
+    lw        t4, 8(a0)     // z1
+    lw        t5, 0(a0)     // z3
+    lw        t8, 24(a0)    // z2
+    li        s0, 10033     // FIX(1.224744871)
+    li        s1, 11190     // FIX(1.366025404)
+    mul       t2, t2, s0    // z4
+    mul       s0, t4, s1    // z4
+    addiu     t5, t5, 0x10
+    sll       t5, t5, 13    // z3
+    sll       t4, t4, 13    // z1
+    sll       t8, t8, 13    // z2
+    subu      s1, t4, t8    // tmp12
+    addu      s2, t5, t2    // tmp10
+    subu      t2, t5, t2    // tmp11
+    addu      s3, t5, s1    // tmp21
+    subu      s1, t5, s1    // tmp24
+    addu      t5, s0, t8    // tmp12
+    addu      v0, s2, t5    // tmp20
+    subu      t5, s2, t5    // tmp25
+    subu      t4, s0, t4
+    subu      t4, t4, t8    // tmp12
+    addu      t8, t2, t4    // tmp22
+    subu      t2, t2, t4    // tmp23
+    // increment counter and pointers
+    addiu     a3, a3, -1
+    addiu     a0, a0, 32
+    // Final stage
+    addu      t4, v0, t7
+    subu      v0, v0, t7
+    addu      t7, s3, t1
+    subu      s3, s3, t1
+    addu      t1, t8, t3
+    subu      t8, t8, t3
+    addu      t3, t2, t9
+    subu      t2, t2, t9
+    addu      t9, s1, t0
+    subu      s1, s1, t0
+    addu      t0, t5, t6
+    subu      t5, t5, t6
+    sll       t4, t4, 4
+    sll       t7, t7, 4
+    sll       t1, t1, 4
+    sll       t3, t3, 4
+    sll       t9, t9, 4
+    sll       t0, t0, 4
+    sll       t5, t5, 4
+    sll       s1, s1, 4
+    sll       t2, t2, 4
+    sll       t8, t8, 4
+    sll       s3, s3, 4
+    sll       v0, v0, 4
+    shll_s.w  t4, t4, 2
+    shll_s.w  t7, t7, 2
+    shll_s.w  t1, t1, 2
+    shll_s.w  t3, t3, 2
+    shll_s.w  t9, t9, 2
+    shll_s.w  t0, t0, 2
+    shll_s.w  t5, t5, 2
+    shll_s.w  s1, s1, 2
+    shll_s.w  t2, t2, 2
+    shll_s.w  t8, t8, 2
+    shll_s.w  s3, s3, 2
+    shll_s.w  v0, v0, 2
+    srl       t4, t4, 24
+    srl       t7, t7, 24
+    srl       t1, t1, 24
+    srl       t3, t3, 24
+    srl       t9, t9, 24
+    srl       t0, t0, 24
+    srl       t5, t5, 24
+    srl       s1, s1, 24
+    srl       t2, t2, 24
+    srl       t8, t8, 24
+    srl       s3, s3, 24
+    srl       v0, v0, 24
+    lw        t6, 0(a1)
+    addiu     t4, t4, 0x80
+    addiu     t7, t7, 0x80
+    addiu     t1, t1, 0x80
+    addiu     t3, t3, 0x80
+    addiu     t9, t9, 0x80
+    addiu     t0, t0, 0x80
+    addiu     t5, t5, 0x80
+    addiu     s1, s1, 0x80
+    addiu     t2, t2, 0x80
+    addiu     t8, t8, 0x80
+    addiu     s3, s3, 0x80
+    addiu     v0, v0, 0x80
+    sb        t4, 0(t6)
+    sb        t7, 1(t6)
+    sb        t1, 2(t6)
+    sb        t3, 3(t6)
+    sb        t9, 4(t6)
+    sb        t0, 5(t6)
+    sb        t5, 6(t6)
+    sb        s1, 7(t6)
+    sb        t2, 8(t6)
+    sb        t8, 9(t6)
+    sb        s3, 10(t6)
+    sb        v0, 11(t6)
+    bgtz      a3, 1b
+     addiu    a1, a1, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    jr        ra
+     nop
+
+END(jsimd_idct_12x12_pass2_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_convsamp_mips_dspr2)
+/*
+ * a0     - sample_data
+ * a1     - start_col
+ * a2     - workspace
+ */
+
+    lw             t0, 0(a0)
+    li             t7, 0xff80ff80
+    addu           t0, t0, a1
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    lw             t0, 4(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 0(a2)
+    usw            t4, 4(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 8(a2)
+    usw            t6, 12(a2)
+
+    lw             t0, 8(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 16(a2)
+    usw            t4, 20(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 24(a2)
+    usw            t6, 28(a2)
+
+    lw             t0, 12(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 32(a2)
+    usw            t4, 36(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 40(a2)
+    usw            t6, 44(a2)
+
+    lw             t0, 16(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 48(a2)
+    usw            t4, 52(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 56(a2)
+    usw            t6, 60(a2)
+
+    lw             t0, 20(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 64(a2)
+    usw            t4, 68(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 72(a2)
+    usw            t6, 76(a2)
+
+    lw             t0, 24(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 80(a2)
+    usw            t4, 84(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 88(a2)
+    usw            t6, 92(a2)
+
+    lw             t0, 28(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 96(a2)
+    usw            t4, 100(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 104(a2)
+    usw            t6, 108(a2)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 112(a2)
+    usw            t4, 116(a2)
+    usw            t5, 120(a2)
+    usw            t6, 124(a2)
+
+    j              ra
+     nop
+
+END(jsimd_convsamp_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_convsamp_float_mips_dspr2)
+/*
+ * a0     - sample_data
+ * a1     - start_col
+ * a2     - workspace
+ */
+
+    .set at
+
+    lw       t0, 0(a0)
+    addu     t0, t0, a1
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 4(a0)
+    swc1     f2, 0(a2)
+    swc1     f4, 4(a2)
+    swc1     f6, 8(a2)
+    addu     t0, t0, a1
+    swc1     f8, 12(a2)
+    swc1     f10, 16(a2)
+    swc1     f12, 20(a2)
+    swc1     f14, 24(a2)
+    swc1     f16, 28(a2)
+    //elemr 1
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 8(a0)
+    swc1     f2, 32(a2)
+    swc1     f4, 36(a2)
+    swc1     f6, 40(a2)
+    addu     t0, t0, a1
+    swc1     f8, 44(a2)
+    swc1     f10, 48(a2)
+    swc1     f12, 52(a2)
+    swc1     f14, 56(a2)
+    swc1     f16, 60(a2)
+    //elemr 2
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 12(a0)
+    swc1     f2, 64(a2)
+    swc1     f4, 68(a2)
+    swc1     f6, 72(a2)
+    addu     t0, t0, a1
+    swc1     f8, 76(a2)
+    swc1     f10, 80(a2)
+    swc1     f12, 84(a2)
+    swc1     f14, 88(a2)
+    swc1     f16, 92(a2)
+    //elemr 3
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 16(a0)
+    swc1     f2, 96(a2)
+    swc1     f4, 100(a2)
+    swc1     f6, 104(a2)
+    addu     t0, t0, a1
+    swc1     f8, 108(a2)
+    swc1     f10, 112(a2)
+    swc1     f12, 116(a2)
+    swc1     f14, 120(a2)
+    swc1     f16, 124(a2)
+    //elemr 4
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 20(a0)
+    swc1     f2, 128(a2)
+    swc1     f4, 132(a2)
+    swc1     f6, 136(a2)
+    addu     t0, t0, a1
+    swc1     f8, 140(a2)
+    swc1     f10, 144(a2)
+    swc1     f12, 148(a2)
+    swc1     f14, 152(a2)
+    swc1     f16, 156(a2)
+    //elemr 5
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 24(a0)
+    swc1     f2, 160(a2)
+    swc1     f4, 164(a2)
+    swc1     f6, 168(a2)
+    addu     t0, t0, a1
+    swc1     f8, 172(a2)
+    swc1     f10, 176(a2)
+    swc1     f12, 180(a2)
+    swc1     f14, 184(a2)
+    swc1     f16, 188(a2)
+    //elemr 6
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 28(a0)
+    swc1     f2, 192(a2)
+    swc1     f4, 196(a2)
+    swc1     f6, 200(a2)
+    addu     t0, t0, a1
+    swc1     f8, 204(a2)
+    swc1     f10, 208(a2)
+    swc1     f12, 212(a2)
+    swc1     f14, 216(a2)
+    swc1     f16, 220(a2)
+    //elemr 7
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    swc1     f2, 224(a2)
+    swc1     f4, 228(a2)
+    swc1     f6, 232(a2)
+    swc1     f8, 236(a2)
+    swc1     f10, 240(a2)
+    swc1     f12, 244(a2)
+    swc1     f14, 248(a2)
+    swc1     f16, 252(a2)
+
+    j        ra
+     nop
+
+END(jsimd_convsamp_float_mips_dspr2)
+
+/*****************************************************************************/
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_powerpc.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_powerpc.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_powerpc.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_powerpc.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,741 @@
+/*
+ * jsimd_powerpc.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014-2015 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * PowerPC architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+static unsigned int simd_support = ~0;
+
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = JSIMD_ALTIVEC;
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_extrgb_ycc_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_extrgbx_ycc_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_extbgr_ycc_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_extbgrx_ycc_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_extxbgr_ycc_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_extxrgb_ycc_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_rgb_ycc_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_extrgb_gray_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_extrgbx_gray_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_extbgr_gray_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_extbgrx_gray_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_extxbgr_gray_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_extxrgb_gray_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_rgb_gray_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_ycc_extrgb_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_ycc_extrgbx_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_ycc_extbgr_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_ycc_extbgrx_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_ycc_extxbgr_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_ycc_extxrgb_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_ycc_rgb_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks,
+                                input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks,
+                                input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_h2v2_extrgb_merged_upsample_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_h2v2_extrgbx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_h2v2_extbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_h2v2_extbgrx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_h2v2_extxbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_h2v2_extxrgb_merged_upsample_altivec;
+      break;
+    default:
+      altivecfct=jsimd_h2v2_merged_upsample_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_h2v1_extrgb_merged_upsample_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_h2v1_extrgbx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_h2v1_extbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_h2v1_extbgrx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_h2v1_extxbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_h2v1_extxrgb_merged_upsample_altivec;
+      break;
+    default:
+      altivecfct=jsimd_h2v1_merged_upsample_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_altivec(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_altivec(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_x86_64.c glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_x86_64.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/jsimd_x86_64.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/jsimd_x86_64.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,887 @@
+/*
+ * jsimd_x86_64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014, 2016 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order) (((size_t)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = JSIMD_SSE2 | JSIMD_SSE;
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_ycc_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_ycc_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_ycc_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_ycc_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_ycc_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_ycc_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_rgb_ycc_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_ycc_extrgb_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_ycc_extrgbx_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_ycc_extbgr_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_ycc_extbgrx_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_ycc_extxbgr_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_ycc_extxrgb_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_ycc_rgb_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2;
+      break;
+    default:
+      sse2fct=jsimd_h2v2_merged_upsample_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2;
+      break;
+    default:
+      sse2fct=jsimd_h2v1_merged_upsample_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+  jsimd_fdct_float_sse(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/Makefile.am glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/Makefile.am
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/Makefile.am	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/Makefile.am	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,97 @@
+noinst_LTLIBRARIES = libsimd.la
+
+BUILT_SOURCES = jsimdcfg.inc
+
+EXTRA_DIST = nasm_lt.sh CMakeLists.txt \
+	jccolext-mmx.asm   jcgryext-mmx.asm   jdcolext-mmx.asm   jdmrgext-mmx.asm \
+	jccolext-sse2.asm  jcgryext-sse2.asm  jdcolext-sse2.asm  jdmrgext-sse2.asm \
+	jccolext-sse2-64.asm  jcgryext-sse2-64.asm  jdcolext-sse2-64.asm \
+	jdmrgext-sse2-64.asm  jccolext-altivec.c    jcgryext-altivec.c \
+	jdcolext-altivec.c    jdmrgext-altivec.c
+
+if SIMD_X86_64
+
+libsimd_la_SOURCES = jsimd_x86_64.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jfdctflt-sse-64.asm \
+	jccolor-sse2-64.asm   jcgray-sse2-64.asm    jchuff-sse2-64.asm \
+	jcsample-sse2-64.asm  jdcolor-sse2-64.asm   jdmerge-sse2-64.asm \
+	jdsample-sse2-64.asm  jfdctfst-sse2-64.asm  jfdctint-sse2-64.asm \
+	jidctflt-sse2-64.asm  jidctfst-sse2-64.asm  jidctint-sse2-64.asm \
+	jidctred-sse2-64.asm  jquantf-sse2-64.asm   jquanti-sse2-64.asm
+
+jccolor-sse2-64.lo:  jccolext-sse2-64.asm
+jcgray-sse2-64.lo:   jcgryext-sse2-64.asm
+jdcolor-sse2-64.lo:  jdcolext-sse2-64.asm
+jdmerge-sse2-64.lo:  jdmrgext-sse2-64.asm
+
+endif
+
+if SIMD_I386
+
+libsimd_la_SOURCES = jsimd_i386.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jsimdcpu.asm \
+	jfdctflt-3dn.asm   jidctflt-3dn.asm   jquant-3dn.asm \
+	jccolor-mmx.asm    jcgray-mmx.asm     jcsample-mmx.asm \
+	jdcolor-mmx.asm    jdmerge-mmx.asm    jdsample-mmx.asm \
+	jfdctfst-mmx.asm   jfdctint-mmx.asm   jidctfst-mmx.asm \
+	jidctint-mmx.asm   jidctred-mmx.asm   jquant-mmx.asm \
+	jfdctflt-sse.asm   jidctflt-sse.asm   jquant-sse.asm \
+	jccolor-sse2.asm   jcgray-sse2.asm    jchuff-sse2.asm \
+	jcsample-sse2.asm  jdcolor-sse2.asm   jdmerge-sse2.asm \
+	jdsample-sse2.asm  jfdctfst-sse2.asm  jfdctint-sse2.asm \
+	jidctflt-sse2.asm  jidctfst-sse2.asm  jidctint-sse2.asm \
+	jidctred-sse2.asm  jquantf-sse2.asm   jquanti-sse2.asm
+
+jccolor-mmx.lo:   jccolext-mmx.asm
+jcgray.-mmx.lo:   jcgryext-mmx.asm
+jdcolor-mmx.lo:   jdcolext-mmx.asm
+jdmerge-mmx.lo:   jdmrgext-mmx.asm
+jccolor-sse2.lo:  jccolext-sse2.asm
+jcgray-sse2.lo:   jcgryext-sse2.asm
+jdcolor-sse2.lo:  jdcolext-sse2.asm
+jdmerge-sse2.lo:  jdmrgext-sse2.asm
+
+endif
+
+if SIMD_ARM
+
+libsimd_la_SOURCES = jsimd_arm.c jsimd_arm_neon.S
+
+endif
+
+if SIMD_ARM_64
+
+libsimd_la_SOURCES = jsimd_arm64.c jsimd_arm64_neon.S
+
+endif
+
+if SIMD_MIPS
+
+libsimd_la_SOURCES = jsimd_mips.c jsimd_mips_dspr2_asm.h jsimd_mips_dspr2.S
+
+endif
+
+if SIMD_POWERPC
+
+libsimd_la_SOURCES = jsimd_powerpc.c jsimd_altivec.h jcsample.h \
+	jccolor-altivec.c     jcgray-altivec.c      jcsample-altivec.c \
+	jdcolor-altivec.c     jdmerge-altivec.c     jdsample-altivec.c \
+	jfdctfst-altivec.c    jfdctint-altivec.c \
+	jidctfst-altivec.c    jidctint-altivec.c \
+	jquanti-altivec.c
+libsimd_la_CFLAGS = -maltivec
+
+jccolor-altivec.lo:  jccolext-altivec.c
+jcgray-altivec.lo:   jcgryext-altivec.c
+jdcolor-altivec.lo:  jdcolext-altivec.c
+jdmerge-altivec.lo:  jdmrgext-altivec.c
+
+endif
+
+AM_CPPFLAGS = -I$(top_srcdir)
+
+.asm.lo:
+	$(AM_V_GEN) $(LIBTOOL) $(AM_V_lt) --mode=compile --tag NASM $(srcdir)/nasm_lt.sh $(AM_V_lt) $(NASM) $(NAFLAGS) -I$(srcdir) -I. $< -o $@
+
+jsimdcfg.inc: $(srcdir)/jsimdcfg.inc.h ../jpeglib.h ../jconfig.h ../jmorecfg.h
+	$(AM_V_GEN) $(CPP) -I$(top_builddir) -I$(top_builddir)/simd $(srcdir)/jsimdcfg.inc.h | $(EGREP) "^[\;%]|^\ %" | sed 's%_cpp_protection_%%' | sed 's@% define@%define@g' > $@
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/nasm_lt.sh glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/nasm_lt.sh
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/simd/nasm_lt.sh	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/simd/nasm_lt.sh	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,60 @@
+#! /bin/sh
+command=""
+infile=""
+o_opt=no
+pic=no
+while [ $# -gt 0 ]; do
+    case "$1" in
+        --silent)
+            exec > /dev/null
+            ;;
+        -DPIC|-fPIC|-fpic|-Kpic|-KPIC)
+            if [ "$pic" != "yes" ] ; then
+                command="$command -DPIC"
+                pic=yes
+            fi
+            ;;
+        -f|-fbin|-faout|-faoutb|-fcoff|-felf|-felf64|-fas86| \
+        -fobj|-fwin32|-fwin64|-frdf|-fieee|-fmacho|-fmacho64)
+            # it's a file format specifier for nasm.
+            command="$command $1"
+            ;;
+        -f*)
+            # maybe a code-generation flag for gcc.
+            ;;
+        -[Ii]*)
+            incdir=`echo "$1" | sed 's/^-[Ii]//'`
+            if [ "x$incdir" = x -a "x$2" != x ] ; then
+                case "$2" in
+                    -*) ;;
+                    *) incdir="$2"; shift;;
+                esac
+            fi
+            if [ "x$incdir" != x ] ; then
+                # In the case of NASM, the trailing slash is necessary.
+                incdir=`echo "$incdir" | sed 's%/*$%/%'`
+                command="$command -I$incdir"
+            fi
+            ;;
+        -o*)
+            o_opt=yes
+            command="$command $1"
+            ;;
+        *.asm)
+            infile=$1
+            command="$command $1"
+            ;;
+        *)
+            command="$command $1"
+            ;;
+    esac
+    shift
+done
+if [ "$o_opt" != yes ] ; then
+    # By default, NASM creates an output file
+    # in the same directory as the input file.
+    outfile="-o `echo $infile | sed -e 's%^.*/%%' -e 's%\.[^.]*$%%'`.o"
+    command="$command $outfile"
+fi
+echo $command
+exec $command
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/structure.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/structure.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/structure.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/structure.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,906 @@
+IJG JPEG LIBRARY:  SYSTEM ARCHITECTURE
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file provides an overview of the architecture of the IJG JPEG software;
+that is, the functions of the various modules in the system and the interfaces
+between modules.  For more precise details about any data structure or calling
+convention, see the include files and comments in the source code.
+
+We assume that the reader is already somewhat familiar with the JPEG standard.
+The README.ijg file includes references for learning about JPEG.  The file
+libjpeg.txt describes the library from the viewpoint of an application
+programmer using the library; it's best to read that file before this one.
+Also, the file coderules.txt describes the coding style conventions we use.
+
+In this document, JPEG-specific terminology follows the JPEG standard:
+  A "component" means a color channel, e.g., Red or Luminance.
+  A "sample" is a single component value (i.e., one number in the image data).
+  A "coefficient" is a frequency coefficient (a DCT transform output number).
+  A "block" is an 8x8 group of samples or coefficients.
+  An "MCU" (minimum coded unit) is an interleaved set of blocks of size
+        determined by the sampling factors, or a single block in a
+        noninterleaved scan.
+We do not use the terms "pixel" and "sample" interchangeably.  When we say
+pixel, we mean an element of the full-size image, while a sample is an element
+of the downsampled image.  Thus the number of samples may vary across
+components while the number of pixels does not.  (This terminology is not used
+rigorously throughout the code, but it is used in places where confusion would
+otherwise result.)
+
+
+*** System features ***
+
+The IJG distribution contains two parts:
+  * A subroutine library for JPEG compression and decompression.
+  * cjpeg/djpeg, two sample applications that use the library to transform
+    JFIF JPEG files to and from several other image formats.
+cjpeg/djpeg are of no great intellectual complexity: they merely add a simple
+command-line user interface and I/O routines for several uncompressed image
+formats.  This document concentrates on the library itself.
+
+We desire the library to be capable of supporting all JPEG baseline, extended
+sequential, and progressive DCT processes.  Hierarchical processes are not
+supported.
+
+The library does not support the lossless (spatial) JPEG process.  Lossless
+JPEG shares little or no code with lossy JPEG, and would normally be used
+without the extensive pre- and post-processing provided by this library.
+We feel that lossless JPEG is better handled by a separate library.
+
+Within these limits, any set of compression parameters allowed by the JPEG
+spec should be readable for decompression.  (We can be more restrictive about
+what formats we can generate.)  Although the system design allows for all
+parameter values, some uncommon settings are not yet implemented and may
+never be; nonintegral sampling ratios are the prime example.  Furthermore,
+we treat 8-bit vs. 12-bit data precision as a compile-time switch, not a
+run-time option, because most machines can store 8-bit pixels much more
+compactly than 12-bit.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, libtiff uses this
+library to implement JPEG compression within the TIFF file format.)
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  This code can be omitted if not needed.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+
+*** System overview ***
+
+The compressor and decompressor are each divided into two main sections:
+the JPEG compressor or decompressor proper, and the preprocessing or
+postprocessing functions.  The interface between these two sections is the
+image data that the official JPEG spec regards as its input or output: this
+data is in the colorspace to be used for compression, and it is downsampled
+to the sampling factors to be used.  The preprocessing and postprocessing
+steps are responsible for converting a normal image representation to or from
+this form.  (Those few applications that want to deal with YCbCr downsampled
+data can skip the preprocessing or postprocessing step.)
+
+Looking more closely, the compressor library contains the following main
+elements:
+
+  Preprocessing:
+    * Color space conversion (e.g., RGB to YCbCr).
+    * Edge expansion and downsampling.  Optionally, this step can do simple
+      smoothing --- this is often helpful for low-quality source data.
+  JPEG proper:
+    * MCU assembly, DCT, quantization.
+    * Entropy coding (sequential or progressive, Huffman or arithmetic).
+
+In addition to these modules we need overall control, marker generation,
+and support code (memory management & error handling).  There is also a
+module responsible for physically writing the output data --- typically
+this is just an interface to fwrite(), but some applications may need to
+do something else with the data.
+
+The decompressor library contains the following main elements:
+
+  JPEG proper:
+    * Entropy decoding (sequential or progressive, Huffman or arithmetic).
+    * Dequantization, inverse DCT, MCU disassembly.
+  Postprocessing:
+    * Upsampling.  Optionally, this step may be able to do more general
+      rescaling of the image.
+    * Color space conversion (e.g., YCbCr to RGB).  This step may also
+      provide gamma adjustment [ currently it does not ].
+    * Optional color quantization (e.g., reduction to 256 colors).
+    * Optional color precision reduction (e.g., 24-bit to 15-bit color).
+      [This feature is not currently implemented.]
+
+We also need overall control, marker parsing, and a data source module.
+The support code (memory management & error handling) can be shared with
+the compression half of the library.
+
+There may be several implementations of each of these elements, particularly
+in the decompressor, where a wide range of speed/quality tradeoffs is very
+useful.  It must be understood that some of the best speedups involve
+merging adjacent steps in the pipeline.  For example, upsampling, color space
+conversion, and color quantization might all be done at once when using a
+low-quality ordered-dither technique.  The system architecture is designed to
+allow such merging where appropriate.
+
+
+Note: it is convenient to regard edge expansion (padding to block boundaries)
+as a preprocessing/postprocessing function, even though the JPEG spec includes
+it in compression/decompression.  We do this because downsampling/upsampling
+can be simplified a little if they work on padded data: it's not necessary to
+have special cases at the right and bottom edges.  Therefore the interface
+buffer is always an integral number of blocks wide and high, and we expect
+compression preprocessing to pad the source data properly.  Padding will occur
+only to the next block (8-sample) boundary.  In an interleaved-scan situation,
+additional dummy blocks may be used to fill out MCUs, but the MCU assembly and
+disassembly logic will create or discard these blocks internally.  (This is
+advantageous for speed reasons, since we avoid DCTing the dummy blocks.
+It also permits a small reduction in file size, because the compressor can
+choose dummy block contents so as to minimize their size in compressed form.
+Finally, it makes the interface buffer specification independent of whether
+the file is actually interleaved or not.)  Applications that wish to deal
+directly with the downsampled data must provide similar buffering and padding
+for odd-sized images.
+
+
+*** Poor man's object-oriented programming ***
+
+It should be clear by now that we have a lot of quasi-independent processing
+steps, many of which have several possible behaviors.  To avoid cluttering the
+code with lots of switch statements, we use a simple form of object-style
+programming to separate out the different possibilities.
+
+For example, two different color quantization algorithms could be implemented
+as two separate modules that present the same external interface; at runtime,
+the calling code will access the proper module indirectly through an "object".
+
+We can get the limited features we need while staying within portable C.
+The basic tool is a function pointer.  An "object" is just a struct
+containing one or more function pointer fields, each of which corresponds to
+a method name in real object-oriented languages.  During initialization we
+fill in the function pointers with references to whichever module we have
+determined we need to use in this run.  Then invocation of the module is done
+by indirecting through a function pointer; on most machines this is no more
+expensive than a switch statement, which would be the only other way of
+making the required run-time choice.  The really significant benefit, of
+course, is keeping the source code clean and well structured.
+
+We can also arrange to have private storage that varies between different
+implementations of the same kind of object.  We do this by making all the
+module-specific object structs be separately allocated entities, which will
+be accessed via pointers in the master compression or decompression struct.
+The "public" fields or methods for a given kind of object are specified by
+a commonly known struct.  But a module's initialization code can allocate
+a larger struct that contains the common struct as its first member, plus
+additional private fields.  With appropriate pointer casting, the module's
+internal functions can access these private fields.  (For a simple example,
+see jdatadst.c, which implements the external interface specified by struct
+jpeg_destination_mgr, but adds extra fields.)
+
+(Of course this would all be a lot easier if we were using C++, but we are
+not yet prepared to assume that everyone has a C++ compiler.)
+
+An important benefit of this scheme is that it is easy to provide multiple
+versions of any method, each tuned to a particular case.  While a lot of
+precalculation might be done to select an optimal implementation of a method,
+the cost per invocation is constant.  For example, the upsampling step might
+have a "generic" method, plus one or more "hardwired" methods for the most
+popular sampling factors; the hardwired methods would be faster because they'd
+use straight-line code instead of for-loops.  The cost to determine which
+method to use is paid only once, at startup, and the selection criteria are
+hidden from the callers of the method.
+
+This plan differs a little bit from usual object-oriented structures, in that
+only one instance of each object class will exist during execution.  The
+reason for having the class structure is that on different runs we may create
+different instances (choose to execute different modules).  You can think of
+the term "method" as denoting the common interface presented by a particular
+set of interchangeable functions, and "object" as denoting a group of related
+methods, or the total shared interface behavior of a group of modules.
+
+
+*** Overall control structure ***
+
+We previously mentioned the need for overall control logic in the compression
+and decompression libraries.  In IJG implementations prior to v5, overall
+control was mostly provided by "pipeline control" modules, which proved to be
+large, unwieldy, and hard to understand.  To improve the situation, the
+control logic has been subdivided into multiple modules.  The control modules
+consist of:
+
+1. Master control for module selection and initialization.  This has two
+responsibilities:
+
+   1A.  Startup initialization at the beginning of image processing.
+        The individual processing modules to be used in this run are selected
+        and given initialization calls.
+
+   1B.  Per-pass control.  This determines how many passes will be performed
+        and calls each active processing module to configure itself
+        appropriately at the beginning of each pass.  End-of-pass processing,
+        where necessary, is also invoked from the master control module.
+
+   Method selection is partially distributed, in that a particular processing
+   module may contain several possible implementations of a particular method,
+   which it will select among when given its initialization call.  The master
+   control code need only be concerned with decisions that affect more than
+   one module.
+
+2. Data buffering control.  A separate control module exists for each
+   inter-processing-step data buffer.  This module is responsible for
+   invoking the processing steps that write or read that data buffer.
+
+Each buffer controller sees the world as follows:
+
+input data => processing step A => buffer => processing step B => output data
+                      |              |               |
+              ------------------ controller ------------------
+
+The controller knows the dataflow requirements of steps A and B: how much data
+they want to accept in one chunk and how much they output in one chunk.  Its
+function is to manage its buffer and call A and B at the proper times.
+
+A data buffer control module may itself be viewed as a processing step by a
+higher-level control module; thus the control modules form a binary tree with
+elementary processing steps at the leaves of the tree.
+
+The control modules are objects.  A considerable amount of flexibility can
+be had by replacing implementations of a control module.  For example:
+* Merging of adjacent steps in the pipeline is done by replacing a control
+  module and its pair of processing-step modules with a single processing-
+  step module.  (Hence the possible merges are determined by the tree of
+  control modules.)
+* In some processing modes, a given interstep buffer need only be a "strip"
+  buffer large enough to accommodate the desired data chunk sizes.  In other
+  modes, a full-image buffer is needed and several passes are required.
+  The control module determines which kind of buffer is used and manipulates
+  virtual array buffers as needed.  One or both processing steps may be
+  unaware of the multi-pass behavior.
+
+In theory, we might be able to make all of the data buffer controllers
+interchangeable and provide just one set of implementations for all.  In
+practice, each one contains considerable special-case processing for its
+particular job.  The buffer controller concept should be regarded as an
+overall system structuring principle, not as a complete description of the
+task performed by any one controller.
+
+
+*** Compression object structure ***
+
+Here is a sketch of the logical structure of the JPEG compression library:
+
+                                                 |-- Colorspace conversion
+                  |-- Preprocessing controller --|
+                  |                              |-- Downsampling
+Main controller --|
+                  |                            |-- Forward DCT, quantize
+                  |-- Coefficient controller --|
+                                               |-- Entropy encoding
+
+This sketch also describes the flow of control (subroutine calls) during
+typical image data processing.  Each of the components shown in the diagram is
+an "object" which may have several different implementations available.  One
+or more source code files contain the actual implementation(s) of each object.
+
+The objects shown above are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the preprocessed input data.  This controller invokes preprocessing to
+  fill the subsampled-data buffer, and JPEG compression to empty it.  There is
+  usually no need for a full-image buffer here; a strip buffer is adequate.
+
+* Preprocessing controller: buffer controller for the downsampling input data
+  buffer, which lies between colorspace conversion and downsampling.  Note
+  that a unified conversion/downsampling module would probably replace this
+  controller entirely.
+
+* Colorspace conversion: converts application image data into the desired
+  JPEG color space; also changes the data from pixel-interleaved layout to
+  separate component planes.  Processes one pixel row at a time.
+
+* Downsampling: performs reduction of chroma components as required.
+  Optionally may perform pixel-level smoothing as well.  Processes a "row
+  group" at a time, where a row group is defined as Vmax pixel rows of each
+  component before downsampling, and Vk sample rows afterwards (remember Vk
+  differs across components).  Some downsampling or smoothing algorithms may
+  require context rows above and below the current row group; the
+  preprocessing controller is responsible for supplying these rows via proper
+  buffering.  The downsampler is responsible for edge expansion at the right
+  edge (i.e., extending each sample row to a multiple of 8 samples); but the
+  preprocessing controller is responsible for vertical edge expansion (i.e.,
+  duplicating the bottom sample row as needed to make a multiple of 8 rows).
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU assembly, including insertion of dummy DCT
+  blocks when needed at the right or bottom edge.  When performing
+  Huffman-code optimization or emitting a multiscan JPEG file, this
+  controller is responsible for buffering the full image.  The equivalent of
+  one fully interleaved MCU row of subsampled data is processed per call,
+  even when the JPEG file is noninterleaved.
+
+* Forward DCT and quantization: Perform DCT, quantize, and emit coefficients.
+  Works on one or more DCT blocks at a time.  (Note: the coefficients are now
+  emitted in normal array order, which the entropy encoder is expected to
+  convert to zigzag order as necessary.  Prior versions of the IJG code did
+  the conversion to zigzag order within the quantization step.)
+
+* Entropy encoding: Perform Huffman or arithmetic entropy coding and emit the
+  coded data to the data destination module.  Works on one MCU per call.
+  For progressive JPEG, the same DCT blocks are fed to the entropy coder
+  during each pass, and the coder must emit the appropriate subset of
+  coefficients.
+
+In addition to the above objects, the compression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.
+
+* Marker writing: generates JPEG markers (except for RSTn, which is emitted
+  by the entropy encoder when needed).
+
+* Data destination manager: writes the output JPEG datastream to its final
+  destination (e.g., a file).  The destination manager supplied with the
+  library knows how to write to a stdio stream or to a memory buffer;
+  for other behaviors, the surrounding application may provide its own
+  destination manager.
+
+* Memory manager: allocates and releases memory, controls virtual arrays
+  (with backing store management, where required).
+
+* Error handler: performs formatting and output of error and trace messages;
+  determines handling of nonfatal errors.  The surrounding application may
+  override some or all of this object's methods to change error handling.
+
+* Progress monitor: supports output of "percent-done" progress reports.
+  This object represents an optional callback to the surrounding application:
+  if wanted, it must be supplied by the application.
+
+The error handler, destination manager, and progress monitor objects are
+defined as separate objects in order to simplify application-specific
+customization of the JPEG library.  A surrounding application may override
+individual methods or supply its own all-new implementation of one of these
+objects.  The object interfaces for these objects are therefore treated as
+part of the application interface of the library, whereas the other objects
+are internal to the library.
+
+The error handler and memory manager are shared by JPEG compression and
+decompression; the progress monitor, if used, may be shared as well.
+
+
+*** Decompression object structure ***
+
+Here is a sketch of the logical structure of the JPEG decompression library:
+
+                                               |-- Entropy decoding
+                  |-- Coefficient controller --|
+                  |                            |-- Dequantize, Inverse DCT
+Main controller --|
+                  |                               |-- Upsampling
+                  |-- Postprocessing controller --|   |-- Colorspace conversion
+                                                  |-- Color quantization
+                                                  |-- Color precision reduction
+
+As before, this diagram also represents typical control flow.  The objects
+shown are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the output of JPEG decompression proper.  This controller's primary
+  task is to feed the postprocessing procedure.  Some upsampling algorithms
+  may require context rows above and below the current row group; when this
+  is true, the main controller is responsible for managing its buffer so as
+  to make context rows available.  In the current design, the main buffer is
+  always a strip buffer; a full-image buffer is never required.
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU disassembly, including deletion of any dummy
+  DCT blocks at the right or bottom edge.  When reading a multiscan JPEG
+  file, this controller is responsible for buffering the full image.
+  (Buffering DCT coefficients, rather than samples, is necessary to support
+  progressive JPEG.)  The equivalent of one fully interleaved MCU row of
+  subsampled data is processed per call, even when the source JPEG file is
+  noninterleaved.
+
+* Entropy decoding: Read coded data from the data source module and perform
+  Huffman or arithmetic entropy decoding.  Works on one MCU per call.
+  For progressive JPEG decoding, the coefficient controller supplies the prior
+  coefficients of each MCU (initially all zeroes), which the entropy decoder
+  modifies in each scan.
+
+* Dequantization and inverse DCT: like it says.  Note that the coefficients
+  buffered by the coefficient controller have NOT been dequantized; we
+  merge dequantization and inverse DCT into a single step for speed reasons.
+  When scaled-down output is asked for, simplified DCT algorithms may be used
+  that emit fewer samples per DCT block, not the full 8x8.  Works on one DCT
+  block at a time.
+
+* Postprocessing controller: buffer controller for the color quantization
+  input buffer, when quantization is in use.  (Without quantization, this
+  controller just calls the upsampler.)  For two-pass quantization, this
+  controller is responsible for buffering the full-image data.
+
+* Upsampling: restores chroma components to full size.  (May support more
+  general output rescaling, too.  Note that if undersized DCT outputs have
+  been emitted by the DCT module, this module must adjust so that properly
+  sized outputs are created.)  Works on one row group at a time.  This module
+  also calls the color conversion module, so its top level is effectively a
+  buffer controller for the upsampling->color conversion buffer.  However, in
+  all but the highest-quality operating modes, upsampling and color
+  conversion are likely to be merged into a single step.
+
+* Colorspace conversion: convert from JPEG color space to output color space,
+  and change data layout from separate component planes to pixel-interleaved.
+  Works on one pixel row at a time.
+
+* Color quantization: reduce the data to colormapped form, using either an
+  externally specified colormap or an internally generated one.  This module
+  is not used for full-color output.  Works on one pixel row at a time; may
+  require two passes to generate a color map.  Note that the output will
+  always be a single component representing colormap indexes.  In the current
+  design, the output values are JSAMPLEs, so an 8-bit compilation cannot
+  quantize to more than 256 colors.  This is unlikely to be a problem in
+  practice.
+
+* Color reduction: this module handles color precision reduction, e.g.,
+  generating 15-bit color (5 bits/primary) from JPEG's 24-bit output.
+  Not quite clear yet how this should be handled... should we merge it with
+  colorspace conversion???
+
+Note that some high-speed operating modes might condense the entire
+postprocessing sequence to a single module (upsample, color convert, and
+quantize in one step).
+
+In addition to the above objects, the decompression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.  This is subdivided into
+  input and output control: jdinput.c controls only input-side processing,
+  while jdmaster.c handles overall initialization and output-side control.
+
+* Marker reading: decodes JPEG markers (except for RSTn).
+
+* Data source manager: supplies the input JPEG datastream.  The source
+  manager supplied with the library knows how to read from a stdio stream
+  or from a memory buffer;  for other behaviors, the surrounding application
+  may provide its own source manager.
+
+* Memory manager: same as for compression library.
+
+* Error handler: same as for compression library.
+
+* Progress monitor: same as for compression library.
+
+As with compression, the data source manager, error handler, and progress
+monitor are candidates for replacement by a surrounding application.
+
+
+*** Decompression input and output separation ***
+
+To support efficient incremental display of progressive JPEG files, the
+decompressor is divided into two sections that can run independently:
+
+1. Data input includes marker parsing, entropy decoding, and input into the
+   coefficient controller's DCT coefficient buffer.  Note that this
+   processing is relatively cheap and fast.
+
+2. Data output reads from the DCT coefficient buffer and performs the IDCT
+   and all postprocessing steps.
+
+For a progressive JPEG file, the data input processing is allowed to get
+arbitrarily far ahead of the data output processing.  (This occurs only
+if the application calls jpeg_consume_input(); otherwise input and output
+run in lockstep, since the input section is called only when the output
+section needs more data.)  In this way the application can avoid making
+extra display passes when data is arriving faster than the display pass
+can run.  Furthermore, it is possible to abort an output pass without
+losing anything, since the coefficient buffer is read-only as far as the
+output section is concerned.  See libjpeg.txt for more detail.
+
+A full-image coefficient array is only created if the JPEG file has multiple
+scans (or if the application specifies buffered-image mode anyway).  When
+reading a single-scan file, the coefficient controller normally creates only
+a one-MCU buffer, so input and output processing must run in lockstep in this
+case.  jpeg_consume_input() is effectively a no-op in this situation.
+
+The main impact of dividing the decompressor in this fashion is that we must
+be very careful with shared variables in the cinfo data structure.  Each
+variable that can change during the course of decompression must be
+classified as belonging to data input or data output, and each section must
+look only at its own variables.  For example, the data output section may not
+depend on any of the variables that describe the current scan in the JPEG
+file, because these may change as the data input section advances into a new
+scan.
+
+The progress monitor is (somewhat arbitrarily) defined to treat input of the
+file as one pass when buffered-image mode is not used, and to ignore data
+input work completely when buffered-image mode is used.  Note that the
+library has no reliable way to predict the number of passes when dealing
+with a progressive JPEG file, nor can it predict the number of output passes
+in buffered-image mode.  So the work estimate is inherently bogus anyway.
+
+No comparable division is currently made in the compression library, because
+there isn't any real need for it.
+
+
+*** Data formats ***
+
+Arrays of pixel sample values use the following data structure:
+
+    typedef something JSAMPLE;          a pixel component value, 0..MAXJSAMPLE
+    typedef JSAMPLE *JSAMPROW;          ptr to a row of samples
+    typedef JSAMPROW *JSAMPARRAY;       ptr to a list of rows
+    typedef JSAMPARRAY *JSAMPIMAGE;     ptr to a list of color-component arrays
+
+The basic element type JSAMPLE will typically be one of unsigned char,
+(signed) char, or short.  Short will be used if samples wider than 8 bits are
+to be supported (this is a compile-time option).  Otherwise, unsigned char is
+used if possible.  If the compiler only supports signed chars, then it is
+necessary to mask off the value when reading.  Thus, all reads of JSAMPLE
+values must be coded as "GETJSAMPLE(value)", where the macro will be defined
+as "((value) & 0xFF)" on signed-char machines and "((int) (value))" elsewhere.
+
+With these conventions, JSAMPLE values can be assumed to be >= 0.  This helps
+simplify correct rounding during downsampling, etc.  The JPEG standard's
+specification that sample values run from -128..127 is accommodated by
+subtracting 128 from the sample value in the DCT step.  Similarly, during
+decompression the output of the IDCT step will be immediately shifted back to
+0..255.  (NB: different values are required when 12-bit samples are in use.
+The code is written in terms of MAXJSAMPLE and CENTERJSAMPLE, which will be
+defined as 255 and 128 respectively in an 8-bit implementation, and as 4095
+and 2048 in a 12-bit implementation.)
+
+We use a pointer per row, rather than a two-dimensional JSAMPLE array.  This
+choice costs only a small amount of memory and has several benefits:
+* Code using the data structure doesn't need to know the allocated width of
+  the rows.  This simplifies edge expansion/compression, since we can work
+  in an array that's wider than the logical picture width.
+* Indexing doesn't require multiplication; this is a performance win on many
+  machines.
+* Arrays with more than 64K total elements can be supported even on machines
+  where malloc() cannot allocate chunks larger than 64K.
+* The rows forming a component array may be allocated at different times
+  without extra copying.  This trick allows some speedups in smoothing steps
+  that need access to the previous and next rows.
+
+Note that each color component is stored in a separate array; we don't use the
+traditional layout in which the components of a pixel are stored together.
+This simplifies coding of modules that work on each component independently,
+because they don't need to know how many components there are.  Furthermore,
+we can read or write each component to a temporary file independently, which
+is helpful when dealing with noninterleaved JPEG files.
+
+In general, a specific sample value is accessed by code such as
+        GETJSAMPLE(image[colorcomponent][row][col])
+where col is measured from the image left edge, but row is measured from the
+first sample row currently in memory.  Either of the first two indexings can
+be precomputed by copying the relevant pointer.
+
+
+Since most image-processing applications prefer to work on images in which
+the components of a pixel are stored together, the data passed to or from the
+surrounding application uses the traditional convention: a single pixel is
+represented by N consecutive JSAMPLE values, and an image row is an array of
+(# of color components)*(image width) JSAMPLEs.  One or more rows of data can
+be represented by a pointer of type JSAMPARRAY in this scheme.  This scheme is
+converted to component-wise storage inside the JPEG library.  (Applications
+that want to skip JPEG preprocessing or postprocessing will have to contend
+with component-wise storage.)
+
+
+Arrays of DCT-coefficient values use the following data structure:
+
+    typedef short JCOEF;                a 16-bit signed integer
+    typedef JCOEF JBLOCK[DCTSIZE2];     an 8x8 block of coefficients
+    typedef JBLOCK *JBLOCKROW;          ptr to one horizontal row of 8x8 blocks
+    typedef JBLOCKROW *JBLOCKARRAY;     ptr to a list of such rows
+    typedef JBLOCKARRAY *JBLOCKIMAGE;   ptr to a list of color component arrays
+
+The underlying type is at least a 16-bit signed integer; while "short" is big
+enough on all machines of interest, on some machines it is preferable to use
+"int" for speed reasons, despite the storage cost.  Coefficients are grouped
+into 8x8 blocks (but we always use #defines DCTSIZE and DCTSIZE2 rather than
+"8" and "64").
+
+The contents of a coefficient block may be in either "natural" or zigzagged
+order, and may be true values or divided by the quantization coefficients,
+depending on where the block is in the processing pipeline.  In the current
+library, coefficient blocks are kept in natural order everywhere; the entropy
+codecs zigzag or dezigzag the data as it is written or read.  The blocks
+contain quantized coefficients everywhere outside the DCT/IDCT subsystems.
+(This latter decision may need to be revisited to support variable
+quantization a la JPEG Part 3.)
+
+Notice that the allocation unit is now a row of 8x8 blocks, corresponding to
+eight rows of samples.  Otherwise the structure is much the same as for
+samples, and for the same reasons.
+
+
+*** Suspendable processing ***
+
+In some applications it is desirable to use the JPEG library as an
+incremental, memory-to-memory filter.  In this situation the data source or
+destination may be a limited-size buffer, and we can't rely on being able to
+empty or refill the buffer at arbitrary times.  Instead the application would
+like to have control return from the library at buffer overflow/underrun, and
+then resume compression or decompression at a later time.
+
+This scenario is supported for simple cases.  (For anything more complex, we
+recommend that the application "bite the bullet" and develop real multitasking
+capability.)  The libjpeg.txt file goes into more detail about the usage and
+limitations of this capability; here we address the implications for library
+structure.
+
+The essence of the problem is that the entropy codec (coder or decoder) must
+be prepared to stop at arbitrary times.  In turn, the controllers that call
+the entropy codec must be able to stop before having produced or consumed all
+the data that they normally would handle in one call.  That part is reasonably
+straightforward: we make the controller call interfaces include "progress
+counters" which indicate the number of data chunks successfully processed, and
+we require callers to test the counter rather than just assume all of the data
+was processed.
+
+Rather than trying to restart at an arbitrary point, the current Huffman
+codecs are designed to restart at the beginning of the current MCU after a
+suspension due to buffer overflow/underrun.  At the start of each call, the
+codec's internal state is loaded from permanent storage (in the JPEG object
+structures) into local variables.  On successful completion of the MCU, the
+permanent state is updated.  (This copying is not very expensive, and may even
+lead to *improved* performance if the local variables can be registerized.)
+If a suspension occurs, the codec simply returns without updating the state,
+thus effectively reverting to the start of the MCU.  Note that this implies
+leaving some data unprocessed in the source/destination buffer (ie, the
+compressed partial MCU).  The data source/destination module interfaces are
+specified so as to make this possible.  This also implies that the data buffer
+must be large enough to hold a worst-case compressed MCU; a couple thousand
+bytes should be enough.
+
+In a successive-approximation AC refinement scan, the progressive Huffman
+decoder has to be able to undo assignments of newly nonzero coefficients if it
+suspends before the MCU is complete, since decoding requires distinguishing
+previously-zero and previously-nonzero coefficients.  This is a bit tedious
+but probably won't have much effect on performance.  Other variants of Huffman
+decoding need not worry about this, since they will just store the same values
+again if forced to repeat the MCU.
+
+This approach would probably not work for an arithmetic codec, since its
+modifiable state is quite large and couldn't be copied cheaply.  Instead it
+would have to suspend and resume exactly at the point of the buffer end.
+
+The JPEG marker reader is designed to cope with suspension at an arbitrary
+point.  It does so by backing up to the start of the marker parameter segment,
+so the data buffer must be big enough to hold the largest marker of interest.
+Again, a couple KB should be adequate.  (A special "skip" convention is used
+to bypass COM and APPn markers, so these can be larger than the buffer size
+without causing problems; otherwise a 64K buffer would be needed in the worst
+case.)
+
+The JPEG marker writer currently does *not* cope with suspension.
+We feel that this is not necessary; it is much easier simply to require
+the application to ensure there is enough buffer space before starting.  (An
+empty 2K buffer is more than sufficient for the header markers; and ensuring
+there are a dozen or two bytes available before calling jpeg_finish_compress()
+will suffice for the trailer.)  This would not work for writing multi-scan
+JPEG files, but we simply do not intend to support that capability with
+suspension.
+
+
+*** Memory manager services ***
+
+The JPEG library's memory manager controls allocation and deallocation of
+memory, and it manages large "virtual" data arrays on machines where the
+operating system does not provide virtual memory.  Note that the same
+memory manager serves both compression and decompression operations.
+
+In all cases, allocated objects are tied to a particular compression or
+decompression master record, and they will be released when that master
+record is destroyed.
+
+The memory manager does not provide explicit deallocation of objects.
+Instead, objects are created in "pools" of free storage, and a whole pool
+can be freed at once.  This approach helps prevent storage-leak bugs, and
+it speeds up operations whenever malloc/free are slow (as they often are).
+The pools can be regarded as lifetime identifiers for objects.  Two
+pools/lifetimes are defined:
+  * JPOOL_PERMANENT     lasts until master record is destroyed
+  * JPOOL_IMAGE         lasts until done with image (JPEG datastream)
+Permanent lifetime is used for parameters and tables that should be carried
+across from one datastream to another; this includes all application-visible
+parameters.  Image lifetime is used for everything else.  (A third lifetime,
+JPOOL_PASS = one processing pass, was originally planned.  However it was
+dropped as not being worthwhile.  The actual usage patterns are such that the
+peak memory usage would be about the same anyway; and having per-pass storage
+substantially complicates the virtual memory allocation rules --- see below.)
+
+The memory manager deals with three kinds of object:
+1. "Small" objects.  Typically these require no more than 10K-20K total.
+2. "Large" objects.  These may require tens to hundreds of K depending on
+   image size.  Semantically they behave the same as small objects, but we
+   distinguish them because pool allocation heuristics may differ for large and
+   small objects (historically, large objects were also referenced by far
+   pointers on MS-DOS machines.)  Note that individual "large" objects cannot
+   exceed the size allowed by type size_t, which may be 64K or less on some
+   machines.
+3. "Virtual" objects.  These are large 2-D arrays of JSAMPLEs or JBLOCKs
+   (typically large enough for the entire image being processed).  The
+   memory manager provides stripwise access to these arrays.  On machines
+   without virtual memory, the rest of the array may be swapped out to a
+   temporary file.
+
+(Note: JSAMPARRAY and JBLOCKARRAY data structures are a combination of large
+objects for the data proper and small objects for the row pointers.  For
+convenience and speed, the memory manager provides single routines to create
+these structures.  Similarly, virtual arrays include a small control block
+and a JSAMPARRAY or JBLOCKARRAY working buffer, all created with one call.)
+
+In the present implementation, virtual arrays are only permitted to have image
+lifespan.  (Permanent lifespan would not be reasonable, and pass lifespan is
+not very useful since a virtual array's raison d'etre is to store data for
+multiple passes through the image.)  We also expect that only "small" objects
+will be given permanent lifespan, though this restriction is not required by
+the memory manager.
+
+In a non-virtual-memory machine, some performance benefit can be gained by
+making the in-memory buffers for virtual arrays be as large as possible.
+(For small images, the buffers might fit entirely in memory, so blind
+swapping would be very wasteful.)  The memory manager will adjust the height
+of the buffers to fit within a prespecified maximum memory usage.  In order
+to do this in a reasonably optimal fashion, the manager needs to allocate all
+of the virtual arrays at once.  Therefore, there isn't a one-step allocation
+routine for virtual arrays; instead, there is a "request" routine that simply
+allocates the control block, and a "realize" routine (called just once) that
+determines space allocation and creates all of the actual buffers.  The
+realize routine must allow for space occupied by non-virtual large objects.
+(We don't bother to factor in the space needed for small objects, on the
+grounds that it isn't worth the trouble.)
+
+To support all this, we establish the following protocol for doing business
+with the memory manager:
+  1. Modules must request virtual arrays (which may have only image lifespan)
+     during the initial setup phase, i.e., in their jinit_xxx routines.
+  2. All "large" objects (including JSAMPARRAYs and JBLOCKARRAYs) must also be
+     allocated during initial setup.
+  3. realize_virt_arrays will be called at the completion of initial setup.
+     The above conventions ensure that sufficient information is available
+     for it to choose a good size for virtual array buffers.
+Small objects of any lifespan may be allocated at any time.  We expect that
+the total space used for small objects will be small enough to be negligible
+in the realize_virt_arrays computation.
+
+In a virtual-memory machine, we simply pretend that the available space is
+infinite, thus causing realize_virt_arrays to decide that it can allocate all
+the virtual arrays as full-size in-memory buffers.  The overhead of the
+virtual-array access protocol is very small when no swapping occurs.
+
+A virtual array can be specified to be "pre-zeroed"; when this flag is set,
+never-yet-written sections of the array are set to zero before being made
+available to the caller.  If this flag is not set, never-written sections
+of the array contain garbage.  (This feature exists primarily because the
+equivalent logic would otherwise be needed in jdcoefct.c for progressive
+JPEG mode; we may as well make it available for possible other uses.)
+
+The first write pass on a virtual array is required to occur in top-to-bottom
+order; read passes, as well as any write passes after the first one, may
+access the array in any order.  This restriction exists partly to simplify
+the virtual array control logic, and partly because some file systems may not
+support seeking beyond the current end-of-file in a temporary file.  The main
+implication of this restriction is that rearrangement of rows (such as
+converting top-to-bottom data order to bottom-to-top) must be handled while
+reading data out of the virtual array, not while putting it in.
+
+
+*** Memory manager internal structure ***
+
+To isolate system dependencies as much as possible, we have broken the
+memory manager into two parts.  There is a reasonably system-independent
+"front end" (jmemmgr.c) and a "back end" that contains only the code
+likely to change across systems.  All of the memory management methods
+outlined above are implemented by the front end.  The back end provides
+the following routines for use by the front end (none of these routines
+are known to the rest of the JPEG code):
+
+jpeg_mem_init, jpeg_mem_term    system-dependent initialization/shutdown
+
+jpeg_get_small, jpeg_free_small interface to malloc and free library routines
+                                (or their equivalents)
+
+jpeg_get_large, jpeg_free_large historically was used to interface with
+                                FAR malloc/free on MS-DOS machines;  now the
+                                same as jpeg_get_small/jpeg_free_small
+
+jpeg_mem_available              estimate available memory
+
+jpeg_open_backing_store         create a backing-store object
+
+read_backing_store,             manipulate a backing-store object
+write_backing_store,
+close_backing_store
+
+On some systems there will be more than one type of backing-store object
+(specifically, in MS-DOS a backing store file might be an area of extended
+memory as well as a disk file).  jpeg_open_backing_store is responsible for
+choosing how to implement a given object.  The read/write/close routines
+are method pointers in the structure that describes a given object; this
+lets them be different for different object types.
+
+It may be necessary to ensure that backing store objects are explicitly
+released upon abnormal program termination.  For example, MS-DOS won't free
+extended memory by itself.  To support this, we will expect the main program
+or surrounding application to arrange to call self_destruct (typically via
+jpeg_destroy) upon abnormal termination.  This may require a SIGINT signal
+handler or equivalent.  We don't want to have the back end module install its
+own signal handler, because that would pre-empt the surrounding application's
+ability to control signal handling.
+
+The IJG distribution includes several memory manager back end implementations.
+Usually the same back end should be suitable for all applications on a given
+system, but it is possible for an application to supply its own back end at
+need.
+
+
+*** Implications of DNL marker ***
+
+Some JPEG files may use a DNL marker to postpone definition of the image
+height (this would be useful for a fax-like scanner's output, for instance).
+In these files the SOF marker claims the image height is 0, and you only
+find out the true image height at the end of the first scan.
+
+We could read these files as follows:
+1. Upon seeing zero image height, replace it by 65535 (the maximum allowed).
+2. When the DNL is found, update the image height in the global image
+   descriptor.
+This implies that control modules must avoid making copies of the image
+height, and must re-test for termination after each MCU row.  This would
+be easy enough to do.
+
+In cases where image-size data structures are allocated, this approach will
+result in very inefficient use of virtual memory or much-larger-than-necessary
+temporary files.  This seems acceptable for something that probably won't be a
+mainstream usage.  People might have to forgo use of memory-hogging options
+(such as two-pass color quantization or noninterleaved JPEG files) if they
+want efficient conversion of such files.  (One could improve efficiency by
+demanding a user-supplied upper bound for the height, less than 65536; in most
+cases it could be much less.)
+
+The standard also permits the SOF marker to overestimate the image height,
+with a DNL to give the true, smaller height at the end of the first scan.
+This would solve the space problems if the overestimate wasn't too great.
+However, it implies that you don't even know whether DNL will be used.
+
+This leads to a couple of very serious objections:
+1. Testing for a DNL marker must occur in the inner loop of the decompressor's
+   Huffman decoder; this implies a speed penalty whether the feature is used
+   or not.
+2. There is no way to hide the last-minute change in image height from an
+   application using the decoder.  Thus *every* application using the IJG
+   library would suffer a complexity penalty whether it cared about DNL or
+   not.
+We currently do not support DNL because of these problems.
+
+A different approach is to insist that DNL-using files be preprocessed by a
+separate program that reads ahead to the DNL, then goes back and fixes the SOF
+marker.  This is a much simpler solution and is probably far more efficient.
+Even if one wants piped input, buffering the first scan of the JPEG file needs
+a lot smaller temp file than is implied by the maximum-height method.  For
+this approach we'd simply treat DNL as a no-op in the decompressor (at most,
+check that it matches the SOF image height).
+
+We will not worry about making the compressor capable of outputting DNL.
+Something similar to the first scheme above could be applied if anyone ever
+wants to make that work.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbench.c glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbench.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbench.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbench.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,990 @@
+/*
+ * Copyright (C)2009-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+#include <math.h>
+#include <errno.h>
+#include <cdjpeg.h>
+#include "./bmp.h"
+#include "./tjutil.h"
+#include "./turbojpeg.h"
+
+
+#define _throw(op, err) {  \
+	printf("ERROR in line %d while %s:\n%s\n", __LINE__, op, err);  \
+  retval=-1;  goto bailout;}
+#define _throwunix(m) _throw(m, strerror(errno))
+#define _throwtj(m) _throw(m, tjGetErrorStr())
+#define _throwbmp(m) _throw(m, bmpgeterr())
+
+int flags=TJFLAG_NOREALLOC, componly=0, decomponly=0, doyuv=0, quiet=0,
+	dotile=0, pf=TJPF_BGR, yuvpad=1, warmup=1, dowrite=1;
+char *ext="ppm";
+const char *pixFormatStr[TJ_NUMPF]=
+{
+	"RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "GRAY", "", "", "", "", "CMYK"
+};
+const char *subNameLong[TJ_NUMSAMP]=
+{
+	"4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *csName[TJ_NUMCS]=
+{
+	"RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+};
+const char *subName[TJ_NUMSAMP]={"444", "422", "420", "GRAY", "440", "411"};
+tjscalingfactor *scalingfactors=NULL, sf={1, 1};  int nsf=0;
+int xformop=TJXOP_NONE, xformopt=0;
+int (*customFilter)(short *, tjregion, tjregion, int, int, tjtransform *);
+double benchtime=5.0;
+
+
+char *formatName(int subsamp, int cs, char *buf)
+{
+	if(cs==TJCS_YCbCr) return (char *)subNameLong[subsamp];
+	else if(cs==TJCS_YCCK)
+	{
+		snprintf(buf, 80, "%s %s", csName[cs], subNameLong[subsamp]);
+		return buf;
+	}
+	else return (char *)csName[cs];
+}
+
+
+char *sigfig(double val, int figs, char *buf, int len)
+{
+	char format[80];
+	int digitsafterdecimal=figs-(int)ceil(log10(fabs(val)));
+	if(digitsafterdecimal<1) snprintf(format, 80, "%%.0f");
+	else snprintf(format, 80, "%%.%df", digitsafterdecimal);
+	snprintf(buf, len, format, val);
+	return buf;
+}
+
+
+/* Custom DCT filter which produces a negative of the image */
+int dummyDCTFilter(short *coeffs, tjregion arrayRegion, tjregion planeRegion,
+	int componentIndex, int transformIndex, tjtransform *transform)
+{
+	int i;
+	for(i=0; i<arrayRegion.w*arrayRegion.h; i++) coeffs[i]=-coeffs[i];
+	return 0;
+}
+
+
+/* Decompression test */
+int decomp(unsigned char *srcbuf, unsigned char **jpegbuf,
+	unsigned long *jpegsize, unsigned char *dstbuf, int w, int h,
+	int subsamp, int jpegqual, char *filename, int tilew, int tileh)
+{
+	char tempstr[1024], sizestr[20]="\0", qualstr[6]="\0", *ptr;
+	FILE *file=NULL;  tjhandle handle=NULL;
+	int row, col, iter=0, dstbufalloc=0, retval=0;
+	double elapsed, elapsedDecode;
+	int ps=tjPixelSize[pf];
+	int scaledw=TJSCALED(w, sf);
+	int scaledh=TJSCALED(h, sf);
+	int pitch=scaledw*ps;
+	int ntilesw=(w+tilew-1)/tilew, ntilesh=(h+tileh-1)/tileh;
+	unsigned char *dstptr, *dstptr2, *yuvbuf=NULL;
+
+	if(jpegqual>0)
+	{
+		snprintf(qualstr, 6, "_Q%d", jpegqual);
+		qualstr[5]=0;
+	}
+
+	if((handle=tjInitDecompress())==NULL)
+		_throwtj("executing tjInitDecompress()");
+
+	if(dstbuf==NULL)
+	{
+		if((dstbuf=(unsigned char *)malloc(pitch*scaledh))==NULL)
+			_throwunix("allocating destination buffer");
+		dstbufalloc=1;
+	}
+	/* Set the destination buffer to gray so we know whether the decompressor
+	   attempted to write to it */
+	memset(dstbuf, 127, pitch*scaledh);
+
+	if(doyuv)
+	{
+		int width=dotile? tilew:scaledw;
+		int height=dotile? tileh:scaledh;
+		int yuvsize=tjBufSizeYUV2(width, yuvpad, height, subsamp);
+		if((yuvbuf=(unsigned char *)malloc(yuvsize))==NULL)
+			_throwunix("allocating YUV buffer");
+		memset(yuvbuf, 127, yuvsize);
+	}
+
+	/* Benchmark */
+	iter=-warmup;
+	elapsed=elapsedDecode=0.;
+	while(1)
+	{
+		int tile=0;
+		double start=gettime();
+		for(row=0, dstptr=dstbuf; row<ntilesh; row++, dstptr+=pitch*tileh)
+		{
+			for(col=0, dstptr2=dstptr; col<ntilesw; col++, tile++, dstptr2+=ps*tilew)
+			{
+				int width=dotile? min(tilew, w-col*tilew):scaledw;
+				int height=dotile? min(tileh, h-row*tileh):scaledh;
+				if(doyuv)
+				{
+					double startDecode;
+					if(tjDecompressToYUV2(handle, jpegbuf[tile], jpegsize[tile], yuvbuf,
+						width, yuvpad, height, flags)==-1)
+						_throwtj("executing tjDecompressToYUV2()");
+					startDecode=gettime();
+					if(tjDecodeYUV(handle, yuvbuf, yuvpad, subsamp, dstptr2, width,
+						pitch, height, pf, flags)==-1)
+						_throwtj("executing tjDecodeYUV()");
+					if(iter>=0) elapsedDecode+=gettime()-startDecode;
+				}
+				else
+					if(tjDecompress2(handle, jpegbuf[tile], jpegsize[tile], dstptr2,
+						width, pitch, height, pf, flags)==-1)
+						_throwtj("executing tjDecompress2()");
+			}
+		}
+		iter++;
+		if(iter>=1)
+		{
+			elapsed+=gettime()-start;
+			if(elapsed>=benchtime) break;
+		}
+	}
+	if(doyuv) elapsed-=elapsedDecode;
+
+	if(tjDestroy(handle)==-1) _throwtj("executing tjDestroy()");
+	handle=NULL;
+
+	if(quiet)
+	{
+		printf("%-6s%s",
+			sigfig((double)(w*h)/1000000.*(double)iter/elapsed, 4, tempstr, 1024),
+			quiet==2? "\n":"  ");
+		if(doyuv)
+			printf("%s\n",
+				sigfig((double)(w*h)/1000000.*(double)iter/elapsedDecode, 4, tempstr,
+					1024));
+		else if(quiet!=2) printf("\n");
+	}
+	else
+	{
+		printf("%s --> Frame rate:         %f fps\n",
+			doyuv? "Decomp to YUV":"Decompress   ", (double)iter/elapsed);
+		printf("                  Throughput:         %f Megapixels/sec\n",
+			(double)(w*h)/1000000.*(double)iter/elapsed);
+		if(doyuv)
+		{
+			printf("YUV Decode    --> Frame rate:         %f fps\n",
+				(double)iter/elapsedDecode);
+			printf("                  Throughput:         %f Megapixels/sec\n",
+				(double)(w*h)/1000000.*(double)iter/elapsedDecode);
+		}
+	}
+
+	if (!dowrite) goto bailout;
+
+	if(sf.num!=1 || sf.denom!=1)
+		snprintf(sizestr, 20, "%d_%d", sf.num, sf.denom);
+	else if(tilew!=w || tileh!=h)
+		snprintf(sizestr, 20, "%dx%d", tilew, tileh);
+	else snprintf(sizestr, 20, "full");
+	if(decomponly)
+		snprintf(tempstr, 1024, "%s_%s.%s", filename, sizestr, ext);
+	else
+		snprintf(tempstr, 1024, "%s_%s%s_%s.%s", filename, subName[subsamp],
+			qualstr, sizestr, ext);
+
+	if(savebmp(tempstr, dstbuf, scaledw, scaledh, pf,
+		(flags&TJFLAG_BOTTOMUP)!=0)==-1)
+		_throwbmp("saving bitmap");
+	ptr=strrchr(tempstr, '.');
+	snprintf(ptr, 1024-(ptr-tempstr), "-err.%s", ext);
+	if(srcbuf && sf.num==1 && sf.denom==1)
+	{
+		if(!quiet) printf("Compression error written to %s.\n", tempstr);
+		if(subsamp==TJ_GRAYSCALE)
+		{
+			int index, index2;
+			for(row=0, index=0; row<h; row++, index+=pitch)
+			{
+				for(col=0, index2=index; col<w; col++, index2+=ps)
+				{
+					int rindex=index2+tjRedOffset[pf];
+					int gindex=index2+tjGreenOffset[pf];
+					int bindex=index2+tjBlueOffset[pf];
+					int y=(int)((double)srcbuf[rindex]*0.299
+						+ (double)srcbuf[gindex]*0.587
+						+ (double)srcbuf[bindex]*0.114 + 0.5);
+					if(y>255) y=255;  if(y<0) y=0;
+					dstbuf[rindex]=abs(dstbuf[rindex]-y);
+					dstbuf[gindex]=abs(dstbuf[gindex]-y);
+					dstbuf[bindex]=abs(dstbuf[bindex]-y);
+				}
+			}
+		}
+		else
+		{
+			for(row=0; row<h; row++)
+				for(col=0; col<w*ps; col++)
+					dstbuf[pitch*row+col]
+						=abs(dstbuf[pitch*row+col]-srcbuf[pitch*row+col]);
+		}
+		if(savebmp(tempstr, dstbuf, w, h, pf,
+			(flags&TJFLAG_BOTTOMUP)!=0)==-1)
+			_throwbmp("saving bitmap");
+	}
+
+	bailout:
+	if(file) fclose(file);
+	if(handle) tjDestroy(handle);
+	if(dstbuf && dstbufalloc) free(dstbuf);
+	if(yuvbuf) free(yuvbuf);
+	return retval;
+}
+
+
+int fullTest(unsigned char *srcbuf, int w, int h, int subsamp, int jpegqual,
+	char *filename)
+{
+	char tempstr[1024], tempstr2[80];
+	FILE *file=NULL;  tjhandle handle=NULL;
+	unsigned char **jpegbuf=NULL, *yuvbuf=NULL, *tmpbuf=NULL, *srcptr, *srcptr2;
+	double start, elapsed, elapsedEncode;
+	int totaljpegsize=0, row, col, i, tilew=w, tileh=h, retval=0;
+	int iter, yuvsize=0;
+	unsigned long *jpegsize=NULL;
+	int ps=tjPixelSize[pf];
+	int ntilesw=1, ntilesh=1, pitch=w*ps;
+	const char *pfStr=pixFormatStr[pf];
+
+	if((tmpbuf=(unsigned char *)malloc(pitch*h)) == NULL)
+		_throwunix("allocating temporary image buffer");
+
+	if(!quiet)
+		printf(">>>>>  %s (%s) <--> JPEG %s Q%d  <<<<<\n", pfStr,
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-up":"Top-down", subNameLong[subsamp],
+			jpegqual);
+
+	for(tilew=dotile? 8:w, tileh=dotile? 8:h; ; tilew*=2, tileh*=2)
+	{
+		if(tilew>w) tilew=w;  if(tileh>h) tileh=h;
+		ntilesw=(w+tilew-1)/tilew;  ntilesh=(h+tileh-1)/tileh;
+
+		if((jpegbuf=(unsigned char **)malloc(sizeof(unsigned char *)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG tile array");
+		memset(jpegbuf, 0, sizeof(unsigned char *)*ntilesw*ntilesh);
+		if((jpegsize=(unsigned long *)malloc(sizeof(unsigned long)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG size array");
+		memset(jpegsize, 0, sizeof(unsigned long)*ntilesw*ntilesh);
+
+		if((flags&TJFLAG_NOREALLOC)!=0)
+			for(i=0; i<ntilesw*ntilesh; i++)
+			{
+				if((jpegbuf[i]=(unsigned char *)tjAlloc(tjBufSize(tilew, tileh,
+					subsamp)))==NULL)
+					_throwunix("allocating JPEG tiles");
+			}
+
+		/* Compression test */
+		if(quiet==1)
+			printf("%-4s (%s)  %-5s    %-3d   ", pfStr,
+				(flags&TJFLAG_BOTTOMUP)? "BU":"TD", subNameLong[subsamp], jpegqual);
+		for(i=0; i<h; i++)
+			memcpy(&tmpbuf[pitch*i], &srcbuf[w*ps*i], w*ps);
+		if((handle=tjInitCompress())==NULL)
+			_throwtj("executing tjInitCompress()");
+
+		if(doyuv)
+		{
+			yuvsize=tjBufSizeYUV2(tilew, yuvpad, tileh, subsamp);
+			if((yuvbuf=(unsigned char *)malloc(yuvsize))==NULL)
+				_throwunix("allocating YUV buffer");
+			memset(yuvbuf, 127, yuvsize);
+		}
+
+		/* Benchmark */
+		iter=-warmup;
+		elapsed=elapsedEncode=0.;
+		while(1)
+		{
+			int tile=0;
+			totaljpegsize=0;
+			start=gettime();
+			for(row=0, srcptr=srcbuf; row<ntilesh; row++, srcptr+=pitch*tileh)
+			{
+				for(col=0, srcptr2=srcptr; col<ntilesw; col++, tile++,
+					srcptr2+=ps*tilew)
+				{
+					int width=min(tilew, w-col*tilew);
+					int height=min(tileh, h-row*tileh);
+					if(doyuv)
+					{
+						double startEncode=gettime();
+						if(tjEncodeYUV3(handle, srcptr2, width, pitch, height, pf, yuvbuf,
+							yuvpad, subsamp, flags)==-1)
+							_throwtj("executing tjEncodeYUV3()");
+						if(iter>=0) elapsedEncode+=gettime()-startEncode;
+						if(tjCompressFromYUV(handle, yuvbuf, width, yuvpad, height,
+							subsamp, &jpegbuf[tile], &jpegsize[tile], jpegqual, flags)==-1)
+							_throwtj("executing tjCompressFromYUV()");
+					}
+					else
+					{
+						if(tjCompress2(handle, srcptr2, width, pitch, height, pf,
+							&jpegbuf[tile], &jpegsize[tile], subsamp, jpegqual, flags)==-1)
+							_throwtj("executing tjCompress2()");
+					}
+					totaljpegsize+=jpegsize[tile];
+				}
+			}
+			iter++;
+			if(iter>=1)
+			{
+				elapsed+=gettime()-start;
+				if(elapsed>=benchtime) break;
+			}
+		}
+		if(doyuv) elapsed-=elapsedEncode;
+
+		if(tjDestroy(handle)==-1) _throwtj("executing tjDestroy()");
+		handle=NULL;
+
+		if(quiet==1) printf("%-5d  %-5d   ", tilew, tileh);
+		if(quiet)
+		{
+			if(doyuv)
+				printf("%-6s%s",
+					sigfig((double)(w*h)/1000000.*(double)iter/elapsedEncode, 4, tempstr,
+						1024), quiet==2? "\n":"  ");
+			printf("%-6s%s",
+				sigfig((double)(w*h)/1000000.*(double)iter/elapsed, 4,	tempstr, 1024),
+				quiet==2? "\n":"  ");
+			printf("%-6s%s",
+				sigfig((double)(w*h*ps)/(double)totaljpegsize, 4, tempstr2, 80),
+				quiet==2? "\n":"  ");
+		}
+		else
+		{
+			printf("\n%s size: %d x %d\n", dotile? "Tile":"Image", tilew,
+				tileh);
+			if(doyuv)
+			{
+				printf("Encode YUV    --> Frame rate:         %f fps\n",
+					(double)iter/elapsedEncode);
+				printf("                  Output image size:  %d bytes\n", yuvsize);
+				printf("                  Compression ratio:  %f:1\n",
+					(double)(w*h*ps)/(double)yuvsize);
+				printf("                  Throughput:         %f Megapixels/sec\n",
+					(double)(w*h)/1000000.*(double)iter/elapsedEncode);
+				printf("                  Output bit stream:  %f Megabits/sec\n",
+					(double)yuvsize*8./1000000.*(double)iter/elapsedEncode);
+			}
+			printf("%s --> Frame rate:         %f fps\n",
+				doyuv? "Comp from YUV":"Compress     ", (double)iter/elapsed);
+			printf("                  Output image size:  %d bytes\n",
+				totaljpegsize);
+			printf("                  Compression ratio:  %f:1\n",
+				(double)(w*h*ps)/(double)totaljpegsize);
+			printf("                  Throughput:         %f Megapixels/sec\n",
+				(double)(w*h)/1000000.*(double)iter/elapsed);
+			printf("                  Output bit stream:  %f Megabits/sec\n",
+				(double)totaljpegsize*8./1000000.*(double)iter/elapsed);
+		}
+		if(tilew==w && tileh==h && dowrite)
+		{
+			snprintf(tempstr, 1024, "%s_%s_Q%d.jpg", filename, subName[subsamp],
+				jpegqual);
+			if((file=fopen(tempstr, "wb"))==NULL)
+				_throwunix("opening reference image");
+			if(fwrite(jpegbuf[0], jpegsize[0], 1, file)!=1)
+				_throwunix("writing reference image");
+			fclose(file);  file=NULL;
+			if(!quiet) printf("Reference image written to %s\n", tempstr);
+		}
+
+		/* Decompression test */
+		if(!componly)
+		{
+			if(decomp(srcbuf, jpegbuf, jpegsize, tmpbuf, w, h, subsamp, jpegqual,
+				filename, tilew, tileh)==-1)
+				goto bailout;
+		}
+
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+		free(jpegsize);  jpegsize=NULL;
+		if(doyuv)
+		{
+			free(yuvbuf);  yuvbuf=NULL;
+		}
+
+		if(tilew==w && tileh==h) break;
+	}
+
+	bailout:
+	if(file) {fclose(file);  file=NULL;}
+	if(jpegbuf)
+	{
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+	}
+	if(yuvbuf) {free(yuvbuf);  yuvbuf=NULL;}
+	if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+	if(tmpbuf) {free(tmpbuf);  tmpbuf=NULL;}
+	if(handle) {tjDestroy(handle);  handle=NULL;}
+	return retval;
+}
+
+
+int decompTest(char *filename)
+{
+	FILE *file=NULL;  tjhandle handle=NULL;
+	unsigned char **jpegbuf=NULL, *srcbuf=NULL;
+	unsigned long *jpegsize=NULL, srcsize, totaljpegsize;
+	tjtransform *t=NULL;
+	int w=0, h=0, subsamp=-1, cs=-1, _w, _h, _tilew, _tileh,
+		_ntilesw, _ntilesh, _subsamp;
+	char *temp=NULL, tempstr[80], tempstr2[80];
+	int row, col, i, iter, tilew, tileh, ntilesw=1, ntilesh=1, retval=0;
+	double start, elapsed;
+	int ps=tjPixelSize[pf], tile;
+
+	if((file=fopen(filename, "rb"))==NULL)
+		_throwunix("opening file");
+	if(fseek(file, 0, SEEK_END)<0 || (srcsize=ftell(file))==(unsigned long)-1)
+		_throwunix("determining file size");
+	if((srcbuf=(unsigned char *)malloc(srcsize))==NULL)
+		_throwunix("allocating memory");
+	if(fseek(file, 0, SEEK_SET)<0)
+		_throwunix("setting file position");
+	if(fread(srcbuf, srcsize, 1, file)<1)
+		_throwunix("reading JPEG data");
+	fclose(file);  file=NULL;
+
+	temp=strrchr(filename, '.');
+	if(temp!=NULL) *temp='\0';
+
+	if((handle=tjInitTransform())==NULL)
+		_throwtj("executing tjInitTransform()");
+	if(tjDecompressHeader3(handle, srcbuf, srcsize, &w, &h, &subsamp, &cs)==-1)
+		_throwtj("executing tjDecompressHeader3()");
+	if(cs==TJCS_YCCK || cs==TJCS_CMYK)
+	{
+		pf=TJPF_CMYK;  ps=tjPixelSize[pf];
+	}
+
+	if(quiet==1)
+	{
+		printf("All performance values in Mpixels/sec\n\n");
+		printf("Bitmap     JPEG   JPEG     %s  %s   Xform   Comp    Decomp  ",
+			dotile? "Tile ":"Image", dotile? "Tile ":"Image");
+		if(doyuv) printf("Decode");
+		printf("\n");
+		printf("Format     CS     Subsamp  Width  Height  Perf    Ratio   Perf    ");
+		if(doyuv) printf("Perf");
+		printf("\n\n");
+	}
+	else if(!quiet)
+		printf(">>>>>  JPEG %s --> %s (%s)  <<<<<\n",
+			formatName(subsamp, cs, tempstr), pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-up":"Top-down");
+
+	for(tilew=dotile? 16:w, tileh=dotile? 16:h; ; tilew*=2, tileh*=2)
+	{
+		if(tilew>w) tilew=w;  if(tileh>h) tileh=h;
+		ntilesw=(w+tilew-1)/tilew;  ntilesh=(h+tileh-1)/tileh;
+
+		if((jpegbuf=(unsigned char **)malloc(sizeof(unsigned char *)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG tile array");
+		memset(jpegbuf, 0, sizeof(unsigned char *)*ntilesw*ntilesh);
+		if((jpegsize=(unsigned long *)malloc(sizeof(unsigned long)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG size array");
+		memset(jpegsize, 0, sizeof(unsigned long)*ntilesw*ntilesh);
+
+		if((flags&TJFLAG_NOREALLOC)!=0 || !dotile)
+			for(i=0; i<ntilesw*ntilesh; i++)
+			{
+				if((jpegbuf[i]=(unsigned char *)tjAlloc(tjBufSize(tilew, tileh,
+					subsamp)))==NULL)
+					_throwunix("allocating JPEG tiles");
+			}
+
+		_w=w;  _h=h;  _tilew=tilew;  _tileh=tileh;
+		if(!quiet)
+		{
+			printf("\n%s size: %d x %d", dotile? "Tile":"Image", _tilew,
+				_tileh);
+			if(sf.num!=1 || sf.denom!=1)
+				printf(" --> %d x %d", TJSCALED(_w, sf), TJSCALED(_h, sf));
+			printf("\n");
+		}
+		else if(quiet==1)
+		{
+			printf("%-4s (%s)  %-5s  %-5s    ", pixFormatStr[pf],
+				(flags&TJFLAG_BOTTOMUP)? "BU":"TD", csName[cs], subNameLong[subsamp]);
+			printf("%-5d  %-5d   ", tilew, tileh);
+		}
+
+		_subsamp=subsamp;
+		if(dotile || xformop!=TJXOP_NONE || xformopt!=0 || customFilter)
+		{
+			if((t=(tjtransform *)malloc(sizeof(tjtransform)*ntilesw*ntilesh))
+				==NULL)
+				_throwunix("allocating image transform array");
+
+			if(xformop==TJXOP_TRANSPOSE || xformop==TJXOP_TRANSVERSE
+				|| xformop==TJXOP_ROT90 || xformop==TJXOP_ROT270)
+			{
+				_w=h;  _h=w;  _tilew=tileh;  _tileh=tilew;
+			}
+
+			if(xformopt&TJXOPT_GRAY) _subsamp=TJ_GRAYSCALE;
+			if(xformop==TJXOP_HFLIP || xformop==TJXOP_ROT180)
+				_w=_w-(_w%tjMCUWidth[_subsamp]);
+			if(xformop==TJXOP_VFLIP || xformop==TJXOP_ROT180)
+				_h=_h-(_h%tjMCUHeight[_subsamp]);
+			if(xformop==TJXOP_TRANSVERSE || xformop==TJXOP_ROT90)
+				_w=_w-(_w%tjMCUHeight[_subsamp]);
+			if(xformop==TJXOP_TRANSVERSE || xformop==TJXOP_ROT270)
+				_h=_h-(_h%tjMCUWidth[_subsamp]);
+			_ntilesw=(_w+_tilew-1)/_tilew;
+			_ntilesh=(_h+_tileh-1)/_tileh;
+
+			if(xformop==TJXOP_TRANSPOSE || xformop==TJXOP_TRANSVERSE
+				|| xformop==TJXOP_ROT90 || xformop==TJXOP_ROT270)
+			{
+				if(_subsamp==TJSAMP_422) _subsamp=TJSAMP_440;
+				else if(_subsamp==TJSAMP_440) _subsamp=TJSAMP_422;
+			}
+
+			for(row=0, tile=0; row<_ntilesh; row++)
+			{
+				for(col=0; col<_ntilesw; col++, tile++)
+				{
+					t[tile].r.w=min(_tilew, _w-col*_tilew);
+					t[tile].r.h=min(_tileh, _h-row*_tileh);
+					t[tile].r.x=col*_tilew;
+					t[tile].r.y=row*_tileh;
+					t[tile].op=xformop;
+					t[tile].options=xformopt|TJXOPT_TRIM;
+					t[tile].customFilter=customFilter;
+					if(t[tile].options&TJXOPT_NOOUTPUT && jpegbuf[tile])
+					{
+						tjFree(jpegbuf[tile]);  jpegbuf[tile]=NULL;
+					}
+				}
+			}
+
+			iter=-warmup;
+			elapsed=0.;
+			while(1)
+			{
+				start=gettime();
+				if(tjTransform(handle, srcbuf, srcsize, _ntilesw*_ntilesh, jpegbuf,
+					jpegsize, t, flags)==-1)
+					_throwtj("executing tjTransform()");
+				iter++;
+				if(iter>=1)
+				{
+					elapsed+=gettime()-start;
+					if(elapsed>=benchtime) break;
+				}
+			}
+
+			free(t);  t=NULL;
+
+			for(tile=0, totaljpegsize=0; tile<_ntilesw*_ntilesh; tile++)
+				totaljpegsize+=jpegsize[tile];
+
+			if(quiet)
+			{
+				printf("%-6s%s%-6s%s",
+					sigfig((double)(w*h)/1000000./elapsed, 4, tempstr, 80),
+					quiet==2? "\n":"  ",
+					sigfig((double)(w*h*ps)/(double)totaljpegsize, 4, tempstr2, 80),
+					quiet==2? "\n":"  ");
+			}
+			else if(!quiet)
+			{
+				printf("Transform     --> Frame rate:         %f fps\n", 1.0/elapsed);
+				printf("                  Output image size:  %lu bytes\n", totaljpegsize);
+				printf("                  Compression ratio:  %f:1\n",
+					(double)(w*h*ps)/(double)totaljpegsize);
+				printf("                  Throughput:         %f Megapixels/sec\n",
+					(double)(w*h)/1000000./elapsed);
+				printf("                  Output bit stream:  %f Megabits/sec\n",
+					(double)totaljpegsize*8./1000000./elapsed);
+			}
+		}
+		else
+		{
+			if(quiet==1) printf("N/A     N/A     ");
+			jpegsize[0]=srcsize;
+			memcpy(jpegbuf[0], srcbuf, srcsize);
+		}
+
+		if(w==tilew) _tilew=_w;
+		if(h==tileh) _tileh=_h;
+		if(!(xformopt&TJXOPT_NOOUTPUT))
+		{
+			if(decomp(NULL, jpegbuf, jpegsize, NULL, _w, _h, _subsamp, 0,
+				filename, _tilew, _tileh)==-1)
+				goto bailout;
+		}
+		else if(quiet==1) printf("N/A\n");
+
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+		if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+
+		if(tilew==w && tileh==h) break;
+	}
+
+	bailout:
+	if(file) {fclose(file);  file=NULL;}
+	if(jpegbuf)
+	{
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+	}
+	if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+	if(srcbuf) {free(srcbuf);  srcbuf=NULL;}
+	if(t) {free(t);  t=NULL;}
+	if(handle) {tjDestroy(handle);  handle=NULL;}
+	return retval;
+}
+
+
+void usage(char *progname)
+{
+	int i;
+	printf("USAGE: %s\n", progname);
+	printf("       <Inputfile (BMP|PPM)> <Quality> [options]\n\n");
+	printf("       %s\n", progname);
+	printf("       <Inputfile (JPG)> [options]\n\n");
+	printf("Options:\n\n");
+	printf("-alloc = Dynamically allocate JPEG image buffers\n");
+	printf("-bmp = Generate output images in Windows Bitmap format (default = PPM)\n");
+	printf("-bottomup = Test bottom-up compression/decompression\n");
+	printf("-tile = Test performance of the codec when the image is encoded as separate\n");
+	printf("     tiles of varying sizes.\n");
+	printf("-rgb, -bgr, -rgbx, -bgrx, -xbgr, -xrgb =\n");
+	printf("     Test the specified color conversion path in the codec (default = BGR)\n");
+	printf("-cmyk = Indirectly test YCCK JPEG compression/decompression (the source\n");
+	printf("     and destination bitmaps are still RGB.  The conversion is done\n");
+	printf("     internally prior to compression or after decompression.)\n");
+	printf("-fastupsample = Use the fastest chrominance upsampling algorithm available in\n");
+	printf("     the underlying codec\n");
+	printf("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying\n");
+	printf("     codec\n");
+	printf("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the\n");
+	printf("     underlying codec\n");
+	printf("-subsamp <s> = When testing JPEG compression, this option specifies the level\n");
+	printf("     of chrominance subsampling to use (<s> = 444, 422, 440, 420, 411, or\n");
+	printf("     GRAY).  The default is to test Grayscale, 4:2:0, 4:2:2, and 4:4:4 in\n");
+	printf("     sequence.\n");
+	printf("-quiet = Output results in tabular rather than verbose format\n");
+	printf("-yuv = Test YUV encoding/decoding functions\n");
+	printf("-yuvpad <p> = If testing YUV encoding/decoding, this specifies the number of\n");
+	printf("     bytes to which each row of each plane in the intermediate YUV image is\n");
+	printf("     padded (default = 1)\n");
+	printf("-scale M/N = Scale down the width/height of the decompressed JPEG image by a\n");
+	printf("     factor of M/N (M/N = ");
+	for(i=0; i<nsf; i++)
+	{
+		printf("%d/%d", scalingfactors[i].num, scalingfactors[i].denom);
+		if(nsf==2 && i!=nsf-1) printf(" or ");
+		else if(nsf>2)
+		{
+			if(i!=nsf-1) printf(", ");
+			if(i==nsf-2) printf("or ");
+		}
+		if(i%8==0 && i!=0) printf("\n     ");
+	}
+	printf(")\n");
+	printf("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =\n");
+	printf("     Perform the corresponding lossless transform prior to\n");
+	printf("     decompression (these options are mutually exclusive)\n");
+	printf("-grayscale = Perform lossless grayscale conversion prior to decompression\n");
+	printf("     test (can be combined with the other transforms above)\n");
+	printf("-benchtime <t> = Run each benchmark for at least <t> seconds (default = 5.0)\n");
+	printf("-warmup <w> = Execute each benchmark <w> times to prime the cache before\n");
+	printf("     taking performance measurements (default = 1)\n");
+	printf("-componly = Stop after running compression tests.  Do not test decompression.\n");
+	printf("-nowrite = Do not write reference or output images (improves consistency of\n");
+	printf("     performance measurements.)\n\n");
+	printf("NOTE:  If the quality is specified as a range (e.g. 90-100), a separate\n");
+	printf("test will be performed for all quality values in the range.\n\n");
+	exit(1);
+}
+
+
+int main(int argc, char *argv[])
+{
+	unsigned char *srcbuf=NULL;  int w=0, h=0, i, j;
+	int minqual=-1, maxqual=-1;  char *temp;
+	int minarg=2, retval=0, subsamp=-1;
+
+	if((scalingfactors=tjGetScalingFactors(&nsf))==NULL || nsf==0)
+		_throwtj("executing tjGetScalingFactors()");
+
+	if(argc<minarg) usage(argv[0]);
+
+	temp=strrchr(argv[1], '.');
+	if(temp!=NULL)
+	{
+		if(!strcasecmp(temp, ".bmp")) ext="bmp";
+		if(!strcasecmp(temp, ".jpg") || !strcasecmp(temp, ".jpeg")) decomponly=1;
+	}
+
+	printf("\n");
+
+	if(!decomponly)
+	{
+		minarg=3;
+		if(argc<minarg) usage(argv[0]);
+		if((minqual=atoi(argv[2]))<1 || minqual>100)
+		{
+			puts("ERROR: Quality must be between 1 and 100.");
+			exit(1);
+		}
+		if((temp=strchr(argv[2], '-'))!=NULL && strlen(temp)>1
+			&& sscanf(&temp[1], "%d", &maxqual)==1 && maxqual>minqual && maxqual>=1
+			&& maxqual<=100) {}
+		else maxqual=minqual;
+	}
+
+	if(argc>minarg)
+	{
+		for(i=minarg; i<argc; i++)
+		{
+			if(!strcasecmp(argv[i], "-tile"))
+			{
+				dotile=1;  xformopt|=TJXOPT_CROP;
+			}
+			if(!strcasecmp(argv[i], "-fastupsample"))
+			{
+				printf("Using fast upsampling code\n\n");
+				flags|=TJFLAG_FASTUPSAMPLE;
+			}
+			if(!strcasecmp(argv[i], "-fastdct"))
+			{
+				printf("Using fastest DCT/IDCT algorithm\n\n");
+				flags|=TJFLAG_FASTDCT;
+			}
+			if(!strcasecmp(argv[i], "-accuratedct"))
+			{
+				printf("Using most accurate DCT/IDCT algorithm\n\n");
+				flags|=TJFLAG_ACCURATEDCT;
+			}
+			if(!strcasecmp(argv[i], "-rgb")) pf=TJPF_RGB;
+			if(!strcasecmp(argv[i], "-rgbx")) pf=TJPF_RGBX;
+			if(!strcasecmp(argv[i], "-bgr")) pf=TJPF_BGR;
+			if(!strcasecmp(argv[i], "-bgrx")) pf=TJPF_BGRX;
+			if(!strcasecmp(argv[i], "-xbgr")) pf=TJPF_XBGR;
+			if(!strcasecmp(argv[i], "-xrgb")) pf=TJPF_XRGB;
+			if(!strcasecmp(argv[i], "-cmyk")) pf=TJPF_CMYK;
+			if(!strcasecmp(argv[i], "-bottomup")) flags|=TJFLAG_BOTTOMUP;
+			if(!strcasecmp(argv[i], "-quiet")) quiet=1;
+			if(!strcasecmp(argv[i], "-qq")) quiet=2;
+			if(!strcasecmp(argv[i], "-scale") && i<argc-1)
+			{
+				int temp1=0, temp2=0, match=0;
+				if(sscanf(argv[++i], "%d/%d", &temp1, &temp2)==2)
+				{
+					for(j=0; j<nsf; j++)
+					{
+						if((double)temp1/(double)temp2
+							== (double)scalingfactors[j].num/(double)scalingfactors[j].denom)
+						{
+							sf=scalingfactors[j];
+							match=1;  break;
+						}
+					}
+					if(!match) usage(argv[0]);
+				}
+				else usage(argv[0]);
+			}
+			if(!strcasecmp(argv[i], "-hflip")) xformop=TJXOP_HFLIP;
+			if(!strcasecmp(argv[i], "-vflip")) xformop=TJXOP_VFLIP;
+			if(!strcasecmp(argv[i], "-transpose")) xformop=TJXOP_TRANSPOSE;
+			if(!strcasecmp(argv[i], "-transverse")) xformop=TJXOP_TRANSVERSE;
+			if(!strcasecmp(argv[i], "-rot90")) xformop=TJXOP_ROT90;
+			if(!strcasecmp(argv[i], "-rot180")) xformop=TJXOP_ROT180;
+			if(!strcasecmp(argv[i], "-rot270")) xformop=TJXOP_ROT270;
+			if(!strcasecmp(argv[i], "-grayscale")) xformopt|=TJXOPT_GRAY;
+			if(!strcasecmp(argv[i], "-custom")) customFilter=dummyDCTFilter;
+			if(!strcasecmp(argv[i], "-nooutput")) xformopt|=TJXOPT_NOOUTPUT;
+			if(!strcasecmp(argv[i], "-benchtime") && i<argc-1)
+			{
+				double temp=atof(argv[++i]);
+				if(temp>0.0) benchtime=temp;
+				else usage(argv[0]);
+			}
+			if(!strcasecmp(argv[i], "-warmup") && i<argc-1)
+			{
+				int temp=atoi(argv[++i]);
+				if(temp>=0)
+				{
+					warmup=temp;
+					printf("Warmup runs = %d\n\n", warmup);
+				}
+				else usage(argv[0]);
+			}
+			if(!strcmp(argv[i], "-?")) usage(argv[0]);
+			if(!strcasecmp(argv[i], "-alloc")) flags&=(~TJFLAG_NOREALLOC);
+			if(!strcasecmp(argv[i], "-bmp")) ext="bmp";
+			if(!strcasecmp(argv[i], "-yuv"))
+			{
+				printf("Testing YUV planar encoding/decoding\n\n");
+				doyuv=1;
+			}
+			if(!strcasecmp(argv[i], "-yuvpad") && i<argc-1)
+			{
+				int temp=atoi(argv[++i]);
+				if(temp>=1) yuvpad=temp;
+			}
+			if(!strcasecmp(argv[i], "-subsamp") && i<argc-1)
+			{
+				i++;
+				if(toupper(argv[i][0])=='G') subsamp=TJSAMP_GRAY;
+				else
+				{
+					int temp=atoi(argv[i]);
+					switch(temp)
+					{
+						case 444:  subsamp=TJSAMP_444;  break;
+						case 422:  subsamp=TJSAMP_422;  break;
+						case 440:  subsamp=TJSAMP_440;  break;
+						case 420:  subsamp=TJSAMP_420;  break;
+						case 411:  subsamp=TJSAMP_411;  break;
+					}
+				}
+			}
+			if(!strcasecmp(argv[i], "-componly")) componly=1;
+			if(!strcasecmp(argv[i], "-nowrite")) dowrite=0;
+		}
+	}
+
+	if((sf.num!=1 || sf.denom!=1) && dotile)
+	{
+		printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+		printf("work when scaled decompression is enabled.\n");
+		dotile=0;
+	}
+
+	if((flags&TJFLAG_NOREALLOC)==0 && dotile)
+	{
+		printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+		printf("work when dynamic JPEG buffer allocation is enabled.\n\n");
+		dotile=0;
+	}
+
+	if(!decomponly)
+	{
+		if(loadbmp(argv[1], &srcbuf, &w, &h, pf, (flags&TJFLAG_BOTTOMUP)!=0)==-1)
+			_throwbmp("loading bitmap");
+		temp=strrchr(argv[1], '.');
+		if(temp!=NULL) *temp='\0';
+	}
+
+	if(quiet==1 && !decomponly)
+	{
+		printf("All performance values in Mpixels/sec\n\n");
+		printf("Bitmap     JPEG     JPEG  %s  %s   ",
+			dotile? "Tile ":"Image", dotile? "Tile ":"Image");
+		if(doyuv) printf("Encode  ");
+		printf("Comp    Comp    Decomp  ");
+		if(doyuv) printf("Decode");
+		printf("\n");
+		printf("Format     Subsamp  Qual  Width  Height  ");
+		if(doyuv) printf("Perf    ");
+		printf("Perf    Ratio   Perf    ");
+		if(doyuv) printf("Perf");
+		printf("\n\n");
+	}
+
+	if(decomponly)
+	{
+		decompTest(argv[1]);
+		printf("\n");
+		goto bailout;
+	}
+	if(subsamp>=0 && subsamp<TJ_NUMSAMP)
+	{
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, subsamp, i, argv[1]);
+		printf("\n");
+	}
+	else
+	{
+		if(pf!=TJPF_CMYK)
+		{
+			for(i=maxqual; i>=minqual; i--)
+				fullTest(srcbuf, w, h, TJSAMP_GRAY, i, argv[1]);
+			printf("\n");
+		}
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_420, i, argv[1]);
+		printf("\n");
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_422, i, argv[1]);
+		printf("\n");
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_444, i, argv[1]);
+		printf("\n");
+	}
+
+	bailout:
+	if(srcbuf) free(srcbuf);
+	return retval;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbenchtest.in glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbenchtest.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbenchtest.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbenchtest.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,249 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+EXT=bmp
+IMAGES="vgl_5674_0098.${EXT} vgl_6434_0018a.${EXT} vgl_6548_0026a.${EXT} nightshot_iso_100.${EXT}"
+IMGDIR=@srcdir@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_output.XXXXXX`
+EXEDIR=.
+BMPARG=
+NSARG=
+YUVARG=
+ALLOC=0
+ALLOCARG=
+if [ "$EXT" = "bmp" ]; then BMPARG=-bmp; fi
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjbenchtest.log
+
+if [ $# -gt 0 ]; then
+	if [ "$1" = "-yuv" ]; then
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.${EXT}"
+	fi
+	if [ "$1" = "-alloc" ]; then
+		ALLOCARG=-alloc
+		ALLOC=1
+	fi
+fi
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .${EXT}`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG $ALLOCARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG
+		for samp in GRAY 444; do
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG
+		for samp in 420 422; do
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+			runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT}
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbenchtest.java.in glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbenchtest.java.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjbenchtest.java.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjbenchtest.java.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,207 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@srcdir@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_java_output.XXXXXX`
+EXEDIR=.
+JAVA="@JAVA@ -cp java/turbojpeg.jar -Djava.library.path=.libs"
+BMPARG=
+NSARG=
+YUVARG=
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjbenchtest-java.log
+
+if [ $# -gt 0 ]; then
+	if [ "$1" = "-yuv" ]; then
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.bmp"
+	fi
+fi
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_default_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG
+		for samp in GRAY 444; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG
+		for samp in 420 422; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp $i -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjexampletest.in glmark2-2021.02/android/jni/src/libjpeg-turbo/tjexampletest.in
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjexampletest.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjexampletest.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,150 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@srcdir@/testimages
+OUTDIR=__tjexampletest_output
+EXEDIR=.
+JAVA="@JAVA@ -cp java/turbojpeg.jar -Djava.library.path=.libs"
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjexampletest.log
+
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	$EXEDIR/cjpeg -quality 95 -dct fast -grayscale $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_GRAY_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_420_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_422_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_444_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -grayscale $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_420_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_422_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_444_accurate_cjpeg.jpg
+	for samp in GRAY 420 422 444; do
+		$EXEDIR/djpeg -rgb -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_default_djpeg.bmp
+		$EXEDIR/djpeg -dct fast -rgb -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_fast_djpeg.bmp
+		$EXEDIR/djpeg -dct int -rgb -bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg >$OUTDIR/${basename}_${samp}_accurate_djpeg.bmp
+	done
+	for samp in 420 422; do
+		$EXEDIR/djpeg -nosmooth -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp
+		$EXEDIR/djpeg -dct fast -nosmooth -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp
+		$EXEDIR/djpeg -dct int -nosmooth -bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg >$OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp
+	done
+
+	# Compression
+	for dct in fast accurate; do
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/$image $OUTDIR/${basename}_${samp}_${dct}.jpg -q 95 -samp ${samp} -${dct}dct
+			runme cmp $OUTDIR/${basename}_${samp}_${dct}.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	# Decompression
+	for dct in fast accurate default; do
+		srcdct=${dct}
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			srcdct=fast
+			dctarg=
+		fi
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}.bmp ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}.bmp $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}.bmp
+		done
+		for samp in 420 422; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp -fastupsample ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			$EXEDIR/djpeg -rgb -bmp -scale ${scalearg} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${scale}.bmp -scale ${scalearg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		$EXEDIR/jpegtran -crop 70x60+16+16 -flip horizontal -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -flip vertical -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -transpose -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -transverse -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 90 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 180 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 270 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -crop 16,16,70x60
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			$EXEDIR/djpeg -rgb -bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 16,16,70x60
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+		for samp in 420 422; do
+			$EXEDIR/djpeg -nosmooth -rgb -bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 16,16,70x60 -fastupsample
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -grayscale -crop 16,16,70x60
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_GRAY_${xform}_jpegtran.jpg
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -grayscale -crop 16,16,70x60
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				$EXEDIR/djpeg -rgb -bmp -scale ${scalearg} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp -$xform -scale ${scalearg} -crop 16,16,70x60
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjunittest.c glmark2-2021.02/android/jni/src/libjpeg-turbo/tjunittest.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjunittest.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjunittest.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,735 @@
+/*
+ * Copyright (C)2009-2014 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program tests the various code paths in the TurboJPEG C Wrapper
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include "./tjutil.h"
+#include "./turbojpeg.h"
+#ifdef _WIN32
+ #include <time.h>
+ #define random() rand()
+#endif
+
+
+void usage(char *progName)
+{
+	printf("\nUSAGE: %s [options]\n", progName);
+	printf("Options:\n");
+	printf("-yuv = test YUV encoding/decoding support\n");
+	printf("-noyuvpad = do not pad each line of each Y, U, and V plane to the nearest\n");
+	printf("            4-byte boundary\n");
+	printf("-alloc = test automatic buffer allocation\n");
+	exit(1);
+}
+
+
+#define _throwtj() {printf("TurboJPEG ERROR:\n%s\n", tjGetErrorStr());  \
+	bailout();}
+#define _tj(f) {if((f)==-1) _throwtj();}
+#define _throw(m) {printf("ERROR: %s\n", m);  bailout();}
+
+const char *subNameLong[TJ_NUMSAMP]=
+{
+	"4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *subName[TJ_NUMSAMP]={"444", "422", "420", "GRAY", "440", "411"};
+
+const char *pixFormatStr[TJ_NUMPF]=
+{
+	"RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "Grayscale",
+	"RGBA", "BGRA", "ABGR", "ARGB", "CMYK"
+};
+
+const int alphaOffset[TJ_NUMPF] = {-1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1};
+
+const int _3byteFormats[]={TJPF_RGB, TJPF_BGR};
+const int _4byteFormats[]={TJPF_RGBX, TJPF_BGRX, TJPF_XBGR, TJPF_XRGB,
+	TJPF_CMYK};
+const int _onlyGray[]={TJPF_GRAY};
+const int _onlyRGB[]={TJPF_RGB};
+
+int doyuv=0, alloc=0, pad=4;
+
+int exitStatus=0;
+#define bailout() {exitStatus=-1;  goto bailout;}
+
+
+void initBuf(unsigned char *buf, int w, int h, int pf, int flags)
+{
+	int roffset=tjRedOffset[pf];
+	int goffset=tjGreenOffset[pf];
+	int boffset=tjBlueOffset[pf];
+	int ps=tjPixelSize[pf];
+	int index, row, col, halfway=16;
+
+	if(pf==TJPF_GRAY)
+	{
+		memset(buf, 0, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0) buf[index]=(row<halfway)? 255:0;
+				else buf[index]=(row<halfway)? 76:226;
+			}
+		}
+	}
+	else if(pf==TJPF_CMYK)
+	{
+		memset(buf, 255, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0)
+				{
+					if(row>=halfway) buf[index*ps+3]=0;
+				}
+				else
+				{
+					buf[index*ps+2]=0;
+					if(row<halfway) buf[index*ps+1]=0;
+				}
+			}
+		}
+	}
+	else
+	{
+		memset(buf, 0, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0)
+				{
+					if(row<halfway)
+					{
+						buf[index*ps+roffset]=255;
+						buf[index*ps+goffset]=255;
+						buf[index*ps+boffset]=255;
+					}
+				}
+				else
+				{
+					buf[index*ps+roffset]=255;
+					if(row>=halfway) buf[index*ps+goffset]=255;
+				}
+			}
+		}
+	}
+}
+
+
+#define checkval(v, cv) { \
+	if(v<cv-1 || v>cv+1) { \
+		printf("\nComp. %s at %d,%d should be %d, not %d\n",  \
+			#v, row, col, cv, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+#define checkval0(v) { \
+	if(v>1) { \
+		printf("\nComp. %s at %d,%d should be 0, not %d\n", #v, row, col, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+#define checkval255(v) { \
+	if(v<254) { \
+		printf("\nComp. %s at %d,%d should be 255, not %d\n", #v, row, col, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+
+int checkBuf(unsigned char *buf, int w, int h, int pf, int subsamp,
+	tjscalingfactor sf, int flags)
+{
+	int roffset=tjRedOffset[pf];
+	int goffset=tjGreenOffset[pf];
+	int boffset=tjBlueOffset[pf];
+	int aoffset=alphaOffset[pf];
+	int ps=tjPixelSize[pf];
+	int index, row, col, retval=1;
+	int halfway=16*sf.num/sf.denom;
+	int blocksize=8*sf.num/sf.denom;
+
+	if(pf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				unsigned char c, m, y, k;
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				c=buf[index*ps];
+				m=buf[index*ps+1];
+				y=buf[index*ps+2];
+				k=buf[index*ps+3];
+				if(((row/blocksize)+(col/blocksize))%2==0)
+				{
+					checkval255(c);  checkval255(m);  checkval255(y);
+					if(row<halfway) checkval255(k)
+					else checkval0(k)
+				}
+				else
+				{
+					checkval255(c);  checkval0(y);  checkval255(k);
+					if(row<halfway) checkval0(m)
+					else checkval255(m)
+				}
+			}
+		}
+		return 1;
+	}
+
+	for(row=0; row<h; row++)
+	{
+		for(col=0; col<w; col++)
+		{
+			unsigned char r, g, b, a;
+			if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+			else index=row*w+col;
+			r=buf[index*ps+roffset];
+			g=buf[index*ps+goffset];
+			b=buf[index*ps+boffset];
+			a=aoffset>=0? buf[index*ps+aoffset]:0xFF;
+			if(((row/blocksize)+(col/blocksize))%2==0)
+			{
+				if(row<halfway)
+				{
+					checkval255(r);  checkval255(g);  checkval255(b);
+				}
+				else
+				{
+					checkval0(r);  checkval0(g);  checkval0(b);
+				}
+			}
+			else
+			{
+				if(subsamp==TJSAMP_GRAY)
+				{
+					if(row<halfway)
+					{
+						checkval(r, 76);  checkval(g, 76);  checkval(b, 76);
+					}
+					else
+					{
+						checkval(r, 226);  checkval(g, 226);  checkval(b, 226);
+					}
+				}
+				else
+				{
+					if(row<halfway)
+					{
+						checkval255(r);  checkval0(g);  checkval0(b);
+					}
+					else
+					{
+						checkval255(r);  checkval255(g);  checkval0(b);
+					}
+				}
+			}
+			checkval255(a);
+		}
+	}
+
+	bailout:
+	if(retval==0)
+	{
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(pf==TJPF_CMYK)
+					printf("%.3d/%.3d/%.3d/%.3d ", buf[(row*w+col)*ps],
+						buf[(row*w+col)*ps+1], buf[(row*w+col)*ps+2],
+						buf[(row*w+col)*ps+3]);
+				else
+					printf("%.3d/%.3d/%.3d ", buf[(row*w+col)*ps+roffset],
+						buf[(row*w+col)*ps+goffset], buf[(row*w+col)*ps+boffset]);
+			}
+			printf("\n");
+		}
+	}
+	return retval;
+}
+
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+
+int checkBufYUV(unsigned char *buf, int w, int h, int subsamp,
+	tjscalingfactor sf)
+{
+	int row, col;
+	int hsf=tjMCUWidth[subsamp]/8, vsf=tjMCUHeight[subsamp]/8;
+	int pw=PAD(w, hsf), ph=PAD(h, vsf);
+	int cw=pw/hsf, ch=ph/vsf;
+	int ypitch=PAD(pw, pad), uvpitch=PAD(cw, pad);
+	int retval=1;
+	int halfway=16*sf.num/sf.denom;
+	int blocksize=8*sf.num/sf.denom;
+
+	for(row=0; row<ph; row++)
+	{
+		for(col=0; col<pw; col++)
+		{
+			unsigned char y=buf[ypitch*row+col];
+			if(((row/blocksize)+(col/blocksize))%2==0)
+			{
+				if(row<halfway) checkval255(y)  else checkval0(y);
+			}
+			else
+			{
+				if(row<halfway) checkval(y, 76)  else checkval(y, 226);
+			}
+		}
+	}
+	if(subsamp!=TJSAMP_GRAY)
+	{
+		int halfway=16/vsf*sf.num/sf.denom;
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+			{
+				unsigned char u=buf[ypitch*ph + (uvpitch*row+col)],
+					v=buf[ypitch*ph + uvpitch*ch + (uvpitch*row+col)];
+				if(((row*vsf/blocksize)+(col*hsf/blocksize))%2==0)
+				{
+					checkval(u, 128);  checkval(v, 128);
+				}
+				else
+				{
+					if(row<halfway)
+					{
+						checkval(u, 85);  checkval255(v);
+					}
+					else
+					{
+						checkval0(u);  checkval(v, 149);
+					}
+				}
+			}
+		}
+	}
+
+	bailout:
+	if(retval==0)
+	{
+		for(row=0; row<ph; row++)
+		{
+			for(col=0; col<pw; col++)
+				printf("%.3d ", buf[ypitch*row+col]);
+			printf("\n");
+		}
+		printf("\n");
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+				printf("%.3d ", buf[ypitch*ph + (uvpitch*row+col)]);
+			printf("\n");
+		}
+		printf("\n");
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+				printf("%.3d ", buf[ypitch*ph + uvpitch*ch + (uvpitch*row+col)]);
+			printf("\n");
+		}
+	}
+
+	return retval;
+}
+
+
+void writeJPEG(unsigned char *jpegBuf, unsigned long jpegSize, char *filename)
+{
+	FILE *file=fopen(filename, "wb");
+	if(!file || fwrite(jpegBuf, jpegSize, 1, file)!=1)
+	{
+		printf("ERROR: Could not write to %s.\n%s\n", filename, strerror(errno));
+		bailout();
+	}
+
+	bailout:
+	if(file) fclose(file);
+}
+
+
+void compTest(tjhandle handle, unsigned char **dstBuf,
+	unsigned long *dstSize, int w, int h, int pf, char *basename,
+	int subsamp, int jpegQual, int flags)
+{
+	char tempStr[1024];  unsigned char *srcBuf=NULL, *yuvBuf=NULL;
+	const char *pfStr=pixFormatStr[pf];
+	const char *buStrLong=(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ";
+	const char *buStr=(flags&TJFLAG_BOTTOMUP)? "BU":"TD";
+
+	if((srcBuf=(unsigned char *)malloc(w*h*tjPixelSize[pf]))==NULL)
+		_throw("Memory allocation failure");
+	initBuf(srcBuf, w, h, pf, flags);
+
+	if(*dstBuf && *dstSize>0) memset(*dstBuf, 0, *dstSize);
+
+
+	if(!alloc) flags|=TJFLAG_NOREALLOC;
+	if(doyuv)
+	{
+		unsigned long yuvSize=tjBufSizeYUV2(w, pad, h, subsamp);
+		tjscalingfactor sf={1, 1};
+		tjhandle handle2=tjInitCompress();
+		if(!handle2) _throwtj();
+
+		if((yuvBuf=(unsigned char *)malloc(yuvSize))==NULL)
+			_throw("Memory allocation failure");
+		memset(yuvBuf, 0, yuvSize);
+
+		printf("%s %s -> YUV %s ... ", pfStr, buStrLong, subNameLong[subsamp]);
+		_tj(tjEncodeYUV3(handle2, srcBuf, w, 0, h, pf, yuvBuf, pad, subsamp,
+			flags));
+		tjDestroy(handle2);
+		if(checkBufYUV(yuvBuf, w, h, subsamp, sf)) printf("Passed.\n");
+		else printf("FAILED!\n");
+
+		printf("YUV %s %s -> JPEG Q%d ... ", subNameLong[subsamp], buStrLong,
+			jpegQual);
+		_tj(tjCompressFromYUV(handle, yuvBuf, w, pad, h, subsamp, dstBuf,
+			dstSize, jpegQual, flags));
+	}
+	else
+	{
+		printf("%s %s -> %s Q%d ... ", pfStr, buStrLong, subNameLong[subsamp],
+			jpegQual);
+		_tj(tjCompress2(handle, srcBuf, w, 0, h, pf, dstBuf, dstSize, subsamp,
+			jpegQual, flags));
+	}
+
+	snprintf(tempStr, 1024, "%s_enc_%s_%s_%s_Q%d.jpg", basename, pfStr, buStr,
+		subName[subsamp], jpegQual);
+	writeJPEG(*dstBuf, *dstSize, tempStr);
+	printf("Done.\n  Result in %s\n", tempStr);
+
+	bailout:
+	if(yuvBuf) free(yuvBuf);
+	if(srcBuf) free(srcBuf);
+}
+
+
+void _decompTest(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, int w, int h, int pf, char *basename, int subsamp,
+	int flags, tjscalingfactor sf)
+{
+	unsigned char *dstBuf=NULL, *yuvBuf=NULL;
+	int _hdrw=0, _hdrh=0, _hdrsubsamp=-1;
+	int scaledWidth=TJSCALED(w, sf);
+	int scaledHeight=TJSCALED(h, sf);
+	unsigned long dstSize=0;
+
+	_tj(tjDecompressHeader2(handle, jpegBuf, jpegSize, &_hdrw, &_hdrh,
+		&_hdrsubsamp));
+	if(_hdrw!=w || _hdrh!=h || _hdrsubsamp!=subsamp)
+		_throw("Incorrect JPEG header");
+
+	dstSize=scaledWidth*scaledHeight*tjPixelSize[pf];
+	if((dstBuf=(unsigned char *)malloc(dstSize))==NULL)
+		_throw("Memory allocation failure");
+	memset(dstBuf, 0, dstSize);
+
+	if(doyuv)
+	{
+		unsigned long yuvSize=tjBufSizeYUV2(scaledWidth, pad, scaledHeight,
+			subsamp);
+		tjhandle handle2=tjInitDecompress();
+		if(!handle2) _throwtj();
+
+		if((yuvBuf=(unsigned char *)malloc(yuvSize))==NULL)
+			_throw("Memory allocation failure");
+		memset(yuvBuf, 0, yuvSize);
+
+		printf("JPEG -> YUV %s ", subNameLong[subsamp]);
+		if(sf.num!=1 || sf.denom!=1)
+			printf("%d/%d ... ", sf.num, sf.denom);
+		else printf("... ");
+		_tj(tjDecompressToYUV2(handle, jpegBuf, jpegSize, yuvBuf, scaledWidth,
+			pad, scaledHeight, flags));
+		if(checkBufYUV(yuvBuf, scaledWidth, scaledHeight, subsamp, sf))
+			printf("Passed.\n");
+		else printf("FAILED!\n");
+
+		printf("YUV %s -> %s %s ... ", subNameLong[subsamp], pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ");
+		_tj(tjDecodeYUV(handle2, yuvBuf, pad, subsamp, dstBuf, scaledWidth, 0,
+			scaledHeight, pf, flags));
+		tjDestroy(handle2);
+	}
+	else
+	{
+		printf("JPEG -> %s %s ", pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ");
+		if(sf.num!=1 || sf.denom!=1)
+			printf("%d/%d ... ", sf.num, sf.denom);
+		else printf("... ");
+		_tj(tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, scaledWidth, 0,
+			scaledHeight, pf, flags));
+	}
+
+	if(checkBuf(dstBuf, scaledWidth, scaledHeight, pf, subsamp, sf, flags))
+		printf("Passed.");
+	else printf("FAILED!");
+	printf("\n");
+
+	bailout:
+	if(yuvBuf) free(yuvBuf);
+	if(dstBuf) free(dstBuf);
+}
+
+
+void decompTest(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, int w, int h, int pf, char *basename, int subsamp,
+	int flags)
+{
+	int i, n=0;
+	tjscalingfactor *sf=tjGetScalingFactors(&n);
+	if(!sf || !n) _throwtj();
+
+	for(i=0; i<n; i++)
+	{
+		if(subsamp==TJSAMP_444 || subsamp==TJSAMP_GRAY ||
+			(subsamp==TJSAMP_411 && sf[i].num==1 &&
+				(sf[i].denom==2 || sf[i].denom==1)) ||
+			(subsamp!=TJSAMP_411 && sf[i].num==1 &&
+				(sf[i].denom==4 || sf[i].denom==2 || sf[i].denom==1)))
+			_decompTest(handle, jpegBuf, jpegSize, w, h, pf, basename, subsamp,
+				flags, sf[i]);
+	}
+
+	bailout:
+	return;
+}
+
+
+void doTest(int w, int h, const int *formats, int nformats, int subsamp,
+	char *basename)
+{
+	tjhandle chandle=NULL, dhandle=NULL;
+	unsigned char *dstBuf=NULL;
+	unsigned long size=0;  int pfi, pf, i;
+
+	if(!alloc)
+		size=tjBufSize(w, h, subsamp);
+	if(size!=0)
+		if((dstBuf=(unsigned char *)tjAlloc(size))==NULL)
+			_throw("Memory allocation failure.");
+
+	if((chandle=tjInitCompress())==NULL || (dhandle=tjInitDecompress())==NULL)
+		_throwtj();
+
+	for(pfi=0; pfi<nformats; pfi++)
+	{
+		for(i=0; i<2; i++)
+		{
+			int flags=0;
+			if(subsamp==TJSAMP_422 || subsamp==TJSAMP_420 || subsamp==TJSAMP_440 ||
+				subsamp==TJSAMP_411)
+				flags|=TJFLAG_FASTUPSAMPLE;
+			if(i==1) flags|=TJFLAG_BOTTOMUP;
+			pf=formats[pfi];
+			compTest(chandle, &dstBuf, &size, w, h, pf, basename, subsamp, 100,
+				flags);
+			decompTest(dhandle, dstBuf, size, w, h, pf, basename, subsamp,
+				flags);
+			if(pf>=TJPF_RGBX && pf<=TJPF_XRGB)
+			{
+				printf("\n");
+				decompTest(dhandle, dstBuf, size, w, h, pf+(TJPF_RGBA-TJPF_RGBX),
+					basename, subsamp, flags);
+			}
+			printf("\n");
+		}
+	}
+	printf("--------------------\n\n");
+
+	bailout:
+	if(chandle) tjDestroy(chandle);
+	if(dhandle) tjDestroy(dhandle);
+
+	if(dstBuf) tjFree(dstBuf);
+}
+
+
+void bufSizeTest(void)
+{
+	int w, h, i, subsamp;
+	unsigned char *srcBuf=NULL, *dstBuf=NULL;
+	tjhandle handle=NULL;
+	unsigned long dstSize=0;
+
+	if((handle=tjInitCompress())==NULL) _throwtj();
+
+	printf("Buffer size regression test\n");
+	for(subsamp=0; subsamp<TJ_NUMSAMP; subsamp++)
+	{
+		for(w=1; w<48; w++)
+		{
+			int maxh=(w==1)? 2048:48;
+			for(h=1; h<maxh; h++)
+			{
+				if(h%100==0) printf("%.4d x %.4d\b\b\b\b\b\b\b\b\b\b\b", w, h);
+				if((srcBuf=(unsigned char *)malloc(w*h*4))==NULL)
+					_throw("Memory allocation failure");
+				if(!alloc || doyuv)
+				{
+					if(doyuv) dstSize=tjBufSizeYUV2(w, pad, h, subsamp);
+					else dstSize=tjBufSize(w, h, subsamp);
+					if((dstBuf=(unsigned char *)tjAlloc(dstSize))==NULL)
+						_throw("Memory allocation failure");
+				}
+
+				for(i=0; i<w*h*4; i++)
+				{
+					if(random()<RAND_MAX/2) srcBuf[i]=0;
+					else srcBuf[i]=255;
+				}
+
+				if(doyuv)
+				{
+					_tj(tjEncodeYUV3(handle, srcBuf, w, 0, h, TJPF_BGRX, dstBuf, pad,
+						subsamp, 0));
+				}
+				else
+				{
+					_tj(tjCompress2(handle, srcBuf, w, 0, h, TJPF_BGRX, &dstBuf,
+						&dstSize, subsamp, 100, alloc? 0:TJFLAG_NOREALLOC));
+				}
+				free(srcBuf);  srcBuf=NULL;
+				if(!alloc || doyuv)
+				{
+					tjFree(dstBuf);  dstBuf=NULL;
+				}
+
+				if((srcBuf=(unsigned char *)malloc(h*w*4))==NULL)
+					_throw("Memory allocation failure");
+				if(!alloc || doyuv)
+				{
+					if(doyuv) dstSize=tjBufSizeYUV2(h, pad, w, subsamp);
+					else dstSize=tjBufSize(h, w, subsamp);
+					if((dstBuf=(unsigned char *)tjAlloc(dstSize))==NULL)
+						_throw("Memory allocation failure");
+				}
+
+				for(i=0; i<h*w*4; i++)
+				{
+					if(random()<RAND_MAX/2) srcBuf[i]=0;
+					else srcBuf[i]=255;
+				}
+
+				if(doyuv)
+				{
+					_tj(tjEncodeYUV3(handle, srcBuf, h, 0, w, TJPF_BGRX, dstBuf, pad,
+						subsamp, 0));
+				}
+				else
+				{
+					_tj(tjCompress2(handle, srcBuf, h, 0, w, TJPF_BGRX, &dstBuf,
+						&dstSize, subsamp, 100, alloc? 0:TJFLAG_NOREALLOC));
+				}
+				free(srcBuf);  srcBuf=NULL;
+				if(!alloc || doyuv)
+				{
+					tjFree(dstBuf);  dstBuf=NULL;
+				}
+			}
+		}
+	}
+	printf("Done.      \n");
+
+	bailout:
+	if(srcBuf) free(srcBuf);
+	if(dstBuf) tjFree(dstBuf);
+	if(handle) tjDestroy(handle);
+}
+
+
+int main(int argc, char *argv[])
+{
+	int i, num4bf=5;
+	#ifdef _WIN32
+	srand((unsigned int)time(NULL));
+	#endif
+	if(argc>1)
+	{
+		for(i=1; i<argc; i++)
+		{
+			if(!strcasecmp(argv[i], "-yuv")) doyuv=1;
+			if(!strcasecmp(argv[i], "-noyuvpad")) pad=1;
+			if(!strcasecmp(argv[i], "-alloc")) alloc=1;
+			if(!strncasecmp(argv[i], "-h", 2) || !strcasecmp(argv[i], "-?"))
+				usage(argv[0]);
+		}
+	}
+	if(alloc) printf("Testing automatic buffer allocation\n");
+	if(doyuv) num4bf=4;
+	doTest(35, 39, _3byteFormats, 2, TJSAMP_444, "test");
+	doTest(39, 41, _4byteFormats, num4bf, TJSAMP_444, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_422, "test");
+	doTest(35, 39, _4byteFormats, num4bf, TJSAMP_422, "test");
+	doTest(39, 41, _3byteFormats, 2, TJSAMP_420, "test");
+	doTest(41, 35, _4byteFormats, num4bf, TJSAMP_420, "test");
+	doTest(35, 39, _3byteFormats, 2, TJSAMP_440, "test");
+	doTest(39, 41, _4byteFormats, num4bf, TJSAMP_440, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_411, "test");
+	doTest(35, 39, _4byteFormats, num4bf, TJSAMP_411, "test");
+	doTest(39, 41, _onlyGray, 1, TJSAMP_GRAY, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_GRAY, "test");
+	doTest(35, 39, _4byteFormats, 4, TJSAMP_GRAY, "test");
+	bufSizeTest();
+	if(doyuv)
+	{
+		printf("\n--------------------\n\n");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_444, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_422, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_420, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_440, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_411, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_GRAY, "test_yuv0");
+		doTest(48, 48, _onlyGray, 1, TJSAMP_GRAY, "test_yuv0");
+	}
+
+	return exitStatus;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjutil.c glmark2-2021.02/android/jni/src/libjpeg-turbo/tjutil.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjutil.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjutil.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,66 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+
+#include <windows.h>
+
+static double getfreq(void)
+{
+	LARGE_INTEGER freq;
+	if(!QueryPerformanceFrequency(&freq)) return 0.0;
+	return (double)freq.QuadPart;
+}
+
+static double f=-1.0;
+
+double gettime(void)
+{
+	LARGE_INTEGER t;
+	if(f<0.0) f=getfreq();
+	if(f==0.0) return (double)GetTickCount()/1000.;
+	else
+	{
+		QueryPerformanceCounter(&t);
+		return (double)t.QuadPart/f;
+	}
+}
+
+#else
+
+#include <stdlib.h>
+#include <sys/time.h>
+
+double gettime(void)
+{
+	struct timeval tv;
+	if(gettimeofday(&tv, NULL)<0) return 0.0;
+	else return (double)tv.tv_sec+((double)tv.tv_usec/1000000.);
+}
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjutil.h glmark2-2021.02/android/jni/src/libjpeg-turbo/tjutil.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/tjutil.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/tjutil.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+	#ifndef __MINGW32__
+		#include <stdio.h>
+		#define snprintf(str, n, format, ...)  \
+			_snprintf_s(str, n, _TRUNCATE, format, __VA_ARGS__)
+	#endif
+	#define strcasecmp stricmp
+	#define strncasecmp strnicmp
+#endif
+
+#ifndef min
+ #define min(a,b) ((a)<(b)?(a):(b))
+#endif
+
+#ifndef max
+ #define max(a,b) ((a)>(b)?(a):(b))
+#endif
+
+extern double gettime(void);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/transupp.c glmark2-2021.02/android/jni/src/libjpeg-turbo/transupp.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/transupp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/transupp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1621 @@
+/*
+ * transupp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2011, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains image transformation routines and other utility code
+ * used by the jpegtran sample application.  These are NOT part of the core
+ * JPEG library.  But we keep these routines separate from jpegtran.c to
+ * ease the task of maintaining jpegtran-like programs that have other user
+ * interfaces.
+ */
+
+/* Although this file really shouldn't have access to the library internals,
+ * it's helpful to let it call jround_up() and jcopy_block_row().
+ */
+#define JPEG_INTERNALS
+
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "transupp.h"           /* My own external interface */
+#include "jpegcomp.h"
+#include <ctype.h>              /* to declare isdigit() */
+
+
+#if JPEG_LIB_VERSION >= 70
+#define dstinfo_min_DCT_h_scaled_size dstinfo->min_DCT_h_scaled_size
+#define dstinfo_min_DCT_v_scaled_size dstinfo->min_DCT_v_scaled_size
+#else
+#define dstinfo_min_DCT_h_scaled_size DCTSIZE
+#define dstinfo_min_DCT_v_scaled_size DCTSIZE
+#endif
+
+
+#if TRANSFORMS_SUPPORTED
+
+/*
+ * Lossless image transformation routines.  These routines work on DCT
+ * coefficient arrays and thus do not require any lossy decompression
+ * or recompression of the image.
+ * Thanks to Guido Vollbeding for the initial design and code of this feature,
+ * and to Ben Jackson for introducing the cropping feature.
+ *
+ * Horizontal flipping is done in-place, using a single top-to-bottom
+ * pass through the virtual source array.  It will thus be much the
+ * fastest option for images larger than main memory.
+ *
+ * The other routines require a set of destination virtual arrays, so they
+ * need twice as much memory as jpegtran normally does.  The destination
+ * arrays are always written in normal scan order (top to bottom) because
+ * the virtual array manager expects this.  The source arrays will be scanned
+ * in the corresponding order, which means multiple passes through the source
+ * arrays for most of the transforms.  That could result in much thrashing
+ * if the image is larger than main memory.
+ *
+ * If cropping or trimming is involved, the destination arrays may be smaller
+ * than the source arrays.  Note it is not possible to do horizontal flip
+ * in-place when a nonzero Y crop offset is specified, since we'd have to move
+ * data from one block row to another but the virtual array manager doesn't
+ * guarantee we can touch more than one row at a time.  So in that case,
+ * we have to use a separate destination array.
+ *
+ * Some notes about the operating environment of the individual transform
+ * routines:
+ * 1. Both the source and destination virtual arrays are allocated from the
+ *    source JPEG object, and therefore should be manipulated by calling the
+ *    source's memory manager.
+ * 2. The destination's component count should be used.  It may be smaller
+ *    than the source's when forcing to grayscale.
+ * 3. Likewise the destination's sampling factors should be used.  When
+ *    forcing to grayscale the destination's sampling factors will be all 1,
+ *    and we may as well take that as the effective iMCU size.
+ * 4. When "trim" is in effect, the destination's dimensions will be the
+ *    trimmed values but the source's will be untrimmed.
+ * 5. When "crop" is in effect, the destination's dimensions will be the
+ *    cropped values but the source's will be uncropped.  Each transform
+ *    routine is responsible for picking up source data starting at the
+ *    correct X and Y offset for the crop region.  (The X and Y offsets
+ *    passed to the transform routines are measured in iMCU blocks of the
+ *    destination.)
+ * 6. All the routines assume that the source and destination buffers are
+ *    padded out to a full iMCU boundary.  This is true, although for the
+ *    source buffer it is an undocumented property of jdcoefct.c.
+ */
+
+
+LOCAL(void)
+do_crop (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+         jvirt_barray_ptr *src_coef_arrays,
+         jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop. */
+{
+  JDIMENSION dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  /* We simply have to copy the right amount of data (the destination's
+   * image size) starting at the given X and Y offsets in the source.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+         dst_blk_y + y_crop_blocks,
+         (JDIMENSION) compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        jcopy_block_row(src_buffer[offset_y] + x_crop_blocks,
+                        dst_buffer[offset_y],
+                        compptr->width_in_blocks);
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h_no_crop (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                   JDIMENSION x_crop_offset,
+                   jvirt_barray_ptr *src_coef_arrays)
+/* Horizontal flip; done in-place, so no separate dest array is required.
+ * NB: this only works when y_crop_offset is zero.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, blk_x, blk_y, x_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY buffer;
+  JCOEFPTR ptr1, ptr2;
+  JCOEF temp1, temp2;
+  jpeg_component_info *compptr;
+
+  /* Horizontal mirroring of DCT blocks is accomplished by swapping
+   * pairs of blocks in-place.  Within a DCT block, we perform horizontal
+   * mirroring by changing the signs of odd-numbered columns.
+   * Partial iMCUs at the right edge are left untouched.
+   */
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    for (blk_y = 0; blk_y < compptr->height_in_blocks;
+         blk_y += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci], blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        /* Do the mirroring */
+        for (blk_x = 0; blk_x * 2 < comp_width; blk_x++) {
+          ptr1 = buffer[offset_y][blk_x];
+          ptr2 = buffer[offset_y][comp_width - blk_x - 1];
+          /* this unrolled loop doesn't need to know which row it's on... */
+          for (k = 0; k < DCTSIZE2; k += 2) {
+            temp1 = *ptr1;      /* swap even column */
+            temp2 = *ptr2;
+            *ptr1++ = temp2;
+            *ptr2++ = temp1;
+            temp1 = *ptr1;      /* swap odd column with sign change */
+            temp2 = *ptr2;
+            *ptr1++ = -temp2;
+            *ptr2++ = -temp1;
+          }
+        }
+        if (x_crop_blocks > 0) {
+          /* Now left-justify the portion of the data to be kept.
+           * We can't use a single jcopy_block_row() call because that routine
+           * depends on memcpy(), whose behavior is unspecified for overlapping
+           * source and destination areas.  Sigh.
+           */
+          for (blk_x = 0; blk_x < compptr->width_in_blocks; blk_x++) {
+            jcopy_block_row(buffer[offset_y] + blk_x + x_crop_blocks,
+                            buffer[offset_y] + blk_x,
+                            (JDIMENSION) 1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* Horizontal flip in general cropping case */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Here we must output into a separate array because we can't touch
+   * different rows of a single virtual array simultaneously.  Otherwise,
+   * this is essentially the same as the routine above.
+   */
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+         dst_blk_y + y_crop_blocks,
+         (JDIMENSION) compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_buffer[offset_y];
+        src_row_ptr = src_buffer[offset_y];
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Do the mirrorable blocks */
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+            /* this unrolled loop doesn't need to know which row it's on... */
+            for (k = 0; k < DCTSIZE2; k += 2) {
+              *dst_ptr++ = *src_ptr++;   /* copy even column */
+              *dst_ptr++ = - *src_ptr++; /* copy odd column with sign change */
+            }
+          } else {
+            /* Copy last partial block(s) verbatim */
+            jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                            dst_row_ptr + dst_blk_x,
+                            (JDIMENSION) 1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_v (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* Vertical flip */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* We output into a separate array because we can't touch different
+   * rows of the source virtual array simultaneously.  Otherwise, this
+   * is a pretty straightforward analog of horizontal flip.
+   * Within a DCT block, vertical mirroring is done by changing the signs
+   * of odd-numbered rows.
+   * Partial iMCUs at the bottom edge are copied verbatim.
+   */
+  MCU_rows = srcinfo->output_height /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION) compptr->v_samp_factor,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge blocks will be copied verbatim. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          dst_row_ptr = dst_buffer[offset_y];
+          src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+          src_row_ptr += x_crop_blocks;
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+               dst_blk_x++) {
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            src_ptr = src_row_ptr[dst_blk_x];
+            for (i = 0; i < DCTSIZE; i += 2) {
+              /* copy even row */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = *src_ptr++;
+              /* copy odd row with sign change */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = - *src_ptr++;
+            }
+          }
+        } else {
+          /* Just copy row verbatim. */
+          jcopy_block_row(src_buffer[offset_y] + x_crop_blocks,
+                          dst_buffer[offset_y],
+                          compptr->width_in_blocks);
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transpose (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+              JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+              jvirt_barray_ptr *src_coef_arrays,
+              jvirt_barray_ptr *dst_coef_arrays)
+/* Transpose source into destination */
+{
+  JDIMENSION dst_blk_x, dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Transposing pixels within a block just requires transposing the
+   * DCT coefficients.
+   * Partial iMCUs at the edges require no special treatment; we simply
+   * process all the available DCT blocks for every component.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION) compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            src_ptr = src_buffer[offset_x][dst_blk_y + offset_y + y_crop_blocks];
+            for (i = 0; i < DCTSIZE; i++)
+              for (j = 0; j < DCTSIZE; j++)
+                dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_90 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* 90 degree rotation is equivalent to
+ *   1. Transposing the image;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) right edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_cols = srcinfo->output_height /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION) compptr->h_samp_factor,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          } else {
+            /* Edge blocks are transposed but not mirrored. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                i++;
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_buffer[offset_x]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_270 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+            JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+            jvirt_barray_ptr *src_coef_arrays,
+            jvirt_barray_ptr *dst_coef_arrays)
+/* 270 degree rotation is equivalent to
+ *   1. Horizontal mirroring;
+ *   2. Transposing the image.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) bottom edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_rows = srcinfo->output_width /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION) compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_buffer[offset_x]
+                [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++) {
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  j++;
+                  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                }
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_buffer[offset_x]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_180 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+            JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+            jvirt_barray_ptr *src_coef_arrays,
+            jvirt_barray_ptr *dst_coef_arrays)
+/* 180 degree rotation is equivalent to
+ *   1. Vertical mirroring;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the vertically mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION) compptr->v_samp_factor,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge rows are only mirrored horizontally. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_buffer[offset_y];
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored both ways. */
+              src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+              for (i = 0; i < DCTSIZE; i += 2) {
+                /* For even row, negate every odd column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = *src_ptr++;
+                  *dst_ptr++ = - *src_ptr++;
+                }
+                /* For odd row, negate every even column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = - *src_ptr++;
+                  *dst_ptr++ = *src_ptr++;
+                }
+              }
+            } else {
+              /* Any remaining right-edge blocks are only mirrored vertically. */
+              src_ptr = src_row_ptr[x_crop_blocks + dst_blk_x];
+              for (i = 0; i < DCTSIZE; i += 2) {
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = *src_ptr++;
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = - *src_ptr++;
+              }
+            }
+          }
+        } else {
+          /* Remaining rows are just mirrored horizontally. */
+          src_row_ptr = src_buffer[offset_y];
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored. */
+              dst_ptr = dst_row_ptr[dst_blk_x];
+              src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+              for (i = 0; i < DCTSIZE2; i += 2) {
+                *dst_ptr++ = *src_ptr++;
+                *dst_ptr++ = - *src_ptr++;
+              }
+            } else {
+              /* Any remaining right-edge blocks are only copied. */
+              jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                              dst_row_ptr + dst_blk_x,
+                              (JDIMENSION) 1);
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transverse (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+               JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+               jvirt_barray_ptr *src_coef_arrays,
+               jvirt_barray_ptr *dst_coef_arrays)
+/* Transverse transpose is equivalent to
+ *   1. 180 degree rotation;
+ *   2. Transposition;
+ * or
+ *   1. Horizontal mirroring;
+ *   2. Transposition;
+ *   3. Horizontal mirroring.
+ * These steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_height /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_width /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION) compptr->h_samp_factor,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          } else {
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Block is within the mirrorable area. */
+                src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                  [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                  }
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  }
+                }
+              } else {
+                /* Right-edge blocks are mirrored in y only */
+                src_ptr = src_buffer[offset_x]
+                  [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                  }
+                }
+              }
+            } else {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Bottom-edge blocks are mirrored in x only */
+                src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                  [dst_blk_y + offset_y + y_crop_blocks];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                }
+              } else {
+                /* At lower right corner, just transpose, no mirroring */
+                src_ptr = src_buffer[offset_x]
+                  [dst_blk_y + offset_y + y_crop_blocks];
+                for (i = 0; i < DCTSIZE; i++)
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+/* Parse an unsigned integer: subroutine for jtransform_parse_crop_spec.
+ * Returns TRUE if valid integer found, FALSE if not.
+ * *strptr is advanced over the digit string, and *result is set to its value.
+ */
+
+LOCAL(boolean)
+jt_read_integer (const char **strptr, JDIMENSION *result)
+{
+  const char *ptr = *strptr;
+  JDIMENSION val = 0;
+
+  for (; isdigit(*ptr); ptr++) {
+    val = val * 10 + (JDIMENSION) (*ptr - '0');
+  }
+  *result = val;
+  if (ptr == *strptr)
+    return FALSE;               /* oops, no digits */
+  *strptr = ptr;
+  return TRUE;
+}
+
+
+/* Parse a crop specification (written in X11 geometry style).
+ * The routine returns TRUE if the spec string is valid, FALSE if not.
+ *
+ * The crop spec string should have the format
+ *      <width>[f]x<height>[f]{+-}<xoffset>{+-}<yoffset>
+ * where width, height, xoffset, and yoffset are unsigned integers.
+ * Each of the elements can be omitted to indicate a default value.
+ * (A weakness of this style is that it is not possible to omit xoffset
+ * while specifying yoffset, since they look alike.)
+ *
+ * This code is loosely based on XParseGeometry from the X11 distribution.
+ */
+
+GLOBAL(boolean)
+jtransform_parse_crop_spec (jpeg_transform_info *info, const char *spec)
+{
+  info->crop = FALSE;
+  info->crop_width_set = JCROP_UNSET;
+  info->crop_height_set = JCROP_UNSET;
+  info->crop_xoffset_set = JCROP_UNSET;
+  info->crop_yoffset_set = JCROP_UNSET;
+
+  if (isdigit(*spec)) {
+    /* fetch width */
+    if (! jt_read_integer(&spec, &info->crop_width))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_width_set = JCROP_FORCE;
+    } else
+      info->crop_width_set = JCROP_POS;
+  }
+  if (*spec == 'x' || *spec == 'X') {
+    /* fetch height */
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_height))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_height_set = JCROP_FORCE;
+    } else
+      info->crop_height_set = JCROP_POS;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch xoffset */
+    info->crop_xoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_xoffset))
+      return FALSE;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch yoffset */
+    info->crop_yoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_yoffset))
+      return FALSE;
+  }
+  /* We had better have gotten to the end of the string. */
+  if (*spec != '\0')
+    return FALSE;
+  info->crop = TRUE;
+  return TRUE;
+}
+
+
+/* Trim off any partial iMCUs on the indicated destination edge */
+
+LOCAL(void)
+trim_right_edge (jpeg_transform_info *info, JDIMENSION full_width)
+{
+  JDIMENSION MCU_cols;
+
+  MCU_cols = info->output_width / info->iMCU_sample_width;
+  if (MCU_cols > 0 && info->x_crop_offset + MCU_cols ==
+      full_width / info->iMCU_sample_width)
+    info->output_width = MCU_cols * info->iMCU_sample_width;
+}
+
+LOCAL(void)
+trim_bottom_edge (jpeg_transform_info *info, JDIMENSION full_height)
+{
+  JDIMENSION MCU_rows;
+
+  MCU_rows = info->output_height / info->iMCU_sample_height;
+  if (MCU_rows > 0 && info->y_crop_offset + MCU_rows ==
+      full_height / info->iMCU_sample_height)
+    info->output_height = MCU_rows * info->iMCU_sample_height;
+}
+
+
+/* Request any required workspace.
+ *
+ * This routine figures out the size that the output image will be
+ * (which implies that all the transform parameters must be set before
+ * it is called).
+ *
+ * We allocate the workspace virtual arrays from the source decompression
+ * object, so that all the arrays (both the original data and the workspace)
+ * will be taken into account while making memory management decisions.
+ * Hence, this routine must be called after jpeg_read_header (which reads
+ * the image dimensions) and before jpeg_read_coefficients (which realizes
+ * the source's virtual arrays).
+ *
+ * This function returns FALSE right away if -perfect is given
+ * and transformation is not perfect.  Otherwise returns TRUE.
+ */
+
+GLOBAL(boolean)
+jtransform_request_workspace (j_decompress_ptr srcinfo,
+                              jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *coef_arrays;
+  boolean need_workspace, transpose_it;
+  jpeg_component_info *compptr;
+  JDIMENSION xoffset, yoffset;
+  JDIMENSION width_in_iMCUs, height_in_iMCUs;
+  JDIMENSION width_in_blocks, height_in_blocks;
+  int ci, h_samp_factor, v_samp_factor;
+
+  /* Determine number of components in output image */
+  if (info->force_grayscale &&
+      srcinfo->jpeg_color_space == JCS_YCbCr &&
+      srcinfo->num_components == 3)
+    /* We'll only process the first component */
+    info->num_components = 1;
+  else
+    /* Process all the components */
+    info->num_components = srcinfo->num_components;
+
+  /* Compute output image dimensions and related values. */
+#if JPEG_LIB_VERSION >= 80
+  jpeg_core_output_dimensions(srcinfo);
+#else
+  srcinfo->output_width = srcinfo->image_width;
+  srcinfo->output_height = srcinfo->image_height;
+#endif
+
+  /* Return right away if -perfect is given and transformation is not perfect.
+   */
+  if (info->perfect) {
+    if (info->num_components == 1) {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    } else {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    }
+  }
+
+  /* If there is only one output component, force the iMCU size to be 1;
+   * else use the source iMCU size.  (This allows us to do the right thing
+   * when reducing color to grayscale, and also provides a handy way of
+   * cleaning up "funny" grayscale images whose sampling factors are not 1x1.)
+   */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+    info->output_width = srcinfo->output_height;
+    info->output_height = srcinfo->output_width;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_h_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+    }
+    break;
+  default:
+    info->output_width = srcinfo->output_width;
+    info->output_height = srcinfo->output_height;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_v_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+    }
+    break;
+  }
+
+  /* If cropping has been requested, compute the crop area's position and
+   * dimensions, ensuring that its upper left corner falls at an iMCU boundary.
+   */
+  if (info->crop) {
+    /* Insert default values for unset crop parameters */
+    if (info->crop_xoffset_set == JCROP_UNSET)
+      info->crop_xoffset = 0;   /* default to +0 */
+    if (info->crop_yoffset_set == JCROP_UNSET)
+      info->crop_yoffset = 0;   /* default to +0 */
+    if (info->crop_xoffset >= info->output_width ||
+        info->crop_yoffset >= info->output_height)
+      ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+    if (info->crop_width_set == JCROP_UNSET)
+      info->crop_width = info->output_width - info->crop_xoffset;
+    if (info->crop_height_set == JCROP_UNSET)
+      info->crop_height = info->output_height - info->crop_yoffset;
+    /* Ensure parameters are valid */
+    if (info->crop_width <= 0 || info->crop_width > info->output_width ||
+        info->crop_height <= 0 || info->crop_height > info->output_height ||
+        info->crop_xoffset > info->output_width - info->crop_width ||
+        info->crop_yoffset > info->output_height - info->crop_height)
+      ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+    /* Convert negative crop offsets into regular offsets */
+    if (info->crop_xoffset_set == JCROP_NEG)
+      xoffset = info->output_width - info->crop_width - info->crop_xoffset;
+    else
+      xoffset = info->crop_xoffset;
+    if (info->crop_yoffset_set == JCROP_NEG)
+      yoffset = info->output_height - info->crop_height - info->crop_yoffset;
+    else
+      yoffset = info->crop_yoffset;
+    /* Now adjust so that upper left corner falls at an iMCU boundary */
+    if (info->crop_width_set == JCROP_FORCE)
+      info->output_width = info->crop_width;
+    else
+      info->output_width =
+        info->crop_width + (xoffset % info->iMCU_sample_width);
+    if (info->crop_height_set == JCROP_FORCE)
+      info->output_height = info->crop_height;
+    else
+      info->output_height =
+        info->crop_height + (yoffset % info->iMCU_sample_height);
+    /* Save x/y offsets measured in iMCUs */
+    info->x_crop_offset = xoffset / info->iMCU_sample_width;
+    info->y_crop_offset = yoffset / info->iMCU_sample_height;
+  } else {
+    info->x_crop_offset = 0;
+    info->y_crop_offset = 0;
+  }
+
+  /* Figure out whether we need workspace arrays,
+   * and if so whether they are transposed relative to the source.
+   */
+  need_workspace = FALSE;
+  transpose_it = FALSE;
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->x_crop_offset != 0 || info->y_crop_offset != 0)
+      need_workspace = TRUE;
+    /* No workspace needed if neither cropping nor transforming */
+    break;
+  case JXFORM_FLIP_H:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_width);
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      need_workspace = TRUE;
+    /* do_flip_h_no_crop doesn't need a workspace array */
+    break;
+  case JXFORM_FLIP_V:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_TRANSPOSE:
+    /* transpose does NOT have to trim anything */
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_TRANSVERSE:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_height);
+      trim_bottom_edge(info, srcinfo->output_width);
+    }
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_90:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_180:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_width);
+      trim_bottom_edge(info, srcinfo->output_height);
+    }
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_ROT_270:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_width);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  }
+
+  /* Allocate workspace if needed.
+   * Note that we allocate arrays padded out to the next iMCU boundary,
+   * so that transform routines need not worry about missing edge blocks.
+   */
+  if (need_workspace) {
+    coef_arrays = (jvirt_barray_ptr *)
+      (*srcinfo->mem->alloc_small) ((j_common_ptr) srcinfo, JPOOL_IMAGE,
+                sizeof(jvirt_barray_ptr) * info->num_components);
+    width_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long) info->output_width,
+                    (long) info->iMCU_sample_width);
+    height_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long) info->output_height,
+                    (long) info->iMCU_sample_height);
+    for (ci = 0; ci < info->num_components; ci++) {
+      compptr = srcinfo->comp_info + ci;
+      if (info->num_components == 1) {
+        /* we're going to force samp factors to 1x1 in this case */
+        h_samp_factor = v_samp_factor = 1;
+      } else if (transpose_it) {
+        h_samp_factor = compptr->v_samp_factor;
+        v_samp_factor = compptr->h_samp_factor;
+      } else {
+        h_samp_factor = compptr->h_samp_factor;
+        v_samp_factor = compptr->v_samp_factor;
+      }
+      width_in_blocks = width_in_iMCUs * h_samp_factor;
+      height_in_blocks = height_in_iMCUs * v_samp_factor;
+      coef_arrays[ci] = (*srcinfo->mem->request_virt_barray)
+        ((j_common_ptr) srcinfo, JPOOL_IMAGE, FALSE,
+         width_in_blocks, height_in_blocks, (JDIMENSION) v_samp_factor);
+    }
+    info->workspace_coef_arrays = coef_arrays;
+  } else
+    info->workspace_coef_arrays = NULL;
+
+  return TRUE;
+}
+
+
+/* Transpose destination image parameters */
+
+LOCAL(void)
+transpose_critical_parameters (j_compress_ptr dstinfo)
+{
+  int tblno, i, j, ci, itemp;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtblptr;
+  JDIMENSION jtemp;
+  UINT16 qtemp;
+
+  /* Transpose image dimensions */
+  jtemp = dstinfo->image_width;
+  dstinfo->image_width = dstinfo->image_height;
+  dstinfo->image_height = jtemp;
+#if JPEG_LIB_VERSION >= 70
+  itemp = dstinfo->min_DCT_h_scaled_size;
+  dstinfo->min_DCT_h_scaled_size = dstinfo->min_DCT_v_scaled_size;
+  dstinfo->min_DCT_v_scaled_size = itemp;
+#endif
+
+  /* Transpose sampling factors */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    itemp = compptr->h_samp_factor;
+    compptr->h_samp_factor = compptr->v_samp_factor;
+    compptr->v_samp_factor = itemp;
+  }
+
+  /* Transpose quantization tables */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    qtblptr = dstinfo->quant_tbl_ptrs[tblno];
+    if (qtblptr != NULL) {
+      for (i = 0; i < DCTSIZE; i++) {
+        for (j = 0; j < i; j++) {
+          qtemp = qtblptr->quantval[i*DCTSIZE+j];
+          qtblptr->quantval[i*DCTSIZE+j] = qtblptr->quantval[j*DCTSIZE+i];
+          qtblptr->quantval[j*DCTSIZE+i] = qtemp;
+        }
+      }
+    }
+  }
+}
+
+
+/* Adjust Exif image parameters.
+ *
+ * We try to adjust the Tags ExifImageWidth and ExifImageHeight if possible.
+ */
+
+#if JPEG_LIB_VERSION >= 70
+LOCAL(void)
+adjust_exif_parameters (JOCTET *data, unsigned int length,
+                        JDIMENSION new_width, JDIMENSION new_height)
+{
+  boolean is_motorola; /* Flag for byte order */
+  unsigned int number_of_tags, tagnum;
+  unsigned int firstoffset, offset;
+  JDIMENSION new_value;
+
+  if (length < 12) return; /* Length of an IFD entry */
+
+  /* Discover byte order */
+  if (GETJOCTET(data[0]) == 0x49 && GETJOCTET(data[1]) == 0x49)
+    is_motorola = FALSE;
+  else if (GETJOCTET(data[0]) == 0x4D && GETJOCTET(data[1]) == 0x4D)
+    is_motorola = TRUE;
+  else
+    return;
+
+  /* Check Tag Mark */
+  if (is_motorola) {
+    if (GETJOCTET(data[2]) != 0) return;
+    if (GETJOCTET(data[3]) != 0x2A) return;
+  } else {
+    if (GETJOCTET(data[3]) != 0) return;
+    if (GETJOCTET(data[2]) != 0x2A) return;
+  }
+
+  /* Get first IFD offset (offset to IFD0) */
+  if (is_motorola) {
+    if (GETJOCTET(data[4]) != 0) return;
+    if (GETJOCTET(data[5]) != 0) return;
+    firstoffset = GETJOCTET(data[6]);
+    firstoffset <<= 8;
+    firstoffset += GETJOCTET(data[7]);
+  } else {
+    if (GETJOCTET(data[7]) != 0) return;
+    if (GETJOCTET(data[6]) != 0) return;
+    firstoffset = GETJOCTET(data[5]);
+    firstoffset <<= 8;
+    firstoffset += GETJOCTET(data[4]);
+  }
+  if (firstoffset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this IFD */
+  if (is_motorola) {
+    number_of_tags = GETJOCTET(data[firstoffset]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[firstoffset+1]);
+  } else {
+    number_of_tags = GETJOCTET(data[firstoffset+1]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[firstoffset]);
+  }
+  if (number_of_tags == 0) return;
+  firstoffset += 2;
+
+  /* Search for ExifSubIFD offset Tag in IFD0 */
+  for (;;) {
+    if (firstoffset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = GETJOCTET(data[firstoffset]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[firstoffset+1]);
+    } else {
+      tagnum = GETJOCTET(data[firstoffset+1]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[firstoffset]);
+    }
+    if (tagnum == 0x8769) break; /* found ExifSubIFD offset Tag */
+    if (--number_of_tags == 0) return;
+    firstoffset += 12;
+  }
+
+  /* Get the ExifSubIFD offset */
+  if (is_motorola) {
+    if (GETJOCTET(data[firstoffset+8]) != 0) return;
+    if (GETJOCTET(data[firstoffset+9]) != 0) return;
+    offset = GETJOCTET(data[firstoffset+10]);
+    offset <<= 8;
+    offset += GETJOCTET(data[firstoffset+11]);
+  } else {
+    if (GETJOCTET(data[firstoffset+11]) != 0) return;
+    if (GETJOCTET(data[firstoffset+10]) != 0) return;
+    offset = GETJOCTET(data[firstoffset+9]);
+    offset <<= 8;
+    offset += GETJOCTET(data[firstoffset+8]);
+  }
+  if (offset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this SubIFD */
+  if (is_motorola) {
+    number_of_tags = GETJOCTET(data[offset]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[offset+1]);
+  } else {
+    number_of_tags = GETJOCTET(data[offset+1]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[offset]);
+  }
+  if (number_of_tags < 2) return;
+  offset += 2;
+
+  /* Search for ExifImageWidth and ExifImageHeight Tags in this SubIFD */
+  do {
+    if (offset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = GETJOCTET(data[offset]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[offset+1]);
+    } else {
+      tagnum = GETJOCTET(data[offset+1]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[offset]);
+    }
+    if (tagnum == 0xA002 || tagnum == 0xA003) {
+      if (tagnum == 0xA002)
+        new_value = new_width; /* ExifImageWidth Tag */
+      else
+        new_value = new_height; /* ExifImageHeight Tag */
+      if (is_motorola) {
+        data[offset+2] = 0; /* Format = unsigned long (4 octets) */
+        data[offset+3] = 4;
+        data[offset+4] = 0; /* Number Of Components = 1 */
+        data[offset+5] = 0;
+        data[offset+6] = 0;
+        data[offset+7] = 1;
+        data[offset+8] = 0;
+        data[offset+9] = 0;
+        data[offset+10] = (JOCTET)((new_value >> 8) & 0xFF);
+        data[offset+11] = (JOCTET)(new_value & 0xFF);
+      } else {
+        data[offset+2] = 4; /* Format = unsigned long (4 octets) */
+        data[offset+3] = 0;
+        data[offset+4] = 1; /* Number Of Components = 1 */
+        data[offset+5] = 0;
+        data[offset+6] = 0;
+        data[offset+7] = 0;
+        data[offset+8] = (JOCTET)(new_value & 0xFF);
+        data[offset+9] = (JOCTET)((new_value >> 8) & 0xFF);
+        data[offset+10] = 0;
+        data[offset+11] = 0;
+      }
+    }
+    offset += 12;
+  } while (--number_of_tags);
+}
+#endif
+
+
+/* Adjust output image parameters as needed.
+ *
+ * This must be called after jpeg_copy_critical_parameters()
+ * and before jpeg_write_coefficients().
+ *
+ * The return value is the set of virtual coefficient arrays to be written
+ * (either the ones allocated by jtransform_request_workspace, or the
+ * original source data arrays).  The caller will need to pass this value
+ * to jpeg_write_coefficients().
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jtransform_adjust_parameters (j_decompress_ptr srcinfo,
+                              j_compress_ptr dstinfo,
+                              jvirt_barray_ptr *src_coef_arrays,
+                              jpeg_transform_info *info)
+{
+  /* If force-to-grayscale is requested, adjust destination parameters */
+  if (info->force_grayscale) {
+    /* First, ensure we have YCbCr or grayscale data, and that the source's
+     * Y channel is full resolution.  (No reasonable person would make Y
+     * be less than full resolution, so actually coping with that case
+     * isn't worth extra code space.  But we check it to avoid crashing.)
+     */
+    if (((dstinfo->jpeg_color_space == JCS_YCbCr &&
+          dstinfo->num_components == 3) ||
+         (dstinfo->jpeg_color_space == JCS_GRAYSCALE &&
+          dstinfo->num_components == 1)) &&
+        srcinfo->comp_info[0].h_samp_factor == srcinfo->max_h_samp_factor &&
+        srcinfo->comp_info[0].v_samp_factor == srcinfo->max_v_samp_factor) {
+      /* We use jpeg_set_colorspace to make sure subsidiary settings get fixed
+       * properly.  Among other things, it sets the target h_samp_factor &
+       * v_samp_factor to 1, which typically won't match the source.
+       * We have to preserve the source's quantization table number, however.
+       */
+      int sv_quant_tbl_no = dstinfo->comp_info[0].quant_tbl_no;
+      jpeg_set_colorspace(dstinfo, JCS_GRAYSCALE);
+      dstinfo->comp_info[0].quant_tbl_no = sv_quant_tbl_no;
+    } else {
+      /* Sorry, can't do it */
+      ERREXIT(dstinfo, JERR_CONVERSION_NOTIMPL);
+    }
+  } else if (info->num_components == 1) {
+    /* For a single-component source, we force the destination sampling factors
+     * to 1x1, with or without force_grayscale.  This is useful because some
+     * decoders choke on grayscale images with other sampling factors.
+     */
+    dstinfo->comp_info[0].h_samp_factor = 1;
+    dstinfo->comp_info[0].v_samp_factor = 1;
+  }
+
+  /* Correct the destination's image dimensions as necessary
+   * for rotate/flip, resize, and crop operations.
+   */
+#if JPEG_LIB_VERSION >= 70
+  dstinfo->jpeg_width = info->output_width;
+  dstinfo->jpeg_height = info->output_height;
+#endif
+
+  /* Transpose destination image parameters */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+#if JPEG_LIB_VERSION < 70
+    dstinfo->image_width = info->output_height;
+    dstinfo->image_height = info->output_width;
+#endif
+    transpose_critical_parameters(dstinfo);
+    break;
+  default:
+#if JPEG_LIB_VERSION < 70
+    dstinfo->image_width = info->output_width;
+    dstinfo->image_height = info->output_height;
+#endif
+    break;
+  }
+
+  /* Adjust Exif properties */
+  if (srcinfo->marker_list != NULL &&
+      srcinfo->marker_list->marker == JPEG_APP0+1 &&
+      srcinfo->marker_list->data_length >= 6 &&
+      GETJOCTET(srcinfo->marker_list->data[0]) == 0x45 &&
+      GETJOCTET(srcinfo->marker_list->data[1]) == 0x78 &&
+      GETJOCTET(srcinfo->marker_list->data[2]) == 0x69 &&
+      GETJOCTET(srcinfo->marker_list->data[3]) == 0x66 &&
+      GETJOCTET(srcinfo->marker_list->data[4]) == 0 &&
+      GETJOCTET(srcinfo->marker_list->data[5]) == 0) {
+    /* Suppress output of JFIF marker */
+    dstinfo->write_JFIF_header = FALSE;
+#if JPEG_LIB_VERSION >= 70
+    /* Adjust Exif image parameters */
+    if (dstinfo->jpeg_width != srcinfo->image_width ||
+        dstinfo->jpeg_height != srcinfo->image_height)
+      /* Align data segment to start of TIFF structure for parsing */
+      adjust_exif_parameters(srcinfo->marker_list->data + 6,
+        srcinfo->marker_list->data_length - 6,
+        dstinfo->jpeg_width, dstinfo->jpeg_height);
+#endif
+  }
+
+  /* Return the appropriate output data set */
+  if (info->workspace_coef_arrays != NULL)
+    return info->workspace_coef_arrays;
+  return src_coef_arrays;
+}
+
+
+/* Execute the actual transformation, if any.
+ *
+ * This must be called *after* jpeg_write_coefficients, because it depends
+ * on jpeg_write_coefficients to have computed subsidiary values such as
+ * the per-component width and height fields in the destination object.
+ *
+ * Note that some transformations will modify the source data arrays!
+ */
+
+GLOBAL(void)
+jtransform_execute_transform (j_decompress_ptr srcinfo,
+                              j_compress_ptr dstinfo,
+                              jvirt_barray_ptr *src_coef_arrays,
+                              jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *dst_coef_arrays = info->workspace_coef_arrays;
+
+  /* Note: conditions tested here should match those in switch statement
+   * in jtransform_request_workspace()
+   */
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->x_crop_offset != 0 || info->y_crop_offset != 0)
+      do_crop(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_FLIP_H:
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      do_flip_h(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                src_coef_arrays, dst_coef_arrays);
+    else
+      do_flip_h_no_crop(srcinfo, dstinfo, info->x_crop_offset,
+                        src_coef_arrays);
+    break;
+  case JXFORM_FLIP_V:
+    do_flip_v(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSPOSE:
+    do_transpose(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                 src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSVERSE:
+    do_transverse(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                  src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_90:
+    do_rot_90(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_180:
+    do_rot_180(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_270:
+    do_rot_270(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  }
+}
+
+/* jtransform_perfect_transform
+ *
+ * Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ *
+ * Inputs:
+ *   image_width, image_height: source image dimensions.
+ *   MCU_width, MCU_height: pixel dimensions of MCU.
+ *   transform: transformation identifier.
+ * Parameter sources from initialized jpeg_struct
+ * (after reading source header):
+ *   image_width = cinfo.image_width
+ *   image_height = cinfo.image_height
+ *   MCU_width = cinfo.max_h_samp_factor * cinfo.block_size
+ *   MCU_height = cinfo.max_v_samp_factor * cinfo.block_size
+ * Result:
+ *   TRUE = perfect transformation possible
+ *   FALSE = perfect transformation not possible
+ *           (may use custom action then)
+ */
+
+GLOBAL(boolean)
+jtransform_perfect_transform(JDIMENSION image_width, JDIMENSION image_height,
+                             int MCU_width, int MCU_height,
+                             JXFORM_CODE transform)
+{
+  boolean result = TRUE; /* initialize TRUE */
+
+  switch (transform) {
+  case JXFORM_FLIP_H:
+  case JXFORM_ROT_270:
+    if (image_width % (JDIMENSION) MCU_width)
+      result = FALSE;
+    break;
+  case JXFORM_FLIP_V:
+  case JXFORM_ROT_90:
+    if (image_height % (JDIMENSION) MCU_height)
+      result = FALSE;
+    break;
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_180:
+    if (image_width % (JDIMENSION) MCU_width)
+      result = FALSE;
+    if (image_height % (JDIMENSION) MCU_height)
+      result = FALSE;
+    break;
+  default:
+    break;
+  }
+
+  return result;
+}
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/* Setup decompression object to save desired markers in memory.
+ * This must be called before jpeg_read_header() to have the desired effect.
+ */
+
+GLOBAL(void)
+jcopy_markers_setup (j_decompress_ptr srcinfo, JCOPY_OPTION option)
+{
+#ifdef SAVE_MARKERS_SUPPORTED
+  int m;
+
+  /* Save comments except under NONE option */
+  if (option != JCOPYOPT_NONE) {
+    jpeg_save_markers(srcinfo, JPEG_COM, 0xFFFF);
+  }
+  /* Save all types of APPn markers iff ALL option */
+  if (option == JCOPYOPT_ALL) {
+    for (m = 0; m < 16; m++)
+      jpeg_save_markers(srcinfo, JPEG_APP0 + m, 0xFFFF);
+  }
+#endif /* SAVE_MARKERS_SUPPORTED */
+}
+
+/* Copy markers saved in the given source object to the destination object.
+ * This should be called just after jpeg_start_compress() or
+ * jpeg_write_coefficients().
+ * Note that those routines will have written the SOI, and also the
+ * JFIF APP0 or Adobe APP14 markers if selected.
+ */
+
+GLOBAL(void)
+jcopy_markers_execute (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                       JCOPY_OPTION option)
+{
+  jpeg_saved_marker_ptr marker;
+
+  /* In the current implementation, we don't actually need to examine the
+   * option flag here; we just copy everything that got saved.
+   * But to avoid confusion, we do not output JFIF and Adobe APP14 markers
+   * if the encoder library already wrote one.
+   */
+  for (marker = srcinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (dstinfo->write_JFIF_header &&
+        marker->marker == JPEG_APP0 &&
+        marker->data_length >= 5 &&
+        GETJOCTET(marker->data[0]) == 0x4A &&
+        GETJOCTET(marker->data[1]) == 0x46 &&
+        GETJOCTET(marker->data[2]) == 0x49 &&
+        GETJOCTET(marker->data[3]) == 0x46 &&
+        GETJOCTET(marker->data[4]) == 0)
+      continue;                 /* reject duplicate JFIF */
+    if (dstinfo->write_Adobe_marker &&
+        marker->marker == JPEG_APP0+14 &&
+        marker->data_length >= 5 &&
+        GETJOCTET(marker->data[0]) == 0x41 &&
+        GETJOCTET(marker->data[1]) == 0x64 &&
+        GETJOCTET(marker->data[2]) == 0x6F &&
+        GETJOCTET(marker->data[3]) == 0x62 &&
+        GETJOCTET(marker->data[4]) == 0x65)
+      continue;                 /* reject duplicate Adobe */
+    jpeg_write_marker(dstinfo, marker->marker,
+                      marker->data, marker->data_length);
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/transupp.h glmark2-2021.02/android/jni/src/libjpeg-turbo/transupp.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/transupp.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/transupp.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,207 @@
+/*
+ * transupp.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2011, Thomas G. Lane, Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains declarations for image transformation routines and
+ * other utility code used by the jpegtran sample application.  These are
+ * NOT part of the core JPEG library.  But we keep these routines separate
+ * from jpegtran.c to ease the task of maintaining jpegtran-like programs
+ * that have other user interfaces.
+ *
+ * NOTE: all the routines declared here have very specific requirements
+ * about when they are to be executed during the reading and writing of the
+ * source and destination files.  See the comments in transupp.c, or see
+ * jpegtran.c for an example of correct usage.
+ */
+
+/* If you happen not to want the image transform support, disable it here */
+#ifndef TRANSFORMS_SUPPORTED
+#define TRANSFORMS_SUPPORTED 1          /* 0 disables transform code */
+#endif
+
+/*
+ * Although rotating and flipping data expressed as DCT coefficients is not
+ * hard, there is an asymmetry in the JPEG format specification for images
+ * whose dimensions aren't multiples of the iMCU size.  The right and bottom
+ * image edges are padded out to the next iMCU boundary with junk data; but
+ * no padding is possible at the top and left edges.  If we were to flip
+ * the whole image including the pad data, then pad garbage would become
+ * visible at the top and/or left, and real pixels would disappear into the
+ * pad margins --- perhaps permanently, since encoders & decoders may not
+ * bother to preserve DCT blocks that appear to be completely outside the
+ * nominal image area.  So, we have to exclude any partial iMCUs from the
+ * basic transformation.
+ *
+ * Transpose is the only transformation that can handle partial iMCUs at the
+ * right and bottom edges completely cleanly.  flip_h can flip partial iMCUs
+ * at the bottom, but leaves any partial iMCUs at the right edge untouched.
+ * Similarly flip_v leaves any partial iMCUs at the bottom edge untouched.
+ * The other transforms are defined as combinations of these basic transforms
+ * and process edge blocks in a way that preserves the equivalence.
+ *
+ * The "trim" option causes untransformable partial iMCUs to be dropped;
+ * this is not strictly lossless, but it usually gives the best-looking
+ * result for odd-size images.  Note that when this option is active,
+ * the expected mathematical equivalences between the transforms may not hold.
+ * (For example, -rot 270 -trim trims only the bottom edge, but -rot 90 -trim
+ * followed by -rot 180 -trim trims both edges.)
+ *
+ * We also offer a lossless-crop option, which discards data outside a given
+ * image region but losslessly preserves what is inside.  Like the rotate and
+ * flip transforms, lossless crop is restricted by the JPEG format: the upper
+ * left corner of the selected region must fall on an iMCU boundary.  If this
+ * does not hold for the given crop parameters, we silently move the upper left
+ * corner up and/or left to make it so, simultaneously increasing the region
+ * dimensions to keep the lower right crop corner unchanged.  (Thus, the
+ * output image covers at least the requested region, but may cover more.)
+ * The adjustment of the region dimensions may be optionally disabled.
+ *
+ * We also provide a lossless-resize option, which is kind of a lossless-crop
+ * operation in the DCT coefficient block domain - it discards higher-order
+ * coefficients and losslessly preserves lower-order coefficients of a
+ * sub-block.
+ *
+ * Rotate/flip transform, resize, and crop can be requested together in a
+ * single invocation.  The crop is applied last --- that is, the crop region
+ * is specified in terms of the destination image after transform/resize.
+ *
+ * We also offer a "force to grayscale" option, which simply discards the
+ * chrominance channels of a YCbCr image.  This is lossless in the sense that
+ * the luminance channel is preserved exactly.  It's not the same kind of
+ * thing as the rotate/flip transformations, but it's convenient to handle it
+ * as part of this package, mainly because the transformation routines have to
+ * be aware of the option to know how many components to work on.
+ */
+
+
+/*
+ * Codes for supported types of image transformations.
+ */
+
+typedef enum {
+  JXFORM_NONE,            /* no transformation */
+  JXFORM_FLIP_H,          /* horizontal flip */
+  JXFORM_FLIP_V,          /* vertical flip */
+  JXFORM_TRANSPOSE,       /* transpose across UL-to-LR axis */
+  JXFORM_TRANSVERSE,      /* transpose across UR-to-LL axis */
+  JXFORM_ROT_90,          /* 90-degree clockwise rotation */
+  JXFORM_ROT_180,         /* 180-degree rotation */
+  JXFORM_ROT_270          /* 270-degree clockwise (or 90 ccw) */
+} JXFORM_CODE;
+
+/*
+ * Codes for crop parameters, which can individually be unspecified,
+ * positive or negative for xoffset or yoffset,
+ * positive or forced for width or height.
+ */
+
+typedef enum {
+  JCROP_UNSET,
+  JCROP_POS,
+  JCROP_NEG,
+  JCROP_FORCE
+} JCROP_CODE;
+
+/*
+ * Transform parameters struct.
+ * NB: application must not change any elements of this struct after
+ * calling jtransform_request_workspace.
+ */
+
+typedef struct {
+  /* Options: set by caller */
+  JXFORM_CODE transform;        /* image transform operator */
+  boolean perfect;              /* if TRUE, fail if partial MCUs are requested */
+  boolean trim;                 /* if TRUE, trim partial MCUs as needed */
+  boolean force_grayscale;      /* if TRUE, convert color image to grayscale */
+  boolean crop;                 /* if TRUE, crop source image */
+  boolean slow_hflip;  /* For best performance, the JXFORM_FLIP_H transform
+                          normally modifies the source coefficients in place.
+                          Setting this to TRUE will instead use a slower,
+                          double-buffered algorithm, which leaves the source
+                          coefficients in tact (necessary if other transformed
+                          images must be generated from the same set of
+                          coefficients. */
+
+  /* Crop parameters: application need not set these unless crop is TRUE.
+   * These can be filled in by jtransform_parse_crop_spec().
+   */
+  JDIMENSION crop_width;        /* Width of selected region */
+  JCROP_CODE crop_width_set;    /* (forced disables adjustment) */
+  JDIMENSION crop_height;       /* Height of selected region */
+  JCROP_CODE crop_height_set;   /* (forced disables adjustment) */
+  JDIMENSION crop_xoffset;      /* X offset of selected region */
+  JCROP_CODE crop_xoffset_set;  /* (negative measures from right edge) */
+  JDIMENSION crop_yoffset;      /* Y offset of selected region */
+  JCROP_CODE crop_yoffset_set;  /* (negative measures from bottom edge) */
+
+  /* Internal workspace: caller should not touch these */
+  int num_components;           /* # of components in workspace */
+  jvirt_barray_ptr *workspace_coef_arrays; /* workspace for transformations */
+  JDIMENSION output_width;      /* cropped destination dimensions */
+  JDIMENSION output_height;
+  JDIMENSION x_crop_offset;     /* destination crop offsets measured in iMCUs */
+  JDIMENSION y_crop_offset;
+  int iMCU_sample_width;        /* destination iMCU size */
+  int iMCU_sample_height;
+} jpeg_transform_info;
+
+
+#if TRANSFORMS_SUPPORTED
+
+/* Parse a crop specification (written in X11 geometry style) */
+EXTERN(boolean) jtransform_parse_crop_spec
+        (jpeg_transform_info *info, const char *spec);
+/* Request any required workspace */
+EXTERN(boolean) jtransform_request_workspace
+        (j_decompress_ptr srcinfo, jpeg_transform_info *info);
+/* Adjust output image parameters */
+EXTERN(jvirt_barray_ptr *) jtransform_adjust_parameters
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         jvirt_barray_ptr *src_coef_arrays, jpeg_transform_info *info);
+/* Execute the actual transformation, if any */
+EXTERN(void) jtransform_execute_transform
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         jvirt_barray_ptr *src_coef_arrays, jpeg_transform_info *info);
+/* Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ */
+EXTERN(boolean) jtransform_perfect_transform
+        (JDIMENSION image_width, JDIMENSION image_height, int MCU_width,
+         int MCU_height, JXFORM_CODE transform);
+
+/* jtransform_execute_transform used to be called
+ * jtransform_execute_transformation, but some compilers complain about
+ * routine names that long.  This macro is here to avoid breaking any
+ * old source code that uses the original name...
+ */
+#define jtransform_execute_transformation       jtransform_execute_transform
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/*
+ * Support for copying optional markers from source to destination file.
+ */
+
+typedef enum {
+  JCOPYOPT_NONE,          /* copy no optional markers */
+  JCOPYOPT_COMMENTS,      /* copy only comment (COM) markers */
+  JCOPYOPT_ALL            /* copy all optional markers */
+} JCOPY_OPTION;
+
+#define JCOPYOPT_DEFAULT  JCOPYOPT_COMMENTS     /* recommended default */
+
+/* Setup decompression object to save desired markers in memory */
+EXTERN(void) jcopy_markers_setup
+        (j_decompress_ptr srcinfo, JCOPY_OPTION option);
+/* Copy markers saved in the given source object to the destination object */
+EXTERN(void) jcopy_markers_execute
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         JCOPY_OPTION option);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg.c glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,2121 @@
+/*
+ * Copyright (C)2009-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* TurboJPEG/LJT:  this implements the TurboJPEG API using libjpeg or
+   libjpeg-turbo */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <jinclude.h>
+#define JPEG_INTERNALS
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+#include "./turbojpeg.h"
+#include "./tjutil.h"
+#include "transupp.h"
+#include "./jpegcomp.h"
+
+extern void jpeg_mem_dest_tj(j_compress_ptr, unsigned char **,
+	unsigned long *, boolean);
+extern void jpeg_mem_src_tj(j_decompress_ptr, const unsigned char *,
+	unsigned long);
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+#define isPow2(x) (((x)&(x-1))==0)
+
+
+/* Error handling (based on example in example.c) */
+
+static char errStr[JMSG_LENGTH_MAX]="No error";
+
+struct my_error_mgr
+{
+	struct jpeg_error_mgr pub;
+	jmp_buf setjmp_buffer;
+	void (*emit_message)(j_common_ptr, int);
+	boolean warning;
+};
+typedef struct my_error_mgr *my_error_ptr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	(*cinfo->err->output_message)(cinfo);
+	longjmp(myerr->setjmp_buffer, 1);
+}
+
+/* Based on output_message() in jerror.c */
+
+static void my_output_message(j_common_ptr cinfo)
+{
+	(*cinfo->err->format_message)(cinfo, errStr);
+}
+
+static void my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	myerr->emit_message(cinfo, msg_level);
+	if(msg_level<0) myerr->warning=TRUE;
+}
+
+
+/* Global structures, macros, etc. */
+
+enum {COMPRESS=1, DECOMPRESS=2};
+
+typedef struct _tjinstance
+{
+	struct jpeg_compress_struct cinfo;
+	struct jpeg_decompress_struct dinfo;
+	struct my_error_mgr jerr;
+	int init, headerRead;
+} tjinstance;
+
+static const int pixelsize[TJ_NUMSAMP]={3, 3, 3, 1, 3, 3};
+
+static const JXFORM_CODE xformtypes[TJ_NUMXOP]=
+{
+	JXFORM_NONE, JXFORM_FLIP_H, JXFORM_FLIP_V, JXFORM_TRANSPOSE,
+	JXFORM_TRANSVERSE, JXFORM_ROT_90, JXFORM_ROT_180, JXFORM_ROT_270
+};
+
+#define NUMSF 16
+static const tjscalingfactor sf[NUMSF]={
+	{2, 1},
+	{15, 8},
+	{7, 4},
+	{13, 8},
+	{3, 2},
+	{11, 8},
+	{5, 4},
+	{9, 8},
+	{1, 1},
+	{7, 8},
+	{3, 4},
+	{5, 8},
+	{1, 2},
+	{3, 8},
+	{1, 4},
+	{1, 8}
+};
+
+#define _throw(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s", m);  \
+	retval=-1;  goto bailout;}
+#define getinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_compress_ptr cinfo=NULL;  j_decompress_ptr dinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	cinfo=&this->cinfo;  dinfo=&this->dinfo;  \
+	this->jerr.warning=FALSE;
+#define getcinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_compress_ptr cinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	cinfo=&this->cinfo;  \
+	this->jerr.warning=FALSE;
+#define getdinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_decompress_ptr dinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	dinfo=&this->dinfo;  \
+	this->jerr.warning=FALSE;
+
+static int getPixelFormat(int pixelSize, int flags)
+{
+	if(pixelSize==1) return TJPF_GRAY;
+	if(pixelSize==3)
+	{
+		if(flags&TJ_BGR) return TJPF_BGR;
+		else return TJPF_RGB;
+	}
+	if(pixelSize==4)
+	{
+		if(flags&TJ_ALPHAFIRST)
+		{
+			if(flags&TJ_BGR) return TJPF_XBGR;
+			else return TJPF_XRGB;
+		}
+		else
+		{
+			if(flags&TJ_BGR) return TJPF_BGRX;
+			else return TJPF_RGBX;
+		}
+	}
+	return -1;
+}
+
+static int setCompDefaults(struct jpeg_compress_struct *cinfo,
+	int pixelFormat, int subsamp, int jpegQual, int flags)
+{
+	int retval=0;
+	char *env=NULL;
+
+	switch(pixelFormat)
+	{
+		case TJPF_GRAY:
+			cinfo->in_color_space=JCS_GRAYSCALE;  break;
+		#if JCS_EXTENSIONS==1
+		case TJPF_RGB:
+			cinfo->in_color_space=JCS_EXT_RGB;  break;
+		case TJPF_BGR:
+			cinfo->in_color_space=JCS_EXT_BGR;  break;
+		case TJPF_RGBX:
+		case TJPF_RGBA:
+			cinfo->in_color_space=JCS_EXT_RGBX;  break;
+		case TJPF_BGRX:
+		case TJPF_BGRA:
+			cinfo->in_color_space=JCS_EXT_BGRX;  break;
+		case TJPF_XRGB:
+		case TJPF_ARGB:
+			cinfo->in_color_space=JCS_EXT_XRGB;  break;
+		case TJPF_XBGR:
+		case TJPF_ABGR:
+			cinfo->in_color_space=JCS_EXT_XBGR;  break;
+		#else
+		case TJPF_RGB:
+		case TJPF_BGR:
+		case TJPF_RGBX:
+		case TJPF_BGRX:
+		case TJPF_XRGB:
+		case TJPF_XBGR:
+		case TJPF_RGBA:
+		case TJPF_BGRA:
+		case TJPF_ARGB:
+		case TJPF_ABGR:
+			cinfo->in_color_space=JCS_RGB;  pixelFormat=TJPF_RGB;
+			break;
+		#endif
+		case TJPF_CMYK:
+			cinfo->in_color_space=JCS_CMYK;  break;
+	}
+
+	cinfo->input_components=tjPixelSize[pixelFormat];
+	jpeg_set_defaults(cinfo);
+
+#ifndef NO_GETENV
+	if((env=getenv("TJ_OPTIMIZE"))!=NULL && strlen(env)>0 && !strcmp(env, "1"))
+		cinfo->optimize_coding=TRUE;
+	if((env=getenv("TJ_ARITHMETIC"))!=NULL && strlen(env)>0	&& !strcmp(env, "1"))
+		cinfo->arith_code=TRUE;
+	if((env=getenv("TJ_RESTART"))!=NULL && strlen(env)>0)
+	{
+		int temp=-1;  char tempc=0;
+		if(sscanf(env, "%d%c", &temp, &tempc)>=1 && temp>=0 && temp<=65535)
+		{
+			if(toupper(tempc)=='B')
+			{
+				cinfo->restart_interval=temp;
+				cinfo->restart_in_rows=0;
+			}
+			else
+				cinfo->restart_in_rows=temp;
+		}
+	}
+#endif
+
+	if(jpegQual>=0)
+	{
+		jpeg_set_quality(cinfo, jpegQual, TRUE);
+		if(jpegQual>=96 || flags&TJFLAG_ACCURATEDCT) cinfo->dct_method=JDCT_ISLOW;
+		else cinfo->dct_method=JDCT_FASTEST;
+	}
+	if(subsamp==TJSAMP_GRAY)
+		jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+	else if(pixelFormat==TJPF_CMYK)
+		jpeg_set_colorspace(cinfo, JCS_YCCK);
+	else jpeg_set_colorspace(cinfo, JCS_YCbCr);
+
+#ifndef NO_GETENV
+	if((env=getenv("TJ_PROGRESSIVE"))!=NULL && strlen(env)>0
+		&& !strcmp(env, "1"))
+		jpeg_simple_progression(cinfo);
+#endif
+
+	cinfo->comp_info[0].h_samp_factor=tjMCUWidth[subsamp]/8;
+	cinfo->comp_info[1].h_samp_factor=1;
+	cinfo->comp_info[2].h_samp_factor=1;
+	if(cinfo->num_components>3)
+		cinfo->comp_info[3].h_samp_factor=tjMCUWidth[subsamp]/8;
+	cinfo->comp_info[0].v_samp_factor=tjMCUHeight[subsamp]/8;
+	cinfo->comp_info[1].v_samp_factor=1;
+	cinfo->comp_info[2].v_samp_factor=1;
+	if(cinfo->num_components>3)
+		cinfo->comp_info[3].v_samp_factor=tjMCUHeight[subsamp]/8;
+
+	return retval;
+}
+
+static int setDecompDefaults(struct jpeg_decompress_struct *dinfo,
+	int pixelFormat, int flags)
+{
+	int retval=0;
+
+	switch(pixelFormat)
+	{
+		case TJPF_GRAY:
+			dinfo->out_color_space=JCS_GRAYSCALE;  break;
+		#if JCS_EXTENSIONS==1
+		case TJPF_RGB:
+			dinfo->out_color_space=JCS_EXT_RGB;  break;
+		case TJPF_BGR:
+			dinfo->out_color_space=JCS_EXT_BGR;  break;
+		case TJPF_RGBX:
+			dinfo->out_color_space=JCS_EXT_RGBX;  break;
+		case TJPF_BGRX:
+			dinfo->out_color_space=JCS_EXT_BGRX;  break;
+		case TJPF_XRGB:
+			dinfo->out_color_space=JCS_EXT_XRGB;  break;
+		case TJPF_XBGR:
+			dinfo->out_color_space=JCS_EXT_XBGR;  break;
+		#if JCS_ALPHA_EXTENSIONS==1
+		case TJPF_RGBA:
+			dinfo->out_color_space=JCS_EXT_RGBA;  break;
+		case TJPF_BGRA:
+			dinfo->out_color_space=JCS_EXT_BGRA;  break;
+		case TJPF_ARGB:
+			dinfo->out_color_space=JCS_EXT_ARGB;  break;
+		case TJPF_ABGR:
+			dinfo->out_color_space=JCS_EXT_ABGR;  break;
+		#endif
+		#else
+		case TJPF_RGB:
+		case TJPF_BGR:
+		case TJPF_RGBX:
+		case TJPF_BGRX:
+		case TJPF_XRGB:
+		case TJPF_XBGR:
+		case TJPF_RGBA:
+		case TJPF_BGRA:
+		case TJPF_ARGB:
+		case TJPF_ABGR:
+			dinfo->out_color_space=JCS_RGB;  break;
+		#endif
+		case TJPF_CMYK:
+			dinfo->out_color_space=JCS_CMYK;  break;
+		default:
+			_throw("Unsupported pixel format");
+	}
+
+	if(flags&TJFLAG_FASTDCT) dinfo->dct_method=JDCT_FASTEST;
+
+	bailout:
+	return retval;
+}
+
+
+static int getSubsamp(j_decompress_ptr dinfo)
+{
+	int retval=-1, i, k;
+
+	/* The sampling factors actually have no meaning with grayscale JPEG files,
+	   and in fact it's possible to generate grayscale JPEGs with sampling
+	   factors > 1 (even though those sampling factors are ignored by the
+	   decompressor.)  Thus, we need to treat grayscale as a special case. */
+	if(dinfo->num_components==1 && dinfo->jpeg_color_space==JCS_GRAYSCALE)
+		return TJSAMP_GRAY;
+
+	for(i=0; i<NUMSUBOPT; i++)
+	{
+		if(dinfo->num_components==pixelsize[i]
+			|| ((dinfo->jpeg_color_space==JCS_YCCK
+				|| dinfo->jpeg_color_space==JCS_CMYK)
+					&& pixelsize[i]==3 && dinfo->num_components==4))
+		{
+			if(dinfo->comp_info[0].h_samp_factor==tjMCUWidth[i]/8
+				&& dinfo->comp_info[0].v_samp_factor==tjMCUHeight[i]/8)
+			{
+				int match=0;
+				for(k=1; k<dinfo->num_components; k++)
+				{
+					int href=1, vref=1;
+					if(dinfo->jpeg_color_space==JCS_YCCK && k==3)
+					{
+						href=tjMCUWidth[i]/8;  vref=tjMCUHeight[i]/8;
+					}
+					if(dinfo->comp_info[k].h_samp_factor==href
+						&& dinfo->comp_info[k].v_samp_factor==vref)
+						match++;
+				}
+				if(match==dinfo->num_components-1)
+				{
+					retval=i;  break;
+				}
+			}
+		}
+	}
+	return retval;
+}
+
+
+#ifndef JCS_EXTENSIONS
+
+/* Conversion functions to emulate the colorspace extensions.  This allows the
+   TurboJPEG wrapper to be used with libjpeg */
+
+#define TORGB(PS, ROFFSET, GOFFSET, BOFFSET) {  \
+	int rowPad=pitch-width*PS;  \
+	while(height--)  \
+	{  \
+		unsigned char *endOfRow=src+width*PS;  \
+		while(src<endOfRow)  \
+		{  \
+			dst[RGB_RED]=src[ROFFSET];  \
+			dst[RGB_GREEN]=src[GOFFSET];  \
+			dst[RGB_BLUE]=src[BOFFSET];  \
+			dst+=RGB_PIXELSIZE;  src+=PS;  \
+		}  \
+		src+=rowPad;  \
+	}  \
+}
+
+static unsigned char *toRGB(unsigned char *src, int width, int pitch,
+	int height, int pixelFormat, unsigned char *dst)
+{
+	unsigned char *retval=src;
+	switch(pixelFormat)
+	{
+		case TJPF_RGB:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=3
+			retval=dst;  TORGB(3, 0, 1, 2);
+			#endif
+			break;
+		case TJPF_BGR:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=3
+			retval=dst;  TORGB(3, 2, 1, 0);
+			#endif
+			break;
+		case TJPF_RGBX:
+		case TJPF_RGBA:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 0, 1, 2);
+			#endif
+			break;
+		case TJPF_BGRX:
+		case TJPF_BGRA:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 2, 1, 0);
+			#endif
+			break;
+		case TJPF_XRGB:
+		case TJPF_ARGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 1, 2, 3);
+			#endif
+			break;
+		case TJPF_XBGR:
+		case TJPF_ABGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 3, 2, 1);
+			#endif
+			break;
+	}
+	return retval;
+}
+
+#define FROMRGB(PS, ROFFSET, GOFFSET, BOFFSET, SETALPHA) {  \
+	int rowPad=pitch-width*PS;  \
+	while(height--)  \
+	{  \
+		unsigned char *endOfRow=dst+width*PS;  \
+		while(dst<endOfRow)  \
+		{  \
+			dst[ROFFSET]=src[RGB_RED];  \
+			dst[GOFFSET]=src[RGB_GREEN];  \
+			dst[BOFFSET]=src[RGB_BLUE];  \
+			SETALPHA  \
+			dst+=PS;  src+=RGB_PIXELSIZE;  \
+		}  \
+		dst+=rowPad;  \
+	}  \
+}
+
+static void fromRGB(unsigned char *src, unsigned char *dst, int width,
+	int pitch, int height, int pixelFormat)
+{
+	switch(pixelFormat)
+	{
+		case TJPF_RGB:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=3
+			FROMRGB(3, 0, 1, 2,);
+			#endif
+			break;
+		case TJPF_BGR:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=3
+			FROMRGB(3, 2, 1, 0,);
+			#endif
+			break;
+		case TJPF_RGBX:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 0, 1, 2,);
+			#endif
+			break;
+		case TJPF_RGBA:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 0, 1, 2, dst[3]=0xFF;);
+			#endif
+			break;
+		case TJPF_BGRX:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 2, 1, 0,);
+			#endif
+			break;
+		case TJPF_BGRA:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 2, 1, 0, dst[3]=0xFF;);  return;
+			#endif
+			break;
+		case TJPF_XRGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 1, 2, 3,);  return;
+			#endif
+			break;
+		case TJPF_ARGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 1, 2, 3, dst[0]=0xFF;);  return;
+			#endif
+			break;
+		case TJPF_XBGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 3, 2, 1,);  return;
+			#endif
+			break;
+		case TJPF_ABGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 3, 2, 1, dst[0]=0xFF;);  return;
+			#endif
+			break;
+	}
+}
+
+#endif
+
+
+/* General API functions */
+
+DLLEXPORT char* DLLCALL tjGetErrorStr(void)
+{
+	return errStr;
+}
+
+
+DLLEXPORT int DLLCALL tjDestroy(tjhandle handle)
+{
+	getinstance(handle);
+	if(setjmp(this->jerr.setjmp_buffer)) return -1;
+	if(this->init&COMPRESS) jpeg_destroy_compress(cinfo);
+	if(this->init&DECOMPRESS) jpeg_destroy_decompress(dinfo);
+	free(this);
+	return 0;
+}
+
+
+/* These are exposed mainly because Windows can't malloc() and free() across
+   DLL boundaries except when the CRT DLL is used, and we don't use the CRT DLL
+   with turbojpeg.dll for compatibility reasons.  However, these functions
+   can potentially be used for other purposes by different implementations. */
+
+DLLEXPORT void DLLCALL tjFree(unsigned char *buf)
+{
+	if(buf) free(buf);
+}
+
+
+DLLEXPORT unsigned char *DLLCALL tjAlloc(int bytes)
+{
+	return (unsigned char *)malloc(bytes);
+}
+
+
+/* Compressor  */
+
+static tjhandle _tjInitCompress(tjinstance *this)
+{
+	static unsigned char buffer[1];
+	unsigned char *buf=buffer;  unsigned long size=1;
+
+	/* This is also straight out of example.c */
+	this->cinfo.err=jpeg_std_error(&this->jerr.pub);
+	this->jerr.pub.error_exit=my_error_exit;
+	this->jerr.pub.output_message=my_output_message;
+	this->jerr.emit_message=this->jerr.pub.emit_message;
+	this->jerr.pub.emit_message=my_emit_message;
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		if(this) free(this);  return NULL;
+	}
+
+	jpeg_create_compress(&this->cinfo);
+	/* Make an initial call so it will create the destination manager */
+	jpeg_mem_dest_tj(&this->cinfo, &buf, &size, 0);
+
+	this->init|=COMPRESS;
+	return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle DLLCALL tjInitCompress(void)
+{
+	tjinstance *this=NULL;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitCompress(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	return _tjInitCompress(this);
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjBufSize(int width, int height,
+	int jpegSubsamp)
+{
+	unsigned long retval=0;  int mcuw, mcuh, chromasf;
+	if(width<1 || height<1 || jpegSubsamp<0 || jpegSubsamp>=NUMSUBOPT)
+		_throw("tjBufSize(): Invalid argument");
+
+	/* This allows for rare corner cases in which a JPEG image can actually be
+	   larger than the uncompressed input (we wouldn't mention it if it hadn't
+	   happened before.) */
+	mcuw=tjMCUWidth[jpegSubsamp];
+	mcuh=tjMCUHeight[jpegSubsamp];
+	chromasf=jpegSubsamp==TJSAMP_GRAY? 0: 4*64/(mcuw*mcuh);
+	retval=PAD(width, mcuw) * PAD(height, mcuh) * (2 + chromasf) + 2048;
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZE(int width, int height)
+{
+	unsigned long retval=0;
+	if(width<1 || height<1)
+		_throw("TJBUFSIZE(): Invalid argument");
+
+	/* This allows for rare corner cases in which a JPEG image can actually be
+	   larger than the uncompressed input (we wouldn't mention it if it hadn't
+	   happened before.) */
+	retval=PAD(width, 16) * PAD(height, 16) * 6 + 2048;
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV2(int width, int pad, int height,
+	int subsamp)
+{
+	int retval=0, nc, i;
+
+	if(subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjBufSizeYUV2(): Invalid argument");
+
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	for(i=0; i<nc; i++)
+	{
+		int pw=tjPlaneWidth(i, width, subsamp);
+		int stride=PAD(pw, pad);
+		int ph=tjPlaneHeight(i, height, subsamp);
+		if(pw<0 || ph<0) return -1;
+		else retval+=stride*ph;
+	}
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV(int width, int height,
+	int subsamp)
+{
+	return tjBufSizeYUV2(width, 4, height, subsamp);
+}
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZEYUV(int width, int height,
+	int subsamp)
+{
+	return tjBufSizeYUV(width, height, subsamp);
+}
+
+
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp)
+{
+	int pw, nc, retval=0;
+
+	if(width<1 || subsamp<0 || subsamp>=TJ_NUMSAMP)
+		_throw("tjPlaneWidth(): Invalid argument");
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	if(componentID<0 || componentID>=nc)
+		_throw("tjPlaneWidth(): Invalid argument");
+
+	pw=PAD(width, tjMCUWidth[subsamp]/8);
+	if(componentID==0)
+		retval=pw;
+	else
+		retval=pw*8/tjMCUWidth[subsamp];
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp)
+{
+	int ph, nc, retval=0;
+
+	if(height<1 || subsamp<0 || subsamp>=TJ_NUMSAMP)
+		_throw("tjPlaneHeight(): Invalid argument");
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	if(componentID<0 || componentID>=nc)
+		_throw("tjPlaneHeight(): Invalid argument");
+
+	ph=PAD(height, tjMCUHeight[subsamp]/8);
+	if(componentID==0)
+		retval=ph;
+	else
+		retval=ph*8/tjMCUHeight[subsamp];
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjPlaneSizeYUV(int componentID, int width,
+	int stride, int height, int subsamp)
+{
+	unsigned long retval=0;
+	int pw, ph;
+
+	if(width<1 || height<1 || subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjPlaneSizeYUV(): Invalid argument");
+
+	pw=tjPlaneWidth(componentID, width, subsamp);
+	ph=tjPlaneHeight(componentID, height, subsamp);
+	if(pw<0 || ph<0) return -1;
+
+	if(stride==0) stride=pw;
+	else stride=abs(stride);
+
+	retval=stride*(ph-1)+pw;
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT int DLLCALL tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelFormat, unsigned char **jpegBuf,
+	unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags)
+{
+	int i, retval=0, alloc=1;  JSAMPROW *row_pointer=NULL;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	#endif
+
+	getcinstance(handle)
+	if((this->init&COMPRESS)==0)
+		_throw("tjCompress2(): Instance has not been initialized for compression");
+
+	if(srcBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF || jpegBuf==NULL || jpegSize==NULL
+		|| jpegSubsamp<0 || jpegSubsamp>=NUMSUBOPT || jpegQual<0 || jpegQual>100)
+		_throw("tjCompress2(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK)
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*RGB_PIXELSIZE);
+		if(!rgbBuf) _throw("tjCompress2(): Memory allocation failure");
+		srcBuf=toRGB(srcBuf, width, pitch, height, pixelFormat, rgbBuf);
+		pitch=width*RGB_PIXELSIZE;
+	}
+	#endif
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(flags&TJFLAG_NOREALLOC)
+	{
+		alloc=0;  *jpegSize=tjBufSize(width, height, jpegSubsamp);
+	}
+	jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+	if(setCompDefaults(cinfo, pixelFormat, jpegSubsamp, jpegQual, flags)==-1)
+		return -1;
+
+	jpeg_start_compress(cinfo, TRUE);
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*height))==NULL)
+		_throw("tjCompress2(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=(JSAMPROW)&srcBuf[(height-i-1)*pitch];
+		else row_pointer[i]=(JSAMPROW)&srcBuf[i*pitch];
+	}
+	while(cinfo->next_scanline<cinfo->image_height)
+	{
+		jpeg_write_scanlines(cinfo, &row_pointer[cinfo->next_scanline],
+			cinfo->image_height-cinfo->next_scanline);
+	}
+	jpeg_finish_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjCompress(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelSize, unsigned char *jpegBuf,
+	unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags)
+{
+	int retval=0;  unsigned long size;
+	if(flags&TJ_YUV)
+	{
+		size=tjBufSizeYUV(width, height, jpegSubsamp);
+		retval=tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+			getPixelFormat(pixelSize, flags), jpegBuf, jpegSubsamp, flags);
+	}
+	else
+	{
+		retval=tjCompress2(handle, srcBuf, width, pitch, height,
+			getPixelFormat(pixelSize, flags), &jpegBuf, &size, jpegSubsamp, jpegQual,
+			flags|TJFLAG_NOREALLOC);
+	}
+	*jpegSize=size;
+	return retval;
+}
+
+
+DLLEXPORT int DLLCALL tjEncodeYUVPlanes(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pitch, int height,
+	int pixelFormat, unsigned char **dstPlanes, int *strides, int subsamp,
+	int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	JSAMPLE *_tmpbuf[MAX_COMPONENTS], *_tmpbuf2[MAX_COMPONENTS];
+	JSAMPROW *tmpbuf[MAX_COMPONENTS], *tmpbuf2[MAX_COMPONENTS];
+	JSAMPROW *outbuf[MAX_COMPONENTS];
+	int row, pw0, ph0, pw[MAX_COMPONENTS], ph[MAX_COMPONENTS];
+	JSAMPLE *ptr;
+	jpeg_component_info *compptr;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	#endif
+
+	getcinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  _tmpbuf[i]=NULL;
+		tmpbuf2[i]=NULL;  _tmpbuf2[i]=NULL;  outbuf[i]=NULL;
+	}
+
+	if((this->init&COMPRESS)==0)
+		_throw("tjEncodeYUVPlanes(): Instance has not been initialized for compression");
+
+	if(srcBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF || !dstPlanes || !dstPlanes[0] || subsamp<0
+		|| subsamp>=NUMSUBOPT)
+		_throw("tjEncodeYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!dstPlanes[1] || !dstPlanes[2]))
+		_throw("tjEncodeYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pixelFormat==TJPF_CMYK)
+		_throw("tjEncodeYUVPlanes(): Cannot generate YUV images from CMYK pixels");
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK)
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*RGB_PIXELSIZE);
+		if(!rgbBuf) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		srcBuf=toRGB(srcBuf, width, pitch, height, pixelFormat, rgbBuf);
+		pitch=width*RGB_PIXELSIZE;
+	}
+	#endif
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setCompDefaults(cinfo, pixelFormat, subsamp, -1, flags)==-1) return -1;
+
+	/* Execute only the parts of jpeg_start_compress() that we need.  If we
+	   were to call the whole jpeg_start_compress() function, then it would try
+	   to write the file headers, which could overflow the output buffer if the
+	   YUV image were very small. */
+	if(cinfo->global_state!=CSTATE_START)
+		_throw("tjEncodeYUVPlanes(): libjpeg API is in the wrong state");
+	(*cinfo->err->reset_error_mgr)((j_common_ptr)cinfo);
+	jinit_c_master_control(cinfo, FALSE);
+	jinit_color_converter(cinfo);
+	jinit_downsampler(cinfo);
+	(*cinfo->cconvert->start_pass)(cinfo);
+
+	pw0=PAD(width, cinfo->max_h_samp_factor);
+	ph0=PAD(height, cinfo->max_v_samp_factor);
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph0))==NULL)
+		_throw("tjEncodeYUVPlanes(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=(JSAMPROW)&srcBuf[(height-i-1)*pitch];
+		else row_pointer[i]=(JSAMPROW)&srcBuf[i*pitch];
+	}
+	if(height<ph0)
+		for(i=height; i<ph0; i++) row_pointer[i]=row_pointer[height-1];
+
+	for(i=0; i<cinfo->num_components; i++)
+	{
+		compptr=&cinfo->comp_info[i];
+		_tmpbuf[i]=(JSAMPLE *)malloc(
+			PAD((compptr->width_in_blocks*cinfo->max_h_samp_factor*DCTSIZE)
+				/compptr->h_samp_factor, 16) * cinfo->max_v_samp_factor + 16);
+		if(!_tmpbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*cinfo->max_v_samp_factor);
+		if(!tmpbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<cinfo->max_v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf[i], 16);
+			tmpbuf[i][row]=&_tmpbuf_aligned[
+				PAD((compptr->width_in_blocks*cinfo->max_h_samp_factor*DCTSIZE)
+					/compptr->h_samp_factor, 16) * row];
+		}
+		_tmpbuf2[i]=(JSAMPLE *)malloc(PAD(compptr->width_in_blocks*DCTSIZE, 16)
+			* compptr->v_samp_factor + 16);
+		if(!_tmpbuf2[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		tmpbuf2[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*compptr->v_samp_factor);
+		if(!tmpbuf2[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<compptr->v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf2_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf2[i], 16);
+			tmpbuf2[i][row]=&_tmpbuf2_aligned[
+				PAD(compptr->width_in_blocks*DCTSIZE, 16) * row];
+		}
+		pw[i]=pw0*compptr->h_samp_factor/cinfo->max_h_samp_factor;
+		ph[i]=ph0*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+		outbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]);
+		if(!outbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		ptr=dstPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			outbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+
+	for(row=0; row<ph0; row+=cinfo->max_v_samp_factor)
+	{
+		(*cinfo->cconvert->color_convert)(cinfo, &row_pointer[row], tmpbuf, 0,
+			cinfo->max_v_samp_factor);
+		(cinfo->downsample->downsample)(cinfo, tmpbuf, 0, tmpbuf2, 0);
+		for(i=0, compptr=cinfo->comp_info; i<cinfo->num_components; i++, compptr++)
+			jcopy_sample_rows(tmpbuf2[i], 0, outbuf[i],
+				row*compptr->v_samp_factor/cinfo->max_v_samp_factor,
+				compptr->v_samp_factor, pw[i]);
+	}
+	cinfo->next_scanline+=height;
+	jpeg_abort_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]!=NULL) free(tmpbuf[i]);
+		if(_tmpbuf[i]!=NULL) free(_tmpbuf[i]);
+		if(tmpbuf2[i]!=NULL) free(tmpbuf2[i]);
+		if(_tmpbuf2[i]!=NULL) free(_tmpbuf2[i]);
+		if(outbuf[i]!=NULL) free(outbuf[i]);
+	}
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV3(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pitch, int height,
+	int pixelFormat, unsigned char *dstBuf, int pad, int subsamp, int flags)
+{
+	unsigned char *dstPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(width<=0 || height<=0 || dstBuf==NULL || pad<0 || !isPow2(pad)
+		|| subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjEncodeYUV3(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	dstPlanes[0]=dstBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		dstPlanes[1]=dstPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		dstPlanes[1]=dstPlanes[0]+strides[0]*ph0;
+		dstPlanes[2]=dstPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjEncodeYUVPlanes(handle, srcBuf, width, pitch, height, pixelFormat,
+		dstPlanes, strides, subsamp, flags);
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV2(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelFormat, unsigned char *dstBuf,
+	int subsamp, int flags)
+{
+	return tjEncodeYUV3(handle, srcBuf, width, pitch, height, pixelFormat,
+		dstBuf, 4, subsamp, flags);
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelSize, unsigned char *dstBuf,
+	int subsamp, int flags)
+{
+	return tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+		getPixelFormat(pixelSize, flags), dstBuf, subsamp, flags);
+}
+
+
+DLLEXPORT int DLLCALL tjCompressFromYUVPlanes(tjhandle handle,
+	const unsigned char **srcPlanes, int width, const int *strides, int height,
+	int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual,
+	int flags)
+{
+	int i, row, retval=0, alloc=1;  JSAMPROW *inbuf[MAX_COMPONENTS];
+	int pw[MAX_COMPONENTS], ph[MAX_COMPONENTS], iw[MAX_COMPONENTS],
+		tmpbufsize=0, usetmpbuf=0, th[MAX_COMPONENTS];
+	JSAMPLE *_tmpbuf=NULL, *ptr;  JSAMPROW *tmpbuf[MAX_COMPONENTS];
+
+	getcinstance(handle)
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  inbuf[i]=NULL;
+	}
+
+	if((this->init&COMPRESS)==0)
+		_throw("tjCompressFromYUVPlanes(): Instance has not been initialized for compression");
+
+	if(!srcPlanes || !srcPlanes[0] || width<=0 || height<=0 || subsamp<0
+		|| subsamp>=NUMSUBOPT || jpegBuf==NULL || jpegSize==NULL || jpegQual<0
+		|| jpegQual>100)
+		_throw("tjCompressFromYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!srcPlanes[1] || !srcPlanes[2]))
+		_throw("tjCompressFromYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(flags&TJFLAG_NOREALLOC)
+	{
+		alloc=0;  *jpegSize=tjBufSize(width, height, subsamp);
+	}
+	jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+	if(setCompDefaults(cinfo, TJPF_RGB, subsamp, jpegQual, flags)==-1)
+		return -1;
+	cinfo->raw_data_in=TRUE;
+
+	jpeg_start_compress(cinfo, TRUE);
+	for(i=0; i<cinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&cinfo->comp_info[i];
+		int ih;
+		iw[i]=compptr->width_in_blocks*DCTSIZE;
+		ih=compptr->height_in_blocks*DCTSIZE;
+		pw[i]=PAD(cinfo->image_width, cinfo->max_h_samp_factor)
+			*compptr->h_samp_factor/cinfo->max_h_samp_factor;
+		ph[i]=PAD(cinfo->image_height, cinfo->max_v_samp_factor)
+			*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+		if(iw[i]!=pw[i] || ih!=ph[i]) usetmpbuf=1;
+		th[i]=compptr->v_samp_factor*DCTSIZE;
+		tmpbufsize+=iw[i]*th[i];
+		if((inbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]))==NULL)
+			_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+		ptr=(JSAMPLE *)srcPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			inbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+	if(usetmpbuf)
+	{
+		if((_tmpbuf=(JSAMPLE *)malloc(sizeof(JSAMPLE)*tmpbufsize))==NULL)
+			_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+		ptr=_tmpbuf;
+		for(i=0; i<cinfo->num_components; i++)
+		{
+			if((tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*th[i]))==NULL)
+				_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+			for(row=0; row<th[i]; row++)
+			{
+				tmpbuf[i][row]=ptr;
+				ptr+=iw[i];
+			}
+		}
+	}
+
+	for(row=0; row<(int)cinfo->image_height;
+		row+=cinfo->max_v_samp_factor*DCTSIZE)
+	{
+		JSAMPARRAY yuvptr[MAX_COMPONENTS];
+		int crow[MAX_COMPONENTS];
+		for(i=0; i<cinfo->num_components; i++)
+		{
+			jpeg_component_info *compptr=&cinfo->comp_info[i];
+			crow[i]=row*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+			if(usetmpbuf)
+			{
+				int j, k;
+				for(j=0; j<min(th[i], ph[i]-crow[i]); j++)
+				{
+					memcpy(tmpbuf[i][j], inbuf[i][crow[i]+j], pw[i]);
+					/* Duplicate last sample in row to fill out MCU */
+					for(k=pw[i]; k<iw[i]; k++) tmpbuf[i][j][k]=tmpbuf[i][j][pw[i]-1];
+				}
+				/* Duplicate last row to fill out MCU */
+				for(j=ph[i]-crow[i]; j<th[i]; j++)
+					memcpy(tmpbuf[i][j], tmpbuf[i][ph[i]-crow[i]-1], iw[i]);
+				yuvptr[i]=tmpbuf[i];
+			}
+			else
+				yuvptr[i]=&inbuf[i][crow[i]];
+		}
+		jpeg_write_raw_data(cinfo, yuvptr, cinfo->max_v_samp_factor*DCTSIZE);
+	}
+	jpeg_finish_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]) free(tmpbuf[i]);
+		if(inbuf[i]) free(inbuf[i]);
+	}
+	if(_tmpbuf) free(_tmpbuf);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjCompressFromYUV(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pad, int height, int subsamp,
+	unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags)
+{
+	const unsigned char *srcPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(srcBuf==NULL || width<=0 || pad<1 || height<=0 || subsamp<0
+		|| subsamp>=NUMSUBOPT)
+		_throw("tjCompressFromYUV(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	srcPlanes[0]=srcBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		srcPlanes[1]=srcPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		srcPlanes[1]=srcPlanes[0]+strides[0]*ph0;
+		srcPlanes[2]=srcPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjCompressFromYUVPlanes(handle, srcPlanes, width, strides, height,
+		subsamp, jpegBuf, jpegSize, jpegQual, flags);
+
+	bailout:
+	return retval;
+}
+
+
+/* Decompressor */
+
+static tjhandle _tjInitDecompress(tjinstance *this)
+{
+	static unsigned char buffer[1];
+
+	/* This is also straight out of example.c */
+	this->dinfo.err=jpeg_std_error(&this->jerr.pub);
+	this->jerr.pub.error_exit=my_error_exit;
+	this->jerr.pub.output_message=my_output_message;
+	this->jerr.emit_message=this->jerr.pub.emit_message;
+	this->jerr.pub.emit_message=my_emit_message;
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		if(this) free(this);  return NULL;
+	}
+
+	jpeg_create_decompress(&this->dinfo);
+	/* Make an initial call so it will create the source manager */
+	jpeg_mem_src_tj(&this->dinfo, buffer, 1);
+
+	this->init|=DECOMPRESS;
+	return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle DLLCALL tjInitDecompress(void)
+{
+	tjinstance *this;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitDecompress(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	return _tjInitDecompress(this);
+}
+
+
+DLLEXPORT int DLLCALL tjDecompressHeader3(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, int *width,
+	int *height, int *jpegSubsamp, int *jpegColorspace)
+{
+	int retval=0;
+
+	getdinstance(handle);
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompressHeader3(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || width==NULL || height==NULL
+		|| jpegSubsamp==NULL || jpegColorspace==NULL)
+		_throw("tjDecompressHeader3(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		return -1;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+
+	*width=dinfo->image_width;
+	*height=dinfo->image_height;
+	*jpegSubsamp=getSubsamp(dinfo);
+	switch(dinfo->jpeg_color_space)
+	{
+		case JCS_GRAYSCALE:  *jpegColorspace=TJCS_GRAY;  break;
+		case JCS_RGB:        *jpegColorspace=TJCS_RGB;  break;
+		case JCS_YCbCr:      *jpegColorspace=TJCS_YCbCr;  break;
+		case JCS_CMYK:       *jpegColorspace=TJCS_CMYK;  break;
+		case JCS_YCCK:       *jpegColorspace=TJCS_YCCK;  break;
+		default:             *jpegColorspace=-1;  break;
+	}
+
+	jpeg_abort_decompress(dinfo);
+
+	if(*jpegSubsamp<0)
+		_throw("tjDecompressHeader3(): Could not determine subsampling type for JPEG image");
+	if(*jpegColorspace<0)
+		_throw("tjDecompressHeader3(): Could not determine colorspace of JPEG image");
+	if(*width<1 || *height<1)
+		_throw("tjDecompressHeader3(): Invalid data returned in header");
+
+	bailout:
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressHeader2(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height,
+	int *jpegSubsamp)
+{
+	int jpegColorspace;
+	return tjDecompressHeader3(handle, jpegBuf, jpegSize, width, height,
+		jpegSubsamp, &jpegColorspace);
+}
+
+DLLEXPORT int DLLCALL tjDecompressHeader(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height)
+{
+	int jpegSubsamp;
+	return tjDecompressHeader2(handle, jpegBuf, jpegSize, width, height,
+		&jpegSubsamp);
+}
+
+
+DLLEXPORT tjscalingfactor* DLLCALL tjGetScalingFactors(int *numscalingfactors)
+{
+	if(numscalingfactors==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjGetScalingFactors(): Invalid argument");
+		return NULL;
+	}
+
+	*numscalingfactors=NUMSF;
+	return (tjscalingfactor *)sf;
+}
+
+
+DLLEXPORT int DLLCALL tjDecompress2(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int width, int pitch, int height, int pixelFormat, int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	int jpegwidth, jpegheight, scaledw, scaledh;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	unsigned char *_dstBuf=NULL;  int _pitch=0;
+	#endif
+
+	getdinstance(handle);
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompress2(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || dstBuf==NULL || width<0 || pitch<0
+		|| height<0 || pixelFormat<0 || pixelFormat>=TJ_NUMPF)
+		_throw("tjDecompress2(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+	if(setDecompDefaults(dinfo, pixelFormat, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+
+	if(flags&TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling=FALSE;
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompress2(): Could not scale down to desired image dimensions");
+	width=scaledw;  height=scaledh;
+	dinfo->scale_num=sf[i].num;
+	dinfo->scale_denom=sf[i].denom;
+
+	jpeg_start_decompress(dinfo);
+	if(pitch==0) pitch=dinfo->output_width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK &&
+		(RGB_RED!=tjRedOffset[pixelFormat] ||
+			RGB_GREEN!=tjGreenOffset[pixelFormat] ||
+			RGB_BLUE!=tjBlueOffset[pixelFormat] ||
+			RGB_PIXELSIZE!=tjPixelSize[pixelFormat]))
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*3);
+		if(!rgbBuf) _throw("tjDecompress2(): Memory allocation failure");
+		_pitch=pitch;  pitch=width*3;
+		_dstBuf=dstBuf;  dstBuf=rgbBuf;
+	}
+	#endif
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)
+		*dinfo->output_height))==NULL)
+		_throw("tjDecompress2(): Memory allocation failure");
+	for(i=0; i<(int)dinfo->output_height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=&dstBuf[(dinfo->output_height-i-1)*pitch];
+		else row_pointer[i]=&dstBuf[i*pitch];
+	}
+	while(dinfo->output_scanline<dinfo->output_height)
+	{
+		jpeg_read_scanlines(dinfo, &row_pointer[dinfo->output_scanline],
+			dinfo->output_height-dinfo->output_scanline);
+	}
+	jpeg_finish_decompress(dinfo);
+
+	#ifndef JCS_EXTENSIONS
+	fromRGB(rgbBuf, _dstBuf, width, _pitch, height, pixelFormat);
+	#endif
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompress(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, unsigned char *dstBuf, int width, int pitch,
+	int height, int pixelSize, int flags)
+{
+	if(flags&TJ_YUV)
+		return tjDecompressToYUV(handle, jpegBuf, jpegSize, dstBuf, flags);
+	else
+		return tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, width, pitch,
+			height, getPixelFormat(pixelSize, flags), flags);
+}
+
+
+static int setDecodeDefaults(struct jpeg_decompress_struct *dinfo,
+	int pixelFormat, int subsamp, int flags)
+{
+	int i;
+
+	dinfo->scale_num=dinfo->scale_denom=1;
+
+	if(subsamp==TJSAMP_GRAY)
+	{
+		dinfo->num_components=dinfo->comps_in_scan=1;
+		dinfo->jpeg_color_space=JCS_GRAYSCALE;
+	}
+	else
+	{
+		dinfo->num_components=dinfo->comps_in_scan=3;
+		dinfo->jpeg_color_space=JCS_YCbCr;
+	}
+
+	dinfo->comp_info=(jpeg_component_info *)
+		(*dinfo->mem->alloc_small)((j_common_ptr)dinfo, JPOOL_IMAGE,
+			dinfo->num_components*sizeof(jpeg_component_info));
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&dinfo->comp_info[i];
+		compptr->h_samp_factor=(i==0)? tjMCUWidth[subsamp]/8:1;
+		compptr->v_samp_factor=(i==0)? tjMCUHeight[subsamp]/8:1;
+		compptr->component_index=i;
+		compptr->component_id=i+1;
+		compptr->quant_tbl_no=compptr->dc_tbl_no=compptr->ac_tbl_no=
+			(i==0)? 0:1;
+		dinfo->cur_comp_info[i]=compptr;
+	}
+	dinfo->data_precision=8;
+	for(i=0; i<2; i++)
+	{
+		if(dinfo->quant_tbl_ptrs[i]==NULL)
+			dinfo->quant_tbl_ptrs[i]=jpeg_alloc_quant_table((j_common_ptr)dinfo);
+	}
+
+	return 0;
+}
+
+
+int my_read_markers(j_decompress_ptr dinfo)
+{
+	return JPEG_REACHED_SOS;
+}
+
+void my_reset_marker_reader(j_decompress_ptr dinfo)
+{
+}
+
+DLLEXPORT int DLLCALL tjDecodeYUVPlanes(tjhandle handle,
+	const unsigned char **srcPlanes, const int *strides, int subsamp,
+	unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat,
+	int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	JSAMPLE *_tmpbuf[MAX_COMPONENTS];
+	JSAMPROW *tmpbuf[MAX_COMPONENTS], *inbuf[MAX_COMPONENTS];
+	int row, pw0, ph0, pw[MAX_COMPONENTS], ph[MAX_COMPONENTS];
+	JSAMPLE *ptr;
+	jpeg_component_info *compptr;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	unsigned char *_dstBuf=NULL;  int _pitch=0;
+	#endif
+	int (*old_read_markers)(j_decompress_ptr);
+	void (*old_reset_marker_reader)(j_decompress_ptr);
+
+	getdinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  _tmpbuf[i]=NULL;  inbuf[i]=NULL;
+	}
+
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecodeYUVPlanes(): Instance has not been initialized for decompression");
+
+	if(!srcPlanes || !srcPlanes[0] || subsamp<0 || subsamp>=NUMSUBOPT
+		|| dstBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF)
+		_throw("tjDecodeYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!srcPlanes[1] || !srcPlanes[2]))
+		_throw("tjDecodeYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pixelFormat==TJPF_CMYK)
+		_throw("tjDecodeYUVPlanes(): Cannot decode YUV images into CMYK pixels.");
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+	dinfo->image_width=width;
+	dinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setDecodeDefaults(dinfo, pixelFormat, subsamp, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+	old_read_markers=dinfo->marker->read_markers;
+	dinfo->marker->read_markers=my_read_markers;
+	old_reset_marker_reader=dinfo->marker->reset_marker_reader;
+	dinfo->marker->reset_marker_reader=my_reset_marker_reader;
+	jpeg_read_header(dinfo, TRUE);
+	dinfo->marker->read_markers=old_read_markers;
+	dinfo->marker->reset_marker_reader=old_reset_marker_reader;
+
+	if(setDecompDefaults(dinfo, pixelFormat, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+	dinfo->do_fancy_upsampling=FALSE;
+	dinfo->Se=DCTSIZE2-1;
+	jinit_master_decompress(dinfo);
+	(*dinfo->upsample->start_pass)(dinfo);
+
+	pw0=PAD(width, dinfo->max_h_samp_factor);
+	ph0=PAD(height, dinfo->max_v_samp_factor);
+
+	if(pitch==0) pitch=dinfo->output_width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK &&
+		(RGB_RED!=tjRedOffset[pixelFormat] ||
+			RGB_GREEN!=tjGreenOffset[pixelFormat] ||
+			RGB_BLUE!=tjBlueOffset[pixelFormat] ||
+			RGB_PIXELSIZE!=tjPixelSize[pixelFormat]))
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*3);
+		if(!rgbBuf) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		_pitch=pitch;  pitch=width*3;
+		_dstBuf=dstBuf;  dstBuf=rgbBuf;
+	}
+	#endif
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph0))==NULL)
+		_throw("tjDecodeYUVPlanes(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP) row_pointer[i]=&dstBuf[(height-i-1)*pitch];
+		else row_pointer[i]=&dstBuf[i*pitch];
+	}
+	if(height<ph0)
+		for(i=height; i<ph0; i++) row_pointer[i]=row_pointer[height-1];
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		compptr=&dinfo->comp_info[i];
+		_tmpbuf[i]=(JSAMPLE *)malloc(PAD(compptr->width_in_blocks*DCTSIZE, 16)
+			* compptr->v_samp_factor + 16);
+		if(!_tmpbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*compptr->v_samp_factor);
+		if(!tmpbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<compptr->v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf[i], 16);
+			tmpbuf[i][row]=&_tmpbuf_aligned[
+				PAD(compptr->width_in_blocks*DCTSIZE, 16) * row];
+		}
+		pw[i]=pw0*compptr->h_samp_factor/dinfo->max_h_samp_factor;
+		ph[i]=ph0*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+		inbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]);
+		if(!inbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		ptr=(JSAMPLE *)srcPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			inbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+
+	for(row=0; row<ph0; row+=dinfo->max_v_samp_factor)
+	{
+		JDIMENSION inrow=0, outrow=0;
+		for(i=0, compptr=dinfo->comp_info; i<dinfo->num_components; i++, compptr++)
+			jcopy_sample_rows(inbuf[i],
+				row*compptr->v_samp_factor/dinfo->max_v_samp_factor, tmpbuf[i], 0,
+				compptr->v_samp_factor, pw[i]);
+		(dinfo->upsample->upsample)(dinfo, tmpbuf, &inrow,
+			dinfo->max_v_samp_factor, &row_pointer[row], &outrow,
+			dinfo->max_v_samp_factor);
+	}
+	jpeg_abort_decompress(dinfo);
+
+	#ifndef JCS_EXTENSIONS
+	fromRGB(rgbBuf, _dstBuf, width, _pitch, height, pixelFormat);
+	#endif
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]!=NULL) free(tmpbuf[i]);
+		if(_tmpbuf[i]!=NULL) free(_tmpbuf[i]);
+		if(inbuf[i]!=NULL) free(inbuf[i]);
+	}
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+	int pad, int subsamp, unsigned char *dstBuf, int width, int pitch,
+	int height, int pixelFormat, int flags)
+{
+	const unsigned char *srcPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(srcBuf==NULL || pad<0 || !isPow2(pad) || subsamp<0 || subsamp>=NUMSUBOPT
+		|| width<=0 || height<=0)
+		_throw("tjDecodeYUV(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	srcPlanes[0]=srcBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		srcPlanes[1]=srcPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		srcPlanes[1]=srcPlanes[0]+strides[0]*ph0;
+		srcPlanes[2]=srcPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjDecodeYUVPlanes(handle, srcPlanes, strides, subsamp, dstBuf, width,
+		pitch, height, pixelFormat, flags);
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUVPlanes(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize,
+	unsigned char **dstPlanes, int width, int *strides, int height, int flags)
+{
+	int i, sfi, row, retval=0;  JSAMPROW *outbuf[MAX_COMPONENTS];
+	int jpegwidth, jpegheight, jpegSubsamp, scaledw, scaledh;
+	int pw[MAX_COMPONENTS], ph[MAX_COMPONENTS], iw[MAX_COMPONENTS],
+		tmpbufsize=0, usetmpbuf=0, th[MAX_COMPONENTS];
+	JSAMPLE *_tmpbuf=NULL, *ptr;  JSAMPROW *tmpbuf[MAX_COMPONENTS];
+	int dctsize;
+
+	getdinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  outbuf[i]=NULL;
+	}
+
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompressToYUVPlanes(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || !dstPlanes || !dstPlanes[0] || width<0
+		|| height<0)
+		_throw("tjDecompressToYUVPlanes(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(!this->headerRead)
+	{
+		jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+		jpeg_read_header(dinfo, TRUE);
+	}
+	this->headerRead=0;
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjDecompressToYUVPlanes(): Could not determine subsampling type for JPEG image");
+
+	if(jpegSubsamp!=TJSAMP_GRAY && (!dstPlanes[1] || !dstPlanes[2]))
+		_throw("tjDecompressToYUVPlanes(): Invalid argument");
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompressToYUVPlanes(): Could not scale down to desired image dimensions");
+	if(dinfo->num_components>3)
+		_throw("tjDecompressToYUVPlanes(): JPEG image must have 3 or fewer components");
+
+	width=scaledw;  height=scaledh;
+	dinfo->scale_num=sf[i].num;
+	dinfo->scale_denom=sf[i].denom;
+	sfi=i;
+	jpeg_calc_output_dimensions(dinfo);
+
+	dctsize=DCTSIZE*sf[sfi].num/sf[sfi].denom;
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&dinfo->comp_info[i];
+		int ih;
+		iw[i]=compptr->width_in_blocks*dctsize;
+		ih=compptr->height_in_blocks*dctsize;
+		pw[i]=PAD(dinfo->output_width, dinfo->max_h_samp_factor)
+			*compptr->h_samp_factor/dinfo->max_h_samp_factor;
+		ph[i]=PAD(dinfo->output_height, dinfo->max_v_samp_factor)
+			*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+		if(iw[i]!=pw[i] || ih!=ph[i]) usetmpbuf=1;
+		th[i]=compptr->v_samp_factor*dctsize;
+		tmpbufsize+=iw[i]*th[i];
+		if((outbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]))==NULL)
+			_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+		ptr=dstPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			outbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+	if(usetmpbuf)
+	{
+		if((_tmpbuf=(JSAMPLE *)malloc(sizeof(JSAMPLE)*tmpbufsize))==NULL)
+			_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+		ptr=_tmpbuf;
+		for(i=0; i<dinfo->num_components; i++)
+		{
+			if((tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*th[i]))==NULL)
+				_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+			for(row=0; row<th[i]; row++)
+			{
+				tmpbuf[i][row]=ptr;
+				ptr+=iw[i];
+			}
+		}
+	}
+
+	if(flags&TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling=FALSE;
+	if(flags&TJFLAG_FASTDCT) dinfo->dct_method=JDCT_FASTEST;
+	dinfo->raw_data_out=TRUE;
+
+	jpeg_start_decompress(dinfo);
+	for(row=0; row<(int)dinfo->output_height;
+		row+=dinfo->max_v_samp_factor*dinfo->_min_DCT_scaled_size)
+	{
+		JSAMPARRAY yuvptr[MAX_COMPONENTS];
+		int crow[MAX_COMPONENTS];
+		for(i=0; i<dinfo->num_components; i++)
+		{
+			jpeg_component_info *compptr=&dinfo->comp_info[i];
+			if(jpegSubsamp==TJ_420)
+			{
+				/* When 4:2:0 subsampling is used with IDCT scaling, libjpeg will try
+				   to be clever and use the IDCT to perform upsampling on the U and V
+				   planes.  For instance, if the output image is to be scaled by 1/2
+				   relative to the JPEG image, then the scaling factor and upsampling
+				   effectively cancel each other, so a normal 8x8 IDCT can be used.
+				   However, this is not desirable when using the decompress-to-YUV
+				   functionality in TurboJPEG, since we want to output the U and V
+				   planes in their subsampled form.  Thus, we have to override some
+				   internal libjpeg parameters to force it to use the "scaled" IDCT
+				   functions on the U and V planes. */
+				compptr->_DCT_scaled_size=dctsize;
+				compptr->MCU_sample_width=tjMCUWidth[jpegSubsamp]*
+					sf[sfi].num/sf[sfi].denom*
+					compptr->v_samp_factor/dinfo->max_v_samp_factor;
+				dinfo->idct->inverse_DCT[i] = dinfo->idct->inverse_DCT[0];
+			}
+			crow[i]=row*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+			if(usetmpbuf) yuvptr[i]=tmpbuf[i];
+			else yuvptr[i]=&outbuf[i][crow[i]];
+		}
+		jpeg_read_raw_data(dinfo, yuvptr,
+			dinfo->max_v_samp_factor*dinfo->_min_DCT_scaled_size);
+		if(usetmpbuf)
+		{
+			int j;
+			for(i=0; i<dinfo->num_components; i++)
+			{
+				for(j=0; j<min(th[i], ph[i]-crow[i]); j++)
+				{
+					memcpy(outbuf[i][crow[i]+j], tmpbuf[i][j], pw[i]);
+				}
+			}
+		}
+	}
+	jpeg_finish_decompress(dinfo);
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]) free(tmpbuf[i]);
+		if(outbuf[i]) free(outbuf[i]);
+	}
+	if(_tmpbuf) free(_tmpbuf);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUV2(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int width, int pad, int height, int flags)
+{
+	unsigned char *dstPlanes[3];
+	int pw0, ph0, strides[3], retval=-1, jpegSubsamp=-1;
+	int i, jpegwidth, jpegheight, scaledw, scaledh;
+
+	getdinstance(handle);
+
+	if(jpegBuf==NULL || jpegSize<=0 || dstBuf==NULL || width<0 || pad<1
+		|| !isPow2(pad) || height<0)
+		_throw("tjDecompressToYUV2(): Invalid argument");
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjDecompressToYUV2(): Could not determine subsampling type for JPEG image");
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompressToYUV2(): Could not scale down to desired image dimensions");
+
+	pw0=tjPlaneWidth(0, width, jpegSubsamp);
+	ph0=tjPlaneHeight(0, height, jpegSubsamp);
+	dstPlanes[0]=dstBuf;
+	strides[0]=PAD(pw0, pad);
+	if(jpegSubsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		dstPlanes[1]=dstPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, jpegSubsamp);
+		int ph1=tjPlaneHeight(1, height, jpegSubsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		dstPlanes[1]=dstPlanes[0]+strides[0]*ph0;
+		dstPlanes[2]=dstPlanes[1]+strides[1]*ph1;
+	}
+
+	this->headerRead=1;
+	return tjDecompressToYUVPlanes(handle, jpegBuf, jpegSize, dstPlanes, width,
+		strides, height, flags);
+
+	bailout:
+	return retval;
+
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUV(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int flags)
+{
+	return tjDecompressToYUV2(handle, jpegBuf, jpegSize, dstBuf, 0, 4, 0, flags);
+}
+
+
+/* Transformer */
+
+DLLEXPORT tjhandle DLLCALL tjInitTransform(void)
+{
+	tjinstance *this=NULL;  tjhandle handle=NULL;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitTransform(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	handle=_tjInitCompress(this);
+	if(!handle) return NULL;
+	handle=_tjInitDecompress(this);
+	return handle;
+}
+
+
+DLLEXPORT int DLLCALL tjTransform(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, int n,
+	unsigned char **dstBufs, unsigned long *dstSizes, tjtransform *t, int flags)
+{
+	jpeg_transform_info *xinfo=NULL;
+	jvirt_barray_ptr *srccoefs, *dstcoefs;
+	int retval=0, i, jpegSubsamp;
+
+	getinstance(handle);
+	if((this->init&COMPRESS)==0 || (this->init&DECOMPRESS)==0)
+		_throw("tjTransform(): Instance has not been initialized for transformation");
+
+	if(jpegBuf==NULL || jpegSize<=0 || n<1 || dstBufs==NULL || dstSizes==NULL
+		|| t==NULL || flags<0)
+		_throw("tjTransform(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+
+	if((xinfo=(jpeg_transform_info *)malloc(sizeof(jpeg_transform_info)*n))
+		==NULL)
+		_throw("tjTransform(): Memory allocation failure");
+	MEMZERO(xinfo, sizeof(jpeg_transform_info)*n);
+
+	for(i=0; i<n; i++)
+	{
+		xinfo[i].transform=xformtypes[t[i].op];
+		xinfo[i].perfect=(t[i].options&TJXOPT_PERFECT)? 1:0;
+		xinfo[i].trim=(t[i].options&TJXOPT_TRIM)? 1:0;
+		xinfo[i].force_grayscale=(t[i].options&TJXOPT_GRAY)? 1:0;
+		xinfo[i].crop=(t[i].options&TJXOPT_CROP)? 1:0;
+		if(n!=1 && t[i].op==TJXOP_HFLIP) xinfo[i].slow_hflip=1;
+		else xinfo[i].slow_hflip=0;
+
+		if(xinfo[i].crop)
+		{
+			xinfo[i].crop_xoffset=t[i].r.x;  xinfo[i].crop_xoffset_set=JCROP_POS;
+			xinfo[i].crop_yoffset=t[i].r.y;  xinfo[i].crop_yoffset_set=JCROP_POS;
+			if(t[i].r.w!=0)
+			{
+				xinfo[i].crop_width=t[i].r.w;  xinfo[i].crop_width_set=JCROP_POS;
+			}
+			else xinfo[i].crop_width=JCROP_UNSET;
+			if(t[i].r.h!=0)
+			{
+				xinfo[i].crop_height=t[i].r.h;  xinfo[i].crop_height_set=JCROP_POS;
+			}
+			else xinfo[i].crop_height=JCROP_UNSET;
+		}
+	}
+
+	jcopy_markers_setup(dinfo, JCOPYOPT_ALL);
+	jpeg_read_header(dinfo, TRUE);
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjTransform(): Could not determine subsampling type for JPEG image");
+
+	for(i=0; i<n; i++)
+	{
+		if(!jtransform_request_workspace(dinfo, &xinfo[i]))
+			_throw("tjTransform(): Transform is not perfect");
+
+		if(xinfo[i].crop)
+		{
+			if((t[i].r.x%xinfo[i].iMCU_sample_width)!=0
+				|| (t[i].r.y%xinfo[i].iMCU_sample_height)!=0)
+			{
+				snprintf(errStr, JMSG_LENGTH_MAX,
+					"To crop this JPEG image, x must be a multiple of %d\n"
+					"and y must be a multiple of %d.\n",
+					xinfo[i].iMCU_sample_width, xinfo[i].iMCU_sample_height);
+				retval=-1;  goto bailout;
+			}
+		}
+	}
+
+	srccoefs=jpeg_read_coefficients(dinfo);
+
+	for(i=0; i<n; i++)
+	{
+		int w, h, alloc=1;
+		if(!xinfo[i].crop)
+		{
+			w=dinfo->image_width;  h=dinfo->image_height;
+		}
+		else
+		{
+			w=xinfo[i].crop_width;  h=xinfo[i].crop_height;
+		}
+		if(flags&TJFLAG_NOREALLOC)
+		{
+			alloc=0;  dstSizes[i]=tjBufSize(w, h, jpegSubsamp);
+		}
+		if(!(t[i].options&TJXOPT_NOOUTPUT))
+			jpeg_mem_dest_tj(cinfo, &dstBufs[i], &dstSizes[i], alloc);
+		jpeg_copy_critical_parameters(dinfo, cinfo);
+		dstcoefs=jtransform_adjust_parameters(dinfo, cinfo, srccoefs,
+			&xinfo[i]);
+		if(!(t[i].options&TJXOPT_NOOUTPUT))
+		{
+			jpeg_write_coefficients(cinfo, dstcoefs);
+			jcopy_markers_execute(dinfo, cinfo, JCOPYOPT_ALL);
+		}
+		else jinit_c_master_control(cinfo, TRUE);
+		jtransform_execute_transformation(dinfo, cinfo, srccoefs,
+			&xinfo[i]);
+		if(t[i].customFilter)
+		{
+			int ci, y;  JDIMENSION by;
+			for(ci=0; ci<cinfo->num_components; ci++)
+			{
+				jpeg_component_info *compptr=&cinfo->comp_info[ci];
+				tjregion arrayRegion={0, 0, compptr->width_in_blocks*DCTSIZE,
+					DCTSIZE};
+				tjregion planeRegion={0, 0, compptr->width_in_blocks*DCTSIZE,
+					compptr->height_in_blocks*DCTSIZE};
+				for(by=0; by<compptr->height_in_blocks; by+=compptr->v_samp_factor)
+				{
+					JBLOCKARRAY barray=(dinfo->mem->access_virt_barray)
+						((j_common_ptr)dinfo, dstcoefs[ci], by, compptr->v_samp_factor,
+						TRUE);
+					for(y=0; y<compptr->v_samp_factor; y++)
+					{
+						if(t[i].customFilter(barray[y][0], arrayRegion, planeRegion,
+							ci, i, &t[i])==-1)
+							_throw("tjTransform(): Error in custom filter");
+						arrayRegion.y+=DCTSIZE;
+					}
+				}
+			}
+		}
+		if(!(t[i].options&TJXOPT_NOOUTPUT)) jpeg_finish_compress(cinfo);
+	}
+
+	jpeg_finish_decompress(dinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	if(xinfo) free(xinfo);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg.h glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1542 @@
+/*
+ * Copyright (C)2009-2015 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __TURBOJPEG_H__
+#define __TURBOJPEG_H__
+
+#if defined(_WIN32) && defined(DLLDEFINE)
+#define DLLEXPORT __declspec(dllexport)
+#else
+#define DLLEXPORT
+#endif
+#define DLLCALL
+
+
+/**
+ * @addtogroup TurboJPEG
+ * TurboJPEG API.  This API provides an interface for generating, decoding, and
+ * transforming planar YUV and JPEG images in memory.
+ *
+ * @anchor YUVnotes
+ * YUV Image Format Notes
+ * ----------------------
+ * Technically, the JPEG format uses the YCbCr colorspace (which is technically
+ * not a colorspace but a color transform), but per the convention of the
+ * digital video community, the TurboJPEG API uses "YUV" to refer to an image
+ * format consisting of Y, Cb, and Cr image planes.
+ *
+ * Each plane is simply a 2D array of bytes, each byte representing the value
+ * of one of the components (Y, Cb, or Cr) at a particular location in the
+ * image.  The width and height of each plane are determined by the image
+ * width, height, and level of chrominance subsampling.   The luminance plane
+ * width is the image width padded to the nearest multiple of the horizontal
+ * subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of
+ * 4:1:1, 1 in the case of 4:4:4 or grayscale.)  Similarly, the luminance plane
+ * height is the image height padded to the nearest multiple of the vertical
+ * subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4
+ * or grayscale.)  This is irrespective of any additional padding that may be
+ * specified as an argument to the various YUV functions.  The chrominance
+ * plane width is equal to the luminance plane width divided by the horizontal
+ * subsampling factor, and the chrominance plane height is equal to the
+ * luminance plane height divided by the vertical subsampling factor.
+ *
+ * For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is
+ * used, then the luminance plane would be 36 x 35 bytes, and each of the
+ * chrominance planes would be 18 x 35 bytes.  If you specify a line padding of
+ * 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and
+ * each of the chrominance planes would be 20 x 35 bytes.
+ *
+ * @{
+ */
+
+
+/**
+ * The number of chrominance subsampling options
+ */
+#define TJ_NUMSAMP 6
+
+/**
+ * Chrominance subsampling options.
+ * When pixels are converted from RGB to YCbCr (see #TJCS_YCbCr) or from CMYK
+ * to YCCK (see #TJCS_YCCK) as part of the JPEG compression process, some of
+ * the Cb and Cr (chrominance) components can be discarded or averaged together
+ * to produce a smaller image with little perceptible loss of image clarity
+ * (the human eye is more sensitive to small changes in brightness than to
+ * small changes in color.)  This is called "chrominance subsampling".
+ */
+enum TJSAMP
+{
+  /**
+   * 4:4:4 chrominance subsampling (no chrominance subsampling).  The JPEG or
+   * YUV image will contain one chrominance component for every pixel in the
+   * source image.
+   */
+  TJSAMP_444=0,
+  /**
+   * 4:2:2 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x1 block of pixels in the source image.
+   */
+  TJSAMP_422,
+  /**
+   * 4:2:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x2 block of pixels in the source image.
+   */
+  TJSAMP_420,
+  /**
+   * Grayscale.  The JPEG or YUV image will contain no chrominance components.
+   */
+  TJSAMP_GRAY,
+  /**
+   * 4:4:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 1x2 block of pixels in the source image.
+   *
+   * @note 4:4:0 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_440,
+  /**
+   * 4:1:1 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 4x1 block of pixels in the source image.
+   * JPEG images compressed with 4:1:1 subsampling will be almost exactly the
+   * same size as those compressed with 4:2:0 subsampling, and in the
+   * aggregate, both subsampling methods produce approximately the same
+   * perceptual quality.  However, 4:1:1 is better able to reproduce sharp
+   * horizontal features.
+   *
+   * @note 4:1:1 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_411
+};
+
+/**
+ * MCU block width (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUWidth[TJ_NUMSAMP]  = {8, 16, 16, 8, 8, 32};
+
+/**
+ * MCU block height (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUHeight[TJ_NUMSAMP] = {8, 8, 16, 8, 16, 8};
+
+
+/**
+ * The number of pixel formats
+ */
+#define TJ_NUMPF 12
+
+/**
+ * Pixel formats
+ */
+enum TJPF
+{
+  /**
+   * RGB pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_RGB=0,
+  /**
+   * BGR pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_BGR,
+  /**
+   * RGBX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_RGBX,
+  /**
+   * BGRX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_BGRX,
+  /**
+   * XBGR pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XBGR,
+  /**
+   * XRGB pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XRGB,
+  /**
+   * Grayscale pixel format.  Each 1-byte pixel represents a luminance
+   * (brightness) level from 0 to 255.
+   */
+  TJPF_GRAY,
+  /**
+   * RGBA pixel format.  This is the same as @ref TJPF_RGBX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_RGBA,
+  /**
+   * BGRA pixel format.  This is the same as @ref TJPF_BGRX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_BGRA,
+  /**
+   * ABGR pixel format.  This is the same as @ref TJPF_XBGR, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ABGR,
+  /**
+   * ARGB pixel format.  This is the same as @ref TJPF_XRGB, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ARGB,
+  /**
+   * CMYK pixel format.  Unlike RGB, which is an additive color model used
+   * primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive
+   * color model used primarily for printing.  In the CMYK color model, the
+   * value of each color component typically corresponds to an amount of cyan,
+   * magenta, yellow, or black ink that is applied to a white background.  In
+   * order to convert between CMYK and RGB, it is necessary to use a color
+   * management system (CMS.)  A CMS will attempt to map colors within the
+   * printer's gamut to perceptually similar colors in the display's gamut and
+   * vice versa, but the mapping is typically not 1:1 or reversible, nor can it
+   * be defined with a simple formula.  Thus, such a conversion is out of scope
+   * for a codec library.  However, the TurboJPEG API allows for compressing
+   * CMYK pixels into a YCCK JPEG image (see #TJCS_YCCK) and decompressing YCCK
+   * JPEG images into CMYK pixels.
+   */
+  TJPF_CMYK
+};
+
+
+/**
+ * Red offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the red component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel[]</tt>,
+ * then the red component will be <tt>pixel[tjRedOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjRedOffset[TJ_NUMPF] = {0, 2, 0, 2, 3, 1, 0, 0, 2, 3, 1, -1};
+/**
+ * Green offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the green component is offset from the start of the pixel.
+ * For instance, if a pixel of format TJ_BGRX is stored in
+ * <tt>char pixel[]</tt>, then the green component will be
+ * <tt>pixel[tjGreenOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjGreenOffset[TJ_NUMPF] = {1, 1, 1, 1, 2, 2, 0, 1, 1, 2, 2, -1};
+/**
+ * Blue offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the Blue component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel[]</tt>,
+ * then the blue component will be <tt>pixel[tjBlueOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjBlueOffset[TJ_NUMPF] = {2, 0, 2, 0, 1, 3, 0, 2, 0, 1, 3, -1};
+
+/**
+ * Pixel size (in bytes) for a given pixel format.
+ */
+static const int tjPixelSize[TJ_NUMPF] = {3, 3, 4, 4, 4, 4, 1, 4, 4, 4, 4, 4};
+
+
+/**
+ * The number of JPEG colorspaces
+ */
+#define TJ_NUMCS 5
+
+/**
+ * JPEG colorspaces
+ */
+enum TJCS
+{
+  /**
+   * RGB colorspace.  When compressing the JPEG image, the R, G, and B
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  RGB JPEG images can be
+   * decompressed to any of the extended RGB pixel formats or grayscale, but
+   * they cannot be decompressed to YUV images.
+   */
+  TJCS_RGB=0,
+  /**
+   * YCbCr colorspace.  YCbCr is not an absolute colorspace but rather a
+   * mathematical transformation of RGB designed solely for storage and
+   * transmission.  YCbCr images must be converted to RGB before they can
+   * actually be displayed.  In the YCbCr colorspace, the Y (luminance)
+   * component represents the black & white portion of the original image, and
+   * the Cb and Cr (chrominance) components represent the color portion of the
+   * original image.  Originally, the analog equivalent of this transformation
+   * allowed the same signal to drive both black & white and color televisions,
+   * but JPEG images use YCbCr primarily because it allows the color data to be
+   * optionally subsampled for the purposes of reducing bandwidth or disk
+   * space.  YCbCr is the most common JPEG colorspace, and YCbCr JPEG images
+   * can be compressed from and decompressed to any of the extended RGB pixel
+   * formats or grayscale, or they can be decompressed to YUV planar images.
+   */
+  TJCS_YCbCr,
+  /**
+   * Grayscale colorspace.  The JPEG image retains only the luminance data (Y
+   * component), and any color data from the source image is discarded.
+   * Grayscale JPEG images can be compressed from and decompressed to any of
+   * the extended RGB pixel formats or grayscale, or they can be decompressed
+   * to YUV planar images.
+   */
+  TJCS_GRAY,
+  /**
+   * CMYK colorspace.  When compressing the JPEG image, the C, M, Y, and K
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  CMYK JPEG images can
+   * only be decompressed to CMYK pixels.
+   */
+  TJCS_CMYK,
+  /**
+   * YCCK colorspace.  YCCK (AKA "YCbCrK") is not an absolute colorspace but
+   * rather a mathematical transformation of CMYK designed solely for storage
+   * and transmission.  It is to CMYK as YCbCr is to RGB.  CMYK pixels can be
+   * reversibly transformed into YCCK, and as with YCbCr, the chrominance
+   * components in the YCCK pixels can be subsampled without incurring major
+   * perceptual loss.  YCCK JPEG images can only be compressed from and
+   * decompressed to CMYK pixels.
+   */
+  TJCS_YCCK
+};
+
+
+/**
+ * The uncompressed source/destination image is stored in bottom-up (Windows,
+ * OpenGL) order, not top-down (X11) order.
+ */
+#define TJFLAG_BOTTOMUP        2
+/**
+ * When decompressing an image that was compressed using chrominance
+ * subsampling, use the fastest chrominance upsampling algorithm available in
+ * the underlying codec.  The default is to use smooth upsampling, which
+ * creates a smooth transition between neighboring chrominance components in
+ * order to reduce upsampling artifacts in the decompressed image.
+ */
+#define TJFLAG_FASTUPSAMPLE  256
+/**
+ * Disable buffer (re)allocation.  If passed to #tjCompress2() or
+ * #tjTransform(), this flag will cause those functions to generate an error if
+ * the JPEG image buffer is invalid or too small rather than attempting to
+ * allocate or reallocate that buffer.  This reproduces the behavior of earlier
+ * versions of TurboJPEG.
+ */
+#define TJFLAG_NOREALLOC     1024
+/**
+ * Use the fastest DCT/IDCT algorithm available in the underlying codec.  The
+ * default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg[-turbo] uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_FASTDCT       2048
+/**
+ * Use the most accurate DCT/IDCT algorithm available in the underlying codec.
+ * The default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg[-turbo] uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_ACCURATEDCT   4096
+
+
+/**
+ * The number of transform operations
+ */
+#define TJ_NUMXOP 8
+
+/**
+ * Transform operations for #tjTransform()
+ */
+enum TJXOP
+{
+  /**
+   * Do not transform the position of the image pixels
+   */
+  TJXOP_NONE=0,
+  /**
+   * Flip (mirror) image horizontally.  This transform is imperfect if there
+   * are any partial MCU blocks on the right edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_HFLIP,
+  /**
+   * Flip (mirror) image vertically.  This transform is imperfect if there are
+   * any partial MCU blocks on the bottom edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_VFLIP,
+  /**
+   * Transpose image (flip/mirror along upper left to lower right axis.)  This
+   * transform is always perfect.
+   */
+  TJXOP_TRANSPOSE,
+  /**
+   * Transverse transpose image (flip/mirror along upper right to lower left
+   * axis.)  This transform is imperfect if there are any partial MCU blocks in
+   * the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_TRANSVERSE,
+  /**
+   * Rotate image clockwise by 90 degrees.  This transform is imperfect if
+   * there are any partial MCU blocks on the bottom edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT90,
+  /**
+   * Rotate image 180 degrees.  This transform is imperfect if there are any
+   * partial MCU blocks in the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT180,
+  /**
+   * Rotate image counter-clockwise by 90 degrees.  This transform is imperfect
+   * if there are any partial MCU blocks on the right edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT270
+};
+
+
+/**
+ * This option will cause #tjTransform() to return an error if the transform is
+ * not perfect.  Lossless transforms operate on MCU blocks, whose size depends
+ * on the level of chrominance subsampling used (see #tjMCUWidth
+ * and #tjMCUHeight.)  If the image's width or height is not evenly divisible
+ * by the MCU block size, then there will be partial MCU blocks on the right
+ * and/or bottom edges.  It is not possible to move these partial MCU blocks to
+ * the top or left of the image, so any transform that would require that is
+ * "imperfect."  If this option is not specified, then any partial MCU blocks
+ * that cannot be transformed will be left in place, which will create
+ * odd-looking strips on the right or bottom edge of the image.
+ */
+#define TJXOPT_PERFECT  1
+/**
+ * This option will cause #tjTransform() to discard any partial MCU blocks that
+ * cannot be transformed.
+ */
+#define TJXOPT_TRIM     2
+/**
+ * This option will enable lossless cropping.  See #tjTransform() for more
+ * information.
+ */
+#define TJXOPT_CROP     4
+/**
+ * This option will discard the color data in the input image and produce
+ * a grayscale output image.
+ */
+#define TJXOPT_GRAY     8
+/**
+ * This option will prevent #tjTransform() from outputting a JPEG image for
+ * this particular transform (this can be used in conjunction with a custom
+ * filter to capture the transformed DCT coefficients without transcoding
+ * them.)
+ */
+#define TJXOPT_NOOUTPUT 16
+
+
+/**
+ * Scaling factor
+ */
+typedef struct
+{
+  /**
+   * Numerator
+   */
+  int num;
+  /**
+   * Denominator
+   */
+  int denom;
+} tjscalingfactor;
+
+/**
+ * Cropping region
+ */
+typedef struct
+{
+  /**
+   * The left boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block width (see #tjMCUWidth.)
+   */
+  int x;
+  /**
+   * The upper boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block height (see #tjMCUHeight.)
+   */
+  int y;
+  /**
+   * The width of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the width of the source JPEG image - x.
+   */
+  int w;
+  /**
+   * The height of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the height of the source JPEG image - y.
+   */
+  int h;
+} tjregion;
+
+/**
+ * Lossless transform
+ */
+typedef struct tjtransform
+{
+  /**
+   * Cropping region
+   */
+  tjregion r;
+  /**
+   * One of the @ref TJXOP "transform operations"
+   */
+  int op;
+  /**
+   * The bitwise OR of one of more of the @ref TJXOPT_CROP "transform options"
+   */
+  int options;
+  /**
+   * Arbitrary data that can be accessed within the body of the callback
+   * function
+   */
+  void *data;
+  /**
+   * A callback function that can be used to modify the DCT coefficients
+   * after they are losslessly transformed but before they are transcoded to a
+   * new JPEG image.  This allows for custom filters or other transformations
+   * to be applied in the frequency domain.
+   *
+   * @param coeffs pointer to an array of transformed DCT coefficients.  (NOTE:
+   * this pointer is not guaranteed to be valid once the callback returns, so
+   * applications wishing to hand off the DCT coefficients to another function
+   * or library should make a copy of them within the body of the callback.)
+   *
+   * @param arrayRegion #tjregion structure containing the width and height of
+   * the array pointed to by <tt>coeffs</tt> as well as its offset relative to
+   * the component plane.  TurboJPEG implementations may choose to split each
+   * component plane into multiple DCT coefficient arrays and call the callback
+   * function once for each array.
+   *
+   * @param planeRegion #tjregion structure containing the width and height of
+   * the component plane to which <tt>coeffs</tt> belongs
+   *
+   * @param componentID ID number of the component plane to which
+   * <tt>coeffs</tt> belongs (Y, Cb, and Cr have, respectively, ID's of 0, 1,
+   * and 2 in typical JPEG images.)
+   *
+   * @param transformID ID number of the transformed image to which
+   * <tt>coeffs</tt> belongs.  This is the same as the index of the transform
+   * in the <tt>transforms</tt> array that was passed to #tjTransform().
+   *
+   * @param transform a pointer to a #tjtransform structure that specifies the
+   * parameters and/or cropping region for this transform
+   *
+   * @return 0 if the callback was successful, or -1 if an error occurred.
+   */
+  int (*customFilter)(short *coeffs, tjregion arrayRegion,
+    tjregion planeRegion, int componentIndex, int transformIndex,
+    struct tjtransform *transform);
+} tjtransform;
+
+/**
+ * TurboJPEG instance handle
+ */
+typedef void* tjhandle;
+
+
+/**
+ * Pad the given width to the nearest 32-bit boundary
+ */
+#define TJPAD(width) (((width)+3)&(~3))
+
+/**
+ * Compute the scaled value of <tt>dimension</tt> using the given scaling
+ * factor.  This macro performs the integer equivalent of <tt>ceil(dimension *
+ * scalingFactor)</tt>.
+ */
+#define TJSCALED(dimension, scalingFactor) ((dimension * scalingFactor.num \
+  + scalingFactor.denom - 1) / scalingFactor.denom)
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Create a TurboJPEG compressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT tjhandle DLLCALL tjInitCompress(void);
+
+
+/**
+ * Compress an RGB, grayscale, or CMYK image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB, grayscale, or
+ * CMYK pixels to be compressed
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer
+ * to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+  int width, int pitch, int height, int pixelFormat, unsigned char **jpegBuf,
+  unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags);
+
+
+/**
+ * Compress a YUV planar image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * compressed.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param pad the line padding used in the source image.  For instance, if each
+ * line in each plane of the YUV image is padded to the nearest multiple of 4
+ * bytes, then <tt>pad</tt> should be set to 4.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompressFromYUV(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pad, int height, int subsamp,
+  unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags);
+
+
+/**
+ * Compress a set of Y, U (Cb), and V (Cr) image planes into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if compressing a grayscale image) that contain a YUV
+ * image to be compressed.  These planes can be contiguous or non-contiguous in
+ * memory.  The size of each plane should match the value returned by
+ * #tjPlaneSizeYUV() for the given image width, height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to create a JPEG image from a subregion of a larger
+ * YUV planar image.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompressFromYUVPlanes(tjhandle handle,
+  const unsigned char **srcPlanes, int width, const int *strides, int height,
+  int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual,
+  int flags);
+
+
+/**
+ * The maximum size of the buffer (in bytes) required to hold a JPEG image with
+ * the given parameters.  The number of bytes returned by this function is
+ * larger than the size of the uncompressed source image.  The reason for this
+ * is that the JPEG format uses 16-bit coefficients, and it is thus possible
+ * for a very high-quality JPEG image with very high-frequency content to
+ * expand rather than compress when converted to the JPEG format.  Such images
+ * represent a very rare corner case, but since there is no way to predict the
+ * size of a JPEG image prior to compression, the corner case has to be
+ * handled.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @return the maximum size of the buffer (in bytes) required to hold the
+ * image, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjBufSize(int width, int height,
+  int jpegSubsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV planar image with
+ * the given parameters.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param pad the width of each line in each plane of the image is padded to
+ * the nearest multiple of this number of bytes (must be a power of 2.)
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the image, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV2(int width, int pad, int height,
+  int subsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV image plane with
+ * the given parameters.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image.  NOTE: this is the width of
+ * the whole image, not the plane width.
+ *
+ * @param stride bytes per line in the image plane.  Setting this to 0 is the
+ * equivalent of setting it to the plane width.
+ *
+ * @param height height (in pixels) of the YUV image.  NOTE: this is the height
+ * of the whole image, not the plane height.
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the YUV image
+ * plane, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjPlaneSizeYUV(int componentID, int width,
+  int stride, int height, int subsamp);
+
+
+/**
+ * The plane width of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane width.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane width of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp);
+
+
+/**
+ * The plane height of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane height.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param height height (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane height of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp);
+
+
+/**
+ * Encode an RGB or grayscale image into a YUV planar image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG compression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of chrominance subsampling.
+ * The Y, U (Cb), and V (Cr) image planes will be stored sequentially in the
+ * buffer (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjEncodeYUV3(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pitch, int height,
+  int pixelFormat, unsigned char *dstBuf, int pad, int subsamp, int flags);
+
+
+/**
+ * Encode an RGB or grayscale image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * compression process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if generating a grayscale image) that will receive the
+ * encoded image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the image width, height, strides, and level of chrominance subsampling.
+ * Refer to @ref YUVnotes "YUV Image Format Notes" for more details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to add an arbitrary amount of line
+ * padding to each plane or to encode an RGB or grayscale image into a
+ * subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjEncodeYUVPlanes(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pitch, int height,
+  int pixelFormat, unsigned char **dstPlanes, int *strides, int subsamp,
+  int flags);
+
+
+/**
+ * Create a TurboJPEG decompressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT tjhandle DLLCALL tjInitDecompress(void);
+
+
+/**
+ * Retrieve information about a JPEG image without decompressing it.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing a JPEG image
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param width pointer to an integer variable that will receive the width (in
+ * pixels) of the JPEG image
+ *
+ * @param height pointer to an integer variable that will receive the height
+ * (in pixels) of the JPEG image
+ *
+ * @param jpegSubsamp pointer to an integer variable that will receive the
+ * level of chrominance subsampling used when the JPEG image was compressed
+ * (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegColorspace pointer to an integer variable that will receive one
+ * of the JPEG colorspace constants, indicating the colorspace of the JPEG
+ * image (see @ref TJCS "JPEG colorspaces".)
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjDecompressHeader3(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, int *width,
+  int *height, int *jpegSubsamp, int *jpegColorspace);
+
+
+/**
+ * Returns a list of fractional scaling factors that the JPEG decompressor in
+ * this implementation of TurboJPEG supports.
+ *
+ * @param numscalingfactors pointer to an integer variable that will receive
+ * the number of elements in the list
+ *
+ * @return a pointer to a list of fractional scaling factors, or NULL if an
+ * error is encountered (see #tjGetErrorStr().)
+*/
+DLLEXPORT tjscalingfactor* DLLCALL tjGetScalingFactors(int *numscalingfactors);
+
+
+/**
+ * Decompress a JPEG image to an RGB, grayscale, or CMYK image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decompressed
+ * image.  This buffer should normally be <tt>pitch * scaledHeight</tt> bytes
+ * in size, where <tt>scaledHeight</tt> can be determined by calling
+ * #TJSCALED() with the JPEG image height and one of the scaling factors
+ * returned by #tjGetScalingFactors().  The <tt>dstBuf</tt> pointer may also be
+ * used to decompress into a specific region of a larger buffer.
+ *
+ * @param width desired width (in pixels) of the destination image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this is
+ * <tt>scaledWidth * #tjPixelSize[pixelFormat]</tt> if the decompressed image
+ * is unpadded, else <tt>#TJPAD(scaledWidth * #tjPixelSize[pixelFormat])</tt>
+ * if each line of the decompressed image is padded to the nearest 32-bit
+ * boundary, as is the case for Windows bitmaps.  (NOTE: <tt>scaledWidth</tt>
+ * can be determined by calling #TJSCALED() with the JPEG image width and one
+ * of the scaling factors returned by #tjGetScalingFactors().)  You can also be
+ * clever and use the pitch parameter to skip lines, etc.  Setting this
+ * parameter to 0 is the equivalent of setting it to
+ * <tt>scaledWidth * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height desired height (in pixels) of the destination image.  If this
+ * is different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref
+ * TJPF "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompress2(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pitch, int height, int pixelFormat, int flags);
+
+
+/**
+ * Decompress a JPEG image to a YUV planar image.  This function performs JPEG
+ * decompression but leaves out the color conversion step, so a planar YUV
+ * image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of subsampling.  The Y,
+ * U (Cb), and V (Cr) image planes will be stored sequentially in the buffer
+ * (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompressToYUV2(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pad, int height, int flags);
+
+
+/**
+ * Decompress a JPEG image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function performs JPEG decompression but leaves out the color
+ * conversion step, so a planar YUV image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decompressing a grayscale image) that will receive
+ * the YUV image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the scaled image width, scaled image height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the scaled plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective scaled plane
+ * widths.  You can adjust the strides in order to add an arbitrary amount of
+ * line padding to each plane or to decompress the JPEG image into a subregion
+ * of a larger YUV planar image.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompressToYUVPlanes(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize,
+  unsigned char **dstPlanes, int width, int *strides, int height, int flags);
+
+
+/**
+ * Decode a YUV planar image into an RGB or grayscale image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG decompression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * decoded.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param pad Use this parameter to specify that the width of each line in each
+ * plane of the YUV source image is padded to the nearest multiple of this
+ * number of bytes (must be a power of 2.)
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSize[pixelFormat]</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+  int pad, int subsamp, unsigned char *dstBuf, int width, int pitch,
+  int height, int pixelFormat, int flags);
+
+
+/**
+ * Decode a set of Y, U (Cb), and V (Cr) image planes into an RGB or grayscale
+ * image.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * decompression process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decoding a grayscale image) that contain a YUV image
+ * to be decoded.  These planes can be contiguous or non-contiguous in memory.
+ * The size of each plane should match the value returned by #tjPlaneSizeYUV()
+ * for the given image width, height, strides, and level of chrominance
+ * subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes" for more
+ * details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to decode a subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSize[pixelFormat]</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecodeYUVPlanes(tjhandle handle,
+  const unsigned char **srcPlanes, const int *strides, int subsamp,
+  unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat,
+  int flags);
+
+
+/**
+ * Create a new TurboJPEG transformer instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT tjhandle DLLCALL tjInitTransform(void);
+
+
+/**
+ * Losslessly transform a JPEG image into another JPEG image.  Lossless
+ * transforms work by moving the raw DCT coefficients from one JPEG image
+ * structure to another without altering the values of the coefficients.  While
+ * this is typically faster than decompressing the image, transforming it, and
+ * re-compressing it, lossless transforms are not free.  Each lossless
+ * transform requires reading and performing Huffman decoding on all of the
+ * coefficients in the source image, regardless of the size of the destination
+ * image.  Thus, this function provides a means of generating multiple
+ * transformed images from the same source or  applying multiple
+ * transformations simultaneously, in order to eliminate the need to read the
+ * source coefficients multiple times.
+ *
+ * @param handle a handle to a TurboJPEG transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG source image to
+ * transform
+ *
+ * @param jpegSize size of the JPEG source image (in bytes)
+ *
+ * @param n the number of transformed JPEG images to generate
+ *
+ * @param dstBufs pointer to an array of n image buffers.  <tt>dstBufs[i]</tt>
+ * will receive a JPEG image that has been transformed using the parameters in
+ * <tt>transforms[i]</tt>.  TurboJPEG has the ability to reallocate the JPEG
+ * buffer to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>dstBufs[i]</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize() with the transformed or cropped width and height.  This should
+ * ensure that the buffer never has to be re-allocated (setting
+ * #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>dstSizes[i]</tt> should be set to the size of
+ * your pre-allocated buffer.  In any case, unless you have set
+ * #TJFLAG_NOREALLOC, you should always check <tt>dstBufs[i]</tt> upon return
+ * from this function, as it may have changed.
+ *
+ * @param dstSizes pointer to an array of n unsigned long variables that will
+ * receive the actual sizes (in bytes) of each transformed JPEG image.  If
+ * <tt>dstBufs[i]</tt> points to a pre-allocated buffer, then
+ * <tt>dstSizes[i]</tt> should be set to the size of the buffer.  Upon return,
+ * <tt>dstSizes[i]</tt> will contain the size of the JPEG image (in bytes.)
+ *
+ * @param transforms pointer to an array of n #tjtransform structures, each of
+ * which specifies the transform parameters and/or cropping region for the
+ * corresponding transformed output image.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjTransform(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, int n,
+  unsigned char **dstBufs, unsigned long *dstSizes, tjtransform *transforms,
+  int flags);
+
+
+/**
+ * Destroy a TurboJPEG compressor, decompressor, or transformer instance.
+ *
+ * @param handle a handle to a TurboJPEG compressor, decompressor or
+ * transformer instance
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDestroy(tjhandle handle);
+
+
+/**
+ * Allocate an image buffer for use with TurboJPEG.  You should always use
+ * this function to allocate the JPEG destination buffer(s) for #tjCompress2()
+ * and #tjTransform() unless you are disabling automatic buffer
+ * (re)allocation (by setting #TJFLAG_NOREALLOC.)
+ *
+ * @param bytes the number of bytes to allocate
+ *
+ * @return a pointer to a newly-allocated buffer with the specified number of
+ * bytes.
+ *
+ * @sa tjFree()
+ */
+DLLEXPORT unsigned char* DLLCALL tjAlloc(int bytes);
+
+
+/**
+ * Free an image buffer previously allocated by TurboJPEG.  You should always
+ * use this function to free JPEG destination buffer(s) that were automatically
+ * (re)allocated by #tjCompress2() or #tjTransform() or that were manually
+ * allocated using #tjAlloc().
+ *
+ * @param buffer address of the buffer to free
+ *
+ * @sa tjAlloc()
+ */
+DLLEXPORT void DLLCALL tjFree(unsigned char *buffer);
+
+
+/**
+ * Returns a descriptive error message explaining why the last command failed.
+ *
+ * @return a descriptive error message explaining why the last command failed.
+ */
+DLLEXPORT char* DLLCALL tjGetErrorStr(void);
+
+
+/* Deprecated functions and macros */
+#define TJFLAG_FORCEMMX        8
+#define TJFLAG_FORCESSE       16
+#define TJFLAG_FORCESSE2      32
+#define TJFLAG_FORCESSE3     128
+
+
+/* Backward compatibility functions and macros (nothing to see here) */
+#define NUMSUBOPT TJ_NUMSAMP
+#define TJ_444 TJSAMP_444
+#define TJ_422 TJSAMP_422
+#define TJ_420 TJSAMP_420
+#define TJ_411 TJSAMP_420
+#define TJ_GRAYSCALE TJSAMP_GRAY
+
+#define TJ_BGR 1
+#define TJ_BOTTOMUP TJFLAG_BOTTOMUP
+#define TJ_FORCEMMX TJFLAG_FORCEMMX
+#define TJ_FORCESSE TJFLAG_FORCESSE
+#define TJ_FORCESSE2 TJFLAG_FORCESSE2
+#define TJ_ALPHAFIRST 64
+#define TJ_FORCESSE3 TJFLAG_FORCESSE3
+#define TJ_FASTUPSAMPLE TJFLAG_FASTUPSAMPLE
+#define TJ_YUV 512
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZE(int width, int height);
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZEYUV(int width, int height,
+  int jpegSubsamp);
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV(int width, int height,
+  int subsamp);
+
+DLLEXPORT int DLLCALL tjCompress(tjhandle handle, unsigned char *srcBuf,
+  int width, int pitch, int height, int pixelSize, unsigned char *dstBuf,
+  unsigned long *compressedSize, int jpegSubsamp, int jpegQual, int flags);
+
+DLLEXPORT int DLLCALL tjEncodeYUV(tjhandle handle,
+  unsigned char *srcBuf, int width, int pitch, int height, int pixelSize,
+  unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int DLLCALL tjEncodeYUV2(tjhandle handle,
+  unsigned char *srcBuf, int width, int pitch, int height, int pixelFormat,
+  unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int DLLCALL tjDecompressHeader(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height);
+
+DLLEXPORT int DLLCALL tjDecompressHeader2(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height,
+  int *jpegSubsamp);
+
+DLLEXPORT int DLLCALL tjDecompress(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pitch, int height, int pixelSize, int flags);
+
+DLLEXPORT int DLLCALL tjDecompressToYUV(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int flags);
+
+
+/**
+ * @}
+ */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-jni.c glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-jni.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-jni.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-jni.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1166 @@
+/*
+ * Copyright (C)2011-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "turbojpeg.h"
+#ifdef WIN32
+#include "tjutil.h"
+#endif
+#include <jni.h>
+#include "java/org_libjpegturbo_turbojpeg_TJCompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJDecompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJ.h"
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+
+#define _throw(msg, exceptionClass) {  \
+	jclass _exccls=(*env)->FindClass(env, exceptionClass);  \
+	if(!_exccls || (*env)->ExceptionCheck(env)) goto bailout;  \
+	(*env)->ThrowNew(env, _exccls, msg);  \
+	goto bailout;  \
+}
+
+#define _throwtj() _throw(tjGetErrorStr(), "org/libjpegturbo/turbojpeg/TJException")
+
+#define _throwarg(msg) _throw(msg, "java/lang/IllegalArgumentException")
+
+#define _throwmem() _throw("Memory allocation failure", "java/lang/OutOfMemoryError");
+
+#define bailif0(f) {if(!(f) || (*env)->ExceptionCheck(env)) {  \
+	goto bailout;  \
+}}
+
+#define gethandle()  \
+	jclass _cls=(*env)->GetObjectClass(env, obj);  \
+	jfieldID _fid;  \
+	if(!_cls || (*env)->ExceptionCheck(env)) goto bailout;  \
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "handle", "J"));  \
+	handle=(tjhandle)(size_t)(*env)->GetLongField(env, obj, _fid);  \
+
+#ifdef _WIN32
+#define setenv(envvar, value, dummy) _putenv_s(envvar, value)
+#endif
+
+#define prop2env(property, envvar)  \
+{  \
+	if((jName=(*env)->NewStringUTF(env, property))!=NULL  \
+		&& (jValue=(*env)->CallStaticObjectMethod(env, cls, mid, jName))!=NULL)  \
+	{  \
+		if((value=(*env)->GetStringUTFChars(env, jValue, 0))!=NULL)  \
+		{  \
+			setenv(envvar, value, 1);  \
+			(*env)->ReleaseStringUTFChars(env, jValue, value);  \
+		}  \
+	}  \
+}
+
+int ProcessSystemProperties(JNIEnv *env)
+{
+	jclass cls;  jmethodID mid;
+	jstring jName, jValue;
+	const char *value;
+
+	bailif0(cls=(*env)->FindClass(env, "java/lang/System"));
+	bailif0(mid=(*env)->GetStaticMethodID(env, cls, "getProperty",
+		"(Ljava/lang/String;)Ljava/lang/String;"));
+
+	prop2env("turbojpeg.optimize", "TJ_OPTIMIZE");
+	prop2env("turbojpeg.arithmetic", "TJ_ARITHMETIC");
+	prop2env("turbojpeg.restart", "TJ_RESTART");
+	prop2env("turbojpeg.progressive", "TJ_PROGRESSIVE");
+	return 0;
+
+	bailout:
+	return -1;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSize() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSize
+	(JNIEnv *env, jclass cls, jint width, jint height, jint jpegSubsamp)
+{
+	jint retval=(jint)tjBufSize(width, height, jpegSubsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII
+	(JNIEnv *env, jclass cls, jint width, jint pad, jint height, jint subsamp)
+{
+	jint retval=(jint)tjBufSizeYUV2(width, pad, height, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III
+	(JNIEnv *env, jclass cls, jint width, jint height, jint subsamp)
+{
+	return Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII(env, cls, width,
+		4, height, subsamp);
+}
+
+/* TurboJPEG 1.4.x: TJ::planeSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII
+	(JNIEnv *env, jclass cls, jint componentID, jint width, jint stride,
+		jint height, jint subsamp)
+{
+	jint retval=(jint)tjPlaneSizeYUV(componentID, width, stride, height,
+		subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeWidth() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III
+	(JNIEnv *env, jclass cls, jint componentID, jint width, jint subsamp)
+{
+	jint retval=(jint)tjPlaneWidth(componentID, width, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeHeight() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III
+	(JNIEnv *env, jclass cls, jint componentID, jint height, jint subsamp)
+{
+	jint retval=(jint)tjPlaneHeight(componentID, height, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitCompress())==NULL)
+		_throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+static jint TJCompressor_compress
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jbyteArray dst,
+		jint jpegSubsamp, jint jpegQual, jint flags)
+{
+	tjhandle handle=0;
+	unsigned long jpegSize=0;
+	jsize arraySize=0, actualPitch;
+	unsigned char *srcBuf=NULL, *jpegBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0)
+		_throwarg("Invalid argument in compress()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+	jpegSize=tjBufSize(width, height, jpegSubsamp);
+	if((*env)->GetArrayLength(env, dst)<(jsize)jpegSize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(ProcessSystemProperties(env)<0) goto bailout;
+
+	if(tjCompress2(handle, &srcBuf[y*actualPitch + x*tjPixelSize[pf]], width,
+		pitch, height, pf, &jpegBuf, &jpegSize, jpegSubsamp, jpegQual,
+		flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, jpegBuf, 0);
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	return (jint)jpegSize;
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+		jint pitch, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+		jint jpegQual, jint flags)
+{
+	return TJCompressor_compress(env, obj, src, 1, x, y, width, pitch, height,
+		pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+		jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+		jint flags)
+{
+	return TJCompressor_compress(env, obj, src, 1, 0, 0, width, pitch, height,
+		pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII
+	(JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+		jint stride, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+		jint jpegQual, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in compress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+	return TJCompressor_compress(env, obj, src, sizeof(jint), x, y, width,
+		stride*sizeof(jint), height, pf, dst, jpegSubsamp, jpegQual, flags);
+
+	bailout:
+	return 0;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII
+	(JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+		jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+		jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in compress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+	return TJCompressor_compress(env, obj, src, sizeof(jint), 0, 0, width,
+		stride*sizeof(jint), height, pf, dst, jpegSubsamp, jpegQual, flags);
+
+	bailout:
+	return 0;
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::compressFromYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jint width, jintArray jSrcStrides, jint height, jint subsamp,
+		jbyteArray dst, jint jpegQual, jint flags)
+{
+	tjhandle handle=0;
+	unsigned long jpegSize=0;
+	jbyteArray jSrcPlanes[3]={NULL, NULL, NULL};
+	const unsigned char *srcPlanes[3];
+	unsigned char *jpegBuf=NULL;
+	int *srcOffsets=NULL, *srcStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(subsamp<0 || subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in compressFromYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, srcobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	jpegSize=tjBufSize(width, height, subsamp);
+	if((*env)->GetArrayLength(env, dst)<(jsize)jpegSize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcOffsets=(*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+	bailif0(srcStrides=(*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, srcStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(srcOffsets[i]<0)
+			_throwarg("Invalid argument in compressFromYUV()");
+		if(srcStrides[i]<0 && srcOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jSrcPlanes[i]=(*env)->GetObjectArrayElement(env, srcobjs, i));
+		if((*env)->GetArrayLength(env, jSrcPlanes[i])<srcOffsets[i]+planeSize)
+			_throwarg("Source plane is not large enough");
+
+		bailif0(srcPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jSrcPlanes[i],
+			0));
+		srcPlanes[i]=&srcPlanes[i][srcOffsets[i]];
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(ProcessSystemProperties(env)<0) goto bailout;
+
+	if(tjCompressFromYUVPlanes(handle, srcPlanes, width, srcStrides, height,
+		subsamp, &jpegBuf, &jpegSize, jpegQual, flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, jpegBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(srcPlanes[i] && jSrcPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jSrcPlanes[i],
+				(unsigned char *)srcPlanes[i], 0);
+	}
+	if(srcStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcStrides, srcStrides, 0);
+	if(srcOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcOffsets, srcOffsets, 0);
+	return (jint)jpegSize;
+}
+
+static void TJCompressor_encodeYUV
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	jbyteArray jDstPlanes[3]={NULL, NULL, NULL};
+	unsigned char *srcBuf=NULL, *dstPlanes[3];
+	int *dstOffsets=NULL, *dstStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0 || subsamp<0 || subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF
+		|| org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, dstobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jDstOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jDstStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+
+	bailif0(dstOffsets=(*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+	bailif0(dstStrides=(*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, dstStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(dstOffsets[i]<0)
+			_throwarg("Invalid argument in encodeYUV()");
+		if(dstStrides[i]<0 && dstOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jDstPlanes[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((*env)->GetArrayLength(env, jDstPlanes[i])<dstOffsets[i]+planeSize)
+			_throwarg("Destination plane is not large enough");
+
+		bailif0(dstPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jDstPlanes[i],
+			0));
+		dstPlanes[i]=&dstPlanes[i][dstOffsets[i]];
+	}
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjEncodeYUVPlanes(handle, &srcBuf[y*actualPitch + x*tjPixelSize[pf]],
+		width, pitch, height, pf, dstPlanes, dstStrides, subsamp, flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(dstPlanes[i] && jDstPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jDstPlanes[i], dstPlanes[i],
+				0);
+	}
+	if(dstStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstStrides, dstStrides, 0);
+	if(dstOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstOffsets, dstOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III
+	(JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+		jint pitch, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	TJCompressor_encodeYUV(env, obj, src, 1, x, y, width, pitch, height, pf,
+		dstobjs, jDstOffsets, jDstStrides, subsamp, flags);
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III
+	(JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+		jint stride, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+	TJCompressor_encodeYUV(env, obj, src, sizeof(jint), x, y, width,
+		stride*sizeof(jint), height, pf, dstobjs, jDstOffsets, jDstStrides,
+		subsamp, flags);
+
+	bailout:
+	return;
+}
+
+JNIEXPORT void JNICALL TJCompressor_encodeYUV_12
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint width,
+		jint pitch, jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0;
+	unsigned char *srcBuf=NULL, *dstBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	arraySize=(pitch==0)? width*tjPixelSize[pf]*height:pitch*height;
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+	if((*env)->GetArrayLength(env, dst)
+		<(jsize)tjBufSizeYUV(width, height, subsamp))
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjEncodeYUV2(handle, srcBuf, width, pitch, height, pf, dstBuf, subsamp,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+		jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	TJCompressor_encodeYUV_12(env, obj, src, 1, width, pitch, height, pf, dst,
+		subsamp, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII
+	(JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+		jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+	TJCompressor_encodeYUV_12(env, obj, src, sizeof(jint), width,
+		stride*sizeof(jint), height, pf, dst, subsamp, flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy
+	(JNIEnv *env, jobject obj)
+{
+	tjhandle handle=0;
+
+	gethandle();
+
+	if(tjDestroy(handle)==-1) _throwtj();
+	(*env)->SetLongField(env, obj, _fid, 0);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitDecompress())==NULL) _throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::getScalingFactors() */
+JNIEXPORT jobjectArray JNICALL Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors
+	(JNIEnv *env, jclass cls)
+{
+	jclass sfcls=NULL;  jfieldID fid=0;
+	tjscalingfactor *sf=NULL;  int n=0, i;
+	jobject sfobj=NULL;
+	jobjectArray sfjava=NULL;
+
+	if((sf=tjGetScalingFactors(&n))==NULL || n==0)
+		_throwarg(tjGetErrorStr());
+
+	bailif0(sfcls=(*env)->FindClass(env, "org/libjpegturbo/turbojpeg/TJScalingFactor"));
+	bailif0(sfjava=(jobjectArray)(*env)->NewObjectArray(env, n, sfcls, 0));
+
+	for(i=0; i<n; i++)
+	{
+		bailif0(sfobj=(*env)->AllocObject(env, sfcls));
+		bailif0(fid=(*env)->GetFieldID(env, sfcls, "num", "I"));
+		(*env)->SetIntField(env, sfobj, fid, sf[i].num);
+		bailif0(fid=(*env)->GetFieldID(env, sfcls, "denom", "I"));
+		(*env)->SetIntField(env, sfobj, fid, sf[i].denom);
+		(*env)->SetObjectArrayElement(env, sfjava, i, sfobj);
+	}
+
+	bailout:
+	return sfjava;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressHeader() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize)
+{
+	tjhandle handle=0;
+	unsigned char *jpegBuf=NULL;
+	int width=0, height=0, jpegSubsamp=-1, jpegColorspace=-1;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjDecompressHeader3(handle, jpegBuf, (unsigned long)jpegSize,
+		&width, &height, &jpegSubsamp, &jpegColorspace)==-1)
+		_throwtj();
+
+	(*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);  jpegBuf=NULL;
+
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	(*env)->SetIntField(env, obj, _fid, jpegSubsamp);
+	if((_fid=(*env)->GetFieldID(env, _cls, "jpegColorspace", "I"))==0)
+		(*env)->ExceptionClear(env);
+	else
+		(*env)->SetIntField(env, obj, _fid, jpegColorspace);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	(*env)->SetIntField(env, obj, _fid, width);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	(*env)->SetIntField(env, obj, _fid, height);
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+static void TJDecompressor_decompress
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jarray dst,
+		jint dstElementSize, jint x, jint y, jint width, jint pitch, jint height,
+		jint pf, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	unsigned char *jpegBuf=NULL, *dstBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, dst)*dstElementSize<arraySize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecompress2(handle, jpegBuf, (unsigned long)jpegSize,
+		&dstBuf[y*actualPitch + x*tjPixelSize[pf]], width, pitch, height, pf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, x, y, width,
+		pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, 0, 0, width,
+		pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+		jint x, jint y, jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), x, y,
+		width, stride*sizeof(jint), height, pf, flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+		jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), 0, 0,
+		width, stride*sizeof(jint), height, pf, flags);
+
+	bailout:
+	return;
+
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize,
+		jobjectArray dstobjs, jintArray jDstOffsets, jint desiredWidth,
+		jintArray jDstStrides, jint desiredHeight, jint flags)
+{
+	tjhandle handle=0;
+	jbyteArray jDstPlanes[3]={NULL, NULL, NULL};
+	unsigned char *jpegBuf=NULL, *dstPlanes[3];
+	int *dstOffsets=NULL, *dstStrides=NULL;
+	int jpegSubsamp=-1, jpegWidth=0, jpegHeight=0;
+	int nc=0, i, width, height, scaledWidth, scaledHeight, nsf=0;
+	tjscalingfactor *sf;
+
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+
+	nc=(jpegSubsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3);
+
+	width=desiredWidth;  height=desiredHeight;
+	if(width==0) width=jpegWidth;
+	if(height==0) height=jpegHeight;
+	sf=tjGetScalingFactors(&nsf);
+	if(!sf || nsf<1)
+		_throwarg(tjGetErrorStr());
+	for(i=0; i<nsf; i++)
+	{
+		scaledWidth=TJSCALED(jpegWidth, sf[i]);
+		scaledHeight=TJSCALED(jpegHeight, sf[i]);
+		if(scaledWidth<=width && scaledHeight<=height)
+			break;
+	}
+	if(i>=nsf)
+		_throwarg("Could not scale down to desired image dimensions");
+
+	bailif0(dstOffsets=(*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+	bailif0(dstStrides=(*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, scaledWidth, dstStrides[i], scaledHeight,
+			jpegSubsamp);
+		int pw=tjPlaneWidth(i, scaledWidth, jpegSubsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(dstOffsets[i]<0)
+			_throwarg("Invalid argument in decompressToYUV()");
+		if(dstStrides[i]<0 && dstOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jDstPlanes[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((*env)->GetArrayLength(env, jDstPlanes[i])<dstOffsets[i]+planeSize)
+			_throwarg("Destination plane is not large enough");
+
+		bailif0(dstPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jDstPlanes[i],
+			0));
+		dstPlanes[i]=&dstPlanes[i][dstOffsets[i]];
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjDecompressToYUVPlanes(handle, jpegBuf, (unsigned long)jpegSize,
+		dstPlanes, desiredWidth, dstStrides, desiredHeight, flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(dstPlanes[i] && jDstPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jDstPlanes[i], dstPlanes[i],
+				0);
+	}
+	if(dstStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstStrides, dstStrides, 0);
+	if(dstOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstOffsets, dstOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint flags)
+{
+	tjhandle handle=0;
+	unsigned char *jpegBuf=NULL, *dstBuf=NULL;
+	int jpegSubsamp=-1, jpegWidth=0, jpegHeight=0;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+	if((*env)->GetArrayLength(env, dst)
+		<(jsize)tjBufSizeYUV(jpegWidth, jpegHeight, jpegSubsamp))
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecompressToYUV(handle, jpegBuf, (unsigned long)jpegSize, dstBuf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+static void TJDecompressor_decodeYUV
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jarray dst, jint dstElementSize,
+		jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	jbyteArray jSrcPlanes[3]={NULL, NULL, NULL};
+	const unsigned char *srcPlanes[3];
+	unsigned char *dstBuf=NULL;
+	int *srcOffsets=NULL, *srcStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || subsamp<0
+		|| subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in decodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF
+		|| org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, srcobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, dst)*dstElementSize<arraySize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcOffsets=(*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+	bailif0(srcStrides=(*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, srcStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(srcOffsets[i]<0)
+			_throwarg("Invalid argument in decodeYUV()");
+		if(srcStrides[i]<0 && srcOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jSrcPlanes[i]=(*env)->GetObjectArrayElement(env, srcobjs, i));
+		if((*env)->GetArrayLength(env, jSrcPlanes[i])<srcOffsets[i]+planeSize)
+			_throwarg("Source plane is not large enough");
+
+		bailif0(srcPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jSrcPlanes[i],
+			0));
+		srcPlanes[i]=&srcPlanes[i][srcOffsets[i]];
+	}
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecodeYUVPlanes(handle, srcPlanes, srcStrides, subsamp,
+		&dstBuf[y*actualPitch + x*tjPixelSize[pf]], width, pitch, height, pf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(srcPlanes[i] && jSrcPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jSrcPlanes[i],
+				(unsigned char *)srcPlanes[i], 0);
+	}
+	if(srcStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcStrides, srcStrides, 0);
+	if(srcOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcOffsets, srcOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jbyteArray dst, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+		subsamp, dst, 1, x, y, width, pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jintArray dst, jint x, jint y,
+		jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decoding to an integer buffer.");
+
+	TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+		subsamp, dst, sizeof(jint), x, y, width, stride*sizeof(jint), height, pf,
+		flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitTransform())==NULL) _throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+typedef struct _JNICustomFilterParams
+{
+	JNIEnv *env;
+	jobject tobj;
+	jobject cfobj;
+} JNICustomFilterParams;
+
+static int JNICustomFilter(short *coeffs, tjregion arrayRegion,
+	tjregion planeRegion, int componentIndex, int transformIndex,
+	tjtransform *transform)
+{
+	JNICustomFilterParams *params=(JNICustomFilterParams *)transform->data;
+	JNIEnv *env=params->env;
+	jobject tobj=params->tobj, cfobj=params->cfobj;
+	jobject arrayRegionObj, planeRegionObj, bufobj, borobj;
+	jclass cls;  jmethodID mid;  jfieldID fid;
+
+	bailif0(bufobj=(*env)->NewDirectByteBuffer(env, coeffs,
+		sizeof(short)*arrayRegion.w*arrayRegion.h));
+	bailif0(cls=(*env)->FindClass(env, "java/nio/ByteOrder"));
+	bailif0(mid=(*env)->GetStaticMethodID(env, cls, "nativeOrder",
+		"()Ljava/nio/ByteOrder;"));
+	bailif0(borobj=(*env)->CallStaticObjectMethod(env, cls, mid));
+	bailif0(cls=(*env)->GetObjectClass(env, bufobj));
+	bailif0(mid=(*env)->GetMethodID(env, cls, "order",
+		"(Ljava/nio/ByteOrder;)Ljava/nio/ByteBuffer;"));
+	(*env)->CallObjectMethod(env, bufobj, mid, borobj);
+	bailif0(mid=(*env)->GetMethodID(env, cls, "asShortBuffer",
+		"()Ljava/nio/ShortBuffer;"));
+	bailif0(bufobj=(*env)->CallObjectMethod(env, bufobj, mid));
+
+	bailif0(cls=(*env)->FindClass(env, "java/awt/Rectangle"));
+	bailif0(arrayRegionObj=(*env)->AllocObject(env, cls));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "x", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.x);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "y", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.y);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "width", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.w);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "height", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.h);
+
+	bailif0(planeRegionObj=(*env)->AllocObject(env, cls));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "x", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.x);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "y", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.y);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "width", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.w);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "height", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.h);
+
+	bailif0(cls=(*env)->GetObjectClass(env, cfobj));
+	bailif0(mid=(*env)->GetMethodID(env, cls, "customFilter",
+		"(Ljava/nio/ShortBuffer;Ljava/awt/Rectangle;Ljava/awt/Rectangle;IILorg/libjpegturbo/turbojpeg/TJTransform;)V"));
+	(*env)->CallVoidMethod(env, cfobj, mid, bufobj, arrayRegionObj,
+		planeRegionObj, componentIndex, transformIndex, tobj);
+
+	return 0;
+
+	bailout:
+	return -1;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::transform() */
+JNIEXPORT jintArray JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_transform
+	(JNIEnv *env, jobject obj, jbyteArray jsrcBuf, jint jpegSize,
+		jobjectArray dstobjs, jobjectArray tobjs, jint flags)
+{
+	tjhandle handle=0;  int i;
+	unsigned char *jpegBuf=NULL, **dstBufs=NULL;  jsize n=0;
+	unsigned long *dstSizes=NULL;  tjtransform *t=NULL;
+	jbyteArray *jdstBufs=NULL;
+	int jpegWidth=0, jpegHeight=0, jpegSubsamp;
+	jintArray jdstSizes=0;  jint *dstSizesi=NULL;
+	JNICustomFilterParams *params=NULL;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, jsrcBuf)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+
+	n=(*env)->GetArrayLength(env, dstobjs);
+	if(n!=(*env)->GetArrayLength(env, tobjs))
+		_throwarg("Mismatch between size of transforms array and destination buffers array");
+
+	if((dstBufs=(unsigned char **)malloc(sizeof(unsigned char *)*n))==NULL)
+		_throwmem();
+	if((jdstBufs=(jbyteArray *)malloc(sizeof(jbyteArray)*n))==NULL)
+		_throwmem();
+	if((dstSizes=(unsigned long *)malloc(sizeof(unsigned long)*n))==NULL)
+		_throwmem();
+	if((t=(tjtransform *)malloc(sizeof(tjtransform)*n))==NULL)
+		_throwmem();
+	if((params=(JNICustomFilterParams *)malloc(sizeof(JNICustomFilterParams)*n))
+		==NULL)
+		_throwmem();
+	for(i=0; i<n; i++)
+	{
+		dstBufs[i]=NULL;  jdstBufs[i]=NULL;  dstSizes[i]=0;
+		memset(&t[i], 0, sizeof(tjtransform));
+		memset(&params[i], 0, sizeof(JNICustomFilterParams));
+	}
+
+	for(i=0; i<n; i++)
+	{
+		jobject tobj, cfobj;
+
+		bailif0(tobj=(*env)->GetObjectArrayElement(env, tobjs, i));
+		bailif0(_cls=(*env)->GetObjectClass(env, tobj));
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "op", "I"));
+		t[i].op=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "options", "I"));
+		t[i].options=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "x", "I"));
+		t[i].r.x=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "y", "I"));
+		t[i].r.y=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "width", "I"));
+		t[i].r.w=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "height", "I"));
+		t[i].r.h=(*env)->GetIntField(env, tobj, _fid);
+
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "cf",
+			"Lorg/libjpegturbo/turbojpeg/TJCustomFilter;"));
+		cfobj=(*env)->GetObjectField(env, tobj, _fid);
+		if(cfobj)
+		{
+			params[i].env=env;
+			params[i].tobj=tobj;
+			params[i].cfobj=cfobj;
+			t[i].customFilter=JNICustomFilter;
+			t[i].data=(void *)&params[i];
+		}
+	}
+
+	for(i=0; i<n; i++)
+	{
+		int w=jpegWidth, h=jpegHeight;
+		if(t[i].r.w!=0) w=t[i].r.w;
+		if(t[i].r.h!=0) h=t[i].r.h;
+		bailif0(jdstBufs[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((unsigned long)(*env)->GetArrayLength(env, jdstBufs[i])
+			<tjBufSize(w, h, jpegSubsamp))
+			_throwarg("Destination buffer is not large enough");
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, jsrcBuf, 0));
+	for(i=0; i<n; i++)
+		bailif0(dstBufs[i]=(*env)->GetPrimitiveArrayCritical(env, jdstBufs[i], 0));
+
+	if(tjTransform(handle, jpegBuf, jpegSize, n, dstBufs, dstSizes, t,
+		flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	for(i=0; i<n; i++)
+	{
+		(*env)->ReleasePrimitiveArrayCritical(env, jdstBufs[i], dstBufs[i], 0);
+		dstBufs[i]=NULL;
+	}
+	(*env)->ReleasePrimitiveArrayCritical(env, jsrcBuf, jpegBuf, 0);
+	jpegBuf=NULL;
+
+	jdstSizes=(*env)->NewIntArray(env, n);
+	bailif0(dstSizesi=(*env)->GetIntArrayElements(env, jdstSizes, 0));
+	for(i=0; i<n; i++) dstSizesi[i]=(int)dstSizes[i];
+
+	bailout:
+	if(dstSizesi) (*env)->ReleaseIntArrayElements(env, jdstSizes, dstSizesi, 0);
+	if(dstBufs)
+	{
+		for(i=0; i<n; i++)
+		{
+			if(dstBufs[i] && jdstBufs && jdstBufs[i])
+				(*env)->ReleasePrimitiveArrayCritical(env, jdstBufs[i], dstBufs[i], 0);
+		}
+		free(dstBufs);
+	}
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, jsrcBuf, jpegBuf, 0);
+	if(jdstBufs) free(jdstBufs);
+	if(dstSizes) free(dstSizes);
+	if(t) free(t);
+	return jdstSizes;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy
+	(JNIEnv *env, jobject obj)
+{
+	Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy(env, obj);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-mapfile glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-mapfile
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-mapfile	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-mapfile	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+TURBOJPEG_1.0
+{
+	global:
+		tjInitCompress;
+		tjCompress;
+		TJBUFSIZE;
+		tjInitDecompress;
+		tjDecompressHeader;
+		tjDecompress;
+		tjDestroy;
+		tjGetErrorStr;
+	local:
+		*;
+};
+
+TURBOJPEG_1.1
+{
+	global:
+		TJBUFSIZEYUV;
+		tjDecompressHeader2;
+		tjDecompressToYUV;
+		tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+	global:
+		tjAlloc;
+		tjBufSize;
+		tjBufSizeYUV;
+		tjCompress2;
+		tjDecompress2;
+		tjEncodeYUV2;
+		tjFree;
+		tjGetScalingFactors;
+		tjInitTransform;
+		tjTransform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.4
+{
+	global:
+		tjBufSizeYUV2;
+		tjCompressFromYUV;
+		tjCompressFromYUVPlanes;
+		tjDecodeYUV;
+		tjDecodeYUVPlanes;
+		tjDecompressHeader3;
+		tjDecompressToYUV2;
+		tjDecompressToYUVPlanes;
+		tjEncodeYUV3;
+		tjEncodeYUVPlanes;
+		tjPlaneHeight;
+		tjPlaneSizeYUV;
+		tjPlaneWidth;
+} TURBOJPEG_1.2;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-mapfile.jni glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-mapfile.jni
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/turbojpeg-mapfile.jni	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/turbojpeg-mapfile.jni	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,92 @@
+TURBOJPEG_1.0
+{
+	global:
+		tjInitCompress;
+		tjCompress;
+		TJBUFSIZE;
+		tjInitDecompress;
+		tjDecompressHeader;
+		tjDecompress;
+		tjDestroy;
+		tjGetErrorStr;
+	local:
+		*;
+};
+
+TURBOJPEG_1.1
+{
+	global:
+		TJBUFSIZEYUV;
+		tjDecompressHeader2;
+		tjDecompressToYUV;
+		tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+	global:
+		tjAlloc;
+		tjBufSize;
+		tjBufSizeYUV;
+		tjCompress2;
+		tjDecompress2;
+		tjEncodeYUV2;
+		tjFree;
+		tjGetScalingFactors;
+		tjInitTransform;
+		tjTransform;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSize;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III;
+		Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_init;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_init;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy;
+		Java_org_libjpegturbo_turbojpeg_TJTransformer_init;
+		Java_org_libjpegturbo_turbojpeg_TJTransformer_transform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.3
+{
+	global:
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII;
+} TURBOJPEG_1.2;
+
+TURBOJPEG_1.4
+{
+	global:
+		tjBufSizeYUV2;
+		tjCompressFromYUV;
+		tjCompressFromYUVPlanes;
+		tjDecodeYUV;
+		tjDecodeYUVPlanes;
+		tjDecompressHeader3;
+		tjDecompressToYUV2;
+		tjDecompressToYUVPlanes;
+		tjEncodeYUV3;
+		tjEncodeYUVPlanes;
+		tjPlaneHeight;
+		tjPlaneSizeYUV;
+		tjPlaneWidth;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III;
+} TURBOJPEG_1.3;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/usage.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/usage.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/usage.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/usage.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,660 @@
+NOTE:  This file was modified by The libjpeg-turbo Project to include only
+information relevant to libjpeg-turbo and to wordsmith certain sections.
+
+USAGE instructions for the Independent JPEG Group's JPEG software
+=================================================================
+
+This file describes usage of the JPEG conversion programs cjpeg and djpeg,
+as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
+the other documentation files if you wish to use the JPEG library within
+your own programs.)
+
+If you are on a Unix machine you may prefer to read the Unix-style manual
+pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.
+
+
+INTRODUCTION
+
+These programs implement JPEG image encoding, decoding, and transcoding.
+JPEG (pronounced "jay-peg") is a standardized compression method for
+full-color and grayscale images.
+
+
+GENERAL USAGE
+
+We provide two programs, cjpeg to compress an image file into JPEG format,
+and djpeg to decompress a JPEG file back into a conventional image format.
+
+On Unix-like systems, you say:
+        cjpeg [switches] [imagefile] >jpegfile
+or
+        djpeg [switches] [jpegfile]  >imagefile
+The programs read the specified input file, or standard input if none is
+named.  They always write to standard output (with trace/error messages to
+standard error).  These conventions are handy for piping images between
+programs.
+
+On most non-Unix systems, you say:
+        cjpeg [switches] imagefile jpegfile
+or
+        djpeg [switches] jpegfile  imagefile
+i.e., both the input and output files are named on the command line.  This
+style is a little more foolproof, and it loses no functionality if you don't
+have pipes.  (You can get this style on Unix too, if you prefer, by defining
+TWO_FILE_COMMANDLINE when you compile the programs; see install.txt.)
+
+You can also say:
+        cjpeg [switches] -outfile jpegfile  imagefile
+or
+        djpeg [switches] -outfile imagefile  jpegfile
+This syntax works on all systems, so it is useful for scripts.
+
+The currently supported image file formats are: PPM (PBMPLUS color format),
+PGM (PBMPLUS grayscale format), BMP, Targa, and RLE (Utah Raster Toolkit
+format).  (RLE is supported only if the URT library is available, which it
+isn't on most non-Unix systems.)  cjpeg recognizes the input image format
+automatically, with the exception of some Targa files.  You have to tell djpeg
+which format to generate.
+
+JPEG files are in the defacto standard JFIF file format.  There are other,
+less widely used JPEG-based file formats, but we don't support them.
+
+All switch names may be abbreviated; for example, -grayscale may be written
+-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
+one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
+British spellings are also accepted (e.g., -greyscale), though for brevity
+these are not mentioned below.
+
+
+CJPEG DETAILS
+
+The basic command line switches for cjpeg are:
+
+        -quality N[,...]  Scale quantization tables to adjust image quality.
+                          Quality is 0 (worst) to 100 (best); default is 75.
+                          (See below for more info.)
+
+        -grayscale      Create monochrome JPEG file from color input.
+                        Be sure to use this switch when compressing a grayscale
+                        BMP file, because cjpeg isn't bright enough to notice
+                        whether a BMP file uses only shades of gray.  By
+                        saying -grayscale, you'll get a smaller JPEG file that
+                        takes less time to process.
+
+        -rgb            Create RGB JPEG file.
+                        Using this switch suppresses the conversion from RGB
+                        colorspace input to the default YCbCr JPEG colorspace.
+
+        -optimize       Perform optimization of entropy encoding parameters.
+                        Without this, default encoding parameters are used.
+                        -optimize usually makes the JPEG file a little smaller,
+                        but cjpeg runs somewhat slower and needs much more
+                        memory.  Image quality and speed of decompression are
+                        unaffected by -optimize.
+
+        -progressive    Create progressive JPEG file (see below).
+
+        -targa          Input file is Targa format.  Targa files that contain
+                        an "identification" field will not be automatically
+                        recognized by cjpeg; for such files you must specify
+                        -targa to make cjpeg treat the input as Targa format.
+                        For most Targa files, you won't need this switch.
+
+The -quality switch lets you trade off compressed file size against quality of
+the reconstructed image: the higher the quality setting, the larger the JPEG
+file, and the closer the output image will be to the original input.  Normally
+you want to use the lowest quality setting (smallest file) that decompresses
+into something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at -quality 75, then go up
+5 or 10 counts at a time until you are happy with the output image.  (The
+optimal setting will vary from one image to another.)
+
+-quality 100 will generate a quantization table of all 1's, minimizing loss
+in the quantization step (but there is still information loss in subsampling,
+as well as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try -quality 2 (or so) for some
+amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
+quantization tables, which are considered optional in the JPEG standard.
+cjpeg emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use -baseline
+if you need to ensure compatibility at low quality values.)
+
+The -quality option has been extended in this version of cjpeg to support
+separate quality settings for luminance and chrominance (or, in general,
+separate settings for every quantization table slot.)  The principle is the
+same as chrominance subsampling:  since the human eye is more sensitive to
+spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+
+The -quality option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.  More (or
+customized) quantization tables can be set with the -qtables option and
+assigned to components with the -qslots option (see the "wizard" switches
+below.)
+
+JPEG  files  generated  with separate luminance and chrominance quality are
+fully compliant with standard JPEG decoders.
+
+CAUTION: For this setting to be useful, be sure to pass an argument of
+-sample 1x1 to cjpeg to disable chrominance subsampling.  Otherwise, the
+default subsampling level (2x2, AKA "4:2:0") will be used.
+
+The -progressive switch creates a "progressive JPEG" file.  In this type of
+JPEG file, the data is stored in multiple scans of increasing quality.  If the
+file is being transmitted over a slow communications link, the decoder can use
+the first scan to display a low-quality image very quickly, and can then
+improve the display with each subsequent scan.  The final image is exactly
+equivalent to a standard JPEG file of the same quality setting, and the total
+file size is about the same --- often a little smaller.
+
+Switches for advanced users:
+
+        -arithmetic     Use arithmetic coding.  CAUTION: arithmetic coded JPEG
+                        is not yet widely implemented, so many decoders will
+                        be unable to view an arithmetic coded JPEG file at
+                        all.
+
+        -dct int        Use integer DCT method (default).
+        -dct fast       Use fast integer DCT (less accurate).
+                        In libjpeg-turbo, the fast method is generally about
+                        5-15% faster than the int method when using the
+                        x86/x86-64 SIMD extensions (results may vary with other
+                        SIMD implementations, or when using libjpeg-turbo
+                        without SIMD extensions.)  For quality levels of 90 and
+                        below, there should be little or no perceptible
+                        difference between the two algorithms.  For quality
+                        levels above 90, however, the difference between
+                        the fast and the int methods becomes more pronounced.
+                        With quality=97, for instance, the fast method incurs
+                        generally about a 1-3 dB loss (in PSNR) relative to
+                        the int method, but this can be larger for some images.
+                        Do not use the fast method with quality levels above
+                        97.  The algorithm often degenerates at quality=98 and
+                        above and can actually produce a more lossy image than
+                        if lower quality levels had been used.  Also, in
+                        libjpeg-turbo, the fast method is not fully accerated
+                        for quality levels above 97, so it will be slower than
+                        the int method.
+        -dct float      Use floating-point DCT method.
+                        The float method is mainly a legacy feature.  It does
+                        not produce significantly more accurate results than
+                        the int method, and it is much slower.  The float
+                        method may also give different results on different
+                        machines due to varying roundoff behavior, whereas the
+                        integer methods should give the same results on all
+                        machines.
+
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+                        -restart 0 (the default) means no restart markers.
+
+        -smooth N       Smooth the input image to eliminate dithering noise.
+                        N, ranging from 1 to 100, indicates the strength of
+                        smoothing.  0 (the default) means no smoothing.
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, temporary files will be used.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+The -restart option inserts extra markers that allow a JPEG decoder to
+resynchronize after a transmission error.  Without restart markers, any damage
+to a compressed file will usually ruin the image from the point of the error
+to the end of the image; with restart markers, the damage is usually confined
+to the portion of the image up to the next restart marker.  Of course, the
+restart markers occupy extra space.  We recommend -restart 1 for images that
+will be transmitted across unreliable networks such as Usenet.
+
+The -smooth option filters the input to eliminate fine-scale noise.  This is
+often useful when converting dithered images to JPEG: a moderate smoothing
+factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
+in a smaller JPEG file and a better-looking image.  Too large a smoothing
+factor will visibly blur the image, however.
+
+Switches for wizards:
+
+        -baseline       Force baseline-compatible quantization tables to be
+                        generated.  This clamps quantization values to 8 bits
+                        even at low quality settings.  (This switch is poorly
+                        named, since it does not ensure that the output is
+                        actually baseline JPEG.  For example, you can use
+                        -baseline and -progressive together.)
+
+        -qtables file   Use the quantization tables given in the specified
+                        text file.
+
+        -qslots N[,...] Select which quantization table to use for each color
+                        component.
+
+        -sample HxV[,...]  Set JPEG sampling factors for each color component.
+
+        -scans file     Use the scan script given in the specified text file.
+
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, DON'T USE THEM.  These switches are documented
+further in the file wizard.txt.
+
+
+DJPEG DETAILS
+
+The basic command line switches for djpeg are:
+
+        -colors N       Reduce image to at most N colors.  This reduces the
+        or -quantize N  number of colors used in the output image, so that it
+                        can be displayed on a colormapped display or stored in
+                        a colormapped file format.  For example, if you have
+                        an 8-bit display, you'd need to reduce to 256 or fewer
+                        colors.  (-colors is the recommended name, -quantize
+                        is provided only for backwards compatibility.)
+
+        -fast           Select recommended processing options for fast, low
+                        quality output.  (The default options are chosen for
+                        highest quality output.)  Currently, this is equivalent
+                        to "-dct fast -nosmooth -onepass -dither ordered".
+
+        -grayscale      Force grayscale output even if JPEG file is color.
+                        Useful for viewing on monochrome displays; also,
+                        djpeg runs noticeably faster in this mode.
+
+        -rgb            Force RGB output even if JPEG file is grayscale.
+
+        -scale M/N      Scale the output image by a factor M/N.  Currently
+                        the scale factor must be M/8, where M is an integer
+                        between 1 and 16 inclusive, or any reduced fraction
+                        thereof (such as 1/2, 3/4, etc.  Scaling is handy if
+                        the image is larger than your screen; also, djpeg runs
+                        much faster when scaling down the output.
+
+        -bmp            Select BMP output format (Windows flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -gif            Select GIF output format.  Since GIF does not support
+                        more than 256 colors, -colors 256 is assumed (unless
+                        you specify a smaller number of colors).  If you
+                        specify -fast, the default number of colors is 216.
+
+        -os2            Select BMP output format (OS/2 1.x flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -pnm            Select PBMPLUS (PPM/PGM) output format (this is the
+                        default format).  PGM is emitted if the JPEG file is
+                        grayscale or if -grayscale is specified; otherwise
+                        PPM is emitted.
+
+        -rle            Select RLE output format.  (Requires URT library.)
+
+        -targa          Select Targa output format.  Grayscale format is
+                        emitted if the JPEG file is grayscale or if
+                        -grayscale is specified; otherwise, colormapped format
+                        is emitted if -colors is specified; otherwise, 24-bit
+                        full-color format is emitted.
+
+Switches for advanced users:
+
+        -dct int        Use integer DCT method (default).
+        -dct fast       Use fast integer DCT (less accurate).
+                        In libjpeg-turbo, the fast method is generally about
+                        5-15% faster than the int method when using the
+                        x86/x86-64 SIMD extensions (results may vary with other
+                        SIMD implementations, or when using libjpeg-turbo
+                        without SIMD extensions.)  If the JPEG image was
+                        compressed using a quality level of 85 or below, then
+                        there should be little or no perceptible difference
+                        between the two algorithms.  When decompressing images
+                        that were compressed using quality levels above 85,
+                        however, the difference between the fast and int
+                        methods becomes more pronounced.  With images
+                        compressed using quality=97, for instance, the fast
+                        method incurs generally about a 4-6 dB loss (in PSNR)
+                        relative to the int method, but this can be larger for
+                        some images.  If you can avoid it, do not use the fast
+                        method when decompressing images that were compressed
+                        using quality levels above 97.  The algorithm often
+                        degenerates for such images and can actually produce
+                        a more lossy output image than if the JPEG image had
+                        been compressed using lower quality levels.
+        -dct float      Use floating-point DCT method.
+                        The float method is mainly a legacy feature.  It does
+                        not produce significantly more accurate results than
+                        the int method, and it is much slower.  The float
+                        method may also give different results on different
+                        machines due to varying roundoff behavior, whereas the
+                        integer methods should give the same results on all
+                        machines.
+
+        -dither fs      Use Floyd-Steinberg dithering in color quantization.
+        -dither ordered Use ordered dithering in color quantization.
+        -dither none    Do not use dithering in color quantization.
+                        By default, Floyd-Steinberg dithering is applied when
+                        quantizing colors; this is slow but usually produces
+                        the best results.  Ordered dither is a compromise
+                        between speed and quality; no dithering is fast but
+                        usually looks awful.  Note that these switches have
+                        no effect unless color quantization is being done.
+                        Ordered dither is only available in -onepass mode.
+
+        -map FILE       Quantize to the colors used in the specified image
+                        file.  This is useful for producing multiple files
+                        with identical color maps, or for forcing a predefined
+                        set of colors to be used.  The FILE must be a GIF
+                        or PPM file.  This option overrides -colors and
+                        -onepass.
+
+        -nosmooth       Use a faster, lower-quality upsampling routine.
+
+        -onepass        Use one-pass instead of two-pass color quantization.
+                        The one-pass method is faster and needs less memory,
+                        but it produces a lower-quality image.  -onepass is
+                        ignored unless you also say -colors N.  Also,
+                        the one-pass method is always used for grayscale
+                        output (the two-pass method is no improvement then).
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, temporary files will be used.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+
+HINTS FOR CJPEG
+
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with cjpeg's -quality and -smooth options
+to get a satisfactory conversion.  -smooth 10 or so is often helpful.
+
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+
+The -optimize option to cjpeg is worth using when you are making a "final"
+version for posting or archiving.  It's also a win when you are using low
+quality settings to make very small JPEG files; the percentage improvement
+is often a lot more than it is on larger files.  (At present, -optimize
+mode is always selected when generating progressive JPEG files.)
+
+Support for GIF input files was removed in cjpeg v6b due to concerns over
+the Unisys LZW patent.  Although this patent expired in 2006, cjpeg still
+lacks GIF support, for these historical reasons.  (Conversion of GIF files to
+JPEG is usually a bad idea anyway.)
+
+
+HINTS FOR DJPEG
+
+To get a quick preview of an image, use the -grayscale and/or -scale switches.
+"-grayscale -scale 1/8" is the fastest case.
+
+Several options are available that trade off image quality to gain speed.
+"-fast" turns on the recommended settings.
+
+"-dct fast" and/or "-nosmooth" gain speed at a small sacrifice in quality.
+When producing a color-quantized image, "-onepass -dither ordered" is fast but
+much lower quality than the default behavior.  "-dither none" may give
+acceptable results in two-pass mode, but is seldom tolerable in one-pass mode.
+
+Two-pass color quantization requires a good deal of memory; on MS-DOS machines
+it may run out of memory even with -maxmemory 0.  In that case you can still
+decompress, with some loss of image quality, by specifying -onepass for
+one-pass quantization.
+
+To avoid the Unisys LZW patent (now expired), djpeg produces uncompressed GIF
+files.  These are larger than they should be, but are readable by standard GIF
+decoders.
+
+
+HINTS FOR BOTH PROGRAMS
+
+If more space is needed than will fit in the available main memory (as
+determined by -maxmemory), temporary files will be used.  (MS-DOS versions
+will try to get extended or expanded memory first.)  The temporary files are
+often rather large: in typical cases they occupy three bytes per pixel, for
+example 3*800*600 = 1.44Mb for an 800x600 image.  If you don't have enough
+free disk space, leave out -progressive and -optimize (for cjpeg) or specify
+-onepass (for djpeg).
+
+On MS-DOS, the temporary files are created in the directory named by the TMP
+or TEMP environment variable, or in the current directory if neither of those
+exist.  Amiga implementations put the temp files in the directory named by
+JPEGTMP:, so be sure to assign JPEGTMP: to a disk partition with adequate free
+space.
+
+The default memory usage limit (-maxmemory) is set when the software is
+compiled.  If you get an "insufficient memory" error, try specifying a smaller
+-maxmemory value, even -maxmemory 0 to use the absolute minimum space.  You
+may want to recompile with a smaller default value if this happens often.
+
+On machines that have "environment" variables, you can define the environment
+variable JPEGMEM to set the default memory limit.  The value is specified as
+described for the -maxmemory switch.  JPEGMEM overrides the default value
+specified when the program was compiled, and itself is overridden by an
+explicit -maxmemory switch.
+
+On MS-DOS machines, -maxmemory is the amount of main (conventional) memory to
+use.  (Extended or expanded memory is also used if available.)  Most
+DOS-specific versions of this software do their own memory space estimation
+and do not need you to specify -maxmemory.
+
+
+JPEGTRAN
+
+jpegtran performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.  For EXIF files and JPEG files
+containing Exif data, you may prefer to use exiftran instead.
+
+jpegtran works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+djpeg followed by cjpeg to accomplish the same conversion.  But by the same
+token, jpegtran cannot perform lossy operations such as changing the image
+quality.  However, while the image data is losslessly transformed, metadata
+can be removed.  See the -copy option for specifics.
+
+jpegtran uses a command line syntax similar to cjpeg or djpeg.
+On Unix-like systems, you say:
+        jpegtran [switches] [inputfile] >outputfile
+On most non-Unix systems, you say:
+        jpegtran [switches] inputfile outputfile
+where both the input and output files are JPEG files.
+
+To specify the coded JPEG representation used in the output file,
+jpegtran accepts a subset of the switches recognized by cjpeg:
+        -optimize       Perform optimization of entropy encoding parameters.
+        -progressive    Create progressive JPEG file.
+        -arithmetic     Use arithmetic coding.
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+        -scans file     Use the scan script given in the specified text file.
+See the previous discussion of cjpeg for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+
+The image can be losslessly transformed by giving one of these switches:
+        -flip horizontal        Mirror image horizontally (left-right).
+        -flip vertical          Mirror image vertically (top-bottom).
+        -rotate 90              Rotate image 90 degrees clockwise.
+        -rotate 180             Rotate image 180 degrees.
+        -rotate 270             Rotate image 270 degrees clockwise (or 90 ccw).
+        -transpose              Transpose image (across UL-to-LR axis).
+        -transverse             Transverse transpose (across UR-to-LL axis).
+
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+
+jpegtran's default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the -trim switch:
+        -trim           Drop non-transformable edge blocks.
+Obviously, a transformation with -trim is not reversible, so strictly speaking
+jpegtran with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
+"-rot 180 -trim" trims both edges.
+
+If you are only interested in perfect transformations, add the -perfect switch:
+        -perfect        Fail with an error if the transformation is not
+                        perfect.
+For example, you may want to do
+  jpegtran -rot 90 -perfect foo.jpg || djpeg foo.jpg | pnmflip -r90 | cjpeg
+to do a perfect rotation, if available, or an approximated one if not.
+
+This version of jpegtran also offers a lossless crop option, which discards
+data outside of a given image region but losslessly preserves what is inside.
+Like the rotate and flip transforms, lossless crop is restricted by the current
+JPEG format; the upper left corner of the selected region must fall on an iMCU
+boundary.  If it doesn't, then it is silently moved up and/or left to the
+nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the output
+image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching an
+'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+        -crop WxH+X+Y   Crop to a rectangular region of width W and height H,
+                        starting at point X,Y.
+
+Other not-strictly-lossless transformation switches are:
+
+        -grayscale      Force grayscale output.
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+
+jpegtran also recognizes these switches that control what to do with "extra"
+markers, such as comment blocks:
+        -copy none      Copy no extra markers from source file.  This setting
+                        suppresses all comments and other metadata in the
+                        source file.
+        -copy comments  Copy only comment markers.  This setting copies
+                        comments from the source file but discards any other
+                        metadata.
+        -copy all       Copy all extra markers.  This setting preserves
+                        miscellaneous markers found in the source file, such
+                        as JFIF thumbnails, Exif data, and Photoshop settings.
+                        In some files, these extra markers can be sizable.
+                        Note that this option will copy thumbnails as-is;
+                        they will not be transformed.
+The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
+jpegtran always did the equivalent of -copy none.)
+
+Additional switches recognized by jpegtran are:
+        -outfile filename
+        -maxmemory N
+        -verbose
+        -debug
+These work the same as in cjpeg or djpeg.
+
+
+THE COMMENT UTILITIES
+
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+
+We provide two utility programs to display COM block contents and add COM
+blocks to a JPEG file.
+
+rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
+standard output.  The command line syntax is
+        rdjpgcom [-raw] [-verbose] [inputfilename]
+The switch "-raw" (or just "-r") causes rdjpgcom to output non-printable
+characters in JPEG comments.  These characters are normally escaped for
+security reasons.
+The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
+image dimensions.  If you omit the input file name from the command line,
+the JPEG file is read from standard input.  (This may not work on some
+operating systems, if binary data can't be read from stdin.)
+
+wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks, but you
+can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
+file; it does not modify the input file.  DO NOT try to overwrite the input
+file by directing wrjpgcom's output back into it; on most systems this will
+just destroy your file.
+
+The command line syntax for wrjpgcom is similar to cjpeg's.  On Unix-like
+systems, it is
+        wrjpgcom [switches] [inputfilename]
+The output file is written to standard output.  The input file comes from
+the named file, or from standard input if no input file is named.
+
+On most non-Unix systems, the syntax is
+        wrjpgcom [switches] inputfilename outputfilename
+where both input and output file names must be given explicitly.
+
+wrjpgcom understands three switches:
+        -replace                 Delete any existing COM blocks from the file.
+        -comment "Comment text"  Supply new COM text on command line.
+        -cfile name              Read text for new COM block from named file.
+(Switch names can be abbreviated.)  If you have only one line of comment text
+to add, you can provide it on the command line with -comment.  The comment
+text must be surrounded with quotes so that it is treated as a single
+argument.  Longer comments can be read from a text file.
+
+If you give neither -comment nor -cfile, then wrjpgcom will read the comment
+text from standard input.  (In this case an input image file name MUST be
+supplied, so that the source JPEG file comes from somewhere else.)  You can
+enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
+(usually control-D or control-Z) to terminate the comment text entry.
+
+wrjpgcom will not add a COM block if the provided comment string is empty.
+Therefore -replace -comment "" can be used to delete all COM blocks from a
+file.
+
+These utility programs do not depend on the IJG JPEG library.  In
+particular, the source code for rdjpgcom is intended as an illustration of
+the minimum amount of code required to parse a JPEG file header correctly.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wizard.txt glmark2-2021.02/android/jni/src/libjpeg-turbo/wizard.txt
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wizard.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wizard.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,211 @@
+Advanced usage instructions for the Independent JPEG Group's JPEG software
+==========================================================================
+
+This file describes cjpeg's "switches for wizards".
+
+The "wizard" switches are intended for experimentation with JPEG by persons
+who are reasonably knowledgeable about the JPEG standard.  If you don't know
+what you are doing, DON'T USE THESE SWITCHES.  You'll likely produce files
+with worse image quality and/or poorer compression than you'd get from the
+default settings.  Furthermore, these switches must be used with caution
+when making files intended for general use, because not all JPEG decoders
+will support unusual JPEG parameter settings.
+
+
+Quantization Table Adjustment
+-----------------------------
+
+Ordinarily, cjpeg starts with a default set of tables (the same ones given
+as examples in the JPEG standard) and scales them up or down according to
+the -quality setting.  The details of the scaling algorithm can be found in
+jcparam.c.  At very low quality settings, some quantization table entries
+can get scaled up to values exceeding 255.  Although 2-byte quantization
+values are supported by the IJG software, this feature is not in baseline
+JPEG and is not supported by all implementations.  If you need to ensure
+wide compatibility of low-quality files, you can constrain the scaled
+quantization values to no more than 255 by giving the -baseline switch.
+Note that use of -baseline will result in poorer quality for the same file
+size, since more bits than necessary are expended on higher AC coefficients.
+
+You can substitute a different set of quantization values by using the
+-qtables switch:
+
+        -qtables file   Use the quantization tables given in the named file.
+
+The specified file should be a text file containing decimal quantization
+values.  The file should contain one to four tables, each of 64 elements.
+The tables are implicitly numbered 0,1,etc. in order of appearance.  Table
+entries appear in normal array order (NOT in the zigzag order in which they
+will be stored in the JPEG file).
+
+Quantization table files are free format, in that arbitrary whitespace can
+appear between numbers.  Also, comments can be included: a comment starts
+with '#' and extends to the end of the line.  Here is an example file that
+duplicates the default quantization tables:
+
+        # Quantization tables given in JPEG spec, section K.1
+
+        # This is table 0 (the luminance table):
+          16  11  10  16  24  40  51  61
+          12  12  14  19  26  58  60  55
+          14  13  16  24  40  57  69  56
+          14  17  22  29  51  87  80  62
+          18  22  37  56  68 109 103  77
+          24  35  55  64  81 104 113  92
+          49  64  78  87 103 121 120 101
+          72  92  95  98 112 100 103  99
+
+        # This is table 1 (the chrominance table):
+          17  18  24  47  99  99  99  99
+          18  21  26  66  99  99  99  99
+          24  26  56  99  99  99  99  99
+          47  66  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+
+If the -qtables switch is used without -quality, then the specified tables
+are used exactly as-is.  If both -qtables and -quality are used, then the
+tables taken from the file are scaled in the same fashion that the default
+tables would be scaled for that quality setting.  If -baseline appears, then
+the quantization values are constrained to the range 1-255.
+
+By default, cjpeg will use quantization table 0 for luminance components and
+table 1 for chrominance components.  To override this choice, use the -qslots
+switch:
+
+        -qslots N[,...]         Select which quantization table to use for
+                                each color component.
+
+The -qslots switch specifies a quantization table number for each color
+component, in the order in which the components appear in the JPEG SOF marker.
+For example, to create a separate table for each of Y,Cb,Cr, you could
+provide a -qtables file that defines three quantization tables and say
+"-qslots 0,1,2".  If -qslots gives fewer table numbers than there are color
+components, then the last table number is repeated as necessary.
+
+
+Sampling Factor Adjustment
+--------------------------
+
+By default, cjpeg uses 2:1 horizontal and vertical downsampling when
+compressing YCbCr data, and no downsampling for all other color spaces.
+You can override this default with the -sample switch:
+
+        -sample HxV[,...]       Set JPEG sampling factors for each color
+                                component.
+
+The -sample switch specifies the JPEG sampling factors for each color
+component, in the order in which they appear in the JPEG SOF marker.
+If you specify fewer HxV pairs than there are components, the remaining
+components are set to 1x1 sampling.  For example, the default YCbCr setting
+is equivalent to "-sample 2x2,1x1,1x1", which can be abbreviated to
+"-sample 2x2".
+
+There are still some JPEG decoders in existence that support only 2x1
+sampling (also called 4:2:2 sampling).  Compatibility with such decoders can
+be achieved by specifying "-sample 2x1".  This is not recommended unless
+really necessary, since it increases file size and encoding/decoding time
+with very little quality gain.
+
+
+Multiple Scan / Progression Control
+-----------------------------------
+
+By default, cjpeg emits a single-scan sequential JPEG file.  The
+-progressive switch generates a progressive JPEG file using a default series
+of progression parameters.  You can create multiple-scan sequential JPEG
+files or progressive JPEG files with custom progression parameters by using
+the -scans switch:
+
+        -scans file     Use the scan sequence given in the named file.
+
+The specified file should be a text file containing a "scan script".
+The script specifies the contents and ordering of the scans to be emitted.
+Each entry in the script defines one scan.  A scan definition specifies
+the components to be included in the scan, and for progressive JPEG it also
+specifies the progression parameters Ss,Se,Ah,Al for the scan.  Scan
+definitions are separated by semicolons (';').  A semicolon after the last
+scan definition is optional.
+
+Each scan definition contains one to four component indexes, optionally
+followed by a colon (':') and the four progressive-JPEG parameters.  The
+component indexes denote which color component(s) are to be transmitted in
+the scan.  Components are numbered in the order in which they appear in the
+JPEG SOF marker, with the first component being numbered 0.  (Note that these
+indexes are not the "component ID" codes assigned to the components, just
+positional indexes.)
+
+The progression parameters for each scan are:
+        Ss      Zigzag index of first coefficient included in scan
+        Se      Zigzag index of last coefficient included in scan
+        Ah      Zero for first scan of a coefficient, else Al of prior scan
+        Al      Successive approximation low bit position for scan
+If the progression parameters are omitted, the values 0,63,0,0 are used,
+producing a sequential JPEG file.  cjpeg automatically determines whether
+the script represents a progressive or sequential file, by observing whether
+Ss and Se values other than 0 and 63 appear.  (The -progressive switch is
+not needed to specify this; in fact, it is ignored when -scans appears.)
+The scan script must meet the JPEG restrictions on progression sequences.
+(cjpeg checks that the spec's requirements are obeyed.)
+
+Scan script files are free format, in that arbitrary whitespace can appear
+between numbers and around punctuation.  Also, comments can be included: a
+comment starts with '#' and extends to the end of the line.  For additional
+legibility, commas or dashes can be placed between values.  (Actually, any
+single punctuation character other than ':' or ';' can be inserted.)  For
+example, the following two scan definitions are equivalent:
+        0 1 2: 0 63 0 0;
+        0,1,2 : 0-63, 0,0 ;
+
+Here is an example of a scan script that generates a partially interleaved
+sequential JPEG file:
+
+        0;                      # Y only in first scan
+        1 2;                    # Cb and Cr in second scan
+
+Here is an example of a progressive scan script using only spectral selection
+(no successive approximation):
+
+        # Interleaved DC scan for Y,Cb,Cr:
+        0,1,2: 0-0,   0, 0 ;
+        # AC scans:
+        0:     1-2,   0, 0 ;    # First two Y AC coefficients
+        0:     3-5,   0, 0 ;    # Three more
+        1:     1-63,  0, 0 ;    # All AC coefficients for Cb
+        2:     1-63,  0, 0 ;    # All AC coefficients for Cr
+        0:     6-9,   0, 0 ;    # More Y coefficients
+        0:     10-63, 0, 0 ;    # Remaining Y coefficients
+
+Here is an example of a successive-approximation script.  This is equivalent
+to the default script used by "cjpeg -progressive" for YCbCr images:
+
+        # Initial DC scan for Y,Cb,Cr (lowest bit not sent)
+        0,1,2: 0-0,   0, 1 ;
+        # First AC scan: send first 5 Y AC coefficients, minus 2 lowest bits:
+        0:     1-5,   0, 2 ;
+        # Send all Cr,Cb AC coefficients, minus lowest bit:
+        # (chroma data is usually too small to be worth subdividing further;
+        #  but note we send Cr first since eye is least sensitive to Cb)
+        2:     1-63,  0, 1 ;
+        1:     1-63,  0, 1 ;
+        # Send remaining Y AC coefficients, minus 2 lowest bits:
+        0:     6-63,  0, 2 ;
+        # Send next-to-lowest bit of all Y AC coefficients:
+        0:     1-63,  2, 1 ;
+        # At this point we've sent all but the lowest bit of all coefficients.
+        # Send lowest bit of DC coefficients
+        0,1,2: 0-0,   1, 0 ;
+        # Send lowest bit of AC coefficients
+        2:     1-63,  1, 0 ;
+        1:     1-63,  1, 0 ;
+        # Y AC lowest bit scan is last; it's usually the largest scan
+        0:     1-63,  1, 0 ;
+
+It may be worth pointing out that this script is tuned for quality settings
+of around 50 to 75.  For lower quality settings, you'd probably want to use
+a script with fewer stages of successive approximation (otherwise the
+initial scans will be really bad).  For higher quality settings, you might
+want to use more stages of successive approximation (so that the initial
+scans are not too large).
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrbmp.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrbmp.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrbmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrbmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,493 @@
+/*
+ * wrbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Microsoft "BMP"
+ * format (MS Windows 3.x and OS/2 1.x flavors).
+ * Either 8-bit colormapped or 24-bit full-color format can be written.
+ * No compression is supported.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jconfigint.h"
+
+#ifdef BMP_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * Since BMP stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * BMP file during finish_output.  The virtual array contains one JSAMPLE per
+ * pixel if the output is grayscale or colormapped, three if it is full color.
+ */
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  boolean is_os2;               /* saves the OS2 format request flag */
+
+  jvirt_sarray_ptr whole_image; /* needed to reverse row order */
+  JDIMENSION data_width;        /* JSAMPLEs per row */
+  JDIMENSION row_width;         /* physical width of one row in the BMP file */
+  int pad_bytes;                /* number of padding bytes needed per row */
+  JDIMENSION cur_output_row;    /* next row# to write to virtual array */
+} bmp_dest_struct;
+
+typedef bmp_dest_struct *bmp_dest_ptr;
+
+
+/* Forward declarations */
+LOCAL(void) write_colormap
+        (j_decompress_ptr cinfo, bmp_dest_ptr dest, int map_colors,
+         int map_entry_size);
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+/* This version is for writing 24-bit pixels */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data.  Note destination values must be in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+
+  if(cinfo->out_color_space == JCS_RGB565) {
+    boolean big_endian = is_big_endian();
+    unsigned short *inptr2 = (unsigned short *)inptr;
+    for (col = cinfo->output_width; col > 0; col--) {
+      if (big_endian) {
+        outptr[0] = (*inptr2 >> 5) & 0xF8;
+        outptr[1] = ((*inptr2 << 5) & 0xE0) | ((*inptr2 >> 11) & 0x1C);
+        outptr[2] = *inptr2 & 0xF8;
+      } else {
+        outptr[0] = (*inptr2 << 3) & 0xF8;
+        outptr[1] = (*inptr2 >> 3) & 0xFC;
+        outptr[2] = (*inptr2 >> 8) & 0xF8;
+      }
+      outptr += 3;
+      inptr2++;
+    }
+  } else {
+    for (col = cinfo->output_width; col > 0; col--) {
+      outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+      outptr[1] = *inptr++;
+      outptr[0] = *inptr++;
+      outptr += 3;
+    }
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* This version is for grayscale OR quantized color output */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data. */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = *inptr++;       /* can omit GETJSAMPLE() safely */
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+
+/*
+ * Startup: normally writes the file header.
+ * In this module we may as well postpone everything until finish_output.
+ */
+
+METHODDEF(void)
+start_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* no work here */
+}
+
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the BMP file.
+ *
+ * First, routines to write the Windows and OS/2 variants of the file header.
+ */
+
+LOCAL(void)
+write_bmp_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write a Windows-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpinfoheader[40];
+#define PUT_2B(array,offset,value)  \
+        (array[offset] = (char) ((value) & 0xFF), \
+         array[offset+1] = (char) (((value) >> 8) & 0xFF))
+#define PUT_4B(array,offset,value)  \
+        (array[offset] = (char) ((value) & 0xFF), \
+         array[offset+1] = (char) (((value) >> 8) & 0xFF), \
+         array[offset+2] = (char) (((value) >> 16) & 0xFF), \
+         array[offset+3] = (char) (((value) >> 24) & 0xFF))
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 40 + cmap_entries * 4; /* Header and colormap */
+  bfSize = headersize + (long) dest->row_width * (long) cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpinfoheader, sizeof(bmpinfoheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPINFOHEADER) */
+  PUT_2B(bmpinfoheader, 0, 40); /* biSize */
+  PUT_4B(bmpinfoheader, 4, cinfo->output_width); /* biWidth */
+  PUT_4B(bmpinfoheader, 8, cinfo->output_height); /* biHeight */
+  PUT_2B(bmpinfoheader, 12, 1); /* biPlanes - must be 1 */
+  PUT_2B(bmpinfoheader, 14, bits_per_pixel); /* biBitCount */
+  /* we leave biCompression = 0, for none */
+  /* we leave biSizeImage = 0; this is correct for uncompressed data */
+  if (cinfo->density_unit == 2) { /* if have density in dots/cm, then */
+    PUT_4B(bmpinfoheader, 24, (long) (cinfo->X_density*100)); /* XPels/M */
+    PUT_4B(bmpinfoheader, 28, (long) (cinfo->Y_density*100)); /* XPels/M */
+  }
+  PUT_2B(bmpinfoheader, 32, cmap_entries); /* biClrUsed */
+  /* we leave biClrImportant = 0 */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpinfoheader, 40) != (size_t) 40)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 4);
+}
+
+
+LOCAL(void)
+write_os2_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write an OS2-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpcoreheader[12];
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 12 + cmap_entries * 3; /* Header and colormap */
+  bfSize = headersize + (long) dest->row_width * (long) cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpcoreheader, sizeof(bmpcoreheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPCOREHEADER) */
+  PUT_2B(bmpcoreheader, 0, 12); /* bcSize */
+  PUT_2B(bmpcoreheader, 4, cinfo->output_width); /* bcWidth */
+  PUT_2B(bmpcoreheader, 6, cinfo->output_height); /* bcHeight */
+  PUT_2B(bmpcoreheader, 8, 1);  /* bcPlanes - must be 1 */
+  PUT_2B(bmpcoreheader, 10, bits_per_pixel); /* bcBitCount */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpcoreheader, 12) != (size_t) 12)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 3);
+}
+
+
+/*
+ * Write the colormap.
+ * Windows uses BGR0 map entries; OS/2 uses BGR entries.
+ */
+
+LOCAL(void)
+write_colormap (j_decompress_ptr cinfo, bmp_dest_ptr dest,
+                int map_colors, int map_entry_size)
+{
+  JSAMPARRAY colormap = cinfo->colormap;
+  int num_colors = cinfo->actual_number_of_colors;
+  FILE *outfile = dest->pub.output_file;
+  int i;
+
+  if (colormap != NULL) {
+    if (cinfo->out_color_components == 3) {
+      /* Normal case with RGB colormap */
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(colormap[2][i]), outfile);
+        putc(GETJSAMPLE(colormap[1][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    } else {
+      /* Grayscale colormap (only happens with grayscale quantization) */
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    }
+  } else {
+    /* If no colormap, must be grayscale data.  Generate a linear "map". */
+    for (i = 0; i < 256; i++) {
+      putc(i, outfile);
+      putc(i, outfile);
+      putc(i, outfile);
+      if (map_entry_size == 4)
+        putc(0, outfile);
+    }
+  }
+  /* Pad colormap with zeros to ensure specified number of colormap entries */
+  if (i > map_colors)
+    ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, i);
+  for (; i < map_colors; i++) {
+    putc(0, outfile);
+    putc(0, outfile);
+    putc(0, outfile);
+    if (map_entry_size == 4)
+      putc(0, outfile);
+  }
+}
+
+
+METHODDEF(void)
+finish_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  register FILE *outfile = dest->pub.output_file;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW data_ptr;
+  JDIMENSION row;
+  register JDIMENSION col;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Write the header and colormap */
+  if (dest->is_os2)
+    write_os2_header(cinfo, dest);
+  else
+    write_bmp_header(cinfo, dest);
+
+  /* Write the file body from our virtual array */
+  for (row = cinfo->output_height; row > 0; row--) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) (cinfo->output_height - row);
+      progress->pub.pass_limit = (long) cinfo->output_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->whole_image, row-1, (JDIMENSION) 1, FALSE);
+    data_ptr = image_ptr[0];
+    for (col = dest->row_width; col > 0; col--) {
+      putc(GETJSAMPLE(*data_ptr), outfile);
+      data_ptr++;
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Make sure we wrote the output file OK */
+  fflush(outfile);
+  if (ferror(outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for BMP format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_bmp (j_decompress_ptr cinfo, boolean is_os2)
+{
+  bmp_dest_ptr dest;
+  JDIMENSION row_width;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (bmp_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(bmp_dest_struct));
+  dest->pub.start_output = start_output_bmp;
+  dest->pub.finish_output = finish_output_bmp;
+  dest->is_os2 = is_os2;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_gray_rows;
+    else
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else if(cinfo->out_color_space == JCS_RGB565 ) {
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else {
+    ERREXIT(cinfo, JERR_BMP_COLORSPACE);
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Determine width of rows in the BMP file (padded to 4-byte boundary). */
+  if (cinfo->out_color_space == JCS_RGB565) {
+    row_width = cinfo->output_width * 2;
+    dest->row_width = dest->data_width = cinfo->output_width * 3;
+  } else {
+    row_width = cinfo->output_width * cinfo->output_components;
+    dest->row_width = dest->data_width = row_width;
+  }
+  while ((dest->row_width & 3) != 0) dest->row_width++;
+  dest->pad_bytes = (int) (dest->row_width - dest->data_width);
+  if (cinfo->out_color_space == JCS_RGB565) {
+    while ((row_width & 3) != 0) row_width++;
+  } else {
+    row_width = dest->row_width;
+  }
+
+
+  /* Allocate space for inversion array, prepare for write pass */
+  dest->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     dest->row_width, cinfo->output_height, (JDIMENSION) 1);
+  dest->cur_output_row = 0;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, row_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* BMP_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrgif.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrgif.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrgif.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrgif.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,402 @@
+/*
+ * wrgif.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in GIF format.
+ *
+ **************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,   *
+ * this code has been modified to output "uncompressed GIF" files.        *
+ * There is no trace of the LZW algorithm in this file.                   *
+ **************************************************************************
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+/*
+ * This code is loosely based on ppmtogif from the PBMPLUS distribution
+ * of Feb. 1991.  That file contains the following copyright notice:
+ *    Based on GIFENCODE by David Rowley <mgardi@watdscu.waterloo.edu>.
+ *    Lempel-Ziv compression based on "compress" by Spencer W. Thomas et al.
+ *    Copyright (C) 1989 by Jef Poskanzer.
+ *    Permission to use, copy, modify, and distribute this software and its
+ *    documentation for any purpose and without fee is hereby granted, provided
+ *    that the above copyright notice appear in all copies and that both that
+ *    copyright notice and this permission notice appear in supporting
+ *    documentation.  This software is provided "as is" without express or
+ *    implied warranty.
+ *
+ * We are also required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  j_decompress_ptr cinfo;       /* back link saves passing separate parm */
+
+  /* State for packing variable-width codes into a bitstream */
+  int n_bits;                   /* current number of bits/code */
+  int maxcode;                  /* maximum code, given n_bits */
+  long cur_accum;               /* holds bits not yet output */
+  int cur_bits;                 /* # of bits in cur_accum */
+
+  /* State for GIF code assignment */
+  int ClearCode;                /* clear code (doesn't change) */
+  int EOFCode;                  /* EOF code (ditto) */
+  int code_counter;             /* counts output symbols */
+
+  /* GIF data packet construction buffer */
+  int bytesinpkt;               /* # of bytes in current packet */
+  char packetbuf[256];          /* workspace for accumulating packet */
+
+} gif_dest_struct;
+
+typedef gif_dest_struct *gif_dest_ptr;
+
+/* Largest value that will fit in N bits */
+#define MAXCODE(n_bits) ((1 << (n_bits)) - 1)
+
+
+/*
+ * Routines to package finished data bytes into GIF data blocks.
+ * A data block consists of a count byte (1..255) and that many data bytes.
+ */
+
+LOCAL(void)
+flush_packet (gif_dest_ptr dinfo)
+/* flush any accumulated data */
+{
+  if (dinfo->bytesinpkt > 0) {  /* never write zero-length packet */
+    dinfo->packetbuf[0] = (char) dinfo->bytesinpkt++;
+    if (JFWRITE(dinfo->pub.output_file, dinfo->packetbuf, dinfo->bytesinpkt)
+        != (size_t) dinfo->bytesinpkt)
+      ERREXIT(dinfo->cinfo, JERR_FILE_WRITE);
+    dinfo->bytesinpkt = 0;
+  }
+}
+
+
+/* Add a character to current packet; flush to disk if necessary */
+#define CHAR_OUT(dinfo,c)  \
+        { (dinfo)->packetbuf[++(dinfo)->bytesinpkt] = (char) (c);  \
+            if ((dinfo)->bytesinpkt >= 255)  \
+              flush_packet(dinfo);  \
+        }
+
+
+/* Routine to convert variable-width codes into a byte stream */
+
+LOCAL(void)
+output (gif_dest_ptr dinfo, int code)
+/* Emit a code of n_bits bits */
+/* Uses cur_accum and cur_bits to reblock into 8-bit bytes */
+{
+  dinfo->cur_accum |= ((long) code) << dinfo->cur_bits;
+  dinfo->cur_bits += dinfo->n_bits;
+
+  while (dinfo->cur_bits >= 8) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+    dinfo->cur_accum >>= 8;
+    dinfo->cur_bits -= 8;
+  }
+}
+
+
+/* The pseudo-compression algorithm.
+ *
+ * In this module we simply output each pixel value as a separate symbol;
+ * thus, no compression occurs.  In fact, there is expansion of one bit per
+ * pixel, because we use a symbol width one bit wider than the pixel width.
+ *
+ * GIF ordinarily uses variable-width symbols, and the decoder will expect
+ * to ratchet up the symbol width after a fixed number of symbols.
+ * To simplify the logic and keep the expansion penalty down, we emit a
+ * GIF Clear code to reset the decoder just before the width would ratchet up.
+ * Thus, all the symbols in the output file will have the same bit width.
+ * Note that emitting the Clear codes at the right times is a mere matter of
+ * counting output symbols and is in no way dependent on the LZW patent.
+ *
+ * With a small basic pixel width (low color count), Clear codes will be
+ * needed very frequently, causing the file to expand even more.  So this
+ * simplistic approach wouldn't work too well on bilevel images, for example.
+ * But for output of JPEG conversions the pixel width will usually be 8 bits
+ * (129 to 256 colors), so the overhead added by Clear symbols is only about
+ * one symbol in every 256.
+ */
+
+LOCAL(void)
+compress_init (gif_dest_ptr dinfo, int i_bits)
+/* Initialize pseudo-compressor */
+{
+  /* init all the state variables */
+  dinfo->n_bits = i_bits;
+  dinfo->maxcode = MAXCODE(dinfo->n_bits);
+  dinfo->ClearCode = (1 << (i_bits - 1));
+  dinfo->EOFCode = dinfo->ClearCode + 1;
+  dinfo->code_counter = dinfo->ClearCode + 2;
+  /* init output buffering vars */
+  dinfo->bytesinpkt = 0;
+  dinfo->cur_accum = 0;
+  dinfo->cur_bits = 0;
+  /* GIF specifies an initial Clear code */
+  output(dinfo, dinfo->ClearCode);
+}
+
+
+LOCAL(void)
+compress_pixel (gif_dest_ptr dinfo, int c)
+/* Accept and "compress" one pixel value.
+ * The given value must be less than n_bits wide.
+ */
+{
+  /* Output the given pixel value as a symbol. */
+  output(dinfo, c);
+  /* Issue Clear codes often enough to keep the reader from ratcheting up
+   * its symbol size.
+   */
+  if (dinfo->code_counter < dinfo->maxcode) {
+    dinfo->code_counter++;
+  } else {
+    output(dinfo, dinfo->ClearCode);
+    dinfo->code_counter = dinfo->ClearCode + 2; /* reset the counter */
+  }
+}
+
+
+LOCAL(void)
+compress_term (gif_dest_ptr dinfo)
+/* Clean up at end */
+{
+  /* Send an EOF code */
+  output(dinfo, dinfo->EOFCode);
+  /* Flush the bit-packing buffer */
+  if (dinfo->cur_bits > 0) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+  }
+  /* Flush the packet buffer */
+  flush_packet(dinfo);
+}
+
+
+/* GIF header construction */
+
+
+LOCAL(void)
+put_word (gif_dest_ptr dinfo, unsigned int w)
+/* Emit a 16-bit word, LSB first */
+{
+  putc(w & 0xFF, dinfo->pub.output_file);
+  putc((w >> 8) & 0xFF, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+put_3bytes (gif_dest_ptr dinfo, int val)
+/* Emit 3 copies of same byte value --- handy subr for colormap construction */
+{
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+emit_header (gif_dest_ptr dinfo, int num_colors, JSAMPARRAY colormap)
+/* Output the GIF file header, including color map */
+/* If colormap==NULL, synthesize a grayscale colormap */
+{
+  int BitsPerPixel, ColorMapSize, InitCodeSize, FlagByte;
+  int cshift = dinfo->cinfo->data_precision - 8;
+  int i;
+
+  if (num_colors > 256)
+    ERREXIT1(dinfo->cinfo, JERR_TOO_MANY_COLORS, num_colors);
+  /* Compute bits/pixel and related values */
+  BitsPerPixel = 1;
+  while (num_colors > (1 << BitsPerPixel))
+    BitsPerPixel++;
+  ColorMapSize = 1 << BitsPerPixel;
+  if (BitsPerPixel <= 1)
+    InitCodeSize = 2;
+  else
+    InitCodeSize = BitsPerPixel;
+  /*
+   * Write the GIF header.
+   * Note that we generate a plain GIF87 header for maximum compatibility.
+   */
+  putc('G', dinfo->pub.output_file);
+  putc('I', dinfo->pub.output_file);
+  putc('F', dinfo->pub.output_file);
+  putc('8', dinfo->pub.output_file);
+  putc('7', dinfo->pub.output_file);
+  putc('a', dinfo->pub.output_file);
+  /* Write the Logical Screen Descriptor */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  FlagByte = 0x80;              /* Yes, there is a global color table */
+  FlagByte |= (BitsPerPixel-1) << 4; /* color resolution */
+  FlagByte |= (BitsPerPixel-1); /* size of global color table */
+  putc(FlagByte, dinfo->pub.output_file);
+  putc(0, dinfo->pub.output_file); /* Background color index */
+  putc(0, dinfo->pub.output_file); /* Reserved (aspect ratio in GIF89) */
+  /* Write the Global Color Map */
+  /* If the color map is more than 8 bits precision, */
+  /* we reduce it to 8 bits by shifting */
+  for (i=0; i < ColorMapSize; i++) {
+    if (i < num_colors) {
+      if (colormap != NULL) {
+        if (dinfo->cinfo->out_color_space == JCS_RGB) {
+          /* Normal case: RGB color map */
+          putc(GETJSAMPLE(colormap[0][i]) >> cshift, dinfo->pub.output_file);
+          putc(GETJSAMPLE(colormap[1][i]) >> cshift, dinfo->pub.output_file);
+          putc(GETJSAMPLE(colormap[2][i]) >> cshift, dinfo->pub.output_file);
+        } else {
+          /* Grayscale "color map": possible if quantizing grayscale image */
+          put_3bytes(dinfo, GETJSAMPLE(colormap[0][i]) >> cshift);
+        }
+      } else {
+        /* Create a grayscale map of num_colors values, range 0..255 */
+        put_3bytes(dinfo, (i * 255 + (num_colors-1)/2) / (num_colors-1));
+      }
+    } else {
+      /* fill out the map to a power of 2 */
+      put_3bytes(dinfo, 0);
+    }
+  }
+  /* Write image separator and Image Descriptor */
+  putc(',', dinfo->pub.output_file); /* separator */
+  put_word(dinfo, 0);           /* left/top offset */
+  put_word(dinfo, 0);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width); /* image size */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  /* flag byte: not interlaced, no local color map */
+  putc(0x00, dinfo->pub.output_file);
+  /* Write Initial Code Size byte */
+  putc(InitCodeSize, dinfo->pub.output_file);
+
+  /* Initialize for "compression" of image data */
+  compress_init(dinfo, InitCodeSize+1);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  if (cinfo->quantize_colors)
+    emit_header(dest, cinfo->actual_number_of_colors, cinfo->colormap);
+  else
+    emit_header(dest, 256, (JSAMPARRAY) NULL);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    compress_pixel(dest, GETJSAMPLE(*ptr++));
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  /* Flush "compression" mechanism */
+  compress_term(dest);
+  /* Write a zero-length data block to end the series */
+  putc(0, dest->pub.output_file);
+  /* Write the GIF terminator mark */
+  putc(';', dest->pub.output_file);
+  /* Make sure we wrote the output file OK */
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for GIF format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_gif (j_decompress_ptr cinfo)
+{
+  gif_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (gif_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(gif_dest_struct));
+  dest->cinfo = cinfo;          /* make back link for subroutines */
+  dest->pub.start_output = start_output_gif;
+  dest->pub.put_pixel_rows = put_pixel_rows;
+  dest->pub.finish_output = finish_output_gif;
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_GIF_COLORSPACE);
+
+  /* Force quantization if color or if > 8 bits input */
+  if (cinfo->out_color_space != JCS_GRAYSCALE || cinfo->data_precision > 8) {
+    /* Force quantization to at most 256 colors */
+    cinfo->quantize_colors = TRUE;
+    if (cinfo->desired_number_of_colors > 256)
+      cinfo->desired_number_of_colors = 256;
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  if (cinfo->output_components != 1) /* safety check: just one component? */
+    ERREXIT(cinfo, JERR_GIF_BUG);
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, cinfo->output_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* GIF_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrjpgcom.1 glmark2-2021.02/android/jni/src/libjpeg-turbo/wrjpgcom.1
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrjpgcom.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrjpgcom.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,103 @@
+.TH WRJPGCOM 1 "15 June 1995"
+.SH NAME
+wrjpgcom \- insert text comments into a JPEG file
+.SH SYNOPSIS
+.B wrjpgcom
+[
+.B \-replace
+]
+[
+.BI \-comment " text"
+]
+[
+.BI \-cfile " name"
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B wrjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and generates a new JPEG/JFIF file on standard output.  A comment block is
+added to the file.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.PP
+.B wrjpgcom
+adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks; but you
+can delete the old COM blocks if you wish.
+.SH OPTIONS
+Switch names may be abbreviated, and are not case sensitive.
+.TP
+.B \-replace
+Delete any existing COM blocks from the file.
+.TP
+.BI \-comment " text"
+Supply text for new COM block on command line.
+.TP
+.BI \-cfile " name"
+Read text for new COM block from named file.
+.PP
+If you have only one line of comment text to add, you can provide it on the
+command line with
+.BR \-comment .
+The comment text must be surrounded with quotes so that it is treated as a
+single argument.  Longer comments can be read from a text file.
+.PP
+If you give neither
+.B \-comment
+nor
+.BR \-cfile ,
+then
+.B wrjpgcom
+will read the comment text from standard input.  (In this case an input image
+file name MUST be supplied, so that the source JPEG file comes from somewhere
+else.)  You can enter multiple lines, up to 64KB worth.  Type an end-of-file
+indicator (usually control-D) to terminate the comment text entry.
+.PP
+.B wrjpgcom
+will not add a COM block if the provided comment string is empty.  Therefore
+\fB\-replace \-comment ""\fR can be used to delete all COM blocks from a file.
+.SH EXAMPLES
+.LP
+Add a short comment to in.jpg, producing out.jpg:
+.IP
+.B wrjpgcom \-c
+\fI"View of my back yard" in.jpg
+.B >
+.I out.jpg
+.PP
+Attach a long comment previously stored in comment.txt:
+.IP
+.B wrjpgcom
+.I in.jpg
+.B <
+.I comment.txt
+.B >
+.I out.jpg
+.PP
+or equivalently
+.IP
+.B wrjpgcom
+.B -cfile
+.I comment.txt
+.B <
+.I in.jpg
+.B >
+.I out.jpg
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrjpgcom.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrjpgcom.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrjpgcom.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrjpgcom.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,592 @@
+/*
+ * wrjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that inserts
+ * user-supplied text as a COM (comment) marker in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc() */
+extern void *malloc ();
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+/* Reduce this value if your malloc() can't allocate blocks up to 64K.
+ * On DOS, compiling in large model is usually a better solution.
+ */
+
+#ifndef MAX_COM_LENGTH
+#define MAX_COM_LENGTH 65000L   /* must be <= 65533 in any case */
+#endif
+
+
+/*
+ * These macros are used to read the input file and write the output file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+static FILE *outfile;           /* output JPEG file */
+
+/* Emit an output byte */
+#define PUTBYTE(x)  putc((x), outfile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/* Routines to write data to output file */
+
+static void
+write_1_byte (int c)
+{
+  PUTBYTE(c);
+}
+
+static void
+write_2_bytes (unsigned int val)
+{
+  PUTBYTE((val >> 8) & 0xFF);
+  PUTBYTE(val & 0xFF);
+}
+
+static void
+write_marker (int marker)
+{
+  PUTBYTE(0xFF);
+  PUTBYTE(marker);
+}
+
+static void
+copy_rest_of_file (void)
+{
+  int c;
+
+  while ((c = NEXTBYTE()) != EOF)
+    PUTBYTE(c);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0            /* Start Of Frame N */
+#define M_SOF1  0xC1            /* N indicates which compression process */
+#define M_SOF2  0xC2            /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5            /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8            /* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9            /* End Of Image (end of datastream) */
+#define M_SOS   0xDA            /* Start Of Scan (begins compressed data) */
+#define M_COM   0xFE            /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.  (Padding FFs will NOT be replicated in the output file.)
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+copy_variable (void)
+/* Copy an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  write_2_bytes(length);
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    write_1_byte(read_1_byte());
+    length--;
+  }
+}
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOFn or EOI is seen;
+ * copy data to output, but discard COM markers unless keep_COM is true.
+ */
+
+static int
+scan_JPEG_header (int keep_COM)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+  write_marker(M_SOI);
+
+  /* Scan miscellaneous markers until we reach SOFn. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      return marker;
+
+    case M_SOS:                 /* should not see compressed data before SOF */
+      ERREXIT("SOS without prior SOFn");
+      break;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:                 /* Existing COM: conditionally discard */
+      if (keep_COM) {
+        write_marker(marker);
+        copy_variable();
+      } else {
+        skip_variable();
+      }
+      break;
+
+    default:                    /* Anything else just gets copied */
+      write_marker(marker);
+      copy_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "wrjpgcom inserts a textual comment in a JPEG file.\n");
+  fprintf(stderr, "You can add to or replace any existing comment(s).\n");
+
+  fprintf(stderr, "Usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -replace         Delete any existing comments\n");
+  fprintf(stderr, "  -comment \"text\"  Insert comment with given text\n");
+  fprintf(stderr, "  -cfile name      Read comment from named file\n");
+  fprintf(stderr, "Notice that you must put quotes around the comment text\n");
+  fprintf(stderr, "when you use -comment.\n");
+  fprintf(stderr, "If you do not give either -comment or -cfile on the command line,\n");
+  fprintf(stderr, "then the comment text is read from standard input.\n");
+  fprintf(stderr, "It can be multiple lines, up to %u characters total.\n",
+          (unsigned int) MAX_COM_LENGTH);
+#ifndef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "You must specify an input JPEG file name when supplying\n");
+  fprintf(stderr, "comment text from standard input.\n");
+#endif
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int keep_COM = 1;
+  char *comment_arg = NULL;
+  FILE *comment_file = NULL;
+  unsigned int comment_length = 0;
+  int marker;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "wrjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "replace", 1)) {
+      keep_COM = 0;
+    } else if (keymatch(arg, "cfile", 2)) {
+      if (++argn >= argc) usage();
+      if ((comment_file = fopen(argv[argn], "r")) == NULL) {
+        fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+        exit(EXIT_FAILURE);
+      }
+    } else if (keymatch(arg, "comment", 1)) {
+      if (++argn >= argc) usage();
+      comment_arg = argv[argn];
+      /* If the comment text starts with '"', then we are probably running
+       * under MS-DOG and must parse out the quoted string ourselves.  Sigh.
+       */
+      if (comment_arg[0] == '"') {
+        comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+        if (comment_arg == NULL)
+          ERREXIT("Insufficient memory");
+        if (strlen(argv[argn]) + 2 >= (size_t) MAX_COM_LENGTH) {
+          fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                  (unsigned int) MAX_COM_LENGTH);
+          exit(EXIT_FAILURE);
+        }
+        strcpy(comment_arg, argv[argn]+1);
+        for (;;) {
+          comment_length = (unsigned int) strlen(comment_arg);
+          if (comment_length > 0 && comment_arg[comment_length-1] == '"') {
+            comment_arg[comment_length-1] = '\0'; /* zap terminating quote */
+            break;
+          }
+          if (++argn >= argc)
+            ERREXIT("Missing ending quote mark");
+          if (strlen(comment_arg) + strlen(argv[argn]) + 2 >=
+              (size_t) MAX_COM_LENGTH) {
+            fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                    (unsigned int) MAX_COM_LENGTH);
+            exit(EXIT_FAILURE);
+          }
+          strcat(comment_arg, " ");
+          strcat(comment_arg, argv[argn]);
+        }
+      } else if (strlen(argv[argn]) >= (size_t) MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int) MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_length = (unsigned int) strlen(comment_arg);
+    } else
+      usage();
+  }
+
+  /* Cannot use both -comment and -cfile. */
+  if (comment_arg != NULL && comment_file != NULL)
+    usage();
+  /* If there is neither -comment nor -cfile, we will read the comment text
+   * from stdin; in this case there MUST be an input JPEG file name.
+   */
+  if (comment_arg == NULL && comment_file == NULL && argn >= argc)
+    usage();
+
+  /* Open the input file. */
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Open the output file. */
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have explicit output file name */
+  if (argn != argc-2) {
+    fprintf(stderr, "%s: must name one input and one output file\n",
+            progname);
+    usage();
+  }
+  if ((outfile = fopen(argv[argn+1], WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open %s\n", progname, argv[argn+1]);
+    exit(EXIT_FAILURE);
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  /* default output file is stdout */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((outfile = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open stdout\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#else
+  outfile = stdout;
+#endif
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Collect comment text from comment_file or stdin, if necessary */
+  if (comment_arg == NULL) {
+    FILE *src_file;
+    int c;
+
+    comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+    if (comment_arg == NULL)
+      ERREXIT("Insufficient memory");
+    comment_length = 0;
+    src_file = (comment_file != NULL ? comment_file : stdin);
+    while ((c = getc(src_file)) != EOF) {
+      if (comment_length >= (unsigned int) MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int) MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_arg[comment_length++] = (char) c;
+    }
+    if (comment_file != NULL)
+      fclose(comment_file);
+  }
+
+  /* Copy JPEG headers until SOFn marker;
+   * we will insert the new comment marker just before SOFn.
+   * This (a) causes the new comment to appear after, rather than before,
+   * existing comments; and (b) ensures that comments come after any JFIF
+   * or JFXX markers, as required by the JFIF specification.
+   */
+  marker = scan_JPEG_header(keep_COM);
+  /* Insert the new COM marker, but only if nonempty text has been supplied */
+  if (comment_length > 0) {
+    write_marker(M_COM);
+    write_2_bytes(comment_length + 2);
+    while (comment_length > 0) {
+      write_1_byte(*comment_arg++);
+      comment_length--;
+    }
+  }
+  /* Duplicate the remainder of the source file.
+   * Note that any COM markers occuring after SOF will not be touched.
+   */
+  write_marker(marker);
+  copy_rest_of_file();
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrppm.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrppm.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrppm.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrppm.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,252 @@
+/*
+ * wrppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * Modified 2009 by Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "wrppm.h"
+
+#ifdef PPM_SUPPORTED
+
+
+/*
+ * For 12-bit JPEG data, we either downscale the values to 8 bits
+ * (to write standard byte-per-sample PPM/PGM files), or output
+ * nonstandard word-per-sample PPM/PGM files.  Downscaling is done
+ * if PPM_NORAWWORD is defined (this can be done in the Makefile
+ * or in jconfig.h).
+ * (When the core library supports data precision reduction, a cleaner
+ * implementation will be to ask for that instead.)
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) (v)
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+#ifdef PPM_NORAWWORD
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) ((v) >> (BITS_IN_JSAMPLE-8))
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+/* The word-per-sample format always puts the MSB first. */
+#define PUTPPMSAMPLE(ptr,v)                     \
+        { register int val_ = v;                \
+          *ptr++ = (char) ((val_ >> 8) & 0xFF); \
+          *ptr++ = (char) (val_ & 0xFF);        \
+        }
+#define BYTESPERSAMPLE 2
+#define PPM_MAXVAL ((1<<BITS_IN_JSAMPLE)-1)
+#endif
+#endif
+
+
+/*
+ * When JSAMPLE is the same size as char, we can just fwrite() the
+ * decompressed data to the PPM or PGM file.
+ */
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ *
+ * put_pixel_rows handles the "normal" 8-bit case where the decompressor
+ * output buffer is physically the same as the fwrite buffer.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * This code is used when we have to copy the data and apply a pixel
+ * format translation.  Typically this only happens in 12-bit mode.
+ */
+
+METHODDEF(void)
+copy_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                 JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = dest->samples_per_row; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(*ptr++));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some pixel data when color quantization is in effect.
+ * We have to demap the color index values to straight data.
+ */
+
+METHODDEF(void)
+put_demapped_rgb (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                  JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register int pixval;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JSAMPROW color_map1 = cinfo->colormap[1];
+  register JSAMPROW color_map2 = cinfo->colormap[2];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    pixval = GETJSAMPLE(*ptr++);
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map0[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map1[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map2[pixval]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map[GETJSAMPLE(*ptr++)]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  /* Emit file header */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    /* emit header for raw PGM format */
+    fprintf(dest->pub.output_file, "P5\n%ld %ld\n%d\n",
+            (long) cinfo->output_width, (long) cinfo->output_height,
+            PPM_MAXVAL);
+    break;
+  case JCS_RGB:
+    /* emit header for raw PPM format */
+    fprintf(dest->pub.output_file, "P6\n%ld %ld\n%d\n",
+            (long) cinfo->output_width, (long) cinfo->output_height,
+            PPM_MAXVAL);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for PPM format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_ppm (j_decompress_ptr cinfo)
+{
+  ppm_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (ppm_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(ppm_dest_struct));
+  dest->pub.start_output = start_output_ppm;
+  dest->pub.finish_output = finish_output_ppm;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create physical I/O buffer */
+  dest->samples_per_row = cinfo->output_width * cinfo->out_color_components;
+  dest->buffer_width = dest->samples_per_row * (BYTESPERSAMPLE * sizeof(char));
+  dest->iobuffer = (char *) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width);
+
+  if (cinfo->quantize_colors || BITS_IN_JSAMPLE != 8 ||
+      sizeof(JSAMPLE) != sizeof(char)) {
+    /* When quantizing, we need an output buffer for colormap indexes
+     * that's separate from the physical I/O buffer.  We also need a
+     * separate buffer if pixel format translation must take place.
+     */
+    dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       cinfo->output_width * cinfo->output_components, (JDIMENSION) 1);
+    dest->pub.buffer_height = 1;
+    if (! cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = copy_pixel_rows;
+    else if (cinfo->out_color_space == JCS_GRAYSCALE)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_demapped_rgb;
+  } else {
+    /* We will fwrite() directly from decompressor output buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    dest->pixrow = (JSAMPROW) dest->iobuffer;
+    dest->pub.buffer = & dest->pixrow;
+    dest->pub.buffer_height = 1;
+    dest->pub.put_pixel_rows = put_pixel_rows;
+  }
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* PPM_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrppm.h glmark2-2021.02/android/jni/src/libjpeg-turbo/wrppm.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrppm.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrppm.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,26 @@
+/*
+ * wrppm.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#ifdef PPM_SUPPORTED
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  char *iobuffer;               /* fwrite's I/O buffer */
+  JSAMPROW pixrow;              /* decompressor output buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JDIMENSION samples_per_row;   /* JSAMPLEs per output row */
+} ppm_dest_struct;
+
+typedef ppm_dest_struct *ppm_dest_ptr;
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrrle.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrrle.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrrle.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrrle.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,307 @@
+/*
+ * wrrle.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * RLE file during finish_output.
+ */
+
+
+/*
+ * For now, if we emit an RLE color map then it is always 256 entries long,
+ * though not all of the entries need be used.
+ */
+
+#define CMAPBITS        8
+#define CMAPLENGTH      (1<<(CMAPBITS))
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  jvirt_sarray_ptr image;       /* virtual array to store the output image */
+  rle_map *colormap;            /* RLE-style color map, or NULL if none */
+  rle_pixel **rle_row;          /* To pass rows to rle_putrow() */
+
+} rle_dest_struct;
+
+typedef rle_dest_struct *rle_dest_ptr;
+
+/* Forward declarations */
+METHODDEF(void) rle_put_pixel_rows
+        (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+         JDIMENSION rows_supplied);
+
+
+/*
+ * Write the file header.
+ *
+ * In this module it's easier to wait till finish_output to write anything.
+ */
+
+METHODDEF(void)
+start_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  size_t cmapsize;
+  int i, ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /*
+   * Make sure the image can be stored in RLE format.
+   *
+   * - RLE stores image dimensions as *signed* 16 bit integers.  JPEG
+   *   uses unsigned, so we have to check the width.
+   *
+   * - Colorspace is expected to be grayscale or RGB.
+   *
+   * - The number of channels (components) is expected to be 1 (grayscale/
+   *   pseudocolor) or 3 (truecolor/directcolor).
+   *   (could be 2 or 4 if using an alpha channel, but we aren't)
+   */
+
+  if (cinfo->output_width > 32767 || cinfo->output_height > 32767)
+    ERREXIT2(cinfo, JERR_RLE_DIMENSIONS, cinfo->output_width,
+             cinfo->output_height);
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_RLE_COLORSPACE);
+
+  if (cinfo->output_components != 1 && cinfo->output_components != 3)
+    ERREXIT1(cinfo, JERR_RLE_TOOMANYCHANNELS, cinfo->num_components);
+
+  /* Convert colormap, if any, to RLE format. */
+
+  dest->colormap = NULL;
+
+  if (cinfo->quantize_colors) {
+    /* Allocate storage for RLE-style cmap, zero any extra entries */
+    cmapsize = cinfo->out_color_components * CMAPLENGTH * sizeof(rle_map);
+    dest->colormap = (rle_map *) (*cinfo->mem->alloc_small)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, cmapsize);
+    MEMZERO(dest->colormap, cmapsize);
+
+    /* Save away data in RLE format --- note 8-bit left shift! */
+    /* Shifting would need adjustment for JSAMPLEs wider than 8 bits. */
+    for (ci = 0; ci < cinfo->out_color_components; ci++) {
+      for (i = 0; i < cinfo->actual_number_of_colors; i++) {
+        dest->colormap[ci * CMAPLENGTH + i] =
+          GETJSAMPLE(cinfo->colormap[ci][i]) << 8;
+      }
+    }
+  }
+
+  /* Set the output buffer to the first row */
+  dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->image, (JDIMENSION) 0, (JDIMENSION) 1, TRUE);
+  dest->pub.buffer_height = 1;
+
+  dest->pub.put_pixel_rows = rle_put_pixel_rows;
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->total_extra_passes++;  /* count file writing as separate pass */
+  }
+#endif
+}
+
+
+/*
+ * Write some pixel data.
+ *
+ * This routine just saves the data away in a virtual array.
+ */
+
+METHODDEF(void)
+rle_put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                    JDIMENSION rows_supplied)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+
+  if (cinfo->output_scanline < cinfo->output_height) {
+    dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->image,
+       cinfo->output_scanline, (JDIMENSION) 1, TRUE);
+  }
+}
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the RLE file.
+ */
+
+METHODDEF(void)
+finish_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  rle_hdr header;               /* Output file information */
+  rle_pixel **rle_row, *red, *green, *blue;
+  JSAMPROW output_row;
+  char cmapcomment[80];
+  int row, col;
+  int ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Initialize the header info */
+  header = *rle_hdr_init(NULL);
+  header.rle_file = dest->pub.output_file;
+  header.xmin     = 0;
+  header.xmax     = cinfo->output_width  - 1;
+  header.ymin     = 0;
+  header.ymax     = cinfo->output_height - 1;
+  header.alpha    = 0;
+  header.ncolors  = cinfo->output_components;
+  for (ci = 0; ci < cinfo->output_components; ci++) {
+    RLE_SET_BIT(header, ci);
+  }
+  if (cinfo->quantize_colors) {
+    header.ncmap   = cinfo->out_color_components;
+    header.cmaplen = CMAPBITS;
+    header.cmap    = dest->colormap;
+    /* Add a comment to the output image with the true colormap length. */
+    sprintf(cmapcomment, "color_map_length=%d", cinfo->actual_number_of_colors);
+    rle_putcom(cmapcomment, &header);
+  }
+
+  /* Emit the RLE header and color map (if any) */
+  rle_put_setup(&header);
+
+  /* Now output the RLE data from our virtual array.
+   * We assume here that rle_pixel is represented the same as JSAMPLE.
+   */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->output_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  if (cinfo->output_components == 1) {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+         (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  } else {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) dest->rle_row;
+      output_row = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+         (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      red = rle_row[0];
+      green = rle_row[1];
+      blue = rle_row[2];
+      for (col = cinfo->output_width; col > 0; col--) {
+        *red++ = GETJSAMPLE(*output_row++);
+        *green++ = GETJSAMPLE(*output_row++);
+        *blue++ = GETJSAMPLE(*output_row++);
+      }
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Emit file trailer */
+  rle_puteof(&header);
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for RLE format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_rle (j_decompress_ptr cinfo)
+{
+  rle_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (rle_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(rle_dest_struct));
+  dest->pub.start_output = start_output_rle;
+  dest->pub.finish_output = finish_output_rle;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Allocate a work array for output to the RLE library. */
+  dest->rle_row = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     cinfo->output_width, (JDIMENSION) cinfo->output_components);
+
+  /* Allocate a virtual array to hold the image. */
+  dest->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (cinfo->output_width * cinfo->output_components),
+     cinfo->output_height, (JDIMENSION) 1);
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* RLE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrtarga.c glmark2-2021.02/android/jni/src/libjpeg-turbo/wrtarga.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libjpeg-turbo/wrtarga.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libjpeg-turbo/wrtarga.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,247 @@
+/*
+ * wrtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  char *iobuffer;               /* physical I/O buffer */
+  JDIMENSION buffer_width;      /* width of one row */
+} tga_dest_struct;
+
+typedef tga_dest_struct *tga_dest_ptr;
+
+
+LOCAL(void)
+write_header (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo, int num_colors)
+/* Create and write a Targa header */
+{
+  char targaheader[18];
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(targaheader, sizeof(targaheader));
+
+  if (num_colors > 0) {
+    targaheader[1] = 1;         /* color map type 1 */
+    targaheader[5] = (char) (num_colors & 0xFF);
+    targaheader[6] = (char) (num_colors >> 8);
+    targaheader[7] = 24;        /* 24 bits per cmap entry */
+  }
+
+  targaheader[12] = (char) (cinfo->output_width & 0xFF);
+  targaheader[13] = (char) (cinfo->output_width >> 8);
+  targaheader[14] = (char) (cinfo->output_height & 0xFF);
+  targaheader[15] = (char) (cinfo->output_height >> 8);
+  targaheader[17] = 0x20;       /* Top-down, non-interlaced */
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    targaheader[2] = 3;         /* image type = uncompressed grayscale */
+    targaheader[16] = 8;        /* bits per pixel */
+  } else {                      /* must be RGB */
+    if (num_colors > 0) {
+      targaheader[2] = 1;       /* image type = colormapped RGB */
+      targaheader[16] = 8;
+    } else {
+      targaheader[2] = 2;       /* image type = uncompressed RGB */
+      targaheader[16] = 24;
+    }
+  }
+
+  if (JFWRITE(dinfo->output_file, targaheader, 18) != (size_t) 18)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+/* used for unquantized full-color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    outptr[0] = (char) GETJSAMPLE(inptr[2]); /* RGB to BGR order */
+    outptr[1] = (char) GETJSAMPLE(inptr[1]);
+    outptr[2] = (char) GETJSAMPLE(inptr[0]);
+    inptr += 3, outptr += 3;
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* used for grayscale OR quantized color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(*inptr++);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some demapped pixel data when color quantization is in effect.
+ * For Targa, this is only applied to grayscale data.
+ */
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(color_map0[GETJSAMPLE(*inptr++)]);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  int num_colors, i;
+  FILE *outfile;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    /* Targa doesn't have a mapped grayscale format, so we will */
+    /* demap quantized gray output.  Never emit a colormap. */
+    write_header(cinfo, dinfo, 0);
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* We only support 8-bit colormap indexes, so only 256 colors */
+      num_colors = cinfo->actual_number_of_colors;
+      if (num_colors > 256)
+        ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, num_colors);
+      write_header(cinfo, dinfo, num_colors);
+      /* Write the colormap.  Note Targa uses BGR byte order */
+      outfile = dest->pub.output_file;
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(cinfo->colormap[2][i]), outfile);
+        putc(GETJSAMPLE(cinfo->colormap[1][i]), outfile);
+        putc(GETJSAMPLE(cinfo->colormap[0][i]), outfile);
+      }
+      dest->pub.put_pixel_rows = put_gray_rows;
+    } else {
+      write_header(cinfo, dinfo, 0);
+      dest->pub.put_pixel_rows = put_pixel_rows;
+    }
+  } else {
+    ERREXIT(cinfo, JERR_TGA_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for Targa format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_targa (j_decompress_ptr cinfo)
+{
+  tga_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (tga_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(tga_dest_struct));
+  dest->pub.start_output = start_output_tga;
+  dest->pub.finish_output = finish_output_tga;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create I/O buffer. */
+  dest->buffer_width = cinfo->output_width * cinfo->output_components;
+  dest->iobuffer = (char *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (size_t) (dest->buffer_width * sizeof(char)));
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* TARGA_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/log.cc glmark2-2021.02/android/jni/src/libmatrix/log.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/log.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/log.cc	2021-04-25 01:47:22.000000000 +0800
@@ -10,7 +10,6 @@
 //     Alexandros Frantzis <alexandros.frantzis@linaro.org>
 //     Jesse Barker <jesse.barker@linaro.org>
 //
-#include <unistd.h>
 #include <cstdio>
 #include <cstdarg>
 #include <string>
@@ -22,6 +21,13 @@
 #include <android/log.h>
 #endif
 
+#ifdef _WIN32
+// On windows 'isatty' is found in <io.h>
+#include <io.h>
+#else
+#include <unistd.h>
+#endif
+
 using std::string;
 
 const string Log::continuation_prefix("\x10");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/mat.h glmark2-2021.02/android/jni/src/libmatrix/mat.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/mat.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/mat.h	2021-04-25 01:47:22.000000000 +0800
@@ -109,9 +109,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat2& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
@@ -411,9 +408,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat3& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
@@ -793,9 +787,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat4& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/program.cc glmark2-2021.02/android/jni/src/libmatrix/program.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/program.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/program.cc	2021-04-25 01:47:22.000000000 +0800
@@ -190,7 +190,7 @@
     }
 
     shader.attach(handle_);
-    shaders_.push_back(shader);
+    shaders_.push_back(std::move(shader));
     return;
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/program.h glmark2-2021.02/android/jni/src/libmatrix/program.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/program.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/program.h	2021-04-25 01:47:22.000000000 +0800
@@ -35,6 +35,13 @@
         message_(shader.message_),
         ready_(shader.ready_),
         valid_(shader.valid_) {}
+    Shader(Shader&& shader) noexcept:
+        handle_(std::exchange(shader.handle_, 0)),
+        type_(std::exchange(shader.type_, 0)),
+        source_(std::move(shader.source_)),
+        message_(std::move(shader.message_)),
+        ready_(std::exchange(shader.ready_, false)),
+        valid_(std::exchange(shader.valid_, false)) {}
     Shader(unsigned int type, const std::string& source);
     ~Shader();
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/shader-source.cc glmark2-2021.02/android/jni/src/libmatrix/shader-source.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/shader-source.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/shader-source.cc	2021-04-25 01:47:22.000000000 +0800
@@ -18,6 +18,36 @@
 #include "vec.h"
 #include "util.h"
 
+namespace
+{
+
+bool is_valid_precision_value(ShaderSource::PrecisionValue precision_value)
+{
+    switch(precision_value) {
+        case ShaderSource::PrecisionValueLow:
+        case ShaderSource::PrecisionValueMedium:
+        case ShaderSource::PrecisionValueHigh:
+        case ShaderSource::PrecisionValueDefault:
+            return true;
+        default:
+            return false;
+    }
+}
+
+bool is_valid_shader_type(ShaderSource::ShaderType shader_type)
+{
+    switch(shader_type) {
+        case ShaderSource::ShaderTypeVertex:
+        case ShaderSource::ShaderTypeFragment:
+        case ShaderSource::ShaderTypeUnknown:
+            return true;
+        default:
+            return false;
+    }
+}
+
+}
+
 /**
  * Holds default precision values for all shader types
  * (even the unknown type, which is hardwired to default precision values)
@@ -34,7 +64,7 @@
 bool
 ShaderSource::load_file(const std::string& filename, std::string& str)
 {
-    std::auto_ptr<std::istream> is_ptr(Util::get_resource(filename));
+    std::unique_ptr<std::istream> is_ptr(Util::get_resource(filename));
     std::istream& inputFile(*is_ptr);
 
     if (!inputFile)
@@ -421,7 +451,9 @@
             ss << "#endif" << std::endl;
         }
     }
-    else if (val >= 0 && val < ShaderSource::PrecisionValueDefault) {
+    else if (is_valid_precision_value(val) &&
+             val != ShaderSource::PrecisionValueDefault)
+    {
         ss << "precision " << precision_map[val] << " ";
         ss << type_str << ";" << std::endl;
     }
@@ -512,7 +544,7 @@
 ShaderSource::default_precision(const ShaderSource::Precision& precision,
                                 ShaderSource::ShaderType type)
 {
-    if (type < 0 || type > ShaderSource::ShaderTypeUnknown)
+    if (!is_valid_shader_type(type))
         type = ShaderSource::ShaderTypeUnknown;
 
     if (type == ShaderSource::ShaderTypeUnknown) {
@@ -537,7 +569,7 @@
 const ShaderSource::Precision&
 ShaderSource::default_precision(ShaderSource::ShaderType type)
 {
-    if (type < 0 || type > ShaderSource::ShaderTypeUnknown)
+    if (!is_valid_shader_type(type))
         type = ShaderSource::ShaderTypeUnknown;
 
     return default_precision_[type];
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/util.cc glmark2-2021.02/android/jni/src/libmatrix/util.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/libmatrix/util.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/libmatrix/util.cc	2021-04-25 01:47:22.000000000 +0800
@@ -220,10 +220,10 @@
 uint64_t
 Util::get_timestamp_us()
 {
-    struct timeval tv;
-    gettimeofday(&tv, NULL);
-    uint64_t now = static_cast<uint64_t>(tv.tv_sec) * 1000000 +
-                   static_cast<double>(tv.tv_usec);
+    struct timespec ts;
+    clock_gettime(CLOCK_MONOTONIC, &ts);
+    uint64_t now = static_cast<uint64_t>(ts.tv_sec) * 1000000 +
+                   static_cast<double>(ts.tv_nsec)/1000.0;
     return now;
 }
 
@@ -244,7 +244,7 @@
 std::istream *
 Util::get_resource(const std::string &path)
 {
-    std::ifstream *ifs = new std::ifstream(path.c_str());
+    std::ifstream *ifs = new std::ifstream(path.c_str(), std::ios::binary);
 
     return static_cast<std::istream *>(ifs);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/main.cpp glmark2-2021.02/android/jni/src/main.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/main.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/main.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -42,12 +42,18 @@
 #include "native-state-mir.h"
 #elif GLMARK2_USE_WAYLAND
 #include "native-state-wayland.h"
+#elif GLMARK2_USE_DISPMANX
+#include "native-state-dispmanx.h"
+#elif GLMARK2_USE_WIN32
+#include "native-state-win32.h"
 #endif
 
 #if GLMARK2_USE_EGL
 #include "gl-state-egl.h"
 #elif GLMARK2_USE_GLX
 #include "gl-state-glx.h"
+#elif GLMARK2_USE_WGL
+#include "gl-state-wgl.h"
 #endif
 
 using std::vector;
@@ -138,7 +144,6 @@
 int
 main(int argc, char *argv[])
 {
-
     if (!Options::parse_args(argc, argv))
         return 1;
 
@@ -168,12 +173,18 @@
     NativeStateMir native_state;
 #elif GLMARK2_USE_WAYLAND
     NativeStateWayland native_state;
+#elif GLMARK2_USE_DISPMANX
+    NativeStateDispmanx native_state;
+#elif GLMARK2_USE_WIN32
+    NativeStateWin32 native_state;
 #endif
 
 #if GLMARK2_USE_EGL
     GLStateEGL gl_state;
 #elif GLMARK2_USE_GLX
     GLStateGLX gl_state;
+#elif GLMARK2_USE_WGL
+    GLStateWGL gl_state;
 #endif
 
     CanvasGeneric canvas(native_state, gl_state, Options::size.first, Options::size.second);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/main-loop.cpp glmark2-2021.02/android/jni/src/main-loop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/main-loop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/main-loop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -80,9 +80,9 @@
 
         /* If we have found a valid scene, set it up */
         if (bench_iter_ != benchmarks_.end()) {
+            before_scene_setup();
             if (!Options::reuse_context)
                 canvas_.reset();
-            before_scene_setup();
             scene_ = &(*bench_iter_)->setup_scene();
             if (!scene_->running()) {
                 if (!scene_->supported(false))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/mesh.cpp glmark2-2021.02/android/jni/src/mesh.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/mesh.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/mesh.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -396,8 +396,9 @@
         /* Update the current range from the vertex data */
         float *dest(array + nfloats * ri->first);
         for (size_t n = ri->first; n <= ri->second; n++) {
-            for (size_t i = 0; i < nfloats; i++)
-                *dest++ = vertices_[n][offset + i];
+            float *src(vertices_[n].data() + offset);
+            std::copy(src, src + nfloats, dest);
+            dest += nfloats;
         }
 
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/meson.build glmark2-2021.02/android/jni/src/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,263 @@
+common_sources = [
+    'benchmark-collection.cpp',
+    'benchmark.cpp',
+    'canvas-generic.cpp',
+    'gl-headers.cpp',
+    'gl-visual-config.cpp',
+    'image-reader.cpp',
+    'libmatrix/log.cc',
+    'libmatrix/mat.cc',
+    'libmatrix/program.cc',
+    'libmatrix/shader-source.cc',
+    'libmatrix/util.cc',
+    'main-loop.cpp',
+    'mesh.cpp',
+    'model.cpp',
+    'options.cpp',
+    'scene-buffer.cpp',
+    'scene-build.cpp',
+    'scene-bump.cpp',
+    'scene-clear.cpp',
+    'scene-conditionals.cpp',
+    'scene.cpp',
+    'scene-default-options.cpp',
+    'scene-desktop.cpp',
+    'scene-effect-2d.cpp',
+    'scene-function.cpp',
+    'scene-grid.cpp',
+    'scene-ideas/a.cc',
+    'scene-ideas.cpp',
+    'scene-ideas/d.cc',
+    'scene-ideas/e.cc',
+    'scene-ideas/i.cc',
+    'scene-ideas/lamp.cc',
+    'scene-ideas/logo.cc',
+    'scene-ideas/m.cc',
+    'scene-ideas/n.cc',
+    'scene-ideas/o.cc',
+    'scene-ideas/s.cc',
+    'scene-ideas/splines.cc',
+    'scene-ideas/table.cc',
+    'scene-ideas/t.cc',
+    'scene-jellyfish.cpp',
+    'scene-loop.cpp',
+    'scene-pulsar.cpp',
+    'scene-refract.cpp',
+    'scene-shading.cpp',
+    'scene-shadow.cpp',
+    'scene-terrain/base-renderer.cpp',
+    'scene-terrain/blur-renderer.cpp',
+    'scene-terrain/copy-renderer.cpp',
+    'scene-terrain.cpp',
+    'scene-terrain/luminance-renderer.cpp',
+    'scene-terrain/normal-from-height-renderer.cpp',
+    'scene-terrain/overlay-renderer.cpp',
+    'scene-terrain/renderer-chain.cpp',
+    'scene-terrain/simplex-noise-renderer.cpp',
+    'scene-terrain/terrain-renderer.cpp',
+    'scene-terrain/texture-renderer.cpp',
+    'scene-texture.cpp',
+    'shared-library.cpp',
+    'text-renderer.cpp',
+    'texture.cpp'
+]
+
+libmatrix_headers_dep = declare_dependency(
+    include_directories: include_directories('libmatrix'),
+    compile_args: ['-DUSE_EXCEPTIONS'],
+)
+
+common_deps = [
+    m_dep,
+    dl_dep,
+    libjpeg_dep,
+    libpng_dep,
+    libmatrix_headers_dep,
+]
+    
+if need_drm
+    native_drm_lib = static_library(
+        'native-drm',
+        'native-state-drm.cpp',
+        dependencies: [libdrm_dep, libudev_dep, gbm_dep, libmatrix_headers_dep],
+        )
+    native_drm_dep = declare_dependency(
+        link_with: native_drm_lib,
+        dependencies: libdrm_dep,
+        compile_args: ['-DGLMARK2_USE_DRM', '-D__GBM__'],
+        )
+else
+    native_drm_dep = declare_dependency()
+endif
+
+if need_wayland
+    wayland_scanner = find_program(wayland_scanner_dep.get_pkgconfig_variable('wayland_scanner'))
+    wayland_protocols_dir = wayland_protocols_dep.get_pkgconfig_variable('pkgdatadir')
+
+    xdg_shell_xml_path = wayland_protocols_dir + '/stable/xdg-shell/xdg-shell.xml'
+    xdg_shell_client_header = custom_target(
+        'xdg-shell client-header',
+        command: [ wayland_scanner, 'client-header', '@INPUT@', '@OUTPUT@' ],
+        input: xdg_shell_xml_path,
+        output: 'xdg-shell-client-protocol.h',
+        )
+    xdg_shell_private_code = custom_target(
+        'xdg-shell private-code',
+        command: [ wayland_scanner, 'private-code', '@INPUT@', '@OUTPUT@' ],
+        input: xdg_shell_xml_path,
+        output: 'xdg-shell-protocol.c',
+        )
+
+    native_wayland_lib = static_library(
+        'native-wayland',
+        'native-state-wayland.cpp',
+        xdg_shell_client_header,
+        xdg_shell_private_code,
+        dependencies: [libmatrix_headers_dep, wayland_client_dep, wayland_cursor_dep, wayland_egl_dep],
+        )
+
+    native_wayland_dep = declare_dependency(
+        link_with: native_wayland_lib,
+        dependencies: [wayland_client_dep, wayland_cursor_dep, wayland_egl_dep],
+        compile_args: ['-DGLMARK2_USE_WAYLAND', '-DWL_EGL_PLATFORM'],
+        )
+else
+    native_wayland_dep = declare_dependency()
+endif
+
+if need_x11
+    native_x11_lib = static_library(
+        'native-x11',
+        'native-state-x11.cpp',
+        dependencies: [x11_dep, libmatrix_headers_dep],
+        )
+
+    native_x11_dep = declare_dependency(
+        link_with: native_x11_lib,
+        compile_args: ['-DGLMARK2_USE_X11'],
+        )
+else
+    native_x11_dep = declare_dependency()
+endif
+
+if need_gl
+    gl_gl_lib = static_library(
+        'gl-gl',
+        'glad/src/gl.c',
+        include_directories: include_directories('glad/include'),
+        )
+    gl_gl_dep = declare_dependency(
+        link_with: gl_gl_lib,
+        include_directories: include_directories('glad/include'),
+        compile_args: ['-DGLMARK2_USE_GL'],
+        )
+
+    lib_common_gl = static_library(
+        'common-gl',
+        common_sources,
+        dependencies: [gl_gl_dep, common_deps],
+        include_directories: [
+            include_directories('scene-ideas'),
+            include_directories('scene-terrain'),
+            ]
+        )
+    common_gl_dep = declare_dependency(
+        link_with: lib_common_gl,
+        include_directories: include_directories('libmatrix'),
+        )
+else
+    gl_gl_dep = declare_dependency()
+    common_gl_dep = declare_dependency()
+endif
+
+if need_glesv2
+    gl_glesv2_lib = static_library(
+        'gl-glesv2',
+        'glad/src/gles2.c',
+        include_directories: include_directories('glad/include'),
+        )
+    gl_glesv2_dep = declare_dependency(
+        link_with: gl_glesv2_lib,
+        include_directories: include_directories('glad/include'),
+        compile_args: ['-DGLMARK2_USE_GLESv2'],
+        )
+
+    lib_common_glesv2 = static_library(
+        'common-glesv2',
+        common_sources,
+        dependencies: [gl_glesv2_dep, common_deps],
+        include_directories: [
+            include_directories('scene-ideas'),
+            include_directories('scene-terrain'),
+            ]
+        )
+    common_glesv2_dep = declare_dependency(
+        link_with: lib_common_glesv2,
+        include_directories: include_directories('libmatrix'),
+        )
+else
+    gl_glesv2_dep = declare_dependency()
+    common_glesv2_dep = declare_dependency()
+endif
+
+if need_egl
+    # gl-state-egl builds depend on both the GL type and the native type. To
+    # avoid creating all the required combinations explicitly, make this a
+    # source dependency which will be built properly as part of the final
+    # glmark2 executable.
+    wsi_egl_dep = declare_dependency(
+        sources: ['gl-state-egl.cpp', 'glad/src/egl.c'],
+        include_directories: [
+            include_directories('glad/include'),
+            include_directories('libmatrix'),
+            ],
+        compile_args: ['-DGLMARK2_USE_EGL'],
+        )
+else
+    wsi_egl_dep = declare_dependency()
+endif
+
+if need_glx
+    # Although there is a single valid flavor for glx (x11-gl) and we could
+    # build gl-state-glx here, make this a source dependency for convenience
+    # and parity with gl-state-egl.
+    wsi_glx_dep = declare_dependency(
+        sources: ['gl-state-glx.cpp', 'glad/src/glx.c'],
+        include_directories: [
+            include_directories('glad/include'),
+            include_directories('libmatrix'),
+            ],
+        compile_args: ['-DGLMARK2_USE_GLX'],
+        )
+else
+    wsi_glx_dep = declare_dependency()
+endif
+
+# Flavor info : [ binary, native, gl, wsi ]
+flavor_info = {
+  'drm-gl' : ['glmark2-drm', native_drm_dep, gl_gl_dep, wsi_egl_dep],
+  'drm-glesv2' : ['glmark2-es2-drm', native_drm_dep, gl_glesv2_dep, wsi_egl_dep],
+  'wayland-gl' : ['glmark2-wayland', native_wayland_dep, gl_gl_dep, wsi_egl_dep],
+  'wayland-glesv2' : ['glmark2-es2-wayland', native_wayland_dep, gl_glesv2_dep, wsi_egl_dep],
+  'x11-gl' : ['glmark2', native_x11_dep, gl_gl_dep, wsi_glx_dep],
+  'x11-glesv2' : ['glmark2-es2', native_x11_dep, gl_glesv2_dep, wsi_egl_dep],
+}
+
+foreach flavor : flavors
+    info = flavor_info[flavor]
+    binary = info[0]
+    native_dep = info[1]
+    gl_dep = info[2]
+    wsi_dep = info[3]
+    if gl_dep == gl_gl_dep
+        common_dep = common_gl_dep
+    elif gl_dep == gl_glesv2_dep
+        common_dep = common_glesv2_dep
+    endif
+    executable(
+        binary,
+        ['main.cpp', 'canvas-generic.cpp'],
+        dependencies: [common_dep, native_dep, gl_dep, wsi_dep],
+        install: true,
+    )
+endforeach
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/model.cpp glmark2-2021.02/android/jni/src/model.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/model.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/model.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -361,7 +361,7 @@
 
     Log::debug("Loading model from 3ds file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& input_file(*input_file_ptr);
 
     if (!input_file) {
@@ -385,7 +385,7 @@
             return false;
         }
 
-        //Read the lenght of the chunk
+        //Read the length of the chunk
         read_or_fail(input_file, &chunk_length, 4);
 
         switch (chunk_id)
@@ -393,7 +393,7 @@
             //----------------- MAIN3DS -----------------
             // Description: Main chunk, contains all the other chunks
             // Chunk ID: 4d4d
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x4d4d:
                 break;
@@ -401,7 +401,7 @@
             //----------------- EDIT3DS -----------------
             // Description: 3D Editor chunk, objects layout info
             // Chunk ID: 3d3d (hex)
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x3d3d:
                 break;
@@ -409,7 +409,7 @@
             //--------------- EDIT_OBJECT ---------------
             // Description: Object block, info for each object
             // Chunk ID: 4000 (hex)
-            // Chunk Lenght: len(object name) + sub chunks
+            // Chunk Length: len(object name) + sub chunks
             //-------------------------------------------
             case 0x4000:
                 {
@@ -429,7 +429,7 @@
             //--------------- OBJ_TRIMESH ---------------
             // Description: Triangular mesh, contains chunks for 3d mesh info
             // Chunk ID: 4100 (hex)
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x4100:
                 break;
@@ -437,7 +437,7 @@
             //--------------- TRI_VERTEXL ---------------
             // Description: Vertices list
             // Chunk ID: 4110 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of vertices)
+            // Chunk Length: 1 x unsigned short (number of vertices)
             //             + 3 x float (vertex coordinates) x (number of vertices)
             //             + sub chunks
             //-------------------------------------------
@@ -461,7 +461,7 @@
             //--------------- TRI_FACEL1 ----------------
             // Description: Polygons (faces) list
             // Chunk ID: 4120 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of polygons)
+            // Chunk Length: 1 x unsigned short (number of polygons)
             //             + 3 x unsigned short (polygon points) x (number of polygons)
             //             + sub chunks
             //-------------------------------------------
@@ -484,7 +484,7 @@
             //------------- TRI_MAPPINGCOORS ------------
             // Description: Vertices list
             // Chunk ID: 4140 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of mapping points)
+            // Chunk Length: 1 x unsigned short (number of mapping points)
             //             + 2 x float (mapping coordinates) x (number of mapping points)
             //             + sub chunks
             //-------------------------------------------
@@ -505,7 +505,7 @@
 
             //----------- Skip unknow chunks ------------
             //We need to skip all the chunks that currently we don't use
-            //We use the chunk lenght information to set the file pointer
+            //We use the chunk length information to set the file pointer
             //to the same level next chunk
             //-------------------------------------------
             default:
@@ -681,7 +681,7 @@
 {
     Log::debug("Loading model from obj file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& inputFile(*input_file_ptr);
     if (!inputFile)
     {
@@ -814,7 +814,7 @@
         return ModelPrivate::modelMap;
     }
     vector<string> pathVec;
-    string dataDir(GLMARK_DATA_PATH"/models");
+    string dataDir(Options::data_path + "/models");
     Util::list_files(dataDir, pathVec);
 #ifdef GLMARK_EXTRAS_PATH
     string extrasDir(GLMARK_EXTRAS_PATH"/models");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-dispmanx.cpp glmark2-2021.02/android/jni/src/native-state-dispmanx.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-dispmanx.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-dispmanx.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,132 @@
+/*
+ * Copyright © 2013 Rafal Mielniczuk
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Rafal Mielniczuk <rafal.mielniczuk2@gmail.com>
+ */
+
+#include "native-state-dispmanx.h"
+#include "log.h"
+
+#include <cstring>
+#include <csignal>
+
+NativeStateDispmanx::NativeStateDispmanx()
+{
+    memset(&properties_, 0, sizeof(properties_));
+    memset(&egl_dispmanx_window, 0, sizeof(egl_dispmanx_window));
+}
+
+NativeStateDispmanx::~NativeStateDispmanx()
+{
+}
+
+bool
+NativeStateDispmanx::init_display()
+{
+    bcm_host_init();
+
+    return true;
+}
+
+void*
+NativeStateDispmanx::display()
+{
+    return EGL_DEFAULT_DISPLAY;
+}
+
+bool
+NativeStateDispmanx::create_window(WindowProperties const& properties)
+{
+    int dispmanx_output = 0; /* LCD */
+
+    if (!properties.fullscreen) {
+        Log::error("Error: Dispmanx only supports full screen rendering.\n");
+	return false;
+    }
+
+    unsigned screen_width, screen_height;
+    if (graphics_get_display_size(dispmanx_output,
+				  &screen_width, &screen_height) < 0) {
+        Log::error("Error: Couldn't get screen width/height.\n");
+	return false;
+    }
+
+    properties_.fullscreen = properties.fullscreen;
+    properties_.visual_id = properties.visual_id;
+    properties_.width = screen_width;
+    properties_.height = screen_height;
+
+    dispmanx_display = vc_dispmanx_display_open(dispmanx_output);
+
+    VC_RECT_T dst_rect;
+    VC_RECT_T src_rect;
+    dst_rect.x = 0;
+    dst_rect.y = 0;
+    dst_rect.width = screen_width;
+    dst_rect.height = screen_height;
+
+    src_rect.x = 0;
+    src_rect.y = 0;
+    src_rect.width = screen_width << 16;
+    src_rect.height = screen_height << 16;
+
+    dispmanx_update = vc_dispmanx_update_start(0);
+
+    dispmanx_element = vc_dispmanx_element_add(dispmanx_update,
+					       dispmanx_display,
+					       0 /*layer*/,
+					       &dst_rect,
+					       0 /*src*/,
+					       &src_rect,
+					       DISPMANX_PROTECTION_NONE,
+					       0 /*alpha*/,
+					       0 /*clamp*/,
+					       DISPMANX_NO_ROTATE);
+
+    egl_dispmanx_window.element = dispmanx_element;
+    egl_dispmanx_window.width = dst_rect.width;
+    egl_dispmanx_window.height = dst_rect.height;
+    vc_dispmanx_update_submit_sync(dispmanx_update);
+
+    return true;
+}
+
+void*
+NativeStateDispmanx::window(WindowProperties &properties)
+{
+    properties = properties_;
+    return &egl_dispmanx_window;
+}
+
+void
+NativeStateDispmanx::visible(bool /*v*/)
+{
+}
+
+bool
+NativeStateDispmanx::should_quit()
+{
+    return false;
+}
+
+void
+NativeStateDispmanx::flip()
+{
+}
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-dispmanx.h glmark2-2021.02/android/jni/src/native-state-dispmanx.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-dispmanx.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-dispmanx.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+/*
+ * Copyright © 2013 Rafal Mielniczuk
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Rafal Mielniczuk <rafal.mielniczuk2@gmail.com>
+ */
+
+#ifndef GLMARK2_NATIVE_STATE_DISPMANX_H_
+#define GLMARK2_NATIVE_STATE_DISPMANX_H_
+
+#include <vector>
+#include "bcm_host.h"
+#include "GLES/gl.h"
+#include "EGL/egl.h"
+#include "EGL/eglext.h"
+
+#include "native-state.h"
+
+class NativeStateDispmanx : public NativeState
+{
+public:
+    NativeStateDispmanx();
+    ~NativeStateDispmanx();
+
+    bool init_display();
+    void* display();
+    bool create_window(WindowProperties const& properties);
+    void* window(WindowProperties& properties);
+    void visible(bool v);
+    bool should_quit();
+    void flip();
+
+private:
+    DISPMANX_DISPLAY_HANDLE_T dispmanx_display;
+    DISPMANX_UPDATE_HANDLE_T dispmanx_update;
+    DISPMANX_ELEMENT_HANDLE_T dispmanx_element;
+    EGL_DISPMANX_WINDOW_T egl_dispmanx_window;
+    WindowProperties properties_;
+};
+
+#endif /* GLMARK2_NATIVE_STATE_DISPMANX_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-drm.cpp glmark2-2021.02/android/jni/src/native-state-drm.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-drm.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-drm.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -25,6 +25,11 @@
 #include "native-state-drm.h"
 #include "log.h"
 
+#include <fcntl.h>
+#include <libudev.h>
+#include <cstring>
+#include <string>
+
 /******************
  * Public methods *
  ******************/
@@ -106,7 +111,8 @@
     FD_ZERO(&fds);
     FD_SET(fd_, &fds);
     drmEventContext evCtx;
-    evCtx.version = DRM_EVENT_CONTEXT_VERSION;
+    memset(&evCtx, 0, sizeof(evCtx));
+    evCtx.version = 2;
     evCtx.page_flip_handler = page_flip_handler;
 
     while (waiting) {
@@ -134,6 +140,269 @@
  * Private methods *
  *******************/
 
+/* Simple helpers */
+
+inline static bool valid_fd(int fd)
+{
+    return fd >= 0;
+}
+
+inline static bool valid_drm_node_path(std::string const& provided_node_path)
+{
+    return !provided_node_path.empty();
+}
+
+inline static bool invalid_drm_node_path(std::string const& provided_node_path)
+{
+    return !(valid_drm_node_path(provided_node_path));
+}
+
+/* Udev methods */
+// Udev detection functions
+#define UDEV_TEST_FUNC_SIGNATURE(udev_identifier, device_identifier, syspath_identifier) \
+    struct udev * __restrict const udev_identifier, \
+    struct udev_device * __restrict const device_identifier, \
+    char const * __restrict syspath_identifier
+
+/* Omitting the parameter names is kind of ugly but is the only way
+ * to force G++ to forget about the unused parameters.
+ * Having big warnings during the compilation isn't very nice.
+ *
+ * These functions will be used as function pointers and should have
+ * the same signature to avoid weird stack related issues.
+ *
+ * Another way to deal with that issue will be to mark unused parameters
+ * with __attribute__((unused))
+ */
+static bool udev_drm_test_virtual(
+    UDEV_TEST_FUNC_SIGNATURE(,,tested_node_syspath))
+{
+    return strstr(tested_node_syspath, "virtual") != NULL;
+}
+
+static bool udev_drm_test_not_virtual(
+    UDEV_TEST_FUNC_SIGNATURE(udev, current_device, tested_node_syspath))
+{
+    return !udev_drm_test_virtual(udev,
+                                  current_device,
+                                  tested_node_syspath);
+}
+
+static bool
+udev_drm_test_primary_gpu(UDEV_TEST_FUNC_SIGNATURE(, current_device,))
+{
+    bool is_main_gpu = false;
+
+    auto const drm_node_parent = udev_device_get_parent(current_device);
+
+    /* While tempting, using udev_device_unref will generate issues
+     * when unreferencing the child in udev_get_node_that_pass_in_enum
+     *
+     * udev_device_unref WILL unreference the parent, so avoid doing
+     * that here.
+     *
+     * ( See udev sources : src/libudev/libudev-device.c )
+     */
+    if (drm_node_parent != NULL) {
+        is_main_gpu =
+            (udev_device_get_sysattr_value(drm_node_parent, "boot_vga")
+             != NULL);
+    }
+
+    return is_main_gpu;
+}
+
+/* Test if the drm-device is actually modeset capable.
+ * Render-only devices cannot drive an actual display,
+ * so the GETRESOURCES ioctl will fail in that case.
+ */
+static bool udev_drm_test_modeset(std::string const& dev_path)
+{
+    struct drm_mode_card_res res {};
+    int fd, ret;
+
+    fd = open(dev_path.c_str(), O_RDWR);
+    if (!valid_fd(fd))
+        return false;
+
+    ret = drmIoctl(fd, DRM_IOCTL_MODE_GETRESOURCES, &res);
+    drmClose(fd);
+
+    return !ret;
+}
+
+static std::string
+udev_get_node_that_pass_in_enum(
+    struct udev * __restrict const udev,
+    struct udev_enumerate * __restrict const dev_enum,
+    bool (* check_function)(UDEV_TEST_FUNC_SIGNATURE(,,)))
+{
+    std::string result;
+
+    auto current_element = udev_enumerate_get_list_entry(dev_enum);
+
+    while (current_element && result.empty()) {
+        char const * __restrict current_element_sys_path =
+            udev_list_entry_get_name(current_element);
+
+        if (current_element_sys_path) {
+            struct udev_device * current_device =
+                udev_device_new_from_syspath(udev,
+                                             current_element_sys_path);
+            auto check_passed = check_function(
+                udev, current_device, current_element_sys_path);
+
+            if (check_passed) {
+                const char * device_node_path =
+                    udev_device_get_devnode(current_device);
+
+                if (device_node_path &&
+                    udev_drm_test_modeset(device_node_path)) {
+                        result = device_node_path;
+                }
+
+            }
+
+            udev_device_unref(current_device);
+        }
+
+        current_element = udev_list_entry_get_next(current_element);
+    }
+
+    return result;
+}
+
+/* Inspired by KWin detection mechanism */
+/* And yet KWin got it wrong too, it seems.
+ * 1 - Looking for the primary GPU by checking the flag 'boot_vga'
+ *     won't get you far with some embedded chipsets, like Rockchip.
+ * 2 - Looking for a GPU plugged in PCI will fail on various embedded
+ *     devices !
+ * 3 - Looking for a render node is not guaranteed to work on some
+ *     poorly maintained DRM drivers, which plague some embedded
+ *     devices too...
+ *
+ * So, we won't play too smart here.
+ * - We first check for a primary GPU plugged in PCI with the 'boot_vga'
+ *   attribute, to take care of Desktop users using multiple GPU.
+ * - Then, we just check for a DRM node that is not virtual
+ * - At least, we use the first virtual node we get, if we didn't find
+ *   anything yet.
+ * This should take care of almost every potential use case.
+ *
+ * The remaining ones will be dealt with an additional option to
+ * specify the DRM dev node manually.
+ */
+static std::string udev_main_gpu_drm_node_path()
+{
+    Log::debug("Using Udev to detect the right DRM node to use\n");
+    auto udev = udev_new();
+    auto dev_enumeration = udev_enumerate_new(udev);
+
+    udev_enumerate_add_match_subsystem(dev_enumeration, "drm");
+    udev_enumerate_add_match_sysname(dev_enumeration, "card[0-9]*");
+    udev_enumerate_scan_devices(dev_enumeration);
+
+    Log::debug("Looking for the main GPU DRM node...\n");
+    std::string node_path = udev_get_node_that_pass_in_enum(
+        udev, dev_enumeration, udev_drm_test_primary_gpu);
+
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found!\n");
+        Log::debug("Looking for a concrete GPU DRM node...\n");
+        node_path = udev_get_node_that_pass_in_enum(
+            udev, dev_enumeration, udev_drm_test_not_virtual);
+    }
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found!?\n");
+        Log::debug("Looking for a virtual GPU DRM node...\n");
+        node_path = udev_get_node_that_pass_in_enum(
+            udev, dev_enumeration, udev_drm_test_virtual);
+    }
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found.\n");
+        Log::debug("Cannot find a single DRM node using UDEV...\n");
+    }
+
+    if (valid_drm_node_path(node_path)) {
+        Log::debug("Success!\n");
+    }
+
+    udev_enumerate_unref(dev_enumeration);
+    udev_unref(udev);
+
+    return node_path;
+}
+
+static int open_using_udev_scan()
+{
+    auto dev_path = udev_main_gpu_drm_node_path();
+
+    int fd = -1;
+    if (valid_drm_node_path(dev_path)) {
+        Log::debug("Trying to use the DRM node %s\n", dev_path.c_str());
+        fd = open(dev_path.c_str(), O_RDWR);
+    }
+    else {
+        Log::error("Can't determine the main graphic card "
+                   "DRM device node\n");
+    }
+
+    if (!valid_fd(fd)) {
+        // %m is GLIBC specific... Maybe use strerror here...
+        Log::error("Tried to use '%s' but failed.\nReason : %m\n",
+                   dev_path.c_str());
+    }
+    else
+        Log::debug("Success!\n");
+
+    return fd;
+}
+
+/* -- End of Udev helpers -- */
+
+/*
+ * This method is there to take care of cases that would not be handled
+ * by open_using_udev_scan. This should not happen.
+ *
+ * If your driver defines a /dev/dri/cardX node and open_using_udev_scan
+ * were not able to detect it, you should probably file an issue.
+ *
+ */
+static int open_using_module_checking()
+{
+    static const char* drm_modules[] = {
+        "i915",
+        "imx-drm",
+        "nouveau",
+        "radeon",
+        "vmgfx",
+        "omapdrm",
+        "exynos",
+        "pl111",
+        "vc4",
+        "msm",
+        "meson",
+        "rockchip",
+        "sun4i-drm",
+        "stm",
+    };
+
+    int fd = -1;
+    unsigned int num_modules(sizeof(drm_modules)/sizeof(drm_modules[0]));
+    for (unsigned int m = 0; m < num_modules; m++) {
+        fd = drmOpen(drm_modules[m], 0);
+        if (fd < 0) {
+            Log::debug("Failed to open DRM module '%s'\n", drm_modules[m]);
+            continue;
+        }
+        Log::debug("Opened DRM module '%s'\n", drm_modules[m]);
+        break;
+    }
+
+    return fd;
+}
+
 void
 NativeStateDRM::fb_destroy_callback(gbm_bo* bo, void* data)
 {
@@ -197,34 +466,23 @@
 bool
 NativeStateDRM::init()
 {
-    // TODO: Replace this with something that explicitly probes for the loaded
-    // driver (udev?).
-    static const char* drm_modules[] = {
-        "i915",
-        "nouveau",
-        "radeon",
-        "vmgfx",
-        "omapdrm",
-        "exynos"
-    };
+    // TODO: The user should be able to define *exactly* which device
+    //       node to open and the program should try to open only
+    //       this node, in order to take care of unknown use cases.
+    int fd = open_using_udev_scan();
 
-    unsigned int num_modules(sizeof(drm_modules)/sizeof(drm_modules[0]));
-    for (unsigned int m = 0; m < num_modules; m++) {
-        fd_ = drmOpen(drm_modules[m], 0);
-        if (fd_ < 0) {
-            Log::debug("Failed to open DRM module '%s'\n", drm_modules[m]);
-            continue;
-        }
-        Log::debug("Opened DRM module '%s'\n", drm_modules[m]);
-        break;
+    if (!valid_fd(fd)) {
+        fd = open_using_module_checking();
     }
 
-    if (fd_ < 0) {
+    if (!valid_fd(fd)) {
         Log::error("Failed to find a suitable DRM device\n");
         return false;
     }
 
-    resources_ = drmModeGetResources(fd_);
+    fd_ = fd;
+
+    resources_ = drmModeGetResources(fd);
     if (!resources_) {
         Log::error("drmModeGetResources failed\n");
         return false;
@@ -232,7 +490,7 @@
 
     // Find a connected connector
     for (int c = 0; c < resources_->count_connectors; c++) {
-        connector_ = drmModeGetConnector(fd_, resources_->connectors[c]);
+        connector_ = drmModeGetConnector(fd, resources_->connectors[c]);
         if (DRM_MODE_CONNECTED == connector_->connection) {
             break;
         }
@@ -263,14 +521,40 @@
 
     // Find a suitable encoder
     for (int e = 0; e < resources_->count_encoders; e++) {
+        bool found = false;
         encoder_ = drmModeGetEncoder(fd_, resources_->encoders[e]);
-        if (encoder_ && encoder_->encoder_id == connector_->encoder_id) {
-            break;
+        for (int ce = 0; ce < connector_->count_encoders; ce++) {
+            if (encoder_ && encoder_->encoder_id == connector_->encoders[ce]) {
+                found = true;
+                break;
+            }
         }
+        if (found)
+            break;
         drmModeFreeEncoder(encoder_);
         encoder_ = 0;
     }
 
+    // If encoder is not connected to the connector try to find
+    // a suitable one
+    if (!encoder_) {
+        for (int e = 0; e < connector_->count_encoders; e++) {
+            encoder_ = drmModeGetEncoder(fd, connector_->encoders[e]);
+            for (int c = 0; c < resources_->count_crtcs; c++) {
+                if (encoder_->possible_crtcs & (1 << c)) {
+                    encoder_->crtc_id = resources_->crtcs[c];
+                    break;
+                }
+            }
+            if (encoder_->crtc_id) {
+                break;
+            }
+
+            drmModeFreeEncoder(encoder_);
+            encoder_ = 0;
+        }
+    }
+
     if (!encoder_) {
         Log::error("Failed to find a suitable encoder\n");
         return false;
@@ -282,8 +566,13 @@
 
     crtc_ = drmModeGetCrtc(fd_, encoder_->crtc_id);
     if (!crtc_) {
-        Log::error("Failed to get current CRTC\n");
-        return false;
+        // if there is no current CRTC, make sure to attach a suitable one
+        for (int c = 0; c < resources_->count_crtcs; c++) {
+            if (encoder_->possible_crtcs & (1 << c)) {
+                encoder_->crtc_id = resources_->crtcs[c];
+                break;
+            }
+        }
     }
 
     signal(SIGINT, &NativeStateDRM::quit_handler);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-drm.h glmark2-2021.02/android/jni/src/native-state-drm.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-drm.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-drm.h	2021-04-25 01:47:22.000000000 +0800
@@ -41,6 +41,7 @@
         resources_(0),
         connector_(0),
         encoder_(0),
+        crtc_(0),
         mode_(0),
         dev_(0),
         surface_(0),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state.h glmark2-2021.02/android/jni/src/native-state.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,12 +22,14 @@
 #ifndef GLMARK2_NATIVE_STATE_H_
 #define GLMARK2_NATIVE_STATE_H_
 
+#include <stdint.h>
+
 class NativeState
 {
 public:
     struct WindowProperties
     {
-        WindowProperties(int w, int h, bool f, int v)
+        WindowProperties(int w, int h, bool f, intptr_t v)
             : width(w), height(h), fullscreen(f), visual_id(v) {}
         WindowProperties()
             : width(0), height(0), fullscreen(false), visual_id(0) {}
@@ -35,7 +37,7 @@
         int width;
         int height;
         bool fullscreen;
-        int visual_id;
+        intptr_t visual_id;
     };
 
     virtual ~NativeState() {}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-mir.cpp glmark2-2021.02/android/jni/src/native-state-mir.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-mir.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-mir.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -157,7 +157,7 @@
 bool
 NativeStateMir::create_window(WindowProperties const& properties)
 {
-    static const char *win_name("glmark2 "GLMARK_VERSION);
+    static const char *win_name("glmark2 " GLMARK_VERSION);
 
     if (!mir_connection_is_valid(mir_connection_)) {
         Log::error("Cannot create a Mir surface without a valid connection "
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-wayland.cpp glmark2-2021.02/android/jni/src/native-state-wayland.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-wayland.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-wayland.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -21,19 +21,29 @@
  */
 
 #include "native-state-wayland.h"
+#include "log.h"
 
+#include <linux/input.h>
 #include <cstring>
 #include <csignal>
+#include <unistd.h>
 
 const struct wl_registry_listener NativeStateWayland::registry_listener_ = {
     NativeStateWayland::registry_handle_global,
     NativeStateWayland::registry_handle_global_remove
 };
 
-const struct wl_shell_surface_listener NativeStateWayland::shell_surface_listener_ = {
-    NativeStateWayland::shell_surface_handle_ping,
-    NativeStateWayland::shell_surface_handle_configure,
-    NativeStateWayland::shell_surface_handle_popup_done
+const struct xdg_toplevel_listener NativeStateWayland::xdg_toplevel_listener_ = {
+    NativeStateWayland::xdg_toplevel_handle_configure,
+    NativeStateWayland::xdg_toplevel_handle_close
+};
+
+const struct xdg_surface_listener NativeStateWayland::xdg_surface_listener_ = {
+    NativeStateWayland::xdg_surface_handle_configure
+};
+
+const struct xdg_wm_base_listener NativeStateWayland::xdg_wm_base_listener_ = {
+    NativeStateWayland::xdg_wm_base_handle_ping
 };
 
 const struct wl_output_listener NativeStateWayland::output_listener_ = {
@@ -43,29 +53,58 @@
     NativeStateWayland::output_handle_scale
 };
 
+const struct wl_seat_listener NativeStateWayland::seat_listener_ = {
+    NativeStateWayland::seat_handle_capabilities,
+};
+
+const struct wl_pointer_listener NativeStateWayland::pointer_listener_ = {
+    NativeStateWayland::pointer_handle_enter,
+    NativeStateWayland::pointer_handle_leave,
+    NativeStateWayland::pointer_handle_motion,
+    NativeStateWayland::pointer_handle_button,
+    NativeStateWayland::pointer_handle_axis,
+};
+
+const struct wl_keyboard_listener NativeStateWayland::keyboard_listener_ = {
+    NativeStateWayland::keyboard_handle_keymap,
+    NativeStateWayland::keyboard_handle_enter,
+    NativeStateWayland::keyboard_handle_leave,
+    NativeStateWayland::keyboard_handle_key,
+    NativeStateWayland::keyboard_handle_modifiers,
+};
+
 volatile bool NativeStateWayland::should_quit_ = false;
 
-NativeStateWayland::NativeStateWayland() : display_(0), window_(0)
+NativeStateWayland::NativeStateWayland() : cursor_(0), display_(0), window_(0)
 {
 }
 
 NativeStateWayland::~NativeStateWayland()
 {
     if (window_) {
-        if (window_->shell_surface)
-            wl_shell_surface_destroy(window_->shell_surface);
-        if (window_->opaque_reqion)
-            wl_region_destroy(window_->opaque_reqion);
-        if (window_->surface)
-            wl_surface_destroy(window_->surface);
+        if (window_->xdg_toplevel)
+            xdg_toplevel_destroy(window_->xdg_toplevel);
+        if (window_->xdg_surface)
+            xdg_surface_destroy(window_->xdg_surface);
         if (window_->native)
             wl_egl_window_destroy(window_->native);
+        if (window_->surface)
+            wl_surface_destroy(window_->surface);
         delete window_;
     }
 
+    if (cursor_) {
+        if (cursor_->cursor_surface)
+            wl_surface_destroy(cursor_->cursor_surface);
+        if (cursor_->cursor_theme)
+            wl_cursor_theme_destroy(cursor_->cursor_theme);
+
+        delete cursor_;
+    }
+
     if (display_) {
-        if (display_->shell)
-            wl_shell_destroy(display_->shell);
+        if (display_->xdg_wm_base)
+            xdg_wm_base_destroy(display_->xdg_wm_base);
 
         for (OutputsVector::iterator it = display_->outputs.begin();
              it != display_->outputs.end(); ++it) {
@@ -88,30 +127,66 @@
 void
 NativeStateWayland::registry_handle_global(void *data, struct wl_registry *registry,
                                            uint32_t id, const char *interface,
-                                           uint32_t /*version*/)
+                                           uint32_t version)
 {
     NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
     if (strcmp(interface, "wl_compositor") == 0) {
         that->display_->compositor =
                 static_cast<struct wl_compositor *>(
                     wl_registry_bind(registry,
-                                     id, &wl_compositor_interface, 1));
-    } else if (strcmp(interface, "wl_shell") == 0) {
-        that->display_->shell =
-                static_cast<struct wl_shell *>(
+                                     id, &wl_compositor_interface, std::min(version, 4U)));
+    } else if (strcmp(interface, "xdg_wm_base") == 0) {
+        that->display_->xdg_wm_base =
+                static_cast<struct xdg_wm_base *>(
                     wl_registry_bind(registry,
-                                     id, &wl_shell_interface, 1));
+                                     id, &xdg_wm_base_interface, std::min(version, 2U)));
+        xdg_wm_base_add_listener(that->display_->xdg_wm_base, &xdg_wm_base_listener_, that);
     } else if (strcmp(interface, "wl_output") == 0) {
-        struct my_output *my_output = new struct my_output;
+        struct my_output *my_output = new struct my_output();
         memset(my_output, 0, sizeof(*my_output));
+        my_output->scale = 1;
         my_output->output =
                 static_cast<struct wl_output *>(
                     wl_registry_bind(registry,
-                                     id, &wl_output_interface, 2));
+                                     id, &wl_output_interface, std::min(version, 2U)));
         that->display_->outputs.push_back(my_output);
 
         wl_output_add_listener(my_output->output, &output_listener_, my_output);
         wl_display_roundtrip(that->display_->display);
+    } else if (strcmp(interface, "wl_seat") == 0) {
+        that->display_->seat =
+            static_cast<struct wl_seat *>(
+            wl_registry_bind(registry, id, &wl_seat_interface, 1));
+        wl_seat_add_listener(that->display_->seat, &seat_listener_, that);
+    } else if (strcmp(interface, "wl_shm") == 0) {
+        that->display_->shm =
+            static_cast<struct wl_shm *>(
+                wl_registry_bind(registry, id, &wl_shm_interface, 1));
+
+        struct my_cursor *my_cursor = new struct my_cursor();
+        my_cursor->cursor_surface =
+             wl_compositor_create_surface(that->display_->compositor);
+        my_cursor->cursor_theme = wl_cursor_theme_load(NULL, 32, that->display_->shm);
+
+        if (!my_cursor->cursor_theme) {
+            Log::error("unable to load default theme\n");
+            wl_surface_destroy(my_cursor->cursor_surface);
+            delete my_cursor;
+            return;
+        }
+
+        my_cursor->default_cursor =
+            wl_cursor_theme_get_cursor(my_cursor->cursor_theme, "left_ptr");
+
+        if (!my_cursor->default_cursor) {
+            wl_surface_destroy(my_cursor->cursor_surface);
+            // assume above cursor_theme was set
+            wl_cursor_theme_destroy(my_cursor->cursor_theme);
+            delete my_cursor;
+            return;
+        }
+
+        that->cursor_ = my_cursor;
     }
 }
 
@@ -152,32 +227,84 @@
 }
 
 void
-NativeStateWayland::output_handle_scale(void * /*data*/, struct wl_output * /*wl_output*/,
-                                        int32_t /*factor*/)
+NativeStateWayland::output_handle_scale(void *data, struct wl_output * /*wl_output*/,
+                                        int32_t factor)
 {
+    struct my_output *my_output = static_cast<struct my_output *>(data);
+    my_output->scale = factor;
 }
 
 void
-NativeStateWayland::shell_surface_handle_ping(void * /*data*/, struct wl_shell_surface *shell_surface,
-                                              uint32_t serial)
+NativeStateWayland::xdg_wm_base_handle_ping(void * /*data*/, struct xdg_wm_base *xdg_wm_base,
+                                            uint32_t serial)
 {
-    wl_shell_surface_pong(shell_surface, serial);
+    xdg_wm_base_pong(xdg_wm_base, serial);
 }
 
 void
-NativeStateWayland::shell_surface_handle_popup_done(void * /*data*/,
-                                                    struct wl_shell_surface * /*shell_surface*/)
+NativeStateWayland::xdg_toplevel_handle_configure(void *data, struct xdg_toplevel * /*xdg_toplevel*/,
+                                                  int32_t width, int32_t height,
+                                                  struct wl_array *states)
 {
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+    uint32_t *state;
+    bool want_fullscreen = false;
+    uint32_t scale = 1;
+
+    that->window_->waiting_for_configure = false;
+
+    if (!that->display_->outputs.empty()) scale = that->display_->outputs.at(0)->scale;
+
+    for (state = (uint32_t *) states->data;
+         state < (uint32_t *) ((char *) states->data + states->size);
+         state++) {
+        if (*state == XDG_TOPLEVEL_STATE_FULLSCREEN)
+            want_fullscreen = true;
+    }
+
+    if (want_fullscreen) {
+        width *= scale;
+        height *= scale;
+    }
+
+    if (want_fullscreen == that->window_->properties.fullscreen &&
+        (width == 0 || width == that->window_->properties.width) &&
+        (height == 0 || height == that->window_->properties.fullscreen)) {
+        return;
+    }
+
+    that->window_->properties.fullscreen = want_fullscreen;
+    that->window_->properties.width = width;
+    that->window_->properties.height = height;
+
+    if (!that->window_->native)
+        return;
+
+    wl_egl_window_resize(that->window_->native, width, height, 0, 0);
+
+    struct wl_region *opaque_reqion = wl_compositor_create_region(that->display_->compositor);
+    wl_region_add(opaque_reqion, 0, 0,
+                  that->window_->properties.width,
+                  that->window_->properties.height);
+    wl_surface_set_opaque_region(that->window_->surface, opaque_reqion);
+    wl_region_destroy(opaque_reqion);
 }
 
 void
-NativeStateWayland::shell_surface_handle_configure(void *data, struct wl_shell_surface * /*shell_surface*/,
-                                                   uint32_t /*edges*/, int32_t width, int32_t height)
+NativeStateWayland::xdg_toplevel_handle_close(void *data,
+                                              struct xdg_toplevel * /*xdg_toplevel */)
 {
     NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
-    that->window_->properties.width = width;
-    that->window_->properties.height = height;
-    wl_egl_window_resize(that->window_->native, width, height, 0, 0);
+
+    that->should_quit_ = true;
+}
+
+void
+NativeStateWayland::xdg_surface_handle_configure(void * /*data*/,
+                                                 struct xdg_surface *xdg_surface,
+                                                 uint32_t serial)
+{
+    xdg_surface_ack_configure(xdg_surface, serial);
 }
 
 bool
@@ -191,7 +318,7 @@
     sigaction(SIGINT, &sa, NULL);
     sigaction(SIGTERM, &sa, NULL);
 
-    display_ = new struct my_display;
+    display_ = new struct my_display();
 
     if (!display_) {
         return false;
@@ -223,42 +350,51 @@
 {
     struct my_output *output = 0;
     if (!display_->outputs.empty()) output = display_->outputs.at(0);
-    window_ = new struct my_window;
+    window_ = new struct my_window();
     window_->properties = properties;
-    window_->surface = wl_compositor_create_surface(display_->compositor);
-    if (window_->properties.fullscreen && output) {
-        window_->native = wl_egl_window_create(window_->surface,
-                                               output->width, output->height);
-        window_->properties.width = output->width;
-        window_->properties.height = output->height;
-    } else {
-        window_->native = wl_egl_window_create(window_->surface,
-                                               properties.width, properties.height);
-    }
 
-    window_->opaque_reqion = wl_compositor_create_region(display_->compositor);
-    wl_region_add(window_->opaque_reqion, 0, 0,
-                  window_->properties.width,
-                  window_->properties.height);
-    wl_surface_set_opaque_region(window_->surface, window_->opaque_reqion);
-
-    window_->shell_surface = wl_shell_get_shell_surface(display_->shell,
-                                                        window_->surface);
+    window_->surface = wl_compositor_create_surface(display_->compositor);
+    window_->xdg_surface = xdg_wm_base_get_xdg_surface(display_->xdg_wm_base,
+                                                       window_->surface);
+    xdg_surface_add_listener(window_->xdg_surface, &xdg_surface_listener_, this);
+    window_->xdg_toplevel = xdg_surface_get_toplevel(window_->xdg_surface);
+    xdg_toplevel_add_listener(window_->xdg_toplevel, &xdg_toplevel_listener_, this);
+
+    xdg_toplevel_set_app_id(window_->xdg_toplevel, "com.github.glmark2.glmark2");
+    xdg_toplevel_set_title(window_->xdg_toplevel, "glmark2");
+    if (window_->properties.fullscreen && output)
+        xdg_toplevel_set_fullscreen(window_->xdg_toplevel, output->output);
+    wl_surface_commit(window_->surface);
+    window_->waiting_for_configure = true;
+
+    /* xdg-shell requires us to wait for the compositor to tell us what its
+     * desired size for us is */
+    while (window_->waiting_for_configure)
+        wl_display_roundtrip(display_->display);
+
+    if (output && window_->properties.fullscreen) {
+        if (window_->properties.width <= 0 ||
+             window_->properties.height <= 0) {
+            window_->properties.width = output->width * output->scale;
+            window_->properties.height = output->height * output->scale;
+        }
 
-    if (window_->shell_surface) {
-        wl_shell_surface_add_listener(window_->shell_surface,
-                                      &shell_surface_listener_, this);
+        if (wl_proxy_get_version((struct wl_proxy *) output) >=
+            WL_OUTPUT_SCALE_SINCE_VERSION) {
+            wl_surface_set_buffer_scale(window_->surface, output->scale);
+        }
     }
 
-    wl_shell_surface_set_title(window_->shell_surface, "glmark2");
+    window_->native = wl_egl_window_create(window_->surface,
+                                           window_->properties.width,
+                                           window_->properties.height);
 
-    if (window_->properties.fullscreen) {
-        wl_shell_surface_set_fullscreen(window_->shell_surface,
-                                        WL_SHELL_SURFACE_FULLSCREEN_METHOD_DRIVER,
-                                        output->refresh, output->output);
-    } else {
-        wl_shell_surface_set_toplevel(window_->shell_surface);
-    }
+    struct wl_region *opaque_reqion = wl_compositor_create_region(display_->compositor);
+    wl_region_add(opaque_reqion, 0, 0,
+                  window_->properties.width,
+                  window_->properties.height);
+    wl_surface_set_opaque_region(window_->surface, opaque_reqion);
+    wl_region_destroy(opaque_reqion);
 
     return true;
 }
@@ -298,3 +434,135 @@
     should_quit_ = true;
 }
 
+void
+NativeStateWayland::seat_handle_capabilities(void *data, struct wl_seat *seat, uint32_t caps)
+{
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    if ((caps & WL_SEAT_CAPABILITY_POINTER) && !that->display_->pointer) {
+        that->display_->pointer = wl_seat_get_pointer(seat);
+        wl_pointer_add_listener(that->display_->pointer, &pointer_listener_, that);
+    } else if (!(caps & WL_SEAT_CAPABILITY_POINTER) && that->display_->pointer) {
+        wl_pointer_destroy(that->display_->pointer);
+        that->display_->pointer = NULL;
+    }
+
+    if ((caps & WL_SEAT_CAPABILITY_KEYBOARD) && !that->display_->keyboard) {
+        that->display_->keyboard = wl_seat_get_keyboard(seat);
+        wl_keyboard_add_listener(that->display_->keyboard, &keyboard_listener_, that);
+    } else if (!(caps & WL_SEAT_CAPABILITY_KEYBOARD) && that->display_->keyboard) {
+        wl_keyboard_destroy(that->display_->keyboard);
+        that->display_->keyboard = NULL;
+    }
+}
+
+
+void
+NativeStateWayland::pointer_handle_enter(void *data, struct wl_pointer *pointer,
+                                         uint32_t serial, struct wl_surface *surface,
+                                         wl_fixed_t sx, wl_fixed_t sy)
+{
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    struct wl_buffer *buffer;
+    struct my_cursor *my_cursor = that->cursor_;
+    struct wl_cursor *cursor;
+    struct wl_cursor_image *image;
+
+    if (!my_cursor)
+        return;
+
+    cursor = my_cursor->default_cursor;
+
+    if (that->window_->properties.fullscreen) {
+        wl_pointer_set_cursor(pointer, serial, NULL, 0, 0);
+    } else {
+        if (cursor) {
+            image = my_cursor->default_cursor->images[0];
+            buffer = wl_cursor_image_get_buffer(image);
+            if (!buffer)
+                return;
+
+            wl_pointer_set_cursor(pointer, serial,
+                                  my_cursor->cursor_surface,
+                                  image->hotspot_x,
+                                  image->hotspot_y);
+            wl_surface_attach(my_cursor->cursor_surface, buffer, 0, 0);
+            wl_surface_damage(my_cursor->cursor_surface, 0, 0,
+                              image->width, image->height);
+            wl_surface_commit(my_cursor->cursor_surface);
+        }
+    }
+}
+
+void
+NativeStateWayland::pointer_handle_leave(void *data, struct wl_pointer *pointer,
+                     uint32_t serial, struct wl_surface *surface)
+{
+}
+
+void
+NativeStateWayland::pointer_handle_motion(void *data, struct wl_pointer *pointer,
+                      uint32_t time, wl_fixed_t sx, wl_fixed_t sy)
+{
+}
+
+void
+NativeStateWayland::pointer_handle_button(void *data, struct wl_pointer *wl_pointer,
+                                          uint32_t serial, uint32_t time, uint32_t button,
+                                          uint32_t state)
+{
+    /* just show that window_ can be used */
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    if (!that->window_->xdg_toplevel)
+        return;
+
+    if (state == WL_POINTER_BUTTON_STATE_PRESSED && button == BTN_LEFT)
+        xdg_toplevel_move(that->window_->xdg_toplevel,
+                          that->display_->seat, serial);
+}
+
+void
+NativeStateWayland::pointer_handle_axis(void *data, struct wl_pointer *wl_pointer,
+                                        uint32_t time, uint32_t axis, wl_fixed_t value)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_keymap(void *data, struct wl_keyboard *keyboard,
+                                           uint32_t format, int fd, uint32_t size)
+{
+    close(fd);
+}
+
+void
+NativeStateWayland::keyboard_handle_enter(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, struct wl_surface *surface,
+                                          struct wl_array *keys)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_leave(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, struct wl_surface *surface)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_key(void *data, struct wl_keyboard *keyboard,
+                                        uint32_t serial, uint32_t time, uint32_t key,
+                                        uint32_t state)
+{
+    if (state == WL_KEYBOARD_KEY_STATE_PRESSED &&
+        (key == KEY_ESC || key == KEY_Q))
+        should_quit_ = true;
+}
+
+void
+NativeStateWayland::keyboard_handle_modifiers(void *data, struct wl_keyboard *keyboard,
+                                              uint32_t serial, uint32_t mods_depressed,
+                                              uint32_t mods_latched, uint32_t mods_locked,
+                                              uint32_t group)
+{
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-wayland.h glmark2-2021.02/android/jni/src/native-state-wayland.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-wayland.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-wayland.h	2021-04-25 01:47:47.000000000 +0800
@@ -26,6 +26,8 @@
 #include <vector>
 #include <wayland-client.h>
 #include <wayland-egl.h>
+#include <wayland-cursor.h>
+#include "xdg-shell-client-protocol.h"
 
 #include "native-state.h"
 
@@ -47,8 +49,13 @@
     static void quit_handler(int signum);
 
     static const struct wl_registry_listener registry_listener_;
-    static const struct wl_shell_surface_listener shell_surface_listener_;
+    static const struct xdg_wm_base_listener xdg_wm_base_listener_;
+    static const struct xdg_surface_listener xdg_surface_listener_;
+    static const struct xdg_toplevel_listener xdg_toplevel_listener_;
     static const struct wl_output_listener output_listener_;
+    static const struct wl_seat_listener seat_listener_;
+    static const struct wl_pointer_listener pointer_listener_;
+    static const struct wl_keyboard_listener keyboard_listener_;
 
     static void
     registry_handle_global(void *data, struct wl_registry *registry,
@@ -76,39 +83,87 @@
     output_handle_scale(void *data, struct wl_output *wl_output, int32_t factor);
 
     static void
-    shell_surface_handle_ping(void *data, struct wl_shell_surface *shell_surface,
-                              uint32_t serial);
+    xdg_wm_base_handle_ping(void *data, struct xdg_wm_base *xdg_wm_base,
+                            uint32_t serial);
     static void
-    shell_surface_handle_configure(void *data,
-                                   struct wl_shell_surface *shell_surface,
-                                   uint32_t edges,
-                                   int32_t width, int32_t height);
-    static void
-    shell_surface_handle_popup_done(void *data,
-                                    struct wl_shell_surface *shell_surface);
-
+    xdg_toplevel_handle_configure(void *data,
+                                  struct xdg_toplevel *xdg_toplevel,
+                                  int32_t width, int32_t height,
+                                  struct wl_array *states);
+    static void
+    xdg_toplevel_handle_close(void *data, struct xdg_toplevel *xdg_toplevel);
+    static void
+    xdg_surface_handle_configure(void *data, struct xdg_surface *xdg_surface,
+                                 uint32_t serial);
+
+    static void seat_handle_capabilities(void *data,
+                                         struct wl_seat *seat, uint32_t caps);
+
+    static void pointer_handle_enter(void *data, struct wl_pointer *pointer,
+                                     uint32_t serial, struct wl_surface *surface,
+                                     wl_fixed_t sx, wl_fixed_t sy);
+
+    static void pointer_handle_leave(void *data, struct wl_pointer *pointer,
+                                     uint32_t serial, struct wl_surface *surface);
+
+    static void pointer_handle_motion(void *data, struct wl_pointer *pointer,
+                                      uint32_t time, wl_fixed_t sx, wl_fixed_t sy);
+
+    static void pointer_handle_button(void *data, struct wl_pointer *wl_pointer,
+                                      uint32_t serial, uint32_t time,
+                                      uint32_t button, uint32_t state);
+
+    static void pointer_handle_axis(void *data, struct wl_pointer *wl_pointer,
+                                    uint32_t time, uint32_t axis, wl_fixed_t value);
+
+    static void keyboard_handle_keymap(void *data, struct wl_keyboard *keyboard,
+                                       uint32_t format, int fd, uint32_t size);
+    static void keyboard_handle_enter(void *data, struct wl_keyboard *keyboard,
+                                      uint32_t serial, struct wl_surface *surface,
+                                      struct wl_array *keys);
+    static void keyboard_handle_leave(void *data, struct wl_keyboard *keyboard,
+                                      uint32_t serial, struct wl_surface *surface);
+    static void keyboard_handle_key(void *data, struct wl_keyboard *keyboard,
+                                    uint32_t serial, uint32_t time, uint32_t key,
+                                    uint32_t state);
+    static void keyboard_handle_modifiers(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, uint32_t mods_depressed,
+                                          uint32_t mods_latched, uint32_t mods_locked,
+                                          uint32_t group);
     struct my_output {
         wl_output *output;
         int32_t width, height;
         int32_t refresh;
+        int32_t scale;
     };
 
     typedef std::vector<struct my_output *> OutputsVector;
 
+    struct my_cursor {
+        wl_cursor_theme *cursor_theme;
+        wl_cursor *default_cursor;
+        wl_surface *cursor_surface;
+    } *cursor_;
+
     struct my_display {
         wl_display *display;
         wl_registry *registry;
         wl_compositor *compositor;
-        wl_shell *shell;
+        wl_shm *shm;
+        wl_seat *seat;
+        wl_pointer *pointer;
+        wl_keyboard *keyboard;
+        struct xdg_wm_base *xdg_wm_base;
         OutputsVector outputs;
     } *display_;
 
     struct my_window {
         WindowProperties properties;
+        bool waiting_for_configure;
         struct wl_surface *surface;
-        struct wl_region *opaque_reqion;
         struct wl_egl_window *native;
-        struct wl_shell_surface *shell_surface;
+        struct xdg_surface *xdg_surface;
+        struct xdg_toplevel *xdg_toplevel;
     } *window_;
 
     static volatile bool should_quit_;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-win32.cpp glmark2-2021.02/android/jni/src/native-state-win32.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-win32.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-win32.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,248 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#include "native-state-win32.h"
+
+static const char *win_name("glmark2 " GLMARK_VERSION);
+static const char *child_win_name("glmark2 " GLMARK_VERSION " child");
+
+/******************
+ * Public methods *
+ ******************/
+
+NativeStateWin32::NativeStateWin32() : parent_window_(0), child_window_(0), native_display_(0), should_quit_(false) {}
+
+NativeStateWin32::~NativeStateWin32() { cleanup(); }
+
+bool
+NativeStateWin32::init_display()
+{
+    cleanup();
+
+    static const LONG initial_width = 128;
+    static const LONG initial_height = 128;
+
+    // Work around compile error from not defining "UNICODE" while Chromium does
+    const LPSTR idcArrow = MAKEINTRESOURCEA(32512);
+
+    WNDCLASSEXA parentWindowClass = {};
+    parentWindowClass.cbSize = sizeof(WNDCLASSEXA);
+    parentWindowClass.style = 0;
+    parentWindowClass.lpfnWndProc = wnd_proc;
+    parentWindowClass.cbClsExtra = 0;
+    parentWindowClass.cbWndExtra = 0;
+    parentWindowClass.hInstance = GetModuleHandle(nullptr);
+    parentWindowClass.hIcon = nullptr;
+    parentWindowClass.hCursor = LoadCursorA(nullptr, idcArrow);
+    parentWindowClass.hbrBackground = 0;
+    parentWindowClass.lpszMenuName = nullptr;
+    parentWindowClass.lpszClassName = win_name;
+    if (!RegisterClassExA(&parentWindowClass))
+    {
+        return false;
+    }
+
+    WNDCLASSEXA childWindowClass = {};
+    childWindowClass.cbSize = sizeof(WNDCLASSEXA);
+    childWindowClass.style = CS_OWNDC;
+    childWindowClass.lpfnWndProc = wnd_proc;
+    childWindowClass.cbClsExtra = 0;
+    childWindowClass.cbWndExtra = 0;
+    childWindowClass.hInstance = GetModuleHandle(nullptr);
+    childWindowClass.hIcon = nullptr;
+    childWindowClass.hCursor = LoadCursorA(nullptr, idcArrow);
+    childWindowClass.hbrBackground = 0;
+    childWindowClass.lpszMenuName = nullptr;
+    childWindowClass.lpszClassName = child_win_name;
+    if (!RegisterClassExA(&childWindowClass))
+    {
+        return false;
+    }
+
+    DWORD parentStyle = WS_CAPTION | WS_THICKFRAME | WS_MINIMIZEBOX | WS_MAXIMIZEBOX | WS_SYSMENU;
+    DWORD parentExtendedStyle = WS_EX_APPWINDOW;
+
+    RECT sizeRect = { 0, 0, initial_width, initial_height };
+    AdjustWindowRectEx(&sizeRect, parentStyle, FALSE, parentExtendedStyle);
+
+    parent_window_ = CreateWindowExA(parentExtendedStyle, win_name, win_name,
+        parentStyle, CW_USEDEFAULT, CW_USEDEFAULT,
+        sizeRect.right - sizeRect.left, sizeRect.bottom - sizeRect.top,
+        nullptr, nullptr, GetModuleHandle(nullptr), this);
+
+    child_window_ = CreateWindowExA(0, child_win_name, win_name, WS_CHILD, 0, 0,
+        static_cast<int>(initial_width), static_cast<int>(initial_height),
+        parent_window_, nullptr, GetModuleHandle(nullptr), this);
+
+    native_display_ = GetDC(child_window_);
+    if (!native_display_)
+    {
+        cleanup();
+    }
+
+    return native_display_;
+}
+
+void*
+NativeStateWin32::display()
+{
+    return native_display_;
+}
+
+bool
+NativeStateWin32::create_window(WindowProperties const& properties)
+{
+    properties_ = properties;
+
+    RECT windowRect;
+    if (!GetWindowRect(parent_window_, &windowRect))
+    {
+        return false;
+    }
+
+    RECT clientRect;
+    if (!GetClientRect(parent_window_, &clientRect))
+    {
+        return false;
+    }
+
+    LONG diffX = (windowRect.right - windowRect.left) - clientRect.right;
+    LONG diffY = (windowRect.bottom - windowRect.top) - clientRect.bottom;
+    if (!MoveWindow(parent_window_, windowRect.left, windowRect.top, properties.width + diffX, properties.height + diffY,
+        TRUE))
+    {
+        return false;
+    }
+
+    if (!MoveWindow(child_window_, 0, 0, properties.width, properties.height, FALSE))
+    {
+        return false;
+    }
+
+    return true;
+}
+
+void*
+NativeStateWin32::window(WindowProperties& properties)
+{
+    properties = properties_;
+    return child_window_;
+}
+
+void
+NativeStateWin32::visible(bool visible)
+{
+    int flag = (visible ? SW_SHOW : SW_HIDE);
+
+    if (parent_window_)
+    {
+        ShowWindow(parent_window_, flag);
+    }
+
+    if (child_window_)
+    {
+        ShowWindow(child_window_, flag);
+    }
+}
+
+bool
+NativeStateWin32::should_quit()
+{
+    pump_messages();
+    return should_quit_;
+}
+
+void
+NativeStateWin32::flip()
+{
+    pump_messages();
+}
+
+/*******************
+ * Private methods *
+ *******************/
+
+void
+NativeStateWin32::cleanup()
+{
+    if (native_display_)
+    {
+        ReleaseDC(child_window_, native_display_);
+        native_display_ = 0;
+    }
+
+    if (child_window_)
+    {
+        DestroyWindow(child_window_);
+        child_window_ = 0;
+    }
+
+    if (parent_window_)
+    {
+        DestroyWindow(parent_window_);
+        parent_window_ = 0;
+    }
+
+    UnregisterClassA(win_name, nullptr);
+    UnregisterClassA(child_win_name, nullptr);
+}
+
+void
+NativeStateWin32::pump_messages()
+{
+    MSG msg;
+    while (PeekMessage(&msg, nullptr, 0, 0, PM_REMOVE))
+    {
+        TranslateMessage(&msg);
+        DispatchMessage(&msg);
+    }
+}
+
+LRESULT CALLBACK
+NativeStateWin32::wnd_proc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam)
+{
+    switch (message)
+    {
+        case WM_NCCREATE:
+        {
+            LPCREATESTRUCT pCreateStruct = reinterpret_cast<LPCREATESTRUCT>(lParam);
+            SetWindowLongPtr(hWnd, GWLP_USERDATA,
+                reinterpret_cast<LONG_PTR>(pCreateStruct->lpCreateParams));
+            return DefWindowProcA(hWnd, message, wParam, lParam);
+        }
+    }
+
+    NativeStateWin32 *window = reinterpret_cast<NativeStateWin32 *>(GetWindowLongPtr(hWnd, GWLP_USERDATA));
+    if (window)
+    {
+        switch (message)
+        {
+            case WM_CLOSE:
+            case WM_DESTROY:
+                window->should_quit_ = true;
+                break;
+            default:
+                break;
+        }
+    }
+
+    return DefWindowProcA(hWnd, message, wParam, lParam);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-win32.h glmark2-2021.02/android/jni/src/native-state-win32.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-win32.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-win32.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#ifndef GLMARK2_NATIVE_STATE_WIN32_H_
+#define GLMARK2_NATIVE_STATE_WIN32_H_
+
+#include "native-state.h"
+
+#include <windows.h>
+
+class NativeStateWin32 : public NativeState
+{
+public:
+    NativeStateWin32();
+    ~NativeStateWin32();
+
+    bool init_display();
+    void* display();
+    bool create_window(WindowProperties const& properties);
+    void* window(WindowProperties& properties);
+    void visible(bool v);
+    bool should_quit();
+    void flip();
+
+private:
+    void cleanup();
+    void pump_messages();
+
+    static LRESULT CALLBACK wnd_proc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam);
+
+    HWND parent_window_;
+    HWND child_window_;
+    HDC native_display_;
+    WindowProperties properties_;
+    bool should_quit_;
+};
+
+#endif /* GLMARK2_NATIVE_STATE_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-x11.cpp glmark2-2021.02/android/jni/src/native-state-x11.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/native-state-x11.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/native-state-x11.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -60,7 +60,7 @@
 bool
 NativeStateX11::create_window(WindowProperties const& properties)
 {
-    static const char *win_name("glmark2 "GLMARK_VERSION);
+    static const char *win_name("glmark2 " GLMARK_VERSION);
 
     if (!xdpy_) {
         Log::error("Error: X11 Display has not been initialized!\n");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/options.cpp glmark2-2021.02/android/jni/src/options.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/options.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/options.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -31,6 +31,7 @@
 std::vector<std::string> Options::benchmarks;
 std::vector<std::string> Options::benchmark_files;
 bool Options::validate = false;
+std::string Options::data_path = std::string(GLMARK_DATA_PATH);
 Options::FrameEnd Options::frame_end = Options::FrameEndDefault;
 std::pair<int,int> Options::size(800, 600);
 bool Options::list_scenes = false;
@@ -48,6 +49,7 @@
     {"benchmark", 1, 0, 0},
     {"benchmark-file", 1, 0, 0},
     {"validate", 0, 0, 0},
+    {"data-path", 1, 0, 0},
     {"frame-end", 1, 0, 0},
     {"off-screen", 0, 0, 0},
     {"visual-config", 1, 0, 0},
@@ -116,13 +118,15 @@
     printf("A benchmark for Open GL (ES) 2.0\n"
            "\n"
            "Options:\n"
-           "  -b, --benchmark BENCH  A benchmark to run: 'scene(:opt1=val1)*'\n"
+           "  -b, --benchmark BENCH  A benchmark or options to run: '(scene)?(:opt1=val1)*'\n"
            "                         (the option can be used multiple times)\n"
            "  -f, --benchmark-file F Load benchmarks to run from a file containing a\n"
            "                         list of benchmark descriptions (one per line)\n"
            "                         (the option can be used multiple times)\n"
            "      --validate         Run a quick output validation test instead of \n"
            "                         running the benchmarks\n"
+           "      --data-path PATH   Path to glmark2 models, shaders and textures\n"
+           "                         Default: " GLMARK_DATA_PATH "\n"
            "      --frame-end METHOD How to end a frame [default,none,swap,finish,readpixels]\n"
            "      --off-screen       Render to an off-screen surface\n"
            "      --visual-config C  The visual configuration to use for the rendering\n"
@@ -171,6 +175,8 @@
             Options::benchmark_files.push_back(optarg);
         else if (!strcmp(optname, "validate"))
             Options::validate = true;
+        else if (!strcmp(optname, "data-path"))
+            Options::data_path = std::string(optarg);
         else if (!strcmp(optname, "frame-end"))
             Options::frame_end = frame_end_from_str(optarg);
         else if (!strcmp(optname, "off-screen"))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/options.h glmark2-2021.02/android/jni/src/options.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/options.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/options.h	2021-04-25 01:47:22.000000000 +0800
@@ -43,6 +43,7 @@
     static std::vector<std::string> benchmarks;
     static std::vector<std::string> benchmark_files;
     static bool validate;
+    static std::string data_path;
     static FrameEnd frame_end;
     static std::pair<int,int> size;
     static bool list_scenes;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-buffer.cpp glmark2-2021.02/android/jni/src/scene-buffer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-buffer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-buffer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "util.h"
@@ -244,9 +245,9 @@
     {
         /* Set up shaders */
         static const std::string vtx_shader_filename(
-                GLMARK_DATA_PATH"/shaders/buffer-wireframe.vert");
+                Options::data_path + "/shaders/buffer-wireframe.vert");
         static const std::string frg_shader_filename(
-                GLMARK_DATA_PATH"/shaders/buffer-wireframe.frag");
+                Options::data_path + "/shaders/buffer-wireframe.frag");
 
         ShaderSource vtx_source(vtx_shader_filename);
         ShaderSource frg_source(frg_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-build.cpp glmark2-2021.02/android/jni/src/scene-build.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-build.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-build.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -25,6 +25,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "model.h"
@@ -89,8 +90,8 @@
         return false;
 
     /* Set up shaders */
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/light-basic.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     static const LibMatrix::vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-bump.cpp glmark2-2021.02/android/jni/src/scene-bump.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-bump.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-bump.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "model.h"
@@ -61,8 +62,8 @@
 bool
 SceneBump::setup_model_plain(const std::string &type)
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-poly.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-poly.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-poly.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-poly.frag");
     static const std::string low_poly_filename("asteroid-low");
     static const std::string high_poly_filename("asteroid-high");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
@@ -115,8 +116,8 @@
 bool
 SceneBump::setup_model_normals()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-normals.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-normals.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
@@ -170,8 +171,8 @@
 bool
 SceneBump::setup_model_normals_tangent()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals-tangent.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals-tangent.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-normals-tangent.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-normals-tangent.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
@@ -228,8 +229,8 @@
 bool
 SceneBump::setup_model_height()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-height.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-height.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-height.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-height.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-clear.cpp glmark2-2021.02/android/jni/src/scene-clear.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-clear.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-clear.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2013 Linaro Limited
+// Copyright © 2013 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-collection.h glmark2-2021.02/android/jni/src/scene-collection.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-collection.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-collection.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2013 Linaro Limited
+// Copyright © 2013 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-conditionals.cpp glmark2-2021.02/android/jni/src/scene-conditionals.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-conditionals.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-conditionals.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,6 +21,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -29,7 +30,7 @@
 
 #include <cmath>
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/conditionals");
+static const std::string shader_file_base("/shaders/conditionals");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -56,14 +57,14 @@
 static std::string
 get_vertex_shader_source(int steps, bool conditionals)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
 
     for (int i = 0; i < steps; i++) {
         if (conditionals)
-            source_main.append_file(step_conditional_file);
+            source_main.append_file(Options::data_path + step_conditional_file);
         else
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
@@ -74,14 +75,14 @@
 static std::string
 get_fragment_shader_source(int steps, bool conditionals)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
 
     for (int i = 0; i < steps; i++) {
         if (conditionals)
-            source_main.append_file(step_conditional_file);
+            source_main.append_file(Options::data_path + step_conditional_file);
         else
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene.cpp glmark2-2021.02/android/jni/src/scene.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -43,10 +43,12 @@
 Scene::Scene(Canvas &pCanvas, const string &name) :
     canvas_(pCanvas), name_(name),
     startTime_(0), lastUpdateTime_(0), currentFrame_(0),
-    running_(0), duration_(0)
+    running_(0), duration_(0), nframes_(0)
 {
     options_["duration"] = Scene::Option("duration", "10.0",
                                          "The duration of each benchmark in seconds");
+    options_["nframes"] = Scene::Option("nframes", "",
+                                         "The number of frames to render");
     options_["vertex-precision"] = Scene::Option("vertex-precision",
                                                  "default,default,default,default",
                                                  "The precision values for the vertex shader (\"int,float,sampler2d,samplercube\")");
@@ -96,6 +98,7 @@
 Scene::setup()
 {
     duration_ = Util::fromString<double>(options_["duration"].value);
+    nframes_ = Util::fromString<unsigned>(options_["nframes"].value);
 
     ShaderSource::default_precision(
             ShaderSource::Precision(options_["vertex-precision"].value),
@@ -132,6 +135,9 @@
 
     if (elapsed_time >= duration_)
         running_ = false;
+
+    if (nframes_ > 0 && currentFrame_ >= nframes_)
+        running_ = false;
 }
 
 void
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-desktop.cpp glmark2-2021.02/android/jni/src/scene-desktop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-desktop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-desktop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -20,11 +20,13 @@
  *  Alexandros Frantzis (glmark2)
  *  Jesse Barker (glmark2)
  */
+#include <algorithm>
 #include <cmath>
 #include <cstdlib>
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -43,8 +45,8 @@
 create_blur_shaders(ShaderSource& vtx_source, ShaderSource& frg_source,
                     unsigned int radius, float sigma, BlurDirection direction)
 {
-    vtx_source.append_file(GLMARK_DATA_PATH"/shaders/desktop.vert");
-    frg_source.append_file(GLMARK_DATA_PATH"/shaders/desktop-blur.frag");
+    vtx_source.append_file(Options::data_path + "/shaders/desktop.vert");
+    frg_source.append_file(Options::data_path + "/shaders/desktop-blur.frag");
 
     /* Don't let the gaussian curve become too narrow */
     if (sigma < 1.0)
@@ -122,6 +124,9 @@
         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 
+        GLint prev_fbo;
+        glGetIntegerv(GL_FRAMEBUFFER_BINDING, &prev_fbo);
+
         /* Create a FBO */
         glGenFramebuffers(1, &fbo_);
         glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
@@ -144,12 +149,12 @@
             }
         }
 
-        glBindFramebuffer(GL_FRAMEBUFFER, 0);
+        glBindFramebuffer(GL_FRAMEBUFFER, prev_fbo);
 
         /* Load the shader program when this class if first used */
         if (RenderObject::use_count == 0) {
-            ShaderSource vtx_source(GLMARK_DATA_PATH"/shaders/desktop.vert");
-            ShaderSource frg_source(GLMARK_DATA_PATH"/shaders/desktop.frag");
+            ShaderSource vtx_source(Options::data_path + "/shaders/desktop.vert");
+            ShaderSource frg_source(Options::data_path + "/shaders/desktop.frag");
             Scene::load_shaders_from_strings(main_program, vtx_source.str(),
                                              frg_source.str());
         }
@@ -196,6 +201,8 @@
         /* Recreate the backing texture with correct size */
         if (size_.x() != size.x() || size_.y() != size.y()) {
             size_ = size;
+            if (fbo_ == 0)
+                return;
             /* If we're resizing the texture, we need to tell the framebuffer*/
             glBindTexture(GL_TEXTURE_2D, texture_);
             glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, size_.x(), size_.y(), 0,
@@ -377,9 +384,11 @@
  */
 class RenderScreen : public RenderObject
 {
+    Canvas &canvas_;
 public:
-    RenderScreen(Canvas &canvas) { fbo_ = canvas.fbo(); }
-    virtual void init() {}
+    RenderScreen(Canvas &canvas) : canvas_(canvas) {}
+    virtual void init() { fbo_ = canvas_.fbo(); }
+    virtual void size(const LibMatrix::vec2& size) { size_ = size; clear(); }
     virtual void release() {}
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-effect-2d.cpp glmark2-2021.02/android/jni/src/scene-effect-2d.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-effect-2d.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-effect-2d.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,6 +26,7 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -96,7 +97,7 @@
 create_convolution_fragment_shader(Canvas &canvas, std::vector<float> &array,
                                    unsigned int width, unsigned int height)
 {
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/effect-2d-convolution.frag");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/effect-2d-convolution.frag");
     ShaderSource source(frg_shader_filename);
 
     if (width * height != array.size()) {
@@ -303,7 +304,7 @@
 
     Texture::find_textures();
 
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/effect-2d.vert");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/effect-2d.vert");
 
     std::vector<float> kernel;
     unsigned int kernel_width = 0;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-function.cpp glmark2-2021.02/android/jni/src/scene-function.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-function.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-function.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,13 +23,14 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
 #include "shader-source.h"
 #include "util.h"
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/function");
+static const std::string shader_file_base("/shaders/function");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -61,7 +62,7 @@
 static std::string
 get_vertex_shader_source(int steps, bool function, std::string &complexity)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
     std::string step_file;
 
@@ -72,13 +73,13 @@
 
     for (int i = 0; i < steps; i++) {
         if (function)
-            source_main.append_file(call_file);
+            source_main.append_file(Options::data_path + call_file);
         else
-            source_main.append_file(step_file);
+            source_main.append_file(Options::data_path + step_file);
     }
 
     if (function)
-        source.replace_with_file("$PROCESS$", step_file);
+        source.replace_with_file("$PROCESS$", Options::data_path + step_file);
     else
         source.replace("$PROCESS$", "");
 
@@ -90,7 +91,7 @@
 static std::string
 get_fragment_shader_source(int steps, bool function, std::string &complexity)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
     std::string step_file;
 
@@ -101,13 +102,13 @@
 
     for (int i = 0; i < steps; i++) {
         if (function)
-            source_main.append_file(call_file);
+            source_main.append_file(Options::data_path + call_file);
         else
-            source_main.append_file(step_file);
+            source_main.append_file(Options::data_path + step_file);
     }
 
     if (function)
-        source.replace_with_file("$PROCESS$", step_file);
+        source.replace_with_file("$PROCESS$", Options::data_path + step_file);
     else
         source.replace("$PROCESS$", "");
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene.h glmark2-2021.02/android/jni/src/scene.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene.h	2021-04-25 01:47:22.000000000 +0800
@@ -234,6 +234,7 @@
     unsigned currentFrame_;
     bool running_;
     double duration_;      // Duration of run in seconds
+    unsigned nframes_;
 };
 
 /*
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/a.cc glmark2-2021.02/android/jni/src/scene-ideas/a.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/a.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/a.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'a'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/characters.h glmark2-2021.02/android/jni/src/scene-ideas/characters.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/characters.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/characters.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/d.cc glmark2-2021.02/android/jni/src/scene-ideas/d.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/d.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/d.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'd'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/e.cc glmark2-2021.02/android/jni/src/scene-ideas/e.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/e.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/e.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'e'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/i.cc glmark2-2021.02/android/jni/src/scene-ideas/i.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/i.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/i.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'i'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/lamp.cc glmark2-2021.02/android/jni/src/scene-ideas/lamp.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/lamp.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/lamp.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the lamp
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -23,6 +23,7 @@
  *  Jesse Barker
  */
 #include "lamp.h"
+#include "options.h"
 #include "shader-source.h"
 #include "log.h"
 #include "scene.h"
@@ -170,8 +171,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // The program for handling lighting...
-    string lit_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-lit.vert");
-    string lit_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-lit.frag");
+    string lit_vtx_filename(Options::data_path + "/shaders/ideas-lamp-lit.vert");
+    string lit_frg_filename(Options::data_path + "/shaders/ideas-lamp-lit.frag");
     ShaderSource lit_vtx_source(lit_vtx_filename);
     ShaderSource lit_frg_source(lit_frg_filename);
     if (!Scene::load_shaders_from_strings(litProgram_, lit_vtx_source.str(),
@@ -182,8 +183,8 @@
     }
 
     // The simple program with no lighting...
-    string unlit_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-unlit.vert");
-    string unlit_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-unlit.frag");
+    string unlit_vtx_filename(Options::data_path + "/shaders/ideas-lamp-unlit.vert");
+    string unlit_frg_filename(Options::data_path + "/shaders/ideas-lamp-unlit.frag");
     ShaderSource unlit_vtx_source(unlit_vtx_filename);
     ShaderSource unlit_frg_source(unlit_frg_filename);
     if (!Scene::load_shaders_from_strings(unlitProgram_, unlit_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/lamp.h glmark2-2021.02/android/jni/src/scene-ideas/lamp.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/lamp.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/lamp.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/logo.cc glmark2-2021.02/android/jni/src/scene-ideas/logo.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/logo.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/logo.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the old Silicon Graphics logo
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -23,6 +23,7 @@
  *  Jesse Barker
  */
 #include "logo.h"
+#include "options.h"
 #include "scene.h"
 #include "shader-source.h"
 #include "log.h"
@@ -423,8 +424,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // The program for handling the main object with lighting...
-    string logo_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo.vert");
-    string logo_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo.frag");
+    string logo_vtx_filename(Options::data_path + "/shaders/ideas-logo.vert");
+    string logo_frg_filename(Options::data_path + "/shaders/ideas-logo.frag");
     ShaderSource logo_vtx_source(logo_vtx_filename);
     ShaderSource logo_frg_source(logo_frg_filename);
     if (!Scene::load_shaders_from_strings(normalProgram_, logo_vtx_source.str(),
@@ -437,8 +438,8 @@
     normalNormalIndex_ = normalProgram_[normalAttribName_].location();
 
     // The program for handling the flat object...
-    string logo_flat_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-flat.vert");
-    string logo_flat_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-flat.frag");
+    string logo_flat_vtx_filename(Options::data_path + "/shaders/ideas-logo-flat.vert");
+    string logo_flat_frg_filename(Options::data_path + "/shaders/ideas-logo-flat.frag");
     ShaderSource logo_flat_vtx_source(logo_flat_vtx_filename);
     ShaderSource logo_flat_frg_source(logo_flat_frg_filename);
     if (!Scene::load_shaders_from_strings(flatProgram_, logo_flat_vtx_source.str(),
@@ -450,8 +451,8 @@
     flatVertexIndex_ = flatProgram_[vertexAttribName_].location();
 
     // The program for handling the shadow object with texturing...
-    string logo_shadow_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-shadow.vert");
-    string logo_shadow_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-shadow.frag");
+    string logo_shadow_vtx_filename(Options::data_path + "/shaders/ideas-logo-shadow.vert");
+    string logo_shadow_frg_filename(Options::data_path + "/shaders/ideas-logo-shadow.frag");
     ShaderSource logo_shadow_vtx_source(logo_shadow_vtx_filename);
     ShaderSource logo_shadow_frg_source(logo_shadow_frg_filename);
     if (!Scene::load_shaders_from_strings(shadowProgram_, logo_shadow_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/logo.h glmark2-2021.02/android/jni/src/scene-ideas/logo.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/logo.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/logo.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/m.cc glmark2-2021.02/android/jni/src/scene-ideas/m.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/m.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/m.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'm'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/n.cc glmark2-2021.02/android/jni/src/scene-ideas/n.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/n.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/n.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'o'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/o.cc glmark2-2021.02/android/jni/src/scene-ideas/o.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/o.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/o.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'o'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/s.cc glmark2-2021.02/android/jni/src/scene-ideas/s.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/s.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/s.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 's'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/splines.cc glmark2-2021.02/android/jni/src/scene-ideas/splines.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/splines.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/splines.cc	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/splines.h glmark2-2021.02/android/jni/src/scene-ideas/splines.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/splines.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/splines.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/table.cc glmark2-2021.02/android/jni/src/scene-ideas/table.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/table.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/table.cc	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -20,6 +20,7 @@
  * Authors:
  *  Jesse Barker
  */
+#include "options.h"
 #include "table.h"
 #include "scene.h"
 #include "shader-source.h"
@@ -101,8 +102,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // Program to render the table with lighting and a time-based fade...
-    string table_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-table.vert");
-    string table_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-table.frag");
+    string table_vtx_filename(Options::data_path + "/shaders/ideas-table.vert");
+    string table_frg_filename(Options::data_path + "/shaders/ideas-table.frag");
     ShaderSource table_vtx_source(table_vtx_filename);
     ShaderSource table_frg_source(table_frg_filename);
     if (!Scene::load_shaders_from_strings(tableProgram_, table_vtx_source.str(),
@@ -111,11 +112,11 @@
         Log::error("No valid program for table rendering.\n");
         return;
     }
-    textVertexIndex_ = tableProgram_[vertexAttribName_].location();
+    tableVertexIndex_ = tableProgram_[vertexAttribName_].location();
 
     // Program to render the paper with lighting and a time-based fade...
-    string paper_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-paper.vert");
-    string paper_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-paper.frag");
+    string paper_vtx_filename(Options::data_path + "/shaders/ideas-paper.vert");
+    string paper_frg_filename(Options::data_path + "/shaders/ideas-paper.frag");
     ShaderSource paper_vtx_source(paper_vtx_filename);
     ShaderSource paper_frg_source(paper_frg_filename);
     if (!Scene::load_shaders_from_strings(paperProgram_, paper_vtx_source.str(),
@@ -127,8 +128,8 @@
     paperVertexIndex_ = paperProgram_[vertexAttribName_].location();
 
     // Program to handle the text (time-based color fade)...
-    string text_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-text.vert");
-    string text_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-text.frag");
+    string text_vtx_filename(Options::data_path + "/shaders/ideas-text.vert");
+    string text_frg_filename(Options::data_path + "/shaders/ideas-text.frag");
     ShaderSource text_vtx_source(text_vtx_filename);
     ShaderSource text_frg_source(text_frg_filename);
     if (!Scene::load_shaders_from_strings(textProgram_, text_vtx_source.str(),
@@ -140,8 +141,8 @@
     textVertexIndex_ = textProgram_[vertexAttribName_].location();
 
     // Program for the drawUnder functionality (just paint it black)...
-    string under_table_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-under-table.vert");
-    string under_table_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-under-table.frag");
+    string under_table_vtx_filename(Options::data_path + "/shaders/ideas-under-table.vert");
+    string under_table_frg_filename(Options::data_path + "/shaders/ideas-under-table.frag");
     ShaderSource under_table_vtx_source(under_table_vtx_filename);
     ShaderSource under_table_frg_source(under_table_frg_filename);
     if (!Scene::load_shaders_from_strings(underProgram_, under_table_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/table.h glmark2-2021.02/android/jni/src/scene-ideas/table.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/table.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/table.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/t.cc glmark2-2021.02/android/jni/src/scene-ideas/t.cc
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas/t.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas/t.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 't'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas.cpp glmark2-2021.02/android/jni/src/scene-ideas.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-ideas.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-ideas.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -48,7 +48,7 @@
         timeOffset_(START_TIME_)
     {
         startTime_.tv_sec = 0;
-        startTime_.tv_usec = 0; 
+        startTime_.tv_nsec = 0; 
     }
     ~SceneIdeasPrivate()
     {
@@ -69,7 +69,7 @@
     float currentSpeed_;
     float currentTime_;
     float timeOffset_;
-    struct timeval startTime_;
+    struct timespec startTime_;
     static const float CYCLE_TIME_;
     static const float TIME_;
     static const float START_TIME_;
@@ -163,17 +163,17 @@
 SceneIdeasPrivate::reset_time()
 {
     timeOffset_ = START_TIME_;
-    gettimeofday(&startTime_, NULL);
+    clock_gettime(CLOCK_MONOTONIC, &startTime_);
 }
 
 void
 SceneIdeasPrivate::update_time()
 {
     // Compute new time
-    struct timeval current = {0, 0};
-    gettimeofday(&current, NULL);
+    struct timespec current = {0, 0};
+    clock_gettime(CLOCK_MONOTONIC, &current);
     float timediff = (current.tv_sec - startTime_.tv_sec) + 
-        static_cast<double>(current.tv_usec - startTime_.tv_usec) / 1000000.0;
+        static_cast<double>(current.tv_nsec - startTime_.tv_nsec) / 1000000000.0;
     float sceneTime = timediff * currentSpeed_ + timeOffset_;
 
     // Keep the current time in [START_TIME_..CYCLE_TIME_)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-jellyfish.cpp glmark2-2021.02/android/jni/src/scene-jellyfish.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-jellyfish.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-jellyfish.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -24,6 +24,7 @@
 #include <fstream>
 #include <memory>
 #include <iomanip>
+#include "options.h"
 #include "scene.h"
 #include "scene-jellyfish.h"
 #include "log.h"
@@ -115,8 +116,8 @@
 GradientRenderer::init()
 {
     // Program set up
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/gradient.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/gradient.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/gradient.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/gradient.frag");
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
     if (!Scene::load_shaders_from_strings(program_, vtx_source.str(),
@@ -272,7 +273,7 @@
 {
     Log::debug("Loading model from file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& inputFile(*input_file_ptr);
     if (!inputFile)
     {
@@ -375,7 +376,7 @@
 bool
 JellyfishPrivate::initialize()
 {
-    static const string modelFilename(GLMARK_DATA_PATH"/models/jellyfish.jobj");
+    static const string modelFilename(Options::data_path + "/models/jellyfish.jobj");
     if (!load_obj(modelFilename))
     {
         return false;
@@ -410,8 +411,8 @@
     // Set up program first so we can store attribute and uniform locations
     // away for the 
     using std::string;
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/jellyfish.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/jellyfish.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/jellyfish.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/jellyfish.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-jellyfish.h glmark2-2021.02/android/jni/src/scene-jellyfish.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-jellyfish.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-jellyfish.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-loop.cpp glmark2-2021.02/android/jni/src/scene-loop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-loop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-loop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,13 +23,14 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
 #include "shader-source.h"
 #include "util.h"
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/loop");
+static const std::string shader_file_base("/shaders/loop");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -62,11 +63,11 @@
 static std::string
 get_fragment_shader_source(int steps, bool loop, bool uniform)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
 
     if (loop) {
-        source_main.append_file(step_loop_file);
+        source_main.append_file(Options::data_path + step_loop_file);
         if (uniform) {
             source_main.replace("$NLOOPS$", "FragmentLoops");
         }
@@ -76,7 +77,7 @@
     }
     else {
         for (int i = 0; i < steps; i++)
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
@@ -87,11 +88,11 @@
 static std::string
 get_vertex_shader_source(int steps, bool loop, bool uniform)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
 
     if (loop) {
-        source_main.append_file(step_loop_file);
+        source_main.append_file(Options::data_path + step_loop_file);
         if (uniform) {
             source_main.replace("$NLOOPS$", "VertexLoops");
         }
@@ -101,7 +102,7 @@
     }
     else {
         for (int i = 0; i < steps; i++)
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-pulsar.cpp glmark2-2021.02/android/jni/src/scene-pulsar.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-pulsar.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-pulsar.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,6 +26,7 @@
 #include <stdlib.h>
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -112,19 +113,19 @@
     std::string frg_shader_filename;
     static const vec4 lightPosition(-20.0f, 20.0f,-20.0f, 1.0f);
     if (options_["light"].value == "true") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/pulsar-light.vert";
+        vtx_shader_filename = Options::data_path + "/shaders/pulsar-light.vert";
     } else {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/pulsar.vert";
+        vtx_shader_filename = Options::data_path + "/shaders/pulsar.vert";
     }
 
     if (options_["texture"].value == "true") {
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic-tex.frag";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic-tex.frag";
         Texture::find_textures();
         if (!Texture::load("crate-base", &texture_, GL_NEAREST, GL_NEAREST, 0))
             return false;
 
     } else {
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.frag";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic.frag";
     }
 
     ShaderSource vtx_source(vtx_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-refract.cpp glmark2-2021.02/android/jni/src/scene-refract.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-refract.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-refract.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -21,6 +21,7 @@
 //
 #include "scene-refract.h"
 #include "model.h"
+#include "options.h"
 #include "texture.h"
 #include "util.h"
 #include "log.h"
@@ -176,10 +177,10 @@
 //
 
 bool
-DistanceRenderTarget::setup(unsigned int width, unsigned int height)
+DistanceRenderTarget::setup(unsigned int canvas_fbo, unsigned int width, unsigned int height)
 {
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/depth.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/depth.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/depth.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/depth.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
@@ -192,6 +193,7 @@
     canvas_height_ = height;
     width_ = canvas_width_ * 2;
     height_ = canvas_height_ * 2;
+    canvas_fbo_ = canvas_fbo;
 
     // If the texture will be too large for the implemnetation, we need to
     // clamp the dimensions but maintain the aspect ratio.
@@ -235,7 +237,7 @@
         Log::error("DistanceRenderTarget::setup: glCheckFramebufferStatus failed (0x%x)\n", status);
         return false;
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
 
     return true;
 }
@@ -272,7 +274,7 @@
 
 void DistanceRenderTarget::disable()
 {
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
     glViewport(0, 0, canvas_width_, canvas_height_);
     glCullFace(GL_BACK);
 }
@@ -281,8 +283,8 @@
 RefractPrivate::setup(map<string, Scene::Option>& options)
 {
     // Program object setup
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-refract.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-refract.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/light-refract.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/light-refract.frag");
     static const vec4 lightColor(0.4, 0.4, 0.4, 1.0);
 
     ShaderSource vtx_source(vtx_shader_filename);
@@ -382,7 +384,7 @@
                                       0.0, 0.0, 0.0,
                                       0.0, 1.0, 0.0);
 
-    if (!depthTarget_.setup(canvas_.width(), canvas_.height())) {
+    if (!depthTarget_.setup(canvas_.fbo(), canvas_.width(), canvas_.height())) {
         Log::error("Failed to set up the render target for the depth pass\n");
         return false;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-refract.h glmark2-2021.02/android/jni/src/scene-refract.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-refract.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-refract.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -46,6 +46,7 @@
     unsigned int height_;
     unsigned int tex_[2];
     unsigned int fbo_;
+    unsigned int canvas_fbo_;
 public:
     DistanceRenderTarget() :
         canvas_width_(0),
@@ -57,7 +58,7 @@
         tex_[DEPTH] = tex_[COLOR] = 0;
     }
     ~DistanceRenderTarget() {}
-    bool setup(unsigned int width, unsigned int height);
+    bool setup(unsigned int canvas_fbo, unsigned int width, unsigned int height);
     void teardown();
     void enable(const LibMatrix::mat4& mvp);
     void disable();
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-shading.cpp glmark2-2021.02/android/jni/src/scene-shading.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-shading.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-shading.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,6 +24,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -154,24 +155,24 @@
     ShaderSource vtx_source;
     ShaderSource frg_source;
     if (shading == "gouraud") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-basic.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic.frag";
         frg_source.append_file(frg_shader_filename);
         vtx_source.append_file(vtx_shader_filename);
         vtx_source.add_const("LightSourcePosition", lightPosition);
         vtx_source.add_const("MaterialDiffuse", materialDiffuse);
     }
     else if (shading == "blinn-phong-inf") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-advanced.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-advanced.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-advanced.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-advanced.frag";
         frg_source.append_file(frg_shader_filename);
         frg_source.add_const("LightSourcePosition", lightPosition);
         frg_source.add_const("LightSourceHalfVector", halfVector);
         vtx_source.append_file(vtx_shader_filename);
     }
     else if (shading == "phong") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-phong.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-phong.frag";
         unsigned int num_lights = Util::fromString<unsigned int>(options_["num-lights"].value);
         string fragsource = get_fragment_shader_source(frg_shader_filename, num_lights);
         frg_source.append(fragsource);
@@ -179,8 +180,8 @@
         vtx_source.append_file(vtx_shader_filename);
     }
     else if (shading == "cel") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-cel.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-phong.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-cel.frag";
         vtx_source.append_file(vtx_shader_filename);
         frg_source.append_file(frg_shader_filename);
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-shadow.cpp glmark2-2021.02/android/jni/src/scene-shadow.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-shadow.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-shadow.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -21,6 +21,7 @@
 //
 #include "scene.h"
 #include "model.h"
+#include "options.h"
 #include "util.h"
 #include "log.h"
 #include "shader-source.h"
@@ -53,6 +54,7 @@
     unsigned int height_;
     unsigned int tex_;
     unsigned int fbo_;
+    unsigned int canvas_fbo_;
 public:
     DepthRenderTarget() :
         canvas_width_(0),
@@ -62,7 +64,7 @@
         tex_(0),
         fbo_(0) {}
     ~DepthRenderTarget() {}
-    bool setup(unsigned int width, unsigned int height);
+    bool setup(unsigned int canvas_fbo, unsigned int width, unsigned int height);
     void teardown();
     void enable(const mat4& mvp);
     void disable();
@@ -71,10 +73,10 @@
 };
 
 bool
-DepthRenderTarget::setup(unsigned int width, unsigned int height)
+DepthRenderTarget::setup(unsigned int canvas_fbo, unsigned int width, unsigned int height)
 {
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/depth.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/depth.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/depth.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/depth.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
@@ -85,6 +87,7 @@
 
     canvas_width_ = width;
     canvas_height_ = height;
+    canvas_fbo_ = canvas_fbo;
     width_ = canvas_width_ * 2;
     height_ = canvas_height_ * 2;
 
@@ -120,7 +123,7 @@
         Log::error("DepthRenderTarget::setup: glCheckFramebufferStatus failed (0x%x)\n", status);
         return false;
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
 
     return true;
 }
@@ -155,7 +158,7 @@
 
 void DepthRenderTarget::disable()
 {
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
     glViewport(0, 0, canvas_width_, canvas_height_);
     glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
 }
@@ -201,8 +204,8 @@
 
     // Program set up
     static const vec4 materialDiffuse(0.3f, 0.3f, 0.3f, 1.0f);
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/shadow.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/shadow.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/shadow.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/shadow.frag");
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
 
@@ -314,8 +317,8 @@
 ShadowPrivate::setup(map<string, Scene::Option>& options)
 {
     // Program object setup
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/light-basic.frag");
     static const vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
     ShaderSource vtx_source(vtx_shader_filename);
@@ -372,7 +375,7 @@
     float aspect(static_cast<float>(canvas_.width())/static_cast<float>(canvas_.height()));
     projection_.perspective(fovy, aspect, 2.0, 50.0);
 
-    if (!depthTarget_.setup(canvas_.width(), canvas_.height())) {
+    if (!depthTarget_.setup(canvas_.fbo(), canvas_.width(), canvas_.height())) {
         Log::error("Failed to set up the render target for the depth pass\n");
         return false;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/base-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/base-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/base-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/base-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,26 +21,33 @@
  */
 #include "renderer.h"
 
-BaseRenderer::BaseRenderer(const LibMatrix::vec2 &size) :
+BaseRenderer::BaseRenderer() :
     texture_(0), input_texture_(0), fbo_(0), depth_renderbuffer_(0),
-    min_filter_(GL_LINEAR), mag_filter_(GL_LINEAR),
+    owns_fbo_(false), min_filter_(GL_LINEAR), mag_filter_(GL_LINEAR),
     wrap_s_(GL_CLAMP_TO_EDGE), wrap_t_(GL_CLAMP_TO_EDGE)
 {
-    setup(size, true, true);
 }
 
 BaseRenderer::~BaseRenderer()
 {
     glDeleteTextures(1, &texture_);
     glDeleteRenderbuffers(1, &depth_renderbuffer_);
-    glDeleteFramebuffers(1, &fbo_);
+    if (owns_fbo_)
+        glDeleteFramebuffers(1, &fbo_);
+}
+
+void
+BaseRenderer::setup_onscreen(Canvas& canvas)
+{
+    size_ = LibMatrix::vec2(canvas.width(), canvas.height());
+    recreate(&canvas, false);
 }
 
 void
-BaseRenderer::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+BaseRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     size_ = size;
-    recreate(onscreen, has_depth);
+    recreate(nullptr, has_depth);
 }
 
 void
@@ -59,7 +66,7 @@
 {
     glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
     glViewport(0, 0, size_.x(), size_.y());
-    if (!fbo_ || depth_renderbuffer_) {
+    if (!owns_fbo_ || depth_renderbuffer_) {
         glEnable(GL_DEPTH_TEST);
         glDepthMask(GL_TRUE);
     }
@@ -79,21 +86,26 @@
 }
 
 void
-BaseRenderer::recreate(bool onscreen, bool has_depth)
+BaseRenderer::recreate(Canvas* canvas, bool has_depth)
 {
     if (texture_) {
         glDeleteTextures(1, &texture_);
         texture_ = 0;
     }
-    if (fbo_) {
+    if (owns_fbo_ && fbo_) {
         glDeleteRenderbuffers(1, &depth_renderbuffer_);
         depth_renderbuffer_ = 0;
         glDeleteFramebuffers(1, &fbo_);
         fbo_ = 0;
     }
-    if (!onscreen) {
+
+    if (canvas) {
+        fbo_ = canvas->fbo();
+        owns_fbo_ = false;
+    } else {
         create_texture();
         create_fbo(has_depth);
+        owns_fbo_ = true;
     }
 }
 
@@ -131,6 +143,9 @@
                 size_.x(), size_.y());
     }
 
+    GLint prev_fbo;
+    glGetIntegerv(GL_FRAMEBUFFER_BINDING, &prev_fbo);
+
     /* Create the FBO and attach the texture and the renderebuffer */
     glGenFramebuffers(1, &fbo_);
     glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
@@ -140,5 +155,6 @@
         glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT,
                 GL_RENDERBUFFER, depth_renderbuffer_);
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+
+    glBindFramebuffer(GL_FRAMEBUFFER, prev_fbo);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/blur-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/blur-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/blur-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/blur-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,18 +19,21 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
+#include <cstdlib>
+
 void
 create_blur_shaders(ShaderSource& vtx_source, ShaderSource& frg_source,
                     unsigned int radius, float sigma, BlurRenderer::BlurDirection direction,
                     float tilt_shift);
 
-BlurRenderer::BlurRenderer(const LibMatrix::vec2 &size, int radius, float sigma,
+BlurRenderer::BlurRenderer(int radius, float sigma,
                            BlurDirection dir, const LibMatrix::vec2 &step, float tilt_shift) :
-        TextureRenderer(size, *blur_program(true, radius, sigma, dir, step, tilt_shift))
+        TextureRenderer(*blur_program(true, radius, sigma, dir, step, tilt_shift))
 {
     blur_program_ = blur_program(false, radius, sigma, dir, step, tilt_shift);
 }
@@ -74,15 +77,15 @@
                     unsigned int radius, float sigma, BlurRenderer::BlurDirection direction,
                     float tilt_shift)
 {
-    vtx_source.append_file(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-    frg_source.append_file(GLMARK_DATA_PATH"/shaders/terrain-blur.frag");
+    vtx_source.append_file(Options::data_path + "/shaders/terrain-texture.vert");
+    frg_source.append_file(Options::data_path + "/shaders/terrain-blur.frag");
 
     /* Don't let the gaussian curve become too narrow */
     if (sigma < 1.0)
         sigma = 1.0;
 
     unsigned int side = 2 * radius + 1;
-    float values[radius];
+    std::vector<float> values(radius + 1);
     float sum = 0.0;
 
     for (unsigned int i = 0; i < radius + 1; i++) {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/copy-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/copy-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/copy-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/copy-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,12 +19,13 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-CopyRenderer::CopyRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *copy_program(true))
+CopyRenderer::CopyRenderer() :
+    TextureRenderer(*copy_program(true))
 {
     copy_program_ = copy_program(false);
 }
@@ -38,8 +39,8 @@
 
     if (!copy_program) {
         copy_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-overlay.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-overlay.frag");
 
         Scene::load_shaders_from_strings(*copy_program, vtx_shader.str(), frg_shader.str());
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/luminance-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/luminance-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/luminance-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/luminance-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,12 +19,13 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-LuminanceRenderer::LuminanceRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *luminance_program(true))
+LuminanceRenderer::LuminanceRenderer() :
+    TextureRenderer(*luminance_program(true))
 {
     luminance_program_ = luminance_program(false);
 }
@@ -38,8 +39,8 @@
 
     if (!luminance_program) {
         luminance_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-luminance.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-luminance.frag");
 
         Scene::load_shaders_from_strings(*luminance_program,
                                          vtx_shader.str(), frg_shader.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/normal-from-height-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/normal-from-height-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/normal-from-height-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/normal-from-height-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,19 +19,40 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-NormalFromHeightRenderer::NormalFromHeightRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *normal_from_height_program(size, true))
+NormalFromHeightRenderer::NormalFromHeightRenderer() :
+    TextureRenderer(*normal_from_height_program(true))
 {
-    normal_from_height_program_ = normal_from_height_program(size, false);
+    normal_from_height_program_ = normal_from_height_program(false);
+}
+
+void
+NormalFromHeightRenderer::setup_onscreen(Canvas& canvas)
+{
+    TextureRenderer::setup_onscreen(canvas);
+
+    normal_from_height_program_->start();
+    (*normal_from_height_program_)["resolution"] =
+        LibMatrix::vec2(canvas.width(), canvas.height());
+    normal_from_height_program_->stop();
+}
+
+void
+NormalFromHeightRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
+{
+    TextureRenderer::setup_offscreen(size, has_depth);
+
+    normal_from_height_program_->start();
+    (*normal_from_height_program_)["resolution"] = size;
+    normal_from_height_program_->stop();
 }
 
 Program *
-NormalFromHeightRenderer::normal_from_height_program(const LibMatrix::vec2 &size,
-                                                     bool create_new)
+NormalFromHeightRenderer::normal_from_height_program(bool create_new)
 {
     static Program *normal_from_height_program(0);
     if (create_new)
@@ -39,15 +60,14 @@
 
     if (!normal_from_height_program) {
         normal_from_height_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-normalmap.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-normalmap.frag");
 
         Scene::load_shaders_from_strings(*normal_from_height_program,
                                          vtx_shader.str(), frg_shader.str());
 
         normal_from_height_program->start();
         (*normal_from_height_program)["heightMap"] = 0;
-        (*normal_from_height_program)["resolution"] = size;
         (*normal_from_height_program)["height"] = 0.05f;
         (*normal_from_height_program)["uvOffset"] = LibMatrix::vec2(0.0f, 0.0f);
         (*normal_from_height_program)["uvScale"] = LibMatrix::vec2(1.0f, 1.0f);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/overlay-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/overlay-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/overlay-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/overlay-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,6 +19,7 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
@@ -33,10 +34,14 @@
 
 
 void
-OverlayRenderer::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+OverlayRenderer::setup_onscreen(Canvas &canvas)
+{
+}
+
+void
+OverlayRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     static_cast<void>(size);
-    static_cast<void>(onscreen);
     static_cast<void>(has_depth);
 }
 
@@ -107,8 +112,8 @@
 void
 OverlayRenderer::create_program()
 {
-    ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-    ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-overlay.frag");
+    ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+    ShaderSource frg_shader(Options::data_path + "/shaders/terrain-overlay.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_shader.str(), frg_shader.str()))
         return;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/renderer-chain.cpp glmark2-2021.02/android/jni/src/scene-terrain/renderer-chain.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/renderer-chain.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/renderer-chain.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -22,10 +22,15 @@
 #include "renderer.h"
 
 void
-RendererChain::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+RendererChain::setup_onscreen(Canvas &canvas)
+{
+    static_cast<void>(canvas);
+}
+
+void
+RendererChain::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     static_cast<void>(size);
-    static_cast<void>(onscreen);
     static_cast<void>(has_depth);
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/renderer.h glmark2-2021.02/android/jni/src/scene-terrain/renderer.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/renderer.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/renderer.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,6 +22,7 @@
 #include <stdint.h>
 #include <vector>
 
+#include "canvas.h"
 #include "mesh.h"
 #include "vec.h"
 #include "program.h"
@@ -36,7 +37,8 @@
     /** 
      * Sets up the renderer's target.
      */
-    virtual void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth) = 0;
+    virtual void setup_onscreen(Canvas& canvas) = 0;
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth) = 0;
     /** 
      * Sets up the renderer's target texture (if any).
      */
@@ -81,7 +83,8 @@
     virtual ~RendererChain() {}
 
     /* IRenderer methods */
-    void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    void setup_onscreen(Canvas& canvas);
+    void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     void setup_texture(GLint min_filter, GLint mag_filter,
                        GLint wrap_s, GLint wrap_t);
     void input_texture(GLuint t);
@@ -108,11 +111,12 @@
 class BaseRenderer : public IRenderer
 {
 public:
-    BaseRenderer(const LibMatrix::vec2 &size);
+    BaseRenderer();
     virtual ~BaseRenderer();
 
     /* IRenderer methods */
-    virtual void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     virtual void setup_texture(GLint min_filter, GLint mag_filter,
                                GLint wrap_s, GLint wrap_t);
     virtual void input_texture(GLuint t) { input_texture_ = t; }
@@ -123,7 +127,7 @@
     virtual void render() = 0;
 
 protected:
-    void recreate(bool onscreen, bool has_depth);
+    void recreate(Canvas* canvas, bool has_depth);
     void create_texture();
     void update_texture_parameters();
     void create_fbo(bool has_depth);
@@ -133,6 +137,7 @@
     GLuint input_texture_;
     GLuint fbo_;
     GLuint depth_renderbuffer_;
+    bool owns_fbo_;
     GLint min_filter_;
     GLint mag_filter_;
     GLint wrap_s_;
@@ -146,7 +151,7 @@
 class TextureRenderer : public BaseRenderer
 {
 public:
-    TextureRenderer(const LibMatrix::vec2 &size, Program &program);
+    TextureRenderer(Program &program);
     virtual ~TextureRenderer() { }
 
     /* IRenderer/BaseRenderer methods */
@@ -170,7 +175,7 @@
 class CopyRenderer : public TextureRenderer
 {
 public:
-    CopyRenderer(const LibMatrix::vec2 &size);
+    CopyRenderer();
 
     virtual ~CopyRenderer() { delete copy_program_; }
 
@@ -186,7 +191,7 @@
 class SimplexNoiseRenderer : public TextureRenderer
 {
 public:
-    SimplexNoiseRenderer(const LibMatrix::vec2 &size);
+    SimplexNoiseRenderer();
     virtual ~SimplexNoiseRenderer() { delete noise_program_; }
 
     LibMatrix::vec2 uv_scale() { return uv_scale_; }
@@ -204,12 +209,14 @@
 class NormalFromHeightRenderer : public TextureRenderer
 {
 public:
-    NormalFromHeightRenderer(const LibMatrix::vec2 &size);
+    NormalFromHeightRenderer();
     virtual ~NormalFromHeightRenderer() { delete normal_from_height_program_; }
 
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
+
 private:
-    static Program *normal_from_height_program(const LibMatrix::vec2 &size, 
-                                               bool create_new);
+    static Program *normal_from_height_program(bool create_new);
 
     Program *normal_from_height_program_;
 };
@@ -220,7 +227,7 @@
 class LuminanceRenderer : public TextureRenderer
 {
 public:
-    LuminanceRenderer(const LibMatrix::vec2 &size);
+    LuminanceRenderer();
     virtual ~LuminanceRenderer() { delete luminance_program_; }
 
 private:
@@ -242,8 +249,8 @@
         BlurDirectionBoth
     };
 
-    BlurRenderer(const LibMatrix::vec2 &size, int radius, float sigma,
-                 BlurDirection dir, const LibMatrix::vec2 &step, float tilt_shift);
+    BlurRenderer(int radius, float sigma, BlurDirection dir,
+                 const LibMatrix::vec2 &step, float tilt_shift);
     virtual ~BlurRenderer() { delete blur_program_; }
 
 private:
@@ -265,7 +272,8 @@
     virtual ~OverlayRenderer() { }
 
     /* IRenderable Methods */
-    void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     void setup_texture(GLint min_filter, GLint mag_filter,
                        GLint wrap_s, GLint wrap_t);
     void input_texture(GLuint t) { input_texture_ = t; }
@@ -293,7 +301,7 @@
 class TerrainRenderer : public BaseRenderer
 {
 public:
-    TerrainRenderer(const LibMatrix::vec2 &size, const LibMatrix::vec2 &repeat_overlay);
+    TerrainRenderer(const LibMatrix::vec2 &repeat_overlay);
     virtual ~TerrainRenderer();
 
     /* IRenderable Methods */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/simplex-noise-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/simplex-noise-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/simplex-noise-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/simplex-noise-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,14 +19,15 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
 LibMatrix::vec2 SimplexNoiseRenderer::uv_scale_(1.5f, 1.5);
 
-SimplexNoiseRenderer::SimplexNoiseRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *noise_program(true))
+SimplexNoiseRenderer::SimplexNoiseRenderer() :
+    TextureRenderer(*noise_program(true))
 {
     noise_program_ = noise_program(false);
 }
@@ -40,8 +41,8 @@
 
     if (!noise_program) {
         noise_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-noise.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-noise.frag");
 
         Scene::load_shaders_from_strings(*noise_program, vtx_shader.str(), frg_shader.str());
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/terrain-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/terrain-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/terrain-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/terrain-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,14 +19,14 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "texture.h"
 #include "shader-source.h"
 
-TerrainRenderer::TerrainRenderer(const LibMatrix::vec2 &size,
-                                 const LibMatrix::vec2 &repeat_overlay) :
-    BaseRenderer(size), height_map_tex_(0), normal_map_tex_(0),
+TerrainRenderer::TerrainRenderer(const LibMatrix::vec2 &repeat_overlay) :
+    BaseRenderer(), height_map_tex_(0), normal_map_tex_(0),
     specular_map_tex_(0), repeat_overlay_(repeat_overlay)
 {
     create_mesh();
@@ -85,8 +85,8 @@
 void
 TerrainRenderer::init_program()
 {
-    ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain.vert");
-    ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain.frag");
+    ShaderSource vtx_shader(Options::data_path + "/shaders/terrain.vert");
+    ShaderSource frg_shader(Options::data_path + "/shaders/terrain.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_shader.str(), frg_shader.str()))
         return;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/texture-renderer.cpp glmark2-2021.02/android/jni/src/scene-terrain/texture-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain/texture-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain/texture-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,8 +21,8 @@
  */
 #include "renderer.h"
 
-TextureRenderer::TextureRenderer(const LibMatrix::vec2 &size, Program &program) :
-    BaseRenderer(size), program_(program)
+TextureRenderer::TextureRenderer(Program &program) :
+    BaseRenderer(), program_(program)
 {
     /* Create the mesh (quad) used for rendering */
     create_mesh();
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain.cpp glmark2-2021.02/android/jni/src/scene-terrain.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-terrain.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-terrain.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -64,63 +64,67 @@
         const vec2 bloom_res(256.0f, 256.0f);
         const vec2 grass_res(512.0f, 512.0f);
 
-        height_map_renderer = new SimplexNoiseRenderer(map_res);
-        height_map_renderer->setup(height_map_renderer->size(), false, false);
+        height_map_renderer = new SimplexNoiseRenderer();
+        height_map_renderer->setup_offscreen(map_res, false);
 
-        normal_map_renderer = new NormalFromHeightRenderer(map_res);
-        normal_map_renderer->setup(normal_map_renderer->size(), false, false);
+        normal_map_renderer = new NormalFromHeightRenderer();
+        normal_map_renderer->setup_offscreen(map_res, false);
 
-        specular_map_renderer = new LuminanceRenderer(grass_res);
+        specular_map_renderer = new LuminanceRenderer();
+        specular_map_renderer->setup_offscreen(grass_res, false);
         specular_map_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                              GL_REPEAT, GL_REPEAT);
-        specular_map_renderer->setup(specular_map_renderer->size(), false, false);
 
-        terrain_renderer = new TerrainRenderer(screen_res, repeat_overlay);
-        terrain_renderer->setup(terrain_renderer->size(),
-                                !use_bloom && !use_tilt_shift,
-                                true);
+        terrain_renderer = new TerrainRenderer(repeat_overlay);
+        if (!use_bloom && !use_tilt_shift)
+            terrain_renderer->setup_onscreen(canvas);
+        else
+            terrain_renderer->setup_offscreen(screen_res, true);
         terrain_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                         GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
 
         /* Bloom */
         if (use_bloom) {
-            bloom_h_renderer = new BlurRenderer(bloom_res, 2, 4.0,
+            bloom_h_renderer = new BlurRenderer(2, 4.0,
                                                 BlurRenderer::BlurDirectionHorizontal,
                                                 vec2(1.0, 1.0) / screen_res,
                                                 0.0);
+            bloom_h_renderer->setup_offscreen(bloom_res, false);
             bloom_h_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                             GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            bloom_h_renderer->setup(bloom_h_renderer->size(), false, false);
 
-            bloom_v_renderer = new BlurRenderer(bloom_res, 2, 4.0,
+            bloom_v_renderer = new BlurRenderer(2, 4.0,
                                                 BlurRenderer::BlurDirectionVertical,
                                                 vec2(1.0, 1.0) / bloom_res,
                                                 0.0);
+            bloom_v_renderer->setup_offscreen(bloom_res, false);
             bloom_v_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                             GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            bloom_v_renderer->setup(bloom_v_renderer->size(), false, false);
             overlay_renderer = new OverlayRenderer(*terrain_renderer, 0.6);
         }
 
         /* Tilt-shift */
         if (use_tilt_shift) {
-            tilt_h_renderer = new BlurRenderer(screen_res, 4, 2.7,
+            tilt_h_renderer = new BlurRenderer(4, 2.7,
                                                BlurRenderer::BlurDirectionHorizontal,
                                                vec2(1.0, 1.0) / screen_res,
                                                0.5);
+            tilt_h_renderer->setup_offscreen(screen_res, false);
             tilt_h_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                            GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            tilt_h_renderer->setup(tilt_h_renderer->size(), false, false);
 
-            tilt_v_renderer = new BlurRenderer(screen_res, 4, 2.7,
+            tilt_v_renderer = new BlurRenderer(4, 2.7,
                                                BlurRenderer::BlurDirectionVertical,
                                                vec2(1.0, 1.0) / screen_res,
                                                0.5);
+            tilt_v_renderer->setup_onscreen(canvas);
         }
 
         /* Copy renderer */
-        if (use_bloom && !use_tilt_shift)
-            copy_renderer = new CopyRenderer(screen_res);
+        if (use_bloom && !use_tilt_shift) {
+            copy_renderer = new CopyRenderer();
+            copy_renderer->setup_onscreen(canvas);
+        }
 
         /* Height normal chain */
         height_normal_chain = new RendererChain();
@@ -310,7 +314,7 @@
     /* Create the specular map */
     priv_->specular_map_renderer->render();
 
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_.fbo());
     glViewport(0, 0, canvas_.width(), canvas_.height());
 
     currentFrame_ = 0;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-texture.cpp glmark2-2021.02/android/jni/src/scene-texture.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/scene-texture.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/scene-texture.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,6 +24,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -109,10 +110,10 @@
     if (!Scene::setup())
         return false;
 
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const std::string vtx_shader_texgen_filename(GLMARK_DATA_PATH"/shaders/light-basic-texgen.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic-tex.frag");
-    static const std::string frg_shader_bilinear_filename(GLMARK_DATA_PATH"/shaders/light-basic-tex-bilinear.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const std::string vtx_shader_texgen_filename(Options::data_path + "/shaders/light-basic-texgen.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/light-basic-tex.frag");
+    static const std::string frg_shader_bilinear_filename(Options::data_path + "/shaders/light-basic-tex-bilinear.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     static const LibMatrix::vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/shared-library.cpp glmark2-2021.02/android/jni/src/shared-library.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/shared-library.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/shared-library.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,90 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#include "shared-library.h"
+
+#include <stdio.h>
+
+#if defined(WIN32)
+#include <windows.h>
+#else
+#include <dlfcn.h>
+#endif
+
+/******************
+ * Public methods *
+ ******************/
+SharedLibrary::SharedLibrary() : handle_(nullptr) {}
+
+SharedLibrary::~SharedLibrary()
+{
+    close();
+}
+
+bool SharedLibrary::open(const char *libName)
+{
+#if defined(WIN32)
+    handle_ = LoadLibraryA(libName);
+#else
+    handle_ = dlopen(libName, RTLD_NOW | RTLD_NODELETE);
+#endif
+    return (handle_ != nullptr);
+}
+
+bool SharedLibrary::open_from_alternatives(std::initializer_list<const char*> alt_names)
+{
+    for (const char *name : alt_names) {
+        if (open(name))
+            return true;
+    }
+
+    return false;
+}
+
+void SharedLibrary::close()
+{
+    if (handle_)
+    {
+#if defined(WIN32)
+        FreeLibrary(reinterpret_cast<HMODULE>(handle_));
+#else
+        dlclose(handle_);
+#endif
+    }
+}
+
+void *SharedLibrary::handle() const
+{
+    return handle_;
+}
+
+void *SharedLibrary::load(const char *symbol) const
+{
+#if defined(WIN32)
+    return reinterpret_cast<void*>(GetProcAddress(reinterpret_cast<HMODULE>(handle_), symbol));
+#else
+    return dlsym(handle_, symbol);
+#endif
+}
+
+/*******************
+ * Private methods *
+ *******************/
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/shared-library.h glmark2-2021.02/android/jni/src/shared-library.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/shared-library.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/shared-library.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,44 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#ifndef GLMARK2_SHARED_LIBRARY_H_
+#define GLMARK2_SHARED_LIBRARY_H_
+
+#include <initializer_list>
+
+class SharedLibrary
+{
+public:
+    SharedLibrary();
+    ~SharedLibrary();
+
+    bool open(const char *name);
+    bool open_from_alternatives(std::initializer_list<const char*> alt_names);
+    void close();
+
+    void *handle() const;
+    void *load(const char *symbol) const;
+
+private:
+    void *handle_;
+};
+
+#endif /* GLMARK2_SHARED_LIBRARY_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/text-renderer.cpp glmark2-2021.02/android/jni/src/text-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/text-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/text-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,6 +21,7 @@
  */
 #include "text-renderer.h"
 #include "gl-headers.h"
+#include "options.h"
 #include "scene.h"
 #include "shader-source.h"
 #include "vec.h"
@@ -49,8 +50,8 @@
     size(0.03);
 
     glGenBuffers(2, vbo_);
-    ShaderSource vtx_source(GLMARK_DATA_PATH"/shaders/text-renderer.vert");
-    ShaderSource frg_source(GLMARK_DATA_PATH"/shaders/text-renderer.frag");
+    ShaderSource vtx_source(Options::data_path + "/shaders/text-renderer.vert");
+    ShaderSource frg_source(Options::data_path + "/shaders/text-renderer.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_source.str(),
                                           frg_source.str()))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/texture.cpp glmark2-2021.02/android/jni/src/texture.cpp
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/texture.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/jni/src/texture.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,9 +23,11 @@
  */
 #include "texture.h"
 #include "log.h"
+#include "options.h"
 #include "util.h"
 #include "image-reader.h"
 
+#include <algorithm>
 #include <cstdarg>
 #include <vector>
 
@@ -149,7 +151,7 @@
         return TexturePrivate::textureMap;
     }
     vector<string> pathVec;
-    string dataDir(GLMARK_DATA_PATH"/textures");
+    string dataDir(Options::data_path + "/textures");
     Util::list_files(dataDir, pathVec);
     // Now that we have a list of all of the image files available to us,
     // let's go through and pull out the names and what format they're in
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/wscript_build glmark2-2021.02/android/jni/src/wscript_build
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/wscript_build	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/android/jni/src/wscript_build	2021-04-25 01:47:22.000000000 +0800
@@ -1,70 +1,251 @@
+# Detect a Windows build.
+is_win = 'WIN32' in ' '.join(bld.env.keys())
+
+# Detect MSVC
+is_msvc = bld.env.CXX_NAME == 'msvc'
+
 all_sources = bld.path.ant_glob('*.cpp scene-ideas/*.cc scene-terrain/*.cpp')
 common_sources = [f for f in all_sources if f.name.find('canvas-') == -1 and
                                             f.name.find('android') == -1 and
                                             f.name.find('native-state-') == -1 and
-                                            f.name.find('gl-state-') == -1]
-common_uselibs = ['libpng']
+                                            f.name.find('gl-state-') == -1 and
+                                            f.name.find('main.cpp') == -1]
+
 common_defines = ['USE_EXCEPTIONS']
 
 libmatrix_sources = [f for f in bld.path.ant_glob('libmatrix/*.cc')
                      if not f.name.endswith('test.cc')]
+
+common_flavor_sources = ['main.cpp', 'canvas-generic.cpp']
+
+libpng_local_sources = [f for f in bld.path.ant_glob('libpng/*.c')]
+zlib_local_sources = [f for f in bld.path.ant_glob('zlib/*.c')]
+
+# Some c sources in libjpeg-turbo are included like headers. So we cannot use globs.
+libjpeg_local_sources = [
+  'jcapimin', 'jcapistd', 'jccoefct', 'jccolor', 'jcdctmgr', 'jchuff', 'jcinit',
+  'jcmainct', 'jcmarker', 'jcmaster', 'jcomapi', 'jcparam', 'jcphuff', 'jcprepct',
+  'jcsample', 'jdapimin', 'jdapistd', 'jdatadst', 'jdatasrc', 'jdcoefct', 'jdcolor',
+  'jddctmgr', 'jdhuff', 'jdinput', 'jdmainct', 'jdmarker', 'jdmaster', 'jdmerge',
+  'jdphuff', 'jdpostct', 'jdsample', 'jerror', 'jfdctflt', 'jfdctfst', 'jfdctint',
+  'jidctflt', 'jidctfst', 'jidctint', 'jidctred', 'jmemmgr', 'jmemnobs', 'jquant1',
+  'jquant2', 'jsimd_none', 'jutils',
+]
+
+libjpeg_turbo_local_sources = ['libjpeg-turbo/%s.c' % f for f in libjpeg_local_sources]
+
+platform_includes = []
+platform_libs = []
+platform_uselibs = []
+
+if is_win:
+  platform_uselibs += ['libpng-local', 'zlib-local', 'libjpeg-turbo-local']
+  platform_libs = ['user32']
+
+  # On windows include some overrides for posix headers.
+  if is_msvc:
+    platform_includes += ['include']
+else:
+  platform_uselibs += ['libpng']
+  platform_libs = ['m', 'jpeg', 'dl']
+
+if 'WAYLAND_SCANNER_wayland_scanner' in bld.env.keys():
+    def wayland_scanner_cmd(arg, src):
+        return '%s %s < %s > ${TGT}' % (bld.env['WAYLAND_SCANNER_wayland_scanner'], arg, src)
+
+    def wayland_proto_src_path(proto, ver):
+        wp_dir = bld.env['WAYLAND_PROTOCOLS_pkgdatadir']
+        if ver == 'stable':
+            return '%s/stable/%s/%s.xml' % (wp_dir, proto, proto)
+        else:
+            return '%s/unstable/%s/%s-unstable-%s.xml' % (wp_dir, proto, proto, ver)
+
+    def wayland_client_protocol(proto, ver, dst_base):
+        src_path = wayland_proto_src_path(proto, ver)
+        out = '%s-client-protocol.h' % dst_base
+        return bld(rule = wayland_scanner_cmd('client-header', src_path),
+                   target = out,
+                   name = out,
+                   ext_out = ['.h'])
+
+    def wayland_protocol_code(proto, ver, dst_base):
+        src_path = wayland_proto_src_path(proto, ver)
+        out = '%s-protocol.c' % dst_base
+        return bld(rule = wayland_scanner_cmd('private-code', src_path),
+                   target = out,
+                   name = out)
+
+    wayland_client_protocol('xdg-shell', 'stable', 'xdg-shell')
+    wayland_protocol_code('xdg-shell', 'stable', 'xdg-shell')
+
 flavor_sources = {
-  'x11-gl' : ['canvas-generic.cpp', 'native-state-x11.cpp', 'gl-state-glx.cpp'],
-  'x11-glesv2' : ['canvas-generic.cpp', 'native-state-x11.cpp', 'gl-state-egl.cpp'],
-  'drm-gl' : ['canvas-generic.cpp', 'native-state-drm.cpp', 'gl-state-egl.cpp'],
-  'drm-glesv2' : ['canvas-generic.cpp', 'native-state-drm.cpp', 'gl-state-egl.cpp'],
-  'mir-gl' : ['canvas-generic.cpp', 'native-state-mir.cpp', 'gl-state-egl.cpp'],
-  'mir-glesv2' : ['canvas-generic.cpp', 'native-state-mir.cpp', 'gl-state-egl.cpp'],
-  'wayland-gl' : ['canvas-generic.cpp', 'native-state-wayland.cpp', 'gl-state-egl.cpp'],
-  'wayland-glesv2' : ['canvas-generic.cpp', 'native-state-wayland.cpp', 'gl-state-egl.cpp']
+  'dispmanx-glesv2' : common_flavor_sources + ['native-state-dispmanx.cpp', 'gl-state-egl.cpp'],
+  'drm-gl' : common_flavor_sources + ['native-state-drm.cpp', 'gl-state-egl.cpp'],
+  'drm-glesv2' : common_flavor_sources + ['native-state-drm.cpp', 'gl-state-egl.cpp'],
+  'mir-gl' : common_flavor_sources + ['native-state-mir.cpp', 'gl-state-egl.cpp'],
+  'mir-glesv2' : common_flavor_sources + ['native-state-mir.cpp', 'gl-state-egl.cpp'],
+  'wayland-gl' : common_flavor_sources + ['native-state-wayland.cpp', 'gl-state-egl.cpp'],
+  'wayland-glesv2' : common_flavor_sources + ['native-state-wayland.cpp', 'gl-state-egl.cpp'],
+  'win32-gl': common_flavor_sources + ['native-state-win32.cpp', 'gl-state-wgl.cpp'],
+  'win32-glesv2': common_flavor_sources + ['native-state-win32.cpp', 'gl-state-egl.cpp'],
+  'x11-gl' : common_flavor_sources + ['native-state-x11.cpp', 'gl-state-glx.cpp'],
+  'x11-glesv2' : common_flavor_sources + ['native-state-x11.cpp', 'gl-state-egl.cpp'],
 }
 flavor_uselibs = {
-  'x11-gl' : ['x11', 'gl', 'matrix-gl'],
-  'x11-glesv2' : ['x11', 'egl', 'glesv2', 'matrix-glesv2'],
-  'drm-gl' : ['drm', 'gbm', 'egl', 'gl', 'matrix-gl'],
-  'drm-glesv2' : ['drm', 'gbm', 'egl', 'glesv2', 'matrix-glesv2'],
-  'mir-gl' : ['mirclient', 'egl', 'gl', 'matrix-gl'],
-  'mir-glesv2' : ['mirclient', 'egl', 'glesv2', 'matrix-glesv2'],
-  'wayland-gl' : ['wayland-client', 'wayland-egl', 'egl', 'gl', 'matrix-gl'],
-  'wayland-glesv2' : ['wayland-client', 'wayland-egl', 'egl', 'glesv2', 'matrix-glesv2']
+  'dispmanx-glesv2' : ['glad-egl-dispmanx', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2',  'dispmanx'],
+  'drm-gl' : ['drm', 'gbm', 'udev', 'glad-egl-drm', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'drm-glesv2' : ['drm', 'gbm', 'udev', 'glad-egl-drm', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'mir-gl' : ['mirclient', 'glad-egl-mir', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'mir-glesv2' : ['mirclient', 'glad-egl-mir', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'wayland-gl' : ['wayland-client', 'wayland-egl', 'wayland-cursor', 'glad-egl-wayland', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'wayland-glesv2' : ['wayland-client', 'wayland-egl', 'wayland-cursor', 'glad-egl-wayland', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'win32-gl': ['glad-gl', 'glad-wgl', 'matrix-gl', 'common-gl'],
+  'win32-glesv2': ['glad-egl-win32', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'x11-gl' : ['x11', 'glad-gl', 'glad-glx', 'matrix-gl', 'common-gl'],
+  'x11-glesv2' : ['x11', 'glad-egl-x11', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
 }
+
 flavor_defines = {
-  'x11-gl' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GL', 'GLMARK2_USE_GLX'],
-  'x11-glesv2' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
-  'drm-gl' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL', '__GBM__'],
-  'drm-glesv2' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL', '__GBM__'],
+  'dispmanx-glesv2' : ['GLMARK2_USE_DISPMANX', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'drm-gl' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
+  'drm-glesv2' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
   'mir-gl' : ['GLMARK2_USE_MIR', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
   'mir-glesv2' : ['GLMARK2_USE_MIR', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
   'wayland-gl' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
-  'wayland-glesv2' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL']
+  'wayland-glesv2' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'win32-gl': ['GLMARK2_USE_WIN32', 'GLMARK2_USE_WGL', 'GLMARK2_USE_GL'],
+  'win32-glesv2': ['GLMARK2_USE_WIN32', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'x11-gl' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GL', 'GLMARK2_USE_GLX'],
+  'x11-glesv2' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+}
+flavor_libs = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : [],
+  'wayland-glesv2' : [],
+  'win32-gl': ['opengl32', 'gdi32'],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+flavor_depends_on = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : ['xdg-shell-client-protocol.h', 'xdg-shell-protocol.c'],
+  'wayland-glesv2' : ['xdg-shell-client-protocol.h', 'xdg-shell-protocol.c'],
+  'win32-gl': [],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+flavor_sources_gen = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : [bld.path.find_or_declare('xdg-shell-protocol.c')],
+  'wayland-glesv2' : [bld.path.find_or_declare('xdg-shell-protocol.c')],
+  'win32-gl': [],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+egl_platform_defines = {
+  'dispmanx' : ['MESA_EGL_NO_X11_HEADERS'],
+  'drm' : ['__GBM__'],
+  'mir' : ['MESA_EGL_NO_X11_HEADERS'],
+  'wayland' : ['WL_EGL_PLATFORM'],
+  'win32' : [],
+  'x11' : [],
 }
 
-includes = ['.', 'scene-ideas', 'scene-terrain']
+includes = ['.', 'scene-ideas', 'scene-terrain'] + platform_includes
 
 all_uselibs = set()
 
 for name in bld.env.keys():
     if name.startswith('FLAVOR_') and bld.env[name]:
         flavor = name.replace('FLAVOR_', '').lower().replace('_', '-')
+        egl_platform = flavor.split('-')[0]
         target = bld.env[name]
-        bld(
+        node = bld(
             features     = ['cxx', 'cprogram'],
-            source       = common_sources + flavor_sources[flavor],
+            source       = flavor_sources[flavor],
             target       = target,
-            use          = common_uselibs + flavor_uselibs[flavor],
-            lib          = ['m', 'jpeg', 'dl'],
-            includes     = includes,
-            defines      = common_defines + flavor_defines[flavor]
+            use          = platform_uselibs + flavor_uselibs[flavor],
+            lib          = platform_libs + flavor_libs[flavor],
+            includes     = ['.'] + platform_includes,
+            defines      = common_defines + flavor_defines[flavor] +
+                           egl_platform_defines[egl_platform],
+            depends_on   = flavor_depends_on[flavor]
             )
-        all_uselibs |= set(flavor_uselibs[flavor])
+        if flavor_sources_gen[flavor]:
+            node.source.extend(flavor_sources_gen[flavor])
+        all_uselibs |= set(flavor_uselibs[flavor] + platform_uselibs)
+
+# Build glad-egl for all used EGL platforms
+for egl_target in (v for v in all_uselibs if v.startswith('glad-egl')):
+    egl_platform = egl_target.split('-')[2]
+    bld(
+        features = ['c'],
+        source   = ['glad/src/egl.c'],
+        target   = egl_target,
+        includes = ['glad/include'],
+        export_includes = 'glad/include',
+        defines = egl_platform_defines[egl_platform]
+        )
+
+if 'glad-glx' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/glx.c'],
+        target   = 'glad-glx',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-gl' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/gl.c'],
+        target   = 'glad-gl',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-wgl' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/wgl.c'],
+        target   = 'glad-wgl',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-glesv2' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/gles2.c'],
+        target   = 'glad-glesv2',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
         
 if 'matrix-gl' in all_uselibs:
     bld(
         features = ['cxx', 'cxxstlib'],
         source   = libmatrix_sources,
         target   = 'matrix-gl',
+        use      = ['glad-gl'],
         lib      = ['m'],
-        includes = ['.'],
+        includes = ['.'] + platform_includes,
         export_includes = 'libmatrix',
         defines  = ['GLMARK2_USE_GL', 'USE_EXCEPTIONS']
         )
@@ -74,8 +255,56 @@
         features = ['cxx', 'cxxstlib'],
         source   = libmatrix_sources,
         target   = 'matrix-glesv2',
+        use      = ['glad-glesv2'],
         lib      = ['m'],
-        includes = ['.'],
+        includes = ['.'] + platform_includes,
         export_includes = 'libmatrix',
         defines  = ['GLMARK2_USE_GLESv2', 'USE_EXCEPTIONS']
         )
+
+if 'libpng-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = libpng_local_sources,
+        target    = 'libpng-local',
+        use       = ['zlib-local'],
+        export_includes = 'libpng'
+        )
+
+if 'zlib-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = zlib_local_sources,
+        target    = 'zlib-local',
+        export_includes = 'zlib'
+        )
+
+if 'libjpeg-turbo-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = libjpeg_turbo_local_sources,
+        target    = 'libjpeg-turbo-local',
+        export_includes = 'libjpeg-turbo'
+        )
+
+if 'common-gl' in all_uselibs:
+    bld(
+        features = ['cxx', 'cxxstlib'],
+        source   = common_sources,
+        target   = 'common-gl',
+        use      = platform_uselibs + ['glad-gl', 'matrix-gl'],
+        lib      = ['m', 'jpeg', 'dl'],
+        includes = includes,
+        defines  = ['GLMARK2_USE_GL', 'USE_EXCEPTIONS']
+        )
+
+if 'common-glesv2' in all_uselibs:
+    bld(
+        features = ['cxx', 'cxxstlib'],
+        source   = common_sources,
+        target   = 'common-glesv2',
+        use      = platform_uselibs + ['glad-glesv2', 'matrix-glesv2'],
+        lib      = ['m', 'jpeg', 'dl'],
+        includes = includes,
+        defines  = ['GLMARK2_USE_GLESv2', 'USE_EXCEPTIONS']
+        )
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/adler32.c glmark2-2021.02/android/jni/src/zlib/adler32.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/adler32.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/adler32.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,186 @@
+/* adler32.c -- compute the Adler-32 checksum of a data stream
+ * Copyright (C) 1995-2011, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#include "zutil.h"
+
+local uLong adler32_combine_ OF((uLong adler1, uLong adler2, z_off64_t len2));
+
+#define BASE 65521U     /* largest prime smaller than 65536 */
+#define NMAX 5552
+/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
+
+#define DO1(buf,i)  {adler += (buf)[i]; sum2 += adler;}
+#define DO2(buf,i)  DO1(buf,i); DO1(buf,i+1);
+#define DO4(buf,i)  DO2(buf,i); DO2(buf,i+2);
+#define DO8(buf,i)  DO4(buf,i); DO4(buf,i+4);
+#define DO16(buf)   DO8(buf,0); DO8(buf,8);
+
+/* use NO_DIVIDE if your processor does not do division in hardware --
+   try it both ways to see which is faster */
+#ifdef NO_DIVIDE
+/* note that this assumes BASE is 65521, where 65536 % 65521 == 15
+   (thank you to John Reiser for pointing this out) */
+#  define CHOP(a) \
+    do { \
+        unsigned long tmp = a >> 16; \
+        a &= 0xffffUL; \
+        a += (tmp << 4) - tmp; \
+    } while (0)
+#  define MOD28(a) \
+    do { \
+        CHOP(a); \
+        if (a >= BASE) a -= BASE; \
+    } while (0)
+#  define MOD(a) \
+    do { \
+        CHOP(a); \
+        MOD28(a); \
+    } while (0)
+#  define MOD63(a) \
+    do { /* this assumes a is not negative */ \
+        z_off64_t tmp = a >> 32; \
+        a &= 0xffffffffL; \
+        a += (tmp << 8) - (tmp << 5) + tmp; \
+        tmp = a >> 16; \
+        a &= 0xffffL; \
+        a += (tmp << 4) - tmp; \
+        tmp = a >> 16; \
+        a &= 0xffffL; \
+        a += (tmp << 4) - tmp; \
+        if (a >= BASE) a -= BASE; \
+    } while (0)
+#else
+#  define MOD(a) a %= BASE
+#  define MOD28(a) a %= BASE
+#  define MOD63(a) a %= BASE
+#endif
+
+/* ========================================================================= */
+uLong ZEXPORT adler32_z(adler, buf, len)
+    uLong adler;
+    const Bytef *buf;
+    z_size_t len;
+{
+    unsigned long sum2;
+    unsigned n;
+
+    /* split Adler-32 into component sums */
+    sum2 = (adler >> 16) & 0xffff;
+    adler &= 0xffff;
+
+    /* in case user likes doing a byte at a time, keep it fast */
+    if (len == 1) {
+        adler += buf[0];
+        if (adler >= BASE)
+            adler -= BASE;
+        sum2 += adler;
+        if (sum2 >= BASE)
+            sum2 -= BASE;
+        return adler | (sum2 << 16);
+    }
+
+    /* initial Adler-32 value (deferred check for len == 1 speed) */
+    if (buf == Z_NULL)
+        return 1L;
+
+    /* in case short lengths are provided, keep it somewhat fast */
+    if (len < 16) {
+        while (len--) {
+            adler += *buf++;
+            sum2 += adler;
+        }
+        if (adler >= BASE)
+            adler -= BASE;
+        MOD28(sum2);            /* only added so many BASE's */
+        return adler | (sum2 << 16);
+    }
+
+    /* do length NMAX blocks -- requires just one modulo operation */
+    while (len >= NMAX) {
+        len -= NMAX;
+        n = NMAX / 16;          /* NMAX is divisible by 16 */
+        do {
+            DO16(buf);          /* 16 sums unrolled */
+            buf += 16;
+        } while (--n);
+        MOD(adler);
+        MOD(sum2);
+    }
+
+    /* do remaining bytes (less than NMAX, still just one modulo) */
+    if (len) {                  /* avoid modulos if none remaining */
+        while (len >= 16) {
+            len -= 16;
+            DO16(buf);
+            buf += 16;
+        }
+        while (len--) {
+            adler += *buf++;
+            sum2 += adler;
+        }
+        MOD(adler);
+        MOD(sum2);
+    }
+
+    /* return recombined sums */
+    return adler | (sum2 << 16);
+}
+
+/* ========================================================================= */
+uLong ZEXPORT adler32(adler, buf, len)
+    uLong adler;
+    const Bytef *buf;
+    uInt len;
+{
+    return adler32_z(adler, buf, len);
+}
+
+/* ========================================================================= */
+local uLong adler32_combine_(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off64_t len2;
+{
+    unsigned long sum1;
+    unsigned long sum2;
+    unsigned rem;
+
+    /* for negative len, return invalid adler32 as a clue for debugging */
+    if (len2 < 0)
+        return 0xffffffffUL;
+
+    /* the derivation of this formula is left as an exercise for the reader */
+    MOD63(len2);                /* assumes len2 >= 0 */
+    rem = (unsigned)len2;
+    sum1 = adler1 & 0xffff;
+    sum2 = rem * sum1;
+    MOD(sum2);
+    sum1 += (adler2 & 0xffff) + BASE - 1;
+    sum2 += ((adler1 >> 16) & 0xffff) + ((adler2 >> 16) & 0xffff) + BASE - rem;
+    if (sum1 >= BASE) sum1 -= BASE;
+    if (sum1 >= BASE) sum1 -= BASE;
+    if (sum2 >= ((unsigned long)BASE << 1)) sum2 -= ((unsigned long)BASE << 1);
+    if (sum2 >= BASE) sum2 -= BASE;
+    return sum1 | (sum2 << 16);
+}
+
+/* ========================================================================= */
+uLong ZEXPORT adler32_combine(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off_t len2;
+{
+    return adler32_combine_(adler1, adler2, len2);
+}
+
+uLong ZEXPORT adler32_combine64(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off64_t len2;
+{
+    return adler32_combine_(adler1, adler2, len2);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/compress.c glmark2-2021.02/android/jni/src/zlib/compress.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/compress.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/compress.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,86 @@
+/* compress.c -- compress a memory buffer
+ * Copyright (C) 1995-2005, 2014, 2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#define ZLIB_INTERNAL
+#include "zlib.h"
+
+/* ===========================================================================
+     Compresses the source buffer into the destination buffer. The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer. Upon entry, destLen is the total size of the
+   destination buffer, which must be at least 0.1% larger than sourceLen plus
+   12 bytes. Upon exit, destLen is the actual size of the compressed buffer.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+int ZEXPORT compress2 (dest, destLen, source, sourceLen, level)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+    int level;
+{
+    z_stream stream;
+    int err;
+    const uInt max = (uInt)-1;
+    uLong left;
+
+    left = *destLen;
+    *destLen = 0;
+
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+    stream.opaque = (voidpf)0;
+
+    err = deflateInit(&stream, level);
+    if (err != Z_OK) return err;
+
+    stream.next_out = dest;
+    stream.avail_out = 0;
+    stream.next_in = (z_const Bytef *)source;
+    stream.avail_in = 0;
+
+    do {
+        if (stream.avail_out == 0) {
+            stream.avail_out = left > (uLong)max ? max : (uInt)left;
+            left -= stream.avail_out;
+        }
+        if (stream.avail_in == 0) {
+            stream.avail_in = sourceLen > (uLong)max ? max : (uInt)sourceLen;
+            sourceLen -= stream.avail_in;
+        }
+        err = deflate(&stream, sourceLen ? Z_NO_FLUSH : Z_FINISH);
+    } while (err == Z_OK);
+
+    *destLen = stream.total_out;
+    deflateEnd(&stream);
+    return err == Z_STREAM_END ? Z_OK : err;
+}
+
+/* ===========================================================================
+ */
+int ZEXPORT compress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    return compress2(dest, destLen, source, sourceLen, Z_DEFAULT_COMPRESSION);
+}
+
+/* ===========================================================================
+     If the default memLevel or windowBits for deflateInit() is changed, then
+   this function needs to be updated.
+ */
+uLong ZEXPORT compressBound (sourceLen)
+    uLong sourceLen;
+{
+    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
+           (sourceLen >> 25) + 13;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/crc32.c glmark2-2021.02/android/jni/src/zlib/crc32.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/crc32.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/crc32.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,442 @@
+/* crc32.c -- compute the CRC-32 of a data stream
+ * Copyright (C) 1995-2006, 2010, 2011, 2012, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ *
+ * Thanks to Rodney Brown <rbrown64@csc.com.au> for his contribution of faster
+ * CRC methods: exclusive-oring 32 bits of data at a time, and pre-computing
+ * tables for updating the shift register in one step with three exclusive-ors
+ * instead of four steps with four exclusive-ors.  This results in about a
+ * factor of two increase in speed on a Power PC G4 (PPC7455) using gcc -O3.
+ */
+
+/* @(#) $Id$ */
+
+/*
+  Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore
+  protection on the static variables used to control the first-use generation
+  of the crc tables.  Therefore, if you #define DYNAMIC_CRC_TABLE, you should
+  first call get_crc_table() to initialize the tables before allowing more than
+  one thread to use crc32().
+
+  DYNAMIC_CRC_TABLE and MAKECRCH can be #defined to write out crc32.h.
+ */
+
+#ifdef MAKECRCH
+#  include <stdio.h>
+#  ifndef DYNAMIC_CRC_TABLE
+#    define DYNAMIC_CRC_TABLE
+#  endif /* !DYNAMIC_CRC_TABLE */
+#endif /* MAKECRCH */
+
+#include "zutil.h"      /* for STDC and FAR definitions */
+
+/* Definitions for doing the crc four data bytes at a time. */
+#if !defined(NOBYFOUR) && defined(Z_U4)
+#  define BYFOUR
+#endif
+#ifdef BYFOUR
+   local unsigned long crc32_little OF((unsigned long,
+                        const unsigned char FAR *, z_size_t));
+   local unsigned long crc32_big OF((unsigned long,
+                        const unsigned char FAR *, z_size_t));
+#  define TBLS 8
+#else
+#  define TBLS 1
+#endif /* BYFOUR */
+
+/* Local functions for crc concatenation */
+local unsigned long gf2_matrix_times OF((unsigned long *mat,
+                                         unsigned long vec));
+local void gf2_matrix_square OF((unsigned long *square, unsigned long *mat));
+local uLong crc32_combine_ OF((uLong crc1, uLong crc2, z_off64_t len2));
+
+
+#ifdef DYNAMIC_CRC_TABLE
+
+local volatile int crc_table_empty = 1;
+local z_crc_t FAR crc_table[TBLS][256];
+local void make_crc_table OF((void));
+#ifdef MAKECRCH
+   local void write_table OF((FILE *, const z_crc_t FAR *));
+#endif /* MAKECRCH */
+/*
+  Generate tables for a byte-wise 32-bit CRC calculation on the polynomial:
+  x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1.
+
+  Polynomials over GF(2) are represented in binary, one bit per coefficient,
+  with the lowest powers in the most significant bit.  Then adding polynomials
+  is just exclusive-or, and multiplying a polynomial by x is a right shift by
+  one.  If we call the above polynomial p, and represent a byte as the
+  polynomial q, also with the lowest power in the most significant bit (so the
+  byte 0xb1 is the polynomial x^7+x^3+x+1), then the CRC is (q*x^32) mod p,
+  where a mod b means the remainder after dividing a by b.
+
+  This calculation is done using the shift-register method of multiplying and
+  taking the remainder.  The register is initialized to zero, and for each
+  incoming bit, x^32 is added mod p to the register if the bit is a one (where
+  x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by
+  x (which is shifting right by one and adding x^32 mod p if the bit shifted
+  out is a one).  We start with the highest power (least significant bit) of
+  q and repeat for all eight bits of q.
+
+  The first table is simply the CRC of all possible eight bit values.  This is
+  all the information needed to generate CRCs on data a byte at a time for all
+  combinations of CRC register values and incoming bytes.  The remaining tables
+  allow for word-at-a-time CRC calculation for both big-endian and little-
+  endian machines, where a word is four bytes.
+*/
+local void make_crc_table()
+{
+    z_crc_t c;
+    int n, k;
+    z_crc_t poly;                       /* polynomial exclusive-or pattern */
+    /* terms of polynomial defining this crc (except x^32): */
+    static volatile int first = 1;      /* flag to limit concurrent making */
+    static const unsigned char p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
+
+    /* See if another task is already doing this (not thread-safe, but better
+       than nothing -- significantly reduces duration of vulnerability in
+       case the advice about DYNAMIC_CRC_TABLE is ignored) */
+    if (first) {
+        first = 0;
+
+        /* make exclusive-or pattern from polynomial (0xedb88320UL) */
+        poly = 0;
+        for (n = 0; n < (int)(sizeof(p)/sizeof(unsigned char)); n++)
+            poly |= (z_crc_t)1 << (31 - p[n]);
+
+        /* generate a crc for every 8-bit value */
+        for (n = 0; n < 256; n++) {
+            c = (z_crc_t)n;
+            for (k = 0; k < 8; k++)
+                c = c & 1 ? poly ^ (c >> 1) : c >> 1;
+            crc_table[0][n] = c;
+        }
+
+#ifdef BYFOUR
+        /* generate crc for each value followed by one, two, and three zeros,
+           and then the byte reversal of those as well as the first table */
+        for (n = 0; n < 256; n++) {
+            c = crc_table[0][n];
+            crc_table[4][n] = ZSWAP32(c);
+            for (k = 1; k < 4; k++) {
+                c = crc_table[0][c & 0xff] ^ (c >> 8);
+                crc_table[k][n] = c;
+                crc_table[k + 4][n] = ZSWAP32(c);
+            }
+        }
+#endif /* BYFOUR */
+
+        crc_table_empty = 0;
+    }
+    else {      /* not first */
+        /* wait for the other guy to finish (not efficient, but rare) */
+        while (crc_table_empty)
+            ;
+    }
+
+#ifdef MAKECRCH
+    /* write out CRC tables to crc32.h */
+    {
+        FILE *out;
+
+        out = fopen("crc32.h", "w");
+        if (out == NULL) return;
+        fprintf(out, "/* crc32.h -- tables for rapid CRC calculation\n");
+        fprintf(out, " * Generated automatically by crc32.c\n */\n\n");
+        fprintf(out, "local const z_crc_t FAR ");
+        fprintf(out, "crc_table[TBLS][256] =\n{\n  {\n");
+        write_table(out, crc_table[0]);
+#  ifdef BYFOUR
+        fprintf(out, "#ifdef BYFOUR\n");
+        for (k = 1; k < 8; k++) {
+            fprintf(out, "  },\n  {\n");
+            write_table(out, crc_table[k]);
+        }
+        fprintf(out, "#endif\n");
+#  endif /* BYFOUR */
+        fprintf(out, "  }\n};\n");
+        fclose(out);
+    }
+#endif /* MAKECRCH */
+}
+
+#ifdef MAKECRCH
+local void write_table(out, table)
+    FILE *out;
+    const z_crc_t FAR *table;
+{
+    int n;
+
+    for (n = 0; n < 256; n++)
+        fprintf(out, "%s0x%08lxUL%s", n % 5 ? "" : "    ",
+                (unsigned long)(table[n]),
+                n == 255 ? "\n" : (n % 5 == 4 ? ",\n" : ", "));
+}
+#endif /* MAKECRCH */
+
+#else /* !DYNAMIC_CRC_TABLE */
+/* ========================================================================
+ * Tables of CRC-32s of all single-byte values, made by make_crc_table().
+ */
+#include "crc32.h"
+#endif /* DYNAMIC_CRC_TABLE */
+
+/* =========================================================================
+ * This function can be used by asm versions of crc32()
+ */
+const z_crc_t FAR * ZEXPORT get_crc_table()
+{
+#ifdef DYNAMIC_CRC_TABLE
+    if (crc_table_empty)
+        make_crc_table();
+#endif /* DYNAMIC_CRC_TABLE */
+    return (const z_crc_t FAR *)crc_table;
+}
+
+/* ========================================================================= */
+#define DO1 crc = crc_table[0][((int)crc ^ (*buf++)) & 0xff] ^ (crc >> 8)
+#define DO8 DO1; DO1; DO1; DO1; DO1; DO1; DO1; DO1
+
+/* ========================================================================= */
+unsigned long ZEXPORT crc32_z(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    if (buf == Z_NULL) return 0UL;
+
+#ifdef DYNAMIC_CRC_TABLE
+    if (crc_table_empty)
+        make_crc_table();
+#endif /* DYNAMIC_CRC_TABLE */
+
+#ifdef BYFOUR
+    if (sizeof(void *) == sizeof(ptrdiff_t)) {
+        z_crc_t endian;
+
+        endian = 1;
+        if (*((unsigned char *)(&endian)))
+            return crc32_little(crc, buf, len);
+        else
+            return crc32_big(crc, buf, len);
+    }
+#endif /* BYFOUR */
+    crc = crc ^ 0xffffffffUL;
+    while (len >= 8) {
+        DO8;
+        len -= 8;
+    }
+    if (len) do {
+        DO1;
+    } while (--len);
+    return crc ^ 0xffffffffUL;
+}
+
+/* ========================================================================= */
+unsigned long ZEXPORT crc32(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    uInt len;
+{
+    return crc32_z(crc, buf, len);
+}
+
+#ifdef BYFOUR
+
+/*
+   This BYFOUR code accesses the passed unsigned char * buffer with a 32-bit
+   integer pointer type. This violates the strict aliasing rule, where a
+   compiler can assume, for optimization purposes, that two pointers to
+   fundamentally different types won't ever point to the same memory. This can
+   manifest as a problem only if one of the pointers is written to. This code
+   only reads from those pointers. So long as this code remains isolated in
+   this compilation unit, there won't be a problem. For this reason, this code
+   should not be copied and pasted into a compilation unit in which other code
+   writes to the buffer that is passed to these routines.
+ */
+
+/* ========================================================================= */
+#define DOLIT4 c ^= *buf4++; \
+        c = crc_table[3][c & 0xff] ^ crc_table[2][(c >> 8) & 0xff] ^ \
+            crc_table[1][(c >> 16) & 0xff] ^ crc_table[0][c >> 24]
+#define DOLIT32 DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4
+
+/* ========================================================================= */
+local unsigned long crc32_little(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    register z_crc_t c;
+    register const z_crc_t FAR *buf4;
+
+    c = (z_crc_t)crc;
+    c = ~c;
+    while (len && ((ptrdiff_t)buf & 3)) {
+        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
+        len--;
+    }
+
+    buf4 = (const z_crc_t FAR *)(const void FAR *)buf;
+    while (len >= 32) {
+        DOLIT32;
+        len -= 32;
+    }
+    while (len >= 4) {
+        DOLIT4;
+        len -= 4;
+    }
+    buf = (const unsigned char FAR *)buf4;
+
+    if (len) do {
+        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
+    } while (--len);
+    c = ~c;
+    return (unsigned long)c;
+}
+
+/* ========================================================================= */
+#define DOBIG4 c ^= *buf4++; \
+        c = crc_table[4][c & 0xff] ^ crc_table[5][(c >> 8) & 0xff] ^ \
+            crc_table[6][(c >> 16) & 0xff] ^ crc_table[7][c >> 24]
+#define DOBIG32 DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4
+
+/* ========================================================================= */
+local unsigned long crc32_big(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    register z_crc_t c;
+    register const z_crc_t FAR *buf4;
+
+    c = ZSWAP32((z_crc_t)crc);
+    c = ~c;
+    while (len && ((ptrdiff_t)buf & 3)) {
+        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
+        len--;
+    }
+
+    buf4 = (const z_crc_t FAR *)(const void FAR *)buf;
+    while (len >= 32) {
+        DOBIG32;
+        len -= 32;
+    }
+    while (len >= 4) {
+        DOBIG4;
+        len -= 4;
+    }
+    buf = (const unsigned char FAR *)buf4;
+
+    if (len) do {
+        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
+    } while (--len);
+    c = ~c;
+    return (unsigned long)(ZSWAP32(c));
+}
+
+#endif /* BYFOUR */
+
+#define GF2_DIM 32      /* dimension of GF(2) vectors (length of CRC) */
+
+/* ========================================================================= */
+local unsigned long gf2_matrix_times(mat, vec)
+    unsigned long *mat;
+    unsigned long vec;
+{
+    unsigned long sum;
+
+    sum = 0;
+    while (vec) {
+        if (vec & 1)
+            sum ^= *mat;
+        vec >>= 1;
+        mat++;
+    }
+    return sum;
+}
+
+/* ========================================================================= */
+local void gf2_matrix_square(square, mat)
+    unsigned long *square;
+    unsigned long *mat;
+{
+    int n;
+
+    for (n = 0; n < GF2_DIM; n++)
+        square[n] = gf2_matrix_times(mat, mat[n]);
+}
+
+/* ========================================================================= */
+local uLong crc32_combine_(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off64_t len2;
+{
+    int n;
+    unsigned long row;
+    unsigned long even[GF2_DIM];    /* even-power-of-two zeros operator */
+    unsigned long odd[GF2_DIM];     /* odd-power-of-two zeros operator */
+
+    /* degenerate case (also disallow negative lengths) */
+    if (len2 <= 0)
+        return crc1;
+
+    /* put operator for one zero bit in odd */
+    odd[0] = 0xedb88320UL;          /* CRC-32 polynomial */
+    row = 1;
+    for (n = 1; n < GF2_DIM; n++) {
+        odd[n] = row;
+        row <<= 1;
+    }
+
+    /* put operator for two zero bits in even */
+    gf2_matrix_square(even, odd);
+
+    /* put operator for four zero bits in odd */
+    gf2_matrix_square(odd, even);
+
+    /* apply len2 zeros to crc1 (first square will put the operator for one
+       zero byte, eight zero bits, in even) */
+    do {
+        /* apply zeros operator for this bit of len2 */
+        gf2_matrix_square(even, odd);
+        if (len2 & 1)
+            crc1 = gf2_matrix_times(even, crc1);
+        len2 >>= 1;
+
+        /* if no more bits set, then done */
+        if (len2 == 0)
+            break;
+
+        /* another iteration of the loop with odd and even swapped */
+        gf2_matrix_square(odd, even);
+        if (len2 & 1)
+            crc1 = gf2_matrix_times(odd, crc1);
+        len2 >>= 1;
+
+        /* if no more bits set, then done */
+    } while (len2 != 0);
+
+    /* return combined crc */
+    crc1 ^= crc2;
+    return crc1;
+}
+
+/* ========================================================================= */
+uLong ZEXPORT crc32_combine(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off_t len2;
+{
+    return crc32_combine_(crc1, crc2, len2);
+}
+
+uLong ZEXPORT crc32_combine64(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off64_t len2;
+{
+    return crc32_combine_(crc1, crc2, len2);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/crc32.h glmark2-2021.02/android/jni/src/zlib/crc32.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/crc32.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/crc32.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,441 @@
+/* crc32.h -- tables for rapid CRC calculation
+ * Generated automatically by crc32.c
+ */
+
+local const z_crc_t FAR crc_table[TBLS][256] =
+{
+  {
+    0x00000000UL, 0x77073096UL, 0xee0e612cUL, 0x990951baUL, 0x076dc419UL,
+    0x706af48fUL, 0xe963a535UL, 0x9e6495a3UL, 0x0edb8832UL, 0x79dcb8a4UL,
+    0xe0d5e91eUL, 0x97d2d988UL, 0x09b64c2bUL, 0x7eb17cbdUL, 0xe7b82d07UL,
+    0x90bf1d91UL, 0x1db71064UL, 0x6ab020f2UL, 0xf3b97148UL, 0x84be41deUL,
+    0x1adad47dUL, 0x6ddde4ebUL, 0xf4d4b551UL, 0x83d385c7UL, 0x136c9856UL,
+    0x646ba8c0UL, 0xfd62f97aUL, 0x8a65c9ecUL, 0x14015c4fUL, 0x63066cd9UL,
+    0xfa0f3d63UL, 0x8d080df5UL, 0x3b6e20c8UL, 0x4c69105eUL, 0xd56041e4UL,
+    0xa2677172UL, 0x3c03e4d1UL, 0x4b04d447UL, 0xd20d85fdUL, 0xa50ab56bUL,
+    0x35b5a8faUL, 0x42b2986cUL, 0xdbbbc9d6UL, 0xacbcf940UL, 0x32d86ce3UL,
+    0x45df5c75UL, 0xdcd60dcfUL, 0xabd13d59UL, 0x26d930acUL, 0x51de003aUL,
+    0xc8d75180UL, 0xbfd06116UL, 0x21b4f4b5UL, 0x56b3c423UL, 0xcfba9599UL,
+    0xb8bda50fUL, 0x2802b89eUL, 0x5f058808UL, 0xc60cd9b2UL, 0xb10be924UL,
+    0x2f6f7c87UL, 0x58684c11UL, 0xc1611dabUL, 0xb6662d3dUL, 0x76dc4190UL,
+    0x01db7106UL, 0x98d220bcUL, 0xefd5102aUL, 0x71b18589UL, 0x06b6b51fUL,
+    0x9fbfe4a5UL, 0xe8b8d433UL, 0x7807c9a2UL, 0x0f00f934UL, 0x9609a88eUL,
+    0xe10e9818UL, 0x7f6a0dbbUL, 0x086d3d2dUL, 0x91646c97UL, 0xe6635c01UL,
+    0x6b6b51f4UL, 0x1c6c6162UL, 0x856530d8UL, 0xf262004eUL, 0x6c0695edUL,
+    0x1b01a57bUL, 0x8208f4c1UL, 0xf50fc457UL, 0x65b0d9c6UL, 0x12b7e950UL,
+    0x8bbeb8eaUL, 0xfcb9887cUL, 0x62dd1ddfUL, 0x15da2d49UL, 0x8cd37cf3UL,
+    0xfbd44c65UL, 0x4db26158UL, 0x3ab551ceUL, 0xa3bc0074UL, 0xd4bb30e2UL,
+    0x4adfa541UL, 0x3dd895d7UL, 0xa4d1c46dUL, 0xd3d6f4fbUL, 0x4369e96aUL,
+    0x346ed9fcUL, 0xad678846UL, 0xda60b8d0UL, 0x44042d73UL, 0x33031de5UL,
+    0xaa0a4c5fUL, 0xdd0d7cc9UL, 0x5005713cUL, 0x270241aaUL, 0xbe0b1010UL,
+    0xc90c2086UL, 0x5768b525UL, 0x206f85b3UL, 0xb966d409UL, 0xce61e49fUL,
+    0x5edef90eUL, 0x29d9c998UL, 0xb0d09822UL, 0xc7d7a8b4UL, 0x59b33d17UL,
+    0x2eb40d81UL, 0xb7bd5c3bUL, 0xc0ba6cadUL, 0xedb88320UL, 0x9abfb3b6UL,
+    0x03b6e20cUL, 0x74b1d29aUL, 0xead54739UL, 0x9dd277afUL, 0x04db2615UL,
+    0x73dc1683UL, 0xe3630b12UL, 0x94643b84UL, 0x0d6d6a3eUL, 0x7a6a5aa8UL,
+    0xe40ecf0bUL, 0x9309ff9dUL, 0x0a00ae27UL, 0x7d079eb1UL, 0xf00f9344UL,
+    0x8708a3d2UL, 0x1e01f268UL, 0x6906c2feUL, 0xf762575dUL, 0x806567cbUL,
+    0x196c3671UL, 0x6e6b06e7UL, 0xfed41b76UL, 0x89d32be0UL, 0x10da7a5aUL,
+    0x67dd4accUL, 0xf9b9df6fUL, 0x8ebeeff9UL, 0x17b7be43UL, 0x60b08ed5UL,
+    0xd6d6a3e8UL, 0xa1d1937eUL, 0x38d8c2c4UL, 0x4fdff252UL, 0xd1bb67f1UL,
+    0xa6bc5767UL, 0x3fb506ddUL, 0x48b2364bUL, 0xd80d2bdaUL, 0xaf0a1b4cUL,
+    0x36034af6UL, 0x41047a60UL, 0xdf60efc3UL, 0xa867df55UL, 0x316e8eefUL,
+    0x4669be79UL, 0xcb61b38cUL, 0xbc66831aUL, 0x256fd2a0UL, 0x5268e236UL,
+    0xcc0c7795UL, 0xbb0b4703UL, 0x220216b9UL, 0x5505262fUL, 0xc5ba3bbeUL,
+    0xb2bd0b28UL, 0x2bb45a92UL, 0x5cb36a04UL, 0xc2d7ffa7UL, 0xb5d0cf31UL,
+    0x2cd99e8bUL, 0x5bdeae1dUL, 0x9b64c2b0UL, 0xec63f226UL, 0x756aa39cUL,
+    0x026d930aUL, 0x9c0906a9UL, 0xeb0e363fUL, 0x72076785UL, 0x05005713UL,
+    0x95bf4a82UL, 0xe2b87a14UL, 0x7bb12baeUL, 0x0cb61b38UL, 0x92d28e9bUL,
+    0xe5d5be0dUL, 0x7cdcefb7UL, 0x0bdbdf21UL, 0x86d3d2d4UL, 0xf1d4e242UL,
+    0x68ddb3f8UL, 0x1fda836eUL, 0x81be16cdUL, 0xf6b9265bUL, 0x6fb077e1UL,
+    0x18b74777UL, 0x88085ae6UL, 0xff0f6a70UL, 0x66063bcaUL, 0x11010b5cUL,
+    0x8f659effUL, 0xf862ae69UL, 0x616bffd3UL, 0x166ccf45UL, 0xa00ae278UL,
+    0xd70dd2eeUL, 0x4e048354UL, 0x3903b3c2UL, 0xa7672661UL, 0xd06016f7UL,
+    0x4969474dUL, 0x3e6e77dbUL, 0xaed16a4aUL, 0xd9d65adcUL, 0x40df0b66UL,
+    0x37d83bf0UL, 0xa9bcae53UL, 0xdebb9ec5UL, 0x47b2cf7fUL, 0x30b5ffe9UL,
+    0xbdbdf21cUL, 0xcabac28aUL, 0x53b39330UL, 0x24b4a3a6UL, 0xbad03605UL,
+    0xcdd70693UL, 0x54de5729UL, 0x23d967bfUL, 0xb3667a2eUL, 0xc4614ab8UL,
+    0x5d681b02UL, 0x2a6f2b94UL, 0xb40bbe37UL, 0xc30c8ea1UL, 0x5a05df1bUL,
+    0x2d02ef8dUL
+#ifdef BYFOUR
+  },
+  {
+    0x00000000UL, 0x191b3141UL, 0x32366282UL, 0x2b2d53c3UL, 0x646cc504UL,
+    0x7d77f445UL, 0x565aa786UL, 0x4f4196c7UL, 0xc8d98a08UL, 0xd1c2bb49UL,
+    0xfaefe88aUL, 0xe3f4d9cbUL, 0xacb54f0cUL, 0xb5ae7e4dUL, 0x9e832d8eUL,
+    0x87981ccfUL, 0x4ac21251UL, 0x53d92310UL, 0x78f470d3UL, 0x61ef4192UL,
+    0x2eaed755UL, 0x37b5e614UL, 0x1c98b5d7UL, 0x05838496UL, 0x821b9859UL,
+    0x9b00a918UL, 0xb02dfadbUL, 0xa936cb9aUL, 0xe6775d5dUL, 0xff6c6c1cUL,
+    0xd4413fdfUL, 0xcd5a0e9eUL, 0x958424a2UL, 0x8c9f15e3UL, 0xa7b24620UL,
+    0xbea97761UL, 0xf1e8e1a6UL, 0xe8f3d0e7UL, 0xc3de8324UL, 0xdac5b265UL,
+    0x5d5daeaaUL, 0x44469febUL, 0x6f6bcc28UL, 0x7670fd69UL, 0x39316baeUL,
+    0x202a5aefUL, 0x0b07092cUL, 0x121c386dUL, 0xdf4636f3UL, 0xc65d07b2UL,
+    0xed705471UL, 0xf46b6530UL, 0xbb2af3f7UL, 0xa231c2b6UL, 0x891c9175UL,
+    0x9007a034UL, 0x179fbcfbUL, 0x0e848dbaUL, 0x25a9de79UL, 0x3cb2ef38UL,
+    0x73f379ffUL, 0x6ae848beUL, 0x41c51b7dUL, 0x58de2a3cUL, 0xf0794f05UL,
+    0xe9627e44UL, 0xc24f2d87UL, 0xdb541cc6UL, 0x94158a01UL, 0x8d0ebb40UL,
+    0xa623e883UL, 0xbf38d9c2UL, 0x38a0c50dUL, 0x21bbf44cUL, 0x0a96a78fUL,
+    0x138d96ceUL, 0x5ccc0009UL, 0x45d73148UL, 0x6efa628bUL, 0x77e153caUL,
+    0xbabb5d54UL, 0xa3a06c15UL, 0x888d3fd6UL, 0x91960e97UL, 0xded79850UL,
+    0xc7cca911UL, 0xece1fad2UL, 0xf5facb93UL, 0x7262d75cUL, 0x6b79e61dUL,
+    0x4054b5deUL, 0x594f849fUL, 0x160e1258UL, 0x0f152319UL, 0x243870daUL,
+    0x3d23419bUL, 0x65fd6ba7UL, 0x7ce65ae6UL, 0x57cb0925UL, 0x4ed03864UL,
+    0x0191aea3UL, 0x188a9fe2UL, 0x33a7cc21UL, 0x2abcfd60UL, 0xad24e1afUL,
+    0xb43fd0eeUL, 0x9f12832dUL, 0x8609b26cUL, 0xc94824abUL, 0xd05315eaUL,
+    0xfb7e4629UL, 0xe2657768UL, 0x2f3f79f6UL, 0x362448b7UL, 0x1d091b74UL,
+    0x04122a35UL, 0x4b53bcf2UL, 0x52488db3UL, 0x7965de70UL, 0x607eef31UL,
+    0xe7e6f3feUL, 0xfefdc2bfUL, 0xd5d0917cUL, 0xcccba03dUL, 0x838a36faUL,
+    0x9a9107bbUL, 0xb1bc5478UL, 0xa8a76539UL, 0x3b83984bUL, 0x2298a90aUL,
+    0x09b5fac9UL, 0x10aecb88UL, 0x5fef5d4fUL, 0x46f46c0eUL, 0x6dd93fcdUL,
+    0x74c20e8cUL, 0xf35a1243UL, 0xea412302UL, 0xc16c70c1UL, 0xd8774180UL,
+    0x9736d747UL, 0x8e2de606UL, 0xa500b5c5UL, 0xbc1b8484UL, 0x71418a1aUL,
+    0x685abb5bUL, 0x4377e898UL, 0x5a6cd9d9UL, 0x152d4f1eUL, 0x0c367e5fUL,
+    0x271b2d9cUL, 0x3e001cddUL, 0xb9980012UL, 0xa0833153UL, 0x8bae6290UL,
+    0x92b553d1UL, 0xddf4c516UL, 0xc4eff457UL, 0xefc2a794UL, 0xf6d996d5UL,
+    0xae07bce9UL, 0xb71c8da8UL, 0x9c31de6bUL, 0x852aef2aUL, 0xca6b79edUL,
+    0xd37048acUL, 0xf85d1b6fUL, 0xe1462a2eUL, 0x66de36e1UL, 0x7fc507a0UL,
+    0x54e85463UL, 0x4df36522UL, 0x02b2f3e5UL, 0x1ba9c2a4UL, 0x30849167UL,
+    0x299fa026UL, 0xe4c5aeb8UL, 0xfdde9ff9UL, 0xd6f3cc3aUL, 0xcfe8fd7bUL,
+    0x80a96bbcUL, 0x99b25afdUL, 0xb29f093eUL, 0xab84387fUL, 0x2c1c24b0UL,
+    0x350715f1UL, 0x1e2a4632UL, 0x07317773UL, 0x4870e1b4UL, 0x516bd0f5UL,
+    0x7a468336UL, 0x635db277UL, 0xcbfad74eUL, 0xd2e1e60fUL, 0xf9ccb5ccUL,
+    0xe0d7848dUL, 0xaf96124aUL, 0xb68d230bUL, 0x9da070c8UL, 0x84bb4189UL,
+    0x03235d46UL, 0x1a386c07UL, 0x31153fc4UL, 0x280e0e85UL, 0x674f9842UL,
+    0x7e54a903UL, 0x5579fac0UL, 0x4c62cb81UL, 0x8138c51fUL, 0x9823f45eUL,
+    0xb30ea79dUL, 0xaa1596dcUL, 0xe554001bUL, 0xfc4f315aUL, 0xd7626299UL,
+    0xce7953d8UL, 0x49e14f17UL, 0x50fa7e56UL, 0x7bd72d95UL, 0x62cc1cd4UL,
+    0x2d8d8a13UL, 0x3496bb52UL, 0x1fbbe891UL, 0x06a0d9d0UL, 0x5e7ef3ecUL,
+    0x4765c2adUL, 0x6c48916eUL, 0x7553a02fUL, 0x3a1236e8UL, 0x230907a9UL,
+    0x0824546aUL, 0x113f652bUL, 0x96a779e4UL, 0x8fbc48a5UL, 0xa4911b66UL,
+    0xbd8a2a27UL, 0xf2cbbce0UL, 0xebd08da1UL, 0xc0fdde62UL, 0xd9e6ef23UL,
+    0x14bce1bdUL, 0x0da7d0fcUL, 0x268a833fUL, 0x3f91b27eUL, 0x70d024b9UL,
+    0x69cb15f8UL, 0x42e6463bUL, 0x5bfd777aUL, 0xdc656bb5UL, 0xc57e5af4UL,
+    0xee530937UL, 0xf7483876UL, 0xb809aeb1UL, 0xa1129ff0UL, 0x8a3fcc33UL,
+    0x9324fd72UL
+  },
+  {
+    0x00000000UL, 0x01c26a37UL, 0x0384d46eUL, 0x0246be59UL, 0x0709a8dcUL,
+    0x06cbc2ebUL, 0x048d7cb2UL, 0x054f1685UL, 0x0e1351b8UL, 0x0fd13b8fUL,
+    0x0d9785d6UL, 0x0c55efe1UL, 0x091af964UL, 0x08d89353UL, 0x0a9e2d0aUL,
+    0x0b5c473dUL, 0x1c26a370UL, 0x1de4c947UL, 0x1fa2771eUL, 0x1e601d29UL,
+    0x1b2f0bacUL, 0x1aed619bUL, 0x18abdfc2UL, 0x1969b5f5UL, 0x1235f2c8UL,
+    0x13f798ffUL, 0x11b126a6UL, 0x10734c91UL, 0x153c5a14UL, 0x14fe3023UL,
+    0x16b88e7aUL, 0x177ae44dUL, 0x384d46e0UL, 0x398f2cd7UL, 0x3bc9928eUL,
+    0x3a0bf8b9UL, 0x3f44ee3cUL, 0x3e86840bUL, 0x3cc03a52UL, 0x3d025065UL,
+    0x365e1758UL, 0x379c7d6fUL, 0x35dac336UL, 0x3418a901UL, 0x3157bf84UL,
+    0x3095d5b3UL, 0x32d36beaUL, 0x331101ddUL, 0x246be590UL, 0x25a98fa7UL,
+    0x27ef31feUL, 0x262d5bc9UL, 0x23624d4cUL, 0x22a0277bUL, 0x20e69922UL,
+    0x2124f315UL, 0x2a78b428UL, 0x2bbade1fUL, 0x29fc6046UL, 0x283e0a71UL,
+    0x2d711cf4UL, 0x2cb376c3UL, 0x2ef5c89aUL, 0x2f37a2adUL, 0x709a8dc0UL,
+    0x7158e7f7UL, 0x731e59aeUL, 0x72dc3399UL, 0x7793251cUL, 0x76514f2bUL,
+    0x7417f172UL, 0x75d59b45UL, 0x7e89dc78UL, 0x7f4bb64fUL, 0x7d0d0816UL,
+    0x7ccf6221UL, 0x798074a4UL, 0x78421e93UL, 0x7a04a0caUL, 0x7bc6cafdUL,
+    0x6cbc2eb0UL, 0x6d7e4487UL, 0x6f38fadeUL, 0x6efa90e9UL, 0x6bb5866cUL,
+    0x6a77ec5bUL, 0x68315202UL, 0x69f33835UL, 0x62af7f08UL, 0x636d153fUL,
+    0x612bab66UL, 0x60e9c151UL, 0x65a6d7d4UL, 0x6464bde3UL, 0x662203baUL,
+    0x67e0698dUL, 0x48d7cb20UL, 0x4915a117UL, 0x4b531f4eUL, 0x4a917579UL,
+    0x4fde63fcUL, 0x4e1c09cbUL, 0x4c5ab792UL, 0x4d98dda5UL, 0x46c49a98UL,
+    0x4706f0afUL, 0x45404ef6UL, 0x448224c1UL, 0x41cd3244UL, 0x400f5873UL,
+    0x4249e62aUL, 0x438b8c1dUL, 0x54f16850UL, 0x55330267UL, 0x5775bc3eUL,
+    0x56b7d609UL, 0x53f8c08cUL, 0x523aaabbUL, 0x507c14e2UL, 0x51be7ed5UL,
+    0x5ae239e8UL, 0x5b2053dfUL, 0x5966ed86UL, 0x58a487b1UL, 0x5deb9134UL,
+    0x5c29fb03UL, 0x5e6f455aUL, 0x5fad2f6dUL, 0xe1351b80UL, 0xe0f771b7UL,
+    0xe2b1cfeeUL, 0xe373a5d9UL, 0xe63cb35cUL, 0xe7fed96bUL, 0xe5b86732UL,
+    0xe47a0d05UL, 0xef264a38UL, 0xeee4200fUL, 0xeca29e56UL, 0xed60f461UL,
+    0xe82fe2e4UL, 0xe9ed88d3UL, 0xebab368aUL, 0xea695cbdUL, 0xfd13b8f0UL,
+    0xfcd1d2c7UL, 0xfe976c9eUL, 0xff5506a9UL, 0xfa1a102cUL, 0xfbd87a1bUL,
+    0xf99ec442UL, 0xf85cae75UL, 0xf300e948UL, 0xf2c2837fUL, 0xf0843d26UL,
+    0xf1465711UL, 0xf4094194UL, 0xf5cb2ba3UL, 0xf78d95faUL, 0xf64fffcdUL,
+    0xd9785d60UL, 0xd8ba3757UL, 0xdafc890eUL, 0xdb3ee339UL, 0xde71f5bcUL,
+    0xdfb39f8bUL, 0xddf521d2UL, 0xdc374be5UL, 0xd76b0cd8UL, 0xd6a966efUL,
+    0xd4efd8b6UL, 0xd52db281UL, 0xd062a404UL, 0xd1a0ce33UL, 0xd3e6706aUL,
+    0xd2241a5dUL, 0xc55efe10UL, 0xc49c9427UL, 0xc6da2a7eUL, 0xc7184049UL,
+    0xc25756ccUL, 0xc3953cfbUL, 0xc1d382a2UL, 0xc011e895UL, 0xcb4dafa8UL,
+    0xca8fc59fUL, 0xc8c97bc6UL, 0xc90b11f1UL, 0xcc440774UL, 0xcd866d43UL,
+    0xcfc0d31aUL, 0xce02b92dUL, 0x91af9640UL, 0x906dfc77UL, 0x922b422eUL,
+    0x93e92819UL, 0x96a63e9cUL, 0x976454abUL, 0x9522eaf2UL, 0x94e080c5UL,
+    0x9fbcc7f8UL, 0x9e7eadcfUL, 0x9c381396UL, 0x9dfa79a1UL, 0x98b56f24UL,
+    0x99770513UL, 0x9b31bb4aUL, 0x9af3d17dUL, 0x8d893530UL, 0x8c4b5f07UL,
+    0x8e0de15eUL, 0x8fcf8b69UL, 0x8a809decUL, 0x8b42f7dbUL, 0x89044982UL,
+    0x88c623b5UL, 0x839a6488UL, 0x82580ebfUL, 0x801eb0e6UL, 0x81dcdad1UL,
+    0x8493cc54UL, 0x8551a663UL, 0x8717183aUL, 0x86d5720dUL, 0xa9e2d0a0UL,
+    0xa820ba97UL, 0xaa6604ceUL, 0xaba46ef9UL, 0xaeeb787cUL, 0xaf29124bUL,
+    0xad6fac12UL, 0xacadc625UL, 0xa7f18118UL, 0xa633eb2fUL, 0xa4755576UL,
+    0xa5b73f41UL, 0xa0f829c4UL, 0xa13a43f3UL, 0xa37cfdaaUL, 0xa2be979dUL,
+    0xb5c473d0UL, 0xb40619e7UL, 0xb640a7beUL, 0xb782cd89UL, 0xb2cddb0cUL,
+    0xb30fb13bUL, 0xb1490f62UL, 0xb08b6555UL, 0xbbd72268UL, 0xba15485fUL,
+    0xb853f606UL, 0xb9919c31UL, 0xbcde8ab4UL, 0xbd1ce083UL, 0xbf5a5edaUL,
+    0xbe9834edUL
+  },
+  {
+    0x00000000UL, 0xb8bc6765UL, 0xaa09c88bUL, 0x12b5afeeUL, 0x8f629757UL,
+    0x37def032UL, 0x256b5fdcUL, 0x9dd738b9UL, 0xc5b428efUL, 0x7d084f8aUL,
+    0x6fbde064UL, 0xd7018701UL, 0x4ad6bfb8UL, 0xf26ad8ddUL, 0xe0df7733UL,
+    0x58631056UL, 0x5019579fUL, 0xe8a530faUL, 0xfa109f14UL, 0x42acf871UL,
+    0xdf7bc0c8UL, 0x67c7a7adUL, 0x75720843UL, 0xcdce6f26UL, 0x95ad7f70UL,
+    0x2d111815UL, 0x3fa4b7fbUL, 0x8718d09eUL, 0x1acfe827UL, 0xa2738f42UL,
+    0xb0c620acUL, 0x087a47c9UL, 0xa032af3eUL, 0x188ec85bUL, 0x0a3b67b5UL,
+    0xb28700d0UL, 0x2f503869UL, 0x97ec5f0cUL, 0x8559f0e2UL, 0x3de59787UL,
+    0x658687d1UL, 0xdd3ae0b4UL, 0xcf8f4f5aUL, 0x7733283fUL, 0xeae41086UL,
+    0x525877e3UL, 0x40edd80dUL, 0xf851bf68UL, 0xf02bf8a1UL, 0x48979fc4UL,
+    0x5a22302aUL, 0xe29e574fUL, 0x7f496ff6UL, 0xc7f50893UL, 0xd540a77dUL,
+    0x6dfcc018UL, 0x359fd04eUL, 0x8d23b72bUL, 0x9f9618c5UL, 0x272a7fa0UL,
+    0xbafd4719UL, 0x0241207cUL, 0x10f48f92UL, 0xa848e8f7UL, 0x9b14583dUL,
+    0x23a83f58UL, 0x311d90b6UL, 0x89a1f7d3UL, 0x1476cf6aUL, 0xaccaa80fUL,
+    0xbe7f07e1UL, 0x06c36084UL, 0x5ea070d2UL, 0xe61c17b7UL, 0xf4a9b859UL,
+    0x4c15df3cUL, 0xd1c2e785UL, 0x697e80e0UL, 0x7bcb2f0eUL, 0xc377486bUL,
+    0xcb0d0fa2UL, 0x73b168c7UL, 0x6104c729UL, 0xd9b8a04cUL, 0x446f98f5UL,
+    0xfcd3ff90UL, 0xee66507eUL, 0x56da371bUL, 0x0eb9274dUL, 0xb6054028UL,
+    0xa4b0efc6UL, 0x1c0c88a3UL, 0x81dbb01aUL, 0x3967d77fUL, 0x2bd27891UL,
+    0x936e1ff4UL, 0x3b26f703UL, 0x839a9066UL, 0x912f3f88UL, 0x299358edUL,
+    0xb4446054UL, 0x0cf80731UL, 0x1e4da8dfUL, 0xa6f1cfbaUL, 0xfe92dfecUL,
+    0x462eb889UL, 0x549b1767UL, 0xec277002UL, 0x71f048bbUL, 0xc94c2fdeUL,
+    0xdbf98030UL, 0x6345e755UL, 0x6b3fa09cUL, 0xd383c7f9UL, 0xc1366817UL,
+    0x798a0f72UL, 0xe45d37cbUL, 0x5ce150aeUL, 0x4e54ff40UL, 0xf6e89825UL,
+    0xae8b8873UL, 0x1637ef16UL, 0x048240f8UL, 0xbc3e279dUL, 0x21e91f24UL,
+    0x99557841UL, 0x8be0d7afUL, 0x335cb0caUL, 0xed59b63bUL, 0x55e5d15eUL,
+    0x47507eb0UL, 0xffec19d5UL, 0x623b216cUL, 0xda874609UL, 0xc832e9e7UL,
+    0x708e8e82UL, 0x28ed9ed4UL, 0x9051f9b1UL, 0x82e4565fUL, 0x3a58313aUL,
+    0xa78f0983UL, 0x1f336ee6UL, 0x0d86c108UL, 0xb53aa66dUL, 0xbd40e1a4UL,
+    0x05fc86c1UL, 0x1749292fUL, 0xaff54e4aUL, 0x322276f3UL, 0x8a9e1196UL,
+    0x982bbe78UL, 0x2097d91dUL, 0x78f4c94bUL, 0xc048ae2eUL, 0xd2fd01c0UL,
+    0x6a4166a5UL, 0xf7965e1cUL, 0x4f2a3979UL, 0x5d9f9697UL, 0xe523f1f2UL,
+    0x4d6b1905UL, 0xf5d77e60UL, 0xe762d18eUL, 0x5fdeb6ebUL, 0xc2098e52UL,
+    0x7ab5e937UL, 0x680046d9UL, 0xd0bc21bcUL, 0x88df31eaUL, 0x3063568fUL,
+    0x22d6f961UL, 0x9a6a9e04UL, 0x07bda6bdUL, 0xbf01c1d8UL, 0xadb46e36UL,
+    0x15080953UL, 0x1d724e9aUL, 0xa5ce29ffUL, 0xb77b8611UL, 0x0fc7e174UL,
+    0x9210d9cdUL, 0x2aacbea8UL, 0x38191146UL, 0x80a57623UL, 0xd8c66675UL,
+    0x607a0110UL, 0x72cfaefeUL, 0xca73c99bUL, 0x57a4f122UL, 0xef189647UL,
+    0xfdad39a9UL, 0x45115eccUL, 0x764dee06UL, 0xcef18963UL, 0xdc44268dUL,
+    0x64f841e8UL, 0xf92f7951UL, 0x41931e34UL, 0x5326b1daUL, 0xeb9ad6bfUL,
+    0xb3f9c6e9UL, 0x0b45a18cUL, 0x19f00e62UL, 0xa14c6907UL, 0x3c9b51beUL,
+    0x842736dbUL, 0x96929935UL, 0x2e2efe50UL, 0x2654b999UL, 0x9ee8defcUL,
+    0x8c5d7112UL, 0x34e11677UL, 0xa9362eceUL, 0x118a49abUL, 0x033fe645UL,
+    0xbb838120UL, 0xe3e09176UL, 0x5b5cf613UL, 0x49e959fdUL, 0xf1553e98UL,
+    0x6c820621UL, 0xd43e6144UL, 0xc68bceaaUL, 0x7e37a9cfUL, 0xd67f4138UL,
+    0x6ec3265dUL, 0x7c7689b3UL, 0xc4caeed6UL, 0x591dd66fUL, 0xe1a1b10aUL,
+    0xf3141ee4UL, 0x4ba87981UL, 0x13cb69d7UL, 0xab770eb2UL, 0xb9c2a15cUL,
+    0x017ec639UL, 0x9ca9fe80UL, 0x241599e5UL, 0x36a0360bUL, 0x8e1c516eUL,
+    0x866616a7UL, 0x3eda71c2UL, 0x2c6fde2cUL, 0x94d3b949UL, 0x090481f0UL,
+    0xb1b8e695UL, 0xa30d497bUL, 0x1bb12e1eUL, 0x43d23e48UL, 0xfb6e592dUL,
+    0xe9dbf6c3UL, 0x516791a6UL, 0xccb0a91fUL, 0x740cce7aUL, 0x66b96194UL,
+    0xde0506f1UL
+  },
+  {
+    0x00000000UL, 0x96300777UL, 0x2c610eeeUL, 0xba510999UL, 0x19c46d07UL,
+    0x8ff46a70UL, 0x35a563e9UL, 0xa395649eUL, 0x3288db0eUL, 0xa4b8dc79UL,
+    0x1ee9d5e0UL, 0x88d9d297UL, 0x2b4cb609UL, 0xbd7cb17eUL, 0x072db8e7UL,
+    0x911dbf90UL, 0x6410b71dUL, 0xf220b06aUL, 0x4871b9f3UL, 0xde41be84UL,
+    0x7dd4da1aUL, 0xebe4dd6dUL, 0x51b5d4f4UL, 0xc785d383UL, 0x56986c13UL,
+    0xc0a86b64UL, 0x7af962fdUL, 0xecc9658aUL, 0x4f5c0114UL, 0xd96c0663UL,
+    0x633d0ffaUL, 0xf50d088dUL, 0xc8206e3bUL, 0x5e10694cUL, 0xe44160d5UL,
+    0x727167a2UL, 0xd1e4033cUL, 0x47d4044bUL, 0xfd850dd2UL, 0x6bb50aa5UL,
+    0xfaa8b535UL, 0x6c98b242UL, 0xd6c9bbdbUL, 0x40f9bcacUL, 0xe36cd832UL,
+    0x755cdf45UL, 0xcf0dd6dcUL, 0x593dd1abUL, 0xac30d926UL, 0x3a00de51UL,
+    0x8051d7c8UL, 0x1661d0bfUL, 0xb5f4b421UL, 0x23c4b356UL, 0x9995bacfUL,
+    0x0fa5bdb8UL, 0x9eb80228UL, 0x0888055fUL, 0xb2d90cc6UL, 0x24e90bb1UL,
+    0x877c6f2fUL, 0x114c6858UL, 0xab1d61c1UL, 0x3d2d66b6UL, 0x9041dc76UL,
+    0x0671db01UL, 0xbc20d298UL, 0x2a10d5efUL, 0x8985b171UL, 0x1fb5b606UL,
+    0xa5e4bf9fUL, 0x33d4b8e8UL, 0xa2c90778UL, 0x34f9000fUL, 0x8ea80996UL,
+    0x18980ee1UL, 0xbb0d6a7fUL, 0x2d3d6d08UL, 0x976c6491UL, 0x015c63e6UL,
+    0xf4516b6bUL, 0x62616c1cUL, 0xd8306585UL, 0x4e0062f2UL, 0xed95066cUL,
+    0x7ba5011bUL, 0xc1f40882UL, 0x57c40ff5UL, 0xc6d9b065UL, 0x50e9b712UL,
+    0xeab8be8bUL, 0x7c88b9fcUL, 0xdf1ddd62UL, 0x492dda15UL, 0xf37cd38cUL,
+    0x654cd4fbUL, 0x5861b24dUL, 0xce51b53aUL, 0x7400bca3UL, 0xe230bbd4UL,
+    0x41a5df4aUL, 0xd795d83dUL, 0x6dc4d1a4UL, 0xfbf4d6d3UL, 0x6ae96943UL,
+    0xfcd96e34UL, 0x468867adUL, 0xd0b860daUL, 0x732d0444UL, 0xe51d0333UL,
+    0x5f4c0aaaUL, 0xc97c0dddUL, 0x3c710550UL, 0xaa410227UL, 0x10100bbeUL,
+    0x86200cc9UL, 0x25b56857UL, 0xb3856f20UL, 0x09d466b9UL, 0x9fe461ceUL,
+    0x0ef9de5eUL, 0x98c9d929UL, 0x2298d0b0UL, 0xb4a8d7c7UL, 0x173db359UL,
+    0x810db42eUL, 0x3b5cbdb7UL, 0xad6cbac0UL, 0x2083b8edUL, 0xb6b3bf9aUL,
+    0x0ce2b603UL, 0x9ad2b174UL, 0x3947d5eaUL, 0xaf77d29dUL, 0x1526db04UL,
+    0x8316dc73UL, 0x120b63e3UL, 0x843b6494UL, 0x3e6a6d0dUL, 0xa85a6a7aUL,
+    0x0bcf0ee4UL, 0x9dff0993UL, 0x27ae000aUL, 0xb19e077dUL, 0x44930ff0UL,
+    0xd2a30887UL, 0x68f2011eUL, 0xfec20669UL, 0x5d5762f7UL, 0xcb676580UL,
+    0x71366c19UL, 0xe7066b6eUL, 0x761bd4feUL, 0xe02bd389UL, 0x5a7ada10UL,
+    0xcc4add67UL, 0x6fdfb9f9UL, 0xf9efbe8eUL, 0x43beb717UL, 0xd58eb060UL,
+    0xe8a3d6d6UL, 0x7e93d1a1UL, 0xc4c2d838UL, 0x52f2df4fUL, 0xf167bbd1UL,
+    0x6757bca6UL, 0xdd06b53fUL, 0x4b36b248UL, 0xda2b0dd8UL, 0x4c1b0aafUL,
+    0xf64a0336UL, 0x607a0441UL, 0xc3ef60dfUL, 0x55df67a8UL, 0xef8e6e31UL,
+    0x79be6946UL, 0x8cb361cbUL, 0x1a8366bcUL, 0xa0d26f25UL, 0x36e26852UL,
+    0x95770cccUL, 0x03470bbbUL, 0xb9160222UL, 0x2f260555UL, 0xbe3bbac5UL,
+    0x280bbdb2UL, 0x925ab42bUL, 0x046ab35cUL, 0xa7ffd7c2UL, 0x31cfd0b5UL,
+    0x8b9ed92cUL, 0x1daede5bUL, 0xb0c2649bUL, 0x26f263ecUL, 0x9ca36a75UL,
+    0x0a936d02UL, 0xa906099cUL, 0x3f360eebUL, 0x85670772UL, 0x13570005UL,
+    0x824abf95UL, 0x147ab8e2UL, 0xae2bb17bUL, 0x381bb60cUL, 0x9b8ed292UL,
+    0x0dbed5e5UL, 0xb7efdc7cUL, 0x21dfdb0bUL, 0xd4d2d386UL, 0x42e2d4f1UL,
+    0xf8b3dd68UL, 0x6e83da1fUL, 0xcd16be81UL, 0x5b26b9f6UL, 0xe177b06fUL,
+    0x7747b718UL, 0xe65a0888UL, 0x706a0fffUL, 0xca3b0666UL, 0x5c0b0111UL,
+    0xff9e658fUL, 0x69ae62f8UL, 0xd3ff6b61UL, 0x45cf6c16UL, 0x78e20aa0UL,
+    0xeed20dd7UL, 0x5483044eUL, 0xc2b30339UL, 0x612667a7UL, 0xf71660d0UL,
+    0x4d476949UL, 0xdb776e3eUL, 0x4a6ad1aeUL, 0xdc5ad6d9UL, 0x660bdf40UL,
+    0xf03bd837UL, 0x53aebca9UL, 0xc59ebbdeUL, 0x7fcfb247UL, 0xe9ffb530UL,
+    0x1cf2bdbdUL, 0x8ac2bacaUL, 0x3093b353UL, 0xa6a3b424UL, 0x0536d0baUL,
+    0x9306d7cdUL, 0x2957de54UL, 0xbf67d923UL, 0x2e7a66b3UL, 0xb84a61c4UL,
+    0x021b685dUL, 0x942b6f2aUL, 0x37be0bb4UL, 0xa18e0cc3UL, 0x1bdf055aUL,
+    0x8def022dUL
+  },
+  {
+    0x00000000UL, 0x41311b19UL, 0x82623632UL, 0xc3532d2bUL, 0x04c56c64UL,
+    0x45f4777dUL, 0x86a75a56UL, 0xc796414fUL, 0x088ad9c8UL, 0x49bbc2d1UL,
+    0x8ae8effaUL, 0xcbd9f4e3UL, 0x0c4fb5acUL, 0x4d7eaeb5UL, 0x8e2d839eUL,
+    0xcf1c9887UL, 0x5112c24aUL, 0x1023d953UL, 0xd370f478UL, 0x9241ef61UL,
+    0x55d7ae2eUL, 0x14e6b537UL, 0xd7b5981cUL, 0x96848305UL, 0x59981b82UL,
+    0x18a9009bUL, 0xdbfa2db0UL, 0x9acb36a9UL, 0x5d5d77e6UL, 0x1c6c6cffUL,
+    0xdf3f41d4UL, 0x9e0e5acdUL, 0xa2248495UL, 0xe3159f8cUL, 0x2046b2a7UL,
+    0x6177a9beUL, 0xa6e1e8f1UL, 0xe7d0f3e8UL, 0x2483dec3UL, 0x65b2c5daUL,
+    0xaaae5d5dUL, 0xeb9f4644UL, 0x28cc6b6fUL, 0x69fd7076UL, 0xae6b3139UL,
+    0xef5a2a20UL, 0x2c09070bUL, 0x6d381c12UL, 0xf33646dfUL, 0xb2075dc6UL,
+    0x715470edUL, 0x30656bf4UL, 0xf7f32abbUL, 0xb6c231a2UL, 0x75911c89UL,
+    0x34a00790UL, 0xfbbc9f17UL, 0xba8d840eUL, 0x79dea925UL, 0x38efb23cUL,
+    0xff79f373UL, 0xbe48e86aUL, 0x7d1bc541UL, 0x3c2ade58UL, 0x054f79f0UL,
+    0x447e62e9UL, 0x872d4fc2UL, 0xc61c54dbUL, 0x018a1594UL, 0x40bb0e8dUL,
+    0x83e823a6UL, 0xc2d938bfUL, 0x0dc5a038UL, 0x4cf4bb21UL, 0x8fa7960aUL,
+    0xce968d13UL, 0x0900cc5cUL, 0x4831d745UL, 0x8b62fa6eUL, 0xca53e177UL,
+    0x545dbbbaUL, 0x156ca0a3UL, 0xd63f8d88UL, 0x970e9691UL, 0x5098d7deUL,
+    0x11a9ccc7UL, 0xd2fae1ecUL, 0x93cbfaf5UL, 0x5cd76272UL, 0x1de6796bUL,
+    0xdeb55440UL, 0x9f844f59UL, 0x58120e16UL, 0x1923150fUL, 0xda703824UL,
+    0x9b41233dUL, 0xa76bfd65UL, 0xe65ae67cUL, 0x2509cb57UL, 0x6438d04eUL,
+    0xa3ae9101UL, 0xe29f8a18UL, 0x21cca733UL, 0x60fdbc2aUL, 0xafe124adUL,
+    0xeed03fb4UL, 0x2d83129fUL, 0x6cb20986UL, 0xab2448c9UL, 0xea1553d0UL,
+    0x29467efbUL, 0x687765e2UL, 0xf6793f2fUL, 0xb7482436UL, 0x741b091dUL,
+    0x352a1204UL, 0xf2bc534bUL, 0xb38d4852UL, 0x70de6579UL, 0x31ef7e60UL,
+    0xfef3e6e7UL, 0xbfc2fdfeUL, 0x7c91d0d5UL, 0x3da0cbccUL, 0xfa368a83UL,
+    0xbb07919aUL, 0x7854bcb1UL, 0x3965a7a8UL, 0x4b98833bUL, 0x0aa99822UL,
+    0xc9fab509UL, 0x88cbae10UL, 0x4f5def5fUL, 0x0e6cf446UL, 0xcd3fd96dUL,
+    0x8c0ec274UL, 0x43125af3UL, 0x022341eaUL, 0xc1706cc1UL, 0x804177d8UL,
+    0x47d73697UL, 0x06e62d8eUL, 0xc5b500a5UL, 0x84841bbcUL, 0x1a8a4171UL,
+    0x5bbb5a68UL, 0x98e87743UL, 0xd9d96c5aUL, 0x1e4f2d15UL, 0x5f7e360cUL,
+    0x9c2d1b27UL, 0xdd1c003eUL, 0x120098b9UL, 0x533183a0UL, 0x9062ae8bUL,
+    0xd153b592UL, 0x16c5f4ddUL, 0x57f4efc4UL, 0x94a7c2efUL, 0xd596d9f6UL,
+    0xe9bc07aeUL, 0xa88d1cb7UL, 0x6bde319cUL, 0x2aef2a85UL, 0xed796bcaUL,
+    0xac4870d3UL, 0x6f1b5df8UL, 0x2e2a46e1UL, 0xe136de66UL, 0xa007c57fUL,
+    0x6354e854UL, 0x2265f34dUL, 0xe5f3b202UL, 0xa4c2a91bUL, 0x67918430UL,
+    0x26a09f29UL, 0xb8aec5e4UL, 0xf99fdefdUL, 0x3accf3d6UL, 0x7bfde8cfUL,
+    0xbc6ba980UL, 0xfd5ab299UL, 0x3e099fb2UL, 0x7f3884abUL, 0xb0241c2cUL,
+    0xf1150735UL, 0x32462a1eUL, 0x73773107UL, 0xb4e17048UL, 0xf5d06b51UL,
+    0x3683467aUL, 0x77b25d63UL, 0x4ed7facbUL, 0x0fe6e1d2UL, 0xccb5ccf9UL,
+    0x8d84d7e0UL, 0x4a1296afUL, 0x0b238db6UL, 0xc870a09dUL, 0x8941bb84UL,
+    0x465d2303UL, 0x076c381aUL, 0xc43f1531UL, 0x850e0e28UL, 0x42984f67UL,
+    0x03a9547eUL, 0xc0fa7955UL, 0x81cb624cUL, 0x1fc53881UL, 0x5ef42398UL,
+    0x9da70eb3UL, 0xdc9615aaUL, 0x1b0054e5UL, 0x5a314ffcUL, 0x996262d7UL,
+    0xd85379ceUL, 0x174fe149UL, 0x567efa50UL, 0x952dd77bUL, 0xd41ccc62UL,
+    0x138a8d2dUL, 0x52bb9634UL, 0x91e8bb1fUL, 0xd0d9a006UL, 0xecf37e5eUL,
+    0xadc26547UL, 0x6e91486cUL, 0x2fa05375UL, 0xe836123aUL, 0xa9070923UL,
+    0x6a542408UL, 0x2b653f11UL, 0xe479a796UL, 0xa548bc8fUL, 0x661b91a4UL,
+    0x272a8abdUL, 0xe0bccbf2UL, 0xa18dd0ebUL, 0x62defdc0UL, 0x23efe6d9UL,
+    0xbde1bc14UL, 0xfcd0a70dUL, 0x3f838a26UL, 0x7eb2913fUL, 0xb924d070UL,
+    0xf815cb69UL, 0x3b46e642UL, 0x7a77fd5bUL, 0xb56b65dcUL, 0xf45a7ec5UL,
+    0x370953eeUL, 0x763848f7UL, 0xb1ae09b8UL, 0xf09f12a1UL, 0x33cc3f8aUL,
+    0x72fd2493UL
+  },
+  {
+    0x00000000UL, 0x376ac201UL, 0x6ed48403UL, 0x59be4602UL, 0xdca80907UL,
+    0xebc2cb06UL, 0xb27c8d04UL, 0x85164f05UL, 0xb851130eUL, 0x8f3bd10fUL,
+    0xd685970dUL, 0xe1ef550cUL, 0x64f91a09UL, 0x5393d808UL, 0x0a2d9e0aUL,
+    0x3d475c0bUL, 0x70a3261cUL, 0x47c9e41dUL, 0x1e77a21fUL, 0x291d601eUL,
+    0xac0b2f1bUL, 0x9b61ed1aUL, 0xc2dfab18UL, 0xf5b56919UL, 0xc8f23512UL,
+    0xff98f713UL, 0xa626b111UL, 0x914c7310UL, 0x145a3c15UL, 0x2330fe14UL,
+    0x7a8eb816UL, 0x4de47a17UL, 0xe0464d38UL, 0xd72c8f39UL, 0x8e92c93bUL,
+    0xb9f80b3aUL, 0x3cee443fUL, 0x0b84863eUL, 0x523ac03cUL, 0x6550023dUL,
+    0x58175e36UL, 0x6f7d9c37UL, 0x36c3da35UL, 0x01a91834UL, 0x84bf5731UL,
+    0xb3d59530UL, 0xea6bd332UL, 0xdd011133UL, 0x90e56b24UL, 0xa78fa925UL,
+    0xfe31ef27UL, 0xc95b2d26UL, 0x4c4d6223UL, 0x7b27a022UL, 0x2299e620UL,
+    0x15f32421UL, 0x28b4782aUL, 0x1fdeba2bUL, 0x4660fc29UL, 0x710a3e28UL,
+    0xf41c712dUL, 0xc376b32cUL, 0x9ac8f52eUL, 0xada2372fUL, 0xc08d9a70UL,
+    0xf7e75871UL, 0xae591e73UL, 0x9933dc72UL, 0x1c259377UL, 0x2b4f5176UL,
+    0x72f11774UL, 0x459bd575UL, 0x78dc897eUL, 0x4fb64b7fUL, 0x16080d7dUL,
+    0x2162cf7cUL, 0xa4748079UL, 0x931e4278UL, 0xcaa0047aUL, 0xfdcac67bUL,
+    0xb02ebc6cUL, 0x87447e6dUL, 0xdefa386fUL, 0xe990fa6eUL, 0x6c86b56bUL,
+    0x5bec776aUL, 0x02523168UL, 0x3538f369UL, 0x087faf62UL, 0x3f156d63UL,
+    0x66ab2b61UL, 0x51c1e960UL, 0xd4d7a665UL, 0xe3bd6464UL, 0xba032266UL,
+    0x8d69e067UL, 0x20cbd748UL, 0x17a11549UL, 0x4e1f534bUL, 0x7975914aUL,
+    0xfc63de4fUL, 0xcb091c4eUL, 0x92b75a4cUL, 0xa5dd984dUL, 0x989ac446UL,
+    0xaff00647UL, 0xf64e4045UL, 0xc1248244UL, 0x4432cd41UL, 0x73580f40UL,
+    0x2ae64942UL, 0x1d8c8b43UL, 0x5068f154UL, 0x67023355UL, 0x3ebc7557UL,
+    0x09d6b756UL, 0x8cc0f853UL, 0xbbaa3a52UL, 0xe2147c50UL, 0xd57ebe51UL,
+    0xe839e25aUL, 0xdf53205bUL, 0x86ed6659UL, 0xb187a458UL, 0x3491eb5dUL,
+    0x03fb295cUL, 0x5a456f5eUL, 0x6d2fad5fUL, 0x801b35e1UL, 0xb771f7e0UL,
+    0xeecfb1e2UL, 0xd9a573e3UL, 0x5cb33ce6UL, 0x6bd9fee7UL, 0x3267b8e5UL,
+    0x050d7ae4UL, 0x384a26efUL, 0x0f20e4eeUL, 0x569ea2ecUL, 0x61f460edUL,
+    0xe4e22fe8UL, 0xd388ede9UL, 0x8a36abebUL, 0xbd5c69eaUL, 0xf0b813fdUL,
+    0xc7d2d1fcUL, 0x9e6c97feUL, 0xa90655ffUL, 0x2c101afaUL, 0x1b7ad8fbUL,
+    0x42c49ef9UL, 0x75ae5cf8UL, 0x48e900f3UL, 0x7f83c2f2UL, 0x263d84f0UL,
+    0x115746f1UL, 0x944109f4UL, 0xa32bcbf5UL, 0xfa958df7UL, 0xcdff4ff6UL,
+    0x605d78d9UL, 0x5737bad8UL, 0x0e89fcdaUL, 0x39e33edbUL, 0xbcf571deUL,
+    0x8b9fb3dfUL, 0xd221f5ddUL, 0xe54b37dcUL, 0xd80c6bd7UL, 0xef66a9d6UL,
+    0xb6d8efd4UL, 0x81b22dd5UL, 0x04a462d0UL, 0x33cea0d1UL, 0x6a70e6d3UL,
+    0x5d1a24d2UL, 0x10fe5ec5UL, 0x27949cc4UL, 0x7e2adac6UL, 0x494018c7UL,
+    0xcc5657c2UL, 0xfb3c95c3UL, 0xa282d3c1UL, 0x95e811c0UL, 0xa8af4dcbUL,
+    0x9fc58fcaUL, 0xc67bc9c8UL, 0xf1110bc9UL, 0x740744ccUL, 0x436d86cdUL,
+    0x1ad3c0cfUL, 0x2db902ceUL, 0x4096af91UL, 0x77fc6d90UL, 0x2e422b92UL,
+    0x1928e993UL, 0x9c3ea696UL, 0xab546497UL, 0xf2ea2295UL, 0xc580e094UL,
+    0xf8c7bc9fUL, 0xcfad7e9eUL, 0x9613389cUL, 0xa179fa9dUL, 0x246fb598UL,
+    0x13057799UL, 0x4abb319bUL, 0x7dd1f39aUL, 0x3035898dUL, 0x075f4b8cUL,
+    0x5ee10d8eUL, 0x698bcf8fUL, 0xec9d808aUL, 0xdbf7428bUL, 0x82490489UL,
+    0xb523c688UL, 0x88649a83UL, 0xbf0e5882UL, 0xe6b01e80UL, 0xd1dadc81UL,
+    0x54cc9384UL, 0x63a65185UL, 0x3a181787UL, 0x0d72d586UL, 0xa0d0e2a9UL,
+    0x97ba20a8UL, 0xce0466aaUL, 0xf96ea4abUL, 0x7c78ebaeUL, 0x4b1229afUL,
+    0x12ac6fadUL, 0x25c6adacUL, 0x1881f1a7UL, 0x2feb33a6UL, 0x765575a4UL,
+    0x413fb7a5UL, 0xc429f8a0UL, 0xf3433aa1UL, 0xaafd7ca3UL, 0x9d97bea2UL,
+    0xd073c4b5UL, 0xe71906b4UL, 0xbea740b6UL, 0x89cd82b7UL, 0x0cdbcdb2UL,
+    0x3bb10fb3UL, 0x620f49b1UL, 0x55658bb0UL, 0x6822d7bbUL, 0x5f4815baUL,
+    0x06f653b8UL, 0x319c91b9UL, 0xb48adebcUL, 0x83e01cbdUL, 0xda5e5abfUL,
+    0xed3498beUL
+  },
+  {
+    0x00000000UL, 0x6567bcb8UL, 0x8bc809aaUL, 0xeeafb512UL, 0x5797628fUL,
+    0x32f0de37UL, 0xdc5f6b25UL, 0xb938d79dUL, 0xef28b4c5UL, 0x8a4f087dUL,
+    0x64e0bd6fUL, 0x018701d7UL, 0xb8bfd64aUL, 0xddd86af2UL, 0x3377dfe0UL,
+    0x56106358UL, 0x9f571950UL, 0xfa30a5e8UL, 0x149f10faUL, 0x71f8ac42UL,
+    0xc8c07bdfUL, 0xada7c767UL, 0x43087275UL, 0x266fcecdUL, 0x707fad95UL,
+    0x1518112dUL, 0xfbb7a43fUL, 0x9ed01887UL, 0x27e8cf1aUL, 0x428f73a2UL,
+    0xac20c6b0UL, 0xc9477a08UL, 0x3eaf32a0UL, 0x5bc88e18UL, 0xb5673b0aUL,
+    0xd00087b2UL, 0x6938502fUL, 0x0c5fec97UL, 0xe2f05985UL, 0x8797e53dUL,
+    0xd1878665UL, 0xb4e03addUL, 0x5a4f8fcfUL, 0x3f283377UL, 0x8610e4eaUL,
+    0xe3775852UL, 0x0dd8ed40UL, 0x68bf51f8UL, 0xa1f82bf0UL, 0xc49f9748UL,
+    0x2a30225aUL, 0x4f579ee2UL, 0xf66f497fUL, 0x9308f5c7UL, 0x7da740d5UL,
+    0x18c0fc6dUL, 0x4ed09f35UL, 0x2bb7238dUL, 0xc518969fUL, 0xa07f2a27UL,
+    0x1947fdbaUL, 0x7c204102UL, 0x928ff410UL, 0xf7e848a8UL, 0x3d58149bUL,
+    0x583fa823UL, 0xb6901d31UL, 0xd3f7a189UL, 0x6acf7614UL, 0x0fa8caacUL,
+    0xe1077fbeUL, 0x8460c306UL, 0xd270a05eUL, 0xb7171ce6UL, 0x59b8a9f4UL,
+    0x3cdf154cUL, 0x85e7c2d1UL, 0xe0807e69UL, 0x0e2fcb7bUL, 0x6b4877c3UL,
+    0xa20f0dcbUL, 0xc768b173UL, 0x29c70461UL, 0x4ca0b8d9UL, 0xf5986f44UL,
+    0x90ffd3fcUL, 0x7e5066eeUL, 0x1b37da56UL, 0x4d27b90eUL, 0x284005b6UL,
+    0xc6efb0a4UL, 0xa3880c1cUL, 0x1ab0db81UL, 0x7fd76739UL, 0x9178d22bUL,
+    0xf41f6e93UL, 0x03f7263bUL, 0x66909a83UL, 0x883f2f91UL, 0xed589329UL,
+    0x546044b4UL, 0x3107f80cUL, 0xdfa84d1eUL, 0xbacff1a6UL, 0xecdf92feUL,
+    0x89b82e46UL, 0x67179b54UL, 0x027027ecUL, 0xbb48f071UL, 0xde2f4cc9UL,
+    0x3080f9dbUL, 0x55e74563UL, 0x9ca03f6bUL, 0xf9c783d3UL, 0x176836c1UL,
+    0x720f8a79UL, 0xcb375de4UL, 0xae50e15cUL, 0x40ff544eUL, 0x2598e8f6UL,
+    0x73888baeUL, 0x16ef3716UL, 0xf8408204UL, 0x9d273ebcUL, 0x241fe921UL,
+    0x41785599UL, 0xafd7e08bUL, 0xcab05c33UL, 0x3bb659edUL, 0x5ed1e555UL,
+    0xb07e5047UL, 0xd519ecffUL, 0x6c213b62UL, 0x094687daUL, 0xe7e932c8UL,
+    0x828e8e70UL, 0xd49eed28UL, 0xb1f95190UL, 0x5f56e482UL, 0x3a31583aUL,
+    0x83098fa7UL, 0xe66e331fUL, 0x08c1860dUL, 0x6da63ab5UL, 0xa4e140bdUL,
+    0xc186fc05UL, 0x2f294917UL, 0x4a4ef5afUL, 0xf3762232UL, 0x96119e8aUL,
+    0x78be2b98UL, 0x1dd99720UL, 0x4bc9f478UL, 0x2eae48c0UL, 0xc001fdd2UL,
+    0xa566416aUL, 0x1c5e96f7UL, 0x79392a4fUL, 0x97969f5dUL, 0xf2f123e5UL,
+    0x05196b4dUL, 0x607ed7f5UL, 0x8ed162e7UL, 0xebb6de5fUL, 0x528e09c2UL,
+    0x37e9b57aUL, 0xd9460068UL, 0xbc21bcd0UL, 0xea31df88UL, 0x8f566330UL,
+    0x61f9d622UL, 0x049e6a9aUL, 0xbda6bd07UL, 0xd8c101bfUL, 0x366eb4adUL,
+    0x53090815UL, 0x9a4e721dUL, 0xff29cea5UL, 0x11867bb7UL, 0x74e1c70fUL,
+    0xcdd91092UL, 0xa8beac2aUL, 0x46111938UL, 0x2376a580UL, 0x7566c6d8UL,
+    0x10017a60UL, 0xfeaecf72UL, 0x9bc973caUL, 0x22f1a457UL, 0x479618efUL,
+    0xa939adfdUL, 0xcc5e1145UL, 0x06ee4d76UL, 0x6389f1ceUL, 0x8d2644dcUL,
+    0xe841f864UL, 0x51792ff9UL, 0x341e9341UL, 0xdab12653UL, 0xbfd69aebUL,
+    0xe9c6f9b3UL, 0x8ca1450bUL, 0x620ef019UL, 0x07694ca1UL, 0xbe519b3cUL,
+    0xdb362784UL, 0x35999296UL, 0x50fe2e2eUL, 0x99b95426UL, 0xfcdee89eUL,
+    0x12715d8cUL, 0x7716e134UL, 0xce2e36a9UL, 0xab498a11UL, 0x45e63f03UL,
+    0x208183bbUL, 0x7691e0e3UL, 0x13f65c5bUL, 0xfd59e949UL, 0x983e55f1UL,
+    0x2106826cUL, 0x44613ed4UL, 0xaace8bc6UL, 0xcfa9377eUL, 0x38417fd6UL,
+    0x5d26c36eUL, 0xb389767cUL, 0xd6eecac4UL, 0x6fd61d59UL, 0x0ab1a1e1UL,
+    0xe41e14f3UL, 0x8179a84bUL, 0xd769cb13UL, 0xb20e77abUL, 0x5ca1c2b9UL,
+    0x39c67e01UL, 0x80fea99cUL, 0xe5991524UL, 0x0b36a036UL, 0x6e511c8eUL,
+    0xa7166686UL, 0xc271da3eUL, 0x2cde6f2cUL, 0x49b9d394UL, 0xf0810409UL,
+    0x95e6b8b1UL, 0x7b490da3UL, 0x1e2eb11bUL, 0x483ed243UL, 0x2d596efbUL,
+    0xc3f6dbe9UL, 0xa6916751UL, 0x1fa9b0ccUL, 0x7ace0c74UL, 0x9461b966UL,
+    0xf10605deUL
+#endif
+  }
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/deflate.c glmark2-2021.02/android/jni/src/zlib/deflate.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/deflate.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/deflate.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,2163 @@
+/* deflate.c -- compress data using the deflation algorithm
+ * Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process depends on being able to identify portions
+ *      of the input text which are identical to earlier input (within a
+ *      sliding window trailing behind the input currently being processed).
+ *
+ *      The most straightforward technique turns out to be the fastest for
+ *      most input files: try all possible matches and select the longest.
+ *      The key feature of this algorithm is that insertions into the string
+ *      dictionary are very simple and thus fast, and deletions are avoided
+ *      completely. Insertions are performed at each input character, whereas
+ *      string matches are performed only when the previous match ends. So it
+ *      is preferable to spend more time in matches to allow very fast string
+ *      insertions and avoid deletions. The matching algorithm for small
+ *      strings is inspired from that of Rabin & Karp. A brute force approach
+ *      is used to find longer strings when a small match has been found.
+ *      A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
+ *      (by Leonid Broukhis).
+ *         A previous version of this file used a more sophisticated algorithm
+ *      (by Fiala and Greene) which is guaranteed to run in linear amortized
+ *      time, but has a larger average cost, uses more memory and is patented.
+ *      However the F&G algorithm may be faster for some highly redundant
+ *      files if the parameter max_chain_length (described below) is too large.
+ *
+ *  ACKNOWLEDGEMENTS
+ *
+ *      The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
+ *      I found it in 'freeze' written by Leonid Broukhis.
+ *      Thanks to many people for bug reports and testing.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
+ *      Available in http://tools.ietf.org/html/rfc1951
+ *
+ *      A description of the Rabin and Karp algorithm is given in the book
+ *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
+ *
+ *      Fiala,E.R., and Greene,D.H.
+ *         Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
+ *
+ */
+
+/* @(#) $Id$ */
+
+#include "deflate.h"
+
+const char deflate_copyright[] =
+   " deflate 1.2.11 Copyright 1995-2017 Jean-loup Gailly and Mark Adler ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+
+/* ===========================================================================
+ *  Function prototypes.
+ */
+typedef enum {
+    need_more,      /* block not completed, need more input or more output */
+    block_done,     /* block flush performed */
+    finish_started, /* finish started, need only more output at next deflate */
+    finish_done     /* finish done, accept no more input or output */
+} block_state;
+
+typedef block_state (*compress_func) OF((deflate_state *s, int flush));
+/* Compression function. Returns the block state after the call. */
+
+local int deflateStateCheck      OF((z_streamp strm));
+local void slide_hash     OF((deflate_state *s));
+local void fill_window    OF((deflate_state *s));
+local block_state deflate_stored OF((deflate_state *s, int flush));
+local block_state deflate_fast   OF((deflate_state *s, int flush));
+#ifndef FASTEST
+local block_state deflate_slow   OF((deflate_state *s, int flush));
+#endif
+local block_state deflate_rle    OF((deflate_state *s, int flush));
+local block_state deflate_huff   OF((deflate_state *s, int flush));
+local void lm_init        OF((deflate_state *s));
+local void putShortMSB    OF((deflate_state *s, uInt b));
+local void flush_pending  OF((z_streamp strm));
+local unsigned read_buf   OF((z_streamp strm, Bytef *buf, unsigned size));
+#ifdef ASMV
+#  pragma message("Assembler code may have bugs -- use at your own risk")
+      void match_init OF((void)); /* asm code initialization */
+      uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#else
+local uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#endif
+
+#ifdef ZLIB_DEBUG
+local  void check_match OF((deflate_state *s, IPos start, IPos match,
+                            int length));
+#endif
+
+/* ===========================================================================
+ * Local data
+ */
+
+#define NIL 0
+/* Tail of hash chains */
+
+#ifndef TOO_FAR
+#  define TOO_FAR 4096
+#endif
+/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
+
+/* Values for max_lazy_match, good_match and max_chain_length, depending on
+ * the desired pack level (0..9). The values given below have been tuned to
+ * exclude worst case performance for pathological files. Better values may be
+ * found for specific files.
+ */
+typedef struct config_s {
+   ush good_length; /* reduce lazy search above this match length */
+   ush max_lazy;    /* do not perform lazy search above this match length */
+   ush nice_length; /* quit search above this match length */
+   ush max_chain;
+   compress_func func;
+} config;
+
+#ifdef FASTEST
+local const config configuration_table[2] = {
+/*      good lazy nice chain */
+/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
+/* 1 */ {4,    4,  8,    4, deflate_fast}}; /* max speed, no lazy matches */
+#else
+local const config configuration_table[10] = {
+/*      good lazy nice chain */
+/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
+/* 1 */ {4,    4,  8,    4, deflate_fast}, /* max speed, no lazy matches */
+/* 2 */ {4,    5, 16,    8, deflate_fast},
+/* 3 */ {4,    6, 32,   32, deflate_fast},
+
+/* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
+/* 5 */ {8,   16, 32,   32, deflate_slow},
+/* 6 */ {8,   16, 128, 128, deflate_slow},
+/* 7 */ {8,   32, 128, 256, deflate_slow},
+/* 8 */ {32, 128, 258, 1024, deflate_slow},
+/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
+#endif
+
+/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
+ * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
+ * meaning.
+ */
+
+/* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */
+#define RANK(f) (((f) * 2) - ((f) > 4 ? 9 : 0))
+
+/* ===========================================================================
+ * Update a hash value with the given input byte
+ * IN  assertion: all calls to UPDATE_HASH are made with consecutive input
+ *    characters, so that a running hash key can be computed from the previous
+ *    key instead of complete recalculation each time.
+ */
+#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
+
+
+/* ===========================================================================
+ * Insert string str in the dictionary and set match_head to the previous head
+ * of the hash chain (the most recent string with same hash key). Return
+ * the previous length of the hash chain.
+ * If this file is compiled with -DFASTEST, the compression level is forced
+ * to 1, and no hash chains are maintained.
+ * IN  assertion: all calls to INSERT_STRING are made with consecutive input
+ *    characters and the first MIN_MATCH bytes of str are valid (except for
+ *    the last MIN_MATCH-1 bytes of the input file).
+ */
+#ifdef FASTEST
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    match_head = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#else
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#endif
+
+/* ===========================================================================
+ * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
+ * prev[] will be initialized on the fly.
+ */
+#define CLEAR_HASH(s) \
+    s->head[s->hash_size-1] = NIL; \
+    zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
+
+/* ===========================================================================
+ * Slide the hash table when sliding the window down (could be avoided with 32
+ * bit values at the expense of memory usage). We slide even when level == 0 to
+ * keep the hash table consistent if we switch back to level > 0 later.
+ */
+local void slide_hash(s)
+    deflate_state *s;
+{
+    unsigned n, m;
+    Posf *p;
+    uInt wsize = s->w_size;
+
+    n = s->hash_size;
+    p = &s->head[n];
+    do {
+        m = *--p;
+        *p = (Pos)(m >= wsize ? m - wsize : NIL);
+    } while (--n);
+    n = wsize;
+#ifndef FASTEST
+    p = &s->prev[n];
+    do {
+        m = *--p;
+        *p = (Pos)(m >= wsize ? m - wsize : NIL);
+        /* If n is not on any hash chain, prev[n] is garbage but
+         * its value will never be used.
+         */
+    } while (--n);
+#endif
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateInit_(strm, level, version, stream_size)
+    z_streamp strm;
+    int level;
+    const char *version;
+    int stream_size;
+{
+    return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
+                         Z_DEFAULT_STRATEGY, version, stream_size);
+    /* To do: ignore strm->next_in if we use it as window */
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
+                  version, stream_size)
+    z_streamp strm;
+    int  level;
+    int  method;
+    int  windowBits;
+    int  memLevel;
+    int  strategy;
+    const char *version;
+    int stream_size;
+{
+    deflate_state *s;
+    int wrap = 1;
+    static const char my_version[] = ZLIB_VERSION;
+
+    ushf *overlay;
+    /* We overlay pending_buf and d_buf+l_buf. This works since the average
+     * output size for (length,distance) codes is <= 24 bits.
+     */
+
+    if (version == Z_NULL || version[0] != my_version[0] ||
+        stream_size != sizeof(z_stream)) {
+        return Z_VERSION_ERROR;
+    }
+    if (strm == Z_NULL) return Z_STREAM_ERROR;
+
+    strm->msg = Z_NULL;
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zfree = zcfree;
+#endif
+
+#ifdef FASTEST
+    if (level != 0) level = 1;
+#else
+    if (level == Z_DEFAULT_COMPRESSION) level = 6;
+#endif
+
+    if (windowBits < 0) { /* suppress zlib wrapper */
+        wrap = 0;
+        windowBits = -windowBits;
+    }
+#ifdef GZIP
+    else if (windowBits > 15) {
+        wrap = 2;       /* write gzip wrapper instead */
+        windowBits -= 16;
+    }
+#endif
+    if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
+        windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
+        strategy < 0 || strategy > Z_FIXED || (windowBits == 8 && wrap != 1)) {
+        return Z_STREAM_ERROR;
+    }
+    if (windowBits == 8) windowBits = 9;  /* until 256-byte window bug fixed */
+    s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
+    if (s == Z_NULL) return Z_MEM_ERROR;
+    strm->state = (struct internal_state FAR *)s;
+    s->strm = strm;
+    s->status = INIT_STATE;     /* to pass state test in deflateReset() */
+
+    s->wrap = wrap;
+    s->gzhead = Z_NULL;
+    s->w_bits = (uInt)windowBits;
+    s->w_size = 1 << s->w_bits;
+    s->w_mask = s->w_size - 1;
+
+    s->hash_bits = (uInt)memLevel + 7;
+    s->hash_size = 1 << s->hash_bits;
+    s->hash_mask = s->hash_size - 1;
+    s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
+
+    s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
+    s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
+    s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
+
+    s->high_water = 0;      /* nothing written to s->window yet */
+
+    s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
+
+    overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
+    s->pending_buf = (uchf *) overlay;
+    s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
+
+    if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
+        s->pending_buf == Z_NULL) {
+        s->status = FINISH_STATE;
+        strm->msg = ERR_MSG(Z_MEM_ERROR);
+        deflateEnd (strm);
+        return Z_MEM_ERROR;
+    }
+    s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
+    s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
+
+    s->level = level;
+    s->strategy = strategy;
+    s->method = (Byte)method;
+
+    return deflateReset(strm);
+}
+
+/* =========================================================================
+ * Check for a valid deflate stream state. Return 0 if ok, 1 if not.
+ */
+local int deflateStateCheck (strm)
+    z_streamp strm;
+{
+    deflate_state *s;
+    if (strm == Z_NULL ||
+        strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0)
+        return 1;
+    s = strm->state;
+    if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE &&
+#ifdef GZIP
+                                           s->status != GZIP_STATE &&
+#endif
+                                           s->status != EXTRA_STATE &&
+                                           s->status != NAME_STATE &&
+                                           s->status != COMMENT_STATE &&
+                                           s->status != HCRC_STATE &&
+                                           s->status != BUSY_STATE &&
+                                           s->status != FINISH_STATE))
+        return 1;
+    return 0;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
+    z_streamp strm;
+    const Bytef *dictionary;
+    uInt  dictLength;
+{
+    deflate_state *s;
+    uInt str, n;
+    int wrap;
+    unsigned avail;
+    z_const unsigned char *next;
+
+    if (deflateStateCheck(strm) || dictionary == Z_NULL)
+        return Z_STREAM_ERROR;
+    s = strm->state;
+    wrap = s->wrap;
+    if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead)
+        return Z_STREAM_ERROR;
+
+    /* when using zlib wrappers, compute Adler-32 for provided dictionary */
+    if (wrap == 1)
+        strm->adler = adler32(strm->adler, dictionary, dictLength);
+    s->wrap = 0;                    /* avoid computing Adler-32 in read_buf */
+
+    /* if dictionary would fill window, just replace the history */
+    if (dictLength >= s->w_size) {
+        if (wrap == 0) {            /* already empty otherwise */
+            CLEAR_HASH(s);
+            s->strstart = 0;
+            s->block_start = 0L;
+            s->insert = 0;
+        }
+        dictionary += dictLength - s->w_size;  /* use the tail */
+        dictLength = s->w_size;
+    }
+
+    /* insert dictionary into window and hash */
+    avail = strm->avail_in;
+    next = strm->next_in;
+    strm->avail_in = dictLength;
+    strm->next_in = (z_const Bytef *)dictionary;
+    fill_window(s);
+    while (s->lookahead >= MIN_MATCH) {
+        str = s->strstart;
+        n = s->lookahead - (MIN_MATCH-1);
+        do {
+            UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+            s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+            s->head[s->ins_h] = (Pos)str;
+            str++;
+        } while (--n);
+        s->strstart = str;
+        s->lookahead = MIN_MATCH-1;
+        fill_window(s);
+    }
+    s->strstart += s->lookahead;
+    s->block_start = (long)s->strstart;
+    s->insert = s->lookahead;
+    s->lookahead = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    strm->next_in = next;
+    strm->avail_in = avail;
+    s->wrap = wrap;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateGetDictionary (strm, dictionary, dictLength)
+    z_streamp strm;
+    Bytef *dictionary;
+    uInt  *dictLength;
+{
+    deflate_state *s;
+    uInt len;
+
+    if (deflateStateCheck(strm))
+        return Z_STREAM_ERROR;
+    s = strm->state;
+    len = s->strstart + s->lookahead;
+    if (len > s->w_size)
+        len = s->w_size;
+    if (dictionary != Z_NULL && len)
+        zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len);
+    if (dictLength != Z_NULL)
+        *dictLength = len;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateResetKeep (strm)
+    z_streamp strm;
+{
+    deflate_state *s;
+
+    if (deflateStateCheck(strm)) {
+        return Z_STREAM_ERROR;
+    }
+
+    strm->total_in = strm->total_out = 0;
+    strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
+    strm->data_type = Z_UNKNOWN;
+
+    s = (deflate_state *)strm->state;
+    s->pending = 0;
+    s->pending_out = s->pending_buf;
+
+    if (s->wrap < 0) {
+        s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
+    }
+    s->status =
+#ifdef GZIP
+        s->wrap == 2 ? GZIP_STATE :
+#endif
+        s->wrap ? INIT_STATE : BUSY_STATE;
+    strm->adler =
+#ifdef GZIP
+        s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
+#endif
+        adler32(0L, Z_NULL, 0);
+    s->last_flush = Z_NO_FLUSH;
+
+    _tr_init(s);
+
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateReset (strm)
+    z_streamp strm;
+{
+    int ret;
+
+    ret = deflateResetKeep(strm);
+    if (ret == Z_OK)
+        lm_init(strm->state);
+    return ret;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateSetHeader (strm, head)
+    z_streamp strm;
+    gz_headerp head;
+{
+    if (deflateStateCheck(strm) || strm->state->wrap != 2)
+        return Z_STREAM_ERROR;
+    strm->state->gzhead = head;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflatePending (strm, pending, bits)
+    unsigned *pending;
+    int *bits;
+    z_streamp strm;
+{
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    if (pending != Z_NULL)
+        *pending = strm->state->pending;
+    if (bits != Z_NULL)
+        *bits = strm->state->bi_valid;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflatePrime (strm, bits, value)
+    z_streamp strm;
+    int bits;
+    int value;
+{
+    deflate_state *s;
+    int put;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+    if ((Bytef *)(s->d_buf) < s->pending_out + ((Buf_size + 7) >> 3))
+        return Z_BUF_ERROR;
+    do {
+        put = Buf_size - s->bi_valid;
+        if (put > bits)
+            put = bits;
+        s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid);
+        s->bi_valid += put;
+        _tr_flush_bits(s);
+        value >>= put;
+        bits -= put;
+    } while (bits);
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateParams(strm, level, strategy)
+    z_streamp strm;
+    int level;
+    int strategy;
+{
+    deflate_state *s;
+    compress_func func;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+
+#ifdef FASTEST
+    if (level != 0) level = 1;
+#else
+    if (level == Z_DEFAULT_COMPRESSION) level = 6;
+#endif
+    if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
+        return Z_STREAM_ERROR;
+    }
+    func = configuration_table[s->level].func;
+
+    if ((strategy != s->strategy || func != configuration_table[level].func) &&
+        s->high_water) {
+        /* Flush the last buffer: */
+        int err = deflate(strm, Z_BLOCK);
+        if (err == Z_STREAM_ERROR)
+            return err;
+        if (strm->avail_out == 0)
+            return Z_BUF_ERROR;
+    }
+    if (s->level != level) {
+        if (s->level == 0 && s->matches != 0) {
+            if (s->matches == 1)
+                slide_hash(s);
+            else
+                CLEAR_HASH(s);
+            s->matches = 0;
+        }
+        s->level = level;
+        s->max_lazy_match   = configuration_table[level].max_lazy;
+        s->good_match       = configuration_table[level].good_length;
+        s->nice_match       = configuration_table[level].nice_length;
+        s->max_chain_length = configuration_table[level].max_chain;
+    }
+    s->strategy = strategy;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
+    z_streamp strm;
+    int good_length;
+    int max_lazy;
+    int nice_length;
+    int max_chain;
+{
+    deflate_state *s;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+    s->good_match = (uInt)good_length;
+    s->max_lazy_match = (uInt)max_lazy;
+    s->nice_match = nice_length;
+    s->max_chain_length = (uInt)max_chain;
+    return Z_OK;
+}
+
+/* =========================================================================
+ * For the default windowBits of 15 and memLevel of 8, this function returns
+ * a close to exact, as well as small, upper bound on the compressed size.
+ * They are coded as constants here for a reason--if the #define's are
+ * changed, then this function needs to be changed as well.  The return
+ * value for 15 and 8 only works for those exact settings.
+ *
+ * For any setting other than those defaults for windowBits and memLevel,
+ * the value returned is a conservative worst case for the maximum expansion
+ * resulting from using fixed blocks instead of stored blocks, which deflate
+ * can emit on compressed data for some combinations of the parameters.
+ *
+ * This function could be more sophisticated to provide closer upper bounds for
+ * every combination of windowBits and memLevel.  But even the conservative
+ * upper bound of about 14% expansion does not seem onerous for output buffer
+ * allocation.
+ */
+uLong ZEXPORT deflateBound(strm, sourceLen)
+    z_streamp strm;
+    uLong sourceLen;
+{
+    deflate_state *s;
+    uLong complen, wraplen;
+
+    /* conservative upper bound for compressed data */
+    complen = sourceLen +
+              ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5;
+
+    /* if can't get parameters, return conservative bound plus zlib wrapper */
+    if (deflateStateCheck(strm))
+        return complen + 6;
+
+    /* compute wrapper length */
+    s = strm->state;
+    switch (s->wrap) {
+    case 0:                                 /* raw deflate */
+        wraplen = 0;
+        break;
+    case 1:                                 /* zlib wrapper */
+        wraplen = 6 + (s->strstart ? 4 : 0);
+        break;
+#ifdef GZIP
+    case 2:                                 /* gzip wrapper */
+        wraplen = 18;
+        if (s->gzhead != Z_NULL) {          /* user-supplied gzip header */
+            Bytef *str;
+            if (s->gzhead->extra != Z_NULL)
+                wraplen += 2 + s->gzhead->extra_len;
+            str = s->gzhead->name;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            str = s->gzhead->comment;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            if (s->gzhead->hcrc)
+                wraplen += 2;
+        }
+        break;
+#endif
+    default:                                /* for compiler happiness */
+        wraplen = 6;
+    }
+
+    /* if not default parameters, return conservative bound */
+    if (s->w_bits != 15 || s->hash_bits != 8 + 7)
+        return complen + wraplen;
+
+    /* default settings: return tight bound for that case */
+    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
+           (sourceLen >> 25) + 13 - 6 + wraplen;
+}
+
+/* =========================================================================
+ * Put a short in the pending buffer. The 16-bit value is put in MSB order.
+ * IN assertion: the stream state is correct and there is enough room in
+ * pending_buf.
+ */
+local void putShortMSB (s, b)
+    deflate_state *s;
+    uInt b;
+{
+    put_byte(s, (Byte)(b >> 8));
+    put_byte(s, (Byte)(b & 0xff));
+}
+
+/* =========================================================================
+ * Flush as much pending output as possible. All deflate() output, except for
+ * some deflate_stored() output, goes through this function so some
+ * applications may wish to modify it to avoid allocating a large
+ * strm->next_out buffer and copying into it. (See also read_buf()).
+ */
+local void flush_pending(strm)
+    z_streamp strm;
+{
+    unsigned len;
+    deflate_state *s = strm->state;
+
+    _tr_flush_bits(s);
+    len = s->pending;
+    if (len > strm->avail_out) len = strm->avail_out;
+    if (len == 0) return;
+
+    zmemcpy(strm->next_out, s->pending_out, len);
+    strm->next_out  += len;
+    s->pending_out  += len;
+    strm->total_out += len;
+    strm->avail_out -= len;
+    s->pending      -= len;
+    if (s->pending == 0) {
+        s->pending_out = s->pending_buf;
+    }
+}
+
+/* ===========================================================================
+ * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1].
+ */
+#define HCRC_UPDATE(beg) \
+    do { \
+        if (s->gzhead->hcrc && s->pending > (beg)) \
+            strm->adler = crc32(strm->adler, s->pending_buf + (beg), \
+                                s->pending - (beg)); \
+    } while (0)
+
+/* ========================================================================= */
+int ZEXPORT deflate (strm, flush)
+    z_streamp strm;
+    int flush;
+{
+    int old_flush; /* value of flush param for previous deflate call */
+    deflate_state *s;
+
+    if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) {
+        return Z_STREAM_ERROR;
+    }
+    s = strm->state;
+
+    if (strm->next_out == Z_NULL ||
+        (strm->avail_in != 0 && strm->next_in == Z_NULL) ||
+        (s->status == FINISH_STATE && flush != Z_FINISH)) {
+        ERR_RETURN(strm, Z_STREAM_ERROR);
+    }
+    if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
+
+    old_flush = s->last_flush;
+    s->last_flush = flush;
+
+    /* Flush as much pending output as possible */
+    if (s->pending != 0) {
+        flush_pending(strm);
+        if (strm->avail_out == 0) {
+            /* Since avail_out is 0, deflate will be called again with
+             * more output space, but possibly with both pending and
+             * avail_in equal to zero. There won't be anything to do,
+             * but this is not an error situation so make sure we
+             * return OK instead of BUF_ERROR at next call of deflate:
+             */
+            s->last_flush = -1;
+            return Z_OK;
+        }
+
+    /* Make sure there is something to do and avoid duplicate consecutive
+     * flushes. For repeated and useless calls with Z_FINISH, we keep
+     * returning Z_STREAM_END instead of Z_BUF_ERROR.
+     */
+    } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) &&
+               flush != Z_FINISH) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* User must not provide more input after the first FINISH: */
+    if (s->status == FINISH_STATE && strm->avail_in != 0) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* Write the header */
+    if (s->status == INIT_STATE) {
+        /* zlib header */
+        uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
+        uInt level_flags;
+
+        if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
+            level_flags = 0;
+        else if (s->level < 6)
+            level_flags = 1;
+        else if (s->level == 6)
+            level_flags = 2;
+        else
+            level_flags = 3;
+        header |= (level_flags << 6);
+        if (s->strstart != 0) header |= PRESET_DICT;
+        header += 31 - (header % 31);
+
+        putShortMSB(s, header);
+
+        /* Save the adler32 of the preset dictionary: */
+        if (s->strstart != 0) {
+            putShortMSB(s, (uInt)(strm->adler >> 16));
+            putShortMSB(s, (uInt)(strm->adler & 0xffff));
+        }
+        strm->adler = adler32(0L, Z_NULL, 0);
+        s->status = BUSY_STATE;
+
+        /* Compression must start with an empty pending buffer */
+        flush_pending(strm);
+        if (s->pending != 0) {
+            s->last_flush = -1;
+            return Z_OK;
+        }
+    }
+#ifdef GZIP
+    if (s->status == GZIP_STATE) {
+        /* gzip header */
+        strm->adler = crc32(0L, Z_NULL, 0);
+        put_byte(s, 31);
+        put_byte(s, 139);
+        put_byte(s, 8);
+        if (s->gzhead == Z_NULL) {
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, s->level == 9 ? 2 :
+                     (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
+                      4 : 0));
+            put_byte(s, OS_CODE);
+            s->status = BUSY_STATE;
+
+            /* Compression must start with an empty pending buffer */
+            flush_pending(strm);
+            if (s->pending != 0) {
+                s->last_flush = -1;
+                return Z_OK;
+            }
+        }
+        else {
+            put_byte(s, (s->gzhead->text ? 1 : 0) +
+                     (s->gzhead->hcrc ? 2 : 0) +
+                     (s->gzhead->extra == Z_NULL ? 0 : 4) +
+                     (s->gzhead->name == Z_NULL ? 0 : 8) +
+                     (s->gzhead->comment == Z_NULL ? 0 : 16)
+                     );
+            put_byte(s, (Byte)(s->gzhead->time & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
+            put_byte(s, s->level == 9 ? 2 :
+                     (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
+                      4 : 0));
+            put_byte(s, s->gzhead->os & 0xff);
+            if (s->gzhead->extra != Z_NULL) {
+                put_byte(s, s->gzhead->extra_len & 0xff);
+                put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
+            }
+            if (s->gzhead->hcrc)
+                strm->adler = crc32(strm->adler, s->pending_buf,
+                                    s->pending);
+            s->gzindex = 0;
+            s->status = EXTRA_STATE;
+        }
+    }
+    if (s->status == EXTRA_STATE) {
+        if (s->gzhead->extra != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex;
+            while (s->pending + left > s->pending_buf_size) {
+                uInt copy = s->pending_buf_size - s->pending;
+                zmemcpy(s->pending_buf + s->pending,
+                        s->gzhead->extra + s->gzindex, copy);
+                s->pending = s->pending_buf_size;
+                HCRC_UPDATE(beg);
+                s->gzindex += copy;
+                flush_pending(strm);
+                if (s->pending != 0) {
+                    s->last_flush = -1;
+                    return Z_OK;
+                }
+                beg = 0;
+                left -= copy;
+            }
+            zmemcpy(s->pending_buf + s->pending,
+                    s->gzhead->extra + s->gzindex, left);
+            s->pending += left;
+            HCRC_UPDATE(beg);
+            s->gzindex = 0;
+        }
+        s->status = NAME_STATE;
+    }
+    if (s->status == NAME_STATE) {
+        if (s->gzhead->name != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            int val;
+            do {
+                if (s->pending == s->pending_buf_size) {
+                    HCRC_UPDATE(beg);
+                    flush_pending(strm);
+                    if (s->pending != 0) {
+                        s->last_flush = -1;
+                        return Z_OK;
+                    }
+                    beg = 0;
+                }
+                val = s->gzhead->name[s->gzindex++];
+                put_byte(s, val);
+            } while (val != 0);
+            HCRC_UPDATE(beg);
+            s->gzindex = 0;
+        }
+        s->status = COMMENT_STATE;
+    }
+    if (s->status == COMMENT_STATE) {
+        if (s->gzhead->comment != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            int val;
+            do {
+                if (s->pending == s->pending_buf_size) {
+                    HCRC_UPDATE(beg);
+                    flush_pending(strm);
+                    if (s->pending != 0) {
+                        s->last_flush = -1;
+                        return Z_OK;
+                    }
+                    beg = 0;
+                }
+                val = s->gzhead->comment[s->gzindex++];
+                put_byte(s, val);
+            } while (val != 0);
+            HCRC_UPDATE(beg);
+        }
+        s->status = HCRC_STATE;
+    }
+    if (s->status == HCRC_STATE) {
+        if (s->gzhead->hcrc) {
+            if (s->pending + 2 > s->pending_buf_size) {
+                flush_pending(strm);
+                if (s->pending != 0) {
+                    s->last_flush = -1;
+                    return Z_OK;
+                }
+            }
+            put_byte(s, (Byte)(strm->adler & 0xff));
+            put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
+            strm->adler = crc32(0L, Z_NULL, 0);
+        }
+        s->status = BUSY_STATE;
+
+        /* Compression must start with an empty pending buffer */
+        flush_pending(strm);
+        if (s->pending != 0) {
+            s->last_flush = -1;
+            return Z_OK;
+        }
+    }
+#endif
+
+    /* Start a new block or continue the current one.
+     */
+    if (strm->avail_in != 0 || s->lookahead != 0 ||
+        (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
+        block_state bstate;
+
+        bstate = s->level == 0 ? deflate_stored(s, flush) :
+                 s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) :
+                 s->strategy == Z_RLE ? deflate_rle(s, flush) :
+                 (*(configuration_table[s->level].func))(s, flush);
+
+        if (bstate == finish_started || bstate == finish_done) {
+            s->status = FINISH_STATE;
+        }
+        if (bstate == need_more || bstate == finish_started) {
+            if (strm->avail_out == 0) {
+                s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
+            }
+            return Z_OK;
+            /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
+             * of deflate should use the same flush parameter to make sure
+             * that the flush is complete. So we don't have to output an
+             * empty block here, this will be done at next call. This also
+             * ensures that for a very small output buffer, we emit at most
+             * one empty block.
+             */
+        }
+        if (bstate == block_done) {
+            if (flush == Z_PARTIAL_FLUSH) {
+                _tr_align(s);
+            } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */
+                _tr_stored_block(s, (char*)0, 0L, 0);
+                /* For a full flush, this empty block will be recognized
+                 * as a special marker by inflate_sync().
+                 */
+                if (flush == Z_FULL_FLUSH) {
+                    CLEAR_HASH(s);             /* forget history */
+                    if (s->lookahead == 0) {
+                        s->strstart = 0;
+                        s->block_start = 0L;
+                        s->insert = 0;
+                    }
+                }
+            }
+            flush_pending(strm);
+            if (strm->avail_out == 0) {
+              s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
+              return Z_OK;
+            }
+        }
+    }
+
+    if (flush != Z_FINISH) return Z_OK;
+    if (s->wrap <= 0) return Z_STREAM_END;
+
+    /* Write the trailer */
+#ifdef GZIP
+    if (s->wrap == 2) {
+        put_byte(s, (Byte)(strm->adler & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
+        put_byte(s, (Byte)(strm->total_in & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
+    }
+    else
+#endif
+    {
+        putShortMSB(s, (uInt)(strm->adler >> 16));
+        putShortMSB(s, (uInt)(strm->adler & 0xffff));
+    }
+    flush_pending(strm);
+    /* If avail_out is zero, the application will call deflate again
+     * to flush the rest.
+     */
+    if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
+    return s->pending != 0 ? Z_OK : Z_STREAM_END;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateEnd (strm)
+    z_streamp strm;
+{
+    int status;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+
+    status = strm->state->status;
+
+    /* Deallocate in reverse order of allocations: */
+    TRY_FREE(strm, strm->state->pending_buf);
+    TRY_FREE(strm, strm->state->head);
+    TRY_FREE(strm, strm->state->prev);
+    TRY_FREE(strm, strm->state->window);
+
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+
+    return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
+}
+
+/* =========================================================================
+ * Copy the source state to the destination state.
+ * To simplify the source, this is not supported for 16-bit MSDOS (which
+ * doesn't have enough memory anyway to duplicate compression states).
+ */
+int ZEXPORT deflateCopy (dest, source)
+    z_streamp dest;
+    z_streamp source;
+{
+#ifdef MAXSEG_64K
+    return Z_STREAM_ERROR;
+#else
+    deflate_state *ds;
+    deflate_state *ss;
+    ushf *overlay;
+
+
+    if (deflateStateCheck(source) || dest == Z_NULL) {
+        return Z_STREAM_ERROR;
+    }
+
+    ss = source->state;
+
+    zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
+
+    ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
+    if (ds == Z_NULL) return Z_MEM_ERROR;
+    dest->state = (struct internal_state FAR *) ds;
+    zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state));
+    ds->strm = dest;
+
+    ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
+    ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
+    ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
+    overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
+    ds->pending_buf = (uchf *) overlay;
+
+    if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
+        ds->pending_buf == Z_NULL) {
+        deflateEnd (dest);
+        return Z_MEM_ERROR;
+    }
+    /* following zmemcpy do not work for 16-bit MSDOS */
+    zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
+    zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos));
+    zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos));
+    zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
+
+    ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
+    ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
+    ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
+
+    ds->l_desc.dyn_tree = ds->dyn_ltree;
+    ds->d_desc.dyn_tree = ds->dyn_dtree;
+    ds->bl_desc.dyn_tree = ds->bl_tree;
+
+    return Z_OK;
+#endif /* MAXSEG_64K */
+}
+
+/* ===========================================================================
+ * Read a new buffer from the current input stream, update the adler32
+ * and total number of bytes read.  All deflate() input goes through
+ * this function so some applications may wish to modify it to avoid
+ * allocating a large strm->next_in buffer and copying from it.
+ * (See also flush_pending()).
+ */
+local unsigned read_buf(strm, buf, size)
+    z_streamp strm;
+    Bytef *buf;
+    unsigned size;
+{
+    unsigned len = strm->avail_in;
+
+    if (len > size) len = size;
+    if (len == 0) return 0;
+
+    strm->avail_in  -= len;
+
+    zmemcpy(buf, strm->next_in, len);
+    if (strm->state->wrap == 1) {
+        strm->adler = adler32(strm->adler, buf, len);
+    }
+#ifdef GZIP
+    else if (strm->state->wrap == 2) {
+        strm->adler = crc32(strm->adler, buf, len);
+    }
+#endif
+    strm->next_in  += len;
+    strm->total_in += len;
+
+    return len;
+}
+
+/* ===========================================================================
+ * Initialize the "longest match" routines for a new zlib stream
+ */
+local void lm_init (s)
+    deflate_state *s;
+{
+    s->window_size = (ulg)2L*s->w_size;
+
+    CLEAR_HASH(s);
+
+    /* Set the default configuration parameters:
+     */
+    s->max_lazy_match   = configuration_table[s->level].max_lazy;
+    s->good_match       = configuration_table[s->level].good_length;
+    s->nice_match       = configuration_table[s->level].nice_length;
+    s->max_chain_length = configuration_table[s->level].max_chain;
+
+    s->strstart = 0;
+    s->block_start = 0L;
+    s->lookahead = 0;
+    s->insert = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    s->ins_h = 0;
+#ifndef FASTEST
+#ifdef ASMV
+    match_init(); /* initialize the asm code */
+#endif
+#endif
+}
+
+#ifndef FASTEST
+/* ===========================================================================
+ * Set match_start to the longest match starting at the given string and
+ * return its length. Matches shorter or equal to prev_length are discarded,
+ * in which case the result is equal to prev_length and match_start is
+ * garbage.
+ * IN assertions: cur_match is the head of the hash chain for the current
+ *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
+ * OUT assertion: the match length is not greater than s->lookahead.
+ */
+#ifndef ASMV
+/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
+ * match.S. The code will be functionally equivalent.
+ */
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    unsigned chain_length = s->max_chain_length;/* max hash chain length */
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                      /* matched string */
+    register int len;                           /* length of current match */
+    int best_len = (int)s->prev_length;         /* best match length so far */
+    int nice_match = s->nice_match;             /* stop if match long enough */
+    IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
+        s->strstart - (IPos)MAX_DIST(s) : NIL;
+    /* Stop when cur_match becomes <= limit. To simplify the code,
+     * we prevent matches with the string of window index 0.
+     */
+    Posf *prev = s->prev;
+    uInt wmask = s->w_mask;
+
+#ifdef UNALIGNED_OK
+    /* Compare two bytes at a time. Note: this is not always beneficial.
+     * Try with and without -DUNALIGNED_OK to check.
+     */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
+    register ush scan_start = *(ushf*)scan;
+    register ush scan_end   = *(ushf*)(scan+best_len-1);
+#else
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+    register Byte scan_end1  = scan[best_len-1];
+    register Byte scan_end   = scan[best_len];
+#endif
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    /* Do not waste too much time if we already have a good match: */
+    if (s->prev_length >= s->good_match) {
+        chain_length >>= 2;
+    }
+    /* Do not look for matches beyond the end of the input. This is necessary
+     * to make deflate deterministic.
+     */
+    if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead;
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    do {
+        Assert(cur_match < s->strstart, "no future");
+        match = s->window + cur_match;
+
+        /* Skip to next match if the match length cannot increase
+         * or if the match length is less than 2.  Note that the checks below
+         * for insufficient lookahead only occur occasionally for performance
+         * reasons.  Therefore uninitialized memory will be accessed, and
+         * conditional jumps will be made that depend on those values.
+         * However the length of the match is limited to the lookahead, so
+         * the output of deflate is not affected by the uninitialized values.
+         */
+#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
+        /* This code assumes sizeof(unsigned short) == 2. Do not use
+         * UNALIGNED_OK if your compiler uses a different size.
+         */
+        if (*(ushf*)(match+best_len-1) != scan_end ||
+            *(ushf*)match != scan_start) continue;
+
+        /* It is not necessary to compare scan[2] and match[2] since they are
+         * always equal when the other bytes match, given that the hash keys
+         * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
+         * strstart+3, +5, ... up to strstart+257. We check for insufficient
+         * lookahead only every 4th comparison; the 128th check will be made
+         * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
+         * necessary to put more guard bytes at the end of the window, or
+         * to check more often for insufficient lookahead.
+         */
+        Assert(scan[2] == match[2], "scan[2]?");
+        scan++, match++;
+        do {
+        } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 scan < strend);
+        /* The funny "do {}" generates better code on most compilers */
+
+        /* Here, scan <= window+strstart+257 */
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+        if (*scan == *match) scan++;
+
+        len = (MAX_MATCH - 1) - (int)(strend-scan);
+        scan = strend - (MAX_MATCH-1);
+
+#else /* UNALIGNED_OK */
+
+        if (match[best_len]   != scan_end  ||
+            match[best_len-1] != scan_end1 ||
+            *match            != *scan     ||
+            *++match          != scan[1])      continue;
+
+        /* The check at best_len-1 can be removed because it will be made
+         * again later. (This heuristic is not always a win.)
+         * It is not necessary to compare scan[2] and match[2] since they
+         * are always equal when the other bytes match, given that
+         * the hash keys are equal and that HASH_BITS >= 8.
+         */
+        scan += 2, match++;
+        Assert(*scan == *match, "match[2]?");
+
+        /* We check for insufficient lookahead only every 8th comparison;
+         * the 256th check will be made at strstart+258.
+         */
+        do {
+        } while (*++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 scan < strend);
+
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+        len = MAX_MATCH - (int)(strend - scan);
+        scan = strend - MAX_MATCH;
+
+#endif /* UNALIGNED_OK */
+
+        if (len > best_len) {
+            s->match_start = cur_match;
+            best_len = len;
+            if (len >= nice_match) break;
+#ifdef UNALIGNED_OK
+            scan_end = *(ushf*)(scan+best_len-1);
+#else
+            scan_end1  = scan[best_len-1];
+            scan_end   = scan[best_len];
+#endif
+        }
+    } while ((cur_match = prev[cur_match & wmask]) > limit
+             && --chain_length != 0);
+
+    if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
+    return s->lookahead;
+}
+#endif /* ASMV */
+
+#else /* FASTEST */
+
+/* ---------------------------------------------------------------------------
+ * Optimized version for FASTEST only
+ */
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                       /* matched string */
+    register int len;                           /* length of current match */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    Assert(cur_match < s->strstart, "no future");
+
+    match = s->window + cur_match;
+
+    /* Return failure if the match length is less than 2:
+     */
+    if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
+
+    /* The check at best_len-1 can be removed because it will be made
+     * again later. (This heuristic is not always a win.)
+     * It is not necessary to compare scan[2] and match[2] since they
+     * are always equal when the other bytes match, given that
+     * the hash keys are equal and that HASH_BITS >= 8.
+     */
+    scan += 2, match += 2;
+    Assert(*scan == *match, "match[2]?");
+
+    /* We check for insufficient lookahead only every 8th comparison;
+     * the 256th check will be made at strstart+258.
+     */
+    do {
+    } while (*++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             scan < strend);
+
+    Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+    len = MAX_MATCH - (int)(strend - scan);
+
+    if (len < MIN_MATCH) return MIN_MATCH - 1;
+
+    s->match_start = cur_match;
+    return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
+}
+
+#endif /* FASTEST */
+
+#ifdef ZLIB_DEBUG
+
+#define EQUAL 0
+/* result of memcmp for equal strings */
+
+/* ===========================================================================
+ * Check that the match at match_start is indeed a match.
+ */
+local void check_match(s, start, match, length)
+    deflate_state *s;
+    IPos start, match;
+    int length;
+{
+    /* check that the match is indeed a match */
+    if (zmemcmp(s->window + match,
+                s->window + start, length) != EQUAL) {
+        fprintf(stderr, " start %u, match %u, length %d\n",
+                start, match, length);
+        do {
+            fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
+        } while (--length != 0);
+        z_error("invalid match");
+    }
+    if (z_verbose > 1) {
+        fprintf(stderr,"\\[%d,%d]", start-match, length);
+        do { putc(s->window[start++], stderr); } while (--length != 0);
+    }
+}
+#else
+#  define check_match(s, start, match, length)
+#endif /* ZLIB_DEBUG */
+
+/* ===========================================================================
+ * Fill the window when the lookahead becomes insufficient.
+ * Updates strstart and lookahead.
+ *
+ * IN assertion: lookahead < MIN_LOOKAHEAD
+ * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
+ *    At least one byte has been read, or avail_in == 0; reads are
+ *    performed for at least two bytes (required for the zip translate_eol
+ *    option -- not supported here).
+ */
+local void fill_window(s)
+    deflate_state *s;
+{
+    unsigned n;
+    unsigned more;    /* Amount of free space at the end of the window. */
+    uInt wsize = s->w_size;
+
+    Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead");
+
+    do {
+        more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
+
+        /* Deal with !@#$% 64K limit: */
+        if (sizeof(int) <= 2) {
+            if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
+                more = wsize;
+
+            } else if (more == (unsigned)(-1)) {
+                /* Very unlikely, but possible on 16 bit machine if
+                 * strstart == 0 && lookahead == 1 (input done a byte at time)
+                 */
+                more--;
+            }
+        }
+
+        /* If the window is almost full and there is insufficient lookahead,
+         * move the upper half to the lower one to make room in the upper half.
+         */
+        if (s->strstart >= wsize+MAX_DIST(s)) {
+
+            zmemcpy(s->window, s->window+wsize, (unsigned)wsize - more);
+            s->match_start -= wsize;
+            s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
+            s->block_start -= (long) wsize;
+            slide_hash(s);
+            more += wsize;
+        }
+        if (s->strm->avail_in == 0) break;
+
+        /* If there was no sliding:
+         *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
+         *    more == window_size - lookahead - strstart
+         * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
+         * => more >= window_size - 2*WSIZE + 2
+         * In the BIG_MEM or MMAP case (not yet supported),
+         *   window_size == input_size + MIN_LOOKAHEAD  &&
+         *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
+         * Otherwise, window_size == 2*WSIZE so more >= 2.
+         * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
+         */
+        Assert(more >= 2, "more < 2");
+
+        n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
+        s->lookahead += n;
+
+        /* Initialize the hash value now that we have some input: */
+        if (s->lookahead + s->insert >= MIN_MATCH) {
+            uInt str = s->strstart - s->insert;
+            s->ins_h = s->window[str];
+            UPDATE_HASH(s, s->ins_h, s->window[str + 1]);
+#if MIN_MATCH != 3
+            Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+            while (s->insert) {
+                UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+                s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+                s->head[s->ins_h] = (Pos)str;
+                str++;
+                s->insert--;
+                if (s->lookahead + s->insert < MIN_MATCH)
+                    break;
+            }
+        }
+        /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
+         * but this is not important since only literal bytes will be emitted.
+         */
+
+    } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
+
+    /* If the WIN_INIT bytes after the end of the current data have never been
+     * written, then zero those bytes in order to avoid memory check reports of
+     * the use of uninitialized (or uninitialised as Julian writes) bytes by
+     * the longest match routines.  Update the high water mark for the next
+     * time through here.  WIN_INIT is set to MAX_MATCH since the longest match
+     * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead.
+     */
+    if (s->high_water < s->window_size) {
+        ulg curr = s->strstart + (ulg)(s->lookahead);
+        ulg init;
+
+        if (s->high_water < curr) {
+            /* Previous high water mark below current data -- zero WIN_INIT
+             * bytes or up to end of window, whichever is less.
+             */
+            init = s->window_size - curr;
+            if (init > WIN_INIT)
+                init = WIN_INIT;
+            zmemzero(s->window + curr, (unsigned)init);
+            s->high_water = curr + init;
+        }
+        else if (s->high_water < (ulg)curr + WIN_INIT) {
+            /* High water mark at or above current data, but below current data
+             * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up
+             * to end of window, whichever is less.
+             */
+            init = (ulg)curr + WIN_INIT - s->high_water;
+            if (init > s->window_size - s->high_water)
+                init = s->window_size - s->high_water;
+            zmemzero(s->window + s->high_water, (unsigned)init);
+            s->high_water += init;
+        }
+    }
+
+    Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD,
+           "not enough room for search");
+}
+
+/* ===========================================================================
+ * Flush the current block, with given end-of-file flag.
+ * IN assertion: strstart is set to the end of the current match.
+ */
+#define FLUSH_BLOCK_ONLY(s, last) { \
+   _tr_flush_block(s, (s->block_start >= 0L ? \
+                   (charf *)&s->window[(unsigned)s->block_start] : \
+                   (charf *)Z_NULL), \
+                (ulg)((long)s->strstart - s->block_start), \
+                (last)); \
+   s->block_start = s->strstart; \
+   flush_pending(s->strm); \
+   Tracev((stderr,"[FLUSH]")); \
+}
+
+/* Same but force premature exit if necessary. */
+#define FLUSH_BLOCK(s, last) { \
+   FLUSH_BLOCK_ONLY(s, last); \
+   if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \
+}
+
+/* Maximum stored block length in deflate format (not including header). */
+#define MAX_STORED 65535
+
+/* Minimum of a and b. */
+#define MIN(a, b) ((a) > (b) ? (b) : (a))
+
+/* ===========================================================================
+ * Copy without compression as much as possible from the input stream, return
+ * the current block state.
+ *
+ * In case deflateParams() is used to later switch to a non-zero compression
+ * level, s->matches (otherwise unused when storing) keeps track of the number
+ * of hash table slides to perform. If s->matches is 1, then one hash table
+ * slide will be done when switching. If s->matches is 2, the maximum value
+ * allowed here, then the hash table will be cleared, since two or more slides
+ * is the same as a clear.
+ *
+ * deflate_stored() is written to minimize the number of times an input byte is
+ * copied. It is most efficient with large input and output buffers, which
+ * maximizes the opportunites to have a single copy from next_in to next_out.
+ */
+local block_state deflate_stored(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    /* Smallest worthy block size when not flushing or finishing. By default
+     * this is 32K. This can be as small as 507 bytes for memLevel == 1. For
+     * large input and output buffers, the stored block size will be larger.
+     */
+    unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size);
+
+    /* Copy as many min_block or larger stored blocks directly to next_out as
+     * possible. If flushing, copy the remaining available input to next_out as
+     * stored blocks, if there is enough space.
+     */
+    unsigned len, left, have, last = 0;
+    unsigned used = s->strm->avail_in;
+    do {
+        /* Set len to the maximum size block that we can copy directly with the
+         * available input data and output space. Set left to how much of that
+         * would be copied from what's left in the window.
+         */
+        len = MAX_STORED;       /* maximum deflate stored block length */
+        have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
+        if (s->strm->avail_out < have)          /* need room for header */
+            break;
+            /* maximum stored block length that will fit in avail_out: */
+        have = s->strm->avail_out - have;
+        left = s->strstart - s->block_start;    /* bytes left in window */
+        if (len > (ulg)left + s->strm->avail_in)
+            len = left + s->strm->avail_in;     /* limit len to the input */
+        if (len > have)
+            len = have;                         /* limit len to the output */
+
+        /* If the stored block would be less than min_block in length, or if
+         * unable to copy all of the available input when flushing, then try
+         * copying to the window and the pending buffer instead. Also don't
+         * write an empty block when flushing -- deflate() does that.
+         */
+        if (len < min_block && ((len == 0 && flush != Z_FINISH) ||
+                                flush == Z_NO_FLUSH ||
+                                len != left + s->strm->avail_in))
+            break;
+
+        /* Make a dummy stored block in pending to get the header bytes,
+         * including any pending bits. This also updates the debugging counts.
+         */
+        last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0;
+        _tr_stored_block(s, (char *)0, 0L, last);
+
+        /* Replace the lengths in the dummy stored block with len. */
+        s->pending_buf[s->pending - 4] = len;
+        s->pending_buf[s->pending - 3] = len >> 8;
+        s->pending_buf[s->pending - 2] = ~len;
+        s->pending_buf[s->pending - 1] = ~len >> 8;
+
+        /* Write the stored block header bytes. */
+        flush_pending(s->strm);
+
+#ifdef ZLIB_DEBUG
+        /* Update debugging counts for the data about to be copied. */
+        s->compressed_len += len << 3;
+        s->bits_sent += len << 3;
+#endif
+
+        /* Copy uncompressed bytes from the window to next_out. */
+        if (left) {
+            if (left > len)
+                left = len;
+            zmemcpy(s->strm->next_out, s->window + s->block_start, left);
+            s->strm->next_out += left;
+            s->strm->avail_out -= left;
+            s->strm->total_out += left;
+            s->block_start += left;
+            len -= left;
+        }
+
+        /* Copy uncompressed bytes directly from next_in to next_out, updating
+         * the check value.
+         */
+        if (len) {
+            read_buf(s->strm, s->strm->next_out, len);
+            s->strm->next_out += len;
+            s->strm->avail_out -= len;
+            s->strm->total_out += len;
+        }
+    } while (last == 0);
+
+    /* Update the sliding window with the last s->w_size bytes of the copied
+     * data, or append all of the copied data to the existing window if less
+     * than s->w_size bytes were copied. Also update the number of bytes to
+     * insert in the hash tables, in the event that deflateParams() switches to
+     * a non-zero compression level.
+     */
+    used -= s->strm->avail_in;      /* number of input bytes directly copied */
+    if (used) {
+        /* If any input was used, then no unused input remains in the window,
+         * therefore s->block_start == s->strstart.
+         */
+        if (used >= s->w_size) {    /* supplant the previous history */
+            s->matches = 2;         /* clear hash */
+            zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size);
+            s->strstart = s->w_size;
+        }
+        else {
+            if (s->window_size - s->strstart <= used) {
+                /* Slide the window down. */
+                s->strstart -= s->w_size;
+                zmemcpy(s->window, s->window + s->w_size, s->strstart);
+                if (s->matches < 2)
+                    s->matches++;   /* add a pending slide_hash() */
+            }
+            zmemcpy(s->window + s->strstart, s->strm->next_in - used, used);
+            s->strstart += used;
+        }
+        s->block_start = s->strstart;
+        s->insert += MIN(used, s->w_size - s->insert);
+    }
+    if (s->high_water < s->strstart)
+        s->high_water = s->strstart;
+
+    /* If the last block was written to next_out, then done. */
+    if (last)
+        return finish_done;
+
+    /* If flushing and all input has been consumed, then done. */
+    if (flush != Z_NO_FLUSH && flush != Z_FINISH &&
+        s->strm->avail_in == 0 && (long)s->strstart == s->block_start)
+        return block_done;
+
+    /* Fill the window with any remaining input. */
+    have = s->window_size - s->strstart - 1;
+    if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) {
+        /* Slide the window down. */
+        s->block_start -= s->w_size;
+        s->strstart -= s->w_size;
+        zmemcpy(s->window, s->window + s->w_size, s->strstart);
+        if (s->matches < 2)
+            s->matches++;           /* add a pending slide_hash() */
+        have += s->w_size;          /* more space now */
+    }
+    if (have > s->strm->avail_in)
+        have = s->strm->avail_in;
+    if (have) {
+        read_buf(s->strm, s->window + s->strstart, have);
+        s->strstart += have;
+    }
+    if (s->high_water < s->strstart)
+        s->high_water = s->strstart;
+
+    /* There was not enough avail_out to write a complete worthy or flushed
+     * stored block to next_out. Write a stored block to pending instead, if we
+     * have enough input for a worthy block, or if flushing and there is enough
+     * room for the remaining input as a stored block in the pending buffer.
+     */
+    have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
+        /* maximum stored block length that will fit in pending: */
+    have = MIN(s->pending_buf_size - have, MAX_STORED);
+    min_block = MIN(have, s->w_size);
+    left = s->strstart - s->block_start;
+    if (left >= min_block ||
+        ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH &&
+         s->strm->avail_in == 0 && left <= have)) {
+        len = MIN(left, have);
+        last = flush == Z_FINISH && s->strm->avail_in == 0 &&
+               len == left ? 1 : 0;
+        _tr_stored_block(s, (charf *)s->window + s->block_start, len, last);
+        s->block_start += len;
+        flush_pending(s->strm);
+    }
+
+    /* We've done all we can with the available input and output. */
+    return last ? finish_started : need_more;
+}
+
+/* ===========================================================================
+ * Compress as much as possible from the input stream, return the current
+ * block state.
+ * This function does not perform lazy evaluation of matches and inserts
+ * new strings in the dictionary only for unmatched strings or for short
+ * matches. It is used only for the fast compression options.
+ */
+local block_state deflate_fast(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head;       /* head of the hash chain */
+    int bflush;           /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        hash_head = NIL;
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         * At this point we have always match_length < MIN_MATCH
+         */
+        if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
+        }
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->match_start, s->match_length);
+
+            _tr_tally_dist(s, s->strstart - s->match_start,
+                           s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+
+            /* Insert new strings in the hash table only if the match length
+             * is not too large. This saves time but degrades compression.
+             */
+#ifndef FASTEST
+            if (s->match_length <= s->max_insert_length &&
+                s->lookahead >= MIN_MATCH) {
+                s->match_length--; /* string at strstart already in table */
+                do {
+                    s->strstart++;
+                    INSERT_STRING(s, s->strstart, hash_head);
+                    /* strstart never exceeds WSIZE-MAX_MATCH, so there are
+                     * always MIN_MATCH bytes ahead.
+                     */
+                } while (--s->match_length != 0);
+                s->strstart++;
+            } else
+#endif
+            {
+                s->strstart += s->match_length;
+                s->match_length = 0;
+                s->ins_h = s->window[s->strstart];
+                UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
+#if MIN_MATCH != 3
+                Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+                /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
+                 * matter since it will be recomputed at next deflate call.
+                 */
+            }
+        } else {
+            /* No match, output a literal byte */
+            Tracevv((stderr,"%c", s->window[s->strstart]));
+            _tr_tally_lit (s, s->window[s->strstart], bflush);
+            s->lookahead--;
+            s->strstart++;
+        }
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+
+#ifndef FASTEST
+/* ===========================================================================
+ * Same as above, but achieves better compression. We use a lazy
+ * evaluation for matches: a match is finally adopted only if there is
+ * no better match at the next window position.
+ */
+local block_state deflate_slow(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head;          /* head of hash chain */
+    int bflush;              /* set if current block must be flushed */
+
+    /* Process the input block. */
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        hash_head = NIL;
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         */
+        s->prev_length = s->match_length, s->prev_match = s->match_start;
+        s->match_length = MIN_MATCH-1;
+
+        if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
+            s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
+
+            if (s->match_length <= 5 && (s->strategy == Z_FILTERED
+#if TOO_FAR <= 32767
+                || (s->match_length == MIN_MATCH &&
+                    s->strstart - s->match_start > TOO_FAR)
+#endif
+                )) {
+
+                /* If prev_match is also MIN_MATCH, match_start is garbage
+                 * but we will ignore the current match anyway.
+                 */
+                s->match_length = MIN_MATCH-1;
+            }
+        }
+        /* If there was a match at the previous step and the current
+         * match is not better, output the previous match:
+         */
+        if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
+            uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
+            /* Do not insert strings in hash table beyond this. */
+
+            check_match(s, s->strstart-1, s->prev_match, s->prev_length);
+
+            _tr_tally_dist(s, s->strstart -1 - s->prev_match,
+                           s->prev_length - MIN_MATCH, bflush);
+
+            /* Insert in hash table all strings up to the end of the match.
+             * strstart-1 and strstart are already inserted. If there is not
+             * enough lookahead, the last two strings are not inserted in
+             * the hash table.
+             */
+            s->lookahead -= s->prev_length-1;
+            s->prev_length -= 2;
+            do {
+                if (++s->strstart <= max_insert) {
+                    INSERT_STRING(s, s->strstart, hash_head);
+                }
+            } while (--s->prev_length != 0);
+            s->match_available = 0;
+            s->match_length = MIN_MATCH-1;
+            s->strstart++;
+
+            if (bflush) FLUSH_BLOCK(s, 0);
+
+        } else if (s->match_available) {
+            /* If there was no match at the previous position, output a
+             * single literal. If there was a match but the current match
+             * is longer, truncate the previous match to a single literal.
+             */
+            Tracevv((stderr,"%c", s->window[s->strstart-1]));
+            _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+            if (bflush) {
+                FLUSH_BLOCK_ONLY(s, 0);
+            }
+            s->strstart++;
+            s->lookahead--;
+            if (s->strm->avail_out == 0) return need_more;
+        } else {
+            /* There is no previous match to compare with, wait for
+             * the next step to decide.
+             */
+            s->match_available = 1;
+            s->strstart++;
+            s->lookahead--;
+        }
+    }
+    Assert (flush != Z_NO_FLUSH, "no flush?");
+    if (s->match_available) {
+        Tracevv((stderr,"%c", s->window[s->strstart-1]));
+        _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+        s->match_available = 0;
+    }
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+#endif /* FASTEST */
+
+/* ===========================================================================
+ * For Z_RLE, simply look for runs of bytes, generate matches only of distance
+ * one.  Do not maintain a hash table.  (It will be regenerated if this run of
+ * deflate switches away from Z_RLE.)
+ */
+local block_state deflate_rle(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    int bflush;             /* set if current block must be flushed */
+    uInt prev;              /* byte at distance one to match */
+    Bytef *scan, *strend;   /* scan goes up to strend for length of run */
+
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the longest run, plus one for the unrolled loop.
+         */
+        if (s->lookahead <= MAX_MATCH) {
+            fill_window(s);
+            if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* See how many times the previous byte repeats */
+        s->match_length = 0;
+        if (s->lookahead >= MIN_MATCH && s->strstart > 0) {
+            scan = s->window + s->strstart - 1;
+            prev = *scan;
+            if (prev == *++scan && prev == *++scan && prev == *++scan) {
+                strend = s->window + s->strstart + MAX_MATCH;
+                do {
+                } while (prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         scan < strend);
+                s->match_length = MAX_MATCH - (uInt)(strend - scan);
+                if (s->match_length > s->lookahead)
+                    s->match_length = s->lookahead;
+            }
+            Assert(scan <= s->window+(uInt)(s->window_size-1), "wild scan");
+        }
+
+        /* Emit match if have run of MIN_MATCH or longer, else emit literal */
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->strstart - 1, s->match_length);
+
+            _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+            s->strstart += s->match_length;
+            s->match_length = 0;
+        } else {
+            /* No match, output a literal byte */
+            Tracevv((stderr,"%c", s->window[s->strstart]));
+            _tr_tally_lit (s, s->window[s->strstart], bflush);
+            s->lookahead--;
+            s->strstart++;
+        }
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+
+/* ===========================================================================
+ * For Z_HUFFMAN_ONLY, do not look for matches.  Do not maintain a hash table.
+ * (It will be regenerated if this run of deflate switches away from Huffman.)
+ */
+local block_state deflate_huff(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    int bflush;             /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we have a literal to write. */
+        if (s->lookahead == 0) {
+            fill_window(s);
+            if (s->lookahead == 0) {
+                if (flush == Z_NO_FLUSH)
+                    return need_more;
+                break;      /* flush the current block */
+            }
+        }
+
+        /* Output a literal byte */
+        s->match_length = 0;
+        Tracevv((stderr,"%c", s->window[s->strstart]));
+        _tr_tally_lit (s, s->window[s->strstart], bflush);
+        s->lookahead--;
+        s->strstart++;
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/deflate.h glmark2-2021.02/android/jni/src/zlib/deflate.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/deflate.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/deflate.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,349 @@
+/* deflate.h -- internal compression state
+ * Copyright (C) 1995-2016 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id$ */
+
+#ifndef DEFLATE_H
+#define DEFLATE_H
+
+#include "zutil.h"
+
+/* define NO_GZIP when compiling if you want to disable gzip header and
+   trailer creation by deflate().  NO_GZIP would be used to avoid linking in
+   the crc code when it is not needed.  For shared libraries, gzip encoding
+   should be left enabled. */
+#ifndef NO_GZIP
+#  define GZIP
+#endif
+
+/* ===========================================================================
+ * Internal compression state.
+ */
+
+#define LENGTH_CODES 29
+/* number of length codes, not counting the special END_BLOCK code */
+
+#define LITERALS  256
+/* number of literal bytes 0..255 */
+
+#define L_CODES (LITERALS+1+LENGTH_CODES)
+/* number of Literal or Length codes, including the END_BLOCK code */
+
+#define D_CODES   30
+/* number of distance codes */
+
+#define BL_CODES  19
+/* number of codes used to transfer the bit lengths */
+
+#define HEAP_SIZE (2*L_CODES+1)
+/* maximum heap size */
+
+#define MAX_BITS 15
+/* All codes must not exceed MAX_BITS bits */
+
+#define Buf_size 16
+/* size of bit buffer in bi_buf */
+
+#define INIT_STATE    42    /* zlib header -> BUSY_STATE */
+#ifdef GZIP
+#  define GZIP_STATE  57    /* gzip header -> BUSY_STATE | EXTRA_STATE */
+#endif
+#define EXTRA_STATE   69    /* gzip extra block -> NAME_STATE */
+#define NAME_STATE    73    /* gzip file name -> COMMENT_STATE */
+#define COMMENT_STATE 91    /* gzip comment -> HCRC_STATE */
+#define HCRC_STATE   103    /* gzip header CRC -> BUSY_STATE */
+#define BUSY_STATE   113    /* deflate -> FINISH_STATE */
+#define FINISH_STATE 666    /* stream complete */
+/* Stream status */
+
+
+/* Data structure describing a single value and its code string. */
+typedef struct ct_data_s {
+    union {
+        ush  freq;       /* frequency count */
+        ush  code;       /* bit string */
+    } fc;
+    union {
+        ush  dad;        /* father node in Huffman tree */
+        ush  len;        /* length of bit string */
+    } dl;
+} FAR ct_data;
+
+#define Freq fc.freq
+#define Code fc.code
+#define Dad  dl.dad
+#define Len  dl.len
+
+typedef struct static_tree_desc_s  static_tree_desc;
+
+typedef struct tree_desc_s {
+    ct_data *dyn_tree;           /* the dynamic tree */
+    int     max_code;            /* largest code with non zero frequency */
+    const static_tree_desc *stat_desc;  /* the corresponding static tree */
+} FAR tree_desc;
+
+typedef ush Pos;
+typedef Pos FAR Posf;
+typedef unsigned IPos;
+
+/* A Pos is an index in the character window. We use short instead of int to
+ * save space in the various tables. IPos is used only for parameter passing.
+ */
+
+typedef struct internal_state {
+    z_streamp strm;      /* pointer back to this zlib stream */
+    int   status;        /* as the name implies */
+    Bytef *pending_buf;  /* output still pending */
+    ulg   pending_buf_size; /* size of pending_buf */
+    Bytef *pending_out;  /* next pending byte to output to the stream */
+    ulg   pending;       /* nb of bytes in the pending buffer */
+    int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
+    gz_headerp  gzhead;  /* gzip header information to write */
+    ulg   gzindex;       /* where in extra, name, or comment */
+    Byte  method;        /* can only be DEFLATED */
+    int   last_flush;    /* value of flush param for previous deflate call */
+
+                /* used by deflate.c: */
+
+    uInt  w_size;        /* LZ77 window size (32K by default) */
+    uInt  w_bits;        /* log2(w_size)  (8..16) */
+    uInt  w_mask;        /* w_size - 1 */
+
+    Bytef *window;
+    /* Sliding window. Input bytes are read into the second half of the window,
+     * and move to the first half later to keep a dictionary of at least wSize
+     * bytes. With this organization, matches are limited to a distance of
+     * wSize-MAX_MATCH bytes, but this ensures that IO is always
+     * performed with a length multiple of the block size. Also, it limits
+     * the window size to 64K, which is quite useful on MSDOS.
+     * To do: use the user input buffer as sliding window.
+     */
+
+    ulg window_size;
+    /* Actual size of window: 2*wSize, except when the user input buffer
+     * is directly used as sliding window.
+     */
+
+    Posf *prev;
+    /* Link to older string with same hash index. To limit the size of this
+     * array to 64K, this link is maintained only for the last 32K strings.
+     * An index in this array is thus a window index modulo 32K.
+     */
+
+    Posf *head; /* Heads of the hash chains or NIL. */
+
+    uInt  ins_h;          /* hash index of string to be inserted */
+    uInt  hash_size;      /* number of elements in hash table */
+    uInt  hash_bits;      /* log2(hash_size) */
+    uInt  hash_mask;      /* hash_size-1 */
+
+    uInt  hash_shift;
+    /* Number of bits by which ins_h must be shifted at each input
+     * step. It must be such that after MIN_MATCH steps, the oldest
+     * byte no longer takes part in the hash key, that is:
+     *   hash_shift * MIN_MATCH >= hash_bits
+     */
+
+    long block_start;
+    /* Window position at the beginning of the current output block. Gets
+     * negative when the window is moved backwards.
+     */
+
+    uInt match_length;           /* length of best match */
+    IPos prev_match;             /* previous match */
+    int match_available;         /* set if previous match exists */
+    uInt strstart;               /* start of string to insert */
+    uInt match_start;            /* start of matching string */
+    uInt lookahead;              /* number of valid bytes ahead in window */
+
+    uInt prev_length;
+    /* Length of the best match at previous step. Matches not greater than this
+     * are discarded. This is used in the lazy match evaluation.
+     */
+
+    uInt max_chain_length;
+    /* To speed up deflation, hash chains are never searched beyond this
+     * length.  A higher limit improves compression ratio but degrades the
+     * speed.
+     */
+
+    uInt max_lazy_match;
+    /* Attempt to find a better match only when the current match is strictly
+     * smaller than this value. This mechanism is used only for compression
+     * levels >= 4.
+     */
+#   define max_insert_length  max_lazy_match
+    /* Insert new strings in the hash table only if the match length is not
+     * greater than this length. This saves time but degrades compression.
+     * max_insert_length is used only for compression levels <= 3.
+     */
+
+    int level;    /* compression level (1..9) */
+    int strategy; /* favor or force Huffman coding*/
+
+    uInt good_match;
+    /* Use a faster search when the previous match is longer than this */
+
+    int nice_match; /* Stop searching when current match exceeds this */
+
+                /* used by trees.c: */
+    /* Didn't use ct_data typedef below to suppress compiler warning */
+    struct ct_data_s dyn_ltree[HEAP_SIZE];   /* literal and length tree */
+    struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
+    struct ct_data_s bl_tree[2*BL_CODES+1];  /* Huffman tree for bit lengths */
+
+    struct tree_desc_s l_desc;               /* desc. for literal tree */
+    struct tree_desc_s d_desc;               /* desc. for distance tree */
+    struct tree_desc_s bl_desc;              /* desc. for bit length tree */
+
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    int heap[2*L_CODES+1];      /* heap used to build the Huffman trees */
+    int heap_len;               /* number of elements in the heap */
+    int heap_max;               /* element of largest frequency */
+    /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
+     * The same heap array is used to build all trees.
+     */
+
+    uch depth[2*L_CODES+1];
+    /* Depth of each subtree used as tie breaker for trees of equal frequency
+     */
+
+    uchf *l_buf;          /* buffer for literals or lengths */
+
+    uInt  lit_bufsize;
+    /* Size of match buffer for literals/lengths.  There are 4 reasons for
+     * limiting lit_bufsize to 64K:
+     *   - frequencies can be kept in 16 bit counters
+     *   - if compression is not successful for the first block, all input
+     *     data is still in the window so we can still emit a stored block even
+     *     when input comes from standard input.  (This can also be done for
+     *     all blocks if lit_bufsize is not greater than 32K.)
+     *   - if compression is not successful for a file smaller than 64K, we can
+     *     even emit a stored file instead of a stored block (saving 5 bytes).
+     *     This is applicable only for zip (not gzip or zlib).
+     *   - creating new Huffman trees less frequently may not provide fast
+     *     adaptation to changes in the input data statistics. (Take for
+     *     example a binary file with poorly compressible code followed by
+     *     a highly compressible string table.) Smaller buffer sizes give
+     *     fast adaptation but have of course the overhead of transmitting
+     *     trees more frequently.
+     *   - I can't count above 4
+     */
+
+    uInt last_lit;      /* running index in l_buf */
+
+    ushf *d_buf;
+    /* Buffer for distances. To simplify the code, d_buf and l_buf have
+     * the same number of elements. To use different lengths, an extra flag
+     * array would be necessary.
+     */
+
+    ulg opt_len;        /* bit length of current block with optimal trees */
+    ulg static_len;     /* bit length of current block with static trees */
+    uInt matches;       /* number of string matches in current block */
+    uInt insert;        /* bytes at end of window left to insert */
+
+#ifdef ZLIB_DEBUG
+    ulg compressed_len; /* total bit length of compressed file mod 2^32 */
+    ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
+#endif
+
+    ush bi_buf;
+    /* Output buffer. bits are inserted starting at the bottom (least
+     * significant bits).
+     */
+    int bi_valid;
+    /* Number of valid bits in bi_buf.  All bits above the last valid bit
+     * are always zero.
+     */
+
+    ulg high_water;
+    /* High water mark offset in window for initialized bytes -- bytes above
+     * this are set to zero in order to avoid memory check warnings when
+     * longest match routines access bytes past the input.  This is then
+     * updated to the new high water mark.
+     */
+
+} FAR deflate_state;
+
+/* Output a byte on the stream.
+ * IN assertion: there is enough room in pending_buf.
+ */
+#define put_byte(s, c) {s->pending_buf[s->pending++] = (Bytef)(c);}
+
+
+#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
+/* Minimum amount of lookahead, except at the end of the input file.
+ * See deflate.c for comments about the MIN_MATCH+1.
+ */
+
+#define MAX_DIST(s)  ((s)->w_size-MIN_LOOKAHEAD)
+/* In order to simplify the code, particularly on 16 bit machines, match
+ * distances are limited to MAX_DIST instead of WSIZE.
+ */
+
+#define WIN_INIT MAX_MATCH
+/* Number of bytes after end of data in window to initialize in order to avoid
+   memory checker errors from longest match routines */
+
+        /* in trees.c */
+void ZLIB_INTERNAL _tr_init OF((deflate_state *s));
+int ZLIB_INTERNAL _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc));
+void ZLIB_INTERNAL _tr_flush_block OF((deflate_state *s, charf *buf,
+                        ulg stored_len, int last));
+void ZLIB_INTERNAL _tr_flush_bits OF((deflate_state *s));
+void ZLIB_INTERNAL _tr_align OF((deflate_state *s));
+void ZLIB_INTERNAL _tr_stored_block OF((deflate_state *s, charf *buf,
+                        ulg stored_len, int last));
+
+#define d_code(dist) \
+   ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
+/* Mapping from a distance to a distance code. dist is the distance - 1 and
+ * must not have side effects. _dist_code[256] and _dist_code[257] are never
+ * used.
+ */
+
+#ifndef ZLIB_DEBUG
+/* Inline versions of _tr_tally for speed: */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+  extern uch ZLIB_INTERNAL _length_code[];
+  extern uch ZLIB_INTERNAL _dist_code[];
+#else
+  extern const uch ZLIB_INTERNAL _length_code[];
+  extern const uch ZLIB_INTERNAL _dist_code[];
+#endif
+
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->d_buf[s->last_lit] = 0; \
+    s->l_buf[s->last_lit++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (uch)(length); \
+    ush dist = (ush)(distance); \
+    s->d_buf[s->last_lit] = dist; \
+    s->l_buf[s->last_lit++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+  }
+#else
+# define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
+# define _tr_tally_dist(s, distance, length, flush) \
+              flush = _tr_tally(s, distance, length)
+#endif
+
+#endif /* DEFLATE_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzclose.c glmark2-2021.02/android/jni/src/zlib/gzclose.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzclose.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/gzclose.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,25 @@
+/* gzclose.c -- zlib gzclose() function
+ * Copyright (C) 2004, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* gzclose() is in a separate file so that it is linked in only if it is used.
+   That way the other gzclose functions can be used instead to avoid linking in
+   unneeded compression or decompression routines. */
+int ZEXPORT gzclose(file)
+    gzFile file;
+{
+#ifndef NO_GZCOMPRESS
+    gz_statep state;
+
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    return state->mode == GZ_READ ? gzclose_r(file) : gzclose_w(file);
+#else
+    return gzclose_r(file);
+#endif
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzguts.h glmark2-2021.02/android/jni/src/zlib/gzguts.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzguts.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/gzguts.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,218 @@
+/* gzguts.h -- zlib internal header definitions for gz* operations
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#ifdef _LARGEFILE64_SOURCE
+#  ifndef _LARGEFILE_SOURCE
+#    define _LARGEFILE_SOURCE 1
+#  endif
+#  ifdef _FILE_OFFSET_BITS
+#    undef _FILE_OFFSET_BITS
+#  endif
+#endif
+
+#ifdef HAVE_HIDDEN
+#  define ZLIB_INTERNAL __attribute__((visibility ("hidden")))
+#else
+#  define ZLIB_INTERNAL
+#endif
+
+#include <stdio.h>
+#include "zlib.h"
+#ifdef STDC
+#  include <string.h>
+#  include <stdlib.h>
+#  include <limits.h>
+#endif
+
+#ifndef _POSIX_SOURCE
+#  define _POSIX_SOURCE
+#endif
+#include <fcntl.h>
+
+#ifdef _WIN32
+#  include <stddef.h>
+#endif
+
+#if defined(__TURBOC__) || defined(_MSC_VER) || defined(_WIN32)
+#  include <io.h>
+#endif
+
+#if defined(_WIN32) || defined(__CYGWIN__)
+#  define WIDECHAR
+#endif
+
+#ifdef WINAPI_FAMILY
+#  define open _open
+#  define read _read
+#  define write _write
+#  define close _close
+#endif
+
+#ifdef NO_DEFLATE       /* for compatibility with old definition */
+#  define NO_GZCOMPRESS
+#endif
+
+#if defined(STDC99) || (defined(__TURBOC__) && __TURBOC__ >= 0x550)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#if defined(__CYGWIN__)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#if defined(MSDOS) && defined(__BORLANDC__) && (BORLANDC > 0x410)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#ifndef HAVE_VSNPRINTF
+#  ifdef MSDOS
+/* vsnprintf may exist on some MS-DOS compilers (DJGPP?),
+   but for now we just assume it doesn't. */
+#    define NO_vsnprintf
+#  endif
+#  ifdef __TURBOC__
+#    define NO_vsnprintf
+#  endif
+#  ifdef WIN32
+/* In Win32, vsnprintf is available as the "non-ANSI" _vsnprintf. */
+#    if !defined(vsnprintf) && !defined(NO_vsnprintf)
+#      if !defined(_MSC_VER) || ( defined(_MSC_VER) && _MSC_VER < 1500 )
+#         define vsnprintf _vsnprintf
+#      endif
+#    endif
+#  endif
+#  ifdef __SASC
+#    define NO_vsnprintf
+#  endif
+#  ifdef VMS
+#    define NO_vsnprintf
+#  endif
+#  ifdef __OS400__
+#    define NO_vsnprintf
+#  endif
+#  ifdef __MVS__
+#    define NO_vsnprintf
+#  endif
+#endif
+
+/* unlike snprintf (which is required in C99), _snprintf does not guarantee
+   null termination of the result -- however this is only used in gzlib.c where
+   the result is assured to fit in the space provided */
+#if defined(_MSC_VER) && _MSC_VER < 1900
+#  define snprintf _snprintf
+#endif
+
+#ifndef local
+#  define local static
+#endif
+/* since "static" is used to mean two completely different things in C, we
+   define "local" for the non-static meaning of "static", for readability
+   (compile with -Dlocal if your debugger can't find static symbols) */
+
+/* gz* functions always use library allocation functions */
+#ifndef STDC
+  extern voidp  malloc OF((uInt size));
+  extern void   free   OF((voidpf ptr));
+#endif
+
+/* get errno and strerror definition */
+#if defined UNDER_CE
+#  include <windows.h>
+#  define zstrerror() gz_strwinerror((DWORD)GetLastError())
+#else
+#  ifndef NO_STRERROR
+#    include <errno.h>
+#    define zstrerror() strerror(errno)
+#  else
+#    define zstrerror() "stdio error (consult errno)"
+#  endif
+#endif
+
+/* provide prototypes for these when building zlib without LFS */
+#if !defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0
+    ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+    ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int));
+    ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile));
+    ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile));
+#endif
+
+/* default memLevel */
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+
+/* default i/o buffer size -- double this for output when reading (this and
+   twice this must be able to fit in an unsigned type) */
+#define GZBUFSIZE 8192
+
+/* gzip modes, also provide a little integrity check on the passed structure */
+#define GZ_NONE 0
+#define GZ_READ 7247
+#define GZ_WRITE 31153
+#define GZ_APPEND 1     /* mode set to GZ_WRITE after the file is opened */
+
+/* values for gz_state how */
+#define LOOK 0      /* look for a gzip header */
+#define COPY 1      /* copy input directly */
+#define GZIP 2      /* decompress a gzip stream */
+
+/* internal gzip file state data structure */
+typedef struct {
+        /* exposed contents for gzgetc() macro */
+    struct gzFile_s x;      /* "x" for exposed */
+                            /* x.have: number of bytes available at x.next */
+                            /* x.next: next output data to deliver or write */
+                            /* x.pos: current position in uncompressed data */
+        /* used for both reading and writing */
+    int mode;               /* see gzip modes above */
+    int fd;                 /* file descriptor */
+    char *path;             /* path or fd for error messages */
+    unsigned size;          /* buffer size, zero if not allocated yet */
+    unsigned want;          /* requested buffer size, default is GZBUFSIZE */
+    unsigned char *in;      /* input buffer (double-sized when writing) */
+    unsigned char *out;     /* output buffer (double-sized when reading) */
+    int direct;             /* 0 if processing gzip, 1 if transparent */
+        /* just for reading */
+    int how;                /* 0: get header, 1: copy, 2: decompress */
+    z_off64_t start;        /* where the gzip data started, for rewinding */
+    int eof;                /* true if end of input file reached */
+    int past;               /* true if read requested past end */
+        /* just for writing */
+    int level;              /* compression level */
+    int strategy;           /* compression strategy */
+        /* seek request */
+    z_off64_t skip;         /* amount to skip (already rewound if backwards) */
+    int seek;               /* true if seek request pending */
+        /* error information */
+    int err;                /* error code */
+    char *msg;              /* error message */
+        /* zlib inflate or deflate stream */
+    z_stream strm;          /* stream structure in-place (not a pointer) */
+} gz_state;
+typedef gz_state FAR *gz_statep;
+
+/* shared functions */
+void ZLIB_INTERNAL gz_error OF((gz_statep, int, const char *));
+#if defined UNDER_CE
+char ZLIB_INTERNAL *gz_strwinerror OF((DWORD error));
+#endif
+
+/* GT_OFF(x), where x is an unsigned value, is true if x > maximum z_off64_t
+   value -- needed when comparing unsigned to z_off64_t, which is signed
+   (possible z_off64_t types off_t, off64_t, and long are all signed) */
+#ifdef INT_MAX
+#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > INT_MAX)
+#else
+unsigned ZLIB_INTERNAL gz_intmax OF((void));
+#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > gz_intmax())
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzlib.c glmark2-2021.02/android/jni/src/zlib/gzlib.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzlib.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/gzlib.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,637 @@
+/* gzlib.c -- zlib functions common to reading and writing gzip files
+ * Copyright (C) 2004-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+#if defined(_WIN32) && !defined(__BORLANDC__) && !defined(__MINGW32__)
+#  define LSEEK _lseeki64
+#else
+#if defined(_LARGEFILE64_SOURCE) && _LFS64_LARGEFILE-0
+#  define LSEEK lseek64
+#else
+#  define LSEEK lseek
+#endif
+#endif
+
+/* Local functions */
+local void gz_reset OF((gz_statep));
+local gzFile gz_open OF((const void *, int, const char *));
+
+#if defined UNDER_CE
+
+/* Map the Windows error number in ERROR to a locale-dependent error message
+   string and return a pointer to it.  Typically, the values for ERROR come
+   from GetLastError.
+
+   The string pointed to shall not be modified by the application, but may be
+   overwritten by a subsequent call to gz_strwinerror
+
+   The gz_strwinerror function does not change the current setting of
+   GetLastError. */
+char ZLIB_INTERNAL *gz_strwinerror (error)
+     DWORD error;
+{
+    static char buf[1024];
+
+    wchar_t *msgbuf;
+    DWORD lasterr = GetLastError();
+    DWORD chars = FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM
+        | FORMAT_MESSAGE_ALLOCATE_BUFFER,
+        NULL,
+        error,
+        0, /* Default language */
+        (LPVOID)&msgbuf,
+        0,
+        NULL);
+    if (chars != 0) {
+        /* If there is an \r\n appended, zap it.  */
+        if (chars >= 2
+            && msgbuf[chars - 2] == '\r' && msgbuf[chars - 1] == '\n') {
+            chars -= 2;
+            msgbuf[chars] = 0;
+        }
+
+        if (chars > sizeof (buf) - 1) {
+            chars = sizeof (buf) - 1;
+            msgbuf[chars] = 0;
+        }
+
+        wcstombs(buf, msgbuf, chars + 1);
+        LocalFree(msgbuf);
+    }
+    else {
+        sprintf(buf, "unknown win32 error (%ld)", error);
+    }
+
+    SetLastError(lasterr);
+    return buf;
+}
+
+#endif /* UNDER_CE */
+
+/* Reset gzip file state */
+local void gz_reset(state)
+    gz_statep state;
+{
+    state->x.have = 0;              /* no output data available */
+    if (state->mode == GZ_READ) {   /* for reading ... */
+        state->eof = 0;             /* not at end of file */
+        state->past = 0;            /* have not read past end yet */
+        state->how = LOOK;          /* look for gzip header */
+    }
+    state->seek = 0;                /* no seek request pending */
+    gz_error(state, Z_OK, NULL);    /* clear error */
+    state->x.pos = 0;               /* no uncompressed data yet */
+    state->strm.avail_in = 0;       /* no input data yet */
+}
+
+/* Open a gzip file either by name or file descriptor. */
+local gzFile gz_open(path, fd, mode)
+    const void *path;
+    int fd;
+    const char *mode;
+{
+    gz_statep state;
+    z_size_t len;
+    int oflag;
+#ifdef O_CLOEXEC
+    int cloexec = 0;
+#endif
+#ifdef O_EXCL
+    int exclusive = 0;
+#endif
+
+    /* check input */
+    if (path == NULL)
+        return NULL;
+
+    /* allocate gzFile structure to return */
+    state = (gz_statep)malloc(sizeof(gz_state));
+    if (state == NULL)
+        return NULL;
+    state->size = 0;            /* no buffers allocated yet */
+    state->want = GZBUFSIZE;    /* requested buffer size */
+    state->msg = NULL;          /* no error message yet */
+
+    /* interpret mode */
+    state->mode = GZ_NONE;
+    state->level = Z_DEFAULT_COMPRESSION;
+    state->strategy = Z_DEFAULT_STRATEGY;
+    state->direct = 0;
+    while (*mode) {
+        if (*mode >= '0' && *mode <= '9')
+            state->level = *mode - '0';
+        else
+            switch (*mode) {
+            case 'r':
+                state->mode = GZ_READ;
+                break;
+#ifndef NO_GZCOMPRESS
+            case 'w':
+                state->mode = GZ_WRITE;
+                break;
+            case 'a':
+                state->mode = GZ_APPEND;
+                break;
+#endif
+            case '+':       /* can't read and write at the same time */
+                free(state);
+                return NULL;
+            case 'b':       /* ignore -- will request binary anyway */
+                break;
+#ifdef O_CLOEXEC
+            case 'e':
+                cloexec = 1;
+                break;
+#endif
+#ifdef O_EXCL
+            case 'x':
+                exclusive = 1;
+                break;
+#endif
+            case 'f':
+                state->strategy = Z_FILTERED;
+                break;
+            case 'h':
+                state->strategy = Z_HUFFMAN_ONLY;
+                break;
+            case 'R':
+                state->strategy = Z_RLE;
+                break;
+            case 'F':
+                state->strategy = Z_FIXED;
+                break;
+            case 'T':
+                state->direct = 1;
+                break;
+            default:        /* could consider as an error, but just ignore */
+                ;
+            }
+        mode++;
+    }
+
+    /* must provide an "r", "w", or "a" */
+    if (state->mode == GZ_NONE) {
+        free(state);
+        return NULL;
+    }
+
+    /* can't force transparent read */
+    if (state->mode == GZ_READ) {
+        if (state->direct) {
+            free(state);
+            return NULL;
+        }
+        state->direct = 1;      /* for empty file */
+    }
+
+    /* save the path name for error messages */
+#ifdef WIDECHAR
+    if (fd == -2) {
+        len = wcstombs(NULL, path, 0);
+        if (len == (z_size_t)-1)
+            len = 0;
+    }
+    else
+#endif
+        len = strlen((const char *)path);
+    state->path = (char *)malloc(len + 1);
+    if (state->path == NULL) {
+        free(state);
+        return NULL;
+    }
+#ifdef WIDECHAR
+    if (fd == -2)
+        if (len)
+            wcstombs(state->path, path, len + 1);
+        else
+            *(state->path) = 0;
+    else
+#endif
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+        (void)snprintf(state->path, len + 1, "%s", (const char *)path);
+#else
+        strcpy(state->path, path);
+#endif
+
+    /* compute the flags for open() */
+    oflag =
+#ifdef O_LARGEFILE
+        O_LARGEFILE |
+#endif
+#ifdef O_BINARY
+        O_BINARY |
+#endif
+#ifdef O_CLOEXEC
+        (cloexec ? O_CLOEXEC : 0) |
+#endif
+        (state->mode == GZ_READ ?
+         O_RDONLY :
+         (O_WRONLY | O_CREAT |
+#ifdef O_EXCL
+          (exclusive ? O_EXCL : 0) |
+#endif
+          (state->mode == GZ_WRITE ?
+           O_TRUNC :
+           O_APPEND)));
+
+    /* open the file with the appropriate flags (or just use fd) */
+    state->fd = fd > -1 ? fd : (
+#ifdef WIDECHAR
+        fd == -2 ? _wopen(path, oflag, 0666) :
+#endif
+        open((const char *)path, oflag, 0666));
+    if (state->fd == -1) {
+        free(state->path);
+        free(state);
+        return NULL;
+    }
+    if (state->mode == GZ_APPEND) {
+        LSEEK(state->fd, 0, SEEK_END);  /* so gzoffset() is correct */
+        state->mode = GZ_WRITE;         /* simplify later checks */
+    }
+
+    /* save the current position for rewinding (only if reading) */
+    if (state->mode == GZ_READ) {
+        state->start = LSEEK(state->fd, 0, SEEK_CUR);
+        if (state->start == -1) state->start = 0;
+    }
+
+    /* initialize stream */
+    gz_reset(state);
+
+    /* return stream */
+    return (gzFile)state;
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzopen(path, mode)
+    const char *path;
+    const char *mode;
+{
+    return gz_open(path, -1, mode);
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzopen64(path, mode)
+    const char *path;
+    const char *mode;
+{
+    return gz_open(path, -1, mode);
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzdopen(fd, mode)
+    int fd;
+    const char *mode;
+{
+    char *path;         /* identifier for error messages */
+    gzFile gz;
+
+    if (fd == -1 || (path = (char *)malloc(7 + 3 * sizeof(int))) == NULL)
+        return NULL;
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    (void)snprintf(path, 7 + 3 * sizeof(int), "<fd:%d>", fd);
+#else
+    sprintf(path, "<fd:%d>", fd);   /* for debugging */
+#endif
+    gz = gz_open(path, fd, mode);
+    free(path);
+    return gz;
+}
+
+/* -- see zlib.h -- */
+#ifdef WIDECHAR
+gzFile ZEXPORT gzopen_w(path, mode)
+    const wchar_t *path;
+    const char *mode;
+{
+    return gz_open(path, -2, mode);
+}
+#endif
+
+/* -- see zlib.h -- */
+int ZEXPORT gzbuffer(file, size)
+    gzFile file;
+    unsigned size;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* make sure we haven't already allocated memory */
+    if (state->size != 0)
+        return -1;
+
+    /* check and set requested size */
+    if ((size << 1) < size)
+        return -1;              /* need to be able to double it */
+    if (size < 2)
+        size = 2;               /* need two bytes to check magic header */
+    state->want = size;
+    return 0;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzrewind(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* back up and start over */
+    if (LSEEK(state->fd, state->start, SEEK_SET) == -1)
+        return -1;
+    gz_reset(state);
+    return 0;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gzseek64(file, offset, whence)
+    gzFile file;
+    z_off64_t offset;
+    int whence;
+{
+    unsigned n;
+    z_off64_t ret;
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* check that there's no error */
+    if (state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+
+    /* can only seek from start or relative to current position */
+    if (whence != SEEK_SET && whence != SEEK_CUR)
+        return -1;
+
+    /* normalize offset to a SEEK_CUR specification */
+    if (whence == SEEK_SET)
+        offset -= state->x.pos;
+    else if (state->seek)
+        offset += state->skip;
+    state->seek = 0;
+
+    /* if within raw area while reading, just go there */
+    if (state->mode == GZ_READ && state->how == COPY &&
+            state->x.pos + offset >= 0) {
+        ret = LSEEK(state->fd, offset - state->x.have, SEEK_CUR);
+        if (ret == -1)
+            return -1;
+        state->x.have = 0;
+        state->eof = 0;
+        state->past = 0;
+        state->seek = 0;
+        gz_error(state, Z_OK, NULL);
+        state->strm.avail_in = 0;
+        state->x.pos += offset;
+        return state->x.pos;
+    }
+
+    /* calculate skip amount, rewinding if needed for back seek when reading */
+    if (offset < 0) {
+        if (state->mode != GZ_READ)         /* writing -- can't go backwards */
+            return -1;
+        offset += state->x.pos;
+        if (offset < 0)                     /* before start of file! */
+            return -1;
+        if (gzrewind(file) == -1)           /* rewind, then skip to offset */
+            return -1;
+    }
+
+    /* if reading, skip what's in output buffer (one less gzgetc() check) */
+    if (state->mode == GZ_READ) {
+        n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > offset ?
+            (unsigned)offset : state->x.have;
+        state->x.have -= n;
+        state->x.next += n;
+        state->x.pos += n;
+        offset -= n;
+    }
+
+    /* request skip (if not zero) */
+    if (offset) {
+        state->seek = 1;
+        state->skip = offset;
+    }
+    return state->x.pos + offset;
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gzseek(file, offset, whence)
+    gzFile file;
+    z_off_t offset;
+    int whence;
+{
+    z_off64_t ret;
+
+    ret = gzseek64(file, (z_off64_t)offset, whence);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gztell64(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* return position */
+    return state->x.pos + (state->seek ? state->skip : 0);
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gztell(file)
+    gzFile file;
+{
+    z_off64_t ret;
+
+    ret = gztell64(file);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gzoffset64(file)
+    gzFile file;
+{
+    z_off64_t offset;
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* compute and return effective offset in file */
+    offset = LSEEK(state->fd, 0, SEEK_CUR);
+    if (offset == -1)
+        return -1;
+    if (state->mode == GZ_READ)             /* reading */
+        offset -= state->strm.avail_in;     /* don't count buffered input */
+    return offset;
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gzoffset(file)
+    gzFile file;
+{
+    z_off64_t ret;
+
+    ret = gzoffset64(file);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzeof(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return 0;
+
+    /* return end-of-file state */
+    return state->mode == GZ_READ ? state->past : 0;
+}
+
+/* -- see zlib.h -- */
+const char * ZEXPORT gzerror(file, errnum)
+    gzFile file;
+    int *errnum;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return NULL;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return NULL;
+
+    /* return error information */
+    if (errnum != NULL)
+        *errnum = state->err;
+    return state->err == Z_MEM_ERROR ? "out of memory" :
+                                       (state->msg == NULL ? "" : state->msg);
+}
+
+/* -- see zlib.h -- */
+void ZEXPORT gzclearerr(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return;
+
+    /* clear error and end-of-file */
+    if (state->mode == GZ_READ) {
+        state->eof = 0;
+        state->past = 0;
+    }
+    gz_error(state, Z_OK, NULL);
+}
+
+/* Create an error message in allocated memory and set state->err and
+   state->msg accordingly.  Free any previous error message already there.  Do
+   not try to free or allocate space if the error is Z_MEM_ERROR (out of
+   memory).  Simply save the error message as a static string.  If there is an
+   allocation failure constructing the error message, then convert the error to
+   out of memory. */
+void ZLIB_INTERNAL gz_error(state, err, msg)
+    gz_statep state;
+    int err;
+    const char *msg;
+{
+    /* free previously allocated message and clear */
+    if (state->msg != NULL) {
+        if (state->err != Z_MEM_ERROR)
+            free(state->msg);
+        state->msg = NULL;
+    }
+
+    /* if fatal, set state->x.have to 0 so that the gzgetc() macro fails */
+    if (err != Z_OK && err != Z_BUF_ERROR)
+        state->x.have = 0;
+
+    /* set error code, and if no message, then done */
+    state->err = err;
+    if (msg == NULL)
+        return;
+
+    /* for an out of memory error, return literal string when requested */
+    if (err == Z_MEM_ERROR)
+        return;
+
+    /* construct error message with path */
+    if ((state->msg = (char *)malloc(strlen(state->path) + strlen(msg) + 3)) ==
+            NULL) {
+        state->err = Z_MEM_ERROR;
+        return;
+    }
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    (void)snprintf(state->msg, strlen(state->path) + strlen(msg) + 3,
+                   "%s%s%s", state->path, ": ", msg);
+#else
+    strcpy(state->msg, state->path);
+    strcat(state->msg, ": ");
+    strcat(state->msg, msg);
+#endif
+}
+
+#ifndef INT_MAX
+/* portably return maximum value for an int (when limits.h presumed not
+   available) -- we need to do this to cover cases where 2's complement not
+   used, since C standard permits 1's complement and sign-bit representations,
+   otherwise we could just use ((unsigned)-1) >> 1 */
+unsigned ZLIB_INTERNAL gz_intmax()
+{
+    unsigned p, q;
+
+    p = 1;
+    do {
+        q = p;
+        p <<= 1;
+        p++;
+    } while (p > q);
+    return q >> 1;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzread.c glmark2-2021.02/android/jni/src/zlib/gzread.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzread.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/gzread.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,654 @@
+/* gzread.c -- zlib functions for reading gzip files
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* Local functions */
+local int gz_load OF((gz_statep, unsigned char *, unsigned, unsigned *));
+local int gz_avail OF((gz_statep));
+local int gz_look OF((gz_statep));
+local int gz_decomp OF((gz_statep));
+local int gz_fetch OF((gz_statep));
+local int gz_skip OF((gz_statep, z_off64_t));
+local z_size_t gz_read OF((gz_statep, voidp, z_size_t));
+
+/* Use read() to load a buffer -- return -1 on error, otherwise 0.  Read from
+   state->fd, and update state->eof, state->err, and state->msg as appropriate.
+   This function needs to loop on read(), since read() is not guaranteed to
+   read the number of bytes requested, depending on the type of descriptor. */
+local int gz_load(state, buf, len, have)
+    gz_statep state;
+    unsigned char *buf;
+    unsigned len;
+    unsigned *have;
+{
+    int ret;
+    unsigned get, max = ((unsigned)-1 >> 2) + 1;
+
+    *have = 0;
+    do {
+        get = len - *have;
+        if (get > max)
+            get = max;
+        ret = read(state->fd, buf + *have, get);
+        if (ret <= 0)
+            break;
+        *have += (unsigned)ret;
+    } while (*have < len);
+    if (ret < 0) {
+        gz_error(state, Z_ERRNO, zstrerror());
+        return -1;
+    }
+    if (ret == 0)
+        state->eof = 1;
+    return 0;
+}
+
+/* Load up input buffer and set eof flag if last data loaded -- return -1 on
+   error, 0 otherwise.  Note that the eof flag is set when the end of the input
+   file is reached, even though there may be unused data in the buffer.  Once
+   that data has been used, no more attempts will be made to read the file.
+   If strm->avail_in != 0, then the current data is moved to the beginning of
+   the input buffer, and then the remainder of the buffer is loaded with the
+   available data from the input file. */
+local int gz_avail(state)
+    gz_statep state;
+{
+    unsigned got;
+    z_streamp strm = &(state->strm);
+
+    if (state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+    if (state->eof == 0) {
+        if (strm->avail_in) {       /* copy what's there to the start */
+            unsigned char *p = state->in;
+            unsigned const char *q = strm->next_in;
+            unsigned n = strm->avail_in;
+            do {
+                *p++ = *q++;
+            } while (--n);
+        }
+        if (gz_load(state, state->in + strm->avail_in,
+                    state->size - strm->avail_in, &got) == -1)
+            return -1;
+        strm->avail_in += got;
+        strm->next_in = state->in;
+    }
+    return 0;
+}
+
+/* Look for gzip header, set up for inflate or copy.  state->x.have must be 0.
+   If this is the first time in, allocate required memory.  state->how will be
+   left unchanged if there is no more input data available, will be set to COPY
+   if there is no gzip header and direct copying will be performed, or it will
+   be set to GZIP for decompression.  If direct copying, then leftover input
+   data from the input buffer will be copied to the output buffer.  In that
+   case, all further file reads will be directly to either the output buffer or
+   a user buffer.  If decompressing, the inflate state will be initialized.
+   gz_look() will return 0 on success or -1 on failure. */
+local int gz_look(state)
+    gz_statep state;
+{
+    z_streamp strm = &(state->strm);
+
+    /* allocate read buffers and inflate memory */
+    if (state->size == 0) {
+        /* allocate buffers */
+        state->in = (unsigned char *)malloc(state->want);
+        state->out = (unsigned char *)malloc(state->want << 1);
+        if (state->in == NULL || state->out == NULL) {
+            free(state->out);
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        state->size = state->want;
+
+        /* allocate inflate memory */
+        state->strm.zalloc = Z_NULL;
+        state->strm.zfree = Z_NULL;
+        state->strm.opaque = Z_NULL;
+        state->strm.avail_in = 0;
+        state->strm.next_in = Z_NULL;
+        if (inflateInit2(&(state->strm), 15 + 16) != Z_OK) {    /* gunzip */
+            free(state->out);
+            free(state->in);
+            state->size = 0;
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+    }
+
+    /* get at least the magic bytes in the input buffer */
+    if (strm->avail_in < 2) {
+        if (gz_avail(state) == -1)
+            return -1;
+        if (strm->avail_in == 0)
+            return 0;
+    }
+
+    /* look for gzip magic bytes -- if there, do gzip decoding (note: there is
+       a logical dilemma here when considering the case of a partially written
+       gzip file, to wit, if a single 31 byte is written, then we cannot tell
+       whether this is a single-byte file, or just a partially written gzip
+       file -- for here we assume that if a gzip file is being written, then
+       the header will be written in a single operation, so that reading a
+       single byte is sufficient indication that it is not a gzip file) */
+    if (strm->avail_in > 1 &&
+            strm->next_in[0] == 31 && strm->next_in[1] == 139) {
+        inflateReset(strm);
+        state->how = GZIP;
+        state->direct = 0;
+        return 0;
+    }
+
+    /* no gzip header -- if we were decoding gzip before, then this is trailing
+       garbage.  Ignore the trailing garbage and finish. */
+    if (state->direct == 0) {
+        strm->avail_in = 0;
+        state->eof = 1;
+        state->x.have = 0;
+        return 0;
+    }
+
+    /* doing raw i/o, copy any leftover input to output -- this assumes that
+       the output buffer is larger than the input buffer, which also assures
+       space for gzungetc() */
+    state->x.next = state->out;
+    if (strm->avail_in) {
+        memcpy(state->x.next, strm->next_in, strm->avail_in);
+        state->x.have = strm->avail_in;
+        strm->avail_in = 0;
+    }
+    state->how = COPY;
+    state->direct = 1;
+    return 0;
+}
+
+/* Decompress from input to the provided next_out and avail_out in the state.
+   On return, state->x.have and state->x.next point to the just decompressed
+   data.  If the gzip stream completes, state->how is reset to LOOK to look for
+   the next gzip stream or raw data, once state->x.have is depleted.  Returns 0
+   on success, -1 on failure. */
+local int gz_decomp(state)
+    gz_statep state;
+{
+    int ret = Z_OK;
+    unsigned had;
+    z_streamp strm = &(state->strm);
+
+    /* fill output buffer up to end of deflate stream */
+    had = strm->avail_out;
+    do {
+        /* get more input for inflate() */
+        if (strm->avail_in == 0 && gz_avail(state) == -1)
+            return -1;
+        if (strm->avail_in == 0) {
+            gz_error(state, Z_BUF_ERROR, "unexpected end of file");
+            break;
+        }
+
+        /* decompress and handle errors */
+        ret = inflate(strm, Z_NO_FLUSH);
+        if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
+            gz_error(state, Z_STREAM_ERROR,
+                     "internal error: inflate stream corrupt");
+            return -1;
+        }
+        if (ret == Z_MEM_ERROR) {
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        if (ret == Z_DATA_ERROR) {              /* deflate stream invalid */
+            gz_error(state, Z_DATA_ERROR,
+                     strm->msg == NULL ? "compressed data error" : strm->msg);
+            return -1;
+        }
+    } while (strm->avail_out && ret != Z_STREAM_END);
+
+    /* update available output */
+    state->x.have = had - strm->avail_out;
+    state->x.next = strm->next_out - state->x.have;
+
+    /* if the gzip stream completed successfully, look for another */
+    if (ret == Z_STREAM_END)
+        state->how = LOOK;
+
+    /* good decompression */
+    return 0;
+}
+
+/* Fetch data and put it in the output buffer.  Assumes state->x.have is 0.
+   Data is either copied from the input file or decompressed from the input
+   file depending on state->how.  If state->how is LOOK, then a gzip header is
+   looked for to determine whether to copy or decompress.  Returns -1 on error,
+   otherwise 0.  gz_fetch() will leave state->how as COPY or GZIP unless the
+   end of the input file has been reached and all data has been processed.  */
+local int gz_fetch(state)
+    gz_statep state;
+{
+    z_streamp strm = &(state->strm);
+
+    do {
+        switch(state->how) {
+        case LOOK:      /* -> LOOK, COPY (only if never GZIP), or GZIP */
+            if (gz_look(state) == -1)
+                return -1;
+            if (state->how == LOOK)
+                return 0;
+            break;
+        case COPY:      /* -> COPY */
+            if (gz_load(state, state->out, state->size << 1, &(state->x.have))
+                    == -1)
+                return -1;
+            state->x.next = state->out;
+            return 0;
+        case GZIP:      /* -> GZIP or LOOK (if end of gzip stream) */
+            strm->avail_out = state->size << 1;
+            strm->next_out = state->out;
+            if (gz_decomp(state) == -1)
+                return -1;
+        }
+    } while (state->x.have == 0 && (!state->eof || strm->avail_in));
+    return 0;
+}
+
+/* Skip len uncompressed bytes of output.  Return -1 on error, 0 on success. */
+local int gz_skip(state, len)
+    gz_statep state;
+    z_off64_t len;
+{
+    unsigned n;
+
+    /* skip over len bytes or reach end-of-file, whichever comes first */
+    while (len)
+        /* skip over whatever is in output buffer */
+        if (state->x.have) {
+            n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > len ?
+                (unsigned)len : state->x.have;
+            state->x.have -= n;
+            state->x.next += n;
+            state->x.pos += n;
+            len -= n;
+        }
+
+        /* output buffer empty -- return if we're at the end of the input */
+        else if (state->eof && state->strm.avail_in == 0)
+            break;
+
+        /* need more data to skip -- load up output buffer */
+        else {
+            /* get more output, looking for header if required */
+            if (gz_fetch(state) == -1)
+                return -1;
+        }
+    return 0;
+}
+
+/* Read len bytes into buf from file, or less than len up to the end of the
+   input.  Return the number of bytes read.  If zero is returned, either the
+   end of file was reached, or there was an error.  state->err must be
+   consulted in that case to determine which. */
+local z_size_t gz_read(state, buf, len)
+    gz_statep state;
+    voidp buf;
+    z_size_t len;
+{
+    z_size_t got;
+    unsigned n;
+
+    /* if len is zero, avoid unnecessary operations */
+    if (len == 0)
+        return 0;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return 0;
+    }
+
+    /* get len bytes to buf, or less than len if at the end */
+    got = 0;
+    do {
+        /* set n to the maximum amount of len that fits in an unsigned int */
+        n = -1;
+        if (n > len)
+            n = len;
+
+        /* first just try copying data from the output buffer */
+        if (state->x.have) {
+            if (state->x.have < n)
+                n = state->x.have;
+            memcpy(buf, state->x.next, n);
+            state->x.next += n;
+            state->x.have -= n;
+        }
+
+        /* output buffer empty -- return if we're at the end of the input */
+        else if (state->eof && state->strm.avail_in == 0) {
+            state->past = 1;        /* tried to read past end */
+            break;
+        }
+
+        /* need output data -- for small len or new stream load up our output
+           buffer */
+        else if (state->how == LOOK || n < (state->size << 1)) {
+            /* get more output, looking for header if required */
+            if (gz_fetch(state) == -1)
+                return 0;
+            continue;       /* no progress yet -- go back to copy above */
+            /* the copy above assures that we will leave with space in the
+               output buffer, allowing at least one gzungetc() to succeed */
+        }
+
+        /* large len -- read directly into user buffer */
+        else if (state->how == COPY) {      /* read directly */
+            if (gz_load(state, (unsigned char *)buf, n, &n) == -1)
+                return 0;
+        }
+
+        /* large len -- decompress directly into user buffer */
+        else {  /* state->how == GZIP */
+            state->strm.avail_out = n;
+            state->strm.next_out = (unsigned char *)buf;
+            if (gz_decomp(state) == -1)
+                return 0;
+            n = state->x.have;
+            state->x.have = 0;
+        }
+
+        /* update progress */
+        len -= n;
+        buf = (char *)buf + n;
+        got += n;
+        state->x.pos += n;
+    } while (len);
+
+    /* return number of bytes read into user buffer */
+    return got;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzread(file, buf, len)
+    gzFile file;
+    voidp buf;
+    unsigned len;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* since an int is returned, make sure len fits in one, otherwise return
+       with an error (this avoids a flaw in the interface) */
+    if ((int)len < 0) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in an int");
+        return -1;
+    }
+
+    /* read len or fewer bytes to buf */
+    len = gz_read(state, buf, len);
+
+    /* check for an error */
+    if (len == 0 && state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+
+    /* return the number of bytes read (this is assured to fit in an int) */
+    return (int)len;
+}
+
+/* -- see zlib.h -- */
+z_size_t ZEXPORT gzfread(buf, size, nitems, file)
+    voidp buf;
+    z_size_t size;
+    z_size_t nitems;
+    gzFile file;
+{
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return 0;
+
+    /* compute bytes to read -- error on overflow */
+    len = nitems * size;
+    if (size && len / size != nitems) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
+        return 0;
+    }
+
+    /* read len or fewer bytes to buf, return the number of full items read */
+    return len ? gz_read(state, buf, len) / size : 0;
+}
+
+/* -- see zlib.h -- */
+#ifdef Z_PREFIX_SET
+#  undef z_gzgetc
+#else
+#  undef gzgetc
+#endif
+int ZEXPORT gzgetc(file)
+    gzFile file;
+{
+    int ret;
+    unsigned char buf[1];
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* try output buffer (no need to check for skip request) */
+    if (state->x.have) {
+        state->x.have--;
+        state->x.pos++;
+        return *(state->x.next)++;
+    }
+
+    /* nothing there -- try gz_read() */
+    ret = gz_read(state, buf, 1);
+    return ret < 1 ? -1 : buf[0];
+}
+
+int ZEXPORT gzgetc_(file)
+gzFile file;
+{
+    return gzgetc(file);
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzungetc(c, file)
+    int c;
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return -1;
+    }
+
+    /* can't push EOF */
+    if (c < 0)
+        return -1;
+
+    /* if output buffer empty, put byte at end (allows more pushing) */
+    if (state->x.have == 0) {
+        state->x.have = 1;
+        state->x.next = state->out + (state->size << 1) - 1;
+        state->x.next[0] = (unsigned char)c;
+        state->x.pos--;
+        state->past = 0;
+        return c;
+    }
+
+    /* if no room, give up (must have already done a gzungetc()) */
+    if (state->x.have == (state->size << 1)) {
+        gz_error(state, Z_DATA_ERROR, "out of room to push characters");
+        return -1;
+    }
+
+    /* slide output data if needed and insert byte before existing data */
+    if (state->x.next == state->out) {
+        unsigned char *src = state->out + state->x.have;
+        unsigned char *dest = state->out + (state->size << 1);
+        while (src > state->out)
+            *--dest = *--src;
+        state->x.next = dest;
+    }
+    state->x.have++;
+    state->x.next--;
+    state->x.next[0] = (unsigned char)c;
+    state->x.pos--;
+    state->past = 0;
+    return c;
+}
+
+/* -- see zlib.h -- */
+char * ZEXPORT gzgets(file, buf, len)
+    gzFile file;
+    char *buf;
+    int len;
+{
+    unsigned left, n;
+    char *str;
+    unsigned char *eol;
+    gz_statep state;
+
+    /* check parameters and get internal structure */
+    if (file == NULL || buf == NULL || len < 1)
+        return NULL;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return NULL;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return NULL;
+    }
+
+    /* copy output bytes up to new line or len - 1, whichever comes first --
+       append a terminating zero to the string (we don't check for a zero in
+       the contents, let the user worry about that) */
+    str = buf;
+    left = (unsigned)len - 1;
+    if (left) do {
+        /* assure that something is in the output buffer */
+        if (state->x.have == 0 && gz_fetch(state) == -1)
+            return NULL;                /* error */
+        if (state->x.have == 0) {       /* end of file */
+            state->past = 1;            /* read past end */
+            break;                      /* return what we have */
+        }
+
+        /* look for end-of-line in current output buffer */
+        n = state->x.have > left ? left : state->x.have;
+        eol = (unsigned char *)memchr(state->x.next, '\n', n);
+        if (eol != NULL)
+            n = (unsigned)(eol - state->x.next) + 1;
+
+        /* copy through end-of-line, or remainder if not found */
+        memcpy(buf, state->x.next, n);
+        state->x.have -= n;
+        state->x.next += n;
+        state->x.pos += n;
+        left -= n;
+        buf += n;
+    } while (left && eol == NULL);
+
+    /* return terminated string, or if nothing, end of file */
+    if (buf == str)
+        return NULL;
+    buf[0] = 0;
+    return str;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzdirect(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* if the state is not known, but we can find out, then do so (this is
+       mainly for right after a gzopen() or gzdopen()) */
+    if (state->mode == GZ_READ && state->how == LOOK && state->x.have == 0)
+        (void)gz_look(state);
+
+    /* return 1 if transparent, 0 if processing a gzip stream */
+    return state->direct;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzclose_r(file)
+    gzFile file;
+{
+    int ret, err;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're reading */
+    if (state->mode != GZ_READ)
+        return Z_STREAM_ERROR;
+
+    /* free memory and close file */
+    if (state->size) {
+        inflateEnd(&(state->strm));
+        free(state->out);
+        free(state->in);
+    }
+    err = state->err == Z_BUF_ERROR ? Z_BUF_ERROR : Z_OK;
+    gz_error(state, Z_OK, NULL);
+    free(state->path);
+    ret = close(state->fd);
+    free(state);
+    return ret ? Z_ERRNO : err;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzwrite.c glmark2-2021.02/android/jni/src/zlib/gzwrite.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/gzwrite.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/gzwrite.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,665 @@
+/* gzwrite.c -- zlib functions for writing gzip files
+ * Copyright (C) 2004-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* Local functions */
+local int gz_init OF((gz_statep));
+local int gz_comp OF((gz_statep, int));
+local int gz_zero OF((gz_statep, z_off64_t));
+local z_size_t gz_write OF((gz_statep, voidpc, z_size_t));
+
+/* Initialize state for writing a gzip file.  Mark initialization by setting
+   state->size to non-zero.  Return -1 on a memory allocation failure, or 0 on
+   success. */
+local int gz_init(state)
+    gz_statep state;
+{
+    int ret;
+    z_streamp strm = &(state->strm);
+
+    /* allocate input buffer (double size for gzprintf) */
+    state->in = (unsigned char *)malloc(state->want << 1);
+    if (state->in == NULL) {
+        gz_error(state, Z_MEM_ERROR, "out of memory");
+        return -1;
+    }
+
+    /* only need output buffer and deflate state if compressing */
+    if (!state->direct) {
+        /* allocate output buffer */
+        state->out = (unsigned char *)malloc(state->want);
+        if (state->out == NULL) {
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+
+        /* allocate deflate memory, set up for gzip compression */
+        strm->zalloc = Z_NULL;
+        strm->zfree = Z_NULL;
+        strm->opaque = Z_NULL;
+        ret = deflateInit2(strm, state->level, Z_DEFLATED,
+                           MAX_WBITS + 16, DEF_MEM_LEVEL, state->strategy);
+        if (ret != Z_OK) {
+            free(state->out);
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        strm->next_in = NULL;
+    }
+
+    /* mark state as initialized */
+    state->size = state->want;
+
+    /* initialize write buffer if compressing */
+    if (!state->direct) {
+        strm->avail_out = state->size;
+        strm->next_out = state->out;
+        state->x.next = strm->next_out;
+    }
+    return 0;
+}
+
+/* Compress whatever is at avail_in and next_in and write to the output file.
+   Return -1 if there is an error writing to the output file or if gz_init()
+   fails to allocate memory, otherwise 0.  flush is assumed to be a valid
+   deflate() flush value.  If flush is Z_FINISH, then the deflate() state is
+   reset to start a new gzip stream.  If gz->direct is true, then simply write
+   to the output file without compressing, and ignore flush. */
+local int gz_comp(state, flush)
+    gz_statep state;
+    int flush;
+{
+    int ret, writ;
+    unsigned have, put, max = ((unsigned)-1 >> 2) + 1;
+    z_streamp strm = &(state->strm);
+
+    /* allocate memory if this is the first time through */
+    if (state->size == 0 && gz_init(state) == -1)
+        return -1;
+
+    /* write directly if requested */
+    if (state->direct) {
+        while (strm->avail_in) {
+            put = strm->avail_in > max ? max : strm->avail_in;
+            writ = write(state->fd, strm->next_in, put);
+            if (writ < 0) {
+                gz_error(state, Z_ERRNO, zstrerror());
+                return -1;
+            }
+            strm->avail_in -= (unsigned)writ;
+            strm->next_in += writ;
+        }
+        return 0;
+    }
+
+    /* run deflate() on provided input until it produces no more output */
+    ret = Z_OK;
+    do {
+        /* write out current buffer contents if full, or if flushing, but if
+           doing Z_FINISH then don't write until we get to Z_STREAM_END */
+        if (strm->avail_out == 0 || (flush != Z_NO_FLUSH &&
+            (flush != Z_FINISH || ret == Z_STREAM_END))) {
+            while (strm->next_out > state->x.next) {
+                put = strm->next_out - state->x.next > (int)max ? max :
+                      (unsigned)(strm->next_out - state->x.next);
+                writ = write(state->fd, state->x.next, put);
+                if (writ < 0) {
+                    gz_error(state, Z_ERRNO, zstrerror());
+                    return -1;
+                }
+                state->x.next += writ;
+            }
+            if (strm->avail_out == 0) {
+                strm->avail_out = state->size;
+                strm->next_out = state->out;
+                state->x.next = state->out;
+            }
+        }
+
+        /* compress */
+        have = strm->avail_out;
+        ret = deflate(strm, flush);
+        if (ret == Z_STREAM_ERROR) {
+            gz_error(state, Z_STREAM_ERROR,
+                      "internal error: deflate stream corrupt");
+            return -1;
+        }
+        have -= strm->avail_out;
+    } while (have);
+
+    /* if that completed a deflate stream, allow another to start */
+    if (flush == Z_FINISH)
+        deflateReset(strm);
+
+    /* all done, no errors */
+    return 0;
+}
+
+/* Compress len zeros to output.  Return -1 on a write error or memory
+   allocation failure by gz_comp(), or 0 on success. */
+local int gz_zero(state, len)
+    gz_statep state;
+    z_off64_t len;
+{
+    int first;
+    unsigned n;
+    z_streamp strm = &(state->strm);
+
+    /* consume whatever's left in the input buffer */
+    if (strm->avail_in && gz_comp(state, Z_NO_FLUSH) == -1)
+        return -1;
+
+    /* compress len zeros (len guaranteed > 0) */
+    first = 1;
+    while (len) {
+        n = GT_OFF(state->size) || (z_off64_t)state->size > len ?
+            (unsigned)len : state->size;
+        if (first) {
+            memset(state->in, 0, n);
+            first = 0;
+        }
+        strm->avail_in = n;
+        strm->next_in = state->in;
+        state->x.pos += n;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return -1;
+        len -= n;
+    }
+    return 0;
+}
+
+/* Write len bytes from buf to file.  Return the number of bytes written.  If
+   the returned value is less than len, then there was an error. */
+local z_size_t gz_write(state, buf, len)
+    gz_statep state;
+    voidpc buf;
+    z_size_t len;
+{
+    z_size_t put = len;
+
+    /* if len is zero, avoid unnecessary operations */
+    if (len == 0)
+        return 0;
+
+    /* allocate memory if this is the first time through */
+    if (state->size == 0 && gz_init(state) == -1)
+        return 0;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return 0;
+    }
+
+    /* for small len, copy to input buffer, otherwise compress directly */
+    if (len < state->size) {
+        /* copy to input buffer, compress when full */
+        do {
+            unsigned have, copy;
+
+            if (state->strm.avail_in == 0)
+                state->strm.next_in = state->in;
+            have = (unsigned)((state->strm.next_in + state->strm.avail_in) -
+                              state->in);
+            copy = state->size - have;
+            if (copy > len)
+                copy = len;
+            memcpy(state->in + have, buf, copy);
+            state->strm.avail_in += copy;
+            state->x.pos += copy;
+            buf = (const char *)buf + copy;
+            len -= copy;
+            if (len && gz_comp(state, Z_NO_FLUSH) == -1)
+                return 0;
+        } while (len);
+    }
+    else {
+        /* consume whatever's left in the input buffer */
+        if (state->strm.avail_in && gz_comp(state, Z_NO_FLUSH) == -1)
+            return 0;
+
+        /* directly compress user buffer to file */
+        state->strm.next_in = (z_const Bytef *)buf;
+        do {
+            unsigned n = (unsigned)-1;
+            if (n > len)
+                n = len;
+            state->strm.avail_in = n;
+            state->x.pos += n;
+            if (gz_comp(state, Z_NO_FLUSH) == -1)
+                return 0;
+            len -= n;
+        } while (len);
+    }
+
+    /* input was all buffered or compressed */
+    return put;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzwrite(file, buf, len)
+    gzFile file;
+    voidpc buf;
+    unsigned len;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return 0;
+
+    /* since an int is returned, make sure len fits in one, otherwise return
+       with an error (this avoids a flaw in the interface) */
+    if ((int)len < 0) {
+        gz_error(state, Z_DATA_ERROR, "requested length does not fit in int");
+        return 0;
+    }
+
+    /* write len bytes from buf (the return value will fit in an int) */
+    return (int)gz_write(state, buf, len);
+}
+
+/* -- see zlib.h -- */
+z_size_t ZEXPORT gzfwrite(buf, size, nitems, file)
+    voidpc buf;
+    z_size_t size;
+    z_size_t nitems;
+    gzFile file;
+{
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return 0;
+
+    /* compute bytes to read -- error on overflow */
+    len = nitems * size;
+    if (size && len / size != nitems) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
+        return 0;
+    }
+
+    /* write len bytes to buf, return the number of full items written */
+    return len ? gz_write(state, buf, len) / size : 0;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzputc(file, c)
+    gzFile file;
+    int c;
+{
+    unsigned have;
+    unsigned char buf[1];
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return -1;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return -1;
+    }
+
+    /* try writing to input buffer for speed (state->size == 0 if buffer not
+       initialized) */
+    if (state->size) {
+        if (strm->avail_in == 0)
+            strm->next_in = state->in;
+        have = (unsigned)((strm->next_in + strm->avail_in) - state->in);
+        if (have < state->size) {
+            state->in[have] = (unsigned char)c;
+            strm->avail_in++;
+            state->x.pos++;
+            return c & 0xff;
+        }
+    }
+
+    /* no room in buffer or not initialized, use gz_write() */
+    buf[0] = (unsigned char)c;
+    if (gz_write(state, buf, 1) != 1)
+        return -1;
+    return c & 0xff;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzputs(file, str)
+    gzFile file;
+    const char *str;
+{
+    int ret;
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return -1;
+
+    /* write string */
+    len = strlen(str);
+    ret = gz_write(state, str, len);
+    return ret == 0 && len != 0 ? -1 : ret;
+}
+
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#include <stdarg.h>
+
+/* -- see zlib.h -- */
+int ZEXPORTVA gzvprintf(gzFile file, const char *format, va_list va)
+{
+    int len;
+    unsigned left;
+    char *next;
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* make sure we have some buffer space */
+    if (state->size == 0 && gz_init(state) == -1)
+        return state->err;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* do the printf() into the input buffer, put length in len -- the input
+       buffer is double-sized just for this function, so there is guaranteed to
+       be state->size bytes available after the current contents */
+    if (strm->avail_in == 0)
+        strm->next_in = state->in;
+    next = (char *)(state->in + (strm->next_in - state->in) + strm->avail_in);
+    next[state->size - 1] = 0;
+#ifdef NO_vsnprintf
+#  ifdef HAS_vsprintf_void
+    (void)vsprintf(next, format, va);
+    for (len = 0; len < state->size; len++)
+        if (next[len] == 0) break;
+#  else
+    len = vsprintf(next, format, va);
+#  endif
+#else
+#  ifdef HAS_vsnprintf_void
+    (void)vsnprintf(next, state->size, format, va);
+    len = strlen(next);
+#  else
+    len = vsnprintf(next, state->size, format, va);
+#  endif
+#endif
+
+    /* check that printf() results fit in buffer */
+    if (len == 0 || (unsigned)len >= state->size || next[state->size - 1] != 0)
+        return 0;
+
+    /* update buffer and position, compress first half if past that */
+    strm->avail_in += (unsigned)len;
+    state->x.pos += len;
+    if (strm->avail_in >= state->size) {
+        left = strm->avail_in - state->size;
+        strm->avail_in = state->size;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return state->err;
+        memcpy(state->in, state->in + state->size, left);
+        strm->next_in = state->in;
+        strm->avail_in = left;
+    }
+    return len;
+}
+
+int ZEXPORTVA gzprintf(gzFile file, const char *format, ...)
+{
+    va_list va;
+    int ret;
+
+    va_start(va, format);
+    ret = gzvprintf(file, format, va);
+    va_end(va);
+    return ret;
+}
+
+#else /* !STDC && !Z_HAVE_STDARG_H */
+
+/* -- see zlib.h -- */
+int ZEXPORTVA gzprintf (file, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+                       a11, a12, a13, a14, a15, a16, a17, a18, a19, a20)
+    gzFile file;
+    const char *format;
+    int a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+        a11, a12, a13, a14, a15, a16, a17, a18, a19, a20;
+{
+    unsigned len, left;
+    char *next;
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that can really pass pointer in ints */
+    if (sizeof(int) != sizeof(void *))
+        return Z_STREAM_ERROR;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* make sure we have some buffer space */
+    if (state->size == 0 && gz_init(state) == -1)
+        return state->error;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->error;
+    }
+
+    /* do the printf() into the input buffer, put length in len -- the input
+       buffer is double-sized just for this function, so there is guaranteed to
+       be state->size bytes available after the current contents */
+    if (strm->avail_in == 0)
+        strm->next_in = state->in;
+    next = (char *)(strm->next_in + strm->avail_in);
+    next[state->size - 1] = 0;
+#ifdef NO_snprintf
+#  ifdef HAS_sprintf_void
+    sprintf(next, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11, a12,
+            a13, a14, a15, a16, a17, a18, a19, a20);
+    for (len = 0; len < size; len++)
+        if (next[len] == 0)
+            break;
+#  else
+    len = sprintf(next, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11,
+                  a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#  endif
+#else
+#  ifdef HAS_snprintf_void
+    snprintf(next, state->size, format, a1, a2, a3, a4, a5, a6, a7, a8, a9,
+             a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+    len = strlen(next);
+#  else
+    len = snprintf(next, state->size, format, a1, a2, a3, a4, a5, a6, a7, a8,
+                   a9, a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#  endif
+#endif
+
+    /* check that printf() results fit in buffer */
+    if (len == 0 || len >= state->size || next[state->size - 1] != 0)
+        return 0;
+
+    /* update buffer and position, compress first half if past that */
+    strm->avail_in += len;
+    state->x.pos += len;
+    if (strm->avail_in >= state->size) {
+        left = strm->avail_in - state->size;
+        strm->avail_in = state->size;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return state->err;
+        memcpy(state->in, state->in + state->size, left);
+        strm->next_in = state->in;
+        strm->avail_in = left;
+    }
+    return (int)len;
+}
+
+#endif
+
+/* -- see zlib.h -- */
+int ZEXPORT gzflush(file, flush)
+    gzFile file;
+    int flush;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* check flush parameter */
+    if (flush < 0 || flush > Z_FINISH)
+        return Z_STREAM_ERROR;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* compress remaining data with requested flush */
+    (void)gz_comp(state, flush);
+    return state->err;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzsetparams(file, level, strategy)
+    gzFile file;
+    int level;
+    int strategy;
+{
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* if no change is requested, then do nothing */
+    if (level == state->level && strategy == state->strategy)
+        return Z_OK;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* change compression parameters for subsequent input */
+    if (state->size) {
+        /* flush previous input with previous parameters before changing */
+        if (strm->avail_in && gz_comp(state, Z_BLOCK) == -1)
+            return state->err;
+        deflateParams(strm, level, strategy);
+    }
+    state->level = level;
+    state->strategy = strategy;
+    return Z_OK;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzclose_w(file)
+    gzFile file;
+{
+    int ret = Z_OK;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're writing */
+    if (state->mode != GZ_WRITE)
+        return Z_STREAM_ERROR;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            ret = state->err;
+    }
+
+    /* flush, free memory, and close file */
+    if (gz_comp(state, Z_FINISH) == -1)
+        ret = state->err;
+    if (state->size) {
+        if (!state->direct) {
+            (void)deflateEnd(&(state->strm));
+            free(state->out);
+        }
+        free(state->in);
+    }
+    gz_error(state, Z_OK, NULL);
+    free(state->path);
+    if (close(state->fd) == -1)
+        ret = Z_ERRNO;
+    free(state);
+    return ret;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/infback.c glmark2-2021.02/android/jni/src/zlib/infback.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/infback.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/infback.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,640 @@
+/* infback.c -- inflate using a call-back interface
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+   This code is largely copied from inflate.c.  Normally either infback.o or
+   inflate.o would be linked into an application--not both.  The interface
+   with inffast.c is retained so that optimized assembler-coded versions of
+   inflate_fast() can be used with either inflate.c or infback.c.
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+/* function prototypes */
+local void fixedtables OF((struct inflate_state FAR *state));
+
+/*
+   strm provides memory allocation functions in zalloc and zfree, or
+   Z_NULL to use the library memory allocation functions.
+
+   windowBits is in the range 8..15, and window is a user-supplied
+   window and output buffer that is 2**windowBits bytes.
+ */
+int ZEXPORT inflateBackInit_(strm, windowBits, window, version, stream_size)
+z_streamp strm;
+int windowBits;
+unsigned char FAR *window;
+const char *version;
+int stream_size;
+{
+    struct inflate_state FAR *state;
+
+    if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
+        stream_size != (int)(sizeof(z_stream)))
+        return Z_VERSION_ERROR;
+    if (strm == Z_NULL || window == Z_NULL ||
+        windowBits < 8 || windowBits > 15)
+        return Z_STREAM_ERROR;
+    strm->msg = Z_NULL;                 /* in case we return an error */
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+    strm->zfree = zcfree;
+#endif
+    state = (struct inflate_state FAR *)ZALLOC(strm, 1,
+                                               sizeof(struct inflate_state));
+    if (state == Z_NULL) return Z_MEM_ERROR;
+    Tracev((stderr, "inflate: allocated\n"));
+    strm->state = (struct internal_state FAR *)state;
+    state->dmax = 32768U;
+    state->wbits = (uInt)windowBits;
+    state->wsize = 1U << windowBits;
+    state->window = window;
+    state->wnext = 0;
+    state->whave = 0;
+    return Z_OK;
+}
+
+/*
+   Return state with length and distance decoding tables and index sizes set to
+   fixed code decoding.  Normally this returns fixed tables from inffixed.h.
+   If BUILDFIXED is defined, then instead this routine builds the tables the
+   first time it's called, and returns those tables the first time and
+   thereafter.  This reduces the size of the code by about 2K bytes, in
+   exchange for a little execution time.  However, BUILDFIXED should not be
+   used for threaded applications, since the rewriting of the tables and virgin
+   may not be thread-safe.
+ */
+local void fixedtables(state)
+struct inflate_state FAR *state;
+{
+#ifdef BUILDFIXED
+    static int virgin = 1;
+    static code *lenfix, *distfix;
+    static code fixed[544];
+
+    /* build fixed huffman tables if first call (may not be thread safe) */
+    if (virgin) {
+        unsigned sym, bits;
+        static code *next;
+
+        /* literal/length table */
+        sym = 0;
+        while (sym < 144) state->lens[sym++] = 8;
+        while (sym < 256) state->lens[sym++] = 9;
+        while (sym < 280) state->lens[sym++] = 7;
+        while (sym < 288) state->lens[sym++] = 8;
+        next = fixed;
+        lenfix = next;
+        bits = 9;
+        inflate_table(LENS, state->lens, 288, &(next), &(bits), state->work);
+
+        /* distance table */
+        sym = 0;
+        while (sym < 32) state->lens[sym++] = 5;
+        distfix = next;
+        bits = 5;
+        inflate_table(DISTS, state->lens, 32, &(next), &(bits), state->work);
+
+        /* do this just once */
+        virgin = 0;
+    }
+#else /* !BUILDFIXED */
+#   include "inffixed.h"
+#endif /* BUILDFIXED */
+    state->lencode = lenfix;
+    state->lenbits = 9;
+    state->distcode = distfix;
+    state->distbits = 5;
+}
+
+/* Macros for inflateBack(): */
+
+/* Load returned state from inflate_fast() */
+#define LOAD() \
+    do { \
+        put = strm->next_out; \
+        left = strm->avail_out; \
+        next = strm->next_in; \
+        have = strm->avail_in; \
+        hold = state->hold; \
+        bits = state->bits; \
+    } while (0)
+
+/* Set state from registers for inflate_fast() */
+#define RESTORE() \
+    do { \
+        strm->next_out = put; \
+        strm->avail_out = left; \
+        strm->next_in = next; \
+        strm->avail_in = have; \
+        state->hold = hold; \
+        state->bits = bits; \
+    } while (0)
+
+/* Clear the input bit accumulator */
+#define INITBITS() \
+    do { \
+        hold = 0; \
+        bits = 0; \
+    } while (0)
+
+/* Assure that some input is available.  If input is requested, but denied,
+   then return a Z_BUF_ERROR from inflateBack(). */
+#define PULL() \
+    do { \
+        if (have == 0) { \
+            have = in(in_desc, &next); \
+            if (have == 0) { \
+                next = Z_NULL; \
+                ret = Z_BUF_ERROR; \
+                goto inf_leave; \
+            } \
+        } \
+    } while (0)
+
+/* Get a byte of input into the bit accumulator, or return from inflateBack()
+   with an error if there is no input available. */
+#define PULLBYTE() \
+    do { \
+        PULL(); \
+        have--; \
+        hold += (unsigned long)(*next++) << bits; \
+        bits += 8; \
+    } while (0)
+
+/* Assure that there are at least n bits in the bit accumulator.  If there is
+   not enough available input to do that, then return from inflateBack() with
+   an error. */
+#define NEEDBITS(n) \
+    do { \
+        while (bits < (unsigned)(n)) \
+            PULLBYTE(); \
+    } while (0)
+
+/* Return the low n bits of the bit accumulator (n < 16) */
+#define BITS(n) \
+    ((unsigned)hold & ((1U << (n)) - 1))
+
+/* Remove n bits from the bit accumulator */
+#define DROPBITS(n) \
+    do { \
+        hold >>= (n); \
+        bits -= (unsigned)(n); \
+    } while (0)
+
+/* Remove zero to seven bits as needed to go to a byte boundary */
+#define BYTEBITS() \
+    do { \
+        hold >>= bits & 7; \
+        bits -= bits & 7; \
+    } while (0)
+
+/* Assure that some output space is available, by writing out the window
+   if it's full.  If the write fails, return from inflateBack() with a
+   Z_BUF_ERROR. */
+#define ROOM() \
+    do { \
+        if (left == 0) { \
+            put = state->window; \
+            left = state->wsize; \
+            state->whave = left; \
+            if (out(out_desc, put, left)) { \
+                ret = Z_BUF_ERROR; \
+                goto inf_leave; \
+            } \
+        } \
+    } while (0)
+
+/*
+   strm provides the memory allocation functions and window buffer on input,
+   and provides information on the unused input on return.  For Z_DATA_ERROR
+   returns, strm will also provide an error message.
+
+   in() and out() are the call-back input and output functions.  When
+   inflateBack() needs more input, it calls in().  When inflateBack() has
+   filled the window with output, or when it completes with data in the
+   window, it calls out() to write out the data.  The application must not
+   change the provided input until in() is called again or inflateBack()
+   returns.  The application must not change the window/output buffer until
+   inflateBack() returns.
+
+   in() and out() are called with a descriptor parameter provided in the
+   inflateBack() call.  This parameter can be a structure that provides the
+   information required to do the read or write, as well as accumulated
+   information on the input and output such as totals and check values.
+
+   in() should return zero on failure.  out() should return non-zero on
+   failure.  If either in() or out() fails, than inflateBack() returns a
+   Z_BUF_ERROR.  strm->next_in can be checked for Z_NULL to see whether it
+   was in() or out() that caused in the error.  Otherwise,  inflateBack()
+   returns Z_STREAM_END on success, Z_DATA_ERROR for an deflate format
+   error, or Z_MEM_ERROR if it could not allocate memory for the state.
+   inflateBack() can also return Z_STREAM_ERROR if the input parameters
+   are not correct, i.e. strm is Z_NULL or the state was not initialized.
+ */
+int ZEXPORT inflateBack(strm, in, in_desc, out, out_desc)
+z_streamp strm;
+in_func in;
+void FAR *in_desc;
+out_func out;
+void FAR *out_desc;
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *next;    /* next input */
+    unsigned char FAR *put;     /* next output */
+    unsigned have, left;        /* available input and output */
+    unsigned long hold;         /* bit buffer */
+    unsigned bits;              /* bits in bit buffer */
+    unsigned copy;              /* number of stored or match bytes to copy */
+    unsigned char FAR *from;    /* where to copy match bytes from */
+    code here;                  /* current decoding table entry */
+    code last;                  /* parent table entry */
+    unsigned len;               /* length to copy for repeats, bits to drop */
+    int ret;                    /* return code */
+    static const unsigned short order[19] = /* permutation of code lengths */
+        {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+    /* Check that the strm exists and that the state was initialized */
+    if (strm == Z_NULL || strm->state == Z_NULL)
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* Reset the state */
+    strm->msg = Z_NULL;
+    state->mode = TYPE;
+    state->last = 0;
+    state->whave = 0;
+    next = strm->next_in;
+    have = next != Z_NULL ? strm->avail_in : 0;
+    hold = 0;
+    bits = 0;
+    put = state->window;
+    left = state->wsize;
+
+    /* Inflate until end of block marked as last */
+    for (;;)
+        switch (state->mode) {
+        case TYPE:
+            /* determine and dispatch block type */
+            if (state->last) {
+                BYTEBITS();
+                state->mode = DONE;
+                break;
+            }
+            NEEDBITS(3);
+            state->last = BITS(1);
+            DROPBITS(1);
+            switch (BITS(2)) {
+            case 0:                             /* stored block */
+                Tracev((stderr, "inflate:     stored block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = STORED;
+                break;
+            case 1:                             /* fixed block */
+                fixedtables(state);
+                Tracev((stderr, "inflate:     fixed codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = LEN;              /* decode codes */
+                break;
+            case 2:                             /* dynamic block */
+                Tracev((stderr, "inflate:     dynamic codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = TABLE;
+                break;
+            case 3:
+                strm->msg = (char *)"invalid block type";
+                state->mode = BAD;
+            }
+            DROPBITS(2);
+            break;
+
+        case STORED:
+            /* get and verify stored block length */
+            BYTEBITS();                         /* go to byte boundary */
+            NEEDBITS(32);
+            if ((hold & 0xffff) != ((hold >> 16) ^ 0xffff)) {
+                strm->msg = (char *)"invalid stored block lengths";
+                state->mode = BAD;
+                break;
+            }
+            state->length = (unsigned)hold & 0xffff;
+            Tracev((stderr, "inflate:       stored length %u\n",
+                    state->length));
+            INITBITS();
+
+            /* copy stored block from input to output */
+            while (state->length != 0) {
+                copy = state->length;
+                PULL();
+                ROOM();
+                if (copy > have) copy = have;
+                if (copy > left) copy = left;
+                zmemcpy(put, next, copy);
+                have -= copy;
+                next += copy;
+                left -= copy;
+                put += copy;
+                state->length -= copy;
+            }
+            Tracev((stderr, "inflate:       stored end\n"));
+            state->mode = TYPE;
+            break;
+
+        case TABLE:
+            /* get dynamic table entries descriptor */
+            NEEDBITS(14);
+            state->nlen = BITS(5) + 257;
+            DROPBITS(5);
+            state->ndist = BITS(5) + 1;
+            DROPBITS(5);
+            state->ncode = BITS(4) + 4;
+            DROPBITS(4);
+#ifndef PKZIP_BUG_WORKAROUND
+            if (state->nlen > 286 || state->ndist > 30) {
+                strm->msg = (char *)"too many length or distance symbols";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracev((stderr, "inflate:       table sizes ok\n"));
+
+            /* get code length code lengths (not a typo) */
+            state->have = 0;
+            while (state->have < state->ncode) {
+                NEEDBITS(3);
+                state->lens[order[state->have++]] = (unsigned short)BITS(3);
+                DROPBITS(3);
+            }
+            while (state->have < 19)
+                state->lens[order[state->have++]] = 0;
+            state->next = state->codes;
+            state->lencode = (code const FAR *)(state->next);
+            state->lenbits = 7;
+            ret = inflate_table(CODES, state->lens, 19, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid code lengths set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       code lengths ok\n"));
+
+            /* get length and distance code code lengths */
+            state->have = 0;
+            while (state->have < state->nlen + state->ndist) {
+                for (;;) {
+                    here = state->lencode[BITS(state->lenbits)];
+                    if ((unsigned)(here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                if (here.val < 16) {
+                    DROPBITS(here.bits);
+                    state->lens[state->have++] = here.val;
+                }
+                else {
+                    if (here.val == 16) {
+                        NEEDBITS(here.bits + 2);
+                        DROPBITS(here.bits);
+                        if (state->have == 0) {
+                            strm->msg = (char *)"invalid bit length repeat";
+                            state->mode = BAD;
+                            break;
+                        }
+                        len = (unsigned)(state->lens[state->have - 1]);
+                        copy = 3 + BITS(2);
+                        DROPBITS(2);
+                    }
+                    else if (here.val == 17) {
+                        NEEDBITS(here.bits + 3);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 3 + BITS(3);
+                        DROPBITS(3);
+                    }
+                    else {
+                        NEEDBITS(here.bits + 7);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 11 + BITS(7);
+                        DROPBITS(7);
+                    }
+                    if (state->have + copy > state->nlen + state->ndist) {
+                        strm->msg = (char *)"invalid bit length repeat";
+                        state->mode = BAD;
+                        break;
+                    }
+                    while (copy--)
+                        state->lens[state->have++] = (unsigned short)len;
+                }
+            }
+
+            /* handle error breaks in while */
+            if (state->mode == BAD) break;
+
+            /* check for end-of-block code (better have one) */
+            if (state->lens[256] == 0) {
+                strm->msg = (char *)"invalid code -- missing end-of-block";
+                state->mode = BAD;
+                break;
+            }
+
+            /* build code tables -- note: do not change the lenbits or distbits
+               values here (9 and 6) without reading the comments in inftrees.h
+               concerning the ENOUGH constants, which depend on those values */
+            state->next = state->codes;
+            state->lencode = (code const FAR *)(state->next);
+            state->lenbits = 9;
+            ret = inflate_table(LENS, state->lens, state->nlen, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid literal/lengths set";
+                state->mode = BAD;
+                break;
+            }
+            state->distcode = (code const FAR *)(state->next);
+            state->distbits = 6;
+            ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist,
+                            &(state->next), &(state->distbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid distances set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       codes ok\n"));
+            state->mode = LEN;
+
+        case LEN:
+            /* use inflate_fast() if we have enough input and output */
+            if (have >= 6 && left >= 258) {
+                RESTORE();
+                if (state->whave < state->wsize)
+                    state->whave = state->wsize - left;
+                inflate_fast(strm, state->wsize);
+                LOAD();
+                break;
+            }
+
+            /* get a literal, length, or end-of-block code */
+            for (;;) {
+                here = state->lencode[BITS(state->lenbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if (here.op && (here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->lencode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+            }
+            DROPBITS(here.bits);
+            state->length = (unsigned)here.val;
+
+            /* process literal */
+            if (here.op == 0) {
+                Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                        "inflate:         literal '%c'\n" :
+                        "inflate:         literal 0x%02x\n", here.val));
+                ROOM();
+                *put++ = (unsigned char)(state->length);
+                left--;
+                state->mode = LEN;
+                break;
+            }
+
+            /* process end of block */
+            if (here.op & 32) {
+                Tracevv((stderr, "inflate:         end of block\n"));
+                state->mode = TYPE;
+                break;
+            }
+
+            /* invalid code */
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid literal/length code";
+                state->mode = BAD;
+                break;
+            }
+
+            /* length code -- get extra bits, if any */
+            state->extra = (unsigned)(here.op) & 15;
+            if (state->extra != 0) {
+                NEEDBITS(state->extra);
+                state->length += BITS(state->extra);
+                DROPBITS(state->extra);
+            }
+            Tracevv((stderr, "inflate:         length %u\n", state->length));
+
+            /* get distance code */
+            for (;;) {
+                here = state->distcode[BITS(state->distbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if ((here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->distcode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+            }
+            DROPBITS(here.bits);
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+            state->offset = (unsigned)here.val;
+
+            /* get distance extra bits, if any */
+            state->extra = (unsigned)(here.op) & 15;
+            if (state->extra != 0) {
+                NEEDBITS(state->extra);
+                state->offset += BITS(state->extra);
+                DROPBITS(state->extra);
+            }
+            if (state->offset > state->wsize - (state->whave < state->wsize ?
+                                                left : 0)) {
+                strm->msg = (char *)"invalid distance too far back";
+                state->mode = BAD;
+                break;
+            }
+            Tracevv((stderr, "inflate:         distance %u\n", state->offset));
+
+            /* copy match from window to output */
+            do {
+                ROOM();
+                copy = state->wsize - state->offset;
+                if (copy < left) {
+                    from = put + copy;
+                    copy = left - copy;
+                }
+                else {
+                    from = put - state->offset;
+                    copy = left;
+                }
+                if (copy > state->length) copy = state->length;
+                state->length -= copy;
+                left -= copy;
+                do {
+                    *put++ = *from++;
+                } while (--copy);
+            } while (state->length != 0);
+            break;
+
+        case DONE:
+            /* inflate stream terminated properly -- write leftover output */
+            ret = Z_STREAM_END;
+            if (left < state->wsize) {
+                if (out(out_desc, state->window, state->wsize - left))
+                    ret = Z_BUF_ERROR;
+            }
+            goto inf_leave;
+
+        case BAD:
+            ret = Z_DATA_ERROR;
+            goto inf_leave;
+
+        default:                /* can't happen, but makes compilers happy */
+            ret = Z_STREAM_ERROR;
+            goto inf_leave;
+        }
+
+    /* Return unused input */
+  inf_leave:
+    strm->next_in = next;
+    strm->avail_in = have;
+    return ret;
+}
+
+int ZEXPORT inflateBackEnd(strm)
+z_streamp strm;
+{
+    if (strm == Z_NULL || strm->state == Z_NULL || strm->zfree == (free_func)0)
+        return Z_STREAM_ERROR;
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+    Tracev((stderr, "inflate: end\n"));
+    return Z_OK;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffast.c glmark2-2021.02/android/jni/src/zlib/inffast.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffast.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inffast.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,323 @@
+/* inffast.c -- fast decoding
+ * Copyright (C) 1995-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+#ifdef ASMINF
+#  pragma message("Assembler code may have bugs -- use at your own risk")
+#else
+
+/*
+   Decode literal, length, and distance codes and write out the resulting
+   literal and match bytes until either not enough input or output is
+   available, an end-of-block is encountered, or a data error is encountered.
+   When large enough input and output buffers are supplied to inflate(), for
+   example, a 16K input buffer and a 64K output buffer, more than 95% of the
+   inflate execution time is spent in this routine.
+
+   Entry assumptions:
+
+        state->mode == LEN
+        strm->avail_in >= 6
+        strm->avail_out >= 258
+        start >= strm->avail_out
+        state->bits < 8
+
+   On return, state->mode is one of:
+
+        LEN -- ran out of enough output space or enough available input
+        TYPE -- reached end of block code, inflate() to interpret next block
+        BAD -- error in block data
+
+   Notes:
+
+    - The maximum input bits used by a length/distance pair is 15 bits for the
+      length code, 5 bits for the length extra, 15 bits for the distance code,
+      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
+      Therefore if strm->avail_in >= 6, then there is enough input to avoid
+      checking for available input while decoding.
+
+    - The maximum bytes that a single length/distance pair can output is 258
+      bytes, which is the maximum length that can be coded.  inflate_fast()
+      requires strm->avail_out >= 258 for each loop to avoid checking for
+      output space.
+ */
+void ZLIB_INTERNAL inflate_fast(strm, start)
+z_streamp strm;
+unsigned start;         /* inflate()'s starting value for strm->avail_out */
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *in;      /* local strm->next_in */
+    z_const unsigned char FAR *last;    /* have enough input while in < last */
+    unsigned char FAR *out;     /* local strm->next_out */
+    unsigned char FAR *beg;     /* inflate()'s initial strm->next_out */
+    unsigned char FAR *end;     /* while out < end, enough space available */
+#ifdef INFLATE_STRICT
+    unsigned dmax;              /* maximum distance from zlib header */
+#endif
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned wnext;             /* window write index */
+    unsigned char FAR *window;  /* allocated sliding window, if wsize != 0 */
+    unsigned long hold;         /* local strm->hold */
+    unsigned bits;              /* local strm->bits */
+    code const FAR *lcode;      /* local strm->lencode */
+    code const FAR *dcode;      /* local strm->distcode */
+    unsigned lmask;             /* mask for first level of length codes */
+    unsigned dmask;             /* mask for first level of distance codes */
+    code here;                  /* retrieved table entry */
+    unsigned op;                /* code bits, operation, extra bits, or */
+                                /*  window position, window bytes to copy */
+    unsigned len;               /* match length, unused bytes */
+    unsigned dist;              /* match distance */
+    unsigned char FAR *from;    /* where to copy match from */
+
+    /* copy state to local variables */
+    state = (struct inflate_state FAR *)strm->state;
+    in = strm->next_in;
+    last = in + (strm->avail_in - 5);
+    out = strm->next_out;
+    beg = out - (start - strm->avail_out);
+    end = out + (strm->avail_out - 257);
+#ifdef INFLATE_STRICT
+    dmax = state->dmax;
+#endif
+    wsize = state->wsize;
+    whave = state->whave;
+    wnext = state->wnext;
+    window = state->window;
+    hold = state->hold;
+    bits = state->bits;
+    lcode = state->lencode;
+    dcode = state->distcode;
+    lmask = (1U << state->lenbits) - 1;
+    dmask = (1U << state->distbits) - 1;
+
+    /* decode literals and length/distances until end-of-block or not enough
+       input data or output space */
+    do {
+        if (bits < 15) {
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+        }
+        here = lcode[hold & lmask];
+      dolen:
+        op = (unsigned)(here.bits);
+        hold >>= op;
+        bits -= op;
+        op = (unsigned)(here.op);
+        if (op == 0) {                          /* literal */
+            Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                    "inflate:         literal '%c'\n" :
+                    "inflate:         literal 0x%02x\n", here.val));
+            *out++ = (unsigned char)(here.val);
+        }
+        else if (op & 16) {                     /* length base */
+            len = (unsigned)(here.val);
+            op &= 15;                           /* number of extra bits */
+            if (op) {
+                if (bits < op) {
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                }
+                len += (unsigned)hold & ((1U << op) - 1);
+                hold >>= op;
+                bits -= op;
+            }
+            Tracevv((stderr, "inflate:         length %u\n", len));
+            if (bits < 15) {
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+            }
+            here = dcode[hold & dmask];
+          dodist:
+            op = (unsigned)(here.bits);
+            hold >>= op;
+            bits -= op;
+            op = (unsigned)(here.op);
+            if (op & 16) {                      /* distance base */
+                dist = (unsigned)(here.val);
+                op &= 15;                       /* number of extra bits */
+                if (bits < op) {
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                    if (bits < op) {
+                        hold += (unsigned long)(*in++) << bits;
+                        bits += 8;
+                    }
+                }
+                dist += (unsigned)hold & ((1U << op) - 1);
+#ifdef INFLATE_STRICT
+                if (dist > dmax) {
+                    strm->msg = (char *)"invalid distance too far back";
+                    state->mode = BAD;
+                    break;
+                }
+#endif
+                hold >>= op;
+                bits -= op;
+                Tracevv((stderr, "inflate:         distance %u\n", dist));
+                op = (unsigned)(out - beg);     /* max distance in output */
+                if (dist > op) {                /* see if copy from window */
+                    op = dist - op;             /* distance back in window */
+                    if (op > whave) {
+                        if (state->sane) {
+                            strm->msg =
+                                (char *)"invalid distance too far back";
+                            state->mode = BAD;
+                            break;
+                        }
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                        if (len <= op - whave) {
+                            do {
+                                *out++ = 0;
+                            } while (--len);
+                            continue;
+                        }
+                        len -= op - whave;
+                        do {
+                            *out++ = 0;
+                        } while (--op > whave);
+                        if (op == 0) {
+                            from = out - dist;
+                            do {
+                                *out++ = *from++;
+                            } while (--len);
+                            continue;
+                        }
+#endif
+                    }
+                    from = window;
+                    if (wnext == 0) {           /* very common case */
+                        from += wsize - op;
+                        if (op < len) {         /* some from window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = out - dist;  /* rest from output */
+                        }
+                    }
+                    else if (wnext < op) {      /* wrap around window */
+                        from += wsize + wnext - op;
+                        op -= wnext;
+                        if (op < len) {         /* some from end of window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = window;
+                            if (wnext < len) {  /* some from start of window */
+                                op = wnext;
+                                len -= op;
+                                do {
+                                    *out++ = *from++;
+                                } while (--op);
+                                from = out - dist;      /* rest from output */
+                            }
+                        }
+                    }
+                    else {                      /* contiguous in window */
+                        from += wnext - op;
+                        if (op < len) {         /* some from window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = out - dist;  /* rest from output */
+                        }
+                    }
+                    while (len > 2) {
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        len -= 3;
+                    }
+                    if (len) {
+                        *out++ = *from++;
+                        if (len > 1)
+                            *out++ = *from++;
+                    }
+                }
+                else {
+                    from = out - dist;          /* copy direct from output */
+                    do {                        /* minimum length is three */
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        len -= 3;
+                    } while (len > 2);
+                    if (len) {
+                        *out++ = *from++;
+                        if (len > 1)
+                            *out++ = *from++;
+                    }
+                }
+            }
+            else if ((op & 64) == 0) {          /* 2nd level distance code */
+                here = dcode[here.val + (hold & ((1U << op) - 1))];
+                goto dodist;
+            }
+            else {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+        }
+        else if ((op & 64) == 0) {              /* 2nd level length code */
+            here = lcode[here.val + (hold & ((1U << op) - 1))];
+            goto dolen;
+        }
+        else if (op & 32) {                     /* end-of-block */
+            Tracevv((stderr, "inflate:         end of block\n"));
+            state->mode = TYPE;
+            break;
+        }
+        else {
+            strm->msg = (char *)"invalid literal/length code";
+            state->mode = BAD;
+            break;
+        }
+    } while (in < last && out < end);
+
+    /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
+    len = bits >> 3;
+    in -= len;
+    bits -= len << 3;
+    hold &= (1U << bits) - 1;
+
+    /* update state and return */
+    strm->next_in = in;
+    strm->next_out = out;
+    strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last));
+    strm->avail_out = (unsigned)(out < end ?
+                                 257 + (end - out) : 257 - (out - end));
+    state->hold = hold;
+    state->bits = bits;
+    return;
+}
+
+/*
+   inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe):
+   - Using bit fields for code structure
+   - Different op definition to avoid & for extra bits (do & for table bits)
+   - Three separate decoding do-loops for direct, window, and wnext == 0
+   - Special case for distance > 1 copies to do overlapped load and store copy
+   - Explicit branch predictions (based on measured branch probabilities)
+   - Deferring match copy and interspersed it with decoding subsequent codes
+   - Swapping literal/length else
+   - Swapping window/direct else
+   - Larger unrolled copy loops (three is about right)
+   - Moving len -= 3 statement into middle of loop
+ */
+
+#endif /* !ASMINF */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffast.h glmark2-2021.02/android/jni/src/zlib/inffast.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffast.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inffast.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+/* inffast.h -- header to use inffast.c
+ * Copyright (C) 1995-2003, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+void ZLIB_INTERNAL inflate_fast OF((z_streamp strm, unsigned start));
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffixed.h glmark2-2021.02/android/jni/src/zlib/inffixed.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inffixed.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inffixed.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,94 @@
+    /* inffixed.h -- table for decoding fixed codes
+     * Generated automatically by makefixed().
+     */
+
+    /* WARNING: this file should *not* be used by applications.
+       It is part of the implementation of this library and is
+       subject to change. Applications should only use zlib.h.
+     */
+
+    static const code lenfix[512] = {
+        {96,7,0},{0,8,80},{0,8,16},{20,8,115},{18,7,31},{0,8,112},{0,8,48},
+        {0,9,192},{16,7,10},{0,8,96},{0,8,32},{0,9,160},{0,8,0},{0,8,128},
+        {0,8,64},{0,9,224},{16,7,6},{0,8,88},{0,8,24},{0,9,144},{19,7,59},
+        {0,8,120},{0,8,56},{0,9,208},{17,7,17},{0,8,104},{0,8,40},{0,9,176},
+        {0,8,8},{0,8,136},{0,8,72},{0,9,240},{16,7,4},{0,8,84},{0,8,20},
+        {21,8,227},{19,7,43},{0,8,116},{0,8,52},{0,9,200},{17,7,13},{0,8,100},
+        {0,8,36},{0,9,168},{0,8,4},{0,8,132},{0,8,68},{0,9,232},{16,7,8},
+        {0,8,92},{0,8,28},{0,9,152},{20,7,83},{0,8,124},{0,8,60},{0,9,216},
+        {18,7,23},{0,8,108},{0,8,44},{0,9,184},{0,8,12},{0,8,140},{0,8,76},
+        {0,9,248},{16,7,3},{0,8,82},{0,8,18},{21,8,163},{19,7,35},{0,8,114},
+        {0,8,50},{0,9,196},{17,7,11},{0,8,98},{0,8,34},{0,9,164},{0,8,2},
+        {0,8,130},{0,8,66},{0,9,228},{16,7,7},{0,8,90},{0,8,26},{0,9,148},
+        {20,7,67},{0,8,122},{0,8,58},{0,9,212},{18,7,19},{0,8,106},{0,8,42},
+        {0,9,180},{0,8,10},{0,8,138},{0,8,74},{0,9,244},{16,7,5},{0,8,86},
+        {0,8,22},{64,8,0},{19,7,51},{0,8,118},{0,8,54},{0,9,204},{17,7,15},
+        {0,8,102},{0,8,38},{0,9,172},{0,8,6},{0,8,134},{0,8,70},{0,9,236},
+        {16,7,9},{0,8,94},{0,8,30},{0,9,156},{20,7,99},{0,8,126},{0,8,62},
+        {0,9,220},{18,7,27},{0,8,110},{0,8,46},{0,9,188},{0,8,14},{0,8,142},
+        {0,8,78},{0,9,252},{96,7,0},{0,8,81},{0,8,17},{21,8,131},{18,7,31},
+        {0,8,113},{0,8,49},{0,9,194},{16,7,10},{0,8,97},{0,8,33},{0,9,162},
+        {0,8,1},{0,8,129},{0,8,65},{0,9,226},{16,7,6},{0,8,89},{0,8,25},
+        {0,9,146},{19,7,59},{0,8,121},{0,8,57},{0,9,210},{17,7,17},{0,8,105},
+        {0,8,41},{0,9,178},{0,8,9},{0,8,137},{0,8,73},{0,9,242},{16,7,4},
+        {0,8,85},{0,8,21},{16,8,258},{19,7,43},{0,8,117},{0,8,53},{0,9,202},
+        {17,7,13},{0,8,101},{0,8,37},{0,9,170},{0,8,5},{0,8,133},{0,8,69},
+        {0,9,234},{16,7,8},{0,8,93},{0,8,29},{0,9,154},{20,7,83},{0,8,125},
+        {0,8,61},{0,9,218},{18,7,23},{0,8,109},{0,8,45},{0,9,186},{0,8,13},
+        {0,8,141},{0,8,77},{0,9,250},{16,7,3},{0,8,83},{0,8,19},{21,8,195},
+        {19,7,35},{0,8,115},{0,8,51},{0,9,198},{17,7,11},{0,8,99},{0,8,35},
+        {0,9,166},{0,8,3},{0,8,131},{0,8,67},{0,9,230},{16,7,7},{0,8,91},
+        {0,8,27},{0,9,150},{20,7,67},{0,8,123},{0,8,59},{0,9,214},{18,7,19},
+        {0,8,107},{0,8,43},{0,9,182},{0,8,11},{0,8,139},{0,8,75},{0,9,246},
+        {16,7,5},{0,8,87},{0,8,23},{64,8,0},{19,7,51},{0,8,119},{0,8,55},
+        {0,9,206},{17,7,15},{0,8,103},{0,8,39},{0,9,174},{0,8,7},{0,8,135},
+        {0,8,71},{0,9,238},{16,7,9},{0,8,95},{0,8,31},{0,9,158},{20,7,99},
+        {0,8,127},{0,8,63},{0,9,222},{18,7,27},{0,8,111},{0,8,47},{0,9,190},
+        {0,8,15},{0,8,143},{0,8,79},{0,9,254},{96,7,0},{0,8,80},{0,8,16},
+        {20,8,115},{18,7,31},{0,8,112},{0,8,48},{0,9,193},{16,7,10},{0,8,96},
+        {0,8,32},{0,9,161},{0,8,0},{0,8,128},{0,8,64},{0,9,225},{16,7,6},
+        {0,8,88},{0,8,24},{0,9,145},{19,7,59},{0,8,120},{0,8,56},{0,9,209},
+        {17,7,17},{0,8,104},{0,8,40},{0,9,177},{0,8,8},{0,8,136},{0,8,72},
+        {0,9,241},{16,7,4},{0,8,84},{0,8,20},{21,8,227},{19,7,43},{0,8,116},
+        {0,8,52},{0,9,201},{17,7,13},{0,8,100},{0,8,36},{0,9,169},{0,8,4},
+        {0,8,132},{0,8,68},{0,9,233},{16,7,8},{0,8,92},{0,8,28},{0,9,153},
+        {20,7,83},{0,8,124},{0,8,60},{0,9,217},{18,7,23},{0,8,108},{0,8,44},
+        {0,9,185},{0,8,12},{0,8,140},{0,8,76},{0,9,249},{16,7,3},{0,8,82},
+        {0,8,18},{21,8,163},{19,7,35},{0,8,114},{0,8,50},{0,9,197},{17,7,11},
+        {0,8,98},{0,8,34},{0,9,165},{0,8,2},{0,8,130},{0,8,66},{0,9,229},
+        {16,7,7},{0,8,90},{0,8,26},{0,9,149},{20,7,67},{0,8,122},{0,8,58},
+        {0,9,213},{18,7,19},{0,8,106},{0,8,42},{0,9,181},{0,8,10},{0,8,138},
+        {0,8,74},{0,9,245},{16,7,5},{0,8,86},{0,8,22},{64,8,0},{19,7,51},
+        {0,8,118},{0,8,54},{0,9,205},{17,7,15},{0,8,102},{0,8,38},{0,9,173},
+        {0,8,6},{0,8,134},{0,8,70},{0,9,237},{16,7,9},{0,8,94},{0,8,30},
+        {0,9,157},{20,7,99},{0,8,126},{0,8,62},{0,9,221},{18,7,27},{0,8,110},
+        {0,8,46},{0,9,189},{0,8,14},{0,8,142},{0,8,78},{0,9,253},{96,7,0},
+        {0,8,81},{0,8,17},{21,8,131},{18,7,31},{0,8,113},{0,8,49},{0,9,195},
+        {16,7,10},{0,8,97},{0,8,33},{0,9,163},{0,8,1},{0,8,129},{0,8,65},
+        {0,9,227},{16,7,6},{0,8,89},{0,8,25},{0,9,147},{19,7,59},{0,8,121},
+        {0,8,57},{0,9,211},{17,7,17},{0,8,105},{0,8,41},{0,9,179},{0,8,9},
+        {0,8,137},{0,8,73},{0,9,243},{16,7,4},{0,8,85},{0,8,21},{16,8,258},
+        {19,7,43},{0,8,117},{0,8,53},{0,9,203},{17,7,13},{0,8,101},{0,8,37},
+        {0,9,171},{0,8,5},{0,8,133},{0,8,69},{0,9,235},{16,7,8},{0,8,93},
+        {0,8,29},{0,9,155},{20,7,83},{0,8,125},{0,8,61},{0,9,219},{18,7,23},
+        {0,8,109},{0,8,45},{0,9,187},{0,8,13},{0,8,141},{0,8,77},{0,9,251},
+        {16,7,3},{0,8,83},{0,8,19},{21,8,195},{19,7,35},{0,8,115},{0,8,51},
+        {0,9,199},{17,7,11},{0,8,99},{0,8,35},{0,9,167},{0,8,3},{0,8,131},
+        {0,8,67},{0,9,231},{16,7,7},{0,8,91},{0,8,27},{0,9,151},{20,7,67},
+        {0,8,123},{0,8,59},{0,9,215},{18,7,19},{0,8,107},{0,8,43},{0,9,183},
+        {0,8,11},{0,8,139},{0,8,75},{0,9,247},{16,7,5},{0,8,87},{0,8,23},
+        {64,8,0},{19,7,51},{0,8,119},{0,8,55},{0,9,207},{17,7,15},{0,8,103},
+        {0,8,39},{0,9,175},{0,8,7},{0,8,135},{0,8,71},{0,9,239},{16,7,9},
+        {0,8,95},{0,8,31},{0,9,159},{20,7,99},{0,8,127},{0,8,63},{0,9,223},
+        {18,7,27},{0,8,111},{0,8,47},{0,9,191},{0,8,15},{0,8,143},{0,8,79},
+        {0,9,255}
+    };
+
+    static const code distfix[32] = {
+        {16,5,1},{23,5,257},{19,5,17},{27,5,4097},{17,5,5},{25,5,1025},
+        {21,5,65},{29,5,16385},{16,5,3},{24,5,513},{20,5,33},{28,5,8193},
+        {18,5,9},{26,5,2049},{22,5,129},{64,5,0},{16,5,2},{23,5,385},
+        {19,5,25},{27,5,6145},{17,5,7},{25,5,1537},{21,5,97},{29,5,24577},
+        {16,5,4},{24,5,769},{20,5,49},{28,5,12289},{18,5,13},{26,5,3073},
+        {22,5,193},{64,5,0}
+    };
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inflate.c glmark2-2021.02/android/jni/src/zlib/inflate.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inflate.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inflate.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1561 @@
+/* inflate.c -- zlib decompression
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ * Change history:
+ *
+ * 1.2.beta0    24 Nov 2002
+ * - First version -- complete rewrite of inflate to simplify code, avoid
+ *   creation of window when not needed, minimize use of window when it is
+ *   needed, make inffast.c even faster, implement gzip decoding, and to
+ *   improve code readability and style over the previous zlib inflate code
+ *
+ * 1.2.beta1    25 Nov 2002
+ * - Use pointers for available input and output checking in inffast.c
+ * - Remove input and output counters in inffast.c
+ * - Change inffast.c entry and loop from avail_in >= 7 to >= 6
+ * - Remove unnecessary second byte pull from length extra in inffast.c
+ * - Unroll direct copy to three copies per loop in inffast.c
+ *
+ * 1.2.beta2    4 Dec 2002
+ * - Change external routine names to reduce potential conflicts
+ * - Correct filename to inffixed.h for fixed tables in inflate.c
+ * - Make hbuf[] unsigned char to match parameter type in inflate.c
+ * - Change strm->next_out[-state->offset] to *(strm->next_out - state->offset)
+ *   to avoid negation problem on Alphas (64 bit) in inflate.c
+ *
+ * 1.2.beta3    22 Dec 2002
+ * - Add comments on state->bits assertion in inffast.c
+ * - Add comments on op field in inftrees.h
+ * - Fix bug in reuse of allocated window after inflateReset()
+ * - Remove bit fields--back to byte structure for speed
+ * - Remove distance extra == 0 check in inflate_fast()--only helps for lengths
+ * - Change post-increments to pre-increments in inflate_fast(), PPC biased?
+ * - Add compile time option, POSTINC, to use post-increments instead (Intel?)
+ * - Make MATCH copy in inflate() much faster for when inflate_fast() not used
+ * - Use local copies of stream next and avail values, as well as local bit
+ *   buffer and bit count in inflate()--for speed when inflate_fast() not used
+ *
+ * 1.2.beta4    1 Jan 2003
+ * - Split ptr - 257 statements in inflate_table() to avoid compiler warnings
+ * - Move a comment on output buffer sizes from inffast.c to inflate.c
+ * - Add comments in inffast.c to introduce the inflate_fast() routine
+ * - Rearrange window copies in inflate_fast() for speed and simplification
+ * - Unroll last copy for window match in inflate_fast()
+ * - Use local copies of window variables in inflate_fast() for speed
+ * - Pull out common wnext == 0 case for speed in inflate_fast()
+ * - Make op and len in inflate_fast() unsigned for consistency
+ * - Add FAR to lcode and dcode declarations in inflate_fast()
+ * - Simplified bad distance check in inflate_fast()
+ * - Added inflateBackInit(), inflateBack(), and inflateBackEnd() in new
+ *   source file infback.c to provide a call-back interface to inflate for
+ *   programs like gzip and unzip -- uses window as output buffer to avoid
+ *   window copying
+ *
+ * 1.2.beta5    1 Jan 2003
+ * - Improved inflateBack() interface to allow the caller to provide initial
+ *   input in strm.
+ * - Fixed stored blocks bug in inflateBack()
+ *
+ * 1.2.beta6    4 Jan 2003
+ * - Added comments in inffast.c on effectiveness of POSTINC
+ * - Typecasting all around to reduce compiler warnings
+ * - Changed loops from while (1) or do {} while (1) to for (;;), again to
+ *   make compilers happy
+ * - Changed type of window in inflateBackInit() to unsigned char *
+ *
+ * 1.2.beta7    27 Jan 2003
+ * - Changed many types to unsigned or unsigned short to avoid warnings
+ * - Added inflateCopy() function
+ *
+ * 1.2.0        9 Mar 2003
+ * - Changed inflateBack() interface to provide separate opaque descriptors
+ *   for the in() and out() functions
+ * - Changed inflateBack() argument and in_func typedef to swap the length
+ *   and buffer address return values for the input function
+ * - Check next_in and next_out for Z_NULL on entry to inflate()
+ *
+ * The history for versions after 1.2.0 are in ChangeLog in zlib distribution.
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+#ifdef MAKEFIXED
+#  ifndef BUILDFIXED
+#    define BUILDFIXED
+#  endif
+#endif
+
+/* function prototypes */
+local int inflateStateCheck OF((z_streamp strm));
+local void fixedtables OF((struct inflate_state FAR *state));
+local int updatewindow OF((z_streamp strm, const unsigned char FAR *end,
+                           unsigned copy));
+#ifdef BUILDFIXED
+   void makefixed OF((void));
+#endif
+local unsigned syncsearch OF((unsigned FAR *have, const unsigned char FAR *buf,
+                              unsigned len));
+
+local int inflateStateCheck(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (strm == Z_NULL ||
+        strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0)
+        return 1;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state == Z_NULL || state->strm != strm ||
+        state->mode < HEAD || state->mode > SYNC)
+        return 1;
+    return 0;
+}
+
+int ZEXPORT inflateResetKeep(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    strm->total_in = strm->total_out = state->total = 0;
+    strm->msg = Z_NULL;
+    if (state->wrap)        /* to support ill-conceived Java test suite */
+        strm->adler = state->wrap & 1;
+    state->mode = HEAD;
+    state->last = 0;
+    state->havedict = 0;
+    state->dmax = 32768U;
+    state->head = Z_NULL;
+    state->hold = 0;
+    state->bits = 0;
+    state->lencode = state->distcode = state->next = state->codes;
+    state->sane = 1;
+    state->back = -1;
+    Tracev((stderr, "inflate: reset\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateReset(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    state->wsize = 0;
+    state->whave = 0;
+    state->wnext = 0;
+    return inflateResetKeep(strm);
+}
+
+int ZEXPORT inflateReset2(strm, windowBits)
+z_streamp strm;
+int windowBits;
+{
+    int wrap;
+    struct inflate_state FAR *state;
+
+    /* get the state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* extract wrap request from windowBits parameter */
+    if (windowBits < 0) {
+        wrap = 0;
+        windowBits = -windowBits;
+    }
+    else {
+        wrap = (windowBits >> 4) + 5;
+#ifdef GUNZIP
+        if (windowBits < 48)
+            windowBits &= 15;
+#endif
+    }
+
+    /* set number of window bits, free window if different */
+    if (windowBits && (windowBits < 8 || windowBits > 15))
+        return Z_STREAM_ERROR;
+    if (state->window != Z_NULL && state->wbits != (unsigned)windowBits) {
+        ZFREE(strm, state->window);
+        state->window = Z_NULL;
+    }
+
+    /* update state and reset the rest of it */
+    state->wrap = wrap;
+    state->wbits = (unsigned)windowBits;
+    return inflateReset(strm);
+}
+
+int ZEXPORT inflateInit2_(strm, windowBits, version, stream_size)
+z_streamp strm;
+int windowBits;
+const char *version;
+int stream_size;
+{
+    int ret;
+    struct inflate_state FAR *state;
+
+    if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
+        stream_size != (int)(sizeof(z_stream)))
+        return Z_VERSION_ERROR;
+    if (strm == Z_NULL) return Z_STREAM_ERROR;
+    strm->msg = Z_NULL;                 /* in case we return an error */
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zfree = zcfree;
+#endif
+    state = (struct inflate_state FAR *)
+            ZALLOC(strm, 1, sizeof(struct inflate_state));
+    if (state == Z_NULL) return Z_MEM_ERROR;
+    Tracev((stderr, "inflate: allocated\n"));
+    strm->state = (struct internal_state FAR *)state;
+    state->strm = strm;
+    state->window = Z_NULL;
+    state->mode = HEAD;     /* to pass state test in inflateReset2() */
+    ret = inflateReset2(strm, windowBits);
+    if (ret != Z_OK) {
+        ZFREE(strm, state);
+        strm->state = Z_NULL;
+    }
+    return ret;
+}
+
+int ZEXPORT inflateInit_(strm, version, stream_size)
+z_streamp strm;
+const char *version;
+int stream_size;
+{
+    return inflateInit2_(strm, DEF_WBITS, version, stream_size);
+}
+
+int ZEXPORT inflatePrime(strm, bits, value)
+z_streamp strm;
+int bits;
+int value;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (bits < 0) {
+        state->hold = 0;
+        state->bits = 0;
+        return Z_OK;
+    }
+    if (bits > 16 || state->bits + (uInt)bits > 32) return Z_STREAM_ERROR;
+    value &= (1L << bits) - 1;
+    state->hold += (unsigned)value << state->bits;
+    state->bits += (uInt)bits;
+    return Z_OK;
+}
+
+/*
+   Return state with length and distance decoding tables and index sizes set to
+   fixed code decoding.  Normally this returns fixed tables from inffixed.h.
+   If BUILDFIXED is defined, then instead this routine builds the tables the
+   first time it's called, and returns those tables the first time and
+   thereafter.  This reduces the size of the code by about 2K bytes, in
+   exchange for a little execution time.  However, BUILDFIXED should not be
+   used for threaded applications, since the rewriting of the tables and virgin
+   may not be thread-safe.
+ */
+local void fixedtables(state)
+struct inflate_state FAR *state;
+{
+#ifdef BUILDFIXED
+    static int virgin = 1;
+    static code *lenfix, *distfix;
+    static code fixed[544];
+
+    /* build fixed huffman tables if first call (may not be thread safe) */
+    if (virgin) {
+        unsigned sym, bits;
+        static code *next;
+
+        /* literal/length table */
+        sym = 0;
+        while (sym < 144) state->lens[sym++] = 8;
+        while (sym < 256) state->lens[sym++] = 9;
+        while (sym < 280) state->lens[sym++] = 7;
+        while (sym < 288) state->lens[sym++] = 8;
+        next = fixed;
+        lenfix = next;
+        bits = 9;
+        inflate_table(LENS, state->lens, 288, &(next), &(bits), state->work);
+
+        /* distance table */
+        sym = 0;
+        while (sym < 32) state->lens[sym++] = 5;
+        distfix = next;
+        bits = 5;
+        inflate_table(DISTS, state->lens, 32, &(next), &(bits), state->work);
+
+        /* do this just once */
+        virgin = 0;
+    }
+#else /* !BUILDFIXED */
+#   include "inffixed.h"
+#endif /* BUILDFIXED */
+    state->lencode = lenfix;
+    state->lenbits = 9;
+    state->distcode = distfix;
+    state->distbits = 5;
+}
+
+#ifdef MAKEFIXED
+#include <stdio.h>
+
+/*
+   Write out the inffixed.h that is #include'd above.  Defining MAKEFIXED also
+   defines BUILDFIXED, so the tables are built on the fly.  makefixed() writes
+   those tables to stdout, which would be piped to inffixed.h.  A small program
+   can simply call makefixed to do this:
+
+    void makefixed(void);
+
+    int main(void)
+    {
+        makefixed();
+        return 0;
+    }
+
+   Then that can be linked with zlib built with MAKEFIXED defined and run:
+
+    a.out > inffixed.h
+ */
+void makefixed()
+{
+    unsigned low, size;
+    struct inflate_state state;
+
+    fixedtables(&state);
+    puts("    /* inffixed.h -- table for decoding fixed codes");
+    puts("     * Generated automatically by makefixed().");
+    puts("     */");
+    puts("");
+    puts("    /* WARNING: this file should *not* be used by applications.");
+    puts("       It is part of the implementation of this library and is");
+    puts("       subject to change. Applications should only use zlib.h.");
+    puts("     */");
+    puts("");
+    size = 1U << 9;
+    printf("    static const code lenfix[%u] = {", size);
+    low = 0;
+    for (;;) {
+        if ((low % 7) == 0) printf("\n        ");
+        printf("{%u,%u,%d}", (low & 127) == 99 ? 64 : state.lencode[low].op,
+               state.lencode[low].bits, state.lencode[low].val);
+        if (++low == size) break;
+        putchar(',');
+    }
+    puts("\n    };");
+    size = 1U << 5;
+    printf("\n    static const code distfix[%u] = {", size);
+    low = 0;
+    for (;;) {
+        if ((low % 6) == 0) printf("\n        ");
+        printf("{%u,%u,%d}", state.distcode[low].op, state.distcode[low].bits,
+               state.distcode[low].val);
+        if (++low == size) break;
+        putchar(',');
+    }
+    puts("\n    };");
+}
+#endif /* MAKEFIXED */
+
+/*
+   Update the window with the last wsize (normally 32K) bytes written before
+   returning.  If window does not exist yet, create it.  This is only called
+   when a window is already in use, or when output has been written during this
+   inflate call, but the end of the deflate stream has not been reached yet.
+   It is also called to create a window for dictionary data when a dictionary
+   is loaded.
+
+   Providing output buffers larger than 32K to inflate() should provide a speed
+   advantage, since only the last 32K of output is copied to the sliding window
+   upon return from inflate(), and since all distances after the first 32K of
+   output will fall in the output data, making match copies simpler and faster.
+   The advantage may be dependent on the size of the processor's data caches.
+ */
+local int updatewindow(strm, end, copy)
+z_streamp strm;
+const Bytef *end;
+unsigned copy;
+{
+    struct inflate_state FAR *state;
+    unsigned dist;
+
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* if it hasn't been done already, allocate space for the window */
+    if (state->window == Z_NULL) {
+        state->window = (unsigned char FAR *)
+                        ZALLOC(strm, 1U << state->wbits,
+                               sizeof(unsigned char));
+        if (state->window == Z_NULL) return 1;
+    }
+
+    /* if window not in use yet, initialize */
+    if (state->wsize == 0) {
+        state->wsize = 1U << state->wbits;
+        state->wnext = 0;
+        state->whave = 0;
+    }
+
+    /* copy state->wsize or less output bytes into the circular window */
+    if (copy >= state->wsize) {
+        zmemcpy(state->window, end - state->wsize, state->wsize);
+        state->wnext = 0;
+        state->whave = state->wsize;
+    }
+    else {
+        dist = state->wsize - state->wnext;
+        if (dist > copy) dist = copy;
+        zmemcpy(state->window + state->wnext, end - copy, dist);
+        copy -= dist;
+        if (copy) {
+            zmemcpy(state->window, end - copy, copy);
+            state->wnext = copy;
+            state->whave = state->wsize;
+        }
+        else {
+            state->wnext += dist;
+            if (state->wnext == state->wsize) state->wnext = 0;
+            if (state->whave < state->wsize) state->whave += dist;
+        }
+    }
+    return 0;
+}
+
+/* Macros for inflate(): */
+
+/* check function to use adler32() for zlib or crc32() for gzip */
+#ifdef GUNZIP
+#  define UPDATE(check, buf, len) \
+    (state->flags ? crc32(check, buf, len) : adler32(check, buf, len))
+#else
+#  define UPDATE(check, buf, len) adler32(check, buf, len)
+#endif
+
+/* check macros for header crc */
+#ifdef GUNZIP
+#  define CRC2(check, word) \
+    do { \
+        hbuf[0] = (unsigned char)(word); \
+        hbuf[1] = (unsigned char)((word) >> 8); \
+        check = crc32(check, hbuf, 2); \
+    } while (0)
+
+#  define CRC4(check, word) \
+    do { \
+        hbuf[0] = (unsigned char)(word); \
+        hbuf[1] = (unsigned char)((word) >> 8); \
+        hbuf[2] = (unsigned char)((word) >> 16); \
+        hbuf[3] = (unsigned char)((word) >> 24); \
+        check = crc32(check, hbuf, 4); \
+    } while (0)
+#endif
+
+/* Load registers with state in inflate() for speed */
+#define LOAD() \
+    do { \
+        put = strm->next_out; \
+        left = strm->avail_out; \
+        next = strm->next_in; \
+        have = strm->avail_in; \
+        hold = state->hold; \
+        bits = state->bits; \
+    } while (0)
+
+/* Restore state from registers in inflate() */
+#define RESTORE() \
+    do { \
+        strm->next_out = put; \
+        strm->avail_out = left; \
+        strm->next_in = next; \
+        strm->avail_in = have; \
+        state->hold = hold; \
+        state->bits = bits; \
+    } while (0)
+
+/* Clear the input bit accumulator */
+#define INITBITS() \
+    do { \
+        hold = 0; \
+        bits = 0; \
+    } while (0)
+
+/* Get a byte of input into the bit accumulator, or return from inflate()
+   if there is no input available. */
+#define PULLBYTE() \
+    do { \
+        if (have == 0) goto inf_leave; \
+        have--; \
+        hold += (unsigned long)(*next++) << bits; \
+        bits += 8; \
+    } while (0)
+
+/* Assure that there are at least n bits in the bit accumulator.  If there is
+   not enough available input to do that, then return from inflate(). */
+#define NEEDBITS(n) \
+    do { \
+        while (bits < (unsigned)(n)) \
+            PULLBYTE(); \
+    } while (0)
+
+/* Return the low n bits of the bit accumulator (n < 16) */
+#define BITS(n) \
+    ((unsigned)hold & ((1U << (n)) - 1))
+
+/* Remove n bits from the bit accumulator */
+#define DROPBITS(n) \
+    do { \
+        hold >>= (n); \
+        bits -= (unsigned)(n); \
+    } while (0)
+
+/* Remove zero to seven bits as needed to go to a byte boundary */
+#define BYTEBITS() \
+    do { \
+        hold >>= bits & 7; \
+        bits -= bits & 7; \
+    } while (0)
+
+/*
+   inflate() uses a state machine to process as much input data and generate as
+   much output data as possible before returning.  The state machine is
+   structured roughly as follows:
+
+    for (;;) switch (state) {
+    ...
+    case STATEn:
+        if (not enough input data or output space to make progress)
+            return;
+        ... make progress ...
+        state = STATEm;
+        break;
+    ...
+    }
+
+   so when inflate() is called again, the same case is attempted again, and
+   if the appropriate resources are provided, the machine proceeds to the
+   next state.  The NEEDBITS() macro is usually the way the state evaluates
+   whether it can proceed or should return.  NEEDBITS() does the return if
+   the requested bits are not available.  The typical use of the BITS macros
+   is:
+
+        NEEDBITS(n);
+        ... do something with BITS(n) ...
+        DROPBITS(n);
+
+   where NEEDBITS(n) either returns from inflate() if there isn't enough
+   input left to load n bits into the accumulator, or it continues.  BITS(n)
+   gives the low n bits in the accumulator.  When done, DROPBITS(n) drops
+   the low n bits off the accumulator.  INITBITS() clears the accumulator
+   and sets the number of available bits to zero.  BYTEBITS() discards just
+   enough bits to put the accumulator on a byte boundary.  After BYTEBITS()
+   and a NEEDBITS(8), then BITS(8) would return the next byte in the stream.
+
+   NEEDBITS(n) uses PULLBYTE() to get an available byte of input, or to return
+   if there is no input available.  The decoding of variable length codes uses
+   PULLBYTE() directly in order to pull just enough bytes to decode the next
+   code, and no more.
+
+   Some states loop until they get enough input, making sure that enough
+   state information is maintained to continue the loop where it left off
+   if NEEDBITS() returns in the loop.  For example, want, need, and keep
+   would all have to actually be part of the saved state in case NEEDBITS()
+   returns:
+
+    case STATEw:
+        while (want < need) {
+            NEEDBITS(n);
+            keep[want++] = BITS(n);
+            DROPBITS(n);
+        }
+        state = STATEx;
+    case STATEx:
+
+   As shown above, if the next state is also the next case, then the break
+   is omitted.
+
+   A state may also return if there is not enough output space available to
+   complete that state.  Those states are copying stored data, writing a
+   literal byte, and copying a matching string.
+
+   When returning, a "goto inf_leave" is used to update the total counters,
+   update the check value, and determine whether any progress has been made
+   during that inflate() call in order to return the proper return code.
+   Progress is defined as a change in either strm->avail_in or strm->avail_out.
+   When there is a window, goto inf_leave will update the window with the last
+   output written.  If a goto inf_leave occurs in the middle of decompression
+   and there is no window currently, goto inf_leave will create one and copy
+   output to the window for the next call of inflate().
+
+   In this implementation, the flush parameter of inflate() only affects the
+   return code (per zlib.h).  inflate() always writes as much as possible to
+   strm->next_out, given the space available and the provided input--the effect
+   documented in zlib.h of Z_SYNC_FLUSH.  Furthermore, inflate() always defers
+   the allocation of and copying into a sliding window until necessary, which
+   provides the effect documented in zlib.h for Z_FINISH when the entire input
+   stream available.  So the only thing the flush parameter actually does is:
+   when flush is set to Z_FINISH, inflate() cannot return Z_OK.  Instead it
+   will return Z_BUF_ERROR if it has not reached the end of the stream.
+ */
+
+int ZEXPORT inflate(strm, flush)
+z_streamp strm;
+int flush;
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *next;    /* next input */
+    unsigned char FAR *put;     /* next output */
+    unsigned have, left;        /* available input and output */
+    unsigned long hold;         /* bit buffer */
+    unsigned bits;              /* bits in bit buffer */
+    unsigned in, out;           /* save starting available input and output */
+    unsigned copy;              /* number of stored or match bytes to copy */
+    unsigned char FAR *from;    /* where to copy match bytes from */
+    code here;                  /* current decoding table entry */
+    code last;                  /* parent table entry */
+    unsigned len;               /* length to copy for repeats, bits to drop */
+    int ret;                    /* return code */
+#ifdef GUNZIP
+    unsigned char hbuf[4];      /* buffer for gzip header crc calculation */
+#endif
+    static const unsigned short order[19] = /* permutation of code lengths */
+        {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+    if (inflateStateCheck(strm) || strm->next_out == Z_NULL ||
+        (strm->next_in == Z_NULL && strm->avail_in != 0))
+        return Z_STREAM_ERROR;
+
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->mode == TYPE) state->mode = TYPEDO;      /* skip check */
+    LOAD();
+    in = have;
+    out = left;
+    ret = Z_OK;
+    for (;;)
+        switch (state->mode) {
+        case HEAD:
+            if (state->wrap == 0) {
+                state->mode = TYPEDO;
+                break;
+            }
+            NEEDBITS(16);
+#ifdef GUNZIP
+            if ((state->wrap & 2) && hold == 0x8b1f) {  /* gzip header */
+                if (state->wbits == 0)
+                    state->wbits = 15;
+                state->check = crc32(0L, Z_NULL, 0);
+                CRC2(state->check, hold);
+                INITBITS();
+                state->mode = FLAGS;
+                break;
+            }
+            state->flags = 0;           /* expect zlib header */
+            if (state->head != Z_NULL)
+                state->head->done = -1;
+            if (!(state->wrap & 1) ||   /* check if zlib header allowed */
+#else
+            if (
+#endif
+                ((BITS(8) << 8) + (hold >> 8)) % 31) {
+                strm->msg = (char *)"incorrect header check";
+                state->mode = BAD;
+                break;
+            }
+            if (BITS(4) != Z_DEFLATED) {
+                strm->msg = (char *)"unknown compression method";
+                state->mode = BAD;
+                break;
+            }
+            DROPBITS(4);
+            len = BITS(4) + 8;
+            if (state->wbits == 0)
+                state->wbits = len;
+            if (len > 15 || len > state->wbits) {
+                strm->msg = (char *)"invalid window size";
+                state->mode = BAD;
+                break;
+            }
+            state->dmax = 1U << len;
+            Tracev((stderr, "inflate:   zlib header ok\n"));
+            strm->adler = state->check = adler32(0L, Z_NULL, 0);
+            state->mode = hold & 0x200 ? DICTID : TYPE;
+            INITBITS();
+            break;
+#ifdef GUNZIP
+        case FLAGS:
+            NEEDBITS(16);
+            state->flags = (int)(hold);
+            if ((state->flags & 0xff) != Z_DEFLATED) {
+                strm->msg = (char *)"unknown compression method";
+                state->mode = BAD;
+                break;
+            }
+            if (state->flags & 0xe000) {
+                strm->msg = (char *)"unknown header flags set";
+                state->mode = BAD;
+                break;
+            }
+            if (state->head != Z_NULL)
+                state->head->text = (int)((hold >> 8) & 1);
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC2(state->check, hold);
+            INITBITS();
+            state->mode = TIME;
+        case TIME:
+            NEEDBITS(32);
+            if (state->head != Z_NULL)
+                state->head->time = hold;
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC4(state->check, hold);
+            INITBITS();
+            state->mode = OS;
+        case OS:
+            NEEDBITS(16);
+            if (state->head != Z_NULL) {
+                state->head->xflags = (int)(hold & 0xff);
+                state->head->os = (int)(hold >> 8);
+            }
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC2(state->check, hold);
+            INITBITS();
+            state->mode = EXLEN;
+        case EXLEN:
+            if (state->flags & 0x0400) {
+                NEEDBITS(16);
+                state->length = (unsigned)(hold);
+                if (state->head != Z_NULL)
+                    state->head->extra_len = (unsigned)hold;
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    CRC2(state->check, hold);
+                INITBITS();
+            }
+            else if (state->head != Z_NULL)
+                state->head->extra = Z_NULL;
+            state->mode = EXTRA;
+        case EXTRA:
+            if (state->flags & 0x0400) {
+                copy = state->length;
+                if (copy > have) copy = have;
+                if (copy) {
+                    if (state->head != Z_NULL &&
+                        state->head->extra != Z_NULL) {
+                        len = state->head->extra_len - state->length;
+                        zmemcpy(state->head->extra + len, next,
+                                len + copy > state->head->extra_max ?
+                                state->head->extra_max - len : copy);
+                    }
+                    if ((state->flags & 0x0200) && (state->wrap & 4))
+                        state->check = crc32(state->check, next, copy);
+                    have -= copy;
+                    next += copy;
+                    state->length -= copy;
+                }
+                if (state->length) goto inf_leave;
+            }
+            state->length = 0;
+            state->mode = NAME;
+        case NAME:
+            if (state->flags & 0x0800) {
+                if (have == 0) goto inf_leave;
+                copy = 0;
+                do {
+                    len = (unsigned)(next[copy++]);
+                    if (state->head != Z_NULL &&
+                            state->head->name != Z_NULL &&
+                            state->length < state->head->name_max)
+                        state->head->name[state->length++] = (Bytef)len;
+                } while (len && copy < have);
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    state->check = crc32(state->check, next, copy);
+                have -= copy;
+                next += copy;
+                if (len) goto inf_leave;
+            }
+            else if (state->head != Z_NULL)
+                state->head->name = Z_NULL;
+            state->length = 0;
+            state->mode = COMMENT;
+        case COMMENT:
+            if (state->flags & 0x1000) {
+                if (have == 0) goto inf_leave;
+                copy = 0;
+                do {
+                    len = (unsigned)(next[copy++]);
+                    if (state->head != Z_NULL &&
+                            state->head->comment != Z_NULL &&
+                            state->length < state->head->comm_max)
+                        state->head->comment[state->length++] = (Bytef)len;
+                } while (len && copy < have);
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    state->check = crc32(state->check, next, copy);
+                have -= copy;
+                next += copy;
+                if (len) goto inf_leave;
+            }
+            else if (state->head != Z_NULL)
+                state->head->comment = Z_NULL;
+            state->mode = HCRC;
+        case HCRC:
+            if (state->flags & 0x0200) {
+                NEEDBITS(16);
+                if ((state->wrap & 4) && hold != (state->check & 0xffff)) {
+                    strm->msg = (char *)"header crc mismatch";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+            }
+            if (state->head != Z_NULL) {
+                state->head->hcrc = (int)((state->flags >> 9) & 1);
+                state->head->done = 1;
+            }
+            strm->adler = state->check = crc32(0L, Z_NULL, 0);
+            state->mode = TYPE;
+            break;
+#endif
+        case DICTID:
+            NEEDBITS(32);
+            strm->adler = state->check = ZSWAP32(hold);
+            INITBITS();
+            state->mode = DICT;
+        case DICT:
+            if (state->havedict == 0) {
+                RESTORE();
+                return Z_NEED_DICT;
+            }
+            strm->adler = state->check = adler32(0L, Z_NULL, 0);
+            state->mode = TYPE;
+        case TYPE:
+            if (flush == Z_BLOCK || flush == Z_TREES) goto inf_leave;
+        case TYPEDO:
+            if (state->last) {
+                BYTEBITS();
+                state->mode = CHECK;
+                break;
+            }
+            NEEDBITS(3);
+            state->last = BITS(1);
+            DROPBITS(1);
+            switch (BITS(2)) {
+            case 0:                             /* stored block */
+                Tracev((stderr, "inflate:     stored block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = STORED;
+                break;
+            case 1:                             /* fixed block */
+                fixedtables(state);
+                Tracev((stderr, "inflate:     fixed codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = LEN_;             /* decode codes */
+                if (flush == Z_TREES) {
+                    DROPBITS(2);
+                    goto inf_leave;
+                }
+                break;
+            case 2:                             /* dynamic block */
+                Tracev((stderr, "inflate:     dynamic codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = TABLE;
+                break;
+            case 3:
+                strm->msg = (char *)"invalid block type";
+                state->mode = BAD;
+            }
+            DROPBITS(2);
+            break;
+        case STORED:
+            BYTEBITS();                         /* go to byte boundary */
+            NEEDBITS(32);
+            if ((hold & 0xffff) != ((hold >> 16) ^ 0xffff)) {
+                strm->msg = (char *)"invalid stored block lengths";
+                state->mode = BAD;
+                break;
+            }
+            state->length = (unsigned)hold & 0xffff;
+            Tracev((stderr, "inflate:       stored length %u\n",
+                    state->length));
+            INITBITS();
+            state->mode = COPY_;
+            if (flush == Z_TREES) goto inf_leave;
+        case COPY_:
+            state->mode = COPY;
+        case COPY:
+            copy = state->length;
+            if (copy) {
+                if (copy > have) copy = have;
+                if (copy > left) copy = left;
+                if (copy == 0) goto inf_leave;
+                zmemcpy(put, next, copy);
+                have -= copy;
+                next += copy;
+                left -= copy;
+                put += copy;
+                state->length -= copy;
+                break;
+            }
+            Tracev((stderr, "inflate:       stored end\n"));
+            state->mode = TYPE;
+            break;
+        case TABLE:
+            NEEDBITS(14);
+            state->nlen = BITS(5) + 257;
+            DROPBITS(5);
+            state->ndist = BITS(5) + 1;
+            DROPBITS(5);
+            state->ncode = BITS(4) + 4;
+            DROPBITS(4);
+#ifndef PKZIP_BUG_WORKAROUND
+            if (state->nlen > 286 || state->ndist > 30) {
+                strm->msg = (char *)"too many length or distance symbols";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracev((stderr, "inflate:       table sizes ok\n"));
+            state->have = 0;
+            state->mode = LENLENS;
+        case LENLENS:
+            while (state->have < state->ncode) {
+                NEEDBITS(3);
+                state->lens[order[state->have++]] = (unsigned short)BITS(3);
+                DROPBITS(3);
+            }
+            while (state->have < 19)
+                state->lens[order[state->have++]] = 0;
+            state->next = state->codes;
+            state->lencode = (const code FAR *)(state->next);
+            state->lenbits = 7;
+            ret = inflate_table(CODES, state->lens, 19, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid code lengths set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       code lengths ok\n"));
+            state->have = 0;
+            state->mode = CODELENS;
+        case CODELENS:
+            while (state->have < state->nlen + state->ndist) {
+                for (;;) {
+                    here = state->lencode[BITS(state->lenbits)];
+                    if ((unsigned)(here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                if (here.val < 16) {
+                    DROPBITS(here.bits);
+                    state->lens[state->have++] = here.val;
+                }
+                else {
+                    if (here.val == 16) {
+                        NEEDBITS(here.bits + 2);
+                        DROPBITS(here.bits);
+                        if (state->have == 0) {
+                            strm->msg = (char *)"invalid bit length repeat";
+                            state->mode = BAD;
+                            break;
+                        }
+                        len = state->lens[state->have - 1];
+                        copy = 3 + BITS(2);
+                        DROPBITS(2);
+                    }
+                    else if (here.val == 17) {
+                        NEEDBITS(here.bits + 3);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 3 + BITS(3);
+                        DROPBITS(3);
+                    }
+                    else {
+                        NEEDBITS(here.bits + 7);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 11 + BITS(7);
+                        DROPBITS(7);
+                    }
+                    if (state->have + copy > state->nlen + state->ndist) {
+                        strm->msg = (char *)"invalid bit length repeat";
+                        state->mode = BAD;
+                        break;
+                    }
+                    while (copy--)
+                        state->lens[state->have++] = (unsigned short)len;
+                }
+            }
+
+            /* handle error breaks in while */
+            if (state->mode == BAD) break;
+
+            /* check for end-of-block code (better have one) */
+            if (state->lens[256] == 0) {
+                strm->msg = (char *)"invalid code -- missing end-of-block";
+                state->mode = BAD;
+                break;
+            }
+
+            /* build code tables -- note: do not change the lenbits or distbits
+               values here (9 and 6) without reading the comments in inftrees.h
+               concerning the ENOUGH constants, which depend on those values */
+            state->next = state->codes;
+            state->lencode = (const code FAR *)(state->next);
+            state->lenbits = 9;
+            ret = inflate_table(LENS, state->lens, state->nlen, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid literal/lengths set";
+                state->mode = BAD;
+                break;
+            }
+            state->distcode = (const code FAR *)(state->next);
+            state->distbits = 6;
+            ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist,
+                            &(state->next), &(state->distbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid distances set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       codes ok\n"));
+            state->mode = LEN_;
+            if (flush == Z_TREES) goto inf_leave;
+        case LEN_:
+            state->mode = LEN;
+        case LEN:
+            if (have >= 6 && left >= 258) {
+                RESTORE();
+                inflate_fast(strm, out);
+                LOAD();
+                if (state->mode == TYPE)
+                    state->back = -1;
+                break;
+            }
+            state->back = 0;
+            for (;;) {
+                here = state->lencode[BITS(state->lenbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if (here.op && (here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->lencode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+                state->back += last.bits;
+            }
+            DROPBITS(here.bits);
+            state->back += here.bits;
+            state->length = (unsigned)here.val;
+            if ((int)(here.op) == 0) {
+                Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                        "inflate:         literal '%c'\n" :
+                        "inflate:         literal 0x%02x\n", here.val));
+                state->mode = LIT;
+                break;
+            }
+            if (here.op & 32) {
+                Tracevv((stderr, "inflate:         end of block\n"));
+                state->back = -1;
+                state->mode = TYPE;
+                break;
+            }
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid literal/length code";
+                state->mode = BAD;
+                break;
+            }
+            state->extra = (unsigned)(here.op) & 15;
+            state->mode = LENEXT;
+        case LENEXT:
+            if (state->extra) {
+                NEEDBITS(state->extra);
+                state->length += BITS(state->extra);
+                DROPBITS(state->extra);
+                state->back += state->extra;
+            }
+            Tracevv((stderr, "inflate:         length %u\n", state->length));
+            state->was = state->length;
+            state->mode = DIST;
+        case DIST:
+            for (;;) {
+                here = state->distcode[BITS(state->distbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if ((here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->distcode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+                state->back += last.bits;
+            }
+            DROPBITS(here.bits);
+            state->back += here.bits;
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+            state->offset = (unsigned)here.val;
+            state->extra = (unsigned)(here.op) & 15;
+            state->mode = DISTEXT;
+        case DISTEXT:
+            if (state->extra) {
+                NEEDBITS(state->extra);
+                state->offset += BITS(state->extra);
+                DROPBITS(state->extra);
+                state->back += state->extra;
+            }
+#ifdef INFLATE_STRICT
+            if (state->offset > state->dmax) {
+                strm->msg = (char *)"invalid distance too far back";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracevv((stderr, "inflate:         distance %u\n", state->offset));
+            state->mode = MATCH;
+        case MATCH:
+            if (left == 0) goto inf_leave;
+            copy = out - left;
+            if (state->offset > copy) {         /* copy from window */
+                copy = state->offset - copy;
+                if (copy > state->whave) {
+                    if (state->sane) {
+                        strm->msg = (char *)"invalid distance too far back";
+                        state->mode = BAD;
+                        break;
+                    }
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                    Trace((stderr, "inflate.c too far\n"));
+                    copy -= state->whave;
+                    if (copy > state->length) copy = state->length;
+                    if (copy > left) copy = left;
+                    left -= copy;
+                    state->length -= copy;
+                    do {
+                        *put++ = 0;
+                    } while (--copy);
+                    if (state->length == 0) state->mode = LEN;
+                    break;
+#endif
+                }
+                if (copy > state->wnext) {
+                    copy -= state->wnext;
+                    from = state->window + (state->wsize - copy);
+                }
+                else
+                    from = state->window + (state->wnext - copy);
+                if (copy > state->length) copy = state->length;
+            }
+            else {                              /* copy from output */
+                from = put - state->offset;
+                copy = state->length;
+            }
+            if (copy > left) copy = left;
+            left -= copy;
+            state->length -= copy;
+            do {
+                *put++ = *from++;
+            } while (--copy);
+            if (state->length == 0) state->mode = LEN;
+            break;
+        case LIT:
+            if (left == 0) goto inf_leave;
+            *put++ = (unsigned char)(state->length);
+            left--;
+            state->mode = LEN;
+            break;
+        case CHECK:
+            if (state->wrap) {
+                NEEDBITS(32);
+                out -= left;
+                strm->total_out += out;
+                state->total += out;
+                if ((state->wrap & 4) && out)
+                    strm->adler = state->check =
+                        UPDATE(state->check, put - out, out);
+                out = left;
+                if ((state->wrap & 4) && (
+#ifdef GUNZIP
+                     state->flags ? hold :
+#endif
+                     ZSWAP32(hold)) != state->check) {
+                    strm->msg = (char *)"incorrect data check";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+                Tracev((stderr, "inflate:   check matches trailer\n"));
+            }
+#ifdef GUNZIP
+            state->mode = LENGTH;
+        case LENGTH:
+            if (state->wrap && state->flags) {
+                NEEDBITS(32);
+                if (hold != (state->total & 0xffffffffUL)) {
+                    strm->msg = (char *)"incorrect length check";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+                Tracev((stderr, "inflate:   length matches trailer\n"));
+            }
+#endif
+            state->mode = DONE;
+        case DONE:
+            ret = Z_STREAM_END;
+            goto inf_leave;
+        case BAD:
+            ret = Z_DATA_ERROR;
+            goto inf_leave;
+        case MEM:
+            return Z_MEM_ERROR;
+        case SYNC:
+        default:
+            return Z_STREAM_ERROR;
+        }
+
+    /*
+       Return from inflate(), updating the total counts and the check value.
+       If there was no progress during the inflate() call, return a buffer
+       error.  Call updatewindow() to create and/or update the window state.
+       Note: a memory error from inflate() is non-recoverable.
+     */
+  inf_leave:
+    RESTORE();
+    if (state->wsize || (out != strm->avail_out && state->mode < BAD &&
+            (state->mode < CHECK || flush != Z_FINISH)))
+        if (updatewindow(strm, strm->next_out, out - strm->avail_out)) {
+            state->mode = MEM;
+            return Z_MEM_ERROR;
+        }
+    in -= strm->avail_in;
+    out -= strm->avail_out;
+    strm->total_in += in;
+    strm->total_out += out;
+    state->total += out;
+    if ((state->wrap & 4) && out)
+        strm->adler = state->check =
+            UPDATE(state->check, strm->next_out - out, out);
+    strm->data_type = (int)state->bits + (state->last ? 64 : 0) +
+                      (state->mode == TYPE ? 128 : 0) +
+                      (state->mode == LEN_ || state->mode == COPY_ ? 256 : 0);
+    if (((in == 0 && out == 0) || flush == Z_FINISH) && ret == Z_OK)
+        ret = Z_BUF_ERROR;
+    return ret;
+}
+
+int ZEXPORT inflateEnd(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (inflateStateCheck(strm))
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->window != Z_NULL) ZFREE(strm, state->window);
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+    Tracev((stderr, "inflate: end\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateGetDictionary(strm, dictionary, dictLength)
+z_streamp strm;
+Bytef *dictionary;
+uInt *dictLength;
+{
+    struct inflate_state FAR *state;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* copy dictionary */
+    if (state->whave && dictionary != Z_NULL) {
+        zmemcpy(dictionary, state->window + state->wnext,
+                state->whave - state->wnext);
+        zmemcpy(dictionary + state->whave - state->wnext,
+                state->window, state->wnext);
+    }
+    if (dictLength != Z_NULL)
+        *dictLength = state->whave;
+    return Z_OK;
+}
+
+int ZEXPORT inflateSetDictionary(strm, dictionary, dictLength)
+z_streamp strm;
+const Bytef *dictionary;
+uInt dictLength;
+{
+    struct inflate_state FAR *state;
+    unsigned long dictid;
+    int ret;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->wrap != 0 && state->mode != DICT)
+        return Z_STREAM_ERROR;
+
+    /* check for correct dictionary identifier */
+    if (state->mode == DICT) {
+        dictid = adler32(0L, Z_NULL, 0);
+        dictid = adler32(dictid, dictionary, dictLength);
+        if (dictid != state->check)
+            return Z_DATA_ERROR;
+    }
+
+    /* copy dictionary to window using updatewindow(), which will amend the
+       existing dictionary if appropriate */
+    ret = updatewindow(strm, dictionary + dictLength, dictLength);
+    if (ret) {
+        state->mode = MEM;
+        return Z_MEM_ERROR;
+    }
+    state->havedict = 1;
+    Tracev((stderr, "inflate:   dictionary set\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateGetHeader(strm, head)
+z_streamp strm;
+gz_headerp head;
+{
+    struct inflate_state FAR *state;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if ((state->wrap & 2) == 0) return Z_STREAM_ERROR;
+
+    /* save header structure */
+    state->head = head;
+    head->done = 0;
+    return Z_OK;
+}
+
+/*
+   Search buf[0..len-1] for the pattern: 0, 0, 0xff, 0xff.  Return when found
+   or when out of input.  When called, *have is the number of pattern bytes
+   found in order so far, in 0..3.  On return *have is updated to the new
+   state.  If on return *have equals four, then the pattern was found and the
+   return value is how many bytes were read including the last byte of the
+   pattern.  If *have is less than four, then the pattern has not been found
+   yet and the return value is len.  In the latter case, syncsearch() can be
+   called again with more data and the *have state.  *have is initialized to
+   zero for the first call.
+ */
+local unsigned syncsearch(have, buf, len)
+unsigned FAR *have;
+const unsigned char FAR *buf;
+unsigned len;
+{
+    unsigned got;
+    unsigned next;
+
+    got = *have;
+    next = 0;
+    while (next < len && got < 4) {
+        if ((int)(buf[next]) == (got < 2 ? 0 : 0xff))
+            got++;
+        else if (buf[next])
+            got = 0;
+        else
+            got = 4 - got;
+        next++;
+    }
+    *have = got;
+    return next;
+}
+
+int ZEXPORT inflateSync(strm)
+z_streamp strm;
+{
+    unsigned len;               /* number of bytes to look at or looked at */
+    unsigned long in, out;      /* temporary to save total_in and total_out */
+    unsigned char buf[4];       /* to restore bit buffer to byte string */
+    struct inflate_state FAR *state;
+
+    /* check parameters */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (strm->avail_in == 0 && state->bits < 8) return Z_BUF_ERROR;
+
+    /* if first time, start search in bit buffer */
+    if (state->mode != SYNC) {
+        state->mode = SYNC;
+        state->hold <<= state->bits & 7;
+        state->bits -= state->bits & 7;
+        len = 0;
+        while (state->bits >= 8) {
+            buf[len++] = (unsigned char)(state->hold);
+            state->hold >>= 8;
+            state->bits -= 8;
+        }
+        state->have = 0;
+        syncsearch(&(state->have), buf, len);
+    }
+
+    /* search available input */
+    len = syncsearch(&(state->have), strm->next_in, strm->avail_in);
+    strm->avail_in -= len;
+    strm->next_in += len;
+    strm->total_in += len;
+
+    /* return no joy or set up to restart inflate() on a new block */
+    if (state->have != 4) return Z_DATA_ERROR;
+    in = strm->total_in;  out = strm->total_out;
+    inflateReset(strm);
+    strm->total_in = in;  strm->total_out = out;
+    state->mode = TYPE;
+    return Z_OK;
+}
+
+/*
+   Returns true if inflate is currently at the end of a block generated by
+   Z_SYNC_FLUSH or Z_FULL_FLUSH. This function is used by one PPP
+   implementation to provide an additional safety check. PPP uses
+   Z_SYNC_FLUSH but removes the length bytes of the resulting empty stored
+   block. When decompressing, PPP checks that at the end of input packet,
+   inflate is waiting for these length bytes.
+ */
+int ZEXPORT inflateSyncPoint(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    return state->mode == STORED && state->bits == 0;
+}
+
+int ZEXPORT inflateCopy(dest, source)
+z_streamp dest;
+z_streamp source;
+{
+    struct inflate_state FAR *state;
+    struct inflate_state FAR *copy;
+    unsigned char FAR *window;
+    unsigned wsize;
+
+    /* check input */
+    if (inflateStateCheck(source) || dest == Z_NULL)
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)source->state;
+
+    /* allocate space */
+    copy = (struct inflate_state FAR *)
+           ZALLOC(source, 1, sizeof(struct inflate_state));
+    if (copy == Z_NULL) return Z_MEM_ERROR;
+    window = Z_NULL;
+    if (state->window != Z_NULL) {
+        window = (unsigned char FAR *)
+                 ZALLOC(source, 1U << state->wbits, sizeof(unsigned char));
+        if (window == Z_NULL) {
+            ZFREE(source, copy);
+            return Z_MEM_ERROR;
+        }
+    }
+
+    /* copy state */
+    zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
+    zmemcpy((voidpf)copy, (voidpf)state, sizeof(struct inflate_state));
+    copy->strm = dest;
+    if (state->lencode >= state->codes &&
+        state->lencode <= state->codes + ENOUGH - 1) {
+        copy->lencode = copy->codes + (state->lencode - state->codes);
+        copy->distcode = copy->codes + (state->distcode - state->codes);
+    }
+    copy->next = copy->codes + (state->next - state->codes);
+    if (window != Z_NULL) {
+        wsize = 1U << state->wbits;
+        zmemcpy(window, state->window, wsize);
+    }
+    copy->window = window;
+    dest->state = (struct internal_state FAR *)copy;
+    return Z_OK;
+}
+
+int ZEXPORT inflateUndermine(strm, subvert)
+z_streamp strm;
+int subvert;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+    state->sane = !subvert;
+    return Z_OK;
+#else
+    (void)subvert;
+    state->sane = 1;
+    return Z_DATA_ERROR;
+#endif
+}
+
+int ZEXPORT inflateValidate(strm, check)
+z_streamp strm;
+int check;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (check)
+        state->wrap |= 4;
+    else
+        state->wrap &= ~4;
+    return Z_OK;
+}
+
+long ZEXPORT inflateMark(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm))
+        return -(1L << 16);
+    state = (struct inflate_state FAR *)strm->state;
+    return (long)(((unsigned long)((long)state->back)) << 16) +
+        (state->mode == COPY ? state->length :
+            (state->mode == MATCH ? state->was - state->length : 0));
+}
+
+unsigned long ZEXPORT inflateCodesUsed(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (inflateStateCheck(strm)) return (unsigned long)-1;
+    state = (struct inflate_state FAR *)strm->state;
+    return (unsigned long)(state->next - state->codes);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inflate.h glmark2-2021.02/android/jni/src/zlib/inflate.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inflate.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inflate.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,125 @@
+/* inflate.h -- internal inflate state definition
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* define NO_GZIP when compiling if you want to disable gzip header and
+   trailer decoding by inflate().  NO_GZIP would be used to avoid linking in
+   the crc code when it is not needed.  For shared libraries, gzip decoding
+   should be left enabled. */
+#ifndef NO_GZIP
+#  define GUNZIP
+#endif
+
+/* Possible inflate modes between inflate() calls */
+typedef enum {
+    HEAD = 16180,   /* i: waiting for magic header */
+    FLAGS,      /* i: waiting for method and flags (gzip) */
+    TIME,       /* i: waiting for modification time (gzip) */
+    OS,         /* i: waiting for extra flags and operating system (gzip) */
+    EXLEN,      /* i: waiting for extra length (gzip) */
+    EXTRA,      /* i: waiting for extra bytes (gzip) */
+    NAME,       /* i: waiting for end of file name (gzip) */
+    COMMENT,    /* i: waiting for end of comment (gzip) */
+    HCRC,       /* i: waiting for header crc (gzip) */
+    DICTID,     /* i: waiting for dictionary check value */
+    DICT,       /* waiting for inflateSetDictionary() call */
+        TYPE,       /* i: waiting for type bits, including last-flag bit */
+        TYPEDO,     /* i: same, but skip check to exit inflate on new block */
+        STORED,     /* i: waiting for stored size (length and complement) */
+        COPY_,      /* i/o: same as COPY below, but only first time in */
+        COPY,       /* i/o: waiting for input or output to copy stored block */
+        TABLE,      /* i: waiting for dynamic block table lengths */
+        LENLENS,    /* i: waiting for code length code lengths */
+        CODELENS,   /* i: waiting for length/lit and distance code lengths */
+            LEN_,       /* i: same as LEN below, but only first time in */
+            LEN,        /* i: waiting for length/lit/eob code */
+            LENEXT,     /* i: waiting for length extra bits */
+            DIST,       /* i: waiting for distance code */
+            DISTEXT,    /* i: waiting for distance extra bits */
+            MATCH,      /* o: waiting for output space to copy string */
+            LIT,        /* o: waiting for output space to write literal */
+    CHECK,      /* i: waiting for 32-bit check value */
+    LENGTH,     /* i: waiting for 32-bit length (gzip) */
+    DONE,       /* finished check, done -- remain here until reset */
+    BAD,        /* got a data error -- remain here until reset */
+    MEM,        /* got an inflate() memory error -- remain here until reset */
+    SYNC        /* looking for synchronization bytes to restart inflate() */
+} inflate_mode;
+
+/*
+    State transitions between above modes -
+
+    (most modes can go to BAD or MEM on error -- not shown for clarity)
+
+    Process header:
+        HEAD -> (gzip) or (zlib) or (raw)
+        (gzip) -> FLAGS -> TIME -> OS -> EXLEN -> EXTRA -> NAME -> COMMENT ->
+                  HCRC -> TYPE
+        (zlib) -> DICTID or TYPE
+        DICTID -> DICT -> TYPE
+        (raw) -> TYPEDO
+    Read deflate blocks:
+            TYPE -> TYPEDO -> STORED or TABLE or LEN_ or CHECK
+            STORED -> COPY_ -> COPY -> TYPE
+            TABLE -> LENLENS -> CODELENS -> LEN_
+            LEN_ -> LEN
+    Read deflate codes in fixed or dynamic block:
+                LEN -> LENEXT or LIT or TYPE
+                LENEXT -> DIST -> DISTEXT -> MATCH -> LEN
+                LIT -> LEN
+    Process trailer:
+        CHECK -> LENGTH -> DONE
+ */
+
+/* State maintained between inflate() calls -- approximately 7K bytes, not
+   including the allocated sliding window, which is up to 32K bytes. */
+struct inflate_state {
+    z_streamp strm;             /* pointer back to this zlib stream */
+    inflate_mode mode;          /* current inflate mode */
+    int last;                   /* true if processing last block */
+    int wrap;                   /* bit 0 true for zlib, bit 1 true for gzip,
+                                   bit 2 true to validate check value */
+    int havedict;               /* true if dictionary provided */
+    int flags;                  /* gzip header method and flags (0 if zlib) */
+    unsigned dmax;              /* zlib header max distance (INFLATE_STRICT) */
+    unsigned long check;        /* protected copy of check value */
+    unsigned long total;        /* protected copy of output count */
+    gz_headerp head;            /* where to save gzip header information */
+        /* sliding window */
+    unsigned wbits;             /* log base 2 of requested window size */
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned wnext;             /* window write index */
+    unsigned char FAR *window;  /* allocated sliding window, if needed */
+        /* bit accumulator */
+    unsigned long hold;         /* input bit accumulator */
+    unsigned bits;              /* number of bits in "in" */
+        /* for string and stored block copying */
+    unsigned length;            /* literal or length of data to copy */
+    unsigned offset;            /* distance back to copy string from */
+        /* for table and code decoding */
+    unsigned extra;             /* extra bits needed */
+        /* fixed and dynamic code tables */
+    code const FAR *lencode;    /* starting table for length/literal codes */
+    code const FAR *distcode;   /* starting table for distance codes */
+    unsigned lenbits;           /* index bits for lencode */
+    unsigned distbits;          /* index bits for distcode */
+        /* dynamic table building */
+    unsigned ncode;             /* number of code length code lengths */
+    unsigned nlen;              /* number of length code lengths */
+    unsigned ndist;             /* number of distance code lengths */
+    unsigned have;              /* number of code lengths in lens[] */
+    code FAR *next;             /* next available space in codes[] */
+    unsigned short lens[320];   /* temporary storage for code lengths */
+    unsigned short work[288];   /* work area for code table building */
+    code codes[ENOUGH];         /* space for code tables */
+    int sane;                   /* if false, allow invalid distance too far */
+    int back;                   /* bits back of last unprocessed length/lit */
+    unsigned was;               /* initial length of match */
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inftrees.c glmark2-2021.02/android/jni/src/zlib/inftrees.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inftrees.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inftrees.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,304 @@
+/* inftrees.c -- generate Huffman trees for efficient decoding
+ * Copyright (C) 1995-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+
+#define MAXBITS 15
+
+const char inflate_copyright[] =
+   " inflate 1.2.11 Copyright 1995-2017 Mark Adler ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+
+/*
+   Build a set of tables to decode the provided canonical Huffman code.
+   The code lengths are lens[0..codes-1].  The result starts at *table,
+   whose indices are 0..2^bits-1.  work is a writable array of at least
+   lens shorts, which is used as a work area.  type is the type of code
+   to be generated, CODES, LENS, or DISTS.  On return, zero is success,
+   -1 is an invalid code, and +1 means that ENOUGH isn't enough.  table
+   on return points to the next available entry's address.  bits is the
+   requested root table index bits, and on return it is the actual root
+   table index bits.  It will differ if the request is greater than the
+   longest code or if it is less than the shortest code.
+ */
+int ZLIB_INTERNAL inflate_table(type, lens, codes, table, bits, work)
+codetype type;
+unsigned short FAR *lens;
+unsigned codes;
+code FAR * FAR *table;
+unsigned FAR *bits;
+unsigned short FAR *work;
+{
+    unsigned len;               /* a code's length in bits */
+    unsigned sym;               /* index of code symbols */
+    unsigned min, max;          /* minimum and maximum code lengths */
+    unsigned root;              /* number of index bits for root table */
+    unsigned curr;              /* number of index bits for current table */
+    unsigned drop;              /* code bits to drop for sub-table */
+    int left;                   /* number of prefix codes available */
+    unsigned used;              /* code entries in table used */
+    unsigned huff;              /* Huffman code */
+    unsigned incr;              /* for incrementing code, index */
+    unsigned fill;              /* index for replicating entries */
+    unsigned low;               /* low bits for current root entry */
+    unsigned mask;              /* mask for low root bits */
+    code here;                  /* table entry for duplication */
+    code FAR *next;             /* next available space in table */
+    const unsigned short FAR *base;     /* base value table to use */
+    const unsigned short FAR *extra;    /* extra bits table to use */
+    unsigned match;             /* use base and extra for symbol >= match */
+    unsigned short count[MAXBITS+1];    /* number of codes of each length */
+    unsigned short offs[MAXBITS+1];     /* offsets in table for each length */
+    static const unsigned short lbase[31] = { /* Length codes 257..285 base */
+        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+        35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
+    static const unsigned short lext[31] = { /* Length codes 257..285 extra */
+        16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
+        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 77, 202};
+    static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
+        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+        257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+        8193, 12289, 16385, 24577, 0, 0};
+    static const unsigned short dext[32] = { /* Distance codes 0..29 extra */
+        16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22,
+        23, 23, 24, 24, 25, 25, 26, 26, 27, 27,
+        28, 28, 29, 29, 64, 64};
+
+    /*
+       Process a set of code lengths to create a canonical Huffman code.  The
+       code lengths are lens[0..codes-1].  Each length corresponds to the
+       symbols 0..codes-1.  The Huffman code is generated by first sorting the
+       symbols by length from short to long, and retaining the symbol order
+       for codes with equal lengths.  Then the code starts with all zero bits
+       for the first code of the shortest length, and the codes are integer
+       increments for the same length, and zeros are appended as the length
+       increases.  For the deflate format, these bits are stored backwards
+       from their more natural integer increment ordering, and so when the
+       decoding tables are built in the large loop below, the integer codes
+       are incremented backwards.
+
+       This routine assumes, but does not check, that all of the entries in
+       lens[] are in the range 0..MAXBITS.  The caller must assure this.
+       1..MAXBITS is interpreted as that code length.  zero means that that
+       symbol does not occur in this code.
+
+       The codes are sorted by computing a count of codes for each length,
+       creating from that a table of starting indices for each length in the
+       sorted table, and then entering the symbols in order in the sorted
+       table.  The sorted table is work[], with that space being provided by
+       the caller.
+
+       The length counts are used for other purposes as well, i.e. finding
+       the minimum and maximum length codes, determining if there are any
+       codes at all, checking for a valid set of lengths, and looking ahead
+       at length counts to determine sub-table sizes when building the
+       decoding tables.
+     */
+
+    /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
+    for (len = 0; len <= MAXBITS; len++)
+        count[len] = 0;
+    for (sym = 0; sym < codes; sym++)
+        count[lens[sym]]++;
+
+    /* bound code lengths, force root to be within code lengths */
+    root = *bits;
+    for (max = MAXBITS; max >= 1; max--)
+        if (count[max] != 0) break;
+    if (root > max) root = max;
+    if (max == 0) {                     /* no symbols to code at all */
+        here.op = (unsigned char)64;    /* invalid code marker */
+        here.bits = (unsigned char)1;
+        here.val = (unsigned short)0;
+        *(*table)++ = here;             /* make a table to force an error */
+        *(*table)++ = here;
+        *bits = 1;
+        return 0;     /* no symbols, but wait for decoding to report error */
+    }
+    for (min = 1; min < max; min++)
+        if (count[min] != 0) break;
+    if (root < min) root = min;
+
+    /* check for an over-subscribed or incomplete set of lengths */
+    left = 1;
+    for (len = 1; len <= MAXBITS; len++) {
+        left <<= 1;
+        left -= count[len];
+        if (left < 0) return -1;        /* over-subscribed */
+    }
+    if (left > 0 && (type == CODES || max != 1))
+        return -1;                      /* incomplete set */
+
+    /* generate offsets into symbol table for each length for sorting */
+    offs[1] = 0;
+    for (len = 1; len < MAXBITS; len++)
+        offs[len + 1] = offs[len] + count[len];
+
+    /* sort symbols by length, by symbol order within each length */
+    for (sym = 0; sym < codes; sym++)
+        if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym;
+
+    /*
+       Create and fill in decoding tables.  In this loop, the table being
+       filled is at next and has curr index bits.  The code being used is huff
+       with length len.  That code is converted to an index by dropping drop
+       bits off of the bottom.  For codes where len is less than drop + curr,
+       those top drop + curr - len bits are incremented through all values to
+       fill the table with replicated entries.
+
+       root is the number of index bits for the root table.  When len exceeds
+       root, sub-tables are created pointed to by the root entry with an index
+       of the low root bits of huff.  This is saved in low to check for when a
+       new sub-table should be started.  drop is zero when the root table is
+       being filled, and drop is root when sub-tables are being filled.
+
+       When a new sub-table is needed, it is necessary to look ahead in the
+       code lengths to determine what size sub-table is needed.  The length
+       counts are used for this, and so count[] is decremented as codes are
+       entered in the tables.
+
+       used keeps track of how many table entries have been allocated from the
+       provided *table space.  It is checked for LENS and DIST tables against
+       the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
+       the initial root table size constants.  See the comments in inftrees.h
+       for more information.
+
+       sym increments through all symbols, and the loop terminates when
+       all codes of length max, i.e. all codes, have been processed.  This
+       routine permits incomplete codes, so another loop after this one fills
+       in the rest of the decoding tables with invalid code markers.
+     */
+
+    /* set up for code type */
+    switch (type) {
+    case CODES:
+        base = extra = work;    /* dummy value--not used */
+        match = 20;
+        break;
+    case LENS:
+        base = lbase;
+        extra = lext;
+        match = 257;
+        break;
+    default:    /* DISTS */
+        base = dbase;
+        extra = dext;
+        match = 0;
+    }
+
+    /* initialize state for loop */
+    huff = 0;                   /* starting code */
+    sym = 0;                    /* starting code symbol */
+    len = min;                  /* starting code length */
+    next = *table;              /* current table to fill in */
+    curr = root;                /* current table index bits */
+    drop = 0;                   /* current bits to drop from code for index */
+    low = (unsigned)(-1);       /* trigger new sub-table when len > root */
+    used = 1U << root;          /* use root table entries */
+    mask = used - 1;            /* mask for comparing low */
+
+    /* check available table space */
+    if ((type == LENS && used > ENOUGH_LENS) ||
+        (type == DISTS && used > ENOUGH_DISTS))
+        return 1;
+
+    /* process all codes and make table entries */
+    for (;;) {
+        /* create table entry */
+        here.bits = (unsigned char)(len - drop);
+        if (work[sym] + 1U < match) {
+            here.op = (unsigned char)0;
+            here.val = work[sym];
+        }
+        else if (work[sym] >= match) {
+            here.op = (unsigned char)(extra[work[sym] - match]);
+            here.val = base[work[sym] - match];
+        }
+        else {
+            here.op = (unsigned char)(32 + 64);         /* end of block */
+            here.val = 0;
+        }
+
+        /* replicate for those indices with low len bits equal to huff */
+        incr = 1U << (len - drop);
+        fill = 1U << curr;
+        min = fill;                 /* save offset to next table */
+        do {
+            fill -= incr;
+            next[(huff >> drop) + fill] = here;
+        } while (fill != 0);
+
+        /* backwards increment the len-bit code huff */
+        incr = 1U << (len - 1);
+        while (huff & incr)
+            incr >>= 1;
+        if (incr != 0) {
+            huff &= incr - 1;
+            huff += incr;
+        }
+        else
+            huff = 0;
+
+        /* go to next symbol, update count, len */
+        sym++;
+        if (--(count[len]) == 0) {
+            if (len == max) break;
+            len = lens[work[sym]];
+        }
+
+        /* create new sub-table if needed */
+        if (len > root && (huff & mask) != low) {
+            /* if first time, transition to sub-tables */
+            if (drop == 0)
+                drop = root;
+
+            /* increment past last table */
+            next += min;            /* here min is 1 << curr */
+
+            /* determine length of next table */
+            curr = len - drop;
+            left = (int)(1 << curr);
+            while (curr + drop < max) {
+                left -= count[curr + drop];
+                if (left <= 0) break;
+                curr++;
+                left <<= 1;
+            }
+
+            /* check for enough space */
+            used += 1U << curr;
+            if ((type == LENS && used > ENOUGH_LENS) ||
+                (type == DISTS && used > ENOUGH_DISTS))
+                return 1;
+
+            /* point entry in root table to sub-table */
+            low = huff & mask;
+            (*table)[low].op = (unsigned char)curr;
+            (*table)[low].bits = (unsigned char)root;
+            (*table)[low].val = (unsigned short)(next - *table);
+        }
+    }
+
+    /* fill in remaining table entry if code is incomplete (guaranteed to have
+       at most one remaining entry, since if the code is incomplete, the
+       maximum code length that was allowed to get this far is one bit) */
+    if (huff != 0) {
+        here.op = (unsigned char)64;            /* invalid code marker */
+        here.bits = (unsigned char)(len - drop);
+        here.val = (unsigned short)0;
+        next[huff] = here;
+    }
+
+    /* set return parameters */
+    *table += used;
+    *bits = root;
+    return 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inftrees.h glmark2-2021.02/android/jni/src/zlib/inftrees.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/inftrees.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/inftrees.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,62 @@
+/* inftrees.h -- header to use inftrees.c
+ * Copyright (C) 1995-2005, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* Structure for decoding tables.  Each entry provides either the
+   information needed to do the operation requested by the code that
+   indexed that table entry, or it provides a pointer to another
+   table that indexes more bits of the code.  op indicates whether
+   the entry is a pointer to another table, a literal, a length or
+   distance, an end-of-block, or an invalid code.  For a table
+   pointer, the low four bits of op is the number of index bits of
+   that table.  For a length or distance, the low four bits of op
+   is the number of extra bits to get after the code.  bits is
+   the number of bits in this code or part of the code to drop off
+   of the bit buffer.  val is the actual byte to output in the case
+   of a literal, the base length or distance, or the offset from
+   the current table to the next table.  Each entry is four bytes. */
+typedef struct {
+    unsigned char op;           /* operation, extra bits, table bits */
+    unsigned char bits;         /* bits in this part of the code */
+    unsigned short val;         /* offset in table or code value */
+} code;
+
+/* op values as set by inflate_table():
+    00000000 - literal
+    0000tttt - table link, tttt != 0 is the number of table index bits
+    0001eeee - length or distance, eeee is the number of extra bits
+    01100000 - end of block
+    01000000 - invalid code
+ */
+
+/* Maximum size of the dynamic table.  The maximum number of code structures is
+   1444, which is the sum of 852 for literal/length codes and 592 for distance
+   codes.  These values were found by exhaustive searches using the program
+   examples/enough.c found in the zlib distribtution.  The arguments to that
+   program are the number of symbols, the initial root table size, and the
+   maximum bit length of a code.  "enough 286 9 15" for literal/length codes
+   returns returns 852, and "enough 30 6 15" for distance codes returns 592.
+   The initial root table size (9 or 6) is found in the fifth argument of the
+   inflate_table() calls in inflate.c and infback.c.  If the root table size is
+   changed, then these maximum sizes would be need to be recalculated and
+   updated. */
+#define ENOUGH_LENS 852
+#define ENOUGH_DISTS 592
+#define ENOUGH (ENOUGH_LENS+ENOUGH_DISTS)
+
+/* Type of code to build for inflate_table() */
+typedef enum {
+    CODES,
+    LENS,
+    DISTS
+} codetype;
+
+int ZLIB_INTERNAL inflate_table OF((codetype type, unsigned short FAR *lens,
+                             unsigned codes, code FAR * FAR *table,
+                             unsigned FAR *bits, unsigned short FAR *work));
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/README glmark2-2021.02/android/jni/src/zlib/README
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/README	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/README	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,115 @@
+ZLIB DATA COMPRESSION LIBRARY
+
+zlib 1.2.11 is a general purpose data compression library.  All the code is
+thread safe.  The data format used by the zlib library is described by RFCs
+(Request for Comments) 1950 to 1952 in the files
+http://tools.ietf.org/html/rfc1950 (zlib format), rfc1951 (deflate format) and
+rfc1952 (gzip format).
+
+All functions of the compression library are documented in the file zlib.h
+(volunteer to write man pages welcome, contact zlib@gzip.org).  A usage example
+of the library is given in the file test/example.c which also tests that
+the library is working correctly.  Another example is given in the file
+test/minigzip.c.  The compression library itself is composed of all source
+files in the root directory.
+
+To compile all files and run the test program, follow the instructions given at
+the top of Makefile.in.  In short "./configure; make test", and if that goes
+well, "make install" should work for most flavors of Unix.  For Windows, use
+one of the special makefiles in win32/ or contrib/vstudio/ .  For VMS, use
+make_vms.com.
+
+Questions about zlib should be sent to <zlib@gzip.org>, or to Gilles Vollant
+<info@winimage.com> for the Windows DLL version.  The zlib home page is
+http://zlib.net/ .  Before reporting a problem, please check this site to
+verify that you have the latest version of zlib; otherwise get the latest
+version and check whether the problem still exists or not.
+
+PLEASE read the zlib FAQ http://zlib.net/zlib_faq.html before asking for help.
+
+Mark Nelson <markn@ieee.org> wrote an article about zlib for the Jan.  1997
+issue of Dr.  Dobb's Journal; a copy of the article is available at
+http://marknelson.us/1997/01/01/zlib-engine/ .
+
+The changes made in version 1.2.11 are documented in the file ChangeLog.
+
+Unsupported third party contributions are provided in directory contrib/ .
+
+zlib is available in Java using the java.util.zip package, documented at
+http://java.sun.com/developer/technicalArticles/Programming/compression/ .
+
+A Perl interface to zlib written by Paul Marquess <pmqs@cpan.org> is available
+at CPAN (Comprehensive Perl Archive Network) sites, including
+http://search.cpan.org/~pmqs/IO-Compress-Zlib/ .
+
+A Python interface to zlib written by A.M. Kuchling <amk@amk.ca> is
+available in Python 1.5 and later versions, see
+http://docs.python.org/library/zlib.html .
+
+zlib is built into tcl: http://wiki.tcl.tk/4610 .
+
+An experimental package to read and write files in .zip format, written on top
+of zlib by Gilles Vollant <info@winimage.com>, is available in the
+contrib/minizip directory of zlib.
+
+
+Notes for some targets:
+
+- For Windows DLL versions, please see win32/DLL_FAQ.txt
+
+- For 64-bit Irix, deflate.c must be compiled without any optimization. With
+  -O, one libpng test fails. The test works in 32 bit mode (with the -n32
+  compiler flag). The compiler bug has been reported to SGI.
+
+- zlib doesn't work with gcc 2.6.3 on a DEC 3000/300LX under OSF/1 2.1 it works
+  when compiled with cc.
+
+- On Digital Unix 4.0D (formely OSF/1) on AlphaServer, the cc option -std1 is
+  necessary to get gzprintf working correctly. This is done by configure.
+
+- zlib doesn't work on HP-UX 9.05 with some versions of /bin/cc. It works with
+  other compilers. Use "make test" to check your compiler.
+
+- gzdopen is not supported on RISCOS or BEOS.
+
+- For PalmOs, see http://palmzlib.sourceforge.net/
+
+
+Acknowledgments:
+
+  The deflate format used by zlib was defined by Phil Katz.  The deflate and
+  zlib specifications were written by L.  Peter Deutsch.  Thanks to all the
+  people who reported problems and suggested various improvements in zlib; they
+  are too numerous to cite here.
+
+Copyright notice:
+
+ (C) 1995-2017 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+If you use the zlib library in a product, we would appreciate *not* receiving
+lengthy legal documents to sign.  The sources are provided for free but without
+warranty of any kind.  The library has been entirely written by Jean-loup
+Gailly and Mark Adler; it does not include third-party code.
+
+If you redistribute modified sources, we would appreciate that you include in
+the file ChangeLog history information documenting your changes.  Please read
+the FAQ for more information on the distribution of modified source versions.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/trees.c glmark2-2021.02/android/jni/src/zlib/trees.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/trees.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/trees.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1203 @@
+/* trees.c -- output deflated data using Huffman coding
+ * Copyright (C) 1995-2017 Jean-loup Gailly
+ * detect_data_type() function provided freely by Cosmin Truta, 2006
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process uses several Huffman trees. The more
+ *      common source values are represented by shorter bit sequences.
+ *
+ *      Each code tree is stored in a compressed form which is itself
+ * a Huffman encoding of the lengths of all the code strings (in
+ * ascending order by source values).  The actual code strings are
+ * reconstructed from the lengths in the inflate process, as described
+ * in the deflate specification.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"'Deflate' Compressed Data Format Specification".
+ *      Available in ftp.uu.net:/pub/archiving/zip/doc/deflate-1.1.doc
+ *
+ *      Storer, James A.
+ *          Data Compression:  Methods and Theory, pp. 49-50.
+ *          Computer Science Press, 1988.  ISBN 0-7167-8156-5.
+ *
+ *      Sedgewick, R.
+ *          Algorithms, p290.
+ *          Addison-Wesley, 1983. ISBN 0-201-06672-6.
+ */
+
+/* @(#) $Id$ */
+
+/* #define GEN_TREES_H */
+
+#include "deflate.h"
+
+#ifdef ZLIB_DEBUG
+#  include <ctype.h>
+#endif
+
+/* ===========================================================================
+ * Constants
+ */
+
+#define MAX_BL_BITS 7
+/* Bit length codes must not exceed MAX_BL_BITS bits */
+
+#define END_BLOCK 256
+/* end of block literal code */
+
+#define REP_3_6      16
+/* repeat previous bit length 3-6 times (2 bits of repeat count) */
+
+#define REPZ_3_10    17
+/* repeat a zero length 3-10 times  (3 bits of repeat count) */
+
+#define REPZ_11_138  18
+/* repeat a zero length 11-138 times  (7 bits of repeat count) */
+
+local const int extra_lbits[LENGTH_CODES] /* extra bits for each length code */
+   = {0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0};
+
+local const int extra_dbits[D_CODES] /* extra bits for each distance code */
+   = {0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+local const int extra_blbits[BL_CODES]/* extra bits for each bit length code */
+   = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,3,7};
+
+local const uch bl_order[BL_CODES]
+   = {16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15};
+/* The lengths of the bit length codes are sent in order of decreasing
+ * probability, to avoid transmitting the lengths for unused bit length codes.
+ */
+
+/* ===========================================================================
+ * Local data. These are initialized only once.
+ */
+
+#define DIST_CODE_LEN  512 /* see definition of array dist_code below */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+/* non ANSI compilers may not accept trees.h */
+
+local ct_data static_ltree[L_CODES+2];
+/* The static literal tree. Since the bit lengths are imposed, there is no
+ * need for the L_CODES extra codes used during heap construction. However
+ * The codes 286 and 287 are needed to build a canonical tree (see _tr_init
+ * below).
+ */
+
+local ct_data static_dtree[D_CODES];
+/* The static distance tree. (Actually a trivial tree since all codes use
+ * 5 bits.)
+ */
+
+uch _dist_code[DIST_CODE_LEN];
+/* Distance codes. The first 256 values correspond to the distances
+ * 3 .. 258, the last 256 values correspond to the top 8 bits of
+ * the 15 bit distances.
+ */
+
+uch _length_code[MAX_MATCH-MIN_MATCH+1];
+/* length code for each normalized match length (0 == MIN_MATCH) */
+
+local int base_length[LENGTH_CODES];
+/* First normalized length for each code (0 = MIN_MATCH) */
+
+local int base_dist[D_CODES];
+/* First normalized distance for each code (0 = distance of 1) */
+
+#else
+#  include "trees.h"
+#endif /* GEN_TREES_H */
+
+struct static_tree_desc_s {
+    const ct_data *static_tree;  /* static tree or NULL */
+    const intf *extra_bits;      /* extra bits for each code or NULL */
+    int     extra_base;          /* base index for extra_bits */
+    int     elems;               /* max number of elements in the tree */
+    int     max_length;          /* max bit length for the codes */
+};
+
+local const static_tree_desc  static_l_desc =
+{static_ltree, extra_lbits, LITERALS+1, L_CODES, MAX_BITS};
+
+local const static_tree_desc  static_d_desc =
+{static_dtree, extra_dbits, 0,          D_CODES, MAX_BITS};
+
+local const static_tree_desc  static_bl_desc =
+{(const ct_data *)0, extra_blbits, 0,   BL_CODES, MAX_BL_BITS};
+
+/* ===========================================================================
+ * Local (static) routines in this file.
+ */
+
+local void tr_static_init OF((void));
+local void init_block     OF((deflate_state *s));
+local void pqdownheap     OF((deflate_state *s, ct_data *tree, int k));
+local void gen_bitlen     OF((deflate_state *s, tree_desc *desc));
+local void gen_codes      OF((ct_data *tree, int max_code, ushf *bl_count));
+local void build_tree     OF((deflate_state *s, tree_desc *desc));
+local void scan_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local void send_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local int  build_bl_tree  OF((deflate_state *s));
+local void send_all_trees OF((deflate_state *s, int lcodes, int dcodes,
+                              int blcodes));
+local void compress_block OF((deflate_state *s, const ct_data *ltree,
+                              const ct_data *dtree));
+local int  detect_data_type OF((deflate_state *s));
+local unsigned bi_reverse OF((unsigned value, int length));
+local void bi_windup      OF((deflate_state *s));
+local void bi_flush       OF((deflate_state *s));
+
+#ifdef GEN_TREES_H
+local void gen_trees_header OF((void));
+#endif
+
+#ifndef ZLIB_DEBUG
+#  define send_code(s, c, tree) send_bits(s, tree[c].Code, tree[c].Len)
+   /* Send a code of the given tree. c and tree must not have side effects */
+
+#else /* !ZLIB_DEBUG */
+#  define send_code(s, c, tree) \
+     { if (z_verbose>2) fprintf(stderr,"\ncd %3d ",(c)); \
+       send_bits(s, tree[c].Code, tree[c].Len); }
+#endif
+
+/* ===========================================================================
+ * Output a short LSB first on the stream.
+ * IN assertion: there is enough room in pendingBuf.
+ */
+#define put_short(s, w) { \
+    put_byte(s, (uch)((w) & 0xff)); \
+    put_byte(s, (uch)((ush)(w) >> 8)); \
+}
+
+/* ===========================================================================
+ * Send a value on a given number of bits.
+ * IN assertion: length <= 16 and value fits in length bits.
+ */
+#ifdef ZLIB_DEBUG
+local void send_bits      OF((deflate_state *s, int value, int length));
+
+local void send_bits(s, value, length)
+    deflate_state *s;
+    int value;  /* value to send */
+    int length; /* number of bits */
+{
+    Tracevv((stderr," l %2d v %4x ", length, value));
+    Assert(length > 0 && length <= 15, "invalid length");
+    s->bits_sent += (ulg)length;
+
+    /* If not enough room in bi_buf, use (valid) bits from bi_buf and
+     * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
+     * unused bits in value.
+     */
+    if (s->bi_valid > (int)Buf_size - length) {
+        s->bi_buf |= (ush)value << s->bi_valid;
+        put_short(s, s->bi_buf);
+        s->bi_buf = (ush)value >> (Buf_size - s->bi_valid);
+        s->bi_valid += length - Buf_size;
+    } else {
+        s->bi_buf |= (ush)value << s->bi_valid;
+        s->bi_valid += length;
+    }
+}
+#else /* !ZLIB_DEBUG */
+
+#define send_bits(s, value, length) \
+{ int len = length;\
+  if (s->bi_valid > (int)Buf_size - len) {\
+    int val = (int)value;\
+    s->bi_buf |= (ush)val << s->bi_valid;\
+    put_short(s, s->bi_buf);\
+    s->bi_buf = (ush)val >> (Buf_size - s->bi_valid);\
+    s->bi_valid += len - Buf_size;\
+  } else {\
+    s->bi_buf |= (ush)(value) << s->bi_valid;\
+    s->bi_valid += len;\
+  }\
+}
+#endif /* ZLIB_DEBUG */
+
+
+/* the arguments must not have side effects */
+
+/* ===========================================================================
+ * Initialize the various 'constant' tables.
+ */
+local void tr_static_init()
+{
+#if defined(GEN_TREES_H) || !defined(STDC)
+    static int static_init_done = 0;
+    int n;        /* iterates over tree elements */
+    int bits;     /* bit counter */
+    int length;   /* length value */
+    int code;     /* code value */
+    int dist;     /* distance index */
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    if (static_init_done) return;
+
+    /* For some embedded targets, global variables are not initialized: */
+#ifdef NO_INIT_GLOBAL_POINTERS
+    static_l_desc.static_tree = static_ltree;
+    static_l_desc.extra_bits = extra_lbits;
+    static_d_desc.static_tree = static_dtree;
+    static_d_desc.extra_bits = extra_dbits;
+    static_bl_desc.extra_bits = extra_blbits;
+#endif
+
+    /* Initialize the mapping length (0..255) -> length code (0..28) */
+    length = 0;
+    for (code = 0; code < LENGTH_CODES-1; code++) {
+        base_length[code] = length;
+        for (n = 0; n < (1<<extra_lbits[code]); n++) {
+            _length_code[length++] = (uch)code;
+        }
+    }
+    Assert (length == 256, "tr_static_init: length != 256");
+    /* Note that the length 255 (match length 258) can be represented
+     * in two different ways: code 284 + 5 bits or code 285, so we
+     * overwrite length_code[255] to use the best encoding:
+     */
+    _length_code[length-1] = (uch)code;
+
+    /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
+    dist = 0;
+    for (code = 0 ; code < 16; code++) {
+        base_dist[code] = dist;
+        for (n = 0; n < (1<<extra_dbits[code]); n++) {
+            _dist_code[dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: dist != 256");
+    dist >>= 7; /* from now on, all distances are divided by 128 */
+    for ( ; code < D_CODES; code++) {
+        base_dist[code] = dist << 7;
+        for (n = 0; n < (1<<(extra_dbits[code]-7)); n++) {
+            _dist_code[256 + dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: 256+dist != 512");
+
+    /* Construct the codes of the static literal tree */
+    for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0;
+    n = 0;
+    while (n <= 143) static_ltree[n++].Len = 8, bl_count[8]++;
+    while (n <= 255) static_ltree[n++].Len = 9, bl_count[9]++;
+    while (n <= 279) static_ltree[n++].Len = 7, bl_count[7]++;
+    while (n <= 287) static_ltree[n++].Len = 8, bl_count[8]++;
+    /* Codes 286 and 287 do not exist, but we must include them in the
+     * tree construction to get a canonical Huffman tree (longest code
+     * all ones)
+     */
+    gen_codes((ct_data *)static_ltree, L_CODES+1, bl_count);
+
+    /* The static distance tree is trivial: */
+    for (n = 0; n < D_CODES; n++) {
+        static_dtree[n].Len = 5;
+        static_dtree[n].Code = bi_reverse((unsigned)n, 5);
+    }
+    static_init_done = 1;
+
+#  ifdef GEN_TREES_H
+    gen_trees_header();
+#  endif
+#endif /* defined(GEN_TREES_H) || !defined(STDC) */
+}
+
+/* ===========================================================================
+ * Genererate the file trees.h describing the static trees.
+ */
+#ifdef GEN_TREES_H
+#  ifndef ZLIB_DEBUG
+#    include <stdio.h>
+#  endif
+
+#  define SEPARATOR(i, last, width) \
+      ((i) == (last)? "\n};\n\n" :    \
+       ((i) % (width) == (width)-1 ? ",\n" : ", "))
+
+void gen_trees_header()
+{
+    FILE *header = fopen("trees.h", "w");
+    int i;
+
+    Assert (header != NULL, "Can't open trees.h");
+    fprintf(header,
+            "/* header created automatically with -DGEN_TREES_H */\n\n");
+
+    fprintf(header, "local const ct_data static_ltree[L_CODES+2] = {\n");
+    for (i = 0; i < L_CODES+2; i++) {
+        fprintf(header, "{{%3u},{%3u}}%s", static_ltree[i].Code,
+                static_ltree[i].Len, SEPARATOR(i, L_CODES+1, 5));
+    }
+
+    fprintf(header, "local const ct_data static_dtree[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+        fprintf(header, "{{%2u},{%2u}}%s", static_dtree[i].Code,
+                static_dtree[i].Len, SEPARATOR(i, D_CODES-1, 5));
+    }
+
+    fprintf(header, "const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = {\n");
+    for (i = 0; i < DIST_CODE_LEN; i++) {
+        fprintf(header, "%2u%s", _dist_code[i],
+                SEPARATOR(i, DIST_CODE_LEN-1, 20));
+    }
+
+    fprintf(header,
+        "const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= {\n");
+    for (i = 0; i < MAX_MATCH-MIN_MATCH+1; i++) {
+        fprintf(header, "%2u%s", _length_code[i],
+                SEPARATOR(i, MAX_MATCH-MIN_MATCH, 20));
+    }
+
+    fprintf(header, "local const int base_length[LENGTH_CODES] = {\n");
+    for (i = 0; i < LENGTH_CODES; i++) {
+        fprintf(header, "%1u%s", base_length[i],
+                SEPARATOR(i, LENGTH_CODES-1, 20));
+    }
+
+    fprintf(header, "local const int base_dist[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+        fprintf(header, "%5u%s", base_dist[i],
+                SEPARATOR(i, D_CODES-1, 10));
+    }
+
+    fclose(header);
+}
+#endif /* GEN_TREES_H */
+
+/* ===========================================================================
+ * Initialize the tree data structures for a new zlib stream.
+ */
+void ZLIB_INTERNAL _tr_init(s)
+    deflate_state *s;
+{
+    tr_static_init();
+
+    s->l_desc.dyn_tree = s->dyn_ltree;
+    s->l_desc.stat_desc = &static_l_desc;
+
+    s->d_desc.dyn_tree = s->dyn_dtree;
+    s->d_desc.stat_desc = &static_d_desc;
+
+    s->bl_desc.dyn_tree = s->bl_tree;
+    s->bl_desc.stat_desc = &static_bl_desc;
+
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+#ifdef ZLIB_DEBUG
+    s->compressed_len = 0L;
+    s->bits_sent = 0L;
+#endif
+
+    /* Initialize the first block of the first file: */
+    init_block(s);
+}
+
+/* ===========================================================================
+ * Initialize a new block.
+ */
+local void init_block(s)
+    deflate_state *s;
+{
+    int n; /* iterates over tree elements */
+
+    /* Initialize the trees. */
+    for (n = 0; n < L_CODES;  n++) s->dyn_ltree[n].Freq = 0;
+    for (n = 0; n < D_CODES;  n++) s->dyn_dtree[n].Freq = 0;
+    for (n = 0; n < BL_CODES; n++) s->bl_tree[n].Freq = 0;
+
+    s->dyn_ltree[END_BLOCK].Freq = 1;
+    s->opt_len = s->static_len = 0L;
+    s->last_lit = s->matches = 0;
+}
+
+#define SMALLEST 1
+/* Index within the heap array of least frequent node in the Huffman tree */
+
+
+/* ===========================================================================
+ * Remove the smallest element from the heap and recreate the heap with
+ * one less element. Updates heap and heap_len.
+ */
+#define pqremove(s, tree, top) \
+{\
+    top = s->heap[SMALLEST]; \
+    s->heap[SMALLEST] = s->heap[s->heap_len--]; \
+    pqdownheap(s, tree, SMALLEST); \
+}
+
+/* ===========================================================================
+ * Compares to subtrees, using the tree depth as tie breaker when
+ * the subtrees have equal frequency. This minimizes the worst case length.
+ */
+#define smaller(tree, n, m, depth) \
+   (tree[n].Freq < tree[m].Freq || \
+   (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
+
+/* ===========================================================================
+ * Restore the heap property by moving down the tree starting at node k,
+ * exchanging a node with the smallest of its two sons if necessary, stopping
+ * when the heap property is re-established (each father smaller than its
+ * two sons).
+ */
+local void pqdownheap(s, tree, k)
+    deflate_state *s;
+    ct_data *tree;  /* the tree to restore */
+    int k;               /* node to move down */
+{
+    int v = s->heap[k];
+    int j = k << 1;  /* left son of k */
+    while (j <= s->heap_len) {
+        /* Set j to the smallest of the two sons: */
+        if (j < s->heap_len &&
+            smaller(tree, s->heap[j+1], s->heap[j], s->depth)) {
+            j++;
+        }
+        /* Exit if v is smaller than both sons */
+        if (smaller(tree, v, s->heap[j], s->depth)) break;
+
+        /* Exchange v with the smallest son */
+        s->heap[k] = s->heap[j];  k = j;
+
+        /* And continue down the tree, setting j to the left son of k */
+        j <<= 1;
+    }
+    s->heap[k] = v;
+}
+
+/* ===========================================================================
+ * Compute the optimal bit lengths for a tree and update the total bit length
+ * for the current block.
+ * IN assertion: the fields freq and dad are set, heap[heap_max] and
+ *    above are the tree nodes sorted by increasing frequency.
+ * OUT assertions: the field len is set to the optimal bit length, the
+ *     array bl_count contains the frequencies for each bit length.
+ *     The length opt_len is updated; static_len is also updated if stree is
+ *     not null.
+ */
+local void gen_bitlen(s, desc)
+    deflate_state *s;
+    tree_desc *desc;    /* the tree descriptor */
+{
+    ct_data *tree        = desc->dyn_tree;
+    int max_code         = desc->max_code;
+    const ct_data *stree = desc->stat_desc->static_tree;
+    const intf *extra    = desc->stat_desc->extra_bits;
+    int base             = desc->stat_desc->extra_base;
+    int max_length       = desc->stat_desc->max_length;
+    int h;              /* heap index */
+    int n, m;           /* iterate over the tree elements */
+    int bits;           /* bit length */
+    int xbits;          /* extra bits */
+    ush f;              /* frequency */
+    int overflow = 0;   /* number of elements with bit length too large */
+
+    for (bits = 0; bits <= MAX_BITS; bits++) s->bl_count[bits] = 0;
+
+    /* In a first pass, compute the optimal bit lengths (which may
+     * overflow in the case of the bit length tree).
+     */
+    tree[s->heap[s->heap_max]].Len = 0; /* root of the heap */
+
+    for (h = s->heap_max+1; h < HEAP_SIZE; h++) {
+        n = s->heap[h];
+        bits = tree[tree[n].Dad].Len + 1;
+        if (bits > max_length) bits = max_length, overflow++;
+        tree[n].Len = (ush)bits;
+        /* We overwrite tree[n].Dad which is no longer needed */
+
+        if (n > max_code) continue; /* not a leaf node */
+
+        s->bl_count[bits]++;
+        xbits = 0;
+        if (n >= base) xbits = extra[n-base];
+        f = tree[n].Freq;
+        s->opt_len += (ulg)f * (unsigned)(bits + xbits);
+        if (stree) s->static_len += (ulg)f * (unsigned)(stree[n].Len + xbits);
+    }
+    if (overflow == 0) return;
+
+    Tracev((stderr,"\nbit length overflow\n"));
+    /* This happens for example on obj2 and pic of the Calgary corpus */
+
+    /* Find the first bit length which could increase: */
+    do {
+        bits = max_length-1;
+        while (s->bl_count[bits] == 0) bits--;
+        s->bl_count[bits]--;      /* move one leaf down the tree */
+        s->bl_count[bits+1] += 2; /* move one overflow item as its brother */
+        s->bl_count[max_length]--;
+        /* The brother of the overflow item also moves one step up,
+         * but this does not affect bl_count[max_length]
+         */
+        overflow -= 2;
+    } while (overflow > 0);
+
+    /* Now recompute all bit lengths, scanning in increasing frequency.
+     * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
+     * lengths instead of fixing only the wrong ones. This idea is taken
+     * from 'ar' written by Haruhiko Okumura.)
+     */
+    for (bits = max_length; bits != 0; bits--) {
+        n = s->bl_count[bits];
+        while (n != 0) {
+            m = s->heap[--h];
+            if (m > max_code) continue;
+            if ((unsigned) tree[m].Len != (unsigned) bits) {
+                Tracev((stderr,"code %d bits %d->%d\n", m, tree[m].Len, bits));
+                s->opt_len += ((ulg)bits - tree[m].Len) * tree[m].Freq;
+                tree[m].Len = (ush)bits;
+            }
+            n--;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Generate the codes for a given tree and bit counts (which need not be
+ * optimal).
+ * IN assertion: the array bl_count contains the bit length statistics for
+ * the given tree and the field len is set for all tree elements.
+ * OUT assertion: the field code is set for all tree elements of non
+ *     zero code length.
+ */
+local void gen_codes (tree, max_code, bl_count)
+    ct_data *tree;             /* the tree to decorate */
+    int max_code;              /* largest code with non zero frequency */
+    ushf *bl_count;            /* number of codes at each bit length */
+{
+    ush next_code[MAX_BITS+1]; /* next code value for each bit length */
+    unsigned code = 0;         /* running code value */
+    int bits;                  /* bit index */
+    int n;                     /* code index */
+
+    /* The distribution counts are first used to generate the code values
+     * without bit reversal.
+     */
+    for (bits = 1; bits <= MAX_BITS; bits++) {
+        code = (code + bl_count[bits-1]) << 1;
+        next_code[bits] = (ush)code;
+    }
+    /* Check that the bit counts in bl_count are consistent. The last code
+     * must be all ones.
+     */
+    Assert (code + bl_count[MAX_BITS]-1 == (1<<MAX_BITS)-1,
+            "inconsistent bit counts");
+    Tracev((stderr,"\ngen_codes: max_code %d ", max_code));
+
+    for (n = 0;  n <= max_code; n++) {
+        int len = tree[n].Len;
+        if (len == 0) continue;
+        /* Now reverse the bits */
+        tree[n].Code = (ush)bi_reverse(next_code[len]++, len);
+
+        Tracecv(tree != static_ltree, (stderr,"\nn %3d %c l %2d c %4x (%x) ",
+             n, (isgraph(n) ? n : ' '), len, tree[n].Code, next_code[len]-1));
+    }
+}
+
+/* ===========================================================================
+ * Construct one Huffman tree and assigns the code bit strings and lengths.
+ * Update the total bit length for the current block.
+ * IN assertion: the field freq is set for all tree elements.
+ * OUT assertions: the fields len and code are set to the optimal bit length
+ *     and corresponding code. The length opt_len is updated; static_len is
+ *     also updated if stree is not null. The field max_code is set.
+ */
+local void build_tree(s, desc)
+    deflate_state *s;
+    tree_desc *desc; /* the tree descriptor */
+{
+    ct_data *tree         = desc->dyn_tree;
+    const ct_data *stree  = desc->stat_desc->static_tree;
+    int elems             = desc->stat_desc->elems;
+    int n, m;          /* iterate over heap elements */
+    int max_code = -1; /* largest code with non zero frequency */
+    int node;          /* new node being created */
+
+    /* Construct the initial heap, with least frequent element in
+     * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
+     * heap[0] is not used.
+     */
+    s->heap_len = 0, s->heap_max = HEAP_SIZE;
+
+    for (n = 0; n < elems; n++) {
+        if (tree[n].Freq != 0) {
+            s->heap[++(s->heap_len)] = max_code = n;
+            s->depth[n] = 0;
+        } else {
+            tree[n].Len = 0;
+        }
+    }
+
+    /* The pkzip format requires that at least one distance code exists,
+     * and that at least one bit should be sent even if there is only one
+     * possible code. So to avoid special checks later on we force at least
+     * two codes of non zero frequency.
+     */
+    while (s->heap_len < 2) {
+        node = s->heap[++(s->heap_len)] = (max_code < 2 ? ++max_code : 0);
+        tree[node].Freq = 1;
+        s->depth[node] = 0;
+        s->opt_len--; if (stree) s->static_len -= stree[node].Len;
+        /* node is 0 or 1 so it does not have extra bits */
+    }
+    desc->max_code = max_code;
+
+    /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
+     * establish sub-heaps of increasing lengths:
+     */
+    for (n = s->heap_len/2; n >= 1; n--) pqdownheap(s, tree, n);
+
+    /* Construct the Huffman tree by repeatedly combining the least two
+     * frequent nodes.
+     */
+    node = elems;              /* next internal node of the tree */
+    do {
+        pqremove(s, tree, n);  /* n = node of least frequency */
+        m = s->heap[SMALLEST]; /* m = node of next least frequency */
+
+        s->heap[--(s->heap_max)] = n; /* keep the nodes sorted by frequency */
+        s->heap[--(s->heap_max)] = m;
+
+        /* Create a new node father of n and m */
+        tree[node].Freq = tree[n].Freq + tree[m].Freq;
+        s->depth[node] = (uch)((s->depth[n] >= s->depth[m] ?
+                                s->depth[n] : s->depth[m]) + 1);
+        tree[n].Dad = tree[m].Dad = (ush)node;
+#ifdef DUMP_BL_TREE
+        if (tree == s->bl_tree) {
+            fprintf(stderr,"\nnode %d(%d), sons %d(%d) %d(%d)",
+                    node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
+        }
+#endif
+        /* and insert the new node in the heap */
+        s->heap[SMALLEST] = node++;
+        pqdownheap(s, tree, SMALLEST);
+
+    } while (s->heap_len >= 2);
+
+    s->heap[--(s->heap_max)] = s->heap[SMALLEST];
+
+    /* At this point, the fields freq and dad are set. We can now
+     * generate the bit lengths.
+     */
+    gen_bitlen(s, (tree_desc *)desc);
+
+    /* The field len is now set, we can generate the bit codes */
+    gen_codes ((ct_data *)tree, max_code, s->bl_count);
+}
+
+/* ===========================================================================
+ * Scan a literal or distance tree to determine the frequencies of the codes
+ * in the bit length tree.
+ */
+local void scan_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree;   /* the tree to be scanned */
+    int max_code;    /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    if (nextlen == 0) max_count = 138, min_count = 3;
+    tree[max_code+1].Len = (ush)0xffff; /* guard */
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            s->bl_tree[curlen].Freq += count;
+        } else if (curlen != 0) {
+            if (curlen != prevlen) s->bl_tree[curlen].Freq++;
+            s->bl_tree[REP_3_6].Freq++;
+        } else if (count <= 10) {
+            s->bl_tree[REPZ_3_10].Freq++;
+        } else {
+            s->bl_tree[REPZ_11_138].Freq++;
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Send a literal or distance tree in compressed form, using the codes in
+ * bl_tree.
+ */
+local void send_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree; /* the tree to be scanned */
+    int max_code;       /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    /* tree[max_code+1].Len = -1; */  /* guard already set */
+    if (nextlen == 0) max_count = 138, min_count = 3;
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            do { send_code(s, curlen, s->bl_tree); } while (--count != 0);
+
+        } else if (curlen != 0) {
+            if (curlen != prevlen) {
+                send_code(s, curlen, s->bl_tree); count--;
+            }
+            Assert(count >= 3 && count <= 6, " 3_6?");
+            send_code(s, REP_3_6, s->bl_tree); send_bits(s, count-3, 2);
+
+        } else if (count <= 10) {
+            send_code(s, REPZ_3_10, s->bl_tree); send_bits(s, count-3, 3);
+
+        } else {
+            send_code(s, REPZ_11_138, s->bl_tree); send_bits(s, count-11, 7);
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Construct the Huffman tree for the bit lengths and return the index in
+ * bl_order of the last bit length code to send.
+ */
+local int build_bl_tree(s)
+    deflate_state *s;
+{
+    int max_blindex;  /* index of last bit length code of non zero freq */
+
+    /* Determine the bit length frequencies for literal and distance trees */
+    scan_tree(s, (ct_data *)s->dyn_ltree, s->l_desc.max_code);
+    scan_tree(s, (ct_data *)s->dyn_dtree, s->d_desc.max_code);
+
+    /* Build the bit length tree: */
+    build_tree(s, (tree_desc *)(&(s->bl_desc)));
+    /* opt_len now includes the length of the tree representations, except
+     * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
+     */
+
+    /* Determine the number of bit length codes to send. The pkzip format
+     * requires that at least 4 bit length codes be sent. (appnote.txt says
+     * 3 but the actual value used is 4.)
+     */
+    for (max_blindex = BL_CODES-1; max_blindex >= 3; max_blindex--) {
+        if (s->bl_tree[bl_order[max_blindex]].Len != 0) break;
+    }
+    /* Update opt_len to include the bit length tree and counts */
+    s->opt_len += 3*((ulg)max_blindex+1) + 5+5+4;
+    Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld",
+            s->opt_len, s->static_len));
+
+    return max_blindex;
+}
+
+/* ===========================================================================
+ * Send the header for a block using dynamic Huffman trees: the counts, the
+ * lengths of the bit length codes, the literal tree and the distance tree.
+ * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
+ */
+local void send_all_trees(s, lcodes, dcodes, blcodes)
+    deflate_state *s;
+    int lcodes, dcodes, blcodes; /* number of codes for each tree */
+{
+    int rank;                    /* index in bl_order */
+
+    Assert (lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
+    Assert (lcodes <= L_CODES && dcodes <= D_CODES && blcodes <= BL_CODES,
+            "too many codes");
+    Tracev((stderr, "\nbl counts: "));
+    send_bits(s, lcodes-257, 5); /* not +255 as stated in appnote.txt */
+    send_bits(s, dcodes-1,   5);
+    send_bits(s, blcodes-4,  4); /* not -3 as stated in appnote.txt */
+    for (rank = 0; rank < blcodes; rank++) {
+        Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
+        send_bits(s, s->bl_tree[bl_order[rank]].Len, 3);
+    }
+    Tracev((stderr, "\nbl tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_ltree, lcodes-1); /* literal tree */
+    Tracev((stderr, "\nlit tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_dtree, dcodes-1); /* distance tree */
+    Tracev((stderr, "\ndist tree: sent %ld", s->bits_sent));
+}
+
+/* ===========================================================================
+ * Send a stored block
+ */
+void ZLIB_INTERNAL _tr_stored_block(s, buf, stored_len, last)
+    deflate_state *s;
+    charf *buf;       /* input block */
+    ulg stored_len;   /* length of input block */
+    int last;         /* one if this is the last block for a file */
+{
+    send_bits(s, (STORED_BLOCK<<1)+last, 3);    /* send block type */
+    bi_windup(s);        /* align on byte boundary */
+    put_short(s, (ush)stored_len);
+    put_short(s, (ush)~stored_len);
+    zmemcpy(s->pending_buf + s->pending, (Bytef *)buf, stored_len);
+    s->pending += stored_len;
+#ifdef ZLIB_DEBUG
+    s->compressed_len = (s->compressed_len + 3 + 7) & (ulg)~7L;
+    s->compressed_len += (stored_len + 4) << 3;
+    s->bits_sent += 2*16;
+    s->bits_sent += stored_len<<3;
+#endif
+}
+
+/* ===========================================================================
+ * Flush the bits in the bit buffer to pending output (leaves at most 7 bits)
+ */
+void ZLIB_INTERNAL _tr_flush_bits(s)
+    deflate_state *s;
+{
+    bi_flush(s);
+}
+
+/* ===========================================================================
+ * Send one empty static block to give enough lookahead for inflate.
+ * This takes 10 bits, of which 7 may remain in the bit buffer.
+ */
+void ZLIB_INTERNAL _tr_align(s)
+    deflate_state *s;
+{
+    send_bits(s, STATIC_TREES<<1, 3);
+    send_code(s, END_BLOCK, static_ltree);
+#ifdef ZLIB_DEBUG
+    s->compressed_len += 10L; /* 3 for block type, 7 for EOB */
+#endif
+    bi_flush(s);
+}
+
+/* ===========================================================================
+ * Determine the best encoding for the current block: dynamic trees, static
+ * trees or store, and write out the encoded block.
+ */
+void ZLIB_INTERNAL _tr_flush_block(s, buf, stored_len, last)
+    deflate_state *s;
+    charf *buf;       /* input block, or NULL if too old */
+    ulg stored_len;   /* length of input block */
+    int last;         /* one if this is the last block for a file */
+{
+    ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
+    int max_blindex = 0;  /* index of last bit length code of non zero freq */
+
+    /* Build the Huffman trees unless a stored block is forced */
+    if (s->level > 0) {
+
+        /* Check if the file is binary or text */
+        if (s->strm->data_type == Z_UNKNOWN)
+            s->strm->data_type = detect_data_type(s);
+
+        /* Construct the literal and distance trees */
+        build_tree(s, (tree_desc *)(&(s->l_desc)));
+        Tracev((stderr, "\nlit data: dyn %ld, stat %ld", s->opt_len,
+                s->static_len));
+
+        build_tree(s, (tree_desc *)(&(s->d_desc)));
+        Tracev((stderr, "\ndist data: dyn %ld, stat %ld", s->opt_len,
+                s->static_len));
+        /* At this point, opt_len and static_len are the total bit lengths of
+         * the compressed block data, excluding the tree representations.
+         */
+
+        /* Build the bit length tree for the above two trees, and get the index
+         * in bl_order of the last bit length code to send.
+         */
+        max_blindex = build_bl_tree(s);
+
+        /* Determine the best encoding. Compute the block lengths in bytes. */
+        opt_lenb = (s->opt_len+3+7)>>3;
+        static_lenb = (s->static_len+3+7)>>3;
+
+        Tracev((stderr, "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u ",
+                opt_lenb, s->opt_len, static_lenb, s->static_len, stored_len,
+                s->last_lit));
+
+        if (static_lenb <= opt_lenb) opt_lenb = static_lenb;
+
+    } else {
+        Assert(buf != (char*)0, "lost buf");
+        opt_lenb = static_lenb = stored_len + 5; /* force a stored block */
+    }
+
+#ifdef FORCE_STORED
+    if (buf != (char*)0) { /* force stored block */
+#else
+    if (stored_len+4 <= opt_lenb && buf != (char*)0) {
+                       /* 4: two words for the lengths */
+#endif
+        /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
+         * Otherwise we can't have processed more than WSIZE input bytes since
+         * the last block flush, because compression would have been
+         * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
+         * transform a block into a stored block.
+         */
+        _tr_stored_block(s, buf, stored_len, last);
+
+#ifdef FORCE_STATIC
+    } else if (static_lenb >= 0) { /* force static trees */
+#else
+    } else if (s->strategy == Z_FIXED || static_lenb == opt_lenb) {
+#endif
+        send_bits(s, (STATIC_TREES<<1)+last, 3);
+        compress_block(s, (const ct_data *)static_ltree,
+                       (const ct_data *)static_dtree);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 3 + s->static_len;
+#endif
+    } else {
+        send_bits(s, (DYN_TREES<<1)+last, 3);
+        send_all_trees(s, s->l_desc.max_code+1, s->d_desc.max_code+1,
+                       max_blindex+1);
+        compress_block(s, (const ct_data *)s->dyn_ltree,
+                       (const ct_data *)s->dyn_dtree);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 3 + s->opt_len;
+#endif
+    }
+    Assert (s->compressed_len == s->bits_sent, "bad compressed size");
+    /* The above check is made mod 2^32, for files larger than 512 MB
+     * and uLong implemented on 32 bits.
+     */
+    init_block(s);
+
+    if (last) {
+        bi_windup(s);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 7;  /* align on byte boundary */
+#endif
+    }
+    Tracev((stderr,"\ncomprlen %lu(%lu) ", s->compressed_len>>3,
+           s->compressed_len-7*last));
+}
+
+/* ===========================================================================
+ * Save the match info and tally the frequency counts. Return true if
+ * the current block must be flushed.
+ */
+int ZLIB_INTERNAL _tr_tally (s, dist, lc)
+    deflate_state *s;
+    unsigned dist;  /* distance of matched string */
+    unsigned lc;    /* match length-MIN_MATCH or unmatched char (if dist==0) */
+{
+    s->d_buf[s->last_lit] = (ush)dist;
+    s->l_buf[s->last_lit++] = (uch)lc;
+    if (dist == 0) {
+        /* lc is the unmatched char */
+        s->dyn_ltree[lc].Freq++;
+    } else {
+        s->matches++;
+        /* Here, lc is the match length - MIN_MATCH */
+        dist--;             /* dist = match distance - 1 */
+        Assert((ush)dist < (ush)MAX_DIST(s) &&
+               (ush)lc <= (ush)(MAX_MATCH-MIN_MATCH) &&
+               (ush)d_code(dist) < (ush)D_CODES,  "_tr_tally: bad match");
+
+        s->dyn_ltree[_length_code[lc]+LITERALS+1].Freq++;
+        s->dyn_dtree[d_code(dist)].Freq++;
+    }
+
+#ifdef TRUNCATE_BLOCK
+    /* Try to guess if it is profitable to stop the current block here */
+    if ((s->last_lit & 0x1fff) == 0 && s->level > 2) {
+        /* Compute an upper bound for the compressed length */
+        ulg out_length = (ulg)s->last_lit*8L;
+        ulg in_length = (ulg)((long)s->strstart - s->block_start);
+        int dcode;
+        for (dcode = 0; dcode < D_CODES; dcode++) {
+            out_length += (ulg)s->dyn_dtree[dcode].Freq *
+                (5L+extra_dbits[dcode]);
+        }
+        out_length >>= 3;
+        Tracev((stderr,"\nlast_lit %u, in %ld, out ~%ld(%ld%%) ",
+               s->last_lit, in_length, out_length,
+               100L - out_length*100L/in_length));
+        if (s->matches < s->last_lit/2 && out_length < in_length/2) return 1;
+    }
+#endif
+    return (s->last_lit == s->lit_bufsize-1);
+    /* We avoid equality with lit_bufsize because of wraparound at 64K
+     * on 16 bit machines and because stored blocks are restricted to
+     * 64K-1 bytes.
+     */
+}
+
+/* ===========================================================================
+ * Send the block data compressed using the given Huffman trees
+ */
+local void compress_block(s, ltree, dtree)
+    deflate_state *s;
+    const ct_data *ltree; /* literal tree */
+    const ct_data *dtree; /* distance tree */
+{
+    unsigned dist;      /* distance of matched string */
+    int lc;             /* match length or unmatched char (if dist == 0) */
+    unsigned lx = 0;    /* running index in l_buf */
+    unsigned code;      /* the code to send */
+    int extra;          /* number of extra bits to send */
+
+    if (s->last_lit != 0) do {
+        dist = s->d_buf[lx];
+        lc = s->l_buf[lx++];
+        if (dist == 0) {
+            send_code(s, lc, ltree); /* send a literal byte */
+            Tracecv(isgraph(lc), (stderr," '%c' ", lc));
+        } else {
+            /* Here, lc is the match length - MIN_MATCH */
+            code = _length_code[lc];
+            send_code(s, code+LITERALS+1, ltree); /* send the length code */
+            extra = extra_lbits[code];
+            if (extra != 0) {
+                lc -= base_length[code];
+                send_bits(s, lc, extra);       /* send the extra length bits */
+            }
+            dist--; /* dist is now the match distance - 1 */
+            code = d_code(dist);
+            Assert (code < D_CODES, "bad d_code");
+
+            send_code(s, code, dtree);       /* send the distance code */
+            extra = extra_dbits[code];
+            if (extra != 0) {
+                dist -= (unsigned)base_dist[code];
+                send_bits(s, dist, extra);   /* send the extra distance bits */
+            }
+        } /* literal or match pair ? */
+
+        /* Check that the overlay between pending_buf and d_buf+l_buf is ok: */
+        Assert((uInt)(s->pending) < s->lit_bufsize + 2*lx,
+               "pendingBuf overflow");
+
+    } while (lx < s->last_lit);
+
+    send_code(s, END_BLOCK, ltree);
+}
+
+/* ===========================================================================
+ * Check if the data type is TEXT or BINARY, using the following algorithm:
+ * - TEXT if the two conditions below are satisfied:
+ *    a) There are no non-portable control characters belonging to the
+ *       "black list" (0..6, 14..25, 28..31).
+ *    b) There is at least one printable character belonging to the
+ *       "white list" (9 {TAB}, 10 {LF}, 13 {CR}, 32..255).
+ * - BINARY otherwise.
+ * - The following partially-portable control characters form a
+ *   "gray list" that is ignored in this detection algorithm:
+ *   (7 {BEL}, 8 {BS}, 11 {VT}, 12 {FF}, 26 {SUB}, 27 {ESC}).
+ * IN assertion: the fields Freq of dyn_ltree are set.
+ */
+local int detect_data_type(s)
+    deflate_state *s;
+{
+    /* black_mask is the bit mask of black-listed bytes
+     * set bits 0..6, 14..25, and 28..31
+     * 0xf3ffc07f = binary 11110011111111111100000001111111
+     */
+    unsigned long black_mask = 0xf3ffc07fUL;
+    int n;
+
+    /* Check for non-textual ("black-listed") bytes. */
+    for (n = 0; n <= 31; n++, black_mask >>= 1)
+        if ((black_mask & 1) && (s->dyn_ltree[n].Freq != 0))
+            return Z_BINARY;
+
+    /* Check for textual ("white-listed") bytes. */
+    if (s->dyn_ltree[9].Freq != 0 || s->dyn_ltree[10].Freq != 0
+            || s->dyn_ltree[13].Freq != 0)
+        return Z_TEXT;
+    for (n = 32; n < LITERALS; n++)
+        if (s->dyn_ltree[n].Freq != 0)
+            return Z_TEXT;
+
+    /* There are no "black-listed" or "white-listed" bytes:
+     * this stream either is empty or has tolerated ("gray-listed") bytes only.
+     */
+    return Z_BINARY;
+}
+
+/* ===========================================================================
+ * Reverse the first len bits of a code, using straightforward code (a faster
+ * method would use a table)
+ * IN assertion: 1 <= len <= 15
+ */
+local unsigned bi_reverse(code, len)
+    unsigned code; /* the value to invert */
+    int len;       /* its bit length */
+{
+    register unsigned res = 0;
+    do {
+        res |= code & 1;
+        code >>= 1, res <<= 1;
+    } while (--len > 0);
+    return res >> 1;
+}
+
+/* ===========================================================================
+ * Flush the bit buffer, keeping at most 7 bits in it.
+ */
+local void bi_flush(s)
+    deflate_state *s;
+{
+    if (s->bi_valid == 16) {
+        put_short(s, s->bi_buf);
+        s->bi_buf = 0;
+        s->bi_valid = 0;
+    } else if (s->bi_valid >= 8) {
+        put_byte(s, (Byte)s->bi_buf);
+        s->bi_buf >>= 8;
+        s->bi_valid -= 8;
+    }
+}
+
+/* ===========================================================================
+ * Flush the bit buffer and align the output on a byte boundary
+ */
+local void bi_windup(s)
+    deflate_state *s;
+{
+    if (s->bi_valid > 8) {
+        put_short(s, s->bi_buf);
+    } else if (s->bi_valid > 0) {
+        put_byte(s, (Byte)s->bi_buf);
+    }
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+#ifdef ZLIB_DEBUG
+    s->bits_sent = (s->bits_sent+7) & ~7;
+#endif
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/trees.h glmark2-2021.02/android/jni/src/zlib/trees.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/trees.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/trees.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,128 @@
+/* header created automatically with -DGEN_TREES_H */
+
+local const ct_data static_ltree[L_CODES+2] = {
+{{ 12},{  8}}, {{140},{  8}}, {{ 76},{  8}}, {{204},{  8}}, {{ 44},{  8}},
+{{172},{  8}}, {{108},{  8}}, {{236},{  8}}, {{ 28},{  8}}, {{156},{  8}},
+{{ 92},{  8}}, {{220},{  8}}, {{ 60},{  8}}, {{188},{  8}}, {{124},{  8}},
+{{252},{  8}}, {{  2},{  8}}, {{130},{  8}}, {{ 66},{  8}}, {{194},{  8}},
+{{ 34},{  8}}, {{162},{  8}}, {{ 98},{  8}}, {{226},{  8}}, {{ 18},{  8}},
+{{146},{  8}}, {{ 82},{  8}}, {{210},{  8}}, {{ 50},{  8}}, {{178},{  8}},
+{{114},{  8}}, {{242},{  8}}, {{ 10},{  8}}, {{138},{  8}}, {{ 74},{  8}},
+{{202},{  8}}, {{ 42},{  8}}, {{170},{  8}}, {{106},{  8}}, {{234},{  8}},
+{{ 26},{  8}}, {{154},{  8}}, {{ 90},{  8}}, {{218},{  8}}, {{ 58},{  8}},
+{{186},{  8}}, {{122},{  8}}, {{250},{  8}}, {{  6},{  8}}, {{134},{  8}},
+{{ 70},{  8}}, {{198},{  8}}, {{ 38},{  8}}, {{166},{  8}}, {{102},{  8}},
+{{230},{  8}}, {{ 22},{  8}}, {{150},{  8}}, {{ 86},{  8}}, {{214},{  8}},
+{{ 54},{  8}}, {{182},{  8}}, {{118},{  8}}, {{246},{  8}}, {{ 14},{  8}},
+{{142},{  8}}, {{ 78},{  8}}, {{206},{  8}}, {{ 46},{  8}}, {{174},{  8}},
+{{110},{  8}}, {{238},{  8}}, {{ 30},{  8}}, {{158},{  8}}, {{ 94},{  8}},
+{{222},{  8}}, {{ 62},{  8}}, {{190},{  8}}, {{126},{  8}}, {{254},{  8}},
+{{  1},{  8}}, {{129},{  8}}, {{ 65},{  8}}, {{193},{  8}}, {{ 33},{  8}},
+{{161},{  8}}, {{ 97},{  8}}, {{225},{  8}}, {{ 17},{  8}}, {{145},{  8}},
+{{ 81},{  8}}, {{209},{  8}}, {{ 49},{  8}}, {{177},{  8}}, {{113},{  8}},
+{{241},{  8}}, {{  9},{  8}}, {{137},{  8}}, {{ 73},{  8}}, {{201},{  8}},
+{{ 41},{  8}}, {{169},{  8}}, {{105},{  8}}, {{233},{  8}}, {{ 25},{  8}},
+{{153},{  8}}, {{ 89},{  8}}, {{217},{  8}}, {{ 57},{  8}}, {{185},{  8}},
+{{121},{  8}}, {{249},{  8}}, {{  5},{  8}}, {{133},{  8}}, {{ 69},{  8}},
+{{197},{  8}}, {{ 37},{  8}}, {{165},{  8}}, {{101},{  8}}, {{229},{  8}},
+{{ 21},{  8}}, {{149},{  8}}, {{ 85},{  8}}, {{213},{  8}}, {{ 53},{  8}},
+{{181},{  8}}, {{117},{  8}}, {{245},{  8}}, {{ 13},{  8}}, {{141},{  8}},
+{{ 77},{  8}}, {{205},{  8}}, {{ 45},{  8}}, {{173},{  8}}, {{109},{  8}},
+{{237},{  8}}, {{ 29},{  8}}, {{157},{  8}}, {{ 93},{  8}}, {{221},{  8}},
+{{ 61},{  8}}, {{189},{  8}}, {{125},{  8}}, {{253},{  8}}, {{ 19},{  9}},
+{{275},{  9}}, {{147},{  9}}, {{403},{  9}}, {{ 83},{  9}}, {{339},{  9}},
+{{211},{  9}}, {{467},{  9}}, {{ 51},{  9}}, {{307},{  9}}, {{179},{  9}},
+{{435},{  9}}, {{115},{  9}}, {{371},{  9}}, {{243},{  9}}, {{499},{  9}},
+{{ 11},{  9}}, {{267},{  9}}, {{139},{  9}}, {{395},{  9}}, {{ 75},{  9}},
+{{331},{  9}}, {{203},{  9}}, {{459},{  9}}, {{ 43},{  9}}, {{299},{  9}},
+{{171},{  9}}, {{427},{  9}}, {{107},{  9}}, {{363},{  9}}, {{235},{  9}},
+{{491},{  9}}, {{ 27},{  9}}, {{283},{  9}}, {{155},{  9}}, {{411},{  9}},
+{{ 91},{  9}}, {{347},{  9}}, {{219},{  9}}, {{475},{  9}}, {{ 59},{  9}},
+{{315},{  9}}, {{187},{  9}}, {{443},{  9}}, {{123},{  9}}, {{379},{  9}},
+{{251},{  9}}, {{507},{  9}}, {{  7},{  9}}, {{263},{  9}}, {{135},{  9}},
+{{391},{  9}}, {{ 71},{  9}}, {{327},{  9}}, {{199},{  9}}, {{455},{  9}},
+{{ 39},{  9}}, {{295},{  9}}, {{167},{  9}}, {{423},{  9}}, {{103},{  9}},
+{{359},{  9}}, {{231},{  9}}, {{487},{  9}}, {{ 23},{  9}}, {{279},{  9}},
+{{151},{  9}}, {{407},{  9}}, {{ 87},{  9}}, {{343},{  9}}, {{215},{  9}},
+{{471},{  9}}, {{ 55},{  9}}, {{311},{  9}}, {{183},{  9}}, {{439},{  9}},
+{{119},{  9}}, {{375},{  9}}, {{247},{  9}}, {{503},{  9}}, {{ 15},{  9}},
+{{271},{  9}}, {{143},{  9}}, {{399},{  9}}, {{ 79},{  9}}, {{335},{  9}},
+{{207},{  9}}, {{463},{  9}}, {{ 47},{  9}}, {{303},{  9}}, {{175},{  9}},
+{{431},{  9}}, {{111},{  9}}, {{367},{  9}}, {{239},{  9}}, {{495},{  9}},
+{{ 31},{  9}}, {{287},{  9}}, {{159},{  9}}, {{415},{  9}}, {{ 95},{  9}},
+{{351},{  9}}, {{223},{  9}}, {{479},{  9}}, {{ 63},{  9}}, {{319},{  9}},
+{{191},{  9}}, {{447},{  9}}, {{127},{  9}}, {{383},{  9}}, {{255},{  9}},
+{{511},{  9}}, {{  0},{  7}}, {{ 64},{  7}}, {{ 32},{  7}}, {{ 96},{  7}},
+{{ 16},{  7}}, {{ 80},{  7}}, {{ 48},{  7}}, {{112},{  7}}, {{  8},{  7}},
+{{ 72},{  7}}, {{ 40},{  7}}, {{104},{  7}}, {{ 24},{  7}}, {{ 88},{  7}},
+{{ 56},{  7}}, {{120},{  7}}, {{  4},{  7}}, {{ 68},{  7}}, {{ 36},{  7}},
+{{100},{  7}}, {{ 20},{  7}}, {{ 84},{  7}}, {{ 52},{  7}}, {{116},{  7}},
+{{  3},{  8}}, {{131},{  8}}, {{ 67},{  8}}, {{195},{  8}}, {{ 35},{  8}},
+{{163},{  8}}, {{ 99},{  8}}, {{227},{  8}}
+};
+
+local const ct_data static_dtree[D_CODES] = {
+{{ 0},{ 5}}, {{16},{ 5}}, {{ 8},{ 5}}, {{24},{ 5}}, {{ 4},{ 5}},
+{{20},{ 5}}, {{12},{ 5}}, {{28},{ 5}}, {{ 2},{ 5}}, {{18},{ 5}},
+{{10},{ 5}}, {{26},{ 5}}, {{ 6},{ 5}}, {{22},{ 5}}, {{14},{ 5}},
+{{30},{ 5}}, {{ 1},{ 5}}, {{17},{ 5}}, {{ 9},{ 5}}, {{25},{ 5}},
+{{ 5},{ 5}}, {{21},{ 5}}, {{13},{ 5}}, {{29},{ 5}}, {{ 3},{ 5}},
+{{19},{ 5}}, {{11},{ 5}}, {{27},{ 5}}, {{ 7},{ 5}}, {{23},{ 5}}
+};
+
+const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = {
+ 0,  1,  2,  3,  4,  4,  5,  5,  6,  6,  6,  6,  7,  7,  7,  7,  8,  8,  8,  8,
+ 8,  8,  8,  8,  9,  9,  9,  9,  9,  9,  9,  9, 10, 10, 10, 10, 10, 10, 10, 10,
+10, 10, 10, 10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  0,  0, 16, 17,
+18, 18, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29
+};
+
+const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= {
+ 0,  1,  2,  3,  4,  5,  6,  7,  8,  8,  9,  9, 10, 10, 11, 11, 12, 12, 12, 12,
+13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16,
+17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, 19, 19,
+19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20,
+21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22,
+22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 28
+};
+
+local const int base_length[LENGTH_CODES] = {
+0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
+64, 80, 96, 112, 128, 160, 192, 224, 0
+};
+
+local const int base_dist[D_CODES] = {
+    0,     1,     2,     3,     4,     6,     8,    12,    16,    24,
+   32,    48,    64,    96,   128,   192,   256,   384,   512,   768,
+ 1024,  1536,  2048,  3072,  4096,  6144,  8192, 12288, 16384, 24576
+};
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/uncompr.c glmark2-2021.02/android/jni/src/zlib/uncompr.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/uncompr.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/uncompr.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,93 @@
+/* uncompr.c -- decompress a memory buffer
+ * Copyright (C) 1995-2003, 2010, 2014, 2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#define ZLIB_INTERNAL
+#include "zlib.h"
+
+/* ===========================================================================
+     Decompresses the source buffer into the destination buffer.  *sourceLen is
+   the byte length of the source buffer. Upon entry, *destLen is the total size
+   of the destination buffer, which must be large enough to hold the entire
+   uncompressed data. (The size of the uncompressed data must have been saved
+   previously by the compressor and transmitted to the decompressor by some
+   mechanism outside the scope of this compression library.) Upon exit,
+   *destLen is the size of the decompressed data and *sourceLen is the number
+   of source bytes consumed. Upon return, source + *sourceLen points to the
+   first unused input byte.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer, or
+   Z_DATA_ERROR if the input data was corrupted, including if the input data is
+   an incomplete zlib stream.
+*/
+int ZEXPORT uncompress2 (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong *sourceLen;
+{
+    z_stream stream;
+    int err;
+    const uInt max = (uInt)-1;
+    uLong len, left;
+    Byte buf[1];    /* for detection of incomplete stream when *destLen == 0 */
+
+    len = *sourceLen;
+    if (*destLen) {
+        left = *destLen;
+        *destLen = 0;
+    }
+    else {
+        left = 1;
+        dest = buf;
+    }
+
+    stream.next_in = (z_const Bytef *)source;
+    stream.avail_in = 0;
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+    stream.opaque = (voidpf)0;
+
+    err = inflateInit(&stream);
+    if (err != Z_OK) return err;
+
+    stream.next_out = dest;
+    stream.avail_out = 0;
+
+    do {
+        if (stream.avail_out == 0) {
+            stream.avail_out = left > (uLong)max ? max : (uInt)left;
+            left -= stream.avail_out;
+        }
+        if (stream.avail_in == 0) {
+            stream.avail_in = len > (uLong)max ? max : (uInt)len;
+            len -= stream.avail_in;
+        }
+        err = inflate(&stream, Z_NO_FLUSH);
+    } while (err == Z_OK);
+
+    *sourceLen -= len + stream.avail_in;
+    if (dest != buf)
+        *destLen = stream.total_out;
+    else if (stream.total_out && err == Z_BUF_ERROR)
+        left = 1;
+
+    inflateEnd(&stream);
+    return err == Z_STREAM_END ? Z_OK :
+           err == Z_NEED_DICT ? Z_DATA_ERROR  :
+           err == Z_BUF_ERROR && left + stream.avail_out ? Z_DATA_ERROR :
+           err;
+}
+
+int ZEXPORT uncompress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    return uncompress2(dest, destLen, source, &sourceLen);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zconf.h glmark2-2021.02/android/jni/src/zlib/zconf.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zconf.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/zconf.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,534 @@
+/* zconf.h -- configuration of the zlib compression library
+ * Copyright (C) 1995-2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#ifndef ZCONF_H
+#define ZCONF_H
+
+/*
+ * If you *really* need a unique prefix for all types and library functions,
+ * compile with -DZ_PREFIX. The "standard" zlib should be compiled without it.
+ * Even better than compiling with -DZ_PREFIX would be to use configure to set
+ * this permanently in zconf.h using "./configure --zprefix".
+ */
+#ifdef Z_PREFIX     /* may be set to #if 1 by ./configure */
+#  define Z_PREFIX_SET
+
+/* all linked symbols and init macros */
+#  define _dist_code            z__dist_code
+#  define _length_code          z__length_code
+#  define _tr_align             z__tr_align
+#  define _tr_flush_bits        z__tr_flush_bits
+#  define _tr_flush_block       z__tr_flush_block
+#  define _tr_init              z__tr_init
+#  define _tr_stored_block      z__tr_stored_block
+#  define _tr_tally             z__tr_tally
+#  define adler32               z_adler32
+#  define adler32_combine       z_adler32_combine
+#  define adler32_combine64     z_adler32_combine64
+#  define adler32_z             z_adler32_z
+#  ifndef Z_SOLO
+#    define compress              z_compress
+#    define compress2             z_compress2
+#    define compressBound         z_compressBound
+#  endif
+#  define crc32                 z_crc32
+#  define crc32_combine         z_crc32_combine
+#  define crc32_combine64       z_crc32_combine64
+#  define crc32_z               z_crc32_z
+#  define deflate               z_deflate
+#  define deflateBound          z_deflateBound
+#  define deflateCopy           z_deflateCopy
+#  define deflateEnd            z_deflateEnd
+#  define deflateGetDictionary  z_deflateGetDictionary
+#  define deflateInit           z_deflateInit
+#  define deflateInit2          z_deflateInit2
+#  define deflateInit2_         z_deflateInit2_
+#  define deflateInit_          z_deflateInit_
+#  define deflateParams         z_deflateParams
+#  define deflatePending        z_deflatePending
+#  define deflatePrime          z_deflatePrime
+#  define deflateReset          z_deflateReset
+#  define deflateResetKeep      z_deflateResetKeep
+#  define deflateSetDictionary  z_deflateSetDictionary
+#  define deflateSetHeader      z_deflateSetHeader
+#  define deflateTune           z_deflateTune
+#  define deflate_copyright     z_deflate_copyright
+#  define get_crc_table         z_get_crc_table
+#  ifndef Z_SOLO
+#    define gz_error              z_gz_error
+#    define gz_intmax             z_gz_intmax
+#    define gz_strwinerror        z_gz_strwinerror
+#    define gzbuffer              z_gzbuffer
+#    define gzclearerr            z_gzclearerr
+#    define gzclose               z_gzclose
+#    define gzclose_r             z_gzclose_r
+#    define gzclose_w             z_gzclose_w
+#    define gzdirect              z_gzdirect
+#    define gzdopen               z_gzdopen
+#    define gzeof                 z_gzeof
+#    define gzerror               z_gzerror
+#    define gzflush               z_gzflush
+#    define gzfread               z_gzfread
+#    define gzfwrite              z_gzfwrite
+#    define gzgetc                z_gzgetc
+#    define gzgetc_               z_gzgetc_
+#    define gzgets                z_gzgets
+#    define gzoffset              z_gzoffset
+#    define gzoffset64            z_gzoffset64
+#    define gzopen                z_gzopen
+#    define gzopen64              z_gzopen64
+#    ifdef _WIN32
+#      define gzopen_w              z_gzopen_w
+#    endif
+#    define gzprintf              z_gzprintf
+#    define gzputc                z_gzputc
+#    define gzputs                z_gzputs
+#    define gzread                z_gzread
+#    define gzrewind              z_gzrewind
+#    define gzseek                z_gzseek
+#    define gzseek64              z_gzseek64
+#    define gzsetparams           z_gzsetparams
+#    define gztell                z_gztell
+#    define gztell64              z_gztell64
+#    define gzungetc              z_gzungetc
+#    define gzvprintf             z_gzvprintf
+#    define gzwrite               z_gzwrite
+#  endif
+#  define inflate               z_inflate
+#  define inflateBack           z_inflateBack
+#  define inflateBackEnd        z_inflateBackEnd
+#  define inflateBackInit       z_inflateBackInit
+#  define inflateBackInit_      z_inflateBackInit_
+#  define inflateCodesUsed      z_inflateCodesUsed
+#  define inflateCopy           z_inflateCopy
+#  define inflateEnd            z_inflateEnd
+#  define inflateGetDictionary  z_inflateGetDictionary
+#  define inflateGetHeader      z_inflateGetHeader
+#  define inflateInit           z_inflateInit
+#  define inflateInit2          z_inflateInit2
+#  define inflateInit2_         z_inflateInit2_
+#  define inflateInit_          z_inflateInit_
+#  define inflateMark           z_inflateMark
+#  define inflatePrime          z_inflatePrime
+#  define inflateReset          z_inflateReset
+#  define inflateReset2         z_inflateReset2
+#  define inflateResetKeep      z_inflateResetKeep
+#  define inflateSetDictionary  z_inflateSetDictionary
+#  define inflateSync           z_inflateSync
+#  define inflateSyncPoint      z_inflateSyncPoint
+#  define inflateUndermine      z_inflateUndermine
+#  define inflateValidate       z_inflateValidate
+#  define inflate_copyright     z_inflate_copyright
+#  define inflate_fast          z_inflate_fast
+#  define inflate_table         z_inflate_table
+#  ifndef Z_SOLO
+#    define uncompress            z_uncompress
+#    define uncompress2           z_uncompress2
+#  endif
+#  define zError                z_zError
+#  ifndef Z_SOLO
+#    define zcalloc               z_zcalloc
+#    define zcfree                z_zcfree
+#  endif
+#  define zlibCompileFlags      z_zlibCompileFlags
+#  define zlibVersion           z_zlibVersion
+
+/* all zlib typedefs in zlib.h and zconf.h */
+#  define Byte                  z_Byte
+#  define Bytef                 z_Bytef
+#  define alloc_func            z_alloc_func
+#  define charf                 z_charf
+#  define free_func             z_free_func
+#  ifndef Z_SOLO
+#    define gzFile                z_gzFile
+#  endif
+#  define gz_header             z_gz_header
+#  define gz_headerp            z_gz_headerp
+#  define in_func               z_in_func
+#  define intf                  z_intf
+#  define out_func              z_out_func
+#  define uInt                  z_uInt
+#  define uIntf                 z_uIntf
+#  define uLong                 z_uLong
+#  define uLongf                z_uLongf
+#  define voidp                 z_voidp
+#  define voidpc                z_voidpc
+#  define voidpf                z_voidpf
+
+/* all zlib structs in zlib.h and zconf.h */
+#  define gz_header_s           z_gz_header_s
+#  define internal_state        z_internal_state
+
+#endif
+
+#if defined(__MSDOS__) && !defined(MSDOS)
+#  define MSDOS
+#endif
+#if (defined(OS_2) || defined(__OS2__)) && !defined(OS2)
+#  define OS2
+#endif
+#if defined(_WINDOWS) && !defined(WINDOWS)
+#  define WINDOWS
+#endif
+#if defined(_WIN32) || defined(_WIN32_WCE) || defined(__WIN32__)
+#  ifndef WIN32
+#    define WIN32
+#  endif
+#endif
+#if (defined(MSDOS) || defined(OS2) || defined(WINDOWS)) && !defined(WIN32)
+#  if !defined(__GNUC__) && !defined(__FLAT__) && !defined(__386__)
+#    ifndef SYS16BIT
+#      define SYS16BIT
+#    endif
+#  endif
+#endif
+
+/*
+ * Compile with -DMAXSEG_64K if the alloc function cannot allocate more
+ * than 64k bytes at a time (needed on systems with 16-bit int).
+ */
+#ifdef SYS16BIT
+#  define MAXSEG_64K
+#endif
+#ifdef MSDOS
+#  define UNALIGNED_OK
+#endif
+
+#ifdef __STDC_VERSION__
+#  ifndef STDC
+#    define STDC
+#  endif
+#  if __STDC_VERSION__ >= 199901L
+#    ifndef STDC99
+#      define STDC99
+#    endif
+#  endif
+#endif
+#if !defined(STDC) && (defined(__STDC__) || defined(__cplusplus))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(__GNUC__) || defined(__BORLANDC__))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(MSDOS) || defined(WINDOWS) || defined(WIN32))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(OS2) || defined(__HOS_AIX__))
+#  define STDC
+#endif
+
+#if defined(__OS400__) && !defined(STDC)    /* iSeries (formerly AS/400). */
+#  define STDC
+#endif
+
+#ifndef STDC
+#  ifndef const /* cannot use !defined(STDC) && !defined(const) on Mac */
+#    define const       /* note: need a more gentle solution here */
+#  endif
+#endif
+
+#if defined(ZLIB_CONST) && !defined(z_const)
+#  define z_const const
+#else
+#  define z_const
+#endif
+
+#ifdef Z_SOLO
+   typedef unsigned long z_size_t;
+#else
+#  define z_longlong long long
+#  if defined(NO_SIZE_T)
+     typedef unsigned NO_SIZE_T z_size_t;
+#  elif defined(STDC)
+#    include <stddef.h>
+     typedef size_t z_size_t;
+#  else
+     typedef unsigned long z_size_t;
+#  endif
+#  undef z_longlong
+#endif
+
+/* Maximum value for memLevel in deflateInit2 */
+#ifndef MAX_MEM_LEVEL
+#  ifdef MAXSEG_64K
+#    define MAX_MEM_LEVEL 8
+#  else
+#    define MAX_MEM_LEVEL 9
+#  endif
+#endif
+
+/* Maximum value for windowBits in deflateInit2 and inflateInit2.
+ * WARNING: reducing MAX_WBITS makes minigzip unable to extract .gz files
+ * created by gzip. (Files created by minigzip can still be extracted by
+ * gzip.)
+ */
+#ifndef MAX_WBITS
+#  define MAX_WBITS   15 /* 32K LZ77 window */
+#endif
+
+/* The memory requirements for deflate are (in bytes):
+            (1 << (windowBits+2)) +  (1 << (memLevel+9))
+ that is: 128K for windowBits=15  +  128K for memLevel = 8  (default values)
+ plus a few kilobytes for small objects. For example, if you want to reduce
+ the default memory requirements from 256K to 128K, compile with
+     make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7"
+ Of course this will generally degrade compression (there's no free lunch).
+
+   The memory requirements for inflate are (in bytes) 1 << windowBits
+ that is, 32K for windowBits=15 (default value) plus about 7 kilobytes
+ for small objects.
+*/
+
+                        /* Type declarations */
+
+#ifndef OF /* function prototypes */
+#  ifdef STDC
+#    define OF(args)  args
+#  else
+#    define OF(args)  ()
+#  endif
+#endif
+
+#ifndef Z_ARG /* function prototypes for stdarg */
+#  if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#    define Z_ARG(args)  args
+#  else
+#    define Z_ARG(args)  ()
+#  endif
+#endif
+
+/* The following definitions for FAR are needed only for MSDOS mixed
+ * model programming (small or medium model with some far allocations).
+ * This was tested only with MSC; for other MSDOS compilers you may have
+ * to define NO_MEMCPY in zutil.h.  If you don't need the mixed model,
+ * just define FAR to be empty.
+ */
+#ifdef SYS16BIT
+#  if defined(M_I86SM) || defined(M_I86MM)
+     /* MSC small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef _MSC_VER
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#  if (defined(__SMALL__) || defined(__MEDIUM__))
+     /* Turbo C small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef __BORLANDC__
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#endif
+
+#if defined(WINDOWS) || defined(WIN32)
+   /* If building or using zlib as a DLL, define ZLIB_DLL.
+    * This is not mandatory, but it offers a little performance increase.
+    */
+#  ifdef ZLIB_DLL
+#    if defined(WIN32) && (!defined(__BORLANDC__) || (__BORLANDC__ >= 0x500))
+#      ifdef ZLIB_INTERNAL
+#        define ZEXTERN extern __declspec(dllexport)
+#      else
+#        define ZEXTERN extern __declspec(dllimport)
+#      endif
+#    endif
+#  endif  /* ZLIB_DLL */
+   /* If building or using zlib with the WINAPI/WINAPIV calling convention,
+    * define ZLIB_WINAPI.
+    * Caution: the standard ZLIB1.DLL is NOT compiled using ZLIB_WINAPI.
+    */
+#  ifdef ZLIB_WINAPI
+#    ifdef FAR
+#      undef FAR
+#    endif
+#    include <windows.h>
+     /* No need for _export, use ZLIB.DEF instead. */
+     /* For complete Windows compatibility, use WINAPI, not __stdcall. */
+#    define ZEXPORT WINAPI
+#    ifdef WIN32
+#      define ZEXPORTVA WINAPIV
+#    else
+#      define ZEXPORTVA FAR CDECL
+#    endif
+#  endif
+#endif
+
+#if defined (__BEOS__)
+#  ifdef ZLIB_DLL
+#    ifdef ZLIB_INTERNAL
+#      define ZEXPORT   __declspec(dllexport)
+#      define ZEXPORTVA __declspec(dllexport)
+#    else
+#      define ZEXPORT   __declspec(dllimport)
+#      define ZEXPORTVA __declspec(dllimport)
+#    endif
+#  endif
+#endif
+
+#ifndef ZEXTERN
+#  define ZEXTERN extern
+#endif
+#ifndef ZEXPORT
+#  define ZEXPORT
+#endif
+#ifndef ZEXPORTVA
+#  define ZEXPORTVA
+#endif
+
+#ifndef FAR
+#  define FAR
+#endif
+
+#if !defined(__MACTYPES__)
+typedef unsigned char  Byte;  /* 8 bits */
+#endif
+typedef unsigned int   uInt;  /* 16 bits or more */
+typedef unsigned long  uLong; /* 32 bits or more */
+
+#ifdef SMALL_MEDIUM
+   /* Borland C/C++ and some old MSC versions ignore FAR inside typedef */
+#  define Bytef Byte FAR
+#else
+   typedef Byte  FAR Bytef;
+#endif
+typedef char  FAR charf;
+typedef int   FAR intf;
+typedef uInt  FAR uIntf;
+typedef uLong FAR uLongf;
+
+#ifdef STDC
+   typedef void const *voidpc;
+   typedef void FAR   *voidpf;
+   typedef void       *voidp;
+#else
+   typedef Byte const *voidpc;
+   typedef Byte FAR   *voidpf;
+   typedef Byte       *voidp;
+#endif
+
+#if !defined(Z_U4) && !defined(Z_SOLO) && defined(STDC)
+#  include <limits.h>
+#  if (UINT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned
+#  elif (ULONG_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned long
+#  elif (USHRT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned short
+#  endif
+#endif
+
+#ifdef Z_U4
+   typedef Z_U4 z_crc_t;
+#else
+   typedef unsigned long z_crc_t;
+#endif
+
+#ifdef HAVE_UNISTD_H    /* may be set to #if 1 by ./configure */
+#  define Z_HAVE_UNISTD_H
+#endif
+
+#ifdef HAVE_STDARG_H    /* may be set to #if 1 by ./configure */
+#  define Z_HAVE_STDARG_H
+#endif
+
+#ifdef STDC
+#  ifndef Z_SOLO
+#    include <sys/types.h>      /* for off_t */
+#  endif
+#endif
+
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+#    include <stdarg.h>         /* for va_list */
+#  endif
+#endif
+
+#ifdef _WIN32
+#  ifndef Z_SOLO
+#    include <stddef.h>         /* for wchar_t */
+#  endif
+#endif
+
+/* a little trick to accommodate both "#define _LARGEFILE64_SOURCE" and
+ * "#define _LARGEFILE64_SOURCE 1" as requesting 64-bit operations, (even
+ * though the former does not conform to the LFS document), but considering
+ * both "#undef _LARGEFILE64_SOURCE" and "#define _LARGEFILE64_SOURCE 0" as
+ * equivalently requesting no 64-bit operations
+ */
+#if defined(_LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1
+#  undef _LARGEFILE64_SOURCE
+#endif
+
+#if defined(__WATCOMC__) && !defined(Z_HAVE_UNISTD_H)
+#  define Z_HAVE_UNISTD_H
+#endif
+#ifndef Z_SOLO
+#  if defined(Z_HAVE_UNISTD_H) || defined(_LARGEFILE64_SOURCE)
+#    include <unistd.h>         /* for SEEK_*, off_t, and _LFS64_LARGEFILE */
+#    ifdef VMS
+#      include <unixio.h>       /* for off_t */
+#    endif
+#    ifndef z_off_t
+#      define z_off_t off_t
+#    endif
+#  endif
+#endif
+
+#if defined(_LFS64_LARGEFILE) && _LFS64_LARGEFILE-0
+#  define Z_LFS64
+#endif
+
+#if defined(_LARGEFILE64_SOURCE) && defined(Z_LFS64)
+#  define Z_LARGE64
+#endif
+
+#if defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS-0 == 64 && defined(Z_LFS64)
+#  define Z_WANT64
+#endif
+
+#if !defined(SEEK_SET) && !defined(Z_SOLO)
+#  define SEEK_SET        0       /* Seek from beginning of file.  */
+#  define SEEK_CUR        1       /* Seek from current position.  */
+#  define SEEK_END        2       /* Set file pointer to EOF plus "offset" */
+#endif
+
+#ifndef z_off_t
+#  define z_off_t long
+#endif
+
+#if !defined(_WIN32) && defined(Z_LARGE64)
+#  define z_off64_t off64_t
+#else
+#  if defined(_WIN32) && !defined(__GNUC__) && !defined(Z_SOLO)
+#    define z_off64_t __int64
+#  else
+#    define z_off64_t z_off_t
+#  endif
+#endif
+
+/* MVS linker does not support external names larger than 8 bytes */
+#if defined(__MVS__)
+  #pragma map(deflateInit_,"DEIN")
+  #pragma map(deflateInit2_,"DEIN2")
+  #pragma map(deflateEnd,"DEEND")
+  #pragma map(deflateBound,"DEBND")
+  #pragma map(inflateInit_,"ININ")
+  #pragma map(inflateInit2_,"ININ2")
+  #pragma map(inflateEnd,"INEND")
+  #pragma map(inflateSync,"INSY")
+  #pragma map(inflateSetDictionary,"INSEDI")
+  #pragma map(compressBound,"CMBND")
+  #pragma map(inflate_table,"INTABL")
+  #pragma map(inflate_fast,"INFA")
+  #pragma map(inflate_copyright,"INCOPY")
+#endif
+
+#endif /* ZCONF_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zlib.h glmark2-2021.02/android/jni/src/zlib/zlib.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zlib.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/zlib.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1912 @@
+/* zlib.h -- interface of the 'zlib' general purpose compression library
+  version 1.2.11, January 15th, 2017
+
+  Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+
+  The data format used by the zlib library is described by RFCs (Request for
+  Comments) 1950 to 1952 in the files http://tools.ietf.org/html/rfc1950
+  (zlib format), rfc1951 (deflate format) and rfc1952 (gzip format).
+*/
+
+#ifndef ZLIB_H
+#define ZLIB_H
+
+#include "zconf.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZLIB_VERSION "1.2.11"
+#define ZLIB_VERNUM 0x12b0
+#define ZLIB_VER_MAJOR 1
+#define ZLIB_VER_MINOR 2
+#define ZLIB_VER_REVISION 11
+#define ZLIB_VER_SUBREVISION 0
+
+/*
+    The 'zlib' compression library provides in-memory compression and
+  decompression functions, including integrity checks of the uncompressed data.
+  This version of the library supports only one compression method (deflation)
+  but other algorithms will be added later and will have the same stream
+  interface.
+
+    Compression can be done in a single step if the buffers are large enough,
+  or can be done by repeated calls of the compression function.  In the latter
+  case, the application must provide more input and/or consume the output
+  (providing more output space) before each call.
+
+    The compressed data format used by default by the in-memory functions is
+  the zlib format, which is a zlib wrapper documented in RFC 1950, wrapped
+  around a deflate stream, which is itself documented in RFC 1951.
+
+    The library also supports reading and writing files in gzip (.gz) format
+  with an interface similar to that of stdio using the functions that start
+  with "gz".  The gzip format is different from the zlib format.  gzip is a
+  gzip wrapper, documented in RFC 1952, wrapped around a deflate stream.
+
+    This library can optionally read and write gzip and raw deflate streams in
+  memory as well.
+
+    The zlib format was designed to be compact and fast for use in memory
+  and on communications channels.  The gzip format was designed for single-
+  file compression on file systems, has a larger header than zlib to maintain
+  directory information, and uses a different, slower check method than zlib.
+
+    The library does not install any signal handler.  The decoder checks
+  the consistency of the compressed data, so the library should never crash
+  even in the case of corrupted input.
+*/
+
+typedef voidpf (*alloc_func) OF((voidpf opaque, uInt items, uInt size));
+typedef void   (*free_func)  OF((voidpf opaque, voidpf address));
+
+struct internal_state;
+
+typedef struct z_stream_s {
+    z_const Bytef *next_in;     /* next input byte */
+    uInt     avail_in;  /* number of bytes available at next_in */
+    uLong    total_in;  /* total number of input bytes read so far */
+
+    Bytef    *next_out; /* next output byte will go here */
+    uInt     avail_out; /* remaining free space at next_out */
+    uLong    total_out; /* total number of bytes output so far */
+
+    z_const char *msg;  /* last error message, NULL if no error */
+    struct internal_state FAR *state; /* not visible by applications */
+
+    alloc_func zalloc;  /* used to allocate the internal state */
+    free_func  zfree;   /* used to free the internal state */
+    voidpf     opaque;  /* private data object passed to zalloc and zfree */
+
+    int     data_type;  /* best guess about the data type: binary or text
+                           for deflate, or the decoding state for inflate */
+    uLong   adler;      /* Adler-32 or CRC-32 value of the uncompressed data */
+    uLong   reserved;   /* reserved for future use */
+} z_stream;
+
+typedef z_stream FAR *z_streamp;
+
+/*
+     gzip header information passed to and from zlib routines.  See RFC 1952
+  for more details on the meanings of these fields.
+*/
+typedef struct gz_header_s {
+    int     text;       /* true if compressed data believed to be text */
+    uLong   time;       /* modification time */
+    int     xflags;     /* extra flags (not used when writing a gzip file) */
+    int     os;         /* operating system */
+    Bytef   *extra;     /* pointer to extra field or Z_NULL if none */
+    uInt    extra_len;  /* extra field length (valid if extra != Z_NULL) */
+    uInt    extra_max;  /* space at extra (only when reading header) */
+    Bytef   *name;      /* pointer to zero-terminated file name or Z_NULL */
+    uInt    name_max;   /* space at name (only when reading header) */
+    Bytef   *comment;   /* pointer to zero-terminated comment or Z_NULL */
+    uInt    comm_max;   /* space at comment (only when reading header) */
+    int     hcrc;       /* true if there was or will be a header crc */
+    int     done;       /* true when done reading gzip header (not used
+                           when writing a gzip file) */
+} gz_header;
+
+typedef gz_header FAR *gz_headerp;
+
+/*
+     The application must update next_in and avail_in when avail_in has dropped
+   to zero.  It must update next_out and avail_out when avail_out has dropped
+   to zero.  The application must initialize zalloc, zfree and opaque before
+   calling the init function.  All other fields are set by the compression
+   library and must not be updated by the application.
+
+     The opaque value provided by the application will be passed as the first
+   parameter for calls of zalloc and zfree.  This can be useful for custom
+   memory management.  The compression library attaches no meaning to the
+   opaque value.
+
+     zalloc must return Z_NULL if there is not enough memory for the object.
+   If zlib is used in a multi-threaded application, zalloc and zfree must be
+   thread safe.  In that case, zlib is thread-safe.  When zalloc and zfree are
+   Z_NULL on entry to the initialization function, they are set to internal
+   routines that use the standard library functions malloc() and free().
+
+     On 16-bit systems, the functions zalloc and zfree must be able to allocate
+   exactly 65536 bytes, but will not be required to allocate more than this if
+   the symbol MAXSEG_64K is defined (see zconf.h).  WARNING: On MSDOS, pointers
+   returned by zalloc for objects of exactly 65536 bytes *must* have their
+   offset normalized to zero.  The default allocation function provided by this
+   library ensures this (see zutil.c).  To reduce memory requirements and avoid
+   any allocation of 64K objects, at the expense of compression ratio, compile
+   the library with -DMAX_WBITS=14 (see zconf.h).
+
+     The fields total_in and total_out can be used for statistics or progress
+   reports.  After compression, total_in holds the total size of the
+   uncompressed data and may be saved for use by the decompressor (particularly
+   if the decompressor wants to decompress everything in a single step).
+*/
+
+                        /* constants */
+
+#define Z_NO_FLUSH      0
+#define Z_PARTIAL_FLUSH 1
+#define Z_SYNC_FLUSH    2
+#define Z_FULL_FLUSH    3
+#define Z_FINISH        4
+#define Z_BLOCK         5
+#define Z_TREES         6
+/* Allowed flush values; see deflate() and inflate() below for details */
+
+#define Z_OK            0
+#define Z_STREAM_END    1
+#define Z_NEED_DICT     2
+#define Z_ERRNO        (-1)
+#define Z_STREAM_ERROR (-2)
+#define Z_DATA_ERROR   (-3)
+#define Z_MEM_ERROR    (-4)
+#define Z_BUF_ERROR    (-5)
+#define Z_VERSION_ERROR (-6)
+/* Return codes for the compression/decompression functions. Negative values
+ * are errors, positive values are used for special but normal events.
+ */
+
+#define Z_NO_COMPRESSION         0
+#define Z_BEST_SPEED             1
+#define Z_BEST_COMPRESSION       9
+#define Z_DEFAULT_COMPRESSION  (-1)
+/* compression levels */
+
+#define Z_FILTERED            1
+#define Z_HUFFMAN_ONLY        2
+#define Z_RLE                 3
+#define Z_FIXED               4
+#define Z_DEFAULT_STRATEGY    0
+/* compression strategy; see deflateInit2() below for details */
+
+#define Z_BINARY   0
+#define Z_TEXT     1
+#define Z_ASCII    Z_TEXT   /* for compatibility with 1.2.2 and earlier */
+#define Z_UNKNOWN  2
+/* Possible values of the data_type field for deflate() */
+
+#define Z_DEFLATED   8
+/* The deflate compression method (the only one supported in this version) */
+
+#define Z_NULL  0  /* for initializing zalloc, zfree, opaque */
+
+#define zlib_version zlibVersion()
+/* for compatibility with versions < 1.0.2 */
+
+
+                        /* basic functions */
+
+ZEXTERN const char * ZEXPORT zlibVersion OF((void));
+/* The application can compare zlibVersion and ZLIB_VERSION for consistency.
+   If the first character differs, the library code actually used is not
+   compatible with the zlib.h header file used by the application.  This check
+   is automatically made by deflateInit and inflateInit.
+ */
+
+/*
+ZEXTERN int ZEXPORT deflateInit OF((z_streamp strm, int level));
+
+     Initializes the internal stream state for compression.  The fields
+   zalloc, zfree and opaque must be initialized before by the caller.  If
+   zalloc and zfree are set to Z_NULL, deflateInit updates them to use default
+   allocation functions.
+
+     The compression level must be Z_DEFAULT_COMPRESSION, or between 0 and 9:
+   1 gives best speed, 9 gives best compression, 0 gives no compression at all
+   (the input data is simply copied a block at a time).  Z_DEFAULT_COMPRESSION
+   requests a default compromise between speed and compression (currently
+   equivalent to level 6).
+
+     deflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if level is not a valid compression level, or
+   Z_VERSION_ERROR if the zlib library version (zlib_version) is incompatible
+   with the version assumed by the caller (ZLIB_VERSION).  msg is set to null
+   if there is no error message.  deflateInit does not perform any compression:
+   this will be done by deflate().
+*/
+
+
+ZEXTERN int ZEXPORT deflate OF((z_streamp strm, int flush));
+/*
+    deflate compresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full.  It may introduce
+  some output latency (reading input without producing any output) except when
+  forced to flush.
+
+    The detailed semantics are as follows.  deflate performs one or both of the
+  following actions:
+
+  - Compress more input starting at next_in and update next_in and avail_in
+    accordingly.  If not all input can be processed (because there is not
+    enough room in the output buffer), next_in and avail_in are updated and
+    processing will resume at this point for the next call of deflate().
+
+  - Generate more output starting at next_out and update next_out and avail_out
+    accordingly.  This action is forced if the parameter flush is non zero.
+    Forcing flush frequently degrades the compression ratio, so this parameter
+    should be set only when necessary.  Some output may be provided even if
+    flush is zero.
+
+    Before the call of deflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming more
+  output, and updating avail_in or avail_out accordingly; avail_out should
+  never be zero before the call.  The application can consume the compressed
+  output when it wants, for example when the output buffer is full (avail_out
+  == 0), or after each call of deflate().  If deflate returns Z_OK and with
+  zero avail_out, it must be called again after making room in the output
+  buffer because there might be more output pending. See deflatePending(),
+  which can be used if desired to determine whether or not there is more ouput
+  in that case.
+
+    Normally the parameter flush is set to Z_NO_FLUSH, which allows deflate to
+  decide how much data to accumulate before producing output, in order to
+  maximize compression.
+
+    If the parameter flush is set to Z_SYNC_FLUSH, all pending output is
+  flushed to the output buffer and the output is aligned on a byte boundary, so
+  that the decompressor can get all input data available so far.  (In
+  particular avail_in is zero after the call if enough output space has been
+  provided before the call.) Flushing may degrade compression for some
+  compression algorithms and so it should be used only when necessary.  This
+  completes the current deflate block and follows it with an empty stored block
+  that is three bits plus filler bits to the next byte, followed by four bytes
+  (00 00 ff ff).
+
+    If flush is set to Z_PARTIAL_FLUSH, all pending output is flushed to the
+  output buffer, but the output is not aligned to a byte boundary.  All of the
+  input data so far will be available to the decompressor, as for Z_SYNC_FLUSH.
+  This completes the current deflate block and follows it with an empty fixed
+  codes block that is 10 bits long.  This assures that enough bytes are output
+  in order for the decompressor to finish the block before the empty fixed
+  codes block.
+
+    If flush is set to Z_BLOCK, a deflate block is completed and emitted, as
+  for Z_SYNC_FLUSH, but the output is not aligned on a byte boundary, and up to
+  seven bits of the current block are held to be written as the next byte after
+  the next deflate block is completed.  In this case, the decompressor may not
+  be provided enough bits at this point in order to complete decompression of
+  the data provided so far to the compressor.  It may need to wait for the next
+  block to be emitted.  This is for advanced applications that need to control
+  the emission of deflate blocks.
+
+    If flush is set to Z_FULL_FLUSH, all output is flushed as with
+  Z_SYNC_FLUSH, and the compression state is reset so that decompression can
+  restart from this point if previous compressed data has been damaged or if
+  random access is desired.  Using Z_FULL_FLUSH too often can seriously degrade
+  compression.
+
+    If deflate returns with avail_out == 0, this function must be called again
+  with the same value of the flush parameter and more output space (updated
+  avail_out), until the flush is complete (deflate returns with non-zero
+  avail_out).  In the case of a Z_FULL_FLUSH or Z_SYNC_FLUSH, make sure that
+  avail_out is greater than six to avoid repeated flush markers due to
+  avail_out == 0 on return.
+
+    If the parameter flush is set to Z_FINISH, pending input is processed,
+  pending output is flushed and deflate returns with Z_STREAM_END if there was
+  enough output space.  If deflate returns with Z_OK or Z_BUF_ERROR, this
+  function must be called again with Z_FINISH and more output space (updated
+  avail_out) but no more input data, until it returns with Z_STREAM_END or an
+  error.  After deflate has returned Z_STREAM_END, the only possible operations
+  on the stream are deflateReset or deflateEnd.
+
+    Z_FINISH can be used in the first deflate call after deflateInit if all the
+  compression is to be done in a single step.  In order to complete in one
+  call, avail_out must be at least the value returned by deflateBound (see
+  below).  Then deflate is guaranteed to return Z_STREAM_END.  If not enough
+  output space is provided, deflate will not return Z_STREAM_END, and it must
+  be called again as described above.
+
+    deflate() sets strm->adler to the Adler-32 checksum of all input read
+  so far (that is, total_in bytes).  If a gzip stream is being generated, then
+  strm->adler will be the CRC-32 checksum of the input read so far.  (See
+  deflateInit2 below.)
+
+    deflate() may update strm->data_type if it can make a good guess about
+  the input data type (Z_BINARY or Z_TEXT).  If in doubt, the data is
+  considered binary.  This field is only for information purposes and does not
+  affect the compression algorithm in any manner.
+
+    deflate() returns Z_OK if some progress has been made (more input
+  processed or more output produced), Z_STREAM_END if all input has been
+  consumed and all output has been produced (only when flush is set to
+  Z_FINISH), Z_STREAM_ERROR if the stream state was inconsistent (for example
+  if next_in or next_out was Z_NULL or the state was inadvertently written over
+  by the application), or Z_BUF_ERROR if no progress is possible (for example
+  avail_in or avail_out was zero).  Note that Z_BUF_ERROR is not fatal, and
+  deflate() can be called again with more input and more output space to
+  continue compressing.
+*/
+
+
+ZEXTERN int ZEXPORT deflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any pending
+   output.
+
+     deflateEnd returns Z_OK if success, Z_STREAM_ERROR if the
+   stream state was inconsistent, Z_DATA_ERROR if the stream was freed
+   prematurely (some input or output was discarded).  In the error case, msg
+   may be set but then points to a static string (which must not be
+   deallocated).
+*/
+
+
+/*
+ZEXTERN int ZEXPORT inflateInit OF((z_streamp strm));
+
+     Initializes the internal stream state for decompression.  The fields
+   next_in, avail_in, zalloc, zfree and opaque must be initialized before by
+   the caller.  In the current version of inflate, the provided input is not
+   read or consumed.  The allocation of a sliding window will be deferred to
+   the first call of inflate (if the decompression does not complete on the
+   first call).  If zalloc and zfree are set to Z_NULL, inflateInit updates
+   them to use default allocation functions.
+
+     inflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_VERSION_ERROR if the zlib library version is incompatible with the
+   version assumed by the caller, or Z_STREAM_ERROR if the parameters are
+   invalid, such as a null pointer to the structure.  msg is set to null if
+   there is no error message.  inflateInit does not perform any decompression.
+   Actual decompression will be done by inflate().  So next_in, and avail_in,
+   next_out, and avail_out are unused and unchanged.  The current
+   implementation of inflateInit() does not process any header information --
+   that is deferred until inflate() is called.
+*/
+
+
+ZEXTERN int ZEXPORT inflate OF((z_streamp strm, int flush));
+/*
+    inflate decompresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full.  It may introduce
+  some output latency (reading input without producing any output) except when
+  forced to flush.
+
+  The detailed semantics are as follows.  inflate performs one or both of the
+  following actions:
+
+  - Decompress more input starting at next_in and update next_in and avail_in
+    accordingly.  If not all input can be processed (because there is not
+    enough room in the output buffer), then next_in and avail_in are updated
+    accordingly, and processing will resume at this point for the next call of
+    inflate().
+
+  - Generate more output starting at next_out and update next_out and avail_out
+    accordingly.  inflate() provides as much output as possible, until there is
+    no more input data or no more space in the output buffer (see below about
+    the flush parameter).
+
+    Before the call of inflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming more
+  output, and updating the next_* and avail_* values accordingly.  If the
+  caller of inflate() does not provide both available input and available
+  output space, it is possible that there will be no progress made.  The
+  application can consume the uncompressed output when it wants, for example
+  when the output buffer is full (avail_out == 0), or after each call of
+  inflate().  If inflate returns Z_OK and with zero avail_out, it must be
+  called again after making room in the output buffer because there might be
+  more output pending.
+
+    The flush parameter of inflate() can be Z_NO_FLUSH, Z_SYNC_FLUSH, Z_FINISH,
+  Z_BLOCK, or Z_TREES.  Z_SYNC_FLUSH requests that inflate() flush as much
+  output as possible to the output buffer.  Z_BLOCK requests that inflate()
+  stop if and when it gets to the next deflate block boundary.  When decoding
+  the zlib or gzip format, this will cause inflate() to return immediately
+  after the header and before the first block.  When doing a raw inflate,
+  inflate() will go ahead and process the first block, and will return when it
+  gets to the end of that block, or when it runs out of data.
+
+    The Z_BLOCK option assists in appending to or combining deflate streams.
+  To assist in this, on return inflate() always sets strm->data_type to the
+  number of unused bits in the last byte taken from strm->next_in, plus 64 if
+  inflate() is currently decoding the last block in the deflate stream, plus
+  128 if inflate() returned immediately after decoding an end-of-block code or
+  decoding the complete header up to just before the first byte of the deflate
+  stream.  The end-of-block will not be indicated until all of the uncompressed
+  data from that block has been written to strm->next_out.  The number of
+  unused bits may in general be greater than seven, except when bit 7 of
+  data_type is set, in which case the number of unused bits will be less than
+  eight.  data_type is set as noted here every time inflate() returns for all
+  flush options, and so can be used to determine the amount of currently
+  consumed input in bits.
+
+    The Z_TREES option behaves as Z_BLOCK does, but it also returns when the
+  end of each deflate block header is reached, before any actual data in that
+  block is decoded.  This allows the caller to determine the length of the
+  deflate block header for later use in random access within a deflate block.
+  256 is added to the value of strm->data_type when inflate() returns
+  immediately after reaching the end of the deflate block header.
+
+    inflate() should normally be called until it returns Z_STREAM_END or an
+  error.  However if all decompression is to be performed in a single step (a
+  single call of inflate), the parameter flush should be set to Z_FINISH.  In
+  this case all pending input is processed and all pending output is flushed;
+  avail_out must be large enough to hold all of the uncompressed data for the
+  operation to complete.  (The size of the uncompressed data may have been
+  saved by the compressor for this purpose.)  The use of Z_FINISH is not
+  required to perform an inflation in one step.  However it may be used to
+  inform inflate that a faster approach can be used for the single inflate()
+  call.  Z_FINISH also informs inflate to not maintain a sliding window if the
+  stream completes, which reduces inflate's memory footprint.  If the stream
+  does not complete, either because not all of the stream is provided or not
+  enough output space is provided, then a sliding window will be allocated and
+  inflate() can be called again to continue the operation as if Z_NO_FLUSH had
+  been used.
+
+     In this implementation, inflate() always flushes as much output as
+  possible to the output buffer, and always uses the faster approach on the
+  first call.  So the effects of the flush parameter in this implementation are
+  on the return value of inflate() as noted below, when inflate() returns early
+  when Z_BLOCK or Z_TREES is used, and when inflate() avoids the allocation of
+  memory for a sliding window when Z_FINISH is used.
+
+     If a preset dictionary is needed after this call (see inflateSetDictionary
+  below), inflate sets strm->adler to the Adler-32 checksum of the dictionary
+  chosen by the compressor and returns Z_NEED_DICT; otherwise it sets
+  strm->adler to the Adler-32 checksum of all output produced so far (that is,
+  total_out bytes) and returns Z_OK, Z_STREAM_END or an error code as described
+  below.  At the end of the stream, inflate() checks that its computed Adler-32
+  checksum is equal to that saved by the compressor and returns Z_STREAM_END
+  only if the checksum is correct.
+
+    inflate() can decompress and check either zlib-wrapped or gzip-wrapped
+  deflate data.  The header type is detected automatically, if requested when
+  initializing with inflateInit2().  Any information contained in the gzip
+  header is not retained unless inflateGetHeader() is used.  When processing
+  gzip-wrapped deflate data, strm->adler32 is set to the CRC-32 of the output
+  produced so far.  The CRC-32 is checked against the gzip trailer, as is the
+  uncompressed length, modulo 2^32.
+
+    inflate() returns Z_OK if some progress has been made (more input processed
+  or more output produced), Z_STREAM_END if the end of the compressed data has
+  been reached and all uncompressed output has been produced, Z_NEED_DICT if a
+  preset dictionary is needed at this point, Z_DATA_ERROR if the input data was
+  corrupted (input stream not conforming to the zlib format or incorrect check
+  value, in which case strm->msg points to a string with a more specific
+  error), Z_STREAM_ERROR if the stream structure was inconsistent (for example
+  next_in or next_out was Z_NULL, or the state was inadvertently written over
+  by the application), Z_MEM_ERROR if there was not enough memory, Z_BUF_ERROR
+  if no progress was possible or if there was not enough room in the output
+  buffer when Z_FINISH is used.  Note that Z_BUF_ERROR is not fatal, and
+  inflate() can be called again with more input and more output space to
+  continue decompressing.  If Z_DATA_ERROR is returned, the application may
+  then call inflateSync() to look for a good compression block if a partial
+  recovery of the data is to be attempted.
+*/
+
+
+ZEXTERN int ZEXPORT inflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any pending
+   output.
+
+     inflateEnd returns Z_OK if success, or Z_STREAM_ERROR if the stream state
+   was inconsistent.
+*/
+
+
+                        /* Advanced functions */
+
+/*
+    The following functions are needed only in some special applications.
+*/
+
+/*
+ZEXTERN int ZEXPORT deflateInit2 OF((z_streamp strm,
+                                     int  level,
+                                     int  method,
+                                     int  windowBits,
+                                     int  memLevel,
+                                     int  strategy));
+
+     This is another version of deflateInit with more compression options.  The
+   fields next_in, zalloc, zfree and opaque must be initialized before by the
+   caller.
+
+     The method parameter is the compression method.  It must be Z_DEFLATED in
+   this version of the library.
+
+     The windowBits parameter is the base two logarithm of the window size
+   (the size of the history buffer).  It should be in the range 8..15 for this
+   version of the library.  Larger values of this parameter result in better
+   compression at the expense of memory usage.  The default value is 15 if
+   deflateInit is used instead.
+
+     For the current implementation of deflate(), a windowBits value of 8 (a
+   window size of 256 bytes) is not supported.  As a result, a request for 8
+   will result in 9 (a 512-byte window).  In that case, providing 8 to
+   inflateInit2() will result in an error when the zlib header with 9 is
+   checked against the initialization of inflate().  The remedy is to not use 8
+   with deflateInit2() with this initialization, or at least in that case use 9
+   with inflateInit2().
+
+     windowBits can also be -8..-15 for raw deflate.  In this case, -windowBits
+   determines the window size.  deflate() will then generate raw deflate data
+   with no zlib header or trailer, and will not compute a check value.
+
+     windowBits can also be greater than 15 for optional gzip encoding.  Add
+   16 to windowBits to write a simple gzip header and trailer around the
+   compressed data instead of a zlib wrapper.  The gzip header will have no
+   file name, no extra data, no comment, no modification time (set to zero), no
+   header crc, and the operating system will be set to the appropriate value,
+   if the operating system was determined at compile time.  If a gzip stream is
+   being written, strm->adler is a CRC-32 instead of an Adler-32.
+
+     For raw deflate or gzip encoding, a request for a 256-byte window is
+   rejected as invalid, since only the zlib header provides a means of
+   transmitting the window size to the decompressor.
+
+     The memLevel parameter specifies how much memory should be allocated
+   for the internal compression state.  memLevel=1 uses minimum memory but is
+   slow and reduces compression ratio; memLevel=9 uses maximum memory for
+   optimal speed.  The default value is 8.  See zconf.h for total memory usage
+   as a function of windowBits and memLevel.
+
+     The strategy parameter is used to tune the compression algorithm.  Use the
+   value Z_DEFAULT_STRATEGY for normal data, Z_FILTERED for data produced by a
+   filter (or predictor), Z_HUFFMAN_ONLY to force Huffman encoding only (no
+   string match), or Z_RLE to limit match distances to one (run-length
+   encoding).  Filtered data consists mostly of small values with a somewhat
+   random distribution.  In this case, the compression algorithm is tuned to
+   compress them better.  The effect of Z_FILTERED is to force more Huffman
+   coding and less string matching; it is somewhat intermediate between
+   Z_DEFAULT_STRATEGY and Z_HUFFMAN_ONLY.  Z_RLE is designed to be almost as
+   fast as Z_HUFFMAN_ONLY, but give better compression for PNG image data.  The
+   strategy parameter only affects the compression ratio but not the
+   correctness of the compressed output even if it is not set appropriately.
+   Z_FIXED prevents the use of dynamic Huffman codes, allowing for a simpler
+   decoder for special applications.
+
+     deflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if any parameter is invalid (such as an invalid
+   method), or Z_VERSION_ERROR if the zlib library version (zlib_version) is
+   incompatible with the version assumed by the caller (ZLIB_VERSION).  msg is
+   set to null if there is no error message.  deflateInit2 does not perform any
+   compression: this will be done by deflate().
+*/
+
+ZEXTERN int ZEXPORT deflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the compression dictionary from the given byte sequence
+   without producing any compressed output.  When using the zlib format, this
+   function must be called immediately after deflateInit, deflateInit2 or
+   deflateReset, and before any call of deflate.  When doing raw deflate, this
+   function must be called either before any call of deflate, or immediately
+   after the completion of a deflate block, i.e. after all input has been
+   consumed and all output has been delivered when using any of the flush
+   options Z_BLOCK, Z_PARTIAL_FLUSH, Z_SYNC_FLUSH, or Z_FULL_FLUSH.  The
+   compressor and decompressor must use exactly the same dictionary (see
+   inflateSetDictionary).
+
+     The dictionary should consist of strings (byte sequences) that are likely
+   to be encountered later in the data to be compressed, with the most commonly
+   used strings preferably put towards the end of the dictionary.  Using a
+   dictionary is most useful when the data to be compressed is short and can be
+   predicted with good accuracy; the data can then be compressed better than
+   with the default empty dictionary.
+
+     Depending on the size of the compression data structures selected by
+   deflateInit or deflateInit2, a part of the dictionary may in effect be
+   discarded, for example if the dictionary is larger than the window size
+   provided in deflateInit or deflateInit2.  Thus the strings most likely to be
+   useful should be put at the end of the dictionary, not at the front.  In
+   addition, the current implementation of deflate will use at most the window
+   size minus 262 bytes of the provided dictionary.
+
+     Upon return of this function, strm->adler is set to the Adler-32 value
+   of the dictionary; the decompressor may later use this value to determine
+   which dictionary has been used by the compressor.  (The Adler-32 value
+   applies to the whole dictionary even if only a subset of the dictionary is
+   actually used by the compressor.) If a raw deflate was requested, then the
+   Adler-32 value is not computed and strm->adler is not set.
+
+     deflateSetDictionary returns Z_OK if success, or Z_STREAM_ERROR if a
+   parameter is invalid (e.g.  dictionary being Z_NULL) or the stream state is
+   inconsistent (for example if deflate has already been called for this stream
+   or if not at a block boundary for raw deflate).  deflateSetDictionary does
+   not perform any compression: this will be done by deflate().
+*/
+
+ZEXTERN int ZEXPORT deflateGetDictionary OF((z_streamp strm,
+                                             Bytef *dictionary,
+                                             uInt  *dictLength));
+/*
+     Returns the sliding dictionary being maintained by deflate.  dictLength is
+   set to the number of bytes in the dictionary, and that many bytes are copied
+   to dictionary.  dictionary must have enough space, where 32768 bytes is
+   always enough.  If deflateGetDictionary() is called with dictionary equal to
+   Z_NULL, then only the dictionary length is returned, and nothing is copied.
+   Similary, if dictLength is Z_NULL, then it is not set.
+
+     deflateGetDictionary() may return a length less than the window size, even
+   when more than the window size in input has been provided. It may return up
+   to 258 bytes less in that case, due to how zlib's implementation of deflate
+   manages the sliding window and lookahead for matches, where matches can be
+   up to 258 bytes long. If the application needs the last window-size bytes of
+   input, then that would need to be saved by the application outside of zlib.
+
+     deflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the
+   stream state is inconsistent.
+*/
+
+ZEXTERN int ZEXPORT deflateCopy OF((z_streamp dest,
+                                    z_streamp source));
+/*
+     Sets the destination stream as a complete copy of the source stream.
+
+     This function can be useful when several compression strategies will be
+   tried, for example when there are several ways of pre-processing the input
+   data with a filter.  The streams that will be discarded should then be freed
+   by calling deflateEnd.  Note that deflateCopy duplicates the internal
+   compression state which can be quite large, so this strategy is slow and can
+   consume lots of memory.
+
+     deflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if the source stream state was inconsistent
+   (such as zalloc being Z_NULL).  msg is left unchanged in both source and
+   destination.
+*/
+
+ZEXTERN int ZEXPORT deflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to deflateEnd followed by deflateInit, but
+   does not free and reallocate the internal compression state.  The stream
+   will leave the compression level and any other attributes that may have been
+   set unchanged.
+
+     deflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL).
+*/
+
+ZEXTERN int ZEXPORT deflateParams OF((z_streamp strm,
+                                      int level,
+                                      int strategy));
+/*
+     Dynamically update the compression level and compression strategy.  The
+   interpretation of level and strategy is as in deflateInit2().  This can be
+   used to switch between compression and straight copy of the input data, or
+   to switch to a different kind of input data requiring a different strategy.
+   If the compression approach (which is a function of the level) or the
+   strategy is changed, and if any input has been consumed in a previous
+   deflate() call, then the input available so far is compressed with the old
+   level and strategy using deflate(strm, Z_BLOCK).  There are three approaches
+   for the compression levels 0, 1..3, and 4..9 respectively.  The new level
+   and strategy will take effect at the next call of deflate().
+
+     If a deflate(strm, Z_BLOCK) is performed by deflateParams(), and it does
+   not have enough output space to complete, then the parameter change will not
+   take effect.  In this case, deflateParams() can be called again with the
+   same parameters and more output space to try again.
+
+     In order to assure a change in the parameters on the first try, the
+   deflate stream should be flushed using deflate() with Z_BLOCK or other flush
+   request until strm.avail_out is not zero, before calling deflateParams().
+   Then no more input data should be provided before the deflateParams() call.
+   If this is done, the old level and strategy will be applied to the data
+   compressed before deflateParams(), and the new level and strategy will be
+   applied to the the data compressed after deflateParams().
+
+     deflateParams returns Z_OK on success, Z_STREAM_ERROR if the source stream
+   state was inconsistent or if a parameter was invalid, or Z_BUF_ERROR if
+   there was not enough output space to complete the compression of the
+   available input data before a change in the strategy or approach.  Note that
+   in the case of a Z_BUF_ERROR, the parameters are not changed.  A return
+   value of Z_BUF_ERROR is not fatal, in which case deflateParams() can be
+   retried with more output space.
+*/
+
+ZEXTERN int ZEXPORT deflateTune OF((z_streamp strm,
+                                    int good_length,
+                                    int max_lazy,
+                                    int nice_length,
+                                    int max_chain));
+/*
+     Fine tune deflate's internal compression parameters.  This should only be
+   used by someone who understands the algorithm used by zlib's deflate for
+   searching for the best matching string, and even then only by the most
+   fanatic optimizer trying to squeeze out the last compressed bit for their
+   specific input data.  Read the deflate.c source code for the meaning of the
+   max_lazy, good_length, nice_length, and max_chain parameters.
+
+     deflateTune() can be called after deflateInit() or deflateInit2(), and
+   returns Z_OK on success, or Z_STREAM_ERROR for an invalid deflate stream.
+ */
+
+ZEXTERN uLong ZEXPORT deflateBound OF((z_streamp strm,
+                                       uLong sourceLen));
+/*
+     deflateBound() returns an upper bound on the compressed size after
+   deflation of sourceLen bytes.  It must be called after deflateInit() or
+   deflateInit2(), and after deflateSetHeader(), if used.  This would be used
+   to allocate an output buffer for deflation in a single pass, and so would be
+   called before deflate().  If that first deflate() call is provided the
+   sourceLen input bytes, an output buffer allocated to the size returned by
+   deflateBound(), and the flush value Z_FINISH, then deflate() is guaranteed
+   to return Z_STREAM_END.  Note that it is possible for the compressed size to
+   be larger than the value returned by deflateBound() if flush options other
+   than Z_FINISH or Z_NO_FLUSH are used.
+*/
+
+ZEXTERN int ZEXPORT deflatePending OF((z_streamp strm,
+                                       unsigned *pending,
+                                       int *bits));
+/*
+     deflatePending() returns the number of bytes and bits of output that have
+   been generated, but not yet provided in the available output.  The bytes not
+   provided would be due to the available output space having being consumed.
+   The number of bits of output not provided are between 0 and 7, where they
+   await more bits to join them in order to fill out a full byte.  If pending
+   or bits are Z_NULL, then those values are not set.
+
+     deflatePending returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+ */
+
+ZEXTERN int ZEXPORT deflatePrime OF((z_streamp strm,
+                                     int bits,
+                                     int value));
+/*
+     deflatePrime() inserts bits in the deflate output stream.  The intent
+   is that this function is used to start off the deflate output with the bits
+   leftover from a previous deflate stream when appending to it.  As such, this
+   function can only be used for raw deflate, and must be used before the first
+   deflate() call after a deflateInit2() or deflateReset().  bits must be less
+   than or equal to 16, and that many of the least significant bits of value
+   will be inserted in the output.
+
+     deflatePrime returns Z_OK if success, Z_BUF_ERROR if there was not enough
+   room in the internal buffer to insert the bits, or Z_STREAM_ERROR if the
+   source stream state was inconsistent.
+*/
+
+ZEXTERN int ZEXPORT deflateSetHeader OF((z_streamp strm,
+                                         gz_headerp head));
+/*
+     deflateSetHeader() provides gzip header information for when a gzip
+   stream is requested by deflateInit2().  deflateSetHeader() may be called
+   after deflateInit2() or deflateReset() and before the first call of
+   deflate().  The text, time, os, extra field, name, and comment information
+   in the provided gz_header structure are written to the gzip header (xflag is
+   ignored -- the extra flags are set according to the compression level).  The
+   caller must assure that, if not Z_NULL, name and comment are terminated with
+   a zero byte, and that if extra is not Z_NULL, that extra_len bytes are
+   available there.  If hcrc is true, a gzip header crc is included.  Note that
+   the current versions of the command-line version of gzip (up through version
+   1.3.x) do not support header crc's, and will report that it is a "multi-part
+   gzip file" and give up.
+
+     If deflateSetHeader is not used, the default gzip header has text false,
+   the time set to zero, and os set to 255, with no extra, name, or comment
+   fields.  The gzip header is returned to the default state by deflateReset().
+
+     deflateSetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+/*
+ZEXTERN int ZEXPORT inflateInit2 OF((z_streamp strm,
+                                     int  windowBits));
+
+     This is another version of inflateInit with an extra parameter.  The
+   fields next_in, avail_in, zalloc, zfree and opaque must be initialized
+   before by the caller.
+
+     The windowBits parameter is the base two logarithm of the maximum window
+   size (the size of the history buffer).  It should be in the range 8..15 for
+   this version of the library.  The default value is 15 if inflateInit is used
+   instead.  windowBits must be greater than or equal to the windowBits value
+   provided to deflateInit2() while compressing, or it must be equal to 15 if
+   deflateInit2() was not used.  If a compressed stream with a larger window
+   size is given as input, inflate() will return with the error code
+   Z_DATA_ERROR instead of trying to allocate a larger window.
+
+     windowBits can also be zero to request that inflate use the window size in
+   the zlib header of the compressed stream.
+
+     windowBits can also be -8..-15 for raw inflate.  In this case, -windowBits
+   determines the window size.  inflate() will then process raw deflate data,
+   not looking for a zlib or gzip header, not generating a check value, and not
+   looking for any check values for comparison at the end of the stream.  This
+   is for use with other formats that use the deflate compressed data format
+   such as zip.  Those formats provide their own check values.  If a custom
+   format is developed using the raw deflate format for compressed data, it is
+   recommended that a check value such as an Adler-32 or a CRC-32 be applied to
+   the uncompressed data as is done in the zlib, gzip, and zip formats.  For
+   most applications, the zlib format should be used as is.  Note that comments
+   above on the use in deflateInit2() applies to the magnitude of windowBits.
+
+     windowBits can also be greater than 15 for optional gzip decoding.  Add
+   32 to windowBits to enable zlib and gzip decoding with automatic header
+   detection, or add 16 to decode only the gzip format (the zlib format will
+   return a Z_DATA_ERROR).  If a gzip stream is being decoded, strm->adler is a
+   CRC-32 instead of an Adler-32.  Unlike the gunzip utility and gzread() (see
+   below), inflate() will not automatically decode concatenated gzip streams.
+   inflate() will return Z_STREAM_END at the end of the gzip stream.  The state
+   would need to be reset to continue decoding a subsequent gzip stream.
+
+     inflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_VERSION_ERROR if the zlib library version is incompatible with the
+   version assumed by the caller, or Z_STREAM_ERROR if the parameters are
+   invalid, such as a null pointer to the structure.  msg is set to null if
+   there is no error message.  inflateInit2 does not perform any decompression
+   apart from possibly reading the zlib header if present: actual decompression
+   will be done by inflate().  (So next_in and avail_in may be modified, but
+   next_out and avail_out are unused and unchanged.) The current implementation
+   of inflateInit2() does not process any header information -- that is
+   deferred until inflate() is called.
+*/
+
+ZEXTERN int ZEXPORT inflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the decompression dictionary from the given uncompressed byte
+   sequence.  This function must be called immediately after a call of inflate,
+   if that call returned Z_NEED_DICT.  The dictionary chosen by the compressor
+   can be determined from the Adler-32 value returned by that call of inflate.
+   The compressor and decompressor must use exactly the same dictionary (see
+   deflateSetDictionary).  For raw inflate, this function can be called at any
+   time to set the dictionary.  If the provided dictionary is smaller than the
+   window and there is already data in the window, then the provided dictionary
+   will amend what's there.  The application must insure that the dictionary
+   that was used for compression is provided.
+
+     inflateSetDictionary returns Z_OK if success, Z_STREAM_ERROR if a
+   parameter is invalid (e.g.  dictionary being Z_NULL) or the stream state is
+   inconsistent, Z_DATA_ERROR if the given dictionary doesn't match the
+   expected one (incorrect Adler-32 value).  inflateSetDictionary does not
+   perform any decompression: this will be done by subsequent calls of
+   inflate().
+*/
+
+ZEXTERN int ZEXPORT inflateGetDictionary OF((z_streamp strm,
+                                             Bytef *dictionary,
+                                             uInt  *dictLength));
+/*
+     Returns the sliding dictionary being maintained by inflate.  dictLength is
+   set to the number of bytes in the dictionary, and that many bytes are copied
+   to dictionary.  dictionary must have enough space, where 32768 bytes is
+   always enough.  If inflateGetDictionary() is called with dictionary equal to
+   Z_NULL, then only the dictionary length is returned, and nothing is copied.
+   Similary, if dictLength is Z_NULL, then it is not set.
+
+     inflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the
+   stream state is inconsistent.
+*/
+
+ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm));
+/*
+     Skips invalid compressed data until a possible full flush point (see above
+   for the description of deflate with Z_FULL_FLUSH) can be found, or until all
+   available input is skipped.  No output is provided.
+
+     inflateSync searches for a 00 00 FF FF pattern in the compressed data.
+   All full flush points have this pattern, but not all occurrences of this
+   pattern are full flush points.
+
+     inflateSync returns Z_OK if a possible full flush point has been found,
+   Z_BUF_ERROR if no more input was provided, Z_DATA_ERROR if no flush point
+   has been found, or Z_STREAM_ERROR if the stream structure was inconsistent.
+   In the success case, the application may save the current current value of
+   total_in which indicates where valid compressed data was found.  In the
+   error case, the application may repeatedly call inflateSync, providing more
+   input each time, until success or end of the input data.
+*/
+
+ZEXTERN int ZEXPORT inflateCopy OF((z_streamp dest,
+                                    z_streamp source));
+/*
+     Sets the destination stream as a complete copy of the source stream.
+
+     This function can be useful when randomly accessing a large stream.  The
+   first pass through the stream can periodically record the inflate state,
+   allowing restarting inflate at those points when randomly accessing the
+   stream.
+
+     inflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if the source stream state was inconsistent
+   (such as zalloc being Z_NULL).  msg is left unchanged in both source and
+   destination.
+*/
+
+ZEXTERN int ZEXPORT inflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to inflateEnd followed by inflateInit,
+   but does not free and reallocate the internal decompression state.  The
+   stream will keep attributes that may have been set by inflateInit2.
+
+     inflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL).
+*/
+
+ZEXTERN int ZEXPORT inflateReset2 OF((z_streamp strm,
+                                      int windowBits));
+/*
+     This function is the same as inflateReset, but it also permits changing
+   the wrap and window size requests.  The windowBits parameter is interpreted
+   the same as it is for inflateInit2.  If the window size is changed, then the
+   memory allocated for the window is freed, and the window will be reallocated
+   by inflate() if needed.
+
+     inflateReset2 returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL), or if
+   the windowBits parameter is invalid.
+*/
+
+ZEXTERN int ZEXPORT inflatePrime OF((z_streamp strm,
+                                     int bits,
+                                     int value));
+/*
+     This function inserts bits in the inflate input stream.  The intent is
+   that this function is used to start inflating at a bit position in the
+   middle of a byte.  The provided bits will be used before any bytes are used
+   from next_in.  This function should only be used with raw inflate, and
+   should be used before the first inflate() call after inflateInit2() or
+   inflateReset().  bits must be less than or equal to 16, and that many of the
+   least significant bits of value will be inserted in the input.
+
+     If bits is negative, then the input stream bit buffer is emptied.  Then
+   inflatePrime() can be called again to put bits in the buffer.  This is used
+   to clear out bits leftover after feeding inflate a block description prior
+   to feeding inflate codes.
+
+     inflatePrime returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+ZEXTERN long ZEXPORT inflateMark OF((z_streamp strm));
+/*
+     This function returns two values, one in the lower 16 bits of the return
+   value, and the other in the remaining upper bits, obtained by shifting the
+   return value down 16 bits.  If the upper value is -1 and the lower value is
+   zero, then inflate() is currently decoding information outside of a block.
+   If the upper value is -1 and the lower value is non-zero, then inflate is in
+   the middle of a stored block, with the lower value equaling the number of
+   bytes from the input remaining to copy.  If the upper value is not -1, then
+   it is the number of bits back from the current bit position in the input of
+   the code (literal or length/distance pair) currently being processed.  In
+   that case the lower value is the number of bytes already emitted for that
+   code.
+
+     A code is being processed if inflate is waiting for more input to complete
+   decoding of the code, or if it has completed decoding but is waiting for
+   more output space to write the literal or match data.
+
+     inflateMark() is used to mark locations in the input data for random
+   access, which may be at bit positions, and to note those cases where the
+   output of a code may span boundaries of random access blocks.  The current
+   location in the input stream can be determined from avail_in and data_type
+   as noted in the description for the Z_BLOCK flush parameter for inflate.
+
+     inflateMark returns the value noted above, or -65536 if the provided
+   source stream state was inconsistent.
+*/
+
+ZEXTERN int ZEXPORT inflateGetHeader OF((z_streamp strm,
+                                         gz_headerp head));
+/*
+     inflateGetHeader() requests that gzip header information be stored in the
+   provided gz_header structure.  inflateGetHeader() may be called after
+   inflateInit2() or inflateReset(), and before the first call of inflate().
+   As inflate() processes the gzip stream, head->done is zero until the header
+   is completed, at which time head->done is set to one.  If a zlib stream is
+   being decoded, then head->done is set to -1 to indicate that there will be
+   no gzip header information forthcoming.  Note that Z_BLOCK or Z_TREES can be
+   used to force inflate() to return immediately after header processing is
+   complete and before any actual data is decompressed.
+
+     The text, time, xflags, and os fields are filled in with the gzip header
+   contents.  hcrc is set to true if there is a header CRC.  (The header CRC
+   was valid if done is set to one.) If extra is not Z_NULL, then extra_max
+   contains the maximum number of bytes to write to extra.  Once done is true,
+   extra_len contains the actual extra field length, and extra contains the
+   extra field, or that field truncated if extra_max is less than extra_len.
+   If name is not Z_NULL, then up to name_max characters are written there,
+   terminated with a zero unless the length is greater than name_max.  If
+   comment is not Z_NULL, then up to comm_max characters are written there,
+   terminated with a zero unless the length is greater than comm_max.  When any
+   of extra, name, or comment are not Z_NULL and the respective field is not
+   present in the header, then that field is set to Z_NULL to signal its
+   absence.  This allows the use of deflateSetHeader() with the returned
+   structure to duplicate the header.  However if those fields are set to
+   allocated memory, then the application will need to save those pointers
+   elsewhere so that they can be eventually freed.
+
+     If inflateGetHeader is not used, then the header information is simply
+   discarded.  The header is always checked for validity, including the header
+   CRC if present.  inflateReset() will reset the process to discard the header
+   information.  The application would need to call inflateGetHeader() again to
+   retrieve the header from the next gzip stream.
+
+     inflateGetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+/*
+ZEXTERN int ZEXPORT inflateBackInit OF((z_streamp strm, int windowBits,
+                                        unsigned char FAR *window));
+
+     Initialize the internal stream state for decompression using inflateBack()
+   calls.  The fields zalloc, zfree and opaque in strm must be initialized
+   before the call.  If zalloc and zfree are Z_NULL, then the default library-
+   derived memory allocation routines are used.  windowBits is the base two
+   logarithm of the window size, in the range 8..15.  window is a caller
+   supplied buffer of that size.  Except for special applications where it is
+   assured that deflate was used with small window sizes, windowBits must be 15
+   and a 32K byte window must be supplied to be able to decompress general
+   deflate streams.
+
+     See inflateBack() for the usage of these routines.
+
+     inflateBackInit will return Z_OK on success, Z_STREAM_ERROR if any of
+   the parameters are invalid, Z_MEM_ERROR if the internal state could not be
+   allocated, or Z_VERSION_ERROR if the version of the library does not match
+   the version of the header file.
+*/
+
+typedef unsigned (*in_func) OF((void FAR *,
+                                z_const unsigned char FAR * FAR *));
+typedef int (*out_func) OF((void FAR *, unsigned char FAR *, unsigned));
+
+ZEXTERN int ZEXPORT inflateBack OF((z_streamp strm,
+                                    in_func in, void FAR *in_desc,
+                                    out_func out, void FAR *out_desc));
+/*
+     inflateBack() does a raw inflate with a single call using a call-back
+   interface for input and output.  This is potentially more efficient than
+   inflate() for file i/o applications, in that it avoids copying between the
+   output and the sliding window by simply making the window itself the output
+   buffer.  inflate() can be faster on modern CPUs when used with large
+   buffers.  inflateBack() trusts the application to not change the output
+   buffer passed by the output function, at least until inflateBack() returns.
+
+     inflateBackInit() must be called first to allocate the internal state
+   and to initialize the state with the user-provided window buffer.
+   inflateBack() may then be used multiple times to inflate a complete, raw
+   deflate stream with each call.  inflateBackEnd() is then called to free the
+   allocated state.
+
+     A raw deflate stream is one with no zlib or gzip header or trailer.
+   This routine would normally be used in a utility that reads zip or gzip
+   files and writes out uncompressed files.  The utility would decode the
+   header and process the trailer on its own, hence this routine expects only
+   the raw deflate stream to decompress.  This is different from the default
+   behavior of inflate(), which expects a zlib header and trailer around the
+   deflate stream.
+
+     inflateBack() uses two subroutines supplied by the caller that are then
+   called by inflateBack() for input and output.  inflateBack() calls those
+   routines until it reads a complete deflate stream and writes out all of the
+   uncompressed data, or until it encounters an error.  The function's
+   parameters and return types are defined above in the in_func and out_func
+   typedefs.  inflateBack() will call in(in_desc, &buf) which should return the
+   number of bytes of provided input, and a pointer to that input in buf.  If
+   there is no input available, in() must return zero -- buf is ignored in that
+   case -- and inflateBack() will return a buffer error.  inflateBack() will
+   call out(out_desc, buf, len) to write the uncompressed data buf[0..len-1].
+   out() should return zero on success, or non-zero on failure.  If out()
+   returns non-zero, inflateBack() will return with an error.  Neither in() nor
+   out() are permitted to change the contents of the window provided to
+   inflateBackInit(), which is also the buffer that out() uses to write from.
+   The length written by out() will be at most the window size.  Any non-zero
+   amount of input may be provided by in().
+
+     For convenience, inflateBack() can be provided input on the first call by
+   setting strm->next_in and strm->avail_in.  If that input is exhausted, then
+   in() will be called.  Therefore strm->next_in must be initialized before
+   calling inflateBack().  If strm->next_in is Z_NULL, then in() will be called
+   immediately for input.  If strm->next_in is not Z_NULL, then strm->avail_in
+   must also be initialized, and then if strm->avail_in is not zero, input will
+   initially be taken from strm->next_in[0 ..  strm->avail_in - 1].
+
+     The in_desc and out_desc parameters of inflateBack() is passed as the
+   first parameter of in() and out() respectively when they are called.  These
+   descriptors can be optionally used to pass any information that the caller-
+   supplied in() and out() functions need to do their job.
+
+     On return, inflateBack() will set strm->next_in and strm->avail_in to
+   pass back any unused input that was provided by the last in() call.  The
+   return values of inflateBack() can be Z_STREAM_END on success, Z_BUF_ERROR
+   if in() or out() returned an error, Z_DATA_ERROR if there was a format error
+   in the deflate stream (in which case strm->msg is set to indicate the nature
+   of the error), or Z_STREAM_ERROR if the stream was not properly initialized.
+   In the case of Z_BUF_ERROR, an input or output error can be distinguished
+   using strm->next_in which will be Z_NULL only if in() returned an error.  If
+   strm->next_in is not Z_NULL, then the Z_BUF_ERROR was due to out() returning
+   non-zero.  (in() will always be called before out(), so strm->next_in is
+   assured to be defined if out() returns non-zero.)  Note that inflateBack()
+   cannot return Z_OK.
+*/
+
+ZEXTERN int ZEXPORT inflateBackEnd OF((z_streamp strm));
+/*
+     All memory allocated by inflateBackInit() is freed.
+
+     inflateBackEnd() returns Z_OK on success, or Z_STREAM_ERROR if the stream
+   state was inconsistent.
+*/
+
+ZEXTERN uLong ZEXPORT zlibCompileFlags OF((void));
+/* Return flags indicating compile-time options.
+
+    Type sizes, two bits each, 00 = 16 bits, 01 = 32, 10 = 64, 11 = other:
+     1.0: size of uInt
+     3.2: size of uLong
+     5.4: size of voidpf (pointer)
+     7.6: size of z_off_t
+
+    Compiler, assembler, and debug options:
+     8: ZLIB_DEBUG
+     9: ASMV or ASMINF -- use ASM code
+     10: ZLIB_WINAPI -- exported functions use the WINAPI calling convention
+     11: 0 (reserved)
+
+    One-time table building (smaller code, but not thread-safe if true):
+     12: BUILDFIXED -- build static block decoding tables when needed
+     13: DYNAMIC_CRC_TABLE -- build CRC calculation tables when needed
+     14,15: 0 (reserved)
+
+    Library content (indicates missing functionality):
+     16: NO_GZCOMPRESS -- gz* functions cannot compress (to avoid linking
+                          deflate code when not needed)
+     17: NO_GZIP -- deflate can't write gzip streams, and inflate can't detect
+                    and decode gzip streams (to avoid linking crc code)
+     18-19: 0 (reserved)
+
+    Operation variations (changes in library functionality):
+     20: PKZIP_BUG_WORKAROUND -- slightly more permissive inflate
+     21: FASTEST -- deflate algorithm with only one, lowest compression level
+     22,23: 0 (reserved)
+
+    The sprintf variant used by gzprintf (zero is best):
+     24: 0 = vs*, 1 = s* -- 1 means limited to 20 arguments after the format
+     25: 0 = *nprintf, 1 = *printf -- 1 means gzprintf() not secure!
+     26: 0 = returns value, 1 = void -- 1 means inferred string length returned
+
+    Remainder:
+     27-31: 0 (reserved)
+ */
+
+#ifndef Z_SOLO
+
+                        /* utility functions */
+
+/*
+     The following utility functions are implemented on top of the basic
+   stream-oriented functions.  To simplify the interface, some default options
+   are assumed (compression level and memory usage, standard memory allocation
+   functions).  The source code of these utility functions can be modified if
+   you need special options.
+*/
+
+ZEXTERN int ZEXPORT compress OF((Bytef *dest,   uLongf *destLen,
+                                 const Bytef *source, uLong sourceLen));
+/*
+     Compresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer.  Upon entry, destLen is the total size
+   of the destination buffer, which must be at least the value returned by
+   compressBound(sourceLen).  Upon exit, destLen is the actual size of the
+   compressed data.  compress() is equivalent to compress2() with a level
+   parameter of Z_DEFAULT_COMPRESSION.
+
+     compress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer.
+*/
+
+ZEXTERN int ZEXPORT compress2 OF((Bytef *dest,   uLongf *destLen,
+                                  const Bytef *source, uLong sourceLen,
+                                  int level));
+/*
+     Compresses the source buffer into the destination buffer.  The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer.  Upon entry, destLen is the total size of the
+   destination buffer, which must be at least the value returned by
+   compressBound(sourceLen).  Upon exit, destLen is the actual size of the
+   compressed data.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+
+ZEXTERN uLong ZEXPORT compressBound OF((uLong sourceLen));
+/*
+     compressBound() returns an upper bound on the compressed size after
+   compress() or compress2() on sourceLen bytes.  It would be used before a
+   compress() or compress2() call to allocate the destination buffer.
+*/
+
+ZEXTERN int ZEXPORT uncompress OF((Bytef *dest,   uLongf *destLen,
+                                   const Bytef *source, uLong sourceLen));
+/*
+     Decompresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer.  Upon entry, destLen is the total size
+   of the destination buffer, which must be large enough to hold the entire
+   uncompressed data.  (The size of the uncompressed data must have been saved
+   previously by the compressor and transmitted to the decompressor by some
+   mechanism outside the scope of this compression library.) Upon exit, destLen
+   is the actual size of the uncompressed data.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer, or Z_DATA_ERROR if the input data was corrupted or incomplete.  In
+   the case where there is not enough room, uncompress() will fill the output
+   buffer with the uncompressed data up to that point.
+*/
+
+ZEXTERN int ZEXPORT uncompress2 OF((Bytef *dest,   uLongf *destLen,
+                                    const Bytef *source, uLong *sourceLen));
+/*
+     Same as uncompress, except that sourceLen is a pointer, where the
+   length of the source is *sourceLen.  On return, *sourceLen is the number of
+   source bytes consumed.
+*/
+
+                        /* gzip file access functions */
+
+/*
+     This library supports reading and writing files in gzip (.gz) format with
+   an interface similar to that of stdio, using the functions that start with
+   "gz".  The gzip format is different from the zlib format.  gzip is a gzip
+   wrapper, documented in RFC 1952, wrapped around a deflate stream.
+*/
+
+typedef struct gzFile_s *gzFile;    /* semi-opaque gzip file descriptor */
+
+/*
+ZEXTERN gzFile ZEXPORT gzopen OF((const char *path, const char *mode));
+
+     Opens a gzip (.gz) file for reading or writing.  The mode parameter is as
+   in fopen ("rb" or "wb") but can also include a compression level ("wb9") or
+   a strategy: 'f' for filtered data as in "wb6f", 'h' for Huffman-only
+   compression as in "wb1h", 'R' for run-length encoding as in "wb1R", or 'F'
+   for fixed code compression as in "wb9F".  (See the description of
+   deflateInit2 for more information about the strategy parameter.)  'T' will
+   request transparent writing or appending with no compression and not using
+   the gzip format.
+
+     "a" can be used instead of "w" to request that the gzip stream that will
+   be written be appended to the file.  "+" will result in an error, since
+   reading and writing to the same gzip file is not supported.  The addition of
+   "x" when writing will create the file exclusively, which fails if the file
+   already exists.  On systems that support it, the addition of "e" when
+   reading or writing will set the flag to close the file on an execve() call.
+
+     These functions, as well as gzip, will read and decode a sequence of gzip
+   streams in a file.  The append function of gzopen() can be used to create
+   such a file.  (Also see gzflush() for another way to do this.)  When
+   appending, gzopen does not test whether the file begins with a gzip stream,
+   nor does it look for the end of the gzip streams to begin appending.  gzopen
+   will simply append a gzip stream to the existing file.
+
+     gzopen can be used to read a file which is not in gzip format; in this
+   case gzread will directly read from the file without decompression.  When
+   reading, this will be detected automatically by looking for the magic two-
+   byte gzip header.
+
+     gzopen returns NULL if the file could not be opened, if there was
+   insufficient memory to allocate the gzFile state, or if an invalid mode was
+   specified (an 'r', 'w', or 'a' was not provided, or '+' was provided).
+   errno can be checked to determine if the reason gzopen failed was that the
+   file could not be opened.
+*/
+
+ZEXTERN gzFile ZEXPORT gzdopen OF((int fd, const char *mode));
+/*
+     gzdopen associates a gzFile with the file descriptor fd.  File descriptors
+   are obtained from calls like open, dup, creat, pipe or fileno (if the file
+   has been previously opened with fopen).  The mode parameter is as in gzopen.
+
+     The next call of gzclose on the returned gzFile will also close the file
+   descriptor fd, just like fclose(fdopen(fd, mode)) closes the file descriptor
+   fd.  If you want to keep fd open, use fd = dup(fd_keep); gz = gzdopen(fd,
+   mode);.  The duplicated descriptor should be saved to avoid a leak, since
+   gzdopen does not close fd if it fails.  If you are using fileno() to get the
+   file descriptor from a FILE *, then you will have to use dup() to avoid
+   double-close()ing the file descriptor.  Both gzclose() and fclose() will
+   close the associated file descriptor, so they need to have different file
+   descriptors.
+
+     gzdopen returns NULL if there was insufficient memory to allocate the
+   gzFile state, if an invalid mode was specified (an 'r', 'w', or 'a' was not
+   provided, or '+' was provided), or if fd is -1.  The file descriptor is not
+   used until the next gz* read, write, seek, or close operation, so gzdopen
+   will not detect if fd is invalid (unless fd is -1).
+*/
+
+ZEXTERN int ZEXPORT gzbuffer OF((gzFile file, unsigned size));
+/*
+     Set the internal buffer size used by this library's functions.  The
+   default buffer size is 8192 bytes.  This function must be called after
+   gzopen() or gzdopen(), and before any other calls that read or write the
+   file.  The buffer memory allocation is always deferred to the first read or
+   write.  Three times that size in buffer space is allocated.  A larger buffer
+   size of, for example, 64K or 128K bytes will noticeably increase the speed
+   of decompression (reading).
+
+     The new buffer size also affects the maximum length for gzprintf().
+
+     gzbuffer() returns 0 on success, or -1 on failure, such as being called
+   too late.
+*/
+
+ZEXTERN int ZEXPORT gzsetparams OF((gzFile file, int level, int strategy));
+/*
+     Dynamically update the compression level or strategy.  See the description
+   of deflateInit2 for the meaning of these parameters.  Previously provided
+   data is flushed before the parameter change.
+
+     gzsetparams returns Z_OK if success, Z_STREAM_ERROR if the file was not
+   opened for writing, Z_ERRNO if there is an error writing the flushed data,
+   or Z_MEM_ERROR if there is a memory allocation error.
+*/
+
+ZEXTERN int ZEXPORT gzread OF((gzFile file, voidp buf, unsigned len));
+/*
+     Reads the given number of uncompressed bytes from the compressed file.  If
+   the input file is not in gzip format, gzread copies the given number of
+   bytes into the buffer directly from the file.
+
+     After reaching the end of a gzip stream in the input, gzread will continue
+   to read, looking for another gzip stream.  Any number of gzip streams may be
+   concatenated in the input file, and will all be decompressed by gzread().
+   If something other than a gzip stream is encountered after a gzip stream,
+   that remaining trailing garbage is ignored (and no error is returned).
+
+     gzread can be used to read a gzip file that is being concurrently written.
+   Upon reaching the end of the input, gzread will return with the available
+   data.  If the error code returned by gzerror is Z_OK or Z_BUF_ERROR, then
+   gzclearerr can be used to clear the end of file indicator in order to permit
+   gzread to be tried again.  Z_OK indicates that a gzip stream was completed
+   on the last gzread.  Z_BUF_ERROR indicates that the input file ended in the
+   middle of a gzip stream.  Note that gzread does not return -1 in the event
+   of an incomplete gzip stream.  This error is deferred until gzclose(), which
+   will return Z_BUF_ERROR if the last gzread ended in the middle of a gzip
+   stream.  Alternatively, gzerror can be used before gzclose to detect this
+   case.
+
+     gzread returns the number of uncompressed bytes actually read, less than
+   len for end of file, or -1 for error.  If len is too large to fit in an int,
+   then nothing is read, -1 is returned, and the error state is set to
+   Z_STREAM_ERROR.
+*/
+
+ZEXTERN z_size_t ZEXPORT gzfread OF((voidp buf, z_size_t size, z_size_t nitems,
+                                     gzFile file));
+/*
+     Read up to nitems items of size size from file to buf, otherwise operating
+   as gzread() does.  This duplicates the interface of stdio's fread(), with
+   size_t request and return types.  If the library defines size_t, then
+   z_size_t is identical to size_t.  If not, then z_size_t is an unsigned
+   integer type that can contain a pointer.
+
+     gzfread() returns the number of full items read of size size, or zero if
+   the end of the file was reached and a full item could not be read, or if
+   there was an error.  gzerror() must be consulted if zero is returned in
+   order to determine if there was an error.  If the multiplication of size and
+   nitems overflows, i.e. the product does not fit in a z_size_t, then nothing
+   is read, zero is returned, and the error state is set to Z_STREAM_ERROR.
+
+     In the event that the end of file is reached and only a partial item is
+   available at the end, i.e. the remaining uncompressed data length is not a
+   multiple of size, then the final partial item is nevetheless read into buf
+   and the end-of-file flag is set.  The length of the partial item read is not
+   provided, but could be inferred from the result of gztell().  This behavior
+   is the same as the behavior of fread() implementations in common libraries,
+   but it prevents the direct use of gzfread() to read a concurrently written
+   file, reseting and retrying on end-of-file, when size is not 1.
+*/
+
+ZEXTERN int ZEXPORT gzwrite OF((gzFile file,
+                                voidpc buf, unsigned len));
+/*
+     Writes the given number of uncompressed bytes into the compressed file.
+   gzwrite returns the number of uncompressed bytes written or 0 in case of
+   error.
+*/
+
+ZEXTERN z_size_t ZEXPORT gzfwrite OF((voidpc buf, z_size_t size,
+                                      z_size_t nitems, gzFile file));
+/*
+     gzfwrite() writes nitems items of size size from buf to file, duplicating
+   the interface of stdio's fwrite(), with size_t request and return types.  If
+   the library defines size_t, then z_size_t is identical to size_t.  If not,
+   then z_size_t is an unsigned integer type that can contain a pointer.
+
+     gzfwrite() returns the number of full items written of size size, or zero
+   if there was an error.  If the multiplication of size and nitems overflows,
+   i.e. the product does not fit in a z_size_t, then nothing is written, zero
+   is returned, and the error state is set to Z_STREAM_ERROR.
+*/
+
+ZEXTERN int ZEXPORTVA gzprintf Z_ARG((gzFile file, const char *format, ...));
+/*
+     Converts, formats, and writes the arguments to the compressed file under
+   control of the format string, as in fprintf.  gzprintf returns the number of
+   uncompressed bytes actually written, or a negative zlib error code in case
+   of error.  The number of uncompressed bytes written is limited to 8191, or
+   one less than the buffer size given to gzbuffer().  The caller should assure
+   that this limit is not exceeded.  If it is exceeded, then gzprintf() will
+   return an error (0) with nothing written.  In this case, there may also be a
+   buffer overflow with unpredictable consequences, which is possible only if
+   zlib was compiled with the insecure functions sprintf() or vsprintf()
+   because the secure snprintf() or vsnprintf() functions were not available.
+   This can be determined using zlibCompileFlags().
+*/
+
+ZEXTERN int ZEXPORT gzputs OF((gzFile file, const char *s));
+/*
+     Writes the given null-terminated string to the compressed file, excluding
+   the terminating null character.
+
+     gzputs returns the number of characters written, or -1 in case of error.
+*/
+
+ZEXTERN char * ZEXPORT gzgets OF((gzFile file, char *buf, int len));
+/*
+     Reads bytes from the compressed file until len-1 characters are read, or a
+   newline character is read and transferred to buf, or an end-of-file
+   condition is encountered.  If any characters are read or if len == 1, the
+   string is terminated with a null character.  If no characters are read due
+   to an end-of-file or len < 1, then the buffer is left untouched.
+
+     gzgets returns buf which is a null-terminated string, or it returns NULL
+   for end-of-file or in case of error.  If there was an error, the contents at
+   buf are indeterminate.
+*/
+
+ZEXTERN int ZEXPORT gzputc OF((gzFile file, int c));
+/*
+     Writes c, converted to an unsigned char, into the compressed file.  gzputc
+   returns the value that was written, or -1 in case of error.
+*/
+
+ZEXTERN int ZEXPORT gzgetc OF((gzFile file));
+/*
+     Reads one byte from the compressed file.  gzgetc returns this byte or -1
+   in case of end of file or error.  This is implemented as a macro for speed.
+   As such, it does not do all of the checking the other functions do.  I.e.
+   it does not check to see if file is NULL, nor whether the structure file
+   points to has been clobbered or not.
+*/
+
+ZEXTERN int ZEXPORT gzungetc OF((int c, gzFile file));
+/*
+     Push one character back onto the stream to be read as the first character
+   on the next read.  At least one character of push-back is allowed.
+   gzungetc() returns the character pushed, or -1 on failure.  gzungetc() will
+   fail if c is -1, and may fail if a character has been pushed but not read
+   yet.  If gzungetc is used immediately after gzopen or gzdopen, at least the
+   output buffer size of pushed characters is allowed.  (See gzbuffer above.)
+   The pushed character will be discarded if the stream is repositioned with
+   gzseek() or gzrewind().
+*/
+
+ZEXTERN int ZEXPORT gzflush OF((gzFile file, int flush));
+/*
+     Flushes all pending output into the compressed file.  The parameter flush
+   is as in the deflate() function.  The return value is the zlib error number
+   (see function gzerror below).  gzflush is only permitted when writing.
+
+     If the flush parameter is Z_FINISH, the remaining data is written and the
+   gzip stream is completed in the output.  If gzwrite() is called again, a new
+   gzip stream will be started in the output.  gzread() is able to read such
+   concatenated gzip streams.
+
+     gzflush should be called only when strictly necessary because it will
+   degrade compression if called too often.
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile file,
+                                   z_off_t offset, int whence));
+
+     Sets the starting position for the next gzread or gzwrite on the given
+   compressed file.  The offset represents a number of bytes in the
+   uncompressed data stream.  The whence parameter is defined as in lseek(2);
+   the value SEEK_END is not supported.
+
+     If the file is opened for reading, this function is emulated but can be
+   extremely slow.  If the file is opened for writing, only forward seeks are
+   supported; gzseek then compresses a sequence of zeroes up to the new
+   starting position.
+
+     gzseek returns the resulting offset location as measured in bytes from
+   the beginning of the uncompressed stream, or -1 in case of error, in
+   particular if the file is opened for writing and the new starting position
+   would be before the current position.
+*/
+
+ZEXTERN int ZEXPORT    gzrewind OF((gzFile file));
+/*
+     Rewinds the given file. This function is supported only for reading.
+
+     gzrewind(file) is equivalent to (int)gzseek(file, 0L, SEEK_SET)
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT    gztell OF((gzFile file));
+
+     Returns the starting position for the next gzread or gzwrite on the given
+   compressed file.  This position represents a number of bytes in the
+   uncompressed data stream, and is zero when starting, even if appending or
+   reading a gzip stream from the middle of a file using gzdopen().
+
+     gztell(file) is equivalent to gzseek(file, 0L, SEEK_CUR)
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile file));
+
+     Returns the current offset in the file being read or written.  This offset
+   includes the count of bytes that precede the gzip stream, for example when
+   appending or when using gzdopen() for reading.  When reading, the offset
+   does not include as yet unused buffered input.  This information can be used
+   for a progress indicator.  On error, gzoffset() returns -1.
+*/
+
+ZEXTERN int ZEXPORT gzeof OF((gzFile file));
+/*
+     Returns true (1) if the end-of-file indicator has been set while reading,
+   false (0) otherwise.  Note that the end-of-file indicator is set only if the
+   read tried to go past the end of the input, but came up short.  Therefore,
+   just like feof(), gzeof() may return false even if there is no more data to
+   read, in the event that the last read request was for the exact number of
+   bytes remaining in the input file.  This will happen if the input file size
+   is an exact multiple of the buffer size.
+
+     If gzeof() returns true, then the read functions will return no more data,
+   unless the end-of-file indicator is reset by gzclearerr() and the input file
+   has grown since the previous end of file was detected.
+*/
+
+ZEXTERN int ZEXPORT gzdirect OF((gzFile file));
+/*
+     Returns true (1) if file is being copied directly while reading, or false
+   (0) if file is a gzip stream being decompressed.
+
+     If the input file is empty, gzdirect() will return true, since the input
+   does not contain a gzip stream.
+
+     If gzdirect() is used immediately after gzopen() or gzdopen() it will
+   cause buffers to be allocated to allow reading the file to determine if it
+   is a gzip file.  Therefore if gzbuffer() is used, it should be called before
+   gzdirect().
+
+     When writing, gzdirect() returns true (1) if transparent writing was
+   requested ("wT" for the gzopen() mode), or false (0) otherwise.  (Note:
+   gzdirect() is not needed when writing.  Transparent writing must be
+   explicitly requested, so the application already knows the answer.  When
+   linking statically, using gzdirect() will include all of the zlib code for
+   gzip file reading and decompression, which may not be desired.)
+*/
+
+ZEXTERN int ZEXPORT    gzclose OF((gzFile file));
+/*
+     Flushes all pending output if necessary, closes the compressed file and
+   deallocates the (de)compression state.  Note that once file is closed, you
+   cannot call gzerror with file, since its structures have been deallocated.
+   gzclose must not be called more than once on the same file, just as free
+   must not be called more than once on the same allocation.
+
+     gzclose will return Z_STREAM_ERROR if file is not valid, Z_ERRNO on a
+   file operation error, Z_MEM_ERROR if out of memory, Z_BUF_ERROR if the
+   last read ended in the middle of a gzip stream, or Z_OK on success.
+*/
+
+ZEXTERN int ZEXPORT gzclose_r OF((gzFile file));
+ZEXTERN int ZEXPORT gzclose_w OF((gzFile file));
+/*
+     Same as gzclose(), but gzclose_r() is only for use when reading, and
+   gzclose_w() is only for use when writing or appending.  The advantage to
+   using these instead of gzclose() is that they avoid linking in zlib
+   compression or decompression code that is not used when only reading or only
+   writing respectively.  If gzclose() is used, then both compression and
+   decompression code will be included the application when linking to a static
+   zlib library.
+*/
+
+ZEXTERN const char * ZEXPORT gzerror OF((gzFile file, int *errnum));
+/*
+     Returns the error message for the last error which occurred on the given
+   compressed file.  errnum is set to zlib error number.  If an error occurred
+   in the file system and not in the compression library, errnum is set to
+   Z_ERRNO and the application may consult errno to get the exact error code.
+
+     The application must not modify the returned string.  Future calls to
+   this function may invalidate the previously returned string.  If file is
+   closed, then the string previously returned by gzerror will no longer be
+   available.
+
+     gzerror() should be used to distinguish errors from end-of-file for those
+   functions above that do not distinguish those cases in their return values.
+*/
+
+ZEXTERN void ZEXPORT gzclearerr OF((gzFile file));
+/*
+     Clears the error and end-of-file flags for file.  This is analogous to the
+   clearerr() function in stdio.  This is useful for continuing to read a gzip
+   file that is being written concurrently.
+*/
+
+#endif /* !Z_SOLO */
+
+                        /* checksum functions */
+
+/*
+     These functions are not related to compression but are exported
+   anyway because they might be useful in applications using the compression
+   library.
+*/
+
+ZEXTERN uLong ZEXPORT adler32 OF((uLong adler, const Bytef *buf, uInt len));
+/*
+     Update a running Adler-32 checksum with the bytes buf[0..len-1] and
+   return the updated checksum.  If buf is Z_NULL, this function returns the
+   required initial value for the checksum.
+
+     An Adler-32 checksum is almost as reliable as a CRC-32 but can be computed
+   much faster.
+
+   Usage example:
+
+     uLong adler = adler32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       adler = adler32(adler, buffer, length);
+     }
+     if (adler != original_adler) error();
+*/
+
+ZEXTERN uLong ZEXPORT adler32_z OF((uLong adler, const Bytef *buf,
+                                    z_size_t len));
+/*
+     Same as adler32(), but with a size_t length.
+*/
+
+/*
+ZEXTERN uLong ZEXPORT adler32_combine OF((uLong adler1, uLong adler2,
+                                          z_off_t len2));
+
+     Combine two Adler-32 checksums into one.  For two sequences of bytes, seq1
+   and seq2 with lengths len1 and len2, Adler-32 checksums were calculated for
+   each, adler1 and adler2.  adler32_combine() returns the Adler-32 checksum of
+   seq1 and seq2 concatenated, requiring only adler1, adler2, and len2.  Note
+   that the z_off_t type (like off_t) is a signed integer.  If len2 is
+   negative, the result has no meaning or utility.
+*/
+
+ZEXTERN uLong ZEXPORT crc32   OF((uLong crc, const Bytef *buf, uInt len));
+/*
+     Update a running CRC-32 with the bytes buf[0..len-1] and return the
+   updated CRC-32.  If buf is Z_NULL, this function returns the required
+   initial value for the crc.  Pre- and post-conditioning (one's complement) is
+   performed within this function so it shouldn't be done by the application.
+
+   Usage example:
+
+     uLong crc = crc32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       crc = crc32(crc, buffer, length);
+     }
+     if (crc != original_crc) error();
+*/
+
+ZEXTERN uLong ZEXPORT crc32_z OF((uLong adler, const Bytef *buf,
+                                  z_size_t len));
+/*
+     Same as crc32(), but with a size_t length.
+*/
+
+/*
+ZEXTERN uLong ZEXPORT crc32_combine OF((uLong crc1, uLong crc2, z_off_t len2));
+
+     Combine two CRC-32 check values into one.  For two sequences of bytes,
+   seq1 and seq2 with lengths len1 and len2, CRC-32 check values were
+   calculated for each, crc1 and crc2.  crc32_combine() returns the CRC-32
+   check value of seq1 and seq2 concatenated, requiring only crc1, crc2, and
+   len2.
+*/
+
+
+                        /* various hacks, don't look :) */
+
+/* deflateInit and inflateInit are macros to allow checking the zlib version
+ * and the compiler's view of z_stream:
+ */
+ZEXTERN int ZEXPORT deflateInit_ OF((z_streamp strm, int level,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT inflateInit_ OF((z_streamp strm,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT deflateInit2_ OF((z_streamp strm, int  level, int  method,
+                                      int windowBits, int memLevel,
+                                      int strategy, const char *version,
+                                      int stream_size));
+ZEXTERN int ZEXPORT inflateInit2_ OF((z_streamp strm, int  windowBits,
+                                      const char *version, int stream_size));
+ZEXTERN int ZEXPORT inflateBackInit_ OF((z_streamp strm, int windowBits,
+                                         unsigned char FAR *window,
+                                         const char *version,
+                                         int stream_size));
+#ifdef Z_PREFIX_SET
+#  define z_deflateInit(strm, level) \
+          deflateInit_((strm), (level), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_inflateInit(strm) \
+          inflateInit_((strm), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_deflateInit2(strm, level, method, windowBits, memLevel, strategy) \
+          deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\
+                        (strategy), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_inflateInit2(strm, windowBits) \
+          inflateInit2_((strm), (windowBits), ZLIB_VERSION, \
+                        (int)sizeof(z_stream))
+#  define z_inflateBackInit(strm, windowBits, window) \
+          inflateBackInit_((strm), (windowBits), (window), \
+                           ZLIB_VERSION, (int)sizeof(z_stream))
+#else
+#  define deflateInit(strm, level) \
+          deflateInit_((strm), (level), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define inflateInit(strm) \
+          inflateInit_((strm), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define deflateInit2(strm, level, method, windowBits, memLevel, strategy) \
+          deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\
+                        (strategy), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define inflateInit2(strm, windowBits) \
+          inflateInit2_((strm), (windowBits), ZLIB_VERSION, \
+                        (int)sizeof(z_stream))
+#  define inflateBackInit(strm, windowBits, window) \
+          inflateBackInit_((strm), (windowBits), (window), \
+                           ZLIB_VERSION, (int)sizeof(z_stream))
+#endif
+
+#ifndef Z_SOLO
+
+/* gzgetc() macro and its supporting function and exposed data structure.  Note
+ * that the real internal state is much larger than the exposed structure.
+ * This abbreviated structure exposes just enough for the gzgetc() macro.  The
+ * user should not mess with these exposed elements, since their names or
+ * behavior could change in the future, perhaps even capriciously.  They can
+ * only be used by the gzgetc() macro.  You have been warned.
+ */
+struct gzFile_s {
+    unsigned have;
+    unsigned char *next;
+    z_off64_t pos;
+};
+ZEXTERN int ZEXPORT gzgetc_ OF((gzFile file));  /* backward compatibility */
+#ifdef Z_PREFIX_SET
+#  undef z_gzgetc
+#  define z_gzgetc(g) \
+          ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : (gzgetc)(g))
+#else
+#  define gzgetc(g) \
+          ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : (gzgetc)(g))
+#endif
+
+/* provide 64-bit offset functions if _LARGEFILE64_SOURCE defined, and/or
+ * change the regular functions to 64 bits if _FILE_OFFSET_BITS is 64 (if
+ * both are true, the application gets the *64 functions, and the regular
+ * functions are changed to 64 bits) -- in case these are set on systems
+ * without large file support, _LFS64_LARGEFILE must also be true
+ */
+#ifdef Z_LARGE64
+   ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+   ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int));
+   ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile));
+   ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile));
+   ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off64_t));
+   ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off64_t));
+#endif
+
+#if !defined(ZLIB_INTERNAL) && defined(Z_WANT64)
+#  ifdef Z_PREFIX_SET
+#    define z_gzopen z_gzopen64
+#    define z_gzseek z_gzseek64
+#    define z_gztell z_gztell64
+#    define z_gzoffset z_gzoffset64
+#    define z_adler32_combine z_adler32_combine64
+#    define z_crc32_combine z_crc32_combine64
+#  else
+#    define gzopen gzopen64
+#    define gzseek gzseek64
+#    define gztell gztell64
+#    define gzoffset gzoffset64
+#    define adler32_combine adler32_combine64
+#    define crc32_combine crc32_combine64
+#  endif
+#  ifndef Z_LARGE64
+     ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+     ZEXTERN z_off_t ZEXPORT gzseek64 OF((gzFile, z_off_t, int));
+     ZEXTERN z_off_t ZEXPORT gztell64 OF((gzFile));
+     ZEXTERN z_off_t ZEXPORT gzoffset64 OF((gzFile));
+     ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t));
+     ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t));
+#  endif
+#else
+   ZEXTERN gzFile ZEXPORT gzopen OF((const char *, const char *));
+   ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile, z_off_t, int));
+   ZEXTERN z_off_t ZEXPORT gztell OF((gzFile));
+   ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile));
+   ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t));
+   ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t));
+#endif
+
+#else /* Z_SOLO */
+
+   ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t));
+   ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t));
+
+#endif /* !Z_SOLO */
+
+/* undocumented functions */
+ZEXTERN const char   * ZEXPORT zError           OF((int));
+ZEXTERN int            ZEXPORT inflateSyncPoint OF((z_streamp));
+ZEXTERN const z_crc_t FAR * ZEXPORT get_crc_table    OF((void));
+ZEXTERN int            ZEXPORT inflateUndermine OF((z_streamp, int));
+ZEXTERN int            ZEXPORT inflateValidate OF((z_streamp, int));
+ZEXTERN unsigned long  ZEXPORT inflateCodesUsed OF ((z_streamp));
+ZEXTERN int            ZEXPORT inflateResetKeep OF((z_streamp));
+ZEXTERN int            ZEXPORT deflateResetKeep OF((z_streamp));
+#if (defined(_WIN32) || defined(__CYGWIN__)) && !defined(Z_SOLO)
+ZEXTERN gzFile         ZEXPORT gzopen_w OF((const wchar_t *path,
+                                            const char *mode));
+#endif
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+ZEXTERN int            ZEXPORTVA gzvprintf Z_ARG((gzFile file,
+                                                  const char *format,
+                                                  va_list va));
+#  endif
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* ZLIB_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zutil.c glmark2-2021.02/android/jni/src/zlib/zutil.c
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zutil.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/zutil.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,325 @@
+/* zutil.c -- target dependent utility functions for the compression library
+ * Copyright (C) 1995-2017 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#include "zutil.h"
+#ifndef Z_SOLO
+#  include "gzguts.h"
+#endif
+
+z_const char * const z_errmsg[10] = {
+    (z_const char *)"need dictionary",     /* Z_NEED_DICT       2  */
+    (z_const char *)"stream end",          /* Z_STREAM_END      1  */
+    (z_const char *)"",                    /* Z_OK              0  */
+    (z_const char *)"file error",          /* Z_ERRNO         (-1) */
+    (z_const char *)"stream error",        /* Z_STREAM_ERROR  (-2) */
+    (z_const char *)"data error",          /* Z_DATA_ERROR    (-3) */
+    (z_const char *)"insufficient memory", /* Z_MEM_ERROR     (-4) */
+    (z_const char *)"buffer error",        /* Z_BUF_ERROR     (-5) */
+    (z_const char *)"incompatible version",/* Z_VERSION_ERROR (-6) */
+    (z_const char *)""
+};
+
+
+const char * ZEXPORT zlibVersion()
+{
+    return ZLIB_VERSION;
+}
+
+uLong ZEXPORT zlibCompileFlags()
+{
+    uLong flags;
+
+    flags = 0;
+    switch ((int)(sizeof(uInt))) {
+    case 2:     break;
+    case 4:     flags += 1;     break;
+    case 8:     flags += 2;     break;
+    default:    flags += 3;
+    }
+    switch ((int)(sizeof(uLong))) {
+    case 2:     break;
+    case 4:     flags += 1 << 2;        break;
+    case 8:     flags += 2 << 2;        break;
+    default:    flags += 3 << 2;
+    }
+    switch ((int)(sizeof(voidpf))) {
+    case 2:     break;
+    case 4:     flags += 1 << 4;        break;
+    case 8:     flags += 2 << 4;        break;
+    default:    flags += 3 << 4;
+    }
+    switch ((int)(sizeof(z_off_t))) {
+    case 2:     break;
+    case 4:     flags += 1 << 6;        break;
+    case 8:     flags += 2 << 6;        break;
+    default:    flags += 3 << 6;
+    }
+#ifdef ZLIB_DEBUG
+    flags += 1 << 8;
+#endif
+#if defined(ASMV) || defined(ASMINF)
+    flags += 1 << 9;
+#endif
+#ifdef ZLIB_WINAPI
+    flags += 1 << 10;
+#endif
+#ifdef BUILDFIXED
+    flags += 1 << 12;
+#endif
+#ifdef DYNAMIC_CRC_TABLE
+    flags += 1 << 13;
+#endif
+#ifdef NO_GZCOMPRESS
+    flags += 1L << 16;
+#endif
+#ifdef NO_GZIP
+    flags += 1L << 17;
+#endif
+#ifdef PKZIP_BUG_WORKAROUND
+    flags += 1L << 20;
+#endif
+#ifdef FASTEST
+    flags += 1L << 21;
+#endif
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifdef NO_vsnprintf
+    flags += 1L << 25;
+#    ifdef HAS_vsprintf_void
+    flags += 1L << 26;
+#    endif
+#  else
+#    ifdef HAS_vsnprintf_void
+    flags += 1L << 26;
+#    endif
+#  endif
+#else
+    flags += 1L << 24;
+#  ifdef NO_snprintf
+    flags += 1L << 25;
+#    ifdef HAS_sprintf_void
+    flags += 1L << 26;
+#    endif
+#  else
+#    ifdef HAS_snprintf_void
+    flags += 1L << 26;
+#    endif
+#  endif
+#endif
+    return flags;
+}
+
+#ifdef ZLIB_DEBUG
+#include <stdlib.h>
+#  ifndef verbose
+#    define verbose 0
+#  endif
+int ZLIB_INTERNAL z_verbose = verbose;
+
+void ZLIB_INTERNAL z_error (m)
+    char *m;
+{
+    fprintf(stderr, "%s\n", m);
+    exit(1);
+}
+#endif
+
+/* exported to allow conversion of error code to string for compress() and
+ * uncompress()
+ */
+const char * ZEXPORT zError(err)
+    int err;
+{
+    return ERR_MSG(err);
+}
+
+#if defined(_WIN32_WCE)
+    /* The Microsoft C Run-Time Library for Windows CE doesn't have
+     * errno.  We define it as a global variable to simplify porting.
+     * Its value is always 0 and should not be used.
+     */
+    int errno = 0;
+#endif
+
+#ifndef HAVE_MEMCPY
+
+void ZLIB_INTERNAL zmemcpy(dest, source, len)
+    Bytef* dest;
+    const Bytef* source;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = *source++; /* ??? to be unrolled */
+    } while (--len != 0);
+}
+
+int ZLIB_INTERNAL zmemcmp(s1, s2, len)
+    const Bytef* s1;
+    const Bytef* s2;
+    uInt  len;
+{
+    uInt j;
+
+    for (j = 0; j < len; j++) {
+        if (s1[j] != s2[j]) return 2*(s1[j] > s2[j])-1;
+    }
+    return 0;
+}
+
+void ZLIB_INTERNAL zmemzero(dest, len)
+    Bytef* dest;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = 0;  /* ??? to be unrolled */
+    } while (--len != 0);
+}
+#endif
+
+#ifndef Z_SOLO
+
+#ifdef SYS16BIT
+
+#ifdef __TURBOC__
+/* Turbo C in 16-bit mode */
+
+#  define MY_ZCALLOC
+
+/* Turbo C malloc() does not allow dynamic allocation of 64K bytes
+ * and farmalloc(64K) returns a pointer with an offset of 8, so we
+ * must fix the pointer. Warning: the pointer must be put back to its
+ * original form in order to free it, use zcfree().
+ */
+
+#define MAX_PTR 10
+/* 10*64K = 640K */
+
+local int next_ptr = 0;
+
+typedef struct ptr_table_s {
+    voidpf org_ptr;
+    voidpf new_ptr;
+} ptr_table;
+
+local ptr_table table[MAX_PTR];
+/* This table is used to remember the original form of pointers
+ * to large buffers (64K). Such pointers are normalized with a zero offset.
+ * Since MSDOS is not a preemptive multitasking OS, this table is not
+ * protected from concurrent access. This hack doesn't work anyway on
+ * a protected system like OS/2. Use Microsoft C instead.
+ */
+
+voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, unsigned items, unsigned size)
+{
+    voidpf buf;
+    ulg bsize = (ulg)items*size;
+
+    (void)opaque;
+
+    /* If we allocate less than 65520 bytes, we assume that farmalloc
+     * will return a usable pointer which doesn't have to be normalized.
+     */
+    if (bsize < 65520L) {
+        buf = farmalloc(bsize);
+        if (*(ush*)&buf != 0) return buf;
+    } else {
+        buf = farmalloc(bsize + 16L);
+    }
+    if (buf == NULL || next_ptr >= MAX_PTR) return NULL;
+    table[next_ptr].org_ptr = buf;
+
+    /* Normalize the pointer to seg:0 */
+    *((ush*)&buf+1) += ((ush)((uch*)buf-0) + 15) >> 4;
+    *(ush*)&buf = 0;
+    table[next_ptr++].new_ptr = buf;
+    return buf;
+}
+
+void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr)
+{
+    int n;
+
+    (void)opaque;
+
+    if (*(ush*)&ptr != 0) { /* object < 64K */
+        farfree(ptr);
+        return;
+    }
+    /* Find the original pointer */
+    for (n = 0; n < next_ptr; n++) {
+        if (ptr != table[n].new_ptr) continue;
+
+        farfree(table[n].org_ptr);
+        while (++n < next_ptr) {
+            table[n-1] = table[n];
+        }
+        next_ptr--;
+        return;
+    }
+    Assert(0, "zcfree: ptr not found");
+}
+
+#endif /* __TURBOC__ */
+
+
+#ifdef M_I86
+/* Microsoft C in 16-bit mode */
+
+#  define MY_ZCALLOC
+
+#if (!defined(_MSC_VER) || (_MSC_VER <= 600))
+#  define _halloc  halloc
+#  define _hfree   hfree
+#endif
+
+voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, uInt items, uInt size)
+{
+    (void)opaque;
+    return _halloc((long)items, size);
+}
+
+void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr)
+{
+    (void)opaque;
+    _hfree(ptr);
+}
+
+#endif /* M_I86 */
+
+#endif /* SYS16BIT */
+
+
+#ifndef MY_ZCALLOC /* Any system without a special alloc function */
+
+#ifndef STDC
+extern voidp  malloc OF((uInt size));
+extern voidp  calloc OF((uInt items, uInt size));
+extern void   free   OF((voidpf ptr));
+#endif
+
+voidpf ZLIB_INTERNAL zcalloc (opaque, items, size)
+    voidpf opaque;
+    unsigned items;
+    unsigned size;
+{
+    (void)opaque;
+    return sizeof(uInt) > 2 ? (voidpf)malloc(items * size) :
+                              (voidpf)calloc(items, size);
+}
+
+void ZLIB_INTERNAL zcfree (opaque, ptr)
+    voidpf opaque;
+    voidpf ptr;
+{
+    (void)opaque;
+    free(ptr);
+}
+
+#endif /* MY_ZCALLOC */
+
+#endif /* !Z_SOLO */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zutil.h glmark2-2021.02/android/jni/src/zlib/zutil.h
--- glmark2-2014.03+git20150611.fa71af2d/android/jni/src/zlib/zutil.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/jni/src/zlib/zutil.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,271 @@
+/* zutil.h -- internal interface and configuration of the compression library
+ * Copyright (C) 1995-2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id$ */
+
+#ifndef ZUTIL_H
+#define ZUTIL_H
+
+#ifdef HAVE_HIDDEN
+#  define ZLIB_INTERNAL __attribute__((visibility ("hidden")))
+#else
+#  define ZLIB_INTERNAL
+#endif
+
+#include "zlib.h"
+
+#if defined(STDC) && !defined(Z_SOLO)
+#  if !(defined(_WIN32_WCE) && defined(_MSC_VER))
+#    include <stddef.h>
+#  endif
+#  include <string.h>
+#  include <stdlib.h>
+#endif
+
+#ifdef Z_SOLO
+   typedef long ptrdiff_t;  /* guess -- will be caught if guess is wrong */
+#endif
+
+#ifndef local
+#  define local static
+#endif
+/* since "static" is used to mean two completely different things in C, we
+   define "local" for the non-static meaning of "static", for readability
+   (compile with -Dlocal if your debugger can't find static symbols) */
+
+typedef unsigned char  uch;
+typedef uch FAR uchf;
+typedef unsigned short ush;
+typedef ush FAR ushf;
+typedef unsigned long  ulg;
+
+extern z_const char * const z_errmsg[10]; /* indexed by 2-zlib_error */
+/* (size given to avoid silly warnings with Visual C++) */
+
+#define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)]
+
+#define ERR_RETURN(strm,err) \
+  return (strm->msg = ERR_MSG(err), (err))
+/* To be used only when the state is known to be valid */
+
+        /* common constants */
+
+#ifndef DEF_WBITS
+#  define DEF_WBITS MAX_WBITS
+#endif
+/* default windowBits for decompression. MAX_WBITS is for compression only */
+
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+/* default memLevel */
+
+#define STORED_BLOCK 0
+#define STATIC_TREES 1
+#define DYN_TREES    2
+/* The three kinds of block type */
+
+#define MIN_MATCH  3
+#define MAX_MATCH  258
+/* The minimum and maximum match lengths */
+
+#define PRESET_DICT 0x20 /* preset dictionary flag in zlib header */
+
+        /* target dependencies */
+
+#if defined(MSDOS) || (defined(WINDOWS) && !defined(WIN32))
+#  define OS_CODE  0x00
+#  ifndef Z_SOLO
+#    if defined(__TURBOC__) || defined(__BORLANDC__)
+#      if (__STDC__ == 1) && (defined(__LARGE__) || defined(__COMPACT__))
+         /* Allow compilation with ANSI keywords only enabled */
+         void _Cdecl farfree( void *block );
+         void *_Cdecl farmalloc( unsigned long nbytes );
+#      else
+#        include <alloc.h>
+#      endif
+#    else /* MSC or DJGPP */
+#      include <malloc.h>
+#    endif
+#  endif
+#endif
+
+#ifdef AMIGA
+#  define OS_CODE  1
+#endif
+
+#if defined(VAXC) || defined(VMS)
+#  define OS_CODE  2
+#  define F_OPEN(name, mode) \
+     fopen((name), (mode), "mbc=60", "ctx=stm", "rfm=fix", "mrs=512")
+#endif
+
+#ifdef __370__
+#  if __TARGET_LIB__ < 0x20000000
+#    define OS_CODE 4
+#  elif __TARGET_LIB__ < 0x40000000
+#    define OS_CODE 11
+#  else
+#    define OS_CODE 8
+#  endif
+#endif
+
+#if defined(ATARI) || defined(atarist)
+#  define OS_CODE  5
+#endif
+
+#ifdef OS2
+#  define OS_CODE  6
+#  if defined(M_I86) && !defined(Z_SOLO)
+#    include <malloc.h>
+#  endif
+#endif
+
+#if defined(MACOS) || defined(TARGET_OS_MAC)
+#  define OS_CODE  7
+#  ifndef Z_SOLO
+#    if defined(__MWERKS__) && __dest_os != __be_os && __dest_os != __win32_os
+#      include <unix.h> /* for fdopen */
+#    else
+#      ifndef fdopen
+#        define fdopen(fd,mode) NULL /* No fdopen() */
+#      endif
+#    endif
+#  endif
+#endif
+
+#ifdef __acorn
+#  define OS_CODE 13
+#endif
+
+#if defined(WIN32) && !defined(__CYGWIN__)
+#  define OS_CODE  10
+#endif
+
+#ifdef _BEOS_
+#  define OS_CODE  16
+#endif
+
+#ifdef __TOS_OS400__
+#  define OS_CODE 18
+#endif
+
+#ifdef __APPLE__
+#  define OS_CODE 19
+#endif
+
+#if defined(_BEOS_) || defined(RISCOS)
+#  define fdopen(fd,mode) NULL /* No fdopen() */
+#endif
+
+#if (defined(_MSC_VER) && (_MSC_VER > 600)) && !defined __INTERIX
+#  if defined(_WIN32_WCE)
+#    define fdopen(fd,mode) NULL /* No fdopen() */
+#    ifndef _PTRDIFF_T_DEFINED
+       typedef int ptrdiff_t;
+#      define _PTRDIFF_T_DEFINED
+#    endif
+#  else
+#    define fdopen(fd,type)  _fdopen(fd,type)
+#  endif
+#endif
+
+#if defined(__BORLANDC__) && !defined(MSDOS)
+  #pragma warn -8004
+  #pragma warn -8008
+  #pragma warn -8066
+#endif
+
+/* provide prototypes for these when building zlib without LFS */
+#if !defined(_WIN32) && \
+    (!defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0)
+    ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t));
+    ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t));
+#endif
+
+        /* common defaults */
+
+#ifndef OS_CODE
+#  define OS_CODE  3     /* assume Unix */
+#endif
+
+#ifndef F_OPEN
+#  define F_OPEN(name, mode) fopen((name), (mode))
+#endif
+
+         /* functions */
+
+#if defined(pyr) || defined(Z_SOLO)
+#  define NO_MEMCPY
+#endif
+#if defined(SMALL_MEDIUM) && !defined(_MSC_VER) && !defined(__SC__)
+ /* Use our own functions for small and medium model with MSC <= 5.0.
+  * You may have to use the same strategy for Borland C (untested).
+  * The __SC__ check is for Symantec.
+  */
+#  define NO_MEMCPY
+#endif
+#if defined(STDC) && !defined(HAVE_MEMCPY) && !defined(NO_MEMCPY)
+#  define HAVE_MEMCPY
+#endif
+#ifdef HAVE_MEMCPY
+#  ifdef SMALL_MEDIUM /* MSDOS small or medium model */
+#    define zmemcpy _fmemcpy
+#    define zmemcmp _fmemcmp
+#    define zmemzero(dest, len) _fmemset(dest, 0, len)
+#  else
+#    define zmemcpy memcpy
+#    define zmemcmp memcmp
+#    define zmemzero(dest, len) memset(dest, 0, len)
+#  endif
+#else
+   void ZLIB_INTERNAL zmemcpy OF((Bytef* dest, const Bytef* source, uInt len));
+   int ZLIB_INTERNAL zmemcmp OF((const Bytef* s1, const Bytef* s2, uInt len));
+   void ZLIB_INTERNAL zmemzero OF((Bytef* dest, uInt len));
+#endif
+
+/* Diagnostic functions */
+#ifdef ZLIB_DEBUG
+#  include <stdio.h>
+   extern int ZLIB_INTERNAL z_verbose;
+   extern void ZLIB_INTERNAL z_error OF((char *m));
+#  define Assert(cond,msg) {if(!(cond)) z_error(msg);}
+#  define Trace(x) {if (z_verbose>=0) fprintf x ;}
+#  define Tracev(x) {if (z_verbose>0) fprintf x ;}
+#  define Tracevv(x) {if (z_verbose>1) fprintf x ;}
+#  define Tracec(c,x) {if (z_verbose>0 && (c)) fprintf x ;}
+#  define Tracecv(c,x) {if (z_verbose>1 && (c)) fprintf x ;}
+#else
+#  define Assert(cond,msg)
+#  define Trace(x)
+#  define Tracev(x)
+#  define Tracevv(x)
+#  define Tracec(c,x)
+#  define Tracecv(c,x)
+#endif
+
+#ifndef Z_SOLO
+   voidpf ZLIB_INTERNAL zcalloc OF((voidpf opaque, unsigned items,
+                                    unsigned size));
+   void ZLIB_INTERNAL zcfree  OF((voidpf opaque, voidpf ptr));
+#endif
+
+#define ZALLOC(strm, items, size) \
+           (*((strm)->zalloc))((strm)->opaque, (items), (size))
+#define ZFREE(strm, addr)  (*((strm)->zfree))((strm)->opaque, (voidpf)(addr))
+#define TRY_FREE(s, p) {if (p) ZFREE(s, p);}
+
+/* Reverse the bytes in a 32-bit value */
+#define ZSWAP32(q) ((((q) >> 24) & 0xff) + (((q) >> 8) & 0xff00) + \
+                    (((q) & 0xff00) << 8) + (((q) & 0xff) << 24))
+
+#endif /* ZUTIL_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/ic_launcher_background.xml glmark2-2021.02/android/res/drawable/ic_launcher_background.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/ic_launcher_background.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/drawable/ic_launcher_background.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,12 @@
+<?xml version="1.0" encoding="utf-8"?>
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="682.66669dp"
+    android:height="682.66669dp"
+    android:viewportWidth="682.66669"
+    android:viewportHeight="682.66669">
+
+    <path
+        android:fillColor="#008779"
+        android:strokeWidth="1.33333337"
+        android:pathData="M 0,341.33334 V 0 H 341.33334 682.66669 V 341.33334 682.66669 H 341.33334 0 Z" />
+</vector>
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/ic_launcher_foreground.xml glmark2-2021.02/android/res/drawable/ic_launcher_foreground.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/ic_launcher_foreground.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/drawable/ic_launcher_foreground.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+<?xml version="1.0" encoding="utf-8"?>
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="682.66669dp"
+    android:height="682.66669dp"
+    android:viewportWidth="682.66669"
+    android:viewportHeight="682.66669">
+
+    <path
+        android:fillColor="#ffffff"
+        android:strokeWidth="0.8031373"
+        android:pathData="M 186.65658,494.96505 c -3.01898,-0.461 -6.24782,-3.99821 -13.40839,-14.68896 -16.52297,-24.66884 -26.80818,-52.18639 -31.51973,-84.32941 -1.67292,-11.41294 -1.26293,-39.35875 0.77176,-52.60549 3.52378,-22.94131 8.13502,-38.08441 17.36406,-57.02275 25.60931,-52.55127 72.33488,-91.06229 127.6241,-105.18722 18.64988,-4.76455 26.41649,-5.62059 51.1143,-5.63386 24.11773,-0.013 31.09507,0.72932 48.99137,5.2119 37.46202,9.38331 71.75536,29.83891 96.9381,57.82251 32.92513,36.58711 50.20661,79.51017 51.41348,127.69883 0.62101,24.7962 -2.5037,47.76835 -9.09724,66.88067 -5.70161,16.52697 -18.85384,41.46961 -29.31392,55.59262 l -5.05992,6.83181 -12.0826,-0.0704 c -6.64544,-0.0387 -13.88966,-0.30976 -16.09829,-0.60236 l -4.01569,-0.53199 5.69473,-6.04715 c 18.28752,-19.41929 34.32792,-49.22025 40.62249,-75.47129 2.83491,-11.82281 3.9505,-20.83258 4.43378,-35.8083 0.7133,-22.10308 -1.6335,-40.23004 -7.5036,-57.95869 -19.22235,-58.05466 -63.98749,-99.64287 -123.49833,-114.73378 -51.33466,-13.01757 -107.89698,-0.81599 -150.16101,32.39263 -8.69952,6.83558 -22.22728,20.87268 -29.31535,30.41913 -15.29218,20.596 -25.12497,41.54557 -31.99264,68.16295 -1.95922,7.59343 -2.09804,9.44292 -2.4241,32.2968 -0.22528,15.78948 -0.0287,26.29076 0.56391,30.11765 1.57352,10.16209 7.94241,31.71876 12.08191,40.89339 9.33199,20.68308 18.53904,34.5884 36.77167,55.53587 1.15948,1.33213 0.49376,1.39921 -12.75526,1.28516 -7.68823,-0.0662 -14.95104,-0.26883 -16.13959,-0.45032 z" />
+    <path
+        android:fillColor="#ffffff"
+        android:strokeWidth="0.8031373"
+        android:pathData="M 328.56346,392.32894 c -3.69652,-1.98783 -8.02481,-6.54921 -10.55306,-11.12138 -3.16546,-5.72455 -3.22291,-13.51364 -0.14304,-19.39422 3.3164,-6.33219 7.34022,-9.80975 13.44849,-11.62274 15.00354,-4.45318 137.06277,-34.80227 138.30452,-34.38835 0.86219,0.2874 -112.93916,71.56543 -119.77377,75.01878 -5.60462,2.83186 -17.28256,3.65925 -21.28314,1.50791 z m 14.49298,-16.02262 c 3.58953,-3.58954 2.36713,-9.29041 -2.32407,-10.83864 -5.08184,-1.67716 -8.55479,1.05647 -8.55479,6.73365 0,5.78084 6.68174,8.3021 10.87886,4.10499 z" />
+    <path
+        android:fillColor="#ffffff"
+        android:strokeWidth="0.20078433"
+        android:pathData="M 289.21697,477.78148 c -0.18611,-0.20564 -0.42824,-0.68 -0.53806,-1.05412 -0.40999,-1.39663 -0.58185,-8.14102 -0.5865,-23.01747 l -0.005,-15.3098 7.07764,0.008 c 9.62802,0.0108 16.61829,0.27907 17.96675,0.68963 1.14353,0.34816 1.56765,0.85063 1.37291,1.62653 -0.31929,1.27214 -2.07241,1.47686 -12.71377,1.48463 l -8.48314,0.006 v 6.80916 6.80915 l 8.08157,0.0848 c 8.7948,0.0923 9.4873,0.17349 10.25563,1.20259 0.34322,0.4597 0.36923,0.62266 0.16938,1.06129 -0.12986,0.285 -0.5551,0.65148 -0.94499,0.81438 -1.07139,0.44766 -4.96181,0.67726 -11.56142,0.68231 l -5.94649,0.005 -0.13549,3.36313 c -0.48173,11.95696 -0.84016,14.15735 -2.41326,14.81463 -0.92085,0.38476 -1.18786,0.37145 -1.59597,-0.0795 z" />
+    <path
+        android:fillColor="#ffffff"
+        android:strokeWidth="0.20078433"
+        android:pathData="M 324.02801,478.23067 c -1.47636,-0.50264 -1.78166,-0.73649 -1.89956,-1.45497 -0.4016,-2.44729 -0.54592,-7.34646 -0.64694,-21.96149 l -0.11345,-16.41412 4.99429,0.004 c 9.18454,0.007 15.94542,0.51305 18.80591,1.40861 2.13419,0.66817 3.89276,2.37363 5.06545,4.91249 0.90289,1.95475 1.18397,3.29128 1.1823,5.62196 -0.002,3.13292 -0.98927,5.79545 -2.83132,7.6375 -2.15344,2.15344 -6.27343,3.19955 -16.56516,4.20609 -2.83793,0.27756 -5.19701,0.50464 -5.24242,0.50464 -0.0454,0 -0.13706,2.82353 -0.20367,6.27451 -0.1249,6.47072 -0.27537,7.92302 -0.91012,8.78432 -0.36061,0.48931 -1.02752,0.68378 -1.63531,0.47685 z m 15.57072,-21.37251 c 2.66735,-0.37099 4.03399,-0.89856 5.1068,-1.97137 1.31968,-1.31968 1.87676,-2.80708 1.89599,-5.06224 0.0136,-1.59218 -0.0401,-1.83432 -0.68802,-3.10371 -0.86331,-1.69137 -1.73625,-2.34965 -3.8149,-2.8768 -2.47622,-0.62796 -4.85788,-0.82475 -9.98902,-0.82535 -3.75958,-4.3e-4 -4.86992,0.0566 -4.87295,0.25042 -0.002,0.13804 -0.18109,3.18745 -0.39764,6.77647 -0.21655,3.58902 -0.39549,6.63843 -0.39765,6.77647 -0.005,0.3243 10.87075,0.35415 13.15739,0.0361 z" />
+    <path
+        android:fillColor="#ffffff"
+        android:strokeWidth="0.20078433"
+        android:pathData="M 369.36791,477.93894 c -1.90948,-0.28011 -3.60226,-0.78091 -5.1661,-1.52838 -3.29583,-1.57531 -6.45699,-6.02117 -6.68021,-9.39509 -0.0874,-1.32031 0.50212,-1.77533 2.3267,-1.796 1.48921,-0.0169 2.30799,0.42686 2.60418,1.41133 0.61323,2.03825 0.84263,2.43901 2.11547,3.69571 0.97514,0.96278 1.82575,1.55758 3.23104,2.25938 2.31926,1.15822 3.80992,1.55381 5.85505,1.55381 2.11561,0 3.37881,-0.29287 5.37518,-1.2462 3.84074,-1.83409 5.58636,-5.10482 4.16706,-7.80771 -0.84206,-1.6036 -3.77867,-3.08396 -10.57847,-5.33266 -7.7415,-2.56012 -11.73177,-5.07253 -13.36805,-8.41698 -1.41913,-2.90058 -0.47099,-6.30295 2.61696,-9.39089 4.24666,-4.24666 12.49334,-4.93527 19.35156,-1.61587 3.18824,1.54312 5.56301,4.29284 5.80826,6.72535 0.0529,0.52423 -0.012,1.12103 -0.14981,1.37853 -0.21668,0.40487 -0.35965,0.44442 -1.3188,0.36488 -1.31324,-0.1089 -2.7375,-0.99017 -4.80835,-2.97519 -1.73056,-1.65883 -2.68551,-2.26366 -4.56412,-2.89077 -4.03065,-1.34549 -8.68037,-0.52921 -10.87223,1.90866 -0.97903,1.08892 -1.32914,2.06138 -1.32914,3.6918 0,1.48903 0.43885,2.38507 1.53876,3.14183 1.35842,0.93462 2.68863,1.46165 9.36905,3.71202 6.89776,2.32359 9.29199,3.41986 10.90858,4.99483 0.78995,0.76962 1.11224,1.2702 1.36034,2.11288 0.86208,2.92799 0.90509,6.53999 0.10851,9.11121 -0.75222,2.428 -2.89775,4.16958 -6.60331,5.36005 -3.33199,1.07046 -7.93392,1.46698 -11.29811,0.97347 z" />
+</vector>
\ No newline at end of file
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_about.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_about.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_delete.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_delete.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_load.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_load.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_results.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_results.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_save.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_save.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable/menu_settings.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable/menu_settings.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_archive.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_archive.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_delete.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_delete.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_info_details.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_info_details.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_mark.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_mark.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_preferences.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_preferences.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-hdpi/ic_menu_save.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-hdpi/ic_menu_save.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_archive.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_archive.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_delete.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_delete.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_info_details.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_info_details.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_mark.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_mark.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_preferences.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_preferences.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-mdpi/ic_menu_save.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-mdpi/ic_menu_save.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_archive.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_archive.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_delete.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_delete.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_info_details.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_info_details.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_matk.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_matk.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_preferences.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_preferences.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/drawable-xhdpi/ic_menu_save.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/drawable-xhdpi/ic_menu_save.png differ
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/menu/main_options_menu.xml glmark2-2021.02/android/res/menu/main_options_menu.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/menu/main_options_menu.xml	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/android/res/menu/main_options_menu.xml	2021-04-25 01:47:22.000000000 +0800
@@ -2,25 +2,25 @@
 <menu xmlns:android="http://schemas.android.com/apk/res/android">
     <item android:id="@+id/save_benchmark_list"
           android:title="@string/saveMainOptionsText"
-          android:icon="@drawable/menu_save" />
+          android:icon="@drawable/ic_menu_save" />
 
     <item android:id="@+id/load_benchmark_list"
           android:title="@string/loadMainOptionsText"
-          android:icon="@drawable/menu_load" />
+          android:icon="@drawable/ic_menu_archive" />
 
     <item android:id="@+id/delete_benchmark_list"
           android:title="@string/deleteMainOptionsText"
-          android:icon="@drawable/menu_delete" />
+          android:icon="@drawable/ic_menu_delete" />
 
     <item android:id="@+id/settings"
           android:title="@string/settingsMainOptionsText"
-          android:icon="@drawable/menu_settings" />
+          android:icon="@drawable/ic_menu_preferences" />
 
     <item android:id="@+id/results"
           android:title="@string/resultsMainOptionsText"
-          android:icon="@drawable/menu_results" />
+          android:icon="@drawable/ic_menu_mark" />
 
     <item android:id="@+id/about"
           android:title="@string/aboutMainOptionsText"
-          android:icon="@drawable/menu_about" />
+          android:icon="@drawable/ic_menu_info_details" />
 </menu>
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap/ic_launcher_round.xml glmark2-2021.02/android/res/mipmap/ic_launcher_round.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap/ic_launcher_round.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/mipmap/ic_launcher_round.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,36 @@
+<?xml version="1.0" encoding="utf-8"?>
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="256dp"
+    android:height="256dp"
+    android:viewportWidth="256"
+    android:viewportHeight="256">
+
+    <path
+        android:fillColor="#6e6e6e"
+        android:strokeWidth="1.33333337"
+        android:pathData="M 118,245.20092 C 68.694377,241.76996 26.472724,206.11052 13.173269,156.66667 9.1103097,141.56165 9.1036777,114.48495 13.159226,99.333333 24.401048,57.333641 57.887772,23.895083 99.333333,13.28317 114.76476,9.3320377 141.99289,9.2665532 156.66667,13.145282 c 23.43709,6.195143 48.67534,22.350386 63.32134,40.532605 8.56636,10.634672 19.24586,31.957237 22.86671,45.655446 3.68585,13.944117 4.03917,40.941587 0.71244,54.438547 -10.95927,44.46313 -45.38429,79.06689 -88.23383,88.69192 -14.66293,3.29364 -21.80776,3.81747 -37.33333,2.73712 z M 78.746265,182.33333 C 68.500019,171.38531 61.796753,151.35807 63.190772,135.85843 64.659353,119.52975 70.587427,106.76645 82.008531,95.343299 93.585981,83.763773 111.72249,76.039416 127.33333,76.039416 c 15.75038,0 33.72514,7.636335 45.35894,19.27014 14.72305,14.723044 21.51038,35.000424 18.19468,54.357114 -1.97303,11.51824 -8.35166,26.01621 -13.70119,31.14138 -2.11883,2.02998 -3.85243,4.1786 -3.85243,4.77474 0,0.59613 2.95376,1.08388 6.5639,1.08388 6.54356,0 6.57882,-0.0237 11.38002,-7.66667 9.95058,-15.84003 13.67664,-39.85777 8.84288,-57 C 192.236,94.039995 170.19648,73.329464 141.96175,67.348585 130.28424,64.874973 124.41266,64.873092 112.66667,67.339195 93.232081,71.419539 73.553804,84.882445 64.163757,100.52249 55.888687,114.30544 53.099305,124.25453 53.095631,140 c -0.0037,16.05973 3.257437,28.08329 10.809032,39.85143 l 4.373325,6.81524 h 7.26192 7.261921 z m 35.263605,-7.66666 c 1.74135,0 3.5369,-0.6 3.99013,-1.33334 C 118.48077,172.55544 117.12496,172 114.74536,172 c -2.73662,0 -4.07869,-0.65809 -4.07869,-2 0,-1.33333 1.33333,-2 4,-2 2.2,0 4,-0.6 4,-1.33333 0,-0.73334 -2.4,-1.33334 -5.33334,-1.33334 H 108 v 7.11111 c 0,7.52677 1.70851,11.53613 2.4219,5.68348 0.32129,-2.63591 1.17685,-3.46125 3.58797,-3.46125 z M 124,177.34321 c 0,-1.7085 1.11772,-2.63292 3.53218,-2.92132 8.29606,-0.9909 6.78546,-9.08856 -1.69544,-9.08856 h -4.50341 v 8.0787 c 0,5.04626 0.50048,7.76937 1.33334,7.25464 C 123.4,180.21344 124,178.71788 124,177.34321 Z M 124,170 c 0,-1.11111 1.18519,-2 2.66667,-2 1.48148,0 2.66666,0.88889 2.66666,2 0,1.11111 -1.18518,2 -2.66666,2 C 125.18519,172 124,171.11111 124,170 Z m 20.55681,9.57853 c 3.56476,-2.60661 2.12154,-6.61752 -2.8993,-8.05746 -4.43427,-1.27174 -5.40892,-2.69328 -2.90683,-4.23966 0.77955,-0.48178 2.51999,0.12187 3.86764,1.3415 1.38815,1.25626 2.79027,1.66734 3.23463,0.94836 1.2852,-2.0795 -2.74823,-4.46236 -6.77232,-4.00095 -2.97672,0.34132 -3.7473,1.10487 -3.7473,3.71311 0,2.56057 0.95402,3.55704 4.33334,4.52621 4.65648,1.33547 5.58721,3.12623 2.28366,4.39392 -1.1273,0.43259 -3.16774,-0.22533 -4.53429,-1.46204 -1.50031,-1.35776 -2.81003,-1.72208 -3.30597,-0.9196 -0.85699,1.38663 4.08133,5.51141 6.59844,5.51141 0.79666,0 2.5284,-0.78965 3.8483,-1.7548 z m 9.93802,-45.20825 c 13.21293,-8.22969 23.83118,-15.15538 23.59613,-15.39043 -0.23505,-0.23505 -12.65168,2.57727 -27.59249,6.2496 -23.21545,5.70616 -27.60836,7.1874 -30.21353,10.1877 -3.26276,3.75761 -3.03615,7.27565 0.74608,11.58285 4.00686,4.563 8.71098,2.7876 33.46381,-12.62972 z m -28.30886,7.37125 c -1.50158,-1.50158 -0.91452,-3.07486 1.14736,-3.07486 1.1,0 2,0.56458 2,1.25464 0,1.6813 -2.08212,2.88546 -3.14736,1.82022 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.07529411"
+        android:pathData="M 56.989172,222.07454 c -3.573786,-2.74992 -6.691217,-5.38956 -9.598201,-8.12715 -1.482508,-1.39611 -4.885205,-4.82953 -6.110384,-6.16555 -7.05864,-7.69721 -12.906672,-15.95702 -17.862055,-25.22851 l -0.688184,-1.28759 V 149.66863 118.07151 H 39.32132 55.912292 l -0.366056,1.29882 c -1.853396,6.57614 -2.598223,13.44338 -2.43973,22.49412 0.03155,1.80141 0.09248,3.81741 0.135412,4.48 0.467901,7.22152 1.572929,13.12235 3.546165,18.93647 1.050127,3.09419 1.91166,5.1372 3.45442,8.19171 1.612354,3.19229 2.058122,3.9427 5.889476,9.91434 l 2.102808,3.27748 h 7.299824 7.299823 l -0.630866,-0.68293 c -0.346977,-0.37561 -1.410161,-1.51664 -2.362631,-2.53563 -2.421213,-2.59031 -3.384719,-3.74563 -4.870595,-5.84026 -6.183198,-8.71638 -10.700865,-21.26644 -11.686799,-32.46589 -0.239339,-2.7187 -0.286851,-6.61635 -0.106929,-8.77176 0.518613,-6.21284 1.807907,-12.38272 3.685674,-17.63769 l 0.235435,-0.65887 60.241097,0.0189 60.24111,0.0189 0.54771,1.54353 c 1.96463,5.53662 3.09549,11.06616 3.44928,16.86588 0.10747,1.7618 0.0586,6.44466 -0.0844,8.09412 -0.46584,5.37142 -1.58111,10.3996 -3.62401,16.33882 -2.08867,6.0723 -4.84258,11.8705 -7.55979,15.91669 -1.15635,1.72191 -1.74128,2.43275 -3.52665,4.28577 -1.71976,1.78492 -2.47904,2.6617 -3.044,3.51504 -0.54982,0.83047 -0.5553,1.03223 -0.035,1.29169 0.41648,0.20769 1.66619,0.46168 2.85319,0.57988 0.51537,0.0513 2.16126,0.0834 4.06588,0.0793 3.45942,-0.007 3.74609,-0.0344 4.78075,-0.44942 1.59464,-0.63962 2.88264,-2.2283 6.21219,-7.66246 5.91668,-9.65659 9.70865,-22.45149 10.50286,-35.43886 0.11839,-1.936 0.0662,-8.6436 -0.08,-10.27765 -0.45498,-5.08583 -1.27181,-9.12152 -2.77742,-13.72235 l -0.32648,-0.99765 h 19.2678 19.26781 l -0.019,27.08706 -0.019,27.08706 -0.69695,1.58117 c -2.23675,5.07458 -4.79797,10.00574 -7.58541,14.60427 -3.87609,6.39452 -8.4447,12.55968 -13.44257,18.14023 -3.36023,3.75199 -7.78557,8.07239 -11.49993,11.22727 -2.01016,1.70737 -5.12175,4.18737 -6.08334,4.84854 l -0.23054,0.15852 -69.98122,-0.003 -69.981222,-0.003 -0.941176,-0.72421 z m 84.705878,-40.92993 c 1.22892,-0.40123 2.85413,-1.40406 3.67367,-2.26683 1.57762,-1.66082 1.57625,-3.62374 -0.004,-5.28699 -0.96106,-1.01175 -1.8067,-1.47328 -4.05016,-2.2105 -2.38347,-0.78324 -3.41021,-1.34647 -3.75577,-2.06031 -0.14479,-0.29908 -0.15418,-0.37824 -0.0788,-0.66458 0.1016,-0.38608 0.77918,-1.08171 1.2985,-1.33311 0.81063,-0.39241 2.123,-0.003 3.40994,1.01159 1.19091,0.93896 1.4479,1.11954 1.89235,1.32963 0.89763,0.4243 1.59984,0.35084 1.8586,-0.19444 0.40168,-0.84649 0.0861,-1.70839 -0.91273,-2.49294 -1.50467,-1.18186 -4.10032,-1.72506 -6.49427,-1.35909 -1.86445,0.28503 -2.67063,0.80453 -3.03219,1.95395 -0.14287,0.45415 -0.16755,0.70137 -0.16604,1.66287 0.002,1.31536 0.1163,1.82218 0.56839,2.52167 0.55203,0.85409 1.40687,1.3272 3.75925,2.08053 3.06964,0.98303 4.36779,1.87112 4.18208,2.86103 -0.12763,0.6803 -1.70794,1.60971 -2.73707,1.60971 -0.95856,0 -2.53232,-0.64744 -3.45618,-1.42186 -1.36879,-1.14738 -1.96364,-1.46984 -2.71954,-1.47421 -0.44464,-0.003 -0.49134,0.0137 -0.69647,0.24314 -0.53407,0.59726 -0.0961,1.58444 1.29639,2.92178 1.73187,1.66332 4.14513,2.91635 5.3355,2.77032 0.18635,-0.0229 0.55906,-0.11347 0.82823,-0.20136 z m -31.87452,-0.4905 c 0.21482,-0.31955 0.36923,-0.85386 0.51758,-1.79097 0.36221,-2.2879 0.58842,-2.93729 1.21545,-3.48923 0.50382,-0.44348 1.07659,-0.59845 2.51812,-0.6813 0.64714,-0.0372 1.39563,-0.11843 1.66331,-0.18051 1.11869,-0.25944 2.05661,-0.74261 2.27997,-1.17452 0.15524,-0.30021 0.14312,-0.40464 -0.0773,-0.66663 -0.36582,-0.43474 -1.41083,-0.65169 -3.46612,-0.71956 -2.46107,-0.0813 -3.54686,-0.57987 -3.7541,-1.72392 -0.0753,-0.41551 0.12018,-1.01781 0.424,-1.30659 0.59529,-0.56583 1.31629,-0.75983 3.18422,-0.8568 0.78683,-0.0408 1.68471,-0.11079 1.9953,-0.15544 1.32812,-0.19094 2.35294,-0.72837 2.35294,-1.23393 0,-0.31083 -0.36842,-0.59515 -1.06838,-0.82453 -1.36821,-0.44835 -2.79484,-0.56926 -6.71667,-0.56926 h -2.92554 v 4.18173 c 0,2.29996 0.0349,4.7649 0.0776,5.47765 0.17423,2.90813 0.59136,5.08707 1.10296,5.7614 0.23566,0.3106 0.44578,0.29583 0.67664,-0.0476 z m 13.20636,-0.29062 c 0.17044,-0.18491 0.39156,-0.51458 0.49137,-0.7326 0.23972,-0.52362 0.48253,-1.48818 0.48325,-1.91966 9.8e-4,-0.59784 0.15122,-1.30627 0.35734,-1.68499 0.47473,-0.87222 1.3941,-1.31194 3.34263,-1.59872 2.61838,-0.38537 4.30138,-1.37331 5.06377,-2.97248 0.21929,-0.45995 0.23333,-0.54188 0.23333,-1.36097 v -0.87156 l -0.31526,-0.63432 c -0.80325,-1.61619 -2.36856,-2.62643 -4.84238,-3.12522 -0.51754,-0.10436 -1.20695,-0.13563 -3.60278,-0.16345 l -2.96279,-0.0344 0.0368,5.67378 c 0.0361,5.57541 0.0741,6.52863 0.3172,7.97026 0.15721,0.9322 0.32421,1.42774 0.56094,1.66447 0.26816,0.26816 0.43415,0.22646 0.8366,-0.21014 z m 4.40838,-31.38723 c 4.29344,-0.96294 10.61486,-4.35165 26.30684,-14.10224 12.3786,-7.69174 23.99282,-15.23337 24.35025,-15.81171 0.0261,-0.0423 0.0201,-0.0938 -0.0134,-0.11448 -0.37345,-0.23082 -15.89756,3.3351 -32.09217,7.37163 -19.08121,4.75601 -23.22765,6.2198 -25.72738,9.08239 -0.96747,1.10789 -1.81702,2.60165 -2.09106,3.67672 -0.32003,1.25544 -0.20584,2.72012 0.31404,4.02823 0.97612,2.45607 3.27636,5.11902 4.98048,5.76581 1.09716,0.41642 2.42351,0.45103 3.97237,0.10365 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.07529411"
+        android:pathData="M 125.96199,171.89985 c -0.88189,-0.17287 -1.72931,-0.87752 -1.87559,-1.55959 -0.1799,-0.83886 0.28268,-1.58285 1.2664,-2.0368 0.54544,-0.2517 1.6497,-0.31597 2.28009,-0.13271 1.61407,0.46922 2.1646,2.04503 1.06769,3.05612 -0.66565,0.61357 -1.71995,0.87266 -2.73859,0.67298 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.07529411"
+        android:pathData="M 126.595,141.98449 c -0.29283,-0.12755 -0.80582,-0.70887 -1.02536,-1.16193 -0.21245,-0.43843 -0.22559,-1.00974 -0.0308,-1.33849 0.56897,-0.96017 2.70874,-1.03939 3.50733,-0.12984 0.18897,0.21522 0.22522,0.32001 0.2246,0.64931 -0.002,1.30689 -1.56594,2.46438 -2.67579,1.98095 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.07529411"
+        android:pathData="M 67.153878,117.99281 c 0,-0.0433 0.242045,-0.67858 0.537877,-1.41177 3.192684,-7.91267 8.006234,-14.96067 14.717833,-21.549844 3.503974,-3.440061 7.361676,-6.406196 11.774878,-9.053545 9.026964,-5.415007 18.948894,-8.779561 28.574114,-9.689566 9.04876,-0.855503 19.22053,0.957922 29.10118,5.188167 1.34825,0.577233 4.5714,2.148383 5.98588,2.917873 5.63075,3.063161 10.70331,6.826176 14.94588,11.087418 4.84037,4.861667 8.69847,10.128197 11.91543,16.265257 0.80499,1.5357 2.13571,4.46625 2.65829,5.85412 l 0.17719,0.47059 h -60.19428 c -41.817389,0 -60.194272,-0.024 -60.194272,-0.0787 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.30117646"
+        android:pathData="M 20.901298,177.22279 C 18.808311,172.84125 17.099947,168.5701 15.367781,163.38824 11.676754,152.34633 10.376505,143.13626 10.371783,128 10.366276,110.34928 12.101843,100.46678 17.72386,86.136471 28.591515,58.435232 50.491461,34.67778 77.131294,21.690231 89.596916,15.612944 100.00926,12.568077 113.42306,11.077485 c 5.56711,-0.618639 26.13988,-0.515213 31.62353,0.158981 8.8459,1.087568 15.8133,2.897783 24.2447,6.29906 16.34325,6.592964 31.70891,16.681111 43.6794,28.677212 7.95324,7.970246 13.04993,15.285769 19.63427,28.181958 6.46585,12.664125 10.36624,23.652614 11.651,32.824124 0.94186,6.72363 1.11988,9.27908 1.2903,18.52236 0.23895,12.96009 -0.38157,20.60141 -2.33297,28.72902 -0.90985,3.78951 -3.94899,13.22372 -5.05584,15.69451 l -0.60714,1.35529 -0.0213,-26.72941 -0.0213,-26.72941 h -19.21614 -19.21614 l -1.31148,-3.3876 c -2.78349,-7.18987 -7.28197,-14.977744 -12.06586,-20.888721 -2.67181,-3.301295 -8.02693,-8.602997 -11.43743,-11.323344 -9.94072,-7.929097 -21.19406,-13.048169 -34.31564,-15.609946 -10.17101,-1.985732 -15.20069,-1.975248 -25.61848,0.0534 C 97.884502,70.106703 81.56167,79.36191 70.582559,91.708239 65.110237,97.862025 59.566596,107.34003 56.791638,115.28669 l -0.968846,2.77449 H 39.297286 22.771729 l -0.07669,31.37808 -0.07669,31.37808 z" />
+    <path
+        android:fillColor="#008476"
+        android:strokeWidth="0.30117646"
+        android:pathData="M 120.80188,245.27885 c -0.66259,-0.0621 -2.89882,-0.25398 -4.96941,-0.42637 -19.060639,-1.58694 -38.951714,-8.87416 -54.964705,-20.13668 l -2.409412,-1.69462 34.885316,-0.0766 c 19.186921,-0.0421 50.479461,-0.0421 69.538971,0 l 34.65366,0.0766 -3.12956,2.1395 c -13.27889,9.07801 -27.25199,14.87167 -43.78839,18.15595 -8.83827,1.75537 -10.27985,1.90084 -19.72706,1.99067 -4.88658,0.0465 -9.42682,0.0337 -10.08941,-0.0284 z" />
+</vector>
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap/ic_launcher.xml glmark2-2021.02/android/res/mipmap/ic_launcher.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap/ic_launcher.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/mipmap/ic_launcher.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,16 @@
+<?xml version="1.0" encoding="utf-8"?>
+<vector xmlns:android="http://schemas.android.com/apk/res/android"
+    android:width="682.66669dp"
+    android:height="682.66669dp"
+    android:viewportWidth="682.66669"
+    android:viewportHeight="682.66669">
+
+    <path
+        android:fillColor="#239387"
+        android:strokeWidth="1.33333337"
+        android:pathData="M 103.9058,610.17686 C 90.811563,606.64629 80.958343,598.5039 75.455785,586.66669 l -2.789116,-6 V 342.00001 103.33334 l 3.143408,-6.951689 c 3.915115,-8.658325 13.518334,-17.799662 22.227481,-21.158412 6.444492,-2.48537 13.220182,-2.55657 243.295782,-2.55657 229.57308,0 236.86416,0.07616 243.25515,2.541047 8.35962,3.224149 17.30633,11.43754 21.8642,20.072093 L 610.00002,102 v 240.00001 240.00001 l -2.93627,5.48863 c -6.59844,12.33413 -15.08873,19.16261 -28.01112,22.52846 -6.60253,1.71975 -38.23278,1.97096 -238.33275,1.89292 -178.39529,-0.0696 -232.10197,-0.46265 -236.81408,-1.73317 z M 216.55762,499.1907 c 3.2456,-1.24546 2.67727,-2.45482 -5.39203,-11.47362 -21.44697,-23.97061 -35.42592,-53.67003 -41.28678,-87.71707 -2.61314,-15.18034 -1.90914,-44.52906 1.4312,-59.66432 7.47466,-33.86809 22.90065,-62.43489 46.56768,-86.23693 51.87684,-52.17277 131.37223,-65.38529 197.61308,-32.84425 54.86629,26.95325 89.74897,78.26993 95.38486,140.32276 2.50268,27.55509 -3.23687,58.74214 -15.63501,84.95631 -6.93782,14.66904 -17.60435,31.05137 -27.13791,41.68012 -3.72315,4.15085 -6.76936,8.13665 -6.76936,8.85734 0,2.30446 3.88936,2.92898 18.24077,2.92898 13.89296,0 14.05759,-0.0373 17.79732,-4.04078 5.55319,-5.94472 14.86714,-20.50869 21.32723,-33.34888 44.39062,-88.23136 15.85067,-196.61532 -66.48673,-252.49161 -20.89405,-14.17922 -46.8999,-25.26489 -72.21193,-30.78217 -15.8521,-3.45531 -53.38092,-4.18496 -71.33333,-1.38691 -27.77784,4.32944 -56.34279,15.50087 -80.66667,31.54782 -15.3233,10.10908 -42.92603,37.16309 -52.8872,51.83585 -35.68902,52.56975 -44.81449,120.06469 -24.15082,178.62724 5.18735,14.70138 12.83567,29.94949 21.89082,43.64278 10.51269,15.89739 11.36176,16.39666 27.88462,16.39666 7.54114,0 14.66022,-0.36419 15.82019,-0.80932 z m 112.77572,-25.178 v -7.03536 l 8.6147,-0.92255 c 5.34889,-0.57283 9.85925,-1.90152 11.89763,-3.50492 6.64236,-5.22487 5.54449,-16.72827 -1.96724,-20.61273 -4.63803,-2.39842 -22.6405,-2.64171 -24.94509,-0.33712 -2.56906,2.56906 -2.18905,36.28713 0.43859,38.91477 3.67023,3.67023 5.96141,1.17124 5.96141,-6.50209 z m 0.8889,-16.23491 c -0.4889,-0.48888 -0.8889,-2.87745 -0.8889,-5.30792 0,-3.60476 0.61584,-4.57361 3.3421,-5.25787 6.00605,-1.50741 12.54243,0.53034 13.46272,4.19707 1.24611,4.96487 -1.51679,7.25761 -8.74586,7.25761 -3.45465,0 -6.68118,-0.4 -7.17006,-0.88889 z m 53.80447,22.87509 c 4.69158,-2.42612 6.63997,-5.7397 6.63997,-11.29248 0,-5.44229 -4.73228,-10.0968 -12.18007,-11.97992 -8.50152,-2.14953 -11.81993,-3.76436 -11.81993,-5.75188 0,-5.73022 10.28861,-6.89153 14.42845,-1.62858 1.44211,1.83333 3.88566,3.33333 5.43011,3.33333 4.39107,0 3.64073,-5.72901 -1.2942,-9.88148 -3.32519,-2.79796 -5.49151,-3.45185 -11.43589,-3.45185 -6.32778,0 -7.86814,0.53454 -11.23103,3.89744 -7.858,7.85798 -3.61621,15.81838 10.70101,20.08224 5.0206,1.4952 9.44862,3.55312 9.84004,4.57317 1.38052,3.59757 -2.536,6.71988 -8.42921,6.71988 -4.75545,0 -6.14765,-0.68747 -9.38916,-4.63637 -3.47133,-4.22887 -3.99292,-4.45158 -5.93444,-2.53395 -1.67037,1.64981 -1.83693,2.86957 -0.77371,5.66605 2.9433,7.74146 16.63039,11.44419 25.44806,6.8844 z m -88.0267,-7.9862 0.66667,-8 8.66666,-0.66666 c 7.55556,-0.5812 8.66667,-1.00855 8.66667,-3.33334 0,-2.32478 -1.11111,-2.75213 -8.66667,-3.33333 l -8.66666,-0.66667 v -4.66666 -4.66667 l 10,-0.66667 c 8.88889,-0.59258 10,-0.96296 10,-3.33333 0,-2.44908 -1.01727,-2.69809 -12.46763,-3.05181 -8.04664,-0.24858 -13.12919,0.16392 -14.33333,1.16326 -1.43272,1.18906 -1.86571,5.79106 -1.86571,19.82959 0,10.05465 0.42531,18.70649 0.94513,19.22632 0.51983,0.51981 2.16983,0.76981 3.66667,0.55555 2.30427,-0.32984 2.82375,-1.61612 3.3882,-8.38958 z m 74,-88.23135 c 21.94998,-13.7545 88.21777,-54.93456 96.85011,-60.18451 5.41676,-3.2943 8.58583,-6.59265 7.43755,-7.74094 -0.76019,-0.76019 -138.52142,33.08149 -145.13102,35.65211 -7.32131,2.8474 -14.48997,13.84174 -14.48997,22.22278 0,2.09206 1.65,7.03891 3.66666,10.993 4.8808,9.56984 12.29083,13.74596 23,12.96228 6.57279,-0.48098 9.54585,-1.92306 28.66667,-13.90472 z m -33.93909,-6.36193 c -2.54758,-3.06963 -1.18603,-8.0084 2.34581,-8.50907 6.15599,-0.87268 9.57851,6.25594 4.3476,9.05544 -3.45739,1.85034 -4.79493,1.74116 -6.69341,-0.54637 z" />
+    <path
+        android:fillColor="#08796d"
+        android:strokeWidth="1.33333337"
+        android:pathData="M 101.76557,608.75489 c -10.069663,-3.1622 -17.070628,-8.56875 -22.559266,-17.42154 -2.417729,-3.89962 -2.339737,-3.87573 3.108165,0.95254 3.073469,2.72388 8.023074,6.48964 10.999124,8.36833 12.282247,7.75344 0.125746,7.41444 252.124437,7.03099 l 231.22866,-0.35186 6.23422,-2.88118 c 8.81086,-4.07199 17.59626,-12.6653 21.28142,-20.81608 l 3.15102,-6.9694 V 340.00001 c 0,-220.71268 -0.15458,-237.0523 -2.2932,-242.387225 -1.26126,-3.146304 -1.75713,-5.389247 -1.10194,-4.984318 0.65518,0.40493 2.28713,3.879178 3.62653,7.720553 2.32247,6.66078 2.43528,17.85476 2.43528,241.65099 0,223.51537 -0.11533,234.99742 -2.42731,241.627 -3.22109,9.23648 -12.30238,19.17101 -21.20012,23.19194 l -7.03924,3.18107 -236,0.25167 c -191.51486,0.20422 -237.04952,-0.0779 -241.56778,-1.4968 z" />
+</vector>
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-anydpi-v26/ic_launcher_round.xml glmark2-2021.02/android/res/mipmap-anydpi-v26/ic_launcher_round.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-anydpi-v26/ic_launcher_round.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/mipmap-anydpi-v26/ic_launcher_round.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+    <background android:drawable="@drawable/ic_launcher_background" />
+    <foreground android:drawable="@drawable/ic_launcher_foreground" />
+</adaptive-icon>
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-anydpi-v26/ic_launcher.xml glmark2-2021.02/android/res/mipmap-anydpi-v26/ic_launcher.xml
--- glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-anydpi-v26/ic_launcher.xml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/android/res/mipmap-anydpi-v26/ic_launcher.xml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,5 @@
+<?xml version="1.0" encoding="utf-8"?>
+<adaptive-icon xmlns:android="http://schemas.android.com/apk/res/android">
+    <background android:drawable="@drawable/ic_launcher_background" />
+    <foreground android:drawable="@drawable/ic_launcher_foreground" />
+</adaptive-icon>
\ No newline at end of file
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-hdpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-hdpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-hdpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-hdpi/ic_launcher_round.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-ldpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-ldpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-ldpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-ldpi/ic_launcher_round.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-mdpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-mdpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-mdpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-mdpi/ic_launcher_round.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xhdpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xhdpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xhdpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xhdpi/ic_launcher_round.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xxhdpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xxhdpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xxhdpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xxhdpi/ic_launcher_round.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xxxhdpi/ic_launcher.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xxxhdpi/ic_launcher.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/android/res/mipmap-xxxhdpi/ic_launcher_round.png and /tmp/VEvcU31eDG/glmark2-2021.02/android/res/mipmap-xxxhdpi/ic_launcher_round.png differ
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/meson.build glmark2-2021.02/data/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/data/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/data/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,14 @@
+install_subdir(
+    'shaders',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
+
+install_subdir(
+    'models',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
+
+install_subdir(
+    'textures',
+    install_dir : join_paths([get_option('datadir'), 'glmark2'])
+    )
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/bump-height.frag glmark2-2021.02/data/shaders/bump-height.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/bump-height.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/bump-height.frag	2021-04-25 01:47:22.000000000 +0800
@@ -4,7 +4,12 @@
 
 uniform sampler2D HeightMap;
 
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+varying highp vec2 TextureCoord;
+#else
 varying vec2 TextureCoord;
+#endif
+
 varying vec3 NormalEye;
 varying vec3 TangentEye;
 varying vec3 BitangentEye;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/conditionals.frag glmark2-2021.02/data/shaders/conditionals.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/conditionals.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/conditionals.frag	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/function.frag glmark2-2021.02/data/shaders/function.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/function.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/function.frag	2021-04-25 01:47:22.000000000 +0800
@@ -8,7 +8,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/ideas-lamp-lit.frag glmark2-2021.02/data/shaders/ideas-lamp-lit.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/ideas-lamp-lit.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/ideas-lamp-lit.frag	2021-04-25 01:47:22.000000000 +0800
@@ -6,7 +6,6 @@
     vec4 position;
 };
 LightSourceParameters lightSource[3];
-uniform mat4 modelview;
 uniform vec4 light0Position;
 uniform vec4 light1Position;
 uniform vec4 light2Position;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/light-refract.frag glmark2-2021.02/data/shaders/light-refract.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/light-refract.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/light-refract.frag	2021-04-25 01:47:22.000000000 +0800
@@ -23,7 +23,7 @@
     vec3 mc_perspective = (MapCoord.xyz / MapCoord.w) + front_refraction;
     vec2 dcoord = mc_perspective.st * point_five + point_five;
     vec4 distance_value = texture2D(DistanceMap, dcoord);
-    vec3 back_position = vertex_position.xyz + front_refraction * distance_value.z;
+    vec3 back_position = vertex_position.xyz + front_refraction * distance_value.x;
     // Use the exit point to index the map of back-side normals, and use the
     // back-side position and normal to find the transmitted vector out of the
     // object.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/loop.frag glmark2-2021.02/data/shaders/loop.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/loop.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/loop.frag	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,14 @@
 
 void main(void)
 {
-    float d = fract(gl_FragCoord.x * gl_FragCoord.y * 0.0001);
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+    // should be declared highp since the multiplication can overflow in
+    // mediump, particularly if mediump is implemented as fp16
+    highp vec2 FragCoord = gl_FragCoord.xy;
+#else
+    vec2 FragCoord = gl_FragCoord.xy;
+#endif
+    float d = fract(FragCoord.x * FragCoord.y * 0.0001);
 
 $MAIN$
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/shadow.frag glmark2-2021.02/data/shaders/shadow.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/shadow.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/shadow.frag	2021-04-25 01:47:22.000000000 +0800
@@ -8,7 +8,7 @@
     vec4 sc_perspective = ShadowCoord / ShadowCoord.w;
     sc_perspective.z += 0.1505;
     vec4 shadow_value = texture2D(ShadowMap, sc_perspective.st);
-    float light_distance = shadow_value.z;
+    float light_distance = shadow_value.x;
     float shadow = 1.0;
     if (ShadowCoord.w > 0.0 && light_distance < sc_perspective.z) {
         shadow = 0.5;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain.frag glmark2-2021.02/data/shaders/terrain.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/terrain.frag	2021-04-25 01:47:22.000000000 +0800
@@ -3,9 +3,6 @@
 uniform vec3 uSpecularColor;
 uniform float uShininess;
 uniform float uOpacity;
-uniform bool enableDiffuse1;
-uniform bool enableDiffuse2;
-uniform bool enableSpecular;
 uniform sampler2D tDiffuse1;
 uniform sampler2D tDiffuse2;
 uniform sampler2D tDetail;
@@ -70,7 +67,12 @@
     vec3 pointSpecular = vec3( 0.0 );
     for ( int i = 0; i < MAX_POINT_LIGHTS; i ++ ) {
         vec4 lPosition = viewMatrix * vec4( pointLightPosition[ i ], 1.0 );
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+        // should be highp for correct behaviour if mediump is implemented as fp16
+        highp vec3 lVector = lPosition.xyz + vViewPosition.xyz;
+#else
         vec3 lVector = lPosition.xyz + vViewPosition.xyz;
+#endif
         float lDistance = 1.0;
         if ( pointLightDistance[ i ] > 0.0 )
             lDistance = 1.0 - min( ( length( lVector ) / pointLightDistance[ i ] ), 1.0 );
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain-noise.frag glmark2-2021.02/data/shaders/terrain-noise.frag
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain-noise.frag	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/terrain-noise.frag	2021-04-25 01:47:22.000000000 +0800
@@ -17,7 +17,13 @@
 uniform MEDIUMP vec2 uvScale;
 varying vec2 vUv;
 
+#ifdef GL_FRAGMENT_PRECISION_HIGH
+// x should be passed as highp since the intermediate multiplications can
+// overflow with mediump
+vec4 permute(highp vec4 x)
+#else
 vec4 permute(vec4 x)
+#endif
 {
     return mod(((x * 34.0) + 1.0) * x, 289.0);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain.vert glmark2-2021.02/data/shaders/terrain.vert
--- glmark2-2014.03+git20150611.fa71af2d/data/shaders/terrain.vert	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/data/shaders/terrain.vert	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,6 @@
 uniform mat4 modelViewMatrix;
 uniform mat4 normalMatrix;
 uniform mat4 projectionMatrix;
-uniform mat4 viewMatrix;
 
 // Vertex attributes
 attribute vec3 position;
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/data/textures/effect-2d.png and /tmp/VEvcU31eDG/glmark2-2021.02/data/textures/effect-2d.png differ
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/changelog glmark2-2021.02/debian/changelog
--- glmark2-2014.03+git20150611.fa71af2d/debian/changelog	2021-01-04 05:21:54.000000000 +0800
+++ glmark2-2021.02/debian/changelog	2021-05-22 23:30:19.000000000 +0800
@@ -1,3 +1,11 @@
+glmark2 (2021.02-0ubuntu1) hirsute; urgency=medium
+
+  * New upstream release 2021.02
+    https://github.com/glmark2/glmark2/tree/2021.02
+  * Build using python3
+
+ -- Soar Huang <soar.huang@canonical.com>  Sat, 22 May 2021 23:30:19 +0800
+
 glmark2 (2014.03+git20150611.fa71af2d-0ubuntu8) hirsute; urgency=medium
 
   * debian/rules:
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/control glmark2-2021.02/debian/control
--- glmark2-2014.03+git20150611.fa71af2d/debian/control	2021-01-04 04:43:15.000000000 +0800
+++ glmark2-2021.02/debian/control	2021-05-22 23:30:19.000000000 +0800
@@ -4,7 +4,7 @@
 Maintainer: Ubuntu Developers <ubuntu-devel-discuss@lists.ubuntu.com>
 XSBC-Original-Maintainer: Alexandros Frantzis <alexandros.frantzis@linaro.org>
 Build-Depends: debhelper (>= 9),
-               python2,
+               python3,
                pkg-config,
                libx11-dev,
                libjpeg-dev,
@@ -15,6 +15,8 @@
                libwayland-dev,
                libdrm-dev,
                libgbm-dev,
+               libudev-dev,
+               wayland-protocols,
 Standards-Version: 3.9.6
 Homepage: https://launchpad.net/glmark2
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-drm.install glmark2-2021.02/debian/glmark2-drm.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-drm.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-drm.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2-drm
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-drm.manpages glmark2-2021.02/debian/glmark2-drm.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-drm.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-drm.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2-drm.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-drm.install glmark2-2021.02/debian/glmark2-es2-drm.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-drm.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2-drm.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2-es2-drm
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-drm.manpages glmark2-2021.02/debian/glmark2-es2-drm.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-drm.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2-drm.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2-es2-drm.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2.install glmark2-2021.02/debian/glmark2-es2.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2-es2
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2.manpages glmark2-2021.02/debian/glmark2-es2.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2-es2.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-wayland.install glmark2-2021.02/debian/glmark2-es2-wayland.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-wayland.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2-wayland.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2-es2-wayland
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-wayland.manpages glmark2-2021.02/debian/glmark2-es2-wayland.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-es2-wayland.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-es2-wayland.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2-es2-wayland.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2.install glmark2-2021.02/debian/glmark2.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2.manpages glmark2-2021.02/debian/glmark2.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-wayland.install glmark2-2021.02/debian/glmark2-wayland.install
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-wayland.install	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-wayland.install	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+usr/bin/glmark2-wayland
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-wayland.manpages glmark2-2021.02/debian/glmark2-wayland.manpages
--- glmark2-2014.03+git20150611.fa71af2d/debian/glmark2-wayland.manpages	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/debian/glmark2-wayland.manpages	2021-05-22 23:30:19.000000000 +0800
@@ -0,0 +1 @@
+debian/tmp/usr/share/man/man1/glmark2-wayland.1
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/patches/c-library-check.patch glmark2-2021.02/debian/patches/c-library-check.patch
--- glmark2-2014.03+git20150611.fa71af2d/debian/patches/c-library-check.patch	2019-01-11 10:14:51.000000000 +0800
+++ glmark2-2021.02/debian/patches/c-library-check.patch	1970-01-01 08:00:00.000000000 +0800
@@ -1,29 +0,0 @@
-Description: wscript: check C libraries using the C compiler
-Author: Alexandre Courbot <acourbot@nvidia.com>
-Applied-Upstream: https://github.com/glmark2/glmark2/commit/88bbc8383fbf8c0ddf5cd133c086df5173975d82
-Last-Update: 2019-01-10
----
-This patch header follows DEP-3: http://dep.debian.net/deps/dep3/
---- a/wscript
-+++ b/wscript
-@@ -82,17 +82,17 @@
-     # Check required headers
-     req_headers = ['stdlib.h', 'string.h', 'unistd.h', 'stdint.h', 'stdio.h', 'jpeglib.h']
-     for header in req_headers:
--        ctx.check_cxx(header_name = header, auto_add_header_name = True, mandatory = True)
-+        ctx.check_cc(header_name = header, auto_add_header_name = True, mandatory = True)
- 
-     # Check for required libs
-     req_libs = [('m', 'm'), ('jpeg', 'jpeg')]
-     for (lib, uselib) in req_libs:
--        ctx.check_cxx(lib = lib, uselib_store = uselib)
-+        ctx.check_cc(lib = lib, uselib_store = uselib)
- 
-     # Check required functions
-     req_funcs = [('memset', 'string.h', []) ,('sqrt', 'math.h', ['m'])]
-     for func, header, uselib in req_funcs:
--        ctx.check_cxx(function_name = func, header_name = header,
-+        ctx.check_cc(function_name = func, header_name = header,
-                       uselib = uselib, mandatory = True)
- 
-     # Check for a supported version of libpng
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/patches/series glmark2-2021.02/debian/patches/series
--- glmark2-2014.03+git20150611.fa71af2d/debian/patches/series	2019-01-11 10:14:42.000000000 +0800
+++ glmark2-2021.02/debian/patches/series	1970-01-01 08:00:00.000000000 +0800
@@ -1 +0,0 @@
-c-library-check.patch
diff -Nru glmark2-2014.03+git20150611.fa71af2d/debian/rules glmark2-2021.02/debian/rules
--- glmark2-2014.03+git20150611.fa71af2d/debian/rules	2021-01-04 04:42:54.000000000 +0800
+++ glmark2-2021.02/debian/rules	2021-05-22 23:30:19.000000000 +0800
@@ -19,7 +19,7 @@
 	dh $@
 
 override_dh_auto_configure:
-	./waf configure --with-flavors=$(FLAVORS) --prefix=/usr --no-werror
+	./waf configure --with-flavors=$(FLAVORS) --prefix=/usr
 
 override_dh_auto_build:
 	./waf
diff -Nru glmark2-2014.03+git20150611.fa71af2d/doc/glmark2.1.in glmark2-2021.02/doc/glmark2.1.in
--- glmark2-2014.03+git20150611.fa71af2d/doc/glmark2.1.in	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/doc/glmark2.1.in	2021-04-25 01:47:22.000000000 +0800
@@ -1,4 +1,4 @@
-.TH @APPNAME@ "1" "March 2014" "@appname@ @appversion@"
+.TH @APPNAME@ "1" "February 2021" "@appname@ @appversion@"
 .SH NAME
 @appname@ \- OpenGL (ES) 2.0 benchmark suite
 .SH SYNOPSIS
@@ -21,6 +21,9 @@
 Run a quick output validation test instead of
 running the benchmarks
 .TP
+\fB\-\-data-path\fR PATH
+Path to glmark2 models, shaders and textures
+.TP
 \fB\-\-frame-end\fR METHOD
 How to end a frame [default,none,swap,finish,readpixels]
 .TP
diff -Nru glmark2-2014.03+git20150611.fa71af2d/doc/meson.build glmark2-2021.02/doc/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/doc/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/doc/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,13 @@
+foreach flavor : flavors
+    binary = flavor_info[flavor][0]
+    config = configuration_data()
+    config.set('APPNAME', binary.to_upper())
+    config.set('appname', binary)
+    config.set('appversion', full_version)
+    f = configure_file(
+        input: 'glmark2.1.in',
+        output: binary + '.1',
+        configuration: config
+        )
+    install_man(f)
+endforeach
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20100715 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20100715
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20100715	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20100715	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,12 @@
+glmark2 10.07 (20100715)
+========================
+
+* First release.
+* Included benchmarks:
+  - Rendering using vertex arrays
+  - Rendering using VBOs
+  - Texturing using nearest filtering
+  - Texturing using linear filtering
+  - Texturing using trilinear mipmapped filtering
+  - Lighting per vertex using simple GLSL shaders
+  - Lighting per pixel using elaborate GLSL shaders
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20100728 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20100728
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20100728	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20100728	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4 @@
+glmark2 10.07.1 (20100728)
+==========================
+
+* Define the precision of fragment shader variables for OpenGL ES 2.0.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110125 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110125
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110125	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110125	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,8 @@
+glmark2 11.01 (20110125)
+==========================
+
+* Fix visual corruption in glmark2-es2 due to missing depth buffer.
+* Fix linking issues with gcc 4.5.
+* Use correct GL functions to manipulate shaders vs programs.
+* Make result reporting more parser-friendly.
+* Upgrade build system to waf 1.6.2.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110530 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110530
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110530	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110530	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,8 @@
+glmark2 11.05 (20110530)
+==========================
+
+* Don't use the SDL_OPENGL flag for GLESv2 (LP: #761848).
+* Ensure our screen updates are not synchronized with the vertical
+  retrace (LP: 761855).
+* Use the correct GL headers depending on the flavor (desktop vs ES2).
+* Query the correct GL object for shader linking status.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110624 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110624
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110624	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110624	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,16 @@
+glmark2 2011.06 (20110624)
+==========================
+
+* Improve benchmark versatility by allowing runtime-configurable,
+  per-scene options.
+* Add command line option to list avalaible scenes and their
+  supported options (-l,--list-scenes).
+* Allow specifying the scenes to run and their options from the
+  command line (-b,--benchmark).
+* Add basic output validation functionality (--validate).
+* Add command line option to call glFinish() instead of swapping
+  the front and back buffers (--no-swap-buffers).
+* Manually disable VSync for GL/GLX (work around an SDL bug).
+* Replace custom math and shader infrastructure with functionality
+  provided by LibMatrix (lp:libmatrix).
+* Improve user documentation (--help and man page).
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110719 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110719
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110719	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110719	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+glmark2 2011.07 (20110719)
+==========================
+
+* Replace SDL with custom window handling code.
+* Add benchmark for shader conditionals.
+* Add benchmark for shader function calls.
+* Add benchmark for shader loops.
+* Add benchmark for real phong (vs blinn-phong) lighting model.
+* Add benchmark for normal mapping.
+* Refactor Mesh class to increase flexibility in vertex attribute
+  handling.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110818 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110818
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110818	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110818	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2011.08 (20110818)
+==========================
+
+* Port to Android (see INSTALL.android).
+* Add benchmark based on pulsar X11 GL screensaver.
+* Add benchmark for 2D image processing using the GPU.
+* Add command line option to set the size of the output window (-s, --size).
+* Implement ShaderSource object to simplify complex shader creation.
+* Implement utility functions to access resources in an abstract way.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110921 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110921
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20110921	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20110921	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2011.09 (20110921)
+==========================
+
+* Add benchmark for blur desktop effect.
+* Add support for multiple lights in the phong shading benchmark.
+* Add support for loading models from OBJ geometry files.
+* Add Stanford Bunny model and make it available in the build benchmark.
+* Add per-scene options to set shader precision at runtime.
+* Add command-line option to specify the benchmarks to run using a text file.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111018 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111018
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111018	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111018	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2011.10 (20111018)
+==========================
+
+* Add benchmark for buffer (VBO) updates.
+* Add benchmark for drop-shadow desktop effect.
+* Add support for glmark2 extra large models.
+* Enable the selection of additional models in the shading benchmark.
+* Gracefully handle unsupported OpenGL versions, on both X11 and Android.
+* Disable screen dimming and screen rotation on Android.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111116 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111116
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111116	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111116	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2011.11 (20111116)
+==========================
+
+* Add benchmark for bump mapping using a height map.
+* Add benchmark for bump mapping using a tangent space normal map
+  (not included in the default benchmarks).
+* Implement validation support for all default benchmarks.
+* Add a colored prefix to log messages in debug mode.
+* Clean up and refactor code.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111215 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111215
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20111215	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20111215	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2011.12 (20111215)
+==========================
+
+* Add benchmark for bilinear filtering implemented in the fragment
+  shader (not included in the default benchmarks).
+* Ensure we don't call any GL functions before binding a GL context.
+* Fix bug in the ShaderSource object that could lead to shader
+  compilation errors in strict OpenGL ES 2.0 implementations.
+* Refactor scene update code to reduce duplication.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120119 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120119
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120119	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120119	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,15 @@
+glmark2 2012.01 (20120119)
+==========================
+
+* X11:
+  - Run each benchmark in a fresh GL context. Use --reuse-context
+    to revert to the old default behavior of using the same context
+    for all benchmarks.
+  - Add option for showing a live FPS counter on screen (--show-fps).
+  - If the list of benchmarks to run contains only option-setting
+    descriptions, run the default benchmarks.
+* Android:
+  - Fix crash on platforms not supporting glMapBufferOES.
+  - Log the glmark2 score when finishing.
+  - Log an error message if we fail to find a suitable EGLContext.
+  - Implement various stability improvements.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120216 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120216
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120216	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120216	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,15 @@
+glmark2 2012.02 (20120216)
+==========================
+
+* Consolidate X11 and Android main loops.
+* Accept command-line options on Android through either a file
+  ('/data/glmark2/args') or an extra intent key ('args').
+* Support per-scene options for displaying an FPS count on screen
+  (show-fps, fps-pos, fps-size), and remove --show-fps command-line option.
+* Support per-scene options for displaying a benchmark title on screen
+  (title, title-pos, title-size).
+* Add command-line option to run benchmarks indefinitely (--run-forever).
+* Add command-line option to annotate the benchmarks with on-screen
+  information (--annotate == -b :show-fps=true:title=#info#).
+* Move various utility classes to libmatrix.
+* Fix build issues on Android ICS.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120322 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120322
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120322	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120322	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,7 @@
+glmark2 2012.03 (20120322)
+==========================
+
+* X11:
+  - Support rendering to an off-screen surface (--off-screen).
+  - Add command-line option to select which method to use to "end"
+    a frame (--frame-end).
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120524 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120524
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120524	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120524	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,8 @@
+glmark2 2012.05 (20120524)
+==========================
+
+* Add benchmark based on the SGI "Ideas in Motion" demo.
+* Ensure that the framebuffer is drawn opaquely.
+* Add command-line option to configure the visual used for rendering
+  (--visual-config).
+* Add support for additional models and textures in the 'texture' scene.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120621 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120621
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120621	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120621	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,5 @@
+glmark2 2012.06 (20120621)
+==========================
+
+* Add command-line option to render in fullscreen mode (--fullscreen).
+* Display the frame time in addition to the FPS for each benchmark.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120719 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120719
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120719	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120719	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,10 @@
+glmark2 2012.07 (20120719)
+==========================
+
+* Add Android GUI for defining and running benchmarks.
+* Add benchmark based on the WebGL jellyfish demo (scene 'jellyfish').
+* Add benchmark based on the WebGL dynamic terrain demo (scene 'terrain').
+* Extend texture scene with the option to compute texture coordinates
+  in the shader (option 'texgen').
+* Add support for reading texture data from JPEG files.
+* Properly support options that have a finite set of acceptable values.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120817 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120817
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20120817	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20120817	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,12 @@
+glmark2 2012.08 (20120817)
+==========================
+
+* Display benchmark results in the Android GUI.
+* Add an Android option menu offering load/save/delete benchmark list
+  functionality, access to last benchmark results, settings and
+  information about glmark2.
+* Properly handle spaces in scene options on Android.
+* Exclude from score calculation benchmarks whose set up wasn't successful.
+* Display an appropriate message (instead of a 0 FPS value) if benchmark
+  set up isn't successful.
+* Add support for standard GNU installation directories.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20121126 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20121126
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20121126	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20121126	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,5 @@
+glmark2 2012.11 (20121126)
+==========================
+
+* Add a shadow mapping benchmark (scene 'shadow').
+* Consolidate EGL interactions into a separate state object.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20121218 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20121218
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20121218	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20121218	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,8 @@
+glmark2 2012.12 (20121218)
+==========================
+
+* Add a refraction (2 interface) and reflection benchmark (scene 'reflect').
+* Add a new canvas object which allows rendering to GBM managed surfaces
+  and page flipping to DRM managed framebuffer from a console.
+* Enhancements to visual config mechanism to consider stencil size in the
+  selection criteria and to query all config attribs.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20140310 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20140310
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20140310	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20140310	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2014.03 (20140310)
+==========================
+
+* Refactor canvas code to make it easier to support new platforms.
+* Add Wayland backend.
+* Add Mir backend.
+* Add a new 'cel' (or 'toon') shading model (scene 'shading').
+* Add a new 'clear' scene, which just clears the canvas.
+* Improve support for the Wavefront .obj file format.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20170728 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20170728
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20170728	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20170728	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,8 @@
+glmark2 2017.07 (20170728)
+==========================
+
+* Use monotonic clock for benchmark timings.
+* Use udev to find the best node path to use in the DRM backend.
+* Don't rebuild common source needlessly.
+* Build in C++14 mode.
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20200428 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20200428
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20200428	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20200428	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+glmark2 2020.04 (20200428)
+==========================
+
+* Port Wayland flavor to xdg-shell window management.
+* Support recent Android SDK/NDK versions.
+* Add support for Windows via WGL and ANGLE-EGL.
+* Support Raspberry Pi's dispmanx.
+* Use glad for GL headers and dynamic GL library loading.
+* Add --data-path command-line option to set data path at runtime.
+* Add 'nframes' scene option to limit the number of rendered frames.
+* Add F-Droid/fastlane metadata.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20210215 glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20210215
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/changelogs/20210215	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/changelogs/20210215	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,9 @@
+glmark2 2021.02 (20210215)
+==========================
+
+* Meson build system support for DRM/X11/Wayland builds.
+* Fix precision handling in various fragment shaders.
+* Fix spurious failures when using --validate.
+* Always draw to the correct, offscreen buffer when using --off-screen.
+* Don't prefer framebuffer formats with more than 8 bytes per component.
+* Add basic mouse and keyboard support on Wayland.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/full_description.txt glmark2-2021.02/fastlane/metadata/android/en-US/full_description.txt
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/full_description.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/full_description.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,5 @@
+This benchmark contains tests for standard OpenGL (ES) 2.0 features, such as vertex arrays, VBOs, texturing and shaders.
+
+glmark2 is developed by Alexandros Frantzis and Jesse Barker based on the original glmark benchmark by Ben Smith.
+
+It is licensed under the GPLv3.
\ No newline at end of file
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/icon.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/icon.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/phoneScreenshots/1.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/phoneScreenshots/1.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/phoneScreenshots/2.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/phoneScreenshots/2.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/phoneScreenshots/3.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/phoneScreenshots/3.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/phoneScreenshots/4.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/phoneScreenshots/4.png differ
Binary files /tmp/XNsyvrPWB7/glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/images/phoneScreenshots/5.png and /tmp/VEvcU31eDG/glmark2-2021.02/fastlane/metadata/android/en-US/images/phoneScreenshots/5.png differ
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/short_description.txt glmark2-2021.02/fastlane/metadata/android/en-US/short_description.txt
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/short_description.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/short_description.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1 @@
+glmark2 is an OpenGL 2.0 and ES 2.0 benchmark.
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/title.txt glmark2-2021.02/fastlane/metadata/android/en-US/title.txt
--- glmark2-2014.03+git20150611.fa71af2d/fastlane/metadata/android/en-US/title.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/fastlane/metadata/android/en-US/title.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1 @@
+glmark2
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/.github/workflows/build.yml glmark2-2021.02/.github/workflows/build.yml
--- glmark2-2014.03+git20150611.fa71af2d/.github/workflows/build.yml	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/.github/workflows/build.yml	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,79 @@
+name: build
+
+on:
+  push:
+    branches:
+    - master
+  pull_request:
+    branches:
+    - master
+
+jobs:
+  build-meson:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v1
+    - name: Install dependencies
+      run: >
+        sudo apt install meson libdrm-dev libgbm-dev libudev-dev
+        libwayland-dev wayland-protocols libx11-dev
+    - name: Setup
+      run: meson setup build -Dflavors=x11-gl,x11-glesv2,wayland-gl,wayland-glesv2,drm-gl,drm-glesv2
+    - name: Build
+      run: ninja -C build
+    - name: Install
+      run: DESTDIR=/tmp/glmark2-install ninja -C build install
+
+  build-meson-only-drm:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v1
+    - name: Install dependencies
+      run: sudo apt install meson libdrm-dev libgbm-dev libudev-dev
+    - name: Setup
+      run: meson setup build -Dflavors=drm-gl,drm-glesv2
+    - name: Build
+      run: ninja -C build
+    - name: Install
+      run: DESTDIR=/tmp/glmark2-install ninja -C build install
+
+  build-meson-only-wayland:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v1
+    - name: Install dependencies
+      run: sudo apt install meson libwayland-dev wayland-protocols
+    - name: Setup
+      run: meson setup build -Dflavors=wayland-gl,wayland-glesv2
+    - name: Build
+      run: ninja -C build
+    - name: Install
+      run: DESTDIR=/tmp/glmark2-install ninja -C build install
+
+  build-meson-only-x11:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v1
+    - name: Install dependencies
+      run: sudo apt install meson libx11-dev
+    - name: Setup
+      run: meson setup build -Dflavors=x11-gl,x11-glesv2
+    - name: Build
+      run: ninja -C build
+    - name: Install
+      run: DESTDIR=/tmp/glmark2-install ninja -C build install
+
+  build-waf:
+    runs-on: ubuntu-20.04
+    steps:
+    - uses: actions/checkout@v1
+    - name: Install dependencies
+      run: >
+        sudo apt install libdrm-dev libgbm-dev libudev-dev
+        libwayland-dev wayland-protocols libx11-dev
+    - name: Setup
+      run: ./waf configure --with-flavors=x11-gl,x11-glesv2,wayland-gl,wayland-glesv2,drm-gl,drm-glesv2
+    - name: Build
+      run: ./waf build
+    - name: Install
+      run: DESTDIR=/tmp/glmark2-install ./waf install
diff -Nru glmark2-2014.03+git20150611.fa71af2d/INSTALL glmark2-2021.02/INSTALL
--- glmark2-2014.03+git20150611.fa71af2d/INSTALL	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/INSTALL	2021-04-25 01:47:22.000000000 +0800
@@ -1,8 +1,31 @@
-glmark2 uses the WAF build system.
+Meson build system
+------------------
+
+glmark2 uses the meson build system for the most common build flavors (X11,
+Wayland, DRM).
+
+To configure glmark2 use:
+
+$ meson setup build -Dflavors=drm-gl,drm-glesv2,wayland-gl,wayland-glesv2,x11-gl,x11-glesv2 [-Ddata-path=DATA_PATH --prefix=PREFIX]
+
+To build use:
+
+$ ninja -C build
+
+To install use:
+
+$ [DESTDIR=<destdir>] ninja -C build install
+
+WAF build system
+----------------
+
+glmark2 previously used the WAF build system, and the builds for some flavors
+(dispmanx, win32, mir) have not been ported to the meson build system. For such
+cases follow the instructions below.
 
 To configure glmark2 use:
 
-$ ./waf configure --with-flavors=<drm-gl,drm-glesv2,mir-gl,mir-glesv2,wayland-gl,wayland-glesv2,x11-gl,x11-glesv2> [--data-path=DATA_PATH --prefix=PREFIX]
+$ ./waf configure --with-flavors=dispmanx-glesv2,win32-gl,win32-glesv2,mir-gl,mir-glesv2... [--data-path=DATA_PATH --prefix=PREFIX]
 
 To build use:
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/INSTALL.android glmark2-2021.02/INSTALL.android
--- glmark2-2014.03+git20150611.fa71af2d/INSTALL.android	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/INSTALL.android	2021-04-25 01:47:22.000000000 +0800
@@ -1,8 +1,26 @@
-The minimum Android API level for glmark2 is 9 (>= Android 2.3).
+A build script for Linux has been added which simplifies the process of
+creating an APK.
+
+Any recent toolchain (released in 2018 or later) should work.  The steps
+using build-tools have been hard coded to 26.0.1, but could be made
+configurable, or update the script to point to installed version.
+
+Tested with NDK r18b.
+
+$ sudo apt-get install openjdk-8-jdk
+$ export ANDROID_SDK=/path/to/Android/Sdk
+$ export ANDROID_NDK=/path/to/Android/Sdk/ndk-bundle
+$ export JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64
+$ cd android
+$ ./build.sh
+$ adb install glmark2.apk
 
 Building using the SDK and NDK
 ------------------------------
 
+Note: These build steps require older SDK and NDK.  Later revisions have
+removed the android and ant scripts.
+
 To build and install glmark2 you need the Android SDK and NDK. The 'android',
 'adb' and 'ndk-build' tools used below are included there.
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/meson.build glmark2-2021.02/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,67 @@
+project(
+    'glmark2',
+    ['cpp', 'c'],
+    default_options : ['cpp_std=c++14'],
+    version : '2021.02',
+    meson_version: '>=0.47'
+    )
+
+cpp = meson.get_compiler('cpp')
+
+full_version = meson.project_version() + get_option('version-suffix')
+add_global_arguments('-DGLMARK_VERSION="@0@"'.format(full_version), language : 'cpp')
+
+data_path = get_option('data-path')
+if data_path == ''
+    data_path = join_paths([get_option('prefix'), get_option('datadir'), 'glmark2'])
+endif
+add_global_arguments('-DGLMARK_DATA_PATH="@0@"'.format(data_path), language : 'cpp')
+
+extras_path = get_option('extras-path')
+if extras_path != ''
+    add_global_arguments('-DGLMARK_EXTRAS_PATH="@0@"'.format(extras_path), language : 'cpp')
+endif
+
+m_dep = cpp.find_library('m', required : false)
+dl_dep = cpp.find_library('dl')
+libjpeg_dep = dependency('libjpeg')
+libpng_dep = dependency('libpng')
+
+flavors = get_option('flavors')
+if flavors.length() == 0
+    error('Need at least one flavor with -Dflavors=flavor1[,flavors2]...\n'+
+          'See meson_options.txt contents for available flavors')
+endif
+
+flavors_str = ', '.join(flavors)
+need_x11 = flavors_str.contains('x11-')
+need_drm = flavors_str.contains('drm-')
+need_wayland = flavors_str.contains('wayland-')
+need_gl = flavors_str.contains('-gl')
+need_glesv2 = flavors_str.contains('-glesv2')
+need_egl = need_drm or need_wayland or need_glesv2
+need_glx = flavors.contains('x11-gl')
+
+if need_x11
+    x11_dep = dependency('x11')
+endif
+
+if need_drm
+    libdrm_dep = dependency('libdrm')
+    gbm_dep = dependency('gbm')
+    libudev_dep = dependency('libudev')
+endif
+
+if need_wayland
+    wayland_client_dep = dependency('wayland-client')
+    wayland_cursor_dep = dependency('wayland-cursor')
+    wayland_egl_dep = dependency('wayland-egl')
+    wayland_protocols_dep = dependency('wayland-protocols', version : '>= 1.12')
+    wayland_scanner_dep = dependency('wayland-scanner', native: true)
+endif
+
+subdir('src')
+subdir('data')
+subdir('doc')
+
+message('Building with flavors: ' + flavors_str)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/meson_options.txt glmark2-2021.02/meson_options.txt
--- glmark2-2014.03+git20150611.fa71af2d/meson_options.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/meson_options.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,15 @@
+option('flavors', type : 'array',
+       choices : [
+           'drm-gl',
+           'drm-glesv2',
+           'wayland-gl',
+           'wayland-glesv2',
+           'x11-gl',
+           'x11-glesv2'
+        ],
+        value : []
+)
+
+option('data-path', type : 'string', value : '')
+option('extras-path', type : 'string', value : '')
+option('version-suffix', type : 'string', value : '')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/NEWS glmark2-2021.02/NEWS
--- glmark2-2014.03+git20150611.fa71af2d/NEWS	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/NEWS	2021-04-25 01:47:22.000000000 +0800
@@ -1,3 +1,33 @@
+glmark2 2021.02 (20210215)
+==========================
+
+* Meson build system support for DRM/X11/Wayland builds.
+* Fix precision handling in various fragment shaders.
+* Fix spurious failures when using --validate.
+* Always draw to the correct, offscreen buffer when using --off-screen.
+* Don't prefer framebuffer formats with more than 8 bytes per component.
+* Add basic mouse and keyboard support on Wayland.
+
+glmark2 2020.04 (20200428)
+==========================
+
+* Port Wayland flavor to xdg-shell window management.
+* Support recent Android SDK/NDK versions.
+* Add support for Windows via WGL and ANGLE-EGL.
+* Support Raspberry Pi's dispmanx.
+* Use glad for GL headers and dynamic GL library loading.
+* Add --data-path command-line option to set data path at runtime.
+* Add 'nframes' scene option to limit the number of rendered frames.
+* Add F-Droid/fastlane metadata.
+
+glmark2 2017.07 (20170728)
+==========================
+
+* Use monotonic clock for benchmark timings.
+* Use udev to find the best node path to use in the DRM backend.
+* Don't rebuild common source needlessly.
+* Build in C++14 mode.
+
 glmark2 2014.03 (20140310)
 ==========================
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/README glmark2-2021.02/README
--- glmark2-2014.03+git20150611.fa71af2d/README	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/README	2021-04-25 01:47:22.000000000 +0800
@@ -5,19 +5,14 @@
 
 It is licensed under the GPLv3 (see COPYING).
 
-To build glmark2 for X11/GL(ES2) you need:
+To build glmark2 you need:
 
- * python 2.x (>= 2.4) for the build system (waf)
- * libpng 1.2
-
-and for OpenGL 2.0:
-
- * libGL
-
-or for OpenGL ES 2.0:
-
- * libEGL
- * libGLESv2
+ * meson (>= 0.47) or python3 (if you are using WAF for building)
+ * libpng 1.6
+ * Window system development files for the flavors you want to build
+   (e.g, X11, Wayland, drm)
+ * libGL (for desktop GL)
+ * libEGL and libGLESv2 (for GLESv2)
 
 Read the INSTALL file for building/installation instructions,
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/canvas-android.cpp glmark2-2021.02/src/canvas-android.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/canvas-android.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/canvas-android.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,7 +24,6 @@
 #include "log.h"
 #include "options.h"
 #include "gl-headers.h"
-#include <EGL/egl.h>
 
 #include <fstream>
 #include <sstream>
@@ -41,8 +40,26 @@
         EGL_NONE
     };
 
+    /* Init the GLAD entrypoints */
+    if (!egl_lib_.open_from_alternatives({"libEGL.so", "libEGL.so.1" })) {
+        Log::error("Error loading EGL library\n");
+        return false;
+    }
+
+    if (gladLoadEGLUserPtr(EGL_NO_DISPLAY, load_proc, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
     /* Get the current EGL config */
     EGLDisplay egl_display(eglGetCurrentDisplay());
+
+    /* Reinitialize GLAD with a known display */
+    if (gladLoadEGLUserPtr(egl_display, load_proc, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points with display failed\n");
+        return false;
+    }
+
     EGLContext egl_context(eglGetCurrentContext());
     EGLConfig egl_config(0);
     EGLint num_configs;
@@ -51,6 +68,17 @@
 
     eglChooseConfig(egl_display, attribs, &egl_config, 1, &num_configs);
 
+    /* Before calling GLES functions, init GLAD GLES */
+    if (!gles_lib_.open("libGLESv2.so")) {
+        Log::error("Error loading GLES library\n");
+        return false;
+    }
+
+    if (!gladLoadGLES2UserPtr(load_proc, &gles_lib_)) {
+        Log::error("Loading GLESv2 entry points failed.");
+        return false;
+    }
+
     resize(width_, height_);
 
     if (!eglSwapInterval(egl_display, 0))
@@ -168,6 +196,20 @@
  * Private methods *
  *******************/
 
+GLADapiproc
+CanvasAndroid::load_proc(void *userdata, const char *name)
+{
+    if (eglGetProcAddress) {
+        GLADapiproc sym = reinterpret_cast<GLADapiproc>(eglGetProcAddress(name));
+        if (sym) {
+            return sym;
+        }
+    }
+
+    SharedLibrary* lib = reinterpret_cast<SharedLibrary*>(userdata);
+    return reinterpret_cast<GLADapiproc>(lib->load(name));
+}
+
 void
 CanvasAndroid::init_gl_extensions()
 {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/canvas-android.h glmark2-2021.02/src/canvas-android.h
--- glmark2-2014.03+git20150611.fa71af2d/src/canvas-android.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/canvas-android.h	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,8 @@
 #define GLMARK2_CANVAS_ANDROID_H_
 
 #include "canvas.h"
+#include "shared-library.h"
+#include "glad/egl.h"
 
 /**
  * Canvas for rendering to Android surfaces.
@@ -49,6 +51,10 @@
     void resize(int width, int height);
 
 private:
+    SharedLibrary egl_lib_;
+    SharedLibrary gles_lib_;
+
+    static GLADapiproc load_proc(void *userdata, const char *name);
     void init_gl_extensions();
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/canvas-generic.cpp glmark2-2021.02/src/canvas-generic.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/canvas-generic.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/canvas-generic.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -103,7 +103,7 @@
     Options::FrameEnd m = Options::frame_end;
 
     if (m == Options::FrameEndDefault) {
-        if (offscreen_)
+        if (offscreen_ || Options::validate)
             m = Options::FrameEndFinish;
         else
             m = Options::FrameEndSwap;
@@ -194,7 +194,8 @@
 bool
 CanvasGeneric::supports_gl2()
 {
-    std::string gl_version_str(reinterpret_cast<const char*>(glGetString(GL_VERSION)));
+    const char *version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
+    std::string gl_version_str(version ? version : "");
     int gl_major(0);
 
     size_t point_pos(gl_version_str.find('.'));
@@ -221,7 +222,7 @@
 {
     bool request_fullscreen = (width == -1 && height == -1);
 
-    int vid;
+    intptr_t vid;
     if (!gl_state_.gotNativeConfig(vid))
     {
         Log::error("Error: Couldn't get GL visual config!\n");
@@ -249,6 +250,7 @@
     }
 
     native_window_ = native_state_.window(cur_properties);
+    window_initialized_ = true;
 
     width_ = cur_properties.width;
     height_ = cur_properties.height;
@@ -274,6 +276,12 @@
 bool
 CanvasGeneric::do_make_current()
 {
+    if (!window_initialized_) {
+        Log::error("glwindow has never been initialized, check native-state code\n"
+                   "it should not return a valid window until create_window() is called\n");
+        return false;
+    }
+
     gl_state_.init_surface(native_window_);
 
     if (!gl_state_.valid()) {
@@ -281,7 +289,8 @@
         return false;
     }
 
-    gl_state_.init_gl_extensions();
+    if (!gl_state_.init_gl_extensions())
+        return false;
 
     if (offscreen_) {
         if (!ensure_fbo())
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/canvas-generic.h glmark2-2021.02/src/canvas-generic.h
--- glmark2-2014.03+git20150611.fa71af2d/src/canvas-generic.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/canvas-generic.h	2021-04-25 01:47:22.000000000 +0800
@@ -37,9 +37,10 @@
     CanvasGeneric(NativeState& native_state, GLState& gl_state,
                   int width, int height)
         : Canvas(width, height),
-          native_state_(native_state), gl_state_(gl_state),
+          native_state_(native_state), gl_state_(gl_state), native_window_(0),
           gl_color_format_(0), gl_depth_format_(0),
-          color_renderbuffer_(0), depth_renderbuffer_(0), fbo_(0) {}
+          color_renderbuffer_(0), depth_renderbuffer_(0), fbo_(0),
+          window_initialized_(false) {}
 
     bool init();
     bool reset();
@@ -70,6 +71,7 @@
     GLuint color_renderbuffer_;
     GLuint depth_renderbuffer_;
     GLuint fbo_;
+    bool window_initialized_;
 };
 
 #endif /* GLMARK2_CANVAS_GENERIC_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/EGL/eglplatform.h glmark2-2021.02/src/glad/include/EGL/eglplatform.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/EGL/eglplatform.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/EGL/eglplatform.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,176 @@
+#ifndef __eglplatform_h_
+#define __eglplatform_h_
+
+/*
+** Copyright (c) 2007-2016 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+/* Platform-specific types and definitions for egl.h
+ * $Revision: 30994 $ on $Date: 2015-04-30 13:36:48 -0700 (Thu, 30 Apr 2015) $
+ *
+ * Adopters may modify khrplatform.h and this file to suit their platform.
+ * You are encouraged to submit all modifications to the Khronos group so that
+ * they can be included in future versions of this file.  Please submit changes
+ * by sending them to the public Khronos Bugzilla (http://khronos.org/bugzilla)
+ * by filing a bug against product "EGL" component "Registry".
+ */
+
+#include <KHR/khrplatform.h>
+
+/* Macros used in EGL function prototype declarations.
+ *
+ * EGL functions should be prototyped as:
+ *
+ * EGLAPI return-type EGLAPIENTRY eglFunction(arguments);
+ * typedef return-type (EXPAPIENTRYP PFNEGLFUNCTIONPROC) (arguments);
+ *
+ * KHRONOS_APICALL and KHRONOS_APIENTRY are defined in KHR/khrplatform.h
+ */
+
+#ifndef EGLAPI
+#define EGLAPI KHRONOS_APICALL
+#endif
+
+#ifndef EGLAPIENTRY
+#define EGLAPIENTRY  KHRONOS_APIENTRY
+#endif
+#define EGLAPIENTRYP EGLAPIENTRY*
+
+/* The types NativeDisplayType, NativeWindowType, and NativePixmapType
+ * are aliases of window-system-dependent types, such as X Display * or
+ * Windows Device Context. They must be defined in platform-specific
+ * code below. The EGL-prefixed versions of Native*Type are the same
+ * types, renamed in EGL 1.3 so all types in the API start with "EGL".
+ *
+ * Khronos STRONGLY RECOMMENDS that you use the default definitions
+ * provided below, since these changes affect both binary and source
+ * portability of applications using EGL running on different EGL
+ * implementations.
+ */
+
+#if defined(_WIN32) || defined(__VC32__) && !defined(__CYGWIN__) && !defined(__SCITECH_SNAP__) /* Win32 and WinCE */
+#ifndef WIN32_LEAN_AND_MEAN
+#define WIN32_LEAN_AND_MEAN 1
+#endif
+#include <windows.h>
+
+typedef HDC     EGLNativeDisplayType;
+typedef HBITMAP EGLNativePixmapType;
+typedef HWND    EGLNativeWindowType;
+
+#elif defined(__EMSCRIPTEN__)
+
+typedef int EGLNativeDisplayType;
+typedef int EGLNativePixmapType;
+typedef int EGLNativeWindowType;
+
+#elif defined(__WINSCW__) || defined(__SYMBIAN32__)  /* Symbian */
+
+typedef int   EGLNativeDisplayType;
+typedef void *EGLNativePixmapType;
+typedef void *EGLNativeWindowType;
+
+#elif defined(WL_EGL_PLATFORM)
+
+typedef struct wl_display     *EGLNativeDisplayType;
+typedef struct wl_egl_pixmap  *EGLNativePixmapType;
+typedef struct wl_egl_window  *EGLNativeWindowType;
+
+#elif defined(__GBM__)
+
+typedef struct gbm_device  *EGLNativeDisplayType;
+typedef struct gbm_bo      *EGLNativePixmapType;
+typedef void               *EGLNativeWindowType;
+
+#elif defined(__ANDROID__) || defined(ANDROID)
+
+struct ANativeWindow;
+struct egl_native_pixmap_t;
+
+typedef void*                           EGLNativeDisplayType;
+typedef struct egl_native_pixmap_t*     EGLNativePixmapType;
+typedef struct ANativeWindow*           EGLNativeWindowType;
+
+#elif defined(USE_OZONE)
+
+typedef intptr_t EGLNativeDisplayType;
+typedef intptr_t EGLNativePixmapType;
+typedef intptr_t EGLNativeWindowType;
+
+#elif defined(__unix__) && defined(EGL_NO_X11)
+
+typedef void             *EGLNativeDisplayType;
+typedef khronos_uintptr_t EGLNativePixmapType;
+typedef khronos_uintptr_t EGLNativeWindowType;
+
+#elif defined(__unix__) || defined(USE_X11)
+
+/* X11 (tentative)  */
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+
+typedef Display *EGLNativeDisplayType;
+typedef Pixmap   EGLNativePixmapType;
+typedef Window   EGLNativeWindowType;
+
+#elif defined(__APPLE__)
+
+typedef int   EGLNativeDisplayType;
+typedef void *EGLNativePixmapType;
+typedef void *EGLNativeWindowType;
+
+#elif defined(__HAIKU__)
+
+#include <kernel/image.h>
+
+typedef void              *EGLNativeDisplayType;
+typedef khronos_uintptr_t  EGLNativePixmapType;
+typedef khronos_uintptr_t  EGLNativeWindowType;
+
+#else
+#error "Platform not recognized"
+#endif
+
+/* EGL 1.2 types, renamed for consistency in EGL 1.3 */
+typedef EGLNativeDisplayType NativeDisplayType;
+typedef EGLNativePixmapType  NativePixmapType;
+typedef EGLNativeWindowType  NativeWindowType;
+
+
+/* Define EGLint. This must be a signed integral type large enough to contain
+ * all legal attribute names and values passed into and out of EGL, whether
+ * their type is boolean, bitmask, enumerant (symbolic constant), integer,
+ * handle, or other.  While in general a 32-bit integer will suffice, if
+ * handles are 64 bit types, then EGLint should be defined as a signed 64-bit
+ * integer type.
+ */
+typedef khronos_int32_t EGLint;
+
+
+/* C++ / C typecast macros for special EGL handle values */
+#if defined(__cplusplus)
+#define EGL_CAST(type, value) (static_cast<type>(value))
+#else
+#define EGL_CAST(type, value) ((type) (value))
+#endif
+
+#endif /* __eglplatform_h */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/egl.h glmark2-2021.02/src/glad/include/glad/egl.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/egl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/glad/egl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,554 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:15 2019
+ *
+ * Generator: C/C++
+ * Specification: egl
+ * Extensions: 4
+ *
+ * APIs:
+ *  - egl=1.5
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='egl=1.5' --extensions='EGL_EXT_platform_base,EGL_KHR_platform_gbm,EGL_KHR_platform_wayland,EGL_KHR_platform_x11' c
+ *
+ * Online:
+ *    http://glad.sh/#api=egl%3D1.5&extensions=EGL_EXT_platform_base%2CEGL_KHR_platform_gbm%2CEGL_KHR_platform_wayland%2CEGL_KHR_platform_x11&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_EGL_H_
+#define GLAD_EGL_H_
+
+
+#define GLAD_EGL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define EGL_ALPHA_FORMAT 0x3088
+#define EGL_ALPHA_FORMAT_NONPRE 0x308B
+#define EGL_ALPHA_FORMAT_PRE 0x308C
+#define EGL_ALPHA_MASK_SIZE 0x303E
+#define EGL_ALPHA_SIZE 0x3021
+#define EGL_BACK_BUFFER 0x3084
+#define EGL_BAD_ACCESS 0x3002
+#define EGL_BAD_ALLOC 0x3003
+#define EGL_BAD_ATTRIBUTE 0x3004
+#define EGL_BAD_CONFIG 0x3005
+#define EGL_BAD_CONTEXT 0x3006
+#define EGL_BAD_CURRENT_SURFACE 0x3007
+#define EGL_BAD_DISPLAY 0x3008
+#define EGL_BAD_MATCH 0x3009
+#define EGL_BAD_NATIVE_PIXMAP 0x300A
+#define EGL_BAD_NATIVE_WINDOW 0x300B
+#define EGL_BAD_PARAMETER 0x300C
+#define EGL_BAD_SURFACE 0x300D
+#define EGL_BIND_TO_TEXTURE_RGB 0x3039
+#define EGL_BIND_TO_TEXTURE_RGBA 0x303A
+#define EGL_BLUE_SIZE 0x3022
+#define EGL_BUFFER_DESTROYED 0x3095
+#define EGL_BUFFER_PRESERVED 0x3094
+#define EGL_BUFFER_SIZE 0x3020
+#define EGL_CLIENT_APIS 0x308D
+#define EGL_CL_EVENT_HANDLE 0x309C
+#define EGL_COLORSPACE 0x3087
+#define EGL_COLORSPACE_LINEAR 0x308A
+#define EGL_COLORSPACE_sRGB 0x3089
+#define EGL_COLOR_BUFFER_TYPE 0x303F
+#define EGL_CONDITION_SATISFIED 0x30F6
+#define EGL_CONFIG_CAVEAT 0x3027
+#define EGL_CONFIG_ID 0x3028
+#define EGL_CONFORMANT 0x3042
+#define EGL_CONTEXT_CLIENT_TYPE 0x3097
+#define EGL_CONTEXT_CLIENT_VERSION 0x3098
+#define EGL_CONTEXT_LOST 0x300E
+#define EGL_CONTEXT_MAJOR_VERSION 0x3098
+#define EGL_CONTEXT_MINOR_VERSION 0x30FB
+#define EGL_CONTEXT_OPENGL_COMPATIBILITY_PROFILE_BIT 0x00000002
+#define EGL_CONTEXT_OPENGL_CORE_PROFILE_BIT 0x00000001
+#define EGL_CONTEXT_OPENGL_DEBUG 0x31B0
+#define EGL_CONTEXT_OPENGL_FORWARD_COMPATIBLE 0x31B1
+#define EGL_CONTEXT_OPENGL_PROFILE_MASK 0x30FD
+#define EGL_CONTEXT_OPENGL_RESET_NOTIFICATION_STRATEGY 0x31BD
+#define EGL_CONTEXT_OPENGL_ROBUST_ACCESS 0x31B2
+#define EGL_CORE_NATIVE_ENGINE 0x305B
+#define EGL_DEFAULT_DISPLAY EGL_CAST(EGLNativeDisplayType,0)
+#define EGL_DEPTH_SIZE 0x3025
+#define EGL_DISPLAY_SCALING 10000
+#define EGL_DONT_CARE EGL_CAST(EGLint,-1)
+#define EGL_DRAW 0x3059
+#define EGL_EXTENSIONS 0x3055
+#define EGL_FALSE 0
+#define EGL_FOREVER 0xFFFFFFFFFFFFFFFF
+#define EGL_GL_COLORSPACE 0x309D
+#define EGL_GL_COLORSPACE_LINEAR 0x308A
+#define EGL_GL_COLORSPACE_SRGB 0x3089
+#define EGL_GL_RENDERBUFFER 0x30B9
+#define EGL_GL_TEXTURE_2D 0x30B1
+#define EGL_GL_TEXTURE_3D 0x30B2
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x30B4
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x30B6
+#define EGL_GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x30B8
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x30B3
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x30B5
+#define EGL_GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x30B7
+#define EGL_GL_TEXTURE_LEVEL 0x30BC
+#define EGL_GL_TEXTURE_ZOFFSET 0x30BD
+#define EGL_GREEN_SIZE 0x3023
+#define EGL_HEIGHT 0x3056
+#define EGL_HORIZONTAL_RESOLUTION 0x3090
+#define EGL_IMAGE_PRESERVED 0x30D2
+#define EGL_LARGEST_PBUFFER 0x3058
+#define EGL_LEVEL 0x3029
+#define EGL_LOSE_CONTEXT_ON_RESET 0x31BF
+#define EGL_LUMINANCE_BUFFER 0x308F
+#define EGL_LUMINANCE_SIZE 0x303D
+#define EGL_MATCH_NATIVE_PIXMAP 0x3041
+#define EGL_MAX_PBUFFER_HEIGHT 0x302A
+#define EGL_MAX_PBUFFER_PIXELS 0x302B
+#define EGL_MAX_PBUFFER_WIDTH 0x302C
+#define EGL_MAX_SWAP_INTERVAL 0x303C
+#define EGL_MIN_SWAP_INTERVAL 0x303B
+#define EGL_MIPMAP_LEVEL 0x3083
+#define EGL_MIPMAP_TEXTURE 0x3082
+#define EGL_MULTISAMPLE_RESOLVE 0x3099
+#define EGL_MULTISAMPLE_RESOLVE_BOX 0x309B
+#define EGL_MULTISAMPLE_RESOLVE_BOX_BIT 0x0200
+#define EGL_MULTISAMPLE_RESOLVE_DEFAULT 0x309A
+#define EGL_NATIVE_RENDERABLE 0x302D
+#define EGL_NATIVE_VISUAL_ID 0x302E
+#define EGL_NATIVE_VISUAL_TYPE 0x302F
+#define EGL_NONE 0x3038
+#define EGL_NON_CONFORMANT_CONFIG 0x3051
+#define EGL_NOT_INITIALIZED 0x3001
+#define EGL_NO_CONTEXT EGL_CAST(EGLContext,0)
+#define EGL_NO_DISPLAY EGL_CAST(EGLDisplay,0)
+#define EGL_NO_IMAGE EGL_CAST(EGLImage,0)
+#define EGL_NO_RESET_NOTIFICATION 0x31BE
+#define EGL_NO_SURFACE EGL_CAST(EGLSurface,0)
+#define EGL_NO_SYNC EGL_CAST(EGLSync,0)
+#define EGL_NO_TEXTURE 0x305C
+#define EGL_OPENGL_API 0x30A2
+#define EGL_OPENGL_BIT 0x0008
+#define EGL_OPENGL_ES2_BIT 0x0004
+#define EGL_OPENGL_ES3_BIT 0x00000040
+#define EGL_OPENGL_ES_API 0x30A0
+#define EGL_OPENGL_ES_BIT 0x0001
+#define EGL_OPENVG_API 0x30A1
+#define EGL_OPENVG_BIT 0x0002
+#define EGL_OPENVG_IMAGE 0x3096
+#define EGL_PBUFFER_BIT 0x0001
+#define EGL_PIXEL_ASPECT_RATIO 0x3092
+#define EGL_PIXMAP_BIT 0x0002
+#define EGL_PLATFORM_GBM_KHR 0x31D7
+#define EGL_PLATFORM_WAYLAND_KHR 0x31D8
+#define EGL_PLATFORM_X11_KHR 0x31D5
+#define EGL_PLATFORM_X11_SCREEN_KHR 0x31D6
+#define EGL_READ 0x305A
+#define EGL_RED_SIZE 0x3024
+#define EGL_RENDERABLE_TYPE 0x3040
+#define EGL_RENDER_BUFFER 0x3086
+#define EGL_RGB_BUFFER 0x308E
+#define EGL_SAMPLES 0x3031
+#define EGL_SAMPLE_BUFFERS 0x3032
+#define EGL_SIGNALED 0x30F2
+#define EGL_SINGLE_BUFFER 0x3085
+#define EGL_SLOW_CONFIG 0x3050
+#define EGL_STENCIL_SIZE 0x3026
+#define EGL_SUCCESS 0x3000
+#define EGL_SURFACE_TYPE 0x3033
+#define EGL_SWAP_BEHAVIOR 0x3093
+#define EGL_SWAP_BEHAVIOR_PRESERVED_BIT 0x0400
+#define EGL_SYNC_CL_EVENT 0x30FE
+#define EGL_SYNC_CL_EVENT_COMPLETE 0x30FF
+#define EGL_SYNC_CONDITION 0x30F8
+#define EGL_SYNC_FENCE 0x30F9
+#define EGL_SYNC_FLUSH_COMMANDS_BIT 0x0001
+#define EGL_SYNC_PRIOR_COMMANDS_COMPLETE 0x30F0
+#define EGL_SYNC_STATUS 0x30F1
+#define EGL_SYNC_TYPE 0x30F7
+#define EGL_TEXTURE_2D 0x305F
+#define EGL_TEXTURE_FORMAT 0x3080
+#define EGL_TEXTURE_RGB 0x305D
+#define EGL_TEXTURE_RGBA 0x305E
+#define EGL_TEXTURE_TARGET 0x3081
+#define EGL_TIMEOUT_EXPIRED 0x30F5
+#define EGL_TRANSPARENT_BLUE_VALUE 0x3035
+#define EGL_TRANSPARENT_GREEN_VALUE 0x3036
+#define EGL_TRANSPARENT_RED_VALUE 0x3037
+#define EGL_TRANSPARENT_RGB 0x3052
+#define EGL_TRANSPARENT_TYPE 0x3034
+#define EGL_TRUE 1
+#define EGL_UNKNOWN EGL_CAST(EGLint,-1)
+#define EGL_UNSIGNALED 0x30F3
+#define EGL_VENDOR 0x3053
+#define EGL_VERSION 0x3054
+#define EGL_VERTICAL_RESOLUTION 0x3091
+#define EGL_VG_ALPHA_FORMAT 0x3088
+#define EGL_VG_ALPHA_FORMAT_NONPRE 0x308B
+#define EGL_VG_ALPHA_FORMAT_PRE 0x308C
+#define EGL_VG_ALPHA_FORMAT_PRE_BIT 0x0040
+#define EGL_VG_COLORSPACE 0x3087
+#define EGL_VG_COLORSPACE_LINEAR 0x308A
+#define EGL_VG_COLORSPACE_LINEAR_BIT 0x0020
+#define EGL_VG_COLORSPACE_sRGB 0x3089
+#define EGL_WIDTH 0x3057
+#define EGL_WINDOW_BIT 0x0004
+
+
+#include <KHR/khrplatform.h>
+#include <EGL/eglplatform.h>
+
+
+
+
+
+
+
+
+
+
+
+struct AHardwareBuffer;
+
+typedef unsigned int EGLBoolean;
+typedef unsigned int EGLenum;
+typedef intptr_t EGLAttribKHR;
+typedef intptr_t EGLAttrib;
+typedef void *EGLClientBuffer;
+typedef void *EGLConfig;
+typedef void *EGLContext;
+typedef void *EGLDeviceEXT;
+typedef void *EGLDisplay;
+typedef void *EGLImage;
+typedef void *EGLImageKHR;
+typedef void *EGLLabelKHR;
+typedef void *EGLObjectKHR;
+typedef void *EGLOutputLayerEXT;
+typedef void *EGLOutputPortEXT;
+typedef void *EGLStreamKHR;
+typedef void *EGLSurface;
+typedef void *EGLSync;
+typedef void *EGLSyncKHR;
+typedef void *EGLSyncNV;
+typedef void (*__eglMustCastToProperFunctionPointerType)(void);
+typedef khronos_utime_nanoseconds_t EGLTimeKHR;
+typedef khronos_utime_nanoseconds_t EGLTime;
+typedef khronos_utime_nanoseconds_t EGLTimeNV;
+typedef khronos_utime_nanoseconds_t EGLuint64NV;
+typedef khronos_uint64_t EGLuint64KHR;
+typedef khronos_stime_nanoseconds_t EGLnsecsANDROID;
+typedef int EGLNativeFileDescriptorKHR;
+typedef khronos_ssize_t EGLsizeiANDROID;
+typedef void (*EGLSetBlobFuncANDROID) (const void *key, EGLsizeiANDROID keySize, const void *value, EGLsizeiANDROID valueSize);
+typedef EGLsizeiANDROID (*EGLGetBlobFuncANDROID) (const void *key, EGLsizeiANDROID keySize, void *value, EGLsizeiANDROID valueSize);
+struct EGLClientPixmapHI {
+    void  *pData;
+    EGLint iWidth;
+    EGLint iHeight;
+    EGLint iStride;
+};
+typedef void (GLAD_API_PTR *EGLDEBUGPROCKHR)(EGLenum error,const char *command,EGLint messageType,EGLLabelKHR threadLabel,EGLLabelKHR objectLabel,const char* message);
+
+
+#define EGL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_0;
+#define EGL_VERSION_1_1 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_1;
+#define EGL_VERSION_1_2 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_2;
+#define EGL_VERSION_1_3 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_3;
+#define EGL_VERSION_1_4 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_4;
+#define EGL_VERSION_1_5 1
+GLAD_API_CALL int GLAD_EGL_VERSION_1_5;
+#define EGL_EXT_platform_base 1
+GLAD_API_CALL int GLAD_EGL_EXT_platform_base;
+#define EGL_KHR_platform_gbm 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_gbm;
+#define EGL_KHR_platform_wayland 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_wayland;
+#define EGL_KHR_platform_x11 1
+GLAD_API_CALL int GLAD_EGL_KHR_platform_x11;
+
+
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLBINDAPIPROC)(EGLenum api);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLBINDTEXIMAGEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLCHOOSECONFIGPROC)(EGLDisplay dpy, const EGLint * attrib_list, EGLConfig * configs, EGLint config_size, EGLint * num_config);
+typedef EGLint (GLAD_API_PTR *PFNEGLCLIENTWAITSYNCPROC)(EGLDisplay dpy, EGLSync sync, EGLint flags, EGLTime timeout);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLCOPYBUFFERSPROC)(EGLDisplay dpy, EGLSurface surface, EGLNativePixmapType target);
+typedef EGLContext (GLAD_API_PTR *PFNEGLCREATECONTEXTPROC)(EGLDisplay dpy, EGLConfig config, EGLContext share_context, const EGLint * attrib_list);
+typedef EGLImage (GLAD_API_PTR *PFNEGLCREATEIMAGEPROC)(EGLDisplay dpy, EGLContext ctx, EGLenum target, EGLClientBuffer buffer, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC)(EGLDisplay dpy, EGLenum buftype, EGLClientBuffer buffer, EGLConfig config, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPBUFFERSURFACEPROC)(EGLDisplay dpy, EGLConfig config, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPIXMAPSURFACEPROC)(EGLDisplay dpy, EGLConfig config, EGLNativePixmapType pixmap, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC)(EGLDisplay dpy, EGLConfig config, void * native_pixmap, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC)(EGLDisplay dpy, EGLConfig config, void * native_pixmap, const EGLint * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMWINDOWSURFACEPROC)(EGLDisplay dpy, EGLConfig config, void * native_window, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC)(EGLDisplay dpy, EGLConfig config, void * native_window, const EGLint * attrib_list);
+typedef EGLSync (GLAD_API_PTR *PFNEGLCREATESYNCPROC)(EGLDisplay dpy, EGLenum type, const EGLAttrib * attrib_list);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLCREATEWINDOWSURFACEPROC)(EGLDisplay dpy, EGLConfig config, EGLNativeWindowType win, const EGLint * attrib_list);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYCONTEXTPROC)(EGLDisplay dpy, EGLContext ctx);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYIMAGEPROC)(EGLDisplay dpy, EGLImage image);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYSURFACEPROC)(EGLDisplay dpy, EGLSurface surface);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLDESTROYSYNCPROC)(EGLDisplay dpy, EGLSync sync);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETCONFIGATTRIBPROC)(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETCONFIGSPROC)(EGLDisplay dpy, EGLConfig * configs, EGLint config_size, EGLint * num_config);
+typedef EGLContext (GLAD_API_PTR *PFNEGLGETCURRENTCONTEXTPROC)(void);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETCURRENTDISPLAYPROC)(void);
+typedef EGLSurface (GLAD_API_PTR *PFNEGLGETCURRENTSURFACEPROC)(EGLint readdraw);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETDISPLAYPROC)(EGLNativeDisplayType display_id);
+typedef EGLint (GLAD_API_PTR *PFNEGLGETERRORPROC)(void);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETPLATFORMDISPLAYPROC)(EGLenum platform, void * native_display, const EGLAttrib * attrib_list);
+typedef EGLDisplay (GLAD_API_PTR *PFNEGLGETPLATFORMDISPLAYEXTPROC)(EGLenum platform, void * native_display, const EGLint * attrib_list);
+typedef __eglMustCastToProperFunctionPointerType (GLAD_API_PTR *PFNEGLGETPROCADDRESSPROC)(const char * procname);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLGETSYNCATTRIBPROC)(EGLDisplay dpy, EGLSync sync, EGLint attribute, EGLAttrib * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLINITIALIZEPROC)(EGLDisplay dpy, EGLint * major, EGLint * minor);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLMAKECURRENTPROC)(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx);
+typedef EGLenum (GLAD_API_PTR *PFNEGLQUERYAPIPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLQUERYCONTEXTPROC)(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint * value);
+typedef const char * (GLAD_API_PTR *PFNEGLQUERYSTRINGPROC)(EGLDisplay dpy, EGLint name);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLQUERYSURFACEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint * value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLRELEASETEXIMAGEPROC)(EGLDisplay dpy, EGLSurface surface, EGLint buffer);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLRELEASETHREADPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSURFACEATTRIBPROC)(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSWAPBUFFERSPROC)(EGLDisplay dpy, EGLSurface surface);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLSWAPINTERVALPROC)(EGLDisplay dpy, EGLint interval);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLTERMINATEPROC)(EGLDisplay dpy);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITCLIENTPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITGLPROC)(void);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITNATIVEPROC)(EGLint engine);
+typedef EGLBoolean (GLAD_API_PTR *PFNEGLWAITSYNCPROC)(EGLDisplay dpy, EGLSync sync, EGLint flags);
+
+GLAD_API_CALL PFNEGLBINDAPIPROC glad_eglBindAPI;
+#define eglBindAPI glad_eglBindAPI
+GLAD_API_CALL PFNEGLBINDTEXIMAGEPROC glad_eglBindTexImage;
+#define eglBindTexImage glad_eglBindTexImage
+GLAD_API_CALL PFNEGLCHOOSECONFIGPROC glad_eglChooseConfig;
+#define eglChooseConfig glad_eglChooseConfig
+GLAD_API_CALL PFNEGLCLIENTWAITSYNCPROC glad_eglClientWaitSync;
+#define eglClientWaitSync glad_eglClientWaitSync
+GLAD_API_CALL PFNEGLCOPYBUFFERSPROC glad_eglCopyBuffers;
+#define eglCopyBuffers glad_eglCopyBuffers
+GLAD_API_CALL PFNEGLCREATECONTEXTPROC glad_eglCreateContext;
+#define eglCreateContext glad_eglCreateContext
+GLAD_API_CALL PFNEGLCREATEIMAGEPROC glad_eglCreateImage;
+#define eglCreateImage glad_eglCreateImage
+GLAD_API_CALL PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC glad_eglCreatePbufferFromClientBuffer;
+#define eglCreatePbufferFromClientBuffer glad_eglCreatePbufferFromClientBuffer
+GLAD_API_CALL PFNEGLCREATEPBUFFERSURFACEPROC glad_eglCreatePbufferSurface;
+#define eglCreatePbufferSurface glad_eglCreatePbufferSurface
+GLAD_API_CALL PFNEGLCREATEPIXMAPSURFACEPROC glad_eglCreatePixmapSurface;
+#define eglCreatePixmapSurface glad_eglCreatePixmapSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC glad_eglCreatePlatformPixmapSurface;
+#define eglCreatePlatformPixmapSurface glad_eglCreatePlatformPixmapSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC glad_eglCreatePlatformPixmapSurfaceEXT;
+#define eglCreatePlatformPixmapSurfaceEXT glad_eglCreatePlatformPixmapSurfaceEXT
+GLAD_API_CALL PFNEGLCREATEPLATFORMWINDOWSURFACEPROC glad_eglCreatePlatformWindowSurface;
+#define eglCreatePlatformWindowSurface glad_eglCreatePlatformWindowSurface
+GLAD_API_CALL PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC glad_eglCreatePlatformWindowSurfaceEXT;
+#define eglCreatePlatformWindowSurfaceEXT glad_eglCreatePlatformWindowSurfaceEXT
+GLAD_API_CALL PFNEGLCREATESYNCPROC glad_eglCreateSync;
+#define eglCreateSync glad_eglCreateSync
+GLAD_API_CALL PFNEGLCREATEWINDOWSURFACEPROC glad_eglCreateWindowSurface;
+#define eglCreateWindowSurface glad_eglCreateWindowSurface
+GLAD_API_CALL PFNEGLDESTROYCONTEXTPROC glad_eglDestroyContext;
+#define eglDestroyContext glad_eglDestroyContext
+GLAD_API_CALL PFNEGLDESTROYIMAGEPROC glad_eglDestroyImage;
+#define eglDestroyImage glad_eglDestroyImage
+GLAD_API_CALL PFNEGLDESTROYSURFACEPROC glad_eglDestroySurface;
+#define eglDestroySurface glad_eglDestroySurface
+GLAD_API_CALL PFNEGLDESTROYSYNCPROC glad_eglDestroySync;
+#define eglDestroySync glad_eglDestroySync
+GLAD_API_CALL PFNEGLGETCONFIGATTRIBPROC glad_eglGetConfigAttrib;
+#define eglGetConfigAttrib glad_eglGetConfigAttrib
+GLAD_API_CALL PFNEGLGETCONFIGSPROC glad_eglGetConfigs;
+#define eglGetConfigs glad_eglGetConfigs
+GLAD_API_CALL PFNEGLGETCURRENTCONTEXTPROC glad_eglGetCurrentContext;
+#define eglGetCurrentContext glad_eglGetCurrentContext
+GLAD_API_CALL PFNEGLGETCURRENTDISPLAYPROC glad_eglGetCurrentDisplay;
+#define eglGetCurrentDisplay glad_eglGetCurrentDisplay
+GLAD_API_CALL PFNEGLGETCURRENTSURFACEPROC glad_eglGetCurrentSurface;
+#define eglGetCurrentSurface glad_eglGetCurrentSurface
+GLAD_API_CALL PFNEGLGETDISPLAYPROC glad_eglGetDisplay;
+#define eglGetDisplay glad_eglGetDisplay
+GLAD_API_CALL PFNEGLGETERRORPROC glad_eglGetError;
+#define eglGetError glad_eglGetError
+GLAD_API_CALL PFNEGLGETPLATFORMDISPLAYPROC glad_eglGetPlatformDisplay;
+#define eglGetPlatformDisplay glad_eglGetPlatformDisplay
+GLAD_API_CALL PFNEGLGETPLATFORMDISPLAYEXTPROC glad_eglGetPlatformDisplayEXT;
+#define eglGetPlatformDisplayEXT glad_eglGetPlatformDisplayEXT
+GLAD_API_CALL PFNEGLGETPROCADDRESSPROC glad_eglGetProcAddress;
+#define eglGetProcAddress glad_eglGetProcAddress
+GLAD_API_CALL PFNEGLGETSYNCATTRIBPROC glad_eglGetSyncAttrib;
+#define eglGetSyncAttrib glad_eglGetSyncAttrib
+GLAD_API_CALL PFNEGLINITIALIZEPROC glad_eglInitialize;
+#define eglInitialize glad_eglInitialize
+GLAD_API_CALL PFNEGLMAKECURRENTPROC glad_eglMakeCurrent;
+#define eglMakeCurrent glad_eglMakeCurrent
+GLAD_API_CALL PFNEGLQUERYAPIPROC glad_eglQueryAPI;
+#define eglQueryAPI glad_eglQueryAPI
+GLAD_API_CALL PFNEGLQUERYCONTEXTPROC glad_eglQueryContext;
+#define eglQueryContext glad_eglQueryContext
+GLAD_API_CALL PFNEGLQUERYSTRINGPROC glad_eglQueryString;
+#define eglQueryString glad_eglQueryString
+GLAD_API_CALL PFNEGLQUERYSURFACEPROC glad_eglQuerySurface;
+#define eglQuerySurface glad_eglQuerySurface
+GLAD_API_CALL PFNEGLRELEASETEXIMAGEPROC glad_eglReleaseTexImage;
+#define eglReleaseTexImage glad_eglReleaseTexImage
+GLAD_API_CALL PFNEGLRELEASETHREADPROC glad_eglReleaseThread;
+#define eglReleaseThread glad_eglReleaseThread
+GLAD_API_CALL PFNEGLSURFACEATTRIBPROC glad_eglSurfaceAttrib;
+#define eglSurfaceAttrib glad_eglSurfaceAttrib
+GLAD_API_CALL PFNEGLSWAPBUFFERSPROC glad_eglSwapBuffers;
+#define eglSwapBuffers glad_eglSwapBuffers
+GLAD_API_CALL PFNEGLSWAPINTERVALPROC glad_eglSwapInterval;
+#define eglSwapInterval glad_eglSwapInterval
+GLAD_API_CALL PFNEGLTERMINATEPROC glad_eglTerminate;
+#define eglTerminate glad_eglTerminate
+GLAD_API_CALL PFNEGLWAITCLIENTPROC glad_eglWaitClient;
+#define eglWaitClient glad_eglWaitClient
+GLAD_API_CALL PFNEGLWAITGLPROC glad_eglWaitGL;
+#define eglWaitGL glad_eglWaitGL
+GLAD_API_CALL PFNEGLWAITNATIVEPROC glad_eglWaitNative;
+#define eglWaitNative glad_eglWaitNative
+GLAD_API_CALL PFNEGLWAITSYNCPROC glad_eglWaitSync;
+#define eglWaitSync glad_eglWaitSync
+
+
+
+
+
+GLAD_API_CALL int gladLoadEGLUserPtr(EGLDisplay display, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadEGL(EGLDisplay display, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/gles2.h glmark2-2021.02/src/glad/include/glad/gles2.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/gles2.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/glad/gles2.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1015 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:14 2019
+ *
+ * Generator: C/C++
+ * Specification: gl
+ * Extensions: 2
+ *
+ * APIs:
+ *  - gles2=2.0
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='gles2=2.0' --extensions='GL_OES_mapbuffer,GL_OES_required_internalformat' c
+ *
+ * Online:
+ *    http://glad.sh/#api=gles2%3D2.0&extensions=GL_OES_mapbuffer%2CGL_OES_required_internalformat&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GLES2_H_
+#define GLAD_GLES2_H_
+
+#ifdef __gl2_h_
+    #error OpenGL ES 2 header already included (API: gles2), remove previous include!
+#endif
+#define __gl2_h_ 1
+
+#ifdef __gl3_h_
+    #error OpenGL ES 3 header already included (API: gles2), remove previous include!
+#endif
+#define __gl3_h_ 1
+
+
+#define GLAD_GLES2
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GL_ACTIVE_ATTRIBUTES 0x8B89
+#define GL_ACTIVE_ATTRIBUTE_MAX_LENGTH 0x8B8A
+#define GL_ACTIVE_TEXTURE 0x84E0
+#define GL_ACTIVE_UNIFORMS 0x8B86
+#define GL_ACTIVE_UNIFORM_MAX_LENGTH 0x8B87
+#define GL_ALIASED_LINE_WIDTH_RANGE 0x846E
+#define GL_ALIASED_POINT_SIZE_RANGE 0x846D
+#define GL_ALPHA 0x1906
+#define GL_ALPHA8_OES 0x803C
+#define GL_ALPHA_BITS 0x0D55
+#define GL_ALWAYS 0x0207
+#define GL_ARRAY_BUFFER 0x8892
+#define GL_ARRAY_BUFFER_BINDING 0x8894
+#define GL_ATTACHED_SHADERS 0x8B85
+#define GL_BACK 0x0405
+#define GL_BLEND 0x0BE2
+#define GL_BLEND_COLOR 0x8005
+#define GL_BLEND_DST_ALPHA 0x80CA
+#define GL_BLEND_DST_RGB 0x80C8
+#define GL_BLEND_EQUATION 0x8009
+#define GL_BLEND_EQUATION_ALPHA 0x883D
+#define GL_BLEND_EQUATION_RGB 0x8009
+#define GL_BLEND_SRC_ALPHA 0x80CB
+#define GL_BLEND_SRC_RGB 0x80C9
+#define GL_BLUE_BITS 0x0D54
+#define GL_BOOL 0x8B56
+#define GL_BOOL_VEC2 0x8B57
+#define GL_BOOL_VEC3 0x8B58
+#define GL_BOOL_VEC4 0x8B59
+#define GL_BUFFER_ACCESS_OES 0x88BB
+#define GL_BUFFER_MAPPED_OES 0x88BC
+#define GL_BUFFER_MAP_POINTER_OES 0x88BD
+#define GL_BUFFER_SIZE 0x8764
+#define GL_BUFFER_USAGE 0x8765
+#define GL_BYTE 0x1400
+#define GL_CCW 0x0901
+#define GL_CLAMP_TO_EDGE 0x812F
+#define GL_COLOR_ATTACHMENT0 0x8CE0
+#define GL_COLOR_BUFFER_BIT 0x00004000
+#define GL_COLOR_CLEAR_VALUE 0x0C22
+#define GL_COLOR_WRITEMASK 0x0C23
+#define GL_COMPILE_STATUS 0x8B81
+#define GL_COMPRESSED_TEXTURE_FORMATS 0x86A3
+#define GL_CONSTANT_ALPHA 0x8003
+#define GL_CONSTANT_COLOR 0x8001
+#define GL_CULL_FACE 0x0B44
+#define GL_CULL_FACE_MODE 0x0B45
+#define GL_CURRENT_PROGRAM 0x8B8D
+#define GL_CURRENT_VERTEX_ATTRIB 0x8626
+#define GL_CW 0x0900
+#define GL_DECR 0x1E03
+#define GL_DECR_WRAP 0x8508
+#define GL_DELETE_STATUS 0x8B80
+#define GL_DEPTH24_STENCIL8_OES 0x88F0
+#define GL_DEPTH_ATTACHMENT 0x8D00
+#define GL_DEPTH_BITS 0x0D56
+#define GL_DEPTH_BUFFER_BIT 0x00000100
+#define GL_DEPTH_CLEAR_VALUE 0x0B73
+#define GL_DEPTH_COMPONENT 0x1902
+#define GL_DEPTH_COMPONENT16 0x81A5
+#define GL_DEPTH_COMPONENT16_OES 0x81A5
+#define GL_DEPTH_COMPONENT24_OES 0x81A6
+#define GL_DEPTH_COMPONENT32_OES 0x81A7
+#define GL_DEPTH_FUNC 0x0B74
+#define GL_DEPTH_RANGE 0x0B70
+#define GL_DEPTH_TEST 0x0B71
+#define GL_DEPTH_WRITEMASK 0x0B72
+#define GL_DITHER 0x0BD0
+#define GL_DONT_CARE 0x1100
+#define GL_DST_ALPHA 0x0304
+#define GL_DST_COLOR 0x0306
+#define GL_DYNAMIC_DRAW 0x88E8
+#define GL_ELEMENT_ARRAY_BUFFER 0x8893
+#define GL_ELEMENT_ARRAY_BUFFER_BINDING 0x8895
+#define GL_EQUAL 0x0202
+#define GL_EXTENSIONS 0x1F03
+#define GL_FALSE 0
+#define GL_FASTEST 0x1101
+#define GL_FIXED 0x140C
+#define GL_FLOAT 0x1406
+#define GL_FLOAT_MAT2 0x8B5A
+#define GL_FLOAT_MAT3 0x8B5B
+#define GL_FLOAT_MAT4 0x8B5C
+#define GL_FLOAT_VEC2 0x8B50
+#define GL_FLOAT_VEC3 0x8B51
+#define GL_FLOAT_VEC4 0x8B52
+#define GL_FRAGMENT_SHADER 0x8B30
+#define GL_FRAMEBUFFER 0x8D40
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME 0x8CD1
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE 0x8CD0
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE 0x8CD3
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL 0x8CD2
+#define GL_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_FRAMEBUFFER_COMPLETE 0x8CD5
+#define GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT 0x8CD6
+#define GL_FRAMEBUFFER_INCOMPLETE_DIMENSIONS 0x8CD9
+#define GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT 0x8CD7
+#define GL_FRAMEBUFFER_UNSUPPORTED 0x8CDD
+#define GL_FRONT 0x0404
+#define GL_FRONT_AND_BACK 0x0408
+#define GL_FRONT_FACE 0x0B46
+#define GL_FUNC_ADD 0x8006
+#define GL_FUNC_REVERSE_SUBTRACT 0x800B
+#define GL_FUNC_SUBTRACT 0x800A
+#define GL_GENERATE_MIPMAP_HINT 0x8192
+#define GL_GEQUAL 0x0206
+#define GL_GREATER 0x0204
+#define GL_GREEN_BITS 0x0D53
+#define GL_HIGH_FLOAT 0x8DF2
+#define GL_HIGH_INT 0x8DF5
+#define GL_IMPLEMENTATION_COLOR_READ_FORMAT 0x8B9B
+#define GL_IMPLEMENTATION_COLOR_READ_TYPE 0x8B9A
+#define GL_INCR 0x1E02
+#define GL_INCR_WRAP 0x8507
+#define GL_INFO_LOG_LENGTH 0x8B84
+#define GL_INT 0x1404
+#define GL_INT_VEC2 0x8B53
+#define GL_INT_VEC3 0x8B54
+#define GL_INT_VEC4 0x8B55
+#define GL_INVALID_ENUM 0x0500
+#define GL_INVALID_FRAMEBUFFER_OPERATION 0x0506
+#define GL_INVALID_OPERATION 0x0502
+#define GL_INVALID_VALUE 0x0501
+#define GL_INVERT 0x150A
+#define GL_KEEP 0x1E00
+#define GL_LEQUAL 0x0203
+#define GL_LESS 0x0201
+#define GL_LINEAR 0x2601
+#define GL_LINEAR_MIPMAP_LINEAR 0x2703
+#define GL_LINEAR_MIPMAP_NEAREST 0x2701
+#define GL_LINES 0x0001
+#define GL_LINE_LOOP 0x0002
+#define GL_LINE_STRIP 0x0003
+#define GL_LINE_WIDTH 0x0B21
+#define GL_LINK_STATUS 0x8B82
+#define GL_LOW_FLOAT 0x8DF0
+#define GL_LOW_INT 0x8DF3
+#define GL_LUMINANCE 0x1909
+#define GL_LUMINANCE4_ALPHA4_OES 0x8043
+#define GL_LUMINANCE8_ALPHA8_OES 0x8045
+#define GL_LUMINANCE8_OES 0x8040
+#define GL_LUMINANCE_ALPHA 0x190A
+#define GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS 0x8B4D
+#define GL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C
+#define GL_MAX_FRAGMENT_UNIFORM_VECTORS 0x8DFD
+#define GL_MAX_RENDERBUFFER_SIZE 0x84E8
+#define GL_MAX_TEXTURE_IMAGE_UNITS 0x8872
+#define GL_MAX_TEXTURE_SIZE 0x0D33
+#define GL_MAX_VARYING_VECTORS 0x8DFC
+#define GL_MAX_VERTEX_ATTRIBS 0x8869
+#define GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS 0x8B4C
+#define GL_MAX_VERTEX_UNIFORM_VECTORS 0x8DFB
+#define GL_MAX_VIEWPORT_DIMS 0x0D3A
+#define GL_MEDIUM_FLOAT 0x8DF1
+#define GL_MEDIUM_INT 0x8DF4
+#define GL_MIRRORED_REPEAT 0x8370
+#define GL_NEAREST 0x2600
+#define GL_NEAREST_MIPMAP_LINEAR 0x2702
+#define GL_NEAREST_MIPMAP_NEAREST 0x2700
+#define GL_NEVER 0x0200
+#define GL_NICEST 0x1102
+#define GL_NONE 0
+#define GL_NOTEQUAL 0x0205
+#define GL_NO_ERROR 0
+#define GL_NUM_COMPRESSED_TEXTURE_FORMATS 0x86A2
+#define GL_NUM_SHADER_BINARY_FORMATS 0x8DF9
+#define GL_ONE 1
+#define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004
+#define GL_ONE_MINUS_CONSTANT_COLOR 0x8002
+#define GL_ONE_MINUS_DST_ALPHA 0x0305
+#define GL_ONE_MINUS_DST_COLOR 0x0307
+#define GL_ONE_MINUS_SRC_ALPHA 0x0303
+#define GL_ONE_MINUS_SRC_COLOR 0x0301
+#define GL_OUT_OF_MEMORY 0x0505
+#define GL_PACK_ALIGNMENT 0x0D05
+#define GL_POINTS 0x0000
+#define GL_POLYGON_OFFSET_FACTOR 0x8038
+#define GL_POLYGON_OFFSET_FILL 0x8037
+#define GL_POLYGON_OFFSET_UNITS 0x2A00
+#define GL_RED_BITS 0x0D52
+#define GL_RENDERBUFFER 0x8D41
+#define GL_RENDERBUFFER_ALPHA_SIZE 0x8D53
+#define GL_RENDERBUFFER_BINDING 0x8CA7
+#define GL_RENDERBUFFER_BLUE_SIZE 0x8D52
+#define GL_RENDERBUFFER_DEPTH_SIZE 0x8D54
+#define GL_RENDERBUFFER_GREEN_SIZE 0x8D51
+#define GL_RENDERBUFFER_HEIGHT 0x8D43
+#define GL_RENDERBUFFER_INTERNAL_FORMAT 0x8D44
+#define GL_RENDERBUFFER_RED_SIZE 0x8D50
+#define GL_RENDERBUFFER_STENCIL_SIZE 0x8D55
+#define GL_RENDERBUFFER_WIDTH 0x8D42
+#define GL_RENDERER 0x1F01
+#define GL_REPEAT 0x2901
+#define GL_REPLACE 0x1E01
+#define GL_RGB 0x1907
+#define GL_RGB10_A2_EXT 0x8059
+#define GL_RGB10_EXT 0x8052
+#define GL_RGB565 0x8D62
+#define GL_RGB565_OES 0x8D62
+#define GL_RGB5_A1 0x8057
+#define GL_RGB5_A1_OES 0x8057
+#define GL_RGB8_OES 0x8051
+#define GL_RGBA 0x1908
+#define GL_RGBA4 0x8056
+#define GL_RGBA4_OES 0x8056
+#define GL_RGBA8_OES 0x8058
+#define GL_SAMPLER_2D 0x8B5E
+#define GL_SAMPLER_CUBE 0x8B60
+#define GL_SAMPLES 0x80A9
+#define GL_SAMPLE_ALPHA_TO_COVERAGE 0x809E
+#define GL_SAMPLE_BUFFERS 0x80A8
+#define GL_SAMPLE_COVERAGE 0x80A0
+#define GL_SAMPLE_COVERAGE_INVERT 0x80AB
+#define GL_SAMPLE_COVERAGE_VALUE 0x80AA
+#define GL_SCISSOR_BOX 0x0C10
+#define GL_SCISSOR_TEST 0x0C11
+#define GL_SHADER_BINARY_FORMATS 0x8DF8
+#define GL_SHADER_COMPILER 0x8DFA
+#define GL_SHADER_SOURCE_LENGTH 0x8B88
+#define GL_SHADER_TYPE 0x8B4F
+#define GL_SHADING_LANGUAGE_VERSION 0x8B8C
+#define GL_SHORT 0x1402
+#define GL_SRC_ALPHA 0x0302
+#define GL_SRC_ALPHA_SATURATE 0x0308
+#define GL_SRC_COLOR 0x0300
+#define GL_STATIC_DRAW 0x88E4
+#define GL_STENCIL_ATTACHMENT 0x8D20
+#define GL_STENCIL_BACK_FAIL 0x8801
+#define GL_STENCIL_BACK_FUNC 0x8800
+#define GL_STENCIL_BACK_PASS_DEPTH_FAIL 0x8802
+#define GL_STENCIL_BACK_PASS_DEPTH_PASS 0x8803
+#define GL_STENCIL_BACK_REF 0x8CA3
+#define GL_STENCIL_BACK_VALUE_MASK 0x8CA4
+#define GL_STENCIL_BACK_WRITEMASK 0x8CA5
+#define GL_STENCIL_BITS 0x0D57
+#define GL_STENCIL_BUFFER_BIT 0x00000400
+#define GL_STENCIL_CLEAR_VALUE 0x0B91
+#define GL_STENCIL_FAIL 0x0B94
+#define GL_STENCIL_FUNC 0x0B92
+#define GL_STENCIL_INDEX8 0x8D48
+#define GL_STENCIL_PASS_DEPTH_FAIL 0x0B95
+#define GL_STENCIL_PASS_DEPTH_PASS 0x0B96
+#define GL_STENCIL_REF 0x0B97
+#define GL_STENCIL_TEST 0x0B90
+#define GL_STENCIL_VALUE_MASK 0x0B93
+#define GL_STENCIL_WRITEMASK 0x0B98
+#define GL_STREAM_DRAW 0x88E0
+#define GL_SUBPIXEL_BITS 0x0D50
+#define GL_TEXTURE 0x1702
+#define GL_TEXTURE0 0x84C0
+#define GL_TEXTURE1 0x84C1
+#define GL_TEXTURE10 0x84CA
+#define GL_TEXTURE11 0x84CB
+#define GL_TEXTURE12 0x84CC
+#define GL_TEXTURE13 0x84CD
+#define GL_TEXTURE14 0x84CE
+#define GL_TEXTURE15 0x84CF
+#define GL_TEXTURE16 0x84D0
+#define GL_TEXTURE17 0x84D1
+#define GL_TEXTURE18 0x84D2
+#define GL_TEXTURE19 0x84D3
+#define GL_TEXTURE2 0x84C2
+#define GL_TEXTURE20 0x84D4
+#define GL_TEXTURE21 0x84D5
+#define GL_TEXTURE22 0x84D6
+#define GL_TEXTURE23 0x84D7
+#define GL_TEXTURE24 0x84D8
+#define GL_TEXTURE25 0x84D9
+#define GL_TEXTURE26 0x84DA
+#define GL_TEXTURE27 0x84DB
+#define GL_TEXTURE28 0x84DC
+#define GL_TEXTURE29 0x84DD
+#define GL_TEXTURE3 0x84C3
+#define GL_TEXTURE30 0x84DE
+#define GL_TEXTURE31 0x84DF
+#define GL_TEXTURE4 0x84C4
+#define GL_TEXTURE5 0x84C5
+#define GL_TEXTURE6 0x84C6
+#define GL_TEXTURE7 0x84C7
+#define GL_TEXTURE8 0x84C8
+#define GL_TEXTURE9 0x84C9
+#define GL_TEXTURE_2D 0x0DE1
+#define GL_TEXTURE_BINDING_2D 0x8069
+#define GL_TEXTURE_BINDING_CUBE_MAP 0x8514
+#define GL_TEXTURE_CUBE_MAP 0x8513
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519
+#define GL_TEXTURE_MAG_FILTER 0x2800
+#define GL_TEXTURE_MIN_FILTER 0x2801
+#define GL_TEXTURE_WRAP_S 0x2802
+#define GL_TEXTURE_WRAP_T 0x2803
+#define GL_TRIANGLES 0x0004
+#define GL_TRIANGLE_FAN 0x0006
+#define GL_TRIANGLE_STRIP 0x0005
+#define GL_TRUE 1
+#define GL_UNPACK_ALIGNMENT 0x0CF5
+#define GL_UNSIGNED_BYTE 0x1401
+#define GL_UNSIGNED_INT 0x1405
+#define GL_UNSIGNED_SHORT 0x1403
+#define GL_UNSIGNED_SHORT_4_4_4_4 0x8033
+#define GL_UNSIGNED_SHORT_5_5_5_1 0x8034
+#define GL_UNSIGNED_SHORT_5_6_5 0x8363
+#define GL_VALIDATE_STATUS 0x8B83
+#define GL_VENDOR 0x1F00
+#define GL_VERSION 0x1F02
+#define GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING 0x889F
+#define GL_VERTEX_ATTRIB_ARRAY_ENABLED 0x8622
+#define GL_VERTEX_ATTRIB_ARRAY_NORMALIZED 0x886A
+#define GL_VERTEX_ATTRIB_ARRAY_POINTER 0x8645
+#define GL_VERTEX_ATTRIB_ARRAY_SIZE 0x8623
+#define GL_VERTEX_ATTRIB_ARRAY_STRIDE 0x8624
+#define GL_VERTEX_ATTRIB_ARRAY_TYPE 0x8625
+#define GL_VERTEX_SHADER 0x8B31
+#define GL_VIEWPORT 0x0BA2
+#define GL_WRITE_ONLY_OES 0x88B9
+#define GL_ZERO 0
+
+
+#include <KHR/khrplatform.h>
+typedef unsigned int GLenum;
+typedef unsigned char GLboolean;
+typedef unsigned int GLbitfield;
+typedef void GLvoid;
+typedef khronos_int8_t GLbyte;
+typedef khronos_uint8_t GLubyte;
+typedef khronos_int16_t GLshort;
+typedef khronos_uint16_t GLushort;
+typedef int GLint;
+typedef unsigned int GLuint;
+typedef khronos_int32_t GLclampx;
+typedef int GLsizei;
+typedef khronos_float_t GLfloat;
+typedef khronos_float_t GLclampf;
+typedef double GLdouble;
+typedef double GLclampd;
+typedef void *GLeglClientBufferEXT;
+typedef void *GLeglImageOES;
+typedef char GLchar;
+typedef char GLcharARB;
+#ifdef __APPLE__
+typedef void *GLhandleARB;
+#else
+typedef unsigned int GLhandleARB;
+#endif
+typedef khronos_uint16_t GLhalf;
+typedef khronos_uint16_t GLhalfARB;
+typedef khronos_int32_t GLfixed;
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptr;
+#else
+typedef khronos_intptr_t GLintptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptrARB;
+#else
+typedef khronos_intptr_t GLintptrARB;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptr;
+#else
+typedef khronos_ssize_t GLsizeiptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptrARB;
+#else
+typedef khronos_ssize_t GLsizeiptrARB;
+#endif
+typedef khronos_int64_t GLint64;
+typedef khronos_int64_t GLint64EXT;
+typedef khronos_uint64_t GLuint64;
+typedef khronos_uint64_t GLuint64EXT;
+typedef struct __GLsync *GLsync;
+struct _cl_context;
+struct _cl_event;
+typedef void (GLAD_API_PTR *GLDEBUGPROC)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCARB)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCKHR)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCAMD)(GLuint id,GLenum category,GLenum severity,GLsizei length,const GLchar *message,void *userParam);
+typedef unsigned short GLhalfNV;
+typedef GLintptr GLvdpauSurfaceNV;
+typedef void (GLAD_API_PTR *GLVULKANPROCNV)(void);
+
+
+#define GL_ES_VERSION_2_0 1
+GLAD_API_CALL int GLAD_GL_ES_VERSION_2_0;
+#define GL_OES_mapbuffer 1
+GLAD_API_CALL int GLAD_GL_OES_mapbuffer;
+#define GL_OES_required_internalformat 1
+GLAD_API_CALL int GLAD_GL_OES_required_internalformat;
+
+
+typedef void (GLAD_API_PTR *PFNGLACTIVETEXTUREPROC)(GLenum texture);
+typedef void (GLAD_API_PTR *PFNGLATTACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLBINDATTRIBLOCATIONPROC)(GLuint program, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERPROC)(GLenum target, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAMEBUFFERPROC)(GLenum target, GLuint framebuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDRENDERBUFFERPROC)(GLenum target, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDTEXTUREPROC)(GLenum target, GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLBLENDCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONSEPARATEPROC)(GLenum modeRGB, GLenum modeAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCPROC)(GLenum sfactor, GLenum dfactor);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCSEPARATEPROC)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+typedef void (GLAD_API_PTR *PFNGLBUFFERDATAPROC)(GLenum target, GLsizeiptr size, const void * data, GLenum usage);
+typedef void (GLAD_API_PTR *PFNGLBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, const void * data);
+typedef GLenum (GLAD_API_PTR *PFNGLCHECKFRAMEBUFFERSTATUSPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLCLEARPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLCLEARCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARDEPTHFPROC)(GLfloat d);
+typedef void (GLAD_API_PTR *PFNGLCLEARSTENCILPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKPROC)(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
+typedef void (GLAD_API_PTR *PFNGLCOMPILESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLsizei height, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATEPROGRAMPROC)(void);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATESHADERPROC)(GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCULLFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLDELETEBUFFERSPROC)(GLsizei n, const GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEFRAMEBUFFERSPROC)(GLsizei n, const GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLDELETERENDERBUFFERSPROC)(GLsizei n, const GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDELETETEXTURESPROC)(GLsizei n, const GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLDEPTHFUNCPROC)(GLenum func);
+typedef void (GLAD_API_PTR *PFNGLDEPTHMASKPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLDEPTHRANGEFPROC)(GLfloat n, GLfloat f);
+typedef void (GLAD_API_PTR *PFNGLDETACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDISABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLDISABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSPROC)(GLenum mode, GLint first, GLsizei count);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLENABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLENABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLFINISHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERRENDERBUFFERPROC)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE2DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRONTFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLGENBUFFERSPROC)(GLsizei n, GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLGENFRAMEBUFFERSPROC)(GLsizei n, GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLGENRENDERBUFFERSPROC)(GLsizei n, GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLGENTEXTURESPROC)(GLsizei n, GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLGENERATEMIPMAPPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEATTRIBPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETATTACHEDSHADERSPROC)(GLuint program, GLsizei maxCount, GLsizei * count, GLuint * shaders);
+typedef GLint (GLAD_API_PTR *PFNGLGETATTRIBLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANVPROC)(GLenum pname, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVPROC)(GLenum target, GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVOESPROC)(GLenum target, GLenum pname, void ** params);
+typedef GLenum (GLAD_API_PTR *PFNGLGETERRORPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLGETFLOATVPROC)(GLenum pname, GLfloat * data);
+typedef void (GLAD_API_PTR *PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC)(GLenum target, GLenum attachment, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERVPROC)(GLenum pname, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMINFOLOGPROC)(GLuint program, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMIVPROC)(GLuint program, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETRENDERBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERINFOLOGPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERPRECISIONFORMATPROC)(GLenum shadertype, GLenum precisiontype, GLint * range, GLint * precision);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERSOURCEPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * source);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERIVPROC)(GLuint shader, GLenum pname, GLint * params);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGPROC)(GLenum name);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef GLint (GLAD_API_PTR *PFNGLGETUNIFORMLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMFVPROC)(GLuint program, GLint location, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMIVPROC)(GLuint program, GLint location, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBPOINTERVPROC)(GLuint index, GLenum pname, void ** pointer);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBFVPROC)(GLuint index, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLHINTPROC)(GLenum target, GLenum mode);
+typedef GLboolean (GLAD_API_PTR *PFNGLISBUFFERPROC)(GLuint buffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDPROC)(GLenum cap);
+typedef GLboolean (GLAD_API_PTR *PFNGLISFRAMEBUFFERPROC)(GLuint framebuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISPROGRAMPROC)(GLuint program);
+typedef GLboolean (GLAD_API_PTR *PFNGLISRENDERBUFFERPROC)(GLuint renderbuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSHADERPROC)(GLuint shader);
+typedef GLboolean (GLAD_API_PTR *PFNGLISTEXTUREPROC)(GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLLINEWIDTHPROC)(GLfloat width);
+typedef void (GLAD_API_PTR *PFNGLLINKPROGRAMPROC)(GLuint program);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERPROC)(GLenum target, GLenum access);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFEROESPROC)(GLenum target, GLenum access);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONOFFSETPROC)(GLfloat factor, GLfloat units);
+typedef void (GLAD_API_PTR *PFNGLREADPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLRELEASESHADERCOMPILERPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEPROC)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSAMPLECOVERAGEPROC)(GLfloat value, GLboolean invert);
+typedef void (GLAD_API_PTR *PFNGLSCISSORPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSHADERBINARYPROC)(GLsizei count, const GLuint * shaders, GLenum binaryformat, const void * binary, GLsizei length);
+typedef void (GLAD_API_PTR *PFNGLSHADERSOURCEPROC)(GLuint shader, GLsizei count, const GLchar *const* string, const GLint * length);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCPROC)(GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCSEPARATEPROC)(GLenum face, GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKSEPARATEPROC)(GLenum face, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPPROC)(GLenum fail, GLenum zfail, GLenum zpass);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPSEPARATEPROC)(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FPROC)(GLint location, GLfloat v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IPROC)(GLint location, GLint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FPROC)(GLint location, GLfloat v0, GLfloat v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IPROC)(GLint location, GLint v0, GLint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IPROC)(GLint location, GLint v0, GLint v1, GLint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2, GLfloat v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IPROC)(GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFERPROC)(GLenum target);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFEROESPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLUSEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVALIDATEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FPROC)(GLuint index, GLfloat x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FPROC)(GLuint index, GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBPOINTERPROC)(GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVIEWPORTPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+
+GLAD_API_CALL PFNGLACTIVETEXTUREPROC glad_glActiveTexture;
+#define glActiveTexture glad_glActiveTexture
+GLAD_API_CALL PFNGLATTACHSHADERPROC glad_glAttachShader;
+#define glAttachShader glad_glAttachShader
+GLAD_API_CALL PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation;
+#define glBindAttribLocation glad_glBindAttribLocation
+GLAD_API_CALL PFNGLBINDBUFFERPROC glad_glBindBuffer;
+#define glBindBuffer glad_glBindBuffer
+GLAD_API_CALL PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer;
+#define glBindFramebuffer glad_glBindFramebuffer
+GLAD_API_CALL PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer;
+#define glBindRenderbuffer glad_glBindRenderbuffer
+GLAD_API_CALL PFNGLBINDTEXTUREPROC glad_glBindTexture;
+#define glBindTexture glad_glBindTexture
+GLAD_API_CALL PFNGLBLENDCOLORPROC glad_glBlendColor;
+#define glBlendColor glad_glBlendColor
+GLAD_API_CALL PFNGLBLENDEQUATIONPROC glad_glBlendEquation;
+#define glBlendEquation glad_glBlendEquation
+GLAD_API_CALL PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate;
+#define glBlendEquationSeparate glad_glBlendEquationSeparate
+GLAD_API_CALL PFNGLBLENDFUNCPROC glad_glBlendFunc;
+#define glBlendFunc glad_glBlendFunc
+GLAD_API_CALL PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate;
+#define glBlendFuncSeparate glad_glBlendFuncSeparate
+GLAD_API_CALL PFNGLBUFFERDATAPROC glad_glBufferData;
+#define glBufferData glad_glBufferData
+GLAD_API_CALL PFNGLBUFFERSUBDATAPROC glad_glBufferSubData;
+#define glBufferSubData glad_glBufferSubData
+GLAD_API_CALL PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus;
+#define glCheckFramebufferStatus glad_glCheckFramebufferStatus
+GLAD_API_CALL PFNGLCLEARPROC glad_glClear;
+#define glClear glad_glClear
+GLAD_API_CALL PFNGLCLEARCOLORPROC glad_glClearColor;
+#define glClearColor glad_glClearColor
+GLAD_API_CALL PFNGLCLEARDEPTHFPROC glad_glClearDepthf;
+#define glClearDepthf glad_glClearDepthf
+GLAD_API_CALL PFNGLCLEARSTENCILPROC glad_glClearStencil;
+#define glClearStencil glad_glClearStencil
+GLAD_API_CALL PFNGLCOLORMASKPROC glad_glColorMask;
+#define glColorMask glad_glColorMask
+GLAD_API_CALL PFNGLCOMPILESHADERPROC glad_glCompileShader;
+#define glCompileShader glad_glCompileShader
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D;
+#define glCompressedTexImage2D glad_glCompressedTexImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D;
+#define glCompressedTexSubImage2D glad_glCompressedTexSubImage2D
+GLAD_API_CALL PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D;
+#define glCopyTexImage2D glad_glCopyTexImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D;
+#define glCopyTexSubImage2D glad_glCopyTexSubImage2D
+GLAD_API_CALL PFNGLCREATEPROGRAMPROC glad_glCreateProgram;
+#define glCreateProgram glad_glCreateProgram
+GLAD_API_CALL PFNGLCREATESHADERPROC glad_glCreateShader;
+#define glCreateShader glad_glCreateShader
+GLAD_API_CALL PFNGLCULLFACEPROC glad_glCullFace;
+#define glCullFace glad_glCullFace
+GLAD_API_CALL PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers;
+#define glDeleteBuffers glad_glDeleteBuffers
+GLAD_API_CALL PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers;
+#define glDeleteFramebuffers glad_glDeleteFramebuffers
+GLAD_API_CALL PFNGLDELETEPROGRAMPROC glad_glDeleteProgram;
+#define glDeleteProgram glad_glDeleteProgram
+GLAD_API_CALL PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers;
+#define glDeleteRenderbuffers glad_glDeleteRenderbuffers
+GLAD_API_CALL PFNGLDELETESHADERPROC glad_glDeleteShader;
+#define glDeleteShader glad_glDeleteShader
+GLAD_API_CALL PFNGLDELETETEXTURESPROC glad_glDeleteTextures;
+#define glDeleteTextures glad_glDeleteTextures
+GLAD_API_CALL PFNGLDEPTHFUNCPROC glad_glDepthFunc;
+#define glDepthFunc glad_glDepthFunc
+GLAD_API_CALL PFNGLDEPTHMASKPROC glad_glDepthMask;
+#define glDepthMask glad_glDepthMask
+GLAD_API_CALL PFNGLDEPTHRANGEFPROC glad_glDepthRangef;
+#define glDepthRangef glad_glDepthRangef
+GLAD_API_CALL PFNGLDETACHSHADERPROC glad_glDetachShader;
+#define glDetachShader glad_glDetachShader
+GLAD_API_CALL PFNGLDISABLEPROC glad_glDisable;
+#define glDisable glad_glDisable
+GLAD_API_CALL PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray;
+#define glDisableVertexAttribArray glad_glDisableVertexAttribArray
+GLAD_API_CALL PFNGLDRAWARRAYSPROC glad_glDrawArrays;
+#define glDrawArrays glad_glDrawArrays
+GLAD_API_CALL PFNGLDRAWELEMENTSPROC glad_glDrawElements;
+#define glDrawElements glad_glDrawElements
+GLAD_API_CALL PFNGLENABLEPROC glad_glEnable;
+#define glEnable glad_glEnable
+GLAD_API_CALL PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray;
+#define glEnableVertexAttribArray glad_glEnableVertexAttribArray
+GLAD_API_CALL PFNGLFINISHPROC glad_glFinish;
+#define glFinish glad_glFinish
+GLAD_API_CALL PFNGLFLUSHPROC glad_glFlush;
+#define glFlush glad_glFlush
+GLAD_API_CALL PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer;
+#define glFramebufferRenderbuffer glad_glFramebufferRenderbuffer
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D;
+#define glFramebufferTexture2D glad_glFramebufferTexture2D
+GLAD_API_CALL PFNGLFRONTFACEPROC glad_glFrontFace;
+#define glFrontFace glad_glFrontFace
+GLAD_API_CALL PFNGLGENBUFFERSPROC glad_glGenBuffers;
+#define glGenBuffers glad_glGenBuffers
+GLAD_API_CALL PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers;
+#define glGenFramebuffers glad_glGenFramebuffers
+GLAD_API_CALL PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers;
+#define glGenRenderbuffers glad_glGenRenderbuffers
+GLAD_API_CALL PFNGLGENTEXTURESPROC glad_glGenTextures;
+#define glGenTextures glad_glGenTextures
+GLAD_API_CALL PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap;
+#define glGenerateMipmap glad_glGenerateMipmap
+GLAD_API_CALL PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib;
+#define glGetActiveAttrib glad_glGetActiveAttrib
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform;
+#define glGetActiveUniform glad_glGetActiveUniform
+GLAD_API_CALL PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders;
+#define glGetAttachedShaders glad_glGetAttachedShaders
+GLAD_API_CALL PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation;
+#define glGetAttribLocation glad_glGetAttribLocation
+GLAD_API_CALL PFNGLGETBOOLEANVPROC glad_glGetBooleanv;
+#define glGetBooleanv glad_glGetBooleanv
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv;
+#define glGetBufferParameteriv glad_glGetBufferParameteriv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv;
+#define glGetBufferPointerv glad_glGetBufferPointerv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVOESPROC glad_glGetBufferPointervOES;
+#define glGetBufferPointervOES glad_glGetBufferPointervOES
+GLAD_API_CALL PFNGLGETERRORPROC glad_glGetError;
+#define glGetError glad_glGetError
+GLAD_API_CALL PFNGLGETFLOATVPROC glad_glGetFloatv;
+#define glGetFloatv glad_glGetFloatv
+GLAD_API_CALL PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv;
+#define glGetFramebufferAttachmentParameteriv glad_glGetFramebufferAttachmentParameteriv
+GLAD_API_CALL PFNGLGETINTEGERVPROC glad_glGetIntegerv;
+#define glGetIntegerv glad_glGetIntegerv
+GLAD_API_CALL PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog;
+#define glGetProgramInfoLog glad_glGetProgramInfoLog
+GLAD_API_CALL PFNGLGETPROGRAMIVPROC glad_glGetProgramiv;
+#define glGetProgramiv glad_glGetProgramiv
+GLAD_API_CALL PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv;
+#define glGetRenderbufferParameteriv glad_glGetRenderbufferParameteriv
+GLAD_API_CALL PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog;
+#define glGetShaderInfoLog glad_glGetShaderInfoLog
+GLAD_API_CALL PFNGLGETSHADERPRECISIONFORMATPROC glad_glGetShaderPrecisionFormat;
+#define glGetShaderPrecisionFormat glad_glGetShaderPrecisionFormat
+GLAD_API_CALL PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource;
+#define glGetShaderSource glad_glGetShaderSource
+GLAD_API_CALL PFNGLGETSHADERIVPROC glad_glGetShaderiv;
+#define glGetShaderiv glad_glGetShaderiv
+GLAD_API_CALL PFNGLGETSTRINGPROC glad_glGetString;
+#define glGetString glad_glGetString
+GLAD_API_CALL PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv;
+#define glGetTexParameterfv glad_glGetTexParameterfv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv;
+#define glGetTexParameteriv glad_glGetTexParameteriv
+GLAD_API_CALL PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation;
+#define glGetUniformLocation glad_glGetUniformLocation
+GLAD_API_CALL PFNGLGETUNIFORMFVPROC glad_glGetUniformfv;
+#define glGetUniformfv glad_glGetUniformfv
+GLAD_API_CALL PFNGLGETUNIFORMIVPROC glad_glGetUniformiv;
+#define glGetUniformiv glad_glGetUniformiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv;
+#define glGetVertexAttribPointerv glad_glGetVertexAttribPointerv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv;
+#define glGetVertexAttribfv glad_glGetVertexAttribfv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv;
+#define glGetVertexAttribiv glad_glGetVertexAttribiv
+GLAD_API_CALL PFNGLHINTPROC glad_glHint;
+#define glHint glad_glHint
+GLAD_API_CALL PFNGLISBUFFERPROC glad_glIsBuffer;
+#define glIsBuffer glad_glIsBuffer
+GLAD_API_CALL PFNGLISENABLEDPROC glad_glIsEnabled;
+#define glIsEnabled glad_glIsEnabled
+GLAD_API_CALL PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer;
+#define glIsFramebuffer glad_glIsFramebuffer
+GLAD_API_CALL PFNGLISPROGRAMPROC glad_glIsProgram;
+#define glIsProgram glad_glIsProgram
+GLAD_API_CALL PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer;
+#define glIsRenderbuffer glad_glIsRenderbuffer
+GLAD_API_CALL PFNGLISSHADERPROC glad_glIsShader;
+#define glIsShader glad_glIsShader
+GLAD_API_CALL PFNGLISTEXTUREPROC glad_glIsTexture;
+#define glIsTexture glad_glIsTexture
+GLAD_API_CALL PFNGLLINEWIDTHPROC glad_glLineWidth;
+#define glLineWidth glad_glLineWidth
+GLAD_API_CALL PFNGLLINKPROGRAMPROC glad_glLinkProgram;
+#define glLinkProgram glad_glLinkProgram
+GLAD_API_CALL PFNGLMAPBUFFERPROC glad_glMapBuffer;
+#define glMapBuffer glad_glMapBuffer
+GLAD_API_CALL PFNGLMAPBUFFEROESPROC glad_glMapBufferOES;
+#define glMapBufferOES glad_glMapBufferOES
+GLAD_API_CALL PFNGLPIXELSTOREIPROC glad_glPixelStorei;
+#define glPixelStorei glad_glPixelStorei
+GLAD_API_CALL PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset;
+#define glPolygonOffset glad_glPolygonOffset
+GLAD_API_CALL PFNGLREADPIXELSPROC glad_glReadPixels;
+#define glReadPixels glad_glReadPixels
+GLAD_API_CALL PFNGLRELEASESHADERCOMPILERPROC glad_glReleaseShaderCompiler;
+#define glReleaseShaderCompiler glad_glReleaseShaderCompiler
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage;
+#define glRenderbufferStorage glad_glRenderbufferStorage
+GLAD_API_CALL PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage;
+#define glSampleCoverage glad_glSampleCoverage
+GLAD_API_CALL PFNGLSCISSORPROC glad_glScissor;
+#define glScissor glad_glScissor
+GLAD_API_CALL PFNGLSHADERBINARYPROC glad_glShaderBinary;
+#define glShaderBinary glad_glShaderBinary
+GLAD_API_CALL PFNGLSHADERSOURCEPROC glad_glShaderSource;
+#define glShaderSource glad_glShaderSource
+GLAD_API_CALL PFNGLSTENCILFUNCPROC glad_glStencilFunc;
+#define glStencilFunc glad_glStencilFunc
+GLAD_API_CALL PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate;
+#define glStencilFuncSeparate glad_glStencilFuncSeparate
+GLAD_API_CALL PFNGLSTENCILMASKPROC glad_glStencilMask;
+#define glStencilMask glad_glStencilMask
+GLAD_API_CALL PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate;
+#define glStencilMaskSeparate glad_glStencilMaskSeparate
+GLAD_API_CALL PFNGLSTENCILOPPROC glad_glStencilOp;
+#define glStencilOp glad_glStencilOp
+GLAD_API_CALL PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate;
+#define glStencilOpSeparate glad_glStencilOpSeparate
+GLAD_API_CALL PFNGLTEXIMAGE2DPROC glad_glTexImage2D;
+#define glTexImage2D glad_glTexImage2D
+GLAD_API_CALL PFNGLTEXPARAMETERFPROC glad_glTexParameterf;
+#define glTexParameterf glad_glTexParameterf
+GLAD_API_CALL PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv;
+#define glTexParameterfv glad_glTexParameterfv
+GLAD_API_CALL PFNGLTEXPARAMETERIPROC glad_glTexParameteri;
+#define glTexParameteri glad_glTexParameteri
+GLAD_API_CALL PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv;
+#define glTexParameteriv glad_glTexParameteriv
+GLAD_API_CALL PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D;
+#define glTexSubImage2D glad_glTexSubImage2D
+GLAD_API_CALL PFNGLUNIFORM1FPROC glad_glUniform1f;
+#define glUniform1f glad_glUniform1f
+GLAD_API_CALL PFNGLUNIFORM1FVPROC glad_glUniform1fv;
+#define glUniform1fv glad_glUniform1fv
+GLAD_API_CALL PFNGLUNIFORM1IPROC glad_glUniform1i;
+#define glUniform1i glad_glUniform1i
+GLAD_API_CALL PFNGLUNIFORM1IVPROC glad_glUniform1iv;
+#define glUniform1iv glad_glUniform1iv
+GLAD_API_CALL PFNGLUNIFORM2FPROC glad_glUniform2f;
+#define glUniform2f glad_glUniform2f
+GLAD_API_CALL PFNGLUNIFORM2FVPROC glad_glUniform2fv;
+#define glUniform2fv glad_glUniform2fv
+GLAD_API_CALL PFNGLUNIFORM2IPROC glad_glUniform2i;
+#define glUniform2i glad_glUniform2i
+GLAD_API_CALL PFNGLUNIFORM2IVPROC glad_glUniform2iv;
+#define glUniform2iv glad_glUniform2iv
+GLAD_API_CALL PFNGLUNIFORM3FPROC glad_glUniform3f;
+#define glUniform3f glad_glUniform3f
+GLAD_API_CALL PFNGLUNIFORM3FVPROC glad_glUniform3fv;
+#define glUniform3fv glad_glUniform3fv
+GLAD_API_CALL PFNGLUNIFORM3IPROC glad_glUniform3i;
+#define glUniform3i glad_glUniform3i
+GLAD_API_CALL PFNGLUNIFORM3IVPROC glad_glUniform3iv;
+#define glUniform3iv glad_glUniform3iv
+GLAD_API_CALL PFNGLUNIFORM4FPROC glad_glUniform4f;
+#define glUniform4f glad_glUniform4f
+GLAD_API_CALL PFNGLUNIFORM4FVPROC glad_glUniform4fv;
+#define glUniform4fv glad_glUniform4fv
+GLAD_API_CALL PFNGLUNIFORM4IPROC glad_glUniform4i;
+#define glUniform4i glad_glUniform4i
+GLAD_API_CALL PFNGLUNIFORM4IVPROC glad_glUniform4iv;
+#define glUniform4iv glad_glUniform4iv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv;
+#define glUniformMatrix2fv glad_glUniformMatrix2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv;
+#define glUniformMatrix3fv glad_glUniformMatrix3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv;
+#define glUniformMatrix4fv glad_glUniformMatrix4fv
+GLAD_API_CALL PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer;
+#define glUnmapBuffer glad_glUnmapBuffer
+GLAD_API_CALL PFNGLUNMAPBUFFEROESPROC glad_glUnmapBufferOES;
+#define glUnmapBufferOES glad_glUnmapBufferOES
+GLAD_API_CALL PFNGLUSEPROGRAMPROC glad_glUseProgram;
+#define glUseProgram glad_glUseProgram
+GLAD_API_CALL PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram;
+#define glValidateProgram glad_glValidateProgram
+GLAD_API_CALL PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f;
+#define glVertexAttrib1f glad_glVertexAttrib1f
+GLAD_API_CALL PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv;
+#define glVertexAttrib1fv glad_glVertexAttrib1fv
+GLAD_API_CALL PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f;
+#define glVertexAttrib2f glad_glVertexAttrib2f
+GLAD_API_CALL PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv;
+#define glVertexAttrib2fv glad_glVertexAttrib2fv
+GLAD_API_CALL PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f;
+#define glVertexAttrib3f glad_glVertexAttrib3f
+GLAD_API_CALL PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv;
+#define glVertexAttrib3fv glad_glVertexAttrib3fv
+GLAD_API_CALL PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f;
+#define glVertexAttrib4f glad_glVertexAttrib4f
+GLAD_API_CALL PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv;
+#define glVertexAttrib4fv glad_glVertexAttrib4fv
+GLAD_API_CALL PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer;
+#define glVertexAttribPointer glad_glVertexAttribPointer
+GLAD_API_CALL PFNGLVIEWPORTPROC glad_glViewport;
+#define glViewport glad_glViewport
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLES2UserPtr( GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGLES2( GLADloadfunc load);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/gl.h glmark2-2021.02/src/glad/include/glad/gl.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/gl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/glad/gl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3686 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:14 2019
+ *
+ * Generator: C/C++
+ * Specification: gl
+ * Extensions: 0
+ *
+ * APIs:
+ *  - gl:compatibility=3.3
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='gl:compatibility=3.3' --extensions='' c
+ *
+ * Online:
+ *    http://glad.sh/#api=gl%3Acompatibility%3D3.3&extensions=&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GL_H_
+#define GLAD_GL_H_
+
+#ifdef __gl_h_
+    #error OpenGL header already included (API: gl), remove previous include!
+#endif
+#define __gl_h_ 1
+
+
+#define GLAD_GL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GL_2D 0x0600
+#define GL_2_BYTES 0x1407
+#define GL_3D 0x0601
+#define GL_3D_COLOR 0x0602
+#define GL_3D_COLOR_TEXTURE 0x0603
+#define GL_3_BYTES 0x1408
+#define GL_4D_COLOR_TEXTURE 0x0604
+#define GL_4_BYTES 0x1409
+#define GL_ACCUM 0x0100
+#define GL_ACCUM_ALPHA_BITS 0x0D5B
+#define GL_ACCUM_BLUE_BITS 0x0D5A
+#define GL_ACCUM_BUFFER_BIT 0x00000200
+#define GL_ACCUM_CLEAR_VALUE 0x0B80
+#define GL_ACCUM_GREEN_BITS 0x0D59
+#define GL_ACCUM_RED_BITS 0x0D58
+#define GL_ACTIVE_ATTRIBUTES 0x8B89
+#define GL_ACTIVE_ATTRIBUTE_MAX_LENGTH 0x8B8A
+#define GL_ACTIVE_TEXTURE 0x84E0
+#define GL_ACTIVE_UNIFORMS 0x8B86
+#define GL_ACTIVE_UNIFORM_BLOCKS 0x8A36
+#define GL_ACTIVE_UNIFORM_BLOCK_MAX_NAME_LENGTH 0x8A35
+#define GL_ACTIVE_UNIFORM_MAX_LENGTH 0x8B87
+#define GL_ADD 0x0104
+#define GL_ADD_SIGNED 0x8574
+#define GL_ALIASED_LINE_WIDTH_RANGE 0x846E
+#define GL_ALIASED_POINT_SIZE_RANGE 0x846D
+#define GL_ALL_ATTRIB_BITS 0xFFFFFFFF
+#define GL_ALPHA 0x1906
+#define GL_ALPHA12 0x803D
+#define GL_ALPHA16 0x803E
+#define GL_ALPHA4 0x803B
+#define GL_ALPHA8 0x803C
+#define GL_ALPHA_BIAS 0x0D1D
+#define GL_ALPHA_BITS 0x0D55
+#define GL_ALPHA_INTEGER 0x8D97
+#define GL_ALPHA_SCALE 0x0D1C
+#define GL_ALPHA_TEST 0x0BC0
+#define GL_ALPHA_TEST_FUNC 0x0BC1
+#define GL_ALPHA_TEST_REF 0x0BC2
+#define GL_ALREADY_SIGNALED 0x911A
+#define GL_ALWAYS 0x0207
+#define GL_AMBIENT 0x1200
+#define GL_AMBIENT_AND_DIFFUSE 0x1602
+#define GL_AND 0x1501
+#define GL_AND_INVERTED 0x1504
+#define GL_AND_REVERSE 0x1502
+#define GL_ANY_SAMPLES_PASSED 0x8C2F
+#define GL_ARRAY_BUFFER 0x8892
+#define GL_ARRAY_BUFFER_BINDING 0x8894
+#define GL_ATTACHED_SHADERS 0x8B85
+#define GL_ATTRIB_STACK_DEPTH 0x0BB0
+#define GL_AUTO_NORMAL 0x0D80
+#define GL_AUX0 0x0409
+#define GL_AUX1 0x040A
+#define GL_AUX2 0x040B
+#define GL_AUX3 0x040C
+#define GL_AUX_BUFFERS 0x0C00
+#define GL_BACK 0x0405
+#define GL_BACK_LEFT 0x0402
+#define GL_BACK_RIGHT 0x0403
+#define GL_BGR 0x80E0
+#define GL_BGRA 0x80E1
+#define GL_BGRA_INTEGER 0x8D9B
+#define GL_BGR_INTEGER 0x8D9A
+#define GL_BITMAP 0x1A00
+#define GL_BITMAP_TOKEN 0x0704
+#define GL_BLEND 0x0BE2
+#define GL_BLEND_COLOR 0x8005
+#define GL_BLEND_DST 0x0BE0
+#define GL_BLEND_DST_ALPHA 0x80CA
+#define GL_BLEND_DST_RGB 0x80C8
+#define GL_BLEND_EQUATION 0x8009
+#define GL_BLEND_EQUATION_ALPHA 0x883D
+#define GL_BLEND_EQUATION_RGB 0x8009
+#define GL_BLEND_SRC 0x0BE1
+#define GL_BLEND_SRC_ALPHA 0x80CB
+#define GL_BLEND_SRC_RGB 0x80C9
+#define GL_BLUE 0x1905
+#define GL_BLUE_BIAS 0x0D1B
+#define GL_BLUE_BITS 0x0D54
+#define GL_BLUE_INTEGER 0x8D96
+#define GL_BLUE_SCALE 0x0D1A
+#define GL_BOOL 0x8B56
+#define GL_BOOL_VEC2 0x8B57
+#define GL_BOOL_VEC3 0x8B58
+#define GL_BOOL_VEC4 0x8B59
+#define GL_BUFFER_ACCESS 0x88BB
+#define GL_BUFFER_ACCESS_FLAGS 0x911F
+#define GL_BUFFER_MAPPED 0x88BC
+#define GL_BUFFER_MAP_LENGTH 0x9120
+#define GL_BUFFER_MAP_OFFSET 0x9121
+#define GL_BUFFER_MAP_POINTER 0x88BD
+#define GL_BUFFER_SIZE 0x8764
+#define GL_BUFFER_USAGE 0x8765
+#define GL_BYTE 0x1400
+#define GL_C3F_V3F 0x2A24
+#define GL_C4F_N3F_V3F 0x2A26
+#define GL_C4UB_V2F 0x2A22
+#define GL_C4UB_V3F 0x2A23
+#define GL_CCW 0x0901
+#define GL_CLAMP 0x2900
+#define GL_CLAMP_FRAGMENT_COLOR 0x891B
+#define GL_CLAMP_READ_COLOR 0x891C
+#define GL_CLAMP_TO_BORDER 0x812D
+#define GL_CLAMP_TO_EDGE 0x812F
+#define GL_CLAMP_VERTEX_COLOR 0x891A
+#define GL_CLEAR 0x1500
+#define GL_CLIENT_ACTIVE_TEXTURE 0x84E1
+#define GL_CLIENT_ALL_ATTRIB_BITS 0xFFFFFFFF
+#define GL_CLIENT_ATTRIB_STACK_DEPTH 0x0BB1
+#define GL_CLIENT_PIXEL_STORE_BIT 0x00000001
+#define GL_CLIENT_VERTEX_ARRAY_BIT 0x00000002
+#define GL_CLIP_DISTANCE0 0x3000
+#define GL_CLIP_DISTANCE1 0x3001
+#define GL_CLIP_DISTANCE2 0x3002
+#define GL_CLIP_DISTANCE3 0x3003
+#define GL_CLIP_DISTANCE4 0x3004
+#define GL_CLIP_DISTANCE5 0x3005
+#define GL_CLIP_DISTANCE6 0x3006
+#define GL_CLIP_DISTANCE7 0x3007
+#define GL_CLIP_PLANE0 0x3000
+#define GL_CLIP_PLANE1 0x3001
+#define GL_CLIP_PLANE2 0x3002
+#define GL_CLIP_PLANE3 0x3003
+#define GL_CLIP_PLANE4 0x3004
+#define GL_CLIP_PLANE5 0x3005
+#define GL_COEFF 0x0A00
+#define GL_COLOR 0x1800
+#define GL_COLOR_ARRAY 0x8076
+#define GL_COLOR_ARRAY_BUFFER_BINDING 0x8898
+#define GL_COLOR_ARRAY_POINTER 0x8090
+#define GL_COLOR_ARRAY_SIZE 0x8081
+#define GL_COLOR_ARRAY_STRIDE 0x8083
+#define GL_COLOR_ARRAY_TYPE 0x8082
+#define GL_COLOR_ATTACHMENT0 0x8CE0
+#define GL_COLOR_ATTACHMENT1 0x8CE1
+#define GL_COLOR_ATTACHMENT10 0x8CEA
+#define GL_COLOR_ATTACHMENT11 0x8CEB
+#define GL_COLOR_ATTACHMENT12 0x8CEC
+#define GL_COLOR_ATTACHMENT13 0x8CED
+#define GL_COLOR_ATTACHMENT14 0x8CEE
+#define GL_COLOR_ATTACHMENT15 0x8CEF
+#define GL_COLOR_ATTACHMENT16 0x8CF0
+#define GL_COLOR_ATTACHMENT17 0x8CF1
+#define GL_COLOR_ATTACHMENT18 0x8CF2
+#define GL_COLOR_ATTACHMENT19 0x8CF3
+#define GL_COLOR_ATTACHMENT2 0x8CE2
+#define GL_COLOR_ATTACHMENT20 0x8CF4
+#define GL_COLOR_ATTACHMENT21 0x8CF5
+#define GL_COLOR_ATTACHMENT22 0x8CF6
+#define GL_COLOR_ATTACHMENT23 0x8CF7
+#define GL_COLOR_ATTACHMENT24 0x8CF8
+#define GL_COLOR_ATTACHMENT25 0x8CF9
+#define GL_COLOR_ATTACHMENT26 0x8CFA
+#define GL_COLOR_ATTACHMENT27 0x8CFB
+#define GL_COLOR_ATTACHMENT28 0x8CFC
+#define GL_COLOR_ATTACHMENT29 0x8CFD
+#define GL_COLOR_ATTACHMENT3 0x8CE3
+#define GL_COLOR_ATTACHMENT30 0x8CFE
+#define GL_COLOR_ATTACHMENT31 0x8CFF
+#define GL_COLOR_ATTACHMENT4 0x8CE4
+#define GL_COLOR_ATTACHMENT5 0x8CE5
+#define GL_COLOR_ATTACHMENT6 0x8CE6
+#define GL_COLOR_ATTACHMENT7 0x8CE7
+#define GL_COLOR_ATTACHMENT8 0x8CE8
+#define GL_COLOR_ATTACHMENT9 0x8CE9
+#define GL_COLOR_BUFFER_BIT 0x00004000
+#define GL_COLOR_CLEAR_VALUE 0x0C22
+#define GL_COLOR_INDEX 0x1900
+#define GL_COLOR_INDEXES 0x1603
+#define GL_COLOR_LOGIC_OP 0x0BF2
+#define GL_COLOR_MATERIAL 0x0B57
+#define GL_COLOR_MATERIAL_FACE 0x0B55
+#define GL_COLOR_MATERIAL_PARAMETER 0x0B56
+#define GL_COLOR_SUM 0x8458
+#define GL_COLOR_WRITEMASK 0x0C23
+#define GL_COMBINE 0x8570
+#define GL_COMBINE_ALPHA 0x8572
+#define GL_COMBINE_RGB 0x8571
+#define GL_COMPARE_REF_TO_TEXTURE 0x884E
+#define GL_COMPARE_R_TO_TEXTURE 0x884E
+#define GL_COMPILE 0x1300
+#define GL_COMPILE_AND_EXECUTE 0x1301
+#define GL_COMPILE_STATUS 0x8B81
+#define GL_COMPRESSED_ALPHA 0x84E9
+#define GL_COMPRESSED_INTENSITY 0x84EC
+#define GL_COMPRESSED_LUMINANCE 0x84EA
+#define GL_COMPRESSED_LUMINANCE_ALPHA 0x84EB
+#define GL_COMPRESSED_RED 0x8225
+#define GL_COMPRESSED_RED_RGTC1 0x8DBB
+#define GL_COMPRESSED_RG 0x8226
+#define GL_COMPRESSED_RGB 0x84ED
+#define GL_COMPRESSED_RGBA 0x84EE
+#define GL_COMPRESSED_RG_RGTC2 0x8DBD
+#define GL_COMPRESSED_SIGNED_RED_RGTC1 0x8DBC
+#define GL_COMPRESSED_SIGNED_RG_RGTC2 0x8DBE
+#define GL_COMPRESSED_SLUMINANCE 0x8C4A
+#define GL_COMPRESSED_SLUMINANCE_ALPHA 0x8C4B
+#define GL_COMPRESSED_SRGB 0x8C48
+#define GL_COMPRESSED_SRGB_ALPHA 0x8C49
+#define GL_COMPRESSED_TEXTURE_FORMATS 0x86A3
+#define GL_CONDITION_SATISFIED 0x911C
+#define GL_CONSTANT 0x8576
+#define GL_CONSTANT_ALPHA 0x8003
+#define GL_CONSTANT_ATTENUATION 0x1207
+#define GL_CONSTANT_COLOR 0x8001
+#define GL_CONTEXT_COMPATIBILITY_PROFILE_BIT 0x00000002
+#define GL_CONTEXT_CORE_PROFILE_BIT 0x00000001
+#define GL_CONTEXT_FLAGS 0x821E
+#define GL_CONTEXT_FLAG_FORWARD_COMPATIBLE_BIT 0x00000001
+#define GL_CONTEXT_PROFILE_MASK 0x9126
+#define GL_COORD_REPLACE 0x8862
+#define GL_COPY 0x1503
+#define GL_COPY_INVERTED 0x150C
+#define GL_COPY_PIXEL_TOKEN 0x0706
+#define GL_COPY_READ_BUFFER 0x8F36
+#define GL_COPY_WRITE_BUFFER 0x8F37
+#define GL_CULL_FACE 0x0B44
+#define GL_CULL_FACE_MODE 0x0B45
+#define GL_CURRENT_BIT 0x00000001
+#define GL_CURRENT_COLOR 0x0B00
+#define GL_CURRENT_FOG_COORD 0x8453
+#define GL_CURRENT_FOG_COORDINATE 0x8453
+#define GL_CURRENT_INDEX 0x0B01
+#define GL_CURRENT_NORMAL 0x0B02
+#define GL_CURRENT_PROGRAM 0x8B8D
+#define GL_CURRENT_QUERY 0x8865
+#define GL_CURRENT_RASTER_COLOR 0x0B04
+#define GL_CURRENT_RASTER_DISTANCE 0x0B09
+#define GL_CURRENT_RASTER_INDEX 0x0B05
+#define GL_CURRENT_RASTER_POSITION 0x0B07
+#define GL_CURRENT_RASTER_POSITION_VALID 0x0B08
+#define GL_CURRENT_RASTER_SECONDARY_COLOR 0x845F
+#define GL_CURRENT_RASTER_TEXTURE_COORDS 0x0B06
+#define GL_CURRENT_SECONDARY_COLOR 0x8459
+#define GL_CURRENT_TEXTURE_COORDS 0x0B03
+#define GL_CURRENT_VERTEX_ATTRIB 0x8626
+#define GL_CW 0x0900
+#define GL_DECAL 0x2101
+#define GL_DECR 0x1E03
+#define GL_DECR_WRAP 0x8508
+#define GL_DELETE_STATUS 0x8B80
+#define GL_DEPTH 0x1801
+#define GL_DEPTH24_STENCIL8 0x88F0
+#define GL_DEPTH32F_STENCIL8 0x8CAD
+#define GL_DEPTH_ATTACHMENT 0x8D00
+#define GL_DEPTH_BIAS 0x0D1F
+#define GL_DEPTH_BITS 0x0D56
+#define GL_DEPTH_BUFFER_BIT 0x00000100
+#define GL_DEPTH_CLAMP 0x864F
+#define GL_DEPTH_CLEAR_VALUE 0x0B73
+#define GL_DEPTH_COMPONENT 0x1902
+#define GL_DEPTH_COMPONENT16 0x81A5
+#define GL_DEPTH_COMPONENT24 0x81A6
+#define GL_DEPTH_COMPONENT32 0x81A7
+#define GL_DEPTH_COMPONENT32F 0x8CAC
+#define GL_DEPTH_FUNC 0x0B74
+#define GL_DEPTH_RANGE 0x0B70
+#define GL_DEPTH_SCALE 0x0D1E
+#define GL_DEPTH_STENCIL 0x84F9
+#define GL_DEPTH_STENCIL_ATTACHMENT 0x821A
+#define GL_DEPTH_TEST 0x0B71
+#define GL_DEPTH_TEXTURE_MODE 0x884B
+#define GL_DEPTH_WRITEMASK 0x0B72
+#define GL_DIFFUSE 0x1201
+#define GL_DITHER 0x0BD0
+#define GL_DOMAIN 0x0A02
+#define GL_DONT_CARE 0x1100
+#define GL_DOT3_RGB 0x86AE
+#define GL_DOT3_RGBA 0x86AF
+#define GL_DOUBLE 0x140A
+#define GL_DOUBLEBUFFER 0x0C32
+#define GL_DRAW_BUFFER 0x0C01
+#define GL_DRAW_BUFFER0 0x8825
+#define GL_DRAW_BUFFER1 0x8826
+#define GL_DRAW_BUFFER10 0x882F
+#define GL_DRAW_BUFFER11 0x8830
+#define GL_DRAW_BUFFER12 0x8831
+#define GL_DRAW_BUFFER13 0x8832
+#define GL_DRAW_BUFFER14 0x8833
+#define GL_DRAW_BUFFER15 0x8834
+#define GL_DRAW_BUFFER2 0x8827
+#define GL_DRAW_BUFFER3 0x8828
+#define GL_DRAW_BUFFER4 0x8829
+#define GL_DRAW_BUFFER5 0x882A
+#define GL_DRAW_BUFFER6 0x882B
+#define GL_DRAW_BUFFER7 0x882C
+#define GL_DRAW_BUFFER8 0x882D
+#define GL_DRAW_BUFFER9 0x882E
+#define GL_DRAW_FRAMEBUFFER 0x8CA9
+#define GL_DRAW_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_DRAW_PIXEL_TOKEN 0x0705
+#define GL_DST_ALPHA 0x0304
+#define GL_DST_COLOR 0x0306
+#define GL_DYNAMIC_COPY 0x88EA
+#define GL_DYNAMIC_DRAW 0x88E8
+#define GL_DYNAMIC_READ 0x88E9
+#define GL_EDGE_FLAG 0x0B43
+#define GL_EDGE_FLAG_ARRAY 0x8079
+#define GL_EDGE_FLAG_ARRAY_BUFFER_BINDING 0x889B
+#define GL_EDGE_FLAG_ARRAY_POINTER 0x8093
+#define GL_EDGE_FLAG_ARRAY_STRIDE 0x808C
+#define GL_ELEMENT_ARRAY_BUFFER 0x8893
+#define GL_ELEMENT_ARRAY_BUFFER_BINDING 0x8895
+#define GL_EMISSION 0x1600
+#define GL_ENABLE_BIT 0x00002000
+#define GL_EQUAL 0x0202
+#define GL_EQUIV 0x1509
+#define GL_EVAL_BIT 0x00010000
+#define GL_EXP 0x0800
+#define GL_EXP2 0x0801
+#define GL_EXTENSIONS 0x1F03
+#define GL_EYE_LINEAR 0x2400
+#define GL_EYE_PLANE 0x2502
+#define GL_FALSE 0
+#define GL_FASTEST 0x1101
+#define GL_FEEDBACK 0x1C01
+#define GL_FEEDBACK_BUFFER_POINTER 0x0DF0
+#define GL_FEEDBACK_BUFFER_SIZE 0x0DF1
+#define GL_FEEDBACK_BUFFER_TYPE 0x0DF2
+#define GL_FILL 0x1B02
+#define GL_FIRST_VERTEX_CONVENTION 0x8E4D
+#define GL_FIXED_ONLY 0x891D
+#define GL_FLAT 0x1D00
+#define GL_FLOAT 0x1406
+#define GL_FLOAT_32_UNSIGNED_INT_24_8_REV 0x8DAD
+#define GL_FLOAT_MAT2 0x8B5A
+#define GL_FLOAT_MAT2x3 0x8B65
+#define GL_FLOAT_MAT2x4 0x8B66
+#define GL_FLOAT_MAT3 0x8B5B
+#define GL_FLOAT_MAT3x2 0x8B67
+#define GL_FLOAT_MAT3x4 0x8B68
+#define GL_FLOAT_MAT4 0x8B5C
+#define GL_FLOAT_MAT4x2 0x8B69
+#define GL_FLOAT_MAT4x3 0x8B6A
+#define GL_FLOAT_VEC2 0x8B50
+#define GL_FLOAT_VEC3 0x8B51
+#define GL_FLOAT_VEC4 0x8B52
+#define GL_FOG 0x0B60
+#define GL_FOG_BIT 0x00000080
+#define GL_FOG_COLOR 0x0B66
+#define GL_FOG_COORD 0x8451
+#define GL_FOG_COORDINATE 0x8451
+#define GL_FOG_COORDINATE_ARRAY 0x8457
+#define GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING 0x889D
+#define GL_FOG_COORDINATE_ARRAY_POINTER 0x8456
+#define GL_FOG_COORDINATE_ARRAY_STRIDE 0x8455
+#define GL_FOG_COORDINATE_ARRAY_TYPE 0x8454
+#define GL_FOG_COORDINATE_SOURCE 0x8450
+#define GL_FOG_COORD_ARRAY 0x8457
+#define GL_FOG_COORD_ARRAY_BUFFER_BINDING 0x889D
+#define GL_FOG_COORD_ARRAY_POINTER 0x8456
+#define GL_FOG_COORD_ARRAY_STRIDE 0x8455
+#define GL_FOG_COORD_ARRAY_TYPE 0x8454
+#define GL_FOG_COORD_SRC 0x8450
+#define GL_FOG_DENSITY 0x0B62
+#define GL_FOG_END 0x0B64
+#define GL_FOG_HINT 0x0C54
+#define GL_FOG_INDEX 0x0B61
+#define GL_FOG_MODE 0x0B65
+#define GL_FOG_START 0x0B63
+#define GL_FRAGMENT_DEPTH 0x8452
+#define GL_FRAGMENT_SHADER 0x8B30
+#define GL_FRAGMENT_SHADER_DERIVATIVE_HINT 0x8B8B
+#define GL_FRAMEBUFFER 0x8D40
+#define GL_FRAMEBUFFER_ATTACHMENT_ALPHA_SIZE 0x8215
+#define GL_FRAMEBUFFER_ATTACHMENT_BLUE_SIZE 0x8214
+#define GL_FRAMEBUFFER_ATTACHMENT_COLOR_ENCODING 0x8210
+#define GL_FRAMEBUFFER_ATTACHMENT_COMPONENT_TYPE 0x8211
+#define GL_FRAMEBUFFER_ATTACHMENT_DEPTH_SIZE 0x8216
+#define GL_FRAMEBUFFER_ATTACHMENT_GREEN_SIZE 0x8213
+#define GL_FRAMEBUFFER_ATTACHMENT_LAYERED 0x8DA7
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_NAME 0x8CD1
+#define GL_FRAMEBUFFER_ATTACHMENT_OBJECT_TYPE 0x8CD0
+#define GL_FRAMEBUFFER_ATTACHMENT_RED_SIZE 0x8212
+#define GL_FRAMEBUFFER_ATTACHMENT_STENCIL_SIZE 0x8217
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE 0x8CD3
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LAYER 0x8CD4
+#define GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_LEVEL 0x8CD2
+#define GL_FRAMEBUFFER_BINDING 0x8CA6
+#define GL_FRAMEBUFFER_COMPLETE 0x8CD5
+#define GL_FRAMEBUFFER_DEFAULT 0x8218
+#define GL_FRAMEBUFFER_INCOMPLETE_ATTACHMENT 0x8CD6
+#define GL_FRAMEBUFFER_INCOMPLETE_DRAW_BUFFER 0x8CDB
+#define GL_FRAMEBUFFER_INCOMPLETE_LAYER_TARGETS 0x8DA8
+#define GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT 0x8CD7
+#define GL_FRAMEBUFFER_INCOMPLETE_MULTISAMPLE 0x8D56
+#define GL_FRAMEBUFFER_INCOMPLETE_READ_BUFFER 0x8CDC
+#define GL_FRAMEBUFFER_SRGB 0x8DB9
+#define GL_FRAMEBUFFER_UNDEFINED 0x8219
+#define GL_FRAMEBUFFER_UNSUPPORTED 0x8CDD
+#define GL_FRONT 0x0404
+#define GL_FRONT_AND_BACK 0x0408
+#define GL_FRONT_FACE 0x0B46
+#define GL_FRONT_LEFT 0x0400
+#define GL_FRONT_RIGHT 0x0401
+#define GL_FUNC_ADD 0x8006
+#define GL_FUNC_REVERSE_SUBTRACT 0x800B
+#define GL_FUNC_SUBTRACT 0x800A
+#define GL_GENERATE_MIPMAP 0x8191
+#define GL_GENERATE_MIPMAP_HINT 0x8192
+#define GL_GEOMETRY_INPUT_TYPE 0x8917
+#define GL_GEOMETRY_OUTPUT_TYPE 0x8918
+#define GL_GEOMETRY_SHADER 0x8DD9
+#define GL_GEOMETRY_VERTICES_OUT 0x8916
+#define GL_GEQUAL 0x0206
+#define GL_GREATER 0x0204
+#define GL_GREEN 0x1904
+#define GL_GREEN_BIAS 0x0D19
+#define GL_GREEN_BITS 0x0D53
+#define GL_GREEN_INTEGER 0x8D95
+#define GL_GREEN_SCALE 0x0D18
+#define GL_HALF_FLOAT 0x140B
+#define GL_HINT_BIT 0x00008000
+#define GL_INCR 0x1E02
+#define GL_INCR_WRAP 0x8507
+#define GL_INDEX 0x8222
+#define GL_INDEX_ARRAY 0x8077
+#define GL_INDEX_ARRAY_BUFFER_BINDING 0x8899
+#define GL_INDEX_ARRAY_POINTER 0x8091
+#define GL_INDEX_ARRAY_STRIDE 0x8086
+#define GL_INDEX_ARRAY_TYPE 0x8085
+#define GL_INDEX_BITS 0x0D51
+#define GL_INDEX_CLEAR_VALUE 0x0C20
+#define GL_INDEX_LOGIC_OP 0x0BF1
+#define GL_INDEX_MODE 0x0C30
+#define GL_INDEX_OFFSET 0x0D13
+#define GL_INDEX_SHIFT 0x0D12
+#define GL_INDEX_WRITEMASK 0x0C21
+#define GL_INFO_LOG_LENGTH 0x8B84
+#define GL_INT 0x1404
+#define GL_INTENSITY 0x8049
+#define GL_INTENSITY12 0x804C
+#define GL_INTENSITY16 0x804D
+#define GL_INTENSITY4 0x804A
+#define GL_INTENSITY8 0x804B
+#define GL_INTERLEAVED_ATTRIBS 0x8C8C
+#define GL_INTERPOLATE 0x8575
+#define GL_INT_2_10_10_10_REV 0x8D9F
+#define GL_INT_SAMPLER_1D 0x8DC9
+#define GL_INT_SAMPLER_1D_ARRAY 0x8DCE
+#define GL_INT_SAMPLER_2D 0x8DCA
+#define GL_INT_SAMPLER_2D_ARRAY 0x8DCF
+#define GL_INT_SAMPLER_2D_MULTISAMPLE 0x9109
+#define GL_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910C
+#define GL_INT_SAMPLER_2D_RECT 0x8DCD
+#define GL_INT_SAMPLER_3D 0x8DCB
+#define GL_INT_SAMPLER_BUFFER 0x8DD0
+#define GL_INT_SAMPLER_CUBE 0x8DCC
+#define GL_INT_VEC2 0x8B53
+#define GL_INT_VEC3 0x8B54
+#define GL_INT_VEC4 0x8B55
+#define GL_INVALID_ENUM 0x0500
+#define GL_INVALID_FRAMEBUFFER_OPERATION 0x0506
+#define GL_INVALID_INDEX 0xFFFFFFFF
+#define GL_INVALID_OPERATION 0x0502
+#define GL_INVALID_VALUE 0x0501
+#define GL_INVERT 0x150A
+#define GL_KEEP 0x1E00
+#define GL_LAST_VERTEX_CONVENTION 0x8E4E
+#define GL_LEFT 0x0406
+#define GL_LEQUAL 0x0203
+#define GL_LESS 0x0201
+#define GL_LIGHT0 0x4000
+#define GL_LIGHT1 0x4001
+#define GL_LIGHT2 0x4002
+#define GL_LIGHT3 0x4003
+#define GL_LIGHT4 0x4004
+#define GL_LIGHT5 0x4005
+#define GL_LIGHT6 0x4006
+#define GL_LIGHT7 0x4007
+#define GL_LIGHTING 0x0B50
+#define GL_LIGHTING_BIT 0x00000040
+#define GL_LIGHT_MODEL_AMBIENT 0x0B53
+#define GL_LIGHT_MODEL_COLOR_CONTROL 0x81F8
+#define GL_LIGHT_MODEL_LOCAL_VIEWER 0x0B51
+#define GL_LIGHT_MODEL_TWO_SIDE 0x0B52
+#define GL_LINE 0x1B01
+#define GL_LINEAR 0x2601
+#define GL_LINEAR_ATTENUATION 0x1208
+#define GL_LINEAR_MIPMAP_LINEAR 0x2703
+#define GL_LINEAR_MIPMAP_NEAREST 0x2701
+#define GL_LINES 0x0001
+#define GL_LINES_ADJACENCY 0x000A
+#define GL_LINE_BIT 0x00000004
+#define GL_LINE_LOOP 0x0002
+#define GL_LINE_RESET_TOKEN 0x0707
+#define GL_LINE_SMOOTH 0x0B20
+#define GL_LINE_SMOOTH_HINT 0x0C52
+#define GL_LINE_STIPPLE 0x0B24
+#define GL_LINE_STIPPLE_PATTERN 0x0B25
+#define GL_LINE_STIPPLE_REPEAT 0x0B26
+#define GL_LINE_STRIP 0x0003
+#define GL_LINE_STRIP_ADJACENCY 0x000B
+#define GL_LINE_TOKEN 0x0702
+#define GL_LINE_WIDTH 0x0B21
+#define GL_LINE_WIDTH_GRANULARITY 0x0B23
+#define GL_LINE_WIDTH_RANGE 0x0B22
+#define GL_LINK_STATUS 0x8B82
+#define GL_LIST_BASE 0x0B32
+#define GL_LIST_BIT 0x00020000
+#define GL_LIST_INDEX 0x0B33
+#define GL_LIST_MODE 0x0B30
+#define GL_LOAD 0x0101
+#define GL_LOGIC_OP 0x0BF1
+#define GL_LOGIC_OP_MODE 0x0BF0
+#define GL_LOWER_LEFT 0x8CA1
+#define GL_LUMINANCE 0x1909
+#define GL_LUMINANCE12 0x8041
+#define GL_LUMINANCE12_ALPHA12 0x8047
+#define GL_LUMINANCE12_ALPHA4 0x8046
+#define GL_LUMINANCE16 0x8042
+#define GL_LUMINANCE16_ALPHA16 0x8048
+#define GL_LUMINANCE4 0x803F
+#define GL_LUMINANCE4_ALPHA4 0x8043
+#define GL_LUMINANCE6_ALPHA2 0x8044
+#define GL_LUMINANCE8 0x8040
+#define GL_LUMINANCE8_ALPHA8 0x8045
+#define GL_LUMINANCE_ALPHA 0x190A
+#define GL_MAJOR_VERSION 0x821B
+#define GL_MAP1_COLOR_4 0x0D90
+#define GL_MAP1_GRID_DOMAIN 0x0DD0
+#define GL_MAP1_GRID_SEGMENTS 0x0DD1
+#define GL_MAP1_INDEX 0x0D91
+#define GL_MAP1_NORMAL 0x0D92
+#define GL_MAP1_TEXTURE_COORD_1 0x0D93
+#define GL_MAP1_TEXTURE_COORD_2 0x0D94
+#define GL_MAP1_TEXTURE_COORD_3 0x0D95
+#define GL_MAP1_TEXTURE_COORD_4 0x0D96
+#define GL_MAP1_VERTEX_3 0x0D97
+#define GL_MAP1_VERTEX_4 0x0D98
+#define GL_MAP2_COLOR_4 0x0DB0
+#define GL_MAP2_GRID_DOMAIN 0x0DD2
+#define GL_MAP2_GRID_SEGMENTS 0x0DD3
+#define GL_MAP2_INDEX 0x0DB1
+#define GL_MAP2_NORMAL 0x0DB2
+#define GL_MAP2_TEXTURE_COORD_1 0x0DB3
+#define GL_MAP2_TEXTURE_COORD_2 0x0DB4
+#define GL_MAP2_TEXTURE_COORD_3 0x0DB5
+#define GL_MAP2_TEXTURE_COORD_4 0x0DB6
+#define GL_MAP2_VERTEX_3 0x0DB7
+#define GL_MAP2_VERTEX_4 0x0DB8
+#define GL_MAP_COLOR 0x0D10
+#define GL_MAP_FLUSH_EXPLICIT_BIT 0x0010
+#define GL_MAP_INVALIDATE_BUFFER_BIT 0x0008
+#define GL_MAP_INVALIDATE_RANGE_BIT 0x0004
+#define GL_MAP_READ_BIT 0x0001
+#define GL_MAP_STENCIL 0x0D11
+#define GL_MAP_UNSYNCHRONIZED_BIT 0x0020
+#define GL_MAP_WRITE_BIT 0x0002
+#define GL_MATRIX_MODE 0x0BA0
+#define GL_MAX 0x8008
+#define GL_MAX_3D_TEXTURE_SIZE 0x8073
+#define GL_MAX_ARRAY_TEXTURE_LAYERS 0x88FF
+#define GL_MAX_ATTRIB_STACK_DEPTH 0x0D35
+#define GL_MAX_CLIENT_ATTRIB_STACK_DEPTH 0x0D3B
+#define GL_MAX_CLIP_DISTANCES 0x0D32
+#define GL_MAX_CLIP_PLANES 0x0D32
+#define GL_MAX_COLOR_ATTACHMENTS 0x8CDF
+#define GL_MAX_COLOR_TEXTURE_SAMPLES 0x910E
+#define GL_MAX_COMBINED_FRAGMENT_UNIFORM_COMPONENTS 0x8A33
+#define GL_MAX_COMBINED_GEOMETRY_UNIFORM_COMPONENTS 0x8A32
+#define GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS 0x8B4D
+#define GL_MAX_COMBINED_UNIFORM_BLOCKS 0x8A2E
+#define GL_MAX_COMBINED_VERTEX_UNIFORM_COMPONENTS 0x8A31
+#define GL_MAX_CUBE_MAP_TEXTURE_SIZE 0x851C
+#define GL_MAX_DEPTH_TEXTURE_SAMPLES 0x910F
+#define GL_MAX_DRAW_BUFFERS 0x8824
+#define GL_MAX_DUAL_SOURCE_DRAW_BUFFERS 0x88FC
+#define GL_MAX_ELEMENTS_INDICES 0x80E9
+#define GL_MAX_ELEMENTS_VERTICES 0x80E8
+#define GL_MAX_EVAL_ORDER 0x0D30
+#define GL_MAX_FRAGMENT_INPUT_COMPONENTS 0x9125
+#define GL_MAX_FRAGMENT_UNIFORM_BLOCKS 0x8A2D
+#define GL_MAX_FRAGMENT_UNIFORM_COMPONENTS 0x8B49
+#define GL_MAX_GEOMETRY_INPUT_COMPONENTS 0x9123
+#define GL_MAX_GEOMETRY_OUTPUT_COMPONENTS 0x9124
+#define GL_MAX_GEOMETRY_OUTPUT_VERTICES 0x8DE0
+#define GL_MAX_GEOMETRY_TEXTURE_IMAGE_UNITS 0x8C29
+#define GL_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS 0x8DE1
+#define GL_MAX_GEOMETRY_UNIFORM_BLOCKS 0x8A2C
+#define GL_MAX_GEOMETRY_UNIFORM_COMPONENTS 0x8DDF
+#define GL_MAX_INTEGER_SAMPLES 0x9110
+#define GL_MAX_LIGHTS 0x0D31
+#define GL_MAX_LIST_NESTING 0x0B31
+#define GL_MAX_MODELVIEW_STACK_DEPTH 0x0D36
+#define GL_MAX_NAME_STACK_DEPTH 0x0D37
+#define GL_MAX_PIXEL_MAP_TABLE 0x0D34
+#define GL_MAX_PROGRAM_TEXEL_OFFSET 0x8905
+#define GL_MAX_PROJECTION_STACK_DEPTH 0x0D38
+#define GL_MAX_RECTANGLE_TEXTURE_SIZE 0x84F8
+#define GL_MAX_RENDERBUFFER_SIZE 0x84E8
+#define GL_MAX_SAMPLES 0x8D57
+#define GL_MAX_SAMPLE_MASK_WORDS 0x8E59
+#define GL_MAX_SERVER_WAIT_TIMEOUT 0x9111
+#define GL_MAX_TEXTURE_BUFFER_SIZE 0x8C2B
+#define GL_MAX_TEXTURE_COORDS 0x8871
+#define GL_MAX_TEXTURE_IMAGE_UNITS 0x8872
+#define GL_MAX_TEXTURE_LOD_BIAS 0x84FD
+#define GL_MAX_TEXTURE_SIZE 0x0D33
+#define GL_MAX_TEXTURE_STACK_DEPTH 0x0D39
+#define GL_MAX_TEXTURE_UNITS 0x84E2
+#define GL_MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS 0x8C8A
+#define GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_ATTRIBS 0x8C8B
+#define GL_MAX_TRANSFORM_FEEDBACK_SEPARATE_COMPONENTS 0x8C80
+#define GL_MAX_UNIFORM_BLOCK_SIZE 0x8A30
+#define GL_MAX_UNIFORM_BUFFER_BINDINGS 0x8A2F
+#define GL_MAX_VARYING_COMPONENTS 0x8B4B
+#define GL_MAX_VARYING_FLOATS 0x8B4B
+#define GL_MAX_VERTEX_ATTRIBS 0x8869
+#define GL_MAX_VERTEX_OUTPUT_COMPONENTS 0x9122
+#define GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS 0x8B4C
+#define GL_MAX_VERTEX_UNIFORM_BLOCKS 0x8A2B
+#define GL_MAX_VERTEX_UNIFORM_COMPONENTS 0x8B4A
+#define GL_MAX_VIEWPORT_DIMS 0x0D3A
+#define GL_MIN 0x8007
+#define GL_MINOR_VERSION 0x821C
+#define GL_MIN_PROGRAM_TEXEL_OFFSET 0x8904
+#define GL_MIRRORED_REPEAT 0x8370
+#define GL_MODELVIEW 0x1700
+#define GL_MODELVIEW_MATRIX 0x0BA6
+#define GL_MODELVIEW_STACK_DEPTH 0x0BA3
+#define GL_MODULATE 0x2100
+#define GL_MULT 0x0103
+#define GL_MULTISAMPLE 0x809D
+#define GL_MULTISAMPLE_BIT 0x20000000
+#define GL_N3F_V3F 0x2A25
+#define GL_NAME_STACK_DEPTH 0x0D70
+#define GL_NAND 0x150E
+#define GL_NEAREST 0x2600
+#define GL_NEAREST_MIPMAP_LINEAR 0x2702
+#define GL_NEAREST_MIPMAP_NEAREST 0x2700
+#define GL_NEVER 0x0200
+#define GL_NICEST 0x1102
+#define GL_NONE 0
+#define GL_NOOP 0x1505
+#define GL_NOR 0x1508
+#define GL_NORMALIZE 0x0BA1
+#define GL_NORMAL_ARRAY 0x8075
+#define GL_NORMAL_ARRAY_BUFFER_BINDING 0x8897
+#define GL_NORMAL_ARRAY_POINTER 0x808F
+#define GL_NORMAL_ARRAY_STRIDE 0x807F
+#define GL_NORMAL_ARRAY_TYPE 0x807E
+#define GL_NORMAL_MAP 0x8511
+#define GL_NOTEQUAL 0x0205
+#define GL_NO_ERROR 0
+#define GL_NUM_COMPRESSED_TEXTURE_FORMATS 0x86A2
+#define GL_NUM_EXTENSIONS 0x821D
+#define GL_OBJECT_LINEAR 0x2401
+#define GL_OBJECT_PLANE 0x2501
+#define GL_OBJECT_TYPE 0x9112
+#define GL_ONE 1
+#define GL_ONE_MINUS_CONSTANT_ALPHA 0x8004
+#define GL_ONE_MINUS_CONSTANT_COLOR 0x8002
+#define GL_ONE_MINUS_DST_ALPHA 0x0305
+#define GL_ONE_MINUS_DST_COLOR 0x0307
+#define GL_ONE_MINUS_SRC1_ALPHA 0x88FB
+#define GL_ONE_MINUS_SRC1_COLOR 0x88FA
+#define GL_ONE_MINUS_SRC_ALPHA 0x0303
+#define GL_ONE_MINUS_SRC_COLOR 0x0301
+#define GL_OPERAND0_ALPHA 0x8598
+#define GL_OPERAND0_RGB 0x8590
+#define GL_OPERAND1_ALPHA 0x8599
+#define GL_OPERAND1_RGB 0x8591
+#define GL_OPERAND2_ALPHA 0x859A
+#define GL_OPERAND2_RGB 0x8592
+#define GL_OR 0x1507
+#define GL_ORDER 0x0A01
+#define GL_OR_INVERTED 0x150D
+#define GL_OR_REVERSE 0x150B
+#define GL_OUT_OF_MEMORY 0x0505
+#define GL_PACK_ALIGNMENT 0x0D05
+#define GL_PACK_IMAGE_HEIGHT 0x806C
+#define GL_PACK_LSB_FIRST 0x0D01
+#define GL_PACK_ROW_LENGTH 0x0D02
+#define GL_PACK_SKIP_IMAGES 0x806B
+#define GL_PACK_SKIP_PIXELS 0x0D04
+#define GL_PACK_SKIP_ROWS 0x0D03
+#define GL_PACK_SWAP_BYTES 0x0D00
+#define GL_PASS_THROUGH_TOKEN 0x0700
+#define GL_PERSPECTIVE_CORRECTION_HINT 0x0C50
+#define GL_PIXEL_MAP_A_TO_A 0x0C79
+#define GL_PIXEL_MAP_A_TO_A_SIZE 0x0CB9
+#define GL_PIXEL_MAP_B_TO_B 0x0C78
+#define GL_PIXEL_MAP_B_TO_B_SIZE 0x0CB8
+#define GL_PIXEL_MAP_G_TO_G 0x0C77
+#define GL_PIXEL_MAP_G_TO_G_SIZE 0x0CB7
+#define GL_PIXEL_MAP_I_TO_A 0x0C75
+#define GL_PIXEL_MAP_I_TO_A_SIZE 0x0CB5
+#define GL_PIXEL_MAP_I_TO_B 0x0C74
+#define GL_PIXEL_MAP_I_TO_B_SIZE 0x0CB4
+#define GL_PIXEL_MAP_I_TO_G 0x0C73
+#define GL_PIXEL_MAP_I_TO_G_SIZE 0x0CB3
+#define GL_PIXEL_MAP_I_TO_I 0x0C70
+#define GL_PIXEL_MAP_I_TO_I_SIZE 0x0CB0
+#define GL_PIXEL_MAP_I_TO_R 0x0C72
+#define GL_PIXEL_MAP_I_TO_R_SIZE 0x0CB2
+#define GL_PIXEL_MAP_R_TO_R 0x0C76
+#define GL_PIXEL_MAP_R_TO_R_SIZE 0x0CB6
+#define GL_PIXEL_MAP_S_TO_S 0x0C71
+#define GL_PIXEL_MAP_S_TO_S_SIZE 0x0CB1
+#define GL_PIXEL_MODE_BIT 0x00000020
+#define GL_PIXEL_PACK_BUFFER 0x88EB
+#define GL_PIXEL_PACK_BUFFER_BINDING 0x88ED
+#define GL_PIXEL_UNPACK_BUFFER 0x88EC
+#define GL_PIXEL_UNPACK_BUFFER_BINDING 0x88EF
+#define GL_POINT 0x1B00
+#define GL_POINTS 0x0000
+#define GL_POINT_BIT 0x00000002
+#define GL_POINT_DISTANCE_ATTENUATION 0x8129
+#define GL_POINT_FADE_THRESHOLD_SIZE 0x8128
+#define GL_POINT_SIZE 0x0B11
+#define GL_POINT_SIZE_GRANULARITY 0x0B13
+#define GL_POINT_SIZE_MAX 0x8127
+#define GL_POINT_SIZE_MIN 0x8126
+#define GL_POINT_SIZE_RANGE 0x0B12
+#define GL_POINT_SMOOTH 0x0B10
+#define GL_POINT_SMOOTH_HINT 0x0C51
+#define GL_POINT_SPRITE 0x8861
+#define GL_POINT_SPRITE_COORD_ORIGIN 0x8CA0
+#define GL_POINT_TOKEN 0x0701
+#define GL_POLYGON 0x0009
+#define GL_POLYGON_BIT 0x00000008
+#define GL_POLYGON_MODE 0x0B40
+#define GL_POLYGON_OFFSET_FACTOR 0x8038
+#define GL_POLYGON_OFFSET_FILL 0x8037
+#define GL_POLYGON_OFFSET_LINE 0x2A02
+#define GL_POLYGON_OFFSET_POINT 0x2A01
+#define GL_POLYGON_OFFSET_UNITS 0x2A00
+#define GL_POLYGON_SMOOTH 0x0B41
+#define GL_POLYGON_SMOOTH_HINT 0x0C53
+#define GL_POLYGON_STIPPLE 0x0B42
+#define GL_POLYGON_STIPPLE_BIT 0x00000010
+#define GL_POLYGON_TOKEN 0x0703
+#define GL_POSITION 0x1203
+#define GL_PREVIOUS 0x8578
+#define GL_PRIMARY_COLOR 0x8577
+#define GL_PRIMITIVES_GENERATED 0x8C87
+#define GL_PRIMITIVE_RESTART 0x8F9D
+#define GL_PRIMITIVE_RESTART_INDEX 0x8F9E
+#define GL_PROGRAM_POINT_SIZE 0x8642
+#define GL_PROJECTION 0x1701
+#define GL_PROJECTION_MATRIX 0x0BA7
+#define GL_PROJECTION_STACK_DEPTH 0x0BA4
+#define GL_PROVOKING_VERTEX 0x8E4F
+#define GL_PROXY_TEXTURE_1D 0x8063
+#define GL_PROXY_TEXTURE_1D_ARRAY 0x8C19
+#define GL_PROXY_TEXTURE_2D 0x8064
+#define GL_PROXY_TEXTURE_2D_ARRAY 0x8C1B
+#define GL_PROXY_TEXTURE_2D_MULTISAMPLE 0x9101
+#define GL_PROXY_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9103
+#define GL_PROXY_TEXTURE_3D 0x8070
+#define GL_PROXY_TEXTURE_CUBE_MAP 0x851B
+#define GL_PROXY_TEXTURE_RECTANGLE 0x84F7
+#define GL_Q 0x2003
+#define GL_QUADRATIC_ATTENUATION 0x1209
+#define GL_QUADS 0x0007
+#define GL_QUADS_FOLLOW_PROVOKING_VERTEX_CONVENTION 0x8E4C
+#define GL_QUAD_STRIP 0x0008
+#define GL_QUERY_BY_REGION_NO_WAIT 0x8E16
+#define GL_QUERY_BY_REGION_WAIT 0x8E15
+#define GL_QUERY_COUNTER_BITS 0x8864
+#define GL_QUERY_NO_WAIT 0x8E14
+#define GL_QUERY_RESULT 0x8866
+#define GL_QUERY_RESULT_AVAILABLE 0x8867
+#define GL_QUERY_WAIT 0x8E13
+#define GL_R 0x2002
+#define GL_R11F_G11F_B10F 0x8C3A
+#define GL_R16 0x822A
+#define GL_R16F 0x822D
+#define GL_R16I 0x8233
+#define GL_R16UI 0x8234
+#define GL_R16_SNORM 0x8F98
+#define GL_R32F 0x822E
+#define GL_R32I 0x8235
+#define GL_R32UI 0x8236
+#define GL_R3_G3_B2 0x2A10
+#define GL_R8 0x8229
+#define GL_R8I 0x8231
+#define GL_R8UI 0x8232
+#define GL_R8_SNORM 0x8F94
+#define GL_RASTERIZER_DISCARD 0x8C89
+#define GL_READ_BUFFER 0x0C02
+#define GL_READ_FRAMEBUFFER 0x8CA8
+#define GL_READ_FRAMEBUFFER_BINDING 0x8CAA
+#define GL_READ_ONLY 0x88B8
+#define GL_READ_WRITE 0x88BA
+#define GL_RED 0x1903
+#define GL_RED_BIAS 0x0D15
+#define GL_RED_BITS 0x0D52
+#define GL_RED_INTEGER 0x8D94
+#define GL_RED_SCALE 0x0D14
+#define GL_REFLECTION_MAP 0x8512
+#define GL_RENDER 0x1C00
+#define GL_RENDERBUFFER 0x8D41
+#define GL_RENDERBUFFER_ALPHA_SIZE 0x8D53
+#define GL_RENDERBUFFER_BINDING 0x8CA7
+#define GL_RENDERBUFFER_BLUE_SIZE 0x8D52
+#define GL_RENDERBUFFER_DEPTH_SIZE 0x8D54
+#define GL_RENDERBUFFER_GREEN_SIZE 0x8D51
+#define GL_RENDERBUFFER_HEIGHT 0x8D43
+#define GL_RENDERBUFFER_INTERNAL_FORMAT 0x8D44
+#define GL_RENDERBUFFER_RED_SIZE 0x8D50
+#define GL_RENDERBUFFER_SAMPLES 0x8CAB
+#define GL_RENDERBUFFER_STENCIL_SIZE 0x8D55
+#define GL_RENDERBUFFER_WIDTH 0x8D42
+#define GL_RENDERER 0x1F01
+#define GL_RENDER_MODE 0x0C40
+#define GL_REPEAT 0x2901
+#define GL_REPLACE 0x1E01
+#define GL_RESCALE_NORMAL 0x803A
+#define GL_RETURN 0x0102
+#define GL_RG 0x8227
+#define GL_RG16 0x822C
+#define GL_RG16F 0x822F
+#define GL_RG16I 0x8239
+#define GL_RG16UI 0x823A
+#define GL_RG16_SNORM 0x8F99
+#define GL_RG32F 0x8230
+#define GL_RG32I 0x823B
+#define GL_RG32UI 0x823C
+#define GL_RG8 0x822B
+#define GL_RG8I 0x8237
+#define GL_RG8UI 0x8238
+#define GL_RG8_SNORM 0x8F95
+#define GL_RGB 0x1907
+#define GL_RGB10 0x8052
+#define GL_RGB10_A2 0x8059
+#define GL_RGB10_A2UI 0x906F
+#define GL_RGB12 0x8053
+#define GL_RGB16 0x8054
+#define GL_RGB16F 0x881B
+#define GL_RGB16I 0x8D89
+#define GL_RGB16UI 0x8D77
+#define GL_RGB16_SNORM 0x8F9A
+#define GL_RGB32F 0x8815
+#define GL_RGB32I 0x8D83
+#define GL_RGB32UI 0x8D71
+#define GL_RGB4 0x804F
+#define GL_RGB5 0x8050
+#define GL_RGB5_A1 0x8057
+#define GL_RGB8 0x8051
+#define GL_RGB8I 0x8D8F
+#define GL_RGB8UI 0x8D7D
+#define GL_RGB8_SNORM 0x8F96
+#define GL_RGB9_E5 0x8C3D
+#define GL_RGBA 0x1908
+#define GL_RGBA12 0x805A
+#define GL_RGBA16 0x805B
+#define GL_RGBA16F 0x881A
+#define GL_RGBA16I 0x8D88
+#define GL_RGBA16UI 0x8D76
+#define GL_RGBA16_SNORM 0x8F9B
+#define GL_RGBA2 0x8055
+#define GL_RGBA32F 0x8814
+#define GL_RGBA32I 0x8D82
+#define GL_RGBA32UI 0x8D70
+#define GL_RGBA4 0x8056
+#define GL_RGBA8 0x8058
+#define GL_RGBA8I 0x8D8E
+#define GL_RGBA8UI 0x8D7C
+#define GL_RGBA8_SNORM 0x8F97
+#define GL_RGBA_INTEGER 0x8D99
+#define GL_RGBA_MODE 0x0C31
+#define GL_RGB_INTEGER 0x8D98
+#define GL_RGB_SCALE 0x8573
+#define GL_RG_INTEGER 0x8228
+#define GL_RIGHT 0x0407
+#define GL_S 0x2000
+#define GL_SAMPLER_1D 0x8B5D
+#define GL_SAMPLER_1D_ARRAY 0x8DC0
+#define GL_SAMPLER_1D_ARRAY_SHADOW 0x8DC3
+#define GL_SAMPLER_1D_SHADOW 0x8B61
+#define GL_SAMPLER_2D 0x8B5E
+#define GL_SAMPLER_2D_ARRAY 0x8DC1
+#define GL_SAMPLER_2D_ARRAY_SHADOW 0x8DC4
+#define GL_SAMPLER_2D_MULTISAMPLE 0x9108
+#define GL_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910B
+#define GL_SAMPLER_2D_RECT 0x8B63
+#define GL_SAMPLER_2D_RECT_SHADOW 0x8B64
+#define GL_SAMPLER_2D_SHADOW 0x8B62
+#define GL_SAMPLER_3D 0x8B5F
+#define GL_SAMPLER_BINDING 0x8919
+#define GL_SAMPLER_BUFFER 0x8DC2
+#define GL_SAMPLER_CUBE 0x8B60
+#define GL_SAMPLER_CUBE_SHADOW 0x8DC5
+#define GL_SAMPLES 0x80A9
+#define GL_SAMPLES_PASSED 0x8914
+#define GL_SAMPLE_ALPHA_TO_COVERAGE 0x809E
+#define GL_SAMPLE_ALPHA_TO_ONE 0x809F
+#define GL_SAMPLE_BUFFERS 0x80A8
+#define GL_SAMPLE_COVERAGE 0x80A0
+#define GL_SAMPLE_COVERAGE_INVERT 0x80AB
+#define GL_SAMPLE_COVERAGE_VALUE 0x80AA
+#define GL_SAMPLE_MASK 0x8E51
+#define GL_SAMPLE_MASK_VALUE 0x8E52
+#define GL_SAMPLE_POSITION 0x8E50
+#define GL_SCISSOR_BIT 0x00080000
+#define GL_SCISSOR_BOX 0x0C10
+#define GL_SCISSOR_TEST 0x0C11
+#define GL_SECONDARY_COLOR_ARRAY 0x845E
+#define GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING 0x889C
+#define GL_SECONDARY_COLOR_ARRAY_POINTER 0x845D
+#define GL_SECONDARY_COLOR_ARRAY_SIZE 0x845A
+#define GL_SECONDARY_COLOR_ARRAY_STRIDE 0x845C
+#define GL_SECONDARY_COLOR_ARRAY_TYPE 0x845B
+#define GL_SELECT 0x1C02
+#define GL_SELECTION_BUFFER_POINTER 0x0DF3
+#define GL_SELECTION_BUFFER_SIZE 0x0DF4
+#define GL_SEPARATE_ATTRIBS 0x8C8D
+#define GL_SEPARATE_SPECULAR_COLOR 0x81FA
+#define GL_SET 0x150F
+#define GL_SHADER_SOURCE_LENGTH 0x8B88
+#define GL_SHADER_TYPE 0x8B4F
+#define GL_SHADE_MODEL 0x0B54
+#define GL_SHADING_LANGUAGE_VERSION 0x8B8C
+#define GL_SHININESS 0x1601
+#define GL_SHORT 0x1402
+#define GL_SIGNALED 0x9119
+#define GL_SIGNED_NORMALIZED 0x8F9C
+#define GL_SINGLE_COLOR 0x81F9
+#define GL_SLUMINANCE 0x8C46
+#define GL_SLUMINANCE8 0x8C47
+#define GL_SLUMINANCE8_ALPHA8 0x8C45
+#define GL_SLUMINANCE_ALPHA 0x8C44
+#define GL_SMOOTH 0x1D01
+#define GL_SMOOTH_LINE_WIDTH_GRANULARITY 0x0B23
+#define GL_SMOOTH_LINE_WIDTH_RANGE 0x0B22
+#define GL_SMOOTH_POINT_SIZE_GRANULARITY 0x0B13
+#define GL_SMOOTH_POINT_SIZE_RANGE 0x0B12
+#define GL_SOURCE0_ALPHA 0x8588
+#define GL_SOURCE0_RGB 0x8580
+#define GL_SOURCE1_ALPHA 0x8589
+#define GL_SOURCE1_RGB 0x8581
+#define GL_SOURCE2_ALPHA 0x858A
+#define GL_SOURCE2_RGB 0x8582
+#define GL_SPECULAR 0x1202
+#define GL_SPHERE_MAP 0x2402
+#define GL_SPOT_CUTOFF 0x1206
+#define GL_SPOT_DIRECTION 0x1204
+#define GL_SPOT_EXPONENT 0x1205
+#define GL_SRC0_ALPHA 0x8588
+#define GL_SRC0_RGB 0x8580
+#define GL_SRC1_ALPHA 0x8589
+#define GL_SRC1_COLOR 0x88F9
+#define GL_SRC1_RGB 0x8581
+#define GL_SRC2_ALPHA 0x858A
+#define GL_SRC2_RGB 0x8582
+#define GL_SRC_ALPHA 0x0302
+#define GL_SRC_ALPHA_SATURATE 0x0308
+#define GL_SRC_COLOR 0x0300
+#define GL_SRGB 0x8C40
+#define GL_SRGB8 0x8C41
+#define GL_SRGB8_ALPHA8 0x8C43
+#define GL_SRGB_ALPHA 0x8C42
+#define GL_STACK_OVERFLOW 0x0503
+#define GL_STACK_UNDERFLOW 0x0504
+#define GL_STATIC_COPY 0x88E6
+#define GL_STATIC_DRAW 0x88E4
+#define GL_STATIC_READ 0x88E5
+#define GL_STENCIL 0x1802
+#define GL_STENCIL_ATTACHMENT 0x8D20
+#define GL_STENCIL_BACK_FAIL 0x8801
+#define GL_STENCIL_BACK_FUNC 0x8800
+#define GL_STENCIL_BACK_PASS_DEPTH_FAIL 0x8802
+#define GL_STENCIL_BACK_PASS_DEPTH_PASS 0x8803
+#define GL_STENCIL_BACK_REF 0x8CA3
+#define GL_STENCIL_BACK_VALUE_MASK 0x8CA4
+#define GL_STENCIL_BACK_WRITEMASK 0x8CA5
+#define GL_STENCIL_BITS 0x0D57
+#define GL_STENCIL_BUFFER_BIT 0x00000400
+#define GL_STENCIL_CLEAR_VALUE 0x0B91
+#define GL_STENCIL_FAIL 0x0B94
+#define GL_STENCIL_FUNC 0x0B92
+#define GL_STENCIL_INDEX 0x1901
+#define GL_STENCIL_INDEX1 0x8D46
+#define GL_STENCIL_INDEX16 0x8D49
+#define GL_STENCIL_INDEX4 0x8D47
+#define GL_STENCIL_INDEX8 0x8D48
+#define GL_STENCIL_PASS_DEPTH_FAIL 0x0B95
+#define GL_STENCIL_PASS_DEPTH_PASS 0x0B96
+#define GL_STENCIL_REF 0x0B97
+#define GL_STENCIL_TEST 0x0B90
+#define GL_STENCIL_VALUE_MASK 0x0B93
+#define GL_STENCIL_WRITEMASK 0x0B98
+#define GL_STEREO 0x0C33
+#define GL_STREAM_COPY 0x88E2
+#define GL_STREAM_DRAW 0x88E0
+#define GL_STREAM_READ 0x88E1
+#define GL_SUBPIXEL_BITS 0x0D50
+#define GL_SUBTRACT 0x84E7
+#define GL_SYNC_CONDITION 0x9113
+#define GL_SYNC_FENCE 0x9116
+#define GL_SYNC_FLAGS 0x9115
+#define GL_SYNC_FLUSH_COMMANDS_BIT 0x00000001
+#define GL_SYNC_GPU_COMMANDS_COMPLETE 0x9117
+#define GL_SYNC_STATUS 0x9114
+#define GL_T 0x2001
+#define GL_T2F_C3F_V3F 0x2A2A
+#define GL_T2F_C4F_N3F_V3F 0x2A2C
+#define GL_T2F_C4UB_V3F 0x2A29
+#define GL_T2F_N3F_V3F 0x2A2B
+#define GL_T2F_V3F 0x2A27
+#define GL_T4F_C4F_N3F_V4F 0x2A2D
+#define GL_T4F_V4F 0x2A28
+#define GL_TEXTURE 0x1702
+#define GL_TEXTURE0 0x84C0
+#define GL_TEXTURE1 0x84C1
+#define GL_TEXTURE10 0x84CA
+#define GL_TEXTURE11 0x84CB
+#define GL_TEXTURE12 0x84CC
+#define GL_TEXTURE13 0x84CD
+#define GL_TEXTURE14 0x84CE
+#define GL_TEXTURE15 0x84CF
+#define GL_TEXTURE16 0x84D0
+#define GL_TEXTURE17 0x84D1
+#define GL_TEXTURE18 0x84D2
+#define GL_TEXTURE19 0x84D3
+#define GL_TEXTURE2 0x84C2
+#define GL_TEXTURE20 0x84D4
+#define GL_TEXTURE21 0x84D5
+#define GL_TEXTURE22 0x84D6
+#define GL_TEXTURE23 0x84D7
+#define GL_TEXTURE24 0x84D8
+#define GL_TEXTURE25 0x84D9
+#define GL_TEXTURE26 0x84DA
+#define GL_TEXTURE27 0x84DB
+#define GL_TEXTURE28 0x84DC
+#define GL_TEXTURE29 0x84DD
+#define GL_TEXTURE3 0x84C3
+#define GL_TEXTURE30 0x84DE
+#define GL_TEXTURE31 0x84DF
+#define GL_TEXTURE4 0x84C4
+#define GL_TEXTURE5 0x84C5
+#define GL_TEXTURE6 0x84C6
+#define GL_TEXTURE7 0x84C7
+#define GL_TEXTURE8 0x84C8
+#define GL_TEXTURE9 0x84C9
+#define GL_TEXTURE_1D 0x0DE0
+#define GL_TEXTURE_1D_ARRAY 0x8C18
+#define GL_TEXTURE_2D 0x0DE1
+#define GL_TEXTURE_2D_ARRAY 0x8C1A
+#define GL_TEXTURE_2D_MULTISAMPLE 0x9100
+#define GL_TEXTURE_2D_MULTISAMPLE_ARRAY 0x9102
+#define GL_TEXTURE_3D 0x806F
+#define GL_TEXTURE_ALPHA_SIZE 0x805F
+#define GL_TEXTURE_ALPHA_TYPE 0x8C13
+#define GL_TEXTURE_BASE_LEVEL 0x813C
+#define GL_TEXTURE_BINDING_1D 0x8068
+#define GL_TEXTURE_BINDING_1D_ARRAY 0x8C1C
+#define GL_TEXTURE_BINDING_2D 0x8069
+#define GL_TEXTURE_BINDING_2D_ARRAY 0x8C1D
+#define GL_TEXTURE_BINDING_2D_MULTISAMPLE 0x9104
+#define GL_TEXTURE_BINDING_2D_MULTISAMPLE_ARRAY 0x9105
+#define GL_TEXTURE_BINDING_3D 0x806A
+#define GL_TEXTURE_BINDING_BUFFER 0x8C2C
+#define GL_TEXTURE_BINDING_CUBE_MAP 0x8514
+#define GL_TEXTURE_BINDING_RECTANGLE 0x84F6
+#define GL_TEXTURE_BIT 0x00040000
+#define GL_TEXTURE_BLUE_SIZE 0x805E
+#define GL_TEXTURE_BLUE_TYPE 0x8C12
+#define GL_TEXTURE_BORDER 0x1005
+#define GL_TEXTURE_BORDER_COLOR 0x1004
+#define GL_TEXTURE_BUFFER 0x8C2A
+#define GL_TEXTURE_BUFFER_DATA_STORE_BINDING 0x8C2D
+#define GL_TEXTURE_COMPARE_FUNC 0x884D
+#define GL_TEXTURE_COMPARE_MODE 0x884C
+#define GL_TEXTURE_COMPONENTS 0x1003
+#define GL_TEXTURE_COMPRESSED 0x86A1
+#define GL_TEXTURE_COMPRESSED_IMAGE_SIZE 0x86A0
+#define GL_TEXTURE_COMPRESSION_HINT 0x84EF
+#define GL_TEXTURE_COORD_ARRAY 0x8078
+#define GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING 0x889A
+#define GL_TEXTURE_COORD_ARRAY_POINTER 0x8092
+#define GL_TEXTURE_COORD_ARRAY_SIZE 0x8088
+#define GL_TEXTURE_COORD_ARRAY_STRIDE 0x808A
+#define GL_TEXTURE_COORD_ARRAY_TYPE 0x8089
+#define GL_TEXTURE_CUBE_MAP 0x8513
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_X 0x8516
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Y 0x8518
+#define GL_TEXTURE_CUBE_MAP_NEGATIVE_Z 0x851A
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_X 0x8515
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Y 0x8517
+#define GL_TEXTURE_CUBE_MAP_POSITIVE_Z 0x8519
+#define GL_TEXTURE_CUBE_MAP_SEAMLESS 0x884F
+#define GL_TEXTURE_DEPTH 0x8071
+#define GL_TEXTURE_DEPTH_SIZE 0x884A
+#define GL_TEXTURE_DEPTH_TYPE 0x8C16
+#define GL_TEXTURE_ENV 0x2300
+#define GL_TEXTURE_ENV_COLOR 0x2201
+#define GL_TEXTURE_ENV_MODE 0x2200
+#define GL_TEXTURE_FILTER_CONTROL 0x8500
+#define GL_TEXTURE_FIXED_SAMPLE_LOCATIONS 0x9107
+#define GL_TEXTURE_GEN_MODE 0x2500
+#define GL_TEXTURE_GEN_Q 0x0C63
+#define GL_TEXTURE_GEN_R 0x0C62
+#define GL_TEXTURE_GEN_S 0x0C60
+#define GL_TEXTURE_GEN_T 0x0C61
+#define GL_TEXTURE_GREEN_SIZE 0x805D
+#define GL_TEXTURE_GREEN_TYPE 0x8C11
+#define GL_TEXTURE_HEIGHT 0x1001
+#define GL_TEXTURE_INTENSITY_SIZE 0x8061
+#define GL_TEXTURE_INTENSITY_TYPE 0x8C15
+#define GL_TEXTURE_INTERNAL_FORMAT 0x1003
+#define GL_TEXTURE_LOD_BIAS 0x8501
+#define GL_TEXTURE_LUMINANCE_SIZE 0x8060
+#define GL_TEXTURE_LUMINANCE_TYPE 0x8C14
+#define GL_TEXTURE_MAG_FILTER 0x2800
+#define GL_TEXTURE_MATRIX 0x0BA8
+#define GL_TEXTURE_MAX_LEVEL 0x813D
+#define GL_TEXTURE_MAX_LOD 0x813B
+#define GL_TEXTURE_MIN_FILTER 0x2801
+#define GL_TEXTURE_MIN_LOD 0x813A
+#define GL_TEXTURE_PRIORITY 0x8066
+#define GL_TEXTURE_RECTANGLE 0x84F5
+#define GL_TEXTURE_RED_SIZE 0x805C
+#define GL_TEXTURE_RED_TYPE 0x8C10
+#define GL_TEXTURE_RESIDENT 0x8067
+#define GL_TEXTURE_SAMPLES 0x9106
+#define GL_TEXTURE_SHARED_SIZE 0x8C3F
+#define GL_TEXTURE_STACK_DEPTH 0x0BA5
+#define GL_TEXTURE_STENCIL_SIZE 0x88F1
+#define GL_TEXTURE_SWIZZLE_A 0x8E45
+#define GL_TEXTURE_SWIZZLE_B 0x8E44
+#define GL_TEXTURE_SWIZZLE_G 0x8E43
+#define GL_TEXTURE_SWIZZLE_R 0x8E42
+#define GL_TEXTURE_SWIZZLE_RGBA 0x8E46
+#define GL_TEXTURE_WIDTH 0x1000
+#define GL_TEXTURE_WRAP_R 0x8072
+#define GL_TEXTURE_WRAP_S 0x2802
+#define GL_TEXTURE_WRAP_T 0x2803
+#define GL_TIMEOUT_EXPIRED 0x911B
+#define GL_TIMEOUT_IGNORED 0xFFFFFFFFFFFFFFFF
+#define GL_TIMESTAMP 0x8E28
+#define GL_TIME_ELAPSED 0x88BF
+#define GL_TRANSFORM_BIT 0x00001000
+#define GL_TRANSFORM_FEEDBACK_BUFFER 0x8C8E
+#define GL_TRANSFORM_FEEDBACK_BUFFER_BINDING 0x8C8F
+#define GL_TRANSFORM_FEEDBACK_BUFFER_MODE 0x8C7F
+#define GL_TRANSFORM_FEEDBACK_BUFFER_SIZE 0x8C85
+#define GL_TRANSFORM_FEEDBACK_BUFFER_START 0x8C84
+#define GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN 0x8C88
+#define GL_TRANSFORM_FEEDBACK_VARYINGS 0x8C83
+#define GL_TRANSFORM_FEEDBACK_VARYING_MAX_LENGTH 0x8C76
+#define GL_TRANSPOSE_COLOR_MATRIX 0x84E6
+#define GL_TRANSPOSE_MODELVIEW_MATRIX 0x84E3
+#define GL_TRANSPOSE_PROJECTION_MATRIX 0x84E4
+#define GL_TRANSPOSE_TEXTURE_MATRIX 0x84E5
+#define GL_TRIANGLES 0x0004
+#define GL_TRIANGLES_ADJACENCY 0x000C
+#define GL_TRIANGLE_FAN 0x0006
+#define GL_TRIANGLE_STRIP 0x0005
+#define GL_TRIANGLE_STRIP_ADJACENCY 0x000D
+#define GL_TRUE 1
+#define GL_UNIFORM_ARRAY_STRIDE 0x8A3C
+#define GL_UNIFORM_BLOCK_ACTIVE_UNIFORMS 0x8A42
+#define GL_UNIFORM_BLOCK_ACTIVE_UNIFORM_INDICES 0x8A43
+#define GL_UNIFORM_BLOCK_BINDING 0x8A3F
+#define GL_UNIFORM_BLOCK_DATA_SIZE 0x8A40
+#define GL_UNIFORM_BLOCK_INDEX 0x8A3A
+#define GL_UNIFORM_BLOCK_NAME_LENGTH 0x8A41
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_FRAGMENT_SHADER 0x8A46
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_GEOMETRY_SHADER 0x8A45
+#define GL_UNIFORM_BLOCK_REFERENCED_BY_VERTEX_SHADER 0x8A44
+#define GL_UNIFORM_BUFFER 0x8A11
+#define GL_UNIFORM_BUFFER_BINDING 0x8A28
+#define GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT 0x8A34
+#define GL_UNIFORM_BUFFER_SIZE 0x8A2A
+#define GL_UNIFORM_BUFFER_START 0x8A29
+#define GL_UNIFORM_IS_ROW_MAJOR 0x8A3E
+#define GL_UNIFORM_MATRIX_STRIDE 0x8A3D
+#define GL_UNIFORM_NAME_LENGTH 0x8A39
+#define GL_UNIFORM_OFFSET 0x8A3B
+#define GL_UNIFORM_SIZE 0x8A38
+#define GL_UNIFORM_TYPE 0x8A37
+#define GL_UNPACK_ALIGNMENT 0x0CF5
+#define GL_UNPACK_IMAGE_HEIGHT 0x806E
+#define GL_UNPACK_LSB_FIRST 0x0CF1
+#define GL_UNPACK_ROW_LENGTH 0x0CF2
+#define GL_UNPACK_SKIP_IMAGES 0x806D
+#define GL_UNPACK_SKIP_PIXELS 0x0CF4
+#define GL_UNPACK_SKIP_ROWS 0x0CF3
+#define GL_UNPACK_SWAP_BYTES 0x0CF0
+#define GL_UNSIGNALED 0x9118
+#define GL_UNSIGNED_BYTE 0x1401
+#define GL_UNSIGNED_BYTE_2_3_3_REV 0x8362
+#define GL_UNSIGNED_BYTE_3_3_2 0x8032
+#define GL_UNSIGNED_INT 0x1405
+#define GL_UNSIGNED_INT_10F_11F_11F_REV 0x8C3B
+#define GL_UNSIGNED_INT_10_10_10_2 0x8036
+#define GL_UNSIGNED_INT_24_8 0x84FA
+#define GL_UNSIGNED_INT_2_10_10_10_REV 0x8368
+#define GL_UNSIGNED_INT_5_9_9_9_REV 0x8C3E
+#define GL_UNSIGNED_INT_8_8_8_8 0x8035
+#define GL_UNSIGNED_INT_8_8_8_8_REV 0x8367
+#define GL_UNSIGNED_INT_SAMPLER_1D 0x8DD1
+#define GL_UNSIGNED_INT_SAMPLER_1D_ARRAY 0x8DD6
+#define GL_UNSIGNED_INT_SAMPLER_2D 0x8DD2
+#define GL_UNSIGNED_INT_SAMPLER_2D_ARRAY 0x8DD7
+#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE 0x910A
+#define GL_UNSIGNED_INT_SAMPLER_2D_MULTISAMPLE_ARRAY 0x910D
+#define GL_UNSIGNED_INT_SAMPLER_2D_RECT 0x8DD5
+#define GL_UNSIGNED_INT_SAMPLER_3D 0x8DD3
+#define GL_UNSIGNED_INT_SAMPLER_BUFFER 0x8DD8
+#define GL_UNSIGNED_INT_SAMPLER_CUBE 0x8DD4
+#define GL_UNSIGNED_INT_VEC2 0x8DC6
+#define GL_UNSIGNED_INT_VEC3 0x8DC7
+#define GL_UNSIGNED_INT_VEC4 0x8DC8
+#define GL_UNSIGNED_NORMALIZED 0x8C17
+#define GL_UNSIGNED_SHORT 0x1403
+#define GL_UNSIGNED_SHORT_1_5_5_5_REV 0x8366
+#define GL_UNSIGNED_SHORT_4_4_4_4 0x8033
+#define GL_UNSIGNED_SHORT_4_4_4_4_REV 0x8365
+#define GL_UNSIGNED_SHORT_5_5_5_1 0x8034
+#define GL_UNSIGNED_SHORT_5_6_5 0x8363
+#define GL_UNSIGNED_SHORT_5_6_5_REV 0x8364
+#define GL_UPPER_LEFT 0x8CA2
+#define GL_V2F 0x2A20
+#define GL_V3F 0x2A21
+#define GL_VALIDATE_STATUS 0x8B83
+#define GL_VENDOR 0x1F00
+#define GL_VERSION 0x1F02
+#define GL_VERTEX_ARRAY 0x8074
+#define GL_VERTEX_ARRAY_BINDING 0x85B5
+#define GL_VERTEX_ARRAY_BUFFER_BINDING 0x8896
+#define GL_VERTEX_ARRAY_POINTER 0x808E
+#define GL_VERTEX_ARRAY_SIZE 0x807A
+#define GL_VERTEX_ARRAY_STRIDE 0x807C
+#define GL_VERTEX_ARRAY_TYPE 0x807B
+#define GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING 0x889F
+#define GL_VERTEX_ATTRIB_ARRAY_DIVISOR 0x88FE
+#define GL_VERTEX_ATTRIB_ARRAY_ENABLED 0x8622
+#define GL_VERTEX_ATTRIB_ARRAY_INTEGER 0x88FD
+#define GL_VERTEX_ATTRIB_ARRAY_NORMALIZED 0x886A
+#define GL_VERTEX_ATTRIB_ARRAY_POINTER 0x8645
+#define GL_VERTEX_ATTRIB_ARRAY_SIZE 0x8623
+#define GL_VERTEX_ATTRIB_ARRAY_STRIDE 0x8624
+#define GL_VERTEX_ATTRIB_ARRAY_TYPE 0x8625
+#define GL_VERTEX_PROGRAM_POINT_SIZE 0x8642
+#define GL_VERTEX_PROGRAM_TWO_SIDE 0x8643
+#define GL_VERTEX_SHADER 0x8B31
+#define GL_VIEWPORT 0x0BA2
+#define GL_VIEWPORT_BIT 0x00000800
+#define GL_WAIT_FAILED 0x911D
+#define GL_WEIGHT_ARRAY_BUFFER_BINDING 0x889E
+#define GL_WRITE_ONLY 0x88B9
+#define GL_XOR 0x1506
+#define GL_ZERO 0
+#define GL_ZOOM_X 0x0D16
+#define GL_ZOOM_Y 0x0D17
+
+
+#include <KHR/khrplatform.h>
+typedef unsigned int GLenum;
+typedef unsigned char GLboolean;
+typedef unsigned int GLbitfield;
+typedef void GLvoid;
+typedef khronos_int8_t GLbyte;
+typedef khronos_uint8_t GLubyte;
+typedef khronos_int16_t GLshort;
+typedef khronos_uint16_t GLushort;
+typedef int GLint;
+typedef unsigned int GLuint;
+typedef khronos_int32_t GLclampx;
+typedef int GLsizei;
+typedef khronos_float_t GLfloat;
+typedef khronos_float_t GLclampf;
+typedef double GLdouble;
+typedef double GLclampd;
+typedef void *GLeglClientBufferEXT;
+typedef void *GLeglImageOES;
+typedef char GLchar;
+typedef char GLcharARB;
+#ifdef __APPLE__
+typedef void *GLhandleARB;
+#else
+typedef unsigned int GLhandleARB;
+#endif
+typedef khronos_uint16_t GLhalf;
+typedef khronos_uint16_t GLhalfARB;
+typedef khronos_int32_t GLfixed;
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptr;
+#else
+typedef khronos_intptr_t GLintptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_intptr_t GLintptrARB;
+#else
+typedef khronos_intptr_t GLintptrARB;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptr;
+#else
+typedef khronos_ssize_t GLsizeiptr;
+#endif
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+typedef khronos_ssize_t GLsizeiptrARB;
+#else
+typedef khronos_ssize_t GLsizeiptrARB;
+#endif
+typedef khronos_int64_t GLint64;
+typedef khronos_int64_t GLint64EXT;
+typedef khronos_uint64_t GLuint64;
+typedef khronos_uint64_t GLuint64EXT;
+typedef struct __GLsync *GLsync;
+struct _cl_context;
+struct _cl_event;
+typedef void (GLAD_API_PTR *GLDEBUGPROC)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCARB)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCKHR)(GLenum source,GLenum type,GLuint id,GLenum severity,GLsizei length,const GLchar *message,const void *userParam);
+typedef void (GLAD_API_PTR *GLDEBUGPROCAMD)(GLuint id,GLenum category,GLenum severity,GLsizei length,const GLchar *message,void *userParam);
+typedef unsigned short GLhalfNV;
+typedef GLintptr GLvdpauSurfaceNV;
+typedef void (GLAD_API_PTR *GLVULKANPROCNV)(void);
+
+
+#define GL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_0;
+#define GL_VERSION_1_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_1;
+#define GL_VERSION_1_2 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_2;
+#define GL_VERSION_1_3 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_3;
+#define GL_VERSION_1_4 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_4;
+#define GL_VERSION_1_5 1
+GLAD_API_CALL int GLAD_GL_VERSION_1_5;
+#define GL_VERSION_2_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_2_0;
+#define GL_VERSION_2_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_2_1;
+#define GL_VERSION_3_0 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_0;
+#define GL_VERSION_3_1 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_1;
+#define GL_VERSION_3_2 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_2;
+#define GL_VERSION_3_3 1
+GLAD_API_CALL int GLAD_GL_VERSION_3_3;
+
+
+typedef void (GLAD_API_PTR *PFNGLACCUMPROC)(GLenum op, GLfloat value);
+typedef void (GLAD_API_PTR *PFNGLACTIVETEXTUREPROC)(GLenum texture);
+typedef void (GLAD_API_PTR *PFNGLALPHAFUNCPROC)(GLenum func, GLfloat ref);
+typedef GLboolean (GLAD_API_PTR *PFNGLARETEXTURESRESIDENTPROC)(GLsizei n, const GLuint * textures, GLboolean * residences);
+typedef void (GLAD_API_PTR *PFNGLARRAYELEMENTPROC)(GLint i);
+typedef void (GLAD_API_PTR *PFNGLATTACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLBEGINPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBEGINCONDITIONALRENDERPROC)(GLuint id, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBEGINQUERYPROC)(GLenum target, GLuint id);
+typedef void (GLAD_API_PTR *PFNGLBEGINTRANSFORMFEEDBACKPROC)(GLenum primitiveMode);
+typedef void (GLAD_API_PTR *PFNGLBINDATTRIBLOCATIONPROC)(GLuint program, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERPROC)(GLenum target, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERBASEPROC)(GLenum target, GLuint index, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLBINDBUFFERRANGEPROC)(GLenum target, GLuint index, GLuint buffer, GLintptr offset, GLsizeiptr size);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAGDATALOCATIONPROC)(GLuint program, GLuint color, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAGDATALOCATIONINDEXEDPROC)(GLuint program, GLuint colorNumber, GLuint index, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLBINDFRAMEBUFFERPROC)(GLenum target, GLuint framebuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDRENDERBUFFERPROC)(GLenum target, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLBINDSAMPLERPROC)(GLuint unit, GLuint sampler);
+typedef void (GLAD_API_PTR *PFNGLBINDTEXTUREPROC)(GLenum target, GLuint texture);
+typedef void (GLAD_API_PTR *PFNGLBINDVERTEXARRAYPROC)(GLuint array);
+typedef void (GLAD_API_PTR *PFNGLBITMAPPROC)(GLsizei width, GLsizei height, GLfloat xorig, GLfloat yorig, GLfloat xmove, GLfloat ymove, const GLubyte * bitmap);
+typedef void (GLAD_API_PTR *PFNGLBLENDCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLBLENDEQUATIONSEPARATEPROC)(GLenum modeRGB, GLenum modeAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCPROC)(GLenum sfactor, GLenum dfactor);
+typedef void (GLAD_API_PTR *PFNGLBLENDFUNCSEPARATEPROC)(GLenum sfactorRGB, GLenum dfactorRGB, GLenum sfactorAlpha, GLenum dfactorAlpha);
+typedef void (GLAD_API_PTR *PFNGLBLITFRAMEBUFFERPROC)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter);
+typedef void (GLAD_API_PTR *PFNGLBUFFERDATAPROC)(GLenum target, GLsizeiptr size, const void * data, GLenum usage);
+typedef void (GLAD_API_PTR *PFNGLBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCALLLISTPROC)(GLuint list);
+typedef void (GLAD_API_PTR *PFNGLCALLLISTSPROC)(GLsizei n, GLenum type, const void * lists);
+typedef GLenum (GLAD_API_PTR *PFNGLCHECKFRAMEBUFFERSTATUSPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLCLAMPCOLORPROC)(GLenum target, GLenum clamp);
+typedef void (GLAD_API_PTR *PFNGLCLEARPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLCLEARACCUMPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERFIPROC)(GLenum buffer, GLint drawbuffer, GLfloat depth, GLint stencil);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERFVPROC)(GLenum buffer, GLint drawbuffer, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERIVPROC)(GLenum buffer, GLint drawbuffer, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARBUFFERUIVPROC)(GLenum buffer, GLint drawbuffer, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLCLEARCOLORPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCLEARDEPTHPROC)(GLdouble depth);
+typedef void (GLAD_API_PTR *PFNGLCLEARINDEXPROC)(GLfloat c);
+typedef void (GLAD_API_PTR *PFNGLCLEARSTENCILPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLCLIENTACTIVETEXTUREPROC)(GLenum texture);
+typedef GLenum (GLAD_API_PTR *PFNGLCLIENTWAITSYNCPROC)(GLsync sync, GLbitfield flags, GLuint64 timeout);
+typedef void (GLAD_API_PTR *PFNGLCLIPPLANEPROC)(GLenum plane, const GLdouble * equation);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3BPROC)(GLbyte red, GLbyte green, GLbyte blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3DPROC)(GLdouble red, GLdouble green, GLdouble blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3FPROC)(GLfloat red, GLfloat green, GLfloat blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3IPROC)(GLint red, GLint green, GLint blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3SPROC)(GLshort red, GLshort green, GLshort blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UBPROC)(GLubyte red, GLubyte green, GLubyte blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UIPROC)(GLuint red, GLuint green, GLuint blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3USPROC)(GLushort red, GLushort green, GLushort blue);
+typedef void (GLAD_API_PTR *PFNGLCOLOR3USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4BPROC)(GLbyte red, GLbyte green, GLbyte blue, GLbyte alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4DPROC)(GLdouble red, GLdouble green, GLdouble blue, GLdouble alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4FPROC)(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4IPROC)(GLint red, GLint green, GLint blue, GLint alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4SPROC)(GLshort red, GLshort green, GLshort blue, GLshort alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UBPROC)(GLubyte red, GLubyte green, GLubyte blue, GLubyte alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UIPROC)(GLuint red, GLuint green, GLuint blue, GLuint alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4USPROC)(GLushort red, GLushort green, GLushort blue, GLushort alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLOR4USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKPROC)(GLboolean red, GLboolean green, GLboolean blue, GLboolean alpha);
+typedef void (GLAD_API_PTR *PFNGLCOLORMASKIPROC)(GLuint index, GLboolean r, GLboolean g, GLboolean b, GLboolean a);
+typedef void (GLAD_API_PTR *PFNGLCOLORMATERIALPROC)(GLenum face, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLCOLORP3UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP3UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP4UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLCOLORP4UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLCOLORPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLCOMPILESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE1DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXIMAGE3DPROC)(GLenum target, GLint level, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLsizei imageSize, const void * data);
+typedef void (GLAD_API_PTR *PFNGLCOPYBUFFERSUBDATAPROC)(GLenum readTarget, GLenum writeTarget, GLintptr readOffset, GLintptr writeOffset, GLsizeiptr size);
+typedef void (GLAD_API_PTR *PFNGLCOPYPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE1DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXIMAGE2DPROC)(GLenum target, GLint level, GLenum internalformat, GLint x, GLint y, GLsizei width, GLsizei height, GLint border);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLCOPYTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATEPROGRAMPROC)(void);
+typedef GLuint (GLAD_API_PTR *PFNGLCREATESHADERPROC)(GLenum type);
+typedef void (GLAD_API_PTR *PFNGLCULLFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLDELETEBUFFERSPROC)(GLsizei n, const GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLDELETEFRAMEBUFFERSPROC)(GLsizei n, const GLuint * framebuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETELISTSPROC)(GLuint list, GLsizei range);
+typedef void (GLAD_API_PTR *PFNGLDELETEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLDELETEQUERIESPROC)(GLsizei n, const GLuint * ids);
+typedef void (GLAD_API_PTR *PFNGLDELETERENDERBUFFERSPROC)(GLsizei n, const GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLDELETESAMPLERSPROC)(GLsizei count, const GLuint * samplers);
+typedef void (GLAD_API_PTR *PFNGLDELETESHADERPROC)(GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDELETESYNCPROC)(GLsync sync);
+typedef void (GLAD_API_PTR *PFNGLDELETETEXTURESPROC)(GLsizei n, const GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLDELETEVERTEXARRAYSPROC)(GLsizei n, const GLuint * arrays);
+typedef void (GLAD_API_PTR *PFNGLDEPTHFUNCPROC)(GLenum func);
+typedef void (GLAD_API_PTR *PFNGLDEPTHMASKPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLDEPTHRANGEPROC)(GLdouble n, GLdouble f);
+typedef void (GLAD_API_PTR *PFNGLDETACHSHADERPROC)(GLuint program, GLuint shader);
+typedef void (GLAD_API_PTR *PFNGLDISABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLDISABLECLIENTSTATEPROC)(GLenum array);
+typedef void (GLAD_API_PTR *PFNGLDISABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDISABLEIPROC)(GLenum target, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSPROC)(GLenum mode, GLint first, GLsizei count);
+typedef void (GLAD_API_PTR *PFNGLDRAWARRAYSINSTANCEDPROC)(GLenum mode, GLint first, GLsizei count, GLsizei instancecount);
+typedef void (GLAD_API_PTR *PFNGLDRAWBUFFERPROC)(GLenum buf);
+typedef void (GLAD_API_PTR *PFNGLDRAWBUFFERSPROC)(GLsizei n, const GLenum * bufs);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSBASEVERTEXPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSINSTANCEDPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLsizei instancecount);
+typedef void (GLAD_API_PTR *PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC)(GLenum mode, GLsizei count, GLenum type, const void * indices, GLsizei instancecount, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLDRAWPIXELSPROC)(GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLDRAWRANGEELEMENTSPROC)(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void * indices);
+typedef void (GLAD_API_PTR *PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC)(GLenum mode, GLuint start, GLuint end, GLsizei count, GLenum type, const void * indices, GLint basevertex);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGPROC)(GLboolean flag);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGPOINTERPROC)(GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLEDGEFLAGVPROC)(const GLboolean * flag);
+typedef void (GLAD_API_PTR *PFNGLENABLEPROC)(GLenum cap);
+typedef void (GLAD_API_PTR *PFNGLENABLECLIENTSTATEPROC)(GLenum array);
+typedef void (GLAD_API_PTR *PFNGLENABLEVERTEXATTRIBARRAYPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLENABLEIPROC)(GLenum target, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLENDPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDCONDITIONALRENDERPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDLISTPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLENDQUERYPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLENDTRANSFORMFEEDBACKPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1DPROC)(GLdouble u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1DVPROC)(const GLdouble * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1FPROC)(GLfloat u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD1FVPROC)(const GLfloat * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2DPROC)(GLdouble u, GLdouble v);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2DVPROC)(const GLdouble * u);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2FPROC)(GLfloat u, GLfloat v);
+typedef void (GLAD_API_PTR *PFNGLEVALCOORD2FVPROC)(const GLfloat * u);
+typedef void (GLAD_API_PTR *PFNGLEVALMESH1PROC)(GLenum mode, GLint i1, GLint i2);
+typedef void (GLAD_API_PTR *PFNGLEVALMESH2PROC)(GLenum mode, GLint i1, GLint i2, GLint j1, GLint j2);
+typedef void (GLAD_API_PTR *PFNGLEVALPOINT1PROC)(GLint i);
+typedef void (GLAD_API_PTR *PFNGLEVALPOINT2PROC)(GLint i, GLint j);
+typedef void (GLAD_API_PTR *PFNGLFEEDBACKBUFFERPROC)(GLsizei size, GLenum type, GLfloat * buffer);
+typedef GLsync (GLAD_API_PTR *PFNGLFENCESYNCPROC)(GLenum condition, GLbitfield flags);
+typedef void (GLAD_API_PTR *PFNGLFINISHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLFLUSHMAPPEDBUFFERRANGEPROC)(GLenum target, GLintptr offset, GLsizeiptr length);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDDPROC)(GLdouble coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDDVPROC)(const GLdouble * coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDFPROC)(GLfloat coord);
+typedef void (GLAD_API_PTR *PFNGLFOGCOORDFVPROC)(const GLfloat * coord);
+typedef void (GLAD_API_PTR *PFNGLFOGFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLFOGFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLFOGIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLFOGIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERRENDERBUFFERPROC)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTUREPROC)(GLenum target, GLenum attachment, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE1DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE2DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURE3DPROC)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset);
+typedef void (GLAD_API_PTR *PFNGLFRAMEBUFFERTEXTURELAYERPROC)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer);
+typedef void (GLAD_API_PTR *PFNGLFRONTFACEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLFRUSTUMPROC)(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble zNear, GLdouble zFar);
+typedef void (GLAD_API_PTR *PFNGLGENBUFFERSPROC)(GLsizei n, GLuint * buffers);
+typedef void (GLAD_API_PTR *PFNGLGENFRAMEBUFFERSPROC)(GLsizei n, GLuint * framebuffers);
+typedef GLuint (GLAD_API_PTR *PFNGLGENLISTSPROC)(GLsizei range);
+typedef void (GLAD_API_PTR *PFNGLGENQUERIESPROC)(GLsizei n, GLuint * ids);
+typedef void (GLAD_API_PTR *PFNGLGENRENDERBUFFERSPROC)(GLsizei n, GLuint * renderbuffers);
+typedef void (GLAD_API_PTR *PFNGLGENSAMPLERSPROC)(GLsizei count, GLuint * samplers);
+typedef void (GLAD_API_PTR *PFNGLGENTEXTURESPROC)(GLsizei n, GLuint * textures);
+typedef void (GLAD_API_PTR *PFNGLGENVERTEXARRAYSPROC)(GLsizei n, GLuint * arrays);
+typedef void (GLAD_API_PTR *PFNGLGENERATEMIPMAPPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEATTRIBPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLint * size, GLenum * type, GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC)(GLuint program, GLuint uniformBlockIndex, GLsizei bufSize, GLsizei * length, GLchar * uniformBlockName);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMBLOCKIVPROC)(GLuint program, GLuint uniformBlockIndex, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMNAMEPROC)(GLuint program, GLuint uniformIndex, GLsizei bufSize, GLsizei * length, GLchar * uniformName);
+typedef void (GLAD_API_PTR *PFNGLGETACTIVEUNIFORMSIVPROC)(GLuint program, GLsizei uniformCount, const GLuint * uniformIndices, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETATTACHEDSHADERSPROC)(GLuint program, GLsizei maxCount, GLsizei * count, GLuint * shaders);
+typedef GLint (GLAD_API_PTR *PFNGLGETATTRIBLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANI_VPROC)(GLenum target, GLuint index, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBOOLEANVPROC)(GLenum pname, GLboolean * data);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERI64VPROC)(GLenum target, GLenum pname, GLint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERPOINTERVPROC)(GLenum target, GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETBUFFERSUBDATAPROC)(GLenum target, GLintptr offset, GLsizeiptr size, void * data);
+typedef void (GLAD_API_PTR *PFNGLGETCLIPPLANEPROC)(GLenum plane, GLdouble * equation);
+typedef void (GLAD_API_PTR *PFNGLGETCOMPRESSEDTEXIMAGEPROC)(GLenum target, GLint level, void * img);
+typedef void (GLAD_API_PTR *PFNGLGETDOUBLEVPROC)(GLenum pname, GLdouble * data);
+typedef GLenum (GLAD_API_PTR *PFNGLGETERRORPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLGETFLOATVPROC)(GLenum pname, GLfloat * data);
+typedef GLint (GLAD_API_PTR *PFNGLGETFRAGDATAINDEXPROC)(GLuint program, const GLchar * name);
+typedef GLint (GLAD_API_PTR *PFNGLGETFRAGDATALOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC)(GLenum target, GLenum attachment, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGER64I_VPROC)(GLenum target, GLuint index, GLint64 * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGER64VPROC)(GLenum pname, GLint64 * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERI_VPROC)(GLenum target, GLuint index, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETINTEGERVPROC)(GLenum pname, GLint * data);
+typedef void (GLAD_API_PTR *PFNGLGETLIGHTFVPROC)(GLenum light, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETLIGHTIVPROC)(GLenum light, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETMAPDVPROC)(GLenum target, GLenum query, GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLGETMAPFVPROC)(GLenum target, GLenum query, GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLGETMAPIVPROC)(GLenum target, GLenum query, GLint * v);
+typedef void (GLAD_API_PTR *PFNGLGETMATERIALFVPROC)(GLenum face, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETMATERIALIVPROC)(GLenum face, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETMULTISAMPLEFVPROC)(GLenum pname, GLuint index, GLfloat * val);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPFVPROC)(GLenum map, GLfloat * values);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPUIVPROC)(GLenum map, GLuint * values);
+typedef void (GLAD_API_PTR *PFNGLGETPIXELMAPUSVPROC)(GLenum map, GLushort * values);
+typedef void (GLAD_API_PTR *PFNGLGETPOINTERVPROC)(GLenum pname, void ** params);
+typedef void (GLAD_API_PTR *PFNGLGETPOLYGONSTIPPLEPROC)(GLubyte * mask);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMINFOLOGPROC)(GLuint program, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETPROGRAMIVPROC)(GLuint program, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTI64VPROC)(GLuint id, GLenum pname, GLint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTIVPROC)(GLuint id, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTUI64VPROC)(GLuint id, GLenum pname, GLuint64 * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYOBJECTUIVPROC)(GLuint id, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETQUERYIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETRENDERBUFFERPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIIVPROC)(GLuint sampler, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIUIVPROC)(GLuint sampler, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERFVPROC)(GLuint sampler, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETSAMPLERPARAMETERIVPROC)(GLuint sampler, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERINFOLOGPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * infoLog);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERSOURCEPROC)(GLuint shader, GLsizei bufSize, GLsizei * length, GLchar * source);
+typedef void (GLAD_API_PTR *PFNGLGETSHADERIVPROC)(GLuint shader, GLenum pname, GLint * params);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGPROC)(GLenum name);
+typedef const GLubyte * (GLAD_API_PTR *PFNGLGETSTRINGIPROC)(GLenum name, GLuint index);
+typedef void (GLAD_API_PTR *PFNGLGETSYNCIVPROC)(GLsync sync, GLenum pname, GLsizei count, GLsizei * length, GLint * values);
+typedef void (GLAD_API_PTR *PFNGLGETTEXENVFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXENVIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENDVPROC)(GLenum coord, GLenum pname, GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENFVPROC)(GLenum coord, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXGENIVPROC)(GLenum coord, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXIMAGEPROC)(GLenum target, GLint level, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLGETTEXLEVELPARAMETERFVPROC)(GLenum target, GLint level, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXLEVELPARAMETERIVPROC)(GLenum target, GLint level, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIUIVPROC)(GLenum target, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERFVPROC)(GLenum target, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETTEXPARAMETERIVPROC)(GLenum target, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETTRANSFORMFEEDBACKVARYINGPROC)(GLuint program, GLuint index, GLsizei bufSize, GLsizei * length, GLsizei * size, GLenum * type, GLchar * name);
+typedef GLuint (GLAD_API_PTR *PFNGLGETUNIFORMBLOCKINDEXPROC)(GLuint program, const GLchar * uniformBlockName);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMINDICESPROC)(GLuint program, GLsizei uniformCount, const GLchar *const* uniformNames, GLuint * uniformIndices);
+typedef GLint (GLAD_API_PTR *PFNGLGETUNIFORMLOCATIONPROC)(GLuint program, const GLchar * name);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMFVPROC)(GLuint program, GLint location, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMIVPROC)(GLuint program, GLint location, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETUNIFORMUIVPROC)(GLuint program, GLint location, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIUIVPROC)(GLuint index, GLenum pname, GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBPOINTERVPROC)(GLuint index, GLenum pname, void ** pointer);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBDVPROC)(GLuint index, GLenum pname, GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBFVPROC)(GLuint index, GLenum pname, GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLGETVERTEXATTRIBIVPROC)(GLuint index, GLenum pname, GLint * params);
+typedef void (GLAD_API_PTR *PFNGLHINTPROC)(GLenum target, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLINDEXMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLINDEXPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLINDEXDPROC)(GLdouble c);
+typedef void (GLAD_API_PTR *PFNGLINDEXDVPROC)(const GLdouble * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXFPROC)(GLfloat c);
+typedef void (GLAD_API_PTR *PFNGLINDEXFVPROC)(const GLfloat * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXIPROC)(GLint c);
+typedef void (GLAD_API_PTR *PFNGLINDEXIVPROC)(const GLint * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXSPROC)(GLshort c);
+typedef void (GLAD_API_PTR *PFNGLINDEXSVPROC)(const GLshort * c);
+typedef void (GLAD_API_PTR *PFNGLINDEXUBPROC)(GLubyte c);
+typedef void (GLAD_API_PTR *PFNGLINDEXUBVPROC)(const GLubyte * c);
+typedef void (GLAD_API_PTR *PFNGLINITNAMESPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLINTERLEAVEDARRAYSPROC)(GLenum format, GLsizei stride, const void * pointer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISBUFFERPROC)(GLuint buffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDPROC)(GLenum cap);
+typedef GLboolean (GLAD_API_PTR *PFNGLISENABLEDIPROC)(GLenum target, GLuint index);
+typedef GLboolean (GLAD_API_PTR *PFNGLISFRAMEBUFFERPROC)(GLuint framebuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISLISTPROC)(GLuint list);
+typedef GLboolean (GLAD_API_PTR *PFNGLISPROGRAMPROC)(GLuint program);
+typedef GLboolean (GLAD_API_PTR *PFNGLISQUERYPROC)(GLuint id);
+typedef GLboolean (GLAD_API_PTR *PFNGLISRENDERBUFFERPROC)(GLuint renderbuffer);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSAMPLERPROC)(GLuint sampler);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSHADERPROC)(GLuint shader);
+typedef GLboolean (GLAD_API_PTR *PFNGLISSYNCPROC)(GLsync sync);
+typedef GLboolean (GLAD_API_PTR *PFNGLISTEXTUREPROC)(GLuint texture);
+typedef GLboolean (GLAD_API_PTR *PFNGLISVERTEXARRAYPROC)(GLuint array);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTMODELIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTFPROC)(GLenum light, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTFVPROC)(GLenum light, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLLIGHTIPROC)(GLenum light, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLLIGHTIVPROC)(GLenum light, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLLINESTIPPLEPROC)(GLint factor, GLushort pattern);
+typedef void (GLAD_API_PTR *PFNGLLINEWIDTHPROC)(GLfloat width);
+typedef void (GLAD_API_PTR *PFNGLLINKPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLLISTBASEPROC)(GLuint base);
+typedef void (GLAD_API_PTR *PFNGLLOADIDENTITYPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLLOADMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLLOADMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLLOADNAMEPROC)(GLuint name);
+typedef void (GLAD_API_PTR *PFNGLLOADTRANSPOSEMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLLOADTRANSPOSEMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLLOGICOPPROC)(GLenum opcode);
+typedef void (GLAD_API_PTR *PFNGLMAP1DPROC)(GLenum target, GLdouble u1, GLdouble u2, GLint stride, GLint order, const GLdouble * points);
+typedef void (GLAD_API_PTR *PFNGLMAP1FPROC)(GLenum target, GLfloat u1, GLfloat u2, GLint stride, GLint order, const GLfloat * points);
+typedef void (GLAD_API_PTR *PFNGLMAP2DPROC)(GLenum target, GLdouble u1, GLdouble u2, GLint ustride, GLint uorder, GLdouble v1, GLdouble v2, GLint vstride, GLint vorder, const GLdouble * points);
+typedef void (GLAD_API_PTR *PFNGLMAP2FPROC)(GLenum target, GLfloat u1, GLfloat u2, GLint ustride, GLint uorder, GLfloat v1, GLfloat v2, GLint vstride, GLint vorder, const GLfloat * points);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERPROC)(GLenum target, GLenum access);
+typedef void * (GLAD_API_PTR *PFNGLMAPBUFFERRANGEPROC)(GLenum target, GLintptr offset, GLsizeiptr length, GLbitfield access);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID1DPROC)(GLint un, GLdouble u1, GLdouble u2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID1FPROC)(GLint un, GLfloat u1, GLfloat u2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID2DPROC)(GLint un, GLdouble u1, GLdouble u2, GLint vn, GLdouble v1, GLdouble v2);
+typedef void (GLAD_API_PTR *PFNGLMAPGRID2FPROC)(GLint un, GLfloat u1, GLfloat u2, GLint vn, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLMATERIALFPROC)(GLenum face, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLMATERIALFVPROC)(GLenum face, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLMATERIALIPROC)(GLenum face, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLMATERIALIVPROC)(GLenum face, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLMATRIXMODEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLMULTMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLMULTMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLMULTTRANSPOSEMATRIXDPROC)(const GLdouble * m);
+typedef void (GLAD_API_PTR *PFNGLMULTTRANSPOSEMATRIXFPROC)(const GLfloat * m);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWARRAYSPROC)(GLenum mode, const GLint * first, const GLsizei * count, GLsizei drawcount);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWELEMENTSPROC)(GLenum mode, const GLsizei * count, GLenum type, const void *const* indices, GLsizei drawcount);
+typedef void (GLAD_API_PTR *PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC)(GLenum mode, const GLsizei * count, GLenum type, const void *const* indices, GLsizei drawcount, const GLint * basevertex);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1DPROC)(GLenum target, GLdouble s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1FPROC)(GLenum target, GLfloat s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1IPROC)(GLenum target, GLint s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1SPROC)(GLenum target, GLshort s);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD1SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2DPROC)(GLenum target, GLdouble s, GLdouble t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2FPROC)(GLenum target, GLfloat s, GLfloat t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2IPROC)(GLenum target, GLint s, GLint t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2SPROC)(GLenum target, GLshort s, GLshort t);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD2SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3DPROC)(GLenum target, GLdouble s, GLdouble t, GLdouble r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3FPROC)(GLenum target, GLfloat s, GLfloat t, GLfloat r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3IPROC)(GLenum target, GLint s, GLint t, GLint r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3SPROC)(GLenum target, GLshort s, GLshort t, GLshort r);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD3SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4DPROC)(GLenum target, GLdouble s, GLdouble t, GLdouble r, GLdouble q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4DVPROC)(GLenum target, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4FPROC)(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4FVPROC)(GLenum target, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4IPROC)(GLenum target, GLint s, GLint t, GLint r, GLint q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4IVPROC)(GLenum target, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4SPROC)(GLenum target, GLshort s, GLshort t, GLshort r, GLshort q);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORD4SVPROC)(GLenum target, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP1UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP1UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP2UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP2UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP3UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP3UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP4UIPROC)(GLenum texture, GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLMULTITEXCOORDP4UIVPROC)(GLenum texture, GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLNEWLISTPROC)(GLuint list, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3BPROC)(GLbyte nx, GLbyte ny, GLbyte nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3DPROC)(GLdouble nx, GLdouble ny, GLdouble nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3FPROC)(GLfloat nx, GLfloat ny, GLfloat nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3IPROC)(GLint nx, GLint ny, GLint nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3SPROC)(GLshort nx, GLshort ny, GLshort nz);
+typedef void (GLAD_API_PTR *PFNGLNORMAL3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLNORMALP3UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLNORMALP3UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLNORMALPOINTERPROC)(GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLORTHOPROC)(GLdouble left, GLdouble right, GLdouble bottom, GLdouble top, GLdouble zNear, GLdouble zFar);
+typedef void (GLAD_API_PTR *PFNGLPASSTHROUGHPROC)(GLfloat token);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPFVPROC)(GLenum map, GLsizei mapsize, const GLfloat * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPUIVPROC)(GLenum map, GLsizei mapsize, const GLuint * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELMAPUSVPROC)(GLenum map, GLsizei mapsize, const GLushort * values);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPIXELSTOREIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPIXELTRANSFERFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPIXELTRANSFERIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPIXELZOOMPROC)(GLfloat xfactor, GLfloat yfactor);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERFPROC)(GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERFVPROC)(GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERIPROC)(GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLPOINTPARAMETERIVPROC)(GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLPOINTSIZEPROC)(GLfloat size);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONMODEPROC)(GLenum face, GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONOFFSETPROC)(GLfloat factor, GLfloat units);
+typedef void (GLAD_API_PTR *PFNGLPOLYGONSTIPPLEPROC)(const GLubyte * mask);
+typedef void (GLAD_API_PTR *PFNGLPOPATTRIBPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPCLIENTATTRIBPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPMATRIXPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPOPNAMEPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPRIMITIVERESTARTINDEXPROC)(GLuint index);
+typedef void (GLAD_API_PTR *PFNGLPRIORITIZETEXTURESPROC)(GLsizei n, const GLuint * textures, const GLfloat * priorities);
+typedef void (GLAD_API_PTR *PFNGLPROVOKINGVERTEXPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLPUSHATTRIBPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLPUSHCLIENTATTRIBPROC)(GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLPUSHMATRIXPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLPUSHNAMEPROC)(GLuint name);
+typedef void (GLAD_API_PTR *PFNGLQUERYCOUNTERPROC)(GLuint id, GLenum target);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4DPROC)(GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4FPROC)(GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4IPROC)(GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4SPROC)(GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLRASTERPOS4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLREADBUFFERPROC)(GLenum src);
+typedef void (GLAD_API_PTR *PFNGLREADPIXELSPROC)(GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, void * pixels);
+typedef void (GLAD_API_PTR *PFNGLRECTDPROC)(GLdouble x1, GLdouble y1, GLdouble x2, GLdouble y2);
+typedef void (GLAD_API_PTR *PFNGLRECTDVPROC)(const GLdouble * v1, const GLdouble * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTFPROC)(GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2);
+typedef void (GLAD_API_PTR *PFNGLRECTFVPROC)(const GLfloat * v1, const GLfloat * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTIPROC)(GLint x1, GLint y1, GLint x2, GLint y2);
+typedef void (GLAD_API_PTR *PFNGLRECTIVPROC)(const GLint * v1, const GLint * v2);
+typedef void (GLAD_API_PTR *PFNGLRECTSPROC)(GLshort x1, GLshort y1, GLshort x2, GLshort y2);
+typedef void (GLAD_API_PTR *PFNGLRECTSVPROC)(const GLshort * v1, const GLshort * v2);
+typedef GLint (GLAD_API_PTR *PFNGLRENDERMODEPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEPROC)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLROTATEDPROC)(GLdouble angle, GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLROTATEFPROC)(GLfloat angle, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLSAMPLECOVERAGEPROC)(GLfloat value, GLboolean invert);
+typedef void (GLAD_API_PTR *PFNGLSAMPLEMASKIPROC)(GLuint maskNumber, GLbitfield mask);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIIVPROC)(GLuint sampler, GLenum pname, const GLint * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIUIVPROC)(GLuint sampler, GLenum pname, const GLuint * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERFPROC)(GLuint sampler, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERFVPROC)(GLuint sampler, GLenum pname, const GLfloat * param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIPROC)(GLuint sampler, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLSAMPLERPARAMETERIVPROC)(GLuint sampler, GLenum pname, const GLint * param);
+typedef void (GLAD_API_PTR *PFNGLSCALEDPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLSCALEFPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLSCISSORPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3BPROC)(GLbyte red, GLbyte green, GLbyte blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3BVPROC)(const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3DPROC)(GLdouble red, GLdouble green, GLdouble blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3FPROC)(GLfloat red, GLfloat green, GLfloat blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3IPROC)(GLint red, GLint green, GLint blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3SPROC)(GLshort red, GLshort green, GLshort blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UBPROC)(GLubyte red, GLubyte green, GLubyte blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UBVPROC)(const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UIPROC)(GLuint red, GLuint green, GLuint blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3UIVPROC)(const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3USPROC)(GLushort red, GLushort green, GLushort blue);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLOR3USVPROC)(const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORP3UIPROC)(GLenum type, GLuint color);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORP3UIVPROC)(GLenum type, const GLuint * color);
+typedef void (GLAD_API_PTR *PFNGLSECONDARYCOLORPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLSELECTBUFFERPROC)(GLsizei size, GLuint * buffer);
+typedef void (GLAD_API_PTR *PFNGLSHADEMODELPROC)(GLenum mode);
+typedef void (GLAD_API_PTR *PFNGLSHADERSOURCEPROC)(GLuint shader, GLsizei count, const GLchar *const* string, const GLint * length);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCPROC)(GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILFUNCSEPARATEPROC)(GLenum face, GLenum func, GLint ref, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKPROC)(GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILMASKSEPARATEPROC)(GLenum face, GLuint mask);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPPROC)(GLenum fail, GLenum zfail, GLenum zpass);
+typedef void (GLAD_API_PTR *PFNGLSTENCILOPSEPARATEPROC)(GLenum face, GLenum sfail, GLenum dpfail, GLenum dppass);
+typedef void (GLAD_API_PTR *PFNGLTEXBUFFERPROC)(GLenum target, GLenum internalformat, GLuint buffer);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1DPROC)(GLdouble s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1FPROC)(GLfloat s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1IPROC)(GLint s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1SPROC)(GLshort s);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD1SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2DPROC)(GLdouble s, GLdouble t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2FPROC)(GLfloat s, GLfloat t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2IPROC)(GLint s, GLint t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2SPROC)(GLshort s, GLshort t);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3DPROC)(GLdouble s, GLdouble t, GLdouble r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3FPROC)(GLfloat s, GLfloat t, GLfloat r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3IPROC)(GLint s, GLint t, GLint r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3SPROC)(GLshort s, GLshort t, GLshort r);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4DPROC)(GLdouble s, GLdouble t, GLdouble r, GLdouble q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4FPROC)(GLfloat s, GLfloat t, GLfloat r, GLfloat q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4IPROC)(GLint s, GLint t, GLint r, GLint q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4SPROC)(GLshort s, GLshort t, GLshort r, GLshort q);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORD4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP1UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP1UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP2UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP2UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP3UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP3UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP4UIPROC)(GLenum type, GLuint coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDP4UIVPROC)(GLenum type, const GLuint * coords);
+typedef void (GLAD_API_PTR *PFNGLTEXCOORDPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLTEXENVFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXENVFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXENVIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXENVIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENDPROC)(GLenum coord, GLenum pname, GLdouble param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENDVPROC)(GLenum coord, GLenum pname, const GLdouble * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENFPROC)(GLenum coord, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENFVPROC)(GLenum coord, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXGENIPROC)(GLenum coord, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXGENIVPROC)(GLenum coord, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE1DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE2DMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLboolean fixedsamplelocations);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE3DPROC)(GLenum target, GLint level, GLint internalformat, GLsizei width, GLsizei height, GLsizei depth, GLint border, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXIMAGE3DMULTISAMPLEPROC)(GLenum target, GLsizei samples, GLenum internalformat, GLsizei width, GLsizei height, GLsizei depth, GLboolean fixedsamplelocations);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIUIVPROC)(GLenum target, GLenum pname, const GLuint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFPROC)(GLenum target, GLenum pname, GLfloat param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERFVPROC)(GLenum target, GLenum pname, const GLfloat * params);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIPROC)(GLenum target, GLenum pname, GLint param);
+typedef void (GLAD_API_PTR *PFNGLTEXPARAMETERIVPROC)(GLenum target, GLenum pname, const GLint * params);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE1DPROC)(GLenum target, GLint level, GLint xoffset, GLsizei width, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE2DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTEXSUBIMAGE3DPROC)(GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const void * pixels);
+typedef void (GLAD_API_PTR *PFNGLTRANSFORMFEEDBACKVARYINGSPROC)(GLuint program, GLsizei count, const GLchar *const* varyings, GLenum bufferMode);
+typedef void (GLAD_API_PTR *PFNGLTRANSLATEDPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLTRANSLATEFPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FPROC)(GLint location, GLfloat v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IPROC)(GLint location, GLint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1UIPROC)(GLint location, GLuint v0);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM1UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FPROC)(GLint location, GLfloat v0, GLfloat v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IPROC)(GLint location, GLint v0, GLint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2UIPROC)(GLint location, GLuint v0, GLuint v1);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM2UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IPROC)(GLint location, GLint v0, GLint v1, GLint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3UIPROC)(GLint location, GLuint v0, GLuint v1, GLuint v2);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM3UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FPROC)(GLint location, GLfloat v0, GLfloat v1, GLfloat v2, GLfloat v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4FVPROC)(GLint location, GLsizei count, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IPROC)(GLint location, GLint v0, GLint v1, GLint v2, GLint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4IVPROC)(GLint location, GLsizei count, const GLint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4UIPROC)(GLint location, GLuint v0, GLuint v1, GLuint v2, GLuint v3);
+typedef void (GLAD_API_PTR *PFNGLUNIFORM4UIVPROC)(GLint location, GLsizei count, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMBLOCKBINDINGPROC)(GLuint program, GLuint uniformBlockIndex, GLuint uniformBlockBinding);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2X3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX2X4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3X2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX3X4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4X2FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef void (GLAD_API_PTR *PFNGLUNIFORMMATRIX4X3FVPROC)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value);
+typedef GLboolean (GLAD_API_PTR *PFNGLUNMAPBUFFERPROC)(GLenum target);
+typedef void (GLAD_API_PTR *PFNGLUSEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVALIDATEPROGRAMPROC)(GLuint program);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLVERTEX2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLVERTEX3SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4DPROC)(GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4FPROC)(GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4IPROC)(GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4SPROC)(GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLVERTEX4SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1DPROC)(GLuint index, GLdouble x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FPROC)(GLuint index, GLfloat x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1SPROC)(GLuint index, GLshort x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB1SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2DPROC)(GLuint index, GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FPROC)(GLuint index, GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2SPROC)(GLuint index, GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB2SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3DPROC)(GLuint index, GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3SPROC)(GLuint index, GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB3SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NBVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NIVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NSVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUBPROC)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4NUSVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4BVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4DPROC)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4DVPROC)(GLuint index, const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FPROC)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4FVPROC)(GLuint index, const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4SPROC)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4UBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIB4USVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBDIVISORPROC)(GLuint index, GLuint divisor);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1IPROC)(GLuint index, GLint x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1UIPROC)(GLuint index, GLuint x);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI1UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2IPROC)(GLuint index, GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2UIPROC)(GLuint index, GLuint x, GLuint y);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI2UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3IPROC)(GLuint index, GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3UIPROC)(GLuint index, GLuint x, GLuint y, GLuint z);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI3UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4BVPROC)(GLuint index, const GLbyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4IPROC)(GLuint index, GLint x, GLint y, GLint z, GLint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4IVPROC)(GLuint index, const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4SVPROC)(GLuint index, const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UBVPROC)(GLuint index, const GLubyte * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UIPROC)(GLuint index, GLuint x, GLuint y, GLuint z, GLuint w);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4UIVPROC)(GLuint index, const GLuint * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBI4USVPROC)(GLuint index, const GLushort * v);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBIPOINTERPROC)(GLuint index, GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP1UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP1UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP2UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP2UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP3UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP3UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP4UIPROC)(GLuint index, GLenum type, GLboolean normalized, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBP4UIVPROC)(GLuint index, GLenum type, GLboolean normalized, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXATTRIBPOINTERPROC)(GLuint index, GLint size, GLenum type, GLboolean normalized, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP2UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP2UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP3UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP3UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP4UIPROC)(GLenum type, GLuint value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXP4UIVPROC)(GLenum type, const GLuint * value);
+typedef void (GLAD_API_PTR *PFNGLVERTEXPOINTERPROC)(GLint size, GLenum type, GLsizei stride, const void * pointer);
+typedef void (GLAD_API_PTR *PFNGLVIEWPORTPROC)(GLint x, GLint y, GLsizei width, GLsizei height);
+typedef void (GLAD_API_PTR *PFNGLWAITSYNCPROC)(GLsync sync, GLbitfield flags, GLuint64 timeout);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2DPROC)(GLdouble x, GLdouble y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2FPROC)(GLfloat x, GLfloat y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2IPROC)(GLint x, GLint y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2SPROC)(GLshort x, GLshort y);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS2SVPROC)(const GLshort * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3DPROC)(GLdouble x, GLdouble y, GLdouble z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3DVPROC)(const GLdouble * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3FPROC)(GLfloat x, GLfloat y, GLfloat z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3FVPROC)(const GLfloat * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3IPROC)(GLint x, GLint y, GLint z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3IVPROC)(const GLint * v);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3SPROC)(GLshort x, GLshort y, GLshort z);
+typedef void (GLAD_API_PTR *PFNGLWINDOWPOS3SVPROC)(const GLshort * v);
+
+GLAD_API_CALL PFNGLACCUMPROC glad_glAccum;
+#define glAccum glad_glAccum
+GLAD_API_CALL PFNGLACTIVETEXTUREPROC glad_glActiveTexture;
+#define glActiveTexture glad_glActiveTexture
+GLAD_API_CALL PFNGLALPHAFUNCPROC glad_glAlphaFunc;
+#define glAlphaFunc glad_glAlphaFunc
+GLAD_API_CALL PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident;
+#define glAreTexturesResident glad_glAreTexturesResident
+GLAD_API_CALL PFNGLARRAYELEMENTPROC glad_glArrayElement;
+#define glArrayElement glad_glArrayElement
+GLAD_API_CALL PFNGLATTACHSHADERPROC glad_glAttachShader;
+#define glAttachShader glad_glAttachShader
+GLAD_API_CALL PFNGLBEGINPROC glad_glBegin;
+#define glBegin glad_glBegin
+GLAD_API_CALL PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender;
+#define glBeginConditionalRender glad_glBeginConditionalRender
+GLAD_API_CALL PFNGLBEGINQUERYPROC glad_glBeginQuery;
+#define glBeginQuery glad_glBeginQuery
+GLAD_API_CALL PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback;
+#define glBeginTransformFeedback glad_glBeginTransformFeedback
+GLAD_API_CALL PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation;
+#define glBindAttribLocation glad_glBindAttribLocation
+GLAD_API_CALL PFNGLBINDBUFFERPROC glad_glBindBuffer;
+#define glBindBuffer glad_glBindBuffer
+GLAD_API_CALL PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase;
+#define glBindBufferBase glad_glBindBufferBase
+GLAD_API_CALL PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange;
+#define glBindBufferRange glad_glBindBufferRange
+GLAD_API_CALL PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation;
+#define glBindFragDataLocation glad_glBindFragDataLocation
+GLAD_API_CALL PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed;
+#define glBindFragDataLocationIndexed glad_glBindFragDataLocationIndexed
+GLAD_API_CALL PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer;
+#define glBindFramebuffer glad_glBindFramebuffer
+GLAD_API_CALL PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer;
+#define glBindRenderbuffer glad_glBindRenderbuffer
+GLAD_API_CALL PFNGLBINDSAMPLERPROC glad_glBindSampler;
+#define glBindSampler glad_glBindSampler
+GLAD_API_CALL PFNGLBINDTEXTUREPROC glad_glBindTexture;
+#define glBindTexture glad_glBindTexture
+GLAD_API_CALL PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray;
+#define glBindVertexArray glad_glBindVertexArray
+GLAD_API_CALL PFNGLBITMAPPROC glad_glBitmap;
+#define glBitmap glad_glBitmap
+GLAD_API_CALL PFNGLBLENDCOLORPROC glad_glBlendColor;
+#define glBlendColor glad_glBlendColor
+GLAD_API_CALL PFNGLBLENDEQUATIONPROC glad_glBlendEquation;
+#define glBlendEquation glad_glBlendEquation
+GLAD_API_CALL PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate;
+#define glBlendEquationSeparate glad_glBlendEquationSeparate
+GLAD_API_CALL PFNGLBLENDFUNCPROC glad_glBlendFunc;
+#define glBlendFunc glad_glBlendFunc
+GLAD_API_CALL PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate;
+#define glBlendFuncSeparate glad_glBlendFuncSeparate
+GLAD_API_CALL PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer;
+#define glBlitFramebuffer glad_glBlitFramebuffer
+GLAD_API_CALL PFNGLBUFFERDATAPROC glad_glBufferData;
+#define glBufferData glad_glBufferData
+GLAD_API_CALL PFNGLBUFFERSUBDATAPROC glad_glBufferSubData;
+#define glBufferSubData glad_glBufferSubData
+GLAD_API_CALL PFNGLCALLLISTPROC glad_glCallList;
+#define glCallList glad_glCallList
+GLAD_API_CALL PFNGLCALLLISTSPROC glad_glCallLists;
+#define glCallLists glad_glCallLists
+GLAD_API_CALL PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus;
+#define glCheckFramebufferStatus glad_glCheckFramebufferStatus
+GLAD_API_CALL PFNGLCLAMPCOLORPROC glad_glClampColor;
+#define glClampColor glad_glClampColor
+GLAD_API_CALL PFNGLCLEARPROC glad_glClear;
+#define glClear glad_glClear
+GLAD_API_CALL PFNGLCLEARACCUMPROC glad_glClearAccum;
+#define glClearAccum glad_glClearAccum
+GLAD_API_CALL PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi;
+#define glClearBufferfi glad_glClearBufferfi
+GLAD_API_CALL PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv;
+#define glClearBufferfv glad_glClearBufferfv
+GLAD_API_CALL PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv;
+#define glClearBufferiv glad_glClearBufferiv
+GLAD_API_CALL PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv;
+#define glClearBufferuiv glad_glClearBufferuiv
+GLAD_API_CALL PFNGLCLEARCOLORPROC glad_glClearColor;
+#define glClearColor glad_glClearColor
+GLAD_API_CALL PFNGLCLEARDEPTHPROC glad_glClearDepth;
+#define glClearDepth glad_glClearDepth
+GLAD_API_CALL PFNGLCLEARINDEXPROC glad_glClearIndex;
+#define glClearIndex glad_glClearIndex
+GLAD_API_CALL PFNGLCLEARSTENCILPROC glad_glClearStencil;
+#define glClearStencil glad_glClearStencil
+GLAD_API_CALL PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture;
+#define glClientActiveTexture glad_glClientActiveTexture
+GLAD_API_CALL PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync;
+#define glClientWaitSync glad_glClientWaitSync
+GLAD_API_CALL PFNGLCLIPPLANEPROC glad_glClipPlane;
+#define glClipPlane glad_glClipPlane
+GLAD_API_CALL PFNGLCOLOR3BPROC glad_glColor3b;
+#define glColor3b glad_glColor3b
+GLAD_API_CALL PFNGLCOLOR3BVPROC glad_glColor3bv;
+#define glColor3bv glad_glColor3bv
+GLAD_API_CALL PFNGLCOLOR3DPROC glad_glColor3d;
+#define glColor3d glad_glColor3d
+GLAD_API_CALL PFNGLCOLOR3DVPROC glad_glColor3dv;
+#define glColor3dv glad_glColor3dv
+GLAD_API_CALL PFNGLCOLOR3FPROC glad_glColor3f;
+#define glColor3f glad_glColor3f
+GLAD_API_CALL PFNGLCOLOR3FVPROC glad_glColor3fv;
+#define glColor3fv glad_glColor3fv
+GLAD_API_CALL PFNGLCOLOR3IPROC glad_glColor3i;
+#define glColor3i glad_glColor3i
+GLAD_API_CALL PFNGLCOLOR3IVPROC glad_glColor3iv;
+#define glColor3iv glad_glColor3iv
+GLAD_API_CALL PFNGLCOLOR3SPROC glad_glColor3s;
+#define glColor3s glad_glColor3s
+GLAD_API_CALL PFNGLCOLOR3SVPROC glad_glColor3sv;
+#define glColor3sv glad_glColor3sv
+GLAD_API_CALL PFNGLCOLOR3UBPROC glad_glColor3ub;
+#define glColor3ub glad_glColor3ub
+GLAD_API_CALL PFNGLCOLOR3UBVPROC glad_glColor3ubv;
+#define glColor3ubv glad_glColor3ubv
+GLAD_API_CALL PFNGLCOLOR3UIPROC glad_glColor3ui;
+#define glColor3ui glad_glColor3ui
+GLAD_API_CALL PFNGLCOLOR3UIVPROC glad_glColor3uiv;
+#define glColor3uiv glad_glColor3uiv
+GLAD_API_CALL PFNGLCOLOR3USPROC glad_glColor3us;
+#define glColor3us glad_glColor3us
+GLAD_API_CALL PFNGLCOLOR3USVPROC glad_glColor3usv;
+#define glColor3usv glad_glColor3usv
+GLAD_API_CALL PFNGLCOLOR4BPROC glad_glColor4b;
+#define glColor4b glad_glColor4b
+GLAD_API_CALL PFNGLCOLOR4BVPROC glad_glColor4bv;
+#define glColor4bv glad_glColor4bv
+GLAD_API_CALL PFNGLCOLOR4DPROC glad_glColor4d;
+#define glColor4d glad_glColor4d
+GLAD_API_CALL PFNGLCOLOR4DVPROC glad_glColor4dv;
+#define glColor4dv glad_glColor4dv
+GLAD_API_CALL PFNGLCOLOR4FPROC glad_glColor4f;
+#define glColor4f glad_glColor4f
+GLAD_API_CALL PFNGLCOLOR4FVPROC glad_glColor4fv;
+#define glColor4fv glad_glColor4fv
+GLAD_API_CALL PFNGLCOLOR4IPROC glad_glColor4i;
+#define glColor4i glad_glColor4i
+GLAD_API_CALL PFNGLCOLOR4IVPROC glad_glColor4iv;
+#define glColor4iv glad_glColor4iv
+GLAD_API_CALL PFNGLCOLOR4SPROC glad_glColor4s;
+#define glColor4s glad_glColor4s
+GLAD_API_CALL PFNGLCOLOR4SVPROC glad_glColor4sv;
+#define glColor4sv glad_glColor4sv
+GLAD_API_CALL PFNGLCOLOR4UBPROC glad_glColor4ub;
+#define glColor4ub glad_glColor4ub
+GLAD_API_CALL PFNGLCOLOR4UBVPROC glad_glColor4ubv;
+#define glColor4ubv glad_glColor4ubv
+GLAD_API_CALL PFNGLCOLOR4UIPROC glad_glColor4ui;
+#define glColor4ui glad_glColor4ui
+GLAD_API_CALL PFNGLCOLOR4UIVPROC glad_glColor4uiv;
+#define glColor4uiv glad_glColor4uiv
+GLAD_API_CALL PFNGLCOLOR4USPROC glad_glColor4us;
+#define glColor4us glad_glColor4us
+GLAD_API_CALL PFNGLCOLOR4USVPROC glad_glColor4usv;
+#define glColor4usv glad_glColor4usv
+GLAD_API_CALL PFNGLCOLORMASKPROC glad_glColorMask;
+#define glColorMask glad_glColorMask
+GLAD_API_CALL PFNGLCOLORMASKIPROC glad_glColorMaski;
+#define glColorMaski glad_glColorMaski
+GLAD_API_CALL PFNGLCOLORMATERIALPROC glad_glColorMaterial;
+#define glColorMaterial glad_glColorMaterial
+GLAD_API_CALL PFNGLCOLORP3UIPROC glad_glColorP3ui;
+#define glColorP3ui glad_glColorP3ui
+GLAD_API_CALL PFNGLCOLORP3UIVPROC glad_glColorP3uiv;
+#define glColorP3uiv glad_glColorP3uiv
+GLAD_API_CALL PFNGLCOLORP4UIPROC glad_glColorP4ui;
+#define glColorP4ui glad_glColorP4ui
+GLAD_API_CALL PFNGLCOLORP4UIVPROC glad_glColorP4uiv;
+#define glColorP4uiv glad_glColorP4uiv
+GLAD_API_CALL PFNGLCOLORPOINTERPROC glad_glColorPointer;
+#define glColorPointer glad_glColorPointer
+GLAD_API_CALL PFNGLCOMPILESHADERPROC glad_glCompileShader;
+#define glCompileShader glad_glCompileShader
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D;
+#define glCompressedTexImage1D glad_glCompressedTexImage1D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D;
+#define glCompressedTexImage2D glad_glCompressedTexImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D;
+#define glCompressedTexImage3D glad_glCompressedTexImage3D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D;
+#define glCompressedTexSubImage1D glad_glCompressedTexSubImage1D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D;
+#define glCompressedTexSubImage2D glad_glCompressedTexSubImage2D
+GLAD_API_CALL PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D;
+#define glCompressedTexSubImage3D glad_glCompressedTexSubImage3D
+GLAD_API_CALL PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData;
+#define glCopyBufferSubData glad_glCopyBufferSubData
+GLAD_API_CALL PFNGLCOPYPIXELSPROC glad_glCopyPixels;
+#define glCopyPixels glad_glCopyPixels
+GLAD_API_CALL PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D;
+#define glCopyTexImage1D glad_glCopyTexImage1D
+GLAD_API_CALL PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D;
+#define glCopyTexImage2D glad_glCopyTexImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D;
+#define glCopyTexSubImage1D glad_glCopyTexSubImage1D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D;
+#define glCopyTexSubImage2D glad_glCopyTexSubImage2D
+GLAD_API_CALL PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D;
+#define glCopyTexSubImage3D glad_glCopyTexSubImage3D
+GLAD_API_CALL PFNGLCREATEPROGRAMPROC glad_glCreateProgram;
+#define glCreateProgram glad_glCreateProgram
+GLAD_API_CALL PFNGLCREATESHADERPROC glad_glCreateShader;
+#define glCreateShader glad_glCreateShader
+GLAD_API_CALL PFNGLCULLFACEPROC glad_glCullFace;
+#define glCullFace glad_glCullFace
+GLAD_API_CALL PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers;
+#define glDeleteBuffers glad_glDeleteBuffers
+GLAD_API_CALL PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers;
+#define glDeleteFramebuffers glad_glDeleteFramebuffers
+GLAD_API_CALL PFNGLDELETELISTSPROC glad_glDeleteLists;
+#define glDeleteLists glad_glDeleteLists
+GLAD_API_CALL PFNGLDELETEPROGRAMPROC glad_glDeleteProgram;
+#define glDeleteProgram glad_glDeleteProgram
+GLAD_API_CALL PFNGLDELETEQUERIESPROC glad_glDeleteQueries;
+#define glDeleteQueries glad_glDeleteQueries
+GLAD_API_CALL PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers;
+#define glDeleteRenderbuffers glad_glDeleteRenderbuffers
+GLAD_API_CALL PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers;
+#define glDeleteSamplers glad_glDeleteSamplers
+GLAD_API_CALL PFNGLDELETESHADERPROC glad_glDeleteShader;
+#define glDeleteShader glad_glDeleteShader
+GLAD_API_CALL PFNGLDELETESYNCPROC glad_glDeleteSync;
+#define glDeleteSync glad_glDeleteSync
+GLAD_API_CALL PFNGLDELETETEXTURESPROC glad_glDeleteTextures;
+#define glDeleteTextures glad_glDeleteTextures
+GLAD_API_CALL PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays;
+#define glDeleteVertexArrays glad_glDeleteVertexArrays
+GLAD_API_CALL PFNGLDEPTHFUNCPROC glad_glDepthFunc;
+#define glDepthFunc glad_glDepthFunc
+GLAD_API_CALL PFNGLDEPTHMASKPROC glad_glDepthMask;
+#define glDepthMask glad_glDepthMask
+GLAD_API_CALL PFNGLDEPTHRANGEPROC glad_glDepthRange;
+#define glDepthRange glad_glDepthRange
+GLAD_API_CALL PFNGLDETACHSHADERPROC glad_glDetachShader;
+#define glDetachShader glad_glDetachShader
+GLAD_API_CALL PFNGLDISABLEPROC glad_glDisable;
+#define glDisable glad_glDisable
+GLAD_API_CALL PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState;
+#define glDisableClientState glad_glDisableClientState
+GLAD_API_CALL PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray;
+#define glDisableVertexAttribArray glad_glDisableVertexAttribArray
+GLAD_API_CALL PFNGLDISABLEIPROC glad_glDisablei;
+#define glDisablei glad_glDisablei
+GLAD_API_CALL PFNGLDRAWARRAYSPROC glad_glDrawArrays;
+#define glDrawArrays glad_glDrawArrays
+GLAD_API_CALL PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced;
+#define glDrawArraysInstanced glad_glDrawArraysInstanced
+GLAD_API_CALL PFNGLDRAWBUFFERPROC glad_glDrawBuffer;
+#define glDrawBuffer glad_glDrawBuffer
+GLAD_API_CALL PFNGLDRAWBUFFERSPROC glad_glDrawBuffers;
+#define glDrawBuffers glad_glDrawBuffers
+GLAD_API_CALL PFNGLDRAWELEMENTSPROC glad_glDrawElements;
+#define glDrawElements glad_glDrawElements
+GLAD_API_CALL PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex;
+#define glDrawElementsBaseVertex glad_glDrawElementsBaseVertex
+GLAD_API_CALL PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced;
+#define glDrawElementsInstanced glad_glDrawElementsInstanced
+GLAD_API_CALL PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex;
+#define glDrawElementsInstancedBaseVertex glad_glDrawElementsInstancedBaseVertex
+GLAD_API_CALL PFNGLDRAWPIXELSPROC glad_glDrawPixels;
+#define glDrawPixels glad_glDrawPixels
+GLAD_API_CALL PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements;
+#define glDrawRangeElements glad_glDrawRangeElements
+GLAD_API_CALL PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex;
+#define glDrawRangeElementsBaseVertex glad_glDrawRangeElementsBaseVertex
+GLAD_API_CALL PFNGLEDGEFLAGPROC glad_glEdgeFlag;
+#define glEdgeFlag glad_glEdgeFlag
+GLAD_API_CALL PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer;
+#define glEdgeFlagPointer glad_glEdgeFlagPointer
+GLAD_API_CALL PFNGLEDGEFLAGVPROC glad_glEdgeFlagv;
+#define glEdgeFlagv glad_glEdgeFlagv
+GLAD_API_CALL PFNGLENABLEPROC glad_glEnable;
+#define glEnable glad_glEnable
+GLAD_API_CALL PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState;
+#define glEnableClientState glad_glEnableClientState
+GLAD_API_CALL PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray;
+#define glEnableVertexAttribArray glad_glEnableVertexAttribArray
+GLAD_API_CALL PFNGLENABLEIPROC glad_glEnablei;
+#define glEnablei glad_glEnablei
+GLAD_API_CALL PFNGLENDPROC glad_glEnd;
+#define glEnd glad_glEnd
+GLAD_API_CALL PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender;
+#define glEndConditionalRender glad_glEndConditionalRender
+GLAD_API_CALL PFNGLENDLISTPROC glad_glEndList;
+#define glEndList glad_glEndList
+GLAD_API_CALL PFNGLENDQUERYPROC glad_glEndQuery;
+#define glEndQuery glad_glEndQuery
+GLAD_API_CALL PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback;
+#define glEndTransformFeedback glad_glEndTransformFeedback
+GLAD_API_CALL PFNGLEVALCOORD1DPROC glad_glEvalCoord1d;
+#define glEvalCoord1d glad_glEvalCoord1d
+GLAD_API_CALL PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv;
+#define glEvalCoord1dv glad_glEvalCoord1dv
+GLAD_API_CALL PFNGLEVALCOORD1FPROC glad_glEvalCoord1f;
+#define glEvalCoord1f glad_glEvalCoord1f
+GLAD_API_CALL PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv;
+#define glEvalCoord1fv glad_glEvalCoord1fv
+GLAD_API_CALL PFNGLEVALCOORD2DPROC glad_glEvalCoord2d;
+#define glEvalCoord2d glad_glEvalCoord2d
+GLAD_API_CALL PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv;
+#define glEvalCoord2dv glad_glEvalCoord2dv
+GLAD_API_CALL PFNGLEVALCOORD2FPROC glad_glEvalCoord2f;
+#define glEvalCoord2f glad_glEvalCoord2f
+GLAD_API_CALL PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv;
+#define glEvalCoord2fv glad_glEvalCoord2fv
+GLAD_API_CALL PFNGLEVALMESH1PROC glad_glEvalMesh1;
+#define glEvalMesh1 glad_glEvalMesh1
+GLAD_API_CALL PFNGLEVALMESH2PROC glad_glEvalMesh2;
+#define glEvalMesh2 glad_glEvalMesh2
+GLAD_API_CALL PFNGLEVALPOINT1PROC glad_glEvalPoint1;
+#define glEvalPoint1 glad_glEvalPoint1
+GLAD_API_CALL PFNGLEVALPOINT2PROC glad_glEvalPoint2;
+#define glEvalPoint2 glad_glEvalPoint2
+GLAD_API_CALL PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer;
+#define glFeedbackBuffer glad_glFeedbackBuffer
+GLAD_API_CALL PFNGLFENCESYNCPROC glad_glFenceSync;
+#define glFenceSync glad_glFenceSync
+GLAD_API_CALL PFNGLFINISHPROC glad_glFinish;
+#define glFinish glad_glFinish
+GLAD_API_CALL PFNGLFLUSHPROC glad_glFlush;
+#define glFlush glad_glFlush
+GLAD_API_CALL PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange;
+#define glFlushMappedBufferRange glad_glFlushMappedBufferRange
+GLAD_API_CALL PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer;
+#define glFogCoordPointer glad_glFogCoordPointer
+GLAD_API_CALL PFNGLFOGCOORDDPROC glad_glFogCoordd;
+#define glFogCoordd glad_glFogCoordd
+GLAD_API_CALL PFNGLFOGCOORDDVPROC glad_glFogCoorddv;
+#define glFogCoorddv glad_glFogCoorddv
+GLAD_API_CALL PFNGLFOGCOORDFPROC glad_glFogCoordf;
+#define glFogCoordf glad_glFogCoordf
+GLAD_API_CALL PFNGLFOGCOORDFVPROC glad_glFogCoordfv;
+#define glFogCoordfv glad_glFogCoordfv
+GLAD_API_CALL PFNGLFOGFPROC glad_glFogf;
+#define glFogf glad_glFogf
+GLAD_API_CALL PFNGLFOGFVPROC glad_glFogfv;
+#define glFogfv glad_glFogfv
+GLAD_API_CALL PFNGLFOGIPROC glad_glFogi;
+#define glFogi glad_glFogi
+GLAD_API_CALL PFNGLFOGIVPROC glad_glFogiv;
+#define glFogiv glad_glFogiv
+GLAD_API_CALL PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer;
+#define glFramebufferRenderbuffer glad_glFramebufferRenderbuffer
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture;
+#define glFramebufferTexture glad_glFramebufferTexture
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D;
+#define glFramebufferTexture1D glad_glFramebufferTexture1D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D;
+#define glFramebufferTexture2D glad_glFramebufferTexture2D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D;
+#define glFramebufferTexture3D glad_glFramebufferTexture3D
+GLAD_API_CALL PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer;
+#define glFramebufferTextureLayer glad_glFramebufferTextureLayer
+GLAD_API_CALL PFNGLFRONTFACEPROC glad_glFrontFace;
+#define glFrontFace glad_glFrontFace
+GLAD_API_CALL PFNGLFRUSTUMPROC glad_glFrustum;
+#define glFrustum glad_glFrustum
+GLAD_API_CALL PFNGLGENBUFFERSPROC glad_glGenBuffers;
+#define glGenBuffers glad_glGenBuffers
+GLAD_API_CALL PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers;
+#define glGenFramebuffers glad_glGenFramebuffers
+GLAD_API_CALL PFNGLGENLISTSPROC glad_glGenLists;
+#define glGenLists glad_glGenLists
+GLAD_API_CALL PFNGLGENQUERIESPROC glad_glGenQueries;
+#define glGenQueries glad_glGenQueries
+GLAD_API_CALL PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers;
+#define glGenRenderbuffers glad_glGenRenderbuffers
+GLAD_API_CALL PFNGLGENSAMPLERSPROC glad_glGenSamplers;
+#define glGenSamplers glad_glGenSamplers
+GLAD_API_CALL PFNGLGENTEXTURESPROC glad_glGenTextures;
+#define glGenTextures glad_glGenTextures
+GLAD_API_CALL PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays;
+#define glGenVertexArrays glad_glGenVertexArrays
+GLAD_API_CALL PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap;
+#define glGenerateMipmap glad_glGenerateMipmap
+GLAD_API_CALL PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib;
+#define glGetActiveAttrib glad_glGetActiveAttrib
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform;
+#define glGetActiveUniform glad_glGetActiveUniform
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName;
+#define glGetActiveUniformBlockName glad_glGetActiveUniformBlockName
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv;
+#define glGetActiveUniformBlockiv glad_glGetActiveUniformBlockiv
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName;
+#define glGetActiveUniformName glad_glGetActiveUniformName
+GLAD_API_CALL PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv;
+#define glGetActiveUniformsiv glad_glGetActiveUniformsiv
+GLAD_API_CALL PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders;
+#define glGetAttachedShaders glad_glGetAttachedShaders
+GLAD_API_CALL PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation;
+#define glGetAttribLocation glad_glGetAttribLocation
+GLAD_API_CALL PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v;
+#define glGetBooleani_v glad_glGetBooleani_v
+GLAD_API_CALL PFNGLGETBOOLEANVPROC glad_glGetBooleanv;
+#define glGetBooleanv glad_glGetBooleanv
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v;
+#define glGetBufferParameteri64v glad_glGetBufferParameteri64v
+GLAD_API_CALL PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv;
+#define glGetBufferParameteriv glad_glGetBufferParameteriv
+GLAD_API_CALL PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv;
+#define glGetBufferPointerv glad_glGetBufferPointerv
+GLAD_API_CALL PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData;
+#define glGetBufferSubData glad_glGetBufferSubData
+GLAD_API_CALL PFNGLGETCLIPPLANEPROC glad_glGetClipPlane;
+#define glGetClipPlane glad_glGetClipPlane
+GLAD_API_CALL PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage;
+#define glGetCompressedTexImage glad_glGetCompressedTexImage
+GLAD_API_CALL PFNGLGETDOUBLEVPROC glad_glGetDoublev;
+#define glGetDoublev glad_glGetDoublev
+GLAD_API_CALL PFNGLGETERRORPROC glad_glGetError;
+#define glGetError glad_glGetError
+GLAD_API_CALL PFNGLGETFLOATVPROC glad_glGetFloatv;
+#define glGetFloatv glad_glGetFloatv
+GLAD_API_CALL PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex;
+#define glGetFragDataIndex glad_glGetFragDataIndex
+GLAD_API_CALL PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation;
+#define glGetFragDataLocation glad_glGetFragDataLocation
+GLAD_API_CALL PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv;
+#define glGetFramebufferAttachmentParameteriv glad_glGetFramebufferAttachmentParameteriv
+GLAD_API_CALL PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v;
+#define glGetInteger64i_v glad_glGetInteger64i_v
+GLAD_API_CALL PFNGLGETINTEGER64VPROC glad_glGetInteger64v;
+#define glGetInteger64v glad_glGetInteger64v
+GLAD_API_CALL PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v;
+#define glGetIntegeri_v glad_glGetIntegeri_v
+GLAD_API_CALL PFNGLGETINTEGERVPROC glad_glGetIntegerv;
+#define glGetIntegerv glad_glGetIntegerv
+GLAD_API_CALL PFNGLGETLIGHTFVPROC glad_glGetLightfv;
+#define glGetLightfv glad_glGetLightfv
+GLAD_API_CALL PFNGLGETLIGHTIVPROC glad_glGetLightiv;
+#define glGetLightiv glad_glGetLightiv
+GLAD_API_CALL PFNGLGETMAPDVPROC glad_glGetMapdv;
+#define glGetMapdv glad_glGetMapdv
+GLAD_API_CALL PFNGLGETMAPFVPROC glad_glGetMapfv;
+#define glGetMapfv glad_glGetMapfv
+GLAD_API_CALL PFNGLGETMAPIVPROC glad_glGetMapiv;
+#define glGetMapiv glad_glGetMapiv
+GLAD_API_CALL PFNGLGETMATERIALFVPROC glad_glGetMaterialfv;
+#define glGetMaterialfv glad_glGetMaterialfv
+GLAD_API_CALL PFNGLGETMATERIALIVPROC glad_glGetMaterialiv;
+#define glGetMaterialiv glad_glGetMaterialiv
+GLAD_API_CALL PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv;
+#define glGetMultisamplefv glad_glGetMultisamplefv
+GLAD_API_CALL PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv;
+#define glGetPixelMapfv glad_glGetPixelMapfv
+GLAD_API_CALL PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv;
+#define glGetPixelMapuiv glad_glGetPixelMapuiv
+GLAD_API_CALL PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv;
+#define glGetPixelMapusv glad_glGetPixelMapusv
+GLAD_API_CALL PFNGLGETPOINTERVPROC glad_glGetPointerv;
+#define glGetPointerv glad_glGetPointerv
+GLAD_API_CALL PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple;
+#define glGetPolygonStipple glad_glGetPolygonStipple
+GLAD_API_CALL PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog;
+#define glGetProgramInfoLog glad_glGetProgramInfoLog
+GLAD_API_CALL PFNGLGETPROGRAMIVPROC glad_glGetProgramiv;
+#define glGetProgramiv glad_glGetProgramiv
+GLAD_API_CALL PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v;
+#define glGetQueryObjecti64v glad_glGetQueryObjecti64v
+GLAD_API_CALL PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv;
+#define glGetQueryObjectiv glad_glGetQueryObjectiv
+GLAD_API_CALL PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v;
+#define glGetQueryObjectui64v glad_glGetQueryObjectui64v
+GLAD_API_CALL PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv;
+#define glGetQueryObjectuiv glad_glGetQueryObjectuiv
+GLAD_API_CALL PFNGLGETQUERYIVPROC glad_glGetQueryiv;
+#define glGetQueryiv glad_glGetQueryiv
+GLAD_API_CALL PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv;
+#define glGetRenderbufferParameteriv glad_glGetRenderbufferParameteriv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv;
+#define glGetSamplerParameterIiv glad_glGetSamplerParameterIiv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv;
+#define glGetSamplerParameterIuiv glad_glGetSamplerParameterIuiv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv;
+#define glGetSamplerParameterfv glad_glGetSamplerParameterfv
+GLAD_API_CALL PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv;
+#define glGetSamplerParameteriv glad_glGetSamplerParameteriv
+GLAD_API_CALL PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog;
+#define glGetShaderInfoLog glad_glGetShaderInfoLog
+GLAD_API_CALL PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource;
+#define glGetShaderSource glad_glGetShaderSource
+GLAD_API_CALL PFNGLGETSHADERIVPROC glad_glGetShaderiv;
+#define glGetShaderiv glad_glGetShaderiv
+GLAD_API_CALL PFNGLGETSTRINGPROC glad_glGetString;
+#define glGetString glad_glGetString
+GLAD_API_CALL PFNGLGETSTRINGIPROC glad_glGetStringi;
+#define glGetStringi glad_glGetStringi
+GLAD_API_CALL PFNGLGETSYNCIVPROC glad_glGetSynciv;
+#define glGetSynciv glad_glGetSynciv
+GLAD_API_CALL PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv;
+#define glGetTexEnvfv glad_glGetTexEnvfv
+GLAD_API_CALL PFNGLGETTEXENVIVPROC glad_glGetTexEnviv;
+#define glGetTexEnviv glad_glGetTexEnviv
+GLAD_API_CALL PFNGLGETTEXGENDVPROC glad_glGetTexGendv;
+#define glGetTexGendv glad_glGetTexGendv
+GLAD_API_CALL PFNGLGETTEXGENFVPROC glad_glGetTexGenfv;
+#define glGetTexGenfv glad_glGetTexGenfv
+GLAD_API_CALL PFNGLGETTEXGENIVPROC glad_glGetTexGeniv;
+#define glGetTexGeniv glad_glGetTexGeniv
+GLAD_API_CALL PFNGLGETTEXIMAGEPROC glad_glGetTexImage;
+#define glGetTexImage glad_glGetTexImage
+GLAD_API_CALL PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv;
+#define glGetTexLevelParameterfv glad_glGetTexLevelParameterfv
+GLAD_API_CALL PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv;
+#define glGetTexLevelParameteriv glad_glGetTexLevelParameteriv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv;
+#define glGetTexParameterIiv glad_glGetTexParameterIiv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv;
+#define glGetTexParameterIuiv glad_glGetTexParameterIuiv
+GLAD_API_CALL PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv;
+#define glGetTexParameterfv glad_glGetTexParameterfv
+GLAD_API_CALL PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv;
+#define glGetTexParameteriv glad_glGetTexParameteriv
+GLAD_API_CALL PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying;
+#define glGetTransformFeedbackVarying glad_glGetTransformFeedbackVarying
+GLAD_API_CALL PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex;
+#define glGetUniformBlockIndex glad_glGetUniformBlockIndex
+GLAD_API_CALL PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices;
+#define glGetUniformIndices glad_glGetUniformIndices
+GLAD_API_CALL PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation;
+#define glGetUniformLocation glad_glGetUniformLocation
+GLAD_API_CALL PFNGLGETUNIFORMFVPROC glad_glGetUniformfv;
+#define glGetUniformfv glad_glGetUniformfv
+GLAD_API_CALL PFNGLGETUNIFORMIVPROC glad_glGetUniformiv;
+#define glGetUniformiv glad_glGetUniformiv
+GLAD_API_CALL PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv;
+#define glGetUniformuiv glad_glGetUniformuiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv;
+#define glGetVertexAttribIiv glad_glGetVertexAttribIiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv;
+#define glGetVertexAttribIuiv glad_glGetVertexAttribIuiv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv;
+#define glGetVertexAttribPointerv glad_glGetVertexAttribPointerv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv;
+#define glGetVertexAttribdv glad_glGetVertexAttribdv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv;
+#define glGetVertexAttribfv glad_glGetVertexAttribfv
+GLAD_API_CALL PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv;
+#define glGetVertexAttribiv glad_glGetVertexAttribiv
+GLAD_API_CALL PFNGLHINTPROC glad_glHint;
+#define glHint glad_glHint
+GLAD_API_CALL PFNGLINDEXMASKPROC glad_glIndexMask;
+#define glIndexMask glad_glIndexMask
+GLAD_API_CALL PFNGLINDEXPOINTERPROC glad_glIndexPointer;
+#define glIndexPointer glad_glIndexPointer
+GLAD_API_CALL PFNGLINDEXDPROC glad_glIndexd;
+#define glIndexd glad_glIndexd
+GLAD_API_CALL PFNGLINDEXDVPROC glad_glIndexdv;
+#define glIndexdv glad_glIndexdv
+GLAD_API_CALL PFNGLINDEXFPROC glad_glIndexf;
+#define glIndexf glad_glIndexf
+GLAD_API_CALL PFNGLINDEXFVPROC glad_glIndexfv;
+#define glIndexfv glad_glIndexfv
+GLAD_API_CALL PFNGLINDEXIPROC glad_glIndexi;
+#define glIndexi glad_glIndexi
+GLAD_API_CALL PFNGLINDEXIVPROC glad_glIndexiv;
+#define glIndexiv glad_glIndexiv
+GLAD_API_CALL PFNGLINDEXSPROC glad_glIndexs;
+#define glIndexs glad_glIndexs
+GLAD_API_CALL PFNGLINDEXSVPROC glad_glIndexsv;
+#define glIndexsv glad_glIndexsv
+GLAD_API_CALL PFNGLINDEXUBPROC glad_glIndexub;
+#define glIndexub glad_glIndexub
+GLAD_API_CALL PFNGLINDEXUBVPROC glad_glIndexubv;
+#define glIndexubv glad_glIndexubv
+GLAD_API_CALL PFNGLINITNAMESPROC glad_glInitNames;
+#define glInitNames glad_glInitNames
+GLAD_API_CALL PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays;
+#define glInterleavedArrays glad_glInterleavedArrays
+GLAD_API_CALL PFNGLISBUFFERPROC glad_glIsBuffer;
+#define glIsBuffer glad_glIsBuffer
+GLAD_API_CALL PFNGLISENABLEDPROC glad_glIsEnabled;
+#define glIsEnabled glad_glIsEnabled
+GLAD_API_CALL PFNGLISENABLEDIPROC glad_glIsEnabledi;
+#define glIsEnabledi glad_glIsEnabledi
+GLAD_API_CALL PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer;
+#define glIsFramebuffer glad_glIsFramebuffer
+GLAD_API_CALL PFNGLISLISTPROC glad_glIsList;
+#define glIsList glad_glIsList
+GLAD_API_CALL PFNGLISPROGRAMPROC glad_glIsProgram;
+#define glIsProgram glad_glIsProgram
+GLAD_API_CALL PFNGLISQUERYPROC glad_glIsQuery;
+#define glIsQuery glad_glIsQuery
+GLAD_API_CALL PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer;
+#define glIsRenderbuffer glad_glIsRenderbuffer
+GLAD_API_CALL PFNGLISSAMPLERPROC glad_glIsSampler;
+#define glIsSampler glad_glIsSampler
+GLAD_API_CALL PFNGLISSHADERPROC glad_glIsShader;
+#define glIsShader glad_glIsShader
+GLAD_API_CALL PFNGLISSYNCPROC glad_glIsSync;
+#define glIsSync glad_glIsSync
+GLAD_API_CALL PFNGLISTEXTUREPROC glad_glIsTexture;
+#define glIsTexture glad_glIsTexture
+GLAD_API_CALL PFNGLISVERTEXARRAYPROC glad_glIsVertexArray;
+#define glIsVertexArray glad_glIsVertexArray
+GLAD_API_CALL PFNGLLIGHTMODELFPROC glad_glLightModelf;
+#define glLightModelf glad_glLightModelf
+GLAD_API_CALL PFNGLLIGHTMODELFVPROC glad_glLightModelfv;
+#define glLightModelfv glad_glLightModelfv
+GLAD_API_CALL PFNGLLIGHTMODELIPROC glad_glLightModeli;
+#define glLightModeli glad_glLightModeli
+GLAD_API_CALL PFNGLLIGHTMODELIVPROC glad_glLightModeliv;
+#define glLightModeliv glad_glLightModeliv
+GLAD_API_CALL PFNGLLIGHTFPROC glad_glLightf;
+#define glLightf glad_glLightf
+GLAD_API_CALL PFNGLLIGHTFVPROC glad_glLightfv;
+#define glLightfv glad_glLightfv
+GLAD_API_CALL PFNGLLIGHTIPROC glad_glLighti;
+#define glLighti glad_glLighti
+GLAD_API_CALL PFNGLLIGHTIVPROC glad_glLightiv;
+#define glLightiv glad_glLightiv
+GLAD_API_CALL PFNGLLINESTIPPLEPROC glad_glLineStipple;
+#define glLineStipple glad_glLineStipple
+GLAD_API_CALL PFNGLLINEWIDTHPROC glad_glLineWidth;
+#define glLineWidth glad_glLineWidth
+GLAD_API_CALL PFNGLLINKPROGRAMPROC glad_glLinkProgram;
+#define glLinkProgram glad_glLinkProgram
+GLAD_API_CALL PFNGLLISTBASEPROC glad_glListBase;
+#define glListBase glad_glListBase
+GLAD_API_CALL PFNGLLOADIDENTITYPROC glad_glLoadIdentity;
+#define glLoadIdentity glad_glLoadIdentity
+GLAD_API_CALL PFNGLLOADMATRIXDPROC glad_glLoadMatrixd;
+#define glLoadMatrixd glad_glLoadMatrixd
+GLAD_API_CALL PFNGLLOADMATRIXFPROC glad_glLoadMatrixf;
+#define glLoadMatrixf glad_glLoadMatrixf
+GLAD_API_CALL PFNGLLOADNAMEPROC glad_glLoadName;
+#define glLoadName glad_glLoadName
+GLAD_API_CALL PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd;
+#define glLoadTransposeMatrixd glad_glLoadTransposeMatrixd
+GLAD_API_CALL PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf;
+#define glLoadTransposeMatrixf glad_glLoadTransposeMatrixf
+GLAD_API_CALL PFNGLLOGICOPPROC glad_glLogicOp;
+#define glLogicOp glad_glLogicOp
+GLAD_API_CALL PFNGLMAP1DPROC glad_glMap1d;
+#define glMap1d glad_glMap1d
+GLAD_API_CALL PFNGLMAP1FPROC glad_glMap1f;
+#define glMap1f glad_glMap1f
+GLAD_API_CALL PFNGLMAP2DPROC glad_glMap2d;
+#define glMap2d glad_glMap2d
+GLAD_API_CALL PFNGLMAP2FPROC glad_glMap2f;
+#define glMap2f glad_glMap2f
+GLAD_API_CALL PFNGLMAPBUFFERPROC glad_glMapBuffer;
+#define glMapBuffer glad_glMapBuffer
+GLAD_API_CALL PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange;
+#define glMapBufferRange glad_glMapBufferRange
+GLAD_API_CALL PFNGLMAPGRID1DPROC glad_glMapGrid1d;
+#define glMapGrid1d glad_glMapGrid1d
+GLAD_API_CALL PFNGLMAPGRID1FPROC glad_glMapGrid1f;
+#define glMapGrid1f glad_glMapGrid1f
+GLAD_API_CALL PFNGLMAPGRID2DPROC glad_glMapGrid2d;
+#define glMapGrid2d glad_glMapGrid2d
+GLAD_API_CALL PFNGLMAPGRID2FPROC glad_glMapGrid2f;
+#define glMapGrid2f glad_glMapGrid2f
+GLAD_API_CALL PFNGLMATERIALFPROC glad_glMaterialf;
+#define glMaterialf glad_glMaterialf
+GLAD_API_CALL PFNGLMATERIALFVPROC glad_glMaterialfv;
+#define glMaterialfv glad_glMaterialfv
+GLAD_API_CALL PFNGLMATERIALIPROC glad_glMateriali;
+#define glMateriali glad_glMateriali
+GLAD_API_CALL PFNGLMATERIALIVPROC glad_glMaterialiv;
+#define glMaterialiv glad_glMaterialiv
+GLAD_API_CALL PFNGLMATRIXMODEPROC glad_glMatrixMode;
+#define glMatrixMode glad_glMatrixMode
+GLAD_API_CALL PFNGLMULTMATRIXDPROC glad_glMultMatrixd;
+#define glMultMatrixd glad_glMultMatrixd
+GLAD_API_CALL PFNGLMULTMATRIXFPROC glad_glMultMatrixf;
+#define glMultMatrixf glad_glMultMatrixf
+GLAD_API_CALL PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd;
+#define glMultTransposeMatrixd glad_glMultTransposeMatrixd
+GLAD_API_CALL PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf;
+#define glMultTransposeMatrixf glad_glMultTransposeMatrixf
+GLAD_API_CALL PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays;
+#define glMultiDrawArrays glad_glMultiDrawArrays
+GLAD_API_CALL PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements;
+#define glMultiDrawElements glad_glMultiDrawElements
+GLAD_API_CALL PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex;
+#define glMultiDrawElementsBaseVertex glad_glMultiDrawElementsBaseVertex
+GLAD_API_CALL PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d;
+#define glMultiTexCoord1d glad_glMultiTexCoord1d
+GLAD_API_CALL PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv;
+#define glMultiTexCoord1dv glad_glMultiTexCoord1dv
+GLAD_API_CALL PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f;
+#define glMultiTexCoord1f glad_glMultiTexCoord1f
+GLAD_API_CALL PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv;
+#define glMultiTexCoord1fv glad_glMultiTexCoord1fv
+GLAD_API_CALL PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i;
+#define glMultiTexCoord1i glad_glMultiTexCoord1i
+GLAD_API_CALL PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv;
+#define glMultiTexCoord1iv glad_glMultiTexCoord1iv
+GLAD_API_CALL PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s;
+#define glMultiTexCoord1s glad_glMultiTexCoord1s
+GLAD_API_CALL PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv;
+#define glMultiTexCoord1sv glad_glMultiTexCoord1sv
+GLAD_API_CALL PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d;
+#define glMultiTexCoord2d glad_glMultiTexCoord2d
+GLAD_API_CALL PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv;
+#define glMultiTexCoord2dv glad_glMultiTexCoord2dv
+GLAD_API_CALL PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f;
+#define glMultiTexCoord2f glad_glMultiTexCoord2f
+GLAD_API_CALL PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv;
+#define glMultiTexCoord2fv glad_glMultiTexCoord2fv
+GLAD_API_CALL PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i;
+#define glMultiTexCoord2i glad_glMultiTexCoord2i
+GLAD_API_CALL PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv;
+#define glMultiTexCoord2iv glad_glMultiTexCoord2iv
+GLAD_API_CALL PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s;
+#define glMultiTexCoord2s glad_glMultiTexCoord2s
+GLAD_API_CALL PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv;
+#define glMultiTexCoord2sv glad_glMultiTexCoord2sv
+GLAD_API_CALL PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d;
+#define glMultiTexCoord3d glad_glMultiTexCoord3d
+GLAD_API_CALL PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv;
+#define glMultiTexCoord3dv glad_glMultiTexCoord3dv
+GLAD_API_CALL PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f;
+#define glMultiTexCoord3f glad_glMultiTexCoord3f
+GLAD_API_CALL PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv;
+#define glMultiTexCoord3fv glad_glMultiTexCoord3fv
+GLAD_API_CALL PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i;
+#define glMultiTexCoord3i glad_glMultiTexCoord3i
+GLAD_API_CALL PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv;
+#define glMultiTexCoord3iv glad_glMultiTexCoord3iv
+GLAD_API_CALL PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s;
+#define glMultiTexCoord3s glad_glMultiTexCoord3s
+GLAD_API_CALL PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv;
+#define glMultiTexCoord3sv glad_glMultiTexCoord3sv
+GLAD_API_CALL PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d;
+#define glMultiTexCoord4d glad_glMultiTexCoord4d
+GLAD_API_CALL PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv;
+#define glMultiTexCoord4dv glad_glMultiTexCoord4dv
+GLAD_API_CALL PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f;
+#define glMultiTexCoord4f glad_glMultiTexCoord4f
+GLAD_API_CALL PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv;
+#define glMultiTexCoord4fv glad_glMultiTexCoord4fv
+GLAD_API_CALL PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i;
+#define glMultiTexCoord4i glad_glMultiTexCoord4i
+GLAD_API_CALL PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv;
+#define glMultiTexCoord4iv glad_glMultiTexCoord4iv
+GLAD_API_CALL PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s;
+#define glMultiTexCoord4s glad_glMultiTexCoord4s
+GLAD_API_CALL PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv;
+#define glMultiTexCoord4sv glad_glMultiTexCoord4sv
+GLAD_API_CALL PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui;
+#define glMultiTexCoordP1ui glad_glMultiTexCoordP1ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv;
+#define glMultiTexCoordP1uiv glad_glMultiTexCoordP1uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui;
+#define glMultiTexCoordP2ui glad_glMultiTexCoordP2ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv;
+#define glMultiTexCoordP2uiv glad_glMultiTexCoordP2uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui;
+#define glMultiTexCoordP3ui glad_glMultiTexCoordP3ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv;
+#define glMultiTexCoordP3uiv glad_glMultiTexCoordP3uiv
+GLAD_API_CALL PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui;
+#define glMultiTexCoordP4ui glad_glMultiTexCoordP4ui
+GLAD_API_CALL PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv;
+#define glMultiTexCoordP4uiv glad_glMultiTexCoordP4uiv
+GLAD_API_CALL PFNGLNEWLISTPROC glad_glNewList;
+#define glNewList glad_glNewList
+GLAD_API_CALL PFNGLNORMAL3BPROC glad_glNormal3b;
+#define glNormal3b glad_glNormal3b
+GLAD_API_CALL PFNGLNORMAL3BVPROC glad_glNormal3bv;
+#define glNormal3bv glad_glNormal3bv
+GLAD_API_CALL PFNGLNORMAL3DPROC glad_glNormal3d;
+#define glNormal3d glad_glNormal3d
+GLAD_API_CALL PFNGLNORMAL3DVPROC glad_glNormal3dv;
+#define glNormal3dv glad_glNormal3dv
+GLAD_API_CALL PFNGLNORMAL3FPROC glad_glNormal3f;
+#define glNormal3f glad_glNormal3f
+GLAD_API_CALL PFNGLNORMAL3FVPROC glad_glNormal3fv;
+#define glNormal3fv glad_glNormal3fv
+GLAD_API_CALL PFNGLNORMAL3IPROC glad_glNormal3i;
+#define glNormal3i glad_glNormal3i
+GLAD_API_CALL PFNGLNORMAL3IVPROC glad_glNormal3iv;
+#define glNormal3iv glad_glNormal3iv
+GLAD_API_CALL PFNGLNORMAL3SPROC glad_glNormal3s;
+#define glNormal3s glad_glNormal3s
+GLAD_API_CALL PFNGLNORMAL3SVPROC glad_glNormal3sv;
+#define glNormal3sv glad_glNormal3sv
+GLAD_API_CALL PFNGLNORMALP3UIPROC glad_glNormalP3ui;
+#define glNormalP3ui glad_glNormalP3ui
+GLAD_API_CALL PFNGLNORMALP3UIVPROC glad_glNormalP3uiv;
+#define glNormalP3uiv glad_glNormalP3uiv
+GLAD_API_CALL PFNGLNORMALPOINTERPROC glad_glNormalPointer;
+#define glNormalPointer glad_glNormalPointer
+GLAD_API_CALL PFNGLORTHOPROC glad_glOrtho;
+#define glOrtho glad_glOrtho
+GLAD_API_CALL PFNGLPASSTHROUGHPROC glad_glPassThrough;
+#define glPassThrough glad_glPassThrough
+GLAD_API_CALL PFNGLPIXELMAPFVPROC glad_glPixelMapfv;
+#define glPixelMapfv glad_glPixelMapfv
+GLAD_API_CALL PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv;
+#define glPixelMapuiv glad_glPixelMapuiv
+GLAD_API_CALL PFNGLPIXELMAPUSVPROC glad_glPixelMapusv;
+#define glPixelMapusv glad_glPixelMapusv
+GLAD_API_CALL PFNGLPIXELSTOREFPROC glad_glPixelStoref;
+#define glPixelStoref glad_glPixelStoref
+GLAD_API_CALL PFNGLPIXELSTOREIPROC glad_glPixelStorei;
+#define glPixelStorei glad_glPixelStorei
+GLAD_API_CALL PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf;
+#define glPixelTransferf glad_glPixelTransferf
+GLAD_API_CALL PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi;
+#define glPixelTransferi glad_glPixelTransferi
+GLAD_API_CALL PFNGLPIXELZOOMPROC glad_glPixelZoom;
+#define glPixelZoom glad_glPixelZoom
+GLAD_API_CALL PFNGLPOINTPARAMETERFPROC glad_glPointParameterf;
+#define glPointParameterf glad_glPointParameterf
+GLAD_API_CALL PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv;
+#define glPointParameterfv glad_glPointParameterfv
+GLAD_API_CALL PFNGLPOINTPARAMETERIPROC glad_glPointParameteri;
+#define glPointParameteri glad_glPointParameteri
+GLAD_API_CALL PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv;
+#define glPointParameteriv glad_glPointParameteriv
+GLAD_API_CALL PFNGLPOINTSIZEPROC glad_glPointSize;
+#define glPointSize glad_glPointSize
+GLAD_API_CALL PFNGLPOLYGONMODEPROC glad_glPolygonMode;
+#define glPolygonMode glad_glPolygonMode
+GLAD_API_CALL PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset;
+#define glPolygonOffset glad_glPolygonOffset
+GLAD_API_CALL PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple;
+#define glPolygonStipple glad_glPolygonStipple
+GLAD_API_CALL PFNGLPOPATTRIBPROC glad_glPopAttrib;
+#define glPopAttrib glad_glPopAttrib
+GLAD_API_CALL PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib;
+#define glPopClientAttrib glad_glPopClientAttrib
+GLAD_API_CALL PFNGLPOPMATRIXPROC glad_glPopMatrix;
+#define glPopMatrix glad_glPopMatrix
+GLAD_API_CALL PFNGLPOPNAMEPROC glad_glPopName;
+#define glPopName glad_glPopName
+GLAD_API_CALL PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex;
+#define glPrimitiveRestartIndex glad_glPrimitiveRestartIndex
+GLAD_API_CALL PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures;
+#define glPrioritizeTextures glad_glPrioritizeTextures
+GLAD_API_CALL PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex;
+#define glProvokingVertex glad_glProvokingVertex
+GLAD_API_CALL PFNGLPUSHATTRIBPROC glad_glPushAttrib;
+#define glPushAttrib glad_glPushAttrib
+GLAD_API_CALL PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib;
+#define glPushClientAttrib glad_glPushClientAttrib
+GLAD_API_CALL PFNGLPUSHMATRIXPROC glad_glPushMatrix;
+#define glPushMatrix glad_glPushMatrix
+GLAD_API_CALL PFNGLPUSHNAMEPROC glad_glPushName;
+#define glPushName glad_glPushName
+GLAD_API_CALL PFNGLQUERYCOUNTERPROC glad_glQueryCounter;
+#define glQueryCounter glad_glQueryCounter
+GLAD_API_CALL PFNGLRASTERPOS2DPROC glad_glRasterPos2d;
+#define glRasterPos2d glad_glRasterPos2d
+GLAD_API_CALL PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv;
+#define glRasterPos2dv glad_glRasterPos2dv
+GLAD_API_CALL PFNGLRASTERPOS2FPROC glad_glRasterPos2f;
+#define glRasterPos2f glad_glRasterPos2f
+GLAD_API_CALL PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv;
+#define glRasterPos2fv glad_glRasterPos2fv
+GLAD_API_CALL PFNGLRASTERPOS2IPROC glad_glRasterPos2i;
+#define glRasterPos2i glad_glRasterPos2i
+GLAD_API_CALL PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv;
+#define glRasterPos2iv glad_glRasterPos2iv
+GLAD_API_CALL PFNGLRASTERPOS2SPROC glad_glRasterPos2s;
+#define glRasterPos2s glad_glRasterPos2s
+GLAD_API_CALL PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv;
+#define glRasterPos2sv glad_glRasterPos2sv
+GLAD_API_CALL PFNGLRASTERPOS3DPROC glad_glRasterPos3d;
+#define glRasterPos3d glad_glRasterPos3d
+GLAD_API_CALL PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv;
+#define glRasterPos3dv glad_glRasterPos3dv
+GLAD_API_CALL PFNGLRASTERPOS3FPROC glad_glRasterPos3f;
+#define glRasterPos3f glad_glRasterPos3f
+GLAD_API_CALL PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv;
+#define glRasterPos3fv glad_glRasterPos3fv
+GLAD_API_CALL PFNGLRASTERPOS3IPROC glad_glRasterPos3i;
+#define glRasterPos3i glad_glRasterPos3i
+GLAD_API_CALL PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv;
+#define glRasterPos3iv glad_glRasterPos3iv
+GLAD_API_CALL PFNGLRASTERPOS3SPROC glad_glRasterPos3s;
+#define glRasterPos3s glad_glRasterPos3s
+GLAD_API_CALL PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv;
+#define glRasterPos3sv glad_glRasterPos3sv
+GLAD_API_CALL PFNGLRASTERPOS4DPROC glad_glRasterPos4d;
+#define glRasterPos4d glad_glRasterPos4d
+GLAD_API_CALL PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv;
+#define glRasterPos4dv glad_glRasterPos4dv
+GLAD_API_CALL PFNGLRASTERPOS4FPROC glad_glRasterPos4f;
+#define glRasterPos4f glad_glRasterPos4f
+GLAD_API_CALL PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv;
+#define glRasterPos4fv glad_glRasterPos4fv
+GLAD_API_CALL PFNGLRASTERPOS4IPROC glad_glRasterPos4i;
+#define glRasterPos4i glad_glRasterPos4i
+GLAD_API_CALL PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv;
+#define glRasterPos4iv glad_glRasterPos4iv
+GLAD_API_CALL PFNGLRASTERPOS4SPROC glad_glRasterPos4s;
+#define glRasterPos4s glad_glRasterPos4s
+GLAD_API_CALL PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv;
+#define glRasterPos4sv glad_glRasterPos4sv
+GLAD_API_CALL PFNGLREADBUFFERPROC glad_glReadBuffer;
+#define glReadBuffer glad_glReadBuffer
+GLAD_API_CALL PFNGLREADPIXELSPROC glad_glReadPixels;
+#define glReadPixels glad_glReadPixels
+GLAD_API_CALL PFNGLRECTDPROC glad_glRectd;
+#define glRectd glad_glRectd
+GLAD_API_CALL PFNGLRECTDVPROC glad_glRectdv;
+#define glRectdv glad_glRectdv
+GLAD_API_CALL PFNGLRECTFPROC glad_glRectf;
+#define glRectf glad_glRectf
+GLAD_API_CALL PFNGLRECTFVPROC glad_glRectfv;
+#define glRectfv glad_glRectfv
+GLAD_API_CALL PFNGLRECTIPROC glad_glRecti;
+#define glRecti glad_glRecti
+GLAD_API_CALL PFNGLRECTIVPROC glad_glRectiv;
+#define glRectiv glad_glRectiv
+GLAD_API_CALL PFNGLRECTSPROC glad_glRects;
+#define glRects glad_glRects
+GLAD_API_CALL PFNGLRECTSVPROC glad_glRectsv;
+#define glRectsv glad_glRectsv
+GLAD_API_CALL PFNGLRENDERMODEPROC glad_glRenderMode;
+#define glRenderMode glad_glRenderMode
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage;
+#define glRenderbufferStorage glad_glRenderbufferStorage
+GLAD_API_CALL PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample;
+#define glRenderbufferStorageMultisample glad_glRenderbufferStorageMultisample
+GLAD_API_CALL PFNGLROTATEDPROC glad_glRotated;
+#define glRotated glad_glRotated
+GLAD_API_CALL PFNGLROTATEFPROC glad_glRotatef;
+#define glRotatef glad_glRotatef
+GLAD_API_CALL PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage;
+#define glSampleCoverage glad_glSampleCoverage
+GLAD_API_CALL PFNGLSAMPLEMASKIPROC glad_glSampleMaski;
+#define glSampleMaski glad_glSampleMaski
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv;
+#define glSamplerParameterIiv glad_glSamplerParameterIiv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv;
+#define glSamplerParameterIuiv glad_glSamplerParameterIuiv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf;
+#define glSamplerParameterf glad_glSamplerParameterf
+GLAD_API_CALL PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv;
+#define glSamplerParameterfv glad_glSamplerParameterfv
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri;
+#define glSamplerParameteri glad_glSamplerParameteri
+GLAD_API_CALL PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv;
+#define glSamplerParameteriv glad_glSamplerParameteriv
+GLAD_API_CALL PFNGLSCALEDPROC glad_glScaled;
+#define glScaled glad_glScaled
+GLAD_API_CALL PFNGLSCALEFPROC glad_glScalef;
+#define glScalef glad_glScalef
+GLAD_API_CALL PFNGLSCISSORPROC glad_glScissor;
+#define glScissor glad_glScissor
+GLAD_API_CALL PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b;
+#define glSecondaryColor3b glad_glSecondaryColor3b
+GLAD_API_CALL PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv;
+#define glSecondaryColor3bv glad_glSecondaryColor3bv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d;
+#define glSecondaryColor3d glad_glSecondaryColor3d
+GLAD_API_CALL PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv;
+#define glSecondaryColor3dv glad_glSecondaryColor3dv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f;
+#define glSecondaryColor3f glad_glSecondaryColor3f
+GLAD_API_CALL PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv;
+#define glSecondaryColor3fv glad_glSecondaryColor3fv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i;
+#define glSecondaryColor3i glad_glSecondaryColor3i
+GLAD_API_CALL PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv;
+#define glSecondaryColor3iv glad_glSecondaryColor3iv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s;
+#define glSecondaryColor3s glad_glSecondaryColor3s
+GLAD_API_CALL PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv;
+#define glSecondaryColor3sv glad_glSecondaryColor3sv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub;
+#define glSecondaryColor3ub glad_glSecondaryColor3ub
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv;
+#define glSecondaryColor3ubv glad_glSecondaryColor3ubv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui;
+#define glSecondaryColor3ui glad_glSecondaryColor3ui
+GLAD_API_CALL PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv;
+#define glSecondaryColor3uiv glad_glSecondaryColor3uiv
+GLAD_API_CALL PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us;
+#define glSecondaryColor3us glad_glSecondaryColor3us
+GLAD_API_CALL PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv;
+#define glSecondaryColor3usv glad_glSecondaryColor3usv
+GLAD_API_CALL PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui;
+#define glSecondaryColorP3ui glad_glSecondaryColorP3ui
+GLAD_API_CALL PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv;
+#define glSecondaryColorP3uiv glad_glSecondaryColorP3uiv
+GLAD_API_CALL PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer;
+#define glSecondaryColorPointer glad_glSecondaryColorPointer
+GLAD_API_CALL PFNGLSELECTBUFFERPROC glad_glSelectBuffer;
+#define glSelectBuffer glad_glSelectBuffer
+GLAD_API_CALL PFNGLSHADEMODELPROC glad_glShadeModel;
+#define glShadeModel glad_glShadeModel
+GLAD_API_CALL PFNGLSHADERSOURCEPROC glad_glShaderSource;
+#define glShaderSource glad_glShaderSource
+GLAD_API_CALL PFNGLSTENCILFUNCPROC glad_glStencilFunc;
+#define glStencilFunc glad_glStencilFunc
+GLAD_API_CALL PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate;
+#define glStencilFuncSeparate glad_glStencilFuncSeparate
+GLAD_API_CALL PFNGLSTENCILMASKPROC glad_glStencilMask;
+#define glStencilMask glad_glStencilMask
+GLAD_API_CALL PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate;
+#define glStencilMaskSeparate glad_glStencilMaskSeparate
+GLAD_API_CALL PFNGLSTENCILOPPROC glad_glStencilOp;
+#define glStencilOp glad_glStencilOp
+GLAD_API_CALL PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate;
+#define glStencilOpSeparate glad_glStencilOpSeparate
+GLAD_API_CALL PFNGLTEXBUFFERPROC glad_glTexBuffer;
+#define glTexBuffer glad_glTexBuffer
+GLAD_API_CALL PFNGLTEXCOORD1DPROC glad_glTexCoord1d;
+#define glTexCoord1d glad_glTexCoord1d
+GLAD_API_CALL PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv;
+#define glTexCoord1dv glad_glTexCoord1dv
+GLAD_API_CALL PFNGLTEXCOORD1FPROC glad_glTexCoord1f;
+#define glTexCoord1f glad_glTexCoord1f
+GLAD_API_CALL PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv;
+#define glTexCoord1fv glad_glTexCoord1fv
+GLAD_API_CALL PFNGLTEXCOORD1IPROC glad_glTexCoord1i;
+#define glTexCoord1i glad_glTexCoord1i
+GLAD_API_CALL PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv;
+#define glTexCoord1iv glad_glTexCoord1iv
+GLAD_API_CALL PFNGLTEXCOORD1SPROC glad_glTexCoord1s;
+#define glTexCoord1s glad_glTexCoord1s
+GLAD_API_CALL PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv;
+#define glTexCoord1sv glad_glTexCoord1sv
+GLAD_API_CALL PFNGLTEXCOORD2DPROC glad_glTexCoord2d;
+#define glTexCoord2d glad_glTexCoord2d
+GLAD_API_CALL PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv;
+#define glTexCoord2dv glad_glTexCoord2dv
+GLAD_API_CALL PFNGLTEXCOORD2FPROC glad_glTexCoord2f;
+#define glTexCoord2f glad_glTexCoord2f
+GLAD_API_CALL PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv;
+#define glTexCoord2fv glad_glTexCoord2fv
+GLAD_API_CALL PFNGLTEXCOORD2IPROC glad_glTexCoord2i;
+#define glTexCoord2i glad_glTexCoord2i
+GLAD_API_CALL PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv;
+#define glTexCoord2iv glad_glTexCoord2iv
+GLAD_API_CALL PFNGLTEXCOORD2SPROC glad_glTexCoord2s;
+#define glTexCoord2s glad_glTexCoord2s
+GLAD_API_CALL PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv;
+#define glTexCoord2sv glad_glTexCoord2sv
+GLAD_API_CALL PFNGLTEXCOORD3DPROC glad_glTexCoord3d;
+#define glTexCoord3d glad_glTexCoord3d
+GLAD_API_CALL PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv;
+#define glTexCoord3dv glad_glTexCoord3dv
+GLAD_API_CALL PFNGLTEXCOORD3FPROC glad_glTexCoord3f;
+#define glTexCoord3f glad_glTexCoord3f
+GLAD_API_CALL PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv;
+#define glTexCoord3fv glad_glTexCoord3fv
+GLAD_API_CALL PFNGLTEXCOORD3IPROC glad_glTexCoord3i;
+#define glTexCoord3i glad_glTexCoord3i
+GLAD_API_CALL PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv;
+#define glTexCoord3iv glad_glTexCoord3iv
+GLAD_API_CALL PFNGLTEXCOORD3SPROC glad_glTexCoord3s;
+#define glTexCoord3s glad_glTexCoord3s
+GLAD_API_CALL PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv;
+#define glTexCoord3sv glad_glTexCoord3sv
+GLAD_API_CALL PFNGLTEXCOORD4DPROC glad_glTexCoord4d;
+#define glTexCoord4d glad_glTexCoord4d
+GLAD_API_CALL PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv;
+#define glTexCoord4dv glad_glTexCoord4dv
+GLAD_API_CALL PFNGLTEXCOORD4FPROC glad_glTexCoord4f;
+#define glTexCoord4f glad_glTexCoord4f
+GLAD_API_CALL PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv;
+#define glTexCoord4fv glad_glTexCoord4fv
+GLAD_API_CALL PFNGLTEXCOORD4IPROC glad_glTexCoord4i;
+#define glTexCoord4i glad_glTexCoord4i
+GLAD_API_CALL PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv;
+#define glTexCoord4iv glad_glTexCoord4iv
+GLAD_API_CALL PFNGLTEXCOORD4SPROC glad_glTexCoord4s;
+#define glTexCoord4s glad_glTexCoord4s
+GLAD_API_CALL PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv;
+#define glTexCoord4sv glad_glTexCoord4sv
+GLAD_API_CALL PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui;
+#define glTexCoordP1ui glad_glTexCoordP1ui
+GLAD_API_CALL PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv;
+#define glTexCoordP1uiv glad_glTexCoordP1uiv
+GLAD_API_CALL PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui;
+#define glTexCoordP2ui glad_glTexCoordP2ui
+GLAD_API_CALL PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv;
+#define glTexCoordP2uiv glad_glTexCoordP2uiv
+GLAD_API_CALL PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui;
+#define glTexCoordP3ui glad_glTexCoordP3ui
+GLAD_API_CALL PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv;
+#define glTexCoordP3uiv glad_glTexCoordP3uiv
+GLAD_API_CALL PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui;
+#define glTexCoordP4ui glad_glTexCoordP4ui
+GLAD_API_CALL PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv;
+#define glTexCoordP4uiv glad_glTexCoordP4uiv
+GLAD_API_CALL PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer;
+#define glTexCoordPointer glad_glTexCoordPointer
+GLAD_API_CALL PFNGLTEXENVFPROC glad_glTexEnvf;
+#define glTexEnvf glad_glTexEnvf
+GLAD_API_CALL PFNGLTEXENVFVPROC glad_glTexEnvfv;
+#define glTexEnvfv glad_glTexEnvfv
+GLAD_API_CALL PFNGLTEXENVIPROC glad_glTexEnvi;
+#define glTexEnvi glad_glTexEnvi
+GLAD_API_CALL PFNGLTEXENVIVPROC glad_glTexEnviv;
+#define glTexEnviv glad_glTexEnviv
+GLAD_API_CALL PFNGLTEXGENDPROC glad_glTexGend;
+#define glTexGend glad_glTexGend
+GLAD_API_CALL PFNGLTEXGENDVPROC glad_glTexGendv;
+#define glTexGendv glad_glTexGendv
+GLAD_API_CALL PFNGLTEXGENFPROC glad_glTexGenf;
+#define glTexGenf glad_glTexGenf
+GLAD_API_CALL PFNGLTEXGENFVPROC glad_glTexGenfv;
+#define glTexGenfv glad_glTexGenfv
+GLAD_API_CALL PFNGLTEXGENIPROC glad_glTexGeni;
+#define glTexGeni glad_glTexGeni
+GLAD_API_CALL PFNGLTEXGENIVPROC glad_glTexGeniv;
+#define glTexGeniv glad_glTexGeniv
+GLAD_API_CALL PFNGLTEXIMAGE1DPROC glad_glTexImage1D;
+#define glTexImage1D glad_glTexImage1D
+GLAD_API_CALL PFNGLTEXIMAGE2DPROC glad_glTexImage2D;
+#define glTexImage2D glad_glTexImage2D
+GLAD_API_CALL PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample;
+#define glTexImage2DMultisample glad_glTexImage2DMultisample
+GLAD_API_CALL PFNGLTEXIMAGE3DPROC glad_glTexImage3D;
+#define glTexImage3D glad_glTexImage3D
+GLAD_API_CALL PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample;
+#define glTexImage3DMultisample glad_glTexImage3DMultisample
+GLAD_API_CALL PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv;
+#define glTexParameterIiv glad_glTexParameterIiv
+GLAD_API_CALL PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv;
+#define glTexParameterIuiv glad_glTexParameterIuiv
+GLAD_API_CALL PFNGLTEXPARAMETERFPROC glad_glTexParameterf;
+#define glTexParameterf glad_glTexParameterf
+GLAD_API_CALL PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv;
+#define glTexParameterfv glad_glTexParameterfv
+GLAD_API_CALL PFNGLTEXPARAMETERIPROC glad_glTexParameteri;
+#define glTexParameteri glad_glTexParameteri
+GLAD_API_CALL PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv;
+#define glTexParameteriv glad_glTexParameteriv
+GLAD_API_CALL PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D;
+#define glTexSubImage1D glad_glTexSubImage1D
+GLAD_API_CALL PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D;
+#define glTexSubImage2D glad_glTexSubImage2D
+GLAD_API_CALL PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D;
+#define glTexSubImage3D glad_glTexSubImage3D
+GLAD_API_CALL PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings;
+#define glTransformFeedbackVaryings glad_glTransformFeedbackVaryings
+GLAD_API_CALL PFNGLTRANSLATEDPROC glad_glTranslated;
+#define glTranslated glad_glTranslated
+GLAD_API_CALL PFNGLTRANSLATEFPROC glad_glTranslatef;
+#define glTranslatef glad_glTranslatef
+GLAD_API_CALL PFNGLUNIFORM1FPROC glad_glUniform1f;
+#define glUniform1f glad_glUniform1f
+GLAD_API_CALL PFNGLUNIFORM1FVPROC glad_glUniform1fv;
+#define glUniform1fv glad_glUniform1fv
+GLAD_API_CALL PFNGLUNIFORM1IPROC glad_glUniform1i;
+#define glUniform1i glad_glUniform1i
+GLAD_API_CALL PFNGLUNIFORM1IVPROC glad_glUniform1iv;
+#define glUniform1iv glad_glUniform1iv
+GLAD_API_CALL PFNGLUNIFORM1UIPROC glad_glUniform1ui;
+#define glUniform1ui glad_glUniform1ui
+GLAD_API_CALL PFNGLUNIFORM1UIVPROC glad_glUniform1uiv;
+#define glUniform1uiv glad_glUniform1uiv
+GLAD_API_CALL PFNGLUNIFORM2FPROC glad_glUniform2f;
+#define glUniform2f glad_glUniform2f
+GLAD_API_CALL PFNGLUNIFORM2FVPROC glad_glUniform2fv;
+#define glUniform2fv glad_glUniform2fv
+GLAD_API_CALL PFNGLUNIFORM2IPROC glad_glUniform2i;
+#define glUniform2i glad_glUniform2i
+GLAD_API_CALL PFNGLUNIFORM2IVPROC glad_glUniform2iv;
+#define glUniform2iv glad_glUniform2iv
+GLAD_API_CALL PFNGLUNIFORM2UIPROC glad_glUniform2ui;
+#define glUniform2ui glad_glUniform2ui
+GLAD_API_CALL PFNGLUNIFORM2UIVPROC glad_glUniform2uiv;
+#define glUniform2uiv glad_glUniform2uiv
+GLAD_API_CALL PFNGLUNIFORM3FPROC glad_glUniform3f;
+#define glUniform3f glad_glUniform3f
+GLAD_API_CALL PFNGLUNIFORM3FVPROC glad_glUniform3fv;
+#define glUniform3fv glad_glUniform3fv
+GLAD_API_CALL PFNGLUNIFORM3IPROC glad_glUniform3i;
+#define glUniform3i glad_glUniform3i
+GLAD_API_CALL PFNGLUNIFORM3IVPROC glad_glUniform3iv;
+#define glUniform3iv glad_glUniform3iv
+GLAD_API_CALL PFNGLUNIFORM3UIPROC glad_glUniform3ui;
+#define glUniform3ui glad_glUniform3ui
+GLAD_API_CALL PFNGLUNIFORM3UIVPROC glad_glUniform3uiv;
+#define glUniform3uiv glad_glUniform3uiv
+GLAD_API_CALL PFNGLUNIFORM4FPROC glad_glUniform4f;
+#define glUniform4f glad_glUniform4f
+GLAD_API_CALL PFNGLUNIFORM4FVPROC glad_glUniform4fv;
+#define glUniform4fv glad_glUniform4fv
+GLAD_API_CALL PFNGLUNIFORM4IPROC glad_glUniform4i;
+#define glUniform4i glad_glUniform4i
+GLAD_API_CALL PFNGLUNIFORM4IVPROC glad_glUniform4iv;
+#define glUniform4iv glad_glUniform4iv
+GLAD_API_CALL PFNGLUNIFORM4UIPROC glad_glUniform4ui;
+#define glUniform4ui glad_glUniform4ui
+GLAD_API_CALL PFNGLUNIFORM4UIVPROC glad_glUniform4uiv;
+#define glUniform4uiv glad_glUniform4uiv
+GLAD_API_CALL PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding;
+#define glUniformBlockBinding glad_glUniformBlockBinding
+GLAD_API_CALL PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv;
+#define glUniformMatrix2fv glad_glUniformMatrix2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv;
+#define glUniformMatrix2x3fv glad_glUniformMatrix2x3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv;
+#define glUniformMatrix2x4fv glad_glUniformMatrix2x4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv;
+#define glUniformMatrix3fv glad_glUniformMatrix3fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv;
+#define glUniformMatrix3x2fv glad_glUniformMatrix3x2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv;
+#define glUniformMatrix3x4fv glad_glUniformMatrix3x4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv;
+#define glUniformMatrix4fv glad_glUniformMatrix4fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv;
+#define glUniformMatrix4x2fv glad_glUniformMatrix4x2fv
+GLAD_API_CALL PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv;
+#define glUniformMatrix4x3fv glad_glUniformMatrix4x3fv
+GLAD_API_CALL PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer;
+#define glUnmapBuffer glad_glUnmapBuffer
+GLAD_API_CALL PFNGLUSEPROGRAMPROC glad_glUseProgram;
+#define glUseProgram glad_glUseProgram
+GLAD_API_CALL PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram;
+#define glValidateProgram glad_glValidateProgram
+GLAD_API_CALL PFNGLVERTEX2DPROC glad_glVertex2d;
+#define glVertex2d glad_glVertex2d
+GLAD_API_CALL PFNGLVERTEX2DVPROC glad_glVertex2dv;
+#define glVertex2dv glad_glVertex2dv
+GLAD_API_CALL PFNGLVERTEX2FPROC glad_glVertex2f;
+#define glVertex2f glad_glVertex2f
+GLAD_API_CALL PFNGLVERTEX2FVPROC glad_glVertex2fv;
+#define glVertex2fv glad_glVertex2fv
+GLAD_API_CALL PFNGLVERTEX2IPROC glad_glVertex2i;
+#define glVertex2i glad_glVertex2i
+GLAD_API_CALL PFNGLVERTEX2IVPROC glad_glVertex2iv;
+#define glVertex2iv glad_glVertex2iv
+GLAD_API_CALL PFNGLVERTEX2SPROC glad_glVertex2s;
+#define glVertex2s glad_glVertex2s
+GLAD_API_CALL PFNGLVERTEX2SVPROC glad_glVertex2sv;
+#define glVertex2sv glad_glVertex2sv
+GLAD_API_CALL PFNGLVERTEX3DPROC glad_glVertex3d;
+#define glVertex3d glad_glVertex3d
+GLAD_API_CALL PFNGLVERTEX3DVPROC glad_glVertex3dv;
+#define glVertex3dv glad_glVertex3dv
+GLAD_API_CALL PFNGLVERTEX3FPROC glad_glVertex3f;
+#define glVertex3f glad_glVertex3f
+GLAD_API_CALL PFNGLVERTEX3FVPROC glad_glVertex3fv;
+#define glVertex3fv glad_glVertex3fv
+GLAD_API_CALL PFNGLVERTEX3IPROC glad_glVertex3i;
+#define glVertex3i glad_glVertex3i
+GLAD_API_CALL PFNGLVERTEX3IVPROC glad_glVertex3iv;
+#define glVertex3iv glad_glVertex3iv
+GLAD_API_CALL PFNGLVERTEX3SPROC glad_glVertex3s;
+#define glVertex3s glad_glVertex3s
+GLAD_API_CALL PFNGLVERTEX3SVPROC glad_glVertex3sv;
+#define glVertex3sv glad_glVertex3sv
+GLAD_API_CALL PFNGLVERTEX4DPROC glad_glVertex4d;
+#define glVertex4d glad_glVertex4d
+GLAD_API_CALL PFNGLVERTEX4DVPROC glad_glVertex4dv;
+#define glVertex4dv glad_glVertex4dv
+GLAD_API_CALL PFNGLVERTEX4FPROC glad_glVertex4f;
+#define glVertex4f glad_glVertex4f
+GLAD_API_CALL PFNGLVERTEX4FVPROC glad_glVertex4fv;
+#define glVertex4fv glad_glVertex4fv
+GLAD_API_CALL PFNGLVERTEX4IPROC glad_glVertex4i;
+#define glVertex4i glad_glVertex4i
+GLAD_API_CALL PFNGLVERTEX4IVPROC glad_glVertex4iv;
+#define glVertex4iv glad_glVertex4iv
+GLAD_API_CALL PFNGLVERTEX4SPROC glad_glVertex4s;
+#define glVertex4s glad_glVertex4s
+GLAD_API_CALL PFNGLVERTEX4SVPROC glad_glVertex4sv;
+#define glVertex4sv glad_glVertex4sv
+GLAD_API_CALL PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d;
+#define glVertexAttrib1d glad_glVertexAttrib1d
+GLAD_API_CALL PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv;
+#define glVertexAttrib1dv glad_glVertexAttrib1dv
+GLAD_API_CALL PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f;
+#define glVertexAttrib1f glad_glVertexAttrib1f
+GLAD_API_CALL PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv;
+#define glVertexAttrib1fv glad_glVertexAttrib1fv
+GLAD_API_CALL PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s;
+#define glVertexAttrib1s glad_glVertexAttrib1s
+GLAD_API_CALL PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv;
+#define glVertexAttrib1sv glad_glVertexAttrib1sv
+GLAD_API_CALL PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d;
+#define glVertexAttrib2d glad_glVertexAttrib2d
+GLAD_API_CALL PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv;
+#define glVertexAttrib2dv glad_glVertexAttrib2dv
+GLAD_API_CALL PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f;
+#define glVertexAttrib2f glad_glVertexAttrib2f
+GLAD_API_CALL PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv;
+#define glVertexAttrib2fv glad_glVertexAttrib2fv
+GLAD_API_CALL PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s;
+#define glVertexAttrib2s glad_glVertexAttrib2s
+GLAD_API_CALL PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv;
+#define glVertexAttrib2sv glad_glVertexAttrib2sv
+GLAD_API_CALL PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d;
+#define glVertexAttrib3d glad_glVertexAttrib3d
+GLAD_API_CALL PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv;
+#define glVertexAttrib3dv glad_glVertexAttrib3dv
+GLAD_API_CALL PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f;
+#define glVertexAttrib3f glad_glVertexAttrib3f
+GLAD_API_CALL PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv;
+#define glVertexAttrib3fv glad_glVertexAttrib3fv
+GLAD_API_CALL PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s;
+#define glVertexAttrib3s glad_glVertexAttrib3s
+GLAD_API_CALL PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv;
+#define glVertexAttrib3sv glad_glVertexAttrib3sv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv;
+#define glVertexAttrib4Nbv glad_glVertexAttrib4Nbv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv;
+#define glVertexAttrib4Niv glad_glVertexAttrib4Niv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv;
+#define glVertexAttrib4Nsv glad_glVertexAttrib4Nsv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub;
+#define glVertexAttrib4Nub glad_glVertexAttrib4Nub
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv;
+#define glVertexAttrib4Nubv glad_glVertexAttrib4Nubv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv;
+#define glVertexAttrib4Nuiv glad_glVertexAttrib4Nuiv
+GLAD_API_CALL PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv;
+#define glVertexAttrib4Nusv glad_glVertexAttrib4Nusv
+GLAD_API_CALL PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv;
+#define glVertexAttrib4bv glad_glVertexAttrib4bv
+GLAD_API_CALL PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d;
+#define glVertexAttrib4d glad_glVertexAttrib4d
+GLAD_API_CALL PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv;
+#define glVertexAttrib4dv glad_glVertexAttrib4dv
+GLAD_API_CALL PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f;
+#define glVertexAttrib4f glad_glVertexAttrib4f
+GLAD_API_CALL PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv;
+#define glVertexAttrib4fv glad_glVertexAttrib4fv
+GLAD_API_CALL PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv;
+#define glVertexAttrib4iv glad_glVertexAttrib4iv
+GLAD_API_CALL PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s;
+#define glVertexAttrib4s glad_glVertexAttrib4s
+GLAD_API_CALL PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv;
+#define glVertexAttrib4sv glad_glVertexAttrib4sv
+GLAD_API_CALL PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv;
+#define glVertexAttrib4ubv glad_glVertexAttrib4ubv
+GLAD_API_CALL PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv;
+#define glVertexAttrib4uiv glad_glVertexAttrib4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv;
+#define glVertexAttrib4usv glad_glVertexAttrib4usv
+GLAD_API_CALL PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor;
+#define glVertexAttribDivisor glad_glVertexAttribDivisor
+GLAD_API_CALL PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i;
+#define glVertexAttribI1i glad_glVertexAttribI1i
+GLAD_API_CALL PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv;
+#define glVertexAttribI1iv glad_glVertexAttribI1iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui;
+#define glVertexAttribI1ui glad_glVertexAttribI1ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv;
+#define glVertexAttribI1uiv glad_glVertexAttribI1uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i;
+#define glVertexAttribI2i glad_glVertexAttribI2i
+GLAD_API_CALL PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv;
+#define glVertexAttribI2iv glad_glVertexAttribI2iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui;
+#define glVertexAttribI2ui glad_glVertexAttribI2ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv;
+#define glVertexAttribI2uiv glad_glVertexAttribI2uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i;
+#define glVertexAttribI3i glad_glVertexAttribI3i
+GLAD_API_CALL PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv;
+#define glVertexAttribI3iv glad_glVertexAttribI3iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui;
+#define glVertexAttribI3ui glad_glVertexAttribI3ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv;
+#define glVertexAttribI3uiv glad_glVertexAttribI3uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv;
+#define glVertexAttribI4bv glad_glVertexAttribI4bv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i;
+#define glVertexAttribI4i glad_glVertexAttribI4i
+GLAD_API_CALL PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv;
+#define glVertexAttribI4iv glad_glVertexAttribI4iv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv;
+#define glVertexAttribI4sv glad_glVertexAttribI4sv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv;
+#define glVertexAttribI4ubv glad_glVertexAttribI4ubv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui;
+#define glVertexAttribI4ui glad_glVertexAttribI4ui
+GLAD_API_CALL PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv;
+#define glVertexAttribI4uiv glad_glVertexAttribI4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv;
+#define glVertexAttribI4usv glad_glVertexAttribI4usv
+GLAD_API_CALL PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer;
+#define glVertexAttribIPointer glad_glVertexAttribIPointer
+GLAD_API_CALL PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui;
+#define glVertexAttribP1ui glad_glVertexAttribP1ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv;
+#define glVertexAttribP1uiv glad_glVertexAttribP1uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui;
+#define glVertexAttribP2ui glad_glVertexAttribP2ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv;
+#define glVertexAttribP2uiv glad_glVertexAttribP2uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui;
+#define glVertexAttribP3ui glad_glVertexAttribP3ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv;
+#define glVertexAttribP3uiv glad_glVertexAttribP3uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui;
+#define glVertexAttribP4ui glad_glVertexAttribP4ui
+GLAD_API_CALL PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv;
+#define glVertexAttribP4uiv glad_glVertexAttribP4uiv
+GLAD_API_CALL PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer;
+#define glVertexAttribPointer glad_glVertexAttribPointer
+GLAD_API_CALL PFNGLVERTEXP2UIPROC glad_glVertexP2ui;
+#define glVertexP2ui glad_glVertexP2ui
+GLAD_API_CALL PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv;
+#define glVertexP2uiv glad_glVertexP2uiv
+GLAD_API_CALL PFNGLVERTEXP3UIPROC glad_glVertexP3ui;
+#define glVertexP3ui glad_glVertexP3ui
+GLAD_API_CALL PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv;
+#define glVertexP3uiv glad_glVertexP3uiv
+GLAD_API_CALL PFNGLVERTEXP4UIPROC glad_glVertexP4ui;
+#define glVertexP4ui glad_glVertexP4ui
+GLAD_API_CALL PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv;
+#define glVertexP4uiv glad_glVertexP4uiv
+GLAD_API_CALL PFNGLVERTEXPOINTERPROC glad_glVertexPointer;
+#define glVertexPointer glad_glVertexPointer
+GLAD_API_CALL PFNGLVIEWPORTPROC glad_glViewport;
+#define glViewport glad_glViewport
+GLAD_API_CALL PFNGLWAITSYNCPROC glad_glWaitSync;
+#define glWaitSync glad_glWaitSync
+GLAD_API_CALL PFNGLWINDOWPOS2DPROC glad_glWindowPos2d;
+#define glWindowPos2d glad_glWindowPos2d
+GLAD_API_CALL PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv;
+#define glWindowPos2dv glad_glWindowPos2dv
+GLAD_API_CALL PFNGLWINDOWPOS2FPROC glad_glWindowPos2f;
+#define glWindowPos2f glad_glWindowPos2f
+GLAD_API_CALL PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv;
+#define glWindowPos2fv glad_glWindowPos2fv
+GLAD_API_CALL PFNGLWINDOWPOS2IPROC glad_glWindowPos2i;
+#define glWindowPos2i glad_glWindowPos2i
+GLAD_API_CALL PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv;
+#define glWindowPos2iv glad_glWindowPos2iv
+GLAD_API_CALL PFNGLWINDOWPOS2SPROC glad_glWindowPos2s;
+#define glWindowPos2s glad_glWindowPos2s
+GLAD_API_CALL PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv;
+#define glWindowPos2sv glad_glWindowPos2sv
+GLAD_API_CALL PFNGLWINDOWPOS3DPROC glad_glWindowPos3d;
+#define glWindowPos3d glad_glWindowPos3d
+GLAD_API_CALL PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv;
+#define glWindowPos3dv glad_glWindowPos3dv
+GLAD_API_CALL PFNGLWINDOWPOS3FPROC glad_glWindowPos3f;
+#define glWindowPos3f glad_glWindowPos3f
+GLAD_API_CALL PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv;
+#define glWindowPos3fv glad_glWindowPos3fv
+GLAD_API_CALL PFNGLWINDOWPOS3IPROC glad_glWindowPos3i;
+#define glWindowPos3i glad_glWindowPos3i
+GLAD_API_CALL PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv;
+#define glWindowPos3iv glad_glWindowPos3iv
+GLAD_API_CALL PFNGLWINDOWPOS3SPROC glad_glWindowPos3s;
+#define glWindowPos3s glad_glWindowPos3s
+GLAD_API_CALL PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv;
+#define glWindowPos3sv glad_glWindowPos3sv
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLUserPtr( GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGL( GLADloadfunc load);
+
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/glx.h glmark2-2021.02/src/glad/include/glad/glx.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/glx.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/glad/glx.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,570 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:13 2019
+ *
+ * Generator: C/C++
+ * Specification: glx
+ * Extensions: 2
+ *
+ * APIs:
+ *  - glx=1.4
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='glx=1.4' --extensions='GLX_EXT_swap_control,GLX_MESA_swap_control' c
+ *
+ * Online:
+ *    http://glad.sh/#api=glx%3D1.4&extensions=GLX_EXT_swap_control%2CGLX_MESA_swap_control&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_GLX_H_
+#define GLAD_GLX_H_
+
+#ifdef GLX_H
+    #error GLX header already included (API: glx), remove previous include!
+#endif
+#define GLX_H 1
+
+
+#include <X11/X.h>
+#include <X11/Xlib.h>
+#include <X11/Xutil.h>
+
+#include <glad/gl.h>
+
+#define GLAD_GLX
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define GLX_ACCUM_ALPHA_SIZE 17
+#define GLX_ACCUM_BLUE_SIZE 16
+#define GLX_ACCUM_BUFFER_BIT 0x00000080
+#define GLX_ACCUM_GREEN_SIZE 15
+#define GLX_ACCUM_RED_SIZE 14
+#define GLX_ALPHA_SIZE 11
+#define GLX_AUX_BUFFERS 7
+#define GLX_AUX_BUFFERS_BIT 0x00000010
+#define GLX_BACK_LEFT_BUFFER_BIT 0x00000004
+#define GLX_BACK_RIGHT_BUFFER_BIT 0x00000008
+#define GLX_BAD_ATTRIBUTE 2
+#define GLX_BAD_CONTEXT 5
+#define GLX_BAD_ENUM 7
+#define GLX_BAD_SCREEN 1
+#define GLX_BAD_VALUE 6
+#define GLX_BAD_VISUAL 4
+#define GLX_BLUE_SIZE 10
+#define GLX_BUFFER_SIZE 2
+#define GLX_BufferSwapComplete 1
+#define GLX_COLOR_INDEX_BIT 0x00000002
+#define GLX_COLOR_INDEX_TYPE 0x8015
+#define GLX_CONFIG_CAVEAT 0x20
+#define GLX_DAMAGED 0x8020
+#define GLX_DEPTH_BUFFER_BIT 0x00000020
+#define GLX_DEPTH_SIZE 12
+#define GLX_DIRECT_COLOR 0x8003
+#define GLX_DONT_CARE 0xFFFFFFFF
+#define GLX_DOUBLEBUFFER 5
+#define GLX_DRAWABLE_TYPE 0x8010
+#define GLX_EVENT_MASK 0x801F
+#define GLX_EXTENSIONS 0x3
+#define GLX_EXTENSION_NAME "GLX"
+#define GLX_FBCONFIG_ID 0x8013
+#define GLX_FRONT_LEFT_BUFFER_BIT 0x00000001
+#define GLX_FRONT_RIGHT_BUFFER_BIT 0x00000002
+#define GLX_GRAY_SCALE 0x8006
+#define GLX_GREEN_SIZE 9
+#define GLX_HEIGHT 0x801E
+#define GLX_LARGEST_PBUFFER 0x801C
+#define GLX_LEVEL 3
+#define GLX_MAX_PBUFFER_HEIGHT 0x8017
+#define GLX_MAX_PBUFFER_PIXELS 0x8018
+#define GLX_MAX_PBUFFER_WIDTH 0x8016
+#define GLX_MAX_SWAP_INTERVAL_EXT 0x20F2
+#define GLX_NONE 0x8000
+#define GLX_NON_CONFORMANT_CONFIG 0x800D
+#define GLX_NO_EXTENSION 3
+#define GLX_PBUFFER 0x8023
+#define GLX_PBUFFER_BIT 0x00000004
+#define GLX_PBUFFER_CLOBBER_MASK 0x08000000
+#define GLX_PBUFFER_HEIGHT 0x8040
+#define GLX_PBUFFER_WIDTH 0x8041
+#define GLX_PIXMAP_BIT 0x00000002
+#define GLX_PRESERVED_CONTENTS 0x801B
+#define GLX_PSEUDO_COLOR 0x8004
+#define GLX_PbufferClobber 0
+#define GLX_RED_SIZE 8
+#define GLX_RENDER_TYPE 0x8011
+#define GLX_RGBA 4
+#define GLX_RGBA_BIT 0x00000001
+#define GLX_RGBA_TYPE 0x8014
+#define GLX_SAMPLES 100001
+#define GLX_SAMPLE_BUFFERS 100000
+#define GLX_SAVED 0x8021
+#define GLX_SCREEN 0x800C
+#define GLX_SLOW_CONFIG 0x8001
+#define GLX_STATIC_COLOR 0x8005
+#define GLX_STATIC_GRAY 0x8007
+#define GLX_STENCIL_BUFFER_BIT 0x00000040
+#define GLX_STENCIL_SIZE 13
+#define GLX_STEREO 6
+#define GLX_SWAP_INTERVAL_EXT 0x20F1
+#define GLX_TRANSPARENT_ALPHA_VALUE 0x28
+#define GLX_TRANSPARENT_BLUE_VALUE 0x27
+#define GLX_TRANSPARENT_GREEN_VALUE 0x26
+#define GLX_TRANSPARENT_INDEX 0x8009
+#define GLX_TRANSPARENT_INDEX_VALUE 0x24
+#define GLX_TRANSPARENT_RED_VALUE 0x25
+#define GLX_TRANSPARENT_RGB 0x8008
+#define GLX_TRANSPARENT_TYPE 0x23
+#define GLX_TRUE_COLOR 0x8002
+#define GLX_USE_GL 1
+#define GLX_VENDOR 0x1
+#define GLX_VERSION 0x2
+#define GLX_VISUAL_ID 0x800B
+#define GLX_WIDTH 0x801D
+#define GLX_WINDOW 0x8022
+#define GLX_WINDOW_BIT 0x00000001
+#define GLX_X_RENDERABLE 0x8012
+#define GLX_X_VISUAL_TYPE 0x22
+#define __GLX_NUMBER_EVENTS 17
+
+
+#ifndef GLEXT_64_TYPES_DEFINED
+/* This code block is duplicated in glext.h, so must be protected */
+#define GLEXT_64_TYPES_DEFINED
+/* Define int32_t, int64_t, and uint64_t types for UST/MSC */
+/* (as used in the GLX_OML_sync_control extension). */
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
+#include <inttypes.h>
+#elif defined(__sun__) || defined(__digital__)
+#include <inttypes.h>
+#if defined(__STDC__)
+#if defined(__arch64__) || defined(_LP64)
+typedef long int int64_t;
+typedef unsigned long int uint64_t;
+#else
+typedef long long int int64_t;
+typedef unsigned long long int uint64_t;
+#endif /* __arch64__ */
+#endif /* __STDC__ */
+#elif defined( __VMS ) || defined(__sgi)
+#include <inttypes.h>
+#elif defined(__SCO__) || defined(__USLC__)
+#include <stdint.h>
+#elif defined(__UNIXOS2__) || defined(__SOL64__)
+typedef long int int32_t;
+typedef long long int int64_t;
+typedef unsigned long long int uint64_t;
+#elif defined(_WIN32) && defined(__GNUC__)
+#include <stdint.h>
+#elif defined(_WIN32)
+typedef __int32 int32_t;
+typedef __int64 int64_t;
+typedef unsigned __int64 uint64_t;
+#else
+/* Fallback if nothing above works */
+#include <inttypes.h>
+#endif
+#endif
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+
+#else
+
+#endif
+
+#if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && (__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ > 1060)
+
+#else
+
+#endif
+
+
+
+
+
+
+
+typedef XID GLXFBConfigID;
+typedef struct __GLXFBConfigRec *GLXFBConfig;
+typedef XID GLXContextID;
+typedef struct __GLXcontextRec *GLXContext;
+typedef XID GLXPixmap;
+typedef XID GLXDrawable;
+typedef XID GLXWindow;
+typedef XID GLXPbuffer;
+typedef void (GLAD_API_PTR *__GLXextFuncPtr)(void);
+typedef XID GLXVideoCaptureDeviceNV;
+typedef unsigned int GLXVideoDeviceNV;
+typedef XID GLXVideoSourceSGIX;
+typedef XID GLXFBConfigIDSGIX;
+typedef struct __GLXFBConfigRec *GLXFBConfigSGIX;
+typedef XID GLXPbufferSGIX;
+typedef struct {
+    int event_type;             /* GLX_DAMAGED or GLX_SAVED */
+    int draw_type;              /* GLX_WINDOW or GLX_PBUFFER */
+    unsigned long serial;       /* # of last request processed by server */
+    Bool send_event;            /* true if this came for SendEvent request */
+    Display *display;           /* display the event was read from */
+    GLXDrawable drawable;       /* XID of Drawable */
+    unsigned int buffer_mask;   /* mask indicating which buffers are affected */
+    unsigned int aux_buffer;    /* which aux buffer was affected */
+    int x, y;
+    int width, height;
+    int count;                  /* if nonzero, at least this many more */
+} GLXPbufferClobberEvent;
+typedef struct {
+    int type;
+    unsigned long serial;       /* # of last request processed by server */
+    Bool send_event;            /* true if this came from a SendEvent request */
+    Display *display;           /* Display the event was read from */
+    GLXDrawable drawable;       /* drawable on which event was requested in event mask */
+    int event_type;
+    int64_t ust;
+    int64_t msc;
+    int64_t sbc;
+} GLXBufferSwapComplete;
+typedef union __GLXEvent {
+    GLXPbufferClobberEvent glxpbufferclobber;
+    GLXBufferSwapComplete glxbufferswapcomplete;
+    long pad[24];
+} GLXEvent;
+typedef struct {
+    int type;
+    unsigned long serial;
+    Bool send_event;
+    Display *display;
+    int extension;
+    int evtype;
+    GLXDrawable window;
+    Bool stereo_tree;
+} GLXStereoNotifyEventEXT;
+typedef struct {
+    int type;
+    unsigned long serial;   /* # of last request processed by server */
+    Bool send_event;        /* true if this came for SendEvent request */
+    Display *display;       /* display the event was read from */
+    GLXDrawable drawable;   /* i.d. of Drawable */
+    int event_type;         /* GLX_DAMAGED_SGIX or GLX_SAVED_SGIX */
+    int draw_type;          /* GLX_WINDOW_SGIX or GLX_PBUFFER_SGIX */
+    unsigned int mask;      /* mask indicating which buffers are affected*/
+    int x, y;
+    int width, height;
+    int count;              /* if nonzero, at least this many more */
+} GLXBufferClobberEventSGIX;
+typedef struct {
+    char    pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int     networkId;
+} GLXHyperpipeNetworkSGIX;
+typedef struct {
+    char    pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int     channel;
+    unsigned int participationType;
+    int     timeSlice;
+} GLXHyperpipeConfigSGIX;
+typedef struct {
+    char pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int srcXOrigin, srcYOrigin, srcWidth, srcHeight;
+    int destXOrigin, destYOrigin, destWidth, destHeight;
+} GLXPipeRect;
+typedef struct {
+    char pipeName[80]; /* Should be [GLX_HYPERPIPE_PIPE_NAME_LENGTH_SGIX] */
+    int XOrigin, YOrigin, maxHeight, maxWidth;
+} GLXPipeRectLimits;
+
+
+#define GLX_VERSION_1_0 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_0;
+#define GLX_VERSION_1_1 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_1;
+#define GLX_VERSION_1_2 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_2;
+#define GLX_VERSION_1_3 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_3;
+#define GLX_VERSION_1_4 1
+GLAD_API_CALL int GLAD_GLX_VERSION_1_4;
+#define GLX_EXT_swap_control 1
+GLAD_API_CALL int GLAD_GLX_EXT_swap_control;
+#define GLX_MESA_swap_control 1
+GLAD_API_CALL int GLAD_GLX_MESA_swap_control;
+
+
+typedef GLXFBConfig * (GLAD_API_PTR *PFNGLXCHOOSEFBCONFIGPROC)(Display * dpy, int screen, const int * attrib_list, int * nelements);
+typedef XVisualInfo * (GLAD_API_PTR *PFNGLXCHOOSEVISUALPROC)(Display * dpy, int screen, int * attribList);
+typedef void (GLAD_API_PTR *PFNGLXCOPYCONTEXTPROC)(Display * dpy, GLXContext src, GLXContext dst, unsigned long mask);
+typedef GLXContext (GLAD_API_PTR *PFNGLXCREATECONTEXTPROC)(Display * dpy, XVisualInfo * vis, GLXContext shareList, Bool direct);
+typedef GLXPixmap (GLAD_API_PTR *PFNGLXCREATEGLXPIXMAPPROC)(Display * dpy, XVisualInfo * visual, Pixmap pixmap);
+typedef GLXContext (GLAD_API_PTR *PFNGLXCREATENEWCONTEXTPROC)(Display * dpy, GLXFBConfig config, int render_type, GLXContext share_list, Bool direct);
+typedef GLXPbuffer (GLAD_API_PTR *PFNGLXCREATEPBUFFERPROC)(Display * dpy, GLXFBConfig config, const int * attrib_list);
+typedef GLXPixmap (GLAD_API_PTR *PFNGLXCREATEPIXMAPPROC)(Display * dpy, GLXFBConfig config, Pixmap pixmap, const int * attrib_list);
+typedef GLXWindow (GLAD_API_PTR *PFNGLXCREATEWINDOWPROC)(Display * dpy, GLXFBConfig config, Window win, const int * attrib_list);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYCONTEXTPROC)(Display * dpy, GLXContext ctx);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYGLXPIXMAPPROC)(Display * dpy, GLXPixmap pixmap);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYPBUFFERPROC)(Display * dpy, GLXPbuffer pbuf);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYPIXMAPPROC)(Display * dpy, GLXPixmap pixmap);
+typedef void (GLAD_API_PTR *PFNGLXDESTROYWINDOWPROC)(Display * dpy, GLXWindow win);
+typedef const char * (GLAD_API_PTR *PFNGLXGETCLIENTSTRINGPROC)(Display * dpy, int name);
+typedef int (GLAD_API_PTR *PFNGLXGETCONFIGPROC)(Display * dpy, XVisualInfo * visual, int attrib, int * value);
+typedef GLXContext (GLAD_API_PTR *PFNGLXGETCURRENTCONTEXTPROC)(void);
+typedef Display * (GLAD_API_PTR *PFNGLXGETCURRENTDISPLAYPROC)(void);
+typedef GLXDrawable (GLAD_API_PTR *PFNGLXGETCURRENTDRAWABLEPROC)(void);
+typedef GLXDrawable (GLAD_API_PTR *PFNGLXGETCURRENTREADDRAWABLEPROC)(void);
+typedef int (GLAD_API_PTR *PFNGLXGETFBCONFIGATTRIBPROC)(Display * dpy, GLXFBConfig config, int attribute, int * value);
+typedef GLXFBConfig * (GLAD_API_PTR *PFNGLXGETFBCONFIGSPROC)(Display * dpy, int screen, int * nelements);
+typedef __GLXextFuncPtr (GLAD_API_PTR *PFNGLXGETPROCADDRESSPROC)(const GLubyte * procName);
+typedef void (GLAD_API_PTR *PFNGLXGETSELECTEDEVENTPROC)(Display * dpy, GLXDrawable draw, unsigned long * event_mask);
+typedef int (GLAD_API_PTR *PFNGLXGETSWAPINTERVALMESAPROC)(void);
+typedef XVisualInfo * (GLAD_API_PTR *PFNGLXGETVISUALFROMFBCONFIGPROC)(Display * dpy, GLXFBConfig config);
+typedef Bool (GLAD_API_PTR *PFNGLXISDIRECTPROC)(Display * dpy, GLXContext ctx);
+typedef Bool (GLAD_API_PTR *PFNGLXMAKECONTEXTCURRENTPROC)(Display * dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx);
+typedef Bool (GLAD_API_PTR *PFNGLXMAKECURRENTPROC)(Display * dpy, GLXDrawable drawable, GLXContext ctx);
+typedef int (GLAD_API_PTR *PFNGLXQUERYCONTEXTPROC)(Display * dpy, GLXContext ctx, int attribute, int * value);
+typedef void (GLAD_API_PTR *PFNGLXQUERYDRAWABLEPROC)(Display * dpy, GLXDrawable draw, int attribute, unsigned int * value);
+typedef Bool (GLAD_API_PTR *PFNGLXQUERYEXTENSIONPROC)(Display * dpy, int * errorb, int * event);
+typedef const char * (GLAD_API_PTR *PFNGLXQUERYEXTENSIONSSTRINGPROC)(Display * dpy, int screen);
+typedef const char * (GLAD_API_PTR *PFNGLXQUERYSERVERSTRINGPROC)(Display * dpy, int screen, int name);
+typedef Bool (GLAD_API_PTR *PFNGLXQUERYVERSIONPROC)(Display * dpy, int * maj, int * min);
+typedef void (GLAD_API_PTR *PFNGLXSELECTEVENTPROC)(Display * dpy, GLXDrawable draw, unsigned long event_mask);
+typedef void (GLAD_API_PTR *PFNGLXSWAPBUFFERSPROC)(Display * dpy, GLXDrawable drawable);
+typedef void (GLAD_API_PTR *PFNGLXSWAPINTERVALEXTPROC)(Display * dpy, GLXDrawable drawable, int interval);
+typedef int (GLAD_API_PTR *PFNGLXSWAPINTERVALMESAPROC)(unsigned int interval);
+typedef void (GLAD_API_PTR *PFNGLXUSEXFONTPROC)(Font font, int first, int count, int list);
+typedef void (GLAD_API_PTR *PFNGLXWAITGLPROC)(void);
+typedef void (GLAD_API_PTR *PFNGLXWAITXPROC)(void);
+
+GLAD_API_CALL PFNGLXCHOOSEFBCONFIGPROC glad_glXChooseFBConfig;
+#define glXChooseFBConfig glad_glXChooseFBConfig
+GLAD_API_CALL PFNGLXCHOOSEVISUALPROC glad_glXChooseVisual;
+#define glXChooseVisual glad_glXChooseVisual
+GLAD_API_CALL PFNGLXCOPYCONTEXTPROC glad_glXCopyContext;
+#define glXCopyContext glad_glXCopyContext
+GLAD_API_CALL PFNGLXCREATECONTEXTPROC glad_glXCreateContext;
+#define glXCreateContext glad_glXCreateContext
+GLAD_API_CALL PFNGLXCREATEGLXPIXMAPPROC glad_glXCreateGLXPixmap;
+#define glXCreateGLXPixmap glad_glXCreateGLXPixmap
+GLAD_API_CALL PFNGLXCREATENEWCONTEXTPROC glad_glXCreateNewContext;
+#define glXCreateNewContext glad_glXCreateNewContext
+GLAD_API_CALL PFNGLXCREATEPBUFFERPROC glad_glXCreatePbuffer;
+#define glXCreatePbuffer glad_glXCreatePbuffer
+GLAD_API_CALL PFNGLXCREATEPIXMAPPROC glad_glXCreatePixmap;
+#define glXCreatePixmap glad_glXCreatePixmap
+GLAD_API_CALL PFNGLXCREATEWINDOWPROC glad_glXCreateWindow;
+#define glXCreateWindow glad_glXCreateWindow
+GLAD_API_CALL PFNGLXDESTROYCONTEXTPROC glad_glXDestroyContext;
+#define glXDestroyContext glad_glXDestroyContext
+GLAD_API_CALL PFNGLXDESTROYGLXPIXMAPPROC glad_glXDestroyGLXPixmap;
+#define glXDestroyGLXPixmap glad_glXDestroyGLXPixmap
+GLAD_API_CALL PFNGLXDESTROYPBUFFERPROC glad_glXDestroyPbuffer;
+#define glXDestroyPbuffer glad_glXDestroyPbuffer
+GLAD_API_CALL PFNGLXDESTROYPIXMAPPROC glad_glXDestroyPixmap;
+#define glXDestroyPixmap glad_glXDestroyPixmap
+GLAD_API_CALL PFNGLXDESTROYWINDOWPROC glad_glXDestroyWindow;
+#define glXDestroyWindow glad_glXDestroyWindow
+GLAD_API_CALL PFNGLXGETCLIENTSTRINGPROC glad_glXGetClientString;
+#define glXGetClientString glad_glXGetClientString
+GLAD_API_CALL PFNGLXGETCONFIGPROC glad_glXGetConfig;
+#define glXGetConfig glad_glXGetConfig
+GLAD_API_CALL PFNGLXGETCURRENTCONTEXTPROC glad_glXGetCurrentContext;
+#define glXGetCurrentContext glad_glXGetCurrentContext
+GLAD_API_CALL PFNGLXGETCURRENTDISPLAYPROC glad_glXGetCurrentDisplay;
+#define glXGetCurrentDisplay glad_glXGetCurrentDisplay
+GLAD_API_CALL PFNGLXGETCURRENTDRAWABLEPROC glad_glXGetCurrentDrawable;
+#define glXGetCurrentDrawable glad_glXGetCurrentDrawable
+GLAD_API_CALL PFNGLXGETCURRENTREADDRAWABLEPROC glad_glXGetCurrentReadDrawable;
+#define glXGetCurrentReadDrawable glad_glXGetCurrentReadDrawable
+GLAD_API_CALL PFNGLXGETFBCONFIGATTRIBPROC glad_glXGetFBConfigAttrib;
+#define glXGetFBConfigAttrib glad_glXGetFBConfigAttrib
+GLAD_API_CALL PFNGLXGETFBCONFIGSPROC glad_glXGetFBConfigs;
+#define glXGetFBConfigs glad_glXGetFBConfigs
+GLAD_API_CALL PFNGLXGETPROCADDRESSPROC glad_glXGetProcAddress;
+#define glXGetProcAddress glad_glXGetProcAddress
+GLAD_API_CALL PFNGLXGETSELECTEDEVENTPROC glad_glXGetSelectedEvent;
+#define glXGetSelectedEvent glad_glXGetSelectedEvent
+GLAD_API_CALL PFNGLXGETSWAPINTERVALMESAPROC glad_glXGetSwapIntervalMESA;
+#define glXGetSwapIntervalMESA glad_glXGetSwapIntervalMESA
+GLAD_API_CALL PFNGLXGETVISUALFROMFBCONFIGPROC glad_glXGetVisualFromFBConfig;
+#define glXGetVisualFromFBConfig glad_glXGetVisualFromFBConfig
+GLAD_API_CALL PFNGLXISDIRECTPROC glad_glXIsDirect;
+#define glXIsDirect glad_glXIsDirect
+GLAD_API_CALL PFNGLXMAKECONTEXTCURRENTPROC glad_glXMakeContextCurrent;
+#define glXMakeContextCurrent glad_glXMakeContextCurrent
+GLAD_API_CALL PFNGLXMAKECURRENTPROC glad_glXMakeCurrent;
+#define glXMakeCurrent glad_glXMakeCurrent
+GLAD_API_CALL PFNGLXQUERYCONTEXTPROC glad_glXQueryContext;
+#define glXQueryContext glad_glXQueryContext
+GLAD_API_CALL PFNGLXQUERYDRAWABLEPROC glad_glXQueryDrawable;
+#define glXQueryDrawable glad_glXQueryDrawable
+GLAD_API_CALL PFNGLXQUERYEXTENSIONPROC glad_glXQueryExtension;
+#define glXQueryExtension glad_glXQueryExtension
+GLAD_API_CALL PFNGLXQUERYEXTENSIONSSTRINGPROC glad_glXQueryExtensionsString;
+#define glXQueryExtensionsString glad_glXQueryExtensionsString
+GLAD_API_CALL PFNGLXQUERYSERVERSTRINGPROC glad_glXQueryServerString;
+#define glXQueryServerString glad_glXQueryServerString
+GLAD_API_CALL PFNGLXQUERYVERSIONPROC glad_glXQueryVersion;
+#define glXQueryVersion glad_glXQueryVersion
+GLAD_API_CALL PFNGLXSELECTEVENTPROC glad_glXSelectEvent;
+#define glXSelectEvent glad_glXSelectEvent
+GLAD_API_CALL PFNGLXSWAPBUFFERSPROC glad_glXSwapBuffers;
+#define glXSwapBuffers glad_glXSwapBuffers
+GLAD_API_CALL PFNGLXSWAPINTERVALEXTPROC glad_glXSwapIntervalEXT;
+#define glXSwapIntervalEXT glad_glXSwapIntervalEXT
+GLAD_API_CALL PFNGLXSWAPINTERVALMESAPROC glad_glXSwapIntervalMESA;
+#define glXSwapIntervalMESA glad_glXSwapIntervalMESA
+GLAD_API_CALL PFNGLXUSEXFONTPROC glad_glXUseXFont;
+#define glXUseXFont glad_glXUseXFont
+GLAD_API_CALL PFNGLXWAITGLPROC glad_glXWaitGL;
+#define glXWaitGL glad_glXWaitGL
+GLAD_API_CALL PFNGLXWAITXPROC glad_glXWaitX;
+#define glXWaitX glad_glXWaitX
+
+
+
+
+
+GLAD_API_CALL int gladLoadGLXUserPtr(Display *display, int screen, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadGLX(Display *display, int screen, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/wgl.h glmark2-2021.02/src/glad/include/glad/wgl.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/glad/wgl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/glad/wgl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,298 @@
+/**
+ * Loader generated by glad 2.0.0-beta on Wed Dec 25 23:13:13 2019
+ *
+ * Generator: C/C++
+ * Specification: wgl
+ * Extensions: 3
+ *
+ * APIs:
+ *  - wgl=1.0
+ *
+ * Options:
+ *  - MX_GLOBAL = False
+ *  - ON_DEMAND = False
+ *  - LOADER = False
+ *  - ALIAS = False
+ *  - HEADER_ONLY = False
+ *  - DEBUG = False
+ *  - MX = False
+ *
+ * Commandline:
+ *    --api='wgl=1.0' --extensions='WGL_ARB_extensions_string,WGL_EXT_extensions_string,WGL_EXT_swap_control' c
+ *
+ * Online:
+ *    http://glad.sh/#api=wgl%3D1.0&extensions=WGL_ARB_extensions_string%2CWGL_EXT_extensions_string%2CWGL_EXT_swap_control&generator=c&options=
+ *
+ */
+
+#ifndef GLAD_WGL_H_
+#define GLAD_WGL_H_
+
+#include <windows.h>
+#include <glad/gl.h>
+
+#define GLAD_WGL
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef GLAD_PLATFORM_H_
+#define GLAD_PLATFORM_H_
+
+#ifndef GLAD_PLATFORM_WIN32
+  #if defined(_WIN32) || defined(__WIN32__) || defined(WIN32) || defined(__MINGW32__)
+    #define GLAD_PLATFORM_WIN32 1
+  #else
+    #define GLAD_PLATFORM_WIN32 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_APPLE
+  #ifdef __APPLE__
+    #define GLAD_PLATFORM_APPLE 1
+  #else
+    #define GLAD_PLATFORM_APPLE 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_EMSCRIPTEN
+  #ifdef __EMSCRIPTEN__
+    #define GLAD_PLATFORM_EMSCRIPTEN 1
+  #else
+    #define GLAD_PLATFORM_EMSCRIPTEN 0
+  #endif
+#endif
+
+#ifndef GLAD_PLATFORM_UWP
+  #if defined(_MSC_VER) && !defined(GLAD_INTERNAL_HAVE_WINAPIFAMILY)
+    #ifdef __has_include
+      #if __has_include(<winapifamily.h>)
+        #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+      #endif
+    #elif _MSC_VER >= 1700 && !_USING_V110_SDK71_
+      #define GLAD_INTERNAL_HAVE_WINAPIFAMILY 1
+    #endif
+  #endif
+
+  #ifdef GLAD_INTERNAL_HAVE_WINAPIFAMILY
+    #include <winapifamily.h>
+    #if !WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_DESKTOP) && WINAPI_FAMILY_PARTITION(WINAPI_PARTITION_APP)
+      #define GLAD_PLATFORM_UWP 1
+    #endif
+  #endif
+
+  #ifndef GLAD_PLATFORM_UWP
+    #define GLAD_PLATFORM_UWP 0
+  #endif
+#endif
+
+#ifdef __GNUC__
+  #define GLAD_GNUC_EXTENSION __extension__
+#else
+  #define GLAD_GNUC_EXTENSION
+#endif
+
+#ifndef GLAD_API_CALL
+  #if defined(GLAD_API_CALL_EXPORT)
+    #if GLAD_PLATFORM_WIN32 || defined(__CYGWIN__)
+      #if defined(GLAD_API_CALL_EXPORT_BUILD)
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllexport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllexport) extern
+        #endif
+      #else
+        #if defined(__GNUC__)
+          #define GLAD_API_CALL __attribute__ ((dllimport)) extern
+        #else
+          #define GLAD_API_CALL __declspec(dllimport) extern
+        #endif
+      #endif
+    #elif defined(__GNUC__) && defined(GLAD_API_CALL_EXPORT_BUILD)
+      #define GLAD_API_CALL __attribute__ ((visibility ("default"))) extern
+    #else
+      #define GLAD_API_CALL extern
+    #endif
+  #else
+    #define GLAD_API_CALL extern
+  #endif
+#endif
+
+#ifdef APIENTRY
+  #define GLAD_API_PTR APIENTRY
+#elif GLAD_PLATFORM_WIN32
+  #define GLAD_API_PTR __stdcall
+#else
+  #define GLAD_API_PTR
+#endif
+
+#ifndef GLAPI
+#define GLAPI GLAD_API_CALL
+#endif
+
+#ifndef GLAPIENTRY
+#define GLAPIENTRY GLAD_API_PTR
+#endif
+
+#define GLAD_MAKE_VERSION(major, minor) (major * 10000 + minor)
+#define GLAD_VERSION_MAJOR(version) (version / 10000)
+#define GLAD_VERSION_MINOR(version) (version % 10000)
+
+#define GLAD_GENERATOR_VERSION "2.0.0-beta"
+
+typedef void (*GLADapiproc)(void);
+
+typedef GLADapiproc (*GLADloadfunc)(const char *name);
+typedef GLADapiproc (*GLADuserptrloadfunc)(void *userptr, const char *name);
+
+typedef void (*GLADprecallback)(const char *name, GLADapiproc apiproc, int len_args, ...);
+typedef void (*GLADpostcallback)(void *ret, const char *name, GLADapiproc apiproc, int len_args, ...);
+
+#endif /* GLAD_PLATFORM_H_ */
+
+#define WGL_FONT_LINES 0
+#define WGL_FONT_POLYGONS 1
+#define WGL_SWAP_MAIN_PLANE 0x00000001
+#define WGL_SWAP_OVERLAY1 0x00000002
+#define WGL_SWAP_OVERLAY10 0x00000400
+#define WGL_SWAP_OVERLAY11 0x00000800
+#define WGL_SWAP_OVERLAY12 0x00001000
+#define WGL_SWAP_OVERLAY13 0x00002000
+#define WGL_SWAP_OVERLAY14 0x00004000
+#define WGL_SWAP_OVERLAY15 0x00008000
+#define WGL_SWAP_OVERLAY2 0x00000004
+#define WGL_SWAP_OVERLAY3 0x00000008
+#define WGL_SWAP_OVERLAY4 0x00000010
+#define WGL_SWAP_OVERLAY5 0x00000020
+#define WGL_SWAP_OVERLAY6 0x00000040
+#define WGL_SWAP_OVERLAY7 0x00000080
+#define WGL_SWAP_OVERLAY8 0x00000100
+#define WGL_SWAP_OVERLAY9 0x00000200
+#define WGL_SWAP_UNDERLAY1 0x00010000
+#define WGL_SWAP_UNDERLAY10 0x02000000
+#define WGL_SWAP_UNDERLAY11 0x04000000
+#define WGL_SWAP_UNDERLAY12 0x08000000
+#define WGL_SWAP_UNDERLAY13 0x10000000
+#define WGL_SWAP_UNDERLAY14 0x20000000
+#define WGL_SWAP_UNDERLAY15 0x40000000
+#define WGL_SWAP_UNDERLAY2 0x00020000
+#define WGL_SWAP_UNDERLAY3 0x00040000
+#define WGL_SWAP_UNDERLAY4 0x00080000
+#define WGL_SWAP_UNDERLAY5 0x00100000
+#define WGL_SWAP_UNDERLAY6 0x00200000
+#define WGL_SWAP_UNDERLAY7 0x00400000
+#define WGL_SWAP_UNDERLAY8 0x00800000
+#define WGL_SWAP_UNDERLAY9 0x01000000
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+struct _GPU_DEVICE {
+    DWORD  cb;
+    CHAR   DeviceName[32];
+    CHAR   DeviceString[128];
+    DWORD  Flags;
+    RECT   rcVirtualScreen;
+};
+DECLARE_HANDLE(HPBUFFERARB);
+DECLARE_HANDLE(HPBUFFEREXT);
+DECLARE_HANDLE(HVIDEOOUTPUTDEVICENV);
+DECLARE_HANDLE(HPVIDEODEV);
+DECLARE_HANDLE(HPGPUNV);
+DECLARE_HANDLE(HGPUNV);
+DECLARE_HANDLE(HVIDEOINPUTDEVICENV);
+typedef struct _GPU_DEVICE GPU_DEVICE;
+typedef struct _GPU_DEVICE *PGPU_DEVICE;
+
+
+#define WGL_VERSION_1_0 1
+GLAD_API_CALL int GLAD_WGL_VERSION_1_0;
+#define WGL_ARB_extensions_string 1
+GLAD_API_CALL int GLAD_WGL_ARB_extensions_string;
+#define WGL_EXT_extensions_string 1
+GLAD_API_CALL int GLAD_WGL_EXT_extensions_string;
+#define WGL_EXT_swap_control 1
+GLAD_API_CALL int GLAD_WGL_EXT_swap_control;
+
+
+typedef int (GLAD_API_PTR *PFNCHOOSEPIXELFORMATPROC)(HDC hDc, const PIXELFORMATDESCRIPTOR * pPfd);
+typedef int (GLAD_API_PTR *PFNDESCRIBEPIXELFORMATPROC)(HDC hdc, int ipfd, UINT cjpfd, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef UINT (GLAD_API_PTR *PFNGETENHMETAFILEPIXELFORMATPROC)(HENHMETAFILE hemf, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef int (GLAD_API_PTR *PFNGETPIXELFORMATPROC)(HDC hdc);
+typedef BOOL (GLAD_API_PTR *PFNSETPIXELFORMATPROC)(HDC hdc, int ipfd, const PIXELFORMATDESCRIPTOR * ppfd);
+typedef BOOL (GLAD_API_PTR *PFNSWAPBUFFERSPROC)(HDC hdc);
+typedef BOOL (GLAD_API_PTR *PFNWGLCOPYCONTEXTPROC)(HGLRC hglrcSrc, HGLRC hglrcDst, UINT mask);
+typedef HGLRC (GLAD_API_PTR *PFNWGLCREATECONTEXTPROC)(HDC hDc);
+typedef HGLRC (GLAD_API_PTR *PFNWGLCREATELAYERCONTEXTPROC)(HDC hDc, int level);
+typedef BOOL (GLAD_API_PTR *PFNWGLDELETECONTEXTPROC)(HGLRC oldContext);
+typedef BOOL (GLAD_API_PTR *PFNWGLDESCRIBELAYERPLANEPROC)(HDC hDc, int pixelFormat, int layerPlane, UINT nBytes, const LAYERPLANEDESCRIPTOR * plpd);
+typedef HGLRC (GLAD_API_PTR *PFNWGLGETCURRENTCONTEXTPROC)(void);
+typedef HDC (GLAD_API_PTR *PFNWGLGETCURRENTDCPROC)(void);
+typedef const char * (GLAD_API_PTR *PFNWGLGETEXTENSIONSSTRINGARBPROC)(HDC hdc);
+typedef const char * (GLAD_API_PTR *PFNWGLGETEXTENSIONSSTRINGEXTPROC)(void);
+typedef int (GLAD_API_PTR *PFNWGLGETLAYERPALETTEENTRIESPROC)(HDC hdc, int iLayerPlane, int iStart, int cEntries, const COLORREF * pcr);
+typedef PROC (GLAD_API_PTR *PFNWGLGETPROCADDRESSPROC)(LPCSTR lpszProc);
+typedef int (GLAD_API_PTR *PFNWGLGETSWAPINTERVALEXTPROC)(void);
+typedef BOOL (GLAD_API_PTR *PFNWGLMAKECURRENTPROC)(HDC hDc, HGLRC newContext);
+typedef BOOL (GLAD_API_PTR *PFNWGLREALIZELAYERPALETTEPROC)(HDC hdc, int iLayerPlane, BOOL bRealize);
+typedef int (GLAD_API_PTR *PFNWGLSETLAYERPALETTEENTRIESPROC)(HDC hdc, int iLayerPlane, int iStart, int cEntries, const COLORREF * pcr);
+typedef BOOL (GLAD_API_PTR *PFNWGLSHARELISTSPROC)(HGLRC hrcSrvShare, HGLRC hrcSrvSource);
+typedef BOOL (GLAD_API_PTR *PFNWGLSWAPINTERVALEXTPROC)(int interval);
+typedef BOOL (GLAD_API_PTR *PFNWGLSWAPLAYERBUFFERSPROC)(HDC hdc, UINT fuFlags);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSAPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTBITMAPSWPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESAPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+typedef BOOL (GLAD_API_PTR *PFNWGLUSEFONTOUTLINESWPROC)(HDC hDC, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf);
+
+GLAD_API_CALL PFNWGLGETEXTENSIONSSTRINGARBPROC glad_wglGetExtensionsStringARB;
+#define wglGetExtensionsStringARB glad_wglGetExtensionsStringARB
+GLAD_API_CALL PFNWGLGETEXTENSIONSSTRINGEXTPROC glad_wglGetExtensionsStringEXT;
+#define wglGetExtensionsStringEXT glad_wglGetExtensionsStringEXT
+GLAD_API_CALL PFNWGLGETSWAPINTERVALEXTPROC glad_wglGetSwapIntervalEXT;
+#define wglGetSwapIntervalEXT glad_wglGetSwapIntervalEXT
+GLAD_API_CALL PFNWGLSWAPINTERVALEXTPROC glad_wglSwapIntervalEXT;
+#define wglSwapIntervalEXT glad_wglSwapIntervalEXT
+
+
+
+
+
+GLAD_API_CALL int gladLoadWGLUserPtr(HDC hdc, GLADuserptrloadfunc load, void *userptr);
+GLAD_API_CALL int gladLoadWGL(HDC hdc, GLADloadfunc load);
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/include/KHR/khrplatform.h glmark2-2021.02/src/glad/include/KHR/khrplatform.h
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/include/KHR/khrplatform.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/include/KHR/khrplatform.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,290 @@
+#ifndef __khrplatform_h_
+#define __khrplatform_h_
+
+/*
+** Copyright (c) 2008-2018 The Khronos Group Inc.
+**
+** Permission is hereby granted, free of charge, to any person obtaining a
+** copy of this software and/or associated documentation files (the
+** "Materials"), to deal in the Materials without restriction, including
+** without limitation the rights to use, copy, modify, merge, publish,
+** distribute, sublicense, and/or sell copies of the Materials, and to
+** permit persons to whom the Materials are furnished to do so, subject to
+** the following conditions:
+**
+** The above copyright notice and this permission notice shall be included
+** in all copies or substantial portions of the Materials.
+**
+** THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+** EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+** MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+** IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+** CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+** TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+** MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.
+*/
+
+/* Khronos platform-specific types and definitions.
+ *
+ * The master copy of khrplatform.h is maintained in the Khronos EGL
+ * Registry repository at https://github.com/KhronosGroup/EGL-Registry
+ * The last semantic modification to khrplatform.h was at commit ID:
+ *      67a3e0864c2d75ea5287b9f3d2eb74a745936692
+ *
+ * Adopters may modify this file to suit their platform. Adopters are
+ * encouraged to submit platform specific modifications to the Khronos
+ * group so that they can be included in future versions of this file.
+ * Please submit changes by filing pull requests or issues on
+ * the EGL Registry repository linked above.
+ *
+ *
+ * See the Implementer's Guidelines for information about where this file
+ * should be located on your system and for more details of its use:
+ *    http://www.khronos.org/registry/implementers_guide.pdf
+ *
+ * This file should be included as
+ *        #include <KHR/khrplatform.h>
+ * by Khronos client API header files that use its types and defines.
+ *
+ * The types in khrplatform.h should only be used to define API-specific types.
+ *
+ * Types defined in khrplatform.h:
+ *    khronos_int8_t              signed   8  bit
+ *    khronos_uint8_t             unsigned 8  bit
+ *    khronos_int16_t             signed   16 bit
+ *    khronos_uint16_t            unsigned 16 bit
+ *    khronos_int32_t             signed   32 bit
+ *    khronos_uint32_t            unsigned 32 bit
+ *    khronos_int64_t             signed   64 bit
+ *    khronos_uint64_t            unsigned 64 bit
+ *    khronos_intptr_t            signed   same number of bits as a pointer
+ *    khronos_uintptr_t           unsigned same number of bits as a pointer
+ *    khronos_ssize_t             signed   size
+ *    khronos_usize_t             unsigned size
+ *    khronos_float_t             signed   32 bit floating point
+ *    khronos_time_ns_t           unsigned 64 bit time in nanoseconds
+ *    khronos_utime_nanoseconds_t unsigned time interval or absolute time in
+ *                                         nanoseconds
+ *    khronos_stime_nanoseconds_t signed time interval in nanoseconds
+ *    khronos_boolean_enum_t      enumerated boolean type. This should
+ *      only be used as a base type when a client API's boolean type is
+ *      an enum. Client APIs which use an integer or other type for
+ *      booleans cannot use this as the base type for their boolean.
+ *
+ * Tokens defined in khrplatform.h:
+ *
+ *    KHRONOS_FALSE, KHRONOS_TRUE Enumerated boolean false/true values.
+ *
+ *    KHRONOS_SUPPORT_INT64 is 1 if 64 bit integers are supported; otherwise 0.
+ *    KHRONOS_SUPPORT_FLOAT is 1 if floats are supported; otherwise 0.
+ *
+ * Calling convention macros defined in this file:
+ *    KHRONOS_APICALL
+ *    KHRONOS_APIENTRY
+ *    KHRONOS_APIATTRIBUTES
+ *
+ * These may be used in function prototypes as:
+ *
+ *      KHRONOS_APICALL void KHRONOS_APIENTRY funcname(
+ *                                  int arg1,
+ *                                  int arg2) KHRONOS_APIATTRIBUTES;
+ */
+
+#if defined(__SCITECH_SNAP__) && !defined(KHRONOS_STATIC)
+#   define KHRONOS_STATIC 1
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APICALL
+ *-------------------------------------------------------------------------
+ * This precedes the return type of the function in the function prototype.
+ */
+#if defined(KHRONOS_STATIC)
+    /* If the preprocessor constant KHRONOS_STATIC is defined, make the
+     * header compatible with static linking. */
+#   define KHRONOS_APICALL
+#elif defined(_WIN32)
+#   define KHRONOS_APICALL __declspec(dllimport)
+#elif defined (__SYMBIAN32__)
+#   define KHRONOS_APICALL IMPORT_C
+#elif defined(__ANDROID__)
+#   define KHRONOS_APICALL __attribute__((visibility("default")))
+#else
+#   define KHRONOS_APICALL
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APIENTRY
+ *-------------------------------------------------------------------------
+ * This follows the return type of the function  and precedes the function
+ * name in the function prototype.
+ */
+#if defined(_WIN32) && !defined(_WIN32_WCE) && !defined(KHRONOS_STATIC)
+    /* Win32 but not WinCE */
+#   define KHRONOS_APIENTRY __stdcall
+#else
+#   define KHRONOS_APIENTRY
+#endif
+
+/*-------------------------------------------------------------------------
+ * Definition of KHRONOS_APIATTRIBUTES
+ *-------------------------------------------------------------------------
+ * This follows the closing parenthesis of the function prototype arguments.
+ */
+#if defined (__ARMCC_2__)
+#define KHRONOS_APIATTRIBUTES __softfp
+#else
+#define KHRONOS_APIATTRIBUTES
+#endif
+
+/*-------------------------------------------------------------------------
+ * basic type definitions
+ *-----------------------------------------------------------------------*/
+#if (defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L) || defined(__GNUC__) || defined(__SCO__) || defined(__USLC__)
+
+
+/*
+ * Using <stdint.h>
+ */
+#include <stdint.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(__VMS ) || defined(__sgi)
+
+/*
+ * Using <inttypes.h>
+ */
+#include <inttypes.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(_WIN32) && !defined(__SCITECH_SNAP__)
+
+/*
+ * Win32
+ */
+typedef __int32                 khronos_int32_t;
+typedef unsigned __int32        khronos_uint32_t;
+typedef __int64                 khronos_int64_t;
+typedef unsigned __int64        khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif defined(__sun__) || defined(__digital__)
+
+/*
+ * Sun or Digital
+ */
+typedef int                     khronos_int32_t;
+typedef unsigned int            khronos_uint32_t;
+#if defined(__arch64__) || defined(_LP64)
+typedef long int                khronos_int64_t;
+typedef unsigned long int       khronos_uint64_t;
+#else
+typedef long long int           khronos_int64_t;
+typedef unsigned long long int  khronos_uint64_t;
+#endif /* __arch64__ */
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#elif 0
+
+/*
+ * Hypothetical platform with no float or int64 support
+ */
+typedef int                     khronos_int32_t;
+typedef unsigned int            khronos_uint32_t;
+#define KHRONOS_SUPPORT_INT64   0
+#define KHRONOS_SUPPORT_FLOAT   0
+
+#else
+
+/*
+ * Generic fallback
+ */
+#include <stdint.h>
+typedef int32_t                 khronos_int32_t;
+typedef uint32_t                khronos_uint32_t;
+typedef int64_t                 khronos_int64_t;
+typedef uint64_t                khronos_uint64_t;
+#define KHRONOS_SUPPORT_INT64   1
+#define KHRONOS_SUPPORT_FLOAT   1
+
+#endif
+
+
+/*
+ * Types that are (so far) the same on all platforms
+ */
+typedef signed   char          khronos_int8_t;
+typedef unsigned char          khronos_uint8_t;
+typedef signed   short int     khronos_int16_t;
+typedef unsigned short int     khronos_uint16_t;
+
+/*
+ * Types that differ between LLP64 and LP64 architectures - in LLP64,
+ * pointers are 64 bits, but 'long' is still 32 bits. Win64 appears
+ * to be the only LLP64 architecture in current use.
+ */
+#ifdef _WIN64
+typedef signed   long long int khronos_intptr_t;
+typedef unsigned long long int khronos_uintptr_t;
+typedef signed   long long int khronos_ssize_t;
+typedef unsigned long long int khronos_usize_t;
+#else
+typedef signed   long  int     khronos_intptr_t;
+typedef unsigned long  int     khronos_uintptr_t;
+typedef signed   long  int     khronos_ssize_t;
+typedef unsigned long  int     khronos_usize_t;
+#endif
+
+#if KHRONOS_SUPPORT_FLOAT
+/*
+ * Float type
+ */
+typedef          float         khronos_float_t;
+#endif
+
+#if KHRONOS_SUPPORT_INT64
+/* Time types
+ *
+ * These types can be used to represent a time interval in nanoseconds or
+ * an absolute Unadjusted System Time.  Unadjusted System Time is the number
+ * of nanoseconds since some arbitrary system event (e.g. since the last
+ * time the system booted).  The Unadjusted System Time is an unsigned
+ * 64 bit value that wraps back to 0 every 584 years.  Time intervals
+ * may be either signed or unsigned.
+ */
+typedef khronos_uint64_t       khronos_utime_nanoseconds_t;
+typedef khronos_int64_t        khronos_stime_nanoseconds_t;
+#endif
+
+/*
+ * Dummy value used to pad enum types to 32 bits.
+ */
+#ifndef KHRONOS_MAX_ENUM
+#define KHRONOS_MAX_ENUM 0x7FFFFFFF
+#endif
+
+/*
+ * Enumerated boolean type
+ *
+ * Values other than zero should be considered to be true.  Therefore
+ * comparisons should not be made against KHRONOS_TRUE.
+ */
+typedef enum {
+    KHRONOS_FALSE = 0,
+    KHRONOS_TRUE  = 1,
+    KHRONOS_BOOLEAN_ENUM_FORCE_SIZE = KHRONOS_MAX_ENUM
+} khronos_boolean_enum_t;
+
+#endif /* __khrplatform_h_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/README.md glmark2-2021.02/src/glad/README.md
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/README.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/README.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,7 @@
+This directory contains a copy of the
+[glad2](https://github.com/Dav1dde/glad/tree/glad2) library.
+
+The contents of this directory can be recreated (at the latest available
+upstream version) with [this][permalink] permalink.
+
+[permalink]: http://gen.glad.sh/#profile=gl%3Dcompatibility%2Cgles1%3Dcommon&api=egl%3D1.5%2Cgl%3D3.3%2Cgles2%3D2.0%2Cglx%3D1.4%2Cwgl%3D1.0&extensions=EGL_EXT_platform_base%2CEGL_KHR_platform_gbm%2CEGL_KHR_platform_wayland%2CEGL_KHR_platform_x11%2CGLX_EXT_swap_control%2CGLX_MESA_swap_control%2CGL_OES_mapbuffer%2CGL_OES_required_internalformat%2CWGL_ARB_extensions_string%2CWGL_EXT_extensions_string%2CWGL_EXT_swap_control&generator=c
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/src/egl.c glmark2-2021.02/src/glad/src/egl.c
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/src/egl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/src/egl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,258 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/egl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_EGL_VERSION_1_0 = 0;
+int GLAD_EGL_VERSION_1_1 = 0;
+int GLAD_EGL_VERSION_1_2 = 0;
+int GLAD_EGL_VERSION_1_3 = 0;
+int GLAD_EGL_VERSION_1_4 = 0;
+int GLAD_EGL_VERSION_1_5 = 0;
+int GLAD_EGL_EXT_platform_base = 0;
+int GLAD_EGL_KHR_platform_gbm = 0;
+int GLAD_EGL_KHR_platform_wayland = 0;
+int GLAD_EGL_KHR_platform_x11 = 0;
+
+
+
+PFNEGLBINDAPIPROC glad_eglBindAPI = NULL;
+PFNEGLBINDTEXIMAGEPROC glad_eglBindTexImage = NULL;
+PFNEGLCHOOSECONFIGPROC glad_eglChooseConfig = NULL;
+PFNEGLCLIENTWAITSYNCPROC glad_eglClientWaitSync = NULL;
+PFNEGLCOPYBUFFERSPROC glad_eglCopyBuffers = NULL;
+PFNEGLCREATECONTEXTPROC glad_eglCreateContext = NULL;
+PFNEGLCREATEIMAGEPROC glad_eglCreateImage = NULL;
+PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC glad_eglCreatePbufferFromClientBuffer = NULL;
+PFNEGLCREATEPBUFFERSURFACEPROC glad_eglCreatePbufferSurface = NULL;
+PFNEGLCREATEPIXMAPSURFACEPROC glad_eglCreatePixmapSurface = NULL;
+PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC glad_eglCreatePlatformPixmapSurface = NULL;
+PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC glad_eglCreatePlatformPixmapSurfaceEXT = NULL;
+PFNEGLCREATEPLATFORMWINDOWSURFACEPROC glad_eglCreatePlatformWindowSurface = NULL;
+PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC glad_eglCreatePlatformWindowSurfaceEXT = NULL;
+PFNEGLCREATESYNCPROC glad_eglCreateSync = NULL;
+PFNEGLCREATEWINDOWSURFACEPROC glad_eglCreateWindowSurface = NULL;
+PFNEGLDESTROYCONTEXTPROC glad_eglDestroyContext = NULL;
+PFNEGLDESTROYIMAGEPROC glad_eglDestroyImage = NULL;
+PFNEGLDESTROYSURFACEPROC glad_eglDestroySurface = NULL;
+PFNEGLDESTROYSYNCPROC glad_eglDestroySync = NULL;
+PFNEGLGETCONFIGATTRIBPROC glad_eglGetConfigAttrib = NULL;
+PFNEGLGETCONFIGSPROC glad_eglGetConfigs = NULL;
+PFNEGLGETCURRENTCONTEXTPROC glad_eglGetCurrentContext = NULL;
+PFNEGLGETCURRENTDISPLAYPROC glad_eglGetCurrentDisplay = NULL;
+PFNEGLGETCURRENTSURFACEPROC glad_eglGetCurrentSurface = NULL;
+PFNEGLGETDISPLAYPROC glad_eglGetDisplay = NULL;
+PFNEGLGETERRORPROC glad_eglGetError = NULL;
+PFNEGLGETPLATFORMDISPLAYPROC glad_eglGetPlatformDisplay = NULL;
+PFNEGLGETPLATFORMDISPLAYEXTPROC glad_eglGetPlatformDisplayEXT = NULL;
+PFNEGLGETPROCADDRESSPROC glad_eglGetProcAddress = NULL;
+PFNEGLGETSYNCATTRIBPROC glad_eglGetSyncAttrib = NULL;
+PFNEGLINITIALIZEPROC glad_eglInitialize = NULL;
+PFNEGLMAKECURRENTPROC glad_eglMakeCurrent = NULL;
+PFNEGLQUERYAPIPROC glad_eglQueryAPI = NULL;
+PFNEGLQUERYCONTEXTPROC glad_eglQueryContext = NULL;
+PFNEGLQUERYSTRINGPROC glad_eglQueryString = NULL;
+PFNEGLQUERYSURFACEPROC glad_eglQuerySurface = NULL;
+PFNEGLRELEASETEXIMAGEPROC glad_eglReleaseTexImage = NULL;
+PFNEGLRELEASETHREADPROC glad_eglReleaseThread = NULL;
+PFNEGLSURFACEATTRIBPROC glad_eglSurfaceAttrib = NULL;
+PFNEGLSWAPBUFFERSPROC glad_eglSwapBuffers = NULL;
+PFNEGLSWAPINTERVALPROC glad_eglSwapInterval = NULL;
+PFNEGLTERMINATEPROC glad_eglTerminate = NULL;
+PFNEGLWAITCLIENTPROC glad_eglWaitClient = NULL;
+PFNEGLWAITGLPROC glad_eglWaitGL = NULL;
+PFNEGLWAITNATIVEPROC glad_eglWaitNative = NULL;
+PFNEGLWAITSYNCPROC glad_eglWaitSync = NULL;
+
+
+static void glad_egl_load_EGL_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_0) return;
+    glad_eglChooseConfig = (PFNEGLCHOOSECONFIGPROC) load(userptr, "eglChooseConfig");
+    glad_eglCopyBuffers = (PFNEGLCOPYBUFFERSPROC) load(userptr, "eglCopyBuffers");
+    glad_eglCreateContext = (PFNEGLCREATECONTEXTPROC) load(userptr, "eglCreateContext");
+    glad_eglCreatePbufferSurface = (PFNEGLCREATEPBUFFERSURFACEPROC) load(userptr, "eglCreatePbufferSurface");
+    glad_eglCreatePixmapSurface = (PFNEGLCREATEPIXMAPSURFACEPROC) load(userptr, "eglCreatePixmapSurface");
+    glad_eglCreateWindowSurface = (PFNEGLCREATEWINDOWSURFACEPROC) load(userptr, "eglCreateWindowSurface");
+    glad_eglDestroyContext = (PFNEGLDESTROYCONTEXTPROC) load(userptr, "eglDestroyContext");
+    glad_eglDestroySurface = (PFNEGLDESTROYSURFACEPROC) load(userptr, "eglDestroySurface");
+    glad_eglGetConfigAttrib = (PFNEGLGETCONFIGATTRIBPROC) load(userptr, "eglGetConfigAttrib");
+    glad_eglGetConfigs = (PFNEGLGETCONFIGSPROC) load(userptr, "eglGetConfigs");
+    glad_eglGetCurrentDisplay = (PFNEGLGETCURRENTDISPLAYPROC) load(userptr, "eglGetCurrentDisplay");
+    glad_eglGetCurrentSurface = (PFNEGLGETCURRENTSURFACEPROC) load(userptr, "eglGetCurrentSurface");
+    glad_eglGetDisplay = (PFNEGLGETDISPLAYPROC) load(userptr, "eglGetDisplay");
+    glad_eglGetError = (PFNEGLGETERRORPROC) load(userptr, "eglGetError");
+    glad_eglGetProcAddress = (PFNEGLGETPROCADDRESSPROC) load(userptr, "eglGetProcAddress");
+    glad_eglInitialize = (PFNEGLINITIALIZEPROC) load(userptr, "eglInitialize");
+    glad_eglMakeCurrent = (PFNEGLMAKECURRENTPROC) load(userptr, "eglMakeCurrent");
+    glad_eglQueryContext = (PFNEGLQUERYCONTEXTPROC) load(userptr, "eglQueryContext");
+    glad_eglQueryString = (PFNEGLQUERYSTRINGPROC) load(userptr, "eglQueryString");
+    glad_eglQuerySurface = (PFNEGLQUERYSURFACEPROC) load(userptr, "eglQuerySurface");
+    glad_eglSwapBuffers = (PFNEGLSWAPBUFFERSPROC) load(userptr, "eglSwapBuffers");
+    glad_eglTerminate = (PFNEGLTERMINATEPROC) load(userptr, "eglTerminate");
+    glad_eglWaitGL = (PFNEGLWAITGLPROC) load(userptr, "eglWaitGL");
+    glad_eglWaitNative = (PFNEGLWAITNATIVEPROC) load(userptr, "eglWaitNative");
+}
+static void glad_egl_load_EGL_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_1) return;
+    glad_eglBindTexImage = (PFNEGLBINDTEXIMAGEPROC) load(userptr, "eglBindTexImage");
+    glad_eglReleaseTexImage = (PFNEGLRELEASETEXIMAGEPROC) load(userptr, "eglReleaseTexImage");
+    glad_eglSurfaceAttrib = (PFNEGLSURFACEATTRIBPROC) load(userptr, "eglSurfaceAttrib");
+    glad_eglSwapInterval = (PFNEGLSWAPINTERVALPROC) load(userptr, "eglSwapInterval");
+}
+static void glad_egl_load_EGL_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_2) return;
+    glad_eglBindAPI = (PFNEGLBINDAPIPROC) load(userptr, "eglBindAPI");
+    glad_eglCreatePbufferFromClientBuffer = (PFNEGLCREATEPBUFFERFROMCLIENTBUFFERPROC) load(userptr, "eglCreatePbufferFromClientBuffer");
+    glad_eglQueryAPI = (PFNEGLQUERYAPIPROC) load(userptr, "eglQueryAPI");
+    glad_eglReleaseThread = (PFNEGLRELEASETHREADPROC) load(userptr, "eglReleaseThread");
+    glad_eglWaitClient = (PFNEGLWAITCLIENTPROC) load(userptr, "eglWaitClient");
+}
+static void glad_egl_load_EGL_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_4) return;
+    glad_eglGetCurrentContext = (PFNEGLGETCURRENTCONTEXTPROC) load(userptr, "eglGetCurrentContext");
+}
+static void glad_egl_load_EGL_VERSION_1_5( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_VERSION_1_5) return;
+    glad_eglClientWaitSync = (PFNEGLCLIENTWAITSYNCPROC) load(userptr, "eglClientWaitSync");
+    glad_eglCreateImage = (PFNEGLCREATEIMAGEPROC) load(userptr, "eglCreateImage");
+    glad_eglCreatePlatformPixmapSurface = (PFNEGLCREATEPLATFORMPIXMAPSURFACEPROC) load(userptr, "eglCreatePlatformPixmapSurface");
+    glad_eglCreatePlatformWindowSurface = (PFNEGLCREATEPLATFORMWINDOWSURFACEPROC) load(userptr, "eglCreatePlatformWindowSurface");
+    glad_eglCreateSync = (PFNEGLCREATESYNCPROC) load(userptr, "eglCreateSync");
+    glad_eglDestroyImage = (PFNEGLDESTROYIMAGEPROC) load(userptr, "eglDestroyImage");
+    glad_eglDestroySync = (PFNEGLDESTROYSYNCPROC) load(userptr, "eglDestroySync");
+    glad_eglGetPlatformDisplay = (PFNEGLGETPLATFORMDISPLAYPROC) load(userptr, "eglGetPlatformDisplay");
+    glad_eglGetSyncAttrib = (PFNEGLGETSYNCATTRIBPROC) load(userptr, "eglGetSyncAttrib");
+    glad_eglWaitSync = (PFNEGLWAITSYNCPROC) load(userptr, "eglWaitSync");
+}
+static void glad_egl_load_EGL_EXT_platform_base( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_EGL_EXT_platform_base) return;
+    glad_eglCreatePlatformPixmapSurfaceEXT = (PFNEGLCREATEPLATFORMPIXMAPSURFACEEXTPROC) load(userptr, "eglCreatePlatformPixmapSurfaceEXT");
+    glad_eglCreatePlatformWindowSurfaceEXT = (PFNEGLCREATEPLATFORMWINDOWSURFACEEXTPROC) load(userptr, "eglCreatePlatformWindowSurfaceEXT");
+    glad_eglGetPlatformDisplayEXT = (PFNEGLGETPLATFORMDISPLAYEXTPROC) load(userptr, "eglGetPlatformDisplayEXT");
+}
+
+
+
+static int glad_egl_get_extensions(EGLDisplay display, const char **extensions) {
+    *extensions = eglQueryString(display, EGL_EXTENSIONS);
+
+    return extensions != NULL;
+}
+
+static int glad_egl_has_extension(const char *extensions, const char *ext) {
+    const char *loc;
+    const char *terminator;
+    if(extensions == NULL) {
+        return 0;
+    }
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL) {
+            return 0;
+        }
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0')) {
+            return 1;
+        }
+        extensions = terminator;
+    }
+}
+
+static GLADapiproc glad_egl_get_proc_from_userptr(void *userptr, const char *name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_egl_find_extensions_egl(EGLDisplay display) {
+    const char *extensions;
+    if (!glad_egl_get_extensions(display, &extensions)) return 0;
+
+    GLAD_EGL_EXT_platform_base = glad_egl_has_extension(extensions, "EGL_EXT_platform_base");
+    GLAD_EGL_KHR_platform_gbm = glad_egl_has_extension(extensions, "EGL_KHR_platform_gbm");
+    GLAD_EGL_KHR_platform_wayland = glad_egl_has_extension(extensions, "EGL_KHR_platform_wayland");
+    GLAD_EGL_KHR_platform_x11 = glad_egl_has_extension(extensions, "EGL_KHR_platform_x11");
+
+    return 1;
+}
+
+static int glad_egl_find_core_egl(EGLDisplay display) {
+    int major, minor;
+    const char *version;
+
+    if (display == NULL) {
+        display = EGL_NO_DISPLAY; /* this is usually NULL, better safe than sorry */
+    }
+    if (display == EGL_NO_DISPLAY) {
+        display = eglGetCurrentDisplay();
+    }
+#ifdef EGL_VERSION_1_4
+    if (display == EGL_NO_DISPLAY) {
+        display = eglGetDisplay(EGL_DEFAULT_DISPLAY);
+    }
+#endif
+#ifndef EGL_VERSION_1_5
+    if (display == EGL_NO_DISPLAY) {
+        return 0;
+    }
+#endif
+
+    version = eglQueryString(display, EGL_VERSION);
+    (void) eglGetError();
+
+    if (version == NULL) {
+        major = 1;
+        minor = 0;
+    } else {
+        GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+    }
+
+    GLAD_EGL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_EGL_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_EGL_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_EGL_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_EGL_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    GLAD_EGL_VERSION_1_5 = (major == 1 && minor >= 5) || major > 1;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadEGLUserPtr(EGLDisplay display, GLADuserptrloadfunc load, void* userptr) {
+    int version;
+    eglGetDisplay = (PFNEGLGETDISPLAYPROC) load(userptr, "eglGetDisplay");
+    eglGetCurrentDisplay = (PFNEGLGETCURRENTDISPLAYPROC) load(userptr, "eglGetCurrentDisplay");
+    eglQueryString = (PFNEGLQUERYSTRINGPROC) load(userptr, "eglQueryString");
+    eglGetError = (PFNEGLGETERRORPROC) load(userptr, "eglGetError");
+    if (eglGetDisplay == NULL || eglGetCurrentDisplay == NULL || eglQueryString == NULL || eglGetError == NULL) return 0;
+
+    version = glad_egl_find_core_egl(display);
+    if (!version) return 0;
+    glad_egl_load_EGL_VERSION_1_0(load, userptr);
+    glad_egl_load_EGL_VERSION_1_1(load, userptr);
+    glad_egl_load_EGL_VERSION_1_2(load, userptr);
+    glad_egl_load_EGL_VERSION_1_4(load, userptr);
+    glad_egl_load_EGL_VERSION_1_5(load, userptr);
+
+    if (!glad_egl_find_extensions_egl(display)) return 0;
+    glad_egl_load_EGL_EXT_platform_base(load, userptr);
+
+    return version;
+}
+
+int gladLoadEGL(EGLDisplay display, GLADloadfunc load) {
+    return gladLoadEGLUserPtr(display, glad_egl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/src/gl.c glmark2-2021.02/src/glad/src/gl.c
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/src/gl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/src/gl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1709 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/gl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GL_VERSION_1_0 = 0;
+int GLAD_GL_VERSION_1_1 = 0;
+int GLAD_GL_VERSION_1_2 = 0;
+int GLAD_GL_VERSION_1_3 = 0;
+int GLAD_GL_VERSION_1_4 = 0;
+int GLAD_GL_VERSION_1_5 = 0;
+int GLAD_GL_VERSION_2_0 = 0;
+int GLAD_GL_VERSION_2_1 = 0;
+int GLAD_GL_VERSION_3_0 = 0;
+int GLAD_GL_VERSION_3_1 = 0;
+int GLAD_GL_VERSION_3_2 = 0;
+int GLAD_GL_VERSION_3_3 = 0;
+
+
+
+PFNGLACCUMPROC glad_glAccum = NULL;
+PFNGLACTIVETEXTUREPROC glad_glActiveTexture = NULL;
+PFNGLALPHAFUNCPROC glad_glAlphaFunc = NULL;
+PFNGLARETEXTURESRESIDENTPROC glad_glAreTexturesResident = NULL;
+PFNGLARRAYELEMENTPROC glad_glArrayElement = NULL;
+PFNGLATTACHSHADERPROC glad_glAttachShader = NULL;
+PFNGLBEGINPROC glad_glBegin = NULL;
+PFNGLBEGINCONDITIONALRENDERPROC glad_glBeginConditionalRender = NULL;
+PFNGLBEGINQUERYPROC glad_glBeginQuery = NULL;
+PFNGLBEGINTRANSFORMFEEDBACKPROC glad_glBeginTransformFeedback = NULL;
+PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation = NULL;
+PFNGLBINDBUFFERPROC glad_glBindBuffer = NULL;
+PFNGLBINDBUFFERBASEPROC glad_glBindBufferBase = NULL;
+PFNGLBINDBUFFERRANGEPROC glad_glBindBufferRange = NULL;
+PFNGLBINDFRAGDATALOCATIONPROC glad_glBindFragDataLocation = NULL;
+PFNGLBINDFRAGDATALOCATIONINDEXEDPROC glad_glBindFragDataLocationIndexed = NULL;
+PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer = NULL;
+PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer = NULL;
+PFNGLBINDSAMPLERPROC glad_glBindSampler = NULL;
+PFNGLBINDTEXTUREPROC glad_glBindTexture = NULL;
+PFNGLBINDVERTEXARRAYPROC glad_glBindVertexArray = NULL;
+PFNGLBITMAPPROC glad_glBitmap = NULL;
+PFNGLBLENDCOLORPROC glad_glBlendColor = NULL;
+PFNGLBLENDEQUATIONPROC glad_glBlendEquation = NULL;
+PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate = NULL;
+PFNGLBLENDFUNCPROC glad_glBlendFunc = NULL;
+PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate = NULL;
+PFNGLBLITFRAMEBUFFERPROC glad_glBlitFramebuffer = NULL;
+PFNGLBUFFERDATAPROC glad_glBufferData = NULL;
+PFNGLBUFFERSUBDATAPROC glad_glBufferSubData = NULL;
+PFNGLCALLLISTPROC glad_glCallList = NULL;
+PFNGLCALLLISTSPROC glad_glCallLists = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus = NULL;
+PFNGLCLAMPCOLORPROC glad_glClampColor = NULL;
+PFNGLCLEARPROC glad_glClear = NULL;
+PFNGLCLEARACCUMPROC glad_glClearAccum = NULL;
+PFNGLCLEARBUFFERFIPROC glad_glClearBufferfi = NULL;
+PFNGLCLEARBUFFERFVPROC glad_glClearBufferfv = NULL;
+PFNGLCLEARBUFFERIVPROC glad_glClearBufferiv = NULL;
+PFNGLCLEARBUFFERUIVPROC glad_glClearBufferuiv = NULL;
+PFNGLCLEARCOLORPROC glad_glClearColor = NULL;
+PFNGLCLEARDEPTHPROC glad_glClearDepth = NULL;
+PFNGLCLEARINDEXPROC glad_glClearIndex = NULL;
+PFNGLCLEARSTENCILPROC glad_glClearStencil = NULL;
+PFNGLCLIENTACTIVETEXTUREPROC glad_glClientActiveTexture = NULL;
+PFNGLCLIENTWAITSYNCPROC glad_glClientWaitSync = NULL;
+PFNGLCLIPPLANEPROC glad_glClipPlane = NULL;
+PFNGLCOLOR3BPROC glad_glColor3b = NULL;
+PFNGLCOLOR3BVPROC glad_glColor3bv = NULL;
+PFNGLCOLOR3DPROC glad_glColor3d = NULL;
+PFNGLCOLOR3DVPROC glad_glColor3dv = NULL;
+PFNGLCOLOR3FPROC glad_glColor3f = NULL;
+PFNGLCOLOR3FVPROC glad_glColor3fv = NULL;
+PFNGLCOLOR3IPROC glad_glColor3i = NULL;
+PFNGLCOLOR3IVPROC glad_glColor3iv = NULL;
+PFNGLCOLOR3SPROC glad_glColor3s = NULL;
+PFNGLCOLOR3SVPROC glad_glColor3sv = NULL;
+PFNGLCOLOR3UBPROC glad_glColor3ub = NULL;
+PFNGLCOLOR3UBVPROC glad_glColor3ubv = NULL;
+PFNGLCOLOR3UIPROC glad_glColor3ui = NULL;
+PFNGLCOLOR3UIVPROC glad_glColor3uiv = NULL;
+PFNGLCOLOR3USPROC glad_glColor3us = NULL;
+PFNGLCOLOR3USVPROC glad_glColor3usv = NULL;
+PFNGLCOLOR4BPROC glad_glColor4b = NULL;
+PFNGLCOLOR4BVPROC glad_glColor4bv = NULL;
+PFNGLCOLOR4DPROC glad_glColor4d = NULL;
+PFNGLCOLOR4DVPROC glad_glColor4dv = NULL;
+PFNGLCOLOR4FPROC glad_glColor4f = NULL;
+PFNGLCOLOR4FVPROC glad_glColor4fv = NULL;
+PFNGLCOLOR4IPROC glad_glColor4i = NULL;
+PFNGLCOLOR4IVPROC glad_glColor4iv = NULL;
+PFNGLCOLOR4SPROC glad_glColor4s = NULL;
+PFNGLCOLOR4SVPROC glad_glColor4sv = NULL;
+PFNGLCOLOR4UBPROC glad_glColor4ub = NULL;
+PFNGLCOLOR4UBVPROC glad_glColor4ubv = NULL;
+PFNGLCOLOR4UIPROC glad_glColor4ui = NULL;
+PFNGLCOLOR4UIVPROC glad_glColor4uiv = NULL;
+PFNGLCOLOR4USPROC glad_glColor4us = NULL;
+PFNGLCOLOR4USVPROC glad_glColor4usv = NULL;
+PFNGLCOLORMASKPROC glad_glColorMask = NULL;
+PFNGLCOLORMASKIPROC glad_glColorMaski = NULL;
+PFNGLCOLORMATERIALPROC glad_glColorMaterial = NULL;
+PFNGLCOLORP3UIPROC glad_glColorP3ui = NULL;
+PFNGLCOLORP3UIVPROC glad_glColorP3uiv = NULL;
+PFNGLCOLORP4UIPROC glad_glColorP4ui = NULL;
+PFNGLCOLORP4UIVPROC glad_glColorP4uiv = NULL;
+PFNGLCOLORPOINTERPROC glad_glColorPointer = NULL;
+PFNGLCOMPILESHADERPROC glad_glCompileShader = NULL;
+PFNGLCOMPRESSEDTEXIMAGE1DPROC glad_glCompressedTexImage1D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D = NULL;
+PFNGLCOMPRESSEDTEXIMAGE3DPROC glad_glCompressedTexImage3D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC glad_glCompressedTexSubImage1D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC glad_glCompressedTexSubImage3D = NULL;
+PFNGLCOPYBUFFERSUBDATAPROC glad_glCopyBufferSubData = NULL;
+PFNGLCOPYPIXELSPROC glad_glCopyPixels = NULL;
+PFNGLCOPYTEXIMAGE1DPROC glad_glCopyTexImage1D = NULL;
+PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE1DPROC glad_glCopyTexSubImage1D = NULL;
+PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE3DPROC glad_glCopyTexSubImage3D = NULL;
+PFNGLCREATEPROGRAMPROC glad_glCreateProgram = NULL;
+PFNGLCREATESHADERPROC glad_glCreateShader = NULL;
+PFNGLCULLFACEPROC glad_glCullFace = NULL;
+PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers = NULL;
+PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers = NULL;
+PFNGLDELETELISTSPROC glad_glDeleteLists = NULL;
+PFNGLDELETEPROGRAMPROC glad_glDeleteProgram = NULL;
+PFNGLDELETEQUERIESPROC glad_glDeleteQueries = NULL;
+PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers = NULL;
+PFNGLDELETESAMPLERSPROC glad_glDeleteSamplers = NULL;
+PFNGLDELETESHADERPROC glad_glDeleteShader = NULL;
+PFNGLDELETESYNCPROC glad_glDeleteSync = NULL;
+PFNGLDELETETEXTURESPROC glad_glDeleteTextures = NULL;
+PFNGLDELETEVERTEXARRAYSPROC glad_glDeleteVertexArrays = NULL;
+PFNGLDEPTHFUNCPROC glad_glDepthFunc = NULL;
+PFNGLDEPTHMASKPROC glad_glDepthMask = NULL;
+PFNGLDEPTHRANGEPROC glad_glDepthRange = NULL;
+PFNGLDETACHSHADERPROC glad_glDetachShader = NULL;
+PFNGLDISABLEPROC glad_glDisable = NULL;
+PFNGLDISABLECLIENTSTATEPROC glad_glDisableClientState = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray = NULL;
+PFNGLDISABLEIPROC glad_glDisablei = NULL;
+PFNGLDRAWARRAYSPROC glad_glDrawArrays = NULL;
+PFNGLDRAWARRAYSINSTANCEDPROC glad_glDrawArraysInstanced = NULL;
+PFNGLDRAWBUFFERPROC glad_glDrawBuffer = NULL;
+PFNGLDRAWBUFFERSPROC glad_glDrawBuffers = NULL;
+PFNGLDRAWELEMENTSPROC glad_glDrawElements = NULL;
+PFNGLDRAWELEMENTSBASEVERTEXPROC glad_glDrawElementsBaseVertex = NULL;
+PFNGLDRAWELEMENTSINSTANCEDPROC glad_glDrawElementsInstanced = NULL;
+PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC glad_glDrawElementsInstancedBaseVertex = NULL;
+PFNGLDRAWPIXELSPROC glad_glDrawPixels = NULL;
+PFNGLDRAWRANGEELEMENTSPROC glad_glDrawRangeElements = NULL;
+PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC glad_glDrawRangeElementsBaseVertex = NULL;
+PFNGLEDGEFLAGPROC glad_glEdgeFlag = NULL;
+PFNGLEDGEFLAGPOINTERPROC glad_glEdgeFlagPointer = NULL;
+PFNGLEDGEFLAGVPROC glad_glEdgeFlagv = NULL;
+PFNGLENABLEPROC glad_glEnable = NULL;
+PFNGLENABLECLIENTSTATEPROC glad_glEnableClientState = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray = NULL;
+PFNGLENABLEIPROC glad_glEnablei = NULL;
+PFNGLENDPROC glad_glEnd = NULL;
+PFNGLENDCONDITIONALRENDERPROC glad_glEndConditionalRender = NULL;
+PFNGLENDLISTPROC glad_glEndList = NULL;
+PFNGLENDQUERYPROC glad_glEndQuery = NULL;
+PFNGLENDTRANSFORMFEEDBACKPROC glad_glEndTransformFeedback = NULL;
+PFNGLEVALCOORD1DPROC glad_glEvalCoord1d = NULL;
+PFNGLEVALCOORD1DVPROC glad_glEvalCoord1dv = NULL;
+PFNGLEVALCOORD1FPROC glad_glEvalCoord1f = NULL;
+PFNGLEVALCOORD1FVPROC glad_glEvalCoord1fv = NULL;
+PFNGLEVALCOORD2DPROC glad_glEvalCoord2d = NULL;
+PFNGLEVALCOORD2DVPROC glad_glEvalCoord2dv = NULL;
+PFNGLEVALCOORD2FPROC glad_glEvalCoord2f = NULL;
+PFNGLEVALCOORD2FVPROC glad_glEvalCoord2fv = NULL;
+PFNGLEVALMESH1PROC glad_glEvalMesh1 = NULL;
+PFNGLEVALMESH2PROC glad_glEvalMesh2 = NULL;
+PFNGLEVALPOINT1PROC glad_glEvalPoint1 = NULL;
+PFNGLEVALPOINT2PROC glad_glEvalPoint2 = NULL;
+PFNGLFEEDBACKBUFFERPROC glad_glFeedbackBuffer = NULL;
+PFNGLFENCESYNCPROC glad_glFenceSync = NULL;
+PFNGLFINISHPROC glad_glFinish = NULL;
+PFNGLFLUSHPROC glad_glFlush = NULL;
+PFNGLFLUSHMAPPEDBUFFERRANGEPROC glad_glFlushMappedBufferRange = NULL;
+PFNGLFOGCOORDPOINTERPROC glad_glFogCoordPointer = NULL;
+PFNGLFOGCOORDDPROC glad_glFogCoordd = NULL;
+PFNGLFOGCOORDDVPROC glad_glFogCoorddv = NULL;
+PFNGLFOGCOORDFPROC glad_glFogCoordf = NULL;
+PFNGLFOGCOORDFVPROC glad_glFogCoordfv = NULL;
+PFNGLFOGFPROC glad_glFogf = NULL;
+PFNGLFOGFVPROC glad_glFogfv = NULL;
+PFNGLFOGIPROC glad_glFogi = NULL;
+PFNGLFOGIVPROC glad_glFogiv = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer = NULL;
+PFNGLFRAMEBUFFERTEXTUREPROC glad_glFramebufferTexture = NULL;
+PFNGLFRAMEBUFFERTEXTURE1DPROC glad_glFramebufferTexture1D = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D = NULL;
+PFNGLFRAMEBUFFERTEXTURE3DPROC glad_glFramebufferTexture3D = NULL;
+PFNGLFRAMEBUFFERTEXTURELAYERPROC glad_glFramebufferTextureLayer = NULL;
+PFNGLFRONTFACEPROC glad_glFrontFace = NULL;
+PFNGLFRUSTUMPROC glad_glFrustum = NULL;
+PFNGLGENBUFFERSPROC glad_glGenBuffers = NULL;
+PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers = NULL;
+PFNGLGENLISTSPROC glad_glGenLists = NULL;
+PFNGLGENQUERIESPROC glad_glGenQueries = NULL;
+PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers = NULL;
+PFNGLGENSAMPLERSPROC glad_glGenSamplers = NULL;
+PFNGLGENTEXTURESPROC glad_glGenTextures = NULL;
+PFNGLGENVERTEXARRAYSPROC glad_glGenVertexArrays = NULL;
+PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap = NULL;
+PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib = NULL;
+PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform = NULL;
+PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC glad_glGetActiveUniformBlockName = NULL;
+PFNGLGETACTIVEUNIFORMBLOCKIVPROC glad_glGetActiveUniformBlockiv = NULL;
+PFNGLGETACTIVEUNIFORMNAMEPROC glad_glGetActiveUniformName = NULL;
+PFNGLGETACTIVEUNIFORMSIVPROC glad_glGetActiveUniformsiv = NULL;
+PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders = NULL;
+PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation = NULL;
+PFNGLGETBOOLEANI_VPROC glad_glGetBooleani_v = NULL;
+PFNGLGETBOOLEANVPROC glad_glGetBooleanv = NULL;
+PFNGLGETBUFFERPARAMETERI64VPROC glad_glGetBufferParameteri64v = NULL;
+PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv = NULL;
+PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv = NULL;
+PFNGLGETBUFFERSUBDATAPROC glad_glGetBufferSubData = NULL;
+PFNGLGETCLIPPLANEPROC glad_glGetClipPlane = NULL;
+PFNGLGETCOMPRESSEDTEXIMAGEPROC glad_glGetCompressedTexImage = NULL;
+PFNGLGETDOUBLEVPROC glad_glGetDoublev = NULL;
+PFNGLGETERRORPROC glad_glGetError = NULL;
+PFNGLGETFLOATVPROC glad_glGetFloatv = NULL;
+PFNGLGETFRAGDATAINDEXPROC glad_glGetFragDataIndex = NULL;
+PFNGLGETFRAGDATALOCATIONPROC glad_glGetFragDataLocation = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv = NULL;
+PFNGLGETINTEGER64I_VPROC glad_glGetInteger64i_v = NULL;
+PFNGLGETINTEGER64VPROC glad_glGetInteger64v = NULL;
+PFNGLGETINTEGERI_VPROC glad_glGetIntegeri_v = NULL;
+PFNGLGETINTEGERVPROC glad_glGetIntegerv = NULL;
+PFNGLGETLIGHTFVPROC glad_glGetLightfv = NULL;
+PFNGLGETLIGHTIVPROC glad_glGetLightiv = NULL;
+PFNGLGETMAPDVPROC glad_glGetMapdv = NULL;
+PFNGLGETMAPFVPROC glad_glGetMapfv = NULL;
+PFNGLGETMAPIVPROC glad_glGetMapiv = NULL;
+PFNGLGETMATERIALFVPROC glad_glGetMaterialfv = NULL;
+PFNGLGETMATERIALIVPROC glad_glGetMaterialiv = NULL;
+PFNGLGETMULTISAMPLEFVPROC glad_glGetMultisamplefv = NULL;
+PFNGLGETPIXELMAPFVPROC glad_glGetPixelMapfv = NULL;
+PFNGLGETPIXELMAPUIVPROC glad_glGetPixelMapuiv = NULL;
+PFNGLGETPIXELMAPUSVPROC glad_glGetPixelMapusv = NULL;
+PFNGLGETPOINTERVPROC glad_glGetPointerv = NULL;
+PFNGLGETPOLYGONSTIPPLEPROC glad_glGetPolygonStipple = NULL;
+PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog = NULL;
+PFNGLGETPROGRAMIVPROC glad_glGetProgramiv = NULL;
+PFNGLGETQUERYOBJECTI64VPROC glad_glGetQueryObjecti64v = NULL;
+PFNGLGETQUERYOBJECTIVPROC glad_glGetQueryObjectiv = NULL;
+PFNGLGETQUERYOBJECTUI64VPROC glad_glGetQueryObjectui64v = NULL;
+PFNGLGETQUERYOBJECTUIVPROC glad_glGetQueryObjectuiv = NULL;
+PFNGLGETQUERYIVPROC glad_glGetQueryiv = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv = NULL;
+PFNGLGETSAMPLERPARAMETERIIVPROC glad_glGetSamplerParameterIiv = NULL;
+PFNGLGETSAMPLERPARAMETERIUIVPROC glad_glGetSamplerParameterIuiv = NULL;
+PFNGLGETSAMPLERPARAMETERFVPROC glad_glGetSamplerParameterfv = NULL;
+PFNGLGETSAMPLERPARAMETERIVPROC glad_glGetSamplerParameteriv = NULL;
+PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog = NULL;
+PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource = NULL;
+PFNGLGETSHADERIVPROC glad_glGetShaderiv = NULL;
+PFNGLGETSTRINGPROC glad_glGetString = NULL;
+PFNGLGETSTRINGIPROC glad_glGetStringi = NULL;
+PFNGLGETSYNCIVPROC glad_glGetSynciv = NULL;
+PFNGLGETTEXENVFVPROC glad_glGetTexEnvfv = NULL;
+PFNGLGETTEXENVIVPROC glad_glGetTexEnviv = NULL;
+PFNGLGETTEXGENDVPROC glad_glGetTexGendv = NULL;
+PFNGLGETTEXGENFVPROC glad_glGetTexGenfv = NULL;
+PFNGLGETTEXGENIVPROC glad_glGetTexGeniv = NULL;
+PFNGLGETTEXIMAGEPROC glad_glGetTexImage = NULL;
+PFNGLGETTEXLEVELPARAMETERFVPROC glad_glGetTexLevelParameterfv = NULL;
+PFNGLGETTEXLEVELPARAMETERIVPROC glad_glGetTexLevelParameteriv = NULL;
+PFNGLGETTEXPARAMETERIIVPROC glad_glGetTexParameterIiv = NULL;
+PFNGLGETTEXPARAMETERIUIVPROC glad_glGetTexParameterIuiv = NULL;
+PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv = NULL;
+PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv = NULL;
+PFNGLGETTRANSFORMFEEDBACKVARYINGPROC glad_glGetTransformFeedbackVarying = NULL;
+PFNGLGETUNIFORMBLOCKINDEXPROC glad_glGetUniformBlockIndex = NULL;
+PFNGLGETUNIFORMINDICESPROC glad_glGetUniformIndices = NULL;
+PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation = NULL;
+PFNGLGETUNIFORMFVPROC glad_glGetUniformfv = NULL;
+PFNGLGETUNIFORMIVPROC glad_glGetUniformiv = NULL;
+PFNGLGETUNIFORMUIVPROC glad_glGetUniformuiv = NULL;
+PFNGLGETVERTEXATTRIBIIVPROC glad_glGetVertexAttribIiv = NULL;
+PFNGLGETVERTEXATTRIBIUIVPROC glad_glGetVertexAttribIuiv = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv = NULL;
+PFNGLGETVERTEXATTRIBDVPROC glad_glGetVertexAttribdv = NULL;
+PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv = NULL;
+PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv = NULL;
+PFNGLHINTPROC glad_glHint = NULL;
+PFNGLINDEXMASKPROC glad_glIndexMask = NULL;
+PFNGLINDEXPOINTERPROC glad_glIndexPointer = NULL;
+PFNGLINDEXDPROC glad_glIndexd = NULL;
+PFNGLINDEXDVPROC glad_glIndexdv = NULL;
+PFNGLINDEXFPROC glad_glIndexf = NULL;
+PFNGLINDEXFVPROC glad_glIndexfv = NULL;
+PFNGLINDEXIPROC glad_glIndexi = NULL;
+PFNGLINDEXIVPROC glad_glIndexiv = NULL;
+PFNGLINDEXSPROC glad_glIndexs = NULL;
+PFNGLINDEXSVPROC glad_glIndexsv = NULL;
+PFNGLINDEXUBPROC glad_glIndexub = NULL;
+PFNGLINDEXUBVPROC glad_glIndexubv = NULL;
+PFNGLINITNAMESPROC glad_glInitNames = NULL;
+PFNGLINTERLEAVEDARRAYSPROC glad_glInterleavedArrays = NULL;
+PFNGLISBUFFERPROC glad_glIsBuffer = NULL;
+PFNGLISENABLEDPROC glad_glIsEnabled = NULL;
+PFNGLISENABLEDIPROC glad_glIsEnabledi = NULL;
+PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer = NULL;
+PFNGLISLISTPROC glad_glIsList = NULL;
+PFNGLISPROGRAMPROC glad_glIsProgram = NULL;
+PFNGLISQUERYPROC glad_glIsQuery = NULL;
+PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer = NULL;
+PFNGLISSAMPLERPROC glad_glIsSampler = NULL;
+PFNGLISSHADERPROC glad_glIsShader = NULL;
+PFNGLISSYNCPROC glad_glIsSync = NULL;
+PFNGLISTEXTUREPROC glad_glIsTexture = NULL;
+PFNGLISVERTEXARRAYPROC glad_glIsVertexArray = NULL;
+PFNGLLIGHTMODELFPROC glad_glLightModelf = NULL;
+PFNGLLIGHTMODELFVPROC glad_glLightModelfv = NULL;
+PFNGLLIGHTMODELIPROC glad_glLightModeli = NULL;
+PFNGLLIGHTMODELIVPROC glad_glLightModeliv = NULL;
+PFNGLLIGHTFPROC glad_glLightf = NULL;
+PFNGLLIGHTFVPROC glad_glLightfv = NULL;
+PFNGLLIGHTIPROC glad_glLighti = NULL;
+PFNGLLIGHTIVPROC glad_glLightiv = NULL;
+PFNGLLINESTIPPLEPROC glad_glLineStipple = NULL;
+PFNGLLINEWIDTHPROC glad_glLineWidth = NULL;
+PFNGLLINKPROGRAMPROC glad_glLinkProgram = NULL;
+PFNGLLISTBASEPROC glad_glListBase = NULL;
+PFNGLLOADIDENTITYPROC glad_glLoadIdentity = NULL;
+PFNGLLOADMATRIXDPROC glad_glLoadMatrixd = NULL;
+PFNGLLOADMATRIXFPROC glad_glLoadMatrixf = NULL;
+PFNGLLOADNAMEPROC glad_glLoadName = NULL;
+PFNGLLOADTRANSPOSEMATRIXDPROC glad_glLoadTransposeMatrixd = NULL;
+PFNGLLOADTRANSPOSEMATRIXFPROC glad_glLoadTransposeMatrixf = NULL;
+PFNGLLOGICOPPROC glad_glLogicOp = NULL;
+PFNGLMAP1DPROC glad_glMap1d = NULL;
+PFNGLMAP1FPROC glad_glMap1f = NULL;
+PFNGLMAP2DPROC glad_glMap2d = NULL;
+PFNGLMAP2FPROC glad_glMap2f = NULL;
+PFNGLMAPBUFFERPROC glad_glMapBuffer = NULL;
+PFNGLMAPBUFFERRANGEPROC glad_glMapBufferRange = NULL;
+PFNGLMAPGRID1DPROC glad_glMapGrid1d = NULL;
+PFNGLMAPGRID1FPROC glad_glMapGrid1f = NULL;
+PFNGLMAPGRID2DPROC glad_glMapGrid2d = NULL;
+PFNGLMAPGRID2FPROC glad_glMapGrid2f = NULL;
+PFNGLMATERIALFPROC glad_glMaterialf = NULL;
+PFNGLMATERIALFVPROC glad_glMaterialfv = NULL;
+PFNGLMATERIALIPROC glad_glMateriali = NULL;
+PFNGLMATERIALIVPROC glad_glMaterialiv = NULL;
+PFNGLMATRIXMODEPROC glad_glMatrixMode = NULL;
+PFNGLMULTMATRIXDPROC glad_glMultMatrixd = NULL;
+PFNGLMULTMATRIXFPROC glad_glMultMatrixf = NULL;
+PFNGLMULTTRANSPOSEMATRIXDPROC glad_glMultTransposeMatrixd = NULL;
+PFNGLMULTTRANSPOSEMATRIXFPROC glad_glMultTransposeMatrixf = NULL;
+PFNGLMULTIDRAWARRAYSPROC glad_glMultiDrawArrays = NULL;
+PFNGLMULTIDRAWELEMENTSPROC glad_glMultiDrawElements = NULL;
+PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC glad_glMultiDrawElementsBaseVertex = NULL;
+PFNGLMULTITEXCOORD1DPROC glad_glMultiTexCoord1d = NULL;
+PFNGLMULTITEXCOORD1DVPROC glad_glMultiTexCoord1dv = NULL;
+PFNGLMULTITEXCOORD1FPROC glad_glMultiTexCoord1f = NULL;
+PFNGLMULTITEXCOORD1FVPROC glad_glMultiTexCoord1fv = NULL;
+PFNGLMULTITEXCOORD1IPROC glad_glMultiTexCoord1i = NULL;
+PFNGLMULTITEXCOORD1IVPROC glad_glMultiTexCoord1iv = NULL;
+PFNGLMULTITEXCOORD1SPROC glad_glMultiTexCoord1s = NULL;
+PFNGLMULTITEXCOORD1SVPROC glad_glMultiTexCoord1sv = NULL;
+PFNGLMULTITEXCOORD2DPROC glad_glMultiTexCoord2d = NULL;
+PFNGLMULTITEXCOORD2DVPROC glad_glMultiTexCoord2dv = NULL;
+PFNGLMULTITEXCOORD2FPROC glad_glMultiTexCoord2f = NULL;
+PFNGLMULTITEXCOORD2FVPROC glad_glMultiTexCoord2fv = NULL;
+PFNGLMULTITEXCOORD2IPROC glad_glMultiTexCoord2i = NULL;
+PFNGLMULTITEXCOORD2IVPROC glad_glMultiTexCoord2iv = NULL;
+PFNGLMULTITEXCOORD2SPROC glad_glMultiTexCoord2s = NULL;
+PFNGLMULTITEXCOORD2SVPROC glad_glMultiTexCoord2sv = NULL;
+PFNGLMULTITEXCOORD3DPROC glad_glMultiTexCoord3d = NULL;
+PFNGLMULTITEXCOORD3DVPROC glad_glMultiTexCoord3dv = NULL;
+PFNGLMULTITEXCOORD3FPROC glad_glMultiTexCoord3f = NULL;
+PFNGLMULTITEXCOORD3FVPROC glad_glMultiTexCoord3fv = NULL;
+PFNGLMULTITEXCOORD3IPROC glad_glMultiTexCoord3i = NULL;
+PFNGLMULTITEXCOORD3IVPROC glad_glMultiTexCoord3iv = NULL;
+PFNGLMULTITEXCOORD3SPROC glad_glMultiTexCoord3s = NULL;
+PFNGLMULTITEXCOORD3SVPROC glad_glMultiTexCoord3sv = NULL;
+PFNGLMULTITEXCOORD4DPROC glad_glMultiTexCoord4d = NULL;
+PFNGLMULTITEXCOORD4DVPROC glad_glMultiTexCoord4dv = NULL;
+PFNGLMULTITEXCOORD4FPROC glad_glMultiTexCoord4f = NULL;
+PFNGLMULTITEXCOORD4FVPROC glad_glMultiTexCoord4fv = NULL;
+PFNGLMULTITEXCOORD4IPROC glad_glMultiTexCoord4i = NULL;
+PFNGLMULTITEXCOORD4IVPROC glad_glMultiTexCoord4iv = NULL;
+PFNGLMULTITEXCOORD4SPROC glad_glMultiTexCoord4s = NULL;
+PFNGLMULTITEXCOORD4SVPROC glad_glMultiTexCoord4sv = NULL;
+PFNGLMULTITEXCOORDP1UIPROC glad_glMultiTexCoordP1ui = NULL;
+PFNGLMULTITEXCOORDP1UIVPROC glad_glMultiTexCoordP1uiv = NULL;
+PFNGLMULTITEXCOORDP2UIPROC glad_glMultiTexCoordP2ui = NULL;
+PFNGLMULTITEXCOORDP2UIVPROC glad_glMultiTexCoordP2uiv = NULL;
+PFNGLMULTITEXCOORDP3UIPROC glad_glMultiTexCoordP3ui = NULL;
+PFNGLMULTITEXCOORDP3UIVPROC glad_glMultiTexCoordP3uiv = NULL;
+PFNGLMULTITEXCOORDP4UIPROC glad_glMultiTexCoordP4ui = NULL;
+PFNGLMULTITEXCOORDP4UIVPROC glad_glMultiTexCoordP4uiv = NULL;
+PFNGLNEWLISTPROC glad_glNewList = NULL;
+PFNGLNORMAL3BPROC glad_glNormal3b = NULL;
+PFNGLNORMAL3BVPROC glad_glNormal3bv = NULL;
+PFNGLNORMAL3DPROC glad_glNormal3d = NULL;
+PFNGLNORMAL3DVPROC glad_glNormal3dv = NULL;
+PFNGLNORMAL3FPROC glad_glNormal3f = NULL;
+PFNGLNORMAL3FVPROC glad_glNormal3fv = NULL;
+PFNGLNORMAL3IPROC glad_glNormal3i = NULL;
+PFNGLNORMAL3IVPROC glad_glNormal3iv = NULL;
+PFNGLNORMAL3SPROC glad_glNormal3s = NULL;
+PFNGLNORMAL3SVPROC glad_glNormal3sv = NULL;
+PFNGLNORMALP3UIPROC glad_glNormalP3ui = NULL;
+PFNGLNORMALP3UIVPROC glad_glNormalP3uiv = NULL;
+PFNGLNORMALPOINTERPROC glad_glNormalPointer = NULL;
+PFNGLORTHOPROC glad_glOrtho = NULL;
+PFNGLPASSTHROUGHPROC glad_glPassThrough = NULL;
+PFNGLPIXELMAPFVPROC glad_glPixelMapfv = NULL;
+PFNGLPIXELMAPUIVPROC glad_glPixelMapuiv = NULL;
+PFNGLPIXELMAPUSVPROC glad_glPixelMapusv = NULL;
+PFNGLPIXELSTOREFPROC glad_glPixelStoref = NULL;
+PFNGLPIXELSTOREIPROC glad_glPixelStorei = NULL;
+PFNGLPIXELTRANSFERFPROC glad_glPixelTransferf = NULL;
+PFNGLPIXELTRANSFERIPROC glad_glPixelTransferi = NULL;
+PFNGLPIXELZOOMPROC glad_glPixelZoom = NULL;
+PFNGLPOINTPARAMETERFPROC glad_glPointParameterf = NULL;
+PFNGLPOINTPARAMETERFVPROC glad_glPointParameterfv = NULL;
+PFNGLPOINTPARAMETERIPROC glad_glPointParameteri = NULL;
+PFNGLPOINTPARAMETERIVPROC glad_glPointParameteriv = NULL;
+PFNGLPOINTSIZEPROC glad_glPointSize = NULL;
+PFNGLPOLYGONMODEPROC glad_glPolygonMode = NULL;
+PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset = NULL;
+PFNGLPOLYGONSTIPPLEPROC glad_glPolygonStipple = NULL;
+PFNGLPOPATTRIBPROC glad_glPopAttrib = NULL;
+PFNGLPOPCLIENTATTRIBPROC glad_glPopClientAttrib = NULL;
+PFNGLPOPMATRIXPROC glad_glPopMatrix = NULL;
+PFNGLPOPNAMEPROC glad_glPopName = NULL;
+PFNGLPRIMITIVERESTARTINDEXPROC glad_glPrimitiveRestartIndex = NULL;
+PFNGLPRIORITIZETEXTURESPROC glad_glPrioritizeTextures = NULL;
+PFNGLPROVOKINGVERTEXPROC glad_glProvokingVertex = NULL;
+PFNGLPUSHATTRIBPROC glad_glPushAttrib = NULL;
+PFNGLPUSHCLIENTATTRIBPROC glad_glPushClientAttrib = NULL;
+PFNGLPUSHMATRIXPROC glad_glPushMatrix = NULL;
+PFNGLPUSHNAMEPROC glad_glPushName = NULL;
+PFNGLQUERYCOUNTERPROC glad_glQueryCounter = NULL;
+PFNGLRASTERPOS2DPROC glad_glRasterPos2d = NULL;
+PFNGLRASTERPOS2DVPROC glad_glRasterPos2dv = NULL;
+PFNGLRASTERPOS2FPROC glad_glRasterPos2f = NULL;
+PFNGLRASTERPOS2FVPROC glad_glRasterPos2fv = NULL;
+PFNGLRASTERPOS2IPROC glad_glRasterPos2i = NULL;
+PFNGLRASTERPOS2IVPROC glad_glRasterPos2iv = NULL;
+PFNGLRASTERPOS2SPROC glad_glRasterPos2s = NULL;
+PFNGLRASTERPOS2SVPROC glad_glRasterPos2sv = NULL;
+PFNGLRASTERPOS3DPROC glad_glRasterPos3d = NULL;
+PFNGLRASTERPOS3DVPROC glad_glRasterPos3dv = NULL;
+PFNGLRASTERPOS3FPROC glad_glRasterPos3f = NULL;
+PFNGLRASTERPOS3FVPROC glad_glRasterPos3fv = NULL;
+PFNGLRASTERPOS3IPROC glad_glRasterPos3i = NULL;
+PFNGLRASTERPOS3IVPROC glad_glRasterPos3iv = NULL;
+PFNGLRASTERPOS3SPROC glad_glRasterPos3s = NULL;
+PFNGLRASTERPOS3SVPROC glad_glRasterPos3sv = NULL;
+PFNGLRASTERPOS4DPROC glad_glRasterPos4d = NULL;
+PFNGLRASTERPOS4DVPROC glad_glRasterPos4dv = NULL;
+PFNGLRASTERPOS4FPROC glad_glRasterPos4f = NULL;
+PFNGLRASTERPOS4FVPROC glad_glRasterPos4fv = NULL;
+PFNGLRASTERPOS4IPROC glad_glRasterPos4i = NULL;
+PFNGLRASTERPOS4IVPROC glad_glRasterPos4iv = NULL;
+PFNGLRASTERPOS4SPROC glad_glRasterPos4s = NULL;
+PFNGLRASTERPOS4SVPROC glad_glRasterPos4sv = NULL;
+PFNGLREADBUFFERPROC glad_glReadBuffer = NULL;
+PFNGLREADPIXELSPROC glad_glReadPixels = NULL;
+PFNGLRECTDPROC glad_glRectd = NULL;
+PFNGLRECTDVPROC glad_glRectdv = NULL;
+PFNGLRECTFPROC glad_glRectf = NULL;
+PFNGLRECTFVPROC glad_glRectfv = NULL;
+PFNGLRECTIPROC glad_glRecti = NULL;
+PFNGLRECTIVPROC glad_glRectiv = NULL;
+PFNGLRECTSPROC glad_glRects = NULL;
+PFNGLRECTSVPROC glad_glRectsv = NULL;
+PFNGLRENDERMODEPROC glad_glRenderMode = NULL;
+PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage = NULL;
+PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC glad_glRenderbufferStorageMultisample = NULL;
+PFNGLROTATEDPROC glad_glRotated = NULL;
+PFNGLROTATEFPROC glad_glRotatef = NULL;
+PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage = NULL;
+PFNGLSAMPLEMASKIPROC glad_glSampleMaski = NULL;
+PFNGLSAMPLERPARAMETERIIVPROC glad_glSamplerParameterIiv = NULL;
+PFNGLSAMPLERPARAMETERIUIVPROC glad_glSamplerParameterIuiv = NULL;
+PFNGLSAMPLERPARAMETERFPROC glad_glSamplerParameterf = NULL;
+PFNGLSAMPLERPARAMETERFVPROC glad_glSamplerParameterfv = NULL;
+PFNGLSAMPLERPARAMETERIPROC glad_glSamplerParameteri = NULL;
+PFNGLSAMPLERPARAMETERIVPROC glad_glSamplerParameteriv = NULL;
+PFNGLSCALEDPROC glad_glScaled = NULL;
+PFNGLSCALEFPROC glad_glScalef = NULL;
+PFNGLSCISSORPROC glad_glScissor = NULL;
+PFNGLSECONDARYCOLOR3BPROC glad_glSecondaryColor3b = NULL;
+PFNGLSECONDARYCOLOR3BVPROC glad_glSecondaryColor3bv = NULL;
+PFNGLSECONDARYCOLOR3DPROC glad_glSecondaryColor3d = NULL;
+PFNGLSECONDARYCOLOR3DVPROC glad_glSecondaryColor3dv = NULL;
+PFNGLSECONDARYCOLOR3FPROC glad_glSecondaryColor3f = NULL;
+PFNGLSECONDARYCOLOR3FVPROC glad_glSecondaryColor3fv = NULL;
+PFNGLSECONDARYCOLOR3IPROC glad_glSecondaryColor3i = NULL;
+PFNGLSECONDARYCOLOR3IVPROC glad_glSecondaryColor3iv = NULL;
+PFNGLSECONDARYCOLOR3SPROC glad_glSecondaryColor3s = NULL;
+PFNGLSECONDARYCOLOR3SVPROC glad_glSecondaryColor3sv = NULL;
+PFNGLSECONDARYCOLOR3UBPROC glad_glSecondaryColor3ub = NULL;
+PFNGLSECONDARYCOLOR3UBVPROC glad_glSecondaryColor3ubv = NULL;
+PFNGLSECONDARYCOLOR3UIPROC glad_glSecondaryColor3ui = NULL;
+PFNGLSECONDARYCOLOR3UIVPROC glad_glSecondaryColor3uiv = NULL;
+PFNGLSECONDARYCOLOR3USPROC glad_glSecondaryColor3us = NULL;
+PFNGLSECONDARYCOLOR3USVPROC glad_glSecondaryColor3usv = NULL;
+PFNGLSECONDARYCOLORP3UIPROC glad_glSecondaryColorP3ui = NULL;
+PFNGLSECONDARYCOLORP3UIVPROC glad_glSecondaryColorP3uiv = NULL;
+PFNGLSECONDARYCOLORPOINTERPROC glad_glSecondaryColorPointer = NULL;
+PFNGLSELECTBUFFERPROC glad_glSelectBuffer = NULL;
+PFNGLSHADEMODELPROC glad_glShadeModel = NULL;
+PFNGLSHADERSOURCEPROC glad_glShaderSource = NULL;
+PFNGLSTENCILFUNCPROC glad_glStencilFunc = NULL;
+PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate = NULL;
+PFNGLSTENCILMASKPROC glad_glStencilMask = NULL;
+PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate = NULL;
+PFNGLSTENCILOPPROC glad_glStencilOp = NULL;
+PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate = NULL;
+PFNGLTEXBUFFERPROC glad_glTexBuffer = NULL;
+PFNGLTEXCOORD1DPROC glad_glTexCoord1d = NULL;
+PFNGLTEXCOORD1DVPROC glad_glTexCoord1dv = NULL;
+PFNGLTEXCOORD1FPROC glad_glTexCoord1f = NULL;
+PFNGLTEXCOORD1FVPROC glad_glTexCoord1fv = NULL;
+PFNGLTEXCOORD1IPROC glad_glTexCoord1i = NULL;
+PFNGLTEXCOORD1IVPROC glad_glTexCoord1iv = NULL;
+PFNGLTEXCOORD1SPROC glad_glTexCoord1s = NULL;
+PFNGLTEXCOORD1SVPROC glad_glTexCoord1sv = NULL;
+PFNGLTEXCOORD2DPROC glad_glTexCoord2d = NULL;
+PFNGLTEXCOORD2DVPROC glad_glTexCoord2dv = NULL;
+PFNGLTEXCOORD2FPROC glad_glTexCoord2f = NULL;
+PFNGLTEXCOORD2FVPROC glad_glTexCoord2fv = NULL;
+PFNGLTEXCOORD2IPROC glad_glTexCoord2i = NULL;
+PFNGLTEXCOORD2IVPROC glad_glTexCoord2iv = NULL;
+PFNGLTEXCOORD2SPROC glad_glTexCoord2s = NULL;
+PFNGLTEXCOORD2SVPROC glad_glTexCoord2sv = NULL;
+PFNGLTEXCOORD3DPROC glad_glTexCoord3d = NULL;
+PFNGLTEXCOORD3DVPROC glad_glTexCoord3dv = NULL;
+PFNGLTEXCOORD3FPROC glad_glTexCoord3f = NULL;
+PFNGLTEXCOORD3FVPROC glad_glTexCoord3fv = NULL;
+PFNGLTEXCOORD3IPROC glad_glTexCoord3i = NULL;
+PFNGLTEXCOORD3IVPROC glad_glTexCoord3iv = NULL;
+PFNGLTEXCOORD3SPROC glad_glTexCoord3s = NULL;
+PFNGLTEXCOORD3SVPROC glad_glTexCoord3sv = NULL;
+PFNGLTEXCOORD4DPROC glad_glTexCoord4d = NULL;
+PFNGLTEXCOORD4DVPROC glad_glTexCoord4dv = NULL;
+PFNGLTEXCOORD4FPROC glad_glTexCoord4f = NULL;
+PFNGLTEXCOORD4FVPROC glad_glTexCoord4fv = NULL;
+PFNGLTEXCOORD4IPROC glad_glTexCoord4i = NULL;
+PFNGLTEXCOORD4IVPROC glad_glTexCoord4iv = NULL;
+PFNGLTEXCOORD4SPROC glad_glTexCoord4s = NULL;
+PFNGLTEXCOORD4SVPROC glad_glTexCoord4sv = NULL;
+PFNGLTEXCOORDP1UIPROC glad_glTexCoordP1ui = NULL;
+PFNGLTEXCOORDP1UIVPROC glad_glTexCoordP1uiv = NULL;
+PFNGLTEXCOORDP2UIPROC glad_glTexCoordP2ui = NULL;
+PFNGLTEXCOORDP2UIVPROC glad_glTexCoordP2uiv = NULL;
+PFNGLTEXCOORDP3UIPROC glad_glTexCoordP3ui = NULL;
+PFNGLTEXCOORDP3UIVPROC glad_glTexCoordP3uiv = NULL;
+PFNGLTEXCOORDP4UIPROC glad_glTexCoordP4ui = NULL;
+PFNGLTEXCOORDP4UIVPROC glad_glTexCoordP4uiv = NULL;
+PFNGLTEXCOORDPOINTERPROC glad_glTexCoordPointer = NULL;
+PFNGLTEXENVFPROC glad_glTexEnvf = NULL;
+PFNGLTEXENVFVPROC glad_glTexEnvfv = NULL;
+PFNGLTEXENVIPROC glad_glTexEnvi = NULL;
+PFNGLTEXENVIVPROC glad_glTexEnviv = NULL;
+PFNGLTEXGENDPROC glad_glTexGend = NULL;
+PFNGLTEXGENDVPROC glad_glTexGendv = NULL;
+PFNGLTEXGENFPROC glad_glTexGenf = NULL;
+PFNGLTEXGENFVPROC glad_glTexGenfv = NULL;
+PFNGLTEXGENIPROC glad_glTexGeni = NULL;
+PFNGLTEXGENIVPROC glad_glTexGeniv = NULL;
+PFNGLTEXIMAGE1DPROC glad_glTexImage1D = NULL;
+PFNGLTEXIMAGE2DPROC glad_glTexImage2D = NULL;
+PFNGLTEXIMAGE2DMULTISAMPLEPROC glad_glTexImage2DMultisample = NULL;
+PFNGLTEXIMAGE3DPROC glad_glTexImage3D = NULL;
+PFNGLTEXIMAGE3DMULTISAMPLEPROC glad_glTexImage3DMultisample = NULL;
+PFNGLTEXPARAMETERIIVPROC glad_glTexParameterIiv = NULL;
+PFNGLTEXPARAMETERIUIVPROC glad_glTexParameterIuiv = NULL;
+PFNGLTEXPARAMETERFPROC glad_glTexParameterf = NULL;
+PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv = NULL;
+PFNGLTEXPARAMETERIPROC glad_glTexParameteri = NULL;
+PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv = NULL;
+PFNGLTEXSUBIMAGE1DPROC glad_glTexSubImage1D = NULL;
+PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D = NULL;
+PFNGLTEXSUBIMAGE3DPROC glad_glTexSubImage3D = NULL;
+PFNGLTRANSFORMFEEDBACKVARYINGSPROC glad_glTransformFeedbackVaryings = NULL;
+PFNGLTRANSLATEDPROC glad_glTranslated = NULL;
+PFNGLTRANSLATEFPROC glad_glTranslatef = NULL;
+PFNGLUNIFORM1FPROC glad_glUniform1f = NULL;
+PFNGLUNIFORM1FVPROC glad_glUniform1fv = NULL;
+PFNGLUNIFORM1IPROC glad_glUniform1i = NULL;
+PFNGLUNIFORM1IVPROC glad_glUniform1iv = NULL;
+PFNGLUNIFORM1UIPROC glad_glUniform1ui = NULL;
+PFNGLUNIFORM1UIVPROC glad_glUniform1uiv = NULL;
+PFNGLUNIFORM2FPROC glad_glUniform2f = NULL;
+PFNGLUNIFORM2FVPROC glad_glUniform2fv = NULL;
+PFNGLUNIFORM2IPROC glad_glUniform2i = NULL;
+PFNGLUNIFORM2IVPROC glad_glUniform2iv = NULL;
+PFNGLUNIFORM2UIPROC glad_glUniform2ui = NULL;
+PFNGLUNIFORM2UIVPROC glad_glUniform2uiv = NULL;
+PFNGLUNIFORM3FPROC glad_glUniform3f = NULL;
+PFNGLUNIFORM3FVPROC glad_glUniform3fv = NULL;
+PFNGLUNIFORM3IPROC glad_glUniform3i = NULL;
+PFNGLUNIFORM3IVPROC glad_glUniform3iv = NULL;
+PFNGLUNIFORM3UIPROC glad_glUniform3ui = NULL;
+PFNGLUNIFORM3UIVPROC glad_glUniform3uiv = NULL;
+PFNGLUNIFORM4FPROC glad_glUniform4f = NULL;
+PFNGLUNIFORM4FVPROC glad_glUniform4fv = NULL;
+PFNGLUNIFORM4IPROC glad_glUniform4i = NULL;
+PFNGLUNIFORM4IVPROC glad_glUniform4iv = NULL;
+PFNGLUNIFORM4UIPROC glad_glUniform4ui = NULL;
+PFNGLUNIFORM4UIVPROC glad_glUniform4uiv = NULL;
+PFNGLUNIFORMBLOCKBINDINGPROC glad_glUniformBlockBinding = NULL;
+PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv = NULL;
+PFNGLUNIFORMMATRIX2X3FVPROC glad_glUniformMatrix2x3fv = NULL;
+PFNGLUNIFORMMATRIX2X4FVPROC glad_glUniformMatrix2x4fv = NULL;
+PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv = NULL;
+PFNGLUNIFORMMATRIX3X2FVPROC glad_glUniformMatrix3x2fv = NULL;
+PFNGLUNIFORMMATRIX3X4FVPROC glad_glUniformMatrix3x4fv = NULL;
+PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv = NULL;
+PFNGLUNIFORMMATRIX4X2FVPROC glad_glUniformMatrix4x2fv = NULL;
+PFNGLUNIFORMMATRIX4X3FVPROC glad_glUniformMatrix4x3fv = NULL;
+PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer = NULL;
+PFNGLUSEPROGRAMPROC glad_glUseProgram = NULL;
+PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram = NULL;
+PFNGLVERTEX2DPROC glad_glVertex2d = NULL;
+PFNGLVERTEX2DVPROC glad_glVertex2dv = NULL;
+PFNGLVERTEX2FPROC glad_glVertex2f = NULL;
+PFNGLVERTEX2FVPROC glad_glVertex2fv = NULL;
+PFNGLVERTEX2IPROC glad_glVertex2i = NULL;
+PFNGLVERTEX2IVPROC glad_glVertex2iv = NULL;
+PFNGLVERTEX2SPROC glad_glVertex2s = NULL;
+PFNGLVERTEX2SVPROC glad_glVertex2sv = NULL;
+PFNGLVERTEX3DPROC glad_glVertex3d = NULL;
+PFNGLVERTEX3DVPROC glad_glVertex3dv = NULL;
+PFNGLVERTEX3FPROC glad_glVertex3f = NULL;
+PFNGLVERTEX3FVPROC glad_glVertex3fv = NULL;
+PFNGLVERTEX3IPROC glad_glVertex3i = NULL;
+PFNGLVERTEX3IVPROC glad_glVertex3iv = NULL;
+PFNGLVERTEX3SPROC glad_glVertex3s = NULL;
+PFNGLVERTEX3SVPROC glad_glVertex3sv = NULL;
+PFNGLVERTEX4DPROC glad_glVertex4d = NULL;
+PFNGLVERTEX4DVPROC glad_glVertex4dv = NULL;
+PFNGLVERTEX4FPROC glad_glVertex4f = NULL;
+PFNGLVERTEX4FVPROC glad_glVertex4fv = NULL;
+PFNGLVERTEX4IPROC glad_glVertex4i = NULL;
+PFNGLVERTEX4IVPROC glad_glVertex4iv = NULL;
+PFNGLVERTEX4SPROC glad_glVertex4s = NULL;
+PFNGLVERTEX4SVPROC glad_glVertex4sv = NULL;
+PFNGLVERTEXATTRIB1DPROC glad_glVertexAttrib1d = NULL;
+PFNGLVERTEXATTRIB1DVPROC glad_glVertexAttrib1dv = NULL;
+PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f = NULL;
+PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv = NULL;
+PFNGLVERTEXATTRIB1SPROC glad_glVertexAttrib1s = NULL;
+PFNGLVERTEXATTRIB1SVPROC glad_glVertexAttrib1sv = NULL;
+PFNGLVERTEXATTRIB2DPROC glad_glVertexAttrib2d = NULL;
+PFNGLVERTEXATTRIB2DVPROC glad_glVertexAttrib2dv = NULL;
+PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f = NULL;
+PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv = NULL;
+PFNGLVERTEXATTRIB2SPROC glad_glVertexAttrib2s = NULL;
+PFNGLVERTEXATTRIB2SVPROC glad_glVertexAttrib2sv = NULL;
+PFNGLVERTEXATTRIB3DPROC glad_glVertexAttrib3d = NULL;
+PFNGLVERTEXATTRIB3DVPROC glad_glVertexAttrib3dv = NULL;
+PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f = NULL;
+PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv = NULL;
+PFNGLVERTEXATTRIB3SPROC glad_glVertexAttrib3s = NULL;
+PFNGLVERTEXATTRIB3SVPROC glad_glVertexAttrib3sv = NULL;
+PFNGLVERTEXATTRIB4NBVPROC glad_glVertexAttrib4Nbv = NULL;
+PFNGLVERTEXATTRIB4NIVPROC glad_glVertexAttrib4Niv = NULL;
+PFNGLVERTEXATTRIB4NSVPROC glad_glVertexAttrib4Nsv = NULL;
+PFNGLVERTEXATTRIB4NUBPROC glad_glVertexAttrib4Nub = NULL;
+PFNGLVERTEXATTRIB4NUBVPROC glad_glVertexAttrib4Nubv = NULL;
+PFNGLVERTEXATTRIB4NUIVPROC glad_glVertexAttrib4Nuiv = NULL;
+PFNGLVERTEXATTRIB4NUSVPROC glad_glVertexAttrib4Nusv = NULL;
+PFNGLVERTEXATTRIB4BVPROC glad_glVertexAttrib4bv = NULL;
+PFNGLVERTEXATTRIB4DPROC glad_glVertexAttrib4d = NULL;
+PFNGLVERTEXATTRIB4DVPROC glad_glVertexAttrib4dv = NULL;
+PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f = NULL;
+PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv = NULL;
+PFNGLVERTEXATTRIB4IVPROC glad_glVertexAttrib4iv = NULL;
+PFNGLVERTEXATTRIB4SPROC glad_glVertexAttrib4s = NULL;
+PFNGLVERTEXATTRIB4SVPROC glad_glVertexAttrib4sv = NULL;
+PFNGLVERTEXATTRIB4UBVPROC glad_glVertexAttrib4ubv = NULL;
+PFNGLVERTEXATTRIB4UIVPROC glad_glVertexAttrib4uiv = NULL;
+PFNGLVERTEXATTRIB4USVPROC glad_glVertexAttrib4usv = NULL;
+PFNGLVERTEXATTRIBDIVISORPROC glad_glVertexAttribDivisor = NULL;
+PFNGLVERTEXATTRIBI1IPROC glad_glVertexAttribI1i = NULL;
+PFNGLVERTEXATTRIBI1IVPROC glad_glVertexAttribI1iv = NULL;
+PFNGLVERTEXATTRIBI1UIPROC glad_glVertexAttribI1ui = NULL;
+PFNGLVERTEXATTRIBI1UIVPROC glad_glVertexAttribI1uiv = NULL;
+PFNGLVERTEXATTRIBI2IPROC glad_glVertexAttribI2i = NULL;
+PFNGLVERTEXATTRIBI2IVPROC glad_glVertexAttribI2iv = NULL;
+PFNGLVERTEXATTRIBI2UIPROC glad_glVertexAttribI2ui = NULL;
+PFNGLVERTEXATTRIBI2UIVPROC glad_glVertexAttribI2uiv = NULL;
+PFNGLVERTEXATTRIBI3IPROC glad_glVertexAttribI3i = NULL;
+PFNGLVERTEXATTRIBI3IVPROC glad_glVertexAttribI3iv = NULL;
+PFNGLVERTEXATTRIBI3UIPROC glad_glVertexAttribI3ui = NULL;
+PFNGLVERTEXATTRIBI3UIVPROC glad_glVertexAttribI3uiv = NULL;
+PFNGLVERTEXATTRIBI4BVPROC glad_glVertexAttribI4bv = NULL;
+PFNGLVERTEXATTRIBI4IPROC glad_glVertexAttribI4i = NULL;
+PFNGLVERTEXATTRIBI4IVPROC glad_glVertexAttribI4iv = NULL;
+PFNGLVERTEXATTRIBI4SVPROC glad_glVertexAttribI4sv = NULL;
+PFNGLVERTEXATTRIBI4UBVPROC glad_glVertexAttribI4ubv = NULL;
+PFNGLVERTEXATTRIBI4UIPROC glad_glVertexAttribI4ui = NULL;
+PFNGLVERTEXATTRIBI4UIVPROC glad_glVertexAttribI4uiv = NULL;
+PFNGLVERTEXATTRIBI4USVPROC glad_glVertexAttribI4usv = NULL;
+PFNGLVERTEXATTRIBIPOINTERPROC glad_glVertexAttribIPointer = NULL;
+PFNGLVERTEXATTRIBP1UIPROC glad_glVertexAttribP1ui = NULL;
+PFNGLVERTEXATTRIBP1UIVPROC glad_glVertexAttribP1uiv = NULL;
+PFNGLVERTEXATTRIBP2UIPROC glad_glVertexAttribP2ui = NULL;
+PFNGLVERTEXATTRIBP2UIVPROC glad_glVertexAttribP2uiv = NULL;
+PFNGLVERTEXATTRIBP3UIPROC glad_glVertexAttribP3ui = NULL;
+PFNGLVERTEXATTRIBP3UIVPROC glad_glVertexAttribP3uiv = NULL;
+PFNGLVERTEXATTRIBP4UIPROC glad_glVertexAttribP4ui = NULL;
+PFNGLVERTEXATTRIBP4UIVPROC glad_glVertexAttribP4uiv = NULL;
+PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer = NULL;
+PFNGLVERTEXP2UIPROC glad_glVertexP2ui = NULL;
+PFNGLVERTEXP2UIVPROC glad_glVertexP2uiv = NULL;
+PFNGLVERTEXP3UIPROC glad_glVertexP3ui = NULL;
+PFNGLVERTEXP3UIVPROC glad_glVertexP3uiv = NULL;
+PFNGLVERTEXP4UIPROC glad_glVertexP4ui = NULL;
+PFNGLVERTEXP4UIVPROC glad_glVertexP4uiv = NULL;
+PFNGLVERTEXPOINTERPROC glad_glVertexPointer = NULL;
+PFNGLVIEWPORTPROC glad_glViewport = NULL;
+PFNGLWAITSYNCPROC glad_glWaitSync = NULL;
+PFNGLWINDOWPOS2DPROC glad_glWindowPos2d = NULL;
+PFNGLWINDOWPOS2DVPROC glad_glWindowPos2dv = NULL;
+PFNGLWINDOWPOS2FPROC glad_glWindowPos2f = NULL;
+PFNGLWINDOWPOS2FVPROC glad_glWindowPos2fv = NULL;
+PFNGLWINDOWPOS2IPROC glad_glWindowPos2i = NULL;
+PFNGLWINDOWPOS2IVPROC glad_glWindowPos2iv = NULL;
+PFNGLWINDOWPOS2SPROC glad_glWindowPos2s = NULL;
+PFNGLWINDOWPOS2SVPROC glad_glWindowPos2sv = NULL;
+PFNGLWINDOWPOS3DPROC glad_glWindowPos3d = NULL;
+PFNGLWINDOWPOS3DVPROC glad_glWindowPos3dv = NULL;
+PFNGLWINDOWPOS3FPROC glad_glWindowPos3f = NULL;
+PFNGLWINDOWPOS3FVPROC glad_glWindowPos3fv = NULL;
+PFNGLWINDOWPOS3IPROC glad_glWindowPos3i = NULL;
+PFNGLWINDOWPOS3IVPROC glad_glWindowPos3iv = NULL;
+PFNGLWINDOWPOS3SPROC glad_glWindowPos3s = NULL;
+PFNGLWINDOWPOS3SVPROC glad_glWindowPos3sv = NULL;
+
+
+static void glad_gl_load_GL_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_0) return;
+    glad_glAccum = (PFNGLACCUMPROC) load(userptr, "glAccum");
+    glad_glAlphaFunc = (PFNGLALPHAFUNCPROC) load(userptr, "glAlphaFunc");
+    glad_glBegin = (PFNGLBEGINPROC) load(userptr, "glBegin");
+    glad_glBitmap = (PFNGLBITMAPPROC) load(userptr, "glBitmap");
+    glad_glBlendFunc = (PFNGLBLENDFUNCPROC) load(userptr, "glBlendFunc");
+    glad_glCallList = (PFNGLCALLLISTPROC) load(userptr, "glCallList");
+    glad_glCallLists = (PFNGLCALLLISTSPROC) load(userptr, "glCallLists");
+    glad_glClear = (PFNGLCLEARPROC) load(userptr, "glClear");
+    glad_glClearAccum = (PFNGLCLEARACCUMPROC) load(userptr, "glClearAccum");
+    glad_glClearColor = (PFNGLCLEARCOLORPROC) load(userptr, "glClearColor");
+    glad_glClearDepth = (PFNGLCLEARDEPTHPROC) load(userptr, "glClearDepth");
+    glad_glClearIndex = (PFNGLCLEARINDEXPROC) load(userptr, "glClearIndex");
+    glad_glClearStencil = (PFNGLCLEARSTENCILPROC) load(userptr, "glClearStencil");
+    glad_glClipPlane = (PFNGLCLIPPLANEPROC) load(userptr, "glClipPlane");
+    glad_glColor3b = (PFNGLCOLOR3BPROC) load(userptr, "glColor3b");
+    glad_glColor3bv = (PFNGLCOLOR3BVPROC) load(userptr, "glColor3bv");
+    glad_glColor3d = (PFNGLCOLOR3DPROC) load(userptr, "glColor3d");
+    glad_glColor3dv = (PFNGLCOLOR3DVPROC) load(userptr, "glColor3dv");
+    glad_glColor3f = (PFNGLCOLOR3FPROC) load(userptr, "glColor3f");
+    glad_glColor3fv = (PFNGLCOLOR3FVPROC) load(userptr, "glColor3fv");
+    glad_glColor3i = (PFNGLCOLOR3IPROC) load(userptr, "glColor3i");
+    glad_glColor3iv = (PFNGLCOLOR3IVPROC) load(userptr, "glColor3iv");
+    glad_glColor3s = (PFNGLCOLOR3SPROC) load(userptr, "glColor3s");
+    glad_glColor3sv = (PFNGLCOLOR3SVPROC) load(userptr, "glColor3sv");
+    glad_glColor3ub = (PFNGLCOLOR3UBPROC) load(userptr, "glColor3ub");
+    glad_glColor3ubv = (PFNGLCOLOR3UBVPROC) load(userptr, "glColor3ubv");
+    glad_glColor3ui = (PFNGLCOLOR3UIPROC) load(userptr, "glColor3ui");
+    glad_glColor3uiv = (PFNGLCOLOR3UIVPROC) load(userptr, "glColor3uiv");
+    glad_glColor3us = (PFNGLCOLOR3USPROC) load(userptr, "glColor3us");
+    glad_glColor3usv = (PFNGLCOLOR3USVPROC) load(userptr, "glColor3usv");
+    glad_glColor4b = (PFNGLCOLOR4BPROC) load(userptr, "glColor4b");
+    glad_glColor4bv = (PFNGLCOLOR4BVPROC) load(userptr, "glColor4bv");
+    glad_glColor4d = (PFNGLCOLOR4DPROC) load(userptr, "glColor4d");
+    glad_glColor4dv = (PFNGLCOLOR4DVPROC) load(userptr, "glColor4dv");
+    glad_glColor4f = (PFNGLCOLOR4FPROC) load(userptr, "glColor4f");
+    glad_glColor4fv = (PFNGLCOLOR4FVPROC) load(userptr, "glColor4fv");
+    glad_glColor4i = (PFNGLCOLOR4IPROC) load(userptr, "glColor4i");
+    glad_glColor4iv = (PFNGLCOLOR4IVPROC) load(userptr, "glColor4iv");
+    glad_glColor4s = (PFNGLCOLOR4SPROC) load(userptr, "glColor4s");
+    glad_glColor4sv = (PFNGLCOLOR4SVPROC) load(userptr, "glColor4sv");
+    glad_glColor4ub = (PFNGLCOLOR4UBPROC) load(userptr, "glColor4ub");
+    glad_glColor4ubv = (PFNGLCOLOR4UBVPROC) load(userptr, "glColor4ubv");
+    glad_glColor4ui = (PFNGLCOLOR4UIPROC) load(userptr, "glColor4ui");
+    glad_glColor4uiv = (PFNGLCOLOR4UIVPROC) load(userptr, "glColor4uiv");
+    glad_glColor4us = (PFNGLCOLOR4USPROC) load(userptr, "glColor4us");
+    glad_glColor4usv = (PFNGLCOLOR4USVPROC) load(userptr, "glColor4usv");
+    glad_glColorMask = (PFNGLCOLORMASKPROC) load(userptr, "glColorMask");
+    glad_glColorMaterial = (PFNGLCOLORMATERIALPROC) load(userptr, "glColorMaterial");
+    glad_glCopyPixels = (PFNGLCOPYPIXELSPROC) load(userptr, "glCopyPixels");
+    glad_glCullFace = (PFNGLCULLFACEPROC) load(userptr, "glCullFace");
+    glad_glDeleteLists = (PFNGLDELETELISTSPROC) load(userptr, "glDeleteLists");
+    glad_glDepthFunc = (PFNGLDEPTHFUNCPROC) load(userptr, "glDepthFunc");
+    glad_glDepthMask = (PFNGLDEPTHMASKPROC) load(userptr, "glDepthMask");
+    glad_glDepthRange = (PFNGLDEPTHRANGEPROC) load(userptr, "glDepthRange");
+    glad_glDisable = (PFNGLDISABLEPROC) load(userptr, "glDisable");
+    glad_glDrawBuffer = (PFNGLDRAWBUFFERPROC) load(userptr, "glDrawBuffer");
+    glad_glDrawPixels = (PFNGLDRAWPIXELSPROC) load(userptr, "glDrawPixels");
+    glad_glEdgeFlag = (PFNGLEDGEFLAGPROC) load(userptr, "glEdgeFlag");
+    glad_glEdgeFlagv = (PFNGLEDGEFLAGVPROC) load(userptr, "glEdgeFlagv");
+    glad_glEnable = (PFNGLENABLEPROC) load(userptr, "glEnable");
+    glad_glEnd = (PFNGLENDPROC) load(userptr, "glEnd");
+    glad_glEndList = (PFNGLENDLISTPROC) load(userptr, "glEndList");
+    glad_glEvalCoord1d = (PFNGLEVALCOORD1DPROC) load(userptr, "glEvalCoord1d");
+    glad_glEvalCoord1dv = (PFNGLEVALCOORD1DVPROC) load(userptr, "glEvalCoord1dv");
+    glad_glEvalCoord1f = (PFNGLEVALCOORD1FPROC) load(userptr, "glEvalCoord1f");
+    glad_glEvalCoord1fv = (PFNGLEVALCOORD1FVPROC) load(userptr, "glEvalCoord1fv");
+    glad_glEvalCoord2d = (PFNGLEVALCOORD2DPROC) load(userptr, "glEvalCoord2d");
+    glad_glEvalCoord2dv = (PFNGLEVALCOORD2DVPROC) load(userptr, "glEvalCoord2dv");
+    glad_glEvalCoord2f = (PFNGLEVALCOORD2FPROC) load(userptr, "glEvalCoord2f");
+    glad_glEvalCoord2fv = (PFNGLEVALCOORD2FVPROC) load(userptr, "glEvalCoord2fv");
+    glad_glEvalMesh1 = (PFNGLEVALMESH1PROC) load(userptr, "glEvalMesh1");
+    glad_glEvalMesh2 = (PFNGLEVALMESH2PROC) load(userptr, "glEvalMesh2");
+    glad_glEvalPoint1 = (PFNGLEVALPOINT1PROC) load(userptr, "glEvalPoint1");
+    glad_glEvalPoint2 = (PFNGLEVALPOINT2PROC) load(userptr, "glEvalPoint2");
+    glad_glFeedbackBuffer = (PFNGLFEEDBACKBUFFERPROC) load(userptr, "glFeedbackBuffer");
+    glad_glFinish = (PFNGLFINISHPROC) load(userptr, "glFinish");
+    glad_glFlush = (PFNGLFLUSHPROC) load(userptr, "glFlush");
+    glad_glFogf = (PFNGLFOGFPROC) load(userptr, "glFogf");
+    glad_glFogfv = (PFNGLFOGFVPROC) load(userptr, "glFogfv");
+    glad_glFogi = (PFNGLFOGIPROC) load(userptr, "glFogi");
+    glad_glFogiv = (PFNGLFOGIVPROC) load(userptr, "glFogiv");
+    glad_glFrontFace = (PFNGLFRONTFACEPROC) load(userptr, "glFrontFace");
+    glad_glFrustum = (PFNGLFRUSTUMPROC) load(userptr, "glFrustum");
+    glad_glGenLists = (PFNGLGENLISTSPROC) load(userptr, "glGenLists");
+    glad_glGetBooleanv = (PFNGLGETBOOLEANVPROC) load(userptr, "glGetBooleanv");
+    glad_glGetClipPlane = (PFNGLGETCLIPPLANEPROC) load(userptr, "glGetClipPlane");
+    glad_glGetDoublev = (PFNGLGETDOUBLEVPROC) load(userptr, "glGetDoublev");
+    glad_glGetError = (PFNGLGETERRORPROC) load(userptr, "glGetError");
+    glad_glGetFloatv = (PFNGLGETFLOATVPROC) load(userptr, "glGetFloatv");
+    glad_glGetIntegerv = (PFNGLGETINTEGERVPROC) load(userptr, "glGetIntegerv");
+    glad_glGetLightfv = (PFNGLGETLIGHTFVPROC) load(userptr, "glGetLightfv");
+    glad_glGetLightiv = (PFNGLGETLIGHTIVPROC) load(userptr, "glGetLightiv");
+    glad_glGetMapdv = (PFNGLGETMAPDVPROC) load(userptr, "glGetMapdv");
+    glad_glGetMapfv = (PFNGLGETMAPFVPROC) load(userptr, "glGetMapfv");
+    glad_glGetMapiv = (PFNGLGETMAPIVPROC) load(userptr, "glGetMapiv");
+    glad_glGetMaterialfv = (PFNGLGETMATERIALFVPROC) load(userptr, "glGetMaterialfv");
+    glad_glGetMaterialiv = (PFNGLGETMATERIALIVPROC) load(userptr, "glGetMaterialiv");
+    glad_glGetPixelMapfv = (PFNGLGETPIXELMAPFVPROC) load(userptr, "glGetPixelMapfv");
+    glad_glGetPixelMapuiv = (PFNGLGETPIXELMAPUIVPROC) load(userptr, "glGetPixelMapuiv");
+    glad_glGetPixelMapusv = (PFNGLGETPIXELMAPUSVPROC) load(userptr, "glGetPixelMapusv");
+    glad_glGetPolygonStipple = (PFNGLGETPOLYGONSTIPPLEPROC) load(userptr, "glGetPolygonStipple");
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    glad_glGetTexEnvfv = (PFNGLGETTEXENVFVPROC) load(userptr, "glGetTexEnvfv");
+    glad_glGetTexEnviv = (PFNGLGETTEXENVIVPROC) load(userptr, "glGetTexEnviv");
+    glad_glGetTexGendv = (PFNGLGETTEXGENDVPROC) load(userptr, "glGetTexGendv");
+    glad_glGetTexGenfv = (PFNGLGETTEXGENFVPROC) load(userptr, "glGetTexGenfv");
+    glad_glGetTexGeniv = (PFNGLGETTEXGENIVPROC) load(userptr, "glGetTexGeniv");
+    glad_glGetTexImage = (PFNGLGETTEXIMAGEPROC) load(userptr, "glGetTexImage");
+    glad_glGetTexLevelParameterfv = (PFNGLGETTEXLEVELPARAMETERFVPROC) load(userptr, "glGetTexLevelParameterfv");
+    glad_glGetTexLevelParameteriv = (PFNGLGETTEXLEVELPARAMETERIVPROC) load(userptr, "glGetTexLevelParameteriv");
+    glad_glGetTexParameterfv = (PFNGLGETTEXPARAMETERFVPROC) load(userptr, "glGetTexParameterfv");
+    glad_glGetTexParameteriv = (PFNGLGETTEXPARAMETERIVPROC) load(userptr, "glGetTexParameteriv");
+    glad_glHint = (PFNGLHINTPROC) load(userptr, "glHint");
+    glad_glIndexMask = (PFNGLINDEXMASKPROC) load(userptr, "glIndexMask");
+    glad_glIndexd = (PFNGLINDEXDPROC) load(userptr, "glIndexd");
+    glad_glIndexdv = (PFNGLINDEXDVPROC) load(userptr, "glIndexdv");
+    glad_glIndexf = (PFNGLINDEXFPROC) load(userptr, "glIndexf");
+    glad_glIndexfv = (PFNGLINDEXFVPROC) load(userptr, "glIndexfv");
+    glad_glIndexi = (PFNGLINDEXIPROC) load(userptr, "glIndexi");
+    glad_glIndexiv = (PFNGLINDEXIVPROC) load(userptr, "glIndexiv");
+    glad_glIndexs = (PFNGLINDEXSPROC) load(userptr, "glIndexs");
+    glad_glIndexsv = (PFNGLINDEXSVPROC) load(userptr, "glIndexsv");
+    glad_glInitNames = (PFNGLINITNAMESPROC) load(userptr, "glInitNames");
+    glad_glIsEnabled = (PFNGLISENABLEDPROC) load(userptr, "glIsEnabled");
+    glad_glIsList = (PFNGLISLISTPROC) load(userptr, "glIsList");
+    glad_glLightModelf = (PFNGLLIGHTMODELFPROC) load(userptr, "glLightModelf");
+    glad_glLightModelfv = (PFNGLLIGHTMODELFVPROC) load(userptr, "glLightModelfv");
+    glad_glLightModeli = (PFNGLLIGHTMODELIPROC) load(userptr, "glLightModeli");
+    glad_glLightModeliv = (PFNGLLIGHTMODELIVPROC) load(userptr, "glLightModeliv");
+    glad_glLightf = (PFNGLLIGHTFPROC) load(userptr, "glLightf");
+    glad_glLightfv = (PFNGLLIGHTFVPROC) load(userptr, "glLightfv");
+    glad_glLighti = (PFNGLLIGHTIPROC) load(userptr, "glLighti");
+    glad_glLightiv = (PFNGLLIGHTIVPROC) load(userptr, "glLightiv");
+    glad_glLineStipple = (PFNGLLINESTIPPLEPROC) load(userptr, "glLineStipple");
+    glad_glLineWidth = (PFNGLLINEWIDTHPROC) load(userptr, "glLineWidth");
+    glad_glListBase = (PFNGLLISTBASEPROC) load(userptr, "glListBase");
+    glad_glLoadIdentity = (PFNGLLOADIDENTITYPROC) load(userptr, "glLoadIdentity");
+    glad_glLoadMatrixd = (PFNGLLOADMATRIXDPROC) load(userptr, "glLoadMatrixd");
+    glad_glLoadMatrixf = (PFNGLLOADMATRIXFPROC) load(userptr, "glLoadMatrixf");
+    glad_glLoadName = (PFNGLLOADNAMEPROC) load(userptr, "glLoadName");
+    glad_glLogicOp = (PFNGLLOGICOPPROC) load(userptr, "glLogicOp");
+    glad_glMap1d = (PFNGLMAP1DPROC) load(userptr, "glMap1d");
+    glad_glMap1f = (PFNGLMAP1FPROC) load(userptr, "glMap1f");
+    glad_glMap2d = (PFNGLMAP2DPROC) load(userptr, "glMap2d");
+    glad_glMap2f = (PFNGLMAP2FPROC) load(userptr, "glMap2f");
+    glad_glMapGrid1d = (PFNGLMAPGRID1DPROC) load(userptr, "glMapGrid1d");
+    glad_glMapGrid1f = (PFNGLMAPGRID1FPROC) load(userptr, "glMapGrid1f");
+    glad_glMapGrid2d = (PFNGLMAPGRID2DPROC) load(userptr, "glMapGrid2d");
+    glad_glMapGrid2f = (PFNGLMAPGRID2FPROC) load(userptr, "glMapGrid2f");
+    glad_glMaterialf = (PFNGLMATERIALFPROC) load(userptr, "glMaterialf");
+    glad_glMaterialfv = (PFNGLMATERIALFVPROC) load(userptr, "glMaterialfv");
+    glad_glMateriali = (PFNGLMATERIALIPROC) load(userptr, "glMateriali");
+    glad_glMaterialiv = (PFNGLMATERIALIVPROC) load(userptr, "glMaterialiv");
+    glad_glMatrixMode = (PFNGLMATRIXMODEPROC) load(userptr, "glMatrixMode");
+    glad_glMultMatrixd = (PFNGLMULTMATRIXDPROC) load(userptr, "glMultMatrixd");
+    glad_glMultMatrixf = (PFNGLMULTMATRIXFPROC) load(userptr, "glMultMatrixf");
+    glad_glNewList = (PFNGLNEWLISTPROC) load(userptr, "glNewList");
+    glad_glNormal3b = (PFNGLNORMAL3BPROC) load(userptr, "glNormal3b");
+    glad_glNormal3bv = (PFNGLNORMAL3BVPROC) load(userptr, "glNormal3bv");
+    glad_glNormal3d = (PFNGLNORMAL3DPROC) load(userptr, "glNormal3d");
+    glad_glNormal3dv = (PFNGLNORMAL3DVPROC) load(userptr, "glNormal3dv");
+    glad_glNormal3f = (PFNGLNORMAL3FPROC) load(userptr, "glNormal3f");
+    glad_glNormal3fv = (PFNGLNORMAL3FVPROC) load(userptr, "glNormal3fv");
+    glad_glNormal3i = (PFNGLNORMAL3IPROC) load(userptr, "glNormal3i");
+    glad_glNormal3iv = (PFNGLNORMAL3IVPROC) load(userptr, "glNormal3iv");
+    glad_glNormal3s = (PFNGLNORMAL3SPROC) load(userptr, "glNormal3s");
+    glad_glNormal3sv = (PFNGLNORMAL3SVPROC) load(userptr, "glNormal3sv");
+    glad_glOrtho = (PFNGLORTHOPROC) load(userptr, "glOrtho");
+    glad_glPassThrough = (PFNGLPASSTHROUGHPROC) load(userptr, "glPassThrough");
+    glad_glPixelMapfv = (PFNGLPIXELMAPFVPROC) load(userptr, "glPixelMapfv");
+    glad_glPixelMapuiv = (PFNGLPIXELMAPUIVPROC) load(userptr, "glPixelMapuiv");
+    glad_glPixelMapusv = (PFNGLPIXELMAPUSVPROC) load(userptr, "glPixelMapusv");
+    glad_glPixelStoref = (PFNGLPIXELSTOREFPROC) load(userptr, "glPixelStoref");
+    glad_glPixelStorei = (PFNGLPIXELSTOREIPROC) load(userptr, "glPixelStorei");
+    glad_glPixelTransferf = (PFNGLPIXELTRANSFERFPROC) load(userptr, "glPixelTransferf");
+    glad_glPixelTransferi = (PFNGLPIXELTRANSFERIPROC) load(userptr, "glPixelTransferi");
+    glad_glPixelZoom = (PFNGLPIXELZOOMPROC) load(userptr, "glPixelZoom");
+    glad_glPointSize = (PFNGLPOINTSIZEPROC) load(userptr, "glPointSize");
+    glad_glPolygonMode = (PFNGLPOLYGONMODEPROC) load(userptr, "glPolygonMode");
+    glad_glPolygonStipple = (PFNGLPOLYGONSTIPPLEPROC) load(userptr, "glPolygonStipple");
+    glad_glPopAttrib = (PFNGLPOPATTRIBPROC) load(userptr, "glPopAttrib");
+    glad_glPopMatrix = (PFNGLPOPMATRIXPROC) load(userptr, "glPopMatrix");
+    glad_glPopName = (PFNGLPOPNAMEPROC) load(userptr, "glPopName");
+    glad_glPushAttrib = (PFNGLPUSHATTRIBPROC) load(userptr, "glPushAttrib");
+    glad_glPushMatrix = (PFNGLPUSHMATRIXPROC) load(userptr, "glPushMatrix");
+    glad_glPushName = (PFNGLPUSHNAMEPROC) load(userptr, "glPushName");
+    glad_glRasterPos2d = (PFNGLRASTERPOS2DPROC) load(userptr, "glRasterPos2d");
+    glad_glRasterPos2dv = (PFNGLRASTERPOS2DVPROC) load(userptr, "glRasterPos2dv");
+    glad_glRasterPos2f = (PFNGLRASTERPOS2FPROC) load(userptr, "glRasterPos2f");
+    glad_glRasterPos2fv = (PFNGLRASTERPOS2FVPROC) load(userptr, "glRasterPos2fv");
+    glad_glRasterPos2i = (PFNGLRASTERPOS2IPROC) load(userptr, "glRasterPos2i");
+    glad_glRasterPos2iv = (PFNGLRASTERPOS2IVPROC) load(userptr, "glRasterPos2iv");
+    glad_glRasterPos2s = (PFNGLRASTERPOS2SPROC) load(userptr, "glRasterPos2s");
+    glad_glRasterPos2sv = (PFNGLRASTERPOS2SVPROC) load(userptr, "glRasterPos2sv");
+    glad_glRasterPos3d = (PFNGLRASTERPOS3DPROC) load(userptr, "glRasterPos3d");
+    glad_glRasterPos3dv = (PFNGLRASTERPOS3DVPROC) load(userptr, "glRasterPos3dv");
+    glad_glRasterPos3f = (PFNGLRASTERPOS3FPROC) load(userptr, "glRasterPos3f");
+    glad_glRasterPos3fv = (PFNGLRASTERPOS3FVPROC) load(userptr, "glRasterPos3fv");
+    glad_glRasterPos3i = (PFNGLRASTERPOS3IPROC) load(userptr, "glRasterPos3i");
+    glad_glRasterPos3iv = (PFNGLRASTERPOS3IVPROC) load(userptr, "glRasterPos3iv");
+    glad_glRasterPos3s = (PFNGLRASTERPOS3SPROC) load(userptr, "glRasterPos3s");
+    glad_glRasterPos3sv = (PFNGLRASTERPOS3SVPROC) load(userptr, "glRasterPos3sv");
+    glad_glRasterPos4d = (PFNGLRASTERPOS4DPROC) load(userptr, "glRasterPos4d");
+    glad_glRasterPos4dv = (PFNGLRASTERPOS4DVPROC) load(userptr, "glRasterPos4dv");
+    glad_glRasterPos4f = (PFNGLRASTERPOS4FPROC) load(userptr, "glRasterPos4f");
+    glad_glRasterPos4fv = (PFNGLRASTERPOS4FVPROC) load(userptr, "glRasterPos4fv");
+    glad_glRasterPos4i = (PFNGLRASTERPOS4IPROC) load(userptr, "glRasterPos4i");
+    glad_glRasterPos4iv = (PFNGLRASTERPOS4IVPROC) load(userptr, "glRasterPos4iv");
+    glad_glRasterPos4s = (PFNGLRASTERPOS4SPROC) load(userptr, "glRasterPos4s");
+    glad_glRasterPos4sv = (PFNGLRASTERPOS4SVPROC) load(userptr, "glRasterPos4sv");
+    glad_glReadBuffer = (PFNGLREADBUFFERPROC) load(userptr, "glReadBuffer");
+    glad_glReadPixels = (PFNGLREADPIXELSPROC) load(userptr, "glReadPixels");
+    glad_glRectd = (PFNGLRECTDPROC) load(userptr, "glRectd");
+    glad_glRectdv = (PFNGLRECTDVPROC) load(userptr, "glRectdv");
+    glad_glRectf = (PFNGLRECTFPROC) load(userptr, "glRectf");
+    glad_glRectfv = (PFNGLRECTFVPROC) load(userptr, "glRectfv");
+    glad_glRecti = (PFNGLRECTIPROC) load(userptr, "glRecti");
+    glad_glRectiv = (PFNGLRECTIVPROC) load(userptr, "glRectiv");
+    glad_glRects = (PFNGLRECTSPROC) load(userptr, "glRects");
+    glad_glRectsv = (PFNGLRECTSVPROC) load(userptr, "glRectsv");
+    glad_glRenderMode = (PFNGLRENDERMODEPROC) load(userptr, "glRenderMode");
+    glad_glRotated = (PFNGLROTATEDPROC) load(userptr, "glRotated");
+    glad_glRotatef = (PFNGLROTATEFPROC) load(userptr, "glRotatef");
+    glad_glScaled = (PFNGLSCALEDPROC) load(userptr, "glScaled");
+    glad_glScalef = (PFNGLSCALEFPROC) load(userptr, "glScalef");
+    glad_glScissor = (PFNGLSCISSORPROC) load(userptr, "glScissor");
+    glad_glSelectBuffer = (PFNGLSELECTBUFFERPROC) load(userptr, "glSelectBuffer");
+    glad_glShadeModel = (PFNGLSHADEMODELPROC) load(userptr, "glShadeModel");
+    glad_glStencilFunc = (PFNGLSTENCILFUNCPROC) load(userptr, "glStencilFunc");
+    glad_glStencilMask = (PFNGLSTENCILMASKPROC) load(userptr, "glStencilMask");
+    glad_glStencilOp = (PFNGLSTENCILOPPROC) load(userptr, "glStencilOp");
+    glad_glTexCoord1d = (PFNGLTEXCOORD1DPROC) load(userptr, "glTexCoord1d");
+    glad_glTexCoord1dv = (PFNGLTEXCOORD1DVPROC) load(userptr, "glTexCoord1dv");
+    glad_glTexCoord1f = (PFNGLTEXCOORD1FPROC) load(userptr, "glTexCoord1f");
+    glad_glTexCoord1fv = (PFNGLTEXCOORD1FVPROC) load(userptr, "glTexCoord1fv");
+    glad_glTexCoord1i = (PFNGLTEXCOORD1IPROC) load(userptr, "glTexCoord1i");
+    glad_glTexCoord1iv = (PFNGLTEXCOORD1IVPROC) load(userptr, "glTexCoord1iv");
+    glad_glTexCoord1s = (PFNGLTEXCOORD1SPROC) load(userptr, "glTexCoord1s");
+    glad_glTexCoord1sv = (PFNGLTEXCOORD1SVPROC) load(userptr, "glTexCoord1sv");
+    glad_glTexCoord2d = (PFNGLTEXCOORD2DPROC) load(userptr, "glTexCoord2d");
+    glad_glTexCoord2dv = (PFNGLTEXCOORD2DVPROC) load(userptr, "glTexCoord2dv");
+    glad_glTexCoord2f = (PFNGLTEXCOORD2FPROC) load(userptr, "glTexCoord2f");
+    glad_glTexCoord2fv = (PFNGLTEXCOORD2FVPROC) load(userptr, "glTexCoord2fv");
+    glad_glTexCoord2i = (PFNGLTEXCOORD2IPROC) load(userptr, "glTexCoord2i");
+    glad_glTexCoord2iv = (PFNGLTEXCOORD2IVPROC) load(userptr, "glTexCoord2iv");
+    glad_glTexCoord2s = (PFNGLTEXCOORD2SPROC) load(userptr, "glTexCoord2s");
+    glad_glTexCoord2sv = (PFNGLTEXCOORD2SVPROC) load(userptr, "glTexCoord2sv");
+    glad_glTexCoord3d = (PFNGLTEXCOORD3DPROC) load(userptr, "glTexCoord3d");
+    glad_glTexCoord3dv = (PFNGLTEXCOORD3DVPROC) load(userptr, "glTexCoord3dv");
+    glad_glTexCoord3f = (PFNGLTEXCOORD3FPROC) load(userptr, "glTexCoord3f");
+    glad_glTexCoord3fv = (PFNGLTEXCOORD3FVPROC) load(userptr, "glTexCoord3fv");
+    glad_glTexCoord3i = (PFNGLTEXCOORD3IPROC) load(userptr, "glTexCoord3i");
+    glad_glTexCoord3iv = (PFNGLTEXCOORD3IVPROC) load(userptr, "glTexCoord3iv");
+    glad_glTexCoord3s = (PFNGLTEXCOORD3SPROC) load(userptr, "glTexCoord3s");
+    glad_glTexCoord3sv = (PFNGLTEXCOORD3SVPROC) load(userptr, "glTexCoord3sv");
+    glad_glTexCoord4d = (PFNGLTEXCOORD4DPROC) load(userptr, "glTexCoord4d");
+    glad_glTexCoord4dv = (PFNGLTEXCOORD4DVPROC) load(userptr, "glTexCoord4dv");
+    glad_glTexCoord4f = (PFNGLTEXCOORD4FPROC) load(userptr, "glTexCoord4f");
+    glad_glTexCoord4fv = (PFNGLTEXCOORD4FVPROC) load(userptr, "glTexCoord4fv");
+    glad_glTexCoord4i = (PFNGLTEXCOORD4IPROC) load(userptr, "glTexCoord4i");
+    glad_glTexCoord4iv = (PFNGLTEXCOORD4IVPROC) load(userptr, "glTexCoord4iv");
+    glad_glTexCoord4s = (PFNGLTEXCOORD4SPROC) load(userptr, "glTexCoord4s");
+    glad_glTexCoord4sv = (PFNGLTEXCOORD4SVPROC) load(userptr, "glTexCoord4sv");
+    glad_glTexEnvf = (PFNGLTEXENVFPROC) load(userptr, "glTexEnvf");
+    glad_glTexEnvfv = (PFNGLTEXENVFVPROC) load(userptr, "glTexEnvfv");
+    glad_glTexEnvi = (PFNGLTEXENVIPROC) load(userptr, "glTexEnvi");
+    glad_glTexEnviv = (PFNGLTEXENVIVPROC) load(userptr, "glTexEnviv");
+    glad_glTexGend = (PFNGLTEXGENDPROC) load(userptr, "glTexGend");
+    glad_glTexGendv = (PFNGLTEXGENDVPROC) load(userptr, "glTexGendv");
+    glad_glTexGenf = (PFNGLTEXGENFPROC) load(userptr, "glTexGenf");
+    glad_glTexGenfv = (PFNGLTEXGENFVPROC) load(userptr, "glTexGenfv");
+    glad_glTexGeni = (PFNGLTEXGENIPROC) load(userptr, "glTexGeni");
+    glad_glTexGeniv = (PFNGLTEXGENIVPROC) load(userptr, "glTexGeniv");
+    glad_glTexImage1D = (PFNGLTEXIMAGE1DPROC) load(userptr, "glTexImage1D");
+    glad_glTexImage2D = (PFNGLTEXIMAGE2DPROC) load(userptr, "glTexImage2D");
+    glad_glTexParameterf = (PFNGLTEXPARAMETERFPROC) load(userptr, "glTexParameterf");
+    glad_glTexParameterfv = (PFNGLTEXPARAMETERFVPROC) load(userptr, "glTexParameterfv");
+    glad_glTexParameteri = (PFNGLTEXPARAMETERIPROC) load(userptr, "glTexParameteri");
+    glad_glTexParameteriv = (PFNGLTEXPARAMETERIVPROC) load(userptr, "glTexParameteriv");
+    glad_glTranslated = (PFNGLTRANSLATEDPROC) load(userptr, "glTranslated");
+    glad_glTranslatef = (PFNGLTRANSLATEFPROC) load(userptr, "glTranslatef");
+    glad_glVertex2d = (PFNGLVERTEX2DPROC) load(userptr, "glVertex2d");
+    glad_glVertex2dv = (PFNGLVERTEX2DVPROC) load(userptr, "glVertex2dv");
+    glad_glVertex2f = (PFNGLVERTEX2FPROC) load(userptr, "glVertex2f");
+    glad_glVertex2fv = (PFNGLVERTEX2FVPROC) load(userptr, "glVertex2fv");
+    glad_glVertex2i = (PFNGLVERTEX2IPROC) load(userptr, "glVertex2i");
+    glad_glVertex2iv = (PFNGLVERTEX2IVPROC) load(userptr, "glVertex2iv");
+    glad_glVertex2s = (PFNGLVERTEX2SPROC) load(userptr, "glVertex2s");
+    glad_glVertex2sv = (PFNGLVERTEX2SVPROC) load(userptr, "glVertex2sv");
+    glad_glVertex3d = (PFNGLVERTEX3DPROC) load(userptr, "glVertex3d");
+    glad_glVertex3dv = (PFNGLVERTEX3DVPROC) load(userptr, "glVertex3dv");
+    glad_glVertex3f = (PFNGLVERTEX3FPROC) load(userptr, "glVertex3f");
+    glad_glVertex3fv = (PFNGLVERTEX3FVPROC) load(userptr, "glVertex3fv");
+    glad_glVertex3i = (PFNGLVERTEX3IPROC) load(userptr, "glVertex3i");
+    glad_glVertex3iv = (PFNGLVERTEX3IVPROC) load(userptr, "glVertex3iv");
+    glad_glVertex3s = (PFNGLVERTEX3SPROC) load(userptr, "glVertex3s");
+    glad_glVertex3sv = (PFNGLVERTEX3SVPROC) load(userptr, "glVertex3sv");
+    glad_glVertex4d = (PFNGLVERTEX4DPROC) load(userptr, "glVertex4d");
+    glad_glVertex4dv = (PFNGLVERTEX4DVPROC) load(userptr, "glVertex4dv");
+    glad_glVertex4f = (PFNGLVERTEX4FPROC) load(userptr, "glVertex4f");
+    glad_glVertex4fv = (PFNGLVERTEX4FVPROC) load(userptr, "glVertex4fv");
+    glad_glVertex4i = (PFNGLVERTEX4IPROC) load(userptr, "glVertex4i");
+    glad_glVertex4iv = (PFNGLVERTEX4IVPROC) load(userptr, "glVertex4iv");
+    glad_glVertex4s = (PFNGLVERTEX4SPROC) load(userptr, "glVertex4s");
+    glad_glVertex4sv = (PFNGLVERTEX4SVPROC) load(userptr, "glVertex4sv");
+    glad_glViewport = (PFNGLVIEWPORTPROC) load(userptr, "glViewport");
+}
+static void glad_gl_load_GL_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_1) return;
+    glad_glAreTexturesResident = (PFNGLARETEXTURESRESIDENTPROC) load(userptr, "glAreTexturesResident");
+    glad_glArrayElement = (PFNGLARRAYELEMENTPROC) load(userptr, "glArrayElement");
+    glad_glBindTexture = (PFNGLBINDTEXTUREPROC) load(userptr, "glBindTexture");
+    glad_glColorPointer = (PFNGLCOLORPOINTERPROC) load(userptr, "glColorPointer");
+    glad_glCopyTexImage1D = (PFNGLCOPYTEXIMAGE1DPROC) load(userptr, "glCopyTexImage1D");
+    glad_glCopyTexImage2D = (PFNGLCOPYTEXIMAGE2DPROC) load(userptr, "glCopyTexImage2D");
+    glad_glCopyTexSubImage1D = (PFNGLCOPYTEXSUBIMAGE1DPROC) load(userptr, "glCopyTexSubImage1D");
+    glad_glCopyTexSubImage2D = (PFNGLCOPYTEXSUBIMAGE2DPROC) load(userptr, "glCopyTexSubImage2D");
+    glad_glDeleteTextures = (PFNGLDELETETEXTURESPROC) load(userptr, "glDeleteTextures");
+    glad_glDisableClientState = (PFNGLDISABLECLIENTSTATEPROC) load(userptr, "glDisableClientState");
+    glad_glDrawArrays = (PFNGLDRAWARRAYSPROC) load(userptr, "glDrawArrays");
+    glad_glDrawElements = (PFNGLDRAWELEMENTSPROC) load(userptr, "glDrawElements");
+    glad_glEdgeFlagPointer = (PFNGLEDGEFLAGPOINTERPROC) load(userptr, "glEdgeFlagPointer");
+    glad_glEnableClientState = (PFNGLENABLECLIENTSTATEPROC) load(userptr, "glEnableClientState");
+    glad_glGenTextures = (PFNGLGENTEXTURESPROC) load(userptr, "glGenTextures");
+    glad_glGetPointerv = (PFNGLGETPOINTERVPROC) load(userptr, "glGetPointerv");
+    glad_glIndexPointer = (PFNGLINDEXPOINTERPROC) load(userptr, "glIndexPointer");
+    glad_glIndexub = (PFNGLINDEXUBPROC) load(userptr, "glIndexub");
+    glad_glIndexubv = (PFNGLINDEXUBVPROC) load(userptr, "glIndexubv");
+    glad_glInterleavedArrays = (PFNGLINTERLEAVEDARRAYSPROC) load(userptr, "glInterleavedArrays");
+    glad_glIsTexture = (PFNGLISTEXTUREPROC) load(userptr, "glIsTexture");
+    glad_glNormalPointer = (PFNGLNORMALPOINTERPROC) load(userptr, "glNormalPointer");
+    glad_glPolygonOffset = (PFNGLPOLYGONOFFSETPROC) load(userptr, "glPolygonOffset");
+    glad_glPopClientAttrib = (PFNGLPOPCLIENTATTRIBPROC) load(userptr, "glPopClientAttrib");
+    glad_glPrioritizeTextures = (PFNGLPRIORITIZETEXTURESPROC) load(userptr, "glPrioritizeTextures");
+    glad_glPushClientAttrib = (PFNGLPUSHCLIENTATTRIBPROC) load(userptr, "glPushClientAttrib");
+    glad_glTexCoordPointer = (PFNGLTEXCOORDPOINTERPROC) load(userptr, "glTexCoordPointer");
+    glad_glTexSubImage1D = (PFNGLTEXSUBIMAGE1DPROC) load(userptr, "glTexSubImage1D");
+    glad_glTexSubImage2D = (PFNGLTEXSUBIMAGE2DPROC) load(userptr, "glTexSubImage2D");
+    glad_glVertexPointer = (PFNGLVERTEXPOINTERPROC) load(userptr, "glVertexPointer");
+}
+static void glad_gl_load_GL_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_2) return;
+    glad_glCopyTexSubImage3D = (PFNGLCOPYTEXSUBIMAGE3DPROC) load(userptr, "glCopyTexSubImage3D");
+    glad_glDrawRangeElements = (PFNGLDRAWRANGEELEMENTSPROC) load(userptr, "glDrawRangeElements");
+    glad_glTexImage3D = (PFNGLTEXIMAGE3DPROC) load(userptr, "glTexImage3D");
+    glad_glTexSubImage3D = (PFNGLTEXSUBIMAGE3DPROC) load(userptr, "glTexSubImage3D");
+}
+static void glad_gl_load_GL_VERSION_1_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_3) return;
+    glad_glActiveTexture = (PFNGLACTIVETEXTUREPROC) load(userptr, "glActiveTexture");
+    glad_glClientActiveTexture = (PFNGLCLIENTACTIVETEXTUREPROC) load(userptr, "glClientActiveTexture");
+    glad_glCompressedTexImage1D = (PFNGLCOMPRESSEDTEXIMAGE1DPROC) load(userptr, "glCompressedTexImage1D");
+    glad_glCompressedTexImage2D = (PFNGLCOMPRESSEDTEXIMAGE2DPROC) load(userptr, "glCompressedTexImage2D");
+    glad_glCompressedTexImage3D = (PFNGLCOMPRESSEDTEXIMAGE3DPROC) load(userptr, "glCompressedTexImage3D");
+    glad_glCompressedTexSubImage1D = (PFNGLCOMPRESSEDTEXSUBIMAGE1DPROC) load(userptr, "glCompressedTexSubImage1D");
+    glad_glCompressedTexSubImage2D = (PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) load(userptr, "glCompressedTexSubImage2D");
+    glad_glCompressedTexSubImage3D = (PFNGLCOMPRESSEDTEXSUBIMAGE3DPROC) load(userptr, "glCompressedTexSubImage3D");
+    glad_glGetCompressedTexImage = (PFNGLGETCOMPRESSEDTEXIMAGEPROC) load(userptr, "glGetCompressedTexImage");
+    glad_glLoadTransposeMatrixd = (PFNGLLOADTRANSPOSEMATRIXDPROC) load(userptr, "glLoadTransposeMatrixd");
+    glad_glLoadTransposeMatrixf = (PFNGLLOADTRANSPOSEMATRIXFPROC) load(userptr, "glLoadTransposeMatrixf");
+    glad_glMultTransposeMatrixd = (PFNGLMULTTRANSPOSEMATRIXDPROC) load(userptr, "glMultTransposeMatrixd");
+    glad_glMultTransposeMatrixf = (PFNGLMULTTRANSPOSEMATRIXFPROC) load(userptr, "glMultTransposeMatrixf");
+    glad_glMultiTexCoord1d = (PFNGLMULTITEXCOORD1DPROC) load(userptr, "glMultiTexCoord1d");
+    glad_glMultiTexCoord1dv = (PFNGLMULTITEXCOORD1DVPROC) load(userptr, "glMultiTexCoord1dv");
+    glad_glMultiTexCoord1f = (PFNGLMULTITEXCOORD1FPROC) load(userptr, "glMultiTexCoord1f");
+    glad_glMultiTexCoord1fv = (PFNGLMULTITEXCOORD1FVPROC) load(userptr, "glMultiTexCoord1fv");
+    glad_glMultiTexCoord1i = (PFNGLMULTITEXCOORD1IPROC) load(userptr, "glMultiTexCoord1i");
+    glad_glMultiTexCoord1iv = (PFNGLMULTITEXCOORD1IVPROC) load(userptr, "glMultiTexCoord1iv");
+    glad_glMultiTexCoord1s = (PFNGLMULTITEXCOORD1SPROC) load(userptr, "glMultiTexCoord1s");
+    glad_glMultiTexCoord1sv = (PFNGLMULTITEXCOORD1SVPROC) load(userptr, "glMultiTexCoord1sv");
+    glad_glMultiTexCoord2d = (PFNGLMULTITEXCOORD2DPROC) load(userptr, "glMultiTexCoord2d");
+    glad_glMultiTexCoord2dv = (PFNGLMULTITEXCOORD2DVPROC) load(userptr, "glMultiTexCoord2dv");
+    glad_glMultiTexCoord2f = (PFNGLMULTITEXCOORD2FPROC) load(userptr, "glMultiTexCoord2f");
+    glad_glMultiTexCoord2fv = (PFNGLMULTITEXCOORD2FVPROC) load(userptr, "glMultiTexCoord2fv");
+    glad_glMultiTexCoord2i = (PFNGLMULTITEXCOORD2IPROC) load(userptr, "glMultiTexCoord2i");
+    glad_glMultiTexCoord2iv = (PFNGLMULTITEXCOORD2IVPROC) load(userptr, "glMultiTexCoord2iv");
+    glad_glMultiTexCoord2s = (PFNGLMULTITEXCOORD2SPROC) load(userptr, "glMultiTexCoord2s");
+    glad_glMultiTexCoord2sv = (PFNGLMULTITEXCOORD2SVPROC) load(userptr, "glMultiTexCoord2sv");
+    glad_glMultiTexCoord3d = (PFNGLMULTITEXCOORD3DPROC) load(userptr, "glMultiTexCoord3d");
+    glad_glMultiTexCoord3dv = (PFNGLMULTITEXCOORD3DVPROC) load(userptr, "glMultiTexCoord3dv");
+    glad_glMultiTexCoord3f = (PFNGLMULTITEXCOORD3FPROC) load(userptr, "glMultiTexCoord3f");
+    glad_glMultiTexCoord3fv = (PFNGLMULTITEXCOORD3FVPROC) load(userptr, "glMultiTexCoord3fv");
+    glad_glMultiTexCoord3i = (PFNGLMULTITEXCOORD3IPROC) load(userptr, "glMultiTexCoord3i");
+    glad_glMultiTexCoord3iv = (PFNGLMULTITEXCOORD3IVPROC) load(userptr, "glMultiTexCoord3iv");
+    glad_glMultiTexCoord3s = (PFNGLMULTITEXCOORD3SPROC) load(userptr, "glMultiTexCoord3s");
+    glad_glMultiTexCoord3sv = (PFNGLMULTITEXCOORD3SVPROC) load(userptr, "glMultiTexCoord3sv");
+    glad_glMultiTexCoord4d = (PFNGLMULTITEXCOORD4DPROC) load(userptr, "glMultiTexCoord4d");
+    glad_glMultiTexCoord4dv = (PFNGLMULTITEXCOORD4DVPROC) load(userptr, "glMultiTexCoord4dv");
+    glad_glMultiTexCoord4f = (PFNGLMULTITEXCOORD4FPROC) load(userptr, "glMultiTexCoord4f");
+    glad_glMultiTexCoord4fv = (PFNGLMULTITEXCOORD4FVPROC) load(userptr, "glMultiTexCoord4fv");
+    glad_glMultiTexCoord4i = (PFNGLMULTITEXCOORD4IPROC) load(userptr, "glMultiTexCoord4i");
+    glad_glMultiTexCoord4iv = (PFNGLMULTITEXCOORD4IVPROC) load(userptr, "glMultiTexCoord4iv");
+    glad_glMultiTexCoord4s = (PFNGLMULTITEXCOORD4SPROC) load(userptr, "glMultiTexCoord4s");
+    glad_glMultiTexCoord4sv = (PFNGLMULTITEXCOORD4SVPROC) load(userptr, "glMultiTexCoord4sv");
+    glad_glSampleCoverage = (PFNGLSAMPLECOVERAGEPROC) load(userptr, "glSampleCoverage");
+}
+static void glad_gl_load_GL_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_4) return;
+    glad_glBlendColor = (PFNGLBLENDCOLORPROC) load(userptr, "glBlendColor");
+    glad_glBlendEquation = (PFNGLBLENDEQUATIONPROC) load(userptr, "glBlendEquation");
+    glad_glBlendFuncSeparate = (PFNGLBLENDFUNCSEPARATEPROC) load(userptr, "glBlendFuncSeparate");
+    glad_glFogCoordPointer = (PFNGLFOGCOORDPOINTERPROC) load(userptr, "glFogCoordPointer");
+    glad_glFogCoordd = (PFNGLFOGCOORDDPROC) load(userptr, "glFogCoordd");
+    glad_glFogCoorddv = (PFNGLFOGCOORDDVPROC) load(userptr, "glFogCoorddv");
+    glad_glFogCoordf = (PFNGLFOGCOORDFPROC) load(userptr, "glFogCoordf");
+    glad_glFogCoordfv = (PFNGLFOGCOORDFVPROC) load(userptr, "glFogCoordfv");
+    glad_glMultiDrawArrays = (PFNGLMULTIDRAWARRAYSPROC) load(userptr, "glMultiDrawArrays");
+    glad_glMultiDrawElements = (PFNGLMULTIDRAWELEMENTSPROC) load(userptr, "glMultiDrawElements");
+    glad_glPointParameterf = (PFNGLPOINTPARAMETERFPROC) load(userptr, "glPointParameterf");
+    glad_glPointParameterfv = (PFNGLPOINTPARAMETERFVPROC) load(userptr, "glPointParameterfv");
+    glad_glPointParameteri = (PFNGLPOINTPARAMETERIPROC) load(userptr, "glPointParameteri");
+    glad_glPointParameteriv = (PFNGLPOINTPARAMETERIVPROC) load(userptr, "glPointParameteriv");
+    glad_glSecondaryColor3b = (PFNGLSECONDARYCOLOR3BPROC) load(userptr, "glSecondaryColor3b");
+    glad_glSecondaryColor3bv = (PFNGLSECONDARYCOLOR3BVPROC) load(userptr, "glSecondaryColor3bv");
+    glad_glSecondaryColor3d = (PFNGLSECONDARYCOLOR3DPROC) load(userptr, "glSecondaryColor3d");
+    glad_glSecondaryColor3dv = (PFNGLSECONDARYCOLOR3DVPROC) load(userptr, "glSecondaryColor3dv");
+    glad_glSecondaryColor3f = (PFNGLSECONDARYCOLOR3FPROC) load(userptr, "glSecondaryColor3f");
+    glad_glSecondaryColor3fv = (PFNGLSECONDARYCOLOR3FVPROC) load(userptr, "glSecondaryColor3fv");
+    glad_glSecondaryColor3i = (PFNGLSECONDARYCOLOR3IPROC) load(userptr, "glSecondaryColor3i");
+    glad_glSecondaryColor3iv = (PFNGLSECONDARYCOLOR3IVPROC) load(userptr, "glSecondaryColor3iv");
+    glad_glSecondaryColor3s = (PFNGLSECONDARYCOLOR3SPROC) load(userptr, "glSecondaryColor3s");
+    glad_glSecondaryColor3sv = (PFNGLSECONDARYCOLOR3SVPROC) load(userptr, "glSecondaryColor3sv");
+    glad_glSecondaryColor3ub = (PFNGLSECONDARYCOLOR3UBPROC) load(userptr, "glSecondaryColor3ub");
+    glad_glSecondaryColor3ubv = (PFNGLSECONDARYCOLOR3UBVPROC) load(userptr, "glSecondaryColor3ubv");
+    glad_glSecondaryColor3ui = (PFNGLSECONDARYCOLOR3UIPROC) load(userptr, "glSecondaryColor3ui");
+    glad_glSecondaryColor3uiv = (PFNGLSECONDARYCOLOR3UIVPROC) load(userptr, "glSecondaryColor3uiv");
+    glad_glSecondaryColor3us = (PFNGLSECONDARYCOLOR3USPROC) load(userptr, "glSecondaryColor3us");
+    glad_glSecondaryColor3usv = (PFNGLSECONDARYCOLOR3USVPROC) load(userptr, "glSecondaryColor3usv");
+    glad_glSecondaryColorPointer = (PFNGLSECONDARYCOLORPOINTERPROC) load(userptr, "glSecondaryColorPointer");
+    glad_glWindowPos2d = (PFNGLWINDOWPOS2DPROC) load(userptr, "glWindowPos2d");
+    glad_glWindowPos2dv = (PFNGLWINDOWPOS2DVPROC) load(userptr, "glWindowPos2dv");
+    glad_glWindowPos2f = (PFNGLWINDOWPOS2FPROC) load(userptr, "glWindowPos2f");
+    glad_glWindowPos2fv = (PFNGLWINDOWPOS2FVPROC) load(userptr, "glWindowPos2fv");
+    glad_glWindowPos2i = (PFNGLWINDOWPOS2IPROC) load(userptr, "glWindowPos2i");
+    glad_glWindowPos2iv = (PFNGLWINDOWPOS2IVPROC) load(userptr, "glWindowPos2iv");
+    glad_glWindowPos2s = (PFNGLWINDOWPOS2SPROC) load(userptr, "glWindowPos2s");
+    glad_glWindowPos2sv = (PFNGLWINDOWPOS2SVPROC) load(userptr, "glWindowPos2sv");
+    glad_glWindowPos3d = (PFNGLWINDOWPOS3DPROC) load(userptr, "glWindowPos3d");
+    glad_glWindowPos3dv = (PFNGLWINDOWPOS3DVPROC) load(userptr, "glWindowPos3dv");
+    glad_glWindowPos3f = (PFNGLWINDOWPOS3FPROC) load(userptr, "glWindowPos3f");
+    glad_glWindowPos3fv = (PFNGLWINDOWPOS3FVPROC) load(userptr, "glWindowPos3fv");
+    glad_glWindowPos3i = (PFNGLWINDOWPOS3IPROC) load(userptr, "glWindowPos3i");
+    glad_glWindowPos3iv = (PFNGLWINDOWPOS3IVPROC) load(userptr, "glWindowPos3iv");
+    glad_glWindowPos3s = (PFNGLWINDOWPOS3SPROC) load(userptr, "glWindowPos3s");
+    glad_glWindowPos3sv = (PFNGLWINDOWPOS3SVPROC) load(userptr, "glWindowPos3sv");
+}
+static void glad_gl_load_GL_VERSION_1_5( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_1_5) return;
+    glad_glBeginQuery = (PFNGLBEGINQUERYPROC) load(userptr, "glBeginQuery");
+    glad_glBindBuffer = (PFNGLBINDBUFFERPROC) load(userptr, "glBindBuffer");
+    glad_glBufferData = (PFNGLBUFFERDATAPROC) load(userptr, "glBufferData");
+    glad_glBufferSubData = (PFNGLBUFFERSUBDATAPROC) load(userptr, "glBufferSubData");
+    glad_glDeleteBuffers = (PFNGLDELETEBUFFERSPROC) load(userptr, "glDeleteBuffers");
+    glad_glDeleteQueries = (PFNGLDELETEQUERIESPROC) load(userptr, "glDeleteQueries");
+    glad_glEndQuery = (PFNGLENDQUERYPROC) load(userptr, "glEndQuery");
+    glad_glGenBuffers = (PFNGLGENBUFFERSPROC) load(userptr, "glGenBuffers");
+    glad_glGenQueries = (PFNGLGENQUERIESPROC) load(userptr, "glGenQueries");
+    glad_glGetBufferParameteriv = (PFNGLGETBUFFERPARAMETERIVPROC) load(userptr, "glGetBufferParameteriv");
+    glad_glGetBufferPointerv = (PFNGLGETBUFFERPOINTERVPROC) load(userptr, "glGetBufferPointerv");
+    glad_glGetBufferSubData = (PFNGLGETBUFFERSUBDATAPROC) load(userptr, "glGetBufferSubData");
+    glad_glGetQueryObjectiv = (PFNGLGETQUERYOBJECTIVPROC) load(userptr, "glGetQueryObjectiv");
+    glad_glGetQueryObjectuiv = (PFNGLGETQUERYOBJECTUIVPROC) load(userptr, "glGetQueryObjectuiv");
+    glad_glGetQueryiv = (PFNGLGETQUERYIVPROC) load(userptr, "glGetQueryiv");
+    glad_glIsBuffer = (PFNGLISBUFFERPROC) load(userptr, "glIsBuffer");
+    glad_glIsQuery = (PFNGLISQUERYPROC) load(userptr, "glIsQuery");
+    glad_glMapBuffer = (PFNGLMAPBUFFERPROC) load(userptr, "glMapBuffer");
+    glad_glUnmapBuffer = (PFNGLUNMAPBUFFERPROC) load(userptr, "glUnmapBuffer");
+}
+static void glad_gl_load_GL_VERSION_2_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_2_0) return;
+    glad_glAttachShader = (PFNGLATTACHSHADERPROC) load(userptr, "glAttachShader");
+    glad_glBindAttribLocation = (PFNGLBINDATTRIBLOCATIONPROC) load(userptr, "glBindAttribLocation");
+    glad_glBlendEquationSeparate = (PFNGLBLENDEQUATIONSEPARATEPROC) load(userptr, "glBlendEquationSeparate");
+    glad_glCompileShader = (PFNGLCOMPILESHADERPROC) load(userptr, "glCompileShader");
+    glad_glCreateProgram = (PFNGLCREATEPROGRAMPROC) load(userptr, "glCreateProgram");
+    glad_glCreateShader = (PFNGLCREATESHADERPROC) load(userptr, "glCreateShader");
+    glad_glDeleteProgram = (PFNGLDELETEPROGRAMPROC) load(userptr, "glDeleteProgram");
+    glad_glDeleteShader = (PFNGLDELETESHADERPROC) load(userptr, "glDeleteShader");
+    glad_glDetachShader = (PFNGLDETACHSHADERPROC) load(userptr, "glDetachShader");
+    glad_glDisableVertexAttribArray = (PFNGLDISABLEVERTEXATTRIBARRAYPROC) load(userptr, "glDisableVertexAttribArray");
+    glad_glDrawBuffers = (PFNGLDRAWBUFFERSPROC) load(userptr, "glDrawBuffers");
+    glad_glEnableVertexAttribArray = (PFNGLENABLEVERTEXATTRIBARRAYPROC) load(userptr, "glEnableVertexAttribArray");
+    glad_glGetActiveAttrib = (PFNGLGETACTIVEATTRIBPROC) load(userptr, "glGetActiveAttrib");
+    glad_glGetActiveUniform = (PFNGLGETACTIVEUNIFORMPROC) load(userptr, "glGetActiveUniform");
+    glad_glGetAttachedShaders = (PFNGLGETATTACHEDSHADERSPROC) load(userptr, "glGetAttachedShaders");
+    glad_glGetAttribLocation = (PFNGLGETATTRIBLOCATIONPROC) load(userptr, "glGetAttribLocation");
+    glad_glGetProgramInfoLog = (PFNGLGETPROGRAMINFOLOGPROC) load(userptr, "glGetProgramInfoLog");
+    glad_glGetProgramiv = (PFNGLGETPROGRAMIVPROC) load(userptr, "glGetProgramiv");
+    glad_glGetShaderInfoLog = (PFNGLGETSHADERINFOLOGPROC) load(userptr, "glGetShaderInfoLog");
+    glad_glGetShaderSource = (PFNGLGETSHADERSOURCEPROC) load(userptr, "glGetShaderSource");
+    glad_glGetShaderiv = (PFNGLGETSHADERIVPROC) load(userptr, "glGetShaderiv");
+    glad_glGetUniformLocation = (PFNGLGETUNIFORMLOCATIONPROC) load(userptr, "glGetUniformLocation");
+    glad_glGetUniformfv = (PFNGLGETUNIFORMFVPROC) load(userptr, "glGetUniformfv");
+    glad_glGetUniformiv = (PFNGLGETUNIFORMIVPROC) load(userptr, "glGetUniformiv");
+    glad_glGetVertexAttribPointerv = (PFNGLGETVERTEXATTRIBPOINTERVPROC) load(userptr, "glGetVertexAttribPointerv");
+    glad_glGetVertexAttribdv = (PFNGLGETVERTEXATTRIBDVPROC) load(userptr, "glGetVertexAttribdv");
+    glad_glGetVertexAttribfv = (PFNGLGETVERTEXATTRIBFVPROC) load(userptr, "glGetVertexAttribfv");
+    glad_glGetVertexAttribiv = (PFNGLGETVERTEXATTRIBIVPROC) load(userptr, "glGetVertexAttribiv");
+    glad_glIsProgram = (PFNGLISPROGRAMPROC) load(userptr, "glIsProgram");
+    glad_glIsShader = (PFNGLISSHADERPROC) load(userptr, "glIsShader");
+    glad_glLinkProgram = (PFNGLLINKPROGRAMPROC) load(userptr, "glLinkProgram");
+    glad_glShaderSource = (PFNGLSHADERSOURCEPROC) load(userptr, "glShaderSource");
+    glad_glStencilFuncSeparate = (PFNGLSTENCILFUNCSEPARATEPROC) load(userptr, "glStencilFuncSeparate");
+    glad_glStencilMaskSeparate = (PFNGLSTENCILMASKSEPARATEPROC) load(userptr, "glStencilMaskSeparate");
+    glad_glStencilOpSeparate = (PFNGLSTENCILOPSEPARATEPROC) load(userptr, "glStencilOpSeparate");
+    glad_glUniform1f = (PFNGLUNIFORM1FPROC) load(userptr, "glUniform1f");
+    glad_glUniform1fv = (PFNGLUNIFORM1FVPROC) load(userptr, "glUniform1fv");
+    glad_glUniform1i = (PFNGLUNIFORM1IPROC) load(userptr, "glUniform1i");
+    glad_glUniform1iv = (PFNGLUNIFORM1IVPROC) load(userptr, "glUniform1iv");
+    glad_glUniform2f = (PFNGLUNIFORM2FPROC) load(userptr, "glUniform2f");
+    glad_glUniform2fv = (PFNGLUNIFORM2FVPROC) load(userptr, "glUniform2fv");
+    glad_glUniform2i = (PFNGLUNIFORM2IPROC) load(userptr, "glUniform2i");
+    glad_glUniform2iv = (PFNGLUNIFORM2IVPROC) load(userptr, "glUniform2iv");
+    glad_glUniform3f = (PFNGLUNIFORM3FPROC) load(userptr, "glUniform3f");
+    glad_glUniform3fv = (PFNGLUNIFORM3FVPROC) load(userptr, "glUniform3fv");
+    glad_glUniform3i = (PFNGLUNIFORM3IPROC) load(userptr, "glUniform3i");
+    glad_glUniform3iv = (PFNGLUNIFORM3IVPROC) load(userptr, "glUniform3iv");
+    glad_glUniform4f = (PFNGLUNIFORM4FPROC) load(userptr, "glUniform4f");
+    glad_glUniform4fv = (PFNGLUNIFORM4FVPROC) load(userptr, "glUniform4fv");
+    glad_glUniform4i = (PFNGLUNIFORM4IPROC) load(userptr, "glUniform4i");
+    glad_glUniform4iv = (PFNGLUNIFORM4IVPROC) load(userptr, "glUniform4iv");
+    glad_glUniformMatrix2fv = (PFNGLUNIFORMMATRIX2FVPROC) load(userptr, "glUniformMatrix2fv");
+    glad_glUniformMatrix3fv = (PFNGLUNIFORMMATRIX3FVPROC) load(userptr, "glUniformMatrix3fv");
+    glad_glUniformMatrix4fv = (PFNGLUNIFORMMATRIX4FVPROC) load(userptr, "glUniformMatrix4fv");
+    glad_glUseProgram = (PFNGLUSEPROGRAMPROC) load(userptr, "glUseProgram");
+    glad_glValidateProgram = (PFNGLVALIDATEPROGRAMPROC) load(userptr, "glValidateProgram");
+    glad_glVertexAttrib1d = (PFNGLVERTEXATTRIB1DPROC) load(userptr, "glVertexAttrib1d");
+    glad_glVertexAttrib1dv = (PFNGLVERTEXATTRIB1DVPROC) load(userptr, "glVertexAttrib1dv");
+    glad_glVertexAttrib1f = (PFNGLVERTEXATTRIB1FPROC) load(userptr, "glVertexAttrib1f");
+    glad_glVertexAttrib1fv = (PFNGLVERTEXATTRIB1FVPROC) load(userptr, "glVertexAttrib1fv");
+    glad_glVertexAttrib1s = (PFNGLVERTEXATTRIB1SPROC) load(userptr, "glVertexAttrib1s");
+    glad_glVertexAttrib1sv = (PFNGLVERTEXATTRIB1SVPROC) load(userptr, "glVertexAttrib1sv");
+    glad_glVertexAttrib2d = (PFNGLVERTEXATTRIB2DPROC) load(userptr, "glVertexAttrib2d");
+    glad_glVertexAttrib2dv = (PFNGLVERTEXATTRIB2DVPROC) load(userptr, "glVertexAttrib2dv");
+    glad_glVertexAttrib2f = (PFNGLVERTEXATTRIB2FPROC) load(userptr, "glVertexAttrib2f");
+    glad_glVertexAttrib2fv = (PFNGLVERTEXATTRIB2FVPROC) load(userptr, "glVertexAttrib2fv");
+    glad_glVertexAttrib2s = (PFNGLVERTEXATTRIB2SPROC) load(userptr, "glVertexAttrib2s");
+    glad_glVertexAttrib2sv = (PFNGLVERTEXATTRIB2SVPROC) load(userptr, "glVertexAttrib2sv");
+    glad_glVertexAttrib3d = (PFNGLVERTEXATTRIB3DPROC) load(userptr, "glVertexAttrib3d");
+    glad_glVertexAttrib3dv = (PFNGLVERTEXATTRIB3DVPROC) load(userptr, "glVertexAttrib3dv");
+    glad_glVertexAttrib3f = (PFNGLVERTEXATTRIB3FPROC) load(userptr, "glVertexAttrib3f");
+    glad_glVertexAttrib3fv = (PFNGLVERTEXATTRIB3FVPROC) load(userptr, "glVertexAttrib3fv");
+    glad_glVertexAttrib3s = (PFNGLVERTEXATTRIB3SPROC) load(userptr, "glVertexAttrib3s");
+    glad_glVertexAttrib3sv = (PFNGLVERTEXATTRIB3SVPROC) load(userptr, "glVertexAttrib3sv");
+    glad_glVertexAttrib4Nbv = (PFNGLVERTEXATTRIB4NBVPROC) load(userptr, "glVertexAttrib4Nbv");
+    glad_glVertexAttrib4Niv = (PFNGLVERTEXATTRIB4NIVPROC) load(userptr, "glVertexAttrib4Niv");
+    glad_glVertexAttrib4Nsv = (PFNGLVERTEXATTRIB4NSVPROC) load(userptr, "glVertexAttrib4Nsv");
+    glad_glVertexAttrib4Nub = (PFNGLVERTEXATTRIB4NUBPROC) load(userptr, "glVertexAttrib4Nub");
+    glad_glVertexAttrib4Nubv = (PFNGLVERTEXATTRIB4NUBVPROC) load(userptr, "glVertexAttrib4Nubv");
+    glad_glVertexAttrib4Nuiv = (PFNGLVERTEXATTRIB4NUIVPROC) load(userptr, "glVertexAttrib4Nuiv");
+    glad_glVertexAttrib4Nusv = (PFNGLVERTEXATTRIB4NUSVPROC) load(userptr, "glVertexAttrib4Nusv");
+    glad_glVertexAttrib4bv = (PFNGLVERTEXATTRIB4BVPROC) load(userptr, "glVertexAttrib4bv");
+    glad_glVertexAttrib4d = (PFNGLVERTEXATTRIB4DPROC) load(userptr, "glVertexAttrib4d");
+    glad_glVertexAttrib4dv = (PFNGLVERTEXATTRIB4DVPROC) load(userptr, "glVertexAttrib4dv");
+    glad_glVertexAttrib4f = (PFNGLVERTEXATTRIB4FPROC) load(userptr, "glVertexAttrib4f");
+    glad_glVertexAttrib4fv = (PFNGLVERTEXATTRIB4FVPROC) load(userptr, "glVertexAttrib4fv");
+    glad_glVertexAttrib4iv = (PFNGLVERTEXATTRIB4IVPROC) load(userptr, "glVertexAttrib4iv");
+    glad_glVertexAttrib4s = (PFNGLVERTEXATTRIB4SPROC) load(userptr, "glVertexAttrib4s");
+    glad_glVertexAttrib4sv = (PFNGLVERTEXATTRIB4SVPROC) load(userptr, "glVertexAttrib4sv");
+    glad_glVertexAttrib4ubv = (PFNGLVERTEXATTRIB4UBVPROC) load(userptr, "glVertexAttrib4ubv");
+    glad_glVertexAttrib4uiv = (PFNGLVERTEXATTRIB4UIVPROC) load(userptr, "glVertexAttrib4uiv");
+    glad_glVertexAttrib4usv = (PFNGLVERTEXATTRIB4USVPROC) load(userptr, "glVertexAttrib4usv");
+    glad_glVertexAttribPointer = (PFNGLVERTEXATTRIBPOINTERPROC) load(userptr, "glVertexAttribPointer");
+}
+static void glad_gl_load_GL_VERSION_2_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_2_1) return;
+    glad_glUniformMatrix2x3fv = (PFNGLUNIFORMMATRIX2X3FVPROC) load(userptr, "glUniformMatrix2x3fv");
+    glad_glUniformMatrix2x4fv = (PFNGLUNIFORMMATRIX2X4FVPROC) load(userptr, "glUniformMatrix2x4fv");
+    glad_glUniformMatrix3x2fv = (PFNGLUNIFORMMATRIX3X2FVPROC) load(userptr, "glUniformMatrix3x2fv");
+    glad_glUniformMatrix3x4fv = (PFNGLUNIFORMMATRIX3X4FVPROC) load(userptr, "glUniformMatrix3x4fv");
+    glad_glUniformMatrix4x2fv = (PFNGLUNIFORMMATRIX4X2FVPROC) load(userptr, "glUniformMatrix4x2fv");
+    glad_glUniformMatrix4x3fv = (PFNGLUNIFORMMATRIX4X3FVPROC) load(userptr, "glUniformMatrix4x3fv");
+}
+static void glad_gl_load_GL_VERSION_3_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_0) return;
+    glad_glBeginConditionalRender = (PFNGLBEGINCONDITIONALRENDERPROC) load(userptr, "glBeginConditionalRender");
+    glad_glBeginTransformFeedback = (PFNGLBEGINTRANSFORMFEEDBACKPROC) load(userptr, "glBeginTransformFeedback");
+    glad_glBindBufferBase = (PFNGLBINDBUFFERBASEPROC) load(userptr, "glBindBufferBase");
+    glad_glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC) load(userptr, "glBindBufferRange");
+    glad_glBindFragDataLocation = (PFNGLBINDFRAGDATALOCATIONPROC) load(userptr, "glBindFragDataLocation");
+    glad_glBindFramebuffer = (PFNGLBINDFRAMEBUFFERPROC) load(userptr, "glBindFramebuffer");
+    glad_glBindRenderbuffer = (PFNGLBINDRENDERBUFFERPROC) load(userptr, "glBindRenderbuffer");
+    glad_glBindVertexArray = (PFNGLBINDVERTEXARRAYPROC) load(userptr, "glBindVertexArray");
+    glad_glBlitFramebuffer = (PFNGLBLITFRAMEBUFFERPROC) load(userptr, "glBlitFramebuffer");
+    glad_glCheckFramebufferStatus = (PFNGLCHECKFRAMEBUFFERSTATUSPROC) load(userptr, "glCheckFramebufferStatus");
+    glad_glClampColor = (PFNGLCLAMPCOLORPROC) load(userptr, "glClampColor");
+    glad_glClearBufferfi = (PFNGLCLEARBUFFERFIPROC) load(userptr, "glClearBufferfi");
+    glad_glClearBufferfv = (PFNGLCLEARBUFFERFVPROC) load(userptr, "glClearBufferfv");
+    glad_glClearBufferiv = (PFNGLCLEARBUFFERIVPROC) load(userptr, "glClearBufferiv");
+    glad_glClearBufferuiv = (PFNGLCLEARBUFFERUIVPROC) load(userptr, "glClearBufferuiv");
+    glad_glColorMaski = (PFNGLCOLORMASKIPROC) load(userptr, "glColorMaski");
+    glad_glDeleteFramebuffers = (PFNGLDELETEFRAMEBUFFERSPROC) load(userptr, "glDeleteFramebuffers");
+    glad_glDeleteRenderbuffers = (PFNGLDELETERENDERBUFFERSPROC) load(userptr, "glDeleteRenderbuffers");
+    glad_glDeleteVertexArrays = (PFNGLDELETEVERTEXARRAYSPROC) load(userptr, "glDeleteVertexArrays");
+    glad_glDisablei = (PFNGLDISABLEIPROC) load(userptr, "glDisablei");
+    glad_glEnablei = (PFNGLENABLEIPROC) load(userptr, "glEnablei");
+    glad_glEndConditionalRender = (PFNGLENDCONDITIONALRENDERPROC) load(userptr, "glEndConditionalRender");
+    glad_glEndTransformFeedback = (PFNGLENDTRANSFORMFEEDBACKPROC) load(userptr, "glEndTransformFeedback");
+    glad_glFlushMappedBufferRange = (PFNGLFLUSHMAPPEDBUFFERRANGEPROC) load(userptr, "glFlushMappedBufferRange");
+    glad_glFramebufferRenderbuffer = (PFNGLFRAMEBUFFERRENDERBUFFERPROC) load(userptr, "glFramebufferRenderbuffer");
+    glad_glFramebufferTexture1D = (PFNGLFRAMEBUFFERTEXTURE1DPROC) load(userptr, "glFramebufferTexture1D");
+    glad_glFramebufferTexture2D = (PFNGLFRAMEBUFFERTEXTURE2DPROC) load(userptr, "glFramebufferTexture2D");
+    glad_glFramebufferTexture3D = (PFNGLFRAMEBUFFERTEXTURE3DPROC) load(userptr, "glFramebufferTexture3D");
+    glad_glFramebufferTextureLayer = (PFNGLFRAMEBUFFERTEXTURELAYERPROC) load(userptr, "glFramebufferTextureLayer");
+    glad_glGenFramebuffers = (PFNGLGENFRAMEBUFFERSPROC) load(userptr, "glGenFramebuffers");
+    glad_glGenRenderbuffers = (PFNGLGENRENDERBUFFERSPROC) load(userptr, "glGenRenderbuffers");
+    glad_glGenVertexArrays = (PFNGLGENVERTEXARRAYSPROC) load(userptr, "glGenVertexArrays");
+    glad_glGenerateMipmap = (PFNGLGENERATEMIPMAPPROC) load(userptr, "glGenerateMipmap");
+    glad_glGetBooleani_v = (PFNGLGETBOOLEANI_VPROC) load(userptr, "glGetBooleani_v");
+    glad_glGetFragDataLocation = (PFNGLGETFRAGDATALOCATIONPROC) load(userptr, "glGetFragDataLocation");
+    glad_glGetFramebufferAttachmentParameteriv = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC) load(userptr, "glGetFramebufferAttachmentParameteriv");
+    glad_glGetIntegeri_v = (PFNGLGETINTEGERI_VPROC) load(userptr, "glGetIntegeri_v");
+    glad_glGetRenderbufferParameteriv = (PFNGLGETRENDERBUFFERPARAMETERIVPROC) load(userptr, "glGetRenderbufferParameteriv");
+    glad_glGetStringi = (PFNGLGETSTRINGIPROC) load(userptr, "glGetStringi");
+    glad_glGetTexParameterIiv = (PFNGLGETTEXPARAMETERIIVPROC) load(userptr, "glGetTexParameterIiv");
+    glad_glGetTexParameterIuiv = (PFNGLGETTEXPARAMETERIUIVPROC) load(userptr, "glGetTexParameterIuiv");
+    glad_glGetTransformFeedbackVarying = (PFNGLGETTRANSFORMFEEDBACKVARYINGPROC) load(userptr, "glGetTransformFeedbackVarying");
+    glad_glGetUniformuiv = (PFNGLGETUNIFORMUIVPROC) load(userptr, "glGetUniformuiv");
+    glad_glGetVertexAttribIiv = (PFNGLGETVERTEXATTRIBIIVPROC) load(userptr, "glGetVertexAttribIiv");
+    glad_glGetVertexAttribIuiv = (PFNGLGETVERTEXATTRIBIUIVPROC) load(userptr, "glGetVertexAttribIuiv");
+    glad_glIsEnabledi = (PFNGLISENABLEDIPROC) load(userptr, "glIsEnabledi");
+    glad_glIsFramebuffer = (PFNGLISFRAMEBUFFERPROC) load(userptr, "glIsFramebuffer");
+    glad_glIsRenderbuffer = (PFNGLISRENDERBUFFERPROC) load(userptr, "glIsRenderbuffer");
+    glad_glIsVertexArray = (PFNGLISVERTEXARRAYPROC) load(userptr, "glIsVertexArray");
+    glad_glMapBufferRange = (PFNGLMAPBUFFERRANGEPROC) load(userptr, "glMapBufferRange");
+    glad_glRenderbufferStorage = (PFNGLRENDERBUFFERSTORAGEPROC) load(userptr, "glRenderbufferStorage");
+    glad_glRenderbufferStorageMultisample = (PFNGLRENDERBUFFERSTORAGEMULTISAMPLEPROC) load(userptr, "glRenderbufferStorageMultisample");
+    glad_glTexParameterIiv = (PFNGLTEXPARAMETERIIVPROC) load(userptr, "glTexParameterIiv");
+    glad_glTexParameterIuiv = (PFNGLTEXPARAMETERIUIVPROC) load(userptr, "glTexParameterIuiv");
+    glad_glTransformFeedbackVaryings = (PFNGLTRANSFORMFEEDBACKVARYINGSPROC) load(userptr, "glTransformFeedbackVaryings");
+    glad_glUniform1ui = (PFNGLUNIFORM1UIPROC) load(userptr, "glUniform1ui");
+    glad_glUniform1uiv = (PFNGLUNIFORM1UIVPROC) load(userptr, "glUniform1uiv");
+    glad_glUniform2ui = (PFNGLUNIFORM2UIPROC) load(userptr, "glUniform2ui");
+    glad_glUniform2uiv = (PFNGLUNIFORM2UIVPROC) load(userptr, "glUniform2uiv");
+    glad_glUniform3ui = (PFNGLUNIFORM3UIPROC) load(userptr, "glUniform3ui");
+    glad_glUniform3uiv = (PFNGLUNIFORM3UIVPROC) load(userptr, "glUniform3uiv");
+    glad_glUniform4ui = (PFNGLUNIFORM4UIPROC) load(userptr, "glUniform4ui");
+    glad_glUniform4uiv = (PFNGLUNIFORM4UIVPROC) load(userptr, "glUniform4uiv");
+    glad_glVertexAttribI1i = (PFNGLVERTEXATTRIBI1IPROC) load(userptr, "glVertexAttribI1i");
+    glad_glVertexAttribI1iv = (PFNGLVERTEXATTRIBI1IVPROC) load(userptr, "glVertexAttribI1iv");
+    glad_glVertexAttribI1ui = (PFNGLVERTEXATTRIBI1UIPROC) load(userptr, "glVertexAttribI1ui");
+    glad_glVertexAttribI1uiv = (PFNGLVERTEXATTRIBI1UIVPROC) load(userptr, "glVertexAttribI1uiv");
+    glad_glVertexAttribI2i = (PFNGLVERTEXATTRIBI2IPROC) load(userptr, "glVertexAttribI2i");
+    glad_glVertexAttribI2iv = (PFNGLVERTEXATTRIBI2IVPROC) load(userptr, "glVertexAttribI2iv");
+    glad_glVertexAttribI2ui = (PFNGLVERTEXATTRIBI2UIPROC) load(userptr, "glVertexAttribI2ui");
+    glad_glVertexAttribI2uiv = (PFNGLVERTEXATTRIBI2UIVPROC) load(userptr, "glVertexAttribI2uiv");
+    glad_glVertexAttribI3i = (PFNGLVERTEXATTRIBI3IPROC) load(userptr, "glVertexAttribI3i");
+    glad_glVertexAttribI3iv = (PFNGLVERTEXATTRIBI3IVPROC) load(userptr, "glVertexAttribI3iv");
+    glad_glVertexAttribI3ui = (PFNGLVERTEXATTRIBI3UIPROC) load(userptr, "glVertexAttribI3ui");
+    glad_glVertexAttribI3uiv = (PFNGLVERTEXATTRIBI3UIVPROC) load(userptr, "glVertexAttribI3uiv");
+    glad_glVertexAttribI4bv = (PFNGLVERTEXATTRIBI4BVPROC) load(userptr, "glVertexAttribI4bv");
+    glad_glVertexAttribI4i = (PFNGLVERTEXATTRIBI4IPROC) load(userptr, "glVertexAttribI4i");
+    glad_glVertexAttribI4iv = (PFNGLVERTEXATTRIBI4IVPROC) load(userptr, "glVertexAttribI4iv");
+    glad_glVertexAttribI4sv = (PFNGLVERTEXATTRIBI4SVPROC) load(userptr, "glVertexAttribI4sv");
+    glad_glVertexAttribI4ubv = (PFNGLVERTEXATTRIBI4UBVPROC) load(userptr, "glVertexAttribI4ubv");
+    glad_glVertexAttribI4ui = (PFNGLVERTEXATTRIBI4UIPROC) load(userptr, "glVertexAttribI4ui");
+    glad_glVertexAttribI4uiv = (PFNGLVERTEXATTRIBI4UIVPROC) load(userptr, "glVertexAttribI4uiv");
+    glad_glVertexAttribI4usv = (PFNGLVERTEXATTRIBI4USVPROC) load(userptr, "glVertexAttribI4usv");
+    glad_glVertexAttribIPointer = (PFNGLVERTEXATTRIBIPOINTERPROC) load(userptr, "glVertexAttribIPointer");
+}
+static void glad_gl_load_GL_VERSION_3_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_1) return;
+    glad_glBindBufferBase = (PFNGLBINDBUFFERBASEPROC) load(userptr, "glBindBufferBase");
+    glad_glBindBufferRange = (PFNGLBINDBUFFERRANGEPROC) load(userptr, "glBindBufferRange");
+    glad_glCopyBufferSubData = (PFNGLCOPYBUFFERSUBDATAPROC) load(userptr, "glCopyBufferSubData");
+    glad_glDrawArraysInstanced = (PFNGLDRAWARRAYSINSTANCEDPROC) load(userptr, "glDrawArraysInstanced");
+    glad_glDrawElementsInstanced = (PFNGLDRAWELEMENTSINSTANCEDPROC) load(userptr, "glDrawElementsInstanced");
+    glad_glGetActiveUniformBlockName = (PFNGLGETACTIVEUNIFORMBLOCKNAMEPROC) load(userptr, "glGetActiveUniformBlockName");
+    glad_glGetActiveUniformBlockiv = (PFNGLGETACTIVEUNIFORMBLOCKIVPROC) load(userptr, "glGetActiveUniformBlockiv");
+    glad_glGetActiveUniformName = (PFNGLGETACTIVEUNIFORMNAMEPROC) load(userptr, "glGetActiveUniformName");
+    glad_glGetActiveUniformsiv = (PFNGLGETACTIVEUNIFORMSIVPROC) load(userptr, "glGetActiveUniformsiv");
+    glad_glGetIntegeri_v = (PFNGLGETINTEGERI_VPROC) load(userptr, "glGetIntegeri_v");
+    glad_glGetUniformBlockIndex = (PFNGLGETUNIFORMBLOCKINDEXPROC) load(userptr, "glGetUniformBlockIndex");
+    glad_glGetUniformIndices = (PFNGLGETUNIFORMINDICESPROC) load(userptr, "glGetUniformIndices");
+    glad_glPrimitiveRestartIndex = (PFNGLPRIMITIVERESTARTINDEXPROC) load(userptr, "glPrimitiveRestartIndex");
+    glad_glTexBuffer = (PFNGLTEXBUFFERPROC) load(userptr, "glTexBuffer");
+    glad_glUniformBlockBinding = (PFNGLUNIFORMBLOCKBINDINGPROC) load(userptr, "glUniformBlockBinding");
+}
+static void glad_gl_load_GL_VERSION_3_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_2) return;
+    glad_glClientWaitSync = (PFNGLCLIENTWAITSYNCPROC) load(userptr, "glClientWaitSync");
+    glad_glDeleteSync = (PFNGLDELETESYNCPROC) load(userptr, "glDeleteSync");
+    glad_glDrawElementsBaseVertex = (PFNGLDRAWELEMENTSBASEVERTEXPROC) load(userptr, "glDrawElementsBaseVertex");
+    glad_glDrawElementsInstancedBaseVertex = (PFNGLDRAWELEMENTSINSTANCEDBASEVERTEXPROC) load(userptr, "glDrawElementsInstancedBaseVertex");
+    glad_glDrawRangeElementsBaseVertex = (PFNGLDRAWRANGEELEMENTSBASEVERTEXPROC) load(userptr, "glDrawRangeElementsBaseVertex");
+    glad_glFenceSync = (PFNGLFENCESYNCPROC) load(userptr, "glFenceSync");
+    glad_glFramebufferTexture = (PFNGLFRAMEBUFFERTEXTUREPROC) load(userptr, "glFramebufferTexture");
+    glad_glGetBufferParameteri64v = (PFNGLGETBUFFERPARAMETERI64VPROC) load(userptr, "glGetBufferParameteri64v");
+    glad_glGetInteger64i_v = (PFNGLGETINTEGER64I_VPROC) load(userptr, "glGetInteger64i_v");
+    glad_glGetInteger64v = (PFNGLGETINTEGER64VPROC) load(userptr, "glGetInteger64v");
+    glad_glGetMultisamplefv = (PFNGLGETMULTISAMPLEFVPROC) load(userptr, "glGetMultisamplefv");
+    glad_glGetSynciv = (PFNGLGETSYNCIVPROC) load(userptr, "glGetSynciv");
+    glad_glIsSync = (PFNGLISSYNCPROC) load(userptr, "glIsSync");
+    glad_glMultiDrawElementsBaseVertex = (PFNGLMULTIDRAWELEMENTSBASEVERTEXPROC) load(userptr, "glMultiDrawElementsBaseVertex");
+    glad_glProvokingVertex = (PFNGLPROVOKINGVERTEXPROC) load(userptr, "glProvokingVertex");
+    glad_glSampleMaski = (PFNGLSAMPLEMASKIPROC) load(userptr, "glSampleMaski");
+    glad_glTexImage2DMultisample = (PFNGLTEXIMAGE2DMULTISAMPLEPROC) load(userptr, "glTexImage2DMultisample");
+    glad_glTexImage3DMultisample = (PFNGLTEXIMAGE3DMULTISAMPLEPROC) load(userptr, "glTexImage3DMultisample");
+    glad_glWaitSync = (PFNGLWAITSYNCPROC) load(userptr, "glWaitSync");
+}
+static void glad_gl_load_GL_VERSION_3_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_VERSION_3_3) return;
+    glad_glBindFragDataLocationIndexed = (PFNGLBINDFRAGDATALOCATIONINDEXEDPROC) load(userptr, "glBindFragDataLocationIndexed");
+    glad_glBindSampler = (PFNGLBINDSAMPLERPROC) load(userptr, "glBindSampler");
+    glad_glColorP3ui = (PFNGLCOLORP3UIPROC) load(userptr, "glColorP3ui");
+    glad_glColorP3uiv = (PFNGLCOLORP3UIVPROC) load(userptr, "glColorP3uiv");
+    glad_glColorP4ui = (PFNGLCOLORP4UIPROC) load(userptr, "glColorP4ui");
+    glad_glColorP4uiv = (PFNGLCOLORP4UIVPROC) load(userptr, "glColorP4uiv");
+    glad_glDeleteSamplers = (PFNGLDELETESAMPLERSPROC) load(userptr, "glDeleteSamplers");
+    glad_glGenSamplers = (PFNGLGENSAMPLERSPROC) load(userptr, "glGenSamplers");
+    glad_glGetFragDataIndex = (PFNGLGETFRAGDATAINDEXPROC) load(userptr, "glGetFragDataIndex");
+    glad_glGetQueryObjecti64v = (PFNGLGETQUERYOBJECTI64VPROC) load(userptr, "glGetQueryObjecti64v");
+    glad_glGetQueryObjectui64v = (PFNGLGETQUERYOBJECTUI64VPROC) load(userptr, "glGetQueryObjectui64v");
+    glad_glGetSamplerParameterIiv = (PFNGLGETSAMPLERPARAMETERIIVPROC) load(userptr, "glGetSamplerParameterIiv");
+    glad_glGetSamplerParameterIuiv = (PFNGLGETSAMPLERPARAMETERIUIVPROC) load(userptr, "glGetSamplerParameterIuiv");
+    glad_glGetSamplerParameterfv = (PFNGLGETSAMPLERPARAMETERFVPROC) load(userptr, "glGetSamplerParameterfv");
+    glad_glGetSamplerParameteriv = (PFNGLGETSAMPLERPARAMETERIVPROC) load(userptr, "glGetSamplerParameteriv");
+    glad_glIsSampler = (PFNGLISSAMPLERPROC) load(userptr, "glIsSampler");
+    glad_glMultiTexCoordP1ui = (PFNGLMULTITEXCOORDP1UIPROC) load(userptr, "glMultiTexCoordP1ui");
+    glad_glMultiTexCoordP1uiv = (PFNGLMULTITEXCOORDP1UIVPROC) load(userptr, "glMultiTexCoordP1uiv");
+    glad_glMultiTexCoordP2ui = (PFNGLMULTITEXCOORDP2UIPROC) load(userptr, "glMultiTexCoordP2ui");
+    glad_glMultiTexCoordP2uiv = (PFNGLMULTITEXCOORDP2UIVPROC) load(userptr, "glMultiTexCoordP2uiv");
+    glad_glMultiTexCoordP3ui = (PFNGLMULTITEXCOORDP3UIPROC) load(userptr, "glMultiTexCoordP3ui");
+    glad_glMultiTexCoordP3uiv = (PFNGLMULTITEXCOORDP3UIVPROC) load(userptr, "glMultiTexCoordP3uiv");
+    glad_glMultiTexCoordP4ui = (PFNGLMULTITEXCOORDP4UIPROC) load(userptr, "glMultiTexCoordP4ui");
+    glad_glMultiTexCoordP4uiv = (PFNGLMULTITEXCOORDP4UIVPROC) load(userptr, "glMultiTexCoordP4uiv");
+    glad_glNormalP3ui = (PFNGLNORMALP3UIPROC) load(userptr, "glNormalP3ui");
+    glad_glNormalP3uiv = (PFNGLNORMALP3UIVPROC) load(userptr, "glNormalP3uiv");
+    glad_glQueryCounter = (PFNGLQUERYCOUNTERPROC) load(userptr, "glQueryCounter");
+    glad_glSamplerParameterIiv = (PFNGLSAMPLERPARAMETERIIVPROC) load(userptr, "glSamplerParameterIiv");
+    glad_glSamplerParameterIuiv = (PFNGLSAMPLERPARAMETERIUIVPROC) load(userptr, "glSamplerParameterIuiv");
+    glad_glSamplerParameterf = (PFNGLSAMPLERPARAMETERFPROC) load(userptr, "glSamplerParameterf");
+    glad_glSamplerParameterfv = (PFNGLSAMPLERPARAMETERFVPROC) load(userptr, "glSamplerParameterfv");
+    glad_glSamplerParameteri = (PFNGLSAMPLERPARAMETERIPROC) load(userptr, "glSamplerParameteri");
+    glad_glSamplerParameteriv = (PFNGLSAMPLERPARAMETERIVPROC) load(userptr, "glSamplerParameteriv");
+    glad_glSecondaryColorP3ui = (PFNGLSECONDARYCOLORP3UIPROC) load(userptr, "glSecondaryColorP3ui");
+    glad_glSecondaryColorP3uiv = (PFNGLSECONDARYCOLORP3UIVPROC) load(userptr, "glSecondaryColorP3uiv");
+    glad_glTexCoordP1ui = (PFNGLTEXCOORDP1UIPROC) load(userptr, "glTexCoordP1ui");
+    glad_glTexCoordP1uiv = (PFNGLTEXCOORDP1UIVPROC) load(userptr, "glTexCoordP1uiv");
+    glad_glTexCoordP2ui = (PFNGLTEXCOORDP2UIPROC) load(userptr, "glTexCoordP2ui");
+    glad_glTexCoordP2uiv = (PFNGLTEXCOORDP2UIVPROC) load(userptr, "glTexCoordP2uiv");
+    glad_glTexCoordP3ui = (PFNGLTEXCOORDP3UIPROC) load(userptr, "glTexCoordP3ui");
+    glad_glTexCoordP3uiv = (PFNGLTEXCOORDP3UIVPROC) load(userptr, "glTexCoordP3uiv");
+    glad_glTexCoordP4ui = (PFNGLTEXCOORDP4UIPROC) load(userptr, "glTexCoordP4ui");
+    glad_glTexCoordP4uiv = (PFNGLTEXCOORDP4UIVPROC) load(userptr, "glTexCoordP4uiv");
+    glad_glVertexAttribDivisor = (PFNGLVERTEXATTRIBDIVISORPROC) load(userptr, "glVertexAttribDivisor");
+    glad_glVertexAttribP1ui = (PFNGLVERTEXATTRIBP1UIPROC) load(userptr, "glVertexAttribP1ui");
+    glad_glVertexAttribP1uiv = (PFNGLVERTEXATTRIBP1UIVPROC) load(userptr, "glVertexAttribP1uiv");
+    glad_glVertexAttribP2ui = (PFNGLVERTEXATTRIBP2UIPROC) load(userptr, "glVertexAttribP2ui");
+    glad_glVertexAttribP2uiv = (PFNGLVERTEXATTRIBP2UIVPROC) load(userptr, "glVertexAttribP2uiv");
+    glad_glVertexAttribP3ui = (PFNGLVERTEXATTRIBP3UIPROC) load(userptr, "glVertexAttribP3ui");
+    glad_glVertexAttribP3uiv = (PFNGLVERTEXATTRIBP3UIVPROC) load(userptr, "glVertexAttribP3uiv");
+    glad_glVertexAttribP4ui = (PFNGLVERTEXATTRIBP4UIPROC) load(userptr, "glVertexAttribP4ui");
+    glad_glVertexAttribP4uiv = (PFNGLVERTEXATTRIBP4UIVPROC) load(userptr, "glVertexAttribP4uiv");
+    glad_glVertexP2ui = (PFNGLVERTEXP2UIPROC) load(userptr, "glVertexP2ui");
+    glad_glVertexP2uiv = (PFNGLVERTEXP2UIVPROC) load(userptr, "glVertexP2uiv");
+    glad_glVertexP3ui = (PFNGLVERTEXP3UIPROC) load(userptr, "glVertexP3ui");
+    glad_glVertexP3uiv = (PFNGLVERTEXP3UIVPROC) load(userptr, "glVertexP3uiv");
+    glad_glVertexP4ui = (PFNGLVERTEXP4UIPROC) load(userptr, "glVertexP4ui");
+    glad_glVertexP4uiv = (PFNGLVERTEXP4UIVPROC) load(userptr, "glVertexP4uiv");
+}
+
+
+
+#if defined(GL_ES_VERSION_3_0) || defined(GL_VERSION_3_0)
+#define GLAD_GL_IS_SOME_NEW_VERSION 1
+#else
+#define GLAD_GL_IS_SOME_NEW_VERSION 0
+#endif
+
+static int glad_gl_get_extensions( int version, const char **out_exts, unsigned int *out_num_exts_i, char ***out_exts_i) {
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    if(GLAD_VERSION_MAJOR(version) < 3) {
+#else
+    (void) version;
+    (void) out_num_exts_i;
+    (void) out_exts_i;
+#endif
+        if (glad_glGetString == NULL) {
+            return 0;
+        }
+        *out_exts = (const char *)glad_glGetString(GL_EXTENSIONS);
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    } else {
+        unsigned int index = 0;
+        unsigned int num_exts_i = 0;
+        char **exts_i = NULL;
+        if (glad_glGetStringi == NULL || glad_glGetIntegerv == NULL) {
+            return 0;
+        }
+        glad_glGetIntegerv(GL_NUM_EXTENSIONS, (int*) &num_exts_i);
+        if (num_exts_i > 0) {
+            exts_i = (char **) malloc(num_exts_i * (sizeof *exts_i));
+        }
+        if (exts_i == NULL) {
+            return 0;
+        }
+        for(index = 0; index < num_exts_i; index++) {
+            const char *gl_str_tmp = (const char*) glad_glGetStringi(GL_EXTENSIONS, index);
+            size_t len = strlen(gl_str_tmp) + 1;
+
+            char *local_str = (char*) malloc(len * sizeof(char));
+            if(local_str != NULL) {
+                memcpy(local_str, gl_str_tmp, len * sizeof(char));
+            }
+
+            exts_i[index] = local_str;
+        }
+
+        *out_num_exts_i = num_exts_i;
+        *out_exts_i = exts_i;
+    }
+#endif
+    return 1;
+}
+static void glad_gl_free_extensions(char **exts_i, unsigned int num_exts_i) {
+    if (exts_i != NULL) {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            free((void *) (exts_i[index]));
+        }
+        free((void *)exts_i);
+        exts_i = NULL;
+    }
+}
+static int glad_gl_has_extension(int version, const char *exts, unsigned int num_exts_i, char **exts_i, const char *ext) {
+    if(GLAD_VERSION_MAJOR(version) < 3 || !GLAD_GL_IS_SOME_NEW_VERSION) {
+        const char *extensions;
+        const char *loc;
+        const char *terminator;
+        extensions = exts;
+        if(extensions == NULL || ext == NULL) {
+            return 0;
+        }
+        while(1) {
+            loc = strstr(extensions, ext);
+            if(loc == NULL) {
+                return 0;
+            }
+            terminator = loc + strlen(ext);
+            if((loc == extensions || *(loc - 1) == ' ') &&
+                (*terminator == ' ' || *terminator == '\0')) {
+                return 1;
+            }
+            extensions = terminator;
+        }
+    } else {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            const char *e = exts_i[index];
+            if(strcmp(e, ext) == 0) {
+                return 1;
+            }
+        }
+    }
+    return 0;
+}
+
+static GLADapiproc glad_gl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_gl_find_extensions_gl( int version) {
+    const char *exts = NULL;
+    unsigned int num_exts_i = 0;
+    char **exts_i = NULL;
+    if (!glad_gl_get_extensions(version, &exts, &num_exts_i, &exts_i)) return 0;
+
+    (void) glad_gl_has_extension;
+
+    glad_gl_free_extensions(exts_i, num_exts_i);
+
+    return 1;
+}
+
+static int glad_gl_find_core_gl(void) {
+    int i, major, minor;
+    const char* version;
+    const char* prefixes[] = {
+        "OpenGL ES-CM ",
+        "OpenGL ES-CL ",
+        "OpenGL ES ",
+        NULL
+    };
+    version = (const char*) glad_glGetString(GL_VERSION);
+    if (!version) return 0;
+    for (i = 0;  prefixes[i];  i++) {
+        const size_t length = strlen(prefixes[i]);
+        if (strncmp(version, prefixes[i], length) == 0) {
+            version += length;
+            break;
+        }
+    }
+
+    GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+
+    GLAD_GL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_GL_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_GL_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_GL_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_GL_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    GLAD_GL_VERSION_1_5 = (major == 1 && minor >= 5) || major > 1;
+    GLAD_GL_VERSION_2_0 = (major == 2 && minor >= 0) || major > 2;
+    GLAD_GL_VERSION_2_1 = (major == 2 && minor >= 1) || major > 2;
+    GLAD_GL_VERSION_3_0 = (major == 3 && minor >= 0) || major > 3;
+    GLAD_GL_VERSION_3_1 = (major == 3 && minor >= 1) || major > 3;
+    GLAD_GL_VERSION_3_2 = (major == 3 && minor >= 2) || major > 3;
+    GLAD_GL_VERSION_3_3 = (major == 3 && minor >= 3) || major > 3;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLUserPtr( GLADuserptrloadfunc load, void *userptr) {
+    int version;
+
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    if(glad_glGetString == NULL) return 0;
+    if(glad_glGetString(GL_VERSION) == NULL) return 0;
+    version = glad_gl_find_core_gl();
+
+    glad_gl_load_GL_VERSION_1_0(load, userptr);
+    glad_gl_load_GL_VERSION_1_1(load, userptr);
+    glad_gl_load_GL_VERSION_1_2(load, userptr);
+    glad_gl_load_GL_VERSION_1_3(load, userptr);
+    glad_gl_load_GL_VERSION_1_4(load, userptr);
+    glad_gl_load_GL_VERSION_1_5(load, userptr);
+    glad_gl_load_GL_VERSION_2_0(load, userptr);
+    glad_gl_load_GL_VERSION_2_1(load, userptr);
+    glad_gl_load_GL_VERSION_3_0(load, userptr);
+    glad_gl_load_GL_VERSION_3_1(load, userptr);
+    glad_gl_load_GL_VERSION_3_2(load, userptr);
+    glad_gl_load_GL_VERSION_3_3(load, userptr);
+
+    if (!glad_gl_find_extensions_gl(version)) return 0;
+
+
+
+    return version;
+}
+
+
+int gladLoadGL( GLADloadfunc load) {
+    return gladLoadGLUserPtr( glad_gl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/src/gles2.c glmark2-2021.02/src/glad/src/gles2.c
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/src/gles2.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/src/gles2.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,495 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/gles2.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GL_ES_VERSION_2_0 = 0;
+int GLAD_GL_OES_mapbuffer = 0;
+int GLAD_GL_OES_required_internalformat = 0;
+
+
+
+PFNGLACTIVETEXTUREPROC glad_glActiveTexture = NULL;
+PFNGLATTACHSHADERPROC glad_glAttachShader = NULL;
+PFNGLBINDATTRIBLOCATIONPROC glad_glBindAttribLocation = NULL;
+PFNGLBINDBUFFERPROC glad_glBindBuffer = NULL;
+PFNGLBINDFRAMEBUFFERPROC glad_glBindFramebuffer = NULL;
+PFNGLBINDRENDERBUFFERPROC glad_glBindRenderbuffer = NULL;
+PFNGLBINDTEXTUREPROC glad_glBindTexture = NULL;
+PFNGLBLENDCOLORPROC glad_glBlendColor = NULL;
+PFNGLBLENDEQUATIONPROC glad_glBlendEquation = NULL;
+PFNGLBLENDEQUATIONSEPARATEPROC glad_glBlendEquationSeparate = NULL;
+PFNGLBLENDFUNCPROC glad_glBlendFunc = NULL;
+PFNGLBLENDFUNCSEPARATEPROC glad_glBlendFuncSeparate = NULL;
+PFNGLBUFFERDATAPROC glad_glBufferData = NULL;
+PFNGLBUFFERSUBDATAPROC glad_glBufferSubData = NULL;
+PFNGLCHECKFRAMEBUFFERSTATUSPROC glad_glCheckFramebufferStatus = NULL;
+PFNGLCLEARPROC glad_glClear = NULL;
+PFNGLCLEARCOLORPROC glad_glClearColor = NULL;
+PFNGLCLEARDEPTHFPROC glad_glClearDepthf = NULL;
+PFNGLCLEARSTENCILPROC glad_glClearStencil = NULL;
+PFNGLCOLORMASKPROC glad_glColorMask = NULL;
+PFNGLCOMPILESHADERPROC glad_glCompileShader = NULL;
+PFNGLCOMPRESSEDTEXIMAGE2DPROC glad_glCompressedTexImage2D = NULL;
+PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC glad_glCompressedTexSubImage2D = NULL;
+PFNGLCOPYTEXIMAGE2DPROC glad_glCopyTexImage2D = NULL;
+PFNGLCOPYTEXSUBIMAGE2DPROC glad_glCopyTexSubImage2D = NULL;
+PFNGLCREATEPROGRAMPROC glad_glCreateProgram = NULL;
+PFNGLCREATESHADERPROC glad_glCreateShader = NULL;
+PFNGLCULLFACEPROC glad_glCullFace = NULL;
+PFNGLDELETEBUFFERSPROC glad_glDeleteBuffers = NULL;
+PFNGLDELETEFRAMEBUFFERSPROC glad_glDeleteFramebuffers = NULL;
+PFNGLDELETEPROGRAMPROC glad_glDeleteProgram = NULL;
+PFNGLDELETERENDERBUFFERSPROC glad_glDeleteRenderbuffers = NULL;
+PFNGLDELETESHADERPROC glad_glDeleteShader = NULL;
+PFNGLDELETETEXTURESPROC glad_glDeleteTextures = NULL;
+PFNGLDEPTHFUNCPROC glad_glDepthFunc = NULL;
+PFNGLDEPTHMASKPROC glad_glDepthMask = NULL;
+PFNGLDEPTHRANGEFPROC glad_glDepthRangef = NULL;
+PFNGLDETACHSHADERPROC glad_glDetachShader = NULL;
+PFNGLDISABLEPROC glad_glDisable = NULL;
+PFNGLDISABLEVERTEXATTRIBARRAYPROC glad_glDisableVertexAttribArray = NULL;
+PFNGLDRAWARRAYSPROC glad_glDrawArrays = NULL;
+PFNGLDRAWELEMENTSPROC glad_glDrawElements = NULL;
+PFNGLENABLEPROC glad_glEnable = NULL;
+PFNGLENABLEVERTEXATTRIBARRAYPROC glad_glEnableVertexAttribArray = NULL;
+PFNGLFINISHPROC glad_glFinish = NULL;
+PFNGLFLUSHPROC glad_glFlush = NULL;
+PFNGLFRAMEBUFFERRENDERBUFFERPROC glad_glFramebufferRenderbuffer = NULL;
+PFNGLFRAMEBUFFERTEXTURE2DPROC glad_glFramebufferTexture2D = NULL;
+PFNGLFRONTFACEPROC glad_glFrontFace = NULL;
+PFNGLGENBUFFERSPROC glad_glGenBuffers = NULL;
+PFNGLGENFRAMEBUFFERSPROC glad_glGenFramebuffers = NULL;
+PFNGLGENRENDERBUFFERSPROC glad_glGenRenderbuffers = NULL;
+PFNGLGENTEXTURESPROC glad_glGenTextures = NULL;
+PFNGLGENERATEMIPMAPPROC glad_glGenerateMipmap = NULL;
+PFNGLGETACTIVEATTRIBPROC glad_glGetActiveAttrib = NULL;
+PFNGLGETACTIVEUNIFORMPROC glad_glGetActiveUniform = NULL;
+PFNGLGETATTACHEDSHADERSPROC glad_glGetAttachedShaders = NULL;
+PFNGLGETATTRIBLOCATIONPROC glad_glGetAttribLocation = NULL;
+PFNGLGETBOOLEANVPROC glad_glGetBooleanv = NULL;
+PFNGLGETBUFFERPARAMETERIVPROC glad_glGetBufferParameteriv = NULL;
+PFNGLGETBUFFERPOINTERVPROC glad_glGetBufferPointerv = NULL;
+PFNGLGETBUFFERPOINTERVOESPROC glad_glGetBufferPointervOES = NULL;
+PFNGLGETERRORPROC glad_glGetError = NULL;
+PFNGLGETFLOATVPROC glad_glGetFloatv = NULL;
+PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC glad_glGetFramebufferAttachmentParameteriv = NULL;
+PFNGLGETINTEGERVPROC glad_glGetIntegerv = NULL;
+PFNGLGETPROGRAMINFOLOGPROC glad_glGetProgramInfoLog = NULL;
+PFNGLGETPROGRAMIVPROC glad_glGetProgramiv = NULL;
+PFNGLGETRENDERBUFFERPARAMETERIVPROC glad_glGetRenderbufferParameteriv = NULL;
+PFNGLGETSHADERINFOLOGPROC glad_glGetShaderInfoLog = NULL;
+PFNGLGETSHADERPRECISIONFORMATPROC glad_glGetShaderPrecisionFormat = NULL;
+PFNGLGETSHADERSOURCEPROC glad_glGetShaderSource = NULL;
+PFNGLGETSHADERIVPROC glad_glGetShaderiv = NULL;
+PFNGLGETSTRINGPROC glad_glGetString = NULL;
+PFNGLGETTEXPARAMETERFVPROC glad_glGetTexParameterfv = NULL;
+PFNGLGETTEXPARAMETERIVPROC glad_glGetTexParameteriv = NULL;
+PFNGLGETUNIFORMLOCATIONPROC glad_glGetUniformLocation = NULL;
+PFNGLGETUNIFORMFVPROC glad_glGetUniformfv = NULL;
+PFNGLGETUNIFORMIVPROC glad_glGetUniformiv = NULL;
+PFNGLGETVERTEXATTRIBPOINTERVPROC glad_glGetVertexAttribPointerv = NULL;
+PFNGLGETVERTEXATTRIBFVPROC glad_glGetVertexAttribfv = NULL;
+PFNGLGETVERTEXATTRIBIVPROC glad_glGetVertexAttribiv = NULL;
+PFNGLHINTPROC glad_glHint = NULL;
+PFNGLISBUFFERPROC glad_glIsBuffer = NULL;
+PFNGLISENABLEDPROC glad_glIsEnabled = NULL;
+PFNGLISFRAMEBUFFERPROC glad_glIsFramebuffer = NULL;
+PFNGLISPROGRAMPROC glad_glIsProgram = NULL;
+PFNGLISRENDERBUFFERPROC glad_glIsRenderbuffer = NULL;
+PFNGLISSHADERPROC glad_glIsShader = NULL;
+PFNGLISTEXTUREPROC glad_glIsTexture = NULL;
+PFNGLLINEWIDTHPROC glad_glLineWidth = NULL;
+PFNGLLINKPROGRAMPROC glad_glLinkProgram = NULL;
+PFNGLMAPBUFFERPROC glad_glMapBuffer = NULL;
+PFNGLMAPBUFFEROESPROC glad_glMapBufferOES = NULL;
+PFNGLPIXELSTOREIPROC glad_glPixelStorei = NULL;
+PFNGLPOLYGONOFFSETPROC glad_glPolygonOffset = NULL;
+PFNGLREADPIXELSPROC glad_glReadPixels = NULL;
+PFNGLRELEASESHADERCOMPILERPROC glad_glReleaseShaderCompiler = NULL;
+PFNGLRENDERBUFFERSTORAGEPROC glad_glRenderbufferStorage = NULL;
+PFNGLSAMPLECOVERAGEPROC glad_glSampleCoverage = NULL;
+PFNGLSCISSORPROC glad_glScissor = NULL;
+PFNGLSHADERBINARYPROC glad_glShaderBinary = NULL;
+PFNGLSHADERSOURCEPROC glad_glShaderSource = NULL;
+PFNGLSTENCILFUNCPROC glad_glStencilFunc = NULL;
+PFNGLSTENCILFUNCSEPARATEPROC glad_glStencilFuncSeparate = NULL;
+PFNGLSTENCILMASKPROC glad_glStencilMask = NULL;
+PFNGLSTENCILMASKSEPARATEPROC glad_glStencilMaskSeparate = NULL;
+PFNGLSTENCILOPPROC glad_glStencilOp = NULL;
+PFNGLSTENCILOPSEPARATEPROC glad_glStencilOpSeparate = NULL;
+PFNGLTEXIMAGE2DPROC glad_glTexImage2D = NULL;
+PFNGLTEXPARAMETERFPROC glad_glTexParameterf = NULL;
+PFNGLTEXPARAMETERFVPROC glad_glTexParameterfv = NULL;
+PFNGLTEXPARAMETERIPROC glad_glTexParameteri = NULL;
+PFNGLTEXPARAMETERIVPROC glad_glTexParameteriv = NULL;
+PFNGLTEXSUBIMAGE2DPROC glad_glTexSubImage2D = NULL;
+PFNGLUNIFORM1FPROC glad_glUniform1f = NULL;
+PFNGLUNIFORM1FVPROC glad_glUniform1fv = NULL;
+PFNGLUNIFORM1IPROC glad_glUniform1i = NULL;
+PFNGLUNIFORM1IVPROC glad_glUniform1iv = NULL;
+PFNGLUNIFORM2FPROC glad_glUniform2f = NULL;
+PFNGLUNIFORM2FVPROC glad_glUniform2fv = NULL;
+PFNGLUNIFORM2IPROC glad_glUniform2i = NULL;
+PFNGLUNIFORM2IVPROC glad_glUniform2iv = NULL;
+PFNGLUNIFORM3FPROC glad_glUniform3f = NULL;
+PFNGLUNIFORM3FVPROC glad_glUniform3fv = NULL;
+PFNGLUNIFORM3IPROC glad_glUniform3i = NULL;
+PFNGLUNIFORM3IVPROC glad_glUniform3iv = NULL;
+PFNGLUNIFORM4FPROC glad_glUniform4f = NULL;
+PFNGLUNIFORM4FVPROC glad_glUniform4fv = NULL;
+PFNGLUNIFORM4IPROC glad_glUniform4i = NULL;
+PFNGLUNIFORM4IVPROC glad_glUniform4iv = NULL;
+PFNGLUNIFORMMATRIX2FVPROC glad_glUniformMatrix2fv = NULL;
+PFNGLUNIFORMMATRIX3FVPROC glad_glUniformMatrix3fv = NULL;
+PFNGLUNIFORMMATRIX4FVPROC glad_glUniformMatrix4fv = NULL;
+PFNGLUNMAPBUFFERPROC glad_glUnmapBuffer = NULL;
+PFNGLUNMAPBUFFEROESPROC glad_glUnmapBufferOES = NULL;
+PFNGLUSEPROGRAMPROC glad_glUseProgram = NULL;
+PFNGLVALIDATEPROGRAMPROC glad_glValidateProgram = NULL;
+PFNGLVERTEXATTRIB1FPROC glad_glVertexAttrib1f = NULL;
+PFNGLVERTEXATTRIB1FVPROC glad_glVertexAttrib1fv = NULL;
+PFNGLVERTEXATTRIB2FPROC glad_glVertexAttrib2f = NULL;
+PFNGLVERTEXATTRIB2FVPROC glad_glVertexAttrib2fv = NULL;
+PFNGLVERTEXATTRIB3FPROC glad_glVertexAttrib3f = NULL;
+PFNGLVERTEXATTRIB3FVPROC glad_glVertexAttrib3fv = NULL;
+PFNGLVERTEXATTRIB4FPROC glad_glVertexAttrib4f = NULL;
+PFNGLVERTEXATTRIB4FVPROC glad_glVertexAttrib4fv = NULL;
+PFNGLVERTEXATTRIBPOINTERPROC glad_glVertexAttribPointer = NULL;
+PFNGLVIEWPORTPROC glad_glViewport = NULL;
+
+
+static void glad_gl_load_GL_ES_VERSION_2_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_ES_VERSION_2_0) return;
+    glad_glActiveTexture = (PFNGLACTIVETEXTUREPROC) load(userptr, "glActiveTexture");
+    glad_glAttachShader = (PFNGLATTACHSHADERPROC) load(userptr, "glAttachShader");
+    glad_glBindAttribLocation = (PFNGLBINDATTRIBLOCATIONPROC) load(userptr, "glBindAttribLocation");
+    glad_glBindBuffer = (PFNGLBINDBUFFERPROC) load(userptr, "glBindBuffer");
+    glad_glBindFramebuffer = (PFNGLBINDFRAMEBUFFERPROC) load(userptr, "glBindFramebuffer");
+    glad_glBindRenderbuffer = (PFNGLBINDRENDERBUFFERPROC) load(userptr, "glBindRenderbuffer");
+    glad_glBindTexture = (PFNGLBINDTEXTUREPROC) load(userptr, "glBindTexture");
+    glad_glBlendColor = (PFNGLBLENDCOLORPROC) load(userptr, "glBlendColor");
+    glad_glBlendEquation = (PFNGLBLENDEQUATIONPROC) load(userptr, "glBlendEquation");
+    glad_glBlendEquationSeparate = (PFNGLBLENDEQUATIONSEPARATEPROC) load(userptr, "glBlendEquationSeparate");
+    glad_glBlendFunc = (PFNGLBLENDFUNCPROC) load(userptr, "glBlendFunc");
+    glad_glBlendFuncSeparate = (PFNGLBLENDFUNCSEPARATEPROC) load(userptr, "glBlendFuncSeparate");
+    glad_glBufferData = (PFNGLBUFFERDATAPROC) load(userptr, "glBufferData");
+    glad_glBufferSubData = (PFNGLBUFFERSUBDATAPROC) load(userptr, "glBufferSubData");
+    glad_glCheckFramebufferStatus = (PFNGLCHECKFRAMEBUFFERSTATUSPROC) load(userptr, "glCheckFramebufferStatus");
+    glad_glClear = (PFNGLCLEARPROC) load(userptr, "glClear");
+    glad_glClearColor = (PFNGLCLEARCOLORPROC) load(userptr, "glClearColor");
+    glad_glClearDepthf = (PFNGLCLEARDEPTHFPROC) load(userptr, "glClearDepthf");
+    glad_glClearStencil = (PFNGLCLEARSTENCILPROC) load(userptr, "glClearStencil");
+    glad_glColorMask = (PFNGLCOLORMASKPROC) load(userptr, "glColorMask");
+    glad_glCompileShader = (PFNGLCOMPILESHADERPROC) load(userptr, "glCompileShader");
+    glad_glCompressedTexImage2D = (PFNGLCOMPRESSEDTEXIMAGE2DPROC) load(userptr, "glCompressedTexImage2D");
+    glad_glCompressedTexSubImage2D = (PFNGLCOMPRESSEDTEXSUBIMAGE2DPROC) load(userptr, "glCompressedTexSubImage2D");
+    glad_glCopyTexImage2D = (PFNGLCOPYTEXIMAGE2DPROC) load(userptr, "glCopyTexImage2D");
+    glad_glCopyTexSubImage2D = (PFNGLCOPYTEXSUBIMAGE2DPROC) load(userptr, "glCopyTexSubImage2D");
+    glad_glCreateProgram = (PFNGLCREATEPROGRAMPROC) load(userptr, "glCreateProgram");
+    glad_glCreateShader = (PFNGLCREATESHADERPROC) load(userptr, "glCreateShader");
+    glad_glCullFace = (PFNGLCULLFACEPROC) load(userptr, "glCullFace");
+    glad_glDeleteBuffers = (PFNGLDELETEBUFFERSPROC) load(userptr, "glDeleteBuffers");
+    glad_glDeleteFramebuffers = (PFNGLDELETEFRAMEBUFFERSPROC) load(userptr, "glDeleteFramebuffers");
+    glad_glDeleteProgram = (PFNGLDELETEPROGRAMPROC) load(userptr, "glDeleteProgram");
+    glad_glDeleteRenderbuffers = (PFNGLDELETERENDERBUFFERSPROC) load(userptr, "glDeleteRenderbuffers");
+    glad_glDeleteShader = (PFNGLDELETESHADERPROC) load(userptr, "glDeleteShader");
+    glad_glDeleteTextures = (PFNGLDELETETEXTURESPROC) load(userptr, "glDeleteTextures");
+    glad_glDepthFunc = (PFNGLDEPTHFUNCPROC) load(userptr, "glDepthFunc");
+    glad_glDepthMask = (PFNGLDEPTHMASKPROC) load(userptr, "glDepthMask");
+    glad_glDepthRangef = (PFNGLDEPTHRANGEFPROC) load(userptr, "glDepthRangef");
+    glad_glDetachShader = (PFNGLDETACHSHADERPROC) load(userptr, "glDetachShader");
+    glad_glDisable = (PFNGLDISABLEPROC) load(userptr, "glDisable");
+    glad_glDisableVertexAttribArray = (PFNGLDISABLEVERTEXATTRIBARRAYPROC) load(userptr, "glDisableVertexAttribArray");
+    glad_glDrawArrays = (PFNGLDRAWARRAYSPROC) load(userptr, "glDrawArrays");
+    glad_glDrawElements = (PFNGLDRAWELEMENTSPROC) load(userptr, "glDrawElements");
+    glad_glEnable = (PFNGLENABLEPROC) load(userptr, "glEnable");
+    glad_glEnableVertexAttribArray = (PFNGLENABLEVERTEXATTRIBARRAYPROC) load(userptr, "glEnableVertexAttribArray");
+    glad_glFinish = (PFNGLFINISHPROC) load(userptr, "glFinish");
+    glad_glFlush = (PFNGLFLUSHPROC) load(userptr, "glFlush");
+    glad_glFramebufferRenderbuffer = (PFNGLFRAMEBUFFERRENDERBUFFERPROC) load(userptr, "glFramebufferRenderbuffer");
+    glad_glFramebufferTexture2D = (PFNGLFRAMEBUFFERTEXTURE2DPROC) load(userptr, "glFramebufferTexture2D");
+    glad_glFrontFace = (PFNGLFRONTFACEPROC) load(userptr, "glFrontFace");
+    glad_glGenBuffers = (PFNGLGENBUFFERSPROC) load(userptr, "glGenBuffers");
+    glad_glGenFramebuffers = (PFNGLGENFRAMEBUFFERSPROC) load(userptr, "glGenFramebuffers");
+    glad_glGenRenderbuffers = (PFNGLGENRENDERBUFFERSPROC) load(userptr, "glGenRenderbuffers");
+    glad_glGenTextures = (PFNGLGENTEXTURESPROC) load(userptr, "glGenTextures");
+    glad_glGenerateMipmap = (PFNGLGENERATEMIPMAPPROC) load(userptr, "glGenerateMipmap");
+    glad_glGetActiveAttrib = (PFNGLGETACTIVEATTRIBPROC) load(userptr, "glGetActiveAttrib");
+    glad_glGetActiveUniform = (PFNGLGETACTIVEUNIFORMPROC) load(userptr, "glGetActiveUniform");
+    glad_glGetAttachedShaders = (PFNGLGETATTACHEDSHADERSPROC) load(userptr, "glGetAttachedShaders");
+    glad_glGetAttribLocation = (PFNGLGETATTRIBLOCATIONPROC) load(userptr, "glGetAttribLocation");
+    glad_glGetBooleanv = (PFNGLGETBOOLEANVPROC) load(userptr, "glGetBooleanv");
+    glad_glGetBufferParameteriv = (PFNGLGETBUFFERPARAMETERIVPROC) load(userptr, "glGetBufferParameteriv");
+    glad_glGetError = (PFNGLGETERRORPROC) load(userptr, "glGetError");
+    glad_glGetFloatv = (PFNGLGETFLOATVPROC) load(userptr, "glGetFloatv");
+    glad_glGetFramebufferAttachmentParameteriv = (PFNGLGETFRAMEBUFFERATTACHMENTPARAMETERIVPROC) load(userptr, "glGetFramebufferAttachmentParameteriv");
+    glad_glGetIntegerv = (PFNGLGETINTEGERVPROC) load(userptr, "glGetIntegerv");
+    glad_glGetProgramInfoLog = (PFNGLGETPROGRAMINFOLOGPROC) load(userptr, "glGetProgramInfoLog");
+    glad_glGetProgramiv = (PFNGLGETPROGRAMIVPROC) load(userptr, "glGetProgramiv");
+    glad_glGetRenderbufferParameteriv = (PFNGLGETRENDERBUFFERPARAMETERIVPROC) load(userptr, "glGetRenderbufferParameteriv");
+    glad_glGetShaderInfoLog = (PFNGLGETSHADERINFOLOGPROC) load(userptr, "glGetShaderInfoLog");
+    glad_glGetShaderPrecisionFormat = (PFNGLGETSHADERPRECISIONFORMATPROC) load(userptr, "glGetShaderPrecisionFormat");
+    glad_glGetShaderSource = (PFNGLGETSHADERSOURCEPROC) load(userptr, "glGetShaderSource");
+    glad_glGetShaderiv = (PFNGLGETSHADERIVPROC) load(userptr, "glGetShaderiv");
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    glad_glGetTexParameterfv = (PFNGLGETTEXPARAMETERFVPROC) load(userptr, "glGetTexParameterfv");
+    glad_glGetTexParameteriv = (PFNGLGETTEXPARAMETERIVPROC) load(userptr, "glGetTexParameteriv");
+    glad_glGetUniformLocation = (PFNGLGETUNIFORMLOCATIONPROC) load(userptr, "glGetUniformLocation");
+    glad_glGetUniformfv = (PFNGLGETUNIFORMFVPROC) load(userptr, "glGetUniformfv");
+    glad_glGetUniformiv = (PFNGLGETUNIFORMIVPROC) load(userptr, "glGetUniformiv");
+    glad_glGetVertexAttribPointerv = (PFNGLGETVERTEXATTRIBPOINTERVPROC) load(userptr, "glGetVertexAttribPointerv");
+    glad_glGetVertexAttribfv = (PFNGLGETVERTEXATTRIBFVPROC) load(userptr, "glGetVertexAttribfv");
+    glad_glGetVertexAttribiv = (PFNGLGETVERTEXATTRIBIVPROC) load(userptr, "glGetVertexAttribiv");
+    glad_glHint = (PFNGLHINTPROC) load(userptr, "glHint");
+    glad_glIsBuffer = (PFNGLISBUFFERPROC) load(userptr, "glIsBuffer");
+    glad_glIsEnabled = (PFNGLISENABLEDPROC) load(userptr, "glIsEnabled");
+    glad_glIsFramebuffer = (PFNGLISFRAMEBUFFERPROC) load(userptr, "glIsFramebuffer");
+    glad_glIsProgram = (PFNGLISPROGRAMPROC) load(userptr, "glIsProgram");
+    glad_glIsRenderbuffer = (PFNGLISRENDERBUFFERPROC) load(userptr, "glIsRenderbuffer");
+    glad_glIsShader = (PFNGLISSHADERPROC) load(userptr, "glIsShader");
+    glad_glIsTexture = (PFNGLISTEXTUREPROC) load(userptr, "glIsTexture");
+    glad_glLineWidth = (PFNGLLINEWIDTHPROC) load(userptr, "glLineWidth");
+    glad_glLinkProgram = (PFNGLLINKPROGRAMPROC) load(userptr, "glLinkProgram");
+    glad_glPixelStorei = (PFNGLPIXELSTOREIPROC) load(userptr, "glPixelStorei");
+    glad_glPolygonOffset = (PFNGLPOLYGONOFFSETPROC) load(userptr, "glPolygonOffset");
+    glad_glReadPixels = (PFNGLREADPIXELSPROC) load(userptr, "glReadPixels");
+    glad_glReleaseShaderCompiler = (PFNGLRELEASESHADERCOMPILERPROC) load(userptr, "glReleaseShaderCompiler");
+    glad_glRenderbufferStorage = (PFNGLRENDERBUFFERSTORAGEPROC) load(userptr, "glRenderbufferStorage");
+    glad_glSampleCoverage = (PFNGLSAMPLECOVERAGEPROC) load(userptr, "glSampleCoverage");
+    glad_glScissor = (PFNGLSCISSORPROC) load(userptr, "glScissor");
+    glad_glShaderBinary = (PFNGLSHADERBINARYPROC) load(userptr, "glShaderBinary");
+    glad_glShaderSource = (PFNGLSHADERSOURCEPROC) load(userptr, "glShaderSource");
+    glad_glStencilFunc = (PFNGLSTENCILFUNCPROC) load(userptr, "glStencilFunc");
+    glad_glStencilFuncSeparate = (PFNGLSTENCILFUNCSEPARATEPROC) load(userptr, "glStencilFuncSeparate");
+    glad_glStencilMask = (PFNGLSTENCILMASKPROC) load(userptr, "glStencilMask");
+    glad_glStencilMaskSeparate = (PFNGLSTENCILMASKSEPARATEPROC) load(userptr, "glStencilMaskSeparate");
+    glad_glStencilOp = (PFNGLSTENCILOPPROC) load(userptr, "glStencilOp");
+    glad_glStencilOpSeparate = (PFNGLSTENCILOPSEPARATEPROC) load(userptr, "glStencilOpSeparate");
+    glad_glTexImage2D = (PFNGLTEXIMAGE2DPROC) load(userptr, "glTexImage2D");
+    glad_glTexParameterf = (PFNGLTEXPARAMETERFPROC) load(userptr, "glTexParameterf");
+    glad_glTexParameterfv = (PFNGLTEXPARAMETERFVPROC) load(userptr, "glTexParameterfv");
+    glad_glTexParameteri = (PFNGLTEXPARAMETERIPROC) load(userptr, "glTexParameteri");
+    glad_glTexParameteriv = (PFNGLTEXPARAMETERIVPROC) load(userptr, "glTexParameteriv");
+    glad_glTexSubImage2D = (PFNGLTEXSUBIMAGE2DPROC) load(userptr, "glTexSubImage2D");
+    glad_glUniform1f = (PFNGLUNIFORM1FPROC) load(userptr, "glUniform1f");
+    glad_glUniform1fv = (PFNGLUNIFORM1FVPROC) load(userptr, "glUniform1fv");
+    glad_glUniform1i = (PFNGLUNIFORM1IPROC) load(userptr, "glUniform1i");
+    glad_glUniform1iv = (PFNGLUNIFORM1IVPROC) load(userptr, "glUniform1iv");
+    glad_glUniform2f = (PFNGLUNIFORM2FPROC) load(userptr, "glUniform2f");
+    glad_glUniform2fv = (PFNGLUNIFORM2FVPROC) load(userptr, "glUniform2fv");
+    glad_glUniform2i = (PFNGLUNIFORM2IPROC) load(userptr, "glUniform2i");
+    glad_glUniform2iv = (PFNGLUNIFORM2IVPROC) load(userptr, "glUniform2iv");
+    glad_glUniform3f = (PFNGLUNIFORM3FPROC) load(userptr, "glUniform3f");
+    glad_glUniform3fv = (PFNGLUNIFORM3FVPROC) load(userptr, "glUniform3fv");
+    glad_glUniform3i = (PFNGLUNIFORM3IPROC) load(userptr, "glUniform3i");
+    glad_glUniform3iv = (PFNGLUNIFORM3IVPROC) load(userptr, "glUniform3iv");
+    glad_glUniform4f = (PFNGLUNIFORM4FPROC) load(userptr, "glUniform4f");
+    glad_glUniform4fv = (PFNGLUNIFORM4FVPROC) load(userptr, "glUniform4fv");
+    glad_glUniform4i = (PFNGLUNIFORM4IPROC) load(userptr, "glUniform4i");
+    glad_glUniform4iv = (PFNGLUNIFORM4IVPROC) load(userptr, "glUniform4iv");
+    glad_glUniformMatrix2fv = (PFNGLUNIFORMMATRIX2FVPROC) load(userptr, "glUniformMatrix2fv");
+    glad_glUniformMatrix3fv = (PFNGLUNIFORMMATRIX3FVPROC) load(userptr, "glUniformMatrix3fv");
+    glad_glUniformMatrix4fv = (PFNGLUNIFORMMATRIX4FVPROC) load(userptr, "glUniformMatrix4fv");
+    glad_glUseProgram = (PFNGLUSEPROGRAMPROC) load(userptr, "glUseProgram");
+    glad_glValidateProgram = (PFNGLVALIDATEPROGRAMPROC) load(userptr, "glValidateProgram");
+    glad_glVertexAttrib1f = (PFNGLVERTEXATTRIB1FPROC) load(userptr, "glVertexAttrib1f");
+    glad_glVertexAttrib1fv = (PFNGLVERTEXATTRIB1FVPROC) load(userptr, "glVertexAttrib1fv");
+    glad_glVertexAttrib2f = (PFNGLVERTEXATTRIB2FPROC) load(userptr, "glVertexAttrib2f");
+    glad_glVertexAttrib2fv = (PFNGLVERTEXATTRIB2FVPROC) load(userptr, "glVertexAttrib2fv");
+    glad_glVertexAttrib3f = (PFNGLVERTEXATTRIB3FPROC) load(userptr, "glVertexAttrib3f");
+    glad_glVertexAttrib3fv = (PFNGLVERTEXATTRIB3FVPROC) load(userptr, "glVertexAttrib3fv");
+    glad_glVertexAttrib4f = (PFNGLVERTEXATTRIB4FPROC) load(userptr, "glVertexAttrib4f");
+    glad_glVertexAttrib4fv = (PFNGLVERTEXATTRIB4FVPROC) load(userptr, "glVertexAttrib4fv");
+    glad_glVertexAttribPointer = (PFNGLVERTEXATTRIBPOINTERPROC) load(userptr, "glVertexAttribPointer");
+    glad_glViewport = (PFNGLVIEWPORTPROC) load(userptr, "glViewport");
+}
+static void glad_gl_load_GL_OES_mapbuffer( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GL_OES_mapbuffer) return;
+    glad_glGetBufferPointerv = (PFNGLGETBUFFERPOINTERVPROC) load(userptr, "glGetBufferPointerv");
+    glad_glGetBufferPointervOES = (PFNGLGETBUFFERPOINTERVOESPROC) load(userptr, "glGetBufferPointervOES");
+    glad_glMapBuffer = (PFNGLMAPBUFFERPROC) load(userptr, "glMapBuffer");
+    glad_glMapBufferOES = (PFNGLMAPBUFFEROESPROC) load(userptr, "glMapBufferOES");
+    glad_glUnmapBuffer = (PFNGLUNMAPBUFFERPROC) load(userptr, "glUnmapBuffer");
+    glad_glUnmapBufferOES = (PFNGLUNMAPBUFFEROESPROC) load(userptr, "glUnmapBufferOES");
+}
+
+
+
+#if defined(GL_ES_VERSION_3_0) || defined(GL_VERSION_3_0)
+#define GLAD_GL_IS_SOME_NEW_VERSION 1
+#else
+#define GLAD_GL_IS_SOME_NEW_VERSION 0
+#endif
+
+static int glad_gl_get_extensions( int version, const char **out_exts, unsigned int *out_num_exts_i, char ***out_exts_i) {
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    if(GLAD_VERSION_MAJOR(version) < 3) {
+#else
+    (void) version;
+    (void) out_num_exts_i;
+    (void) out_exts_i;
+#endif
+        if (glad_glGetString == NULL) {
+            return 0;
+        }
+        *out_exts = (const char *)glad_glGetString(GL_EXTENSIONS);
+#if GLAD_GL_IS_SOME_NEW_VERSION
+    } else {
+        unsigned int index = 0;
+        unsigned int num_exts_i = 0;
+        char **exts_i = NULL;
+        if (glad_glGetStringi == NULL || glad_glGetIntegerv == NULL) {
+            return 0;
+        }
+        glad_glGetIntegerv(GL_NUM_EXTENSIONS, (int*) &num_exts_i);
+        if (num_exts_i > 0) {
+            exts_i = (char **) malloc(num_exts_i * (sizeof *exts_i));
+        }
+        if (exts_i == NULL) {
+            return 0;
+        }
+        for(index = 0; index < num_exts_i; index++) {
+            const char *gl_str_tmp = (const char*) glad_glGetStringi(GL_EXTENSIONS, index);
+            size_t len = strlen(gl_str_tmp) + 1;
+
+            char *local_str = (char*) malloc(len * sizeof(char));
+            if(local_str != NULL) {
+                memcpy(local_str, gl_str_tmp, len * sizeof(char));
+            }
+
+            exts_i[index] = local_str;
+        }
+
+        *out_num_exts_i = num_exts_i;
+        *out_exts_i = exts_i;
+    }
+#endif
+    return 1;
+}
+static void glad_gl_free_extensions(char **exts_i, unsigned int num_exts_i) {
+    if (exts_i != NULL) {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            free((void *) (exts_i[index]));
+        }
+        free((void *)exts_i);
+        exts_i = NULL;
+    }
+}
+static int glad_gl_has_extension(int version, const char *exts, unsigned int num_exts_i, char **exts_i, const char *ext) {
+    if(GLAD_VERSION_MAJOR(version) < 3 || !GLAD_GL_IS_SOME_NEW_VERSION) {
+        const char *extensions;
+        const char *loc;
+        const char *terminator;
+        extensions = exts;
+        if(extensions == NULL || ext == NULL) {
+            return 0;
+        }
+        while(1) {
+            loc = strstr(extensions, ext);
+            if(loc == NULL) {
+                return 0;
+            }
+            terminator = loc + strlen(ext);
+            if((loc == extensions || *(loc - 1) == ' ') &&
+                (*terminator == ' ' || *terminator == '\0')) {
+                return 1;
+            }
+            extensions = terminator;
+        }
+    } else {
+        unsigned int index;
+        for(index = 0; index < num_exts_i; index++) {
+            const char *e = exts_i[index];
+            if(strcmp(e, ext) == 0) {
+                return 1;
+            }
+        }
+    }
+    return 0;
+}
+
+static GLADapiproc glad_gl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_gl_find_extensions_gles2( int version) {
+    const char *exts = NULL;
+    unsigned int num_exts_i = 0;
+    char **exts_i = NULL;
+    if (!glad_gl_get_extensions(version, &exts, &num_exts_i, &exts_i)) return 0;
+
+    GLAD_GL_OES_mapbuffer = glad_gl_has_extension(version, exts, num_exts_i, exts_i, "GL_OES_mapbuffer");
+    GLAD_GL_OES_required_internalformat = glad_gl_has_extension(version, exts, num_exts_i, exts_i, "GL_OES_required_internalformat");
+
+    glad_gl_free_extensions(exts_i, num_exts_i);
+
+    return 1;
+}
+
+static int glad_gl_find_core_gles2(void) {
+    int i, major, minor;
+    const char* version;
+    const char* prefixes[] = {
+        "OpenGL ES-CM ",
+        "OpenGL ES-CL ",
+        "OpenGL ES ",
+        NULL
+    };
+    version = (const char*) glad_glGetString(GL_VERSION);
+    if (!version) return 0;
+    for (i = 0;  prefixes[i];  i++) {
+        const size_t length = strlen(prefixes[i]);
+        if (strncmp(version, prefixes[i], length) == 0) {
+            version += length;
+            break;
+        }
+    }
+
+    GLAD_IMPL_UTIL_SSCANF(version, "%d.%d", &major, &minor);
+
+    GLAD_GL_ES_VERSION_2_0 = (major == 2 && minor >= 0) || major > 2;
+
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLES2UserPtr( GLADuserptrloadfunc load, void *userptr) {
+    int version;
+
+    glad_glGetString = (PFNGLGETSTRINGPROC) load(userptr, "glGetString");
+    if(glad_glGetString == NULL) return 0;
+    if(glad_glGetString(GL_VERSION) == NULL) return 0;
+    version = glad_gl_find_core_gles2();
+
+    glad_gl_load_GL_ES_VERSION_2_0(load, userptr);
+
+    if (!glad_gl_find_extensions_gles2(version)) return 0;
+    glad_gl_load_GL_OES_mapbuffer(load, userptr);
+
+
+
+    return version;
+}
+
+
+int gladLoadGLES2( GLADloadfunc load) {
+    return gladLoadGLES2UserPtr( glad_gl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/src/glx.c glmark2-2021.02/src/glad/src/glx.c
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/src/glx.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/src/glx.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,232 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/glx.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_GLX_VERSION_1_0 = 0;
+int GLAD_GLX_VERSION_1_1 = 0;
+int GLAD_GLX_VERSION_1_2 = 0;
+int GLAD_GLX_VERSION_1_3 = 0;
+int GLAD_GLX_VERSION_1_4 = 0;
+int GLAD_GLX_EXT_swap_control = 0;
+int GLAD_GLX_MESA_swap_control = 0;
+
+
+
+PFNGLXCHOOSEFBCONFIGPROC glad_glXChooseFBConfig = NULL;
+PFNGLXCHOOSEVISUALPROC glad_glXChooseVisual = NULL;
+PFNGLXCOPYCONTEXTPROC glad_glXCopyContext = NULL;
+PFNGLXCREATECONTEXTPROC glad_glXCreateContext = NULL;
+PFNGLXCREATEGLXPIXMAPPROC glad_glXCreateGLXPixmap = NULL;
+PFNGLXCREATENEWCONTEXTPROC glad_glXCreateNewContext = NULL;
+PFNGLXCREATEPBUFFERPROC glad_glXCreatePbuffer = NULL;
+PFNGLXCREATEPIXMAPPROC glad_glXCreatePixmap = NULL;
+PFNGLXCREATEWINDOWPROC glad_glXCreateWindow = NULL;
+PFNGLXDESTROYCONTEXTPROC glad_glXDestroyContext = NULL;
+PFNGLXDESTROYGLXPIXMAPPROC glad_glXDestroyGLXPixmap = NULL;
+PFNGLXDESTROYPBUFFERPROC glad_glXDestroyPbuffer = NULL;
+PFNGLXDESTROYPIXMAPPROC glad_glXDestroyPixmap = NULL;
+PFNGLXDESTROYWINDOWPROC glad_glXDestroyWindow = NULL;
+PFNGLXGETCLIENTSTRINGPROC glad_glXGetClientString = NULL;
+PFNGLXGETCONFIGPROC glad_glXGetConfig = NULL;
+PFNGLXGETCURRENTCONTEXTPROC glad_glXGetCurrentContext = NULL;
+PFNGLXGETCURRENTDISPLAYPROC glad_glXGetCurrentDisplay = NULL;
+PFNGLXGETCURRENTDRAWABLEPROC glad_glXGetCurrentDrawable = NULL;
+PFNGLXGETCURRENTREADDRAWABLEPROC glad_glXGetCurrentReadDrawable = NULL;
+PFNGLXGETFBCONFIGATTRIBPROC glad_glXGetFBConfigAttrib = NULL;
+PFNGLXGETFBCONFIGSPROC glad_glXGetFBConfigs = NULL;
+PFNGLXGETPROCADDRESSPROC glad_glXGetProcAddress = NULL;
+PFNGLXGETSELECTEDEVENTPROC glad_glXGetSelectedEvent = NULL;
+PFNGLXGETSWAPINTERVALMESAPROC glad_glXGetSwapIntervalMESA = NULL;
+PFNGLXGETVISUALFROMFBCONFIGPROC glad_glXGetVisualFromFBConfig = NULL;
+PFNGLXISDIRECTPROC glad_glXIsDirect = NULL;
+PFNGLXMAKECONTEXTCURRENTPROC glad_glXMakeContextCurrent = NULL;
+PFNGLXMAKECURRENTPROC glad_glXMakeCurrent = NULL;
+PFNGLXQUERYCONTEXTPROC glad_glXQueryContext = NULL;
+PFNGLXQUERYDRAWABLEPROC glad_glXQueryDrawable = NULL;
+PFNGLXQUERYEXTENSIONPROC glad_glXQueryExtension = NULL;
+PFNGLXQUERYEXTENSIONSSTRINGPROC glad_glXQueryExtensionsString = NULL;
+PFNGLXQUERYSERVERSTRINGPROC glad_glXQueryServerString = NULL;
+PFNGLXQUERYVERSIONPROC glad_glXQueryVersion = NULL;
+PFNGLXSELECTEVENTPROC glad_glXSelectEvent = NULL;
+PFNGLXSWAPBUFFERSPROC glad_glXSwapBuffers = NULL;
+PFNGLXSWAPINTERVALEXTPROC glad_glXSwapIntervalEXT = NULL;
+PFNGLXSWAPINTERVALMESAPROC glad_glXSwapIntervalMESA = NULL;
+PFNGLXUSEXFONTPROC glad_glXUseXFont = NULL;
+PFNGLXWAITGLPROC glad_glXWaitGL = NULL;
+PFNGLXWAITXPROC glad_glXWaitX = NULL;
+
+
+static void glad_glx_load_GLX_VERSION_1_0( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_0) return;
+    glad_glXChooseVisual = (PFNGLXCHOOSEVISUALPROC) load(userptr, "glXChooseVisual");
+    glad_glXCopyContext = (PFNGLXCOPYCONTEXTPROC) load(userptr, "glXCopyContext");
+    glad_glXCreateContext = (PFNGLXCREATECONTEXTPROC) load(userptr, "glXCreateContext");
+    glad_glXCreateGLXPixmap = (PFNGLXCREATEGLXPIXMAPPROC) load(userptr, "glXCreateGLXPixmap");
+    glad_glXDestroyContext = (PFNGLXDESTROYCONTEXTPROC) load(userptr, "glXDestroyContext");
+    glad_glXDestroyGLXPixmap = (PFNGLXDESTROYGLXPIXMAPPROC) load(userptr, "glXDestroyGLXPixmap");
+    glad_glXGetConfig = (PFNGLXGETCONFIGPROC) load(userptr, "glXGetConfig");
+    glad_glXGetCurrentContext = (PFNGLXGETCURRENTCONTEXTPROC) load(userptr, "glXGetCurrentContext");
+    glad_glXGetCurrentDrawable = (PFNGLXGETCURRENTDRAWABLEPROC) load(userptr, "glXGetCurrentDrawable");
+    glad_glXIsDirect = (PFNGLXISDIRECTPROC) load(userptr, "glXIsDirect");
+    glad_glXMakeCurrent = (PFNGLXMAKECURRENTPROC) load(userptr, "glXMakeCurrent");
+    glad_glXQueryExtension = (PFNGLXQUERYEXTENSIONPROC) load(userptr, "glXQueryExtension");
+    glad_glXQueryVersion = (PFNGLXQUERYVERSIONPROC) load(userptr, "glXQueryVersion");
+    glad_glXSwapBuffers = (PFNGLXSWAPBUFFERSPROC) load(userptr, "glXSwapBuffers");
+    glad_glXUseXFont = (PFNGLXUSEXFONTPROC) load(userptr, "glXUseXFont");
+    glad_glXWaitGL = (PFNGLXWAITGLPROC) load(userptr, "glXWaitGL");
+    glad_glXWaitX = (PFNGLXWAITXPROC) load(userptr, "glXWaitX");
+}
+static void glad_glx_load_GLX_VERSION_1_1( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_1) return;
+    glad_glXGetClientString = (PFNGLXGETCLIENTSTRINGPROC) load(userptr, "glXGetClientString");
+    glad_glXQueryExtensionsString = (PFNGLXQUERYEXTENSIONSSTRINGPROC) load(userptr, "glXQueryExtensionsString");
+    glad_glXQueryServerString = (PFNGLXQUERYSERVERSTRINGPROC) load(userptr, "glXQueryServerString");
+}
+static void glad_glx_load_GLX_VERSION_1_2( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_2) return;
+    glad_glXGetCurrentDisplay = (PFNGLXGETCURRENTDISPLAYPROC) load(userptr, "glXGetCurrentDisplay");
+}
+static void glad_glx_load_GLX_VERSION_1_3( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_3) return;
+    glad_glXChooseFBConfig = (PFNGLXCHOOSEFBCONFIGPROC) load(userptr, "glXChooseFBConfig");
+    glad_glXCreateNewContext = (PFNGLXCREATENEWCONTEXTPROC) load(userptr, "glXCreateNewContext");
+    glad_glXCreatePbuffer = (PFNGLXCREATEPBUFFERPROC) load(userptr, "glXCreatePbuffer");
+    glad_glXCreatePixmap = (PFNGLXCREATEPIXMAPPROC) load(userptr, "glXCreatePixmap");
+    glad_glXCreateWindow = (PFNGLXCREATEWINDOWPROC) load(userptr, "glXCreateWindow");
+    glad_glXDestroyPbuffer = (PFNGLXDESTROYPBUFFERPROC) load(userptr, "glXDestroyPbuffer");
+    glad_glXDestroyPixmap = (PFNGLXDESTROYPIXMAPPROC) load(userptr, "glXDestroyPixmap");
+    glad_glXDestroyWindow = (PFNGLXDESTROYWINDOWPROC) load(userptr, "glXDestroyWindow");
+    glad_glXGetCurrentReadDrawable = (PFNGLXGETCURRENTREADDRAWABLEPROC) load(userptr, "glXGetCurrentReadDrawable");
+    glad_glXGetFBConfigAttrib = (PFNGLXGETFBCONFIGATTRIBPROC) load(userptr, "glXGetFBConfigAttrib");
+    glad_glXGetFBConfigs = (PFNGLXGETFBCONFIGSPROC) load(userptr, "glXGetFBConfigs");
+    glad_glXGetSelectedEvent = (PFNGLXGETSELECTEDEVENTPROC) load(userptr, "glXGetSelectedEvent");
+    glad_glXGetVisualFromFBConfig = (PFNGLXGETVISUALFROMFBCONFIGPROC) load(userptr, "glXGetVisualFromFBConfig");
+    glad_glXMakeContextCurrent = (PFNGLXMAKECONTEXTCURRENTPROC) load(userptr, "glXMakeContextCurrent");
+    glad_glXQueryContext = (PFNGLXQUERYCONTEXTPROC) load(userptr, "glXQueryContext");
+    glad_glXQueryDrawable = (PFNGLXQUERYDRAWABLEPROC) load(userptr, "glXQueryDrawable");
+    glad_glXSelectEvent = (PFNGLXSELECTEVENTPROC) load(userptr, "glXSelectEvent");
+}
+static void glad_glx_load_GLX_VERSION_1_4( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_VERSION_1_4) return;
+    glad_glXGetProcAddress = (PFNGLXGETPROCADDRESSPROC) load(userptr, "glXGetProcAddress");
+}
+static void glad_glx_load_GLX_EXT_swap_control( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_EXT_swap_control) return;
+    glad_glXSwapIntervalEXT = (PFNGLXSWAPINTERVALEXTPROC) load(userptr, "glXSwapIntervalEXT");
+}
+static void glad_glx_load_GLX_MESA_swap_control( GLADuserptrloadfunc load, void* userptr) {
+    if(!GLAD_GLX_MESA_swap_control) return;
+    glad_glXGetSwapIntervalMESA = (PFNGLXGETSWAPINTERVALMESAPROC) load(userptr, "glXGetSwapIntervalMESA");
+    glad_glXSwapIntervalMESA = (PFNGLXSWAPINTERVALMESAPROC) load(userptr, "glXSwapIntervalMESA");
+}
+
+
+
+static int glad_glx_has_extension(Display *display, int screen, const char *ext) {
+#ifndef GLX_VERSION_1_1
+    (void) display;
+    (void) screen;
+    (void) ext;
+#else
+    const char *terminator;
+    const char *loc;
+    const char *extensions;
+
+    if (glXQueryExtensionsString == NULL) {
+        return 0;
+    }
+
+    extensions = glXQueryExtensionsString(display, screen);
+
+    if(extensions == NULL || ext == NULL) {
+        return 0;
+    }
+
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL)
+            break;
+
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0')) {
+            return 1;
+        }
+        extensions = terminator;
+    }
+#endif
+
+    return 0;
+}
+
+static GLADapiproc glad_glx_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_glx_find_extensions(Display *display, int screen) {
+    GLAD_GLX_EXT_swap_control = glad_glx_has_extension(display, screen, "GLX_EXT_swap_control");
+    GLAD_GLX_MESA_swap_control = glad_glx_has_extension(display, screen, "GLX_MESA_swap_control");
+    return 1;
+}
+
+static int glad_glx_find_core_glx(Display **display, int *screen) {
+    int major = 0, minor = 0;
+    if(*display == NULL) {
+#ifdef GLAD_GLX_NO_X11
+        (void) screen;
+        return 0;
+#else
+        *display = XOpenDisplay(0);
+        if (*display == NULL) {
+            return 0;
+        }
+        *screen = XScreenNumberOfScreen(XDefaultScreenOfDisplay(*display));
+#endif
+    }
+    glXQueryVersion(*display, &major, &minor);
+    GLAD_GLX_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    GLAD_GLX_VERSION_1_1 = (major == 1 && minor >= 1) || major > 1;
+    GLAD_GLX_VERSION_1_2 = (major == 1 && minor >= 2) || major > 1;
+    GLAD_GLX_VERSION_1_3 = (major == 1 && minor >= 3) || major > 1;
+    GLAD_GLX_VERSION_1_4 = (major == 1 && minor >= 4) || major > 1;
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadGLXUserPtr(Display *display, int screen, GLADuserptrloadfunc load, void *userptr) {
+    int version;
+    glXQueryVersion = (PFNGLXQUERYVERSIONPROC) load(userptr, "glXQueryVersion");
+    if(glXQueryVersion == NULL) return 0;
+    version = glad_glx_find_core_glx(&display, &screen);
+
+    glad_glx_load_GLX_VERSION_1_0(load, userptr);
+    glad_glx_load_GLX_VERSION_1_1(load, userptr);
+    glad_glx_load_GLX_VERSION_1_2(load, userptr);
+    glad_glx_load_GLX_VERSION_1_3(load, userptr);
+    glad_glx_load_GLX_VERSION_1_4(load, userptr);
+
+    if (!glad_glx_find_extensions(display, screen)) return 0;
+    glad_glx_load_GLX_EXT_swap_control(load, userptr);
+    glad_glx_load_GLX_MESA_swap_control(load, userptr);
+
+    return version;
+}
+
+int gladLoadGLX(Display *display, int screen, GLADloadfunc load) {
+    return gladLoadGLXUserPtr(display, screen, glad_glx_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/glad/src/wgl.c glmark2-2021.02/src/glad/src/wgl.c
--- glmark2-2014.03+git20150611.fa71af2d/src/glad/src/wgl.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/glad/src/wgl.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,117 @@
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <glad/wgl.h>
+
+#ifndef GLAD_IMPL_UTIL_C_
+#define GLAD_IMPL_UTIL_C_
+
+#ifdef _MSC_VER
+#define GLAD_IMPL_UTIL_SSCANF sscanf_s
+#else
+#define GLAD_IMPL_UTIL_SSCANF sscanf
+#endif
+
+#endif /* GLAD_IMPL_UTIL_C_ */
+
+
+int GLAD_WGL_VERSION_1_0 = 0;
+int GLAD_WGL_ARB_extensions_string = 0;
+int GLAD_WGL_EXT_extensions_string = 0;
+int GLAD_WGL_EXT_swap_control = 0;
+
+
+
+PFNWGLGETEXTENSIONSSTRINGARBPROC glad_wglGetExtensionsStringARB = NULL;
+PFNWGLGETEXTENSIONSSTRINGEXTPROC glad_wglGetExtensionsStringEXT = NULL;
+PFNWGLGETSWAPINTERVALEXTPROC glad_wglGetSwapIntervalEXT = NULL;
+PFNWGLSWAPINTERVALEXTPROC glad_wglSwapIntervalEXT = NULL;
+
+
+static void glad_wgl_load_WGL_ARB_extensions_string(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_ARB_extensions_string) return;
+    glad_wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC) load(userptr, "wglGetExtensionsStringARB");
+}
+static void glad_wgl_load_WGL_EXT_extensions_string(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_EXT_extensions_string) return;
+    glad_wglGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC) load(userptr, "wglGetExtensionsStringEXT");
+}
+static void glad_wgl_load_WGL_EXT_swap_control(GLADuserptrloadfunc load, void *userptr) {
+    if(!GLAD_WGL_EXT_swap_control) return;
+    glad_wglGetSwapIntervalEXT = (PFNWGLGETSWAPINTERVALEXTPROC) load(userptr, "wglGetSwapIntervalEXT");
+    glad_wglSwapIntervalEXT = (PFNWGLSWAPINTERVALEXTPROC) load(userptr, "wglSwapIntervalEXT");
+}
+
+
+
+static int glad_wgl_has_extension(HDC hdc, const char *ext) {
+    const char *terminator;
+    const char *loc;
+    const char *extensions;
+
+    if(wglGetExtensionsStringEXT == NULL && wglGetExtensionsStringARB == NULL)
+        return 0;
+
+    if(wglGetExtensionsStringARB == NULL || hdc == INVALID_HANDLE_VALUE)
+        extensions = wglGetExtensionsStringEXT();
+    else
+        extensions = wglGetExtensionsStringARB(hdc);
+
+    if(extensions == NULL || ext == NULL)
+        return 0;
+
+    while(1) {
+        loc = strstr(extensions, ext);
+        if(loc == NULL)
+            break;
+
+        terminator = loc + strlen(ext);
+        if((loc == extensions || *(loc - 1) == ' ') &&
+            (*terminator == ' ' || *terminator == '\0'))
+        {
+            return 1;
+        }
+        extensions = terminator;
+    }
+
+    return 0;
+}
+
+static GLADapiproc glad_wgl_get_proc_from_userptr(void *userptr, const char* name) {
+    return (GLAD_GNUC_EXTENSION (GLADapiproc (*)(const char *name)) userptr)(name);
+}
+
+static int glad_wgl_find_extensions_wgl(HDC hdc) {
+    GLAD_WGL_ARB_extensions_string = glad_wgl_has_extension(hdc, "WGL_ARB_extensions_string");
+    GLAD_WGL_EXT_extensions_string = glad_wgl_has_extension(hdc, "WGL_EXT_extensions_string");
+    GLAD_WGL_EXT_swap_control = glad_wgl_has_extension(hdc, "WGL_EXT_swap_control");
+    return 1;
+}
+
+static int glad_wgl_find_core_wgl(void) {
+    int major = 1, minor = 0;
+    GLAD_WGL_VERSION_1_0 = (major == 1 && minor >= 0) || major > 1;
+    return GLAD_MAKE_VERSION(major, minor);
+}
+
+int gladLoadWGLUserPtr(HDC hdc, GLADuserptrloadfunc load, void *userptr) {
+    int version;
+    wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC) load(userptr, "wglGetExtensionsStringARB");
+    wglGetExtensionsStringEXT = (PFNWGLGETEXTENSIONSSTRINGEXTPROC) load(userptr, "wglGetExtensionsStringEXT");
+    if(wglGetExtensionsStringARB == NULL && wglGetExtensionsStringEXT == NULL) return 0;
+    version = glad_wgl_find_core_wgl();
+
+
+    if (!glad_wgl_find_extensions_wgl(hdc)) return 0;
+    glad_wgl_load_WGL_ARB_extensions_string(load, userptr);
+    glad_wgl_load_WGL_EXT_extensions_string(load, userptr);
+    glad_wgl_load_WGL_EXT_swap_control(load, userptr);
+
+    return version;
+}
+
+int gladLoadWGL(HDC hdc, GLADloadfunc load) {
+    return gladLoadWGLUserPtr(hdc, glad_wgl_get_proc_from_userptr, GLAD_GNUC_EXTENSION (void*) load);
+}
+ 
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-headers.cpp glmark2-2021.02/src/gl-headers.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-headers.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-headers.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,8 +21,8 @@
  */
 #include "gl-headers.h"
 
-void* (*GLExtensions::MapBuffer) (GLenum target, GLenum access) = 0;
-GLboolean (*GLExtensions::UnmapBuffer) (GLenum target) = 0;
+void* (GLAD_API_PTR *GLExtensions::MapBuffer) (GLenum target, GLenum access) = 0;
+GLboolean (GLAD_API_PTR *GLExtensions::UnmapBuffer) (GLenum target) = 0;
 
 bool
 GLExtensions::support(const std::string &ext)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-headers.h glmark2-2021.02/src/gl-headers.h
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-headers.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-headers.h	2021-04-25 01:47:22.000000000 +0800
@@ -25,14 +25,12 @@
 #define GL_GLEXT_PROTOTYPES
 
 #if GLMARK2_USE_GL
-#include <GL/gl.h>
-#include <GL/glext.h>
+#include <glad/gl.h>
 #ifndef GL_RGB565
 #define GL_RGB565 0x8D62
 #endif
 #elif GLMARK2_USE_GLESv2
-#include <GLES2/gl2.h>
-#include <GLES2/gl2ext.h>
+#include <glad/gles2.h>
 #ifndef GL_WRITE_ONLY
 #define GL_WRITE_ONLY GL_WRITE_ONLY_OES
 #endif
@@ -64,8 +62,8 @@
      */
     static bool support(const std::string &ext);
 
-    static void* (*MapBuffer) (GLenum target, GLenum access);
-    static GLboolean (*UnmapBuffer) (GLenum target);
+    static void* (GLAD_API_PTR *MapBuffer) (GLenum target, GLenum access);
+    static GLboolean (GLAD_API_PTR *UnmapBuffer) (GLenum target);
 };
 
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-egl.cpp glmark2-2021.02/src/gl-state-egl.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-egl.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-state-egl.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -27,10 +27,17 @@
 #include "gl-headers.h"
 #include <iomanip>
 #include <sstream>
+#include <cstring>
 
 using std::vector;
 using std::string;
 
+GLADapiproc load_egl_func(void *userdata, const char *name)
+{
+    SharedLibrary *lib = reinterpret_cast<SharedLibrary *>(userdata);
+    return reinterpret_cast<GLADapiproc>(lib->load(name));
+}
+
 /****************************
  * EGLConfig public methods *
  ****************************/
@@ -287,9 +294,34 @@
  * GLStateEGL public methods *
  ****************************/
 
+GLStateEGL::~GLStateEGL()
+{
+    if(egl_display_ != nullptr){
+        if(!eglTerminate(egl_display_))
+            Log::error("eglTerminate failed\n");
+    }
+
+    if(eglReleaseThread && !eglReleaseThread())
+       Log::error("eglReleaseThread failed\n");
+}
+
 bool
 GLStateEGL::init_display(void* native_display, GLVisualConfig& visual_config)
 {
+#if defined(WIN32)
+    if (!egl_lib_.open("libEGL.dll")) {
+#else
+    if (!egl_lib_.open_from_alternatives({"libEGL.so", "libEGL.so.1" })) {
+#endif
+        Log::error("Error loading EGL library\n");
+        return false;
+    }
+
+    if (gladLoadEGLUserPtr(EGL_NO_DISPLAY, load_egl_func, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
     native_display_ = reinterpret_cast<EGLNativeDisplayType>(native_display);
     requested_visual_config_ = visual_config;
 
@@ -304,20 +336,28 @@
     return gotValidSurface();
 }
 
-void
+bool
 GLStateEGL::init_gl_extensions()
 {
 #if GLMARK2_USE_GLESv2
+    if (!gladLoadGLES2UserPtr(load_proc, this)) {
+        Log::error("Loading GLESv2 entry points failed.");
+        return false;
+    }
+
     if (GLExtensions::support("GL_OES_mapbuffer")) {
-        GLExtensions::MapBuffer =
-            reinterpret_cast<PFNGLMAPBUFFEROESPROC>(eglGetProcAddress("glMapBufferOES"));
-        GLExtensions::UnmapBuffer =
-            reinterpret_cast<PFNGLUNMAPBUFFEROESPROC>(eglGetProcAddress("glUnmapBufferOES"));
+        GLExtensions::MapBuffer = glMapBufferOES;
+        GLExtensions::UnmapBuffer = glUnmapBufferOES;
     }
 #elif GLMARK2_USE_GL
+    if (!gladLoadGLUserPtr(load_proc, this)) {
+        Log::error("Loading GL entry points failed.");
+        return false;
+    }
     GLExtensions::MapBuffer = glMapBuffer;
     GLExtensions::UnmapBuffer = glUnmapBuffer;
 #endif
+    return true;
 }
 
 bool
@@ -347,7 +387,9 @@
         Log::info("** Failed to set swap interval. Results may be bounded above by refresh rate.\n");
     }
 
-    init_gl_extensions();
+    if (!init_gl_extensions()) {
+        return false;
+    }
 
     return true;
 }
@@ -379,7 +421,7 @@
 }
 
 bool
-GLStateEGL::gotNativeConfig(int& vid)
+GLStateEGL::gotNativeConfig(intptr_t& vid)
 {
     if (!gotValidConfig())
         return false;
@@ -409,17 +451,64 @@
  * GLStateEGL private methods *
  *****************************/
 
+#ifdef GLMARK2_USE_X11
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_X11_KHR
+#elif  GLMARK2_USE_WAYLAND
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_WAYLAND_KHR
+#elif  GLMARK2_USE_DRM
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_GBM_KHR
+#elif  GLMARK2_USE_MIR
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM EGL_PLATFORM_MIR_KHR
+#else
+// Platforms not in the above platform enums fall back to eglGetDisplay.
+#define GLMARK2_NATIVE_EGL_DISPLAY_ENUM 0
+#endif
+
 bool
 GLStateEGL::gotValidDisplay()
 {
     if (egl_display_)
         return true;
 
-    egl_display_ = eglGetDisplay(native_display_);
+    char const * __restrict const supported_extensions =
+        eglQueryString(EGL_NO_DISPLAY, EGL_EXTENSIONS);
+
+    if (GLMARK2_NATIVE_EGL_DISPLAY_ENUM != 0 && supported_extensions
+        && strstr(supported_extensions, "EGL_EXT_platform_base"))
+    {
+        Log::debug("Using eglGetPlatformDisplayEXT()\n");
+        PFNEGLGETPLATFORMDISPLAYEXTPROC get_platform_display =
+            reinterpret_cast<PFNEGLGETPLATFORMDISPLAYEXTPROC>(
+                eglGetProcAddress("eglGetPlatformDisplayEXT"));
+
+        if (get_platform_display != nullptr) {
+            egl_display_ = get_platform_display(
+                GLMARK2_NATIVE_EGL_DISPLAY_ENUM,
+                reinterpret_cast<void*>(native_display_),
+                nullptr);
+        }
+
+        if (!egl_display_) {
+            Log::debug("eglGetPlatformDisplayEXT() failed with error: 0x%x\n",
+                       eglGetError());
+        }
+    }
+    else
+    {
+        Log::debug("eglGetPlatformDisplayEXT() seems unsupported\n");
+    }
+
+    /* Just in case get_platform_display failed... */
+    if (!egl_display_) {
+        Log::debug("Falling back to eglGetDisplay()\n");
+        egl_display_ = eglGetDisplay(native_display_);
+    }
+
     if (!egl_display_) {
         Log::error("eglGetDisplay() failed with error: 0x%x\n", eglGetError());
         return false;
     }
+
     int egl_major(-1);
     int egl_minor(-1);
     if (!eglInitialize(egl_display_, &egl_major, &egl_minor)) {
@@ -428,16 +517,33 @@
         return false;
     }
 
+    /* Reinitialize GLAD with a known display */
+    if (gladLoadEGLUserPtr(egl_display_, load_egl_func, &egl_lib_) == 0) {
+        Log::error("Loading EGL entry points failed\n");
+        return false;
+    }
+
 #if GLMARK2_USE_GLESv2
     EGLenum apiType(EGL_OPENGL_ES_API);
+#if defined(WIN32)
+    std::initializer_list<const char *> libNames = { "libGLESv2.dll" };
+#else
+    std::initializer_list<const char *> libNames = { "libGLESv2.so", "libGLESv2.so.2" };
+#endif
 #elif GLMARK2_USE_GL
     EGLenum apiType(EGL_OPENGL_API);
+    std::initializer_list<const char *> libNames = { "libGL.so", "libGL.so.1" };
 #endif
     if (!eglBindAPI(apiType)) {
         Log::error("Failed to bind api EGL_OPENGL_ES_API\n");
         return false;
     }
 
+    if (!gl_lib_.open_from_alternatives(libNames)) {
+        Log::error("Error loading GL library\n");
+        return false;
+    }
+
     return true;
 }
 
@@ -617,3 +723,16 @@
     return true;
 }
 
+GLADapiproc
+GLStateEGL::load_proc(void *userptr, const char* name)
+{
+    if (eglGetProcAddress) {
+        GLADapiproc sym = reinterpret_cast<GLADapiproc>(eglGetProcAddress(name));
+        if (sym) {
+            return sym;
+        }
+    }
+
+    GLStateEGL* state = reinterpret_cast<GLStateEGL*>(userptr);
+    return reinterpret_cast<GLADapiproc>(state->gl_lib_.load(name));
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-egl.h glmark2-2021.02/src/gl-state-egl.h
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-egl.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-state-egl.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -23,9 +23,10 @@
 #define GLMARK2_GL_STATE_EGL_H_
 
 #include <vector>
-#include <EGL/egl.h>
+#include <glad/egl.h>
 #include "gl-state.h"
 #include "gl-visual-config.h"
+#include "shared-library.h"
 
 class EglConfig
 {
@@ -124,12 +125,17 @@
     EGLSurface egl_surface_;
     GLVisualConfig requested_visual_config_;
     EglConfig best_config_;
+    SharedLibrary egl_lib_;
+    SharedLibrary gl_lib_;
     bool gotValidDisplay();
     bool gotValidConfig();
     bool gotValidSurface();
     bool gotValidContext();
     void get_glvisualconfig(EGLConfig config, GLVisualConfig& visual_config);
     EGLConfig select_best_config(std::vector<EGLConfig>& configs);
+
+    static GLADapiproc load_proc(void *userptr, const char* name);
+
 public:
     GLStateEGL() :
         native_display_(0),
@@ -138,15 +144,16 @@
         egl_config_(0),
         egl_context_(0),
         egl_surface_(0) {}
+    ~GLStateEGL();
 
     bool init_display(void* native_display, GLVisualConfig& config_pref);
     bool init_surface(void* native_window);
-    void init_gl_extensions();
+    bool init_gl_extensions();
     bool valid();
     bool reset();
     void swap();
     // Performs a config search, returning a native visual ID on success
-    bool gotNativeConfig(int& vid);
+    bool gotNativeConfig(intptr_t& vid);
     void getVisualConfig(GLVisualConfig& vc);
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-glx.cpp glmark2-2021.02/src/gl-state-glx.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-glx.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-state-glx.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,13 +26,6 @@
 
 #include <climits>
 
-namespace
-{
-PFNGLXSWAPINTERVALEXTPROC glXSwapIntervalEXT_;
-PFNGLXSWAPINTERVALMESAPROC glXSwapIntervalMESA_;
-PFNGLXGETSWAPINTERVALMESAPROC glXGetSwapIntervalMESA_;
-}
-
 /******************
  * Public methods *
  ******************/
@@ -43,6 +36,12 @@
     xdpy_ = reinterpret_cast<Display*>(native_display);
     requested_visual_config_ = visual_config;
 
+    if (!lib_.open_from_alternatives({"libGL.so", "libGL.so.1"})) {
+        Log::error("Failed to load libGL\n");
+        return false;
+    }
+
+    gladLoadGLXUserPtr(xdpy_, DefaultScreen(xdpy_), load_proc, this);
     return (xdpy_ != 0);
 }
 
@@ -54,11 +53,12 @@
     return (xwin_ != 0);
 }
 
-void
+bool
 GLStateGLX::init_gl_extensions()
 {
     GLExtensions::MapBuffer = glMapBuffer;
     GLExtensions::UnmapBuffer = glUnmapBuffer;
+    return true;
 }
 
 bool
@@ -80,20 +80,26 @@
         return false;
     }
 
-    init_gl_extensions();
+    if (gladLoadGLUserPtr(load_proc, this) == 0) {
+        Log::error("Failed to load GL entry points\n");
+        return false;
+    }
+
+    if (!init_gl_extensions())
+        return false;
 
     unsigned int desired_swap(0);
     unsigned int actual_swap(-1);
-    if (glXSwapIntervalEXT_) {
-        glXSwapIntervalEXT_(xdpy_, xwin_, desired_swap);
+    if (glXSwapIntervalEXT) {
+        glXSwapIntervalEXT(xdpy_, xwin_, desired_swap);
         glXQueryDrawable(xdpy_, xwin_, GLX_SWAP_INTERVAL_EXT, &actual_swap);
         if (actual_swap == desired_swap)
             return true;
     }
 
-    if (glXSwapIntervalMESA_) {
-        glXSwapIntervalMESA_(desired_swap);
-        actual_swap = glXGetSwapIntervalMESA_();
+    if (glXSwapIntervalMESA) {
+        glXSwapIntervalMESA(desired_swap);
+        actual_swap = glXGetSwapIntervalMESA();
         if (actual_swap == desired_swap)
             return true;
     }
@@ -123,7 +129,7 @@
 }
 
 bool
-GLStateGLX::gotNativeConfig(int& vid)
+GLStateGLX::gotNativeConfig(intptr_t& vid)
 {
     if (!ensure_glx_fbconfig())
         return false;
@@ -186,31 +192,7 @@
      * GLX_SGI_swap_control is not enough because it doesn't allow 0 as a valid
      * value (i.e. you can't turn off VSync).
      */
-    if (extString.find("GLX_EXT_swap_control") != std::string::npos) {
-        glXSwapIntervalEXT_ =
-            reinterpret_cast<PFNGLXSWAPINTERVALEXTPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXSwapIntervalEXT")
-                )
-            );
-    }
-    else if (extString.find("GLX_MESA_swap_control") != std::string::npos) {
-        glXSwapIntervalMESA_ =
-            reinterpret_cast<PFNGLXSWAPINTERVALMESAPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXSwapIntervalMESA")
-                )
-            );
-        glXGetSwapIntervalMESA_ =
-            reinterpret_cast<PFNGLXGETSWAPINTERVALMESAPROC>(
-                glXGetProcAddress(
-                    reinterpret_cast<const GLubyte *>("glXGetSwapIntervalMESA")
-                )
-            );
-    }
-
-
-    if (!glXSwapIntervalEXT_ && !glXSwapIntervalMESA_) {
+    if (!glXSwapIntervalEXT && !glXSwapIntervalMESA) {
         Log::info("** GLX does not support GLX_EXT_swap_control or GLX_MESA_swap_control!\n");
     }
 }
@@ -346,3 +328,18 @@
 
     return best_config;
 }
+
+GLADapiproc
+GLStateGLX::load_proc(void *userptr, const char* name)
+{
+    if (glXGetProcAddress) {
+        const GLubyte* bytes = reinterpret_cast<const GLubyte*>(name);
+        GLADapiproc sym = glXGetProcAddress(bytes);
+        if (sym) {
+            return sym;
+        }
+    }
+
+    GLStateGLX* state = reinterpret_cast<GLStateGLX*>(userptr);
+    return reinterpret_cast<GLADapiproc>(state->lib_.load(name));
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-glx.h glmark2-2021.02/src/gl-state-glx.h
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-glx.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-state-glx.h	2021-04-25 01:47:22.000000000 +0800
@@ -26,13 +26,11 @@
 #include "gl-state.h"
 #include "gl-visual-config.h"
 #include "gl-headers.h"
+#include "shared-library.h"
 
 #include <vector>
 
-#include <X11/Xlib.h>
-#define GLX_GLXEXT_PROTOTYPES
-#include <GL/glx.h>
-#include <GL/glxext.h>
+#include <glad/glx.h>
 
 class GLStateGLX : public GLState
 {
@@ -43,10 +41,10 @@
     bool valid();
     bool init_display(void* native_display, GLVisualConfig& config_pref);
     bool init_surface(void* native_window);
-    void init_gl_extensions();
+    bool init_gl_extensions();
     bool reset();
     void swap();
-    bool gotNativeConfig(int& vid);
+    bool gotNativeConfig(intptr_t& vid);
     void getVisualConfig(GLVisualConfig& vc);
 
 private:
@@ -57,11 +55,14 @@
     void get_glvisualconfig_glx(GLXFBConfig config, GLVisualConfig &visual_config);
     GLXFBConfig select_best_config(std::vector<GLXFBConfig> configs);
 
+    static GLADapiproc load_proc(void* userptr, const char* name);
+
     Display* xdpy_;
     Window xwin_;
     GLXFBConfig glx_fbconfig_;
     GLXContext glx_context_;
     GLVisualConfig requested_visual_config_;
+    SharedLibrary lib_;
 };
 
 #endif /* GLMARK2_GL_STATE_GLX_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state.h glmark2-2021.02/src/gl-state.h
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-state.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,6 +22,8 @@
 #ifndef GLMARK2_GL_STATE_H_
 #define GLMARK2_GL_STATE_H_
 
+#include <stdint.h>
+
 class GLVisualConfig;
 
 class GLState
@@ -31,11 +33,11 @@
 
     virtual bool init_display(void *native_display, GLVisualConfig& config_pref) = 0;
     virtual bool init_surface(void *native_window) = 0;
-    virtual void init_gl_extensions() = 0;
+    virtual bool init_gl_extensions() = 0;
     virtual bool valid() = 0;
     virtual bool reset() = 0;
     virtual void swap() = 0;
-    virtual bool gotNativeConfig(int& vid) = 0;
+    virtual bool gotNativeConfig(intptr_t& vid) = 0;
     virtual void getVisualConfig(GLVisualConfig& vc) = 0;
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-wgl.cpp glmark2-2021.02/src/gl-state-wgl.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-wgl.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/gl-state-wgl.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,178 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+#include "gl-state-wgl.h"
+
+#include <glad/wgl.h>
+#include "gl-headers.h"
+#include "log.h"
+#include "options.h"
+
+/******************
+ * Public methods *
+ ******************/
+
+GLStateWGL::GLStateWGL() : hdc_(nullptr), wgl_context_(nullptr), get_proc_addr_(nullptr) {}
+
+GLStateWGL::~GLStateWGL()
+{
+    if (wgl_context_) {
+        wglDeleteContext(wgl_context_);
+        wgl_context_ = nullptr;
+    }
+}
+
+bool
+GLStateWGL::init_display(void* native_display, GLVisualConfig& /*visual_config*/)
+{
+    if (!wgl_library_.open("opengl32.dll")) {
+        Log::error("Failed to load opengl32.dll\n");
+        return false;
+    }
+
+    hdc_ = reinterpret_cast<HDC>(native_display);
+
+    PIXELFORMATDESCRIPTOR pixel_format_desc = {};
+    pixel_format_desc.nSize = sizeof(pixel_format_desc);
+    pixel_format_desc.nVersion = 1;
+    pixel_format_desc.dwFlags =
+        PFD_DRAW_TO_WINDOW | PFD_GENERIC_ACCELERATED | PFD_SUPPORT_OPENGL | PFD_DOUBLEBUFFER;
+    pixel_format_desc.iPixelType = PFD_TYPE_RGBA;
+    pixel_format_desc.cColorBits = 24;
+    pixel_format_desc.cAlphaBits = 8;
+    pixel_format_desc.cDepthBits = 24;
+    pixel_format_desc.cStencilBits = 8;
+    pixel_format_desc.iLayerType = PFD_MAIN_PLANE;
+
+    int pixelFormat = ChoosePixelFormat(hdc_, &pixel_format_desc);
+    if (pixelFormat == 0) {
+        Log::error("Could not find a compatible pixel format\n");
+        return false;
+    }
+
+    if (SetPixelFormat(hdc_, pixelFormat, &pixel_format_desc) != TRUE) {
+        Log::error("Failed to set the pixel format\n");
+        return false;
+    }
+
+    wgl_context_ = wglCreateContext(hdc_);
+    if (!wgl_context_) {
+        Log::error("Failed to create a WGL context\n");
+        return false;
+    }
+
+    if (wglMakeCurrent(hdc_, wgl_context_) == FALSE) {
+        Log::error("Error during wglMakeCurrent\n");
+        return false;
+    }
+
+    get_proc_addr_ = wgl_library_.load("wglGetProcAddress");
+    if (!get_proc_addr_) {
+        Log::error("Failed to find wglGetProcAddress\n");
+        return false;
+    }
+
+    if (gladLoadGLUserPtr(load_proc, this) == 0) {
+        Log::error("Failed to load GL entry points\n");
+        return false;
+    }
+
+    if (gladLoadWGLUserPtr(hdc_, load_proc, this) == 0) {
+        Log::error("Failed to load WGL entry points\n");
+        return false;
+    }
+
+    return true;
+}
+
+bool
+GLStateWGL::init_surface(void* /*native_window*/)
+{
+    return true;
+}
+
+bool
+GLStateWGL::init_gl_extensions()
+{
+    GLExtensions::MapBuffer = glMapBuffer;
+    GLExtensions::UnmapBuffer = glUnmapBuffer;
+    return true;
+}
+
+bool
+GLStateWGL::valid()
+{
+    if (!wglSwapIntervalEXT || wglSwapIntervalEXT(0) == FALSE) {
+        Log::info("** Failed to set swap interval. Results may be bounded above by refresh rate.\n");
+    }
+
+    return wgl_context_ != nullptr;
+}
+
+
+bool
+GLStateWGL::reset()
+{
+    return true;
+}
+
+void
+GLStateWGL::swap()
+{
+    if (SwapBuffers(hdc_) == FALSE) {
+        Log::error("Error during SwapBuffers\n");
+    }
+}
+
+bool
+GLStateWGL::gotNativeConfig(intptr_t& vid)
+{
+    vid = 1;
+    return true;
+}
+
+void
+GLStateWGL::getVisualConfig(GLVisualConfig& vc)
+{
+    vc.buffer = 32;
+    vc.red = 8;
+    vc.green = 8;
+    vc.blue = 8;
+    vc.alpha = 8;
+    vc.depth = 24;
+    vc.stencil = 8;
+}
+
+/*******************
+ * Private methods *
+ *******************/
+
+GLStateWGL::api_proc GLStateWGL::load_proc(void *userptr, const char *name)
+{
+    GLStateWGL* state = reinterpret_cast<GLStateWGL*>(userptr);
+    auto get_proc_addr = reinterpret_cast<decltype(&wglGetProcAddress)>(state->get_proc_addr_);
+    auto proc = reinterpret_cast<GLStateWGL::api_proc>(get_proc_addr(name));
+    if (proc) {
+        return proc;
+    }
+    return reinterpret_cast<GLStateWGL::api_proc>(state->wgl_library_.load(name));
+}
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-state-wgl.h glmark2-2021.02/src/gl-state-wgl.h
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-state-wgl.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/gl-state-wgl.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,54 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+#ifndef GLMARK2_GL_STATE_WGL_H_
+#define GLMARK2_GL_STATE_WGL_H_
+
+#include "gl-state.h"
+#include "shared-library.h"
+#include <windows.h>
+
+class GLStateWGL : public GLState
+{
+public:
+    GLStateWGL();
+    ~GLStateWGL();
+
+    bool init_display(void* native_display, GLVisualConfig& config_pref);
+    bool init_surface(void* native_window);
+    bool init_gl_extensions();
+    bool valid();
+    bool reset();
+    void swap();
+    bool gotNativeConfig(intptr_t& vid);
+    void getVisualConfig(GLVisualConfig& vc);
+
+private:
+    using api_proc = void (*)();
+    static api_proc load_proc(void *userptr, const char *name);
+
+    SharedLibrary wgl_library_;
+    HDC hdc_;
+    HGLRC wgl_context_;
+    void* get_proc_addr_;
+};
+
+#endif // GLMARK2_GL_STATE_WGL_H_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/gl-visual-config.cpp glmark2-2021.02/src/gl-visual-config.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/gl-visual-config.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/gl-visual-config.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -108,6 +108,14 @@
         /* Reward exact matches with the maximum per component score */
         score = MAXIMUM_COMPONENT_SCORE;
     }
+    else if (component > 8 && target <= 8 && scale > 1)
+    {
+        /* Penalize RGBA component widths larger than 8, since they are
+         * unlikely to be what the users want or properly supported for
+         * display. Such widths can still be used, but only if explicitly
+         * requested. */
+        score = UNACCEPTABLE_COMPONENT_PENALTY;
+    }
     else
     {
         /* 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/image-reader.cpp glmark2-2021.02/src/image-reader.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/image-reader.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/image-reader.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -120,7 +120,7 @@
 
     Log::debug("Reading PNG file %s\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> is_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> is_ptr(Util::get_resource(filename));
     if (!(*is_ptr)) {
         Log::error("Cannot open file %s!\n", filename.c_str());
         return false;
@@ -268,7 +268,7 @@
             size_t n = static_cast<size_t>(num_bytes);
             while (n > src->pub.bytes_in_buffer) {
                 n -= src->pub.bytes_in_buffer;
-                (*src->fill_input_buffer)(cinfo);
+                src->fill_input_buffer(cinfo);
             }
             src->pub.next_input_byte += n;
             src->pub.bytes_in_buffer -= n;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/include/dirent.h glmark2-2021.02/src/include/dirent.h
--- glmark2-2014.03+git20150611.fa71af2d/src/include/dirent.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/include/dirent.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1224 @@
+/*
+ * Dirent interface for Microsoft Visual Studio
+ *
+ * Copyright (C) 2006-2012 Toni Ronkko
+ * This file is part of dirent.  Dirent may be freely distributed
+ * under the MIT license.  For all details and documentation, see
+ * https://github.com/tronkko/dirent
+ */
+#ifndef DIRENT_H
+#define DIRENT_H
+ 
+/*
+ * Include windows.h without Windows Sockets 1.1 to prevent conflicts with
+ * Windows Sockets 2.0.
+ */
+#ifndef WIN32_LEAN_AND_MEAN
+#   define WIN32_LEAN_AND_MEAN
+#endif
+#include <windows.h>
+
+#include <errno.h>
+#include <malloc.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <wchar.h>
+
+/* Indicates that d_type field is available in dirent structure */
+#define _DIRENT_HAVE_D_TYPE
+
+/* Indicates that d_namlen field is available in dirent structure */
+#define _DIRENT_HAVE_D_NAMLEN
+
+/* Entries missing from MSVC 6.0 */
+#if !defined(FILE_ATTRIBUTE_DEVICE)
+#   define FILE_ATTRIBUTE_DEVICE 0x40
+#endif
+
+/* File type and permission flags for stat(), general mask */
+#if !defined(S_IFMT)
+#   define S_IFMT _S_IFMT
+#endif
+
+/* Directory bit */
+#if !defined(S_IFDIR)
+#   define S_IFDIR _S_IFDIR
+#endif
+
+/* Character device bit */
+#if !defined(S_IFCHR)
+#   define S_IFCHR _S_IFCHR
+#endif
+
+/* Pipe bit */
+#if !defined(S_IFFIFO)
+#   define S_IFFIFO _S_IFFIFO
+#endif
+
+/* Regular file bit */
+#if !defined(S_IFREG)
+#   define S_IFREG _S_IFREG
+#endif
+
+/* Read permission */
+#if !defined(S_IREAD)
+#   define S_IREAD _S_IREAD
+#endif
+
+/* Write permission */
+#if !defined(S_IWRITE)
+#   define S_IWRITE _S_IWRITE
+#endif
+
+/* Execute permission */
+#if !defined(S_IEXEC)
+#   define S_IEXEC _S_IEXEC
+#endif
+
+/* Pipe */
+#if !defined(S_IFIFO)
+#   define S_IFIFO _S_IFIFO
+#endif
+
+/* Block device */
+#if !defined(S_IFBLK)
+#   define S_IFBLK 0
+#endif
+
+/* Link */
+#if !defined(S_IFLNK)
+#   define S_IFLNK 0
+#endif
+
+/* Socket */
+#if !defined(S_IFSOCK)
+#   define S_IFSOCK 0
+#endif
+
+/* Read user permission */
+#if !defined(S_IRUSR)
+#   define S_IRUSR S_IREAD
+#endif
+
+/* Write user permission */
+#if !defined(S_IWUSR)
+#   define S_IWUSR S_IWRITE
+#endif
+
+/* Execute user permission */
+#if !defined(S_IXUSR)
+#   define S_IXUSR 0
+#endif
+
+/* Read group permission */
+#if !defined(S_IRGRP)
+#   define S_IRGRP 0
+#endif
+
+/* Write group permission */
+#if !defined(S_IWGRP)
+#   define S_IWGRP 0
+#endif
+
+/* Execute group permission */
+#if !defined(S_IXGRP)
+#   define S_IXGRP 0
+#endif
+
+/* Read others permission */
+#if !defined(S_IROTH)
+#   define S_IROTH 0
+#endif
+
+/* Write others permission */
+#if !defined(S_IWOTH)
+#   define S_IWOTH 0
+#endif
+
+/* Execute others permission */
+#if !defined(S_IXOTH)
+#   define S_IXOTH 0
+#endif
+
+/* Maximum length of file name */
+#if !defined(PATH_MAX)
+#   define PATH_MAX MAX_PATH
+#endif
+#if !defined(FILENAME_MAX)
+#   define FILENAME_MAX MAX_PATH
+#endif
+#if !defined(NAME_MAX)
+#   define NAME_MAX FILENAME_MAX
+#endif
+
+/* File type flags for d_type */
+#define DT_UNKNOWN 0
+#define DT_REG S_IFREG
+#define DT_DIR S_IFDIR
+#define DT_FIFO S_IFIFO
+#define DT_SOCK S_IFSOCK
+#define DT_CHR S_IFCHR
+#define DT_BLK S_IFBLK
+#define DT_LNK S_IFLNK
+
+/* Macros for converting between st_mode and d_type */
+#define IFTODT(mode) ((mode) & S_IFMT)
+#define DTTOIF(type) (type)
+
+/*
+ * File type macros.  Note that block devices, sockets and links cannot be
+ * distinguished on Windows and the macros S_ISBLK, S_ISSOCK and S_ISLNK are
+ * only defined for compatibility.  These macros should always return false
+ * on Windows.
+ */
+#if !defined(S_ISFIFO)
+#   define S_ISFIFO(mode) (((mode) & S_IFMT) == S_IFIFO)
+#endif
+#if !defined(S_ISDIR)
+#   define S_ISDIR(mode) (((mode) & S_IFMT) == S_IFDIR)
+#endif
+#if !defined(S_ISREG)
+#   define S_ISREG(mode) (((mode) & S_IFMT) == S_IFREG)
+#endif
+#if !defined(S_ISLNK)
+#   define S_ISLNK(mode) (((mode) & S_IFMT) == S_IFLNK)
+#endif
+#if !defined(S_ISSOCK)
+#   define S_ISSOCK(mode) (((mode) & S_IFMT) == S_IFSOCK)
+#endif
+#if !defined(S_ISCHR)
+#   define S_ISCHR(mode) (((mode) & S_IFMT) == S_IFCHR)
+#endif
+#if !defined(S_ISBLK)
+#   define S_ISBLK(mode) (((mode) & S_IFMT) == S_IFBLK)
+#endif
+
+/* Return the exact length of the file name without zero terminator */
+#define _D_EXACT_NAMLEN(p) ((p)->d_namlen)
+
+/* Return the maximum size of a file name */
+#define _D_ALLOC_NAMLEN(p) ((PATH_MAX)+1)
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/* Wide-character version */
+struct _wdirent {
+    /* Always zero */
+    long d_ino;
+
+    /* File position within stream */
+    long d_off;
+
+    /* Structure size */
+    unsigned short d_reclen;
+
+    /* Length of name without \0 */
+    size_t d_namlen;
+
+    /* File type */
+    int d_type;
+
+    /* File name */
+    wchar_t d_name[PATH_MAX+1];
+};
+typedef struct _wdirent _wdirent;
+
+struct _WDIR {
+    /* Current directory entry */
+    struct _wdirent ent;
+
+    /* Private file data */
+    WIN32_FIND_DATAW data;
+
+    /* True if data is valid */
+    int cached;
+
+    /* Win32 search handle */
+    HANDLE handle;
+
+    /* Initial directory name */
+    wchar_t *patt;
+};
+typedef struct _WDIR _WDIR;
+
+/* Multi-byte character version */
+struct dirent {
+    /* Always zero */
+    long d_ino;
+
+    /* File position within stream */
+    long d_off;
+
+    /* Structure size */
+    unsigned short d_reclen;
+
+    /* Length of name without \0 */
+    size_t d_namlen;
+
+    /* File type */
+    int d_type;
+
+    /* File name */
+    char d_name[PATH_MAX+1];
+};
+typedef struct dirent dirent;
+
+struct DIR {
+    struct dirent ent;
+    struct _WDIR *wdirp;
+};
+typedef struct DIR DIR;
+
+
+/* Dirent functions */
+static DIR *opendir (const char *dirname);
+static _WDIR *_wopendir (const wchar_t *dirname);
+
+static struct dirent *readdir (DIR *dirp);
+static struct _wdirent *_wreaddir (_WDIR *dirp);
+
+static int readdir_r(
+    DIR *dirp, struct dirent *entry, struct dirent **result);
+static int _wreaddir_r(
+    _WDIR *dirp, struct _wdirent *entry, struct _wdirent **result);
+
+static int closedir (DIR *dirp);
+static int _wclosedir (_WDIR *dirp);
+
+static void rewinddir (DIR* dirp);
+static void _wrewinddir (_WDIR* dirp);
+
+static int scandir (const char *dirname, struct dirent ***namelist,
+    int (*filter)(const struct dirent*),
+    int (*compare)(const struct dirent**, const struct dirent**));
+
+static int alphasort (const struct dirent **a, const struct dirent **b);
+
+static int versionsort (const struct dirent **a, const struct dirent **b);
+
+
+/* For compatibility with Symbian */
+#define wdirent _wdirent
+#define WDIR _WDIR
+#define wopendir _wopendir
+#define wreaddir _wreaddir
+#define wclosedir _wclosedir
+#define wrewinddir _wrewinddir
+
+
+/* Internal utility functions */
+static WIN32_FIND_DATAW *dirent_first (_WDIR *dirp);
+static WIN32_FIND_DATAW *dirent_next (_WDIR *dirp);
+
+static int dirent_mbstowcs_s(
+    size_t *pReturnValue,
+    wchar_t *wcstr,
+    size_t sizeInWords,
+    const char *mbstr,
+    size_t count);
+
+static int dirent_wcstombs_s(
+    size_t *pReturnValue,
+    char *mbstr,
+    size_t sizeInBytes,
+    const wchar_t *wcstr,
+    size_t count);
+
+static void dirent_set_errno (int error);
+
+
+/*
+ * Open directory stream DIRNAME for read and return a pointer to the
+ * internal working area that is used to retrieve individual directory
+ * entries.
+ */
+static _WDIR*
+_wopendir(
+    const wchar_t *dirname)
+{
+    _WDIR *dirp = NULL;
+    int error;
+
+    /* Must have directory name */
+    if (dirname == NULL  ||  dirname[0] == '\0') {
+        dirent_set_errno (ENOENT);
+        return NULL;
+    }
+
+    /* Allocate new _WDIR structure */
+    dirp = (_WDIR*) malloc (sizeof (struct _WDIR));
+    if (dirp != NULL) {
+        DWORD n;
+
+        /* Reset _WDIR structure */
+        dirp->handle = INVALID_HANDLE_VALUE;
+        dirp->patt = NULL;
+        dirp->cached = 0;
+
+        /* Compute the length of full path plus zero terminator
+         *
+         * Note that on WinRT there's no way to convert relative paths
+         * into absolute paths, so just assume it is an absolute path.
+         */
+#       if defined(WINAPI_FAMILY) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP)
+            n = wcslen(dirname);
+#       else
+            n = GetFullPathNameW (dirname, 0, NULL, NULL);
+#       endif
+
+        /* Allocate room for absolute directory name and search pattern */
+        dirp->patt = (wchar_t*) malloc (sizeof (wchar_t) * n + 16);
+        if (dirp->patt) {
+
+            /*
+             * Convert relative directory name to an absolute one.  This
+             * allows rewinddir() to function correctly even when current
+             * working directory is changed between opendir() and rewinddir().
+             *
+             * Note that on WinRT there's no way to convert relative paths
+             * into absolute paths, so just assume it is an absolute path.
+             */
+#           if defined(WINAPI_FAMILY) && (WINAPI_FAMILY == WINAPI_FAMILY_PHONE_APP)
+                wcsncpy_s(dirp->patt, n+1, dirname, n);
+#           else
+                n = GetFullPathNameW (dirname, n, dirp->patt, NULL);
+#           endif
+            if (n > 0) {
+                wchar_t *p;
+
+                /* Append search pattern \* to the directory name */
+                p = dirp->patt + n;
+                if (dirp->patt < p) {
+                    switch (p[-1]) {
+                    case '\\':
+                    case '/':
+                    case ':':
+                        /* Directory ends in path separator, e.g. c:\temp\ */
+                        /*NOP*/;
+                        break;
+
+                    default:
+                        /* Directory name doesn't end in path separator */
+                        *p++ = '\\';
+                    }
+                }
+                *p++ = '*';
+                *p = '\0';
+
+                /* Open directory stream and retrieve the first entry */
+                if (dirent_first (dirp)) {
+                    /* Directory stream opened successfully */
+                    error = 0;
+                } else {
+                    /* Cannot retrieve first entry */
+                    error = 1;
+                    dirent_set_errno (ENOENT);
+                }
+
+            } else {
+                /* Cannot retrieve full path name */
+                dirent_set_errno (ENOENT);
+                error = 1;
+            }
+
+        } else {
+            /* Cannot allocate memory for search pattern */
+            error = 1;
+        }
+
+    } else {
+        /* Cannot allocate _WDIR structure */
+        error = 1;
+    }
+
+    /* Clean up in case of error */
+    if (error  &&  dirp) {
+        _wclosedir (dirp);
+        dirp = NULL;
+    }
+
+    return dirp;
+}
+
+/*
+ * Read next directory entry.
+ *
+ * Returns pointer to static directory entry which may be overwritten by
+ * subsequent calls to _wreaddir().
+ */
+static struct _wdirent*
+_wreaddir(
+    _WDIR *dirp)
+{
+    struct _wdirent *entry;
+
+    /*
+     * Read directory entry to buffer.  We can safely ignore the return value
+     * as entry will be set to NULL in case of error.
+     */
+    (void) _wreaddir_r (dirp, &dirp->ent, &entry);
+
+    /* Return pointer to statically allocated directory entry */
+    return entry;
+}
+
+/*
+ * Read next directory entry.
+ *
+ * Returns zero on success.  If end of directory stream is reached, then sets
+ * result to NULL and returns zero.
+ */
+static int
+_wreaddir_r(
+    _WDIR *dirp,
+    struct _wdirent *entry,
+    struct _wdirent **result)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Read next directory entry */
+    datap = dirent_next (dirp);
+    if (datap) {
+        size_t n;
+        DWORD attr;
+
+        /*
+         * Copy file name as wide-character string.  If the file name is too
+         * long to fit in to the destination buffer, then truncate file name
+         * to PATH_MAX characters and zero-terminate the buffer.
+         */
+        n = 0;
+        while (n < PATH_MAX  &&  datap->cFileName[n] != 0) {
+            entry->d_name[n] = datap->cFileName[n];
+            n++;
+        }
+        entry->d_name[n] = 0;
+
+        /* Length of file name excluding zero terminator */
+        entry->d_namlen = n;
+
+        /* File type */
+        attr = datap->dwFileAttributes;
+        if ((attr & FILE_ATTRIBUTE_DEVICE) != 0) {
+            entry->d_type = DT_CHR;
+        } else if ((attr & FILE_ATTRIBUTE_DIRECTORY) != 0) {
+            entry->d_type = DT_DIR;
+        } else {
+            entry->d_type = DT_REG;
+        }
+
+        /* Reset dummy fields */
+        entry->d_ino = 0;
+        entry->d_off = 0;
+        entry->d_reclen = sizeof (struct _wdirent);
+
+        /* Set result address */
+        *result = entry;
+
+    } else {
+
+        /* Return NULL to indicate end of directory */
+        *result = NULL;
+
+    }
+
+    return /*OK*/0;
+}
+
+/*
+ * Close directory stream opened by opendir() function.  This invalidates the
+ * DIR structure as well as any directory entry read previously by
+ * _wreaddir().
+ */
+static int
+_wclosedir(
+    _WDIR *dirp)
+{
+    int ok;
+    if (dirp) {
+
+        /* Release search handle */
+        if (dirp->handle != INVALID_HANDLE_VALUE) {
+            FindClose (dirp->handle);
+            dirp->handle = INVALID_HANDLE_VALUE;
+        }
+
+        /* Release search pattern */
+        if (dirp->patt) {
+            free (dirp->patt);
+            dirp->patt = NULL;
+        }
+
+        /* Release directory structure */
+        free (dirp);
+        ok = /*success*/0;
+
+    } else {
+
+        /* Invalid directory stream */
+        dirent_set_errno (EBADF);
+        ok = /*failure*/-1;
+
+    }
+    return ok;
+}
+
+/*
+ * Rewind directory stream such that _wreaddir() returns the very first
+ * file name again.
+ */
+static void
+_wrewinddir(
+    _WDIR* dirp)
+{
+    if (dirp) {
+        /* Release existing search handle */
+        if (dirp->handle != INVALID_HANDLE_VALUE) {
+            FindClose (dirp->handle);
+        }
+
+        /* Open new search handle */
+        dirent_first (dirp);
+    }
+}
+
+/* Get first directory entry (internal) */
+static WIN32_FIND_DATAW*
+dirent_first(
+    _WDIR *dirp)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Open directory and retrieve the first entry */
+    dirp->handle = FindFirstFileExW(
+        dirp->patt, FindExInfoStandard, &dirp->data,
+        FindExSearchNameMatch, NULL, 0);
+    if (dirp->handle != INVALID_HANDLE_VALUE) {
+
+        /* a directory entry is now waiting in memory */
+        datap = &dirp->data;
+        dirp->cached = 1;
+
+    } else {
+
+        /* Failed to re-open directory: no directory entry in memory */
+        dirp->cached = 0;
+        datap = NULL;
+
+    }
+    return datap;
+}
+
+/*
+ * Get next directory entry (internal).
+ *
+ * Returns
+ */
+static WIN32_FIND_DATAW*
+dirent_next(
+    _WDIR *dirp)
+{
+    WIN32_FIND_DATAW *p;
+
+    /* Get next directory entry */
+    if (dirp->cached != 0) {
+
+        /* A valid directory entry already in memory */
+        p = &dirp->data;
+        dirp->cached = 0;
+
+    } else if (dirp->handle != INVALID_HANDLE_VALUE) {
+
+        /* Get the next directory entry from stream */
+        if (FindNextFileW (dirp->handle, &dirp->data) != FALSE) {
+            /* Got a file */
+            p = &dirp->data;
+        } else {
+            /* The very last entry has been processed or an error occurred */
+            FindClose (dirp->handle);
+            dirp->handle = INVALID_HANDLE_VALUE;
+            p = NULL;
+        }
+
+    } else {
+
+        /* End of directory stream reached */
+        p = NULL;
+
+    }
+
+    return p;
+}
+
+/*
+ * Open directory stream using plain old C-string.
+ */
+static DIR*
+opendir(
+    const char *dirname)
+{
+    struct DIR *dirp;
+    int error;
+
+    /* Must have directory name */
+    if (dirname == NULL  ||  dirname[0] == '\0') {
+        dirent_set_errno (ENOENT);
+        return NULL;
+    }
+
+    /* Allocate memory for DIR structure */
+    dirp = (DIR*) malloc (sizeof (struct DIR));
+    if (dirp) {
+        wchar_t wname[PATH_MAX + 1];
+        size_t n;
+
+        /* Convert directory name to wide-character string */
+        error = dirent_mbstowcs_s(
+            &n, wname, PATH_MAX + 1, dirname, PATH_MAX + 1);
+        if (!error) {
+
+            /* Open directory stream using wide-character name */
+            dirp->wdirp = _wopendir (wname);
+            if (dirp->wdirp) {
+                /* Directory stream opened */
+                error = 0;
+            } else {
+                /* Failed to open directory stream */
+                error = 1;
+            }
+
+        } else {
+            /*
+             * Cannot convert file name to wide-character string.  This
+             * occurs if the string contains invalid multi-byte sequences or
+             * the output buffer is too small to contain the resulting
+             * string.
+             */
+            error = 1;
+        }
+
+    } else {
+        /* Cannot allocate DIR structure */
+        error = 1;
+    }
+
+    /* Clean up in case of error */
+    if (error  &&  dirp) {
+        free (dirp);
+        dirp = NULL;
+    }
+
+    return dirp;
+}
+
+/*
+ * Read next directory entry.
+ */
+static struct dirent*
+readdir(
+    DIR *dirp)
+{
+    struct dirent *entry;
+
+    /*
+     * Read directory entry to buffer.  We can safely ignore the return value
+     * as entry will be set to NULL in case of error.
+     */
+    (void) readdir_r (dirp, &dirp->ent, &entry);
+
+    /* Return pointer to statically allocated directory entry */
+    return entry;
+}
+
+/*
+ * Read next directory entry into called-allocated buffer.
+ *
+ * Returns zero on success.  If the end of directory stream is reached, then
+ * sets result to NULL and returns zero.
+ */
+static int
+readdir_r(
+    DIR *dirp,
+    struct dirent *entry,
+    struct dirent **result)
+{
+    WIN32_FIND_DATAW *datap;
+
+    /* Read next directory entry */
+    datap = dirent_next (dirp->wdirp);
+    if (datap) {
+        size_t n;
+        int error;
+
+        /* Attempt to convert file name to multi-byte string */
+        error = dirent_wcstombs_s(
+            &n, entry->d_name, PATH_MAX + 1, datap->cFileName, PATH_MAX + 1);
+
+        /*
+         * If the file name cannot be represented by a multi-byte string,
+         * then attempt to use old 8+3 file name.  This allows traditional
+         * Unix-code to access some file names despite of unicode
+         * characters, although file names may seem unfamiliar to the user.
+         *
+         * Be ware that the code below cannot come up with a short file
+         * name unless the file system provides one.  At least
+         * VirtualBox shared folders fail to do this.
+         */
+        if (error  &&  datap->cAlternateFileName[0] != '\0') {
+            error = dirent_wcstombs_s(
+                &n, entry->d_name, PATH_MAX + 1,
+                datap->cAlternateFileName, PATH_MAX + 1);
+        }
+
+        if (!error) {
+            DWORD attr;
+
+            /* Length of file name excluding zero terminator */
+            entry->d_namlen = n - 1;
+
+            /* File attributes */
+            attr = datap->dwFileAttributes;
+            if ((attr & FILE_ATTRIBUTE_DEVICE) != 0) {
+                entry->d_type = DT_CHR;
+            } else if ((attr & FILE_ATTRIBUTE_DIRECTORY) != 0) {
+                entry->d_type = DT_DIR;
+            } else {
+                entry->d_type = DT_REG;
+            }
+
+            /* Reset dummy fields */
+            entry->d_ino = 0;
+            entry->d_off = 0;
+            entry->d_reclen = sizeof (struct dirent);
+
+        } else {
+
+            /*
+             * Cannot convert file name to multi-byte string so construct
+             * an erroneous directory entry and return that.  Note that
+             * we cannot return NULL as that would stop the processing
+             * of directory entries completely.
+             */
+            entry->d_name[0] = '?';
+            entry->d_name[1] = '\0';
+            entry->d_namlen = 1;
+            entry->d_type = DT_UNKNOWN;
+            entry->d_ino = 0;
+            entry->d_off = -1;
+            entry->d_reclen = 0;
+
+        }
+
+        /* Return pointer to directory entry */
+        *result = entry;
+
+    } else {
+
+        /* No more directory entries */
+        *result = NULL;
+
+    }
+
+    return /*OK*/0;
+}
+
+/*
+ * Close directory stream.
+ */
+static int
+closedir(
+    DIR *dirp)
+{
+    int ok;
+    if (dirp) {
+
+        /* Close wide-character directory stream */
+        ok = _wclosedir (dirp->wdirp);
+        dirp->wdirp = NULL;
+
+        /* Release multi-byte character version */
+        free (dirp);
+
+    } else {
+
+        /* Invalid directory stream */
+        dirent_set_errno (EBADF);
+        ok = /*failure*/-1;
+
+    }
+    return ok;
+}
+
+/*
+ * Rewind directory stream to beginning.
+ */
+static void
+rewinddir(
+    DIR* dirp)
+{
+    /* Rewind wide-character string directory stream */
+    _wrewinddir (dirp->wdirp);
+}
+
+/*
+ * Scan directory for entries.
+ */
+static int
+scandir(
+    const char *dirname,
+    struct dirent ***namelist,
+    int (*filter)(const struct dirent*),
+    int (*compare)(const struct dirent**, const struct dirent**))
+{
+    struct dirent **files = NULL;
+    size_t size = 0;
+    size_t allocated = 0;
+    const size_t init_size = 1;
+    DIR *dir = NULL;
+    struct dirent *entry;
+    struct dirent *tmp = NULL;
+    size_t i;
+    int result = 0;
+
+    /* Open directory stream */
+    dir = opendir (dirname);
+    if (dir) {
+
+        /* Read directory entries to memory */
+        while (1) {
+
+            /* Enlarge pointer table to make room for another pointer */
+            if (size >= allocated) {
+                void *p;
+                size_t num_entries;
+
+                /* Compute number of entries in the enlarged pointer table */
+                if (size < init_size) {
+                    /* Allocate initial pointer table */
+                    num_entries = init_size;
+                } else {
+                    /* Double the size */
+                    num_entries = size * 2;
+                }
+
+                /* Allocate first pointer table or enlarge existing table */
+                p = realloc (files, sizeof (void*) * num_entries);
+                if (p != NULL) {
+                    /* Got the memory */
+                    files = (dirent**) p;
+                    allocated = num_entries;
+                } else {
+                    /* Out of memory */
+                    result = -1;
+                    break;
+                }
+
+            }
+
+            /* Allocate room for temporary directory entry */
+            if (tmp == NULL) {
+                tmp = (struct dirent*) malloc (sizeof (struct dirent));
+                if (tmp == NULL) {
+                    /* Cannot allocate temporary directory entry */
+                    result = -1;
+                    break;
+                }
+            }
+
+            /* Read directory entry to temporary area */
+            if (readdir_r (dir, tmp, &entry) == /*OK*/0) {
+
+                /* Did we get an entry? */
+                if (entry != NULL) {
+                    int pass;
+
+                    /* Determine whether to include the entry in result */
+                    if (filter) {
+                        /* Let the filter function decide */
+                        pass = filter (tmp);
+                    } else {
+                        /* No filter function, include everything */
+                        pass = 1;
+                    }
+
+                    if (pass) {
+                        /* Store the temporary entry to pointer table */
+                        files[size++] = tmp;
+                        tmp = NULL;
+
+                        /* Keep up with the number of files */
+                        result++;
+                    }
+
+                } else {
+
+                    /*
+                     * End of directory stream reached => sort entries and
+                     * exit.
+                     */
+                    qsort (files, size, sizeof (void*),
+                        (int (*) (const void*, const void*)) compare);
+                    break;
+
+                }
+
+            } else {
+                /* Error reading directory entry */
+                result = /*Error*/ -1;
+                break;
+            }
+
+        }
+
+    } else {
+        /* Cannot open directory */
+        result = /*Error*/ -1;
+    }
+
+    /* Release temporary directory entry */
+    if (tmp) {
+        free (tmp);
+    }
+
+    /* Release allocated memory on error */
+    if (result < 0) {
+        for (i = 0; i < size; i++) {
+            free (files[i]);
+        }
+        free (files);
+        files = NULL;
+    }
+
+    /* Close directory stream */
+    if (dir) {
+        closedir (dir);
+    }
+
+    /* Pass pointer table to caller */
+    if (namelist) {
+        *namelist = files;
+    }
+    return result;
+}
+
+/* Alphabetical sorting */
+static int
+alphasort(
+    const struct dirent **a, const struct dirent **b)
+{
+    return strcoll ((*a)->d_name, (*b)->d_name);
+}
+
+/* Sort versions */
+static int
+versionsort(
+    const struct dirent **a, const struct dirent **b)
+{
+    /* FIXME: implement strverscmp and use that */
+    return alphasort (a, b);
+}
+
+/* Convert multi-byte string to wide character string */
+static int
+dirent_mbstowcs_s(
+    size_t *pReturnValue,
+    wchar_t *wcstr,
+    size_t sizeInWords,
+    const char *mbstr,
+    size_t count)
+{
+    int error;
+    int n;
+    size_t len;
+    UINT cp;
+    DWORD flags;
+
+    /* Determine code page for multi-byte string */
+    if (AreFileApisANSI ()) {
+        /* Default ANSI code page */
+        cp = GetACP ();
+    } else {
+        /* Default OEM code page */
+        cp = GetOEMCP ();
+    }
+
+    /*
+     * Determine flags based on the character set.  For more information,
+     * please see https://docs.microsoft.com/fi-fi/windows/desktop/api/stringapiset/nf-stringapiset-multibytetowidechar
+     */
+    switch (cp) {
+    case 42:
+    case 50220:
+    case 50221:
+    case 50222:
+    case 50225:
+    case 50227:
+    case 50229:
+    case 57002:
+    case 57003:
+    case 57004:
+    case 57005:
+    case 57006:
+    case 57007:
+    case 57008:
+    case 57009:
+    case 57010:
+    case 57011:
+    case 65000:
+        /* MultiByteToWideChar does not support MB_ERR_INVALID_CHARS */
+        flags = 0;
+        break;
+
+    default:
+        /*
+         * Ask MultiByteToWideChar to return an error if a multi-byte
+         * character cannot be converted to a wide-character.
+         */
+        flags = MB_ERR_INVALID_CHARS;
+    }
+
+    /* Compute the length of input string without zero-terminator */
+    len = 0;
+    while (mbstr[len] != '\0'  &&  len < count) {
+        len++;
+    }
+
+    /* Convert to wide-character string */
+    n = MultiByteToWideChar(
+        /* Source code page */ cp,
+        /* Flags */ flags,
+        /* Pointer to string to convert */ mbstr,
+        /* Size of multi-byte string */ (int) len,
+        /* Pointer to output buffer */ wcstr,
+        /* Size of output buffer */ (int)sizeInWords - 1
+    );
+
+    /* Ensure that output buffer is zero-terminated */
+    wcstr[n] = '\0';
+
+    /* Return length of wide-character string with zero-terminator */
+    *pReturnValue = (size_t) (n + 1);
+
+    /* Return zero if conversion succeeded */
+    if (n > 0) {
+        error = 0;
+    } else {
+        error = 1;
+    }
+    return error;
+}
+
+/* Convert wide-character string to multi-byte string */
+static int
+dirent_wcstombs_s(
+    size_t *pReturnValue,
+    char *mbstr,
+    size_t sizeInBytes, /* max size of mbstr */
+    const wchar_t *wcstr,
+    size_t count)
+{
+    int n;
+    int error;
+    UINT cp;
+    size_t len;
+    BOOL flag = 0;
+    LPBOOL pflag;
+
+    /* Determine code page for multi-byte string */
+    if (AreFileApisANSI ()) {
+        /* Default ANSI code page */
+        cp = GetACP ();
+    } else {
+        /* Default OEM code page */
+        cp = GetOEMCP ();
+    }
+
+    /* Compute the length of input string without zero-terminator */
+    len = 0;
+    while (wcstr[len] != '\0'  &&  len < count) {
+        len++;
+    }
+
+    /*
+     * Determine if we can ask WideCharToMultiByte to return information on
+     * broken characters.  For more information, please see
+     * https://docs.microsoft.com/en-us/windows/desktop/api/stringapiset/nf-stringapiset-widechartomultibyte
+     */
+    switch (cp) {
+    case CP_UTF7:
+    case CP_UTF8:
+        /*
+         * WideCharToMultiByte fails if we request information on default
+         * characters.  This is just a nuisance but doesn't affect the
+         * conversion: if Windows is configured to use UTF-8, then the default
+         * character should not be needed anyway.
+         */
+        pflag = NULL;
+        break;
+
+    default:
+        /*
+         * Request that WideCharToMultiByte sets the flag if it uses the
+         * default character.
+         */
+        pflag = &flag;
+    }
+
+    /* Convert wide-character string to multi-byte character string */
+    n = WideCharToMultiByte(
+        /* Target code page */ cp,
+        /* Flags */ 0,
+        /* Pointer to unicode string */ wcstr,
+        /* Length of unicode string */ (int) len,
+        /* Pointer to output buffer */ mbstr,
+        /* Size of output buffer */ (int)sizeInBytes - 1,
+        /* Default character */ NULL,
+        /* Whether default character was used or not */ pflag
+    );
+
+    /* Ensure that output buffer is zero-terminated */
+    mbstr[n] = '\0';
+
+    /* Return length of multi-byte string with zero-terminator */
+    *pReturnValue = (size_t) (n + 1);
+
+    /* Return zero if conversion succeeded without using default characters */
+    if (n > 0  &&  flag == 0) {
+        error = 0;
+    } else {
+        error = 1;
+    }
+    return error;
+}
+
+/* Set errno variable */
+static void
+dirent_set_errno(
+    int error)
+{
+#if defined(_MSC_VER)  &&  _MSC_VER >= 1400
+
+    /* Microsoft Visual Studio 2005 and later */
+    _set_errno (error);
+
+#else
+
+    /* Non-Microsoft compiler or older Microsoft compiler */
+    errno = error;
+
+#endif
+}
+
+
+#ifdef __cplusplus
+}
+#endif
+#endif /*DIRENT_H*/
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/include/getopt.h glmark2-2021.02/src/include/getopt.h
--- glmark2-2014.03+git20150611.fa71af2d/src/include/getopt.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/include/getopt.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,653 @@
+#ifndef __GETOPT_H__
+/**
+ * DISCLAIMER
+ * This file is part of the mingw-w64 runtime package.
+ *
+ * The mingw-w64 runtime package and its code is distributed in the hope that it 
+ * will be useful but WITHOUT ANY WARRANTY.  ALL WARRANTIES, EXPRESSED OR 
+ * IMPLIED ARE HEREBY DISCLAIMED.  This includes but is not limited to 
+ * warranties of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+ */
+ /*
+ * Copyright (c) 2002 Todd C. Miller <Todd.Miller@courtesan.com>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ *
+ * Sponsored in part by the Defense Advanced Research Projects
+ * Agency (DARPA) and Air Force Research Laboratory, Air Force
+ * Materiel Command, USAF, under agreement number F39502-99-1-0512.
+ */
+/*-
+ * Copyright (c) 2000 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Dieter Baron and Thomas Klausner.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#pragma warning(disable:4996)
+
+#define __GETOPT_H__
+
+/* All the headers include this file. */
+#include <crtdefs.h>
+#include <errno.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <windows.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define	REPLACE_GETOPT		/* use this getopt as the system getopt(3) */
+
+#ifdef REPLACE_GETOPT
+int	opterr = 1;		/* if error message should be printed */
+int	optind = 1;		/* index into parent argv vector */
+int	optopt = '?';		/* character checked for validity */
+#undef	optreset		/* see getopt.h */
+#define	optreset		__mingw_optreset
+int	optreset;		/* reset getopt */
+char    *optarg;		/* argument associated with option */
+#endif
+
+//extern int optind;		/* index of first non-option in argv      */
+//extern int optopt;		/* single option character, as parsed     */
+//extern int opterr;		/* flag to enable built-in diagnostics... */
+//				/* (user may set to zero, to suppress)    */
+//
+//extern char *optarg;		/* pointer to argument of current option  */
+
+#define PRINT_ERROR	((opterr) && (*options != ':'))
+
+#define FLAG_PERMUTE	0x01	/* permute non-options to the end of argv */
+#define FLAG_ALLARGS	0x02	/* treat non-options as args to option "-1" */
+#define FLAG_LONGONLY	0x04	/* operate as getopt_long_only */
+
+/* return values */
+#define	BADCH		(int)'?'
+#define	BADARG		((*options == ':') ? (int)':' : (int)'?')
+#define	INORDER 	(int)1
+
+#ifndef __CYGWIN__
+#define __progname __argv[0]
+#else
+extern char __declspec(dllimport) *__progname;
+#endif
+
+#ifdef __CYGWIN__
+static char EMSG[] = "";
+#else
+#define	EMSG		""
+#endif
+
+static int getopt_internal(int, char * const *, const char *,
+			   const struct option *, int *, int);
+static int parse_long_options(char * const *, const char *,
+			      const struct option *, int *, int);
+static int gcd(int, int);
+static void permute_args(int, int, int, char * const *);
+
+static char *place = EMSG; /* option letter processing */
+
+/* XXX: set optreset to 1 rather than these two */
+static int nonopt_start = -1; /* first non option argument (for permute) */
+static int nonopt_end = -1;   /* first option after non options (for permute) */
+
+/* Error messages */
+static const char recargchar[] = "option requires an argument -- %c";
+static const char recargstring[] = "option requires an argument -- %s";
+static const char ambig[] = "ambiguous option -- %.*s";
+static const char noarg[] = "option doesn't take an argument -- %.*s";
+static const char illoptchar[] = "unknown option -- %c";
+static const char illoptstring[] = "unknown option -- %s";
+
+static void
+_vwarnx(const char *fmt,va_list ap)
+{
+  (void)fprintf(stderr,"%s: ",__progname);
+  if (fmt != NULL)
+    (void)vfprintf(stderr,fmt,ap);
+  (void)fprintf(stderr,"\n");
+}
+
+static void
+warnx(const char *fmt,...)
+{
+  va_list ap;
+  va_start(ap,fmt);
+  _vwarnx(fmt,ap);
+  va_end(ap);
+}
+
+/*
+ * Compute the greatest common divisor of a and b.
+ */
+static int
+gcd(int a, int b)
+{
+	int c;
+
+	c = a % b;
+	while (c != 0) {
+		a = b;
+		b = c;
+		c = a % b;
+	}
+
+	return (b);
+}
+
+/*
+ * Exchange the block from nonopt_start to nonopt_end with the block
+ * from nonopt_end to opt_end (keeping the same order of arguments
+ * in each block).
+ */
+static void
+permute_args(int panonopt_start, int panonopt_end, int opt_end,
+	char * const *nargv)
+{
+	int cstart, cyclelen, i, j, ncycle, nnonopts, nopts, pos;
+	char *swap;
+
+	/*
+	 * compute lengths of blocks and number and size of cycles
+	 */
+	nnonopts = panonopt_end - panonopt_start;
+	nopts = opt_end - panonopt_end;
+	ncycle = gcd(nnonopts, nopts);
+	cyclelen = (opt_end - panonopt_start) / ncycle;
+
+	for (i = 0; i < ncycle; i++) {
+		cstart = panonopt_end+i;
+		pos = cstart;
+		for (j = 0; j < cyclelen; j++) {
+			if (pos >= panonopt_end)
+				pos -= nnonopts;
+			else
+				pos += nopts;
+			swap = nargv[pos];
+			/* LINTED const cast */
+			((char **) nargv)[pos] = nargv[cstart];
+			/* LINTED const cast */
+			((char **)nargv)[cstart] = swap;
+		}
+	}
+}
+
+#ifdef REPLACE_GETOPT
+/*
+ * getopt --
+ *	Parse argc/argv argument vector.
+ *
+ * [eventually this will replace the BSD getopt]
+ */
+int
+getopt(int nargc, char * const *nargv, const char *options)
+{
+
+	/*
+	 * We don't pass FLAG_PERMUTE to getopt_internal() since
+	 * the BSD getopt(3) (unlike GNU) has never done this.
+	 *
+	 * Furthermore, since many privileged programs call getopt()
+	 * before dropping privileges it makes sense to keep things
+	 * as simple (and bug-free) as possible.
+	 */
+	return (getopt_internal(nargc, nargv, options, NULL, NULL, 0));
+}
+#endif /* REPLACE_GETOPT */
+
+//extern int getopt(int nargc, char * const *nargv, const char *options);
+
+#ifdef _BSD_SOURCE
+/*
+ * BSD adds the non-standard `optreset' feature, for reinitialisation
+ * of `getopt' parsing.  We support this feature, for applications which
+ * proclaim their BSD heritage, before including this header; however,
+ * to maintain portability, developers are advised to avoid it.
+ */
+# define optreset  __mingw_optreset
+extern int optreset;
+#endif
+#ifdef __cplusplus
+}
+#endif
+/*
+ * POSIX requires the `getopt' API to be specified in `unistd.h';
+ * thus, `unistd.h' includes this header.  However, we do not want
+ * to expose the `getopt_long' or `getopt_long_only' APIs, when
+ * included in this manner.  Thus, close the standard __GETOPT_H__
+ * declarations block, and open an additional __GETOPT_LONG_H__
+ * specific block, only when *not* __UNISTD_H_SOURCED__, in which
+ * to declare the extended API.
+ */
+#endif /* !defined(__GETOPT_H__) */
+
+#if !defined(__UNISTD_H_SOURCED__) && !defined(__GETOPT_LONG_H__)
+#define __GETOPT_LONG_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct option		/* specification for a long form option...	*/
+{
+  const char *name;		/* option name, without leading hyphens */
+  int         has_arg;		/* does it take an argument?		*/
+  int        *flag;		/* where to save its status, or NULL	*/
+  int         val;		/* its associated status value		*/
+};
+
+enum    		/* permitted values for its `has_arg' field...	*/
+{
+  no_argument = 0,      	/* option never takes an argument	*/
+  required_argument,		/* option always requires an argument	*/
+  optional_argument		/* option may take an argument		*/
+};
+
+/*
+ * parse_long_options --
+ *	Parse long options in argc/argv argument vector.
+ * Returns -1 if short_too is set and the option does not match long_options.
+ */
+static int
+parse_long_options(char * const *nargv, const char *options,
+	const struct option *long_options, int *idx, int short_too)
+{
+	char *current_argv, *has_equal;
+	size_t current_argv_len;
+	int i, ambiguous, match;
+
+#define IDENTICAL_INTERPRETATION(_x, _y)                                \
+	(long_options[(_x)].has_arg == long_options[(_y)].has_arg &&    \
+	 long_options[(_x)].flag == long_options[(_y)].flag &&          \
+	 long_options[(_x)].val == long_options[(_y)].val)
+
+	current_argv = place;
+	match = -1;
+	ambiguous = 0;
+
+	optind++;
+
+	if ((has_equal = strchr(current_argv, '=')) != NULL) {
+		/* argument found (--option=arg) */
+		current_argv_len = has_equal - current_argv;
+		has_equal++;
+	} else
+		current_argv_len = strlen(current_argv);
+
+	for (i = 0; long_options[i].name; i++) {
+		/* find matching long option */
+		if (strncmp(current_argv, long_options[i].name,
+		    current_argv_len))
+			continue;
+
+		if (strlen(long_options[i].name) == current_argv_len) {
+			/* exact match */
+			match = i;
+			ambiguous = 0;
+			break;
+		}
+		/*
+		 * If this is a known short option, don't allow
+		 * a partial match of a single character.
+		 */
+		if (short_too && current_argv_len == 1)
+			continue;
+
+		if (match == -1)	/* partial match */
+			match = i;
+		else if (!IDENTICAL_INTERPRETATION(i, match))
+			ambiguous = 1;
+	}
+	if (ambiguous) {
+		/* ambiguous abbreviation */
+		if (PRINT_ERROR)
+			warnx(ambig, (int)current_argv_len,
+			     current_argv);
+		optopt = 0;
+		return (BADCH);
+	}
+	if (match != -1) {		/* option found */
+		if (long_options[match].has_arg == no_argument
+		    && has_equal) {
+			if (PRINT_ERROR)
+				warnx(noarg, (int)current_argv_len,
+				     current_argv);
+			/*
+			 * XXX: GNU sets optopt to val regardless of flag
+			 */
+			if (long_options[match].flag == NULL)
+				optopt = long_options[match].val;
+			else
+				optopt = 0;
+			return (BADARG);
+		}
+		if (long_options[match].has_arg == required_argument ||
+		    long_options[match].has_arg == optional_argument) {
+			if (has_equal)
+				optarg = has_equal;
+			else if (long_options[match].has_arg ==
+			    required_argument) {
+				/*
+				 * optional argument doesn't use next nargv
+				 */
+				optarg = nargv[optind++];
+			}
+		}
+		if ((long_options[match].has_arg == required_argument)
+		    && (optarg == NULL)) {
+			/*
+			 * Missing argument; leading ':' indicates no error
+			 * should be generated.
+			 */
+			if (PRINT_ERROR)
+				warnx(recargstring,
+				    current_argv);
+			/*
+			 * XXX: GNU sets optopt to val regardless of flag
+			 */
+			if (long_options[match].flag == NULL)
+				optopt = long_options[match].val;
+			else
+				optopt = 0;
+			--optind;
+			return (BADARG);
+		}
+	} else {			/* unknown option */
+		if (short_too) {
+			--optind;
+			return (-1);
+		}
+		if (PRINT_ERROR)
+			warnx(illoptstring, current_argv);
+		optopt = 0;
+		return (BADCH);
+	}
+	if (idx)
+		*idx = match;
+	if (long_options[match].flag) {
+		*long_options[match].flag = long_options[match].val;
+		return (0);
+	} else
+		return (long_options[match].val);
+#undef IDENTICAL_INTERPRETATION
+}
+
+/*
+ * getopt_internal --
+ *	Parse argc/argv argument vector.  Called by user level routines.
+ */
+static int
+getopt_internal(int nargc, char * const *nargv, const char *options,
+	const struct option *long_options, int *idx, int flags)
+{
+	char *oli;				/* option letter list index */
+	int optchar, short_too;
+	static int posixly_correct = -1;
+
+	if (options == NULL)
+		return (-1);
+
+	/*
+	 * XXX Some GNU programs (like cvs) set optind to 0 instead of
+	 * XXX using optreset.  Work around this braindamage.
+	 */
+	if (optind == 0)
+		optind = optreset = 1;
+
+	/*
+	 * Disable GNU extensions if POSIXLY_CORRECT is set or options
+	 * string begins with a '+'.
+	 *
+	 * CV, 2009-12-14: Check POSIXLY_CORRECT anew if optind == 0 or
+	 *                 optreset != 0 for GNU compatibility.
+	 */
+	if (posixly_correct == -1 || optreset != 0)
+		posixly_correct = (getenv("POSIXLY_CORRECT") != NULL);
+	if (*options == '-')
+		flags |= FLAG_ALLARGS;
+	else if (posixly_correct || *options == '+')
+		flags &= ~FLAG_PERMUTE;
+	if (*options == '+' || *options == '-')
+		options++;
+
+	optarg = NULL;
+	if (optreset)
+		nonopt_start = nonopt_end = -1;
+start:
+	if (optreset || !*place) {		/* update scanning pointer */
+		optreset = 0;
+		if (optind >= nargc) {          /* end of argument vector */
+			place = EMSG;
+			if (nonopt_end != -1) {
+				/* do permutation, if we have to */
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				optind -= nonopt_end - nonopt_start;
+			}
+			else if (nonopt_start != -1) {
+				/*
+				 * If we skipped non-options, set optind
+				 * to the first of them.
+				 */
+				optind = nonopt_start;
+			}
+			nonopt_start = nonopt_end = -1;
+			return (-1);
+		}
+		if (*(place = nargv[optind]) != '-' ||
+		    (place[1] == '\0' && strchr(options, '-') == NULL)) {
+			place = EMSG;		/* found non-option */
+			if (flags & FLAG_ALLARGS) {
+				/*
+				 * GNU extension:
+				 * return non-option as argument to option 1
+				 */
+				optarg = nargv[optind++];
+				return (INORDER);
+			}
+			if (!(flags & FLAG_PERMUTE)) {
+				/*
+				 * If no permutation wanted, stop parsing
+				 * at first non-option.
+				 */
+				return (-1);
+			}
+			/* do permutation */
+			if (nonopt_start == -1)
+				nonopt_start = optind;
+			else if (nonopt_end != -1) {
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				nonopt_start = optind -
+				    (nonopt_end - nonopt_start);
+				nonopt_end = -1;
+			}
+			optind++;
+			/* process next argument */
+			goto start;
+		}
+		if (nonopt_start != -1 && nonopt_end == -1)
+			nonopt_end = optind;
+
+		/*
+		 * If we have "-" do nothing, if "--" we are done.
+		 */
+		if (place[1] != '\0' && *++place == '-' && place[1] == '\0') {
+			optind++;
+			place = EMSG;
+			/*
+			 * We found an option (--), so if we skipped
+			 * non-options, we have to permute.
+			 */
+			if (nonopt_end != -1) {
+				permute_args(nonopt_start, nonopt_end,
+				    optind, nargv);
+				optind -= nonopt_end - nonopt_start;
+			}
+			nonopt_start = nonopt_end = -1;
+			return (-1);
+		}
+	}
+
+	/*
+	 * Check long options if:
+	 *  1) we were passed some
+	 *  2) the arg is not just "-"
+	 *  3) either the arg starts with -- we are getopt_long_only()
+	 */
+	if (long_options != NULL && place != nargv[optind] &&
+	    (*place == '-' || (flags & FLAG_LONGONLY))) {
+		short_too = 0;
+		if (*place == '-')
+			place++;		/* --foo long option */
+		else if (*place != ':' && strchr(options, *place) != NULL)
+			short_too = 1;		/* could be short option too */
+
+		optchar = parse_long_options(nargv, options, long_options,
+		    idx, short_too);
+		if (optchar != -1) {
+			place = EMSG;
+			return (optchar);
+		}
+	}
+
+	if ((optchar = (int)*place++) == (int)':' ||
+	    (optchar == (int)'-' && *place != '\0') ||
+	    (oli = (char*)strchr(options, optchar)) == NULL) {
+		/*
+		 * If the user specified "-" and  '-' isn't listed in
+		 * options, return -1 (non-option) as per POSIX.
+		 * Otherwise, it is an unknown option character (or ':').
+		 */
+		if (optchar == (int)'-' && *place == '\0')
+			return (-1);
+		if (!*place)
+			++optind;
+		if (PRINT_ERROR)
+			warnx(illoptchar, optchar);
+		optopt = optchar;
+		return (BADCH);
+	}
+	if (long_options != NULL && optchar == 'W' && oli[1] == ';') {
+		/* -W long-option */
+		if (*place)			/* no space */
+			/* NOTHING */;
+		else if (++optind >= nargc) {	/* no arg */
+			place = EMSG;
+			if (PRINT_ERROR)
+				warnx(recargchar, optchar);
+			optopt = optchar;
+			return (BADARG);
+		} else				/* white space */
+			place = nargv[optind];
+		optchar = parse_long_options(nargv, options, long_options,
+		    idx, 0);
+		place = EMSG;
+		return (optchar);
+	}
+	if (*++oli != ':') {			/* doesn't take argument */
+		if (!*place)
+			++optind;
+	} else {				/* takes (optional) argument */
+		optarg = NULL;
+		if (*place)			/* no white space */
+			optarg = place;
+		else if (oli[1] != ':') {	/* arg not optional */
+			if (++optind >= nargc) {	/* no arg */
+				place = EMSG;
+				if (PRINT_ERROR)
+					warnx(recargchar, optchar);
+				optopt = optchar;
+				return (BADARG);
+			} else
+				optarg = nargv[optind];
+		}
+		place = EMSG;
+		++optind;
+	}
+	/* dump back option letter */
+	return (optchar);
+}
+
+/*
+ * getopt_long --
+ *	Parse argc/argv argument vector.
+ */
+int
+getopt_long(int nargc, char * const *nargv, const char *options,
+    const struct option *long_options, int *idx)
+{
+
+	return (getopt_internal(nargc, nargv, options, long_options, idx,
+	    FLAG_PERMUTE));
+}
+
+/*
+ * getopt_long_only --
+ *	Parse argc/argv argument vector.
+ */
+int
+getopt_long_only(int nargc, char * const *nargv, const char *options,
+    const struct option *long_options, int *idx)
+{
+
+	return (getopt_internal(nargc, nargv, options, long_options, idx,
+	    FLAG_PERMUTE|FLAG_LONGONLY));
+}
+
+//extern int getopt_long(int nargc, char * const *nargv, const char *options,
+//    const struct option *long_options, int *idx);
+//extern int getopt_long_only(int nargc, char * const *nargv, const char *options,
+//    const struct option *long_options, int *idx);
+/*
+ * Previous MinGW implementation had...
+ */
+#ifndef HAVE_DECL_GETOPT
+/*
+ * ...for the long form API only; keep this for compatibility.
+ */
+# define HAVE_DECL_GETOPT	1
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* !defined(__UNISTD_H_SOURCED__) && !defined(__GETOPT_LONG_H__) */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/include/sys/time.h glmark2-2021.02/src/include/sys/time.h
--- glmark2-2014.03+git20150611.fa71af2d/src/include/sys/time.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/include/sys/time.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,45 @@
+//
+// Copyright © 2019 Jamie Madill
+//
+// This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+//
+// glmark2 is free software: you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the Free Software
+// Foundation, either version 3 of the License, or (at your option) any later
+// version.
+//
+// glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+// FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+// details.
+//
+// You should have received a copy of the GNU General Public License along with
+// glmark2.  If not, see <http://www.gnu.org/licenses/>.
+//
+// Authors:
+//  Jamie Madill
+//
+// From Stack Overflow under Create Commons license:
+// https://stackoverflow.com/a/31335254
+
+#ifndef SYS_TIME_H_FOR_WINDOWS_
+#define SYS_TIME_H_FOR_WINDOWS_
+
+#ifdef _WIN32
+#include <time.h>
+#include <windows.h>
+#define CLOCK_MONOTONIC 0
+inline int clock_gettime(int, struct timespec *spec)
+{
+    __int64 wintime;
+    GetSystemTimeAsFileTime((FILETIME*)&wintime);
+    wintime      -= 116444736000000000i64;           // 1601 to 1970
+    spec->tv_sec  = wintime / 10000000i64;           // seconds
+    spec->tv_nsec = wintime % 10000000i64 *100;      // nanoseconds
+    return 0;
+}
+#else
+#include <sys/time.h>
+#endif
+
+#endif  // SYS_TIME_H_FOR_WINDOWS_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/acinclude.m4 glmark2-2021.02/src/libjpeg-turbo/acinclude.m4
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/acinclude.m4	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/acinclude.m4	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,254 @@
+# AC_PROG_NASM
+# --------------------------
+# Check that NASM exists and determine flags
+AC_DEFUN([AC_PROG_NASM],[
+
+AC_ARG_VAR(NASM, [NASM command (used to build the x86/x86-64 SIMD code)])
+if test "x$NASM" = "x"; then
+  AC_CHECK_PROGS(NASM, [nasm nasmw yasm])
+  test -z "$NASM" && AC_MSG_ERROR([no nasm (Netwide Assembler) found])
+fi
+
+AC_MSG_CHECKING([for object file format of host system])
+case "$host_os" in
+  cygwin* | mingw* | pw32* | interix*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='Win64-COFF'
+        ;;
+      *)
+        objfmt='Win32-COFF'
+        ;;
+    esac
+  ;;
+  msdosdjgpp* | go32*)
+    objfmt='COFF'
+  ;;
+  os2-emx*)			# not tested
+    objfmt='MSOMF'		# obj
+  ;;
+  linux*coff* | linux*oldld*)
+    objfmt='COFF'		# ???
+  ;;
+  linux*aout*)
+    objfmt='a.out'
+  ;;
+  linux*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='ELF64'
+        ;;
+      *)
+        objfmt='ELF'
+        ;;
+    esac
+  ;;
+  kfreebsd* | freebsd* | netbsd* | openbsd*)
+    if echo __ELF__ | $CC -E - | grep __ELF__ > /dev/null; then
+      objfmt='BSD-a.out'
+    else
+      case "$host_cpu" in
+        x86_64 | amd64)
+          objfmt='ELF64'
+          ;;
+        *)
+          objfmt='ELF'
+          ;;
+      esac
+    fi
+  ;;
+  solaris* | sunos* | sysv* | sco*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='ELF64'
+        ;;
+      *)
+        objfmt='ELF'
+        ;;
+    esac
+  ;;
+  darwin* | rhapsody* | nextstep* | openstep* | macos*)
+    case "$host_cpu" in
+      x86_64)
+        objfmt='Mach-O64'
+        ;;
+      *)
+        objfmt='Mach-O'
+        ;;
+    esac
+  ;;
+  *)
+    objfmt='ELF ?'
+  ;;
+esac
+
+AC_MSG_RESULT([$objfmt])
+if test "$objfmt" = 'ELF ?'; then
+  objfmt='ELF'
+  AC_MSG_WARN([unexpected host system. assumed that the format is $objfmt.])
+fi
+
+AC_MSG_CHECKING([for object file format specifier (NAFLAGS) ])
+case "$objfmt" in
+  MSOMF)      NAFLAGS='-fobj -DOBJ32';;
+  Win32-COFF) NAFLAGS='-fwin32 -DWIN32';;
+  Win64-COFF) NAFLAGS='-fwin64 -DWIN64 -D__x86_64__';;
+  COFF)       NAFLAGS='-fcoff -DCOFF';;
+  a.out)      NAFLAGS='-faout -DAOUT';;
+  BSD-a.out)  NAFLAGS='-faoutb -DAOUT';;
+  ELF)        NAFLAGS='-felf -DELF';;
+  ELF64)      NAFLAGS='-felf64 -DELF -D__x86_64__';;
+  RDF)        NAFLAGS='-frdf -DRDF';;
+  Mach-O)     NAFLAGS='-fmacho -DMACHO';;
+  Mach-O64)   NAFLAGS='-fmacho64 -DMACHO -D__x86_64__';;
+esac
+AC_MSG_RESULT([$NAFLAGS])
+AC_SUBST([NAFLAGS])
+
+AC_MSG_CHECKING([whether the assembler ($NASM $NAFLAGS) works])
+cat > conftest.asm <<EOF
+[%line __oline__ "configure"
+        section .text
+        global  _main,main
+_main:
+main:   xor     eax,eax
+        ret
+]EOF
+try_nasm='$NASM $NAFLAGS -o conftest.o conftest.asm'
+if AC_TRY_EVAL(try_nasm) && test -s conftest.o; then
+  AC_MSG_RESULT(yes)
+else
+  echo "configure: failed program was:" >&AC_FD_CC
+  cat conftest.asm >&AC_FD_CC
+  rm -rf conftest*
+  AC_MSG_RESULT(no)
+  AC_MSG_ERROR([installation or configuration problem: assembler cannot create object files.])
+fi
+
+AC_MSG_CHECKING([whether the linker accepts assembler output])
+try_nasm='${CC-cc} -o conftest${ac_exeext} $LDFLAGS conftest.o $LIBS 1>&AC_FD_CC'
+if AC_TRY_EVAL(try_nasm) && test -s conftest${ac_exeext}; then
+  rm -rf conftest*
+  AC_MSG_RESULT(yes)
+else
+  rm -rf conftest*
+  AC_MSG_RESULT(no)
+  AC_MSG_ERROR([configuration problem: maybe object file format mismatch.])
+fi
+
+])
+
+# AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE
+# --------------------------
+# Test whether the assembler is suitable and supports NEON instructions
+AC_DEFUN([AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE],[
+  ac_good_gnu_arm_assembler=no
+  ac_save_CC="$CC"
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -x assembler-with-cpp"
+  CC="$CCAS"
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+    .text
+    .fpu neon
+    .arch armv7a
+    .object_arch armv4
+    .arm
+    pld [r0]
+    vmovn.u16 d0, q0]])], ac_good_gnu_arm_assembler=yes)
+
+  ac_use_gas_preprocessor=no
+  if test "x$ac_good_gnu_arm_assembler" = "xno" ; then
+    CC="gas-preprocessor.pl $CCAS"
+    AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+      .text
+      .fpu neon
+      .arch armv7a
+      .object_arch armv4
+      .arm
+      pld [r0]
+      vmovn.u16 d0, q0]])], ac_use_gas_preprocessor=yes)
+  fi
+  CFLAGS="$ac_save_CFLAGS"
+  CC="$ac_save_CC"
+
+  if test "x$ac_use_gas_preprocessor" = "xyes" ; then
+    CCAS="gas-preprocessor.pl $CCAS"
+    AC_SUBST([CCAS])
+    ac_good_gnu_arm_assembler=yes
+  fi
+
+  if test "x$ac_good_gnu_arm_assembler" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
+
+# AC_CHECK_COMPATIBLE_MIPSEL_ASSEMBLER_IFELSE
+# --------------------------
+# Test whether the assembler is suitable and supports MIPS instructions
+AC_DEFUN([AC_CHECK_COMPATIBLE_MIPS_ASSEMBLER_IFELSE],[
+  have_mips_dspr2=no
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -mdspr2"
+
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+
+  int main ()
+  {
+    int c = 0, a = 0, b = 0;
+    __asm__ __volatile__ (
+        "precr.qb.ph %[c], %[a], %[b]          \n\t"
+        : [c] "=r" (c)
+        : [a] "r" (a), [b] "r" (b)
+    );
+    return c;
+  }
+  ]])], have_mips_dspr2=yes)
+  CFLAGS=$ac_save_CFLAGS
+
+  if test "x$have_mips_dspr2" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
+
+AC_DEFUN([AC_CHECK_COMPATIBLE_ARM64_ASSEMBLER_IFELSE],[
+  ac_good_gnu_arm_assembler=no
+  ac_save_CC="$CC"
+  ac_save_CFLAGS="$CFLAGS"
+  CFLAGS="$CCASFLAGS -x assembler-with-cpp"
+  CC="$CCAS"
+  AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+    .text
+    MYVAR .req x0
+    movi v0.16b, #100
+    mov MYVAR, #100
+    .unreq MYVAR]])], ac_good_gnu_arm_assembler=yes)
+
+  ac_use_gas_preprocessor=no
+  if test "x$ac_good_gnu_arm_assembler" = "xno" ; then
+    CC="gas-preprocessor.pl $CCAS"
+    AC_COMPILE_IFELSE([AC_LANG_SOURCE([[
+      .text
+      MYVAR .req x0
+      movi v0.16b, #100
+      mov MYVAR, #100
+      .unreq MYVAR]])], ac_use_gas_preprocessor=yes)
+  fi
+  CFLAGS="$ac_save_CFLAGS"
+  CC="$ac_save_CC"
+
+  if test "x$ac_use_gas_preprocessor" = "xyes" ; then
+    CCAS="gas-preprocessor.pl $CCAS"
+    AC_SUBST([CCAS])
+    ac_good_gnu_arm_assembler=yes
+  fi
+
+  if test "x$ac_good_gnu_arm_assembler" = "xyes" ; then
+    $1
+  else
+    $2
+  fi
+])
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/bmp.c glmark2-2021.02/src/libjpeg-turbo/bmp.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/bmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/bmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,334 @@
+/*
+ * Copyright (C)2011, 2015 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <string.h>
+#include <setjmp.h>
+#include <errno.h>
+#include "cdjpeg.h"
+#include <jpeglib.h>
+#include <jpegint.h>
+#include "tjutil.h"
+#include "bmp.h"
+
+
+/* This duplicates the functionality of the VirtualGL bitmap library using
+   the components from cjpeg and djpeg */
+
+
+/* Error handling (based on example in example.c) */
+
+static char errStr[JMSG_LENGTH_MAX]="No error";
+
+struct my_error_mgr
+{
+	struct jpeg_error_mgr pub;
+	jmp_buf setjmp_buffer;
+};
+typedef struct my_error_mgr *my_error_ptr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	(*cinfo->err->output_message)(cinfo);
+	longjmp(myerr->setjmp_buffer, 1);
+}
+
+/* Based on output_message() in jerror.c */
+
+static void my_output_message(j_common_ptr cinfo)
+{
+	(*cinfo->err->format_message)(cinfo, errStr);
+}
+
+#define _throw(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s", m);  \
+	retval=-1;  goto bailout;}
+#define _throwunix(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s\n%s", m,  \
+	strerror(errno));  retval=-1;  goto bailout;}
+
+
+static void pixelconvert(unsigned char *srcbuf, int srcpf, int srcbottomup,
+	unsigned char *dstbuf, int dstpf, int dstbottomup, int w, int h)
+{
+	unsigned char *srcrowptr=srcbuf, *srccolptr;
+	int srcps=tjPixelSize[srcpf];
+	int srcstride=srcbottomup? -w*srcps:w*srcps;
+	unsigned char *dstrowptr=dstbuf, *dstcolptr;
+	int dstps=tjPixelSize[dstpf];
+	int dststride=dstbottomup? -w*dstps:w*dstps;
+	int row, col;
+
+	if(srcbottomup) srcrowptr=&srcbuf[w*srcps*(h-1)];
+	if(dstbottomup) dstrowptr=&dstbuf[w*dstps*(h-1)];
+
+	/* NOTE: These quick & dirty CMYK<->RGB conversion routines are for testing
+	   purposes only.  Properly converting between CMYK and RGB requires a color
+	   management system. */
+
+	if(dstpf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, srccolptr+=srcps)
+			{
+				double c=1.0-((double)(srccolptr[tjRedOffset[srcpf]])/255.);
+				double m=1.0-((double)(srccolptr[tjGreenOffset[srcpf]])/255.);
+				double y=1.0-((double)(srccolptr[tjBlueOffset[srcpf]])/255.);
+				double k=min(min(c,m),min(y,1.0));
+				if(k==1.0) c=m=y=0.0;
+				else
+				{
+					c=(c-k)/(1.0-k);
+					m=(m-k)/(1.0-k);
+					y=(y-k)/(1.0-k);
+				}
+				if(c>1.0) c=1.0;  if(c<0.) c=0.;
+				if(m>1.0) m=1.0;  if(m<0.) m=0.;
+				if(y>1.0) y=1.0;  if(y<0.) y=0.;
+				if(k>1.0) k=1.0;  if(k<0.) k=0.;
+				*dstcolptr++=(unsigned char)(255.0-c*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-m*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-y*255.0+0.5);
+				*dstcolptr++=(unsigned char)(255.0-k*255.0+0.5);
+			}
+		}
+	}
+	else if(srcpf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, dstcolptr+=dstps)
+			{
+				double c=(double)(*srccolptr++);
+				double m=(double)(*srccolptr++);
+				double y=(double)(*srccolptr++);
+				double k=(double)(*srccolptr++);
+				double r=c*k/255.;
+				double g=m*k/255.;
+				double b=y*k/255.;
+				if(r>255.0) r=255.0;  if(r<0.) r=0.;
+				if(g>255.0) g=255.0;  if(g<0.) g=0.;
+				if(b>255.0) b=255.0;  if(b<0.) b=0.;
+				dstcolptr[tjRedOffset[dstpf]]=(unsigned char)(r+0.5);
+				dstcolptr[tjGreenOffset[dstpf]]=(unsigned char)(g+0.5);
+				dstcolptr[tjBlueOffset[dstpf]]=(unsigned char)(b+0.5);
+			}
+		}
+	}
+	else
+	{
+		for(row=0; row<h; row++, srcrowptr+=srcstride, dstrowptr+=dststride)
+		{
+			for(col=0, srccolptr=srcrowptr, dstcolptr=dstrowptr;
+				col<w; col++, srccolptr+=srcps, dstcolptr+=dstps)
+			{
+				dstcolptr[tjRedOffset[dstpf]]=srccolptr[tjRedOffset[srcpf]];
+				dstcolptr[tjGreenOffset[dstpf]]=srccolptr[tjGreenOffset[srcpf]];
+				dstcolptr[tjBlueOffset[dstpf]]=srccolptr[tjBlueOffset[srcpf]];
+			}
+		}
+	}
+}
+
+
+int loadbmp(char *filename, unsigned char **buf, int *w, int *h,
+	int dstpf, int bottomup)
+{
+	int retval=0, dstps, srcpf, tempc;
+	struct jpeg_compress_struct cinfo;
+	struct my_error_mgr jerr;
+	cjpeg_source_ptr src;
+	FILE *file=NULL;
+
+	memset(&cinfo, 0, sizeof(struct jpeg_compress_struct));
+
+	if(!filename || !buf || !w || !h || dstpf<0 || dstpf>=TJ_NUMPF)
+		_throw("loadbmp(): Invalid argument");
+
+	if((file=fopen(filename, "rb"))==NULL)
+		_throwunix("loadbmp(): Cannot open input file");
+
+	cinfo.err=jpeg_std_error(&jerr.pub);
+	jerr.pub.error_exit=my_error_exit;
+	jerr.pub.output_message=my_output_message;
+
+	if(setjmp(jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;  goto bailout;
+	}
+
+	jpeg_create_compress(&cinfo);
+	if((tempc=getc(file))<0 || ungetc(tempc, file)==EOF)
+		_throwunix("loadbmp(): Could not read input file")
+	else if(tempc==EOF) _throw("loadbmp(): Input file contains no data");
+
+	if(tempc=='B')
+	{
+		if((src=jinit_read_bmp(&cinfo))==NULL)
+			_throw("loadbmp(): Could not initialize bitmap loader");
+	}
+	else if(tempc=='P')
+	{
+		if((src=jinit_read_ppm(&cinfo))==NULL)
+			_throw("loadbmp(): Could not initialize bitmap loader");
+	}
+	else _throw("loadbmp(): Unsupported file type");
+
+	src->input_file=file;
+	(*src->start_input)(&cinfo, src);
+	(*cinfo.mem->realize_virt_arrays)((j_common_ptr)&cinfo);
+
+	*w=cinfo.image_width;  *h=cinfo.image_height;
+
+	if(cinfo.input_components==1 && cinfo.in_color_space==JCS_RGB)
+		srcpf=TJPF_GRAY;
+	else srcpf=TJPF_RGB;
+
+	dstps=tjPixelSize[dstpf];
+	if((*buf=(unsigned char *)malloc((*w)*(*h)*dstps))==NULL)
+		_throw("loadbmp(): Memory allocation failure");
+
+	while(cinfo.next_scanline<cinfo.image_height)
+	{
+		int i, nlines=(*src->get_pixel_rows)(&cinfo, src);
+		for(i=0; i<nlines; i++)
+		{
+			unsigned char *outbuf;  int row;
+			row=cinfo.next_scanline+i;
+			if(bottomup) outbuf=&(*buf)[((*h)-row-1)*(*w)*dstps];
+			else outbuf=&(*buf)[row*(*w)*dstps];
+			pixelconvert(src->buffer[i], srcpf, 0, outbuf, dstpf, bottomup, *w,
+				nlines);
+		}
+		cinfo.next_scanline+=nlines;
+	}
+
+	(*src->finish_input)(&cinfo, src);
+
+	bailout:
+	jpeg_destroy_compress(&cinfo);
+	if(file) fclose(file);
+	if(retval<0 && buf && *buf) {free(*buf);  *buf=NULL;}
+	return retval;
+}
+
+
+int savebmp(char *filename, unsigned char *buf, int w, int h, int srcpf,
+	int bottomup)
+{
+	int retval=0, srcps, dstpf;
+	struct jpeg_decompress_struct dinfo;
+	struct my_error_mgr jerr;
+	djpeg_dest_ptr dst;
+	FILE *file=NULL;
+	char *ptr=NULL;
+
+	memset(&dinfo, 0, sizeof(struct jpeg_decompress_struct));
+
+	if(!filename || !buf || w<1 || h<1 || srcpf<0 || srcpf>=TJ_NUMPF)
+		_throw("savebmp(): Invalid argument");
+
+	if((file=fopen(filename, "wb"))==NULL)
+		_throwunix("savebmp(): Cannot open output file");
+
+	dinfo.err=jpeg_std_error(&jerr.pub);
+	jerr.pub.error_exit=my_error_exit;
+	jerr.pub.output_message=my_output_message;
+
+	if(setjmp(jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;  goto bailout;
+	}
+
+	jpeg_create_decompress(&dinfo);
+	if(srcpf==TJPF_GRAY)
+	{
+		dinfo.out_color_components=dinfo.output_components=1;
+		dinfo.out_color_space=JCS_GRAYSCALE;
+	}
+	else
+	{
+		dinfo.out_color_components=dinfo.output_components=3;
+		dinfo.out_color_space=JCS_RGB;
+	}
+	dinfo.image_width=w;  dinfo.image_height=h;
+	dinfo.global_state=DSTATE_READY;
+	dinfo.scale_num=dinfo.scale_denom=1;
+
+	ptr=strrchr(filename, '.');
+	if(ptr && !strcasecmp(ptr, ".bmp"))
+	{
+		if((dst=jinit_write_bmp(&dinfo, 0))==NULL)
+			_throw("savebmp(): Could not initialize bitmap writer");
+	}
+	else
+	{
+		if((dst=jinit_write_ppm(&dinfo))==NULL)
+			_throw("savebmp(): Could not initialize PPM writer");
+	}
+
+	dst->output_file=file;
+	(*dst->start_output)(&dinfo, dst);
+	(*dinfo.mem->realize_virt_arrays)((j_common_ptr)&dinfo);
+
+	if(srcpf==TJPF_GRAY) dstpf=srcpf;
+	else dstpf=TJPF_RGB;
+	srcps=tjPixelSize[srcpf];
+
+	while(dinfo.output_scanline<dinfo.output_height)
+	{
+		int i, nlines=dst->buffer_height;
+		for(i=0; i<nlines; i++)
+		{
+			unsigned char *inbuf;  int row;
+			row=dinfo.output_scanline+i;
+			if(bottomup) inbuf=&buf[(h-row-1)*w*srcps];
+			else inbuf=&buf[row*w*srcps];
+			pixelconvert(inbuf, srcpf, bottomup, dst->buffer[i], dstpf, 0, w,
+				nlines);
+		}
+		(*dst->put_pixel_rows)(&dinfo, dst, nlines);
+		dinfo.output_scanline+=nlines;
+	}
+
+	(*dst->finish_output)(&dinfo, dst);
+
+	bailout:
+	jpeg_destroy_decompress(&dinfo);
+	if(file) fclose(file);
+	return retval;
+}
+
+const char *bmpgeterr(void)
+{
+	return errStr;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/bmp.h glmark2-2021.02/src/libjpeg-turbo/bmp.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/bmp.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/bmp.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,42 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __BMP_H__
+#define __BMP_H__
+
+#include "./turbojpeg.h"
+
+int loadbmp(char *filename, unsigned char **buf, int *w, int *h, int pf,
+	int bottomup);
+
+int savebmp(char *filename, unsigned char *buf, int w, int h, int pf,
+	int bottomup);
+
+const char *bmpgeterr(void);
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cderror.h glmark2-2021.02/src/libjpeg-turbo/cderror.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cderror.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/cderror.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,136 @@
+/*
+ * cderror.h
+ *
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009 by Guido Vollbeding.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file defines the error and message codes for the cjpeg/djpeg
+ * applications.  These strings are not needed as part of the JPEG library
+ * proper.
+ * Edit this file to add new codes, or to translate the message strings to
+ * some other language.
+ */
+
+/*
+ * To define the enum list of message codes, include this file without
+ * defining macro JMESSAGE.  To create a message string table, include it
+ * again with a suitable JMESSAGE definition (see jerror.c for an example).
+ */
+#ifndef JMESSAGE
+#ifndef CDERROR_H
+#define CDERROR_H
+/* First time through, define the enum list */
+#define JMAKE_ENUM_LIST
+#else
+/* Repeated inclusions of this file are no-ops unless JMESSAGE is defined */
+#define JMESSAGE(code,string)
+#endif /* CDERROR_H */
+#endif /* JMESSAGE */
+
+#ifdef JMAKE_ENUM_LIST
+
+typedef enum {
+
+#define JMESSAGE(code,string)   code ,
+
+#endif /* JMAKE_ENUM_LIST */
+
+JMESSAGE(JMSG_FIRSTADDONCODE=1000, NULL) /* Must be first entry! */
+
+#ifdef BMP_SUPPORTED
+JMESSAGE(JERR_BMP_BADCMAP, "Unsupported BMP colormap format")
+JMESSAGE(JERR_BMP_BADDEPTH, "Only 8- and 24-bit BMP files are supported")
+JMESSAGE(JERR_BMP_BADHEADER, "Invalid BMP file: bad header length")
+JMESSAGE(JERR_BMP_BADPLANES, "Invalid BMP file: biPlanes not equal to 1")
+JMESSAGE(JERR_BMP_COLORSPACE, "BMP output must be grayscale or RGB")
+JMESSAGE(JERR_BMP_COMPRESSED, "Sorry, compressed BMPs not yet supported")
+JMESSAGE(JERR_BMP_EMPTY, "Empty BMP image")
+JMESSAGE(JERR_BMP_NOT, "Not a BMP file - does not start with BM")
+JMESSAGE(JTRC_BMP, "%ux%u 24-bit BMP image")
+JMESSAGE(JTRC_BMP_MAPPED, "%ux%u 8-bit colormapped BMP image")
+JMESSAGE(JTRC_BMP_OS2, "%ux%u 24-bit OS2 BMP image")
+JMESSAGE(JTRC_BMP_OS2_MAPPED, "%ux%u 8-bit colormapped OS2 BMP image")
+#endif /* BMP_SUPPORTED */
+
+#ifdef GIF_SUPPORTED
+JMESSAGE(JERR_GIF_BUG, "GIF output got confused")
+JMESSAGE(JERR_GIF_CODESIZE, "Bogus GIF codesize %d")
+JMESSAGE(JERR_GIF_COLORSPACE, "GIF output must be grayscale or RGB")
+JMESSAGE(JERR_GIF_IMAGENOTFOUND, "Too few images in GIF file")
+JMESSAGE(JERR_GIF_NOT, "Not a GIF file")
+JMESSAGE(JTRC_GIF, "%ux%ux%d GIF image")
+JMESSAGE(JTRC_GIF_BADVERSION,
+         "Warning: unexpected GIF version number '%c%c%c'")
+JMESSAGE(JTRC_GIF_EXTENSION, "Ignoring GIF extension block of type 0x%02x")
+JMESSAGE(JTRC_GIF_NONSQUARE, "Caution: nonsquare pixels in input")
+JMESSAGE(JWRN_GIF_BADDATA, "Corrupt data in GIF file")
+JMESSAGE(JWRN_GIF_CHAR, "Bogus char 0x%02x in GIF file, ignoring")
+JMESSAGE(JWRN_GIF_ENDCODE, "Premature end of GIF image")
+JMESSAGE(JWRN_GIF_NOMOREDATA, "Ran out of GIF bits")
+#endif /* GIF_SUPPORTED */
+
+#ifdef PPM_SUPPORTED
+JMESSAGE(JERR_PPM_COLORSPACE, "PPM output must be grayscale or RGB")
+JMESSAGE(JERR_PPM_NONNUMERIC, "Nonnumeric data in PPM file")
+JMESSAGE(JERR_PPM_TOOLARGE, "Integer value too large in PPM file")
+JMESSAGE(JERR_PPM_NOT, "Not a PPM/PGM file")
+JMESSAGE(JTRC_PGM, "%ux%u PGM image")
+JMESSAGE(JTRC_PGM_TEXT, "%ux%u text PGM image")
+JMESSAGE(JTRC_PPM, "%ux%u PPM image")
+JMESSAGE(JTRC_PPM_TEXT, "%ux%u text PPM image")
+#endif /* PPM_SUPPORTED */
+
+#ifdef RLE_SUPPORTED
+JMESSAGE(JERR_RLE_BADERROR, "Bogus error code from RLE library")
+JMESSAGE(JERR_RLE_COLORSPACE, "RLE output must be grayscale or RGB")
+JMESSAGE(JERR_RLE_DIMENSIONS, "Image dimensions (%ux%u) too large for RLE")
+JMESSAGE(JERR_RLE_EMPTY, "Empty RLE file")
+JMESSAGE(JERR_RLE_EOF, "Premature EOF in RLE header")
+JMESSAGE(JERR_RLE_MEM, "Insufficient memory for RLE header")
+JMESSAGE(JERR_RLE_NOT, "Not an RLE file")
+JMESSAGE(JERR_RLE_TOOMANYCHANNELS, "Cannot handle %d output channels for RLE")
+JMESSAGE(JERR_RLE_UNSUPPORTED, "Cannot handle this RLE setup")
+JMESSAGE(JTRC_RLE, "%ux%u full-color RLE file")
+JMESSAGE(JTRC_RLE_FULLMAP, "%ux%u full-color RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_GRAY, "%ux%u grayscale RLE file")
+JMESSAGE(JTRC_RLE_MAPGRAY, "%ux%u grayscale RLE file with map of length %d")
+JMESSAGE(JTRC_RLE_MAPPED, "%ux%u colormapped RLE file with map of length %d")
+#endif /* RLE_SUPPORTED */
+
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_TGA_BADCMAP, "Unsupported Targa colormap format")
+JMESSAGE(JERR_TGA_BADPARMS, "Invalid or unsupported Targa file")
+JMESSAGE(JERR_TGA_COLORSPACE, "Targa output must be grayscale or RGB")
+JMESSAGE(JTRC_TGA, "%ux%u RGB Targa image")
+JMESSAGE(JTRC_TGA_GRAY, "%ux%u grayscale Targa image")
+JMESSAGE(JTRC_TGA_MAPPED, "%ux%u colormapped Targa image")
+#else
+JMESSAGE(JERR_TGA_NOTCOMP, "Targa support was not compiled")
+#endif /* TARGA_SUPPORTED */
+
+JMESSAGE(JERR_BAD_CMAP_FILE,
+         "Color map file is invalid or of unsupported format")
+JMESSAGE(JERR_TOO_MANY_COLORS,
+         "Output file format cannot handle %d colormap entries")
+JMESSAGE(JERR_UNGETC_FAILED, "ungetc failed")
+#ifdef TARGA_SUPPORTED
+JMESSAGE(JERR_UNKNOWN_FORMAT,
+         "Unrecognized input file format --- perhaps you need -targa")
+#else
+JMESSAGE(JERR_UNKNOWN_FORMAT, "Unrecognized input file format")
+#endif
+JMESSAGE(JERR_UNSUPPORTED_FORMAT, "Unsupported output file format")
+
+#ifdef JMAKE_ENUM_LIST
+
+  JMSG_LASTADDONCODE
+} ADDON_MESSAGE_CODE;
+
+#undef JMAKE_ENUM_LIST
+#endif /* JMAKE_ENUM_LIST */
+
+/* Zap JMESSAGE macro so that future re-inclusions do nothing by default */
+#undef JMESSAGE
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cdjpeg.c glmark2-2021.02/src/libjpeg-turbo/cdjpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cdjpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/cdjpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,144 @@
+/*
+ * cdjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common support routines used by the IJG application
+ * programs (cjpeg, djpeg, jpegtran).
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+
+/*
+ * Optional progress monitor: display a percent-done figure on stderr.
+ */
+
+#ifdef PROGRESS_REPORT
+
+METHODDEF(void)
+progress_monitor (j_common_ptr cinfo)
+{
+  cd_progress_ptr prog = (cd_progress_ptr) cinfo->progress;
+  int total_passes = prog->pub.total_passes + prog->total_extra_passes;
+  int percent_done = (int) (prog->pub.pass_counter*100L/prog->pub.pass_limit);
+
+  if (percent_done != prog->percent_done) {
+    prog->percent_done = percent_done;
+    if (total_passes > 1) {
+      fprintf(stderr, "\rPass %d/%d: %3d%% ",
+              prog->pub.completed_passes + prog->completed_extra_passes + 1,
+              total_passes, percent_done);
+    } else {
+      fprintf(stderr, "\r %3d%% ", percent_done);
+    }
+    fflush(stderr);
+  }
+}
+
+
+GLOBAL(void)
+start_progress_monitor (j_common_ptr cinfo, cd_progress_ptr progress)
+{
+  /* Enable progress display, unless trace output is on */
+  if (cinfo->err->trace_level == 0) {
+    progress->pub.progress_monitor = progress_monitor;
+    progress->completed_extra_passes = 0;
+    progress->total_extra_passes = 0;
+    progress->percent_done = -1;
+    cinfo->progress = &progress->pub;
+  }
+}
+
+
+GLOBAL(void)
+end_progress_monitor (j_common_ptr cinfo)
+{
+  /* Clear away progress display */
+  if (cinfo->err->trace_level == 0) {
+    fprintf(stderr, "\r                \r");
+    fflush(stderr);
+  }
+}
+
+#endif
+
+
+/*
+ * Case-insensitive matching of possibly-abbreviated keyword switches.
+ * keyword is the constant keyword (must be lower case already),
+ * minchars is length of minimum legal abbreviation.
+ */
+
+GLOBAL(boolean)
+keymatch (char *arg, const char *keyword, int minchars)
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return FALSE;             /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return FALSE;             /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return FALSE;
+  return TRUE;                  /* A-OK */
+}
+
+
+/*
+ * Routines to establish binary I/O mode for stdin and stdout.
+ * Non-Unix systems often require some hacking to get out of text mode.
+ */
+
+GLOBAL(FILE *)
+read_stdin (void)
+{
+  FILE * input_file = stdin;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((input_file = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdin\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return input_file;
+}
+
+
+GLOBAL(FILE *)
+write_stdout (void)
+{
+  FILE * output_file = stdout;
+
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((output_file = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "Cannot reopen stdout\n");
+    exit(EXIT_FAILURE);
+  }
+#endif
+  return output_file;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cdjpeg.h glmark2-2021.02/src/libjpeg-turbo/cdjpeg.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cdjpeg.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/cdjpeg.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,145 @@
+/*
+ * cdjpeg.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains common declarations for the sample applications
+ * cjpeg and djpeg.  It is NOT used by the core JPEG library.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* define proper options in jconfig.h */
+#define JPEG_INTERNAL_OPTIONS   /* cjpeg.c,djpeg.c need to see xxx_SUPPORTED */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"             /* get library error codes too */
+#include "cderror.h"            /* get application-specific error codes */
+
+
+/*
+ * Object interface for cjpeg's source file decoding modules
+ */
+
+typedef struct cjpeg_source_struct *cjpeg_source_ptr;
+
+struct cjpeg_source_struct {
+  void (*start_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+  void (*finish_input) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+
+  FILE *input_file;
+
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * Object interface for djpeg's output file encoding modules
+ */
+
+typedef struct djpeg_dest_struct *djpeg_dest_ptr;
+
+struct djpeg_dest_struct {
+  /* start_output is called after jpeg_start_decompress finishes.
+   * The color map will be ready at this time, if one is needed.
+   */
+  void (*start_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+  /* Emit the specified number of pixel rows from the buffer. */
+  void (*put_pixel_rows) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                          JDIMENSION rows_supplied);
+  /* Finish up at the end of the image. */
+  void (*finish_output) (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo);
+
+  /* Target file spec; filled in by djpeg.c after object is created. */
+  FILE *output_file;
+
+  /* Output pixel-row buffer.  Created by module init or start_output.
+   * Width is cinfo->output_width * cinfo->output_components;
+   * height is buffer_height.
+   */
+  JSAMPARRAY buffer;
+  JDIMENSION buffer_height;
+};
+
+
+/*
+ * cjpeg/djpeg may need to perform extra passes to convert to or from
+ * the source/destination file format.  The JPEG library does not know
+ * about these passes, but we'd like them to be counted by the progress
+ * monitor.  We use an expanded progress monitor object to hold the
+ * additional pass count.
+ */
+
+struct cdjpeg_progress_mgr {
+  struct jpeg_progress_mgr pub; /* fields known to JPEG library */
+  int completed_extra_passes;   /* extra passes completed */
+  int total_extra_passes;       /* total extra */
+  /* last printed percentage stored here to avoid multiple printouts */
+  int percent_done;
+};
+
+typedef struct cdjpeg_progress_mgr *cd_progress_ptr;
+
+
+/* Module selection routines for I/O modules. */
+
+EXTERN(cjpeg_source_ptr) jinit_read_bmp (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_bmp (j_decompress_ptr cinfo,
+                                        boolean is_os2);
+EXTERN(cjpeg_source_ptr) jinit_read_gif (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_gif (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_ppm (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_ppm (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_rle (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_rle (j_decompress_ptr cinfo);
+EXTERN(cjpeg_source_ptr) jinit_read_targa (j_compress_ptr cinfo);
+EXTERN(djpeg_dest_ptr) jinit_write_targa (j_decompress_ptr cinfo);
+
+/* cjpeg support routines (in rdswitch.c) */
+
+EXTERN(boolean) read_quant_tables (j_compress_ptr cinfo, char *filename,
+                                   boolean force_baseline);
+EXTERN(boolean) read_scan_script (j_compress_ptr cinfo, char *filename);
+EXTERN(boolean) set_quality_ratings (j_compress_ptr cinfo, char *arg,
+                                     boolean force_baseline);
+EXTERN(boolean) set_quant_slots (j_compress_ptr cinfo, char *arg);
+EXTERN(boolean) set_sample_factors (j_compress_ptr cinfo, char *arg);
+
+/* djpeg support routines (in rdcolmap.c) */
+
+EXTERN(void) read_color_map (j_decompress_ptr cinfo, FILE *infile);
+
+/* common support routines (in cdjpeg.c) */
+
+EXTERN(void) enable_signal_catcher (j_common_ptr cinfo);
+EXTERN(void) start_progress_monitor (j_common_ptr cinfo,
+                                     cd_progress_ptr progress);
+EXTERN(void) end_progress_monitor (j_common_ptr cinfo);
+EXTERN(boolean) keymatch (char *arg, const char *keyword, int minchars);
+EXTERN(FILE *) read_stdin (void);
+EXTERN(FILE *) write_stdout (void);
+
+/* miscellaneous useful macros */
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+#ifndef EXIT_WARNING
+#define EXIT_WARNING  2
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/change.log glmark2-2021.02/src/libjpeg-turbo/change.log
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/change.log	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/change.log	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,313 @@
+NOTE:  This file was modified by The libjpeg-turbo Project to include only
+information relevant to libjpeg-turbo.
+
+CHANGE LOG for Independent JPEG Group's JPEG software
+
+
+Version 9b  17-Jan-2016
+-----------------------
+
+Document 'f' specifier for jpegtran -crop specification.
+Thank to Michele Martone for suggestion.
+
+
+Version 9  13-Jan-2013
+----------------------
+
+Add remark for jpeg_mem_dest() in jdatadst.c.
+Thank to Elie-Gregoire Khoury for the hint.
+
+Correct argument type in format string, avoid compiler warnings.
+Thank to Vincent Torri for hint.
+
+
+Version 8d  15-Jan-2012
+-----------------------
+
+Add cjpeg -rgb option to create RGB JPEG files.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+Thank to Michael Koch for the initial suggestion.
+
+Add option to disable the region adjustment in the transupp crop code.
+Thank to Jeffrey Friedl for the suggestion.
+
+
+Version 8b  16-May-2010
+-----------------------
+
+Repair problem in new memory source manager with corrupt JPEG data.
+Thank to Ted Campbell and Samuel Chun for the report.
+
+
+Version 8a  28-Feb-2010
+-----------------------
+
+Writing tables-only datastreams via jpeg_write_tables works again.
+
+Support 32-bit BMPs (RGB image with Alpha channel) for read in cjpeg.
+Thank to Brett Blackham for the suggestion.
+
+
+Version 8  10-Jan-2010
+----------------------
+
+Add sanity check in BMP reader module to avoid cjpeg crash for empty input
+image (thank to Isaev Ildar of ISP RAS, Moscow, RU for reporting this error).
+
+Add data source and destination managers for read from and write to
+memory buffers.  New API functions jpeg_mem_src and jpeg_mem_dest.
+Thank to Roberto Boni from Italy for the suggestion.
+
+
+Version 7  27-Jun-2009
+----------------------
+
+New scaled DCTs implemented.
+djpeg now supports scalings N/8 with all N from 1 to 16.
+
+cjpeg -quality option has been extended for support of separate quality
+settings for luminance and chrominance (or in general, for every provided
+quantization table slot).
+New API function jpeg_default_qtables() and q_scale_factor array in library.
+
+Support arithmetic entropy encoding and decoding.
+Added files jaricom.c, jcarith.c, jdarith.c.
+
+jpegtran has a new "lossless" cropping feature.
+
+Implement -perfect option in jpegtran, new API function
+jtransform_perfect_transform() in transupp. (DP 204_perfect.dpatch)
+
+Better error messages for jpegtran fopen failure.
+(DP 203_jpegtran_errmsg.dpatch)
+
+Fix byte order issue with 16bit PPM/PGM files in rdppm.c/wrppm.c:
+according to Netpbm, the de facto standard implementation of the PNM formats,
+the most significant byte is first. (DP 203_rdppm.dpatch)
+
+Add -raw option to rdjpgcom not to mangle the output.
+(DP 205_rdjpgcom_raw.dpatch)
+
+Make rdjpgcom locale aware. (DP 201_rdjpgcom_locale.dpatch)
+
+Add extern "C" to jpeglib.h.
+This avoids the need to put extern "C" { ... } around #include "jpeglib.h"
+in your C++ application.  Defining the symbol DONT_USE_EXTERN_C in the
+configuration prevents this. (DP 202_jpeglib.h_c++.dpatch)
+
+
+Version 6b  27-Mar-1998
+-----------------------
+
+jpegtran has new features for lossless image transformations (rotation
+and flipping) as well as "lossless" reduction to grayscale.
+
+jpegtran now copies comments by default; it has a -copy switch to enable
+copying all APPn blocks as well, or to suppress comments.  (Formerly it
+always suppressed comments and APPn blocks.)  jpegtran now also preserves
+JFIF version and resolution information.
+
+New decompressor library feature: COM and APPn markers found in the input
+file can be saved in memory for later use by the application.  (Before,
+you had to code this up yourself with a custom marker processor.)
+
+There is an unused field "void * client_data" now in compress and decompress
+parameter structs; this may be useful in some applications.
+
+JFIF version number information is now saved by the decoder and accepted by
+the encoder.  jpegtran uses this to copy the source file's version number,
+to ensure "jpegtran -copy all" won't create bogus files that contain JFXX
+extensions but claim to be version 1.01.  Applications that generate their
+own JFXX extension markers also (finally) have a supported way to cause the
+encoder to emit JFIF version number 1.02.
+
+djpeg's trace mode reports JFIF 1.02 thumbnail images as such, rather
+than as unknown APP0 markers.
+
+In -verbose mode, djpeg and rdjpgcom will try to print the contents of
+APP12 markers as text.  Some digital cameras store useful text information
+in APP12 markers.
+
+Handling of truncated data streams is more robust: blocks beyond the one in
+which the error occurs will be output as uniform gray, or left unchanged
+if decoding a progressive JPEG.  The appearance no longer depends on the
+Huffman tables being used.
+
+Huffman tables are checked for validity much more carefully than before.
+
+To avoid the Unisys LZW patent, djpeg's GIF output capability has been
+changed to produce "uncompressed GIFs", and cjpeg's GIF input capability
+has been removed altogether.  We're not happy about it either, but there
+seems to be no good alternative.
+
+The configure script now supports building libjpeg as a shared library
+on many flavors of Unix (all the ones that GNU libtool knows how to
+build shared libraries for).  Use "./configure --enable-shared" to
+try this out.
+
+New jconfig file and makefiles for Microsoft Visual C++ and Developer Studio.
+Also, a jconfig file and a build script for Metrowerks CodeWarrior
+on Apple Macintosh.  makefile.dj has been updated for DJGPP v2, and there
+are miscellaneous other minor improvements in the makefiles.
+
+jmemmac.c now knows how to create temporary files following Mac System 7
+conventions.
+
+djpeg's -map switch is now able to read raw-format PPM files reliably.
+
+cjpeg -progressive -restart no longer generates any unnecessary DRI markers.
+
+Multiple calls to jpeg_simple_progression for a single JPEG object
+no longer leak memory.
+
+
+Version 6a  7-Feb-96
+--------------------
+
+Library initialization sequence modified to detect version mismatches
+and struct field packing mismatches between library and calling application.
+This change requires applications to be recompiled, but does not require
+any application source code change.
+
+All routine declarations changed to the style "GLOBAL(type) name ...",
+that is, GLOBAL, LOCAL, METHODDEF, EXTERN are now macros taking the
+routine's return type as an argument.  This makes it possible to add
+Microsoft-style linkage keywords to all the routines by changing just
+these macros.  Note that any application code that was using these macros
+will have to be changed.
+
+DCT coefficient quantization tables are now stored in normal array order
+rather than zigzag order.  Application code that calls jpeg_add_quant_table,
+or otherwise manipulates quantization tables directly, will need to be
+changed.  If you need to make such code work with either older or newer
+versions of the library, a test like "#if JPEG_LIB_VERSION >= 61" is
+recommended.
+
+djpeg's trace capability now dumps DQT tables in natural order, not zigzag
+order.  This allows the trace output to be made into a "-qtables" file
+more easily.
+
+New system-dependent memory manager module for use on Apple Macintosh.
+
+Fix bug in cjpeg's -smooth option: last one or two scanlines would be
+duplicates of the prior line unless the image height mod 16 was 1 or 2.
+
+Repair minor problems in VMS, BCC, MC6 makefiles.
+
+New configure script based on latest GNU Autoconf.
+
+Correct the list of include files needed by MetroWerks C for ccommand().
+
+Numerous small documentation updates.
+
+
+Version 6  2-Aug-95
+-------------------
+
+Progressive JPEG support: library can read and write full progressive JPEG
+files.  A "buffered image" mode supports incremental decoding for on-the-fly
+display of progressive images.  Simply recompiling an existing IJG-v5-based
+decoder with v6 should allow it to read progressive files, though of course
+without any special progressive display.
+
+New "jpegtran" application performs lossless transcoding between different
+JPEG formats; primarily, it can be used to convert baseline to progressive
+JPEG and vice versa.  In support of jpegtran, the library now allows lossless
+reading and writing of JPEG files as DCT coefficient arrays.  This ability
+may be of use in other applications.
+
+Notes for programmers:
+* We changed jpeg_start_decompress() to be able to suspend; this makes all
+decoding modes available to suspending-input applications.  However,
+existing applications that use suspending input will need to be changed
+to check the return value from jpeg_start_decompress().  You don't need to
+do anything if you don't use a suspending data source.
+* We changed the interface to the virtual array routines: access_virt_array
+routines now take a count of the number of rows to access this time.  The
+last parameter to request_virt_array routines is now interpreted as the
+maximum number of rows that may be accessed at once, but not necessarily
+the height of every access.
+
+
+Version 5b  15-Mar-95
+---------------------
+
+Correct bugs with grayscale images having v_samp_factor > 1.
+
+jpeg_write_raw_data() now supports output suspension.
+
+Correct bugs in "configure" script for case of compiling in
+a directory other than the one containing the source files.
+
+Repair bug in jquant1.c: sometimes didn't use as many colors as it could.
+
+Borland C makefile and jconfig file work under either MS-DOS or OS/2.
+
+Miscellaneous improvements to documentation.
+
+
+Version 5a  7-Dec-94
+--------------------
+
+Changed color conversion roundoff behavior so that grayscale values are
+represented exactly.  (This causes test image files to change.)
+
+Make ordered dither use 16x16 instead of 4x4 pattern for a small quality
+improvement.
+
+New configure script based on latest GNU Autoconf.
+Fix configure script to handle CFLAGS correctly.
+Rename *.auto files to *.cfg, so that configure script still works if
+file names have been truncated for DOS.
+
+Fix bug in rdbmp.c: didn't allow for extra data between header and image.
+
+Modify rdppm.c/wrppm.c to handle 2-byte raw PPM/PGM formats for 12-bit data.
+
+Fix several bugs in rdrle.c.
+
+NEED_SHORT_EXTERNAL_NAMES option was broken.
+
+Revise jerror.h/jerror.c for more flexibility in message table.
+
+Repair oversight in jmemname.c NO_MKTEMP case: file could be there
+but unreadable.
+
+
+Version 5  24-Sep-94
+--------------------
+
+Version 5 represents a nearly complete redesign and rewrite of the IJG
+software.  Major user-visible changes include:
+  * Automatic configuration simplifies installation for most Unix systems.
+  * A range of speed vs. image quality tradeoffs are supported.
+    This includes resizing of an image during decompression: scaling down
+    by a factor of 1/2, 1/4, or 1/8 is handled very efficiently.
+  * New programs rdjpgcom and wrjpgcom allow insertion and extraction
+    of text comments in a JPEG file.
+
+The application programmer's interface to the library has changed completely.
+Notable improvements include:
+  * We have eliminated the use of callback routines for handling the
+    uncompressed image data.  The application now sees the library as a
+    set of routines that it calls to read or write image data on a
+    scanline-by-scanline basis.
+  * The application image data is represented in a conventional interleaved-
+    pixel format, rather than as a separate array for each color channel.
+    This can save a copying step in many programs.
+  * The handling of compressed data has been cleaned up: the application can
+    supply routines to source or sink the compressed data.  It is possible to
+    suspend processing on source/sink buffer overrun, although this is not
+    supported in all operating modes.
+  * All static state has been eliminated from the library, so that multiple
+    instances of compression or decompression can be active concurrently.
+  * JPEG abbreviated datastream formats are supported, ie, quantization and
+    Huffman tables can be stored separately from the image data.
+  * And not only that, but the documentation of the library has improved
+    considerably!
+
+
+The last widely used release before the version 5 rewrite was version 4A of
+18-Feb-93.  Change logs before that point have been discarded, since they
+are not of much interest after the rewrite.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/ChangeLog.txt glmark2-2021.02/src/libjpeg-turbo/ChangeLog.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/ChangeLog.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/ChangeLog.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,861 @@
+1.5 beta1
+=========
+
+[1] Added full SIMD acceleration for PowerPC platforms using AltiVec VMX
+(128-bit SIMD) instructions.  Although the performance of libjpeg-turbo on
+PowerPC was already good, due to the increased number of registers available
+to the compiler vs. x86, it was still possible to speed up compression by about
+3-4x and decompression by about 2-2.5x (relative to libjpeg v6b) through the
+use of AltiVec instructions.
+
+[2] Added two new libjpeg API functions (jpeg_skip_scanlines() and
+jpeg_crop_scanline()) that can be used to partially decode a JPEG image.  See
+libjpeg.txt for more details.
+
+[3] The TJCompressor and TJDecompressor classes in the TurboJPEG Java API now
+implement the Closeable interface, so those classes can be used with a
+try-with-resources statement.
+
+[4] The TurboJPEG Java classes now throw unchecked idiomatic exceptions
+(IllegalArgumentException, IllegalStateException) for unrecoverable errors
+caused by incorrect API usage, and those classes throw a new checked exception
+type (TJException) for errors that are passed through from the C library.
+
+[5] Source buffers for the TurboJPEG C API functions, as well as the
+jpeg_mem_src() function in the libjpeg API, are now declared as const pointers.
+This facilitates passing read-only buffers to those functions and ensures the
+caller that the source buffer will not be modified.  This should not create any
+backward API or ABI incompatibilities with prior libjpeg-turbo releases.
+
+[6] The MIPS DSPr2 SIMD code can now be compiled to support either FR=0 or FR=1
+FPUs.
+
+[7] Fixed additional negative left shifts and other issues reported by the GCC
+and Clang undefined behavior sanitizers.  Most of these issues affected only
+32-bit code, and none of them was known to pose a security threat, but removing
+the warnings makes it easier to detect actual security issues, should they
+arise in the future.
+
+[8] Removed the unnecessary .arch directive from the ARM64 NEON SIMD code.
+This directive was preventing the code from assembling using the clang
+integrated assembler.
+
+[9] Fixed a regression caused by 1.4.1[6] that prevented 32-bit and 64-bit
+libjpeg-turbo RPMs from being installed simultaneously on recent Red Hat/Fedora
+distributions.  This was due to the addition of a macro in jconfig.h that
+allows the Huffman codec to determine the word size at compile time.  Since
+that macro differs between 32-bit and 64-bit builds, this caused a conflict
+between the i386 and x86_64 RPMs (any differing files, other than executables,
+are not allowed when 32-bit and 64-bit RPMs are installed simultaneously.)
+Since the macro is used only internally, it has been moved into jconfigint.h.
+
+[10] The x86-64 SIMD code can now be disabled at run time by setting the
+JSIMD_FORCENONE environment variable to 1 (the other SIMD implementations
+already had this capability.)
+
+[11] Added a new command-line argument to TJBench (-nowrite) that prevents the
+benchmark from outputting any images.  This removes any potential operating
+system overhead that might be caused by lazy writes to disk and thus improves
+the consistency of the performance measurements.
+
+[12] Added SIMD acceleration for Huffman encoding on SSE2-capable x86 and
+x86-64 platforms.  This speeds up the compression of full-color JPEGs by about
+10-15% on average (relative to libjpeg-turbo 1.4.x) when using modern Intel and
+AMD CPUs.  Additionally, this works around an issue in the clang optimizer that
+prevents it (as of this writing) from achieving the same performance as GCC
+when compiling the C version of the Huffman encoder
+(https://llvm.org/bugs/show_bug.cgi?id=16035). For the purposes of benchmarking
+or regression testing, SIMD-accelerated Huffman encoding can be disabled by
+setting the JSIMD_NOHUFFENC environment variable to 1.
+
+[13] Added ARM 64-bit (ARMv8) NEON SIMD implementations of the commonly-used
+compression algorithms (including the slow integer forward DCT and h2v2 & h2v1
+downsampling algorithms, which are not accelerated in the 32-bit NEON
+implementation.)  This speeds up the compression of full-color JPEGs by about
+75% on average on a Cavium ThunderX processor and by about 2-2.5x on average on
+Cortex-A53 and Cortex-A57 cores.
+
+[14] Added SIMD acceleration for Huffman encoding on NEON-capable ARM 32-bit
+and 64-bit platforms.
+
+For 32-bit code, this speeds up the compression of full-color JPEGs by about
+30% on average on a typical iOS device (iPhone 4S, Cortex-A9) and by about 6-7%
+on average on a typical Android device (Nexus 5X, Cortex-A53 and Cortex-A57),
+relative to libjpeg-turbo 1.4.x.  Note that the larger speedup under iOS is due
+to the fact that iOS builds use LLVM, which does not optimize the C Huffman
+encoder as well as GCC does.
+
+For 64-bit code, NEON-accelerated Huffman encoding speeds up the compression of
+full-color JPEGs by about 40% on average on a typical iOS device (iPhone 5S,
+Apple A7) and by about 7-8% on average on a typical Android device (Nexus 5X,
+Cortex-A53 and Cortex-A57), in addition to the speedup described in [13] above.
+
+For the purposes of benchmarking or regression testing, SIMD-accelerated
+Huffman encoding can be disabled by setting the JSIMD_NOHUFFENC environment
+variable to 1.
+
+[15] pkg-config (.pc) scripts are now included for both the libjpeg and
+TurboJPEG API libraries on Un*x systems.  Note that if a project's build system
+relies on these scripts, then it will not be possible to build that project
+with libjpeg or with a prior version of libjpeg-turbo.
+
+[16] Optimized the ARM 64-bit (ARMv8) NEON SIMD decompression routines to
+improve performance on CPUs with in-order pipelines.  This speeds up the
+decompression of full-color JPEGs by nearly 2x on average on a Cavium ThunderX
+processor and by about 15% on average on a Cortex-A53 core.
+
+[17] Fixed an issue in the accelerated Huffman decoder that could have caused
+the decoder to read past the end of the input buffer when a malformed,
+specially-crafted JPEG image was being decompressed.  In prior versions of
+libjpeg-turbo, the accelerated Huffman decoder was invoked (in most cases) only
+if there were > 128 bytes of data in the input buffer.  However, it is possible
+to construct a JPEG image in which a single Huffman block is over 430 bytes
+long, so this version of libjpeg-turbo activates the accelerated Huffman
+decoder only if there are > 512 bytes of data in the input buffer.
+
+[18] Fixed a memory leak in tjunittest encountered when running the program
+with the -yuv option.
+
+
+1.4.2
+=====
+
+[1] Fixed an issue whereby cjpeg would segfault if a Windows bitmap with a
+negative width or height was used as an input image (Windows bitmaps can have
+a negative height if they are stored in top-down order, but such files are
+rare and not supported by libjpeg-turbo.)
+
+[2] Fixed an issue whereby, under certain circumstances, libjpeg-turbo would
+incorrectly encode certain JPEG images when quality=100 and the fast integer
+forward DCT were used.  This was known to cause 'make test' to fail when the
+library was built with '-march=haswell' on x86 systems.
+
+[3] Fixed an issue whereby libjpeg-turbo would crash when built with the latest
+& greatest development version of the Clang/LLVM compiler.  This was caused by
+an x86-64 ABI conformance issue in some of libjpeg-turbo's 64-bit SSE2 SIMD
+routines.  Those routines were incorrectly using a 64-bit mov instruction to
+transfer a 32-bit JDIMENSION argument, whereas the x86-64 ABI allows the upper
+(unused) 32 bits of a 32-bit argument's register to be undefined.  The new
+Clang/LLVM optimizer uses load combining to transfer multiple adjacent 32-bit
+structure members into a single 64-bit register, and this exposed the ABI
+conformance issue.
+
+[4] Fixed a bug in the MIPS DSPr2 4:2:0 "plain" (non-fancy and non-merged)
+upsampling routine that caused a buffer overflow (and subsequent segfault) when
+decompressing a 4:2:0 JPEG image whose scaled output width was less than 16
+pixels.  The "plain" upsampling routines are normally only used when
+decompressing a non-YCbCr JPEG image, but they are also used when decompressing
+a JPEG image whose scaled output height is 1.
+
+[5] Fixed various negative left shifts and other issues reported by the GCC and
+Clang undefined behavior sanitizers.  None of these was known to pose a
+security threat, but removing the warnings makes it easier to detect actual
+security issues, should they arise in the future.
+
+
+1.4.1
+=====
+
+[1] tjbench now properly handles CMYK/YCCK JPEG files.  Passing an argument of
+-cmyk (instead of, for instance, -rgb) will cause tjbench to internally convert
+the source bitmap to CMYK prior to compression, to generate YCCK JPEG files,
+and to internally convert the decompressed CMYK pixels back to RGB after
+decompression (the latter is done automatically if a CMYK or YCCK JPEG is
+passed to tjbench as a source image.)  The CMYK<->RGB conversion operation is
+not benchmarked.  NOTE: The quick & dirty CMYK<->RGB conversions that tjbench
+uses are suitable for testing only.  Proper conversion between CMYK and RGB
+requires a color management system.
+
+[2] 'make test' now performs additional bitwise regression tests using tjbench,
+mainly for the purpose of testing compression from/decompression to a subregion
+of a larger image buffer.
+
+[3] 'make test' no longer tests the regression of the floating point DCT/IDCT
+by default, since the results of those tests can vary if the algorithms in
+question are not implemented using SIMD instructions on a particular platform.
+See the comments in Makefile.am for information on how to re-enable the tests
+and to specify an expected result for them based on the particulars of your
+platform.
+
+[4] The NULL color conversion routines have been significantly optimized,
+which speeds up the compression of RGB and CMYK JPEGs by 5-20% when using
+64-bit code and 0-3% when using 32-bit code, and the decompression of those
+images by 10-30% when using 64-bit code and 3-12% when using 32-bit code.
+
+[5] Fixed an "illegal instruction" error that occurred when djpeg from a
+SIMD-enabled libjpeg-turbo MIPS build was executed with the -nosmooth option on
+a MIPS machine that lacked DSPr2 support.  The MIPS SIMD routines for h2v1 and
+h2v2 merged upsampling were not properly checking for the existence of DSPr2.
+
+[6] Performance has been improved significantly on 64-bit non-Linux and
+non-Windows platforms (generally 10-20% faster compression and 5-10% faster
+decompression.)  Due to an oversight, the 64-bit version of the accelerated
+Huffman codec was not being compiled in when libjpeg-turbo was built on
+platforms other than Windows or Linux.  Oops.
+
+[7] Fixed an extremely rare bug in the Huffman encoder that caused 64-bit
+builds of libjpeg-turbo to incorrectly encode a few specific test images when
+quality=98, an optimized Huffman table, and the slow integer forward DCT were
+used.
+
+[8] The Windows (CMake) build system now supports building only static or only
+shared libraries.  This is accomplished by adding either -DENABLE_STATIC=0 or
+-DENABLE_SHARED=0 to the CMake command line.
+
+[9] TurboJPEG API functions will now return an error code if a warning is
+triggered in the underlying libjpeg API.  For instance, if a JPEG file is
+corrupt, the TurboJPEG decompression functions will attempt to decompress
+as much of the image as possible, but those functions will now return -1 to
+indicate that the decompression was not entirely successful.
+
+[10] Fixed a bug in the MIPS DSPr2 4:2:2 fancy upsampling routine that caused a
+buffer overflow (and subsequent segfault) when decompressing a 4:2:2 JPEG image
+in which the right-most MCU was 5 or 6 pixels wide.
+
+
+1.4.0
+=====
+
+[1] Fixed a build issue on OS X PowerPC platforms (md5cmp failed to build
+because OS X does not provide the le32toh() and htole32() functions.)
+
+[2] The non-SIMD RGB565 color conversion code did not work correctly on big
+endian machines.  This has been fixed.
+
+[3] Fixed an issue in tjPlaneSizeYUV() whereby it would erroneously return 1
+instead of -1 if componentID was > 0 and subsamp was TJSAMP_GRAY.
+
+[3] Fixed an issue in tjBufSizeYUV2() whereby it would erroneously return 0
+instead of -1 if width was < 1.
+
+[5] The Huffman encoder now uses clz and bsr instructions for bit counting on
+ARM64 platforms (see 1.4 beta1 [5].)
+
+[6] The close() method in the TJCompressor and TJDecompressor Java classes is
+now idempotent.  Previously, that method would call the native tjDestroy()
+function even if the TurboJPEG instance had already been destroyed.  This
+caused an exception to be thrown during finalization, if the close() method had
+already been called.  The exception was caught, but it was still an expensive
+operation.
+
+[7] The TurboJPEG API previously generated an error ("Could not determine
+subsampling type for JPEG image") when attempting to decompress grayscale JPEG
+images that were compressed with a sampling factor other than 1 (for instance,
+with 'cjpeg -grayscale -sample 2x2').  Subsampling technically has no meaning
+with grayscale JPEGs, and thus the horizontal and vertical sampling factors
+for such images are ignored by the decompressor.  However, the TurboJPEG API
+was being too rigid and was expecting the sampling factors to be equal to 1
+before it treated the image as a grayscale JPEG.
+
+[8] cjpeg, djpeg, and jpegtran now accept an argument of -version, which will
+print the library version and exit.
+
+[9] Referring to 1.4 beta1 [15], another extremely rare circumstance was
+discovered under which the Huffman encoder's local buffer can be overrun
+when a buffered destination manager is being used and an
+extremely-high-frequency block (basically junk image data) is being encoded.
+Even though the Huffman local buffer was increased from 128 bytes to 136 bytes
+to address the previous issue, the new issue caused even the larger buffer to
+be overrun.  Further analysis reveals that, in the absolute worst case (such as
+setting alternating AC coefficients to 32767 and -32768 in the JPEG scanning
+order), the Huffman encoder can produce encoded blocks that approach double the
+size of the unencoded blocks.  Thus, the Huffman local buffer was increased to
+256 bytes, which should prevent any such issue from re-occurring in the future.
+
+[10] The new tjPlaneSizeYUV(), tjPlaneWidth(), and tjPlaneHeight() functions
+were not actually usable on any platform except OS X and Windows, because
+those functions were not included in the libturbojpeg mapfile.  This has been
+fixed.
+
+[11] Restored the JPP(), JMETHOD(), and FAR macros in the libjpeg-turbo header
+files.  The JPP() and JMETHOD() macros were originally implemented in libjpeg
+as a way of supporting non-ANSI compilers that lacked support for prototype
+parameters.  libjpeg-turbo has never supported such compilers, but some
+software packages still use the macros to define their own prototypes.
+Similarly, libjpeg-turbo has never supported MS-DOS and other platforms that
+have far symbols, but some software packages still use the FAR macro.  A pretty
+good argument can be made that this is a bad practice on the part of the
+software in question, but since this affects more than one package, it's just
+easier to fix it here.
+
+[12] Fixed issues that were preventing the ARM 64-bit SIMD code from compiling
+for iOS, and included an ARMv8 architecture in all of the binaries installed by
+the "official" libjpeg-turbo SDK for OS X.
+
+
+1.3.90 (1.4 beta1)
+==================
+
+[1] New features in the TurboJPEG API:
+-- YUV planar images can now be generated with an arbitrary line padding
+(previously only 4-byte padding, which was compatible with X Video, was
+supported.)
+-- The decompress-to-YUV function has been extended to support image scaling.
+-- JPEG images can now be compressed from YUV planar source images.
+-- YUV planar images can now be decoded into RGB or grayscale images.
+-- 4:1:1 subsampling is now supported.  This is mainly included for
+compatibility, since 4:1:1 is not fully accelerated in libjpeg-turbo and has no
+significant advantages relative to 4:2:0.
+-- CMYK images are now supported.  This feature allows CMYK source images to be
+compressed to YCCK JPEGs and YCCK or CMYK JPEGs to be decompressed to CMYK
+destination images.  Conversion between CMYK/YCCK and RGB or YUV images is not
+supported.  Such conversion requires a color management system and is thus out
+of scope for a codec library.
+-- The handling of YUV images in the Java API has been significantly refactored
+and should now be much more intuitive.
+-- The Java API now supports encoding a YUV image from an arbitrary position in
+a large image buffer.
+-- All of the YUV functions now have a corresponding function that operates on
+separate image planes instead of a unified image buffer.  This allows for
+compressing/decoding from or decompressing/encoding to a subregion of a larger
+YUV image.  It also allows for handling YUV formats that swap the order of the
+U and V planes.
+
+[2] Added SIMD acceleration for DSPr2-capable MIPS platforms.  This speeds up
+the compression of full-color JPEGs by 70-80% on such platforms and
+decompression by 25-35%.
+
+[3] If an application attempts to decompress a Huffman-coded JPEG image whose
+header does not contain Huffman tables, libjpeg-turbo will now insert the
+default Huffman tables.  In order to save space, many motion JPEG video frames
+are encoded without the default Huffman tables, so these frames can now be
+successfully decompressed by libjpeg-turbo without additional work on the part
+of the application.  An application can still override the Huffman tables, for
+instance to re-use tables from a previous frame of the same video.
+
+[4] The Mac packaging system now uses pkgbuild and productbuild rather than
+PackageMaker (which is obsolete and no longer supported.)  This means that
+OS X 10.6 "Snow Leopard" or later must be used when packaging libjpeg-turbo,
+although the packages produced can be installed on OS X 10.5 "Leopard" or
+later.  OS X 10.4 "Tiger" is no longer supported.
+
+[5] The Huffman encoder now uses clz and bsr instructions for bit counting on
+ARM platforms rather than a lookup table.  This reduces the memory footprint
+by 64k, which may be important for some mobile applications.  Out of four
+Android devices that were tested, two demonstrated a small overall performance
+loss (~3-4% on average) with ARMv6 code and a small gain (also ~3-4%) with
+ARMv7 code when enabling this new feature, but the other two devices
+demonstrated a significant overall performance gain with both ARMv6 and ARMv7
+code (~10-20%) when enabling the feature.  Actual mileage may vary.
+
+[6] Worked around an issue with Visual C++ 2010 and later that caused incorrect
+pixels to be generated when decompressing a JPEG image to a 256-color bitmap,
+if compiler optimization was enabled when libjpeg-turbo was built.  This caused
+the regression tests to fail when doing a release build under Visual C++ 2010
+and later.
+
+[7] Improved the accuracy and performance of the non-SIMD implementation of the
+floating point inverse DCT (using code borrowed from libjpeg v8a and later.)
+The accuracy of this implementation now matches the accuracy of the SSE/SSE2
+implementation.  Note, however, that the floating point DCT/IDCT algorithms are
+mainly a legacy feature.  They generally do not produce significantly better
+accuracy than the slow integer DCT/IDCT algorithms, and they are quite a bit
+slower.
+
+[8] Added a new output colorspace (JCS_RGB565) to the libjpeg API that allows
+for decompressing JPEG images into RGB565 (16-bit) pixels.  If dithering is not
+used, then this code path is SIMD-accelerated on ARM platforms.
+
+[9] Numerous obsolete features, such as support for non-ANSI compilers and
+support for the MS-DOS memory model, were removed from the libjpeg code,
+greatly improving its readability and making it easier to maintain and extend.
+
+[10] Fixed a segfault that occurred when calling output_message() with msg_code
+set to JMSG_COPYRIGHT.
+
+[11] Fixed an issue whereby wrjpgcom was allowing comments longer than 65k
+characters to be passed on the command line, which was causing it to generate
+incorrect JPEG files.
+
+[12] Fixed a bug in the build system that was causing the Windows version of
+wrjpgcom to be built using the rdjpgcom source code.
+
+[13] Restored 12-bit-per-component JPEG support.  A 12-bit version of
+libjpeg-turbo can now be built by passing an argument of --with-12bit to
+configure (Unix) or -DWITH_12BIT=1 to cmake (Windows.)  12-bit JPEG support is
+included only for convenience.  Enabling this feature disables all of the
+performance features in libjpeg-turbo, as well as arithmetic coding and the
+TurboJPEG API.  The resulting library still contains the other libjpeg-turbo
+features (such as the colorspace extensions), but in general, it performs no
+faster than libjpeg v6b.
+
+[14] Added ARM 64-bit SIMD acceleration for the YCC-to-RGB color conversion
+and IDCT algorithms (both are used during JPEG decompression.)  For unknown
+reasons (probably related to clang), this code cannot currently be compiled for
+iOS.
+
+[15] Fixed an extremely rare bug that could cause the Huffman encoder's local
+buffer to overrun when a very high-frequency MCU is compressed using quality
+100 and no subsampling, and when the JPEG output buffer is being dynamically
+resized by the destination manager.  This issue was so rare that, even with a
+test program specifically designed to make the bug occur (by injecting random
+high-frequency YUV data into the compressor), it was reproducible only once in
+about every 25 million iterations.
+
+[16] Fixed an oversight in the TurboJPEG C wrapper:  if any of the JPEG
+compression functions was called repeatedly with the same
+automatically-allocated destination buffer, then TurboJPEG would erroneously
+assume that the jpegSize parameter was equal to the size of the buffer, when in
+fact that parameter was probably equal to the size of the most recently
+compressed JPEG image.  If the size of the previous JPEG image was not as large
+as the current JPEG image, then TurboJPEG would unnecessarily reallocate the
+destination buffer.
+
+
+1.3.1
+=====
+
+[1] On Un*x systems, 'make install' now installs the libjpeg-turbo libraries
+into /opt/libjpeg-turbo/lib32 by default on any 32-bit system, not just x86,
+and into /opt/libjpeg-turbo/lib64 by default on any 64-bit system, not just
+x86-64.  You can override this by overriding either the 'prefix' or 'libdir'
+configure variables.
+
+[2] The Windows installer now places a copy of the TurboJPEG DLLs in the same
+directory as the rest of the libjpeg-turbo binaries.  This was mainly done
+to support TurboVNC 1.3, which bundles the DLLs in its Windows installation.
+When using a 32-bit version of CMake on 64-bit Windows, it is impossible to
+access the c:\WINDOWS\system32 directory, which made it impossible for the
+TurboVNC build scripts to bundle the 64-bit TurboJPEG DLL.
+
+[3] Fixed a bug whereby attempting to encode a progressive JPEG with arithmetic
+entropy coding (by passing arguments of -progressive -arithmetic to cjpeg or
+jpegtran, for instance) would result in an error, "Requested feature was
+omitted at compile time".
+
+[4] Fixed a couple of issues whereby malformed JPEG images would cause
+libjpeg-turbo to use uninitialized memory during decompression.
+
+[5] Fixed an error ("Buffer passed to JPEG library is too small") that occurred
+when calling the TurboJPEG YUV encoding function with a very small (< 5x5)
+source image, and added a unit test to check for this error.
+
+[6] The Java classes should now build properly under Visual Studio 2010 and
+later.
+
+[7] Fixed an issue that prevented SRPMs generated using the in-tree packaging
+tools from being rebuilt on certain newer Linux distributions.
+
+[8] Numerous minor fixes to eliminate compilation and build/packaging system
+warnings, fix cosmetic issues, improve documentation clarity, and other general
+source cleanup.
+
+
+1.3.0
+=====
+
+[1] 'make test' now works properly on FreeBSD, and it no longer requires the
+md5sum executable to be present on other Un*x platforms.
+
+[2] Overhauled the packaging system:
+-- To avoid conflict with vendor-supplied libjpeg-turbo packages, the
+official RPMs and DEBs for libjpeg-turbo have been renamed to
+"libjpeg-turbo-official".
+-- The TurboJPEG libraries are now located under /opt/libjpeg-turbo in the
+official Linux and Mac packages, to avoid conflict with vendor-supplied
+packages and also to streamline the packaging system.
+-- Release packages are now created with the directory structure defined
+by the configure variables "prefix", "bindir", "libdir", etc. (Un*x) or by the
+CMAKE_INSTALL_PREFIX variable (Windows.)  The exception is that the docs are
+always located under the system default documentation directory on Un*x and Mac
+systems, and on Windows, the TurboJPEG DLL is always located in the Windows
+system directory.
+-- To avoid confusion, official libjpeg-turbo packages on Linux/Unix platforms
+(except for Mac) will always install the 32-bit libraries in
+/opt/libjpeg-turbo/lib32 and the 64-bit libraries in /opt/libjpeg-turbo/lib64.
+-- Fixed an issue whereby, in some cases, the libjpeg-turbo executables on Un*x
+systems were not properly linking with the shared libraries installed by the
+same package.
+-- Fixed an issue whereby building the "installer" target on Windows when
+WITH_JAVA=1 would fail if the TurboJPEG JAR had not been previously built.
+-- Building the "install" target on Windows now installs files into the same
+places that the installer does.
+
+[3] Fixed a Huffman encoder bug that prevented I/O suspension from working
+properly.
+
+
+1.2.90 (1.3 beta1)
+==================
+
+[1] Added support for additional scaling factors (3/8, 5/8, 3/4, 7/8, 9/8, 5/4,
+11/8, 3/2, 13/8, 7/4, 15/8, and 2) when decompressing.  Note that the IDCT will
+not be SIMD-accelerated when using any of these new scaling factors.
+
+[2] The TurboJPEG dynamic library is now versioned.  It was not strictly
+necessary to do so, because TurboJPEG uses versioned symbols, and if a function
+changes in an ABI-incompatible way, that function is renamed and a legacy
+function is provided to maintain backward compatibility.  However, certain
+Linux distro maintainers have a policy against accepting any library that isn't
+versioned.
+
+[3] Extended the TurboJPEG Java API so that it can be used to compress a JPEG
+image from and decompress a JPEG image to an arbitrary position in a large
+image buffer.
+
+[4] The tjDecompressToYUV() function now supports the TJFLAG_FASTDCT flag.
+
+[5] The 32-bit supplementary package for amd64 Debian systems now provides
+symlinks in /usr/lib/i386-linux-gnu for the TurboJPEG libraries in /usr/lib32.
+This allows those libraries to be used on MultiArch-compatible systems (such as
+Ubuntu 11 and later) without setting the linker path.
+
+[6] The TurboJPEG Java wrapper should now find the JNI library on Mac systems
+without having to pass -Djava.library.path=/usr/lib to java.
+
+[7] TJBench has been ported to Java to provide a convenient way of validating
+the performance of the TurboJPEG Java API.  It can be run with
+'java -cp turbojpeg.jar TJBench'.
+
+[8] cjpeg can now be used to generate JPEG files with the RGB colorspace
+(feature ported from jpeg-8d.)
+
+[9] The width and height in the -crop argument passed to jpegtran can now be
+suffixed with "f" to indicate that, when the upper left corner of the cropping
+region is automatically moved to the nearest iMCU boundary, the bottom right
+corner should be moved by the same amount.  In other words, this feature causes
+jpegtran to strictly honor the specified width/height rather than the specified
+bottom right corner (feature ported from jpeg-8d.)
+
+[10] JPEG files using the RGB colorspace can now be decompressed into grayscale
+images (feature ported from jpeg-8d.)
+
+[11] Fixed a regression caused by 1.2.1[7] whereby the build would fail with
+multiple "Mismatch in operand sizes" errors when attempting to build the x86
+SIMD code with NASM 0.98.
+
+[12] The in-memory source/destination managers (jpeg_mem_src() and
+jpeg_mem_dest()) are now included by default when building libjpeg-turbo with
+libjpeg v6b or v7 emulation, so that programs can take advantage of these
+functions without requiring the use of the backward-incompatible libjpeg v8
+ABI.  The "age number" of the libjpeg-turbo library on Un*x systems has been
+incremented by 1 to reflect this.  You can disable this feature with a
+configure/CMake switch in order to retain strict API/ABI compatibility with the
+libjpeg v6b or v7 API/ABI (or with previous versions of libjpeg-turbo.)  See
+README.md for more details.
+
+[13] Added ARMv7s architecture to libjpeg.a and libturbojpeg.a in the official
+libjpeg-turbo binary package for OS X, so that those libraries can be used to
+build applications that leverage the faster CPUs in the iPhone 5 and iPad 4.
+
+
+1.2.1
+=====
+
+[1] Creating or decoding a JPEG file that uses the RGB colorspace should now
+properly work when the input or output colorspace is one of the libjpeg-turbo
+colorspace extensions.
+
+[2] When libjpeg-turbo was built without SIMD support and merged (non-fancy)
+upsampling was used along with an alpha-enabled colorspace during
+decompression, the unused byte of the decompressed pixels was not being set to
+0xFF.  This has been fixed.  TJUnitTest has also been extended to test for the
+correct behavior of the colorspace extensions when merged upsampling is used.
+
+[3] Fixed a bug whereby the libjpeg-turbo SSE2 SIMD code would not preserve the
+upper 64 bits of xmm6 and xmm7 on Win64 platforms, which violated the Win64
+calling conventions.
+
+[4] Fixed a regression caused by 1.2.0[6] whereby decompressing corrupt JPEG
+images (specifically, images in which the component count was erroneously set
+to a large value) would cause libjpeg-turbo to segfault.
+
+[5] Worked around a severe performance issue with "Bobcat" (AMD Embedded APU)
+processors.  The MASKMOVDQU instruction, which was used by the libjpeg-turbo
+SSE2 SIMD code, is apparently implemented in microcode on AMD processors, and
+it is painfully slow on Bobcat processors in particular.  Eliminating the use
+of this instruction improved performance by an order of magnitude on Bobcat
+processors and by a small amount (typically 5%) on AMD desktop processors.
+
+[6] Added SIMD acceleration for performing 4:2:2 upsampling on NEON-capable ARM
+platforms.  This speeds up the decompression of 4:2:2 JPEGs by 20-25% on such
+platforms.
+
+[7] Fixed a regression caused by 1.2.0[2] whereby, on Linux/x86 platforms
+running the 32-bit SSE2 SIMD code in libjpeg-turbo, decompressing a 4:2:0 or
+4:2:2 JPEG image into a 32-bit (RGBX, BGRX, etc.) buffer without using fancy
+upsampling would produce several incorrect columns of pixels at the right-hand
+side of the output image if each row in the output image was not evenly
+divisible by 16 bytes.
+
+[8] Fixed an issue whereby attempting to build the SIMD extensions with Xcode
+4.3 on OS X platforms would cause NASM to return numerous errors of the form
+"'%define' expects a macro identifier".
+
+[9] Added flags to the TurboJPEG API that allow the caller to force the use of
+either the fast or the accurate DCT/IDCT algorithms in the underlying codec.
+
+
+1.2.0
+=====
+
+[1] Fixed build issue with YASM on Unix systems (the libjpeg-turbo build system
+was not adding the current directory to the assembler include path, so YASM
+was not able to find jsimdcfg.inc.)
+
+[2] Fixed out-of-bounds read in SSE2 SIMD code that occurred when decompressing
+a JPEG image to a bitmap buffer whose size was not a multiple of 16 bytes.
+This was more of an annoyance than an actual bug, since it did not cause any
+actual run-time problems, but the issue showed up when running libjpeg-turbo in
+valgrind.  See http://crbug.com/72399 for more information.
+
+[3] Added a compile-time macro (LIBJPEG_TURBO_VERSION) that can be used to
+check the version of libjpeg-turbo against which an application was compiled.
+
+[4] Added new RGBA/BGRA/ABGR/ARGB colorspace extension constants (libjpeg API)
+and pixel formats (TurboJPEG API), which allow applications to specify that,
+when decompressing to a 4-component RGB buffer, the unused byte should be set
+to 0xFF so that it can be interpreted as an opaque alpha channel.
+
+[5] Fixed regression issue whereby DevIL failed to build against libjpeg-turbo
+because libjpeg-turbo's distributed version of jconfig.h contained an INLINE
+macro, which conflicted with a similar macro in DevIL.  This macro is used only
+internally when building libjpeg-turbo, so it was moved into config.h.
+
+[6] libjpeg-turbo will now correctly decompress erroneous CMYK/YCCK JPEGs whose
+K component is assigned a component ID of 1 instead of 4.  Although these files
+are in violation of the spec, other JPEG implementations handle them
+correctly.
+
+[7] Added ARMv6 and ARMv7 architectures to libjpeg.a and libturbojpeg.a in
+the official libjpeg-turbo binary package for OS X, so that those libraries can
+be used to build both OS X and iOS applications.
+
+
+1.1.90 (1.2 beta1)
+==================
+
+[1] Added a Java wrapper for the TurboJPEG API.  See java/README for more
+details.
+
+[2] The TurboJPEG API can now be used to scale down images during
+decompression.
+
+[3] Added SIMD routines for RGB-to-grayscale color conversion, which
+significantly improves the performance of grayscale JPEG compression from an
+RGB source image.
+
+[4] Improved the performance of the C color conversion routines, which are used
+on platforms for which SIMD acceleration is not available.
+
+[5] Added a function to the TurboJPEG API that performs lossless transforms.
+This function is implemented using the same back end as jpegtran, but it
+performs transcoding entirely in memory and allows multiple transforms and/or
+crop operations to be batched together, so the source coefficients only need to
+be read once.  This is useful when generating image tiles from a single source
+JPEG.
+
+[6] Added tests for the new TurboJPEG scaled decompression and lossless
+transform features to tjbench (the TurboJPEG benchmark, formerly called
+"jpgtest".)
+
+[7] Added support for 4:4:0 (transposed 4:2:2) subsampling in TurboJPEG, which
+was necessary in order for it to read 4:2:2 JPEG files that had been losslessly
+transposed or rotated 90 degrees.
+
+[8] All legacy VirtualGL code has been re-factored, and this has allowed
+libjpeg-turbo, in its entirety, to be re-licensed under a BSD-style license.
+
+[9] libjpeg-turbo can now be built with YASM.
+
+[10] Added SIMD acceleration for ARM Linux and iOS platforms that support
+NEON instructions.
+
+[11] Refactored the TurboJPEG C API and documented it using Doxygen.  The
+TurboJPEG 1.2 API uses pixel formats to define the size and component order of
+the uncompressed source/destination images, and it includes a more efficient
+version of TJBUFSIZE() that computes a worst-case JPEG size based on the level
+of chrominance subsampling.  The refactored implementation of the TurboJPEG API
+now uses the libjpeg memory source and destination managers, which allows the
+TurboJPEG compressor to grow the JPEG buffer as necessary.
+
+[12] Eliminated errors in the output of jpegtran on Windows that occurred when
+the application was invoked using I/O redirection
+(jpegtran <input.jpg >output.jpg).
+
+[13] The inclusion of libjpeg v7 and v8 emulation as well as arithmetic coding
+support in libjpeg-turbo v1.1.0 introduced several new error constants in
+jerror.h, and these were mistakenly enabled for all emulation modes, causing
+the error enum in libjpeg-turbo to sometimes have different values than the
+same enum in libjpeg.  This represents an ABI incompatibility, and it caused
+problems with rare applications that took specific action based on a particular
+error value.  The fix was to include the new error constants conditionally
+based on whether libjpeg v7 or v8 emulation was enabled.
+
+[14] Fixed an issue whereby Windows applications that used libjpeg-turbo would
+fail to compile if the Windows system headers were included before jpeglib.h.
+This issue was caused by a conflict in the definition of the INT32 type.
+
+[15] Fixed 32-bit supplementary package for amd64 Debian systems, which was
+broken by enhancements to the packaging system in 1.1.
+
+[16] When decompressing a JPEG image using an output colorspace of
+JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR, or JCS_EXT_XRGB, libjpeg-turbo will
+now set the unused byte to 0xFF, which allows applications to interpret that
+byte as an alpha channel (0xFF = opaque).
+
+
+1.1.1
+=====
+
+[1] Fixed a 1-pixel error in row 0, column 21 of the luminance plane generated
+by tjEncodeYUV().
+
+[2] libjpeg-turbo's accelerated Huffman decoder previously ignored unexpected
+markers found in the middle of the JPEG data stream during decompression.  It
+will now hand off decoding of a particular block to the unaccelerated Huffman
+decoder if an unexpected marker is found, so that the unaccelerated Huffman
+decoder can generate an appropriate warning.
+
+[3] Older versions of MinGW64 prefixed symbol names with underscores by
+default, which differed from the behavior of 64-bit Visual C++.  MinGW64 1.0
+has adopted the behavior of 64-bit Visual C++ as the default, so to accommodate
+this, the libjpeg-turbo SIMD function names are no longer prefixed with an
+underscore when building with MinGW64.  This means that, when building
+libjpeg-turbo with older versions of MinGW64, you will now have to add
+-fno-leading-underscore to the CFLAGS.
+
+[4] Fixed a regression bug in the NSIS script that caused the Windows installer
+build to fail when using the Visual Studio IDE.
+
+[5] Fixed a bug in jpeg_read_coefficients() whereby it would not initialize
+cinfo->image_width and cinfo->image_height if libjpeg v7 or v8 emulation was
+enabled.  This specifically caused the jpegoptim program to fail if it was
+linked against a version of libjpeg-turbo that was built with libjpeg v7 or v8
+emulation.
+
+[6] Eliminated excessive I/O overhead that occurred when reading BMP files in
+cjpeg.
+
+[7] Eliminated errors in the output of cjpeg on Windows that occurred when the
+application was invoked using I/O redirection (cjpeg <inputfile >output.jpg).
+
+
+1.1.0
+=====
+
+[1] The algorithm used by the SIMD quantization function cannot produce correct
+results when the JPEG quality is >= 98 and the fast integer forward DCT is
+used.  Thus, the non-SIMD quantization function is now used for those cases,
+and libjpeg-turbo should now produce identical output to libjpeg v6b in all
+cases.
+
+[2] Despite the above, the fast integer forward DCT still degrades somewhat for
+JPEG qualities greater than 95, so the TurboJPEG wrapper will now automatically
+use the slow integer forward DCT when generating JPEG images of quality 96 or
+greater.  This reduces compression performance by as much as 15% for these
+high-quality images but is necessary to ensure that the images are perceptually
+lossless.  It also ensures that the library can avoid the performance pitfall
+created by [1].
+
+[3] Ported jpgtest.cxx to pure C to avoid the need for a C++ compiler.
+
+[4] Fixed visual artifacts in grayscale JPEG compression caused by a typo in
+the RGB-to-luminance lookup tables.
+
+[5] The Windows distribution packages now include the libjpeg run-time programs
+(cjpeg, etc.)
+
+[6] All packages now include jpgtest.
+
+[7] The TurboJPEG dynamic library now uses versioned symbols.
+
+[8] Added two new TurboJPEG API functions, tjEncodeYUV() and
+tjDecompressToYUV(), to replace the somewhat hackish TJ_YUV flag.
+
+
+1.0.90 (1.1 beta1)
+==================
+
+[1] Added emulation of the libjpeg v7 and v8 APIs and ABIs.  See
+README.md for more details.  This feature was sponsored by CamTrace SAS.
+
+[2] Created a new CMake-based build system for the Visual C++ and MinGW builds.
+
+[3] Grayscale bitmaps can now be compressed from/decompressed to using the
+TurboJPEG API.
+
+[4] jpgtest can now be used to test decompression performance with existing
+JPEG images.
+
+[5] If the default install prefix (/opt/libjpeg-turbo) is used, then
+'make install' now creates /opt/libjpeg-turbo/lib32 and
+/opt/libjpeg-turbo/lib64 sym links to duplicate the behavior of the binary
+packages.
+
+[6] All symbols in the libjpeg-turbo dynamic library are now versioned, even
+when the library is built with libjpeg v6b emulation.
+
+[7] Added arithmetic encoding and decoding support (can be disabled with
+configure or CMake options)
+
+[8] Added a TJ_YUV flag to the TurboJPEG API, which causes both the compressor
+and decompressor to output planar YUV images.
+
+[9] Added an extended version of tjDecompressHeader() to the TurboJPEG API,
+which allows the caller to determine the type of subsampling used in a JPEG
+image.
+
+[10] Added further protections against invalid Huffman codes.
+
+
+1.0.1
+=====
+
+[1] The Huffman decoder will now handle erroneous Huffman codes (for instance,
+from a corrupt JPEG image.)  Previously, these would cause libjpeg-turbo to
+crash under certain circumstances.
+
+[2] Fixed typo in SIMD dispatch routines that was causing 4:2:2 upsampling to
+be used instead of 4:2:0 when decompressing JPEG images using SSE2 code.
+
+[3] configure script will now automatically determine whether the
+INCOMPLETE_TYPES_BROKEN macro should be defined.
+
+
+1.0.0
+=====
+
+[1] 2983700: Further FreeBSD build tweaks (no longer necessary to specify
+--host when configuring on a 64-bit system)
+
+[2] Created symlinks in the Unix/Linux packages so that the TurboJPEG
+include file can always be found in /opt/libjpeg-turbo/include, the 32-bit
+static libraries can always be found in /opt/libjpeg-turbo/lib32, and the
+64-bit static libraries can always be found in /opt/libjpeg-turbo/lib64.
+
+[3] The Unix/Linux distribution packages now include the libjpeg run-time
+programs (cjpeg, etc.) and man pages.
+
+[4] Created a 32-bit supplementary package for amd64 Debian systems, which
+contains just the 32-bit libjpeg-turbo libraries.
+
+[5] Moved the libraries from */lib32 to */lib in the i386 Debian package.
+
+[6] Include distribution package for Cygwin
+
+[7] No longer necessary to specify --without-simd on non-x86 architectures, and
+unit tests now work on those architectures.
+
+
+0.0.93
+======
+
+[1] 2982659, Fixed x86-64 build on FreeBSD systems
+
+[2] 2988188: Added support for Windows 64-bit systems
+
+
+0.0.91
+======
+
+[1] Added documentation to .deb packages
+
+[2] 2968313: Fixed data corruption issues when decompressing large JPEG images
+and/or using buffered I/O with the libjpeg-turbo decompressor
+
+
+0.0.90
+======
+
+Initial release
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cjpeg.1 glmark2-2021.02/src/libjpeg-turbo/cjpeg.1
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cjpeg.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/cjpeg.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,351 @@
+.TH CJPEG 1 "17 February 2016"
+.SH NAME
+cjpeg \- compress an image file to a JPEG file
+.SH SYNOPSIS
+.B cjpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B cjpeg
+compresses the named image file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+The currently supported input file formats are: PPM (PBMPLUS color
+format), PGM (PBMPLUS grayscale format), BMP, Targa, and RLE (Utah Raster
+Toolkit format).  (RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-quality " N[,...]"
+Scale quantization tables to adjust image quality.  Quality is 0 (worst) to
+100 (best); default is 75.  (See below for more info.)
+.TP
+.B \-grayscale
+Create monochrome JPEG file from color input.  Be sure to use this switch when
+compressing a grayscale BMP file, because
+.B cjpeg
+isn't bright enough to notice whether a BMP file uses only shades of gray.
+By saying
+.BR \-grayscale ,
+you'll get a smaller JPEG file that takes less time to process.
+.TP
+.B \-rgb
+Create RGB JPEG file.
+Using this switch suppresses the conversion from RGB
+colorspace input to the default YCbCr JPEG colorspace.
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.  Without this, default
+encoding parameters are used.
+.B \-optimize
+usually makes the JPEG file a little smaller, but
+.B cjpeg
+runs somewhat slower and needs much more memory.  Image quality and speed of
+decompression are unaffected by
+.BR \-optimize .
+.TP
+.B \-progressive
+Create progressive JPEG file (see below).
+.TP
+.B \-targa
+Input file is Targa format.  Targa files that contain an "identification"
+field will not be automatically recognized by
+.BR cjpeg ;
+for such files you must specify
+.B \-targa
+to make
+.B cjpeg
+treat the input as Targa format.
+For most Targa files, you won't need this switch.
+.PP
+The
+.B \-quality
+switch lets you trade off compressed file size against quality of the
+reconstructed image: the higher the quality setting, the larger the JPEG file,
+and the closer the output image will be to the original input.  Normally you
+want to use the lowest quality setting (smallest file) that decompresses into
+something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at
+.B \-quality
+75, then go up 5 or 10 counts at a time until you are happy with the output
+image.  (The optimal setting will vary from one image to another.)
+.PP
+.B \-quality
+100 will generate a quantization table of all 1's, minimizing loss in the
+quantization step (but there is still information loss in subsampling, as well
+as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+.PP
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try
+.B \-quality
+2 (or so) for some amusing Cubist effects.  (Note: quality
+values below about 25 generate 2-byte quantization tables, which are
+considered optional in the JPEG standard.
+.B cjpeg
+emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use
+.B \-baseline
+if you need to ensure compatibility at low quality values.)
+.PP
+The \fB-quality\fR option has been extended in this version of \fBcjpeg\fR to
+support separate quality settings for luminance and chrominance (or, in
+general, separate settings for every quantization table slot.)  The principle
+is the same as chrominance subsampling:  since the human eye is more sensitive
+to spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+.PP
+The \fB-quality\fR option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.
+More (or customized) quantization tables can be set with the \fB-qtables\fR
+option and assigned to components with the \fB-qslots\fR option (see the
+"wizard" switches below.)
+.PP
+JPEG files generated with separate luminance and chrominance quality are fully
+compliant with standard JPEG decoders.
+.PP
+.BR CAUTION:
+For this setting to be useful, be sure to pass an argument of \fB-sample 1x1\fR
+to \fBcjpeg\fR to disable chrominance subsampling.  Otherwise, the default
+subsampling level (2x2, AKA "4:2:0") will be used.
+.PP
+The
+.B \-progressive
+switch creates a "progressive JPEG" file.  In this type of JPEG file, the data
+is stored in multiple scans of increasing quality.  If the file is being
+transmitted over a slow communications link, the decoder can use the first
+scan to display a low-quality image very quickly, and can then improve the
+display with each subsequent scan.  The final image is exactly equivalent to a
+standard JPEG file of the same quality setting, and the total file size is
+about the same --- often a little smaller.
+.PP
+Switches for advanced users:
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.B Caution:
+arithmetic coded JPEG is not yet widely implemented, so many decoders will be
+unable to view an arithmetic coded JPEG file at all.
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+In libjpeg-turbo, the fast method is generally about 5-15% faster than the int
+method when using the x86/x86-64 SIMD extensions (results may vary with other
+SIMD implementations, or when using libjpeg-turbo without SIMD extensions.)
+For quality levels of 90 and below, there should be little or no perceptible
+difference between the two algorithms.  For quality levels above 90, however,
+the difference between the fast and the int methods becomes more pronounced.
+With quality=97, for instance, the fast method incurs generally about a 1-3 dB
+loss (in PSNR) relative to the int method, but this can be larger for some
+images.  Do not use the fast method with quality levels above 97.  The
+algorithm often degenerates at quality=98 and above and can actually produce a
+more lossy image than if lower quality levels had been used.  Also, in
+libjpeg-turbo, the fast method is not fully accelerated for quality levels
+above 97, so it will be slower than the int method.
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is mainly a legacy feature.  It does not produce significantly
+more accurate results than the int method, and it is much slower.  The float
+method may also give different results on different machines due to varying
+roundoff behavior, whereas the integer methods should give the same results on
+all machines.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.B \-restart 0
+(the default) means no restart markers.
+.TP
+.BI \-smooth " N"
+Smooth the input image to eliminate dithering noise.  N, ranging from 1 to
+100, indicates the strength of smoothing.  0 (the default) means no smoothing.
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memdst
+Compress to memory instead of a file.  This feature was implemented mainly as a
+way of testing the in-memory destination manager (jpeg_mem_dest()), but it is
+also useful for benchmarking, since it reduces the I/O overhead.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.PP
+The
+.B \-restart
+option inserts extra markers that allow a JPEG decoder to resynchronize after
+a transmission error.  Without restart markers, any damage to a compressed
+file will usually ruin the image from the point of the error to the end of the
+image; with restart markers, the damage is usually confined to the portion of
+the image up to the next restart marker.  Of course, the restart markers
+occupy extra space.  We recommend
+.B \-restart 1
+for images that will be transmitted across unreliable networks such as Usenet.
+.PP
+The
+.B \-smooth
+option filters the input to eliminate fine-scale noise.  This is often useful
+when converting dithered images to JPEG: a moderate smoothing factor of 10 to
+50 gets rid of dithering patterns in the input file, resulting in a smaller
+JPEG file and a better-looking image.  Too large a smoothing factor will
+visibly blur the image, however.
+.PP
+Switches for wizards:
+.TP
+.B \-baseline
+Force baseline-compatible quantization tables to be generated.  This clamps
+quantization values to 8 bits even at low quality settings.  (This switch is
+poorly named, since it does not ensure that the output is actually baseline
+JPEG.  For example, you can use
+.B \-baseline
+and
+.B \-progressive
+together.)
+.TP
+.BI \-qtables " file"
+Use the quantization tables given in the specified text file.
+.TP
+.BI \-qslots " N[,...]"
+Select which quantization table to use for each color component.
+.TP
+.BI \-sample " HxV[,...]"
+Set JPEG sampling factors for each color component.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, \fBdon't use them\fR.  These switches are
+documented further in the file wizard.txt.
+.SH EXAMPLES
+.LP
+This example compresses the PPM file foo.ppm with a quality factor of
+60 and saves the output as foo.jpg:
+.IP
+.B cjpeg \-quality
+.I 60 foo.ppm
+.B >
+.I foo.jpg
+.SH HINTS
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with
+.BR cjpeg 's
+.B \-quality
+and
+.B \-smooth
+options to get a satisfactory conversion.
+.B \-smooth 10
+or so is often helpful.
+.PP
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+.PP
+The
+.B \-optimize
+option to
+.B cjpeg
+is worth using when you are making a "final" version for posting or archiving.
+It's also a win when you are using low quality settings to make very small
+JPEG files; the percentage improvement is often a lot more than it is on
+larger files.  (At present,
+.B \-optimize
+mode is always selected when generating progressive JPEG files.)
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.
+.SH ISSUES
+Support for GIF input files was removed in cjpeg v6b due to concerns over
+the Unisys LZW patent.  Although this patent expired in 2006, cjpeg still
+lacks GIF support, for these historical reasons.  (Conversion of GIF files to
+JPEG is usually a bad idea anyway, since GIF is a 256-color format.)
+.PP
+Not all variants of BMP and Targa file formats are supported.
+.PP
+The
+.B \-targa
+switch is not a bug, it's a feature.  (It would be a bug if the Targa format
+designers had not been clueless.)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cjpeg.c glmark2-2021.02/src/libjpeg-turbo/cjpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/cjpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/cjpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,644 @@
+/*
+ * cjpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2013-2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG compressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      cjpeg [options]  inputfile outputfile
+ *      cjpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      cjpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)   string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This routine determines what format the input file is,
+ * and selects the appropriate input-reading module.
+ *
+ * To determine which family of input formats the file belongs to,
+ * we may look only at the first byte of the file, since C does not
+ * guarantee that more than one character can be pushed back with ungetc.
+ * Looking at additional bytes would require one of these approaches:
+ *     1) assume we can fseek() the input file (fails for piped input);
+ *     2) assume we can push back more than one character (works in
+ *        some C implementations, but unportable);
+ *     3) provide our own buffering (breaks input readers that want to use
+ *        stdio directly, such as the RLE library);
+ * or  4) don't put back the data, and modify the input_init methods to assume
+ *        they start reading after the start of file (also breaks RLE library).
+ * #1 is attractive for MS-DOS but is untenable on Unix.
+ *
+ * The most portable solution for file types that can't be identified by their
+ * first byte is to make the user tell us what they are.  This is also the
+ * only approach for "raw" file types that contain only arbitrary values.
+ * We presently apply this method for Targa files.  Most of the time Targa
+ * files start with 0x00, so we recognize that case.  Potentially, however,
+ * a Targa file could start with any byte value (byte 0 is the length of the
+ * seldom-used ID field), so we provide a switch to force Targa input mode.
+ */
+
+static boolean is_targa;        /* records user -targa switch */
+
+
+LOCAL(cjpeg_source_ptr)
+select_file_type (j_compress_ptr cinfo, FILE *infile)
+{
+  int c;
+
+  if (is_targa) {
+#ifdef TARGA_SUPPORTED
+    return jinit_read_targa(cinfo);
+#else
+    ERREXIT(cinfo, JERR_TGA_NOTCOMP);
+#endif
+  }
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+  if (ungetc(c, infile) == EOF)
+    ERREXIT(cinfo, JERR_UNGETC_FAILED);
+
+  switch (c) {
+#ifdef BMP_SUPPORTED
+  case 'B':
+    return jinit_read_bmp(cinfo);
+#endif
+#ifdef GIF_SUPPORTED
+  case 'G':
+    return jinit_read_gif(cinfo);
+#endif
+#ifdef PPM_SUPPORTED
+  case 'P':
+    return jinit_read_ppm(cinfo);
+#endif
+#ifdef RLE_SUPPORTED
+  case 'R':
+    return jinit_read_rle(cinfo);
+#endif
+#ifdef TARGA_SUPPORTED
+  case 0x00:
+    return jinit_read_targa(cinfo);
+#endif
+  default:
+    ERREXIT(cinfo, JERR_UNKNOWN_FORMAT);
+    break;
+  }
+
+  return NULL;                  /* suppress compiler warnings */
+}
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+boolean memdst;                 /* for -memdst switch */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -quality N[,...]   Compression quality (0..100; 5-95 is most useful range,\n");
+  fprintf(stderr, "                     default is 75)\n");
+  fprintf(stderr, "  -grayscale     Create monochrome JPEG file\n");
+  fprintf(stderr, "  -rgb           Create RGB JPEG file\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Input file is Targa format (usually not needed)\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+#ifdef INPUT_SMOOTHING_SUPPORTED
+  fprintf(stderr, "  -smooth N      Smooth dithered input (N=1..100 is strength)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memdst        Compress to memory instead of file (useful for benchmarking)\n");
+#endif
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+  fprintf(stderr, "  -baseline      Force baseline quantization tables\n");
+  fprintf(stderr, "  -qtables file  Use quantization tables given in file\n");
+  fprintf(stderr, "  -qslots N[,...]    Set component quantization tables\n");
+  fprintf(stderr, "  -sample HxV[,...]  Set component sampling factors\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean force_baseline;
+  boolean simple_progressive;
+  char *qualityarg = NULL;      /* saves -quality parm if any */
+  char *qtablefile = NULL;      /* saves -qtables filename if any */
+  char *qslotsarg = NULL;       /* saves -qslots parm if any */
+  char *samplearg = NULL;       /* saves -sample parm if any */
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+
+  force_baseline = FALSE;       /* by default, allow 16-bit quantizers */
+  simple_progressive = FALSE;
+  is_targa = FALSE;
+  outfilename = NULL;
+  memdst = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "baseline", 1)) {
+      /* Force baseline-compatible output (8-bit quantizer values). */
+      force_baseline = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select DCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force a monochrome JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+
+    } else if (keymatch(arg, "rgb", 3)) {
+      /* Force an RGB JPEG file to be generated. */
+      jpeg_set_colorspace(cinfo, JCS_RGB);
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "progressive", 1)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "memdst", 2)) {
+      /* Use in-memory destination manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memdst = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory destination manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "quality", 1)) {
+      /* Quality ratings (quantization table scaling factors). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qualityarg = argv[argn];
+
+    } else if (keymatch(arg, "qslots", 2)) {
+      /* Quantization table slot numbers. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qslotsarg = argv[argn];
+      /* Must delay setting qslots until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default quant table numbers.
+       */
+
+    } else if (keymatch(arg, "qtables", 2)) {
+      /* Quantization tables fetched from file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      qtablefile = argv[argn];
+      /* We postpone actually reading the file in case -quality comes later. */
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int) lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int) lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "sample", 2)) {
+      /* Set sampling factors. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      samplearg = argv[argn];
+      /* Must delay setting sample factors until after we have processed any
+       * colorspace-determining switches, since jpeg_set_colorspace sets
+       * default sampling factors.
+       */
+
+    } else if (keymatch(arg, "scans", 4)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "smooth", 2)) {
+      /* Set input smoothing factor. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+        usage();
+      if (val < 0 || val > 100)
+        usage();
+      cinfo->smoothing_factor = val;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Input file is Targa format. */
+      is_targa = TRUE;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+    /* Set quantization tables for selected quality. */
+    /* Some or all may be overridden if -qtables is present. */
+    if (qualityarg != NULL)     /* process -quality if it was present */
+      if (! set_quality_ratings(cinfo, qualityarg, force_baseline))
+        usage();
+
+    if (qtablefile != NULL)     /* process -qtables if it was present */
+      if (! read_quant_tables(cinfo, qtablefile, force_baseline))
+        usage();
+
+    if (qslotsarg != NULL)      /* process -qslots if it was present */
+      if (! set_quant_slots(cinfo, qslotsarg))
+        usage();
+
+    if (samplearg != NULL)      /* process -sample if it was present */
+      if (! set_sample_factors(cinfo, samplearg))
+        usage();
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_compress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  cjpeg_source_ptr src_mgr;
+  FILE *input_file;
+  FILE *output_file = NULL;
+  unsigned char *outbuffer = NULL;
+  unsigned long outsize = 0;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "cjpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG compression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_compress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Initialize JPEG parameters.
+   * Much of this may be overridden later.
+   * In particular, we don't yet know the input file's color space,
+   * but we need to provide some value for jpeg_set_defaults() to work.
+   */
+
+  cinfo.in_color_space = JCS_RGB; /* arbitrary guess */
+  jpeg_set_defaults(&cinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  if (!memdst) {
+    /* Must have either -outfile switch or explicit output file name */
+    if (outfilename == NULL) {
+      if (file_index != argc-2) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+      outfilename = argv[file_index+1];
+    } else {
+      if (file_index != argc-1) {
+        fprintf(stderr, "%s: must name one input and one output file\n",
+                progname);
+        usage();
+      }
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else if (!memdst) {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Figure out the input file format, and set up to read it. */
+  src_mgr = select_file_type(&cinfo, input_file);
+  src_mgr->input_file = input_file;
+
+  /* Read the input file header to obtain file size & colorspace. */
+  (*src_mgr->start_input) (&cinfo, src_mgr);
+
+  /* Now that we know input colorspace, fix colorspace-dependent defaults */
+  jpeg_default_colorspace(&cinfo);
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memdst)
+    jpeg_mem_dest(&cinfo, &outbuffer, &outsize);
+  else
+#endif
+    jpeg_stdio_dest(&cinfo, output_file);
+
+  /* Start compressor */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Process data */
+  while (cinfo.next_scanline < cinfo.image_height) {
+    num_scanlines = (*src_mgr->get_pixel_rows) (&cinfo, src_mgr);
+    (void) jpeg_write_scanlines(&cinfo, src_mgr->buffer, num_scanlines);
+  }
+
+  /* Finish compression and release memory */
+  (*src_mgr->finish_input) (&cinfo, src_mgr);
+  jpeg_finish_compress(&cinfo);
+  jpeg_destroy_compress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout && output_file != NULL)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  if (memdst) {
+    fprintf(stderr, "Compressed size:  %lu bytes\n", outsize);
+    if (outbuffer != NULL)
+      free(outbuffer);
+  }
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/coderules.txt glmark2-2021.02/src/libjpeg-turbo/coderules.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/coderules.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/coderules.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,78 @@
+IJG JPEG LIBRARY:  CODING RULES
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-1996, Thomas G. Lane.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+Since numerous people will be contributing code and bug fixes, it's important
+to establish a common coding style.  The goal of using similar coding styles
+is much more important than the details of just what that style is.
+
+In general we follow the recommendations of "Recommended C Style and Coding
+Standards" revision 6.1 (Cannon et al. as modified by Spencer, Keppel and
+Brader).  This document is available in the IJG FTP archive (see
+jpeg/doc/cstyle.ms.tbl.Z, or cstyle.txt.Z for those without nroff/tbl).
+
+Block comments should be laid out thusly:
+
+/*
+ *  Block comments in this style.
+ */
+
+We indent statements in K&R style, e.g.,
+        if (test) {
+          then-part;
+        } else {
+          else-part;
+        }
+with two spaces per indentation level.  (This indentation convention is
+handled automatically by GNU Emacs and many other text editors.)
+
+Multi-word names should be written in lower case with underscores, e.g.,
+multi_word_name (not multiWordName).  Preprocessor symbols and enum constants
+are similar but upper case (MULTI_WORD_NAME).  Names should be unique within
+the first fifteen characters.
+
+Note that each function definition must begin with GLOBAL(type), LOCAL(type),
+or METHODDEF(type).  These macros expand to "static type" or just "type" as
+appropriate.  They provide a readable indication of the routine's usage and
+can readily be changed for special needs.  (For instance, special linkage
+keywords can be inserted for use in Windows DLLs.)
+
+A similar solution is used for external function declarations (see the EXTERN
+macro.)
+
+
+The JPEG library is intended to be used within larger programs.  Furthermore,
+we want it to be reentrant so that it can be used by applications that process
+multiple images concurrently.  The following rules support these requirements:
+
+1. Avoid direct use of file I/O, "malloc", error report printouts, etc;
+pass these through the common routines provided.
+
+2. Minimize global namespace pollution.  Functions should be declared static
+wherever possible.  (Note that our method-based calling conventions help this
+a lot: in many modules only the initialization function will ever need to be
+called directly, so only that function need be externally visible.)  All
+global function names should begin with "jpeg_".
+
+3. Don't use global variables; anything that must be used in another module
+should be in the common data structures.
+
+4. Don't use static variables except for read-only constant tables.  Variables
+that should be private to a module can be placed into private structures (see
+the system architecture document, structure.txt).
+
+5. Source file names should begin with "j" for files that are part of the
+library proper; source files that are not part of the library, such as cjpeg.c
+and djpeg.c, do not begin with "j".  Keep compression and decompression code in
+separate source files --- some applications may want only one half of the
+library.
+
+Note: these rules (particularly #4) are not followed religiously in the
+modules that are used in cjpeg/djpeg but are not part of the JPEG library
+proper.  Those modules are not really intended to be used in other
+applications.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/config.h glmark2-2021.02/src/libjpeg-turbo/config.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/config.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/config.h	1970-01-01 08:00:00.000000000 +0800
@@ -1,137 +0,0 @@
-/* config.h.  Generated from config.h.in by configure.  */
-/* config.h.in.  Generated from configure.ac by autoheader.  */
-
-/* Build number */
-#define BUILD "20120626"
-
-/* Support arithmetic encoding */
-#define C_ARITH_CODING_SUPPORTED 1
-
-/* Support arithmetic decoding */
-#define D_ARITH_CODING_SUPPORTED 1
-
-/* Define to 1 if you have the <dlfcn.h> header file. */
-#define HAVE_DLFCN_H 1
-
-/* Define to 1 if you have the <inttypes.h> header file. */
-#define HAVE_INTTYPES_H 1
-
-/* Define to 1 if you have the <jni.h> header file. */
-/* #undef HAVE_JNI_H */
-
-/* Define to 1 if you have the `memcpy' function. */
-#define HAVE_MEMCPY 1
-
-/* Define to 1 if you have the <memory.h> header file. */
-#define HAVE_MEMORY_H 1
-
-/* Define to 1 if you have the `memset' function. */
-#define HAVE_MEMSET 1
-
-/* Define if your compiler supports prototypes */
-#define HAVE_PROTOTYPES 1
-
-/* Define to 1 if you have the <stddef.h> header file. */
-#define HAVE_STDDEF_H 1
-
-/* Define to 1 if you have the <stdint.h> header file. */
-#define HAVE_STDINT_H 1
-
-/* Define to 1 if you have the <stdlib.h> header file. */
-#define HAVE_STDLIB_H 1
-
-/* Define to 1 if you have the <strings.h> header file. */
-#define HAVE_STRINGS_H 1
-
-/* Define to 1 if you have the <string.h> header file. */
-#define HAVE_STRING_H 1
-
-/* Define to 1 if you have the <sys/stat.h> header file. */
-#define HAVE_SYS_STAT_H 1
-
-/* Define to 1 if you have the <sys/types.h> header file. */
-#define HAVE_SYS_TYPES_H 1
-
-/* Define to 1 if you have the <unistd.h> header file. */
-#define HAVE_UNISTD_H 1
-
-/* Define to 1 if the system has the type `unsigned char'. */
-#define HAVE_UNSIGNED_CHAR 1
-
-/* Define to 1 if the system has the type `unsigned short'. */
-#define HAVE_UNSIGNED_SHORT 1
-
-/* Compiler does not support pointers to undefined structures. */
-/* #undef INCOMPLETE_TYPES_BROKEN */
-
-/* How to obtain function inlining. */
-#define INLINE __attribute__((always_inline))
-
-/* libjpeg API version */
-#define JPEG_LIB_VERSION 62
-
-/* libjpeg-turbo version */
-#define LIBJPEG_TURBO_VERSION 1.2.0
-
-/* Define to the sub-directory in which libtool stores uninstalled libraries.
-   */
-#define LT_OBJDIR ".libs/"
-
-/* Define if you have BSD-like bzero and bcopy */
-/* #undef NEED_BSD_STRINGS */
-
-/* Define if you need short function names */
-/* #undef NEED_SHORT_EXTERNAL_NAMES */
-
-/* Define if you have sys/types.h */
-#define NEED_SYS_TYPES_H 1
-
-/* Name of package */
-#define PACKAGE "libjpeg-turbo"
-
-/* Define to the address where bug reports for this package should be sent. */
-#define PACKAGE_BUGREPORT ""
-
-/* Define to the full name of this package. */
-#define PACKAGE_NAME "libjpeg-turbo"
-
-/* Define to the full name and version of this package. */
-#define PACKAGE_STRING "libjpeg-turbo 1.2.0"
-
-/* Define to the one symbol short name of this package. */
-#define PACKAGE_TARNAME "libjpeg-turbo"
-
-/* Define to the home page for this package. */
-#define PACKAGE_URL ""
-
-/* Define to the version of this package. */
-#define PACKAGE_VERSION "1.2.0"
-
-/* Define if shift is unsigned */
-/* #undef RIGHT_SHIFT_IS_UNSIGNED */
-
-/* Define to 1 if you have the ANSI C header files. */
-#define STDC_HEADERS 1
-
-/* Version number of package */
-#define VERSION "1.2.0"
-
-/* Use accelerated SIMD routines. */
-#define WITH_SIMD 1
-
-/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
-#ifndef __CHAR_UNSIGNED__
-/* # undef __CHAR_UNSIGNED__ */
-#endif
-
-/* Define to empty if `const' does not conform to ANSI C. */
-/* #undef const */
-
-/* Define to `__inline__' or `__inline' if that's what the C compiler
-   calls it, or to nothing if 'inline' is not supported under any name.  */
-#ifndef __cplusplus
-/* #undef inline */
-#endif
-
-/* Define to `unsigned int' if <sys/types.h> does not define. */
-/* #undef size_t */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/configure.ac glmark2-2021.02/src/libjpeg-turbo/configure.ac
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/configure.ac	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/configure.ac	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,595 @@
+#                                               -*- Autoconf -*-
+# Process this file with autoconf to produce a configure script.
+
+AC_PREREQ([2.56])
+AC_INIT([libjpeg-turbo], [1.4.90])
+
+AM_INIT_AUTOMAKE([-Wall foreign dist-bzip2])
+AC_PREFIX_DEFAULT(/opt/libjpeg-turbo)
+
+m4_ifdef([AM_SILENT_RULES], [AM_SILENT_RULES([yes])])
+
+# Checks for programs.
+SAVED_CFLAGS=${CFLAGS}
+SAVED_CPPFLAGS=${CPPFLAGS}
+AC_PROG_CPP
+AC_PROG_CC
+m4_ifdef([AM_PROG_AR], [AM_PROG_AR])
+AM_PROG_AS
+AM_PROG_CC_C_O
+AC_PROG_INSTALL
+AC_PROG_LIBTOOL
+AC_PROG_LN_S
+
+AC_ARG_WITH([build-date], [Use custom build string to enable reproducible builds (default: YYMMDD)],
+  [BUILD="$with_build_date"],
+  [BUILD=`date +%Y%m%d`])
+
+PKG_PROG_PKG_CONFIG
+
+# When the prefix is /opt/libjpeg-turbo, we assume that an "official" binary is
+# being created, and thus we install things into specific locations.
+
+old_prefix=${prefix}
+if test "x$prefix" = "xNONE" -a "x$ac_default_prefix" != "x"; then
+  prefix=$ac_default_prefix
+fi
+DATADIR=`eval echo ${datadir}`
+DATADIR=`eval echo $DATADIR`
+if test "$DATADIR" = "/opt/libjpeg-turbo/share"; then
+  datadir='${prefix}'
+fi
+DATADIR=`eval echo ${datarootdir}`
+DATADIR=`eval echo $DATADIR`
+if test "$DATADIR" = "/opt/libjpeg-turbo/share"; then
+  datarootdir='${prefix}'
+fi
+DOCDIR=`eval echo ${docdir}`
+DOCDIR=`eval echo $DOCDIR`
+if test "$DOCDIR" = "/opt/libjpeg-turbo/doc/libjpeg-turbo"; then
+  docdir='${datadir}/doc'
+fi
+
+old_exec_prefix=${exec_prefix}
+if test "x$exec_prefix" = "xNONE"; then
+  exec_prefix=${prefix}
+fi
+
+AC_CHECK_SIZEOF(size_t)
+
+if test "x${libdir}" = 'x${exec_prefix}/lib' -o "x${libdir}" = 'x${prefix}/lib'; then
+  LIBDIR=`eval echo ${libdir}`
+  LIBDIR=`eval echo $LIBDIR`
+  if test "$LIBDIR" = "/opt/libjpeg-turbo/lib"; then
+    case $host_os in
+      darwin*)
+        ;;
+      *)
+        if test "${ac_cv_sizeof_size_t}" = "8"; then
+          libdir='${exec_prefix}/lib64'
+        elif test "${ac_cv_sizeof_size_t}" = "4"; then
+          libdir='${exec_prefix}/lib32'
+        fi
+        ;;
+    esac
+  fi
+fi
+exec_prefix=${old_exec_prefix}
+prefix=${old_prefix}
+
+# Check whether compiler supports pointers to undefined structures
+AC_MSG_CHECKING(whether compiler supports pointers to undefined structures)
+AC_TRY_COMPILE([ typedef struct undefined_structure *undef_struct_ptr; ], ,
+  AC_MSG_RESULT(yes),
+  [AC_MSG_RESULT(no)
+   AC_DEFINE([INCOMPLETE_TYPES_BROKEN], [1],
+     [Compiler does not support pointers to undefined structures.])])
+
+if test "x${GCC}" = "xyes"; then
+  if test "x${SAVED_CFLAGS}" = "x"; then
+    CFLAGS=-O3
+  fi
+  if test "x${SAVED_CPPFLAGS}" = "x"; then
+    CPPFLAGS=-Wall
+  fi
+fi
+
+AC_CHECK_DECL([__SUNPRO_C], [SUNCC="yes"], [SUNCC="no"])
+if test "x${SUNCC}" = "xyes"; then
+  if test "x${SAVED_CFLAGS}" = "x"; then
+    CFLAGS=-xO5
+  fi
+fi
+
+# Checks for libraries.
+
+# Checks for header files.
+AC_HEADER_STDC
+AC_CHECK_HEADERS([stddef.h stdlib.h locale.h string.h])
+AC_CHECK_HEADER([sys/types.h],
+  AC_DEFINE([NEED_SYS_TYPES_H], 1, [Define if you need to include <sys/types.h> to get size_t.]))
+
+# Checks for typedefs, structures, and compiler characteristics.
+AC_C_CONST
+AC_C_CHAR_UNSIGNED
+AC_C_INLINE
+AC_TYPE_SIZE_T
+AC_CHECK_TYPES([unsigned char, unsigned short])
+
+AC_MSG_CHECKING([if right shift is signed])
+AC_TRY_RUN(
+  [#include <stdio.h>
+   int is_shifting_signed (long arg) {
+     long res = arg >> 4;
+
+     if (res == -0x7F7E80CL)
+       return 1; /* right shift is signed */
+
+     /* see if unsigned-shift hack will fix it. */
+     /* we can't just test exact value since it depends on width of long... */
+     res |= (~0L) << (32-4);
+     if (res == -0x7F7E80CL)
+       return 0; /* right shift is unsigned */
+
+     printf("Right shift isn't acting as I expect it to.\n");
+     printf("I fear the JPEG software will not work at all.\n\n");
+     return 0; /* try it with unsigned anyway */
+   }
+   int main (void) {
+     exit(is_shifting_signed(-0x7F7E80B1L));
+   }],
+  [AC_MSG_RESULT(no)
+   AC_DEFINE([RIGHT_SHIFT_IS_UNSIGNED], 1,
+     [Define if your (broken) compiler shifts signed values as if they were unsigned.])],
+  [AC_MSG_RESULT(yes)],
+  [AC_MSG_RESULT(Assuming that right shift is signed on target machine.)])
+
+# Checks for library functions.
+AC_CHECK_FUNCS([memset memcpy], [],
+  [AC_DEFINE([NEED_BSD_STRINGS], 1,
+     [Define if you have BSD-like bzero and bcopy in <strings.h> rather than memset/memcpy in <string.h>.])])
+
+AC_MSG_CHECKING([libjpeg API version])
+AC_ARG_VAR(JPEG_LIB_VERSION, [libjpeg API version (62, 70, or 80)])
+if test "x$JPEG_LIB_VERSION" = "x"; then
+  AC_ARG_WITH([jpeg7],
+    AC_HELP_STRING([--with-jpeg7],
+      [Emulate libjpeg v7 API/ABI (this makes libjpeg-turbo backward incompatible with libjpeg v6b.)]))
+  AC_ARG_WITH([jpeg8],
+    AC_HELP_STRING([--with-jpeg8],
+      [Emulate libjpeg v8 API/ABI (this makes libjpeg-turbo backward incompatible with libjpeg v6b.)]))
+  if test "x${with_jpeg8}" = "xyes"; then
+    JPEG_LIB_VERSION=80
+  else
+    if test "x${with_jpeg7}" = "xyes"; then
+      JPEG_LIB_VERSION=70
+    else
+      JPEG_LIB_VERSION=62
+    fi
+  fi
+fi
+JPEG_LIB_VERSION_DECIMAL=`expr $JPEG_LIB_VERSION / 10`.`expr $JPEG_LIB_VERSION % 10`
+AC_SUBST(JPEG_LIB_VERSION_DECIMAL)
+AC_MSG_RESULT([$JPEG_LIB_VERSION_DECIMAL])
+AC_DEFINE_UNQUOTED(JPEG_LIB_VERSION, [$JPEG_LIB_VERSION],
+  [libjpeg API version])
+
+AC_ARG_VAR(SO_MAJOR_VERSION,
+  [Major version of the libjpeg-turbo shared library (default is determined by the API version)])
+AC_ARG_VAR(SO_MINOR_VERSION,
+  [Minor version of the libjpeg-turbo shared library (default is determined by the API version)])
+if test "x$SO_MAJOR_VERSION" = "x"; then
+  case "$JPEG_LIB_VERSION" in
+    62)  SO_MAJOR_VERSION=$JPEG_LIB_VERSION ;;
+    *)   SO_MAJOR_VERSION=`expr $JPEG_LIB_VERSION / 10` ;;
+  esac
+fi
+if test "x$SO_MINOR_VERSION" = "x"; then
+  case "$JPEG_LIB_VERSION" in
+    80)  SO_MINOR_VERSION=2 ;;
+    *)   SO_MINOR_VERSION=0 ;;
+  esac
+fi
+
+RPM_CONFIG_ARGS=
+
+# Memory source/destination managers
+SO_AGE=1
+MEM_SRCDST_FUNCTIONS=
+if test "x${with_jpeg8}" != "xyes"; then
+  AC_MSG_CHECKING([whether to include in-memory source/destination managers])
+  AC_ARG_WITH([mem-srcdst],
+    AC_HELP_STRING([--without-mem-srcdst],
+      [Do not include in-memory source/destination manager functions when emulating the libjpeg v6b or v7 API/ABI]))
+  if test "x$with_mem_srcdst" != "xno"; then
+    AC_MSG_RESULT(yes)
+    AC_DEFINE([MEM_SRCDST_SUPPORTED], [1],
+      [Support in-memory source/destination managers])
+    SO_AGE=2
+    MEM_SRCDST_FUNCTIONS="global:  jpeg_mem_dest;  jpeg_mem_src;";
+  else
+    AC_MSG_RESULT(no)
+    RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-mem-srcdst"
+  fi
+fi
+
+AC_MSG_CHECKING([libjpeg shared library version])
+AC_MSG_RESULT([$SO_MAJOR_VERSION.$SO_AGE.$SO_MINOR_VERSION])
+LIBTOOL_CURRENT=`expr $SO_MAJOR_VERSION + $SO_AGE`
+AC_SUBST(LIBTOOL_CURRENT)
+AC_SUBST(SO_MAJOR_VERSION)
+AC_SUBST(SO_MINOR_VERSION)
+AC_SUBST(SO_AGE)
+AC_SUBST(MEM_SRCDST_FUNCTIONS)
+
+AC_DEFINE_UNQUOTED(LIBJPEG_TURBO_VERSION, [$VERSION], [libjpeg-turbo version])
+
+VERSION_SCRIPT=yes
+AC_ARG_ENABLE([ld-version-script],
+  AS_HELP_STRING([--disable-ld-version-script],
+    [Disable linker version script for libjpeg-turbo (default is to use linker version script if the linker supports it)]),
+  [VERSION_SCRIPT=$enableval], [])
+
+AC_MSG_CHECKING([whether the linker supports version scripts])
+SAVED_LDFLAGS="$LDFLAGS"
+LDFLAGS="$LDFLAGS -Wl,--version-script,conftest.map"
+cat > conftest.map <<EOF
+VERS_1 {
+  global: *;
+};
+EOF
+AC_LINK_IFELSE([AC_LANG_PROGRAM([], [])],
+  [VERSION_SCRIPT_FLAG=-Wl,--version-script,;
+   AC_MSG_RESULT([yes (GNU style)])],
+  [])
+if test "x$VERSION_SCRIPT_FLAG" = "x"; then
+  LDFLAGS="$SAVED_LDFLAGS -Wl,-M,conftest.map"
+  AC_LINK_IFELSE([AC_LANG_PROGRAM([], [])],
+    [VERSION_SCRIPT_FLAG=-Wl,-M,;
+     AC_MSG_RESULT([yes (Sun style)])],
+    [])
+fi
+if test "x$VERSION_SCRIPT_FLAG" = "x"; then
+  VERSION_SCRIPT=no
+  AC_MSG_RESULT(no)
+fi
+LDFLAGS="$SAVED_LDFLAGS"
+
+AC_MSG_CHECKING([whether to use version script when building libjpeg-turbo])
+AC_MSG_RESULT($VERSION_SCRIPT)
+
+AM_CONDITIONAL(VERSION_SCRIPT, test "x$VERSION_SCRIPT" = "xyes")
+AC_SUBST(VERSION_SCRIPT_FLAG)
+
+# Check for non-broken inline under various spellings
+AC_MSG_CHECKING(for inline)
+ljt_cv_inline=""
+AC_TRY_COMPILE(, [} inline __attribute__((always_inline)) int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="inline __attribute__((always_inline))",
+AC_TRY_COMPILE(, [} __inline__ int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline__",
+AC_TRY_COMPILE(, [} __inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline",
+AC_TRY_COMPILE(, [} inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="inline"))))
+AC_MSG_RESULT($ljt_cv_inline)
+AC_DEFINE_UNQUOTED([INLINE],[$ljt_cv_inline],[How to obtain function inlining.])
+
+# Arithmetic coding support
+AC_MSG_CHECKING([whether to include arithmetic encoding support])
+AC_ARG_WITH([arith-enc],
+  AC_HELP_STRING([--without-arith-enc],
+    [Do not include arithmetic encoding support]))
+if test "x$with_12bit" = "xyes"; then
+  with_arith_enc=no
+fi
+if test "x$with_arith_enc" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-arith-enc"
+else
+  AC_DEFINE([C_ARITH_CODING_SUPPORTED], [1], [Support arithmetic encoding])
+  AC_MSG_RESULT(yes)
+fi
+AM_CONDITIONAL([WITH_ARITH_ENC], [test "x$with_arith_enc" != "xno"])
+
+AC_MSG_CHECKING([whether to include arithmetic decoding support])
+AC_ARG_WITH([arith-dec],
+  AC_HELP_STRING([--without-arith-dec],
+    [Do not include arithmetic decoding support]))
+if test "x$with_12bit" = "xyes"; then
+  with_arith_dec=no
+fi
+if test "x$with_arith_dec" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-arith-dec"
+else
+  AC_DEFINE([D_ARITH_CODING_SUPPORTED], [1], [Support arithmetic decoding])
+  AC_MSG_RESULT(yes)
+fi
+AM_CONDITIONAL([WITH_ARITH_DEC], [test "x$with_arith_dec" != "xno"])
+
+AM_CONDITIONAL([WITH_ARITH],
+  [test "x$with_arith_dec" != "xno" -o "x$with_arith_enc" != "xno"])
+
+# 12-bit component support
+AC_MSG_CHECKING([whether to use 12-bit samples])
+AC_ARG_WITH([12bit],
+  AC_HELP_STRING([--with-12bit], [Encode/decode JPEG images with 12-bit samples (implies --without-simd --without-turbojpeg --without-arith-dec --without-arith-enc)]))
+if test "x$with_12bit" = "xyes"; then
+  AC_DEFINE([BITS_IN_JSAMPLE], [12], [use 8 or 12])
+  AC_MSG_RESULT(yes)
+else
+  AC_MSG_RESULT(no)
+fi
+AM_CONDITIONAL([WITH_12BIT], [test "x$with_12bit" = "xyes"])
+
+# TurboJPEG support
+AC_MSG_CHECKING([whether to build TurboJPEG C wrapper])
+AC_ARG_WITH([turbojpeg],
+  AC_HELP_STRING([--without-turbojpeg],
+    [Do not include the TurboJPEG wrapper library and associated test programs]))
+if test "x$with_12bit" = "xyes"; then
+  with_turbojpeg=no
+fi
+if test "x$with_turbojpeg" = "xno"; then
+  AC_MSG_RESULT(no)
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-turbojpeg"
+else
+  AC_MSG_RESULT(yes)
+fi
+
+# Java support
+AC_ARG_VAR(JAVAC, [Java compiler command (default: javac)])
+if test "x$JAVAC" = "x"; then
+  JAVAC=javac
+fi
+AC_SUBST(JAVAC)
+AC_ARG_VAR(JAVACFLAGS, [Java compiler flags])
+AC_SUBST(JAVACFLAGS)
+AC_ARG_VAR(JAR, [Java archive command (default: jar)])
+if test "x$JAR" = "x"; then
+  JAR=jar
+fi
+AC_SUBST(JAR)
+AC_ARG_VAR(JAVA, [Java runtime command (default: java)])
+if test "x$JAVA" = "x"; then
+  JAVA=java
+fi
+AC_SUBST(JAVA)
+AC_ARG_VAR(JNI_CFLAGS,
+  [C compiler flags needed to include jni.h (default: -I/System/Library/Frameworks/JavaVM.framework/Headers on OS X, '-I/usr/java/include -I/usr/java/include/solaris' on Solaris, and '-I/usr/java/default/include -I/usr/java/default/include/linux' on Linux)])
+
+AC_MSG_CHECKING([whether to build TurboJPEG Java wrapper])
+AC_ARG_WITH([java],
+  AC_HELP_STRING([--with-java], [Build Java wrapper for the TurboJPEG library]))
+if test "x$with_12bit" = "xyes" -o "x$with_turbojpeg" = "xno"; then
+  with_java=no
+fi
+
+WITH_JAVA=0
+if test "x$with_java" = "xyes"; then
+  AC_MSG_RESULT(yes)
+
+  case $host_os in
+    darwin*)
+      DEFAULT_JNI_CFLAGS=-I/System/Library/Frameworks/JavaVM.framework/Headers
+      ;;
+    solaris*)
+      DEFAULT_JNI_CFLAGS='-I/usr/java/include -I/usr/java/include/solaris'
+      ;;
+    linux*)
+      DEFAULT_JNI_CFLAGS='-I/usr/java/default/include -I/usr/java/default/include/linux'
+      ;;
+  esac
+  if test "x$JNI_CFLAGS" = "x"; then
+    JNI_CFLAGS=$DEFAULT_JNI_CFLAGS
+  fi
+
+  SAVE_CPPFLAGS=${CPPFLAGS}
+  CPPFLAGS="${CPPFLAGS} ${JNI_CFLAGS}"
+  AC_CHECK_HEADERS([jni.h], [DUMMY=1],
+    [AC_MSG_ERROR([Could not find JNI header file])])
+  CPPFLAGS=${SAVE_CPPFLAGS}
+  AC_SUBST(JNI_CFLAGS)
+
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --with-java"
+  JAVA_RPM_CONTENTS_1='%dir %{_datadir}/classes'
+  JAVA_RPM_CONTENTS_2=%{_datadir}/classes/turbojpeg.jar
+  WITH_JAVA=1
+else
+  AC_MSG_RESULT(no)
+fi
+AM_CONDITIONAL([WITH_JAVA], [test "x$with_java" = "xyes"])
+AC_SUBST(WITH_JAVA)
+AC_SUBST(JAVA_RPM_CONTENTS_1)
+AC_SUBST(JAVA_RPM_CONTENTS_2)
+
+# optionally force using gas-preprocessor.pl for compatibility testing
+AC_ARG_WITH([gas-preprocessor],
+  AC_HELP_STRING([--with-gas-preprocessor],
+    [Force using gas-preprocessor.pl on ARM.]))
+if test "x${with_gas_preprocessor}" = "xyes"; then
+  case $host_os in
+    darwin*)
+      CCAS="gas-preprocessor.pl -fix-unreq $CC"
+      ;;
+    *)
+      CCAS="gas-preprocessor.pl -no-fix-unreq $CC"
+      ;;
+  esac
+  AC_SUBST([CCAS])
+fi
+
+# SIMD is optional
+AC_ARG_WITH([simd],
+  AC_HELP_STRING([--without-simd], [Do not include SIMD extensions]))
+if test "x$with_12bit" = "xyes"; then
+  with_simd=no
+fi
+if test "x${with_simd}" != "xno"; then
+  require_simd=no
+  if test "x${with_simd}" = "xyes"; then
+    require_simd=yes
+  fi
+  # Check if we're on a supported CPU
+  AC_MSG_CHECKING([if we have SIMD optimisations for cpu type])
+  case "$host_cpu" in
+    x86_64 | amd64)
+      AC_MSG_RESULT([yes (x86_64)])
+      AC_PROG_NASM
+      simd_arch=x86_64
+      ;;
+    i*86 | x86 | ia32)
+      AC_MSG_RESULT([yes (i386)])
+      AC_PROG_NASM
+      simd_arch=i386
+      ;;
+    arm*)
+      AC_MSG_RESULT([yes (arm)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_ARM_ASSEMBLER_IFELSE(
+        [if test "x$ac_use_gas_preprocessor" = "xyes"; then
+           AC_MSG_RESULT([yes (with gas-preprocessor)])
+         else
+           AC_MSG_RESULT([yes])
+         fi
+         simd_arch=arm],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    aarch64*)
+      AC_MSG_RESULT([yes (arm64)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_ARM64_ASSEMBLER_IFELSE(
+        [if test "x$ac_use_gas_preprocessor" = "xyes"; then
+           AC_MSG_RESULT([yes (with gas-preprocessor)])
+         else
+           AC_MSG_RESULT([yes])
+         fi
+         simd_arch=aarch64],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    mips*)
+      AC_MSG_RESULT([yes (mips)])
+      AC_MSG_CHECKING([if the assembler is GNU-compatible and can be used])
+      AC_CHECK_COMPATIBLE_MIPS_ASSEMBLER_IFELSE(
+        [AC_MSG_RESULT([yes])
+         simd_arch=mips],
+        [AC_MSG_RESULT([no])
+         with_simd=no])
+      if test "x${with_simd}" = "xno"; then
+        if test "x${require_simd}" = "xyes"; then
+          AC_MSG_ERROR([SIMD support can't be enabled.])
+        else
+          AC_MSG_WARN([SIMD support can't be enabled.  Performance will suffer.])
+        fi
+      fi
+      ;;
+    powerpc*)
+      AC_MSG_RESULT([yes (powerpc)])
+      simd_arch=powerpc
+      ;;
+    *)
+      AC_MSG_RESULT([no ("$host_cpu")])
+      with_simd=no;
+      if test "x${require_simd}" = "xyes"; then
+        AC_MSG_ERROR([SIMD support not available for this CPU.])
+      else
+        AC_MSG_WARN([SIMD support not available for this CPU.  Performance will suffer.])
+      fi
+      ;;
+  esac
+
+  if test "x${with_simd}" != "xno"; then
+    AC_DEFINE([WITH_SIMD], [1], [Use accelerated SIMD routines.])
+  fi
+else
+  RPM_CONFIG_ARGS="$RPM_CONFIG_ARGS --without-simd"
+fi
+
+AM_CONDITIONAL([WITH_SIMD], [test "x$with_simd" != "xno"])
+AM_CONDITIONAL([WITH_SSE_FLOAT_DCT], [test "x$simd_arch" = "xx86_64" -o "x$simd_arch" = "xi386"])
+AM_CONDITIONAL([SIMD_I386], [test "x$simd_arch" = "xi386"])
+AM_CONDITIONAL([SIMD_X86_64], [test "x$simd_arch" = "xx86_64"])
+AM_CONDITIONAL([SIMD_ARM], [test "x$simd_arch" = "xarm"])
+AM_CONDITIONAL([SIMD_ARM_64], [test "x$simd_arch" = "xaarch64"])
+AM_CONDITIONAL([SIMD_MIPS], [test "x$simd_arch" = "xmips"])
+AM_CONDITIONAL([SIMD_POWERPC], [test "x$simd_arch" = "xpowerpc"])
+AM_CONDITIONAL([X86_64], [test "x$host_cpu" = "xx86_64" -o "x$host_cpu" = "xamd64"])
+AM_CONDITIONAL([WITH_TURBOJPEG], [test "x$with_turbojpeg" != "xno"])
+AM_CONDITIONAL([CROSS_COMPILING], [test "x$cross_compiling" = "xyes"])
+
+AC_ARG_VAR(PKGNAME, [distribution package name (default: libjpeg-turbo)])
+if test "x$PKGNAME" = "x"; then
+  PKGNAME=$PACKAGE_NAME
+fi
+AC_SUBST(PKGNAME)
+
+case "$host_cpu" in
+  x86_64)
+    RPMARCH=x86_64
+    DEBARCH=amd64
+    ;;
+  i*86 | x86 | ia32)
+    RPMARCH=i386
+    DEBARCH=i386
+    ;;
+  *)
+    RPMARCH=`uname -m`
+    DEBARCH=$RPMARCH
+    ;;
+esac
+
+if test "${docdir}" = ""; then
+  docdir=${datadir}/doc
+  AC_SUBST(docdir)
+fi
+
+AC_SUBST(RPMARCH)
+AC_SUBST(RPM_CONFIG_ARGS)
+AC_SUBST(DEBARCH)
+AC_SUBST(BUILD)
+AC_DEFINE_UNQUOTED([BUILD], "$BUILD", [libjpeg-turbo build number])
+
+# NOTE: autoheader automatically modifies the input file of the first
+# invocation of AC_CONFIG_HEADERS, so we put config.h first to prevent
+# jconfig.h.in from being clobbered.  config.h is used only internally, whereas
+# jconfig.h contains macros that are relevant to external programs (macros that
+# specify which features were built into the library.)
+AC_CONFIG_HEADERS([config.h])
+AC_CONFIG_HEADERS([jconfig.h])
+AC_CONFIG_HEADERS([jconfigint.h])
+AC_CONFIG_FILES([pkgscripts/libjpeg-turbo.spec.tmpl:release/libjpeg-turbo.spec.in])
+AC_CONFIG_FILES([pkgscripts/makecygwinpkg.tmpl:release/makecygwinpkg.in])
+AC_CONFIG_FILES([pkgscripts/makedpkg.tmpl:release/makedpkg.in])
+AC_CONFIG_FILES([pkgscripts/makemacpkg.tmpl:release/makemacpkg.in])
+AC_CONFIG_FILES([pkgscripts/uninstall.tmpl:release/uninstall.in])
+AC_CONFIG_FILES([pkgscripts/libjpeg.pc:release/libjpeg.pc.in])
+AC_CONFIG_FILES([pkgscripts/libturbojpeg.pc:release/libturbojpeg.pc.in])
+if test "x$with_turbojpeg" != "xno"; then
+  AC_CONFIG_FILES([tjbenchtest])
+fi
+if test "x$with_java" = "xyes"; then
+  AC_CONFIG_FILES([tjbenchtest.java])
+  AC_CONFIG_FILES([tjexampletest])
+fi
+AC_CONFIG_FILES([libjpeg.map])
+AC_CONFIG_FILES([Makefile simd/Makefile])
+AC_CONFIG_FILES([java/Makefile])
+AC_CONFIG_FILES([md5/Makefile])
+AC_OUTPUT
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/djpeg.1 glmark2-2021.02/src/libjpeg-turbo/djpeg.1
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/djpeg.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/djpeg.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,292 @@
+.TH DJPEG 1 "18 February 2016"
+.SH NAME
+djpeg \- decompress a JPEG file to an image file
+.SH SYNOPSIS
+.B djpeg
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B djpeg
+decompresses the named JPEG file, or the standard input if no file is named,
+and produces an image file on the standard output.  PBMPLUS (PPM/PGM), BMP,
+GIF, Targa, or RLE (Utah Raster Toolkit) output format can be selected.
+(RLE is supported only if the URT library is available.)
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-grayscale
+may be written
+.B \-gray
+or
+.BR \-gr .
+Most of the "basic" switches can be abbreviated to as little as one letter.
+Upper and lower case are equivalent (thus
+.B \-BMP
+is the same as
+.BR \-bmp ).
+British spellings are also accepted (e.g.,
+.BR \-greyscale ),
+though for brevity these are not mentioned below.
+.PP
+The basic switches are:
+.TP
+.BI \-colors " N"
+Reduce image to at most N colors.  This reduces the number of colors used in
+the output image, so that it can be displayed on a colormapped display or
+stored in a colormapped file format.  For example, if you have an 8-bit
+display, you'd need to reduce to 256 or fewer colors.
+.TP
+.BI \-quantize " N"
+Same as
+.BR \-colors .
+.B \-colors
+is the recommended name,
+.B \-quantize
+is provided only for backwards compatibility.
+.TP
+.B \-fast
+Select recommended processing options for fast, low quality output.  (The
+default options are chosen for highest quality output.)  Currently, this is
+equivalent to \fB\-dct fast \-nosmooth \-onepass \-dither ordered\fR.
+.TP
+.B \-grayscale
+Force grayscale output even if JPEG file is color.  Useful for viewing on
+monochrome displays; also,
+.B djpeg
+runs noticeably faster in this mode.
+.TP
+.B \-rgb
+Force RGB output even if JPEG file is grayscale.
+.TP
+.BI \-scale " M/N"
+Scale the output image by a factor M/N.  Currently the scale factor must be
+M/8, where M is an integer between 1 and 16 inclusive, or any reduced fraction
+thereof (such as 1/2, 3/4, etc.)  Scaling is handy if the image is larger than
+your screen; also,
+.B djpeg
+runs much faster when scaling down the output.
+.TP
+.B \-bmp
+Select BMP output format (Windows flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-gif
+Select GIF output format.  Since GIF does not support more than 256 colors,
+.B \-colors 256
+is assumed (unless you specify a smaller number of colors).
+.TP
+.B \-os2
+Select BMP output format (OS/2 1.x flavor).  8-bit colormapped format is
+emitted if
+.B \-colors
+or
+.B \-grayscale
+is specified, or if the JPEG file is grayscale; otherwise, 24-bit full-color
+format is emitted.
+.TP
+.B \-pnm
+Select PBMPLUS (PPM/PGM) output format (this is the default format).
+PGM is emitted if the JPEG file is grayscale or if
+.B \-grayscale
+is specified; otherwise PPM is emitted.
+.TP
+.B \-rle
+Select RLE output format.  (Requires URT library.)
+.TP
+.B \-targa
+Select Targa output format.  Grayscale format is emitted if the JPEG file is
+grayscale or if
+.B \-grayscale
+is specified; otherwise, colormapped format is emitted if
+.B \-colors
+is specified; otherwise, 24-bit full-color format is emitted.
+.PP
+Switches for advanced users:
+.TP
+.B \-dct int
+Use integer DCT method (default).
+.TP
+.B \-dct fast
+Use fast integer DCT (less accurate).
+In libjpeg-turbo, the fast method is generally about 5-15% faster than the int
+method when using the x86/x86-64 SIMD extensions (results may vary with other
+SIMD implementations, or when using libjpeg-turbo without SIMD extensions.)  If
+the JPEG image was compressed using a quality level of 85 or below, then there
+should be little or no perceptible difference between the two algorithms.  When
+decompressing images that were compressed using quality levels above 85,
+however, the difference between the fast and int methods becomes more
+pronounced.  With images compressed using quality=97, for instance, the fast
+method incurs generally about a 4-6 dB loss (in PSNR) relative to the int
+method, but this can be larger for some images.  If you can avoid it, do not
+use the fast method when decompressing images that were compressed using
+quality levels above 97.  The algorithm often degenerates for such images and
+can actually produce a more lossy output image than if the JPEG image had been
+compressed using lower quality levels.
+.TP
+.B \-dct float
+Use floating-point DCT method.
+The float method is mainly a legacy feature.  It does not produce significantly
+more accurate results than the int method, and it is much slower.  The float
+method may also give different results on different machines due to varying
+roundoff behavior, whereas the integer methods should give the same results on
+all machines.
+.TP
+.B \-dither fs
+Use Floyd-Steinberg dithering in color quantization.
+.TP
+.B \-dither ordered
+Use ordered dithering in color quantization.
+.TP
+.B \-dither none
+Do not use dithering in color quantization.
+By default, Floyd-Steinberg dithering is applied when quantizing colors; this
+is slow but usually produces the best results.  Ordered dither is a compromise
+between speed and quality; no dithering is fast but usually looks awful.  Note
+that these switches have no effect unless color quantization is being done.
+Ordered dither is only available in
+.B \-onepass
+mode.
+.TP
+.BI \-map " file"
+Quantize to the colors used in the specified image file.  This is useful for
+producing multiple files with identical color maps, or for forcing a
+predefined set of colors to be used.  The
+.I file
+must be a GIF or PPM file. This option overrides
+.B \-colors
+and
+.BR \-onepass .
+.TP
+.B \-nosmooth
+Use a faster, lower-quality upsampling routine.
+.TP
+.B \-onepass
+Use one-pass instead of two-pass color quantization.  The one-pass method is
+faster and needs less memory, but it produces a lower-quality image.
+.B \-onepass
+is ignored unless you also say
+.B \-colors
+.IR N .
+Also, the one-pass method is always used for grayscale output (the two-pass
+method is no improvement then).
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.BI \-memsrc
+Load input file into memory before decompressing.  This feature was implemented
+mainly as a way of testing the in-memory source manager (jpeg_mem_src().)
+.TP
+.BI \-skip " Y0,Y1"
+Decompress all rows of the JPEG image except those between Y0 and Y1
+(inclusive.)  Note that if decompression scaling is being used, then Y0 and Y1
+are relative to the scaled image dimensions.
+.TP
+.BI \-crop " WxH+X+Y"
+Decompress only a rectangular subregion of the image, starting at point X,Y
+with width W and height H.  If necessary, X will be shifted left to the nearest
+iMCU boundary, and the width will be increased accordingly.  Note that if
+decompression scaling is being used, then X, Y, W, and H are relative to the
+scaled image dimensions.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example decompresses the JPEG file foo.jpg, quantizes it to
+256 colors, and saves the output in 8-bit BMP format in foo.bmp:
+.IP
+.B djpeg \-colors 256 \-bmp
+.I foo.jpg
+.B >
+.I foo.bmp
+.SH HINTS
+To get a quick preview of an image, use the
+.B \-grayscale
+and/or
+.B \-scale
+switches.
+.B \-grayscale \-scale 1/8
+is the fastest case.
+.PP
+Several options are available that trade off image quality to gain speed.
+.B \-fast
+turns on the recommended settings.
+.PP
+.B \-dct fast
+and/or
+.B \-nosmooth
+gain speed at a small sacrifice in quality.
+When producing a color-quantized image,
+.B \-onepass \-dither ordered
+is fast but much lower quality than the default behavior.
+.B \-dither none
+may give acceptable results in two-pass mode, but is seldom tolerable in
+one-pass mode.
+.PP
+If you are fortunate enough to have very fast floating point hardware,
+\fB\-dct float\fR may be even faster than \fB\-dct fast\fR.  But on most
+machines \fB\-dct float\fR is slower than \fB\-dct int\fR; in this case it is
+not worth using, because its theoretical accuracy advantage is too small to be
+significant in practice.
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+.BR ppm (5),
+.BR pgm (5)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo, to wordsmith certain sections, and to describe
+features not present in libjpeg.
+.SH ISSUES
+Support for compressed GIF output files was removed in djpeg v6b due to
+concerns over the Unisys LZW patent.  Although this patent expired in 2006,
+djpeg still lacks compressed GIF support, for these historical reasons.
+(Conversion of JPEG files to GIF is usually a bad idea anyway, since GIF is a
+256-color format.)  The uncompressed GIF files that djpeg generates are larger
+than they should be, but they are readable by standard GIF decoders.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/djpeg.c glmark2-2021.02/src/libjpeg-turbo/djpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/djpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/djpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,781 @@
+/*
+ * djpeg.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2013 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2013-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for the JPEG decompressor.
+ * It should work on any system with Unix- or MS-DOS-style command lines.
+ *
+ * Two different command line styles are permitted, depending on the
+ * compile-time switch TWO_FILE_COMMANDLINE:
+ *      djpeg [options]  inputfile outputfile
+ *      djpeg [options]  [inputfile]
+ * In the second style, output is always to standard output, which you'd
+ * normally redirect to a file or pipe to some other program.  Input is
+ * either from a named file or from standard input (typically redirected).
+ * The second style is convenient on Unix but is unhelpful on systems that
+ * don't support pipes.  Also, you MUST use the first style if your system
+ * doesn't do binary I/O to stdin/stdout.
+ * To simplify script writing, the "-outfile" switch is provided.  The syntax
+ *      djpeg [options]  -outfile outputfile  inputfile
+ * works regardless of which command line style is used.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+#include "wrppm.h"
+
+#include <ctype.h>              /* to declare isprint() */
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/* Create the add-on message string table. */
+
+#define JMESSAGE(code,string)   string ,
+
+static const char * const cdjpeg_message_table[] = {
+#include "cderror.h"
+  NULL
+};
+
+
+/*
+ * This list defines the known output image formats
+ * (not all of which need be supported by a given version).
+ * You can change the default output format by defining DEFAULT_FMT;
+ * indeed, you had better do so if you undefine PPM_SUPPORTED.
+ */
+
+typedef enum {
+        FMT_BMP,                /* BMP format (Windows flavor) */
+        FMT_GIF,                /* GIF format */
+        FMT_OS2,                /* BMP format (OS/2 flavor) */
+        FMT_PPM,                /* PPM/PGM (PBMPLUS formats) */
+        FMT_RLE,                /* RLE format */
+        FMT_TARGA,              /* Targa format */
+        FMT_TIFF                /* TIFF format */
+} IMAGE_FORMATS;
+
+#ifndef DEFAULT_FMT             /* so can override from CFLAGS in Makefile */
+#define DEFAULT_FMT     FMT_PPM
+#endif
+
+static IMAGE_FORMATS requested_fmt;
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+boolean memsrc;                 /* for -memsrc switch */
+boolean skip, crop;
+JDIMENSION skip_start, skip_end;
+JDIMENSION crop_x, crop_y, crop_width, crop_height;
+#define INPUT_BUF_SIZE  4096
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -colors N      Reduce image to no more than N colors\n");
+  fprintf(stderr, "  -fast          Fast, low-quality processing\n");
+  fprintf(stderr, "  -grayscale     Force grayscale output\n");
+  fprintf(stderr, "  -rgb           Force RGB output\n");
+  fprintf(stderr, "  -rgb565        Force RGB565 output\n");
+#ifdef IDCT_SCALING_SUPPORTED
+  fprintf(stderr, "  -scale M/N     Scale output image by fraction M/N, eg, 1/8\n");
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -bmp           Select BMP output format (Windows style)%s\n",
+          (DEFAULT_FMT == FMT_BMP ? " (default)" : ""));
+#endif
+#ifdef GIF_SUPPORTED
+  fprintf(stderr, "  -gif           Select GIF output format%s\n",
+          (DEFAULT_FMT == FMT_GIF ? " (default)" : ""));
+#endif
+#ifdef BMP_SUPPORTED
+  fprintf(stderr, "  -os2           Select BMP output format (OS/2 style)%s\n",
+          (DEFAULT_FMT == FMT_OS2 ? " (default)" : ""));
+#endif
+#ifdef PPM_SUPPORTED
+  fprintf(stderr, "  -pnm           Select PBMPLUS (PPM/PGM) output format%s\n",
+          (DEFAULT_FMT == FMT_PPM ? " (default)" : ""));
+#endif
+#ifdef RLE_SUPPORTED
+  fprintf(stderr, "  -rle           Select Utah RLE output format%s\n",
+          (DEFAULT_FMT == FMT_RLE ? " (default)" : ""));
+#endif
+#ifdef TARGA_SUPPORTED
+  fprintf(stderr, "  -targa         Select Targa output format%s\n",
+          (DEFAULT_FMT == FMT_TARGA ? " (default)" : ""));
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef DCT_ISLOW_SUPPORTED
+  fprintf(stderr, "  -dct int       Use integer DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_ISLOW ? " (default)" : ""));
+#endif
+#ifdef DCT_IFAST_SUPPORTED
+  fprintf(stderr, "  -dct fast      Use fast integer DCT (less accurate)%s\n",
+          (JDCT_DEFAULT == JDCT_IFAST ? " (default)" : ""));
+#endif
+#ifdef DCT_FLOAT_SUPPORTED
+  fprintf(stderr, "  -dct float     Use floating-point DCT method%s\n",
+          (JDCT_DEFAULT == JDCT_FLOAT ? " (default)" : ""));
+#endif
+  fprintf(stderr, "  -dither fs     Use F-S dithering (default)\n");
+  fprintf(stderr, "  -dither none   Don't use dithering in quantization\n");
+  fprintf(stderr, "  -dither ordered  Use ordered dither (medium speed, quality)\n");
+#ifdef QUANT_2PASS_SUPPORTED
+  fprintf(stderr, "  -map FILE      Map to colors used in named image file\n");
+#endif
+  fprintf(stderr, "  -nosmooth      Don't use high-quality upsampling\n");
+#ifdef QUANT_1PASS_SUPPORTED
+  fprintf(stderr, "  -onepass       Use 1-pass quantization (fast, low quality)\n");
+#endif
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  fprintf(stderr, "  -memsrc        Load input file into memory before decompressing\n");
+#endif
+
+  fprintf(stderr, "  -skip Y0,Y1    Decompress all rows except those between Y0 and Y1 (inclusive)\n");
+  fprintf(stderr, "  -crop WxH+X+Y  Decompress only a rectangular subregion of the image\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(int)
+parse_switches (j_decompress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+
+  /* Set up default JPEG parameters. */
+  requested_fmt = DEFAULT_FMT;  /* set default output file format */
+  outfilename = NULL;
+  memsrc = FALSE;
+  skip = FALSE;
+  crop = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "bmp", 1)) {
+      /* BMP output format. */
+      requested_fmt = FMT_BMP;
+
+    } else if (keymatch(arg, "colors", 1) || keymatch(arg, "colours", 1) ||
+               keymatch(arg, "quantize", 1) || keymatch(arg, "quantise", 1)) {
+      /* Do color quantization. */
+      int val;
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%d", &val) != 1)
+        usage();
+      cinfo->desired_number_of_colors = val;
+      cinfo->quantize_colors = TRUE;
+
+    } else if (keymatch(arg, "dct", 2)) {
+      /* Select IDCT algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "int", 1)) {
+        cinfo->dct_method = JDCT_ISLOW;
+      } else if (keymatch(argv[argn], "fast", 2)) {
+        cinfo->dct_method = JDCT_IFAST;
+      } else if (keymatch(argv[argn], "float", 2)) {
+        cinfo->dct_method = JDCT_FLOAT;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "dither", 2)) {
+      /* Select dithering algorithm. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "fs", 2)) {
+        cinfo->dither_mode = JDITHER_FS;
+      } else if (keymatch(argv[argn], "none", 2)) {
+        cinfo->dither_mode = JDITHER_NONE;
+      } else if (keymatch(argv[argn], "ordered", 2)) {
+        cinfo->dither_mode = JDITHER_ORDERED;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "fast", 1)) {
+      /* Select recommended processing options for quick-and-dirty output. */
+      cinfo->two_pass_quantize = FALSE;
+      cinfo->dither_mode = JDITHER_ORDERED;
+      if (! cinfo->quantize_colors) /* don't override an earlier -colors */
+        cinfo->desired_number_of_colors = 216;
+      cinfo->dct_method = JDCT_FASTEST;
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "gif", 1)) {
+      /* GIF output format. */
+      requested_fmt = FMT_GIF;
+
+    } else if (keymatch(arg, "grayscale", 2) || keymatch(arg, "greyscale",2)) {
+      /* Force monochrome output. */
+      cinfo->out_color_space = JCS_GRAYSCALE;
+
+    } else if (keymatch(arg, "rgb", 2)) {
+      /* Force RGB output. */
+      cinfo->out_color_space = JCS_RGB;
+
+    } else if (keymatch(arg, "rgb565", 2)) {
+      /* Force RGB565 output. */
+      cinfo->out_color_space = JCS_RGB565;
+
+    } else if (keymatch(arg, "map", 3)) {
+      /* Quantize to a color map taken from an input file. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (for_real) {           /* too expensive to do twice! */
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+        FILE *mapfile;
+
+        if ((mapfile = fopen(argv[argn], READ_BINARY)) == NULL) {
+          fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+          exit(EXIT_FAILURE);
+        }
+        read_color_map(cinfo, mapfile);
+        fclose(mapfile);
+        cinfo->quantize_colors = TRUE;
+#else
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+#endif
+      }
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "nosmooth", 3)) {
+      /* Suppress fancy upsampling */
+      cinfo->do_fancy_upsampling = FALSE;
+
+    } else if (keymatch(arg, "onepass", 3)) {
+      /* Use fast one-pass quantization. */
+      cinfo->two_pass_quantize = FALSE;
+
+    } else if (keymatch(arg, "os2", 3)) {
+      /* BMP output format (OS/2 flavor). */
+      requested_fmt = FMT_OS2;
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "memsrc", 2)) {
+      /* Use in-memory source manager */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+      memsrc = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, in-memory source manager was not compiled in\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "pnm", 1) || keymatch(arg, "ppm", 1)) {
+      /* PPM/PGM output format. */
+      requested_fmt = FMT_PPM;
+
+    } else if (keymatch(arg, "rle", 1)) {
+      /* RLE output format. */
+      requested_fmt = FMT_RLE;
+
+    } else if (keymatch(arg, "scale", 2)) {
+      /* Scale the output image by a fraction M/N. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%u/%u",
+                 &cinfo->scale_num, &cinfo->scale_denom) != 2)
+        usage();
+
+    } else if (keymatch(arg, "skip", 2)) {
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argv[argn], "%u,%u", &skip_start, &skip_end) != 2 ||
+          skip_start > skip_end)
+        usage();
+      skip = TRUE;
+
+    } else if (keymatch(arg, "crop", 2)) {
+      char c;
+      if (++argn >= argc)
+        usage();
+      if (sscanf(argv[argn], "%u%c%u+%u+%u", &crop_width, &c, &crop_height,
+                 &crop_x, &crop_y) != 5 ||
+          (c != 'X' && c != 'x') || crop_width < 1 || crop_height < 1)
+        usage();
+      crop = TRUE;
+
+    } else if (keymatch(arg, "targa", 1)) {
+      /* Targa output format. */
+      requested_fmt = FMT_TARGA;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * Marker processor for COM and interesting APPn markers.
+ * This replaces the library's built-in processor, which just skips the marker.
+ * We want to print out the marker as text, to the extent possible.
+ * Note this code relies on a non-suspending data source.
+ */
+
+LOCAL(unsigned int)
+jpeg_getc (j_decompress_ptr cinfo)
+/* Read next byte */
+{
+  struct jpeg_source_mgr *datasrc = cinfo->src;
+
+  if (datasrc->bytes_in_buffer == 0) {
+    if (! (*datasrc->fill_input_buffer) (cinfo))
+      ERREXIT(cinfo, JERR_CANT_SUSPEND);
+  }
+  datasrc->bytes_in_buffer--;
+  return GETJOCTET(*datasrc->next_input_byte++);
+}
+
+
+METHODDEF(boolean)
+print_text_marker (j_decompress_ptr cinfo)
+{
+  boolean traceit = (cinfo->err->trace_level >= 1);
+  long length;
+  unsigned int ch;
+  unsigned int lastch = 0;
+
+  length = jpeg_getc(cinfo) << 8;
+  length += jpeg_getc(cinfo);
+  length -= 2;                  /* discount the length word itself */
+
+  if (traceit) {
+    if (cinfo->unread_marker == JPEG_COM)
+      fprintf(stderr, "Comment, length %ld:\n", (long) length);
+    else                        /* assume it is an APPn otherwise */
+      fprintf(stderr, "APP%d, length %ld:\n",
+              cinfo->unread_marker - JPEG_APP0, (long) length);
+  }
+
+  while (--length >= 0) {
+    ch = jpeg_getc(cinfo);
+    if (traceit) {
+      /* Emit the character in a readable form.
+       * Nonprintables are converted to \nnn form,
+       * while \ is converted to \\.
+       * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+       */
+      if (ch == '\r') {
+        fprintf(stderr, "\n");
+      } else if (ch == '\n') {
+        if (lastch != '\r')
+          fprintf(stderr, "\n");
+      } else if (ch == '\\') {
+        fprintf(stderr, "\\\\");
+      } else if (isprint(ch)) {
+        putc(ch, stderr);
+      } else {
+        fprintf(stderr, "\\%03o", ch);
+      }
+      lastch = ch;
+    }
+  }
+
+  if (traceit)
+    fprintf(stderr, "\n");
+
+  return TRUE;
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct cinfo;
+  struct jpeg_error_mgr jerr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  int file_index;
+  djpeg_dest_ptr dest_mgr = NULL;
+  FILE *input_file;
+  FILE *output_file;
+  unsigned char *inbuffer = NULL;
+  unsigned long insize = 0;
+  JDIMENSION num_scanlines;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "djpeg";         /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  cinfo.err = jpeg_std_error(&jerr);
+  jpeg_create_decompress(&cinfo);
+  /* Add some application-specific error messages (from cderror.h) */
+  jerr.addon_message_table = cdjpeg_message_table;
+  jerr.first_addon_message = JMSG_FIRSTADDONCODE;
+  jerr.last_addon_message = JMSG_LASTADDONCODE;
+
+  /* Insert custom marker processor for COM and APP12.
+   * APP12 is used by some digital camera makers for textual info,
+   * so we provide the ability to display it as text.
+   * If you like, additional APPn marker types can be selected for display,
+   * but don't try to override APP0 or APP14 this way (see libjpeg.txt).
+   */
+  jpeg_set_marker_processor(&cinfo, JPEG_COM, print_text_marker);
+  jpeg_set_marker_processor(&cinfo, JPEG_APP0+12, print_text_marker);
+
+  /* Scan command line to find file names. */
+  /* It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are ignored; we will rescan the switches after opening
+   * the input file.
+   * (Exception: tracing level set here controls verbosity for COM markers
+   * found during jpeg_read_header...)
+   */
+
+  file_index = parse_switches(&cinfo, argc, argv, 0, FALSE);
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((input_file = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    input_file = read_stdin();
+  }
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((output_file = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    output_file = write_stdout();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &cinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+  if (memsrc) {
+    size_t nbytes;
+    do {
+      inbuffer = (unsigned char *)realloc(inbuffer, insize + INPUT_BUF_SIZE);
+      if (inbuffer == NULL) {
+        fprintf(stderr, "%s: memory allocation failure\n", progname);
+        exit(EXIT_FAILURE);
+      }
+      nbytes = JFREAD(input_file, &inbuffer[insize], INPUT_BUF_SIZE);
+      if (nbytes < INPUT_BUF_SIZE && ferror(input_file)) {
+        if (file_index < argc)
+          fprintf(stderr, "%s: can't read from %s\n", progname,
+                  argv[file_index]);
+        else
+          fprintf(stderr, "%s: can't read from stdin\n", progname);
+      }
+      insize += (unsigned long)nbytes;
+    } while (nbytes == INPUT_BUF_SIZE);
+    fprintf(stderr, "Compressed size:  %lu bytes\n", insize);
+    jpeg_mem_src(&cinfo, inbuffer, insize);
+  } else
+#endif
+    jpeg_stdio_src(&cinfo, input_file);
+
+  /* Read file header, set default decompression parameters */
+  (void) jpeg_read_header(&cinfo, TRUE);
+
+  /* Adjust default decompression parameters by re-parsing the options */
+  file_index = parse_switches(&cinfo, argc, argv, 0, TRUE);
+
+  /* Initialize the output module now to let it override any crucial
+   * option settings (for instance, GIF wants to force color quantization).
+   */
+  switch (requested_fmt) {
+#ifdef BMP_SUPPORTED
+  case FMT_BMP:
+    dest_mgr = jinit_write_bmp(&cinfo, FALSE);
+    break;
+  case FMT_OS2:
+    dest_mgr = jinit_write_bmp(&cinfo, TRUE);
+    break;
+#endif
+#ifdef GIF_SUPPORTED
+  case FMT_GIF:
+    dest_mgr = jinit_write_gif(&cinfo);
+    break;
+#endif
+#ifdef PPM_SUPPORTED
+  case FMT_PPM:
+    dest_mgr = jinit_write_ppm(&cinfo);
+    break;
+#endif
+#ifdef RLE_SUPPORTED
+  case FMT_RLE:
+    dest_mgr = jinit_write_rle(&cinfo);
+    break;
+#endif
+#ifdef TARGA_SUPPORTED
+  case FMT_TARGA:
+    dest_mgr = jinit_write_targa(&cinfo);
+    break;
+#endif
+  default:
+    ERREXIT(&cinfo, JERR_UNSUPPORTED_FORMAT);
+    break;
+  }
+  dest_mgr->output_file = output_file;
+
+  /* Start decompressor */
+  (void) jpeg_start_decompress(&cinfo);
+
+  /* Skip rows */
+  if (skip) {
+    JDIMENSION tmp;
+
+    /* Check for valid skip_end.  We cannot check this value until after
+     * jpeg_start_decompress() is called.  Note that we have already verified
+     * that skip_start <= skip_end.
+     */
+    if (skip_end > cinfo.output_height - 1) {
+      fprintf(stderr, "%s: skip region exceeds image height %d\n", progname,
+              cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height -= (skip_end - skip_start + 1);
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    while (cinfo.output_scanline < skip_start) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    jpeg_skip_scanlines(&cinfo, skip_end - skip_start + 1);
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+
+  /* Decompress a subregion */
+  } else if (crop) {
+    JDIMENSION tmp;
+
+    /* Check for valid crop dimensions.  We cannot check these values until
+     * after jpeg_start_decompress() is called.
+     */
+    if (crop_x + crop_width > cinfo.output_width ||
+        crop_y + crop_height > cinfo.output_height) {
+      fprintf(stderr, "%s: crop dimensions exceed image dimensions %d x %d\n",
+              progname, cinfo.output_width, cinfo.output_height);
+      exit(EXIT_FAILURE);
+    }
+
+    jpeg_crop_scanline(&cinfo, &crop_x, &crop_width);
+    ((ppm_dest_ptr) dest_mgr)->buffer_width = cinfo.output_width *
+                                              cinfo.out_color_components *
+                                              sizeof(JSAMPLE);
+
+    /* Write output file header.  This is a hack to ensure that the destination
+     * manager creates an output image of the proper size.
+     */
+    tmp = cinfo.output_height;
+    cinfo.output_height = crop_height;
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+    cinfo.output_height = tmp;
+
+    /* Process data */
+    jpeg_skip_scanlines(&cinfo, crop_y);
+    while (cinfo.output_scanline < crop_y + crop_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+    jpeg_skip_scanlines(&cinfo, cinfo.output_height - crop_y - crop_height);
+
+  /* Normal full-image decompress */
+  } else {
+    /* Write output file header */
+    (*dest_mgr->start_output) (&cinfo, dest_mgr);
+
+    /* Process data */
+    while (cinfo.output_scanline < cinfo.output_height) {
+      num_scanlines = jpeg_read_scanlines(&cinfo, dest_mgr->buffer,
+                                          dest_mgr->buffer_height);
+      (*dest_mgr->put_pixel_rows) (&cinfo, dest_mgr, num_scanlines);
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  /* Hack: count final pass as done in case finish_output does an extra pass.
+   * The library won't have updated completed_passes.
+   */
+  progress.pub.completed_passes = progress.pub.total_passes;
+#endif
+
+  /* Finish decompression and release memory.
+   * I must do it in this order because output module has allocated memory
+   * of lifespan JPOOL_IMAGE; it needs to finish before releasing memory.
+   */
+  (*dest_mgr->finish_output) (&cinfo, dest_mgr);
+  (void) jpeg_finish_decompress(&cinfo);
+  jpeg_destroy_decompress(&cinfo);
+
+  /* Close files, if we opened them */
+  if (input_file != stdin)
+    fclose(input_file);
+  if (output_file != stdout)
+    fclose(output_file);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &cinfo);
+#endif
+
+  if (memsrc && inbuffer != NULL)
+    free(inbuffer);
+
+  /* All done. */
+  exit(jerr.num_warnings ? EXIT_WARNING : EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/doxygen.config glmark2-2021.02/src/libjpeg-turbo/doxygen.config
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/doxygen.config	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/doxygen.config	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,16 @@
+PROJECT_NAME = TurboJPEG
+PROJECT_NUMBER = 1.5
+OUTPUT_DIRECTORY = doc/
+USE_WINDOWS_ENCODING = NO
+OPTIMIZE_OUTPUT_FOR_C = YES
+WARN_NO_PARAMDOC = YES
+GENERATE_LATEX = NO
+FILE_PATTERNS = turbojpeg.h
+HIDE_UNDOC_MEMBERS = YES
+VERBATIM_HEADERS = NO
+EXTRACT_STATIC = YES
+JAVADOC_AUTOBRIEF = YES
+MAX_INITIALIZER_LINES = 0
+ALWAYS_DETAILED_SEC = YES
+HTML_TIMESTAMP = NO
+HTML_EXTRA_STYLESHEET = doxygen-extra.css
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/doxygen-extra.css glmark2-2021.02/src/libjpeg-turbo/doxygen-extra.css
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/doxygen-extra.css	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/doxygen-extra.css	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3 @@
+code {
+	color: #4665A2; 
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/example.c glmark2-2021.02/src/libjpeg-turbo/example.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/example.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/example.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,433 @@
+/*
+ * example.c
+ *
+ * This file illustrates how to use the IJG code as a subroutine library
+ * to read or write JPEG image files.  You should look at this code in
+ * conjunction with the documentation file libjpeg.txt.
+ *
+ * This code will not do anything useful as-is, but it may be helpful as a
+ * skeleton for constructing routines that call the JPEG library.
+ *
+ * We present these routines in the same coding style used in the JPEG code
+ * (ANSI function definitions, etc); but you are of course free to code your
+ * routines in a different style if you prefer.
+ */
+
+#include <stdio.h>
+
+/*
+ * Include file for users of JPEG library.
+ * You will need to have included system headers that define at least
+ * the typedefs FILE and size_t before you can include jpeglib.h.
+ * (stdio.h is sufficient on ANSI-conforming systems.)
+ * You may also wish to include "jerror.h".
+ */
+
+#include "jpeglib.h"
+
+/*
+ * <setjmp.h> is used for the optional error recovery mechanism shown in
+ * the second part of the example.
+ */
+
+#include <setjmp.h>
+
+
+
+/******************** JPEG COMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to feed data into the JPEG compressor.
+ * We present a minimal version that does not worry about refinements such
+ * as error recovery (the JPEG code will just exit() if it gets an error).
+ */
+
+
+/*
+ * IMAGE DATA FORMATS:
+ *
+ * The standard input image format is a rectangular array of pixels, with
+ * each pixel having the same number of "component" values (color channels).
+ * Each pixel row is an array of JSAMPLEs (which typically are unsigned chars).
+ * If you are working with color data, then the color values for each pixel
+ * must be adjacent in the row; for example, R,G,B,R,G,B,R,G,B,... for 24-bit
+ * RGB color.
+ *
+ * For this example, we'll assume that this data structure matches the way
+ * our application has stored the image in memory, so we can just pass a
+ * pointer to our image buffer.  In particular, let's say that the image is
+ * RGB color and is described by:
+ */
+
+extern JSAMPLE *image_buffer;   /* Points to large array of R,G,B-order data */
+extern int image_height;        /* Number of rows in image */
+extern int image_width;         /* Number of columns in image */
+
+
+/*
+ * Sample routine for JPEG compression.  We assume that the target file name
+ * and a compression quality factor are passed in.
+ */
+
+GLOBAL(void)
+write_JPEG_file (char *filename, int quality)
+{
+  /* This struct contains the JPEG compression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   * It is possible to have several such structures, representing multiple
+   * compression/decompression processes, in existence at once.  We refer
+   * to any one struct (and its associated working data) as a "JPEG object".
+   */
+  struct jpeg_compress_struct cinfo;
+  /* This struct represents a JPEG error handler.  It is declared separately
+   * because applications often want to supply a specialized error handler
+   * (see the second half of this file for an example).  But here we just
+   * take the easy way out and use the standard error handler, which will
+   * print a message on stderr and call exit() if compression fails.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct jpeg_error_mgr jerr;
+  /* More stuff */
+  FILE *outfile;                /* target file */
+  JSAMPROW row_pointer[1];      /* pointer to JSAMPLE row[s] */
+  int row_stride;               /* physical row width in image buffer */
+
+  /* Step 1: allocate and initialize JPEG compression object */
+
+  /* We have to set up the error handler first, in case the initialization
+   * step fails.  (Unlikely, but it could happen if you are out of memory.)
+   * This routine fills in the contents of struct jerr, and returns jerr's
+   * address which we place into the link field in cinfo.
+   */
+  cinfo.err = jpeg_std_error(&jerr);
+  /* Now we can initialize the JPEG compression object. */
+  jpeg_create_compress(&cinfo);
+
+  /* Step 2: specify data destination (eg, a file) */
+  /* Note: steps 2 and 3 can be done in either order. */
+
+  /* Here we use the library-supplied code to send compressed data to a
+   * stdio stream.  You can also write your own code to do something else.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to write binary files.
+   */
+  if ((outfile = fopen(filename, "wb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    exit(1);
+  }
+  jpeg_stdio_dest(&cinfo, outfile);
+
+  /* Step 3: set parameters for compression */
+
+  /* First we supply a description of the input image.
+   * Four fields of the cinfo struct must be filled in:
+   */
+  cinfo.image_width = image_width;      /* image width and height, in pixels */
+  cinfo.image_height = image_height;
+  cinfo.input_components = 3;           /* # of color components per pixel */
+  cinfo.in_color_space = JCS_RGB;       /* colorspace of input image */
+  /* Now use the library's routine to set default compression parameters.
+   * (You must set at least cinfo.in_color_space before calling this,
+   * since the defaults depend on the source color space.)
+   */
+  jpeg_set_defaults(&cinfo);
+  /* Now you can set any non-default parameters you wish to.
+   * Here we just illustrate the use of quality (quantization table) scaling:
+   */
+  jpeg_set_quality(&cinfo, quality, TRUE /* limit to baseline-JPEG values */);
+
+  /* Step 4: Start compressor */
+
+  /* TRUE ensures that we will write a complete interchange-JPEG file.
+   * Pass TRUE unless you are very sure of what you're doing.
+   */
+  jpeg_start_compress(&cinfo, TRUE);
+
+  /* Step 5: while (scan lines remain to be written) */
+  /*           jpeg_write_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.next_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   * To keep things simple, we pass one scanline per call; you can pass
+   * more if you wish, though.
+   */
+  row_stride = image_width * 3; /* JSAMPLEs per row in image_buffer */
+
+  while (cinfo.next_scanline < cinfo.image_height) {
+    /* jpeg_write_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could pass
+     * more than one scanline at a time if that's more convenient.
+     */
+    row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+    (void) jpeg_write_scanlines(&cinfo, row_pointer, 1);
+  }
+
+  /* Step 6: Finish compression */
+
+  jpeg_finish_compress(&cinfo);
+  /* After finish_compress, we can close the output file. */
+  fclose(outfile);
+
+  /* Step 7: release JPEG compression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_compress(&cinfo);
+
+  /* And we're done! */
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above loop, we ignored the return value of jpeg_write_scanlines,
+ * which is the number of scanlines actually written.  We could get away
+ * with this because we were only relying on the value of cinfo.next_scanline,
+ * which will be incremented correctly.  If you maintain additional loop
+ * variables then you should be careful to increment them properly.
+ * Actually, for output to a stdio stream you needn't worry, because
+ * then jpeg_write_scanlines will write all the lines passed (or else exit
+ * with a fatal error).  Partial writes can only occur if you use a data
+ * destination module that can demand suspension of the compressor.
+ * (If you don't know what that's for, you don't need it.)
+ *
+ * If the compressor requires full-image buffers (for entropy-coding
+ * optimization or a multi-scan JPEG file), it will create temporary
+ * files for anything that doesn't fit within the maximum-memory setting.
+ * (Note that temp files are NOT needed if you use the default parameters.)
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ *
+ * Scanlines MUST be supplied in top-to-bottom order if you want your JPEG
+ * files to be compatible with everyone else's.  If you cannot readily read
+ * your data in that order, you'll need an intermediate array to hold the
+ * image.  See rdtarga.c or rdbmp.c for examples of handling bottom-to-top
+ * source data using the JPEG code's internal virtual-array mechanisms.
+ */
+
+
+
+/******************** JPEG DECOMPRESSION SAMPLE INTERFACE *******************/
+
+/* This half of the example shows how to read data from the JPEG decompressor.
+ * It's a bit more refined than the above, in that we show:
+ *   (a) how to modify the JPEG library's standard error-reporting behavior;
+ *   (b) how to allocate workspace using the library's memory manager.
+ *
+ * Just to make this example a little different from the first one, we'll
+ * assume that we do not intend to put the whole image into an in-memory
+ * buffer, but to send it line-by-line someplace else.  We need a one-
+ * scanline-high JSAMPLE array as a work buffer, and we will let the JPEG
+ * memory manager allocate it for us.  This approach is actually quite useful
+ * because we don't need to remember to deallocate the buffer separately: it
+ * will go away automatically when the JPEG object is cleaned up.
+ */
+
+
+/*
+ * ERROR HANDLING:
+ *
+ * The JPEG library's standard error handler (jerror.c) is divided into
+ * several "methods" which you can override individually.  This lets you
+ * adjust the behavior without duplicating a lot of code, which you might
+ * have to update with each future release.
+ *
+ * Our example here shows how to override the "error_exit" method so that
+ * control is returned to the library's caller when a fatal error occurs,
+ * rather than calling exit() as the standard error_exit method does.
+ *
+ * We use C's setjmp/longjmp facility to return control.  This means that the
+ * routine which calls the JPEG library must first execute a setjmp() call to
+ * establish the return point.  We want the replacement error_exit to do a
+ * longjmp().  But we need to make the setjmp buffer accessible to the
+ * error_exit routine.  To do this, we make a private extension of the
+ * standard JPEG error handler object.  (If we were using C++, we'd say we
+ * were making a subclass of the regular error handler.)
+ *
+ * Here's the extended error handler struct:
+ */
+
+struct my_error_mgr {
+  struct jpeg_error_mgr pub;    /* "public" fields */
+
+  jmp_buf setjmp_buffer;        /* for return to caller */
+};
+
+typedef struct my_error_mgr *my_error_ptr;
+
+/*
+ * Here's the routine that will replace the standard error_exit method:
+ */
+
+METHODDEF(void)
+my_error_exit (j_common_ptr cinfo)
+{
+  /* cinfo->err really points to a my_error_mgr struct, so coerce pointer */
+  my_error_ptr myerr = (my_error_ptr) cinfo->err;
+
+  /* Always display the message. */
+  /* We could postpone this until after returning, if we chose. */
+  (*cinfo->err->output_message) (cinfo);
+
+  /* Return control to the setjmp point */
+  longjmp(myerr->setjmp_buffer, 1);
+}
+
+
+/*
+ * Sample routine for JPEG decompression.  We assume that the source file name
+ * is passed in.  We want to return 1 on success, 0 on error.
+ */
+
+
+GLOBAL(int)
+read_JPEG_file (char *filename)
+{
+  /* This struct contains the JPEG decompression parameters and pointers to
+   * working space (which is allocated as needed by the JPEG library).
+   */
+  struct jpeg_decompress_struct cinfo;
+  /* We use our private extension JPEG error handler.
+   * Note that this struct must live as long as the main JPEG parameter
+   * struct, to avoid dangling-pointer problems.
+   */
+  struct my_error_mgr jerr;
+  /* More stuff */
+  FILE *infile;                 /* source file */
+  JSAMPARRAY buffer;            /* Output row buffer */
+  int row_stride;               /* physical row width in output buffer */
+
+  /* In this example we want to open the input file before doing anything else,
+   * so that the setjmp() error recovery below can assume the file is open.
+   * VERY IMPORTANT: use "b" option to fopen() if you are on a machine that
+   * requires it in order to read binary files.
+   */
+
+  if ((infile = fopen(filename, "rb")) == NULL) {
+    fprintf(stderr, "can't open %s\n", filename);
+    return 0;
+  }
+
+  /* Step 1: allocate and initialize JPEG decompression object */
+
+  /* We set up the normal JPEG error routines, then override error_exit. */
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  /* Establish the setjmp return context for my_error_exit to use. */
+  if (setjmp(jerr.setjmp_buffer)) {
+    /* If we get here, the JPEG code has signaled an error.
+     * We need to clean up the JPEG object, close the input file, and return.
+     */
+    jpeg_destroy_decompress(&cinfo);
+    fclose(infile);
+    return 0;
+  }
+  /* Now we can initialize the JPEG decompression object. */
+  jpeg_create_decompress(&cinfo);
+
+  /* Step 2: specify data source (eg, a file) */
+
+  jpeg_stdio_src(&cinfo, infile);
+
+  /* Step 3: read file parameters with jpeg_read_header() */
+
+  (void) jpeg_read_header(&cinfo, TRUE);
+  /* We can ignore the return value from jpeg_read_header since
+   *   (a) suspension is not possible with the stdio data source, and
+   *   (b) we passed TRUE to reject a tables-only JPEG file as an error.
+   * See libjpeg.txt for more info.
+   */
+
+  /* Step 4: set parameters for decompression */
+
+  /* In this example, we don't need to change any of the defaults set by
+   * jpeg_read_header(), so we do nothing here.
+   */
+
+  /* Step 5: Start decompressor */
+
+  (void) jpeg_start_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* We may need to do some setup of our own at this point before reading
+   * the data.  After jpeg_start_decompress() we have the correct scaled
+   * output image dimensions available, as well as the output colormap
+   * if we asked for color quantization.
+   * In this example, we need to make an output work buffer of the right size.
+   */
+  /* JSAMPLEs per row in output buffer */
+  row_stride = cinfo.output_width * cinfo.output_components;
+  /* Make a one-row-high sample array that will go away when done with image */
+  buffer = (*cinfo.mem->alloc_sarray)
+                ((j_common_ptr) &cinfo, JPOOL_IMAGE, row_stride, 1);
+
+  /* Step 6: while (scan lines remain to be read) */
+  /*           jpeg_read_scanlines(...); */
+
+  /* Here we use the library's state variable cinfo.output_scanline as the
+   * loop counter, so that we don't have to keep track ourselves.
+   */
+  while (cinfo.output_scanline < cinfo.output_height) {
+    /* jpeg_read_scanlines expects an array of pointers to scanlines.
+     * Here the array is only one element long, but you could ask for
+     * more than one scanline at a time if that's more convenient.
+     */
+    (void) jpeg_read_scanlines(&cinfo, buffer, 1);
+    /* Assume put_scanline_someplace wants a pointer and sample count. */
+    put_scanline_someplace(buffer[0], row_stride);
+  }
+
+  /* Step 7: Finish decompression */
+
+  (void) jpeg_finish_decompress(&cinfo);
+  /* We can ignore the return value since suspension is not possible
+   * with the stdio data source.
+   */
+
+  /* Step 8: Release JPEG decompression object */
+
+  /* This is an important step since it will release a good deal of memory. */
+  jpeg_destroy_decompress(&cinfo);
+
+  /* After finish_decompress, we can close the input file.
+   * Here we postpone it until after no more JPEG errors are possible,
+   * so as to simplify the setjmp error logic above.  (Actually, I don't
+   * think that jpeg_destroy can do an error exit, but why assume anything...)
+   */
+  fclose(infile);
+
+  /* At this point you may want to check to see whether any corrupt-data
+   * warnings occurred (test whether jerr.pub.num_warnings is nonzero).
+   */
+
+  /* And we're done! */
+  return 1;
+}
+
+
+/*
+ * SOME FINE POINTS:
+ *
+ * In the above code, we ignored the return value of jpeg_read_scanlines,
+ * which is the number of scanlines actually read.  We could get away with
+ * this because we asked for only one line at a time and we weren't using
+ * a suspending data source.  See libjpeg.txt for more info.
+ *
+ * We cheated a bit by calling alloc_sarray() after jpeg_start_decompress();
+ * we should have done it beforehand to ensure that the space would be
+ * counted against the JPEG max_memory setting.  In some systems the above
+ * code would risk an out-of-memory error.  However, in general we don't
+ * know the output image dimensions before jpeg_start_decompress(), unless we
+ * call jpeg_calc_output_dimensions().  See libjpeg.txt for more about this.
+ *
+ * Scanlines are returned in the same order as they appear in the JPEG file,
+ * which is standardly top-to-bottom.  If you must emit data bottom-to-top,
+ * you can use one of the virtual arrays provided by the JPEG memory manager
+ * to invert the data.  See wrbmp.c for an example.
+ *
+ * As with compression, some operating modes may require temporary files.
+ * On some systems you may need to set up a signal handler to ensure that
+ * temporary files are deleted if the program is interrupted.  See libjpeg.txt.
+ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jaricom.c glmark2-2021.02/src/libjpeg-turbo/jaricom.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jaricom.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jaricom.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jaricom.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains probability estimation tables for common use in
  * arithmetic entropy encoding and decoding routines.
@@ -18,7 +21,7 @@
 #include "jpeglib.h"
 
 /* The following #define specifies the packing of the four components
- * into the compact INT32 representation.
+ * into the compact JLONG representation.
  * Note that this formula must match the actual arithmetic encoder
  * and decoder implementation.  The implementation has to be changed
  * if this formula is changed.
@@ -26,9 +29,9 @@
  * implementation (jbig_tab.c).
  */
 
-#define V(i,a,b,c,d) (((INT32)a << 16) | ((INT32)c << 8) | ((INT32)d << 7) | b)
+#define V(i,a,b,c,d) (((JLONG)a << 16) | ((JLONG)c << 8) | ((JLONG)d << 7) | b)
 
-const INT32 jpeg_aritab[113+1] = {
+const JLONG jpeg_aritab[113+1] = {
 /*
  * Index, Qe_Value, Next_Index_LPS, Next_Index_MPS, Switch_MPS
  */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcapimin.c glmark2-2021.02/src/libjpeg-turbo/jcapimin.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcapimin.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcapimin.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jcapimin.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
  * Modified 2003-2010 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the compression half
  * of the JPEG library.  These are the "minimum" API routines that may be
@@ -33,12 +36,12 @@
   int i;
 
   /* Guard against version mismatches between library and caller. */
-  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
   if (version != JPEG_LIB_VERSION)
     ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
-  if (structsize != SIZEOF(struct jpeg_compress_struct))
-    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
-	     (int) SIZEOF(struct jpeg_compress_struct), (int) structsize);
+  if (structsize != sizeof(struct jpeg_compress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int) sizeof(struct jpeg_compress_struct), (int) structsize);
 
   /* For debugging purposes, we zero the whole master structure.
    * But the application has already set the err pointer, and may have set
@@ -47,9 +50,9 @@
    * complain here.
    */
   {
-    struct jpeg_error_mgr * err = cinfo->err;
-    void * client_data = cinfo->client_data; /* ignore Purify complaint here */
-    MEMZERO(cinfo, SIZEOF(struct jpeg_compress_struct));
+    struct jpeg_error_mgr *err = cinfo->err;
+    void *client_data = cinfo->client_data; /* ignore Purify complaint here */
+    MEMZERO(cinfo, sizeof(struct jpeg_compress_struct));
     cinfo->err = err;
     cinfo->client_data = client_data;
   }
@@ -85,7 +88,7 @@
 
   cinfo->script_space = NULL;
 
-  cinfo->input_gamma = 1.0;	/* in case application forgets */
+  cinfo->input_gamma = 1.0;     /* in case application forgets */
 
   /* OK, I'm ready */
   cinfo->global_state = CSTATE_START;
@@ -131,8 +134,8 @@
 jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress)
 {
   int i;
-  JQUANT_TBL * qtbl;
-  JHUFF_TBL * htbl;
+  JQUANT_TBL *qtbl;
+  JHUFF_TBL *htbl;
 
   for (i = 0; i < NUM_QUANT_TBLS; i++) {
     if ((qtbl = cinfo->quant_tbl_ptrs[i]) != NULL)
@@ -173,15 +176,15 @@
     (*cinfo->master->prepare_for_pass) (cinfo);
     for (iMCU_row = 0; iMCU_row < cinfo->total_iMCU_rows; iMCU_row++) {
       if (cinfo->progress != NULL) {
-	cinfo->progress->pass_counter = (long) iMCU_row;
-	cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        cinfo->progress->pass_counter = (long) iMCU_row;
+        cinfo->progress->pass_limit = (long) cinfo->total_iMCU_rows;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       }
       /* We bypass the main controller and invoke coef controller directly;
        * all work is being done from the coefficient buffer.
        */
       if (! (*cinfo->coef->compress_data) (cinfo, (JSAMPIMAGE) NULL))
-	ERREXIT(cinfo, JERR_CANT_SUSPEND);
+        ERREXIT(cinfo, JERR_CANT_SUSPEND);
     }
     (*cinfo->master->finish_pass) (cinfo);
   }
@@ -202,9 +205,9 @@
 
 GLOBAL(void)
 jpeg_write_marker (j_compress_ptr cinfo, int marker,
-		   const JOCTET *dataptr, unsigned int datalen)
+                   const JOCTET *dataptr, unsigned int datalen)
 {
-  JMETHOD(void, write_marker_byte, (j_compress_ptr info, int val));
+  void (*write_marker_byte) (j_compress_ptr info, int val);
 
   if (cinfo->next_scanline != 0 ||
       (cinfo->global_state != CSTATE_SCANNING &&
@@ -213,7 +216,7 @@
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
   (*cinfo->marker->write_marker_header) (cinfo, marker, datalen);
-  write_marker_byte = cinfo->marker->write_marker_byte;	/* copy for speed */
+  write_marker_byte = cinfo->marker->write_marker_byte; /* copy for speed */
   while (datalen--) {
     (*write_marker_byte) (cinfo, *dataptr);
     dataptr++;
@@ -248,14 +251,14 @@
  * To produce a pair of files containing abbreviated tables and abbreviated
  * image data, one would proceed as follows:
  *
- *		initialize JPEG object
- *		set JPEG parameters
- *		set destination to table file
- *		jpeg_write_tables(cinfo);
- *		set destination to image file
- *		jpeg_start_compress(cinfo, FALSE);
- *		write data...
- *		jpeg_finish_compress(cinfo);
+ *              initialize JPEG object
+ *              set JPEG parameters
+ *              set destination to table file
+ *              jpeg_write_tables(cinfo);
+ *              set destination to image file
+ *              jpeg_start_compress(cinfo, FALSE);
+ *              write data...
+ *              jpeg_finish_compress(cinfo);
  *
  * jpeg_write_tables has the side effect of marking all tables written
  * (same as jpeg_suppress_tables(..., TRUE)).  Thus a subsequent start_compress
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcapistd.c glmark2-2021.02/src/libjpeg-turbo/jcapistd.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcapistd.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcapistd.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1994-1996, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the compression half
  * of the JPEG library.  These are the "standard" API routines that are
@@ -41,7 +42,7 @@
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
   if (write_all_tables)
-    jpeg_suppress_tables(cinfo, FALSE);	/* mark all tables to be written */
+    jpeg_suppress_tables(cinfo, FALSE); /* mark all tables to be written */
 
   /* (Re)initialize error mgr and destination modules */
   (*cinfo->err->reset_error_mgr) ((j_common_ptr) cinfo);
@@ -75,7 +76,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_write_scanlines (j_compress_ptr cinfo, JSAMPARRAY scanlines,
-		      JDIMENSION num_lines)
+                      JDIMENSION num_lines)
 {
   JDIMENSION row_ctr, rows_left;
 
@@ -118,7 +119,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
-		     JDIMENSION num_lines)
+                     JDIMENSION num_lines)
 {
   JDIMENSION lines_per_iMCU_row;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcarith.c glmark2-2021.02/src/libjpeg-turbo/jcarith.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcarith.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcarith.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcarith.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains portable arithmetic entropy encoding routines for JPEG
  * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
@@ -23,10 +26,10 @@
 typedef struct {
   struct jpeg_entropy_encoder pub; /* public fields */
 
-  INT32 c; /* C register, base of coding interval, layout as in sec. D.1.3 */
-  INT32 a;               /* A register, normalized size of coding interval */
-  INT32 sc;        /* counter for stacked 0xFF values which might overflow */
-  INT32 zc;          /* counter for pending 0x00 output values which might *
+  JLONG c; /* C register, base of coding interval, layout as in sec. D.1.3 */
+  JLONG a;               /* A register, normalized size of coding interval */
+  JLONG sc;        /* counter for stacked 0xFF values which might overflow */
+  JLONG zc;          /* counter for pending 0x00 output values which might *
                           * be discarded at the end ("Pacman" termination) */
   int ct;  /* bit shift counter, determines when next byte will be written */
   int buffer;                /* buffer for most recent output byte != 0xFF */
@@ -34,18 +37,18 @@
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
   int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to statistics areas (these workspaces have image lifespan) */
-  unsigned char * dc_stats[NUM_ARITH_TBLS];
-  unsigned char * ac_stats[NUM_ARITH_TBLS];
+  unsigned char *dc_stats[NUM_ARITH_TBLS];
+  unsigned char *ac_stats[NUM_ARITH_TBLS];
 
   /* Statistics bin for coding with fixed probability 0.5 */
   unsigned char fixed_bin[4];
 } arith_entropy_encoder;
 
-typedef arith_entropy_encoder * arith_entropy_ptr;
+typedef arith_entropy_encoder *arith_entropy_ptr;
 
 /* The following two definitions specify the allocation chunk size
  * for the statistics area.
@@ -95,20 +98,20 @@
 #define CALCULATE_SPECTRAL_CONDITIONING
  */
 
-/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
- * We assume that int right shift is unsigned if INT32 right shift is,
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
  * which should be safe.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	int ishift_temp;
+#define ISHIFT_TEMPS    int ishift_temp;
 #define IRIGHT_SHIFT(x,shft)  \
-	((ishift_temp = (x)) < 0 ? \
-	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
-	 (ishift_temp >> (shft)))
+        ((ishift_temp = (x)) < 0 ? \
+         (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+         (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 
@@ -116,7 +119,7 @@
 emit_byte (int val, j_compress_ptr cinfo)
 /* Write next output byte; we do not support suspension in this module. */
 {
-  struct jpeg_destination_mgr * dest = cinfo->dest;
+  struct jpeg_destination_mgr *dest = cinfo->dest;
 
   *dest->next_output_byte++ = (JOCTET) val;
   if (--dest->free_in_buffer == 0)
@@ -133,7 +136,7 @@
 finish_pass (j_compress_ptr cinfo)
 {
   arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
-  INT32 temp;
+  JLONG temp;
 
   /* Section D.1.8: Termination of encoding */
 
@@ -149,11 +152,11 @@
     /* One final overflow has to be handled */
     if (e->buffer >= 0) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       emit_byte(e->buffer + 1, cinfo);
       if (e->buffer + 1 == 0xFF)
-	emit_byte(0x00, cinfo);
+        emit_byte(0x00, cinfo);
     }
     e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
     e->sc = 0;
@@ -162,17 +165,17 @@
       ++e->zc;
     else if (e->buffer >= 0) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       emit_byte(e->buffer, cinfo);
     }
     if (e->sc) {
       if (e->zc)
-	do emit_byte(0x00, cinfo);
-	while (--e->zc);
+        do emit_byte(0x00, cinfo);
+        while (--e->zc);
       do {
-	emit_byte(0xFF, cinfo);
-	emit_byte(0x00, cinfo);
+        emit_byte(0xFF, cinfo);
+        emit_byte(0x00, cinfo);
       } while (--e->sc);
     }
   }
@@ -187,7 +190,7 @@
     if (e->c & 0x7F800L) {
       emit_byte((e->c >> 11) & 0xFF, cinfo);
       if (((e->c >> 11) & 0xFF) == 0xFF)
-	emit_byte(0x00, cinfo);
+        emit_byte(0x00, cinfo);
     }
   }
 }
@@ -216,20 +219,20 @@
  */
 
 LOCAL(void)
-arith_encode (j_compress_ptr cinfo, unsigned char *st, int val) 
+arith_encode (j_compress_ptr cinfo, unsigned char *st, int val)
 {
   register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
   register unsigned char nl, nm;
-  register INT32 qe, temp;
+  register JLONG qe, temp;
   register int sv;
 
   /* Fetch values from our compact representation of Table D.2:
    * Qe values and probability estimation state machine
    */
   sv = *st;
-  qe = jpeg_aritab[sv & 0x7F];	/* => Qe_Value */
-  nl = qe & 0xFF; qe >>= 8;	/* Next_Index_LPS + Switch_MPS */
-  nm = qe & 0xFF; qe >>= 8;	/* Next_Index_MPS */
+  qe = jpeg_aritab[sv & 0x7F];  /* => Qe_Value */
+  nl = qe & 0xFF; qe >>= 8;     /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF; qe >>= 8;     /* Next_Index_MPS */
 
   /* Encode & estimation procedures per sections D.1.4 & D.1.5 */
   e->a -= qe;
@@ -243,7 +246,7 @@
       e->c += e->a;
       e->a = qe;
     }
-    *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
+    *st = (sv & 0x80) ^ nl;     /* Estimate_after_LPS */
   } else {
     /* Encode the more probable symbol */
     if (e->a >= 0x8000L)
@@ -255,7 +258,7 @@
       e->c += e->a;
       e->a = qe;
     }
-    *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+    *st = (sv & 0x80) ^ nm;     /* Estimate_after_MPS */
   }
 
   /* Renormalization & data output per section D.1.6 */
@@ -266,43 +269,43 @@
       /* Another byte is ready for output */
       temp = e->c >> 19;
       if (temp > 0xFF) {
-	/* Handle overflow over all stacked 0xFF bytes */
-	if (e->buffer >= 0) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  emit_byte(e->buffer + 1, cinfo);
-	  if (e->buffer + 1 == 0xFF)
-	    emit_byte(0x00, cinfo);
-	}
-	e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
-	e->sc = 0;
-	/* Note: The 3 spacer bits in the C register guarantee
-	 * that the new buffer byte can't be 0xFF here
-	 * (see page 160 in the P&M JPEG book). */
-	e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
+        /* Handle overflow over all stacked 0xFF bytes */
+        if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer + 1, cinfo);
+          if (e->buffer + 1 == 0xFF)
+            emit_byte(0x00, cinfo);
+        }
+        e->zc += e->sc;  /* carry-over converts stacked 0xFF bytes to 0x00 */
+        e->sc = 0;
+        /* Note: The 3 spacer bits in the C register guarantee
+         * that the new buffer byte can't be 0xFF here
+         * (see page 160 in the P&M JPEG book). */
+        e->buffer = temp & 0xFF;  /* new output byte, might overflow later */
       } else if (temp == 0xFF) {
-	++e->sc;  /* stack 0xFF byte (which might overflow later) */
+        ++e->sc;  /* stack 0xFF byte (which might overflow later) */
       } else {
-	/* Output all stacked 0xFF bytes, they will not overflow any more */
-	if (e->buffer == 0)
-	  ++e->zc;
-	else if (e->buffer >= 0) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  emit_byte(e->buffer, cinfo);
-	}
-	if (e->sc) {
-	  if (e->zc)
-	    do emit_byte(0x00, cinfo);
-	    while (--e->zc);
-	  do {
-	    emit_byte(0xFF, cinfo);
-	    emit_byte(0x00, cinfo);
-	  } while (--e->sc);
-	}
-	e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
+        /* Output all stacked 0xFF bytes, they will not overflow any more */
+        if (e->buffer == 0)
+          ++e->zc;
+        else if (e->buffer >= 0) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          emit_byte(e->buffer, cinfo);
+        }
+        if (e->sc) {
+          if (e->zc)
+            do emit_byte(0x00, cinfo);
+            while (--e->zc);
+          do {
+            emit_byte(0xFF, cinfo);
+            emit_byte(0x00, cinfo);
+          } while (--e->sc);
+        }
+        e->buffer = temp & 0xFF;  /* new output byte (can still overflow) */
       }
       e->c &= 0x7FFFFL;
       e->ct += 8;
@@ -320,7 +323,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   finish_pass(cinfo);
 
@@ -398,45 +401,45 @@
     /* Figure F.4: Encode_DC_DIFF */
     if ((v = m - entropy->last_dc_val[ci]) == 0) {
       arith_encode(cinfo, st, 0);
-      entropy->dc_context[ci] = 0;	/* zero diff category */
+      entropy->dc_context[ci] = 0;      /* zero diff category */
     } else {
       entropy->last_dc_val[ci] = m;
       arith_encode(cinfo, st, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, st + 1, 0);	/* Table F.4: SS = S0 + 1 */
-	st += 2;			/* Table F.4: SP = S0 + 2 */
-	entropy->dc_context[ci] = 4;	/* small positive diff category */
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_context[ci] = 4;    /* small positive diff category */
       } else {
-	v = -v;
-	arith_encode(cinfo, st + 1, 1);	/* Table F.4: SS = S0 + 1 */
-	st += 3;			/* Table F.4: SN = S0 + 3 */
-	entropy->dc_context[ci] = 8;	/* small negative diff category */
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_context[ci] = 8;    /* small negative diff category */
       }
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;	/* zero diff category */
+        entropy->dc_context[ci] = 0;    /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] += 8;	/* large diff category */
+        entropy->dc_context[ci] += 8;   /* large diff category */
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
   }
 
@@ -491,21 +494,21 @@
   /* Figure F.5: Encode_AC_Coefficients */
   for (k = cinfo->Ss; k <= ke; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
-    arith_encode(cinfo, st, 0);		/* EOB decision */
+    arith_encode(cinfo, st, 0);         /* EOB decision */
     for (;;) {
       if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
-	if (v >>= cinfo->Al) {
-	  arith_encode(cinfo, st + 1, 1);
-	  arith_encode(cinfo, entropy->fixed_bin, 0);
-	  break;
-	}
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 0);
+          break;
+        }
       } else {
-	v = -v;
-	if (v >>= cinfo->Al) {
-	  arith_encode(cinfo, st + 1, 1);
-	  arith_encode(cinfo, entropy->fixed_bin, 1);
-	  break;
-	}
+        v = -v;
+        if (v >>= cinfo->Al) {
+          arith_encode(cinfo, st + 1, 1);
+          arith_encode(cinfo, entropy->fixed_bin, 1);
+          break;
+        }
       }
       arith_encode(cinfo, st + 1, 0); st += 3; k++;
     }
@@ -517,15 +520,15 @@
       m = 1;
       v2 = v;
       if (v2 >>= 1) {
-	arith_encode(cinfo, st, 1);
-	m <<= 1;
-	st = entropy->ac_stats[tbl] +
-	     (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m <<= 1;
+        st = entropy->ac_stats[tbl] +
+             (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
     }
     arith_encode(cinfo, st, 0);
@@ -566,7 +569,7 @@
     entropy->restarts_to_go--;
   }
 
-  st = entropy->fixed_bin;	/* use fixed probability estimation */
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
   Al = cinfo->Al;
 
   /* Encode the MCU data blocks */
@@ -635,29 +638,29 @@
   for (k = cinfo->Ss; k <= ke; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
     if (k > kex)
-      arith_encode(cinfo, st, 0);	/* EOB decision */
+      arith_encode(cinfo, st, 0);       /* EOB decision */
     for (;;) {
       if ((v = (*block)[jpeg_natural_order[k]]) >= 0) {
-	if (v >>= cinfo->Al) {
-	  if (v >> 1)			/* previously nonzero coef */
-	    arith_encode(cinfo, st + 2, (v & 1));
-	  else {			/* newly nonzero coef */
-	    arith_encode(cinfo, st + 1, 1);
-	    arith_encode(cinfo, entropy->fixed_bin, 0);
-	  }
-	  break;
-	}
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 0);
+          }
+          break;
+        }
       } else {
-	v = -v;
-	if (v >>= cinfo->Al) {
-	  if (v >> 1)			/* previously nonzero coef */
-	    arith_encode(cinfo, st + 2, (v & 1));
-	  else {			/* newly nonzero coef */
-	    arith_encode(cinfo, st + 1, 1);
-	    arith_encode(cinfo, entropy->fixed_bin, 1);
-	  }
-	  break;
-	}
+        v = -v;
+        if (v >>= cinfo->Al) {
+          if (v >> 1)                   /* previously nonzero coef */
+            arith_encode(cinfo, st + 2, (v & 1));
+          else {                        /* newly nonzero coef */
+            arith_encode(cinfo, st + 1, 1);
+            arith_encode(cinfo, entropy->fixed_bin, 1);
+          }
+          break;
+        }
       }
       arith_encode(cinfo, st + 1, 0); st += 3; k++;
     }
@@ -680,7 +683,7 @@
 encode_mcu (j_compress_ptr cinfo, JBLOCKROW *MCU_data)
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JBLOCKROW block;
   unsigned char *st;
   int blkn, ci, tbl, k, ke;
@@ -713,45 +716,45 @@
     /* Figure F.4: Encode_DC_DIFF */
     if ((v = (*block)[0] - entropy->last_dc_val[ci]) == 0) {
       arith_encode(cinfo, st, 0);
-      entropy->dc_context[ci] = 0;	/* zero diff category */
+      entropy->dc_context[ci] = 0;      /* zero diff category */
     } else {
       entropy->last_dc_val[ci] = (*block)[0];
       arith_encode(cinfo, st, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, st + 1, 0);	/* Table F.4: SS = S0 + 1 */
-	st += 2;			/* Table F.4: SP = S0 + 2 */
-	entropy->dc_context[ci] = 4;	/* small positive diff category */
+        arith_encode(cinfo, st + 1, 0); /* Table F.4: SS = S0 + 1 */
+        st += 2;                        /* Table F.4: SP = S0 + 2 */
+        entropy->dc_context[ci] = 4;    /* small positive diff category */
       } else {
-	v = -v;
-	arith_encode(cinfo, st + 1, 1);	/* Table F.4: SS = S0 + 1 */
-	st += 3;			/* Table F.4: SN = S0 + 3 */
-	entropy->dc_context[ci] = 8;	/* small negative diff category */
+        v = -v;
+        arith_encode(cinfo, st + 1, 1); /* Table F.4: SS = S0 + 1 */
+        st += 3;                        /* Table F.4: SN = S0 + 3 */
+        entropy->dc_context[ci] = 8;    /* small negative diff category */
       }
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
-	while (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st += 1;
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        st = entropy->dc_stats[tbl] + 20; /* Table F.4: X1 = 20 */
+        while (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st += 1;
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;	/* zero diff category */
+        entropy->dc_context[ci] = 0;    /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] += 8;	/* large diff category */
+        entropy->dc_context[ci] += 8;   /* large diff category */
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
 
     /* Sections F.1.4.2 & F.1.4.4.2: Encoding of AC coefficients */
@@ -765,43 +768,43 @@
     /* Figure F.5: Encode_AC_Coefficients */
     for (k = 1; k <= ke; k++) {
       st = entropy->ac_stats[tbl] + 3 * (k - 1);
-      arith_encode(cinfo, st, 0);	/* EOB decision */
+      arith_encode(cinfo, st, 0);       /* EOB decision */
       while ((v = (*block)[jpeg_natural_order[k]]) == 0) {
-	arith_encode(cinfo, st + 1, 0); st += 3; k++;
+        arith_encode(cinfo, st + 1, 0); st += 3; k++;
       }
       arith_encode(cinfo, st + 1, 1);
       /* Figure F.6: Encoding nonzero value v */
       /* Figure F.7: Encoding the sign of v */
       if (v > 0) {
-	arith_encode(cinfo, entropy->fixed_bin, 0);
+        arith_encode(cinfo, entropy->fixed_bin, 0);
       } else {
-	v = -v;
-	arith_encode(cinfo, entropy->fixed_bin, 1);
+        v = -v;
+        arith_encode(cinfo, entropy->fixed_bin, 1);
       }
       st += 2;
       /* Figure F.8: Encoding the magnitude category of v */
       m = 0;
       if (v -= 1) {
-	arith_encode(cinfo, st, 1);
-	m = 1;
-	v2 = v;
-	if (v2 >>= 1) {
-	  arith_encode(cinfo, st, 1);
-	  m <<= 1;
-	  st = entropy->ac_stats[tbl] +
-	       (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	  while (v2 >>= 1) {
-	    arith_encode(cinfo, st, 1);
-	    m <<= 1;
-	    st += 1;
-	  }
-	}
+        arith_encode(cinfo, st, 1);
+        m = 1;
+        v2 = v;
+        if (v2 >>= 1) {
+          arith_encode(cinfo, st, 1);
+          m <<= 1;
+          st = entropy->ac_stats[tbl] +
+               (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+          while (v2 >>= 1) {
+            arith_encode(cinfo, st, 1);
+            m <<= 1;
+            st += 1;
+          }
+        }
       }
       arith_encode(cinfo, st, 0);
       /* Figure F.9: Encoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	arith_encode(cinfo, st, (m & v) ? 1 : 0);
+        arith_encode(cinfo, st, (m & v) ? 1 : 0);
     }
     /* Encode EOB decision only if k <= DCTSIZE2 - 1 */
     if (k <= DCTSIZE2 - 1) {
@@ -823,7 +826,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (gather_statistics)
     /* Make sure to avoid that in the master control logic!
@@ -838,14 +841,14 @@
   if (cinfo->progressive_mode) {
     if (cinfo->Ah == 0) {
       if (cinfo->Ss == 0)
-	entropy->pub.encode_mcu = encode_mcu_DC_first;
+        entropy->pub.encode_mcu = encode_mcu_DC_first;
       else
-	entropy->pub.encode_mcu = encode_mcu_AC_first;
+        entropy->pub.encode_mcu = encode_mcu_AC_first;
     } else {
       if (cinfo->Ss == 0)
-	entropy->pub.encode_mcu = encode_mcu_DC_refine;
+        entropy->pub.encode_mcu = encode_mcu_DC_refine;
       else
-	entropy->pub.encode_mcu = encode_mcu_AC_refine;
+        entropy->pub.encode_mcu = encode_mcu_AC_refine;
     }
   } else
     entropy->pub.encode_mcu = encode_mcu;
@@ -857,10 +860,10 @@
     if (cinfo->progressive_mode == 0 || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       tbl = compptr->dc_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->dc_stats[tbl] == NULL)
-	entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+        entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
       MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
       /* Initialize DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
@@ -870,15 +873,15 @@
     if (cinfo->progressive_mode == 0 || cinfo->Se) {
       tbl = compptr->ac_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->ac_stats[tbl] == NULL)
-	entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+        entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
       MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
 #ifdef CALCULATE_SPECTRAL_CONDITIONING
       if (cinfo->progressive_mode)
-	/* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
-	cinfo->arith_ac_K[tbl] = cinfo->Ss + ((8 + cinfo->Se - cinfo->Ss) >> 4);
+        /* Section G.1.3.2: Set appropriate arithmetic conditioning value Kx */
+        cinfo->arith_ac_K[tbl] = cinfo->Ss + ((8 + cinfo->Se - cinfo->Ss) >> 4);
 #endif
     }
   }
@@ -909,7 +912,7 @@
 
   entropy = (arith_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(arith_entropy_encoder));
+                                sizeof(arith_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass;
   entropy->pub.finish_pass = finish_pass;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccoefct.c glmark2-2021.02/src/libjpeg-turbo/jccoefct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccoefct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jccoefct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jccoefct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the coefficient buffer controller for compression.
  * This controller is the top level of the JPEG compressor proper.
@@ -34,19 +37,16 @@
 typedef struct {
   struct jpeg_c_coef_controller pub; /* public fields */
 
-  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
-  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
 
   /* For single-pass compression, it's sufficient to buffer just one MCU
    * (although this may prove a bit slow in practice).  We allocate a
    * workspace of C_MAX_BLOCKS_IN_MCU coefficient blocks, and reuse it for each
-   * MCU constructed and sent.  (On 80x86, the workspace is FAR even though
-   * it's not really very big; this is to keep the module interfaces unchanged
-   * when a large coefficient buffer is necessary.)
-   * In multi-pass modes, this array points to the current MCU's blocks
-   * within the virtual arrays.
+   * MCU constructed and sent.  In multi-pass modes, this array points to the
+   * current MCU's blocks within the virtual arrays.
    */
   JBLOCKROW MCU_buffer[C_MAX_BLOCKS_IN_MCU];
 
@@ -54,17 +54,17 @@
   jvirt_barray_ptr whole_image[MAX_COMPONENTS];
 } my_coef_controller;
 
-typedef my_coef_controller * my_coef_ptr;
+typedef my_coef_controller *my_coef_ptr;
 
 
 /* Forward declarations */
 METHODDEF(boolean) compress_data
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 #ifdef FULL_COEF_BUFFER_SUPPORTED
 METHODDEF(boolean) compress_first_pass
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 METHODDEF(boolean) compress_output
-    JPP((j_compress_ptr cinfo, JSAMPIMAGE input_buf));
+        (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 #endif
 
 
@@ -143,7 +143,7 @@
 compress_data (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, bi, ci, yindex, yoffset, blockcnt;
@@ -154,7 +154,7 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num <= last_MCU_col;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Determine where data comes from in input_buf and do the DCT thing.
        * Each call on forward_DCT processes a horizontal row of DCT blocks
        * as wide as an MCU; we rely on having allocated the MCU_buffer[] blocks
@@ -166,46 +166,46 @@
        */
       blkn = 0;
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						: compptr->last_col_width;
-	xpos = MCU_col_num * compptr->MCU_sample_width;
-	ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (coef->iMCU_row_num < last_iMCU_row ||
-	      yoffset+yindex < compptr->last_row_height) {
-	    (*cinfo->fdct->forward_DCT) (cinfo, compptr,
-					 input_buf[compptr->component_index],
-					 coef->MCU_buffer[blkn],
-					 ypos, xpos, (JDIMENSION) blockcnt);
-	    if (blockcnt < compptr->MCU_width) {
-	      /* Create some dummy blocks at the right edge of the image. */
-	      jzero_far((void FAR *) coef->MCU_buffer[blkn + blockcnt],
-			(compptr->MCU_width - blockcnt) * SIZEOF(JBLOCK));
-	      for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
-		coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
-	      }
-	    }
-	  } else {
-	    /* Create a row of dummy blocks at the bottom of the image. */
-	    jzero_far((void FAR *) coef->MCU_buffer[blkn],
-		      compptr->MCU_width * SIZEOF(JBLOCK));
-	    for (bi = 0; bi < compptr->MCU_width; bi++) {
-	      coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
-	    }
-	  }
-	  blkn += compptr->MCU_width;
-	  ypos += DCTSIZE;
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                : compptr->last_col_width;
+        xpos = MCU_col_num * compptr->MCU_sample_width;
+        ypos = yoffset * DCTSIZE; /* ypos == (yoffset+yindex) * DCTSIZE */
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yoffset+yindex < compptr->last_row_height) {
+            (*cinfo->fdct->forward_DCT) (cinfo, compptr,
+                                         input_buf[compptr->component_index],
+                                         coef->MCU_buffer[blkn],
+                                         ypos, xpos, (JDIMENSION) blockcnt);
+            if (blockcnt < compptr->MCU_width) {
+              /* Create some dummy blocks at the right edge of the image. */
+              jzero_far((void *) coef->MCU_buffer[blkn + blockcnt],
+                        (compptr->MCU_width - blockcnt) * sizeof(JBLOCK));
+              for (bi = blockcnt; bi < compptr->MCU_width; bi++) {
+                coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn+bi-1][0][0];
+              }
+            }
+          } else {
+            /* Create a row of dummy blocks at the bottom of the image. */
+            jzero_far((void *) coef->MCU_buffer[blkn],
+                      compptr->MCU_width * sizeof(JBLOCK));
+            for (bi = 0; bi < compptr->MCU_width; bi++) {
+              coef->MCU_buffer[blkn+bi][0][0] = coef->MCU_buffer[blkn-1][0][0];
+            }
+          }
+          blkn += compptr->MCU_width;
+          ypos += DCTSIZE;
+        }
       }
       /* Try to write the MCU.  In event of a suspension failure, we will
        * re-DCT the MCU on restart (a bit inefficient, could be fixed...)
        */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -280,17 +280,17 @@
     for (block_row = 0; block_row < block_rows; block_row++) {
       thisblockrow = buffer[block_row];
       (*cinfo->fdct->forward_DCT) (cinfo, compptr,
-				   input_buf[ci], thisblockrow,
-				   (JDIMENSION) (block_row * DCTSIZE),
-				   (JDIMENSION) 0, blocks_across);
+                                   input_buf[ci], thisblockrow,
+                                   (JDIMENSION) (block_row * DCTSIZE),
+                                   (JDIMENSION) 0, blocks_across);
       if (ndummy > 0) {
-	/* Create dummy blocks at the right edge of the image. */
-	thisblockrow += blocks_across; /* => first dummy block */
-	jzero_far((void FAR *) thisblockrow, ndummy * SIZEOF(JBLOCK));
-	lastDC = thisblockrow[-1][0];
-	for (bi = 0; bi < ndummy; bi++) {
-	  thisblockrow[bi][0] = lastDC;
-	}
+        /* Create dummy blocks at the right edge of the image. */
+        thisblockrow += blocks_across; /* => first dummy block */
+        jzero_far((void *) thisblockrow, ndummy * sizeof(JBLOCK));
+        lastDC = thisblockrow[-1][0];
+        for (bi = 0; bi < ndummy; bi++) {
+          thisblockrow[bi][0] = lastDC;
+        }
       }
     }
     /* If at end of image, create dummy block rows as needed.
@@ -299,22 +299,22 @@
      * This squeezes a few more bytes out of the resulting file...
      */
     if (coef->iMCU_row_num == last_iMCU_row) {
-      blocks_across += ndummy;	/* include lower right corner */
+      blocks_across += ndummy;  /* include lower right corner */
       MCUs_across = blocks_across / h_samp_factor;
       for (block_row = block_rows; block_row < compptr->v_samp_factor;
-	   block_row++) {
-	thisblockrow = buffer[block_row];
-	lastblockrow = buffer[block_row-1];
-	jzero_far((void FAR *) thisblockrow,
-		  (size_t) (blocks_across * SIZEOF(JBLOCK)));
-	for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
-	  lastDC = lastblockrow[h_samp_factor-1][0];
-	  for (bi = 0; bi < h_samp_factor; bi++) {
-	    thisblockrow[bi][0] = lastDC;
-	  }
-	  thisblockrow += h_samp_factor; /* advance to next MCU in row */
-	  lastblockrow += h_samp_factor;
-	}
+           block_row++) {
+        thisblockrow = buffer[block_row];
+        lastblockrow = buffer[block_row-1];
+        jzero_far((void *) thisblockrow,
+                  (size_t) (blocks_across * sizeof(JBLOCK)));
+        for (MCUindex = 0; MCUindex < MCUs_across; MCUindex++) {
+          lastDC = lastblockrow[h_samp_factor-1][0];
+          for (bi = 0; bi < h_samp_factor; bi++) {
+            thisblockrow[bi][0] = lastDC;
+          }
+          thisblockrow += h_samp_factor; /* advance to next MCU in row */
+          lastblockrow += h_samp_factor;
+        }
       }
     }
   }
@@ -341,7 +341,7 @@
 compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   int blkn, ci, xindex, yindex, yoffset;
   JDIMENSION start_col;
   JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
@@ -364,25 +364,25 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
-	    coef->MCU_buffer[blkn++] = buffer_ptr++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+          }
+        }
       }
       /* Try to write the MCU. */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -408,7 +408,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_c_coef_controller *) coef;
   coef->pub.start_pass = start_pass_coef;
 
@@ -421,14 +421,14 @@
     jpeg_component_info *compptr;
 
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
-				(long) compptr->h_samp_factor),
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor),
-	 (JDIMENSION) compptr->v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+         (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+                                (long) compptr->h_samp_factor),
+         (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+                                (long) compptr->v_samp_factor),
+         (JDIMENSION) compptr->v_samp_factor);
     }
 #else
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
@@ -440,7 +440,7 @@
 
     buffer = (JBLOCKROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                  C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
     for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
       coef->MCU_buffer[i] = buffer + i;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolext.c glmark2-2021.02/src/libjpeg-turbo/jccolext.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jccolext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,148 @@
+/*
+ * jccolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+
+/* This file is included by jccolor.c */
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+rgb_ycc_convert_internal (j_compress_ptr cinfo,
+                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                          JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+                ((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+                 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+                ((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+INLINE
+LOCAL(void)
+rgb_gray_convert_internal (j_compress_ptr cinfo,
+                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                           JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptr[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles extended RGB->plain RGB conversion
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal (j_compress_ptr cinfo,
+                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                          JDIMENSION output_row, int num_rows)
+{
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      outptr0[col] = GETJSAMPLE(inptr[RGB_RED]);
+      outptr1[col] = GETJSAMPLE(inptr[RGB_GREEN]);
+      outptr2[col] = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolext.c.inc glmark2-2021.02/src/libjpeg-turbo/jccolext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jccolext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,114 +0,0 @@
-/*
- * jccolext.c
- *
- * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains input colorspace conversion routines.
- */
-
-
-/* This file is included by jccolor.c */
-
-
-/*
- * Convert some rows of samples to the JPEG colorspace.
- *
- * Note that we change from the application's interleaved-pixel format
- * to our internal noninterleaved, one-plane-per-component format.
- * The input buffer is therefore three times as wide as the output buffer.
- *
- * A starting row offset is provided only for the output buffer.  The caller
- * can easily adjust the passed input_buf value to accommodate any row
- * offset required on that side.
- */
-
-INLINE
-LOCAL(void)
-rgb_ycc_convert_internal (j_compress_ptr cinfo,
-                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-                          JDIMENSION output_row, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr0, outptr1, outptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr0 = output_buf[0][output_row];
-    outptr1 = output_buf[1][output_row];
-    outptr2 = output_buf[2][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
-      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
-       * must be too; we do not need an explicit range-limiting operation.
-       * Hence the value being shifted is never negative, and we don't
-       * need the general RIGHT_SHIFT macro.
-       */
-      /* Y */
-      outptr0[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-      /* Cb */
-      outptr1[col] = (JSAMPLE)
-		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
-		 >> SCALEBITS);
-      /* Cr */
-      outptr2[col] = (JSAMPLE)
-		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
-		 >> SCALEBITS);
-    }
-  }
-}
-
-
-/**************** Cases other than RGB -> YCbCr **************/
-
-
-/*
- * Convert some rows of samples to the JPEG colorspace.
- * This version handles RGB->grayscale conversion, which is the same
- * as the RGB->Y portion of RGB->YCbCr.
- * We assume rgb_ycc_start has been called (we only use the Y tables).
- */
-
-INLINE
-LOCAL(void)
-rgb_gray_convert_internal (j_compress_ptr cinfo,
-                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-                           JDIMENSION output_row, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr = output_buf[0][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[RGB_RED]);
-      g = GETJSAMPLE(inptr[RGB_GREEN]);
-      b = GETJSAMPLE(inptr[RGB_BLUE]);
-      inptr += RGB_PIXELSIZE;
-      /* Y */
-      outptr[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-    }
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolor.c glmark2-2021.02/src/libjpeg-turbo/jccolor.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jccolor.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jccolor.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jccolor.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009-2012, 2015 D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains input colorspace conversion routines.
  */
@@ -14,7 +17,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 
 /* Private subobject */
@@ -23,10 +26,10 @@
   struct jpeg_color_converter pub; /* public fields */
 
   /* Private state for RGB->YCC conversion */
-  INT32 * rgb_ycc_tab;		/* => table for RGB to YCbCr conversion */
+  JLONG *rgb_ycc_tab;           /* => table for RGB to YCbCr conversion */
 } my_color_converter;
 
-typedef my_color_converter * my_cconvert_ptr;
+typedef my_color_converter *my_cconvert_ptr;
 
 
 /**************** RGB -> YCbCr conversion: most common case **************/
@@ -35,9 +38,9 @@
  * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
  * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
  * The conversion equations to be implemented are therefore
- *	Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
- *	Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
- *	Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
+ *      Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+ *      Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B  + CENTERJSAMPLE
+ *      Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B  + CENTERJSAMPLE
  * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
  * Note: older versions of the IJG code used a zero offset of MAXJSAMPLE/2,
  * rather than CENTERJSAMPLE, for Cb and Cr.  This gave equal positive and
@@ -59,10 +62,10 @@
  * in the tables to save adding them separately in the inner loop.
  */
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define CBCR_OFFSET	((INT32) CENTERJSAMPLE << SCALEBITS)
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define CBCR_OFFSET     ((JLONG) CENTERJSAMPLE << SCALEBITS)
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
 
 /* We allocate one big table and divide it up into eight parts, instead of
  * doing eight alloc_small requests.  This lets us use a single table base
@@ -70,21 +73,21 @@
  * machines (more than can hold all eight addresses, anyway).
  */
 
-#define R_Y_OFF		0			/* offset to R => Y section */
-#define G_Y_OFF		(1*(MAXJSAMPLE+1))	/* offset to G => Y section */
-#define B_Y_OFF		(2*(MAXJSAMPLE+1))	/* etc. */
-#define R_CB_OFF	(3*(MAXJSAMPLE+1))
-#define G_CB_OFF	(4*(MAXJSAMPLE+1))
-#define B_CB_OFF	(5*(MAXJSAMPLE+1))
-#define R_CR_OFF	B_CB_OFF		/* B=>Cb, R=>Cr are the same */
-#define G_CR_OFF	(6*(MAXJSAMPLE+1))
-#define B_CR_OFF	(7*(MAXJSAMPLE+1))
-#define TABLE_SIZE	(8*(MAXJSAMPLE+1))
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1*(MAXJSAMPLE+1))      /* offset to G => Y section */
+#define B_Y_OFF         (2*(MAXJSAMPLE+1))      /* etc. */
+#define R_CB_OFF        (3*(MAXJSAMPLE+1))
+#define G_CB_OFF        (4*(MAXJSAMPLE+1))
+#define B_CB_OFF        (5*(MAXJSAMPLE+1))
+#define R_CR_OFF        B_CB_OFF                /* B=>Cb, R=>Cr are the same */
+#define G_CR_OFF        (6*(MAXJSAMPLE+1))
+#define B_CR_OFF        (7*(MAXJSAMPLE+1))
+#define TABLE_SIZE      (8*(MAXJSAMPLE+1))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jccolext.c.inc"
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -96,13 +99,15 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define rgb_ycc_convert_internal extrgb_ycc_convert_internal
 #define rgb_gray_convert_internal extrgb_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extrgb_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
@@ -110,13 +115,15 @@
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define rgb_ycc_convert_internal extrgbx_ycc_convert_internal
 #define rgb_gray_convert_internal extrgbx_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extrgbx_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGR_RED
 #define RGB_GREEN EXT_BGR_GREEN
@@ -124,13 +131,15 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define rgb_ycc_convert_internal extbgr_ycc_convert_internal
 #define rgb_gray_convert_internal extbgr_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extbgr_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
@@ -138,13 +147,15 @@
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define rgb_ycc_convert_internal extbgrx_ycc_convert_internal
 #define rgb_gray_convert_internal extbgrx_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extbgrx_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
@@ -152,13 +163,15 @@
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define rgb_ycc_convert_internal extxbgr_ycc_convert_internal
 #define rgb_gray_convert_internal extxbgr_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extxbgr_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
@@ -166,13 +179,15 @@
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define rgb_ycc_convert_internal extxrgb_ycc_convert_internal
 #define rgb_gray_convert_internal extxrgb_gray_convert_internal
-#include "jccolext.c.inc"
+#define rgb_rgb_convert_internal extxrgb_rgb_convert_internal
+#include "jccolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef rgb_ycc_convert_internal
 #undef rgb_gray_convert_internal
+#undef rgb_rgb_convert_internal
 
 
 /*
@@ -183,13 +198,13 @@
 rgb_ycc_start (j_compress_ptr cinfo)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  INT32 * rgb_ycc_tab;
-  INT32 i;
+  JLONG *rgb_ycc_tab;
+  JLONG i;
 
   /* Allocate and fill in the conversion tables. */
-  cconvert->rgb_ycc_tab = rgb_ycc_tab = (INT32 *)
+  cconvert->rgb_ycc_tab = rgb_ycc_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(TABLE_SIZE * SIZEOF(INT32)));
+                                (TABLE_SIZE * sizeof(JLONG)));
 
   for (i = 0; i <= MAXJSAMPLE; i++) {
     rgb_ycc_tab[i+R_Y_OFF] = FIX(0.29900) * i;
@@ -217,8 +232,8 @@
 
 METHODDEF(void)
 rgb_ycc_convert (j_compress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		 JDIMENSION output_row, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                 JDIMENSION output_row, int num_rows)
 {
   switch (cinfo->in_color_space) {
     case JCS_EXT_RGB:
@@ -266,8 +281,8 @@
 
 METHODDEF(void)
 rgb_gray_convert (j_compress_ptr cinfo,
-		  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		  JDIMENSION output_row, int num_rows)
+                  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                  JDIMENSION output_row, int num_rows)
 {
   switch (cinfo->in_color_space) {
     case JCS_EXT_RGB:
@@ -307,6 +322,52 @@
 
 
 /*
+ * Extended RGB to plain RGB conversion
+ */
+
+METHODDEF(void)
+rgb_rgb_convert (j_compress_ptr cinfo,
+                  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                  JDIMENSION output_row, int num_rows)
+{
+  switch (cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      extrgbx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      extbgrx_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      extxbgr_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      extxrgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    default:
+      rgb_rgb_convert_internal(cinfo, input_buf, output_buf, output_row,
+                               num_rows);
+      break;
+  }
+}
+
+
+/*
  * Convert some rows of samples to the JPEG colorspace.
  * This version handles Adobe-style CMYK->YCCK conversion,
  * where we convert R=1-C, G=1-M, and B=1-Y to YCbCr using the same
@@ -316,12 +377,12 @@
 
 METHODDEF(void)
 cmyk_ycck_convert (j_compress_ptr cinfo,
-		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		   JDIMENSION output_row, int num_rows)
+                   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                   JDIMENSION output_row, int num_rows)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JLONG *ctab = cconvert->rgb_ycc_tab;
   register JSAMPROW inptr;
   register JSAMPROW outptr0, outptr1, outptr2, outptr3;
   register JDIMENSION col;
@@ -339,7 +400,7 @@
       g = MAXJSAMPLE - GETJSAMPLE(inptr[1]);
       b = MAXJSAMPLE - GETJSAMPLE(inptr[2]);
       /* K passes through as-is */
-      outptr3[col] = inptr[3];	/* don't need GETJSAMPLE here */
+      outptr3[col] = inptr[3];  /* don't need GETJSAMPLE here */
       inptr += 4;
       /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
        * must be too; we do not need an explicit range-limiting operation.
@@ -348,16 +409,16 @@
        */
       /* Y */
       outptr0[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
       /* Cb */
       outptr1[col] = (JSAMPLE)
-		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+                 >> SCALEBITS);
       /* Cr */
       outptr2[col] = (JSAMPLE)
-		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
-		 >> SCALEBITS);
+                ((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+                 >> SCALEBITS);
     }
   }
 }
@@ -371,8 +432,8 @@
 
 METHODDEF(void)
 grayscale_convert (j_compress_ptr cinfo,
-		   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-		   JDIMENSION output_row, int num_rows)
+                   JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                   JDIMENSION output_row, int num_rows)
 {
   register JSAMPROW inptr;
   register JSAMPROW outptr;
@@ -385,7 +446,7 @@
     outptr = output_buf[0][output_row];
     output_row++;
     for (col = 0; col < num_cols; col++) {
-      outptr[col] = inptr[0];	/* don't need GETJSAMPLE() here */
+      outptr[col] = inptr[0];   /* don't need GETJSAMPLE() here */
       inptr += instride;
     }
   }
@@ -400,28 +461,58 @@
 
 METHODDEF(void)
 null_convert (j_compress_ptr cinfo,
-	      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-	      JDIMENSION output_row, int num_rows)
+              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+              JDIMENSION output_row, int num_rows)
 {
   register JSAMPROW inptr;
-  register JSAMPROW outptr;
+  register JSAMPROW outptr, outptr0, outptr1, outptr2, outptr3;
   register JDIMENSION col;
   register int ci;
   int nc = cinfo->num_components;
   JDIMENSION num_cols = cinfo->image_width;
 
-  while (--num_rows >= 0) {
-    /* It seems fastest to make a separate pass for each component. */
-    for (ci = 0; ci < nc; ci++) {
-      inptr = *input_buf;
-      outptr = output_buf[ci][output_row];
+  if (nc == 3) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf[0][output_row];
+      outptr1 = output_buf[1][output_row];
+      outptr2 = output_buf[2][output_row];
+      output_row++;
       for (col = 0; col < num_cols; col++) {
-	outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
-	inptr += nc;
+        outptr0[col] = *inptr++;
+        outptr1[col] = *inptr++;
+        outptr2[col] = *inptr++;
       }
     }
-    input_buf++;
-    output_row++;
+  } else if (nc == 4) {
+    while (--num_rows >= 0) {
+      inptr = *input_buf++;
+      outptr0 = output_buf[0][output_row];
+      outptr1 = output_buf[1][output_row];
+      outptr2 = output_buf[2][output_row];
+      outptr3 = output_buf[3][output_row];
+      output_row++;
+      for (col = 0; col < num_cols; col++) {
+        outptr0[col] = *inptr++;
+        outptr1[col] = *inptr++;
+        outptr2[col] = *inptr++;
+        outptr3[col] = *inptr++;
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      /* It seems fastest to make a separate pass for each component. */
+      for (ci = 0; ci < nc; ci++) {
+        inptr = *input_buf;
+        outptr = output_buf[ci][output_row];
+        for (col = 0; col < num_cols; col++) {
+          outptr[col] = inptr[ci]; /* don't need GETJSAMPLE() here */
+          inptr += nc;
+        }
+      }
+      input_buf++;
+      output_row++;
+    }
   }
 }
 
@@ -448,7 +539,7 @@
 
   cconvert = (my_cconvert_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_color_converter));
+                                sizeof(my_color_converter));
   cinfo->cconvert = (struct jpeg_color_converter *) cconvert;
   /* set start_pass to null method until we find out differently */
   cconvert->pub.start_pass = null_method;
@@ -486,7 +577,7 @@
       ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
     break;
 
-  default:			/* JCS_UNKNOWN can be anything */
+  default:                      /* JCS_UNKNOWN can be anything */
     if (cinfo->input_components < 1)
       ERREXIT(cinfo, JERR_BAD_IN_COLORSPACE);
     break;
@@ -523,21 +614,30 @@
     break;
 
   case JCS_RGB:
-  case JCS_EXT_RGB:
-  case JCS_EXT_RGBX:
-  case JCS_EXT_BGR:
-  case JCS_EXT_BGRX:
-  case JCS_EXT_XBGR:
-  case JCS_EXT_XRGB:
-  case JCS_EXT_RGBA:
-  case JCS_EXT_BGRA:
-  case JCS_EXT_ABGR:
-  case JCS_EXT_ARGB:
     if (cinfo->num_components != 3)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space == cinfo->jpeg_color_space &&
-      rgb_pixelsize[cinfo->in_color_space] == 3)
-      cconvert->pub.color_convert = null_convert;
+    if (rgb_red[cinfo->in_color_space] == 0 &&
+        rgb_green[cinfo->in_color_space] == 1 &&
+        rgb_blue[cinfo->in_color_space] == 2 &&
+        rgb_pixelsize[cinfo->in_color_space] == 3) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else if (cinfo->in_color_space == JCS_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGB ||
+               cinfo->in_color_space == JCS_EXT_RGBX ||
+               cinfo->in_color_space == JCS_EXT_BGR ||
+               cinfo->in_color_space == JCS_EXT_BGRX ||
+               cinfo->in_color_space == JCS_EXT_XBGR ||
+               cinfo->in_color_space == JCS_EXT_XRGB ||
+               cinfo->in_color_space == JCS_EXT_RGBA ||
+               cinfo->in_color_space == JCS_EXT_BGRA ||
+               cinfo->in_color_space == JCS_EXT_ABGR ||
+               cinfo->in_color_space == JCS_EXT_ARGB)
+      cconvert->pub.color_convert = rgb_rgb_convert;
     else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
@@ -562,18 +662,28 @@
         cconvert->pub.start_pass = rgb_ycc_start;
         cconvert->pub.color_convert = rgb_ycc_convert;
       }
-    } else if (cinfo->in_color_space == JCS_YCbCr)
-      cconvert->pub.color_convert = null_convert;
-    else
+    } else if (cinfo->in_color_space == JCS_YCbCr) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
   case JCS_CMYK:
     if (cinfo->num_components != 4)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
-    if (cinfo->in_color_space == JCS_CMYK)
-      cconvert->pub.color_convert = null_convert;
-    else
+    if (cinfo->in_color_space == JCS_CMYK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
@@ -583,17 +693,27 @@
     if (cinfo->in_color_space == JCS_CMYK) {
       cconvert->pub.start_pass = rgb_ycc_start;
       cconvert->pub.color_convert = cmyk_ycck_convert;
-    } else if (cinfo->in_color_space == JCS_YCCK)
-      cconvert->pub.color_convert = null_convert;
-    else
+    } else if (cinfo->in_color_space == JCS_YCCK) {
+#if defined(__mips__)
+      if (jsimd_c_can_null_convert())
+        cconvert->pub.color_convert = jsimd_c_null_convert;
+      else
+#endif
+        cconvert->pub.color_convert = null_convert;
+    } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
-  default:			/* allow null conversion of JCS_UNKNOWN */
+  default:                      /* allow null conversion of JCS_UNKNOWN */
     if (cinfo->jpeg_color_space != cinfo->in_color_space ||
-	cinfo->num_components != cinfo->input_components)
+        cinfo->num_components != cinfo->input_components)
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
-    cconvert->pub.color_convert = null_convert;
+#if defined(__mips__)
+    if (jsimd_c_can_null_convert())
+      cconvert->pub.color_convert = jsimd_c_null_convert;
+    else
+#endif
+      cconvert->pub.color_convert = null_convert;
     break;
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcdctmgr.c glmark2-2021.02/src/libjpeg-turbo/jcdctmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcdctmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcdctmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,14 @@
 /*
  * jcdctmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2011 D. R. Commander
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2011, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the forward-DCT management logic.
  * This code selects a particular DCT implementation to be used,
@@ -17,33 +19,32 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 #include "jsimddct.h"
 
 
 /* Private subobject for this module */
 
-typedef JMETHOD(void, forward_DCT_method_ptr, (DCTELEM * data));
-typedef JMETHOD(void, float_DCT_method_ptr, (FAST_FLOAT * data));
+typedef void (*forward_DCT_method_ptr) (DCTELEM *data);
+typedef void (*float_DCT_method_ptr) (FAST_FLOAT *data);
 
-typedef JMETHOD(void, convsamp_method_ptr,
-                (JSAMPARRAY sample_data, JDIMENSION start_col,
-                 DCTELEM * workspace));
-typedef JMETHOD(void, float_convsamp_method_ptr,
-                (JSAMPARRAY sample_data, JDIMENSION start_col,
-                 FAST_FLOAT *workspace));
-
-typedef JMETHOD(void, quantize_method_ptr,
-                (JCOEFPTR coef_block, DCTELEM * divisors,
-                 DCTELEM * workspace));
-typedef JMETHOD(void, float_quantize_method_ptr,
-                (JCOEFPTR coef_block, FAST_FLOAT * divisors,
-                 FAST_FLOAT * workspace));
+typedef void (*convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                     JDIMENSION start_col,
+                                     DCTELEM *workspace);
+typedef void (*float_convsamp_method_ptr) (JSAMPARRAY sample_data,
+                                           JDIMENSION start_col,
+                                           FAST_FLOAT *workspace);
+
+typedef void (*quantize_method_ptr) (JCOEFPTR coef_block, DCTELEM *divisors,
+                                     DCTELEM *workspace);
+typedef void (*float_quantize_method_ptr) (JCOEFPTR coef_block,
+                                           FAST_FLOAT *divisors,
+                                           FAST_FLOAT *workspace);
 
 METHODDEF(void) quantize (JCOEFPTR, DCTELEM *, DCTELEM *);
 
 typedef struct {
-  struct jpeg_forward_dct pub;	/* public fields */
+  struct jpeg_forward_dct pub;  /* public fields */
 
   /* Pointer to the DCT routine actually in use */
   forward_DCT_method_ptr dct;
@@ -54,27 +55,30 @@
    * entries, because of scaling (especially for an unnormalized DCT).
    * Each table is given in normal array order.
    */
-  DCTELEM * divisors[NUM_QUANT_TBLS];
+  DCTELEM *divisors[NUM_QUANT_TBLS];
 
   /* work area for FDCT subroutine */
-  DCTELEM * workspace;
+  DCTELEM *workspace;
 
 #ifdef DCT_FLOAT_SUPPORTED
   /* Same as above for the floating-point case. */
   float_DCT_method_ptr float_dct;
   float_convsamp_method_ptr float_convsamp;
   float_quantize_method_ptr float_quantize;
-  FAST_FLOAT * float_divisors[NUM_QUANT_TBLS];
-  FAST_FLOAT * float_workspace;
+  FAST_FLOAT *float_divisors[NUM_QUANT_TBLS];
+  FAST_FLOAT *float_workspace;
 #endif
 } my_fdct_controller;
 
-typedef my_fdct_controller * my_fdct_ptr;
+typedef my_fdct_controller *my_fdct_ptr;
 
 
+#if BITS_IN_JSAMPLE == 8
+
 /*
  * Find the highest bit in an integer through binary search.
  */
+
 LOCAL(int)
 flss (UINT16 val)
 {
@@ -105,6 +109,7 @@
   return bit;
 }
 
+
 /*
  * Compute values to do a division using reciprocal.
  *
@@ -146,7 +151,7 @@
  *
  * In order to allow SIMD implementations we also tweak the values to
  * allow the same calculation to be made at all times:
- * 
+ *
  *   dctbl[0] = f rounded to nearest integer
  *   dctbl[1] = divisor / 2 (+ 1 if fractional part of f < 0.5)
  *   dctbl[2] = 1 << ((word size) * 2 - r)
@@ -163,13 +168,27 @@
  * of in a consecutive manner, yet again in order to allow SIMD
  * routines.
  */
+
 LOCAL(int)
-compute_reciprocal (UINT16 divisor, DCTELEM * dtbl)
+compute_reciprocal (UINT16 divisor, DCTELEM *dtbl)
 {
   UDCTELEM2 fq, fr;
   UDCTELEM c;
   int b, r;
 
+  if (divisor == 1) {
+    /* divisor == 1 means unquantized, so these reciprocal/correction/shift
+     * values will cause the C quantization algorithm to act like the
+     * identity function.  Since only the C quantization algorithm is used in
+     * these cases, the scale value is irrelevant.
+     */
+    dtbl[DCTSIZE2 * 0] = (DCTELEM) 1;                       /* reciprocal */
+    dtbl[DCTSIZE2 * 1] = (DCTELEM) 0;                       /* correction */
+    dtbl[DCTSIZE2 * 2] = (DCTELEM) 1;                       /* scale */
+    dtbl[DCTSIZE2 * 3] = -(DCTELEM) (sizeof(DCTELEM) * 8);  /* shift */
+    return 0;
+  }
+
   b = flss(divisor) - 1;
   r  = sizeof(DCTELEM) * 8 + b;
 
@@ -190,13 +209,20 @@
 
   dtbl[DCTSIZE2 * 0] = (DCTELEM) fq;      /* reciprocal */
   dtbl[DCTSIZE2 * 1] = (DCTELEM) c;       /* correction + roundfactor */
+#ifdef WITH_SIMD
   dtbl[DCTSIZE2 * 2] = (DCTELEM) (1 << (sizeof(DCTELEM)*8*2 - r));  /* scale */
+#else
+  dtbl[DCTSIZE2 * 2] = 1;
+#endif
   dtbl[DCTSIZE2 * 3] = (DCTELEM) r - sizeof(DCTELEM)*8; /* shift */
 
   if(r <= 16) return 0;
   else return 1;
 }
 
+#endif
+
+
 /*
  * Initialize for a processing pass.
  * Verify that all referenced Q-tables are present, and set up
@@ -212,15 +238,15 @@
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
   int ci, qtblno, i;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtbl;
-  DCTELEM * dtbl;
+  JQUANT_TBL *qtbl;
+  DCTELEM *dtbl;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     qtblno = compptr->quant_tbl_no;
     /* Make sure specified quantization table is present */
     if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
-	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+        cinfo->quant_tbl_ptrs[qtblno] == NULL)
       ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
     qtbl = cinfo->quant_tbl_ptrs[qtblno];
     /* Compute divisors for this quant table */
@@ -232,91 +258,102 @@
        * coefficients multiplied by 8 (to counteract scaling).
        */
       if (fdct->divisors[qtblno] == NULL) {
-	fdct->divisors[qtblno] = (DCTELEM *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      (DCTSIZE2 * 4) * SIZEOF(DCTELEM));
+        fdct->divisors[qtblno] = (DCTELEM *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      (DCTSIZE2 * 4) * sizeof(DCTELEM));
       }
       dtbl = fdct->divisors[qtblno];
       for (i = 0; i < DCTSIZE2; i++) {
-	if(!compute_reciprocal(qtbl->quantval[i] << 3, &dtbl[i])
-	  && fdct->quantize == jsimd_quantize)
-	  fdct->quantize = quantize;
+#if BITS_IN_JSAMPLE == 8
+        if (!compute_reciprocal(qtbl->quantval[i] << 3, &dtbl[i]) &&
+            fdct->quantize == jsimd_quantize)
+          fdct->quantize = quantize;
+#else
+        dtbl[i] = ((DCTELEM) qtbl->quantval[i]) << 3;
+#endif
       }
       break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
     case JDCT_IFAST:
       {
-	/* For AA&N IDCT method, divisors are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * We apply a further scale factor of 8.
-	 */
+        /* For AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         */
 #define CONST_BITS 14
-	static const INT16 aanscales[DCTSIZE2] = {
-	  /* precomputed values scaled up by 14 bits */
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
-	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
-	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
-	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
-	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
-	};
-	SHIFT_TEMPS
-
-	if (fdct->divisors[qtblno] == NULL) {
-	  fdct->divisors[qtblno] = (DCTELEM *)
-	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-					(DCTSIZE2 * 4) * SIZEOF(DCTELEM));
-	}
-	dtbl = fdct->divisors[qtblno];
-	for (i = 0; i < DCTSIZE2; i++) {
-	  if(!compute_reciprocal(
-	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
-				  (INT32) aanscales[i]),
-		    CONST_BITS-3), &dtbl[i])
-	    && fdct->quantize == jsimd_quantize)
-	    fdct->quantize = quantize;
-	}
+        static const INT16 aanscales[DCTSIZE2] = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        if (fdct->divisors[qtblno] == NULL) {
+          fdct->divisors[qtblno] = (DCTELEM *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                        (DCTSIZE2 * 4) * sizeof(DCTELEM));
+        }
+        dtbl = fdct->divisors[qtblno];
+        for (i = 0; i < DCTSIZE2; i++) {
+#if BITS_IN_JSAMPLE == 8
+          if (!compute_reciprocal(
+                DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                      (JLONG) aanscales[i]),
+                        CONST_BITS-3), &dtbl[i]) &&
+              fdct->quantize == jsimd_quantize)
+            fdct->quantize = quantize;
+#else
+           dtbl[i] = (DCTELEM)
+             DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                   (JLONG) aanscales[i]),
+                     CONST_BITS-3);
+#endif
+        }
       }
       break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
     case JDCT_FLOAT:
       {
-	/* For float AA&N IDCT method, divisors are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * We apply a further scale factor of 8.
-	 * What's actually stored is 1/divisor so that the inner loop can
-	 * use a multiplication rather than a division.
-	 */
-	FAST_FLOAT * fdtbl;
-	int row, col;
-	static const double aanscalefactor[DCTSIZE] = {
-	  1.0, 1.387039845, 1.306562965, 1.175875602,
-	  1.0, 0.785694958, 0.541196100, 0.275899379
-	};
-
-	if (fdct->float_divisors[qtblno] == NULL) {
-	  fdct->float_divisors[qtblno] = (FAST_FLOAT *)
-	    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-					DCTSIZE2 * SIZEOF(FAST_FLOAT));
-	}
-	fdtbl = fdct->float_divisors[qtblno];
-	i = 0;
-	for (row = 0; row < DCTSIZE; row++) {
-	  for (col = 0; col < DCTSIZE; col++) {
-	    fdtbl[i] = (FAST_FLOAT)
-	      (1.0 / (((double) qtbl->quantval[i] *
-		       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
-	    i++;
-	  }
-	}
+        /* For float AA&N IDCT method, divisors are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * We apply a further scale factor of 8.
+         * What's actually stored is 1/divisor so that the inner loop can
+         * use a multiplication rather than a division.
+         */
+        FAST_FLOAT *fdtbl;
+        int row, col;
+        static const double aanscalefactor[DCTSIZE] = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        if (fdct->float_divisors[qtblno] == NULL) {
+          fdct->float_divisors[qtblno] = (FAST_FLOAT *)
+            (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                        DCTSIZE2 * sizeof(FAST_FLOAT));
+        }
+        fdtbl = fdct->float_divisors[qtblno];
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fdtbl[i] = (FAST_FLOAT)
+              (1.0 / (((double) qtbl->quantval[i] *
+                       aanscalefactor[row] * aanscalefactor[col] * 8.0)));
+            i++;
+          }
+        }
       }
       break;
 #endif
@@ -333,7 +370,7 @@
  */
 
 METHODDEF(void)
-convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM * workspace)
+convsamp (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace)
 {
   register DCTELEM *workspaceptr;
   register JSAMPROW elemptr;
@@ -343,7 +380,7 @@
   for (elemr = 0; elemr < DCTSIZE; elemr++) {
     elemptr = sample_data[elemr] + start_col;
 
-#if DCTSIZE == 8		/* unroll the inner loop */
+#if DCTSIZE == 8                /* unroll the inner loop */
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
     *workspaceptr++ = GETJSAMPLE(*elemptr++) - CENTERJSAMPLE;
@@ -368,14 +405,18 @@
  */
 
 METHODDEF(void)
-quantize (JCOEFPTR coef_block, DCTELEM * divisors, DCTELEM * workspace)
+quantize (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace)
 {
   int i;
   DCTELEM temp;
-  UDCTELEM recip, corr, shift;
-  UDCTELEM2 product;
   JCOEFPTR output_ptr = coef_block;
 
+#if BITS_IN_JSAMPLE == 8
+
+  UDCTELEM recip, corr;
+  int shift;
+  UDCTELEM2 product;
+
   for (i = 0; i < DCTSIZE2; i++) {
     temp = workspace[i];
     recip = divisors[i + DCTSIZE2 * 0];
@@ -386,16 +427,54 @@
       temp = -temp;
       product = (UDCTELEM2)(temp + corr) * recip;
       product >>= shift + sizeof(DCTELEM)*8;
-      temp = product;
+      temp = (DCTELEM)product;
       temp = -temp;
     } else {
       product = (UDCTELEM2)(temp + corr) * recip;
       product >>= shift + sizeof(DCTELEM)*8;
-      temp = product;
+      temp = (DCTELEM)product;
     }
+    output_ptr[i] = (JCOEF) temp;
+  }
+
+#else
+
+  register DCTELEM qval;
 
+  for (i = 0; i < DCTSIZE2; i++) {
+    qval = divisors[i];
+    temp = workspace[i];
+    /* Divide the coefficient value by qval, ensuring proper rounding.
+     * Since C does not specify the direction of rounding for negative
+     * quotients, we have to force the dividend positive for portability.
+     *
+     * In most files, at least half of the output values will be zero
+     * (at default quantization settings, more like three-quarters...)
+     * so we should ensure that this case is fast.  On many machines,
+     * a comparison is enough cheaper than a divide to make a special test
+     * a win.  Since both inputs will be nonnegative, we need only test
+     * for a < b to discover whether a/b is 0.
+     * If your machine's division is fast enough, define FAST_DIVIDE.
+     */
+#ifdef FAST_DIVIDE
+#define DIVIDE_BY(a,b)  a /= b
+#else
+#define DIVIDE_BY(a,b)  if (a >= b) a /= b; else a = 0
+#endif
+    if (temp < 0) {
+      temp = -temp;
+      temp += qval>>1;  /* for rounding */
+      DIVIDE_BY(temp, qval);
+      temp = -temp;
+    } else {
+      temp += qval>>1;  /* for rounding */
+      DIVIDE_BY(temp, qval);
+    }
     output_ptr[i] = (JCOEF) temp;
   }
+
+#endif
+
 }
 
 
@@ -408,16 +487,16 @@
  */
 
 METHODDEF(void)
-forward_DCT (j_compress_ptr cinfo, jpeg_component_info * compptr,
-	     JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-	     JDIMENSION start_row, JDIMENSION start_col,
-	     JDIMENSION num_blocks)
+forward_DCT (j_compress_ptr cinfo, jpeg_component_info *compptr,
+             JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+             JDIMENSION start_row, JDIMENSION start_col,
+             JDIMENSION num_blocks)
 /* This version is used for integer DCT implementations. */
 {
   /* This routine is heavily used, so it's worth coding it tightly. */
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
-  DCTELEM * divisors = fdct->divisors[compptr->quant_tbl_no];
-  DCTELEM * workspace;
+  DCTELEM *divisors = fdct->divisors[compptr->quant_tbl_no];
+  DCTELEM *workspace;
   JDIMENSION bi;
 
   /* Make sure the compiler doesn't look up these every pass */
@@ -426,7 +505,7 @@
   quantize_method_ptr do_quantize = fdct->quantize;
   workspace = fdct->workspace;
 
-  sample_data += start_row;	/* fold in the vertical offset once */
+  sample_data += start_row;     /* fold in the vertical offset once */
 
   for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
     /* Load data into workspace, applying unsigned->signed conversion */
@@ -445,7 +524,7 @@
 
 
 METHODDEF(void)
-convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT * workspace)
+convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace)
 {
   register FAST_FLOAT *workspaceptr;
   register JSAMPROW elemptr;
@@ -454,7 +533,7 @@
   workspaceptr = workspace;
   for (elemr = 0; elemr < DCTSIZE; elemr++) {
     elemptr = sample_data[elemr] + start_col;
-#if DCTSIZE == 8		/* unroll the inner loop */
+#if DCTSIZE == 8                /* unroll the inner loop */
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
     *workspaceptr++ = (FAST_FLOAT)(GETJSAMPLE(*elemptr++) - CENTERJSAMPLE);
@@ -476,7 +555,7 @@
 
 
 METHODDEF(void)
-quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors, FAST_FLOAT * workspace)
+quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace)
 {
   register FAST_FLOAT temp;
   register int i;
@@ -498,16 +577,16 @@
 
 
 METHODDEF(void)
-forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-		   JDIMENSION start_row, JDIMENSION start_col,
-		   JDIMENSION num_blocks)
+forward_DCT_float (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                   JDIMENSION start_row, JDIMENSION start_col,
+                   JDIMENSION num_blocks)
 /* This version is used for floating-point DCT implementations. */
 {
   /* This routine is heavily used, so it's worth coding it tightly. */
   my_fdct_ptr fdct = (my_fdct_ptr) cinfo->fdct;
-  FAST_FLOAT * divisors = fdct->float_divisors[compptr->quant_tbl_no];
-  FAST_FLOAT * workspace;
+  FAST_FLOAT *divisors = fdct->float_divisors[compptr->quant_tbl_no];
+  FAST_FLOAT *workspace;
   JDIMENSION bi;
 
 
@@ -517,7 +596,7 @@
   float_quantize_method_ptr do_quantize = fdct->float_quantize;
   workspace = fdct->float_workspace;
 
-  sample_data += start_row;	/* fold in the vertical offset once */
+  sample_data += start_row;     /* fold in the vertical offset once */
 
   for (bi = 0; bi < num_blocks; bi++, start_col += DCTSIZE) {
     /* Load data into workspace, applying unsigned->signed conversion */
@@ -546,7 +625,7 @@
 
   fdct = (my_fdct_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_fdct_controller));
+                                sizeof(my_fdct_controller));
   cinfo->fdct = (struct jpeg_forward_dct *) fdct;
   fdct->pub.start_pass = start_pass_fdctmgr;
 
@@ -625,12 +704,12 @@
   if (cinfo->dct_method == JDCT_FLOAT)
     fdct->float_workspace = (FAST_FLOAT *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(FAST_FLOAT) * DCTSIZE2);
+                                  sizeof(FAST_FLOAT) * DCTSIZE2);
   else
 #endif
     fdct->workspace = (DCTELEM *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(DCTELEM) * DCTSIZE2);
+                                  sizeof(DCTELEM) * DCTSIZE2);
 
   /* Mark divisor tables unallocated */
   for (i = 0; i < NUM_QUANT_TBLS; i++) {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jchuff.c glmark2-2021.02/src/libjpeg-turbo/jchuff.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jchuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jchuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jchuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2014-2016 D. R. Commander.
+ * Copyright (C) 2015 Matthieu Darbois.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy encoding routines.
  *
@@ -18,11 +21,39 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jchuff.h"		/* Declarations shared with jcphuff.c */
+#include "jsimd.h"
+#include "jconfigint.h"
 #include <limits.h>
 
-static unsigned char jpeg_nbits_table[65536];
-static int jpeg_nbits_table_init = 0;
+/*
+ * NOTE: If USE_CLZ_INTRINSIC is defined, then clz/bsr instructions will be
+ * used for bit counting rather than the lookup table.  This will reduce the
+ * memory footprint by 64k, which is important for some mobile applications
+ * that create many isolated instances of libjpeg-turbo (web browsers, for
+ * instance.)  This may improve performance on some mobile platforms as well.
+ * This feature is enabled by default only on ARM processors, because some x86
+ * chips have a slow implementation of bsr, and the use of clz/bsr cannot be
+ * shown to have a significant performance impact even on the x86 chips that
+ * have a fast implementation of it.  When building for ARMv6, you can
+ * explicitly disable the use of clz/bsr by adding -mthumb to the compiler
+ * flags (this defines __thumb__).
+ */
+
+/* NOTE: Both GCC and Clang define __GNUC__ */
+#if defined __GNUC__ && (defined __arm__ || defined __aarch64__)
+#if !defined __thumb__ || defined __thumb2__
+#define USE_CLZ_INTRINSIC
+#endif
+#endif
+
+#ifdef USE_CLZ_INTRINSIC
+#define JPEG_NBITS_NONZERO(x) (32 - __builtin_clz(x))
+#define JPEG_NBITS(x) (x ? JPEG_NBITS_NONZERO(x) : 0)
+#else
+#include "jpeg_nbits_table.h"
+#define JPEG_NBITS(x) (jpeg_nbits_table[x])
+#define JPEG_NBITS_NONZERO(x) JPEG_NBITS(x)
+#endif
 
 #ifndef min
  #define min(a,b) ((a)<(b)?(a):(b))
@@ -36,8 +67,8 @@
  */
 
 typedef struct {
-  size_t put_buffer;		/* current bit-accumulation buffer */
-  int put_bits;			/* # of bits now in it */
+  size_t put_buffer;            /* current bit-accumulation buffer */
+  int put_bits;                 /* # of bits now in it */
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 } savable_state;
 
@@ -51,12 +82,12 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).put_buffer = (src).put_buffer, \
-	 (dest).put_bits = (src).put_bits, \
-	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).put_buffer = (src).put_buffer, \
+         (dest).put_bits = (src).put_bits, \
+         (dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -64,44 +95,45 @@
 typedef struct {
   struct jpeg_entropy_encoder pub; /* public fields */
 
-  savable_state saved;		/* Bit buffer & DC state at start of MCU */
+  savable_state saved;          /* Bit buffer & DC state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  c_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
-  c_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *dc_derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *ac_derived_tbls[NUM_HUFF_TBLS];
 
-#ifdef ENTROPY_OPT_SUPPORTED	/* Statistics tables for optimization */
-  long * dc_count_ptrs[NUM_HUFF_TBLS];
-  long * ac_count_ptrs[NUM_HUFF_TBLS];
+#ifdef ENTROPY_OPT_SUPPORTED    /* Statistics tables for optimization */
+  long *dc_count_ptrs[NUM_HUFF_TBLS];
+  long *ac_count_ptrs[NUM_HUFF_TBLS];
 #endif
+
+  int simd;
 } huff_entropy_encoder;
 
-typedef huff_entropy_encoder * huff_entropy_ptr;
+typedef huff_entropy_encoder *huff_entropy_ptr;
 
 /* Working state while writing an MCU.
  * This struct contains all the fields that are needed by subroutines.
  */
 
 typedef struct {
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
-  savable_state cur;		/* Current bit buffer & DC state */
-  j_compress_ptr cinfo;		/* dump_buffer needs access to this */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  savable_state cur;            /* Current bit buffer & DC state */
+  j_compress_ptr cinfo;         /* dump_buffer needs access to this */
 } working_state;
 
 
 /* Forward declarations */
-METHODDEF(boolean) encode_mcu_huff JPP((j_compress_ptr cinfo,
-					JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_huff JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_huff (j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_huff (j_compress_ptr cinfo);
 #ifdef ENTROPY_OPT_SUPPORTED
-METHODDEF(boolean) encode_mcu_gather JPP((j_compress_ptr cinfo,
-					  JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_gather JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_gather (j_compress_ptr cinfo,
+                                      JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_gather (j_compress_ptr cinfo);
 #endif
 
 
@@ -116,7 +148,7 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, dctbl, actbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (gather_statistics) {
 #ifdef ENTROPY_OPT_SUPPORTED
@@ -130,6 +162,8 @@
     entropy->pub.finish_pass = finish_pass_huff;
   }
 
+  entropy->simd = jsimd_can_huff_encode_one_block();
+
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
     dctbl = compptr->dc_tbl_no;
@@ -139,29 +173,29 @@
       /* Check for invalid table indexes */
       /* (make_c_derived_tbl does this in the other path) */
       if (dctbl < 0 || dctbl >= NUM_HUFF_TBLS)
-	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, dctbl);
       if (actbl < 0 || actbl >= NUM_HUFF_TBLS)
-	ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
+        ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, actbl);
       /* Allocate and zero the statistics tables */
       /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
       if (entropy->dc_count_ptrs[dctbl] == NULL)
-	entropy->dc_count_ptrs[dctbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * SIZEOF(long));
+        entropy->dc_count_ptrs[dctbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->dc_count_ptrs[dctbl], 257 * sizeof(long));
       if (entropy->ac_count_ptrs[actbl] == NULL)
-	entropy->ac_count_ptrs[actbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * SIZEOF(long));
+        entropy->ac_count_ptrs[actbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->ac_count_ptrs[actbl], 257 * sizeof(long));
 #endif
     } else {
       /* Compute derived values for Huffman tables */
       /* We may do this more than once for a table, but it's not expensive */
       jpeg_make_c_derived_tbl(cinfo, TRUE, dctbl,
-			      & entropy->dc_derived_tbls[dctbl]);
+                              & entropy->dc_derived_tbls[dctbl]);
       jpeg_make_c_derived_tbl(cinfo, FALSE, actbl,
-			      & entropy->ac_derived_tbls[actbl]);
+                              & entropy->ac_derived_tbls[actbl]);
     }
     /* Initialize DC predictions to 0 */
     entropy->saved.last_dc_val[ci] = 0;
@@ -186,7 +220,7 @@
 
 GLOBAL(void)
 jpeg_make_c_derived_tbl (j_compress_ptr cinfo, boolean isDC, int tblno,
-			 c_derived_tbl ** pdtbl)
+                         c_derived_tbl **pdtbl)
 {
   JHUFF_TBL *htbl;
   c_derived_tbl *dtbl;
@@ -211,22 +245,22 @@
   if (*pdtbl == NULL)
     *pdtbl = (c_derived_tbl *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(c_derived_tbl));
+                                  sizeof(c_derived_tbl));
   dtbl = *pdtbl;
-  
+
   /* Figure C.1: make table of Huffman code length for each symbol */
 
   p = 0;
   for (l = 1; l <= 16; l++) {
     i = (int) htbl->bits[l];
-    if (i < 0 || p + i > 256)	/* protect against table overrun */
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     while (i--)
       huffsize[p++] = (char) l;
   }
   huffsize[p] = 0;
   lastp = p;
-  
+
   /* Figure C.2: generate the codes themselves */
   /* We also validate that the counts represent a legal Huffman code tree. */
 
@@ -241,12 +275,12 @@
     /* code is now 1 more than the last code used for codelength si; but
      * it must still fit in si bits, since no code is allowed to be all ones.
      */
-    if (((INT32) code) >= (((INT32) 1) << si))
+    if (((JLONG) code) >= (((JLONG) 1) << si))
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     code <<= 1;
     si++;
   }
-  
+
   /* Figure C.3: generate encoding tables */
   /* These are code and size indexed by symbol value */
 
@@ -254,7 +288,7 @@
    * this lets us detect duplicate VAL entries here, and later
    * allows emit_bits to detect any attempt to emit such symbols.
    */
-  MEMZERO(dtbl->ehufsi, SIZEOF(dtbl->ehufsi));
+  MEMZERO(dtbl->ehufsi, sizeof(dtbl->ehufsi));
 
   /* This is also a convenient place to check for out-of-range
    * and duplicated VAL entries.  We allow 0..255 for AC symbols
@@ -270,15 +304,6 @@
     dtbl->ehufco[i] = huffcode[p];
     dtbl->ehufsi[i] = huffsize[p];
   }
-
-  if(!jpeg_nbits_table_init) {
-    for(i = 0; i < 65536; i++) {
-      int nbits = 0, temp = i;
-      while (temp) {temp >>= 1;  nbits++;}
-      jpeg_nbits_table[i] = nbits;
-    }
-    jpeg_nbits_table_init = 1;
-  }
 }
 
 
@@ -286,19 +311,17 @@
 
 /* Emit a byte, taking 'action' if must suspend. */
 #define emit_byte(state,val,action)  \
-	{ *(state)->next_output_byte++ = (JOCTET) (val);  \
-	  if (--(state)->free_in_buffer == 0)  \
-	    if (! dump_buffer(state))  \
-	      { action; } }
+        { *(state)->next_output_byte++ = (JOCTET) (val);  \
+          if (--(state)->free_in_buffer == 0)  \
+            if (! dump_buffer(state))  \
+              { action; } }
 
 
 LOCAL(boolean)
-dump_buffer (working_state * state)
+dump_buffer (working_state *state)
 /* Empty the output buffer; return TRUE if successful, FALSE if must suspend */
 {
-  struct jpeg_destination_mgr * dest = state->cinfo->dest;
-
-  dest->free_in_buffer = state->free_in_buffer;
+  struct jpeg_destination_mgr *dest = state->cinfo->dest;
 
   if (! (*dest->empty_output_buffer) (state->cinfo))
     return FALSE;
@@ -360,7 +383,11 @@
   } \
 }
 
-#if __WORDSIZE==64 || defined(_WIN64)
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
 #define EMIT_BITS(code, size) { \
   CHECKBUF47() \
@@ -368,7 +395,7 @@
 }
 
 #define EMIT_CODE(code, size) { \
-  temp2 &= (((INT32) 1)<<nbits) - 1; \
+  temp2 &= (((JLONG) 1)<<nbits) - 1; \
   CHECKBUF31() \
   PUT_BITS(code, size) \
   PUT_BITS(temp2, nbits) \
@@ -382,7 +409,7 @@
 }
 
 #define EMIT_CODE(code, size) { \
-  temp2 &= (((INT32) 1)<<nbits) - 1; \
+  temp2 &= (((JLONG) 1)<<nbits) - 1; \
   PUT_BITS(code, size) \
   CHECKBUF15() \
   PUT_BITS(temp2, nbits) \
@@ -392,7 +419,16 @@
 #endif
 
 
-#define BUFSIZE (DCTSIZE2 * 2)
+/* Although it is exceedingly rare, it is possible for a Huffman-encoded
+ * coefficient block to be larger than the 128-byte unencoded block.  For each
+ * of the 64 coefficients, PUT_BITS is invoked twice, and each invocation can
+ * theoretically store 16 bits (for a maximum of 2048 bits or 256 bytes per
+ * encoded block.)  If, for instance, one artificially sets the AC
+ * coefficients to alternating values of 32767 and -32768 (using the JPEG
+ * scanning order-- 1, 8, 16, etc.), then this will produce an encoded block
+ * larger than 200 bytes.
+ */
+#define BUFSIZE (DCTSIZE2 * 4)
 
 #define LOAD_BUFFER() { \
   if (state->free_in_buffer < BUFSIZE) { \
@@ -425,7 +461,7 @@
 
 
 LOCAL(boolean)
-flush_bits (working_state * state)
+flush_bits (working_state *state)
 {
   JOCTET _buffer[BUFSIZE], *buffer;
   size_t put_buffer;  int put_bits;
@@ -439,7 +475,7 @@
   PUT_BITS(0x7F, 7)
   while (put_bits >= 8) EMIT_BYTE()
 
-  state->cur.put_buffer = 0;	/* and reset bit-buffer to empty */
+  state->cur.put_buffer = 0;    /* and reset bit-buffer to empty */
   state->cur.put_bits = 0;
   STORE_BUFFER()
 
@@ -450,8 +486,25 @@
 /* Encode a single block's worth of coefficients */
 
 LOCAL(boolean)
-encode_one_block (working_state * state, JCOEFPTR block, int last_dc_val,
-		  c_derived_tbl *dctbl, c_derived_tbl *actbl)
+encode_one_block_simd (working_state *state, JCOEFPTR block, int last_dc_val,
+                       c_derived_tbl *dctbl, c_derived_tbl *actbl)
+{
+  JOCTET _buffer[BUFSIZE], *buffer;
+  size_t bytes, bytestocopy;  int localbuf = 0;
+
+  LOAD_BUFFER()
+
+  buffer = jsimd_huff_encode_one_block(state, buffer, block, last_dc_val,
+                                       dctbl, actbl);
+
+  STORE_BUFFER()
+
+  return TRUE;
+}
+
+LOCAL(boolean)
+encode_one_block (working_state *state, JCOEFPTR block, int last_dc_val,
+                  c_derived_tbl *dctbl, c_derived_tbl *actbl)
 {
   int temp, temp2, temp3;
   int nbits;
@@ -466,7 +519,7 @@
   LOAD_BUFFER()
 
   /* Encode the DC coefficient difference per section F.1.2.1 */
-  
+
   temp = temp2 = block[0] - last_dc_val;
 
  /* This is a well-known technique for obtaining the absolute value without a
@@ -483,25 +536,23 @@
   temp2 += temp3;
 
   /* Find the number of bits needed for the magnitude of the coefficient */
-  nbits = jpeg_nbits_table[temp];
+  nbits = JPEG_NBITS(temp);
 
   /* Emit the Huffman-coded symbol for the number of bits */
   code = dctbl->ehufco[nbits];
   size = dctbl->ehufsi[nbits];
-  PUT_BITS(code, size)
-  CHECKBUF15()
+  EMIT_BITS(code, size)
 
   /* Mask off any extra bits in code */
-  temp2 &= (((INT32) 1)<<nbits) - 1;
+  temp2 &= (((JLONG) 1)<<nbits) - 1;
 
   /* Emit that number of bits of the value, if positive, */
   /* or the complement of its magnitude, if negative. */
-  PUT_BITS(temp2, nbits)
-  CHECKBUF15()
+  EMIT_BITS(temp2, nbits)
 
   /* Encode the AC coefficients per section F.1.2.2 */
-  
-  r = 0;			/* r = run length of zeros */
+
+  r = 0;                        /* r = run length of zeros */
 
 /* Manually unroll the k loop to eliminate the counter variable.  This
  * improves performance greatly on systems with a limited number of
@@ -517,7 +568,7 @@
     temp ^= temp3; \
     temp -= temp3; \
     temp2 += temp3; \
-    nbits = jpeg_nbits_table[temp]; \
+    nbits = JPEG_NBITS_NONZERO(temp); \
     /* if run length > 15, must emit special run-length-16 codes (0xF0) */ \
     while (r > 15) { \
       EMIT_BITS(code_0xf0, size_0xf0) \
@@ -565,7 +616,7 @@
  */
 
 LOCAL(boolean)
-emit_restart (working_state * state, int restart_num)
+emit_restart (working_state *state, int restart_num)
 {
   int ci;
 
@@ -595,7 +646,7 @@
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   working_state state;
   int blkn, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Load up working state */
   state.next_output_byte = cinfo->dest->next_output_byte;
@@ -607,20 +658,34 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! emit_restart(&state, entropy->next_restart_num))
-	return FALSE;
+        return FALSE;
   }
 
   /* Encode the MCU data blocks */
-  for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    ci = cinfo->MCU_membership[blkn];
-    compptr = cinfo->cur_comp_info[ci];
-    if (! encode_one_block(&state,
-			   MCU_data[blkn][0], state.cur.last_dc_val[ci],
-			   entropy->dc_derived_tbls[compptr->dc_tbl_no],
-			   entropy->ac_derived_tbls[compptr->ac_tbl_no]))
-      return FALSE;
-    /* Update last_dc_val */
-    state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+  if (entropy->simd) {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      if (! encode_one_block_simd(&state,
+                                  MCU_data[blkn][0], state.cur.last_dc_val[ci],
+                                  entropy->dc_derived_tbls[compptr->dc_tbl_no],
+                                  entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+    }
+  } else {
+    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
+      ci = cinfo->MCU_membership[blkn];
+      compptr = cinfo->cur_comp_info[ci];
+      if (! encode_one_block(&state,
+                             MCU_data[blkn][0], state.cur.last_dc_val[ci],
+                             entropy->dc_derived_tbls[compptr->dc_tbl_no],
+                             entropy->ac_derived_tbls[compptr->ac_tbl_no]))
+        return FALSE;
+      /* Update last_dc_val */
+      state.cur.last_dc_val[ci] = MCU_data[blkn][0][0];
+    }
   }
 
   /* Completed MCU, so update state */
@@ -687,18 +752,18 @@
 
 LOCAL(void)
 htest_one_block (j_compress_ptr cinfo, JCOEFPTR block, int last_dc_val,
-		 long dc_counts[], long ac_counts[])
+                 long dc_counts[], long ac_counts[])
 {
   register int temp;
   register int nbits;
   register int k, r;
-  
+
   /* Encode the DC coefficient difference per section F.1.2.1 */
-  
+
   temp = block[0] - last_dc_val;
   if (temp < 0)
     temp = -temp;
-  
+
   /* Find the number of bits needed for the magnitude of the coefficient */
   nbits = 0;
   while (temp) {
@@ -713,36 +778,36 @@
 
   /* Count the Huffman symbol for the number of bits */
   dc_counts[nbits]++;
-  
+
   /* Encode the AC coefficients per section F.1.2.2 */
-  
-  r = 0;			/* r = run length of zeros */
-  
+
+  r = 0;                        /* r = run length of zeros */
+
   for (k = 1; k < DCTSIZE2; k++) {
     if ((temp = block[jpeg_natural_order[k]]) == 0) {
       r++;
     } else {
       /* if run length > 15, must emit special run-length-16 codes (0xF0) */
       while (r > 15) {
-	ac_counts[0xF0]++;
-	r -= 16;
+        ac_counts[0xF0]++;
+        r -= 16;
       }
-      
+
       /* Find the number of bits needed for the magnitude of the coefficient */
       if (temp < 0)
-	temp = -temp;
-      
+        temp = -temp;
+
       /* Find the number of bits needed for the magnitude of the coefficient */
-      nbits = 1;		/* there must be at least one 1 bit */
+      nbits = 1;                /* there must be at least one 1 bit */
       while ((temp >>= 1))
-	nbits++;
+        nbits++;
       /* Check for out-of-range coefficient values */
       if (nbits > MAX_COEF_BITS)
-	ERREXIT(cinfo, JERR_BAD_DCT_COEF);
-      
+        ERREXIT(cinfo, JERR_BAD_DCT_COEF);
+
       /* Count Huffman symbol for run length / number of bits */
       ac_counts[(r << 4) + nbits]++;
-      
+
       r = 0;
     }
   }
@@ -763,14 +828,14 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int blkn, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Take care of restart intervals if needed */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0) {
       /* Re-initialize DC predictions to 0 */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++)
-	entropy->saved.last_dc_val[ci] = 0;
+        entropy->saved.last_dc_val[ci] = 0;
       /* Update restart state */
       entropy->restarts_to_go = cinfo->restart_interval;
     }
@@ -781,8 +846,8 @@
     ci = cinfo->MCU_membership[blkn];
     compptr = cinfo->cur_comp_info[ci];
     htest_one_block(cinfo, MCU_data[blkn][0], entropy->saved.last_dc_val[ci],
-		    entropy->dc_count_ptrs[compptr->dc_tbl_no],
-		    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
+                    entropy->dc_count_ptrs[compptr->dc_tbl_no],
+                    entropy->ac_count_ptrs[compptr->ac_tbl_no]);
     entropy->saved.last_dc_val[ci] = MCU_data[blkn][0][0];
   }
 
@@ -819,24 +884,24 @@
  */
 
 GLOBAL(void)
-jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[])
+jpeg_gen_optimal_table (j_compress_ptr cinfo, JHUFF_TBL *htbl, long freq[])
 {
-#define MAX_CLEN 32		/* assumed maximum initial code length */
-  UINT8 bits[MAX_CLEN+1];	/* bits[k] = # of symbols with code length k */
-  int codesize[257];		/* codesize[k] = code length of symbol k */
-  int others[257];		/* next symbol in current branch of tree */
+#define MAX_CLEN 32             /* assumed maximum initial code length */
+  UINT8 bits[MAX_CLEN+1];       /* bits[k] = # of symbols with code length k */
+  int codesize[257];            /* codesize[k] = code length of symbol k */
+  int others[257];              /* next symbol in current branch of tree */
   int c1, c2;
   int p, i, j;
   long v;
 
   /* This algorithm is explained in section K.2 of the JPEG standard */
 
-  MEMZERO(bits, SIZEOF(bits));
-  MEMZERO(codesize, SIZEOF(codesize));
+  MEMZERO(bits, sizeof(bits));
+  MEMZERO(codesize, sizeof(codesize));
   for (i = 0; i < 257; i++)
-    others[i] = -1;		/* init links to empty */
-  
-  freq[256] = 1;		/* make sure 256 has a nonzero count */
+    others[i] = -1;             /* init links to empty */
+
+  freq[256] = 1;                /* make sure 256 has a nonzero count */
   /* Including the pseudo-symbol 256 in the Huffman procedure guarantees
    * that no real symbol is given code-value of all ones, because 256
    * will be placed last in the largest codeword category.
@@ -851,8 +916,8 @@
     v = 1000000000L;
     for (i = 0; i <= 256; i++) {
       if (freq[i] && freq[i] <= v) {
-	v = freq[i];
-	c1 = i;
+        v = freq[i];
+        c1 = i;
       }
     }
 
@@ -862,15 +927,15 @@
     v = 1000000000L;
     for (i = 0; i <= 256; i++) {
       if (freq[i] && freq[i] <= v && i != c1) {
-	v = freq[i];
-	c2 = i;
+        v = freq[i];
+        c2 = i;
       }
     }
 
     /* Done if we've merged everything into one frequency */
     if (c2 < 0)
       break;
-    
+
     /* Else merge the two counts/trees */
     freq[c1] += freq[c2];
     freq[c2] = 0;
@@ -881,9 +946,9 @@
       c1 = others[c1];
       codesize[c1]++;
     }
-    
-    others[c1] = c2;		/* chain c2 onto c1's tree branch */
-    
+
+    others[c1] = c2;            /* chain c2 onto c1's tree branch */
+
     /* Increment the codesize of everything in c2's tree branch */
     codesize[c2]++;
     while (others[c2] >= 0) {
@@ -898,7 +963,7 @@
       /* The JPEG standard seems to think that this can't happen, */
       /* but I'm paranoid... */
       if (codesize[i] > MAX_CLEN)
-	ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
+        ERREXIT(cinfo, JERR_HUFF_CLEN_OVERFLOW);
 
       bits[codesize[i]]++;
     }
@@ -914,28 +979,28 @@
    * shortest nonzero BITS entry is converted into a prefix for two code words
    * one bit longer.
    */
-  
+
   for (i = MAX_CLEN; i > 16; i--) {
     while (bits[i] > 0) {
-      j = i - 2;		/* find length of new prefix to be used */
+      j = i - 2;                /* find length of new prefix to be used */
       while (bits[j] == 0)
-	j--;
-      
-      bits[i] -= 2;		/* remove two symbols */
-      bits[i-1]++;		/* one goes in this length */
-      bits[j+1] += 2;		/* two new symbols in this length */
-      bits[j]--;		/* symbol of this length is now a prefix */
+        j--;
+
+      bits[i] -= 2;             /* remove two symbols */
+      bits[i-1]++;              /* one goes in this length */
+      bits[j+1] += 2;           /* two new symbols in this length */
+      bits[j]--;                /* symbol of this length is now a prefix */
     }
   }
 
   /* Remove the count for the pseudo-symbol 256 from the largest codelength */
-  while (bits[i] == 0)		/* find largest codelength still in use */
+  while (bits[i] == 0)          /* find largest codelength still in use */
     i--;
   bits[i]--;
-  
+
   /* Return final symbol counts (only for lengths 0..16) */
-  MEMCOPY(htbl->bits, bits, SIZEOF(htbl->bits));
-  
+  MEMCOPY(htbl->bits, bits, sizeof(htbl->bits));
+
   /* Return a list of the symbols sorted by code length */
   /* It's not real clear to me why we don't need to consider the codelength
    * changes made above, but the JPEG spec seems to think this works.
@@ -944,8 +1009,8 @@
   for (i = 1; i <= MAX_CLEN; i++) {
     for (j = 0; j <= 255; j++) {
       if (codesize[j] == i) {
-	htbl->huffval[p] = (UINT8) j;
-	p++;
+        htbl->huffval[p] = (UINT8) j;
+        p++;
       }
     }
   }
@@ -964,7 +1029,7 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, dctbl, actbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JHUFF_TBL **htblptr;
   boolean did_dc[NUM_HUFF_TBLS];
   boolean did_ac[NUM_HUFF_TBLS];
@@ -972,8 +1037,8 @@
   /* It's important not to apply jpeg_gen_optimal_table more than once
    * per table, because it clobbers the input frequency counts!
    */
-  MEMZERO(did_dc, SIZEOF(did_dc));
-  MEMZERO(did_ac, SIZEOF(did_ac));
+  MEMZERO(did_dc, sizeof(did_dc));
+  MEMZERO(did_ac, sizeof(did_ac));
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
@@ -982,14 +1047,14 @@
     if (! did_dc[dctbl]) {
       htblptr = & cinfo->dc_huff_tbl_ptrs[dctbl];
       if (*htblptr == NULL)
-	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
       jpeg_gen_optimal_table(cinfo, *htblptr, entropy->dc_count_ptrs[dctbl]);
       did_dc[dctbl] = TRUE;
     }
     if (! did_ac[actbl]) {
       htblptr = & cinfo->ac_huff_tbl_ptrs[actbl];
       if (*htblptr == NULL)
-	*htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
+        *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
       jpeg_gen_optimal_table(cinfo, *htblptr, entropy->ac_count_ptrs[actbl]);
       did_ac[actbl] = TRUE;
     }
@@ -1012,7 +1077,7 @@
 
   entropy = (huff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(huff_entropy_encoder));
+                                sizeof(huff_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass_huff;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jchuff.h glmark2-2021.02/src/libjpeg-turbo/jchuff.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jchuff.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jchuff.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jchuff.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains declarations for Huffman entropy encoding routines
  * that are shared between the sequential encoder (jchuff.c) and the
@@ -25,23 +28,16 @@
 /* Derived data constructed for each Huffman table */
 
 typedef struct {
-  unsigned int ehufco[256];	/* code for each symbol */
-  char ehufsi[256];		/* length of code for each symbol */
+  unsigned int ehufco[256];     /* code for each symbol */
+  char ehufsi[256];             /* length of code for each symbol */
   /* If no code has been allocated for a symbol S, ehufsi[S] contains 0 */
 } c_derived_tbl;
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_make_c_derived_tbl	jMkCDerived
-#define jpeg_gen_optimal_table	jGenOptTbl
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
 /* Expand a Huffman table definition into the derived format */
 EXTERN(void) jpeg_make_c_derived_tbl
-	JPP((j_compress_ptr cinfo, boolean isDC, int tblno,
-	     c_derived_tbl ** pdtbl));
+        (j_compress_ptr cinfo, boolean isDC, int tblno,
+         c_derived_tbl ** pdtbl);
 
 /* Generate an optimal table definition given the specified counts */
 EXTERN(void) jpeg_gen_optimal_table
-	JPP((j_compress_ptr cinfo, JHUFF_TBL * htbl, long freq[]));
+        (j_compress_ptr cinfo, JHUFF_TBL *htbl, long freq[]);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcinit.c glmark2-2021.02/src/libjpeg-turbo/jcinit.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcinit.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcinit.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains initialization logic for the JPEG compressor.
  * This routine is in charge of selecting the modules to be executed and
@@ -60,7 +61,7 @@
 
   /* Need a full-image coefficient buffer in any multi-pass mode. */
   jinit_c_coef_controller(cinfo,
-		(boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
+                (boolean) (cinfo->num_scans > 1 || cinfo->optimize_coding));
   jinit_c_main_controller(cinfo, FALSE /* never need full buffer here */);
 
   jinit_marker_writer(cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmainct.c glmark2-2021.02/src/libjpeg-turbo/jcmainct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmainct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcmainct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcmainct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the main buffer controller for compression.
  * The main buffer lies between the pre-processor and the JPEG
@@ -15,50 +18,30 @@
 #include "jpeglib.h"
 
 
-/* Note: currently, there is no operating mode in which a full-image buffer
- * is needed at this step.  If there were, that mode could not be used with
- * "raw data" input, since this module is bypassed in that case.  However,
- * we've left the code here for possible use in special applications.
- */
-#undef FULL_MAIN_BUFFER_SUPPORTED
-
-
 /* Private buffer controller object */
 
 typedef struct {
   struct jpeg_c_main_controller pub; /* public fields */
 
-  JDIMENSION cur_iMCU_row;	/* number of current iMCU row */
-  JDIMENSION rowgroup_ctr;	/* counts row groups received in iMCU row */
-  boolean suspended;		/* remember if we suspended output */
-  J_BUF_MODE pass_mode;		/* current operating mode */
+  JDIMENSION cur_iMCU_row;      /* number of current iMCU row */
+  JDIMENSION rowgroup_ctr;      /* counts row groups received in iMCU row */
+  boolean suspended;            /* remember if we suspended output */
+  J_BUF_MODE pass_mode;         /* current operating mode */
 
   /* If using just a strip buffer, this points to the entire set of buffers
    * (we allocate one for each component).  In the full-image case, this
    * points to the currently accessible strips of the virtual arrays.
    */
   JSAMPARRAY buffer[MAX_COMPONENTS];
-
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-  /* If using full-image storage, this array holds pointers to virtual-array
-   * control blocks for each component.  Unused if not full-image storage.
-   */
-  jvirt_sarray_ptr whole_image[MAX_COMPONENTS];
-#endif
 } my_main_controller;
 
-typedef my_main_controller * my_main_ptr;
+typedef my_main_controller *my_main_ptr;
 
 
 /* Forward declarations */
 METHODDEF(void) process_data_simple_main
-	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
-	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-METHODDEF(void) process_data_buffer_main
-	JPP((j_compress_ptr cinfo, JSAMPARRAY input_buf,
-	     JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail));
-#endif
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+         JDIMENSION in_rows_avail);
 
 
 /*
@@ -74,32 +57,14 @@
   if (cinfo->raw_data_in)
     return;
 
-  main_ptr->cur_iMCU_row = 0;	/* initialize counters */
+  if (pass_mode != JBUF_PASS_THRU)
+    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
+
+  main_ptr->cur_iMCU_row = 0;   /* initialize counters */
   main_ptr->rowgroup_ctr = 0;
   main_ptr->suspended = FALSE;
-  main_ptr->pass_mode = pass_mode;	/* save mode for use by process_data */
-
-  switch (pass_mode) {
-  case JBUF_PASS_THRU:
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    if (main_ptr->whole_image[0] != NULL)
-      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-#endif
-    main_ptr->pub.process_data = process_data_simple_main;
-    break;
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-  case JBUF_SAVE_SOURCE:
-  case JBUF_CRANK_DEST:
-  case JBUF_SAVE_AND_PASS:
-    if (main_ptr->whole_image[0] == NULL)
-      ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-    main_ptr->pub.process_data = process_data_buffer_main;
-    break;
-#endif
-  default:
-    ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-    break;
-  }
+  main_ptr->pass_mode = pass_mode;      /* save mode for use by process_data */
+  main_ptr->pub.process_data = process_data_simple_main;
 }
 
 
@@ -111,8 +76,8 @@
 
 METHODDEF(void)
 process_data_simple_main (j_compress_ptr cinfo,
-			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			  JDIMENSION in_rows_avail)
+                          JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                          JDIMENSION in_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
 
@@ -120,9 +85,9 @@
     /* Read input data if we haven't filled the main buffer yet */
     if (main_ptr->rowgroup_ctr < DCTSIZE)
       (*cinfo->prep->pre_process_data) (cinfo,
-					input_buf, in_row_ctr, in_rows_avail,
-					main_ptr->buffer, &main_ptr->rowgroup_ctr,
-					(JDIMENSION) DCTSIZE);
+                                        input_buf, in_row_ctr, in_rows_avail,
+                                        main_ptr->buffer, &main_ptr->rowgroup_ctr,
+                                        (JDIMENSION) DCTSIZE);
 
     /* If we don't have a full iMCU row buffered, return to application for
      * more data.  Note that preprocessor will always pad to fill the iMCU row
@@ -140,8 +105,8 @@
        * think we were done.
        */
       if (! main_ptr->suspended) {
-	(*in_row_ctr)--;
-	main_ptr->suspended = TRUE;
+        (*in_row_ctr)--;
+        main_ptr->suspended = TRUE;
       }
       return;
     }
@@ -158,85 +123,6 @@
 }
 
 
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-
-/*
- * Process some data.
- * This routine handles all of the modes that use a full-size buffer.
- */
-
-METHODDEF(void)
-process_data_buffer_main (j_compress_ptr cinfo,
-			  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			  JDIMENSION in_rows_avail)
-{
-  my_main_ptr main = (my_main_ptr) cinfo->main;
-  int ci;
-  jpeg_component_info *compptr;
-  boolean writing = (main_ptr->pass_mode != JBUF_CRANK_DEST);
-
-  while (main_ptr->cur_iMCU_row < cinfo->total_iMCU_rows) {
-    /* Realign the virtual buffers if at the start of an iMCU row. */
-    if (main_ptr->rowgroup_ctr == 0) {
-      for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	   ci++, compptr++) {
-	main_ptr->buffer[ci] = (*cinfo->mem->access_virt_sarray)
-	  ((j_common_ptr) cinfo, main_ptr->whole_image[ci],
-	   main_ptr->cur_iMCU_row * (compptr->v_samp_factor * DCTSIZE),
-	   (JDIMENSION) (compptr->v_samp_factor * DCTSIZE), writing);
-      }
-      /* In a read pass, pretend we just read some source data. */
-      if (! writing) {
-	*in_row_ctr += cinfo->max_v_samp_factor * DCTSIZE;
-	main_ptr->rowgroup_ctr = DCTSIZE;
-      }
-    }
-
-    /* If a write pass, read input data until the current iMCU row is full. */
-    /* Note: preprocessor will pad if necessary to fill the last iMCU row. */
-    if (writing) {
-      (*cinfo->prep->pre_process_data) (cinfo,
-					input_buf, in_row_ctr, in_rows_avail,
-					main_ptr->buffer, &main_ptr->rowgroup_ctr,
-					(JDIMENSION) DCTSIZE);
-      /* Return to application if we need more data to fill the iMCU row. */
-      if (main_ptr->rowgroup_ctr < DCTSIZE)
-	return;
-    }
-
-    /* Emit data, unless this is a sink-only pass. */
-    if (main_ptr->pass_mode != JBUF_SAVE_SOURCE) {
-      if (! (*cinfo->coef->compress_data) (cinfo, main_ptr->buffer)) {
-	/* If compressor did not consume the whole row, then we must need to
-	 * suspend processing and return to the application.  In this situation
-	 * we pretend we didn't yet consume the last input row; otherwise, if
-	 * it happened to be the last row of the image, the application would
-	 * think we were done.
-	 */
-	if (! main_ptr->suspended) {
-	  (*in_row_ctr)--;
-	  main_ptr->suspended = TRUE;
-	}
-	return;
-      }
-      /* We did finish the row.  Undo our little suspension hack if a previous
-       * call suspended; then mark the main buffer empty.
-       */
-      if (main_ptr->suspended) {
-	(*in_row_ctr)++;
-	main_ptr->suspended = FALSE;
-      }
-    }
-
-    /* If get here, we are done with this iMCU row.  Mark buffer empty. */
-    main_ptr->rowgroup_ctr = 0;
-    main_ptr->cur_iMCU_row++;
-  }
-}
-
-#endif /* FULL_MAIN_BUFFER_SUPPORTED */
-
-
 /*
  * Initialize main buffer controller.
  */
@@ -250,7 +136,7 @@
 
   main_ptr = (my_main_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_main_controller));
+                                sizeof(my_main_controller));
   cinfo->main = (struct jpeg_c_main_controller *) main_ptr;
   main_ptr->pub.start_pass = start_pass_main;
 
@@ -262,32 +148,15 @@
    * may be of a different size.
    */
   if (need_full_buffer) {
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    /* Allocate a full-image virtual array for each component */
-    /* Note we pad the bottom to a multiple of the iMCU height */
-    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
-      main_ptr->whole_image[ci] = (*cinfo->mem->request_virt_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 compptr->width_in_blocks * DCTSIZE,
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor) * DCTSIZE,
-	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
-    }
-#else
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
-#endif
   } else {
-#ifdef FULL_MAIN_BUFFER_SUPPORTED
-    main_ptr->whole_image[0] = NULL; /* flag for no virtual arrays */
-#endif
     /* Allocate a strip buffer for each component */
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 compptr->width_in_blocks * DCTSIZE,
-	 (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         compptr->width_in_blocks * DCTSIZE,
+         (JDIMENSION) (compptr->v_samp_factor * DCTSIZE));
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmarker.c glmark2-2021.02/src/libjpeg-turbo/jcmarker.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmarker.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcmarker.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jcmarker.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
+ * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains routines to write JPEG datastream markers.
  */
@@ -15,29 +18,29 @@
 #include "jpegcomp.h"
 
 
-typedef enum {			/* JPEG marker codes */
+typedef enum {                  /* JPEG marker codes */
   M_SOF0  = 0xc0,
   M_SOF1  = 0xc1,
   M_SOF2  = 0xc2,
   M_SOF3  = 0xc3,
-  
+
   M_SOF5  = 0xc5,
   M_SOF6  = 0xc6,
   M_SOF7  = 0xc7,
-  
+
   M_JPG   = 0xc8,
   M_SOF9  = 0xc9,
   M_SOF10 = 0xca,
   M_SOF11 = 0xcb,
-  
+
   M_SOF13 = 0xcd,
   M_SOF14 = 0xce,
   M_SOF15 = 0xcf,
-  
+
   M_DHT   = 0xc4,
-  
+
   M_DAC   = 0xcc,
-  
+
   M_RST0  = 0xd0,
   M_RST1  = 0xd1,
   M_RST2  = 0xd2,
@@ -46,7 +49,7 @@
   M_RST5  = 0xd5,
   M_RST6  = 0xd6,
   M_RST7  = 0xd7,
-  
+
   M_SOI   = 0xd8,
   M_EOI   = 0xd9,
   M_SOS   = 0xda,
@@ -55,7 +58,7 @@
   M_DRI   = 0xdd,
   M_DHP   = 0xde,
   M_EXP   = 0xdf,
-  
+
   M_APP0  = 0xe0,
   M_APP1  = 0xe1,
   M_APP2  = 0xe2,
@@ -72,13 +75,13 @@
   M_APP13 = 0xed,
   M_APP14 = 0xee,
   M_APP15 = 0xef,
-  
+
   M_JPG0  = 0xf0,
   M_JPG13 = 0xfd,
   M_COM   = 0xfe,
-  
+
   M_TEM   = 0x01,
-  
+
   M_ERROR = 0x100
 } JPEG_MARKER;
 
@@ -91,7 +94,7 @@
   unsigned int last_restart_interval; /* last DRI value emitted; 0 after SOI */
 } my_marker_writer;
 
-typedef my_marker_writer * my_marker_ptr;
+typedef my_marker_writer *my_marker_ptr;
 
 
 /*
@@ -110,7 +113,7 @@
 emit_byte (j_compress_ptr cinfo, int val)
 /* Emit a byte */
 {
-  struct jpeg_destination_mgr * dest = cinfo->dest;
+  struct jpeg_destination_mgr *dest = cinfo->dest;
 
   *(dest->next_output_byte)++ = (JOCTET) val;
   if (--dest->free_in_buffer == 0) {
@@ -147,7 +150,7 @@
 /* Emit a DQT marker */
 /* Returns the precision used (0 = 8bits, 1 = 16bits) for baseline checking */
 {
-  JQUANT_TBL * qtbl = cinfo->quant_tbl_ptrs[index];
+  JQUANT_TBL *qtbl = cinfo->quant_tbl_ptrs[index];
   int prec;
   int i;
 
@@ -171,7 +174,7 @@
       /* The table entries must be emitted in zigzag order. */
       unsigned int qval = qtbl->quantval[jpeg_natural_order[i]];
       if (prec)
-	emit_byte(cinfo, (int) (qval >> 8));
+        emit_byte(cinfo, (int) (qval >> 8));
       emit_byte(cinfo, (int) (qval & 0xFF));
     }
 
@@ -186,35 +189,35 @@
 emit_dht (j_compress_ptr cinfo, int index, boolean is_ac)
 /* Emit a DHT marker */
 {
-  JHUFF_TBL * htbl;
+  JHUFF_TBL *htbl;
   int length, i;
-  
+
   if (is_ac) {
     htbl = cinfo->ac_huff_tbl_ptrs[index];
-    index += 0x10;		/* output index has AC bit set */
+    index += 0x10;              /* output index has AC bit set */
   } else {
     htbl = cinfo->dc_huff_tbl_ptrs[index];
   }
 
   if (htbl == NULL)
     ERREXIT1(cinfo, JERR_NO_HUFF_TABLE, index);
-  
+
   if (! htbl->sent_table) {
     emit_marker(cinfo, M_DHT);
-    
+
     length = 0;
     for (i = 1; i <= 16; i++)
       length += htbl->bits[i];
-    
+
     emit_2bytes(cinfo, length + 2 + 1 + 16);
     emit_byte(cinfo, index);
-    
+
     for (i = 1; i <= 16; i++)
       emit_byte(cinfo, htbl->bits[i]);
-    
+
     for (i = 0; i < length; i++)
       emit_byte(cinfo, htbl->huffval[i]);
-    
+
     htbl->sent_table = TRUE;
   }
 }
@@ -231,32 +234,38 @@
   char ac_in_use[NUM_ARITH_TBLS];
   int length, i;
   jpeg_component_info *compptr;
-  
+
   for (i = 0; i < NUM_ARITH_TBLS; i++)
     dc_in_use[i] = ac_in_use[i] = 0;
-  
+
   for (i = 0; i < cinfo->comps_in_scan; i++) {
     compptr = cinfo->cur_comp_info[i];
-    dc_in_use[compptr->dc_tbl_no] = 1;
-    ac_in_use[compptr->ac_tbl_no] = 1;
+    /* DC needs no table for refinement scan */
+    if (cinfo->Ss == 0 && cinfo->Ah == 0)
+      dc_in_use[compptr->dc_tbl_no] = 1;
+    /* AC needs no table when not present */
+    if (cinfo->Se)
+      ac_in_use[compptr->ac_tbl_no] = 1;
   }
-  
+
   length = 0;
   for (i = 0; i < NUM_ARITH_TBLS; i++)
     length += dc_in_use[i] + ac_in_use[i];
-  
-  emit_marker(cinfo, M_DAC);
-  
-  emit_2bytes(cinfo, length*2 + 2);
-  
-  for (i = 0; i < NUM_ARITH_TBLS; i++) {
-    if (dc_in_use[i]) {
-      emit_byte(cinfo, i);
-      emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
-    }
-    if (ac_in_use[i]) {
-      emit_byte(cinfo, i + 0x10);
-      emit_byte(cinfo, cinfo->arith_ac_K[i]);
+
+  if (length) {
+    emit_marker(cinfo, M_DAC);
+
+    emit_2bytes(cinfo, length*2 + 2);
+
+    for (i = 0; i < NUM_ARITH_TBLS; i++) {
+      if (dc_in_use[i]) {
+        emit_byte(cinfo, i);
+        emit_byte(cinfo, cinfo->arith_dc_L[i] + (cinfo->arith_dc_U[i]<<4));
+      }
+      if (ac_in_use[i]) {
+        emit_byte(cinfo, i + 0x10);
+        emit_byte(cinfo, cinfo->arith_ac_K[i]);
+      }
     }
   }
 #endif /* C_ARITH_CODING_SUPPORTED */
@@ -268,8 +277,8 @@
 /* Emit a DRI marker */
 {
   emit_marker(cinfo, M_DRI);
-  
-  emit_2bytes(cinfo, 4);	/* fixed length */
+
+  emit_2bytes(cinfo, 4);        /* fixed length */
 
   emit_2bytes(cinfo, (int) cinfo->restart_interval);
 }
@@ -281,9 +290,9 @@
 {
   int ci;
   jpeg_component_info *compptr;
-  
+
   emit_marker(cinfo, code);
-  
+
   emit_2bytes(cinfo, 3 * cinfo->num_components + 2 + 5 + 1); /* length */
 
   /* Make sure image isn't bigger than SOF field can handle */
@@ -312,32 +321,26 @@
 {
   int i, td, ta;
   jpeg_component_info *compptr;
-  
+
   emit_marker(cinfo, M_SOS);
-  
+
   emit_2bytes(cinfo, 2 * cinfo->comps_in_scan + 2 + 1 + 3); /* length */
-  
+
   emit_byte(cinfo, cinfo->comps_in_scan);
-  
+
   for (i = 0; i < cinfo->comps_in_scan; i++) {
     compptr = cinfo->cur_comp_info[i];
     emit_byte(cinfo, compptr->component_id);
-    td = compptr->dc_tbl_no;
-    ta = compptr->ac_tbl_no;
-    if (cinfo->progressive_mode) {
-      /* Progressive mode: only DC or only AC tables are used in one scan;
-       * furthermore, Huffman coding of DC refinement uses no table at all.
-       * We emit 0 for unused field(s); this is recommended by the P&M text
-       * but does not seem to be specified in the standard.
-       */
-      if (cinfo->Ss == 0) {
-	ta = 0;			/* DC scan */
-	if (cinfo->Ah != 0 && !cinfo->arith_code)
-	  td = 0;		/* no DC table either */
-      } else {
-	td = 0;			/* AC scan */
-      }
-    }
+
+    /* We emit 0 for unused field(s); this is recommended by the P&M text
+     * but does not seem to be specified in the standard.
+     */
+
+    /* DC needs no table for refinement scan */
+    td = cinfo->Ss == 0 && cinfo->Ah == 0 ? compptr->dc_tbl_no : 0;
+    /* AC needs no table when not present */
+    ta = cinfo->Se ? compptr->ac_tbl_no : 0;
+
     emit_byte(cinfo, (td << 4) + ta);
   }
 
@@ -352,22 +355,22 @@
 /* Emit a JFIF-compliant APP0 marker */
 {
   /*
-   * Length of APP0 block	(2 bytes)
-   * Block ID			(4 bytes - ASCII "JFIF")
-   * Zero byte			(1 byte to terminate the ID string)
-   * Version Major, Minor	(2 bytes - major first)
-   * Units			(1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
-   * Xdpu			(2 bytes - dots per unit horizontal)
-   * Ydpu			(2 bytes - dots per unit vertical)
-   * Thumbnail X size		(1 byte)
-   * Thumbnail Y size		(1 byte)
+   * Length of APP0 block       (2 bytes)
+   * Block ID                   (4 bytes - ASCII "JFIF")
+   * Zero byte                  (1 byte to terminate the ID string)
+   * Version Major, Minor       (2 bytes - major first)
+   * Units                      (1 byte - 0x00 = none, 0x01 = inch, 0x02 = cm)
+   * Xdpu                       (2 bytes - dots per unit horizontal)
+   * Ydpu                       (2 bytes - dots per unit vertical)
+   * Thumbnail X size           (1 byte)
+   * Thumbnail Y size           (1 byte)
    */
-  
+
   emit_marker(cinfo, M_APP0);
-  
+
   emit_2bytes(cinfo, 2 + 4 + 1 + 2 + 1 + 2 + 2 + 1 + 1); /* length */
 
-  emit_byte(cinfo, 0x4A);	/* Identifier: ASCII "JFIF" */
+  emit_byte(cinfo, 0x4A);       /* Identifier: ASCII "JFIF" */
   emit_byte(cinfo, 0x46);
   emit_byte(cinfo, 0x49);
   emit_byte(cinfo, 0x46);
@@ -377,7 +380,7 @@
   emit_byte(cinfo, cinfo->density_unit); /* Pixel size information */
   emit_2bytes(cinfo, (int) cinfo->X_density);
   emit_2bytes(cinfo, (int) cinfo->Y_density);
-  emit_byte(cinfo, 0);		/* No thumbnail image */
+  emit_byte(cinfo, 0);          /* No thumbnail image */
   emit_byte(cinfo, 0);
 }
 
@@ -387,12 +390,12 @@
 /* Emit an Adobe APP14 marker */
 {
   /*
-   * Length of APP14 block	(2 bytes)
-   * Block ID			(5 bytes - ASCII "Adobe")
-   * Version Number		(2 bytes - currently 100)
-   * Flags0			(2 bytes - currently 0)
-   * Flags1			(2 bytes - currently 0)
-   * Color transform		(1 byte)
+   * Length of APP14 block      (2 bytes)
+   * Block ID                   (5 bytes - ASCII "Adobe")
+   * Version Number             (2 bytes - currently 100)
+   * Flags0                     (2 bytes - currently 0)
+   * Flags1                     (2 bytes - currently 0)
+   * Color transform            (1 byte)
    *
    * Although Adobe TN 5116 mentions Version = 101, all the Adobe files
    * now in circulation seem to use Version = 100, so that's what we write.
@@ -401,28 +404,28 @@
    * YCbCr, 2 if it's YCCK, 0 otherwise.  Adobe's definition has to do with
    * whether the encoder performed a transformation, which is pretty useless.
    */
-  
+
   emit_marker(cinfo, M_APP14);
-  
+
   emit_2bytes(cinfo, 2 + 5 + 2 + 2 + 2 + 1); /* length */
 
-  emit_byte(cinfo, 0x41);	/* Identifier: ASCII "Adobe" */
+  emit_byte(cinfo, 0x41);       /* Identifier: ASCII "Adobe" */
   emit_byte(cinfo, 0x64);
   emit_byte(cinfo, 0x6F);
   emit_byte(cinfo, 0x62);
   emit_byte(cinfo, 0x65);
-  emit_2bytes(cinfo, 100);	/* Version */
-  emit_2bytes(cinfo, 0);	/* Flags0 */
-  emit_2bytes(cinfo, 0);	/* Flags1 */
+  emit_2bytes(cinfo, 100);      /* Version */
+  emit_2bytes(cinfo, 0);        /* Flags0 */
+  emit_2bytes(cinfo, 0);        /* Flags1 */
   switch (cinfo->jpeg_color_space) {
   case JCS_YCbCr:
-    emit_byte(cinfo, 1);	/* Color transform = 1 */
+    emit_byte(cinfo, 1);        /* Color transform = 1 */
     break;
   case JCS_YCCK:
-    emit_byte(cinfo, 2);	/* Color transform = 2 */
+    emit_byte(cinfo, 2);        /* Color transform = 2 */
     break;
   default:
-    emit_byte(cinfo, 0);	/* Color transform = 0 */
+    emit_byte(cinfo, 0);        /* Color transform = 0 */
     break;
   }
 }
@@ -440,12 +443,12 @@
 write_marker_header (j_compress_ptr cinfo, int marker, unsigned int datalen)
 /* Emit an arbitrary marker header */
 {
-  if (datalen > (unsigned int) 65533)		/* safety check */
+  if (datalen > (unsigned int) 65533)           /* safety check */
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
   emit_marker(cinfo, (JPEG_MARKER) marker);
 
-  emit_2bytes(cinfo, (int) (datalen + 2));	/* total length */
+  emit_2bytes(cinfo, (int) (datalen + 2));      /* total length */
 }
 
 METHODDEF(void)
@@ -472,12 +475,12 @@
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
-  emit_marker(cinfo, M_SOI);	/* first the SOI */
+  emit_marker(cinfo, M_SOI);    /* first the SOI */
 
   /* SOI is defined to reset restart interval to 0 */
   marker->last_restart_interval = 0;
 
-  if (cinfo->write_JFIF_header)	/* next an optional JFIF APP0 */
+  if (cinfo->write_JFIF_header) /* next an optional JFIF APP0 */
     emit_jfif_app0(cinfo);
   if (cinfo->write_Adobe_marker) /* next an optional Adobe APP14 */
     emit_adobe_app14(cinfo);
@@ -498,7 +501,7 @@
   int ci, prec;
   boolean is_baseline;
   jpeg_component_info *compptr;
-  
+
   /* Emit DQT for each quantization table.
    * Note that emit_dqt() suppresses any duplicate tables.
    */
@@ -518,9 +521,9 @@
   } else {
     is_baseline = TRUE;
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       if (compptr->dc_tbl_no > 1 || compptr->ac_tbl_no > 1)
-	is_baseline = FALSE;
+        is_baseline = FALSE;
     }
     if (prec && is_baseline) {
       is_baseline = FALSE;
@@ -531,14 +534,17 @@
 
   /* Emit the proper SOF marker */
   if (cinfo->arith_code) {
-    emit_sof(cinfo, M_SOF9);	/* SOF code for arithmetic coding */
+    if (cinfo->progressive_mode)
+      emit_sof(cinfo, M_SOF10); /* SOF code for progressive arithmetic */
+    else
+      emit_sof(cinfo, M_SOF9);  /* SOF code for sequential arithmetic */
   } else {
     if (cinfo->progressive_mode)
-      emit_sof(cinfo, M_SOF2);	/* SOF code for progressive Huffman */
+      emit_sof(cinfo, M_SOF2);  /* SOF code for progressive Huffman */
     else if (is_baseline)
-      emit_sof(cinfo, M_SOF0);	/* SOF code for baseline implementation */
+      emit_sof(cinfo, M_SOF0);  /* SOF code for baseline implementation */
     else
-      emit_sof(cinfo, M_SOF1);	/* SOF code for non-baseline Huffman file */
+      emit_sof(cinfo, M_SOF1);  /* SOF code for non-baseline Huffman file */
   }
 }
 
@@ -568,19 +574,12 @@
      */
     for (i = 0; i < cinfo->comps_in_scan; i++) {
       compptr = cinfo->cur_comp_info[i];
-      if (cinfo->progressive_mode) {
-	/* Progressive mode: only DC or only AC tables are used in one scan */
-	if (cinfo->Ss == 0) {
-	  if (cinfo->Ah == 0)	/* DC needs no table for refinement scan */
-	    emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
-	} else {
-	  emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
-	}
-      } else {
-	/* Sequential mode: need both DC and AC tables */
-	emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
-	emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
-      }
+      /* DC needs no table for refinement scan */
+      if (cinfo->Ss == 0 && cinfo->Ah == 0)
+        emit_dht(cinfo, compptr->dc_tbl_no, FALSE);
+      /* AC needs no table when not present */
+      if (cinfo->Se)
+        emit_dht(cinfo, compptr->ac_tbl_no, TRUE);
     }
   }
 
@@ -629,9 +628,9 @@
   if (! cinfo->arith_code) {
     for (i = 0; i < NUM_HUFF_TBLS; i++) {
       if (cinfo->dc_huff_tbl_ptrs[i] != NULL)
-	emit_dht(cinfo, i, FALSE);
+        emit_dht(cinfo, i, FALSE);
       if (cinfo->ac_huff_tbl_ptrs[i] != NULL)
-	emit_dht(cinfo, i, TRUE);
+        emit_dht(cinfo, i, TRUE);
     }
   }
 
@@ -651,7 +650,7 @@
   /* Create the subobject */
   marker = (my_marker_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_marker_writer));
+                                sizeof(my_marker_writer));
   cinfo->marker = (struct jpeg_marker_writer *) marker;
   /* Initialize method pointers */
   marker->pub.write_file_header = write_file_header;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmaster.c glmark2-2021.02/src/libjpeg-turbo/jcmaster.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcmaster.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcmaster.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,17 @@
 /*
  * jcmaster.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * Modified 2003-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains master control logic for the JPEG compressor.
  * These routines are concerned with parameter validation, initial setup,
- * and inter-pass control (determining the number of passes and the work 
+ * and inter-pass control (determining the number of passes and the work
  * to be done in each pass).
  */
 
@@ -22,23 +24,23 @@
 /* Private state */
 
 typedef enum {
-	main_pass,		/* input data, also do first output step */
-	huff_opt_pass,		/* Huffman code optimization pass */
-	output_pass		/* data output pass */
+        main_pass,              /* input data, also do first output step */
+        huff_opt_pass,          /* Huffman code optimization pass */
+        output_pass             /* data output pass */
 } c_pass_type;
 
 typedef struct {
-  struct jpeg_comp_master pub;	/* public fields */
+  struct jpeg_comp_master pub;  /* public fields */
 
-  c_pass_type pass_type;	/* the type of the current pass */
+  c_pass_type pass_type;        /* the type of the current pass */
 
-  int pass_number;		/* # of passes completed */
-  int total_passes;		/* total # of passes needed */
+  int pass_number;              /* # of passes completed */
+  int total_passes;             /* total # of passes needed */
 
-  int scan_number;		/* current index in scan_info[] */
+  int scan_number;              /* current index in scan_info[] */
 } my_comp_master;
 
-typedef my_comp_master * my_master_ptr;
+typedef my_comp_master *my_master_ptr;
 
 
 /*
@@ -104,7 +106,7 @@
   /* Check that number of components won't exceed internal array sizes */
   if (cinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
 
   /* Compute maximum sampling factors; check factor validity */
   cinfo->max_h_samp_factor = 1;
@@ -112,12 +114,12 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
-	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+        compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
       ERREXIT(cinfo, JERR_BAD_SAMPLING);
     cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
-				   compptr->h_samp_factor);
+                                   compptr->h_samp_factor);
     cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
-				   compptr->v_samp_factor);
+                                   compptr->v_samp_factor);
   }
 
   /* Compute dimensions of components */
@@ -134,17 +136,17 @@
     /* Size in DCT blocks */
     compptr->width_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width * (long) compptr->h_samp_factor,
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->height_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height * (long) compptr->v_samp_factor,
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
     /* Size in samples */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width * (long) compptr->h_samp_factor,
-		    (long) cinfo->max_h_samp_factor);
+                    (long) cinfo->max_h_samp_factor);
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height * (long) compptr->v_samp_factor,
-		    (long) cinfo->max_v_samp_factor);
+                    (long) cinfo->max_v_samp_factor);
     /* Mark component needed (this flag isn't actually used for compression) */
     compptr->component_needed = TRUE;
   }
@@ -154,7 +156,7 @@
    */
   cinfo->total_iMCU_rows = (JDIMENSION)
     jdiv_round_up((long) cinfo->_jpeg_height,
-		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+                  (long) (cinfo->max_v_samp_factor*DCTSIZE));
 }
 
 
@@ -166,12 +168,12 @@
  * determine whether it uses progressive JPEG, and set cinfo->progressive_mode.
  */
 {
-  const jpeg_scan_info * scanptr;
+  const jpeg_scan_info *scanptr;
   int scanno, ncomps, ci, coefi, thisi;
   int Ss, Se, Ah, Al;
   boolean component_sent[MAX_COMPONENTS];
 #ifdef C_PROGRESSIVE_SUPPORTED
-  int * last_bitpos_ptr;
+  int *last_bitpos_ptr;
   int last_bitpos[MAX_COMPONENTS][DCTSIZE2];
   /* -1 until that coefficient has been seen; then last Al for it */
 #endif
@@ -187,15 +189,15 @@
 #ifdef C_PROGRESSIVE_SUPPORTED
     cinfo->progressive_mode = TRUE;
     last_bitpos_ptr = & last_bitpos[0][0];
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       for (coefi = 0; coefi < DCTSIZE2; coefi++)
-	*last_bitpos_ptr++ = -1;
+        *last_bitpos_ptr++ = -1;
 #else
     ERREXIT(cinfo, JERR_NOT_COMPILED);
 #endif
   } else {
     cinfo->progressive_mode = FALSE;
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       component_sent[ci] = FALSE;
   }
 
@@ -207,10 +209,10 @@
     for (ci = 0; ci < ncomps; ci++) {
       thisi = scanptr->component_index[ci];
       if (thisi < 0 || thisi >= cinfo->num_components)
-	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
       /* Components must appear in SOF order within each scan */
       if (ci > 0 && thisi <= scanptr->component_index[ci-1])
-	ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
     }
     /* Validate progression parameters */
     Ss = scanptr->Ss;
@@ -232,43 +234,43 @@
 #define MAX_AH_AL 13
 #endif
       if (Ss < 0 || Ss >= DCTSIZE2 || Se < Ss || Se >= DCTSIZE2 ||
-	  Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
-	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          Ah < 0 || Ah > MAX_AH_AL || Al < 0 || Al > MAX_AH_AL)
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       if (Ss == 0) {
-	if (Se != 0)		/* DC and AC together not OK */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        if (Se != 0)            /* DC and AC together not OK */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       } else {
-	if (ncomps != 1)	/* AC scans must be for only one component */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        if (ncomps != 1)        /* AC scans must be for only one component */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       }
       for (ci = 0; ci < ncomps; ci++) {
-	last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
-	if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
-	  ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	for (coefi = Ss; coefi <= Se; coefi++) {
-	  if (last_bitpos_ptr[coefi] < 0) {
-	    /* first scan of this coefficient */
-	    if (Ah != 0)
-	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	  } else {
-	    /* not first scan */
-	    if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
-	      ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
-	  }
-	  last_bitpos_ptr[coefi] = Al;
-	}
+        last_bitpos_ptr = & last_bitpos[scanptr->component_index[ci]][0];
+        if (Ss != 0 && last_bitpos_ptr[0] < 0) /* AC without prior DC scan */
+          ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        for (coefi = Ss; coefi <= Se; coefi++) {
+          if (last_bitpos_ptr[coefi] < 0) {
+            /* first scan of this coefficient */
+            if (Ah != 0)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          } else {
+            /* not first scan */
+            if (Ah != last_bitpos_ptr[coefi] || Al != Ah-1)
+              ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+          }
+          last_bitpos_ptr[coefi] = Al;
+        }
       }
 #endif
     } else {
       /* For sequential JPEG, all progression parameters must be these: */
       if (Ss != 0 || Se != DCTSIZE2-1 || Ah != 0 || Al != 0)
-	ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
+        ERREXIT1(cinfo, JERR_BAD_PROG_SCRIPT, scanno);
       /* Make sure components are not sent twice */
       for (ci = 0; ci < ncomps; ci++) {
-	thisi = scanptr->component_index[ci];
-	if (component_sent[thisi])
-	  ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
-	component_sent[thisi] = TRUE;
+        thisi = scanptr->component_index[ci];
+        if (component_sent[thisi])
+          ERREXIT1(cinfo, JERR_BAD_SCAN_SCRIPT, scanno);
+        component_sent[thisi] = TRUE;
       }
     }
   }
@@ -283,13 +285,13 @@
      */
     for (ci = 0; ci < cinfo->num_components; ci++) {
       if (last_bitpos[ci][0] < 0)
-	ERREXIT(cinfo, JERR_MISSING_DATA);
+        ERREXIT(cinfo, JERR_MISSING_DATA);
     }
 #endif
   } else {
     for (ci = 0; ci < cinfo->num_components; ci++) {
       if (! component_sent[ci])
-	ERREXIT(cinfo, JERR_MISSING_DATA);
+        ERREXIT(cinfo, JERR_MISSING_DATA);
     }
   }
 }
@@ -307,12 +309,12 @@
   if (cinfo->scan_info != NULL) {
     /* Prepare for current scan --- the script is already validated */
     my_master_ptr master = (my_master_ptr) cinfo->master;
-    const jpeg_scan_info * scanptr = cinfo->scan_info + master->scan_number;
+    const jpeg_scan_info *scanptr = cinfo->scan_info + master->scan_number;
 
     cinfo->comps_in_scan = scanptr->comps_in_scan;
     for (ci = 0; ci < scanptr->comps_in_scan; ci++) {
       cinfo->cur_comp_info[ci] =
-	&cinfo->comp_info[scanptr->component_index[ci]];
+        &cinfo->comp_info[scanptr->component_index[ci]];
     }
     cinfo->Ss = scanptr->Ss;
     cinfo->Se = scanptr->Se;
@@ -325,7 +327,7 @@
     /* Prepare for single sequential-JPEG scan containing all components */
     if (cinfo->num_components > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	       MAX_COMPS_IN_SCAN);
+               MAX_COMPS_IN_SCAN);
     cinfo->comps_in_scan = cinfo->num_components;
     for (ci = 0; ci < cinfo->num_components; ci++) {
       cinfo->cur_comp_info[ci] = &cinfo->comp_info[ci];
@@ -345,16 +347,16 @@
 {
   int ci, mcublks, tmp;
   jpeg_component_info *compptr;
-  
+
   if (cinfo->comps_in_scan == 1) {
-    
+
     /* Noninterleaved (single-component) scan */
     compptr = cinfo->cur_comp_info[0];
-    
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = compptr->width_in_blocks;
     cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
-    
+
     /* For noninterleaved scan, always one block per MCU */
     compptr->MCU_width = 1;
     compptr->MCU_height = 1;
@@ -367,28 +369,28 @@
     tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
     if (tmp == 0) tmp = compptr->v_samp_factor;
     compptr->last_row_height = tmp;
-    
+
     /* Prepare array describing MCU composition */
     cinfo->blocks_in_MCU = 1;
     cinfo->MCU_membership[0] = 0;
-    
+
   } else {
-    
+
     /* Interleaved (multi-component) scan */
     if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
-	       MAX_COMPS_IN_SCAN);
-    
+               MAX_COMPS_IN_SCAN);
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_width,
-		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor*DCTSIZE));
     cinfo->MCU_rows_in_scan = (JDIMENSION)
       jdiv_round_up((long) cinfo->_jpeg_height,
-		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
-    
+                    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
     cinfo->blocks_in_MCU = 0;
-    
+
     for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
       compptr = cinfo->cur_comp_info[ci];
       /* Sampling factors give # of blocks of component in each MCU */
@@ -406,12 +408,12 @@
       /* Prepare array describing MCU composition */
       mcublks = compptr->MCU_blocks;
       if (cinfo->blocks_in_MCU + mcublks > C_MAX_BLOCKS_IN_MCU)
-	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
       while (mcublks-- > 0) {
-	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+        cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
       }
     }
-    
+
   }
 
   /* Convert restart specified in rows to actual MCU count. */
@@ -451,8 +453,8 @@
     (*cinfo->fdct->start_pass) (cinfo);
     (*cinfo->entropy->start_pass) (cinfo, cinfo->optimize_coding);
     (*cinfo->coef->start_pass) (cinfo,
-				(master->total_passes > 1 ?
-				 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+                                (master->total_passes > 1 ?
+                                 JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
     (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
     if (cinfo->optimize_coding) {
       /* No immediate data output; postpone writing frame/scan headers */
@@ -580,7 +582,7 @@
 
   master = (my_master_ptr)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(my_comp_master));
+                                  sizeof(my_comp_master));
   cinfo->master = (struct jpeg_comp_master *) master;
   master->pub.prepare_for_pass = prepare_for_pass;
   master->pub.pass_startup = pass_startup;
@@ -601,7 +603,7 @@
     cinfo->num_scans = 1;
   }
 
-  if (cinfo->progressive_mode)	/*  TEMPORARY HACK ??? */
+  if (cinfo->progressive_mode && !cinfo->arith_code)  /*  TEMPORARY HACK ??? */
     cinfo->optimize_coding = TRUE; /* assume default tables no good for progressive mode */
 
   /* Initialize my private state */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcomapi.c glmark2-2021.02/src/libjpeg-turbo/jcomapi.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcomapi.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcomapi.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcomapi.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface routines that are used for both
  * compression and decompression.
@@ -72,8 +75,8 @@
   /* NB: mem pointer is NULL if memory mgr failed to initialize. */
   if (cinfo->mem != NULL)
     (*cinfo->mem->self_destruct) (cinfo);
-  cinfo->mem = NULL;		/* be safe if jpeg_destroy is called twice */
-  cinfo->global_state = 0;	/* mark it destroyed */
+  cinfo->mem = NULL;            /* be safe if jpeg_destroy is called twice */
+  cinfo->global_state = 0;      /* mark it destroyed */
 }
 
 
@@ -88,8 +91,8 @@
   JQUANT_TBL *tbl;
 
   tbl = (JQUANT_TBL *)
-    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JQUANT_TBL));
-  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JQUANT_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
   return tbl;
 }
 
@@ -100,7 +103,7 @@
   JHUFF_TBL *tbl;
 
   tbl = (JHUFF_TBL *)
-    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, SIZEOF(JHUFF_TBL));
-  tbl->sent_table = FALSE;	/* make sure this is false in any new table */
+    (*cinfo->mem->alloc_small) (cinfo, JPOOL_PERMANENT, sizeof(JHUFF_TBL));
+  tbl->sent_table = FALSE;      /* make sure this is false in any new table */
   return tbl;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.h glmark2-2021.02/src/libjpeg-turbo/jconfig.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jconfig.h	2021-04-25 01:47:22.000000000 +0800
@@ -5,16 +5,27 @@
 #define JPEG_LIB_VERSION 62
 
 /* libjpeg-turbo version */
-#define LIBJPEG_TURBO_VERSION 1.2.0
+#define LIBJPEG_TURBO_VERSION 1.4.2
 
 /* Support arithmetic encoding */
-#define C_ARITH_CODING_SUPPORTED 1
+/* #define C_ARITH_CODING_SUPPORTED 1 */
 
 /* Support arithmetic decoding */
-#define D_ARITH_CODING_SUPPORTED 1
+/* #define D_ARITH_CODING_SUPPORTED 1 */
 
-/* Compiler supports function prototypes. */
-#define HAVE_PROTOTYPES 1
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8      /* use 8 or 12 */
+
+/* Define to 1 if you have the <locale.h> header file. */
+#define HAVE_LOCALE_H 1
 
 /* Define to 1 if you have the <stddef.h> header file. */
 #define HAVE_STDDEF_H 1
@@ -22,25 +33,27 @@
 /* Define to 1 if you have the <stdlib.h> header file. */
 #define HAVE_STDLIB_H 1
 
-/* Compiler supports 'unsigned char'. */
+/* Define to 1 if the system has the type `unsigned char'. */
 #define HAVE_UNSIGNED_CHAR 1
 
-/* Compiler supports 'unsigned short'. */
+/* Define to 1 if the system has the type `unsigned short'. */
 #define HAVE_UNSIGNED_SHORT 1
 
-/* Compiler does not support pointers to unspecified structures. */
+/* Compiler does not support pointers to undefined structures. */
 /* #undef INCOMPLETE_TYPES_BROKEN */
 
-/* Compiler has <strings.h> rather than standard <string.h>. */
-/* #undef NEED_BSD_STRINGS */
+/* Support in-memory source/destination managers */
+/* #undef MEM_SRCDST_SUPPORTED */
 
-/* Linker requires that global names be unique in first 15 characters. */
-/* #undef NEED_SHORT_EXTERNAL_NAMES */
+/* Define if you have BSD-like bzero and bcopy in <strings.h> rather than
+   memset/memcpy in <string.h>. */
+/* #undef NEED_BSD_STRINGS */
 
-/* Need to include <sys/types.h> in order to obtain size_t. */
-#define NEED_SYS_TYPES_H 1
+/* Define if you need to include <sys/types.h> to get size_t. */
+/* #undef NEED_SYS_TYPES_H 1 */
 
-/* Broken compiler shifts signed values as an unsigned shift. */
+/* Define if your (broken) compiler shifts signed values as if they were
+   unsigned. */
 /* #undef RIGHT_SHIFT_IS_UNSIGNED */
 
 /* Use accelerated SIMD routines. */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.h.in glmark2-2021.02/src/libjpeg-turbo/jconfig.h.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.h.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jconfig.h.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,70 @@
+/* Version ID for the JPEG library.
+ * Might be useful for tests like "#if JPEG_LIB_VERSION >= 60".
+ */
+#define JPEG_LIB_VERSION  62	/* Version 6b */
+
+/* libjpeg-turbo version */
+#define LIBJPEG_TURBO_VERSION 0
+
+/* Support arithmetic encoding */
+#undef C_ARITH_CODING_SUPPORTED
+
+/* Support arithmetic decoding */
+#undef D_ARITH_CODING_SUPPORTED
+
+/*
+ * Define BITS_IN_JSAMPLE as either
+ *   8   for 8-bit sample values (the usual setting)
+ *   12  for 12-bit sample values
+ * Only 8 and 12 are legal data precisions for lossy JPEG according to the
+ * JPEG standard, and the IJG code does not support anything else!
+ * We do not support run-time selection of data precision, sorry.
+ */
+
+#define BITS_IN_JSAMPLE  8      /* use 8 or 12 */
+
+/* Define to 1 if you have the <locale.h> header file. */
+#undef HAVE_LOCALE_H
+
+/* Define to 1 if you have the <stddef.h> header file. */
+#undef HAVE_STDDEF_H
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#undef HAVE_STDLIB_H
+
+/* Define to 1 if the system has the type `unsigned char'. */
+#undef HAVE_UNSIGNED_CHAR
+
+/* Define to 1 if the system has the type `unsigned short'. */
+#undef HAVE_UNSIGNED_SHORT
+
+/* Compiler does not support pointers to undefined structures. */
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Support in-memory source/destination managers */
+#undef MEM_SRCDST_SUPPORTED
+
+/* Define if you have BSD-like bzero and bcopy in <strings.h> rather than
+   memset/memcpy in <string.h>. */
+#undef NEED_BSD_STRINGS
+
+/* Define if you need to include <sys/types.h> to get size_t. */
+#undef NEED_SYS_TYPES_H
+
+/* Define if your (broken) compiler shifts signed values as if they were
+   unsigned. */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+/* Use accelerated SIMD routines. */
+#undef WITH_SIMD
+
+/* Define to 1 if type `char' is unsigned and you are not using gcc.  */
+#ifndef __CHAR_UNSIGNED__
+# undef __CHAR_UNSIGNED__
+#endif
+
+/* Define to empty if `const' does not conform to ANSI C. */
+#undef const
+
+/* Define to `unsigned int' if <sys/types.h> does not define. */
+#undef size_t
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfigint.h glmark2-2021.02/src/libjpeg-turbo/jconfigint.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfigint.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jconfigint.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,27 @@
+/* jconfigint.h.  Generated from jconfigint.h.in by configure.  */
+/* libjpeg-turbo build number */
+#define BUILD ""
+
+/* How to obtain function inlining. */
+#ifndef INLINE
+#if defined(__GNUC__)
+#define INLINE inline __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#define INLINE __forceinline
+#else
+#define INLINE
+#endif
+#endif
+
+/* Define to the full name of this package. */
+#define PACKAGE_NAME "libjpeg-turbo"
+
+/* Version number of package */
+#define VERSION "1.4.90"
+
+/* The size of `size_t', as computed by sizeof. */
+#if __WORDSIZE==64 || defined(_WIN64)
+#define SIZEOF_SIZE_T 8
+#else
+#define SIZEOF_SIZE_T 4
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfigint.h.in glmark2-2021.02/src/libjpeg-turbo/jconfigint.h.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfigint.h.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jconfigint.h.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,14 @@
+/* libjpeg-turbo build number */
+#undef BUILD
+
+/* How to obtain function inlining. */
+#undef INLINE
+
+/* Define to the full name of this package. */
+#undef PACKAGE_NAME
+
+/* Version number of package */
+#undef VERSION
+
+/* The size of `size_t', as computed by sizeof. */
+#undef SIZEOF_SIZE_T
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.txt glmark2-2021.02/src/libjpeg-turbo/jconfig.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jconfig.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jconfig.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,143 @@
+/*
+ * jconfig.txt
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1994, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file documents the configuration options that are required to
+ * customize the JPEG software for a particular system.
+ *
+ * The actual configuration options for a particular installation are stored
+ * in jconfig.h.  On many machines, jconfig.h can be generated automatically
+ * or copied from one of the "canned" jconfig files that we supply.  But if
+ * you need to generate a jconfig.h file by hand, this file tells you how.
+ *
+ * DO NOT EDIT THIS FILE --- IT WON'T ACCOMPLISH ANYTHING.
+ * EDIT A COPY NAMED JCONFIG.H.
+ */
+
+
+/*
+ * These symbols indicate the properties of your machine or compiler.
+ * #define the symbol if yes, #undef it if no.
+ */
+
+/* Does your compiler support the declaration "unsigned char" ?
+ * How about "unsigned short" ?
+ */
+#define HAVE_UNSIGNED_CHAR
+#define HAVE_UNSIGNED_SHORT
+
+/* Define "void" as "char" if your compiler doesn't know about type void.
+ * NOTE: be sure to define void such that "void *" represents the most general
+ * pointer type, e.g., that returned by malloc().
+ */
+/* #define void char */
+
+/* Define "const" as empty if your compiler doesn't know the "const" keyword.
+ */
+/* #define const */
+
+/* Define this if an ordinary "char" type is unsigned.
+ * If you're not sure, leaving it undefined will work at some cost in speed.
+ * If you defined HAVE_UNSIGNED_CHAR then the speed difference is minimal.
+ */
+#undef __CHAR_UNSIGNED__
+
+/* Define this if your system has an ANSI-conforming <stddef.h> file.
+ */
+#define HAVE_STDDEF_H
+
+/* Define this if your system has an ANSI-conforming <stdlib.h> file.
+ */
+#define HAVE_STDLIB_H
+
+/* Define this if your system does not have an ANSI/SysV <string.h>,
+ * but does have a BSD-style <strings.h>.
+ */
+#undef NEED_BSD_STRINGS
+
+/* Define this if your system does not provide typedef size_t in any of the
+ * ANSI-standard places (stddef.h, stdlib.h, or stdio.h), but places it in
+ * <sys/types.h> instead.
+ */
+#undef NEED_SYS_TYPES_H
+
+/* Although a real ANSI C compiler can deal perfectly well with pointers to
+ * unspecified structures (see "incomplete types" in the spec), a few pre-ANSI
+ * and pseudo-ANSI compilers get confused.  To keep one of these bozos happy,
+ * define INCOMPLETE_TYPES_BROKEN.  This is not recommended unless you
+ * actually get "missing structure definition" warnings or errors while
+ * compiling the JPEG code.
+ */
+#undef INCOMPLETE_TYPES_BROKEN
+
+/* Define "boolean" as unsigned char, not int, on Windows systems.
+ */
+#ifdef _WIN32
+#ifndef __RPCNDR_H__            /* don't conflict if rpcndr.h already read */
+typedef unsigned char boolean;
+#endif
+#define HAVE_BOOLEAN            /* prevent jmorecfg.h from redefining it */
+#endif
+
+
+/*
+ * The following options affect code selection within the JPEG library,
+ * but they don't need to be visible to applications using the library.
+ * To minimize application namespace pollution, the symbols won't be
+ * defined unless JPEG_INTERNALS has been defined.
+ */
+
+#ifdef JPEG_INTERNALS
+
+/* Define this if your compiler implements ">>" on signed values as a logical
+ * (unsigned) shift; leave it undefined if ">>" is a signed (arithmetic) shift,
+ * which is the normal and rational definition.
+ */
+#undef RIGHT_SHIFT_IS_UNSIGNED
+
+
+#endif /* JPEG_INTERNALS */
+
+
+/*
+ * The remaining options do not affect the JPEG library proper,
+ * but only the sample applications cjpeg/djpeg (see cjpeg.c, djpeg.c).
+ * Other applications can ignore these.
+ */
+
+#ifdef JPEG_CJPEG_DJPEG
+
+/* These defines indicate which image (non-JPEG) file formats are allowed. */
+
+#define BMP_SUPPORTED           /* BMP image file format */
+#define GIF_SUPPORTED           /* GIF image file format */
+#define PPM_SUPPORTED           /* PBMPLUS PPM/PGM image file format */
+#undef RLE_SUPPORTED            /* Utah RLE image file format */
+#define TARGA_SUPPORTED         /* Targa image file format */
+
+/* Define this if you want to name both input and output files on the command
+ * line, rather than using stdout and optionally stdin.  You MUST do this if
+ * your system can't cope with binary I/O to stdin/stdout.  See comments at
+ * head of cjpeg.c or djpeg.c.
+ */
+#undef TWO_FILE_COMMANDLINE
+
+/* By default, we open image files with fopen(...,"rb") or fopen(...,"wb").
+ * This is necessary on systems that distinguish text files from binary files,
+ * and is harmless on most systems that don't.  If you have one of the rare
+ * systems that complains about the "b" spec, define this symbol.
+ */
+#undef DONT_USE_B_MODE
+
+/* Define this if you want percent-done progress reports from cjpeg/djpeg.
+ */
+#undef PROGRESS_REPORT
+
+
+#endif /* JPEG_CJPEG_DJPEG */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcparam.c glmark2-2021.02/src/libjpeg-turbo/jcparam.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcparam.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcparam.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,13 @@
 /*
  * jcparam.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
  * Modified 2003-2008 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains optional default-setting code for the JPEG compressor.
  * Applications do not have to use this file, but those that don't use it
@@ -15,6 +17,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
+#include "jstdhuff.c"
 
 
 /*
@@ -23,15 +26,15 @@
 
 GLOBAL(void)
 jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
-		      const unsigned int *basic_table,
-		      int scale_factor, boolean force_baseline)
+                      const unsigned int *basic_table,
+                      int scale_factor, boolean force_baseline)
 /* Define a quantization table equal to the basic_table times
  * a scale factor (given as a percentage).
  * If force_baseline is TRUE, the computed quantization table entries
  * are limited to 1..255 for JPEG baseline compatibility.
  */
 {
-  JQUANT_TBL ** qtblptr;
+  JQUANT_TBL **qtblptr;
   int i;
   long temp;
 
@@ -53,7 +56,7 @@
     if (temp <= 0L) temp = 1L;
     if (temp > 32767L) temp = 32767L; /* max quantizer needed for 12 bits */
     if (force_baseline && temp > 255L)
-      temp = 255L;		/* limit to baseline range if requested */
+      temp = 255L;              /* limit to baseline range if requested */
     (*qtblptr)->quantval[i] = (UINT16) temp;
   }
 
@@ -98,16 +101,16 @@
 {
   /* Set up two quantization tables using the specified scaling */
   jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
-		       cinfo->q_scale_factor[0], force_baseline);
+                       cinfo->q_scale_factor[0], force_baseline);
   jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
-		       cinfo->q_scale_factor[1], force_baseline);
+                       cinfo->q_scale_factor[1], force_baseline);
 }
 #endif
 
 
 GLOBAL(void)
 jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
-			 boolean force_baseline)
+                         boolean force_baseline)
 /* Set or change the 'quality' (quantization) setting, using default tables
  * and a straight percentage-scaling quality scale.  In most cases it's better
  * to use jpeg_set_quality (below); this entry point is provided for
@@ -116,9 +119,9 @@
 {
   /* Set up two quantization tables using the specified scaling */
   jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
-		       scale_factor, force_baseline);
+                       scale_factor, force_baseline);
   jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
-		       scale_factor, force_baseline);
+                       scale_factor, force_baseline);
 }
 
 
@@ -165,116 +168,6 @@
 
 
 /*
- * Huffman table setup routines
- */
-
-LOCAL(void)
-add_huff_table (j_compress_ptr cinfo,
-		JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
-/* Define a Huffman table */
-{
-  int nsymbols, len;
-
-  if (*htblptr == NULL)
-    *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
-
-  /* Copy the number-of-symbols-of-each-code-length counts */
-  MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
-
-  /* Validate the counts.  We do this here mainly so we can copy the right
-   * number of symbols from the val[] array, without risking marching off
-   * the end of memory.  jchuff.c will do a more thorough test later.
-   */
-  nsymbols = 0;
-  for (len = 1; len <= 16; len++)
-    nsymbols += bits[len];
-  if (nsymbols < 1 || nsymbols > 256)
-    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
-
-  MEMCOPY((*htblptr)->huffval, val, nsymbols * SIZEOF(UINT8));
-
-  /* Initialize sent_table FALSE so table will be written to JPEG file. */
-  (*htblptr)->sent_table = FALSE;
-}
-
-
-LOCAL(void)
-std_huff_tables (j_compress_ptr cinfo)
-/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
-/* IMPORTANT: these are only valid for 8-bit data precision! */
-{
-  static const UINT8 bits_dc_luminance[17] =
-    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_luminance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-  
-  static const UINT8 bits_dc_chrominance[17] =
-    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
-  static const UINT8 val_dc_chrominance[] =
-    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
-  
-  static const UINT8 bits_ac_luminance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
-  static const UINT8 val_ac_luminance[] =
-    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
-      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
-      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
-      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
-      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
-      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
-      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
-      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
-      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
-      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
-      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
-      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
-      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
-      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
-      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
-      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
-      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
-      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
-      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
-      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-  
-  static const UINT8 bits_ac_chrominance[17] =
-    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
-  static const UINT8 val_ac_chrominance[] =
-    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
-      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
-      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
-      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
-      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
-      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
-      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
-      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
-      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
-      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
-      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
-      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
-      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
-      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
-      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
-      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
-      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
-      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
-      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
-      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
-      0xf9, 0xfa };
-  
-  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[0],
-		 bits_dc_luminance, val_dc_luminance);
-  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[0],
-		 bits_ac_luminance, val_ac_luminance);
-  add_huff_table(cinfo, &cinfo->dc_huff_tbl_ptrs[1],
-		 bits_dc_chrominance, val_dc_chrominance);
-  add_huff_table(cinfo, &cinfo->ac_huff_tbl_ptrs[1],
-		 bits_ac_chrominance, val_ac_chrominance);
-}
-
-
-/*
  * Default parameter setup for compression.
  *
  * Applications that don't choose to use this routine must do their
@@ -300,19 +193,19 @@
   if (cinfo->comp_info == NULL)
     cinfo->comp_info = (jpeg_component_info *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  MAX_COMPONENTS * SIZEOF(jpeg_component_info));
+                                  MAX_COMPONENTS * sizeof(jpeg_component_info));
 
   /* Initialize everything not dependent on the color space */
 
 #if JPEG_LIB_VERSION >= 70
-  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_num = 1;         /* 1:1 scaling */
   cinfo->scale_denom = 1;
 #endif
   cinfo->data_precision = BITS_IN_JSAMPLE;
   /* Set up two quantization tables using default quality of 75 */
   jpeg_set_quality(cinfo, 75, TRUE);
   /* Set up two Huffman tables */
-  std_huff_tables(cinfo);
+  std_huff_tables((j_common_ptr) cinfo);
 
   /* Initialize default arithmetic coding conditioning */
   for (i = 0; i < NUM_ARITH_TBLS; i++) {
@@ -370,8 +263,8 @@
    */
   cinfo->JFIF_major_version = 1; /* Default JFIF version = 1.01 */
   cinfo->JFIF_minor_version = 1;
-  cinfo->density_unit = 0;	/* Pixel size is unknown by default */
-  cinfo->X_density = 1;		/* Pixel aspect ratio is square by default */
+  cinfo->density_unit = 0;      /* Pixel size is unknown by default */
+  cinfo->X_density = 1;         /* Pixel aspect ratio is square by default */
   cinfo->Y_density = 1;
 
   /* Choose JPEG colorspace based on input space, set defaults accordingly */
@@ -429,7 +322,7 @@
 GLOBAL(void)
 jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
 {
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   int ci;
 
 #define SET_COMP(index,id,hsamp,vsamp,quant,dctbl,actbl)  \
@@ -497,7 +390,7 @@
     cinfo->num_components = cinfo->input_components;
     if (cinfo->num_components < 1 || cinfo->num_components > MAX_COMPONENTS)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	       MAX_COMPONENTS);
+               MAX_COMPONENTS);
     for (ci = 0; ci < cinfo->num_components; ci++) {
       SET_COMP(ci, ci, 1,1, 0, 0,0);
     }
@@ -511,8 +404,8 @@
 #ifdef C_PROGRESSIVE_SUPPORTED
 
 LOCAL(jpeg_scan_info *)
-fill_a_scan (jpeg_scan_info * scanptr, int ci,
-	     int Ss, int Se, int Ah, int Al)
+fill_a_scan (jpeg_scan_info *scanptr, int ci,
+             int Ss, int Se, int Ah, int Al)
 /* Support routine: generate one scan for specified component */
 {
   scanptr->comps_in_scan = 1;
@@ -526,8 +419,8 @@
 }
 
 LOCAL(jpeg_scan_info *)
-fill_scans (jpeg_scan_info * scanptr, int ncomps,
-	    int Ss, int Se, int Ah, int Al)
+fill_scans (jpeg_scan_info *scanptr, int ncomps,
+            int Ss, int Se, int Ah, int Al)
 /* Support routine: generate one scan for each component */
 {
   int ci;
@@ -545,7 +438,7 @@
 }
 
 LOCAL(jpeg_scan_info *)
-fill_dc_scans (jpeg_scan_info * scanptr, int ncomps, int Ah, int Al)
+fill_dc_scans (jpeg_scan_info *scanptr, int ncomps, int Ah, int Al)
 /* Support routine: generate interleaved DC scan if possible, else N scans */
 {
   int ci;
@@ -577,7 +470,7 @@
 {
   int ncomps = cinfo->num_components;
   int nscans;
-  jpeg_scan_info * scanptr;
+  jpeg_scan_info *scanptr;
 
   /* Safety check to ensure start_compress not called yet. */
   if (cinfo->global_state != CSTATE_START)
@@ -590,9 +483,9 @@
   } else {
     /* All-purpose script for other color spaces. */
     if (ncomps > MAX_COMPS_IN_SCAN)
-      nscans = 6 * ncomps;	/* 2 DC + 4 AC scans per component */
+      nscans = 6 * ncomps;      /* 2 DC + 4 AC scans per component */
     else
-      nscans = 2 + 4 * ncomps;	/* 2 DC scans; 4 AC scans per component */
+      nscans = 2 + 4 * ncomps;  /* 2 DC scans; 4 AC scans per component */
   }
 
   /* Allocate space for script.
@@ -606,7 +499,7 @@
     cinfo->script_space_size = MAX(nscans, 10);
     cinfo->script_space = (jpeg_scan_info *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-			cinfo->script_space_size * SIZEOF(jpeg_scan_info));
+                        cinfo->script_space_size * sizeof(jpeg_scan_info));
   }
   scanptr = cinfo->script_space;
   cinfo->scan_info = scanptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcphuff.c glmark2-2021.02/src/libjpeg-turbo/jcphuff.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcphuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcphuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcphuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy encoding routines for progressive JPEG.
  *
@@ -15,7 +18,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jchuff.h"		/* Declarations shared with jchuff.c */
+#include "jchuff.h"             /* Declarations shared with jchuff.c */
 
 #ifdef C_PROGRESSIVE_SUPPORTED
 
@@ -30,36 +33,36 @@
   /* Bit-level coding status.
    * next_output_byte/free_in_buffer are local copies of cinfo->dest fields.
    */
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
-  INT32 put_buffer;		/* current bit-accumulation buffer */
-  int put_bits;			/* # of bits now in it */
-  j_compress_ptr cinfo;		/* link to cinfo (needed for dump_buffer) */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
+  size_t put_buffer;            /* current bit-accumulation buffer */
+  int put_bits;                 /* # of bits now in it */
+  j_compress_ptr cinfo;         /* link to cinfo (needed for dump_buffer) */
 
   /* Coding status for DC components */
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
 
   /* Coding status for AC components */
-  int ac_tbl_no;		/* the table number of the single component */
-  unsigned int EOBRUN;		/* run length of EOBs */
-  unsigned int BE;		/* # of buffered correction bits before MCU */
-  char * bit_buffer;		/* buffer for correction bits (1 per char) */
+  int ac_tbl_no;                /* the table number of the single component */
+  unsigned int EOBRUN;          /* run length of EOBs */
+  unsigned int BE;              /* # of buffered correction bits before MCU */
+  char *bit_buffer;             /* buffer for correction bits (1 per char) */
   /* packing correction bits tightly would save some space but cost time... */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
-  int next_restart_num;		/* next restart number to write (0-7) */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
+  int next_restart_num;         /* next restart number to write (0-7) */
 
   /* Pointers to derived tables (these workspaces have image lifespan).
    * Since any one scan codes only DC or only AC, we only need one set
    * of tables, not one for DC and one for AC.
    */
-  c_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+  c_derived_tbl *derived_tbls[NUM_HUFF_TBLS];
 
   /* Statistics tables for optimization; again, one set is enough */
-  long * count_ptrs[NUM_HUFF_TBLS];
+  long *count_ptrs[NUM_HUFF_TBLS];
 } phuff_entropy_encoder;
 
-typedef phuff_entropy_encoder * phuff_entropy_ptr;
+typedef phuff_entropy_encoder *phuff_entropy_ptr;
 
 /* MAX_CORR_BITS is the number of bits the AC refinement correction-bit
  * buffer can hold.  Larger sizes may slightly improve compression, but
@@ -67,35 +70,35 @@
  * The minimum safe size is 64 bits.
  */
 
-#define MAX_CORR_BITS  1000	/* Max # of correction bits I can buffer */
+#define MAX_CORR_BITS  1000     /* Max # of correction bits I can buffer */
 
-/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than INT32.
- * We assume that int right shift is unsigned if INT32 right shift is,
+/* IRIGHT_SHIFT is like RIGHT_SHIFT, but works on int rather than JLONG.
+ * We assume that int right shift is unsigned if JLONG right shift is,
  * which should be safe.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	int ishift_temp;
+#define ISHIFT_TEMPS    int ishift_temp;
 #define IRIGHT_SHIFT(x,shft)  \
-	((ishift_temp = (x)) < 0 ? \
-	 (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
-	 (ishift_temp >> (shft)))
+        ((ishift_temp = (x)) < 0 ? \
+         (ishift_temp >> (shft)) | ((~0) << (16-(shft))) : \
+         (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 /* Forward declarations */
-METHODDEF(boolean) encode_mcu_DC_first JPP((j_compress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_AC_first JPP((j_compress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_DC_refine JPP((j_compress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(boolean) encode_mcu_AC_refine JPP((j_compress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(void) finish_pass_phuff JPP((j_compress_ptr cinfo));
-METHODDEF(void) finish_pass_gather_phuff JPP((j_compress_ptr cinfo));
+METHODDEF(boolean) encode_mcu_DC_first (j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_AC_first (j_compress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_DC_refine (j_compress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(boolean) encode_mcu_AC_refine (j_compress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(void) finish_pass_phuff (j_compress_ptr cinfo);
+METHODDEF(void) finish_pass_gather_phuff (j_compress_ptr cinfo);
 
 
 /*
@@ -104,11 +107,11 @@
 
 METHODDEF(void)
 start_pass_phuff (j_compress_ptr cinfo, boolean gather_statistics)
-{  
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   entropy->cinfo = cinfo;
   entropy->gather_statistics = gather_statistics;
@@ -130,9 +133,9 @@
       entropy->pub.encode_mcu = encode_mcu_AC_refine;
       /* AC refinement needs a correction bit buffer */
       if (entropy->bit_buffer == NULL)
-	entropy->bit_buffer = (char *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      MAX_CORR_BITS * SIZEOF(char));
+        entropy->bit_buffer = (char *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      MAX_CORR_BITS * sizeof(char));
     }
   }
   if (gather_statistics)
@@ -149,8 +152,8 @@
     entropy->last_dc_val[ci] = 0;
     /* Get table index */
     if (is_DC_band) {
-      if (cinfo->Ah != 0)	/* DC refinement needs no table */
-	continue;
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
       tbl = compptr->dc_tbl_no;
     } else {
       entropy->ac_tbl_no = tbl = compptr->ac_tbl_no;
@@ -163,15 +166,15 @@
       /* Allocate and zero the statistics tables */
       /* Note that jpeg_gen_optimal_table expects 257 entries in each table! */
       if (entropy->count_ptrs[tbl] == NULL)
-	entropy->count_ptrs[tbl] = (long *)
-	  (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				      257 * SIZEOF(long));
-      MEMZERO(entropy->count_ptrs[tbl], 257 * SIZEOF(long));
+        entropy->count_ptrs[tbl] = (long *)
+          (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                      257 * sizeof(long));
+      MEMZERO(entropy->count_ptrs[tbl], 257 * sizeof(long));
     } else {
       /* Compute derived values for Huffman table */
       /* We may do this more than once for a table, but it's not expensive */
       jpeg_make_c_derived_tbl(cinfo, is_DC_band, tbl,
-			      & entropy->derived_tbls[tbl]);
+                              & entropy->derived_tbls[tbl]);
     }
   }
 
@@ -196,16 +199,16 @@
 
 /* Emit a byte */
 #define emit_byte(entropy,val)  \
-	{ *(entropy)->next_output_byte++ = (JOCTET) (val);  \
-	  if (--(entropy)->free_in_buffer == 0)  \
-	    dump_buffer(entropy); }
+        { *(entropy)->next_output_byte++ = (JOCTET) (val);  \
+          if (--(entropy)->free_in_buffer == 0)  \
+            dump_buffer(entropy); }
 
 
 LOCAL(void)
 dump_buffer (phuff_entropy_ptr entropy)
 /* Empty the output buffer; we do not support suspension in this module. */
 {
-  struct jpeg_destination_mgr * dest = entropy->cinfo->dest;
+  struct jpeg_destination_mgr *dest = entropy->cinfo->dest;
 
   if (! (*dest->empty_output_buffer) (entropy->cinfo))
     ERREXIT(entropy->cinfo, JERR_CANT_SUSPEND);
@@ -228,7 +231,7 @@
 /* Emit some bits, unless we are in gather mode */
 {
   /* This routine is heavily used, so it's worth coding tightly. */
-  register INT32 put_buffer = (INT32) code;
+  register size_t put_buffer = (size_t) code;
   register int put_bits = entropy->put_bits;
 
   /* if size is 0, caller used an invalid Huffman table entry */
@@ -236,21 +239,21 @@
     ERREXIT(entropy->cinfo, JERR_HUFF_MISSING_CODE);
 
   if (entropy->gather_statistics)
-    return;			/* do nothing if we're only getting stats */
+    return;                     /* do nothing if we're only getting stats */
+
+  put_buffer &= (((size_t) 1)<<size) - 1; /* mask off any extra bits in code */
+
+  put_bits += size;             /* new number of bits in buffer */
 
-  put_buffer &= (((INT32) 1)<<size) - 1; /* mask off any extra bits in code */
-  
-  put_bits += size;		/* new number of bits in buffer */
-  
   put_buffer <<= 24 - put_bits; /* align incoming bits */
 
   put_buffer |= entropy->put_buffer; /* and merge with old buffer contents */
 
   while (put_bits >= 8) {
     int c = (int) ((put_buffer >> 16) & 0xFF);
-    
+
     emit_byte(entropy, c);
-    if (c == 0xFF) {		/* need to stuff a zero byte? */
+    if (c == 0xFF) {            /* need to stuff a zero byte? */
       emit_byte(entropy, 0);
     }
     put_buffer <<= 8;
@@ -281,7 +284,7 @@
   if (entropy->gather_statistics)
     entropy->count_ptrs[tbl_no][symbol]++;
   else {
-    c_derived_tbl * tbl = entropy->derived_tbls[tbl_no];
+    c_derived_tbl *tbl = entropy->derived_tbls[tbl_no];
     emit_bits(entropy, tbl->ehufco[symbol], tbl->ehufsi[symbol]);
   }
 }
@@ -292,11 +295,11 @@
  */
 
 LOCAL(void)
-emit_buffered_bits (phuff_entropy_ptr entropy, char * bufstart,
-		    unsigned int nbits)
+emit_buffered_bits (phuff_entropy_ptr entropy, char *bufstart,
+                    unsigned int nbits)
 {
   if (entropy->gather_statistics)
-    return;			/* no real work */
+    return;                     /* no real work */
 
   while (nbits > 0) {
     emit_bits(entropy, (unsigned int) (*bufstart), 1);
@@ -315,7 +318,7 @@
 {
   register int temp, nbits;
 
-  if (entropy->EOBRUN > 0) {	/* if there is any pending EOBRUN */
+  if (entropy->EOBRUN > 0) {    /* if there is any pending EOBRUN */
     temp = entropy->EOBRUN;
     nbits = 0;
     while ((temp >>= 1))
@@ -380,7 +383,7 @@
   int blkn, ci;
   int Al = cinfo->Al;
   JBLOCKROW block;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   ISHIFT_TEMPS
 
   entropy->next_output_byte = cinfo->dest->next_output_byte;
@@ -409,12 +412,12 @@
     /* Encode the DC coefficient difference per section G.1.2.1 */
     temp2 = temp;
     if (temp < 0) {
-      temp = -temp;		/* temp is abs value of input */
+      temp = -temp;             /* temp is abs value of input */
       /* For a negative input, want temp2 = bitwise complement of abs(input) */
       /* This code assumes we are on a two's complement machine */
       temp2--;
     }
-    
+
     /* Find the number of bits needed for the magnitude of the coefficient */
     nbits = 0;
     while (temp) {
@@ -426,13 +429,13 @@
      */
     if (nbits > MAX_COEF_BITS+1)
       ERREXIT(cinfo, JERR_BAD_DCT_COEF);
-    
+
     /* Count/emit the Huffman-coded symbol for the number of bits */
     emit_symbol(entropy, compptr->dc_tbl_no, nbits);
-    
+
     /* Emit that number of bits of the value, if positive, */
     /* or the complement of its magnitude, if negative. */
-    if (nbits)			/* emit_bits rejects calls with size 0 */
+    if (nbits)                  /* emit_bits rejects calls with size 0 */
       emit_bits(entropy, (unsigned int) temp2, nbits);
   }
 
@@ -481,9 +484,9 @@
   block = MCU_data[0];
 
   /* Encode the AC coefficients per section G.1.2.2, fig. G.3 */
-  
-  r = 0;			/* r = run length of zeros */
-   
+
+  r = 0;                        /* r = run length of zeros */
+
   for (k = cinfo->Ss; k <= Se; k++) {
     if ((temp = (*block)[jpeg_natural_order[k]]) == 0) {
       r++;
@@ -495,12 +498,12 @@
      * interwoven with finding the abs value (temp) and output bits (temp2).
      */
     if (temp < 0) {
-      temp = -temp;		/* temp is abs value of input */
-      temp >>= Al;		/* apply the point transform */
+      temp = -temp;             /* temp is abs value of input */
+      temp >>= Al;              /* apply the point transform */
       /* For a negative coef, want temp2 = bitwise complement of abs(coef) */
       temp2 = ~temp;
     } else {
-      temp >>= Al;		/* apply the point transform */
+      temp >>= Al;              /* apply the point transform */
       temp2 = temp;
     }
     /* Watch out for case that nonzero coef is zero after point transform */
@@ -519,7 +522,7 @@
     }
 
     /* Find the number of bits needed for the magnitude of the coefficient */
-    nbits = 1;			/* there must be at least one 1 bit */
+    nbits = 1;                  /* there must be at least one 1 bit */
     while ((temp >>= 1))
       nbits++;
     /* Check for out-of-range coefficient values */
@@ -533,13 +536,13 @@
     /* or the complement of its magnitude, if negative. */
     emit_bits(entropy, (unsigned int) temp2, nbits);
 
-    r = 0;			/* reset zero run length */
+    r = 0;                      /* reset zero run length */
   }
 
-  if (r > 0) {			/* If there are trailing zeroes, */
-    entropy->EOBRUN++;		/* count an EOB */
+  if (r > 0) {                  /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
     if (entropy->EOBRUN == 0x7FFF)
-      emit_eobrun(entropy);	/* force it out to avoid overflow */
+      emit_eobrun(entropy);     /* force it out to avoid overflow */
   }
 
   cinfo->dest->next_output_byte = entropy->next_output_byte;
@@ -648,17 +651,17 @@
      * in C, we shift after obtaining the absolute value.
      */
     if (temp < 0)
-      temp = -temp;		/* temp is abs value of input */
-    temp >>= Al;		/* apply the point transform */
-    absvalues[k] = temp;	/* save abs value for main pass */
+      temp = -temp;             /* temp is abs value of input */
+    temp >>= Al;                /* apply the point transform */
+    absvalues[k] = temp;        /* save abs value for main pass */
     if (temp == 1)
-      EOB = k;			/* EOB = index of last newly-nonzero coef */
+      EOB = k;                  /* EOB = index of last newly-nonzero coef */
   }
 
   /* Encode the AC coefficients per section G.1.2.3, fig. G.7 */
-  
-  r = 0;			/* r = run length of zeros */
-  BR = 0;			/* BR = count of buffered bits added now */
+
+  r = 0;                        /* r = run length of zeros */
+  BR = 0;                       /* BR = count of buffered bits added now */
   BR_buffer = entropy->bit_buffer + entropy->BE; /* Append bits to buffer */
 
   for (k = cinfo->Ss; k <= Se; k++) {
@@ -705,12 +708,12 @@
     emit_buffered_bits(entropy, BR_buffer, BR);
     BR_buffer = entropy->bit_buffer; /* BE bits are gone now */
     BR = 0;
-    r = 0;			/* reset zero run length */
+    r = 0;                      /* reset zero run length */
   }
 
-  if (r > 0 || BR > 0) {	/* If there are trailing zeroes, */
-    entropy->EOBRUN++;		/* count an EOB */
-    entropy->BE += BR;		/* concat my correction bits to older ones */
+  if (r > 0 || BR > 0) {        /* If there are trailing zeroes, */
+    entropy->EOBRUN++;          /* count an EOB */
+    entropy->BE += BR;          /* concat my correction bits to older ones */
     /* We force out the EOB if we risk either:
      * 1. overflow of the EOB counter;
      * 2. overflow of the correction bit buffer during the next MCU.
@@ -742,7 +745,7 @@
 
 METHODDEF(void)
 finish_pass_phuff (j_compress_ptr cinfo)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
 
   entropy->next_output_byte = cinfo->dest->next_output_byte;
@@ -767,7 +770,7 @@
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JHUFF_TBL **htblptr;
   boolean did[NUM_HUFF_TBLS];
 
@@ -779,13 +782,13 @@
   /* It's important not to apply jpeg_gen_optimal_table more than once
    * per table, because it clobbers the input frequency counts!
    */
-  MEMZERO(did, SIZEOF(did));
+  MEMZERO(did, sizeof(did));
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
     if (is_DC_band) {
-      if (cinfo->Ah != 0)	/* DC refinement needs no table */
-	continue;
+      if (cinfo->Ah != 0)       /* DC refinement needs no table */
+        continue;
       tbl = compptr->dc_tbl_no;
     } else {
       tbl = compptr->ac_tbl_no;
@@ -816,7 +819,7 @@
 
   entropy = (phuff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(phuff_entropy_encoder));
+                                sizeof(phuff_entropy_encoder));
   cinfo->entropy = (struct jpeg_entropy_encoder *) entropy;
   entropy->pub.start_pass = start_pass_phuff;
 
@@ -825,7 +828,7 @@
     entropy->derived_tbls[i] = NULL;
     entropy->count_ptrs[i] = NULL;
   }
-  entropy->bit_buffer = NULL;	/* needed only in AC refinement scan */
+  entropy->bit_buffer = NULL;   /* needed only in AC refinement scan */
 }
 
 #endif /* C_PROGRESSIVE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcprepct.c glmark2-2021.02/src/libjpeg-turbo/jcprepct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcprepct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcprepct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jcprepct.c
  *
+ * This file is part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the compression preprocessing controller.
  * This controller manages the color conversion, downsampling,
@@ -58,16 +61,16 @@
    */
   JSAMPARRAY color_buf[MAX_COMPONENTS];
 
-  JDIMENSION rows_to_go;	/* counts rows remaining in source image */
-  int next_buf_row;		/* index of next row to store in color_buf */
+  JDIMENSION rows_to_go;        /* counts rows remaining in source image */
+  int next_buf_row;             /* index of next row to store in color_buf */
 
-#ifdef CONTEXT_ROWS_SUPPORTED	/* only needed for context case */
-  int this_row_group;		/* starting row index of group to process */
-  int next_buf_stop;		/* downsample when we reach this index */
+#ifdef CONTEXT_ROWS_SUPPORTED   /* only needed for context case */
+  int this_row_group;           /* starting row index of group to process */
+  int next_buf_stop;            /* downsample when we reach this index */
 #endif
 } my_prep_controller;
 
-typedef my_prep_controller * my_prep_ptr;
+typedef my_prep_controller *my_prep_ptr;
 
 
 /*
@@ -104,13 +107,13 @@
 
 LOCAL(void)
 expand_bottom_edge (JSAMPARRAY image_data, JDIMENSION num_cols,
-		    int input_rows, int output_rows)
+                    int input_rows, int output_rows)
 {
   register int row;
 
   for (row = input_rows; row < output_rows; row++) {
     jcopy_sample_rows(image_data, input_rows-1, image_data, row,
-		      1, num_cols);
+                      1, num_cols);
   }
 }
 
@@ -126,43 +129,43 @@
 
 METHODDEF(void)
 pre_process_data (j_compress_ptr cinfo,
-		  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-		  JDIMENSION in_rows_avail,
-		  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
-		  JDIMENSION out_row_groups_avail)
+                  JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                  JDIMENSION in_rows_avail,
+                  JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                  JDIMENSION out_row_groups_avail)
 {
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int numrows, ci;
   JDIMENSION inrows;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   while (*in_row_ctr < in_rows_avail &&
-	 *out_row_group_ctr < out_row_groups_avail) {
+         *out_row_group_ctr < out_row_groups_avail) {
     /* Do color conversion to fill the conversion buffer. */
     inrows = in_rows_avail - *in_row_ctr;
     numrows = cinfo->max_v_samp_factor - prep->next_buf_row;
     numrows = (int) MIN((JDIMENSION) numrows, inrows);
     (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
-				       prep->color_buf,
-				       (JDIMENSION) prep->next_buf_row,
-				       numrows);
+                                       prep->color_buf,
+                                       (JDIMENSION) prep->next_buf_row,
+                                       numrows);
     *in_row_ctr += numrows;
     prep->next_buf_row += numrows;
     prep->rows_to_go -= numrows;
     /* If at bottom of image, pad to fill the conversion buffer. */
     if (prep->rows_to_go == 0 &&
-	prep->next_buf_row < cinfo->max_v_samp_factor) {
+        prep->next_buf_row < cinfo->max_v_samp_factor) {
       for (ci = 0; ci < cinfo->num_components; ci++) {
-	expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
-			   prep->next_buf_row, cinfo->max_v_samp_factor);
+        expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+                           prep->next_buf_row, cinfo->max_v_samp_factor);
       }
       prep->next_buf_row = cinfo->max_v_samp_factor;
     }
     /* If we've filled the conversion buffer, empty it. */
     if (prep->next_buf_row == cinfo->max_v_samp_factor) {
       (*cinfo->downsample->downsample) (cinfo,
-					prep->color_buf, (JDIMENSION) 0,
-					output_buf, *out_row_group_ctr);
+                                        prep->color_buf, (JDIMENSION) 0,
+                                        output_buf, *out_row_group_ctr);
       prep->next_buf_row = 0;
       (*out_row_group_ctr)++;
     }
@@ -170,16 +173,16 @@
      * Note we assume the caller is providing a one-iMCU-height output buffer!
      */
     if (prep->rows_to_go == 0 &&
-	*out_row_group_ctr < out_row_groups_avail) {
+        *out_row_group_ctr < out_row_groups_avail) {
       for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	   ci++, compptr++) {
-	expand_bottom_edge(output_buf[ci],
-			   compptr->width_in_blocks * DCTSIZE,
-			   (int) (*out_row_group_ctr * compptr->v_samp_factor),
-			   (int) (out_row_groups_avail * compptr->v_samp_factor));
+           ci++, compptr++) {
+        expand_bottom_edge(output_buf[ci],
+                           compptr->width_in_blocks * DCTSIZE,
+                           (int) (*out_row_group_ctr * compptr->v_samp_factor),
+                           (int) (out_row_groups_avail * compptr->v_samp_factor));
       }
       *out_row_group_ctr = out_row_groups_avail;
-      break;			/* can exit outer loop without test */
+      break;                    /* can exit outer loop without test */
     }
   }
 }
@@ -193,10 +196,10 @@
 
 METHODDEF(void)
 pre_process_context (j_compress_ptr cinfo,
-		     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-		     JDIMENSION in_rows_avail,
-		     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
-		     JDIMENSION out_row_groups_avail)
+                     JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
+                     JDIMENSION in_rows_avail,
+                     JSAMPIMAGE output_buf, JDIMENSION *out_row_group_ctr,
+                     JDIMENSION out_row_groups_avail)
 {
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int numrows, ci;
@@ -210,19 +213,19 @@
       numrows = prep->next_buf_stop - prep->next_buf_row;
       numrows = (int) MIN((JDIMENSION) numrows, inrows);
       (*cinfo->cconvert->color_convert) (cinfo, input_buf + *in_row_ctr,
-					 prep->color_buf,
-					 (JDIMENSION) prep->next_buf_row,
-					 numrows);
+                                         prep->color_buf,
+                                         (JDIMENSION) prep->next_buf_row,
+                                         numrows);
       /* Pad at top of image, if first time through */
       if (prep->rows_to_go == cinfo->image_height) {
-	for (ci = 0; ci < cinfo->num_components; ci++) {
-	  int row;
-	  for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
-	    jcopy_sample_rows(prep->color_buf[ci], 0,
-			      prep->color_buf[ci], -row,
-			      1, cinfo->image_width);
-	  }
-	}
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          int row;
+          for (row = 1; row <= cinfo->max_v_samp_factor; row++) {
+            jcopy_sample_rows(prep->color_buf[ci], 0,
+                              prep->color_buf[ci], -row,
+                              1, cinfo->image_width);
+          }
+        }
       }
       *in_row_ctr += numrows;
       prep->next_buf_row += numrows;
@@ -230,29 +233,29 @@
     } else {
       /* Return for more data, unless we are at the bottom of the image. */
       if (prep->rows_to_go != 0)
-	break;
+        break;
       /* When at bottom of image, pad to fill the conversion buffer. */
       if (prep->next_buf_row < prep->next_buf_stop) {
-	for (ci = 0; ci < cinfo->num_components; ci++) {
-	  expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
-			     prep->next_buf_row, prep->next_buf_stop);
-	}
-	prep->next_buf_row = prep->next_buf_stop;
+        for (ci = 0; ci < cinfo->num_components; ci++) {
+          expand_bottom_edge(prep->color_buf[ci], cinfo->image_width,
+                             prep->next_buf_row, prep->next_buf_stop);
+        }
+        prep->next_buf_row = prep->next_buf_stop;
       }
     }
     /* If we've gotten enough data, downsample a row group. */
     if (prep->next_buf_row == prep->next_buf_stop) {
       (*cinfo->downsample->downsample) (cinfo,
-					prep->color_buf,
-					(JDIMENSION) prep->this_row_group,
-					output_buf, *out_row_group_ctr);
+                                        prep->color_buf,
+                                        (JDIMENSION) prep->this_row_group,
+                                        output_buf, *out_row_group_ctr);
       (*out_row_group_ctr)++;
       /* Advance pointers with wraparound as necessary. */
       prep->this_row_group += cinfo->max_v_samp_factor;
       if (prep->this_row_group >= buf_height)
-	prep->this_row_group = 0;
+        prep->this_row_group = 0;
       if (prep->next_buf_row >= buf_height)
-	prep->next_buf_row = 0;
+        prep->next_buf_row = 0;
       prep->next_buf_stop = prep->next_buf_row + cinfo->max_v_samp_factor;
     }
   }
@@ -269,7 +272,7 @@
   my_prep_ptr prep = (my_prep_ptr) cinfo->prep;
   int rgroup_height = cinfo->max_v_samp_factor;
   int ci, i;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JSAMPARRAY true_buffer, fake_buffer;
 
   /* Grab enough space for fake row pointers for all the components;
@@ -277,8 +280,8 @@
    */
   fake_buffer = (JSAMPARRAY)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(cinfo->num_components * 5 * rgroup_height) *
-				SIZEOF(JSAMPROW));
+                                (cinfo->num_components * 5 * rgroup_height) *
+                                sizeof(JSAMPROW));
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
@@ -289,11 +292,11 @@
     true_buffer = (*cinfo->mem->alloc_sarray)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
        (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
-		      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+                      cinfo->max_h_samp_factor) / compptr->h_samp_factor),
        (JDIMENSION) (3 * rgroup_height));
     /* Copy true buffer row pointers into the middle of the fake row array */
     MEMCOPY(fake_buffer + rgroup_height, true_buffer,
-	    3 * rgroup_height * SIZEOF(JSAMPROW));
+            3 * rgroup_height * sizeof(JSAMPROW));
     /* Fill in the above and below wraparound pointers */
     for (i = 0; i < rgroup_height; i++) {
       fake_buffer[i] = true_buffer[2 * rgroup_height + i];
@@ -316,14 +319,14 @@
 {
   my_prep_ptr prep;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
-  if (need_full_buffer)		/* safety check */
+  if (need_full_buffer)         /* safety check */
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 
   prep = (my_prep_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_prep_controller));
+                                sizeof(my_prep_controller));
   cinfo->prep = (struct jpeg_c_prep_controller *) prep;
   prep->pub.start_pass = start_pass_prep;
 
@@ -343,12 +346,12 @@
     /* No context, just make it tall enough for one row group */
     prep->pub.pre_process_data = pre_process_data;
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       prep->color_buf[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
-			cinfo->max_h_samp_factor) / compptr->h_samp_factor),
-	 (JDIMENSION) cinfo->max_v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         (JDIMENSION) (((long) compptr->width_in_blocks * DCTSIZE *
+                        cinfo->max_h_samp_factor) / compptr->h_samp_factor),
+         (JDIMENSION) cinfo->max_v_samp_factor);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcsample.c glmark2-2021.02/src/libjpeg-turbo/jcsample.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcsample.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcsample.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jcsample.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains downsampling routines.
  *
@@ -53,20 +57,21 @@
 
 
 /* Pointer to routine to downsample a single component */
-typedef JMETHOD(void, downsample1_ptr,
-		(j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data));
+typedef void (*downsample1_ptr) (j_compress_ptr cinfo,
+                                 jpeg_component_info *compptr,
+                                 JSAMPARRAY input_data,
+                                 JSAMPARRAY output_data);
 
 /* Private subobject */
 
 typedef struct {
-  struct jpeg_downsampler pub;	/* public fields */
+  struct jpeg_downsampler pub;  /* public fields */
 
   /* Downsampling method pointers, one per component */
   downsample1_ptr methods[MAX_COMPONENTS];
 } my_downsampler;
 
-typedef my_downsampler * my_downsample_ptr;
+typedef my_downsampler *my_downsample_ptr;
 
 
 /*
@@ -87,7 +92,7 @@
 
 LOCAL(void)
 expand_right_edge (JSAMPARRAY image_data, int num_rows,
-		   JDIMENSION input_cols, JDIMENSION output_cols)
+                   JDIMENSION input_cols, JDIMENSION output_cols)
 {
   register JSAMPROW ptr;
   register JSAMPLE pixval;
@@ -98,9 +103,9 @@
   if (numcols > 0) {
     for (row = 0; row < num_rows; row++) {
       ptr = image_data[row] + input_cols;
-      pixval = ptr[-1];		/* don't need GETJSAMPLE() here */
+      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
       for (count = numcols; count > 0; count--)
-	*ptr++ = pixval;
+        *ptr++ = pixval;
     }
   }
 }
@@ -114,12 +119,12 @@
 
 METHODDEF(void)
 sep_downsample (j_compress_ptr cinfo,
-		JSAMPIMAGE input_buf, JDIMENSION in_row_index,
-		JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
+                JSAMPIMAGE input_buf, JDIMENSION in_row_index,
+                JSAMPIMAGE output_buf, JDIMENSION out_row_group_index)
 {
   my_downsample_ptr downsample = (my_downsample_ptr) cinfo->downsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JSAMPARRAY in_ptr, out_ptr;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -139,14 +144,14 @@
  */
 
 METHODDEF(void)
-int_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		JSAMPARRAY input_data, JSAMPARRAY output_data)
+int_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow, h_expand, v_expand, numpix, numpix2, h, v;
-  JDIMENSION outcol, outcol_h;	/* outcol_h == outcol*h_expand */
+  JDIMENSION outcol, outcol_h;  /* outcol_h == outcol*h_expand */
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   JSAMPROW inptr, outptr;
-  INT32 outvalue;
+  JLONG outvalue;
 
   h_expand = cinfo->max_h_samp_factor / compptr->h_samp_factor;
   v_expand = cinfo->max_v_samp_factor / compptr->v_samp_factor;
@@ -158,19 +163,19 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * h_expand);
+                    cinfo->image_width, output_cols * h_expand);
 
   inrow = 0;
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     for (outcol = 0, outcol_h = 0; outcol < output_cols;
-	 outcol++, outcol_h += h_expand) {
+         outcol++, outcol_h += h_expand) {
       outvalue = 0;
       for (v = 0; v < v_expand; v++) {
-	inptr = input_data[inrow+v] + outcol_h;
-	for (h = 0; h < h_expand; h++) {
-	  outvalue += (INT32) GETJSAMPLE(*inptr++);
-	}
+        inptr = input_data[inrow+v] + outcol_h;
+        for (h = 0; h < h_expand; h++) {
+          outvalue += (JLONG) GETJSAMPLE(*inptr++);
+        }
       }
       *outptr++ = (JSAMPLE) ((outvalue + numpix2) / numpix);
     }
@@ -186,15 +191,15 @@
  */
 
 METHODDEF(void)
-fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY output_data)
+fullsize_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   /* Copy the data */
   jcopy_sample_rows(input_data, 0, output_data, 0,
-		    cinfo->max_v_samp_factor, cinfo->image_width);
+                    cinfo->max_v_samp_factor, cinfo->image_width);
   /* Edge-expand */
   expand_right_edge(output_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
+                    cinfo->image_width, compptr->width_in_blocks * DCTSIZE);
 }
 
 
@@ -211,8 +216,8 @@
  */
 
 METHODDEF(void)
-h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                 JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int outrow;
   JDIMENSION outcol;
@@ -225,16 +230,16 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     inptr = input_data[outrow];
-    bias = 0;			/* bias = 0,1,0,1,... for successive samples */
+    bias = 0;                   /* bias = 0,1,0,1,... for successive samples */
     for (outcol = 0; outcol < output_cols; outcol++) {
       *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr) + GETJSAMPLE(inptr[1])
-			      + bias) >> 1);
-      bias ^= 1;		/* 0=>1, 1=>0 */
+                              + bias) >> 1);
+      bias ^= 1;                /* 0=>1, 1=>0 */
       inptr += 2;
     }
   }
@@ -248,8 +253,8 @@
  */
 
 METHODDEF(void)
-h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                 JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow;
   JDIMENSION outcol;
@@ -262,19 +267,19 @@
    * efficient.
    */
   expand_right_edge(input_data, cinfo->max_v_samp_factor,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   inrow = 0;
   for (outrow = 0; outrow < compptr->v_samp_factor; outrow++) {
     outptr = output_data[outrow];
     inptr0 = input_data[inrow];
     inptr1 = input_data[inrow+1];
-    bias = 1;			/* bias = 1,2,1,2,... for successive samples */
+    bias = 1;                   /* bias = 1,2,1,2,... for successive samples */
     for (outcol = 0; outcol < output_cols; outcol++) {
       *outptr++ = (JSAMPLE) ((GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-			      GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
-			      + bias) >> 2);
-      bias ^= 3;		/* 1=>2, 2=>1 */
+                              GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1])
+                              + bias) >> 2);
+      bias ^= 3;                /* 1=>2, 2=>1 */
       inptr0 += 2; inptr1 += 2;
     }
     inrow += 2;
@@ -291,21 +296,21 @@
  */
 
 METHODDEF(void)
-h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
-			JSAMPARRAY input_data, JSAMPARRAY output_data)
+h2v2_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int inrow, outrow;
   JDIMENSION colctr;
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   register JSAMPROW inptr0, inptr1, above_ptr, below_ptr, outptr;
-  INT32 membersum, neighsum, memberscale, neighscale;
+  JLONG membersum, neighsum, memberscale, neighscale;
 
   /* Expand input data enough to let all the output samples be generated
    * by the standard loop.  Special-casing padded output would be more
    * efficient.
    */
   expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
-		    cinfo->image_width, output_cols * 2);
+                    cinfo->image_width, output_cols * 2);
 
   /* We don't bother to form the individual "smoothed" input pixel values;
    * we can directly compute the output which is the average of the four
@@ -333,14 +338,14 @@
 
     /* Special case for first column: pretend column -1 is same as column 0 */
     membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
     neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-	       GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
-	       GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[2]) +
+               GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[2]);
     neighsum += neighsum;
     neighsum += GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[2]) +
-		GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
+                GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[2]);
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
     inptr0 += 2; inptr1 += 2; above_ptr += 2; below_ptr += 2;
@@ -348,17 +353,17 @@
     for (colctr = output_cols - 2; colctr > 0; colctr--) {
       /* sum of pixels directly mapped to this output element */
       membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                  GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
       /* sum of edge-neighbor pixels */
       neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-		 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-		 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
-		 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
+                 GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+                 GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[2]) +
+                 GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[2]);
       /* The edge-neighbors count twice as much as corner-neighbors */
       neighsum += neighsum;
       /* Add in the corner-neighbors */
       neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[2]) +
-		  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
+                  GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[2]);
       /* form final output scaled up by 2^16 */
       membersum = membersum * memberscale + neighsum * neighscale;
       /* round, descale and output it */
@@ -368,14 +373,14 @@
 
     /* Special case for last column */
     membersum = GETJSAMPLE(*inptr0) + GETJSAMPLE(inptr0[1]) +
-		GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
+                GETJSAMPLE(*inptr1) + GETJSAMPLE(inptr1[1]);
     neighsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(above_ptr[1]) +
-	       GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
-	       GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
-	       GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
+               GETJSAMPLE(*below_ptr) + GETJSAMPLE(below_ptr[1]) +
+               GETJSAMPLE(inptr0[-1]) + GETJSAMPLE(inptr0[1]) +
+               GETJSAMPLE(inptr1[-1]) + GETJSAMPLE(inptr1[1]);
     neighsum += neighsum;
     neighsum += GETJSAMPLE(above_ptr[-1]) + GETJSAMPLE(above_ptr[1]) +
-		GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
+                GETJSAMPLE(below_ptr[-1]) + GETJSAMPLE(below_ptr[1]);
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr = (JSAMPLE) ((membersum + 32768) >> 16);
 
@@ -392,13 +397,13 @@
 
 METHODDEF(void)
 fullsize_smooth_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
-			    JSAMPARRAY input_data, JSAMPARRAY output_data)
+                            JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
   int outrow;
   JDIMENSION colctr;
   JDIMENSION output_cols = compptr->width_in_blocks * DCTSIZE;
   register JSAMPROW inptr, above_ptr, below_ptr, outptr;
-  INT32 membersum, neighsum, memberscale, neighscale;
+  JLONG membersum, neighsum, memberscale, neighscale;
   int colsum, lastcolsum, nextcolsum;
 
   /* Expand input data enough to let all the output samples be generated
@@ -406,7 +411,7 @@
    * efficient.
    */
   expand_right_edge(input_data - 1, cinfo->max_v_samp_factor + 2,
-		    cinfo->image_width, output_cols);
+                    cinfo->image_width, output_cols);
 
   /* Each of the eight neighbor pixels contributes a fraction SF to the
    * smoothed pixel, while the main pixel contributes (1-8*SF).  In order
@@ -425,10 +430,10 @@
 
     /* Special case for first column */
     colsum = GETJSAMPLE(*above_ptr++) + GETJSAMPLE(*below_ptr++) +
-	     GETJSAMPLE(*inptr);
+             GETJSAMPLE(*inptr);
     membersum = GETJSAMPLE(*inptr++);
     nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-		 GETJSAMPLE(*inptr);
+                 GETJSAMPLE(*inptr);
     neighsum = colsum + (colsum - membersum) + nextcolsum;
     membersum = membersum * memberscale + neighsum * neighscale;
     *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
@@ -438,7 +443,7 @@
       membersum = GETJSAMPLE(*inptr++);
       above_ptr++; below_ptr++;
       nextcolsum = GETJSAMPLE(*above_ptr) + GETJSAMPLE(*below_ptr) +
-		   GETJSAMPLE(*inptr);
+                   GETJSAMPLE(*inptr);
       neighsum = lastcolsum + (colsum - membersum) + nextcolsum;
       membersum = membersum * memberscale + neighsum * neighscale;
       *outptr++ = (JSAMPLE) ((membersum + 32768) >> 16);
@@ -467,12 +472,12 @@
 {
   my_downsample_ptr downsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   boolean smoothok = TRUE;
 
   downsample = (my_downsample_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_downsampler));
+                                sizeof(my_downsampler));
   cinfo->downsample = (struct jpeg_downsampler *) downsample;
   downsample->pub.start_pass = start_pass_downsample;
   downsample->pub.downsample = sep_downsample;
@@ -485,35 +490,42 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor == cinfo->max_h_samp_factor &&
-	compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+        compptr->v_samp_factor == cinfo->max_v_samp_factor) {
 #ifdef INPUT_SMOOTHING_SUPPORTED
       if (cinfo->smoothing_factor) {
-	downsample->methods[ci] = fullsize_smooth_downsample;
-	downsample->pub.need_context_rows = TRUE;
+        downsample->methods[ci] = fullsize_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
       } else
 #endif
-	downsample->methods[ci] = fullsize_downsample;
+        downsample->methods[ci] = fullsize_downsample;
     } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
-	       compptr->v_samp_factor == cinfo->max_v_samp_factor) {
+               compptr->v_samp_factor == cinfo->max_v_samp_factor) {
       smoothok = FALSE;
       if (jsimd_can_h2v1_downsample())
         downsample->methods[ci] = jsimd_h2v1_downsample;
       else
         downsample->methods[ci] = h2v1_downsample;
     } else if (compptr->h_samp_factor * 2 == cinfo->max_h_samp_factor &&
-	       compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
+               compptr->v_samp_factor * 2 == cinfo->max_v_samp_factor) {
 #ifdef INPUT_SMOOTHING_SUPPORTED
       if (cinfo->smoothing_factor) {
-	downsample->methods[ci] = h2v2_smooth_downsample;
-	downsample->pub.need_context_rows = TRUE;
+#if defined(__mips__)
+        if (jsimd_can_h2v2_smooth_downsample())
+          downsample->methods[ci] = jsimd_h2v2_smooth_downsample;
+        else
+#endif
+          downsample->methods[ci] = h2v2_smooth_downsample;
+        downsample->pub.need_context_rows = TRUE;
       } else
 #endif
-	if (jsimd_can_h2v2_downsample())
-	  downsample->methods[ci] = jsimd_h2v2_downsample;
-	else
-	  downsample->methods[ci] = h2v2_downsample;
+      {
+        if (jsimd_can_h2v2_downsample())
+          downsample->methods[ci] = jsimd_h2v2_downsample;
+        else
+          downsample->methods[ci] = h2v2_downsample;
+      }
     } else if ((cinfo->max_h_samp_factor % compptr->h_samp_factor) == 0 &&
-	       (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
+               (cinfo->max_v_samp_factor % compptr->v_samp_factor) == 0) {
       smoothok = FALSE;
       downsample->methods[ci] = int_downsample;
     } else
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcstest.c glmark2-2021.02/src/libjpeg-turbo/jcstest.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jcstest.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jcstest.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* This program demonstrates how to check for the colorspace extension
+   capabilities of libjpeg-turbo at both compile time and run time. */
+
+#include <stdio.h>
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+
+#ifndef JCS_EXTENSIONS
+#define JCS_EXT_RGB 6
+#endif
+#if !defined(JCS_EXTENSIONS) || !defined(JCS_ALPHA_EXTENSIONS)
+#define JCS_EXT_RGBA 12
+#endif
+
+static char lasterror[JMSG_LENGTH_MAX] = "No error";
+
+typedef struct _error_mgr {
+  struct jpeg_error_mgr pub;
+  jmp_buf jb;
+} error_mgr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+  error_mgr *myerr = (error_mgr *)cinfo->err;
+  (*cinfo->err->output_message)(cinfo);
+  longjmp(myerr->jb, 1);
+}
+
+static void my_output_message(j_common_ptr cinfo)
+{
+  (*cinfo->err->format_message)(cinfo, lasterror);
+}
+
+int main(void)
+{
+  int jcs_valid = -1, jcs_alpha_valid = -1;
+  struct jpeg_compress_struct cinfo;
+  error_mgr jerr;
+
+  printf("libjpeg-turbo colorspace extensions:\n");
+  #if JCS_EXTENSIONS
+  printf("  Present at compile time\n");
+  #else
+  printf("  Not present at compile time\n");
+  #endif
+
+  cinfo.err = jpeg_std_error(&jerr.pub);
+  jerr.pub.error_exit = my_error_exit;
+  jerr.pub.output_message = my_output_message;
+
+  if(setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_valid = 0;
+    goto done;
+  }
+
+  jpeg_create_compress(&cinfo);
+  cinfo.input_components = 3;
+  jpeg_set_defaults(&cinfo);
+  cinfo.in_color_space = JCS_EXT_RGB;
+  jpeg_default_colorspace(&cinfo);
+  jcs_valid = 1;
+
+  done:
+  if (jcs_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  printf("libjpeg-turbo alpha colorspace extensions:\n");
+  #if JCS_ALPHA_EXTENSIONS
+  printf("  Present at compile time\n");
+  #else
+  printf("  Not present at compile time\n");
+  #endif
+
+  if(setjmp(jerr.jb)) {
+    /* this will execute if libjpeg has an error */
+    jcs_alpha_valid = 0;
+    goto done2;
+  }
+
+  cinfo.in_color_space = JCS_EXT_RGBA;
+  jpeg_default_colorspace(&cinfo);
+  jcs_alpha_valid = 1;
+
+  done2:
+  if (jcs_alpha_valid)
+    printf("  Working properly\n");
+  else
+    printf("  Not working properly.  Error returned was:\n    %s\n",
+           lasterror);
+
+  jpeg_destroy_compress(&cinfo);
+  return 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jctrans.c glmark2-2021.02/src/libjpeg-turbo/jctrans.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jctrans.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jctrans.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jctrans.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1998, Thomas G. Lane.
  * Modified 2000-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains library routines for transcoding compression,
  * that is, writing raw DCT coefficient arrays to an output JPEG file.
@@ -18,9 +21,9 @@
 
 /* Forward declarations */
 LOCAL(void) transencode_master_selection
-	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+        (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays);
 LOCAL(void) transencode_coef_controller
-	JPP((j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays));
+        (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays);
 
 
 /*
@@ -36,7 +39,7 @@
  */
 
 GLOBAL(void)
-jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr * coef_arrays)
+jpeg_write_coefficients (j_compress_ptr cinfo, jvirt_barray_ptr *coef_arrays)
 {
   if (cinfo->global_state != CSTATE_START)
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
@@ -48,7 +51,7 @@
   /* Perform master selection of active modules */
   transencode_master_selection(cinfo, coef_arrays);
   /* Wait for jpeg_finish_compress() call */
-  cinfo->next_scanline = 0;	/* so jpeg_write_marker works */
+  cinfo->next_scanline = 0;     /* so jpeg_write_marker works */
   cinfo->global_state = CSTATE_WRCOEFS;
 }
 
@@ -62,9 +65,9 @@
 
 GLOBAL(void)
 jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
-			       j_compress_ptr dstinfo)
+                               j_compress_ptr dstinfo)
 {
-  JQUANT_TBL ** qtblptr;
+  JQUANT_TBL **qtblptr;
   jpeg_component_info *incomp, *outcomp;
   JQUANT_TBL *c_quant, *slot_quant;
   int tblno, ci, coefi;
@@ -96,10 +99,10 @@
     if (srcinfo->quant_tbl_ptrs[tblno] != NULL) {
       qtblptr = & dstinfo->quant_tbl_ptrs[tblno];
       if (*qtblptr == NULL)
-	*qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
+        *qtblptr = jpeg_alloc_quant_table((j_common_ptr) dstinfo);
       MEMCOPY((*qtblptr)->quantval,
-	      srcinfo->quant_tbl_ptrs[tblno]->quantval,
-	      SIZEOF((*qtblptr)->quantval));
+              srcinfo->quant_tbl_ptrs[tblno]->quantval,
+              sizeof((*qtblptr)->quantval));
       (*qtblptr)->sent_table = FALSE;
     }
   }
@@ -109,7 +112,7 @@
   dstinfo->num_components = srcinfo->num_components;
   if (dstinfo->num_components < 1 || dstinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(dstinfo, JERR_COMPONENT_COUNT, dstinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
   for (ci = 0, incomp = srcinfo->comp_info, outcomp = dstinfo->comp_info;
        ci < dstinfo->num_components; ci++, incomp++, outcomp++) {
     outcomp->component_id = incomp->component_id;
@@ -122,14 +125,14 @@
      */
     tblno = outcomp->quant_tbl_no;
     if (tblno < 0 || tblno >= NUM_QUANT_TBLS ||
-	srcinfo->quant_tbl_ptrs[tblno] == NULL)
+        srcinfo->quant_tbl_ptrs[tblno] == NULL)
       ERREXIT1(dstinfo, JERR_NO_QUANT_TABLE, tblno);
     slot_quant = srcinfo->quant_tbl_ptrs[tblno];
     c_quant = incomp->quant_table;
     if (c_quant != NULL) {
       for (coefi = 0; coefi < DCTSIZE2; coefi++) {
-	if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
-	  ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
+        if (c_quant->quantval[coefi] != slot_quant->quantval[coefi])
+          ERREXIT1(dstinfo, JERR_MISMATCHED_QUANT_TABLE, tblno);
       }
     }
     /* Note: we do not copy the source's Huffman table assignments;
@@ -163,7 +166,7 @@
 
 LOCAL(void)
 transencode_master_selection (j_compress_ptr cinfo,
-			      jvirt_barray_ptr * coef_arrays)
+                              jvirt_barray_ptr *coef_arrays)
 {
   /* Although we don't actually use input_components for transcoding,
    * jcmaster.c's initial_setup will complain if input_components is 0.
@@ -219,19 +222,19 @@
 typedef struct {
   struct jpeg_c_coef_controller pub; /* public fields */
 
-  JDIMENSION iMCU_row_num;	/* iMCU row # within image */
-  JDIMENSION mcu_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
+  JDIMENSION iMCU_row_num;      /* iMCU row # within image */
+  JDIMENSION mcu_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
 
   /* Virtual block array for each component. */
-  jvirt_barray_ptr * whole_image;
+  jvirt_barray_ptr *whole_image;
 
   /* Workspace for constructing dummy blocks at right/bottom edges. */
   JBLOCKROW dummy_buffer[C_MAX_BLOCKS_IN_MCU];
 } my_coef_controller;
 
-typedef my_coef_controller * my_coef_ptr;
+typedef my_coef_controller *my_coef_ptr;
 
 
 LOCAL(void)
@@ -289,7 +292,7 @@
 compress_output (j_compress_ptr cinfo, JSAMPIMAGE input_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, ci, xindex, yindex, yoffset, blockcnt;
@@ -312,44 +315,44 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->mcu_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						: compptr->last_col_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (coef->iMCU_row_num < last_iMCU_row ||
-	      yindex+yoffset < compptr->last_row_height) {
-	    /* Fill in pointers to real blocks in this row */
-	    buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	    for (xindex = 0; xindex < blockcnt; xindex++)
-	      MCU_buffer[blkn++] = buffer_ptr++;
-	  } else {
-	    /* At bottom of image, need a whole row of dummy blocks */
-	    xindex = 0;
-	  }
-	  /* Fill in any dummy blocks needed in this row.
-	   * Dummy blocks are filled in the same way as in jccoefct.c:
-	   * all zeroes in the AC entries, DC entries equal to previous
-	   * block's DC value.  The init routine has already zeroed the
-	   * AC entries, so we need only set the DC entries correctly.
-	   */
-	  for (; xindex < compptr->MCU_width; xindex++) {
-	    MCU_buffer[blkn] = coef->dummy_buffer[blkn];
-	    MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
-	    blkn++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        blockcnt = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                : compptr->last_col_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          if (coef->iMCU_row_num < last_iMCU_row ||
+              yindex+yoffset < compptr->last_row_height) {
+            /* Fill in pointers to real blocks in this row */
+            buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+            for (xindex = 0; xindex < blockcnt; xindex++)
+              MCU_buffer[blkn++] = buffer_ptr++;
+          } else {
+            /* At bottom of image, need a whole row of dummy blocks */
+            xindex = 0;
+          }
+          /* Fill in any dummy blocks needed in this row.
+           * Dummy blocks are filled in the same way as in jccoefct.c:
+           * all zeroes in the AC entries, DC entries equal to previous
+           * block's DC value.  The init routine has already zeroed the
+           * AC entries, so we need only set the DC entries correctly.
+           */
+          for (; xindex < compptr->MCU_width; xindex++) {
+            MCU_buffer[blkn] = coef->dummy_buffer[blkn];
+            MCU_buffer[blkn][0][0] = MCU_buffer[blkn-1][0][0];
+            blkn++;
+          }
+        }
       }
       /* Try to write the MCU. */
       if (! (*cinfo->entropy->encode_mcu) (cinfo, MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->mcu_ctr = MCU_col_num;
-	return FALSE;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->mcu_ctr = MCU_col_num;
+        return FALSE;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -372,7 +375,7 @@
 
 LOCAL(void)
 transencode_coef_controller (j_compress_ptr cinfo,
-			     jvirt_barray_ptr * coef_arrays)
+                             jvirt_barray_ptr *coef_arrays)
 {
   my_coef_ptr coef;
   JBLOCKROW buffer;
@@ -380,7 +383,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_c_coef_controller *) coef;
   coef->pub.start_pass = start_pass_coef;
   coef->pub.compress_data = compress_output;
@@ -391,8 +394,8 @@
   /* Allocate and pre-zero space for dummy DCT blocks. */
   buffer = (JBLOCKROW)
     (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
-  jzero_far((void FAR *) buffer, C_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
+  jzero_far((void *) buffer, C_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
   for (i = 0; i < C_MAX_BLOCKS_IN_MCU; i++) {
     coef->dummy_buffer[i] = buffer + i;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdapimin.c glmark2-2021.02/src/libjpeg-turbo/jdapimin.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdapimin.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdapimin.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdapimin.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the decompression half
  * of the JPEG library.  These are the "minimum" API routines that may be
@@ -19,6 +22,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
+#include "jdmaster.h"
 
 
 /*
@@ -32,12 +36,12 @@
   int i;
 
   /* Guard against version mismatches between library and caller. */
-  cinfo->mem = NULL;		/* so jpeg_destroy knows mem mgr not called */
+  cinfo->mem = NULL;            /* so jpeg_destroy knows mem mgr not called */
   if (version != JPEG_LIB_VERSION)
     ERREXIT2(cinfo, JERR_BAD_LIB_VERSION, JPEG_LIB_VERSION, version);
-  if (structsize != SIZEOF(struct jpeg_decompress_struct))
-    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE, 
-	     (int) SIZEOF(struct jpeg_decompress_struct), (int) structsize);
+  if (structsize != sizeof(struct jpeg_decompress_struct))
+    ERREXIT2(cinfo, JERR_BAD_STRUCT_SIZE,
+             (int) sizeof(struct jpeg_decompress_struct), (int) structsize);
 
   /* For debugging purposes, we zero the whole master structure.
    * But the application has already set the err pointer, and may have set
@@ -48,7 +52,7 @@
   {
     struct jpeg_error_mgr * err = cinfo->err;
     void * client_data = cinfo->client_data; /* ignore Purify complaint here */
-    MEMZERO(cinfo, SIZEOF(struct jpeg_decompress_struct));
+    MEMZERO(cinfo, sizeof(struct jpeg_decompress_struct));
     cinfo->err = err;
     cinfo->client_data = client_data;
   }
@@ -80,6 +84,14 @@
 
   /* OK, I'm ready */
   cinfo->global_state = DSTATE_START;
+
+  /* The master struct is used to store extension parameters, so we allocate it
+   * here.
+   */
+  cinfo->master = (struct jpeg_decomp_master *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_decomp_master));
+  MEMZERO(cinfo->master, sizeof(my_decomp_master));
 }
 
 
@@ -121,22 +133,22 @@
     cinfo->jpeg_color_space = JCS_GRAYSCALE;
     cinfo->out_color_space = JCS_GRAYSCALE;
     break;
-    
+
   case 3:
     if (cinfo->saw_JFIF_marker) {
       cinfo->jpeg_color_space = JCS_YCbCr; /* JFIF implies YCbCr */
     } else if (cinfo->saw_Adobe_marker) {
       switch (cinfo->Adobe_transform) {
       case 0:
-	cinfo->jpeg_color_space = JCS_RGB;
-	break;
+        cinfo->jpeg_color_space = JCS_RGB;
+        break;
       case 1:
-	cinfo->jpeg_color_space = JCS_YCbCr;
-	break;
+        cinfo->jpeg_color_space = JCS_YCbCr;
+        break;
       default:
-	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
-	break;
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+        break;
       }
     } else {
       /* Saw no special markers, try to guess from the component IDs */
@@ -145,31 +157,31 @@
       int cid2 = cinfo->comp_info[2].component_id;
 
       if (cid0 == 1 && cid1 == 2 && cid2 == 3)
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume JFIF w/out marker */
       else if (cid0 == 82 && cid1 == 71 && cid2 == 66)
-	cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
+        cinfo->jpeg_color_space = JCS_RGB; /* ASCII 'R', 'G', 'B' */
       else {
-	TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
-	cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
+        TRACEMS3(cinfo, 1, JTRC_UNKNOWN_IDS, cid0, cid1, cid2);
+        cinfo->jpeg_color_space = JCS_YCbCr; /* assume it's YCbCr */
       }
     }
     /* Always guess RGB is proper output colorspace. */
     cinfo->out_color_space = JCS_RGB;
     break;
-    
+
   case 4:
     if (cinfo->saw_Adobe_marker) {
       switch (cinfo->Adobe_transform) {
       case 0:
-	cinfo->jpeg_color_space = JCS_CMYK;
-	break;
+        cinfo->jpeg_color_space = JCS_CMYK;
+        break;
       case 2:
-	cinfo->jpeg_color_space = JCS_YCCK;
-	break;
+        cinfo->jpeg_color_space = JCS_YCCK;
+        break;
       default:
-	WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
-	cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
-	break;
+        WARNMS1(cinfo, JWRN_ADOBE_XFORM, cinfo->Adobe_transform);
+        cinfo->jpeg_color_space = JCS_YCCK; /* assume it's YCCK */
+        break;
       }
     } else {
       /* No special markers, assume straight CMYK. */
@@ -177,7 +189,7 @@
     }
     cinfo->out_color_space = JCS_CMYK;
     break;
-    
+
   default:
     cinfo->jpeg_color_space = JCS_UNKNOWN;
     cinfo->out_color_space = JCS_UNKNOWN;
@@ -185,7 +197,7 @@
   }
 
   /* Set defaults for other decompression parameters. */
-  cinfo->scale_num = 1;		/* 1:1 scaling */
+  cinfo->scale_num = 1;         /* 1:1 scaling */
   cinfo->scale_denom = 1;
   cinfo->output_gamma = 1.0;
   cinfo->buffered_image = FALSE;
@@ -253,7 +265,7 @@
     retcode = JPEG_HEADER_OK;
     break;
   case JPEG_REACHED_EOI:
-    if (require_image)		/* Complain if application wanted an image */
+    if (require_image)          /* Complain if application wanted an image */
       ERREXIT(cinfo, JERR_NO_IMAGE);
     /* Reset to start state; it would be safer to require the application to
      * call jpeg_abort, but we can't change it now for compatibility reasons.
@@ -385,7 +397,7 @@
   /* Read until EOI */
   while (! cinfo->inputctl->eoi_reached) {
     if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
-      return FALSE;		/* Suspend, come back later */
+      return FALSE;             /* Suspend, come back later */
   }
   /* Do final cleanup */
   (*cinfo->src->term_source) (cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdapistd.c glmark2-2021.02/src/libjpeg-turbo/jdapistd.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdapistd.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdapistd.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jdapistd.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains application interface code for the decompression half
  * of the JPEG library.  These are the "standard" API routines that are
@@ -15,14 +18,14 @@
  * whole decompression library into a transcoder.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
-#include "jpegcomp.h"
-
+#include "jdmainct.h"
+#include "jdcoefct.h"
+#include "jdsample.h"
+#include "jmemsys.h"
 
 /* Forward declarations */
-LOCAL(boolean) output_pass_setup JPP((j_decompress_ptr cinfo));
+LOCAL(boolean) output_pass_setup (j_decompress_ptr cinfo);
 
 
 /*
@@ -54,24 +57,24 @@
     if (cinfo->inputctl->has_multiple_scans) {
 #ifdef D_MULTISCAN_FILES_SUPPORTED
       for (;;) {
-	int retcode;
-	/* Call progress monitor hook if present */
-	if (cinfo->progress != NULL)
-	  (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
-	/* Absorb some more input */
-	retcode = (*cinfo->inputctl->consume_input) (cinfo);
-	if (retcode == JPEG_SUSPENDED)
-	  return FALSE;
-	if (retcode == JPEG_REACHED_EOI)
-	  break;
-	/* Advance progress counter if appropriate */
-	if (cinfo->progress != NULL &&
-	    (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
-	  if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
-	    /* jdmaster underestimated number of scans; ratchet up one scan */
-	    cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
-	  }
-	}
+        int retcode;
+        /* Call progress monitor hook if present */
+        if (cinfo->progress != NULL)
+          (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        /* Absorb some more input */
+        retcode = (*cinfo->inputctl->consume_input) (cinfo);
+        if (retcode == JPEG_SUSPENDED)
+          return FALSE;
+        if (retcode == JPEG_REACHED_EOI)
+          break;
+        /* Advance progress counter if appropriate */
+        if (cinfo->progress != NULL &&
+            (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+          if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+            /* jdmaster underestimated number of scans; ratchet up one scan */
+            cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+          }
+        }
       }
 #else
       ERREXIT(cinfo, JERR_NOT_COMPILED);
@@ -110,16 +113,16 @@
       JDIMENSION last_scanline;
       /* Call progress monitor hook if present */
       if (cinfo->progress != NULL) {
-	cinfo->progress->pass_counter = (long) cinfo->output_scanline;
-	cinfo->progress->pass_limit = (long) cinfo->output_height;
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        cinfo->progress->pass_counter = (long) cinfo->output_scanline;
+        cinfo->progress->pass_limit = (long) cinfo->output_height;
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       }
       /* Process some data */
       last_scanline = cinfo->output_scanline;
       (*cinfo->main->process_data) (cinfo, (JSAMPARRAY) NULL,
-				    &cinfo->output_scanline, (JDIMENSION) 0);
+                                    &cinfo->output_scanline, (JDIMENSION) 0);
       if (cinfo->output_scanline == last_scanline)
-	return FALSE;		/* No progress made, must suspend */
+        return FALSE;           /* No progress made, must suspend */
     }
     /* Finish up dummy pass, and set up for another one */
     (*cinfo->master->finish_output_pass) (cinfo);
@@ -138,6 +141,110 @@
 
 
 /*
+ * Enable partial scanline decompression
+ *
+ * Must be called after jpeg_start_decompress() and before any calls to
+ * jpeg_read_scanlines() or jpeg_skip_scanlines().
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(void)
+jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                    JDIMENSION *width)
+{
+  int ci, align, orig_downsampled_width;
+  JDIMENSION input_xoffset;
+  boolean reinit_upsampler = FALSE;
+  jpeg_component_info *compptr;
+
+  if (cinfo->global_state != DSTATE_SCANNING || cinfo->output_scanline != 0)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  if (!xoffset || !width)
+    ERREXIT(cinfo, JERR_BAD_CROP_SPEC);
+
+  /* xoffset and width must fall within the output image dimensions. */
+  if (*width == 0 || *xoffset + *width > cinfo->output_width)
+    ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
+
+  /* No need to do anything if the caller wants the entire width. */
+  if (*width == cinfo->output_width)
+    return;
+
+  /* Ensuring the proper alignment of xoffset is tricky.  At minimum, it
+   * must align with an MCU boundary, because:
+   *
+   *   (1) The IDCT is performed in blocks, and it is not feasible to modify
+   *       the algorithm so that it can transform partial blocks.
+   *   (2) Because of the SIMD extensions, any input buffer passed to the
+   *       upsampling and color conversion routines must be aligned to the
+   *       SIMD word size (for instance, 128-bit in the case of SSE2.)  The
+   *       easiest way to accomplish this without copying data is to ensure
+   *       that upsampling and color conversion begin at the start of the
+   *       first MCU column that will be inverse transformed.
+   *
+   * In practice, we actually impose a stricter alignment requirement.  We
+   * require that xoffset be a multiple of the maximum MCU column width of all
+   * of the components (the "iMCU column width.")  This is to simplify the
+   * single-pass decompression case, allowing us to use the same MCU column
+   * width for all of the components.
+   */
+  align = cinfo->_min_DCT_scaled_size * cinfo->max_h_samp_factor;
+
+  /* Adjust xoffset to the nearest iMCU boundary <= the requested value */
+  input_xoffset = *xoffset;
+  *xoffset = (input_xoffset / align) * align;
+
+  /* Adjust the width so that the right edge of the output image is as
+   * requested (only the left edge is altered.)  It is important that calling
+   * programs check this value after this function returns, so that they can
+   * allocate an output buffer with the appropriate size.
+   */
+  *width = *width + input_xoffset - *xoffset;
+  cinfo->output_width = *width;
+
+  /* Set the first and last iMCU columns that we must decompress.  These values
+   * will be used in single-scan decompressions.
+   */
+  cinfo->master->first_iMCU_col =
+    (JDIMENSION) (long) (*xoffset) / (long) align;
+  cinfo->master->last_iMCU_col =
+    (JDIMENSION) jdiv_round_up((long) (*xoffset + cinfo->output_width),
+                               (long) align) - 1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    /* Set downsampled_width to the new output width. */
+    orig_downsampled_width = compptr->downsampled_width;
+    compptr->downsampled_width =
+      (JDIMENSION) jdiv_round_up((long) (cinfo->output_width *
+                                         compptr->h_samp_factor),
+                                 (long) cinfo->max_h_samp_factor);
+    if (compptr->downsampled_width < 2 && orig_downsampled_width >= 2)
+      reinit_upsampler = TRUE;
+
+    /* Set the first and last iMCU columns that we must decompress.  These
+     * values will be used in multi-scan decompressions.
+     */
+    cinfo->master->first_MCU_col[ci] =
+      (JDIMENSION) (long) (*xoffset * compptr->h_samp_factor) /
+                   (long) align;
+    cinfo->master->last_MCU_col[ci] =
+      (JDIMENSION) jdiv_round_up((long) ((*xoffset + cinfo->output_width) *
+                                         compptr->h_samp_factor),
+                                 (long) align) - 1;
+  }
+
+  if (reinit_upsampler) {
+    cinfo->master->jinit_upsampler_no_alloc = TRUE;
+    jinit_upsampler(cinfo);
+    cinfo->master->jinit_upsampler_no_alloc = FALSE;
+  }
+}
+
+
+/*
  * Read some scanlines of data from the JPEG decompressor.
  *
  * The return value will be the number of lines actually read.
@@ -152,7 +259,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_read_scanlines (j_decompress_ptr cinfo, JSAMPARRAY scanlines,
-		     JDIMENSION max_lines)
+                     JDIMENSION max_lines)
 {
   JDIMENSION row_ctr;
 
@@ -178,6 +285,236 @@
 }
 
 
+/* Dummy color convert function used by jpeg_skip_scanlines() */
+LOCAL(void)
+noop_convert (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+              JDIMENSION input_row, JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+
+/*
+ * In some cases, it is best to call jpeg_read_scanlines() and discard the
+ * output, rather than skipping the scanlines, because this allows us to
+ * maintain the internal state of the context-based upsampler.  In these cases,
+ * we set up and tear down a dummy color converter in order to avoid valgrind
+ * errors and to achieve the best possible performance.
+ */
+
+LOCAL(void)
+read_and_discard_scanlines (j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  JDIMENSION n;
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows);
+
+  color_convert = cinfo->cconvert->color_convert;
+  cinfo->cconvert->color_convert = noop_convert;
+
+  for (n = 0; n < num_lines; n++)
+    jpeg_read_scanlines(cinfo, NULL, 1);
+
+  cinfo->cconvert->color_convert = color_convert;
+}
+
+
+/*
+ * Called by jpeg_skip_scanlines().  This partially skips a decompress block by
+ * incrementing the rowgroup counter.
+ */
+
+LOCAL(void)
+increment_simple_rowgroup_ctr (j_decompress_ptr cinfo, JDIMENSION rows)
+{
+  JDIMENSION rows_left;
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+
+  /* Increment the counter to the next row group after the skipped rows. */
+  main_ptr->rowgroup_ctr += rows / cinfo->max_v_samp_factor;
+
+  /* Partially skipping a row group would involve modifying the internal state
+   * of the upsampler, so read the remaining rows into a dummy buffer instead.
+   */
+  rows_left = rows % cinfo->max_v_samp_factor;
+  cinfo->output_scanline += rows - rows_left;
+
+  read_and_discard_scanlines(cinfo, rows_left);
+}
+
+/*
+ * Skips some scanlines of data from the JPEG decompressor.
+ *
+ * The return value will be the number of lines actually skipped.  If skipping
+ * num_lines would move beyond the end of the image, then the actual number of
+ * lines remaining in the image is returned.  Otherwise, the return value will
+ * be equal to num_lines.
+ *
+ * Refer to libjpeg.txt for more information.
+ */
+
+GLOBAL(JDIMENSION)
+jpeg_skip_scanlines (j_decompress_ptr cinfo, JDIMENSION num_lines)
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  JDIMENSION i, x;
+  int y;
+  JDIMENSION lines_per_iMCU_row, lines_left_in_iMCU_row, lines_after_iMCU_row;
+  JDIMENSION lines_to_skip, lines_to_read;
+
+  if (cinfo->global_state != DSTATE_SCANNING)
+    ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
+
+  /* Do not skip past the bottom of the image. */
+  if (cinfo->output_scanline + num_lines >= cinfo->output_height) {
+    cinfo->output_scanline = cinfo->output_height;
+    return cinfo->output_height - cinfo->output_scanline;
+  }
+
+  if (num_lines == 0)
+    return 0;
+
+  lines_per_iMCU_row = cinfo->_min_DCT_scaled_size * cinfo->max_v_samp_factor;
+  lines_left_in_iMCU_row =
+    (lines_per_iMCU_row - (cinfo->output_scanline % lines_per_iMCU_row)) %
+    lines_per_iMCU_row;
+  lines_after_iMCU_row = num_lines - lines_left_in_iMCU_row;
+
+  /* Skip the lines remaining in the current iMCU row.  When upsampling
+   * requires context rows, we need the previous and next rows in order to read
+   * the current row.  This adds some complexity.
+   */
+  if (cinfo->upsample->need_context_rows) {
+    /* If the skipped lines would not move us past the current iMCU row, we
+     * read the lines and ignore them.  There might be a faster way of doing
+     * this, but we are facing increasing complexity for diminishing returns.
+     * The increasing complexity would be a by-product of meddling with the
+     * state machine used to skip context rows.  Near the end of an iMCU row,
+     * the next iMCU row may have already been entropy-decoded.  In this unique
+     * case, we will read the next iMCU row if we cannot skip past it as well.
+     */
+    if ((num_lines < lines_left_in_iMCU_row + 1) ||
+        (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full &&
+         lines_after_iMCU_row < lines_per_iMCU_row + 1)) {
+      read_and_discard_scanlines(cinfo, num_lines);
+      return num_lines;
+    }
+
+    /* If the next iMCU row has already been entropy-decoded, make sure that
+     * we do not skip too far.
+     */
+    if (lines_left_in_iMCU_row <= 1 && main_ptr->buffer_full) {
+      cinfo->output_scanline += lines_left_in_iMCU_row + lines_per_iMCU_row;
+      lines_after_iMCU_row -= lines_per_iMCU_row;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+    }
+
+    /* If we have just completed the first block, adjust the buffer pointers */
+    if (main_ptr->iMCU_row_ctr == 0 ||
+        (main_ptr->iMCU_row_ctr == 1 && lines_left_in_iMCU_row > 2))
+      set_wraparound_pointers(cinfo);
+    main_ptr->buffer_full = FALSE;
+    main_ptr->rowgroup_ctr = 0;
+    main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
+    upsample->next_row_out = cinfo->max_v_samp_factor;
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+  }
+
+  /* Skipping is much simpler when context rows are not required. */
+  else {
+    if (num_lines < lines_left_in_iMCU_row) {
+      increment_simple_rowgroup_ctr(cinfo, num_lines);
+      return num_lines;
+    } else {
+      cinfo->output_scanline += lines_left_in_iMCU_row;
+      main_ptr->buffer_full = FALSE;
+      main_ptr->rowgroup_ctr = 0;
+      upsample->next_row_out = cinfo->max_v_samp_factor;
+      upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    }
+  }
+
+  /* Calculate how many full iMCU rows we can skip. */
+  if (cinfo->upsample->need_context_rows)
+    lines_to_skip = ((lines_after_iMCU_row - 1) / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  else
+    lines_to_skip = (lines_after_iMCU_row / lines_per_iMCU_row) *
+                    lines_per_iMCU_row;
+  /* Calculate the number of lines that remain to be skipped after skipping all
+   * of the full iMCU rows that we can.  We will not read these lines unless we
+   * have to.
+   */
+  lines_to_read = lines_after_iMCU_row - lines_to_skip;
+
+  /* For images requiring multiple scans (progressive, non-interleaved, etc.),
+   * all of the entropy decoding occurs in jpeg_start_decompress(), assuming
+   * that the input data source is non-suspending.  This makes skipping easy.
+   */
+  if (cinfo->inputctl->has_multiple_scans) {
+    if (cinfo->upsample->need_context_rows) {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      main_ptr->iMCU_row_ctr += lines_after_iMCU_row / lines_per_iMCU_row;
+      /* It is complex to properly move to the middle of a context block, so
+       * read the remaining lines instead of skipping them.
+       */
+      read_and_discard_scanlines(cinfo, lines_to_read);
+    } else {
+      cinfo->output_scanline += lines_to_skip;
+      cinfo->output_iMCU_row += lines_to_skip / lines_per_iMCU_row;
+      increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+    }
+    upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+    return num_lines;
+  }
+
+  /* Skip the iMCU rows that we can safely skip. */
+  for (i = 0; i < lines_to_skip; i += lines_per_iMCU_row) {
+    for (y = 0; y < coef->MCU_rows_per_iMCU_row; y++) {
+      for (x = 0; x < cinfo->MCUs_per_row; x++) {
+        /* Calling decode_mcu() with a NULL pointer causes it to discard the
+         * decoded coefficients.  This is ~5% faster for large subsets, but
+         * it's tough to tell a difference for smaller images.
+         */
+        (*cinfo->entropy->decode_mcu) (cinfo, NULL);
+      }
+    }
+    cinfo->input_iMCU_row++;
+    cinfo->output_iMCU_row++;
+    if (cinfo->input_iMCU_row < cinfo->total_iMCU_rows)
+      start_iMCU_row(cinfo);
+    else
+      (*cinfo->inputctl->finish_input_pass) (cinfo);
+  }
+  cinfo->output_scanline += lines_to_skip;
+
+  if (cinfo->upsample->need_context_rows) {
+    /* Context-based upsampling keeps track of iMCU rows. */
+    main_ptr->iMCU_row_ctr += lines_to_skip / lines_per_iMCU_row;
+
+    /* It is complex to properly move to the middle of a context block, so
+     * read the remaining lines instead of skipping them.
+     */
+    read_and_discard_scanlines(cinfo, lines_to_read);
+  } else {
+    increment_simple_rowgroup_ctr(cinfo, lines_to_read);
+  }
+
+  /* Since skipping lines involves skipping the upsampling step, the value of
+   * "rows_to_go" will become invalid unless we set it here.  NOTE: This is a
+   * bit odd, since "rows_to_go" seems to be redundantly keeping track of
+   * output_scanline.
+   */
+  upsample->rows_to_go = cinfo->output_height - cinfo->output_scanline;
+
+  /* Always skip the requested number of lines. */
+  return num_lines;
+}
+
 /*
  * Alternate entry point to read raw data.
  * Processes exactly one iMCU row per call, unless suspended.
@@ -185,7 +522,7 @@
 
 GLOBAL(JDIMENSION)
 jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
-		    JDIMENSION max_lines)
+                    JDIMENSION max_lines)
 {
   JDIMENSION lines_per_iMCU_row;
 
@@ -210,7 +547,7 @@
 
   /* Decompress directly into user's buffer. */
   if (! (*cinfo->coef->decompress_data) (cinfo, data))
-    return 0;			/* suspension forced, can do nothing more */
+    return 0;                   /* suspension forced, can do nothing more */
 
   /* OK, we processed one iMCU row. */
   cinfo->output_scanline += lines_per_iMCU_row;
@@ -266,9 +603,9 @@
   }
   /* Read markers looking for SOS or EOI */
   while (cinfo->input_scan_number <= cinfo->output_scan_number &&
-	 ! cinfo->inputctl->eoi_reached) {
+         ! cinfo->inputctl->eoi_reached) {
     if ((*cinfo->inputctl->consume_input) (cinfo) == JPEG_SUSPENDED)
-      return FALSE;		/* Suspend, come back later */
+      return FALSE;             /* Suspend, come back later */
   }
   cinfo->global_state = DSTATE_BUFIMAGE;
   return TRUE;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdarith.c glmark2-2021.02/src/libjpeg-turbo/jdarith.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdarith.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdarith.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdarith.c
  *
- * Developed 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * This file was part of the Independent JPEG Group's software:
+ * Developed 1997-2015 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains portable arithmetic entropy decoding routines for JPEG
  * (implementing the ISO/IEC IS 10918-1 and CCITT Recommendation ITU-T T.81).
@@ -23,8 +26,8 @@
 typedef struct {
   struct jpeg_entropy_decoder pub; /* public fields */
 
-  INT32 c;       /* C register, base of coding interval + input bit buffer */
-  INT32 a;               /* A register, normalized size of coding interval */
+  JLONG c;       /* C register, base of coding interval + input bit buffer */
+  JLONG a;               /* A register, normalized size of coding interval */
   int ct;     /* bit shift counter, # of bits left in bit buffer part of C */
                                                          /* init: ct = -16 */
                                                          /* run: ct = 0..7 */
@@ -32,17 +35,17 @@
   int last_dc_val[MAX_COMPS_IN_SCAN]; /* last DC coef for each component */
   int dc_context[MAX_COMPS_IN_SCAN]; /* context index for DC conditioning */
 
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to statistics areas (these workspaces have image lifespan) */
-  unsigned char * dc_stats[NUM_ARITH_TBLS];
-  unsigned char * ac_stats[NUM_ARITH_TBLS];
+  unsigned char *dc_stats[NUM_ARITH_TBLS];
+  unsigned char *ac_stats[NUM_ARITH_TBLS];
 
   /* Statistics bin for coding with fixed probability 0.5 */
   unsigned char fixed_bin[4];
 } arith_entropy_decoder;
 
-typedef arith_entropy_decoder * arith_entropy_ptr;
+typedef arith_entropy_decoder *arith_entropy_ptr;
 
 /* The following two definitions specify the allocation chunk size
  * for the statistics area.
@@ -65,7 +68,7 @@
 get_byte (j_decompress_ptr cinfo)
 /* Read next input byte; we do not support suspension in this module. */
 {
-  struct jpeg_source_mgr * src = cinfo->src;
+  struct jpeg_source_mgr *src = cinfo->src;
 
   if (src->bytes_in_buffer == 0)
     if (! (*src->fill_input_buffer) (cinfo))
@@ -94,7 +97,7 @@
  * (instead of fixed) with the bit shift counter CT.
  * Thus, we also need only one (variable instead of
  * fixed size) shift for the LPS/MPS decision, and
- * we can get away with any renormalization update
+ * we can do away with any renormalization update
  * of C (except for new data insertion, of course).
  *
  * I've also introduced a new scheme for accessing
@@ -107,7 +110,7 @@
 {
   register arith_entropy_ptr e = (arith_entropy_ptr) cinfo->entropy;
   register unsigned char nl, nm;
-  register INT32 qe, temp;
+  register JLONG qe, temp;
   register int sv, data;
 
   /* Renormalization & data input per section D.2.6 */
@@ -115,32 +118,32 @@
     if (--e->ct < 0) {
       /* Need to fetch next data byte */
       if (cinfo->unread_marker)
-	data = 0;		/* stuff zero data */
+        data = 0;               /* stuff zero data */
       else {
-	data = get_byte(cinfo);	/* read next input byte */
-	if (data == 0xFF) {	/* zero stuff or marker code */
-	  do data = get_byte(cinfo);
-	  while (data == 0xFF);	/* swallow extra 0xFF bytes */
-	  if (data == 0)
-	    data = 0xFF;	/* discard stuffed zero byte */
-	  else {
-	    /* Note: Different from the Huffman decoder, hitting
-	     * a marker while processing the compressed data
-	     * segment is legal in arithmetic coding.
-	     * The convention is to supply zero data
-	     * then until decoding is complete.
-	     */
-	    cinfo->unread_marker = data;
-	    data = 0;
-	  }
-	}
+        data = get_byte(cinfo); /* read next input byte */
+        if (data == 0xFF) {     /* zero stuff or marker code */
+          do data = get_byte(cinfo);
+          while (data == 0xFF); /* swallow extra 0xFF bytes */
+          if (data == 0)
+            data = 0xFF;        /* discard stuffed zero byte */
+          else {
+            /* Note: Different from the Huffman decoder, hitting
+             * a marker while processing the compressed data
+             * segment is legal in arithmetic coding.
+             * The convention is to supply zero data
+             * then until decoding is complete.
+             */
+            cinfo->unread_marker = data;
+            data = 0;
+          }
+        }
       }
       e->c = (e->c << 8) | data; /* insert data into C register */
-      if ((e->ct += 8) < 0)	 /* update bit shift counter */
-	/* Need more initial bytes */
-	if (++e->ct == 0)
-	  /* Got 2 initial bytes -> re-init A and exit loop */
-	  e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
+      if ((e->ct += 8) < 0)      /* update bit shift counter */
+        /* Need more initial bytes */
+        if (++e->ct == 0)
+          /* Got 2 initial bytes -> re-init A and exit loop */
+          e->a = 0x8000L; /* => e->a = 0x10000L after loop exit */
     }
     e->a <<= 1;
   }
@@ -149,9 +152,9 @@
    * Qe values and probability estimation state machine
    */
   sv = *st;
-  qe = jpeg_aritab[sv & 0x7F];	/* => Qe_Value */
-  nl = qe & 0xFF; qe >>= 8;	/* Next_Index_LPS + Switch_MPS */
-  nm = qe & 0xFF; qe >>= 8;	/* Next_Index_MPS */
+  qe = jpeg_aritab[sv & 0x7F];  /* => Qe_Value */
+  nl = qe & 0xFF; qe >>= 8;     /* Next_Index_LPS + Switch_MPS */
+  nm = qe & 0xFF; qe >>= 8;     /* Next_Index_MPS */
 
   /* Decode & estimation procedures per sections D.2.4 & D.2.5 */
   temp = e->a - qe;
@@ -162,19 +165,19 @@
     /* Conditional LPS (less probable symbol) exchange */
     if (e->a < qe) {
       e->a = qe;
-      *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
     } else {
       e->a = qe;
-      *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
-      sv ^= 0x80;		/* Exchange LPS/MPS */
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
     }
   } else if (e->a < 0x8000L) {
     /* Conditional MPS (more probable symbol) exchange */
     if (e->a < qe) {
-      *st = (sv & 0x80) ^ nl;	/* Estimate_after_LPS */
-      sv ^= 0x80;		/* Exchange LPS/MPS */
+      *st = (sv & 0x80) ^ nl;   /* Estimate_after_LPS */
+      sv ^= 0x80;               /* Exchange LPS/MPS */
     } else {
-      *st = (sv & 0x80) ^ nm;	/* Estimate_after_MPS */
+      *st = (sv & 0x80) ^ nm;   /* Estimate_after_MPS */
     }
   }
 
@@ -191,7 +194,7 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   /* Advance past the RSTn marker */
   if (! (*cinfo->marker->read_restart_marker) (cinfo))
@@ -200,13 +203,13 @@
   /* Re-initialize statistics areas */
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
-    if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       MEMZERO(entropy->dc_stats[compptr->dc_tbl_no], DC_STAT_BINS);
       /* Reset DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
       entropy->dc_context[ci] = 0;
     }
-    if (! cinfo->progressive_mode || cinfo->Ss) {
+    if (!cinfo->progressive_mode || cinfo->Ss) {
       MEMZERO(entropy->ac_stats[compptr->ac_tbl_no], AC_STAT_BINS);
     }
   }
@@ -214,7 +217,7 @@
   /* Reset arithmetic decoding variables */
   entropy->c = 0;
   entropy->a = 0;
-  entropy->ct = -16;	/* force reading 2 initial bytes to fill C */
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
 
   /* Reset restart counter */
   entropy->restarts_to_go = cinfo->restart_interval;
@@ -253,7 +256,7 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* Outer loop handles each block in the MCU */
 
@@ -277,34 +280,34 @@
       st += 2; st += sign;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        st = entropy->dc_stats[tbl] + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;		   /* zero diff category */
+        entropy->dc_context[ci] = 0;               /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+        entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
       else
-	entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
+        entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
       entropy->last_dc_val[ci] += v;
     }
 
     /* Scale and output the DC coefficient (assumes jpeg_natural_order[0]=0) */
-    (*block)[0] = (JCOEF) (entropy->last_dc_val[ci] << cinfo->Al);
+    (*block)[0] = (JCOEF) LEFT_SHIFT(entropy->last_dc_val[ci], cinfo->Al);
   }
 
   return TRUE;
@@ -332,7 +335,7 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* There is always only one block per MCU */
   block = MCU_data[0];
@@ -343,13 +346,13 @@
   /* Figure F.20: Decode_AC_coefficients */
   for (k = cinfo->Ss; k <= cinfo->Se; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
-    if (arith_decode(cinfo, st)) break;		/* EOB flag */
+    if (arith_decode(cinfo, st)) break;         /* EOB flag */
     while (arith_decode(cinfo, st + 1) == 0) {
       st += 3; k++;
       if (k > cinfo->Se) {
-	WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	entropy->ct = -1;			/* spectral overflow */
-	return TRUE;
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
       }
     }
     /* Figure F.21: Decoding nonzero value v */
@@ -359,17 +362,17 @@
     /* Figure F.23: Decoding the magnitude category of v */
     if ((m = arith_decode(cinfo, st)) != 0) {
       if (arith_decode(cinfo, st)) {
-	m <<= 1;
-	st = entropy->ac_stats[tbl] +
-	     (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        m <<= 1;
+        st = entropy->ac_stats[tbl] +
+             (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
     }
     v = m;
@@ -404,8 +407,8 @@
     entropy->restarts_to_go--;
   }
 
-  st = entropy->fixed_bin;	/* use fixed probability estimation */
-  p1 = 1 << cinfo->Al;		/* 1 in the bit position being coded */
+  st = entropy->fixed_bin;      /* use fixed probability estimation */
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
 
   /* Outer loop handles each block in the MCU */
 
@@ -440,14 +443,14 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* There is always only one block per MCU */
   block = MCU_data[0];
   tbl = cinfo->cur_comp_info[0]->ac_tbl_no;
 
-  p1 = 1 << cinfo->Al;		/* 1 in the bit position being coded */
-  m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  p1 = 1 << cinfo->Al;          /* 1 in the bit position being coded */
+  m1 = (-1) << cinfo->Al;       /* -1 in the bit position being coded */
 
   /* Establish EOBx (previous stage end-of-block) index */
   for (kex = cinfo->Se; kex > 0; kex--)
@@ -456,30 +459,30 @@
   for (k = cinfo->Ss; k <= cinfo->Se; k++) {
     st = entropy->ac_stats[tbl] + 3 * (k - 1);
     if (k > kex)
-      if (arith_decode(cinfo, st)) break;	/* EOB flag */
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
     for (;;) {
       thiscoef = *block + jpeg_natural_order[k];
-      if (*thiscoef) {				/* previously nonzero coef */
-	if (arith_decode(cinfo, st + 2)) {
-	  if (*thiscoef < 0)
-	    *thiscoef += m1;
-	  else
-	    *thiscoef += p1;
-	}
-	break;
-      }
-      if (arith_decode(cinfo, st + 1)) {	/* newly nonzero coef */
-	if (arith_decode(cinfo, entropy->fixed_bin))
-	  *thiscoef = m1;
-	else
-	  *thiscoef = p1;
-	break;
+      if (*thiscoef) {                          /* previously nonzero coef */
+        if (arith_decode(cinfo, st + 2)) {
+          if (*thiscoef < 0)
+            *thiscoef += m1;
+          else
+            *thiscoef += p1;
+        }
+        break;
+      }
+      if (arith_decode(cinfo, st + 1)) {        /* newly nonzero coef */
+        if (arith_decode(cinfo, entropy->fixed_bin))
+          *thiscoef = m1;
+        else
+          *thiscoef = p1;
+        break;
       }
       st += 3; k++;
       if (k > cinfo->Se) {
-	WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	entropy->ct = -1;			/* spectral overflow */
-	return TRUE;
+        WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+        entropy->ct = -1;                       /* spectral overflow */
+        return TRUE;
       }
     }
   }
@@ -496,7 +499,7 @@
 decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JBLOCKROW block;
   unsigned char *st;
   int blkn, ci, tbl, sign, k;
@@ -509,12 +512,12 @@
     entropy->restarts_to_go--;
   }
 
-  if (entropy->ct == -1) return TRUE;	/* if error do nothing */
+  if (entropy->ct == -1) return TRUE;   /* if error do nothing */
 
   /* Outer loop handles each block in the MCU */
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    block = MCU_data[blkn];
+    block = MCU_data ? MCU_data[blkn] : NULL;
     ci = cinfo->MCU_membership[blkn];
     compptr = cinfo->cur_comp_info[ci];
 
@@ -535,33 +538,34 @@
       st += 2; st += sign;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	st = entropy->dc_stats[tbl] + 20;	/* Table F.4: X1 = 20 */
-	while (arith_decode(cinfo, st)) {
-	  if ((m <<= 1) == 0x8000) {
-	    WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	    entropy->ct = -1;			/* magnitude overflow */
-	    return TRUE;
-	  }
-	  st += 1;
-	}
+        st = entropy->dc_stats[tbl] + 20;       /* Table F.4: X1 = 20 */
+        while (arith_decode(cinfo, st)) {
+          if ((m <<= 1) == 0x8000) {
+            WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+            entropy->ct = -1;                   /* magnitude overflow */
+            return TRUE;
+          }
+          st += 1;
+        }
       }
       /* Section F.1.4.4.1.2: Establish dc_context conditioning category */
       if (m < (int) ((1L << cinfo->arith_dc_L[tbl]) >> 1))
-	entropy->dc_context[ci] = 0;		   /* zero diff category */
+        entropy->dc_context[ci] = 0;               /* zero diff category */
       else if (m > (int) ((1L << cinfo->arith_dc_U[tbl]) >> 1))
-	entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
+        entropy->dc_context[ci] = 12 + (sign * 4); /* large diff category */
       else
-	entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
+        entropy->dc_context[ci] = 4 + (sign * 4);  /* small diff category */
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
       entropy->last_dc_val[ci] += v;
     }
 
-    (*block)[0] = (JCOEF) entropy->last_dc_val[ci];
+    if (block)
+      (*block)[0] = (JCOEF) entropy->last_dc_val[ci];
 
     /* Sections F.2.4.2 & F.1.4.4.2: Decoding of AC coefficients */
 
@@ -570,14 +574,14 @@
     /* Figure F.20: Decode_AC_coefficients */
     for (k = 1; k <= DCTSIZE2 - 1; k++) {
       st = entropy->ac_stats[tbl] + 3 * (k - 1);
-      if (arith_decode(cinfo, st)) break;	/* EOB flag */
+      if (arith_decode(cinfo, st)) break;       /* EOB flag */
       while (arith_decode(cinfo, st + 1) == 0) {
-	st += 3; k++;
-	if (k > DCTSIZE2 - 1) {
-	  WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	  entropy->ct = -1;			/* spectral overflow */
-	  return TRUE;
-	}
+        st += 3; k++;
+        if (k > DCTSIZE2 - 1) {
+          WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+          entropy->ct = -1;                     /* spectral overflow */
+          return TRUE;
+        }
       }
       /* Figure F.21: Decoding nonzero value v */
       /* Figure F.22: Decoding the sign of v */
@@ -585,27 +589,28 @@
       st += 2;
       /* Figure F.23: Decoding the magnitude category of v */
       if ((m = arith_decode(cinfo, st)) != 0) {
-	if (arith_decode(cinfo, st)) {
-	  m <<= 1;
-	  st = entropy->ac_stats[tbl] +
-	       (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
-	  while (arith_decode(cinfo, st)) {
-	    if ((m <<= 1) == 0x8000) {
-	      WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
-	      entropy->ct = -1;			/* magnitude overflow */
-	      return TRUE;
-	    }
-	    st += 1;
-	  }
-	}
+        if (arith_decode(cinfo, st)) {
+          m <<= 1;
+          st = entropy->ac_stats[tbl] +
+               (k <= cinfo->arith_ac_K[tbl] ? 189 : 217);
+          while (arith_decode(cinfo, st)) {
+            if ((m <<= 1) == 0x8000) {
+              WARNMS(cinfo, JWRN_ARITH_BAD_CODE);
+              entropy->ct = -1;                 /* magnitude overflow */
+              return TRUE;
+            }
+            st += 1;
+          }
+        }
       }
       v = m;
       /* Figure F.24: Decoding the magnitude bit pattern of v */
       st += 14;
       while (m >>= 1)
-	if (arith_decode(cinfo, st)) v |= m;
+        if (arith_decode(cinfo, st)) v |= m;
       v += 1; if (sign) v = -v;
-      (*block)[jpeg_natural_order[k]] = (JCOEF) v;
+      if (block)
+        (*block)[jpeg_natural_order[k]] = (JCOEF) v;
     }
   }
 
@@ -622,30 +627,30 @@
 {
   arith_entropy_ptr entropy = (arith_entropy_ptr) cinfo->entropy;
   int ci, tbl;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   if (cinfo->progressive_mode) {
     /* Validate progressive scan parameters */
     if (cinfo->Ss == 0) {
       if (cinfo->Se != 0)
-	goto bad;
+        goto bad;
     } else {
       /* need not check Ss/Se < 0 since they came from unsigned bytes */
       if (cinfo->Se < cinfo->Ss || cinfo->Se > DCTSIZE2 - 1)
-	goto bad;
+        goto bad;
       /* AC scans may have only one component */
       if (cinfo->comps_in_scan != 1)
-	goto bad;
+        goto bad;
     }
     if (cinfo->Ah != 0) {
       /* Successive approximation refinement scan: must have Al = Ah-1. */
       if (cinfo->Ah-1 != cinfo->Al)
-	goto bad;
+        goto bad;
     }
-    if (cinfo->Al > 13) {	/* need not check for < 0 */
+    if (cinfo->Al > 13) {       /* need not check for < 0 */
       bad:
       ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
-	       cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+               cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
     }
     /* Update progression status, and verify that scan order is legal.
      * Note that inter-scan inconsistencies are treated as warnings
@@ -655,32 +660,32 @@
       int coefi, cindex = cinfo->cur_comp_info[ci]->component_index;
       int *coef_bit_ptr = & cinfo->coef_bits[cindex][0];
       if (cinfo->Ss && coef_bit_ptr[0] < 0) /* AC without prior DC scan */
-	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, 0);
       for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
-	int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
-	if (cinfo->Ah != expected)
-	  WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
-	coef_bit_ptr[coefi] = cinfo->Al;
+        int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
+        if (cinfo->Ah != expected)
+          WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+        coef_bit_ptr[coefi] = cinfo->Al;
       }
     }
     /* Select MCU decoding routine */
     if (cinfo->Ah == 0) {
       if (cinfo->Ss == 0)
-	entropy->pub.decode_mcu = decode_mcu_DC_first;
+        entropy->pub.decode_mcu = decode_mcu_DC_first;
       else
-	entropy->pub.decode_mcu = decode_mcu_AC_first;
+        entropy->pub.decode_mcu = decode_mcu_AC_first;
     } else {
       if (cinfo->Ss == 0)
-	entropy->pub.decode_mcu = decode_mcu_DC_refine;
+        entropy->pub.decode_mcu = decode_mcu_DC_refine;
       else
-	entropy->pub.decode_mcu = decode_mcu_AC_refine;
+        entropy->pub.decode_mcu = decode_mcu_AC_refine;
     }
   } else {
     /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
      * This ought to be an error condition, but we make it a warning.
      */
     if (cinfo->Ss != 0 || cinfo->Ah != 0 || cinfo->Al != 0 ||
-	(cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1))
+        (cinfo->Se < DCTSIZE2 && cinfo->Se != DCTSIZE2 - 1))
       WARNMS(cinfo, JWRN_NOT_SEQUENTIAL);
     /* Select MCU decoding routine */
     entropy->pub.decode_mcu = decode_mcu;
@@ -689,25 +694,25 @@
   /* Allocate & initialize requested statistics areas */
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
-    if (! cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
+    if (!cinfo->progressive_mode || (cinfo->Ss == 0 && cinfo->Ah == 0)) {
       tbl = compptr->dc_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->dc_stats[tbl] == NULL)
-	entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
+        entropy->dc_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, DC_STAT_BINS);
       MEMZERO(entropy->dc_stats[tbl], DC_STAT_BINS);
       /* Initialize DC predictions to 0 */
       entropy->last_dc_val[ci] = 0;
       entropy->dc_context[ci] = 0;
     }
-    if (! cinfo->progressive_mode || cinfo->Ss) {
+    if (!cinfo->progressive_mode || cinfo->Ss) {
       tbl = compptr->ac_tbl_no;
       if (tbl < 0 || tbl >= NUM_ARITH_TBLS)
-	ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
+        ERREXIT1(cinfo, JERR_NO_ARITH_TABLE, tbl);
       if (entropy->ac_stats[tbl] == NULL)
-	entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
+        entropy->ac_stats[tbl] = (unsigned char *) (*cinfo->mem->alloc_small)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, AC_STAT_BINS);
       MEMZERO(entropy->ac_stats[tbl], AC_STAT_BINS);
     }
   }
@@ -715,7 +720,7 @@
   /* Initialize arithmetic decoding variables */
   entropy->c = 0;
   entropy->a = 0;
-  entropy->ct = -16;	/* force reading 2 initial bytes to fill C */
+  entropy->ct = -16;    /* force reading 2 initial bytes to fill C */
 
   /* Initialize restart counter */
   entropy->restarts_to_go = cinfo->restart_interval;
@@ -734,7 +739,7 @@
 
   entropy = (arith_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(arith_entropy_decoder));
+                                sizeof(arith_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass;
 
@@ -752,10 +757,10 @@
     int *coef_bit_ptr, ci;
     cinfo->coef_bits = (int (*)[DCTSIZE2])
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  cinfo->num_components*DCTSIZE2*SIZEOF(int));
+                                  cinfo->num_components*DCTSIZE2*sizeof(int));
     coef_bit_ptr = & cinfo->coef_bits[0][0];
-    for (ci = 0; ci < cinfo->num_components; ci++) 
+    for (ci = 0; ci < cinfo->num_components; ci++)
       for (i = 0; i < DCTSIZE2; i++)
-	*coef_bit_ptr++ = -1;
+        *coef_bit_ptr++ = -1;
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatadst.c glmark2-2021.02/src/libjpeg-turbo/jdatadst.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatadst.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdatadst.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,283 @@
+/*
+ * jdatadst.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains compression data destination routines for the case of
+ * emitting JPEG data to memory or to a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different destination manager.
+ * IMPORTANT: we assume that fwrite() will correctly transcribe an array of
+ * JOCTETs into 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
+#endif
+
+
+/* Expanded data destination object for stdio output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  FILE *outfile;                /* target stream */
+  JOCTET *buffer;               /* start of buffer */
+} my_destination_mgr;
+
+typedef my_destination_mgr *my_dest_ptr;
+
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/* Expanded data destination object for memory output */
+
+typedef struct {
+  struct jpeg_destination_mgr pub; /* public fields */
+
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
+  size_t bufsize;
+} my_mem_destination_mgr;
+
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
+#endif
+
+
+/*
+ * Initialize destination --- called by jpeg_start_compress
+ * before any data is actually written.
+ */
+
+METHODDEF(void)
+init_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  /* Allocate the output buffer --- it will be released when done with image */
+  dest->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  OUTPUT_BUF_SIZE * sizeof(JOCTET));
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_destination (j_compress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Empty the output buffer --- called whenever buffer fills up.
+ *
+ * In typical applications, this should write the entire output buffer
+ * (ignoring the current state of next_output_byte & free_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been dumped.
+ *
+ * In applications that need to be able to suspend compression due to output
+ * overrun, a FALSE return indicates that the buffer cannot be emptied now.
+ * In this situation, the compressor will return to its caller (possibly with
+ * an indication that it has not accepted all the supplied scanlines).  The
+ * application should resume compression after it has made more room in the
+ * output buffer.  Note that there are substantial restrictions on the use of
+ * suspension --- see the documentation.
+ *
+ * When suspending, the compressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_output_byte & free_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point will be regenerated after resumption, so do not
+ * write it out when emptying the buffer externally.
+ */
+
+METHODDEF(boolean)
+empty_output_buffer (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+
+  if (JFWRITE(dest->outfile, dest->buffer, OUTPUT_BUF_SIZE) !=
+      (size_t) OUTPUT_BUF_SIZE)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  dest->pub.next_output_byte = dest->buffer;
+  dest->pub.free_in_buffer = OUTPUT_BUF_SIZE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+empty_mem_output_buffer (j_compress_ptr cinfo)
+{
+  size_t nextsize;
+  JOCTET *nextbuffer;
+  my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
+
+  /* Try to allocate new buffer with double size */
+  nextsize = dest->bufsize * 2;
+  nextbuffer = (JOCTET *) malloc(nextsize);
+
+  if (nextbuffer == NULL)
+    ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+
+  MEMCOPY(nextbuffer, dest->buffer, dest->bufsize);
+
+  if (dest->newbuffer != NULL)
+    free(dest->newbuffer);
+
+  dest->newbuffer = nextbuffer;
+
+  dest->pub.next_output_byte = nextbuffer + dest->bufsize;
+  dest->pub.free_in_buffer = dest->bufsize;
+
+  dest->buffer = nextbuffer;
+  dest->bufsize = nextsize;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Terminate destination --- called by jpeg_finish_compress
+ * after all data has been written.  Usually needs to flush buffer.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_destination (j_compress_ptr cinfo)
+{
+  my_dest_ptr dest = (my_dest_ptr) cinfo->dest;
+  size_t datacount = OUTPUT_BUF_SIZE - dest->pub.free_in_buffer;
+
+  /* Write any data remaining in the buffer */
+  if (datacount > 0) {
+    if (JFWRITE(dest->outfile, dest->buffer, datacount) != datacount)
+      ERREXIT(cinfo, JERR_FILE_WRITE);
+  }
+  fflush(dest->outfile);
+  /* Make sure we wrote the output file OK */
+  if (ferror(dest->outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+term_mem_destination (j_compress_ptr cinfo)
+{
+  my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
+
+  *dest->outbuffer = dest->buffer;
+  *dest->outsize = (unsigned long)(dest->bufsize - dest->pub.free_in_buffer);
+}
+#endif
+
+
+/*
+ * Prepare for output to a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing compression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_dest (j_compress_ptr cinfo, FILE *outfile)
+{
+  my_dest_ptr dest;
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same file without re-executing jpeg_stdio_dest.
+   * This makes it dangerous to use this manager and a different destination
+   * manager serially with the same JPEG object, because their private object
+   * sizes may be different.  Caveat programmer.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_destination_mgr));
+  }
+
+  dest = (my_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_destination;
+  dest->pub.empty_output_buffer = empty_output_buffer;
+  dest->pub.term_destination = term_destination;
+  dest->outfile = outfile;
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for output to a memory buffer.
+ * The caller may supply an own initial buffer with appropriate size.
+ * Otherwise, or when the actual data output exceeds the given size,
+ * the library adapts the buffer size as necessary.
+ * The standard library functions malloc/free are used for allocating
+ * larger memory, so the buffer is available to the application after
+ * finishing compression, and then the application is responsible for
+ * freeing the requested memory.
+ * Note:  An initial buffer supplied by the caller is expected to be
+ * managed by the application.  The library does not free such buffer
+ * when allocating a larger buffer.
+ */
+
+GLOBAL(void)
+jpeg_mem_dest (j_compress_ptr cinfo,
+               unsigned char **outbuffer, unsigned long *outsize)
+{
+  my_mem_dest_ptr dest;
+
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
+    ERREXIT(cinfo, JERR_BUFFER_SIZE);
+
+  /* The destination object is made permanent so that multiple JPEG images
+   * can be written to the same buffer without re-executing jpeg_mem_dest.
+   */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
+    cinfo->dest = (struct jpeg_destination_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_mem_destination_mgr));
+  }
+
+  dest = (my_mem_dest_ptr) cinfo->dest;
+  dest->pub.init_destination = init_mem_destination;
+  dest->pub.empty_output_buffer = empty_mem_output_buffer;
+  dest->pub.term_destination = term_mem_destination;
+  dest->outbuffer = outbuffer;
+  dest->outsize = outsize;
+  dest->newbuffer = NULL;
+
+  if (*outbuffer == NULL || *outsize == 0) {
+    /* Allocate initial buffer */
+    dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE);
+    if (dest->newbuffer == NULL)
+      ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
+    *outsize = OUTPUT_BUF_SIZE;
+  }
+
+  dest->pub.next_output_byte = dest->buffer = *outbuffer;
+  dest->pub.free_in_buffer = dest->bufsize = *outsize;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatadst-tj.c glmark2-2021.02/src/libjpeg-turbo/jdatadst-tj.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatadst-tj.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdatadst-tj.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
- * jdatadst.c
+ * jdatadst-tj.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2009-2012 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2014 D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains compression data destination routines for the case of
  * emitting JPEG data to memory or to a file (or any stdio stream).
@@ -20,13 +23,13 @@
 #include "jpeglib.h"
 #include "jerror.h"
 
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
-extern void * malloc JPP((size_t size));
-extern void free JPP((void *ptr));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
 #endif
 
 
-#define OUTPUT_BUF_SIZE  4096	/* choose an efficiently fwrite'able size */
+#define OUTPUT_BUF_SIZE  4096   /* choose an efficiently fwrite'able size */
 
 
 /* Expanded data destination object for memory output */
@@ -34,15 +37,15 @@
 typedef struct {
   struct jpeg_destination_mgr pub; /* public fields */
 
-  unsigned char ** outbuffer;	/* target buffer */
-  unsigned long * outsize;
-  unsigned char * newbuffer;	/* newly allocated buffer */
-  JOCTET * buffer;		/* start of buffer */
+  unsigned char **outbuffer;    /* target buffer */
+  unsigned long *outsize;
+  unsigned char *newbuffer;     /* newly allocated buffer */
+  JOCTET *buffer;               /* start of buffer */
   size_t bufsize;
   boolean alloc;
 } my_mem_destination_mgr;
 
-typedef my_mem_destination_mgr * my_mem_dest_ptr;
+typedef my_mem_destination_mgr *my_mem_dest_ptr;
 
 
 /*
@@ -84,14 +87,14 @@
 empty_mem_output_buffer (j_compress_ptr cinfo)
 {
   size_t nextsize;
-  JOCTET * nextbuffer;
+  JOCTET *nextbuffer;
   my_mem_dest_ptr dest = (my_mem_dest_ptr) cinfo->dest;
 
   if (!dest->alloc) ERREXIT(cinfo, JERR_BUFFER_SIZE);
 
   /* Try to allocate new buffer with double size */
   nextsize = dest->bufsize * 2;
-  nextbuffer = malloc(nextsize);
+  nextbuffer = (JOCTET *) malloc(nextsize);
 
   if (nextbuffer == NULL)
     ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
@@ -145,29 +148,33 @@
 
 GLOBAL(void)
 jpeg_mem_dest_tj (j_compress_ptr cinfo,
-	       unsigned char ** outbuffer, unsigned long * outsize,
-	       boolean alloc)
+               unsigned char **outbuffer, unsigned long *outsize,
+               boolean alloc)
 {
+  boolean reused = FALSE;
   my_mem_dest_ptr dest;
 
-  if (outbuffer == NULL || outsize == NULL)	/* sanity check */
+  if (outbuffer == NULL || outsize == NULL)     /* sanity check */
     ERREXIT(cinfo, JERR_BUFFER_SIZE);
 
   /* The destination object is made permanent so that multiple JPEG images
    * can be written to the same buffer without re-executing jpeg_mem_dest.
    */
-  if (cinfo->dest == NULL) {	/* first time for this JPEG object? */
+  if (cinfo->dest == NULL) {    /* first time for this JPEG object? */
     cinfo->dest = (struct jpeg_destination_mgr *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  SIZEOF(my_mem_destination_mgr));
+                                  sizeof(my_mem_destination_mgr));
     dest = (my_mem_dest_ptr) cinfo->dest;
     dest->newbuffer = NULL;
+    dest->buffer = NULL;
   }
 
   dest = (my_mem_dest_ptr) cinfo->dest;
   dest->pub.init_destination = init_mem_destination;
   dest->pub.empty_output_buffer = empty_mem_output_buffer;
   dest->pub.term_destination = term_mem_destination;
+  if (dest->buffer == *outbuffer && *outbuffer != NULL && alloc)
+    reused = TRUE;
   dest->outbuffer = outbuffer;
   dest->outsize = outsize;
   dest->alloc = alloc;
@@ -175,7 +182,7 @@
   if (*outbuffer == NULL || *outsize == 0) {
     if (alloc) {
       /* Allocate initial buffer */
-      dest->newbuffer = *outbuffer = malloc(OUTPUT_BUF_SIZE);
+      dest->newbuffer = *outbuffer = (unsigned char *) malloc(OUTPUT_BUF_SIZE);
       if (dest->newbuffer == NULL)
         ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 10);
       *outsize = OUTPUT_BUF_SIZE;
@@ -184,5 +191,7 @@
   }
 
   dest->pub.next_output_byte = dest->buffer = *outbuffer;
-  dest->pub.free_in_buffer = dest->bufsize = *outsize;
+  if (!reused)
+    dest->bufsize = *outsize;
+  dest->pub.free_in_buffer = dest->bufsize;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatasrc.c glmark2-2021.02/src/libjpeg-turbo/jdatasrc.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatasrc.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdatasrc.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,284 @@
+/*
+ * jdatasrc.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains decompression data source routines for the case of
+ * reading JPEG data from memory or from a file (or any stdio stream).
+ * While these routines are sufficient for most applications,
+ * some will want to use a different source manager.
+ * IMPORTANT: we assume that fread() will correctly transcribe an array of
+ * JOCTETs from 8-bit-wide elements on external storage.  If char is wider
+ * than 8 bits on your machine, you may need to do some tweaking.
+ */
+
+/* this is not a core library module, so it doesn't define JPEG_INTERNALS */
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jerror.h"
+
+
+/* Expanded data source object for stdio input */
+
+typedef struct {
+  struct jpeg_source_mgr pub;   /* public fields */
+
+  FILE *infile;                 /* source stream */
+  JOCTET *buffer;               /* start of buffer */
+  boolean start_of_file;        /* have we gotten any data yet? */
+} my_source_mgr;
+
+typedef my_source_mgr *my_src_ptr;
+
+#define INPUT_BUF_SIZE  4096    /* choose an efficiently fread'able size */
+
+
+/*
+ * Initialize source --- called by jpeg_read_header
+ * before any data is actually read.
+ */
+
+METHODDEF(void)
+init_source (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+
+  /* We reset the empty-input-file flag for each image,
+   * but we don't clear the input buffer.
+   * This is correct behavior for reading a series of images from one source.
+   */
+  src->start_of_file = TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(void)
+init_mem_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+#endif
+
+
+/*
+ * Fill the input buffer --- called whenever buffer is emptied.
+ *
+ * In typical applications, this should read fresh data into the buffer
+ * (ignoring the current state of next_input_byte & bytes_in_buffer),
+ * reset the pointer & count to the start of the buffer, and return TRUE
+ * indicating that the buffer has been reloaded.  It is not necessary to
+ * fill the buffer entirely, only to obtain at least one more byte.
+ *
+ * There is no such thing as an EOF return.  If the end of the file has been
+ * reached, the routine has a choice of ERREXIT() or inserting fake data into
+ * the buffer.  In most cases, generating a warning message and inserting a
+ * fake EOI marker is the best course of action --- this will allow the
+ * decompressor to output however much of the image is there.  However,
+ * the resulting error message is misleading if the real problem is an empty
+ * input file, so we handle that case specially.
+ *
+ * In applications that need to be able to suspend compression due to input
+ * not being available yet, a FALSE return indicates that no more data can be
+ * obtained right now, but more may be forthcoming later.  In this situation,
+ * the decompressor will return to its caller (with an indication of the
+ * number of scanlines it has read, if any).  The application should resume
+ * decompression after it has loaded more data into the input buffer.  Note
+ * that there are substantial restrictions on the use of suspension --- see
+ * the documentation.
+ *
+ * When suspending, the decompressor will back up to a convenient restart point
+ * (typically the start of the current MCU). next_input_byte & bytes_in_buffer
+ * indicate where the restart point will be if the current call returns FALSE.
+ * Data beyond this point must be rescanned after resumption, so move it to
+ * the front of the buffer rather than discarding it.
+ */
+
+METHODDEF(boolean)
+fill_input_buffer (j_decompress_ptr cinfo)
+{
+  my_src_ptr src = (my_src_ptr) cinfo->src;
+  size_t nbytes;
+
+  nbytes = JFREAD(src->infile, src->buffer, INPUT_BUF_SIZE);
+
+  if (nbytes <= 0) {
+    if (src->start_of_file)     /* Treat empty input file as fatal error */
+      ERREXIT(cinfo, JERR_INPUT_EMPTY);
+    WARNMS(cinfo, JWRN_JPEG_EOF);
+    /* Insert a fake EOI marker */
+    src->buffer[0] = (JOCTET) 0xFF;
+    src->buffer[1] = (JOCTET) JPEG_EOI;
+    nbytes = 2;
+  }
+
+  src->pub.next_input_byte = src->buffer;
+  src->pub.bytes_in_buffer = nbytes;
+  src->start_of_file = FALSE;
+
+  return TRUE;
+}
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+METHODDEF(boolean)
+fill_mem_input_buffer (j_decompress_ptr cinfo)
+{
+  static const JOCTET mybuffer[4] = {
+    (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0
+  };
+
+  /* The whole JPEG data is expected to reside in the supplied memory
+   * buffer, so any request for more data beyond the given buffer size
+   * is treated as an error.
+   */
+  WARNMS(cinfo, JWRN_JPEG_EOF);
+
+  /* Insert a fake EOI marker */
+
+  cinfo->src->next_input_byte = mybuffer;
+  cinfo->src->bytes_in_buffer = 2;
+
+  return TRUE;
+}
+#endif
+
+
+/*
+ * Skip data --- used to skip over a potentially large amount of
+ * uninteresting data (such as an APPn marker).
+ *
+ * Writers of suspendable-input applications must note that skip_input_data
+ * is not granted the right to give a suspension return.  If the skip extends
+ * beyond the data currently in the buffer, the buffer can be marked empty so
+ * that the next read will cause a fill_input_buffer call that can suspend.
+ * Arranging for additional bytes to be discarded before reloading the input
+ * buffer is the application writer's problem.
+ */
+
+METHODDEF(void)
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+{
+  struct jpeg_source_mgr *src = cinfo->src;
+
+  /* Just a dumb implementation for now.  Could use fseek() except
+   * it doesn't work on pipes.  Not clear that being smart is worth
+   * any trouble anyway --- large skips are infrequent.
+   */
+  if (num_bytes > 0) {
+    while (num_bytes > (long) src->bytes_in_buffer) {
+      num_bytes -= (long) src->bytes_in_buffer;
+      (void) (*src->fill_input_buffer) (cinfo);
+      /* note we assume that fill_input_buffer will never return FALSE,
+       * so suspension need not be handled.
+       */
+    }
+    src->next_input_byte += (size_t) num_bytes;
+    src->bytes_in_buffer -= (size_t) num_bytes;
+  }
+}
+
+
+/*
+ * An additional method that can be provided by data source modules is the
+ * resync_to_restart method for error recovery in the presence of RST markers.
+ * For the moment, this source module just uses the default resync method
+ * provided by the JPEG library.  That method assumes that no backtracking
+ * is possible.
+ */
+
+
+/*
+ * Terminate source --- called by jpeg_finish_decompress
+ * after all data has been read.  Often a no-op.
+ *
+ * NB: *not* called by jpeg_abort or jpeg_destroy; surrounding
+ * application must deal with any cleanup that should happen even
+ * for error exit.
+ */
+
+METHODDEF(void)
+term_source (j_decompress_ptr cinfo)
+{
+  /* no work necessary here */
+}
+
+
+/*
+ * Prepare for input from a stdio stream.
+ * The caller must have already opened the stream, and is responsible
+ * for closing it after finishing decompression.
+ */
+
+GLOBAL(void)
+jpeg_stdio_src (j_decompress_ptr cinfo, FILE *infile)
+{
+  my_src_ptr src;
+
+  /* The source object and input buffer are made permanent so that a series
+   * of JPEG images can be read from the same file by calling jpeg_stdio_src
+   * only before the first one.  (If we discarded the buffer at the end of
+   * one image, we'd likely lose the start of the next one.)
+   * This makes it unsafe to use this manager and a different source
+   * manager serially with the same JPEG object.  Caveat programmer.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(my_source_mgr));
+    src = (my_src_ptr) cinfo->src;
+    src->buffer = (JOCTET *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  INPUT_BUF_SIZE * sizeof(JOCTET));
+  }
+
+  src = (my_src_ptr) cinfo->src;
+  src->pub.init_source = init_source;
+  src->pub.fill_input_buffer = fill_input_buffer;
+  src->pub.skip_input_data = skip_input_data;
+  src->pub.resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->pub.term_source = term_source;
+  src->infile = infile;
+  src->pub.bytes_in_buffer = 0; /* forces fill_input_buffer on first read */
+  src->pub.next_input_byte = NULL; /* until buffer loaded */
+}
+
+
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
+/*
+ * Prepare for input from a supplied memory buffer.
+ * The buffer must contain the whole JPEG data.
+ */
+
+GLOBAL(void)
+jpeg_mem_src (j_decompress_ptr cinfo,
+              const unsigned char *inbuffer, unsigned long insize)
+{
+  struct jpeg_source_mgr *src;
+
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
+    ERREXIT(cinfo, JERR_INPUT_EMPTY);
+
+  /* The source object is made permanent so that a series of JPEG images
+   * can be read from the same buffer by calling jpeg_mem_src only before
+   * the first one.
+   */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
+    cinfo->src = (struct jpeg_source_mgr *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
+                                  sizeof(struct jpeg_source_mgr));
+  }
+
+  src = cinfo->src;
+  src->init_source = init_mem_source;
+  src->fill_input_buffer = fill_mem_input_buffer;
+  src->skip_input_data = skip_input_data;
+  src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
+  src->term_source = term_source;
+  src->bytes_in_buffer = (size_t) insize;
+  src->next_input_byte = (const JOCTET *) inbuffer;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatasrc-tj.c glmark2-2021.02/src/libjpeg-turbo/jdatasrc-tj.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdatasrc-tj.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdatasrc-tj.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
- * jdatasrc.c
+ * jdatasrc-tj.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Modified 2009-2010 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2009-2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains decompression data source routines for the case of
  * reading JPEG data from memory or from a file (or any stdio stream).
@@ -69,16 +72,17 @@
 METHODDEF(boolean)
 fill_mem_input_buffer (j_decompress_ptr cinfo)
 {
-  static JOCTET mybuffer[4];
+  static const JOCTET mybuffer[4] = {
+    (JOCTET) 0xFF, (JOCTET) JPEG_EOI, 0, 0
+  };
 
   /* The whole JPEG data is expected to reside in the supplied memory
    * buffer, so any request for more data beyond the given buffer size
    * is treated as an error.
    */
   WARNMS(cinfo, JWRN_JPEG_EOF);
+
   /* Insert a fake EOI marker */
-  mybuffer[0] = (JOCTET) 0xFF;
-  mybuffer[1] = (JOCTET) JPEG_EOI;
 
   cinfo->src->next_input_byte = mybuffer;
   cinfo->src->bytes_in_buffer = 2;
@@ -102,7 +106,7 @@
 METHODDEF(void)
 skip_input_data (j_decompress_ptr cinfo, long num_bytes)
 {
-  struct jpeg_source_mgr * src = cinfo->src;
+  struct jpeg_source_mgr *src = cinfo->src;
 
   /* Just a dumb implementation for now.  Could use fseek() except
    * it doesn't work on pipes.  Not clear that being smart is worth
@@ -154,21 +158,21 @@
 
 GLOBAL(void)
 jpeg_mem_src_tj (j_decompress_ptr cinfo,
-	      unsigned char * inbuffer, unsigned long insize)
+                 const unsigned char *inbuffer, unsigned long insize)
 {
-  struct jpeg_source_mgr * src;
+  struct jpeg_source_mgr *src;
 
-  if (inbuffer == NULL || insize == 0)	/* Treat empty input as fatal error */
+  if (inbuffer == NULL || insize == 0)  /* Treat empty input as fatal error */
     ERREXIT(cinfo, JERR_INPUT_EMPTY);
 
   /* The source object is made permanent so that a series of JPEG images
    * can be read from the same buffer by calling jpeg_mem_src only before
    * the first one.
    */
-  if (cinfo->src == NULL) {	/* first time for this JPEG object? */
+  if (cinfo->src == NULL) {     /* first time for this JPEG object? */
     cinfo->src = (struct jpeg_source_mgr *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				  SIZEOF(struct jpeg_source_mgr));
+                                  sizeof(struct jpeg_source_mgr));
   }
 
   src = cinfo->src;
@@ -178,5 +182,5 @@
   src->resync_to_restart = jpeg_resync_to_restart; /* use default method */
   src->term_source = term_source;
   src->bytes_in_buffer = (size_t) insize;
-  src->next_input_byte = (JOCTET *) inbuffer;
+  src->next_input_byte = (const JOCTET *) inbuffer;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcoefct.c glmark2-2021.02/src/libjpeg-turbo/jdcoefct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcoefct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcoefct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jdcoefct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the coefficient buffer controller for decompression.
  * This controller is the top level of the JPEG decompressor proper.
@@ -15,95 +19,25 @@
  * Also, the input side (only) is used when reading a file for transcoding.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
+#include "jdcoefct.h"
 #include "jpegcomp.h"
 
-/* Block smoothing is only applicable for progressive JPEG, so: */
-#ifndef D_PROGRESSIVE_SUPPORTED
-#undef BLOCK_SMOOTHING_SUPPORTED
-#endif
-
-/* Private buffer controller object */
-
-typedef struct {
-  struct jpeg_d_coef_controller pub; /* public fields */
-
-  /* These variables keep track of the current location of the input side. */
-  /* cinfo->input_iMCU_row is also used for this. */
-  JDIMENSION MCU_ctr;		/* counts MCUs processed in current row */
-  int MCU_vert_offset;		/* counts MCU rows within iMCU row */
-  int MCU_rows_per_iMCU_row;	/* number of such rows needed */
-
-  /* The output side's location is represented by cinfo->output_iMCU_row. */
-
-  /* In single-pass modes, it's sufficient to buffer just one MCU.
-   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
-   * and let the entropy decoder write into that workspace each time.
-   * (On 80x86, the workspace is FAR even though it's not really very big;
-   * this is to keep the module interfaces unchanged when a large coefficient
-   * buffer is necessary.)
-   * In multi-pass modes, this array points to the current MCU's blocks
-   * within the virtual arrays; it is used only by the input side.
-   */
-  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
-
-  /* Temporary workspace for one MCU */
-  JCOEF * workspace;
-
-#ifdef D_MULTISCAN_FILES_SUPPORTED
-  /* In multi-pass modes, we need a virtual block array for each component. */
-  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
-#endif
-
-#ifdef BLOCK_SMOOTHING_SUPPORTED
-  /* When doing block smoothing, we latch coefficient Al values here */
-  int * coef_bits_latch;
-#define SAVED_COEFS  6		/* we save coef_bits[0..5] */
-#endif
-} my_coef_controller;
-
-typedef my_coef_controller * my_coef_ptr;
 
 /* Forward declarations */
 METHODDEF(int) decompress_onepass
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #ifdef D_MULTISCAN_FILES_SUPPORTED
 METHODDEF(int) decompress_data
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #endif
 #ifdef BLOCK_SMOOTHING_SUPPORTED
-LOCAL(boolean) smoothing_ok JPP((j_decompress_ptr cinfo));
+LOCAL(boolean) smoothing_ok (j_decompress_ptr cinfo);
 METHODDEF(int) decompress_smooth_data
-	JPP((j_decompress_ptr cinfo, JSAMPIMAGE output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
 #endif
 
 
-LOCAL(void)
-start_iMCU_row (j_decompress_ptr cinfo)
-/* Reset within-iMCU-row counters for a new row (input side) */
-{
-  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-
-  /* In an interleaved scan, an MCU row is the same as an iMCU row.
-   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
-   * But at the bottom of the image, process only what's left.
-   */
-  if (cinfo->comps_in_scan > 1) {
-    coef->MCU_rows_per_iMCU_row = 1;
-  } else {
-    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
-      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
-    else
-      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
-  }
-
-  coef->MCU_ctr = 0;
-  coef->MCU_vert_offset = 0;
-}
-
-
 /*
  * Initialize for an input processing pass.
  */
@@ -152,7 +86,7 @@
 decompress_onepass (j_decompress_ptr cinfo, JSAMPIMAGE output_buf)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   JDIMENSION last_MCU_col = cinfo->MCUs_per_row - 1;
   JDIMENSION last_iMCU_row = cinfo->total_iMCU_rows - 1;
   int blkn, ci, xindex, yindex, yoffset, useful_width;
@@ -165,49 +99,57 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->MCU_ctr; MCU_col_num <= last_MCU_col;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Try to fetch an MCU.  Entropy decoder expects buffer to be zeroed. */
-      jzero_far((void FAR *) coef->MCU_buffer[0],
-		(size_t) (cinfo->blocks_in_MCU * SIZEOF(JBLOCK)));
+      jzero_far((void *) coef->MCU_buffer[0],
+                (size_t) (cinfo->blocks_in_MCU * sizeof(JBLOCK)));
       if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->MCU_ctr = MCU_col_num;
-	return JPEG_SUSPENDED;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
       }
-      /* Determine where data should go in output_buf and do the IDCT thing.
-       * We skip dummy blocks at the right and bottom edges (but blkn gets
-       * incremented past them!).  Note the inner loop relies on having
-       * allocated the MCU_buffer[] blocks sequentially.
+
+      /* Only perform the IDCT on blocks that are contained within the desired
+       * cropping region.
        */
-      blkn = 0;			/* index of current DCT block within MCU */
-      for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	/* Don't bother to IDCT an uninteresting component. */
-	if (! compptr->component_needed) {
-	  blkn += compptr->MCU_blocks;
-	  continue;
-	}
-	inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
-	useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
-						    : compptr->last_col_width;
-	output_ptr = output_buf[compptr->component_index] +
-	  yoffset * compptr->_DCT_scaled_size;
-	start_col = MCU_col_num * compptr->MCU_sample_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  if (cinfo->input_iMCU_row < last_iMCU_row ||
-	      yoffset+yindex < compptr->last_row_height) {
-	    output_col = start_col;
-	    for (xindex = 0; xindex < useful_width; xindex++) {
-	      (*inverse_DCT) (cinfo, compptr,
-			      (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
-			      output_ptr, output_col);
-	      output_col += compptr->_DCT_scaled_size;
-	    }
-	  }
-	  blkn += compptr->MCU_width;
-	  output_ptr += compptr->_DCT_scaled_size;
-	}
+      if (MCU_col_num >= cinfo->master->first_iMCU_col &&
+          MCU_col_num <= cinfo->master->last_iMCU_col) {
+        /* Determine where data should go in output_buf and do the IDCT thing.
+         * We skip dummy blocks at the right and bottom edges (but blkn gets
+         * incremented past them!).  Note the inner loop relies on having
+         * allocated the MCU_buffer[] blocks sequentially.
+         */
+        blkn = 0;                 /* index of current DCT block within MCU */
+        for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
+          compptr = cinfo->cur_comp_info[ci];
+          /* Don't bother to IDCT an uninteresting component. */
+          if (! compptr->component_needed) {
+            blkn += compptr->MCU_blocks;
+            continue;
+          }
+          inverse_DCT = cinfo->idct->inverse_DCT[compptr->component_index];
+          useful_width = (MCU_col_num < last_MCU_col) ? compptr->MCU_width
+                                                      : compptr->last_col_width;
+          output_ptr = output_buf[compptr->component_index] +
+            yoffset * compptr->_DCT_scaled_size;
+          start_col = (MCU_col_num - cinfo->master->first_iMCU_col) *
+              compptr->MCU_sample_width;
+          for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+            if (cinfo->input_iMCU_row < last_iMCU_row ||
+                yoffset+yindex < compptr->last_row_height) {
+              output_col = start_col;
+              for (xindex = 0; xindex < useful_width; xindex++) {
+                (*inverse_DCT) (cinfo, compptr,
+                                (JCOEFPTR) coef->MCU_buffer[blkn+xindex],
+                                output_ptr, output_col);
+                output_col += compptr->_DCT_scaled_size;
+              }
+            }
+            blkn += compptr->MCU_width;
+            output_ptr += compptr->_DCT_scaled_size;
+          }
+        }
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -232,7 +174,7 @@
 METHODDEF(int)
 dummy_consume_data (j_decompress_ptr cinfo)
 {
-  return JPEG_SUSPENDED;	/* Always indicate nothing was done */
+  return JPEG_SUSPENDED;        /* Always indicate nothing was done */
 }
 
 
@@ -249,7 +191,7 @@
 consume_data (j_decompress_ptr cinfo)
 {
   my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
-  JDIMENSION MCU_col_num;	/* index of current MCU within row */
+  JDIMENSION MCU_col_num;       /* index of current MCU within row */
   int blkn, ci, xindex, yindex, yoffset;
   JDIMENSION start_col;
   JBLOCKARRAY buffer[MAX_COMPS_IN_SCAN];
@@ -273,25 +215,25 @@
   for (yoffset = coef->MCU_vert_offset; yoffset < coef->MCU_rows_per_iMCU_row;
        yoffset++) {
     for (MCU_col_num = coef->MCU_ctr; MCU_col_num < cinfo->MCUs_per_row;
-	 MCU_col_num++) {
+         MCU_col_num++) {
       /* Construct list of pointers to DCT blocks belonging to this MCU */
-      blkn = 0;			/* index of current DCT block within MCU */
+      blkn = 0;                 /* index of current DCT block within MCU */
       for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
-	compptr = cinfo->cur_comp_info[ci];
-	start_col = MCU_col_num * compptr->MCU_width;
-	for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
-	  buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
-	  for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
-	    coef->MCU_buffer[blkn++] = buffer_ptr++;
-	  }
-	}
+        compptr = cinfo->cur_comp_info[ci];
+        start_col = MCU_col_num * compptr->MCU_width;
+        for (yindex = 0; yindex < compptr->MCU_height; yindex++) {
+          buffer_ptr = buffer[ci][yindex+yoffset] + start_col;
+          for (xindex = 0; xindex < compptr->MCU_width; xindex++) {
+            coef->MCU_buffer[blkn++] = buffer_ptr++;
+          }
+        }
       }
       /* Try to fetch the MCU. */
       if (! (*cinfo->entropy->decode_mcu) (cinfo, coef->MCU_buffer)) {
-	/* Suspension forced; update state counters and exit */
-	coef->MCU_vert_offset = yoffset;
-	coef->MCU_ctr = MCU_col_num;
-	return JPEG_SUSPENDED;
+        /* Suspension forced; update state counters and exit */
+        coef->MCU_vert_offset = yoffset;
+        coef->MCU_ctr = MCU_col_num;
+        return JPEG_SUSPENDED;
       }
     }
     /* Completed an MCU row, but perhaps not an iMCU row */
@@ -332,8 +274,8 @@
 
   /* Force some input to be done if we are getting ahead of the input. */
   while (cinfo->input_scan_number < cinfo->output_scan_number ||
-	 (cinfo->input_scan_number == cinfo->output_scan_number &&
-	  cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
+         (cinfo->input_scan_number == cinfo->output_scan_number &&
+          cinfo->input_iMCU_row <= cinfo->output_iMCU_row)) {
     if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
       return JPEG_SUSPENDED;
   }
@@ -361,13 +303,14 @@
     output_ptr = output_buf[ci];
     /* Loop over all DCT blocks to be processed. */
     for (block_row = 0; block_row < block_rows; block_row++) {
-      buffer_ptr = buffer[block_row];
+      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
       output_col = 0;
-      for (block_num = 0; block_num < compptr->width_in_blocks; block_num++) {
-	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
-			output_ptr, output_col);
-	buffer_ptr++;
-	output_col += compptr->_DCT_scaled_size;
+      for (block_num = cinfo->master->first_MCU_col[ci];
+           block_num <= cinfo->master->last_MCU_col[ci]; block_num++) {
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) buffer_ptr,
+                        output_ptr, output_col);
+        buffer_ptr++;
+        output_col += compptr->_DCT_scaled_size;
       }
       output_ptr += compptr->_DCT_scaled_size;
     }
@@ -413,9 +356,9 @@
   boolean smoothing_useful = FALSE;
   int ci, coefi;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtable;
-  int * coef_bits;
-  int * coef_bits_latch;
+  JQUANT_TBL *qtable;
+  int *coef_bits;
+  int *coef_bits_latch;
 
   if (! cinfo->progressive_mode || cinfo->coef_bits == NULL)
     return FALSE;
@@ -424,8 +367,8 @@
   if (coef->coef_bits_latch == NULL)
     coef->coef_bits_latch = (int *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  cinfo->num_components *
-				  (SAVED_COEFS * SIZEOF(int)));
+                                  cinfo->num_components *
+                                  (SAVED_COEFS * sizeof(int)));
   coef_bits_latch = coef->coef_bits_latch;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -435,11 +378,11 @@
       return FALSE;
     /* Verify DC & first 5 AC quantizers are nonzero to avoid zero-divide. */
     if (qtable->quantval[0] == 0 ||
-	qtable->quantval[Q01_POS] == 0 ||
-	qtable->quantval[Q10_POS] == 0 ||
-	qtable->quantval[Q20_POS] == 0 ||
-	qtable->quantval[Q11_POS] == 0 ||
-	qtable->quantval[Q02_POS] == 0)
+        qtable->quantval[Q01_POS] == 0 ||
+        qtable->quantval[Q10_POS] == 0 ||
+        qtable->quantval[Q20_POS] == 0 ||
+        qtable->quantval[Q11_POS] == 0 ||
+        qtable->quantval[Q02_POS] == 0)
       return FALSE;
     /* DC values must be at least partly known for all components. */
     coef_bits = cinfo->coef_bits[ci];
@@ -449,7 +392,7 @@
     for (coefi = 1; coefi <= 5; coefi++) {
       coef_bits_latch[coefi] = coef_bits[coefi];
       if (coef_bits[coefi] != 0)
-	smoothing_useful = TRUE;
+        smoothing_useful = TRUE;
     }
     coef_bits_latch += SAVED_COEFS;
   }
@@ -476,10 +419,10 @@
   jpeg_component_info *compptr;
   inverse_DCT_method_ptr inverse_DCT;
   boolean first_row, last_row;
-  JCOEF * workspace;
+  JCOEF *workspace;
   int *coef_bits;
   JQUANT_TBL *quanttbl;
-  INT32 Q00,Q01,Q02,Q10,Q11,Q20, num;
+  JLONG Q00,Q01,Q02,Q10,Q11,Q20, num;
   int DC1,DC2,DC3,DC4,DC5,DC6,DC7,DC8,DC9;
   int Al, pred;
 
@@ -488,7 +431,7 @@
 
   /* Force some input to be done if we are getting ahead of the input. */
   while (cinfo->input_scan_number <= cinfo->output_scan_number &&
-	 ! cinfo->inputctl->eoi_reached) {
+         ! cinfo->inputctl->eoi_reached) {
     if (cinfo->input_scan_number == cinfo->output_scan_number) {
       /* If input is working on current scan, we ordinarily want it to
        * have completed the current row.  But if input scan is DC,
@@ -497,7 +440,7 @@
        */
       JDIMENSION delta = (cinfo->Ss == 0) ? 1 : 0;
       if (cinfo->input_iMCU_row > cinfo->output_iMCU_row+delta)
-	break;
+        break;
     }
     if ((*cinfo->inputctl->consume_input)(cinfo) == JPEG_SUSPENDED)
       return JPEG_SUSPENDED;
@@ -525,15 +468,15 @@
     if (cinfo->output_iMCU_row > 0) {
       access_rows += compptr->v_samp_factor; /* prior iMCU row too */
       buffer = (*cinfo->mem->access_virt_barray)
-	((j_common_ptr) cinfo, coef->whole_image[ci],
-	 (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
-	 (JDIMENSION) access_rows, FALSE);
-      buffer += compptr->v_samp_factor;	/* point to current iMCU row */
+        ((j_common_ptr) cinfo, coef->whole_image[ci],
+         (cinfo->output_iMCU_row - 1) * compptr->v_samp_factor,
+         (JDIMENSION) access_rows, FALSE);
+      buffer += compptr->v_samp_factor; /* point to current iMCU row */
       first_row = FALSE;
     } else {
       buffer = (*cinfo->mem->access_virt_barray)
-	((j_common_ptr) cinfo, coef->whole_image[ci],
-	 (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
+        ((j_common_ptr) cinfo, coef->whole_image[ci],
+         (JDIMENSION) 0, (JDIMENSION) access_rows, FALSE);
       first_row = TRUE;
     }
     /* Fetch component-dependent info */
@@ -549,15 +492,15 @@
     output_ptr = output_buf[ci];
     /* Loop over all DCT blocks to be processed. */
     for (block_row = 0; block_row < block_rows; block_row++) {
-      buffer_ptr = buffer[block_row];
+      buffer_ptr = buffer[block_row] + cinfo->master->first_MCU_col[ci];
       if (first_row && block_row == 0)
-	prev_block_row = buffer_ptr;
+        prev_block_row = buffer_ptr;
       else
-	prev_block_row = buffer[block_row-1];
+        prev_block_row = buffer[block_row-1];
       if (last_row && block_row == block_rows-1)
-	next_block_row = buffer_ptr;
+        next_block_row = buffer_ptr;
       else
-	next_block_row = buffer[block_row+1];
+        next_block_row = buffer[block_row+1];
       /* We fetch the surrounding DC values using a sliding-register approach.
        * Initialize all nine here so as to do the right thing on narrow pics.
        */
@@ -566,103 +509,104 @@
       DC7 = DC8 = DC9 = (int) next_block_row[0][0];
       output_col = 0;
       last_block_column = compptr->width_in_blocks - 1;
-      for (block_num = 0; block_num <= last_block_column; block_num++) {
-	/* Fetch current DCT block into workspace so we can modify it. */
-	jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
-	/* Update DC values */
-	if (block_num < last_block_column) {
-	  DC3 = (int) prev_block_row[1][0];
-	  DC6 = (int) buffer_ptr[1][0];
-	  DC9 = (int) next_block_row[1][0];
-	}
-	/* Compute coefficient estimates per K.8.
-	 * An estimate is applied only if coefficient is still zero,
-	 * and is not known to be fully accurate.
-	 */
-	/* AC01 */
-	if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
-	  num = 36 * Q00 * (DC4 - DC6);
-	  if (num >= 0) {
-	    pred = (int) (((Q01<<7) + num) / (Q01<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q01<<7) - num) / (Q01<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[1] = (JCOEF) pred;
-	}
-	/* AC10 */
-	if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
-	  num = 36 * Q00 * (DC2 - DC8);
-	  if (num >= 0) {
-	    pred = (int) (((Q10<<7) + num) / (Q10<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q10<<7) - num) / (Q10<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[8] = (JCOEF) pred;
-	}
-	/* AC20 */
-	if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
-	  num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
-	  if (num >= 0) {
-	    pred = (int) (((Q20<<7) + num) / (Q20<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q20<<7) - num) / (Q20<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[16] = (JCOEF) pred;
-	}
-	/* AC11 */
-	if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
-	  num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
-	  if (num >= 0) {
-	    pred = (int) (((Q11<<7) + num) / (Q11<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q11<<7) - num) / (Q11<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[9] = (JCOEF) pred;
-	}
-	/* AC02 */
-	if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
-	  num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
-	  if (num >= 0) {
-	    pred = (int) (((Q02<<7) + num) / (Q02<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	  } else {
-	    pred = (int) (((Q02<<7) - num) / (Q02<<8));
-	    if (Al > 0 && pred >= (1<<Al))
-	      pred = (1<<Al)-1;
-	    pred = -pred;
-	  }
-	  workspace[2] = (JCOEF) pred;
-	}
-	/* OK, do the IDCT */
-	(*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
-			output_ptr, output_col);
-	/* Advance for next column */
-	DC1 = DC2; DC2 = DC3;
-	DC4 = DC5; DC5 = DC6;
-	DC7 = DC8; DC8 = DC9;
-	buffer_ptr++, prev_block_row++, next_block_row++;
-	output_col += compptr->_DCT_scaled_size;
+      for (block_num = cinfo->master->first_MCU_col[ci];
+           block_num <= cinfo->master->last_MCU_col[ci]; block_num++) {
+        /* Fetch current DCT block into workspace so we can modify it. */
+        jcopy_block_row(buffer_ptr, (JBLOCKROW) workspace, (JDIMENSION) 1);
+        /* Update DC values */
+        if (block_num < last_block_column) {
+          DC3 = (int) prev_block_row[1][0];
+          DC6 = (int) buffer_ptr[1][0];
+          DC9 = (int) next_block_row[1][0];
+        }
+        /* Compute coefficient estimates per K.8.
+         * An estimate is applied only if coefficient is still zero,
+         * and is not known to be fully accurate.
+         */
+        /* AC01 */
+        if ((Al=coef_bits[1]) != 0 && workspace[1] == 0) {
+          num = 36 * Q00 * (DC4 - DC6);
+          if (num >= 0) {
+            pred = (int) (((Q01<<7) + num) / (Q01<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q01<<7) - num) / (Q01<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[1] = (JCOEF) pred;
+        }
+        /* AC10 */
+        if ((Al=coef_bits[2]) != 0 && workspace[8] == 0) {
+          num = 36 * Q00 * (DC2 - DC8);
+          if (num >= 0) {
+            pred = (int) (((Q10<<7) + num) / (Q10<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q10<<7) - num) / (Q10<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[8] = (JCOEF) pred;
+        }
+        /* AC20 */
+        if ((Al=coef_bits[3]) != 0 && workspace[16] == 0) {
+          num = 9 * Q00 * (DC2 + DC8 - 2*DC5);
+          if (num >= 0) {
+            pred = (int) (((Q20<<7) + num) / (Q20<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q20<<7) - num) / (Q20<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[16] = (JCOEF) pred;
+        }
+        /* AC11 */
+        if ((Al=coef_bits[4]) != 0 && workspace[9] == 0) {
+          num = 5 * Q00 * (DC1 - DC3 - DC7 + DC9);
+          if (num >= 0) {
+            pred = (int) (((Q11<<7) + num) / (Q11<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q11<<7) - num) / (Q11<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[9] = (JCOEF) pred;
+        }
+        /* AC02 */
+        if ((Al=coef_bits[5]) != 0 && workspace[2] == 0) {
+          num = 9 * Q00 * (DC4 + DC6 - 2*DC5);
+          if (num >= 0) {
+            pred = (int) (((Q02<<7) + num) / (Q02<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+          } else {
+            pred = (int) (((Q02<<7) - num) / (Q02<<8));
+            if (Al > 0 && pred >= (1<<Al))
+              pred = (1<<Al)-1;
+            pred = -pred;
+          }
+          workspace[2] = (JCOEF) pred;
+        }
+        /* OK, do the IDCT */
+        (*inverse_DCT) (cinfo, compptr, (JCOEFPTR) workspace,
+                        output_ptr, output_col);
+        /* Advance for next column */
+        DC1 = DC2; DC2 = DC3;
+        DC4 = DC5; DC5 = DC6;
+        DC7 = DC8; DC8 = DC9;
+        buffer_ptr++, prev_block_row++, next_block_row++;
+        output_col += compptr->_DCT_scaled_size;
       }
       output_ptr += compptr->_DCT_scaled_size;
     }
@@ -687,7 +631,7 @@
 
   coef = (my_coef_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_coef_controller));
+                                sizeof(my_coef_controller));
   cinfo->coef = (struct jpeg_d_coef_controller *) coef;
   coef->pub.start_input_pass = start_input_pass;
   coef->pub.start_output_pass = start_output_pass;
@@ -705,20 +649,20 @@
     jpeg_component_info *compptr;
 
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       access_rows = compptr->v_samp_factor;
 #ifdef BLOCK_SMOOTHING_SUPPORTED
       /* If block smoothing could be used, need a bigger window */
       if (cinfo->progressive_mode)
-	access_rows *= 3;
+        access_rows *= 3;
 #endif
       coef->whole_image[ci] = (*cinfo->mem->request_virt_barray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
-	 (JDIMENSION) jround_up((long) compptr->width_in_blocks,
-				(long) compptr->h_samp_factor),
-	 (JDIMENSION) jround_up((long) compptr->height_in_blocks,
-				(long) compptr->v_samp_factor),
-	 (JDIMENSION) access_rows);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, TRUE,
+         (JDIMENSION) jround_up((long) compptr->width_in_blocks,
+                                (long) compptr->h_samp_factor),
+         (JDIMENSION) jround_up((long) compptr->height_in_blocks,
+                                (long) compptr->v_samp_factor),
+         (JDIMENSION) access_rows);
     }
     coef->pub.consume_data = consume_data;
     coef->pub.decompress_data = decompress_data;
@@ -733,7 +677,7 @@
 
     buffer = (JBLOCKROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  D_MAX_BLOCKS_IN_MCU * SIZEOF(JBLOCK));
+                                  D_MAX_BLOCKS_IN_MCU * sizeof(JBLOCK));
     for (i = 0; i < D_MAX_BLOCKS_IN_MCU; i++) {
       coef->MCU_buffer[i] = buffer + i;
     }
@@ -745,5 +689,5 @@
   /* Allocate the workspace buffer */
   coef->workspace = (JCOEF *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-                                SIZEOF(JCOEF) * DCTSIZE2);
+                                sizeof(JCOEF) * DCTSIZE2);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcoefct.h glmark2-2021.02/src/libjpeg-turbo/jdcoefct.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcoefct.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcoefct.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,82 @@
+/*
+ * jdcoefct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Block smoothing is only applicable for progressive JPEG, so: */
+#ifndef D_PROGRESSIVE_SUPPORTED
+#undef BLOCK_SMOOTHING_SUPPORTED
+#endif
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_coef_controller pub; /* public fields */
+
+  /* These variables keep track of the current location of the input side. */
+  /* cinfo->input_iMCU_row is also used for this. */
+  JDIMENSION MCU_ctr;           /* counts MCUs processed in current row */
+  int MCU_vert_offset;          /* counts MCU rows within iMCU row */
+  int MCU_rows_per_iMCU_row;    /* number of such rows needed */
+
+  /* The output side's location is represented by cinfo->output_iMCU_row. */
+
+  /* In single-pass modes, it's sufficient to buffer just one MCU.
+   * We allocate a workspace of D_MAX_BLOCKS_IN_MCU coefficient blocks,
+   * and let the entropy decoder write into that workspace each time.
+   * In multi-pass modes, this array points to the current MCU's blocks
+   * within the virtual arrays; it is used only by the input side.
+   */
+  JBLOCKROW MCU_buffer[D_MAX_BLOCKS_IN_MCU];
+
+  /* Temporary workspace for one MCU */
+  JCOEF *workspace;
+
+#ifdef D_MULTISCAN_FILES_SUPPORTED
+  /* In multi-pass modes, we need a virtual block array for each component. */
+  jvirt_barray_ptr whole_image[MAX_COMPONENTS];
+#endif
+
+#ifdef BLOCK_SMOOTHING_SUPPORTED
+  /* When doing block smoothing, we latch coefficient Al values here */
+  int *coef_bits_latch;
+#define SAVED_COEFS  6          /* we save coef_bits[0..5] */
+#endif
+} my_coef_controller;
+
+typedef my_coef_controller *my_coef_ptr;
+
+
+LOCAL(void)
+start_iMCU_row (j_decompress_ptr cinfo)
+/* Reset within-iMCU-row counters for a new row (input side) */
+{
+  my_coef_ptr coef = (my_coef_ptr) cinfo->coef;
+
+  /* In an interleaved scan, an MCU row is the same as an iMCU row.
+   * In a noninterleaved scan, an iMCU row has v_samp_factor MCU rows.
+   * But at the bottom of the image, process only what's left.
+   */
+  if (cinfo->comps_in_scan > 1) {
+    coef->MCU_rows_per_iMCU_row = 1;
+  } else {
+    if (cinfo->input_iMCU_row < (cinfo->total_iMCU_rows-1))
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->v_samp_factor;
+    else
+      coef->MCU_rows_per_iMCU_row = cinfo->cur_comp_info[0]->last_row_height;
+  }
+
+  coef->MCU_ctr = 0;
+  coef->MCU_vert_offset = 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcol565.c glmark2-2021.02/src/libjpeg-turbo/jdcol565.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcol565.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcol565.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,384 @@
+/*
+ * jdcol565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+/* This file is included by jdcolor.c */
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565_convert_internal (j_decompress_ptr cinfo,
+                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+                             JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[y + Crrtab[cr]];
+      g = range_limit[y + ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                            SCALEBITS))];
+      b = range_limit[y + Cbbtab[cb]];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+ycc_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      y  = GETJSAMPLE(*inptr0++);
+      cb = GETJSAMPLE(*inptr1++);
+      cr = GETJSAMPLE(*inptr2++);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      y  = GETJSAMPLE(*inptr0);
+      cb = GETJSAMPLE(*inptr1);
+      cr = GETJSAMPLE(*inptr2);
+      r = range_limit[DITHER_565_R(y + Crrtab[cr], d0)];
+      g = range_limit[DITHER_565_G(y +
+                                   ((int)RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                     SCALEBITS)), d0)];
+      b = range_limit[DITHER_565_B(y + Cbbtab[cb], d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565_convert_internal (j_decompress_ptr cinfo,
+                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+                             JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = GETJSAMPLE(*inptr0++);
+      g = GETJSAMPLE(*inptr1++);
+      b = GETJSAMPLE(*inptr2++);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = GETJSAMPLE(*inptr0);
+      g = GETJSAMPLE(*inptr1);
+      b = GETJSAMPLE(*inptr2);
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+rgb_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int r, g, b;
+
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_SHORT_565(r, g, b);
+
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0++), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1++), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2++), d0)];
+      d0 = DITHER_ROTATE(d0);
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      r = range_limit[DITHER_565_R(GETJSAMPLE(*inptr0), d0)];
+      g = range_limit[DITHER_565_G(GETJSAMPLE(*inptr1), d0)];
+      b = range_limit[DITHER_565_B(GETJSAMPLE(*inptr2), d0)];
+      rgb = PACK_SHORT_565(r, g, b);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565_convert_internal (j_decompress_ptr cinfo,
+                              JSAMPIMAGE input_buf, JDIMENSION input_row,
+                              JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      rgb = PACK_SHORT_565(g, g, g);
+      g = *inptr++;
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
+
+
+INLINE
+LOCAL(void)
+gray_rgb565D_convert_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf, JDIMENSION input_row,
+                               JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  JDIMENSION num_cols = cinfo->output_width;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+
+  while (--num_rows >= 0) {
+    JLONG rgb;
+    unsigned int g;
+
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    if (PACK_NEED_ALIGNMENT(outptr)) {
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+      outptr += 2;
+      num_cols--;
+    }
+    for (col = 0; col < (num_cols >> 1); col++) {
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      d0 = DITHER_ROTATE(d0);
+
+      g = *inptr++;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(g, g, g));
+      d0 = DITHER_ROTATE(d0);
+
+      WRITE_TWO_ALIGNED_PIXELS(outptr, rgb);
+      outptr += 4;
+    }
+    if (num_cols & 1) {
+      g = *inptr;
+      g = range_limit[DITHER_565_R(g, d0)];
+      rgb = PACK_SHORT_565(g, g, g);
+      *(INT16*)outptr = (INT16)rgb;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolext.c glmark2-2021.02/src/libjpeg-turbo/jdcolext.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcolext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,143 @@
+/*
+ * jdcolext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+
+/* This file is included by jdcolor.c */
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+ycc_rgb_convert_internal (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register JLONG * Crgtab = cconvert->Cr_g_tab;
+  register JLONG * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+INLINE
+LOCAL(void)
+gray_rgb_convert_internal (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION input_row,
+                           JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert RGB to extended RGB: just swap the order of source pixels
+ */
+
+INLINE
+LOCAL(void)
+rgb_rgb_convert_internal (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = inptr0[col];
+      outptr[RGB_GREEN] = inptr1[col];
+      outptr[RGB_BLUE] = inptr2[col];
+      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
+      /* alpha channel value */
+#ifdef RGB_ALPHA
+      outptr[RGB_ALPHA] = 0xFF;
+#endif
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolext.c.inc glmark2-2021.02/src/libjpeg-turbo/jdcolext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcolext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,104 +0,0 @@
-/*
- * jdcolext.c
- *
- * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains output colorspace conversion routines.
- */
-
-
-/* This file is included by jdcolor.c */
-
-
-/*
- * Convert some rows of samples to the output colorspace.
- *
- * Note that we change from noninterleaved, one-plane-per-component format
- * to interleaved-pixel format.  The output buffer is therefore three times
- * as wide as the input buffer.
- * A starting row offset is provided only for the input buffer.  The caller
- * can easily adjust the passed output_buf value to accommodate any row
- * offset required on that side.
- */
-
-INLINE
-LOCAL(void)
-ycc_rgb_convert_internal (j_decompress_ptr cinfo,
-                          JSAMPIMAGE input_buf, JDIMENSION input_row,
-                          JSAMPARRAY output_buf, int num_rows)
-{
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int y, cb, cr;
-  register JSAMPROW outptr;
-  register JSAMPROW inptr0, inptr1, inptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->output_width;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  register int * Crrtab = cconvert->Cr_r_tab;
-  register int * Cbbtab = cconvert->Cb_b_tab;
-  register INT32 * Crgtab = cconvert->Cr_g_tab;
-  register INT32 * Cbgtab = cconvert->Cb_g_tab;
-  SHIFT_TEMPS
-
-  while (--num_rows >= 0) {
-    inptr0 = input_buf[0][input_row];
-    inptr1 = input_buf[1][input_row];
-    inptr2 = input_buf[2][input_row];
-    input_row++;
-    outptr = *output_buf++;
-    for (col = 0; col < num_cols; col++) {
-      y  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
-      /* Range-limiting is essential due to noise introduced by DCT losses. */
-      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
-      outptr[RGB_GREEN] = range_limit[y +
-			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
-						 SCALEBITS))];
-      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
-      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
-      /* alpha channel value */
-#ifdef RGB_ALPHA
-      outptr[RGB_ALPHA] = 0xFF;
-#endif
-      outptr += RGB_PIXELSIZE;
-    }
-  }
-}
-
-
-/*
- * Convert grayscale to RGB: just duplicate the graylevel three times.
- * This is provided to support applications that don't want to cope
- * with grayscale as a separate case.
- */
-
-INLINE
-LOCAL(void)
-gray_rgb_convert_internal (j_decompress_ptr cinfo,
-                           JSAMPIMAGE input_buf, JDIMENSION input_row,
-                           JSAMPARRAY output_buf, int num_rows)
-{
-  register JSAMPROW inptr, outptr;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->output_width;
-
-  while (--num_rows >= 0) {
-    inptr = input_buf[0][input_row++];
-    outptr = *output_buf++;
-    for (col = 0; col < num_cols; col++) {
-      /* We can dispense with GETJSAMPLE() here */
-      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
-      /* Set unused byte to 0xFF so it can be interpreted as an opaque */
-      /* alpha channel value */
-#ifdef RGB_ALPHA
-      outptr[RGB_ALPHA] = 0xFF;
-#endif
-      outptr += RGB_PIXELSIZE;
-    }
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolor.c glmark2-2021.02/src/libjpeg-turbo/jdcolor.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdcolor.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdcolor.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jdcolor.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2011 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009, 2011-2012, 2014-2015, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains output colorspace conversion routines.
  */
@@ -14,7 +18,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 
 /* Private subobject */
@@ -23,24 +27,32 @@
   struct jpeg_color_deconverter pub; /* public fields */
 
   /* Private state for YCC->RGB conversion */
-  int * Cr_r_tab;		/* => table for Cr to R conversion */
-  int * Cb_b_tab;		/* => table for Cb to B conversion */
-  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
-  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
+
+  /* Private state for RGB->Y conversion */
+  JLONG *rgb_y_tab;             /* => table for RGB to Y conversion */
 } my_color_deconverter;
 
-typedef my_color_deconverter * my_cconvert_ptr;
+typedef my_color_deconverter *my_cconvert_ptr;
 
 
 /**************** YCbCr -> RGB conversion: most common case **************/
+/****************   RGB -> Y   conversion: less common case **************/
 
 /*
  * YCbCr is defined per CCIR 601-1, except that Cb and Cr are
  * normalized to the range 0..MAXJSAMPLE rather than -0.5 .. 0.5.
  * The conversion equations to be implemented are therefore
- *	R = Y                + 1.40200 * Cr
- *	G = Y - 0.34414 * Cb - 0.71414 * Cr
- *	B = Y + 1.77200 * Cb
+ *
+ *      R = Y                + 1.40200 * Cr
+ *      G = Y - 0.34414 * Cb - 0.71414 * Cr
+ *      B = Y + 1.77200 * Cb
+ *
+ *      Y = 0.29900 * R + 0.58700 * G + 0.11400 * B
+ *
  * where Cb and Cr represent the incoming values less CENTERJSAMPLE.
  * (These numbers are derived from TIFF 6.0 section 21, dated 3-June-92.)
  *
@@ -61,14 +73,26 @@
  * together before rounding.
  */
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
+
+/* We allocate one big table for RGB->Y conversion and divide it up into
+ * three parts, instead of doing three alloc_small requests.  This lets us
+ * use a single table base address, which can be held in a register in the
+ * inner loops on many machines (more than can hold all three addresses,
+ * anyway).
+ */
+
+#define R_Y_OFF         0                       /* offset to R => Y section */
+#define G_Y_OFF         (1*(MAXJSAMPLE+1))      /* offset to G => Y section */
+#define B_Y_OFF         (2*(MAXJSAMPLE+1))      /* etc. */
+#define TABLE_SIZE      (3*(MAXJSAMPLE+1))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jdcolext.c.inc"
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -80,13 +104,15 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extrgb_convert_internal
 #define gray_rgb_convert_internal gray_extrgb_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extrgb_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
@@ -95,7 +121,8 @@
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extrgbx_convert_internal
 #define gray_rgb_convert_internal gray_extrgbx_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extrgbx_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -103,6 +130,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGR_RED
 #define RGB_GREEN EXT_BGR_GREEN
@@ -110,13 +138,15 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extbgr_convert_internal
 #define gray_rgb_convert_internal gray_extbgr_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extbgr_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
@@ -125,7 +155,8 @@
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extbgrx_convert_internal
 #define gray_rgb_convert_internal gray_extbgrx_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extbgrx_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -133,6 +164,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
@@ -141,7 +173,8 @@
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extxbgr_convert_internal
 #define gray_rgb_convert_internal gray_extxbgr_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extxbgr_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -149,6 +182,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
@@ -157,7 +191,8 @@
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define ycc_rgb_convert_internal ycc_extxrgb_convert_internal
 #define gray_rgb_convert_internal gray_extxrgb_convert_internal
-#include "jdcolext.c.inc"
+#define rgb_rgb_convert_internal rgb_extxrgb_convert_internal
+#include "jdcolext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -165,6 +200,7 @@
 #undef RGB_PIXELSIZE
 #undef ycc_rgb_convert_internal
 #undef gray_rgb_convert_internal
+#undef rgb_rgb_convert_internal
 
 
 /*
@@ -176,31 +212,31 @@
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   int i;
-  INT32 x;
+  JLONG x;
   SHIFT_TEMPS
 
   cconvert->Cr_r_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
+                                (MAXJSAMPLE+1) * sizeof(int));
   cconvert->Cb_b_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
-  cconvert->Cr_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(int));
+  cconvert->Cr_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
-  cconvert->Cb_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
+  cconvert->Cb_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
 
   for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
     /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
     /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
     /* Cr=>R value is nearest int to 1.40200 * x */
     cconvert->Cr_r_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
     /* Cb=>B value is nearest int to 1.77200 * x */
     cconvert->Cb_b_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
     /* Cr=>G value is scaled-up -0.71414 * x */
     cconvert->Cr_g_tab[i] = (- FIX(0.71414)) * x;
     /* Cb=>G value is scaled-up -0.34414 * x */
@@ -216,8 +252,8 @@
 
 METHODDEF(void)
 ycc_rgb_convert (j_decompress_ptr cinfo,
-		 JSAMPIMAGE input_buf, JDIMENSION input_row,
-		 JSAMPARRAY output_buf, int num_rows)
+                 JSAMPIMAGE input_buf, JDIMENSION input_row,
+                 JSAMPARRAY output_buf, int num_rows)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -260,32 +296,122 @@
 
 
 /*
+ * Initialize for RGB->grayscale colorspace conversion.
+ */
+
+LOCAL(void)
+build_rgb_y_table (j_decompress_ptr cinfo)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  JLONG *rgb_y_tab;
+  JLONG i;
+
+  /* Allocate and fill in the conversion tables. */
+  cconvert->rgb_y_tab = rgb_y_tab = (JLONG *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (TABLE_SIZE * sizeof(JLONG)));
+
+  for (i = 0; i <= MAXJSAMPLE; i++) {
+    rgb_y_tab[i+R_Y_OFF] = FIX(0.29900) * i;
+    rgb_y_tab[i+G_Y_OFF] = FIX(0.58700) * i;
+    rgb_y_tab[i+B_Y_OFF] = FIX(0.11400) * i + ONE_HALF;
+  }
+}
+
+
+/*
+ * Convert RGB to grayscale.
+ */
+
+METHODDEF(void)
+rgb_gray_convert (j_decompress_ptr cinfo,
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register JLONG *ctab = cconvert->rgb_y_tab;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr0[col]);
+      g = GETJSAMPLE(inptr1[col]);
+      b = GETJSAMPLE(inptr2[col]);
+      /* Y */
+      outptr[col] = (JSAMPLE)
+                ((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+                 >> SCALEBITS);
+    }
+  }
+}
+
+
+/*
  * Color conversion for no colorspace change: just copy the data,
  * converting from separate-planes to interleaved representation.
  */
 
 METHODDEF(void)
 null_convert (j_decompress_ptr cinfo,
-	      JSAMPIMAGE input_buf, JDIMENSION input_row,
-	      JSAMPARRAY output_buf, int num_rows)
+              JSAMPIMAGE input_buf, JDIMENSION input_row,
+              JSAMPARRAY output_buf, int num_rows)
 {
-  register JSAMPROW inptr, outptr;
-  register JDIMENSION count;
+  register JSAMPROW inptr, inptr0, inptr1, inptr2, inptr3, outptr;
+  register JDIMENSION col;
   register int num_components = cinfo->num_components;
   JDIMENSION num_cols = cinfo->output_width;
   int ci;
 
-  while (--num_rows >= 0) {
-    for (ci = 0; ci < num_components; ci++) {
-      inptr = input_buf[ci][input_row];
-      outptr = output_buf[0] + ci;
-      for (count = num_cols; count > 0; count--) {
-	*outptr = *inptr++;	/* needn't bother with GETJSAMPLE() here */
-	outptr += num_components;
+  if (num_components == 3) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf[0][input_row];
+      inptr1 = input_buf[1][input_row];
+      inptr2 = input_buf[2][input_row];
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0[col];
+        *outptr++ = inptr1[col];
+        *outptr++ = inptr2[col];
       }
     }
-    input_row++;
-    output_buf++;
+  } else if (num_components == 4) {
+    while (--num_rows >= 0) {
+      inptr0 = input_buf[0][input_row];
+      inptr1 = input_buf[1][input_row];
+      inptr2 = input_buf[2][input_row];
+      inptr3 = input_buf[3][input_row];
+      input_row++;
+      outptr = *output_buf++;
+      for (col = 0; col < num_cols; col++) {
+        *outptr++ = inptr0[col];
+        *outptr++ = inptr1[col];
+        *outptr++ = inptr2[col];
+        *outptr++ = inptr3[col];
+      }
+    }
+  } else {
+    while (--num_rows >= 0) {
+      for (ci = 0; ci < num_components; ci++) {
+        inptr = input_buf[ci][input_row];
+        outptr = *output_buf;
+        for (col = 0; col < num_cols; col++) {
+          outptr[ci] = inptr[col];
+          outptr += num_components;
+        }
+      }
+      output_buf++;
+      input_row++;
+    }
   }
 }
 
@@ -298,11 +424,11 @@
 
 METHODDEF(void)
 grayscale_convert (j_decompress_ptr cinfo,
-		   JSAMPIMAGE input_buf, JDIMENSION input_row,
-		   JSAMPARRAY output_buf, int num_rows)
+                   JSAMPIMAGE input_buf, JDIMENSION input_row,
+                   JSAMPARRAY output_buf, int num_rows)
 {
   jcopy_sample_rows(input_buf[0], (int) input_row, output_buf, 0,
-		    num_rows, cinfo->output_width);
+                    num_rows, cinfo->output_width);
 }
 
 
@@ -312,8 +438,8 @@
 
 METHODDEF(void)
 gray_rgb_convert (j_decompress_ptr cinfo,
-		  JSAMPIMAGE input_buf, JDIMENSION input_row,
-		  JSAMPARRAY output_buf, int num_rows)
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -353,6 +479,52 @@
 
 
 /*
+ * Convert plain RGB to extended RGB
+ */
+
+METHODDEF(void)
+rgb_rgb_convert (j_decompress_ptr cinfo,
+                  JSAMPIMAGE input_buf, JDIMENSION input_row,
+                  JSAMPARRAY output_buf, int num_rows)
+{
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      rgb_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      rgb_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      rgb_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      rgb_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      rgb_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      rgb_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    default:
+      rgb_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                               num_rows);
+      break;
+  }
+}
+
+
+/*
  * Adobe-style YCCK->CMYK conversion.
  * We convert YCbCr to R=1-C, G=1-M, and B=1-Y using the same
  * conversion as above, while passing K (black) unchanged.
@@ -361,8 +533,8 @@
 
 METHODDEF(void)
 ycck_cmyk_convert (j_decompress_ptr cinfo,
-		   JSAMPIMAGE input_buf, JDIMENSION input_row,
-		   JSAMPARRAY output_buf, int num_rows)
+                   JSAMPIMAGE input_buf, JDIMENSION input_row,
+                   JSAMPARRAY output_buf, int num_rows)
 {
   my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
   register int y, cb, cr;
@@ -371,11 +543,11 @@
   register JDIMENSION col;
   JDIMENSION num_cols = cinfo->output_width;
   /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  register int * Crrtab = cconvert->Cr_r_tab;
-  register int * Cbbtab = cconvert->Cb_b_tab;
-  register INT32 * Crgtab = cconvert->Cr_g_tab;
-  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  register JSAMPLE *range_limit = cinfo->sample_range_limit;
+  register int *Crrtab = cconvert->Cr_r_tab;
+  register int *Cbbtab = cconvert->Cb_b_tab;
+  register JLONG *Crgtab = cconvert->Cr_g_tab;
+  register JLONG *Cbgtab = cconvert->Cb_g_tab;
   SHIFT_TEMPS
 
   while (--num_rows >= 0) {
@@ -390,13 +562,13 @@
       cb = GETJSAMPLE(inptr1[col]);
       cr = GETJSAMPLE(inptr2[col]);
       /* Range-limiting is essential due to noise introduced by DCT losses. */
-      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];	/* red */
-      outptr[1] = range_limit[MAXJSAMPLE - (y +			/* green */
-			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
-						 SCALEBITS)))];
-      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];	/* blue */
+      outptr[0] = range_limit[MAXJSAMPLE - (y + Crrtab[cr])];   /* red */
+      outptr[1] = range_limit[MAXJSAMPLE - (y +                 /* green */
+                              ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+                                                 SCALEBITS)))];
+      outptr[2] = range_limit[MAXJSAMPLE - (y + Cbbtab[cb])];   /* blue */
       /* K passes through unchanged */
-      outptr[3] = inptr3[col];	/* don't need GETJSAMPLE here */
+      outptr[3] = inptr3[col];  /* don't need GETJSAMPLE here */
       outptr += 4;
     }
   }
@@ -404,6 +576,164 @@
 
 
 /*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b)   ((((r) << 8) & 0xF800) |  \
+                                      (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)   (((r) & 0xF8) | ((g) >> 5) |  \
+                                      (((g) << 11) & 0xE000) |  \
+                                      (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)     ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)     ((l << 16) | r)
+
+#define PACK_NEED_ALIGNMENT(ptr)     (((size_t)(ptr)) & 3)
+
+#define WRITE_TWO_ALIGNED_PIXELS(addr, pixels)  ((*(int *)(addr)) = pixels)
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficent for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565 PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE
+#define ycc_rgb565_convert_internal ycc_rgb565_convert_le
+#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_le
+#define rgb_rgb565_convert_internal rgb_rgb565_convert_le
+#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_le
+#define gray_rgb565_convert_internal gray_rgb565_convert_le
+#define gray_rgb565D_convert_internal gray_rgb565D_convert_le
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+#define PACK_SHORT_565 PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE
+#define ycc_rgb565_convert_internal ycc_rgb565_convert_be
+#define ycc_rgb565D_convert_internal ycc_rgb565D_convert_be
+#define rgb_rgb565_convert_internal rgb_rgb565_convert_be
+#define rgb_rgb565D_convert_internal rgb_rgb565D_convert_be
+#define gray_rgb565_convert_internal gray_rgb565_convert_be
+#define gray_rgb565D_convert_internal gray_rgb565D_convert_be
+#include "jdcol565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef ycc_rgb565_convert_internal
+#undef ycc_rgb565D_convert_internal
+#undef rgb_rgb565_convert_internal
+#undef rgb_rgb565D_convert_internal
+#undef gray_rgb565_convert_internal
+#undef gray_rgb565D_convert_internal
+
+
+METHODDEF(void)
+ycc_rgb565_convert (j_decompress_ptr cinfo,
+                    JSAMPIMAGE input_buf, JDIMENSION input_row,
+                    JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+ycc_rgb565D_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    ycc_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    ycc_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565_convert (j_decompress_ptr cinfo,
+                    JSAMPIMAGE input_buf, JDIMENSION input_row,
+                    JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+rgb_rgb565D_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    rgb_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    rgb_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565_convert (j_decompress_ptr cinfo,
+                     JSAMPIMAGE input_buf, JDIMENSION input_row,
+                     JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+METHODDEF(void)
+gray_rgb565D_convert (j_decompress_ptr cinfo,
+                      JSAMPIMAGE input_buf, JDIMENSION input_row,
+                      JSAMPARRAY output_buf, int num_rows)
+{
+  if (is_big_endian())
+    gray_rgb565D_convert_be(cinfo, input_buf, input_row, output_buf, num_rows);
+  else
+    gray_rgb565D_convert_le(cinfo, input_buf, input_row, output_buf, num_rows);
+}
+
+
+/*
  * Empty method for start_pass.
  */
 
@@ -426,7 +756,7 @@
 
   cconvert = (my_cconvert_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_color_deconverter));
+                                sizeof(my_color_deconverter));
   cinfo->cconvert = (struct jpeg_color_deconverter *) cconvert;
   cconvert->pub.start_pass = start_pass_dcolor;
 
@@ -449,7 +779,7 @@
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
     break;
 
-  default:			/* JCS_UNKNOWN can be anything */
+  default:                      /* JCS_UNKNOWN can be anything */
     if (cinfo->num_components < 1)
       ERREXIT(cinfo, JERR_BAD_J_COLORSPACE);
     break;
@@ -464,11 +794,14 @@
   case JCS_GRAYSCALE:
     cinfo->out_color_components = 1;
     if (cinfo->jpeg_color_space == JCS_GRAYSCALE ||
-	cinfo->jpeg_color_space == JCS_YCbCr) {
+        cinfo->jpeg_color_space == JCS_YCbCr) {
       cconvert->pub.color_convert = grayscale_convert;
       /* For color->grayscale conversion, only the Y (0) component is needed */
       for (ci = 1; ci < cinfo->num_components; ci++)
-	cinfo->comp_info[ci].component_needed = FALSE;
+        cinfo->comp_info[ci].component_needed = FALSE;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      cconvert->pub.color_convert = rgb_gray_convert;
+      build_rgb_y_table(cinfo);
     } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
@@ -494,13 +827,48 @@
       }
     } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
       cconvert->pub.color_convert = gray_rgb_convert;
-    } else if (cinfo->jpeg_color_space == cinfo->out_color_space &&
-      rgb_pixelsize[cinfo->out_color_space] == 3) {
-      cconvert->pub.color_convert = null_convert;
+    } else if (cinfo->jpeg_color_space == JCS_RGB) {
+      if (rgb_red[cinfo->out_color_space] == 0 &&
+          rgb_green[cinfo->out_color_space] == 1 &&
+          rgb_blue[cinfo->out_color_space] == 2 &&
+          rgb_pixelsize[cinfo->out_color_space] == 3)
+        cconvert->pub.color_convert = null_convert;
+      else
+        cconvert->pub.color_convert = rgb_rgb_convert;
     } else
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
 
+  case JCS_RGB565:
+    cinfo->out_color_components = 3;
+    if (cinfo->dither_mode == JDITHER_NONE) {
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+         if (jsimd_can_ycc_rgb565())
+           cconvert->pub.color_convert = jsimd_ycc_rgb565_convert;
+         else {
+           cconvert->pub.color_convert = ycc_rgb565_convert;
+           build_ycc_rgb_table(cinfo);
+        }
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    } else {
+      /* only ordered dithering is supported */
+      if (cinfo->jpeg_color_space == JCS_YCbCr) {
+        cconvert->pub.color_convert = ycc_rgb565D_convert;
+        build_ycc_rgb_table(cinfo);
+      } else if (cinfo->jpeg_color_space == JCS_GRAYSCALE) {
+        cconvert->pub.color_convert = gray_rgb565D_convert;
+      } else if (cinfo->jpeg_color_space == JCS_RGB) {
+        cconvert->pub.color_convert = rgb_rgb565D_convert;
+      } else
+        ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
+    }
+    break;
+
   case JCS_CMYK:
     cinfo->out_color_components = 4;
     if (cinfo->jpeg_color_space == JCS_YCCK) {
@@ -517,7 +885,7 @@
     if (cinfo->out_color_space == cinfo->jpeg_color_space) {
       cinfo->out_color_components = cinfo->num_components;
       cconvert->pub.color_convert = null_convert;
-    } else			/* unsupported non-null conversion */
+    } else                      /* unsupported non-null conversion */
       ERREXIT(cinfo, JERR_CONVERSION_NOTIMPL);
     break;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdct.h glmark2-2021.02/src/libjpeg-turbo/jdct.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdct.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdct.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,14 +1,17 @@
 /*
  * jdct.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This include file contains common declarations for the forward and
  * inverse DCT modules.  These declarations are private to the DCT managers
  * (jcdctmgr.c, jddctmgr.c) and the individual DCT algorithms.
- * The individual DCT algorithms are kept in separate files to ease 
+ * The individual DCT algorithms are kept in separate files to ease
  * machine-dependent tuning (e.g., assembly coding).
  */
 
@@ -16,7 +19,7 @@
 /*
  * A forward DCT routine is given a pointer to a work area of type DCTELEM[];
  * the DCT is to be performed in-place in that buffer.  Type DCTELEM is int
- * for 8-bit samples, INT32 for 12-bit samples.  (NOTE: Floating-point DCT
+ * for 8-bit samples, JLONG for 12-bit samples.  (NOTE: Floating-point DCT
  * implementations use an array of type FAST_FLOAT, instead.)
  * The DCT inputs are expected to be signed (range +-CENTERJSAMPLE).
  * The DCT outputs are returned scaled up by a factor of 8; they therefore
@@ -29,7 +32,7 @@
 
 #if BITS_IN_JSAMPLE == 8
 #ifndef WITH_SIMD
-typedef int DCTELEM;		/* 16 or 32 bits is fine */
+typedef int DCTELEM;            /* 16 or 32 bits is fine */
 typedef unsigned int UDCTELEM;
 typedef unsigned long long UDCTELEM2;
 #else
@@ -38,8 +41,7 @@
 typedef unsigned int UDCTELEM2;
 #endif
 #else
-typedef INT32 DCTELEM;		/* must have 32 bits */
-typedef UINT32 UDCTELEM;
+typedef JLONG DCTELEM;          /* must have 32 bits */
 typedef unsigned long long UDCTELEM2;
 #endif
 
@@ -64,10 +66,10 @@
 typedef MULTIPLIER ISLOW_MULT_TYPE; /* short or int, whichever is faster */
 #if BITS_IN_JSAMPLE == 8
 typedef MULTIPLIER IFAST_MULT_TYPE; /* 16 bits is OK, use short if faster */
-#define IFAST_SCALE_BITS  2	/* fractional bits in scale factors */
+#define IFAST_SCALE_BITS  2     /* fractional bits in scale factors */
 #else
-typedef INT32 IFAST_MULT_TYPE;	/* need 32 bits for scaled quantizers */
-#define IFAST_SCALE_BITS  13	/* fractional bits in scale factors */
+typedef JLONG IFAST_MULT_TYPE;  /* need 32 bits for scaled quantizers */
+#define IFAST_SCALE_BITS  13    /* fractional bits in scale factors */
 #endif
 typedef FAST_FLOAT FLOAT_MULT_TYPE; /* preferred floating type */
 
@@ -86,57 +88,79 @@
 #define RANGE_MASK  (MAXJSAMPLE * 4 + 3) /* 2 bits wider than legal samples */
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_fdct_islow		jFDislow
-#define jpeg_fdct_ifast		jFDifast
-#define jpeg_fdct_float		jFDfloat
-#define jpeg_idct_islow		jRDislow
-#define jpeg_idct_ifast		jRDifast
-#define jpeg_idct_float		jRDfloat
-#define jpeg_idct_4x4		jRD4x4
-#define jpeg_idct_2x2		jRD2x2
-#define jpeg_idct_1x1		jRD1x1
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
 /* Extern declarations for the forward and inverse DCT routines. */
 
-EXTERN(void) jpeg_fdct_islow JPP((DCTELEM * data));
-EXTERN(void) jpeg_fdct_ifast JPP((DCTELEM * data));
-EXTERN(void) jpeg_fdct_float JPP((FAST_FLOAT * data));
+EXTERN(void) jpeg_fdct_islow (DCTELEM *data);
+EXTERN(void) jpeg_fdct_ifast (DCTELEM *data);
+EXTERN(void) jpeg_fdct_float (FAST_FLOAT *data);
 
 EXTERN(void) jpeg_idct_islow
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_ifast
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_float
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_7x7
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_6x6
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_5x5
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_4x4
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_3x3
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_2x2
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 EXTERN(void) jpeg_idct_1x1
-    JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	 JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_9x9
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_10x10
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_11x11
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_12x12
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_13x13
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_14x14
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_15x15
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
+EXTERN(void) jpeg_idct_16x16
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JCOEFPTR coef_block, JSAMPARRAY output_buf, JDIMENSION output_col);
 
 
 /*
  * Macros for handling fixed-point arithmetic; these are used by many
  * but not all of the DCT/IDCT modules.
  *
- * All values are expected to be of type INT32.
+ * All values are expected to be of type JLONG.
  * Fractional constants are scaled left by CONST_BITS bits.
  * CONST_BITS is defined within each module using these macros,
  * and may differ from one module to the next.
  */
 
-#define ONE	((INT32) 1)
+#define ONE     ((JLONG) 1)
 #define CONST_SCALE (ONE << CONST_BITS)
 
 /* Convert a positive real constant to an integer scaled by CONST_SCALE.
@@ -144,16 +168,16 @@
  * thus causing a lot of useless floating-point operations at run time.
  */
 
-#define FIX(x)	((INT32) ((x) * CONST_SCALE + 0.5))
+#define FIX(x)  ((JLONG) ((x) * CONST_SCALE + 0.5))
 
-/* Descale and correctly round an INT32 value that's scaled by N bits.
+/* Descale and correctly round a JLONG value that's scaled by N bits.
  * We assume RIGHT_SHIFT rounds towards minus infinity, so adding
  * the fudge factor is correct for either sign of X.
  */
 
 #define DESCALE(x,n)  RIGHT_SHIFT((x) + (ONE << ((n)-1)), n)
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
  * This macro is used only when the two inputs will actually be no more than
  * 16 bits wide, so that a 16x16->32 bit multiply can be used instead of a
  * full 32x32 multiply.  This provides a useful speedup on many machines.
@@ -162,23 +186,23 @@
  * correct combination of casts.
  */
 
-#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
 #define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT16) (const)))
 #endif
-#ifdef SHORTxLCONST_32		/* known to work with Microsoft C 6.0 */
-#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((INT32) (const)))
+#ifdef SHORTxLCONST_32          /* known to work with Microsoft C 6.0 */
+#define MULTIPLY16C16(var,const)  (((INT16) (var)) * ((JLONG) (const)))
 #endif
 
-#ifndef MULTIPLY16C16		/* default definition */
+#ifndef MULTIPLY16C16           /* default definition */
 #define MULTIPLY16C16(var,const)  ((var) * (const))
 #endif
 
 /* Same except both inputs are variables. */
 
-#ifdef SHORTxSHORT_32		/* may work if 'int' is 32 bits */
+#ifdef SHORTxSHORT_32           /* may work if 'int' is 32 bits */
 #define MULTIPLY16V16(var1,var2)  (((INT16) (var1)) * ((INT16) (var2)))
 #endif
 
-#ifndef MULTIPLY16V16		/* default definition */
+#ifndef MULTIPLY16V16           /* default definition */
 #define MULTIPLY16V16(var1,var2)  ((var1) * (var2))
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jddctmgr.c glmark2-2021.02/src/libjpeg-turbo/jddctmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jddctmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jddctmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jddctmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2002-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2010, 2015, D. R. Commander.
+ * Copyright (C) 2013, MIPS Technologies, Inc., California
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the inverse-DCT management logic.
  * This code selects a particular IDCT implementation to be used,
@@ -20,7 +24,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 #include "jsimddct.h"
 #include "jpegcomp.h"
 
@@ -45,7 +49,7 @@
 /* Private subobject for this module */
 
 typedef struct {
-  struct jpeg_inverse_dct pub;	/* public fields */
+  struct jpeg_inverse_dct pub;  /* public fields */
 
   /* This array contains the IDCT method code that each multiplier table
    * is currently set up for, or -1 if it's not yet set up.
@@ -55,7 +59,7 @@
   int cur_method[MAX_COMPONENTS];
 } my_idct_controller;
 
-typedef my_idct_controller * my_idct_ptr;
+typedef my_idct_controller *my_idct_ptr;
 
 
 /* Allocated multiplier tables: big enough for any supported variant */
@@ -97,7 +101,7 @@
   jpeg_component_info *compptr;
   int method = 0;
   inverse_DCT_method_ptr method_ptr = NULL;
-  JQUANT_TBL * qtbl;
+  JQUANT_TBL *qtbl;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
@@ -106,57 +110,117 @@
 #ifdef IDCT_SCALING_SUPPORTED
     case 1:
       method_ptr = jpeg_idct_1x1;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
       break;
     case 2:
       if (jsimd_can_idct_2x2())
         method_ptr = jsimd_idct_2x2;
       else
         method_ptr = jpeg_idct_2x2;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 3:
+      method_ptr = jpeg_idct_3x3;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
       break;
     case 4:
       if (jsimd_can_idct_4x4())
         method_ptr = jsimd_idct_4x4;
       else
         method_ptr = jpeg_idct_4x4;
-      method = JDCT_ISLOW;	/* jidctred uses islow-style table */
+      method = JDCT_ISLOW;      /* jidctred uses islow-style table */
+      break;
+    case 5:
+      method_ptr = jpeg_idct_5x5;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 6:
+#if defined(__mips__)
+      if (jsimd_can_idct_6x6())
+        method_ptr = jsimd_idct_6x6;
+      else
+#endif
+      method_ptr = jpeg_idct_6x6;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 7:
+      method_ptr = jpeg_idct_7x7;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
       break;
 #endif
     case DCTSIZE:
       switch (cinfo->dct_method) {
 #ifdef DCT_ISLOW_SUPPORTED
       case JDCT_ISLOW:
-	if (jsimd_can_idct_islow())
-	  method_ptr = jsimd_idct_islow;
-	else
-	  method_ptr = jpeg_idct_islow;
-	method = JDCT_ISLOW;
-	break;
+        if (jsimd_can_idct_islow())
+          method_ptr = jsimd_idct_islow;
+        else
+          method_ptr = jpeg_idct_islow;
+        method = JDCT_ISLOW;
+        break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
       case JDCT_IFAST:
-	if (jsimd_can_idct_ifast())
-	  method_ptr = jsimd_idct_ifast;
-	else
-	  method_ptr = jpeg_idct_ifast;
-	method = JDCT_IFAST;
-	break;
+        if (jsimd_can_idct_ifast())
+          method_ptr = jsimd_idct_ifast;
+        else
+          method_ptr = jpeg_idct_ifast;
+        method = JDCT_IFAST;
+        break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
       case JDCT_FLOAT:
-	if (jsimd_can_idct_float())
-	  method_ptr = jsimd_idct_float;
-	else
-	  method_ptr = jpeg_idct_float;
-	method = JDCT_FLOAT;
-	break;
+        if (jsimd_can_idct_float())
+          method_ptr = jsimd_idct_float;
+        else
+          method_ptr = jpeg_idct_float;
+        method = JDCT_FLOAT;
+        break;
 #endif
       default:
-	ERREXIT(cinfo, JERR_NOT_COMPILED);
-	break;
+        ERREXIT(cinfo, JERR_NOT_COMPILED);
+        break;
       }
       break;
+#ifdef IDCT_SCALING_SUPPORTED
+    case 9:
+      method_ptr = jpeg_idct_9x9;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 10:
+      method_ptr = jpeg_idct_10x10;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 11:
+      method_ptr = jpeg_idct_11x11;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 12:
+#if defined(__mips__)
+      if (jsimd_can_idct_12x12())
+        method_ptr = jsimd_idct_12x12;
+      else
+#endif
+      method_ptr = jpeg_idct_12x12;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 13:
+      method_ptr = jpeg_idct_13x13;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 14:
+      method_ptr = jpeg_idct_14x14;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 15:
+      method_ptr = jpeg_idct_15x15;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+    case 16:
+      method_ptr = jpeg_idct_16x16;
+      method = JDCT_ISLOW;      /* jidctint uses islow-style table */
+      break;
+#endif
     default:
       ERREXIT1(cinfo, JERR_BAD_DCTSIZE, compptr->_DCT_scaled_size);
       break;
@@ -172,81 +236,81 @@
     if (! compptr->component_needed || idct->cur_method[ci] == method)
       continue;
     qtbl = compptr->quant_table;
-    if (qtbl == NULL)		/* happens if no data yet for component */
+    if (qtbl == NULL)           /* happens if no data yet for component */
       continue;
     idct->cur_method[ci] = method;
     switch (method) {
 #ifdef PROVIDE_ISLOW_TABLES
     case JDCT_ISLOW:
       {
-	/* For LL&M IDCT method, multipliers are equal to raw quantization
-	 * coefficients, but are stored as ints to ensure access efficiency.
-	 */
-	ISLOW_MULT_TYPE * ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
-	for (i = 0; i < DCTSIZE2; i++) {
-	  ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
-	}
+        /* For LL&M IDCT method, multipliers are equal to raw quantization
+         * coefficients, but are stored as ints to ensure access efficiency.
+         */
+        ISLOW_MULT_TYPE *ismtbl = (ISLOW_MULT_TYPE *) compptr->dct_table;
+        for (i = 0; i < DCTSIZE2; i++) {
+          ismtbl[i] = (ISLOW_MULT_TYPE) qtbl->quantval[i];
+        }
       }
       break;
 #endif
 #ifdef DCT_IFAST_SUPPORTED
     case JDCT_IFAST:
       {
-	/* For AA&N IDCT method, multipliers are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 * For integer operation, the multiplier table is to be scaled by
-	 * IFAST_SCALE_BITS.
-	 */
-	IFAST_MULT_TYPE * ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
+        /* For AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         * For integer operation, the multiplier table is to be scaled by
+         * IFAST_SCALE_BITS.
+         */
+        IFAST_MULT_TYPE *ifmtbl = (IFAST_MULT_TYPE *) compptr->dct_table;
 #define CONST_BITS 14
-	static const INT16 aanscales[DCTSIZE2] = {
-	  /* precomputed values scaled up by 14 bits */
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
-	  21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
-	  19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
-	  16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
-	  12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
-	   8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
-	   4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
-	};
-	SHIFT_TEMPS
-
-	for (i = 0; i < DCTSIZE2; i++) {
-	  ifmtbl[i] = (IFAST_MULT_TYPE)
-	    DESCALE(MULTIPLY16V16((INT32) qtbl->quantval[i],
-				  (INT32) aanscales[i]),
-		    CONST_BITS-IFAST_SCALE_BITS);
-	}
+        static const INT16 aanscales[DCTSIZE2] = {
+          /* precomputed values scaled up by 14 bits */
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
+          21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
+          19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
+          16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
+          12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
+           8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
+           4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
+        };
+        SHIFT_TEMPS
+
+        for (i = 0; i < DCTSIZE2; i++) {
+          ifmtbl[i] = (IFAST_MULT_TYPE)
+            DESCALE(MULTIPLY16V16((JLONG) qtbl->quantval[i],
+                                  (JLONG) aanscales[i]),
+                    CONST_BITS-IFAST_SCALE_BITS);
+        }
       }
       break;
 #endif
 #ifdef DCT_FLOAT_SUPPORTED
     case JDCT_FLOAT:
       {
-	/* For float AA&N IDCT method, multipliers are equal to quantization
-	 * coefficients scaled by scalefactor[row]*scalefactor[col], where
-	 *   scalefactor[0] = 1
-	 *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
-	 */
-	FLOAT_MULT_TYPE * fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
-	int row, col;
-	static const double aanscalefactor[DCTSIZE] = {
-	  1.0, 1.387039845, 1.306562965, 1.175875602,
-	  1.0, 0.785694958, 0.541196100, 0.275899379
-	};
-
-	i = 0;
-	for (row = 0; row < DCTSIZE; row++) {
-	  for (col = 0; col < DCTSIZE; col++) {
-	    fmtbl[i] = (FLOAT_MULT_TYPE)
-	      ((double) qtbl->quantval[i] *
-	       aanscalefactor[row] * aanscalefactor[col]);
-	    i++;
-	  }
-	}
+        /* For float AA&N IDCT method, multipliers are equal to quantization
+         * coefficients scaled by scalefactor[row]*scalefactor[col], where
+         *   scalefactor[0] = 1
+         *   scalefactor[k] = cos(k*PI/16) * sqrt(2)    for k=1..7
+         */
+        FLOAT_MULT_TYPE *fmtbl = (FLOAT_MULT_TYPE *) compptr->dct_table;
+        int row, col;
+        static const double aanscalefactor[DCTSIZE] = {
+          1.0, 1.387039845, 1.306562965, 1.175875602,
+          1.0, 0.785694958, 0.541196100, 0.275899379
+        };
+
+        i = 0;
+        for (row = 0; row < DCTSIZE; row++) {
+          for (col = 0; col < DCTSIZE; col++) {
+            fmtbl[i] = (FLOAT_MULT_TYPE)
+              ((double) qtbl->quantval[i] *
+               aanscalefactor[row] * aanscalefactor[col]);
+            i++;
+          }
+        }
       }
       break;
 #endif
@@ -271,7 +335,7 @@
 
   idct = (my_idct_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_idct_controller));
+                                sizeof(my_idct_controller));
   cinfo->idct = (struct jpeg_inverse_dct *) idct;
   idct->pub.start_pass = start_pass;
 
@@ -280,8 +344,8 @@
     /* Allocate and pre-zero a multiplier table for each component */
     compptr->dct_table =
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(multiplier_table));
-    MEMZERO(compptr->dct_table, SIZEOF(multiplier_table));
+                                  sizeof(multiplier_table));
+    MEMZERO(compptr->dct_table, sizeof(multiplier_table));
     /* Mark multiplier table not yet set up for any method */
     idct->cur_method[ci] = -1;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdhuff.c glmark2-2021.02/src/libjpeg-turbo/jdhuff.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdhuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdhuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdhuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy decoding routines.
  *
@@ -18,8 +20,9 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdhuff.h"		/* Declarations shared with jdphuff.c */
+#include "jdhuff.h"             /* Declarations shared with jdphuff.c */
 #include "jpegcomp.h"
+#include "jstdhuff.c"
 
 
 /*
@@ -43,10 +46,10 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -57,27 +60,27 @@
   /* These fields are loaded into local variables at start of each MCU.
    * In case of suspension, we exit WITHOUT updating them.
    */
-  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
-  savable_state saved;		/* Other state at start of MCU */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  d_derived_tbl * dc_derived_tbls[NUM_HUFF_TBLS];
-  d_derived_tbl * ac_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *dc_derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *ac_derived_tbls[NUM_HUFF_TBLS];
 
   /* Precalculated info set up by start_pass for use in decode_mcu: */
 
   /* Pointers to derived tables to be used for each block within an MCU */
-  d_derived_tbl * dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
-  d_derived_tbl * ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl *dc_cur_tbls[D_MAX_BLOCKS_IN_MCU];
+  d_derived_tbl *ac_cur_tbls[D_MAX_BLOCKS_IN_MCU];
   /* Whether we care about the DC and AC coefficient values for each block */
   boolean dc_needed[D_MAX_BLOCKS_IN_MCU];
   boolean ac_needed[D_MAX_BLOCKS_IN_MCU];
 } huff_entropy_decoder;
 
-typedef huff_entropy_decoder * huff_entropy_ptr;
+typedef huff_entropy_decoder *huff_entropy_ptr;
 
 
 /*
@@ -89,7 +92,8 @@
 {
   huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
   int ci, blkn, dctbl, actbl;
-  jpeg_component_info * compptr;
+  d_derived_tbl **pdtbl;
+  jpeg_component_info *compptr;
 
   /* Check that the scan parameters Ss, Se, Ah/Al are OK for sequential JPEG.
    * This ought to be an error condition, but we make it a warning because
@@ -105,10 +109,10 @@
     actbl = compptr->ac_tbl_no;
     /* Compute derived values for Huffman tables */
     /* We may do this more than once for a table, but it's not expensive */
-    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl,
-			    & entropy->dc_derived_tbls[dctbl]);
-    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl,
-			    & entropy->ac_derived_tbls[actbl]);
+    pdtbl = entropy->dc_derived_tbls + dctbl;
+    jpeg_make_d_derived_tbl(cinfo, TRUE, dctbl, pdtbl);
+    pdtbl = entropy->ac_derived_tbls + actbl;
+    jpeg_make_d_derived_tbl(cinfo, FALSE, actbl, pdtbl);
     /* Initialize DC predictions to 0 */
     entropy->saved.last_dc_val[ci] = 0;
   }
@@ -149,7 +153,7 @@
 
 GLOBAL(void)
 jpeg_make_d_derived_tbl (j_decompress_ptr cinfo, boolean isDC, int tblno,
-			 d_derived_tbl ** pdtbl)
+                         d_derived_tbl **pdtbl)
 {
   JHUFF_TBL *htbl;
   d_derived_tbl *dtbl;
@@ -175,26 +179,26 @@
   if (*pdtbl == NULL)
     *pdtbl = (d_derived_tbl *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(d_derived_tbl));
+                                  sizeof(d_derived_tbl));
   dtbl = *pdtbl;
-  dtbl->pub = htbl;		/* fill in back link */
-  
+  dtbl->pub = htbl;             /* fill in back link */
+
   /* Figure C.1: make table of Huffman code length for each symbol */
 
   p = 0;
   for (l = 1; l <= 16; l++) {
     i = (int) htbl->bits[l];
-    if (i < 0 || p + i > 256)	/* protect against table overrun */
+    if (i < 0 || p + i > 256)   /* protect against table overrun */
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     while (i--)
       huffsize[p++] = (char) l;
   }
   huffsize[p] = 0;
   numsymbols = p;
-  
+
   /* Figure C.2: generate the codes themselves */
   /* We also validate that the counts represent a legal Huffman code tree. */
-  
+
   code = 0;
   si = huffsize[0];
   p = 0;
@@ -206,7 +210,7 @@
     /* code is now 1 more than the last code used for codelength si; but
      * it must still fit in si bits, since no code is allowed to be all ones.
      */
-    if (((INT32) code) >= (((INT32) 1) << si))
+    if (((JLONG) code) >= (((JLONG) 1) << si))
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     code <<= 1;
     si++;
@@ -220,11 +224,11 @@
       /* valoffset[l] = huffval[] index of 1st symbol of code length l,
        * minus the minimum code of length l
        */
-      dtbl->valoffset[l] = (INT32) p - (INT32) huffcode[p];
+      dtbl->valoffset[l] = (JLONG) p - (JLONG) huffcode[p];
       p += htbl->bits[l];
       dtbl->maxcode[l] = huffcode[p-1]; /* maximum code of length l */
     } else {
-      dtbl->maxcode[l] = -1;	/* -1 if no codes of this length */
+      dtbl->maxcode[l] = -1;    /* -1 if no codes of this length */
     }
   }
   dtbl->valoffset[17] = 0;
@@ -247,8 +251,8 @@
       /* Generate left-justified code followed by all possible bit sequences */
       lookbits = huffcode[p] << (HUFF_LOOKAHEAD-l);
       for (ctr = 1 << (HUFF_LOOKAHEAD-l); ctr > 0; ctr--) {
-	dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p];
-	lookbits++;
+        dtbl->lookup[lookbits] = (l << HUFF_LOOKAHEAD) | htbl->huffval[p];
+        lookbits++;
       }
     }
   }
@@ -263,7 +267,7 @@
     for (i = 0; i < numsymbols; i++) {
       int sym = htbl->huffval[i];
       if (sym < 0 || sym > 15)
-	ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+        ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
     }
   }
 }
@@ -285,20 +289,20 @@
  */
 
 #ifdef SLOW_SHIFT_32
-#define MIN_GET_BITS  15	/* minimum allowable value */
+#define MIN_GET_BITS  15        /* minimum allowable value */
 #else
 #define MIN_GET_BITS  (BIT_BUF_SIZE-7)
 #endif
 
 
 GLOBAL(boolean)
-jpeg_fill_bit_buffer (bitread_working_state * state,
-		      register bit_buf_type get_buffer, register int bits_left,
-		      int nbits)
+jpeg_fill_bit_buffer (bitread_working_state *state,
+                      register bit_buf_type get_buffer, register int bits_left,
+                      int nbits)
 /* Load up the bit buffer to a depth of at least nbits */
 {
   /* Copy heavily used state fields into locals (hopefully registers) */
-  register const JOCTET * next_input_byte = state->next_input_byte;
+  register const JOCTET *next_input_byte = state->next_input_byte;
   register size_t bytes_in_buffer = state->bytes_in_buffer;
   j_decompress_ptr cinfo = state->cinfo;
 
@@ -306,54 +310,54 @@
   /* (It is assumed that no request will be for more than that many bits.) */
   /* We fail to do so only if we hit a marker or are forced to suspend. */
 
-  if (cinfo->unread_marker == 0) {	/* cannot advance past a marker */
+  if (cinfo->unread_marker == 0) {      /* cannot advance past a marker */
     while (bits_left < MIN_GET_BITS) {
       register int c;
 
       /* Attempt to read a byte */
       if (bytes_in_buffer == 0) {
-	if (! (*cinfo->src->fill_input_buffer) (cinfo))
-	  return FALSE;
-	next_input_byte = cinfo->src->next_input_byte;
-	bytes_in_buffer = cinfo->src->bytes_in_buffer;
+        if (! (*cinfo->src->fill_input_buffer) (cinfo))
+          return FALSE;
+        next_input_byte = cinfo->src->next_input_byte;
+        bytes_in_buffer = cinfo->src->bytes_in_buffer;
       }
       bytes_in_buffer--;
       c = GETJOCTET(*next_input_byte++);
 
       /* If it's 0xFF, check and discard stuffed zero byte */
       if (c == 0xFF) {
-	/* Loop here to discard any padding FF's on terminating marker,
-	 * so that we can save a valid unread_marker value.  NOTE: we will
-	 * accept multiple FF's followed by a 0 as meaning a single FF data
-	 * byte.  This data pattern is not valid according to the standard.
-	 */
-	do {
-	  if (bytes_in_buffer == 0) {
-	    if (! (*cinfo->src->fill_input_buffer) (cinfo))
-	      return FALSE;
-	    next_input_byte = cinfo->src->next_input_byte;
-	    bytes_in_buffer = cinfo->src->bytes_in_buffer;
-	  }
-	  bytes_in_buffer--;
-	  c = GETJOCTET(*next_input_byte++);
-	} while (c == 0xFF);
-
-	if (c == 0) {
-	  /* Found FF/00, which represents an FF data byte */
-	  c = 0xFF;
-	} else {
-	  /* Oops, it's actually a marker indicating end of compressed data.
-	   * Save the marker code for later use.
-	   * Fine point: it might appear that we should save the marker into
-	   * bitread working state, not straight into permanent state.  But
-	   * once we have hit a marker, we cannot need to suspend within the
-	   * current MCU, because we will read no more bytes from the data
-	   * source.  So it is OK to update permanent state right away.
-	   */
-	  cinfo->unread_marker = c;
-	  /* See if we need to insert some fake zero bits. */
-	  goto no_more_bytes;
-	}
+        /* Loop here to discard any padding FF's on terminating marker,
+         * so that we can save a valid unread_marker value.  NOTE: we will
+         * accept multiple FF's followed by a 0 as meaning a single FF data
+         * byte.  This data pattern is not valid according to the standard.
+         */
+        do {
+          if (bytes_in_buffer == 0) {
+            if (! (*cinfo->src->fill_input_buffer) (cinfo))
+              return FALSE;
+            next_input_byte = cinfo->src->next_input_byte;
+            bytes_in_buffer = cinfo->src->bytes_in_buffer;
+          }
+          bytes_in_buffer--;
+          c = GETJOCTET(*next_input_byte++);
+        } while (c == 0xFF);
+
+        if (c == 0) {
+          /* Found FF/00, which represents an FF data byte */
+          c = 0xFF;
+        } else {
+          /* Oops, it's actually a marker indicating end of compressed data.
+           * Save the marker code for later use.
+           * Fine point: it might appear that we should save the marker into
+           * bitread working state, not straight into permanent state.  But
+           * once we have hit a marker, we cannot need to suspend within the
+           * current MCU, because we will read no more bytes from the data
+           * source.  So it is OK to update permanent state right away.
+           */
+          cinfo->unread_marker = c;
+          /* See if we need to insert some fake zero bits. */
+          goto no_more_bytes;
+        }
       }
 
       /* OK, load c into get_buffer */
@@ -373,8 +377,8 @@
        * appears per data segment.
        */
       if (! cinfo->entropy->insufficient_data) {
-	WARNMS(cinfo, JWRN_HIT_MARKER);
-	cinfo->entropy->insufficient_data = TRUE;
+        WARNMS(cinfo, JWRN_HIT_MARKER);
+        cinfo->entropy->insufficient_data = TRUE;
       }
       /* Fill the buffer with zero bits */
       get_buffer <<= MIN_GET_BITS - bits_left;
@@ -417,11 +421,11 @@
   } \
 }
 
-#if __WORDSIZE == 64 || defined(_WIN64)
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
 /* Pre-fetch 48 bytes, because the holding register is 64-bit */
 #define FILL_BIT_BUFFER_FAST \
-  if (bits_left < 16) { \
+  if (bits_left <= 16) { \
     GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE GET_BYTE \
   }
 
@@ -429,7 +433,7 @@
 
 /* Pre-fetch 16 bytes, because the holding register is 32-bit */
 #define FILL_BIT_BUFFER_FAST \
-  if (bits_left < 16) { \
+  if (bits_left <= 16) { \
     GET_BYTE GET_BYTE \
   }
 
@@ -442,12 +446,12 @@
  */
 
 GLOBAL(int)
-jpeg_huff_decode (bitread_working_state * state,
-		  register bit_buf_type get_buffer, register int bits_left,
-		  d_derived_tbl * htbl, int min_bits)
+jpeg_huff_decode (bitread_working_state *state,
+                  register bit_buf_type get_buffer, register int bits_left,
+                  d_derived_tbl *htbl, int min_bits)
 {
   register int l = min_bits;
-  register INT32 code;
+  register JLONG code;
 
   /* HUFF_DECODE has determined that the code is at least min_bits */
   /* bits long, so fetch that many bits in one swoop. */
@@ -473,7 +477,7 @@
 
   if (l > 16) {
     WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
-    return 0;			/* fake a zero as the safest result */
+    return 0;                   /* fake a zero as the safest result */
   }
 
   return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ];
@@ -488,7 +492,8 @@
 #define AVOID_TABLES
 #ifdef AVOID_TABLES
 
-#define HUFF_EXTEND(x,s)  ((x) + ((((x) - (1<<((s)-1))) >> 31) & (((-1)<<(s)) + 1)))
+#define NEG_1 ((unsigned int)-1)
+#define HUFF_EXTEND(x,s)  ((x) + ((((x) - (1<<((s)-1))) >> 31) & (((NEG_1)<<(s)) + 1)))
 
 #else
 
@@ -560,9 +565,9 @@
   ASSIGN_STATE(state, entropy->saved);
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    JBLOCKROW block = MCU_data[blkn];
-    d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
-    d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
+    d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
     register int s, k, r;
 
     /* Decode a single block's worth of coefficients */
@@ -580,11 +585,13 @@
       int ci = cinfo->MCU_membership[blkn];
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
-      /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
-      (*block)[0] = (JCOEF) s;
+      if (block) {
+        /* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
+        (*block)[0] = (JCOEF) s;
+      }
     }
 
-    if (entropy->ac_needed[blkn]) {
+    if (entropy->ac_needed[blkn] && block) {
 
       /* Section F.2.2.2: decode the AC coefficients */
       /* Since zeroes are skipped, output area must be cleared beforehand */
@@ -593,7 +600,7 @@
 
         r = s >> 4;
         s &= 15;
-      
+
         if (s) {
           k += r;
           CHECK_BIT_BUFFER(br_state, s, return FALSE);
@@ -657,12 +664,12 @@
   ASSIGN_STATE(state, entropy->saved);
 
   for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
-    JBLOCKROW block = MCU_data[blkn];
-    d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
-    d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
+    JBLOCKROW block = MCU_data ? MCU_data[blkn] : NULL;
+    d_derived_tbl *dctbl = entropy->dc_cur_tbls[blkn];
+    d_derived_tbl *actbl = entropy->ac_cur_tbls[blkn];
     register int s, k, r, l;
 
-    HUFF_DECODE_FAST(s, l, dctbl);
+    HUFF_DECODE_FAST(s, l, dctbl, slow_decode_mcu);
     if (s) {
       FILL_BIT_BUFFER_FAST
       r = GET_BITS(s);
@@ -673,16 +680,17 @@
       int ci = cinfo->MCU_membership[blkn];
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
-      (*block)[0] = (JCOEF) s;
+      if (block)
+        (*block)[0] = (JCOEF) s;
     }
 
-    if (entropy->ac_needed[blkn]) {
+    if (entropy->ac_needed[blkn] && block) {
 
       for (k = 1; k < DCTSIZE2; k++) {
-        HUFF_DECODE_FAST(s, l, actbl);
+        HUFF_DECODE_FAST(s, l, actbl, slow_decode_mcu);
         r = s >> 4;
         s &= 15;
-      
+
         if (s) {
           k += r;
           FILL_BIT_BUFFER_FAST
@@ -698,7 +706,7 @@
     } else {
 
       for (k = 1; k < DCTSIZE2; k++) {
-        HUFF_DECODE_FAST(s, l, actbl);
+        HUFF_DECODE_FAST(s, l, actbl, slow_decode_mcu);
         r = s >> 4;
         s &= 15;
 
@@ -715,6 +723,7 @@
   }
 
   if (cinfo->unread_marker != 0) {
+slow_decode_mcu:
     cinfo->unread_marker = 0;
     return FALSE;
   }
@@ -742,7 +751,7 @@
  * this module, since we'll just re-assign them on the next call.)
  */
 
-#define BUFSIZE (DCTSIZE2 * 2)
+#define BUFSIZE (DCTSIZE2 * 8)
 
 METHODDEF(boolean)
 decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
@@ -754,7 +763,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
     usefast = 0;
   }
 
@@ -794,9 +803,15 @@
   huff_entropy_ptr entropy;
   int i;
 
+  /* Motion JPEG frames typically do not include the Huffman tables if they
+     are the default tables.  Thus, if the tables are not set by the time
+     the Huffman decoder is initialized (usually within the body of
+     jpeg_start_decompress()), we set them to default values. */
+  std_huff_tables((j_common_ptr) cinfo);
+
   entropy = (huff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(huff_entropy_decoder));
+                                sizeof(huff_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass_huff_decoder;
   entropy->pub.decode_mcu = decode_mcu;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdhuff.h glmark2-2021.02/src/libjpeg-turbo/jdhuff.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdhuff.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdhuff.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,34 +1,30 @@
 /*
  * jdhuff.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2010-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010-2011, 2015-2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains declarations for Huffman entropy decoding routines
  * that are shared between the sequential decoder (jdhuff.c) and the
  * progressive decoder (jdphuff.c).  No other modules need to see these.
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_make_d_derived_tbl	jMkDDerived
-#define jpeg_fill_bit_buffer	jFilBitBuf
-#define jpeg_huff_decode	jHufDecode
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#include "jconfigint.h"
 
 
 /* Derived data constructed for each Huffman table */
 
-#define HUFF_LOOKAHEAD	8	/* # of bits of lookahead */
+#define HUFF_LOOKAHEAD  8       /* # of bits of lookahead */
 
 typedef struct {
   /* Basic tables: (element [0] of each array is unused) */
-  INT32 maxcode[18];		/* largest code of length k (-1 if none) */
+  JLONG maxcode[18];            /* largest code of length k (-1 if none) */
   /* (maxcode[17] is a sentinel to ensure jpeg_huff_decode terminates) */
-  INT32 valoffset[18];		/* huffval[] offset for codes of length k */
+  JLONG valoffset[18];          /* huffval[] offset for codes of length k */
   /* valoffset[k] = huffval[] index of 1st symbol of code length k, less
    * the smallest code of length k; so given a code of length k, the
    * corresponding symbol is huffval[code + valoffset[k]]
@@ -52,8 +48,8 @@
 
 /* Expand a Huffman table definition into the derived format */
 EXTERN(void) jpeg_make_d_derived_tbl
-	JPP((j_decompress_ptr cinfo, boolean isDC, int tblno,
-	     d_derived_tbl ** pdtbl));
+        (j_decompress_ptr cinfo, boolean isDC, int tblno,
+         d_derived_tbl ** pdtbl);
 
 
 /*
@@ -74,15 +70,19 @@
  * necessary.
  */
 
-#if __WORDSIZE == 64 || defined(_WIN64)
+#if !defined(_WIN32) && !defined(SIZEOF_SIZE_T)
+#error Cannot determine word size
+#endif
+
+#if SIZEOF_SIZE_T==8 || defined(_WIN64)
 
-typedef size_t bit_buf_type;	/* type of bit-extraction buffer */
-#define BIT_BUF_SIZE  64		/* size of buffer in bits */
+typedef size_t bit_buf_type;            /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  64                /* size of buffer in bits */
 
 #else
 
-typedef INT32 bit_buf_type;	/* type of bit-extraction buffer */
-#define BIT_BUF_SIZE  32		/* size of buffer in bits */
+typedef unsigned long bit_buf_type;     /* type of bit-extraction buffer */
+#define BIT_BUF_SIZE  32                /* size of buffer in bits */
 
 #endif
 
@@ -93,43 +93,43 @@
  * because not all machines measure sizeof in 8-bit bytes.
  */
 
-typedef struct {		/* Bitreading state saved across MCUs */
-  bit_buf_type get_buffer;	/* current bit-extraction buffer */
-  int bits_left;		/* # of unused bits in it */
+typedef struct {                /* Bitreading state saved across MCUs */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
 } bitread_perm_state;
 
-typedef struct {		/* Bitreading working state within an MCU */
+typedef struct {                /* Bitreading working state within an MCU */
   /* Current data source location */
   /* We need a copy, rather than munging the original, in case of suspension */
-  const JOCTET * next_input_byte; /* => next byte to read from source */
-  size_t bytes_in_buffer;	/* # of bytes remaining in source buffer */
+  const JOCTET *next_input_byte; /* => next byte to read from source */
+  size_t bytes_in_buffer;       /* # of bytes remaining in source buffer */
   /* Bit input buffer --- note these values are kept in register variables,
    * not in this struct, inside the inner loops.
    */
-  bit_buf_type get_buffer;	/* current bit-extraction buffer */
-  int bits_left;		/* # of unused bits in it */
+  bit_buf_type get_buffer;      /* current bit-extraction buffer */
+  int bits_left;                /* # of unused bits in it */
   /* Pointer needed by jpeg_fill_bit_buffer. */
-  j_decompress_ptr cinfo;	/* back link to decompress master record */
+  j_decompress_ptr cinfo;       /* back link to decompress master record */
 } bitread_working_state;
 
 /* Macros to declare and load/save bitread local variables. */
 #define BITREAD_STATE_VARS  \
-	register bit_buf_type get_buffer;  \
-	register int bits_left;  \
-	bitread_working_state br_state
+        register bit_buf_type get_buffer;  \
+        register int bits_left;  \
+        bitread_working_state br_state
 
 #define BITREAD_LOAD_STATE(cinfop,permstate)  \
-	br_state.cinfo = cinfop; \
-	br_state.next_input_byte = cinfop->src->next_input_byte; \
-	br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
-	get_buffer = permstate.get_buffer; \
-	bits_left = permstate.bits_left;
+        br_state.cinfo = cinfop; \
+        br_state.next_input_byte = cinfop->src->next_input_byte; \
+        br_state.bytes_in_buffer = cinfop->src->bytes_in_buffer; \
+        get_buffer = permstate.get_buffer; \
+        bits_left = permstate.bits_left;
 
 #define BITREAD_SAVE_STATE(cinfop,permstate)  \
-	cinfop->src->next_input_byte = br_state.next_input_byte; \
-	cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
-	permstate.get_buffer = get_buffer; \
-	permstate.bits_left = bits_left
+        cinfop->src->next_input_byte = br_state.next_input_byte; \
+        cinfop->src->bytes_in_buffer = br_state.bytes_in_buffer; \
+        permstate.get_buffer = get_buffer; \
+        permstate.bits_left = bits_left
 
 /*
  * These macros provide the in-line portion of bit fetching.
@@ -137,37 +137,37 @@
  * before using GET_BITS, PEEK_BITS, or DROP_BITS.
  * The variables get_buffer and bits_left are assumed to be locals,
  * but the state struct might not be (jpeg_huff_decode needs this).
- *	CHECK_BIT_BUFFER(state,n,action);
- *		Ensure there are N bits in get_buffer; if suspend, take action.
+ *      CHECK_BIT_BUFFER(state,n,action);
+ *              Ensure there are N bits in get_buffer; if suspend, take action.
  *      val = GET_BITS(n);
- *		Fetch next N bits.
+ *              Fetch next N bits.
  *      val = PEEK_BITS(n);
- *		Fetch next N bits without removing them from the buffer.
- *	DROP_BITS(n);
- *		Discard next N bits.
+ *              Fetch next N bits without removing them from the buffer.
+ *      DROP_BITS(n);
+ *              Discard next N bits.
  * The value N should be a simple variable, not an expression, because it
  * is evaluated multiple times.
  */
 
 #define CHECK_BIT_BUFFER(state,nbits,action) \
-	{ if (bits_left < (nbits)) {  \
-	    if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
-	      { action; }  \
-	    get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
+        { if (bits_left < (nbits)) {  \
+            if (! jpeg_fill_bit_buffer(&(state),get_buffer,bits_left,nbits))  \
+              { action; }  \
+            get_buffer = (state).get_buffer; bits_left = (state).bits_left; } }
 
 #define GET_BITS(nbits) \
-	(((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
+        (((int) (get_buffer >> (bits_left -= (nbits)))) & ((1<<(nbits))-1))
 
 #define PEEK_BITS(nbits) \
-	(((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
+        (((int) (get_buffer >> (bits_left -  (nbits)))) & ((1<<(nbits))-1))
 
 #define DROP_BITS(nbits) \
-	(bits_left -= (nbits))
+        (bits_left -= (nbits))
 
 /* Load up the bit buffer to a depth of at least nbits */
 EXTERN(boolean) jpeg_fill_bit_buffer
-	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
-	     register int bits_left, int nbits));
+        (bitread_working_state *state, register bit_buf_type get_buffer,
+         register int bits_left, int nbits);
 
 
 /*
@@ -203,12 +203,12 @@
   } else { \
 slowlabel: \
     if ((result=jpeg_huff_decode(&state,get_buffer,bits_left,htbl,nb)) < 0) \
-	{ failaction; } \
+        { failaction; } \
     get_buffer = state.get_buffer; bits_left = state.bits_left; \
   } \
 }
 
-#define HUFF_DECODE_FAST(s,nb,htbl) \
+#define HUFF_DECODE_FAST(s,nb,htbl,slowlabel) \
   FILL_BIT_BUFFER_FAST; \
   s = PEEK_BITS(HUFF_LOOKAHEAD); \
   s = htbl->lookup[s]; \
@@ -225,10 +225,12 @@
       s |= GET_BITS(1); \
       nb++; \
     } \
-    s = htbl->pub->huffval[ (int) (s + htbl->valoffset[nb]) & 0xFF ]; \
+    if (nb > 16) \
+      goto slowlabel; \
+    s = htbl->pub->huffval[ (int) (s + htbl->valoffset[nb]) ]; \
   }
 
 /* Out-of-line case for Huffman code fetching */
 EXTERN(int) jpeg_huff_decode
-	JPP((bitread_working_state * state, register bit_buf_type get_buffer,
-	     register int bits_left, d_derived_tbl * htbl, int min_bits));
+        (bitread_working_state *state, register bit_buf_type get_buffer,
+         register int bits_left, d_derived_tbl *htbl, int min_bits);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdinput.c glmark2-2021.02/src/libjpeg-turbo/jdinput.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdinput.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdinput.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,13 @@
 /*
  * jdinput.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Modified 2002-2009 by Guido Vollbeding.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains input control logic for the JPEG decompressor.
  * These routines are concerned with controlling the decompressor's input
@@ -24,93 +26,20 @@
 typedef struct {
   struct jpeg_input_controller pub; /* public fields */
 
-  boolean inheaders;		/* TRUE until first SOS is reached */
+  boolean inheaders;            /* TRUE until first SOS is reached */
 } my_input_controller;
 
-typedef my_input_controller * my_inputctl_ptr;
+typedef my_input_controller *my_inputctl_ptr;
 
 
 /* Forward declarations */
-METHODDEF(int) consume_markers JPP((j_decompress_ptr cinfo));
+METHODDEF(int) consume_markers (j_decompress_ptr cinfo);
 
 
 /*
  * Routines to calculate various quantities related to the size of the image.
  */
 
-
-#if JPEG_LIB_VERSION >= 80
-/*
- * Compute output image dimensions and related values.
- * NOTE: this is exported for possible use by application.
- * Hence it mustn't do anything that can't be done twice.
- */
-
-GLOBAL(void)
-jpeg_core_output_dimensions (j_decompress_ptr cinfo)
-/* Do computations that are needed before master selection phase.
- * This function is used for transcoding and full decompression.
- */
-{
-#ifdef IDCT_SCALING_SUPPORTED
-  int ci;
-  jpeg_component_info *compptr;
-
-  /* Compute actual output image dimensions and DCT scaling choices. */
-  if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom) {
-    /* Provide 1/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 1;
-    cinfo->min_DCT_v_scaled_size = 1;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 2) {
-    /* Provide 2/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 2L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 2L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 2;
-    cinfo->min_DCT_v_scaled_size = 2;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 4) {
-    /* Provide 4/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 4L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 4L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 4;
-    cinfo->min_DCT_v_scaled_size = 4;
-  } else if (cinfo->scale_num * cinfo->block_size <= cinfo->scale_denom * 8) {
-    /* Provide 8/block_size scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width * 8L, (long) cinfo->block_size);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height * 8L, (long) cinfo->block_size);
-    cinfo->min_DCT_h_scaled_size = 8;
-    cinfo->min_DCT_v_scaled_size = 8;
-  }
-  /* Recompute dimensions of components */
-  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-       ci++, compptr++) {
-    compptr->DCT_h_scaled_size = cinfo->min_DCT_h_scaled_size;
-    compptr->DCT_v_scaled_size = cinfo->min_DCT_v_scaled_size;
-  }
-
-#else /* !IDCT_SCALING_SUPPORTED */
-
-  /* Hardwire it to "no scaling" */
-  cinfo->output_width = cinfo->image_width;
-  cinfo->output_height = cinfo->image_height;
-  /* jdinput.c has already initialized DCT_scaled_size,
-   * and has computed unscaled downsampled_width and downsampled_height.
-   */
-
-#endif /* IDCT_SCALING_SUPPORTED */
-}
-#endif
-
-
 LOCAL(void)
 initial_setup (j_decompress_ptr cinfo)
 /* Called once, when first SOS marker is reached */
@@ -130,7 +59,7 @@
   /* Check that number of components won't exceed internal array sizes */
   if (cinfo->num_components > MAX_COMPONENTS)
     ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->num_components,
-	     MAX_COMPONENTS);
+             MAX_COMPONENTS);
 
   /* Compute maximum sampling factors; check factor validity */
   cinfo->max_h_samp_factor = 1;
@@ -138,12 +67,12 @@
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     if (compptr->h_samp_factor<=0 || compptr->h_samp_factor>MAX_SAMP_FACTOR ||
-	compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
+        compptr->v_samp_factor<=0 || compptr->v_samp_factor>MAX_SAMP_FACTOR)
       ERREXIT(cinfo, JERR_BAD_SAMPLING);
     cinfo->max_h_samp_factor = MAX(cinfo->max_h_samp_factor,
-				   compptr->h_samp_factor);
+                                   compptr->h_samp_factor);
     cinfo->max_v_samp_factor = MAX(cinfo->max_v_samp_factor,
-				   compptr->v_samp_factor);
+                                   compptr->v_samp_factor);
   }
 
 #if JPEG_LIB_VERSION >=80
@@ -173,10 +102,15 @@
     /* Size in DCT blocks */
     compptr->width_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->height_in_blocks = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+    /* Set the first and last MCU columns to decompress from multi-scan images.
+     * By default, decompress all of the MCU columns.
+     */
+    cinfo->master->first_MCU_col[ci] = 0;
+    cinfo->master->last_MCU_col[ci] = compptr->width_in_blocks - 1;
     /* downsampled_width and downsampled_height will also be overridden by
      * jdmaster.c if we are doing full decompression.  The transcoder library
      * doesn't use these values, but the calling application might.
@@ -184,10 +118,10 @@
     /* Size in samples */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width * (long) compptr->h_samp_factor,
-		    (long) cinfo->max_h_samp_factor);
+                    (long) cinfo->max_h_samp_factor);
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height * (long) compptr->v_samp_factor,
-		    (long) cinfo->max_v_samp_factor);
+                    (long) cinfo->max_v_samp_factor);
     /* Mark component needed, until color conversion says otherwise */
     compptr->component_needed = TRUE;
     /* Mark no quantization table yet saved for component */
@@ -197,7 +131,7 @@
   /* Compute number of fully interleaved MCU rows. */
   cinfo->total_iMCU_rows = (JDIMENSION)
     jdiv_round_up((long) cinfo->image_height,
-		  (long) (cinfo->max_v_samp_factor*DCTSIZE));
+                  (long) (cinfo->max_v_samp_factor*DCTSIZE));
 
   /* Decide whether file contains multiple scans */
   if (cinfo->comps_in_scan < cinfo->num_components || cinfo->progressive_mode)
@@ -214,16 +148,16 @@
 {
   int ci, mcublks, tmp;
   jpeg_component_info *compptr;
-  
+
   if (cinfo->comps_in_scan == 1) {
-    
+
     /* Noninterleaved (single-component) scan */
     compptr = cinfo->cur_comp_info[0];
-    
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = compptr->width_in_blocks;
     cinfo->MCU_rows_in_scan = compptr->height_in_blocks;
-    
+
     /* For noninterleaved scan, always one block per MCU */
     compptr->MCU_width = 1;
     compptr->MCU_height = 1;
@@ -236,28 +170,28 @@
     tmp = (int) (compptr->height_in_blocks % compptr->v_samp_factor);
     if (tmp == 0) tmp = compptr->v_samp_factor;
     compptr->last_row_height = tmp;
-    
+
     /* Prepare array describing MCU composition */
     cinfo->blocks_in_MCU = 1;
     cinfo->MCU_membership[0] = 0;
-    
+
   } else {
-    
+
     /* Interleaved (multi-component) scan */
     if (cinfo->comps_in_scan <= 0 || cinfo->comps_in_scan > MAX_COMPS_IN_SCAN)
       ERREXIT2(cinfo, JERR_COMPONENT_COUNT, cinfo->comps_in_scan,
-	       MAX_COMPS_IN_SCAN);
-    
+               MAX_COMPS_IN_SCAN);
+
     /* Overall image size in MCUs */
     cinfo->MCUs_per_row = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width,
-		    (long) (cinfo->max_h_samp_factor*DCTSIZE));
+                    (long) (cinfo->max_h_samp_factor*DCTSIZE));
     cinfo->MCU_rows_in_scan = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height,
-		    (long) (cinfo->max_v_samp_factor*DCTSIZE));
-    
+                    (long) (cinfo->max_v_samp_factor*DCTSIZE));
+
     cinfo->blocks_in_MCU = 0;
-    
+
     for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
       compptr = cinfo->cur_comp_info[ci];
       /* Sampling factors give # of blocks of component in each MCU */
@@ -275,12 +209,12 @@
       /* Prepare array describing MCU composition */
       mcublks = compptr->MCU_blocks;
       if (cinfo->blocks_in_MCU + mcublks > D_MAX_BLOCKS_IN_MCU)
-	ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
+        ERREXIT(cinfo, JERR_BAD_MCU_SIZE);
       while (mcublks-- > 0) {
-	cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
+        cinfo->MCU_membership[cinfo->blocks_in_MCU++] = ci;
       }
     }
-    
+
   }
 }
 
@@ -311,7 +245,7 @@
 {
   int ci, qtblno;
   jpeg_component_info *compptr;
-  JQUANT_TBL * qtbl;
+  JQUANT_TBL *qtbl;
 
   for (ci = 0; ci < cinfo->comps_in_scan; ci++) {
     compptr = cinfo->cur_comp_info[ci];
@@ -321,13 +255,13 @@
     /* Make sure specified quantization table is present */
     qtblno = compptr->quant_tbl_no;
     if (qtblno < 0 || qtblno >= NUM_QUANT_TBLS ||
-	cinfo->quant_tbl_ptrs[qtblno] == NULL)
+        cinfo->quant_tbl_ptrs[qtblno] == NULL)
       ERREXIT1(cinfo, JERR_NO_QUANT_TABLE, qtblno);
     /* OK, save away the quantization table */
     qtbl = (JQUANT_TBL *)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(JQUANT_TBL));
-    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], SIZEOF(JQUANT_TBL));
+                                  sizeof(JQUANT_TBL));
+    MEMCOPY(qtbl, cinfo->quant_tbl_ptrs[qtblno], sizeof(JQUANT_TBL));
     compptr->quant_table = qtbl;
   }
 }
@@ -386,31 +320,31 @@
   val = (*cinfo->marker->read_markers) (cinfo);
 
   switch (val) {
-  case JPEG_REACHED_SOS:	/* Found SOS */
-    if (inputctl->inheaders) {	/* 1st SOS */
+  case JPEG_REACHED_SOS:        /* Found SOS */
+    if (inputctl->inheaders) {  /* 1st SOS */
       initial_setup(cinfo);
       inputctl->inheaders = FALSE;
       /* Note: start_input_pass must be called by jdmaster.c
        * before any more input can be consumed.  jdapimin.c is
        * responsible for enforcing this sequencing.
        */
-    } else {			/* 2nd or later SOS marker */
+    } else {                    /* 2nd or later SOS marker */
       if (! inputctl->pub.has_multiple_scans)
-	ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
+        ERREXIT(cinfo, JERR_EOI_EXPECTED); /* Oops, I wasn't expecting this! */
       start_input_pass(cinfo);
     }
     break;
-  case JPEG_REACHED_EOI:	/* Found EOI */
+  case JPEG_REACHED_EOI:        /* Found EOI */
     inputctl->pub.eoi_reached = TRUE;
-    if (inputctl->inheaders) {	/* Tables-only datastream, apparently */
+    if (inputctl->inheaders) {  /* Tables-only datastream, apparently */
       if (cinfo->marker->saw_SOF)
-	ERREXIT(cinfo, JERR_SOF_NO_SOS);
+        ERREXIT(cinfo, JERR_SOF_NO_SOS);
     } else {
       /* Prevent infinite loop in coef ctlr's decompress_data routine
        * if user set output_scan_number larger than number of scans.
        */
       if (cinfo->output_scan_number > cinfo->input_scan_number)
-	cinfo->output_scan_number = cinfo->input_scan_number;
+        cinfo->output_scan_number = cinfo->input_scan_number;
     }
     break;
   case JPEG_SUSPENDED:
@@ -455,7 +389,7 @@
   /* Create subobject in permanent pool */
   inputctl = (my_inputctl_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				SIZEOF(my_input_controller));
+                                sizeof(my_input_controller));
   cinfo->inputctl = (struct jpeg_input_controller *) inputctl;
   /* Initialize method pointers */
   inputctl->pub.consume_input = consume_markers;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmainct.c glmark2-2021.02/src/libjpeg-turbo/jdmainct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmainct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmainct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdmainct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the main buffer controller for decompression.
  * The main buffer lies between the JPEG decompressor proper and the
@@ -14,10 +16,8 @@
  * supplies the equivalent of the main buffer in that case.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
-#include "jpegcomp.h"
+#include "jdmainct.h"
 
 
 /*
@@ -111,47 +111,17 @@
  */
 
 
-/* Private buffer controller object */
-
-typedef struct {
-  struct jpeg_d_main_controller pub; /* public fields */
-
-  /* Pointer to allocated workspace (M or M+2 row groups). */
-  JSAMPARRAY buffer[MAX_COMPONENTS];
-
-  boolean buffer_full;		/* Have we gotten an iMCU row from decoder? */
-  JDIMENSION rowgroup_ctr;	/* counts row groups output to postprocessor */
-
-  /* Remaining fields are only used in the context case. */
-
-  /* These are the master pointers to the funny-order pointer lists. */
-  JSAMPIMAGE xbuffer[2];	/* pointers to weird pointer lists */
-
-  int whichptr;			/* indicates which pointer set is now in use */
-  int context_state;		/* process_data state machine status */
-  JDIMENSION rowgroups_avail;	/* row groups available to postprocessor */
-  JDIMENSION iMCU_row_ctr;	/* counts iMCU rows to detect image top/bot */
-} my_main_controller;
-
-typedef my_main_controller * my_main_ptr;
-
-/* context_state values: */
-#define CTX_PREPARE_FOR_IMCU	0	/* need to prepare for MCU row */
-#define CTX_PROCESS_IMCU	1	/* feeding iMCU to postprocessor */
-#define CTX_POSTPONED_ROW	2	/* feeding postponed row group */
-
-
 /* Forward declarations */
 METHODDEF(void) process_data_simple_main
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 METHODDEF(void) process_data_context_main
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 #ifdef QUANT_2PASS_SUPPORTED
 METHODDEF(void) process_data_crank_post
-	JPP((j_decompress_ptr cinfo, JSAMPARRAY output_buf,
-	     JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+         JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 #endif
 
 
@@ -172,7 +142,7 @@
    */
   main_ptr->xbuffer[0] = (JSAMPIMAGE)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				cinfo->num_components * 2 * SIZEOF(JSAMPARRAY));
+                                cinfo->num_components * 2 * sizeof(JSAMPARRAY));
   main_ptr->xbuffer[1] = main_ptr->xbuffer[0] + cinfo->num_components;
 
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
@@ -184,8 +154,8 @@
      */
     xbuf = (JSAMPARRAY)
       (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  2 * (rgroup * (M + 4)) * SIZEOF(JSAMPROW));
-    xbuf += rgroup;		/* want one row group at negative offsets */
+                                  2 * (rgroup * (M + 4)) * sizeof(JSAMPROW));
+    xbuf += rgroup;             /* want one row group at negative offsets */
     main_ptr->xbuffer[0][ci] = xbuf;
     xbuf += rgroup * (M + 4);
     main_ptr->xbuffer[1][ci] = xbuf;
@@ -237,34 +207,6 @@
 
 
 LOCAL(void)
-set_wraparound_pointers (j_decompress_ptr cinfo)
-/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
- * This changes the pointer list state from top-of-image to the normal state.
- */
-{
-  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
-  int ci, i, rgroup;
-  int M = cinfo->_min_DCT_scaled_size;
-  jpeg_component_info *compptr;
-  JSAMPARRAY xbuf0, xbuf1;
-
-  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-       ci++, compptr++) {
-    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
-      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
-    xbuf0 = main_ptr->xbuffer[0][ci];
-    xbuf1 = main_ptr->xbuffer[1][ci];
-    for (i = 0; i < rgroup; i++) {
-      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
-      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
-      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
-      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
-    }
-  }
-}
-
-
-LOCAL(void)
 set_bottom_pointers (j_decompress_ptr cinfo)
 /* Change the pointer lists to duplicate the last sample row at the bottom
  * of the image.  whichptr indicates which xbuffer holds the final iMCU row.
@@ -315,14 +257,14 @@
     if (cinfo->upsample->need_context_rows) {
       main_ptr->pub.process_data = process_data_context_main;
       make_funny_pointers(cinfo); /* Create the xbuffer[] lists */
-      main_ptr->whichptr = 0;	/* Read first iMCU row into xbuffer[0] */
+      main_ptr->whichptr = 0;   /* Read first iMCU row into xbuffer[0] */
       main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
       main_ptr->iMCU_row_ctr = 0;
     } else {
       /* Simple case with no context needed */
       main_ptr->pub.process_data = process_data_simple_main;
     }
-    main_ptr->buffer_full = FALSE;	/* Mark buffer empty */
+    main_ptr->buffer_full = FALSE;      /* Mark buffer empty */
     main_ptr->rowgroup_ctr = 0;
     break;
 #ifdef QUANT_2PASS_SUPPORTED
@@ -345,8 +287,8 @@
 
 METHODDEF(void)
 process_data_simple_main (j_decompress_ptr cinfo,
-			  JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			  JDIMENSION out_rows_avail)
+                          JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                          JDIMENSION out_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
   JDIMENSION rowgroups_avail;
@@ -354,8 +296,8 @@
   /* Read input data if we haven't filled the main buffer yet */
   if (! main_ptr->buffer_full) {
     if (! (*cinfo->coef->decompress_data) (cinfo, main_ptr->buffer))
-      return;			/* suspension forced, can do nothing more */
-    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
   }
 
   /* There are always min_DCT_scaled_size row groups in an iMCU row. */
@@ -367,8 +309,8 @@
 
   /* Feed the postprocessor */
   (*cinfo->post->post_process_data) (cinfo, main_ptr->buffer,
-				     &main_ptr->rowgroup_ctr, rowgroups_avail,
-				     output_buf, out_row_ctr, out_rows_avail);
+                                     &main_ptr->rowgroup_ctr, rowgroups_avail,
+                                     output_buf, out_row_ctr, out_rows_avail);
 
   /* Has postprocessor consumed all the data yet? If so, mark buffer empty */
   if (main_ptr->rowgroup_ctr >= rowgroups_avail) {
@@ -385,18 +327,18 @@
 
 METHODDEF(void)
 process_data_context_main (j_decompress_ptr cinfo,
-			   JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			   JDIMENSION out_rows_avail)
+                           JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                           JDIMENSION out_rows_avail)
 {
   my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
 
   /* Read input data if we haven't filled the main buffer yet */
   if (! main_ptr->buffer_full) {
     if (! (*cinfo->coef->decompress_data) (cinfo,
-					   main_ptr->xbuffer[main_ptr->whichptr]))
-      return;			/* suspension forced, can do nothing more */
-    main_ptr->buffer_full = TRUE;	/* OK, we have an iMCU row to work with */
-    main_ptr->iMCU_row_ctr++;	/* count rows received */
+                                           main_ptr->xbuffer[main_ptr->whichptr]))
+      return;                   /* suspension forced, can do nothing more */
+    main_ptr->buffer_full = TRUE;       /* OK, we have an iMCU row to work with */
+    main_ptr->iMCU_row_ctr++;   /* count rows received */
   }
 
   /* Postprocessor typically will not swallow all the input data it is handed
@@ -408,13 +350,13 @@
   case CTX_POSTPONED_ROW:
     /* Call postprocessor using previously set pointers for postponed row */
     (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
-			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
-			output_buf, out_row_ctr, out_rows_avail);
+                        &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+                        output_buf, out_row_ctr, out_rows_avail);
     if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
-      return;			/* Need to suspend */
+      return;                   /* Need to suspend */
     main_ptr->context_state = CTX_PREPARE_FOR_IMCU;
     if (*out_row_ctr >= out_rows_avail)
-      return;			/* Postprocessor exactly filled output buf */
+      return;                   /* Postprocessor exactly filled output buf */
     /*FALLTHROUGH*/
   case CTX_PREPARE_FOR_IMCU:
     /* Prepare to process first M-1 row groups of this iMCU row */
@@ -430,15 +372,15 @@
   case CTX_PROCESS_IMCU:
     /* Call postprocessor using previously set pointers */
     (*cinfo->post->post_process_data) (cinfo, main_ptr->xbuffer[main_ptr->whichptr],
-			&main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
-			output_buf, out_row_ctr, out_rows_avail);
+                        &main_ptr->rowgroup_ctr, main_ptr->rowgroups_avail,
+                        output_buf, out_row_ctr, out_rows_avail);
     if (main_ptr->rowgroup_ctr < main_ptr->rowgroups_avail)
-      return;			/* Need to suspend */
+      return;                   /* Need to suspend */
     /* After the first iMCU, change wraparound pointers to normal state */
     if (main_ptr->iMCU_row_ctr == 1)
       set_wraparound_pointers(cinfo);
     /* Prepare to load new iMCU row using other xbuffer list */
-    main_ptr->whichptr ^= 1;	/* 0=>1 or 1=>0 */
+    main_ptr->whichptr ^= 1;    /* 0=>1 or 1=>0 */
     main_ptr->buffer_full = FALSE;
     /* Still need to process last row group of this iMCU row, */
     /* which is saved at index M+1 of the other xbuffer */
@@ -459,12 +401,12 @@
 
 METHODDEF(void)
 process_data_crank_post (j_decompress_ptr cinfo,
-			 JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			 JDIMENSION out_rows_avail)
+                         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                         JDIMENSION out_rows_avail)
 {
   (*cinfo->post->post_process_data) (cinfo, (JSAMPIMAGE) NULL,
-				     (JDIMENSION *) NULL, (JDIMENSION) 0,
-				     output_buf, out_row_ctr, out_rows_avail);
+                                     (JDIMENSION *) NULL, (JDIMENSION) 0,
+                                     output_buf, out_row_ctr, out_rows_avail);
 }
 
 #endif /* QUANT_2PASS_SUPPORTED */
@@ -483,11 +425,11 @@
 
   main_ptr = (my_main_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_main_controller));
+                                sizeof(my_main_controller));
   cinfo->main = (struct jpeg_d_main_controller *) main_ptr;
   main_ptr->pub.start_pass = start_pass_main;
 
-  if (need_full_buffer)		/* shouldn't happen */
+  if (need_full_buffer)         /* shouldn't happen */
     ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 
   /* Allocate the workspace.
@@ -507,8 +449,8 @@
     rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
       cinfo->_min_DCT_scaled_size; /* height of a row group of component */
     main_ptr->buffer[ci] = (*cinfo->mem->alloc_sarray)
-			((j_common_ptr) cinfo, JPOOL_IMAGE,
-			 compptr->width_in_blocks * compptr->_DCT_scaled_size,
-			 (JDIMENSION) (rgroup * ngroups));
+                        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                         compptr->width_in_blocks * compptr->_DCT_scaled_size,
+                         (JDIMENSION) (rgroup * ngroups));
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmainct.h glmark2-2021.02/src/libjpeg-turbo/jdmainct.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmainct.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmainct.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,71 @@
+/*
+ * jdmainct.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+#include "jpegcomp.h"
+
+
+/* Private buffer controller object */
+
+typedef struct {
+  struct jpeg_d_main_controller pub; /* public fields */
+
+  /* Pointer to allocated workspace (M or M+2 row groups). */
+  JSAMPARRAY buffer[MAX_COMPONENTS];
+
+  boolean buffer_full;          /* Have we gotten an iMCU row from decoder? */
+  JDIMENSION rowgroup_ctr;      /* counts row groups output to postprocessor */
+
+  /* Remaining fields are only used in the context case. */
+
+  /* These are the master pointers to the funny-order pointer lists. */
+  JSAMPIMAGE xbuffer[2];        /* pointers to weird pointer lists */
+
+  int whichptr;                 /* indicates which pointer set is now in use */
+  int context_state;            /* process_data state machine status */
+  JDIMENSION rowgroups_avail;   /* row groups available to postprocessor */
+  JDIMENSION iMCU_row_ctr;      /* counts iMCU rows to detect image top/bot */
+} my_main_controller;
+
+typedef my_main_controller *my_main_ptr;
+
+
+/* context_state values: */
+#define CTX_PREPARE_FOR_IMCU    0       /* need to prepare for MCU row */
+#define CTX_PROCESS_IMCU        1       /* feeding iMCU to postprocessor */
+#define CTX_POSTPONED_ROW       2       /* feeding postponed row group */
+
+
+LOCAL(void)
+set_wraparound_pointers (j_decompress_ptr cinfo)
+/* Set up the "wraparound" pointers at top and bottom of the pointer lists.
+ * This changes the pointer list state from top-of-image to the normal state.
+ */
+{
+  my_main_ptr main_ptr = (my_main_ptr) cinfo->main;
+  int ci, i, rgroup;
+  int M = cinfo->_min_DCT_scaled_size;
+  jpeg_component_info *compptr;
+  JSAMPARRAY xbuf0, xbuf1;
+
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    rgroup = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
+      cinfo->_min_DCT_scaled_size; /* height of a row group of component */
+    xbuf0 = main_ptr->xbuffer[0][ci];
+    xbuf1 = main_ptr->xbuffer[1][ci];
+    for (i = 0; i < rgroup; i++) {
+      xbuf0[i - rgroup] = xbuf0[rgroup*(M+1) + i];
+      xbuf1[i - rgroup] = xbuf1[rgroup*(M+1) + i];
+      xbuf0[rgroup*(M+2) + i] = xbuf0[i];
+      xbuf1[rgroup*(M+2) + i] = xbuf1[i];
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmarker.c glmark2-2021.02/src/libjpeg-turbo/jdmarker.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmarker.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmarker.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jdmarker.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * Copyright (C) 2012, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2012, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains routines to decode JPEG datastream markers.
  * Most of the complexity arises from our desire to support input
@@ -18,29 +20,29 @@
 #include "jpeglib.h"
 
 
-typedef enum {			/* JPEG marker codes */
+typedef enum {                  /* JPEG marker codes */
   M_SOF0  = 0xc0,
   M_SOF1  = 0xc1,
   M_SOF2  = 0xc2,
   M_SOF3  = 0xc3,
-  
+
   M_SOF5  = 0xc5,
   M_SOF6  = 0xc6,
   M_SOF7  = 0xc7,
-  
+
   M_JPG   = 0xc8,
   M_SOF9  = 0xc9,
   M_SOF10 = 0xca,
   M_SOF11 = 0xcb,
-  
+
   M_SOF13 = 0xcd,
   M_SOF14 = 0xce,
   M_SOF15 = 0xcf,
-  
+
   M_DHT   = 0xc4,
-  
+
   M_DAC   = 0xcc,
-  
+
   M_RST0  = 0xd0,
   M_RST1  = 0xd1,
   M_RST2  = 0xd2,
@@ -49,7 +51,7 @@
   M_RST5  = 0xd5,
   M_RST6  = 0xd6,
   M_RST7  = 0xd7,
-  
+
   M_SOI   = 0xd8,
   M_EOI   = 0xd9,
   M_SOS   = 0xda,
@@ -58,7 +60,7 @@
   M_DRI   = 0xdd,
   M_DHP   = 0xde,
   M_EXP   = 0xdf,
-  
+
   M_APP0  = 0xe0,
   M_APP1  = 0xe1,
   M_APP2  = 0xe2,
@@ -75,13 +77,13 @@
   M_APP13 = 0xed,
   M_APP14 = 0xee,
   M_APP15 = 0xef,
-  
+
   M_JPG0  = 0xf0,
   M_JPG13 = 0xfd,
   M_COM   = 0xfe,
-  
+
   M_TEM   = 0x01,
-  
+
   M_ERROR = 0x100
 } JPEG_MARKER;
 
@@ -100,12 +102,12 @@
   unsigned int length_limit_APPn[16];
 
   /* Status of COM/APPn marker saving */
-  jpeg_saved_marker_ptr cur_marker;	/* NULL if not processing a marker */
-  unsigned int bytes_read;		/* data bytes read so far in marker */
+  jpeg_saved_marker_ptr cur_marker;     /* NULL if not processing a marker */
+  unsigned int bytes_read;              /* data bytes read so far in marker */
   /* Note: cur_marker is not linked into marker_list until it's all read. */
 } my_marker_reader;
 
-typedef my_marker_reader * my_marker_ptr;
+typedef my_marker_reader *my_marker_ptr;
 
 
 /*
@@ -118,49 +120,49 @@
 
 /* Declare and initialize local copies of input pointer/count */
 #define INPUT_VARS(cinfo)  \
-	struct jpeg_source_mgr * datasrc = (cinfo)->src;  \
-	const JOCTET * next_input_byte = datasrc->next_input_byte;  \
-	size_t bytes_in_buffer = datasrc->bytes_in_buffer
+        struct jpeg_source_mgr *datasrc = (cinfo)->src;  \
+        const JOCTET *next_input_byte = datasrc->next_input_byte;  \
+        size_t bytes_in_buffer = datasrc->bytes_in_buffer
 
 /* Unload the local copies --- do this only at a restart boundary */
 #define INPUT_SYNC(cinfo)  \
-	( datasrc->next_input_byte = next_input_byte,  \
-	  datasrc->bytes_in_buffer = bytes_in_buffer )
+        ( datasrc->next_input_byte = next_input_byte,  \
+          datasrc->bytes_in_buffer = bytes_in_buffer )
 
 /* Reload the local copies --- used only in MAKE_BYTE_AVAIL */
 #define INPUT_RELOAD(cinfo)  \
-	( next_input_byte = datasrc->next_input_byte,  \
-	  bytes_in_buffer = datasrc->bytes_in_buffer )
+        ( next_input_byte = datasrc->next_input_byte,  \
+          bytes_in_buffer = datasrc->bytes_in_buffer )
 
 /* Internal macro for INPUT_BYTE and INPUT_2BYTES: make a byte available.
  * Note we do *not* do INPUT_SYNC before calling fill_input_buffer,
  * but we must reload the local copies after a successful fill.
  */
 #define MAKE_BYTE_AVAIL(cinfo,action)  \
-	if (bytes_in_buffer == 0) {  \
-	  if (! (*datasrc->fill_input_buffer) (cinfo))  \
-	    { action; }  \
-	  INPUT_RELOAD(cinfo);  \
-	}
+        if (bytes_in_buffer == 0) {  \
+          if (! (*datasrc->fill_input_buffer) (cinfo))  \
+            { action; }  \
+          INPUT_RELOAD(cinfo);  \
+        }
 
 /* Read a byte into variable V.
  * If must suspend, take the specified action (typically "return FALSE").
  */
 #define INPUT_BYTE(cinfo,V,action)  \
-	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V = GETJOCTET(*next_input_byte++); )
+        MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V = GETJOCTET(*next_input_byte++); )
 
 /* As above, but read two bytes interpreted as an unsigned 16-bit integer.
- * V should be declared unsigned int or perhaps INT32.
+ * V should be declared unsigned int or perhaps JLONG.
  */
 #define INPUT_2BYTES(cinfo,V,action)  \
-	MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
-		  MAKE_BYTE_AVAIL(cinfo,action); \
-		  bytes_in_buffer--; \
-		  V += GETJOCTET(*next_input_byte++); )
+        MAKESTMT( MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V = ((unsigned int) GETJOCTET(*next_input_byte++)) << 8; \
+                  MAKE_BYTE_AVAIL(cinfo,action); \
+                  bytes_in_buffer--; \
+                  V += GETJOCTET(*next_input_byte++); )
 
 
 /*
@@ -199,7 +201,7 @@
 /* Process an SOI marker */
 {
   int i;
-  
+
   TRACEMS(cinfo, 1, JTRC_SOI);
 
   if (cinfo->marker->saw_SOI)
@@ -238,9 +240,9 @@
 get_sof (j_decompress_ptr cinfo, boolean is_prog, boolean is_arith)
 /* Process a SOFn marker */
 {
-  INT32 length;
+  JLONG length;
   int c, ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   INPUT_VARS(cinfo);
 
   cinfo->progressive_mode = is_prog;
@@ -256,8 +258,8 @@
   length -= 8;
 
   TRACEMS4(cinfo, 1, JTRC_SOF, cinfo->unread_marker,
-	   (int) cinfo->image_width, (int) cinfo->image_height,
-	   cinfo->num_components);
+           (int) cinfo->image_width, (int) cinfo->image_height,
+           cinfo->num_components);
 
   if (cinfo->marker->saw_SOF)
     ERREXIT(cinfo, JERR_SOF_DUPLICATE);
@@ -272,11 +274,11 @@
   if (length != (cinfo->num_components * 3))
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
-  if (cinfo->comp_info == NULL)	/* do only once, even if suspend */
+  if (cinfo->comp_info == NULL) /* do only once, even if suspend */
     cinfo->comp_info = (jpeg_component_info *) (*cinfo->mem->alloc_small)
-			((j_common_ptr) cinfo, JPOOL_IMAGE,
-			 cinfo->num_components * SIZEOF(jpeg_component_info));
-  
+                        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                         cinfo->num_components * sizeof(jpeg_component_info));
+
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     compptr->component_index = ci;
@@ -287,8 +289,8 @@
     INPUT_BYTE(cinfo, compptr->quant_tbl_no, return FALSE);
 
     TRACEMS4(cinfo, 1, JTRC_SOF_COMPONENT,
-	     compptr->component_id, compptr->h_samp_factor,
-	     compptr->v_samp_factor, compptr->quant_tbl_no);
+             compptr->component_id, compptr->h_samp_factor,
+             compptr->v_samp_factor, compptr->quant_tbl_no);
   }
 
   cinfo->marker->saw_SOF = TRUE;
@@ -302,9 +304,9 @@
 get_sos (j_decompress_ptr cinfo)
 /* Process a SOS marker */
 {
-  INT32 length;
-  int i, ci, n, c, cc;
-  jpeg_component_info * compptr;
+  JLONG length;
+  int i, ci, n, c, cc, pi;
+  jpeg_component_info *compptr;
   INPUT_VARS(cinfo);
 
   if (! cinfo->marker->saw_SOF)
@@ -323,17 +325,18 @@
 
   /* Collect the component-spec parameters */
 
-  for (i = 0; i < cinfo->num_components; i++)
+  for (i = 0; i < MAX_COMPS_IN_SCAN; i++)
     cinfo->cur_comp_info[i] = NULL;
 
   for (i = 0; i < n; i++) {
     INPUT_BYTE(cinfo, cc, return FALSE);
     INPUT_BYTE(cinfo, c, return FALSE);
-    
-    for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+
+    for (ci = 0, compptr = cinfo->comp_info;
+         ci < cinfo->num_components && ci < MAX_COMPS_IN_SCAN;
+         ci++, compptr++) {
       if (cc == compptr->component_id && !cinfo->cur_comp_info[ci])
-	goto id_found;
+        goto id_found;
     }
 
     ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
@@ -343,9 +346,16 @@
     cinfo->cur_comp_info[i] = compptr;
     compptr->dc_tbl_no = (c >> 4) & 15;
     compptr->ac_tbl_no = (c     ) & 15;
-    
+
     TRACEMS3(cinfo, 1, JTRC_SOS_COMPONENT, cc,
-	     compptr->dc_tbl_no, compptr->ac_tbl_no);
+             compptr->dc_tbl_no, compptr->ac_tbl_no);
+
+    /* This CSi (cc) should differ from the previous CSi */
+    for (pi = 0; pi < i; pi++) {
+      if (cinfo->cur_comp_info[pi] == compptr) {
+        ERREXIT1(cinfo, JERR_BAD_COMPONENT_ID, cc);
+      }
+    }
   }
 
   /* Collect the additional scan parameters Ss, Se, Ah/Al. */
@@ -358,7 +368,7 @@
   cinfo->Al = (c     ) & 15;
 
   TRACEMS4(cinfo, 1, JTRC_SOS_PARAMS, cinfo->Ss, cinfo->Se,
-	   cinfo->Ah, cinfo->Al);
+           cinfo->Ah, cinfo->Al);
 
   /* Prepare to scan data & restart markers */
   cinfo->marker->next_restart_num = 0;
@@ -377,13 +387,13 @@
 get_dac (j_decompress_ptr cinfo)
 /* Process a DAC marker */
 {
-  INT32 length;
+  JLONG length;
   int index, val;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   while (length > 0) {
     INPUT_BYTE(cinfo, index, return FALSE);
     INPUT_BYTE(cinfo, val, return FALSE);
@@ -397,11 +407,11 @@
 
     if (index >= NUM_ARITH_TBLS) { /* define AC table */
       cinfo->arith_ac_K[index-NUM_ARITH_TBLS] = (UINT8) val;
-    } else {			/* define DC table */
+    } else {                    /* define DC table */
       cinfo->arith_dc_L[index] = (UINT8) (val & 0x0F);
       cinfo->arith_dc_U[index] = (UINT8) (val >> 4);
       if (cinfo->arith_dc_L[index] > cinfo->arith_dc_U[index])
-	ERREXIT1(cinfo, JERR_DAC_VALUE, val);
+        ERREXIT1(cinfo, JERR_DAC_VALUE, val);
     }
   }
 
@@ -423,7 +433,7 @@
 get_dht (j_decompress_ptr cinfo)
 /* Process a DHT marker */
 {
-  INT32 length;
+  JLONG length;
   UINT8 bits[17];
   UINT8 huffval[256];
   int i, index, count;
@@ -432,12 +442,12 @@
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   while (length > 16) {
     INPUT_BYTE(cinfo, index, return FALSE);
 
     TRACEMS1(cinfo, 1, JTRC_DHT, index);
-      
+
     bits[0] = 0;
     count = 0;
     for (i = 1; i <= 16; i++) {
@@ -448,38 +458,41 @@
     length -= 1 + 16;
 
     TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
-	     bits[1], bits[2], bits[3], bits[4],
-	     bits[5], bits[6], bits[7], bits[8]);
+             bits[1], bits[2], bits[3], bits[4],
+             bits[5], bits[6], bits[7], bits[8]);
     TRACEMS8(cinfo, 2, JTRC_HUFFBITS,
-	     bits[9], bits[10], bits[11], bits[12],
-	     bits[13], bits[14], bits[15], bits[16]);
+             bits[9], bits[10], bits[11], bits[12],
+             bits[13], bits[14], bits[15], bits[16]);
 
     /* Here we just do minimal validation of the counts to avoid walking
      * off the end of our table space.  jdhuff.c will check more carefully.
      */
-    if (count > 256 || ((INT32) count) > length)
+    if (count > 256 || ((JLONG) count) > length)
       ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
 
     for (i = 0; i < count; i++)
       INPUT_BYTE(cinfo, huffval[i], return FALSE);
 
+    MEMZERO(&huffval[count], (256 - count) * sizeof(UINT8));
+
     length -= count;
 
-    if (index & 0x10) {		/* AC table definition */
+    if (index & 0x10) {         /* AC table definition */
       index -= 0x10;
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
       htblptr = &cinfo->ac_huff_tbl_ptrs[index];
-    } else {			/* DC table definition */
+    } else {                    /* DC table definition */
+      if (index < 0 || index >= NUM_HUFF_TBLS)
+        ERREXIT1(cinfo, JERR_DHT_INDEX, index);
       htblptr = &cinfo->dc_huff_tbl_ptrs[index];
     }
 
-    if (index < 0 || index >= NUM_HUFF_TBLS)
-      ERREXIT1(cinfo, JERR_DHT_INDEX, index);
-
     if (*htblptr == NULL)
       *htblptr = jpeg_alloc_huff_table((j_common_ptr) cinfo);
-  
-    MEMCOPY((*htblptr)->bits, bits, SIZEOF((*htblptr)->bits));
-    MEMCOPY((*htblptr)->huffval, huffval, SIZEOF((*htblptr)->huffval));
+
+    MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+    MEMCOPY((*htblptr)->huffval, huffval, sizeof((*htblptr)->huffval));
   }
 
   if (length != 0)
@@ -494,7 +507,7 @@
 get_dqt (j_decompress_ptr cinfo)
 /* Process a DQT marker */
 {
-  INT32 length;
+  JLONG length;
   int n, i, prec;
   unsigned int tmp;
   JQUANT_TBL *quant_ptr;
@@ -512,27 +525,27 @@
 
     if (n >= NUM_QUANT_TBLS)
       ERREXIT1(cinfo, JERR_DQT_INDEX, n);
-      
+
     if (cinfo->quant_tbl_ptrs[n] == NULL)
       cinfo->quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) cinfo);
     quant_ptr = cinfo->quant_tbl_ptrs[n];
 
     for (i = 0; i < DCTSIZE2; i++) {
       if (prec)
-	INPUT_2BYTES(cinfo, tmp, return FALSE);
+        INPUT_2BYTES(cinfo, tmp, return FALSE);
       else
-	INPUT_BYTE(cinfo, tmp, return FALSE);
+        INPUT_BYTE(cinfo, tmp, return FALSE);
       /* We convert the zigzag-order table to natural array order. */
       quant_ptr->quantval[jpeg_natural_order[i]] = (UINT16) tmp;
     }
 
     if (cinfo->err->trace_level >= 2) {
       for (i = 0; i < DCTSIZE2; i += 8) {
-	TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
-		 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
-		 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
-		 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
-		 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
+        TRACEMS8(cinfo, 2, JTRC_QUANTVALS,
+                 quant_ptr->quantval[i],   quant_ptr->quantval[i+1],
+                 quant_ptr->quantval[i+2], quant_ptr->quantval[i+3],
+                 quant_ptr->quantval[i+4], quant_ptr->quantval[i+5],
+                 quant_ptr->quantval[i+6], quant_ptr->quantval[i+7]);
       }
     }
 
@@ -552,12 +565,12 @@
 get_dri (j_decompress_ptr cinfo)
 /* Process a DRI marker */
 {
-  INT32 length;
+  JLONG length;
   unsigned int tmp;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
-  
+
   if (length != 4)
     ERREXIT(cinfo, JERR_BAD_LENGTH);
 
@@ -579,20 +592,20 @@
  * JFIF and Adobe markers, respectively.
  */
 
-#define APP0_DATA_LEN	14	/* Length of interesting data in APP0 */
-#define APP14_DATA_LEN	12	/* Length of interesting data in APP14 */
-#define APPN_DATA_LEN	14	/* Must be the largest of the above!! */
+#define APP0_DATA_LEN   14      /* Length of interesting data in APP0 */
+#define APP14_DATA_LEN  12      /* Length of interesting data in APP14 */
+#define APPN_DATA_LEN   14      /* Must be the largest of the above!! */
 
 
 LOCAL(void)
-examine_app0 (j_decompress_ptr cinfo, JOCTET FAR * data,
-	      unsigned int datalen, INT32 remaining)
+examine_app0 (j_decompress_ptr cinfo, JOCTET *data,
+              unsigned int datalen, JLONG remaining)
 /* Examine first few bytes from an APP0.
  * Take appropriate action if it is a JFIF marker.
  * datalen is # of bytes at data[], remaining is length of rest of marker data.
  */
 {
-  INT32 totallen = (INT32) datalen + remaining;
+  JLONG totallen = (JLONG) datalen + remaining;
 
   if (datalen >= APP0_DATA_LEN &&
       GETJOCTET(data[0]) == 0x4A &&
@@ -615,18 +628,18 @@
      */
     if (cinfo->JFIF_major_version != 1)
       WARNMS2(cinfo, JWRN_JFIF_MAJOR,
-	      cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
+              cinfo->JFIF_major_version, cinfo->JFIF_minor_version);
     /* Generate trace messages */
     TRACEMS5(cinfo, 1, JTRC_JFIF,
-	     cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
-	     cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
+             cinfo->JFIF_major_version, cinfo->JFIF_minor_version,
+             cinfo->X_density, cinfo->Y_density, cinfo->density_unit);
     /* Validate thumbnail dimensions and issue appropriate messages */
     if (GETJOCTET(data[12]) | GETJOCTET(data[13]))
       TRACEMS2(cinfo, 1, JTRC_JFIF_THUMBNAIL,
-	       GETJOCTET(data[12]), GETJOCTET(data[13]));
+               GETJOCTET(data[12]), GETJOCTET(data[13]));
     totallen -= APP0_DATA_LEN;
     if (totallen !=
-	((INT32)GETJOCTET(data[12]) * (INT32)GETJOCTET(data[13]) * (INT32) 3))
+        ((JLONG)GETJOCTET(data[12]) * (JLONG)GETJOCTET(data[13]) * (JLONG) 3))
       TRACEMS1(cinfo, 1, JTRC_JFIF_BADTHUMBNAILSIZE, (int) totallen);
   } else if (datalen >= 6 &&
       GETJOCTET(data[0]) == 0x4A &&
@@ -650,7 +663,7 @@
       break;
     default:
       TRACEMS2(cinfo, 1, JTRC_JFIF_EXTENSION,
-	       GETJOCTET(data[5]), (int) totallen);
+               GETJOCTET(data[5]), (int) totallen);
       break;
     }
   } else {
@@ -661,8 +674,8 @@
 
 
 LOCAL(void)
-examine_app14 (j_decompress_ptr cinfo, JOCTET FAR * data,
-	       unsigned int datalen, INT32 remaining)
+examine_app14 (j_decompress_ptr cinfo, JOCTET *data,
+               unsigned int datalen, JLONG remaining)
 /* Examine first few bytes from an APP14.
  * Take appropriate action if it is an Adobe marker.
  * datalen is # of bytes at data[], remaining is length of rest of marker data.
@@ -695,7 +708,7 @@
 get_interesting_appn (j_decompress_ptr cinfo)
 /* Process an APP0 or APP14 marker without saving it */
 {
-  INT32 length;
+  JLONG length;
   JOCTET b[APPN_DATA_LEN];
   unsigned int i, numtoread;
   INPUT_VARS(cinfo);
@@ -717,10 +730,10 @@
   /* process it */
   switch (cinfo->unread_marker) {
   case M_APP0:
-    examine_app0(cinfo, (JOCTET FAR *) b, numtoread, length);
+    examine_app0(cinfo, (JOCTET *) b, numtoread, length);
     break;
   case M_APP14:
-    examine_app14(cinfo, (JOCTET FAR *) b, numtoread, length);
+    examine_app14(cinfo, (JOCTET *) b, numtoread, length);
     break;
   default:
     /* can't get here unless jpeg_save_markers chooses wrong processor */
@@ -746,33 +759,33 @@
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
   jpeg_saved_marker_ptr cur_marker = marker->cur_marker;
   unsigned int bytes_read, data_length;
-  JOCTET FAR * data;
-  INT32 length = 0;
+  JOCTET *data;
+  JLONG length = 0;
   INPUT_VARS(cinfo);
 
   if (cur_marker == NULL) {
     /* begin reading a marker */
     INPUT_2BYTES(cinfo, length, return FALSE);
     length -= 2;
-    if (length >= 0) {		/* watch out for bogus length word */
+    if (length >= 0) {          /* watch out for bogus length word */
       /* figure out how much we want to save */
       unsigned int limit;
       if (cinfo->unread_marker == (int) M_COM)
-	limit = marker->length_limit_COM;
+        limit = marker->length_limit_COM;
       else
-	limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
+        limit = marker->length_limit_APPn[cinfo->unread_marker - (int) M_APP0];
       if ((unsigned int) length < limit)
-	limit = (unsigned int) length;
+        limit = (unsigned int) length;
       /* allocate and initialize the marker item */
       cur_marker = (jpeg_saved_marker_ptr)
-	(*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				    SIZEOF(struct jpeg_marker_struct) + limit);
+        (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                    sizeof(struct jpeg_marker_struct) + limit);
       cur_marker->next = NULL;
       cur_marker->marker = (UINT8) cinfo->unread_marker;
       cur_marker->original_length = (unsigned int) length;
       cur_marker->data_length = limit;
       /* data area is just beyond the jpeg_marker_struct */
-      data = cur_marker->data = (JOCTET FAR *) (cur_marker + 1);
+      data = cur_marker->data = (JOCTET *) (cur_marker + 1);
       marker->cur_marker = cur_marker;
       marker->bytes_read = 0;
       bytes_read = 0;
@@ -790,7 +803,7 @@
   }
 
   while (bytes_read < data_length) {
-    INPUT_SYNC(cinfo);		/* move the restart point to here */
+    INPUT_SYNC(cinfo);          /* move the restart point to here */
     marker->bytes_read = bytes_read;
     /* If there's not at least one byte in buffer, suspend */
     MAKE_BYTE_AVAIL(cinfo, return FALSE);
@@ -803,14 +816,14 @@
   }
 
   /* Done reading what we want to read */
-  if (cur_marker != NULL) {	/* will be NULL if bogus length word */
+  if (cur_marker != NULL) {     /* will be NULL if bogus length word */
     /* Add new marker to end of list */
     if (cinfo->marker_list == NULL) {
       cinfo->marker_list = cur_marker;
     } else {
       jpeg_saved_marker_ptr prev = cinfo->marker_list;
       while (prev->next != NULL)
-	prev = prev->next;
+        prev = prev->next;
       prev->next = cur_marker;
     }
     /* Reset pointer & calc remaining data length */
@@ -830,12 +843,12 @@
     break;
   default:
     TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker,
-	     (int) (data_length + length));
+             (int) (data_length + length));
     break;
   }
 
   /* skip any remaining data -- could be lots */
-  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
   if (length > 0)
     (*cinfo->src->skip_input_data) (cinfo, (long) length);
 
@@ -849,15 +862,15 @@
 skip_variable (j_decompress_ptr cinfo)
 /* Skip over an unknown or uninteresting variable-length marker */
 {
-  INT32 length;
+  JLONG length;
   INPUT_VARS(cinfo);
 
   INPUT_2BYTES(cinfo, length, return FALSE);
   length -= 2;
-  
+
   TRACEMS2(cinfo, 1, JTRC_MISC_MARKER, cinfo->unread_marker, (int) length);
 
-  INPUT_SYNC(cinfo);		/* do before skip_input_data */
+  INPUT_SYNC(cinfo);            /* do before skip_input_data */
   if (length > 0)
     (*cinfo->src->skip_input_data) (cinfo, (long) length);
 
@@ -901,7 +914,7 @@
       INPUT_BYTE(cinfo, c, return FALSE);
     } while (c == 0xFF);
     if (c != 0)
-      break;			/* found a valid marker, exit loop */
+      break;                    /* found a valid marker, exit loop */
     /* Reach here if we found a stuffed-zero data sequence (FF/00).
      * Discard it and loop back to try again.
      */
@@ -961,11 +974,11 @@
     /* NB: first_marker() enforces the requirement that SOI appear first. */
     if (cinfo->unread_marker == 0) {
       if (! cinfo->marker->saw_SOI) {
-	if (! first_marker(cinfo))
-	  return JPEG_SUSPENDED;
+        if (! first_marker(cinfo))
+          return JPEG_SUSPENDED;
       } else {
-	if (! next_marker(cinfo))
-	  return JPEG_SUSPENDED;
+        if (! next_marker(cinfo))
+          return JPEG_SUSPENDED;
       }
     }
     /* At this point cinfo->unread_marker contains the marker code and the
@@ -975,74 +988,74 @@
     switch (cinfo->unread_marker) {
     case M_SOI:
       if (! get_soi(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF0:		/* Baseline */
-    case M_SOF1:		/* Extended sequential, Huffman */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
       if (! get_sof(cinfo, FALSE, FALSE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF2:		/* Progressive, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
       if (! get_sof(cinfo, TRUE, FALSE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF9:		/* Extended sequential, arithmetic */
+    case M_SOF9:                /* Extended sequential, arithmetic */
       if (! get_sof(cinfo, FALSE, TRUE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_SOF10:		/* Progressive, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
       if (! get_sof(cinfo, TRUE, TRUE))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
     /* Currently unsupported SOFn types */
-    case M_SOF3:		/* Lossless, Huffman */
-    case M_SOF5:		/* Differential sequential, Huffman */
-    case M_SOF6:		/* Differential progressive, Huffman */
-    case M_SOF7:		/* Differential lossless, Huffman */
-    case M_JPG:			/* Reserved for JPEG extensions */
-    case M_SOF11:		/* Lossless, arithmetic */
-    case M_SOF13:		/* Differential sequential, arithmetic */
-    case M_SOF14:		/* Differential progressive, arithmetic */
-    case M_SOF15:		/* Differential lossless, arithmetic */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_JPG:                 /* Reserved for JPEG extensions */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
       ERREXIT1(cinfo, JERR_SOF_UNSUPPORTED, cinfo->unread_marker);
       break;
 
     case M_SOS:
       if (! get_sos(cinfo))
-	return JPEG_SUSPENDED;
-      cinfo->unread_marker = 0;	/* processed the marker */
+        return JPEG_SUSPENDED;
+      cinfo->unread_marker = 0; /* processed the marker */
       return JPEG_REACHED_SOS;
-    
+
     case M_EOI:
       TRACEMS(cinfo, 1, JTRC_EOI);
-      cinfo->unread_marker = 0;	/* processed the marker */
+      cinfo->unread_marker = 0; /* processed the marker */
       return JPEG_REACHED_EOI;
-      
+
     case M_DAC:
       if (! get_dac(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DHT:
       if (! get_dht(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DQT:
       if (! get_dqt(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_DRI:
       if (! get_dri(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_APP0:
     case M_APP1:
     case M_APP2:
@@ -1060,16 +1073,16 @@
     case M_APP14:
     case M_APP15:
       if (! (*((my_marker_ptr) cinfo->marker)->process_APPn[
-		cinfo->unread_marker - (int) M_APP0]) (cinfo))
-	return JPEG_SUSPENDED;
+                cinfo->unread_marker - (int) M_APP0]) (cinfo))
+        return JPEG_SUSPENDED;
       break;
-      
+
     case M_COM:
       if (! (*((my_marker_ptr) cinfo->marker)->process_COM) (cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    case M_RST0:		/* these are all parameterless */
+    case M_RST0:                /* these are all parameterless */
     case M_RST1:
     case M_RST2:
     case M_RST3:
@@ -1081,12 +1094,12 @@
       TRACEMS1(cinfo, 1, JTRC_PARMLESS_MARKER, cinfo->unread_marker);
       break;
 
-    case M_DNL:			/* Ignore DNL ... perhaps the wrong thing */
+    case M_DNL:                 /* Ignore DNL ... perhaps the wrong thing */
       if (! skip_variable(cinfo))
-	return JPEG_SUSPENDED;
+        return JPEG_SUSPENDED;
       break;
 
-    default:			/* must be DHP, EXP, JPGn, or RESn */
+    default:                    /* must be DHP, EXP, JPGn, or RESn */
       /* For now, we treat the reserved markers as fatal errors since they are
        * likely to be used to signal incompatible JPEG Part 3 extensions.
        * Once the JPEG 3 version-number marker is well defined, this code
@@ -1132,7 +1145,7 @@
     /* Uh-oh, the restart markers have been messed up. */
     /* Let the data source manager determine how to resync. */
     if (! (*cinfo->src->resync_to_restart) (cinfo,
-					    cinfo->marker->next_restart_num))
+                                            cinfo->marker->next_restart_num))
       return FALSE;
   }
 
@@ -1197,25 +1210,25 @@
 {
   int marker = cinfo->unread_marker;
   int action = 1;
-  
+
   /* Always put up a warning. */
   WARNMS2(cinfo, JWRN_MUST_RESYNC, marker, desired);
-  
+
   /* Outer loop handles repeated decision after scanning forward. */
   for (;;) {
     if (marker < (int) M_SOF0)
-      action = 2;		/* invalid marker */
+      action = 2;               /* invalid marker */
     else if (marker < (int) M_RST0 || marker > (int) M_RST7)
-      action = 3;		/* valid non-restart marker */
+      action = 3;               /* valid non-restart marker */
     else {
       if (marker == ((int) M_RST0 + ((desired+1) & 7)) ||
-	  marker == ((int) M_RST0 + ((desired+2) & 7)))
-	action = 3;		/* one of the next two expected restarts */
+          marker == ((int) M_RST0 + ((desired+2) & 7)))
+        action = 3;             /* one of the next two expected restarts */
       else if (marker == ((int) M_RST0 + ((desired-1) & 7)) ||
-	       marker == ((int) M_RST0 + ((desired-2) & 7)))
-	action = 2;		/* a prior restart, so advance */
+               marker == ((int) M_RST0 + ((desired-2) & 7)))
+        action = 2;             /* a prior restart, so advance */
       else
-	action = 1;		/* desired restart or too far away */
+        action = 1;             /* desired restart or too far away */
     }
     TRACEMS2(cinfo, 4, JTRC_RECOVERY_ACTION, marker, action);
     switch (action) {
@@ -1226,7 +1239,7 @@
     case 2:
       /* Scan to the next marker, and repeat the decision loop. */
       if (! next_marker(cinfo))
-	return FALSE;
+        return FALSE;
       marker = cinfo->unread_marker;
       break;
     case 3:
@@ -1247,10 +1260,10 @@
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
-  cinfo->comp_info = NULL;		/* until allocated by get_sof */
-  cinfo->input_scan_number = 0;		/* no SOS seen yet */
-  cinfo->unread_marker = 0;		/* no pending marker */
-  marker->pub.saw_SOI = FALSE;		/* set internal state too */
+  cinfo->comp_info = NULL;              /* until allocated by get_sof */
+  cinfo->input_scan_number = 0;         /* no SOS seen yet */
+  cinfo->unread_marker = 0;             /* no pending marker */
+  marker->pub.saw_SOI = FALSE;          /* set internal state too */
   marker->pub.saw_SOF = FALSE;
   marker->pub.discarded_bytes = 0;
   marker->cur_marker = NULL;
@@ -1271,7 +1284,7 @@
   /* Create subobject in permanent pool */
   marker = (my_marker_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_PERMANENT,
-				SIZEOF(my_marker_reader));
+                                sizeof(my_marker_reader));
   cinfo->marker = (struct jpeg_marker_reader *) marker;
   /* Initialize public method pointers */
   marker->pub.reset_marker_reader = reset_marker_reader;
@@ -1302,7 +1315,7 @@
 
 GLOBAL(void)
 jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
-		   unsigned int length_limit)
+                   unsigned int length_limit)
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
   long maxlength;
@@ -1311,7 +1324,7 @@
   /* Length limit mustn't be larger than what we can allocate
    * (should only be a concern in a 16-bit environment).
    */
-  maxlength = cinfo->mem->max_alloc_chunk - SIZEOF(struct jpeg_marker_struct);
+  maxlength = cinfo->mem->max_alloc_chunk - sizeof(struct jpeg_marker_struct);
   if (((long) length_limit) > maxlength)
     length_limit = (unsigned int) maxlength;
 
@@ -1351,7 +1364,7 @@
 
 GLOBAL(void)
 jpeg_set_marker_processor (j_decompress_ptr cinfo, int marker_code,
-			   jpeg_marker_parser_method routine)
+                           jpeg_marker_parser_method routine)
 {
   my_marker_ptr marker = (my_marker_ptr) cinfo->marker;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmaster.c glmark2-2021.02/src/libjpeg-turbo/jdmaster.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmaster.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmaster.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,15 @@
 /*
  * jdmaster.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2016, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains master control logic for the JPEG decompressor.
  * These routines are concerned with selecting the modules to be executed
@@ -16,25 +21,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jpegcomp.h"
-
-
-/* Private state */
-
-typedef struct {
-  struct jpeg_decomp_master pub; /* public fields */
-
-  int pass_number;		/* # of passes completed */
-
-  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
-
-  /* Saved references to initialized quantizer modules,
-   * in case we need to switch modes.
-   */
-  struct jpeg_color_quantizer * quantizer_1pass;
-  struct jpeg_color_quantizer * quantizer_2pass;
-} my_decomp_master;
-
-typedef my_decomp_master * my_master_ptr;
+#include "jdmaster.h"
 
 
 /*
@@ -49,9 +36,10 @@
   /* Merging is the equivalent of plain box-filter upsampling */
   if (cinfo->do_fancy_upsampling || cinfo->CCIR601_sampling)
     return FALSE;
-  /* jdmerge.c only supports YCC=>RGB color conversion */
+  /* jdmerge.c only supports YCC=>RGB and YCC=>RGB565 color conversion */
   if (cinfo->jpeg_color_space != JCS_YCbCr || cinfo->num_components != 3 ||
       (cinfo->out_color_space != JCS_RGB &&
+      cinfo->out_color_space != JCS_RGB565 &&
       cinfo->out_color_space != JCS_EXT_RGB &&
       cinfo->out_color_space != JCS_EXT_RGBX &&
       cinfo->out_color_space != JCS_EXT_BGR &&
@@ -61,8 +49,12 @@
       cinfo->out_color_space != JCS_EXT_RGBA &&
       cinfo->out_color_space != JCS_EXT_BGRA &&
       cinfo->out_color_space != JCS_EXT_ABGR &&
-      cinfo->out_color_space != JCS_EXT_ARGB) ||
-      cinfo->out_color_components != rgb_pixelsize[cinfo->out_color_space])
+      cinfo->out_color_space != JCS_EXT_ARGB))
+    return FALSE;
+  if ((cinfo->out_color_space == JCS_RGB565 &&
+      cinfo->out_color_components != 3) ||
+      (cinfo->out_color_space != JCS_RGB565 &&
+      cinfo->out_color_components != rgb_pixelsize[cinfo->out_color_space]))
     return FALSE;
   /* and it only handles 2h1v or 2h2v sampling ratios */
   if (cinfo->comp_info[0].h_samp_factor != 2 ||
@@ -78,7 +70,7 @@
       cinfo->comp_info[2]._DCT_scaled_size != cinfo->_min_DCT_scaled_size)
     return FALSE;
   /* ??? also need to test for upsample-time rescaling, when & if supported */
-  return TRUE;			/* by golly, it'll work... */
+  return TRUE;                  /* by golly, it'll work... */
 #else
   return FALSE;
 #endif
@@ -89,6 +81,177 @@
  * Compute output image dimensions and related values.
  * NOTE: this is exported for possible use by application.
  * Hence it mustn't do anything that can't be done twice.
+ */
+
+#if JPEG_LIB_VERSION >= 80
+GLOBAL(void)
+#else
+LOCAL(void)
+#endif
+jpeg_core_output_dimensions (j_decompress_ptr cinfo)
+/* Do computations that are needed before master selection phase.
+ * This function is used for transcoding and full decompression.
+ */
+{
+#ifdef IDCT_SCALING_SUPPORTED
+  int ci;
+  jpeg_component_info *compptr;
+
+  /* Compute actual output image dimensions and DCT scaling choices. */
+  if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom) {
+    /* Provide 1/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 1;
+    cinfo->_min_DCT_v_scaled_size = 1;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 2) {
+    /* Provide 2/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 2L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 2L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 2;
+    cinfo->_min_DCT_v_scaled_size = 2;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 3) {
+    /* Provide 3/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 3L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 3L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 3;
+    cinfo->_min_DCT_v_scaled_size = 3;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 4) {
+    /* Provide 4/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 4L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 4L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 4;
+    cinfo->_min_DCT_v_scaled_size = 4;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 5) {
+    /* Provide 5/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 5L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 5L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 5;
+    cinfo->_min_DCT_v_scaled_size = 5;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 6) {
+    /* Provide 6/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 6L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 6L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 6;
+    cinfo->_min_DCT_v_scaled_size = 6;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 7) {
+    /* Provide 7/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 7L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 7L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 7;
+    cinfo->_min_DCT_v_scaled_size = 7;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 8) {
+    /* Provide 8/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 8L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 8L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 8;
+    cinfo->_min_DCT_v_scaled_size = 8;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 9) {
+    /* Provide 9/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 9L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 9L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 9;
+    cinfo->_min_DCT_v_scaled_size = 9;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 10) {
+    /* Provide 10/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 10L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 10L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 10;
+    cinfo->_min_DCT_v_scaled_size = 10;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 11) {
+    /* Provide 11/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 11L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 11L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 11;
+    cinfo->_min_DCT_v_scaled_size = 11;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 12) {
+    /* Provide 12/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 12L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 12L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 12;
+    cinfo->_min_DCT_v_scaled_size = 12;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 13) {
+    /* Provide 13/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 13L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 13L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 13;
+    cinfo->_min_DCT_v_scaled_size = 13;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 14) {
+    /* Provide 14/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 14L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 14L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 14;
+    cinfo->_min_DCT_v_scaled_size = 14;
+  } else if (cinfo->scale_num * DCTSIZE <= cinfo->scale_denom * 15) {
+    /* Provide 15/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 15L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 15L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 15;
+    cinfo->_min_DCT_v_scaled_size = 15;
+  } else {
+    /* Provide 16/block_size scaling */
+    cinfo->output_width = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_width * 16L, (long) DCTSIZE);
+    cinfo->output_height = (JDIMENSION)
+      jdiv_round_up((long) cinfo->image_height * 16L, (long) DCTSIZE);
+    cinfo->_min_DCT_h_scaled_size = 16;
+    cinfo->_min_DCT_v_scaled_size = 16;
+  }
+
+  /* Recompute dimensions of components */
+  for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
+       ci++, compptr++) {
+    compptr->_DCT_h_scaled_size = cinfo->_min_DCT_h_scaled_size;
+    compptr->_DCT_v_scaled_size = cinfo->_min_DCT_v_scaled_size;
+  }
+
+#else /* !IDCT_SCALING_SUPPORTED */
+
+  /* Hardwire it to "no scaling" */
+  cinfo->output_width = cinfo->image_width;
+  cinfo->output_height = cinfo->image_height;
+  /* jdinput.c has already initialized DCT_scaled_size,
+   * and has computed unscaled downsampled_width and downsampled_height.
+   */
+
+#endif /* IDCT_SCALING_SUPPORTED */
+}
+
+
+/*
+ * Compute output image dimensions and related values.
+ * NOTE: this is exported for possible use by application.
+ * Hence it mustn't do anything that can't be done twice.
  * Also note that it may be called before the master module is initialized!
  */
 
@@ -105,65 +268,24 @@
   if (cinfo->global_state != DSTATE_READY)
     ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
 
+  /* Compute core output image dimensions and DCT scaling choices. */
+  jpeg_core_output_dimensions(cinfo);
+
 #ifdef IDCT_SCALING_SUPPORTED
 
-  /* Compute actual output image dimensions and DCT scaling choices. */
-  if (cinfo->scale_num * 8 <= cinfo->scale_denom) {
-    /* Provide 1/8 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 8L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 8L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 1;
-#else
-    cinfo->min_DCT_scaled_size = 1;
-#endif
-  } else if (cinfo->scale_num * 4 <= cinfo->scale_denom) {
-    /* Provide 1/4 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 4L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 4L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 2;
-#else
-    cinfo->min_DCT_scaled_size = 2;
-#endif
-  } else if (cinfo->scale_num * 2 <= cinfo->scale_denom) {
-    /* Provide 1/2 scaling */
-    cinfo->output_width = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_width, 2L);
-    cinfo->output_height = (JDIMENSION)
-      jdiv_round_up((long) cinfo->image_height, 2L);
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = 4;
-#else
-    cinfo->min_DCT_scaled_size = 4;
-#endif
-  } else {
-    /* Provide 1/1 scaling */
-    cinfo->output_width = cinfo->image_width;
-    cinfo->output_height = cinfo->image_height;
-#if JPEG_LIB_VERSION >= 70
-    cinfo->min_DCT_h_scaled_size = cinfo->min_DCT_v_scaled_size = DCTSIZE;
-#else
-    cinfo->min_DCT_scaled_size = DCTSIZE;
-#endif
-  }
   /* In selecting the actual DCT scaling for each component, we try to
    * scale up the chroma components via IDCT scaling rather than upsampling.
    * This saves time if the upsampler gets to use 1:1 scaling.
-   * Note this code assumes that the supported DCT scalings are powers of 2.
+   * Note this code adapts subsampling ratios which are powers of 2.
    */
   for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
        ci++, compptr++) {
     int ssize = cinfo->_min_DCT_scaled_size;
     while (ssize < DCTSIZE &&
-	   (compptr->h_samp_factor * ssize * 2 <=
-	    cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) &&
-	   (compptr->v_samp_factor * ssize * 2 <=
-	    cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size)) {
+           ((cinfo->max_h_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->h_samp_factor * ssize * 2) == 0) &&
+           ((cinfo->max_v_samp_factor * cinfo->_min_DCT_scaled_size) %
+            (compptr->v_samp_factor * ssize * 2) == 0)) {
       ssize = ssize * 2;
     }
 #if JPEG_LIB_VERSION >= 70
@@ -181,12 +303,12 @@
     /* Size in samples, after IDCT scaling */
     compptr->downsampled_width = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_width *
-		    (long) (compptr->h_samp_factor * compptr->_DCT_scaled_size),
-		    (long) (cinfo->max_h_samp_factor * DCTSIZE));
+                    (long) (compptr->h_samp_factor * compptr->_DCT_scaled_size),
+                    (long) (cinfo->max_h_samp_factor * DCTSIZE));
     compptr->downsampled_height = (JDIMENSION)
       jdiv_round_up((long) cinfo->image_height *
-		    (long) (compptr->v_samp_factor * compptr->_DCT_scaled_size),
-		    (long) (cinfo->max_v_samp_factor * DCTSIZE));
+                    (long) (compptr->v_samp_factor * compptr->_DCT_scaled_size),
+                    (long) (cinfo->max_v_samp_factor * DCTSIZE));
   }
 
 #else /* !IDCT_SCALING_SUPPORTED */
@@ -220,18 +342,19 @@
     cinfo->out_color_components = rgb_pixelsize[cinfo->out_color_space];
     break;
   case JCS_YCbCr:
+  case JCS_RGB565:
     cinfo->out_color_components = 3;
     break;
   case JCS_CMYK:
   case JCS_YCCK:
     cinfo->out_color_components = 4;
     break;
-  default:			/* else must be same colorspace as in file */
+  default:                      /* else must be same colorspace as in file */
     cinfo->out_color_components = cinfo->num_components;
     break;
   }
   cinfo->output_components = (cinfo->quantize_colors ? 1 :
-			      cinfo->out_color_components);
+                              cinfo->out_color_components);
 
   /* See if upsampler will want to emit more than one row at a time */
   if (use_merged_upsample(cinfo))
@@ -248,20 +371,20 @@
  * processes are inner loops and need to be as fast as possible.  On most
  * machines, particularly CPUs with pipelines or instruction prefetch,
  * a (subscript-check-less) C table lookup
- *		x = sample_range_limit[x];
+ *              x = sample_range_limit[x];
  * is faster than explicit tests
- *		if (x < 0)  x = 0;
- *		else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
+ *              if (x < 0)  x = 0;
+ *              else if (x > MAXJSAMPLE)  x = MAXJSAMPLE;
  * These processes all use a common table prepared by the routine below.
  *
  * For most steps we can mathematically guarantee that the initial value
  * of x is within MAXJSAMPLE+1 of the legal range, so a table running from
  * -(MAXJSAMPLE+1) to 2*MAXJSAMPLE+1 is sufficient.  But for the initial
- * limiting step (just after the IDCT), a wildly out-of-range value is 
+ * limiting step (just after the IDCT), a wildly out-of-range value is
  * possible if the input data is corrupt.  To avoid any chance of indexing
  * off the end of memory and getting a bad-pointer trap, we perform the
  * post-IDCT limiting thus:
- *		x = range_limit[x & MASK];
+ *              x = range_limit[x & MASK];
  * where MASK is 2 bits wider than legal sample data, ie 10 bits for 8-bit
  * samples.  Under normal circumstances this is more than enough range and
  * a correct output will be generated; with bogus input data the mask will
@@ -279,37 +402,34 @@
  * We can save some space by overlapping the start of the post-IDCT table
  * with the simpler range limiting table.  The post-IDCT table begins at
  * sample_range_limit + CENTERJSAMPLE.
- *
- * Note that the table is allocated in near data space on PCs; it's small
- * enough and used often enough to justify this.
  */
 
 LOCAL(void)
 prepare_range_limit_table (j_decompress_ptr cinfo)
 /* Allocate and fill in the sample_range_limit table */
 {
-  JSAMPLE * table;
+  JSAMPLE *table;
   int i;
 
   table = (JSAMPLE *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-		(5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * SIZEOF(JSAMPLE));
-  table += (MAXJSAMPLE+1);	/* allow negative subscripts of simple table */
+                (5 * (MAXJSAMPLE+1) + CENTERJSAMPLE) * sizeof(JSAMPLE));
+  table += (MAXJSAMPLE+1);      /* allow negative subscripts of simple table */
   cinfo->sample_range_limit = table;
   /* First segment of "simple" table: limit[x] = 0 for x < 0 */
-  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * SIZEOF(JSAMPLE));
+  MEMZERO(table - (MAXJSAMPLE+1), (MAXJSAMPLE+1) * sizeof(JSAMPLE));
   /* Main part of "simple" table: limit[x] = x */
   for (i = 0; i <= MAXJSAMPLE; i++)
     table[i] = (JSAMPLE) i;
-  table += CENTERJSAMPLE;	/* Point to where post-IDCT table starts */
+  table += CENTERJSAMPLE;       /* Point to where post-IDCT table starts */
   /* End of simple table, rest of first half of post-IDCT table */
   for (i = CENTERJSAMPLE; i < 2*(MAXJSAMPLE+1); i++)
     table[i] = MAXJSAMPLE;
   /* Second half of post-IDCT table */
   MEMZERO(table + (2 * (MAXJSAMPLE+1)),
-	  (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * SIZEOF(JSAMPLE));
+          (2 * (MAXJSAMPLE+1) - CENTERJSAMPLE) * sizeof(JSAMPLE));
   MEMCOPY(table + (4 * (MAXJSAMPLE+1) - CENTERJSAMPLE),
-	  cinfo->sample_range_limit, CENTERJSAMPLE * SIZEOF(JSAMPLE));
+          cinfo->sample_range_limit, CENTERJSAMPLE * sizeof(JSAMPLE));
 }
 
 
@@ -442,6 +562,12 @@
   /* Initialize input side of decompressor to consume first scan. */
   (*cinfo->inputctl->start_input_pass) (cinfo);
 
+  /* Set the first and last iMCU columns to decompress from single-scan images.
+   * By default, decompress all of the iMCU columns.
+   */
+  cinfo->master->first_iMCU_col = 0;
+  cinfo->master->last_iMCU_col = cinfo->MCUs_per_row - 1;
+
 #ifdef D_MULTISCAN_FILES_SUPPORTED
   /* If jpeg_start_decompress will read the whole file, initialize
    * progress monitoring appropriately.  The input step is counted
@@ -497,24 +623,24 @@
     if (cinfo->quantize_colors && cinfo->colormap == NULL) {
       /* Select new quantization method */
       if (cinfo->two_pass_quantize && cinfo->enable_2pass_quant) {
-	cinfo->cquantize = master->quantizer_2pass;
-	master->pub.is_dummy_pass = TRUE;
+        cinfo->cquantize = master->quantizer_2pass;
+        master->pub.is_dummy_pass = TRUE;
       } else if (cinfo->enable_1pass_quant) {
-	cinfo->cquantize = master->quantizer_1pass;
+        cinfo->cquantize = master->quantizer_1pass;
       } else {
-	ERREXIT(cinfo, JERR_MODE_CHANGE);
+        ERREXIT(cinfo, JERR_MODE_CHANGE);
       }
     }
     (*cinfo->idct->start_pass) (cinfo);
     (*cinfo->coef->start_output_pass) (cinfo);
     if (! cinfo->raw_data_out) {
       if (! master->using_merged_upsample)
-	(*cinfo->cconvert->start_pass) (cinfo);
+        (*cinfo->cconvert->start_pass) (cinfo);
       (*cinfo->upsample->start_pass) (cinfo);
       if (cinfo->quantize_colors)
-	(*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
+        (*cinfo->cquantize->start_pass) (cinfo, master->pub.is_dummy_pass);
       (*cinfo->post->start_pass) (cinfo,
-	    (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
+            (master->pub.is_dummy_pass ? JBUF_SAVE_AND_PASS : JBUF_PASS_THRU));
       (*cinfo->main->start_pass) (cinfo, JBUF_PASS_THRU);
     }
   }
@@ -523,7 +649,7 @@
   if (cinfo->progress != NULL) {
     cinfo->progress->completed_passes = master->pass_number;
     cinfo->progress->total_passes = master->pass_number +
-				    (master->pub.is_dummy_pass ? 2 : 1);
+                                    (master->pub.is_dummy_pass ? 2 : 1);
     /* In buffered-image mode, we assume one more output pass if EOI not
      * yet reached, but no more passes if EOI has been reached.
      */
@@ -586,16 +712,13 @@
 GLOBAL(void)
 jinit_master_decompress (j_decompress_ptr cinfo)
 {
-  my_master_ptr master;
+  my_master_ptr master = (my_master_ptr) cinfo->master;
 
-  master = (my_master_ptr)
-      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				  SIZEOF(my_decomp_master));
-  cinfo->master = (struct jpeg_decomp_master *) master;
   master->pub.prepare_for_output_pass = prepare_for_output_pass;
   master->pub.finish_output_pass = finish_output_pass;
 
   master->pub.is_dummy_pass = FALSE;
+  master->pub.jinit_upsampler_no_alloc = FALSE;
 
   master_selection(cinfo);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmaster.h glmark2-2021.02/src/libjpeg-turbo/jdmaster.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmaster.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmaster.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+/*
+ * jdmaster.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1995, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains the master control structure for the JPEG decompressor.
+ */
+
+/* Private state */
+
+typedef struct {
+  struct jpeg_decomp_master pub; /* public fields */
+
+  int pass_number;              /* # of passes completed */
+
+  boolean using_merged_upsample; /* TRUE if using merged upsample/cconvert */
+
+  /* Saved references to initialized quantizer modules,
+   * in case we need to switch modes.
+   */
+  struct jpeg_color_quantizer *quantizer_1pass;
+  struct jpeg_color_quantizer *quantizer_2pass;
+} my_decomp_master;
+
+typedef my_decomp_master *my_master_ptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmerge.c glmark2-2021.02/src/libjpeg-turbo/jdmerge.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmerge.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmerge.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jdmerge.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2009, 2011, 2014-2015, D. R. Commander.
+ * Copyright (C) 2013, Linaro Limited.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains code for merged upsampling/color conversion.
  *
@@ -16,19 +19,19 @@
  * (ie, box filtering), we can save some work in color conversion by
  * calculating all the output pixels corresponding to a pair of chroma
  * samples at one time.  In the conversion equations
- *	R = Y           + K1 * Cr
- *	G = Y + K2 * Cb + K3 * Cr
- *	B = Y + K4 * Cb
+ *      R = Y           + K1 * Cr
+ *      G = Y + K2 * Cb + K3 * Cr
+ *      B = Y + K4 * Cb
  * only the Y term varies among the group of pixels corresponding to a pair
  * of chroma samples, so the rest of the terms can be calculated just once.
  * At typical sampling ratios, this eliminates half or three-quarters of the
  * multiplications needed for color conversion.
  *
  * This file currently provides implementations for the following cases:
- *	YCbCr => RGB color conversion only.
- *	Sampling ratios of 2h1v or 2h2v.
- *	No scaling needed at upsample time.
- *	Corner-aligned (non-CCIR601) sampling alignment.
+ *      YCbCr => RGB color conversion only.
+ *      Sampling ratios of 2h1v or 2h2v.
+ *      No scaling needed at upsample time.
+ *      Corner-aligned (non-CCIR601) sampling alignment.
  * Other special cases could be added, but in most applications these are
  * the only common cases.  (For uncommon cases we fall back on the more
  * general code in jdsample.c and jdcolor.c.)
@@ -38,7 +41,7 @@
 #include "jinclude.h"
 #include "jpeglib.h"
 #include "jsimd.h"
-#include "config.h"
+#include "jconfigint.h"
 
 #ifdef UPSAMPLE_MERGING_SUPPORTED
 
@@ -46,18 +49,17 @@
 /* Private subobject */
 
 typedef struct {
-  struct jpeg_upsampler pub;	/* public fields */
+  struct jpeg_upsampler pub;    /* public fields */
 
   /* Pointer to routine to do actual upsampling/conversion of one row group */
-  JMETHOD(void, upmethod, (j_decompress_ptr cinfo,
-			   JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-			   JSAMPARRAY output_buf));
+  void (*upmethod) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
   /* Private state for YCC->RGB conversion */
-  int * Cr_r_tab;		/* => table for Cr to R conversion */
-  int * Cb_b_tab;		/* => table for Cb to B conversion */
-  INT32 * Cr_g_tab;		/* => table for Cr to G conversion */
-  INT32 * Cb_g_tab;		/* => table for Cb to G conversion */
+  int *Cr_r_tab;                /* => table for Cr to R conversion */
+  int *Cb_b_tab;                /* => table for Cb to B conversion */
+  JLONG *Cr_g_tab;              /* => table for Cr to G conversion */
+  JLONG *Cb_g_tab;              /* => table for Cb to G conversion */
 
   /* For 2:1 vertical sampling, we produce two output rows at a time.
    * We need a "spare" row buffer to hold the second output row if the
@@ -65,22 +67,22 @@
    * to discard the dummy last row if the image height is odd.
    */
   JSAMPROW spare_row;
-  boolean spare_full;		/* T if spare buffer is occupied */
+  boolean spare_full;           /* T if spare buffer is occupied */
 
-  JDIMENSION out_row_width;	/* samples per output row */
-  JDIMENSION rows_to_go;	/* counts rows remaining in image */
+  JDIMENSION out_row_width;     /* samples per output row */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
 } my_upsampler;
 
-typedef my_upsampler * my_upsample_ptr;
+typedef my_upsampler *my_upsample_ptr;
 
-#define SCALEBITS	16	/* speediest right-shift on some machines */
-#define ONE_HALF	((INT32) 1 << (SCALEBITS-1))
-#define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
+#define SCALEBITS       16      /* speediest right-shift on some machines */
+#define ONE_HALF        ((JLONG) 1 << (SCALEBITS-1))
+#define FIX(x)          ((JLONG) ((x) * (1L<<SCALEBITS) + 0.5))
 
 
 /* Include inline routines for colorspace extensions */
 
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -92,7 +94,7 @@
 #define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 #define h2v1_merged_upsample_internal extrgb_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extrgb_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -103,13 +105,15 @@
 #define RGB_RED EXT_RGBX_RED
 #define RGB_GREEN EXT_RGBX_GREEN
 #define RGB_BLUE EXT_RGBX_BLUE
+#define RGB_ALPHA 3
 #define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 #define h2v1_merged_upsample_internal extrgbx_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extrgbx_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -120,7 +124,7 @@
 #define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 #define h2v1_merged_upsample_internal extbgr_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extbgr_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
@@ -131,13 +135,15 @@
 #define RGB_RED EXT_BGRX_RED
 #define RGB_GREEN EXT_BGRX_GREEN
 #define RGB_BLUE EXT_BGRX_BLUE
+#define RGB_ALPHA 3
 #define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 #define h2v1_merged_upsample_internal extbgrx_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extbgrx_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -145,13 +151,15 @@
 #define RGB_RED EXT_XBGR_RED
 #define RGB_GREEN EXT_XBGR_GREEN
 #define RGB_BLUE EXT_XBGR_BLUE
+#define RGB_ALPHA 0
 #define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 #define h2v1_merged_upsample_internal extxbgr_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extxbgr_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -159,13 +167,15 @@
 #define RGB_RED EXT_XRGB_RED
 #define RGB_GREEN EXT_XRGB_GREEN
 #define RGB_BLUE EXT_XRGB_BLUE
+#define RGB_ALPHA 0
 #define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 #define h2v1_merged_upsample_internal extxrgb_h2v1_merged_upsample_internal
 #define h2v2_merged_upsample_internal extxrgb_h2v2_merged_upsample_internal
-#include "jdmrgext.c.inc"
+#include "jdmrgext.c"
 #undef RGB_RED
 #undef RGB_GREEN
 #undef RGB_BLUE
+#undef RGB_ALPHA
 #undef RGB_PIXELSIZE
 #undef h2v1_merged_upsample_internal
 #undef h2v2_merged_upsample_internal
@@ -181,31 +191,31 @@
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   int i;
-  INT32 x;
+  JLONG x;
   SHIFT_TEMPS
 
   upsample->Cr_r_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
+                                (MAXJSAMPLE+1) * sizeof(int));
   upsample->Cb_b_tab = (int *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(int));
-  upsample->Cr_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(int));
+  upsample->Cr_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
-  upsample->Cb_g_tab = (INT32 *)
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
+  upsample->Cb_g_tab = (JLONG *)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				(MAXJSAMPLE+1) * SIZEOF(INT32));
+                                (MAXJSAMPLE+1) * sizeof(JLONG));
 
   for (i = 0, x = -CENTERJSAMPLE; i <= MAXJSAMPLE; i++, x++) {
     /* i is the actual input pixel value, in the range 0..MAXJSAMPLE */
     /* The Cb or Cr value we are thinking of is x = i - CENTERJSAMPLE */
     /* Cr=>R value is nearest int to 1.40200 * x */
     upsample->Cr_r_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.40200) * x + ONE_HALF, SCALEBITS);
     /* Cb=>B value is nearest int to 1.77200 * x */
     upsample->Cb_b_tab[i] = (int)
-		    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
+                    RIGHT_SHIFT(FIX(1.77200) * x + ONE_HALF, SCALEBITS);
     /* Cr=>G value is scaled-up -0.71414 * x */
     upsample->Cr_g_tab[i] = (- FIX(0.71414)) * x;
     /* Cb=>G value is scaled-up -0.34414 * x */
@@ -239,20 +249,23 @@
 
 METHODDEF(void)
 merged_2v_upsample (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 /* 2:1 vertical sampling case: may need a spare row. */
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   JSAMPROW work_ptrs[2];
-  JDIMENSION num_rows;		/* number of rows returned to caller */
+  JDIMENSION num_rows;          /* number of rows returned to caller */
 
   if (upsample->spare_full) {
     /* If we have a spare row saved from a previous cycle, just return it. */
+    JDIMENSION size = upsample->out_row_width;
+    if (cinfo->out_color_space == JCS_RGB565)
+      size = cinfo->output_width * 2;
     jcopy_sample_rows(& upsample->spare_row, 0, output_buf + *out_row_ctr, 0,
-		      1, upsample->out_row_width);
+                      1, size);
     num_rows = 1;
     upsample->spare_full = FALSE;
   } else {
@@ -288,17 +301,17 @@
 
 METHODDEF(void)
 merged_1v_upsample (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 /* 1:1 vertical sampling case: much easier, never need a spare row. */
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
 
   /* Just do the upsampling. */
   (*upsample->upmethod) (cinfo, input_buf, *in_row_group_ctr,
-			 output_buf + *out_row_ctr);
+                         output_buf + *out_row_ctr);
   /* Adjust counts */
   (*out_row_ctr)++;
   (*in_row_group_ctr)++;
@@ -321,8 +334,8 @@
 
 METHODDEF(void)
 h2v1_merged_upsample (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-		      JSAMPARRAY output_buf)
+                      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                      JSAMPARRAY output_buf)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -367,8 +380,8 @@
 
 METHODDEF(void)
 h2v2_merged_upsample (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-		      JSAMPARRAY output_buf)
+                      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                      JSAMPARRAY output_buf)
 {
   switch (cinfo->out_color_space) {
     case JCS_EXT_RGB:
@@ -408,6 +421,151 @@
 
 
 /*
+ * RGB565 conversion
+ */
+
+#define PACK_SHORT_565_LE(r, g, b)   ((((r) << 8) & 0xF800) |  \
+                                      (((g) << 3) & 0x7E0) | ((b) >> 3))
+#define PACK_SHORT_565_BE(r, g, b)   (((r) & 0xF8) | ((g) >> 5) |  \
+                                      (((g) << 11) & 0xE000) |  \
+                                      (((b) << 5) & 0x1F00))
+
+#define PACK_TWO_PIXELS_LE(l, r)     ((r << 16) | l)
+#define PACK_TWO_PIXELS_BE(l, r)     ((l << 16) | r)
+
+#define PACK_NEED_ALIGNMENT(ptr)  (((size_t)(ptr)) & 3)
+
+#define WRITE_TWO_PIXELS_LE(addr, pixels) {  \
+  ((INT16*)(addr))[0] = (INT16)(pixels);  \
+  ((INT16*)(addr))[1] = (INT16)((pixels) >> 16);  \
+}
+#define WRITE_TWO_PIXELS_BE(addr, pixels) {  \
+  ((INT16*)(addr))[1] = (INT16)(pixels);  \
+  ((INT16*)(addr))[0] = (INT16)((pixels) >> 16);  \
+}
+
+#define DITHER_565_R(r, dither)  ((r) + ((dither) & 0xFF))
+#define DITHER_565_G(g, dither)  ((g) + (((dither) & 0xFF) >> 1))
+#define DITHER_565_B(b, dither)  ((b) + ((dither) & 0xFF))
+
+
+/* Declarations for ordered dithering
+ *
+ * We use a 4x4 ordered dither array packed into 32 bits.  This array is
+ * sufficent for dithering RGB888 to RGB565.
+ */
+
+#define DITHER_MASK       0x3
+#define DITHER_ROTATE(x)  ((((x) & 0xFF) << 24) | (((x) >> 8) & 0x00FFFFFF))
+static const JLONG dither_matrix[4] = {
+  0x0008020A,
+  0x0C040E06,
+  0x030B0109,
+  0x0F070D05
+};
+
+
+/* Include inline routines for RGB565 conversion */
+
+#define PACK_SHORT_565 PACK_SHORT_565_LE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_LE
+#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_LE
+#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_le
+#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_le
+#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_le
+#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_le
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+#define PACK_SHORT_565 PACK_SHORT_565_BE
+#define PACK_TWO_PIXELS PACK_TWO_PIXELS_BE
+#define WRITE_TWO_PIXELS WRITE_TWO_PIXELS_BE
+#define h2v1_merged_upsample_565_internal h2v1_merged_upsample_565_be
+#define h2v1_merged_upsample_565D_internal h2v1_merged_upsample_565D_be
+#define h2v2_merged_upsample_565_internal h2v2_merged_upsample_565_be
+#define h2v2_merged_upsample_565D_internal h2v2_merged_upsample_565D_be
+#include "jdmrg565.c"
+#undef PACK_SHORT_565
+#undef PACK_TWO_PIXELS
+#undef WRITE_TWO_PIXELS
+#undef h2v1_merged_upsample_565_internal
+#undef h2v1_merged_upsample_565D_internal
+#undef h2v2_merged_upsample_565_internal
+#undef h2v2_merged_upsample_565D_internal
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565 (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                          JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v1_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+ }
+
+
+METHODDEF(void)
+h2v1_merged_upsample_565D (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                           JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v1_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v1_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565 (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                          JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565_be(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+  else
+    h2v2_merged_upsample_565_le(cinfo, input_buf, in_row_group_ctr,
+                                output_buf);
+}
+
+
+METHODDEF(void)
+h2v2_merged_upsample_565D (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
+                           JSAMPARRAY output_buf)
+{
+  if (is_big_endian())
+    h2v2_merged_upsample_565D_be(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+  else
+    h2v2_merged_upsample_565D_le(cinfo, input_buf, in_row_group_ctr,
+                                 output_buf);
+}
+
+
+/*
  * Module initialization routine for merged upsampling/color conversion.
  *
  * NB: this is called under the conditions determined by use_merged_upsample()
@@ -422,7 +580,7 @@
 
   upsample = (my_upsample_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_upsampler));
+                                sizeof(my_upsampler));
   cinfo->upsample = (struct jpeg_upsampler *) upsample;
   upsample->pub.start_pass = start_pass_merged_upsample;
   upsample->pub.need_context_rows = FALSE;
@@ -435,16 +593,30 @@
       upsample->upmethod = jsimd_h2v2_merged_upsample;
     else
       upsample->upmethod = h2v2_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v2_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v2_merged_upsample_565;
+      }
+    }
     /* Allocate a spare row buffer */
     upsample->spare_row = (JSAMPROW)
       (*cinfo->mem->alloc_large) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-		(size_t) (upsample->out_row_width * SIZEOF(JSAMPLE)));
+                (size_t) (upsample->out_row_width * sizeof(JSAMPLE)));
   } else {
     upsample->pub.upsample = merged_1v_upsample;
     if (jsimd_can_h2v1_merged_upsample())
       upsample->upmethod = jsimd_h2v1_merged_upsample;
     else
       upsample->upmethod = h2v1_merged_upsample;
+    if (cinfo->out_color_space == JCS_RGB565) {
+      if (cinfo->dither_mode != JDITHER_NONE) {
+        upsample->upmethod = h2v1_merged_upsample_565D;
+      } else {
+        upsample->upmethod = h2v1_merged_upsample_565;
+      }
+    }
     /* No spare row needed */
     upsample->spare_row = NULL;
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrg565.c glmark2-2021.02/src/libjpeg-turbo/jdmrg565.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrg565.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmrg565.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,356 @@
+/*
+ * jdmrg565.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565_internal (j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr = (INT16)rgb;
+   }
+ }
+
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_565D_internal (j_decompress_ptr cinfo,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr0++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr, rgb);
+    outptr += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565_internal (j_decompress_ptr cinfo,
+                                   JSAMPIMAGE input_buf,
+                                   JDIMENSION in_row_group_ctr,
+                                   JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr * 2];
+  inptr01 = input_buf[0][in_row_group_ctr * 2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr0 = (INT16)rgb;
+
+    y  = GETJSAMPLE(*inptr01);
+    r = range_limit[y + cred];
+    g = range_limit[y + cgreen];
+    b = range_limit[y + cblue];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr1 = (INT16)rgb;
+  }
+}
+
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_565D_internal (j_decompress_ptr cinfo,
+                                    JSAMPIMAGE input_buf,
+                                    JDIMENSION in_row_group_ctr,
+                                    JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  JLONG d0 = dither_matrix[cinfo->output_scanline & DITHER_MASK];
+  JLONG d1 = dither_matrix[(cinfo->output_scanline+1) & DITHER_MASK];
+  unsigned int r, g, b;
+  JLONG rgb;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr00++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr0, rgb);
+    outptr0 += 4;
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    d0 = DITHER_ROTATE(d0);
+    rgb = PACK_SHORT_565(r, g, b);
+
+    y  = GETJSAMPLE(*inptr01++);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    d1 = DITHER_ROTATE(d1);
+    rgb = PACK_TWO_PIXELS(rgb, PACK_SHORT_565(r, g, b));
+
+    WRITE_TWO_PIXELS(outptr1, rgb);
+    outptr1 += 4;
+  }
+
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+
+    y  = GETJSAMPLE(*inptr00);
+    r = range_limit[DITHER_565_R(y + cred, d0)];
+    g = range_limit[DITHER_565_G(y + cgreen, d0)];
+    b = range_limit[DITHER_565_B(y + cblue, d0)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr0 = (INT16)rgb;
+
+    y  = GETJSAMPLE(*inptr01);
+    r = range_limit[DITHER_565_R(y + cred, d1)];
+    g = range_limit[DITHER_565_G(y + cgreen, d1)];
+    b = range_limit[DITHER_565_B(y + cblue, d1)];
+    rgb = PACK_SHORT_565(r, g, b);
+    *(INT16*)outptr1 = (INT16)rgb;
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrgext.c glmark2-2021.02/src/libjpeg-turbo/jdmrgext.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrgext.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmrgext.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,186 @@
+/*
+ * jdmrgext.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2011, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+/* This file is included by jdmerge.c */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+    outptr += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr[RGB_ALPHA] = 0xFF;
+#endif
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  JLONG * Crgtab = upsample->Cr_g_tab;
+  JLONG * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+    outptr1 += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr0[RGB_ALPHA] = 0xFF;
+#endif
+    y  = GETJSAMPLE(*inptr01);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+#ifdef RGB_ALPHA
+    outptr1[RGB_ALPHA] = 0xFF;
+#endif
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrgext.c.inc glmark2-2021.02/src/libjpeg-turbo/jdmrgext.c.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdmrgext.c.inc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdmrgext.c.inc	1970-01-01 08:00:00.000000000 +0800
@@ -1,156 +0,0 @@
-/*
- * jdmrgext.c
- *
- * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
- *
- * This file contains code for merged upsampling/color conversion.
- */
-
-
-/* This file is included by jdmerge.c */
-
-
-/*
- * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
- */
-
-INLINE
-LOCAL(void)
-h2v1_merged_upsample_internal (j_decompress_ptr cinfo,
-                               JSAMPIMAGE input_buf,
-                               JDIMENSION in_row_group_ctr,
-                               JSAMPARRAY output_buf)
-{
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr;
-  JSAMPROW inptr0, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr0 = input_buf[0][in_row_group_ctr];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr = output_buf[0];
-  /* Loop for each pair of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-    outptr += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-    outptr += RGB_PIXELSIZE;
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
-    outptr[RGB_RED] =   range_limit[y + cred];
-    outptr[RGB_GREEN] = range_limit[y + cgreen];
-    outptr[RGB_BLUE] =  range_limit[y + cblue];
-  }
-}
-
-
-/*
- * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
- */
-
-INLINE
-LOCAL(void)
-h2v2_merged_upsample_internal (j_decompress_ptr cinfo,
-                               JSAMPIMAGE input_buf,
-                               JDIMENSION in_row_group_ctr,
-                               JSAMPARRAY output_buf)
-{
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr0, outptr1;
-  JSAMPROW inptr00, inptr01, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr00 = input_buf[0][in_row_group_ctr*2];
-  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr0 = output_buf[0];
-  outptr1 = output_buf[1];
-  /* Loop for each group of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr00);
-    outptr0[RGB_RED] =   range_limit[y + cred];
-    outptr0[RGB_GREEN] = range_limit[y + cgreen];
-    outptr0[RGB_BLUE] =  range_limit[y + cblue];
-    y  = GETJSAMPLE(*inptr01);
-    outptr1[RGB_RED] =   range_limit[y + cred];
-    outptr1[RGB_GREEN] = range_limit[y + cgreen];
-    outptr1[RGB_BLUE] =  range_limit[y + cblue];
-  }
-}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdphuff.c glmark2-2021.02/src/libjpeg-turbo/jdphuff.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdphuff.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdphuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdphuff.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains Huffman entropy decoding routines for progressive JPEG.
  *
@@ -17,7 +20,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdhuff.h"		/* Declarations shared with jdhuff.c */
+#include "jdhuff.h"             /* Declarations shared with jdhuff.c */
 
 
 #ifdef D_PROGRESSIVE_SUPPORTED
@@ -30,8 +33,8 @@
  */
 
 typedef struct {
-  unsigned int EOBRUN;			/* remaining EOBs in EOBRUN */
-  int last_dc_val[MAX_COMPS_IN_SCAN];	/* last DC coef for each component */
+  unsigned int EOBRUN;                  /* remaining EOBs in EOBRUN */
+  int last_dc_val[MAX_COMPS_IN_SCAN];   /* last DC coef for each component */
 } savable_state;
 
 /* This macro is to work around compilers with missing or broken
@@ -44,11 +47,11 @@
 #else
 #if MAX_COMPS_IN_SCAN == 4
 #define ASSIGN_STATE(dest,src)  \
-	((dest).EOBRUN = (src).EOBRUN, \
-	 (dest).last_dc_val[0] = (src).last_dc_val[0], \
-	 (dest).last_dc_val[1] = (src).last_dc_val[1], \
-	 (dest).last_dc_val[2] = (src).last_dc_val[2], \
-	 (dest).last_dc_val[3] = (src).last_dc_val[3])
+        ((dest).EOBRUN = (src).EOBRUN, \
+         (dest).last_dc_val[0] = (src).last_dc_val[0], \
+         (dest).last_dc_val[1] = (src).last_dc_val[1], \
+         (dest).last_dc_val[2] = (src).last_dc_val[2], \
+         (dest).last_dc_val[3] = (src).last_dc_val[3])
 #endif
 #endif
 
@@ -59,29 +62,29 @@
   /* These fields are loaded into local variables at start of each MCU.
    * In case of suspension, we exit WITHOUT updating them.
    */
-  bitread_perm_state bitstate;	/* Bit buffer at start of MCU */
-  savable_state saved;		/* Other state at start of MCU */
+  bitread_perm_state bitstate;  /* Bit buffer at start of MCU */
+  savable_state saved;          /* Other state at start of MCU */
 
   /* These fields are NOT loaded into local working state. */
-  unsigned int restarts_to_go;	/* MCUs left in this restart interval */
+  unsigned int restarts_to_go;  /* MCUs left in this restart interval */
 
   /* Pointers to derived tables (these workspaces have image lifespan) */
-  d_derived_tbl * derived_tbls[NUM_HUFF_TBLS];
+  d_derived_tbl *derived_tbls[NUM_HUFF_TBLS];
 
-  d_derived_tbl * ac_derived_tbl; /* active table during an AC scan */
+  d_derived_tbl *ac_derived_tbl; /* active table during an AC scan */
 } phuff_entropy_decoder;
 
-typedef phuff_entropy_decoder * phuff_entropy_ptr;
+typedef phuff_entropy_decoder *phuff_entropy_ptr;
 
 /* Forward declarations */
-METHODDEF(boolean) decode_mcu_DC_first JPP((j_decompress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_AC_first JPP((j_decompress_ptr cinfo,
-					    JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_DC_refine JPP((j_decompress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
-METHODDEF(boolean) decode_mcu_AC_refine JPP((j_decompress_ptr cinfo,
-					     JBLOCKROW *MCU_data));
+METHODDEF(boolean) decode_mcu_DC_first (j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_first (j_decompress_ptr cinfo,
+                                        JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_DC_refine (j_decompress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
+METHODDEF(boolean) decode_mcu_AC_refine (j_decompress_ptr cinfo,
+                                         JBLOCKROW *MCU_data);
 
 
 /*
@@ -94,8 +97,9 @@
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   boolean is_DC_band, bad;
   int ci, coefi, tbl;
+  d_derived_tbl **pdtbl;
   int *coef_bit_ptr;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
 
   is_DC_band = (cinfo->Ss == 0);
 
@@ -117,7 +121,7 @@
     if (cinfo->Al != cinfo->Ah-1)
       bad = TRUE;
   }
-  if (cinfo->Al > 13)		/* need not check for < 0 */
+  if (cinfo->Al > 13)           /* need not check for < 0 */
     bad = TRUE;
   /* Arguably the maximum Al value should be less than 13 for 8-bit precision,
    * but the spec doesn't say so, and we try to be liberal about what we
@@ -127,7 +131,7 @@
    */
   if (bad)
     ERREXIT4(cinfo, JERR_BAD_PROGRESSION,
-	     cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
+             cinfo->Ss, cinfo->Se, cinfo->Ah, cinfo->Al);
   /* Update progression status, and verify that scan order is legal.
    * Note that inter-scan inconsistencies are treated as warnings
    * not fatal errors ... not clear if this is right way to behave.
@@ -140,7 +144,7 @@
     for (coefi = cinfo->Ss; coefi <= cinfo->Se; coefi++) {
       int expected = (coef_bit_ptr[coefi] < 0) ? 0 : coef_bit_ptr[coefi];
       if (cinfo->Ah != expected)
-	WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
+        WARNMS2(cinfo, JWRN_BOGUS_PROGRESSION, cindex, coefi);
       coef_bit_ptr[coefi] = cinfo->Al;
     }
   }
@@ -164,15 +168,15 @@
      * We may build same derived table more than once, but it's not expensive.
      */
     if (is_DC_band) {
-      if (cinfo->Ah == 0) {	/* DC refinement needs no table */
-	tbl = compptr->dc_tbl_no;
-	jpeg_make_d_derived_tbl(cinfo, TRUE, tbl,
-				& entropy->derived_tbls[tbl]);
+      if (cinfo->Ah == 0) {     /* DC refinement needs no table */
+        tbl = compptr->dc_tbl_no;
+        pdtbl = entropy->derived_tbls + tbl;
+        jpeg_make_d_derived_tbl(cinfo, TRUE, tbl, pdtbl);
       }
     } else {
       tbl = compptr->ac_tbl_no;
-      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl,
-			      & entropy->derived_tbls[tbl]);
+      pdtbl = entropy->derived_tbls + tbl;
+      jpeg_make_d_derived_tbl(cinfo, FALSE, tbl, pdtbl);
       /* remember the single active table */
       entropy->ac_derived_tbl = entropy->derived_tbls[tbl];
     }
@@ -198,9 +202,11 @@
  * On some machines, a shift and add will be faster than a table lookup.
  */
 
+#define AVOID_TABLES
 #ifdef AVOID_TABLES
 
-#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))
+#define NEG_1 ((unsigned)-1)
+#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((NEG_1)<<(s)) + 1) : (x))
 
 #else
 
@@ -263,7 +269,7 @@
 /*
  * Huffman MCU decoding.
  * Each of these routines decodes and returns one MCU's worth of
- * Huffman-compressed coefficients. 
+ * Huffman-compressed coefficients.
  * The coefficients are reordered from zigzag order into natural array order,
  * but are not dequantized.
  *
@@ -284,7 +290,7 @@
 
 METHODDEF(boolean)
 decode_mcu_DC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Al = cinfo->Al;
   register int s, r;
@@ -292,14 +298,14 @@
   JBLOCKROW block;
   BITREAD_STATE_VARS;
   savable_state state;
-  d_derived_tbl * tbl;
-  jpeg_component_info * compptr;
+  d_derived_tbl *tbl;
+  jpeg_component_info *compptr;
 
   /* Process restart marker if needed; may have to suspend */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, just leave the MCU set to zeroes.
@@ -324,16 +330,16 @@
       /* Section F.2.2.1: decode the DC coefficient difference */
       HUFF_DECODE(s, br_state, tbl, return FALSE, label1);
       if (s) {
-	CHECK_BIT_BUFFER(br_state, s, return FALSE);
-	r = GET_BITS(s);
-	s = HUFF_EXTEND(r, s);
+        CHECK_BIT_BUFFER(br_state, s, return FALSE);
+        r = GET_BITS(s);
+        s = HUFF_EXTEND(r, s);
       }
 
       /* Convert DC difference to actual value, update last_dc_val */
       s += state.last_dc_val[ci];
       state.last_dc_val[ci] = s;
       /* Scale and output the coefficient (assumes jpeg_natural_order[0]=0) */
-      (*block)[0] = (JCOEF) (s << Al);
+      (*block)[0] = (JCOEF) LEFT_SHIFT(s, Al);
     }
 
     /* Completed MCU, so update state */
@@ -355,7 +361,7 @@
 
 METHODDEF(boolean)
 decode_mcu_AC_first (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Se = cinfo->Se;
   int Al = cinfo->Al;
@@ -363,13 +369,13 @@
   unsigned int EOBRUN;
   JBLOCKROW block;
   BITREAD_STATE_VARS;
-  d_derived_tbl * tbl;
+  d_derived_tbl *tbl;
 
   /* Process restart marker if needed; may have to suspend */
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, just leave the MCU set to zeroes.
@@ -380,49 +386,49 @@
     /* Load up working state.
      * We can avoid loading/saving bitread state if in an EOB run.
      */
-    EOBRUN = entropy->saved.EOBRUN;	/* only part of saved state we need */
+    EOBRUN = entropy->saved.EOBRUN;     /* only part of saved state we need */
 
     /* There is always only one block per MCU */
 
-    if (EOBRUN > 0)		/* if it's a band of zeroes... */
-      EOBRUN--;			/* ...process it now (we do nothing) */
+    if (EOBRUN > 0)             /* if it's a band of zeroes... */
+      EOBRUN--;                 /* ...process it now (we do nothing) */
     else {
       BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
       block = MCU_data[0];
       tbl = entropy->ac_derived_tbl;
 
       for (k = cinfo->Ss; k <= Se; k++) {
-	HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
-	r = s >> 4;
-	s &= 15;
-	if (s) {
-	  k += r;
-	  CHECK_BIT_BUFFER(br_state, s, return FALSE);
-	  r = GET_BITS(s);
-	  s = HUFF_EXTEND(r, s);
-	  /* Scale and output coefficient in natural (dezigzagged) order */
-	  (*block)[jpeg_natural_order[k]] = (JCOEF) (s << Al);
-	} else {
-	  if (r == 15) {	/* ZRL */
-	    k += 15;		/* skip 15 zeroes in band */
-	  } else {		/* EOBr, run length is 2^r + appended bits */
-	    EOBRUN = 1 << r;
-	    if (r) {		/* EOBr, r > 0 */
-	      CHECK_BIT_BUFFER(br_state, r, return FALSE);
-	      r = GET_BITS(r);
-	      EOBRUN += r;
-	    }
-	    EOBRUN--;		/* this band is processed at this moment */
-	    break;		/* force end-of-band */
-	  }
-	}
+        HUFF_DECODE(s, br_state, tbl, return FALSE, label2);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          k += r;
+          CHECK_BIT_BUFFER(br_state, s, return FALSE);
+          r = GET_BITS(s);
+          s = HUFF_EXTEND(r, s);
+          /* Scale and output coefficient in natural (dezigzagged) order */
+          (*block)[jpeg_natural_order[k]] = (JCOEF) LEFT_SHIFT(s, Al);
+        } else {
+          if (r == 15) {        /* ZRL */
+            k += 15;            /* skip 15 zeroes in band */
+          } else {              /* EOBr, run length is 2^r + appended bits */
+            EOBRUN = 1 << r;
+            if (r) {            /* EOBr, r > 0 */
+              CHECK_BIT_BUFFER(br_state, r, return FALSE);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            EOBRUN--;           /* this band is processed at this moment */
+            break;              /* force end-of-band */
+          }
+        }
       }
 
       BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
     }
 
     /* Completed MCU, so update state */
-    entropy->saved.EOBRUN = EOBRUN;	/* only part of saved state we need */
+    entropy->saved.EOBRUN = EOBRUN;     /* only part of saved state we need */
   }
 
   /* Account for restart interval (no-op if not using restarts) */
@@ -440,9 +446,9 @@
 
 METHODDEF(boolean)
 decode_mcu_DC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
-  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
+  int p1 = 1 << cinfo->Al;      /* 1 in the bit position being coded */
   int blkn;
   JBLOCKROW block;
   BITREAD_STATE_VARS;
@@ -451,7 +457,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* Not worth the cycles to check insufficient_data here,
@@ -489,17 +495,17 @@
 
 METHODDEF(boolean)
 decode_mcu_AC_refine (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
-{   
+{
   phuff_entropy_ptr entropy = (phuff_entropy_ptr) cinfo->entropy;
   int Se = cinfo->Se;
-  int p1 = 1 << cinfo->Al;	/* 1 in the bit position being coded */
-  int m1 = (-1) << cinfo->Al;	/* -1 in the bit position being coded */
+  int p1 = 1 << cinfo->Al;        /* 1 in the bit position being coded */
+  int m1 = (NEG_1) << cinfo->Al;  /* -1 in the bit position being coded */
   register int s, k, r;
   unsigned int EOBRUN;
   JBLOCKROW block;
   JCOEFPTR thiscoef;
   BITREAD_STATE_VARS;
-  d_derived_tbl * tbl;
+  d_derived_tbl *tbl;
   int num_newnz;
   int newnz_pos[DCTSIZE2];
 
@@ -507,7 +513,7 @@
   if (cinfo->restart_interval) {
     if (entropy->restarts_to_go == 0)
       if (! process_restart(cinfo))
-	return FALSE;
+        return FALSE;
   }
 
   /* If we've run out of data, don't modify the MCU.
@@ -535,58 +541,58 @@
 
     if (EOBRUN == 0) {
       for (; k <= Se; k++) {
-	HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
-	r = s >> 4;
-	s &= 15;
-	if (s) {
-	  if (s != 1)		/* size of new coef should always be 1 */
-	    WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
-	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	  if (GET_BITS(1))
-	    s = p1;		/* newly nonzero coef is positive */
-	  else
-	    s = m1;		/* newly nonzero coef is negative */
-	} else {
-	  if (r != 15) {
-	    EOBRUN = 1 << r;	/* EOBr, run length is 2^r + appended bits */
-	    if (r) {
-	      CHECK_BIT_BUFFER(br_state, r, goto undoit);
-	      r = GET_BITS(r);
-	      EOBRUN += r;
-	    }
-	    break;		/* rest of block is handled by EOB logic */
-	  }
-	  /* note s = 0 for processing ZRL */
-	}
-	/* Advance over already-nonzero coefs and r still-zero coefs,
-	 * appending correction bits to the nonzeroes.  A correction bit is 1
-	 * if the absolute value of the coefficient must be increased.
-	 */
-	do {
-	  thiscoef = *block + jpeg_natural_order[k];
-	  if (*thiscoef != 0) {
-	    CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	    if (GET_BITS(1)) {
-	      if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
-		if (*thiscoef >= 0)
-		  *thiscoef += p1;
-		else
-		  *thiscoef += m1;
-	      }
-	    }
-	  } else {
-	    if (--r < 0)
-	      break;		/* reached target zero coefficient */
-	  }
-	  k++;
-	} while (k <= Se);
-	if (s) {
-	  int pos = jpeg_natural_order[k];
-	  /* Output newly nonzero coefficient */
-	  (*block)[pos] = (JCOEF) s;
-	  /* Remember its position in case we have to suspend */
-	  newnz_pos[num_newnz++] = pos;
-	}
+        HUFF_DECODE(s, br_state, tbl, goto undoit, label3);
+        r = s >> 4;
+        s &= 15;
+        if (s) {
+          if (s != 1)           /* size of new coef should always be 1 */
+            WARNMS(cinfo, JWRN_HUFF_BAD_CODE);
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1))
+            s = p1;             /* newly nonzero coef is positive */
+          else
+            s = m1;             /* newly nonzero coef is negative */
+        } else {
+          if (r != 15) {
+            EOBRUN = 1 << r;    /* EOBr, run length is 2^r + appended bits */
+            if (r) {
+              CHECK_BIT_BUFFER(br_state, r, goto undoit);
+              r = GET_BITS(r);
+              EOBRUN += r;
+            }
+            break;              /* rest of block is handled by EOB logic */
+          }
+          /* note s = 0 for processing ZRL */
+        }
+        /* Advance over already-nonzero coefs and r still-zero coefs,
+         * appending correction bits to the nonzeroes.  A correction bit is 1
+         * if the absolute value of the coefficient must be increased.
+         */
+        do {
+          thiscoef = *block + jpeg_natural_order[k];
+          if (*thiscoef != 0) {
+            CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+            if (GET_BITS(1)) {
+              if ((*thiscoef & p1) == 0) { /* do nothing if already set it */
+                if (*thiscoef >= 0)
+                  *thiscoef += p1;
+                else
+                  *thiscoef += m1;
+              }
+            }
+          } else {
+            if (--r < 0)
+              break;            /* reached target zero coefficient */
+          }
+          k++;
+        } while (k <= Se);
+        if (s) {
+          int pos = jpeg_natural_order[k];
+          /* Output newly nonzero coefficient */
+          (*block)[pos] = (JCOEF) s;
+          /* Remember its position in case we have to suspend */
+          newnz_pos[num_newnz++] = pos;
+        }
       }
     }
 
@@ -597,18 +603,18 @@
        * if the absolute value of the coefficient must be increased.
        */
       for (; k <= Se; k++) {
-	thiscoef = *block + jpeg_natural_order[k];
-	if (*thiscoef != 0) {
-	  CHECK_BIT_BUFFER(br_state, 1, goto undoit);
-	  if (GET_BITS(1)) {
-	    if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
-	      if (*thiscoef >= 0)
-		*thiscoef += p1;
-	      else
-		*thiscoef += m1;
-	    }
-	  }
-	}
+        thiscoef = *block + jpeg_natural_order[k];
+        if (*thiscoef != 0) {
+          CHECK_BIT_BUFFER(br_state, 1, goto undoit);
+          if (GET_BITS(1)) {
+            if ((*thiscoef & p1) == 0) { /* do nothing if already changed it */
+              if (*thiscoef >= 0)
+                *thiscoef += p1;
+              else
+                *thiscoef += m1;
+            }
+          }
+        }
       }
       /* Count one block completed in EOB run */
       EOBRUN--;
@@ -646,7 +652,7 @@
 
   entropy = (phuff_entropy_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(phuff_entropy_decoder));
+                                sizeof(phuff_entropy_decoder));
   cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
   entropy->pub.start_pass = start_pass_phuff_decoder;
 
@@ -658,9 +664,9 @@
   /* Create progression status table */
   cinfo->coef_bits = (int (*)[DCTSIZE2])
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				cinfo->num_components*DCTSIZE2*SIZEOF(int));
+                                cinfo->num_components*DCTSIZE2*sizeof(int));
   coef_bit_ptr = & cinfo->coef_bits[0][0];
-  for (ci = 0; ci < cinfo->num_components; ci++) 
+  for (ci = 0; ci < cinfo->num_components; ci++)
     for (i = 0; i < DCTSIZE2; i++)
       *coef_bit_ptr++ = -1;
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdpostct.c glmark2-2021.02/src/libjpeg-turbo/jdpostct.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdpostct.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdpostct.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdpostct.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the decompression postprocessing controller.
  * This controller manages the upsampling, color conversion, and color
@@ -31,37 +34,34 @@
    * For two-pass color quantization, we need a full-image buffer;
    * for one-pass operation, a strip buffer is sufficient.
    */
-  jvirt_sarray_ptr whole_image;	/* virtual array, or NULL if one-pass */
-  JSAMPARRAY buffer;		/* strip buffer, or current strip of virtual */
-  JDIMENSION strip_height;	/* buffer size in rows */
+  jvirt_sarray_ptr whole_image; /* virtual array, or NULL if one-pass */
+  JSAMPARRAY buffer;            /* strip buffer, or current strip of virtual */
+  JDIMENSION strip_height;      /* buffer size in rows */
   /* for two-pass mode only: */
-  JDIMENSION starting_row;	/* row # of first row in current strip */
-  JDIMENSION next_row;		/* index of next row to fill/empty in strip */
+  JDIMENSION starting_row;      /* row # of first row in current strip */
+  JDIMENSION next_row;          /* index of next row to fill/empty in strip */
 } my_post_controller;
 
-typedef my_post_controller * my_post_ptr;
+typedef my_post_controller *my_post_ptr;
 
 
 /* Forward declarations */
 METHODDEF(void) post_process_1pass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 #ifdef QUANT_2PASS_SUPPORTED
 METHODDEF(void) post_process_prepass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 METHODDEF(void) post_process_2pass
-	JPP((j_decompress_ptr cinfo,
-	     JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	     JDIMENSION in_row_groups_avail,
-	     JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	     JDIMENSION out_rows_avail));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION *in_row_group_ctr, JDIMENSION in_row_groups_avail,
+         JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+         JDIMENSION out_rows_avail);
 #endif
 
 
@@ -84,9 +84,9 @@
        * allocate a strip buffer.  Use the virtual-array buffer as workspace.
        */
       if (post->buffer == NULL) {
-	post->buffer = (*cinfo->mem->access_virt_sarray)
-	  ((j_common_ptr) cinfo, post->whole_image,
-	   (JDIMENSION) 0, post->strip_height, TRUE);
+        post->buffer = (*cinfo->mem->access_virt_sarray)
+          ((j_common_ptr) cinfo, post->whole_image,
+           (JDIMENSION) 0, post->strip_height, TRUE);
       }
     } else {
       /* For single-pass processing without color quantization,
@@ -124,10 +124,10 @@
 
 METHODDEF(void)
 post_process_1pass (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION num_rows, max_rows;
@@ -139,11 +139,11 @@
     max_rows = post->strip_height;
   num_rows = 0;
   (*cinfo->upsample->upsample) (cinfo,
-		input_buf, in_row_group_ctr, in_row_groups_avail,
-		post->buffer, &num_rows, max_rows);
+                input_buf, in_row_group_ctr, in_row_groups_avail,
+                post->buffer, &num_rows, max_rows);
   /* Quantize and emit data. */
   (*cinfo->cquantize->color_quantize) (cinfo,
-		post->buffer, output_buf + *out_row_ctr, (int) num_rows);
+                post->buffer, output_buf + *out_row_ctr, (int) num_rows);
   *out_row_ctr += num_rows;
 }
 
@@ -156,10 +156,10 @@
 
 METHODDEF(void)
 post_process_prepass (j_decompress_ptr cinfo,
-		      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		      JDIMENSION in_row_groups_avail,
-		      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		      JDIMENSION out_rows_avail)
+                      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                      JDIMENSION in_row_groups_avail,
+                      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                      JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION old_next_row, num_rows;
@@ -167,22 +167,22 @@
   /* Reposition virtual buffer if at start of strip. */
   if (post->next_row == 0) {
     post->buffer = (*cinfo->mem->access_virt_sarray)
-	((j_common_ptr) cinfo, post->whole_image,
-	 post->starting_row, post->strip_height, TRUE);
+        ((j_common_ptr) cinfo, post->whole_image,
+         post->starting_row, post->strip_height, TRUE);
   }
 
   /* Upsample some data (up to a strip height's worth). */
   old_next_row = post->next_row;
   (*cinfo->upsample->upsample) (cinfo,
-		input_buf, in_row_group_ctr, in_row_groups_avail,
-		post->buffer, &post->next_row, post->strip_height);
+                input_buf, in_row_group_ctr, in_row_groups_avail,
+                post->buffer, &post->next_row, post->strip_height);
 
   /* Allow quantizer to scan new data.  No data is emitted, */
   /* but we advance out_row_ctr so outer loop can tell when we're done. */
   if (post->next_row > old_next_row) {
     num_rows = post->next_row - old_next_row;
     (*cinfo->cquantize->color_quantize) (cinfo, post->buffer + old_next_row,
-					 (JSAMPARRAY) NULL, (int) num_rows);
+                                         (JSAMPARRAY) NULL, (int) num_rows);
     *out_row_ctr += num_rows;
   }
 
@@ -200,10 +200,10 @@
 
 METHODDEF(void)
 post_process_2pass (j_decompress_ptr cinfo,
-		    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-		    JDIMENSION in_row_groups_avail,
-		    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-		    JDIMENSION out_rows_avail)
+                    JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail,
+                    JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                    JDIMENSION out_rows_avail)
 {
   my_post_ptr post = (my_post_ptr) cinfo->post;
   JDIMENSION num_rows, max_rows;
@@ -211,8 +211,8 @@
   /* Reposition virtual buffer if at start of strip. */
   if (post->next_row == 0) {
     post->buffer = (*cinfo->mem->access_virt_sarray)
-	((j_common_ptr) cinfo, post->whole_image,
-	 post->starting_row, post->strip_height, FALSE);
+        ((j_common_ptr) cinfo, post->whole_image,
+         post->starting_row, post->strip_height, FALSE);
   }
 
   /* Determine number of rows to emit. */
@@ -227,8 +227,8 @@
 
   /* Quantize and emit data. */
   (*cinfo->cquantize->color_quantize) (cinfo,
-		post->buffer + post->next_row, output_buf + *out_row_ctr,
-		(int) num_rows);
+                post->buffer + post->next_row, output_buf + *out_row_ctr,
+                (int) num_rows);
   *out_row_ctr += num_rows;
 
   /* Advance if we filled the strip. */
@@ -253,11 +253,11 @@
 
   post = (my_post_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_post_controller));
+                                sizeof(my_post_controller));
   cinfo->post = (struct jpeg_d_post_controller *) post;
   post->pub.start_pass = start_pass_dpost;
-  post->whole_image = NULL;	/* flag for no virtual arrays */
-  post->buffer = NULL;		/* flag for no strip buffer */
+  post->whole_image = NULL;     /* flag for no virtual arrays */
+  post->buffer = NULL;          /* flag for no strip buffer */
 
   /* Create the quantization buffer, if needed */
   if (cinfo->quantize_colors) {
@@ -271,20 +271,20 @@
       /* We round up the number of rows to a multiple of the strip height. */
 #ifdef QUANT_2PASS_SUPPORTED
       post->whole_image = (*cinfo->mem->request_virt_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
-	 cinfo->output_width * cinfo->out_color_components,
-	 (JDIMENSION) jround_up((long) cinfo->output_height,
-				(long) post->strip_height),
-	 post->strip_height);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+         cinfo->output_width * cinfo->out_color_components,
+         (JDIMENSION) jround_up((long) cinfo->output_height,
+                                (long) post->strip_height),
+         post->strip_height);
 #else
       ERREXIT(cinfo, JERR_BAD_BUFFER_MODE);
 #endif /* QUANT_2PASS_SUPPORTED */
     } else {
       /* One-pass color quantization: just make a strip buffer. */
       post->buffer = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 cinfo->output_width * cinfo->out_color_components,
-	 post->strip_height);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         cinfo->output_width * cinfo->out_color_components,
+         post->strip_height);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdsample.c glmark2-2021.02/src/libjpeg-turbo/jdsample.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdsample.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdsample.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,15 @@
 /*
  * jdsample.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2010, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Copyright (C) 2010, 2015-2016, D. R. Commander.
+ * Copyright (C) 2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains upsampling routines.
  *
@@ -20,50 +24,12 @@
  *   Pub. by IEEE Computer Society Press, Los Alamitos, CA. ISBN 0-8186-8944-7.
  */
 
-#define JPEG_INTERNALS
 #include "jinclude.h"
-#include "jpeglib.h"
+#include "jdsample.h"
 #include "jsimd.h"
 #include "jpegcomp.h"
 
 
-/* Pointer to routine to upsample a single component */
-typedef JMETHOD(void, upsample1_ptr,
-		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-
-/* Private subobject */
-
-typedef struct {
-  struct jpeg_upsampler pub;	/* public fields */
-
-  /* Color conversion buffer.  When using separate upsampling and color
-   * conversion steps, this buffer holds one upsampled row group until it
-   * has been color converted and output.
-   * Note: we do not allocate any storage for component(s) which are full-size,
-   * ie do not need rescaling.  The corresponding entry of color_buf[] is
-   * simply set to point to the input data array, thereby avoiding copying.
-   */
-  JSAMPARRAY color_buf[MAX_COMPONENTS];
-
-  /* Per-component upsampling method pointers */
-  upsample1_ptr methods[MAX_COMPONENTS];
-
-  int next_row_out;		/* counts rows emitted from color_buf */
-  JDIMENSION rows_to_go;	/* counts rows remaining in image */
-
-  /* Height of an input row group for each component. */
-  int rowgroup_height[MAX_COMPONENTS];
-
-  /* These arrays save pixel expansion factors so that int_expand need not
-   * recompute them each time.  They are unused for other upsampling methods.
-   */
-  UINT8 h_expand[MAX_COMPONENTS];
-  UINT8 v_expand[MAX_COMPONENTS];
-} my_upsampler;
-
-typedef my_upsampler * my_upsample_ptr;
-
 
 /*
  * Initialize for an upsampling pass.
@@ -91,26 +57,26 @@
 
 METHODDEF(void)
 sep_upsample (j_decompress_ptr cinfo,
-	      JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
-	      JDIMENSION in_row_groups_avail,
-	      JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-	      JDIMENSION out_rows_avail)
+              JSAMPIMAGE input_buf, JDIMENSION *in_row_group_ctr,
+              JDIMENSION in_row_groups_avail,
+              JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+              JDIMENSION out_rows_avail)
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   JDIMENSION num_rows;
 
   /* Fill the conversion buffer, if it's empty */
   if (upsample->next_row_out >= cinfo->max_v_samp_factor) {
     for (ci = 0, compptr = cinfo->comp_info; ci < cinfo->num_components;
-	 ci++, compptr++) {
+         ci++, compptr++) {
       /* Invoke per-component upsample method.  Notice we pass a POINTER
        * to color_buf[ci], so that fullsize_upsample can change it.
        */
       (*upsample->methods[ci]) (cinfo, compptr,
-	input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
-	upsample->color_buf + ci);
+        input_buf[ci] + (*in_row_group_ctr * upsample->rowgroup_height[ci]),
+        upsample->color_buf + ci);
     }
     upsample->next_row_out = 0;
   }
@@ -122,7 +88,7 @@
   /* Not more than the distance to the end of the image.  Need this test
    * in case the image height is not a multiple of max_v_samp_factor:
    */
-  if (num_rows > upsample->rows_to_go) 
+  if (num_rows > upsample->rows_to_go)
     num_rows = upsample->rows_to_go;
   /* And not more than what the client can accept: */
   out_rows_avail -= *out_row_ctr;
@@ -130,9 +96,9 @@
     num_rows = out_rows_avail;
 
   (*cinfo->cconvert->color_convert) (cinfo, upsample->color_buf,
-				     (JDIMENSION) upsample->next_row_out,
-				     output_buf + *out_row_ctr,
-				     (int) num_rows);
+                                     (JDIMENSION) upsample->next_row_out,
+                                     output_buf + *out_row_ctr,
+                                     (int) num_rows);
 
   /* Adjust counts */
   *out_row_ctr += num_rows;
@@ -158,8 +124,8 @@
  */
 
 METHODDEF(void)
-fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		   JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+fullsize_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                   JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   *output_data_ptr = input_data;
 }
@@ -171,10 +137,10 @@
  */
 
 METHODDEF(void)
-noop_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+noop_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
-  *output_data_ptr = NULL;	/* safety check */
+  *output_data_ptr = NULL;      /* safety check */
 }
 
 
@@ -190,8 +156,8 @@
  */
 
 METHODDEF(void)
-int_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	      JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+              JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
   JSAMPARRAY output_data = *output_data_ptr;
@@ -212,15 +178,15 @@
     outptr = output_data[outrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       for (h = h_expand; h > 0; h--) {
-	*outptr++ = invalue;
+        *outptr++ = invalue;
       }
     }
     /* Generate any additional output rows by duplicating the first one */
     if (v_expand > 1) {
       jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
-			v_expand-1, cinfo->output_width);
+                        v_expand-1, cinfo->output_width);
     }
     inrow++;
     outrow += v_expand;
@@ -234,8 +200,8 @@
  */
 
 METHODDEF(void)
-h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v1_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -248,7 +214,7 @@
     outptr = output_data[inrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       *outptr++ = invalue;
       *outptr++ = invalue;
     }
@@ -262,8 +228,8 @@
  */
 
 METHODDEF(void)
-h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v2_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -277,12 +243,12 @@
     outptr = output_data[outrow];
     outend = outptr + cinfo->output_width;
     while (outptr < outend) {
-      invalue = *inptr++;	/* don't need GETJSAMPLE() here */
+      invalue = *inptr++;       /* don't need GETJSAMPLE() here */
       *outptr++ = invalue;
       *outptr++ = invalue;
     }
     jcopy_sample_rows(output_data, outrow, output_data, outrow+1,
-		      1, cinfo->output_width);
+                      1, cinfo->output_width);
     inrow++;
     outrow += 2;
   }
@@ -305,8 +271,8 @@
  */
 
 METHODDEF(void)
-h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v1_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr, outptr;
@@ -346,15 +312,15 @@
  */
 
 METHODDEF(void)
-h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		     JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr)
+h2v2_fancy_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                     JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
 {
   JSAMPARRAY output_data = *output_data_ptr;
   register JSAMPROW inptr0, inptr1, outptr;
 #if BITS_IN_JSAMPLE == 8
   register int thiscolsum, lastcolsum, nextcolsum;
 #else
-  register INT32 thiscolsum, lastcolsum, nextcolsum;
+  register JLONG thiscolsum, lastcolsum, nextcolsum;
 #endif
   register JDIMENSION colctr;
   int inrow, outrow, v;
@@ -364,10 +330,10 @@
     for (v = 0; v < 2; v++) {
       /* inptr0 points to nearest input row, inptr1 points to next nearest */
       inptr0 = input_data[inrow];
-      if (v == 0)		/* next nearest is row above */
-	inptr1 = input_data[inrow-1];
-      else			/* next nearest is row below */
-	inptr1 = input_data[inrow+1];
+      if (v == 0)               /* next nearest is row above */
+        inptr1 = input_data[inrow-1];
+      else                      /* next nearest is row below */
+        inptr1 = input_data[inrow+1];
       outptr = output_data[outrow++];
 
       /* Special case for first column */
@@ -378,12 +344,12 @@
       lastcolsum = thiscolsum; thiscolsum = nextcolsum;
 
       for (colctr = compptr->downsampled_width - 2; colctr > 0; colctr--) {
-	/* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
-	/* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
-	nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
-	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
-	*outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
-	lastcolsum = thiscolsum; thiscolsum = nextcolsum;
+        /* General case: 3/4 * nearer pixel + 1/4 * further pixel in each */
+        /* dimension, thus 9/16, 3/16, 3/16, 1/16 overall */
+        nextcolsum = GETJSAMPLE(*inptr0++) * 3 + GETJSAMPLE(*inptr1++);
+        *outptr++ = (JSAMPLE) ((thiscolsum * 3 + lastcolsum + 8) >> 4);
+        *outptr++ = (JSAMPLE) ((thiscolsum * 3 + nextcolsum + 7) >> 4);
+        lastcolsum = thiscolsum; thiscolsum = nextcolsum;
       }
 
       /* Special case for last column */
@@ -404,19 +370,22 @@
 {
   my_upsample_ptr upsample;
   int ci;
-  jpeg_component_info * compptr;
+  jpeg_component_info *compptr;
   boolean need_buffer, do_fancy;
   int h_in_group, v_in_group, h_out_group, v_out_group;
 
-  upsample = (my_upsample_ptr)
-    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_upsampler));
-  cinfo->upsample = (struct jpeg_upsampler *) upsample;
-  upsample->pub.start_pass = start_pass_upsample;
-  upsample->pub.upsample = sep_upsample;
-  upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+  if (!cinfo->master->jinit_upsampler_no_alloc) {
+    upsample = (my_upsample_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(my_upsampler));
+    cinfo->upsample = (struct jpeg_upsampler *) upsample;
+    upsample->pub.start_pass = start_pass_upsample;
+    upsample->pub.upsample = sep_upsample;
+    upsample->pub.need_context_rows = FALSE; /* until we find out differently */
+  } else
+    upsample = (my_upsample_ptr) cinfo->upsample;
 
-  if (cinfo->CCIR601_sampling)	/* this isn't supported */
+  if (cinfo->CCIR601_sampling)  /* this isn't supported */
     ERREXIT(cinfo, JERR_CCIR601_NOTIMPL);
 
   /* jdmainct.c doesn't support context rows when min_DCT_scaled_size = 1,
@@ -433,9 +402,9 @@
      * are to be converted to max_h_samp_factor * max_v_samp_factor pixels.
      */
     h_in_group = (compptr->h_samp_factor * compptr->_DCT_scaled_size) /
-		 cinfo->_min_DCT_scaled_size;
+                 cinfo->_min_DCT_scaled_size;
     v_in_group = (compptr->v_samp_factor * compptr->_DCT_scaled_size) /
-		 cinfo->_min_DCT_scaled_size;
+                 cinfo->_min_DCT_scaled_size;
     h_out_group = cinfo->max_h_samp_factor;
     v_out_group = cinfo->max_v_samp_factor;
     upsample->rowgroup_height[ci] = v_in_group; /* save for use later */
@@ -449,48 +418,53 @@
       upsample->methods[ci] = fullsize_upsample;
       need_buffer = FALSE;
     } else if (h_in_group * 2 == h_out_group &&
-	       v_in_group == v_out_group) {
+               v_in_group == v_out_group) {
       /* Special cases for 2h1v upsampling */
       if (do_fancy && compptr->downsampled_width > 2) {
-	if (jsimd_can_h2v1_fancy_upsample())
-	  upsample->methods[ci] = jsimd_h2v1_fancy_upsample;
-	else
-	  upsample->methods[ci] = h2v1_fancy_upsample;
+        if (jsimd_can_h2v1_fancy_upsample())
+          upsample->methods[ci] = jsimd_h2v1_fancy_upsample;
+        else
+          upsample->methods[ci] = h2v1_fancy_upsample;
       } else {
-	if (jsimd_can_h2v1_upsample())
-	  upsample->methods[ci] = jsimd_h2v1_upsample;
-	else
-	  upsample->methods[ci] = h2v1_upsample;
+        if (jsimd_can_h2v1_upsample())
+          upsample->methods[ci] = jsimd_h2v1_upsample;
+        else
+          upsample->methods[ci] = h2v1_upsample;
       }
     } else if (h_in_group * 2 == h_out_group &&
-	       v_in_group * 2 == v_out_group) {
+               v_in_group * 2 == v_out_group) {
       /* Special cases for 2h2v upsampling */
       if (do_fancy && compptr->downsampled_width > 2) {
-	if (jsimd_can_h2v2_fancy_upsample())
-	  upsample->methods[ci] = jsimd_h2v2_fancy_upsample;
-	else
-	  upsample->methods[ci] = h2v2_fancy_upsample;
-	upsample->pub.need_context_rows = TRUE;
+        if (jsimd_can_h2v2_fancy_upsample())
+          upsample->methods[ci] = jsimd_h2v2_fancy_upsample;
+        else
+          upsample->methods[ci] = h2v2_fancy_upsample;
+        upsample->pub.need_context_rows = TRUE;
       } else {
-	if (jsimd_can_h2v2_upsample())
-	  upsample->methods[ci] = jsimd_h2v2_upsample;
-	else
-	  upsample->methods[ci] = h2v2_upsample;
+        if (jsimd_can_h2v2_upsample())
+          upsample->methods[ci] = jsimd_h2v2_upsample;
+        else
+          upsample->methods[ci] = h2v2_upsample;
       }
     } else if ((h_out_group % h_in_group) == 0 &&
-	       (v_out_group % v_in_group) == 0) {
+               (v_out_group % v_in_group) == 0) {
       /* Generic integral-factors upsampling method */
-      upsample->methods[ci] = int_upsample;
+#if defined(__mips__)
+      if (jsimd_can_int_upsample())
+        upsample->methods[ci] = jsimd_int_upsample;
+      else
+#endif
+        upsample->methods[ci] = int_upsample;
       upsample->h_expand[ci] = (UINT8) (h_out_group / h_in_group);
       upsample->v_expand[ci] = (UINT8) (v_out_group / v_in_group);
     } else
       ERREXIT(cinfo, JERR_FRACT_SAMPLE_NOTIMPL);
-    if (need_buffer) {
+    if (need_buffer && !cinfo->master->jinit_upsampler_no_alloc) {
       upsample->color_buf[ci] = (*cinfo->mem->alloc_sarray)
-	((j_common_ptr) cinfo, JPOOL_IMAGE,
-	 (JDIMENSION) jround_up((long) cinfo->output_width,
-				(long) cinfo->max_h_samp_factor),
-	 (JDIMENSION) cinfo->max_v_samp_factor);
+        ((j_common_ptr) cinfo, JPOOL_IMAGE,
+         (JDIMENSION) jround_up((long) cinfo->output_width,
+                                (long) cinfo->max_h_samp_factor),
+         (JDIMENSION) cinfo->max_v_samp_factor);
     }
   }
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdsample.h glmark2-2021.02/src/libjpeg-turbo/jdsample.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdsample.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdsample.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,50 @@
+/*
+ * jdsample.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#define JPEG_INTERNALS
+#include "jpeglib.h"
+
+
+/* Pointer to routine to upsample a single component */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+
+/* Private subobject */
+
+typedef struct {
+  struct jpeg_upsampler pub;    /* public fields */
+
+  /* Color conversion buffer.  When using separate upsampling and color
+   * conversion steps, this buffer holds one upsampled row group until it
+   * has been color converted and output.
+   * Note: we do not allocate any storage for component(s) which are full-size,
+   * ie do not need rescaling.  The corresponding entry of color_buf[] is
+   * simply set to point to the input data array, thereby avoiding copying.
+   */
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+
+  /* Per-component upsampling method pointers */
+  upsample1_ptr methods[MAX_COMPONENTS];
+
+  int next_row_out;             /* counts rows emitted from color_buf */
+  JDIMENSION rows_to_go;        /* counts rows remaining in image */
+
+  /* Height of an input row group for each component. */
+  int rowgroup_height[MAX_COMPONENTS];
+
+  /* These arrays save pixel expansion factors so that int_expand need not
+   * recompute them each time.  They are unused for other upsampling methods.
+   */
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdtrans.c glmark2-2021.02/src/libjpeg-turbo/jdtrans.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jdtrans.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jdtrans.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jdtrans.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1995-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains library routines for transcoding decompression,
  * that is, reading raw DCT coefficient arrays from an input JPEG file.
@@ -16,7 +19,7 @@
 
 
 /* Forward declarations */
-LOCAL(void) transdecode_master_selection JPP((j_decompress_ptr cinfo));
+LOCAL(void) transdecode_master_selection (j_decompress_ptr cinfo);
 
 
 /*
@@ -55,20 +58,20 @@
       int retcode;
       /* Call progress monitor hook if present */
       if (cinfo->progress != NULL)
-	(*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
+        (*cinfo->progress->progress_monitor) ((j_common_ptr) cinfo);
       /* Absorb some more input */
       retcode = (*cinfo->inputctl->consume_input) (cinfo);
       if (retcode == JPEG_SUSPENDED)
-	return NULL;
+        return NULL;
       if (retcode == JPEG_REACHED_EOI)
-	break;
+        break;
       /* Advance progress counter if appropriate */
       if (cinfo->progress != NULL &&
-	  (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
-	if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
-	  /* startup underestimated number of scans; ratchet up one scan */
-	  cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
-	}
+          (retcode == JPEG_ROW_COMPLETED || retcode == JPEG_REACHED_SOS)) {
+        if (++cinfo->progress->pass_counter >= cinfo->progress->pass_limit) {
+          /* startup underestimated number of scans; ratchet up one scan */
+          cinfo->progress->pass_limit += (long) cinfo->total_iMCU_rows;
+        }
       }
     }
     /* Set state so that jpeg_finish_decompress does the right thing */
@@ -84,7 +87,7 @@
   }
   /* Oops, improper usage */
   ERREXIT1(cinfo, JERR_BAD_STATE, cinfo->global_state);
-  return NULL;			/* keep compiler happy */
+  return NULL;                  /* keep compiler happy */
 }
 
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jerror.c glmark2-2021.02/src/libjpeg-turbo/jerror.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jerror.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jerror.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jerror.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains simple error-reporting and trace-message routines.
  * These are suitable for Unix-like systems and others where writing to
@@ -28,7 +31,7 @@
 #include <windows.h>
 #endif
 
-#ifndef EXIT_FAILURE		/* define exit() codes if not provided */
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
 #define EXIT_FAILURE  1
 #endif
 
@@ -41,11 +44,7 @@
  * want to refer to it directly.
  */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_std_message_table	jMsgTable
-#endif
-
-#define JMESSAGE(code,string)	string ,
+#define JMESSAGE(code,string)   string ,
 
 const char * const jpeg_std_message_table[] = {
 #include "jerror.h"
@@ -105,7 +104,7 @@
 #ifdef USE_WINDOWS_MESSAGEBOX
   /* Display it in a message dialog box */
   MessageBox(GetActiveWindow(), buffer, "JPEG Library Error",
-	     MB_OK | MB_ICONERROR);
+             MB_OK | MB_ICONERROR);
 #else
   /* Send it to stderr, adding a newline */
   fprintf(stderr, "%s\n", buffer);
@@ -127,7 +126,7 @@
 METHODDEF(void)
 emit_message (j_common_ptr cinfo, int msg_level)
 {
-  struct jpeg_error_mgr * err = cinfo->err;
+  struct jpeg_error_mgr *err = cinfo->err;
 
   if (msg_level < 0) {
     /* It's a warning message.  Since corrupt files may generate many warnings,
@@ -154,12 +153,12 @@
  */
 
 METHODDEF(void)
-format_message (j_common_ptr cinfo, char * buffer)
+format_message (j_common_ptr cinfo, char *buffer)
 {
-  struct jpeg_error_mgr * err = cinfo->err;
+  struct jpeg_error_mgr *err = cinfo->err;
   int msg_code = err->msg_code;
-  const char * msgtext = NULL;
-  const char * msgptr;
+  const char *msgtext = NULL;
+  const char *msgptr;
   char ch;
   boolean isstring;
 
@@ -167,8 +166,8 @@
   if (msg_code > 0 && msg_code <= err->last_jpeg_message) {
     msgtext = err->jpeg_message_table[msg_code];
   } else if (err->addon_message_table != NULL &&
-	     msg_code >= err->first_addon_message &&
-	     msg_code <= err->last_addon_message) {
+             msg_code >= err->first_addon_message &&
+             msg_code <= err->last_addon_message) {
     msgtext = err->addon_message_table[msg_code - err->first_addon_message];
   }
 
@@ -193,10 +192,10 @@
     sprintf(buffer, msgtext, err->msg_parm.s);
   else
     sprintf(buffer, msgtext,
-	    err->msg_parm.i[0], err->msg_parm.i[1],
-	    err->msg_parm.i[2], err->msg_parm.i[3],
-	    err->msg_parm.i[4], err->msg_parm.i[5],
-	    err->msg_parm.i[6], err->msg_parm.i[7]);
+            err->msg_parm.i[0], err->msg_parm.i[1],
+            err->msg_parm.i[2], err->msg_parm.i[3],
+            err->msg_parm.i[4], err->msg_parm.i[5],
+            err->msg_parm.i[6], err->msg_parm.i[7]);
 }
 
 
@@ -213,22 +212,22 @@
 {
   cinfo->err->num_warnings = 0;
   /* trace_level is not reset since it is an application-supplied parameter */
-  cinfo->err->msg_code = 0;	/* may be useful as a flag for "no error" */
+  cinfo->err->msg_code = 0;     /* may be useful as a flag for "no error" */
 }
 
 
 /*
  * Fill in the standard error-handling methods in a jpeg_error_mgr object.
  * Typical call is:
- *	struct jpeg_compress_struct cinfo;
- *	struct jpeg_error_mgr err;
+ *      struct jpeg_compress_struct cinfo;
+ *      struct jpeg_error_mgr err;
  *
- *	cinfo.err = jpeg_std_error(&err);
+ *      cinfo.err = jpeg_std_error(&err);
  * after which the application may override some of the methods.
  */
 
 GLOBAL(struct jpeg_error_mgr *)
-jpeg_std_error (struct jpeg_error_mgr * err)
+jpeg_std_error (struct jpeg_error_mgr *err)
 {
   err->error_exit = error_exit;
   err->emit_message = emit_message;
@@ -236,16 +235,16 @@
   err->format_message = format_message;
   err->reset_error_mgr = reset_error_mgr;
 
-  err->trace_level = 0;		/* default = no tracing */
-  err->num_warnings = 0;	/* no warnings emitted yet */
-  err->msg_code = 0;		/* may be useful as a flag for "no error" */
+  err->trace_level = 0;         /* default = no tracing */
+  err->num_warnings = 0;        /* no warnings emitted yet */
+  err->msg_code = 0;            /* may be useful as a flag for "no error" */
 
   /* Initialize message table pointers */
   err->jpeg_message_table = jpeg_std_message_table;
   err->last_jpeg_message = (int) JMSG_LASTMSGCODE - 1;
 
   err->addon_message_table = NULL;
-  err->first_addon_message = 0;	/* for safety */
+  err->first_addon_message = 0; /* for safety */
   err->last_addon_message = 0;
 
   return err;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jerror.h glmark2-2021.02/src/libjpeg-turbo/jerror.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jerror.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jerror.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jerror.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1997, Thomas G. Lane.
  * Modified 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file defines the error and message codes for the JPEG library.
  * Edit this file to add new codes, or to translate the message strings to
@@ -33,7 +36,7 @@
 
 typedef enum {
 
-#define JMESSAGE(code,string)	code ,
+#define JMESSAGE(code,string)   code ,
 
 #endif /* JMAKE_ENUM_LIST */
 
@@ -42,7 +45,7 @@
 /* For maintenance convenience, list is alphabetical by message code name */
 #if JPEG_LIB_VERSION < 70
 JMESSAGE(JERR_ARITH_NOTIMPL,
-	 "Sorry, arithmetic coding is not implemented")
+         "Sorry, arithmetic coding is not implemented")
 #endif
 JMESSAGE(JERR_BAD_ALIGN_TYPE, "ALIGN_TYPE is wrong, please fix")
 JMESSAGE(JERR_BAD_ALLOC_CHUNK, "MAX_ALLOC_CHUNK is wrong, please fix")
@@ -55,26 +58,26 @@
 JMESSAGE(JERR_BAD_DCTSIZE, "IDCT output block size %d not supported")
 #if JPEG_LIB_VERSION >= 70
 JMESSAGE(JERR_BAD_DROP_SAMPLING,
-	 "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
+         "Component index %d: mismatching sampling ratio %d:%d, %d:%d, %c")
 #endif
 JMESSAGE(JERR_BAD_HUFF_TABLE, "Bogus Huffman table definition")
 JMESSAGE(JERR_BAD_IN_COLORSPACE, "Bogus input colorspace")
 JMESSAGE(JERR_BAD_J_COLORSPACE, "Bogus JPEG colorspace")
 JMESSAGE(JERR_BAD_LENGTH, "Bogus marker length")
 JMESSAGE(JERR_BAD_LIB_VERSION,
-	 "Wrong JPEG library version: library is %d, caller expects %d")
+         "Wrong JPEG library version: library is %d, caller expects %d")
 JMESSAGE(JERR_BAD_MCU_SIZE, "Sampling factors too large for interleaved scan")
 JMESSAGE(JERR_BAD_POOL_ID, "Invalid memory pool code %d")
 JMESSAGE(JERR_BAD_PRECISION, "Unsupported JPEG data precision %d")
 JMESSAGE(JERR_BAD_PROGRESSION,
-	 "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
+         "Invalid progressive parameters Ss=%d Se=%d Ah=%d Al=%d")
 JMESSAGE(JERR_BAD_PROG_SCRIPT,
-	 "Invalid progressive parameters at scan script entry %d")
+         "Invalid progressive parameters at scan script entry %d")
 JMESSAGE(JERR_BAD_SAMPLING, "Bogus sampling factors")
 JMESSAGE(JERR_BAD_SCAN_SCRIPT, "Invalid scan script at entry %d")
 JMESSAGE(JERR_BAD_STATE, "Improper call to JPEG library in state %d")
 JMESSAGE(JERR_BAD_STRUCT_SIZE,
-	 "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
+         "JPEG parameter struct mismatch: library thinks size is %u, caller expects %u")
 JMESSAGE(JERR_BAD_VIRTUAL_ACCESS, "Bogus virtual array access")
 JMESSAGE(JERR_BUFFER_SIZE, "Buffer passed to JPEG library is too small")
 JMESSAGE(JERR_CANT_SUSPEND, "Suspension not allowed here")
@@ -98,7 +101,7 @@
 JMESSAGE(JERR_INPUT_EMPTY, "Empty input file")
 JMESSAGE(JERR_INPUT_EOF, "Premature end of input file")
 JMESSAGE(JERR_MISMATCHED_QUANT_TABLE,
-	 "Cannot transcode due to multiple use of quantization table %d")
+         "Cannot transcode due to multiple use of quantization table %d")
 JMESSAGE(JERR_MISSING_DATA, "Scan script does not transmit all data")
 JMESSAGE(JERR_MODE_CHANGE, "Invalid color quantization mode change")
 JMESSAGE(JERR_NOTIMPL, "Not implemented yet")
@@ -113,7 +116,7 @@
 JMESSAGE(JERR_NO_SOI, "Not a JPEG file: starts with 0x%02x 0x%02x")
 JMESSAGE(JERR_OUT_OF_MEMORY, "Insufficient memory (case %d)")
 JMESSAGE(JERR_QUANT_COMPONENTS,
-	 "Cannot quantize more than %d color components")
+         "Cannot quantize more than %d color components")
 JMESSAGE(JERR_QUANT_FEW_COLORS, "Cannot quantize to fewer than %d colors")
 JMESSAGE(JERR_QUANT_MANY_COLORS, "Cannot quantize to more than %d colors")
 JMESSAGE(JERR_SOF_DUPLICATE, "Invalid JPEG file structure: two SOF markers")
@@ -125,19 +128,19 @@
 JMESSAGE(JERR_TFILE_READ, "Read failed on temporary file")
 JMESSAGE(JERR_TFILE_SEEK, "Seek failed on temporary file")
 JMESSAGE(JERR_TFILE_WRITE,
-	 "Write failed on temporary file --- out of disk space?")
+         "Write failed on temporary file --- out of disk space?")
 JMESSAGE(JERR_TOO_LITTLE_DATA, "Application transferred too few scanlines")
 JMESSAGE(JERR_UNKNOWN_MARKER, "Unsupported marker type 0x%02x")
 JMESSAGE(JERR_VIRTUAL_BUG, "Virtual array controller messed up")
 JMESSAGE(JERR_WIDTH_OVERFLOW, "Image too wide for this implementation")
 JMESSAGE(JERR_XMS_READ, "Read from XMS failed")
 JMESSAGE(JERR_XMS_WRITE, "Write to XMS failed")
-JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT)
+JMESSAGE(JMSG_COPYRIGHT, JCOPYRIGHT_SHORT)
 JMESSAGE(JMSG_VERSION, JVERSION)
 JMESSAGE(JTRC_16BIT_TABLES,
-	 "Caution: quantization tables are too coarse for baseline JPEG")
+         "Caution: quantization tables are too coarse for baseline JPEG")
 JMESSAGE(JTRC_ADOBE,
-	 "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
+         "Adobe APP14 marker: version %d, flags 0x%04x 0x%04x, transform %d")
 JMESSAGE(JTRC_APP0, "Unknown APP0 marker (not JFIF), length %u")
 JMESSAGE(JTRC_APP14, "Unknown APP14 marker (not Adobe), length %u")
 JMESSAGE(JTRC_DAC, "Define Arithmetic Table 0x%02x: 0x%02x")
@@ -150,9 +153,9 @@
 JMESSAGE(JTRC_HUFFBITS, "        %3d %3d %3d %3d %3d %3d %3d %3d")
 JMESSAGE(JTRC_JFIF, "JFIF APP0 marker: version %d.%02d, density %dx%d  %d")
 JMESSAGE(JTRC_JFIF_BADTHUMBNAILSIZE,
-	 "Warning: thumbnail image size does not match data length %u")
+         "Warning: thumbnail image size does not match data length %u")
 JMESSAGE(JTRC_JFIF_EXTENSION,
-	 "JFIF extension marker: type 0x%02x, length %u")
+         "JFIF extension marker: type 0x%02x, length %u")
 JMESSAGE(JTRC_JFIF_THUMBNAIL, "    with %d x %d thumbnail image")
 JMESSAGE(JTRC_MISC_MARKER, "Miscellaneous marker 0x%02x, length %u")
 JMESSAGE(JTRC_PARMLESS_MARKER, "Unexpected marker 0x%02x")
@@ -163,7 +166,7 @@
 JMESSAGE(JTRC_RECOVERY_ACTION, "At marker 0x%02x, recovery action %d")
 JMESSAGE(JTRC_RST, "RST%d")
 JMESSAGE(JTRC_SMOOTH_NOTIMPL,
-	 "Smoothing not supported with nonstandard sampling ratios")
+         "Smoothing not supported with nonstandard sampling ratios")
 JMESSAGE(JTRC_SOF, "Start Of Frame 0x%02x: width=%u, height=%u, components=%d")
 JMESSAGE(JTRC_SOF_COMPONENT, "    Component %d: %dhx%dv q=%d")
 JMESSAGE(JTRC_SOI, "Start of Image")
@@ -173,13 +176,13 @@
 JMESSAGE(JTRC_TFILE_CLOSE, "Closed temporary file %s")
 JMESSAGE(JTRC_TFILE_OPEN, "Opened temporary file %s")
 JMESSAGE(JTRC_THUMB_JPEG,
-	 "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
+         "JFIF extension marker: JPEG-compressed thumbnail image, length %u")
 JMESSAGE(JTRC_THUMB_PALETTE,
-	 "JFIF extension marker: palette thumbnail image, length %u")
+         "JFIF extension marker: palette thumbnail image, length %u")
 JMESSAGE(JTRC_THUMB_RGB,
-	 "JFIF extension marker: RGB thumbnail image, length %u")
+         "JFIF extension marker: RGB thumbnail image, length %u")
 JMESSAGE(JTRC_UNKNOWN_IDS,
-	 "Unrecognized component IDs %d %d %d, assuming YCbCr")
+         "Unrecognized component IDs %d %d %d, assuming YCbCr")
 JMESSAGE(JTRC_XMS_CLOSE, "Freed XMS handle %u")
 JMESSAGE(JTRC_XMS_OPEN, "Obtained XMS handle %u")
 JMESSAGE(JWRN_ADOBE_XFORM, "Unknown Adobe color transform code %d")
@@ -187,15 +190,15 @@
 JMESSAGE(JWRN_ARITH_BAD_CODE, "Corrupt JPEG data: bad arithmetic code")
 #endif
 JMESSAGE(JWRN_BOGUS_PROGRESSION,
-	 "Inconsistent progression sequence for component %d coefficient %d")
+         "Inconsistent progression sequence for component %d coefficient %d")
 JMESSAGE(JWRN_EXTRANEOUS_DATA,
-	 "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
+         "Corrupt JPEG data: %u extraneous bytes before marker 0x%02x")
 JMESSAGE(JWRN_HIT_MARKER, "Corrupt JPEG data: premature end of data segment")
 JMESSAGE(JWRN_HUFF_BAD_CODE, "Corrupt JPEG data: bad Huffman code")
 JMESSAGE(JWRN_JFIF_MAJOR, "Warning: unknown JFIF revision number %d.%02d")
 JMESSAGE(JWRN_JPEG_EOF, "Premature end of JPEG file")
 JMESSAGE(JWRN_MUST_RESYNC,
-	 "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
+         "Corrupt JPEG data: found marker 0x%02x instead of RST%d")
 JMESSAGE(JWRN_NOT_SEQUENTIAL, "Invalid SOS parameters for sequential JPEG")
 JMESSAGE(JWRN_TOO_MUCH_DATA, "Application transferred too many scanlines")
 #if JPEG_LIB_VERSION < 70
@@ -255,7 +258,7 @@
    strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
    (*(cinfo)->err->error_exit) ((j_common_ptr) (cinfo)))
 
-#define MAKESTMT(stuff)		do { stuff } while (0)
+#define MAKESTMT(stuff)         do { stuff } while (0)
 
 /* Nonfatal errors (we can keep going, but the data is probably corrupt) */
 #define WARNMS(cinfo,code)  \
@@ -286,26 +289,26 @@
    (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)))
 #define TRACEMS3(cinfo,lvl,code,p1,p2,p3)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS4(cinfo,lvl,code,p1,p2,p3,p4)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS5(cinfo,lvl,code,p1,p2,p3,p4,p5)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   _mp[4] = (p5); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           _mp[4] = (p5); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMS8(cinfo,lvl,code,p1,p2,p3,p4,p5,p6,p7,p8)  \
   MAKESTMT(int * _mp = (cinfo)->err->msg_parm.i; \
-	   _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
-	   _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
-	   (cinfo)->err->msg_code = (code); \
-	   (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
+           _mp[0] = (p1); _mp[1] = (p2); _mp[2] = (p3); _mp[3] = (p4); \
+           _mp[4] = (p5); _mp[5] = (p6); _mp[6] = (p7); _mp[7] = (p8); \
+           (cinfo)->err->msg_code = (code); \
+           (*(cinfo)->err->emit_message) ((j_common_ptr) (cinfo), (lvl)); )
 #define TRACEMSS(cinfo,lvl,code,str)  \
   ((cinfo)->err->msg_code = (code), \
    strncpy((cinfo)->err->msg_parm.s, (str), JMSG_STR_PARM_MAX), \
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctflt.c glmark2-2021.02/src/libjpeg-turbo/jfdctflt.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctflt.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jfdctflt.c	2021-04-25 01:47:22.000000000 +0800
@@ -3,7 +3,8 @@
  *
  * Copyright (C) 1994-1996, Thomas G. Lane.
  * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a floating-point implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -20,8 +21,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -37,7 +38,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_FLOAT_SUPPORTED
 
@@ -56,7 +57,7 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_float (FAST_FLOAT * data)
+jpeg_fdct_float (FAST_FLOAT *data)
 {
   FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
@@ -76,24 +77,24 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[0] = tmp10 + tmp11; /* phase 3 */
     dataptr[4] = tmp10 - tmp11;
-    
+
     z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
-    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[2] = tmp13 + z1;    /* phase 5 */
     dataptr[6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -103,15 +104,15 @@
     z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
     z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
-    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[5] = z13 + z2;      /* phase 6 */
     dataptr[3] = z13 - z2;
     dataptr[1] = z11 + z4;
     dataptr[7] = z11 - z4;
 
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns. */
@@ -126,24 +127,24 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
     dataptr[DCTSIZE*4] = tmp10 - tmp11;
-    
+
     z1 = (tmp12 + tmp13) * ((FAST_FLOAT) 0.707106781); /* c4 */
     dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
     dataptr[DCTSIZE*6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -153,7 +154,7 @@
     z4 = ((FAST_FLOAT) 1.306562965) * tmp12 + z5; /* c2+c6 */
     z3 = tmp11 * ((FAST_FLOAT) 0.707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
     dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
@@ -161,7 +162,7 @@
     dataptr[DCTSIZE*1] = z11 + z4;
     dataptr[DCTSIZE*7] = z11 - z4;
 
-    dataptr++;			/* advance pointer to next column */
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctfst.c glmark2-2021.02/src/libjpeg-turbo/jfdctfst.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctfst.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jfdctfst.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jfdctfst.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a fast, not so accurate integer implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -15,8 +18,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -33,7 +36,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_IFAST_SUPPORTED
 
@@ -76,10 +79,10 @@
  */
 
 #if CONST_BITS == 8
-#define FIX_0_382683433  ((INT32)   98)		/* FIX(0.382683433) */
-#define FIX_0_541196100  ((INT32)  139)		/* FIX(0.541196100) */
-#define FIX_0_707106781  ((INT32)  181)		/* FIX(0.707106781) */
-#define FIX_1_306562965  ((INT32)  334)		/* FIX(1.306562965) */
+#define FIX_0_382683433  ((JLONG)   98)         /* FIX(0.382683433) */
+#define FIX_0_541196100  ((JLONG)  139)         /* FIX(0.541196100) */
+#define FIX_0_707106781  ((JLONG)  181)         /* FIX(0.707106781) */
+#define FIX_1_306562965  ((JLONG)  334)         /* FIX(1.306562965) */
 #else
 #define FIX_0_382683433  FIX(0.382683433)
 #define FIX_0_541196100  FIX(0.541196100)
@@ -99,7 +102,7 @@
 #endif
 
 
-/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
  * descale to yield a DCTELEM result.
  */
 
@@ -111,7 +114,7 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_ifast (DCTELEM * data)
+jpeg_fdct_ifast (DCTELEM *data)
 {
   DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   DCTELEM tmp10, tmp11, tmp12, tmp13;
@@ -132,24 +135,24 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[0] = tmp10 + tmp11; /* phase 3 */
     dataptr[4] = tmp10 - tmp11;
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
-    dataptr[2] = tmp13 + z1;	/* phase 5 */
+    dataptr[2] = tmp13 + z1;    /* phase 5 */
     dataptr[6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -159,15 +162,15 @@
     z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
     z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
-    dataptr[5] = z13 + z2;	/* phase 6 */
+    dataptr[5] = z13 + z2;      /* phase 6 */
     dataptr[3] = z13 - z2;
     dataptr[1] = z11 + z4;
     dataptr[7] = z11 - z4;
 
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns. */
@@ -182,24 +185,24 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part */
-    
-    tmp10 = tmp0 + tmp3;	/* phase 2 */
+
+    tmp10 = tmp0 + tmp3;        /* phase 2 */
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = tmp10 + tmp11; /* phase 3 */
     dataptr[DCTSIZE*4] = tmp10 - tmp11;
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_707106781); /* c4 */
     dataptr[DCTSIZE*2] = tmp13 + z1; /* phase 5 */
     dataptr[DCTSIZE*6] = tmp13 - z1;
-    
+
     /* Odd part */
 
-    tmp10 = tmp4 + tmp5;	/* phase 2 */
+    tmp10 = tmp4 + tmp5;        /* phase 2 */
     tmp11 = tmp5 + tmp6;
     tmp12 = tmp6 + tmp7;
 
@@ -209,7 +212,7 @@
     z4 = MULTIPLY(tmp12, FIX_1_306562965) + z5; /* c2+c6 */
     z3 = MULTIPLY(tmp11, FIX_0_707106781); /* c4 */
 
-    z11 = tmp7 + z3;		/* phase 5 */
+    z11 = tmp7 + z3;            /* phase 5 */
     z13 = tmp7 - z3;
 
     dataptr[DCTSIZE*5] = z13 + z2; /* phase 6 */
@@ -217,7 +220,7 @@
     dataptr[DCTSIZE*1] = z11 + z4;
     dataptr[DCTSIZE*7] = z11 - z4;
 
-    dataptr++;			/* advance pointer to next column */
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctint.c glmark2-2021.02/src/libjpeg-turbo/jfdctint.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jfdctint.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jfdctint.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jfdctint.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a slow-but-accurate integer implementation of the
  * forward DCT (Discrete Cosine Transform).
@@ -26,7 +29,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_ISLOW_SUPPORTED
 
@@ -67,7 +70,7 @@
  * they are represented to better-than-integral precision.  These outputs
  * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
  * with the recommended scaling.  (For 12-bit sample data, the intermediate
- * array is INT32 anyway.)
+ * array is JLONG anyway.)
  *
  * To avoid overflow of the 32-bit intermediate results in pass 2, we must
  * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
@@ -79,7 +82,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -90,18 +93,18 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
-#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
-#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
-#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
-#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#define FIX_0_298631336  ((JLONG)  2446)        /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)  3196)        /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)  4433)        /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)  9633)        /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)  12299)       /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)  16069)       /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)  16819)       /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)  25172)       /* FIX(3.072711026) */
 #else
 #define FIX_0_298631336  FIX(0.298631336)
 #define FIX_0_390180644  FIX(0.390180644)
@@ -118,7 +121,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -137,11 +140,11 @@
  */
 
 GLOBAL(void)
-jpeg_fdct_islow (DCTELEM * data)
+jpeg_fdct_islow (DCTELEM *data)
 {
-  INT32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
-  INT32 tmp10, tmp11, tmp12, tmp13;
-  INT32 z1, z2, z3, z4, z5;
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
   DCTELEM *dataptr;
   int ctr;
   SHIFT_TEMPS
@@ -160,36 +163,36 @@
     tmp5 = dataptr[2] - dataptr[5];
     tmp3 = dataptr[3] + dataptr[4];
     tmp4 = dataptr[3] - dataptr[4];
-    
+
     /* Even part per LL&M figure 1 --- note that published figure is faulty;
      * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
      */
-    
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
-    dataptr[0] = (DCTELEM) ((tmp10 + tmp11) << PASS1_BITS);
-    dataptr[4] = (DCTELEM) ((tmp10 - tmp11) << PASS1_BITS);
-    
+
+    dataptr[0] = (DCTELEM) LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS);
+    dataptr[4] = (DCTELEM) LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS);
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
-				   CONST_BITS-PASS1_BITS);
+                                   CONST_BITS-PASS1_BITS);
     dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
-				   CONST_BITS-PASS1_BITS);
-    
+                                   CONST_BITS-PASS1_BITS);
+
     /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
      * cK represents cos(K*pi/16).
      * i0..i3 in the paper are tmp4..tmp7 here.
      */
-    
+
     z1 = tmp4 + tmp7;
     z2 = tmp5 + tmp6;
     z3 = tmp4 + tmp6;
     z4 = tmp5 + tmp7;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -198,16 +201,16 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS);
     dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS);
     dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS);
     dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS);
-    
-    dataptr += DCTSIZE;		/* advance pointer to next row */
+
+    dataptr += DCTSIZE;         /* advance pointer to next row */
   }
 
   /* Pass 2: process columns.
@@ -225,36 +228,36 @@
     tmp5 = dataptr[DCTSIZE*2] - dataptr[DCTSIZE*5];
     tmp3 = dataptr[DCTSIZE*3] + dataptr[DCTSIZE*4];
     tmp4 = dataptr[DCTSIZE*3] - dataptr[DCTSIZE*4];
-    
+
     /* Even part per LL&M figure 1 --- note that published figure is faulty;
      * rotator "sqrt(2)*c1" should be "sqrt(2)*c6".
      */
-    
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     dataptr[DCTSIZE*0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS);
     dataptr[DCTSIZE*4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS);
-    
+
     z1 = MULTIPLY(tmp12 + tmp13, FIX_0_541196100);
     dataptr[DCTSIZE*2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, FIX_0_765366865),
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, - FIX_1_847759065),
-					   CONST_BITS+PASS1_BITS);
-    
+                                           CONST_BITS+PASS1_BITS);
+
     /* Odd part per figure 8 --- note paper omits factor of sqrt(2).
      * cK represents cos(K*pi/16).
      * i0..i3 in the paper are tmp4..tmp7 here.
      */
-    
+
     z1 = tmp4 + tmp7;
     z2 = tmp5 + tmp6;
     z3 = tmp4 + tmp6;
     z4 = tmp5 + tmp7;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp4 = MULTIPLY(tmp4, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp5 = MULTIPLY(tmp5, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp6 = MULTIPLY(tmp6, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -263,20 +266,20 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     dataptr[DCTSIZE*7] = (DCTELEM) DESCALE(tmp4 + z1 + z3,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*5] = (DCTELEM) DESCALE(tmp5 + z2 + z4,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*3] = (DCTELEM) DESCALE(tmp6 + z2 + z3,
-					   CONST_BITS+PASS1_BITS);
+                                           CONST_BITS+PASS1_BITS);
     dataptr[DCTSIZE*1] = (DCTELEM) DESCALE(tmp7 + z1 + z4,
-					   CONST_BITS+PASS1_BITS);
-    
-    dataptr++;			/* advance pointer to next column */
+                                           CONST_BITS+PASS1_BITS);
+
+    dataptr++;                  /* advance pointer to next column */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctflt.c glmark2-2021.02/src/libjpeg-turbo/jidctflt.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctflt.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jidctflt.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
  * jidctflt.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a floating-point implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -22,8 +26,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -39,7 +43,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_FLOAT_SUPPORTED
 
@@ -65,21 +69,21 @@
  */
 
 GLOBAL(void)
-jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   FAST_FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   FAST_FLOAT tmp10, tmp11, tmp12, tmp13;
   FAST_FLOAT z5, z10, z11, z12, z13;
   JCOEFPTR inptr;
-  FLOAT_MULT_TYPE * quantptr;
-  FAST_FLOAT * wsptr;
+  FLOAT_MULT_TYPE *quantptr;
+  FAST_FLOAT *wsptr;
   JSAMPROW outptr;
-  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  JSAMPLE *range_limit = cinfo->sample_range_limit;
   int ctr;
   FAST_FLOAT workspace[DCTSIZE2]; /* buffers data between passes */
-  SHIFT_TEMPS
+  #define _0_125 ((FLOAT_MULT_TYPE)0.125)
 
   /* Pass 1: process columns from input, store into work array. */
 
@@ -95,14 +99,15 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
-      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-      
+      FAST_FLOAT dcval = DEQUANTIZE(inptr[DCTSIZE*0],
+                                    quantptr[DCTSIZE*0] * _0_125);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
@@ -111,53 +116,53 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part */
 
-    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
-    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
-    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0] * _0_125);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2] * _0_125);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4] * _0_125);
+    tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6] * _0_125);
 
-    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
     tmp11 = tmp0 - tmp2;
 
-    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
     tmp12 = (tmp1 - tmp3) * ((FAST_FLOAT) 1.414213562) - tmp13; /* 2*c4 */
 
-    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
     tmp3 = tmp10 - tmp13;
     tmp1 = tmp11 + tmp12;
     tmp2 = tmp11 - tmp12;
-    
+
     /* Odd part */
 
-    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
-    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
-    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1] * _0_125);
+    tmp5 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3] * _0_125);
+    tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5] * _0_125);
+    tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7] * _0_125);
 
-    z13 = tmp6 + tmp5;		/* phase 6 */
+    z13 = tmp6 + tmp5;          /* phase 6 */
     z10 = tmp6 - tmp5;
     z11 = tmp4 + tmp7;
     z12 = tmp4 - tmp7;
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562); /* 2*c4 */
 
     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
-    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
-    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+    tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
-    tmp4 = tmp10 + tmp5;
+    tmp4 = tmp10 - tmp5;
 
     wsptr[DCTSIZE*0] = tmp0 + tmp7;
     wsptr[DCTSIZE*7] = tmp0 - tmp7;
@@ -165,16 +170,15 @@
     wsptr[DCTSIZE*6] = tmp1 - tmp6;
     wsptr[DCTSIZE*2] = tmp2 + tmp5;
     wsptr[DCTSIZE*5] = tmp2 - tmp5;
-    wsptr[DCTSIZE*4] = tmp3 + tmp4;
-    wsptr[DCTSIZE*3] = tmp3 - tmp4;
+    wsptr[DCTSIZE*3] = tmp3 + tmp4;
+    wsptr[DCTSIZE*4] = tmp3 - tmp4;
 
-    inptr++;			/* advance pointers to next column */
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
-  /* Note that we must descale the results by a factor of 8 == 2**3. */
 
   wsptr = workspace;
   for (ctr = 0; ctr < DCTSIZE; ctr++) {
@@ -184,11 +188,13 @@
      * the simplification applies less often (typically 5% to 10% of the time).
      * And testing floats for zero is relatively expensive, so we don't bother.
      */
-    
+
     /* Even part */
 
-    tmp10 = wsptr[0] + wsptr[4];
-    tmp11 = wsptr[0] - wsptr[4];
+    /* Apply signed->unsigned and prepare float->int conversion */
+    z5 = wsptr[0] + ((FAST_FLOAT) CENTERJSAMPLE + (FAST_FLOAT) 0.5);
+    tmp10 = z5 + wsptr[4];
+    tmp11 = z5 - wsptr[4];
 
     tmp13 = wsptr[2] + wsptr[6];
     tmp12 = (wsptr[2] - wsptr[6]) * ((FAST_FLOAT) 1.414213562) - tmp13;
@@ -209,33 +215,25 @@
     tmp11 = (z11 - z13) * ((FAST_FLOAT) 1.414213562);
 
     z5 = (z10 + z12) * ((FAST_FLOAT) 1.847759065); /* 2*c2 */
-    tmp10 = ((FAST_FLOAT) 1.082392200) * z12 - z5; /* 2*(c2-c6) */
-    tmp12 = ((FAST_FLOAT) -2.613125930) * z10 + z5; /* -2*(c2+c6) */
+    tmp10 = z5 - z12 * ((FAST_FLOAT) 1.082392200); /* 2*(c2-c6) */
+    tmp12 = z5 - z10 * ((FAST_FLOAT) 2.613125930); /* 2*(c2+c6) */
 
     tmp6 = tmp12 - tmp7;
     tmp5 = tmp11 - tmp6;
-    tmp4 = tmp10 + tmp5;
+    tmp4 = tmp10 - tmp5;
+
+    /* Final output stage: float->int conversion and range-limit */
 
-    /* Final output stage: scale down by a factor of 8 and range-limit */
+    outptr[0] = range_limit[((int) (tmp0 + tmp7)) & RANGE_MASK];
+    outptr[7] = range_limit[((int) (tmp0 - tmp7)) & RANGE_MASK];
+    outptr[1] = range_limit[((int) (tmp1 + tmp6)) & RANGE_MASK];
+    outptr[6] = range_limit[((int) (tmp1 - tmp6)) & RANGE_MASK];
+    outptr[2] = range_limit[((int) (tmp2 + tmp5)) & RANGE_MASK];
+    outptr[5] = range_limit[((int) (tmp2 - tmp5)) & RANGE_MASK];
+    outptr[3] = range_limit[((int) (tmp3 + tmp4)) & RANGE_MASK];
+    outptr[4] = range_limit[((int) (tmp3 - tmp4)) & RANGE_MASK];
 
-    outptr[0] = range_limit[(int) DESCALE((INT32) (tmp0 + tmp7), 3)
-			    & RANGE_MASK];
-    outptr[7] = range_limit[(int) DESCALE((INT32) (tmp0 - tmp7), 3)
-			    & RANGE_MASK];
-    outptr[1] = range_limit[(int) DESCALE((INT32) (tmp1 + tmp6), 3)
-			    & RANGE_MASK];
-    outptr[6] = range_limit[(int) DESCALE((INT32) (tmp1 - tmp6), 3)
-			    & RANGE_MASK];
-    outptr[2] = range_limit[(int) DESCALE((INT32) (tmp2 + tmp5), 3)
-			    & RANGE_MASK];
-    outptr[5] = range_limit[(int) DESCALE((INT32) (tmp2 - tmp5), 3)
-			    & RANGE_MASK];
-    outptr[4] = range_limit[(int) DESCALE((INT32) (tmp3 + tmp4), 3)
-			    & RANGE_MASK];
-    outptr[3] = range_limit[(int) DESCALE((INT32) (tmp3 - tmp4), 3)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctfst.c glmark2-2021.02/src/libjpeg-turbo/jidctfst.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctfst.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jidctfst.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jidctfst.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a fast, not so accurate integer implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -17,8 +20,8 @@
  * This implementation is based on Arai, Agui, and Nakajima's algorithm for
  * scaled DCT.  Their original paper (Trans. IEICE E-71(11):1095) is in
  * Japanese, but the algorithm is described in the Pennebaker & Mitchell
- * JPEG textbook (see REFERENCES section in file README).  The following code
- * is based directly on figure 4-8 in P&M.
+ * JPEG textbook (see REFERENCES section in file README.ijg).  The following
+ * code is based directly on figure 4-8 in P&M.
  * While an 8-point DCT cannot be done in less than 11 multiplies, it is
  * possible to arrange the computation so that many of the multiplies are
  * simple scalings of the final outputs.  These multiplies can then be
@@ -35,7 +38,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_IFAST_SUPPORTED
 
@@ -78,7 +81,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  8
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -89,10 +92,10 @@
  */
 
 #if CONST_BITS == 8
-#define FIX_1_082392200  ((INT32)  277)		/* FIX(1.082392200) */
-#define FIX_1_414213562  ((INT32)  362)		/* FIX(1.414213562) */
-#define FIX_1_847759065  ((INT32)  473)		/* FIX(1.847759065) */
-#define FIX_2_613125930  ((INT32)  669)		/* FIX(2.613125930) */
+#define FIX_1_082392200  ((JLONG)  277)         /* FIX(1.082392200) */
+#define FIX_1_414213562  ((JLONG)  362)         /* FIX(1.414213562) */
+#define FIX_1_847759065  ((JLONG)  473)         /* FIX(1.847759065) */
+#define FIX_2_613125930  ((JLONG)  669)         /* FIX(2.613125930) */
 #else
 #define FIX_1_082392200  FIX(1.082392200)
 #define FIX_1_414213562  FIX(1.414213562)
@@ -112,7 +115,7 @@
 #endif
 
 
-/* Multiply a DCTELEM variable by an INT32 constant, and immediately
+/* Multiply a DCTELEM variable by an JLONG constant, and immediately
  * descale to yield a DCTELEM result.
  */
 
@@ -122,27 +125,27 @@
 /* Dequantize a coefficient by multiplying it by the multiplier-table
  * entry; produce a DCTELEM result.  For 8-bit data a 16x16->16
  * multiplication will do.  For 12-bit data, the multiplier table is
- * declared INT32, so a 32-bit multiply will be used.
+ * declared JLONG, so a 32-bit multiply will be used.
  */
 
 #if BITS_IN_JSAMPLE == 8
 #define DEQUANTIZE(coef,quantval)  (((IFAST_MULT_TYPE) (coef)) * (quantval))
 #else
 #define DEQUANTIZE(coef,quantval)  \
-	DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
+        DESCALE((coef)*(quantval), IFAST_SCALE_BITS-PASS1_BITS)
 #endif
 
 
 /* Like DESCALE, but applies to a DCTELEM and produces an int.
- * We assume that int right shift is unsigned if INT32 right shift is.
+ * We assume that int right shift is unsigned if JLONG right shift is.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define ISHIFT_TEMPS	DCTELEM ishift_temp;
+#define ISHIFT_TEMPS    DCTELEM ishift_temp;
 #if BITS_IN_JSAMPLE == 8
-#define DCTELEMBITS  16		/* DCTELEM may be 16 or 32 bits */
+#define DCTELEMBITS  16         /* DCTELEM may be 16 or 32 bits */
 #else
-#define DCTELEMBITS  32		/* DCTELEM must be 32 bits */
+#define DCTELEMBITS  32         /* DCTELEM must be 32 bits */
 #endif
 #define IRIGHT_SHIFT(x,shft)  \
     ((ishift_temp = (x)) < 0 ? \
@@ -150,7 +153,7 @@
      (ishift_temp >> (shft)))
 #else
 #define ISHIFT_TEMPS
-#define IRIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define IRIGHT_SHIFT(x,shft)    ((x) >> (shft))
 #endif
 
 #ifdef USE_ACCURATE_ROUNDING
@@ -165,22 +168,22 @@
  */
 
 GLOBAL(void)
-jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   DCTELEM tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
   DCTELEM tmp10, tmp11, tmp12, tmp13;
   DCTELEM z5, z10, z11, z12, z13;
   JCOEFPTR inptr;
-  IFAST_MULT_TYPE * quantptr;
-  int * wsptr;
+  IFAST_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE2];	/* buffers data between passes */
-  SHIFT_TEMPS			/* for DESCALE */
-  ISHIFT_TEMPS			/* for IDESCALE */
+  int workspace[DCTSIZE2];      /* buffers data between passes */
+  SHIFT_TEMPS                   /* for DESCALE */
+  ISHIFT_TEMPS                  /* for IDESCALE */
 
   /* Pass 1: process columns from input, store into work array. */
 
@@ -196,11 +199,11 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
       int dcval = (int) DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
 
@@ -212,13 +215,13 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part */
 
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
@@ -226,17 +229,17 @@
     tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
     tmp3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
 
-    tmp10 = tmp0 + tmp2;	/* phase 3 */
+    tmp10 = tmp0 + tmp2;        /* phase 3 */
     tmp11 = tmp0 - tmp2;
 
-    tmp13 = tmp1 + tmp3;	/* phases 5-3 */
+    tmp13 = tmp1 + tmp3;        /* phases 5-3 */
     tmp12 = MULTIPLY(tmp1 - tmp3, FIX_1_414213562) - tmp13; /* 2*c4 */
 
-    tmp0 = tmp10 + tmp13;	/* phase 2 */
+    tmp0 = tmp10 + tmp13;       /* phase 2 */
     tmp3 = tmp10 - tmp13;
     tmp1 = tmp11 + tmp12;
     tmp2 = tmp11 - tmp12;
-    
+
     /* Odd part */
 
     tmp4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
@@ -244,19 +247,19 @@
     tmp6 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     tmp7 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
 
-    z13 = tmp6 + tmp5;		/* phase 6 */
+    z13 = tmp6 + tmp5;          /* phase 6 */
     z10 = tmp6 - tmp5;
     z11 = tmp4 + tmp7;
     z12 = tmp4 - tmp7;
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
 
     z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
     tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
     tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
     tmp4 = tmp10 + tmp5;
 
@@ -269,11 +272,11 @@
     wsptr[DCTSIZE*4] = (int) (tmp3 + tmp4);
     wsptr[DCTSIZE*3] = (int) (tmp3 - tmp4);
 
-    inptr++;			/* advance pointers to next column */
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
   /* Note that we must descale the results by a factor of 8 == 2**3, */
   /* and also undo the PASS1_BITS scaling. */
@@ -288,14 +291,14 @@
      * test takes more time than it's worth.  In that case this section
      * may be commented out.
      */
-    
+
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
       JSAMPLE dcval = range_limit[IDESCALE(wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
@@ -305,11 +308,11 @@
       outptr[6] = dcval;
       outptr[7] = dcval;
 
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
 
     tmp10 = ((DCTELEM) wsptr[0] + (DCTELEM) wsptr[4]);
@@ -317,7 +320,7 @@
 
     tmp13 = ((DCTELEM) wsptr[2] + (DCTELEM) wsptr[6]);
     tmp12 = MULTIPLY((DCTELEM) wsptr[2] - (DCTELEM) wsptr[6], FIX_1_414213562)
-	    - tmp13;
+            - tmp13;
 
     tmp0 = tmp10 + tmp13;
     tmp3 = tmp10 - tmp13;
@@ -331,37 +334,37 @@
     z11 = (DCTELEM) wsptr[1] + (DCTELEM) wsptr[7];
     z12 = (DCTELEM) wsptr[1] - (DCTELEM) wsptr[7];
 
-    tmp7 = z11 + z13;		/* phase 5 */
+    tmp7 = z11 + z13;           /* phase 5 */
     tmp11 = MULTIPLY(z11 - z13, FIX_1_414213562); /* 2*c4 */
 
     z5 = MULTIPLY(z10 + z12, FIX_1_847759065); /* 2*c2 */
     tmp10 = MULTIPLY(z12, FIX_1_082392200) - z5; /* 2*(c2-c6) */
     tmp12 = MULTIPLY(z10, - FIX_2_613125930) + z5; /* -2*(c2+c6) */
 
-    tmp6 = tmp12 - tmp7;	/* phase 2 */
+    tmp6 = tmp12 - tmp7;        /* phase 2 */
     tmp5 = tmp11 - tmp6;
     tmp4 = tmp10 + tmp5;
 
     /* Final output stage: scale down by a factor of 8 and range-limit */
 
     outptr[0] = range_limit[IDESCALE(tmp0 + tmp7, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[7] = range_limit[IDESCALE(tmp0 - tmp7, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[1] = range_limit[IDESCALE(tmp1 + tmp6, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[6] = range_limit[IDESCALE(tmp1 - tmp6, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[2] = range_limit[IDESCALE(tmp2 + tmp5, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[5] = range_limit[IDESCALE(tmp2 - tmp5, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[4] = range_limit[IDESCALE(tmp3 + tmp4, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
     outptr[3] = range_limit[IDESCALE(tmp3 - tmp4, PASS1_BITS+3)
-			    & RANGE_MASK];
+                            & RANGE_MASK];
 
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctint.c glmark2-2021.02/src/libjpeg-turbo/jidctint.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctint.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jidctint.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
  * jidctint.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1991-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modification developed 2002-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains a slow-but-accurate integer implementation of the
  * inverse DCT (Discrete Cosine Transform).  In the IJG code, this routine
@@ -23,12 +27,33 @@
  * The advantage of this method is that no data path contains more than one
  * multiplication; this allows a very simple and accurate implementation in
  * scaled fixed-point arithmetic, with a minimal number of shifts.
+ *
+ * We also provide IDCT routines with various output sample block sizes for
+ * direct resolution reduction or enlargement without additional resampling:
+ * NxN (N=1...16) pixels for one 8x8 input DCT block.
+ *
+ * For N<8 we simply take the corresponding low-frequency coefficients of
+ * the 8x8 input DCT block and apply an NxN point IDCT on the sub-block
+ * to yield the downscaled outputs.
+ * This can be seen as direct low-pass downsampling from the DCT domain
+ * point of view rather than the usual spatial domain point of view,
+ * yielding significant computational savings and results at least
+ * as good as common bilinear (averaging) spatial downsampling.
+ *
+ * For N>8 we apply a partial NxN IDCT on the 8 input coefficients as
+ * lower frequencies and higher frequencies assumed to be zero.
+ * It turns out that the computational effort is similar to the 8x8 IDCT
+ * regarding the output size.
+ * Furthermore, the scaling and descaling is the same for all IDCT sizes.
+ *
+ * CAUTION: We rely on the FIX() macro except for the N=1,2,4,8 cases
+ * since there would be too many additional constants to pre-calculate.
  */
 
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef DCT_ISLOW_SUPPORTED
 
@@ -38,7 +63,7 @@
  */
 
 #if DCTSIZE != 8
-  Sorry, this code only copes with 8x8 DCTs. /* deliberate syntax err */
+  Sorry, this code only copes with 8x8 DCT blocks. /* deliberate syntax err */
 #endif
 
 
@@ -67,7 +92,7 @@
  * they are represented to better-than-integral precision.  These outputs
  * require BITS_IN_JSAMPLE + PASS1_BITS + 3 bits; this fits in a 16-bit word
  * with the recommended scaling.  (To scale up 12-bit sample data further, an
- * intermediate INT32 array would be needed.)
+ * intermediate JLONG array would be needed.)
  *
  * To avoid overflow of the 32-bit intermediate results in pass 2, we must
  * have BITS_IN_JSAMPLE + CONST_BITS + PASS1_BITS <= 26.  Error analysis
@@ -79,7 +104,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -90,18 +115,18 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_298631336  ((INT32)  2446)	/* FIX(0.298631336) */
-#define FIX_0_390180644  ((INT32)  3196)	/* FIX(0.390180644) */
-#define FIX_0_541196100  ((INT32)  4433)	/* FIX(0.541196100) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_175875602  ((INT32)  9633)	/* FIX(1.175875602) */
-#define FIX_1_501321110  ((INT32)  12299)	/* FIX(1.501321110) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_1_961570560  ((INT32)  16069)	/* FIX(1.961570560) */
-#define FIX_2_053119869  ((INT32)  16819)	/* FIX(2.053119869) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_072711026  ((INT32)  25172)	/* FIX(3.072711026) */
+#define FIX_0_298631336  ((JLONG)  2446)        /* FIX(0.298631336) */
+#define FIX_0_390180644  ((JLONG)  3196)        /* FIX(0.390180644) */
+#define FIX_0_541196100  ((JLONG)  4433)        /* FIX(0.541196100) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_175875602  ((JLONG)  9633)        /* FIX(1.175875602) */
+#define FIX_1_501321110  ((JLONG)  12299)       /* FIX(1.501321110) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_1_961570560  ((JLONG)  16069)       /* FIX(1.961570560) */
+#define FIX_2_053119869  ((JLONG)  16819)       /* FIX(2.053119869) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_072711026  ((JLONG)  25172)       /* FIX(3.072711026) */
 #else
 #define FIX_0_298631336  FIX(0.298631336)
 #define FIX_0_390180644  FIX(0.390180644)
@@ -118,7 +143,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply an JLONG variable by an JLONG constant to yield an JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -145,20 +170,20 @@
  */
 
 GLOBAL(void)
-jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp1, tmp2, tmp3;
-  INT32 tmp10, tmp11, tmp12, tmp13;
-  INT32 z1, z2, z3, z4, z5;
+  JLONG tmp0, tmp1, tmp2, tmp3;
+  JLONG tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3, z4, z5;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE2];	/* buffers data between passes */
+  int workspace[DCTSIZE2];      /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -177,14 +202,15 @@
      * With typical images and quantization tables, half or more of the
      * column DCT calculations can be simplified this way.
      */
-    
+
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
-	inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*4] == 0 &&
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*6] == 0 &&
+        inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
@@ -193,49 +219,49 @@
       wsptr[DCTSIZE*5] = dcval;
       wsptr[DCTSIZE*6] = dcval;
       wsptr[DCTSIZE*7] = dcval;
-      
-      inptr++;			/* advance pointers to next column */
+
+      inptr++;                  /* advance pointers to next column */
       quantptr++;
       wsptr++;
       continue;
     }
-    
+
     /* Even part: reverse the even part of the forward DCT. */
     /* The rotator is sqrt(2)*c(-6). */
-    
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
-    
+
     z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
     tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
     tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
-    
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
 
-    tmp0 = (z2 + z3) << CONST_BITS;
-    tmp1 = (z2 - z3) << CONST_BITS;
-    
+    tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS);
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     /* Odd part per figure 8; the matrix is unitary and hence its
      * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
      */
-    
+
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
     tmp1 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     tmp2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
     tmp3 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    
+
     z1 = tmp0 + tmp3;
     z2 = tmp1 + tmp2;
     z3 = tmp0 + tmp2;
     z4 = tmp1 + tmp3;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -244,17 +270,17 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     tmp0 += z1 + z3;
     tmp1 += z2 + z4;
     tmp2 += z2 + z3;
     tmp3 += z1 + z4;
-    
+
     /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS);
@@ -263,12 +289,12 @@
     wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS);
     wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS);
-    
-    inptr++;			/* advance pointers to next column */
+
+    inptr++;                    /* advance pointers to next column */
     quantptr++;
     wsptr++;
   }
-  
+
   /* Pass 2: process rows from work array, store into output array. */
   /* Note that we must descale the results by a factor of 8 == 2**3, */
   /* and also undo the PASS1_BITS scaling. */
@@ -283,14 +309,14 @@
      * test takes more time than it's worth.  In that case this section
      * may be commented out.
      */
-    
+
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 && wsptr[4] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
@@ -300,44 +326,44 @@
       outptr[6] = dcval;
       outptr[7] = dcval;
 
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part: reverse the even part of the forward DCT. */
     /* The rotator is sqrt(2)*c(-6). */
-    
-    z2 = (INT32) wsptr[2];
-    z3 = (INT32) wsptr[6];
-    
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[6];
+
     z1 = MULTIPLY(z2 + z3, FIX_0_541196100);
     tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065);
     tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865);
-    
-    tmp0 = ((INT32) wsptr[0] + (INT32) wsptr[4]) << CONST_BITS;
-    tmp1 = ((INT32) wsptr[0] - (INT32) wsptr[4]) << CONST_BITS;
-    
+
+    tmp0 = LEFT_SHIFT((JLONG) wsptr[0] + (JLONG) wsptr[4], CONST_BITS);
+    tmp1 = LEFT_SHIFT((JLONG) wsptr[0] - (JLONG) wsptr[4], CONST_BITS);
+
     tmp10 = tmp0 + tmp3;
     tmp13 = tmp0 - tmp3;
     tmp11 = tmp1 + tmp2;
     tmp12 = tmp1 - tmp2;
-    
+
     /* Odd part per figure 8; the matrix is unitary and hence its
      * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
      */
-    
-    tmp0 = (INT32) wsptr[7];
-    tmp1 = (INT32) wsptr[5];
-    tmp2 = (INT32) wsptr[3];
-    tmp3 = (INT32) wsptr[1];
-    
+
+    tmp0 = (JLONG) wsptr[7];
+    tmp1 = (JLONG) wsptr[5];
+    tmp2 = (JLONG) wsptr[3];
+    tmp3 = (JLONG) wsptr[1];
+
     z1 = tmp0 + tmp3;
     z2 = tmp1 + tmp2;
     z3 = tmp0 + tmp2;
     z4 = tmp1 + tmp3;
     z5 = MULTIPLY(z3 + z4, FIX_1_175875602); /* sqrt(2) * c3 */
-    
+
     tmp0 = MULTIPLY(tmp0, FIX_0_298631336); /* sqrt(2) * (-c1+c3+c5-c7) */
     tmp1 = MULTIPLY(tmp1, FIX_2_053119869); /* sqrt(2) * ( c1+c3-c5+c7) */
     tmp2 = MULTIPLY(tmp2, FIX_3_072711026); /* sqrt(2) * ( c1+c3+c5-c7) */
@@ -346,44 +372,2256 @@
     z2 = MULTIPLY(z2, - FIX_2_562915447); /* sqrt(2) * (-c1-c3) */
     z3 = MULTIPLY(z3, - FIX_1_961570560); /* sqrt(2) * (-c3-c5) */
     z4 = MULTIPLY(z4, - FIX_0_390180644); /* sqrt(2) * (c5-c3) */
-    
+
     z3 += z5;
     z4 += z5;
-    
+
     tmp0 += z1 + z3;
     tmp1 += z2 + z4;
     tmp2 += z2 + z3;
     tmp3 += z1 + z4;
-    
+
     /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp3,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[7] = range_limit[(int) DESCALE(tmp10 - tmp3,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp11 + tmp2,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[6] = range_limit[(int) DESCALE(tmp11 - tmp2,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[2] = range_limit[(int) DESCALE(tmp12 + tmp1,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[5] = range_limit[(int) DESCALE(tmp12 - tmp1,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[3] = range_limit[(int) DESCALE(tmp13 + tmp0,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
     outptr[4] = range_limit[(int) DESCALE(tmp13 - tmp0,
-					  CONST_BITS+PASS1_BITS+3)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
+  }
+}
+
+#ifdef IDCT_SCALING_SUPPORTED
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 7x7 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/14).
+ */
+
+GLOBAL(void)
+jpeg_idct_7x7 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12, tmp13;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[7*7];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp13 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp13 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276));    /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    wsptr[7*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[7*6] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[7*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[7*5] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[7*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[7*4] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[7*3] = (int) RIGHT_SHIFT(tmp13, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 7 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 7; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp13 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp13 = LEFT_SHIFT(tmp13, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp10 = MULTIPLY(z2 - z3, FIX(0.881747734));     /* c4 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.314692123));     /* c6 */
+    tmp11 = tmp10 + tmp12 + tmp13 - MULTIPLY(z2, FIX(1.841218003)); /* c2+c4-c6 */
+    tmp0 = z1 + z3;
+    z2 -= tmp0;
+    tmp0 = MULTIPLY(tmp0, FIX(1.274162392)) + tmp13; /* c2 */
+    tmp10 += tmp0 - MULTIPLY(z3, FIX(0.077722536));  /* c2-c4-c6 */
+    tmp12 += tmp0 - MULTIPLY(z1, FIX(2.470602249));  /* c2+c4+c6 */
+    tmp13 += MULTIPLY(z2, FIX(1.414213562));         /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+
+    tmp1 = MULTIPLY(z1 + z2, FIX(0.935414347));      /* (c3+c1-c5)/2 */
+    tmp2 = MULTIPLY(z1 - z2, FIX(0.170262339));      /* (c3+c5-c1)/2 */
+    tmp0 = tmp1 - tmp2;
+    tmp1 += tmp2;
+    tmp2 = MULTIPLY(z2 + z3, - FIX(1.378756276));    /* -c1 */
+    tmp1 += tmp2;
+    z2 = MULTIPLY(z1 + z3, FIX(0.613604268));        /* c5 */
+    tmp0 += z2;
+    tmp2 += z2 + MULTIPLY(z3, FIX(1.870828693));     /* c3+c1-c5 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 7;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 6x6 output block.
+ *
+ * Optimized algorithm with 3 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/12).
+ */
+
+GLOBAL(void)
+jpeg_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[6*6];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = RIGHT_SHIFT(tmp0 - tmp10 - tmp10, CONST_BITS-PASS1_BITS);
+    tmp10 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr[6*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[6*5] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[6*1] = (int) (tmp11 + tmp1);
+    wsptr[6*4] = (int) (tmp11 - tmp1);
+    wsptr[6*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[6*3] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 6 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 6; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG) wsptr[4];
+    tmp10 = MULTIPLY(tmp2, FIX(0.707106781));   /* c4 */
+    tmp1 = tmp0 + tmp10;
+    tmp11 = tmp0 - tmp10 - tmp10;
+    tmp10 = (JLONG) wsptr[2];
+    tmp0 = MULTIPLY(tmp10, FIX(1.224744871));   /* c2 */
+    tmp10 = tmp1 + tmp0;
+    tmp12 = tmp1 - tmp0;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    tmp1 = MULTIPLY(z1 + z3, FIX(0.366025404)); /* c5 */
+    tmp0 = tmp1 + LEFT_SHIFT(z1 + z2, CONST_BITS);
+    tmp2 = tmp1 + LEFT_SHIFT(z3 - z2, CONST_BITS);
+    tmp1 = LEFT_SHIFT(z1 - z2 - z3, CONST_BITS);
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 6;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 5x5 output block.
+ *
+ * Optimized algorithm with 5 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/10).
+ */
+
+GLOBAL(void)
+jpeg_idct_5x5 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp10, tmp11, tmp12;
+  JLONG z1, z2, z3;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[5*5];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp12 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp12 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    wsptr[5*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[5*4] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[5*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[5*3] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[5*2] = (int) RIGHT_SHIFT(tmp12, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 5 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 5; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp12 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp12 = LEFT_SHIFT(tmp12, CONST_BITS);
+    tmp0 = (JLONG) wsptr[2];
+    tmp1 = (JLONG) wsptr[4];
+    z1 = MULTIPLY(tmp0 + tmp1, FIX(0.790569415)); /* (c2+c4)/2 */
+    z2 = MULTIPLY(tmp0 - tmp1, FIX(0.353553391)); /* (c2-c4)/2 */
+    z3 = tmp12 + z2;
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z1;
+    tmp12 -= LEFT_SHIFT(z2, 2);
+
+    /* Odd part */
+
+    z2 = (JLONG) wsptr[1];
+    z3 = (JLONG) wsptr[3];
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));     /* c3 */
+    tmp0 = z1 + MULTIPLY(z2, FIX(0.513743148));   /* c1-c3 */
+    tmp1 = z1 - MULTIPLY(z3, FIX(2.176250899));   /* c1+c3 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 5;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a reduced-size 3x3 output block.
+ *
+ * Optimized algorithm with 2 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/6).
+ */
+
+GLOBAL(void)
+jpeg_idct_3x3 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[3*3];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+    tmp2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    wsptr[3*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[3*2] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[3*1] = (int) RIGHT_SHIFT(tmp2, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 3 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 3; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    tmp2 = (JLONG) wsptr[2];
+    tmp12 = MULTIPLY(tmp2, FIX(0.707106781)); /* c2 */
+    tmp10 = tmp0 + tmp12;
+    tmp2 = tmp0 - tmp12 - tmp12;
+
+    /* Odd part */
+
+    tmp12 = (JLONG) wsptr[1];
+    tmp0 = MULTIPLY(tmp12, FIX(1.224744871)); /* c1 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 3;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 9x9 output block.
+ *
+ * Optimized algorithm with 10 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/18).
+ */
+
+GLOBAL(void)
+jpeg_idct_9x9 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*9];   /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    wsptr[8*0] = (int) RIGHT_SHIFT(tmp10 + tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[8*8] = (int) RIGHT_SHIFT(tmp10 - tmp0, CONST_BITS-PASS1_BITS);
+    wsptr[8*1] = (int) RIGHT_SHIFT(tmp11 + tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[8*7] = (int) RIGHT_SHIFT(tmp11 - tmp1, CONST_BITS-PASS1_BITS);
+    wsptr[8*2] = (int) RIGHT_SHIFT(tmp12 + tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[8*6] = (int) RIGHT_SHIFT(tmp12 - tmp2, CONST_BITS-PASS1_BITS);
+    wsptr[8*3] = (int) RIGHT_SHIFT(tmp13 + tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[8*5] = (int) RIGHT_SHIFT(tmp13 - tmp3, CONST_BITS-PASS1_BITS);
+    wsptr[8*4] = (int) RIGHT_SHIFT(tmp14, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 9 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 9; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp3 = MULTIPLY(z3, FIX(0.707106781));      /* c6 */
+    tmp1 = tmp0 + tmp3;
+    tmp2 = tmp0 - tmp3 - tmp3;
+
+    tmp0 = MULTIPLY(z1 - z2, FIX(0.707106781)); /* c6 */
+    tmp11 = tmp2 + tmp0;
+    tmp14 = tmp2 - tmp0 - tmp0;
+
+    tmp0 = MULTIPLY(z1 + z2, FIX(1.328926049)); /* c2 */
+    tmp2 = MULTIPLY(z1, FIX(1.083350441));      /* c4 */
+    tmp3 = MULTIPLY(z2, FIX(0.245575608));      /* c8 */
+
+    tmp10 = tmp1 + tmp0 - tmp3;
+    tmp12 = tmp1 - tmp0 + tmp2;
+    tmp13 = tmp1 - tmp2 + tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    z2 = MULTIPLY(z2, - FIX(1.224744871));           /* -c3 */
+
+    tmp2 = MULTIPLY(z1 + z3, FIX(0.909038955));      /* c5 */
+    tmp3 = MULTIPLY(z1 + z4, FIX(0.483689525));      /* c7 */
+    tmp0 = tmp2 + tmp3 - z2;
+    tmp1 = MULTIPLY(z3 - z4, FIX(1.392728481));      /* c1 */
+    tmp2 += z2 - tmp1;
+    tmp3 += z2 + tmp1;
+    tmp1 = MULTIPLY(z1 - z3 - z4, FIX(1.224744871)); /* c3 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp10 + tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp10 - tmp0,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp11 + tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp11 - tmp1,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp12 + tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp12 - tmp2,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp13 + tmp3,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp13 - tmp3,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 10x10 output block.
+ *
+ * Optimized algorithm with 12 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/20).
+ */
+
+GLOBAL(void)
+jpeg_idct_10x10 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24;
+  JLONG z1, z2, z3, z4, z5;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*10];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS-PASS1_BITS-1);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = RIGHT_SHIFT(z3 - LEFT_SHIFT(z1 - z2, 1),
+                        CONST_BITS-PASS1_BITS);  /* c0 = (c4-c8)*2 */
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+    z5 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z5 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z5 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13 - z3, PASS1_BITS);
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    wsptr[8*0] = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*9] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1] = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*8] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2] = (int) (tmp22 + tmp12);
+    wsptr[8*7] = (int) (tmp22 - tmp12);
+    wsptr[8*3] = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*6] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4] = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 10 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 10; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG) wsptr[4];
+    z1 = MULTIPLY(z4, FIX(1.144122806));         /* c4 */
+    z2 = MULTIPLY(z4, FIX(0.437016024));         /* c8 */
+    tmp10 = z3 + z1;
+    tmp11 = z3 - z2;
+
+    tmp22 = z3 - LEFT_SHIFT(z1 - z2, 1);         /* c0 = (c4-c8)*2 */
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[6];
+
+    z1 = MULTIPLY(z2 + z3, FIX(0.831253876));    /* c6 */
+    tmp12 = z1 + MULTIPLY(z2, FIX(0.513743148)); /* c2-c6 */
+    tmp13 = z1 - MULTIPLY(z3, FIX(2.176250899)); /* c2+c6 */
+
+    tmp20 = tmp10 + tmp12;
+    tmp24 = tmp10 - tmp12;
+    tmp21 = tmp11 + tmp13;
+    tmp23 = tmp11 - tmp13;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z2 + z4;
+    tmp13 = z2 - z4;
+
+    tmp12 = MULTIPLY(tmp13, FIX(0.309016994));        /* (c3-c7)/2 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.951056516));           /* (c3+c7)/2 */
+    z4 = z3 + tmp12;
+
+    tmp10 = MULTIPLY(z1, FIX(1.396802247)) + z2 + z4; /* c1 */
+    tmp14 = MULTIPLY(z1, FIX(0.221231742)) - z2 + z4; /* c9 */
+
+    z2 = MULTIPLY(tmp11, FIX(0.587785252));           /* (c1-c9)/2 */
+    z4 = z3 - tmp12 - LEFT_SHIFT(tmp13, CONST_BITS - 1);
+
+    tmp12 = LEFT_SHIFT(z1 - tmp13, CONST_BITS) - z3;
+
+    tmp11 = MULTIPLY(z1, FIX(1.260073511)) - z2 - z4; /* c3 */
+    tmp13 = MULTIPLY(z1, FIX(0.642039522)) - z2 + z4; /* c7 */
+
+    /* Final output stage */
+
+    outptr[0] = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[9] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[1] = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[8] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[2] = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[7] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[3] = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[6] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[4] = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+    outptr[5] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                              CONST_BITS+PASS1_BITS+3)
+                            & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 11x11 output block.
+ *
+ * Optimized algorithm with 24 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/22).
+ */
+
+GLOBAL(void)
+jpeg_idct_11x11 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*11];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp10 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp10 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, - FIX(1.155664402));        /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, - FIX(1.798248910));       /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, - FIX(1.467221301)) +          /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 11 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 11; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp10 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp10 = LEFT_SHIFT(tmp10, CONST_BITS);
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[4];
+    z3 = (JLONG) wsptr[6];
+
+    tmp20 = MULTIPLY(z2 - z3, FIX(2.546640132));     /* c2+c4 */
+    tmp23 = MULTIPLY(z2 - z1, FIX(0.430815045));     /* c2-c6 */
+    z4 = z1 + z3;
+    tmp24 = MULTIPLY(z4, - FIX(1.155664402));        /* -(c2-c10) */
+    z4 -= z2;
+    tmp25 = tmp10 + MULTIPLY(z4, FIX(1.356927976));  /* c2 */
+    tmp21 = tmp20 + tmp23 + tmp25 -
+            MULTIPLY(z2, FIX(1.821790775));          /* c2+c4+c10-c6 */
+    tmp20 += tmp25 + MULTIPLY(z3, FIX(2.115825087)); /* c4+c6 */
+    tmp23 += tmp25 - MULTIPLY(z1, FIX(1.513598477)); /* c6+c8 */
+    tmp24 += tmp25;
+    tmp22 = tmp24 - MULTIPLY(z3, FIX(0.788749120));  /* c8+c10 */
+    tmp24 += MULTIPLY(z2, FIX(1.944413522)) -        /* c2+c8 */
+             MULTIPLY(z1, FIX(1.390975730));         /* c4+c10 */
+    tmp25 = tmp10 - MULTIPLY(z4, FIX(1.414213562));  /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z1 + z2;
+    tmp14 = MULTIPLY(tmp11 + z3 + z4, FIX(0.398430003)); /* c9 */
+    tmp11 = MULTIPLY(tmp11, FIX(0.887983902));           /* c3-c9 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(0.670361295));         /* c5-c9 */
+    tmp13 = tmp14 + MULTIPLY(z1 + z4, FIX(0.366151574)); /* c7-c9 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(0.923107866));              /* c7+c5+c3-c1-2*c9 */
+    z1    = tmp14 - MULTIPLY(z2 + z3, FIX(1.163011579)); /* c7+c9 */
+    tmp11 += z1 + MULTIPLY(z2, FIX(2.073276588));        /* c1+c7+3*c9-c3 */
+    tmp12 += z1 - MULTIPLY(z3, FIX(1.192193623));        /* c3+c5-c7-c9 */
+    z1    = MULTIPLY(z2 + z4, - FIX(1.798248910));       /* -(c1+c9) */
+    tmp11 += z1;
+    tmp13 += z1 + MULTIPLY(z4, FIX(2.102458632));        /* c1+c5+c9-c7 */
+    tmp14 += MULTIPLY(z2, - FIX(1.467221301)) +          /* -(c5+c9) */
+             MULTIPLY(z3, FIX(1.001388905)) -            /* c1-c9 */
+             MULTIPLY(z4, FIX(1.684843907));             /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 12x12 output block.
+ *
+ * Optimized algorithm with 15 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/24).
+ */
+
+GLOBAL(void)
+jpeg_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*12];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z3 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, - FIX_0_541196100);                 /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580));           /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 12 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 12; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z3 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z3 = LEFT_SHIFT(z3, CONST_BITS);
+
+    z4 = (JLONG) wsptr[4];
+    z4 = MULTIPLY(z4, FIX(1.224744871)); /* c4 */
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    z1 = (JLONG) wsptr[2];
+    z4 = MULTIPLY(z1, FIX(1.366025404)); /* c2 */
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z2 = (JLONG) wsptr[6];
+    z2 = LEFT_SHIFT(z2, CONST_BITS);
+
+    tmp12 = z1 - z2;
+
+    tmp21 = z3 + tmp12;
+    tmp24 = z3 - tmp12;
+
+    tmp12 = z4 + z2;
+
+    tmp20 = tmp10 + tmp12;
+    tmp25 = tmp10 - tmp12;
+
+    tmp12 = z4 - z1 - z2;
+
+    tmp22 = tmp11 + tmp12;
+    tmp23 = tmp11 - tmp12;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = MULTIPLY(z2, FIX(1.306562965));                  /* c3 */
+    tmp14 = MULTIPLY(z2, - FIX_0_541196100);                 /* -c9 */
+
+    tmp10 = z1 + z3;
+    tmp15 = MULTIPLY(tmp10 + z4, FIX(0.860918669));          /* c7 */
+    tmp12 = tmp15 + MULTIPLY(tmp10, FIX(0.261052384));       /* c5-c7 */
+    tmp10 = tmp12 + tmp11 + MULTIPLY(z1, FIX(0.280143716));  /* c1-c5 */
+    tmp13 = MULTIPLY(z3 + z4, - FIX(1.045510580));           /* -(c7+c11) */
+    tmp12 += tmp13 + tmp14 - MULTIPLY(z3, FIX(1.478575242)); /* c1+c5-c7-c11 */
+    tmp13 += tmp15 - tmp11 + MULTIPLY(z4, FIX(1.586706681)); /* c1+c11 */
+    tmp15 += tmp14 - MULTIPLY(z1, FIX(0.676326758)) -        /* c7-c11 */
+             MULTIPLY(z4, FIX(1.982889723));                 /* c5+c7 */
+
+    z1 -= z4;
+    z2 -= z3;
+    z3 = MULTIPLY(z1 + z2, FIX_0_541196100);                 /* c9 */
+    tmp11 = z3 + MULTIPLY(z1, FIX_0_765366865);              /* c3-c9 */
+    tmp14 = z3 - MULTIPLY(z2, FIX_1_847759065);              /* c3+c9 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 13x13 output block.
+ *
+ * Optimized algorithm with 29 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/26).
+ */
+
+GLOBAL(void)
+jpeg_idct_13x13 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*13];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */
+    tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458));   /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945));   /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813));   /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 13 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 13; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[4];
+    z4 = (JLONG) wsptr[6];
+
+    tmp10 = z3 + z4;
+    tmp11 = z3 - z4;
+
+    tmp12 = MULTIPLY(tmp10, FIX(1.155388986));                /* (c4+c6)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.096834934)) + z1;           /* (c4-c6)/2 */
+
+    tmp20 = MULTIPLY(z2, FIX(1.373119086)) + tmp12 + tmp13;   /* c2 */
+    tmp22 = MULTIPLY(z2, FIX(0.501487041)) - tmp12 + tmp13;   /* c10 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.316450131));                /* (c8-c12)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.486914739)) + z1;           /* (c8+c12)/2 */
+
+    tmp21 = MULTIPLY(z2, FIX(1.058554052)) - tmp12 + tmp13;   /* c6 */
+    tmp25 = MULTIPLY(z2, - FIX(1.252223920)) + tmp12 + tmp13; /* c4 */
+
+    tmp12 = MULTIPLY(tmp10, FIX(0.435816023));                /* (c2-c10)/2 */
+    tmp13 = MULTIPLY(tmp11, FIX(0.937303064)) - z1;           /* (c2+c10)/2 */
+
+    tmp23 = MULTIPLY(z2, - FIX(0.170464608)) - tmp12 - tmp13; /* c12 */
+    tmp24 = MULTIPLY(z2, - FIX(0.803364869)) + tmp12 - tmp13; /* c8 */
+
+    tmp26 = MULTIPLY(tmp11 - z2, FIX(1.414213562)) + z1;      /* c0 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.322312651));     /* c3 */
+    tmp12 = MULTIPLY(z1 + z3, FIX(1.163874945));     /* c5 */
+    tmp15 = z1 + z4;
+    tmp13 = MULTIPLY(tmp15, FIX(0.937797057));       /* c7 */
+    tmp10 = tmp11 + tmp12 + tmp13 -
+            MULTIPLY(z1, FIX(2.020082300));          /* c7+c5+c3-c1 */
+    tmp14 = MULTIPLY(z2 + z3, - FIX(0.338443458));   /* -c11 */
+    tmp11 += tmp14 + MULTIPLY(z2, FIX(0.837223564)); /* c5+c9+c11-c3 */
+    tmp12 += tmp14 - MULTIPLY(z3, FIX(1.572116027)); /* c1+c5-c9-c11 */
+    tmp14 = MULTIPLY(z2 + z4, - FIX(1.163874945));   /* -c5 */
+    tmp11 += tmp14;
+    tmp13 += tmp14 + MULTIPLY(z4, FIX(2.205608352)); /* c3+c5+c9-c7 */
+    tmp14 = MULTIPLY(z3 + z4, - FIX(0.657217813));   /* -c9 */
+    tmp12 += tmp14;
+    tmp13 += tmp14;
+    tmp15 = MULTIPLY(tmp15, FIX(0.338443458));       /* c11 */
+    tmp14 = tmp15 + MULTIPLY(z1, FIX(0.318774355)) - /* c9-c11 */
+            MULTIPLY(z2, FIX(0.466105296));          /* c1-c7 */
+    z1    = MULTIPLY(z3 - z2, FIX(0.937797057));     /* c7 */
+    tmp14 += z1;
+    tmp15 += z1 + MULTIPLY(z3, FIX(0.384515595)) -   /* c3-c7 */
+             MULTIPLY(z4, FIX(1.742345811));         /* c1+c11 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 14x14 output block.
+ *
+ * Optimized algorithm with 20 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/28).
+ */
+
+GLOBAL(void)
+jpeg_idct_14x14 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*14];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = RIGHT_SHIFT(z1 - LEFT_SHIFT(z2 + z3 - z4, 1),
+                        CONST_BITS-PASS1_BITS);  /* c0 = (c4+c12-c8)*2 */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+    tmp13 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + tmp13 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - tmp13;        /* c11 */
+    tmp16 += tmp15;
+    z1    += z4;
+    z4    = MULTIPLY(z2 + z3, - FIX(0.158341681)) - tmp13; /* -c13 */
+    tmp11 += z4 - MULTIPLY(z2, FIX(0.424103948));          /* c3-c9-c13 */
+    tmp12 += z4 - MULTIPLY(z3, FIX(2.373959773));          /* c3+c5-c13 */
+    z4    = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += z4 + tmp13 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += z4 + MULTIPLY(z2, FIX(0.674957567));          /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, PASS1_BITS);
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) (tmp23 + tmp13);
+    wsptr[8*10] = (int) (tmp23 - tmp13);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 14 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 14; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    z4 = (JLONG) wsptr[4];
+    z2 = MULTIPLY(z4, FIX(1.274162392));         /* c4 */
+    z3 = MULTIPLY(z4, FIX(0.314692123));         /* c12 */
+    z4 = MULTIPLY(z4, FIX(0.881747734));         /* c8 */
+
+    tmp10 = z1 + z2;
+    tmp11 = z1 + z3;
+    tmp12 = z1 - z4;
+
+    tmp23 = z1 - LEFT_SHIFT(z2 + z3 - z4, 1);    /* c0 = (c4+c12-c8)*2 */
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[6];
+
+    z3 = MULTIPLY(z1 + z2, FIX(1.105676686));    /* c6 */
+
+    tmp13 = z3 + MULTIPLY(z1, FIX(0.273079590)); /* c2-c6 */
+    tmp14 = z3 - MULTIPLY(z2, FIX(1.719280954)); /* c6+c10 */
+    tmp15 = MULTIPLY(z1, FIX(0.613604268)) -     /* c10 */
+            MULTIPLY(z2, FIX(1.378756276));      /* c2 */
+
+    tmp20 = tmp10 + tmp13;
+    tmp26 = tmp10 - tmp13;
+    tmp21 = tmp11 + tmp14;
+    tmp25 = tmp11 - tmp14;
+    tmp22 = tmp12 + tmp15;
+    tmp24 = tmp12 - tmp15;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+    z4 = LEFT_SHIFT(z4, CONST_BITS);
+
+    tmp14 = z1 + z3;
+    tmp11 = MULTIPLY(z1 + z2, FIX(1.334852607));           /* c3 */
+    tmp12 = MULTIPLY(tmp14, FIX(1.197448846));             /* c5 */
+    tmp10 = tmp11 + tmp12 + z4 - MULTIPLY(z1, FIX(1.126980169)); /* c3+c5-c1 */
+    tmp14 = MULTIPLY(tmp14, FIX(0.752406978));             /* c9 */
+    tmp16 = tmp14 - MULTIPLY(z1, FIX(1.061150426));        /* c9+c11-c13 */
+    z1    -= z2;
+    tmp15 = MULTIPLY(z1, FIX(0.467085129)) - z4;           /* c11 */
+    tmp16 += tmp15;
+    tmp13 = MULTIPLY(z2 + z3, - FIX(0.158341681)) - z4;    /* -c13 */
+    tmp11 += tmp13 - MULTIPLY(z2, FIX(0.424103948));       /* c3-c9-c13 */
+    tmp12 += tmp13 - MULTIPLY(z3, FIX(2.373959773));       /* c3+c5-c13 */
+    tmp13 = MULTIPLY(z3 - z2, FIX(1.405321284));           /* c1 */
+    tmp14 += tmp13 + z4 - MULTIPLY(z3, FIX(1.6906431334)); /* c1+c9-c11 */
+    tmp15 += tmp13 + MULTIPLY(z2, FIX(0.674957567));       /* c1+c11-c5 */
+
+    tmp13 = LEFT_SHIFT(z1 - z3, CONST_BITS) + z4;
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 15x15 output block.
+ *
+ * Optimized algorithm with 22 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/30).
+ */
+
+GLOBAL(void)
+jpeg_idct_15x15 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp10, tmp11, tmp12, tmp13, tmp14, tmp15, tmp16;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*15];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    z1 += ONE << (CONST_BITS-PASS1_BITS-1);
+
+    z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, - FIX(0.831253876));               /* -c9 */
+    tmp15 = MULTIPLY(z2, - FIX(1.344997024));               /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*14] = (int) RIGHT_SHIFT(tmp20 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp21 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp22 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp23 - tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp24 - tmp14, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp25 - tmp15, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp26 - tmp16, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp27, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 15 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 15; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    z1 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    z1 = LEFT_SHIFT(z1, CONST_BITS);
+
+    z2 = (JLONG) wsptr[2];
+    z3 = (JLONG) wsptr[4];
+    z4 = (JLONG) wsptr[6];
+
+    tmp10 = MULTIPLY(z4, FIX(0.437016024)); /* c12 */
+    tmp11 = MULTIPLY(z4, FIX(1.144122806)); /* c6 */
+
+    tmp12 = z1 - tmp10;
+    tmp13 = z1 + tmp11;
+    z1 -= LEFT_SHIFT(tmp11 - tmp10, 1);     /* c0 = (c6-c12)*2 */
+
+    z4 = z2 - z3;
+    z3 += z2;
+    tmp10 = MULTIPLY(z3, FIX(1.337628990)); /* (c2+c4)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.045680613)); /* (c2-c4)/2 */
+    z2 = MULTIPLY(z2, FIX(1.439773946));    /* c4+c14 */
+
+    tmp20 = tmp13 + tmp10 + tmp11;
+    tmp23 = tmp12 - tmp10 + tmp11 + z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.547059574)); /* (c8+c14)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.399234004)); /* (c8-c14)/2 */
+
+    tmp25 = tmp13 - tmp10 - tmp11;
+    tmp26 = tmp12 + tmp10 - tmp11 - z2;
+
+    tmp10 = MULTIPLY(z3, FIX(0.790569415)); /* (c6+c12)/2 */
+    tmp11 = MULTIPLY(z4, FIX(0.353553391)); /* (c6-c12)/2 */
+
+    tmp21 = tmp12 + tmp10 + tmp11;
+    tmp24 = tmp13 - tmp10 + tmp11;
+    tmp11 += tmp11;
+    tmp22 = z1 + tmp11;                     /* c10 = c6-c12 */
+    tmp27 = z1 - tmp11 - tmp11;             /* c0 = (c6-c12)*2 */
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z4 = (JLONG) wsptr[5];
+    z3 = MULTIPLY(z4, FIX(1.224744871));                    /* c5 */
+    z4 = (JLONG) wsptr[7];
+
+    tmp13 = z2 - z4;
+    tmp15 = MULTIPLY(z1 + tmp13, FIX(0.831253876));         /* c9 */
+    tmp11 = tmp15 + MULTIPLY(z1, FIX(0.513743148));         /* c3-c9 */
+    tmp14 = tmp15 - MULTIPLY(tmp13, FIX(2.176250899));      /* c3+c9 */
+
+    tmp13 = MULTIPLY(z2, - FIX(0.831253876));               /* -c9 */
+    tmp15 = MULTIPLY(z2, - FIX(1.344997024));               /* -c3 */
+    z2 = z1 - z4;
+    tmp12 = z3 + MULTIPLY(z2, FIX(1.406466353));            /* c1 */
+
+    tmp10 = tmp12 + MULTIPLY(z4, FIX(2.457431844)) - tmp15; /* c1+c7 */
+    tmp16 = tmp12 - MULTIPLY(z1, FIX(1.112434820)) + tmp13; /* c1-c13 */
+    tmp12 = MULTIPLY(z2, FIX(1.224744871)) - z3;            /* c5 */
+    z2 = MULTIPLY(z1 + z4, FIX(0.575212477));               /* c11 */
+    tmp13 += z2 + MULTIPLY(z1, FIX(0.475753014)) - z3;      /* c7-c11 */
+    tmp15 += z2 - MULTIPLY(z4, FIX(0.869244010)) + z3;      /* c11+c13 */
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp14,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp15,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp16,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp27,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
+  }
+}
+
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients,
+ * producing a 16x16 output block.
+ *
+ * Optimized algorithm with 28 multiplications in the 1-D kernel.
+ * cK represents sqrt(2) * cos(K*pi/32).
+ */
+
+GLOBAL(void)
+jpeg_idct_16x16 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                 JCOEFPTR coef_block,
+                 JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  JLONG tmp0, tmp1, tmp2, tmp3, tmp10, tmp11, tmp12, tmp13;
+  JLONG tmp20, tmp21, tmp22, tmp23, tmp24, tmp25, tmp26, tmp27;
+  JLONG z1, z2, z3, z4;
+  JCOEFPTR inptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
+  JSAMPROW outptr;
+  JSAMPLE *range_limit = IDCT_range_limit(cinfo);
+  int ctr;
+  int workspace[8*16];  /* buffers data between passes */
+  SHIFT_TEMPS
+
+  /* Pass 1: process columns from input, store into work array. */
+
+  inptr = coef_block;
+  quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
+  wsptr = workspace;
+  for (ctr = 0; ctr < 8; ctr++, inptr++, quantptr++, wsptr++) {
+    /* Even part */
+
+    tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+    /* Add fudge factor here for final descale. */
+    tmp0 += 1 << (CONST_BITS-PASS1_BITS-1);
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]);
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c12[16] = c6[8] */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c14[16] = c7[8] */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c2[16] = c1[8] */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)[16] = (c3+c1)[8] */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)[16] = (c3-c7)[8] */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
+    z2 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
+    z3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
+    z4 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, - FIX(0.666655658));      /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, - FIX(1.247225013));      /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    wsptr[8*0]  = (int) RIGHT_SHIFT(tmp20 + tmp0,  CONST_BITS-PASS1_BITS);
+    wsptr[8*15] = (int) RIGHT_SHIFT(tmp20 - tmp0,  CONST_BITS-PASS1_BITS);
+    wsptr[8*1]  = (int) RIGHT_SHIFT(tmp21 + tmp1,  CONST_BITS-PASS1_BITS);
+    wsptr[8*14] = (int) RIGHT_SHIFT(tmp21 - tmp1,  CONST_BITS-PASS1_BITS);
+    wsptr[8*2]  = (int) RIGHT_SHIFT(tmp22 + tmp2,  CONST_BITS-PASS1_BITS);
+    wsptr[8*13] = (int) RIGHT_SHIFT(tmp22 - tmp2,  CONST_BITS-PASS1_BITS);
+    wsptr[8*3]  = (int) RIGHT_SHIFT(tmp23 + tmp3,  CONST_BITS-PASS1_BITS);
+    wsptr[8*12] = (int) RIGHT_SHIFT(tmp23 - tmp3,  CONST_BITS-PASS1_BITS);
+    wsptr[8*4]  = (int) RIGHT_SHIFT(tmp24 + tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*11] = (int) RIGHT_SHIFT(tmp24 - tmp10, CONST_BITS-PASS1_BITS);
+    wsptr[8*5]  = (int) RIGHT_SHIFT(tmp25 + tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*10] = (int) RIGHT_SHIFT(tmp25 - tmp11, CONST_BITS-PASS1_BITS);
+    wsptr[8*6]  = (int) RIGHT_SHIFT(tmp26 + tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*9]  = (int) RIGHT_SHIFT(tmp26 - tmp12, CONST_BITS-PASS1_BITS);
+    wsptr[8*7]  = (int) RIGHT_SHIFT(tmp27 + tmp13, CONST_BITS-PASS1_BITS);
+    wsptr[8*8]  = (int) RIGHT_SHIFT(tmp27 - tmp13, CONST_BITS-PASS1_BITS);
+  }
+
+  /* Pass 2: process 16 rows from work array, store into output array. */
+
+  wsptr = workspace;
+  for (ctr = 0; ctr < 16; ctr++) {
+    outptr = output_buf[ctr] + output_col;
+
+    /* Even part */
+
+    /* Add fudge factor here for final descale. */
+    tmp0 = (JLONG) wsptr[0] + (ONE << (PASS1_BITS+2));
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS);
+
+    z1 = (JLONG) wsptr[4];
+    tmp1 = MULTIPLY(z1, FIX(1.306562965));      /* c4[16] = c2[8] */
+    tmp2 = MULTIPLY(z1, FIX_0_541196100);       /* c12[16] = c6[8] */
+
+    tmp10 = tmp0 + tmp1;
+    tmp11 = tmp0 - tmp1;
+    tmp12 = tmp0 + tmp2;
+    tmp13 = tmp0 - tmp2;
+
+    z1 = (JLONG) wsptr[2];
+    z2 = (JLONG) wsptr[6];
+    z3 = z1 - z2;
+    z4 = MULTIPLY(z3, FIX(0.275899379));        /* c14[16] = c7[8] */
+    z3 = MULTIPLY(z3, FIX(1.387039845));        /* c2[16] = c1[8] */
+
+    tmp0 = z3 + MULTIPLY(z2, FIX_2_562915447);  /* (c6+c2)[16] = (c3+c1)[8] */
+    tmp1 = z4 + MULTIPLY(z1, FIX_0_899976223);  /* (c6-c14)[16] = (c3-c7)[8] */
+    tmp2 = z3 - MULTIPLY(z1, FIX(0.601344887)); /* (c2-c10)[16] = (c1-c5)[8] */
+    tmp3 = z4 - MULTIPLY(z2, FIX(0.509795579)); /* (c10-c14)[16] = (c5-c7)[8] */
+
+    tmp20 = tmp10 + tmp0;
+    tmp27 = tmp10 - tmp0;
+    tmp21 = tmp12 + tmp1;
+    tmp26 = tmp12 - tmp1;
+    tmp22 = tmp13 + tmp2;
+    tmp25 = tmp13 - tmp2;
+    tmp23 = tmp11 + tmp3;
+    tmp24 = tmp11 - tmp3;
+
+    /* Odd part */
+
+    z1 = (JLONG) wsptr[1];
+    z2 = (JLONG) wsptr[3];
+    z3 = (JLONG) wsptr[5];
+    z4 = (JLONG) wsptr[7];
+
+    tmp11 = z1 + z3;
+
+    tmp1  = MULTIPLY(z1 + z2, FIX(1.353318001));   /* c3 */
+    tmp2  = MULTIPLY(tmp11,   FIX(1.247225013));   /* c5 */
+    tmp3  = MULTIPLY(z1 + z4, FIX(1.093201867));   /* c7 */
+    tmp10 = MULTIPLY(z1 - z4, FIX(0.897167586));   /* c9 */
+    tmp11 = MULTIPLY(tmp11,   FIX(0.666655658));   /* c11 */
+    tmp12 = MULTIPLY(z1 - z2, FIX(0.410524528));   /* c13 */
+    tmp0  = tmp1 + tmp2 + tmp3 -
+            MULTIPLY(z1, FIX(2.286341144));        /* c7+c5+c3-c1 */
+    tmp13 = tmp10 + tmp11 + tmp12 -
+            MULTIPLY(z1, FIX(1.835730603));        /* c9+c11+c13-c15 */
+    z1    = MULTIPLY(z2 + z3, FIX(0.138617169));   /* c15 */
+    tmp1  += z1 + MULTIPLY(z2, FIX(0.071888074));  /* c9+c11-c3-c15 */
+    tmp2  += z1 - MULTIPLY(z3, FIX(1.125726048));  /* c5+c7+c15-c3 */
+    z1    = MULTIPLY(z3 - z2, FIX(1.407403738));   /* c1 */
+    tmp11 += z1 - MULTIPLY(z3, FIX(0.766367282));  /* c1+c11-c9-c13 */
+    tmp12 += z1 + MULTIPLY(z2, FIX(1.971951411));  /* c1+c5+c13-c7 */
+    z2    += z4;
+    z1    = MULTIPLY(z2, - FIX(0.666655658));      /* -c11 */
+    tmp1  += z1;
+    tmp3  += z1 + MULTIPLY(z4, FIX(1.065388962));  /* c3+c11+c15-c7 */
+    z2    = MULTIPLY(z2, - FIX(1.247225013));      /* -c5 */
+    tmp10 += z2 + MULTIPLY(z4, FIX(3.141271809));  /* c1+c5+c9-c13 */
+    tmp12 += z2;
+    z2    = MULTIPLY(z3 + z4, - FIX(1.353318001)); /* -c3 */
+    tmp2  += z2;
+    tmp3  += z2;
+    z2    = MULTIPLY(z4 - z3, FIX(0.410524528));   /* c13 */
+    tmp10 += z2;
+    tmp11 += z2;
+
+    /* Final output stage */
+
+    outptr[0]  = range_limit[(int) RIGHT_SHIFT(tmp20 + tmp0,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[15] = range_limit[(int) RIGHT_SHIFT(tmp20 - tmp0,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[1]  = range_limit[(int) RIGHT_SHIFT(tmp21 + tmp1,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[14] = range_limit[(int) RIGHT_SHIFT(tmp21 - tmp1,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[2]  = range_limit[(int) RIGHT_SHIFT(tmp22 + tmp2,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[13] = range_limit[(int) RIGHT_SHIFT(tmp22 - tmp2,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[3]  = range_limit[(int) RIGHT_SHIFT(tmp23 + tmp3,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[12] = range_limit[(int) RIGHT_SHIFT(tmp23 - tmp3,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[4]  = range_limit[(int) RIGHT_SHIFT(tmp24 + tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[11] = range_limit[(int) RIGHT_SHIFT(tmp24 - tmp10,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[5]  = range_limit[(int) RIGHT_SHIFT(tmp25 + tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[10] = range_limit[(int) RIGHT_SHIFT(tmp25 - tmp11,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[6]  = range_limit[(int) RIGHT_SHIFT(tmp26 + tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[9]  = range_limit[(int) RIGHT_SHIFT(tmp26 - tmp12,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[7]  = range_limit[(int) RIGHT_SHIFT(tmp27 + tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+    outptr[8]  = range_limit[(int) RIGHT_SHIFT(tmp27 - tmp13,
+                                               CONST_BITS+PASS1_BITS+3)
+                             & RANGE_MASK];
+
+    wsptr += 8;         /* advance pointer to next row */
   }
 }
 
+#endif /* IDCT_SCALING_SUPPORTED */
 #endif /* DCT_ISLOW_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctred.c glmark2-2021.02/src/libjpeg-turbo/jidctred.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jidctred.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jidctred.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jidctred.c
  *
+ * This file was part of the Independent JPEG Group's software.
  * Copyright (C) 1994-1998, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains inverse-DCT routines that produce reduced-size output:
  * either 4x4, 2x2, or 1x1 pixels from an 8x8 DCT block.
@@ -23,7 +26,7 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jdct.h"		/* Private declarations for DCT subsystem */
+#include "jdct.h"               /* Private declarations for DCT subsystem */
 
 #ifdef IDCT_SCALING_SUPPORTED
 
@@ -44,7 +47,7 @@
 #define PASS1_BITS  2
 #else
 #define CONST_BITS  13
-#define PASS1_BITS  1		/* lose a little precision to avoid overflow */
+#define PASS1_BITS  1           /* lose a little precision to avoid overflow */
 #endif
 
 /* Some C compilers fail to reduce "FIX(constant)" at compile time, thus
@@ -55,20 +58,20 @@
  */
 
 #if CONST_BITS == 13
-#define FIX_0_211164243  ((INT32)  1730)	/* FIX(0.211164243) */
-#define FIX_0_509795579  ((INT32)  4176)	/* FIX(0.509795579) */
-#define FIX_0_601344887  ((INT32)  4926)	/* FIX(0.601344887) */
-#define FIX_0_720959822  ((INT32)  5906)	/* FIX(0.720959822) */
-#define FIX_0_765366865  ((INT32)  6270)	/* FIX(0.765366865) */
-#define FIX_0_850430095  ((INT32)  6967)	/* FIX(0.850430095) */
-#define FIX_0_899976223  ((INT32)  7373)	/* FIX(0.899976223) */
-#define FIX_1_061594337  ((INT32)  8697)	/* FIX(1.061594337) */
-#define FIX_1_272758580  ((INT32)  10426)	/* FIX(1.272758580) */
-#define FIX_1_451774981  ((INT32)  11893)	/* FIX(1.451774981) */
-#define FIX_1_847759065  ((INT32)  15137)	/* FIX(1.847759065) */
-#define FIX_2_172734803  ((INT32)  17799)	/* FIX(2.172734803) */
-#define FIX_2_562915447  ((INT32)  20995)	/* FIX(2.562915447) */
-#define FIX_3_624509785  ((INT32)  29692)	/* FIX(3.624509785) */
+#define FIX_0_211164243  ((JLONG)  1730)        /* FIX(0.211164243) */
+#define FIX_0_509795579  ((JLONG)  4176)        /* FIX(0.509795579) */
+#define FIX_0_601344887  ((JLONG)  4926)        /* FIX(0.601344887) */
+#define FIX_0_720959822  ((JLONG)  5906)        /* FIX(0.720959822) */
+#define FIX_0_765366865  ((JLONG)  6270)        /* FIX(0.765366865) */
+#define FIX_0_850430095  ((JLONG)  6967)        /* FIX(0.850430095) */
+#define FIX_0_899976223  ((JLONG)  7373)        /* FIX(0.899976223) */
+#define FIX_1_061594337  ((JLONG)  8697)        /* FIX(1.061594337) */
+#define FIX_1_272758580  ((JLONG)  10426)       /* FIX(1.272758580) */
+#define FIX_1_451774981  ((JLONG)  11893)       /* FIX(1.451774981) */
+#define FIX_1_847759065  ((JLONG)  15137)       /* FIX(1.847759065) */
+#define FIX_2_172734803  ((JLONG)  17799)       /* FIX(2.172734803) */
+#define FIX_2_562915447  ((JLONG)  20995)       /* FIX(2.562915447) */
+#define FIX_3_624509785  ((JLONG)  29692)       /* FIX(3.624509785) */
 #else
 #define FIX_0_211164243  FIX(0.211164243)
 #define FIX_0_509795579  FIX(0.509795579)
@@ -87,7 +90,7 @@
 #endif
 
 
-/* Multiply an INT32 variable by an INT32 constant to yield an INT32 result.
+/* Multiply a JLONG variable by a JLONG constant to yield a JLONG result.
  * For 8-bit samples with the recommended scaling, all the variable
  * and constant values involved are no more than 16 bits wide, so a
  * 16x16->32 bit multiply can be used instead of a full 32x32 multiply.
@@ -115,19 +118,19 @@
  */
 
 GLOBAL(void)
-jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp2, tmp10, tmp12;
-  INT32 z1, z2, z3, z4;
+  JLONG tmp0, tmp2, tmp10, tmp12;
+  JLONG z1, z2, z3, z4;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE*4];	/* buffers data between passes */
+  int workspace[DCTSIZE*4];     /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -140,57 +143,58 @@
     if (ctr == DCTSIZE-4)
       continue;
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*2] == 0 &&
-	inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
-	inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*3] == 0 && inptr[DCTSIZE*5] == 0 &&
+        inptr[DCTSIZE*6] == 0 && inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero; we need not examine term 4 for 4x4 output */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
       wsptr[DCTSIZE*2] = dcval;
       wsptr[DCTSIZE*3] = dcval;
-      
+
       continue;
     }
-    
+
     /* Even part */
-    
+
     tmp0 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp0 <<= (CONST_BITS+1);
-    
+    tmp0 = LEFT_SHIFT(tmp0, CONST_BITS+1);
+
     z2 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]);
 
     tmp2 = MULTIPLY(z2, FIX_1_847759065) + MULTIPLY(z3, - FIX_0_765366865);
-    
+
     tmp10 = tmp0 + tmp2;
     tmp12 = tmp0 - tmp2;
-    
+
     /* Odd part */
-    
+
     z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
     z2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]);
     z3 = DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]);
     z4 = DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]);
-    
+
     tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
-	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
-	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
-	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
-    
+         + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+         + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+         + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+
     tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
-	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
-	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
-	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+         + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+         + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+         + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
 
     /* Final output stage */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp2, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*3] = (int) DESCALE(tmp10 - tmp2, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp12 + tmp0, CONST_BITS-PASS1_BITS+1);
     wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 - tmp0, CONST_BITS-PASS1_BITS+1);
   }
-  
+
   /* Pass 2: process 4 rows from work array, store into output array. */
 
   wsptr = workspace;
@@ -200,64 +204,64 @@
 
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[2] == 0 && wsptr[3] == 0 &&
-	wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
+        wsptr[5] == 0 && wsptr[6] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
       outptr[2] = dcval;
       outptr[3] = dcval;
-      
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
-    
-    tmp0 = ((INT32) wsptr[0]) << (CONST_BITS+1);
-    
-    tmp2 = MULTIPLY((INT32) wsptr[2], FIX_1_847759065)
-	 + MULTIPLY((INT32) wsptr[6], - FIX_0_765366865);
-    
+
+    tmp0 = LEFT_SHIFT((JLONG) wsptr[0], CONST_BITS+1);
+
+    tmp2 = MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+         + MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865);
+
     tmp10 = tmp0 + tmp2;
     tmp12 = tmp0 - tmp2;
-    
+
     /* Odd part */
-    
-    z1 = (INT32) wsptr[7];
-    z2 = (INT32) wsptr[5];
-    z3 = (INT32) wsptr[3];
-    z4 = (INT32) wsptr[1];
-    
+
+    z1 = (JLONG) wsptr[7];
+    z2 = (JLONG) wsptr[5];
+    z3 = (JLONG) wsptr[3];
+    z4 = (JLONG) wsptr[1];
+
     tmp0 = MULTIPLY(z1, - FIX_0_211164243) /* sqrt(2) * (c3-c1) */
-	 + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
-	 + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
-	 + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
-    
+         + MULTIPLY(z2, FIX_1_451774981) /* sqrt(2) * (c3+c7) */
+         + MULTIPLY(z3, - FIX_2_172734803) /* sqrt(2) * (-c1-c5) */
+         + MULTIPLY(z4, FIX_1_061594337); /* sqrt(2) * (c5+c7) */
+
     tmp2 = MULTIPLY(z1, - FIX_0_509795579) /* sqrt(2) * (c7-c5) */
-	 + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
-	 + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
-	 + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
+         + MULTIPLY(z2, - FIX_0_601344887) /* sqrt(2) * (c5-c1) */
+         + MULTIPLY(z3, FIX_0_899976223) /* sqrt(2) * (c3-c7) */
+         + MULTIPLY(z4, FIX_2_562915447); /* sqrt(2) * (c1+c3) */
 
     /* Final output stage */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp2,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[3] = range_limit[(int) DESCALE(tmp10 - tmp2,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp12 + tmp0,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
     outptr[2] = range_limit[(int) DESCALE(tmp12 - tmp0,
-					  CONST_BITS+PASS1_BITS+3+1)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3+1)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
@@ -268,18 +272,18 @@
  */
 
 GLOBAL(void)
-jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
-  INT32 tmp0, tmp10, z1;
+  JLONG tmp0, tmp10, z1;
   JCOEFPTR inptr;
-  ISLOW_MULT_TYPE * quantptr;
-  int * wsptr;
+  ISLOW_MULT_TYPE *quantptr;
+  int *wsptr;
   JSAMPROW outptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   int ctr;
-  int workspace[DCTSIZE*2];	/* buffers data between passes */
+  int workspace[DCTSIZE*2];     /* buffers data between passes */
   SHIFT_TEMPS
 
   /* Pass 1: process columns from input, store into work array. */
@@ -292,21 +296,22 @@
     if (ctr == DCTSIZE-2 || ctr == DCTSIZE-4 || ctr == DCTSIZE-6)
       continue;
     if (inptr[DCTSIZE*1] == 0 && inptr[DCTSIZE*3] == 0 &&
-	inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
+        inptr[DCTSIZE*5] == 0 && inptr[DCTSIZE*7] == 0) {
       /* AC terms all zero; we need not examine terms 2,4,6 for 2x2 output */
-      int dcval = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) << PASS1_BITS;
-      
+      int dcval = LEFT_SHIFT(DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]),
+                             PASS1_BITS);
+
       wsptr[DCTSIZE*0] = dcval;
       wsptr[DCTSIZE*1] = dcval;
-      
+
       continue;
     }
-    
+
     /* Even part */
-    
+
     z1 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]);
-    tmp10 = z1 << (CONST_BITS+2);
-    
+    tmp10 = LEFT_SHIFT(z1, CONST_BITS+2);
+
     /* Odd part */
 
     z1 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]);
@@ -319,11 +324,11 @@
     tmp0 += MULTIPLY(z1, FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
 
     /* Final output stage */
-    
+
     wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp0, CONST_BITS-PASS1_BITS+2);
     wsptr[DCTSIZE*1] = (int) DESCALE(tmp10 - tmp0, CONST_BITS-PASS1_BITS+2);
   }
-  
+
   /* Pass 2: process 2 rows from work array, store into output array. */
 
   wsptr = workspace;
@@ -334,38 +339,38 @@
 #ifndef NO_ZERO_ROW_TEST
     if (wsptr[1] == 0 && wsptr[3] == 0 && wsptr[5] == 0 && wsptr[7] == 0) {
       /* AC terms all zero */
-      JSAMPLE dcval = range_limit[(int) DESCALE((INT32) wsptr[0], PASS1_BITS+3)
-				  & RANGE_MASK];
-      
+      JSAMPLE dcval = range_limit[(int) DESCALE((JLONG) wsptr[0], PASS1_BITS+3)
+                                  & RANGE_MASK];
+
       outptr[0] = dcval;
       outptr[1] = dcval;
-      
-      wsptr += DCTSIZE;		/* advance pointer to next row */
+
+      wsptr += DCTSIZE;         /* advance pointer to next row */
       continue;
     }
 #endif
-    
+
     /* Even part */
-    
-    tmp10 = ((INT32) wsptr[0]) << (CONST_BITS+2);
-    
+
+    tmp10 = LEFT_SHIFT((JLONG) wsptr[0], CONST_BITS+2);
+
     /* Odd part */
 
-    tmp0 = MULTIPLY((INT32) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
-	 + MULTIPLY((INT32) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
-	 + MULTIPLY((INT32) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
-	 + MULTIPLY((INT32) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
+    tmp0 = MULTIPLY((JLONG) wsptr[7], - FIX_0_720959822) /* sqrt(2) * (c7-c5+c3-c1) */
+         + MULTIPLY((JLONG) wsptr[5], FIX_0_850430095) /* sqrt(2) * (-c1+c3+c5+c7) */
+         + MULTIPLY((JLONG) wsptr[3], - FIX_1_272758580) /* sqrt(2) * (-c1+c3-c5-c7) */
+         + MULTIPLY((JLONG) wsptr[1], FIX_3_624509785); /* sqrt(2) * (c1+c3+c5+c7) */
 
     /* Final output stage */
-    
+
     outptr[0] = range_limit[(int) DESCALE(tmp10 + tmp0,
-					  CONST_BITS+PASS1_BITS+3+2)
-			    & RANGE_MASK];
+                                          CONST_BITS+PASS1_BITS+3+2)
+                            & RANGE_MASK];
     outptr[1] = range_limit[(int) DESCALE(tmp10 - tmp0,
-					  CONST_BITS+PASS1_BITS+3+2)
-			    & RANGE_MASK];
-    
-    wsptr += DCTSIZE;		/* advance pointer to next row */
+                                          CONST_BITS+PASS1_BITS+3+2)
+                            & RANGE_MASK];
+
+    wsptr += DCTSIZE;           /* advance pointer to next row */
   }
 }
 
@@ -376,12 +381,12 @@
  */
 
 GLOBAL(void)
-jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-	       JCOEFPTR coef_block,
-	       JSAMPARRAY output_buf, JDIMENSION output_col)
+jpeg_idct_1x1 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+               JCOEFPTR coef_block,
+               JSAMPARRAY output_buf, JDIMENSION output_col)
 {
   int dcval;
-  ISLOW_MULT_TYPE * quantptr;
+  ISLOW_MULT_TYPE *quantptr;
   JSAMPLE *range_limit = IDCT_range_limit(cinfo);
   SHIFT_TEMPS
 
@@ -390,7 +395,7 @@
    */
   quantptr = (ISLOW_MULT_TYPE *) compptr->dct_table;
   dcval = DEQUANTIZE(coef_block[0], quantptr[0]);
-  dcval = (int) DESCALE((INT32) dcval, 3);
+  dcval = (int) DESCALE((JLONG) dcval, 3);
 
   output_buf[0][output_col] = range_limit[dcval & RANGE_MASK];
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jinclude.h glmark2-2021.02/src/libjpeg-turbo/jinclude.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jinclude.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jinclude.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jinclude.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1994, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file exists to provide a single place to fix any problems with
  * including the wrong system include files.  (Common problems are taken
@@ -17,8 +20,8 @@
 
 /* Include auto-config file to find out which system include files we need. */
 
-#include "jconfig.h"		/* auto configuration options */
-#define JCONFIG_INCLUDED	/* so that jpeglib.h doesn't do it again */
+#include "jconfig.h"            /* auto configuration options */
+#define JCONFIG_INCLUDED        /* so that jpeglib.h doesn't do it again */
 
 /*
  * We need the NULL macro and size_t typedef.
@@ -58,28 +61,18 @@
 #ifdef NEED_BSD_STRINGS
 
 #include <strings.h>
-#define MEMZERO(target,size)	bzero((void *)(target), (size_t)(size))
-#define MEMCOPY(dest,src,size)	bcopy((const void *)(src), (void *)(dest), (size_t)(size))
+#define MEMZERO(target,size)    bzero((void *)(target), (size_t)(size))
+#define MEMCOPY(dest,src,size)  bcopy((const void *)(src), (void *)(dest), (size_t)(size))
 
 #else /* not BSD, assume ANSI/SysV string lib */
 
 #include <string.h>
-#define MEMZERO(target,size)	memset((void *)(target), 0, (size_t)(size))
-#define MEMCOPY(dest,src,size)	memcpy((void *)(dest), (const void *)(src), (size_t)(size))
+#define MEMZERO(target,size)    memset((void *)(target), 0, (size_t)(size))
+#define MEMCOPY(dest,src,size)  memcpy((void *)(dest), (const void *)(src), (size_t)(size))
 
 #endif
 
 /*
- * In ANSI C, and indeed any rational implementation, size_t is also the
- * type returned by sizeof().  However, it seems there are some irrational
- * implementations out there, in which sizeof() returns an int even though
- * size_t is defined as long or unsigned long.  To ensure consistent results
- * we always use this SIZEOF() macro in place of using sizeof() directly.
- */
-
-#define SIZEOF(object)	((size_t) sizeof(object))
-
-/*
  * The modules that use fread() and fwrite() always invoke them through
  * these macros.  On some systems you may need to twiddle the argument casts.
  * CAUTION: argument order is different from underlying functions!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemmgr.c glmark2-2021.02/src/libjpeg-turbo/jmemmgr.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemmgr.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jmemmgr.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemmgr.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains the JPEG system-independent memory management
  * routines.  This code is usable across a wide variety of machines; most
@@ -25,14 +28,14 @@
  */
 
 #define JPEG_INTERNALS
-#define AM_MEMORY_MANAGER	/* we define jvirt_Xarray_control structs */
+#define AM_MEMORY_MANAGER       /* we define jvirt_Xarray_control structs */
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jmemsys.h"		/* import the system-dependent declarations */
+#include "jmemsys.h"            /* import the system-dependent declarations */
 
 #ifndef NO_GETENV
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare getenv() */
-extern char * getenv JPP((const char * name));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare getenv() */
+extern char *getenv (const char *name);
 #endif
 #endif
 
@@ -67,9 +70,9 @@
  * There isn't any really portable way to determine the worst-case alignment
  * requirement.  This module assumes that the alignment requirement is
  * multiples of ALIGN_SIZE.
- * By default, we define ALIGN_SIZE as sizeof(double).  This is necessary on some
- * workstations (where doubles really do need 8-byte alignment) and will work
- * fine on nearly everything.  If your machine has lesser alignment needs,
+ * By default, we define ALIGN_SIZE as sizeof(double).  This is necessary on
+ * some workstations (where doubles really do need 8-byte alignment) and will
+ * work fine on nearly everything.  If your machine has lesser alignment needs,
  * you can save a few bytes by making ALIGN_SIZE smaller.
  * The only place I know of where this will NOT work is certain Macintosh
  * 680x0 compilers that define double as a 10-byte IEEE extended float.
@@ -78,9 +81,9 @@
  * such a compiler.
  */
 
-#ifndef ALIGN_SIZE		/* so can override from jconfig.h */
+#ifndef ALIGN_SIZE              /* so can override from jconfig.h */
 #ifndef WITH_SIMD
-#define ALIGN_SIZE  SIZEOF(double)
+#define ALIGN_SIZE  sizeof(double)
 #else
 #define ALIGN_SIZE  16 /* Most SIMD implementations require this */
 #endif
@@ -91,24 +94,23 @@
  * request to jpeg_get_small() or jpeg_get_large().  There is no per-object
  * overhead within a pool, except for alignment padding.  Each pool has a
  * header with a link to the next pool of the same class.
- * Small and large pool headers are identical except that the latter's
- * link pointer must be FAR on 80x86 machines.
+ * Small and large pool headers are identical.
  */
 
-typedef struct small_pool_struct * small_pool_ptr;
+typedef struct small_pool_struct *small_pool_ptr;
 
 typedef struct small_pool_struct {
-  small_pool_ptr next;	/* next in list of pools */
-  size_t bytes_used;		/* how many bytes already used within pool */
-  size_t bytes_left;		/* bytes still available in this pool */
+  small_pool_ptr next;  /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
 } small_pool_hdr;
 
-typedef struct large_pool_struct FAR * large_pool_ptr;
+typedef struct large_pool_struct *large_pool_ptr;
 
 typedef struct large_pool_struct {
-  large_pool_ptr next;	/* next in list of pools */
-  size_t bytes_used;		/* how many bytes already used within pool */
-  size_t bytes_left;		/* bytes still available in this pool */
+  large_pool_ptr next;  /* next in list of pools */
+  size_t bytes_used;            /* how many bytes already used within pool */
+  size_t bytes_left;            /* bytes still available in this pool */
 } large_pool_hdr;
 
 /*
@@ -116,7 +118,7 @@
  */
 
 typedef struct {
-  struct jpeg_memory_mgr pub;	/* public fields */
+  struct jpeg_memory_mgr pub;   /* public fields */
 
   /* Each pool identifier (lifetime class) names a linked list of pools. */
   small_pool_ptr small_list[JPOOL_NUMPOOLS];
@@ -136,10 +138,10 @@
   /* alloc_sarray and alloc_barray set this value for use by virtual
    * array routines.
    */
-  JDIMENSION last_rowsperchunk;	/* from most recent alloc_sarray/barray */
+  JDIMENSION last_rowsperchunk; /* from most recent alloc_sarray/barray */
 } my_memory_mgr;
 
-typedef my_memory_mgr * my_mem_ptr;
+typedef my_memory_mgr *my_mem_ptr;
 
 
 /*
@@ -150,39 +152,39 @@
  */
 
 struct jvirt_sarray_control {
-  JSAMPARRAY mem_buffer;	/* => the in-memory buffer */
-  JDIMENSION rows_in_array;	/* total virtual array height */
-  JDIMENSION samplesperrow;	/* width of array (and of memory buffer) */
-  JDIMENSION maxaccess;		/* max rows accessed by access_virt_sarray */
-  JDIMENSION rows_in_mem;	/* height of memory buffer */
-  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
-  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
-  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
-  boolean pre_zero;		/* pre-zero mode requested? */
-  boolean dirty;		/* do current buffer contents need written? */
-  boolean b_s_open;		/* is backing-store data valid? */
-  jvirt_sarray_ptr next;	/* link to next virtual sarray control block */
-  backing_store_info b_s_info;	/* System-dependent control info */
+  JSAMPARRAY mem_buffer;        /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION samplesperrow;     /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_sarray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_sarray_ptr next;        /* link to next virtual sarray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
 };
 
 struct jvirt_barray_control {
-  JBLOCKARRAY mem_buffer;	/* => the in-memory buffer */
-  JDIMENSION rows_in_array;	/* total virtual array height */
-  JDIMENSION blocksperrow;	/* width of array (and of memory buffer) */
-  JDIMENSION maxaccess;		/* max rows accessed by access_virt_barray */
-  JDIMENSION rows_in_mem;	/* height of memory buffer */
-  JDIMENSION rowsperchunk;	/* allocation chunk size in mem_buffer */
-  JDIMENSION cur_start_row;	/* first logical row # in the buffer */
-  JDIMENSION first_undef_row;	/* row # of first uninitialized row */
-  boolean pre_zero;		/* pre-zero mode requested? */
-  boolean dirty;		/* do current buffer contents need written? */
-  boolean b_s_open;		/* is backing-store data valid? */
-  jvirt_barray_ptr next;	/* link to next virtual barray control block */
-  backing_store_info b_s_info;	/* System-dependent control info */
+  JBLOCKARRAY mem_buffer;       /* => the in-memory buffer */
+  JDIMENSION rows_in_array;     /* total virtual array height */
+  JDIMENSION blocksperrow;      /* width of array (and of memory buffer) */
+  JDIMENSION maxaccess;         /* max rows accessed by access_virt_barray */
+  JDIMENSION rows_in_mem;       /* height of memory buffer */
+  JDIMENSION rowsperchunk;      /* allocation chunk size in mem_buffer */
+  JDIMENSION cur_start_row;     /* first logical row # in the buffer */
+  JDIMENSION first_undef_row;   /* row # of first uninitialized row */
+  boolean pre_zero;             /* pre-zero mode requested? */
+  boolean dirty;                /* do current buffer contents need written? */
+  boolean b_s_open;             /* is backing-store data valid? */
+  jvirt_barray_ptr next;        /* link to next virtual barray control block */
+  backing_store_info b_s_info;  /* System-dependent control info */
 };
 
 
-#ifdef MEM_STATS		/* optional extra stuff for statistics */
+#ifdef MEM_STATS                /* optional extra stuff for statistics */
 
 LOCAL(void)
 print_mem_stats (j_common_ptr cinfo, int pool_id)
@@ -196,19 +198,19 @@
    * This is helpful because message parm array can't handle longs.
    */
   fprintf(stderr, "Freeing pool %d, total space = %ld\n",
-	  pool_id, mem->total_space_allocated);
+          pool_id, mem->total_space_allocated);
 
   for (lhdr_ptr = mem->large_list[pool_id]; lhdr_ptr != NULL;
        lhdr_ptr = lhdr_ptr->next) {
     fprintf(stderr, "  Large chunk used %ld\n",
-	    (long) lhdr_ptr->bytes_used);
+            (long) lhdr_ptr->bytes_used);
   }
 
   for (shdr_ptr = mem->small_list[pool_id]; shdr_ptr != NULL;
        shdr_ptr = shdr_ptr->next) {
     fprintf(stderr, "  Small chunk used %ld free %ld\n",
-	    (long) shdr_ptr->bytes_used,
-	    (long) shdr_ptr->bytes_left);
+            (long) shdr_ptr->bytes_used,
+            (long) shdr_ptr->bytes_left);
   }
 }
 
@@ -221,7 +223,7 @@
 /* If we compiled MEM_STATS support, report alloc requests before dying */
 {
 #ifdef MEM_STATS
-  cinfo->err->trace_level = 2;	/* force self_destruct to report stats */
+  cinfo->err->trace_level = 2;  /* force self_destruct to report stats */
 #endif
   ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, which);
 }
@@ -244,19 +246,19 @@
  * adjustment.
  */
 
-static const size_t first_pool_slop[JPOOL_NUMPOOLS] = 
+static const size_t first_pool_slop[JPOOL_NUMPOOLS] =
 {
-	1600,			/* first PERMANENT pool */
-	16000			/* first IMAGE pool */
+        1600,                   /* first PERMANENT pool */
+        16000                   /* first IMAGE pool */
 };
 
-static const size_t extra_pool_slop[JPOOL_NUMPOOLS] = 
+static const size_t extra_pool_slop[JPOOL_NUMPOOLS] =
 {
-	0,			/* additional PERMANENT pools */
-	5000			/* additional IMAGE pools */
+        0,                      /* additional PERMANENT pools */
+        5000                    /* additional IMAGE pools */
 };
 
-#define MIN_SLOP  50		/* greater than 0 to avoid futile looping */
+#define MIN_SLOP  50            /* greater than 0 to avoid futile looping */
 
 
 METHODDEF(void *)
@@ -265,7 +267,7 @@
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
   small_pool_ptr hdr_ptr, prev_hdr_ptr;
-  char * data_ptr;
+  char *data_ptr;
   size_t min_request, slop;
 
   /*
@@ -274,20 +276,26 @@
    * and so that algorithms can straddle outside the proper area up
    * to the next alignment.
    */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 7);
+  }
   sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
 
   /* Check for unsatisfiable request (do now to ensure no overflow below) */
-  if ((SIZEOF(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK)
-    out_of_memory(cinfo, 1);	/* request exceeds malloc's ability */
+  if ((sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 1);    /* request exceeds malloc's ability */
 
   /* See if space is available in any existing pool */
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
   prev_hdr_ptr = NULL;
   hdr_ptr = mem->small_list[pool_id];
   while (hdr_ptr != NULL) {
     if (hdr_ptr->bytes_left >= sizeofobject)
-      break;			/* found pool with enough space */
+      break;                    /* found pool with enough space */
     prev_hdr_ptr = hdr_ptr;
     hdr_ptr = hdr_ptr->next;
   }
@@ -295,8 +303,8 @@
   /* Time to make a new pool? */
   if (hdr_ptr == NULL) {
     /* min_request is what we need now, slop is what will be leftover */
-    min_request = SIZEOF(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1;
-    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+    min_request = sizeof(small_pool_hdr) + sizeofobject + ALIGN_SIZE - 1;
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
       slop = first_pool_slop[pool_id];
     else
       slop = extra_pool_slop[pool_id];
@@ -307,17 +315,17 @@
     for (;;) {
       hdr_ptr = (small_pool_ptr) jpeg_get_small(cinfo, min_request + slop);
       if (hdr_ptr != NULL)
-	break;
+        break;
       slop /= 2;
-      if (slop < MIN_SLOP)	/* give up when it gets real small */
-	out_of_memory(cinfo, 2); /* jpeg_get_small failed */
+      if (slop < MIN_SLOP)      /* give up when it gets real small */
+        out_of_memory(cinfo, 2); /* jpeg_get_small failed */
     }
     mem->total_space_allocated += min_request + slop;
     /* Success, initialize the new pool header and add to end of list */
     hdr_ptr->next = NULL;
     hdr_ptr->bytes_used = 0;
     hdr_ptr->bytes_left = sizeofobject + slop;
-    if (prev_hdr_ptr == NULL)	/* first pool in class? */
+    if (prev_hdr_ptr == NULL)   /* first pool in class? */
       mem->small_list[pool_id] = hdr_ptr;
     else
       prev_hdr_ptr->next = hdr_ptr;
@@ -325,7 +333,7 @@
 
   /* OK, allocate the object from the current pool */
   data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */
-  data_ptr += SIZEOF(small_pool_hdr); /* ...by skipping the header... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
   if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
     data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
   data_ptr += hdr_ptr->bytes_used; /* point to place for object */
@@ -339,9 +347,8 @@
 /*
  * Allocation of "large" objects.
  *
- * The external semantics of these are the same as "small" objects,
- * except that FAR pointers are used on 80x86.  However the pool
- * management heuristics are quite different.  We assume that each
+ * The external semantics of these are the same as "small" objects.  However,
+ * the pool management heuristics are quite different.  We assume that each
  * request is large enough that it may as well be passed directly to
  * jpeg_get_large; the pool management just links everything together
  * so that we can free it all on demand.
@@ -350,35 +357,42 @@
  * deliberately bunch rows together to ensure a large request size.
  */
 
-METHODDEF(void FAR *)
+METHODDEF(void *)
 alloc_large (j_common_ptr cinfo, int pool_id, size_t sizeofobject)
 /* Allocate a "large" object */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
   large_pool_ptr hdr_ptr;
-  char FAR * data_ptr;
+  char *data_ptr;
 
   /*
    * Round up the requested size to a multiple of ALIGN_SIZE so that
    * algorithms can straddle outside the proper area up to the next
    * alignment.
    */
+  if (sizeofobject > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 8);
+  }
   sizeofobject = round_up_pow2(sizeofobject, ALIGN_SIZE);
 
   /* Check for unsatisfiable request (do now to ensure no overflow below) */
-  if ((SIZEOF(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) > MAX_ALLOC_CHUNK)
-    out_of_memory(cinfo, 3);	/* request exceeds malloc's ability */
+  if ((sizeof(large_pool_hdr) + sizeofobject + ALIGN_SIZE - 1) >
+      MAX_ALLOC_CHUNK)
+    out_of_memory(cinfo, 3);    /* request exceeds malloc's ability */
 
   /* Always make a new pool */
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   hdr_ptr = (large_pool_ptr) jpeg_get_large(cinfo, sizeofobject +
-					    SIZEOF(large_pool_hdr) +
-					    ALIGN_SIZE - 1);
+                                            sizeof(large_pool_hdr) +
+                                            ALIGN_SIZE - 1);
   if (hdr_ptr == NULL)
-    out_of_memory(cinfo, 4);	/* jpeg_get_large failed */
-  mem->total_space_allocated += sizeofobject + SIZEOF(large_pool_hdr) + ALIGN_SIZE - 1;
+    out_of_memory(cinfo, 4);    /* jpeg_get_large failed */
+  mem->total_space_allocated += sizeofobject + sizeof(large_pool_hdr) +
+                                ALIGN_SIZE - 1;
 
   /* Success, initialize the new pool header and add to list */
   hdr_ptr->next = mem->large_list[pool_id];
@@ -390,17 +404,16 @@
   mem->large_list[pool_id] = hdr_ptr;
 
   data_ptr = (char *) hdr_ptr; /* point to first data byte in pool... */
-  data_ptr += SIZEOF(small_pool_hdr); /* ...by skipping the header... */
+  data_ptr += sizeof(small_pool_hdr); /* ...by skipping the header... */
   if ((size_t)data_ptr % ALIGN_SIZE) /* ...and adjust for alignment */
     data_ptr += ALIGN_SIZE - (size_t)data_ptr % ALIGN_SIZE;
 
-  return (void FAR *) data_ptr;
+  return (void *) data_ptr;
 }
 
 
 /*
  * Creation of 2-D sample arrays.
- * The pointers are in near heap, the samples themselves in FAR heap.
  *
  * To minimize allocation overhead and to allow I/O of large contiguous
  * blocks, we allocate the sample rows in groups of as many rows as possible
@@ -417,7 +430,7 @@
 
 METHODDEF(JSAMPARRAY)
 alloc_sarray (j_common_ptr cinfo, int pool_id,
-	      JDIMENSION samplesperrow, JDIMENSION numrows)
+              JDIMENSION samplesperrow, JDIMENSION numrows)
 /* Allocate a 2-D sample array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -427,13 +440,20 @@
   long ltemp;
 
   /* Make sure each row is properly aligned */
-  if ((ALIGN_SIZE % SIZEOF(JSAMPLE)) != 0)
-    out_of_memory(cinfo, 5);	/* safety check */
-  samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) / SIZEOF(JSAMPLE));
+  if ((ALIGN_SIZE % sizeof(JSAMPLE)) != 0)
+    out_of_memory(cinfo, 5);    /* safety check */
+
+  if (samplesperrow > MAX_ALLOC_CHUNK) {
+    /* This prevents overflow/wrap-around in round_up_pow2() if sizeofobject
+       is close to SIZE_MAX. */
+    out_of_memory(cinfo, 9);
+  }
+  samplesperrow = (JDIMENSION)round_up_pow2(samplesperrow, (2 * ALIGN_SIZE) /
+                                                           sizeof(JSAMPLE));
 
   /* Calculate max # of rows allowed in one allocation chunk */
-  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
-	  ((long) samplesperrow * SIZEOF(JSAMPLE));
+  ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) /
+          ((long) samplesperrow * sizeof(JSAMPLE));
   if (ltemp <= 0)
     ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
   if (ltemp < (long) numrows)
@@ -444,15 +464,15 @@
 
   /* Get space for row pointers (small object) */
   result = (JSAMPARRAY) alloc_small(cinfo, pool_id,
-				    (size_t) (numrows * SIZEOF(JSAMPROW)));
+                                    (size_t) (numrows * sizeof(JSAMPROW)));
 
   /* Get the rows themselves (large objects) */
   currow = 0;
   while (currow < numrows) {
     rowsperchunk = MIN(rowsperchunk, numrows - currow);
     workspace = (JSAMPROW) alloc_large(cinfo, pool_id,
-	(size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
-		  * SIZEOF(JSAMPLE)));
+        (size_t) ((size_t) rowsperchunk * (size_t) samplesperrow
+                  * sizeof(JSAMPLE)));
     for (i = rowsperchunk; i > 0; i--) {
       result[currow++] = workspace;
       workspace += samplesperrow;
@@ -470,7 +490,7 @@
 
 METHODDEF(JBLOCKARRAY)
 alloc_barray (j_common_ptr cinfo, int pool_id,
-	      JDIMENSION blocksperrow, JDIMENSION numrows)
+              JDIMENSION blocksperrow, JDIMENSION numrows)
 /* Allocate a 2-D coefficient-block array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -480,12 +500,12 @@
   long ltemp;
 
   /* Make sure each row is properly aligned */
-  if ((SIZEOF(JBLOCK) % ALIGN_SIZE) != 0)
-    out_of_memory(cinfo, 6);	/* safety check */
+  if ((sizeof(JBLOCK) % ALIGN_SIZE) != 0)
+    out_of_memory(cinfo, 6);    /* safety check */
 
   /* Calculate max # of rows allowed in one allocation chunk */
-  ltemp = (MAX_ALLOC_CHUNK-SIZEOF(large_pool_hdr)) /
-	  ((long) blocksperrow * SIZEOF(JBLOCK));
+  ltemp = (MAX_ALLOC_CHUNK-sizeof(large_pool_hdr)) /
+          ((long) blocksperrow * sizeof(JBLOCK));
   if (ltemp <= 0)
     ERREXIT(cinfo, JERR_WIDTH_OVERFLOW);
   if (ltemp < (long) numrows)
@@ -496,15 +516,15 @@
 
   /* Get space for row pointers (small object) */
   result = (JBLOCKARRAY) alloc_small(cinfo, pool_id,
-				     (size_t) (numrows * SIZEOF(JBLOCKROW)));
+                                     (size_t) (numrows * sizeof(JBLOCKROW)));
 
   /* Get the rows themselves (large objects) */
   currow = 0;
   while (currow < numrows) {
     rowsperchunk = MIN(rowsperchunk, numrows - currow);
     workspace = (JBLOCKROW) alloc_large(cinfo, pool_id,
-	(size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
-		  * SIZEOF(JBLOCK)));
+        (size_t) ((size_t) rowsperchunk * (size_t) blocksperrow
+                  * sizeof(JBLOCK)));
     for (i = rowsperchunk; i > 0; i--) {
       result[currow++] = workspace;
       workspace += blocksperrow;
@@ -554,8 +574,8 @@
 
 METHODDEF(jvirt_sarray_ptr)
 request_virt_sarray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
-		     JDIMENSION samplesperrow, JDIMENSION numrows,
-		     JDIMENSION maxaccess)
+                     JDIMENSION samplesperrow, JDIMENSION numrows,
+                     JDIMENSION maxaccess)
 /* Request a virtual 2-D sample array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -563,18 +583,18 @@
 
   /* Only IMAGE-lifetime virtual arrays are currently supported */
   if (pool_id != JPOOL_IMAGE)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   /* get control block */
   result = (jvirt_sarray_ptr) alloc_small(cinfo, pool_id,
-					  SIZEOF(struct jvirt_sarray_control));
+                                          sizeof(struct jvirt_sarray_control));
 
-  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->mem_buffer = NULL;    /* marks array not yet realized */
   result->rows_in_array = numrows;
   result->samplesperrow = samplesperrow;
   result->maxaccess = maxaccess;
   result->pre_zero = pre_zero;
-  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->b_s_open = FALSE;     /* no associated backing-store object */
   result->next = mem->virt_sarray_list; /* add to list of virtual arrays */
   mem->virt_sarray_list = result;
 
@@ -584,8 +604,8 @@
 
 METHODDEF(jvirt_barray_ptr)
 request_virt_barray (j_common_ptr cinfo, int pool_id, boolean pre_zero,
-		     JDIMENSION blocksperrow, JDIMENSION numrows,
-		     JDIMENSION maxaccess)
+                     JDIMENSION blocksperrow, JDIMENSION numrows,
+                     JDIMENSION maxaccess)
 /* Request a virtual 2-D coefficient-block array */
 {
   my_mem_ptr mem = (my_mem_ptr) cinfo->mem;
@@ -593,18 +613,18 @@
 
   /* Only IMAGE-lifetime virtual arrays are currently supported */
   if (pool_id != JPOOL_IMAGE)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
   /* get control block */
   result = (jvirt_barray_ptr) alloc_small(cinfo, pool_id,
-					  SIZEOF(struct jvirt_barray_control));
+                                          sizeof(struct jvirt_barray_control));
 
-  result->mem_buffer = NULL;	/* marks array not yet realized */
+  result->mem_buffer = NULL;    /* marks array not yet realized */
   result->rows_in_array = numrows;
   result->blocksperrow = blocksperrow;
   result->maxaccess = maxaccess;
   result->pre_zero = pre_zero;
-  result->b_s_open = FALSE;	/* no associated backing-store object */
+  result->b_s_open = FALSE;     /* no associated backing-store object */
   result->next = mem->virt_barray_list; /* add to list of virtual arrays */
   mem->virt_barray_list = result;
 
@@ -631,26 +651,26 @@
   for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
     if (sptr->mem_buffer == NULL) { /* if not realized yet */
       space_per_minheight += (long) sptr->maxaccess *
-			     (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+                             (long) sptr->samplesperrow * sizeof(JSAMPLE);
       maximum_space += (long) sptr->rows_in_array *
-		       (long) sptr->samplesperrow * SIZEOF(JSAMPLE);
+                       (long) sptr->samplesperrow * sizeof(JSAMPLE);
     }
   }
   for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
     if (bptr->mem_buffer == NULL) { /* if not realized yet */
       space_per_minheight += (long) bptr->maxaccess *
-			     (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+                             (long) bptr->blocksperrow * sizeof(JBLOCK);
       maximum_space += (long) bptr->rows_in_array *
-		       (long) bptr->blocksperrow * SIZEOF(JBLOCK);
+                       (long) bptr->blocksperrow * sizeof(JBLOCK);
     }
   }
 
   if (space_per_minheight <= 0)
-    return;			/* no unrealized arrays, no work */
+    return;                     /* no unrealized arrays, no work */
 
   /* Determine amount of memory to actually use; this is system-dependent. */
   avail_mem = jpeg_mem_available(cinfo, space_per_minheight, maximum_space,
-				 mem->total_space_allocated);
+                                 mem->total_space_allocated);
 
   /* If the maximum space needed is available, make all the buffers full
    * height; otherwise parcel it out with the same number of minheights
@@ -673,19 +693,19 @@
     if (sptr->mem_buffer == NULL) { /* if not realized yet */
       minheights = ((long) sptr->rows_in_array - 1L) / sptr->maxaccess + 1L;
       if (minheights <= max_minheights) {
-	/* This buffer fits in memory */
-	sptr->rows_in_mem = sptr->rows_in_array;
+        /* This buffer fits in memory */
+        sptr->rows_in_mem = sptr->rows_in_array;
       } else {
-	/* It doesn't fit in memory, create backing store. */
-	sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
-	jpeg_open_backing_store(cinfo, & sptr->b_s_info,
-				(long) sptr->rows_in_array *
-				(long) sptr->samplesperrow *
-				(long) SIZEOF(JSAMPLE));
-	sptr->b_s_open = TRUE;
+        /* It doesn't fit in memory, create backing store. */
+        sptr->rows_in_mem = (JDIMENSION) (max_minheights * sptr->maxaccess);
+        jpeg_open_backing_store(cinfo, & sptr->b_s_info,
+                                (long) sptr->rows_in_array *
+                                (long) sptr->samplesperrow *
+                                (long) sizeof(JSAMPLE));
+        sptr->b_s_open = TRUE;
       }
       sptr->mem_buffer = alloc_sarray(cinfo, JPOOL_IMAGE,
-				      sptr->samplesperrow, sptr->rows_in_mem);
+                                      sptr->samplesperrow, sptr->rows_in_mem);
       sptr->rowsperchunk = mem->last_rowsperchunk;
       sptr->cur_start_row = 0;
       sptr->first_undef_row = 0;
@@ -697,19 +717,19 @@
     if (bptr->mem_buffer == NULL) { /* if not realized yet */
       minheights = ((long) bptr->rows_in_array - 1L) / bptr->maxaccess + 1L;
       if (minheights <= max_minheights) {
-	/* This buffer fits in memory */
-	bptr->rows_in_mem = bptr->rows_in_array;
+        /* This buffer fits in memory */
+        bptr->rows_in_mem = bptr->rows_in_array;
       } else {
-	/* It doesn't fit in memory, create backing store. */
-	bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
-	jpeg_open_backing_store(cinfo, & bptr->b_s_info,
-				(long) bptr->rows_in_array *
-				(long) bptr->blocksperrow *
-				(long) SIZEOF(JBLOCK));
-	bptr->b_s_open = TRUE;
+        /* It doesn't fit in memory, create backing store. */
+        bptr->rows_in_mem = (JDIMENSION) (max_minheights * bptr->maxaccess);
+        jpeg_open_backing_store(cinfo, & bptr->b_s_info,
+                                (long) bptr->rows_in_array *
+                                (long) bptr->blocksperrow *
+                                (long) sizeof(JBLOCK));
+        bptr->b_s_open = TRUE;
       }
       bptr->mem_buffer = alloc_barray(cinfo, JPOOL_IMAGE,
-				      bptr->blocksperrow, bptr->rows_in_mem);
+                                      bptr->blocksperrow, bptr->rows_in_mem);
       bptr->rowsperchunk = mem->last_rowsperchunk;
       bptr->cur_start_row = 0;
       bptr->first_undef_row = 0;
@@ -725,7 +745,7 @@
 {
   long bytesperrow, file_offset, byte_count, rows, thisrow, i;
 
-  bytesperrow = (long) ptr->samplesperrow * SIZEOF(JSAMPLE);
+  bytesperrow = (long) ptr->samplesperrow * sizeof(JSAMPLE);
   file_offset = ptr->cur_start_row * bytesperrow;
   /* Loop to read or write each allocation chunk in mem_buffer */
   for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
@@ -736,17 +756,17 @@
     rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
     /* Transfer no more than fits in file */
     rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
-    if (rows <= 0)		/* this chunk might be past end of file! */
+    if (rows <= 0)              /* this chunk might be past end of file! */
       break;
     byte_count = rows * bytesperrow;
     if (writing)
       (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
-					    (void FAR *) ptr->mem_buffer[i],
-					    file_offset, byte_count);
+                                            (void *) ptr->mem_buffer[i],
+                                            file_offset, byte_count);
     else
       (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
-					   (void FAR *) ptr->mem_buffer[i],
-					   file_offset, byte_count);
+                                           (void *) ptr->mem_buffer[i],
+                                           file_offset, byte_count);
     file_offset += byte_count;
   }
 }
@@ -758,7 +778,7 @@
 {
   long bytesperrow, file_offset, byte_count, rows, thisrow, i;
 
-  bytesperrow = (long) ptr->blocksperrow * SIZEOF(JBLOCK);
+  bytesperrow = (long) ptr->blocksperrow * sizeof(JBLOCK);
   file_offset = ptr->cur_start_row * bytesperrow;
   /* Loop to read or write each allocation chunk in mem_buffer */
   for (i = 0; i < (long) ptr->rows_in_mem; i += ptr->rowsperchunk) {
@@ -769,17 +789,17 @@
     rows = MIN(rows, (long) ptr->first_undef_row - thisrow);
     /* Transfer no more than fits in file */
     rows = MIN(rows, (long) ptr->rows_in_array - thisrow);
-    if (rows <= 0)		/* this chunk might be past end of file! */
+    if (rows <= 0)              /* this chunk might be past end of file! */
       break;
     byte_count = rows * bytesperrow;
     if (writing)
       (*ptr->b_s_info.write_backing_store) (cinfo, & ptr->b_s_info,
-					    (void FAR *) ptr->mem_buffer[i],
-					    file_offset, byte_count);
+                                            (void *) ptr->mem_buffer[i],
+                                            file_offset, byte_count);
     else
       (*ptr->b_s_info.read_backing_store) (cinfo, & ptr->b_s_info,
-					   (void FAR *) ptr->mem_buffer[i],
-					   file_offset, byte_count);
+                                           (void *) ptr->mem_buffer[i],
+                                           file_offset, byte_count);
     file_offset += byte_count;
   }
 }
@@ -787,8 +807,8 @@
 
 METHODDEF(JSAMPARRAY)
 access_virt_sarray (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
-		    JDIMENSION start_row, JDIMENSION num_rows,
-		    boolean writable)
+                    JDIMENSION start_row, JDIMENSION num_rows,
+                    boolean writable)
 /* Access the part of a virtual sample array starting at start_row */
 /* and extending for num_rows rows.  writable is true if  */
 /* caller intends to modify the accessed area. */
@@ -826,7 +846,7 @@
 
       ltemp = (long) end_row - (long) ptr->rows_in_mem;
       if (ltemp < 0)
-	ltemp = 0;		/* don't fall off front end of file */
+        ltemp = 0;              /* don't fall off front end of file */
       ptr->cur_start_row = (JDIMENSION) ltemp;
     }
     /* Read in the selected part of the array.
@@ -841,25 +861,25 @@
    */
   if (ptr->first_undef_row < end_row) {
     if (ptr->first_undef_row < start_row) {
-      if (writable)		/* writer skipped over a section of array */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
-      undef_row = start_row;	/* but reader is allowed to read ahead */
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
     } else {
       undef_row = ptr->first_undef_row;
     }
     if (writable)
       ptr->first_undef_row = end_row;
     if (ptr->pre_zero) {
-      size_t bytesperrow = (size_t) ptr->samplesperrow * SIZEOF(JSAMPLE);
+      size_t bytesperrow = (size_t) ptr->samplesperrow * sizeof(JSAMPLE);
       undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
       end_row -= ptr->cur_start_row;
       while (undef_row < end_row) {
-	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
-	undef_row++;
+        jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow);
+        undef_row++;
       }
     } else {
-      if (! writable)		/* reader looking at undefined data */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      if (! writable)           /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
     }
   }
   /* Flag the buffer dirty if caller will write in it */
@@ -872,8 +892,8 @@
 
 METHODDEF(JBLOCKARRAY)
 access_virt_barray (j_common_ptr cinfo, jvirt_barray_ptr ptr,
-		    JDIMENSION start_row, JDIMENSION num_rows,
-		    boolean writable)
+                    JDIMENSION start_row, JDIMENSION num_rows,
+                    boolean writable)
 /* Access the part of a virtual block array starting at start_row */
 /* and extending for num_rows rows.  writable is true if  */
 /* caller intends to modify the accessed area. */
@@ -911,7 +931,7 @@
 
       ltemp = (long) end_row - (long) ptr->rows_in_mem;
       if (ltemp < 0)
-	ltemp = 0;		/* don't fall off front end of file */
+        ltemp = 0;              /* don't fall off front end of file */
       ptr->cur_start_row = (JDIMENSION) ltemp;
     }
     /* Read in the selected part of the array.
@@ -926,25 +946,25 @@
    */
   if (ptr->first_undef_row < end_row) {
     if (ptr->first_undef_row < start_row) {
-      if (writable)		/* writer skipped over a section of array */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
-      undef_row = start_row;	/* but reader is allowed to read ahead */
+      if (writable)             /* writer skipped over a section of array */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      undef_row = start_row;    /* but reader is allowed to read ahead */
     } else {
       undef_row = ptr->first_undef_row;
     }
     if (writable)
       ptr->first_undef_row = end_row;
     if (ptr->pre_zero) {
-      size_t bytesperrow = (size_t) ptr->blocksperrow * SIZEOF(JBLOCK);
+      size_t bytesperrow = (size_t) ptr->blocksperrow * sizeof(JBLOCK);
       undef_row -= ptr->cur_start_row; /* make indexes relative to buffer */
       end_row -= ptr->cur_start_row;
       while (undef_row < end_row) {
-	jzero_far((void FAR *) ptr->mem_buffer[undef_row], bytesperrow);
-	undef_row++;
+        jzero_far((void *) ptr->mem_buffer[undef_row], bytesperrow);
+        undef_row++;
       }
     } else {
-      if (! writable)		/* reader looking at undefined data */
-	ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
+      if (! writable)           /* reader looking at undefined data */
+        ERREXIT(cinfo, JERR_BAD_VIRTUAL_ACCESS);
     }
   }
   /* Flag the buffer dirty if caller will write in it */
@@ -968,7 +988,7 @@
   size_t space_freed;
 
   if (pool_id < 0 || pool_id >= JPOOL_NUMPOOLS)
-    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id);	/* safety check */
+    ERREXIT1(cinfo, JERR_BAD_POOL_ID, pool_id); /* safety check */
 
 #ifdef MEM_STATS
   if (cinfo->err->trace_level > 1)
@@ -981,16 +1001,16 @@
     jvirt_barray_ptr bptr;
 
     for (sptr = mem->virt_sarray_list; sptr != NULL; sptr = sptr->next) {
-      if (sptr->b_s_open) {	/* there may be no backing store */
-	sptr->b_s_open = FALSE;	/* prevent recursive close if error */
-	(*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
+      if (sptr->b_s_open) {     /* there may be no backing store */
+        sptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*sptr->b_s_info.close_backing_store) (cinfo, & sptr->b_s_info);
       }
     }
     mem->virt_sarray_list = NULL;
     for (bptr = mem->virt_barray_list; bptr != NULL; bptr = bptr->next) {
-      if (bptr->b_s_open) {	/* there may be no backing store */
-	bptr->b_s_open = FALSE;	/* prevent recursive close if error */
-	(*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
+      if (bptr->b_s_open) {     /* there may be no backing store */
+        bptr->b_s_open = FALSE; /* prevent recursive close if error */
+        (*bptr->b_s_info.close_backing_store) (cinfo, & bptr->b_s_info);
       }
     }
     mem->virt_barray_list = NULL;
@@ -1003,9 +1023,9 @@
   while (lhdr_ptr != NULL) {
     large_pool_ptr next_lhdr_ptr = lhdr_ptr->next;
     space_freed = lhdr_ptr->bytes_used +
-		  lhdr_ptr->bytes_left +
-		  SIZEOF(large_pool_hdr);
-    jpeg_free_large(cinfo, (void FAR *) lhdr_ptr, space_freed);
+                  lhdr_ptr->bytes_left +
+                  sizeof(large_pool_hdr);
+    jpeg_free_large(cinfo, (void *) lhdr_ptr, space_freed);
     mem->total_space_allocated -= space_freed;
     lhdr_ptr = next_lhdr_ptr;
   }
@@ -1017,8 +1037,8 @@
   while (shdr_ptr != NULL) {
     small_pool_ptr next_shdr_ptr = shdr_ptr->next;
     space_freed = shdr_ptr->bytes_used +
-		  shdr_ptr->bytes_left +
-		  SIZEOF(small_pool_hdr);
+                  shdr_ptr->bytes_left +
+                  sizeof(small_pool_hdr);
     jpeg_free_small(cinfo, (void *) shdr_ptr, space_freed);
     mem->total_space_allocated -= space_freed;
     shdr_ptr = next_shdr_ptr;
@@ -1045,10 +1065,10 @@
   }
 
   /* Release the memory manager control block too. */
-  jpeg_free_small(cinfo, (void *) cinfo->mem, SIZEOF(my_memory_mgr));
-  cinfo->mem = NULL;		/* ensures I will be called only once */
+  jpeg_free_small(cinfo, (void *) cinfo->mem, sizeof(my_memory_mgr));
+  cinfo->mem = NULL;            /* ensures I will be called only once */
 
-  jpeg_mem_term(cinfo);		/* system-dependent cleanup */
+  jpeg_mem_term(cinfo);         /* system-dependent cleanup */
 }
 
 
@@ -1065,10 +1085,10 @@
   int pool;
   size_t test_mac;
 
-  cinfo->mem = NULL;		/* for safety if init fails */
+  cinfo->mem = NULL;            /* for safety if init fails */
 
   /* Check for configuration errors.
-   * SIZEOF(ALIGN_TYPE) should be a power of 2; otherwise, it probably
+   * sizeof(ALIGN_TYPE) should be a power of 2; otherwise, it probably
    * doesn't reflect any real hardware alignment requirement.
    * The test is a little tricky: for X>0, X and X-1 have no one-bits
    * in common if and only if X is a power of 2, ie has only one one-bit.
@@ -1089,10 +1109,10 @@
   max_to_use = jpeg_mem_init(cinfo); /* system-dependent initialization */
 
   /* Attempt to allocate memory manager's control block */
-  mem = (my_mem_ptr) jpeg_get_small(cinfo, SIZEOF(my_memory_mgr));
+  mem = (my_mem_ptr) jpeg_get_small(cinfo, sizeof(my_memory_mgr));
 
   if (mem == NULL) {
-    jpeg_mem_term(cinfo);	/* system-dependent cleanup */
+    jpeg_mem_term(cinfo);       /* system-dependent cleanup */
     ERREXIT1(cinfo, JERR_OUT_OF_MEMORY, 0);
   }
 
@@ -1122,7 +1142,7 @@
   mem->virt_sarray_list = NULL;
   mem->virt_barray_list = NULL;
 
-  mem->total_space_allocated = SIZEOF(my_memory_mgr);
+  mem->total_space_allocated = sizeof(my_memory_mgr);
 
   /* Declare ourselves open for business */
   cinfo->mem = & mem->pub;
@@ -1134,15 +1154,15 @@
    * this feature.
    */
 #ifndef NO_GETENV
-  { char * memenv;
+  { char *memenv;
 
     if ((memenv = getenv("JPEGMEM")) != NULL) {
       char ch = 'x';
 
       if (sscanf(memenv, "%ld%c", &max_to_use, &ch) > 0) {
-	if (ch == 'm' || ch == 'M')
-	  max_to_use *= 1000L;
-	mem->pub.max_memory_to_use = max_to_use * 1000L;
+        if (ch == 'm' || ch == 'M')
+          max_to_use *= 1000L;
+        mem->pub.max_memory_to_use = max_to_use * 1000L;
       }
     }
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemnobs.c glmark2-2021.02/src/libjpeg-turbo/jmemnobs.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemnobs.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jmemnobs.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemnobs.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1992-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file provides a really simple implementation of the system-
  * dependent portion of the JPEG memory manager.  This implementation
@@ -18,11 +21,11 @@
 #define JPEG_INTERNALS
 #include "jinclude.h"
 #include "jpeglib.h"
-#include "jmemsys.h"		/* import the system-dependent declarations */
+#include "jmemsys.h"            /* import the system-dependent declarations */
 
-#ifndef HAVE_STDLIB_H		/* <stdlib.h> should declare malloc(),free() */
-extern void * malloc JPP((size_t size));
-extern void free JPP((void *ptr));
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc(),free() */
+extern void *malloc (size_t size);
+extern void free (void *ptr);
 #endif
 
 
@@ -38,7 +41,7 @@
 }
 
 GLOBAL(void)
-jpeg_free_small (j_common_ptr cinfo, void * object, size_t sizeofobject)
+jpeg_free_small (j_common_ptr cinfo, void *object, size_t sizeofobject)
 {
   free(object);
 }
@@ -46,19 +49,16 @@
 
 /*
  * "Large" objects are treated the same as "small" ones.
- * NB: although we include FAR keywords in the routine declarations,
- * this file won't actually work in 80x86 small/medium model; at least,
- * you probably won't be able to process useful-size images in only 64KB.
  */
 
-GLOBAL(void FAR *)
+GLOBAL(void *)
 jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject)
 {
-  return (void FAR *) malloc(sizeofobject);
+  return (void *) malloc(sizeofobject);
 }
 
 GLOBAL(void)
-jpeg_free_large (j_common_ptr cinfo, void FAR * object, size_t sizeofobject)
+jpeg_free_large (j_common_ptr cinfo, void *object, size_t sizeofobject)
 {
   free(object);
 }
@@ -71,7 +71,7 @@
 
 GLOBAL(size_t)
 jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed,
-		    size_t max_bytes_needed, size_t already_allocated)
+                    size_t max_bytes_needed, size_t already_allocated)
 {
   return max_bytes_needed;
 }
@@ -85,7 +85,7 @@
 
 GLOBAL(void)
 jpeg_open_backing_store (j_common_ptr cinfo, backing_store_ptr info,
-			 long total_bytes_needed)
+                         long total_bytes_needed)
 {
   ERREXIT(cinfo, JERR_NO_BACKING_STORE);
 }
@@ -99,7 +99,7 @@
 GLOBAL(long)
 jpeg_mem_init (j_common_ptr cinfo)
 {
-  return 0;			/* just set max_memory_to_use to 0 */
+  return 0;                     /* just set max_memory_to_use to 0 */
 }
 
 GLOBAL(void)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemsys.h glmark2-2021.02/src/libjpeg-turbo/jmemsys.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmemsys.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jmemsys.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jmemsys.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1992-1997, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This include file defines the interface between the system-independent
  * and system-dependent portions of the JPEG memory manager.  No other
@@ -14,25 +17,10 @@
  * in the IJG distribution.  You may need to modify it if you write a
  * custom memory manager.  If system-dependent changes are needed in
  * this file, the best method is to #ifdef them based on a configuration
- * symbol supplied in jconfig.h, as we have done with USE_MSDOS_MEMMGR
- * and USE_MAC_MEMMGR.
+ * symbol supplied in jconfig.h.
  */
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_get_small		jGetSmall
-#define jpeg_free_small		jFreeSmall
-#define jpeg_get_large		jGetLarge
-#define jpeg_free_large		jFreeLarge
-#define jpeg_mem_available	jMemAvail
-#define jpeg_open_backing_store	jOpenBackStore
-#define jpeg_mem_init		jMemInit
-#define jpeg_mem_term		jMemTerm
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-
 /*
  * These two functions are used to allocate and release small chunks of
  * memory.  (Typically the total amount requested through jpeg_get_small is
@@ -41,40 +29,36 @@
  * and free; in particular, jpeg_get_small must return NULL on failure.
  * On most systems, these ARE malloc and free.  jpeg_free_small is passed the
  * size of the object being freed, just in case it's needed.
- * On an 80x86 machine using small-data memory model, these manage near heap.
  */
 
-EXTERN(void *) jpeg_get_small JPP((j_common_ptr cinfo, size_t sizeofobject));
-EXTERN(void) jpeg_free_small JPP((j_common_ptr cinfo, void * object,
-				  size_t sizeofobject));
+EXTERN(void *) jpeg_get_small (j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_small (j_common_ptr cinfo, void *object,
+                              size_t sizeofobject);
 
 /*
  * These two functions are used to allocate and release large chunks of
  * memory (up to the total free space designated by jpeg_mem_available).
- * The interface is the same as above, except that on an 80x86 machine,
- * far pointers are used.  On most other machines these are identical to
- * the jpeg_get/free_small routines; but we keep them separate anyway,
- * in case a different allocation strategy is desirable for large chunks.
+ * These are identical to the jpeg_get/free_small routines; but we keep them
+ * separate anyway, in case a different allocation strategy is desirable for
+ * large chunks.
  */
 
-EXTERN(void FAR *) jpeg_get_large JPP((j_common_ptr cinfo,
-				       size_t sizeofobject));
-EXTERN(void) jpeg_free_large JPP((j_common_ptr cinfo, void FAR * object,
-				  size_t sizeofobject));
+EXTERN(void *) jpeg_get_large (j_common_ptr cinfo, size_t sizeofobject);
+EXTERN(void) jpeg_free_large (j_common_ptr cinfo, void *object,
+                              size_t sizeofobject);
 
 /*
  * The macro MAX_ALLOC_CHUNK designates the maximum number of bytes that may
  * be requested in a single call to jpeg_get_large (and jpeg_get_small for that
- * matter, but that case should never come into play).  This macro is needed
+ * matter, but that case should never come into play).  This macro was needed
  * to model the 64Kb-segment-size limit of far addressing on 80x86 machines.
- * On those machines, we expect that jconfig.h will provide a proper value.
- * On machines with 32-bit flat address spaces, any large constant may be used.
+ * On machines with flat address spaces, any large constant may be used.
  *
  * NB: jmemmgr.c expects that MAX_ALLOC_CHUNK will be representable as type
  * size_t and will be a multiple of sizeof(align_type).
  */
 
-#ifndef MAX_ALLOC_CHUNK		/* may be overridden in jconfig.h */
+#ifndef MAX_ALLOC_CHUNK         /* may be overridden in jconfig.h */
 #define MAX_ALLOC_CHUNK  1000000000L
 #endif
 
@@ -100,10 +84,9 @@
  * Conversely, zero may be returned to always use the minimum amount of memory.
  */
 
-EXTERN(size_t) jpeg_mem_available JPP((j_common_ptr cinfo,
-				     size_t min_bytes_needed,
-				     size_t max_bytes_needed,
-				     size_t already_allocated));
+EXTERN(size_t) jpeg_mem_available (j_common_ptr cinfo, size_t min_bytes_needed,
+                                   size_t max_bytes_needed,
+                                   size_t already_allocated);
 
 
 /*
@@ -113,56 +96,53 @@
  * are private to the system-dependent backing store routines.
  */
 
-#define TEMP_NAME_LENGTH   64	/* max length of a temporary file's name */
+#define TEMP_NAME_LENGTH   64   /* max length of a temporary file's name */
 
 
-#ifdef USE_MSDOS_MEMMGR		/* DOS-specific junk */
+#ifdef USE_MSDOS_MEMMGR         /* DOS-specific junk */
 
-typedef unsigned short XMSH;	/* type of extended-memory handles */
-typedef unsigned short EMSH;	/* type of expanded-memory handles */
+typedef unsigned short XMSH;    /* type of extended-memory handles */
+typedef unsigned short EMSH;    /* type of expanded-memory handles */
 
 typedef union {
-  short file_handle;		/* DOS file handle if it's a temp file */
-  XMSH xms_handle;		/* handle if it's a chunk of XMS */
-  EMSH ems_handle;		/* handle if it's a chunk of EMS */
+  short file_handle;            /* DOS file handle if it's a temp file */
+  XMSH xms_handle;              /* handle if it's a chunk of XMS */
+  EMSH ems_handle;              /* handle if it's a chunk of EMS */
 } handle_union;
 
 #endif /* USE_MSDOS_MEMMGR */
 
-#ifdef USE_MAC_MEMMGR		/* Mac-specific junk */
+#ifdef USE_MAC_MEMMGR           /* Mac-specific junk */
 #include <Files.h>
 #endif /* USE_MAC_MEMMGR */
 
 
-typedef struct backing_store_struct * backing_store_ptr;
+typedef struct backing_store_struct *backing_store_ptr;
 
 typedef struct backing_store_struct {
   /* Methods for reading/writing/closing this backing-store object */
-  JMETHOD(void, read_backing_store, (j_common_ptr cinfo,
-				     backing_store_ptr info,
-				     void FAR * buffer_address,
-				     long file_offset, long byte_count));
-  JMETHOD(void, write_backing_store, (j_common_ptr cinfo,
-				      backing_store_ptr info,
-				      void FAR * buffer_address,
-				      long file_offset, long byte_count));
-  JMETHOD(void, close_backing_store, (j_common_ptr cinfo,
-				      backing_store_ptr info));
+  void (*read_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                              void *buffer_address, long file_offset,
+                              long byte_count);
+  void (*write_backing_store) (j_common_ptr cinfo, backing_store_ptr info,
+                               void *buffer_address, long file_offset,
+                               long byte_count);
+  void (*close_backing_store) (j_common_ptr cinfo, backing_store_ptr info);
 
   /* Private fields for system-dependent backing-store management */
 #ifdef USE_MSDOS_MEMMGR
   /* For the MS-DOS manager (jmemdos.c), we need: */
-  handle_union handle;		/* reference to backing-store storage object */
+  handle_union handle;          /* reference to backing-store storage object */
   char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
 #else
 #ifdef USE_MAC_MEMMGR
   /* For the Mac manager (jmemmac.c), we need: */
-  short temp_file;		/* file reference number to temp file */
-  FSSpec tempSpec;		/* the FSSpec for the temp file */
+  short temp_file;              /* file reference number to temp file */
+  FSSpec tempSpec;              /* the FSSpec for the temp file */
   char temp_name[TEMP_NAME_LENGTH]; /* name if it's a file */
 #else
   /* For a typical implementation with temp files, we need: */
-  FILE * temp_file;		/* stdio reference to temp file */
+  FILE *temp_file;              /* stdio reference to temp file */
   char temp_name[TEMP_NAME_LENGTH]; /* name of temp file */
 #endif
 #endif
@@ -177,9 +157,9 @@
  * just take an error exit.)
  */
 
-EXTERN(void) jpeg_open_backing_store JPP((j_common_ptr cinfo,
-					  backing_store_ptr info,
-					  long total_bytes_needed));
+EXTERN(void) jpeg_open_backing_store (j_common_ptr cinfo,
+                                      backing_store_ptr info,
+                                      long total_bytes_needed);
 
 
 /*
@@ -194,5 +174,5 @@
  * all opened backing-store objects have been closed.
  */
 
-EXTERN(long) jpeg_mem_init JPP((j_common_ptr cinfo));
-EXTERN(void) jpeg_mem_term JPP((j_common_ptr cinfo));
+EXTERN(long) jpeg_mem_init (j_common_ptr cinfo);
+EXTERN(void) jpeg_mem_term (j_common_ptr cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmorecfg.h glmark2-2021.02/src/libjpeg-turbo/jmorecfg.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jmorecfg.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jmorecfg.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,13 @@
 /*
  * jmorecfg.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
- * Copyright (C) 2009, 2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * Modified 1997-2009 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2011, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains additional configuration options that customize the
  * JPEG software for special applications or support machine-dependent
@@ -13,18 +16,6 @@
 
 
 /*
- * Define BITS_IN_JSAMPLE as either
- *   8   for 8-bit sample values (the usual setting)
- *   12  for 12-bit sample values
- * Only 8 and 12 are legal data precisions for lossy JPEG according to the
- * JPEG standard, and the IJG code does not support anything else!
- * We do not support run-time selection of data precision, sorry.
- */
-
-#define BITS_IN_JSAMPLE  8	/* use 8 or 12 */
-
-
-/*
  * Maximum number of components (color channels) allowed in JPEG image.
  * To meet the letter of the JPEG spec, set this to 255.  However, darn
  * few applications need more than 4 channels (maybe 5 for CMYK + alpha
@@ -33,7 +24,7 @@
  * bytes of storage, whether actually used in an image or not.)
  */
 
-#define MAX_COMPONENTS  10	/* maximum number of image components */
+#define MAX_COMPONENTS  10      /* maximum number of image components */
 
 
 /*
@@ -71,8 +62,8 @@
 
 #endif /* HAVE_UNSIGNED_CHAR */
 
-#define MAXJSAMPLE	255
-#define CENTERJSAMPLE	128
+#define MAXJSAMPLE      255
+#define CENTERJSAMPLE   128
 
 #endif /* BITS_IN_JSAMPLE == 8 */
 
@@ -85,8 +76,8 @@
 typedef short JSAMPLE;
 #define GETJSAMPLE(value)  ((int) (value))
 
-#define MAXJSAMPLE	4095
-#define CENTERJSAMPLE	2048
+#define MAXJSAMPLE      4095
+#define CENTERJSAMPLE   2048
 
 #endif /* BITS_IN_JSAMPLE == 12 */
 
@@ -152,21 +143,52 @@
 
 /* INT16 must hold at least the values -32768..32767. */
 
-#ifndef XMD_H			/* X11/xmd.h correctly defines INT16 */
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT16 */
 typedef short INT16;
 #endif
 
-/* INT32 must hold at least signed 32-bit values. */
+/* INT32 must hold at least signed 32-bit values.
+ *
+ * NOTE: The INT32 typedef dates back to libjpeg v5 (1994.)  Integers were
+ * sometimes 16-bit back then (MS-DOS), which is why INT32 is typedef'd to
+ * long.  It also wasn't common (or at least as common) in 1994 for INT32 to be
+ * defined by platform headers.  Since then, however, INT32 is defined in
+ * several other common places:
+ *
+ * Xmd.h (X11 header) typedefs INT32 to int on 64-bit platforms and long on
+ * 32-bit platforms (i.e always a 32-bit signed type.)
+ *
+ * basetsd.h (Win32 header) typedefs INT32 to int (always a 32-bit signed type
+ * on modern platforms.)
+ *
+ * qglobal.h (Qt header) typedefs INT32 to int (always a 32-bit signed type on
+ * modern platforms.)
+ *
+ * This is a recipe for conflict, since "long" and "int" aren't always
+ * compatible types.  Since the definition of INT32 has technically been part
+ * of the libjpeg API for more than 20 years, we can't remove it, but we do not
+ * use it internally any longer.  We instead define a separate type (JLONG)
+ * for internal use, which ensures that internal behavior will always be the
+ * same regardless of any external headers that may be included.
+ */
 
-#ifndef XMD_H			/* X11/xmd.h correctly defines INT32 */
+#ifndef XMD_H                   /* X11/xmd.h correctly defines INT32 */
+#ifndef _BASETSD_H_		/* Microsoft defines it in basetsd.h */
+#ifndef _BASETSD_H		/* MinGW is slightly different */
+#ifndef QGLOBAL_H		/* Qt defines it in qglobal.h */
 typedef long INT32;
 #endif
+#endif
+#endif
+#endif
 
 /* Datatype used for image dimensions.  The JPEG standard only supports
  * images up to 64K*64K due to 16-bit fields in SOF markers.  Therefore
  * "unsigned int" is sufficient on all machines.  However, if you need to
  * handle larger images and you don't mind deviating from the spec, you
- * can change this datatype.
+ * can change this datatype.  (Note that changing this datatype will
+ * potentially require modifying the SIMD code.  The x86-64 SIMD extensions,
+ * in particular, assume a 32-bit JDIMENSION.)
  */
 
 typedef unsigned int JDIMENSION;
@@ -182,39 +204,31 @@
  */
 
 /* a function called through method pointers: */
-#define METHODDEF(type)		static type
+#define METHODDEF(type)         static type
 /* a function used only in its module: */
-#define LOCAL(type)		static type
+#define LOCAL(type)             static type
 /* a function referenced thru EXTERNs: */
-#define GLOBAL(type)		type
+#define GLOBAL(type)            type
 /* a reference to a GLOBAL function: */
-#define EXTERN(type)		extern type
+#define EXTERN(type)            extern type
 
 
-/* This macro is used to declare a "method", that is, a function pointer.
- * We want to supply prototype parameters if the compiler can cope.
- * Note that the arglist parameter must be parenthesized!
- * Again, you can customize this if you need special linkage keywords.
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
  */
 
-#ifdef HAVE_PROTOTYPES
 #define JMETHOD(type,methodname,arglist)  type (*methodname) arglist
-#else
-#define JMETHOD(type,methodname,arglist)  type (*methodname) ()
-#endif
 
 
-/* Here is the pseudo-keyword for declaring pointers that must be "far"
- * on 80x86 machines.  Most of the specialized coding for 80x86 is handled
- * by just saying "FAR *" where such a pointer is needed.  In a few places
- * explicit coding is needed; see uses of the NEED_FAR_POINTERS symbol.
+/* libjpeg-turbo no longer supports platforms that have far symbols (MS-DOS),
+ * but again, some software relies on this macro.
  */
 
-#ifdef NEED_FAR_POINTERS
-#define FAR  far
-#else
+#undef FAR
 #define FAR
-#endif
 
 
 /*
@@ -227,11 +241,11 @@
 #ifndef HAVE_BOOLEAN
 typedef int boolean;
 #endif
-#ifndef FALSE			/* in case these macros already exist */
-#define FALSE	0		/* values of boolean */
+#ifndef FALSE                   /* in case these macros already exist */
+#define FALSE   0               /* values of boolean */
 #endif
 #ifndef TRUE
-#define TRUE	1
+#define TRUE    1
 #endif
 
 
@@ -259,15 +273,15 @@
 
 /* Capability options common to encoder and decoder: */
 
-#define DCT_ISLOW_SUPPORTED	/* slow but accurate integer algorithm */
-#define DCT_IFAST_SUPPORTED	/* faster, less accurate integer method */
-#define DCT_FLOAT_SUPPORTED	/* floating-point: accurate, fast on fast HW */
+#define DCT_ISLOW_SUPPORTED     /* slow but accurate integer algorithm */
+#define DCT_IFAST_SUPPORTED     /* faster, less accurate integer method */
+#define DCT_FLOAT_SUPPORTED     /* floating-point: accurate, fast on fast HW */
 
 /* Encoder capability options: */
 
 #define C_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
-#define C_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
-#define ENTROPY_OPT_SUPPORTED	    /* Optimization of entropy coding parms? */
+#define C_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define ENTROPY_OPT_SUPPORTED       /* Optimization of entropy coding parms? */
 /* Note: if you selected 12-bit data precision, it is dangerous to turn off
  * ENTROPY_OPT_SUPPORTED.  The standard Huffman tables are only good for 8-bit
  * precision, so jchuff.c normally uses entropy optimization to compute
@@ -281,39 +295,45 @@
 /* Decoder capability options: */
 
 #define D_MULTISCAN_FILES_SUPPORTED /* Multiple-scan JPEG files? */
-#define D_PROGRESSIVE_SUPPORTED	    /* Progressive JPEG? (Requires MULTISCAN)*/
-#define SAVE_MARKERS_SUPPORTED	    /* jpeg_save_markers() needed? */
+#define D_PROGRESSIVE_SUPPORTED     /* Progressive JPEG? (Requires MULTISCAN)*/
+#define SAVE_MARKERS_SUPPORTED      /* jpeg_save_markers() needed? */
 #define BLOCK_SMOOTHING_SUPPORTED   /* Block smoothing? (Progressive only) */
-#define IDCT_SCALING_SUPPORTED	    /* Output rescaling via IDCT? */
+#define IDCT_SCALING_SUPPORTED      /* Output rescaling via IDCT? */
 #undef  UPSAMPLE_SCALING_SUPPORTED  /* Output rescaling at upsample stage? */
 #define UPSAMPLE_MERGING_SUPPORTED  /* Fast path for sloppy upsampling? */
-#define QUANT_1PASS_SUPPORTED	    /* 1-pass color quantization? */
-#define QUANT_2PASS_SUPPORTED	    /* 2-pass color quantization? */
+#define QUANT_1PASS_SUPPORTED       /* 1-pass color quantization? */
+#define QUANT_2PASS_SUPPORTED       /* 2-pass color quantization? */
 
 /* more capability options later, no doubt */
 
 
 /*
- * Ordering of RGB data in scanlines passed to or from the application.
- * If your application wants to deal with data in the order B,G,R, just
- * change these macros.  You can also deal with formats such as R,G,B,X
- * (one extra byte per pixel) by changing RGB_PIXELSIZE.  Note that changing
- * the offsets will also change the order in which colormap data is organized.
- * RESTRICTIONS:
- * 1. The sample applications cjpeg,djpeg do NOT support modified RGB formats.
- * 2. These macros only affect RGB<=>YCbCr color conversion, so they are not
- *    useful if you are using JPEG color spaces other than YCbCr or grayscale.
- * 3. The color quantizer modules will not behave desirably if RGB_PIXELSIZE
- *    is not 3 (they don't understand about dummy color components!).  So you
- *    can't use color quantization if you change that value.
- */
-
-#define RGB_RED		0	/* Offset of Red in an RGB scanline element */
-#define RGB_GREEN	1	/* Offset of Green */
-#define RGB_BLUE	2	/* Offset of Blue */
-#define RGB_PIXELSIZE	3	/* JSAMPLEs per RGB scanline element */
+ * The RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros are a vestigial
+ * feature of libjpeg.  The idea was that, if an application developer needed
+ * to compress from/decompress to a BGR/BGRX/RGBX/XBGR/XRGB buffer, they could
+ * change these macros, rebuild libjpeg, and link their application statically
+ * with it.  In reality, few people ever did this, because there were some
+ * severe restrictions involved (cjpeg and djpeg no longer worked properly,
+ * compressing/decompressing RGB JPEGs no longer worked properly, and the color
+ * quantizer wouldn't work with pixel sizes other than 3.)  Further, since all
+ * of the O/S-supplied versions of libjpeg were built with the default values
+ * of RGB_RED, RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE, many applications have
+ * come to regard these values as immutable.
+ *
+ * The libjpeg-turbo colorspace extensions provide a much cleaner way of
+ * compressing from/decompressing to buffers with arbitrary component orders
+ * and pixel sizes.  Thus, we do not support changing the values of RGB_RED,
+ * RGB_GREEN, RGB_BLUE, or RGB_PIXELSIZE.  In addition to the restrictions
+ * listed above, changing these values will also break the SIMD extensions and
+ * the regression tests.
+ */
+
+#define RGB_RED         0       /* Offset of Red in an RGB scanline element */
+#define RGB_GREEN       1       /* Offset of Green */
+#define RGB_BLUE        2       /* Offset of Blue */
+#define RGB_PIXELSIZE   3       /* JSAMPLEs per RGB scanline element */
 
-#define JPEG_NUMCS 16
+#define JPEG_NUMCS 17
 
 #define EXT_RGB_RED        0
 #define EXT_RGB_GREEN      1
@@ -348,25 +368,29 @@
 static const int rgb_red[JPEG_NUMCS] = {
   -1, -1, RGB_RED, -1, -1, -1, EXT_RGB_RED, EXT_RGBX_RED,
   EXT_BGR_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
-  EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED
+  EXT_RGBX_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED,
+  -1
 };
 
 static const int rgb_green[JPEG_NUMCS] = {
   -1, -1, RGB_GREEN, -1, -1, -1, EXT_RGB_GREEN, EXT_RGBX_GREEN,
   EXT_BGR_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
-  EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN
+  EXT_RGBX_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN,
+  -1
 };
 
 static const int rgb_blue[JPEG_NUMCS] = {
   -1, -1, RGB_BLUE, -1, -1, -1, EXT_RGB_BLUE, EXT_RGBX_BLUE,
   EXT_BGR_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
-  EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE
+  EXT_RGBX_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE,
+  -1
 };
 
 static const int rgb_pixelsize[JPEG_NUMCS] = {
   -1, -1, RGB_PIXELSIZE, -1, -1, -1, EXT_RGB_PIXELSIZE, EXT_RGBX_PIXELSIZE,
   EXT_BGR_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
-  EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE
+  EXT_RGBX_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE,
+  -1
 };
 
 /* Definitions for speed-related optimizations. */
@@ -378,7 +402,7 @@
 
 #ifndef MULTIPLIER
 #ifndef WITH_SIMD
-#define MULTIPLIER  int		/* type for fastest integer multiply */
+#define MULTIPLIER  int         /* type for fastest integer multiply */
 #else
 #define MULTIPLIER short  /* prefer 16-bit with SIMD for parellelism */
 #endif
@@ -388,17 +412,10 @@
 /* FAST_FLOAT should be either float or double, whichever is done faster
  * by your compiler.  (Note that this type is only used in the floating point
  * DCT routines, so it only matters if you've defined DCT_FLOAT_SUPPORTED.)
- * Typically, float is faster in ANSI C compilers, while double is faster in
- * pre-ANSI compilers (because they insist on converting to double anyway).
- * The code below therefore chooses float if we have ANSI-style prototypes.
  */
 
 #ifndef FAST_FLOAT
-#ifdef HAVE_PROTOTYPES
 #define FAST_FLOAT  float
-#else
-#define FAST_FLOAT  double
-#endif
 #endif
 
 #endif /* JPEG_INTERNAL_OPTIONS */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegcomp.h glmark2-2021.02/src/libjpeg-turbo/jpegcomp.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegcomp.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpegcomp.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,8 @@
  * jpegcomp.h
  *
  * Copyright (C) 2010, D. R. Commander
- * For conditions of distribution and use, see the accompanying README file.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * JPEG compatibility macros
  * These declarations are considered internal to the JPEG library; most
@@ -11,6 +12,8 @@
 
 #if JPEG_LIB_VERSION >= 70
 #define _DCT_scaled_size DCT_h_scaled_size
+#define _DCT_h_scaled_size DCT_h_scaled_size
+#define _DCT_v_scaled_size DCT_v_scaled_size
 #define _min_DCT_scaled_size min_DCT_h_scaled_size
 #define _min_DCT_h_scaled_size min_DCT_h_scaled_size
 #define _min_DCT_v_scaled_size min_DCT_v_scaled_size
@@ -18,6 +21,8 @@
 #define _jpeg_height jpeg_height
 #else
 #define _DCT_scaled_size DCT_scaled_size
+#define _DCT_h_scaled_size DCT_scaled_size
+#define _DCT_v_scaled_size DCT_scaled_size
 #define _min_DCT_scaled_size min_DCT_scaled_size
 #define _min_DCT_h_scaled_size min_DCT_scaled_size
 #define _min_DCT_v_scaled_size min_DCT_scaled_size
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegint.h glmark2-2021.02/src/libjpeg-turbo/jpegint.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegint.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpegint.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,14 @@
 /*
  * jpegint.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * Modified 1997-2009 by Guido Vollbeding.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015-2016, D. R. Commander
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file provides common declarations for the various JPEG modules.
  * These declarations are considered internal to the JPEG library; most
@@ -14,121 +18,127 @@
 
 /* Declarations for both compression & decompression */
 
-typedef enum {			/* Operating modes for buffer controllers */
-	JBUF_PASS_THRU,		/* Plain stripwise operation */
-	/* Remaining modes require a full-image buffer to have been created */
-	JBUF_SAVE_SOURCE,	/* Run source subobject only, save output */
-	JBUF_CRANK_DEST,	/* Run dest subobject only, using saved data */
-	JBUF_SAVE_AND_PASS	/* Run both subobjects, save output */
+typedef enum {            /* Operating modes for buffer controllers */
+  JBUF_PASS_THRU,         /* Plain stripwise operation */
+  /* Remaining modes require a full-image buffer to have been created */
+  JBUF_SAVE_SOURCE,       /* Run source subobject only, save output */
+  JBUF_CRANK_DEST,        /* Run dest subobject only, using saved data */
+  JBUF_SAVE_AND_PASS      /* Run both subobjects, save output */
 } J_BUF_MODE;
 
 /* Values of global_state field (jdapi.c has some dependencies on ordering!) */
-#define CSTATE_START	100	/* after create_compress */
-#define CSTATE_SCANNING	101	/* start_compress done, write_scanlines OK */
-#define CSTATE_RAW_OK	102	/* start_compress done, write_raw_data OK */
-#define CSTATE_WRCOEFS	103	/* jpeg_write_coefficients done */
-#define DSTATE_START	200	/* after create_decompress */
-#define DSTATE_INHEADER	201	/* reading header markers, no SOS yet */
-#define DSTATE_READY	202	/* found SOS, ready for start_decompress */
-#define DSTATE_PRELOAD	203	/* reading multiscan file in start_decompress*/
-#define DSTATE_PRESCAN	204	/* performing dummy pass for 2-pass quant */
-#define DSTATE_SCANNING	205	/* start_decompress done, read_scanlines OK */
-#define DSTATE_RAW_OK	206	/* start_decompress done, read_raw_data OK */
-#define DSTATE_BUFIMAGE	207	/* expecting jpeg_start_output */
-#define DSTATE_BUFPOST	208	/* looking for SOS/EOI in jpeg_finish_output */
-#define DSTATE_RDCOEFS	209	/* reading file in jpeg_read_coefficients */
-#define DSTATE_STOPPING	210	/* looking for EOI in jpeg_finish_decompress */
+#define CSTATE_START    100     /* after create_compress */
+#define CSTATE_SCANNING 101     /* start_compress done, write_scanlines OK */
+#define CSTATE_RAW_OK   102     /* start_compress done, write_raw_data OK */
+#define CSTATE_WRCOEFS  103     /* jpeg_write_coefficients done */
+#define DSTATE_START    200     /* after create_decompress */
+#define DSTATE_INHEADER 201     /* reading header markers, no SOS yet */
+#define DSTATE_READY    202     /* found SOS, ready for start_decompress */
+#define DSTATE_PRELOAD  203     /* reading multiscan file in start_decompress*/
+#define DSTATE_PRESCAN  204     /* performing dummy pass for 2-pass quant */
+#define DSTATE_SCANNING 205     /* start_decompress done, read_scanlines OK */
+#define DSTATE_RAW_OK   206     /* start_decompress done, read_raw_data OK */
+#define DSTATE_BUFIMAGE 207     /* expecting jpeg_start_output */
+#define DSTATE_BUFPOST  208     /* looking for SOS/EOI in jpeg_finish_output */
+#define DSTATE_RDCOEFS  209     /* reading file in jpeg_read_coefficients */
+#define DSTATE_STOPPING 210     /* looking for EOI in jpeg_finish_decompress */
+
+
+/* JLONG must hold at least signed 32-bit values. */
+typedef long JLONG;
+
+
+/*
+ * Left shift macro that handles a negative operand without causing any
+ * sanitizer warnings
+ */
+
+#define LEFT_SHIFT(a, b) ((JLONG)((unsigned long)(a) << (b)))
 
 
 /* Declarations for compression modules */
 
 /* Master control module */
 struct jpeg_comp_master {
-  JMETHOD(void, prepare_for_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, pass_startup, (j_compress_ptr cinfo));
-  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+  void (*prepare_for_pass) (j_compress_ptr cinfo);
+  void (*pass_startup) (j_compress_ptr cinfo);
+  void (*finish_pass) (j_compress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean call_pass_startup;	/* True if pass_startup must be called */
-  boolean is_last_pass;		/* True during last pass */
+  boolean call_pass_startup;    /* True if pass_startup must be called */
+  boolean is_last_pass;         /* True during last pass */
 };
 
 /* Main buffer control (downsampled-data buffer) */
 struct jpeg_c_main_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, process_data, (j_compress_ptr cinfo,
-			       JSAMPARRAY input_buf, JDIMENSION *in_row_ctr,
-			       JDIMENSION in_rows_avail));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                        JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail);
 };
 
 /* Compression preprocessing (downsampling input buffer control) */
 struct jpeg_c_prep_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, pre_process_data, (j_compress_ptr cinfo,
-				   JSAMPARRAY input_buf,
-				   JDIMENSION *in_row_ctr,
-				   JDIMENSION in_rows_avail,
-				   JSAMPIMAGE output_buf,
-				   JDIMENSION *out_row_group_ctr,
-				   JDIMENSION out_row_groups_avail));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*pre_process_data) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                            JDIMENSION *in_row_ctr, JDIMENSION in_rows_avail,
+                            JSAMPIMAGE output_buf,
+                            JDIMENSION *out_row_group_ctr,
+                            JDIMENSION out_row_groups_avail);
 };
 
 /* Coefficient buffer control */
 struct jpeg_c_coef_controller {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(boolean, compress_data, (j_compress_ptr cinfo,
-				   JSAMPIMAGE input_buf));
+  void (*start_pass) (j_compress_ptr cinfo, J_BUF_MODE pass_mode);
+  boolean (*compress_data) (j_compress_ptr cinfo, JSAMPIMAGE input_buf);
 };
 
 /* Colorspace conversion */
 struct jpeg_color_converter {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, color_convert, (j_compress_ptr cinfo,
-				JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-				JDIMENSION output_row, int num_rows));
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*color_convert) (j_compress_ptr cinfo, JSAMPARRAY input_buf,
+                         JSAMPIMAGE output_buf, JDIMENSION output_row,
+                         int num_rows);
 };
 
 /* Downsampling */
 struct jpeg_downsampler {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
-  JMETHOD(void, downsample, (j_compress_ptr cinfo,
-			     JSAMPIMAGE input_buf, JDIMENSION in_row_index,
-			     JSAMPIMAGE output_buf,
-			     JDIMENSION out_row_group_index));
+  void (*start_pass) (j_compress_ptr cinfo);
+  void (*downsample) (j_compress_ptr cinfo, JSAMPIMAGE input_buf,
+                      JDIMENSION in_row_index, JSAMPIMAGE output_buf,
+                      JDIMENSION out_row_group_index);
 
-  boolean need_context_rows;	/* TRUE if need rows above & below */
+  boolean need_context_rows;    /* TRUE if need rows above & below */
 };
 
 /* Forward DCT (also controls coefficient quantization) */
 struct jpeg_forward_dct {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo));
+  void (*start_pass) (j_compress_ptr cinfo);
   /* perhaps this should be an array??? */
-  JMETHOD(void, forward_DCT, (j_compress_ptr cinfo,
-			      jpeg_component_info * compptr,
-			      JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
-			      JDIMENSION start_row, JDIMENSION start_col,
-			      JDIMENSION num_blocks));
+  void (*forward_DCT) (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY sample_data, JBLOCKROW coef_blocks,
+                       JDIMENSION start_row, JDIMENSION start_col,
+                       JDIMENSION num_blocks);
 };
 
 /* Entropy encoding */
 struct jpeg_entropy_encoder {
-  JMETHOD(void, start_pass, (j_compress_ptr cinfo, boolean gather_statistics));
-  JMETHOD(boolean, encode_mcu, (j_compress_ptr cinfo, JBLOCKROW *MCU_data));
-  JMETHOD(void, finish_pass, (j_compress_ptr cinfo));
+  void (*start_pass) (j_compress_ptr cinfo, boolean gather_statistics);
+  boolean (*encode_mcu) (j_compress_ptr cinfo, JBLOCKROW *MCU_data);
+  void (*finish_pass) (j_compress_ptr cinfo);
 };
 
 /* Marker writing */
 struct jpeg_marker_writer {
-  JMETHOD(void, write_file_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_frame_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_scan_header, (j_compress_ptr cinfo));
-  JMETHOD(void, write_file_trailer, (j_compress_ptr cinfo));
-  JMETHOD(void, write_tables_only, (j_compress_ptr cinfo));
+  void (*write_file_header) (j_compress_ptr cinfo);
+  void (*write_frame_header) (j_compress_ptr cinfo);
+  void (*write_scan_header) (j_compress_ptr cinfo);
+  void (*write_file_trailer) (j_compress_ptr cinfo);
+  void (*write_tables_only) (j_compress_ptr cinfo);
   /* These routines are exported to allow insertion of extra markers */
   /* Probably only COM and APPn markers should be written this way */
-  JMETHOD(void, write_marker_header, (j_compress_ptr cinfo, int marker,
-				      unsigned int datalen));
-  JMETHOD(void, write_marker_byte, (j_compress_ptr cinfo, int val));
+  void (*write_marker_header) (j_compress_ptr cinfo, int marker,
+                               unsigned int datalen);
+  void (*write_marker_byte) (j_compress_ptr cinfo, int val);
 };
 
 
@@ -136,138 +146,137 @@
 
 /* Master control module */
 struct jpeg_decomp_master {
-  JMETHOD(void, prepare_for_output_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, finish_output_pass, (j_decompress_ptr cinfo));
+  void (*prepare_for_output_pass) (j_decompress_ptr cinfo);
+  void (*finish_output_pass) (j_decompress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean is_dummy_pass;	/* True during 1st pass for 2-pass quant */
+  boolean is_dummy_pass;        /* True during 1st pass for 2-pass quant */
+
+  /* Partial decompression variables */
+  JDIMENSION first_iMCU_col;
+  JDIMENSION last_iMCU_col;
+  JDIMENSION first_MCU_col[MAX_COMPS_IN_SCAN];
+  JDIMENSION last_MCU_col[MAX_COMPS_IN_SCAN];
+  boolean jinit_upsampler_no_alloc;
 };
 
 /* Input control module */
 struct jpeg_input_controller {
-  JMETHOD(int, consume_input, (j_decompress_ptr cinfo));
-  JMETHOD(void, reset_input_controller, (j_decompress_ptr cinfo));
-  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, finish_input_pass, (j_decompress_ptr cinfo));
+  int (*consume_input) (j_decompress_ptr cinfo);
+  void (*reset_input_controller) (j_decompress_ptr cinfo);
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  void (*finish_input_pass) (j_decompress_ptr cinfo);
 
   /* State variables made visible to other modules */
-  boolean has_multiple_scans;	/* True if file has multiple scans */
-  boolean eoi_reached;		/* True when EOI has been consumed */
+  boolean has_multiple_scans;   /* True if file has multiple scans */
+  boolean eoi_reached;          /* True when EOI has been consumed */
 };
 
 /* Main buffer control (downsampled-data buffer) */
 struct jpeg_d_main_controller {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, process_data, (j_decompress_ptr cinfo,
-			       JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
-			       JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*process_data) (j_decompress_ptr cinfo, JSAMPARRAY output_buf,
+                        JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 };
 
 /* Coefficient buffer control */
 struct jpeg_d_coef_controller {
-  JMETHOD(void, start_input_pass, (j_decompress_ptr cinfo));
-  JMETHOD(int, consume_data, (j_decompress_ptr cinfo));
-  JMETHOD(void, start_output_pass, (j_decompress_ptr cinfo));
-  JMETHOD(int, decompress_data, (j_decompress_ptr cinfo,
-				 JSAMPIMAGE output_buf));
+  void (*start_input_pass) (j_decompress_ptr cinfo);
+  int (*consume_data) (j_decompress_ptr cinfo);
+  void (*start_output_pass) (j_decompress_ptr cinfo);
+  int (*decompress_data) (j_decompress_ptr cinfo, JSAMPIMAGE output_buf);
   /* Pointer to array of coefficient virtual arrays, or NULL if none */
   jvirt_barray_ptr *coef_arrays;
 };
 
 /* Decompression postprocessing (color quantization buffer control) */
 struct jpeg_d_post_controller {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, J_BUF_MODE pass_mode));
-  JMETHOD(void, post_process_data, (j_decompress_ptr cinfo,
-				    JSAMPIMAGE input_buf,
-				    JDIMENSION *in_row_group_ctr,
-				    JDIMENSION in_row_groups_avail,
-				    JSAMPARRAY output_buf,
-				    JDIMENSION *out_row_ctr,
-				    JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo, J_BUF_MODE pass_mode);
+  void (*post_process_data) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                             JDIMENSION *in_row_group_ctr,
+                             JDIMENSION in_row_groups_avail,
+                             JSAMPARRAY output_buf, JDIMENSION *out_row_ctr,
+                             JDIMENSION out_rows_avail);
 };
 
 /* Marker reading & parsing */
 struct jpeg_marker_reader {
-  JMETHOD(void, reset_marker_reader, (j_decompress_ptr cinfo));
+  void (*reset_marker_reader) (j_decompress_ptr cinfo);
   /* Read markers until SOS or EOI.
    * Returns same codes as are defined for jpeg_consume_input:
    * JPEG_SUSPENDED, JPEG_REACHED_SOS, or JPEG_REACHED_EOI.
    */
-  JMETHOD(int, read_markers, (j_decompress_ptr cinfo));
+  int (*read_markers) (j_decompress_ptr cinfo);
   /* Read a restart marker --- exported for use by entropy decoder only */
   jpeg_marker_parser_method read_restart_marker;
 
   /* State of marker reader --- nominally internal, but applications
    * supplying COM or APPn handlers might like to know the state.
    */
-  boolean saw_SOI;		/* found SOI? */
-  boolean saw_SOF;		/* found SOF? */
-  int next_restart_num;		/* next restart number expected (0-7) */
-  unsigned int discarded_bytes;	/* # of bytes skipped looking for a marker */
+  boolean saw_SOI;              /* found SOI? */
+  boolean saw_SOF;              /* found SOF? */
+  int next_restart_num;         /* next restart number expected (0-7) */
+  unsigned int discarded_bytes; /* # of bytes skipped looking for a marker */
 };
 
 /* Entropy decoding */
 struct jpeg_entropy_decoder {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(boolean, decode_mcu, (j_decompress_ptr cinfo,
-				JBLOCKROW *MCU_data));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  boolean (*decode_mcu) (j_decompress_ptr cinfo, JBLOCKROW *MCU_data);
 
   /* This is here to share code between baseline and progressive decoders; */
   /* other modules probably should not use it */
-  boolean insufficient_data;	/* set TRUE after emitting warning */
+  boolean insufficient_data;    /* set TRUE after emitting warning */
 };
 
 /* Inverse DCT (also performs dequantization) */
-typedef JMETHOD(void, inverse_DCT_method_ptr,
-		(j_decompress_ptr cinfo, jpeg_component_info * compptr,
-		 JCOEFPTR coef_block,
-		 JSAMPARRAY output_buf, JDIMENSION output_col));
+typedef void (*inverse_DCT_method_ptr) (j_decompress_ptr cinfo,
+                                        jpeg_component_info *compptr,
+                                        JCOEFPTR coef_block,
+                                        JSAMPARRAY output_buf,
+                                        JDIMENSION output_col);
 
 struct jpeg_inverse_dct {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
+  void (*start_pass) (j_decompress_ptr cinfo);
   /* It is useful to allow each component to have a separate IDCT method. */
   inverse_DCT_method_ptr inverse_DCT[MAX_COMPONENTS];
 };
 
 /* Upsampling (note that upsampler must also call color converter) */
 struct jpeg_upsampler {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, upsample, (j_decompress_ptr cinfo,
-			   JSAMPIMAGE input_buf,
-			   JDIMENSION *in_row_group_ctr,
-			   JDIMENSION in_row_groups_avail,
-			   JSAMPARRAY output_buf,
-			   JDIMENSION *out_row_ctr,
-			   JDIMENSION out_rows_avail));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*upsample) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                    JDIMENSION *in_row_group_ctr,
+                    JDIMENSION in_row_groups_avail, JSAMPARRAY output_buf,
+                    JDIMENSION *out_row_ctr, JDIMENSION out_rows_avail);
 
-  boolean need_context_rows;	/* TRUE if need rows above & below */
+  boolean need_context_rows;    /* TRUE if need rows above & below */
 };
 
 /* Colorspace conversion */
 struct jpeg_color_deconverter {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, color_convert, (j_decompress_ptr cinfo,
-				JSAMPIMAGE input_buf, JDIMENSION input_row,
-				JSAMPARRAY output_buf, int num_rows));
+  void (*start_pass) (j_decompress_ptr cinfo);
+  void (*color_convert) (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+                         JDIMENSION input_row, JSAMPARRAY output_buf,
+                         int num_rows);
 };
 
 /* Color quantization or color precision reduction */
 struct jpeg_color_quantizer {
-  JMETHOD(void, start_pass, (j_decompress_ptr cinfo, boolean is_pre_scan));
-  JMETHOD(void, color_quantize, (j_decompress_ptr cinfo,
-				 JSAMPARRAY input_buf, JSAMPARRAY output_buf,
-				 int num_rows));
-  JMETHOD(void, finish_pass, (j_decompress_ptr cinfo));
-  JMETHOD(void, new_color_map, (j_decompress_ptr cinfo));
+  void (*start_pass) (j_decompress_ptr cinfo, boolean is_pre_scan);
+  void (*color_quantize) (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
+                          JSAMPARRAY output_buf, int num_rows);
+  void (*finish_pass) (j_decompress_ptr cinfo);
+  void (*new_color_map) (j_decompress_ptr cinfo);
 };
 
 
 /* Miscellaneous useful macros */
 
 #undef MAX
-#define MAX(a,b)	((a) > (b) ? (a) : (b))
+#define MAX(a,b)        ((a) > (b) ? (a) : (b))
 #undef MIN
-#define MIN(a,b)	((a) < (b) ? (a) : (b))
+#define MIN(a,b)        ((a) < (b) ? (a) : (b))
 
 
 /* We assume that right shift corresponds to signed division by 2 with
@@ -275,126 +284,84 @@
  * shift" instructions that shift in copies of the sign bit.  But some
  * C compilers implement >> with an unsigned shift.  For these machines you
  * must define RIGHT_SHIFT_IS_UNSIGNED.
- * RIGHT_SHIFT provides a proper signed right shift of an INT32 quantity.
+ * RIGHT_SHIFT provides a proper signed right shift of a JLONG quantity.
  * It is only applied with constant shift counts.  SHIFT_TEMPS must be
  * included in the variables of any routine using RIGHT_SHIFT.
  */
 
 #ifdef RIGHT_SHIFT_IS_UNSIGNED
-#define SHIFT_TEMPS	INT32 shift_temp;
+#define SHIFT_TEMPS     JLONG shift_temp;
 #define RIGHT_SHIFT(x,shft)  \
-	((shift_temp = (x)) < 0 ? \
-	 (shift_temp >> (shft)) | ((~((INT32) 0)) << (32-(shft))) : \
-	 (shift_temp >> (shft)))
+        ((shift_temp = (x)) < 0 ? \
+         (shift_temp >> (shft)) | ((~((JLONG) 0)) << (32-(shft))) : \
+         (shift_temp >> (shft)))
 #else
 #define SHIFT_TEMPS
-#define RIGHT_SHIFT(x,shft)	((x) >> (shft))
+#define RIGHT_SHIFT(x,shft)     ((x) >> (shft))
 #endif
 
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jinit_compress_master	jICompress
-#define jinit_c_master_control	jICMaster
-#define jinit_c_main_controller	jICMainC
-#define jinit_c_prep_controller	jICPrepC
-#define jinit_c_coef_controller	jICCoefC
-#define jinit_color_converter	jICColor
-#define jinit_downsampler	jIDownsampler
-#define jinit_forward_dct	jIFDCT
-#define jinit_huff_encoder	jIHEncoder
-#define jinit_phuff_encoder	jIPHEncoder
-#define jinit_arith_encoder	jIAEncoder
-#define jinit_marker_writer	jIMWriter
-#define jinit_master_decompress	jIDMaster
-#define jinit_d_main_controller	jIDMainC
-#define jinit_d_coef_controller	jIDCoefC
-#define jinit_d_post_controller	jIDPostC
-#define jinit_input_controller	jIInCtlr
-#define jinit_marker_reader	jIMReader
-#define jinit_huff_decoder	jIHDecoder
-#define jinit_phuff_decoder	jIPHDecoder
-#define jinit_arith_decoder	jIADecoder
-#define jinit_inverse_dct	jIIDCT
-#define jinit_upsampler		jIUpsampler
-#define jinit_color_deconverter	jIDColor
-#define jinit_1pass_quantizer	jI1Quant
-#define jinit_2pass_quantizer	jI2Quant
-#define jinit_merged_upsampler	jIMUpsampler
-#define jinit_memory_mgr	jIMemMgr
-#define jdiv_round_up		jDivRound
-#define jround_up		jRound
-#define jcopy_sample_rows	jCopySamples
-#define jcopy_block_row		jCopyBlocks
-#define jzero_far		jZeroFar
-#define jpeg_zigzag_order	jZIGTable
-#define jpeg_natural_order	jZAGTable
-#define jpeg_aritab		jAriTab
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-
 /* Compression module initialization routines */
-EXTERN(void) jinit_compress_master JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_c_master_control JPP((j_compress_ptr cinfo,
-					 boolean transcode_only));
-EXTERN(void) jinit_c_main_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_c_prep_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_c_coef_controller JPP((j_compress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_color_converter JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_downsampler JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_forward_dct JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_huff_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_phuff_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_arith_encoder JPP((j_compress_ptr cinfo));
-EXTERN(void) jinit_marker_writer JPP((j_compress_ptr cinfo));
+EXTERN(void) jinit_compress_master (j_compress_ptr cinfo);
+EXTERN(void) jinit_c_master_control (j_compress_ptr cinfo,
+                                     boolean transcode_only);
+EXTERN(void) jinit_c_main_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_c_prep_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_c_coef_controller (j_compress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_color_converter (j_compress_ptr cinfo);
+EXTERN(void) jinit_downsampler (j_compress_ptr cinfo);
+EXTERN(void) jinit_forward_dct (j_compress_ptr cinfo);
+EXTERN(void) jinit_huff_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_phuff_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_arith_encoder (j_compress_ptr cinfo);
+EXTERN(void) jinit_marker_writer (j_compress_ptr cinfo);
 /* Decompression module initialization routines */
-EXTERN(void) jinit_master_decompress JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_d_main_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_d_coef_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_d_post_controller JPP((j_decompress_ptr cinfo,
-					  boolean need_full_buffer));
-EXTERN(void) jinit_input_controller JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_marker_reader JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_huff_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_phuff_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_arith_decoder JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_inverse_dct JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_upsampler JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_color_deconverter JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_1pass_quantizer JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_2pass_quantizer JPP((j_decompress_ptr cinfo));
-EXTERN(void) jinit_merged_upsampler JPP((j_decompress_ptr cinfo));
+EXTERN(void) jinit_master_decompress (j_decompress_ptr cinfo);
+EXTERN(void) jinit_d_main_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_d_coef_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_d_post_controller (j_decompress_ptr cinfo,
+                                      boolean need_full_buffer);
+EXTERN(void) jinit_input_controller (j_decompress_ptr cinfo);
+EXTERN(void) jinit_marker_reader (j_decompress_ptr cinfo);
+EXTERN(void) jinit_huff_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_phuff_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_arith_decoder (j_decompress_ptr cinfo);
+EXTERN(void) jinit_inverse_dct (j_decompress_ptr cinfo);
+EXTERN(void) jinit_upsampler (j_decompress_ptr cinfo);
+EXTERN(void) jinit_color_deconverter (j_decompress_ptr cinfo);
+EXTERN(void) jinit_1pass_quantizer (j_decompress_ptr cinfo);
+EXTERN(void) jinit_2pass_quantizer (j_decompress_ptr cinfo);
+EXTERN(void) jinit_merged_upsampler (j_decompress_ptr cinfo);
 /* Memory manager initialization */
-EXTERN(void) jinit_memory_mgr JPP((j_common_ptr cinfo));
+EXTERN(void) jinit_memory_mgr (j_common_ptr cinfo);
 
 /* Utility routines in jutils.c */
-EXTERN(long) jdiv_round_up JPP((long a, long b));
-EXTERN(long) jround_up JPP((long a, long b));
-EXTERN(void) jcopy_sample_rows JPP((JSAMPARRAY input_array, int source_row,
-				    JSAMPARRAY output_array, int dest_row,
-				    int num_rows, JDIMENSION num_cols));
-EXTERN(void) jcopy_block_row JPP((JBLOCKROW input_row, JBLOCKROW output_row,
-				  JDIMENSION num_blocks));
-EXTERN(void) jzero_far JPP((void FAR * target, size_t bytestozero));
+EXTERN(long) jdiv_round_up (long a, long b);
+EXTERN(long) jround_up (long a, long b);
+EXTERN(void) jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
+                                JSAMPARRAY output_array, int dest_row,
+                                int num_rows, JDIMENSION num_cols);
+EXTERN(void) jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
+                              JDIMENSION num_blocks);
+EXTERN(void) jzero_far (void *target, size_t bytestozero);
 /* Constant tables in jutils.c */
-#if 0				/* This table is not actually needed in v6a */
+#if 0                           /* This table is not actually needed in v6a */
 extern const int jpeg_zigzag_order[]; /* natural coef order to zigzag order */
 #endif
 extern const int jpeg_natural_order[]; /* zigzag coef order to natural order */
 
 /* Arithmetic coding probability estimation tables in jaricom.c */
-extern const INT32 jpeg_aritab[];
+extern const JLONG jpeg_aritab[];
 
 /* Suppress undefined-structure complaints if necessary. */
 
 #ifdef INCOMPLETE_TYPES_BROKEN
-#ifndef AM_MEMORY_MANAGER	/* only jmemmgr.c defines these */
+#ifndef AM_MEMORY_MANAGER       /* only jmemmgr.c defines these */
 struct jvirt_sarray_control { long dummy; };
 struct jvirt_barray_control { long dummy; };
 #endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeglib.h glmark2-2021.02/src/libjpeg-turbo/jpeglib.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeglib.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpeglib.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,14 @@
 /*
  * jpeglib.h
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1998, Thomas G. Lane.
  * Modified 2002-2009 by Guido Vollbeding.
- * Copyright (C) 2009-2011, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009-2011, 2013-2014, 2016, D. R. Commander.
+ * Copyright (C) 2015, Google, Inc.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file defines the application interface for the JPEG library.
  * Most applications using the library need only include this file,
@@ -15,6 +18,10 @@
 #ifndef JPEGLIB_H
 #define JPEGLIB_H
 
+/* Begin chromium edits */
+#include "jpeglibmangler.h"
+/* End chromium edits */
+
 /*
  * First we include the configuration files that record how this
  * installation of the JPEG library is set up.  jconfig.h can be
@@ -22,10 +29,10 @@
  * manual configuration options that most people need not worry about.
  */
 
-#ifndef JCONFIG_INCLUDED	/* in case jinclude.h already did */
-#include "jconfig.h"		/* widely used configuration options */
+#ifndef JCONFIG_INCLUDED        /* in case jinclude.h already did */
+#include "jconfig.h"            /* widely used configuration options */
 #endif
-#include "jmorecfg.h"		/* seldom changed options */
+#include "jmorecfg.h"           /* seldom changed options */
 
 
 #ifdef __cplusplus
@@ -40,13 +47,13 @@
  * if you want to be compatible.
  */
 
-#define DCTSIZE		    8	/* The basic DCT block is 8x8 samples */
-#define DCTSIZE2	    64	/* DCTSIZE squared; # of elements in a block */
-#define NUM_QUANT_TBLS      4	/* Quantization tables are numbered 0..3 */
-#define NUM_HUFF_TBLS       4	/* Huffman tables are numbered 0..3 */
-#define NUM_ARITH_TBLS      16	/* Arith-coding tables are numbered 0..15 */
-#define MAX_COMPS_IN_SCAN   4	/* JPEG limit on # of components in one scan */
-#define MAX_SAMP_FACTOR     4	/* JPEG limit on sampling factors */
+#define DCTSIZE             8   /* The basic DCT block is 8x8 samples */
+#define DCTSIZE2            64  /* DCTSIZE squared; # of elements in a block */
+#define NUM_QUANT_TBLS      4   /* Quantization tables are numbered 0..3 */
+#define NUM_HUFF_TBLS       4   /* Huffman tables are numbered 0..3 */
+#define NUM_ARITH_TBLS      16  /* Arith-coding tables are numbered 0..15 */
+#define MAX_COMPS_IN_SCAN   4   /* JPEG limit on # of components in one scan */
+#define MAX_SAMP_FACTOR     4   /* JPEG limit on sampling factors */
 /* Unfortunately, some bozo at Adobe saw no reason to be bound by the standard;
  * the PostScript DCT filter can emit files with many more than 10 blocks/MCU.
  * If you happen to run across such a file, you can up D_MAX_BLOCKS_IN_MCU
@@ -61,20 +68,18 @@
 
 
 /* Data structures for images (arrays of samples and of DCT coefficients).
- * On 80x86 machines, the image arrays are too big for near pointers,
- * but the pointer arrays can fit in near memory.
  */
 
-typedef JSAMPLE FAR *JSAMPROW;	/* ptr to one image row of pixel samples. */
-typedef JSAMPROW *JSAMPARRAY;	/* ptr to some rows (a 2-D sample array) */
-typedef JSAMPARRAY *JSAMPIMAGE;	/* a 3-D sample array: top index is color */
-
-typedef JCOEF JBLOCK[DCTSIZE2];	/* one block of coefficients */
-typedef JBLOCK FAR *JBLOCKROW;	/* pointer to one row of coefficient blocks */
-typedef JBLOCKROW *JBLOCKARRAY;		/* a 2-D array of coefficient blocks */
-typedef JBLOCKARRAY *JBLOCKIMAGE;	/* a 3-D array of coefficient blocks */
+typedef JSAMPLE *JSAMPROW;      /* ptr to one image row of pixel samples. */
+typedef JSAMPROW *JSAMPARRAY;   /* ptr to some rows (a 2-D sample array) */
+typedef JSAMPARRAY *JSAMPIMAGE; /* a 3-D sample array: top index is color */
+
+typedef JCOEF JBLOCK[DCTSIZE2]; /* one block of coefficients */
+typedef JBLOCK *JBLOCKROW;      /* pointer to one row of coefficient blocks */
+typedef JBLOCKROW *JBLOCKARRAY;         /* a 2-D array of coefficient blocks */
+typedef JBLOCKARRAY *JBLOCKIMAGE;       /* a 3-D array of coefficient blocks */
 
-typedef JCOEF FAR *JCOEFPTR;	/* useful in a couple of places */
+typedef JCOEF *JCOEFPTR;        /* useful in a couple of places */
 
 
 /* Types for JPEG compression parameters and working tables. */
@@ -87,13 +92,13 @@
    * (not the zigzag order in which they are stored in a JPEG DQT marker).
    * CAUTION: IJG versions prior to v6a kept this array in zigzag order.
    */
-  UINT16 quantval[DCTSIZE2];	/* quantization step for each coefficient */
+  UINT16 quantval[DCTSIZE2];    /* quantization step for each coefficient */
   /* This field is used only during compression.  It's initialized FALSE when
    * the table is created, and set TRUE when it's been output to the file.
    * You could suppress output of a table by setting this to TRUE.
    * (See jpeg_suppress_tables for an example.)
    */
-  boolean sent_table;		/* TRUE when table has been output */
+  boolean sent_table;           /* TRUE when table has been output */
 } JQUANT_TBL;
 
 
@@ -101,15 +106,15 @@
 
 typedef struct {
   /* These two fields directly represent the contents of a JPEG DHT marker */
-  UINT8 bits[17];		/* bits[k] = # of symbols with codes of */
-				/* length k bits; bits[0] is unused */
-  UINT8 huffval[256];		/* The symbols, in order of incr code length */
+  UINT8 bits[17];               /* bits[k] = # of symbols with codes of */
+                                /* length k bits; bits[0] is unused */
+  UINT8 huffval[256];           /* The symbols, in order of incr code length */
   /* This field is used only during compression.  It's initialized FALSE when
    * the table is created, and set TRUE when it's been output to the file.
    * You could suppress output of a table by setting this to TRUE.
    * (See jpeg_suppress_tables for an example.)
    */
-  boolean sent_table;		/* TRUE when table has been output */
+  boolean sent_table;           /* TRUE when table has been output */
 } JHUFF_TBL;
 
 
@@ -119,20 +124,20 @@
   /* These values are fixed over the whole image. */
   /* For compression, they must be supplied by parameter setup; */
   /* for decompression, they are read from the SOF marker. */
-  int component_id;		/* identifier for this component (0..255) */
-  int component_index;		/* its index in SOF or cinfo->comp_info[] */
-  int h_samp_factor;		/* horizontal sampling factor (1..4) */
-  int v_samp_factor;		/* vertical sampling factor (1..4) */
-  int quant_tbl_no;		/* quantization table selector (0..3) */
+  int component_id;             /* identifier for this component (0..255) */
+  int component_index;          /* its index in SOF or cinfo->comp_info[] */
+  int h_samp_factor;            /* horizontal sampling factor (1..4) */
+  int v_samp_factor;            /* vertical sampling factor (1..4) */
+  int quant_tbl_no;             /* quantization table selector (0..3) */
   /* These values may vary between scans. */
   /* For compression, they must be supplied by parameter setup; */
   /* for decompression, they are read from the SOS marker. */
   /* The decompressor output side may not use these variables. */
-  int dc_tbl_no;		/* DC entropy table selector (0..3) */
-  int ac_tbl_no;		/* AC entropy table selector (0..3) */
-  
+  int dc_tbl_no;                /* DC entropy table selector (0..3) */
+  int ac_tbl_no;                /* AC entropy table selector (0..3) */
+
   /* Remaining fields should be treated as private by applications. */
-  
+
   /* These values are computed during compression or decompression startup: */
   /* Component's size in DCT blocks.
    * Any dummy blocks added to complete an MCU are not counted; therefore
@@ -143,8 +148,8 @@
   /* Size of a DCT block in samples.  Always DCTSIZE for compression.
    * For decompression this is the size of the output from one DCT block,
    * reflecting any scaling we choose to apply during the IDCT step.
-   * Values of 1,2,4,8 are likely to be supported.  Note that different
-   * components may receive different IDCT scalings.
+   * Values from 1 to 16 are supported.
+   * Note that different components may receive different IDCT scalings.
    */
 #if JPEG_LIB_VERSION >= 70
   int DCT_h_scaled_size;
@@ -158,53 +163,53 @@
    * and similarly for height.  For decompression, IDCT scaling is included, so
    * downsampled_width = ceil(image_width * Hi/Hmax * DCT_[h_]scaled_size/DCTSIZE)
    */
-  JDIMENSION downsampled_width;	 /* actual width in samples */
+  JDIMENSION downsampled_width;  /* actual width in samples */
   JDIMENSION downsampled_height; /* actual height in samples */
   /* This flag is used only for decompression.  In cases where some of the
    * components will be ignored (eg grayscale output from YCbCr image),
    * we can skip most computations for the unused components.
    */
-  boolean component_needed;	/* do we need the value of this component? */
+  boolean component_needed;     /* do we need the value of this component? */
 
   /* These values are computed before starting a scan of the component. */
   /* The decompressor output side may not use these variables. */
-  int MCU_width;		/* number of blocks per MCU, horizontally */
-  int MCU_height;		/* number of blocks per MCU, vertically */
-  int MCU_blocks;		/* MCU_width * MCU_height */
-  int MCU_sample_width;		/* MCU width in samples, MCU_width*DCT_[h_]scaled_size */
-  int last_col_width;		/* # of non-dummy blocks across in last MCU */
-  int last_row_height;		/* # of non-dummy blocks down in last MCU */
+  int MCU_width;                /* number of blocks per MCU, horizontally */
+  int MCU_height;               /* number of blocks per MCU, vertically */
+  int MCU_blocks;               /* MCU_width * MCU_height */
+  int MCU_sample_width;         /* MCU width in samples, MCU_width*DCT_[h_]scaled_size */
+  int last_col_width;           /* # of non-dummy blocks across in last MCU */
+  int last_row_height;          /* # of non-dummy blocks down in last MCU */
 
   /* Saved quantization table for component; NULL if none yet saved.
    * See jdinput.c comments about the need for this information.
    * This field is currently used only for decompression.
    */
-  JQUANT_TBL * quant_table;
+  JQUANT_TBL *quant_table;
 
   /* Private per-component storage for DCT or IDCT subsystem. */
-  void * dct_table;
+  void *dct_table;
 } jpeg_component_info;
 
 
 /* The script for encoding a multiple-scan file is an array of these: */
 
 typedef struct {
-  int comps_in_scan;		/* number of components encoded in this scan */
+  int comps_in_scan;            /* number of components encoded in this scan */
   int component_index[MAX_COMPS_IN_SCAN]; /* their SOF/comp_info[] indexes */
-  int Ss, Se;			/* progressive JPEG spectral selection parms */
-  int Ah, Al;			/* progressive JPEG successive approx. parms */
+  int Ss, Se;                   /* progressive JPEG spectral selection parms */
+  int Ah, Al;                   /* progressive JPEG successive approx. parms */
 } jpeg_scan_info;
 
 /* The decompressor can save APPn and COM markers in a list of these: */
 
-typedef struct jpeg_marker_struct FAR * jpeg_saved_marker_ptr;
+typedef struct jpeg_marker_struct *jpeg_saved_marker_ptr;
 
 struct jpeg_marker_struct {
-  jpeg_saved_marker_ptr next;	/* next in list, or NULL */
-  UINT8 marker;			/* marker code: JPEG_COM, or JPEG_APP0+n */
-  unsigned int original_length;	/* # bytes of data in the file */
-  unsigned int data_length;	/* # bytes of data saved at data[] */
-  JOCTET FAR * data;		/* the data contained in the marker */
+  jpeg_saved_marker_ptr next;   /* next in list, or NULL */
+  UINT8 marker;                 /* marker code: JPEG_COM, or JPEG_APP0+n */
+  unsigned int original_length; /* # bytes of data in the file */
+  unsigned int data_length;     /* # bytes of data saved at data[] */
+  JOCTET *data;                 /* the data contained in the marker */
   /* the marker length word is not counted in data_length or original_length */
 };
 
@@ -214,102 +219,102 @@
 #define JCS_ALPHA_EXTENSIONS 1
 
 typedef enum {
-	JCS_UNKNOWN,		/* error/unspecified */
-	JCS_GRAYSCALE,		/* monochrome */
-	JCS_RGB,		/* red/green/blue as specified by the RGB_RED, RGB_GREEN,
-				   RGB_BLUE, and RGB_PIXELSIZE macros */
-	JCS_YCbCr,		/* Y/Cb/Cr (also known as YUV) */
-	JCS_CMYK,		/* C/M/Y/K */
-	JCS_YCCK,		/* Y/Cb/Cr/K */
-	JCS_EXT_RGB,		/* red/green/blue */
-	JCS_EXT_RGBX,		/* red/green/blue/x */
-	JCS_EXT_BGR,		/* blue/green/red */
-	JCS_EXT_BGRX,		/* blue/green/red/x */
-	JCS_EXT_XBGR,		/* x/blue/green/red */
-	JCS_EXT_XRGB,		/* x/red/green/blue */
-	/* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX,
-	   JCS_EXT_XBGR, or JCS_EXT_XRGB during decompression, the X byte is
-	   undefined, and in order to ensure the best performance,
-	   libjpeg-turbo can set that byte to whatever value it wishes.  Use
-	   the following colorspace constants to ensure that the X byte is set
-	   to 0xFF, so that it can be interpreted as an opaque alpha
-	   channel. */
-	JCS_EXT_RGBA,		/* red/green/blue/alpha */
-	JCS_EXT_BGRA,		/* blue/green/red/alpha */
-	JCS_EXT_ABGR,		/* alpha/blue/green/red */
-	JCS_EXT_ARGB		/* alpha/red/green/blue */
+  JCS_UNKNOWN,            /* error/unspecified */
+  JCS_GRAYSCALE,          /* monochrome */
+  JCS_RGB,                /* red/green/blue as specified by the RGB_RED,
+                             RGB_GREEN, RGB_BLUE, and RGB_PIXELSIZE macros */
+  JCS_YCbCr,              /* Y/Cb/Cr (also known as YUV) */
+  JCS_CMYK,               /* C/M/Y/K */
+  JCS_YCCK,               /* Y/Cb/Cr/K */
+  JCS_EXT_RGB,            /* red/green/blue */
+  JCS_EXT_RGBX,           /* red/green/blue/x */
+  JCS_EXT_BGR,            /* blue/green/red */
+  JCS_EXT_BGRX,           /* blue/green/red/x */
+  JCS_EXT_XBGR,           /* x/blue/green/red */
+  JCS_EXT_XRGB,           /* x/red/green/blue */
+  /* When out_color_space it set to JCS_EXT_RGBX, JCS_EXT_BGRX, JCS_EXT_XBGR,
+     or JCS_EXT_XRGB during decompression, the X byte is undefined, and in
+     order to ensure the best performance, libjpeg-turbo can set that byte to
+     whatever value it wishes.  Use the following colorspace constants to
+     ensure that the X byte is set to 0xFF, so that it can be interpreted as an
+     opaque alpha channel. */
+  JCS_EXT_RGBA,           /* red/green/blue/alpha */
+  JCS_EXT_BGRA,           /* blue/green/red/alpha */
+  JCS_EXT_ABGR,           /* alpha/blue/green/red */
+  JCS_EXT_ARGB,           /* alpha/red/green/blue */
+  JCS_RGB565              /* 5-bit red/6-bit green/5-bit blue */
 } J_COLOR_SPACE;
 
 /* DCT/IDCT algorithm options. */
 
 typedef enum {
-	JDCT_ISLOW,		/* slow but accurate integer algorithm */
-	JDCT_IFAST,		/* faster, less accurate integer method */
-	JDCT_FLOAT		/* floating-point: accurate, fast on fast HW */
+  JDCT_ISLOW,             /* slow but accurate integer algorithm */
+  JDCT_IFAST,             /* faster, less accurate integer method */
+  JDCT_FLOAT              /* floating-point: accurate, fast on fast HW */
 } J_DCT_METHOD;
 
-#ifndef JDCT_DEFAULT		/* may be overridden in jconfig.h */
+#ifndef JDCT_DEFAULT            /* may be overridden in jconfig.h */
 #define JDCT_DEFAULT  JDCT_ISLOW
 #endif
-#ifndef JDCT_FASTEST		/* may be overridden in jconfig.h */
+#ifndef JDCT_FASTEST            /* may be overridden in jconfig.h */
 #define JDCT_FASTEST  JDCT_IFAST
 #endif
 
 /* Dithering options for decompression. */
 
 typedef enum {
-	JDITHER_NONE,		/* no dithering */
-	JDITHER_ORDERED,	/* simple ordered dither */
-	JDITHER_FS		/* Floyd-Steinberg error diffusion dither */
+  JDITHER_NONE,           /* no dithering */
+  JDITHER_ORDERED,        /* simple ordered dither */
+  JDITHER_FS              /* Floyd-Steinberg error diffusion dither */
 } J_DITHER_MODE;
 
 
 /* Common fields between JPEG compression and decompression master structs. */
 
 #define jpeg_common_fields \
-  struct jpeg_error_mgr * err;	/* Error handler module */\
-  struct jpeg_memory_mgr * mem;	/* Memory manager module */\
-  struct jpeg_progress_mgr * progress; /* Progress monitor, or NULL if none */\
-  void * client_data;		/* Available for use by application */\
-  boolean is_decompressor;	/* So common code can tell which is which */\
-  int global_state		/* For checking call sequence validity */
+  struct jpeg_error_mgr *err;   /* Error handler module */\
+  struct jpeg_memory_mgr *mem;  /* Memory manager module */\
+  struct jpeg_progress_mgr *progress; /* Progress monitor, or NULL if none */\
+  void *client_data;            /* Available for use by application */\
+  boolean is_decompressor;      /* So common code can tell which is which */\
+  int global_state              /* For checking call sequence validity */
 
 /* Routines that are to be used by both halves of the library are declared
  * to receive a pointer to this structure.  There are no actual instances of
  * jpeg_common_struct, only of jpeg_compress_struct and jpeg_decompress_struct.
  */
 struct jpeg_common_struct {
-  jpeg_common_fields;		/* Fields common to both master struct types */
+  jpeg_common_fields;           /* Fields common to both master struct types */
   /* Additional fields follow in an actual jpeg_compress_struct or
    * jpeg_decompress_struct.  All three structs must agree on these
    * initial fields!  (This would be a lot cleaner in C++.)
    */
 };
 
-typedef struct jpeg_common_struct * j_common_ptr;
-typedef struct jpeg_compress_struct * j_compress_ptr;
-typedef struct jpeg_decompress_struct * j_decompress_ptr;
+typedef struct jpeg_common_struct *j_common_ptr;
+typedef struct jpeg_compress_struct *j_compress_ptr;
+typedef struct jpeg_decompress_struct *j_decompress_ptr;
 
 
 /* Master record for a compression instance */
 
 struct jpeg_compress_struct {
-  jpeg_common_fields;		/* Fields shared with jpeg_decompress_struct */
+  jpeg_common_fields;           /* Fields shared with jpeg_decompress_struct */
 
   /* Destination for compressed data */
-  struct jpeg_destination_mgr * dest;
+  struct jpeg_destination_mgr *dest;
 
   /* Description of source image --- these fields must be filled in by
    * outer application before starting compression.  in_color_space must
    * be correct before you can even call jpeg_set_defaults().
    */
 
-  JDIMENSION image_width;	/* input image width */
-  JDIMENSION image_height;	/* input image height */
-  int input_components;		/* # of color components in input image */
-  J_COLOR_SPACE in_color_space;	/* colorspace of input image */
+  JDIMENSION image_width;       /* input image width */
+  JDIMENSION image_height;      /* input image height */
+  int input_components;         /* # of color components in input image */
+  J_COLOR_SPACE in_color_space; /* colorspace of input image */
 
-  double input_gamma;		/* image gamma of input image */
+  double input_gamma;           /* image gamma of input image */
 
   /* Compression parameters --- these fields must be set before calling
    * jpeg_start_compress().  We recommend calling jpeg_set_defaults() to
@@ -322,8 +327,8 @@
 #if JPEG_LIB_VERSION >= 70
   unsigned int scale_num, scale_denom; /* fraction by which to scale image */
 
-  JDIMENSION jpeg_width;	/* scaled JPEG image width */
-  JDIMENSION jpeg_height;	/* scaled JPEG image height */
+  JDIMENSION jpeg_width;        /* scaled JPEG image width */
+  JDIMENSION jpeg_height;       /* scaled JPEG image height */
   /* Dimensions of actual JPEG image that will be written to file,
    * derived from input dimensions by scaling factors above.
    * These fields are computed by jpeg_start_compress().
@@ -332,15 +337,15 @@
    */
 #endif
 
-  int data_precision;		/* bits of precision in image data */
+  int data_precision;           /* bits of precision in image data */
 
-  int num_components;		/* # of color components in JPEG image */
+  int num_components;           /* # of color components in JPEG image */
   J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
 
-  jpeg_component_info * comp_info;
+  jpeg_component_info *comp_info;
   /* comp_info[i] describes component that appears i'th in SOF */
 
-  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS];
 #if JPEG_LIB_VERSION >= 70
   int q_scale_factor[NUM_QUANT_TBLS];
 #endif
@@ -348,30 +353,30 @@
    * and corresponding scale factors (percentage, initialized 100).
    */
 
-  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
-  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
   /* ptrs to Huffman coding tables, or NULL if not defined */
 
   UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
   UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
   UINT8 arith_ac_K[NUM_ARITH_TBLS]; /* Kx values for AC arith-coding tables */
 
-  int num_scans;		/* # of entries in scan_info array */
-  const jpeg_scan_info * scan_info; /* script for multi-scan file, or NULL */
+  int num_scans;                /* # of entries in scan_info array */
+  const jpeg_scan_info *scan_info; /* script for multi-scan file, or NULL */
   /* The default value of scan_info is NULL, which causes a single-scan
    * sequential JPEG file to be emitted.  To create a multi-scan file,
    * set num_scans and scan_info to point to an array of scan definitions.
    */
 
-  boolean raw_data_in;		/* TRUE=caller supplies downsampled data */
-  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
-  boolean optimize_coding;	/* TRUE=optimize entropy encoding parms */
-  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  boolean raw_data_in;          /* TRUE=caller supplies downsampled data */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean optimize_coding;      /* TRUE=optimize entropy encoding parms */
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
 #if JPEG_LIB_VERSION >= 70
   boolean do_fancy_downsampling; /* TRUE=apply fancy downsampling */
 #endif
-  int smoothing_factor;		/* 1..100, or 0 for no input smoothing */
-  J_DCT_METHOD dct_method;	/* DCT algorithm selector */
+  int smoothing_factor;         /* 1..100, or 0 for no input smoothing */
+  J_DCT_METHOD dct_method;      /* DCT algorithm selector */
 
   /* The restart interval can be specified in absolute MCUs by setting
    * restart_interval, or in MCU rows by setting restart_in_rows
@@ -379,28 +384,28 @@
    * for each scan).
    */
   unsigned int restart_interval; /* MCUs per restart, or 0 for no restart */
-  int restart_in_rows;		/* if > 0, MCU rows per restart interval */
+  int restart_in_rows;          /* if > 0, MCU rows per restart interval */
 
   /* Parameters controlling emission of special markers. */
 
-  boolean write_JFIF_header;	/* should a JFIF marker be written? */
-  UINT8 JFIF_major_version;	/* What to write for the JFIF version number */
+  boolean write_JFIF_header;    /* should a JFIF marker be written? */
+  UINT8 JFIF_major_version;     /* What to write for the JFIF version number */
   UINT8 JFIF_minor_version;
   /* These three values are not used by the JPEG code, merely copied */
   /* into the JFIF APP0 marker.  density_unit can be 0 for unknown, */
   /* 1 for dots/inch, or 2 for dots/cm.  Note that the pixel aspect */
   /* ratio is defined by X_density/Y_density even when density_unit=0. */
-  UINT8 density_unit;		/* JFIF code for pixel size units */
-  UINT16 X_density;		/* Horizontal pixel density */
-  UINT16 Y_density;		/* Vertical pixel density */
-  boolean write_Adobe_marker;	/* should an Adobe marker be written? */
-  
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean write_Adobe_marker;   /* should an Adobe marker be written? */
+
   /* State variable: index of next scanline to be written to
    * jpeg_write_scanlines().  Application may use this to control its
    * processing loop, e.g., "while (next_scanline < image_height)".
    */
 
-  JDIMENSION next_scanline;	/* 0 .. image_height-1  */
+  JDIMENSION next_scanline;     /* 0 .. image_height-1  */
 
   /* Remaining fields are known throughout compressor, but generally
    * should not be touched by a surrounding application.
@@ -409,59 +414,59 @@
   /*
    * These fields are computed during compression startup
    */
-  boolean progressive_mode;	/* TRUE if scan script uses progressive mode */
-  int max_h_samp_factor;	/* largest h_samp_factor */
-  int max_v_samp_factor;	/* largest v_samp_factor */
+  boolean progressive_mode;     /* TRUE if scan script uses progressive mode */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
 
 #if JPEG_LIB_VERSION >= 70
-  int min_DCT_h_scaled_size;	/* smallest DCT_h_scaled_size of any component */
-  int min_DCT_v_scaled_size;	/* smallest DCT_v_scaled_size of any component */
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
 #endif
 
-  JDIMENSION total_iMCU_rows;	/* # of iMCU rows to be input to coef ctlr */
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows to be input to coef ctlr */
   /* The coefficient controller receives data in units of MCU rows as defined
    * for fully interleaved scans (whether the JPEG file is interleaved or not).
    * There are v_samp_factor * DCTSIZE sample rows of each component in an
    * "iMCU" (interleaved MCU) row.
    */
-  
+
   /*
    * These fields are valid during any one scan.
    * They describe the components and MCUs actually appearing in the scan.
    */
-  int comps_in_scan;		/* # of JPEG components in this scan */
-  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_info[MAX_COMPS_IN_SCAN];
   /* *cur_comp_info[i] describes component that appears i'th in SOS */
-  
-  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
-  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
-  
-  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
+
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
   int MCU_membership[C_MAX_BLOCKS_IN_MCU];
   /* MCU_membership[i] is index in cur_comp_info of component owning */
   /* i'th block in an MCU */
 
-  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
 
 #if JPEG_LIB_VERSION >= 80
-  int block_size;		/* the basic DCT block size: 1..16 */
-  const int * natural_order;	/* natural-order position array */
-  int lim_Se;			/* min( Se, DCTSIZE2-1 ) */
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order;     /* natural-order position array */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) */
 #endif
 
   /*
    * Links to compression subobjects (methods and private variables of modules)
    */
-  struct jpeg_comp_master * master;
-  struct jpeg_c_main_controller * main;
-  struct jpeg_c_prep_controller * prep;
-  struct jpeg_c_coef_controller * coef;
-  struct jpeg_marker_writer * marker;
-  struct jpeg_color_converter * cconvert;
-  struct jpeg_downsampler * downsample;
-  struct jpeg_forward_dct * fdct;
-  struct jpeg_entropy_encoder * entropy;
-  jpeg_scan_info * script_space; /* workspace for jpeg_simple_progression */
+  struct jpeg_comp_master *master;
+  struct jpeg_c_main_controller *main;
+  struct jpeg_c_prep_controller *prep;
+  struct jpeg_c_coef_controller *coef;
+  struct jpeg_marker_writer *marker;
+  struct jpeg_color_converter *cconvert;
+  struct jpeg_downsampler *downsample;
+  struct jpeg_forward_dct *fdct;
+  struct jpeg_entropy_encoder *entropy;
+  jpeg_scan_info *script_space; /* workspace for jpeg_simple_progression */
   int script_space_size;
 };
 
@@ -469,17 +474,17 @@
 /* Master record for a decompression instance */
 
 struct jpeg_decompress_struct {
-  jpeg_common_fields;		/* Fields shared with jpeg_compress_struct */
+  jpeg_common_fields;           /* Fields shared with jpeg_compress_struct */
 
   /* Source of compressed data */
-  struct jpeg_source_mgr * src;
+  struct jpeg_source_mgr *src;
 
   /* Basic description of image --- filled in by jpeg_read_header(). */
   /* Application may inspect these values to decide how to process image. */
 
-  JDIMENSION image_width;	/* nominal image width (from SOF marker) */
-  JDIMENSION image_height;	/* nominal image height */
-  int num_components;		/* # of color components in JPEG image */
+  JDIMENSION image_width;       /* nominal image width (from SOF marker) */
+  JDIMENSION image_height;      /* nominal image height */
+  int num_components;           /* # of color components in JPEG image */
   J_COLOR_SPACE jpeg_color_space; /* colorspace of JPEG image */
 
   /* Decompression processing parameters --- these fields must be set before
@@ -491,24 +496,24 @@
 
   unsigned int scale_num, scale_denom; /* fraction by which to scale image */
 
-  double output_gamma;		/* image gamma wanted in output */
+  double output_gamma;          /* image gamma wanted in output */
 
-  boolean buffered_image;	/* TRUE=multiple output passes */
-  boolean raw_data_out;		/* TRUE=downsampled data wanted */
+  boolean buffered_image;       /* TRUE=multiple output passes */
+  boolean raw_data_out;         /* TRUE=downsampled data wanted */
 
-  J_DCT_METHOD dct_method;	/* IDCT algorithm selector */
-  boolean do_fancy_upsampling;	/* TRUE=apply fancy upsampling */
-  boolean do_block_smoothing;	/* TRUE=apply interblock smoothing */
+  J_DCT_METHOD dct_method;      /* IDCT algorithm selector */
+  boolean do_fancy_upsampling;  /* TRUE=apply fancy upsampling */
+  boolean do_block_smoothing;   /* TRUE=apply interblock smoothing */
 
-  boolean quantize_colors;	/* TRUE=colormapped output wanted */
+  boolean quantize_colors;      /* TRUE=colormapped output wanted */
   /* the following are ignored if not quantize_colors: */
-  J_DITHER_MODE dither_mode;	/* type of color dithering to use */
-  boolean two_pass_quantize;	/* TRUE=use two-pass color quantization */
-  int desired_number_of_colors;	/* max # colors to use in created colormap */
+  J_DITHER_MODE dither_mode;    /* type of color dithering to use */
+  boolean two_pass_quantize;    /* TRUE=use two-pass color quantization */
+  int desired_number_of_colors; /* max # colors to use in created colormap */
   /* these are significant only in buffered-image mode: */
-  boolean enable_1pass_quant;	/* enable future use of 1-pass quantizer */
+  boolean enable_1pass_quant;   /* enable future use of 1-pass quantizer */
   boolean enable_external_quant;/* enable future use of external colormap */
-  boolean enable_2pass_quant;	/* enable future use of 2-pass quantizer */
+  boolean enable_2pass_quant;   /* enable future use of 2-pass quantizer */
 
   /* Description of actual output image that will be returned to application.
    * These fields are computed by jpeg_start_decompress().
@@ -516,14 +521,14 @@
    * in advance of calling jpeg_start_decompress().
    */
 
-  JDIMENSION output_width;	/* scaled image width */
-  JDIMENSION output_height;	/* scaled image height */
-  int out_color_components;	/* # of color components in out_color_space */
-  int output_components;	/* # of color components returned */
+  JDIMENSION output_width;      /* scaled image width */
+  JDIMENSION output_height;     /* scaled image height */
+  int out_color_components;     /* # of color components in out_color_space */
+  int output_components;        /* # of color components returned */
   /* output_components is 1 (a colormap index) when quantizing colors;
    * otherwise it equals out_color_components.
    */
-  int rec_outbuf_height;	/* min recommended height of scanline buffer */
+  int rec_outbuf_height;        /* min recommended height of scanline buffer */
   /* If the buffer passed to jpeg_read_scanlines() is less than this many rows
    * high, space and time will be wasted due to unnecessary data copying.
    * Usually rec_outbuf_height will be 1 or 2, at most 4.
@@ -535,8 +540,8 @@
    * jpeg_start_decompress or jpeg_start_output.
    * The map has out_color_components rows and actual_number_of_colors columns.
    */
-  int actual_number_of_colors;	/* number of entries in use */
-  JSAMPARRAY colormap;		/* The color map as a 2-D pixel array */
+  int actual_number_of_colors;  /* number of entries in use */
+  JSAMPARRAY colormap;          /* The color map as a 2-D pixel array */
 
   /* State variables: these variables indicate the progress of decompression.
    * The application may examine these but must not modify them.
@@ -546,20 +551,20 @@
    * Application may use this to control its processing loop, e.g.,
    * "while (output_scanline < output_height)".
    */
-  JDIMENSION output_scanline;	/* 0 .. output_height-1  */
+  JDIMENSION output_scanline;   /* 0 .. output_height-1  */
 
   /* Current input scan number and number of iMCU rows completed in scan.
    * These indicate the progress of the decompressor input side.
    */
-  int input_scan_number;	/* Number of SOS markers seen so far */
-  JDIMENSION input_iMCU_row;	/* Number of iMCU rows completed */
+  int input_scan_number;        /* Number of SOS markers seen so far */
+  JDIMENSION input_iMCU_row;    /* Number of iMCU rows completed */
 
   /* The "output scan number" is the notional scan being displayed by the
    * output side.  The decompressor will not allow output scan/row number
    * to get ahead of input scan/row, but it can fall arbitrarily far behind.
    */
-  int output_scan_number;	/* Nominal scan number being displayed */
-  JDIMENSION output_iMCU_row;	/* Number of iMCU rows read */
+  int output_scan_number;       /* Nominal scan number being displayed */
+  JDIMENSION output_iMCU_row;   /* Number of iMCU rows read */
 
   /* Current progression status.  coef_bits[c][i] indicates the precision
    * with which component c's DCT coefficient i (in zigzag order) is known.
@@ -568,7 +573,7 @@
    * (thus, 0 at completion of the progression).
    * This pointer is NULL when reading a non-progressive file.
    */
-  int (*coef_bits)[DCTSIZE2];	/* -1 or current Al value for each coef */
+  int (*coef_bits)[DCTSIZE2];   /* -1 or current Al value for each coef */
 
   /* Internal JPEG parameters --- the application usually need not look at
    * these fields.  Note that the decompressor output side may not use
@@ -579,27 +584,27 @@
    * datastreams when processing abbreviated JPEG datastreams.
    */
 
-  JQUANT_TBL * quant_tbl_ptrs[NUM_QUANT_TBLS];
+  JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS];
   /* ptrs to coefficient quantization tables, or NULL if not defined */
 
-  JHUFF_TBL * dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
-  JHUFF_TBL * ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS];
+  JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS];
   /* ptrs to Huffman coding tables, or NULL if not defined */
 
   /* These parameters are never carried across datastreams, since they
    * are given in SOF/SOS markers or defined to be reset by SOI.
    */
 
-  int data_precision;		/* bits of precision in image data */
+  int data_precision;           /* bits of precision in image data */
 
-  jpeg_component_info * comp_info;
+  jpeg_component_info *comp_info;
   /* comp_info[i] describes component that appears i'th in SOF */
 
 #if JPEG_LIB_VERSION >= 80
-  boolean is_baseline;		/* TRUE if Baseline SOF0 encountered */
+  boolean is_baseline;          /* TRUE if Baseline SOF0 encountered */
 #endif
-  boolean progressive_mode;	/* TRUE if SOFn specifies progressive mode */
-  boolean arith_code;		/* TRUE=arithmetic coding, FALSE=Huffman */
+  boolean progressive_mode;     /* TRUE if SOFn specifies progressive mode */
+  boolean arith_code;           /* TRUE=arithmetic coding, FALSE=Huffman */
 
   UINT8 arith_dc_L[NUM_ARITH_TBLS]; /* L values for DC arith-coding tables */
   UINT8 arith_dc_U[NUM_ARITH_TBLS]; /* U values for DC arith-coding tables */
@@ -610,17 +615,17 @@
   /* These fields record data obtained from optional markers recognized by
    * the JPEG library.
    */
-  boolean saw_JFIF_marker;	/* TRUE iff a JFIF APP0 marker was found */
+  boolean saw_JFIF_marker;      /* TRUE iff a JFIF APP0 marker was found */
   /* Data copied from JFIF marker; only valid if saw_JFIF_marker is TRUE: */
-  UINT8 JFIF_major_version;	/* JFIF version number */
+  UINT8 JFIF_major_version;     /* JFIF version number */
   UINT8 JFIF_minor_version;
-  UINT8 density_unit;		/* JFIF code for pixel size units */
-  UINT16 X_density;		/* Horizontal pixel density */
-  UINT16 Y_density;		/* Vertical pixel density */
-  boolean saw_Adobe_marker;	/* TRUE iff an Adobe APP14 marker was found */
-  UINT8 Adobe_transform;	/* Color transform code from Adobe marker */
+  UINT8 density_unit;           /* JFIF code for pixel size units */
+  UINT16 X_density;             /* Horizontal pixel density */
+  UINT16 Y_density;             /* Vertical pixel density */
+  boolean saw_Adobe_marker;     /* TRUE iff an Adobe APP14 marker was found */
+  UINT8 Adobe_transform;        /* Color transform code from Adobe marker */
 
-  boolean CCIR601_sampling;	/* TRUE=first samples are cosited */
+  boolean CCIR601_sampling;     /* TRUE=first samples are cosited */
 
   /* Aside from the specific data retained from APPn markers known to the
    * library, the uninterpreted contents of any or all APPn and COM markers
@@ -635,17 +640,17 @@
   /*
    * These fields are computed during decompression startup
    */
-  int max_h_samp_factor;	/* largest h_samp_factor */
-  int max_v_samp_factor;	/* largest v_samp_factor */
+  int max_h_samp_factor;        /* largest h_samp_factor */
+  int max_v_samp_factor;        /* largest v_samp_factor */
 
 #if JPEG_LIB_VERSION >= 70
-  int min_DCT_h_scaled_size;	/* smallest DCT_h_scaled_size of any component */
-  int min_DCT_v_scaled_size;	/* smallest DCT_v_scaled_size of any component */
+  int min_DCT_h_scaled_size;    /* smallest DCT_h_scaled_size of any component */
+  int min_DCT_v_scaled_size;    /* smallest DCT_v_scaled_size of any component */
 #else
-  int min_DCT_scaled_size;	/* smallest DCT_scaled_size of any component */
+  int min_DCT_scaled_size;      /* smallest DCT_scaled_size of any component */
 #endif
 
-  JDIMENSION total_iMCU_rows;	/* # of iMCU rows in image */
+  JDIMENSION total_iMCU_rows;   /* # of iMCU rows in image */
   /* The coefficient controller's input and output progress is measured in
    * units of "iMCU" (interleaved MCU) rows.  These are the same as MCU rows
    * in fully interleaved JPEG scans, but are used whether the scan is
@@ -654,33 +659,33 @@
    * v_samp_factor*DCT_[v_]scaled_size sample rows of a component per iMCU row.
    */
 
-  JSAMPLE * sample_range_limit; /* table for fast range-limiting */
+  JSAMPLE *sample_range_limit;  /* table for fast range-limiting */
 
   /*
    * These fields are valid during any one scan.
    * They describe the components and MCUs actually appearing in the scan.
    * Note that the decompressor output side must not use these fields.
    */
-  int comps_in_scan;		/* # of JPEG components in this scan */
-  jpeg_component_info * cur_comp_info[MAX_COMPS_IN_SCAN];
+  int comps_in_scan;            /* # of JPEG components in this scan */
+  jpeg_component_info *cur_comp_info[MAX_COMPS_IN_SCAN];
   /* *cur_comp_info[i] describes component that appears i'th in SOS */
 
-  JDIMENSION MCUs_per_row;	/* # of MCUs across the image */
-  JDIMENSION MCU_rows_in_scan;	/* # of MCU rows in the image */
+  JDIMENSION MCUs_per_row;      /* # of MCUs across the image */
+  JDIMENSION MCU_rows_in_scan;  /* # of MCU rows in the image */
 
-  int blocks_in_MCU;		/* # of DCT blocks per MCU */
+  int blocks_in_MCU;            /* # of DCT blocks per MCU */
   int MCU_membership[D_MAX_BLOCKS_IN_MCU];
   /* MCU_membership[i] is index in cur_comp_info of component owning */
   /* i'th block in an MCU */
 
-  int Ss, Se, Ah, Al;		/* progressive JPEG parameters for scan */
+  int Ss, Se, Ah, Al;           /* progressive JPEG parameters for scan */
 
 #if JPEG_LIB_VERSION >= 80
   /* These fields are derived from Se of first SOS marker.
    */
-  int block_size;		/* the basic DCT block size: 1..16 */
-  const int * natural_order; /* natural-order position array for entropy decode */
-  int lim_Se;			/* min( Se, DCTSIZE2-1 ) for entropy decode */
+  int block_size;               /* the basic DCT block size: 1..16 */
+  const int *natural_order; /* natural-order position array for entropy decode */
+  int lim_Se;                   /* min( Se, DCTSIZE2-1 ) for entropy decode */
 #endif
 
   /* This field is shared between entropy decoder and marker parser.
@@ -692,17 +697,17 @@
   /*
    * Links to decompression subobjects (methods, private variables of modules)
    */
-  struct jpeg_decomp_master * master;
-  struct jpeg_d_main_controller * main;
-  struct jpeg_d_coef_controller * coef;
-  struct jpeg_d_post_controller * post;
-  struct jpeg_input_controller * inputctl;
-  struct jpeg_marker_reader * marker;
-  struct jpeg_entropy_decoder * entropy;
-  struct jpeg_inverse_dct * idct;
-  struct jpeg_upsampler * upsample;
-  struct jpeg_color_deconverter * cconvert;
-  struct jpeg_color_quantizer * cquantize;
+  struct jpeg_decomp_master *master;
+  struct jpeg_d_main_controller *main;
+  struct jpeg_d_coef_controller *coef;
+  struct jpeg_d_post_controller *post;
+  struct jpeg_input_controller *inputctl;
+  struct jpeg_marker_reader *marker;
+  struct jpeg_entropy_decoder *entropy;
+  struct jpeg_inverse_dct *idct;
+  struct jpeg_upsampler *upsample;
+  struct jpeg_color_deconverter *cconvert;
+  struct jpeg_color_quantizer *cquantize;
 };
 
 
@@ -718,17 +723,17 @@
 
 struct jpeg_error_mgr {
   /* Error exit handler: does not return to caller */
-  JMETHOD(void, error_exit, (j_common_ptr cinfo));
+  void (*error_exit) (j_common_ptr cinfo);
   /* Conditionally emit a trace or warning message */
-  JMETHOD(void, emit_message, (j_common_ptr cinfo, int msg_level));
+  void (*emit_message) (j_common_ptr cinfo, int msg_level);
   /* Routine that actually outputs a trace or error message */
-  JMETHOD(void, output_message, (j_common_ptr cinfo));
+  void (*output_message) (j_common_ptr cinfo);
   /* Format a message string for the most recent JPEG error or message */
-  JMETHOD(void, format_message, (j_common_ptr cinfo, char * buffer));
-#define JMSG_LENGTH_MAX  200	/* recommended size of format_message buffer */
+  void (*format_message) (j_common_ptr cinfo, char *buffer);
+#define JMSG_LENGTH_MAX  200    /* recommended size of format_message buffer */
   /* Reset error state variables at start of a new image */
-  JMETHOD(void, reset_error_mgr, (j_common_ptr cinfo));
-  
+  void (*reset_error_mgr) (j_common_ptr cinfo);
+
   /* The message ID code and any parameters are saved here.
    * A message can have one string parameter or up to 8 int parameters.
    */
@@ -738,18 +743,18 @@
     int i[8];
     char s[JMSG_STR_PARM_MAX];
   } msg_parm;
-  
+
   /* Standard state variables for error facility */
-  
-  int trace_level;		/* max msg_level that will be displayed */
-  
+
+  int trace_level;              /* max msg_level that will be displayed */
+
   /* For recoverable corrupt-data errors, we emit a warning message,
    * but keep going unless emit_message chooses to abort.  emit_message
    * should count warnings in num_warnings.  The surrounding application
    * can check for bad data by seeing if num_warnings is nonzero at the
    * end of processing.
    */
-  long num_warnings;		/* number of corrupt-data warnings */
+  long num_warnings;            /* number of corrupt-data warnings */
 
   /* These fields point to the table(s) of error message strings.
    * An application can change the table pointer to switch to a different
@@ -761,52 +766,52 @@
    * First table includes all errors generated by JPEG library itself.
    * Error code 0 is reserved for a "no such error string" message.
    */
-  const char * const * jpeg_message_table; /* Library errors */
+  const char * const *jpeg_message_table; /* Library errors */
   int last_jpeg_message;    /* Table contains strings 0..last_jpeg_message */
   /* Second table can be added by application (see cjpeg/djpeg for example).
    * It contains strings numbered first_addon_message..last_addon_message.
    */
-  const char * const * addon_message_table; /* Non-library errors */
-  int first_addon_message;	/* code for first string in addon table */
-  int last_addon_message;	/* code for last string in addon table */
+  const char * const *addon_message_table; /* Non-library errors */
+  int first_addon_message;      /* code for first string in addon table */
+  int last_addon_message;       /* code for last string in addon table */
 };
 
 
 /* Progress monitor object */
 
 struct jpeg_progress_mgr {
-  JMETHOD(void, progress_monitor, (j_common_ptr cinfo));
+  void (*progress_monitor) (j_common_ptr cinfo);
 
-  long pass_counter;		/* work units completed in this pass */
-  long pass_limit;		/* total number of work units in this pass */
-  int completed_passes;		/* passes completed so far */
-  int total_passes;		/* total number of passes expected */
+  long pass_counter;            /* work units completed in this pass */
+  long pass_limit;              /* total number of work units in this pass */
+  int completed_passes;         /* passes completed so far */
+  int total_passes;             /* total number of passes expected */
 };
 
 
 /* Data destination object for compression */
 
 struct jpeg_destination_mgr {
-  JOCTET * next_output_byte;	/* => next byte to write in buffer */
-  size_t free_in_buffer;	/* # of byte spaces remaining in buffer */
+  JOCTET *next_output_byte;     /* => next byte to write in buffer */
+  size_t free_in_buffer;        /* # of byte spaces remaining in buffer */
 
-  JMETHOD(void, init_destination, (j_compress_ptr cinfo));
-  JMETHOD(boolean, empty_output_buffer, (j_compress_ptr cinfo));
-  JMETHOD(void, term_destination, (j_compress_ptr cinfo));
+  void (*init_destination) (j_compress_ptr cinfo);
+  boolean (*empty_output_buffer) (j_compress_ptr cinfo);
+  void (*term_destination) (j_compress_ptr cinfo);
 };
 
 
 /* Data source object for decompression */
 
 struct jpeg_source_mgr {
-  const JOCTET * next_input_byte; /* => next byte to read from buffer */
-  size_t bytes_in_buffer;	/* # of bytes remaining in buffer */
+  const JOCTET *next_input_byte; /* => next byte to read from buffer */
+  size_t bytes_in_buffer;       /* # of bytes remaining in buffer */
 
-  JMETHOD(void, init_source, (j_decompress_ptr cinfo));
-  JMETHOD(boolean, fill_input_buffer, (j_decompress_ptr cinfo));
-  JMETHOD(void, skip_input_data, (j_decompress_ptr cinfo, long num_bytes));
-  JMETHOD(boolean, resync_to_restart, (j_decompress_ptr cinfo, int desired));
-  JMETHOD(void, term_source, (j_decompress_ptr cinfo));
+  void (*init_source) (j_decompress_ptr cinfo);
+  boolean (*fill_input_buffer) (j_decompress_ptr cinfo);
+  void (*skip_input_data) (j_decompress_ptr cinfo, long num_bytes);
+  boolean (*resync_to_restart) (j_decompress_ptr cinfo, int desired);
+  void (*term_source) (j_decompress_ptr cinfo);
 };
 
 
@@ -821,51 +826,42 @@
  * successful.
  */
 
-#define JPOOL_PERMANENT	0	/* lasts until master record is destroyed */
-#define JPOOL_IMAGE	1	/* lasts until done with image/datastream */
-#define JPOOL_NUMPOOLS	2
+#define JPOOL_PERMANENT 0       /* lasts until master record is destroyed */
+#define JPOOL_IMAGE     1       /* lasts until done with image/datastream */
+#define JPOOL_NUMPOOLS  2
 
-typedef struct jvirt_sarray_control * jvirt_sarray_ptr;
-typedef struct jvirt_barray_control * jvirt_barray_ptr;
+typedef struct jvirt_sarray_control *jvirt_sarray_ptr;
+typedef struct jvirt_barray_control *jvirt_barray_ptr;
 
 
 struct jpeg_memory_mgr {
   /* Method pointers */
-  JMETHOD(void *, alloc_small, (j_common_ptr cinfo, int pool_id,
-				size_t sizeofobject));
-  JMETHOD(void FAR *, alloc_large, (j_common_ptr cinfo, int pool_id,
-				     size_t sizeofobject));
-  JMETHOD(JSAMPARRAY, alloc_sarray, (j_common_ptr cinfo, int pool_id,
-				     JDIMENSION samplesperrow,
-				     JDIMENSION numrows));
-  JMETHOD(JBLOCKARRAY, alloc_barray, (j_common_ptr cinfo, int pool_id,
-				      JDIMENSION blocksperrow,
-				      JDIMENSION numrows));
-  JMETHOD(jvirt_sarray_ptr, request_virt_sarray, (j_common_ptr cinfo,
-						  int pool_id,
-						  boolean pre_zero,
-						  JDIMENSION samplesperrow,
-						  JDIMENSION numrows,
-						  JDIMENSION maxaccess));
-  JMETHOD(jvirt_barray_ptr, request_virt_barray, (j_common_ptr cinfo,
-						  int pool_id,
-						  boolean pre_zero,
-						  JDIMENSION blocksperrow,
-						  JDIMENSION numrows,
-						  JDIMENSION maxaccess));
-  JMETHOD(void, realize_virt_arrays, (j_common_ptr cinfo));
-  JMETHOD(JSAMPARRAY, access_virt_sarray, (j_common_ptr cinfo,
-					   jvirt_sarray_ptr ptr,
-					   JDIMENSION start_row,
-					   JDIMENSION num_rows,
-					   boolean writable));
-  JMETHOD(JBLOCKARRAY, access_virt_barray, (j_common_ptr cinfo,
-					    jvirt_barray_ptr ptr,
-					    JDIMENSION start_row,
-					    JDIMENSION num_rows,
-					    boolean writable));
-  JMETHOD(void, free_pool, (j_common_ptr cinfo, int pool_id));
-  JMETHOD(void, self_destruct, (j_common_ptr cinfo));
+  void *(*alloc_small) (j_common_ptr cinfo, int pool_id, size_t sizeofobject);
+  void *(*alloc_large) (j_common_ptr cinfo, int pool_id,
+                        size_t sizeofobject);
+  JSAMPARRAY (*alloc_sarray) (j_common_ptr cinfo, int pool_id,
+                              JDIMENSION samplesperrow, JDIMENSION numrows);
+  JBLOCKARRAY (*alloc_barray) (j_common_ptr cinfo, int pool_id,
+                               JDIMENSION blocksperrow, JDIMENSION numrows);
+  jvirt_sarray_ptr (*request_virt_sarray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION samplesperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  jvirt_barray_ptr (*request_virt_barray) (j_common_ptr cinfo, int pool_id,
+                                           boolean pre_zero,
+                                           JDIMENSION blocksperrow,
+                                           JDIMENSION numrows,
+                                           JDIMENSION maxaccess);
+  void (*realize_virt_arrays) (j_common_ptr cinfo);
+  JSAMPARRAY (*access_virt_sarray) (j_common_ptr cinfo, jvirt_sarray_ptr ptr,
+                                    JDIMENSION start_row, JDIMENSION num_rows,
+                                    boolean writable);
+  JBLOCKARRAY (*access_virt_barray) (j_common_ptr cinfo, jvirt_barray_ptr ptr,
+                                     JDIMENSION start_row, JDIMENSION num_rows,
+                                     boolean writable);
+  void (*free_pool) (j_common_ptr cinfo, int pool_id);
+  void (*self_destruct) (j_common_ptr cinfo);
 
   /* Limit on memory allocation for this JPEG object.  (Note that this is
    * merely advisory, not a guaranteed maximum; it only affects the space
@@ -882,96 +878,21 @@
 /* Routine signature for application-supplied marker processing methods.
  * Need not pass marker code since it is stored in cinfo->unread_marker.
  */
-typedef JMETHOD(boolean, jpeg_marker_parser_method, (j_decompress_ptr cinfo));
-
-
-/* Declarations for routines called by application.
- * The JPP macro hides prototype parameters from compilers that can't cope.
- * Note JPP requires double parentheses.
- */
-
-#ifdef HAVE_PROTOTYPES
-#define JPP(arglist)	arglist
-#else
-#define JPP(arglist)	()
-#endif
+typedef boolean (*jpeg_marker_parser_method) (j_decompress_ptr cinfo);
 
 
-/* Short forms of external names for systems with brain-damaged linkers.
- * We shorten external names to be unique in the first six letters, which
- * is good enough for all known systems.
- * (If your compiler itself needs names to be unique in less than 15 
- * characters, you are out of luck.  Get a better compiler.)
+/* Originally, this macro was used as a way of defining function prototypes
+ * for both modern compilers as well as older compilers that did not support
+ * prototype parameters.  libjpeg-turbo has never supported these older,
+ * non-ANSI compilers, but the macro is still included because there is some
+ * software out there that uses it.
  */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_std_error		jStdError
-#define jpeg_CreateCompress	jCreaCompress
-#define jpeg_CreateDecompress	jCreaDecompress
-#define jpeg_destroy_compress	jDestCompress
-#define jpeg_destroy_decompress	jDestDecompress
-#define jpeg_stdio_dest		jStdDest
-#define jpeg_stdio_src		jStdSrc
-#if JPEG_LIB_VERSION >= 80
-#define jpeg_mem_dest		jMemDest
-#define jpeg_mem_src		jMemSrc
-#endif
-#define jpeg_set_defaults	jSetDefaults
-#define jpeg_set_colorspace	jSetColorspace
-#define jpeg_default_colorspace	jDefColorspace
-#define jpeg_set_quality	jSetQuality
-#define jpeg_set_linear_quality	jSetLQuality
-#if JPEG_LIB_VERSION >= 70
-#define jpeg_default_qtables	jDefQTables
-#endif
-#define jpeg_add_quant_table	jAddQuantTable
-#define jpeg_quality_scaling	jQualityScaling
-#define jpeg_simple_progression	jSimProgress
-#define jpeg_suppress_tables	jSuppressTables
-#define jpeg_alloc_quant_table	jAlcQTable
-#define jpeg_alloc_huff_table	jAlcHTable
-#define jpeg_start_compress	jStrtCompress
-#define jpeg_write_scanlines	jWrtScanlines
-#define jpeg_finish_compress	jFinCompress
-#if JPEG_LIB_VERSION >= 70
-#define jpeg_calc_jpeg_dimensions	jCjpegDimensions
-#endif
-#define jpeg_write_raw_data	jWrtRawData
-#define jpeg_write_marker	jWrtMarker
-#define jpeg_write_m_header	jWrtMHeader
-#define jpeg_write_m_byte	jWrtMByte
-#define jpeg_write_tables	jWrtTables
-#define jpeg_read_header	jReadHeader
-#define jpeg_start_decompress	jStrtDecompress
-#define jpeg_read_scanlines	jReadScanlines
-#define jpeg_finish_decompress	jFinDecompress
-#define jpeg_read_raw_data	jReadRawData
-#define jpeg_has_multiple_scans	jHasMultScn
-#define jpeg_start_output	jStrtOutput
-#define jpeg_finish_output	jFinOutput
-#define jpeg_input_complete	jInComplete
-#define jpeg_new_colormap	jNewCMap
-#define jpeg_consume_input	jConsumeInput
-#if JPEG_LIB_VERSION >= 80
-#define jpeg_core_output_dimensions	jCoreDimensions
-#endif
-#define jpeg_calc_output_dimensions	jCalcDimensions
-#define jpeg_save_markers	jSaveMarkers
-#define jpeg_set_marker_processor	jSetMarker
-#define jpeg_read_coefficients	jReadCoefs
-#define jpeg_write_coefficients	jWrtCoefs
-#define jpeg_copy_critical_parameters	jCopyCrit
-#define jpeg_abort_compress	jAbrtCompress
-#define jpeg_abort_decompress	jAbrtDecompress
-#define jpeg_abort		jAbort
-#define jpeg_destroy		jDestroy
-#define jpeg_resync_to_restart	jResyncRestart
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#define JPP(arglist)    arglist
 
 
 /* Default error-management setup */
-EXTERN(struct jpeg_error_mgr *) jpeg_std_error
-	JPP((struct jpeg_error_mgr * err));
+EXTERN(struct jpeg_error_mgr *) jpeg_std_error (struct jpeg_error_mgr *err);
 
 /* Initialization of JPEG compression objects.
  * jpeg_create_compress() and jpeg_create_decompress() are the exported
@@ -982,97 +903,89 @@
  */
 #define jpeg_create_compress(cinfo) \
     jpeg_CreateCompress((cinfo), JPEG_LIB_VERSION, \
-			(size_t) sizeof(struct jpeg_compress_struct))
+                        (size_t) sizeof(struct jpeg_compress_struct))
 #define jpeg_create_decompress(cinfo) \
     jpeg_CreateDecompress((cinfo), JPEG_LIB_VERSION, \
-			  (size_t) sizeof(struct jpeg_decompress_struct))
-EXTERN(void) jpeg_CreateCompress JPP((j_compress_ptr cinfo,
-				      int version, size_t structsize));
-EXTERN(void) jpeg_CreateDecompress JPP((j_decompress_ptr cinfo,
-					int version, size_t structsize));
+                          (size_t) sizeof(struct jpeg_decompress_struct))
+EXTERN(void) jpeg_CreateCompress (j_compress_ptr cinfo, int version,
+                                  size_t structsize);
+EXTERN(void) jpeg_CreateDecompress (j_decompress_ptr cinfo, int version,
+                                    size_t structsize);
 /* Destruction of JPEG compression objects */
-EXTERN(void) jpeg_destroy_compress JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_destroy_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_destroy_compress (j_compress_ptr cinfo);
+EXTERN(void) jpeg_destroy_decompress (j_decompress_ptr cinfo);
 
 /* Standard data source and destination managers: stdio streams. */
 /* Caller is responsible for opening the file before and closing after. */
-EXTERN(void) jpeg_stdio_dest JPP((j_compress_ptr cinfo, FILE * outfile));
-EXTERN(void) jpeg_stdio_src JPP((j_decompress_ptr cinfo, FILE * infile));
+EXTERN(void) jpeg_stdio_dest (j_compress_ptr cinfo, FILE *outfile);
+EXTERN(void) jpeg_stdio_src (j_decompress_ptr cinfo, FILE *infile);
 
-#if JPEG_LIB_VERSION >= 80
+#if JPEG_LIB_VERSION >= 80 || defined(MEM_SRCDST_SUPPORTED)
 /* Data source and destination managers: memory buffers. */
-EXTERN(void) jpeg_mem_dest JPP((j_compress_ptr cinfo,
-			       unsigned char ** outbuffer,
-			       unsigned long * outsize));
-EXTERN(void) jpeg_mem_src JPP((j_decompress_ptr cinfo,
-			      unsigned char * inbuffer,
-			      unsigned long insize));
+EXTERN(void) jpeg_mem_dest (j_compress_ptr cinfo, unsigned char **outbuffer,
+                            unsigned long *outsize);
+EXTERN(void) jpeg_mem_src (j_decompress_ptr cinfo,
+                           const unsigned char *inbuffer,
+                           unsigned long insize);
 #endif
 
 /* Default parameter setup for compression */
-EXTERN(void) jpeg_set_defaults JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_set_defaults (j_compress_ptr cinfo);
 /* Compression parameter setup aids */
-EXTERN(void) jpeg_set_colorspace JPP((j_compress_ptr cinfo,
-				      J_COLOR_SPACE colorspace));
-EXTERN(void) jpeg_default_colorspace JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_set_quality JPP((j_compress_ptr cinfo, int quality,
-				   boolean force_baseline));
-EXTERN(void) jpeg_set_linear_quality JPP((j_compress_ptr cinfo,
-					  int scale_factor,
-					  boolean force_baseline));
+EXTERN(void) jpeg_set_colorspace (j_compress_ptr cinfo,
+                                  J_COLOR_SPACE colorspace);
+EXTERN(void) jpeg_default_colorspace (j_compress_ptr cinfo);
+EXTERN(void) jpeg_set_quality (j_compress_ptr cinfo, int quality,
+                               boolean force_baseline);
+EXTERN(void) jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+                                      boolean force_baseline);
 #if JPEG_LIB_VERSION >= 70
-EXTERN(void) jpeg_default_qtables JPP((j_compress_ptr cinfo,
-				       boolean force_baseline));
+EXTERN(void) jpeg_default_qtables (j_compress_ptr cinfo,
+                                   boolean force_baseline);
 #endif
-EXTERN(void) jpeg_add_quant_table JPP((j_compress_ptr cinfo, int which_tbl,
-				       const unsigned int *basic_table,
-				       int scale_factor,
-				       boolean force_baseline));
-EXTERN(int) jpeg_quality_scaling JPP((int quality));
-EXTERN(void) jpeg_simple_progression JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_suppress_tables JPP((j_compress_ptr cinfo,
-				       boolean suppress));
-EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table JPP((j_common_ptr cinfo));
-EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+                                   const unsigned int *basic_table,
+                                   int scale_factor, boolean force_baseline);
+EXTERN(int) jpeg_quality_scaling (int quality);
+EXTERN(void) jpeg_simple_progression (j_compress_ptr cinfo);
+EXTERN(void) jpeg_suppress_tables (j_compress_ptr cinfo, boolean suppress);
+EXTERN(JQUANT_TBL *) jpeg_alloc_quant_table (j_common_ptr cinfo);
+EXTERN(JHUFF_TBL *) jpeg_alloc_huff_table (j_common_ptr cinfo);
 
 /* Main entry points for compression */
-EXTERN(void) jpeg_start_compress JPP((j_compress_ptr cinfo,
-				      boolean write_all_tables));
-EXTERN(JDIMENSION) jpeg_write_scanlines JPP((j_compress_ptr cinfo,
-					     JSAMPARRAY scanlines,
-					     JDIMENSION num_lines));
-EXTERN(void) jpeg_finish_compress JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_start_compress (j_compress_ptr cinfo,
+                                  boolean write_all_tables);
+EXTERN(JDIMENSION) jpeg_write_scanlines (j_compress_ptr cinfo,
+                                         JSAMPARRAY scanlines,
+                                         JDIMENSION num_lines);
+EXTERN(void) jpeg_finish_compress (j_compress_ptr cinfo);
 
 #if JPEG_LIB_VERSION >= 70
 /* Precalculate JPEG dimensions for current compression parameters. */
-EXTERN(void) jpeg_calc_jpeg_dimensions JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_calc_jpeg_dimensions (j_compress_ptr cinfo);
 #endif
 
 /* Replaces jpeg_write_scanlines when writing raw downsampled data. */
-EXTERN(JDIMENSION) jpeg_write_raw_data JPP((j_compress_ptr cinfo,
-					    JSAMPIMAGE data,
-					    JDIMENSION num_lines));
+EXTERN(JDIMENSION) jpeg_write_raw_data (j_compress_ptr cinfo, JSAMPIMAGE data,
+                                        JDIMENSION num_lines);
 
 /* Write a special marker.  See libjpeg.txt concerning safe usage. */
-EXTERN(void) jpeg_write_marker
-	JPP((j_compress_ptr cinfo, int marker,
-	     const JOCTET * dataptr, unsigned int datalen));
+EXTERN(void) jpeg_write_marker (j_compress_ptr cinfo, int marker,
+                                const JOCTET *dataptr, unsigned int datalen);
 /* Same, but piecemeal. */
-EXTERN(void) jpeg_write_m_header
-	JPP((j_compress_ptr cinfo, int marker, unsigned int datalen));
-EXTERN(void) jpeg_write_m_byte
-	JPP((j_compress_ptr cinfo, int val));
+EXTERN(void) jpeg_write_m_header (j_compress_ptr cinfo, int marker,
+                                  unsigned int datalen);
+EXTERN(void) jpeg_write_m_byte (j_compress_ptr cinfo, int val);
 
 /* Alternate compression function: just write an abbreviated table file */
-EXTERN(void) jpeg_write_tables JPP((j_compress_ptr cinfo));
+EXTERN(void) jpeg_write_tables (j_compress_ptr cinfo);
 
 /* Decompression startup: read start of JPEG datastream to see what's there */
-EXTERN(int) jpeg_read_header JPP((j_decompress_ptr cinfo,
-				  boolean require_image));
+EXTERN(int) jpeg_read_header (j_decompress_ptr cinfo, boolean require_image);
 /* Return value is one of: */
-#define JPEG_SUSPENDED		0 /* Suspended due to lack of input data */
-#define JPEG_HEADER_OK		1 /* Found valid image datastream */
-#define JPEG_HEADER_TABLES_ONLY	2 /* Found valid table-specs-only datastream */
+#define JPEG_SUSPENDED          0 /* Suspended due to lack of input data */
+#define JPEG_HEADER_OK          1 /* Found valid image datastream */
+#define JPEG_HEADER_TABLES_ONLY 2 /* Found valid table-specs-only datastream */
 /* If you pass require_image = TRUE (normal case), you need not check for
  * a TABLES_ONLY return code; an abbreviated file will cause an error exit.
  * JPEG_SUSPENDED is only possible if you use a data source module that can
@@ -1080,54 +993,55 @@
  */
 
 /* Main entry points for decompression */
-EXTERN(boolean) jpeg_start_decompress JPP((j_decompress_ptr cinfo));
-EXTERN(JDIMENSION) jpeg_read_scanlines JPP((j_decompress_ptr cinfo,
-					    JSAMPARRAY scanlines,
-					    JDIMENSION max_lines));
-EXTERN(boolean) jpeg_finish_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_start_decompress (j_decompress_ptr cinfo);
+EXTERN(JDIMENSION) jpeg_read_scanlines (j_decompress_ptr cinfo,
+                                        JSAMPARRAY scanlines,
+                                        JDIMENSION max_lines);
+EXTERN(JDIMENSION) jpeg_skip_scanlines (j_decompress_ptr cinfo,
+                                        JDIMENSION num_lines);
+EXTERN(void) jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                                 JDIMENSION *width);
+EXTERN(boolean) jpeg_finish_decompress (j_decompress_ptr cinfo);
 
 /* Replaces jpeg_read_scanlines when reading raw downsampled data. */
-EXTERN(JDIMENSION) jpeg_read_raw_data JPP((j_decompress_ptr cinfo,
-					   JSAMPIMAGE data,
-					   JDIMENSION max_lines));
+EXTERN(JDIMENSION) jpeg_read_raw_data (j_decompress_ptr cinfo, JSAMPIMAGE data,
+                                       JDIMENSION max_lines);
 
 /* Additional entry points for buffered-image mode. */
-EXTERN(boolean) jpeg_has_multiple_scans JPP((j_decompress_ptr cinfo));
-EXTERN(boolean) jpeg_start_output JPP((j_decompress_ptr cinfo,
-				       int scan_number));
-EXTERN(boolean) jpeg_finish_output JPP((j_decompress_ptr cinfo));
-EXTERN(boolean) jpeg_input_complete JPP((j_decompress_ptr cinfo));
-EXTERN(void) jpeg_new_colormap JPP((j_decompress_ptr cinfo));
-EXTERN(int) jpeg_consume_input JPP((j_decompress_ptr cinfo));
+EXTERN(boolean) jpeg_has_multiple_scans (j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_start_output (j_decompress_ptr cinfo, int scan_number);
+EXTERN(boolean) jpeg_finish_output (j_decompress_ptr cinfo);
+EXTERN(boolean) jpeg_input_complete (j_decompress_ptr cinfo);
+EXTERN(void) jpeg_new_colormap (j_decompress_ptr cinfo);
+EXTERN(int) jpeg_consume_input (j_decompress_ptr cinfo);
 /* Return value is one of: */
-/* #define JPEG_SUSPENDED	0    Suspended due to lack of input data */
-#define JPEG_REACHED_SOS	1 /* Reached start of new scan */
-#define JPEG_REACHED_EOI	2 /* Reached end of image */
-#define JPEG_ROW_COMPLETED	3 /* Completed one iMCU row */
-#define JPEG_SCAN_COMPLETED	4 /* Completed last iMCU row of a scan */
+/* #define JPEG_SUSPENDED       0    Suspended due to lack of input data */
+#define JPEG_REACHED_SOS        1 /* Reached start of new scan */
+#define JPEG_REACHED_EOI        2 /* Reached end of image */
+#define JPEG_ROW_COMPLETED      3 /* Completed one iMCU row */
+#define JPEG_SCAN_COMPLETED     4 /* Completed last iMCU row of a scan */
 
 /* Precalculate output dimensions for current decompression parameters. */
 #if JPEG_LIB_VERSION >= 80
-EXTERN(void) jpeg_core_output_dimensions JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_core_output_dimensions (j_decompress_ptr cinfo);
 #endif
-EXTERN(void) jpeg_calc_output_dimensions JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_calc_output_dimensions (j_decompress_ptr cinfo);
 
 /* Control saving of COM and APPn markers into marker_list. */
-EXTERN(void) jpeg_save_markers
-	JPP((j_decompress_ptr cinfo, int marker_code,
-	     unsigned int length_limit));
+EXTERN(void) jpeg_save_markers (j_decompress_ptr cinfo, int marker_code,
+                                unsigned int length_limit);
 
 /* Install a special processing method for COM or APPn markers. */
-EXTERN(void) jpeg_set_marker_processor
-	JPP((j_decompress_ptr cinfo, int marker_code,
-	     jpeg_marker_parser_method routine));
+EXTERN(void) jpeg_set_marker_processor (j_decompress_ptr cinfo,
+                                        int marker_code,
+                                        jpeg_marker_parser_method routine);
 
 /* Read or write raw DCT coefficients --- useful for lossless transcoding. */
-EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients JPP((j_decompress_ptr cinfo));
-EXTERN(void) jpeg_write_coefficients JPP((j_compress_ptr cinfo,
-					  jvirt_barray_ptr * coef_arrays));
-EXTERN(void) jpeg_copy_critical_parameters JPP((j_decompress_ptr srcinfo,
-						j_compress_ptr dstinfo));
+EXTERN(jvirt_barray_ptr *) jpeg_read_coefficients (j_decompress_ptr cinfo);
+EXTERN(void) jpeg_write_coefficients (j_compress_ptr cinfo,
+                                      jvirt_barray_ptr *coef_arrays);
+EXTERN(void) jpeg_copy_critical_parameters (j_decompress_ptr srcinfo,
+                                            j_compress_ptr dstinfo);
 
 /* If you choose to abort compression or decompression before completing
  * jpeg_finish_(de)compress, then you need to clean up to release memory,
@@ -1135,28 +1049,27 @@
  * if you're done with the JPEG object, but if you want to clean it up and
  * reuse it, call this:
  */
-EXTERN(void) jpeg_abort_compress JPP((j_compress_ptr cinfo));
-EXTERN(void) jpeg_abort_decompress JPP((j_decompress_ptr cinfo));
+EXTERN(void) jpeg_abort_compress (j_compress_ptr cinfo);
+EXTERN(void) jpeg_abort_decompress (j_decompress_ptr cinfo);
 
 /* Generic versions of jpeg_abort and jpeg_destroy that work on either
  * flavor of JPEG object.  These may be more convenient in some places.
  */
-EXTERN(void) jpeg_abort JPP((j_common_ptr cinfo));
-EXTERN(void) jpeg_destroy JPP((j_common_ptr cinfo));
+EXTERN(void) jpeg_abort (j_common_ptr cinfo);
+EXTERN(void) jpeg_destroy (j_common_ptr cinfo);
 
 /* Default restart-marker-resync procedure for use by data source modules */
-EXTERN(boolean) jpeg_resync_to_restart JPP((j_decompress_ptr cinfo,
-					    int desired));
+EXTERN(boolean) jpeg_resync_to_restart (j_decompress_ptr cinfo, int desired);
 
 
 /* These marker codes are exported since applications and data source modules
  * are likely to want to use them.
  */
 
-#define JPEG_RST0	0xD0	/* RST0 marker code */
-#define JPEG_EOI	0xD9	/* EOI marker code */
-#define JPEG_APP0	0xE0	/* APP0 marker code */
-#define JPEG_COM	0xFE	/* COM marker code */
+#define JPEG_RST0       0xD0    /* RST0 marker code */
+#define JPEG_EOI        0xD9    /* EOI marker code */
+#define JPEG_APP0       0xE0    /* APP0 marker code */
+#define JPEG_COM        0xFE    /* COM marker code */
 
 
 /* If we have a brain-damaged compiler that emits warnings (or worse, errors)
@@ -1165,7 +1078,7 @@
  */
 
 #ifdef INCOMPLETE_TYPES_BROKEN
-#ifndef JPEG_INTERNALS		/* will be defined in jpegint.h */
+#ifndef JPEG_INTERNALS          /* will be defined in jpegint.h */
 struct jvirt_sarray_control { long dummy; };
 struct jvirt_barray_control { long dummy; };
 struct jpeg_comp_master { long dummy; };
@@ -1200,8 +1113,8 @@
  */
 
 #ifdef JPEG_INTERNALS
-#include "jpegint.h"		/* fetch private declarations */
-#include "jerror.h"		/* fetch error codes too */
+#include "jpegint.h"            /* fetch private declarations */
+#include "jerror.h"             /* fetch error codes too */
 #endif
 
 #ifdef __cplusplus
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeglibmangler.h glmark2-2021.02/src/libjpeg-turbo/jpeglibmangler.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeglibmangler.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpeglibmangler.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,117 @@
+// Copyright (c) 2009 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+#define THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
+
+// Mangle all externally visible function names so we can build our own libjpeg
+// without system libraries trying to use it.
+
+#define jpeg_make_c_derived_tbl chromium_jpeg_make_c_derived_tbl
+#define jpeg_gen_optimal_table chromium_jpeg_gen_optimal_table
+#define jpeg_make_d_derived_tbl chromium_jpeg_make_d_derived_tbl
+#define jpeg_fill_bit_buffer chromium_jpeg_fill_bit_buffer
+#define jpeg_huff_decode chromium_jpeg_huff_decode
+#define jpeg_fdct_islow chromium_jpeg_fdct_islow
+#define jpeg_fdct_ifast chromium_jpeg_fdct_ifast
+#define jpeg_fdct_float chromium_jpeg_fdct_float
+#define jpeg_idct_islow chromium_jpeg_idct_islow
+#define jpeg_idct_ifast chromium_jpeg_idct_ifast
+#define jpeg_idct_float chromium_jpeg_idct_float
+#define jpeg_idct_4x4 chromium_jpeg_idct_4x4
+#define jpeg_idct_2x2 chromium_jpeg_idct_2x2
+#define jpeg_idct_1x1 chromium_jpeg_idct_1x1
+#define jinit_compress_master chromium_jinit_compress_master
+#define jinit_c_master_control chromium_jinit_c_master_control
+#define jinit_c_main_controller chromium_jinit_c_main_controller
+#define jinit_c_prep_controller chromium_jinit_c_prep_controller
+#define jinit_c_coef_controller chromium_jinit_c_coef_controller
+#define jinit_color_converter chromium_jinit_color_converter
+#define jinit_downsampler chromium_jinit_downsampler
+#define jinit_forward_dct chromium_jinit_forward_dct
+#define jinit_huff_encoder chromium_jinit_huff_encoder
+#define jinit_phuff_encoder chromium_jinit_phuff_encoder
+#define jinit_marker_writer chromium_jinit_marker_writer
+#define jinit_master_decompress chromium_jinit_master_decompress
+#define jinit_d_main_controller chromium_jinit_d_main_controller
+#define jinit_d_coef_controller chromium_jinit_d_coef_controller
+#define jinit_d_post_controller chromium_jinit_d_post_controller
+#define jinit_input_controller chromium_jinit_input_controller
+#define jinit_marker_reader chromium_jinit_marker_reader
+#define jinit_huff_decoder chromium_jinit_huff_decoder
+#define jinit_phuff_decoder chromium_jinit_phuff_decoder
+#define jinit_inverse_dct chromium_jinit_inverse_dct
+#define jinit_upsampler chromium_jinit_upsampler
+#define jinit_color_deconverter chromium_jinit_color_deconverter
+#define jinit_1pass_quantizer chromium_jinit_1pass_quantizer
+#define jinit_2pass_quantizer chromium_jinit_2pass_quantizer
+#define jinit_merged_upsampler chromium_jinit_merged_upsampler
+#define jinit_memory_mgr chromium_jinit_memory_mgr
+#define jdiv_round_up chromium_jdiv_round_up
+#define jround_up chromium_jround_up
+#define jcopy_sample_rows chromium_jcopy_sample_rows
+#define jcopy_block_row chromium_jcopy_block_row
+#define jzero_far chromium_jzero_far
+#define jpeg_std_error chromium_jpeg_std_error
+#define jpeg_CreateCompress chromium_jpeg_CreateCompress
+#define jpeg_CreateDecompress chromium_jpeg_CreateDecompress
+#define jpeg_destroy_compress chromium_jpeg_destroy_compress
+#define jpeg_destroy_decompress chromium_jpeg_destroy_decompress
+#define jpeg_stdio_dest chromium_jpeg_stdio_dest
+#define jpeg_stdio_src chromium_jpeg_stdio_src
+#define jpeg_set_defaults chromium_jpeg_set_defaults
+#define jpeg_set_colorspace chromium_jpeg_set_colorspace
+#define jpeg_default_colorspace chromium_jpeg_default_colorspace
+#define jpeg_set_quality chromium_jpeg_set_quality
+#define jpeg_set_linear_quality chromium_jpeg_set_linear_quality
+#define jpeg_add_quant_table chromium_jpeg_add_quant_table
+#define jpeg_quality_scaling chromium_jpeg_quality_scaling
+#define jpeg_simple_progression chromium_jpeg_simple_progression
+#define jpeg_suppress_tables chromium_jpeg_suppress_tables
+#define jpeg_alloc_quant_table chromium_jpeg_alloc_quant_table
+#define jpeg_alloc_huff_table chromium_jpeg_alloc_huff_table
+#define jpeg_start_compress chromium_jpeg_start_compress
+#define jpeg_write_scanlines chromium_jpeg_write_scanlines
+#define jpeg_finish_compress chromium_jpeg_finish_compress
+#define jpeg_write_raw_data chromium_jpeg_write_raw_data
+#define jpeg_write_marker chromium_jpeg_write_marker
+#define jpeg_write_m_header chromium_jpeg_write_m_header
+#define jpeg_write_m_byte chromium_jpeg_write_m_byte
+#define jpeg_write_tables chromium_jpeg_write_tables
+#define jpeg_read_header chromium_jpeg_read_header
+#define jpeg_start_decompress chromium_jpeg_start_decompress
+#define jpeg_read_scanlines chromium_jpeg_read_scanlines
+#define jpeg_skip_scanlines chromium_jpeg_skip_scanlines
+#define jpeg_crop_scanline chromium_jpeg_crop_scanline
+#define jpeg_finish_decompress chromium_jpeg_finish_decompress
+#define jpeg_read_raw_data chromium_jpeg_read_raw_data
+#define jpeg_has_multiple_scans chromium_jpeg_has_multiple_scans
+#define jpeg_start_output chromium_jpeg_start_output
+#define jpeg_finish_output chromium_jpeg_finish_output
+#define jpeg_input_complete chromium_jpeg_input_complete
+#define jpeg_new_colormap chromium_jpeg_new_colormap
+#define jpeg_consume_input chromium_jpeg_consume_input
+#define jpeg_calc_output_dimensions chromium_jpeg_calc_output_dimensions
+#define jpeg_save_markers chromium_jpeg_save_markers
+#define jpeg_set_marker_processor chromium_jpeg_set_marker_processor
+#define jpeg_read_coefficients chromium_jpeg_read_coefficients
+#define jpeg_write_coefficients chromium_jpeg_write_coefficients
+#define jpeg_copy_critical_parameters chromium_jpeg_copy_critical_parameters
+#define jpeg_abort_compress chromium_jpeg_abort_compress
+#define jpeg_abort_decompress chromium_jpeg_abort_decompress
+#define jpeg_abort chromium_jpeg_abort
+#define jpeg_destroy chromium_jpeg_destroy
+#define jpeg_resync_to_restart chromium_jpeg_resync_to_restart
+#define jpeg_get_small chromium_jpeg_get_small
+#define jpeg_free_small chromium_jpeg_free_small
+#define jpeg_get_large chromium_jpeg_get_large
+#define jpeg_free_large chromium_jpeg_free_large
+#define jpeg_mem_available chromium_jpeg_mem_available
+#define jpeg_open_backing_store chromium_jpeg_open_backing_store
+#define jpeg_mem_init chromium_jpeg_mem_init
+#define jpeg_mem_term chromium_jpeg_mem_term
+#define jpeg_std_message_table chromium_jpeg_std_message_table
+#define jpeg_natural_order chromium_jpeg_natural_order
+
+#endif  // THIRD_PARTY_LIBJPEG_TURBO_JPEGLIBMANGLER_H_
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeg_nbits_table.h glmark2-2021.02/src/libjpeg-turbo/jpeg_nbits_table.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpeg_nbits_table.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpeg_nbits_table.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4098 @@
+static const unsigned char jpeg_nbits_table[65536] = {
+   0,  1,  2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,
+   5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegtran.1 glmark2-2021.02/src/libjpeg-turbo/jpegtran.1
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegtran.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpegtran.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,290 @@
+.TH JPEGTRAN 1 "18 February 2016"
+.SH NAME
+jpegtran \- lossless transformation of JPEG files
+.SH SYNOPSIS
+.B jpegtran
+[
+.I options
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B jpegtran
+performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.
+.PP
+For EXIF files and JPEG files containing Exif data, you may prefer to use
+.B exiftran
+instead.
+.PP
+.B jpegtran
+works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+.B djpeg
+followed by
+.B cjpeg
+to accomplish the same conversion.  But by the same token,
+.B jpegtran
+cannot perform lossy operations such as changing the image quality.  However,
+while the image data is losslessly transformed, metadata can be removed.  See
+the
+.B \-copy
+option for specifics.
+.PP
+.B jpegtran
+reads the named JPEG/JFIF file, or the standard input if no file is
+named, and produces a JPEG/JFIF file on the standard output.
+.SH OPTIONS
+All switch names may be abbreviated; for example,
+.B \-optimize
+may be written
+.B \-opt
+or
+.BR \-o .
+Upper and lower case are equivalent.
+British spellings are also accepted (e.g.,
+.BR \-optimise ),
+though for brevity these are not mentioned below.
+.PP
+To specify the coded JPEG representation used in the output file,
+.B jpegtran
+accepts a subset of the switches recognized by
+.BR cjpeg :
+.TP
+.B \-optimize
+Perform optimization of entropy encoding parameters.
+.TP
+.B \-progressive
+Create progressive JPEG file.
+.TP
+.BI \-restart " N"
+Emit a JPEG restart marker every N MCU rows, or every N MCU blocks if "B" is
+attached to the number.
+.TP
+.B \-arithmetic
+Use arithmetic coding.
+.TP
+.BI \-scans " file"
+Use the scan script given in the specified text file.
+.PP
+See
+.BR cjpeg (1)
+for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+.PP
+The image can be losslessly transformed by giving one of these switches:
+.TP
+.B \-flip horizontal
+Mirror image horizontally (left-right).
+.TP
+.B \-flip vertical
+Mirror image vertically (top-bottom).
+.TP
+.B \-rotate 90
+Rotate image 90 degrees clockwise.
+.TP
+.B \-rotate 180
+Rotate image 180 degrees.
+.TP
+.B \-rotate 270
+Rotate image 270 degrees clockwise (or 90 ccw).
+.TP
+.B \-transpose
+Transpose image (across UL-to-LR axis).
+.TP
+.B \-transverse
+Transverse transpose (across UR-to-LL axis).
+.PP
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+.PP
+.BR jpegtran 's
+default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+.PP
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the
+.B \-trim
+switch:
+.TP
+.B \-trim
+Drop non-transformable edge blocks.
+.IP
+Obviously, a transformation with
+.B \-trim
+is not reversible, so strictly speaking
+.B jpegtran
+with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+.B \-rot 270 -trim
+trims only the bottom edge, but
+.B \-rot 90 -trim
+followed by
+.B \-rot 180 -trim
+trims both edges.
+.TP
+.B \-perfect
+If you are only interested in perfect transformations, add the
+.B \-perfect
+switch.  This causes
+.B jpegtran
+to fail with an error if the transformation is not perfect.
+.IP
+For example, you may want to do
+.IP
+.B (jpegtran \-rot 90 -perfect
+.I foo.jpg
+.B || djpeg
+.I foo.jpg
+.B | pnmflip \-r90 | cjpeg)
+.IP
+to do a perfect rotation, if available, or an approximated one if not.
+.PP
+This version of \fBjpegtran\fR also offers a lossless crop option, which
+discards data outside of a given image region but losslessly preserves what is
+inside. Like the rotate and flip transforms, lossless crop is restricted by the
+current JPEG format; the upper left corner of the selected region must fall on
+an iMCU boundary.  If it doesn't, then it is silently moved up and/or left to
+the nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the
+output image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching
+an 'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+.TP
+.B \-crop WxH+X+Y
+Crop the image to a rectangular region of width W and height H, starting at
+point X,Y.  The lossless crop feature discards data outside of a given image
+region but losslessly preserves what is inside.  Like the rotate and flip
+transforms, lossless crop is restricted by the current JPEG format; the upper
+left corner of the selected region must fall on an iMCU boundary.  If it
+doesn't, then it is silently moved up and/or left to the nearest iMCU boundary
+(the lower right corner is unchanged.)
+.PP
+Other not-strictly-lossless transformation switches are:
+.TP
+.B \-grayscale
+Force grayscale output.
+.IP
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+.PP
+.B jpegtran
+also recognizes these switches that control what to do with "extra" markers,
+such as comment blocks:
+.TP
+.B \-copy none
+Copy no extra markers from source file.  This setting suppresses all
+comments and other metadata in the source file.
+.TP
+.B \-copy comments
+Copy only comment markers.  This setting copies comments from the source file
+but discards any other metadata.
+.TP
+.B \-copy all
+Copy all extra markers.  This setting preserves miscellaneous markers
+found in the source file, such as JFIF thumbnails, Exif data, and Photoshop
+settings.  In some files, these extra markers can be sizable.  Note that this
+option will copy thumbnails as-is; they will not be transformed.
+.PP
+The default behavior is \fB-copy comments\fR.  (Note: in IJG releases v6 and
+v6a, \fBjpegtran\fR always did the equivalent of \fB-copy none\fR.)
+.PP
+Additional switches recognized by jpegtran are:
+.TP
+.BI \-maxmemory " N"
+Set limit for amount of memory to use in processing large images.  Value is
+in thousands of bytes, or millions of bytes if "M" is attached to the
+number.  For example,
+.B \-max 4m
+selects 4000000 bytes.  If more space is needed, temporary files will be used.
+.TP
+.BI \-outfile " name"
+Send output image to the named file, not to standard output.
+.TP
+.B \-verbose
+Enable debug printout.  More
+.BR \-v 's
+give more output.  Also, version information is printed at startup.
+.TP
+.B \-debug
+Same as
+.BR \-verbose .
+.TP
+.B \-version
+Print version information and exit.
+.SH EXAMPLES
+.LP
+This example converts a baseline JPEG file to progressive form:
+.IP
+.B jpegtran \-progressive
+.I foo.jpg
+.B >
+.I fooprog.jpg
+.PP
+This example rotates an image 90 degrees clockwise, discarding any
+unrotatable edge pixels:
+.IP
+.B jpegtran \-rot 90 -trim
+.I foo.jpg
+.B >
+.I foo90.jpg
+.SH ENVIRONMENT
+.TP
+.B JPEGMEM
+If this environment variable is set, its value is the default memory limit.
+The value is specified as described for the
+.B \-maxmemory
+switch.
+.B JPEGMEM
+overrides the default value specified when the program was compiled, and
+itself is overridden by an explicit
+.BR \-maxmemory .
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR rdjpgcom (1),
+.BR wrjpgcom (1)
+.br
+Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+Communications of the ACM, April 1991 (vol. 34, no. 4), pp. 30-44.
+.SH AUTHOR
+Independent JPEG Group
+.PP
+This file was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo and to wordsmith certain sections.
+.SH BUGS
+The transform options can't transform odd-size images perfectly.  Use
+.B \-trim
+or
+.B \-perfect
+if you don't like the results.
+.PP
+The entire image is read into memory and then written out again, even in
+cases where this isn't really necessary.  Expect swapping on large images,
+especially when using the more complex transform options.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegtran.c glmark2-2021.02/src/libjpeg-turbo/jpegtran.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jpegtran.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jpegtran.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,551 @@
+/*
+ * jpegtran.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1995-2010, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2014, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a command-line user interface for JPEG transcoding.
+ * It is very similar to cjpeg.c, and partly to djpeg.c, but provides
+ * lossless transcoding between different JPEG file formats.  It also
+ * provides some lossless and sort-of-lossless transformations of JPEG data.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "transupp.h"           /* Support routines for jpegtran */
+#include "jversion.h"           /* for version message */
+#include "jconfigint.h"
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+
+/*
+ * Argument-parsing code.
+ * The switch parser is designed to be useful with DOS-style command line
+ * syntax, ie, intermixed switches and file names, where only the switches
+ * to the left of a given file name affect processing of that file.
+ * The main program in this file doesn't actually use this capability...
+ */
+
+
+static const char *progname;    /* program name for error messages */
+static char *outfilename;       /* for -outfile switch */
+static JCOPY_OPTION copyoption; /* -copy switch */
+static jpeg_transform_info transformoption; /* image transformation options */
+
+
+LOCAL(void)
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -copy none     Copy no extra markers from source file\n");
+  fprintf(stderr, "  -copy comments Copy only comment markers (default)\n");
+  fprintf(stderr, "  -copy all      Copy all extra markers\n");
+#ifdef ENTROPY_OPT_SUPPORTED
+  fprintf(stderr, "  -optimize      Optimize Huffman table (smaller file, but slow compression)\n");
+#endif
+#ifdef C_PROGRESSIVE_SUPPORTED
+  fprintf(stderr, "  -progressive   Create progressive JPEG file\n");
+#endif
+  fprintf(stderr, "Switches for modifying the image:\n");
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -crop WxH+X+Y  Crop to a rectangular subarea\n");
+  fprintf(stderr, "  -grayscale     Reduce to grayscale (omit color data)\n");
+  fprintf(stderr, "  -flip [horizontal|vertical]  Mirror image (left-right or top-bottom)\n");
+  fprintf(stderr, "  -perfect       Fail if there is non-transformable edge blocks\n");
+  fprintf(stderr, "  -rotate [90|180|270]         Rotate image (degrees clockwise)\n");
+#endif
+#if TRANSFORMS_SUPPORTED
+  fprintf(stderr, "  -transpose     Transpose image\n");
+  fprintf(stderr, "  -transverse    Transverse transpose image\n");
+  fprintf(stderr, "  -trim          Drop non-transformable edge blocks\n");
+#endif
+  fprintf(stderr, "Switches for advanced users:\n");
+#ifdef C_ARITH_CODING_SUPPORTED
+  fprintf(stderr, "  -arithmetic    Use arithmetic coding\n");
+#endif
+  fprintf(stderr, "  -restart N     Set restart interval in rows, or in blocks with B\n");
+  fprintf(stderr, "  -maxmemory N   Maximum memory to use (in kbytes)\n");
+  fprintf(stderr, "  -outfile name  Specify name for output file\n");
+  fprintf(stderr, "  -verbose  or  -debug   Emit debug output\n");
+  fprintf(stderr, "  -version       Print version information and exit\n");
+  fprintf(stderr, "Switches for wizards:\n");
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+  fprintf(stderr, "  -scans file    Create multi-scan JPEG per script file\n");
+#endif
+  exit(EXIT_FAILURE);
+}
+
+
+LOCAL(void)
+select_transform (JXFORM_CODE transform)
+/* Silly little routine to detect multiple transform options,
+ * which we can't handle.
+ */
+{
+#if TRANSFORMS_SUPPORTED
+  if (transformoption.transform == JXFORM_NONE ||
+      transformoption.transform == transform) {
+    transformoption.transform = transform;
+  } else {
+    fprintf(stderr, "%s: can only do one image transformation at a time\n",
+            progname);
+    usage();
+  }
+#else
+  fprintf(stderr, "%s: sorry, image transformation was not compiled\n",
+          progname);
+  exit(EXIT_FAILURE);
+#endif
+}
+
+
+LOCAL(int)
+parse_switches (j_compress_ptr cinfo, int argc, char **argv,
+                int last_file_arg_seen, boolean for_real)
+/* Parse optional switches.
+ * Returns argv[] index of first file-name argument (== argc if none).
+ * Any file names with indexes <= last_file_arg_seen are ignored;
+ * they have presumably been processed in a previous iteration.
+ * (Pass 0 for last_file_arg_seen on the first or only iteration.)
+ * for_real is FALSE on the first (dummy) pass; we may skip any expensive
+ * processing.
+ */
+{
+  int argn;
+  char *arg;
+  boolean simple_progressive;
+  char *scansarg = NULL;        /* saves -scans parm if any */
+
+  /* Set up default JPEG parameters. */
+  simple_progressive = FALSE;
+  outfilename = NULL;
+  copyoption = JCOPYOPT_DEFAULT;
+  transformoption.transform = JXFORM_NONE;
+  transformoption.perfect = FALSE;
+  transformoption.trim = FALSE;
+  transformoption.force_grayscale = FALSE;
+  transformoption.crop = FALSE;
+  transformoption.slow_hflip = FALSE;
+  cinfo->err->trace_level = 0;
+
+  /* Scan command line options, adjust parameters */
+
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (*arg != '-') {
+      /* Not a switch, must be a file name argument */
+      if (argn <= last_file_arg_seen) {
+        outfilename = NULL;     /* -outfile applies to just one input file */
+        continue;               /* ignore this name if previously processed */
+      }
+      break;                    /* else done parsing switches */
+    }
+    arg++;                      /* advance past switch marker character */
+
+    if (keymatch(arg, "arithmetic", 1)) {
+      /* Use arithmetic coding. */
+#ifdef C_ARITH_CODING_SUPPORTED
+      cinfo->arith_code = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, arithmetic coding not supported\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "copy", 2)) {
+      /* Select which extra markers to copy. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "none", 1)) {
+        copyoption = JCOPYOPT_NONE;
+      } else if (keymatch(argv[argn], "comments", 1)) {
+        copyoption = JCOPYOPT_COMMENTS;
+      } else if (keymatch(argv[argn], "all", 1)) {
+        copyoption = JCOPYOPT_ALL;
+      } else
+        usage();
+
+    } else if (keymatch(arg, "crop", 2)) {
+      /* Perform lossless cropping. */
+#if TRANSFORMS_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (! jtransform_parse_crop_spec(&transformoption, argv[argn])) {
+        fprintf(stderr, "%s: bogus -crop argument '%s'\n",
+                progname, argv[argn]);
+        exit(EXIT_FAILURE);
+      }
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "debug", 1) || keymatch(arg, "verbose", 1)) {
+      /* Enable debug printouts. */
+      /* On first -d, print version identification */
+      static boolean printed_version = FALSE;
+
+      if (! printed_version) {
+        fprintf(stderr, "%s version %s (build %s)\n",
+                PACKAGE_NAME, VERSION, BUILD);
+        fprintf(stderr, "%s\n\n", JCOPYRIGHT);
+        fprintf(stderr, "Emulating The Independent JPEG Group's software, version %s\n\n",
+                JVERSION);
+        printed_version = TRUE;
+      }
+      cinfo->err->trace_level++;
+
+    } else if (keymatch(arg, "version", 4)) {
+      fprintf(stderr, "%s version %s (build %s)\n",
+              PACKAGE_NAME, VERSION, BUILD);
+      exit(EXIT_SUCCESS);
+
+    } else if (keymatch(arg, "flip", 1)) {
+      /* Mirror left-right or top-bottom. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "horizontal", 1))
+        select_transform(JXFORM_FLIP_H);
+      else if (keymatch(argv[argn], "vertical", 1))
+        select_transform(JXFORM_FLIP_V);
+      else
+        usage();
+
+    } else if (keymatch(arg, "grayscale", 1) || keymatch(arg, "greyscale",1)) {
+      /* Force to grayscale. */
+#if TRANSFORMS_SUPPORTED
+      transformoption.force_grayscale = TRUE;
+#else
+      select_transform(JXFORM_NONE);    /* force an error */
+#endif
+
+    } else if (keymatch(arg, "maxmemory", 3)) {
+      /* Maximum memory in Kb (or Mb with 'm'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (ch == 'm' || ch == 'M')
+        lval *= 1000L;
+      cinfo->mem->max_memory_to_use = lval * 1000L;
+
+    } else if (keymatch(arg, "optimize", 1) || keymatch(arg, "optimise", 1)) {
+      /* Enable entropy parm optimization. */
+#ifdef ENTROPY_OPT_SUPPORTED
+      cinfo->optimize_coding = TRUE;
+#else
+      fprintf(stderr, "%s: sorry, entropy optimization was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "outfile", 4)) {
+      /* Set output file name. */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      outfilename = argv[argn]; /* save it away for later use */
+
+    } else if (keymatch(arg, "perfect", 2)) {
+      /* Fail if there is any partial edge MCUs that the transform can't
+       * handle. */
+      transformoption.perfect = TRUE;
+
+    } else if (keymatch(arg, "progressive", 2)) {
+      /* Select simple progressive mode. */
+#ifdef C_PROGRESSIVE_SUPPORTED
+      simple_progressive = TRUE;
+      /* We must postpone execution until num_components is known. */
+#else
+      fprintf(stderr, "%s: sorry, progressive output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "restart", 1)) {
+      /* Restart interval in MCU rows (or in MCUs with 'b'). */
+      long lval;
+      char ch = 'x';
+
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (sscanf(argv[argn], "%ld%c", &lval, &ch) < 1)
+        usage();
+      if (lval < 0 || lval > 65535L)
+        usage();
+      if (ch == 'b' || ch == 'B') {
+        cinfo->restart_interval = (unsigned int) lval;
+        cinfo->restart_in_rows = 0; /* else prior '-restart n' overrides me */
+      } else {
+        cinfo->restart_in_rows = (int) lval;
+        /* restart_interval will be computed during startup */
+      }
+
+    } else if (keymatch(arg, "rotate", 2)) {
+      /* Rotate 90, 180, or 270 degrees (measured clockwise). */
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      if (keymatch(argv[argn], "90", 2))
+        select_transform(JXFORM_ROT_90);
+      else if (keymatch(argv[argn], "180", 3))
+        select_transform(JXFORM_ROT_180);
+      else if (keymatch(argv[argn], "270", 3))
+        select_transform(JXFORM_ROT_270);
+      else
+        usage();
+
+    } else if (keymatch(arg, "scans", 1)) {
+      /* Set scan script. */
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+      if (++argn >= argc)       /* advance to next argument */
+        usage();
+      scansarg = argv[argn];
+      /* We must postpone reading the file in case -progressive appears. */
+#else
+      fprintf(stderr, "%s: sorry, multi-scan output was not compiled\n",
+              progname);
+      exit(EXIT_FAILURE);
+#endif
+
+    } else if (keymatch(arg, "transpose", 1)) {
+      /* Transpose (across UL-to-LR axis). */
+      select_transform(JXFORM_TRANSPOSE);
+
+    } else if (keymatch(arg, "transverse", 6)) {
+      /* Transverse transpose (across UR-to-LL axis). */
+      select_transform(JXFORM_TRANSVERSE);
+
+    } else if (keymatch(arg, "trim", 3)) {
+      /* Trim off any partial edge MCUs that the transform can't handle. */
+      transformoption.trim = TRUE;
+
+    } else {
+      usage();                  /* bogus switch */
+    }
+  }
+
+  /* Post-switch-scanning cleanup */
+
+  if (for_real) {
+
+#ifdef C_PROGRESSIVE_SUPPORTED
+    if (simple_progressive)     /* process -progressive; -scans can override */
+      jpeg_simple_progression(cinfo);
+#endif
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+    if (scansarg != NULL)       /* process -scans if it was present */
+      if (! read_scan_script(cinfo, scansarg))
+        usage();
+#endif
+  }
+
+  return argn;                  /* return index of next arg (file name) */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  struct jpeg_decompress_struct srcinfo;
+  struct jpeg_compress_struct dstinfo;
+  struct jpeg_error_mgr jsrcerr, jdsterr;
+#ifdef PROGRESS_REPORT
+  struct cdjpeg_progress_mgr progress;
+#endif
+  jvirt_barray_ptr *src_coef_arrays;
+  jvirt_barray_ptr *dst_coef_arrays;
+  int file_index;
+  /* We assume all-in-memory processing and can therefore use only a
+   * single file pointer for sequential input and output operation.
+   */
+  FILE *fp;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "jpegtran";      /* in case C library doesn't provide it */
+
+  /* Initialize the JPEG decompression object with default error handling. */
+  srcinfo.err = jpeg_std_error(&jsrcerr);
+  jpeg_create_decompress(&srcinfo);
+  /* Initialize the JPEG compression object with default error handling. */
+  dstinfo.err = jpeg_std_error(&jdsterr);
+  jpeg_create_compress(&dstinfo);
+
+  /* Scan command line to find file names.
+   * It is convenient to use just one switch-parsing routine, but the switch
+   * values read here are mostly ignored; we will rescan the switches after
+   * opening the input file.  Also note that most of the switches affect the
+   * destination JPEG object, so we parse into that and then copy over what
+   * needs to affects the source too.
+   */
+
+  file_index = parse_switches(&dstinfo, argc, argv, 0, FALSE);
+  jsrcerr.trace_level = jdsterr.trace_level;
+  srcinfo.mem->max_memory_to_use = dstinfo.mem->max_memory_to_use;
+
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have either -outfile switch or explicit output file name */
+  if (outfilename == NULL) {
+    if (file_index != argc-2) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+    outfilename = argv[file_index+1];
+  } else {
+    if (file_index != argc-1) {
+      fprintf(stderr, "%s: must name one input and one output file\n",
+              progname);
+      usage();
+    }
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (file_index < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Open the input file. */
+  if (file_index < argc) {
+    if ((fp = fopen(argv[file_index], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for reading\n", progname, argv[file_index]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+    fp = read_stdin();
+  }
+
+#ifdef PROGRESS_REPORT
+  start_progress_monitor((j_common_ptr) &dstinfo, &progress);
+#endif
+
+  /* Specify data source for decompression */
+  jpeg_stdio_src(&srcinfo, fp);
+
+  /* Enable saving of extra markers that we want to copy */
+  jcopy_markers_setup(&srcinfo, copyoption);
+
+  /* Read file header */
+  (void) jpeg_read_header(&srcinfo, TRUE);
+
+  /* Any space needed by a transform option must be requested before
+   * jpeg_read_coefficients so that memory allocation will be done right.
+   */
+#if TRANSFORMS_SUPPORTED
+  /* Fail right away if -perfect is given and transformation is not perfect.
+   */
+  if (!jtransform_request_workspace(&srcinfo, &transformoption)) {
+    fprintf(stderr, "%s: transformation is not perfect\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#endif
+
+  /* Read source file as DCT coefficients */
+  src_coef_arrays = jpeg_read_coefficients(&srcinfo);
+
+  /* Initialize destination compression parameters from source values */
+  jpeg_copy_critical_parameters(&srcinfo, &dstinfo);
+
+  /* Adjust destination parameters if required by transform options;
+   * also find out which set of coefficient arrays will hold the output.
+   */
+#if TRANSFORMS_SUPPORTED
+  dst_coef_arrays = jtransform_adjust_parameters(&srcinfo, &dstinfo,
+                                                 src_coef_arrays,
+                                                 &transformoption);
+#else
+  dst_coef_arrays = src_coef_arrays;
+#endif
+
+  /* Close input file, if we opened it.
+   * Note: we assume that jpeg_read_coefficients consumed all input
+   * until JPEG_REACHED_EOI, and that jpeg_finish_decompress will
+   * only consume more while (! cinfo->inputctl->eoi_reached).
+   * We cannot call jpeg_finish_decompress here since we still need the
+   * virtual arrays allocated from the source object for processing.
+   */
+  if (fp != stdin)
+    fclose(fp);
+
+  /* Open the output file. */
+  if (outfilename != NULL) {
+    if ((fp = fopen(outfilename, WRITE_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s for writing\n", progname, outfilename);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default output file is stdout */
+    fp = write_stdout();
+  }
+
+  /* Adjust default compression parameters by re-parsing the options */
+  file_index = parse_switches(&dstinfo, argc, argv, 0, TRUE);
+
+  /* Specify data destination for compression */
+  jpeg_stdio_dest(&dstinfo, fp);
+
+  /* Start compressor (note no image data is actually written here) */
+  jpeg_write_coefficients(&dstinfo, dst_coef_arrays);
+
+  /* Copy to the output file any extra markers that we want to preserve */
+  jcopy_markers_execute(&srcinfo, &dstinfo, copyoption);
+
+  /* Execute image transformation, if any */
+#if TRANSFORMS_SUPPORTED
+  jtransform_execute_transformation(&srcinfo, &dstinfo,
+                                    src_coef_arrays,
+                                    &transformoption);
+#endif
+
+  /* Finish compression and release memory */
+  jpeg_finish_compress(&dstinfo);
+  jpeg_destroy_compress(&dstinfo);
+  (void) jpeg_finish_decompress(&srcinfo);
+  jpeg_destroy_decompress(&srcinfo);
+
+  /* Close output file, if we opened it */
+  if (fp != stdout)
+    fclose(fp);
+
+#ifdef PROGRESS_REPORT
+  end_progress_monitor((j_common_ptr) &dstinfo);
+#endif
+
+  /* All done. */
+  exit(jsrcerr.num_warnings + jdsterr.num_warnings ?EXIT_WARNING:EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jquant1.c glmark2-2021.02/src/libjpeg-turbo/jquant1.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jquant1.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jquant1.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jquant1.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009, D. R. Commander
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains 1-pass color quantization (color mapping) routines.
  * These routines provide mapping to a fixed color map using equally spaced
@@ -69,9 +71,9 @@
  * table in both directions.
  */
 
-#define ODITHER_SIZE  16	/* dimension of dither matrix */
+#define ODITHER_SIZE  16        /* dimension of dither matrix */
 /* NB: if ODITHER_SIZE is not a power of 2, ODITHER_MASK uses will break */
-#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)	/* # cells in matrix */
+#define ODITHER_CELLS (ODITHER_SIZE*ODITHER_SIZE)       /* # cells in matrix */
 #define ODITHER_MASK  (ODITHER_SIZE-1) /* mask for wrapping around counters */
 
 typedef int ODITHER_MATRIX[ODITHER_SIZE][ODITHER_SIZE];
@@ -106,8 +108,8 @@
  * Errors are accumulated into the array fserrors[], at a resolution of
  * 1/16th of a pixel count.  The error at a given pixel is propagated
  * to its not-yet-processed neighbors using the standard F-S fractions,
- *		...	(here)	7/16
- *		3/16	5/16	1/16
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
  * We work left-to-right on even rows, right-to-left on odd rows.
  *
  * We can get away with a single array (holding one row's worth of errors)
@@ -120,52 +122,49 @@
  * The fserrors[] array is indexed [component#][position].
  * We provide (#columns + 2) entries per component; the extra entry at each
  * end saves us from special-casing the first and last pixels.
- *
- * Note: on a wide image, we might not have enough room in a PC's near data
- * segment to hold the error array; so it is allocated with alloc_large.
  */
 
 #if BITS_IN_JSAMPLE == 8
-typedef INT16 FSERROR;		/* 16 bits should be enough */
-typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
 #else
-typedef INT32 FSERROR;		/* may need more than 16 bits */
-typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
 #endif
 
-typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+typedef FSERROR *FSERRPTR;  /* pointer to error array */
 
 
 /* Private subobject */
 
-#define MAX_Q_COMPS 4		/* max components I can handle */
+#define MAX_Q_COMPS 4           /* max components I can handle */
 
 typedef struct {
   struct jpeg_color_quantizer pub; /* public fields */
 
   /* Initially allocated colormap is saved here */
-  JSAMPARRAY sv_colormap;	/* The color map as a 2-D pixel array */
-  int sv_actual;		/* number of entries in use */
+  JSAMPARRAY sv_colormap;       /* The color map as a 2-D pixel array */
+  int sv_actual;                /* number of entries in use */
 
-  JSAMPARRAY colorindex;	/* Precomputed mapping for speed */
+  JSAMPARRAY colorindex;        /* Precomputed mapping for speed */
   /* colorindex[i][j] = index of color closest to pixel value j in component i,
    * premultiplied as described above.  Since colormap indexes must fit into
    * JSAMPLEs, the entries of this array will too.
    */
-  boolean is_padded;		/* is the colorindex padded for odither? */
+  boolean is_padded;            /* is the colorindex padded for odither? */
 
-  int Ncolors[MAX_Q_COMPS];	/* # of values alloced to each component */
+  int Ncolors[MAX_Q_COMPS];     /* # of values alloced to each component */
 
   /* Variables for ordered dithering */
-  int row_index;		/* cur row's vertical index in dither matrix */
+  int row_index;                /* cur row's vertical index in dither matrix */
   ODITHER_MATRIX_PTR odither[MAX_Q_COMPS]; /* one dither array per component */
 
   /* Variables for Floyd-Steinberg dithering */
   FSERRPTR fserrors[MAX_Q_COMPS]; /* accumulated errors */
-  boolean on_odd_row;		/* flag to remember which row we are on */
+  boolean on_odd_row;           /* flag to remember which row we are on */
 } my_cquantizer;
 
-typedef my_cquantizer * my_cquantize_ptr;
+typedef my_cquantizer *my_cquantize_ptr;
 
 
 /*
@@ -204,11 +203,11 @@
   iroot = 1;
   do {
     iroot++;
-    temp = iroot;		/* set temp = iroot ** nc */
+    temp = iroot;               /* set temp = iroot ** nc */
     for (i = 1; i < nc; i++)
       temp *= iroot;
   } while (temp <= (long) max_colors); /* repeat till iroot exceeds root */
-  iroot--;			/* now iroot = floor(root) */
+  iroot--;                      /* now iroot = floor(root) */
 
   /* Must have at least 2 color values per component */
   if (iroot < 2)
@@ -232,10 +231,10 @@
       j = (cinfo->out_color_space == JCS_RGB ? RGB_order[i] : i);
       /* calculate new total_colors if Ncolors[j] is incremented */
       temp = total_colors / Ncolors[j];
-      temp *= Ncolors[j]+1;	/* done in long arith to avoid oflo */
+      temp *= Ncolors[j]+1;     /* done in long arith to avoid oflo */
       if (temp > (long) max_colors)
-	break;			/* won't fit, done with this pass */
-      Ncolors[j]++;		/* OK, apply the increment */
+        break;                  /* won't fit, done with this pass */
+      Ncolors[j]++;             /* OK, apply the increment */
       total_colors = (int) temp;
       changed = TRUE;
     }
@@ -255,7 +254,7 @@
    * (Forcing the upper and lower values to the limits ensures that
    * dithering can't produce a color outside the selected gamut.)
    */
-  return (int) (((INT32) j * MAXJSAMPLE + maxj/2) / maxj);
+  return (int) (((JLONG) j * MAXJSAMPLE + maxj/2) / maxj);
 }
 
 
@@ -265,7 +264,7 @@
 /* Must have largest(j=0) >= 0, and largest(j=maxj) >= MAXJSAMPLE */
 {
   /* Breakpoints are halfway between values returned by output_value */
-  return (int) (((INT32) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
+  return (int) (((JLONG) (2*j + 1) * MAXJSAMPLE + maxj) / (2*maxj));
 }
 
 
@@ -277,8 +276,8 @@
 create_colormap (j_decompress_ptr cinfo)
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  JSAMPARRAY colormap;		/* Created colormap */
-  int total_colors;		/* Number of distinct output colors */
+  JSAMPARRAY colormap;          /* Created colormap */
+  int total_colors;             /* Number of distinct output colors */
   int i,j,k, nci, blksize, blkdist, ptr, val;
 
   /* Select number of colors for each component */
@@ -287,8 +286,8 @@
   /* Report selected color counts */
   if (cinfo->out_color_components == 3)
     TRACEMS4(cinfo, 1, JTRC_QUANT_3_NCOLORS,
-	     total_colors, cquantize->Ncolors[0],
-	     cquantize->Ncolors[1], cquantize->Ncolors[2]);
+             total_colors, cquantize->Ncolors[0],
+             cquantize->Ncolors[1], cquantize->Ncolors[2]);
   else
     TRACEMS1(cinfo, 1, JTRC_QUANT_NCOLORS, total_colors);
 
@@ -313,12 +312,12 @@
       val = output_value(cinfo, i, j, nci-1);
       /* Fill in all colormap entries that have this value of this component */
       for (ptr = j * blksize; ptr < total_colors; ptr += blkdist) {
-	/* fill in blksize entries beginning at ptr */
-	for (k = 0; k < blksize; k++)
-	  colormap[i][ptr+k] = (JSAMPLE) val;
+        /* fill in blksize entries beginning at ptr */
+        for (k = 0; k < blksize; k++)
+          colormap[i][ptr+k] = (JSAMPLE) val;
       }
     }
-    blkdist = blksize;		/* blksize of this color is blkdist of next */
+    blkdist = blksize;          /* blksize of this color is blkdist of next */
   }
 
   /* Save the colormap in private storage,
@@ -376,16 +375,16 @@
     val = 0;
     k = largest_input_value(cinfo, i, 0, nci-1);
     for (j = 0; j <= MAXJSAMPLE; j++) {
-      while (j > k)		/* advance val if past boundary */
-	k = largest_input_value(cinfo, i, ++val, nci-1);
+      while (j > k)             /* advance val if past boundary */
+        k = largest_input_value(cinfo, i, ++val, nci-1);
       /* premultiply so that no multiplication needed in main processing */
       indexptr[j] = (JSAMPLE) (val * blksize);
     }
     /* Pad at both ends if necessary */
     if (pad)
       for (j = 1; j <= MAXJSAMPLE; j++) {
-	indexptr[-j] = indexptr[0];
-	indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
+        indexptr[-j] = indexptr[0];
+        indexptr[MAXJSAMPLE+j] = indexptr[MAXJSAMPLE];
       }
   }
 }
@@ -401,21 +400,21 @@
 {
   ODITHER_MATRIX_PTR odither;
   int j,k;
-  INT32 num,den;
+  JLONG num,den;
 
   odither = (ODITHER_MATRIX_PTR)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(ODITHER_MATRIX));
+                                sizeof(ODITHER_MATRIX));
   /* The inter-value distance for this color is MAXJSAMPLE/(ncolors-1).
    * Hence the dither value for the matrix cell with fill order f
    * (f=0..N-1) should be (N-1-2*f)/(2*N) * MAXJSAMPLE/(ncolors-1).
    * On 16-bit-int machine, be careful to avoid overflow.
    */
-  den = 2 * ODITHER_CELLS * ((INT32) (ncolors - 1));
+  den = 2 * ODITHER_CELLS * ((JLONG) (ncolors - 1));
   for (j = 0; j < ODITHER_SIZE; j++) {
     for (k = 0; k < ODITHER_SIZE; k++) {
-      num = ((INT32) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
-	    * MAXJSAMPLE;
+      num = ((JLONG) (ODITHER_CELLS-1 - 2*((int)base_dither_matrix[j][k])))
+            * MAXJSAMPLE;
       /* Ensure round towards zero despite C's lack of consistency
        * about rounding negative values in integer division...
        */
@@ -428,7 +427,7 @@
 
 /*
  * Create the ordered-dither tables.
- * Components having the same number of representative colors may 
+ * Components having the same number of representative colors may
  * share a dither table.
  */
 
@@ -441,14 +440,14 @@
 
   for (i = 0; i < cinfo->out_color_components; i++) {
     nci = cquantize->Ncolors[i]; /* # of distinct values for this color */
-    odither = NULL;		/* search for matching prior component */
+    odither = NULL;             /* search for matching prior component */
     for (j = 0; j < i; j++) {
       if (nci == cquantize->Ncolors[j]) {
-	odither = cquantize->odither[j];
-	break;
+        odither = cquantize->odither[j];
+        break;
       }
     }
-    if (odither == NULL)	/* need a new table? */
+    if (odither == NULL)        /* need a new table? */
       odither = make_odither_array(cinfo, nci);
     cquantize->odither[i] = odither;
   }
@@ -461,7 +460,7 @@
 
 METHODDEF(void)
 color_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		JSAMPARRAY output_buf, int num_rows)
+                JSAMPARRAY output_buf, int num_rows)
 /* General case, no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -479,7 +478,7 @@
     for (col = width; col > 0; col--) {
       pixcode = 0;
       for (ci = 0; ci < nc; ci++) {
-	pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
+        pixcode += GETJSAMPLE(colorindex[ci][GETJSAMPLE(*ptrin++)]);
       }
       *ptrout++ = (JSAMPLE) pixcode;
     }
@@ -489,7 +488,7 @@
 
 METHODDEF(void)
 color_quantize3 (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		 JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY output_buf, int num_rows)
 /* Fast path for out_color_components==3, no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -517,15 +516,15 @@
 
 METHODDEF(void)
 quantize_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		     JSAMPARRAY output_buf, int num_rows)
+                     JSAMPARRAY output_buf, int num_rows)
 /* General case, with ordered dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   register JSAMPROW input_ptr;
   register JSAMPROW output_ptr;
   JSAMPROW colorindex_ci;
-  int * dither;			/* points to active row of dither matrix */
-  int row_index, col_index;	/* current indexes into dither matrix */
+  int *dither;                  /* points to active row of dither matrix */
+  int row_index, col_index;     /* current indexes into dither matrix */
   int nc = cinfo->out_color_components;
   int ci;
   int row;
@@ -534,8 +533,7 @@
 
   for (row = 0; row < num_rows; row++) {
     /* Initialize output values to 0 so can process components separately */
-    jzero_far((void FAR *) output_buf[row],
-	      (size_t) (width * SIZEOF(JSAMPLE)));
+    jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE)));
     row_index = cquantize->row_index;
     for (ci = 0; ci < nc; ci++) {
       input_ptr = input_buf[row] + ci;
@@ -545,17 +543,17 @@
       col_index = 0;
 
       for (col = width; col > 0; col--) {
-	/* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
-	 * select output value, accumulate into output code for this pixel.
-	 * Range-limiting need not be done explicitly, as we have extended
-	 * the colorindex table to produce the right answers for out-of-range
-	 * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
-	 * required amount of padding.
-	 */
-	*output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
-	input_ptr += nc;
-	output_ptr++;
-	col_index = (col_index + 1) & ODITHER_MASK;
+        /* Form pixel value + dither, range-limit to 0..MAXJSAMPLE,
+         * select output value, accumulate into output code for this pixel.
+         * Range-limiting need not be done explicitly, as we have extended
+         * the colorindex table to produce the right answers for out-of-range
+         * inputs.  The maximum dither is +- MAXJSAMPLE; this sets the
+         * required amount of padding.
+         */
+        *output_ptr += colorindex_ci[GETJSAMPLE(*input_ptr)+dither[col_index]];
+        input_ptr += nc;
+        output_ptr++;
+        col_index = (col_index + 1) & ODITHER_MASK;
       }
     }
     /* Advance row index for next row */
@@ -567,7 +565,7 @@
 
 METHODDEF(void)
 quantize3_ord_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		      JSAMPARRAY output_buf, int num_rows)
+                      JSAMPARRAY output_buf, int num_rows)
 /* Fast path for out_color_components==3, with ordered dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -577,10 +575,10 @@
   JSAMPROW colorindex0 = cquantize->colorindex[0];
   JSAMPROW colorindex1 = cquantize->colorindex[1];
   JSAMPROW colorindex2 = cquantize->colorindex[2];
-  int * dither0;		/* points to active row of dither matrix */
-  int * dither1;
-  int * dither2;
-  int row_index, col_index;	/* current indexes into dither matrix */
+  int *dither0;                 /* points to active row of dither matrix */
+  int *dither1;
+  int *dither2;
+  int row_index, col_index;     /* current indexes into dither matrix */
   int row;
   JDIMENSION col;
   JDIMENSION width = cinfo->output_width;
@@ -596,11 +594,11 @@
 
     for (col = width; col > 0; col--) {
       pixcode  = GETJSAMPLE(colorindex0[GETJSAMPLE(*input_ptr++) +
-					dither0[col_index]]);
+                                        dither0[col_index]]);
       pixcode += GETJSAMPLE(colorindex1[GETJSAMPLE(*input_ptr++) +
-					dither1[col_index]]);
+                                        dither1[col_index]]);
       pixcode += GETJSAMPLE(colorindex2[GETJSAMPLE(*input_ptr++) +
-					dither2[col_index]]);
+                                        dither2[col_index]]);
       *output_ptr++ = (JSAMPLE) pixcode;
       col_index = (col_index + 1) & ODITHER_MASK;
     }
@@ -612,24 +610,24 @@
 
 METHODDEF(void)
 quantize_fs_dither (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		    JSAMPARRAY output_buf, int num_rows)
+                    JSAMPARRAY output_buf, int num_rows)
 /* General case, with Floyd-Steinberg dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  register LOCFSERROR cur;	/* current error or pixel value */
-  LOCFSERROR belowerr;		/* error for pixel below cur */
-  LOCFSERROR bpreverr;		/* error for below/prev col */
-  LOCFSERROR bnexterr;		/* error for below/next col */
+  register LOCFSERROR cur;      /* current error or pixel value */
+  LOCFSERROR belowerr;          /* error for pixel below cur */
+  LOCFSERROR bpreverr;          /* error for below/prev col */
+  LOCFSERROR bnexterr;          /* error for below/next col */
   LOCFSERROR delta;
-  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
+  register FSERRPTR errorptr;   /* => fserrors[] at column before current */
   register JSAMPROW input_ptr;
   register JSAMPROW output_ptr;
   JSAMPROW colorindex_ci;
   JSAMPROW colormap_ci;
   int pixcode;
   int nc = cinfo->out_color_components;
-  int dir;			/* 1 for left-to-right, -1 for right-to-left */
-  int dirnc;			/* dir * nc */
+  int dir;                      /* 1 for left-to-right, -1 for right-to-left */
+  int dirnc;                    /* dir * nc */
   int ci;
   int row;
   JDIMENSION col;
@@ -639,23 +637,22 @@
 
   for (row = 0; row < num_rows; row++) {
     /* Initialize output values to 0 so can process components separately */
-    jzero_far((void FAR *) output_buf[row],
-	      (size_t) (width * SIZEOF(JSAMPLE)));
+    jzero_far((void *) output_buf[row], (size_t) (width * sizeof(JSAMPLE)));
     for (ci = 0; ci < nc; ci++) {
       input_ptr = input_buf[row] + ci;
       output_ptr = output_buf[row];
       if (cquantize->on_odd_row) {
-	/* work right to left in this row */
-	input_ptr += (width-1) * nc; /* so point to rightmost pixel */
-	output_ptr += width-1;
-	dir = -1;
-	dirnc = -nc;
-	errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
+        /* work right to left in this row */
+        input_ptr += (width-1) * nc; /* so point to rightmost pixel */
+        output_ptr += width-1;
+        dir = -1;
+        dirnc = -nc;
+        errorptr = cquantize->fserrors[ci] + (width+1); /* => entry after last column */
       } else {
-	/* work left to right in this row */
-	dir = 1;
-	dirnc = nc;
-	errorptr = cquantize->fserrors[ci]; /* => entry before first column */
+        /* work left to right in this row */
+        dir = 1;
+        dirnc = nc;
+        errorptr = cquantize->fserrors[ci]; /* => entry before first column */
       }
       colorindex_ci = cquantize->colorindex[ci];
       colormap_ci = cquantize->sv_colormap[ci];
@@ -665,47 +662,47 @@
       belowerr = bpreverr = 0;
 
       for (col = width; col > 0; col--) {
-	/* cur holds the error propagated from the previous pixel on the
-	 * current line.  Add the error propagated from the previous line
-	 * to form the complete error correction term for this pixel, and
-	 * round the error term (which is expressed * 16) to an integer.
-	 * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
-	 * for either sign of the error value.
-	 * Note: errorptr points to *previous* column's array entry.
-	 */
-	cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
-	/* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
-	 * The maximum error is +- MAXJSAMPLE; this sets the required size
-	 * of the range_limit array.
-	 */
-	cur += GETJSAMPLE(*input_ptr);
-	cur = GETJSAMPLE(range_limit[cur]);
-	/* Select output value, accumulate into output code for this pixel */
-	pixcode = GETJSAMPLE(colorindex_ci[cur]);
-	*output_ptr += (JSAMPLE) pixcode;
-	/* Compute actual representation error at this pixel */
-	/* Note: we can do this even though we don't have the final */
-	/* pixel code, because the colormap is orthogonal. */
-	cur -= GETJSAMPLE(colormap_ci[pixcode]);
-	/* Compute error fractions to be propagated to adjacent pixels.
-	 * Add these into the running sums, and simultaneously shift the
-	 * next-line error sums left by 1 column.
-	 */
-	bnexterr = cur;
-	delta = cur * 2;
-	cur += delta;		/* form error * 3 */
-	errorptr[0] = (FSERROR) (bpreverr + cur);
-	cur += delta;		/* form error * 5 */
-	bpreverr = belowerr + cur;
-	belowerr = bnexterr;
-	cur += delta;		/* form error * 7 */
-	/* At this point cur contains the 7/16 error value to be propagated
-	 * to the next pixel on the current line, and all the errors for the
-	 * next line have been shifted over. We are therefore ready to move on.
-	 */
-	input_ptr += dirnc;	/* advance input ptr to next column */
-	output_ptr += dir;	/* advance output ptr to next column */
-	errorptr += dir;	/* advance errorptr to current column */
+        /* cur holds the error propagated from the previous pixel on the
+         * current line.  Add the error propagated from the previous line
+         * to form the complete error correction term for this pixel, and
+         * round the error term (which is expressed * 16) to an integer.
+         * RIGHT_SHIFT rounds towards minus infinity, so adding 8 is correct
+         * for either sign of the error value.
+         * Note: errorptr points to *previous* column's array entry.
+         */
+        cur = RIGHT_SHIFT(cur + errorptr[dir] + 8, 4);
+        /* Form pixel value + error, and range-limit to 0..MAXJSAMPLE.
+         * The maximum error is +- MAXJSAMPLE; this sets the required size
+         * of the range_limit array.
+         */
+        cur += GETJSAMPLE(*input_ptr);
+        cur = GETJSAMPLE(range_limit[cur]);
+        /* Select output value, accumulate into output code for this pixel */
+        pixcode = GETJSAMPLE(colorindex_ci[cur]);
+        *output_ptr += (JSAMPLE) pixcode;
+        /* Compute actual representation error at this pixel */
+        /* Note: we can do this even though we don't have the final */
+        /* pixel code, because the colormap is orthogonal. */
+        cur -= GETJSAMPLE(colormap_ci[pixcode]);
+        /* Compute error fractions to be propagated to adjacent pixels.
+         * Add these into the running sums, and simultaneously shift the
+         * next-line error sums left by 1 column.
+         */
+        bnexterr = cur;
+        delta = cur * 2;
+        cur += delta;           /* form error * 3 */
+        errorptr[0] = (FSERROR) (bpreverr + cur);
+        cur += delta;           /* form error * 5 */
+        bpreverr = belowerr + cur;
+        belowerr = bnexterr;
+        cur += delta;           /* form error * 7 */
+        /* At this point cur contains the 7/16 error value to be propagated
+         * to the next pixel on the current line, and all the errors for the
+         * next line have been shifted over. We are therefore ready to move on.
+         */
+        input_ptr += dirnc;     /* advance input ptr to next column */
+        output_ptr += dir;      /* advance output ptr to next column */
+        errorptr += dir;        /* advance errorptr to current column */
       }
       /* Post-loop cleanup: we must unload the final error value into the
        * final fserrors[] entry.  Note we need not unload belowerr because
@@ -729,7 +726,7 @@
   size_t arraysize;
   int i;
 
-  arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+  arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR));
   for (i = 0; i < cinfo->out_color_components; i++) {
     cquantize->fserrors[i] = (FSERRPTR)
       (*cinfo->mem->alloc_large)((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
@@ -765,7 +762,7 @@
       cquantize->pub.color_quantize = quantize3_ord_dither;
     else
       cquantize->pub.color_quantize = quantize_ord_dither;
-    cquantize->row_index = 0;	/* initialize state for ordered dither */
+    cquantize->row_index = 0;   /* initialize state for ordered dither */
     /* If user changed to ordered dither from another mode,
      * we must recreate the color index table with padding.
      * This will cost extra space, but probably isn't very likely.
@@ -783,9 +780,9 @@
     if (cquantize->fserrors[0] == NULL)
       alloc_fs_workspace(cinfo);
     /* Initialize the propagated errors to zero. */
-    arraysize = (size_t) ((cinfo->output_width + 2) * SIZEOF(FSERROR));
+    arraysize = (size_t) ((cinfo->output_width + 2) * sizeof(FSERROR));
     for (i = 0; i < cinfo->out_color_components; i++)
-      jzero_far((void FAR *) cquantize->fserrors[i], arraysize);
+      jzero_far((void *) cquantize->fserrors[i], arraysize);
     break;
   default:
     ERREXIT(cinfo, JERR_NOT_COMPILED);
@@ -828,13 +825,13 @@
 
   cquantize = (my_cquantize_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_cquantizer));
+                                sizeof(my_cquantizer));
   cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
   cquantize->pub.start_pass = start_pass_1_quant;
   cquantize->pub.finish_pass = finish_pass_1_quant;
   cquantize->pub.new_color_map = new_color_map_1_quant;
   cquantize->fserrors[0] = NULL; /* Flag FS workspace not allocated */
-  cquantize->odither[0] = NULL;	/* Also flag odither arrays not allocated */
+  cquantize->odither[0] = NULL; /* Also flag odither arrays not allocated */
 
   /* Make sure my internal arrays won't overflow */
   if (cinfo->out_color_components > MAX_Q_COMPS)
@@ -848,10 +845,10 @@
   create_colorindex(cinfo);
 
   /* Allocate Floyd-Steinberg workspace now if requested.
-   * We do this now since it is FAR storage and may affect the memory
-   * manager's space calculations.  If the user changes to FS dither
-   * mode in a later pass, we will allocate the space then, and will
-   * possibly overrun the max_memory_to_use setting.
+   * We do this now since it may affect the memory manager's space
+   * calculations.  If the user changes to FS dither mode in a later pass, we
+   * will allocate the space then, and will possibly overrun the
+   * max_memory_to_use setting.
    */
   if (cinfo->dither_mode == JDITHER_FS)
     alloc_fs_workspace(cinfo);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jquant2.c glmark2-2021.02/src/libjpeg-turbo/jquant2.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jquant2.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jquant2.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jquant2.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * Copyright (C) 2009, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2009, 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains 2-pass color quantization (color mapping) routines.
  * These routines provide selection of a custom color map for an image,
@@ -42,7 +44,7 @@
  * color space, and repeatedly splits the "largest" remaining box until we
  * have as many boxes as desired colors.  Then the mean color in each
  * remaining box becomes one of the possible output colors.
- * 
+ *
  * The second pass over the image maps each input pixel to the closest output
  * color (optionally after applying a Floyd-Steinberg dithering correction).
  * This mapping is logically trivial, but making it go fast enough requires
@@ -71,9 +73,9 @@
  * probably need to change these scale factors.
  */
 
-#define R_SCALE 2		/* scale R distances by this much */
-#define G_SCALE 3		/* scale G distances by this much */
-#define B_SCALE 1		/* and B by this much */
+#define R_SCALE 2               /* scale R distances by this much */
+#define G_SCALE 3               /* scale G distances by this much */
+#define B_SCALE 1               /* and B by this much */
 
 static const int c_scales[3]={R_SCALE, G_SCALE, B_SCALE};
 #define C0_SCALE c_scales[rgb_red[cinfo->out_color_space]]
@@ -101,9 +103,7 @@
  * machines, we can't just allocate the histogram in one chunk.  Instead
  * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
  * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
- * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.  Note that
- * on 80x86 machines, the pointer row is in near memory but the actual
- * arrays are in far memory (same arrangement as we use for image arrays).
+ * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.
  */
 
 #define MAXNUMCOLORS  (MAXJSAMPLE+1) /* maximum size of colormap */
@@ -111,9 +111,9 @@
 /* These will do the right thing for either R,G,B or B,G,R color order,
  * but you may not like the results for other color orders.
  */
-#define HIST_C0_BITS  5		/* bits of precision in R/B histogram */
-#define HIST_C1_BITS  6		/* bits of precision in G histogram */
-#define HIST_C2_BITS  5		/* bits of precision in B/R histogram */
+#define HIST_C0_BITS  5         /* bits of precision in R/B histogram */
+#define HIST_C1_BITS  6         /* bits of precision in G histogram */
+#define HIST_C2_BITS  5         /* bits of precision in B/R histogram */
 
 /* Number of elements along histogram axes. */
 #define HIST_C0_ELEMS  (1<<HIST_C0_BITS)
@@ -126,13 +126,13 @@
 #define C2_SHIFT  (BITS_IN_JSAMPLE-HIST_C2_BITS)
 
 
-typedef UINT16 histcell;	/* histogram cell; prefer an unsigned type */
+typedef UINT16 histcell;        /* histogram cell; prefer an unsigned type */
 
-typedef histcell FAR * histptr;	/* for pointers to histogram cells */
+typedef histcell *histptr; /* for pointers to histogram cells */
 
 typedef histcell hist1d[HIST_C2_ELEMS]; /* typedefs for the array */
-typedef hist1d FAR * hist2d;	/* type for the 2nd-level pointers */
-typedef hist2d * hist3d;	/* type for top-level pointer */
+typedef hist1d *hist2d;         /* type for the 2nd-level pointers */
+typedef hist2d *hist3d;         /* type for top-level pointer */
 
 
 /* Declarations for Floyd-Steinberg dithering.
@@ -140,8 +140,8 @@
  * Errors are accumulated into the array fserrors[], at a resolution of
  * 1/16th of a pixel count.  The error at a given pixel is propagated
  * to its not-yet-processed neighbors using the standard F-S fractions,
- *		...	(here)	7/16
- *		3/16	5/16	1/16
+ *              ...     (here)  7/16
+ *              3/16    5/16    1/16
  * We work left-to-right on even rows, right-to-left on odd rows.
  *
  * We can get away with a single array (holding one row's worth of errors)
@@ -154,20 +154,17 @@
  * The fserrors[] array has (#columns + 2) entries; the extra entry at
  * each end saves us from special-casing the first and last pixels.
  * Each entry is three values long, one value for each color component.
- *
- * Note: on a wide image, we might not have enough room in a PC's near data
- * segment to hold the error array; so it is allocated with alloc_large.
  */
 
 #if BITS_IN_JSAMPLE == 8
-typedef INT16 FSERROR;		/* 16 bits should be enough */
-typedef int LOCFSERROR;		/* use 'int' for calculation temps */
+typedef INT16 FSERROR;          /* 16 bits should be enough */
+typedef int LOCFSERROR;         /* use 'int' for calculation temps */
 #else
-typedef INT32 FSERROR;		/* may need more than 16 bits */
-typedef INT32 LOCFSERROR;	/* be sure calculation temps are big enough */
+typedef JLONG FSERROR;          /* may need more than 16 bits */
+typedef JLONG LOCFSERROR;       /* be sure calculation temps are big enough */
 #endif
 
-typedef FSERROR FAR *FSERRPTR;	/* pointer to error array (in FAR storage!) */
+typedef FSERROR *FSERRPTR;      /* pointer to error array */
 
 
 /* Private subobject */
@@ -176,21 +173,21 @@
   struct jpeg_color_quantizer pub; /* public fields */
 
   /* Space for the eventually created colormap is stashed here */
-  JSAMPARRAY sv_colormap;	/* colormap allocated at init time */
-  int desired;			/* desired # of colors = size of colormap */
+  JSAMPARRAY sv_colormap;       /* colormap allocated at init time */
+  int desired;                  /* desired # of colors = size of colormap */
 
   /* Variables for accumulating image statistics */
-  hist3d histogram;		/* pointer to the histogram */
+  hist3d histogram;             /* pointer to the histogram */
 
-  boolean needs_zeroed;		/* TRUE if next pass must zero histogram */
+  boolean needs_zeroed;         /* TRUE if next pass must zero histogram */
 
   /* Variables for Floyd-Steinberg dithering */
-  FSERRPTR fserrors;		/* accumulated errors */
-  boolean on_odd_row;		/* flag to remember which row we are on */
-  int * error_limiter;		/* table for clamping the applied error */
+  FSERRPTR fserrors;            /* accumulated errors */
+  boolean on_odd_row;           /* flag to remember which row we are on */
+  int *error_limiter;           /* table for clamping the applied error */
 } my_cquantizer;
 
-typedef my_cquantizer * my_cquantize_ptr;
+typedef my_cquantizer *my_cquantize_ptr;
 
 
 /*
@@ -204,7 +201,7 @@
 
 METHODDEF(void)
 prescan_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
-		  JSAMPARRAY output_buf, int num_rows)
+                  JSAMPARRAY output_buf, int num_rows)
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   register JSAMPROW ptr;
@@ -219,11 +216,11 @@
     for (col = width; col > 0; col--) {
       /* get pixel value and index into the histogram */
       histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
-			 [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
-			 [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
+                         [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
+                         [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
       /* increment, check for overflow and undo increment if so. */
       if (++(*histp) <= 0)
-	(*histp)--;
+        (*histp)--;
       ptr += 3;
     }
   }
@@ -243,12 +240,12 @@
   int c1min, c1max;
   int c2min, c2max;
   /* The volume (actually 2-norm) of the box */
-  INT32 volume;
+  JLONG volume;
   /* The number of nonzero histogram cells within this box */
   long colorcount;
 } box;
 
-typedef box * boxptr;
+typedef box *boxptr;
 
 
 LOCAL(boxptr)
@@ -260,7 +257,7 @@
   register int i;
   register long maxc = 0;
   boxptr which = NULL;
-  
+
   for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
     if (boxp->colorcount > maxc && boxp->volume > 0) {
       which = boxp;
@@ -278,9 +275,9 @@
 {
   register boxptr boxp;
   register int i;
-  register INT32 maxv = 0;
+  register JLONG maxv = 0;
   boxptr which = NULL;
-  
+
   for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
     if (boxp->volume > maxv) {
       which = boxp;
@@ -301,77 +298,77 @@
   histptr histp;
   int c0,c1,c2;
   int c0min,c0max,c1min,c1max,c2min,c2max;
-  INT32 dist0,dist1,dist2;
+  JLONG dist0,dist1,dist2;
   long ccount;
-  
+
   c0min = boxp->c0min;  c0max = boxp->c0max;
   c1min = boxp->c1min;  c1max = boxp->c1max;
   c2min = boxp->c2min;  c2max = boxp->c2max;
-  
+
   if (c0max > c0min)
     for (c0 = c0min; c0 <= c0max; c0++)
       for (c1 = c1min; c1 <= c1max; c1++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c0min = c0min = c0;
-	    goto have_c0min;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0min = c0min = c0;
+            goto have_c0min;
+          }
       }
  have_c0min:
   if (c0max > c0min)
     for (c0 = c0max; c0 >= c0min; c0--)
       for (c1 = c1min; c1 <= c1max; c1++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c0max = c0max = c0;
-	    goto have_c0max;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c0max = c0max = c0;
+            goto have_c0max;
+          }
       }
  have_c0max:
   if (c1max > c1min)
     for (c1 = c1min; c1 <= c1max; c1++)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c1min = c1min = c1;
-	    goto have_c1min;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1min = c1min = c1;
+            goto have_c1min;
+          }
       }
  have_c1min:
   if (c1max > c1min)
     for (c1 = c1max; c1 >= c1min; c1--)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1][c2min];
-	for (c2 = c2min; c2 <= c2max; c2++)
-	  if (*histp++ != 0) {
-	    boxp->c1max = c1max = c1;
-	    goto have_c1max;
-	  }
+        histp = & histogram[c0][c1][c2min];
+        for (c2 = c2min; c2 <= c2max; c2++)
+          if (*histp++ != 0) {
+            boxp->c1max = c1max = c1;
+            goto have_c1max;
+          }
       }
  have_c1max:
   if (c2max > c2min)
     for (c2 = c2min; c2 <= c2max; c2++)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1min][c2];
-	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
-	  if (*histp != 0) {
-	    boxp->c2min = c2min = c2;
-	    goto have_c2min;
-	  }
+        histp = & histogram[c0][c1min][c2];
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2min = c2min = c2;
+            goto have_c2min;
+          }
       }
  have_c2min:
   if (c2max > c2min)
     for (c2 = c2max; c2 >= c2min; c2--)
       for (c0 = c0min; c0 <= c0max; c0++) {
-	histp = & histogram[c0][c1min][c2];
-	for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
-	  if (*histp != 0) {
-	    boxp->c2max = c2max = c2;
-	    goto have_c2max;
-	  }
+        histp = & histogram[c0][c1min][c2];
+        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
+          if (*histp != 0) {
+            boxp->c2max = c2max = c2;
+            goto have_c2max;
+          }
       }
  have_c2max:
 
@@ -387,16 +384,16 @@
   dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
   dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
   boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2;
-  
+
   /* Now scan remaining volume of box and compute population */
   ccount = 0;
   for (c0 = c0min; c0 <= c0max; c0++)
     for (c1 = c1min; c1 <= c1max; c1++) {
       histp = & histogram[c0][c1][c2min];
       for (c2 = c2min; c2 <= c2max; c2++, histp++)
-	if (*histp != 0) {
-	  ccount++;
-	}
+        if (*histp != 0) {
+          ccount++;
+        }
     }
   boxp->colorcount = ccount;
 }
@@ -404,7 +401,7 @@
 
 LOCAL(int)
 median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
-	    int desired_colors)
+            int desired_colors)
 /* Repeatedly select and split the largest box until we have enough boxes */
 {
   int n,lb;
@@ -420,9 +417,9 @@
     } else {
       b1 = find_biggest_volume(boxlist, numboxes);
     }
-    if (b1 == NULL)		/* no splittable boxes left! */
+    if (b1 == NULL)             /* no splittable boxes left! */
       break;
-    b2 = &boxlist[numboxes];	/* where new box will go */
+    b2 = &boxlist[numboxes];    /* where new box will go */
     /* Copy the color bounds to the new box. */
     b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max;
     b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min;
@@ -494,24 +491,24 @@
   long c0total = 0;
   long c1total = 0;
   long c2total = 0;
-  
+
   c0min = boxp->c0min;  c0max = boxp->c0max;
   c1min = boxp->c1min;  c1max = boxp->c1max;
   c2min = boxp->c2min;  c2max = boxp->c2max;
-  
+
   for (c0 = c0min; c0 <= c0max; c0++)
     for (c1 = c1min; c1 <= c1max; c1++) {
       histp = & histogram[c0][c1][c2min];
       for (c2 = c2min; c2 <= c2max; c2++) {
-	if ((count = *histp++) != 0) {
-	  total += count;
-	  c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
-	  c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
-	  c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
-	}
+        if ((count = *histp++) != 0) {
+          total += count;
+          c0total += ((c0 << C0_SHIFT) + ((1<<C0_SHIFT)>>1)) * count;
+          c1total += ((c1 << C1_SHIFT) + ((1<<C1_SHIFT)>>1)) * count;
+          c2total += ((c2 << C2_SHIFT) + ((1<<C2_SHIFT)>>1)) * count;
+        }
       }
     }
-  
+
   cinfo->colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total);
   cinfo->colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total);
   cinfo->colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total);
@@ -528,7 +525,7 @@
 
   /* Allocate workspace for box list */
   boxlist = (boxptr) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * SIZEOF(box));
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, desired_colors * sizeof(box));
   /* Initialize one box containing whole space */
   numboxes = 1;
   boxlist[0].c0min = 0;
@@ -575,7 +572,7 @@
  * distance from every colormap entry to every histogram cell.  Unfortunately,
  * it needs a work array to hold the best-distance-so-far for each histogram
  * cell (because the inner loop has to be over cells, not colormap entries).
- * The work array elements have to be INT32s, so the work array would need
+ * The work array elements have to be JLONGs, so the work array would need
  * 256Kb at our recommended precision.  This is not feasible in DOS machines.
  *
  * To get around these problems, we apply Thomas' method to compute the
@@ -627,7 +624,7 @@
 
 LOCAL(int)
 find_nearby_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
-		    JSAMPLE colorlist[])
+                    JSAMPLE colorlist[])
 /* Locate the colormap entries close enough to an update box to be candidates
  * for the nearest entry to some cell(s) in the update box.  The update box
  * is specified by the center coordinates of its first cell.  The number of
@@ -641,8 +638,8 @@
   int maxc0, maxc1, maxc2;
   int centerc0, centerc1, centerc2;
   int i, x, ncolors;
-  INT32 minmaxdist, min_dist, max_dist, tdist;
-  INT32 mindist[MAXNUMCOLORS];	/* min distance to colormap entry i */
+  JLONG minmaxdist, min_dist, max_dist, tdist;
+  JLONG mindist[MAXNUMCOLORS];  /* min distance to colormap entry i */
 
   /* Compute true coordinates of update box's upper corner and center.
    * Actually we compute the coordinates of the center of the upper-corner
@@ -684,11 +681,11 @@
       /* within cell range so no contribution to min_dist */
       min_dist = 0;
       if (x <= centerc0) {
-	tdist = (x - maxc0) * C0_SCALE;
-	max_dist = tdist*tdist;
+        tdist = (x - maxc0) * C0_SCALE;
+        max_dist = tdist*tdist;
       } else {
-	tdist = (x - minc0) * C0_SCALE;
-	max_dist = tdist*tdist;
+        tdist = (x - minc0) * C0_SCALE;
+        max_dist = tdist*tdist;
       }
     }
 
@@ -706,11 +703,11 @@
     } else {
       /* within cell range so no contribution to min_dist */
       if (x <= centerc1) {
-	tdist = (x - maxc1) * C1_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - maxc1) * C1_SCALE;
+        max_dist += tdist*tdist;
       } else {
-	tdist = (x - minc1) * C1_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - minc1) * C1_SCALE;
+        max_dist += tdist*tdist;
       }
     }
 
@@ -728,15 +725,15 @@
     } else {
       /* within cell range so no contribution to min_dist */
       if (x <= centerc2) {
-	tdist = (x - maxc2) * C2_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - maxc2) * C2_SCALE;
+        max_dist += tdist*tdist;
       } else {
-	tdist = (x - minc2) * C2_SCALE;
-	max_dist += tdist*tdist;
+        tdist = (x - minc2) * C2_SCALE;
+        max_dist += tdist*tdist;
       }
     }
 
-    mindist[i] = min_dist;	/* save away the results */
+    mindist[i] = min_dist;      /* save away the results */
     if (max_dist < minmaxdist)
       minmaxdist = max_dist;
   }
@@ -756,7 +753,7 @@
 
 LOCAL(void)
 find_best_colors (j_decompress_ptr cinfo, int minc0, int minc1, int minc2,
-		  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
+                  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
 /* Find the closest colormap entry for each cell in the update box,
  * given the list of candidate colors prepared by find_nearby_colors.
  * Return the indexes of the closest entries in the bestcolor[] array.
@@ -766,31 +763,31 @@
 {
   int ic0, ic1, ic2;
   int i, icolor;
-  register INT32 * bptr;	/* pointer into bestdist[] array */
-  JSAMPLE * cptr;		/* pointer into bestcolor[] array */
-  INT32 dist0, dist1;		/* initial distance values */
-  register INT32 dist2;		/* current distance in inner loop */
-  INT32 xx0, xx1;		/* distance increments */
-  register INT32 xx2;
-  INT32 inc0, inc1, inc2;	/* initial values for increments */
+  register JLONG *bptr;         /* pointer into bestdist[] array */
+  JSAMPLE *cptr;                /* pointer into bestcolor[] array */
+  JLONG dist0, dist1;           /* initial distance values */
+  register JLONG dist2;         /* current distance in inner loop */
+  JLONG xx0, xx1;               /* distance increments */
+  register JLONG xx2;
+  JLONG inc0, inc1, inc2;       /* initial values for increments */
   /* This array holds the distance to the nearest-so-far color for each cell */
-  INT32 bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
+  JLONG bestdist[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
 
   /* Initialize best-distance for each cell of the update box */
   bptr = bestdist;
   for (i = BOX_C0_ELEMS*BOX_C1_ELEMS*BOX_C2_ELEMS-1; i >= 0; i--)
     *bptr++ = 0x7FFFFFFFL;
-  
+
   /* For each color selected by find_nearby_colors,
    * compute its distance to the center of each cell in the box.
    * If that's less than best-so-far, update best distance and color number.
    */
-  
+
   /* Nominal steps between cell centers ("x" in Thomas article) */
 #define STEP_C0  ((1 << C0_SHIFT) * C0_SCALE)
 #define STEP_C1  ((1 << C1_SHIFT) * C1_SCALE)
 #define STEP_C2  ((1 << C2_SHIFT) * C2_SCALE)
-  
+
   for (i = 0; i < numcolors; i++) {
     icolor = GETJSAMPLE(colorlist[i]);
     /* Compute (square of) distance from minc0/c1/c2 to this color */
@@ -812,20 +809,20 @@
       dist1 = dist0;
       xx1 = inc1;
       for (ic1 = BOX_C1_ELEMS-1; ic1 >= 0; ic1--) {
-	dist2 = dist1;
-	xx2 = inc2;
-	for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
-	  if (dist2 < *bptr) {
-	    *bptr = dist2;
-	    *cptr = (JSAMPLE) icolor;
-	  }
-	  dist2 += xx2;
-	  xx2 += 2 * STEP_C2 * STEP_C2;
-	  bptr++;
-	  cptr++;
-	}
-	dist1 += xx1;
-	xx1 += 2 * STEP_C1 * STEP_C1;
+        dist2 = dist1;
+        xx2 = inc2;
+        for (ic2 = BOX_C2_ELEMS-1; ic2 >= 0; ic2--) {
+          if (dist2 < *bptr) {
+            *bptr = dist2;
+            *cptr = (JSAMPLE) icolor;
+          }
+          dist2 += xx2;
+          xx2 += 2 * STEP_C2 * STEP_C2;
+          bptr++;
+          cptr++;
+        }
+        dist1 += xx1;
+        xx1 += 2 * STEP_C1 * STEP_C1;
       }
       dist0 += xx0;
       xx0 += 2 * STEP_C0 * STEP_C0;
@@ -842,13 +839,13 @@
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   hist3d histogram = cquantize->histogram;
-  int minc0, minc1, minc2;	/* lower left corner of update box */
+  int minc0, minc1, minc2;      /* lower left corner of update box */
   int ic0, ic1, ic2;
-  register JSAMPLE * cptr;	/* pointer into bestcolor[] array */
-  register histptr cachep;	/* pointer into main cache array */
+  register JSAMPLE *cptr;       /* pointer into bestcolor[] array */
+  register histptr cachep;      /* pointer into main cache array */
   /* This array lists the candidate colormap indexes. */
   JSAMPLE colorlist[MAXNUMCOLORS];
-  int numcolors;		/* number of candidate colors */
+  int numcolors;                /* number of candidate colors */
   /* This array holds the actually closest colormap index for each cell. */
   JSAMPLE bestcolor[BOX_C0_ELEMS * BOX_C1_ELEMS * BOX_C2_ELEMS];
 
@@ -864,7 +861,7 @@
   minc0 = (c0 << BOX_C0_SHIFT) + ((1 << C0_SHIFT) >> 1);
   minc1 = (c1 << BOX_C1_SHIFT) + ((1 << C1_SHIFT) >> 1);
   minc2 = (c2 << BOX_C2_SHIFT) + ((1 << C2_SHIFT) >> 1);
-  
+
   /* Determine which colormap entries are close enough to be candidates
    * for the nearest entry to some cell in the update box.
    */
@@ -872,10 +869,10 @@
 
   /* Determine the actually nearest colors. */
   find_best_colors(cinfo, minc0, minc1, minc2, numcolors, colorlist,
-		   bestcolor);
+                   bestcolor);
 
   /* Save the best color numbers (plus 1) in the main cache array */
-  c0 <<= BOX_C0_LOG;		/* convert ID back to base cell indexes */
+  c0 <<= BOX_C0_LOG;            /* convert ID back to base cell indexes */
   c1 <<= BOX_C1_LOG;
   c2 <<= BOX_C2_LOG;
   cptr = bestcolor;
@@ -883,7 +880,7 @@
     for (ic1 = 0; ic1 < BOX_C1_ELEMS; ic1++) {
       cachep = & histogram[c0+ic0][c1+ic1][c2];
       for (ic2 = 0; ic2 < BOX_C2_ELEMS; ic2++) {
-	*cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
+        *cachep++ = (histcell) (GETJSAMPLE(*cptr++) + 1);
       }
     }
   }
@@ -896,7 +893,7 @@
 
 METHODDEF(void)
 pass2_no_dither (j_decompress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
 /* This version performs no dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
@@ -920,7 +917,7 @@
       /* If we have not seen this color before, find nearest colormap entry */
       /* and update the cache */
       if (*cachep == 0)
-	fill_inverse_cmap(cinfo, c0,c1,c2);
+        fill_inverse_cmap(cinfo, c0,c1,c2);
       /* Now emit the colormap index for this cell */
       *outptr++ = (JSAMPLE) (*cachep - 1);
     }
@@ -930,20 +927,20 @@
 
 METHODDEF(void)
 pass2_fs_dither (j_decompress_ptr cinfo,
-		 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
+                 JSAMPARRAY input_buf, JSAMPARRAY output_buf, int num_rows)
 /* This version performs Floyd-Steinberg dithering */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
   hist3d histogram = cquantize->histogram;
-  register LOCFSERROR cur0, cur1, cur2;	/* current error or pixel value */
+  register LOCFSERROR cur0, cur1, cur2; /* current error or pixel value */
   LOCFSERROR belowerr0, belowerr1, belowerr2; /* error for pixel below cur */
   LOCFSERROR bpreverr0, bpreverr1, bpreverr2; /* error for below/prev col */
-  register FSERRPTR errorptr;	/* => fserrors[] at column before current */
-  JSAMPROW inptr;		/* => current input pixel */
-  JSAMPROW outptr;		/* => current output pixel */
+  register FSERRPTR errorptr;   /* => fserrors[] at column before current */
+  JSAMPROW inptr;               /* => current input pixel */
+  JSAMPROW outptr;              /* => current output pixel */
   histptr cachep;
-  int dir;			/* +1 or -1 depending on direction */
-  int dir3;			/* 3*dir, for advancing inptr & errorptr */
+  int dir;                      /* +1 or -1 depending on direction */
+  int dir3;                     /* 3*dir, for advancing inptr & errorptr */
   int row;
   JDIMENSION col;
   JDIMENSION width = cinfo->output_width;
@@ -959,7 +956,7 @@
     outptr = output_buf[row];
     if (cquantize->on_odd_row) {
       /* work right to left in this row */
-      inptr += (width-1) * 3;	/* so point to rightmost pixel */
+      inptr += (width-1) * 3;   /* so point to rightmost pixel */
       outptr += width-1;
       dir = -1;
       dir3 = -3;
@@ -1011,53 +1008,44 @@
       /* If we have not seen this color before, find nearest colormap */
       /* entry and update the cache */
       if (*cachep == 0)
-	fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
+        fill_inverse_cmap(cinfo, cur0>>C0_SHIFT,cur1>>C1_SHIFT,cur2>>C2_SHIFT);
       /* Now emit the colormap index for this cell */
       { register int pixcode = *cachep - 1;
-	*outptr = (JSAMPLE) pixcode;
-	/* Compute representation error for this pixel */
-	cur0 -= GETJSAMPLE(colormap0[pixcode]);
-	cur1 -= GETJSAMPLE(colormap1[pixcode]);
-	cur2 -= GETJSAMPLE(colormap2[pixcode]);
+        *outptr = (JSAMPLE) pixcode;
+        /* Compute representation error for this pixel */
+        cur0 -= GETJSAMPLE(colormap0[pixcode]);
+        cur1 -= GETJSAMPLE(colormap1[pixcode]);
+        cur2 -= GETJSAMPLE(colormap2[pixcode]);
       }
       /* Compute error fractions to be propagated to adjacent pixels.
        * Add these into the running sums, and simultaneously shift the
        * next-line error sums left by 1 column.
        */
-      { register LOCFSERROR bnexterr, delta;
+      { register LOCFSERROR bnexterr;
 
-	bnexterr = cur0;	/* Process component 0 */
-	delta = cur0 * 2;
-	cur0 += delta;		/* form error * 3 */
-	errorptr[0] = (FSERROR) (bpreverr0 + cur0);
-	cur0 += delta;		/* form error * 5 */
-	bpreverr0 = belowerr0 + cur0;
-	belowerr0 = bnexterr;
-	cur0 += delta;		/* form error * 7 */
-	bnexterr = cur1;	/* Process component 1 */
-	delta = cur1 * 2;
-	cur1 += delta;		/* form error * 3 */
-	errorptr[1] = (FSERROR) (bpreverr1 + cur1);
-	cur1 += delta;		/* form error * 5 */
-	bpreverr1 = belowerr1 + cur1;
-	belowerr1 = bnexterr;
-	cur1 += delta;		/* form error * 7 */
-	bnexterr = cur2;	/* Process component 2 */
-	delta = cur2 * 2;
-	cur2 += delta;		/* form error * 3 */
-	errorptr[2] = (FSERROR) (bpreverr2 + cur2);
-	cur2 += delta;		/* form error * 5 */
-	bpreverr2 = belowerr2 + cur2;
-	belowerr2 = bnexterr;
-	cur2 += delta;		/* form error * 7 */
+        bnexterr = cur0;        /* Process component 0 */
+        errorptr[0] = (FSERROR) (bpreverr0 + cur0 * 3);
+        bpreverr0 = belowerr0 + cur0 * 5;
+        belowerr0 = bnexterr;
+        cur0 *= 7;
+        bnexterr = cur1;        /* Process component 1 */
+        errorptr[1] = (FSERROR) (bpreverr1 + cur1 * 3);
+        bpreverr1 = belowerr1 + cur1 * 5;
+        belowerr1 = bnexterr;
+        cur1 *= 7;
+        bnexterr = cur2;        /* Process component 2 */
+        errorptr[2] = (FSERROR) (bpreverr2 + cur2 * 3);
+        bpreverr2 = belowerr2 + cur2 * 5;
+        belowerr2 = bnexterr;
+        cur2 *= 7;
       }
       /* At this point curN contains the 7/16 error value to be propagated
        * to the next pixel on the current line, and all the errors for the
        * next line have been shifted over.  We are therefore ready to move on.
        */
-      inptr += dir3;		/* Advance pixel pointers to next column */
+      inptr += dir3;            /* Advance pixel pointers to next column */
       outptr += dir;
-      errorptr += dir3;		/* advance errorptr to current column */
+      errorptr += dir3;         /* advance errorptr to current column */
     }
     /* Post-loop cleanup: we must unload the final error values into the
      * final fserrors[] entry.  Note we need not unload belowerrN because
@@ -1092,12 +1080,12 @@
 /* Allocate and fill in the error_limiter table */
 {
   my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
-  int * table;
+  int *table;
   int in, out;
 
   table = (int *) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * SIZEOF(int));
-  table += MAXJSAMPLE;		/* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, (MAXJSAMPLE*2+1) * sizeof(int));
+  table += MAXJSAMPLE;          /* so can index -MAXJSAMPLE .. +MAXJSAMPLE */
   cquantize->error_limiter = table;
 
 #define STEPSIZE ((MAXJSAMPLE+1)/16)
@@ -1180,16 +1168,16 @@
 
     if (cinfo->dither_mode == JDITHER_FS) {
       size_t arraysize = (size_t) ((cinfo->output_width + 2) *
-				   (3 * SIZEOF(FSERROR)));
+                                   (3 * sizeof(FSERROR)));
       /* Allocate Floyd-Steinberg workspace if we didn't already. */
       if (cquantize->fserrors == NULL)
-	cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
-	  ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
+        cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
+          ((j_common_ptr) cinfo, JPOOL_IMAGE, arraysize);
       /* Initialize the propagated errors to zero. */
-      jzero_far((void FAR *) cquantize->fserrors, arraysize);
+      jzero_far((void *) cquantize->fserrors, arraysize);
       /* Make the error-limit table if we didn't already. */
       if (cquantize->error_limiter == NULL)
-	init_error_limit(cinfo);
+        init_error_limit(cinfo);
       cquantize->on_odd_row = FALSE;
     }
 
@@ -1197,8 +1185,8 @@
   /* Zero the histogram or inverse color map, if necessary */
   if (cquantize->needs_zeroed) {
     for (i = 0; i < HIST_C0_ELEMS; i++) {
-      jzero_far((void FAR *) histogram[i],
-		HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+      jzero_far((void *) histogram[i],
+                HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell));
     }
     cquantize->needs_zeroed = FALSE;
   }
@@ -1231,11 +1219,11 @@
 
   cquantize = (my_cquantize_ptr)
     (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
-				SIZEOF(my_cquantizer));
+                                sizeof(my_cquantizer));
   cinfo->cquantize = (struct jpeg_color_quantizer *) cquantize;
   cquantize->pub.start_pass = start_pass_2_quant;
   cquantize->pub.new_color_map = new_color_map_2_quant;
-  cquantize->fserrors = NULL;	/* flag optional arrays not allocated */
+  cquantize->fserrors = NULL;   /* flag optional arrays not allocated */
   cquantize->error_limiter = NULL;
 
   /* Make sure jdmaster didn't give me a case I can't handle */
@@ -1244,17 +1232,17 @@
 
   /* Allocate the histogram/inverse colormap storage */
   cquantize->histogram = (hist3d) (*cinfo->mem->alloc_small)
-    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * SIZEOF(hist2d));
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, HIST_C0_ELEMS * sizeof(hist2d));
   for (i = 0; i < HIST_C0_ELEMS; i++) {
     cquantize->histogram[i] = (hist2d) (*cinfo->mem->alloc_large)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
-       HIST_C1_ELEMS*HIST_C2_ELEMS * SIZEOF(histcell));
+       HIST_C1_ELEMS*HIST_C2_ELEMS * sizeof(histcell));
   }
   cquantize->needs_zeroed = TRUE; /* histogram is garbage now */
 
   /* Allocate storage for the completed colormap, if required.
-   * We do this now since it is FAR storage and may affect
-   * the memory manager's space calculations.
+   * We do this now since it may affect the memory manager's space
+   * calculations.
    */
   if (cinfo->enable_2pass_quant) {
     /* Make sure color count is acceptable */
@@ -1277,14 +1265,15 @@
     cinfo->dither_mode = JDITHER_FS;
 
   /* Allocate Floyd-Steinberg workspace if necessary.
-   * This isn't really needed until pass 2, but again it is FAR storage.
-   * Although we will cope with a later change in dither_mode,
-   * we do not promise to honor max_memory_to_use if dither_mode changes.
+   * This isn't really needed until pass 2, but again it may affect the memory
+   * manager's space calculations.  Although we will cope with a later change
+   * in dither_mode, we do not promise to honor max_memory_to_use if
+   * dither_mode changes.
    */
   if (cinfo->dither_mode == JDITHER_FS) {
     cquantize->fserrors = (FSERRPTR) (*cinfo->mem->alloc_large)
       ((j_common_ptr) cinfo, JPOOL_IMAGE,
-       (size_t) ((cinfo->output_width + 2) * (3 * SIZEOF(FSERROR))));
+       (size_t) ((cinfo->output_width + 2) * (3 * sizeof(FSERROR))));
     /* Might as well create the error-limiting table too. */
     init_error_limit(cinfo);
   }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimddct.h glmark2-2021.02/src/libjpeg-turbo/jsimddct.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimddct.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jsimddct.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,101 +2,73 @@
  * jsimddct.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * 
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jsimd_can_convsamp                jSCanConv
-#define jsimd_can_convsamp_float          jSCanConvF
-#define jsimd_convsamp                    jSConv
-#define jsimd_convsamp_float              jSConvF
-#define jsimd_can_fdct_islow              jSCanFDCTIS
-#define jsimd_can_fdct_ifast              jSCanFDCTIF
-#define jsimd_can_fdct_float              jSCanFDCTFl
-#define jsimd_fdct_islow                  jSFDCTIS
-#define jsimd_fdct_ifast                  jSFDCTIF
-#define jsimd_fdct_float                  jSFDCTFl
-#define jsimd_can_quantize                jSCanQuant
-#define jsimd_can_quantize_float          jSCanQuantF
-#define jsimd_quantize                    jSQuant
-#define jsimd_quantize_float              jSQuantF
-#define jsimd_can_idct_2x2                jSCanIDCT22
-#define jsimd_can_idct_4x4                jSCanIDCT44
-#define jsimd_idct_2x2                    jSIDCT22
-#define jsimd_idct_4x4                    jSIDCT44
-#define jsimd_can_idct_islow              jSCanIDCTIS
-#define jsimd_can_idct_ifast              jSCanIDCTIF
-#define jsimd_can_idct_float              jSCanIDCTFl
-#define jsimd_idct_islow                  jSIDCTIS
-#define jsimd_idct_ifast                  jSIDCTIF
-#define jsimd_idct_float                  jSIDCTFl
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-EXTERN(int) jsimd_can_convsamp JPP((void));
-EXTERN(int) jsimd_can_convsamp_float JPP((void));
-
-EXTERN(void) jsimd_convsamp JPP((JSAMPARRAY sample_data,
-                                 JDIMENSION start_col,
-                                 DCTELEM * workspace));
-EXTERN(void) jsimd_convsamp_float JPP((JSAMPARRAY sample_data,
-                                       JDIMENSION start_col,
-                                       FAST_FLOAT * workspace));
-
-EXTERN(int) jsimd_can_fdct_islow JPP((void));
-EXTERN(int) jsimd_can_fdct_ifast JPP((void));
-EXTERN(int) jsimd_can_fdct_float JPP((void));
-
-EXTERN(void) jsimd_fdct_islow JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_ifast JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_float JPP((FAST_FLOAT * data));
-
-EXTERN(int) jsimd_can_quantize JPP((void));
-EXTERN(int) jsimd_can_quantize_float JPP((void));
-
-EXTERN(void) jsimd_quantize JPP((JCOEFPTR coef_block,
-                                 DCTELEM * divisors,
-                                 DCTELEM * workspace));
-EXTERN(void) jsimd_quantize_float JPP((JCOEFPTR coef_block,
-                                       FAST_FLOAT * divisors,
-                                       FAST_FLOAT * workspace));
-
-EXTERN(int) jsimd_can_idct_2x2 JPP((void));
-EXTERN(int) jsimd_can_idct_4x4 JPP((void));
-
-EXTERN(void) jsimd_idct_2x2 JPP((j_decompress_ptr cinfo,
-                                 jpeg_component_info * compptr,
-                                 JCOEFPTR coef_block,
-                                 JSAMPARRAY output_buf,
-                                 JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4 JPP((j_decompress_ptr cinfo,
-                                 jpeg_component_info * compptr,
-                                 JCOEFPTR coef_block,
-                                 JSAMPARRAY output_buf,
-                                 JDIMENSION output_col));
-
-EXTERN(int) jsimd_can_idct_islow JPP((void));
-EXTERN(int) jsimd_can_idct_ifast JPP((void));
-EXTERN(int) jsimd_can_idct_float JPP((void));
-
-EXTERN(void) jsimd_idct_islow JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
-EXTERN(void) jsimd_idct_float JPP((j_decompress_ptr cinfo,
-                                   jpeg_component_info * compptr,
-                                   JCOEFPTR coef_block,
-                                   JSAMPARRAY output_buf,
-                                   JDIMENSION output_col));
+EXTERN(int) jsimd_can_convsamp (void);
+EXTERN(int) jsimd_can_convsamp_float (void);
 
+EXTERN(void) jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                             DCTELEM *workspace);
+EXTERN(void) jsimd_convsamp_float (JSAMPARRAY sample_data,
+                                   JDIMENSION start_col,
+                                   FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_fdct_islow (void);
+EXTERN(int) jsimd_can_fdct_ifast (void);
+EXTERN(int) jsimd_can_fdct_float (void);
+
+EXTERN(void) jsimd_fdct_islow (DCTELEM *data);
+EXTERN(void) jsimd_fdct_ifast (DCTELEM *data);
+EXTERN(void) jsimd_fdct_float (FAST_FLOAT *data);
+
+EXTERN(int) jsimd_can_quantize (void);
+EXTERN(int) jsimd_can_quantize_float (void);
+
+EXTERN(void) jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                             DCTELEM *workspace);
+EXTERN(void) jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                                   FAST_FLOAT *workspace);
+
+EXTERN(int) jsimd_can_idct_2x2 (void);
+EXTERN(int) jsimd_can_idct_4x4 (void);
+EXTERN(int) jsimd_can_idct_6x6 (void);
+EXTERN(int) jsimd_can_idct_12x12 (void);
+
+EXTERN(void) jsimd_idct_2x2 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_6x6 (j_decompress_ptr cinfo,
+                             jpeg_component_info *compptr,
+                             JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                             JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12 (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+
+EXTERN(int) jsimd_can_idct_islow (void);
+EXTERN(int) jsimd_can_idct_ifast (void);
+EXTERN(int) jsimd_can_idct_float (void);
+
+EXTERN(void) jsimd_idct_islow (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+EXTERN(void) jsimd_idct_ifast (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
+EXTERN(void) jsimd_idct_float (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                               JDIMENSION output_col);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimd.h glmark2-2021.02/src/libjpeg-turbo/jsimd.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimd.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jsimd.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,97 +2,92 @@
  * jsimd.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2011 D. R. Commander
- * 
+ * Copyright 2011, 2014 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  */
 
-/* Short forms of external names for systems with brain-damaged linkers. */
+#include "jchuff.h"             /* Declarations shared with jcphuff.c */
 
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jsimd_can_rgb_ycc                 jSCanRgbYcc
-#define jsimd_can_rgb_gray                jSCanRgbGry
-#define jsimd_can_ycc_rgb                 jSCanYccRgb
-#define jsimd_rgb_ycc_convert             jSRgbYccConv
-#define jsimd_rgb_gray_convert            jSRgbGryConv
-#define jsimd_ycc_rgb_convert             jSYccRgbConv
-#define jsimd_can_h2v2_downsample         jSCanH2V2Down
-#define jsimd_can_h2v1_downsample         jSCanH2V1Down
-#define jsimd_h2v2_downsample             jSH2V2Down
-#define jsimd_h2v1_downsample             jSH2V1Down
-#define jsimd_can_h2v2_upsample           jSCanH2V2Up
-#define jsimd_can_h2v1_upsample           jSCanH2V1Up
-#define jsimd_h2v2_upsample               jSH2V2Up
-#define jsimd_h2v1_upsample               jSH2V1Up
-#define jsimd_can_h2v2_fancy_upsample     jSCanH2V2FUp
-#define jsimd_can_h2v1_fancy_upsample     jSCanH2V1FUp
-#define jsimd_h2v2_fancy_upsample         jSH2V2FUp
-#define jsimd_h2v1_fancy_upsample         jSH2V1FUp
-#define jsimd_can_h2v2_merged_upsample    jSCanH2V2MUp
-#define jsimd_can_h2v1_merged_upsample    jSCanH2V1MUp
-#define jsimd_h2v2_merged_upsample        jSH2V2MUp
-#define jsimd_h2v1_merged_upsample        jSH2V1MUp
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
-
-EXTERN(int) jsimd_can_rgb_ycc JPP((void));
-EXTERN(int) jsimd_can_rgb_gray JPP((void));
-EXTERN(int) jsimd_can_ycc_rgb JPP((void));
+EXTERN(int) jsimd_can_rgb_ycc (void);
+EXTERN(int) jsimd_can_rgb_gray (void);
+EXTERN(int) jsimd_can_ycc_rgb (void);
+EXTERN(int) jsimd_can_ycc_rgb565 (void);
+EXTERN(int) jsimd_c_can_null_convert (void);
 
 EXTERN(void) jsimd_rgb_ycc_convert
-        JPP((j_compress_ptr cinfo,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_rgb_gray_convert
-        JPP((j_compress_ptr cinfo,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_ycc_rgb_convert
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_c_null_convert
+        (j_compress_ptr cinfo, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
-EXTERN(int) jsimd_can_h2v2_downsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_downsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_downsample (void);
+EXTERN(int) jsimd_can_h2v1_downsample (void);
 
 EXTERN(void) jsimd_h2v2_downsample
-        JPP((j_compress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(int) jsimd_can_h2v2_smooth_downsample (void);
+
+EXTERN(void) jsimd_h2v2_smooth_downsample
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
 EXTERN(void) jsimd_h2v1_downsample
-        JPP((j_compress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (j_compress_ptr cinfo, jpeg_component_info *compptr,
+        JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-EXTERN(int) jsimd_can_h2v2_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_upsample (void);
+EXTERN(int) jsimd_can_h2v1_upsample (void);
+EXTERN(int) jsimd_can_int_upsample (void);
 
 EXTERN(void) jsimd_h2v2_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 EXTERN(void) jsimd_h2v1_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_int_upsample
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(int) jsimd_can_h2v2_fancy_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_fancy_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_fancy_upsample (void);
+EXTERN(int) jsimd_can_h2v1_fancy_upsample (void);
 
 EXTERN(void) jsimd_h2v2_fancy_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 EXTERN(void) jsimd_h2v1_fancy_upsample
-        JPP((j_decompress_ptr cinfo, jpeg_component_info * compptr,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(int) jsimd_can_h2v2_merged_upsample JPP((void));
-EXTERN(int) jsimd_can_h2v1_merged_upsample JPP((void));
+EXTERN(int) jsimd_can_h2v2_merged_upsample (void);
+EXTERN(int) jsimd_can_h2v1_merged_upsample (void);
 
 EXTERN(void) jsimd_h2v2_merged_upsample
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-             JSAMPARRAY output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_merged_upsample
-        JPP((j_decompress_ptr cinfo,
-             JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
-             JSAMPARRAY output_buf));
+        (j_decompress_ptr cinfo, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(int) jsimd_can_huff_encode_one_block (void);
 
+EXTERN(JOCTET*) jsimd_huff_encode_one_block
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimd_none.c glmark2-2021.02/src/libjpeg-turbo/jsimd_none.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jsimd_none.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jsimd_none.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,404 @@
+/*
+ * jsimd_none.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains stubs for when there is no SIMD support available.
+ */
+
+#define JPEG_INTERNALS
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "jsimd.h"
+#include "jdct.h"
+#include "jsimddct.h"
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert (j_compress_ptr cinfo,
+                      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                      JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                      JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jstdhuff.c glmark2-2021.02/src/libjpeg-turbo/jstdhuff.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jstdhuff.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jstdhuff.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,135 @@
+/*
+ * jstdhuff.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1998, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to set the default Huffman tables, if they are
+ * not already set.
+ */
+
+/*
+ * Huffman table setup routines
+ */
+
+LOCAL(void)
+add_huff_table (j_common_ptr cinfo,
+                JHUFF_TBL **htblptr, const UINT8 *bits, const UINT8 *val)
+/* Define a Huffman table */
+{
+  int nsymbols, len;
+
+  if (*htblptr == NULL)
+    *htblptr = jpeg_alloc_huff_table(cinfo);
+  else
+    return;
+
+  /* Copy the number-of-symbols-of-each-code-length counts */
+  MEMCOPY((*htblptr)->bits, bits, sizeof((*htblptr)->bits));
+
+  /* Validate the counts.  We do this here mainly so we can copy the right
+   * number of symbols from the val[] array, without risking marching off
+   * the end of memory.  jchuff.c will do a more thorough test later.
+   */
+  nsymbols = 0;
+  for (len = 1; len <= 16; len++)
+    nsymbols += bits[len];
+  if (nsymbols < 1 || nsymbols > 256)
+    ERREXIT(cinfo, JERR_BAD_HUFF_TABLE);
+
+  MEMCOPY((*htblptr)->huffval, val, nsymbols * sizeof(UINT8));
+  MEMZERO(&((*htblptr)->huffval[nsymbols]), (256 - nsymbols) * sizeof(UINT8));
+
+  /* Initialize sent_table FALSE so table will be written to JPEG file. */
+  (*htblptr)->sent_table = FALSE;
+}
+
+
+LOCAL(void)
+std_huff_tables (j_common_ptr cinfo)
+/* Set up the standard Huffman tables (cf. JPEG standard section K.3) */
+/* IMPORTANT: these are only valid for 8-bit data precision! */
+{
+  JHUFF_TBL **dc_huff_tbl_ptrs, **ac_huff_tbl_ptrs;
+
+  static const UINT8 bits_dc_luminance[17] =
+    { /* 0-base */ 0, 0, 1, 5, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_luminance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+  static const UINT8 bits_dc_chrominance[17] =
+    { /* 0-base */ 0, 0, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0 };
+  static const UINT8 val_dc_chrominance[] =
+    { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 };
+
+  static const UINT8 bits_ac_luminance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 3, 3, 2, 4, 3, 5, 5, 4, 4, 0, 0, 1, 0x7d };
+  static const UINT8 val_ac_luminance[] =
+    { 0x01, 0x02, 0x03, 0x00, 0x04, 0x11, 0x05, 0x12,
+      0x21, 0x31, 0x41, 0x06, 0x13, 0x51, 0x61, 0x07,
+      0x22, 0x71, 0x14, 0x32, 0x81, 0x91, 0xa1, 0x08,
+      0x23, 0x42, 0xb1, 0xc1, 0x15, 0x52, 0xd1, 0xf0,
+      0x24, 0x33, 0x62, 0x72, 0x82, 0x09, 0x0a, 0x16,
+      0x17, 0x18, 0x19, 0x1a, 0x25, 0x26, 0x27, 0x28,
+      0x29, 0x2a, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39,
+      0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49,
+      0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59,
+      0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69,
+      0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79,
+      0x7a, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89,
+      0x8a, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98,
+      0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7,
+      0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6,
+      0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3, 0xc4, 0xc5,
+      0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2, 0xd3, 0xd4,
+      0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xe1, 0xe2,
+      0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea,
+      0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+
+  static const UINT8 bits_ac_chrominance[17] =
+    { /* 0-base */ 0, 0, 2, 1, 2, 4, 4, 3, 4, 7, 5, 4, 4, 0, 1, 2, 0x77 };
+  static const UINT8 val_ac_chrominance[] =
+    { 0x00, 0x01, 0x02, 0x03, 0x11, 0x04, 0x05, 0x21,
+      0x31, 0x06, 0x12, 0x41, 0x51, 0x07, 0x61, 0x71,
+      0x13, 0x22, 0x32, 0x81, 0x08, 0x14, 0x42, 0x91,
+      0xa1, 0xb1, 0xc1, 0x09, 0x23, 0x33, 0x52, 0xf0,
+      0x15, 0x62, 0x72, 0xd1, 0x0a, 0x16, 0x24, 0x34,
+      0xe1, 0x25, 0xf1, 0x17, 0x18, 0x19, 0x1a, 0x26,
+      0x27, 0x28, 0x29, 0x2a, 0x35, 0x36, 0x37, 0x38,
+      0x39, 0x3a, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48,
+      0x49, 0x4a, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58,
+      0x59, 0x5a, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
+      0x69, 0x6a, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78,
+      0x79, 0x7a, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87,
+      0x88, 0x89, 0x8a, 0x92, 0x93, 0x94, 0x95, 0x96,
+      0x97, 0x98, 0x99, 0x9a, 0xa2, 0xa3, 0xa4, 0xa5,
+      0xa6, 0xa7, 0xa8, 0xa9, 0xaa, 0xb2, 0xb3, 0xb4,
+      0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xc2, 0xc3,
+      0xc4, 0xc5, 0xc6, 0xc7, 0xc8, 0xc9, 0xca, 0xd2,
+      0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda,
+      0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9,
+      0xea, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8,
+      0xf9, 0xfa };
+
+  if (cinfo->is_decompressor) {
+    dc_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_decompress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  } else {
+    dc_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->dc_huff_tbl_ptrs;
+    ac_huff_tbl_ptrs = ((j_compress_ptr)cinfo)->ac_huff_tbl_ptrs;
+  }
+
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs[0], bits_dc_luminance,
+                 val_dc_luminance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs[0], bits_ac_luminance,
+                 val_ac_luminance);
+  add_huff_table(cinfo, &dc_huff_tbl_ptrs[1], bits_dc_chrominance,
+                 val_dc_chrominance);
+  add_huff_table(cinfo, &ac_huff_tbl_ptrs[1], bits_ac_chrominance,
+                 val_ac_chrominance);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jutils.c glmark2-2021.02/src/libjpeg-turbo/jutils.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jutils.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jutils.c	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,12 @@
 /*
  * jutils.c
  *
+ * This file was part of the Independent JPEG Group's software:
  * Copyright (C) 1991-1996, Thomas G. Lane.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * It was modified by The libjpeg-turbo Project to include only code
+ * relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains tables and miscellaneous utility routines needed
  * for both compression and decompression.
@@ -21,7 +24,7 @@
  * of a DCT block read in natural order (left to right, top to bottom).
  */
 
-#if 0				/* This table is not actually needed in v6a */
+#if 0                           /* This table is not actually needed in v6a */
 
 const int jpeg_zigzag_order[DCTSIZE2] = {
    0,  1,  5,  6, 14, 15, 27, 28,
@@ -87,30 +90,10 @@
 }
 
 
-/* On normal machines we can apply MEMCOPY() and MEMZERO() to sample arrays
- * and coefficient-block arrays.  This won't work on 80x86 because the arrays
- * are FAR and we're assuming a small-pointer memory model.  However, some
- * DOS compilers provide far-pointer versions of memcpy() and memset() even
- * in the small-model libraries.  These will be used if USE_FMEM is defined.
- * Otherwise, the routines below do it the hard way.  (The performance cost
- * is not all that great, because these routines aren't very heavily used.)
- */
-
-#ifndef NEED_FAR_POINTERS	/* normal case, same as regular macros */
-#define FMEMCOPY(dest,src,size)	MEMCOPY(dest,src,size)
-#define FMEMZERO(target,size)	MEMZERO(target,size)
-#else				/* 80x86 case, define if we can */
-#ifdef USE_FMEM
-#define FMEMCOPY(dest,src,size)	_fmemcpy((void FAR *)(dest), (const void FAR *)(src), (size_t)(size))
-#define FMEMZERO(target,size)	_fmemset((void FAR *)(target), 0, (size_t)(size))
-#endif
-#endif
-
-
 GLOBAL(void)
 jcopy_sample_rows (JSAMPARRAY input_array, int source_row,
-		   JSAMPARRAY output_array, int dest_row,
-		   int num_rows, JDIMENSION num_cols)
+                   JSAMPARRAY output_array, int dest_row,
+                   int num_rows, JDIMENSION num_cols)
 /* Copy some rows of samples from one place to another.
  * num_rows rows are copied from input_array[source_row++]
  * to output_array[dest_row++]; these areas may overlap for duplication.
@@ -118,11 +101,7 @@
  */
 {
   register JSAMPROW inptr, outptr;
-#ifdef FMEMCOPY
-  register size_t count = (size_t) (num_cols * SIZEOF(JSAMPLE));
-#else
-  register JDIMENSION count;
-#endif
+  register size_t count = (size_t) (num_cols * sizeof(JSAMPLE));
   register int row;
 
   input_array += source_row;
@@ -131,49 +110,24 @@
   for (row = num_rows; row > 0; row--) {
     inptr = *input_array++;
     outptr = *output_array++;
-#ifdef FMEMCOPY
-    FMEMCOPY(outptr, inptr, count);
-#else
-    for (count = num_cols; count > 0; count--)
-      *outptr++ = *inptr++;	/* needn't bother with GETJSAMPLE() here */
-#endif
+    MEMCOPY(outptr, inptr, count);
   }
 }
 
 
 GLOBAL(void)
 jcopy_block_row (JBLOCKROW input_row, JBLOCKROW output_row,
-		 JDIMENSION num_blocks)
+                 JDIMENSION num_blocks)
 /* Copy a row of coefficient blocks from one place to another. */
 {
-#ifdef FMEMCOPY
-  FMEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * SIZEOF(JCOEF)));
-#else
-  register JCOEFPTR inptr, outptr;
-  register long count;
-
-  inptr = (JCOEFPTR) input_row;
-  outptr = (JCOEFPTR) output_row;
-  for (count = (long) num_blocks * DCTSIZE2; count > 0; count--) {
-    *outptr++ = *inptr++;
-  }
-#endif
+  MEMCOPY(output_row, input_row, num_blocks * (DCTSIZE2 * sizeof(JCOEF)));
 }
 
 
 GLOBAL(void)
-jzero_far (void FAR * target, size_t bytestozero)
-/* Zero out a chunk of FAR memory. */
+jzero_far (void *target, size_t bytestozero)
+/* Zero out a chunk of memory. */
 /* This might be sample-array data, block-array data, or alloc_large data. */
 {
-#ifdef FMEMZERO
-  FMEMZERO(target, bytestozero);
-#else
-  register char FAR * ptr = (char FAR *) target;
-  register size_t count;
-
-  for (count = bytestozero; count > 0; count--) {
-    *ptr++ = 0;
-  }
-#endif
+  MEMZERO(target, bytestozero);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jversion.h glmark2-2021.02/src/libjpeg-turbo/jversion.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/jversion.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/jversion.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,12 @@
 /*
  * jversion.h
  *
- * Copyright (C) 1991-2010, Thomas G. Lane, Guido Vollbeding.
- * Copyright (C) 2010, 2012, D. R. Commander.
- * This file is part of the Independent JPEG Group's software.
- * For conditions of distribution and use, see the accompanying README file.
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, 2012-2016, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
  *
  * This file contains software version identification.
  */
@@ -12,7 +14,7 @@
 
 #if JPEG_LIB_VERSION >= 80
 
-#define JVERSION	"8b  16-May-2010"
+#define JVERSION        "8d  15-Jan-2012"
 
 #elif JPEG_LIB_VERSION >= 70
 
@@ -20,12 +22,28 @@
 
 #else
 
-#define JVERSION	"6b  27-Mar-1998"
+#define JVERSION        "6b  27-Mar-1998"
 
 #endif
 
-#define JCOPYRIGHT	"Copyright (C) 1991-2010 Thomas G. Lane, Guido Vollbeding\n" \
-			"Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
-			"Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \
-			"Copyright (C) 2009-2012 D. R. Commander\n" \
-			"Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)"
+/*
+ * NOTE: It is our convention to place the authors in the following order:
+ * - libjpeg-turbo authors (2009-) in descending order of the date of their
+ *   most recent contribution to the project, then in ascending order of the
+ *   date of their first contribution to the project
+ * - Upstream authors in descending order of the date of the first inclusion of
+ *   their code
+ */
+
+#define JCOPYRIGHT      "Copyright (C) 2009-2016 D. R. Commander\n" \
+                        "Copyright (C) 2011-2016 Siarhei Siamashka\n" \
+                        "Copyright (C) 2015-2016 Matthieu Darbois\n" \
+                        "Copyright (C) 2015 Google, Inc.\n" \
+                        "Copyright (C) 2013-2014 MIPS Technologies, Inc.\n" \
+                        "Copyright (C) 2013 Linaro Limited\n" \
+                        "Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies)\n" \
+                        "Copyright (C) 2009 Pierre Ossman for Cendio AB\n" \
+                        "Copyright (C) 1999-2006 MIYASAKA Masaru\n" \
+                        "Copyright (C) 1991-2016 Thomas G. Lane, Guido Vollbeding" \
+
+#define JCOPYRIGHT_SHORT "Copyright (C) 1991-2016 The libjpeg-turbo Project and many others"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/libjpeg.map.in glmark2-2021.02/src/libjpeg-turbo/libjpeg.map.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/libjpeg.map.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/libjpeg.map.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+LIBJPEGTURBO_@JPEG_LIB_VERSION_DECIMAL@ {
+  @MEM_SRCDST_FUNCTIONS@
+  local:
+    jsimd_*;
+    jconst_*;
+};
+
+LIBJPEG_@JPEG_LIB_VERSION_DECIMAL@ {
+  global:
+    *;
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/libjpeg.txt glmark2-2021.02/src/libjpeg-turbo/libjpeg.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/libjpeg.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/libjpeg.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3105 @@
+USING THE IJG JPEG LIBRARY
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1994-2013, Thomas G. Lane, Guido Vollbeding.
+libjpeg-turbo Modifications:
+Copyright (C) 2010, 2014-2016, D. R. Commander.
+Copyright (C) 2015, Google, Inc.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file describes how to use the IJG JPEG library within an application
+program.  Read it if you want to write a program that uses the library.
+
+The file example.c provides heavily commented skeleton code for calling the
+JPEG library.  Also see jpeglib.h (the include file to be used by application
+programs) for full details about data structures and function parameter lists.
+The library source code, of course, is the ultimate reference.
+
+Note that there have been *major* changes from the application interface
+presented by IJG version 4 and earlier versions.  The old design had several
+inherent limitations, and it had accumulated a lot of cruft as we added
+features while trying to minimize application-interface changes.  We have
+sacrificed backward compatibility in the version 5 rewrite, but we think the
+improvements justify this.
+
+
+TABLE OF CONTENTS
+-----------------
+
+Overview:
+        Functions provided by the library
+        Outline of typical usage
+Basic library usage:
+        Data formats
+        Compression details
+        Decompression details
+        Mechanics of usage: include files, linking, etc
+Advanced features:
+        Compression parameter selection
+        Decompression parameter selection
+        Special color spaces
+        Error handling
+        Compressed data handling (source and destination managers)
+        I/O suspension
+        Progressive JPEG support
+        Buffered-image mode
+        Abbreviated datastreams and multiple images
+        Special markers
+        Raw (downsampled) image data
+        Really raw data: DCT coefficients
+        Progress monitoring
+        Memory management
+        Memory usage
+        Library compile-time options
+        Portability considerations
+
+You should read at least the overview and basic usage sections before trying
+to program with the library.  The sections on advanced features can be read
+if and when you need them.
+
+
+OVERVIEW
+========
+
+Functions provided by the library
+---------------------------------
+
+The IJG JPEG library provides C code to read and write JPEG-compressed image
+files.  The surrounding application program receives or supplies image data a
+scanline at a time, using a straightforward uncompressed image format.  All
+details of color conversion and other preprocessing/postprocessing can be
+handled by the library.
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  The application indirectly selects use of this code
+by specifying the format in which it wishes to supply or receive image data.
+For example, if colormapped output is requested, then the decompression
+library automatically invokes color quantization.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+A word about functions *not* provided by the library.  We handle a subset of
+the ISO JPEG standard; most baseline, extended-sequential, and progressive
+JPEG processes are supported.  (Our subset includes all features now in common
+use.)  Unsupported ISO options include:
+        * Hierarchical storage
+        * Lossless JPEG
+        * DNL marker
+        * Nonintegral subsampling ratios
+We support both 8- and 12-bit data precision, but this is a compile-time
+choice rather than a run-time choice; hence it is difficult to use both
+precisions in a single application.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, this library is
+used by the free LIBTIFF library to support JPEG compression in TIFF.)
+
+
+Outline of typical usage
+------------------------
+
+The rough outline of a JPEG compression operation is:
+
+        Allocate and initialize a JPEG compression object
+        Specify the destination for the compressed data (eg, a file)
+        Set parameters for compression, including image size & colorspace
+        jpeg_start_compress(...);
+        while (scan lines remain to be written)
+                jpeg_write_scanlines(...);
+        jpeg_finish_compress(...);
+        Release the JPEG compression object
+
+A JPEG compression object holds parameters and working state for the JPEG
+library.  We make creation/destruction of the object separate from starting
+or finishing compression of an image; the same object can be re-used for a
+series of image compression operations.  This makes it easy to re-use the
+same parameter settings for a sequence of images.  Re-use of a JPEG object
+also has important implications for processing abbreviated JPEG datastreams,
+as discussed later.
+
+The image data to be compressed is supplied to jpeg_write_scanlines() from
+in-memory buffers.  If the application is doing file-to-file compression,
+reading image data from the source file is the application's responsibility.
+The library emits compressed data by calling a "data destination manager",
+which typically will write the data into a file; but the application can
+provide its own destination manager to do something else.
+
+Similarly, the rough outline of a JPEG decompression operation is:
+
+        Allocate and initialize a JPEG decompression object
+        Specify the source of the compressed data (eg, a file)
+        Call jpeg_read_header() to obtain image info
+        Set parameters for decompression
+        jpeg_start_decompress(...);
+        while (scan lines remain to be read)
+                jpeg_read_scanlines(...);
+        jpeg_finish_decompress(...);
+        Release the JPEG decompression object
+
+This is comparable to the compression outline except that reading the
+datastream header is a separate step.  This is helpful because information
+about the image's size, colorspace, etc is available when the application
+selects decompression parameters.  For example, the application can choose an
+output scaling ratio that will fit the image into the available screen size.
+
+The decompression library obtains compressed data by calling a data source
+manager, which typically will read the data from a file; but other behaviors
+can be obtained with a custom source manager.  Decompressed data is delivered
+into in-memory buffers passed to jpeg_read_scanlines().
+
+It is possible to abort an incomplete compression or decompression operation
+by calling jpeg_abort(); or, if you do not need to retain the JPEG object,
+simply release it by calling jpeg_destroy().
+
+JPEG compression and decompression objects are two separate struct types.
+However, they share some common fields, and certain routines such as
+jpeg_destroy() can work on either type of object.
+
+The JPEG library has no static variables: all state is in the compression
+or decompression object.  Therefore it is possible to process multiple
+compression and decompression operations concurrently, using multiple JPEG
+objects.
+
+Both compression and decompression can be done in an incremental memory-to-
+memory fashion, if suitable source/destination managers are used.  See the
+section on "I/O suspension" for more details.
+
+
+BASIC LIBRARY USAGE
+===================
+
+Data formats
+------------
+
+Before diving into procedural details, it is helpful to understand the
+image data format that the JPEG library expects or returns.
+
+The standard input image format is a rectangular array of pixels, with each
+pixel having the same number of "component" or "sample" values (color
+channels).  You must specify how many components there are and the colorspace
+interpretation of the components.  Most applications will use RGB data
+(three components per pixel) or grayscale data (one component per pixel).
+PLEASE NOTE THAT RGB DATA IS THREE SAMPLES PER PIXEL, GRAYSCALE ONLY ONE.
+A remarkable number of people manage to miss this, only to find that their
+programs don't work with grayscale JPEG files.
+
+There is no provision for colormapped input.  JPEG files are always full-color
+or full grayscale (or sometimes another colorspace such as CMYK).  You can
+feed in a colormapped image by expanding it to full-color format.  However
+JPEG often doesn't work very well with source data that has been colormapped,
+because of dithering noise.  This is discussed in more detail in the JPEG FAQ
+and the other references mentioned in the README.ijg file.
+
+Pixels are stored by scanlines, with each scanline running from left to
+right.  The component values for each pixel are adjacent in the row; for
+example, R,G,B,R,G,B,R,G,B,... for 24-bit RGB color.  Each scanline is an
+array of data type JSAMPLE --- which is typically "unsigned char", unless
+you've changed jmorecfg.h.  (You can also change the RGB pixel layout, say
+to B,G,R order, by modifying jmorecfg.h.  But see the restrictions listed in
+that file before doing so.)
+
+A 2-D array of pixels is formed by making a list of pointers to the starts of
+scanlines; so the scanlines need not be physically adjacent in memory.  Even
+if you process just one scanline at a time, you must make a one-element
+pointer array to conform to this structure.  Pointers to JSAMPLE rows are of
+type JSAMPROW, and the pointer to the pointer array is of type JSAMPARRAY.
+
+The library accepts or supplies one or more complete scanlines per call.
+It is not possible to process part of a row at a time.  Scanlines are always
+processed top-to-bottom.  You can process an entire image in one call if you
+have it all in memory, but usually it's simplest to process one scanline at
+a time.
+
+For best results, source data values should have the precision specified by
+BITS_IN_JSAMPLE (normally 8 bits).  For instance, if you choose to compress
+data that's only 6 bits/channel, you should left-justify each value in a
+byte before passing it to the compressor.  If you need to compress data
+that has more than 8 bits/channel, compile with BITS_IN_JSAMPLE = 12.
+(See "Library compile-time options", later.)
+
+
+The data format returned by the decompressor is the same in all details,
+except that colormapped output is supported.  (Again, a JPEG file is never
+colormapped.  But you can ask the decompressor to perform on-the-fly color
+quantization to deliver colormapped output.)  If you request colormapped
+output then the returned data array contains a single JSAMPLE per pixel;
+its value is an index into a color map.  The color map is represented as
+a 2-D JSAMPARRAY in which each row holds the values of one color component,
+that is, colormap[i][j] is the value of the i'th color component for pixel
+value (map index) j.  Note that since the colormap indexes are stored in
+JSAMPLEs, the maximum number of colors is limited by the size of JSAMPLE
+(ie, at most 256 colors for an 8-bit JPEG library).
+
+
+Compression details
+-------------------
+
+Here we revisit the JPEG compression outline given in the overview.
+
+1. Allocate and initialize a JPEG compression object.
+
+A JPEG compression object is a "struct jpeg_compress_struct".  (It also has
+a bunch of subsidiary structures which are allocated via malloc(), but the
+application doesn't control those directly.)  This struct can be just a local
+variable in the calling routine, if a single routine is going to execute the
+whole JPEG compression sequence.  Otherwise it can be static or allocated
+from malloc().
+
+You will also need a structure representing a JPEG error handler.  The part
+of this that the library cares about is a "struct jpeg_error_mgr".  If you
+are providing your own error handler, you'll typically want to embed the
+jpeg_error_mgr struct in a larger structure; this is discussed later under
+"Error handling".  For now we'll assume you are just using the default error
+handler.  The default error handler will print JPEG error/warning messages
+on stderr, and it will call exit() if a fatal error occurs.
+
+You must initialize the error handler structure, store a pointer to it into
+the JPEG object's "err" field, and then call jpeg_create_compress() to
+initialize the rest of the JPEG object.
+
+Typical code for this step, if you are using the default error handler, is
+
+        struct jpeg_compress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_compress(&cinfo);
+
+jpeg_create_compress allocates a small amount of memory, so it could fail
+if you are out of memory.  In that case it will exit via the error handler;
+that's why the error handler must be initialized first.
+
+
+2. Specify the destination for the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library delivers compressed data to a
+"data destination" module.  The library includes one data destination
+module which knows how to write to a stdio stream.  You can use your own
+destination module if you want to do something else, as discussed later.
+
+If you use the standard destination module, you must open the target stdio
+stream beforehand.  Typical code for this step looks like:
+
+        FILE *outfile;
+        ...
+        if ((outfile = fopen(filename, "wb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_dest(&cinfo, outfile);
+
+where the last line invokes the standard destination module.
+
+WARNING: it is critical that the binary compressed data be delivered to the
+output file unchanged.  On non-Unix systems the stdio library may perform
+newline translation or otherwise corrupt binary data.  To suppress this
+behavior, you may need to use a "b" option to fopen (as shown above), or use
+setmode() or another routine to put the stdio stream in binary mode.  See
+cjpeg.c and djpeg.c for code that has been found to work on many systems.
+
+You can select the data destination after setting other parameters (step 3),
+if that's more convenient.  You may not change the destination between
+calling jpeg_start_compress() and jpeg_finish_compress().
+
+
+3. Set parameters for compression, including image size & colorspace.
+
+You must supply information about the source image by setting the following
+fields in the JPEG object (cinfo structure):
+
+        image_width             Width of image, in pixels
+        image_height            Height of image, in pixels
+        input_components        Number of color channels (samples per pixel)
+        in_color_space          Color space of source image
+
+The image dimensions are, hopefully, obvious.  JPEG supports image dimensions
+of 1 to 64K pixels in either direction.  The input color space is typically
+RGB or grayscale, and input_components is 3 or 1 accordingly.  (See "Special
+color spaces", later, for more info.)  The in_color_space field must be
+assigned one of the J_COLOR_SPACE enum constants, typically JCS_RGB or
+JCS_GRAYSCALE.
+
+JPEG has a large number of compression parameters that determine how the
+image is encoded.  Most applications don't need or want to know about all
+these parameters.  You can set all the parameters to reasonable defaults by
+calling jpeg_set_defaults(); then, if there are particular values you want
+to change, you can do so after that.  The "Compression parameter selection"
+section tells about all the parameters.
+
+You must set in_color_space correctly before calling jpeg_set_defaults(),
+because the defaults depend on the source image colorspace.  However the
+other three source image parameters need not be valid until you call
+jpeg_start_compress().  There's no harm in calling jpeg_set_defaults() more
+than once, if that happens to be convenient.
+
+Typical code for a 24-bit RGB source image is
+
+        cinfo.image_width = Width;      /* image width and height, in pixels */
+        cinfo.image_height = Height;
+        cinfo.input_components = 3;     /* # of color components per pixel */
+        cinfo.in_color_space = JCS_RGB; /* colorspace of input image */
+
+        jpeg_set_defaults(&cinfo);
+        /* Make optional parameter settings here */
+
+
+4. jpeg_start_compress(...);
+
+After you have established the data destination and set all the necessary
+source image info and other parameters, call jpeg_start_compress() to begin
+a compression cycle.  This will initialize internal state, allocate working
+storage, and emit the first few bytes of the JPEG datastream header.
+
+Typical code:
+
+        jpeg_start_compress(&cinfo, TRUE);
+
+The "TRUE" parameter ensures that a complete JPEG interchange datastream
+will be written.  This is appropriate in most cases.  If you think you might
+want to use an abbreviated datastream, read the section on abbreviated
+datastreams, below.
+
+Once you have called jpeg_start_compress(), you may not alter any JPEG
+parameters or other fields of the JPEG object until you have completed
+the compression cycle.
+
+
+5. while (scan lines remain to be written)
+        jpeg_write_scanlines(...);
+
+Now write all the required image data by calling jpeg_write_scanlines()
+one or more times.  You can pass one or more scanlines in each call, up
+to the total image height.  In most applications it is convenient to pass
+just one or a few scanlines at a time.  The expected format for the passed
+data is discussed under "Data formats", above.
+
+Image data should be written in top-to-bottom scanline order.  The JPEG spec
+contains some weasel wording about how top and bottom are application-defined
+terms (a curious interpretation of the English language...) but if you want
+your files to be compatible with everyone else's, you WILL use top-to-bottom
+order.  If the source data must be read in bottom-to-top order, you can use
+the JPEG library's virtual array mechanism to invert the data efficiently.
+Examples of this can be found in the sample application cjpeg.
+
+The library maintains a count of the number of scanlines written so far
+in the next_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.next_scanline < cinfo.image_height)".
+
+Code for this step depends heavily on the way that you store the source data.
+example.c shows the following code for the case of a full-size 2-D source
+array containing 3-byte RGB pixels:
+
+        JSAMPROW row_pointer[1];        /* pointer to a single row */
+        int row_stride;                 /* physical row width in buffer */
+
+        row_stride = image_width * 3;   /* JSAMPLEs per row in image_buffer */
+
+        while (cinfo.next_scanline < cinfo.image_height) {
+            row_pointer[0] = & image_buffer[cinfo.next_scanline * row_stride];
+            jpeg_write_scanlines(&cinfo, row_pointer, 1);
+        }
+
+jpeg_write_scanlines() returns the number of scanlines actually written.
+This will normally be equal to the number passed in, so you can usually
+ignore the return value.  It is different in just two cases:
+  * If you try to write more scanlines than the declared image height,
+    the additional scanlines are ignored.
+  * If you use a suspending data destination manager, output buffer overrun
+    will cause the compressor to return before accepting all the passed lines.
+    This feature is discussed under "I/O suspension", below.  The normal
+    stdio destination manager will NOT cause this to happen.
+In any case, the return value is the same as the change in the value of
+next_scanline.
+
+
+6. jpeg_finish_compress(...);
+
+After all the image data has been written, call jpeg_finish_compress() to
+complete the compression cycle.  This step is ESSENTIAL to ensure that the
+last bufferload of data is written to the data destination.
+jpeg_finish_compress() also releases working memory associated with the JPEG
+object.
+
+Typical code:
+
+        jpeg_finish_compress(&cinfo);
+
+If using the stdio destination manager, don't forget to close the output
+stdio stream (if necessary) afterwards.
+
+If you have requested a multi-pass operating mode, such as Huffman code
+optimization, jpeg_finish_compress() will perform the additional passes using
+data buffered by the first pass.  In this case jpeg_finish_compress() may take
+quite a while to complete.  With the default compression parameters, this will
+not happen.
+
+It is an error to call jpeg_finish_compress() before writing the necessary
+total number of scanlines.  If you wish to abort compression, call
+jpeg_abort() as discussed below.
+
+After completing a compression cycle, you may dispose of the JPEG object
+as discussed next, or you may use it to compress another image.  In that case
+return to step 2, 3, or 4 as appropriate.  If you do not change the
+destination manager, the new datastream will be written to the same target.
+If you do not change any JPEG parameters, the new datastream will be written
+with the same parameters as before.  Note that you can change the input image
+dimensions freely between cycles, but if you change the input colorspace, you
+should call jpeg_set_defaults() to adjust for the new colorspace; and then
+you'll need to repeat all of step 3.
+
+
+7. Release the JPEG compression object.
+
+When you are done with a JPEG compression object, destroy it by calling
+jpeg_destroy_compress().  This will free all subsidiary memory (regardless of
+the previous state of the object).  Or you can call jpeg_destroy(), which
+works for either compression or decompression objects --- this may be more
+convenient if you are sharing code between compression and decompression
+cases.  (Actually, these routines are equivalent except for the declared type
+of the passed pointer.  To avoid gripes from ANSI C compilers, jpeg_destroy()
+should be passed a j_common_ptr.)
+
+If you allocated the jpeg_compress_struct structure from malloc(), freeing
+it is your responsibility --- jpeg_destroy() won't.  Ditto for the error
+handler structure.
+
+Typical code:
+
+        jpeg_destroy_compress(&cinfo);
+
+
+8. Aborting.
+
+If you decide to abort a compression cycle before finishing, you can clean up
+in either of two ways:
+
+* If you don't need the JPEG object any more, just call
+  jpeg_destroy_compress() or jpeg_destroy() to release memory.  This is
+  legitimate at any point after calling jpeg_create_compress() --- in fact,
+  it's safe even if jpeg_create_compress() fails.
+
+* If you want to re-use the JPEG object, call jpeg_abort_compress(), or call
+  jpeg_abort() which works on both compression and decompression objects.
+  This will return the object to an idle state, releasing any working memory.
+  jpeg_abort() is allowed at any time after successful object creation.
+
+Note that cleaning up the data destination, if required, is your
+responsibility; neither of these routines will call term_destination().
+(See "Compressed data handling", below, for more about that.)
+
+jpeg_destroy() and jpeg_abort() are the only safe calls to make on a JPEG
+object that has reported an error by calling error_exit (see "Error handling"
+for more info).  The internal state of such an object is likely to be out of
+whack.  Either of these two routines will return the object to a known state.
+
+
+Decompression details
+---------------------
+
+Here we revisit the JPEG decompression outline given in the overview.
+
+1. Allocate and initialize a JPEG decompression object.
+
+This is just like initialization for compression, as discussed above,
+except that the object is a "struct jpeg_decompress_struct" and you
+call jpeg_create_decompress().  Error handling is exactly the same.
+
+Typical code:
+
+        struct jpeg_decompress_struct cinfo;
+        struct jpeg_error_mgr jerr;
+        ...
+        cinfo.err = jpeg_std_error(&jerr);
+        jpeg_create_decompress(&cinfo);
+
+(Both here and in the IJG code, we usually use variable name "cinfo" for
+both compression and decompression objects.)
+
+
+2. Specify the source of the compressed data (eg, a file).
+
+As previously mentioned, the JPEG library reads compressed data from a "data
+source" module.  The library includes one data source module which knows how
+to read from a stdio stream.  You can use your own source module if you want
+to do something else, as discussed later.
+
+If you use the standard source module, you must open the source stdio stream
+beforehand.  Typical code for this step looks like:
+
+        FILE *infile;
+        ...
+        if ((infile = fopen(filename, "rb")) == NULL) {
+            fprintf(stderr, "can't open %s\n", filename);
+            exit(1);
+        }
+        jpeg_stdio_src(&cinfo, infile);
+
+where the last line invokes the standard source module.
+
+WARNING: it is critical that the binary compressed data be read unchanged.
+On non-Unix systems the stdio library may perform newline translation or
+otherwise corrupt binary data.  To suppress this behavior, you may need to use
+a "b" option to fopen (as shown above), or use setmode() or another routine to
+put the stdio stream in binary mode.  See cjpeg.c and djpeg.c for code that
+has been found to work on many systems.
+
+You may not change the data source between calling jpeg_read_header() and
+jpeg_finish_decompress().  If you wish to read a series of JPEG images from
+a single source file, you should repeat the jpeg_read_header() to
+jpeg_finish_decompress() sequence without reinitializing either the JPEG
+object or the data source module; this prevents buffered input data from
+being discarded.
+
+
+3. Call jpeg_read_header() to obtain image info.
+
+Typical code for this step is just
+
+        jpeg_read_header(&cinfo, TRUE);
+
+This will read the source datastream header markers, up to the beginning
+of the compressed data proper.  On return, the image dimensions and other
+info have been stored in the JPEG object.  The application may wish to
+consult this information before selecting decompression parameters.
+
+More complex code is necessary if
+  * A suspending data source is used --- in that case jpeg_read_header()
+    may return before it has read all the header data.  See "I/O suspension",
+    below.  The normal stdio source manager will NOT cause this to happen.
+  * Abbreviated JPEG files are to be processed --- see the section on
+    abbreviated datastreams.  Standard applications that deal only in
+    interchange JPEG files need not be concerned with this case either.
+
+It is permissible to stop at this point if you just wanted to find out the
+image dimensions and other header info for a JPEG file.  In that case,
+call jpeg_destroy() when you are done with the JPEG object, or call
+jpeg_abort() to return it to an idle state before selecting a new data
+source and reading another header.
+
+
+4. Set parameters for decompression.
+
+jpeg_read_header() sets appropriate default decompression parameters based on
+the properties of the image (in particular, its colorspace).  However, you
+may well want to alter these defaults before beginning the decompression.
+For example, the default is to produce full color output from a color file.
+If you want colormapped output you must ask for it.  Other options allow the
+returned image to be scaled and allow various speed/quality tradeoffs to be
+selected.  "Decompression parameter selection", below, gives details.
+
+If the defaults are appropriate, nothing need be done at this step.
+
+Note that all default values are set by each call to jpeg_read_header().
+If you reuse a decompression object, you cannot expect your parameter
+settings to be preserved across cycles, as you can for compression.
+You must set desired parameter values each time.
+
+
+5. jpeg_start_decompress(...);
+
+Once the parameter values are satisfactory, call jpeg_start_decompress() to
+begin decompression.  This will initialize internal state, allocate working
+memory, and prepare for returning data.
+
+Typical code is just
+
+        jpeg_start_decompress(&cinfo);
+
+If you have requested a multi-pass operating mode, such as 2-pass color
+quantization, jpeg_start_decompress() will do everything needed before data
+output can begin.  In this case jpeg_start_decompress() may take quite a while
+to complete.  With a single-scan (non progressive) JPEG file and default
+decompression parameters, this will not happen; jpeg_start_decompress() will
+return quickly.
+
+After this call, the final output image dimensions, including any requested
+scaling, are available in the JPEG object; so is the selected colormap, if
+colormapped output has been requested.  Useful fields include
+
+        output_width            image width and height, as scaled
+        output_height
+        out_color_components    # of color components in out_color_space
+        output_components       # of color components returned per pixel
+        colormap                the selected colormap, if any
+        actual_number_of_colors         number of entries in colormap
+
+output_components is 1 (a colormap index) when quantizing colors; otherwise it
+equals out_color_components.  It is the number of JSAMPLE values that will be
+emitted per pixel in the output arrays.
+
+Typically you will need to allocate data buffers to hold the incoming image.
+You will need output_width * output_components JSAMPLEs per scanline in your
+output buffer, and a total of output_height scanlines will be returned.
+
+Note: if you are using the JPEG library's internal memory manager to allocate
+data buffers (as djpeg does), then the manager's protocol requires that you
+request large buffers *before* calling jpeg_start_decompress().  This is a
+little tricky since the output_XXX fields are not normally valid then.  You
+can make them valid by calling jpeg_calc_output_dimensions() after setting the
+relevant parameters (scaling, output color space, and quantization flag).
+
+
+6. while (scan lines remain to be read)
+        jpeg_read_scanlines(...);
+
+Now you can read the decompressed image data by calling jpeg_read_scanlines()
+one or more times.  At each call, you pass in the maximum number of scanlines
+to be read (ie, the height of your working buffer); jpeg_read_scanlines()
+will return up to that many lines.  The return value is the number of lines
+actually read.  The format of the returned data is discussed under "Data
+formats", above.  Don't forget that grayscale and color JPEGs will return
+different data formats!
+
+Image data is returned in top-to-bottom scanline order.  If you must write
+out the image in bottom-to-top order, you can use the JPEG library's virtual
+array mechanism to invert the data efficiently.  Examples of this can be
+found in the sample application djpeg.
+
+The library maintains a count of the number of scanlines returned so far
+in the output_scanline field of the JPEG object.  Usually you can just use
+this variable as the loop counter, so that the loop test looks like
+"while (cinfo.output_scanline < cinfo.output_height)".  (Note that the test
+should NOT be against image_height, unless you never use scaling.  The
+image_height field is the height of the original unscaled image.)
+The return value always equals the change in the value of output_scanline.
+
+If you don't use a suspending data source, it is safe to assume that
+jpeg_read_scanlines() reads at least one scanline per call, until the
+bottom of the image has been reached.
+
+If you use a buffer larger than one scanline, it is NOT safe to assume that
+jpeg_read_scanlines() fills it.  (The current implementation returns only a
+few scanlines per call, no matter how large a buffer you pass.)  So you must
+always provide a loop that calls jpeg_read_scanlines() repeatedly until the
+whole image has been read.
+
+
+7. jpeg_finish_decompress(...);
+
+After all the image data has been read, call jpeg_finish_decompress() to
+complete the decompression cycle.  This causes working memory associated
+with the JPEG object to be released.
+
+Typical code:
+
+        jpeg_finish_decompress(&cinfo);
+
+If using the stdio source manager, don't forget to close the source stdio
+stream if necessary.
+
+It is an error to call jpeg_finish_decompress() before reading the correct
+total number of scanlines.  If you wish to abort decompression, call
+jpeg_abort() as discussed below.
+
+After completing a decompression cycle, you may dispose of the JPEG object as
+discussed next, or you may use it to decompress another image.  In that case
+return to step 2 or 3 as appropriate.  If you do not change the source
+manager, the next image will be read from the same source.
+
+
+8. Release the JPEG decompression object.
+
+When you are done with a JPEG decompression object, destroy it by calling
+jpeg_destroy_decompress() or jpeg_destroy().  The previous discussion of
+destroying compression objects applies here too.
+
+Typical code:
+
+        jpeg_destroy_decompress(&cinfo);
+
+
+9. Aborting.
+
+You can abort a decompression cycle by calling jpeg_destroy_decompress() or
+jpeg_destroy() if you don't need the JPEG object any more, or
+jpeg_abort_decompress() or jpeg_abort() if you want to reuse the object.
+The previous discussion of aborting compression cycles applies here too.
+
+
+Partial image decompression
+---------------------------
+
+Partial image decompression is convenient for performance-critical applications
+that wish to view only a portion of a large JPEG image without decompressing
+the whole thing.  It it also useful in memory-constrained environments (such as
+on mobile devices.)  This library provides the following functions to support
+partial image decompression:
+
+1. Skipping rows when decompressing
+
+        jpeg_skip_scanlines(j_decompress_ptr cinfo, JDIMENSION num_lines);
+
+This function provides application programmers with the ability to skip over
+multiple rows in the JPEG image.
+
+Suspending data sources are not supported by this function.  Calling
+jpeg_skip_scanlines() with a suspending data source will result in undefined
+behavior.
+
+jpeg_skip_scanlines() will not allow skipping past the bottom of the image.  If
+the value of num_lines is large enough to skip past the bottom of the image,
+then the function will skip to the end of the image instead.
+
+If the value of num_lines is valid, then jpeg_skip_scanlines() will always
+skip all of the input rows requested.  There is no need to inspect the return
+value of the function in that case.
+
+Best results will be achieved by calling jpeg_skip_scanlines() for large chunks
+of rows.  The function should be viewed as a way to quickly jump to a
+particular vertical offset in the JPEG image in order to decode a subset of the
+image.  Used in this manner, it will provide significant performance
+improvements.
+
+Calling jpeg_skip_scanlines() for small values of num_lines has several
+potential drawbacks:
+    1) JPEG decompression occurs in blocks, so if jpeg_skip_scanlines() is
+       called from the middle of a decompression block, then it is likely that
+       much of the decompression work has already been done for the first
+       couple of rows that need to be skipped.
+    2) When this function returns, it must leave the decompressor in a state
+       such that it is ready to read the next line.  This may involve
+       decompressing a block that must be partially skipped.
+These issues are especially tricky for cases in which upsampling requires
+context rows.  In the worst case, jpeg_skip_scanlines() will perform similarly
+to jpeg_read_scanlines() (since it will actually call jpeg_read_scanlines().)
+
+2. Decompressing partial scanlines
+
+        jpeg_crop_scanline (j_decompress_ptr cinfo, JDIMENSION *xoffset,
+                            JDIMENSION *width)
+
+This function provides application programmers with the ability to decompress
+only a portion of each row in the JPEG image.  It must be called after
+jpeg_start_decompress() and before any calls to jpeg_read_scanlines() or
+jpeg_skip_scanlines().
+
+If xoffset and width do not form a valid subset of the image row, then this
+function will generate an error.  Note that if the output image is scaled, then
+xoffset and width are relative to the scaled image dimensions.
+
+xoffset and width are passed by reference because xoffset must fall on an iMCU
+boundary.  If it doesn't, then it will be moved left to the nearest iMCU
+boundary, and width will be increased accordingly.  If the calling program does
+not like the adjusted values of xoffset and width, then it can call
+jpeg_crop_scanline() again with new values (for instance, if it wants to move
+xoffset to the nearest iMCU boundary to the right instead of to the left.)
+
+After calling this function, cinfo->output_width will be set to the adjusted
+width.  This value should be used when allocating an output buffer to pass to
+jpeg_read_scanlines().
+
+The output image from a partial-width decompression will be identical to the
+corresponding image region from a full decode, with one exception:  The "fancy"
+(smooth) h2v2 (4:2:0) and h2v1 (4:2:2) upsampling algorithms fill in the
+missing chroma components by averaging the chroma components from neighboring
+pixels, except on the right and left edges of the image (where there are no
+neighboring pixels.)  When performing a partial-width decompression, these
+"fancy" upsampling algorithms may treat the left and right edges of the partial
+image region as if they are the left and right edges of the image, meaning that
+the upsampling algorithm may be simplified.  The result is that the pixels on
+the left or right edge of the partial image may not be exactly identical to the
+corresponding pixels in the original image.
+
+
+Mechanics of usage: include files, linking, etc
+-----------------------------------------------
+
+Applications using the JPEG library should include the header file jpeglib.h
+to obtain declarations of data types and routines.  Before including
+jpeglib.h, include system headers that define at least the typedefs FILE and
+size_t.  On ANSI-conforming systems, including <stdio.h> is sufficient; on
+older Unix systems, you may need <sys/types.h> to define size_t.
+
+If the application needs to refer to individual JPEG library error codes, also
+include jerror.h to define those symbols.
+
+jpeglib.h indirectly includes the files jconfig.h and jmorecfg.h.  If you are
+installing the JPEG header files in a system directory, you will want to
+install all four files: jpeglib.h, jerror.h, jconfig.h, jmorecfg.h.
+
+The most convenient way to include the JPEG code into your executable program
+is to prepare a library file ("libjpeg.a", or a corresponding name on non-Unix
+machines) and reference it at your link step.  If you use only half of the
+library (only compression or only decompression), only that much code will be
+included from the library, unless your linker is hopelessly brain-damaged.
+The supplied makefiles build libjpeg.a automatically (see install.txt).
+
+While you can build the JPEG library as a shared library if the whim strikes
+you, we don't really recommend it.  The trouble with shared libraries is that
+at some point you'll probably try to substitute a new version of the library
+without recompiling the calling applications.  That generally doesn't work
+because the parameter struct declarations usually change with each new
+version.  In other words, the library's API is *not* guaranteed binary
+compatible across versions; we only try to ensure source-code compatibility.
+(In hindsight, it might have been smarter to hide the parameter structs from
+applications and introduce a ton of access functions instead.  Too late now,
+however.)
+
+It may be worth pointing out that the core JPEG library does not actually
+require the stdio library: only the default source/destination managers and
+error handler need it.  You can use the library in a stdio-less environment
+if you replace those modules and use jmemnobs.c (or another memory manager of
+your own devising).  More info about the minimum system library requirements
+may be found in jinclude.h.
+
+
+ADVANCED FEATURES
+=================
+
+Compression parameter selection
+-------------------------------
+
+This section describes all the optional parameters you can set for JPEG
+compression, as well as the "helper" routines provided to assist in this
+task.  Proper setting of some parameters requires detailed understanding
+of the JPEG standard; if you don't know what a parameter is for, it's best
+not to mess with it!  See REFERENCES in the README.ijg file for pointers to
+more info about JPEG.
+
+It's a good idea to call jpeg_set_defaults() first, even if you plan to set
+all the parameters; that way your code is more likely to work with future JPEG
+libraries that have additional parameters.  For the same reason, we recommend
+you use a helper routine where one is provided, in preference to twiddling
+cinfo fields directly.
+
+The helper routines are:
+
+jpeg_set_defaults (j_compress_ptr cinfo)
+        This routine sets all JPEG parameters to reasonable defaults, using
+        only the input image's color space (field in_color_space, which must
+        already be set in cinfo).  Many applications will only need to use
+        this routine and perhaps jpeg_set_quality().
+
+jpeg_set_colorspace (j_compress_ptr cinfo, J_COLOR_SPACE colorspace)
+        Sets the JPEG file's colorspace (field jpeg_color_space) as specified,
+        and sets other color-space-dependent parameters appropriately.  See
+        "Special color spaces", below, before using this.  A large number of
+        parameters, including all per-component parameters, are set by this
+        routine; if you want to twiddle individual parameters you should call
+        jpeg_set_colorspace() before rather than after.
+
+jpeg_default_colorspace (j_compress_ptr cinfo)
+        Selects an appropriate JPEG colorspace based on cinfo->in_color_space,
+        and calls jpeg_set_colorspace().  This is actually a subroutine of
+        jpeg_set_defaults().  It's broken out in case you want to change
+        just the colorspace-dependent JPEG parameters.
+
+jpeg_set_quality (j_compress_ptr cinfo, int quality, boolean force_baseline)
+        Constructs JPEG quantization tables appropriate for the indicated
+        quality setting.  The quality value is expressed on the 0..100 scale
+        recommended by IJG (cjpeg's "-quality" switch uses this routine).
+        Note that the exact mapping from quality values to tables may change
+        in future IJG releases as more is learned about DCT quantization.
+        If the force_baseline parameter is TRUE, then the quantization table
+        entries are constrained to the range 1..255 for full JPEG baseline
+        compatibility.  In the current implementation, this only makes a
+        difference for quality settings below 25, and it effectively prevents
+        very small/low quality files from being generated.  The IJG decoder
+        is capable of reading the non-baseline files generated at low quality
+        settings when force_baseline is FALSE, but other decoders may not be.
+
+jpeg_set_linear_quality (j_compress_ptr cinfo, int scale_factor,
+                         boolean force_baseline)
+        Same as jpeg_set_quality() except that the generated tables are the
+        sample tables given in the JPEC spec section K.1, multiplied by the
+        specified scale factor (which is expressed as a percentage; thus
+        scale_factor = 100 reproduces the spec's tables).  Note that larger
+        scale factors give lower quality.  This entry point is useful for
+        conforming to the Adobe PostScript DCT conventions, but we do not
+        recommend linear scaling as a user-visible quality scale otherwise.
+        force_baseline again constrains the computed table entries to 1..255.
+
+int jpeg_quality_scaling (int quality)
+        Converts a value on the IJG-recommended quality scale to a linear
+        scaling percentage.  Note that this routine may change or go away
+        in future releases --- IJG may choose to adopt a scaling method that
+        can't be expressed as a simple scalar multiplier, in which case the
+        premise of this routine collapses.  Caveat user.
+
+jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
+        [libjpeg v7+ API/ABI emulation only]
+        Set default quantization tables with linear q_scale_factor[] values
+        (see below).
+
+jpeg_add_quant_table (j_compress_ptr cinfo, int which_tbl,
+                      const unsigned int *basic_table,
+                      int scale_factor, boolean force_baseline)
+        Allows an arbitrary quantization table to be created.  which_tbl
+        indicates which table slot to fill.  basic_table points to an array
+        of 64 unsigned ints given in normal array order.  These values are
+        multiplied by scale_factor/100 and then clamped to the range 1..65535
+        (or to 1..255 if force_baseline is TRUE).
+        CAUTION: prior to library version 6a, jpeg_add_quant_table expected
+        the basic table to be given in JPEG zigzag order.  If you need to
+        write code that works with either older or newer versions of this
+        routine, you must check the library version number.  Something like
+        "#if JPEG_LIB_VERSION >= 61" is the right test.
+
+jpeg_simple_progression (j_compress_ptr cinfo)
+        Generates a default scan script for writing a progressive-JPEG file.
+        This is the recommended method of creating a progressive file,
+        unless you want to make a custom scan sequence.  You must ensure that
+        the JPEG color space is set correctly before calling this routine.
+
+
+Compression parameters (cinfo fields) include:
+
+boolean arith_code
+	If TRUE, use arithmetic coding.
+	If FALSE, use Huffman coding.
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: slow but accurate integer algorithm
+                JDCT_IFAST: faster, less accurate integer method
+                JDCT_FLOAT: floating-point method
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        In libjpeg-turbo, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW when using the x86/x86-64 SIMD extensions (results may vary
+        with other SIMD implementations, or when using libjpeg-turbo without
+        SIMD extensions.)  For quality levels of 90 and below, there should be
+        little or no perceptible difference between the two algorithms.  For
+        quality levels above 90, however, the difference between JDCT_IFAST and
+        JDCT_ISLOW becomes more pronounced.  With quality=97, for instance,
+        JDCT_IFAST incurs generally about a 1-3 dB loss (in PSNR) relative to
+        JDCT_ISLOW, but this can be larger for some images.  Do not use
+        JDCT_IFAST with quality levels above 97.  The algorithm often
+        degenerates at quality=98 and above and can actually produce a more
+        lossy image than if lower quality levels had been used.  Also, in
+        libjpeg-turbo, JDCT_IFAST is not fully accelerated for quality levels
+        above 97, so it will be slower than JDCT_ISLOW.  JDCT_FLOAT is mainly a
+        legacy feature.  It does not produce significantly more accurate
+        results than the ISLOW method, and it is much slower.  The FLOAT method
+        may also give different results on different machines due to varying
+        roundoff behavior, whereas the integer methods should give the same
+        results on all machines.
+
+J_COLOR_SPACE jpeg_color_space
+int num_components
+        The JPEG color space and corresponding number of components; see
+        "Special color spaces", below, for more info.  We recommend using
+        jpeg_set_color_space() if you want to change these.
+
+boolean optimize_coding
+        TRUE causes the compressor to compute optimal Huffman coding tables
+        for the image.  This requires an extra pass over the data and
+        therefore costs a good deal of space and time.  The default is
+        FALSE, which tells the compressor to use the supplied or default
+        Huffman tables.  In most cases optimal tables save only a few percent
+        of file size compared to the default tables.  Note that when this is
+        TRUE, you need not supply Huffman tables at all, and any you do
+        supply will be overwritten.
+
+unsigned int restart_interval
+int restart_in_rows
+        To emit restart markers in the JPEG file, set one of these nonzero.
+        Set restart_interval to specify the exact interval in MCU blocks.
+        Set restart_in_rows to specify the interval in MCU rows.  (If
+        restart_in_rows is not 0, then restart_interval is set after the
+        image width in MCUs is computed.)  Defaults are zero (no restarts).
+        One restart marker per MCU row is often a good choice.
+        NOTE: the overhead of restart markers is higher in grayscale JPEG
+        files than in color files, and MUCH higher in progressive JPEGs.
+        If you use restarts, you may want to use larger intervals in those
+        cases.
+
+const jpeg_scan_info *scan_info
+int num_scans
+        By default, scan_info is NULL; this causes the compressor to write a
+        single-scan sequential JPEG file.  If not NULL, scan_info points to
+        an array of scan definition records of length num_scans.  The
+        compressor will then write a JPEG file having one scan for each scan
+        definition record.  This is used to generate noninterleaved or
+        progressive JPEG files.  The library checks that the scan array
+        defines a valid JPEG scan sequence.  (jpeg_simple_progression creates
+        a suitable scan definition array for progressive JPEG.)  This is
+        discussed further under "Progressive JPEG support".
+
+int smoothing_factor
+        If non-zero, the input image is smoothed; the value should be 1 for
+        minimal smoothing to 100 for maximum smoothing.  Consult jcsample.c
+        for details of the smoothing algorithm.  The default is zero.
+
+boolean write_JFIF_header
+        If TRUE, a JFIF APP0 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if a JFIF-legal JPEG color space
+        (ie, YCbCr or grayscale) is selected, otherwise FALSE.
+
+UINT8 JFIF_major_version
+UINT8 JFIF_minor_version
+        The version number to be written into the JFIF marker.
+        jpeg_set_defaults() initializes the version to 1.01 (major=minor=1).
+        You should set it to 1.02 (major=1, minor=2) if you plan to write
+        any JFIF 1.02 extension markers.
+
+UINT8 density_unit
+UINT16 X_density
+UINT16 Y_density
+        The resolution information to be written into the JFIF marker;
+        not used otherwise.  density_unit may be 0 for unknown,
+        1 for dots/inch, or 2 for dots/cm.  The default values are 0,1,1
+        indicating square pixels of unknown size.
+
+boolean write_Adobe_marker
+        If TRUE, an Adobe APP14 marker is emitted.  jpeg_set_defaults() and
+        jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
+        or YCCK is selected, otherwise FALSE.  It is generally a bad idea
+        to set both write_JFIF_header and write_Adobe_marker.  In fact,
+        you probably shouldn't change the default settings at all --- the
+        default behavior ensures that the JPEG file's color space can be
+        recognized by the decoder.
+
+JQUANT_TBL *quant_tbl_ptrs[NUM_QUANT_TBLS]
+        Pointers to coefficient quantization tables, one per table slot,
+        or NULL if no table is defined for a slot.  Usually these should
+        be set via one of the above helper routines; jpeg_add_quant_table()
+        is general enough to define any quantization table.  The other
+        routines will set up table slot 0 for luminance quality and table
+        slot 1 for chrominance.
+
+int q_scale_factor[NUM_QUANT_TBLS]
+        [libjpeg v7+ API/ABI emulation only]
+        Linear quantization scaling factors (0-100, default 100)
+        for use with jpeg_default_qtables().
+        See rdswitch.c and cjpeg.c for an example of usage.
+        Note that the q_scale_factor[] values use "linear" scales, so JPEG
+        quality levels chosen by the user must be converted to these scales
+        using jpeg_quality_scaling().  Here is an example that corresponds to
+        cjpeg -quality 90,70:
+
+                jpeg_set_defaults(cinfo);
+
+                /* Set luminance quality 90. */
+                cinfo->q_scale_factor[0] = jpeg_quality_scaling(90);
+                /* Set chrominance quality 70. */
+                cinfo->q_scale_factor[1] = jpeg_quality_scaling(70);
+
+                jpeg_default_qtables(cinfo, force_baseline);
+
+        CAUTION: Setting separate quality levels for chrominance and luminance
+        is mainly only useful if chrominance subsampling is disabled.  2x2
+        chrominance subsampling (AKA "4:2:0") is the default, but you can
+        explicitly disable subsampling as follows:
+
+                cinfo->comp_info[0].v_samp_factor = 1;
+                cinfo->comp_info[0].h_samp_factor = 1;
+
+JHUFF_TBL *dc_huff_tbl_ptrs[NUM_HUFF_TBLS]
+JHUFF_TBL *ac_huff_tbl_ptrs[NUM_HUFF_TBLS]
+        Pointers to Huffman coding tables, one per table slot, or NULL if
+        no table is defined for a slot.  Slots 0 and 1 are filled with the
+        JPEG sample tables by jpeg_set_defaults().  If you need to allocate
+        more table structures, jpeg_alloc_huff_table() may be used.
+        Note that optimal Huffman tables can be computed for an image
+        by setting optimize_coding, as discussed above; there's seldom
+        any need to mess with providing your own Huffman tables.
+
+
+[libjpeg v7+ API/ABI emulation only]
+The actual dimensions of the JPEG image that will be written to the file are
+given by the following fields.  These are computed from the input image
+dimensions and the compression parameters by jpeg_start_compress().  You can
+also call jpeg_calc_jpeg_dimensions() to obtain the values that will result
+from the current parameter settings.  This can be useful if you are trying
+to pick a scaling ratio that will get close to a desired target size.
+
+JDIMENSION jpeg_width           Actual dimensions of output image.
+JDIMENSION jpeg_height
+
+
+Per-component parameters are stored in the struct cinfo.comp_info[i] for
+component number i.  Note that components here refer to components of the
+JPEG color space, *not* the source image color space.  A suitably large
+comp_info[] array is allocated by jpeg_set_defaults(); if you choose not
+to use that routine, it's up to you to allocate the array.
+
+int component_id
+        The one-byte identifier code to be recorded in the JPEG file for
+        this component.  For the standard color spaces, we recommend you
+        leave the default values alone.
+
+int h_samp_factor
+int v_samp_factor
+        Horizontal and vertical sampling factors for the component; must
+        be 1..4 according to the JPEG standard.  Note that larger sampling
+        factors indicate a higher-resolution component; many people find
+        this behavior quite unintuitive.  The default values are 2,2 for
+        luminance components and 1,1 for chrominance components, except
+        for grayscale where 1,1 is used.
+
+int quant_tbl_no
+        Quantization table number for component.  The default value is
+        0 for luminance components and 1 for chrominance components.
+
+int dc_tbl_no
+int ac_tbl_no
+        DC and AC entropy coding table numbers.  The default values are
+        0 for luminance components and 1 for chrominance components.
+
+int component_index
+        Must equal the component's index in comp_info[].  (Beginning in
+        release v6, the compressor library will fill this in automatically;
+        you don't have to.)
+
+
+Decompression parameter selection
+---------------------------------
+
+Decompression parameter selection is somewhat simpler than compression
+parameter selection, since all of the JPEG internal parameters are
+recorded in the source file and need not be supplied by the application.
+(Unless you are working with abbreviated files, in which case see
+"Abbreviated datastreams", below.)  Decompression parameters control
+the postprocessing done on the image to deliver it in a format suitable
+for the application's use.  Many of the parameters control speed/quality
+tradeoffs, in which faster decompression may be obtained at the price of
+a poorer-quality image.  The defaults select the highest quality (slowest)
+processing.
+
+The following fields in the JPEG object are set by jpeg_read_header() and
+may be useful to the application in choosing decompression parameters:
+
+JDIMENSION image_width                  Width and height of image
+JDIMENSION image_height
+int num_components                      Number of color components
+J_COLOR_SPACE jpeg_color_space          Colorspace of image
+boolean saw_JFIF_marker                 TRUE if a JFIF APP0 marker was seen
+  UINT8 JFIF_major_version              Version information from JFIF marker
+  UINT8 JFIF_minor_version
+  UINT8 density_unit                    Resolution data from JFIF marker
+  UINT16 X_density
+  UINT16 Y_density
+boolean saw_Adobe_marker                TRUE if an Adobe APP14 marker was seen
+  UINT8 Adobe_transform                 Color transform code from Adobe marker
+
+The JPEG color space, unfortunately, is something of a guess since the JPEG
+standard proper does not provide a way to record it.  In practice most files
+adhere to the JFIF or Adobe conventions, and the decoder will recognize these
+correctly.  See "Special color spaces", below, for more info.
+
+
+The decompression parameters that determine the basic properties of the
+returned image are:
+
+J_COLOR_SPACE out_color_space
+        Output color space.  jpeg_read_header() sets an appropriate default
+        based on jpeg_color_space; typically it will be RGB or grayscale.
+        The application can change this field to request output in a different
+        colorspace.  For example, set it to JCS_GRAYSCALE to get grayscale
+        output from a color file.  (This is useful for previewing: grayscale
+        output is faster than full color since the color components need not
+        be processed.)  Note that not all possible color space transforms are
+        currently implemented; you may need to extend jdcolor.c if you want an
+        unusual conversion.
+
+unsigned int scale_num, scale_denom
+        Scale the image by the fraction scale_num/scale_denom.  Default is
+        1/1, or no scaling.  Currently, the only supported scaling ratios
+        are M/8 with all M from 1 to 16, or any reduced fraction thereof (such
+        as 1/2, 3/4, etc.)  (The library design allows for arbitrary
+        scaling ratios but this is not likely to be implemented any time soon.)
+        Smaller scaling ratios permit significantly faster decoding since
+        fewer pixels need be processed and a simpler IDCT method can be used.
+
+boolean quantize_colors
+        If set TRUE, colormapped output will be delivered.  Default is FALSE,
+        meaning that full-color output will be delivered.
+
+The next three parameters are relevant only if quantize_colors is TRUE.
+
+int desired_number_of_colors
+        Maximum number of colors to use in generating a library-supplied color
+        map (the actual number of colors is returned in a different field).
+        Default 256.  Ignored when the application supplies its own color map.
+
+boolean two_pass_quantize
+        If TRUE, an extra pass over the image is made to select a custom color
+        map for the image.  This usually looks a lot better than the one-size-
+        fits-all colormap that is used otherwise.  Default is TRUE.  Ignored
+        when the application supplies its own color map.
+
+J_DITHER_MODE dither_mode
+        Selects color dithering method.  Supported values are:
+                JDITHER_NONE    no dithering: fast, very low quality
+                JDITHER_ORDERED ordered dither: moderate speed and quality
+                JDITHER_FS      Floyd-Steinberg dither: slow, high quality
+        Default is JDITHER_FS.  (At present, ordered dither is implemented
+        only in the single-pass, standard-colormap case.  If you ask for
+        ordered dither when two_pass_quantize is TRUE or when you supply
+        an external color map, you'll get F-S dithering.)
+
+When quantize_colors is TRUE, the target color map is described by the next
+two fields.  colormap is set to NULL by jpeg_read_header().  The application
+can supply a color map by setting colormap non-NULL and setting
+actual_number_of_colors to the map size.  Otherwise, jpeg_start_decompress()
+selects a suitable color map and sets these two fields itself.
+[Implementation restriction: at present, an externally supplied colormap is
+only accepted for 3-component output color spaces.]
+
+JSAMPARRAY colormap
+        The color map, represented as a 2-D pixel array of out_color_components
+        rows and actual_number_of_colors columns.  Ignored if not quantizing.
+        CAUTION: if the JPEG library creates its own colormap, the storage
+        pointed to by this field is released by jpeg_finish_decompress().
+        Copy the colormap somewhere else first, if you want to save it.
+
+int actual_number_of_colors
+        The number of colors in the color map.
+
+Additional decompression parameters that the application may set include:
+
+J_DCT_METHOD dct_method
+        Selects the algorithm used for the DCT step.  Choices are:
+                JDCT_ISLOW: slow but accurate integer algorithm
+                JDCT_IFAST: faster, less accurate integer method
+                JDCT_FLOAT: floating-point method
+                JDCT_DEFAULT: default method (normally JDCT_ISLOW)
+                JDCT_FASTEST: fastest method (normally JDCT_IFAST)
+        In libjpeg-turbo, JDCT_IFAST is generally about 5-15% faster than
+        JDCT_ISLOW when using the x86/x86-64 SIMD extensions (results may vary
+        with other SIMD implementations, or when using libjpeg-turbo without
+        SIMD extensions.)  If the JPEG image was compressed using a quality
+        level of 85 or below, then there should be little or no perceptible
+        difference between the two algorithms.  When decompressing images that
+        were compressed using quality levels above 85, however, the difference
+        between JDCT_IFAST and JDCT_ISLOW becomes more pronounced.  With images
+        compressed using quality=97, for instance, JDCT_IFAST incurs generally
+        about a 4-6 dB loss (in PSNR) relative to JDCT_ISLOW, but this can be
+        larger for some images.  If you can avoid it, do not use JDCT_IFAST
+        when decompressing images that were compressed using quality levels
+        above 97.  The algorithm often degenerates for such images and can
+        actually produce a more lossy output image than if the JPEG image had
+        been compressed using lower quality levels.  JDCT_FLOAT is mainly a
+        legacy feature.  It does not produce significantly more accurate
+        results than the ISLOW method, and it is much slower.  The FLOAT method
+        may also give different results on different machines due to varying
+        roundoff behavior, whereas the integer methods should give the same
+        results on all machines.
+
+boolean do_fancy_upsampling
+        If TRUE, do careful upsampling of chroma components.  If FALSE,
+        a faster but sloppier method is used.  Default is TRUE.  The visual
+        impact of the sloppier method is often very small.
+
+boolean do_block_smoothing
+        If TRUE, interblock smoothing is applied in early stages of decoding
+        progressive JPEG files; if FALSE, not.  Default is TRUE.  Early
+        progression stages look "fuzzy" with smoothing, "blocky" without.
+        In any case, block smoothing ceases to be applied after the first few
+        AC coefficients are known to full accuracy, so it is relevant only
+        when using buffered-image mode for progressive images.
+
+boolean enable_1pass_quant
+boolean enable_external_quant
+boolean enable_2pass_quant
+        These are significant only in buffered-image mode, which is
+        described in its own section below.
+
+
+The output image dimensions are given by the following fields.  These are
+computed from the source image dimensions and the decompression parameters
+by jpeg_start_decompress().  You can also call jpeg_calc_output_dimensions()
+to obtain the values that will result from the current parameter settings.
+This can be useful if you are trying to pick a scaling ratio that will get
+close to a desired target size.  It's also important if you are using the
+JPEG library's memory manager to allocate output buffer space, because you
+are supposed to request such buffers *before* jpeg_start_decompress().
+
+JDIMENSION output_width         Actual dimensions of output image.
+JDIMENSION output_height
+int out_color_components        Number of color components in out_color_space.
+int output_components           Number of color components returned.
+int rec_outbuf_height           Recommended height of scanline buffer.
+
+When quantizing colors, output_components is 1, indicating a single color map
+index per pixel.  Otherwise it equals out_color_components.  The output arrays
+are required to be output_width * output_components JSAMPLEs wide.
+
+rec_outbuf_height is the recommended minimum height (in scanlines) of the
+buffer passed to jpeg_read_scanlines().  If the buffer is smaller, the
+library will still work, but time will be wasted due to unnecessary data
+copying.  In high-quality modes, rec_outbuf_height is always 1, but some
+faster, lower-quality modes set it to larger values (typically 2 to 4).
+If you are going to ask for a high-speed processing mode, you may as well
+go to the trouble of honoring rec_outbuf_height so as to avoid data copying.
+(An output buffer larger than rec_outbuf_height lines is OK, but won't
+provide any material speed improvement over that height.)
+
+
+Special color spaces
+--------------------
+
+The JPEG standard itself is "color blind" and doesn't specify any particular
+color space.  It is customary to convert color data to a luminance/chrominance
+color space before compressing, since this permits greater compression.  The
+existing de-facto JPEG file format standards specify YCbCr or grayscale data
+(JFIF), or grayscale, RGB, YCbCr, CMYK, or YCCK (Adobe).  For special
+applications such as multispectral images, other color spaces can be used,
+but it must be understood that such files will be unportable.
+
+The JPEG library can handle the most common colorspace conversions (namely
+RGB <=> YCbCr and CMYK <=> YCCK).  It can also deal with data of an unknown
+color space, passing it through without conversion.  If you deal extensively
+with an unusual color space, you can easily extend the library to understand
+additional color spaces and perform appropriate conversions.
+
+For compression, the source data's color space is specified by field
+in_color_space.  This is transformed to the JPEG file's color space given
+by jpeg_color_space.  jpeg_set_defaults() chooses a reasonable JPEG color
+space depending on in_color_space, but you can override this by calling
+jpeg_set_colorspace().  Of course you must select a supported transformation.
+jccolor.c currently supports the following transformations:
+        RGB => YCbCr
+        RGB => GRAYSCALE
+        YCbCr => GRAYSCALE
+        CMYK => YCCK
+plus the null transforms: GRAYSCALE => GRAYSCALE, RGB => RGB,
+YCbCr => YCbCr, CMYK => CMYK, YCCK => YCCK, and UNKNOWN => UNKNOWN.
+
+The de-facto file format standards (JFIF and Adobe) specify APPn markers that
+indicate the color space of the JPEG file.  It is important to ensure that
+these are written correctly, or omitted if the JPEG file's color space is not
+one of the ones supported by the de-facto standards.  jpeg_set_colorspace()
+will set the compression parameters to include or omit the APPn markers
+properly, so long as it is told the truth about the JPEG color space.
+For example, if you are writing some random 3-component color space without
+conversion, don't try to fake out the library by setting in_color_space and
+jpeg_color_space to JCS_YCbCr; use JCS_UNKNOWN.  You may want to write an
+APPn marker of your own devising to identify the colorspace --- see "Special
+markers", below.
+
+When told that the color space is UNKNOWN, the library will default to using
+luminance-quality compression parameters for all color components.  You may
+well want to change these parameters.  See the source code for
+jpeg_set_colorspace(), in jcparam.c, for details.
+
+For decompression, the JPEG file's color space is given in jpeg_color_space,
+and this is transformed to the output color space out_color_space.
+jpeg_read_header's setting of jpeg_color_space can be relied on if the file
+conforms to JFIF or Adobe conventions, but otherwise it is no better than a
+guess.  If you know the JPEG file's color space for certain, you can override
+jpeg_read_header's guess by setting jpeg_color_space.  jpeg_read_header also
+selects a default output color space based on (its guess of) jpeg_color_space;
+set out_color_space to override this.  Again, you must select a supported
+transformation.  jdcolor.c currently supports
+        YCbCr => RGB
+        YCbCr => GRAYSCALE
+        RGB => GRAYSCALE
+        GRAYSCALE => RGB
+        YCCK => CMYK
+as well as the null transforms.  (Since GRAYSCALE=>RGB is provided, an
+application can force grayscale JPEGs to look like color JPEGs if it only
+wants to handle one case.)
+
+The two-pass color quantizer, jquant2.c, is specialized to handle RGB data
+(it weights distances appropriately for RGB colors).  You'll need to modify
+the code if you want to use it for non-RGB output color spaces.  Note that
+jquant2.c is used to map to an application-supplied colormap as well as for
+the normal two-pass colormap selection process.
+
+CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
+files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
+This is arguably a bug in Photoshop, but if you need to work with Photoshop
+CMYK files, you will have to deal with it in your application.  We cannot
+"fix" this in the library by inverting the data during the CMYK<=>YCCK
+transform, because that would break other applications, notably Ghostscript.
+Photoshop versions prior to 3.0 write EPS files containing JPEG-encoded CMYK
+data in the same inverted-YCCK representation used in bare JPEG files, but
+the surrounding PostScript code performs an inversion using the PS image
+operator.  I am told that Photoshop 3.0 will write uninverted YCCK in
+EPS/JPEG files, and will omit the PS-level inversion.  (But the data
+polarity used in bare JPEG files will not change in 3.0.)  In either case,
+the JPEG library must not invert the data itself, or else Ghostscript would
+read these EPS files incorrectly.
+
+
+Error handling
+--------------
+
+When the default error handler is used, any error detected inside the JPEG
+routines will cause a message to be printed on stderr, followed by exit().
+You can supply your own error handling routines to override this behavior
+and to control the treatment of nonfatal warnings and trace/debug messages.
+The file example.c illustrates the most common case, which is to have the
+application regain control after an error rather than exiting.
+
+The JPEG library never writes any message directly; it always goes through
+the error handling routines.  Three classes of messages are recognized:
+  * Fatal errors: the library cannot continue.
+  * Warnings: the library can continue, but the data is corrupt, and a
+    damaged output image is likely to result.
+  * Trace/informational messages.  These come with a trace level indicating
+    the importance of the message; you can control the verbosity of the
+    program by adjusting the maximum trace level that will be displayed.
+
+You may, if you wish, simply replace the entire JPEG error handling module
+(jerror.c) with your own code.  However, you can avoid code duplication by
+only replacing some of the routines depending on the behavior you need.
+This is accomplished by calling jpeg_std_error() as usual, but then overriding
+some of the method pointers in the jpeg_error_mgr struct, as illustrated by
+example.c.
+
+All of the error handling routines will receive a pointer to the JPEG object
+(a j_common_ptr which points to either a jpeg_compress_struct or a
+jpeg_decompress_struct; if you need to tell which, test the is_decompressor
+field).  This struct includes a pointer to the error manager struct in its
+"err" field.  Frequently, custom error handler routines will need to access
+additional data which is not known to the JPEG library or the standard error
+handler.  The most convenient way to do this is to embed either the JPEG
+object or the jpeg_error_mgr struct in a larger structure that contains
+additional fields; then casting the passed pointer provides access to the
+additional fields.  Again, see example.c for one way to do it.  (Beginning
+with IJG version 6b, there is also a void pointer "client_data" in each
+JPEG object, which the application can also use to find related data.
+The library does not touch client_data at all.)
+
+The individual methods that you might wish to override are:
+
+error_exit (j_common_ptr cinfo)
+        Receives control for a fatal error.  Information sufficient to
+        generate the error message has been stored in cinfo->err; call
+        output_message to display it.  Control must NOT return to the caller;
+        generally this routine will exit() or longjmp() somewhere.
+        Typically you would override this routine to get rid of the exit()
+        default behavior.  Note that if you continue processing, you should
+        clean up the JPEG object with jpeg_abort() or jpeg_destroy().
+
+output_message (j_common_ptr cinfo)
+        Actual output of any JPEG message.  Override this to send messages
+        somewhere other than stderr.  Note that this method does not know
+        how to generate a message, only where to send it.
+
+format_message (j_common_ptr cinfo, char *buffer)
+        Constructs a readable error message string based on the error info
+        stored in cinfo->err.  This method is called by output_message.  Few
+        applications should need to override this method.  One possible
+        reason for doing so is to implement dynamic switching of error message
+        language.
+
+emit_message (j_common_ptr cinfo, int msg_level)
+        Decide whether or not to emit a warning or trace message; if so,
+        calls output_message.  The main reason for overriding this method
+        would be to abort on warnings.  msg_level is -1 for warnings,
+        0 and up for trace messages.
+
+Only error_exit() and emit_message() are called from the rest of the JPEG
+library; the other two are internal to the error handler.
+
+The actual message texts are stored in an array of strings which is pointed to
+by the field err->jpeg_message_table.  The messages are numbered from 0 to
+err->last_jpeg_message, and it is these code numbers that are used in the
+JPEG library code.  You could replace the message texts (for instance, with
+messages in French or German) by changing the message table pointer.  See
+jerror.h for the default texts.  CAUTION: this table will almost certainly
+change or grow from one library version to the next.
+
+It may be useful for an application to add its own message texts that are
+handled by the same mechanism.  The error handler supports a second "add-on"
+message table for this purpose.  To define an addon table, set the pointer
+err->addon_message_table and the message numbers err->first_addon_message and
+err->last_addon_message.  If you number the addon messages beginning at 1000
+or so, you won't have to worry about conflicts with the library's built-in
+messages.  See the sample applications cjpeg/djpeg for an example of using
+addon messages (the addon messages are defined in cderror.h).
+
+Actual invocation of the error handler is done via macros defined in jerror.h:
+        ERREXITn(...)   for fatal errors
+        WARNMSn(...)    for corrupt-data warnings
+        TRACEMSn(...)   for trace and informational messages.
+These macros store the message code and any additional parameters into the
+error handler struct, then invoke the error_exit() or emit_message() method.
+The variants of each macro are for varying numbers of additional parameters.
+The additional parameters are inserted into the generated message using
+standard printf() format codes.
+
+See jerror.h and jerror.c for further details.
+
+
+Compressed data handling (source and destination managers)
+----------------------------------------------------------
+
+The JPEG compression library sends its compressed data to a "destination
+manager" module.  The default destination manager just writes the data to a
+memory buffer or to a stdio stream, but you can provide your own manager to
+do something else.  Similarly, the decompression library calls a "source
+manager" to obtain the compressed data; you can provide your own source
+manager if you want the data to come from somewhere other than a memory
+buffer or a stdio stream.
+
+In both cases, compressed data is processed a bufferload at a time: the
+destination or source manager provides a work buffer, and the library invokes
+the manager only when the buffer is filled or emptied.  (You could define a
+one-character buffer to force the manager to be invoked for each byte, but
+that would be rather inefficient.)  The buffer's size and location are
+controlled by the manager, not by the library.  For example, the memory
+source manager just makes the buffer pointer and length point to the original
+data in memory.  In this case the buffer-reload procedure will be invoked
+only if the decompressor ran off the end of the datastream, which would
+indicate an erroneous datastream.
+
+The work buffer is defined as an array of datatype JOCTET, which is generally
+"char" or "unsigned char".  On a machine where char is not exactly 8 bits
+wide, you must define JOCTET as a wider data type and then modify the data
+source and destination modules to transcribe the work arrays into 8-bit units
+on external storage.
+
+A data destination manager struct contains a pointer and count defining the
+next byte to write in the work buffer and the remaining free space:
+
+        JOCTET *next_output_byte;   /* => next byte to write in buffer */
+        size_t free_in_buffer;      /* # of byte spaces remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is filled.  The manager's empty_output_buffer method must reset the pointer
+and count.  The manager is expected to remember the buffer's starting address
+and total size in private fields not visible to the library.
+
+A data destination manager provides three methods:
+
+init_destination (j_compress_ptr cinfo)
+        Initialize destination.  This is called by jpeg_start_compress()
+        before any data is actually written.  It must initialize
+        next_output_byte and free_in_buffer.  free_in_buffer must be
+        initialized to a positive value.
+
+empty_output_buffer (j_compress_ptr cinfo)
+        This is called whenever the buffer has filled (free_in_buffer
+        reaches zero).  In typical applications, it should write out the
+        *entire* buffer (use the saved start address and buffer length;
+        ignore the current state of next_output_byte and free_in_buffer).
+        Then reset the pointer & count to the start of the buffer, and
+        return TRUE indicating that the buffer has been dumped.
+        free_in_buffer must be set to a positive value when TRUE is
+        returned.  A FALSE return should only be used when I/O suspension is
+        desired (this operating mode is discussed in the next section).
+
+term_destination (j_compress_ptr cinfo)
+        Terminate destination --- called by jpeg_finish_compress() after all
+        data has been written.  In most applications, this must flush any
+        data remaining in the buffer.  Use either next_output_byte or
+        free_in_buffer to determine how much data is in the buffer.
+
+term_destination() is NOT called by jpeg_abort() or jpeg_destroy().  If you
+want the destination manager to be cleaned up during an abort, you must do it
+yourself.
+
+You will also need code to create a jpeg_destination_mgr struct, fill in its
+method pointers, and insert a pointer to the struct into the "dest" field of
+the JPEG compression object.  This can be done in-line in your setup code if
+you like, but it's probably cleaner to provide a separate routine similar to
+the jpeg_stdio_dest() or jpeg_mem_dest() routines of the supplied destination
+managers.
+
+Decompression source managers follow a parallel design, but with some
+additional frammishes.  The source manager struct contains a pointer and count
+defining the next byte to read from the work buffer and the number of bytes
+remaining:
+
+        const JOCTET *next_input_byte;  /* => next byte to read from buffer */
+        size_t bytes_in_buffer;         /* # of bytes remaining in buffer */
+
+The library increments the pointer and decrements the count until the buffer
+is emptied.  The manager's fill_input_buffer method must reset the pointer and
+count.  In most applications, the manager must remember the buffer's starting
+address and total size in private fields not visible to the library.
+
+A data source manager provides five methods:
+
+init_source (j_decompress_ptr cinfo)
+        Initialize source.  This is called by jpeg_read_header() before any
+        data is actually read.  Unlike init_destination(), it may leave
+        bytes_in_buffer set to 0 (in which case a fill_input_buffer() call
+        will occur immediately).
+
+fill_input_buffer (j_decompress_ptr cinfo)
+        This is called whenever bytes_in_buffer has reached zero and more
+        data is wanted.  In typical applications, it should read fresh data
+        into the buffer (ignoring the current state of next_input_byte and
+        bytes_in_buffer), reset the pointer & count to the start of the
+        buffer, and return TRUE indicating that the buffer has been reloaded.
+        It is not necessary to fill the buffer entirely, only to obtain at
+        least one more byte.  bytes_in_buffer MUST be set to a positive value
+        if TRUE is returned.  A FALSE return should only be used when I/O
+        suspension is desired (this mode is discussed in the next section).
+
+skip_input_data (j_decompress_ptr cinfo, long num_bytes)
+        Skip num_bytes worth of data.  The buffer pointer and count should
+        be advanced over num_bytes input bytes, refilling the buffer as
+        needed.  This is used to skip over a potentially large amount of
+        uninteresting data (such as an APPn marker).  In some applications
+        it may be possible to optimize away the reading of the skipped data,
+        but it's not clear that being smart is worth much trouble; large
+        skips are uncommon.  bytes_in_buffer may be zero on return.
+        A zero or negative skip count should be treated as a no-op.
+
+resync_to_restart (j_decompress_ptr cinfo, int desired)
+        This routine is called only when the decompressor has failed to find
+        a restart (RSTn) marker where one is expected.  Its mission is to
+        find a suitable point for resuming decompression.  For most
+        applications, we recommend that you just use the default resync
+        procedure, jpeg_resync_to_restart().  However, if you are able to back
+        up in the input data stream, or if you have a-priori knowledge about
+        the likely location of restart markers, you may be able to do better.
+        Read the read_restart_marker() and jpeg_resync_to_restart() routines
+        in jdmarker.c if you think you'd like to implement your own resync
+        procedure.
+
+term_source (j_decompress_ptr cinfo)
+        Terminate source --- called by jpeg_finish_decompress() after all
+        data has been read.  Often a no-op.
+
+For both fill_input_buffer() and skip_input_data(), there is no such thing
+as an EOF return.  If the end of the file has been reached, the routine has
+a choice of exiting via ERREXIT() or inserting fake data into the buffer.
+In most cases, generating a warning message and inserting a fake EOI marker
+is the best course of action --- this will allow the decompressor to output
+however much of the image is there.  In pathological cases, the decompressor
+may swallow the EOI and again demand data ... just keep feeding it fake EOIs.
+jdatasrc.c illustrates the recommended error recovery behavior.
+
+term_source() is NOT called by jpeg_abort() or jpeg_destroy().  If you want
+the source manager to be cleaned up during an abort, you must do it yourself.
+
+You will also need code to create a jpeg_source_mgr struct, fill in its method
+pointers, and insert a pointer to the struct into the "src" field of the JPEG
+decompression object.  This can be done in-line in your setup code if you
+like, but it's probably cleaner to provide a separate routine similar to the
+jpeg_stdio_src() or jpeg_mem_src() routines of the supplied source managers.
+
+For more information, consult the memory and stdio source and destination
+managers in jdatasrc.c and jdatadst.c.
+
+
+I/O suspension
+--------------
+
+Some applications need to use the JPEG library as an incremental memory-to-
+memory filter: when the compressed data buffer is filled or emptied, they want
+control to return to the outer loop, rather than expecting that the buffer can
+be emptied or reloaded within the data source/destination manager subroutine.
+The library supports this need by providing an "I/O suspension" mode, which we
+describe in this section.
+
+The I/O suspension mode is not a panacea: nothing is guaranteed about the
+maximum amount of time spent in any one call to the library, so it will not
+eliminate response-time problems in single-threaded applications.  If you
+need guaranteed response time, we suggest you "bite the bullet" and implement
+a real multi-tasking capability.
+
+To use I/O suspension, cooperation is needed between the calling application
+and the data source or destination manager; you will always need a custom
+source/destination manager.  (Please read the previous section if you haven't
+already.)  The basic idea is that the empty_output_buffer() or
+fill_input_buffer() routine is a no-op, merely returning FALSE to indicate
+that it has done nothing.  Upon seeing this, the JPEG library suspends
+operation and returns to its caller.  The surrounding application is
+responsible for emptying or refilling the work buffer before calling the
+JPEG library again.
+
+Compression suspension:
+
+For compression suspension, use an empty_output_buffer() routine that returns
+FALSE; typically it will not do anything else.  This will cause the
+compressor to return to the caller of jpeg_write_scanlines(), with the return
+value indicating that not all the supplied scanlines have been accepted.
+The application must make more room in the output buffer, adjust the output
+buffer pointer/count appropriately, and then call jpeg_write_scanlines()
+again, pointing to the first unconsumed scanline.
+
+When forced to suspend, the compressor will backtrack to a convenient stopping
+point (usually the start of the current MCU); it will regenerate some output
+data when restarted.  Therefore, although empty_output_buffer() is only
+called when the buffer is filled, you should NOT write out the entire buffer
+after a suspension.  Write only the data up to the current position of
+next_output_byte/free_in_buffer.  The data beyond that point will be
+regenerated after resumption.
+
+Because of the backtracking behavior, a good-size output buffer is essential
+for efficiency; you don't want the compressor to suspend often.  (In fact, an
+overly small buffer could lead to infinite looping, if a single MCU required
+more data than would fit in the buffer.)  We recommend a buffer of at least
+several Kbytes.  You may want to insert explicit code to ensure that you don't
+call jpeg_write_scanlines() unless there is a reasonable amount of space in
+the output buffer; in other words, flush the buffer before trying to compress
+more data.
+
+The compressor does not allow suspension while it is trying to write JPEG
+markers at the beginning and end of the file.  This means that:
+  * At the beginning of a compression operation, there must be enough free
+    space in the output buffer to hold the header markers (typically 600 or
+    so bytes).  The recommended buffer size is bigger than this anyway, so
+    this is not a problem as long as you start with an empty buffer.  However,
+    this restriction might catch you if you insert large special markers, such
+    as a JFIF thumbnail image, without flushing the buffer afterwards.
+  * When you call jpeg_finish_compress(), there must be enough space in the
+    output buffer to emit any buffered data and the final EOI marker.  In the
+    current implementation, half a dozen bytes should suffice for this, but
+    for safety's sake we recommend ensuring that at least 100 bytes are free
+    before calling jpeg_finish_compress().
+
+A more significant restriction is that jpeg_finish_compress() cannot suspend.
+This means you cannot use suspension with multi-pass operating modes, namely
+Huffman code optimization and multiple-scan output.  Those modes write the
+whole file during jpeg_finish_compress(), which will certainly result in
+buffer overrun.  (Note that this restriction applies only to compression,
+not decompression.  The decompressor supports input suspension in all of its
+operating modes.)
+
+Decompression suspension:
+
+For decompression suspension, use a fill_input_buffer() routine that simply
+returns FALSE (except perhaps during error recovery, as discussed below).
+This will cause the decompressor to return to its caller with an indication
+that suspension has occurred.  This can happen at four places:
+  * jpeg_read_header(): will return JPEG_SUSPENDED.
+  * jpeg_start_decompress(): will return FALSE, rather than its usual TRUE.
+  * jpeg_read_scanlines(): will return the number of scanlines already
+        completed (possibly 0).
+  * jpeg_finish_decompress(): will return FALSE, rather than its usual TRUE.
+The surrounding application must recognize these cases, load more data into
+the input buffer, and repeat the call.  In the case of jpeg_read_scanlines(),
+increment the passed pointers past any scanlines successfully read.
+
+Just as with compression, the decompressor will typically backtrack to a
+convenient restart point before suspending.  When fill_input_buffer() is
+called, next_input_byte/bytes_in_buffer point to the current restart point,
+which is where the decompressor will backtrack to if FALSE is returned.
+The data beyond that position must NOT be discarded if you suspend; it needs
+to be re-read upon resumption.  In most implementations, you'll need to shift
+this data down to the start of your work buffer and then load more data after
+it.  Again, this behavior means that a several-Kbyte work buffer is essential
+for decent performance; furthermore, you should load a reasonable amount of
+new data before resuming decompression.  (If you loaded, say, only one new
+byte each time around, you could waste a LOT of cycles.)
+
+The skip_input_data() source manager routine requires special care in a
+suspension scenario.  This routine is NOT granted the ability to suspend the
+decompressor; it can decrement bytes_in_buffer to zero, but no more.  If the
+requested skip distance exceeds the amount of data currently in the input
+buffer, then skip_input_data() must set bytes_in_buffer to zero and record the
+additional skip distance somewhere else.  The decompressor will immediately
+call fill_input_buffer(), which should return FALSE, which will cause a
+suspension return.  The surrounding application must then arrange to discard
+the recorded number of bytes before it resumes loading the input buffer.
+(Yes, this design is rather baroque, but it avoids complexity in the far more
+common case where a non-suspending source manager is used.)
+
+If the input data has been exhausted, we recommend that you emit a warning
+and insert dummy EOI markers just as a non-suspending data source manager
+would do.  This can be handled either in the surrounding application logic or
+within fill_input_buffer(); the latter is probably more efficient.  If
+fill_input_buffer() knows that no more data is available, it can set the
+pointer/count to point to a dummy EOI marker and then return TRUE just as
+though it had read more data in a non-suspending situation.
+
+The decompressor does not attempt to suspend within standard JPEG markers;
+instead it will backtrack to the start of the marker and reprocess the whole
+marker next time.  Hence the input buffer must be large enough to hold the
+longest standard marker in the file.  Standard JPEG markers should normally
+not exceed a few hundred bytes each (DHT tables are typically the longest).
+We recommend at least a 2K buffer for performance reasons, which is much
+larger than any correct marker is likely to be.  For robustness against
+damaged marker length counts, you may wish to insert a test in your
+application for the case that the input buffer is completely full and yet
+the decoder has suspended without consuming any data --- otherwise, if this
+situation did occur, it would lead to an endless loop.  (The library can't
+provide this test since it has no idea whether "the buffer is full", or
+even whether there is a fixed-size input buffer.)
+
+The input buffer would need to be 64K to allow for arbitrary COM or APPn
+markers, but these are handled specially: they are either saved into allocated
+memory, or skipped over by calling skip_input_data().  In the former case,
+suspension is handled correctly, and in the latter case, the problem of
+buffer overrun is placed on skip_input_data's shoulders, as explained above.
+Note that if you provide your own marker handling routine for large markers,
+you should consider how to deal with buffer overflow.
+
+Multiple-buffer management:
+
+In some applications it is desirable to store the compressed data in a linked
+list of buffer areas, so as to avoid data copying.  This can be handled by
+having empty_output_buffer() or fill_input_buffer() set the pointer and count
+to reference the next available buffer; FALSE is returned only if no more
+buffers are available.  Although seemingly straightforward, there is a
+pitfall in this approach: the backtrack that occurs when FALSE is returned
+could back up into an earlier buffer.  For example, when fill_input_buffer()
+is called, the current pointer & count indicate the backtrack restart point.
+Since fill_input_buffer() will set the pointer and count to refer to a new
+buffer, the restart position must be saved somewhere else.  Suppose a second
+call to fill_input_buffer() occurs in the same library call, and no
+additional input data is available, so fill_input_buffer must return FALSE.
+If the JPEG library has not moved the pointer/count forward in the current
+buffer, then *the correct restart point is the saved position in the prior
+buffer*.  Prior buffers may be discarded only after the library establishes
+a restart point within a later buffer.  Similar remarks apply for output into
+a chain of buffers.
+
+The library will never attempt to backtrack over a skip_input_data() call,
+so any skipped data can be permanently discarded.  You still have to deal
+with the case of skipping not-yet-received data, however.
+
+It's much simpler to use only a single buffer; when fill_input_buffer() is
+called, move any unconsumed data (beyond the current pointer/count) down to
+the beginning of this buffer and then load new data into the remaining buffer
+space.  This approach requires a little more data copying but is far easier
+to get right.
+
+
+Progressive JPEG support
+------------------------
+
+Progressive JPEG rearranges the stored data into a series of scans of
+increasing quality.  In situations where a JPEG file is transmitted across a
+slow communications link, a decoder can generate a low-quality image very
+quickly from the first scan, then gradually improve the displayed quality as
+more scans are received.  The final image after all scans are complete is
+identical to that of a regular (sequential) JPEG file of the same quality
+setting.  Progressive JPEG files are often slightly smaller than equivalent
+sequential JPEG files, but the possibility of incremental display is the main
+reason for using progressive JPEG.
+
+The IJG encoder library generates progressive JPEG files when given a
+suitable "scan script" defining how to divide the data into scans.
+Creation of progressive JPEG files is otherwise transparent to the encoder.
+Progressive JPEG files can also be read transparently by the decoder library.
+If the decoding application simply uses the library as defined above, it
+will receive a final decoded image without any indication that the file was
+progressive.  Of course, this approach does not allow incremental display.
+To perform incremental display, an application needs to use the decoder
+library's "buffered-image" mode, in which it receives a decoded image
+multiple times.
+
+Each displayed scan requires about as much work to decode as a full JPEG
+image of the same size, so the decoder must be fairly fast in relation to the
+data transmission rate in order to make incremental display useful.  However,
+it is possible to skip displaying the image and simply add the incoming bits
+to the decoder's coefficient buffer.  This is fast because only Huffman
+decoding need be done, not IDCT, upsampling, colorspace conversion, etc.
+The IJG decoder library allows the application to switch dynamically between
+displaying the image and simply absorbing the incoming bits.  A properly
+coded application can automatically adapt the number of display passes to
+suit the time available as the image is received.  Also, a final
+higher-quality display cycle can be performed from the buffered data after
+the end of the file is reached.
+
+Progressive compression:
+
+To create a progressive JPEG file (or a multiple-scan sequential JPEG file),
+set the scan_info cinfo field to point to an array of scan descriptors, and
+perform compression as usual.  Instead of constructing your own scan list,
+you can call the jpeg_simple_progression() helper routine to create a
+recommended progression sequence; this method should be used by all
+applications that don't want to get involved in the nitty-gritty of
+progressive scan sequence design.  (If you want to provide user control of
+scan sequences, you may wish to borrow the scan script reading code found
+in rdswitch.c, so that you can read scan script files just like cjpeg's.)
+When scan_info is not NULL, the compression library will store DCT'd data
+into a buffer array as jpeg_write_scanlines() is called, and will emit all
+the requested scans during jpeg_finish_compress().  This implies that
+multiple-scan output cannot be created with a suspending data destination
+manager, since jpeg_finish_compress() does not support suspension.  We
+should also note that the compressor currently forces Huffman optimization
+mode when creating a progressive JPEG file, because the default Huffman
+tables are unsuitable for progressive files.
+
+Progressive decompression:
+
+When buffered-image mode is not used, the decoder library will read all of
+a multi-scan file during jpeg_start_decompress(), so that it can provide a
+final decoded image.  (Here "multi-scan" means either progressive or
+multi-scan sequential.)  This makes multi-scan files transparent to the
+decoding application.  However, existing applications that used suspending
+input with version 5 of the IJG library will need to be modified to check
+for a suspension return from jpeg_start_decompress().
+
+To perform incremental display, an application must use the library's
+buffered-image mode.  This is described in the next section.
+
+
+Buffered-image mode
+-------------------
+
+In buffered-image mode, the library stores the partially decoded image in a
+coefficient buffer, from which it can be read out as many times as desired.
+This mode is typically used for incremental display of progressive JPEG files,
+but it can be used with any JPEG file.  Each scan of a progressive JPEG file
+adds more data (more detail) to the buffered image.  The application can
+display in lockstep with the source file (one display pass per input scan),
+or it can allow input processing to outrun display processing.  By making
+input and display processing run independently, it is possible for the
+application to adapt progressive display to a wide range of data transmission
+rates.
+
+The basic control flow for buffered-image decoding is
+
+        jpeg_create_decompress()
+        set data source
+        jpeg_read_header()
+        set overall decompression parameters
+        cinfo.buffered_image = TRUE;    /* select buffered-image mode */
+        jpeg_start_decompress()
+        for (each output pass) {
+            adjust output decompression parameters if required
+            jpeg_start_output()         /* start a new output pass */
+            for (all scanlines in image) {
+                jpeg_read_scanlines()
+                display scanlines
+            }
+            jpeg_finish_output()        /* terminate output pass */
+        }
+        jpeg_finish_decompress()
+        jpeg_destroy_decompress()
+
+This differs from ordinary unbuffered decoding in that there is an additional
+level of looping.  The application can choose how many output passes to make
+and how to display each pass.
+
+The simplest approach to displaying progressive images is to do one display
+pass for each scan appearing in the input file.  In this case the outer loop
+condition is typically
+        while (! jpeg_input_complete(&cinfo))
+and the start-output call should read
+        jpeg_start_output(&cinfo, cinfo.input_scan_number);
+The second parameter to jpeg_start_output() indicates which scan of the input
+file is to be displayed; the scans are numbered starting at 1 for this
+purpose.  (You can use a loop counter starting at 1 if you like, but using
+the library's input scan counter is easier.)  The library automatically reads
+data as necessary to complete each requested scan, and jpeg_finish_output()
+advances to the next scan or end-of-image marker (hence input_scan_number
+will be incremented by the time control arrives back at jpeg_start_output()).
+With this technique, data is read from the input file only as needed, and
+input and output processing run in lockstep.
+
+After reading the final scan and reaching the end of the input file, the
+buffered image remains available; it can be read additional times by
+repeating the jpeg_start_output()/jpeg_read_scanlines()/jpeg_finish_output()
+sequence.  For example, a useful technique is to use fast one-pass color
+quantization for display passes made while the image is arriving, followed by
+a final display pass using two-pass quantization for highest quality.  This
+is done by changing the library parameters before the final output pass.
+Changing parameters between passes is discussed in detail below.
+
+In general the last scan of a progressive file cannot be recognized as such
+until after it is read, so a post-input display pass is the best approach if
+you want special processing in the final pass.
+
+When done with the image, be sure to call jpeg_finish_decompress() to release
+the buffered image (or just use jpeg_destroy_decompress()).
+
+If input data arrives faster than it can be displayed, the application can
+cause the library to decode input data in advance of what's needed to produce
+output.  This is done by calling the routine jpeg_consume_input().
+The return value is one of the following:
+        JPEG_REACHED_SOS:    reached an SOS marker (the start of a new scan)
+        JPEG_REACHED_EOI:    reached the EOI marker (end of image)
+        JPEG_ROW_COMPLETED:  completed reading one MCU row of compressed data
+        JPEG_SCAN_COMPLETED: completed reading last MCU row of current scan
+        JPEG_SUSPENDED:      suspended before completing any of the above
+(JPEG_SUSPENDED can occur only if a suspending data source is used.)  This
+routine can be called at any time after initializing the JPEG object.  It
+reads some additional data and returns when one of the indicated significant
+events occurs.  (If called after the EOI marker is reached, it will
+immediately return JPEG_REACHED_EOI without attempting to read more data.)
+
+The library's output processing will automatically call jpeg_consume_input()
+whenever the output processing overtakes the input; thus, simple lockstep
+display requires no direct calls to jpeg_consume_input().  But by adding
+calls to jpeg_consume_input(), you can absorb data in advance of what is
+being displayed.  This has two benefits:
+  * You can limit buildup of unprocessed data in your input buffer.
+  * You can eliminate extra display passes by paying attention to the
+    state of the library's input processing.
+
+The first of these benefits only requires interspersing calls to
+jpeg_consume_input() with your display operations and any other processing
+you may be doing.  To avoid wasting cycles due to backtracking, it's best to
+call jpeg_consume_input() only after a hundred or so new bytes have arrived.
+This is discussed further under "I/O suspension", above.  (Note: the JPEG
+library currently is not thread-safe.  You must not call jpeg_consume_input()
+from one thread of control if a different library routine is working on the
+same JPEG object in another thread.)
+
+When input arrives fast enough that more than one new scan is available
+before you start a new output pass, you may as well skip the output pass
+corresponding to the completed scan.  This occurs for free if you pass
+cinfo.input_scan_number as the target scan number to jpeg_start_output().
+The input_scan_number field is simply the index of the scan currently being
+consumed by the input processor.  You can ensure that this is up-to-date by
+emptying the input buffer just before calling jpeg_start_output(): call
+jpeg_consume_input() repeatedly until it returns JPEG_SUSPENDED or
+JPEG_REACHED_EOI.
+
+The target scan number passed to jpeg_start_output() is saved in the
+cinfo.output_scan_number field.  The library's output processing calls
+jpeg_consume_input() whenever the current input scan number and row within
+that scan is less than or equal to the current output scan number and row.
+Thus, input processing can "get ahead" of the output processing but is not
+allowed to "fall behind".  You can achieve several different effects by
+manipulating this interlock rule.  For example, if you pass a target scan
+number greater than the current input scan number, the output processor will
+wait until that scan starts to arrive before producing any output.  (To avoid
+an infinite loop, the target scan number is automatically reset to the last
+scan number when the end of image is reached.  Thus, if you specify a large
+target scan number, the library will just absorb the entire input file and
+then perform an output pass.  This is effectively the same as what
+jpeg_start_decompress() does when you don't select buffered-image mode.)
+When you pass a target scan number equal to the current input scan number,
+the image is displayed no faster than the current input scan arrives.  The
+final possibility is to pass a target scan number less than the current input
+scan number; this disables the input/output interlock and causes the output
+processor to simply display whatever it finds in the image buffer, without
+waiting for input.  (However, the library will not accept a target scan
+number less than one, so you can't avoid waiting for the first scan.)
+
+When data is arriving faster than the output display processing can advance
+through the image, jpeg_consume_input() will store data into the buffered
+image beyond the point at which the output processing is reading data out
+again.  If the input arrives fast enough, it may "wrap around" the buffer to
+the point where the input is more than one whole scan ahead of the output.
+If the output processing simply proceeds through its display pass without
+paying attention to the input, the effect seen on-screen is that the lower
+part of the image is one or more scans better in quality than the upper part.
+Then, when the next output scan is started, you have a choice of what target
+scan number to use.  The recommended choice is to use the current input scan
+number at that time, which implies that you've skipped the output scans
+corresponding to the input scans that were completed while you processed the
+previous output scan.  In this way, the decoder automatically adapts its
+speed to the arriving data, by skipping output scans as necessary to keep up
+with the arriving data.
+
+When using this strategy, you'll want to be sure that you perform a final
+output pass after receiving all the data; otherwise your last display may not
+be full quality across the whole screen.  So the right outer loop logic is
+something like this:
+        do {
+            absorb any waiting input by calling jpeg_consume_input()
+            final_pass = jpeg_input_complete(&cinfo);
+            adjust output decompression parameters if required
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+        } while (! final_pass);
+rather than quitting as soon as jpeg_input_complete() returns TRUE.  This
+arrangement makes it simple to use higher-quality decoding parameters
+for the final pass.  But if you don't want to use special parameters for
+the final pass, the right loop logic is like this:
+        for (;;) {
+            absorb any waiting input by calling jpeg_consume_input()
+            jpeg_start_output(&cinfo, cinfo.input_scan_number);
+            ...
+            jpeg_finish_output()
+            if (jpeg_input_complete(&cinfo) &&
+                cinfo.input_scan_number == cinfo.output_scan_number)
+              break;
+        }
+In this case you don't need to know in advance whether an output pass is to
+be the last one, so it's not necessary to have reached EOF before starting
+the final output pass; rather, what you want to test is whether the output
+pass was performed in sync with the final input scan.  This form of the loop
+will avoid an extra output pass whenever the decoder is able (or nearly able)
+to keep up with the incoming data.
+
+When the data transmission speed is high, you might begin a display pass,
+then find that much or all of the file has arrived before you can complete
+the pass.  (You can detect this by noting the JPEG_REACHED_EOI return code
+from jpeg_consume_input(), or equivalently by testing jpeg_input_complete().)
+In this situation you may wish to abort the current display pass and start a
+new one using the newly arrived information.  To do so, just call
+jpeg_finish_output() and then start a new pass with jpeg_start_output().
+
+A variant strategy is to abort and restart display if more than one complete
+scan arrives during an output pass; this can be detected by noting
+JPEG_REACHED_SOS returns and/or examining cinfo.input_scan_number.  This
+idea should be employed with caution, however, since the display process
+might never get to the bottom of the image before being aborted, resulting
+in the lower part of the screen being several passes worse than the upper.
+In most cases it's probably best to abort an output pass only if the whole
+file has arrived and you want to begin the final output pass immediately.
+
+When receiving data across a communication link, we recommend always using
+the current input scan number for the output target scan number; if a
+higher-quality final pass is to be done, it should be started (aborting any
+incomplete output pass) as soon as the end of file is received.  However,
+many other strategies are possible.  For example, the application can examine
+the parameters of the current input scan and decide whether to display it or
+not.  If the scan contains only chroma data, one might choose not to use it
+as the target scan, expecting that the scan will be small and will arrive
+quickly.  To skip to the next scan, call jpeg_consume_input() until it
+returns JPEG_REACHED_SOS or JPEG_REACHED_EOI.  Or just use the next higher
+number as the target scan for jpeg_start_output(); but that method doesn't
+let you inspect the next scan's parameters before deciding to display it.
+
+
+In buffered-image mode, jpeg_start_decompress() never performs input and
+thus never suspends.  An application that uses input suspension with
+buffered-image mode must be prepared for suspension returns from these
+routines:
+* jpeg_start_output() performs input only if you request 2-pass quantization
+  and the target scan isn't fully read yet.  (This is discussed below.)
+* jpeg_read_scanlines(), as always, returns the number of scanlines that it
+  was able to produce before suspending.
+* jpeg_finish_output() will read any markers following the target scan,
+  up to the end of the file or the SOS marker that begins another scan.
+  (But it reads no input if jpeg_consume_input() has already reached the
+  end of the file or a SOS marker beyond the target output scan.)
+* jpeg_finish_decompress() will read until the end of file, and thus can
+  suspend if the end hasn't already been reached (as can be tested by
+  calling jpeg_input_complete()).
+jpeg_start_output(), jpeg_finish_output(), and jpeg_finish_decompress()
+all return TRUE if they completed their tasks, FALSE if they had to suspend.
+In the event of a FALSE return, the application must load more input data
+and repeat the call.  Applications that use non-suspending data sources need
+not check the return values of these three routines.
+
+
+It is possible to change decoding parameters between output passes in the
+buffered-image mode.  The decoder library currently supports only very
+limited changes of parameters.  ONLY THE FOLLOWING parameter changes are
+allowed after jpeg_start_decompress() is called:
+* dct_method can be changed before each call to jpeg_start_output().
+  For example, one could use a fast DCT method for early scans, changing
+  to a higher quality method for the final scan.
+* dither_mode can be changed before each call to jpeg_start_output();
+  of course this has no impact if not using color quantization.  Typically
+  one would use ordered dither for initial passes, then switch to
+  Floyd-Steinberg dither for the final pass.  Caution: changing dither mode
+  can cause more memory to be allocated by the library.  Although the amount
+  of memory involved is not large (a scanline or so), it may cause the
+  initial max_memory_to_use specification to be exceeded, which in the worst
+  case would result in an out-of-memory failure.
+* do_block_smoothing can be changed before each call to jpeg_start_output().
+  This setting is relevant only when decoding a progressive JPEG image.
+  During the first DC-only scan, block smoothing provides a very "fuzzy" look
+  instead of the very "blocky" look seen without it; which is better seems a
+  matter of personal taste.  But block smoothing is nearly always a win
+  during later stages, especially when decoding a successive-approximation
+  image: smoothing helps to hide the slight blockiness that otherwise shows
+  up on smooth gradients until the lowest coefficient bits are sent.
+* Color quantization mode can be changed under the rules described below.
+  You *cannot* change between full-color and quantized output (because that
+  would alter the required I/O buffer sizes), but you can change which
+  quantization method is used.
+
+When generating color-quantized output, changing quantization method is a
+very useful way of switching between high-speed and high-quality display.
+The library allows you to change among its three quantization methods:
+1. Single-pass quantization to a fixed color cube.
+   Selected by cinfo.two_pass_quantize = FALSE and cinfo.colormap = NULL.
+2. Single-pass quantization to an application-supplied colormap.
+   Selected by setting cinfo.colormap to point to the colormap (the value of
+   two_pass_quantize is ignored); also set cinfo.actual_number_of_colors.
+3. Two-pass quantization to a colormap chosen specifically for the image.
+   Selected by cinfo.two_pass_quantize = TRUE and cinfo.colormap = NULL.
+   (This is the default setting selected by jpeg_read_header, but it is
+   probably NOT what you want for the first pass of progressive display!)
+These methods offer successively better quality and lesser speed.  However,
+only the first method is available for quantizing in non-RGB color spaces.
+
+IMPORTANT: because the different quantizer methods have very different
+working-storage requirements, the library requires you to indicate which
+one(s) you intend to use before you call jpeg_start_decompress().  (If we did
+not require this, the max_memory_to_use setting would be a complete fiction.)
+You do this by setting one or more of these three cinfo fields to TRUE:
+        enable_1pass_quant              Fixed color cube colormap
+        enable_external_quant           Externally-supplied colormap
+        enable_2pass_quant              Two-pass custom colormap
+All three are initialized FALSE by jpeg_read_header().  But
+jpeg_start_decompress() automatically sets TRUE the one selected by the
+current two_pass_quantize and colormap settings, so you only need to set the
+enable flags for any other quantization methods you plan to change to later.
+
+After setting the enable flags correctly at jpeg_start_decompress() time, you
+can change to any enabled quantization method by setting two_pass_quantize
+and colormap properly just before calling jpeg_start_output().  The following
+special rules apply:
+1. You must explicitly set cinfo.colormap to NULL when switching to 1-pass
+   or 2-pass mode from a different mode, or when you want the 2-pass
+   quantizer to be re-run to generate a new colormap.
+2. To switch to an external colormap, or to change to a different external
+   colormap than was used on the prior pass, you must call
+   jpeg_new_colormap() after setting cinfo.colormap.
+NOTE: if you want to use the same colormap as was used in the prior pass,
+you should not do either of these things.  This will save some nontrivial
+switchover costs.
+(These requirements exist because cinfo.colormap will always be non-NULL
+after completing a prior output pass, since both the 1-pass and 2-pass
+quantizers set it to point to their output colormaps.  Thus you have to
+do one of these two things to notify the library that something has changed.
+Yup, it's a bit klugy, but it's necessary to do it this way for backwards
+compatibility.)
+
+Note that in buffered-image mode, the library generates any requested colormap
+during jpeg_start_output(), not during jpeg_start_decompress().
+
+When using two-pass quantization, jpeg_start_output() makes a pass over the
+buffered image to determine the optimum color map; it therefore may take a
+significant amount of time, whereas ordinarily it does little work.  The
+progress monitor hook is called during this pass, if defined.  It is also
+important to realize that if the specified target scan number is greater than
+or equal to the current input scan number, jpeg_start_output() will attempt
+to consume input as it makes this pass.  If you use a suspending data source,
+you need to check for a FALSE return from jpeg_start_output() under these
+conditions.  The combination of 2-pass quantization and a not-yet-fully-read
+target scan is the only case in which jpeg_start_output() will consume input.
+
+
+Application authors who support buffered-image mode may be tempted to use it
+for all JPEG images, even single-scan ones.  This will work, but it is
+inefficient: there is no need to create an image-sized coefficient buffer for
+single-scan images.  Requesting buffered-image mode for such an image wastes
+memory.  Worse, it can cost time on large images, since the buffered data has
+to be swapped out or written to a temporary file.  If you are concerned about
+maximum performance on baseline JPEG files, you should use buffered-image
+mode only when the incoming file actually has multiple scans.  This can be
+tested by calling jpeg_has_multiple_scans(), which will return a correct
+result at any time after jpeg_read_header() completes.
+
+It is also worth noting that when you use jpeg_consume_input() to let input
+processing get ahead of output processing, the resulting pattern of access to
+the coefficient buffer is quite nonsequential.  It's best to use the memory
+manager jmemnobs.c if you can (ie, if you have enough real or virtual main
+memory).  If not, at least make sure that max_memory_to_use is set as high as
+possible.  If the JPEG memory manager has to use a temporary file, you will
+probably see a lot of disk traffic and poor performance.  (This could be
+improved with additional work on the memory manager, but we haven't gotten
+around to it yet.)
+
+In some applications it may be convenient to use jpeg_consume_input() for all
+input processing, including reading the initial markers; that is, you may
+wish to call jpeg_consume_input() instead of jpeg_read_header() during
+startup.  This works, but note that you must check for JPEG_REACHED_SOS and
+JPEG_REACHED_EOI return codes as the equivalent of jpeg_read_header's codes.
+Once the first SOS marker has been reached, you must call
+jpeg_start_decompress() before jpeg_consume_input() will consume more input;
+it'll just keep returning JPEG_REACHED_SOS until you do.  If you read a
+tables-only file this way, jpeg_consume_input() will return JPEG_REACHED_EOI
+without ever returning JPEG_REACHED_SOS; be sure to check for this case.
+If this happens, the decompressor will not read any more input until you call
+jpeg_abort() to reset it.  It is OK to call jpeg_consume_input() even when not
+using buffered-image mode, but in that case it's basically a no-op after the
+initial markers have been read: it will just return JPEG_SUSPENDED.
+
+
+Abbreviated datastreams and multiple images
+-------------------------------------------
+
+A JPEG compression or decompression object can be reused to process multiple
+images.  This saves a small amount of time per image by eliminating the
+"create" and "destroy" operations, but that isn't the real purpose of the
+feature.  Rather, reuse of an object provides support for abbreviated JPEG
+datastreams.  Object reuse can also simplify processing a series of images in
+a single input or output file.  This section explains these features.
+
+A JPEG file normally contains several hundred bytes worth of quantization
+and Huffman tables.  In a situation where many images will be stored or
+transmitted with identical tables, this may represent an annoying overhead.
+The JPEG standard therefore permits tables to be omitted.  The standard
+defines three classes of JPEG datastreams:
+  * "Interchange" datastreams contain an image and all tables needed to decode
+     the image.  These are the usual kind of JPEG file.
+  * "Abbreviated image" datastreams contain an image, but are missing some or
+    all of the tables needed to decode that image.
+  * "Abbreviated table specification" (henceforth "tables-only") datastreams
+    contain only table specifications.
+To decode an abbreviated image, it is necessary to load the missing table(s)
+into the decoder beforehand.  This can be accomplished by reading a separate
+tables-only file.  A variant scheme uses a series of images in which the first
+image is an interchange (complete) datastream, while subsequent ones are
+abbreviated and rely on the tables loaded by the first image.  It is assumed
+that once the decoder has read a table, it will remember that table until a
+new definition for the same table number is encountered.
+
+It is the application designer's responsibility to figure out how to associate
+the correct tables with an abbreviated image.  While abbreviated datastreams
+can be useful in a closed environment, their use is strongly discouraged in
+any situation where data exchange with other applications might be needed.
+Caveat designer.
+
+The JPEG library provides support for reading and writing any combination of
+tables-only datastreams and abbreviated images.  In both compression and
+decompression objects, a quantization or Huffman table will be retained for
+the lifetime of the object, unless it is overwritten by a new table definition.
+
+
+To create abbreviated image datastreams, it is only necessary to tell the
+compressor not to emit some or all of the tables it is using.  Each
+quantization and Huffman table struct contains a boolean field "sent_table",
+which normally is initialized to FALSE.  For each table used by the image, the
+header-writing process emits the table and sets sent_table = TRUE unless it is
+already TRUE.  (In normal usage, this prevents outputting the same table
+definition multiple times, as would otherwise occur because the chroma
+components typically share tables.)  Thus, setting this field to TRUE before
+calling jpeg_start_compress() will prevent the table from being written at
+all.
+
+If you want to create a "pure" abbreviated image file containing no tables,
+just call "jpeg_suppress_tables(&cinfo, TRUE)" after constructing all the
+tables.  If you want to emit some but not all tables, you'll need to set the
+individual sent_table fields directly.
+
+To create an abbreviated image, you must also call jpeg_start_compress()
+with a second parameter of FALSE, not TRUE.  Otherwise jpeg_start_compress()
+will force all the sent_table fields to FALSE.  (This is a safety feature to
+prevent abbreviated images from being created accidentally.)
+
+To create a tables-only file, perform the same parameter setup that you
+normally would, but instead of calling jpeg_start_compress() and so on, call
+jpeg_write_tables(&cinfo).  This will write an abbreviated datastream
+containing only SOI, DQT and/or DHT markers, and EOI.  All the quantization
+and Huffman tables that are currently defined in the compression object will
+be emitted unless their sent_tables flag is already TRUE, and then all the
+sent_tables flags will be set TRUE.
+
+A sure-fire way to create matching tables-only and abbreviated image files
+is to proceed as follows:
+
+        create JPEG compression object
+        set JPEG parameters
+        set destination to tables-only file
+        jpeg_write_tables(&cinfo);
+        set destination to image file
+        jpeg_start_compress(&cinfo, FALSE);
+        write data...
+        jpeg_finish_compress(&cinfo);
+
+Since the JPEG parameters are not altered between writing the table file and
+the abbreviated image file, the same tables are sure to be used.  Of course,
+you can repeat the jpeg_start_compress() ... jpeg_finish_compress() sequence
+many times to produce many abbreviated image files matching the table file.
+
+You cannot suppress output of the computed Huffman tables when Huffman
+optimization is selected.  (If you could, there'd be no way to decode the
+image...)  Generally, you don't want to set optimize_coding = TRUE when
+you are trying to produce abbreviated files.
+
+In some cases you might want to compress an image using tables which are
+not stored in the application, but are defined in an interchange or
+tables-only file readable by the application.  This can be done by setting up
+a JPEG decompression object to read the specification file, then copying the
+tables into your compression object.  See jpeg_copy_critical_parameters()
+for an example of copying quantization tables.
+
+
+To read abbreviated image files, you simply need to load the proper tables
+into the decompression object before trying to read the abbreviated image.
+If the proper tables are stored in the application program, you can just
+allocate the table structs and fill in their contents directly.  For example,
+to load a fixed quantization table into table slot "n":
+
+    if (cinfo.quant_tbl_ptrs[n] == NULL)
+      cinfo.quant_tbl_ptrs[n] = jpeg_alloc_quant_table((j_common_ptr) &cinfo);
+    quant_ptr = cinfo.quant_tbl_ptrs[n];        /* quant_ptr is JQUANT_TBL* */
+    for (i = 0; i < 64; i++) {
+      /* Qtable[] is desired quantization table, in natural array order */
+      quant_ptr->quantval[i] = Qtable[i];
+    }
+
+Code to load a fixed Huffman table is typically (for AC table "n"):
+
+    if (cinfo.ac_huff_tbl_ptrs[n] == NULL)
+      cinfo.ac_huff_tbl_ptrs[n] = jpeg_alloc_huff_table((j_common_ptr) &cinfo);
+    huff_ptr = cinfo.ac_huff_tbl_ptrs[n];       /* huff_ptr is JHUFF_TBL* */
+    for (i = 1; i <= 16; i++) {
+      /* counts[i] is number of Huffman codes of length i bits, i=1..16 */
+      huff_ptr->bits[i] = counts[i];
+    }
+    for (i = 0; i < 256; i++) {
+      /* symbols[] is the list of Huffman symbols, in code-length order */
+      huff_ptr->huffval[i] = symbols[i];
+    }
+
+(Note that trying to set cinfo.quant_tbl_ptrs[n] to point directly at a
+constant JQUANT_TBL object is not safe.  If the incoming file happened to
+contain a quantization table definition, your master table would get
+overwritten!  Instead allocate a working table copy and copy the master table
+into it, as illustrated above.  Ditto for Huffman tables, of course.)
+
+You might want to read the tables from a tables-only file, rather than
+hard-wiring them into your application.  The jpeg_read_header() call is
+sufficient to read a tables-only file.  You must pass a second parameter of
+FALSE to indicate that you do not require an image to be present.  Thus, the
+typical scenario is
+
+        create JPEG decompression object
+        set source to tables-only file
+        jpeg_read_header(&cinfo, FALSE);
+        set source to abbreviated image file
+        jpeg_read_header(&cinfo, TRUE);
+        set decompression parameters
+        jpeg_start_decompress(&cinfo);
+        read data...
+        jpeg_finish_decompress(&cinfo);
+
+In some cases, you may want to read a file without knowing whether it contains
+an image or just tables.  In that case, pass FALSE and check the return value
+from jpeg_read_header(): it will be JPEG_HEADER_OK if an image was found,
+JPEG_HEADER_TABLES_ONLY if only tables were found.  (A third return value,
+JPEG_SUSPENDED, is possible when using a suspending data source manager.)
+Note that jpeg_read_header() will not complain if you read an abbreviated
+image for which you haven't loaded the missing tables; the missing-table check
+occurs later, in jpeg_start_decompress().
+
+
+It is possible to read a series of images from a single source file by
+repeating the jpeg_read_header() ... jpeg_finish_decompress() sequence,
+without releasing/recreating the JPEG object or the data source module.
+(If you did reinitialize, any partial bufferload left in the data source
+buffer at the end of one image would be discarded, causing you to lose the
+start of the next image.)  When you use this method, stored tables are
+automatically carried forward, so some of the images can be abbreviated images
+that depend on tables from earlier images.
+
+If you intend to write a series of images into a single destination file,
+you might want to make a specialized data destination module that doesn't
+flush the output buffer at term_destination() time.  This would speed things
+up by some trifling amount.  Of course, you'd need to remember to flush the
+buffer after the last image.  You can make the later images be abbreviated
+ones by passing FALSE to jpeg_start_compress().
+
+
+Special markers
+---------------
+
+Some applications may need to insert or extract special data in the JPEG
+datastream.  The JPEG standard provides marker types "COM" (comment) and
+"APP0" through "APP15" (application) to hold application-specific data.
+Unfortunately, the use of these markers is not specified by the standard.
+COM markers are fairly widely used to hold user-supplied text.  The JFIF file
+format spec uses APP0 markers with specified initial strings to hold certain
+data.  Adobe applications use APP14 markers beginning with the string "Adobe"
+for miscellaneous data.  Other APPn markers are rarely seen, but might
+contain almost anything.
+
+If you wish to store user-supplied text, we recommend you use COM markers
+and place readable 7-bit ASCII text in them.  Newline conventions are not
+standardized --- expect to find LF (Unix style), CR/LF (DOS style), or CR
+(Mac style).  A robust COM reader should be able to cope with random binary
+garbage, including nulls, since some applications generate COM markers
+containing non-ASCII junk.  (But yours should not be one of them.)
+
+For program-supplied data, use an APPn marker, and be sure to begin it with an
+identifying string so that you can tell whether the marker is actually yours.
+It's probably best to avoid using APP0 or APP14 for any private markers.
+(NOTE: the upcoming SPIFF standard will use APP8 markers; we recommend you
+not use APP8 markers for any private purposes, either.)
+
+Keep in mind that at most 65533 bytes can be put into one marker, but you
+can have as many markers as you like.
+
+By default, the IJG compression library will write a JFIF APP0 marker if the
+selected JPEG colorspace is grayscale or YCbCr, or an Adobe APP14 marker if
+the selected colorspace is RGB, CMYK, or YCCK.  You can disable this, but
+we don't recommend it.  The decompression library will recognize JFIF and
+Adobe markers and will set the JPEG colorspace properly when one is found.
+
+
+You can write special markers immediately following the datastream header by
+calling jpeg_write_marker() after jpeg_start_compress() and before the first
+call to jpeg_write_scanlines().  When you do this, the markers appear after
+the SOI and the JFIF APP0 and Adobe APP14 markers (if written), but before
+all else.  Specify the marker type parameter as "JPEG_COM" for COM or
+"JPEG_APP0 + n" for APPn.  (Actually, jpeg_write_marker will let you write
+any marker type, but we don't recommend writing any other kinds of marker.)
+For example, to write a user comment string pointed to by comment_text:
+        jpeg_write_marker(cinfo, JPEG_COM, comment_text, strlen(comment_text));
+
+If it's not convenient to store all the marker data in memory at once,
+you can instead call jpeg_write_m_header() followed by multiple calls to
+jpeg_write_m_byte().  If you do it this way, it's your responsibility to
+call jpeg_write_m_byte() exactly the number of times given in the length
+parameter to jpeg_write_m_header().  (This method lets you empty the
+output buffer partway through a marker, which might be important when
+using a suspending data destination module.  In any case, if you are using
+a suspending destination, you should flush its buffer after inserting
+any special markers.  See "I/O suspension".)
+
+Or, if you prefer to synthesize the marker byte sequence yourself,
+you can just cram it straight into the data destination module.
+
+If you are writing JFIF 1.02 extension markers (thumbnail images), don't
+forget to set cinfo.JFIF_minor_version = 2 so that the encoder will write the
+correct JFIF version number in the JFIF header marker.  The library's default
+is to write version 1.01, but that's wrong if you insert any 1.02 extension
+markers.  (We could probably get away with just defaulting to 1.02, but there
+used to be broken decoders that would complain about unknown minor version
+numbers.  To reduce compatibility risks it's safest not to write 1.02 unless
+you are actually using 1.02 extensions.)
+
+
+When reading, two methods of handling special markers are available:
+1. You can ask the library to save the contents of COM and/or APPn markers
+into memory, and then examine them at your leisure afterwards.
+2. You can supply your own routine to process COM and/or APPn markers
+on-the-fly as they are read.
+The first method is simpler to use, especially if you are using a suspending
+data source; writing a marker processor that copes with input suspension is
+not easy (consider what happens if the marker is longer than your available
+input buffer).  However, the second method conserves memory since the marker
+data need not be kept around after it's been processed.
+
+For either method, you'd normally set up marker handling after creating a
+decompression object and before calling jpeg_read_header(), because the
+markers of interest will typically be near the head of the file and so will
+be scanned by jpeg_read_header.  Once you've established a marker handling
+method, it will be used for the life of that decompression object
+(potentially many datastreams), unless you change it.  Marker handling is
+determined separately for COM markers and for each APPn marker code.
+
+
+To save the contents of special markers in memory, call
+        jpeg_save_markers(cinfo, marker_code, length_limit)
+where marker_code is the marker type to save, JPEG_COM or JPEG_APP0+n.
+(To arrange to save all the special marker types, you need to call this
+routine 17 times, for COM and APP0-APP15.)  If the incoming marker is longer
+than length_limit data bytes, only length_limit bytes will be saved; this
+parameter allows you to avoid chewing up memory when you only need to see the
+first few bytes of a potentially large marker.  If you want to save all the
+data, set length_limit to 0xFFFF; that is enough since marker lengths are only
+16 bits.  As a special case, setting length_limit to 0 prevents that marker
+type from being saved at all.  (That is the default behavior, in fact.)
+
+After jpeg_read_header() completes, you can examine the special markers by
+following the cinfo->marker_list pointer chain.  All the special markers in
+the file appear in this list, in order of their occurrence in the file (but
+omitting any markers of types you didn't ask for).  Both the original data
+length and the saved data length are recorded for each list entry; the latter
+will not exceed length_limit for the particular marker type.  Note that these
+lengths exclude the marker length word, whereas the stored representation
+within the JPEG file includes it.  (Hence the maximum data length is really
+only 65533.)
+
+It is possible that additional special markers appear in the file beyond the
+SOS marker at which jpeg_read_header stops; if so, the marker list will be
+extended during reading of the rest of the file.  This is not expected to be
+common, however.  If you are short on memory you may want to reset the length
+limit to zero for all marker types after finishing jpeg_read_header, to
+ensure that the max_memory_to_use setting cannot be exceeded due to addition
+of later markers.
+
+The marker list remains stored until you call jpeg_finish_decompress or
+jpeg_abort, at which point the memory is freed and the list is set to empty.
+(jpeg_destroy also releases the storage, of course.)
+
+Note that the library is internally interested in APP0 and APP14 markers;
+if you try to set a small nonzero length limit on these types, the library
+will silently force the length up to the minimum it wants.  (But you can set
+a zero length limit to prevent them from being saved at all.)  Also, in a
+16-bit environment, the maximum length limit may be constrained to less than
+65533 by malloc() limitations.  It is therefore best not to assume that the
+effective length limit is exactly what you set it to be.
+
+
+If you want to supply your own marker-reading routine, you do it by calling
+jpeg_set_marker_processor().  A marker processor routine must have the
+signature
+        boolean jpeg_marker_parser_method (j_decompress_ptr cinfo)
+Although the marker code is not explicitly passed, the routine can find it
+in cinfo->unread_marker.  At the time of call, the marker proper has been
+read from the data source module.  The processor routine is responsible for
+reading the marker length word and the remaining parameter bytes, if any.
+Return TRUE to indicate success.  (FALSE should be returned only if you are
+using a suspending data source and it tells you to suspend.  See the standard
+marker processors in jdmarker.c for appropriate coding methods if you need to
+use a suspending data source.)
+
+If you override the default APP0 or APP14 processors, it is up to you to
+recognize JFIF and Adobe markers if you want colorspace recognition to occur
+properly.  We recommend copying and extending the default processors if you
+want to do that.  (A better idea is to save these marker types for later
+examination by calling jpeg_save_markers(); that method doesn't interfere
+with the library's own processing of these markers.)
+
+jpeg_set_marker_processor() and jpeg_save_markers() are mutually exclusive
+--- if you call one it overrides any previous call to the other, for the
+particular marker type specified.
+
+A simple example of an external COM processor can be found in djpeg.c.
+Also, see jpegtran.c for an example of using jpeg_save_markers.
+
+
+Raw (downsampled) image data
+----------------------------
+
+Some applications need to supply already-downsampled image data to the JPEG
+compressor, or to receive raw downsampled data from the decompressor.  The
+library supports this requirement by allowing the application to write or
+read raw data, bypassing the normal preprocessing or postprocessing steps.
+The interface is different from the standard one and is somewhat harder to
+use.  If your interest is merely in bypassing color conversion, we recommend
+that you use the standard interface and simply set jpeg_color_space =
+in_color_space (or jpeg_color_space = out_color_space for decompression).
+The mechanism described in this section is necessary only to supply or
+receive downsampled image data, in which not all components have the same
+dimensions.
+
+
+To compress raw data, you must supply the data in the colorspace to be used
+in the JPEG file (please read the earlier section on Special color spaces)
+and downsampled to the sampling factors specified in the JPEG parameters.
+You must supply the data in the format used internally by the JPEG library,
+namely a JSAMPIMAGE array.  This is an array of pointers to two-dimensional
+arrays, each of type JSAMPARRAY.  Each 2-D array holds the values for one
+color component.  This structure is necessary since the components are of
+different sizes.  If the image dimensions are not a multiple of the MCU size,
+you must also pad the data correctly (usually, this is done by replicating
+the last column and/or row).  The data must be padded to a multiple of a DCT
+block in each component: that is, each downsampled row must contain a
+multiple of 8 valid samples, and there must be a multiple of 8 sample rows
+for each component.  (For applications such as conversion of digital TV
+images, the standard image size is usually a multiple of the DCT block size,
+so that no padding need actually be done.)
+
+The procedure for compression of raw data is basically the same as normal
+compression, except that you call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  Before calling jpeg_start_compress(), you must do
+the following:
+  * Set cinfo->raw_data_in to TRUE.  (It is set FALSE by jpeg_set_defaults().)
+    This notifies the library that you will be supplying raw data.
+  * Ensure jpeg_color_space is correct --- an explicit jpeg_set_colorspace()
+    call is a good idea.  Note that since color conversion is bypassed,
+    in_color_space is ignored, except that jpeg_set_defaults() uses it to
+    choose the default jpeg_color_space setting.
+  * Ensure the sampling factors, cinfo->comp_info[i].h_samp_factor and
+    cinfo->comp_info[i].v_samp_factor, are correct.  Since these indicate the
+    dimensions of the data you are supplying, it's wise to set them
+    explicitly, rather than assuming the library's defaults are what you want.
+
+To pass raw data to the library, call jpeg_write_raw_data() in place of
+jpeg_write_scanlines().  The two routines work similarly except that
+jpeg_write_raw_data takes a JSAMPIMAGE data array rather than JSAMPARRAY.
+The scanlines count passed to and returned from jpeg_write_raw_data is
+measured in terms of the component with the largest v_samp_factor.
+
+jpeg_write_raw_data() processes one MCU row per call, which is to say
+v_samp_factor*DCTSIZE sample rows of each component.  The passed num_lines
+value must be at least max_v_samp_factor*DCTSIZE, and the return value will
+be exactly that amount (or possibly some multiple of that amount, in future
+library versions).  This is true even on the last call at the bottom of the
+image; don't forget to pad your data as necessary.
+
+The required dimensions of the supplied data can be computed for each
+component as
+        cinfo->comp_info[i].width_in_blocks*DCTSIZE  samples per row
+        cinfo->comp_info[i].height_in_blocks*DCTSIZE rows in image
+after jpeg_start_compress() has initialized those fields.  If the valid data
+is smaller than this, it must be padded appropriately.  For some sampling
+factors and image sizes, additional dummy DCT blocks are inserted to make
+the image a multiple of the MCU dimensions.  The library creates such dummy
+blocks itself; it does not read them from your supplied data.  Therefore you
+need never pad by more than DCTSIZE samples.  An example may help here.
+Assume 2h2v downsampling of YCbCr data, that is
+        cinfo->comp_info[0].h_samp_factor = 2           for Y
+        cinfo->comp_info[0].v_samp_factor = 2
+        cinfo->comp_info[1].h_samp_factor = 1           for Cb
+        cinfo->comp_info[1].v_samp_factor = 1
+        cinfo->comp_info[2].h_samp_factor = 1           for Cr
+        cinfo->comp_info[2].v_samp_factor = 1
+and suppose that the nominal image dimensions (cinfo->image_width and
+cinfo->image_height) are 101x101 pixels.  Then jpeg_start_compress() will
+compute downsampled_width = 101 and width_in_blocks = 13 for Y,
+downsampled_width = 51 and width_in_blocks = 7 for Cb and Cr (and the same
+for the height fields).  You must pad the Y data to at least 13*8 = 104
+columns and rows, the Cb/Cr data to at least 7*8 = 56 columns and rows.  The
+MCU height is max_v_samp_factor = 2 DCT rows so you must pass at least 16
+scanlines on each call to jpeg_write_raw_data(), which is to say 16 actual
+sample rows of Y and 8 each of Cb and Cr.  A total of 7 MCU rows are needed,
+so you must pass a total of 7*16 = 112 "scanlines".  The last DCT block row
+of Y data is dummy, so it doesn't matter what you pass for it in the data
+arrays, but the scanlines count must total up to 112 so that all of the Cb
+and Cr data gets passed.
+
+Output suspension is supported with raw-data compression: if the data
+destination module suspends, jpeg_write_raw_data() will return 0.
+In this case the same data rows must be passed again on the next call.
+
+
+Decompression with raw data output implies bypassing all postprocessing:
+you cannot ask for rescaling or color quantization, for instance.  More
+seriously, you must deal with the color space and sampling factors present in
+the incoming file.  If your application only handles, say, 2h1v YCbCr data,
+you must check for and fail on other color spaces or other sampling factors.
+The library will not convert to a different color space for you.
+
+To obtain raw data output, set cinfo->raw_data_out = TRUE before
+jpeg_start_decompress() (it is set FALSE by jpeg_read_header()).  Be sure to
+verify that the color space and sampling factors are ones you can handle.
+Then call jpeg_read_raw_data() in place of jpeg_read_scanlines().  The
+decompression process is otherwise the same as usual.
+
+jpeg_read_raw_data() returns one MCU row per call, and thus you must pass a
+buffer of at least max_v_samp_factor*DCTSIZE scanlines (scanline counting is
+the same as for raw-data compression).  The buffer you pass must be large
+enough to hold the actual data plus padding to DCT-block boundaries.  As with
+compression, any entirely dummy DCT blocks are not processed so you need not
+allocate space for them, but the total scanline count includes them.  The
+above example of computing buffer dimensions for raw-data compression is
+equally valid for decompression.
+
+Input suspension is supported with raw-data decompression: if the data source
+module suspends, jpeg_read_raw_data() will return 0.  You can also use
+buffered-image mode to read raw data in multiple passes.
+
+
+Really raw data: DCT coefficients
+---------------------------------
+
+It is possible to read or write the contents of a JPEG file as raw DCT
+coefficients.  This facility is mainly intended for use in lossless
+transcoding between different JPEG file formats.  Other possible applications
+include lossless cropping of a JPEG image, lossless reassembly of a
+multi-strip or multi-tile TIFF/JPEG file into a single JPEG datastream, etc.
+
+To read the contents of a JPEG file as DCT coefficients, open the file and do
+jpeg_read_header() as usual.  But instead of calling jpeg_start_decompress()
+and jpeg_read_scanlines(), call jpeg_read_coefficients().  This will read the
+entire image into a set of virtual coefficient-block arrays, one array per
+component.  The return value is a pointer to an array of virtual-array
+descriptors.  Each virtual array can be accessed directly using the JPEG
+memory manager's access_virt_barray method (see Memory management, below,
+and also read structure.txt's discussion of virtual array handling).  Or,
+for simple transcoding to a different JPEG file format, the array list can
+just be handed directly to jpeg_write_coefficients().
+
+Each block in the block arrays contains quantized coefficient values in
+normal array order (not JPEG zigzag order).  The block arrays contain only
+DCT blocks containing real data; any entirely-dummy blocks added to fill out
+interleaved MCUs at the right or bottom edges of the image are discarded
+during reading and are not stored in the block arrays.  (The size of each
+block array can be determined from the width_in_blocks and height_in_blocks
+fields of the component's comp_info entry.)  This is also the data format
+expected by jpeg_write_coefficients().
+
+When you are done using the virtual arrays, call jpeg_finish_decompress()
+to release the array storage and return the decompression object to an idle
+state; or just call jpeg_destroy() if you don't need to reuse the object.
+
+If you use a suspending data source, jpeg_read_coefficients() will return
+NULL if it is forced to suspend; a non-NULL return value indicates successful
+completion.  You need not test for a NULL return value when using a
+non-suspending data source.
+
+It is also possible to call jpeg_read_coefficients() to obtain access to the
+decoder's coefficient arrays during a normal decode cycle in buffered-image
+mode.  This frammish might be useful for progressively displaying an incoming
+image and then re-encoding it without loss.  To do this, decode in buffered-
+image mode as discussed previously, then call jpeg_read_coefficients() after
+the last jpeg_finish_output() call.  The arrays will be available for your use
+until you call jpeg_finish_decompress().
+
+
+To write the contents of a JPEG file as DCT coefficients, you must provide
+the DCT coefficients stored in virtual block arrays.  You can either pass
+block arrays read from an input JPEG file by jpeg_read_coefficients(), or
+allocate virtual arrays from the JPEG compression object and fill them
+yourself.  In either case, jpeg_write_coefficients() is substituted for
+jpeg_start_compress() and jpeg_write_scanlines().  Thus the sequence is
+  * Create compression object
+  * Set all compression parameters as necessary
+  * Request virtual arrays if needed
+  * jpeg_write_coefficients()
+  * jpeg_finish_compress()
+  * Destroy or re-use compression object
+jpeg_write_coefficients() is passed a pointer to an array of virtual block
+array descriptors; the number of arrays is equal to cinfo.num_components.
+
+The virtual arrays need only have been requested, not realized, before
+jpeg_write_coefficients() is called.  A side-effect of
+jpeg_write_coefficients() is to realize any virtual arrays that have been
+requested from the compression object's memory manager.  Thus, when obtaining
+the virtual arrays from the compression object, you should fill the arrays
+after calling jpeg_write_coefficients().  The data is actually written out
+when you call jpeg_finish_compress(); jpeg_write_coefficients() only writes
+the file header.
+
+When writing raw DCT coefficients, it is crucial that the JPEG quantization
+tables and sampling factors match the way the data was encoded, or the
+resulting file will be invalid.  For transcoding from an existing JPEG file,
+we recommend using jpeg_copy_critical_parameters().  This routine initializes
+all the compression parameters to default values (like jpeg_set_defaults()),
+then copies the critical information from a source decompression object.
+The decompression object should have just been used to read the entire
+JPEG input file --- that is, it should be awaiting jpeg_finish_decompress().
+
+jpeg_write_coefficients() marks all tables stored in the compression object
+as needing to be written to the output file (thus, it acts like
+jpeg_start_compress(cinfo, TRUE)).  This is for safety's sake, to avoid
+emitting abbreviated JPEG files by accident.  If you really want to emit an
+abbreviated JPEG file, call jpeg_suppress_tables(), or set the tables'
+individual sent_table flags, between calling jpeg_write_coefficients() and
+jpeg_finish_compress().
+
+
+Progress monitoring
+-------------------
+
+Some applications may need to regain control from the JPEG library every so
+often.  The typical use of this feature is to produce a percent-done bar or
+other progress display.  (For a simple example, see cjpeg.c or djpeg.c.)
+Although you do get control back frequently during the data-transferring pass
+(the jpeg_read_scanlines or jpeg_write_scanlines loop), any additional passes
+will occur inside jpeg_finish_compress or jpeg_start_decompress; those
+routines may take a long time to execute, and you don't get control back
+until they are done.
+
+You can define a progress-monitor routine which will be called periodically
+by the library.  No guarantees are made about how often this call will occur,
+so we don't recommend you use it for mouse tracking or anything like that.
+At present, a call will occur once per MCU row, scanline, or sample row
+group, whichever unit is convenient for the current processing mode; so the
+wider the image, the longer the time between calls.  During the data
+transferring pass, only one call occurs per call of jpeg_read_scanlines or
+jpeg_write_scanlines, so don't pass a large number of scanlines at once if
+you want fine resolution in the progress count.  (If you really need to use
+the callback mechanism for time-critical tasks like mouse tracking, you could
+insert additional calls inside some of the library's inner loops.)
+
+To establish a progress-monitor callback, create a struct jpeg_progress_mgr,
+fill in its progress_monitor field with a pointer to your callback routine,
+and set cinfo->progress to point to the struct.  The callback will be called
+whenever cinfo->progress is non-NULL.  (This pointer is set to NULL by
+jpeg_create_compress or jpeg_create_decompress; the library will not change
+it thereafter.  So if you allocate dynamic storage for the progress struct,
+make sure it will live as long as the JPEG object does.  Allocating from the
+JPEG memory manager with lifetime JPOOL_PERMANENT will work nicely.)  You
+can use the same callback routine for both compression and decompression.
+
+The jpeg_progress_mgr struct contains four fields which are set by the library:
+        long pass_counter;      /* work units completed in this pass */
+        long pass_limit;        /* total number of work units in this pass */
+        int completed_passes;   /* passes completed so far */
+        int total_passes;       /* total number of passes expected */
+During any one pass, pass_counter increases from 0 up to (not including)
+pass_limit; the step size is usually but not necessarily 1.  The pass_limit
+value may change from one pass to another.  The expected total number of
+passes is in total_passes, and the number of passes already completed is in
+completed_passes.  Thus the fraction of work completed may be estimated as
+                completed_passes + (pass_counter/pass_limit)
+                --------------------------------------------
+                                total_passes
+ignoring the fact that the passes may not be equal amounts of work.
+
+When decompressing, pass_limit can even change within a pass, because it
+depends on the number of scans in the JPEG file, which isn't always known in
+advance.  The computed fraction-of-work-done may jump suddenly (if the library
+discovers it has overestimated the number of scans) or even decrease (in the
+opposite case).  It is not wise to put great faith in the work estimate.
+
+When using the decompressor's buffered-image mode, the progress monitor work
+estimate is likely to be completely unhelpful, because the library has no way
+to know how many output passes will be demanded of it.  Currently, the library
+sets total_passes based on the assumption that there will be one more output
+pass if the input file end hasn't yet been read (jpeg_input_complete() isn't
+TRUE), but no more output passes if the file end has been reached when the
+output pass is started.  This means that total_passes will rise as additional
+output passes are requested.  If you have a way of determining the input file
+size, estimating progress based on the fraction of the file that's been read
+will probably be more useful than using the library's value.
+
+
+Memory management
+-----------------
+
+This section covers some key facts about the JPEG library's built-in memory
+manager.  For more info, please read structure.txt's section about the memory
+manager, and consult the source code if necessary.
+
+All memory and temporary file allocation within the library is done via the
+memory manager.  If necessary, you can replace the "back end" of the memory
+manager to control allocation yourself (for example, if you don't want the
+library to use malloc() and free() for some reason).
+
+Some data is allocated "permanently" and will not be freed until the JPEG
+object is destroyed.  Most data is allocated "per image" and is freed by
+jpeg_finish_compress, jpeg_finish_decompress, or jpeg_abort.  You can call the
+memory manager yourself to allocate structures that will automatically be
+freed at these times.  Typical code for this is
+  ptr = (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE, size);
+Use JPOOL_PERMANENT to get storage that lasts as long as the JPEG object.
+Use alloc_large instead of alloc_small for anything bigger than a few Kbytes.
+There are also alloc_sarray and alloc_barray routines that automatically
+build 2-D sample or block arrays.
+
+The library's minimum space requirements to process an image depend on the
+image's width, but not on its height, because the library ordinarily works
+with "strip" buffers that are as wide as the image but just a few rows high.
+Some operating modes (eg, two-pass color quantization) require full-image
+buffers.  Such buffers are treated as "virtual arrays": only the current strip
+need be in memory, and the rest can be swapped out to a temporary file.
+
+If you use the simplest memory manager back end (jmemnobs.c), then no
+temporary files are used; virtual arrays are simply malloc()'d.  Images bigger
+than memory can be processed only if your system supports virtual memory.
+The other memory manager back ends support temporary files of various flavors
+and thus work in machines without virtual memory.  They may also be useful on
+Unix machines if you need to process images that exceed available swap space.
+
+When using temporary files, the library will make the in-memory buffers for
+its virtual arrays just big enough to stay within a "maximum memory" setting.
+Your application can set this limit by setting cinfo->mem->max_memory_to_use
+after creating the JPEG object.  (Of course, there is still a minimum size for
+the buffers, so the max-memory setting is effective only if it is bigger than
+the minimum space needed.)  If you allocate any large structures yourself, you
+must allocate them before jpeg_start_compress() or jpeg_start_decompress() in
+order to have them counted against the max memory limit.  Also keep in mind
+that space allocated with alloc_small() is ignored, on the assumption that
+it's too small to be worth worrying about; so a reasonable safety margin
+should be left when setting max_memory_to_use.
+
+
+Memory usage
+------------
+
+Working memory requirements while performing compression or decompression
+depend on image dimensions, image characteristics (such as colorspace and
+JPEG process), and operating mode (application-selected options).
+
+As of v6b, the decompressor requires:
+ 1. About 24K in more-or-less-fixed-size data.  This varies a bit depending
+    on operating mode and image characteristics (particularly color vs.
+    grayscale), but it doesn't depend on image dimensions.
+ 2. Strip buffers (of size proportional to the image width) for IDCT and
+    upsampling results.  The worst case for commonly used sampling factors
+    is about 34 bytes * width in pixels for a color image.  A grayscale image
+    only needs about 8 bytes per pixel column.
+ 3. A full-image DCT coefficient buffer is needed to decode a multi-scan JPEG
+    file (including progressive JPEGs), or whenever you select buffered-image
+    mode.  This takes 2 bytes/coefficient.  At typical 2x2 sampling, that's
+    3 bytes per pixel for a color image.  Worst case (1x1 sampling) requires
+    6 bytes/pixel.  For grayscale, figure 2 bytes/pixel.
+ 4. To perform 2-pass color quantization, the decompressor also needs a
+    128K color lookup table and a full-image pixel buffer (3 bytes/pixel).
+This does not count any memory allocated by the application, such as a
+buffer to hold the final output image.
+
+The above figures are valid for 8-bit JPEG data precision and a machine with
+32-bit ints.  For 12-bit JPEG data, double the size of the strip buffers and
+quantization pixel buffer.  The "fixed-size" data will be somewhat smaller
+with 16-bit ints, larger with 64-bit ints.  Also, CMYK or other unusual
+color spaces will require different amounts of space.
+
+The full-image coefficient and pixel buffers, if needed at all, do not
+have to be fully RAM resident; you can have the library use temporary
+files instead when the total memory usage would exceed a limit you set.
+(But if your OS supports virtual memory, it's probably better to just use
+jmemnobs and let the OS do the swapping.)
+
+The compressor's memory requirements are similar, except that it has no need
+for color quantization.  Also, it needs a full-image DCT coefficient buffer
+if Huffman-table optimization is asked for, even if progressive mode is not
+requested.
+
+If you need more detailed information about memory usage in a particular
+situation, you can enable the MEM_STATS code in jmemmgr.c.
+
+
+Library compile-time options
+----------------------------
+
+A number of compile-time options are available by modifying jmorecfg.h.
+
+The JPEG standard provides for both the baseline 8-bit DCT process and
+a 12-bit DCT process.  The IJG code supports 12-bit lossy JPEG if you define
+BITS_IN_JSAMPLE as 12 rather than 8.  Note that this causes JSAMPLE to be
+larger than a char, so it affects the surrounding application's image data.
+The sample applications cjpeg and djpeg can support 12-bit mode only for PPM
+and GIF file formats; you must disable the other file formats to compile a
+12-bit cjpeg or djpeg.  (install.txt has more information about that.)
+At present, a 12-bit library can handle *only* 12-bit images, not both
+precisions.
+
+Note that a 12-bit library always compresses in Huffman optimization mode,
+in order to generate valid Huffman tables.  This is necessary because our
+default Huffman tables only cover 8-bit data.  If you need to output 12-bit
+files in one pass, you'll have to supply suitable default Huffman tables.
+You may also want to supply your own DCT quantization tables; the existing
+quality-scaling code has been developed for 8-bit use, and probably doesn't
+generate especially good tables for 12-bit.
+
+The maximum number of components (color channels) in the image is determined
+by MAX_COMPONENTS.  The JPEG standard allows up to 255 components, but we
+expect that few applications will need more than four or so.
+
+On machines with unusual data type sizes, you may be able to improve
+performance or reduce memory space by tweaking the various typedefs in
+jmorecfg.h.  In particular, on some RISC CPUs, access to arrays of "short"s
+is quite slow; consider trading memory for speed by making JCOEF, INT16, and
+UINT16 be "int" or "unsigned int".  UINT8 is also a candidate to become int.
+You probably don't want to make JSAMPLE be int unless you have lots of memory
+to burn.
+
+You can reduce the size of the library by compiling out various optional
+functions.  To do this, undefine xxx_SUPPORTED symbols as necessary.
+
+You can also save a few K by not having text error messages in the library;
+the standard error message table occupies about 5Kb.  This is particularly
+reasonable for embedded applications where there's no good way to display
+a message anyway.  To do this, remove the creation of the message table
+(jpeg_std_message_table[]) from jerror.c, and alter format_message to do
+something reasonable without it.  You could output the numeric value of the
+message code number, for example.  If you do this, you can also save a couple
+more K by modifying the TRACEMSn() macros in jerror.h to expand to nothing;
+you don't need trace capability anyway, right?
+
+
+Portability considerations
+--------------------------
+
+The JPEG library has been written to be extremely portable; the sample
+applications cjpeg and djpeg are slightly less so.  This section summarizes
+the design goals in this area.  (If you encounter any bugs that cause the
+library to be less portable than is claimed here, we'd appreciate hearing
+about them.)
+
+The code works fine on ANSI C and C++ compilers, using any of the popular
+system include file setups, and some not-so-popular ones too.
+
+The code is not dependent on the exact sizes of the C data types.  As
+distributed, we make the assumptions that
+        char    is at least 8 bits wide
+        short   is at least 16 bits wide
+        int     is at least 16 bits wide
+        long    is at least 32 bits wide
+(These are the minimum requirements of the ANSI C standard.)  Wider types will
+work fine, although memory may be used inefficiently if char is much larger
+than 8 bits or short is much bigger than 16 bits.  The code should work
+equally well with 16- or 32-bit ints.
+
+In a system where these assumptions are not met, you may be able to make the
+code work by modifying the typedefs in jmorecfg.h.  However, you will probably
+have difficulty if int is less than 16 bits wide, since references to plain
+int abound in the code.
+
+char can be either signed or unsigned, although the code runs faster if an
+unsigned char type is available.  If char is wider than 8 bits, you will need
+to redefine JOCTET and/or provide custom data source/destination managers so
+that JOCTET represents exactly 8 bits of data on external storage.
+
+The JPEG library proper does not assume ASCII representation of characters.
+But some of the image file I/O modules in cjpeg/djpeg do have ASCII
+dependencies in file-header manipulation; so does cjpeg's select_file_type()
+routine.
+
+The JPEG library does not rely heavily on the C library.  In particular, C
+stdio is used only by the data source/destination modules and the error
+handler, all of which are application-replaceable.  (cjpeg/djpeg are more
+heavily dependent on stdio.)  malloc and free are called only from the memory
+manager "back end" module, so you can use a different memory allocator by
+replacing that one file.
+
+More info about porting the code may be gleaned by reading jconfig.txt,
+jmorecfg.h, and jinclude.h.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/LICENSE.md glmark2-2021.02/src/libjpeg-turbo/LICENSE.md
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/LICENSE.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/LICENSE.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,88 @@
+libjpeg-turbo Licenses
+======================
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses:
+
+- The IJG (Independent JPEG Group) License, which is listed in
+  [README.ijg](README.ijg)
+
+  This license applies to the libjpeg API library and associated programs
+  (any code inherited from libjpeg, and any modifications to that code.)
+
+- The Modified (3-clause) BSD License, which is listed in
+  [turbojpeg.c](turbojpeg.c)
+
+  This license covers the TurboJPEG API library and associated programs.
+
+- The zlib License, which is listed in [simd/jsimdext.inc](simd/jsimdext.inc)
+
+  This license is a subset of the other two, and it covers the libjpeg-turbo
+  SIMD extensions.
+
+
+Complying with the libjpeg-turbo Licenses
+=========================================
+
+This section provides a roll-up of the libjpeg-turbo licensing terms, to the
+best of our understanding.
+
+1.  If you are distributing a modified version of the libjpeg-turbo source,
+    then:
+
+    1.  You cannot alter or remove any existing copyright or license notices
+        from the source.
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 1 of the Modified BSD License
+        - Clauses 1 and 3 of the zlib License
+
+    2.  You must add your own copyright notice to the header of each source
+        file you modified, so others can tell that you modified that file (if
+        there is not an existing copyright header in that file, then you can
+        simply add a notice stating that you modified the file.)
+
+        **Origin**
+        - Clause 1 of the IJG License
+        - Clause 2 of the zlib License
+
+    3.  You must include the IJG README file, and you must not alter any of the
+        copyright or license text in that file.
+
+        **Origin**
+        - Clause 1 of the IJG License
+
+2.  If you are distributing only libjpeg-turbo binaries without the source, or
+    if you are distributing an application that statically links with
+    libjpeg-turbo, then:
+
+    1.  Your product documentation must include a message stating:
+
+        This software is based in part on the work of the Independent JPEG
+        Group.
+
+        **Origin**
+        - Clause 2 of the IJG license
+
+    2.  If your binary distribution includes or uses the TurboJPEG API, then
+        your product documentation must include the text of the Modified BSD
+        License.
+
+        **Origin**
+        - Clause 2 of the Modified BSD License
+
+3.  You cannot use the name of the IJG or The libjpeg-turbo Project or the
+    contributors thereof in advertising, publicity, etc.
+
+    **Origin**
+    - IJG License
+    - Clause 3 of the Modified BSD License
+
+4.  The IJG and The libjpeg-turbo Project do not warrant libjpeg-turbo to be
+    free of defects, nor do we accept any liability for undesirable
+    consequences resulting from your use of the software.
+
+    **Origin**
+    - IJG License
+    - Modified BSD License
+    - zlib License
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/Makefile.am glmark2-2021.02/src/libjpeg-turbo/Makefile.am
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/Makefile.am	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/Makefile.am	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,778 @@
+lib_LTLIBRARIES = libjpeg.la
+libjpeg_la_LDFLAGS = -version-info ${LIBTOOL_CURRENT}:${SO_MINOR_VERSION}:${SO_AGE} -no-undefined
+include_HEADERS = jerror.h jmorecfg.h jpeglib.h
+
+if WITH_TURBOJPEG
+lib_LTLIBRARIES += libturbojpeg.la
+libturbojpeg_la_LDFLAGS = -version-info 1:0:1 -no-undefined
+include_HEADERS += turbojpeg.h
+endif
+
+nodist_include_HEADERS = jconfig.h
+
+pkgconfigdir = $(libdir)/pkgconfig
+pkgconfig_DATA = pkgscripts/libjpeg.pc pkgscripts/libturbojpeg.pc
+
+HDRS = jchuff.h jdct.h jdhuff.h jerror.h jinclude.h jmemsys.h jmorecfg.h \
+	jpegint.h jpeglib.h jversion.h jsimd.h jsimddct.h jpegcomp.h \
+	jpeg_nbits_table.h
+
+libjpeg_la_SOURCES = $(HDRS) jcapimin.c jcapistd.c jccoefct.c jccolor.c \
+	jcdctmgr.c jchuff.c jcinit.c jcmainct.c jcmarker.c jcmaster.c \
+	jcomapi.c jcparam.c jcphuff.c jcprepct.c jcsample.c jctrans.c \
+	jdapimin.c jdapistd.c jdatadst.c jdatasrc.c jdcoefct.c jdcolor.c \
+	jddctmgr.c jdhuff.c jdinput.c jdmainct.c jdmarker.c jdmaster.c \
+	jdmerge.c jdphuff.c jdpostct.c jdsample.c jdtrans.c jerror.c \
+	jfdctflt.c jfdctfst.c jfdctint.c jidctflt.c jidctfst.c jidctint.c \
+	jidctred.c jquant1.c jquant2.c jutils.c jmemmgr.c jmemnobs.c
+
+if WITH_ARITH
+libjpeg_la_SOURCES += jaricom.c
+endif
+
+if WITH_ARITH_ENC
+libjpeg_la_SOURCES += jcarith.c
+endif
+
+if WITH_ARITH_DEC
+libjpeg_la_SOURCES += jdarith.c
+endif
+
+
+SUBDIRS = java
+
+
+if WITH_TURBOJPEG
+
+libturbojpeg_la_SOURCES = $(libjpeg_la_SOURCES) turbojpeg.c turbojpeg.h \
+	transupp.c transupp.h jdatadst-tj.c jdatasrc-tj.c
+
+if WITH_JAVA
+
+libturbojpeg_la_SOURCES += turbojpeg-jni.c
+libturbojpeg_la_CFLAGS = ${JNI_CFLAGS}
+TJMAPFILE = turbojpeg-mapfile.jni
+
+else
+
+TJMAPFILE = turbojpeg-mapfile
+
+endif
+
+libturbojpeg_la_SOURCES += $(TJMAPFILE)
+
+if VERSION_SCRIPT
+libturbojpeg_la_LDFLAGS += $(VERSION_SCRIPT_FLAG)$(srcdir)/$(TJMAPFILE)
+endif
+
+endif
+
+
+if VERSION_SCRIPT
+libjpeg_la_LDFLAGS += $(VERSION_SCRIPT_FLAG)libjpeg.map
+endif
+
+
+if WITH_SIMD
+
+SUBDIRS += simd
+libjpeg_la_LIBADD = simd/libsimd.la
+libturbojpeg_la_LIBADD = simd/libsimd.la
+
+else
+
+libjpeg_la_SOURCES += jsimd_none.c
+
+endif
+
+
+bin_PROGRAMS = cjpeg djpeg jpegtran rdjpgcom wrjpgcom
+noinst_PROGRAMS = jcstest
+
+
+if WITH_TURBOJPEG
+
+bin_PROGRAMS += tjbench
+
+noinst_PROGRAMS += tjunittest
+
+tjbench_SOURCES = tjbench.c bmp.h bmp.c tjutil.h tjutil.c rdbmp.c rdppm.c \
+	wrbmp.c wrppm.c
+
+tjbench_LDADD = libturbojpeg.la libjpeg.la -lm
+
+tjbench_CFLAGS = -DBMP_SUPPORTED -DPPM_SUPPORTED
+
+tjunittest_SOURCES = tjunittest.c tjutil.h tjutil.c
+
+tjunittest_LDADD = libturbojpeg.la
+
+endif
+
+
+cjpeg_SOURCES = cdjpeg.h cderror.h cdjpeg.c cjpeg.c rdgif.c rdppm.c rdswitch.c
+if WITH_12BIT
+else
+cjpeg_SOURCES += rdbmp.c rdtarga.c
+endif
+
+cjpeg_LDADD = libjpeg.la
+
+cjpeg_CFLAGS = -DGIF_SUPPORTED -DPPM_SUPPORTED
+if WITH_12BIT
+else
+cjpeg_CFLAGS += -DBMP_SUPPORTED -DTARGA_SUPPORTED
+endif
+
+djpeg_SOURCES = cdjpeg.h cderror.h cdjpeg.c djpeg.c rdcolmap.c rdswitch.c \
+	wrgif.c wrppm.c
+if WITH_12BIT
+else
+djpeg_SOURCES += wrbmp.c wrtarga.c
+endif
+
+djpeg_LDADD = libjpeg.la
+
+djpeg_CFLAGS = -DGIF_SUPPORTED -DPPM_SUPPORTED
+if WITH_12BIT
+else
+djpeg_CFLAGS += -DBMP_SUPPORTED -DTARGA_SUPPORTED
+endif
+
+jpegtran_SOURCES = jpegtran.c rdswitch.c cdjpeg.c transupp.c transupp.h
+
+jpegtran_LDADD = libjpeg.la
+
+rdjpgcom_SOURCES = rdjpgcom.c
+
+rdjpgcom_LDADD = libjpeg.la
+
+wrjpgcom_SOURCES = wrjpgcom.c
+
+wrjpgcom_LDADD = libjpeg.la
+
+jcstest_SOURCES = jcstest.c
+
+jcstest_LDADD = libjpeg.la
+
+dist_man1_MANS = cjpeg.1 djpeg.1 jpegtran.1 rdjpgcom.1 wrjpgcom.1
+
+DOCS= coderules.txt jconfig.txt change.log rdrle.c wrrle.c BUILDING.md \
+	ChangeLog.txt
+
+dist_doc_DATA = README.ijg README.md libjpeg.txt structure.txt usage.txt \
+	wizard.txt LICENSE.md
+
+exampledir = $(docdir)
+dist_example_DATA = example.c
+
+
+EXTRA_DIST = win release $(DOCS) testimages CMakeLists.txt \
+	sharedlib/CMakeLists.txt cmakescripts libjpeg.map.in doc doxygen.config \
+	doxygen-extra.css jccolext.c jdcolext.c jdcol565.c jdmrgext.c jdmrg565.c \
+	jstdhuff.c jdcoefct.h jdmainct.h jdmaster.h jdsample.h wrppm.h \
+	md5/CMakeLists.txt
+
+dist-hook:
+	rm -rf `find $(distdir) -name .svn`
+
+
+SUBDIRS += md5
+
+if WITH_12BIT
+
+TESTORIG = testorig12.jpg
+MD5_JPEG_RGB_ISLOW = 9620f424569594bb9242b48498ad801f
+MD5_PPM_RGB_ISLOW = f3301d2219783b8b3d942b7239fa50c0
+MD5_JPEG_422_IFAST_OPT = 7322e3bd2f127f7de4b40d4480ce60e4
+MD5_PPM_422_IFAST = 79807fa552899e66a04708f533e16950
+MD5_PPM_422M_IFAST = 07737bfe8a7c1c87aaa393a0098d16b0
+MD5_JPEG_420_IFAST_Q100_PROG = a1da220b5604081863a504297ed59e55
+MD5_PPM_420_Q100_IFAST = 1b3730122709f53d007255e8dfd3305e
+MD5_PPM_420M_Q100_IFAST = 980a1a3c5bf9510022869d30b7d26566
+MD5_JPEG_GRAY_ISLOW = 235c90707b16e2e069f37c888b2636d9
+MD5_PPM_GRAY_ISLOW = 7213c10af507ad467da5578ca5ee1fca
+MD5_PPM_GRAY_ISLOW_RGB = e96ee81c30a6ed422d466338bd3de65d
+MD5_JPEG_420S_IFAST_OPT = 7af8e60be4d9c227ec63ac9b6630855e
+MD5_JPEG_3x2_FLOAT_PROG_SSE = a8c17daf77b457725ec929e215b603f8
+MD5_PPM_3x2_FLOAT_SSE = 42876ab9e5c2f76a87d08db5fbd57956
+MD5_JPEG_3x2_FLOAT_PROG_32BIT = a8c17daf77b457725ec929e215b603f8
+MD5_PPM_3x2_FLOAT_32BIT = 42876ab9e5c2f76a87d08db5fbd57956
+MD5_PPM_3x2_FLOAT_64BIT = d6fbc71153b3d8ded484dbc17c7b9cf4
+MD5_JPEG_3x2_IFAST_PROG = 1396cc2b7185cfe943d408c9d305339e
+MD5_PPM_3x2_IFAST = 3975985ef6eeb0a2cdc58daa651ccc00
+MD5_PPM_420M_ISLOW_2_1 = 4ca6be2a6f326ff9eaab63e70a8259c0
+MD5_PPM_420M_ISLOW_15_8 = 12aa9f9534c1b3d7ba047322226365eb
+MD5_PPM_420M_ISLOW_13_8 = f7e22817c7b25e1393e4ec101e9d4e96
+MD5_PPM_420M_ISLOW_11_8 = 800a16f9f4dc9b293197bfe11be10a82
+MD5_PPM_420M_ISLOW_9_8 = 06b7a92a9bc69f4dc36ec40f1937d55c
+MD5_PPM_420M_ISLOW_7_8 = 3ec444a14a4ab4eab88ffc49c48eca43
+MD5_PPM_420M_ISLOW_3_4 = 3e726b7ea872445b19437d1c1d4f0d93
+MD5_PPM_420M_ISLOW_5_8 = a8a771abdc94301d20ffac119b2caccd
+MD5_PPM_420M_ISLOW_1_2 = b419124dd5568b085787234866102866
+MD5_PPM_420M_ISLOW_3_8 = 343d19015531b7bbe746124127244fa8
+MD5_PPM_420M_ISLOW_1_4 = 35fd59d866e44659edfa3c18db2a3edb
+MD5_PPM_420M_ISLOW_1_8 = ccaed48ac0aedefda5d4abe4013f4ad7
+MD5_PPM_420_ISLOW_SKIP15_31 = 86664cd9dc956536409e44e244d20a97
+MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 = 452a21656115a163029cfba5c04fa76a
+MD5_PPM_444_ISLOW_SKIP1_6 = ef63901f71ef7a75cd78253fc0914f84
+MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 = 15b173fb5872d9575572fbcc1b05956f
+MD5_JPEG_CROP = cdb35ff4b4519392690ea040c56ea99c
+
+else
+
+TESTORIG = testorig.jpg
+MD5_JPEG_RGB_ISLOW = 768e970dd57b340ff1b83c9d3d47c77b
+MD5_PPM_RGB_ISLOW = 00a257f5393fef8821f2b88ac7421291
+MD5_BMP_RGB_ISLOW_565 = f07d2e75073e4bb10f6c6f4d36e2e3be
+MD5_BMP_RGB_ISLOW_565D = 4cfa0928ef3e6bb626d7728c924cfda4
+MD5_JPEG_422_IFAST_OPT = 2540287b79d913f91665e660303ab2c8
+MD5_PPM_422_IFAST = 35bd6b3f833bad23de82acea847129fa
+MD5_PPM_422M_IFAST = 8dbc65323d62cca7c91ba02dd1cfa81d
+MD5_BMP_422M_IFAST_565 = 3294bd4d9a1f2b3d08ea6020d0db7065
+MD5_BMP_422M_IFAST_565D = da98c9c7b6039511be4a79a878a9abc1
+MD5_JPEG_420_IFAST_Q100_PROG = 990cbe0329c882420a2094da7e5adade
+MD5_PPM_420_Q100_IFAST = 5a732542015c278ff43635e473a8a294
+MD5_PPM_420M_Q100_IFAST = ff692ee9323a3b424894862557c092f1
+MD5_JPEG_GRAY_ISLOW = 72b51f894b8f4a10b3ee3066770aa38d
+MD5_PPM_GRAY_ISLOW = 8d3596c56eace32f205deccc229aa5ed
+MD5_PPM_GRAY_ISLOW_RGB = 116424ac07b79e5e801f00508eab48ec
+MD5_BMP_GRAY_ISLOW_565 = 12f78118e56a2f48b966f792fedf23cc
+MD5_BMP_GRAY_ISLOW_565D = bdbbd616441a24354c98553df5dc82db
+MD5_JPEG_420S_IFAST_OPT = 388708217ac46273ca33086b22827ed8
+# See README.md for more details on why this next bit is necessary.
+MD5_JPEG_3x2_FLOAT_PROG_SSE = 343e3f8caf8af5986ebaf0bdc13b5c71
+MD5_PPM_3x2_FLOAT_SSE = 1a75f36e5904d6fc3a85a43da9ad89bb
+MD5_JPEG_3x2_FLOAT_PROG_32BIT = 9bca803d2042bd1eb03819e2bf92b3e5
+MD5_PPM_3x2_FLOAT_32BIT = f6bfab038438ed8f5522fbd33595dcdc
+MD5_PPM_3x2_FLOAT_64BIT = 0e917a34193ef976b679a6b069b1be26
+MD5_JPEG_3x2_IFAST_PROG = 1ee5d2c1a77f2da495f993c8c7cceca5
+MD5_PPM_3x2_IFAST = fd283664b3b49127984af0a7f118fccd
+MD5_JPEG_420_ISLOW_ARI = e986fb0a637a8d833d96e8a6d6d84ea1
+MD5_JPEG_444_ISLOW_PROGARI = 0a8f1c8f66e113c3cf635df0a475a617
+MD5_PPM_420M_IFAST_ARI = 72b59a99bcf1de24c5b27d151bde2437
+MD5_JPEG_420_ISLOW = 9a68f56bc76e466aa7e52f415d0f4a5f
+MD5_PPM_420M_ISLOW_2_1 = 9f9de8c0612f8d06869b960b05abf9c9
+MD5_PPM_420M_ISLOW_15_8 = b6875bc070720b899566cc06459b63b7
+MD5_PPM_420M_ISLOW_13_8 = bc3452573c8152f6ae552939ee19f82f
+MD5_PPM_420M_ISLOW_11_8 = d8cc73c0aaacd4556569b59437ba00a5
+MD5_PPM_420M_ISLOW_9_8 = d25e61bc7eac0002f5b393aa223747b6
+MD5_PPM_420M_ISLOW_7_8 = ddb564b7c74a09494016d6cd7502a946
+MD5_PPM_420M_ISLOW_3_4 = 8ed8e68808c3fbc4ea764fc9d2968646
+MD5_PPM_420M_ISLOW_5_8 = a3363274999da2366a024efae6d16c9b
+MD5_PPM_420M_ISLOW_1_2 = e692a315cea26b988c8e8b29a5dbcd81
+MD5_PPM_420M_ISLOW_3_8 = 79eca9175652ced755155c90e785a996
+MD5_PPM_420M_ISLOW_1_4 = 79cd778f8bf1a117690052cacdd54eca
+MD5_PPM_420M_ISLOW_1_8 = 391b3d4aca640c8567d6f8745eb2142f
+MD5_BMP_420_ISLOW_256 = 4980185e3776e89bd931736e1cddeee6
+MD5_BMP_420_ISLOW_565 = bf9d13e16c4923b92e1faa604d7922cb
+MD5_BMP_420_ISLOW_565D = 6bde71526acc44bcff76f696df8638d2
+MD5_BMP_420M_ISLOW_565 = 8dc0185245353cfa32ad97027342216f
+MD5_BMP_420M_ISLOW_565D =d1be3a3339166255e76fa50a0d70d73e
+MD5_PPM_420_ISLOW_SKIP15_31 = c4c65c1e43d7275cd50328a61e6534f0
+MD5_PPM_420_ISLOW_ARI_SKIP16_139 = 087c6b123db16ac00cb88c5b590bb74a
+MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71 = 26eb36ccc7d1f0cb80cdabb0ac8b5d99
+MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4 = 886c6775af22370257122f8b16207e6d
+MD5_PPM_444_ISLOW_SKIP1_6 = 5606f86874cf26b8fcee1117a0a436a6
+MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13 = db87dc7ce26bcdc7a6b56239ce2b9d6c
+MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0 = cb57b32bd6d03e35432362f7bf184b6d
+MD5_JPEG_CROP = b4197f377e621c4e9b1d20471432610d
+
+endif
+
+.PHONY: test
+test: tjquicktest tjbittest bittest
+
+if CROSS_COMPILING
+tjquicktest: testclean
+else
+tjquicktest: testclean all
+endif
+
+if WITH_TURBOJPEG
+if WITH_JAVA
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -bi
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -noyuvpad
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -bi
+	$(JAVA) -cp java/turbojpeg.jar -Djava.library.path=.libs TJUnitTest -yuv -bi -noyuvpad
+endif
+	./tjunittest
+	./tjunittest -alloc
+	./tjunittest -yuv
+	./tjunittest -yuv -alloc
+	./tjunittest -yuv -noyuvpad
+endif
+	echo GREAT SUCCESS!
+
+if CROSS_COMPILING
+tjbittest: testclean
+else
+tjbittest: testclean all
+endif
+
+if WITH_TURBOJPEG
+
+MD5_PPM_GRAY_TILE = 89d3ca21213d9d864b50b4e4e7de4ca6
+MD5_PPM_420_8x8_TILE = 847fceab15c5b7b911cb986cf0f71de3
+MD5_PPM_420_16x16_TILE = ca45552a93687e078f7137cc4126a7b0
+MD5_PPM_420_32x32_TILE = d8676f1d6b68df358353bba9844f4a00
+MD5_PPM_420_64x64_TILE = 4e4c1a3d7ea4bace4f868bcbe83b7050
+MD5_PPM_420_128x128_TILE = f24c3429c52265832beab9df72a0ceae
+MD5_PPM_420M_8x8_TILE = bc25320e1f4c31ce2e610e43e9fd173c
+MD5_PPM_420M_TILE = 75ffdf14602258c5c189522af57fa605
+MD5_PPM_422_8x8_TILE = d83dacd9fc73b0a6f10c09acad64eb1e
+MD5_PPM_422_16x16_TILE = 35077fb610d72dd743b1eb0cbcfe10fb
+MD5_PPM_422_32x32_TILE = e6902ed8a449ecc0f0d6f2bf945f65f7
+MD5_PPM_422_64x64_TILE = 2b4502a8f316cedbde1da7bce3d2231e
+MD5_PPM_422_128x128_TILE = f0b5617d578f5e13c8eee215d64d4877
+MD5_PPM_422M_8x8_TILE = 828941d7f41cd6283abd6beffb7fd51d
+MD5_PPM_422M_TILE = e877ae1324c4a280b95376f7f018172f
+MD5_PPM_444_TILE = 7964e41e67cfb8d0a587c0aa4798f9c3
+
+# Test compressing from/decompressing to an arbitrary subregion of a larger
+# image buffer
+	cp $(srcdir)/testimages/testorig.ppm testout_tile.ppm
+	./tjbench testout_tile.ppm 95 -rgb -quiet -tile -benchtime 0.01 >/dev/null 2>&1
+	for i in 8 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_GRAY_TILE) testout_tile_GRAY_Q95_$$i\x$$i.ppm; \
+	done
+	md5/md5cmp $(MD5_PPM_420_8x8_TILE) testout_tile_420_Q95_8x8.ppm
+	md5/md5cmp $(MD5_PPM_420_16x16_TILE) testout_tile_420_Q95_16x16.ppm
+	md5/md5cmp $(MD5_PPM_420_32x32_TILE) testout_tile_420_Q95_32x32.ppm
+	md5/md5cmp $(MD5_PPM_420_64x64_TILE) testout_tile_420_Q95_64x64.ppm
+	md5/md5cmp $(MD5_PPM_420_128x128_TILE) testout_tile_420_Q95_128x128.ppm
+	md5/md5cmp $(MD5_PPM_422_8x8_TILE) testout_tile_422_Q95_8x8.ppm
+	md5/md5cmp $(MD5_PPM_422_16x16_TILE) testout_tile_422_Q95_16x16.ppm
+	md5/md5cmp $(MD5_PPM_422_32x32_TILE) testout_tile_422_Q95_32x32.ppm
+	md5/md5cmp $(MD5_PPM_422_64x64_TILE) testout_tile_422_Q95_64x64.ppm
+	md5/md5cmp $(MD5_PPM_422_128x128_TILE) testout_tile_422_Q95_128x128.ppm
+	for i in 8 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_444_TILE) testout_tile_444_Q95_$$i\x$$i.ppm; \
+	done
+	rm -f testout_tile_GRAY_* testout_tile_420_* testout_tile_422_* testout_tile_444_*
+
+	./tjbench testout_tile.ppm 95 -rgb -fastupsample -quiet -tile -benchtime 0.01 >/dev/null 2>&1
+	md5/md5cmp $(MD5_PPM_420M_8x8_TILE) testout_tile_420_Q95_8x8.ppm
+	for i in 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_420M_TILE) testout_tile_420_Q95_$$i\x$$i.ppm; \
+	done
+	md5/md5cmp $(MD5_PPM_422M_8x8_TILE) testout_tile_422_Q95_8x8.ppm
+	for i in 16 32 64 128; do \
+		md5/md5cmp $(MD5_PPM_422M_TILE) testout_tile_422_Q95_$$i\x$$i.ppm; \
+	done
+	rm -f testout_tile_GRAY_* testout_tile_420_* testout_tile_422_* testout_tile_444_* testout_tile.ppm
+	echo GREAT SUCCESS!
+
+endif
+
+if CROSS_COMPILING
+bittest: testclean
+else
+bittest: testclean all
+endif
+
+# These tests are carefully crafted to provide full coverage of as many of the
+# underlying algorithms as possible (including all of the SIMD-accelerated
+# ones.)
+
+# CC: null  SAMP: fullsize  FDCT: islow  ENT: huff
+	./cjpeg -rgb -dct int -outfile testout_rgb_islow.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_RGB_ISLOW) testout_rgb_islow.jpg
+# CC: null  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -ppm -outfile testout_rgb_islow.ppm testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_PPM_RGB_ISLOW) testout_rgb_islow.ppm
+	rm -f testout_rgb_islow.ppm
+if WITH_12BIT
+	rm -f testout_rgb_islow.jpg
+else
+# CC: RGB->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_rgb_islow_565.bmp testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_BMP_RGB_ISLOW_565) testout_rgb_islow_565.bmp
+	rm -f testout_rgb_islow_565.bmp
+# CC: RGB->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_rgb_islow_565D.bmp testout_rgb_islow.jpg
+	md5/md5cmp $(MD5_BMP_RGB_ISLOW_565D) testout_rgb_islow_565D.bmp
+	rm -f testout_rgb_islow_565D.bmp testout_rgb_islow.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize/h2v1  FDCT: ifast  ENT: 2-pass huff
+	./cjpeg -sample 2x1 -dct fast -opt -outfile testout_422_ifast_opt.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_422_IFAST_OPT) testout_422_ifast_opt.jpg
+# CC: YCC->RGB  SAMP: fullsize/h2v1 fancy  IDCT: ifast  ENT: huff
+	./djpeg -dct fast -outfile testout_422_ifast.ppm testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_PPM_422_IFAST) testout_422_ifast.ppm
+	rm -f testout_422_ifast.ppm
+# CC: YCC->RGB  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct fast -nosmooth -outfile testout_422m_ifast.ppm testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_PPM_422M_IFAST) testout_422m_ifast.ppm
+	rm -f testout_422m_ifast.ppm
+if WITH_12BIT
+	rm -f testout_422_ifast_opt.jpg
+else
+# CC: YCC->RGB565  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -dither none -bmp -outfile testout_422m_ifast_565.bmp testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_BMP_422M_IFAST_565) testout_422m_ifast_565.bmp
+	rm -f testout_422m_ifast_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v1 merged  IDCT: ifast  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -bmp -outfile testout_422m_ifast_565D.bmp testout_422_ifast_opt.jpg
+	md5/md5cmp $(MD5_BMP_422M_IFAST_565D) testout_422m_ifast_565D.bmp
+	rm -f testout_422m_ifast_565D.bmp testout_422_ifast_opt.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize/h2v2  FDCT: ifast  ENT: prog huff
+	./cjpeg -sample 2x2 -quality 100 -dct fast -prog -outfile testout_420_q100_ifast_prog.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420_IFAST_Q100_PROG) testout_420_q100_ifast_prog.jpg
+# CC: YCC->RGB  SAMP: fullsize/h2v2 fancy  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -outfile testout_420_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_420_Q100_IFAST) testout_420_q100_ifast.ppm
+	rm -f testout_420_q100_ifast.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -nosmooth -outfile testout_420m_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_420M_Q100_IFAST) testout_420m_q100_ifast.ppm
+	rm -f testout_420m_q100_ifast.ppm testout_420_q100_ifast_prog.jpg
+
+# CC: RGB->Gray  SAMP: fullsize  FDCT: islow  ENT: huff
+	./cjpeg -gray -dct int -outfile testout_gray_islow.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_GRAY_ISLOW) testout_gray_islow.jpg
+# CC: Gray->Gray  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -outfile testout_gray_islow.ppm testout_gray_islow.jpg
+	md5/md5cmp $(MD5_PPM_GRAY_ISLOW) testout_gray_islow.ppm
+	rm -f testout_gray_islow.ppm
+# CC: Gray->RGB  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb -outfile testout_gray_islow_rgb.ppm testout_gray_islow.jpg
+	md5/md5cmp $(MD5_PPM_GRAY_ISLOW_RGB) testout_gray_islow_rgb.ppm
+	rm -f testout_gray_islow_rgb.ppm
+if WITH_12BIT
+	rm -f testout_gray_islow.jpg
+else
+# CC: Gray->RGB565  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_gray_islow_565.bmp testout_gray_islow.jpg
+	md5/md5cmp $(MD5_BMP_GRAY_ISLOW_565) testout_gray_islow_565.bmp
+	rm -f testout_gray_islow_565.bmp
+# CC: Gray->RGB565 (dithered)  SAMP: fullsize  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_gray_islow_565D.bmp testout_gray_islow.jpg
+	md5/md5cmp $(MD5_BMP_GRAY_ISLOW_565D) testout_gray_islow_565D.bmp
+	rm -f testout_gray_islow_565D.bmp testout_gray_islow.jpg
+endif
+
+# CC: RGB->YCC  SAMP: fullsize smooth/h2v2 smooth  FDCT: islow
+# ENT: 2-pass huff
+	./cjpeg -sample 2x2 -smooth 1 -dct int -opt -outfile testout_420s_ifast_opt.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420S_IFAST_OPT) testout_420s_ifast_opt.jpg
+	rm -f testout_420s_ifast_opt.jpg
+
+# The output of the floating point tests is not validated by default, because
+# the output differs depending on the type of floating point math used, and
+# this is only deterministic if the DCT/IDCT are implemented using SIMD
+# instructions on a particular platform.  Pass one of the following on the make
+# command line to validate the floating point tests against one of the expected
+# results:
+#
+# FLOATTEST=sse  validate against the expected results from the libjpeg-turbo
+#                SSE SIMD extensions
+# FLOATTEST=32bit  validate against the expected results from the C code
+#                  when running on a 32-bit FPU (or when SSE is being used for
+#                  floating point math, which is generally the default with
+#                  x86-64 compilers)
+# FLOATTEST=64bit  validate against the exepected results from the C code
+#                  when running on a 64-bit FPU
+
+# CC: RGB->YCC  SAMP: fullsize/int  FDCT: float  ENT: prog huff
+	./cjpeg -sample 3x2 -dct float -prog -outfile testout_3x2_float_prog.jpg $(srcdir)/testimages/testorig.ppm
+	if [ "${FLOATTEST}" = "sse" ]; then \
+		md5/md5cmp $(MD5_JPEG_3x2_FLOAT_PROG_SSE) testout_3x2_float_prog.jpg; \
+	elif [ "${FLOATTEST}" = "32bit" -o "${FLOATTEST}" = "64bit" ]; then \
+		md5/md5cmp $(MD5_JPEG_3x2_FLOAT_PROG_32BIT) testout_3x2_float_prog.jpg; \
+	fi
+# CC: YCC->RGB  SAMP: fullsize/int  IDCT: float  ENT: prog huff
+	./djpeg -dct float -outfile testout_3x2_float.ppm testout_3x2_float_prog.jpg
+	if [ "${FLOATTEST}" = "sse" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_SSE) testout_3x2_float.ppm; \
+	elif [ "${FLOATTEST}" = "32bit" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_32BIT) testout_3x2_float.ppm; \
+	elif [ "${FLOATTEST}" = "64bit" ]; then \
+		md5/md5cmp $(MD5_PPM_3x2_FLOAT_64BIT) testout_3x2_float.ppm; \
+	fi
+	rm -f testout_3x2_float.ppm testout_3x2_float_prog.jpg
+
+# CC: RGB->YCC  SAMP: fullsize/int  FDCT: ifast  ENT: prog huff
+	./cjpeg -sample 3x2 -dct fast -prog -outfile testout_3x2_ifast_prog.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_3x2_IFAST_PROG) testout_3x2_ifast_prog.jpg
+# CC: YCC->RGB  SAMP: fullsize/int  IDCT: ifast  ENT: prog huff
+	./djpeg -dct fast -outfile testout_3x2_ifast.ppm testout_3x2_ifast_prog.jpg
+	md5/md5cmp $(MD5_PPM_3x2_IFAST) testout_3x2_ifast.ppm
+	rm -f testout_3x2_ifast.ppm testout_3x2_ifast_prog.jpg
+
+if WITH_ARITH_ENC
+# CC: YCC->RGB  SAMP: fullsize/h2v2  FDCT: islow  ENT: arith
+	./cjpeg -dct int -arithmetic -outfile testout_420_islow_ari.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_420_ISLOW_ARI) testout_420_islow_ari.jpg
+	rm -f testout_420_islow_ari.jpg
+	./jpegtran -arithmetic -outfile testout_420_islow_ari.jpg $(srcdir)/testimages/testimgint.jpg
+	md5/md5cmp $(MD5_JPEG_420_ISLOW_ARI) testout_420_islow_ari.jpg
+	rm -f testout_420_islow_ari.jpg
+# CC: YCC->RGB  SAMP: fullsize  FDCT: islow  ENT: prog arith
+	./cjpeg -sample 1x1 -dct int -prog -arithmetic -outfile testout_444_islow_progari.jpg $(srcdir)/testimages/testorig.ppm
+	md5/md5cmp $(MD5_JPEG_444_ISLOW_PROGARI) testout_444_islow_progari.jpg
+	rm -f testout_444_islow_progari.jpg
+endif
+if WITH_ARITH_DEC
+# CC: RGB->YCC  SAMP: h2v2 merged  IDCT: ifast  ENT: arith
+	./djpeg -fast -ppm -outfile testout_420m_ifast_ari.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420M_IFAST_ARI) testout_420m_ifast_ari.ppm
+	rm -f testout_420m_ifast_ari.ppm
+	./jpegtran -outfile testout_420_islow.jpg $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_JPEG_420_ISLOW) testout_420_islow.jpg
+	rm -f testout_420_islow.jpg
+endif
+
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 16x16 islow  ENT: huff
+	./djpeg -dct int -scale 2/1 -nosmooth -ppm -outfile testout_420m_islow_2_1.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_2_1) testout_420m_islow_2_1.ppm
+	rm -f testout_420m_islow_2_1.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 15x15 islow  ENT: huff
+	./djpeg -dct int -scale 15/8 -nosmooth -ppm -outfile testout_420m_islow_15_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_15_8) testout_420m_islow_15_8.ppm
+	rm -f testout_420m_islow_15_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 13x13 islow  ENT: huff
+	./djpeg -dct int -scale 13/8 -nosmooth -ppm -outfile testout_420m_islow_13_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_13_8) testout_420m_islow_13_8.ppm
+	rm -f testout_420m_islow_13_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 11x11 islow  ENT: huff
+	./djpeg -dct int -scale 11/8 -nosmooth -ppm -outfile testout_420m_islow_11_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_11_8) testout_420m_islow_11_8.ppm
+	rm -f testout_420m_islow_11_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 9x9 islow  ENT: huff
+	./djpeg -dct int -scale 9/8 -nosmooth -ppm -outfile testout_420m_islow_9_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_9_8) testout_420m_islow_9_8.ppm
+	rm -f testout_420m_islow_9_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 7x7 islow/14x14 islow  ENT: huff
+	./djpeg -dct int -scale 7/8 -nosmooth -ppm -outfile testout_420m_islow_7_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_7_8) testout_420m_islow_7_8.ppm
+	rm -f testout_420m_islow_7_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 6x6 islow/12x12 islow  ENT: huff
+	./djpeg -dct int -scale 3/4 -nosmooth -ppm -outfile testout_420m_islow_3_4.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_3_4) testout_420m_islow_3_4.ppm
+	rm -f testout_420m_islow_3_4.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 5x5 islow/10x10 islow  ENT: huff
+	./djpeg -dct int -scale 5/8 -nosmooth -ppm -outfile testout_420m_islow_5_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_5_8) testout_420m_islow_5_8.ppm
+	rm -f testout_420m_islow_5_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 4x4 islow/8x8 islow  ENT: huff
+	./djpeg -dct int -scale 1/2 -nosmooth -ppm -outfile testout_420m_islow_1_2.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_2) testout_420m_islow_1_2.ppm
+	rm -f testout_420m_islow_1_2.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 3x3 islow/6x6 islow  ENT: huff
+	./djpeg -dct int -scale 3/8 -nosmooth -ppm -outfile testout_420m_islow_3_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_3_8) testout_420m_islow_3_8.ppm
+	rm -f testout_420m_islow_3_8.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 2x2 islow/4x4 islow  ENT: huff
+	./djpeg -dct int -scale 1/4 -nosmooth -ppm -outfile testout_420m_islow_1_4.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_4) testout_420m_islow_1_4.ppm
+	rm -f testout_420m_islow_1_4.ppm
+# CC: YCC->RGB  SAMP: h2v2 merged  IDCT: 1x1 islow/2x2 islow  ENT: huff
+	./djpeg -dct int -scale 1/8 -nosmooth -ppm -outfile testout_420m_islow_1_8.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420M_ISLOW_1_8) testout_420m_islow_1_8.ppm
+	rm -f testout_420m_islow_1_8.ppm
+if WITH_12BIT
+else
+# CC: YCC->RGB (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -colors 256 -bmp -outfile testout_420_islow_256.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_256) testout_420_islow_256.bmp
+	rm -f testout_420_islow_256.bmp
+# CC: YCC->RGB565  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -dither none -bmp -outfile testout_420_islow_565.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_565) testout_420_islow_565.bmp
+	rm -f testout_420_islow_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v2 fancy  IDCT: islow  ENT: huff
+	./djpeg -dct int -rgb565 -bmp -outfile testout_420_islow_565D.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420_ISLOW_565D) testout_420_islow_565D.bmp
+	rm -f testout_420_islow_565D.bmp
+# CC: YCC->RGB565  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -dither none -bmp -outfile testout_420m_islow_565.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420M_ISLOW_565) testout_420m_islow_565.bmp
+	rm -f testout_420m_islow_565.bmp
+# CC: YCC->RGB565 (dithered)  SAMP: h2v2 merged  IDCT: islow  ENT: huff
+	./djpeg -dct int -nosmooth -rgb565 -bmp -outfile testout_420m_islow_565D.bmp $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_BMP_420M_ISLOW_565D) testout_420m_islow_565D.bmp
+	rm -f testout_420m_islow_565D.bmp
+endif
+
+# Partial decode tests.  These tests are designed to cover all of the possible
+# code paths in jpeg_skip_scanlines().
+
+# Context rows: Yes  Intra-iMCU row: Yes  iMCU row prefetch: No   ENT: huff
+	./djpeg -dct int -skip 15,31 -ppm -outfile testout_420_islow_skip15,31.ppm $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_PPM_420_ISLOW_SKIP15_31) testout_420_islow_skip15,31.ppm
+	rm -f testout_420_islow_skip15,31.ppm
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: Yes  ENT: arith
+if WITH_ARITH_DEC
+	./djpeg -dct int -skip 16,139 -ppm -outfile testout_420_islow_ari_skip16,139.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_ARI_SKIP16_139) testout_420_islow_ari_skip16,139.ppm
+	rm -f testout_420_islow_ari_skip16,139.ppm
+endif
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: prog huff
+	./cjpeg -dct int -prog -outfile testout_420_islow_prog.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -crop 62x62+71+71 -ppm -outfile testout_420_islow_prog_crop62x62,71,71.ppm testout_420_islow_prog.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_PROG_CROP62x62_71_71) testout_420_islow_prog_crop62x62,71,71.ppm
+	rm -f testout_420_islow_prog_crop62x62,71,71.ppm testout_420_islow_prog.jpg
+# Context rows: Yes  Intra-iMCU row: No   iMCU row prefetch: No   ENT: arith
+if WITH_ARITH_DEC
+	./djpeg -dct int -crop 53x53+4+4 -ppm -outfile testout_420_islow_ari_crop53x53,4,4.ppm $(srcdir)/testimages/testimgari.jpg
+	md5/md5cmp $(MD5_PPM_420_ISLOW_ARI_CROP53x53_4_4) testout_420_islow_ari_crop53x53,4,4.ppm
+	rm -f testout_420_islow_ari_crop53x53,4,4.ppm
+endif
+# Context rows: No   Intra-iMCU row: Yes  ENT: huff
+	./cjpeg -dct int -sample 1x1 -outfile testout_444_islow.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -skip 1,6 -ppm -outfile testout_444_islow_skip1,6.ppm testout_444_islow.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_SKIP1_6) testout_444_islow_skip1,6.ppm
+	rm -f testout_444_islow_skip1,6.ppm testout_444_islow.jpg
+# Context rows: No   Intra-iMCU row: No   ENT: prog huff
+	./cjpeg -dct int -prog -sample 1x1 -outfile testout_444_islow_prog.jpg $(srcdir)/testimages/testorig.ppm
+	./djpeg -dct int -crop 98x98+13+13 -ppm -outfile testout_444_islow_prog_crop98x98,13,13.ppm testout_444_islow_prog.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_PROG_CROP98x98_13_13) testout_444_islow_prog_crop98x98,13,13.ppm
+	rm -f testout_444_islow_prog_crop98x98,13,13.ppm testout_444_islow_prog.jpg
+# Context rows: No   Intra-iMCU row: No   ENT: arith
+if WITH_ARITH_ENC
+	./cjpeg -dct int -arithmetic -sample 1x1 -outfile testout_444_islow_ari.jpg $(srcdir)/testimages/testorig.ppm
+if WITH_ARITH_DEC
+	./djpeg -dct int -crop 37x37+0+0 -ppm -outfile testout_444_islow_ari_crop37x37,0,0.ppm testout_444_islow_ari.jpg
+	md5/md5cmp $(MD5_PPM_444_ISLOW_ARI_CROP37x37_0_0) testout_444_islow_ari_crop37x37,0,0.ppm
+	rm -f testout_444_islow_ari_crop37x37,0,0.ppm
+endif
+	rm -f testout_444_islow_ari.jpg
+endif
+
+	./jpegtran -crop 120x90+20+50 -transpose -perfect -outfile testout_crop.jpg $(srcdir)/testimages/$(TESTORIG)
+	md5/md5cmp $(MD5_JPEG_CROP) testout_crop.jpg
+	rm -f testout_crop.jpg
+	echo GREAT SUCCESS!
+
+
+testclean:
+	rm -f testout*
+	rm -f *_GRAY_*.bmp
+	rm -f *_GRAY_*.png
+	rm -f *_GRAY_*.ppm
+	rm -f *_GRAY_*.jpg
+	rm -f *_GRAY.yuv
+	rm -f *_420_*.bmp
+	rm -f *_420_*.png
+	rm -f *_420_*.ppm
+	rm -f *_420_*.jpg
+	rm -f *_420.yuv
+	rm -f *_422_*.bmp
+	rm -f *_422_*.png
+	rm -f *_422_*.ppm
+	rm -f *_422_*.jpg
+	rm -f *_422.yuv
+	rm -f *_444_*.bmp
+	rm -f *_444_*.png
+	rm -f *_444_*.ppm
+	rm -f *_444_*.jpg
+	rm -f *_444.yuv
+	rm -f *_440_*.bmp
+	rm -f *_440_*.png
+	rm -f *_440_*.ppm
+	rm -f *_440_*.jpg
+	rm -f *_440.yuv
+	rm -f *_411_*.bmp
+	rm -f *_411_*.png
+	rm -f *_411_*.ppm
+	rm -f *_411_*.jpg
+	rm -f *_411.yuv
+
+
+tjtest:
+	sh ./tjbenchtest
+	sh ./tjbenchtest -alloc
+	sh ./tjbenchtest -yuv
+	sh ./tjbenchtest -yuv -alloc
+if WITH_JAVA
+	sh ./tjbenchtest.java
+	sh ./tjbenchtest.java -yuv
+endif
+
+
+pkgscripts/libjpeg-turbo.spec: pkgscripts/libjpeg-turbo.spec.tmpl
+	cat pkgscripts/libjpeg-turbo.spec.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__datadir}@$(datadir)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__includedir}@$(includedir)@g | \
+		sed s@%{__libdir}@$(libdir)@g | sed s@%{__mandir}@$(mandir)@g \
+		> pkgscripts/libjpeg-turbo.spec
+
+rpm: all pkgscripts/libjpeg-turbo.spec
+	TMPDIR=`mktemp -d /tmp/${PACKAGE_NAME}-build.XXXXXX`; \
+	mkdir -p $$TMPDIR/RPMS; \
+	ln -fs `pwd` $$TMPDIR/BUILD; \
+	rm -f ${PKGNAME}-${VERSION}.${RPMARCH}.rpm; \
+	rpmbuild -bb --define "_blddir $$TMPDIR/buildroot"  \
+		--define "_topdir $$TMPDIR" \
+		--target ${RPMARCH} pkgscripts/libjpeg-turbo.spec; \
+	cp $$TMPDIR/RPMS/${RPMARCH}/${PKGNAME}-${VERSION}-${BUILD}.${RPMARCH}.rpm \
+		${PKGNAME}-${VERSION}.${RPMARCH}.rpm; \
+	rm -rf $$TMPDIR
+
+srpm: dist-gzip pkgscripts/libjpeg-turbo.spec
+	TMPDIR=`mktemp -d /tmp/${PACKAGE_NAME}-build.XXXXXX`; \
+	mkdir -p $$TMPDIR/RPMS; \
+	mkdir -p $$TMPDIR/SRPMS; \
+	mkdir -p $$TMPDIR/BUILD; \
+	mkdir -p $$TMPDIR/SOURCES; \
+	mkdir -p $$TMPDIR/SPECS; \
+	rm -f ${PKGNAME}-${VERSION}.src.rpm; \
+	cp ${PACKAGE_NAME}-${VERSION}.tar.gz $$TMPDIR/SOURCES; \
+	cat pkgscripts/libjpeg-turbo.spec | sed s/%{_blddir}/%{_tmppath}/g \
+		| sed s/#--\>//g \
+		> $$TMPDIR/SPECS/libjpeg-turbo.spec; \
+	rpmbuild -bs --define "_topdir $$TMPDIR" $$TMPDIR/SPECS/libjpeg-turbo.spec; \
+	cp $$TMPDIR/SRPMS/${PKGNAME}-${VERSION}-${BUILD}.src.rpm \
+		${PKGNAME}-${VERSION}.src.rpm; \
+	rm -rf $$TMPDIR
+
+pkgscripts/makedpkg: pkgscripts/makedpkg.tmpl
+	cat pkgscripts/makedpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__libdir}@$(libdir)@g \
+		> pkgscripts/makedpkg
+
+deb: all pkgscripts/makedpkg
+	sh pkgscripts/makedpkg
+
+pkgscripts/uninstall: pkgscripts/uninstall.tmpl
+	cat pkgscripts/uninstall.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__datadir}@$(datadir)@g | \
+		sed s@%{__includedir}@$(includedir)@g | sed s@%{__libdir}@$(libdir)@g | \
+		sed s@%{__mandir}@$(mandir)@g > pkgscripts/uninstall
+
+pkgscripts/makemacpkg: pkgscripts/makemacpkg.tmpl
+	cat pkgscripts/makemacpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__bindir}@$(bindir)@g | sed s@%{__docdir}@$(docdir)@g | \
+		sed s@%{__libdir}@$(libdir)@g > pkgscripts/makemacpkg
+
+if X86_64
+
+udmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -build32 ${BUILDDIR32}
+
+iosdmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -build32 ${BUILDDIR32} -buildarmv6 ${BUILDDIRARMV6} -buildarmv7 ${BUILDDIRARMV7} -buildarmv7s ${BUILDDIRARMV7S} -buildarmv8 ${BUILDDIRARMV8} -lipo "${LIPO}"
+
+else
+
+iosdmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg -buildarmv6 ${BUILDDIRARMV6} -buildarmv7 ${BUILDDIRARMV7} -buildarmv7s ${BUILDDIRARMV7S} -buildarmv8 ${BUILDDIRARMV8} -lipo "${LIPO}"
+
+endif
+
+dmg: all pkgscripts/makemacpkg pkgscripts/uninstall
+	sh pkgscripts/makemacpkg
+
+pkgscripts/makecygwinpkg: pkgscripts/makecygwinpkg.tmpl
+	cat pkgscripts/makecygwinpkg.tmpl | sed s@%{__prefix}@$(prefix)@g | \
+		sed s@%{__docdir}@$(docdir)@g | sed s@%{__libdir}@$(libdir)@g \
+		> pkgscripts/makecygwinpkg
+
+cygwinpkg: all pkgscripts/makecygwinpkg
+	sh pkgscripts/makecygwinpkg
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdbmp.c glmark2-2021.02/src/libjpeg-turbo/rdbmp.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdbmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdbmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,483 @@
+/*
+ * rdbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * Modified 2009-2010 by Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Modified 2011 by Siarhei Siamashka.
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Microsoft "BMP"
+ * format (MS Windows 3.x, OS/2 1.x, and OS/2 2.x flavors).
+ * Currently, only 8-bit and 24-bit images are supported, not 1-bit or
+ * 4-bit (feeding such low-depth images into JPEG would be silly anyway).
+ * Also, we don't support RLE-compressed files.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed BMP format).
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef BMP_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _bmp_source_struct *bmp_source_ptr;
+
+typedef struct _bmp_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* BMP colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed to reverse row order */
+  JDIMENSION source_row;        /* Current source row number */
+  JDIMENSION row_width;         /* Physical width of scanlines in file */
+
+  int bits_per_pixel;           /* remembers 8- or 24-bit format */
+} bmp_source_struct;
+
+
+LOCAL(int)
+read_byte (bmp_source_ptr sinfo)
+/* Read next byte from BMP file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (bmp_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a BMP file */
+{
+  int i;
+
+  switch (mapentrysize) {
+  case 3:
+    /* BGR format (occurs in OS/2 files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+    }
+    break;
+  case 4:
+    /* BGR0 format (occurs in MS Windows files) */
+    for (i = 0; i < cmaplen; i++) {
+      sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+      sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+      (void) read_byte(sinfo);
+    }
+    break;
+  default:
+    ERREXIT(sinfo->cinfo, JERR_BMP_BADCMAP);
+    break;
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ * The image has been read into the whole_image array, but is otherwise
+ * unprocessed.  We must read it out in top-to-bottom row order, and if
+ * it is an 8-bit image, we must expand colormapped pixels to 24bit format.
+ */
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register JSAMPARRAY colormap = source->colormap;
+  JSAMPARRAY image_ptr;
+  register int t;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Expand the colormap indexes to real data */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    t = GETJSAMPLE(*inptr++);
+    *outptr++ = colormap[0][t]; /* can omit GETJSAMPLE() safely */
+    *outptr++ = colormap[1][t];
+    *outptr++ = colormap[2][t];
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    outptr += 3;
+  }
+
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_32bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 32-bit pixels */
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+
+  /* Fetch next row from virtual array */
+  source->source_row--;
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source->source_row, (JDIMENSION) 1, FALSE);
+  /* Transfer data.  Note source values are in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = image_ptr[0];
+  outptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+    outptr[1] = *inptr++;
+    outptr[0] = *inptr++;
+    inptr++;                    /* skip the 4th byte (Alpha channel) */
+    outptr += 3;
+  }
+
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_8bit_row, get_24bit_row, or get_32bit_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  register FILE *infile = source->pub.input_file;
+  register JSAMPROW out_ptr;
+  JSAMPARRAY image_ptr;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image,
+       row, (JDIMENSION) 1, TRUE);
+    out_ptr = image_ptr[0];
+    if (fread(out_ptr, 1, source->row_width, infile) != source->row_width) {
+      if (feof(infile))
+        ERREXIT(cinfo, JERR_INPUT_EOF);
+      else
+        ERREXIT(cinfo, JERR_FILE_READ);
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in top-to-bottom order */
+  switch (source->bits_per_pixel) {
+  case 8:
+    source->pub.get_pixel_rows = get_8bit_row;
+    break;
+  case 24:
+    source->pub.get_pixel_rows = get_24bit_row;
+    break;
+  case 32:
+    source->pub.get_pixel_rows = get_32bit_row;
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+  }
+  source->source_row = cinfo->image_height;
+
+  /* And read the first row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  bmp_source_ptr source = (bmp_source_ptr) sinfo;
+  U_CHAR bmpfileheader[14];
+  U_CHAR bmpinfoheader[64];
+#define GET_2B(array,offset)  ((unsigned short) UCH(array[offset]) + \
+                               (((unsigned short) UCH(array[offset+1])) << 8))
+#define GET_4B(array,offset)  ((unsigned int) UCH(array[offset]) + \
+                               (((unsigned int) UCH(array[offset+1])) << 8) + \
+                               (((unsigned int) UCH(array[offset+2])) << 16) + \
+                               (((unsigned int) UCH(array[offset+3])) << 24))
+  unsigned int bfOffBits;
+  unsigned int headerSize;
+  int biWidth;
+  int biHeight;
+  unsigned short biPlanes;
+  unsigned int biCompression;
+  int biXPelsPerMeter,biYPelsPerMeter;
+  unsigned int biClrUsed = 0;
+  int mapentrysize = 0;         /* 0 indicates no colormap */
+  int bPad;
+  JDIMENSION row_width;
+
+  /* Read and verify the bitmap file header */
+  if (! ReadOK(source->pub.input_file, bmpfileheader, 14))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  if (GET_2B(bmpfileheader,0) != 0x4D42) /* 'BM' */
+    ERREXIT(cinfo, JERR_BMP_NOT);
+  bfOffBits = GET_4B(bmpfileheader,10);
+  /* We ignore the remaining fileheader fields */
+
+  /* The infoheader might be 12 bytes (OS/2 1.x), 40 bytes (Windows),
+   * or 64 bytes (OS/2 2.x).  Check the first 4 bytes to find out which.
+   */
+  if (! ReadOK(source->pub.input_file, bmpinfoheader, 4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  headerSize = GET_4B(bmpinfoheader,0);
+  if (headerSize < 12 || headerSize > 64)
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  if (! ReadOK(source->pub.input_file, bmpinfoheader+4, headerSize-4))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  switch (headerSize) {
+  case 12:
+    /* Decode OS/2 1.x header (Microsoft calls this a BITMAPCOREHEADER) */
+    biWidth = (int) GET_2B(bmpinfoheader,4);
+    biHeight = (int) GET_2B(bmpinfoheader,6);
+    biPlanes = GET_2B(bmpinfoheader,8);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,10);
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 3;         /* OS/2 uses RGBTRIPLE colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP_OS2, biWidth, biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    break;
+  case 40:
+  case 64:
+    /* Decode Windows 3.x header (Microsoft calls this a BITMAPINFOHEADER) */
+    /* or OS/2 2.x header, which has additional fields that we ignore */
+    biWidth = (int) GET_4B(bmpinfoheader,4);
+    biHeight = (int) GET_4B(bmpinfoheader,8);
+    biPlanes = GET_2B(bmpinfoheader,12);
+    source->bits_per_pixel = (int) GET_2B(bmpinfoheader,14);
+    biCompression = GET_4B(bmpinfoheader,16);
+    biXPelsPerMeter = (int) GET_4B(bmpinfoheader,24);
+    biYPelsPerMeter = (int) GET_4B(bmpinfoheader,28);
+    biClrUsed = GET_4B(bmpinfoheader,32);
+    /* biSizeImage, biClrImportant fields are ignored */
+
+    switch (source->bits_per_pixel) {
+    case 8:                     /* colormapped image */
+      mapentrysize = 4;         /* Windows uses RGBQUAD colormap */
+      TRACEMS2(cinfo, 1, JTRC_BMP_MAPPED, biWidth, biHeight);
+      break;
+    case 24:                    /* RGB image */
+      TRACEMS2(cinfo, 1, JTRC_BMP, biWidth, biHeight);
+      break;
+    case 32:                    /* RGB image + Alpha channel */
+      TRACEMS2(cinfo, 1, JTRC_BMP, biWidth, biHeight);
+      break;
+    default:
+      ERREXIT(cinfo, JERR_BMP_BADDEPTH);
+      break;
+    }
+    if (biCompression != 0)
+      ERREXIT(cinfo, JERR_BMP_COMPRESSED);
+
+    if (biXPelsPerMeter > 0 && biYPelsPerMeter > 0) {
+      /* Set JFIF density parameters from the BMP data */
+      cinfo->X_density = (UINT16) (biXPelsPerMeter/100); /* 100 cm per meter */
+      cinfo->Y_density = (UINT16) (biYPelsPerMeter/100);
+      cinfo->density_unit = 2;  /* dots/cm */
+    }
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+    return;
+  }
+
+  if (biWidth <= 0 || biHeight <= 0)
+    ERREXIT(cinfo, JERR_BMP_EMPTY);
+  if (biPlanes != 1)
+    ERREXIT(cinfo, JERR_BMP_BADPLANES);
+
+  /* Compute distance to bitmap data --- will adjust for colormap below */
+  bPad = bfOffBits - (headerSize + 14);
+
+  /* Read the colormap, if any */
+  if (mapentrysize > 0) {
+    if (biClrUsed <= 0)
+      biClrUsed = 256;          /* assume it's 256 */
+    else if (biClrUsed > 256)
+      ERREXIT(cinfo, JERR_BMP_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) biClrUsed, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) biClrUsed, mapentrysize);
+    /* account for size of colormap */
+    bPad -= biClrUsed * mapentrysize;
+  }
+
+  /* Skip any remaining pad bytes */
+  if (bPad < 0)                 /* incorrect bfOffBits value? */
+    ERREXIT(cinfo, JERR_BMP_BADHEADER);
+  while (--bPad >= 0) {
+    (void) read_byte(source);
+  }
+
+  /* Compute row width in file, including padding to 4-byte boundary */
+  if (source->bits_per_pixel == 24)
+    row_width = (JDIMENSION) (biWidth * 3);
+  else if (source->bits_per_pixel == 32)
+    row_width = (JDIMENSION) (biWidth * 4);
+  else
+    row_width = (JDIMENSION) biWidth;
+  while ((row_width & 3) != 0) row_width++;
+  source->row_width = row_width;
+
+  /* Allocate space for inversion array, prepare for preload pass */
+  source->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     row_width, (JDIMENSION) biHeight, (JDIMENSION) 1);
+  source->pub.get_pixel_rows = preload_image;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Allocate one-row buffer for returned data */
+  source->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (biWidth * 3), (JDIMENSION) 1);
+  source->pub.buffer_height = 1;
+
+  cinfo->in_color_space = JCS_RGB;
+  cinfo->input_components = 3;
+  cinfo->data_precision = 8;
+  cinfo->image_width = (JDIMENSION) biWidth;
+  cinfo->image_height = (JDIMENSION) biHeight;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_bmp (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for BMP format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_bmp (j_compress_ptr cinfo)
+{
+  bmp_source_ptr source;
+
+  /* Create module interface object */
+  source = (bmp_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(bmp_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_bmp;
+  source->pub.finish_input = finish_input_bmp;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* BMP_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdcolmap.c glmark2-2021.02/src/libjpeg-turbo/rdcolmap.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdcolmap.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdcolmap.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,254 @@
+/*
+ * rdcolmap.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file implements djpeg's "-map file" switch.  It reads a source image
+ * and constructs a colormap to be supplied to the JPEG decompressor.
+ *
+ * Currently, these file formats are supported for the map file:
+ *   GIF: the contents of the GIF's global colormap are used.
+ *   PPM (either text or raw flavor): the entire file is read and
+ *      each unique pixel value is entered in the map.
+ * Note that reading a large PPM file will be horrendously slow.
+ * Typically, a PPM-format map file should contain just one pixel
+ * of each desired color.  Such a file can be extracted from an
+ * ordinary image PPM file with ppmtomap(1).
+ *
+ * Rescaling a PPM that has a maxval unequal to MAXJSAMPLE is not
+ * currently implemented.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef QUANT_2PASS_SUPPORTED    /* otherwise can't quantize to supplied map */
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/*
+ * Add a (potentially) new color to the color map.
+ */
+
+LOCAL(void)
+add_map_entry (j_decompress_ptr cinfo, int R, int G, int B)
+{
+  JSAMPROW colormap0 = cinfo->colormap[0];
+  JSAMPROW colormap1 = cinfo->colormap[1];
+  JSAMPROW colormap2 = cinfo->colormap[2];
+  int ncolors = cinfo->actual_number_of_colors;
+  int index;
+
+  /* Check for duplicate color. */
+  for (index = 0; index < ncolors; index++) {
+    if (GETJSAMPLE(colormap0[index]) == R &&
+        GETJSAMPLE(colormap1[index]) == G &&
+        GETJSAMPLE(colormap2[index]) == B)
+      return;                   /* color is already in map */
+  }
+
+  /* Check for map overflow. */
+  if (ncolors >= (MAXJSAMPLE+1))
+    ERREXIT1(cinfo, JERR_QUANT_MANY_COLORS, (MAXJSAMPLE+1));
+
+  /* OK, add color to map. */
+  colormap0[ncolors] = (JSAMPLE) R;
+  colormap1[ncolors] = (JSAMPLE) G;
+  colormap2[ncolors] = (JSAMPLE) B;
+  cinfo->actual_number_of_colors++;
+}
+
+
+/*
+ * Extract color map from a GIF file.
+ */
+
+LOCAL(void)
+read_gif_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  int header[13];
+  int i, colormaplen;
+  int R, G, B;
+
+  /* Initial 'G' has already been read by read_color_map */
+  /* Read the rest of the GIF header and logical screen descriptor */
+  for (i = 1; i < 13; i++) {
+    if ((header[i] = getc(infile)) == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  }
+
+  /* Verify GIF Header */
+  if (header[1] != 'I' || header[2] != 'F')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* There must be a global color map. */
+  if ((header[10] & 0x80) == 0)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* OK, fetch it. */
+  colormaplen = 2 << (header[10] & 0x07);
+
+  for (i = 0; i < colormaplen; i++) {
+    R = getc(infile);
+    G = getc(infile);
+    B = getc(infile);
+    if (R == EOF || G == EOF || B == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    add_map_entry(cinfo,
+                  R << (BITS_IN_JSAMPLE-8),
+                  G << (BITS_IN_JSAMPLE-8),
+                  B << (BITS_IN_JSAMPLE-8));
+  }
+}
+
+
+/* Support routines for reading PPM */
+
+
+LOCAL(int)
+pbm_getc (FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_decompress_ptr cinfo, FILE *infile)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+  return val;
+}
+
+
+/*
+ * Extract color map from a PPM file.
+ */
+
+LOCAL(void)
+read_ppm_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  int c;
+  unsigned int w, h, maxval, row, col;
+  int R, G, B;
+
+  /* Initial 'P' has already been read by read_color_map */
+  c = getc(infile);             /* save format discriminator for a sec */
+
+  /* while we fetch the remaining header info */
+  w = read_pbm_integer(cinfo, infile);
+  h = read_pbm_integer(cinfo, infile);
+  maxval = read_pbm_integer(cinfo, infile);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  /* For now, we don't support rescaling from an unusual maxval. */
+  if (maxval != (unsigned int) MAXJSAMPLE)
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+
+  switch (c) {
+  case '3':                     /* it's a text-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = read_pbm_integer(cinfo, infile);
+        G = read_pbm_integer(cinfo, infile);
+        B = read_pbm_integer(cinfo, infile);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    for (row = 0; row < h; row++) {
+      for (col = 0; col < w; col++) {
+        R = getc(infile);
+        G = getc(infile);
+        B = getc(infile);
+        if (R == EOF || G == EOF || B == EOF)
+          ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+        add_map_entry(cinfo, R, G, B);
+      }
+    }
+    break;
+
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+
+/*
+ * Main entry point from djpeg.c.
+ *  Input: opened input file (from file name argument on command line).
+ *  Output: colormap and actual_number_of_colors fields are set in cinfo.
+ */
+
+GLOBAL(void)
+read_color_map (j_decompress_ptr cinfo, FILE *infile)
+{
+  /* Allocate space for a color map of maximum supported size. */
+  cinfo->colormap = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     (JDIMENSION) (MAXJSAMPLE+1), (JDIMENSION) 3);
+  cinfo->actual_number_of_colors = 0; /* initialize map to empty */
+
+  /* Read first byte to determine file format */
+  switch (getc(infile)) {
+  case 'G':
+    read_gif_map(cinfo, infile);
+    break;
+  case 'P':
+    read_ppm_map(cinfo, infile);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_BAD_CMAP_FILE);
+    break;
+  }
+}
+
+#endif /* QUANT_2PASS_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdgif.c glmark2-2021.02/src/libjpeg-turbo/rdgif.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdgif.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdgif.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,39 @@
+/*
+ * rdgif.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in GIF format.
+ *
+ *****************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,      *
+ * the ability to read GIF files has been removed from the IJG distribution. *
+ * Sorry about that.                                                         *
+ *****************************************************************************
+ *
+ * We are required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+/*
+ * The module selection routine for GIF format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_gif (j_compress_ptr cinfo)
+{
+  fprintf(stderr, "GIF input is unsupported for legal reasons.  Sorry.\n");
+  exit(EXIT_FAILURE);
+  return NULL;                  /* keep compiler happy */
+}
+
+#endif /* GIF_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdjpgcom.1 glmark2-2021.02/src/libjpeg-turbo/rdjpgcom.1
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdjpgcom.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdjpgcom.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,63 @@
+.TH RDJPGCOM 1 "02 April 2009"
+.SH NAME
+rdjpgcom \- display text comments from a JPEG file
+.SH SYNOPSIS
+.B rdjpgcom
+[
+.B \-raw
+]
+[
+.B \-verbose
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B rdjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and prints any text comments found in the file on the standard output.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.SH OPTIONS
+.TP
+.B \-raw
+Normally
+.B rdjpgcom
+escapes non-printable characters in comments, for security reasons.
+This option avoids that.
+.PP
+.B \-verbose
+Causes
+.B rdjpgcom
+to also display the JPEG image dimensions.
+.PP
+Switch names may be abbreviated, and are not case sensitive.
+.SH HINTS
+.B rdjpgcom
+does not depend on the IJG JPEG library.  Its source code is intended as an
+illustration of the minimum amount of code required to parse a JPEG file
+header correctly.
+.PP
+In
+.B \-verbose
+mode,
+.B rdjpgcom
+will also attempt to print the contents of any "APP12" markers as text.
+Some digital cameras produce APP12 markers containing useful textual
+information.  If you like, you can modify the source code to print
+other APPn marker types as well.
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR wrjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdjpgcom.c glmark2-2021.02/src/libjpeg-turbo/rdjpgcom.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdjpgcom.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdjpgcom.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,510 @@
+/*
+ * rdjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that displays
+ * the text in COM (comment) markers in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifdef HAVE_LOCALE_H
+#include <locale.h>             /* Bill Allombert: use locale for isprint */
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#else
+#define READ_BINARY     "rb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+
+/*
+ * These macros are used to read the input file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0            /* Start Of Frame N */
+#define M_SOF1  0xC1            /* N indicates which compression process */
+#define M_SOF2  0xC2            /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5            /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8            /* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9            /* End Of Image (end of datastream) */
+#define M_SOS   0xDA            /* Start Of Scan (begins compressed data) */
+#define M_APP0  0xE0            /* Application-specific marker, type N */
+#define M_APP12 0xEC            /* (we don't bother to list all 16 APPn's) */
+#define M_COM   0xFE            /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Process a COM marker.
+ * We want to print out the marker contents as legible text;
+ * we must guard against non-text junk and varying newline representations.
+ */
+
+static void
+process_COM (int raw)
+{
+  unsigned int length;
+  int ch;
+  int lastch = 0;
+
+  /* Bill Allombert: set locale properly for isprint */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "");
+#endif
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+
+  while (length > 0) {
+    ch = read_1_byte();
+    if (raw) {
+      putc(ch, stdout);
+    /* Emit the character in a readable form.
+     * Nonprintables are converted to \nnn form,
+     * while \ is converted to \\.
+     * Newlines in CR, CR/LF, or LF form will be printed as one newline.
+     */
+    } else if (ch == '\r') {
+      printf("\n");
+    } else if (ch == '\n') {
+      if (lastch != '\r')
+        printf("\n");
+    } else if (ch == '\\') {
+      printf("\\\\");
+    } else if (isprint(ch)) {
+      putc(ch, stdout);
+    } else {
+      printf("\\%03o", ch);
+    }
+    lastch = ch;
+    length--;
+  }
+  printf("\n");
+
+  /* Bill Allombert: revert to C locale */
+#ifdef HAVE_LOCALE_H
+  setlocale(LC_CTYPE, "C");
+#endif
+}
+
+
+/*
+ * Process a SOFn marker.
+ * This code is only needed if you want to know the image dimensions...
+ */
+
+static void
+process_SOFn (int marker)
+{
+  unsigned int length;
+  unsigned int image_height, image_width;
+  int data_precision, num_components;
+  const char *process;
+  int ci;
+
+  length = read_2_bytes();      /* usual parameter length count */
+
+  data_precision = read_1_byte();
+  image_height = read_2_bytes();
+  image_width = read_2_bytes();
+  num_components = read_1_byte();
+
+  switch (marker) {
+  case M_SOF0:  process = "Baseline";  break;
+  case M_SOF1:  process = "Extended sequential";  break;
+  case M_SOF2:  process = "Progressive";  break;
+  case M_SOF3:  process = "Lossless";  break;
+  case M_SOF5:  process = "Differential sequential";  break;
+  case M_SOF6:  process = "Differential progressive";  break;
+  case M_SOF7:  process = "Differential lossless";  break;
+  case M_SOF9:  process = "Extended sequential, arithmetic coding";  break;
+  case M_SOF10: process = "Progressive, arithmetic coding";  break;
+  case M_SOF11: process = "Lossless, arithmetic coding";  break;
+  case M_SOF13: process = "Differential sequential, arithmetic coding";  break;
+  case M_SOF14: process = "Differential progressive, arithmetic coding"; break;
+  case M_SOF15: process = "Differential lossless, arithmetic coding";  break;
+  default:      process = "Unknown";  break;
+  }
+
+  printf("JPEG image is %uw * %uh, %d color components, %d bits per sample\n",
+         image_width, image_height, num_components, data_precision);
+  printf("JPEG process: %s\n", process);
+
+  if (length != (unsigned int) (8 + num_components * 3))
+    ERREXIT("Bogus SOF marker length");
+
+  for (ci = 0; ci < num_components; ci++) {
+    (void) read_1_byte();       /* Component ID code */
+    (void) read_1_byte();       /* H, V sampling factors */
+    (void) read_1_byte();       /* Quantization table number */
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOS or EOI is seen;
+ * display any COM markers.
+ * While the companion program wrjpgcom will always insert COM markers before
+ * SOFn, other implementations might not, so we scan to SOS before stopping.
+ * If we were only interested in the image dimensions, we would stop at SOFn.
+ * (Conversely, if we only cared about COM markers, there would be no need
+ * for special code to handle SOFn; we could treat it like other markers.)
+ */
+
+static int
+scan_JPEG_header (int verbose, int raw)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+
+  /* Scan miscellaneous markers until we reach SOS. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      if (verbose)
+        process_SOFn(marker);
+      else
+        skip_variable();
+      break;
+
+    case M_SOS:                 /* stop before hitting compressed data */
+      return marker;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:
+      process_COM(raw);
+      break;
+
+    case M_APP12:
+      /* Some digital camera makers put useful textual information into
+       * APP12 markers, so we print those out too when in -verbose mode.
+       */
+      if (verbose) {
+        printf("APP12 contains:\n");
+        process_COM(raw);
+      } else
+        skip_variable();
+      break;
+
+    default:                    /* Anything else just gets skipped */
+      skip_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "rdjpgcom displays any textual comments in a JPEG file.\n");
+
+  fprintf(stderr, "Usage: %s [switches] [inputfile]\n", progname);
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -raw        Display non-printable characters in comments (unsafe)\n");
+  fprintf(stderr, "  -verbose    Also display dimensions of JPEG image\n");
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int verbose = 0, raw = 0;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "rdjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "verbose", 1)) {
+      verbose++;
+    } else if (keymatch(arg, "raw", 1)) {
+      raw = 1;
+    } else
+      usage();
+  }
+
+  /* Open the input file. */
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Scan the JPEG headers. */
+  (void) scan_JPEG_header(verbose, raw);
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdppm.c glmark2-2021.02/src/libjpeg-turbo/rdppm.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdppm.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdppm.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,461 @@
+/*
+ * rdppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Modified 2009 by Bill Allombert, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed PPM format).
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef PPM_SUPPORTED
+
+
+/* Portions of this code are based on the PBMPLUS library, which is:
+**
+** Copyright (C) 1988 by Jef Poskanzer.
+**
+** Permission to use, copy, modify, and distribute this software and its
+** documentation for any purpose and without fee is hereby granted, provided
+** that the above copyright notice appear in all copies and that both that
+** copyright notice and this permission notice appear in supporting
+** documentation.  This software is provided "as is" without express or
+** implied warranty.
+*/
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  U_CHAR *iobuffer;             /* fread's I/O buffer */
+  JSAMPROW pixrow;              /* compressor input buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JSAMPLE *rescale;             /* => maxval-remapping array, or NULL */
+  int maxval;
+} ppm_source_struct;
+
+typedef ppm_source_struct *ppm_source_ptr;
+
+
+LOCAL(int)
+pbm_getc (FILE *infile)
+/* Read next char, skipping over any comments */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(infile);
+  if (ch == '#') {
+    do {
+      ch = getc(infile);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(unsigned int)
+read_pbm_integer (j_compress_ptr cinfo, FILE *infile, unsigned int maxval)
+/* Read an unsigned decimal integer from the PPM file */
+/* Swallows one trailing character after the integer */
+/* Note that on a 16-bit-int machine, only values up to 64k can be read. */
+/* This should not be a problem in practice. */
+{
+  register int ch;
+  register unsigned int val;
+
+  /* Skip any leading whitespace */
+  do {
+    ch = pbm_getc(infile);
+    if (ch == EOF)
+      ERREXIT(cinfo, JERR_INPUT_EOF);
+  } while (ch == ' ' || ch == '\t' || ch == '\n' || ch == '\r');
+
+  if (ch < '0' || ch > '9')
+    ERREXIT(cinfo, JERR_PPM_NONNUMERIC);
+
+  val = ch - '0';
+  while ((ch = pbm_getc(infile)) >= '0' && ch <= '9') {
+    val *= 10;
+    val += ch - '0';
+  }
+
+  if (val > maxval)
+    ERREXIT(cinfo, JERR_PPM_TOOLARGE);
+
+  return val;
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ * In all cases, input is scaled to the size of JSAMPLE.
+ *
+ * A really fast path is provided for reading byte/sample raw files with
+ * maxval = MAXJSAMPLE, which is the normal case for 8-bit data.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_text_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_text_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading text-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  FILE *infile = source->pub.input_file;
+  register JSAMPROW ptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+  unsigned int maxval = source->maxval;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+    *ptr++ = rescale[read_pbm_integer(cinfo, infile, maxval)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_scaled_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+    *ptr++ = rescale[UCH(*bufferptr++)];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_raw_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-byte-format files with maxval = MAXJSAMPLE.
+ * In this case we just read right into the JSAMPLE buffer!
+ * Note that same code works for PPM and PGM files.
+ */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PGM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+METHODDEF(JDIMENSION)
+get_word_rgb_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading raw-word-format PPM files with any maxval */
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register U_CHAR *bufferptr;
+  register JSAMPLE *rescale = source->rescale;
+  JDIMENSION col;
+
+  if (! ReadOK(source->pub.input_file, source->iobuffer, source->buffer_width))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+  ptr = source->pub.buffer[0];
+  bufferptr = source->iobuffer;
+  for (col = cinfo->image_width; col > 0; col--) {
+    register int temp;
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+    temp  = UCH(*bufferptr++) << 8;
+    temp |= UCH(*bufferptr++);
+    *ptr++ = rescale[temp];
+  }
+  return 1;
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  ppm_source_ptr source = (ppm_source_ptr) sinfo;
+  int c;
+  unsigned int w, h, maxval;
+  boolean need_iobuffer, use_raw_buffer, need_rescale;
+
+  if (getc(source->pub.input_file) != 'P')
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  c = getc(source->pub.input_file); /* subformat discriminator character */
+
+  /* detect unsupported variants (ie, PBM) before trying to read header */
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+  case '3':                     /* it's a text-format PPM file */
+  case '5':                     /* it's a raw-format PGM file */
+  case '6':                     /* it's a raw-format PPM file */
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_NOT);
+    break;
+  }
+
+  /* fetch the remaining header info */
+  w = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  h = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+  maxval = read_pbm_integer(cinfo, source->pub.input_file, 65535);
+
+  if (w <= 0 || h <= 0 || maxval <= 0) /* error check */
+    ERREXIT(cinfo, JERR_PPM_NOT);
+
+  cinfo->data_precision = BITS_IN_JSAMPLE; /* we always rescale data to this */
+  cinfo->image_width = (JDIMENSION) w;
+  cinfo->image_height = (JDIMENSION) h;
+  source->maxval = maxval;
+
+  /* initialize flags to most common settings */
+  need_iobuffer = TRUE;         /* do we need an I/O buffer? */
+  use_raw_buffer = FALSE;       /* do we map input buffer onto I/O buffer? */
+  need_rescale = TRUE;          /* do we need a rescale array? */
+
+  switch (c) {
+  case '2':                     /* it's a text-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_gray_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '3':                     /* it's a text-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM_TEXT, w, h);
+    source->pub.get_pixel_rows = get_text_rgb_row;
+    need_iobuffer = FALSE;
+    break;
+
+  case '5':                     /* it's a raw-format PGM file */
+    cinfo->input_components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_PGM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_gray_row;
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_gray_row;
+    }
+    break;
+
+  case '6':                     /* it's a raw-format PPM file */
+    cinfo->input_components = 3;
+    cinfo->in_color_space = JCS_RGB;
+    TRACEMS2(cinfo, 1, JTRC_PPM, w, h);
+    if (maxval > 255) {
+      source->pub.get_pixel_rows = get_word_rgb_row;
+    } else if (maxval == MAXJSAMPLE && sizeof(JSAMPLE) == sizeof(U_CHAR)) {
+      source->pub.get_pixel_rows = get_raw_row;
+      use_raw_buffer = TRUE;
+      need_rescale = FALSE;
+    } else {
+      source->pub.get_pixel_rows = get_scaled_rgb_row;
+    }
+    break;
+  }
+
+  /* Allocate space for I/O buffer: 1 or 3 bytes or words/pixel. */
+  if (need_iobuffer) {
+    source->buffer_width = (size_t) w * cinfo->input_components *
+      ((maxval <= 255) ? sizeof(U_CHAR) : (2 * sizeof(U_CHAR)));
+    source->iobuffer = (U_CHAR *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  source->buffer_width);
+  }
+
+  /* Create compressor input buffer. */
+  if (use_raw_buffer) {
+    /* For unscaled raw-input case, we can just map it onto the I/O buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    source->pixrow = (JSAMPROW) source->iobuffer;
+    source->pub.buffer = & source->pixrow;
+    source->pub.buffer_height = 1;
+  } else {
+    /* Need to translate anyway, so make a separate sample buffer. */
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) w * cinfo->input_components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+  }
+
+  /* Compute the rescaling array if required. */
+  if (need_rescale) {
+    long val, half_maxval;
+
+    /* On 16-bit-int machines we have to be careful of maxval = 65535 */
+    source->rescale = (JSAMPLE *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  (size_t) (((long) maxval + 1L) *
+                                            sizeof(JSAMPLE)));
+    half_maxval = maxval / 2;
+    for (val = 0; val <= (long) maxval; val++) {
+      /* The multiplication here must be done in 32 bits to avoid overflow */
+      source->rescale[val] = (JSAMPLE) ((val * MAXJSAMPLE + half_maxval) /
+                                        maxval);
+    }
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_ppm (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for PPM format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_ppm (j_compress_ptr cinfo)
+{
+  ppm_source_ptr source;
+
+  /* Create module interface object */
+  source = (ppm_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(ppm_source_struct));
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_ppm;
+  source->pub.finish_input = finish_input_ppm;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* PPM_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdrle.c glmark2-2021.02/src/libjpeg-turbo/rdrle.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdrle.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdrle.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,389 @@
+/*
+ * rdrle.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Utah RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed RLE format).
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * We support the following types of RLE files:
+ *
+ *   GRAYSCALE   - 8 bits, no colormap
+ *   MAPPEDGRAY  - 8 bits, 1 channel colomap
+ *   PSEUDOCOLOR - 8 bits, 3 channel colormap
+ *   TRUECOLOR   - 24 bits, 3 channel colormap
+ *   DIRECTCOLOR - 24 bits, no colormap
+ *
+ * For now, we ignore any alpha channel in the image.
+ */
+
+typedef enum
+  { GRAYSCALE, MAPPEDGRAY, PSEUDOCOLOR, TRUECOLOR, DIRECTCOLOR } rle_kind;
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * to conform to JPEG's top-to-bottom order.  To do this, we read the
+ * incoming image into a virtual array on the first get_pixel_rows call,
+ * then fetch the required row from the virtual array on subsequent calls.
+ */
+
+typedef struct _rle_source_struct *rle_source_ptr;
+
+typedef struct _rle_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  rle_kind visual;              /* actual type of input file */
+  jvirt_sarray_ptr image;       /* virtual array to hold the image */
+  JDIMENSION row;               /* current row # in the virtual array */
+  rle_hdr header;               /* Input file information */
+  rle_pixel **rle_row;          /* holds a row returned by rle_getrow() */
+
+} rle_source_struct;
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION width, height;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Use RLE library routine to get the header info */
+  source->header = *rle_hdr_init(NULL);
+  source->header.rle_file = source->pub.input_file;
+  switch (rle_get_setup(&(source->header))) {
+  case RLE_SUCCESS:
+    /* A-OK */
+    break;
+  case RLE_NOT_RLE:
+    ERREXIT(cinfo, JERR_RLE_NOT);
+    break;
+  case RLE_NO_SPACE:
+    ERREXIT(cinfo, JERR_RLE_MEM);
+    break;
+  case RLE_EMPTY:
+    ERREXIT(cinfo, JERR_RLE_EMPTY);
+    break;
+  case RLE_EOF:
+    ERREXIT(cinfo, JERR_RLE_EOF);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_RLE_BADERROR);
+    break;
+  }
+
+  /* Figure out what we have, set private vars and return values accordingly */
+
+  width  = source->header.xmax - source->header.xmin + 1;
+  height = source->header.ymax - source->header.ymin + 1;
+  source->header.xmin = 0;              /* realign horizontally */
+  source->header.xmax = width-1;
+
+  cinfo->image_width      = width;
+  cinfo->image_height     = height;
+  cinfo->data_precision   = 8;  /* we can only handle 8 bit data */
+
+  if (source->header.ncolors == 1 && source->header.ncmap == 0) {
+    source->visual     = GRAYSCALE;
+    TRACEMS2(cinfo, 1, JTRC_RLE_GRAY, width, height);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 1) {
+    source->visual     = MAPPEDGRAY;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPGRAY, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 1 && source->header.ncmap == 3) {
+    source->visual     = PSEUDOCOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_MAPPED, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 3) {
+    source->visual     = TRUECOLOR;
+    TRACEMS3(cinfo, 1, JTRC_RLE_FULLMAP, width, height,
+             1 << source->header.cmaplen);
+  } else if (source->header.ncolors == 3 && source->header.ncmap == 0) {
+    source->visual     = DIRECTCOLOR;
+    TRACEMS2(cinfo, 1, JTRC_RLE, width, height);
+  } else
+    ERREXIT(cinfo, JERR_RLE_UNSUPPORTED);
+
+  if (source->visual == GRAYSCALE || source->visual == MAPPEDGRAY) {
+    cinfo->in_color_space   = JCS_GRAYSCALE;
+    cinfo->input_components = 1;
+  } else {
+    cinfo->in_color_space   = JCS_RGB;
+    cinfo->input_components = 3;
+  }
+
+  /*
+   * A place to hold each scanline while it's converted.
+   * (GRAYSCALE scanlines don't need converting)
+   */
+  if (source->visual != GRAYSCALE) {
+    source->rle_row = (rle_pixel**) (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width, (JDIMENSION) cinfo->input_components);
+  }
+
+  /* request a virtual array to hold the image */
+  source->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (width * source->header.ncolors),
+     (JDIMENSION) height, (JDIMENSION) 1);
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    /* count file input as separate pass */
+    progress->total_extra_passes++;
+  }
+#endif
+
+  source->pub.buffer_height = 1;
+}
+
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for GRAYSCALE, MAPPEDGRAY, TRUECOLOR, and DIRECTCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_rle_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+
+  source->row--;
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  return 1;
+}
+
+/*
+ * Read one row of pixels.
+ * Called only after load_image has read the image into the virtual array.
+ * Used for PSEUDOCOLOR images.
+ */
+
+METHODDEF(JDIMENSION)
+get_pseudocolor_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JSAMPROW src_row, dest_row;
+  JDIMENSION col;
+  rle_map *colormap;
+  int val;
+
+  colormap = source->header.cmap;
+  dest_row = source->pub.buffer[0];
+  source->row--;
+  src_row = *(*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->image, source->row, (JDIMENSION) 1, FALSE);
+
+  for (col = cinfo->image_width; col > 0; col--) {
+    val = GETJSAMPLE(*src_row++);
+    *dest_row++ = (JSAMPLE) (colormap[val      ] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 256] >> 8);
+    *dest_row++ = (JSAMPLE) (colormap[val + 512] >> 8);
+  }
+
+  return 1;
+}
+
+
+/*
+ * Load the image into a virtual array.  We have to do this because RLE
+ * files start at the lower left while the JPEG standard has them starting
+ * in the upper left.  This is called the first time we want to get a row
+ * of input.  What we do is load the RLE data into the array and then call
+ * the appropriate routine to read one row from the array.  Before returning,
+ * we set source->pub.get_pixel_rows so that subsequent calls go straight to
+ * the appropriate row-reading routine.
+ */
+
+METHODDEF(JDIMENSION)
+load_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  rle_source_ptr source = (rle_source_ptr) sinfo;
+  JDIMENSION row, col;
+  JSAMPROW  scanline, red_ptr, green_ptr, blue_ptr;
+  rle_pixel **rle_row;
+  rle_map *colormap;
+  char channel;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  colormap = source->header.cmap;
+  rle_row = source->rle_row;
+
+  /* Read the RLE data into our virtual array.
+   * We assume here that rle_pixel is represented the same as JSAMPLE.
+   */
+  RLE_CLR_BIT(source->header, RLE_ALPHA); /* don't read the alpha channel */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->image_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  switch (source->visual) {
+
+  case GRAYSCALE:
+  case PSEUDOCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+         ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case MAPPEDGRAY:
+  case TRUECOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_row = source->rle_row;
+      rle_getrow(&source->header, rle_row);
+
+      for (col = 0; col < cinfo->image_width; col++) {
+        for (channel = 0; channel < source->header.ncolors; channel++) {
+          *scanline++ = (JSAMPLE)
+            (colormap[GETJSAMPLE(rle_row[channel][col]) + 256 * channel] >> 8);
+        }
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+    break;
+
+  case DIRECTCOLOR:
+    for (row = 0; row < cinfo->image_height; row++) {
+      scanline = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, source->image, row, (JDIMENSION) 1, TRUE);
+      rle_getrow(&source->header, rle_row);
+
+      red_ptr   = rle_row[0];
+      green_ptr = rle_row[1];
+      blue_ptr  = rle_row[2];
+
+      for (col = cinfo->image_width; col > 0; col--) {
+        *scanline++ = *red_ptr++;
+        *scanline++ = *green_ptr++;
+        *scanline++ = *blue_ptr++;
+      }
+
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Set up to call proper row-extraction routine in future */
+  if (source->visual == PSEUDOCOLOR) {
+    source->pub.buffer = source->rle_row;
+    source->pub.get_pixel_rows = get_pseudocolor_row;
+  } else {
+    source->pub.get_pixel_rows = get_rle_row;
+  }
+  source->row = cinfo->image_height;
+
+  /* And fetch the topmost (bottommost) row */
+  return (*source->pub.get_pixel_rows) (cinfo, sinfo);
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_rle (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for RLE format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_rle (j_compress_ptr cinfo)
+{
+  rle_source_ptr source;
+
+  /* Create module interface object */
+  source = (rle_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(rle_source_struct));
+  /* Fill in method ptrs */
+  source->pub.start_input = start_input_rle;
+  source->pub.finish_input = finish_input_rle;
+  source->pub.get_pixel_rows = load_image;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* RLE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdswitch.c glmark2-2021.02/src/libjpeg-turbo/rdswitch.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdswitch.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdswitch.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,424 @@
+/*
+ * rdswitch.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to process some of cjpeg's more complicated
+ * command-line switches.  Switches processed here are:
+ *      -qtables file           Read quantization tables from text file
+ *      -scans file             Read scan script from text file
+ *      -quality N[,N,...]      Set quality ratings
+ *      -qslots N[,N,...]       Set component quantization table selectors
+ *      -sample HxV[,HxV,...]   Set component sampling factors
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include <ctype.h>              /* to declare isdigit(), isspace() */
+
+
+LOCAL(int)
+text_getc (FILE *file)
+/* Read next char, skipping over any comments (# to end of line) */
+/* A comment/newline sequence is returned as a newline */
+{
+  register int ch;
+
+  ch = getc(file);
+  if (ch == '#') {
+    do {
+      ch = getc(file);
+    } while (ch != '\n' && ch != EOF);
+  }
+  return ch;
+}
+
+
+LOCAL(boolean)
+read_text_integer (FILE *file, long *result, int *termchar)
+/* Read an unsigned decimal integer from a file, store it in result */
+/* Reads one trailing character after the integer; returns it in termchar */
+{
+  register int ch;
+  register long val;
+
+  /* Skip any leading whitespace, detect EOF */
+  do {
+    ch = text_getc(file);
+    if (ch == EOF) {
+      *termchar = ch;
+      return FALSE;
+    }
+  } while (isspace(ch));
+
+  if (! isdigit(ch)) {
+    *termchar = ch;
+    return FALSE;
+  }
+
+  val = ch - '0';
+  while ((ch = text_getc(file)) != EOF) {
+    if (! isdigit(ch))
+      break;
+    val *= 10;
+    val += ch - '0';
+  }
+  *result = val;
+  *termchar = ch;
+  return TRUE;
+}
+
+
+#if JPEG_LIB_VERSION < 70
+static int q_scale_factor[NUM_QUANT_TBLS] = {100, 100, 100, 100};
+#endif
+
+GLOBAL(boolean)
+read_quant_tables (j_compress_ptr cinfo, char *filename,
+                   boolean force_baseline)
+/* Read a set of quantization tables from the specified file.
+ * The file is plain ASCII text: decimal numbers with whitespace between.
+ * Comments preceded by '#' may be included in the file.
+ * There may be one to NUM_QUANT_TBLS tables in the file, each of 64 values.
+ * The tables are implicitly numbered 0,1,etc.
+ * NOTE: does not affect the qslots mapping, which will default to selecting
+ * table 0 for luminance (or primary) components, 1 for chrominance components.
+ * You must use -qslots if you want a different component->table mapping.
+ */
+{
+  FILE *fp;
+  int tblno, i, termchar;
+  long val;
+  unsigned int table[DCTSIZE2];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open table file %s\n", filename);
+    return FALSE;
+  }
+  tblno = 0;
+
+  while (read_text_integer(fp, &val, &termchar)) { /* read 1st element of table */
+    if (tblno >= NUM_QUANT_TBLS) {
+      fprintf(stderr, "Too many tables in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    table[0] = (unsigned int) val;
+    for (i = 1; i < DCTSIZE2; i++) {
+      if (! read_text_integer(fp, &val, &termchar)) {
+        fprintf(stderr, "Invalid table data in file %s\n", filename);
+        fclose(fp);
+        return FALSE;
+      }
+      table[i] = (unsigned int) val;
+    }
+#if JPEG_LIB_VERSION >= 70
+    jpeg_add_quant_table(cinfo, tblno, table, cinfo->q_scale_factor[tblno],
+                         force_baseline);
+#else
+    jpeg_add_quant_table(cinfo, tblno, table, q_scale_factor[tblno],
+                         force_baseline);
+#endif
+    tblno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+
+#ifdef C_MULTISCAN_FILES_SUPPORTED
+
+LOCAL(boolean)
+read_scan_integer (FILE *file, long *result, int *termchar)
+/* Variant of read_text_integer that always looks for a non-space termchar;
+ * this simplifies parsing of punctuation in scan scripts.
+ */
+{
+  register int ch;
+
+  if (! read_text_integer(file, result, termchar))
+    return FALSE;
+  ch = *termchar;
+  while (ch != EOF && isspace(ch))
+    ch = text_getc(file);
+  if (isdigit(ch)) {            /* oops, put it back */
+    if (ungetc(ch, file) == EOF)
+      return FALSE;
+    ch = ' ';
+  } else {
+    /* Any separators other than ';' and ':' are ignored;
+     * this allows user to insert commas, etc, if desired.
+     */
+    if (ch != EOF && ch != ';' && ch != ':')
+      ch = ' ';
+  }
+  *termchar = ch;
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+read_scan_script (j_compress_ptr cinfo, char *filename)
+/* Read a scan script from the specified text file.
+ * Each entry in the file defines one scan to be emitted.
+ * Entries are separated by semicolons ';'.
+ * An entry contains one to four component indexes,
+ * optionally followed by a colon ':' and four progressive-JPEG parameters.
+ * The component indexes denote which component(s) are to be transmitted
+ * in the current scan.  The first component has index 0.
+ * Sequential JPEG is used if the progressive-JPEG parameters are omitted.
+ * The file is free format text: any whitespace may appear between numbers
+ * and the ':' and ';' punctuation marks.  Also, other punctuation (such
+ * as commas or dashes) can be placed between numbers if desired.
+ * Comments preceded by '#' may be included in the file.
+ * Note: we do very little validity checking here;
+ * jcmaster.c will validate the script parameters.
+ */
+{
+  FILE *fp;
+  int scanno, ncomps, termchar;
+  long val;
+  jpeg_scan_info *scanptr;
+#define MAX_SCANS  100          /* quite arbitrary limit */
+  jpeg_scan_info scans[MAX_SCANS];
+
+  if ((fp = fopen(filename, "r")) == NULL) {
+    fprintf(stderr, "Can't open scan definition file %s\n", filename);
+    return FALSE;
+  }
+  scanptr = scans;
+  scanno = 0;
+
+  while (read_scan_integer(fp, &val, &termchar)) {
+    if (scanno >= MAX_SCANS) {
+      fprintf(stderr, "Too many scans defined in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr->component_index[0] = (int) val;
+    ncomps = 1;
+    while (termchar == ' ') {
+      if (ncomps >= MAX_COMPS_IN_SCAN) {
+        fprintf(stderr, "Too many components in one scan in file %s\n",
+                filename);
+        fclose(fp);
+        return FALSE;
+      }
+      if (! read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->component_index[ncomps] = (int) val;
+      ncomps++;
+    }
+    scanptr->comps_in_scan = ncomps;
+    if (termchar == ':') {
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ss = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Se = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar) || termchar != ' ')
+        goto bogus;
+      scanptr->Ah = (int) val;
+      if (! read_scan_integer(fp, &val, &termchar))
+        goto bogus;
+      scanptr->Al = (int) val;
+    } else {
+      /* set non-progressive parameters */
+      scanptr->Ss = 0;
+      scanptr->Se = DCTSIZE2-1;
+      scanptr->Ah = 0;
+      scanptr->Al = 0;
+    }
+    if (termchar != ';' && termchar != EOF) {
+bogus:
+      fprintf(stderr, "Invalid scan entry format in file %s\n", filename);
+      fclose(fp);
+      return FALSE;
+    }
+    scanptr++, scanno++;
+  }
+
+  if (termchar != EOF) {
+    fprintf(stderr, "Non-numeric data in file %s\n", filename);
+    fclose(fp);
+    return FALSE;
+  }
+
+  if (scanno > 0) {
+    /* Stash completed scan list in cinfo structure.
+     * NOTE: for cjpeg's use, JPOOL_IMAGE is the right lifetime for this data,
+     * but if you want to compress multiple images you'd want JPOOL_PERMANENT.
+     */
+    scanptr = (jpeg_scan_info *)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  scanno * sizeof(jpeg_scan_info));
+    MEMCOPY(scanptr, scans, scanno * sizeof(jpeg_scan_info));
+    cinfo->scan_info = scanptr;
+    cinfo->num_scans = scanno;
+  }
+
+  fclose(fp);
+  return TRUE;
+}
+
+#endif /* C_MULTISCAN_FILES_SUPPORTED */
+
+
+#if JPEG_LIB_VERSION < 70
+/* These are the sample quantization tables given in JPEG spec section K.1.
+ * The spec says that the values given produce "good" quality, and
+ * when divided by 2, "very good" quality.
+ */
+static const unsigned int std_luminance_quant_tbl[DCTSIZE2] = {
+  16,  11,  10,  16,  24,  40,  51,  61,
+  12,  12,  14,  19,  26,  58,  60,  55,
+  14,  13,  16,  24,  40,  57,  69,  56,
+  14,  17,  22,  29,  51,  87,  80,  62,
+  18,  22,  37,  56,  68, 109, 103,  77,
+  24,  35,  55,  64,  81, 104, 113,  92,
+  49,  64,  78,  87, 103, 121, 120, 101,
+  72,  92,  95,  98, 112, 100, 103,  99
+};
+static const unsigned int std_chrominance_quant_tbl[DCTSIZE2] = {
+  17,  18,  24,  47,  99,  99,  99,  99,
+  18,  21,  26,  66,  99,  99,  99,  99,
+  24,  26,  56,  99,  99,  99,  99,  99,
+  47,  66,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99,
+  99,  99,  99,  99,  99,  99,  99,  99
+};
+
+
+LOCAL(void)
+jpeg_default_qtables (j_compress_ptr cinfo, boolean force_baseline)
+{
+  jpeg_add_quant_table(cinfo, 0, std_luminance_quant_tbl,
+                       q_scale_factor[0], force_baseline);
+  jpeg_add_quant_table(cinfo, 1, std_chrominance_quant_tbl,
+                       q_scale_factor[1], force_baseline);
+}
+#endif
+
+
+GLOBAL(boolean)
+set_quality_ratings (j_compress_ptr cinfo, char *arg, boolean force_baseline)
+/* Process a quality-ratings parameter string, of the form
+ *     N[,N,...]
+ * If there are more q-table slots than parameters, the last value is replicated.
+ */
+{
+  int val = 75;                 /* default value */
+  int tblno;
+  char ch;
+
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      /* Convert user 0-100 rating to percentage scaling */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#else
+      q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#endif
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining factors to last value */
+#if JPEG_LIB_VERSION >= 70
+      cinfo->q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#else
+      q_scale_factor[tblno] = jpeg_quality_scaling(val);
+#endif
+    }
+  }
+  jpeg_default_qtables(cinfo, force_baseline);
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_quant_slots (j_compress_ptr cinfo, char *arg)
+/* Process a quantization-table-selectors parameter string, of the form
+ *     N[,N,...]
+ * If there are more components than parameters, the last value is replicated.
+ */
+{
+  int val = 0;                  /* default table # */
+  int ci;
+  char ch;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch = ',';                 /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c", &val, &ch) < 1)
+        return FALSE;
+      if (ch != ',')            /* syntax check */
+        return FALSE;
+      if (val < 0 || val >= NUM_QUANT_TBLS) {
+        fprintf(stderr, "JPEG quantization tables are numbered 0..%d\n",
+                NUM_QUANT_TBLS-1);
+        return FALSE;
+      }
+      cinfo->comp_info[ci].quant_tbl_no = val;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining components to last table */
+      cinfo->comp_info[ci].quant_tbl_no = val;
+    }
+  }
+  return TRUE;
+}
+
+
+GLOBAL(boolean)
+set_sample_factors (j_compress_ptr cinfo, char *arg)
+/* Process a sample-factors parameter string, of the form
+ *     HxV[,HxV,...]
+ * If there are more components than parameters, "1x1" is assumed for the rest.
+ */
+{
+  int ci, val1, val2;
+  char ch1, ch2;
+
+  for (ci = 0; ci < MAX_COMPONENTS; ci++) {
+    if (*arg) {
+      ch2 = ',';                /* if not set by sscanf, will be ',' */
+      if (sscanf(arg, "%d%c%d%c", &val1, &ch1, &val2, &ch2) < 3)
+        return FALSE;
+      if ((ch1 != 'x' && ch1 != 'X') || ch2 != ',') /* syntax check */
+        return FALSE;
+      if (val1 <= 0 || val1 > 4 || val2 <= 0 || val2 > 4) {
+        fprintf(stderr, "JPEG sampling factors must be 1..4\n");
+        return FALSE;
+      }
+      cinfo->comp_info[ci].h_samp_factor = val1;
+      cinfo->comp_info[ci].v_samp_factor = val2;
+      while (*arg && *arg++ != ',') /* advance to next segment of arg string */
+        ;
+    } else {
+      /* reached end of parameter, set remaining components to 1x1 sampling */
+      cinfo->comp_info[ci].h_samp_factor = 1;
+      cinfo->comp_info[ci].v_samp_factor = 1;
+    }
+  }
+  return TRUE;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdtarga.c glmark2-2021.02/src/libjpeg-turbo/rdtarga.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/rdtarga.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/rdtarga.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,503 @@
+/*
+ * rdtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to read input images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume input from
+ * an ordinary stdio stream.  They further assume that reading begins
+ * at the start of the file; start_input may need work if the
+ * user interface has already read some data (e.g., to determine that
+ * the file is indeed Targa format).
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/* Macros to deal with unsigned chars as efficiently as compiler allows */
+
+#ifdef HAVE_UNSIGNED_CHAR
+typedef unsigned char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else /* !HAVE_UNSIGNED_CHAR */
+#ifdef __CHAR_UNSIGNED__
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x))
+#else
+typedef char U_CHAR;
+#define UCH(x)  ((int) (x) & 0xFF)
+#endif
+#endif /* HAVE_UNSIGNED_CHAR */
+
+
+#define ReadOK(file,buffer,len) (JFREAD(file,buffer,len) == ((size_t) (len)))
+
+
+/* Private version of data source object */
+
+typedef struct _tga_source_struct *tga_source_ptr;
+
+typedef struct _tga_source_struct {
+  struct cjpeg_source_struct pub; /* public fields */
+
+  j_compress_ptr cinfo;         /* back link saves passing separate parm */
+
+  JSAMPARRAY colormap;          /* Targa colormap (converted to my format) */
+
+  jvirt_sarray_ptr whole_image; /* Needed if funny input row order */
+  JDIMENSION current_row;       /* Current logical row number to read */
+
+  /* Pointer to routine to extract next Targa pixel from input file */
+  void (*read_pixel) (tga_source_ptr sinfo);
+
+  /* Result of read_pixel is delivered here: */
+  U_CHAR tga_pixel[4];
+
+  int pixel_size;               /* Bytes per Targa pixel (1 to 4) */
+
+  /* State info for reading RLE-coded pixels; both counts must be init to 0 */
+  int block_count;              /* # of pixels remaining in RLE block */
+  int dup_pixel_count;          /* # of times to duplicate previous pixel */
+
+  /* This saves the correct pixel-row-expansion method for preload_image */
+  JDIMENSION (*get_pixel_rows) (j_compress_ptr cinfo, cjpeg_source_ptr sinfo);
+} tga_source_struct;
+
+
+/* For expanding 5-bit pixel values to 8-bit with best rounding */
+
+static const UINT8 c5to8bits[32] = {
+    0,   8,  16,  25,  33,  41,  49,  58,
+   66,  74,  82,  90,  99, 107, 115, 123,
+  132, 140, 148, 156, 165, 173, 181, 189,
+  197, 206, 214, 222, 230, 239, 247, 255
+};
+
+
+
+LOCAL(int)
+read_byte (tga_source_ptr sinfo)
+/* Read next byte from Targa file */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int c;
+
+  if ((c = getc(infile)) == EOF)
+    ERREXIT(sinfo->cinfo, JERR_INPUT_EOF);
+  return c;
+}
+
+
+LOCAL(void)
+read_colormap (tga_source_ptr sinfo, int cmaplen, int mapentrysize)
+/* Read the colormap from a Targa file */
+{
+  int i;
+
+  /* Presently only handles 24-bit BGR format */
+  if (mapentrysize != 24)
+    ERREXIT(sinfo->cinfo, JERR_TGA_BADCMAP);
+
+  for (i = 0; i < cmaplen; i++) {
+    sinfo->colormap[2][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[1][i] = (JSAMPLE) read_byte(sinfo);
+    sinfo->colormap[0][i] = (JSAMPLE) read_byte(sinfo);
+  }
+}
+
+
+/*
+ * read_pixel methods: get a single pixel from Targa file into tga_pixel[]
+ */
+
+METHODDEF(void)
+read_non_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file; no RLE expansion */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+METHODDEF(void)
+read_rle_pixel (tga_source_ptr sinfo)
+/* Read one Targa pixel from the input file, expanding RLE data as needed */
+{
+  register FILE *infile = sinfo->pub.input_file;
+  register int i;
+
+  /* Duplicate previously read pixel? */
+  if (sinfo->dup_pixel_count > 0) {
+    sinfo->dup_pixel_count--;
+    return;
+  }
+
+  /* Time to read RLE block header? */
+  if (--sinfo->block_count < 0) { /* decrement pixels remaining in block */
+    i = read_byte(sinfo);
+    if (i & 0x80) {             /* Start of duplicate-pixel block? */
+      sinfo->dup_pixel_count = i & 0x7F; /* number of dups after this one */
+      sinfo->block_count = 0;   /* then read new block header */
+    } else {
+      sinfo->block_count = i & 0x7F; /* number of pixels after this one */
+    }
+  }
+
+  /* Read next pixel */
+  for (i = 0; i < sinfo->pixel_size; i++) {
+    sinfo->tga_pixel[i] = (U_CHAR) getc(infile);
+  }
+}
+
+
+/*
+ * Read one row of pixels.
+ *
+ * We provide several different versions depending on input file format.
+ */
+
+
+METHODDEF(JDIMENSION)
+get_8bit_gray_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit grayscale pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_8bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 8-bit colormap indexes */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+  register JSAMPARRAY colormap = source->colormap;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    *ptr++ = colormap[0][t];
+    *ptr++ = colormap[1][t];
+    *ptr++ = colormap[2][t];
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_16bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 16-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register int t;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    t = UCH(source->tga_pixel[0]);
+    t += UCH(source->tga_pixel[1]) << 8;
+    /* We expand 5 bit data to 8 bit sample width.
+     * The format of the 16-bit (LSB first) input word is
+     *     xRRRRRGGGGGBBBBB
+     */
+    ptr[2] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[1] = (JSAMPLE) c5to8bits[t & 0x1F];
+    t >>= 5;
+    ptr[0] = (JSAMPLE) c5to8bits[t & 0x1F];
+    ptr += 3;
+  }
+  return 1;
+}
+
+METHODDEF(JDIMENSION)
+get_24bit_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+/* This version is for reading 24-bit pixels */
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = source->pub.buffer[0];
+  for (col = cinfo->image_width; col > 0; col--) {
+    (*source->read_pixel) (source); /* Load next pixel into tga_pixel */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[2]); /* change BGR to RGB order */
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[1]);
+    *ptr++ = (JSAMPLE) UCH(source->tga_pixel[0]);
+  }
+  return 1;
+}
+
+/*
+ * Targa also defines a 32-bit pixel format with order B,G,R,A.
+ * We presently ignore the attribute byte, so the code for reading
+ * these pixels is identical to the 24-bit routine above.
+ * This works because the actual pixel length is only known to read_pixel.
+ */
+
+#define get_32bit_row  get_24bit_row
+
+
+/*
+ * This method is for re-reading the input data in standard top-down
+ * row order.  The entire image has already been read into whole_image
+ * with proper conversion of pixel format, but it's in a funny row order.
+ */
+
+METHODDEF(JDIMENSION)
+get_memory_row (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION source_row;
+
+  /* Compute row of source that maps to current_row of normal order */
+  /* For now, assume image is bottom-up and not interlaced. */
+  /* NEEDS WORK to support interlaced images! */
+  source_row = cinfo->image_height - source->current_row - 1;
+
+  /* Fetch that row from virtual array */
+  source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, source->whole_image,
+     source_row, (JDIMENSION) 1, FALSE);
+
+  source->current_row++;
+  return 1;
+}
+
+
+/*
+ * This method loads the image into whole_image during the first call on
+ * get_pixel_rows.  The get_pixel_rows pointer is then adjusted to call
+ * get_memory_row on subsequent calls.
+ */
+
+METHODDEF(JDIMENSION)
+preload_image (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  JDIMENSION row;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Read the data into a virtual array in input-file row order. */
+  for (row = 0; row < cinfo->image_height; row++) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) row;
+      progress->pub.pass_limit = (long) cinfo->image_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    source->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, source->whole_image, row, (JDIMENSION) 1, TRUE);
+    (*source->get_pixel_rows) (cinfo, sinfo);
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Set up to read from the virtual array in unscrambled order */
+  source->pub.get_pixel_rows = get_memory_row;
+  source->current_row = 0;
+  /* And read the first row */
+  return get_memory_row(cinfo, sinfo);
+}
+
+
+/*
+ * Read the file header; return image size and component count.
+ */
+
+METHODDEF(void)
+start_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  tga_source_ptr source = (tga_source_ptr) sinfo;
+  U_CHAR targaheader[18];
+  int idlen, cmaptype, subtype, flags, interlace_type, components;
+  unsigned int width, height, maplen;
+  boolean is_bottom_up;
+
+#define GET_2B(offset)  ((unsigned int) UCH(targaheader[offset]) + \
+                         (((unsigned int) UCH(targaheader[offset+1])) << 8))
+
+  if (! ReadOK(source->pub.input_file, targaheader, 18))
+    ERREXIT(cinfo, JERR_INPUT_EOF);
+
+  /* Pretend "15-bit" pixels are 16-bit --- we ignore attribute bit anyway */
+  if (targaheader[16] == 15)
+    targaheader[16] = 16;
+
+  idlen = UCH(targaheader[0]);
+  cmaptype = UCH(targaheader[1]);
+  subtype = UCH(targaheader[2]);
+  maplen = GET_2B(5);
+  width = GET_2B(12);
+  height = GET_2B(14);
+  source->pixel_size = UCH(targaheader[16]) >> 3;
+  flags = UCH(targaheader[17]); /* Image Descriptor byte */
+
+  is_bottom_up = ((flags & 0x20) == 0); /* bit 5 set => top-down */
+  interlace_type = flags >> 6;  /* bits 6/7 are interlace code */
+
+  if (cmaptype > 1 ||           /* cmaptype must be 0 or 1 */
+      source->pixel_size < 1 || source->pixel_size > 4 ||
+      (UCH(targaheader[16]) & 7) != 0 || /* bits/pixel must be multiple of 8 */
+      interlace_type != 0 ||      /* currently don't allow interlaced image */
+      width == 0 || height == 0)  /* image width/height must be non-zero */
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+
+  if (subtype > 8) {
+    /* It's an RLE-coded file */
+    source->read_pixel = read_rle_pixel;
+    source->block_count = source->dup_pixel_count = 0;
+    subtype -= 8;
+  } else {
+    /* Non-RLE file */
+    source->read_pixel = read_non_rle_pixel;
+  }
+
+  /* Now should have subtype 1, 2, or 3 */
+  components = 3;               /* until proven different */
+  cinfo->in_color_space = JCS_RGB;
+
+  switch (subtype) {
+  case 1:                       /* Colormapped image */
+    if (source->pixel_size == 1 && cmaptype == 1)
+      source->get_pixel_rows = get_8bit_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_MAPPED, width, height);
+    break;
+  case 2:                       /* RGB image */
+    switch (source->pixel_size) {
+    case 2:
+      source->get_pixel_rows = get_16bit_row;
+      break;
+    case 3:
+      source->get_pixel_rows = get_24bit_row;
+      break;
+    case 4:
+      source->get_pixel_rows = get_32bit_row;
+      break;
+    default:
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+      break;
+    }
+    TRACEMS2(cinfo, 1, JTRC_TGA, width, height);
+    break;
+  case 3:                       /* Grayscale image */
+    components = 1;
+    cinfo->in_color_space = JCS_GRAYSCALE;
+    if (source->pixel_size == 1)
+      source->get_pixel_rows = get_8bit_gray_row;
+    else
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    TRACEMS2(cinfo, 1, JTRC_TGA_GRAY, width, height);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    break;
+  }
+
+  if (is_bottom_up) {
+    /* Create a virtual array to buffer the upside-down image. */
+    source->whole_image = (*cinfo->mem->request_virt_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+       (JDIMENSION) width * components, (JDIMENSION) height, (JDIMENSION) 1);
+    if (cinfo->progress != NULL) {
+      cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+      progress->total_extra_passes++; /* count file input as separate pass */
+    }
+    /* source->pub.buffer will point to the virtual array. */
+    source->pub.buffer_height = 1; /* in case anyone looks at it */
+    source->pub.get_pixel_rows = preload_image;
+  } else {
+    /* Don't need a virtual array, but do need a one-row input buffer. */
+    source->whole_image = NULL;
+    source->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       (JDIMENSION) width * components, (JDIMENSION) 1);
+    source->pub.buffer_height = 1;
+    source->pub.get_pixel_rows = source->get_pixel_rows;
+  }
+
+  while (idlen--)               /* Throw away ID field */
+    (void) read_byte(source);
+
+  if (maplen > 0) {
+    if (maplen > 256 || GET_2B(3) != 0)
+      ERREXIT(cinfo, JERR_TGA_BADCMAP);
+    /* Allocate space to store the colormap */
+    source->colormap = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, (JDIMENSION) maplen, (JDIMENSION) 3);
+    /* and read it from the file */
+    read_colormap(source, (int) maplen, UCH(targaheader[7]));
+  } else {
+    if (cmaptype)               /* but you promised a cmap! */
+      ERREXIT(cinfo, JERR_TGA_BADPARMS);
+    source->colormap = NULL;
+  }
+
+  cinfo->input_components = components;
+  cinfo->data_precision = 8;
+  cinfo->image_width = width;
+  cinfo->image_height = height;
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_input_tga (j_compress_ptr cinfo, cjpeg_source_ptr sinfo)
+{
+  /* no work */
+}
+
+
+/*
+ * The module selection routine for Targa format input.
+ */
+
+GLOBAL(cjpeg_source_ptr)
+jinit_read_targa (j_compress_ptr cinfo)
+{
+  tga_source_ptr source;
+
+  /* Create module interface object */
+  source = (tga_source_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(tga_source_struct));
+  source->cinfo = cinfo;        /* make back link for subroutines */
+  /* Fill in method ptrs, except get_pixel_rows which start_input sets */
+  source->pub.start_input = start_input_tga;
+  source->pub.finish_input = finish_input_tga;
+
+  return (cjpeg_source_ptr) source;
+}
+
+#endif /* TARGA_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README glmark2-2021.02/src/libjpeg-turbo/README
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/README	1970-01-01 08:00:00.000000000 +0800
@@ -1,290 +0,0 @@
-libjpeg-turbo note:  This file contains portions of the libjpeg v6b and v8
-README files, with additional wordsmithing by The libjpeg-turbo Project.
-It is included only for reference, as some parts of it may not apply to
-libjpeg-turbo.  Please see README-turbo.txt for information specific to
-libjpeg-turbo.
-
-
-The Independent JPEG Group's JPEG software
-==========================================
-
-This distribution contains a release of the Independent JPEG Group's free JPEG
-software.  You are welcome to redistribute this software and to use it for any
-purpose, subject to the conditions under LEGAL ISSUES, below.
-
-This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
-Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
-Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
-and other members of the Independent JPEG Group.
-
-IJG is not affiliated with the official ISO JPEG standards committee.
-
-
-DOCUMENTATION ROADMAP
-=====================
-
-This file contains the following sections:
-
-OVERVIEW            General description of JPEG and the IJG software.
-LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
-REFERENCES          Where to learn more about JPEG.
-ARCHIVE LOCATIONS   Where to find newer versions of this software.
-FILE FORMAT WARS    Software *not* to get.
-TO DO               Plans for future IJG releases.
-
-Other documentation files in the distribution are:
-
-User documentation:
-  install.txt       How to configure and install the IJG software.
-  usage.txt         Usage instructions for cjpeg, djpeg, jpegtran,
-                    rdjpgcom, and wrjpgcom.
-  *.1               Unix-style man pages for programs (same info as usage.txt).
-  wizard.txt        Advanced usage instructions for JPEG wizards only.
-  change.log        Version-to-version change highlights.
-Programmer and internal documentation:
-  libjpeg.txt       How to use the JPEG library in your own programs.
-  example.c         Sample code for calling the JPEG library.
-  structure.txt     Overview of the JPEG library's internal structure.
-  filelist.txt      Road map of IJG files.
-  coderules.txt     Coding style rules --- please read if you contribute code.
-
-Please read at least the files install.txt and usage.txt.  Some information
-can also be found in the JPEG FAQ (Frequently Asked Questions) article.  See
-ARCHIVE LOCATIONS below to find out where to obtain the FAQ article.
-
-If you want to understand how the JPEG code works, we suggest reading one or
-more of the REFERENCES, then looking at the documentation files (in roughly
-the order listed) before diving into the code.
-
-
-OVERVIEW
-========
-
-This package contains C software to implement JPEG image encoding, decoding,
-and transcoding.  JPEG (pronounced "jay-peg") is a standardized compression
-method for full-color and gray-scale images.  JPEG's strong suit is compressing
-photographic images or other types of images that have smooth color and
-brightness transitions between neighboring pixels.  Images with sharp lines or
-other abrupt features may not compress well with JPEG, and a higher JPEG
-quality may have to be used to avoid visible compression artifacts with such
-images.
-
-JPEG is lossy, meaning that the output pixels are not necessarily identical to
-the input pixels.  However, on photographic content and other "smooth" images,
-very good compression ratios can be obtained with no visible compression
-artifacts, and extremely high compression ratios are possible if you are
-willing to sacrifice image quality (by reducing the "quality" setting in the
-compressor.)
-
-This software implements JPEG baseline, extended-sequential, and progressive
-compression processes.  Provision is made for supporting all variants of these
-processes, although some uncommon parameter settings aren't implemented yet.
-We have made no provision for supporting the hierarchical or lossless
-processes defined in the standard.
-
-We provide a set of library routines for reading and writing JPEG image files,
-plus two sample applications "cjpeg" and "djpeg", which use the library to
-perform conversion between JPEG and some other popular image file formats.
-The library is intended to be reused in other applications.
-
-In order to support file conversion and viewing software, we have included
-considerable functionality beyond the bare JPEG coding/decoding capability;
-for example, the color quantization modules are not strictly part of JPEG
-decoding, but they are essential for output to colormapped file formats or
-colormapped displays.  These extra functions can be compiled out of the
-library if not required for a particular application.
-
-We have also included "jpegtran", a utility for lossless transcoding between
-different JPEG processes, and "rdjpgcom" and "wrjpgcom", two simple
-applications for inserting and extracting textual comments in JFIF files.
-
-The emphasis in designing this software has been on achieving portability and
-flexibility, while also making it fast enough to be useful.  In particular,
-the software is not intended to be read as a tutorial on JPEG.  (See the
-REFERENCES section for introductory material.)  Rather, it is intended to
-be reliable, portable, industrial-strength code.  We do not claim to have
-achieved that goal in every aspect of the software, but we strive for it.
-
-We welcome the use of this software as a component of commercial products.
-No royalty is required, but we do ask for an acknowledgement in product
-documentation, as described under LEGAL ISSUES.
-
-
-LEGAL ISSUES
-============
-
-In plain English:
-
-1. We don't promise that this software works.  (But if you find any bugs,
-   please let us know!)
-2. You can use this software for whatever you want.  You don't have to pay us.
-3. You may not pretend that you wrote this software.  If you use it in a
-   program, you must acknowledge somewhere in your documentation that
-   you've used the IJG code.
-
-In legalese:
-
-The authors make NO WARRANTY or representation, either express or implied,
-with respect to this software, its quality, accuracy, merchantability, or
-fitness for a particular purpose.  This software is provided "AS IS", and you,
-its user, assume the entire risk as to its quality and accuracy.
-
-This software is copyright (C) 1991-2010, Thomas G. Lane, Guido Vollbeding.
-All Rights Reserved except as specified below.
-
-Permission is hereby granted to use, copy, modify, and distribute this
-software (or portions thereof) for any purpose, without fee, subject to these
-conditions:
-(1) If any part of the source code for this software is distributed, then this
-README file must be included, with this copyright and no-warranty notice
-unaltered; and any additions, deletions, or changes to the original files
-must be clearly indicated in accompanying documentation.
-(2) If only executable code is distributed, then the accompanying
-documentation must state that "this software is based in part on the work of
-the Independent JPEG Group".
-(3) Permission for use of this software is granted only if the user accepts
-full responsibility for any undesirable consequences; the authors accept
-NO LIABILITY for damages of any kind.
-
-These conditions apply to any software derived from or based on the IJG code,
-not just to the unmodified library.  If you use our work, you ought to
-acknowledge us.
-
-Permission is NOT granted for the use of any IJG author's name or company name
-in advertising or publicity relating to this software or products derived from
-it.  This software may be referred to only as "the Independent JPEG Group's
-software".
-
-We specifically permit and encourage the use of this software as the basis of
-commercial products, provided that all warranty or liability claims are
-assumed by the product vendor.
-
-
-ansi2knr.c is included in this distribution by permission of L. Peter Deutsch,
-sole proprietor of its copyright holder, Aladdin Enterprises of Menlo Park, CA.
-ansi2knr.c is NOT covered by the above copyright and conditions, but instead
-by the usual distribution terms of the Free Software Foundation; principally,
-that you must include source code if you redistribute it.  (See the file
-ansi2knr.c for full details.)  However, since ansi2knr.c is not needed as part
-of any program generated from the IJG code, this does not limit you more than
-the foregoing paragraphs do.
-
-The Unix configuration script "configure" was produced with GNU Autoconf.
-It is copyright by the Free Software Foundation but is freely distributable.
-The same holds for its supporting scripts (config.guess, config.sub,
-ltmain.sh).  Another support script, install-sh, is copyright by X Consortium
-but is also freely distributable.
-
-The IJG distribution formerly included code to read and write GIF files.
-To avoid entanglement with the Unisys LZW patent, GIF reading support has
-been removed altogether, and the GIF writer has been simplified to produce
-"uncompressed GIFs".  This technique does not use the LZW algorithm; the
-resulting GIF files are larger than usual, but are readable by all standard
-GIF decoders.
-
-We are required to state that
-    "The Graphics Interchange Format(c) is the Copyright property of
-    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
-    CompuServe Incorporated."
-
-
-REFERENCES
-==========
-
-We recommend reading one or more of these references before trying to
-understand the innards of the JPEG software.
-
-The best short technical introduction to the JPEG compression algorithm is
-	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
-	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
-(Adjacent articles in that issue discuss MPEG motion picture compression,
-applications of JPEG, and related topics.)  If you don't have the CACM issue
-handy, a PostScript file containing a revised version of Wallace's article is
-available at http://www.ijg.org/files/wallace.ps.gz.  The file (actually
-a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
-omits the sample images that appeared in CACM, but it includes corrections
-and some added material.  Note: the Wallace article is copyright ACM and IEEE,
-and it may not be used for commercial purposes.
-
-A somewhat less technical, more leisurely introduction to JPEG can be found in
-"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
-M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
-good explanations and example C code for a multitude of compression methods
-including JPEG.  It is an excellent source if you are comfortable reading C
-code but don't know much about data compression in general.  The book's JPEG
-sample code is far from industrial-strength, but when you are ready to look
-at a full implementation, you've got one here...
-
-The best currently available description of JPEG is the textbook "JPEG Still
-Image Data Compression Standard" by William B. Pennebaker and Joan L.
-Mitchell, published by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
-Price US$59.95, 638 pp.  The book includes the complete text of the ISO JPEG
-standards (DIS 10918-1 and draft DIS 10918-2).
-
-The original JPEG standard is divided into two parts, Part 1 being the actual
-specification, while Part 2 covers compliance testing methods.  Part 1 is
-titled "Digital Compression and Coding of Continuous-tone Still Images,
-Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
-10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
-Continuous-tone Still Images, Part 2: Compliance testing" and has document
-numbers ISO/IEC IS 10918-2, ITU-T T.83.
-
-The JPEG standard does not specify all details of an interchangeable file
-format.  For the omitted details we follow the "JFIF" conventions, revision
-1.02.  JFIF 1.02 has been adopted as an Ecma International Technical Report
-and thus received a formal publication status.  It is available as a free
-download in PDF format from
-http://www.ecma-international.org/publications/techreports/E-TR-098.htm.
-A PostScript version of the JFIF document is available at
-http://www.ijg.org/files/jfif.ps.gz.  There is also a plain text version at
-http://www.ijg.org/files/jfif.txt.gz, but it is missing the figures.
-
-The TIFF 6.0 file format specification can be obtained by FTP from
-ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
-found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
-IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
-Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
-(Compression tag 7).  Copies of this Note can be obtained from
-http://www.ijg.org/files/.  It is expected that the next revision
-of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
-Although IJG's own code does not support TIFF/JPEG, the free libtiff library
-uses our library to implement TIFF/JPEG per the Note.
-
-
-ARCHIVE LOCATIONS
-=================
-
-The "official" archive site for this software is www.ijg.org.
-The most recent released version can always be found there in
-directory "files".  This particular version will be archived as
-http://www.ijg.org/files/jpegsrc.v8d.tar.gz, and in Windows-compatible
-"zip" archive format as http://www.ijg.org/files/jpegsr8d.zip.
-
-The JPEG FAQ (Frequently Asked Questions) article is a source of some
-general information about JPEG.
-It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
-and other news.answers archive sites, including the official news.answers
-archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
-If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
-with body
-	send usenet/news.answers/jpeg-faq/part1
-	send usenet/news.answers/jpeg-faq/part2
-
-
-FILE FORMAT WARS
-================
-
-The ISO JPEG standards committee actually promotes different formats like
-"JPEG 2000" or "JPEG XR", which are incompatible with original DCT-based
-JPEG.  IJG therefore does not support these formats (see REFERENCES).  Indeed,
-one of the original reasons for developing this free software was to help
-force convergence on common, interoperable format standards for JPEG files.
-Don't use an incompatible file format!
-(In any case, our decoder will remain capable of reading existing JPEG
-image files indefinitely.)
-
-
-TO DO
-=====
-
-Please send bug reports, offers of help, etc. to jpeg-info@jpegclub.org.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README.ijg glmark2-2021.02/src/libjpeg-turbo/README.ijg
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README.ijg	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/README.ijg	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,279 @@
+libjpeg-turbo note:  This file has been modified by The libjpeg-turbo Project
+to include only information relevant to libjpeg-turbo, to wordsmith certain
+sections, and to remove impolitic language that existed in the libjpeg v8
+README.  It is included only for reference.  Please see README.md for
+information specific to libjpeg-turbo.
+
+
+The Independent JPEG Group's JPEG software
+==========================================
+
+This distribution contains a release of the Independent JPEG Group's free JPEG
+software.  You are welcome to redistribute this software and to use it for any
+purpose, subject to the conditions under LEGAL ISSUES, below.
+
+This software is the work of Tom Lane, Guido Vollbeding, Philip Gladstone,
+Bill Allombert, Jim Boucher, Lee Crocker, Bob Friesenhahn, Ben Jackson,
+Julian Minguillon, Luis Ortiz, George Phillips, Davide Rossi, Ge' Weijers,
+and other members of the Independent JPEG Group.
+
+IJG is not affiliated with the ISO/IEC JTC1/SC29/WG1 standards committee
+(also known as JPEG, together with ITU-T SG16).
+
+
+DOCUMENTATION ROADMAP
+=====================
+
+This file contains the following sections:
+
+OVERVIEW            General description of JPEG and the IJG software.
+LEGAL ISSUES        Copyright, lack of warranty, terms of distribution.
+REFERENCES          Where to learn more about JPEG.
+ARCHIVE LOCATIONS   Where to find newer versions of this software.
+FILE FORMAT WARS    Software *not* to get.
+TO DO               Plans for future IJG releases.
+
+Other documentation files in the distribution are:
+
+User documentation:
+  usage.txt         Usage instructions for cjpeg, djpeg, jpegtran,
+                    rdjpgcom, and wrjpgcom.
+  *.1               Unix-style man pages for programs (same info as usage.txt).
+  wizard.txt        Advanced usage instructions for JPEG wizards only.
+  change.log        Version-to-version change highlights.
+Programmer and internal documentation:
+  libjpeg.txt       How to use the JPEG library in your own programs.
+  example.c         Sample code for calling the JPEG library.
+  structure.txt     Overview of the JPEG library's internal structure.
+  coderules.txt     Coding style rules --- please read if you contribute code.
+
+Please read at least usage.txt.  Some information can also be found in the JPEG
+FAQ (Frequently Asked Questions) article.  See ARCHIVE LOCATIONS below to find
+out where to obtain the FAQ article.
+
+If you want to understand how the JPEG code works, we suggest reading one or
+more of the REFERENCES, then looking at the documentation files (in roughly
+the order listed) before diving into the code.
+
+
+OVERVIEW
+========
+
+This package contains C software to implement JPEG image encoding, decoding,
+and transcoding.  JPEG (pronounced "jay-peg") is a standardized compression
+method for full-color and grayscale images.  JPEG's strong suit is compressing
+photographic images or other types of images that have smooth color and
+brightness transitions between neighboring pixels.  Images with sharp lines or
+other abrupt features may not compress well with JPEG, and a higher JPEG
+quality may have to be used to avoid visible compression artifacts with such
+images.
+
+JPEG is lossy, meaning that the output pixels are not necessarily identical to
+the input pixels.  However, on photographic content and other "smooth" images,
+very good compression ratios can be obtained with no visible compression
+artifacts, and extremely high compression ratios are possible if you are
+willing to sacrifice image quality (by reducing the "quality" setting in the
+compressor.)
+
+This software implements JPEG baseline, extended-sequential, and progressive
+compression processes.  Provision is made for supporting all variants of these
+processes, although some uncommon parameter settings aren't implemented yet.
+We have made no provision for supporting the hierarchical or lossless
+processes defined in the standard.
+
+We provide a set of library routines for reading and writing JPEG image files,
+plus two sample applications "cjpeg" and "djpeg", which use the library to
+perform conversion between JPEG and some other popular image file formats.
+The library is intended to be reused in other applications.
+
+In order to support file conversion and viewing software, we have included
+considerable functionality beyond the bare JPEG coding/decoding capability;
+for example, the color quantization modules are not strictly part of JPEG
+decoding, but they are essential for output to colormapped file formats or
+colormapped displays.  These extra functions can be compiled out of the
+library if not required for a particular application.
+
+We have also included "jpegtran", a utility for lossless transcoding between
+different JPEG processes, and "rdjpgcom" and "wrjpgcom", two simple
+applications for inserting and extracting textual comments in JFIF files.
+
+The emphasis in designing this software has been on achieving portability and
+flexibility, while also making it fast enough to be useful.  In particular,
+the software is not intended to be read as a tutorial on JPEG.  (See the
+REFERENCES section for introductory material.)  Rather, it is intended to
+be reliable, portable, industrial-strength code.  We do not claim to have
+achieved that goal in every aspect of the software, but we strive for it.
+
+We welcome the use of this software as a component of commercial products.
+No royalty is required, but we do ask for an acknowledgement in product
+documentation, as described under LEGAL ISSUES.
+
+
+LEGAL ISSUES
+============
+
+In plain English:
+
+1. We don't promise that this software works.  (But if you find any bugs,
+   please let us know!)
+2. You can use this software for whatever you want.  You don't have to pay us.
+3. You may not pretend that you wrote this software.  If you use it in a
+   program, you must acknowledge somewhere in your documentation that
+   you've used the IJG code.
+
+In legalese:
+
+The authors make NO WARRANTY or representation, either express or implied,
+with respect to this software, its quality, accuracy, merchantability, or
+fitness for a particular purpose.  This software is provided "AS IS", and you,
+its user, assume the entire risk as to its quality and accuracy.
+
+This software is copyright (C) 1991-2016, Thomas G. Lane, Guido Vollbeding.
+All Rights Reserved except as specified below.
+
+Permission is hereby granted to use, copy, modify, and distribute this
+software (or portions thereof) for any purpose, without fee, subject to these
+conditions:
+(1) If any part of the source code for this software is distributed, then this
+README file must be included, with this copyright and no-warranty notice
+unaltered; and any additions, deletions, or changes to the original files
+must be clearly indicated in accompanying documentation.
+(2) If only executable code is distributed, then the accompanying
+documentation must state that "this software is based in part on the work of
+the Independent JPEG Group".
+(3) Permission for use of this software is granted only if the user accepts
+full responsibility for any undesirable consequences; the authors accept
+NO LIABILITY for damages of any kind.
+
+These conditions apply to any software derived from or based on the IJG code,
+not just to the unmodified library.  If you use our work, you ought to
+acknowledge us.
+
+Permission is NOT granted for the use of any IJG author's name or company name
+in advertising or publicity relating to this software or products derived from
+it.  This software may be referred to only as "the Independent JPEG Group's
+software".
+
+We specifically permit and encourage the use of this software as the basis of
+commercial products, provided that all warranty or liability claims are
+assumed by the product vendor.
+
+
+The Unix configuration script "configure" was produced with GNU Autoconf.
+It is copyright by the Free Software Foundation but is freely distributable.
+The same holds for its supporting scripts (config.guess, config.sub,
+ltmain.sh).  Another support script, install-sh, is copyright by X Consortium
+but is also freely distributable.
+
+The IJG distribution formerly included code to read and write GIF files.
+To avoid entanglement with the Unisys LZW patent (now expired), GIF reading
+support has been removed altogether, and the GIF writer has been simplified
+to produce "uncompressed GIFs".  This technique does not use the LZW
+algorithm; the resulting GIF files are larger than usual, but are readable
+by all standard GIF decoders.
+
+We are required to state that
+    "The Graphics Interchange Format(c) is the Copyright property of
+    CompuServe Incorporated.  GIF(sm) is a Service Mark property of
+    CompuServe Incorporated."
+
+
+REFERENCES
+==========
+
+We recommend reading one or more of these references before trying to
+understand the innards of the JPEG software.
+
+The best short technical introduction to the JPEG compression algorithm is
+	Wallace, Gregory K.  "The JPEG Still Picture Compression Standard",
+	Communications of the ACM, April 1991 (vol. 34 no. 4), pp. 30-44.
+(Adjacent articles in that issue discuss MPEG motion picture compression,
+applications of JPEG, and related topics.)  If you don't have the CACM issue
+handy, a PDF file containing a revised version of Wallace's article is
+available at http://www.ijg.org/files/Wallace.JPEG.pdf.  The file (actually
+a preprint for an article that appeared in IEEE Trans. Consumer Electronics)
+omits the sample images that appeared in CACM, but it includes corrections
+and some added material.  Note: the Wallace article is copyright ACM and IEEE,
+and it may not be used for commercial purposes.
+
+A somewhat less technical, more leisurely introduction to JPEG can be found in
+"The Data Compression Book" by Mark Nelson and Jean-loup Gailly, published by
+M&T Books (New York), 2nd ed. 1996, ISBN 1-55851-434-1.  This book provides
+good explanations and example C code for a multitude of compression methods
+including JPEG.  It is an excellent source if you are comfortable reading C
+code but don't know much about data compression in general.  The book's JPEG
+sample code is far from industrial-strength, but when you are ready to look
+at a full implementation, you've got one here...
+
+The best currently available description of JPEG is the textbook "JPEG Still
+Image Data Compression Standard" by William B. Pennebaker and Joan L.
+Mitchell, published by Van Nostrand Reinhold, 1993, ISBN 0-442-01272-1.
+Price US$59.95, 638 pp.  The book includes the complete text of the ISO JPEG
+standards (DIS 10918-1 and draft DIS 10918-2).
+
+The original JPEG standard is divided into two parts, Part 1 being the actual
+specification, while Part 2 covers compliance testing methods.  Part 1 is
+titled "Digital Compression and Coding of Continuous-tone Still Images,
+Part 1: Requirements and guidelines" and has document numbers ISO/IEC IS
+10918-1, ITU-T T.81.  Part 2 is titled "Digital Compression and Coding of
+Continuous-tone Still Images, Part 2: Compliance testing" and has document
+numbers ISO/IEC IS 10918-2, ITU-T T.83.
+
+The JPEG standard does not specify all details of an interchangeable file
+format.  For the omitted details we follow the "JFIF" conventions, revision
+1.02.  JFIF 1.02 has been adopted as an Ecma International Technical Report
+and thus received a formal publication status.  It is available as a free
+download in PDF format from
+http://www.ecma-international.org/publications/techreports/E-TR-098.htm.
+A PostScript version of the JFIF document is available at
+http://www.ijg.org/files/jfif.ps.gz.  There is also a plain text version at
+http://www.ijg.org/files/jfif.txt.gz, but it is missing the figures.
+
+The TIFF 6.0 file format specification can be obtained by FTP from
+ftp://ftp.sgi.com/graphics/tiff/TIFF6.ps.gz.  The JPEG incorporation scheme
+found in the TIFF 6.0 spec of 3-June-92 has a number of serious problems.
+IJG does not recommend use of the TIFF 6.0 design (TIFF Compression tag 6).
+Instead, we recommend the JPEG design proposed by TIFF Technical Note #2
+(Compression tag 7).  Copies of this Note can be obtained from
+http://www.ijg.org/files/.  It is expected that the next revision
+of the TIFF spec will replace the 6.0 JPEG design with the Note's design.
+Although IJG's own code does not support TIFF/JPEG, the free libtiff library
+uses our library to implement TIFF/JPEG per the Note.
+
+
+ARCHIVE LOCATIONS
+=================
+
+The "official" archive site for this software is www.ijg.org.
+The most recent released version can always be found there in
+directory "files".
+
+The JPEG FAQ (Frequently Asked Questions) article is a source of some
+general information about JPEG.
+It is available on the World Wide Web at http://www.faqs.org/faqs/jpeg-faq/
+and other news.answers archive sites, including the official news.answers
+archive at rtfm.mit.edu: ftp://rtfm.mit.edu/pub/usenet/news.answers/jpeg-faq/.
+If you don't have Web or FTP access, send e-mail to mail-server@rtfm.mit.edu
+with body
+	send usenet/news.answers/jpeg-faq/part1
+	send usenet/news.answers/jpeg-faq/part2
+
+
+FILE FORMAT WARS
+================
+
+The ISO/IEC JTC1/SC29/WG1 standards committee (also known as JPEG, together
+with ITU-T SG16) currently promotes different formats containing the name
+"JPEG" which are incompatible with original DCT-based JPEG.  IJG therefore does
+not support these formats (see REFERENCES).  Indeed, one of the original
+reasons for developing this free software was to help force convergence on
+common, interoperable format standards for JPEG files.
+Don't use an incompatible file format!
+(In any case, our decoder will remain capable of reading existing JPEG
+image files indefinitely.)
+
+
+TO DO
+=====
+
+Please send bug reports, offers of help, etc. to jpeg-info@jpegclub.org.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README.md glmark2-2021.02/src/libjpeg-turbo/README.md
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README.md	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/README.md	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,335 @@
+Background
+==========
+
+libjpeg-turbo is a JPEG image codec that uses SIMD instructions (MMX, SSE2,
+NEON, AltiVec) to accelerate baseline JPEG compression and decompression on
+x86, x86-64, ARM, and PowerPC systems.  On such systems, libjpeg-turbo is
+generally 2-6x as fast as libjpeg, all else being equal.  On other types of
+systems, libjpeg-turbo can still outperform libjpeg by a significant amount, by
+virtue of its highly-optimized Huffman coding routines.  In many cases, the
+performance of libjpeg-turbo rivals that of proprietary high-speed JPEG codecs.
+
+libjpeg-turbo implements both the traditional libjpeg API as well as the less
+powerful but more straightforward TurboJPEG API.  libjpeg-turbo also features
+colorspace extensions that allow it to compress from/decompress to 32-bit and
+big-endian pixel buffers (RGBX, XBGR, etc.), as well as a full-featured Java
+interface.
+
+libjpeg-turbo was originally based on libjpeg/SIMD, an MMX-accelerated
+derivative of libjpeg v6b developed by Miyasaka Masaru.  The TigerVNC and
+VirtualGL projects made numerous enhancements to the codec in 2009, and in
+early 2010, libjpeg-turbo spun off into an independent project, with the goal
+of making high-speed JPEG compression/decompression technology available to a
+broader range of users and developers.
+
+
+License
+=======
+
+libjpeg-turbo is covered by three compatible BSD-style open source licenses.
+Refer to [LICENSE.md](LICENSE.md) for a roll-up of license terms.
+
+
+Using libjpeg-turbo
+===================
+
+libjpeg-turbo includes two APIs that can be used to compress and decompress
+JPEG images:
+
+- **TurboJPEG API**
+  This API provides an easy-to-use interface for compressing and decompressing
+  JPEG images in memory.  It also provides some functionality that would not be
+  straightforward to achieve using the underlying libjpeg API, such as
+  generating planar YUV images and performing multiple simultaneous lossless
+  transforms on an image.  The Java interface for libjpeg-turbo is written on
+  top of the TurboJPEG API.
+
+- **libjpeg API**
+  This is the de facto industry-standard API for compressing and decompressing
+  JPEG images.  It is more difficult to use than the TurboJPEG API but also
+  more powerful.  The libjpeg API implementation in libjpeg-turbo is both
+  API/ABI-compatible and mathematically compatible with libjpeg v6b.  It can
+  also optionally be configured to be API/ABI-compatible with libjpeg v7 and v8
+  (see below.)
+
+There is no significant performance advantage to either API when both are used
+to perform similar operations.
+
+Colorspace Extensions
+---------------------
+
+libjpeg-turbo includes extensions that allow JPEG images to be compressed
+directly from (and decompressed directly to) buffers that use BGR, BGRX,
+RGBX, XBGR, and XRGB pixel ordering.  This is implemented with ten new
+colorspace constants:
+
+    JCS_EXT_RGB   /* red/green/blue */
+    JCS_EXT_RGBX  /* red/green/blue/x */
+    JCS_EXT_BGR   /* blue/green/red */
+    JCS_EXT_BGRX  /* blue/green/red/x */
+    JCS_EXT_XBGR  /* x/blue/green/red */
+    JCS_EXT_XRGB  /* x/red/green/blue */
+    JCS_EXT_RGBA  /* red/green/blue/alpha */
+    JCS_EXT_BGRA  /* blue/green/red/alpha */
+    JCS_EXT_ABGR  /* alpha/blue/green/red */
+    JCS_EXT_ARGB  /* alpha/red/green/blue */
+
+Setting `cinfo.in_color_space` (compression) or `cinfo.out_color_space`
+(decompression) to one of these values will cause libjpeg-turbo to read the
+red, green, and blue values from (or write them to) the appropriate position in
+the pixel when compressing from/decompressing to an RGB buffer.
+
+Your application can check for the existence of these extensions at compile
+time with:
+
+    #ifdef JCS_EXTENSIONS
+
+At run time, attempting to use these extensions with a libjpeg implementation
+that does not support them will result in a "Bogus input colorspace" error.
+Applications can trap this error in order to test whether run-time support is
+available for the colorspace extensions.
+
+When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
+X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
+can set that byte to whatever value it wishes.  If an application expects the X
+byte to be used as an alpha channel, then it should specify `JCS_EXT_RGBA`,
+`JCS_EXT_BGRA`, `JCS_EXT_ABGR`, or `JCS_EXT_ARGB`.  When these colorspace
+constants are used, the X byte is guaranteed to be 0xFF, which is interpreted
+as opaque.
+
+Your application can check for the existence of the alpha channel colorspace
+extensions at compile time with:
+
+    #ifdef JCS_ALPHA_EXTENSIONS
+
+[jcstest.c](jcstest.c), located in the libjpeg-turbo source tree, demonstrates
+how to check for the existence of the colorspace extensions at compile time and
+run time.
+
+libjpeg v7 and v8 API/ABI Emulation
+-----------------------------------
+
+With libjpeg v7 and v8, new features were added that necessitated extending the
+compression and decompression structures.  Unfortunately, due to the exposed
+nature of those structures, extending them also necessitated breaking backward
+ABI compatibility with previous libjpeg releases.  Thus, programs that were
+built to use libjpeg v7 or v8 did not work with libjpeg-turbo, since it is
+based on the libjpeg v6b code base.  Although libjpeg v7 and v8 are not
+as widely used as v6b, enough programs (including a few Linux distros) made
+the switch that there was a demand to emulate the libjpeg v7 and v8 ABIs
+in libjpeg-turbo.  It should be noted, however, that this feature was added
+primarily so that applications that had already been compiled to use libjpeg
+v7+ could take advantage of accelerated baseline JPEG encoding/decoding
+without recompiling.  libjpeg-turbo does not claim to support all of the
+libjpeg v7+ features, nor to produce identical output to libjpeg v7+ in all
+cases (see below.)
+
+By passing an argument of `--with-jpeg7` or `--with-jpeg8` to `configure`, or
+an argument of `-DWITH_JPEG7=1` or `-DWITH_JPEG8=1` to `cmake`, you can build a
+version of libjpeg-turbo that emulates the libjpeg v7 or v8 ABI, so that
+programs that are built against libjpeg v7 or v8 can be run with libjpeg-turbo.
+The following section describes which libjpeg v7+ features are supported and
+which aren't.
+
+### Support for libjpeg v7 and v8 Features
+
+#### Fully supported
+
+- **libjpeg: IDCT scaling extensions in decompressor**
+  libjpeg-turbo supports IDCT scaling with scaling factors of 1/8, 1/4, 3/8,
+  1/2, 5/8, 3/4, 7/8, 9/8, 5/4, 11/8, 3/2, 13/8, 7/4, 15/8, and 2/1 (only 1/4
+  and 1/2 are SIMD-accelerated.)
+
+- **libjpeg: Arithmetic coding**
+
+- **libjpeg: In-memory source and destination managers**
+  See notes below.
+
+- **cjpeg: Separate quality settings for luminance and chrominance**
+  Note that the libpjeg v7+ API was extended to accommodate this feature only
+  for convenience purposes.  It has always been possible to implement this
+  feature with libjpeg v6b (see rdswitch.c for an example.)
+
+- **cjpeg: 32-bit BMP support**
+
+- **cjpeg: `-rgb` option**
+
+- **jpegtran: Lossless cropping**
+
+- **jpegtran: `-perfect` option**
+
+- **jpegtran: Forcing width/height when performing lossless crop**
+
+- **rdjpgcom: `-raw` option**
+
+- **rdjpgcom: Locale awareness**
+
+
+#### Not supported
+
+NOTE:  As of this writing, extensive research has been conducted into the
+usefulness of DCT scaling as a means of data reduction and SmartScale as a
+means of quality improvement.  The reader is invited to peruse the research at
+http://www.libjpeg-turbo.org/About/SmartScale and draw his/her own conclusions,
+but it is the general belief of our project that these features have not
+demonstrated sufficient usefulness to justify inclusion in libjpeg-turbo.
+
+- **libjpeg: DCT scaling in compressor**
+  `cinfo.scale_num` and `cinfo.scale_denom` are silently ignored.
+  There is no technical reason why DCT scaling could not be supported when
+  emulating the libjpeg v7+ API/ABI, but without the SmartScale extension (see
+  below), only scaling factors of 1/2, 8/15, 4/7, 8/13, 2/3, 8/11, 4/5, and
+  8/9 would be available, which is of limited usefulness.
+
+- **libjpeg: SmartScale**
+  `cinfo.block_size` is silently ignored.
+  SmartScale is an extension to the JPEG format that allows for DCT block
+  sizes other than 8x8.  Providing support for this new format would be
+  feasible (particularly without full acceleration.)  However, until/unless
+  the format becomes either an official industry standard or, at minimum, an
+  accepted solution in the community, we are hesitant to implement it, as
+  there is no sense of whether or how it might change in the future.  It is
+  our belief that SmartScale has not demonstrated sufficient usefulness as a
+  lossless format nor as a means of quality enhancement, and thus our primary
+  interest in providing this feature would be as a means of supporting
+  additional DCT scaling factors.
+
+- **libjpeg: Fancy downsampling in compressor**
+  `cinfo.do_fancy_downsampling` is silently ignored.
+  This requires the DCT scaling feature, which is not supported.
+
+- **jpegtran: Scaling**
+  This requires both the DCT scaling and SmartScale features, which are not
+  supported.
+
+- **Lossless RGB JPEG files**
+  This requires the SmartScale feature, which is not supported.
+
+### What About libjpeg v9?
+
+libjpeg v9 introduced yet another field to the JPEG compression structure
+(`color_transform`), thus making the ABI backward incompatible with that of
+libjpeg v8.  This new field was introduced solely for the purpose of supporting
+lossless SmartScale encoding.  Furthermore, there was actually no reason to
+extend the API in this manner, as the color transform could have just as easily
+been activated by way of a new JPEG colorspace constant, thus preserving
+backward ABI compatibility.
+
+Our research (see link above) has shown that lossless SmartScale does not
+generally accomplish anything that can't already be accomplished better with
+existing, standard lossless formats.  Therefore, at this time it is our belief
+that there is not sufficient technical justification for software projects to
+upgrade from libjpeg v8 to libjpeg v9, and thus there is not sufficient
+echnical justification for us to emulate the libjpeg v9 ABI.
+
+In-Memory Source/Destination Managers
+-------------------------------------
+
+By default, libjpeg-turbo 1.3 and later includes the `jpeg_mem_src()` and
+`jpeg_mem_dest()` functions, even when not emulating the libjpeg v8 API/ABI.
+Previously, it was necessary to build libjpeg-turbo from source with libjpeg v8
+API/ABI emulation in order to use the in-memory source/destination managers,
+but several projects requested that those functions be included when emulating
+the libjpeg v6b API/ABI as well.  This allows the use of those functions by
+programs that need them, without breaking ABI compatibility for programs that
+don't, and it allows those functions to be provided in the "official"
+libjpeg-turbo binaries.
+
+Those who are concerned about maintaining strict conformance with the libjpeg
+v6b or v7 API can pass an argument of `--without-mem-srcdst` to `configure` or
+an argument of `-DWITH_MEM_SRCDST=0` to `cmake` prior to building
+libjpeg-turbo.  This will restore the pre-1.3 behavior, in which
+`jpeg_mem_src()` and `jpeg_mem_dest()` are only included when emulating the
+libjpeg v8 API/ABI.
+
+On Un*x systems, including the in-memory source/destination managers changes
+the dynamic library version from 62.0.0 to 62.1.0 if using libjpeg v6b API/ABI
+emulation and from 7.0.0 to 7.1.0 if using libjpeg v7 API/ABI emulation.
+
+Note that, on most Un*x systems, the dynamic linker will not look for a
+function in a library until that function is actually used.  Thus, if a program
+is built against libjpeg-turbo 1.3+ and uses `jpeg_mem_src()` or
+`jpeg_mem_dest()`, that program will not fail if run against an older version
+of libjpeg-turbo or against libjpeg v7- until the program actually tries to
+call `jpeg_mem_src()` or `jpeg_mem_dest()`.  Such is not the case on Windows.
+If a program is built against the libjpeg-turbo 1.3+ DLL and uses
+`jpeg_mem_src()` or `jpeg_mem_dest()`, then it must use the libjpeg-turbo 1.3+
+DLL at run time.
+
+Both cjpeg and djpeg have been extended to allow testing the in-memory
+source/destination manager functions.  See their respective man pages for more
+details.
+
+
+Mathematical Compatibility
+==========================
+
+For the most part, libjpeg-turbo should produce identical output to libjpeg
+v6b.  The one exception to this is when using the floating point DCT/IDCT, in
+which case the outputs of libjpeg v6b and libjpeg-turbo can differ for the
+following reasons:
+
+- The SSE/SSE2 floating point DCT implementation in libjpeg-turbo is ever so
+  slightly more accurate than the implementation in libjpeg v6b, but not by
+  any amount perceptible to human vision (generally in the range of 0.01 to
+  0.08 dB gain in PNSR.)
+
+- When not using the SIMD extensions, libjpeg-turbo uses the more accurate
+  (and slightly faster) floating point IDCT algorithm introduced in libjpeg
+  v8a as opposed to the algorithm used in libjpeg v6b.  It should be noted,
+  however, that this algorithm basically brings the accuracy of the floating
+  point IDCT in line with the accuracy of the slow integer IDCT.  The floating
+  point DCT/IDCT algorithms are mainly a legacy feature, and they do not
+  produce significantly more accuracy than the slow integer algorithms (to put
+  numbers on this, the typical difference in PNSR between the two algorithms
+  is less than 0.10 dB, whereas changing the quality level by 1 in the upper
+  range of the quality scale is typically more like a 1.0 dB difference.)
+
+- If the floating point algorithms in libjpeg-turbo are not implemented using
+  SIMD instructions on a particular platform, then the accuracy of the
+  floating point DCT/IDCT can depend on the compiler settings.
+
+While libjpeg-turbo does emulate the libjpeg v8 API/ABI, under the hood it is
+still using the same algorithms as libjpeg v6b, so there are several specific
+cases in which libjpeg-turbo cannot be expected to produce the same output as
+libjpeg v8:
+
+- When decompressing using scaling factors of 1/2 and 1/4, because libjpeg v8
+  implements those scaling algorithms differently than libjpeg v6b does, and
+  libjpeg-turbo's SIMD extensions are based on the libjpeg v6b behavior.
+
+- When using chrominance subsampling, because libjpeg v8 implements this
+  with its DCT/IDCT scaling algorithms rather than with a separate
+  downsampling/upsampling algorithm.  In our testing, the subsampled/upsampled
+  output of libjpeg v8 is less accurate than that of libjpeg v6b for this
+  reason.
+
+- When decompressing using a scaling factor > 1 and merged (AKA "non-fancy" or
+  "non-smooth") chrominance upsampling, because libjpeg v8 does not support
+  merged upsampling with scaling factors > 1.
+
+
+Performance Pitfalls
+====================
+
+Restart Markers
+---------------
+
+The optimized Huffman decoder in libjpeg-turbo does not handle restart markers
+in a way that makes the rest of the libjpeg infrastructure happy, so it is
+necessary to use the slow Huffman decoder when decompressing a JPEG image that
+has restart markers.  This can cause the decompression performance to drop by
+as much as 20%, but the performance will still be much greater than that of
+libjpeg.  Many consumer packages, such as PhotoShop, use restart markers when
+generating JPEG images, so images generated by those programs will experience
+this issue.
+
+Fast Integer Forward DCT at High Quality Levels
+-----------------------------------------------
+
+The algorithm used by the SIMD-accelerated quantization function cannot produce
+correct results whenever the fast integer forward DCT is used along with a JPEG
+quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
+function in those cases.  This causes performance to drop by as much as 40%.
+It is therefore strongly advised that you use the slow integer forward DCT
+whenever encoding images with a JPEG quality of 98 or higher.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README-turbo.txt glmark2-2021.02/src/libjpeg-turbo/README-turbo.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/README-turbo.txt	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/README-turbo.txt	1970-01-01 08:00:00.000000000 +0800
@@ -1,361 +0,0 @@
-*******************************************************************************
-**     Background
-*******************************************************************************
-
-libjpeg-turbo is a derivative of libjpeg that uses SIMD instructions (MMX,
-SSE2, NEON) to accelerate baseline JPEG compression and decompression on x86,
-x86-64, and ARM systems.  On such systems, libjpeg-turbo is generally 2-4x as
-fast as the unmodified version of libjpeg, all else being equal.
-
-libjpeg-turbo was originally based on libjpeg/SIMD by Miyasaka Masaru, but
-the TigerVNC and VirtualGL projects made numerous enhancements to the codec in
-2009, including improved support for Mac OS X, 64-bit support, support for
-32-bit and big-endian pixel formats (RGBX, XBGR, etc.), accelerated Huffman
-encoding/decoding, and various bug fixes.  The goal was to produce a fully
-open-source codec that could replace the partially closed-source TurboJPEG/IPP
-codec used by VirtualGL and TurboVNC.  libjpeg-turbo generally achieves 80-120%
-of the performance of TurboJPEG/IPP.  It is faster in some areas but slower in
-others.
-
-In early 2010, libjpeg-turbo spun off into its own independent project, with
-the goal of making high-speed JPEG compression/decompression technology
-available to a broader range of users and developers.
-
-
-*******************************************************************************
-**     License
-*******************************************************************************
-
-Most of libjpeg-turbo inherits the non-restrictive, BSD-style license used by
-libjpeg (see README.)  The TurboJPEG/OSS wrapper (both C and Java versions) and
-associated test programs bear a similar license, which is reproduced below:
-
-Redistribution and use in source and binary forms, with or without
-modification, are permitted provided that the following conditions are met:
-
-- Redistributions of source code must retain the above copyright notice,
-  this list of conditions and the following disclaimer.
-- Redistributions in binary form must reproduce the above copyright notice,
-  this list of conditions and the following disclaimer in the documentation
-  and/or other materials provided with the distribution.
-- Neither the name of the libjpeg-turbo Project nor the names of its
-  contributors may be used to endorse or promote products derived from this
-  software without specific prior written permission.
-
-THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
-AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
-IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
-ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
-LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
-CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
-SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
-INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
-CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
-ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
-POSSIBILITY OF SUCH DAMAGE.
-
-
-*******************************************************************************
-**     Using libjpeg-turbo
-*******************************************************************************
-
-libjpeg-turbo includes two APIs that can be used to compress and decompress
-JPEG images:
-
-  TurboJPEG API:  This API provides an easy-to-use interface for compressing
-  and decompressing JPEG images in memory.  It also provides some functionality
-  that would not be straightforward to achieve using the underlying libjpeg
-  API, such as generating planar YUV images and performing multiple
-  simultaneous lossless transforms on an image.  The Java interface for
-  libjpeg-turbo is written on top of the TurboJPEG API.
-
-  libjpeg API:  This is the de facto industry-standard API for compressing and
-  decompressing JPEG images.  It is more difficult to use than the TurboJPEG
-  API but also more powerful.  libjpeg-turbo is both API/ABI-compatible and
-  mathematically compatible with libjpeg v6b.  It can also optionally be
-  configured to be API/ABI-compatible with libjpeg v7 and v8 (see below.)
-
-
-=============================
-Replacing libjpeg at Run Time
-=============================
-
-If a Unix application is dynamically linked with libjpeg, then you can replace
-libjpeg with libjpeg-turbo at run time by manipulating LD_LIBRARY_PATH.
-For instance:
-
-  [Using libjpeg]
-  > time cjpeg <vgl_5674_0098.ppm >vgl_5674_0098.jpg
-  real  0m0.392s
-  user  0m0.074s
-  sys   0m0.020s
-
-  [Using libjpeg-turbo]
-  > export LD_LIBRARY_PATH=/opt/libjpeg-turbo/{lib}:$LD_LIBRARY_PATH
-  > time cjpeg <vgl_5674_0098.ppm >vgl_5674_0098.jpg
-  real  0m0.109s
-  user  0m0.029s
-  sys   0m0.010s
-
-NOTE: {lib} can be lib, lib32, lib64, or lib/64, depending on the O/S and
-architecture.
-
-System administrators can also replace the libjpeg sym links in /usr/{lib} with
-links to the libjpeg-turbo dynamic library located in /opt/libjpeg-turbo/{lib}.
-This will effectively accelerate every application that uses the libjpeg
-dynamic library on the system.
-
-The libjpeg-turbo SDK for Visual C++ installs the libjpeg-turbo DLL
-(jpeg62.dll, jpeg7.dll, or jpeg8.dll, depending on whether it was built with
-libjpeg v6b, v7, or v8 emulation) into c:\libjpeg-turbo[64]\bin, and the PATH
-environment variable can be modified such that this directory is searched
-before any others that might contain a libjpeg DLL.  However, if a libjpeg
-DLL exists in an application's install directory, then Windows will load this
-DLL first whenever the application is launched.  Thus, if an application ships
-with jpeg62.dll, jpeg7.dll, or jpeg8.dll, then back up the application's
-version of this DLL and copy c:\libjpeg-turbo[64]\bin\jpeg*.dll into the
-application's install directory to accelerate it.
-
-The version of the libjpeg-turbo DLL distributed in the libjpeg-turbo SDK for
-Visual C++ requires the Visual C++ 2008 C run-time DLL (msvcr90.dll).
-msvcr90.dll ships with more recent versions of Windows, but users of older
-Windows releases can obtain it from the Visual C++ 2008 Redistributable
-Package, which is available as a free download from Microsoft's web site.
-
-NOTE:  Features of libjpeg that require passing a C run-time structure, such
-as a file handle, from an application to libjpeg will probably not work with
-the version of the libjpeg-turbo DLL distributed in the libjpeg-turbo SDK for
-Visual C++, unless the application is also built to use the Visual C++ 2008 C
-run-time DLL.  In particular, this affects jpeg_stdio_dest() and
-jpeg_stdio_src().
-
-Mac applications typically embed their own copies of the libjpeg dylib inside
-the (hidden) application bundle, so it is not possible to globally replace
-libjpeg on OS X systems.  If an application uses a shared library version of
-libjpeg, then it may be possible to replace the application's version of it.
-This would generally involve copying libjpeg.*.dylib from libjpeg-turbo into
-the appropriate place in the application bundle and using install_name_tool to
-repoint the dylib to the new directory.  This requires an advanced knowledge of
-OS X and would not survive an upgrade or a re-install of the application.
-Thus, it is not recommended for most users.
-
-=======================
-Replacing TurboJPEG/IPP
-=======================
-
-libjpeg-turbo is a drop-in replacement for the TurboJPEG/IPP SDK used by
-VirtualGL 2.1.x and TurboVNC 0.6 (and prior.)  libjpeg-turbo contains a wrapper
-library (TurboJPEG/OSS) that emulates the TurboJPEG API using libjpeg-turbo
-instead of the closed-source Intel Performance Primitives.  You can replace the
-TurboJPEG/IPP package on Linux systems with the libjpeg-turbo package in order
-to make existing releases of VirtualGL 2.1.x and TurboVNC 0.x use the new codec
-at run time.  Note that the 64-bit libjpeg-turbo packages contain only 64-bit
-binaries, whereas the TurboJPEG/IPP 64-bit packages contained both 64-bit and
-32-bit binaries.  Thus, to replace a TurboJPEG/IPP 64-bit package, install
-both the 64-bit and 32-bit versions of libjpeg-turbo.
-
-You can also build the VirtualGL 2.1.x and TurboVNC 0.6 source code with
-the libjpeg-turbo SDK instead of TurboJPEG/IPP.  It should work identically.
-libjpeg-turbo also includes static library versions of TurboJPEG/OSS, which
-are used to build VirtualGL 2.2 and TurboVNC 1.0 and later.
-
-========================================
-Using libjpeg-turbo in Your Own Programs
-========================================
-
-For the most part, libjpeg-turbo should work identically to libjpeg, so in
-most cases, an application can be built against libjpeg and then run against
-libjpeg-turbo.  On Unix systems (including Cygwin), you can build against
-libjpeg-turbo instead of libjpeg by setting
-
-  CPATH=/opt/libjpeg-turbo/include
-  and
-  LIBRARY_PATH=/opt/libjpeg-turbo/{lib}
-
-({lib} = lib32 or lib64, depending on whether you are building a 32-bit or a
-64-bit application.)
-
-If using MinGW, then set
-
-  CPATH=/c/libjpeg-turbo-gcc[64]/include
-  and
-  LIBRARY_PATH=/c/libjpeg-turbo-gcc[64]/lib
-
-Building against libjpeg-turbo is useful, for instance, if you want to build an
-application that leverages the libjpeg-turbo colorspace extensions (see below.)
-On Linux and Solaris systems, you would still need to manipulate
-LD_LIBRARY_PATH or create appropriate sym links to use libjpeg-turbo at run
-time.  On such systems, you can pass -R /opt/libjpeg-turbo/{lib} to the linker
-to force the use of libjpeg-turbo at run time rather than libjpeg (also useful
-if you want to leverage the colorspace extensions), or you can link against the
-libjpeg-turbo static library.
-
-To force a Linux, Solaris, or MinGW application to link against the static
-version of libjpeg-turbo, you can use the following linker options:
-
-  -Wl,-Bstatic -ljpeg -Wl,-Bdynamic
-
-On OS X, simply add /opt/libjpeg-turbo/lib/libjpeg.a to the linker command
-line (this also works on Linux and Solaris.)
-
-To build Visual C++ applications using libjpeg-turbo, add
-c:\libjpeg-turbo[64]\include to the system or user INCLUDE environment
-variable and c:\libjpeg-turbo[64]\lib to the system or user LIB environment
-variable, and then link against either jpeg.lib (to use the DLL version of
-libjpeg-turbo) or jpeg-static.lib (to use the static version of libjpeg-turbo.)
-
-=====================
-Colorspace Extensions
-=====================
-
-libjpeg-turbo includes extensions that allow JPEG images to be compressed
-directly from (and decompressed directly to) buffers that use BGR, BGRX,
-RGBX, XBGR, and XRGB pixel ordering.  This is implemented with ten new
-colorspace constants:
-
-  JCS_EXT_RGB   /* red/green/blue */
-  JCS_EXT_RGBX  /* red/green/blue/x */
-  JCS_EXT_BGR   /* blue/green/red */
-  JCS_EXT_BGRX  /* blue/green/red/x */
-  JCS_EXT_XBGR  /* x/blue/green/red */
-  JCS_EXT_XRGB  /* x/red/green/blue */
-  JCS_EXT_RGBA  /* red/green/blue/alpha */
-  JCS_EXT_BGRA  /* blue/green/red/alpha */
-  JCS_EXT_ABGR  /* alpha/blue/green/red */
-  JCS_EXT_ARGB  /* alpha/red/green/blue */
-
-Setting cinfo.in_color_space (compression) or cinfo.out_color_space
-(decompression) to one of these values will cause libjpeg-turbo to read the
-red, green, and blue values from (or write them to) the appropriate position in
-the pixel when compressing from/decompressing to an RGB buffer.
-
-Your application can check for the existence of these extensions at compile
-time with:
-
-  #ifdef JCS_EXTENSIONS
-
-At run time, attempting to use these extensions with a version of libjpeg
-that doesn't support them will result in a "Bogus input colorspace" error.
-
-When using the RGBX, BGRX, XBGR, and XRGB colorspaces during decompression, the
-X byte is undefined, and in order to ensure the best performance, libjpeg-turbo
-can set that byte to whatever value it wishes.  If an application expects the X
-byte to be used as an alpha channel, then it should specify JCS_EXT_RGBA,
-JCS_EXT_BGRA, JCS_EXT_ABGR, or JCS_EXT_ARGB.  When these colorspace constants
-are used, the X byte is guaranteed to be 0xFF, which is interpreted as opaque.
-
-Your application can check for the existence of the alpha channel colorspace
-extensions at compile time with:
-
-  #ifdef JCS_ALPHA_EXTENSIONS
-
-jcstest.c, located in the libjpeg-turbo source tree, demonstrates how to check
-for the existence of the colorspace extensions at compile time and run time.
-
-=================================
-libjpeg v7 and v8 API/ABI support
-=================================
-
-With libjpeg v7 and v8, new features were added that necessitated extending the
-compression and decompression structures.  Unfortunately, due to the exposed
-nature of those structures, extending them also necessitated breaking backward
-ABI compatibility with previous libjpeg releases.  Thus, programs that are
-built to use libjpeg v7 or v8 did not work with libjpeg-turbo, since it is
-based on the libjpeg v6b code base.  Although libjpeg v7 and v8 are still not
-as widely used as v6b, enough programs (including a few Linux distros) have
-made the switch that it was desirable to provide support for the libjpeg v7/v8
-API/ABI in libjpeg-turbo.  Although libjpeg-turbo can now be configured as a
-drop-in replacement for libjpeg v7 or v8, it should be noted that not all of
-the features in libjpeg v7 and v8 are supported (see below.)
-
-By passing an argument of --with-jpeg7 or --with-jpeg8 to configure, or an
-argument of -DWITH_JPEG7=1 or -DWITH_JPEG8=1 to cmake, you can build a version
-of libjpeg-turbo that emulates the libjpeg v7 or v8 API/ABI, so that programs
-that are built against libjpeg v7 or v8 can be run with libjpeg-turbo.  The
-following section describes which libjpeg v7+ features are supported and which
-aren't.
-
-libjpeg v7 and v8 Features:
----------------------------
-
-Fully supported:
-
--- cjpeg: Separate quality settings for luminance and chrominance
-   Note that the libpjeg v7+ API was extended to accommodate this feature only
-   for convenience purposes.  It has always been possible to implement this
-   feature with libjpeg v6b (see rdswitch.c for an example.)
-
--- cjpeg: 32-bit BMP support
-
--- jpegtran: lossless cropping
-
--- jpegtran: -perfect option
-
--- rdjpgcom: -raw option
-
--- rdjpgcom: locale awareness
-
-
-Fully supported when using libjpeg v7/v8 emulation:
-
--- libjpeg: In-memory source and destination managers
-
-
-Not supported:
-
--- libjpeg: DCT scaling in compressor
-   cinfo.scale_num and cinfo.scale_denom are silently ignored.
-   There is no technical reason why DCT scaling cannot be supported, but
-   without the SmartScale extension (see below), it would only be able to
-   down-scale using ratios of 1/2, 8/15, 4/7, 8/13, 2/3, 8/11, 4/5, and 8/9,
-   which is of limited usefulness.
-
--- libjpeg: SmartScale
-   cinfo.block_size is silently ignored.
-   SmartScale is an extension to the JPEG format that allows for DCT block
-   sizes other than 8x8.  It would be difficult to support this feature while
-   retaining backward compatibility with libjpeg v6b.
-
--- libjpeg: IDCT scaling extensions in decompressor
-   libjpeg-turbo still supports IDCT scaling with scaling factors of 1/2, 1/4,
-   and 1/8 (same as libjpeg v6b.)
-
--- libjpeg: Fancy downsampling in compressor
-   cinfo.do_fancy_downsampling is silently ignored.
-   This requires the DCT scaling feature, which is not supported.
-
--- jpegtran: Scaling
-   This requires both the DCT scaling and SmartScale features, which are not
-   supported.
-
--- Lossless RGB JPEG files
-   This requires the SmartScale feature, which is not supported.
-
-
-*******************************************************************************
-**     Performance pitfalls
-*******************************************************************************
-
-===============
-Restart Markers
-===============
-
-The optimized Huffman decoder in libjpeg-turbo does not handle restart markers
-in a way that makes the rest of the libjpeg infrastructure happy, so it is
-necessary to use the slow Huffman decoder when decompressing a JPEG image that
-has restart markers.  This can cause the decompression performance to drop by
-as much as 20%, but the performance will still be much greater than that of
-libjpeg.  Many consumer packages, such as PhotoShop, use restart markers when
-generating JPEG images, so images generated by those programs will experience
-this issue.
-
-===============================================
-Fast Integer Forward DCT at High Quality Levels
-===============================================
-
-The algorithm used by the SIMD-accelerated quantization function cannot produce
-correct results whenever the fast integer forward DCT is used along with a JPEG
-quality of 98-100.  Thus, libjpeg-turbo must use the non-SIMD quantization
-function in those cases.  This causes performance to drop by as much as 40%.
-It is therefore strongly advised that you use the slow integer forward DCT
-whenever encoding images with a JPEG quality of 98 or higher.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/CMakeLists.txt glmark2-2021.02/src/libjpeg-turbo/simd/CMakeLists.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/CMakeLists.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/CMakeLists.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,80 @@
+if(NOT DEFINED NASM)
+  set(NASM nasm CACHE FILEPATH "Path to NASM/YASM executable")
+endif()
+
+if(SIMD_X86_64)
+  set(NAFLAGS -fwin64 -DWIN64 -D__x86_64__)
+else()
+  if(BORLAND)
+    set(NAFLAGS -fobj -DOBJ32)
+  else()
+    set(NAFLAGS -fwin32 -DWIN32)
+  endif()
+endif()
+set(NAFLAGS ${NAFLAGS} -I${CMAKE_SOURCE_DIR}/win/ -I${CMAKE_CURRENT_SOURCE_DIR}/)
+
+# This only works if building from the command line.  There is currently no way
+# to set a variable's value based on the build type when using the MSVC IDE.
+if(CMAKE_BUILD_TYPE STREQUAL "Debug"
+  OR CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
+  set(NAFLAGS ${NAFLAGS} -g)
+endif()
+
+if(SIMD_X86_64)
+  set(SIMD_BASENAMES jfdctflt-sse-64 jccolor-sse2-64 jcgray-sse2-64
+    jchuff-sse2-64 jcsample-sse2-64 jdcolor-sse2-64 jdmerge-sse2-64
+    jdsample-sse2-64 jfdctfst-sse2-64 jfdctint-sse2-64 jidctflt-sse2-64
+    jidctfst-sse2-64 jidctint-sse2-64 jidctred-sse2-64 jquantf-sse2-64
+    jquanti-sse2-64)
+  message(STATUS "Building x86_64 SIMD extensions")
+else()
+  set(SIMD_BASENAMES jsimdcpu jfdctflt-3dn jidctflt-3dn jquant-3dn jccolor-mmx
+    jcgray-mmx jcsample-mmx jdcolor-mmx jdmerge-mmx jdsample-mmx jfdctfst-mmx
+    jfdctint-mmx jidctfst-mmx jidctint-mmx jidctred-mmx jquant-mmx jfdctflt-sse
+    jidctflt-sse jquant-sse jccolor-sse2 jcgray-sse2 jchuff-sse2 jcsample-sse2
+    jdcolor-sse2 jdmerge-sse2 jdsample-sse2 jfdctfst-sse2 jfdctint-sse2
+    jidctflt-sse2 jidctfst-sse2 jidctint-sse2 jidctred-sse2 jquantf-sse2
+    jquanti-sse2)
+  message(STATUS "Building i386 SIMD extensions")
+endif()
+
+if(MSVC_IDE)
+  set(OBJDIR "${CMAKE_CURRENT_BINARY_DIR}/${CMAKE_CFG_INTDIR}")
+else()
+  set(OBJDIR ${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
+file(GLOB INC_FILES *.inc)
+
+foreach(file ${SIMD_BASENAMES})
+  set(DEPFILE "")
+  set(SIMD_SRC ${CMAKE_CURRENT_SOURCE_DIR}/${file}.asm)
+  if(${file} MATCHES jccolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jccolor" "jccolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jcgray)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jcgray" "jcgryext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdcolor)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdcolor" "jdcolext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  if(${file} MATCHES jdmerge)
+    set(DEPFILE ${file})
+    string(REGEX REPLACE "jdmerge" "jdmrgext" DEPFILE ${DEPFILE})
+    set(DEPFILE ${CMAKE_CURRENT_SOURCE_DIR}/${DEPFILE}.asm)
+  endif()
+  set(SIMD_OBJ ${OBJDIR}/${file}.obj)
+  add_custom_command(OUTPUT ${SIMD_OBJ}
+    DEPENDS ${SIMD_SRC} ${DEPFILE} ${INC_FILES}
+    COMMAND ${NASM} ${NAFLAGS} ${SIMD_SRC} -o${SIMD_OBJ})
+  set(SIMD_OBJS ${SIMD_OBJS} ${SIMD_OBJ})
+endforeach()
+
+set(SIMD_OBJS ${SIMD_OBJS} PARENT_SCOPE)
+add_custom_target(simd DEPENDS ${SIMD_OBJS})
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,267 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * Copyright (C) 2014, Jay Foad.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jccolor-altivec.c */
+
+
+void jsimd_rgb_ycc_convert_altivec (JDIMENSION img_width, JSAMPARRAY input_buf,
+                                    JSAMPIMAGE output_buf,
+                                    JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr0, outptr1, outptr2;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1 = {0}, rgb2 = {0},
+    rgbg0, rgbg1, rgbg2, rgbg3, y, cb, cr;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = {0};
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = {0};
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh, crl, crh, cbl, cbh;
+  __vector int y0, y1, y2, y3, cr0, cr1, cr2, cr3, cb0, cb1, cb2, cb3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) },
+    pw_mf016_mf033 = { __4X2(-F_0_168, -F_0_331) },
+    pw_mf008_mf041 = { __4X2(-F_0_081, -F_0_418) };
+  __vector unsigned short pw_f050_f000 = { __4X2(F_0_500, 0) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) },
+    pd_onehalfm1_cj = { __4X(ONE_HALF - 1 + (CENTERJSAMPLE << SCALEBITS)) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr0 += 16, outptr1 += 16, outptr2 += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = VEC_LD(0, tmpbuf);
+          rgb1 = VEC_LD(16, tmpbuf);
+          rgb2 = VEC_LD(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = VEC_LD(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = VEC_LD(0, inptr);
+          if (num_cols > 16)
+            rgb1 = VEC_LD(16, inptr);
+          if (num_cols > 32)
+            rgb2 = VEC_LD(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = VEC_LD(48, inptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       * Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+       * Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr0);
+
+      /* Calculate Cb values */
+      cb0 = vec_msums(rg0, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb1 = vec_msums(rg1, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb2 = vec_msums(rg2, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb3 = vec_msums(rg3, pw_mf016_mf033, pd_onehalfm1_cj);
+      cb0 = (__vector int)vec_msum((__vector unsigned short)bg0, pw_f050_f000,
+                                   (__vector unsigned int)cb0);
+      cb1 = (__vector int)vec_msum((__vector unsigned short)bg1, pw_f050_f000,
+                                   (__vector unsigned int)cb1);
+      cb2 = (__vector int)vec_msum((__vector unsigned short)bg2, pw_f050_f000,
+                                   (__vector unsigned int)cb2);
+      cb3 = (__vector int)vec_msum((__vector unsigned short)bg3, pw_f050_f000,
+                                   (__vector unsigned int)cb3);
+      cbl = vec_perm((__vector unsigned short)cb0,
+                     (__vector unsigned short)cb1, shift_pack_index);
+      cbh = vec_perm((__vector unsigned short)cb2,
+                     (__vector unsigned short)cb3, shift_pack_index);
+      cb = vec_pack(cbl, cbh);
+      vec_st(cb, 0, outptr1);
+
+      /* Calculate Cr values */
+      cr0 = vec_msums(bg0, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr1 = vec_msums(bg1, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr2 = vec_msums(bg2, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr3 = vec_msums(bg3, pw_mf008_mf041, pd_onehalfm1_cj);
+      cr0 = (__vector int)vec_msum((__vector unsigned short)rg0, pw_f050_f000,
+                                   (__vector unsigned int)cr0);
+      cr1 = (__vector int)vec_msum((__vector unsigned short)rg1, pw_f050_f000,
+                                   (__vector unsigned int)cr1);
+      cr2 = (__vector int)vec_msum((__vector unsigned short)rg2, pw_f050_f000,
+                                   (__vector unsigned int)cr2);
+      cr3 = (__vector int)vec_msum((__vector unsigned short)rg3, pw_f050_f000,
+                                   (__vector unsigned int)cr3);
+      crl = vec_perm((__vector unsigned short)cr0,
+                     (__vector unsigned short)cr1, shift_pack_index);
+      crh = vec_perm((__vector unsigned short)cr2,
+                     (__vector unsigned short)cr3, shift_pack_index);
+      cr = vec_pack(crl, crh);
+      vec_st(cr, 0, outptr2);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,477 @@
+;
+; jccolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_mmx (JDIMENSION img_width,
+;                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                           JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          8
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_rgb_ycc_convert_mmx) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edx
+        push    ebx
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        mov     ebx, JSAMPROW [ebx]     ; outptr1
+        mov     edx, JSAMPROW [edx]     ; outptr2
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        xor     eax,eax
+        mov     al, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        xor     edx,edx
+        mov     dx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    mmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    mmG, DWORD [esi+ecx]
+        psllq   mmA, DWORD_BIT
+        por     mmA,mmG
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        movq    mmG,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        mov     ecx, SIZEOF_MMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld16:
+        test    cl, 2*SIZEOF_MMWORD
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_ycc_cnv
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+2*SIZEOF_MMWORD]
+
+.rgb_ycc_cnv:
+        ; mmA=(00 10 20 01 11 21 02 12)
+        ; mmG=(22 03 13 23 04 14 24 05)
+        ; mmF=(15 25 06 16 26 07 17 27)
+
+        movq      mmD,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 10 20 01)
+        psrlq     mmD,4*BYTE_BIT        ; mmD=(11 21 02 12 -- -- -- --)
+
+        punpckhbw mmA,mmG               ; mmA=(00 04 10 14 20 24 01 05)
+        psllq     mmG,4*BYTE_BIT        ; mmG=(-- -- -- -- 22 03 13 23)
+
+        punpcklbw mmD,mmF               ; mmD=(11 15 21 25 02 06 12 16)
+        punpckhbw mmG,mmF               ; mmG=(22 26 03 07 13 17 23 27)
+
+        movq      mmE,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 04 10 14)
+        psrlq     mmE,4*BYTE_BIT        ; mmE=(20 24 01 05 -- -- -- --)
+
+        punpckhbw mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        psllq     mmD,4*BYTE_BIT        ; mmD=(-- -- -- -- 11 15 21 25)
+
+        punpcklbw mmE,mmG               ; mmE=(20 22 24 26 01 03 05 07)
+        punpckhbw mmD,mmG               ; mmD=(11 13 15 17 21 23 25 27)
+
+        pxor      mmH,mmH
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmH               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmH               ; mmC=(10 12 14 16)
+
+        movq      mmB,mmE
+        punpcklbw mmE,mmH               ; mmE=(20 22 24 26)
+        punpckhbw mmB,mmH               ; mmB=(01 03 05 07)
+
+        movq      mmF,mmD
+        punpcklbw mmD,mmH               ; mmD=(11 13 15 17)
+        punpckhbw mmF,mmH               ; mmF=(21 23 25 27)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_MMWORD/8
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_MMWORD/8
+        movd    mmA, DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_MMWORD/4
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_MMWORD/4
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld4:
+        test    cl, SIZEOF_MMWORD/2
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_ycc_cnv
+        movq    mmD,mmA
+        movq    mmC,mmF
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmD, MMWORD [esi+2*SIZEOF_MMWORD]
+        movq    mmC, MMWORD [esi+3*SIZEOF_MMWORD]
+
+.rgb_ycc_cnv:
+        ; mmA=(00 10 20 30 01 11 21 31)
+        ; mmF=(02 12 22 32 03 13 23 33)
+        ; mmD=(04 14 24 34 05 15 25 35)
+        ; mmC=(06 16 26 36 07 17 27 37)
+
+        movq      mmB,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 10 12 20 22 30 32)
+        punpckhbw mmB,mmF               ; mmB=(01 03 11 13 21 23 31 33)
+
+        movq      mmG,mmD
+        punpcklbw mmD,mmC               ; mmD=(04 06 14 16 24 26 34 36)
+        punpckhbw mmG,mmC               ; mmG=(05 07 15 17 25 27 35 37)
+
+        movq      mmE,mmA
+        punpcklwd mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        punpckhwd mmE,mmD               ; mmE=(20 22 24 26 30 32 34 36)
+
+        movq      mmH,mmB
+        punpcklwd mmB,mmG               ; mmB=(01 03 05 07 11 13 15 17)
+        punpckhwd mmH,mmG               ; mmH=(21 23 25 27 31 33 35 37)
+
+        pxor      mmF,mmF
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmF               ; mmC=(10 12 14 16)
+
+        movq      mmD,mmB
+        punpcklbw mmB,mmF               ; mmB=(01 03 05 07)
+        punpckhbw mmD,mmF               ; mmD=(11 13 15 17)
+
+        movq      mmG,mmE
+        punpcklbw mmE,mmF               ; mmE=(20 22 24 26)
+        punpckhbw mmG,mmF               ; mmG=(30 32 34 36)
+
+        punpcklbw mmF,mmH
+        punpckhbw mmH,mmH
+        psrlw     mmF,BYTE_BIT          ; mmF=(21 23 25 27)
+        psrlw     mmH,BYTE_BIT          ; mmH=(31 33 35 37)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+        ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movq      MMWORD [wk(0)], mm0   ; wk(0)=RE
+        movq      MMWORD [wk(1)], mm1   ; wk(1)=RO
+        movq      MMWORD [wk(2)], mm4   ; wk(2)=BE
+        movq      MMWORD [wk(3)], mm5   ; wk(3)=BO
+
+        movq      mm6,mm1
+        punpcklwd mm1,mm3
+        punpckhwd mm6,mm3
+        movq      mm7,mm1
+        movq      mm4,mm6
+        pmaddwd   mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF016_MF033)] ; mm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movq      MMWORD [wk(4)], mm1   ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movq      MMWORD [wk(5)], mm6   ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      mm1,mm1
+        pxor      mm6,mm6
+        punpcklwd mm1,mm5               ; mm1=BOL
+        punpckhwd mm6,mm5               ; mm6=BOH
+        psrld     mm1,1                 ; mm1=BOL*FIX(0.500)
+        psrld     mm6,1                 ; mm6=BOH*FIX(0.500)
+
+        movq      mm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm5=[PD_ONEHALFM1_CJ]
+
+        paddd     mm7,mm1
+        paddd     mm4,mm6
+        paddd     mm7,mm5
+        paddd     mm4,mm5
+        psrld     mm7,SCALEBITS         ; mm7=CbOL
+        psrld     mm4,SCALEBITS         ; mm4=CbOH
+        packssdw  mm7,mm4               ; mm7=CbO
+
+        movq      mm1, MMWORD [wk(2)]   ; mm1=BE
+
+        movq      mm6,mm0
+        punpcklwd mm0,mm2
+        punpckhwd mm6,mm2
+        movq      mm5,mm0
+        movq      mm4,mm6
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF016_MF033)] ; mm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF016_MF033)] ; mm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movq      MMWORD [wk(6)], mm0   ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movq      MMWORD [wk(7)], mm6   ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      mm0,mm0
+        pxor      mm6,mm6
+        punpcklwd mm0,mm1               ; mm0=BEL
+        punpckhwd mm6,mm1               ; mm6=BEH
+        psrld     mm0,1                 ; mm0=BEL*FIX(0.500)
+        psrld     mm6,1                 ; mm6=BEH*FIX(0.500)
+
+        movq      mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ]
+
+        paddd     mm5,mm0
+        paddd     mm4,mm6
+        paddd     mm5,mm1
+        paddd     mm4,mm1
+        psrld     mm5,SCALEBITS         ; mm5=CbEL
+        psrld     mm4,SCALEBITS         ; mm4=CbEH
+        packssdw  mm5,mm4               ; mm5=CbE
+
+        psllw     mm7,BYTE_BIT
+        por       mm5,mm7               ; mm5=Cb
+        movq      MMWORD [ebx], mm5     ; Save Cb
+
+        movq      mm0, MMWORD [wk(3)]   ; mm0=BO
+        movq      mm6, MMWORD [wk(2)]   ; mm6=BE
+        movq      mm1, MMWORD [wk(1)]   ; mm1=RO
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm3
+        punpckhwd mm4,mm3
+        movq      mm7,mm0
+        movq      mm5,mm4
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF008_MF041)] ; mm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movq      mm3,[GOTOFF(eax,PD_ONEHALF)]  ; mm3=[PD_ONEHALF]
+
+        paddd     mm0, MMWORD [wk(4)]
+        paddd     mm4, MMWORD [wk(5)]
+        paddd     mm0,mm3
+        paddd     mm4,mm3
+        psrld     mm0,SCALEBITS         ; mm0=YOL
+        psrld     mm4,SCALEBITS         ; mm4=YOH
+        packssdw  mm0,mm4               ; mm0=YO
+
+        pxor      mm3,mm3
+        pxor      mm4,mm4
+        punpcklwd mm3,mm1               ; mm3=ROL
+        punpckhwd mm4,mm1               ; mm4=ROH
+        psrld     mm3,1                 ; mm3=ROL*FIX(0.500)
+        psrld     mm4,1                 ; mm4=ROH*FIX(0.500)
+
+        movq      mm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm1=[PD_ONEHALFM1_CJ]
+
+        paddd     mm7,mm3
+        paddd     mm5,mm4
+        paddd     mm7,mm1
+        paddd     mm5,mm1
+        psrld     mm7,SCALEBITS         ; mm7=CrOL
+        psrld     mm5,SCALEBITS         ; mm5=CrOH
+        packssdw  mm7,mm5               ; mm7=CrO
+
+        movq      mm3, MMWORD [wk(0)]   ; mm3=RE
+
+        movq      mm4,mm6
+        punpcklwd mm6,mm2
+        punpckhwd mm4,mm2
+        movq      mm1,mm6
+        movq      mm5,mm4
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   mm1,[GOTOFF(eax,PW_MF008_MF041)] ; mm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF008_MF041)] ; mm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movq      mm2,[GOTOFF(eax,PD_ONEHALF)]  ; mm2=[PD_ONEHALF]
+
+        paddd     mm6, MMWORD [wk(6)]
+        paddd     mm4, MMWORD [wk(7)]
+        paddd     mm6,mm2
+        paddd     mm4,mm2
+        psrld     mm6,SCALEBITS         ; mm6=YEL
+        psrld     mm4,SCALEBITS         ; mm4=YEH
+        packssdw  mm6,mm4               ; mm6=YE
+
+        psllw     mm0,BYTE_BIT
+        por       mm6,mm0               ; mm6=Y
+        movq      MMWORD [edi], mm6     ; Save Y
+
+        pxor      mm2,mm2
+        pxor      mm4,mm4
+        punpcklwd mm2,mm3               ; mm2=REL
+        punpckhwd mm4,mm3               ; mm4=REH
+        psrld     mm2,1                 ; mm2=REL*FIX(0.500)
+        psrld     mm4,1                 ; mm4=REH*FIX(0.500)
+
+        movq      mm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; mm0=[PD_ONEHALFM1_CJ]
+
+        paddd     mm1,mm2
+        paddd     mm5,mm4
+        paddd     mm1,mm0
+        paddd     mm5,mm0
+        psrld     mm1,SCALEBITS         ; mm1=CrEL
+        psrld     mm5,SCALEBITS         ; mm5=CrEH
+        packssdw  mm1,mm5               ; mm1=CrE
+
+        psllw     mm7,BYTE_BIT
+        por       mm1,mm7               ; mm1=Cr
+        movq      MMWORD [edx], mm1     ; Save Cr
+
+        sub     ecx, byte SIZEOF_MMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; inptr
+        add     edi, byte SIZEOF_MMWORD                 ; outptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; outptr1
+        add     edx, byte SIZEOF_MMWORD                 ; outptr2
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        pop     ebx
+        pop     edx
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,485 @@
+;
+; jccolext.asm - colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+; r10 = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13 = JDIMENSION output_row
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          8
+
+        align   16
+
+        global  EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov rsi, r12
+        mov ecx, r13d
+        mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rsi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rsi+2*SIZEOF_JSAMPARRAY]
+        lea     rdi, [rdi+rcx*SIZEOF_JSAMPROW]
+        lea     rbx, [rbx+rcx*SIZEOF_JSAMPROW]
+        lea     rdx, [rdx+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov rsi, r11
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rdx
+        push    rbx
+        push    rdi
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr0
+        mov     rbx, JSAMPROW [rbx]     ; outptr1
+        mov     rdx, JSAMPROW [rdx]     ; outptr2
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    rax
+        push    rdx
+        lea     rcx,[rcx+rcx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_BYTE
+        movzx   rax, BYTE [rsi+rcx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_WORD
+        movzx   rdx, WORD [rsi+rcx]
+        shl     rax, WORD_BIT
+        or      rax,rdx
+.column_ld4:
+        movd    xmmA,eax
+        pop     rdx
+        pop     rax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     rcx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=RE
+        movdqa    XMMWORD [wk(1)], xmm1 ; wk(1)=RO
+        movdqa    XMMWORD [wk(2)], xmm4 ; wk(2)=BE
+        movdqa    XMMWORD [wk(3)], xmm5 ; wk(3)=BO
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        movdqa    xmm7,xmm1
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   xmm7,[rel PW_MF016_MF033] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   xmm4,[rel PW_MF016_MF033] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movdqa    XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      xmm1,xmm1
+        pxor      xmm6,xmm6
+        punpcklwd xmm1,xmm5             ; xmm1=BOL
+        punpckhwd xmm6,xmm5             ; xmm6=BOH
+        psrld     xmm1,1                ; xmm1=BOL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BOH*FIX(0.500)
+
+        movdqa    xmm5,[rel PD_ONEHALFM1_CJ] ; xmm5=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm1
+        paddd     xmm4,xmm6
+        paddd     xmm7,xmm5
+        paddd     xmm4,xmm5
+        psrld     xmm7,SCALEBITS        ; xmm7=CbOL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbOH
+        packssdw  xmm7,xmm4             ; xmm7=CbO
+
+        movdqa    xmm1, XMMWORD [wk(2)] ; xmm1=BE
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   xmm5,[rel PW_MF016_MF033] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   xmm4,[rel PW_MF016_MF033] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      xmm0,xmm0
+        pxor      xmm6,xmm6
+        punpcklwd xmm0,xmm1             ; xmm0=BEL
+        punpckhwd xmm6,xmm1             ; xmm6=BEH
+        psrld     xmm0,1                ; xmm0=BEL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BEH*FIX(0.500)
+
+        movdqa    xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm5,xmm0
+        paddd     xmm4,xmm6
+        paddd     xmm5,xmm1
+        paddd     xmm4,xmm1
+        psrld     xmm5,SCALEBITS        ; xmm5=CbEL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbEH
+        packssdw  xmm5,xmm4             ; xmm5=CbE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm5,xmm7             ; xmm5=Cb
+        movdqa    XMMWORD [rbx], xmm5   ; Save Cb
+
+        movdqa    xmm0, XMMWORD [wk(3)] ; xmm0=BO
+        movdqa    xmm6, XMMWORD [wk(2)] ; xmm6=BE
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=RO
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        movdqa    xmm7,xmm0
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   xmm7,[rel PW_MF008_MF041] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   xmm5,[rel PW_MF008_MF041] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movdqa    xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, XMMWORD [wk(4)]
+        paddd     xmm4, XMMWORD [wk(5)]
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        pxor      xmm3,xmm3
+        pxor      xmm4,xmm4
+        punpcklwd xmm3,xmm1             ; xmm3=ROL
+        punpckhwd xmm4,xmm1             ; xmm4=ROH
+        psrld     xmm3,1                ; xmm3=ROL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=ROH*FIX(0.500)
+
+        movdqa    xmm1,[rel PD_ONEHALFM1_CJ] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm3
+        paddd     xmm5,xmm4
+        paddd     xmm7,xmm1
+        paddd     xmm5,xmm1
+        psrld     xmm7,SCALEBITS        ; xmm7=CrOL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrOH
+        packssdw  xmm7,xmm5             ; xmm7=CrO
+
+        movdqa    xmm3, XMMWORD [wk(0)] ; xmm3=RE
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   xmm1,[rel PW_MF008_MF041] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   xmm5,[rel PW_MF008_MF041] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movdqa    xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(6)]
+        paddd     xmm4, XMMWORD [wk(7)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [rdi], xmm6   ; Save Y
+
+        pxor      xmm2,xmm2
+        pxor      xmm4,xmm4
+        punpcklwd xmm2,xmm3             ; xmm2=REL
+        punpckhwd xmm4,xmm3             ; xmm4=REH
+        psrld     xmm2,1                ; xmm2=REL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=REH*FIX(0.500)
+
+        movdqa    xmm0,[rel PD_ONEHALFM1_CJ] ; xmm0=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm1,xmm2
+        paddd     xmm5,xmm4
+        paddd     xmm1,xmm0
+        paddd     xmm5,xmm0
+        psrld     xmm1,SCALEBITS        ; xmm1=CrEL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrEH
+        packssdw  xmm1,xmm5             ; xmm1=CrE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm1,xmm7             ; xmm1=Cr
+        movdqa    XMMWORD [rdx], xmm1   ; Save Cr
+
+        sub     rcx, byte SIZEOF_XMMWORD
+        add     rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     rdi, byte SIZEOF_XMMWORD                ; outptr0
+        add     rbx, byte SIZEOF_XMMWORD                ; outptr1
+        add     rdx, byte SIZEOF_XMMWORD                ; outptr2
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .column_ld1
+
+        pop     rcx                     ; col
+        pop     rsi
+        pop     rdi
+        pop     rbx
+        pop     rdx
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     rdi, byte SIZEOF_JSAMPROW
+        add     rbx, byte SIZEOF_JSAMPROW
+        add     rdx, byte SIZEOF_JSAMPROW
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,503 @@
+;
+; jccolext.asm - colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_ycc_convert_sse2 (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          8
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+
+        global  EXTN(jsimd_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_ycc_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [esi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [esi+2*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edx
+        push    ebx
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        mov     ebx, JSAMPROW [ebx]     ; outptr1
+        mov     edx, JSAMPROW [edx]     ; outptr2
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        movzx   eax, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        movzx   edx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    xmmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     ecx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .rgb_ycc_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_ycc_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_ycc_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD]
+
+.rgb_ycc_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+        ; Cb = -0.16874 * R - 0.33126 * G + 0.50000 * B + CENTERJSAMPLE
+        ; Cr =  0.50000 * R - 0.41869 * G - 0.08131 * B + CENTERJSAMPLE
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=RE
+        movdqa    XMMWORD [wk(1)], xmm1 ; wk(1)=RO
+        movdqa    XMMWORD [wk(2)], xmm4 ; wk(2)=BE
+        movdqa    XMMWORD [wk(3)], xmm5 ; wk(3)=BO
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        movdqa    xmm7,xmm1
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF016_MF033)] ; xmm7=ROL*-FIX(0.168)+GOL*-FIX(0.331)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=ROH*-FIX(0.168)+GOH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(4)], xmm1 ; wk(4)=ROL*FIX(0.299)+GOL*FIX(0.337)
+        movdqa    XMMWORD [wk(5)], xmm6 ; wk(5)=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        pxor      xmm1,xmm1
+        pxor      xmm6,xmm6
+        punpcklwd xmm1,xmm5             ; xmm1=BOL
+        punpckhwd xmm6,xmm5             ; xmm6=BOH
+        psrld     xmm1,1                ; xmm1=BOL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BOH*FIX(0.500)
+
+        movdqa    xmm5,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm5=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm1
+        paddd     xmm4,xmm6
+        paddd     xmm7,xmm5
+        paddd     xmm4,xmm5
+        psrld     xmm7,SCALEBITS        ; xmm7=CbOL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbOH
+        packssdw  xmm7,xmm4             ; xmm7=CbO
+
+        movdqa    xmm1, XMMWORD [wk(2)] ; xmm1=BE
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm6
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF016_MF033)] ; xmm5=REL*-FIX(0.168)+GEL*-FIX(0.331)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF016_MF033)] ; xmm4=REH*-FIX(0.168)+GEH*-FIX(0.331)
+
+        movdqa    XMMWORD [wk(6)], xmm0 ; wk(6)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(7)], xmm6 ; wk(7)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        pxor      xmm0,xmm0
+        pxor      xmm6,xmm6
+        punpcklwd xmm0,xmm1             ; xmm0=BEL
+        punpckhwd xmm6,xmm1             ; xmm6=BEH
+        psrld     xmm0,1                ; xmm0=BEL*FIX(0.500)
+        psrld     xmm6,1                ; xmm6=BEH*FIX(0.500)
+
+        movdqa    xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm5,xmm0
+        paddd     xmm4,xmm6
+        paddd     xmm5,xmm1
+        paddd     xmm4,xmm1
+        psrld     xmm5,SCALEBITS        ; xmm5=CbEL
+        psrld     xmm4,SCALEBITS        ; xmm4=CbEH
+        packssdw  xmm5,xmm4             ; xmm5=CbE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm5,xmm7             ; xmm5=Cb
+        movdqa    XMMWORD [ebx], xmm5   ; Save Cb
+
+        movdqa    xmm0, XMMWORD [wk(3)] ; xmm0=BO
+        movdqa    xmm6, XMMWORD [wk(2)] ; xmm6=BE
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=RO
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        movdqa    xmm7,xmm0
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF008_MF041)] ; xmm7=BOL*-FIX(0.081)+GOL*-FIX(0.418)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BOH*-FIX(0.081)+GOH*-FIX(0.418)
+
+        movdqa    xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, XMMWORD [wk(4)]
+        paddd     xmm4, XMMWORD [wk(5)]
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        pxor      xmm3,xmm3
+        pxor      xmm4,xmm4
+        punpcklwd xmm3,xmm1             ; xmm3=ROL
+        punpckhwd xmm4,xmm1             ; xmm4=ROH
+        psrld     xmm3,1                ; xmm3=ROL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=ROH*FIX(0.500)
+
+        movdqa    xmm1,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm1=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm7,xmm3
+        paddd     xmm5,xmm4
+        paddd     xmm7,xmm1
+        paddd     xmm5,xmm1
+        psrld     xmm7,SCALEBITS        ; xmm7=CrOL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrOH
+        packssdw  xmm7,xmm5             ; xmm7=CrO
+
+        movdqa    xmm3, XMMWORD [wk(0)] ; xmm3=RE
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm5,xmm4
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+        pmaddwd   xmm1,[GOTOFF(eax,PW_MF008_MF041)] ; xmm1=BEL*-FIX(0.081)+GEL*-FIX(0.418)
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF008_MF041)] ; xmm5=BEH*-FIX(0.081)+GEH*-FIX(0.418)
+
+        movdqa    xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(6)]
+        paddd     xmm4, XMMWORD [wk(7)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [edi], xmm6   ; Save Y
+
+        pxor      xmm2,xmm2
+        pxor      xmm4,xmm4
+        punpcklwd xmm2,xmm3             ; xmm2=REL
+        punpckhwd xmm4,xmm3             ; xmm4=REH
+        psrld     xmm2,1                ; xmm2=REL*FIX(0.500)
+        psrld     xmm4,1                ; xmm4=REH*FIX(0.500)
+
+        movdqa    xmm0,[GOTOFF(eax,PD_ONEHALFM1_CJ)] ; xmm0=[PD_ONEHALFM1_CJ]
+
+        paddd     xmm1,xmm2
+        paddd     xmm5,xmm4
+        paddd     xmm1,xmm0
+        paddd     xmm5,xmm0
+        psrld     xmm1,SCALEBITS        ; xmm1=CrEL
+        psrld     xmm5,SCALEBITS        ; xmm5=CrEH
+        packssdw  xmm1,xmm5             ; xmm1=CrE
+
+        psllw     xmm7,BYTE_BIT
+        por       xmm1,xmm7             ; xmm1=Cr
+        movdqa    XMMWORD [edx], xmm1   ; Save Cr
+
+        sub     ecx, byte SIZEOF_XMMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     edi, byte SIZEOF_XMMWORD                ; outptr0
+        add     ebx, byte SIZEOF_XMMWORD                ; outptr1
+        add     edx, byte SIZEOF_XMMWORD                ; outptr2
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        pop     ebx
+        pop     edx
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,104 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> YCC CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_081 5329                 /* FIX(0.08131) */
+#define F_0_114 7471                 /* FIX(0.11400) */
+#define F_0_168 11059                /* FIX(0.16874) */
+#define F_0_250 16384                /* FIX(0.25000) */
+#define F_0_299 19595                /* FIX(0.29900) */
+#define F_0_331 21709                /* FIX(0.33126) */
+#define F_0_418 27439                /* FIX(0.41869) */
+#define F_0_500 32768                /* FIX(0.50000) */
+#define F_0_587 38470                /* FIX(0.58700) */
+#define F_0_337 (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 {0,1,3,4,6,7,9,10,2,1,5,4,8,7,11,10}
+#define RGBG_INDEX1 {12,13,15,16,18,19,21,22,14,13,17,16,20,19,23,22}
+#define RGBG_INDEX2 {8,9,11,12,14,15,17,18,10,9,13,12,16,15,19,18}
+#define RGBG_INDEX3 {4,5,7,8,10,11,13,14,6,5,9,8,12,11,15,14}
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_rgb_ycc_convert_altivec jsimd_extrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX {0,1,4,5,8,9,12,13,2,1,6,5,10,9,14,13}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extrgbx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 {2,1,5,4,8,7,11,10,0,1,3,4,6,7,9,10}
+#define RGBG_INDEX1 {14,13,17,16,20,19,23,22,12,13,15,16,18,19,21,22}
+#define RGBG_INDEX2 {10,9,13,12,16,15,19,18,8,9,11,12,14,15,17,18}
+#define RGBG_INDEX3 {6,5,9,8,12,11,15,14,4,5,7,8,10,11,13,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX {2,1,6,5,10,9,14,13,0,1,4,5,8,9,12,13}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extbgrx_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX {3,2,7,6,11,10,15,14,1,2,5,6,9,10,13,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extxbgr_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX {1,2,5,6,9,10,13,14,3,2,7,6,11,10,15,14}
+#define jsimd_rgb_ycc_convert_altivec jsimd_extxrgb_ycc_convert_altivec
+#include "jccolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_ycc_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,123 @@
+;
+; jccolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_mmx) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_mmx):
+
+PW_F0299_F0337  times 2 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 2 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 2 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 2 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 2 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 2 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extrgbx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extbgrx_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extxbgr_ycc_convert_mmx
+%include "jccolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_mmx jsimd_extxrgb_ycc_convert_mmx
+%include "jccolext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jccolor-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jccolor-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jccolor.asm - colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2009, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_081 equ      5329                   ; FIX(0.08131)
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_168 equ     11059                   ; FIX(0.16874)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_331 equ     21709                   ; FIX(0.33126)
+F_0_418 equ     27439                   ; FIX(0.41869)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_ycc_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_ycc_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PW_MF016_MF033  times 4 dw -F_0_168,-F_0_331
+PW_MF008_MF041  times 4 dw -F_0_081,-F_0_418
+PD_ONEHALFM1_CJ times 4 dd  (1 << (SCALEBITS-1)) - 1 + (CENTERJSAMPLE << SCALEBITS)
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
+%include "jccolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
+%include "jccolext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,99 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* RGB --> GRAYSCALE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_114 7471                 /* FIX(0.11400) */
+#define F_0_250 16384                /* FIX(0.25000) */
+#define F_0_299 19595                /* FIX(0.29900) */
+#define F_0_587 38470                /* FIX(0.58700) */
+#define F_0_337 (F_0_587 - F_0_250)  /* FIX(0.58700) - FIX(0.25000) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+
+#define RGBG_INDEX0 {0,1,3,4,6,7,9,10,2,1,5,4,8,7,11,10}
+#define RGBG_INDEX1 {12,13,15,16,18,19,21,22,14,13,17,16,20,19,23,22}
+#define RGBG_INDEX2 {8,9,11,12,14,15,17,18,10,9,13,12,16,15,19,18}
+#define RGBG_INDEX3 {4,5,7,8,10,11,13,14,6,5,9,8,12,11,15,14}
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_rgb_gray_convert_altivec jsimd_extrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGBG_INDEX {0,1,4,5,8,9,12,13,2,1,6,5,10,9,14,13}
+#define jsimd_rgb_gray_convert_altivec jsimd_extrgbx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGBG_INDEX0 {2,1,5,4,8,7,11,10,0,1,3,4,6,7,9,10}
+#define RGBG_INDEX1 {14,13,17,16,20,19,23,22,12,13,15,16,18,19,21,22}
+#define RGBG_INDEX2 {10,9,13,12,16,15,19,18,8,9,11,12,14,15,17,18}
+#define RGBG_INDEX3 {6,5,9,8,12,11,15,14,4,5,7,8,10,11,13,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX0
+#undef RGBG_INDEX1
+#undef RGBG_INDEX2
+#undef RGBG_INDEX3
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGBG_INDEX {2,1,6,5,10,9,14,13,0,1,4,5,8,9,12,13}
+#define jsimd_rgb_gray_convert_altivec jsimd_extbgrx_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGBG_INDEX {3,2,7,6,11,10,15,14,1,2,5,6,9,10,13,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extxbgr_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGBG_INDEX {1,2,5,6,9,10,13,14,3,2,7,6,11,10,15,14}
+#define jsimd_rgb_gray_convert_altivec jsimd_extxrgb_gray_convert_altivec
+#include "jcgryext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGBG_INDEX
+#undef jsimd_rgb_gray_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,116 @@
+;
+; jcgray.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2011 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_mmx) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_mmx):
+
+PW_F0299_F0337  times 2 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 2 dw  F_0_114, F_0_250
+PD_ONEHALF      times 2 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extrgbx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extbgrx_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extxbgr_gray_convert_mmx
+%include "jcgryext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_mmx jsimd_extxrgb_gray_convert_mmx
+%include "jcgryext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgray-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgray-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,113 @@
+;
+; jcgray.asm - grayscale colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_114 equ      7471                   ; FIX(0.11400)
+F_0_250 equ     16384                   ; FIX(0.25000)
+F_0_299 equ     19595                   ; FIX(0.29900)
+F_0_587 equ     38470                   ; FIX(0.58700)
+F_0_337 equ     (F_0_587 - F_0_250)     ; FIX(0.58700) - FIX(0.25000)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jconst_rgb_gray_convert_sse2):
+
+PW_F0299_F0337  times 4 dw  F_0_299, F_0_337
+PW_F0114_F0250  times 4 dw  F_0_114, F_0_250
+PD_ONEHALF      times 4 dd  (1 << (SCALEBITS-1))
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
+%include "jcgryext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
+%include "jcgryext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,227 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * Copyright (C) 2014, Jay Foad.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jcgray-altivec.c */
+
+
+void jsimd_rgb_gray_convert_altivec (JDIMENSION img_width,
+                                     JSAMPARRAY input_buf,
+                                     JSAMPIMAGE output_buf,
+                                     JDIMENSION output_row, int num_rows)
+{
+  JSAMPROW inptr, outptr;
+  int pitch = img_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+#endif
+
+  __vector unsigned char rgb0, rgb1 = {0}, rgb2 = {0},
+    rgbg0, rgbg1, rgbg2, rgbg3, y;
+#if __BIG_ENDIAN__ || RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3 = {0};
+#endif
+#if __BIG_ENDIAN__ && RGB_PIXELSIZE == 4
+  __vector unsigned char rgb4 = {0};
+#endif
+  __vector short rg0, rg1, rg2, rg3, bg0, bg1, bg2, bg3;
+  __vector unsigned short yl, yh;
+  __vector int y0, y1, y2, y3;
+
+  /* Constants */
+  __vector short pw_f0299_f0337 = { __4X2(F_0_299, F_0_337) },
+    pw_f0114_f0250 = { __4X2(F_0_114, F_0_250) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, inptr += RGB_PIXELSIZE * 16,
+         outptr += 16) {
+
+#if __BIG_ENDIAN__
+      /* Load 16 pixels == 48 or 64 bytes */
+      offset = (size_t)inptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overread.  Since there is no way to
+           * read a partial AltiVec register, overread would occur on the last
+           * chunk of the last image row if the right edge is not on a 16-byte
+           * boundary.  It could also occur on other rows if the bytes per row
+           * is low enough.  Since we can't determine whether we're on the last
+           * image row, we have to assume every row is the last.
+           */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (bytes > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (bytes > 32)
+            rgb2 = vec_ld(32, inptr);
+          if (bytes > 48)
+            rgb3 = vec_ld(48, inptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            rgb4 = vec_ld(64, inptr);
+#endif
+          unaligned_shift_index = vec_lvsl(0, inptr);
+          rgb0 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          rgb1 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+          rgb2 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_perm(rgb3, rgb4, unaligned_shift_index);
+#endif
+        }
+      } else {
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          memcpy(tmpbuf, inptr, min(num_cols, RGB_PIXELSIZE * 16));
+          rgb0 = vec_ld(0, tmpbuf);
+          rgb1 = vec_ld(16, tmpbuf);
+          rgb2 = vec_ld(32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          rgb3 = vec_ld(48, tmpbuf);
+#endif
+        } else {
+          /* Fast path */
+          rgb0 = vec_ld(0, inptr);
+          if (num_cols > 16)
+            rgb1 = vec_ld(16, inptr);
+          if (num_cols > 32)
+            rgb2 = vec_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            rgb3 = vec_ld(48, inptr);
+#endif
+        }
+      }
+#else
+      /* Little endian */
+      rgb0 = vec_vsx_ld(0, inptr);
+      if (num_cols > 16)
+        rgb1 = vec_vsx_ld(16, inptr);
+      if (num_cols > 32)
+        rgb2 = vec_vsx_ld(32, inptr);
+#if RGB_PIXELSIZE == 4
+      if (num_cols > 48)
+        rgb3 = vec_vsx_ld(48, inptr);
+#endif
+#endif
+
+#if RGB_PIXELSIZE == 3
+      /* rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX0);
+      rgbg1 = vec_perm(rgb0, rgb1, (__vector unsigned char)RGBG_INDEX1);
+      rgbg2 = vec_perm(rgb1, rgb2, (__vector unsigned char)RGBG_INDEX2);
+      rgbg3 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX3);
+#else
+      /* rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       *
+       * rgbg0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 G0 B1 G1 B2 G2 B3 G3
+       * rgbg1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 G4 B5 G5 B6 G6 B7 G7
+       * rgbg2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 G8 B9 G9 Ba Ga Bb Gb
+       * rgbg3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Gc Bd Gd Be Ge Bf Gf
+       */
+      rgbg0 = vec_perm(rgb0, rgb0, (__vector unsigned char)RGBG_INDEX);
+      rgbg1 = vec_perm(rgb1, rgb1, (__vector unsigned char)RGBG_INDEX);
+      rgbg2 = vec_perm(rgb2, rgb2, (__vector unsigned char)RGBG_INDEX);
+      rgbg3 = vec_perm(rgb3, rgb3, (__vector unsigned char)RGBG_INDEX);
+#endif
+
+      /* rg0 = R0 G0 R1 G1 R2 G2 R3 G3
+       * bg0 = B0 G0 B1 G1 B2 G2 B3 G3
+       * ...
+       *
+       * NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      rg0 = (__vector signed short)VEC_UNPACKHU(rgbg0);
+      bg0 = (__vector signed short)VEC_UNPACKLU(rgbg0);
+      rg1 = (__vector signed short)VEC_UNPACKHU(rgbg1);
+      bg1 = (__vector signed short)VEC_UNPACKLU(rgbg1);
+      rg2 = (__vector signed short)VEC_UNPACKHU(rgbg2);
+      bg2 = (__vector signed short)VEC_UNPACKLU(rgbg2);
+      rg3 = (__vector signed short)VEC_UNPACKHU(rgbg3);
+      bg3 = (__vector signed short)VEC_UNPACKLU(rgbg3);
+
+      /* (Original)
+       * Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+       *
+       * (This implementation)
+       * Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+       */
+
+      /* Calculate Y values */
+
+      y0 = vec_msums(rg0, pw_f0299_f0337, pd_onehalf);
+      y1 = vec_msums(rg1, pw_f0299_f0337, pd_onehalf);
+      y2 = vec_msums(rg2, pw_f0299_f0337, pd_onehalf);
+      y3 = vec_msums(rg3, pw_f0299_f0337, pd_onehalf);
+      y0 = vec_msums(bg0, pw_f0114_f0250, y0);
+      y1 = vec_msums(bg1, pw_f0114_f0250, y1);
+      y2 = vec_msums(bg2, pw_f0114_f0250, y2);
+      y3 = vec_msums(bg3, pw_f0114_f0250, y3);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      yl = vec_perm((__vector unsigned short)y0, (__vector unsigned short)y1,
+                    shift_pack_index);
+      yh = vec_perm((__vector unsigned short)y2, (__vector unsigned short)y3,
+                    shift_pack_index);
+      y = vec_pack(yl, yh);
+      vec_st(y, 0, outptr);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,357 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2011 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_mmx (JDIMENSION img_width,
+;                             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                             JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_rgb_gray_convert_mmx) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        xor     eax,eax
+        mov     al, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        xor     edx,edx
+        mov     dx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    mmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    mmG, DWORD [esi+ecx]
+        psllq   mmA, DWORD_BIT
+        por     mmA,mmG
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        movq    mmG,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        mov     ecx, SIZEOF_MMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld16:
+        test    cl, 2*SIZEOF_MMWORD
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_gray_cnv
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmG, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+2*SIZEOF_MMWORD]
+
+.rgb_gray_cnv:
+        ; mmA=(00 10 20 01 11 21 02 12)
+        ; mmG=(22 03 13 23 04 14 24 05)
+        ; mmF=(15 25 06 16 26 07 17 27)
+
+        movq      mmD,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 10 20 01)
+        psrlq     mmD,4*BYTE_BIT        ; mmD=(11 21 02 12 -- -- -- --)
+
+        punpckhbw mmA,mmG               ; mmA=(00 04 10 14 20 24 01 05)
+        psllq     mmG,4*BYTE_BIT        ; mmG=(-- -- -- -- 22 03 13 23)
+
+        punpcklbw mmD,mmF               ; mmD=(11 15 21 25 02 06 12 16)
+        punpckhbw mmG,mmF               ; mmG=(22 26 03 07 13 17 23 27)
+
+        movq      mmE,mmA
+        psllq     mmA,4*BYTE_BIT        ; mmA=(-- -- -- -- 00 04 10 14)
+        psrlq     mmE,4*BYTE_BIT        ; mmE=(20 24 01 05 -- -- -- --)
+
+        punpckhbw mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        psllq     mmD,4*BYTE_BIT        ; mmD=(-- -- -- -- 11 15 21 25)
+
+        punpcklbw mmE,mmG               ; mmE=(20 22 24 26 01 03 05 07)
+        punpckhbw mmD,mmG               ; mmD=(11 13 15 17 21 23 25 27)
+
+        pxor      mmH,mmH
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmH               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmH               ; mmC=(10 12 14 16)
+
+        movq      mmB,mmE
+        punpcklbw mmE,mmH               ; mmE=(20 22 24 26)
+        punpckhbw mmB,mmH               ; mmB=(01 03 05 07)
+
+        movq      mmF,mmD
+        punpcklbw mmD,mmH               ; mmD=(11 13 15 17)
+        punpckhbw mmF,mmH               ; mmF=(21 23 25 27)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_MMWORD/8
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_MMWORD/8
+        movd    mmA, DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_MMWORD/4
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_MMWORD/4
+        movq    mmF,mmA
+        movq    mmA, MMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld4:
+        test    cl, SIZEOF_MMWORD/2
+        mov     ecx, SIZEOF_MMWORD
+        jz      short .rgb_gray_cnv
+        movq    mmD,mmA
+        movq    mmC,mmF
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movq    mmA, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mmF, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mmD, MMWORD [esi+2*SIZEOF_MMWORD]
+        movq    mmC, MMWORD [esi+3*SIZEOF_MMWORD]
+
+.rgb_gray_cnv:
+        ; mmA=(00 10 20 30 01 11 21 31)
+        ; mmF=(02 12 22 32 03 13 23 33)
+        ; mmD=(04 14 24 34 05 15 25 35)
+        ; mmC=(06 16 26 36 07 17 27 37)
+
+        movq      mmB,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 10 12 20 22 30 32)
+        punpckhbw mmB,mmF               ; mmB=(01 03 11 13 21 23 31 33)
+
+        movq      mmG,mmD
+        punpcklbw mmD,mmC               ; mmD=(04 06 14 16 24 26 34 36)
+        punpckhbw mmG,mmC               ; mmG=(05 07 15 17 25 27 35 37)
+
+        movq      mmE,mmA
+        punpcklwd mmA,mmD               ; mmA=(00 02 04 06 10 12 14 16)
+        punpckhwd mmE,mmD               ; mmE=(20 22 24 26 30 32 34 36)
+
+        movq      mmH,mmB
+        punpcklwd mmB,mmG               ; mmB=(01 03 05 07 11 13 15 17)
+        punpckhwd mmH,mmG               ; mmH=(21 23 25 27 31 33 35 37)
+
+        pxor      mmF,mmF
+
+        movq      mmC,mmA
+        punpcklbw mmA,mmF               ; mmA=(00 02 04 06)
+        punpckhbw mmC,mmF               ; mmC=(10 12 14 16)
+
+        movq      mmD,mmB
+        punpcklbw mmB,mmF               ; mmB=(01 03 05 07)
+        punpckhbw mmD,mmF               ; mmD=(11 13 15 17)
+
+        movq      mmG,mmE
+        punpcklbw mmE,mmF               ; mmE=(20 22 24 26)
+        punpckhbw mmG,mmF               ; mmG=(30 32 34 36)
+
+        punpcklbw mmF,mmH
+        punpckhbw mmH,mmH
+        psrlw     mmF,BYTE_BIT          ; mmF=(21 23 25 27)
+        psrlw     mmH,BYTE_BIT          ; mmH=(31 33 35 37)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; mm0=(R0 R2 R4 R6)=RE, mm2=(G0 G2 G4 G6)=GE, mm4=(B0 B2 B4 B6)=BE
+        ; mm1=(R1 R3 R5 R7)=RO, mm3=(G1 G3 G5 G7)=GO, mm5=(B1 B3 B5 B7)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movq      mm6,mm1
+        punpcklwd mm1,mm3
+        punpckhwd mm6,mm3
+        pmaddwd   mm1,[GOTOFF(eax,PW_F0299_F0337)] ; mm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movq      mm7, mm6      ; mm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movq      mm6,mm0
+        punpcklwd mm0,mm2
+        punpckhwd mm6,mm2
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0299_F0337)] ; mm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0299_F0337)] ; mm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movq      MMWORD [wk(0)], mm0   ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movq      MMWORD [wk(1)], mm6   ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movq      mm0, mm5      ; mm0=BO
+        movq      mm6, mm4      ; mm6=BE
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm3
+        punpckhwd mm4,mm3
+        pmaddwd   mm0,[GOTOFF(eax,PW_F0114_F0250)] ; mm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movq      mm3,[GOTOFF(eax,PD_ONEHALF)]  ; mm3=[PD_ONEHALF]
+
+        paddd     mm0, mm1
+        paddd     mm4, mm7
+        paddd     mm0,mm3
+        paddd     mm4,mm3
+        psrld     mm0,SCALEBITS         ; mm0=YOL
+        psrld     mm4,SCALEBITS         ; mm4=YOH
+        packssdw  mm0,mm4               ; mm0=YO
+
+        movq      mm4,mm6
+        punpcklwd mm6,mm2
+        punpckhwd mm4,mm2
+        pmaddwd   mm6,[GOTOFF(eax,PW_F0114_F0250)] ; mm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   mm4,[GOTOFF(eax,PW_F0114_F0250)] ; mm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movq      mm2,[GOTOFF(eax,PD_ONEHALF)]  ; mm2=[PD_ONEHALF]
+
+        paddd     mm6, MMWORD [wk(0)]
+        paddd     mm4, MMWORD [wk(1)]
+        paddd     mm6,mm2
+        paddd     mm4,mm2
+        psrld     mm6,SCALEBITS         ; mm6=YEL
+        psrld     mm4,SCALEBITS         ; mm4=YEH
+        packssdw  mm6,mm4               ; mm6=YE
+
+        psllw     mm0,BYTE_BIT
+        por       mm6,mm0               ; mm6=Y
+        movq      MMWORD [edi], mm6     ; Save Y
+
+        sub     ecx, byte SIZEOF_MMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; inptr
+        add     edi, byte SIZEOF_MMWORD                 ; outptr0
+        cmp     ecx, byte SIZEOF_MMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,364 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (64-bit SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width,
+;                              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                              JDIMENSION output_row, int num_rows);
+;
+
+; r10 = JDIMENSION img_width
+; r11 = JSAMPARRAY input_buf
+; r12 = JSAMPIMAGE output_buf
+; r13 = JDIMENSION output_row
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+
+        global  EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov rsi, r12
+        mov ecx, r13d
+        mov     rdi, JSAMPARRAY [rsi+0*SIZEOF_JSAMPARRAY]
+        lea     rdi, [rdi+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov rsi, r11
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rdi
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr0
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    rax
+        push    rdx
+        lea     rcx,[rcx+rcx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_BYTE
+        movzx   rax, BYTE [rsi+rcx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_WORD
+        movzx   rdx, WORD [rsi+rcx]
+        shl     rax, WORD_BIT
+        or      rax,rdx
+.column_ld4:
+        movd    xmmA,eax
+        pop     rdx
+        pop     rax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     rcx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [rsi+rcx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     rcx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [rsi+rcx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     rcx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [rsi+rcx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     rcx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [rsi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [rsi+3*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        pmaddwd   xmm1,[rel PW_F0299_F0337] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm7, xmm6    ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        pmaddwd   xmm0,[rel PW_F0299_F0337] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[rel PW_F0299_F0337] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    xmm0, xmm5    ; xmm0=BO
+        movdqa    xmm6, xmm4    ; xmm6=BE
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        pmaddwd   xmm0,[rel PW_F0114_F0250] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movdqa    xmm3,[rel PD_ONEHALF] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, xmm1
+        paddd     xmm4, xmm7
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        pmaddwd   xmm6,[rel PW_F0114_F0250] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[rel PW_F0114_F0250] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movdqa    xmm2,[rel PD_ONEHALF] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(0)]
+        paddd     xmm4, XMMWORD [wk(1)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [rdi], xmm6   ; Save Y
+
+        sub     rcx, byte SIZEOF_XMMWORD
+        add     rsi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     rdi, byte SIZEOF_XMMWORD                ; outptr0
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .column_ld1
+
+        pop     rcx                     ; col
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     rdi, byte SIZEOF_JSAMPROW
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcgryext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcgryext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,383 @@
+;
+; jcgryext.asm - grayscale colorspace conversion (SSE2)
+;
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; Copyright (C) 2011, D. R. Commander.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_rgb_gray_convert_sse2 (JDIMENSION img_width,
+;                              JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+;                              JDIMENSION output_row, int num_rows);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION img_width
+%define input_buf(b)    (b)+12          ; JSAMPARRAY input_buf
+%define output_buf(b)   (b)+16          ; JSAMPIMAGE output_buf
+%define output_row(b)   (b)+20          ; JDIMENSION output_row
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+
+        global  EXTN(jsimd_rgb_gray_convert_sse2) PRIVATE
+
+EXTN(jsimd_rgb_gray_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [img_width(eax)]
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     esi, JSAMPIMAGE [output_buf(eax)]
+        mov     ecx, JDIMENSION [output_row(eax)]
+        mov     edi, JSAMPARRAY [esi+0*SIZEOF_JSAMPARRAY]
+        lea     edi, [edi+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     esi, JSAMPARRAY [input_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        pushpic eax
+        push    edi
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr0
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        alignx  16,7
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+.column_ld1:
+        push    eax
+        push    edx
+        lea     ecx,[ecx+ecx*2]         ; imul ecx,RGB_PIXELSIZE
+        test    cl, SIZEOF_BYTE
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_BYTE
+        movzx   eax, BYTE [esi+ecx]
+.column_ld2:
+        test    cl, SIZEOF_WORD
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_WORD
+        movzx   edx, WORD [esi+ecx]
+        shl     eax, WORD_BIT
+        or      eax,edx
+.column_ld4:
+        movd    xmmA,eax
+        pop     edx
+        pop     eax
+        test    cl, SIZEOF_DWORD
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_DWORD
+        movd    xmmF, XMM_DWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_DWORD
+        por     xmmA,xmmF
+.column_ld8:
+        test    cl, SIZEOF_MMWORD
+        jz      short .column_ld16
+        sub     ecx, byte SIZEOF_MMWORD
+        movq    xmmB, XMM_MMWORD [esi+ecx]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmB
+.column_ld16:
+        test    cl, SIZEOF_XMMWORD
+        jz      short .column_ld32
+        movdqa  xmmF,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .rgb_gray_cnv
+.column_ld32:
+        test    cl, 2*SIZEOF_XMMWORD
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmB,xmmA
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmB, XMMWORD [esi+2*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        ; xmmF=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        ; xmmB=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        movdqa    xmmG,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 10 20 01 11 21 02 12)
+        psrldq    xmmG,8        ; xmmG=(22 03 13 23 04 14 24 05 -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmF     ; xmmA=(00 08 10 18 20 28 01 09 11 19 21 29 02 0A 12 1A)
+        pslldq    xmmF,8        ; xmmF=(-- -- -- -- -- -- -- -- 15 25 06 16 26 07 17 27)
+
+        punpcklbw xmmG,xmmB     ; xmmG=(22 2A 03 0B 13 1B 23 2B 04 0C 14 1C 24 2C 05 0D)
+        punpckhbw xmmF,xmmB     ; xmmF=(15 1D 25 2D 06 0E 16 1E 26 2E 07 0F 17 1F 27 2F)
+
+        movdqa    xmmD,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 08 10 18 20 28 01 09)
+        psrldq    xmmD,8        ; xmmD=(11 19 21 29 02 0A 12 1A -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmG     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 01 05 09 0D)
+        pslldq    xmmG,8        ; xmmG=(-- -- -- -- -- -- -- -- 22 2A 03 0B 13 1B 23 2B)
+
+        punpcklbw xmmD,xmmF     ; xmmD=(11 15 19 1D 21 25 29 2D 02 06 0A 0E 12 16 1A 1E)
+        punpckhbw xmmG,xmmF     ; xmmG=(22 26 2A 2E 03 07 0B 0F 13 17 1B 1F 23 27 2B 2F)
+
+        movdqa    xmmE,xmmA
+        pslldq    xmmA,8        ; xmmA=(-- -- -- -- -- -- -- -- 00 04 08 0C 10 14 18 1C)
+        psrldq    xmmE,8        ; xmmE=(20 24 28 2C 01 05 09 0D -- -- -- -- -- -- -- --)
+
+        punpckhbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        pslldq    xmmD,8        ; xmmD=(-- -- -- -- -- -- -- -- 11 15 19 1D 21 25 29 2D)
+
+        punpcklbw xmmE,xmmG     ; xmmE=(20 22 24 26 28 2A 2C 2E 01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmG     ; xmmD=(11 13 15 17 19 1B 1D 1F 21 23 25 27 29 2B 2D 2F)
+
+        pxor      xmmH,xmmH
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmH     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmH     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmB,xmmE
+        punpcklbw xmmE,xmmH     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmB,xmmH     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+
+        movdqa    xmmF,xmmD
+        punpcklbw xmmD,xmmH     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmF,xmmH     ; xmmF=(21 23 25 27 29 2B 2D 2F)
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+.column_ld1:
+        test    cl, SIZEOF_XMMWORD/16
+        jz      short .column_ld2
+        sub     ecx, byte SIZEOF_XMMWORD/16
+        movd    xmmA, XMM_DWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld2:
+        test    cl, SIZEOF_XMMWORD/8
+        jz      short .column_ld4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        movq    xmmE, XMM_MMWORD [esi+ecx*RGB_PIXELSIZE]
+        pslldq  xmmA, SIZEOF_MMWORD
+        por     xmmA,xmmE
+.column_ld4:
+        test    cl, SIZEOF_XMMWORD/4
+        jz      short .column_ld8
+        sub     ecx, byte SIZEOF_XMMWORD/4
+        movdqa  xmmE,xmmA
+        movdqu  xmmA, XMMWORD [esi+ecx*RGB_PIXELSIZE]
+.column_ld8:
+        test    cl, SIZEOF_XMMWORD/2
+        mov     ecx, SIZEOF_XMMWORD
+        jz      short .rgb_gray_cnv
+        movdqa  xmmF,xmmA
+        movdqa  xmmH,xmmE
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        jmp     short .rgb_gray_cnv
+        alignx  16,7
+
+.columnloop:
+        movdqu  xmmA, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqu  xmmE, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        movdqu  xmmF, XMMWORD [esi+2*SIZEOF_XMMWORD]
+        movdqu  xmmH, XMMWORD [esi+3*SIZEOF_XMMWORD]
+
+.rgb_gray_cnv:
+        ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        ; xmmE=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        ; xmmF=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpcklbw xmmA,xmmE     ; xmmA=(00 04 10 14 20 24 30 34 01 05 11 15 21 25 31 35)
+        punpckhbw xmmD,xmmE     ; xmmD=(02 06 12 16 22 26 32 36 03 07 13 17 23 27 33 37)
+
+        movdqa    xmmC,xmmF
+        punpcklbw xmmF,xmmH     ; xmmF=(08 0C 18 1C 28 2C 38 3C 09 0D 19 1D 29 2D 39 3D)
+        punpckhbw xmmC,xmmH     ; xmmC=(0A 0E 1A 1E 2A 2E 3A 3E 0B 0F 1B 1F 2B 2F 3B 3F)
+
+        movdqa    xmmB,xmmA
+        punpcklwd xmmA,xmmF     ; xmmA=(00 04 08 0C 10 14 18 1C 20 24 28 2C 30 34 38 3C)
+        punpckhwd xmmB,xmmF     ; xmmB=(01 05 09 0D 11 15 19 1D 21 25 29 2D 31 35 39 3D)
+
+        movdqa    xmmG,xmmD
+        punpcklwd xmmD,xmmC     ; xmmD=(02 06 0A 0E 12 16 1A 1E 22 26 2A 2E 32 36 3A 3E)
+        punpckhwd xmmG,xmmC     ; xmmG=(03 07 0B 0F 13 17 1B 1F 23 27 2B 2F 33 37 3B 3F)
+
+        movdqa    xmmE,xmmA
+        punpcklbw xmmA,xmmD     ; xmmA=(00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E)
+        punpckhbw xmmE,xmmD     ; xmmE=(20 22 24 26 28 2A 2C 2E 30 32 34 36 38 3A 3C 3E)
+
+        movdqa    xmmH,xmmB
+        punpcklbw xmmB,xmmG     ; xmmB=(01 03 05 07 09 0B 0D 0F 11 13 15 17 19 1B 1D 1F)
+        punpckhbw xmmH,xmmG     ; xmmH=(21 23 25 27 29 2B 2D 2F 31 33 35 37 39 3B 3D 3F)
+
+        pxor      xmmF,xmmF
+
+        movdqa    xmmC,xmmA
+        punpcklbw xmmA,xmmF     ; xmmA=(00 02 04 06 08 0A 0C 0E)
+        punpckhbw xmmC,xmmF     ; xmmC=(10 12 14 16 18 1A 1C 1E)
+
+        movdqa    xmmD,xmmB
+        punpcklbw xmmB,xmmF     ; xmmB=(01 03 05 07 09 0B 0D 0F)
+        punpckhbw xmmD,xmmF     ; xmmD=(11 13 15 17 19 1B 1D 1F)
+
+        movdqa    xmmG,xmmE
+        punpcklbw xmmE,xmmF     ; xmmE=(20 22 24 26 28 2A 2C 2E)
+        punpckhbw xmmG,xmmF     ; xmmG=(30 32 34 36 38 3A 3C 3E)
+
+        punpcklbw xmmF,xmmH
+        punpckhbw xmmH,xmmH
+        psrlw     xmmF,BYTE_BIT ; xmmF=(21 23 25 27 29 2B 2D 2F)
+        psrlw     xmmH,BYTE_BIT ; xmmH=(31 33 35 37 39 3B 3D 3F)
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        ; xmm0=R(02468ACE)=RE, xmm2=G(02468ACE)=GE, xmm4=B(02468ACE)=BE
+        ; xmm1=R(13579BDF)=RO, xmm3=G(13579BDF)=GO, xmm5=B(13579BDF)=BO
+
+        ; (Original)
+        ; Y  =  0.29900 * R + 0.58700 * G + 0.11400 * B
+        ;
+        ; (This implementation)
+        ; Y  =  0.29900 * R + 0.33700 * G + 0.11400 * B + 0.25000 * G
+
+        movdqa    xmm6,xmm1
+        punpcklwd xmm1,xmm3
+        punpckhwd xmm6,xmm3
+        pmaddwd   xmm1,[GOTOFF(eax,PW_F0299_F0337)] ; xmm1=ROL*FIX(0.299)+GOL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm7, xmm6    ; xmm7=ROH*FIX(0.299)+GOH*FIX(0.337)
+
+        movdqa    xmm6,xmm0
+        punpcklwd xmm0,xmm2
+        punpckhwd xmm6,xmm2
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0299_F0337)] ; xmm0=REL*FIX(0.299)+GEL*FIX(0.337)
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0299_F0337)] ; xmm6=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    XMMWORD [wk(0)], xmm0 ; wk(0)=REL*FIX(0.299)+GEL*FIX(0.337)
+        movdqa    XMMWORD [wk(1)], xmm6 ; wk(1)=REH*FIX(0.299)+GEH*FIX(0.337)
+
+        movdqa    xmm0, xmm5    ; xmm0=BO
+        movdqa    xmm6, xmm4    ; xmm6=BE
+
+        movdqa    xmm4,xmm0
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm4,xmm3
+        pmaddwd   xmm0,[GOTOFF(eax,PW_F0114_F0250)] ; xmm0=BOL*FIX(0.114)+GOL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BOH*FIX(0.114)+GOH*FIX(0.250)
+
+        movdqa    xmm3,[GOTOFF(eax,PD_ONEHALF)] ; xmm3=[PD_ONEHALF]
+
+        paddd     xmm0, xmm1
+        paddd     xmm4, xmm7
+        paddd     xmm0,xmm3
+        paddd     xmm4,xmm3
+        psrld     xmm0,SCALEBITS        ; xmm0=YOL
+        psrld     xmm4,SCALEBITS        ; xmm4=YOH
+        packssdw  xmm0,xmm4             ; xmm0=YO
+
+        movdqa    xmm4,xmm6
+        punpcklwd xmm6,xmm2
+        punpckhwd xmm4,xmm2
+        pmaddwd   xmm6,[GOTOFF(eax,PW_F0114_F0250)] ; xmm6=BEL*FIX(0.114)+GEL*FIX(0.250)
+        pmaddwd   xmm4,[GOTOFF(eax,PW_F0114_F0250)] ; xmm4=BEH*FIX(0.114)+GEH*FIX(0.250)
+
+        movdqa    xmm2,[GOTOFF(eax,PD_ONEHALF)] ; xmm2=[PD_ONEHALF]
+
+        paddd     xmm6, XMMWORD [wk(0)]
+        paddd     xmm4, XMMWORD [wk(1)]
+        paddd     xmm6,xmm2
+        paddd     xmm4,xmm2
+        psrld     xmm6,SCALEBITS        ; xmm6=YEL
+        psrld     xmm4,SCALEBITS        ; xmm4=YEH
+        packssdw  xmm6,xmm4             ; xmm6=YE
+
+        psllw     xmm0,BYTE_BIT
+        por       xmm6,xmm0             ; xmm6=Y
+        movdqa    XMMWORD [edi], xmm6   ; Save Y
+
+        sub     ecx, byte SIZEOF_XMMWORD
+        add     esi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; inptr
+        add     edi, byte SIZEOF_XMMWORD                ; outptr0
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .column_ld1
+
+        pop     ecx                     ; col
+        pop     esi
+        pop     edi
+        poppic  eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_buf
+        add     edi, byte SIZEOF_JSAMPROW
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jchuff-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jchuff-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jchuff-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jchuff-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,361 @@
+;
+; jchuff-sse2-64.asm - Huffman entropy encoding (64-bit SSE2)
+;
+; Copyright 2009-2011, 2014-2016 D. R. Commander.
+; Copyright 2015 Matthieu Darbois
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based directly on jchuff.c; see jchuff.c for more
+; details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_huff_encode_one_block) PRIVATE
+
+EXTN(jconst_huff_encode_one_block):
+
+%include "jpeg_nbits_table.inc"
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+; These macros perform the same task as the emit_bits() function in the
+; original libjpeg code.  In addition to reducing overhead by explicitly
+; inlining the code, additional performance is achieved by taking into
+; account the size of the bit buffer and waiting until it is almost full
+; before emptying it.  This mostly benefits 64-bit platforms, since 6
+; bytes can be stored in a 64-bit bit buffer before it has to be emptied.
+
+%macro EMIT_BYTE 0
+        sub put_bits, 8  ; put_bits -= 8;
+        mov rdx, put_buffer
+        mov ecx, put_bits
+        shr rdx, cl  ; c = (JOCTET)GETJOCTET(put_buffer >> put_bits);
+        mov byte [buffer], dl  ; *buffer++ = c;
+        add buffer, 1
+        cmp dl, 0xFF  ; need to stuff a zero byte?
+        jne %%.EMIT_BYTE_END
+        mov byte [buffer], 0  ; *buffer++ = 0;
+        add buffer, 1
+%%.EMIT_BYTE_END:
+%endmacro
+
+%macro PUT_BITS 1
+        add put_bits, ecx  ; put_bits += size;
+        shl put_buffer, cl  ; put_buffer = (put_buffer << size);
+        or  put_buffer, %1
+%endmacro
+
+%macro CHECKBUF31 0
+        cmp put_bits, 32  ; if (put_bits > 31) {
+        jl %%.CHECKBUF31_END
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+%%.CHECKBUF31_END:
+%endmacro
+
+%macro CHECKBUF47 0
+        cmp put_bits, 48  ; if (put_bits > 47) {
+        jl %%.CHECKBUF47_END
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+        EMIT_BYTE
+%%.CHECKBUF47_END:
+%endmacro
+
+%macro EMIT_BITS 2
+        CHECKBUF47
+        mov ecx, %2
+        PUT_BITS %1
+%endmacro
+
+%macro kloop_prepare 37  ;(ko, jno0, ..., jno31, xmm0, xmm1, xmm2, xmm3)
+    pxor xmm8, xmm8  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm9, xmm9  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm10, xmm10  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm11, xmm11  ; __m128i neg = _mm_setzero_si128();
+    pinsrw %34, word [r12 + %2  * SIZEOF_WORD], 0  ; xmm_shadow[0] = block[jno0];
+    pinsrw %35, word [r12 + %10 * SIZEOF_WORD], 0  ; xmm_shadow[8] = block[jno8];
+    pinsrw %36, word [r12 + %18 * SIZEOF_WORD], 0  ; xmm_shadow[16] = block[jno16];
+    pinsrw %37, word [r12 + %26 * SIZEOF_WORD], 0  ; xmm_shadow[24] = block[jno24];
+    pinsrw %34, word [r12 + %3  * SIZEOF_WORD], 1  ; xmm_shadow[1] = block[jno1];
+    pinsrw %35, word [r12 + %11 * SIZEOF_WORD], 1  ; xmm_shadow[9] = block[jno9];
+    pinsrw %36, word [r12 + %19 * SIZEOF_WORD], 1  ; xmm_shadow[17] = block[jno17];
+    pinsrw %37, word [r12 + %27 * SIZEOF_WORD], 1  ; xmm_shadow[25] = block[jno25];
+    pinsrw %34, word [r12 + %4  * SIZEOF_WORD], 2  ; xmm_shadow[2] = block[jno2];
+    pinsrw %35, word [r12 + %12 * SIZEOF_WORD], 2  ; xmm_shadow[10] = block[jno10];
+    pinsrw %36, word [r12 + %20 * SIZEOF_WORD], 2  ; xmm_shadow[18] = block[jno18];
+    pinsrw %37, word [r12 + %28 * SIZEOF_WORD], 2  ; xmm_shadow[26] = block[jno26];
+    pinsrw %34, word [r12 + %5  * SIZEOF_WORD], 3  ; xmm_shadow[3] = block[jno3];
+    pinsrw %35, word [r12 + %13 * SIZEOF_WORD], 3  ; xmm_shadow[11] = block[jno11];
+    pinsrw %36, word [r12 + %21 * SIZEOF_WORD], 3  ; xmm_shadow[19] = block[jno19];
+    pinsrw %37, word [r12 + %29 * SIZEOF_WORD], 3  ; xmm_shadow[27] = block[jno27];
+    pinsrw %34, word [r12 + %6  * SIZEOF_WORD], 4  ; xmm_shadow[4] = block[jno4];
+    pinsrw %35, word [r12 + %14 * SIZEOF_WORD], 4  ; xmm_shadow[12] = block[jno12];
+    pinsrw %36, word [r12 + %22 * SIZEOF_WORD], 4  ; xmm_shadow[20] = block[jno20];
+    pinsrw %37, word [r12 + %30 * SIZEOF_WORD], 4  ; xmm_shadow[28] = block[jno28];
+    pinsrw %34, word [r12 + %7  * SIZEOF_WORD], 5  ; xmm_shadow[5] = block[jno5];
+    pinsrw %35, word [r12 + %15 * SIZEOF_WORD], 5  ; xmm_shadow[13] = block[jno13];
+    pinsrw %36, word [r12 + %23 * SIZEOF_WORD], 5  ; xmm_shadow[21] = block[jno21];
+    pinsrw %37, word [r12 + %31 * SIZEOF_WORD], 5  ; xmm_shadow[29] = block[jno29];
+    pinsrw %34, word [r12 + %8  * SIZEOF_WORD], 6  ; xmm_shadow[6] = block[jno6];
+    pinsrw %35, word [r12 + %16 * SIZEOF_WORD], 6  ; xmm_shadow[14] = block[jno14];
+    pinsrw %36, word [r12 + %24 * SIZEOF_WORD], 6  ; xmm_shadow[22] = block[jno22];
+    pinsrw %37, word [r12 + %32 * SIZEOF_WORD], 6  ; xmm_shadow[30] = block[jno30];
+    pinsrw %34, word [r12 + %9  * SIZEOF_WORD], 7  ; xmm_shadow[7] = block[jno7];
+    pinsrw %35, word [r12 + %17 * SIZEOF_WORD], 7  ; xmm_shadow[15] = block[jno15];
+    pinsrw %36, word [r12 + %25 * SIZEOF_WORD], 7  ; xmm_shadow[23] = block[jno23];
+%if %1 != 32
+    pinsrw %37, word [r12 + %33 * SIZEOF_WORD], 7  ; xmm_shadow[31] = block[jno31];
+%else
+    pinsrw %37, ebx, 7  ; xmm_shadow[31] = block[jno31];
+%endif
+    pcmpgtw xmm8, %34  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm9, %35  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm10, %36  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm11, %37  ; neg = _mm_cmpgt_epi16(neg, x1);
+    paddw %34, xmm8   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %35, xmm9   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %36, xmm10  ; x1 = _mm_add_epi16(x1, neg);
+    paddw %37, xmm11  ; x1 = _mm_add_epi16(x1, neg);
+    pxor %34, xmm8    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %35, xmm9    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %36, xmm10   ; x1 = _mm_xor_si128(x1, neg);
+    pxor %37, xmm11   ; x1 = _mm_xor_si128(x1, neg);
+    pxor xmm8, %34    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm9, %35    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm10, %36   ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm11, %37   ; neg = _mm_xor_si128(neg, x1);
+    movdqa XMMWORD [t1 + %1 * SIZEOF_WORD], %34  ; _mm_storeu_si128((__m128i *)(t1 + ko), x1);
+    movdqa XMMWORD [t1 + (%1 + 8) * SIZEOF_WORD], %35  ; _mm_storeu_si128((__m128i *)(t1 + ko + 8), x1);
+    movdqa XMMWORD [t1 + (%1 + 16) * SIZEOF_WORD], %36  ; _mm_storeu_si128((__m128i *)(t1 + ko + 16), x1);
+    movdqa XMMWORD [t1 + (%1 + 24) * SIZEOF_WORD], %37  ; _mm_storeu_si128((__m128i *)(t1 + ko + 24), x1);
+    movdqa XMMWORD [t2 + %1 * SIZEOF_WORD], xmm8  ; _mm_storeu_si128((__m128i *)(t2 + ko), neg);
+    movdqa XMMWORD [t2 + (%1 + 8) * SIZEOF_WORD], xmm9  ; _mm_storeu_si128((__m128i *)(t2 + ko + 8), neg);
+    movdqa XMMWORD [t2 + (%1 + 16) * SIZEOF_WORD], xmm10  ; _mm_storeu_si128((__m128i *)(t2 + ko + 16), neg);
+    movdqa XMMWORD [t2 + (%1 + 24) * SIZEOF_WORD], xmm11  ; _mm_storeu_si128((__m128i *)(t2 + ko + 24), neg);
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET*)
+; jsimd_huff_encode_one_block_sse2 (working_state *state, JOCTET *buffer,
+;                                   JCOEFPTR block, int last_dc_val,
+;                                   c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+
+; r10 = working_state *state
+; r11 = JOCTET *buffer
+; r12 = JCOEFPTR block
+; r13 = int last_dc_val
+; r14 = c_derived_tbl *dctbl
+; r15 = c_derived_tbl *actbl
+
+%define t1              rbp-(DCTSIZE2*SIZEOF_WORD)
+%define t2              t1-(DCTSIZE2*SIZEOF_WORD)
+%define put_buffer      r8
+%define put_bits        r9d
+%define buffer          rax
+
+        align   16
+        global  EXTN(jsimd_huff_encode_one_block_sse2) PRIVATE
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [t2]
+        collect_args
+%ifdef WIN64
+        movaps  XMMWORD [rsp-1*SIZEOF_XMMWORD], xmm8
+        movaps  XMMWORD [rsp-2*SIZEOF_XMMWORD], xmm9
+        movaps  XMMWORD [rsp-3*SIZEOF_XMMWORD], xmm10
+        movaps  XMMWORD [rsp-4*SIZEOF_XMMWORD], xmm11
+        sub     rsp, 4*SIZEOF_XMMWORD
+%endif
+        push rbx
+
+        mov buffer, r11  ; r11 is now sratch
+
+        mov put_buffer, MMWORD [r10+16]  ; put_buffer = state->cur.put_buffer;
+        mov put_bits,    DWORD [r10+24]  ; put_bits = state->cur.put_bits;
+        push r10  ; r10 is now scratch
+
+        ; Encode the DC coefficient difference per section F.1.2.1
+        movsx edi, word [r12]  ; temp = temp2 = block[0] - last_dc_val;
+        sub   edi, r13d  ; r13 is not used anymore
+        mov   ebx, edi
+
+        ; This is a well-known technique for obtaining the absolute value
+        ; without a branch.  It is derived from an assembly language technique
+        ; presented in "How to Optimize for the Pentium Processors",
+        ; Copyright (c) 1996, 1997 by Agner Fog.
+        mov esi, edi
+        sar esi, 31   ; temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+        xor edi, esi  ; temp ^= temp3;
+        sub edi, esi  ; temp -= temp3;
+
+        ; For a negative input, want temp2 = bitwise complement of abs(input)
+        ; This code assumes we are on a two's complement machine
+        add ebx, esi  ; temp2 += temp3;
+
+        ; Find the number of bits needed for the magnitude of the coefficient
+        lea   r11, [rel jpeg_nbits_table]
+        movzx rdi, byte [r11 + rdi]  ; nbits = JPEG_NBITS(temp);
+        ; Emit the Huffman-coded symbol for the number of bits
+        mov   r11d,  INT [r14 + rdi * 4]  ; code = dctbl->ehufco[nbits];
+        movzx  esi, byte [r14 + rdi + 1024]  ; size = dctbl->ehufsi[nbits];
+        EMIT_BITS r11, esi  ; EMIT_BITS(code, size)
+
+        ; Mask off any extra bits in code
+        mov esi, 1
+        mov ecx, edi
+        shl esi, cl
+        dec esi
+        and ebx, esi  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+
+        ; Emit that number of bits of the value, if positive,
+        ; or the complement of its magnitude, if negative.
+        EMIT_BITS rbx, edi  ; EMIT_BITS(temp2, nbits)
+
+        ; Prepare data
+        xor ebx, ebx
+        kloop_prepare  0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, \
+                       18, 11, 4,  5,  12, 19, 26, 33, 40, 48, 41, 34, \
+                       27, 20, 13, 6,  7,  14, 21, 28, 35, \
+                       xmm0, xmm1, xmm2, xmm3
+        kloop_prepare  32, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, \
+                       30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, \
+                       53, 60, 61, 54, 47, 55, 62, 63, 63, \
+                       xmm4, xmm5, xmm6, xmm7
+
+        pxor xmm8, xmm8
+        pcmpeqw xmm0, xmm8  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm8  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm8  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm8  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        pcmpeqw xmm4, xmm8  ; tmp4 = _mm_cmpeq_epi16(tmp4, zero);
+        pcmpeqw xmm5, xmm8  ; tmp5 = _mm_cmpeq_epi16(tmp5, zero);
+        pcmpeqw xmm6, xmm8  ; tmp6 = _mm_cmpeq_epi16(tmp6, zero);
+        pcmpeqw xmm7, xmm8  ; tmp7 = _mm_cmpeq_epi16(tmp7, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        packsswb xmm4, xmm5  ; tmp4 = _mm_packs_epi16(tmp4, tmp5);
+        packsswb xmm6, xmm7  ; tmp6 = _mm_packs_epi16(tmp6, tmp7);
+        pmovmskb r11d, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb r12d, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        pmovmskb r13d, xmm4  ; index  = ((uint64_t)_mm_movemask_epi8(tmp4)) << 32;
+        pmovmskb r14d, xmm6  ; index  = ((uint64_t)_mm_movemask_epi8(tmp6)) << 48;
+        shl r12, 16
+        shl r14, 16
+        or  r11, r12
+        or  r13, r14
+        shl r13, 32
+        or  r11, r13
+        not r11  ; index = ~index;
+
+        ;mov MMWORD [ t1 + DCTSIZE2 * SIZEOF_WORD ], r11
+        ;jmp .EFN
+
+        mov   r13d,  INT [r15 + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx r14d, byte [r15 + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        lea rsi, [t1]
+.BLOOP:
+        bsf r12, r11  ; r = __builtin_ctzl(index);
+        jz .ELOOP
+        mov rcx, r12
+        lea rsi, [rsi+r12*2]  ; k += r;
+        shr r11, cl  ; index >>= r;
+        movzx rdi, word [rsi]  ; temp = t1[k];
+        lea   rbx, [rel jpeg_nbits_table]
+        movzx rdi, byte [rbx + rdi]  ; nbits = JPEG_NBITS(temp);
+.BRLOOP:
+        cmp r12, 16  ; while (r > 15) {
+        jl .ERLOOP
+        EMIT_BITS r13, r14d  ; EMIT_BITS(code_0xf0, size_0xf0)
+        sub r12, 16  ; r -= 16;
+        jmp .BRLOOP
+.ERLOOP:
+        ; Emit Huffman symbol for run length / number of bits
+        CHECKBUF31  ; uses rcx, rdx
+
+        shl r12, 4  ; temp3 = (r << 4) + nbits;
+        add r12, rdi
+        mov   ebx,  INT [r15 + r12 * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [r15 + r12 + 1024]  ; size = actbl->ehufsi[temp3];
+        PUT_BITS rbx
+
+        ;EMIT_CODE(code, size)
+
+        movsx ebx, word [rsi-DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov rcx, rdi
+        mov rdx, 1
+        shl rdx, cl
+        dec rdx
+        and rbx, rdx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        PUT_BITS rbx  ; PUT_BITS(temp2, nbits)
+
+        shr r11, 1  ; index >>= 1;
+        add rsi, 2  ; ++k;
+        jmp .BLOOP
+.ELOOP:
+        ; If the last coef(s) were zero, emit an end-of-block code
+        lea rdi, [t1 + (DCTSIZE2-1) * 2]  ; r = DCTSIZE2-1-k;
+        cmp rdi, rsi  ; if (r > 0) {
+        je .EFN
+        mov   ebx,  INT [r15]  ; code = actbl->ehufco[0];
+        movzx r12d, byte [r15 + 1024]  ; size = actbl->ehufsi[0];
+        EMIT_BITS rbx, r12d
+.EFN:
+        pop r10
+        ; Save put_buffer & put_bits
+        mov MMWORD [r10+16], put_buffer  ; state->cur.put_buffer = put_buffer;
+        mov DWORD  [r10+24], put_bits  ; state->cur.put_bits = put_bits;
+
+        pop rbx
+%ifdef WIN64
+        movaps  xmm11, XMMWORD [rsp+0*SIZEOF_XMMWORD]
+        movaps  xmm10, XMMWORD [rsp+1*SIZEOF_XMMWORD]
+        movaps  xmm9, XMMWORD [rsp+2*SIZEOF_XMMWORD]
+        movaps  xmm8, XMMWORD [rsp+3*SIZEOF_XMMWORD]
+        add     rsp, 4*SIZEOF_XMMWORD
+%endif
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jchuff-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jchuff-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jchuff-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jchuff-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,427 @@
+;
+; jchuff-sse2.asm - Huffman entropy encoding (SSE2)
+;
+; Copyright 2009-2011, 2014-2016 D. R. Commander.
+; Copyright 2015 Matthieu Darbois
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains an SSE2 implementation for Huffman coding of one block.
+; The following code is based directly on jchuff.c; see jchuff.c for more
+; details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_huff_encode_one_block) PRIVATE
+
+EXTN(jconst_huff_encode_one_block):
+
+%include "jpeg_nbits_table.inc"
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+; These macros perform the same task as the emit_bits() function in the
+; original libjpeg code.  In addition to reducing overhead by explicitly
+; inlining the code, additional performance is achieved by taking into
+; account the size of the bit buffer and waiting until it is almost full
+; before emptying it.  This mostly benefits 64-bit platforms, since 6
+; bytes can be stored in a 64-bit bit buffer before it has to be emptied.
+
+%macro EMIT_BYTE 0
+        sub put_bits, 8  ; put_bits -= 8;
+        mov edx, put_buffer
+        mov ecx, put_bits
+        shr edx, cl  ; c = (JOCTET)GETJOCTET(put_buffer >> put_bits);
+        mov byte [eax], dl  ; *buffer++ = c;
+        add eax, 1
+        cmp dl, 0xFF  ; need to stuff a zero byte?
+        jne %%.EMIT_BYTE_END
+        mov byte [eax], 0  ; *buffer++ = 0;
+        add eax, 1
+%%.EMIT_BYTE_END:
+%endmacro
+
+%macro PUT_BITS 1
+        add put_bits, ecx  ; put_bits += size;
+        shl put_buffer, cl  ; put_buffer = (put_buffer << size);
+        or  put_buffer, %1
+%endmacro
+
+%macro CHECKBUF15 0
+        cmp put_bits, 16  ; if (put_bits > 31) {
+        jl %%.CHECKBUF15_END
+        mov eax, POINTER [esp+buffer]
+        EMIT_BYTE
+        EMIT_BYTE
+        mov POINTER [esp+buffer], eax
+%%.CHECKBUF15_END:
+%endmacro
+
+%macro EMIT_BITS 1
+        PUT_BITS %1
+        CHECKBUF15
+%endmacro
+
+%macro kloop_prepare 37  ;(ko, jno0, ..., jno31, xmm0, xmm1, xmm2, xmm3)
+    pxor xmm4, xmm4  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm5, xmm5  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm6, xmm6  ; __m128i neg = _mm_setzero_si128();
+    pxor xmm7, xmm7  ; __m128i neg = _mm_setzero_si128();
+    pinsrw %34, word [esi + %2  * SIZEOF_WORD], 0  ; xmm_shadow[0] = block[jno0];
+    pinsrw %35, word [esi + %10 * SIZEOF_WORD], 0  ; xmm_shadow[8] = block[jno8];
+    pinsrw %36, word [esi + %18 * SIZEOF_WORD], 0  ; xmm_shadow[16] = block[jno16];
+    pinsrw %37, word [esi + %26 * SIZEOF_WORD], 0  ; xmm_shadow[24] = block[jno24];
+    pinsrw %34, word [esi + %3  * SIZEOF_WORD], 1  ; xmm_shadow[1] = block[jno1];
+    pinsrw %35, word [esi + %11 * SIZEOF_WORD], 1  ; xmm_shadow[9] = block[jno9];
+    pinsrw %36, word [esi + %19 * SIZEOF_WORD], 1  ; xmm_shadow[17] = block[jno17];
+    pinsrw %37, word [esi + %27 * SIZEOF_WORD], 1  ; xmm_shadow[25] = block[jno25];
+    pinsrw %34, word [esi + %4  * SIZEOF_WORD], 2  ; xmm_shadow[2] = block[jno2];
+    pinsrw %35, word [esi + %12 * SIZEOF_WORD], 2  ; xmm_shadow[10] = block[jno10];
+    pinsrw %36, word [esi + %20 * SIZEOF_WORD], 2  ; xmm_shadow[18] = block[jno18];
+    pinsrw %37, word [esi + %28 * SIZEOF_WORD], 2  ; xmm_shadow[26] = block[jno26];
+    pinsrw %34, word [esi + %5  * SIZEOF_WORD], 3  ; xmm_shadow[3] = block[jno3];
+    pinsrw %35, word [esi + %13 * SIZEOF_WORD], 3  ; xmm_shadow[11] = block[jno11];
+    pinsrw %36, word [esi + %21 * SIZEOF_WORD], 3  ; xmm_shadow[19] = block[jno19];
+    pinsrw %37, word [esi + %29 * SIZEOF_WORD], 3  ; xmm_shadow[27] = block[jno27];
+    pinsrw %34, word [esi + %6  * SIZEOF_WORD], 4  ; xmm_shadow[4] = block[jno4];
+    pinsrw %35, word [esi + %14 * SIZEOF_WORD], 4  ; xmm_shadow[12] = block[jno12];
+    pinsrw %36, word [esi + %22 * SIZEOF_WORD], 4  ; xmm_shadow[20] = block[jno20];
+    pinsrw %37, word [esi + %30 * SIZEOF_WORD], 4  ; xmm_shadow[28] = block[jno28];
+    pinsrw %34, word [esi + %7  * SIZEOF_WORD], 5  ; xmm_shadow[5] = block[jno5];
+    pinsrw %35, word [esi + %15 * SIZEOF_WORD], 5  ; xmm_shadow[13] = block[jno13];
+    pinsrw %36, word [esi + %23 * SIZEOF_WORD], 5  ; xmm_shadow[21] = block[jno21];
+    pinsrw %37, word [esi + %31 * SIZEOF_WORD], 5  ; xmm_shadow[29] = block[jno29];
+    pinsrw %34, word [esi + %8  * SIZEOF_WORD], 6  ; xmm_shadow[6] = block[jno6];
+    pinsrw %35, word [esi + %16 * SIZEOF_WORD], 6  ; xmm_shadow[14] = block[jno14];
+    pinsrw %36, word [esi + %24 * SIZEOF_WORD], 6  ; xmm_shadow[22] = block[jno22];
+    pinsrw %37, word [esi + %32 * SIZEOF_WORD], 6  ; xmm_shadow[30] = block[jno30];
+    pinsrw %34, word [esi + %9  * SIZEOF_WORD], 7  ; xmm_shadow[7] = block[jno7];
+    pinsrw %35, word [esi + %17 * SIZEOF_WORD], 7  ; xmm_shadow[15] = block[jno15];
+    pinsrw %36, word [esi + %25 * SIZEOF_WORD], 7  ; xmm_shadow[23] = block[jno23];
+%if %1 != 32
+    pinsrw %37, word [esi + %33 * SIZEOF_WORD], 7  ; xmm_shadow[31] = block[jno31];
+%else
+    pinsrw %37, ecx, 7  ; xmm_shadow[31] = block[jno31];
+%endif
+    pcmpgtw xmm4, %34  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm5, %35  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm6, %36  ; neg = _mm_cmpgt_epi16(neg, x1);
+    pcmpgtw xmm7, %37  ; neg = _mm_cmpgt_epi16(neg, x1);
+    paddw %34, xmm4   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %35, xmm5   ; x1 = _mm_add_epi16(x1, neg);
+    paddw %36, xmm6  ; x1 = _mm_add_epi16(x1, neg);
+    paddw %37, xmm7  ; x1 = _mm_add_epi16(x1, neg);
+    pxor %34, xmm4    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %35, xmm5    ; x1 = _mm_xor_si128(x1, neg);
+    pxor %36, xmm6   ; x1 = _mm_xor_si128(x1, neg);
+    pxor %37, xmm7   ; x1 = _mm_xor_si128(x1, neg);
+    pxor xmm4, %34    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm5, %35    ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm6, %36   ; neg = _mm_xor_si128(neg, x1);
+    pxor xmm7, %37   ; neg = _mm_xor_si128(neg, x1);
+    movdqa XMMWORD [esp + t1 + %1 * SIZEOF_WORD], %34  ; _mm_storeu_si128((__m128i *)(t1 + ko), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 8) * SIZEOF_WORD], %35  ; _mm_storeu_si128((__m128i *)(t1 + ko + 8), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 16) * SIZEOF_WORD], %36  ; _mm_storeu_si128((__m128i *)(t1 + ko + 16), x1);
+    movdqa XMMWORD [esp + t1 + (%1 + 24) * SIZEOF_WORD], %37  ; _mm_storeu_si128((__m128i *)(t1 + ko + 24), x1);
+    movdqa XMMWORD [esp + t2 + %1 * SIZEOF_WORD], xmm4  ; _mm_storeu_si128((__m128i *)(t2 + ko), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 8) * SIZEOF_WORD], xmm5  ; _mm_storeu_si128((__m128i *)(t2 + ko + 8), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 16) * SIZEOF_WORD], xmm6  ; _mm_storeu_si128((__m128i *)(t2 + ko + 16), neg);
+    movdqa XMMWORD [esp + t2 + (%1 + 24) * SIZEOF_WORD], xmm7  ; _mm_storeu_si128((__m128i *)(t2 + ko + 24), neg);
+%endmacro
+
+;
+; Encode a single block's worth of coefficients.
+;
+; GLOBAL(JOCTET*)
+; jsimd_huff_encode_one_block_sse2 (working_state *state, JOCTET *buffer,
+;                                   JCOEFPTR block, int last_dc_val,
+;                                   c_derived_tbl *dctbl, c_derived_tbl *actbl)
+;
+
+; eax + 8 = working_state *state
+; eax + 12 = JOCTET *buffer
+; eax + 16 = JCOEFPTR block
+; eax + 20 = int last_dc_val
+; eax + 24 = c_derived_tbl *dctbl
+; eax + 28 = c_derived_tbl *actbl
+
+%define pad             6*SIZEOF_DWORD  ; Align to 16 bytes
+%define t1              pad
+%define t2              t1+(DCTSIZE2*SIZEOF_WORD)
+%define block           t2+(DCTSIZE2*SIZEOF_WORD)
+%define actbl           block+SIZEOF_DWORD
+%define buffer          actbl+SIZEOF_DWORD
+%define temp            buffer+SIZEOF_DWORD
+%define temp2           temp+SIZEOF_DWORD
+%define temp3           temp2+SIZEOF_DWORD
+%define temp4           temp3+SIZEOF_DWORD
+%define temp5           temp4+SIZEOF_DWORD
+%define gotptr          temp5+SIZEOF_DWORD  ; void *gotptr
+%define put_buffer      ebx
+%define put_bits        edi
+
+        align   16
+        global  EXTN(jsimd_huff_encode_one_block_sse2) PRIVATE
+
+EXTN(jsimd_huff_encode_one_block_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        sub     esp, temp5+9*SIZEOF_DWORD-pad
+        push    ebx
+        push    ecx
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+        push    ebp
+
+        mov esi, POINTER [eax+8]        ; (working_state *state)
+        mov put_buffer,  DWORD [esi+8]  ; put_buffer = state->cur.put_buffer;
+        mov put_bits,    DWORD [esi+12]  ; put_bits = state->cur.put_bits;
+        push esi  ; esi is now scratch
+
+        get_GOT edx                       ; get GOT address
+        movpic POINTER [esp+gotptr], edx  ; save GOT address
+
+        mov ecx, POINTER [eax+28]
+        mov edx, POINTER [eax+16]
+        mov esi, POINTER [eax+12]
+        mov POINTER [esp+actbl],  ecx
+        mov POINTER [esp+block],  edx
+        mov POINTER [esp+buffer], esi
+
+        ; Encode the DC coefficient difference per section F.1.2.1
+        mov esi, POINTER [esp+block]        ; block
+        movsx ecx, word [esi]  ; temp = temp2 = block[0] - last_dc_val;
+        sub   ecx, DWORD [eax+20]
+        mov   esi, ecx
+
+        ; This is a well-known technique for obtaining the absolute value
+        ; without a branch.  It is derived from an assembly language technique
+        ; presented in "How to Optimize for the Pentium Processors",
+        ; Copyright (c) 1996, 1997 by Agner Fog.
+        mov edx, ecx
+        sar edx, 31   ; temp3 = temp >> (CHAR_BIT * sizeof(int) - 1);
+        xor ecx, edx ; temp ^= temp3;
+        sub ecx, edx ; temp -= temp3;
+
+        ; For a negative input, want temp2 = bitwise complement of abs(input)
+        ; This code assumes we are on a two's complement machine
+        add esi, edx  ; temp2 += temp3;
+        mov DWORD [esp+temp], esi  ; backup temp2 in temp
+
+        ; Find the number of bits needed for the magnitude of the coefficient
+        movpic ebp, POINTER [esp+gotptr]   ; load GOT address (ebp)
+        movzx edx, byte [GOTOFF(ebp, jpeg_nbits_table + ecx)]  ; nbits = JPEG_NBITS(temp);
+        mov DWORD [esp+temp2], edx  ; backup nbits in temp2
+
+        ; Emit the Huffman-coded symbol for the number of bits
+        mov    ebp, POINTER [eax+24]  ; After this point, arguments are not accessible anymore
+        mov    eax,  INT [ebp + edx * 4]  ; code = dctbl->ehufco[nbits];
+        movzx  ecx, byte [ebp + edx + 1024]  ; size = dctbl->ehufsi[nbits];
+        EMIT_BITS eax  ; EMIT_BITS(code, size)
+
+        mov ecx, DWORD [esp+temp2]  ; restore nbits
+
+        ; Mask off any extra bits in code
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, DWORD [esp+temp]  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+
+        ; Emit that number of bits of the value, if positive,
+        ; or the complement of its magnitude, if negative.
+        EMIT_BITS eax  ; EMIT_BITS(temp2, nbits)
+
+        ; Prepare data
+        xor ecx, ecx
+        mov esi, POINTER [esp+block]
+        kloop_prepare  0,  1,  8,  16, 9,  2,  3,  10, 17, 24, 32, 25, \
+                       18, 11, 4,  5,  12, 19, 26, 33, 40, 48, 41, 34, \
+                       27, 20, 13, 6,  7,  14, 21, 28, 35, \
+                       xmm0, xmm1, xmm2, xmm3
+        kloop_prepare  32, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, \
+                       30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, \
+                       53, 60, 61, 54, 47, 55, 62, 63, 63, \
+                       xmm0, xmm1, xmm2, xmm3
+
+        pxor xmm7, xmm7
+        movdqa xmm0, XMMWORD [esp + t1 + 0 * SIZEOF_WORD]   ; __m128i tmp0 = _mm_loadu_si128((__m128i *)(t1 + 0));
+        movdqa xmm1, XMMWORD [esp + t1 + 8 * SIZEOF_WORD]   ; __m128i tmp1 = _mm_loadu_si128((__m128i *)(t1 + 8));
+        movdqa xmm2, XMMWORD [esp + t1 + 16 * SIZEOF_WORD]  ; __m128i tmp2 = _mm_loadu_si128((__m128i *)(t1 + 16));
+        movdqa xmm3, XMMWORD [esp + t1 + 24 * SIZEOF_WORD]  ; __m128i tmp3 = _mm_loadu_si128((__m128i *)(t1 + 24));
+        pcmpeqw xmm0, xmm7  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm7  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm7  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm7  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        pmovmskb edx, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb ecx, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        shl ecx, 16
+        or  edx, ecx
+        not edx  ; index = ~index;
+
+        lea esi, [esp+t1]
+        mov ebp, POINTER [esp+actbl]  ; ebp = actbl
+
+.BLOOP:
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP
+        lea esi, [esi+ecx*2]  ; k += r;
+        shr edx, cl  ; index >>= r;
+        mov DWORD [esp+temp3], edx
+.BRLOOP:
+        cmp ecx, 16  ; while (r > 15) {
+        jl .ERLOOP
+        sub ecx, 16 ; r -= 16;
+        mov DWORD [esp+temp], ecx
+        mov   eax, INT [ebp + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx ecx, byte [ebp + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        EMIT_BITS eax  ; EMIT_BITS(code_0xf0, size_0xf0)
+        mov ecx, DWORD [esp+temp]
+        jmp .BRLOOP
+.ERLOOP:
+        movsx eax, word [esi]  ; temp = t1[k];
+        movpic edx, POINTER [esp+gotptr]   ; load GOT address (edx)
+        movzx eax, byte [GOTOFF(edx, jpeg_nbits_table + eax)]  ; nbits = JPEG_NBITS(temp);
+        mov DWORD [esp+temp2], eax
+        ; Emit Huffman symbol for run length / number of bits
+        shl ecx, 4  ; temp3 = (r << 4) + nbits;
+        add ecx, eax
+        mov   eax,  INT [ebp + ecx * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [ebp + ecx + 1024]  ; size = actbl->ehufsi[temp3];
+        EMIT_BITS eax
+
+        movsx edx, word [esi+DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov ecx, DWORD [esp+temp2]
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, edx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        EMIT_BITS eax  ; PUT_BITS(temp2, nbits)
+        mov edx, DWORD [esp+temp3]
+        add esi, 2  ; ++k;
+        shr edx, 1  ; index >>= 1;
+
+        jmp .BLOOP
+.ELOOP:
+        movdqa xmm0, XMMWORD [esp + t1 + 32 * SIZEOF_WORD]  ; __m128i tmp0 = _mm_loadu_si128((__m128i *)(t1 + 0));
+        movdqa xmm1, XMMWORD [esp + t1 + 40 * SIZEOF_WORD]  ; __m128i tmp1 = _mm_loadu_si128((__m128i *)(t1 + 8));
+        movdqa xmm2, XMMWORD [esp + t1 + 48 * SIZEOF_WORD]  ; __m128i tmp2 = _mm_loadu_si128((__m128i *)(t1 + 16));
+        movdqa xmm3, XMMWORD [esp + t1 + 56 * SIZEOF_WORD]  ; __m128i tmp3 = _mm_loadu_si128((__m128i *)(t1 + 24));
+        pcmpeqw xmm0, xmm7  ; tmp0 = _mm_cmpeq_epi16(tmp0, zero);
+        pcmpeqw xmm1, xmm7  ; tmp1 = _mm_cmpeq_epi16(tmp1, zero);
+        pcmpeqw xmm2, xmm7  ; tmp2 = _mm_cmpeq_epi16(tmp2, zero);
+        pcmpeqw xmm3, xmm7  ; tmp3 = _mm_cmpeq_epi16(tmp3, zero);
+        packsswb xmm0, xmm1  ; tmp0 = _mm_packs_epi16(tmp0, tmp1);
+        packsswb xmm2, xmm3  ; tmp2 = _mm_packs_epi16(tmp2, tmp3);
+        pmovmskb edx, xmm0  ; index  = ((uint64_t)_mm_movemask_epi8(tmp0)) << 0;
+        pmovmskb ecx, xmm2  ; index  = ((uint64_t)_mm_movemask_epi8(tmp2)) << 16;
+        shl ecx, 16
+        or  edx, ecx
+        not edx  ; index = ~index;
+
+        lea eax, [esp + t1 + (DCTSIZE2/2) * 2]
+        sub eax, esi
+        shr eax, 1
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP2
+        shr edx, cl  ; index >>= r;
+        add ecx, eax
+        lea esi, [esi+ecx*2]  ; k += r;
+        mov DWORD [esp+temp3], edx
+        jmp .BRLOOP2
+.BLOOP2:
+        bsf ecx, edx  ; r = __builtin_ctzl(index);
+        jz .ELOOP2
+        lea esi, [esi+ecx*2]  ; k += r;
+        shr edx, cl  ; index >>= r;
+        mov DWORD [esp+temp3], edx
+.BRLOOP2:
+        cmp ecx, 16  ; while (r > 15) {
+        jl .ERLOOP2
+        sub ecx, 16  ; r -= 16;
+        mov DWORD [esp+temp], ecx
+        mov   eax, INT [ebp + 240 * 4]  ; code_0xf0 = actbl->ehufco[0xf0];
+        movzx ecx, byte [ebp + 1024 + 240]  ; size_0xf0 = actbl->ehufsi[0xf0];
+        EMIT_BITS eax  ; EMIT_BITS(code_0xf0, size_0xf0)
+        mov ecx, DWORD [esp+temp]
+        jmp .BRLOOP2
+.ERLOOP2:
+        movsx eax, word [esi]  ; temp = t1[k];
+        bsr eax, eax  ; nbits = 32 - __builtin_clz(temp);
+        inc eax
+        mov DWORD [esp+temp2], eax
+        ; Emit Huffman symbol for run length / number of bits
+        shl ecx, 4  ; temp3 = (r << 4) + nbits;
+        add ecx, eax
+        mov   eax,  INT [ebp + ecx * 4]  ; code = actbl->ehufco[temp3];
+        movzx ecx, byte [ebp + ecx + 1024]  ; size = actbl->ehufsi[temp3];
+        EMIT_BITS eax
+
+        movsx edx, word [esi+DCTSIZE2*2]  ; temp2 = t2[k];
+        ; Mask off any extra bits in code
+        mov ecx, DWORD [esp+temp2]
+        mov eax, 1
+        shl eax, cl
+        dec eax
+        and eax, edx  ; temp2 &= (((JLONG) 1)<<nbits) - 1;
+        EMIT_BITS eax  ; PUT_BITS(temp2, nbits)
+        mov edx, DWORD [esp+temp3]
+        add esi, 2  ; ++k;
+        shr edx, 1  ; index >>= 1;
+
+        jmp .BLOOP2
+.ELOOP2:
+        ; If the last coef(s) were zero, emit an end-of-block code
+        lea edx, [esp + t1 + (DCTSIZE2-1) * 2]  ; r = DCTSIZE2-1-k;
+        cmp edx, esi  ; if (r > 0) {
+        je .EFN
+        mov   eax,  INT [ebp]  ; code = actbl->ehufco[0];
+        movzx ecx, byte [ebp + 1024]  ; size = actbl->ehufsi[0];
+        EMIT_BITS eax
+.EFN:
+        mov eax, [esp+buffer]
+        pop esi
+        ; Save put_buffer & put_bits
+        mov DWORD [esi+8], put_buffer  ; state->cur.put_buffer = put_buffer;
+        mov DWORD [esi+12], put_bits  ; state->cur.put_bits = put_bits;
+
+        pop     ebp
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+        pop     ecx
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcolsamp.inc glmark2-2021.02/src/libjpeg-turbo/simd/jcolsamp.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcolsamp.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcolsamp.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,105 @@
+;
+; jcolsamp.inc - private declarations for color conversion & up/downsampling
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; [TAB8]
+
+; --------------------------------------------------------------------------
+
+; pseudo-resisters to make ordering of RGB configurable
+;
+%if RGB_RED == 0
+%define  mmA  mm0
+%define  mmB  mm1
+%define xmmA xmm0
+%define xmmB xmm1
+%elif RGB_GREEN == 0
+%define  mmA  mm2
+%define  mmB  mm3
+%define xmmA xmm2
+%define xmmB xmm3
+%elif RGB_BLUE == 0
+%define  mmA  mm4
+%define  mmB  mm5
+%define xmmA xmm4
+%define xmmB xmm5
+%else
+%define  mmA  mm6
+%define  mmB  mm7
+%define xmmA xmm6
+%define xmmB xmm7
+%endif
+
+%if RGB_RED == 1
+%define  mmC  mm0
+%define  mmD  mm1
+%define xmmC xmm0
+%define xmmD xmm1
+%elif RGB_GREEN == 1
+%define  mmC  mm2
+%define  mmD  mm3
+%define xmmC xmm2
+%define xmmD xmm3
+%elif RGB_BLUE == 1
+%define  mmC  mm4
+%define  mmD  mm5
+%define xmmC xmm4
+%define xmmD xmm5
+%else
+%define  mmC  mm6
+%define  mmD  mm7
+%define xmmC xmm6
+%define xmmD xmm7
+%endif
+
+%if RGB_RED == 2
+%define  mmE  mm0
+%define  mmF  mm1
+%define xmmE xmm0
+%define xmmF xmm1
+%elif RGB_GREEN == 2
+%define  mmE  mm2
+%define  mmF  mm3
+%define xmmE xmm2
+%define xmmF xmm3
+%elif RGB_BLUE == 2
+%define  mmE  mm4
+%define  mmF  mm5
+%define xmmE xmm4
+%define xmmF xmm5
+%else
+%define  mmE  mm6
+%define  mmF  mm7
+%define xmmE xmm6
+%define xmmF xmm7
+%endif
+
+%if RGB_RED == 3
+%define  mmG  mm0
+%define  mmH  mm1
+%define xmmG xmm0
+%define xmmH xmm1
+%elif RGB_GREEN == 3
+%define  mmG  mm2
+%define  mmH  mm3
+%define xmmG xmm2
+%define xmmH xmm3
+%elif RGB_BLUE == 3
+%define  mmG  mm4
+%define  mmH  mm5
+%define xmmG xmm4
+%define xmmH xmm5
+%else
+%define  mmG  mm6
+%define  mmH  mm7
+%define xmmG xmm6
+%define xmmH xmm7
+%endif
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,158 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA DOWNSAMPLING */
+
+#include "jsimd_altivec.h"
+#include "jcsample.h"
+
+
+void
+jsimd_h2v1_downsample_altivec (JDIMENSION image_width, int max_v_samp_factor,
+                               JDIMENSION v_samp_factor,
+                               JDIMENSION width_blocks,
+                               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int outrow, outcol;
+  JDIMENSION output_cols = width_blocks * DCTSIZE;
+  JSAMPROW inptr, outptr;
+
+  __vector unsigned char this0, next0, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(0, 1) },
+    pw_one = { __8X(1) };
+  __vector unsigned char even_odd_index =
+    {0,2,4,6,8,10,12,14,1,3,5,7,9,11,13,15},
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (outrow = 0; outrow < v_samp_factor; outrow++) {
+    outptr = output_data[outrow];
+    inptr = input_data[outrow];
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      outl = vec_add(this0e, this0o);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_one);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        outh = vec_add(next0e, next0o);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_one);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_downsample_altivec (JDIMENSION image_width, int max_v_samp_factor,
+                               JDIMENSION v_samp_factor,
+                               JDIMENSION width_blocks,
+                               JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  int inrow, outrow, outcol;
+  JDIMENSION output_cols = width_blocks * DCTSIZE;
+  JSAMPROW inptr0, inptr1, outptr;
+
+  __vector unsigned char this0, next0, this1, next1, out;
+  __vector unsigned short this0e, this0o, next0e, next0o, this1e, this1o,
+    next1e, next1o, out0l, out0h, out1l, out1h, outl, outh;
+
+  /* Constants */
+  __vector unsigned short pw_bias = { __4X2(1, 2) },
+    pw_two = { __8X(2) };
+  __vector unsigned char even_odd_index =
+    { 0, 2, 4, 6, 8, 10, 12, 14, 1, 3, 5, 7, 9, 11, 13, 15 },
+    pb_zero = { __16X(0) };
+
+  expand_right_edge(input_data, max_v_samp_factor, image_width,
+                    output_cols * 2);
+
+  for (inrow = 0, outrow = 0; outrow < v_samp_factor;
+       inrow += 2, outrow++) {
+
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow + 1];
+    outptr = output_data[outrow];
+
+    for (outcol = output_cols; outcol > 0;
+         outcol -= 16, inptr0 += 32, inptr1 += 32, outptr += 16) {
+
+      this0 = vec_ld(0, inptr0);
+      this0 = vec_perm(this0, this0, even_odd_index);
+      this0e = (__vector unsigned short)VEC_UNPACKHU(this0);
+      this0o = (__vector unsigned short)VEC_UNPACKLU(this0);
+      out0l = vec_add(this0e, this0o);
+
+      this1 = vec_ld(0, inptr1);
+      this1 = vec_perm(this1, this1, even_odd_index);
+      this1e = (__vector unsigned short)VEC_UNPACKHU(this1);
+      this1o = (__vector unsigned short)VEC_UNPACKLU(this1);
+      out1l = vec_add(this1e, this1o);
+
+      outl = vec_add(out0l, out1l);
+      outl = vec_add(outl, pw_bias);
+      outl = vec_sr(outl, pw_two);
+
+      if (outcol > 8) {
+        next0 = vec_ld(16, inptr0);
+        next0 = vec_perm(next0, next0, even_odd_index);
+        next0e = (__vector unsigned short)VEC_UNPACKHU(next0);
+        next0o = (__vector unsigned short)VEC_UNPACKLU(next0);
+        out0h = vec_add(next0e, next0o);
+
+        next1 = vec_ld(16, inptr1);
+        next1 = vec_perm(next1, next1, even_odd_index);
+        next1e = (__vector unsigned short)VEC_UNPACKHU(next1);
+        next1o = (__vector unsigned short)VEC_UNPACKLU(next1);
+        out1h = vec_add(next1e, next1o);
+
+        outh = vec_add(out0h, out1h);
+        outh = vec_add(outh, pw_bias);
+        outh = vec_sr(outh, pw_two);
+      } else
+        outh = vec_splat_u16(0);
+
+      out = vec_pack(outl, outh);
+      vec_st(out, 0, outptr);
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample.h glmark2-2021.02/src/libjpeg-turbo/simd/jcsample.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcsample.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+/*
+ * jcsample.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+LOCAL(void)
+expand_right_edge (JSAMPARRAY image_data, int num_rows,
+                   JDIMENSION input_cols, JDIMENSION output_cols)
+{
+  register JSAMPROW ptr;
+  register JSAMPLE pixval;
+  register int count;
+  int row;
+  int numcols = (int) (output_cols - input_cols);
+
+  if (numcols > 0) {
+    for (row = 0; row < num_rows; row++) {
+      ptr = image_data[row] + input_cols;
+      pixval = ptr[-1];         /* don't need GETJSAMPLE() here */
+      for (count = numcols; count > 0; count--)
+        *ptr++ = pixval;
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,324 @@
+;
+; jcsample.asm - downsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                            JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov       edx, 0x00010000       ; bias pattern
+        movd      mm7,edx
+        pcmpeqw   mm6,mm6
+        punpckldq mm7,mm7               ; mm7={0, 1, 0, 1}
+        psrlw     mm6,BYTE_BIT          ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm1, MMWORD [esi+1*SIZEOF_MMWORD]
+        movq    mm2,mm0
+        movq    mm3,mm1
+
+        pand    mm0,mm6
+        psrlw   mm2,BYTE_BIT
+        pand    mm1,mm6
+        psrlw   mm3,BYTE_BIT
+
+        paddw   mm0,mm2
+        paddw   mm1,mm3
+        paddw   mm0,mm7
+        paddw   mm1,mm7
+        psrlw   mm0,1
+        psrlw   mm1,1
+
+        packuswb mm0,mm1
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 1*SIZEOF_MMWORD       ; outptr
+        sub     ecx, byte SIZEOF_MMWORD         ; outcol
+        jnz     short .columnloop
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     eax                             ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_mmx (JDIMENSION image_width, int max_v_samp_factor,
+;                            JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                            JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov       edx, 0x00020001       ; bias pattern
+        movd      mm7,edx
+        pcmpeqw   mm6,mm6
+        punpckldq mm7,mm7               ; mm7={1, 2, 1, 2}
+        psrlw     mm6,BYTE_BIT          ; mm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     edi, JSAMPROW [edi]                     ; outptr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [edx+0*SIZEOF_MMWORD]
+        movq    mm1, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm2, MMWORD [edx+1*SIZEOF_MMWORD]
+        movq    mm3, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pand    mm0,mm6
+        psrlw   mm4,BYTE_BIT
+        pand    mm1,mm6
+        psrlw   mm5,BYTE_BIT
+        paddw   mm0,mm4
+        paddw   mm1,mm5
+
+        movq    mm4,mm2
+        movq    mm5,mm3
+        pand    mm2,mm6
+        psrlw   mm4,BYTE_BIT
+        pand    mm3,mm6
+        psrlw   mm5,BYTE_BIT
+        paddw   mm2,mm4
+        paddw   mm3,mm5
+
+        paddw   mm0,mm1
+        paddw   mm2,mm3
+        paddw   mm0,mm7
+        paddw   mm2,mm7
+        psrlw   mm0,2
+        psrlw   mm2,2
+
+        packuswb mm0,mm2
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+
+        add     edx, byte 2*SIZEOF_MMWORD       ; inptr0
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr1
+        add     edi, byte 1*SIZEOF_MMWORD       ; outptr
+        sub     ecx, byte SIZEOF_MMWORD         ; outcol
+        jnz     near .columnloop
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,330 @@
+;
+; jcsample.asm - downsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+; r10 = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12 = JDIMENSION v_samp_factor
+; r13 = JDIMENSION width_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov ecx, r13d
+        shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
+        jz      near .return
+
+        mov edx, r10d
+
+        ; -- expand_right_edge
+
+        push    rcx
+        shl     rcx,1                           ; output_cols * 2
+        sub     rcx,rdx
+        jle     short .expand_end
+
+        mov     rax, r11
+        test    rax,rax
+        jle     short .expand_end
+
+        cld
+        mov     rsi, r14        ; input_data
+.expandloop:
+        push    rax
+        push    rcx
+
+        mov     rdi, JSAMPROW [rsi]
+        add     rdi,rdx
+        mov     al, JSAMPLE [rdi-1]
+
+        rep stosb
+
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        dec     rax
+        jg      short .expandloop
+
+.expand_end:
+        pop     rcx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, r12d        ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     rdx, 0x00010000         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     rsi, r14        ; input_data
+        mov     rdi, r15        ; output_data
+.rowloop:
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]             ; inptr
+        mov rdi, JSAMPROW [rdi]         ; outptr
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        pxor    xmm1,xmm1
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .downsample
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm2,xmm0
+        movdqa  xmm3,xmm1
+
+        pand    xmm0,xmm6
+        psrlw   xmm2,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm3,BYTE_BIT
+
+        paddw   xmm0,xmm2
+        paddw   xmm1,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+        psrlw   xmm0,1
+        psrlw   xmm1,1
+
+        packuswb xmm0,xmm1
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     rcx, byte SIZEOF_XMMWORD        ; outcol
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        test    rcx,rcx
+        jnz     short .columnloop_r8
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+; r10 = JDIMENSION image_width
+; r11 = int max_v_samp_factor
+; r12 = JDIMENSION v_samp_factor
+; r13 = JDIMENSION width_blocks
+; r14 = JSAMPARRAY input_data
+; r15 = JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     ecx, r13d
+        shl     rcx,3                   ; imul rcx,DCTSIZE (rcx = output_cols)
+        jz      near .return
+
+        mov     edx, r10d
+
+        ; -- expand_right_edge
+
+        push    rcx
+        shl     rcx,1                           ; output_cols * 2
+        sub     rcx,rdx
+        jle     short .expand_end
+
+        mov     rax, r11
+        test    rax,rax
+        jle     short .expand_end
+
+        cld
+        mov     rsi, r14        ; input_data
+.expandloop:
+        push    rax
+        push    rcx
+
+        mov     rdi, JSAMPROW [rsi]
+        add     rdi,rdx
+        mov     al, JSAMPLE [rdi-1]
+
+        rep stosb
+
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        dec     rax
+        jg      short .expandloop
+
+.expand_end:
+        pop     rcx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, r12d        ; rowctr
+        test    rax,rax
+        jle     near .return
+
+        mov     rdx, 0x00020001         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     rsi, r14        ; input_data
+        mov     rdi, r15        ; output_data
+.rowloop:
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rdx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     rdi, JSAMPROW [rdi]                     ; outptr
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        pxor    xmm2,xmm2
+        pxor    xmm3,xmm3
+        mov     rcx, SIZEOF_XMMWORD
+        jmp     short .downsample
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        pand    xmm0,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm0,xmm4
+        paddw   xmm1,xmm5
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm5,xmm3
+        pand    xmm2,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm3,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm2,xmm4
+        paddw   xmm3,xmm5
+
+        paddw   xmm0,xmm1
+        paddw   xmm2,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm7
+        psrlw   xmm0,2
+        psrlw   xmm2,2
+
+        packuswb xmm0,xmm2
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     rcx, byte SIZEOF_XMMWORD        ; outcol
+        add     rdx, byte 2*SIZEOF_XMMWORD      ; inptr0
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr1
+        add     rdi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    rcx,rcx
+        jnz     near .columnloop_r8
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+
+        add     rsi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     rax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jcsample-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jcsample-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,351 @@
+;
+; jcsample.asm - downsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Downsample pixel values of a single component.
+; This version handles the common case of 2:1 horizontal and 1:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v1_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v1_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_downsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v1_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     edx, 0x00010000         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={0, 1, 0, 1, 0, 1, 0, 1}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        pxor    xmm1,xmm1
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .downsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm2,xmm0
+        movdqa  xmm3,xmm1
+
+        pand    xmm0,xmm6
+        psrlw   xmm2,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm3,BYTE_BIT
+
+        paddw   xmm0,xmm2
+        paddw   xmm1,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+        psrlw   xmm0,1
+        psrlw   xmm1,1
+
+        packuswb xmm0,xmm1
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     ecx, byte SIZEOF_XMMWORD        ; outcol
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        test    ecx,ecx
+        jnz     short .columnloop_r8
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Downsample pixel values of a single component.
+; This version handles the standard case of 2:1 horizontal and 2:1 vertical,
+; without smoothing.
+;
+; GLOBAL(void)
+; jsimd_h2v2_downsample_sse2 (JDIMENSION image_width, int max_v_samp_factor,
+;                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+;                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+;
+
+%define img_width(b)    (b)+8           ; JDIMENSION image_width
+%define max_v_samp(b)   (b)+12          ; int max_v_samp_factor
+%define v_samp(b)       (b)+16          ; JDIMENSION v_samp_factor
+%define width_blks(b)   (b)+20          ; JDIMENSION width_blocks
+%define input_data(b)   (b)+24          ; JSAMPARRAY input_data
+%define output_data(b)  (b)+28          ; JSAMPARRAY output_data
+
+        align   16
+        global  EXTN(jsimd_h2v2_downsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_downsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     ecx, JDIMENSION [width_blks(ebp)]
+        shl     ecx,3                   ; imul ecx,DCTSIZE (ecx = output_cols)
+        jz      near .return
+
+        mov     edx, JDIMENSION [img_width(ebp)]
+
+        ; -- expand_right_edge
+
+        push    ecx
+        shl     ecx,1                           ; output_cols * 2
+        sub     ecx,edx
+        jle     short .expand_end
+
+        mov     eax, INT [max_v_samp(ebp)]
+        test    eax,eax
+        jle     short .expand_end
+
+        cld
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        alignx  16,7
+.expandloop:
+        push    eax
+        push    ecx
+
+        mov     edi, JSAMPROW [esi]
+        add     edi,edx
+        mov     al, JSAMPLE [edi-1]
+
+        rep stosb
+
+        pop     ecx
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        dec     eax
+        jg      short .expandloop
+
+.expand_end:
+        pop     ecx                             ; output_cols
+
+        ; -- h2v2_downsample
+
+        mov     eax, JDIMENSION [v_samp(ebp)]   ; rowctr
+        test    eax,eax
+        jle     near .return
+
+        mov     edx, 0x00020001         ; bias pattern
+        movd    xmm7,edx
+        pcmpeqw xmm6,xmm6
+        pshufd  xmm7,xmm7,0x00          ; xmm7={1, 2, 1, 2, 1, 2, 1, 2}
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, JSAMPARRAY [output_data(ebp)]      ; output_data
+        alignx  16,7
+.rowloop:
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     edx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1
+        mov     edi, JSAMPROW [edi]                     ; outptr
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     short .columnloop
+        alignx  16,7
+
+.columnloop_r8:
+        movdqa  xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        pxor    xmm2,xmm2
+        pxor    xmm3,xmm3
+        mov     ecx, SIZEOF_XMMWORD
+        jmp     short .downsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm0, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2, XMMWORD [edx+1*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+.downsample:
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        pand    xmm0,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm1,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm0,xmm4
+        paddw   xmm1,xmm5
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm5,xmm3
+        pand    xmm2,xmm6
+        psrlw   xmm4,BYTE_BIT
+        pand    xmm3,xmm6
+        psrlw   xmm5,BYTE_BIT
+        paddw   xmm2,xmm4
+        paddw   xmm3,xmm5
+
+        paddw   xmm0,xmm1
+        paddw   xmm2,xmm3
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm7
+        psrlw   xmm0,2
+        psrlw   xmm2,2
+
+        packuswb xmm0,xmm2
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+
+        sub     ecx, byte SIZEOF_XMMWORD        ; outcol
+        add     edx, byte 2*SIZEOF_XMMWORD      ; inptr0
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr1
+        add     edi, byte 1*SIZEOF_XMMWORD      ; outptr
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jae     near .columnloop
+        test    ecx,ecx
+        jnz     near .columnloop_r8
+
+        pop     esi
+        pop     edi
+        pop     ecx
+
+        add     esi, byte 2*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 1*SIZEOF_JSAMPROW     ; output_data
+        dec     eax                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,274 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdcolor-altivec.c */
+
+
+void jsimd_ycc_rgb_convert_altivec (JDIMENSION out_width, JSAMPIMAGE input_buf,
+                                    JDIMENSION input_row,
+                                    JSAMPARRAY output_buf, int num_rows)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = out_width * RGB_PIXELSIZE, num_cols;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, yl, yh, cbl, cbh,
+    crl, crh, rl, rh, gl, gh, bl, bh, g0w, g1w, g2w, g3w;
+  __vector int g0, g1, g2, g3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31};
+#endif
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+
+    for (num_cols = pitch; num_cols > 0;
+         num_cols -= RGB_PIXELSIZE * 16, outptr += RGB_PIXELSIZE * 16,
+         inptr0 += 16, inptr1 += 16, inptr2 += 16) {
+
+      y = vec_ld(0, inptr0);
+      /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+       * support unsigned vectors.
+       */
+      yl = (__vector signed short)VEC_UNPACKHU(y);
+      yh = (__vector signed short)VEC_UNPACKLU(y);
+
+      cb = vec_ld(0, inptr1);
+      cbl = (__vector signed short)VEC_UNPACKHU(cb);
+      cbh = (__vector signed short)VEC_UNPACKLU(cb);
+      cbl = vec_sub(cbl, pw_cj);
+      cbh = vec_sub(cbh, pw_cj);
+
+      cr = vec_ld(0, inptr2);
+      crl = (__vector signed short)VEC_UNPACKHU(cr);
+      crh = (__vector signed short)VEC_UNPACKLU(cr);
+      crl = vec_sub(crl, pw_cj);
+      crh = vec_sub(crh, pw_cj);
+
+      /* (Original)
+       * R = Y                + 1.40200 * Cr
+       * G = Y - 0.34414 * Cb - 0.71414 * Cr
+       * B = Y + 1.77200 * Cb
+       *
+       * (This implementation)
+       * R = Y                + 0.40200 * Cr + Cr
+       * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+       * B = Y - 0.22800 * Cb + Cb + Cb
+       */
+      bl = vec_add(cbl, cbl);
+      bh = vec_add(cbh, cbh);
+      bl = vec_madds(bl, pw_mf0228, pw_one);
+      bh = vec_madds(bh, pw_mf0228, pw_one);
+      bl = vec_sra(bl, (__vector unsigned short)pw_one);
+      bh = vec_sra(bh, (__vector unsigned short)pw_one);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, cbl);
+      bh = vec_add(bh, cbh);
+      bl = vec_add(bl, yl);
+      bh = vec_add(bh, yh);
+
+      rl = vec_add(crl, crl);
+      rh = vec_add(crh, crh);
+      rl = vec_madds(rl, pw_f0402, pw_one);
+      rh = vec_madds(rh, pw_f0402, pw_one);
+      rl = vec_sra(rl, (__vector unsigned short)pw_one);
+      rh = vec_sra(rh, (__vector unsigned short)pw_one);
+      rl = vec_add(rl, crl);
+      rh = vec_add(rh, crh);
+      rl = vec_add(rl, yl);
+      rh = vec_add(rh, yh);
+
+      g0w = vec_mergeh(cbl, crl);
+      g1w = vec_mergel(cbl, crl);
+      g0 = vec_msums(g0w, pw_mf0344_f0285, pd_onehalf);
+      g1 = vec_msums(g1w, pw_mf0344_f0285, pd_onehalf);
+      g2w = vec_mergeh(cbh, crh);
+      g3w = vec_mergel(cbh, crh);
+      g2 = vec_msums(g2w, pw_mf0344_f0285, pd_onehalf);
+      g3 = vec_msums(g3w, pw_mf0344_f0285, pd_onehalf);
+      /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+       * each dword into a new 16-bit vector, which is the equivalent of
+       * descaling the 32-bit results (right-shifting by 16 bits) and then
+       * packing them.
+       */
+      gl = vec_perm((__vector short)g0, (__vector short)g1, shift_pack_index);
+      gh = vec_perm((__vector short)g2, (__vector short)g3, shift_pack_index);
+      gl = vec_sub(gl, crl);
+      gh = vec_sub(gh, crh);
+      gl = vec_add(gl, yl);
+      gh = vec_add(gh, yh);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,405 @@
+;
+; jdcolext.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_mmx (JDIMENSION out_width,
+;                            JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                            JSAMPARRAY output_buf, int num_rows)
+;
+
+%define out_width(b)    (b)+8           ; JDIMENSION out_width
+%define input_buf(b)    (b)+12          ; JSAMPIMAGE input_buf
+%define input_row(b)    (b)+16          ; JDIMENSION input_row
+%define output_buf(b)   (b)+20          ; JSAMPARRAY output_buf
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_mmx) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [out_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [input_row(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        push    eax
+        push    edi
+        push    edx
+        push    ebx
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr0
+        mov     ebx, JSAMPROW [ebx]     ; inptr1
+        mov     edx, JSAMPROW [edx]     ; inptr2
+        mov     edi, JSAMPROW [edi]     ; outptr
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+        alignx  16,7
+.columnloop:
+
+        movq    mm5, MMWORD [ebx]       ; mm5=Cb(01234567)
+        movq    mm1, MMWORD [edx]       ; mm1=Cr(01234567)
+
+        pcmpeqw mm4,mm4
+        pcmpeqw mm7,mm7
+        psrlw   mm4,BYTE_BIT
+        psllw   mm7,7                   ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+        movq    mm0,mm4                 ; mm0=mm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    mm4,mm5                 ; mm4=Cb(0246)=CbE
+        psrlw   mm5,BYTE_BIT            ; mm5=Cb(1357)=CbO
+        pand    mm0,mm1                 ; mm0=Cr(0246)=CrE
+        psrlw   mm1,BYTE_BIT            ; mm1=Cr(1357)=CrO
+
+        paddw   mm4,mm7
+        paddw   mm5,mm7
+        paddw   mm0,mm7
+        paddw   mm1,mm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movq    mm2,mm4                 ; mm2=CbE
+        movq    mm3,mm5                 ; mm3=CbO
+        paddw   mm4,mm4                 ; mm4=2*CbE
+        paddw   mm5,mm5                 ; mm5=2*CbO
+        movq    mm6,mm0                 ; mm6=CrE
+        movq    mm7,mm1                 ; mm7=CrO
+        paddw   mm0,mm0                 ; mm0=2*CrE
+        paddw   mm1,mm1                 ; mm1=2*CrO
+
+        pmulhw  mm4,[GOTOFF(eax,PW_MF0228)]     ; mm4=(2*CbE * -FIX(0.22800))
+        pmulhw  mm5,[GOTOFF(eax,PW_MF0228)]     ; mm5=(2*CbO * -FIX(0.22800))
+        pmulhw  mm0,[GOTOFF(eax,PW_F0402)]      ; mm0=(2*CrE * FIX(0.40200))
+        pmulhw  mm1,[GOTOFF(eax,PW_F0402)]      ; mm1=(2*CrO * FIX(0.40200))
+
+        paddw   mm4,[GOTOFF(eax,PW_ONE)]
+        paddw   mm5,[GOTOFF(eax,PW_ONE)]
+        psraw   mm4,1                   ; mm4=(CbE * -FIX(0.22800))
+        psraw   mm5,1                   ; mm5=(CbO * -FIX(0.22800))
+        paddw   mm0,[GOTOFF(eax,PW_ONE)]
+        paddw   mm1,[GOTOFF(eax,PW_ONE)]
+        psraw   mm0,1                   ; mm0=(CrE * FIX(0.40200))
+        psraw   mm1,1                   ; mm1=(CrO * FIX(0.40200))
+
+        paddw   mm4,mm2
+        paddw   mm5,mm3
+        paddw   mm4,mm2                 ; mm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   mm5,mm3                 ; mm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   mm0,mm6                 ; mm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   mm1,mm7                 ; mm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(B-Y)E
+        movq    MMWORD [wk(1)], mm5     ; wk(1)=(B-Y)O
+
+        movq      mm4,mm2
+        movq      mm5,mm3
+        punpcklwd mm2,mm6
+        punpckhwd mm4,mm6
+        pmaddwd   mm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm4,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd mm3,mm7
+        punpckhwd mm5,mm7
+        pmaddwd   mm3,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     mm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm4,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm2,SCALEBITS
+        psrad     mm4,SCALEBITS
+        paddd     mm3,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm5,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm3,SCALEBITS
+        psrad     mm5,SCALEBITS
+
+        packssdw  mm2,mm4       ; mm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  mm3,mm5       ; mm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     mm2,mm6       ; mm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     mm3,mm7       ; mm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movq      mm5, MMWORD [esi]     ; mm5=Y(01234567)
+
+        pcmpeqw   mm4,mm4
+        psrlw     mm4,BYTE_BIT          ; mm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      mm4,mm5               ; mm4=Y(0246)=YE
+        psrlw     mm5,BYTE_BIT          ; mm5=Y(1357)=YO
+
+        paddw     mm0,mm4               ; mm0=((R-Y)E+YE)=RE=(R0 R2 R4 R6)
+        paddw     mm1,mm5               ; mm1=((R-Y)O+YO)=RO=(R1 R3 R5 R7)
+        packuswb  mm0,mm0               ; mm0=(R0 R2 R4 R6 ** ** ** **)
+        packuswb  mm1,mm1               ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+        paddw     mm2,mm4               ; mm2=((G-Y)E+YE)=GE=(G0 G2 G4 G6)
+        paddw     mm3,mm5               ; mm3=((G-Y)O+YO)=GO=(G1 G3 G5 G7)
+        packuswb  mm2,mm2               ; mm2=(G0 G2 G4 G6 ** ** ** **)
+        packuswb  mm3,mm3               ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+        paddw     mm4, MMWORD [wk(0)]   ; mm4=(YE+(B-Y)E)=BE=(B0 B2 B4 B6)
+        paddw     mm5, MMWORD [wk(1)]   ; mm5=(YO+(B-Y)O)=BO=(B1 B3 B5 B7)
+        packuswb  mm4,mm4               ; mm4=(B0 B2 B4 B6 ** ** ** **)
+        packuswb  mm5,mm5               ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmB               ; mmE=(20 01 22 03 24 05 26 07)
+        punpcklbw mmD,mmF               ; mmD=(11 21 13 23 15 25 17 27)
+
+        movq      mmG,mmA
+        movq      mmH,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 01 02 12 22 03)
+        punpckhwd mmG,mmE               ; mmG=(04 14 24 05 06 16 26 07)
+
+        psrlq     mmH,2*BYTE_BIT        ; mmH=(02 12 04 14 06 16 -- --)
+        psrlq     mmE,2*BYTE_BIT        ; mmE=(22 03 24 05 26 07 -- --)
+
+        movq      mmC,mmD
+        movq      mmB,mmD
+        punpcklwd mmD,mmH               ; mmD=(11 21 02 12 13 23 04 14)
+        punpckhwd mmC,mmH               ; mmC=(15 25 06 16 17 27 -- --)
+
+        psrlq     mmB,2*BYTE_BIT        ; mmB=(13 23 15 25 17 27 -- --)
+
+        movq      mmF,mmE
+        punpcklwd mmE,mmB               ; mmE=(22 03 13 23 24 05 15 25)
+        punpckhwd mmF,mmB               ; mmF=(26 07 17 27 -- -- -- --)
+
+        punpckldq mmA,mmD               ; mmA=(00 10 20 01 11 21 02 12)
+        punpckldq mmE,mmG               ; mmE=(22 03 13 23 04 14 24 05)
+        punpckldq mmC,mmF               ; mmC=(15 25 06 16 26 07 17 27)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        lea     ecx, [ecx+ecx*2]        ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_MMWORD
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    mmA,mmC
+        sub     ecx, byte 2*SIZEOF_MMWORD
+        add     edi, byte 2*SIZEOF_MMWORD
+        jmp     short .column_st4
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmE
+        sub     ecx, byte SIZEOF_MMWORD
+        add     edi, byte SIZEOF_MMWORD
+.column_st4:
+        movd    eax,mmA
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st2
+        mov     DWORD [edi+0*SIZEOF_DWORD], eax
+        psrlq   mmA,DWORD_BIT
+        movd    eax,mmA
+        sub     ecx, byte SIZEOF_DWORD
+        add     edi, byte SIZEOF_DWORD
+.column_st2:
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi+0*SIZEOF_WORD], ax
+        shr     eax,WORD_BIT
+        sub     ecx, byte SIZEOF_WORD
+        add     edi, byte SIZEOF_WORD
+.column_st1:
+        cmp     ecx, byte SIZEOF_BYTE
+        jb      short .nextrow
+        mov     BYTE [edi+0*SIZEOF_BYTE], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pcmpeqb   mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+        pxor      mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pxor      mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmG               ; mmE=(20 30 22 32 24 34 26 36)
+        punpcklbw mmB,mmD               ; mmB=(01 11 03 13 05 15 07 17)
+        punpcklbw mmF,mmH               ; mmF=(21 31 23 33 25 35 27 37)
+
+        movq      mmC,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 30 02 12 22 32)
+        punpckhwd mmC,mmE               ; mmC=(04 14 24 34 06 16 26 36)
+        movq      mmG,mmB
+        punpcklwd mmB,mmF               ; mmB=(01 11 21 31 03 13 23 33)
+        punpckhwd mmG,mmF               ; mmG=(05 15 25 35 07 17 27 37)
+
+        movq      mmD,mmA
+        punpckldq mmA,mmB               ; mmA=(00 10 20 30 01 11 21 31)
+        punpckhdq mmD,mmB               ; mmD=(02 12 22 32 03 13 23 33)
+        movq      mmH,mmC
+        punpckldq mmC,mmG               ; mmC=(04 14 24 34 05 15 25 35)
+        punpckhdq mmH,mmG               ; mmH=(06 16 26 36 07 17 27 37)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mmH
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        cmp     ecx, byte SIZEOF_MMWORD/2
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    mmA,mmC
+        movq    mmD,mmH
+        sub     ecx, byte SIZEOF_MMWORD/2
+        add     edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD/4
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmD
+        sub     ecx, byte SIZEOF_MMWORD/4
+        add     edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+        cmp     ecx, byte SIZEOF_MMWORD/8
+        jb      short .nextrow
+        movd    DWORD [edi+0*SIZEOF_DWORD], mmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        alignx  16,7
+
+.nextrow:
+        pop     ecx
+        pop     esi
+        pop     ebx
+        pop     edx
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        add     edi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,441 @@
+;
+; jdcolext.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009, 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width,
+;                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                             JSAMPARRAY output_buf, int num_rows)
+;
+
+; r10 = JDIMENSION out_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION input_row
+; r13 = JSAMPARRAY output_buf
+; r14 = int num_rows
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d        ; num_cols
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        lea     rsi, [rsi+rcx*SIZEOF_JSAMPROW]
+        lea     rbx, [rbx+rcx*SIZEOF_JSAMPROW]
+        lea     rdx, [rdx+rcx*SIZEOF_JSAMPROW]
+
+        pop     rcx
+
+        mov     rdi, r13
+        mov     eax, r14d
+        test    rax,rax
+        jle     near .return
+.rowloop:
+        push    rax
+        push    rdi
+        push    rdx
+        push    rbx
+        push    rsi
+        push    rcx                     ; col
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr0
+        mov     rbx, JSAMPROW [rbx]     ; inptr1
+        mov     rdx, JSAMPROW [rdx]     ; inptr2
+        mov     rdi, JSAMPROW [rdi]     ; outptr
+.columnloop:
+
+        movdqa  xmm5, XMMWORD [rbx]     ; xmm5=Cb(0123456789ABCDEF)
+        movdqa  xmm1, XMMWORD [rdx]     ; xmm1=Cr(0123456789ABCDEF)
+
+        pcmpeqw xmm4,xmm4
+        pcmpeqw xmm7,xmm7
+        psrlw   xmm4,BYTE_BIT
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+        movdqa  xmm0,xmm4               ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    xmm4,xmm5               ; xmm4=Cb(02468ACE)=CbE
+        psrlw   xmm5,BYTE_BIT           ; xmm5=Cb(13579BDF)=CbO
+        pand    xmm0,xmm1               ; xmm0=Cr(02468ACE)=CrE
+        psrlw   xmm1,BYTE_BIT           ; xmm1=Cr(13579BDF)=CrO
+
+        paddw   xmm4,xmm7
+        paddw   xmm5,xmm7
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm2,xmm4               ; xmm2=CbE
+        movdqa  xmm3,xmm5               ; xmm3=CbO
+        paddw   xmm4,xmm4               ; xmm4=2*CbE
+        paddw   xmm5,xmm5               ; xmm5=2*CbO
+        movdqa  xmm6,xmm0               ; xmm6=CrE
+        movdqa  xmm7,xmm1               ; xmm7=CrO
+        paddw   xmm0,xmm0               ; xmm0=2*CrE
+        paddw   xmm1,xmm1               ; xmm1=2*CrO
+
+        pmulhw  xmm4,[rel PW_MF0228]    ; xmm4=(2*CbE * -FIX(0.22800))
+        pmulhw  xmm5,[rel PW_MF0228]    ; xmm5=(2*CbO * -FIX(0.22800))
+        pmulhw  xmm0,[rel PW_F0402]     ; xmm0=(2*CrE * FIX(0.40200))
+        pmulhw  xmm1,[rel PW_F0402]     ; xmm1=(2*CrO * FIX(0.40200))
+
+        paddw   xmm4,[rel PW_ONE]
+        paddw   xmm5,[rel PW_ONE]
+        psraw   xmm4,1                  ; xmm4=(CbE * -FIX(0.22800))
+        psraw   xmm5,1                  ; xmm5=(CbO * -FIX(0.22800))
+        paddw   xmm0,[rel PW_ONE]
+        paddw   xmm1,[rel PW_ONE]
+        psraw   xmm0,1                  ; xmm0=(CrE * FIX(0.40200))
+        psraw   xmm1,1                  ; xmm1=(CrO * FIX(0.40200))
+
+        paddw   xmm4,xmm2
+        paddw   xmm5,xmm3
+        paddw   xmm4,xmm2               ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   xmm5,xmm3               ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   xmm0,xmm6               ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   xmm1,xmm7               ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=(B-Y)E
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(B-Y)O
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm5,xmm3
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm4,xmm6
+        pmaddwd   xmm2,[rel PW_MF0344_F0285]
+        pmaddwd   xmm4,[rel PW_MF0344_F0285]
+        punpcklwd xmm3,xmm7
+        punpckhwd xmm5,xmm7
+        pmaddwd   xmm3,[rel PW_MF0344_F0285]
+        pmaddwd   xmm5,[rel PW_MF0344_F0285]
+
+        paddd     xmm2,[rel PD_ONEHALF]
+        paddd     xmm4,[rel PD_ONEHALF]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm4,SCALEBITS
+        paddd     xmm3,[rel PD_ONEHALF]
+        paddd     xmm5,[rel PD_ONEHALF]
+        psrad     xmm3,SCALEBITS
+        psrad     xmm5,SCALEBITS
+
+        packssdw  xmm2,xmm4     ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  xmm3,xmm5     ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     xmm2,xmm6     ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     xmm3,xmm7     ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movdqa    xmm5, XMMWORD [rsi]   ; xmm5=Y(0123456789ABCDEF)
+
+        pcmpeqw   xmm4,xmm4
+        psrlw     xmm4,BYTE_BIT         ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      xmm4,xmm5             ; xmm4=Y(02468ACE)=YE
+        psrlw     xmm5,BYTE_BIT         ; xmm5=Y(13579BDF)=YO
+
+        paddw     xmm0,xmm4             ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm5             ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm4             ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm5             ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+        paddw     xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        lea     rcx, [rcx+rcx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     rcx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     rcx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_MMWORD
+        sub     rcx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_DWORD
+        sub     rcx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of rax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     rcx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [rdi], ax
+        add     rdi, byte SIZEOF_WORD
+        sub     rcx, byte SIZEOF_WORD
+        shr     rax, 16
+.column_st1:
+        ; Store the lower 1 byte of rax to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .nextrow
+        mov     BYTE [rdi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        cmp     rcx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_XMMWORD/8*4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .nextrow
+        movd    XMM_DWORD [rdi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.nextrow:
+        pop     rcx
+        pop     rsi
+        pop     rbx
+        pop     rdx
+        pop     rdi
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW
+        add     rbx, byte SIZEOF_JSAMPROW
+        add     rdx, byte SIZEOF_JSAMPROW
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     rax                             ; num_rows
+        jg      near .rowloop
+
+        sfence          ; flush the write buffer
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,460 @@
+;
+; jdcolext.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Convert some rows of samples to the output colorspace.
+;
+; GLOBAL(void)
+; jsimd_ycc_rgb_convert_sse2 (JDIMENSION out_width,
+;                             JSAMPIMAGE input_buf, JDIMENSION input_row,
+;                             JSAMPARRAY output_buf, int num_rows)
+;
+
+%define out_width(b)    (b)+8           ; JDIMENSION out_width
+%define input_buf(b)    (b)+12          ; JSAMPIMAGE input_buf
+%define input_row(b)    (b)+16          ; JDIMENSION input_row
+%define output_buf(b)   (b)+20          ; JSAMPARRAY output_buf
+%define num_rows(b)     (b)+24          ; int num_rows
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jsimd_ycc_rgb_convert_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [out_width(eax)]        ; num_cols
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [input_row(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+        lea     ebx, [ebx+ecx*SIZEOF_JSAMPROW]
+        lea     edx, [edx+ecx*SIZEOF_JSAMPROW]
+
+        pop     ecx
+
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     eax, INT [num_rows(eax)]
+        test    eax,eax
+        jle     near .return
+        alignx  16,7
+.rowloop:
+        push    eax
+        push    edi
+        push    edx
+        push    ebx
+        push    esi
+        push    ecx                     ; col
+
+        mov     esi, JSAMPROW [esi]     ; inptr0
+        mov     ebx, JSAMPROW [ebx]     ; inptr1
+        mov     edx, JSAMPROW [edx]     ; inptr2
+        mov     edi, JSAMPROW [edi]     ; outptr
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm5, XMMWORD [ebx]     ; xmm5=Cb(0123456789ABCDEF)
+        movdqa  xmm1, XMMWORD [edx]     ; xmm1=Cr(0123456789ABCDEF)
+
+        pcmpeqw xmm4,xmm4
+        pcmpeqw xmm7,xmm7
+        psrlw   xmm4,BYTE_BIT
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+        movdqa  xmm0,xmm4               ; xmm0=xmm4={0xFF 0x00 0xFF 0x00 ..}
+
+        pand    xmm4,xmm5               ; xmm4=Cb(02468ACE)=CbE
+        psrlw   xmm5,BYTE_BIT           ; xmm5=Cb(13579BDF)=CbO
+        pand    xmm0,xmm1               ; xmm0=Cr(02468ACE)=CrE
+        psrlw   xmm1,BYTE_BIT           ; xmm1=Cr(13579BDF)=CrO
+
+        paddw   xmm4,xmm7
+        paddw   xmm5,xmm7
+        paddw   xmm0,xmm7
+        paddw   xmm1,xmm7
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm2,xmm4               ; xmm2=CbE
+        movdqa  xmm3,xmm5               ; xmm3=CbO
+        paddw   xmm4,xmm4               ; xmm4=2*CbE
+        paddw   xmm5,xmm5               ; xmm5=2*CbO
+        movdqa  xmm6,xmm0               ; xmm6=CrE
+        movdqa  xmm7,xmm1               ; xmm7=CrO
+        paddw   xmm0,xmm0               ; xmm0=2*CrE
+        paddw   xmm1,xmm1               ; xmm1=2*CrO
+
+        pmulhw  xmm4,[GOTOFF(eax,PW_MF0228)]    ; xmm4=(2*CbE * -FIX(0.22800))
+        pmulhw  xmm5,[GOTOFF(eax,PW_MF0228)]    ; xmm5=(2*CbO * -FIX(0.22800))
+        pmulhw  xmm0,[GOTOFF(eax,PW_F0402)]     ; xmm0=(2*CrE * FIX(0.40200))
+        pmulhw  xmm1,[GOTOFF(eax,PW_F0402)]     ; xmm1=(2*CrO * FIX(0.40200))
+
+        paddw   xmm4,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm5,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm4,1                  ; xmm4=(CbE * -FIX(0.22800))
+        psraw   xmm5,1                  ; xmm5=(CbO * -FIX(0.22800))
+        paddw   xmm0,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm1,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm0,1                  ; xmm0=(CrE * FIX(0.40200))
+        psraw   xmm1,1                  ; xmm1=(CrO * FIX(0.40200))
+
+        paddw   xmm4,xmm2
+        paddw   xmm5,xmm3
+        paddw   xmm4,xmm2               ; xmm4=(CbE * FIX(1.77200))=(B-Y)E
+        paddw   xmm5,xmm3               ; xmm5=(CbO * FIX(1.77200))=(B-Y)O
+        paddw   xmm0,xmm6               ; xmm0=(CrE * FIX(1.40200))=(R-Y)E
+        paddw   xmm1,xmm7               ; xmm1=(CrO * FIX(1.40200))=(R-Y)O
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=(B-Y)E
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(B-Y)O
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm5,xmm3
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm4,xmm6
+        pmaddwd   xmm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm4,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd xmm3,xmm7
+        punpckhwd xmm5,xmm7
+        pmaddwd   xmm3,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     xmm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm4,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm4,SCALEBITS
+        paddd     xmm3,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm5,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm3,SCALEBITS
+        psrad     xmm5,SCALEBITS
+
+        packssdw  xmm2,xmm4     ; xmm2=CbE*-FIX(0.344)+CrE*FIX(0.285)
+        packssdw  xmm3,xmm5     ; xmm3=CbO*-FIX(0.344)+CrO*FIX(0.285)
+        psubw     xmm2,xmm6     ; xmm2=CbE*-FIX(0.344)+CrE*-FIX(0.714)=(G-Y)E
+        psubw     xmm3,xmm7     ; xmm3=CbO*-FIX(0.344)+CrO*-FIX(0.714)=(G-Y)O
+
+        movdqa    xmm5, XMMWORD [esi]   ; xmm5=Y(0123456789ABCDEF)
+
+        pcmpeqw   xmm4,xmm4
+        psrlw     xmm4,BYTE_BIT         ; xmm4={0xFF 0x00 0xFF 0x00 ..}
+        pand      xmm4,xmm5             ; xmm4=Y(02468ACE)=YE
+        psrlw     xmm5,BYTE_BIT         ; xmm5=Y(13579BDF)=YO
+
+        paddw     xmm0,xmm4             ; xmm0=((R-Y)E+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm5             ; xmm1=((R-Y)O+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm4             ; xmm2=((G-Y)E+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm5             ; xmm3=((G-Y)O+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4, XMMWORD [wk(0)] ; xmm4=(YE+(B-Y)E)=BE=B(02468ACE)
+        paddw     xmm5, XMMWORD [wk(1)] ; xmm5=(YO+(B-Y)O)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        lea     ecx, [ecx+ecx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     ecx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_MMWORD
+        sub     ecx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [edi], xmmA
+        add     edi, byte SIZEOF_DWORD
+        sub     ecx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of eax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi], ax
+        add     edi, byte SIZEOF_WORD
+        sub     ecx, byte SIZEOF_WORD
+        shr     eax, 16
+.column_st1:
+        ; Store the lower 1 byte of eax to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .nextrow
+        mov     BYTE [edi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .nextrow
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        cmp     ecx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_XMMWORD/8*4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .nextrow
+        movd    XMM_DWORD [edi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+        alignx  16,7
+
+.nextrow:
+        pop     ecx
+        pop     esi
+        pop     ebx
+        pop     edx
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW
+        add     ebx, byte SIZEOF_JSAMPROW
+        add     edx, byte SIZEOF_JSAMPROW
+        add     edi, byte SIZEOF_JSAMPROW       ; output_buf
+        dec     eax                             ; num_rows
+        jg      near .rowloop
+
+        sfence          ; flush the write buffer
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,96 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* YCC --> RGB CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344 22554              /* FIX(0.34414) */
+#define F_0_714 46802              /* FIX(0.71414) */
+#define F_1_402 91881              /* FIX(1.40200) */
+#define F_1_772 116130             /* FIX(1.77200) */
+#define F_0_402 (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285 (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228 (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 {0,1,8,2,3,10,4,5,12,6,7,14,16,17,24,18}
+#define RGB_INDEX1 {3,10,4,5,12,6,7,14,16,17,24,18,19,26,20,21}
+#define RGB_INDEX2 {12,6,7,14,16,17,24,18,19,26,20,21,28,22,23,30}
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGB_INDEX {0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extrgbx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 {8,1,0,10,3,2,12,5,4,14,7,6,24,17,16,26}
+#define RGB_INDEX1 {3,2,12,5,4,14,7,6,24,17,16,26,19,18,28,21}
+#define RGB_INDEX2 {4,14,7,6,24,17,16,26,19,18,28,21,20,30,23,22}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGB_INDEX {8,1,0,9,10,3,2,11,12,5,4,13,14,7,6,15}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extbgrx_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGB_INDEX {9,8,1,0,11,10,3,2,13,12,5,4,15,14,7,6}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extxbgr_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGB_INDEX {9,0,1,8,11,2,3,10,13,4,5,12,15,6,7,14}
+#define jsimd_ycc_rgb_convert_altivec jsimd_ycc_extxrgb_convert_altivec
+#include "jdcolext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_ycc_rgb_convert_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_mmx) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgb_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgbx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgrx_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxbgr_convert_mmx
+%include "jdcolext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxrgb_convert_mmx
+%include "jdcolext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdcolor-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdcolor-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,120 @@
+;
+; jdcolor.asm - colorspace conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_ycc_rgb_convert_sse2) PRIVATE
+
+EXTN(jconst_ycc_rgb_convert_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
+%include "jdcolext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
+%include "jdcolext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdct.inc glmark2-2021.02/src/libjpeg-turbo/simd/jdct.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdct.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdct.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,28 @@
+;
+; jdct.inc - private declarations for forward & reverse DCT subsystems
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; [TAB8]
+
+; Each IDCT routine is responsible for range-limiting its results and
+; converting them to unsigned form (0..MAXJSAMPLE).  The raw outputs could
+; be quite far out of range if the input data is corrupt, so a bulletproof
+; range-limiting step is required.  We use a mask-and-table-lookup method
+; to do the combined operations quickly.
+;
+%define RANGE_MASK  (MAXJSAMPLE * 4 + 3)  ; 2 bits wider than legal samples
+
+%define ROW(n,b,s)              ((b)+(n)*(s))
+%define COL(n,b,s)              ((b)+(n)*(s)*DCTSIZE)
+
+%define DWBLOCK(m,n,b,s)        ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_DWORD)
+%define MMBLOCK(m,n,b,s)        ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_MMWORD)
+%define XMMBLOCK(m,n,b,s)       ((b)+(m)*DCTSIZE*(s)+(n)*SIZEOF_XMMWORD)
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,108 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* MERGED YCC --> RGB CONVERSION AND UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_344 22554              /* FIX(0.34414) */
+#define F_0_714 46802              /* FIX(0.71414) */
+#define F_1_402 91881              /* FIX(1.40200) */
+#define F_1_772 116130             /* FIX(1.77200) */
+#define F_0_402 (F_1_402 - 65536)  /* FIX(1.40200) - FIX(1) */
+#define F_0_285 (65536 - F_0_714)  /* FIX(1) - FIX(0.71414) */
+#define F_0_228 (131072 - F_1_772) /* FIX(2) - FIX(1.77200) */
+
+#define SCALEBITS 16
+#define ONE_HALF (1 << (SCALEBITS - 1))
+
+#define RGB_INDEX0 {0,1,8,2,3,10,4,5,12,6,7,14,16,17,24,18}
+#define RGB_INDEX1 {3,10,4,5,12,6,7,14,16,17,24,18,19,26,20,21}
+#define RGB_INDEX2 {12,6,7,14,16,17,24,18,19,26,20,21,28,22,23,30}
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define RGB_INDEX {0,1,8,9,2,3,10,11,4,5,12,13,6,7,14,15}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extrgbx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extrgbx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define RGB_INDEX0 {8,1,0,10,3,2,12,5,4,14,7,6,24,17,16,26}
+#define RGB_INDEX1 {3,2,12,5,4,14,7,6,24,17,16,26,19,18,28,21}
+#define RGB_INDEX2 {4,14,7,6,24,17,16,26,19,18,28,21,20,30,23,22}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX0
+#undef RGB_INDEX1
+#undef RGB_INDEX2
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define RGB_INDEX {8,1,0,9,10,3,2,11,12,5,4,13,14,7,6,15}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extbgrx_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extbgrx_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define RGB_INDEX {9,8,1,0,11,10,3,2,13,12,5,4,15,14,7,6}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extxbgr_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extxbgr_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
+
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define RGB_INDEX {9,0,1,8,11,2,3,10,13,4,5,12,15,6,7,14}
+#define jsimd_h2v1_merged_upsample_altivec jsimd_h2v1_extxrgb_merged_upsample_altivec
+#define jsimd_h2v2_merged_upsample_altivec jsimd_h2v2_extxrgb_merged_upsample_altivec
+#include "jdmrgext-altivec.c"
+#undef RGB_PIXELSIZE
+#undef RGB_INDEX
+#undef jsimd_h2v1_merged_upsample_altivec
+#undef jsimd_h2v2_merged_upsample_altivec
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_mmx) PRIVATE
+
+EXTN(jconst_merged_upsample_mmx):
+
+PW_F0402        times 4 dw  F_0_402
+PW_MF0228       times 4 dw -F_0_228
+PW_MF0344_F0285 times 2 dw -F_0_344, F_0_285
+PW_ONE          times 4 dw  1
+PD_ONEHALF      times 2 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgbx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgbx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgrx_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgrx_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxbgr_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxbgr_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxrgb_merged_upsample_mmx
+%define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxrgb_merged_upsample_mmx
+%include "jdmrgext-mmx.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_sse2) PRIVATE
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2-64.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmerge-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmerge-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,126 @@
+;
+; jdmerge.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+
+%define SCALEBITS       16
+
+F_0_344 equ      22554                  ; FIX(0.34414)
+F_0_714 equ      46802                  ; FIX(0.71414)
+F_1_402 equ      91881                  ; FIX(1.40200)
+F_1_772 equ     116130                  ; FIX(1.77200)
+F_0_402 equ     (F_1_402 - 65536)       ; FIX(1.40200) - FIX(1)
+F_0_285 equ     ( 65536 - F_0_714)      ; FIX(1) - FIX(0.71414)
+F_0_228 equ     (131072 - F_1_772)      ; FIX(2) - FIX(1.77200)
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_merged_upsample_sse2) PRIVATE
+
+EXTN(jconst_merged_upsample_sse2):
+
+PW_F0402        times 8 dw  F_0_402
+PW_MF0228       times 8 dw -F_0_228
+PW_MF0344_F0285 times 4 dw -F_0_344, F_0_285
+PW_ONE          times 8 dw  1
+PD_ONEHALF      times 4 dd  1 << (SCALEBITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
+
+%undef RGB_RED
+%undef RGB_GREEN
+%undef RGB_BLUE
+%undef RGB_PIXELSIZE
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+%define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
+%define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
+%include "jdmrgext-sse2.asm"
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,323 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* This file is included by jdmerge-altivec.c */
+
+
+void jsimd_h2v1_merged_upsample_altivec (JDIMENSION output_width,
+                                         JSAMPIMAGE input_buf,
+                                         JDIMENSION in_row_group_ctr,
+                                         JSAMPARRAY output_buf)
+{
+  JSAMPROW outptr, inptr0, inptr1, inptr2;
+  int pitch = output_width * RGB_PIXELSIZE, num_cols, yloop;
+#if __BIG_ENDIAN__
+  int offset;
+#endif
+  unsigned char __attribute__((aligned(16))) tmpbuf[RGB_PIXELSIZE * 16];
+
+  __vector unsigned char rgb0, rgb1, rgb2, rgbx0, rgbx1, rgbx2, rgbx3,
+    y, cb, cr;
+#if __BIG_ENDIAN__
+  __vector unsigned char edgel, edgeh, edges, out0, out1, out2, out3;
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char out4;
+#endif
+#endif
+#if RGB_PIXELSIZE == 4
+  __vector unsigned char rgb3;
+#endif
+  __vector short rg0, rg1, rg2, rg3, bx0, bx1, bx2, bx3, ye, yo, cbl, cbh,
+    crl, crh, r_yl, r_yh, g_yl, g_yh, b_yl, b_yh, g_y0w, g_y1w, g_y2w, g_y3w,
+    rl, rh, gl, gh, bl, bh, re, ro, ge, go, be, bo;
+  __vector int g_y0, g_y1, g_y2, g_y3;
+
+  /* Constants
+   * NOTE: The >> 1 is to compensate for the fact that vec_madds() returns 17
+   * high-order bits, not 16.
+   */
+  __vector short pw_f0402 = { __8X(F_0_402 >> 1) },
+    pw_mf0228 = { __8X(-F_0_228 >> 1) },
+    pw_mf0344_f0285 = { __4X2(-F_0_344, F_0_285) },
+    pw_one = { __8X(1) }, pw_255 = { __8X(255) },
+    pw_cj = { __8X(CENTERJSAMPLE) };
+  __vector int pd_onehalf = { __4X(ONE_HALF) };
+  __vector unsigned char pb_zero = { __16X(0) },
+#if __BIG_ENDIAN__
+    shift_pack_index = {0,1,4,5,8,9,12,13,16,17,20,21,24,25,28,29},
+    even_index = {0,16,0,18,0,20,0,22,0,24,0,26,0,28,0,30},
+    odd_index = {0,17,0,19,0,21,0,23,0,25,0,27,0,29,0,31};
+#else
+    shift_pack_index = {2,3,6,7,10,11,14,15,18,19,22,23,26,27,30,31},
+    even_index = {16,0,18,0,20,0,22,0,24,0,26,0,28,0,30,0},
+    odd_index = {17,0,19,0,21,0,23,0,25,0,27,0,29,0,31,0};
+#endif
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  for (num_cols = pitch; num_cols > 0; inptr1 += 16, inptr2 += 16) {
+
+    cb = vec_ld(0, inptr1);
+    /* NOTE: We have to use vec_merge*() here because vec_unpack*() doesn't
+     * support unsigned vectors.
+     */
+    cbl = (__vector signed short)VEC_UNPACKHU(cb);
+    cbh = (__vector signed short)VEC_UNPACKLU(cb);
+    cbl = vec_sub(cbl, pw_cj);
+    cbh = vec_sub(cbh, pw_cj);
+
+    cr = vec_ld(0, inptr2);
+    crl = (__vector signed short)VEC_UNPACKHU(cr);
+    crh = (__vector signed short)VEC_UNPACKLU(cr);
+    crl = vec_sub(crl, pw_cj);
+    crh = vec_sub(crh, pw_cj);
+
+    /* (Original)
+     * R = Y                + 1.40200 * Cr
+     * G = Y - 0.34414 * Cb - 0.71414 * Cr
+     * B = Y + 1.77200 * Cb
+     *
+     * (This implementation)
+     * R = Y                + 0.40200 * Cr + Cr
+     * G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+     * B = Y - 0.22800 * Cb + Cb + Cb
+     */
+    b_yl = vec_add(cbl, cbl);
+    b_yh = vec_add(cbh, cbh);
+    b_yl = vec_madds(b_yl, pw_mf0228, pw_one);
+    b_yh = vec_madds(b_yh, pw_mf0228, pw_one);
+    b_yl = vec_sra(b_yl, (__vector unsigned short)pw_one);
+    b_yh = vec_sra(b_yh, (__vector unsigned short)pw_one);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+    b_yl = vec_add(b_yl, cbl);
+    b_yh = vec_add(b_yh, cbh);
+
+    r_yl = vec_add(crl, crl);
+    r_yh = vec_add(crh, crh);
+    r_yl = vec_madds(r_yl, pw_f0402, pw_one);
+    r_yh = vec_madds(r_yh, pw_f0402, pw_one);
+    r_yl = vec_sra(r_yl, (__vector unsigned short)pw_one);
+    r_yh = vec_sra(r_yh, (__vector unsigned short)pw_one);
+    r_yl = vec_add(r_yl, crl);
+    r_yh = vec_add(r_yh, crh);
+
+    g_y0w = vec_mergeh(cbl, crl);
+    g_y1w = vec_mergel(cbl, crl);
+    g_y0 = vec_msums(g_y0w, pw_mf0344_f0285, pd_onehalf);
+    g_y1 = vec_msums(g_y1w, pw_mf0344_f0285, pd_onehalf);
+    g_y2w = vec_mergeh(cbh, crh);
+    g_y3w = vec_mergel(cbh, crh);
+    g_y2 = vec_msums(g_y2w, pw_mf0344_f0285, pd_onehalf);
+    g_y3 = vec_msums(g_y3w, pw_mf0344_f0285, pd_onehalf);
+    /* Clever way to avoid 4 shifts + 2 packs.  This packs the high word from
+     * each dword into a new 16-bit vector, which is the equivalent of
+     * descaling the 32-bit results (right-shifting by 16 bits) and then
+     * packing them.
+     */
+    g_yl = vec_perm((__vector short)g_y0, (__vector short)g_y1,
+                    shift_pack_index);
+    g_yh = vec_perm((__vector short)g_y2, (__vector short)g_y3,
+                    shift_pack_index);
+    g_yl = vec_sub(g_yl, crl);
+    g_yh = vec_sub(g_yh, crh);
+
+    for (yloop = 0; yloop < 2 && num_cols > 0; yloop++,
+         num_cols -= RGB_PIXELSIZE * 16,
+         outptr += RGB_PIXELSIZE * 16, inptr0 += 16) {
+
+      y = vec_ld(0, inptr0);
+      ye = (__vector signed short)vec_perm(pb_zero, y, even_index);
+      yo = (__vector signed short)vec_perm(pb_zero, y, odd_index);
+
+      if (yloop == 0) {
+        be = vec_add(b_yl, ye);
+        bo = vec_add(b_yl, yo);
+        re = vec_add(r_yl, ye);
+        ro = vec_add(r_yl, yo);
+        ge = vec_add(g_yl, ye);
+        go = vec_add(g_yl, yo);
+      } else {
+        be = vec_add(b_yh, ye);
+        bo = vec_add(b_yh, yo);
+        re = vec_add(r_yh, ye);
+        ro = vec_add(r_yh, yo);
+        ge = vec_add(g_yh, ye);
+        go = vec_add(g_yh, yo);
+      }
+
+      rl = vec_mergeh(re, ro);
+      rh = vec_mergel(re, ro);
+      gl = vec_mergeh(ge, go);
+      gh = vec_mergel(ge, go);
+      bl = vec_mergeh(be, bo);
+      bh = vec_mergel(be, bo);
+
+      rg0 = vec_mergeh(rl, gl);
+      bx0 = vec_mergeh(bl, pw_255);
+      rg1 = vec_mergel(rl, gl);
+      bx1 = vec_mergel(bl, pw_255);
+      rg2 = vec_mergeh(rh, gh);
+      bx2 = vec_mergeh(bh, pw_255);
+      rg3 = vec_mergel(rh, gh);
+      bx3 = vec_mergel(bh, pw_255);
+
+      rgbx0 = vec_packsu(rg0, bx0);
+      rgbx1 = vec_packsu(rg1, bx1);
+      rgbx2 = vec_packsu(rg2, bx2);
+      rgbx3 = vec_packsu(rg3, bx3);
+
+#if RGB_PIXELSIZE == 3
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 R1 G1 B1 R2 G2 B2 R3 G3 B3 R4 G4 B4 R5
+       * rgb1 = G5 B5 R6 G6 B6 R7 G7 B7 R8 G8 B8 R9 G9 B9 Ra Ga
+       * rgb2 = Ba Rb Gb Bb Rc Gc Bc Rd Gd Bd Re Ge Be Rf Gf Bf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx1, (__vector unsigned char)RGB_INDEX0);
+      rgb1 = vec_perm(rgbx1, rgbx2, (__vector unsigned char)RGB_INDEX1);
+      rgb2 = vec_perm(rgbx2, rgbx3, (__vector unsigned char)RGB_INDEX2);
+#else
+      /* rgbx0 = R0 G0 R1 G1 R2 G2 R3 G3 B0 X0 B1 X1 B2 X2 B3 X3
+       * rgbx1 = R4 G4 R5 G5 R6 G6 R7 G7 B4 X4 B5 X5 B6 X6 B7 X7
+       * rgbx2 = R8 G8 R9 G9 Ra Ga Rb Gb B8 X8 B9 X9 Ba Xa Bb Xb
+       * rgbx3 = Rc Gc Rd Gd Re Ge Rf Gf Bc Xc Bd Xd Be Xe Bf Xf
+       *
+       * rgb0 = R0 G0 B0 X0 R1 G1 B1 X1 R2 G2 B2 X2 R3 G3 B3 X3
+       * rgb1 = R4 G4 B4 X4 R5 G5 B5 X5 R6 G6 B6 X6 R7 G7 B7 X7
+       * rgb2 = R8 G8 B8 X8 R9 G9 B9 X9 Ra Ga Ba Xa Rb Gb Bb Xb
+       * rgb3 = Rc Gc Bc Xc Rd Gd Bd Xd Re Ge Be Xe Rf Gf Bf Xf
+       */
+      rgb0 = vec_perm(rgbx0, rgbx0, (__vector unsigned char)RGB_INDEX);
+      rgb1 = vec_perm(rgbx1, rgbx1, (__vector unsigned char)RGB_INDEX);
+      rgb2 = vec_perm(rgbx2, rgbx2, (__vector unsigned char)RGB_INDEX);
+      rgb3 = vec_perm(rgbx3, rgbx3, (__vector unsigned char)RGB_INDEX);
+#endif
+
+#if __BIG_ENDIAN__
+      offset = (size_t)outptr & 15;
+      if (offset) {
+        __vector unsigned char unaligned_shift_index;
+        int bytes = num_cols + offset;
+
+        if (bytes < (RGB_PIXELSIZE + 1) * 16 && (bytes & 15)) {
+          /* Slow path to prevent buffer overwrite.  Since there is no way to
+           * write a partial AltiVec register, overwrite would occur on the
+           * last chunk of the last image row if the right edge is not on a
+           * 16-byte boundary.  It could also occur on other rows if the bytes
+           * per row is low enough.  Since we can't determine whether we're on
+           * the last image row, we have to assume every row is the last.
+           */
+          vec_st(rgb0, 0, tmpbuf);
+          vec_st(rgb1, 16, tmpbuf);
+          vec_st(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          vec_st(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          unaligned_shift_index = vec_lvsl(0, outptr);
+          edgel = vec_ld(0, outptr);
+          edgeh = vec_ld(min(num_cols - 1, RGB_PIXELSIZE * 16), outptr);
+          edges = vec_perm(edgeh, edgel, unaligned_shift_index);
+          unaligned_shift_index = vec_lvsr(0, outptr);
+          out0 = vec_perm(edges, rgb0, unaligned_shift_index);
+          out1 = vec_perm(rgb0, rgb1, unaligned_shift_index);
+          out2 = vec_perm(rgb1, rgb2, unaligned_shift_index);
+#if RGB_PIXELSIZE == 4
+          out3 = vec_perm(rgb2, rgb3, unaligned_shift_index);
+          out4 = vec_perm(rgb3, edges, unaligned_shift_index);
+#else
+          out3 = vec_perm(rgb2, edges, unaligned_shift_index);
+#endif
+          vec_st(out0, 0, outptr);
+          if (bytes > 16)
+            vec_st(out1, 16, outptr);
+          if (bytes > 32)
+            vec_st(out2, 32, outptr);
+          if (bytes > 48)
+            vec_st(out3, 48, outptr);
+#if RGB_PIXELSIZE == 4
+          if (bytes > 64)
+            vec_st(out4, 64, outptr);
+#endif
+        }
+      } else {
+#endif /* __BIG_ENDIAN__ */
+        if (num_cols < RGB_PIXELSIZE * 16 && (num_cols & 15)) {
+          /* Slow path */
+          VEC_ST(rgb0, 0, tmpbuf);
+          VEC_ST(rgb1, 16, tmpbuf);
+          VEC_ST(rgb2, 32, tmpbuf);
+#if RGB_PIXELSIZE == 4
+          VEC_ST(rgb3, 48, tmpbuf);
+#endif
+          memcpy(outptr, tmpbuf, min(num_cols, RGB_PIXELSIZE * 16));
+        } else {
+          /* Fast path */
+          VEC_ST(rgb0, 0, outptr);
+          if (num_cols > 16)
+            VEC_ST(rgb1, 16, outptr);
+          if (num_cols > 32)
+            VEC_ST(rgb2, 32, outptr);
+#if RGB_PIXELSIZE == 4
+          if (num_cols > 48)
+            VEC_ST(rgb3, 48, outptr);
+#endif
+        }
+#if __BIG_ENDIAN__
+      }
+#endif
+    }
+  }
+}
+
+
+void jsimd_h2v2_merged_upsample_altivec (JDIMENSION output_width,
+                                         JSAMPIMAGE input_buf,
+                                         JDIMENSION in_row_group_ctr,
+                                         JSAMPARRAY output_buf)
+{
+  JSAMPROW inptr, outptr;
+
+  inptr = input_buf[0][in_row_group_ctr];
+  outptr = output_buf[0];
+
+  input_buf[0][in_row_group_ctr] = input_buf[0][in_row_group_ctr * 2];
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf[0][in_row_group_ctr] = input_buf[0][in_row_group_ctr * 2 + 1];
+  output_buf[0] = output_buf[1];
+  jsimd_h2v1_merged_upsample_altivec(output_width, input_buf, in_row_group_ctr,
+                                     output_buf);
+
+  input_buf[0][in_row_group_ctr] = inptr;
+  output_buf[0] = outptr;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,464 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_mmx (JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          3
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [output_width(eax)]     ; col
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     edi, JSAMPROW [edi]                             ; outptr
+
+        pop     ecx                     ; col
+
+        alignx  16,7
+.columnloop:
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        movq      mm6, MMWORD [ebx]     ; mm6=Cb(01234567)
+        movq      mm7, MMWORD [edx]     ; mm7=Cr(01234567)
+
+        pxor      mm1,mm1               ; mm1=(all 0's)
+        pcmpeqw   mm3,mm3
+        psllw     mm3,7                 ; mm3={0xFF80 0xFF80 0xFF80 0xFF80}
+
+        movq      mm4,mm6
+        punpckhbw mm6,mm1               ; mm6=Cb(4567)=CbH
+        punpcklbw mm4,mm1               ; mm4=Cb(0123)=CbL
+        movq      mm0,mm7
+        punpckhbw mm7,mm1               ; mm7=Cr(4567)=CrH
+        punpcklbw mm0,mm1               ; mm0=Cr(0123)=CrL
+
+        paddw     mm6,mm3
+        paddw     mm4,mm3
+        paddw     mm7,mm3
+        paddw     mm0,mm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movq    mm5,mm6                 ; mm5=CbH
+        movq    mm2,mm4                 ; mm2=CbL
+        paddw   mm6,mm6                 ; mm6=2*CbH
+        paddw   mm4,mm4                 ; mm4=2*CbL
+        movq    mm1,mm7                 ; mm1=CrH
+        movq    mm3,mm0                 ; mm3=CrL
+        paddw   mm7,mm7                 ; mm7=2*CrH
+        paddw   mm0,mm0                 ; mm0=2*CrL
+
+        pmulhw  mm6,[GOTOFF(eax,PW_MF0228)]     ; mm6=(2*CbH * -FIX(0.22800))
+        pmulhw  mm4,[GOTOFF(eax,PW_MF0228)]     ; mm4=(2*CbL * -FIX(0.22800))
+        pmulhw  mm7,[GOTOFF(eax,PW_F0402)]      ; mm7=(2*CrH * FIX(0.40200))
+        pmulhw  mm0,[GOTOFF(eax,PW_F0402)]      ; mm0=(2*CrL * FIX(0.40200))
+
+        paddw   mm6,[GOTOFF(eax,PW_ONE)]
+        paddw   mm4,[GOTOFF(eax,PW_ONE)]
+        psraw   mm6,1                   ; mm6=(CbH * -FIX(0.22800))
+        psraw   mm4,1                   ; mm4=(CbL * -FIX(0.22800))
+        paddw   mm7,[GOTOFF(eax,PW_ONE)]
+        paddw   mm0,[GOTOFF(eax,PW_ONE)]
+        psraw   mm7,1                   ; mm7=(CrH * FIX(0.40200))
+        psraw   mm0,1                   ; mm0=(CrL * FIX(0.40200))
+
+        paddw   mm6,mm5
+        paddw   mm4,mm2
+        paddw   mm6,mm5                 ; mm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   mm4,mm2                 ; mm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   mm7,mm1                 ; mm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   mm0,mm3                 ; mm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movq    MMWORD [wk(0)], mm6     ; wk(0)=(B-Y)H
+        movq    MMWORD [wk(1)], mm7     ; wk(1)=(R-Y)H
+
+        movq      mm6,mm5
+        movq      mm7,mm2
+        punpcklwd mm5,mm1
+        punpckhwd mm6,mm1
+        pmaddwd   mm5,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm6,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd mm2,mm3
+        punpckhwd mm7,mm3
+        pmaddwd   mm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   mm7,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     mm5,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm6,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm5,SCALEBITS
+        psrad     mm6,SCALEBITS
+        paddd     mm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     mm7,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     mm2,SCALEBITS
+        psrad     mm7,SCALEBITS
+
+        packssdw  mm5,mm6       ; mm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  mm2,mm7       ; mm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     mm5,mm1       ; mm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     mm2,mm3       ; mm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movq    MMWORD [wk(2)], mm5     ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+        alignx  16,7
+
+.Yloop_2nd:
+        movq    mm0, MMWORD [wk(1)]     ; mm0=(R-Y)H
+        movq    mm2, MMWORD [wk(2)]     ; mm2=(G-Y)H
+        movq    mm4, MMWORD [wk(0)]     ; mm4=(B-Y)H
+        alignx  16,7
+
+.Yloop_1st:
+        movq    mm7, MMWORD [esi]       ; mm7=Y(01234567)
+
+        pcmpeqw mm6,mm6
+        psrlw   mm6,BYTE_BIT            ; mm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    mm6,mm7                 ; mm6=Y(0246)=YE
+        psrlw   mm7,BYTE_BIT            ; mm7=Y(1357)=YO
+
+        movq    mm1,mm0                 ; mm1=mm0=(R-Y)(L/H)
+        movq    mm3,mm2                 ; mm3=mm2=(G-Y)(L/H)
+        movq    mm5,mm4                 ; mm5=mm4=(B-Y)(L/H)
+
+        paddw     mm0,mm6               ; mm0=((R-Y)+YE)=RE=(R0 R2 R4 R6)
+        paddw     mm1,mm7               ; mm1=((R-Y)+YO)=RO=(R1 R3 R5 R7)
+        packuswb  mm0,mm0               ; mm0=(R0 R2 R4 R6 ** ** ** **)
+        packuswb  mm1,mm1               ; mm1=(R1 R3 R5 R7 ** ** ** **)
+
+        paddw     mm2,mm6               ; mm2=((G-Y)+YE)=GE=(G0 G2 G4 G6)
+        paddw     mm3,mm7               ; mm3=((G-Y)+YO)=GO=(G1 G3 G5 G7)
+        packuswb  mm2,mm2               ; mm2=(G0 G2 G4 G6 ** ** ** **)
+        packuswb  mm3,mm3               ; mm3=(G1 G3 G5 G7 ** ** ** **)
+
+        paddw     mm4,mm6               ; mm4=((B-Y)+YE)=BE=(B0 B2 B4 B6)
+        paddw     mm5,mm7               ; mm5=((B-Y)+YO)=BO=(B1 B3 B5 B7)
+        packuswb  mm4,mm4               ; mm4=(B0 B2 B4 B6 ** ** ** **)
+        packuswb  mm5,mm5               ; mm5=(B1 B3 B5 B7 ** ** ** **)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(** ** ** ** ** ** ** **), mmH=(** ** ** ** ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmB               ; mmE=(20 01 22 03 24 05 26 07)
+        punpcklbw mmD,mmF               ; mmD=(11 21 13 23 15 25 17 27)
+
+        movq      mmG,mmA
+        movq      mmH,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 01 02 12 22 03)
+        punpckhwd mmG,mmE               ; mmG=(04 14 24 05 06 16 26 07)
+
+        psrlq     mmH,2*BYTE_BIT        ; mmH=(02 12 04 14 06 16 -- --)
+        psrlq     mmE,2*BYTE_BIT        ; mmE=(22 03 24 05 26 07 -- --)
+
+        movq      mmC,mmD
+        movq      mmB,mmD
+        punpcklwd mmD,mmH               ; mmD=(11 21 02 12 13 23 04 14)
+        punpckhwd mmC,mmH               ; mmC=(15 25 06 16 17 27 -- --)
+
+        psrlq     mmB,2*BYTE_BIT        ; mmB=(13 23 15 25 17 27 -- --)
+
+        movq      mmF,mmE
+        punpcklwd mmE,mmB               ; mmE=(22 03 13 23 24 05 15 25)
+        punpckhwd mmF,mmB               ; mmF=(26 07 17 27 -- -- -- --)
+
+        punpckldq mmA,mmD               ; mmA=(00 10 20 01 11 21 02 12)
+        punpckldq mmE,mmG               ; mmE=(22 03 13 23 04 14 24 05)
+        punpckldq mmC,mmF               ; mmC=(15 25 06 16 26 07 17 27)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      near .endcolumn
+
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        lea     ecx, [ecx+ecx*2]        ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_MMWORD
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmE
+        movq    mmA,mmC
+        sub     ecx, byte 2*SIZEOF_MMWORD
+        add     edi, byte 2*SIZEOF_MMWORD
+        jmp     short .column_st4
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmE
+        sub     ecx, byte SIZEOF_MMWORD
+        add     edi, byte SIZEOF_MMWORD
+.column_st4:
+        movd    eax,mmA
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st2
+        mov     DWORD [edi+0*SIZEOF_DWORD], eax
+        psrlq   mmA,DWORD_BIT
+        movd    eax,mmA
+        sub     ecx, byte SIZEOF_DWORD
+        add     edi, byte SIZEOF_DWORD
+.column_st2:
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi+0*SIZEOF_WORD], ax
+        shr     eax,WORD_BIT
+        sub     ecx, byte SIZEOF_WORD
+        add     edi, byte SIZEOF_WORD
+.column_st1:
+        cmp     ecx, byte SIZEOF_BYTE
+        jb      short .endcolumn
+        mov     BYTE [edi+0*SIZEOF_BYTE], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pcmpeqb   mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%else
+        pxor      mm6,mm6               ; mm6=(X0 X2 X4 X6 ** ** ** **)
+        pxor      mm7,mm7               ; mm7=(X1 X3 X5 X7 ** ** ** **)
+%endif
+        ; mmA=(00 02 04 06 ** ** ** **), mmB=(01 03 05 07 ** ** ** **)
+        ; mmC=(10 12 14 16 ** ** ** **), mmD=(11 13 15 17 ** ** ** **)
+        ; mmE=(20 22 24 26 ** ** ** **), mmF=(21 23 25 27 ** ** ** **)
+        ; mmG=(30 32 34 36 ** ** ** **), mmH=(31 33 35 37 ** ** ** **)
+
+        punpcklbw mmA,mmC               ; mmA=(00 10 02 12 04 14 06 16)
+        punpcklbw mmE,mmG               ; mmE=(20 30 22 32 24 34 26 36)
+        punpcklbw mmB,mmD               ; mmB=(01 11 03 13 05 15 07 17)
+        punpcklbw mmF,mmH               ; mmF=(21 31 23 33 25 35 27 37)
+
+        movq      mmC,mmA
+        punpcklwd mmA,mmE               ; mmA=(00 10 20 30 02 12 22 32)
+        punpckhwd mmC,mmE               ; mmC=(04 14 24 34 06 16 26 36)
+        movq      mmG,mmB
+        punpcklwd mmB,mmF               ; mmB=(01 11 21 31 03 13 23 33)
+        punpckhwd mmG,mmF               ; mmG=(05 15 25 35 07 17 27 37)
+
+        movq      mmD,mmA
+        punpckldq mmA,mmB               ; mmA=(00 10 20 30 01 11 21 31)
+        punpckhdq mmD,mmB               ; mmD=(02 12 22 32 03 13 23 33)
+        movq      mmH,mmC
+        punpckldq mmC,mmG               ; mmC=(04 14 24 34 05 15 25 35)
+        punpckhdq mmH,mmG               ; mmH=(06 16 26 36 07 17 27 37)
+
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st16
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mmC
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mmH
+
+        sub     ecx, byte SIZEOF_MMWORD
+        jz      short .endcolumn
+
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_MMWORD   ; outptr
+        add     esi, byte SIZEOF_MMWORD                 ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_MMWORD                 ; inptr1
+        add     edx, byte SIZEOF_MMWORD                 ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st16:
+        cmp     ecx, byte SIZEOF_MMWORD/2
+        jb      short .column_st8
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mmD
+        movq    mmA,mmC
+        movq    mmD,mmH
+        sub     ecx, byte SIZEOF_MMWORD/2
+        add     edi, byte 2*SIZEOF_MMWORD
+.column_st8:
+        cmp     ecx, byte SIZEOF_MMWORD/4
+        jb      short .column_st4
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mmA
+        movq    mmA,mmD
+        sub     ecx, byte SIZEOF_MMWORD/4
+        add     edi, byte 1*SIZEOF_MMWORD
+.column_st4:
+        cmp     ecx, byte SIZEOF_MMWORD/8
+        jb      short .endcolumn
+        movd    DWORD [edi+0*SIZEOF_DWORD], mmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_mmx (JDIMENSION output_width,
+;                                 JSAMPIMAGE input_buf,
+;                                 JDIMENSION in_row_group_ctr,
+;                                 JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     eax, JDIMENSION [output_width(ebp)]
+
+        mov     edi, JSAMPIMAGE [input_buf(ebp)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(ebp)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(ebp)]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+
+        push    edx                     ; inptr2
+        push    ebx                     ; inptr1
+        push    esi                     ; inptr00
+        mov     ebx,esp
+
+        push    edi                     ; output_buf (outptr0)
+        push    ecx                     ; in_row_group_ctr
+        push    ebx                     ; input_buf
+        push    eax                     ; output_width
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+        add     esi, byte SIZEOF_JSAMPROW       ; inptr01
+        add     edi, byte SIZEOF_JSAMPROW       ; outptr1
+        mov     POINTER [ebx+0*SIZEOF_POINTER], esi
+        mov     POINTER [ebx-1*SIZEOF_POINTER], edi
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_mmx)
+
+        add     esp, byte 7*SIZEOF_DWORD
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,538 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (64-bit SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009, 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+; r10 = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          3
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     ecx, r10d        ; col
+        test    rcx,rcx
+        jz      near .return
+
+        push    rcx
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        mov     rdi, r13
+        mov     rsi, JSAMPROW [rsi+rcx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     rbx, JSAMPROW [rbx+rcx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     rdx, JSAMPROW [rdx+rcx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     rdi, JSAMPROW [rdi]                             ; outptr
+
+        pop     rcx                     ; col
+
+.columnloop:
+
+        movdqa    xmm6, XMMWORD [rbx]   ; xmm6=Cb(0123456789ABCDEF)
+        movdqa    xmm7, XMMWORD [rdx]   ; xmm7=Cr(0123456789ABCDEF)
+
+        pxor      xmm1,xmm1             ; xmm1=(all 0's)
+        pcmpeqw   xmm3,xmm3
+        psllw     xmm3,7                ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        movdqa    xmm4,xmm6
+        punpckhbw xmm6,xmm1             ; xmm6=Cb(89ABCDEF)=CbH
+        punpcklbw xmm4,xmm1             ; xmm4=Cb(01234567)=CbL
+        movdqa    xmm0,xmm7
+        punpckhbw xmm7,xmm1             ; xmm7=Cr(89ABCDEF)=CrH
+        punpcklbw xmm0,xmm1             ; xmm0=Cr(01234567)=CrL
+
+        paddw     xmm6,xmm3
+        paddw     xmm4,xmm3
+        paddw     xmm7,xmm3
+        paddw     xmm0,xmm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm5,xmm6               ; xmm5=CbH
+        movdqa  xmm2,xmm4               ; xmm2=CbL
+        paddw   xmm6,xmm6               ; xmm6=2*CbH
+        paddw   xmm4,xmm4               ; xmm4=2*CbL
+        movdqa  xmm1,xmm7               ; xmm1=CrH
+        movdqa  xmm3,xmm0               ; xmm3=CrL
+        paddw   xmm7,xmm7               ; xmm7=2*CrH
+        paddw   xmm0,xmm0               ; xmm0=2*CrL
+
+        pmulhw  xmm6,[rel PW_MF0228]    ; xmm6=(2*CbH * -FIX(0.22800))
+        pmulhw  xmm4,[rel PW_MF0228]    ; xmm4=(2*CbL * -FIX(0.22800))
+        pmulhw  xmm7,[rel PW_F0402]     ; xmm7=(2*CrH * FIX(0.40200))
+        pmulhw  xmm0,[rel PW_F0402]     ; xmm0=(2*CrL * FIX(0.40200))
+
+        paddw   xmm6,[rel PW_ONE]
+        paddw   xmm4,[rel PW_ONE]
+        psraw   xmm6,1                  ; xmm6=(CbH * -FIX(0.22800))
+        psraw   xmm4,1                  ; xmm4=(CbL * -FIX(0.22800))
+        paddw   xmm7,[rel PW_ONE]
+        paddw   xmm0,[rel PW_ONE]
+        psraw   xmm7,1                  ; xmm7=(CrH * FIX(0.40200))
+        psraw   xmm0,1                  ; xmm0=(CrL * FIX(0.40200))
+
+        paddw   xmm6,xmm5
+        paddw   xmm4,xmm2
+        paddw   xmm6,xmm5               ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   xmm4,xmm2               ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   xmm7,xmm1               ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   xmm0,xmm3               ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=(B-Y)H
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(R-Y)H
+
+        movdqa    xmm6,xmm5
+        movdqa    xmm7,xmm2
+        punpcklwd xmm5,xmm1
+        punpckhwd xmm6,xmm1
+        pmaddwd   xmm5,[rel PW_MF0344_F0285]
+        pmaddwd   xmm6,[rel PW_MF0344_F0285]
+        punpcklwd xmm2,xmm3
+        punpckhwd xmm7,xmm3
+        pmaddwd   xmm2,[rel PW_MF0344_F0285]
+        pmaddwd   xmm7,[rel PW_MF0344_F0285]
+
+        paddd     xmm5,[rel PD_ONEHALF]
+        paddd     xmm6,[rel PD_ONEHALF]
+        psrad     xmm5,SCALEBITS
+        psrad     xmm6,SCALEBITS
+        paddd     xmm2,[rel PD_ONEHALF]
+        paddd     xmm7,[rel PD_ONEHALF]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm7,SCALEBITS
+
+        packssdw  xmm5,xmm6     ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  xmm2,xmm7     ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     xmm5,xmm1     ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     xmm2,xmm3     ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movdqa  XMMWORD [wk(2)], xmm5   ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+
+.Yloop_2nd:
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(R-Y)H
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(G-Y)H
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(B-Y)H
+
+.Yloop_1st:
+        movdqa  xmm7, XMMWORD [rsi]     ; xmm7=Y(0123456789ABCDEF)
+
+        pcmpeqw xmm6,xmm6
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    xmm6,xmm7               ; xmm6=Y(02468ACE)=YE
+        psrlw   xmm7,BYTE_BIT           ; xmm7=Y(13579BDF)=YO
+
+        movdqa  xmm1,xmm0               ; xmm1=xmm0=(R-Y)(L/H)
+        movdqa  xmm3,xmm2               ; xmm3=xmm2=(G-Y)(L/H)
+        movdqa  xmm5,xmm4               ; xmm5=xmm4=(B-Y)(L/H)
+
+        paddw     xmm0,xmm6             ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm7             ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm6             ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm7             ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4,xmm6             ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+        paddw     xmm5,xmm7             ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        lea     rcx, [rcx+rcx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     rcx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     rcx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_MMWORD
+        sub     rcx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_DWORD
+        sub     rcx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of rax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     rcx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [rdi], ax
+        add     rdi, byte SIZEOF_WORD
+        sub     rcx, byte SIZEOF_WORD
+        shr     rax, 16
+.column_st1:
+        ; Store the lower 1 byte of rax to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .endcolumn
+        mov     BYTE [rdi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     rcx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    rdi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     rdi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     rcx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     rsi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     rbx, byte SIZEOF_XMMWORD        ; inptr1
+        add     rdx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+
+.column_st32:
+        cmp     rcx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmmD
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     rcx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     rcx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmmA
+        add     rdi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     rcx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     rcx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [rdi], xmmA
+        add     rdi, byte SIZEOF_XMMWORD/8*4
+        sub     rcx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    rcx, rcx
+        jz      short .endcolumn
+        movd    XMM_DWORD [rdi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        sfence          ; flush the write buffer
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+; r10 = JDIMENSION output_width
+; r11 = JSAMPIMAGE input_buf
+; r12 = JDIMENSION in_row_group_ctr
+; r13 = JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        mov     eax, r10d
+
+        mov     rdi, r11
+        mov     ecx, r12d
+        mov     rsi, JSAMPARRAY [rdi+0*SIZEOF_JSAMPARRAY]
+        mov     rbx, JSAMPARRAY [rdi+1*SIZEOF_JSAMPARRAY]
+        mov     rdx, JSAMPARRAY [rdi+2*SIZEOF_JSAMPARRAY]
+        mov     rdi, r13
+        lea     rsi, [rsi+rcx*SIZEOF_JSAMPROW]
+
+        push    rdx                     ; inptr2
+        push    rbx                     ; inptr1
+        push    rsi                     ; inptr00
+        mov     rbx,rsp
+
+        push    rdi
+        push    rcx
+        push    rax
+
+        %ifdef WIN64
+        mov r8, rcx
+        mov r9, rdi
+        mov rcx, rax
+        mov rdx, rbx
+        %else
+        mov rdx, rcx
+        mov rcx, rdi
+        mov     rdi, rax
+        mov rsi, rbx
+        %endif
+
+        call    EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        pop rax
+        pop rcx
+        pop rdi
+        pop rsi
+        pop rbx
+        pop rdx
+
+        add     rdi, byte SIZEOF_JSAMPROW       ; outptr1
+        add     rsi, byte SIZEOF_JSAMPROW       ; inptr01
+
+        push    rdx                     ; inptr2
+        push    rbx                     ; inptr1
+        push    rsi                     ; inptr00
+        mov     rbx,rsp
+
+        push    rdi
+        push    rcx
+        push    rax
+
+        %ifdef WIN64
+        mov r8, rcx
+        mov r9, rdi
+        mov rcx, rax
+        mov rdx, rbx
+        %else
+        mov rdx, rcx
+        mov rcx, rdi
+        mov     rdi, rax
+        mov rsi, rbx
+        %endif
+
+        call    EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        pop rax
+        pop rcx
+        pop rdi
+        pop rsi
+        pop rbx
+        pop rdx
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdmrgext-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdmrgext-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,519 @@
+;
+; jdmrgext.asm - merged upsampling/color conversion (SSE2)
+;
+; Copyright 2009, 2012 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2012 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jcolsamp.inc"
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v1_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          3
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void * gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_merged_upsample_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     ecx, JDIMENSION [output_width(eax)]     ; col
+        test    ecx,ecx
+        jz      near .return
+
+        push    ecx
+
+        mov     edi, JSAMPIMAGE [input_buf(eax)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(eax)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(eax)]
+        mov     esi, JSAMPROW [esi+ecx*SIZEOF_JSAMPROW]         ; inptr0
+        mov     ebx, JSAMPROW [ebx+ecx*SIZEOF_JSAMPROW]         ; inptr1
+        mov     edx, JSAMPROW [edx+ecx*SIZEOF_JSAMPROW]         ; inptr2
+        mov     edi, JSAMPROW [edi]                             ; outptr
+
+        pop     ecx                     ; col
+
+        alignx  16,7
+.columnloop:
+        movpic  eax, POINTER [gotptr]   ; load GOT address (eax)
+
+        movdqa    xmm6, XMMWORD [ebx]   ; xmm6=Cb(0123456789ABCDEF)
+        movdqa    xmm7, XMMWORD [edx]   ; xmm7=Cr(0123456789ABCDEF)
+
+        pxor      xmm1,xmm1             ; xmm1=(all 0's)
+        pcmpeqw   xmm3,xmm3
+        psllw     xmm3,7                ; xmm3={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        movdqa    xmm4,xmm6
+        punpckhbw xmm6,xmm1             ; xmm6=Cb(89ABCDEF)=CbH
+        punpcklbw xmm4,xmm1             ; xmm4=Cb(01234567)=CbL
+        movdqa    xmm0,xmm7
+        punpckhbw xmm7,xmm1             ; xmm7=Cr(89ABCDEF)=CrH
+        punpcklbw xmm0,xmm1             ; xmm0=Cr(01234567)=CrL
+
+        paddw     xmm6,xmm3
+        paddw     xmm4,xmm3
+        paddw     xmm7,xmm3
+        paddw     xmm0,xmm3
+
+        ; (Original)
+        ; R = Y                + 1.40200 * Cr
+        ; G = Y - 0.34414 * Cb - 0.71414 * Cr
+        ; B = Y + 1.77200 * Cb
+        ;
+        ; (This implementation)
+        ; R = Y                + 0.40200 * Cr + Cr
+        ; G = Y - 0.34414 * Cb + 0.28586 * Cr - Cr
+        ; B = Y - 0.22800 * Cb + Cb + Cb
+
+        movdqa  xmm5,xmm6               ; xmm5=CbH
+        movdqa  xmm2,xmm4               ; xmm2=CbL
+        paddw   xmm6,xmm6               ; xmm6=2*CbH
+        paddw   xmm4,xmm4               ; xmm4=2*CbL
+        movdqa  xmm1,xmm7               ; xmm1=CrH
+        movdqa  xmm3,xmm0               ; xmm3=CrL
+        paddw   xmm7,xmm7               ; xmm7=2*CrH
+        paddw   xmm0,xmm0               ; xmm0=2*CrL
+
+        pmulhw  xmm6,[GOTOFF(eax,PW_MF0228)]    ; xmm6=(2*CbH * -FIX(0.22800))
+        pmulhw  xmm4,[GOTOFF(eax,PW_MF0228)]    ; xmm4=(2*CbL * -FIX(0.22800))
+        pmulhw  xmm7,[GOTOFF(eax,PW_F0402)]     ; xmm7=(2*CrH * FIX(0.40200))
+        pmulhw  xmm0,[GOTOFF(eax,PW_F0402)]     ; xmm0=(2*CrL * FIX(0.40200))
+
+        paddw   xmm6,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm4,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm6,1                  ; xmm6=(CbH * -FIX(0.22800))
+        psraw   xmm4,1                  ; xmm4=(CbL * -FIX(0.22800))
+        paddw   xmm7,[GOTOFF(eax,PW_ONE)]
+        paddw   xmm0,[GOTOFF(eax,PW_ONE)]
+        psraw   xmm7,1                  ; xmm7=(CrH * FIX(0.40200))
+        psraw   xmm0,1                  ; xmm0=(CrL * FIX(0.40200))
+
+        paddw   xmm6,xmm5
+        paddw   xmm4,xmm2
+        paddw   xmm6,xmm5               ; xmm6=(CbH * FIX(1.77200))=(B-Y)H
+        paddw   xmm4,xmm2               ; xmm4=(CbL * FIX(1.77200))=(B-Y)L
+        paddw   xmm7,xmm1               ; xmm7=(CrH * FIX(1.40200))=(R-Y)H
+        paddw   xmm0,xmm3               ; xmm0=(CrL * FIX(1.40200))=(R-Y)L
+
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=(B-Y)H
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(R-Y)H
+
+        movdqa    xmm6,xmm5
+        movdqa    xmm7,xmm2
+        punpcklwd xmm5,xmm1
+        punpckhwd xmm6,xmm1
+        pmaddwd   xmm5,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm6,[GOTOFF(eax,PW_MF0344_F0285)]
+        punpcklwd xmm2,xmm3
+        punpckhwd xmm7,xmm3
+        pmaddwd   xmm2,[GOTOFF(eax,PW_MF0344_F0285)]
+        pmaddwd   xmm7,[GOTOFF(eax,PW_MF0344_F0285)]
+
+        paddd     xmm5,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm6,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm5,SCALEBITS
+        psrad     xmm6,SCALEBITS
+        paddd     xmm2,[GOTOFF(eax,PD_ONEHALF)]
+        paddd     xmm7,[GOTOFF(eax,PD_ONEHALF)]
+        psrad     xmm2,SCALEBITS
+        psrad     xmm7,SCALEBITS
+
+        packssdw  xmm5,xmm6     ; xmm5=CbH*-FIX(0.344)+CrH*FIX(0.285)
+        packssdw  xmm2,xmm7     ; xmm2=CbL*-FIX(0.344)+CrL*FIX(0.285)
+        psubw     xmm5,xmm1     ; xmm5=CbH*-FIX(0.344)+CrH*-FIX(0.714)=(G-Y)H
+        psubw     xmm2,xmm3     ; xmm2=CbL*-FIX(0.344)+CrL*-FIX(0.714)=(G-Y)L
+
+        movdqa  XMMWORD [wk(2)], xmm5   ; wk(2)=(G-Y)H
+
+        mov     al,2                    ; Yctr
+        jmp     short .Yloop_1st
+        alignx  16,7
+
+.Yloop_2nd:
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(R-Y)H
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(G-Y)H
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(B-Y)H
+        alignx  16,7
+
+.Yloop_1st:
+        movdqa  xmm7, XMMWORD [esi]     ; xmm7=Y(0123456789ABCDEF)
+
+        pcmpeqw xmm6,xmm6
+        psrlw   xmm6,BYTE_BIT           ; xmm6={0xFF 0x00 0xFF 0x00 ..}
+        pand    xmm6,xmm7               ; xmm6=Y(02468ACE)=YE
+        psrlw   xmm7,BYTE_BIT           ; xmm7=Y(13579BDF)=YO
+
+        movdqa  xmm1,xmm0               ; xmm1=xmm0=(R-Y)(L/H)
+        movdqa  xmm3,xmm2               ; xmm3=xmm2=(G-Y)(L/H)
+        movdqa  xmm5,xmm4               ; xmm5=xmm4=(B-Y)(L/H)
+
+        paddw     xmm0,xmm6             ; xmm0=((R-Y)+YE)=RE=R(02468ACE)
+        paddw     xmm1,xmm7             ; xmm1=((R-Y)+YO)=RO=R(13579BDF)
+        packuswb  xmm0,xmm0             ; xmm0=R(02468ACE********)
+        packuswb  xmm1,xmm1             ; xmm1=R(13579BDF********)
+
+        paddw     xmm2,xmm6             ; xmm2=((G-Y)+YE)=GE=G(02468ACE)
+        paddw     xmm3,xmm7             ; xmm3=((G-Y)+YO)=GO=G(13579BDF)
+        packuswb  xmm2,xmm2             ; xmm2=G(02468ACE********)
+        packuswb  xmm3,xmm3             ; xmm3=G(13579BDF********)
+
+        paddw     xmm4,xmm6             ; xmm4=((B-Y)+YE)=BE=B(02468ACE)
+        paddw     xmm5,xmm7             ; xmm5=((B-Y)+YO)=BO=B(13579BDF)
+        packuswb  xmm4,xmm4             ; xmm4=B(02468ACE********)
+        packuswb  xmm5,xmm5             ; xmm5=B(13579BDF********)
+
+%if RGB_PIXELSIZE == 3 ; ---------------
+
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(** ** ** ** ** ** ** ** **), xmmH=(** ** ** ** ** ** ** ** **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmB     ; xmmE=(20 01 22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F)
+        punpcklbw xmmD,xmmF     ; xmmD=(11 21 13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F)
+
+        movdqa    xmmG,xmmA
+        movdqa    xmmH,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 01 02 12 22 03 04 14 24 05 06 16 26 07)
+        punpckhwd xmmG,xmmE     ; xmmG=(08 18 28 09 0A 1A 2A 0B 0C 1C 2C 0D 0E 1E 2E 0F)
+
+        psrldq    xmmH,2        ; xmmH=(02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E -- --)
+        psrldq    xmmE,2        ; xmmE=(22 03 24 05 26 07 28 09 2A 0B 2C 0D 2E 0F -- --)
+
+        movdqa    xmmC,xmmD
+        movdqa    xmmB,xmmD
+        punpcklwd xmmD,xmmH     ; xmmD=(11 21 02 12 13 23 04 14 15 25 06 16 17 27 08 18)
+        punpckhwd xmmC,xmmH     ; xmmC=(19 29 0A 1A 1B 2B 0C 1C 1D 2D 0E 1E 1F 2F -- --)
+
+        psrldq    xmmB,2        ; xmmB=(13 23 15 25 17 27 19 29 1B 2B 1D 2D 1F 2F -- --)
+
+        movdqa    xmmF,xmmE
+        punpcklwd xmmE,xmmB     ; xmmE=(22 03 13 23 24 05 15 25 26 07 17 27 28 09 19 29)
+        punpckhwd xmmF,xmmB     ; xmmF=(2A 0B 1B 2B 2C 0D 1D 2D 2E 0F 1F 2F -- -- -- --)
+
+        pshufd    xmmH,xmmA,0x4E; xmmH=(04 14 24 05 06 16 26 07 00 10 20 01 02 12 22 03)
+        movdqa    xmmB,xmmE
+        punpckldq xmmA,xmmD     ; xmmA=(00 10 20 01 11 21 02 12 02 12 22 03 13 23 04 14)
+        punpckldq xmmE,xmmH     ; xmmE=(22 03 13 23 04 14 24 05 24 05 15 25 06 16 26 07)
+        punpckhdq xmmD,xmmB     ; xmmD=(15 25 06 16 26 07 17 27 17 27 08 18 28 09 19 29)
+
+        pshufd    xmmH,xmmG,0x4E; xmmH=(0C 1C 2C 0D 0E 1E 2E 0F 08 18 28 09 0A 1A 2A 0B)
+        movdqa    xmmB,xmmF
+        punpckldq xmmG,xmmC     ; xmmG=(08 18 28 09 19 29 0A 1A 0A 1A 2A 0B 1B 2B 0C 1C)
+        punpckldq xmmF,xmmH     ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 2C 0D 1D 2D 0E 1E 2E 0F)
+        punpckhdq xmmC,xmmB     ; xmmC=(1D 2D 0E 1E 2E 0F 1F 2F 1F 2F -- -- -- -- -- --)
+
+        punpcklqdq xmmA,xmmE    ; xmmA=(00 10 20 01 11 21 02 12 22 03 13 23 04 14 24 05)
+        punpcklqdq xmmD,xmmG    ; xmmD=(15 25 06 16 26 07 17 27 08 18 28 09 19 29 0A 1A)
+        punpcklqdq xmmF,xmmC    ; xmmF=(2A 0B 1B 2B 0C 1C 2C 0D 1D 2D 0E 1E 2E 0F 1F 2F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmF
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        lea     ecx, [ecx+ecx*2]                ; imul ecx, RGB_PIXELSIZE
+        cmp     ecx, byte 2*SIZEOF_XMMWORD
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmF
+        sub     ecx, byte 2*SIZEOF_XMMWORD
+        jmp     short .column_st15
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD
+.column_st15:
+        ; Store the lower 8 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_MMWORD
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_MMWORD
+        sub     ecx, byte SIZEOF_MMWORD
+        psrldq  xmmA, SIZEOF_MMWORD
+.column_st7:
+        ; Store the lower 4 bytes of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_DWORD
+        jb      short .column_st3
+        movd    XMM_DWORD [edi], xmmA
+        add     edi, byte SIZEOF_DWORD
+        sub     ecx, byte SIZEOF_DWORD
+        psrldq  xmmA, SIZEOF_DWORD
+.column_st3:
+        ; Store the lower 2 bytes of eax to the output when it has enough
+        ; space.
+        movd    eax, xmmA
+        cmp     ecx, byte SIZEOF_WORD
+        jb      short .column_st1
+        mov     WORD [edi], ax
+        add     edi, byte SIZEOF_WORD
+        sub     ecx, byte SIZEOF_WORD
+        shr     eax, 16
+.column_st1:
+        ; Store the lower 1 byte of eax to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .endcolumn
+        mov     BYTE [edi], al
+
+%else ; RGB_PIXELSIZE == 4 ; -----------
+
+%ifdef RGBX_FILLER_0XFF
+        pcmpeqb   xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pcmpeqb   xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%else
+        pxor      xmm6,xmm6             ; xmm6=XE=X(02468ACE********)
+        pxor      xmm7,xmm7             ; xmm7=XO=X(13579BDF********)
+%endif
+        ; xmmA=(00 02 04 06 08 0A 0C 0E **), xmmB=(01 03 05 07 09 0B 0D 0F **)
+        ; xmmC=(10 12 14 16 18 1A 1C 1E **), xmmD=(11 13 15 17 19 1B 1D 1F **)
+        ; xmmE=(20 22 24 26 28 2A 2C 2E **), xmmF=(21 23 25 27 29 2B 2D 2F **)
+        ; xmmG=(30 32 34 36 38 3A 3C 3E **), xmmH=(31 33 35 37 39 3B 3D 3F **)
+
+        punpcklbw xmmA,xmmC     ; xmmA=(00 10 02 12 04 14 06 16 08 18 0A 1A 0C 1C 0E 1E)
+        punpcklbw xmmE,xmmG     ; xmmE=(20 30 22 32 24 34 26 36 28 38 2A 3A 2C 3C 2E 3E)
+        punpcklbw xmmB,xmmD     ; xmmB=(01 11 03 13 05 15 07 17 09 19 0B 1B 0D 1D 0F 1F)
+        punpcklbw xmmF,xmmH     ; xmmF=(21 31 23 33 25 35 27 37 29 39 2B 3B 2D 3D 2F 3F)
+
+        movdqa    xmmC,xmmA
+        punpcklwd xmmA,xmmE     ; xmmA=(00 10 20 30 02 12 22 32 04 14 24 34 06 16 26 36)
+        punpckhwd xmmC,xmmE     ; xmmC=(08 18 28 38 0A 1A 2A 3A 0C 1C 2C 3C 0E 1E 2E 3E)
+        movdqa    xmmG,xmmB
+        punpcklwd xmmB,xmmF     ; xmmB=(01 11 21 31 03 13 23 33 05 15 25 35 07 17 27 37)
+        punpckhwd xmmG,xmmF     ; xmmG=(09 19 29 39 0B 1B 2B 3B 0D 1D 2D 3D 0F 1F 2F 3F)
+
+        movdqa    xmmD,xmmA
+        punpckldq xmmA,xmmB     ; xmmA=(00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33)
+        punpckhdq xmmD,xmmB     ; xmmD=(04 14 24 34 05 15 25 35 06 16 26 36 07 17 27 37)
+        movdqa    xmmH,xmmC
+        punpckldq xmmC,xmmG     ; xmmC=(08 18 28 38 09 19 29 39 0A 1A 2A 3A 0B 1B 2B 3B)
+        punpckhdq xmmH,xmmG     ; xmmH=(0C 1C 2C 3C 0D 1D 2D 3D 0E 1E 2E 3E 0F 1F 2F 3F)
+
+        cmp     ecx, byte SIZEOF_XMMWORD
+        jb      short .column_st32
+
+        test    edi, SIZEOF_XMMWORD-1
+        jnz     short .out1
+        ; --(aligned)-------------------
+        movntdq XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movntdq XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movntdq XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movntdq XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+        jmp     short .out0
+.out1:  ; --(unaligned)-----------------
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        movdqu  XMMWORD [edi+2*SIZEOF_XMMWORD], xmmC
+        movdqu  XMMWORD [edi+3*SIZEOF_XMMWORD], xmmH
+.out0:
+        add     edi, byte RGB_PIXELSIZE*SIZEOF_XMMWORD  ; outptr
+        sub     ecx, byte SIZEOF_XMMWORD
+        jz      near .endcolumn
+
+        add     esi, byte SIZEOF_XMMWORD        ; inptr0
+        dec     al                      ; Yctr
+        jnz     near .Yloop_2nd
+
+        add     ebx, byte SIZEOF_XMMWORD        ; inptr1
+        add     edx, byte SIZEOF_XMMWORD        ; inptr2
+        jmp     near .columnloop
+        alignx  16,7
+
+.column_st32:
+        cmp     ecx, byte SIZEOF_XMMWORD/2
+        jb      short .column_st16
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        movdqu  XMMWORD [edi+1*SIZEOF_XMMWORD], xmmD
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        movdqa  xmmA,xmmC
+        movdqa  xmmD,xmmH
+        sub     ecx, byte SIZEOF_XMMWORD/2
+.column_st16:
+        cmp     ecx, byte SIZEOF_XMMWORD/4
+        jb      short .column_st15
+        movdqu  XMMWORD [edi+0*SIZEOF_XMMWORD], xmmA
+        add     edi, byte SIZEOF_XMMWORD        ; outptr
+        movdqa  xmmA,xmmD
+        sub     ecx, byte SIZEOF_XMMWORD/4
+.column_st15:
+        ; Store two pixels (8 bytes) of xmmA to the output when it has enough
+        ; space.
+        cmp     ecx, byte SIZEOF_XMMWORD/8
+        jb      short .column_st7
+        movq    XMM_MMWORD [edi], xmmA
+        add     edi, byte SIZEOF_XMMWORD/8*4
+        sub     ecx, byte SIZEOF_XMMWORD/8
+        psrldq  xmmA, SIZEOF_XMMWORD/8*4
+.column_st7:
+        ; Store one pixel (4 bytes) of xmmA to the output when it has enough
+        ; space.
+        test    ecx, ecx
+        jz      short .endcolumn
+        movd    XMM_DWORD [edi], xmmA
+
+%endif ; RGB_PIXELSIZE ; ---------------
+
+.endcolumn:
+        sfence          ; flush the write buffer
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+;
+; GLOBAL(void)
+; jsimd_h2v2_merged_upsample_sse2 (JDIMENSION output_width,
+;                                  JSAMPIMAGE input_buf,
+;                                  JDIMENSION in_row_group_ctr,
+;                                  JSAMPARRAY output_buf);
+;
+
+%define output_width(b) (b)+8                   ; JDIMENSION output_width
+%define input_buf(b)            (b)+12          ; JSAMPIMAGE input_buf
+%define in_row_group_ctr(b)     (b)+16          ; JDIMENSION in_row_group_ctr
+%define output_buf(b)           (b)+20          ; JSAMPARRAY output_buf
+
+        align   16
+        global  EXTN(jsimd_h2v2_merged_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_merged_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     eax, POINTER [output_width(ebp)]
+
+        mov     edi, JSAMPIMAGE [input_buf(ebp)]
+        mov     ecx, JDIMENSION [in_row_group_ctr(ebp)]
+        mov     esi, JSAMPARRAY [edi+0*SIZEOF_JSAMPARRAY]
+        mov     ebx, JSAMPARRAY [edi+1*SIZEOF_JSAMPARRAY]
+        mov     edx, JSAMPARRAY [edi+2*SIZEOF_JSAMPARRAY]
+        mov     edi, JSAMPARRAY [output_buf(ebp)]
+        lea     esi, [esi+ecx*SIZEOF_JSAMPROW]
+
+        push    edx                     ; inptr2
+        push    ebx                     ; inptr1
+        push    esi                     ; inptr00
+        mov     ebx,esp
+
+        push    edi                     ; output_buf (outptr0)
+        push    ecx                     ; in_row_group_ctr
+        push    ebx                     ; input_buf
+        push    eax                     ; output_width
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        add     esi, byte SIZEOF_JSAMPROW       ; inptr01
+        add     edi, byte SIZEOF_JSAMPROW       ; outptr1
+        mov     POINTER [ebx+0*SIZEOF_POINTER], esi
+        mov     POINTER [ebx-1*SIZEOF_POINTER], edi
+
+        call    near EXTN(jsimd_h2v1_merged_upsample_sse2)
+
+        add     esp, byte 7*SIZEOF_DWORD
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,392 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* CHROMA UPSAMPLING */
+
+#include "jsimd_altivec.h"
+
+
+void
+jsimd_h2v1_fancy_upsample_altivec (int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char this0, last0, p_last0, next0 = {0}, p_next0,
+    out;
+  __vector short this0e, this0o, this0l, this0h, last0l, last0h,
+    next0l, next0h, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) }, pb_three = { __16X(3) },
+    last_index_col0 = {0,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14},
+    last_index = {15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30},
+    next_index = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16},
+    next_index_lastcol = {1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,15},
+#if __BIG_ENDIAN__
+    merge_pack_index = {1,17,3,19,5,21,7,23,9,25,11,27,13,29,15,31};
+#else
+    merge_pack_index = {0,16,2,18,4,20,6,22,8,24,10,26,12,28,14,30};
+#endif
+  __vector short pw_one = { __8X(1) }, pw_two = { __8X(2) };
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+
+    if (downsampled_width & 15)
+      inptr[downsampled_width] = inptr[downsampled_width - 1];
+
+    this0 = vec_ld(0, inptr);
+    p_last0 = vec_perm(this0, this0, last_index_col0);
+    last0 = this0;
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr += 16, outptr += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_last0 = vec_perm(last0, this0, last_index);
+        last0 = this0;
+      }
+
+      if (incol <= 16)
+        p_next0 = vec_perm(this0, this0, next_index_lastcol);
+      else {
+        next0 = vec_ld(16, inptr);
+        p_next0 = vec_perm(this0, next0, next_index);
+      }
+
+      this0e = (__vector short)vec_mule(this0, pb_three);
+      this0o = (__vector short)vec_mulo(this0, pb_three);
+      this0l = vec_mergeh(this0e, this0o);
+      this0h = vec_mergel(this0e, this0o);
+
+      last0l = (__vector short)VEC_UNPACKHU(p_last0);
+      last0h = (__vector short)VEC_UNPACKLU(p_last0);
+      last0l = vec_add(last0l, pw_one);
+
+      next0l = (__vector short)VEC_UNPACKHU(p_next0);
+      next0h = (__vector short)VEC_UNPACKLU(p_next0);
+      next0l = vec_add(next0l, pw_two);
+
+      outle = vec_add(this0l, last0l);
+      outlo = vec_add(this0l, next0l);
+      outle = vec_sr(outle, (__vector unsigned short)pw_two);
+      outlo = vec_sr(outlo, (__vector unsigned short)pw_two);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr);
+
+      if (incol > 8) {
+        last0h = vec_add(last0h, pw_one);
+        next0h = vec_add(next0h, pw_two);
+
+        outhe = vec_add(this0h, last0h);
+        outho = vec_add(this0h, next0h);
+        outhe = vec_sr(outhe, (__vector unsigned short)pw_two);
+        outho = vec_sr(outho, (__vector unsigned short)pw_two);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr);
+      }
+
+      this0 = next0;
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_fancy_upsample_altivec (int max_v_samp_factor,
+                                   JDIMENSION downsampled_width,
+                                   JSAMPARRAY input_data,
+                                   JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr_1, inptr0, inptr1, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char this_1, this0, this1, out;
+  __vector short this_1l, this_1h, this0l, this0h, this1l, this1h,
+    lastcolsum_1h, lastcolsum1h,
+    p_lastcolsum_1l, p_lastcolsum_1h, p_lastcolsum1l, p_lastcolsum1h,
+    thiscolsum_1l, thiscolsum_1h, thiscolsum1l, thiscolsum1h,
+    nextcolsum_1l = {0}, nextcolsum_1h = {0},
+    nextcolsum1l = {0}, nextcolsum1h = {0},
+    p_nextcolsum_1l, p_nextcolsum_1h, p_nextcolsum1l, p_nextcolsum1h,
+    tmpl, tmph, outle, outhe, outlo, outho;
+
+  /* Constants */
+  __vector unsigned char pb_zero = { __16X(0) },
+    last_index_col0 = {0,1,0,1,2,3,4,5,6,7,8,9,10,11,12,13},
+    last_index={14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29},
+    next_index = {2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17},
+    next_index_lastcol = {2,3,4,5,6,7,8,9,10,11,12,13,14,15,14,15},
+#if __BIG_ENDIAN__
+    merge_pack_index = {1,17,3,19,5,21,7,23,9,25,11,27,13,29,15,31};
+#else
+    merge_pack_index = {0,16,2,18,4,20,6,22,8,24,10,26,12,28,14,30};
+#endif
+  __vector short pw_zero = { __8X(0) }, pw_three = { __8X(3) },
+    pw_seven = { __8X(7) }, pw_eight = { __8X(8) };
+  __vector unsigned short pw_four = { __8X(4) };
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr_1 = input_data[inrow - 1];
+    inptr0 = input_data[inrow];
+    inptr1 = input_data[inrow + 1];
+    outptr0 = output_data[outrow++];
+    outptr1 = output_data[outrow++];
+
+    if (downsampled_width & 15) {
+      inptr_1[downsampled_width] = inptr_1[downsampled_width - 1];
+      inptr0[downsampled_width] = inptr0[downsampled_width - 1];
+      inptr1[downsampled_width] = inptr1[downsampled_width - 1];
+    }
+
+    this0 = vec_ld(0, inptr0);
+    this0l = (__vector short)VEC_UNPACKHU(this0);
+    this0h = (__vector short)VEC_UNPACKLU(this0);
+    this0l = vec_mladd(this0l, pw_three, pw_zero);
+    this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+    this_1 = vec_ld(0, inptr_1);
+    this_1l = (__vector short)VEC_UNPACKHU(this_1);
+    this_1h = (__vector short)VEC_UNPACKLU(this_1);
+    thiscolsum_1l = vec_add(this0l, this_1l);
+    thiscolsum_1h = vec_add(this0h, this_1h);
+    lastcolsum_1h = thiscolsum_1h;
+    p_lastcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1l, last_index_col0);
+    p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+
+    this1 = vec_ld(0, inptr1);
+    this1l = (__vector short)VEC_UNPACKHU(this1);
+    this1h = (__vector short)VEC_UNPACKLU(this1);
+    thiscolsum1l = vec_add(this0l, this1l);
+    thiscolsum1h = vec_add(this0h, this1h);
+    lastcolsum1h = thiscolsum1h;
+    p_lastcolsum1l = vec_perm(thiscolsum1l, thiscolsum1l, last_index_col0);
+    p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+
+    for (incol = downsampled_width; incol > 0;
+         incol -= 16, inptr_1 += 16, inptr0 += 16, inptr1 += 16,
+         outptr0 += 32, outptr1 += 32) {
+
+      if (downsampled_width - incol > 0) {
+        p_lastcolsum_1l = vec_perm(lastcolsum_1h, thiscolsum_1l, last_index);
+        p_lastcolsum_1h = vec_perm(thiscolsum_1l, thiscolsum_1h, last_index);
+        p_lastcolsum1l = vec_perm(lastcolsum1h, thiscolsum1l, last_index);
+        p_lastcolsum1h = vec_perm(thiscolsum1l, thiscolsum1h, last_index);
+        lastcolsum_1h = thiscolsum_1h;  lastcolsum1h = thiscolsum1h;
+      }
+
+      if (incol <= 16) {
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, thiscolsum_1h,
+                                   next_index_lastcol);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, thiscolsum1h,
+                                  next_index_lastcol);
+      } else {
+        this0 = vec_ld(16, inptr0);
+        this0l = (__vector short)VEC_UNPACKHU(this0);
+        this0h = (__vector short)VEC_UNPACKLU(this0);
+        this0l = vec_mladd(this0l, pw_three, pw_zero);
+        this0h = vec_mladd(this0h, pw_three, pw_zero);
+
+        this_1 = vec_ld(16, inptr_1);
+        this_1l = (__vector short)VEC_UNPACKHU(this_1);
+        this_1h = (__vector short)VEC_UNPACKLU(this_1);
+        nextcolsum_1l = vec_add(this0l, this_1l);
+        nextcolsum_1h = vec_add(this0h, this_1h);
+        p_nextcolsum_1l = vec_perm(thiscolsum_1l, thiscolsum_1h, next_index);
+        p_nextcolsum_1h = vec_perm(thiscolsum_1h, nextcolsum_1l, next_index);
+
+        this1 = vec_ld(16, inptr1);
+        this1l = (__vector short)VEC_UNPACKHU(this1);
+        this1h = (__vector short)VEC_UNPACKLU(this1);
+        nextcolsum1l = vec_add(this0l, this1l);
+        nextcolsum1h = vec_add(this0h, this1h);
+        p_nextcolsum1l = vec_perm(thiscolsum1l, thiscolsum1h, next_index);
+        p_nextcolsum1h = vec_perm(thiscolsum1h, nextcolsum1l, next_index);
+      }
+
+      /* Process the upper row */
+
+      tmpl = vec_mladd(thiscolsum_1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum_1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum_1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr0);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum_1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum_1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum_1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr0);
+      }
+
+      /* Process the lower row */
+
+      tmpl = vec_mladd(thiscolsum1l, pw_three, pw_zero);
+      outle = vec_add(tmpl, p_lastcolsum1l);
+      outle = vec_add(outle, pw_eight);
+      outle = vec_sr(outle, pw_four);
+
+      outlo = vec_add(tmpl, p_nextcolsum1l);
+      outlo = vec_add(outlo, pw_seven);
+      outlo = vec_sr(outlo, pw_four);
+
+      out = vec_perm((__vector unsigned char)outle,
+                     (__vector unsigned char)outlo, merge_pack_index);
+      vec_st(out, 0, outptr1);
+
+      if (incol > 8) {
+        tmph = vec_mladd(thiscolsum1h, pw_three, pw_zero);
+        outhe = vec_add(tmph, p_lastcolsum1h);
+        outhe = vec_add(outhe, pw_eight);
+        outhe = vec_sr(outhe, pw_four);
+
+        outho = vec_add(tmph, p_nextcolsum1h);
+        outho = vec_add(outho, pw_seven);
+        outho = vec_sr(outho, pw_four);
+
+        out = vec_perm((__vector unsigned char)outhe,
+                       (__vector unsigned char)outho, merge_pack_index);
+        vec_st(out, 16, outptr1);
+      }
+
+      thiscolsum_1l = nextcolsum_1l;  thiscolsum_1h = nextcolsum_1h;
+      thiscolsum1l = nextcolsum1l;  thiscolsum1h = nextcolsum1h;
+    }
+  }
+}
+
+
+/* These are rarely used (mainly just for decompressing YCCK images) */
+
+void
+jsimd_h2v1_upsample_altivec (int max_v_samp_factor,
+                             JDIMENSION output_width,
+                             JSAMPARRAY input_data,
+                             JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr;
+  int inrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0; inrow < max_v_samp_factor; inrow++) {
+    inptr = input_data[inrow];
+    outptr = output_data[inrow];
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr);
+      vec_st(inh, 16, outptr);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr);
+        vec_st(inh, 48, outptr);
+      }
+    }
+  }
+}
+
+
+void
+jsimd_h2v2_upsample_altivec (int max_v_samp_factor,
+                             JDIMENSION output_width,
+                             JSAMPARRAY input_data,
+                             JSAMPARRAY *output_data_ptr)
+{
+  JSAMPARRAY output_data = *output_data_ptr;
+  JSAMPROW inptr, outptr0, outptr1;
+  int inrow, outrow, incol;
+
+  __vector unsigned char in, inl, inh;
+
+  for (inrow = 0, outrow = 0; outrow < max_v_samp_factor; inrow++) {
+
+    inptr = input_data[inrow];
+    outptr0 = output_data[outrow++];
+    outptr1 = output_data[outrow++];
+
+    for (incol = (output_width + 31) & (~31); incol > 0;
+         incol -= 64, inptr += 32, outptr0 += 64, outptr1 += 64) {
+
+      in = vec_ld(0, inptr);
+      inl = vec_mergeh(in, in);
+      inh = vec_mergel(in, in);
+
+      vec_st(inl, 0, outptr0);
+      vec_st(inl, 0, outptr1);
+
+      vec_st(inh, 16, outptr0);
+      vec_st(inh, 16, outptr1);
+
+      if (incol > 32) {
+        in = vec_ld(16, inptr);
+        inl = vec_mergeh(in, in);
+        inh = vec_mergel(in, in);
+
+        vec_st(inl, 32, outptr0);
+        vec_st(inl, 32, outptr1);
+
+        vec_st(inh, 48, outptr0);
+        vec_st(inh, 48, outptr1);
+      }
+    }
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,737 @@
+;
+; jdsample.asm - upsampling (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_mmx) PRIVATE
+
+EXTN(jconst_fancy_upsample_mmx):
+
+PW_ONE          times 4 dw  1
+PW_TWO          times 4 dw  2
+PW_THREE        times 4 dw  3
+PW_SEVEN        times 4 dw  7
+PW_EIGHT        times 4 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_mmx (int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        mov     eax, JDIMENSION [downsamp_width(ebp)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                     ; colctr
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr
+
+        test    eax, SIZEOF_MMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    mm0,mm0                 ; mm0=(all 0's)
+        pcmpeqb mm7,mm7
+        psrlq   mm7,(SIZEOF_MMWORD-1)*BYTE_BIT
+        pand    mm7, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        add     eax, byte SIZEOF_MMWORD-1
+        and     eax, byte -SIZEOF_MMWORD
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        pcmpeqb mm6,mm6
+        psllq   mm6,(SIZEOF_MMWORD-1)*BYTE_BIT
+        pand    mm6, MMWORD [esi+0*SIZEOF_MMWORD]
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        movq    mm6, MMWORD [esi+1*SIZEOF_MMWORD]
+        psllq   mm6,(SIZEOF_MMWORD-1)*BYTE_BIT
+
+.upsample:
+        movq    mm1, MMWORD [esi+0*SIZEOF_MMWORD]
+        movq    mm2,mm1
+        movq    mm3,mm1                 ; mm1=( 0 1 2 3 4 5 6 7)
+        psllq   mm2,BYTE_BIT            ; mm2=( - 0 1 2 3 4 5 6)
+        psrlq   mm3,BYTE_BIT            ; mm3=( 1 2 3 4 5 6 7 -)
+
+        por     mm2,mm7                 ; mm2=(-1 0 1 2 3 4 5 6)
+        por     mm3,mm6                 ; mm3=( 1 2 3 4 5 6 7 8)
+
+        movq    mm7,mm1
+        psrlq   mm7,(SIZEOF_MMWORD-1)*BYTE_BIT  ; mm7=( 7 - - - - - - -)
+
+        movq      mm4,mm1
+        punpcklbw mm1,mm0               ; mm1=( 0 1 2 3)
+        punpckhbw mm4,mm0               ; mm4=( 4 5 6 7)
+        movq      mm5,mm2
+        punpcklbw mm2,mm0               ; mm2=(-1 0 1 2)
+        punpckhbw mm5,mm0               ; mm5=( 3 4 5 6)
+        movq      mm6,mm3
+        punpcklbw mm3,mm0               ; mm3=( 1 2 3 4)
+        punpckhbw mm6,mm0               ; mm6=( 5 6 7 8)
+
+        pmullw  mm1,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm2,[GOTOFF(ebx,PW_ONE)]
+        paddw   mm5,[GOTOFF(ebx,PW_ONE)]
+        paddw   mm3,[GOTOFF(ebx,PW_TWO)]
+        paddw   mm6,[GOTOFF(ebx,PW_TWO)]
+
+        paddw   mm2,mm1
+        paddw   mm5,mm4
+        psrlw   mm2,2                   ; mm2=OutLE=( 0  2  4  6)
+        psrlw   mm5,2                   ; mm5=OutHE=( 8 10 12 14)
+        paddw   mm3,mm1
+        paddw   mm6,mm4
+        psrlw   mm3,2                   ; mm3=OutLO=( 1  3  5  7)
+        psrlw   mm6,2                   ; mm6=OutHO=( 9 11 13 15)
+
+        psllw   mm3,BYTE_BIT
+        psllw   mm6,BYTE_BIT
+        por     mm2,mm3                 ; mm2=OutL=( 0  1  2  3  4  5  6  7)
+        por     mm5,mm6                 ; mm5=OutH=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm5
+
+        sub     eax, byte SIZEOF_MMWORD
+        add     esi, byte 1*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 2*SIZEOF_MMWORD       ; outptr
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_mmx (int max_v_samp_factor,
+;                                JDIMENSION downsampled_width,
+;                                JSAMPARRAY input_data,
+;                                JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          4
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void *gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     edx,eax                         ; edx = original ebp
+        mov     eax, JDIMENSION [downsamp_width(edx)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(edx)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(edx)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(edx)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                                     ; colctr
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    eax, SIZEOF_MMWORD-1
+        jz      short .skip
+        push    edx
+        mov     dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     edx
+.skip:
+        ; -- process the first column block
+
+        movq    mm0, MMWORD [ebx+0*SIZEOF_MMWORD]       ; mm0=row[ 0][0]
+        movq    mm1, MMWORD [ecx+0*SIZEOF_MMWORD]       ; mm1=row[-1][0]
+        movq    mm2, MMWORD [esi+0*SIZEOF_MMWORD]       ; mm2=row[+1][0]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      mm3,mm3               ; mm3=(all 0's)
+        movq      mm4,mm0
+        punpcklbw mm0,mm3               ; mm0=row[ 0][0]( 0 1 2 3)
+        punpckhbw mm4,mm3               ; mm4=row[ 0][0]( 4 5 6 7)
+        movq      mm5,mm1
+        punpcklbw mm1,mm3               ; mm1=row[-1][0]( 0 1 2 3)
+        punpckhbw mm5,mm3               ; mm5=row[-1][0]( 4 5 6 7)
+        movq      mm6,mm2
+        punpcklbw mm2,mm3               ; mm2=row[+1][0]( 0 1 2 3)
+        punpckhbw mm6,mm3               ; mm6=row[+1][0]( 4 5 6 7)
+
+        pmullw  mm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+
+        pcmpeqb mm7,mm7
+        psrlq   mm7,(SIZEOF_MMWORD-2)*BYTE_BIT
+
+        paddw   mm1,mm0                 ; mm1=Int0L=( 0 1 2 3)
+        paddw   mm5,mm4                 ; mm5=Int0H=( 4 5 6 7)
+        paddw   mm2,mm0                 ; mm2=Int1L=( 0 1 2 3)
+        paddw   mm6,mm4                 ; mm6=Int1H=( 4 5 6 7)
+
+        movq    MMWORD [edx+0*SIZEOF_MMWORD], mm1       ; temporarily save
+        movq    MMWORD [edx+1*SIZEOF_MMWORD], mm5       ; the intermediate data
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm6
+
+        pand    mm1,mm7                 ; mm1=( 0 - - -)
+        pand    mm2,mm7                 ; mm2=( 0 - - -)
+
+        movq    MMWORD [wk(0)], mm1
+        movq    MMWORD [wk(1)], mm2
+
+        poppic  ebx
+
+        add     eax, byte SIZEOF_MMWORD-1
+        and     eax, byte -SIZEOF_MMWORD
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pcmpeqb mm1,mm1
+        psllq   mm1,(SIZEOF_MMWORD-2)*BYTE_BIT
+        movq    mm2,mm1
+
+        pand    mm1, MMWORD [edx+1*SIZEOF_MMWORD]       ; mm1=( - - - 7)
+        pand    mm2, MMWORD [edi+1*SIZEOF_MMWORD]       ; mm2=( - - - 7)
+
+        movq    MMWORD [wk(2)], mm1
+        movq    MMWORD [wk(3)], mm2
+
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        ; -- process the next column block
+
+        movq    mm0, MMWORD [ebx+1*SIZEOF_MMWORD]       ; mm0=row[ 0][1]
+        movq    mm1, MMWORD [ecx+1*SIZEOF_MMWORD]       ; mm1=row[-1][1]
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]       ; mm2=row[+1][1]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      mm3,mm3               ; mm3=(all 0's)
+        movq      mm4,mm0
+        punpcklbw mm0,mm3               ; mm0=row[ 0][1]( 0 1 2 3)
+        punpckhbw mm4,mm3               ; mm4=row[ 0][1]( 4 5 6 7)
+        movq      mm5,mm1
+        punpcklbw mm1,mm3               ; mm1=row[-1][1]( 0 1 2 3)
+        punpckhbw mm5,mm3               ; mm5=row[-1][1]( 4 5 6 7)
+        movq      mm6,mm2
+        punpcklbw mm2,mm3               ; mm2=row[+1][1]( 0 1 2 3)
+        punpckhbw mm6,mm3               ; mm6=row[+1][1]( 4 5 6 7)
+
+        pmullw  mm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+
+        paddw   mm1,mm0                 ; mm1=Int0L=( 0 1 2 3)
+        paddw   mm5,mm4                 ; mm5=Int0H=( 4 5 6 7)
+        paddw   mm2,mm0                 ; mm2=Int1L=( 0 1 2 3)
+        paddw   mm6,mm4                 ; mm6=Int1H=( 4 5 6 7)
+
+        movq    MMWORD [edx+2*SIZEOF_MMWORD], mm1       ; temporarily save
+        movq    MMWORD [edx+3*SIZEOF_MMWORD], mm5       ; the intermediate data
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm6
+
+        psllq   mm1,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm1=( - - - 0)
+        psllq   mm2,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm2=( - - - 0)
+
+        movq    MMWORD [wk(2)], mm1
+        movq    MMWORD [wk(3)], mm2
+
+.upsample:
+        ; -- process the upper row
+
+        movq    mm7, MMWORD [edx+0*SIZEOF_MMWORD]       ; mm7=Int0L=( 0 1 2 3)
+        movq    mm3, MMWORD [edx+1*SIZEOF_MMWORD]       ; mm3=Int0H=( 4 5 6 7)
+
+        movq    mm0,mm7
+        movq    mm4,mm3
+        psrlq   mm0,2*BYTE_BIT                  ; mm0=( 1 2 3 -)
+        psllq   mm4,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( - - - 4)
+        movq    mm5,mm7
+        movq    mm6,mm3
+        psrlq   mm5,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm5=( 3 - - -)
+        psllq   mm6,2*BYTE_BIT                  ; mm6=( - 4 5 6)
+
+        por     mm0,mm4                         ; mm0=( 1 2 3 4)
+        por     mm5,mm6                         ; mm5=( 3 4 5 6)
+
+        movq    mm1,mm7
+        movq    mm2,mm3
+        psllq   mm1,2*BYTE_BIT                  ; mm1=( - 0 1 2)
+        psrlq   mm2,2*BYTE_BIT                  ; mm2=( 5 6 7 -)
+        movq    mm4,mm3
+        psrlq   mm4,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm4=( 7 - - -)
+
+        por     mm1, MMWORD [wk(0)]             ; mm1=(-1 0 1 2)
+        por     mm2, MMWORD [wk(2)]             ; mm2=( 5 6 7 8)
+
+        movq    MMWORD [wk(0)], mm4
+
+        pmullw  mm7,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm3,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm5,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm0,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   mm2,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   mm1,mm7
+        paddw   mm5,mm3
+        psrlw   mm1,4                   ; mm1=Out0LE=( 0  2  4  6)
+        psrlw   mm5,4                   ; mm5=Out0HE=( 8 10 12 14)
+        paddw   mm0,mm7
+        paddw   mm2,mm3
+        psrlw   mm0,4                   ; mm0=Out0LO=( 1  3  5  7)
+        psrlw   mm2,4                   ; mm2=Out0HO=( 9 11 13 15)
+
+        psllw   mm0,BYTE_BIT
+        psllw   mm2,BYTE_BIT
+        por     mm1,mm0                 ; mm1=Out0L=( 0  1  2  3  4  5  6  7)
+        por     mm5,mm2                 ; mm5=Out0H=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edx+0*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edx+1*SIZEOF_MMWORD], mm5
+
+        ; -- process the lower row
+
+        movq    mm6, MMWORD [edi+0*SIZEOF_MMWORD]       ; mm6=Int1L=( 0 1 2 3)
+        movq    mm4, MMWORD [edi+1*SIZEOF_MMWORD]       ; mm4=Int1H=( 4 5 6 7)
+
+        movq    mm7,mm6
+        movq    mm3,mm4
+        psrlq   mm7,2*BYTE_BIT                  ; mm7=( 1 2 3 -)
+        psllq   mm3,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( - - - 4)
+        movq    mm0,mm6
+        movq    mm2,mm4
+        psrlq   mm0,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm0=( 3 - - -)
+        psllq   mm2,2*BYTE_BIT                  ; mm2=( - 4 5 6)
+
+        por     mm7,mm3                         ; mm7=( 1 2 3 4)
+        por     mm0,mm2                         ; mm0=( 3 4 5 6)
+
+        movq    mm1,mm6
+        movq    mm5,mm4
+        psllq   mm1,2*BYTE_BIT                  ; mm1=( - 0 1 2)
+        psrlq   mm5,2*BYTE_BIT                  ; mm5=( 5 6 7 -)
+        movq    mm3,mm4
+        psrlq   mm3,(SIZEOF_MMWORD-2)*BYTE_BIT  ; mm3=( 7 - - -)
+
+        por     mm1, MMWORD [wk(1)]             ; mm1=(-1 0 1 2)
+        por     mm5, MMWORD [wk(3)]             ; mm5=( 5 6 7 8)
+
+        movq    MMWORD [wk(1)], mm3
+
+        pmullw  mm6,[GOTOFF(ebx,PW_THREE)]
+        pmullw  mm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   mm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm0,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   mm7,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   mm5,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   mm1,mm6
+        paddw   mm0,mm4
+        psrlw   mm1,4                   ; mm1=Out1LE=( 0  2  4  6)
+        psrlw   mm0,4                   ; mm0=Out1HE=( 8 10 12 14)
+        paddw   mm7,mm6
+        paddw   mm5,mm4
+        psrlw   mm7,4                   ; mm7=Out1LO=( 1  3  5  7)
+        psrlw   mm5,4                   ; mm5=Out1HO=( 9 11 13 15)
+
+        psllw   mm7,BYTE_BIT
+        psllw   mm5,BYTE_BIT
+        por     mm1,mm7                 ; mm1=Out1L=( 0  1  2  3  4  5  6  7)
+        por     mm0,mm5                 ; mm0=Out1H=( 8  9 10 11 12 13 14 15)
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm0
+
+        poppic  ebx
+
+        sub     eax, byte SIZEOF_MMWORD
+        add     ecx, byte 1*SIZEOF_MMWORD       ; inptr1(above)
+        add     ebx, byte 1*SIZEOF_MMWORD       ; inptr0
+        add     esi, byte 1*SIZEOF_MMWORD       ; inptr1(below)
+        add     edx, byte 2*SIZEOF_MMWORD       ; outptr0
+        add     edi, byte 2*SIZEOF_MMWORD       ; outptr1
+        cmp     eax, byte SIZEOF_MMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     ecx
+        pop     eax
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_mmx (int max_v_samp_factor,
+;                          JDIMENSION output_width,
+;                          JSAMPARRAY input_data,
+;                          JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_MMWORD)-1
+        and     edx, byte -(2*SIZEOF_MMWORD)
+        jz      short .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        mov     eax,edx                         ; colctr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        movq      mm1,mm0
+        punpcklbw mm0,mm0
+        punpckhbw mm1,mm1
+
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm1
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq      mm3,mm2
+        punpcklbw mm2,mm2
+        punpckhbw mm3,mm3
+
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm3
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     edi, byte 4*SIZEOF_MMWORD       ; outptr
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_mmx (int max_v_samp_factor,
+;                          JDIMENSION output_width,
+;                          JSAMPARRAY input_data,
+;                          JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_mmx) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_MMWORD)-1
+        and     edx, byte -(2*SIZEOF_MMWORD)
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]                     ; inptr
+        mov     ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     eax,edx                                 ; colctr
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [esi+0*SIZEOF_MMWORD]
+
+        movq      mm1,mm0
+        punpcklbw mm0,mm0
+        punpckhbw mm1,mm1
+
+        movq    MMWORD [ebx+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [ebx+1*SIZEOF_MMWORD], mm1
+        movq    MMWORD [edi+0*SIZEOF_MMWORD], mm0
+        movq    MMWORD [edi+1*SIZEOF_MMWORD], mm1
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        movq    mm2, MMWORD [esi+1*SIZEOF_MMWORD]
+
+        movq      mm3,mm2
+        punpcklbw mm2,mm2
+        punpckhbw mm3,mm3
+
+        movq    MMWORD [ebx+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [ebx+3*SIZEOF_MMWORD], mm3
+        movq    MMWORD [edi+2*SIZEOF_MMWORD], mm2
+        movq    MMWORD [edi+3*SIZEOF_MMWORD], mm3
+
+        sub     eax, byte 2*SIZEOF_MMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_MMWORD       ; inptr
+        add     ebx, byte 4*SIZEOF_MMWORD       ; outptr0
+        add     edi, byte 4*SIZEOF_MMWORD       ; outptr1
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      short .rowloop
+
+        emms            ; empty MMX state
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,671 @@
+;
+; jdsample.asm - upsampling (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_sse2) PRIVATE
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE          times 8 dw  1
+PW_TWO          times 8 dw  2
+PW_THREE        times 8 dw  3
+PW_SEVEN        times 8 dw  7
+PW_EIGHT        times 8 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     eax, r11d  ; colctr
+        test    rax,rax
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rax                     ; colctr
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]     ; inptr
+        mov     rdi, JSAMPROW [rdi]     ; outptr
+
+        test    rax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    xmm0,xmm0               ; xmm0=(all 0's)
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-1)
+        pand    xmm7, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        add     rax, byte SIZEOF_XMMWORD-1
+        and     rax, byte -SIZEOF_XMMWORD
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+
+.columnloop_last:
+        pcmpeqb xmm6,xmm6
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+        pand    xmm6, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        jmp     short .upsample
+
+.columnloop:
+        movdqa  xmm6, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+
+.upsample:
+        movdqa  xmm1, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2,xmm1
+        movdqa  xmm3,xmm1               ; xmm1=( 0  1  2 ... 13 14 15)
+        pslldq  xmm2,1                  ; xmm2=(--  0  1 ... 12 13 14)
+        psrldq  xmm3,1                  ; xmm3=( 1  2  3 ... 14 15 --)
+
+        por     xmm2,xmm7               ; xmm2=(-1  0  1 ... 12 13 14)
+        por     xmm3,xmm6               ; xmm3=( 1  2  3 ... 14 15 16)
+
+        movdqa  xmm7,xmm1
+        psrldq  xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --)
+
+        movdqa    xmm4,xmm1
+        punpcklbw xmm1,xmm0             ; xmm1=( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm0             ; xmm4=( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm2
+        punpcklbw xmm2,xmm0             ; xmm2=(-1  0  1  2  3  4  5  6)
+        punpckhbw xmm5,xmm0             ; xmm5=( 7  8  9 10 11 12 13 14)
+        movdqa    xmm6,xmm3
+        punpcklbw xmm3,xmm0             ; xmm3=( 1  2  3  4  5  6  7  8)
+        punpckhbw xmm6,xmm0             ; xmm6=( 9 10 11 12 13 14 15 16)
+
+        pmullw  xmm1,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+        paddw   xmm2,[rel PW_ONE]
+        paddw   xmm5,[rel PW_ONE]
+        paddw   xmm3,[rel PW_TWO]
+        paddw   xmm6,[rel PW_TWO]
+
+        paddw   xmm2,xmm1
+        paddw   xmm5,xmm4
+        psrlw   xmm2,2                  ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,2                  ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+        paddw   xmm3,xmm1
+        paddw   xmm6,xmm4
+        psrlw   xmm3,2                  ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm6,2                  ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm3,BYTE_BIT
+        psllw   xmm6,BYTE_BIT
+        por     xmm2,xmm3               ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm6               ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm5
+
+        sub     rax, byte SIZEOF_XMMWORD
+        add     rsi, byte 1*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     rsi
+        pop     rdi
+        pop     rax
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rcx                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION downsampled_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          4
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+        push    rbx
+
+        mov     eax, r11d  ; colctr
+        test    rax,rax
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rax                                     ; colctr
+        push    rcx
+        push    rdi
+        push    rsi
+
+        mov     rcx, JSAMPROW [rsi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     rsi, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    rax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        push    rdx
+        mov     dl, JSAMPLE [rcx+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rcx+rax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [rbx+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rbx+rax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [rsi+(rax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [rsi+rax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     rdx
+.skip:
+        ; -- process the first column block
+
+        movdqa  xmm0, XMMWORD [rbx+0*SIZEOF_XMMWORD]    ; xmm0=row[ 0][0]
+        movdqa  xmm1, XMMWORD [rcx+0*SIZEOF_XMMWORD]    ; xmm1=row[-1][0]
+        movdqa  xmm2, XMMWORD [rsi+0*SIZEOF_XMMWORD]    ; xmm2=row[+1][0]
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-2)
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm6
+
+        pand    xmm1,xmm7               ; xmm1=( 0 -- -- -- -- -- -- --)
+        pand    xmm2,xmm7               ; xmm2=( 0 -- -- -- -- -- -- --)
+
+        movdqa  XMMWORD [wk(0)], xmm1
+        movdqa  XMMWORD [wk(1)], xmm2
+
+        add     rax, byte SIZEOF_XMMWORD-1
+        and     rax, byte -SIZEOF_XMMWORD
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pcmpeqb xmm1,xmm1
+        pslldq  xmm1,(SIZEOF_XMMWORD-2)
+        movdqa  xmm2,xmm1
+
+        pand    xmm1, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+        pand    xmm2, XMMWORD [rdi+1*SIZEOF_XMMWORD]
+
+        movdqa  XMMWORD [wk(2)], xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+        movdqa  XMMWORD [wk(3)], xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+        jmp     near .upsample
+
+.columnloop:
+        ; -- process the next column block
+
+        movdqa  xmm0, XMMWORD [rbx+1*SIZEOF_XMMWORD]    ; xmm0=row[ 0][1]
+        movdqa  xmm1, XMMWORD [rcx+1*SIZEOF_XMMWORD]    ; xmm1=row[-1][1]
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]    ; xmm2=row[+1][1]
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [rdx+2*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [rdx+3*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm6
+
+        pslldq  xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- --  0)
+        pslldq  xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- --  0)
+
+        movdqa  XMMWORD [wk(2)], xmm1
+        movdqa  XMMWORD [wk(3)], xmm2
+
+.upsample:
+        ; -- process the upper row
+
+        movdqa  xmm7, XMMWORD [rdx+0*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [rdx+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm0,xmm7               ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm4,xmm3               ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm0,2                  ; xmm0=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- --  8)
+        movdqa  xmm5,xmm7
+        movdqa  xmm6,xmm3
+        psrldq  xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm6,2                  ; xmm6=(--  8  9 10 11 12 13 14)
+
+        por     xmm0,xmm4               ; xmm0=( 1  2  3  4  5  6  7  8)
+        por     xmm5,xmm6               ; xmm5=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm2,xmm3
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm2,2                  ; xmm2=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm4,xmm3
+        psrldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(0)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm2, XMMWORD [wk(2)]   ; xmm2=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(0)], xmm4
+
+        pmullw  xmm7,[rel PW_THREE]
+        pmullw  xmm3,[rel PW_THREE]
+        paddw   xmm1,[rel PW_EIGHT]
+        paddw   xmm5,[rel PW_EIGHT]
+        paddw   xmm0,[rel PW_SEVEN]
+        paddw   xmm2,[rel PW_SEVEN]
+
+        paddw   xmm1,xmm7
+        paddw   xmm5,xmm3
+        psrlw   xmm1,4                  ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,4                  ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm3
+        psrlw   xmm0,4                  ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm2,4                  ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm0,BYTE_BIT
+        psllw   xmm2,BYTE_BIT
+        por     xmm1,xmm0               ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm2               ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdx+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdx+1*SIZEOF_XMMWORD], xmm5
+
+        ; -- process the lower row
+
+        movdqa  xmm6, XMMWORD [rdi+0*SIZEOF_XMMWORD]
+        movdqa  xmm4, XMMWORD [rdi+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm7,xmm6               ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm3,xmm4               ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm7,2                  ; xmm7=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- --  8)
+        movdqa  xmm0,xmm6
+        movdqa  xmm2,xmm4
+        psrldq  xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm2,2                  ; xmm2=(--  8  9 10 11 12 13 14)
+
+        por     xmm7,xmm3               ; xmm7=( 1  2  3  4  5  6  7  8)
+        por     xmm0,xmm2               ; xmm0=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm4
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm5,2                  ; xmm5=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm3,xmm4
+        psrldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(1)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm5, XMMWORD [wk(3)]   ; xmm5=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(1)], xmm3
+
+        pmullw  xmm6,[rel PW_THREE]
+        pmullw  xmm4,[rel PW_THREE]
+        paddw   xmm1,[rel PW_EIGHT]
+        paddw   xmm0,[rel PW_EIGHT]
+        paddw   xmm7,[rel PW_SEVEN]
+        paddw   xmm5,[rel PW_SEVEN]
+
+        paddw   xmm1,xmm6
+        paddw   xmm0,xmm4
+        psrlw   xmm1,4                  ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm0,4                  ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm7,xmm6
+        paddw   xmm5,xmm4
+        psrlw   xmm7,4                  ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm5,4                  ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm7,BYTE_BIT
+        psllw   xmm5,BYTE_BIT
+        por     xmm1,xmm7               ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+        por     xmm0,xmm5               ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm0
+
+        sub     rax, byte SIZEOF_XMMWORD
+        add     rcx, byte 1*SIZEOF_XMMWORD      ; inptr1(above)
+        add     rbx, byte 1*SIZEOF_XMMWORD      ; inptr0
+        add     rsi, byte 1*SIZEOF_XMMWORD      ; inptr1(below)
+        add     rdx, byte 2*SIZEOF_XMMWORD      ; outptr0
+        add     rdi, byte 2*SIZEOF_XMMWORD      ; outptr1
+        cmp     rax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    rax,rax
+        jnz     near .columnloop_last
+
+        pop     rsi
+        pop     rdi
+        pop     rcx
+        pop     rax
+
+        add     rsi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     rcx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov     edx, r11d
+        add     rdx, byte (2*SIZEOF_XMMWORD)-1
+        and     rdx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      short .return
+
+        mov     rsi, r12 ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]             ; inptr
+        mov     rdi, JSAMPROW [rdi]             ; outptr
+        mov     rax,rdx                         ; colctr
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rdi, byte 4*SIZEOF_XMMWORD      ; outptr
+        jmp     short .columnloop
+
+.nextrow:
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte SIZEOF_JSAMPROW       ; input_data
+        add     rdi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     rcx                             ; rowctr
+        jg      short .rowloop
+
+.return:
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+; r10 = int max_v_samp_factor
+; r11 = JDIMENSION output_width
+; r12 = JSAMPARRAY input_data
+; r13 = JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        mov     edx, r11d
+        add     rdx, byte (2*SIZEOF_XMMWORD)-1
+        and     rdx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     rcx, r10        ; rowctr
+        test    rcx,rcx
+        jz      near .return
+
+        mov     rsi, r12        ; input_data
+        mov     rdi, r13
+        mov     rdi, JSAMPARRAY [rdi]                   ; output_data
+.rowloop:
+        push    rdi
+        push    rsi
+
+        mov     rsi, JSAMPROW [rsi]                     ; inptr
+        mov     rbx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     rdi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     rax,rdx                                 ; colctr
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [rsi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [rbx+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rbx+1*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [rdi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [rdi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [rsi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [rbx+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rbx+3*SIZEOF_XMMWORD], xmm3
+        movdqa  XMMWORD [rdi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [rdi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     rax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     rsi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     rbx, byte 4*SIZEOF_XMMWORD      ; outptr0
+        add     rdi, byte 4*SIZEOF_XMMWORD      ; outptr1
+        jmp     short .columnloop
+
+.nextrow:
+        pop     rsi
+        pop     rdi
+
+        add     rsi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     rdi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     rcx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jdsample-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jdsample-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,729 @@
+;
+; jdsample.asm - upsampling (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fancy_upsample_sse2) PRIVATE
+
+EXTN(jconst_fancy_upsample_sse2):
+
+PW_ONE          times 8 dw  1
+PW_TWO          times 8 dw  2
+PW_THREE        times 8 dw  3
+PW_SEVEN        times 8 dw  7
+PW_EIGHT        times 8 dw  8
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Fancy processing for the common case of 2:1 horizontal and 1:1 vertical.
+;
+; The upsampling algorithm is linear interpolation between pixel centers,
+; also known as a "triangle filter".  This is a good compromise between
+; speed and visual quality.  The centers of the output pixels are 1/4 and 3/4
+; of the way between input pixel centers.
+;
+; GLOBAL(void)
+; jsimd_h2v1_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_fancy_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        mov     eax, JDIMENSION [downsamp_width(ebp)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                     ; colctr
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]     ; inptr
+        mov     edi, JSAMPROW [edi]     ; outptr
+
+        test    eax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+.skip:
+        pxor    xmm0,xmm0               ; xmm0=(all 0's)
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-1)
+        pand    xmm7, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        add     eax, byte SIZEOF_XMMWORD-1
+        and     eax, byte -SIZEOF_XMMWORD
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        pcmpeqb xmm6,xmm6
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+        pand    xmm6, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        jmp     short .upsample
+        alignx  16,7
+
+.columnloop:
+        movdqa  xmm6, XMMWORD [esi+1*SIZEOF_XMMWORD]
+        pslldq  xmm6,(SIZEOF_XMMWORD-1)
+
+.upsample:
+        movdqa  xmm1, XMMWORD [esi+0*SIZEOF_XMMWORD]
+        movdqa  xmm2,xmm1
+        movdqa  xmm3,xmm1               ; xmm1=( 0  1  2 ... 13 14 15)
+        pslldq  xmm2,1                  ; xmm2=(--  0  1 ... 12 13 14)
+        psrldq  xmm3,1                  ; xmm3=( 1  2  3 ... 14 15 --)
+
+        por     xmm2,xmm7               ; xmm2=(-1  0  1 ... 12 13 14)
+        por     xmm3,xmm6               ; xmm3=( 1  2  3 ... 14 15 16)
+
+        movdqa  xmm7,xmm1
+        psrldq  xmm7,(SIZEOF_XMMWORD-1) ; xmm7=(15 -- -- ... -- -- --)
+
+        movdqa    xmm4,xmm1
+        punpcklbw xmm1,xmm0             ; xmm1=( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm0             ; xmm4=( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm2
+        punpcklbw xmm2,xmm0             ; xmm2=(-1  0  1  2  3  4  5  6)
+        punpckhbw xmm5,xmm0             ; xmm5=( 7  8  9 10 11 12 13 14)
+        movdqa    xmm6,xmm3
+        punpcklbw xmm3,xmm0             ; xmm3=( 1  2  3  4  5  6  7  8)
+        punpckhbw xmm6,xmm0             ; xmm6=( 9 10 11 12 13 14 15 16)
+
+        pmullw  xmm1,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm2,[GOTOFF(ebx,PW_ONE)]
+        paddw   xmm5,[GOTOFF(ebx,PW_ONE)]
+        paddw   xmm3,[GOTOFF(ebx,PW_TWO)]
+        paddw   xmm6,[GOTOFF(ebx,PW_TWO)]
+
+        paddw   xmm2,xmm1
+        paddw   xmm5,xmm4
+        psrlw   xmm2,2                  ; xmm2=OutLE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,2                  ; xmm5=OutHE=(16 18 20 22 24 26 28 30)
+        paddw   xmm3,xmm1
+        paddw   xmm6,xmm4
+        psrlw   xmm3,2                  ; xmm3=OutLO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm6,2                  ; xmm6=OutHO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm3,BYTE_BIT
+        psllw   xmm6,BYTE_BIT
+        por     xmm2,xmm3               ; xmm2=OutL=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm6               ; xmm5=OutH=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm5
+
+        sub     eax, byte SIZEOF_XMMWORD
+        add     esi, byte 1*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     eax
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+; Again a triangle filter; see comments for h2v1 case, above.
+;
+; GLOBAL(void)
+; jsimd_h2v2_fancy_upsample_sse2 (int max_v_samp_factor,
+;                                 JDIMENSION downsampled_width,
+;                                 JSAMPARRAY input_data,
+;                                 JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define downsamp_width(b)       (b)+12          ; JDIMENSION downsampled_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          4
+%define gotptr          wk(0)-SIZEOF_POINTER    ; void *gotptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_fancy_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_fancy_upsample_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic eax             ; make a room for GOT address
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx                     ; get GOT address
+        movpic  POINTER [gotptr], ebx   ; save GOT address
+
+        mov     edx,eax                         ; edx = original ebp
+        mov     eax, JDIMENSION [downsamp_width(edx)]  ; colctr
+        test    eax,eax
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(edx)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(edx)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(edx)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    eax                                     ; colctr
+        push    ecx
+        push    edi
+        push    esi
+
+        mov     ecx, JSAMPROW [esi-1*SIZEOF_JSAMPROW]   ; inptr1(above)
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; inptr0
+        mov     esi, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; inptr1(below)
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+
+        test    eax, SIZEOF_XMMWORD-1
+        jz      short .skip
+        push    edx
+        mov     dl, JSAMPLE [ecx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ecx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [ebx+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [ebx+eax*SIZEOF_JSAMPLE], dl
+        mov     dl, JSAMPLE [esi+(eax-1)*SIZEOF_JSAMPLE]
+        mov     JSAMPLE [esi+eax*SIZEOF_JSAMPLE], dl    ; insert a dummy sample
+        pop     edx
+.skip:
+        ; -- process the first column block
+
+        movdqa  xmm0, XMMWORD [ebx+0*SIZEOF_XMMWORD]    ; xmm0=row[ 0][0]
+        movdqa  xmm1, XMMWORD [ecx+0*SIZEOF_XMMWORD]    ; xmm1=row[-1][0]
+        movdqa  xmm2, XMMWORD [esi+0*SIZEOF_XMMWORD]    ; xmm2=row[+1][0]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+
+        pcmpeqb xmm7,xmm7
+        psrldq  xmm7,(SIZEOF_XMMWORD-2)
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm6
+
+        pand    xmm1,xmm7               ; xmm1=( 0 -- -- -- -- -- -- --)
+        pand    xmm2,xmm7               ; xmm2=( 0 -- -- -- -- -- -- --)
+
+        movdqa  XMMWORD [wk(0)], xmm1
+        movdqa  XMMWORD [wk(1)], xmm2
+
+        poppic  ebx
+
+        add     eax, byte SIZEOF_XMMWORD-1
+        and     eax, byte -SIZEOF_XMMWORD
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      short .columnloop
+        alignx  16,7
+
+.columnloop_last:
+        ; -- process the last column block
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pcmpeqb xmm1,xmm1
+        pslldq  xmm1,(SIZEOF_XMMWORD-2)
+        movdqa  xmm2,xmm1
+
+        pand    xmm1, XMMWORD [edx+1*SIZEOF_XMMWORD]
+        pand    xmm2, XMMWORD [edi+1*SIZEOF_XMMWORD]
+
+        movdqa  XMMWORD [wk(2)], xmm1   ; xmm1=(-- -- -- -- -- -- -- 15)
+        movdqa  XMMWORD [wk(3)], xmm2   ; xmm2=(-- -- -- -- -- -- -- 15)
+
+        jmp     near .upsample
+        alignx  16,7
+
+.columnloop:
+        ; -- process the next column block
+
+        movdqa  xmm0, XMMWORD [ebx+1*SIZEOF_XMMWORD]    ; xmm0=row[ 0][1]
+        movdqa  xmm1, XMMWORD [ecx+1*SIZEOF_XMMWORD]    ; xmm1=row[-1][1]
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]    ; xmm2=row[+1][1]
+
+        pushpic ebx
+        movpic  ebx, POINTER [gotptr]   ; load GOT address
+
+        pxor      xmm3,xmm3             ; xmm3=(all 0's)
+        movdqa    xmm4,xmm0
+        punpcklbw xmm0,xmm3             ; xmm0=row[ 0]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm4,xmm3             ; xmm4=row[ 0]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm5,xmm1
+        punpcklbw xmm1,xmm3             ; xmm1=row[-1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm5,xmm3             ; xmm5=row[-1]( 8  9 10 11 12 13 14 15)
+        movdqa    xmm6,xmm2
+        punpcklbw xmm2,xmm3             ; xmm2=row[+1]( 0  1  2  3  4  5  6  7)
+        punpckhbw xmm6,xmm3             ; xmm6=row[+1]( 8  9 10 11 12 13 14 15)
+
+        pmullw  xmm0,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+
+        paddw   xmm1,xmm0               ; xmm1=Int0L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm5,xmm4               ; xmm5=Int0H=( 8  9 10 11 12 13 14 15)
+        paddw   xmm2,xmm0               ; xmm2=Int1L=( 0  1  2  3  4  5  6  7)
+        paddw   xmm6,xmm4               ; xmm6=Int1H=( 8  9 10 11 12 13 14 15)
+
+        movdqa  XMMWORD [edx+2*SIZEOF_XMMWORD], xmm1    ; temporarily save
+        movdqa  XMMWORD [edx+3*SIZEOF_XMMWORD], xmm5    ; the intermediate data
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm6
+
+        pslldq  xmm1,(SIZEOF_XMMWORD-2) ; xmm1=(-- -- -- -- -- -- --  0)
+        pslldq  xmm2,(SIZEOF_XMMWORD-2) ; xmm2=(-- -- -- -- -- -- --  0)
+
+        movdqa  XMMWORD [wk(2)], xmm1
+        movdqa  XMMWORD [wk(3)], xmm2
+
+.upsample:
+        ; -- process the upper row
+
+        movdqa  xmm7, XMMWORD [edx+0*SIZEOF_XMMWORD]
+        movdqa  xmm3, XMMWORD [edx+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm0,xmm7               ; xmm7=Int0L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm4,xmm3               ; xmm3=Int0H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm0,2                  ; xmm0=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(-- -- -- -- -- -- --  8)
+        movdqa  xmm5,xmm7
+        movdqa  xmm6,xmm3
+        psrldq  xmm5,(SIZEOF_XMMWORD-2) ; xmm5=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm6,2                  ; xmm6=(--  8  9 10 11 12 13 14)
+
+        por     xmm0,xmm4               ; xmm0=( 1  2  3  4  5  6  7  8)
+        por     xmm5,xmm6               ; xmm5=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm2,xmm3
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm2,2                  ; xmm2=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm4,xmm3
+        psrldq  xmm4,(SIZEOF_XMMWORD-2) ; xmm4=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(0)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm2, XMMWORD [wk(2)]   ; xmm2=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(0)], xmm4
+
+        pmullw  xmm7,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm3,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm5,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm0,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   xmm2,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   xmm1,xmm7
+        paddw   xmm5,xmm3
+        psrlw   xmm1,4                  ; xmm1=Out0LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm5,4                  ; xmm5=Out0HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm0,xmm7
+        paddw   xmm2,xmm3
+        psrlw   xmm0,4                  ; xmm0=Out0LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm2,4                  ; xmm2=Out0HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm0,BYTE_BIT
+        psllw   xmm2,BYTE_BIT
+        por     xmm1,xmm0               ; xmm1=Out0L=( 0  1  2 ... 13 14 15)
+        por     xmm5,xmm2               ; xmm5=Out0H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edx+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edx+1*SIZEOF_XMMWORD], xmm5
+
+        ; -- process the lower row
+
+        movdqa  xmm6, XMMWORD [edi+0*SIZEOF_XMMWORD]
+        movdqa  xmm4, XMMWORD [edi+1*SIZEOF_XMMWORD]
+
+        movdqa  xmm7,xmm6               ; xmm6=Int1L=( 0  1  2  3  4  5  6  7)
+        movdqa  xmm3,xmm4               ; xmm4=Int1H=( 8  9 10 11 12 13 14 15)
+        psrldq  xmm7,2                  ; xmm7=( 1  2  3  4  5  6  7 --)
+        pslldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(-- -- -- -- -- -- --  8)
+        movdqa  xmm0,xmm6
+        movdqa  xmm2,xmm4
+        psrldq  xmm0,(SIZEOF_XMMWORD-2) ; xmm0=( 7 -- -- -- -- -- -- --)
+        pslldq  xmm2,2                  ; xmm2=(--  8  9 10 11 12 13 14)
+
+        por     xmm7,xmm3               ; xmm7=( 1  2  3  4  5  6  7  8)
+        por     xmm0,xmm2               ; xmm0=( 7  8  9 10 11 12 13 14)
+
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm4
+        pslldq  xmm1,2                  ; xmm1=(--  0  1  2  3  4  5  6)
+        psrldq  xmm5,2                  ; xmm5=( 9 10 11 12 13 14 15 --)
+        movdqa  xmm3,xmm4
+        psrldq  xmm3,(SIZEOF_XMMWORD-2) ; xmm3=(15 -- -- -- -- -- -- --)
+
+        por     xmm1, XMMWORD [wk(1)]   ; xmm1=(-1  0  1  2  3  4  5  6)
+        por     xmm5, XMMWORD [wk(3)]   ; xmm5=( 9 10 11 12 13 14 15 16)
+
+        movdqa  XMMWORD [wk(1)], xmm3
+
+        pmullw  xmm6,[GOTOFF(ebx,PW_THREE)]
+        pmullw  xmm4,[GOTOFF(ebx,PW_THREE)]
+        paddw   xmm1,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm0,[GOTOFF(ebx,PW_EIGHT)]
+        paddw   xmm7,[GOTOFF(ebx,PW_SEVEN)]
+        paddw   xmm5,[GOTOFF(ebx,PW_SEVEN)]
+
+        paddw   xmm1,xmm6
+        paddw   xmm0,xmm4
+        psrlw   xmm1,4                  ; xmm1=Out1LE=( 0  2  4  6  8 10 12 14)
+        psrlw   xmm0,4                  ; xmm0=Out1HE=(16 18 20 22 24 26 28 30)
+        paddw   xmm7,xmm6
+        paddw   xmm5,xmm4
+        psrlw   xmm7,4                  ; xmm7=Out1LO=( 1  3  5  7  9 11 13 15)
+        psrlw   xmm5,4                  ; xmm5=Out1HO=(17 19 21 23 25 27 29 31)
+
+        psllw   xmm7,BYTE_BIT
+        psllw   xmm5,BYTE_BIT
+        por     xmm1,xmm7               ; xmm1=Out1L=( 0  1  2 ... 13 14 15)
+        por     xmm0,xmm5               ; xmm0=Out1H=(16 17 18 ... 29 30 31)
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm0
+
+        poppic  ebx
+
+        sub     eax, byte SIZEOF_XMMWORD
+        add     ecx, byte 1*SIZEOF_XMMWORD      ; inptr1(above)
+        add     ebx, byte 1*SIZEOF_XMMWORD      ; inptr0
+        add     esi, byte 1*SIZEOF_XMMWORD      ; inptr1(below)
+        add     edx, byte 2*SIZEOF_XMMWORD      ; outptr0
+        add     edi, byte 2*SIZEOF_XMMWORD      ; outptr1
+        cmp     eax, byte SIZEOF_XMMWORD
+        ja      near .columnloop
+        test    eax,eax
+        jnz     near .columnloop_last
+
+        pop     esi
+        pop     edi
+        pop     ecx
+        pop     eax
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      near .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 1:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v1_upsample_sse2 (int max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v1_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v1_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_XMMWORD)-1
+        and     edx, byte -(2*SIZEOF_XMMWORD)
+        jz      short .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      short .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]             ; inptr
+        mov     edi, JSAMPROW [edi]             ; outptr
+        mov     eax,edx                         ; colctr
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     edi, byte 4*SIZEOF_XMMWORD      ; outptr
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte SIZEOF_JSAMPROW       ; input_data
+        add     edi, byte SIZEOF_JSAMPROW       ; output_data
+        dec     ecx                             ; rowctr
+        jg      short .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Fast processing for the common case of 2:1 horizontal and 2:1 vertical.
+; It's still a box filter.
+;
+; GLOBAL(void)
+; jsimd_h2v2_upsample_sse2 (nt max_v_samp_factor,
+;                           JDIMENSION output_width,
+;                           JSAMPARRAY input_data,
+;                           JSAMPARRAY *output_data_ptr);
+;
+
+%define max_v_samp(b)           (b)+8           ; int max_v_samp_factor
+%define output_width(b)         (b)+12          ; JDIMENSION output_width
+%define input_data(b)           (b)+16          ; JSAMPARRAY input_data
+%define output_data_ptr(b)      (b)+20          ; JSAMPARRAY *output_data_ptr
+
+        align   16
+        global  EXTN(jsimd_h2v2_upsample_sse2) PRIVATE
+
+EXTN(jsimd_h2v2_upsample_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     edx, JDIMENSION [output_width(ebp)]
+        add     edx, byte (2*SIZEOF_XMMWORD)-1
+        and     edx, byte -(2*SIZEOF_XMMWORD)
+        jz      near .return
+
+        mov     ecx, INT [max_v_samp(ebp)]      ; rowctr
+        test    ecx,ecx
+        jz      near .return
+
+        mov     esi, JSAMPARRAY [input_data(ebp)]       ; input_data
+        mov     edi, POINTER [output_data_ptr(ebp)]
+        mov     edi, JSAMPARRAY [edi]                   ; output_data
+        alignx  16,7
+.rowloop:
+        push    edi
+        push    esi
+
+        mov     esi, JSAMPROW [esi]                     ; inptr
+        mov     ebx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]   ; outptr0
+        mov     edi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]   ; outptr1
+        mov     eax,edx                                 ; colctr
+        alignx  16,7
+.columnloop:
+
+        movdqa  xmm0, XMMWORD [esi+0*SIZEOF_XMMWORD]
+
+        movdqa    xmm1,xmm0
+        punpcklbw xmm0,xmm0
+        punpckhbw xmm1,xmm1
+
+        movdqa  XMMWORD [ebx+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [ebx+1*SIZEOF_XMMWORD], xmm1
+        movdqa  XMMWORD [edi+0*SIZEOF_XMMWORD], xmm0
+        movdqa  XMMWORD [edi+1*SIZEOF_XMMWORD], xmm1
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        movdqa  xmm2, XMMWORD [esi+1*SIZEOF_XMMWORD]
+
+        movdqa    xmm3,xmm2
+        punpcklbw xmm2,xmm2
+        punpckhbw xmm3,xmm3
+
+        movdqa  XMMWORD [ebx+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [ebx+3*SIZEOF_XMMWORD], xmm3
+        movdqa  XMMWORD [edi+2*SIZEOF_XMMWORD], xmm2
+        movdqa  XMMWORD [edi+3*SIZEOF_XMMWORD], xmm3
+
+        sub     eax, byte 2*SIZEOF_XMMWORD
+        jz      short .nextrow
+
+        add     esi, byte 2*SIZEOF_XMMWORD      ; inptr
+        add     ebx, byte 4*SIZEOF_XMMWORD      ; outptr0
+        add     edi, byte 4*SIZEOF_XMMWORD      ; outptr1
+        jmp     short .columnloop
+        alignx  16,7
+
+.nextrow:
+        pop     esi
+        pop     edi
+
+        add     esi, byte 1*SIZEOF_JSAMPROW     ; input_data
+        add     edi, byte 2*SIZEOF_JSAMPROW     ; output_data
+        sub     ecx, byte 2                     ; rowctr
+        jg      short .rowloop
+
+.return:
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-3dn.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,320 @@
+;
+; jfdctflt.asm - floating-point FDCT (3DNow!)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_3dnow) PRIVATE
+
+EXTN(jconst_fdct_float_3dnow):
+
+PD_0_382        times 2 dd  0.382683432365089771728460
+PD_0_707        times 2 dd  0.707106781186547524400844
+PD_0_541        times 2 dd  0.541196100146196984399723
+PD_1_306        times 2 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_3dnow (FAST_FLOAT *data)
+;
+
+%define data(b)         (b)+8           ; FAST_FLOAT *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_3dnow) PRIVATE
+
+EXTN(jsimd_fdct_float_3dnow):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm0=(00 01), mm1=(10 11), mm2=(06 07), mm3=(16 17)
+
+        movq      mm4,mm0               ; transpose coefficients
+        punpckldq mm0,mm1               ; mm0=(00 10)=data0
+        punpckhdq mm4,mm1               ; mm4=(01 11)=data1
+        movq      mm5,mm2               ; transpose coefficients
+        punpckldq mm2,mm3               ; mm2=(06 16)=data6
+        punpckhdq mm5,mm3               ; mm5=(07 17)=data7
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm2                 ; mm4=data1-data6=tmp6
+        pfsub   mm0,mm5                 ; mm0=data0-data7=tmp7
+        pfadd   mm6,mm2                 ; mm6=data1+data6=tmp1
+        pfadd   mm7,mm5                 ; mm7=data0+data7=tmp0
+
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm1=(02 03), mm3=(12 13), mm2=(04 05), mm5=(14 15)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=tmp7
+
+        movq      mm4,mm1               ; transpose coefficients
+        punpckldq mm1,mm3               ; mm1=(02 12)=data2
+        punpckhdq mm4,mm3               ; mm4=(03 13)=data3
+        movq      mm0,mm2               ; transpose coefficients
+        punpckldq mm2,mm5               ; mm2=(04 14)=data4
+        punpckhdq mm0,mm5               ; mm0=(05 15)=data5
+
+        movq    mm3,mm4
+        movq    mm5,mm1
+        pfadd   mm4,mm2                 ; mm4=data3+data4=tmp3
+        pfadd   mm1,mm0                 ; mm1=data2+data5=tmp2
+        pfsub   mm3,mm2                 ; mm3=data3-data4=tmp4
+        pfsub   mm5,mm0                 ; mm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm2,mm7
+        movq    mm0,mm6
+        pfsub   mm7,mm4                 ; mm7=tmp13
+        pfsub   mm6,mm1                 ; mm6=tmp12
+        pfadd   mm2,mm4                 ; mm2=tmp10
+        pfadd   mm0,mm1                 ; mm0=tmp11
+
+        pfadd   mm6,mm7
+        pfmul   mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1
+
+        movq    mm4,mm2
+        movq    mm1,mm7
+        pfsub   mm2,mm0                 ; mm2=data4
+        pfsub   mm7,mm6                 ; mm7=data6
+        pfadd   mm4,mm0                 ; mm4=data0
+        pfadd   mm1,mm6                 ; mm1=data2
+
+        movq    MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [wk(0)]     ; mm0=tmp6
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp7
+
+        pfadd   mm3,mm5                 ; mm3=tmp10
+        pfadd   mm5,mm0                 ; mm5=tmp11
+        pfadd   mm0,mm6                 ; mm0=tmp12, mm6=tmp7
+
+        pfmul   mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3
+
+        movq    mm2,mm3                 ; mm2=tmp10
+        pfsub   mm3,mm0
+        pfmul   mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5
+        pfmul   mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+        pfmul   mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+        pfadd   mm2,mm3                 ; mm2=z2
+        pfadd   mm0,mm3                 ; mm0=z4
+
+        movq    mm7,mm6
+        pfsub   mm6,mm5                 ; mm6=z13
+        pfadd   mm7,mm5                 ; mm7=z11
+
+        movq    mm4,mm6
+        movq    mm1,mm7
+        pfsub   mm6,mm2                 ; mm6=data3
+        pfsub   mm7,mm0                 ; mm7=data7
+        pfadd   mm4,mm2                 ; mm4=data5
+        pfadd   mm1,mm0                 ; mm1=data1
+
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        add     edx, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm0=(00 10), mm1=(01 11), mm2=(60 70), mm3=(61 71)
+
+        movq      mm4,mm0               ; transpose coefficients
+        punpckldq mm0,mm1               ; mm0=(00 01)=data0
+        punpckhdq mm4,mm1               ; mm4=(10 11)=data1
+        movq      mm5,mm2               ; transpose coefficients
+        punpckldq mm2,mm3               ; mm2=(60 61)=data6
+        punpckhdq mm5,mm3               ; mm5=(70 71)=data7
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm2                 ; mm4=data1-data6=tmp6
+        pfsub   mm0,mm5                 ; mm0=data0-data7=tmp7
+        pfadd   mm6,mm2                 ; mm6=data1+data6=tmp1
+        pfadd   mm7,mm5                 ; mm7=data0+data7=tmp0
+
+        movq    mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; mm1=(20 30), mm3=(21 31), mm2=(40 50), mm5=(41 51)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=tmp7
+
+        movq      mm4,mm1               ; transpose coefficients
+        punpckldq mm1,mm3               ; mm1=(20 21)=data2
+        punpckhdq mm4,mm3               ; mm4=(30 31)=data3
+        movq      mm0,mm2               ; transpose coefficients
+        punpckldq mm2,mm5               ; mm2=(40 41)=data4
+        punpckhdq mm0,mm5               ; mm0=(50 51)=data5
+
+        movq    mm3,mm4
+        movq    mm5,mm1
+        pfadd   mm4,mm2                 ; mm4=data3+data4=tmp3
+        pfadd   mm1,mm0                 ; mm1=data2+data5=tmp2
+        pfsub   mm3,mm2                 ; mm3=data3-data4=tmp4
+        pfsub   mm5,mm0                 ; mm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm2,mm7
+        movq    mm0,mm6
+        pfsub   mm7,mm4                 ; mm7=tmp13
+        pfsub   mm6,mm1                 ; mm6=tmp12
+        pfadd   mm2,mm4                 ; mm2=tmp10
+        pfadd   mm0,mm1                 ; mm0=tmp11
+
+        pfadd   mm6,mm7
+        pfmul   mm6,[GOTOFF(ebx,PD_0_707)] ; mm6=z1
+
+        movq    mm4,mm2
+        movq    mm1,mm7
+        pfsub   mm2,mm0                 ; mm2=data4
+        pfsub   mm7,mm6                 ; mm7=data6
+        pfadd   mm4,mm0                 ; mm4=data0
+        pfadd   mm1,mm6                 ; mm1=data2
+
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [wk(0)]     ; mm0=tmp6
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp7
+
+        pfadd   mm3,mm5                 ; mm3=tmp10
+        pfadd   mm5,mm0                 ; mm5=tmp11
+        pfadd   mm0,mm6                 ; mm0=tmp12, mm6=tmp7
+
+        pfmul   mm5,[GOTOFF(ebx,PD_0_707)] ; mm5=z3
+
+        movq    mm2,mm3                 ; mm2=tmp10
+        pfsub   mm3,mm0
+        pfmul   mm3,[GOTOFF(ebx,PD_0_382)] ; mm3=z5
+        pfmul   mm2,[GOTOFF(ebx,PD_0_541)] ; mm2=MULTIPLY(tmp10,FIX_0_54119610)
+        pfmul   mm0,[GOTOFF(ebx,PD_1_306)] ; mm0=MULTIPLY(tmp12,FIX_1_30656296)
+        pfadd   mm2,mm3                 ; mm2=z2
+        pfadd   mm0,mm3                 ; mm0=z4
+
+        movq    mm7,mm6
+        pfsub   mm6,mm5                 ; mm6=z13
+        pfadd   mm7,mm5                 ; mm7=z11
+
+        movq    mm4,mm6
+        movq    mm1,mm7
+        pfsub   mm6,mm2                 ; mm6=data3
+        pfsub   mm7,mm0                 ; mm7=data7
+        pfadd   mm4,mm2                 ; mm4=data5
+        pfadd   mm1,mm0                 ; mm1=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], mm1
+
+        add     edx, byte 2*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .columnloop
+
+        femms           ; empty MMX/3DNow! state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-sse-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,358 @@
+;
+; jfdctflt.asm - floating-point FDCT (64-bit SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_sse) PRIVATE
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382        times 4 dd  0.382683432365089771728460
+PD_0_707        times 4 dd  0.707106781186547524400844
+PD_0_541        times 4 dd  0.541196100146196984399723
+PD_1_306        times 4 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse (FAST_FLOAT *data)
+;
+
+; r10 = FAST_FLOAT *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_sse) PRIVATE
+
+EXTN(jsimd_fdct_float_sse):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (FAST_FLOAT *)
+        mov     rcx, DCTSIZE/4
+.rowloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+        ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(20 30 21 31)
+        unpckhps xmm4,xmm1              ; xmm4=(22 32 23 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(24 34 25 35)
+        unpckhps xmm5,xmm3              ; xmm5=(26 36 27 37)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+        ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 32 23 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(24 34 25 35)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 10 01 11)
+        unpckhps xmm4,xmm7              ; xmm4=(02 12 03 13)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(04 14 05 15)
+        unpckhps xmm2,xmm3              ; xmm2=(06 16 07 17)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 10 20 30)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(01 11 21 31)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(06 16 26 36)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(07 17 27 37)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 32 23 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(24 34 25 35)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(02 12 22 32)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(03 13 23 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(04 14 24 34)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(05 15 25 35)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[rel PD_0_707] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,1,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[rel PD_0_707] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[rel PD_0_382] ; xmm2=z5
+        mulps   xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(3,1,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     rdx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     rdx, r10        ; (FAST_FLOAT *)
+        mov     rcx, DCTSIZE/4
+.columnloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+        ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(02 03 12 13)
+        unpckhps xmm4,xmm1              ; xmm4=(22 23 32 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(42 43 52 53)
+        unpckhps xmm5,xmm3              ; xmm5=(62 63 72 73)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+        ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 23 32 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(42 43 52 53)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 01 10 11)
+        unpckhps xmm4,xmm7              ; xmm4=(20 21 30 31)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(40 41 50 51)
+        unpckhps xmm2,xmm3              ; xmm2=(60 61 70 71)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 01 02 03)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(10 11 12 13)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(60 61 62 63)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(70 71 72 73)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 23 32 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(20 21 22 23)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(30 31 32 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(40 41 42 43)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(50 51 52 53)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[rel PD_0_707] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[rel PD_0_707] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[rel PD_0_382] ; xmm2=z5
+        mulps   xmm1,[rel PD_0_541] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[rel PD_1_306] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     rdx, byte 4*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     near .columnloop
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-sse.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctflt-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctflt-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,370 @@
+;
+; jfdctflt.asm - floating-point FDCT (SSE)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the forward DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jfdctflt.c; see the jfdctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_float_sse) PRIVATE
+
+EXTN(jconst_fdct_float_sse):
+
+PD_0_382        times 4 dd  0.382683432365089771728460
+PD_0_707        times 4 dd  0.707106781186547524400844
+PD_0_541        times 4 dd  0.541196100146196984399723
+PD_1_306        times 4 dd  1.306562964876376527856643
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_float_sse (FAST_FLOAT *data)
+;
+
+%define data(b)         (b)+8           ; FAST_FLOAT *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_float_sse) PRIVATE
+
+EXTN(jsimd_fdct_float_sse):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(20 21 22 23), xmm2=(24 25 26 27)
+        ; xmm1=(30 31 32 33), xmm3=(34 35 36 37)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(20 30 21 31)
+        unpckhps xmm4,xmm1              ; xmm4=(22 32 23 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(24 34 25 35)
+        unpckhps xmm5,xmm3              ; xmm5=(26 36 27 37)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 01 02 03), xmm1=(04 05 06 07)
+        ; xmm7=(10 11 12 13), xmm3=(14 15 16 17)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 32 23 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(24 34 25 35)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 10 01 11)
+        unpckhps xmm4,xmm7              ; xmm4=(02 12 03 13)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(04 14 05 15)
+        unpckhps xmm2,xmm3              ; xmm2=(06 16 07 17)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 10 20 30)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(01 11 21 31)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(06 16 26 36)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(07 17 27 37)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 32 23 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(24 34 25 35)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(02 12 22 32)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(03 13 23 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(04 14 24 34)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(05 15 25 35)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,1,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5
+        mulps   xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(3,1,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     edx, 4*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (FAST_FLOAT *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm0=(02 12 22 32), xmm2=(42 52 62 72)
+        ; xmm1=(03 13 23 33), xmm3=(43 53 63 73)
+
+        movaps   xmm4,xmm0              ; transpose coefficients(phase 1)
+        unpcklps xmm0,xmm1              ; xmm0=(02 03 12 13)
+        unpckhps xmm4,xmm1              ; xmm4=(22 23 32 33)
+        movaps   xmm5,xmm2              ; transpose coefficients(phase 1)
+        unpcklps xmm2,xmm3              ; xmm2=(42 43 52 53)
+        unpckhps xmm5,xmm3              ; xmm5=(62 63 72 73)
+
+        movaps  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm7, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)]
+
+        ; xmm6=(00 10 20 30), xmm1=(40 50 60 70)
+        ; xmm7=(01 11 21 31), xmm3=(41 51 61 71)
+
+        movaps  XMMWORD [wk(0)], xmm4   ; wk(0)=(22 23 32 33)
+        movaps  XMMWORD [wk(1)], xmm2   ; wk(1)=(42 43 52 53)
+
+        movaps   xmm4,xmm6              ; transpose coefficients(phase 1)
+        unpcklps xmm6,xmm7              ; xmm6=(00 01 10 11)
+        unpckhps xmm4,xmm7              ; xmm4=(20 21 30 31)
+        movaps   xmm2,xmm1              ; transpose coefficients(phase 1)
+        unpcklps xmm1,xmm3              ; xmm1=(40 41 50 51)
+        unpckhps xmm2,xmm3              ; xmm2=(60 61 70 71)
+
+        movaps    xmm7,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm0             ; xmm6=(00 01 02 03)=data0
+        unpckhps2 xmm7,xmm0             ; xmm7=(10 11 12 13)=data1
+        movaps    xmm3,xmm2             ; transpose coefficients(phase 2)
+        unpcklps2 xmm2,xmm5             ; xmm2=(60 61 62 63)=data6
+        unpckhps2 xmm3,xmm5             ; xmm3=(70 71 72 73)=data7
+
+        movaps  xmm0,xmm7
+        movaps  xmm5,xmm6
+        subps   xmm7,xmm2               ; xmm7=data1-data6=tmp6
+        subps   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        addps   xmm0,xmm2               ; xmm0=data1+data6=tmp1
+        addps   xmm5,xmm3               ; xmm5=data0+data7=tmp0
+
+        movaps  xmm2, XMMWORD [wk(0)]   ; xmm2=(22 23 32 33)
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53)
+        movaps  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp6
+        movaps  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movaps    xmm7,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(20 21 22 23)=data2
+        unpckhps2 xmm7,xmm2             ; xmm7=(30 31 32 33)=data3
+        movaps    xmm6,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm3             ; xmm1=(40 41 42 43)=data4
+        unpckhps2 xmm6,xmm3             ; xmm6=(50 51 52 53)=data5
+
+        movaps  xmm2,xmm7
+        movaps  xmm3,xmm4
+        addps   xmm7,xmm1               ; xmm7=data3+data4=tmp3
+        addps   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        subps   xmm2,xmm1               ; xmm2=data3-data4=tmp4
+        subps   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movaps  xmm1,xmm5
+        movaps  xmm6,xmm0
+        subps   xmm5,xmm7               ; xmm5=tmp13
+        subps   xmm0,xmm4               ; xmm0=tmp12
+        addps   xmm1,xmm7               ; xmm1=tmp10
+        addps   xmm6,xmm4               ; xmm6=tmp11
+
+        addps   xmm0,xmm5
+        mulps   xmm0,[GOTOFF(ebx,PD_0_707)] ; xmm0=z1
+
+        movaps  xmm7,xmm1
+        movaps  xmm4,xmm5
+        subps   xmm1,xmm6               ; xmm1=data4
+        subps   xmm5,xmm0               ; xmm5=data6
+        addps   xmm7,xmm6               ; xmm7=data0
+        addps   xmm4,xmm0               ; xmm4=data2
+
+        movaps  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        ; -- Odd part
+
+        movaps  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp6
+        movaps  xmm0, XMMWORD [wk(1)]   ; xmm0=tmp7
+
+        addps   xmm2,xmm3               ; xmm2=tmp10
+        addps   xmm3,xmm6               ; xmm3=tmp11
+        addps   xmm6,xmm0               ; xmm6=tmp12, xmm0=tmp7
+
+        mulps   xmm3,[GOTOFF(ebx,PD_0_707)] ; xmm3=z3
+
+        movaps  xmm1,xmm2               ; xmm1=tmp10
+        subps   xmm2,xmm6
+        mulps   xmm2,[GOTOFF(ebx,PD_0_382)] ; xmm2=z5
+        mulps   xmm1,[GOTOFF(ebx,PD_0_541)] ; xmm1=MULTIPLY(tmp10,FIX_0_541196)
+        mulps   xmm6,[GOTOFF(ebx,PD_1_306)] ; xmm6=MULTIPLY(tmp12,FIX_1_306562)
+        addps   xmm1,xmm2               ; xmm1=z2
+        addps   xmm6,xmm2               ; xmm6=z4
+
+        movaps  xmm5,xmm0
+        subps   xmm0,xmm3               ; xmm0=z13
+        addps   xmm5,xmm3               ; xmm5=z11
+
+        movaps  xmm7,xmm0
+        movaps  xmm4,xmm5
+        subps   xmm0,xmm1               ; xmm0=data3
+        subps   xmm5,xmm6               ; xmm5=data7
+        addps   xmm7,xmm1               ; xmm7=data5
+        addps   xmm4,xmm6               ; xmm4=data1
+
+        movaps  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FAST_FLOAT)], xmm7
+        movaps  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)], xmm4
+
+        add     edx, byte 4*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .columnloop
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,156 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER FORWARD DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_382 98   /* FIX(0.382683433) */
+#define F_0_541 139  /* FIX(0.541196100) */
+#define F_0_707 181  /* FIX(0.707106781) */
+#define F_1_306 334  /* FIX(1.306562965) */
+
+#define CONST_BITS 8
+#define PRE_MULTIPLY_SCALE_BITS 2
+#define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_FDCT()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  \
+  z1 = vec_add(tmp12, tmp13);  \
+  z1 = vec_sl(z1, pre_multiply_scale_bits);  \
+  z1 = vec_madds(z1, pw_0707, pw_zero);  \
+  \
+  out2 = vec_add(tmp13, z1);  \
+  out6 = vec_sub(tmp13, z1);  \
+  \
+  /* Odd part */  \
+  \
+  tmp10 = vec_add(tmp4, tmp5);  \
+  tmp11 = vec_add(tmp5, tmp6);  \
+  tmp12 = vec_add(tmp6, tmp7);  \
+  \
+  tmp10 = vec_sl(tmp10, pre_multiply_scale_bits);  \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits);  \
+  z5 = vec_sub(tmp10, tmp12);  \
+  z5 = vec_madds(z5, pw_0382, pw_zero);  \
+  \
+  z2 = vec_madds(tmp10, pw_0541, z5);  \
+  z4 = vec_madds(tmp12, pw_1306, z5);  \
+  \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits);  \
+  z3 = vec_madds(tmp11, pw_0707, pw_zero);  \
+  \
+  z11 = vec_add(tmp7, z3);  \
+  z13 = vec_sub(tmp7, z3);  \
+  \
+  out5 = vec_add(z13, z2);  \
+  out3 = vec_sub(z13, z2);  \
+  out1 = vec_add(z11, z4);  \
+  out7 = vec_sub(z11, z4);  \
+}
+
+
+void
+jsimd_fdct_ifast_altivec (DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z1, z2, z3, z4, z5, z11, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_0382 = { __8X(F_0_382 << CONST_SHIFT) },
+    pw_0541 = { __8X(F_0_541 << CONST_SHIFT) },
+    pw_0707 = { __8X(F_0_707 << CONST_SHIFT) },
+    pw_1306 = { __8X(F_1_306 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,397 @@
+;
+; jfdctfst.asm - fast integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_mmx) PRIVATE
+
+EXTN(jconst_fdct_ifast_mmx):
+
+PW_F0707        times 4 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 4 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 4 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 4 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_mmx (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_mmx) PRIVATE
+
+EXTN(jsimd_fdct_ifast_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+        ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(24 34 25 35)
+        punpckhwd mm5,mm3               ; mm5=(26 36 27 37)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+        ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 32 23 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(24 34 25 35)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm4,mm7               ; mm4=(02 12 03 13)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(04 14 05 15)
+        punpckhwd mm2,mm3               ; mm2=(06 16 07 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 10 20 30)=data0
+        punpckhdq mm7,mm0               ; mm7=(01 11 21 31)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(06 16 26 36)=data6
+        punpckhdq mm3,mm5               ; mm3=(07 17 27 37)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 32 23 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(24 34 25 35)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(02 12 22 32)=data2
+        punpckhdq mm7,mm2               ; mm7=(03 13 23 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(04 14 24 34)=data4
+        punpckhdq mm6,mm3               ; mm6=(05 15 25 35)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        psubw   mm5,mm7                 ; mm5=tmp13
+        psubw   mm0,mm4                 ; mm0=tmp12
+        paddw   mm1,mm7                 ; mm1=tmp10
+        paddw   mm6,mm4                 ; mm6=tmp11
+
+        paddw   mm0,mm5
+        psllw   mm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1
+
+        movq    mm7,mm1
+        movq    mm4,mm5
+        psubw   mm1,mm6                 ; mm1=data4
+        psubw   mm5,mm0                 ; mm5=data6
+        paddw   mm7,mm6                 ; mm7=data0
+        paddw   mm4,mm0                 ; mm4=data2
+
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+
+        ; -- Odd part
+
+        movq    mm6, MMWORD [wk(0)]     ; mm6=tmp6
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp7
+
+        paddw   mm2,mm3                 ; mm2=tmp10
+        paddw   mm3,mm6                 ; mm3=tmp11
+        paddw   mm6,mm0                 ; mm6=tmp12, mm0=tmp7
+
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm6,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   mm3,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3
+
+        movq    mm1,mm2                 ; mm1=tmp10
+        psubw   mm2,mm6
+        pmulhw  mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5
+        pmulhw  mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+        pmulhw  mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+        paddw   mm1,mm2                 ; mm1=z2
+        paddw   mm6,mm2                 ; mm6=z4
+
+        movq    mm5,mm0
+        psubw   mm0,mm3                 ; mm0=z13
+        paddw   mm5,mm3                 ; mm5=z11
+
+        movq    mm7,mm0
+        movq    mm4,mm5
+        psubw   mm0,mm1                 ; mm0=data3
+        psubw   mm5,mm6                 ; mm5=data7
+        paddw   mm7,mm1                 ; mm7=data5
+        paddw   mm4,mm6                 ; mm4=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4
+
+        add     edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+        ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(02 03 12 13)
+        punpckhwd mm4,mm1               ; mm4=(22 23 32 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(42 43 52 53)
+        punpckhwd mm5,mm3               ; mm5=(62 63 72 73)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+        ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 23 32 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(42 43 52 53)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 01 10 11)
+        punpckhwd mm4,mm7               ; mm4=(20 21 30 31)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(40 41 50 51)
+        punpckhwd mm2,mm3               ; mm2=(60 61 70 71)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03)=data0
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(60 61 62 63)=data6
+        punpckhdq mm3,mm5               ; mm3=(70 71 72 73)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 23 32 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(42 43 52 53)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(20 21 22 23)=data2
+        punpckhdq mm7,mm2               ; mm7=(30 31 32 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(40 41 42 43)=data4
+        punpckhdq mm6,mm3               ; mm6=(50 51 52 53)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        psubw   mm5,mm7                 ; mm5=tmp13
+        psubw   mm0,mm4                 ; mm0=tmp12
+        paddw   mm1,mm7                 ; mm1=tmp10
+        paddw   mm6,mm4                 ; mm6=tmp11
+
+        paddw   mm0,mm5
+        psllw   mm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm0,[GOTOFF(ebx,PW_F0707)] ; mm0=z1
+
+        movq    mm7,mm1
+        movq    mm4,mm5
+        psubw   mm1,mm6                 ; mm1=data4
+        psubw   mm5,mm0                 ; mm5=data6
+        paddw   mm7,mm6                 ; mm7=data0
+        paddw   mm4,mm0                 ; mm4=data2
+
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+
+        ; -- Odd part
+
+        movq    mm6, MMWORD [wk(0)]     ; mm6=tmp6
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp7
+
+        paddw   mm2,mm3                 ; mm2=tmp10
+        paddw   mm3,mm6                 ; mm3=tmp11
+        paddw   mm6,mm0                 ; mm6=tmp12, mm0=tmp7
+
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm6,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   mm3,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm3,[GOTOFF(ebx,PW_F0707)] ; mm3=z3
+
+        movq    mm1,mm2                 ; mm1=tmp10
+        psubw   mm2,mm6
+        pmulhw  mm2,[GOTOFF(ebx,PW_F0382)] ; mm2=z5
+        pmulhw  mm1,[GOTOFF(ebx,PW_F0541)] ; mm1=MULTIPLY(tmp10,FIX_0_54119610)
+        pmulhw  mm6,[GOTOFF(ebx,PW_F1306)] ; mm6=MULTIPLY(tmp12,FIX_1_30656296)
+        paddw   mm1,mm2                 ; mm1=z2
+        paddw   mm6,mm2                 ; mm6=z4
+
+        movq    mm5,mm0
+        psubw   mm0,mm3                 ; mm0=z13
+        paddw   mm5,mm3                 ; mm5=z11
+
+        movq    mm7,mm0
+        movq    mm4,mm5
+        psubw   mm0,mm1                 ; mm0=data3
+        psubw   mm5,mm6                 ; mm5=data7
+        paddw   mm7,mm1                 ; mm7=data5
+        paddw   mm4,mm6                 ; mm4=data1
+
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm4
+
+        add     edx, byte 4*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .columnloop
+
+        emms            ; empty MMX state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,392 @@
+;
+; jfdctfst.asm - fast integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_sse2) PRIVATE
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707        times 8 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 8 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 8 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 8 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2 (DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_fdct_ifast_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(0)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        psubw   xmm3,xmm1               ; xmm3=tmp13
+        psubw   xmm6,xmm7               ; xmm6=tmp12
+        paddw   xmm4,xmm1               ; xmm4=tmp10
+        paddw   xmm0,xmm7               ; xmm0=tmp11
+
+        paddw   xmm6,xmm3
+        psllw   xmm6,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm6,[rel PW_F0707] ; xmm6=z1
+
+        movdqa  xmm1,xmm4
+        movdqa  xmm7,xmm3
+        psubw   xmm4,xmm0               ; xmm4=data4
+        psubw   xmm3,xmm6               ; xmm3=data6
+        paddw   xmm1,xmm0               ; xmm1=data0
+        paddw   xmm7,xmm6               ; xmm7=data2
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=tmp6
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp7
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=data4
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=data6
+
+        ; -- Odd part
+
+        paddw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm5,xmm0               ; xmm5=tmp11
+        paddw   xmm0,xmm6               ; xmm0=tmp12, xmm6=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F0707] ; xmm5=z3
+
+        movdqa  xmm4,xmm2               ; xmm4=tmp10
+        psubw   xmm2,xmm0
+        pmulhw  xmm2,[rel PW_F0382] ; xmm2=z5
+        pmulhw  xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm0,[rel PW_F1306] ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm2               ; xmm4=z2
+        paddw   xmm0,xmm2               ; xmm0=z4
+
+        movdqa  xmm3,xmm6
+        psubw   xmm6,xmm5               ; xmm6=z13
+        paddw   xmm3,xmm5               ; xmm3=z11
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm5,xmm3
+        psubw   xmm6,xmm4               ; xmm6=data3
+        psubw   xmm3,xmm0               ; xmm3=data7
+        paddw   xmm2,xmm4               ; xmm2=data5
+        paddw   xmm5,xmm0               ; xmm5=data1
+
+        ; ---- Pass 2: process columns.
+
+        ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+        ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm4,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm4,xmm5             ; xmm4=(40 41 50 51 60 61 70 71)
+        movdqa    xmm0,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm6             ; xmm7=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm0,xmm6             ; xmm0=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=col4
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=col6
+
+        ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+        ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm7,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm2             ; xmm5=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm7,xmm2             ; xmm7=(44 45 54 55 64 65 74 75)
+        movdqa    xmm0,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm3             ; xmm6=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm0,xmm3             ; xmm0=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm2,xmm5             ; transpose coefficients(phase 2)
+        punpckldq xmm5,xmm6             ; xmm5=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm2,xmm6             ; xmm2=(24 25 26 27 34 35 36 37)
+        movdqa    xmm3,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm0             ; xmm7=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm3,xmm0             ; xmm3=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=(02 03 12 13 22 23 32 33)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm6             ; xmm1=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm2,xmm6             ; xmm2=(20 21 22 23 30 31 32 33)
+        movdqa    xmm7,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm7,xmm0             ; xmm7=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm5            ; xmm1=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm6,xmm5            ; xmm6=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm0,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm3            ; xmm7=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm0,xmm3            ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm3,xmm1
+        psubw   xmm6,xmm7               ; xmm6=data1-data6=tmp6
+        psubw   xmm1,xmm0               ; xmm1=data0-data7=tmp7
+        paddw   xmm5,xmm7               ; xmm5=data1+data6=tmp1
+        paddw   xmm3,xmm0               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(24 25 26 27 34 35 36 37)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=tmp7
+
+        movdqa     xmm6,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm7            ; xmm2=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm6,xmm7            ; xmm6=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm1,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm1,xmm0            ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm7,xmm6
+        movdqa  xmm0,xmm2
+        paddw   xmm6,xmm4               ; xmm6=data3+data4=tmp3
+        paddw   xmm2,xmm1               ; xmm2=data2+data5=tmp2
+        psubw   xmm7,xmm4               ; xmm7=data3-data4=tmp4
+        psubw   xmm0,xmm1               ; xmm0=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm1,xmm5
+        psubw   xmm3,xmm6               ; xmm3=tmp13
+        psubw   xmm5,xmm2               ; xmm5=tmp12
+        paddw   xmm4,xmm6               ; xmm4=tmp10
+        paddw   xmm1,xmm2               ; xmm1=tmp11
+
+        paddw   xmm5,xmm3
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F0707] ; xmm5=z1
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm2,xmm3
+        psubw   xmm4,xmm1               ; xmm4=data4
+        psubw   xmm3,xmm5               ; xmm3=data6
+        paddw   xmm6,xmm1               ; xmm6=data0
+        paddw   xmm2,xmm5               ; xmm2=data2
+
+        movdqa  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm2
+
+        ; -- Odd part
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        paddw   xmm7,xmm0               ; xmm7=tmp10
+        paddw   xmm0,xmm1               ; xmm0=tmp11
+        paddw   xmm1,xmm5               ; xmm1=tmp12, xmm5=tmp7
+
+        psllw   xmm7,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm0,[rel PW_F0707] ; xmm0=z3
+
+        movdqa  xmm4,xmm7               ; xmm4=tmp10
+        psubw   xmm7,xmm1
+        pmulhw  xmm7,[rel PW_F0382] ; xmm7=z5
+        pmulhw  xmm4,[rel PW_F0541] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm1,[rel PW_F1306] ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm7               ; xmm4=z2
+        paddw   xmm1,xmm7               ; xmm1=z4
+
+        movdqa  xmm3,xmm5
+        psubw   xmm5,xmm0               ; xmm5=z13
+        paddw   xmm3,xmm0               ; xmm3=z11
+
+        movdqa  xmm6,xmm5
+        movdqa  xmm2,xmm3
+        psubw   xmm5,xmm4               ; xmm5=data3
+        psubw   xmm3,xmm1               ; xmm3=data7
+        paddw   xmm6,xmm4               ; xmm6=data5
+        paddw   xmm2,xmm1               ; xmm2=data1
+
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm5
+        movdqa  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm2
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctfst-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctfst-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,404 @@
+;
+; jfdctfst.asm - fast integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the forward DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jfdctfst.c; see the jfdctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+
+%if CONST_BITS == 8
+F_0_382 equ      98             ; FIX(0.382683433)
+F_0_541 equ     139             ; FIX(0.541196100)
+F_0_707 equ     181             ; FIX(0.707106781)
+F_1_306 equ     334             ; FIX(1.306562965)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_382 equ     DESCALE( 410903207,30-CONST_BITS)       ; FIX(0.382683433)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_707 equ     DESCALE( 759250124,30-CONST_BITS)       ; FIX(0.707106781)
+F_1_306 equ     DESCALE(1402911301,30-CONST_BITS)       ; FIX(1.306562965)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_fdct_ifast_sse2) PRIVATE
+
+EXTN(jconst_fdct_ifast_sse2):
+
+PW_F0707        times 8 dw  F_0_707 << CONST_SHIFT
+PW_F0382        times 8 dw  F_0_382 << CONST_SHIFT
+PW_F0541        times 8 dw  F_0_541 << CONST_SHIFT
+PW_F1306        times 8 dw  F_1_306 << CONST_SHIFT
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_ifast_sse2 (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_fdct_ifast_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(0)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        psubw   xmm3,xmm1               ; xmm3=tmp13
+        psubw   xmm6,xmm7               ; xmm6=tmp12
+        paddw   xmm4,xmm1               ; xmm4=tmp10
+        paddw   xmm0,xmm7               ; xmm0=tmp11
+
+        paddw   xmm6,xmm3
+        psllw   xmm6,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm6,[GOTOFF(ebx,PW_F0707)] ; xmm6=z1
+
+        movdqa  xmm1,xmm4
+        movdqa  xmm7,xmm3
+        psubw   xmm4,xmm0               ; xmm4=data4
+        psubw   xmm3,xmm6               ; xmm3=data6
+        paddw   xmm1,xmm0               ; xmm1=data0
+        paddw   xmm7,xmm6               ; xmm7=data2
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=tmp6
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp7
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=data4
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=data6
+
+        ; -- Odd part
+
+        paddw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm5,xmm0               ; xmm5=tmp11
+        paddw   xmm0,xmm6               ; xmm0=tmp12, xmm6=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z3
+
+        movdqa  xmm4,xmm2               ; xmm4=tmp10
+        psubw   xmm2,xmm0
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F0382)] ; xmm2=z5
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F1306)] ; xmm0=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm2               ; xmm4=z2
+        paddw   xmm0,xmm2               ; xmm0=z4
+
+        movdqa  xmm3,xmm6
+        psubw   xmm6,xmm5               ; xmm6=z13
+        paddw   xmm3,xmm5               ; xmm3=z11
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm5,xmm3
+        psubw   xmm6,xmm4               ; xmm6=data3
+        psubw   xmm3,xmm0               ; xmm3=data7
+        paddw   xmm2,xmm4               ; xmm2=data5
+        paddw   xmm5,xmm0               ; xmm5=data1
+
+        ; ---- Pass 2: process columns.
+
+;       mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        ; xmm1=(00 10 20 30 40 50 60 70), xmm7=(02 12 22 32 42 52 62 72)
+        ; xmm5=(01 11 21 31 41 51 61 71), xmm6=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm4,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm4,xmm5             ; xmm4=(40 41 50 51 60 61 70 71)
+        movdqa    xmm0,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm6             ; xmm7=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm0,xmm6             ; xmm0=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=col4
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=col6
+
+        ; xmm5=(04 14 24 34 44 54 64 74), xmm6=(06 16 26 36 46 56 66 76)
+        ; xmm2=(05 15 25 35 45 55 65 75), xmm3=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm7,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm2             ; xmm5=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm7,xmm2             ; xmm7=(44 45 54 55 64 65 74 75)
+        movdqa    xmm0,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm3             ; xmm6=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm0,xmm3             ; xmm0=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm2,xmm5             ; transpose coefficients(phase 2)
+        punpckldq xmm5,xmm6             ; xmm5=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm2,xmm6             ; xmm2=(24 25 26 27 34 35 36 37)
+        movdqa    xmm3,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm0             ; xmm7=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm3,xmm0             ; xmm3=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=(02 03 12 13 22 23 32 33)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm6             ; xmm1=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm2,xmm6             ; xmm2=(20 21 22 23 30 31 32 33)
+        movdqa    xmm7,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm7,xmm0             ; xmm7=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm5            ; xmm1=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm6,xmm5            ; xmm6=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm0,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm3            ; xmm7=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm0,xmm3            ; xmm0=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm3,xmm1
+        psubw   xmm6,xmm7               ; xmm6=data1-data6=tmp6
+        psubw   xmm1,xmm0               ; xmm1=data0-data7=tmp7
+        paddw   xmm5,xmm7               ; xmm5=data1+data6=tmp1
+        paddw   xmm3,xmm0               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(24 25 26 27 34 35 36 37)
+        movdqa  xmm0, XMMWORD [wk(1)]   ; xmm0=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm6   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=tmp7
+
+        movdqa     xmm6,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm7            ; xmm2=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm6,xmm7            ; xmm6=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm1,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm1,xmm0            ; xmm1=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm7,xmm6
+        movdqa  xmm0,xmm2
+        paddw   xmm6,xmm4               ; xmm6=data3+data4=tmp3
+        paddw   xmm2,xmm1               ; xmm2=data2+data5=tmp2
+        psubw   xmm7,xmm4               ; xmm7=data3-data4=tmp4
+        psubw   xmm0,xmm1               ; xmm0=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm1,xmm5
+        psubw   xmm3,xmm6               ; xmm3=tmp13
+        psubw   xmm5,xmm2               ; xmm5=tmp12
+        paddw   xmm4,xmm6               ; xmm4=tmp10
+        paddw   xmm1,xmm2               ; xmm1=tmp11
+
+        paddw   xmm5,xmm3
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F0707)] ; xmm5=z1
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm2,xmm3
+        psubw   xmm4,xmm1               ; xmm4=data4
+        psubw   xmm3,xmm5               ; xmm3=data6
+        paddw   xmm6,xmm1               ; xmm6=data0
+        paddw   xmm2,xmm5               ; xmm2=data2
+
+        movdqa  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm2
+
+        ; -- Odd part
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        paddw   xmm7,xmm0               ; xmm7=tmp10
+        paddw   xmm0,xmm1               ; xmm0=tmp11
+        paddw   xmm1,xmm5               ; xmm1=tmp12, xmm5=tmp7
+
+        psllw   xmm7,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F0707)] ; xmm0=z3
+
+        movdqa  xmm4,xmm7               ; xmm4=tmp10
+        psubw   xmm7,xmm1
+        pmulhw  xmm7,[GOTOFF(ebx,PW_F0382)] ; xmm7=z5
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F0541)] ; xmm4=MULTIPLY(tmp10,FIX_0_541196)
+        pmulhw  xmm1,[GOTOFF(ebx,PW_F1306)] ; xmm1=MULTIPLY(tmp12,FIX_1_306562)
+        paddw   xmm4,xmm7               ; xmm4=z2
+        paddw   xmm1,xmm7               ; xmm1=z4
+
+        movdqa  xmm3,xmm5
+        psubw   xmm5,xmm0               ; xmm5=z13
+        paddw   xmm3,xmm0               ; xmm3=z11
+
+        movdqa  xmm6,xmm5
+        movdqa  xmm2,xmm3
+        psubw   xmm5,xmm4               ; xmm5=data3
+        psubw   xmm3,xmm1               ; xmm3=data7
+        paddw   xmm6,xmm4               ; xmm6=data5
+        paddw   xmm2,xmm1               ; xmm2=data1
+
+        movdqa  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm5
+        movdqa  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm3
+        movdqa  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm6
+        movdqa  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm2
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,262 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* SLOW INTEGER FORWARD DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298 2446   /* FIX(0.298631336) */
+#define F_0_390 3196   /* FIX(0.390180644) */
+#define F_0_541 4433   /* FIX(0.541196100) */
+#define F_0_765 6270   /* FIX(0.765366865) */
+#define F_0_899 7373   /* FIX(0.899976223) */
+#define F_1_175 9633   /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+#define DESCALE_P1 (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS + PASS1_BITS)
+
+
+#define DO_FDCT_COMMON(PASS)  \
+{  \
+  /* (Original)  \
+   * z1 = (tmp12 + tmp13) * 0.541196100;  \
+   * data2 = z1 + tmp13 * 0.765366865;  \
+   * data6 = z1 + tmp12 * -1.847759065;  \
+   *  \
+   * (This implementation)  \
+   * data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;  \
+   * data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);  \
+   */  \
+  \
+  tmp1312l = vec_mergeh(tmp13, tmp12);  \
+  tmp1312h = vec_mergel(tmp13, tmp12);  \
+  \
+  out2l = vec_msums(tmp1312l, pw_f130_f054, pd_descale_p##PASS);  \
+  out2h = vec_msums(tmp1312h, pw_f130_f054, pd_descale_p##PASS);  \
+  out6l = vec_msums(tmp1312l, pw_f054_mf130, pd_descale_p##PASS);  \
+  out6h = vec_msums(tmp1312h, pw_f054_mf130, pd_descale_p##PASS);  \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS);  \
+  out2h = vec_sra(out2h, descale_p##PASS);  \
+  out6l = vec_sra(out6l, descale_p##PASS);  \
+  out6h = vec_sra(out6h, descale_p##PASS);  \
+  \
+  out2 = vec_pack(out2l, out2h);  \
+  out6 = vec_pack(out6l, out6h);  \
+  \
+  /* Odd part */  \
+  \
+  z3 = vec_add(tmp4, tmp6);  \
+  z4 = vec_add(tmp5, tmp7);  \
+  \
+  /* (Original)  \
+   * z5 = (z3 + z4) * 1.175875602;  \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;  \
+   * z3 += z5;  z4 += z5;  \
+   *  \
+   * (This implementation)  \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;  \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);  \
+   */  \
+  \
+  z34l = vec_mergeh(z3, z4);  \
+  z34h = vec_mergel(z3, z4);  \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_descale_p##PASS);  \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_descale_p##PASS);  \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_descale_p##PASS);  \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_descale_p##PASS);  \
+  \
+  /* (Original)  \
+   * z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;  \
+   * tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;  \
+   * tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;  \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;  \
+   * data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;  \
+   * data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;  \
+   *  \
+   * (This implementation)  \
+   * tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;  \
+   * tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;  \
+   * tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);  \
+   * tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);  \
+   * data7 = tmp4 + z3;  data5 = tmp5 + z4;  \
+   * data3 = tmp6 + z3;  data1 = tmp7 + z4;  \
+   */  \
+  \
+  tmp47l = vec_mergeh(tmp4, tmp7);  \
+  tmp47h = vec_mergel(tmp4, tmp7);  \
+  \
+  out7l = vec_msums(tmp47l, pw_mf060_mf089, z3l);  \
+  out7h = vec_msums(tmp47h, pw_mf060_mf089, z3h);  \
+  out1l = vec_msums(tmp47l, pw_mf089_f060, z4l);  \
+  out1h = vec_msums(tmp47h, pw_mf089_f060, z4h);  \
+  \
+  out7l = vec_sra(out7l, descale_p##PASS);  \
+  out7h = vec_sra(out7h, descale_p##PASS);  \
+  out1l = vec_sra(out1l, descale_p##PASS);  \
+  out1h = vec_sra(out1h, descale_p##PASS);  \
+  \
+  out7 = vec_pack(out7l, out7h);  \
+  out1 = vec_pack(out1l, out1h);  \
+  \
+  tmp56l = vec_mergeh(tmp5, tmp6);  \
+  tmp56h = vec_mergel(tmp5, tmp6);  \
+  \
+  out5l = vec_msums(tmp56l, pw_mf050_mf256, z4l);  \
+  out5h = vec_msums(tmp56h, pw_mf050_mf256, z4h);  \
+  out3l = vec_msums(tmp56l, pw_mf256_f050, z3l);  \
+  out3h = vec_msums(tmp56h, pw_mf256_f050, z3h);  \
+  \
+  out5l = vec_sra(out5l, descale_p##PASS);  \
+  out5h = vec_sra(out5h, descale_p##PASS);  \
+  out3l = vec_sra(out3l, descale_p##PASS);  \
+  out3h = vec_sra(out3h, descale_p##PASS);  \
+  \
+  out5 = vec_pack(out5l, out5h);  \
+  out3 = vec_pack(out3l, out3h);  \
+}
+
+#define DO_FDCT_PASS1()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out0  = vec_sl(out0, pass1_bits);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  out4  = vec_sl(out4, pass1_bits);  \
+  \
+  DO_FDCT_COMMON(1);  \
+}
+
+#define DO_FDCT_PASS2()  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(tmp0, tmp3);  \
+  tmp13 = vec_sub(tmp0, tmp3);  \
+  tmp11 = vec_add(tmp1, tmp2);  \
+  tmp12 = vec_sub(tmp1, tmp2);  \
+  \
+  out0  = vec_add(tmp10, tmp11);  \
+  out0  = vec_add(out0, pw_descale_p2x);  \
+  out0  = vec_sra(out0, pass1_bits);  \
+  out4  = vec_sub(tmp10, tmp11);  \
+  out4  = vec_add(out4, pw_descale_p2x);  \
+  out4  = vec_sra(out4, pass1_bits);  \
+  \
+  DO_FDCT_COMMON(2);  \
+}
+
+
+void
+jsimd_fdct_islow_altivec (DCTELEM *data)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    tmp47l, tmp47h, tmp56l, tmp56h, tmp1312l, tmp1312h,
+    z3, z4, z34l, z34h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int z3l, z3h, z4l, z4h,
+    out1l, out1h, out2l, out2h, out3l, out3h, out5l, out5h, out6l, out6h,
+    out7l, out7h;
+
+  /* Constants */
+  __vector short
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) },
+    pw_descale_p2x = { __8X(1 << (PASS1_BITS - 1)) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) };
+
+  /* Pass 1: process rows */
+
+  row0 = vec_ld(0, data);
+  row1 = vec_ld(16, data);
+  row2 = vec_ld(32, data);
+  row3 = vec_ld(48, data);
+  row4 = vec_ld(64, data);
+  row5 = vec_ld(80, data);
+  row6 = vec_ld(96, data);
+  row7 = vec_ld(112, data);
+
+  TRANSPOSE(row, col);
+
+  tmp0 = vec_add(col0, col7);
+  tmp7 = vec_sub(col0, col7);
+  tmp1 = vec_add(col1, col6);
+  tmp6 = vec_sub(col1, col6);
+  tmp2 = vec_add(col2, col5);
+  tmp5 = vec_sub(col2, col5);
+  tmp3 = vec_add(col3, col4);
+  tmp4 = vec_sub(col3, col4);
+
+  DO_FDCT_PASS1();
+
+  /* Pass 2: process columns */
+
+  TRANSPOSE(out, row);
+
+  tmp0 = vec_add(row0, row7);
+  tmp7 = vec_sub(row0, row7);
+  tmp1 = vec_add(row1, row6);
+  tmp6 = vec_sub(row1, row6);
+  tmp2 = vec_add(row2, row5);
+  tmp5 = vec_sub(row2, row5);
+  tmp3 = vec_add(row3, row4);
+  tmp4 = vec_sub(row3, row4);
+
+  DO_FDCT_PASS2();
+
+  vec_st(out0, 0, data);
+  vec_st(out1, 16, data);
+  vec_st(out2, 32, data);
+  vec_st(out3, 48, data);
+  vec_st(out4, 64, data);
+  vec_st(out5, 80, data);
+  vec_st(out6, 96, data);
+  vec_st(out7, 112, data);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,622 @@
+;
+; jfdctint.asm - accurate integer FDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_mmx) PRIVATE
+
+EXTN(jconst_fdct_islow_mmx):
+
+PW_F130_F054    times 2 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 2 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 2 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 2 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 2 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 2 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 2 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 2 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 2 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 2 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 4 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_mmx (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_mmx) PRIVATE
+
+EXTN(jsimd_fdct_islow_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.rowloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(20 21 22 23), mm2=(24 25 26 27)
+        ; mm1=(30 31 32 33), mm3=(34 35 36 37)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(24 34 25 35)
+        punpckhwd mm5,mm3               ; mm5=(26 36 27 37)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 01 02 03), mm1=(04 05 06 07)
+        ; mm7=(10 11 12 13), mm3=(14 15 16 17)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 32 23 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(24 34 25 35)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm4,mm7               ; mm4=(02 12 03 13)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(04 14 05 15)
+        punpckhwd mm2,mm3               ; mm2=(06 16 07 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 10 20 30)=data0
+        punpckhdq mm7,mm0               ; mm7=(01 11 21 31)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(06 16 26 36)=data6
+        punpckhdq mm3,mm5               ; mm3=(07 17 27 37)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 32 23 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(24 34 25 35)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(02 12 22 32)=data2
+        punpckhdq mm7,mm2               ; mm7=(03 13 23 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(04 14 24 34)=data4
+        punpckhdq mm6,mm3               ; mm6=(05 15 25 35)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        paddw   mm5,mm7                 ; mm5=tmp10
+        paddw   mm0,mm4                 ; mm0=tmp11
+        psubw   mm1,mm7                 ; mm1=tmp13
+        psubw   mm6,mm4                 ; mm6=tmp12
+
+        movq    mm7,mm5
+        paddw   mm5,mm0                 ; mm5=tmp10+tmp11
+        psubw   mm7,mm0                 ; mm7=tmp10-tmp11
+
+        psllw   mm5,PASS1_BITS          ; mm5=data0
+        psllw   mm7,PASS1_BITS          ; mm7=data4
+
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(0,1,edx,SIZEOF_DCTELEM)], mm7
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movq      mm4,mm1               ; mm1=tmp13
+        movq      mm0,mm1
+        punpcklwd mm4,mm6               ; mm6=tmp12
+        punpckhwd mm0,mm6
+        movq      mm1,mm4
+        movq      mm6,mm0
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=data2L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F130_F054)]        ; mm0=data2H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=data6L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F054_MF130)]       ; mm6=data6H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm0,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm4,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm6,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm1,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm4,mm0               ; mm4=data2
+        packssdw  mm1,mm6               ; mm1=data6
+
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(2,1,edx,SIZEOF_DCTELEM)], mm1
+
+        ; -- Odd part
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp6
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp7
+
+        movq    mm0,mm2                 ; mm2=tmp4
+        movq    mm6,mm3                 ; mm3=tmp5
+        paddw   mm0,mm5                 ; mm0=z3
+        paddw   mm6,mm7                 ; mm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm4,mm0
+        movq      mm1,mm0
+        punpcklwd mm4,mm6
+        punpckhwd mm1,mm6
+        movq      mm0,mm4
+        movq      mm6,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF078_F117)]       ; mm4=z3L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF078_F117)]       ; mm1=z3H
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F117_F078)]        ; mm0=z4L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F117_F078)]        ; mm6=z4H
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=z3L
+        movq    MMWORD [wk(1)], mm1     ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movq      mm4,mm2
+        movq      mm1,mm2
+        punpcklwd mm4,mm7
+        punpckhwd mm1,mm7
+        movq      mm2,mm4
+        movq      mm7,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm4=tmp4L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm1=tmp4H
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF089_F060)]       ; mm2=tmp7L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF089_F060)]       ; mm7=tmp7H
+
+        paddd   mm4, MMWORD [wk(0)]     ; mm4=data7L
+        paddd   mm1, MMWORD [wk(1)]     ; mm1=data7H
+        paddd   mm2,mm0                 ; mm2=data1L
+        paddd   mm7,mm6                 ; mm7=data1H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm4,DESCALE_P1
+        psrad   mm1,DESCALE_P1
+        paddd   mm2,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm2,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+
+        packssdw  mm4,mm1               ; mm4=data7
+        packssdw  mm2,mm7               ; mm2=data1
+
+        movq    MMWORD [MMBLOCK(3,1,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2
+
+        movq      mm1,mm3
+        movq      mm7,mm3
+        punpcklwd mm1,mm5
+        punpckhwd mm7,mm5
+        movq      mm3,mm1
+        movq      mm5,mm7
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm1=tmp5L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm7=tmp5H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF256_F050)]       ; mm3=tmp6L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_MF256_F050)]       ; mm5=tmp6H
+
+        paddd   mm1,mm0                 ; mm1=data5L
+        paddd   mm7,mm6                 ; mm7=data5H
+        paddd   mm3, MMWORD [wk(0)]     ; mm3=data3L
+        paddd   mm5, MMWORD [wk(1)]     ; mm5=data3H
+
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm1,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+        paddd   mm3,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   mm5,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   mm3,DESCALE_P1
+        psrad   mm5,DESCALE_P1
+
+        packssdw  mm1,mm7               ; mm1=data5
+        packssdw  mm3,mm5               ; mm3=data3
+
+        movq    MMWORD [MMBLOCK(1,1,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3
+
+        add     edx, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .rowloop
+
+        ; ---- Pass 2: process columns.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.columnloop:
+
+        movq    mm0, MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+        movq    mm2, MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm0=(02 12 22 32), mm2=(42 52 62 72)
+        ; mm1=(03 13 23 33), mm3=(43 53 63 73)
+
+        movq      mm4,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm1               ; mm0=(02 03 12 13)
+        punpckhwd mm4,mm1               ; mm4=(22 23 32 33)
+        movq      mm5,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm3               ; mm2=(42 43 52 53)
+        punpckhwd mm5,mm3               ; mm5=(62 63 72 73)
+
+        movq    mm6, MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movq    mm7, MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movq    mm1, MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+
+        ; mm6=(00 10 20 30), mm1=(40 50 60 70)
+        ; mm7=(01 11 21 31), mm3=(41 51 61 71)
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=(22 23 32 33)
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=(42 43 52 53)
+
+        movq      mm4,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 01 10 11)
+        punpckhwd mm4,mm7               ; mm4=(20 21 30 31)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm3               ; mm1=(40 41 50 51)
+        punpckhwd mm2,mm3               ; mm2=(60 61 70 71)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03)=data0
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13)=data1
+        movq      mm3,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm5               ; mm2=(60 61 62 63)=data6
+        punpckhdq mm3,mm5               ; mm3=(70 71 72 73)=data7
+
+        movq    mm0,mm7
+        movq    mm5,mm6
+        psubw   mm7,mm2                 ; mm7=data1-data6=tmp6
+        psubw   mm6,mm3                 ; mm6=data0-data7=tmp7
+        paddw   mm0,mm2                 ; mm0=data1+data6=tmp1
+        paddw   mm5,mm3                 ; mm5=data0+data7=tmp0
+
+        movq    mm2, MMWORD [wk(0)]     ; mm2=(22 23 32 33)
+        movq    mm3, MMWORD [wk(1)]     ; mm3=(42 43 52 53)
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp6
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=tmp7
+
+        movq      mm7,mm4               ; transpose coefficients(phase 2)
+        punpckldq mm4,mm2               ; mm4=(20 21 22 23)=data2
+        punpckhdq mm7,mm2               ; mm7=(30 31 32 33)=data3
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm3               ; mm1=(40 41 42 43)=data4
+        punpckhdq mm6,mm3               ; mm6=(50 51 52 53)=data5
+
+        movq    mm2,mm7
+        movq    mm3,mm4
+        paddw   mm7,mm1                 ; mm7=data3+data4=tmp3
+        paddw   mm4,mm6                 ; mm4=data2+data5=tmp2
+        psubw   mm2,mm1                 ; mm2=data3-data4=tmp4
+        psubw   mm3,mm6                 ; mm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movq    mm1,mm5
+        movq    mm6,mm0
+        paddw   mm5,mm7                 ; mm5=tmp10
+        paddw   mm0,mm4                 ; mm0=tmp11
+        psubw   mm1,mm7                 ; mm1=tmp13
+        psubw   mm6,mm4                 ; mm6=tmp12
+
+        movq    mm7,mm5
+        paddw   mm5,mm0                 ; mm5=tmp10+tmp11
+        psubw   mm7,mm0                 ; mm7=tmp10-tmp11
+
+        paddw   mm5,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        paddw   mm7,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        psraw   mm5,PASS1_BITS          ; mm5=data0
+        psraw   mm7,PASS1_BITS          ; mm7=data4
+
+        movq    MMWORD [MMBLOCK(0,0,edx,SIZEOF_DCTELEM)], mm5
+        movq    MMWORD [MMBLOCK(4,0,edx,SIZEOF_DCTELEM)], mm7
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movq      mm4,mm1               ; mm1=tmp13
+        movq      mm0,mm1
+        punpcklwd mm4,mm6               ; mm6=tmp12
+        punpckhwd mm0,mm6
+        movq      mm1,mm4
+        movq      mm6,mm0
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=data2L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F130_F054)]        ; mm0=data2H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=data6L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F054_MF130)]       ; mm6=data6H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm0,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm4,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm6,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm1,DESCALE_P2
+        psrad   mm6,DESCALE_P2
+
+        packssdw  mm4,mm0               ; mm4=data2
+        packssdw  mm1,mm6               ; mm1=data6
+
+        movq    MMWORD [MMBLOCK(2,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(6,0,edx,SIZEOF_DCTELEM)], mm1
+
+        ; -- Odd part
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp6
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp7
+
+        movq    mm0,mm2                 ; mm2=tmp4
+        movq    mm6,mm3                 ; mm3=tmp5
+        paddw   mm0,mm5                 ; mm0=z3
+        paddw   mm6,mm7                 ; mm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm4,mm0
+        movq      mm1,mm0
+        punpcklwd mm4,mm6
+        punpckhwd mm1,mm6
+        movq      mm0,mm4
+        movq      mm6,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF078_F117)]       ; mm4=z3L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF078_F117)]       ; mm1=z3H
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F117_F078)]        ; mm0=z4L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_F117_F078)]        ; mm6=z4H
+
+        movq    MMWORD [wk(0)], mm4     ; wk(0)=z3L
+        movq    MMWORD [wk(1)], mm1     ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movq      mm4,mm2
+        movq      mm1,mm2
+        punpcklwd mm4,mm7
+        punpckhwd mm1,mm7
+        movq      mm2,mm4
+        movq      mm7,mm1
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm4=tmp4L
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm1=tmp4H
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF089_F060)]       ; mm2=tmp7L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF089_F060)]       ; mm7=tmp7H
+
+        paddd   mm4, MMWORD [wk(0)]     ; mm4=data7L
+        paddd   mm1, MMWORD [wk(1)]     ; mm1=data7H
+        paddd   mm2,mm0                 ; mm2=data1L
+        paddd   mm7,mm6                 ; mm7=data1H
+
+        paddd   mm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm4,DESCALE_P2
+        psrad   mm1,DESCALE_P2
+        paddd   mm2,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm2,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+
+        packssdw  mm4,mm1               ; mm4=data7
+        packssdw  mm2,mm7               ; mm2=data1
+
+        movq    MMWORD [MMBLOCK(7,0,edx,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edx,SIZEOF_DCTELEM)], mm2
+
+        movq      mm1,mm3
+        movq      mm7,mm3
+        punpcklwd mm1,mm5
+        punpckhwd mm7,mm5
+        movq      mm3,mm1
+        movq      mm5,mm7
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm1=tmp5L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm7=tmp5H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF256_F050)]       ; mm3=tmp6L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_MF256_F050)]       ; mm5=tmp6H
+
+        paddd   mm1,mm0                 ; mm1=data5L
+        paddd   mm7,mm6                 ; mm7=data5H
+        paddd   mm3, MMWORD [wk(0)]     ; mm3=data3L
+        paddd   mm5, MMWORD [wk(1)]     ; mm5=data3H
+
+        paddd   mm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm1,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+        paddd   mm3,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   mm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   mm3,DESCALE_P2
+        psrad   mm5,DESCALE_P2
+
+        packssdw  mm1,mm7               ; mm1=data5
+        packssdw  mm3,mm5               ; mm3=data3
+
+        movq    MMWORD [MMBLOCK(5,0,edx,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(3,0,edx,SIZEOF_DCTELEM)], mm3
+
+        add     edx, byte 4*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     near .columnloop
+
+        emms            ; empty MMX state
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,622 @@
+;
+; jfdctint.asm - accurate integer FDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_sse2) PRIVATE
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 8 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2 (DCTELEM *data)
+;
+
+; r10 = DCTELEM *data
+
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          6
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_sse2) PRIVATE
+
+EXTN(jsimd_fdct_islow_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process rows.
+
+        mov     rdx, r10        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(3)], xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(3)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        paddw   xmm3,xmm1               ; xmm3=tmp10
+        paddw   xmm6,xmm7               ; xmm6=tmp11
+        psubw   xmm4,xmm1               ; xmm4=tmp13
+        psubw   xmm0,xmm7               ; xmm0=tmp12
+
+        movdqa  xmm1,xmm3
+        paddw   xmm3,xmm6               ; xmm3=tmp10+tmp11
+        psubw   xmm1,xmm6               ; xmm1=tmp10-tmp11
+
+        psllw   xmm3,PASS1_BITS         ; xmm3=data0
+        psllw   xmm1,PASS1_BITS         ; xmm1=data4
+
+        movdqa  XMMWORD [wk(2)], xmm3   ; wk(2)=data0
+        movdqa  XMMWORD [wk(3)], xmm1   ; wk(3)=data4
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm7,xmm4             ; xmm4=tmp13
+        movdqa    xmm6,xmm4
+        punpcklwd xmm7,xmm0             ; xmm0=tmp12
+        punpckhwd xmm6,xmm0
+        movdqa    xmm4,xmm7
+        movdqa    xmm0,xmm6
+        pmaddwd   xmm7,[rel PW_F130_F054]       ; xmm7=data2L
+        pmaddwd   xmm6,[rel PW_F130_F054]       ; xmm6=data2H
+        pmaddwd   xmm4,[rel PW_F054_MF130]      ; xmm4=data6L
+        pmaddwd   xmm0,[rel PW_F054_MF130]      ; xmm0=data6H
+
+        paddd   xmm7,[rel PD_DESCALE_P1]
+        paddd   xmm6,[rel PD_DESCALE_P1]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        paddd   xmm0,[rel PD_DESCALE_P1]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm7,xmm6             ; xmm7=data2
+        packssdw  xmm4,xmm0             ; xmm4=data6
+
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=data2
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=data6
+
+        ; -- Odd part
+
+        movdqa  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp6
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp7
+
+        movdqa  xmm6,xmm2               ; xmm2=tmp4
+        movdqa  xmm0,xmm5               ; xmm5=tmp5
+        paddw   xmm6,xmm3               ; xmm6=z3
+        paddw   xmm0,xmm1               ; xmm0=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm7,xmm6
+        movdqa    xmm4,xmm6
+        punpcklwd xmm7,xmm0
+        punpckhwd xmm4,xmm0
+        movdqa    xmm6,xmm7
+        movdqa    xmm0,xmm4
+        pmaddwd   xmm7,[rel PW_MF078_F117]      ; xmm7=z3L
+        pmaddwd   xmm4,[rel PW_MF078_F117]      ; xmm4=z3H
+        pmaddwd   xmm6,[rel PW_F117_F078]       ; xmm6=z4L
+        pmaddwd   xmm0,[rel PW_F117_F078]       ; xmm0=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm7,xmm2
+        movdqa    xmm4,xmm2
+        punpcklwd xmm7,xmm1
+        punpckhwd xmm4,xmm1
+        movdqa    xmm2,xmm7
+        movdqa    xmm1,xmm4
+        pmaddwd   xmm7,[rel PW_MF060_MF089]     ; xmm7=tmp4L
+        pmaddwd   xmm4,[rel PW_MF060_MF089]     ; xmm4=tmp4H
+        pmaddwd   xmm2,[rel PW_MF089_F060]      ; xmm2=tmp7L
+        pmaddwd   xmm1,[rel PW_MF089_F060]      ; xmm1=tmp7H
+
+        paddd   xmm7, XMMWORD [wk(0)]   ; xmm7=data7L
+        paddd   xmm4, XMMWORD [wk(1)]   ; xmm4=data7H
+        paddd   xmm2,xmm6               ; xmm2=data1L
+        paddd   xmm1,xmm0               ; xmm1=data1H
+
+        paddd   xmm7,[rel PD_DESCALE_P1]
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm2,[rel PD_DESCALE_P1]
+        paddd   xmm1,[rel PD_DESCALE_P1]
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+
+        packssdw  xmm7,xmm4             ; xmm7=data7
+        packssdw  xmm2,xmm1             ; xmm2=data1
+
+        movdqa    xmm4,xmm5
+        movdqa    xmm1,xmm5
+        punpcklwd xmm4,xmm3
+        punpckhwd xmm1,xmm3
+        movdqa    xmm5,xmm4
+        movdqa    xmm3,xmm1
+        pmaddwd   xmm4,[rel PW_MF050_MF256]     ; xmm4=tmp5L
+        pmaddwd   xmm1,[rel PW_MF050_MF256]     ; xmm1=tmp5H
+        pmaddwd   xmm5,[rel PW_MF256_F050]      ; xmm5=tmp6L
+        pmaddwd   xmm3,[rel PW_MF256_F050]      ; xmm3=tmp6H
+
+        paddd   xmm4,xmm6               ; xmm4=data5L
+        paddd   xmm1,xmm0               ; xmm1=data5H
+        paddd   xmm5, XMMWORD [wk(0)]   ; xmm5=data3L
+        paddd   xmm3, XMMWORD [wk(1)]   ; xmm3=data3H
+
+        paddd   xmm4,[rel PD_DESCALE_P1]
+        paddd   xmm1,[rel PD_DESCALE_P1]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+        paddd   xmm5,[rel PD_DESCALE_P1]
+        paddd   xmm3,[rel PD_DESCALE_P1]
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+
+        packssdw  xmm4,xmm1             ; xmm4=data5
+        packssdw  xmm5,xmm3             ; xmm5=data3
+
+        ; ---- Pass 2: process columns.
+
+        movdqa  xmm6, XMMWORD [wk(2)]   ; xmm6=col0
+        movdqa  xmm0, XMMWORD [wk(4)]   ; xmm0=col2
+
+        ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+        ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm1,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm2             ; xmm6=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm1,xmm2             ; xmm1=(40 41 50 51 60 61 70 71)
+        movdqa    xmm3,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm3,xmm5             ; xmm3=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm2, XMMWORD [wk(3)]   ; xmm2=col4
+        movdqa  xmm5, XMMWORD [wk(5)]   ; xmm5=col6
+
+        ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+        ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm4             ; xmm2=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm0,xmm4             ; xmm0=(44 45 54 55 64 65 74 75)
+        movdqa    xmm3,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm7             ; xmm5=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm3,xmm7             ; xmm3=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm4,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm4,xmm5             ; xmm4=(24 25 26 27 34 35 36 37)
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm7,xmm3             ; xmm7=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=(02 03 12 13 22 23 32 33)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(2)], xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(3)], xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm4,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm5             ; xmm6=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm4,xmm5             ; xmm4=(20 21 22 23 30 31 32 33)
+        movdqa    xmm0,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm3             ; xmm1=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm0,xmm3             ; xmm0=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm5,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm2            ; xmm6=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm5,xmm2            ; xmm5=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm7            ; xmm0=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm3,xmm7            ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm7,xmm6
+        psubw   xmm5,xmm0               ; xmm5=data1-data6=tmp6
+        psubw   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        paddw   xmm2,xmm0               ; xmm2=data1+data6=tmp1
+        paddw   xmm7,xmm3               ; xmm7=data0+data7=tmp0
+
+        movdqa  xmm0, XMMWORD [wk(2)]   ; xmm0=(24 25 26 27 34 35 36 37)
+        movdqa  xmm3, XMMWORD [wk(3)]   ; xmm3=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movdqa     xmm5,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm5,xmm0            ; xmm5=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm3            ; xmm1=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm6,xmm3            ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm3,xmm4
+        paddw   xmm5,xmm1               ; xmm5=data3+data4=tmp3
+        paddw   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        psubw   xmm0,xmm1               ; xmm0=data3-data4=tmp4
+        psubw   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm6,xmm2
+        paddw   xmm7,xmm5               ; xmm7=tmp10
+        paddw   xmm2,xmm4               ; xmm2=tmp11
+        psubw   xmm1,xmm5               ; xmm1=tmp13
+        psubw   xmm6,xmm4               ; xmm6=tmp12
+
+        movdqa  xmm5,xmm7
+        paddw   xmm7,xmm2               ; xmm7=tmp10+tmp11
+        psubw   xmm5,xmm2               ; xmm5=tmp10-tmp11
+
+        paddw   xmm7,[rel PW_DESCALE_P2X]
+        paddw   xmm5,[rel PW_DESCALE_P2X]
+        psraw   xmm7,PASS1_BITS         ; xmm7=data0
+        psraw   xmm5,PASS1_BITS         ; xmm5=data4
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_DCTELEM)], xmm7
+        movdqa  XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_DCTELEM)], xmm5
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm4,xmm1             ; xmm1=tmp13
+        movdqa    xmm2,xmm1
+        punpcklwd xmm4,xmm6             ; xmm6=tmp12
+        punpckhwd xmm2,xmm6
+        movdqa    xmm1,xmm4
+        movdqa    xmm6,xmm2
+        pmaddwd   xmm4,[rel PW_F130_F054]       ; xmm4=data2L
+        pmaddwd   xmm2,[rel PW_F130_F054]       ; xmm2=data2H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=data6L
+        pmaddwd   xmm6,[rel PW_F054_MF130]      ; xmm6=data6H
+
+        paddd   xmm4,[rel PD_DESCALE_P2]
+        paddd   xmm2,[rel PD_DESCALE_P2]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        paddd   xmm6,[rel PD_DESCALE_P2]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm6,DESCALE_P2
+
+        packssdw  xmm4,xmm2             ; xmm4=data2
+        packssdw  xmm1,xmm6             ; xmm1=data6
+
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_DCTELEM)], xmm1
+
+        ; -- Odd part
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        movdqa  xmm2,xmm0               ; xmm0=tmp4
+        movdqa  xmm6,xmm3               ; xmm3=tmp5
+        paddw   xmm2,xmm7               ; xmm2=z3
+        paddw   xmm6,xmm5               ; xmm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm1,xmm2
+        punpcklwd xmm4,xmm6
+        punpckhwd xmm1,xmm6
+        movdqa    xmm2,xmm4
+        movdqa    xmm6,xmm1
+        pmaddwd   xmm4,[rel PW_MF078_F117]      ; xmm4=z3L
+        pmaddwd   xmm1,[rel PW_MF078_F117]      ; xmm1=z3H
+        pmaddwd   xmm2,[rel PW_F117_F078]       ; xmm2=z4L
+        pmaddwd   xmm6,[rel PW_F117_F078]       ; xmm6=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm1,xmm0
+        punpcklwd xmm4,xmm5
+        punpckhwd xmm1,xmm5
+        movdqa    xmm0,xmm4
+        movdqa    xmm5,xmm1
+        pmaddwd   xmm4,[rel PW_MF060_MF089]     ; xmm4=tmp4L
+        pmaddwd   xmm1,[rel PW_MF060_MF089]     ; xmm1=tmp4H
+        pmaddwd   xmm0,[rel PW_MF089_F060]      ; xmm0=tmp7L
+        pmaddwd   xmm5,[rel PW_MF089_F060]      ; xmm5=tmp7H
+
+        paddd   xmm4, XMMWORD [wk(0)]   ; xmm4=data7L
+        paddd   xmm1, XMMWORD [wk(1)]   ; xmm1=data7H
+        paddd   xmm0,xmm2               ; xmm0=data1L
+        paddd   xmm5,xmm6               ; xmm5=data1H
+
+        paddd   xmm4,[rel PD_DESCALE_P2]
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm0,[rel PD_DESCALE_P2]
+        paddd   xmm5,[rel PD_DESCALE_P2]
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+
+        packssdw  xmm4,xmm1             ; xmm4=data7
+        packssdw  xmm0,xmm5             ; xmm0=data1
+
+        movdqa  XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_DCTELEM)], xmm0
+
+        movdqa    xmm1,xmm3
+        movdqa    xmm5,xmm3
+        punpcklwd xmm1,xmm7
+        punpckhwd xmm5,xmm7
+        movdqa    xmm3,xmm1
+        movdqa    xmm7,xmm5
+        pmaddwd   xmm1,[rel PW_MF050_MF256]     ; xmm1=tmp5L
+        pmaddwd   xmm5,[rel PW_MF050_MF256]     ; xmm5=tmp5H
+        pmaddwd   xmm3,[rel PW_MF256_F050]      ; xmm3=tmp6L
+        pmaddwd   xmm7,[rel PW_MF256_F050]      ; xmm7=tmp6H
+
+        paddd   xmm1,xmm2               ; xmm1=data5L
+        paddd   xmm5,xmm6               ; xmm5=data5H
+        paddd   xmm3, XMMWORD [wk(0)]   ; xmm3=data3L
+        paddd   xmm7, XMMWORD [wk(1)]   ; xmm7=data3H
+
+        paddd   xmm1,[rel PD_DESCALE_P2]
+        paddd   xmm5,[rel PD_DESCALE_P2]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+        paddd   xmm3,[rel PD_DESCALE_P2]
+        paddd   xmm7,[rel PD_DESCALE_P2]
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm1,xmm5             ; xmm1=data5
+        packssdw  xmm3,xmm7             ; xmm3=data3
+
+        movdqa  XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_DCTELEM)], xmm3
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jfdctint-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jfdctint-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,634 @@
+;
+; jfdctint.asm - accurate integer FDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; forward DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jfdctint.c; see the jfdctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_fdct_islow_sse2) PRIVATE
+
+EXTN(jconst_fdct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PW_DESCALE_P2X  times 8 dw  1 << (PASS1_BITS-1)
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform the forward DCT on one block of samples.
+;
+; GLOBAL(void)
+; jsimd_fdct_islow_sse2 (DCTELEM *data)
+;
+
+%define data(b)         (b)+8           ; DCTELEM *data
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          6
+
+        align   16
+        global  EXTN(jsimd_fdct_islow_sse2) PRIVATE
+
+EXTN(jsimd_fdct_islow_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+;       push    edi             ; unused
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process rows.
+
+        mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm0=(00 01 02 03 04 05 06 07), xmm2=(20 21 22 23 24 25 26 27)
+        ; xmm1=(10 11 12 13 14 15 16 17), xmm3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm1             ; xmm0=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm4,xmm1             ; xmm4=(04 14 05 15 06 16 07 17)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm3             ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm5,xmm3             ; xmm5=(24 34 25 35 26 36 27 37)
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)]
+
+        ; xmm6=( 4 12 20 28 36 44 52 60), xmm1=( 6 14 22 30 38 46 54 62)
+        ; xmm7=( 5 13 21 29 37 45 53 61), xmm3=( 7 15 23 31 39 47 55 63)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=(20 30 21 31 22 32 23 33)
+        movdqa  XMMWORD [wk(1)], xmm5   ; wk(1)=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm2,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm7             ; xmm2=(44 54 45 55 46 56 47 57)
+        movdqa    xmm5,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm3             ; xmm1=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm5,xmm3             ; xmm5=(64 74 65 75 66 76 67 77)
+
+        movdqa    xmm7,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm1             ; xmm6=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm7,xmm1             ; xmm7=(42 52 62 72 43 53 63 73)
+        movdqa    xmm3,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm3,xmm5             ; xmm3=(46 56 66 76 47 57 67 77)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(20 30 21 31 22 32 23 33)
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=(24 34 25 35 26 36 27 37)
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=(42 52 62 72 43 53 63 73)
+        movdqa  XMMWORD [wk(3)], xmm2   ; wk(3)=(44 54 64 74 45 55 65 75)
+
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm1             ; xmm0=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm7,xmm1             ; xmm7=(02 12 22 32 03 13 23 33)
+        movdqa    xmm2,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm5             ; xmm4=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm2,xmm5             ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa     xmm1,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=(00 10 20 30 40 50 60 70)=data0
+        punpckhqdq xmm1,xmm6            ; xmm1=(01 11 21 31 41 51 61 71)=data1
+        movdqa     xmm5,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm3            ; xmm2=(06 16 26 36 46 56 66 76)=data6
+        punpckhqdq xmm5,xmm3            ; xmm5=(07 17 27 37 47 57 67 77)=data7
+
+        movdqa  xmm6,xmm1
+        movdqa  xmm3,xmm0
+        psubw   xmm1,xmm2               ; xmm1=data1-data6=tmp6
+        psubw   xmm0,xmm5               ; xmm0=data0-data7=tmp7
+        paddw   xmm6,xmm2               ; xmm6=data1+data6=tmp1
+        paddw   xmm3,xmm5               ; xmm3=data0+data7=tmp0
+
+        movdqa  xmm2, XMMWORD [wk(2)]   ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa  xmm5, XMMWORD [wk(3)]   ; xmm5=(44 54 64 74 45 55 65 75)
+        movdqa  XMMWORD [wk(0)], xmm1   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp7
+
+        movdqa     xmm1,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm2            ; xmm7=(02 12 22 32 42 52 62 72)=data2
+        punpckhqdq xmm1,xmm2            ; xmm1=(03 13 23 33 43 53 63 73)=data3
+        movdqa     xmm0,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm5            ; xmm4=(04 14 24 34 44 54 64 74)=data4
+        punpckhqdq xmm0,xmm5            ; xmm0=(05 15 25 35 45 55 65 75)=data5
+
+        movdqa  xmm2,xmm1
+        movdqa  xmm5,xmm7
+        paddw   xmm1,xmm4               ; xmm1=data3+data4=tmp3
+        paddw   xmm7,xmm0               ; xmm7=data2+data5=tmp2
+        psubw   xmm2,xmm4               ; xmm2=data3-data4=tmp4
+        psubw   xmm5,xmm0               ; xmm5=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm0,xmm6
+        paddw   xmm3,xmm1               ; xmm3=tmp10
+        paddw   xmm6,xmm7               ; xmm6=tmp11
+        psubw   xmm4,xmm1               ; xmm4=tmp13
+        psubw   xmm0,xmm7               ; xmm0=tmp12
+
+        movdqa  xmm1,xmm3
+        paddw   xmm3,xmm6               ; xmm3=tmp10+tmp11
+        psubw   xmm1,xmm6               ; xmm1=tmp10-tmp11
+
+        psllw   xmm3,PASS1_BITS         ; xmm3=data0
+        psllw   xmm1,PASS1_BITS         ; xmm1=data4
+
+        movdqa  XMMWORD [wk(2)], xmm3   ; wk(2)=data0
+        movdqa  XMMWORD [wk(3)], xmm1   ; wk(3)=data4
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm7,xmm4             ; xmm4=tmp13
+        movdqa    xmm6,xmm4
+        punpcklwd xmm7,xmm0             ; xmm0=tmp12
+        punpckhwd xmm6,xmm0
+        movdqa    xmm4,xmm7
+        movdqa    xmm0,xmm6
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F130_F054)]       ; xmm7=data2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F130_F054)]       ; xmm6=data2H
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm4=data6L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm0=data6H
+
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm6,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm0,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm7,xmm6             ; xmm7=data2
+        packssdw  xmm4,xmm0             ; xmm4=data6
+
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=data2
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=data6
+
+        ; -- Odd part
+
+        movdqa  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp6
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp7
+
+        movdqa  xmm6,xmm2               ; xmm2=tmp4
+        movdqa  xmm0,xmm5               ; xmm5=tmp5
+        paddw   xmm6,xmm3               ; xmm6=z3
+        paddw   xmm0,xmm1               ; xmm0=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm7,xmm6
+        movdqa    xmm4,xmm6
+        punpcklwd xmm7,xmm0
+        punpckhwd xmm4,xmm0
+        movdqa    xmm6,xmm7
+        movdqa    xmm0,xmm4
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm7=z3L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm4=z3H
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F117_F078)]       ; xmm6=z4L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F117_F078)]       ; xmm0=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm7,xmm2
+        movdqa    xmm4,xmm2
+        punpcklwd xmm7,xmm1
+        punpckhwd xmm4,xmm1
+        movdqa    xmm2,xmm7
+        movdqa    xmm1,xmm4
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm7=tmp4L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm4=tmp4H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm2=tmp7L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm1=tmp7H
+
+        paddd   xmm7, XMMWORD [wk(0)]   ; xmm7=data7L
+        paddd   xmm4, XMMWORD [wk(1)]   ; xmm4=data7H
+        paddd   xmm2,xmm6               ; xmm2=data1L
+        paddd   xmm1,xmm0               ; xmm1=data1H
+
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm2,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+
+        packssdw  xmm7,xmm4             ; xmm7=data7
+        packssdw  xmm2,xmm1             ; xmm2=data1
+
+        movdqa    xmm4,xmm5
+        movdqa    xmm1,xmm5
+        punpcklwd xmm4,xmm3
+        punpckhwd xmm1,xmm3
+        movdqa    xmm5,xmm4
+        movdqa    xmm3,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm4=tmp5L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm1=tmp5H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm5=tmp6L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm3=tmp6H
+
+        paddd   xmm4,xmm6               ; xmm4=data5L
+        paddd   xmm1,xmm0               ; xmm1=data5H
+        paddd   xmm5, XMMWORD [wk(0)]   ; xmm5=data3L
+        paddd   xmm3, XMMWORD [wk(1)]   ; xmm3=data3H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm1,DESCALE_P1
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P1)]
+        paddd   xmm3,[GOTOFF(ebx,PD_DESCALE_P1)]
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+
+        packssdw  xmm4,xmm1             ; xmm4=data5
+        packssdw  xmm5,xmm3             ; xmm5=data3
+
+        ; ---- Pass 2: process columns.
+
+;       mov     edx, POINTER [data(eax)]        ; (DCTELEM *)
+
+        movdqa  xmm6, XMMWORD [wk(2)]   ; xmm6=col0
+        movdqa  xmm0, XMMWORD [wk(4)]   ; xmm0=col2
+
+        ; xmm6=(00 10 20 30 40 50 60 70), xmm0=(02 12 22 32 42 52 62 72)
+        ; xmm2=(01 11 21 31 41 51 61 71), xmm5=(03 13 23 33 43 53 63 73)
+
+        movdqa    xmm1,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm2             ; xmm6=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm1,xmm2             ; xmm1=(40 41 50 51 60 61 70 71)
+        movdqa    xmm3,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(02 03 12 13 22 23 32 33)
+        punpckhwd xmm3,xmm5             ; xmm3=(42 43 52 53 62 63 72 73)
+
+        movdqa  xmm2, XMMWORD [wk(3)]   ; xmm2=col4
+        movdqa  xmm5, XMMWORD [wk(5)]   ; xmm5=col6
+
+        ; xmm2=(04 14 24 34 44 54 64 74), xmm5=(06 16 26 36 46 56 66 76)
+        ; xmm4=(05 15 25 35 45 55 65 75), xmm7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=(02 03 12 13 22 23 32 33)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(42 43 52 53 62 63 72 73)
+
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm4             ; xmm2=(04 05 14 15 24 25 34 35)
+        punpckhwd xmm0,xmm4             ; xmm0=(44 45 54 55 64 65 74 75)
+        movdqa    xmm3,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm7             ; xmm5=(06 07 16 17 26 27 36 37)
+        punpckhwd xmm3,xmm7             ; xmm3=(46 47 56 57 66 67 76 77)
+
+        movdqa    xmm4,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm5             ; xmm2=(04 05 06 07 14 15 16 17)
+        punpckhdq xmm4,xmm5             ; xmm4=(24 25 26 27 34 35 36 37)
+        movdqa    xmm7,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(44 45 46 47 54 55 56 57)
+        punpckhdq xmm7,xmm3             ; xmm7=(64 65 66 67 74 75 76 77)
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=(02 03 12 13 22 23 32 33)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(42 43 52 53 62 63 72 73)
+        movdqa  XMMWORD [wk(2)], xmm4   ; wk(2)=(24 25 26 27 34 35 36 37)
+        movdqa  XMMWORD [wk(3)], xmm0   ; wk(3)=(44 45 46 47 54 55 56 57)
+
+        movdqa    xmm4,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm5             ; xmm6=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm4,xmm5             ; xmm4=(20 21 22 23 30 31 32 33)
+        movdqa    xmm0,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm3             ; xmm1=(40 41 42 43 50 51 52 53)
+        punpckhdq xmm0,xmm3             ; xmm0=(60 61 62 63 70 71 72 73)
+
+        movdqa     xmm5,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm2            ; xmm6=(00 01 02 03 04 05 06 07)=data0
+        punpckhqdq xmm5,xmm2            ; xmm5=(10 11 12 13 14 15 16 17)=data1
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm7            ; xmm0=(60 61 62 63 64 65 66 67)=data6
+        punpckhqdq xmm3,xmm7            ; xmm3=(70 71 72 73 74 75 76 77)=data7
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm7,xmm6
+        psubw   xmm5,xmm0               ; xmm5=data1-data6=tmp6
+        psubw   xmm6,xmm3               ; xmm6=data0-data7=tmp7
+        paddw   xmm2,xmm0               ; xmm2=data1+data6=tmp1
+        paddw   xmm7,xmm3               ; xmm7=data0+data7=tmp0
+
+        movdqa  xmm0, XMMWORD [wk(2)]   ; xmm0=(24 25 26 27 34 35 36 37)
+        movdqa  xmm3, XMMWORD [wk(3)]   ; xmm3=(44 45 46 47 54 55 56 57)
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=tmp6
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp7
+
+        movdqa     xmm5,xmm4            ; transpose coefficients(phase 3)
+        punpcklqdq xmm4,xmm0            ; xmm4=(20 21 22 23 24 25 26 27)=data2
+        punpckhqdq xmm5,xmm0            ; xmm5=(30 31 32 33 34 35 36 37)=data3
+        movdqa     xmm6,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm3            ; xmm1=(40 41 42 43 44 45 46 47)=data4
+        punpckhqdq xmm6,xmm3            ; xmm6=(50 51 52 53 54 55 56 57)=data5
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm3,xmm4
+        paddw   xmm5,xmm1               ; xmm5=data3+data4=tmp3
+        paddw   xmm4,xmm6               ; xmm4=data2+data5=tmp2
+        psubw   xmm0,xmm1               ; xmm0=data3-data4=tmp4
+        psubw   xmm3,xmm6               ; xmm3=data2-data5=tmp5
+
+        ; -- Even part
+
+        movdqa  xmm1,xmm7
+        movdqa  xmm6,xmm2
+        paddw   xmm7,xmm5               ; xmm7=tmp10
+        paddw   xmm2,xmm4               ; xmm2=tmp11
+        psubw   xmm1,xmm5               ; xmm1=tmp13
+        psubw   xmm6,xmm4               ; xmm6=tmp12
+
+        movdqa  xmm5,xmm7
+        paddw   xmm7,xmm2               ; xmm7=tmp10+tmp11
+        psubw   xmm5,xmm2               ; xmm5=tmp10-tmp11
+
+        paddw   xmm7,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        paddw   xmm5,[GOTOFF(ebx,PW_DESCALE_P2X)]
+        psraw   xmm7,PASS1_BITS         ; xmm7=data0
+        psraw   xmm5,PASS1_BITS         ; xmm5=data4
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], xmm7
+        movdqa  XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], xmm5
+
+        ; (Original)
+        ; z1 = (tmp12 + tmp13) * 0.541196100;
+        ; data2 = z1 + tmp13 * 0.765366865;
+        ; data6 = z1 + tmp12 * -1.847759065;
+        ;
+        ; (This implementation)
+        ; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
+        ; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
+
+        movdqa    xmm4,xmm1             ; xmm1=tmp13
+        movdqa    xmm2,xmm1
+        punpcklwd xmm4,xmm6             ; xmm6=tmp12
+        punpckhwd xmm2,xmm6
+        movdqa    xmm1,xmm4
+        movdqa    xmm6,xmm2
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F130_F054)]       ; xmm4=data2L
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F130_F054)]       ; xmm2=data2H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=data6L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm6=data6H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm2,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm6,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm6,DESCALE_P2
+
+        packssdw  xmm4,xmm2             ; xmm4=data2
+        packssdw  xmm1,xmm6             ; xmm1=data6
+
+        movdqa  XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], xmm1
+
+        ; -- Odd part
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp6
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp7
+
+        movdqa  xmm2,xmm0               ; xmm0=tmp4
+        movdqa  xmm6,xmm3               ; xmm3=tmp5
+        paddw   xmm2,xmm7               ; xmm2=z3
+        paddw   xmm6,xmm5               ; xmm6=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm4,xmm2
+        movdqa    xmm1,xmm2
+        punpcklwd xmm4,xmm6
+        punpckhwd xmm1,xmm6
+        movdqa    xmm2,xmm4
+        movdqa    xmm6,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm4=z3L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm1=z3H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F117_F078)]       ; xmm2=z4L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F117_F078)]       ; xmm6=z4H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=z3L
+        movdqa  XMMWORD [wk(1)], xmm1   ; wk(1)=z3H
+
+        ; (Original)
+        ; z1 = tmp4 + tmp7;  z2 = tmp5 + tmp6;
+        ; tmp4 = tmp4 * 0.298631336;  tmp5 = tmp5 * 2.053119869;
+        ; tmp6 = tmp6 * 3.072711026;  tmp7 = tmp7 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; data7 = tmp4 + z1 + z3;  data5 = tmp5 + z2 + z4;
+        ; data3 = tmp6 + z2 + z3;  data1 = tmp7 + z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
+        ; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
+        ; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
+        ; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
+        ; data7 = tmp4 + z3;  data5 = tmp5 + z4;
+        ; data3 = tmp6 + z3;  data1 = tmp7 + z4;
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm1,xmm0
+        punpcklwd xmm4,xmm5
+        punpckhwd xmm1,xmm5
+        movdqa    xmm0,xmm4
+        movdqa    xmm5,xmm1
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm4=tmp4L
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm1=tmp4H
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm0=tmp7L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm5=tmp7H
+
+        paddd   xmm4, XMMWORD [wk(0)]   ; xmm4=data7L
+        paddd   xmm1, XMMWORD [wk(1)]   ; xmm1=data7H
+        paddd   xmm0,xmm2               ; xmm0=data1L
+        paddd   xmm5,xmm6               ; xmm5=data1H
+
+        paddd   xmm4,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm0,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+
+        packssdw  xmm4,xmm1             ; xmm4=data7
+        packssdw  xmm0,xmm5             ; xmm0=data1
+
+        movdqa  XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_DCTELEM)], xmm4
+        movdqa  XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_DCTELEM)], xmm0
+
+        movdqa    xmm1,xmm3
+        movdqa    xmm5,xmm3
+        punpcklwd xmm1,xmm7
+        punpckhwd xmm5,xmm7
+        movdqa    xmm3,xmm1
+        movdqa    xmm7,xmm5
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm1=tmp5L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm5=tmp5H
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm3=tmp6L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm7=tmp6H
+
+        paddd   xmm1,xmm2               ; xmm1=data5L
+        paddd   xmm5,xmm6               ; xmm5=data5H
+        paddd   xmm3, XMMWORD [wk(0)]   ; xmm3=data3L
+        paddd   xmm7, XMMWORD [wk(1)]   ; xmm7=data3H
+
+        paddd   xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm1,DESCALE_P2
+        psrad   xmm5,DESCALE_P2
+        paddd   xmm3,[GOTOFF(ebx,PD_DESCALE_P2)]
+        paddd   xmm7,[GOTOFF(ebx,PD_DESCALE_P2)]
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm1,xmm5             ; xmm1=data5
+        packssdw  xmm3,xmm7             ; xmm3=data3
+
+        movdqa  XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_DCTELEM)], xmm3
+
+;       pop     edi             ; unused
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-3dn.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,452 @@
+;
+; jidctflt.asm - floating-point IDCT (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_3dnow) PRIVATE
+
+EXTN(jconst_idct_float_3dnow):
+
+PD_1_414        times 2 dd  1.414213562373095048801689
+PD_1_847        times 2 dd  1.847759065022573512256366
+PD_1_082        times 2 dd  1.082392200292393968799446
+PD_2_613        times 2 dd  2.613125929752753055713286
+PD_RNDINT_MAGIC times 2 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_3dnow (void *dct_table, JCOEFPTR coef_block,
+;                         JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_3dnow) PRIVATE
+
+EXTN(jsimd_idct_float_3dnow):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/2                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_3DNOW
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        pushpic ebx             ; save GOT address
+        mov     ebx, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        mov     eax, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        or      ebx, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        or      ebx, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        or      eax,ebx
+        poppic  ebx             ; restore GOT address
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movd      mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm0,mm0
+        psrad     mm0,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm0,mm0
+
+        pfmul     mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0
+        punpckhdq mm1,mm1
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movd      mm0, DWORD [DWBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movd      mm1, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movd      mm2, DWORD [DWBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movd      mm3, DWORD [DWBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm0,mm0
+        punpcklwd mm1,mm1
+        psrad     mm0,(DWORD_BIT-WORD_BIT)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm0,mm0
+        pi2fd     mm1,mm1
+
+        pfmul     mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        punpcklwd mm2,mm2
+        punpcklwd mm3,mm3
+        psrad     mm2,(DWORD_BIT-WORD_BIT)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm2,mm2
+        pi2fd     mm3,mm3
+
+        pfmul     mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pfsub   mm0,mm2                 ; mm0=tmp11
+        pfsub   mm1,mm3
+        pfadd   mm4,mm2                 ; mm4=tmp10
+        pfadd   mm5,mm3                 ; mm5=tmp13
+
+        pfmul   mm1,[GOTOFF(ebx,PD_1_414)]
+        pfsub   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm5                 ; mm4=tmp3
+        pfsub   mm0,mm1                 ; mm0=tmp2
+        pfadd   mm6,mm5                 ; mm6=tmp0
+        pfadd   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; tmp3
+        movq    MMWORD [wk(0)], mm0     ; tmp2
+
+        ; -- Odd part
+
+        movd      mm2, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movd      mm3, DWORD [DWBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movd      mm5, DWORD [DWBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movd      mm1, DWORD [DWBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd mm2,mm2
+        punpcklwd mm3,mm3
+        psrad     mm2,(DWORD_BIT-WORD_BIT)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm2,mm2
+        pi2fd     mm3,mm3
+
+        pfmul     mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        punpcklwd mm5,mm5
+        punpcklwd mm1,mm1
+        psrad     mm5,(DWORD_BIT-WORD_BIT)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)
+        pi2fd     mm5,mm5
+        pi2fd     mm1,mm1
+
+        pfmul     mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        pfmul     mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        pfadd   mm2,mm1                 ; mm2=z11
+        pfadd   mm5,mm3                 ; mm5=z13
+        pfsub   mm4,mm1                 ; mm4=z12
+        pfsub   mm0,mm3                 ; mm0=z10
+
+        movq    mm1,mm2
+        pfsub   mm2,mm5
+        pfadd   mm1,mm5                 ; mm1=tmp7
+
+        pfmul   mm2,[GOTOFF(ebx,PD_1_414)]      ; mm2=tmp11
+
+        movq    mm3,mm0
+        pfadd   mm0,mm4
+        pfmul   mm0,[GOTOFF(ebx,PD_1_847)]      ; mm0=z5
+        pfmul   mm3,[GOTOFF(ebx,PD_2_613)]      ; mm3=(z10 * 2.613125930)
+        pfmul   mm4,[GOTOFF(ebx,PD_1_082)]      ; mm4=(z12 * 1.082392200)
+        pfsubr  mm3,mm0                 ; mm3=tmp12
+        pfsub   mm4,mm0                 ; mm4=tmp10
+
+        ; -- Final output stage
+
+        pfsub   mm3,mm1                 ; mm3=tmp6
+        movq    mm5,mm6
+        movq    mm0,mm7
+        pfadd   mm6,mm1                 ; mm6=data0=(00 01)
+        pfadd   mm7,mm3                 ; mm7=data1=(10 11)
+        pfsub   mm5,mm1                 ; mm5=data7=(70 71)
+        pfsub   mm0,mm3                 ; mm0=data6=(60 61)
+        pfsub   mm2,mm3                 ; mm2=tmp5
+
+        movq      mm1,mm6               ; transpose coefficients
+        punpckldq mm6,mm7               ; mm6=(00 10)
+        punpckhdq mm1,mm7               ; mm1=(01 11)
+        movq      mm3,mm0               ; transpose coefficients
+        punpckldq mm0,mm5               ; mm0=(60 70)
+        punpckhdq mm3,mm5               ; mm3=(61 71)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm1
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm3
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp2
+        movq    mm5, MMWORD [wk(1)]     ; mm5=tmp3
+
+        pfadd   mm4,mm2                 ; mm4=tmp4
+        movq    mm6,mm7
+        movq    mm1,mm5
+        pfadd   mm7,mm2                 ; mm7=data2=(20 21)
+        pfadd   mm5,mm4                 ; mm5=data4=(40 41)
+        pfsub   mm6,mm2                 ; mm6=data5=(50 51)
+        pfsub   mm1,mm4                 ; mm1=data3=(30 31)
+
+        movq      mm0,mm7               ; transpose coefficients
+        punpckldq mm7,mm1               ; mm7=(20 30)
+        punpckhdq mm0,mm1               ; mm0=(21 31)
+        movq      mm3,mm5               ; transpose coefficients
+        punpckldq mm5,mm6               ; mm5=(40 50)
+        punpckhdq mm3,mm6               ; mm3=(41 51)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm7
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm0
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm3
+
+.nextcolumn:
+        add     esi, byte 2*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 2*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetch [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/2                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pfsub   mm0,mm2                 ; mm0=tmp11
+        pfsub   mm1,mm3
+        pfadd   mm4,mm2                 ; mm4=tmp10
+        pfadd   mm5,mm3                 ; mm5=tmp13
+
+        pfmul   mm1,[GOTOFF(ebx,PD_1_414)]
+        pfsub   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        pfsub   mm4,mm5                 ; mm4=tmp3
+        pfsub   mm0,mm1                 ; mm0=tmp2
+        pfadd   mm6,mm5                 ; mm6=tmp0
+        pfadd   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; tmp3
+        movq    MMWORD [wk(0)], mm0     ; tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        pfadd   mm2,mm1                 ; mm2=z11
+        pfadd   mm5,mm3                 ; mm5=z13
+        pfsub   mm4,mm1                 ; mm4=z12
+        pfsub   mm0,mm3                 ; mm0=z10
+
+        movq    mm1,mm2
+        pfsub   mm2,mm5
+        pfadd   mm1,mm5                 ; mm1=tmp7
+
+        pfmul   mm2,[GOTOFF(ebx,PD_1_414)]      ; mm2=tmp11
+
+        movq    mm3,mm0
+        pfadd   mm0,mm4
+        pfmul   mm0,[GOTOFF(ebx,PD_1_847)]      ; mm0=z5
+        pfmul   mm3,[GOTOFF(ebx,PD_2_613)]      ; mm3=(z10 * 2.613125930)
+        pfmul   mm4,[GOTOFF(ebx,PD_1_082)]      ; mm4=(z12 * 1.082392200)
+        pfsubr  mm3,mm0                 ; mm3=tmp12
+        pfsub   mm4,mm0                 ; mm4=tmp10
+
+        ; -- Final output stage
+
+        pfsub   mm3,mm1                 ; mm3=tmp6
+        movq    mm5,mm6
+        movq    mm0,mm7
+        pfadd   mm6,mm1                 ; mm6=data0=(00 10)
+        pfadd   mm7,mm3                 ; mm7=data1=(01 11)
+        pfsub   mm5,mm1                 ; mm5=data7=(07 17)
+        pfsub   mm0,mm3                 ; mm0=data6=(06 16)
+        pfsub   mm2,mm3                 ; mm2=tmp5
+
+        movq    mm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)]       ; mm1=[PD_RNDINT_MAGIC]
+        pcmpeqd mm3,mm3
+        psrld   mm3,WORD_BIT            ; mm3={0xFFFF 0x0000 0xFFFF 0x0000}
+
+        pfadd   mm6,mm1                 ; mm6=roundint(data0/8)=(00 ** 10 **)
+        pfadd   mm7,mm1                 ; mm7=roundint(data1/8)=(01 ** 11 **)
+        pfadd   mm0,mm1                 ; mm0=roundint(data6/8)=(06 ** 16 **)
+        pfadd   mm5,mm1                 ; mm5=roundint(data7/8)=(07 ** 17 **)
+
+        pand    mm6,mm3                 ; mm6=(00 -- 10 --)
+        pslld   mm7,WORD_BIT            ; mm7=(-- 01 -- 11)
+        pand    mm0,mm3                 ; mm0=(06 -- 16 --)
+        pslld   mm5,WORD_BIT            ; mm5=(-- 07 -- 17)
+        por     mm6,mm7                 ; mm6=(00 01 10 11)
+        por     mm0,mm5                 ; mm0=(06 07 16 17)
+
+        movq    mm1, MMWORD [wk(0)]     ; mm1=tmp2
+        movq    mm3, MMWORD [wk(1)]     ; mm3=tmp3
+
+        pfadd   mm4,mm2                 ; mm4=tmp4
+        movq    mm7,mm1
+        movq    mm5,mm3
+        pfadd   mm1,mm2                 ; mm1=data2=(02 12)
+        pfadd   mm3,mm4                 ; mm3=data4=(04 14)
+        pfsub   mm7,mm2                 ; mm7=data5=(05 15)
+        pfsub   mm5,mm4                 ; mm5=data3=(03 13)
+
+        movq    mm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)]       ; mm2=[PD_RNDINT_MAGIC]
+        pcmpeqd mm4,mm4
+        psrld   mm4,WORD_BIT            ; mm4={0xFFFF 0x0000 0xFFFF 0x0000}
+
+        pfadd   mm3,mm2                 ; mm3=roundint(data4/8)=(04 ** 14 **)
+        pfadd   mm7,mm2                 ; mm7=roundint(data5/8)=(05 ** 15 **)
+        pfadd   mm1,mm2                 ; mm1=roundint(data2/8)=(02 ** 12 **)
+        pfadd   mm5,mm2                 ; mm5=roundint(data3/8)=(03 ** 13 **)
+
+        pand    mm3,mm4                 ; mm3=(04 -- 14 --)
+        pslld   mm7,WORD_BIT            ; mm7=(-- 05 -- 15)
+        pand    mm1,mm4                 ; mm1=(02 -- 12 --)
+        pslld   mm5,WORD_BIT            ; mm5=(-- 03 -- 13)
+        por     mm3,mm7                 ; mm3=(04 05 14 15)
+        por     mm1,mm5                 ; mm1=(02 03 12 13)
+
+        movq      mm2,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm2=[PB_CENTERJSAMP]
+
+        packsswb  mm6,mm3               ; mm6=(00 01 10 11 04 05 14 15)
+        packsswb  mm1,mm0               ; mm1=(02 03 12 13 06 07 16 17)
+        paddb     mm6,mm2
+        paddb     mm1,mm2
+
+        movq      mm4,mm6               ; transpose coefficients(phase 2)
+        punpcklwd mm6,mm1               ; mm6=(00 01 02 03 10 11 12 13)
+        punpckhwd mm4,mm1               ; mm4=(04 05 06 07 14 15 16 17)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 3)
+        punpckldq mm6,mm4               ; mm6=(00 01 02 03 04 05 06 07)
+        punpckhdq mm7,mm4               ; mm7=(10 11 12 13 14 15 16 17)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 2*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 2*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,483 @@
+;
+; jidctflt.asm - floating-point IDCT (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse2) PRIVATE
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse2) PRIVATE
+
+EXTN(jsimd_idct_float_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [workspace]
+        collect_args
+        push    rbx
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+        lea     rdi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     rcx, DCTSIZE/4                          ; ctr
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    xmm1, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm2, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm4, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm6, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        movq    xmm7, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm2
+        por     xmm3,xmm4
+        por     xmm5,xmm6
+        por     xmm1,xmm3
+        por     xmm5,xmm7
+        por     xmm1,xmm5
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm2, XMM_MMWORD [MMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpcklwd xmm1,xmm1             ; xmm1=(20 20 21 21 22 22 23 23)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in2=(20 21 22 23)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in2=(20 21 22 23)
+
+        punpcklwd xmm2,xmm2             ; xmm2=(40 40 41 41 42 42 43 43)
+        punpcklwd xmm3,xmm3             ; xmm3=(60 60 61 61 62 62 63 63)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in4=(40 41 42 43)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in6=(60 61 62 63)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in4=(40 41 42 43)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in6=(60 61 62 63)
+
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[rel PD_1_414]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      xmm2, XMM_MMWORD [MMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm5, XMM_MMWORD [MMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm2,xmm2             ; xmm2=(10 10 11 11 12 12 13 13)
+        punpcklwd xmm3,xmm3             ; xmm3=(30 30 31 31 32 32 33 33)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in1=(10 11 12 13)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in3=(30 31 32 33)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in1=(10 11 12 13)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in3=(30 31 32 33)
+
+        punpcklwd xmm5,xmm5             ; xmm5=(50 50 51 51 52 52 53 53)
+        punpcklwd xmm1,xmm1             ; xmm1=(70 70 71 71 72 72 73 73)
+        psrad     xmm5,(DWORD_BIT-WORD_BIT)     ; xmm5=in5=(50 51 52 53)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in7=(70 71 72 73)
+        cvtdq2ps  xmm5,xmm5                     ; xmm5=in5=(50 51 52 53)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in7=(70 71 72 73)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[rel PD_1_414]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[rel PD_1_847]     ; xmm0=z5
+        mulps   xmm3,[rel PD_M2_613]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[rel PD_1_082]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,rdi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,rdi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     rsi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     rdx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     rdi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     rcx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        lea     rsi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+        mov     rcx, DCTSIZE/4                          ; ctr
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[rel PD_1_414]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[rel PD_1_414]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[rel PD_1_847]     ; xmm0=z5
+        mulps   xmm3,[rel PD_M2_613]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[rel PD_1_082]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[rel PD_RNDINT_MAGIC]      ; xmm1=[rel PD_RNDINT_MAGIC]
+        pcmpeqd xmm3,xmm3
+        psrld   xmm3,WORD_BIT           ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm6,xmm1       ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+        addps   xmm7,xmm1       ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+        addps   xmm0,xmm1       ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+        addps   xmm5,xmm1       ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+        pand    xmm6,xmm3               ; xmm6=(00 -- 10 -- 20 -- 30 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+        pand    xmm0,xmm3               ; xmm0=(06 -- 16 -- 26 -- 36 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+        por     xmm6,xmm7               ; xmm6=(00 01 10 11 20 21 30 31)
+        por     xmm0,xmm5               ; xmm0=(06 07 16 17 26 27 36 37)
+
+        movaps  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp2
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=tmp3
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm7,xmm1
+        movaps  xmm5,xmm3
+        addps   xmm1,xmm2               ; xmm1=data2=(02 12 22 32)
+        addps   xmm3,xmm4               ; xmm3=data4=(04 14 24 34)
+        subps   xmm7,xmm2               ; xmm7=data5=(05 15 25 35)
+        subps   xmm5,xmm4               ; xmm5=data3=(03 13 23 33)
+
+        movaps  xmm2,[rel PD_RNDINT_MAGIC]      ; xmm2=[rel PD_RNDINT_MAGIC]
+        pcmpeqd xmm4,xmm4
+        psrld   xmm4,WORD_BIT           ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm3,xmm2       ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+        addps   xmm7,xmm2       ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+        addps   xmm1,xmm2       ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+        addps   xmm5,xmm2       ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+        pand    xmm3,xmm4               ; xmm3=(04 -- 14 -- 24 -- 34 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+        pand    xmm1,xmm4               ; xmm1=(02 -- 12 -- 22 -- 32 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+        por     xmm3,xmm7               ; xmm3=(04 05 14 15 24 25 34 35)
+        por     xmm1,xmm5               ; xmm1=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm2,[rel PB_CENTERJSAMP]     ; xmm2=[rel PB_CENTERJSAMP]
+
+        packsswb  xmm6,xmm3     ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+        packsswb  xmm1,xmm0     ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+        paddb     xmm6,xmm2
+        paddb     xmm1,xmm2
+
+        movdqa    xmm4,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm1     ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm1     ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+        movdqa    xmm7,xmm6     ; transpose coefficients(phase 3)
+        punpckldq xmm6,xmm4     ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm7,xmm4     ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+        pshufd  xmm5,xmm6,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm3,xmm7,0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rbx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm7
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rbx, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE], xmm3
+
+        add     rsi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     rdi, byte 4*SIZEOF_JSAMPROW
+        dec     rcx                             ; ctr
+        jnz     near .rowloop
+
+        pop     rbx
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,498 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse2) PRIVATE
+
+EXTN(jconst_idct_float_sse2):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_RNDINT_MAGIC times 4 dd  100663296.0 ; (float)(0x00C00000 << 3)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse2) PRIVATE
+
+EXTN(jsimd_idct_float_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    xmm1, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    xmm2, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    xmm4, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    xmm6, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        movq    xmm7, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm2
+        por     xmm3,xmm4
+        por     xmm5,xmm6
+        por     xmm1,xmm3
+        por     xmm5,xmm7
+        por     xmm1,xmm5
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      xmm0, XMM_MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq      xmm2, XMM_MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpcklwd xmm1,xmm1             ; xmm1=(20 20 21 21 22 22 23 23)
+        psrad     xmm0,(DWORD_BIT-WORD_BIT)     ; xmm0=in0=(00 01 02 03)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in2=(20 21 22 23)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=in0=(00 01 02 03)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in2=(20 21 22 23)
+
+        punpcklwd xmm2,xmm2             ; xmm2=(40 40 41 41 42 42 43 43)
+        punpcklwd xmm3,xmm3             ; xmm3=(60 60 61 61 62 62 63 63)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in4=(40 41 42 43)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in6=(60 61 62 63)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in4=(40 41 42 43)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in6=(60 61 62 63)
+
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      xmm2, XMM_MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq      xmm3, XMM_MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq      xmm5, XMM_MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq      xmm1, XMM_MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpcklwd xmm2,xmm2             ; xmm2=(10 10 11 11 12 12 13 13)
+        punpcklwd xmm3,xmm3             ; xmm3=(30 30 31 31 32 32 33 33)
+        psrad     xmm2,(DWORD_BIT-WORD_BIT)     ; xmm2=in1=(10 11 12 13)
+        psrad     xmm3,(DWORD_BIT-WORD_BIT)     ; xmm3=in3=(30 31 32 33)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=in1=(10 11 12 13)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=in3=(30 31 32 33)
+
+        punpcklwd xmm5,xmm5             ; xmm5=(50 50 51 51 52 52 53 53)
+        punpcklwd xmm1,xmm1             ; xmm1=(70 70 71 71 72 72 73 73)
+        psrad     xmm5,(DWORD_BIT-WORD_BIT)     ; xmm5=in5=(50 51 52 53)
+        psrad     xmm1,(DWORD_BIT-WORD_BIT)     ; xmm1=in7=(70 71 72 73)
+        cvtdq2ps  xmm5,xmm5                     ; xmm5=in5=(50 51 52 53)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=in7=(70 71 72 73)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[GOTOFF(ebx,PD_RNDINT_MAGIC)]      ; xmm1=[PD_RNDINT_MAGIC]
+        pcmpeqd xmm3,xmm3
+        psrld   xmm3,WORD_BIT           ; xmm3={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm6,xmm1       ; xmm6=roundint(data0/8)=(00 ** 10 ** 20 ** 30 **)
+        addps   xmm7,xmm1       ; xmm7=roundint(data1/8)=(01 ** 11 ** 21 ** 31 **)
+        addps   xmm0,xmm1       ; xmm0=roundint(data6/8)=(06 ** 16 ** 26 ** 36 **)
+        addps   xmm5,xmm1       ; xmm5=roundint(data7/8)=(07 ** 17 ** 27 ** 37 **)
+
+        pand    xmm6,xmm3               ; xmm6=(00 -- 10 -- 20 -- 30 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 01 -- 11 -- 21 -- 31)
+        pand    xmm0,xmm3               ; xmm0=(06 -- 16 -- 26 -- 36 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 07 -- 17 -- 27 -- 37)
+        por     xmm6,xmm7               ; xmm6=(00 01 10 11 20 21 30 31)
+        por     xmm0,xmm5               ; xmm0=(06 07 16 17 26 27 36 37)
+
+        movaps  xmm1, XMMWORD [wk(0)]   ; xmm1=tmp2
+        movaps  xmm3, XMMWORD [wk(1)]   ; xmm3=tmp3
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm7,xmm1
+        movaps  xmm5,xmm3
+        addps   xmm1,xmm2               ; xmm1=data2=(02 12 22 32)
+        addps   xmm3,xmm4               ; xmm3=data4=(04 14 24 34)
+        subps   xmm7,xmm2               ; xmm7=data5=(05 15 25 35)
+        subps   xmm5,xmm4               ; xmm5=data3=(03 13 23 33)
+
+        movaps  xmm2,[GOTOFF(ebx,PD_RNDINT_MAGIC)]      ; xmm2=[PD_RNDINT_MAGIC]
+        pcmpeqd xmm4,xmm4
+        psrld   xmm4,WORD_BIT           ; xmm4={0xFFFF 0x0000 0xFFFF 0x0000 ..}
+
+        addps   xmm3,xmm2       ; xmm3=roundint(data4/8)=(04 ** 14 ** 24 ** 34 **)
+        addps   xmm7,xmm2       ; xmm7=roundint(data5/8)=(05 ** 15 ** 25 ** 35 **)
+        addps   xmm1,xmm2       ; xmm1=roundint(data2/8)=(02 ** 12 ** 22 ** 32 **)
+        addps   xmm5,xmm2       ; xmm5=roundint(data3/8)=(03 ** 13 ** 23 ** 33 **)
+
+        pand    xmm3,xmm4               ; xmm3=(04 -- 14 -- 24 -- 34 --)
+        pslld   xmm7,WORD_BIT           ; xmm7=(-- 05 -- 15 -- 25 -- 35)
+        pand    xmm1,xmm4               ; xmm1=(02 -- 12 -- 22 -- 32 --)
+        pslld   xmm5,WORD_BIT           ; xmm5=(-- 03 -- 13 -- 23 -- 33)
+        por     xmm3,xmm7               ; xmm3=(04 05 14 15 24 25 34 35)
+        por     xmm1,xmm5               ; xmm1=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm2=[PB_CENTERJSAMP]
+
+        packsswb  xmm6,xmm3     ; xmm6=(00 01 10 11 20 21 30 31 04 05 14 15 24 25 34 35)
+        packsswb  xmm1,xmm0     ; xmm1=(02 03 12 13 22 23 32 33 06 07 16 17 26 27 36 37)
+        paddb     xmm6,xmm2
+        paddb     xmm1,xmm2
+
+        movdqa    xmm4,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm1     ; xmm6=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm1     ; xmm4=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+
+        movdqa    xmm7,xmm6     ; transpose coefficients(phase 3)
+        punpckldq xmm6,xmm4     ; xmm6=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm7,xmm4     ; xmm7=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+
+        pshufd  xmm5,xmm6,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm3,xmm7,0x4E  ; xmm3=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm7
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE], xmm3
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctflt-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctflt-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,572 @@
+;
+; jidctflt.asm - floating-point IDCT (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a floating-point implementation of the inverse DCT
+; (Discrete Cosine Transform). The following code is based directly on
+; the IJG's original jidctflt.c; see the jidctflt.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%macro  unpcklps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(0 1 4 5)
+        shufps  %1,%2,0x44
+%endmacro
+
+%macro  unpckhps2 2     ; %1=(0 1 2 3) / %2=(4 5 6 7) => %1=(2 3 6 7)
+        shufps  %1,%2,0xEE
+%endmacro
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_float_sse) PRIVATE
+
+EXTN(jconst_idct_float_sse):
+
+PD_1_414        times 4 dd  1.414213562373095048801689
+PD_1_847        times 4 dd  1.847759065022573512256366
+PD_1_082        times 4 dd  1.082392200292393968799446
+PD_M2_613       times 4 dd -2.613125929752753055713286
+PD_0_125        times 4 dd  0.125       ; 1/8
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_float_sse (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_FAST_FLOAT
+                                        ; FAST_FLOAT workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_float_sse) PRIVATE
+
+EXTN(jsimd_idct_float_sse):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_FLOAT_SSE
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq      mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm1,mm0                       ; mm1=(** 02 ** 03)
+        punpcklwd mm0,mm0                       ; mm0=(00 00 01 01)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in0H=(02 03)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in0L=(00 01)
+        cvtpi2ps  xmm3,mm1                      ; xmm3=(02 03 ** **)
+        cvtpi2ps  xmm0,mm0                      ; xmm0=(00 01 ** **)
+        movlhps   xmm0,xmm3                     ; xmm0=in0=(00 01 02 03)
+
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm1,xmm0
+        movaps  xmm2,xmm0
+        movaps  xmm3,xmm0
+
+        shufps  xmm0,xmm0,0x00                  ; xmm0=(00 00 00 00)
+        shufps  xmm1,xmm1,0x55                  ; xmm1=(01 01 01 01)
+        shufps  xmm2,xmm2,0xAA                  ; xmm2=(02 02 02 02)
+        shufps  xmm3,xmm3,0xFF                  ; xmm3=(03 03 03 03)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq      mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq      mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq      mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq      mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm4,mm0                       ; mm4=(** 02 ** 03)
+        punpcklwd mm0,mm0                       ; mm0=(00 00 01 01)
+        punpckhwd mm5,mm1                       ; mm5=(** 22 ** 23)
+        punpcklwd mm1,mm1                       ; mm1=(20 20 21 21)
+
+        psrad     mm4,(DWORD_BIT-WORD_BIT)      ; mm4=in0H=(02 03)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in0L=(00 01)
+        cvtpi2ps  xmm4,mm4                      ; xmm4=(02 03 ** **)
+        cvtpi2ps  xmm0,mm0                      ; xmm0=(00 01 ** **)
+        psrad     mm5,(DWORD_BIT-WORD_BIT)      ; mm5=in2H=(22 23)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in2L=(20 21)
+        cvtpi2ps  xmm5,mm5                      ; xmm5=(22 23 ** **)
+        cvtpi2ps  xmm1,mm1                      ; xmm1=(20 21 ** **)
+
+        punpckhwd mm6,mm2                       ; mm6=(** 42 ** 43)
+        punpcklwd mm2,mm2                       ; mm2=(40 40 41 41)
+        punpckhwd mm7,mm3                       ; mm7=(** 62 ** 63)
+        punpcklwd mm3,mm3                       ; mm3=(60 60 61 61)
+
+        psrad     mm6,(DWORD_BIT-WORD_BIT)      ; mm6=in4H=(42 43)
+        psrad     mm2,(DWORD_BIT-WORD_BIT)      ; mm2=in4L=(40 41)
+        cvtpi2ps  xmm6,mm6                      ; xmm6=(42 43 ** **)
+        cvtpi2ps  xmm2,mm2                      ; xmm2=(40 41 ** **)
+        psrad     mm7,(DWORD_BIT-WORD_BIT)      ; mm7=in6H=(62 63)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)      ; mm3=in6L=(60 61)
+        cvtpi2ps  xmm7,mm7                      ; xmm7=(62 63 ** **)
+        cvtpi2ps  xmm3,mm3                      ; xmm3=(60 61 ** **)
+
+        movlhps   xmm0,xmm4                     ; xmm0=in0=(00 01 02 03)
+        movlhps   xmm1,xmm5                     ; xmm1=in2=(20 21 22 23)
+        mulps     xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movlhps   xmm2,xmm6                     ; xmm2=in4=(40 41 42 43)
+        movlhps   xmm3,xmm7                     ; xmm3=in6=(60 61 62 63)
+        mulps     xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movq      mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq      mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq      mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq      mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        punpckhwd mm6,mm4                       ; mm6=(** 12 ** 13)
+        punpcklwd mm4,mm4                       ; mm4=(10 10 11 11)
+        punpckhwd mm2,mm0                       ; mm2=(** 32 ** 33)
+        punpcklwd mm0,mm0                       ; mm0=(30 30 31 31)
+
+        psrad     mm6,(DWORD_BIT-WORD_BIT)      ; mm6=in1H=(12 13)
+        psrad     mm4,(DWORD_BIT-WORD_BIT)      ; mm4=in1L=(10 11)
+        cvtpi2ps  xmm4,mm6                      ; xmm4=(12 13 ** **)
+        cvtpi2ps  xmm2,mm4                      ; xmm2=(10 11 ** **)
+        psrad     mm2,(DWORD_BIT-WORD_BIT)      ; mm2=in3H=(32 33)
+        psrad     mm0,(DWORD_BIT-WORD_BIT)      ; mm0=in3L=(30 31)
+        cvtpi2ps  xmm0,mm2                      ; xmm0=(32 33 ** **)
+        cvtpi2ps  xmm3,mm0                      ; xmm3=(30 31 ** **)
+
+        punpckhwd mm7,mm5                       ; mm7=(** 52 ** 53)
+        punpcklwd mm5,mm5                       ; mm5=(50 50 51 51)
+        punpckhwd mm3,mm1                       ; mm3=(** 72 ** 73)
+        punpcklwd mm1,mm1                       ; mm1=(70 70 71 71)
+
+        movlhps   xmm2,xmm4                     ; xmm2=in1=(10 11 12 13)
+        movlhps   xmm3,xmm0                     ; xmm3=in3=(30 31 32 33)
+
+        psrad     mm7,(DWORD_BIT-WORD_BIT)      ; mm7=in5H=(52 53)
+        psrad     mm5,(DWORD_BIT-WORD_BIT)      ; mm5=in5L=(50 51)
+        cvtpi2ps  xmm4,mm7                      ; xmm4=(52 53 ** **)
+        cvtpi2ps  xmm5,mm5                      ; xmm5=(50 51 ** **)
+        psrad     mm3,(DWORD_BIT-WORD_BIT)      ; mm3=in7H=(72 73)
+        psrad     mm1,(DWORD_BIT-WORD_BIT)      ; mm1=in7L=(70 71)
+        cvtpi2ps  xmm0,mm3                      ; xmm0=(72 73 ** **)
+        cvtpi2ps  xmm1,mm1                      ; xmm1=(70 71 ** **)
+
+        mulps     xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movlhps   xmm5,xmm4                     ; xmm5=in5=(50 51 52 53)
+        movlhps   xmm1,xmm0                     ; xmm1=in7=(70 71 72 73)
+        mulps     xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+        mulps     xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_FLOAT_MULT_TYPE)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 01 02 03)
+        addps   xmm7,xmm3               ; xmm7=data1=(10 11 12 13)
+        subps   xmm5,xmm1               ; xmm5=data7=(70 71 72 73)
+        subps   xmm0,xmm3               ; xmm0=data6=(60 61 62 63)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps    xmm1,xmm6             ; transpose coefficients(phase 1)
+        unpcklps  xmm6,xmm7             ; xmm6=(00 10 01 11)
+        unpckhps  xmm1,xmm7             ; xmm1=(02 12 03 13)
+        movaps    xmm3,xmm0             ; transpose coefficients(phase 1)
+        unpcklps  xmm0,xmm5             ; xmm0=(60 70 61 71)
+        unpckhps  xmm3,xmm5             ; xmm3=(62 72 63 73)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movaps  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp3
+
+        movaps  XMMWORD [wk(0)], xmm0   ; wk(0)=(60 70 61 71)
+        movaps  XMMWORD [wk(1)], xmm3   ; wk(1)=(62 72 63 73)
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm0,xmm7
+        movaps  xmm3,xmm5
+        addps   xmm7,xmm2               ; xmm7=data2=(20 21 22 23)
+        addps   xmm5,xmm4               ; xmm5=data4=(40 41 42 43)
+        subps   xmm0,xmm2               ; xmm0=data5=(50 51 52 53)
+        subps   xmm3,xmm4               ; xmm3=data3=(30 31 32 33)
+
+        movaps    xmm2,xmm7             ; transpose coefficients(phase 1)
+        unpcklps  xmm7,xmm3             ; xmm7=(20 30 21 31)
+        unpckhps  xmm2,xmm3             ; xmm2=(22 32 23 33)
+        movaps    xmm4,xmm5             ; transpose coefficients(phase 1)
+        unpcklps  xmm5,xmm0             ; xmm5=(40 50 41 51)
+        unpckhps  xmm4,xmm0             ; xmm4=(42 52 43 53)
+
+        movaps    xmm3,xmm6             ; transpose coefficients(phase 2)
+        unpcklps2 xmm6,xmm7             ; xmm6=(00 10 20 30)
+        unpckhps2 xmm3,xmm7             ; xmm3=(01 11 21 31)
+        movaps    xmm0,xmm1             ; transpose coefficients(phase 2)
+        unpcklps2 xmm1,xmm2             ; xmm1=(02 12 22 32)
+        unpckhps2 xmm0,xmm2             ; xmm0=(03 13 23 33)
+
+        movaps  xmm7, XMMWORD [wk(0)]   ; xmm7=(60 70 61 71)
+        movaps  xmm2, XMMWORD [wk(1)]   ; xmm2=(62 72 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_FAST_FLOAT)], xmm1
+        movaps  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+
+        movaps    xmm6,xmm5             ; transpose coefficients(phase 2)
+        unpcklps2 xmm5,xmm7             ; xmm5=(40 50 60 70)
+        unpckhps2 xmm6,xmm7             ; xmm6=(41 51 61 71)
+        movaps    xmm3,xmm4             ; transpose coefficients(phase 2)
+        unpcklps2 xmm4,xmm2             ; xmm4=(42 52 62 72)
+        unpckhps2 xmm3,xmm2             ; xmm3=(43 53 63 73)
+
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm5
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+        movaps  XMMWORD [XMMBLOCK(2,1,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(3,1,edi,SIZEOF_FAST_FLOAT)], xmm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_FLOAT_MULT_TYPE      ; quantptr
+        add     edi,      4*DCTSIZE*SIZEOF_FAST_FLOAT   ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + (DCTSIZE2-8)*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; FAST_FLOAT *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm0
+        movaps  xmm5,xmm1
+        subps   xmm0,xmm2               ; xmm0=tmp11
+        subps   xmm1,xmm3
+        addps   xmm4,xmm2               ; xmm4=tmp10
+        addps   xmm5,xmm3               ; xmm5=tmp13
+
+        mulps   xmm1,[GOTOFF(ebx,PD_1_414)]
+        subps   xmm1,xmm5               ; xmm1=tmp12
+
+        movaps  xmm6,xmm4
+        movaps  xmm7,xmm0
+        subps   xmm4,xmm5               ; xmm4=tmp3
+        subps   xmm0,xmm1               ; xmm0=tmp2
+        addps   xmm6,xmm5               ; xmm6=tmp0
+        addps   xmm7,xmm1               ; xmm7=tmp1
+
+        movaps  XMMWORD [wk(1)], xmm4   ; tmp3
+        movaps  XMMWORD [wk(0)], xmm0   ; tmp2
+
+        ; -- Odd part
+
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_FAST_FLOAT)]
+
+        movaps  xmm4,xmm2
+        movaps  xmm0,xmm5
+        addps   xmm2,xmm1               ; xmm2=z11
+        addps   xmm5,xmm3               ; xmm5=z13
+        subps   xmm4,xmm1               ; xmm4=z12
+        subps   xmm0,xmm3               ; xmm0=z10
+
+        movaps  xmm1,xmm2
+        subps   xmm2,xmm5
+        addps   xmm1,xmm5               ; xmm1=tmp7
+
+        mulps   xmm2,[GOTOFF(ebx,PD_1_414)]     ; xmm2=tmp11
+
+        movaps  xmm3,xmm0
+        addps   xmm0,xmm4
+        mulps   xmm0,[GOTOFF(ebx,PD_1_847)]     ; xmm0=z5
+        mulps   xmm3,[GOTOFF(ebx,PD_M2_613)]    ; xmm3=(z10 * -2.613125930)
+        mulps   xmm4,[GOTOFF(ebx,PD_1_082)]     ; xmm4=(z12 * 1.082392200)
+        addps   xmm3,xmm0               ; xmm3=tmp12
+        subps   xmm4,xmm0               ; xmm4=tmp10
+
+        ; -- Final output stage
+
+        subps   xmm3,xmm1               ; xmm3=tmp6
+        movaps  xmm5,xmm6
+        movaps  xmm0,xmm7
+        addps   xmm6,xmm1               ; xmm6=data0=(00 10 20 30)
+        addps   xmm7,xmm3               ; xmm7=data1=(01 11 21 31)
+        subps   xmm5,xmm1               ; xmm5=data7=(07 17 27 37)
+        subps   xmm0,xmm3               ; xmm0=data6=(06 16 26 36)
+        subps   xmm2,xmm3               ; xmm2=tmp5
+
+        movaps  xmm1,[GOTOFF(ebx,PD_0_125)]     ; xmm1=[PD_0_125]
+
+        mulps   xmm6,xmm1               ; descale(1/8)
+        mulps   xmm7,xmm1               ; descale(1/8)
+        mulps   xmm5,xmm1               ; descale(1/8)
+        mulps   xmm0,xmm1               ; descale(1/8)
+
+        movhlps   xmm3,xmm6
+        movhlps   xmm1,xmm7
+        cvtps2pi  mm0,xmm6              ; round to int32, mm0=data0L=(00 10)
+        cvtps2pi  mm1,xmm7              ; round to int32, mm1=data1L=(01 11)
+        cvtps2pi  mm2,xmm3              ; round to int32, mm2=data0H=(20 30)
+        cvtps2pi  mm3,xmm1              ; round to int32, mm3=data1H=(21 31)
+        packssdw  mm0,mm2               ; mm0=data0=(00 10 20 30)
+        packssdw  mm1,mm3               ; mm1=data1=(01 11 21 31)
+
+        movhlps   xmm6,xmm5
+        movhlps   xmm7,xmm0
+        cvtps2pi  mm4,xmm5              ; round to int32, mm4=data7L=(07 17)
+        cvtps2pi  mm5,xmm0              ; round to int32, mm5=data6L=(06 16)
+        cvtps2pi  mm6,xmm6              ; round to int32, mm6=data7H=(27 37)
+        cvtps2pi  mm7,xmm7              ; round to int32, mm7=data6H=(26 36)
+        packssdw  mm4,mm6               ; mm4=data7=(07 17 27 37)
+        packssdw  mm5,mm7               ; mm5=data6=(06 16 26 36)
+
+        packsswb  mm0,mm5               ; mm0=(00 10 20 30 06 16 26 36)
+        packsswb  mm1,mm4               ; mm1=(01 11 21 31 07 17 27 37)
+
+        movaps  xmm3, XMMWORD [wk(0)]   ; xmm3=tmp2
+        movaps  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movaps  xmm6,[GOTOFF(ebx,PD_0_125)]     ; xmm6=[PD_0_125]
+
+        addps   xmm4,xmm2               ; xmm4=tmp4
+        movaps  xmm5,xmm3
+        movaps  xmm0,xmm1
+        addps   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+        addps   xmm1,xmm4               ; xmm1=data4=(04 14 24 34)
+        subps   xmm5,xmm2               ; xmm5=data5=(05 15 25 35)
+        subps   xmm0,xmm4               ; xmm0=data3=(03 13 23 33)
+
+        mulps   xmm3,xmm6               ; descale(1/8)
+        mulps   xmm1,xmm6               ; descale(1/8)
+        mulps   xmm5,xmm6               ; descale(1/8)
+        mulps   xmm0,xmm6               ; descale(1/8)
+
+        movhlps   xmm7,xmm3
+        movhlps   xmm2,xmm1
+        cvtps2pi  mm2,xmm3              ; round to int32, mm2=data2L=(02 12)
+        cvtps2pi  mm3,xmm1              ; round to int32, mm3=data4L=(04 14)
+        cvtps2pi  mm6,xmm7              ; round to int32, mm6=data2H=(22 32)
+        cvtps2pi  mm7,xmm2              ; round to int32, mm7=data4H=(24 34)
+        packssdw  mm2,mm6               ; mm2=data2=(02 12 22 32)
+        packssdw  mm3,mm7               ; mm3=data4=(04 14 24 34)
+
+        movhlps   xmm4,xmm5
+        movhlps   xmm6,xmm0
+        cvtps2pi  mm5,xmm5              ; round to int32, mm5=data5L=(05 15)
+        cvtps2pi  mm4,xmm0              ; round to int32, mm4=data3L=(03 13)
+        cvtps2pi  mm6,xmm4              ; round to int32, mm6=data5H=(25 35)
+        cvtps2pi  mm7,xmm6              ; round to int32, mm7=data3H=(23 33)
+        packssdw  mm5,mm6               ; mm5=data5=(05 15 25 35)
+        packssdw  mm4,mm7               ; mm4=data3=(03 13 23 33)
+
+        movq      mm6,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm6=[PB_CENTERJSAMP]
+
+        packsswb  mm2,mm3               ; mm2=(02 12 22 32 04 14 24 34)
+        packsswb  mm4,mm5               ; mm4=(03 13 23 33 05 15 25 35)
+
+        paddb     mm0,mm6
+        paddb     mm1,mm6
+        paddb     mm2,mm6
+        paddb     mm4,mm6
+
+        movq      mm7,mm0               ; transpose coefficients(phase 1)
+        punpcklbw mm0,mm1               ; mm0=(00 01 10 11 20 21 30 31)
+        punpckhbw mm7,mm1               ; mm7=(06 07 16 17 26 27 36 37)
+        movq      mm3,mm2               ; transpose coefficients(phase 1)
+        punpcklbw mm2,mm4               ; mm2=(02 03 12 13 22 23 32 33)
+        punpckhbw mm3,mm4               ; mm3=(04 05 14 15 24 25 34 35)
+
+        movq      mm5,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm2               ; mm0=(00 01 02 03 10 11 12 13)
+        punpckhwd mm5,mm2               ; mm5=(20 21 22 23 30 31 32 33)
+        movq      mm6,mm3               ; transpose coefficients(phase 2)
+        punpcklwd mm3,mm7               ; mm3=(04 05 06 07 14 15 16 17)
+        punpckhwd mm6,mm7               ; mm6=(24 25 26 27 34 35 36 37)
+
+        movq      mm1,mm0               ; transpose coefficients(phase 3)
+        punpckldq mm0,mm3               ; mm0=(00 01 02 03 04 05 06 07)
+        punpckhdq mm1,mm3               ; mm1=(10 11 12 13 14 15 16 17)
+        movq      mm4,mm5               ; transpose coefficients(phase 3)
+        punpckldq mm5,mm6               ; mm5=(20 21 22 23 24 25 26 27)
+        punpckhdq mm4,mm6               ; mm4=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_FAST_FLOAT   ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,257 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* FAST INTEGER INVERSE DCT
+ *
+ * This is similar to the SSE2 implementation, except that we left-shift the
+ * constants by 1 less bit (the -1 in CONST_SHIFT.)  This is because
+ * vec_madds(arg1, arg2, arg3) generates the 16-bit saturated sum of:
+ *   the elements in arg3 + the most significant 17 bits of
+ *     (the elements in arg1 * the elements in arg2).
+ */
+
+#include "jsimd_altivec.h"
+
+
+#define F_1_082 277              /* FIX(1.082392200) */
+#define F_1_414 362              /* FIX(1.414213562) */
+#define F_1_847 473              /* FIX(1.847759065) */
+#define F_2_613 669              /* FIX(2.613125930) */
+#define F_1_613 (F_2_613 - 256)  /* FIX(2.613125930) - FIX(1) */
+
+#define CONST_BITS 8
+#define PASS1_BITS 2
+#define PRE_MULTIPLY_SCALE_BITS 2
+#define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS - 1)
+
+
+#define DO_IDCT(in)  \
+{  \
+  /* Even part */  \
+  \
+  tmp10 = vec_add(in##0, in##4);  \
+  tmp11 = vec_sub(in##0, in##4);  \
+  tmp13 = vec_add(in##2, in##6);  \
+  \
+  tmp12 = vec_sub(in##2, in##6);  \
+  tmp12 = vec_sl(tmp12, pre_multiply_scale_bits);  \
+  tmp12 = vec_madds(tmp12, pw_F1414, pw_zero);  \
+  tmp12 = vec_sub(tmp12, tmp13);  \
+  \
+  tmp0 = vec_add(tmp10, tmp13);  \
+  tmp3 = vec_sub(tmp10, tmp13);  \
+  tmp1 = vec_add(tmp11, tmp12);  \
+  tmp2 = vec_sub(tmp11, tmp12);  \
+  \
+  /* Odd part */  \
+  \
+  z13 = vec_add(in##5, in##3);  \
+  z10 = vec_sub(in##5, in##3);  \
+  z10s = vec_sl(z10, pre_multiply_scale_bits);  \
+  z11 = vec_add(in##1, in##7);  \
+  z12s = vec_sub(in##1, in##7);  \
+  z12s = vec_sl(z12s, pre_multiply_scale_bits);  \
+  \
+  tmp11 = vec_sub(z11, z13);  \
+  tmp11 = vec_sl(tmp11, pre_multiply_scale_bits);  \
+  tmp11 = vec_madds(tmp11, pw_F1414, pw_zero);  \
+  \
+  tmp7 = vec_add(z11, z13);  \
+  \
+  /* To avoid overflow...  \
+   *  \
+   * (Original)  \
+   * tmp12 = -2.613125930 * z10 + z5;  \
+   *  \
+   * (This implementation)  \
+   * tmp12 = (-1.613125930 - 1) * z10 + z5;  \
+   *       = -1.613125930 * z10 - z10 + z5;  \
+   */  \
+  \
+  z5 = vec_add(z10s, z12s);  \
+  z5 = vec_madds(z5, pw_F1847, pw_zero);  \
+  \
+  tmp10 = vec_madds(z12s, pw_F1082, pw_zero);  \
+  tmp10 = vec_sub(tmp10, z5);  \
+  tmp12 = vec_madds(z10s, pw_MF1613, z5);  \
+  tmp12 = vec_sub(tmp12, z10);  \
+  \
+  tmp6 = vec_sub(tmp12, tmp7);  \
+  tmp5 = vec_sub(tmp11, tmp6);  \
+  tmp4 = vec_add(tmp10, tmp5);  \
+  \
+  out0 = vec_add(tmp0, tmp7);  \
+  out1 = vec_add(tmp1, tmp6);  \
+  out2 = vec_add(tmp2, tmp5);  \
+  out3 = vec_sub(tmp3, tmp4);  \
+  out4 = vec_add(tmp3, tmp4);  \
+  out5 = vec_sub(tmp2, tmp5);  \
+  out6 = vec_sub(tmp1, tmp6);  \
+  out7 = vec_sub(tmp0, tmp7);  \
+}
+
+
+void
+jsimd_idct_ifast_altivec (void *dct_table_, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp10, tmp11, tmp12, tmp13,
+    z5, z10, z10s, z11, z12s, z13,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_F1414 = { __8X(F_1_414 << CONST_SHIFT) },
+    pw_F1847 = { __8X(F_1_847 << CONST_SHIFT) },
+    pw_MF1613 = { __8X(-F_1_613 << CONST_SHIFT) },
+    pw_F1082 = { __8X(F_1_082 << CONST_SHIFT) };
+  __vector unsigned short
+    pre_multiply_scale_bits = { __8X(PRE_MULTIPLY_SCALE_BITS) },
+    pass1_bits3 = { __8X(PASS1_BITS + 3) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row);
+
+  out0 = vec_sra(out0, pass1_bits3);
+  out1 = vec_sra(out1, pass1_bits3);
+  out2 = vec_sra(out2, pass1_bits3);
+  out3 = vec_sra(out3, pass1_bits3);
+  out4 = vec_sra(out4, pass1_bits3);
+  out5 = vec_sra(out5, pass1_bits3);
+  out6 = vec_sra(out6, pass1_bits3);
+  out7 = vec_sra(out7, pass1_bits3);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[0] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[1] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[2] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[3] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[4] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[5] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[6] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[7] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,500 @@
+;
+; jidctfst.asm - fast integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_mmx) PRIVATE
+
+EXTN(jconst_idct_ifast_mmx):
+
+PW_F1414        times 4 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 4 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 4 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 4 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_mmx (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_mmx) PRIVATE
+
+EXTN(jsimd_idct_ifast_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        psubw   mm0,mm2                 ; mm0=tmp11
+        psubw   mm1,mm3
+        paddw   mm4,mm2                 ; mm4=tmp10
+        paddw   mm5,mm3                 ; mm5=tmp13
+
+        psllw   mm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        psubw   mm4,mm5                 ; mm4=tmp3
+        psubw   mm0,mm1                 ; mm0=tmp2
+        paddw   mm6,mm5                 ; mm6=tmp0
+        paddw   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=tmp3
+        movq    MMWORD [wk(0)], mm0     ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm5, MMWORD [MMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        psubw   mm2,mm1                 ; mm2=z12
+        psubw   mm5,mm3                 ; mm5=z10
+        paddw   mm4,mm1                 ; mm4=z11
+        paddw   mm0,mm3                 ; mm0=z13
+
+        movq    mm1,mm5                 ; mm1=z10(unscaled)
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm5,PRE_MULTIPLY_SCALE_BITS
+
+        movq    mm3,mm4
+        psubw   mm4,mm0
+        paddw   mm3,mm0                 ; mm3=tmp7
+
+        psllw   mm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm4,[GOTOFF(ebx,PW_F1414)]      ; mm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movq    mm0,mm5
+        paddw   mm5,mm2
+        pmulhw  mm5,[GOTOFF(ebx,PW_F1847)]      ; mm5=z5
+        pmulhw  mm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  mm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   mm0,mm1
+        psubw   mm2,mm5                 ; mm2=tmp10
+        paddw   mm0,mm5                 ; mm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   mm0,mm3                 ; mm0=tmp6
+        movq    mm1,mm6
+        movq    mm5,mm7
+        paddw   mm6,mm3                 ; mm6=data0=(00 01 02 03)
+        paddw   mm7,mm0                 ; mm7=data1=(10 11 12 13)
+        psubw   mm1,mm3                 ; mm1=data7=(70 71 72 73)
+        psubw   mm5,mm0                 ; mm5=data6=(60 61 62 63)
+        psubw   mm4,mm0                 ; mm4=tmp5
+
+        movq      mm3,mm6               ; transpose coefficients(phase 1)
+        punpcklwd mm6,mm7               ; mm6=(00 10 01 11)
+        punpckhwd mm3,mm7               ; mm3=(02 12 03 13)
+        movq      mm0,mm5               ; transpose coefficients(phase 1)
+        punpcklwd mm5,mm1               ; mm5=(60 70 61 71)
+        punpckhwd mm0,mm1               ; mm0=(62 72 63 73)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp2
+        movq    mm1, MMWORD [wk(1)]     ; mm1=tmp3
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(60 70 61 71)
+        movq    MMWORD [wk(1)], mm0     ; wk(1)=(62 72 63 73)
+
+        paddw   mm2,mm4                 ; mm2=tmp4
+        movq    mm5,mm7
+        movq    mm0,mm1
+        paddw   mm7,mm4                 ; mm7=data2=(20 21 22 23)
+        paddw   mm1,mm2                 ; mm1=data4=(40 41 42 43)
+        psubw   mm5,mm4                 ; mm5=data5=(50 51 52 53)
+        psubw   mm0,mm2                 ; mm0=data3=(30 31 32 33)
+
+        movq      mm4,mm7               ; transpose coefficients(phase 1)
+        punpcklwd mm7,mm0               ; mm7=(20 30 21 31)
+        punpckhwd mm4,mm0               ; mm4=(22 32 23 33)
+        movq      mm2,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm5               ; mm1=(40 50 41 51)
+        punpckhwd mm2,mm5               ; mm2=(42 52 43 53)
+
+        movq      mm0,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm7               ; mm6=(00 10 20 30)
+        punpckhdq mm0,mm7               ; mm0=(01 11 21 31)
+        movq      mm5,mm3               ; transpose coefficients(phase 2)
+        punpckldq mm3,mm4               ; mm3=(02 12 22 32)
+        punpckhdq mm5,mm4               ; mm5=(03 13 23 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=(60 70 61 71)
+        movq    mm4, MMWORD [wk(1)]     ; mm4=(62 72 63 73)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5
+
+        movq      mm6,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm7               ; mm1=(40 50 60 70)
+        punpckhdq mm6,mm7               ; mm6=(41 51 61 71)
+        movq      mm0,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm4               ; mm2=(42 52 62 72)
+        punpckhdq mm0,mm4               ; mm0=(43 53 63 73)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm0
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_IFAST_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm4,mm0
+        movq    mm5,mm1
+        psubw   mm0,mm2                 ; mm0=tmp11
+        psubw   mm1,mm3
+        paddw   mm4,mm2                 ; mm4=tmp10
+        paddw   mm5,mm3                 ; mm5=tmp13
+
+        psllw   mm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   mm1,mm5                 ; mm1=tmp12
+
+        movq    mm6,mm4
+        movq    mm7,mm0
+        psubw   mm4,mm5                 ; mm4=tmp3
+        psubw   mm0,mm1                 ; mm0=tmp2
+        paddw   mm6,mm5                 ; mm6=tmp0
+        paddw   mm7,mm1                 ; mm7=tmp1
+
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=tmp3
+        movq    MMWORD [wk(0)], mm0     ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movq    mm2, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm4,mm2
+        movq    mm0,mm5
+        psubw   mm2,mm1                 ; mm2=z12
+        psubw   mm5,mm3                 ; mm5=z10
+        paddw   mm4,mm1                 ; mm4=z11
+        paddw   mm0,mm3                 ; mm0=z13
+
+        movq    mm1,mm5                 ; mm1=z10(unscaled)
+        psllw   mm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   mm5,PRE_MULTIPLY_SCALE_BITS
+
+        movq    mm3,mm4
+        psubw   mm4,mm0
+        paddw   mm3,mm0                 ; mm3=tmp7
+
+        psllw   mm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  mm4,[GOTOFF(ebx,PW_F1414)]      ; mm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movq    mm0,mm5
+        paddw   mm5,mm2
+        pmulhw  mm5,[GOTOFF(ebx,PW_F1847)]      ; mm5=z5
+        pmulhw  mm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  mm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   mm0,mm1
+        psubw   mm2,mm5                 ; mm2=tmp10
+        paddw   mm0,mm5                 ; mm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   mm0,mm3                 ; mm0=tmp6
+        movq    mm1,mm6
+        movq    mm5,mm7
+        paddw   mm6,mm3                 ; mm6=data0=(00 10 20 30)
+        paddw   mm7,mm0                 ; mm7=data1=(01 11 21 31)
+        psraw   mm6,(PASS1_BITS+3)      ; descale
+        psraw   mm7,(PASS1_BITS+3)      ; descale
+        psubw   mm1,mm3                 ; mm1=data7=(07 17 27 37)
+        psubw   mm5,mm0                 ; mm5=data6=(06 16 26 36)
+        psraw   mm1,(PASS1_BITS+3)      ; descale
+        psraw   mm5,(PASS1_BITS+3)      ; descale
+        psubw   mm4,mm0                 ; mm4=tmp5
+
+        packsswb  mm6,mm5               ; mm6=(00 10 20 30 06 16 26 36)
+        packsswb  mm7,mm1               ; mm7=(01 11 21 31 07 17 27 37)
+
+        movq    mm3, MMWORD [wk(0)]     ; mm3=tmp2
+        movq    mm0, MMWORD [wk(1)]     ; mm0=tmp3
+
+        paddw   mm2,mm4                 ; mm2=tmp4
+        movq    mm5,mm3
+        movq    mm1,mm0
+        paddw   mm3,mm4                 ; mm3=data2=(02 12 22 32)
+        paddw   mm0,mm2                 ; mm0=data4=(04 14 24 34)
+        psraw   mm3,(PASS1_BITS+3)      ; descale
+        psraw   mm0,(PASS1_BITS+3)      ; descale
+        psubw   mm5,mm4                 ; mm5=data5=(05 15 25 35)
+        psubw   mm1,mm2                 ; mm1=data3=(03 13 23 33)
+        psraw   mm5,(PASS1_BITS+3)      ; descale
+        psraw   mm1,(PASS1_BITS+3)      ; descale
+
+        movq      mm4,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm4=[PB_CENTERJSAMP]
+
+        packsswb  mm3,mm0               ; mm3=(02 12 22 32 04 14 24 34)
+        packsswb  mm1,mm5               ; mm1=(03 13 23 33 05 15 25 35)
+
+        paddb     mm6,mm4
+        paddb     mm7,mm4
+        paddb     mm3,mm4
+        paddb     mm1,mm4
+
+        movq      mm2,mm6               ; transpose coefficients(phase 1)
+        punpcklbw mm6,mm7               ; mm6=(00 01 10 11 20 21 30 31)
+        punpckhbw mm2,mm7               ; mm2=(06 07 16 17 26 27 36 37)
+        movq      mm0,mm3               ; transpose coefficients(phase 1)
+        punpcklbw mm3,mm1               ; mm3=(02 03 12 13 22 23 32 33)
+        punpckhbw mm0,mm1               ; mm0=(04 05 14 15 24 25 34 35)
+
+        movq      mm5,mm6               ; transpose coefficients(phase 2)
+        punpcklwd mm6,mm3               ; mm6=(00 01 02 03 10 11 12 13)
+        punpckhwd mm5,mm3               ; mm5=(20 21 22 23 30 31 32 33)
+        movq      mm4,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm2               ; mm0=(04 05 06 07 14 15 16 17)
+        punpckhwd mm4,mm2               ; mm4=(24 25 26 27 34 35 36 37)
+
+        movq      mm7,mm6               ; transpose coefficients(phase 3)
+        punpckldq mm6,mm0               ; mm6=(00 01 02 03 04 05 06 07)
+        punpckhdq mm7,mm0               ; mm7=(10 11 12 13 14 15 16 17)
+        movq      mm1,mm5               ; transpose coefficients(phase 3)
+        punpckldq mm5,mm4               ; mm5=(20 21 22 23 24 25 26 27)
+        punpckhdq mm1,mm4               ; mm1=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm6
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm7
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm5
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm1
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_JCOEF        ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,492 @@
+;
+; jidctfst.asm - fast integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_sse2) PRIVATE
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414        times 8 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 8 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 8 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 8 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_idct_ifast_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm7,xmm0             ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm7,xmm7             ; xmm7=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm6,xmm0,0x00          ; xmm6=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm2,xmm0,0x55          ; xmm2=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm5,xmm0,0xAA          ; xmm5=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm0,xmm0,0xFF          ; xmm0=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm1,xmm7,0x00          ; xmm1=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm4,xmm7,0x55          ; xmm4=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm3,xmm7,0xAA          ; xmm3=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm7,xmm7,0xFF          ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=col3
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        psubw   xmm0,xmm2               ; xmm0=tmp11
+        psubw   xmm1,xmm3
+        paddw   xmm4,xmm2               ; xmm4=tmp10
+        paddw   xmm5,xmm3               ; xmm5=tmp13
+
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm1,[rel PW_F1414]
+        psubw   xmm1,xmm5               ; xmm1=tmp12
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm7,xmm0
+        psubw   xmm4,xmm5               ; xmm4=tmp3
+        psubw   xmm0,xmm1               ; xmm0=tmp2
+        paddw   xmm6,xmm5               ; xmm6=tmp0
+        paddw   xmm7,xmm1               ; xmm7=tmp1
+
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=tmp3
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movdqa  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm0,xmm5
+        psubw   xmm2,xmm1               ; xmm2=z12
+        psubw   xmm5,xmm3               ; xmm5=z10
+        paddw   xmm4,xmm1               ; xmm4=z11
+        paddw   xmm0,xmm3               ; xmm0=z13
+
+        movdqa  xmm1,xmm5               ; xmm1=z10(unscaled)
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm3,xmm4
+        psubw   xmm4,xmm0
+        paddw   xmm3,xmm0               ; xmm3=tmp7
+
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm4,[rel PW_F1414]     ; xmm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm0,xmm5
+        paddw   xmm5,xmm2
+        pmulhw  xmm5,[rel PW_F1847]     ; xmm5=z5
+        pmulhw  xmm0,[rel PW_MF1613]
+        pmulhw  xmm2,[rel PW_F1082]
+        psubw   xmm0,xmm1
+        psubw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm0,xmm5               ; xmm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm0,xmm3               ; xmm0=tmp6
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm7
+        paddw   xmm6,xmm3               ; xmm6=data0=(00 01 02 03 04 05 06 07)
+        paddw   xmm7,xmm0               ; xmm7=data1=(10 11 12 13 14 15 16 17)
+        psubw   xmm1,xmm3               ; xmm1=data7=(70 71 72 73 74 75 76 77)
+        psubw   xmm5,xmm0               ; xmm5=data6=(60 61 62 63 64 65 66 67)
+        psubw   xmm4,xmm0               ; xmm4=tmp5
+
+        movdqa    xmm3,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm3,xmm7             ; xmm3=(04 14 05 15 06 16 07 17)
+        movdqa    xmm0,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm1             ; xmm5=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm0,xmm1             ; xmm0=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+        paddw   xmm2,xmm4               ; xmm2=tmp4
+        movdqa  xmm5,xmm7
+        movdqa  xmm0,xmm1
+        paddw   xmm7,xmm4               ; xmm7=data2=(20 21 22 23 24 25 26 27)
+        paddw   xmm1,xmm2               ; xmm1=data4=(40 41 42 43 44 45 46 47)
+        psubw   xmm5,xmm4               ; xmm5=data5=(50 51 52 53 54 55 56 57)
+        psubw   xmm0,xmm2               ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm0             ; xmm7=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm0             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm5             ; xmm2=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm0,xmm3             ; transpose coefficients(phase 2)
+        punpckldq xmm3,xmm4             ; xmm3=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm0,xmm4             ; xmm0=(06 16 26 36 07 17 27 37)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7             ; xmm6=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm5,xmm7             ; xmm5=(02 12 22 32 03 13 23 33)
+
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(60 70 61 71 62 72 63 73)
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm3,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm4             ; xmm1=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm3,xmm4             ; xmm3=(42 52 62 72 43 53 63 73)
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm7             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm0,xmm7             ; xmm0=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm4,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm1            ; xmm6=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm4,xmm1            ; xmm4=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm7,xmm5            ; transpose coefficients(phase 3)
+        punpcklqdq xmm5,xmm3            ; xmm5=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm7,xmm3            ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(04 14 24 34 05 15 25 35)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=col3
+
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm4,xmm2            ; xmm4=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm7,xmm3            ; transpose coefficients(phase 3)
+        punpcklqdq xmm3,xmm0            ; xmm3=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm7,xmm0            ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm0,xmm5
+        psubw   xmm6,xmm1               ; xmm6=tmp11
+        psubw   xmm5,xmm3
+        paddw   xmm2,xmm1               ; xmm2=tmp10
+        paddw   xmm0,xmm3               ; xmm0=tmp13
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[rel PW_F1414]
+        psubw   xmm5,xmm0               ; xmm5=tmp12
+
+        movdqa  xmm1,xmm2
+        movdqa  xmm3,xmm6
+        psubw   xmm2,xmm0               ; xmm2=tmp3
+        psubw   xmm6,xmm5               ; xmm6=tmp2
+        paddw   xmm1,xmm0               ; xmm1=tmp0
+        paddw   xmm3,xmm5               ; xmm3=tmp1
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=col1
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=col3
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp3
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp2
+
+        ; -- Odd part
+
+        ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+        movdqa  xmm2,xmm0
+        movdqa  xmm6,xmm4
+        psubw   xmm0,xmm7               ; xmm0=z12
+        psubw   xmm4,xmm5               ; xmm4=z10
+        paddw   xmm2,xmm7               ; xmm2=z11
+        paddw   xmm6,xmm5               ; xmm6=z13
+
+        movdqa  xmm7,xmm4               ; xmm7=z10(unscaled)
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm5,xmm2
+        psubw   xmm2,xmm6
+        paddw   xmm5,xmm6               ; xmm5=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm2,[rel PW_F1414]     ; xmm2=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm6,xmm4
+        paddw   xmm4,xmm0
+        pmulhw  xmm4,[rel PW_F1847]     ; xmm4=z5
+        pmulhw  xmm6,[rel PW_MF1613]
+        pmulhw  xmm0,[rel PW_F1082]
+        psubw   xmm6,xmm7
+        psubw   xmm0,xmm4               ; xmm0=tmp10
+        paddw   xmm6,xmm4               ; xmm6=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm6,xmm5               ; xmm6=tmp6
+        movdqa  xmm7,xmm1
+        movdqa  xmm4,xmm3
+        paddw   xmm1,xmm5               ; xmm1=data0=(00 10 20 30 40 50 60 70)
+        paddw   xmm3,xmm6               ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        psraw   xmm1,(PASS1_BITS+3)     ; descale
+        psraw   xmm3,(PASS1_BITS+3)     ; descale
+        psubw   xmm7,xmm5               ; xmm7=data7=(07 17 27 37 47 57 67 77)
+        psubw   xmm4,xmm6               ; xmm4=data6=(06 16 26 36 46 56 66 76)
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psubw   xmm2,xmm6               ; xmm2=tmp5
+
+        packsswb  xmm1,xmm4     ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm7     ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp2
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp3
+
+        paddw   xmm0,xmm2               ; xmm0=tmp4
+        movdqa  xmm4,xmm5
+        movdqa  xmm7,xmm6
+        paddw   xmm5,xmm2               ; xmm5=data2=(02 12 22 32 42 52 62 72)
+        paddw   xmm6,xmm0               ; xmm6=data4=(04 14 24 34 44 54 64 74)
+        psraw   xmm5,(PASS1_BITS+3)     ; descale
+        psraw   xmm6,(PASS1_BITS+3)     ; descale
+        psubw   xmm4,xmm2               ; xmm4=data5=(05 15 25 35 45 55 65 75)
+        psubw   xmm7,xmm0               ; xmm7=data3=(03 13 23 33 43 53 63 73)
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+
+        movdqa    xmm2,[rel PB_CENTERJSAMP]     ; xmm2=[rel PB_CENTERJSAMP]
+
+        packsswb  xmm5,xmm6     ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm7,xmm4     ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm1,xmm2
+        paddb     xmm3,xmm2
+        paddb     xmm5,xmm2
+        paddb     xmm7,xmm2
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 1)
+        punpcklbw xmm1,xmm3     ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm3     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm6,xmm5     ; transpose coefficients(phase 1)
+        punpcklbw xmm5,xmm7     ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm6,xmm7     ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm1     ; transpose coefficients(phase 2)
+        punpcklwd xmm1,xmm5     ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm5     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm0     ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm2,xmm0     ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm3,xmm1     ; transpose coefficients(phase 3)
+        punpckldq xmm1,xmm6     ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm3,xmm6     ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm7,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm2     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm7,xmm2     ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm5,xmm1,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm3,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm6,xmm4,0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm2,xmm7,0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+        mov     rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm7
+
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0
+        mov     rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctfst-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctfst-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,502 @@
+;
+; jidctfst.asm - fast integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a fast, not so accurate integer implementation of
+; the inverse DCT (Discrete Cosine Transform). The following code is
+; based directly on the IJG's original jidctfst.c; see the jidctfst.c
+; for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      8       ; 14 is also OK.
+%define PASS1_BITS      2
+
+%if IFAST_SCALE_BITS != PASS1_BITS
+%error "'IFAST_SCALE_BITS' must be equal to 'PASS1_BITS'."
+%endif
+
+%if CONST_BITS == 8
+F_1_082 equ     277             ; FIX(1.082392200)
+F_1_414 equ     362             ; FIX(1.414213562)
+F_1_847 equ     473             ; FIX(1.847759065)
+F_2_613 equ     669             ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - 256) ; FIX(2.613125930) - FIX(1)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_1_082 equ     DESCALE(1162209775,30-CONST_BITS)       ; FIX(1.082392200)
+F_1_414 equ     DESCALE(1518500249,30-CONST_BITS)       ; FIX(1.414213562)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_613 equ     DESCALE(2805822602,30-CONST_BITS)       ; FIX(2.613125930)
+F_1_613 equ     (F_2_613 - (1 << CONST_BITS))   ; FIX(2.613125930) - FIX(1)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+; PRE_MULTIPLY_SCALE_BITS <= 2 (to avoid overflow)
+; CONST_BITS + CONST_SHIFT + PRE_MULTIPLY_SCALE_BITS == 16 (for pmulhw)
+
+%define PRE_MULTIPLY_SCALE_BITS   2
+%define CONST_SHIFT     (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
+
+        alignz  16
+        global  EXTN(jconst_idct_ifast_sse2) PRIVATE
+
+EXTN(jconst_idct_ifast_sse2):
+
+PW_F1414        times 8 dw  F_1_414 << CONST_SHIFT
+PW_F1847        times 8 dw  F_1_847 << CONST_SHIFT
+PW_MF1613       times 8 dw -F_1_613 << CONST_SHIFT
+PW_F1082        times 8 dw  F_1_082 << CONST_SHIFT
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_ifast_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_ifast_sse2) PRIVATE
+
+EXTN(jsimd_idct_ifast_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_IFAST_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm7,xmm0             ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0             ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm7,xmm7             ; xmm7=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm6,xmm0,0x00          ; xmm6=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm2,xmm0,0x55          ; xmm2=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm5,xmm0,0xAA          ; xmm5=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm0,xmm0,0xFF          ; xmm0=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm1,xmm7,0x00          ; xmm1=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm4,xmm7,0x55          ; xmm4=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm3,xmm7,0xAA          ; xmm3=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm7,xmm7,0xFF          ; xmm7=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=col3
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm0
+        movdqa  xmm5,xmm1
+        psubw   xmm0,xmm2               ; xmm0=tmp11
+        psubw   xmm1,xmm3
+        paddw   xmm4,xmm2               ; xmm4=tmp10
+        paddw   xmm5,xmm3               ; xmm5=tmp13
+
+        psllw   xmm1,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm1,[GOTOFF(ebx,PW_F1414)]
+        psubw   xmm1,xmm5               ; xmm1=tmp12
+
+        movdqa  xmm6,xmm4
+        movdqa  xmm7,xmm0
+        psubw   xmm4,xmm5               ; xmm4=tmp3
+        psubw   xmm0,xmm1               ; xmm0=tmp2
+        paddw   xmm6,xmm5               ; xmm6=tmp0
+        paddw   xmm7,xmm1               ; xmm7=tmp1
+
+        movdqa  XMMWORD [wk(1)], xmm4   ; wk(1)=tmp3
+        movdqa  XMMWORD [wk(0)], xmm0   ; wk(0)=tmp2
+
+        ; -- Odd part
+
+        movdqa  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_IFAST_MULT_TYPE)]
+
+        movdqa  xmm4,xmm2
+        movdqa  xmm0,xmm5
+        psubw   xmm2,xmm1               ; xmm2=z12
+        psubw   xmm5,xmm3               ; xmm5=z10
+        paddw   xmm4,xmm1               ; xmm4=z11
+        paddw   xmm0,xmm3               ; xmm0=z13
+
+        movdqa  xmm1,xmm5               ; xmm1=z10(unscaled)
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm3,xmm4
+        psubw   xmm4,xmm0
+        paddw   xmm3,xmm0               ; xmm3=tmp7
+
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F1414)]     ; xmm4=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm0,xmm5
+        paddw   xmm5,xmm2
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F1847)]     ; xmm5=z5
+        pmulhw  xmm0,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F1082)]
+        psubw   xmm0,xmm1
+        psubw   xmm2,xmm5               ; xmm2=tmp10
+        paddw   xmm0,xmm5               ; xmm0=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm0,xmm3               ; xmm0=tmp6
+        movdqa  xmm1,xmm6
+        movdqa  xmm5,xmm7
+        paddw   xmm6,xmm3               ; xmm6=data0=(00 01 02 03 04 05 06 07)
+        paddw   xmm7,xmm0               ; xmm7=data1=(10 11 12 13 14 15 16 17)
+        psubw   xmm1,xmm3               ; xmm1=data7=(70 71 72 73 74 75 76 77)
+        psubw   xmm5,xmm0               ; xmm5=data6=(60 61 62 63 64 65 66 67)
+        psubw   xmm4,xmm0               ; xmm4=tmp5
+
+        movdqa    xmm3,xmm6             ; transpose coefficients(phase 1)
+        punpcklwd xmm6,xmm7             ; xmm6=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm3,xmm7             ; xmm3=(04 14 05 15 06 16 07 17)
+        movdqa    xmm0,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm1             ; xmm5=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm0,xmm1             ; xmm0=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp2
+        movdqa  xmm1, XMMWORD [wk(1)]   ; xmm1=tmp3
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(64 74 65 75 66 76 67 77)
+
+        paddw   xmm2,xmm4               ; xmm2=tmp4
+        movdqa  xmm5,xmm7
+        movdqa  xmm0,xmm1
+        paddw   xmm7,xmm4               ; xmm7=data2=(20 21 22 23 24 25 26 27)
+        paddw   xmm1,xmm2               ; xmm1=data4=(40 41 42 43 44 45 46 47)
+        psubw   xmm5,xmm4               ; xmm5=data5=(50 51 52 53 54 55 56 57)
+        psubw   xmm0,xmm2               ; xmm0=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa    xmm4,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm0             ; xmm7=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm0             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm2,xmm1             ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm5             ; xmm1=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm2,xmm5             ; xmm2=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm0,xmm3             ; transpose coefficients(phase 2)
+        punpckldq xmm3,xmm4             ; xmm3=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm0,xmm4             ; xmm0=(06 16 26 36 07 17 27 37)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7             ; xmm6=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm5,xmm7             ; xmm5=(02 12 22 32 03 13 23 33)
+
+        movdqa  xmm4, XMMWORD [wk(0)]   ; xmm4=(60 70 61 71 62 72 63 73)
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(0)], xmm3   ; wk(0)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm3,xmm1             ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm4             ; xmm1=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm3,xmm4             ; xmm3=(42 52 62 72 43 53 63 73)
+        movdqa    xmm0,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm7             ; xmm2=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm0,xmm7             ; xmm0=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm4,xmm6            ; transpose coefficients(phase 3)
+        punpcklqdq xmm6,xmm1            ; xmm6=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm4,xmm1            ; xmm4=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm7,xmm5            ; transpose coefficients(phase 3)
+        punpcklqdq xmm5,xmm3            ; xmm5=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm7,xmm3            ; xmm7=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm1, XMMWORD [wk(0)]   ; xmm1=(04 14 24 34 05 15 25 35)
+        movdqa  xmm3, XMMWORD [wk(1)]   ; xmm3=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=col1
+        movdqa  XMMWORD [wk(1)], xmm7   ; wk(1)=col3
+
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm4,xmm2            ; xmm4=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm7,xmm3            ; transpose coefficients(phase 3)
+        punpcklqdq xmm3,xmm0            ; xmm3=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm7,xmm0            ; xmm7=col7=(07 17 27 37 47 57 67 77)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        ; xmm6=col0, xmm5=col2, xmm1=col4, xmm3=col6
+
+        movdqa  xmm2,xmm6
+        movdqa  xmm0,xmm5
+        psubw   xmm6,xmm1               ; xmm6=tmp11
+        psubw   xmm5,xmm3
+        paddw   xmm2,xmm1               ; xmm2=tmp10
+        paddw   xmm0,xmm3               ; xmm0=tmp13
+
+        psllw   xmm5,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm5,[GOTOFF(ebx,PW_F1414)]
+        psubw   xmm5,xmm0               ; xmm5=tmp12
+
+        movdqa  xmm1,xmm2
+        movdqa  xmm3,xmm6
+        psubw   xmm2,xmm0               ; xmm2=tmp3
+        psubw   xmm6,xmm5               ; xmm6=tmp2
+        paddw   xmm1,xmm0               ; xmm1=tmp0
+        paddw   xmm3,xmm5               ; xmm3=tmp1
+
+        movdqa  xmm0, XMMWORD [wk(0)]   ; xmm0=col1
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=col3
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp3
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=tmp2
+
+        ; -- Odd part
+
+        ; xmm0=col1, xmm5=col3, xmm4=col5, xmm7=col7
+
+        movdqa  xmm2,xmm0
+        movdqa  xmm6,xmm4
+        psubw   xmm0,xmm7               ; xmm0=z12
+        psubw   xmm4,xmm5               ; xmm4=z10
+        paddw   xmm2,xmm7               ; xmm2=z11
+        paddw   xmm6,xmm5               ; xmm6=z13
+
+        movdqa  xmm7,xmm4               ; xmm7=z10(unscaled)
+        psllw   xmm0,PRE_MULTIPLY_SCALE_BITS
+        psllw   xmm4,PRE_MULTIPLY_SCALE_BITS
+
+        movdqa  xmm5,xmm2
+        psubw   xmm2,xmm6
+        paddw   xmm5,xmm6               ; xmm5=tmp7
+
+        psllw   xmm2,PRE_MULTIPLY_SCALE_BITS
+        pmulhw  xmm2,[GOTOFF(ebx,PW_F1414)]     ; xmm2=tmp11
+
+        ; To avoid overflow...
+        ;
+        ; (Original)
+        ; tmp12 = -2.613125930 * z10 + z5;
+        ;
+        ; (This implementation)
+        ; tmp12 = (-1.613125930 - 1) * z10 + z5;
+        ;       = -1.613125930 * z10 - z10 + z5;
+
+        movdqa  xmm6,xmm4
+        paddw   xmm4,xmm0
+        pmulhw  xmm4,[GOTOFF(ebx,PW_F1847)]     ; xmm4=z5
+        pmulhw  xmm6,[GOTOFF(ebx,PW_MF1613)]
+        pmulhw  xmm0,[GOTOFF(ebx,PW_F1082)]
+        psubw   xmm6,xmm7
+        psubw   xmm0,xmm4               ; xmm0=tmp10
+        paddw   xmm6,xmm4               ; xmm6=tmp12
+
+        ; -- Final output stage
+
+        psubw   xmm6,xmm5               ; xmm6=tmp6
+        movdqa  xmm7,xmm1
+        movdqa  xmm4,xmm3
+        paddw   xmm1,xmm5               ; xmm1=data0=(00 10 20 30 40 50 60 70)
+        paddw   xmm3,xmm6               ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        psraw   xmm1,(PASS1_BITS+3)     ; descale
+        psraw   xmm3,(PASS1_BITS+3)     ; descale
+        psubw   xmm7,xmm5               ; xmm7=data7=(07 17 27 37 47 57 67 77)
+        psubw   xmm4,xmm6               ; xmm4=data6=(06 16 26 36 46 56 66 76)
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psubw   xmm2,xmm6               ; xmm2=tmp5
+
+        packsswb  xmm1,xmm4     ; xmm1=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm7     ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm5, XMMWORD [wk(1)]   ; xmm5=tmp2
+        movdqa  xmm6, XMMWORD [wk(0)]   ; xmm6=tmp3
+
+        paddw   xmm0,xmm2               ; xmm0=tmp4
+        movdqa  xmm4,xmm5
+        movdqa  xmm7,xmm6
+        paddw   xmm5,xmm2               ; xmm5=data2=(02 12 22 32 42 52 62 72)
+        paddw   xmm6,xmm0               ; xmm6=data4=(04 14 24 34 44 54 64 74)
+        psraw   xmm5,(PASS1_BITS+3)     ; descale
+        psraw   xmm6,(PASS1_BITS+3)     ; descale
+        psubw   xmm4,xmm2               ; xmm4=data5=(05 15 25 35 45 55 65 75)
+        psubw   xmm7,xmm0               ; xmm7=data3=(03 13 23 33 43 53 63 73)
+        psraw   xmm4,(PASS1_BITS+3)     ; descale
+        psraw   xmm7,(PASS1_BITS+3)     ; descale
+
+        movdqa    xmm2,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm2=[PB_CENTERJSAMP]
+
+        packsswb  xmm5,xmm6     ; xmm5=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm7,xmm4     ; xmm7=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm1,xmm2
+        paddb     xmm3,xmm2
+        paddb     xmm5,xmm2
+        paddb     xmm7,xmm2
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 1)
+        punpcklbw xmm1,xmm3     ; xmm1=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm3     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm6,xmm5     ; transpose coefficients(phase 1)
+        punpcklbw xmm5,xmm7     ; xmm5=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm6,xmm7     ; xmm6=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm1     ; transpose coefficients(phase 2)
+        punpcklwd xmm1,xmm5     ; xmm1=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm5     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 2)
+        punpcklwd xmm6,xmm0     ; xmm6=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm2,xmm0     ; xmm2=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm3,xmm1     ; transpose coefficients(phase 3)
+        punpckldq xmm1,xmm6     ; xmm1=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm3,xmm6     ; xmm3=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm7,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm2     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm7,xmm2     ; xmm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm5,xmm1,0x4E  ; xmm5=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm3,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm6,xmm4,0x4E  ; xmm6=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm2,xmm7,0x4E  ; xmm2=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm1
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+        mov     edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm7
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm5
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0
+        mov     edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm2
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,359 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* SLOW INTEGER INVERSE DCT */
+
+#include "jsimd_altivec.h"
+
+
+#define F_0_298 2446   /* FIX(0.298631336) */
+#define F_0_390 3196   /* FIX(0.390180644) */
+#define F_0_541 4433   /* FIX(0.541196100) */
+#define F_0_765 6270   /* FIX(0.765366865) */
+#define F_0_899 7373   /* FIX(0.899976223) */
+#define F_1_175 9633   /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+#define DESCALE_P1 (CONST_BITS - PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS + PASS1_BITS + 3)
+
+
+#define DO_IDCT(in, PASS)  \
+{  \
+  /* Even part  \
+   *  \
+   * (Original)  \
+   * z1 = (z2 + z3) * 0.541196100;  \
+   * tmp2 = z1 + z3 * -1.847759065;  \
+   * tmp3 = z1 + z2 * 0.765366865;  \
+   *  \
+   * (This implementation)  \
+   * tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);  \
+   * tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;  \
+   */  \
+  \
+  in##26l = vec_mergeh(in##2, in##6);  \
+  in##26h = vec_mergel(in##2, in##6);  \
+  \
+  tmp3l = vec_msums(in##26l, pw_f130_f054, pd_zero);  \
+  tmp3h = vec_msums(in##26h, pw_f130_f054, pd_zero);  \
+  tmp2l = vec_msums(in##26l, pw_f054_mf130, pd_zero);  \
+  tmp2h = vec_msums(in##26h, pw_f054_mf130, pd_zero);  \
+  \
+  tmp0 = vec_add(in##0, in##4);  \
+  tmp1 = vec_sub(in##0, in##4);  \
+  \
+  tmp0l = vec_unpackh(tmp0);  \
+  tmp0h = vec_unpackl(tmp0);  \
+  tmp0l = vec_sl(tmp0l, const_bits);  \
+  tmp0h = vec_sl(tmp0h, const_bits);  \
+  tmp0l = vec_add(tmp0l, pd_descale_p##PASS);  \
+  tmp0h = vec_add(tmp0h, pd_descale_p##PASS);  \
+  \
+  tmp10l = vec_add(tmp0l, tmp3l);  \
+  tmp10h = vec_add(tmp0h, tmp3h);  \
+  tmp13l = vec_sub(tmp0l, tmp3l);  \
+  tmp13h = vec_sub(tmp0h, tmp3h);  \
+  \
+  tmp1l = vec_unpackh(tmp1);  \
+  tmp1h = vec_unpackl(tmp1);  \
+  tmp1l = vec_sl(tmp1l, const_bits);  \
+  tmp1h = vec_sl(tmp1h, const_bits);  \
+  tmp1l = vec_add(tmp1l, pd_descale_p##PASS);  \
+  tmp1h = vec_add(tmp1h, pd_descale_p##PASS);  \
+  \
+  tmp11l = vec_add(tmp1l, tmp2l);  \
+  tmp11h = vec_add(tmp1h, tmp2h);  \
+  tmp12l = vec_sub(tmp1l, tmp2l);  \
+  tmp12h = vec_sub(tmp1h, tmp2h);  \
+  \
+  /* Odd part */  \
+  \
+  z3 = vec_add(in##3, in##7);  \
+  z4 = vec_add(in##1, in##5);  \
+  \
+  /* (Original)  \
+   * z5 = (z3 + z4) * 1.175875602;  \
+   * z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;  \
+   * z3 += z5;  z4 += z5;  \
+   *  \
+   * (This implementation)  \
+   * z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;  \
+   * z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);  \
+   */  \
+  \
+  z34l = vec_mergeh(z3, z4);  \
+  z34h = vec_mergel(z3, z4);  \
+  \
+  z3l = vec_msums(z34l, pw_mf078_f117, pd_zero);  \
+  z3h = vec_msums(z34h, pw_mf078_f117, pd_zero);  \
+  z4l = vec_msums(z34l, pw_f117_f078, pd_zero);  \
+  z4h = vec_msums(z34h, pw_f117_f078, pd_zero);  \
+  \
+  /* (Original)  \
+   * z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;  \
+   * tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;  \
+   * tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;  \
+   * z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;  \
+   * tmp0 += z1 + z3;  tmp1 += z2 + z4;  \
+   * tmp2 += z2 + z3;  tmp3 += z1 + z4;  \
+   *  \
+   * (This implementation)  \
+   * tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;  \
+   * tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;  \
+   * tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);  \
+   * tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);  \
+   * tmp0 += z3;  tmp1 += z4;  \
+   * tmp2 += z3;  tmp3 += z4;  \
+   */  \
+  \
+  in##71l = vec_mergeh(in##7, in##1);  \
+  in##71h = vec_mergel(in##7, in##1);  \
+  \
+  tmp0l = vec_msums(in##71l, pw_mf060_mf089, z3l);  \
+  tmp0h = vec_msums(in##71h, pw_mf060_mf089, z3h);  \
+  tmp3l = vec_msums(in##71l, pw_mf089_f060, z4l);  \
+  tmp3h = vec_msums(in##71h, pw_mf089_f060, z4h);  \
+  \
+  in##53l = vec_mergeh(in##5, in##3);  \
+  in##53h = vec_mergel(in##5, in##3);  \
+  \
+  tmp1l = vec_msums(in##53l, pw_mf050_mf256, z4l);  \
+  tmp1h = vec_msums(in##53h, pw_mf050_mf256, z4h);  \
+  tmp2l = vec_msums(in##53l, pw_mf256_f050, z3l);  \
+  tmp2h = vec_msums(in##53h, pw_mf256_f050, z3h);  \
+  \
+  /* Final output stage */  \
+  \
+  out0l = vec_add(tmp10l, tmp3l);  \
+  out0h = vec_add(tmp10h, tmp3h);  \
+  out7l = vec_sub(tmp10l, tmp3l);  \
+  out7h = vec_sub(tmp10h, tmp3h);  \
+  \
+  out0l = vec_sra(out0l, descale_p##PASS);  \
+  out0h = vec_sra(out0h, descale_p##PASS);  \
+  out7l = vec_sra(out7l, descale_p##PASS);  \
+  out7h = vec_sra(out7h, descale_p##PASS);  \
+  \
+  out0 = vec_pack(out0l, out0h);  \
+  out7 = vec_pack(out7l, out7h);  \
+  \
+  out1l = vec_add(tmp11l, tmp2l);  \
+  out1h = vec_add(tmp11h, tmp2h);  \
+  out6l = vec_sub(tmp11l, tmp2l);  \
+  out6h = vec_sub(tmp11h, tmp2h);  \
+  \
+  out1l = vec_sra(out1l, descale_p##PASS);  \
+  out1h = vec_sra(out1h, descale_p##PASS);  \
+  out6l = vec_sra(out6l, descale_p##PASS);  \
+  out6h = vec_sra(out6h, descale_p##PASS);  \
+  \
+  out1 = vec_pack(out1l, out1h);  \
+  out6 = vec_pack(out6l, out6h);  \
+  \
+  out2l = vec_add(tmp12l, tmp1l);  \
+  out2h = vec_add(tmp12h, tmp1h);  \
+  out5l = vec_sub(tmp12l, tmp1l);  \
+  out5h = vec_sub(tmp12h, tmp1h);  \
+  \
+  out2l = vec_sra(out2l, descale_p##PASS);  \
+  out2h = vec_sra(out2h, descale_p##PASS);  \
+  out5l = vec_sra(out5l, descale_p##PASS);  \
+  out5h = vec_sra(out5h, descale_p##PASS);  \
+  \
+  out2 = vec_pack(out2l, out2h);  \
+  out5 = vec_pack(out5l, out5h);  \
+  \
+  out3l = vec_add(tmp13l, tmp0l);  \
+  out3h = vec_add(tmp13h, tmp0h);  \
+  out4l = vec_sub(tmp13l, tmp0l);  \
+  out4h = vec_sub(tmp13h, tmp0h);  \
+  \
+  out3l = vec_sra(out3l, descale_p##PASS);  \
+  out3h = vec_sra(out3h, descale_p##PASS);  \
+  out4l = vec_sra(out4l, descale_p##PASS);  \
+  out4h = vec_sra(out4h, descale_p##PASS);  \
+  \
+  out3 = vec_pack(out3l, out3h);  \
+  out4 = vec_pack(out4l, out4h);  \
+}
+
+
+void
+jsimd_idct_islow_altivec (void *dct_table_, JCOEFPTR coef_block,
+                          JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  short *dct_table = (short *)dct_table_;
+  int *outptr;
+
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    col0, col1, col2, col3, col4, col5, col6, col7,
+    quant0, quant1, quant2, quant3, quant4, quant5, quant6, quant7,
+    tmp0, tmp1, tmp2, tmp3, z3, z4,
+    z34l, z34h, col71l, col71h, col26l, col26h, col53l, col53h,
+    row71l, row71h, row26l, row26h, row53l, row53h,
+    out0, out1, out2, out3, out4, out5, out6, out7;
+  __vector int tmp0l, tmp0h, tmp1l, tmp1h, tmp2l, tmp2h, tmp3l, tmp3h,
+    tmp10l, tmp10h, tmp11l, tmp11h, tmp12l, tmp12h, tmp13l, tmp13h,
+    z3l, z3h, z4l, z4h,
+    out0l, out0h, out1l, out1h, out2l, out2h, out3l, out3h, out4l, out4h,
+    out5l, out5h, out6l, out6h, out7l, out7h;
+  __vector signed char outb;
+
+  /* Constants */
+  __vector short pw_zero = { __8X(0) },
+    pw_f130_f054 = { __4X2(F_0_541 + F_0_765, F_0_541) },
+    pw_f054_mf130 = { __4X2(F_0_541, F_0_541 - F_1_847) },
+    pw_mf078_f117 = { __4X2(F_1_175 - F_1_961, F_1_175) },
+    pw_f117_f078 = { __4X2(F_1_175, F_1_175 - F_0_390) },
+    pw_mf060_mf089 = { __4X2(F_0_298 - F_0_899, -F_0_899) },
+    pw_mf089_f060 = { __4X2(-F_0_899, F_1_501 - F_0_899) },
+    pw_mf050_mf256 = { __4X2(F_2_053 - F_2_562, -F_2_562) },
+    pw_mf256_f050 = { __4X2(-F_2_562, F_3_072 - F_2_562) };
+  __vector unsigned short pass1_bits = { __8X(PASS1_BITS) };
+  __vector int pd_zero = { __4X(0) },
+    pd_descale_p1 = { __4X(1 << (DESCALE_P1 - 1)) },
+    pd_descale_p2 = { __4X(1 << (DESCALE_P2 - 1)) };
+  __vector unsigned int descale_p1 = { __4X(DESCALE_P1) },
+    descale_p2 = { __4X(DESCALE_P2) },
+    const_bits = { __4X(CONST_BITS) };
+  __vector signed char pb_centerjsamp = { __16X(CENTERJSAMPLE) };
+
+  /* Pass 1: process columns */
+
+  col0 = vec_ld(0, coef_block);
+  col1 = vec_ld(16, coef_block);
+  col2 = vec_ld(32, coef_block);
+  col3 = vec_ld(48, coef_block);
+  col4 = vec_ld(64, coef_block);
+  col5 = vec_ld(80, coef_block);
+  col6 = vec_ld(96, coef_block);
+  col7 = vec_ld(112, coef_block);
+
+  tmp1 = vec_or(col1, col2);
+  tmp2 = vec_or(col3, col4);
+  tmp1 = vec_or(tmp1, tmp2);
+  tmp3 = vec_or(col5, col6);
+  tmp3 = vec_or(tmp3, col7);
+  tmp1 = vec_or(tmp1, tmp3);
+
+  quant0 = vec_ld(0, dct_table);
+  col0 = vec_mladd(col0, quant0, pw_zero);
+
+  if (vec_all_eq(tmp1, pw_zero)) {
+    /* AC terms all zero */
+
+    col0 = vec_sl(col0, pass1_bits);
+
+    row0 = vec_splat(col0, 0);
+    row1 = vec_splat(col0, 1);
+    row2 = vec_splat(col0, 2);
+    row3 = vec_splat(col0, 3);
+    row4 = vec_splat(col0, 4);
+    row5 = vec_splat(col0, 5);
+    row6 = vec_splat(col0, 6);
+    row7 = vec_splat(col0, 7);
+
+  } else {
+
+    quant1 = vec_ld(16, dct_table);
+    quant2 = vec_ld(32, dct_table);
+    quant3 = vec_ld(48, dct_table);
+    quant4 = vec_ld(64, dct_table);
+    quant5 = vec_ld(80, dct_table);
+    quant6 = vec_ld(96, dct_table);
+    quant7 = vec_ld(112, dct_table);
+
+    col1 = vec_mladd(col1, quant1, pw_zero);
+    col2 = vec_mladd(col2, quant2, pw_zero);
+    col3 = vec_mladd(col3, quant3, pw_zero);
+    col4 = vec_mladd(col4, quant4, pw_zero);
+    col5 = vec_mladd(col5, quant5, pw_zero);
+    col6 = vec_mladd(col6, quant6, pw_zero);
+    col7 = vec_mladd(col7, quant7, pw_zero);
+
+    DO_IDCT(col, 1);
+
+    TRANSPOSE(out, row);
+  }
+
+  /* Pass 2: process rows */
+
+  DO_IDCT(row, 2);
+
+  TRANSPOSE(out, col);
+
+  outb = vec_packs(col0, col0);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[0] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col1, col1);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[1] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col2, col2);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[2] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col3, col3);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[3] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col4, col4);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[4] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col5, col5);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[5] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col6, col6);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[6] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+
+  outb = vec_packs(col7, col7);
+  outb = vec_add(outb, pb_centerjsamp);
+  outptr = (int *)(output_buf[7] + output_col);
+  vec_ste((__vector int)outb, 0, outptr);
+  vec_ste((__vector int)outb, 4, outptr);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,852 @@
+;
+; jidctint.asm - accurate integer IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_mmx) PRIVATE
+
+EXTN(jconst_idct_islow_mmx):
+
+PW_F130_F054    times 2 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 2 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 2 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 2 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 2 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 2 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 2 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 2 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 2 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 2 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_mmx (void *dct_table, JCOEFPTR coef_block,
+;                       JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          12
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_islow_mmx) PRIVATE
+
+EXTN(jsimd_idct_islow_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm1,mm0
+        packsswb mm1,mm1
+        movd    eax,mm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   mm0,PASS1_BITS
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movq      mm4,mm1               ; mm1=in2=z2
+        movq      mm5,mm1
+        punpcklwd mm4,mm3               ; mm3=in6=z3
+        punpckhwd mm5,mm3
+        movq      mm1,mm4
+        movq      mm3,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=tmp3L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F130_F054)]        ; mm5=tmp3H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F054_MF130)]       ; mm3=tmp2H
+
+        movq      mm6,mm0
+        paddw     mm0,mm2               ; mm0=in0+in4
+        psubw     mm6,mm2               ; mm6=in0-in4
+
+        pxor      mm7,mm7
+        pxor      mm2,mm2
+        punpcklwd mm7,mm0               ; mm7=tmp0L
+        punpckhwd mm2,mm0               ; mm2=tmp0H
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+        psrad     mm2,(16-CONST_BITS)   ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+        movq    mm0,mm7
+        paddd   mm7,mm4                 ; mm7=tmp10L
+        psubd   mm0,mm4                 ; mm0=tmp13L
+        movq    mm4,mm2
+        paddd   mm2,mm5                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp13H
+
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp10L
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=tmp10H
+        movq    MMWORD [wk(2)], mm0     ; wk(2)=tmp13L
+        movq    MMWORD [wk(3)], mm4     ; wk(3)=tmp13H
+
+        pxor      mm5,mm5
+        pxor      mm7,mm7
+        punpcklwd mm5,mm6               ; mm5=tmp1L
+        punpckhwd mm7,mm6               ; mm7=tmp1H
+        psrad     mm5,(16-CONST_BITS)   ; psrad mm5,16 & pslld mm5,CONST_BITS
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+        movq    mm2,mm5
+        paddd   mm5,mm1                 ; mm5=tmp11L
+        psubd   mm2,mm1                 ; mm2=tmp12L
+        movq    mm0,mm7
+        paddd   mm7,mm3                 ; mm7=tmp11H
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        movq    MMWORD [wk(4)], mm5     ; wk(4)=tmp11L
+        movq    MMWORD [wk(5)], mm7     ; wk(5)=tmp11H
+        movq    MMWORD [wk(6)], mm2     ; wk(6)=tmp12L
+        movq    MMWORD [wk(7)], mm0     ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movq    mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm4, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm6, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm1, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movq    mm5,mm6
+        movq    mm7,mm4
+        paddw   mm5,mm3                 ; mm5=z3
+        paddw   mm7,mm1                 ; mm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm2,mm5
+        movq      mm0,mm5
+        punpcklwd mm2,mm7
+        punpckhwd mm0,mm7
+        movq      mm5,mm2
+        movq      mm7,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF078_F117)]       ; mm2=z3L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF078_F117)]       ; mm0=z3H
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F117_F078)]        ; mm5=z4L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_F117_F078)]        ; mm7=z4H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=z3L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movq      mm2,mm3
+        movq      mm0,mm3
+        punpcklwd mm2,mm4
+        punpckhwd mm0,mm4
+        movq      mm3,mm2
+        movq      mm4,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm2=tmp0L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm0=tmp0H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF089_F060)]       ; mm3=tmp3L
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF089_F060)]       ; mm4=tmp3H
+
+        paddd   mm2, MMWORD [wk(10)]    ; mm2=tmp0L
+        paddd   mm0, MMWORD [wk(11)]    ; mm0=tmp0H
+        paddd   mm3,mm5                 ; mm3=tmp3L
+        paddd   mm4,mm7                 ; mm4=tmp3H
+
+        movq    MMWORD [wk(8)], mm2     ; wk(8)=tmp0L
+        movq    MMWORD [wk(9)], mm0     ; wk(9)=tmp0H
+
+        movq      mm2,mm1
+        movq      mm0,mm1
+        punpcklwd mm2,mm6
+        punpckhwd mm0,mm6
+        movq      mm1,mm2
+        movq      mm6,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm2=tmp1L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm0=tmp1H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF256_F050)]       ; mm1=tmp2L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF256_F050)]       ; mm6=tmp2H
+
+        paddd   mm2,mm5                 ; mm2=tmp1L
+        paddd   mm0,mm7                 ; mm0=tmp1H
+        paddd   mm1, MMWORD [wk(10)]    ; mm1=tmp2L
+        paddd   mm6, MMWORD [wk(11)]    ; mm6=tmp2H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=tmp1L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp10L
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp10H
+
+        movq    mm2,mm5
+        movq    mm0,mm7
+        paddd   mm5,mm3                 ; mm5=data0L
+        paddd   mm7,mm4                 ; mm7=data0H
+        psubd   mm2,mm3                 ; mm2=data7L
+        psubd   mm0,mm4                 ; mm0=data7H
+
+        movq    mm3,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm3=[PD_DESCALE_P1]
+
+        paddd   mm5,mm3
+        paddd   mm7,mm3
+        psrad   mm5,DESCALE_P1
+        psrad   mm7,DESCALE_P1
+        paddd   mm2,mm3
+        paddd   mm0,mm3
+        psrad   mm2,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+
+        packssdw  mm5,mm7               ; mm5=data0=(00 01 02 03)
+        packssdw  mm2,mm0               ; mm2=data7=(70 71 72 73)
+
+        movq    mm4, MMWORD [wk(4)]     ; mm4=tmp11L
+        movq    mm3, MMWORD [wk(5)]     ; mm3=tmp11H
+
+        movq    mm7,mm4
+        movq    mm0,mm3
+        paddd   mm4,mm1                 ; mm4=data1L
+        paddd   mm3,mm6                 ; mm3=data1H
+        psubd   mm7,mm1                 ; mm7=data6L
+        psubd   mm0,mm6                 ; mm0=data6H
+
+        movq    mm1,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm1=[PD_DESCALE_P1]
+
+        paddd   mm4,mm1
+        paddd   mm3,mm1
+        psrad   mm4,DESCALE_P1
+        psrad   mm3,DESCALE_P1
+        paddd   mm7,mm1
+        paddd   mm0,mm1
+        psrad   mm7,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+
+        packssdw  mm4,mm3               ; mm4=data1=(10 11 12 13)
+        packssdw  mm7,mm0               ; mm7=data6=(60 61 62 63)
+
+        movq      mm6,mm5               ; transpose coefficients(phase 1)
+        punpcklwd mm5,mm4               ; mm5=(00 10 01 11)
+        punpckhwd mm6,mm4               ; mm6=(02 12 03 13)
+        movq      mm1,mm7               ; transpose coefficients(phase 1)
+        punpcklwd mm7,mm2               ; mm7=(60 70 61 71)
+        punpckhwd mm1,mm2               ; mm1=(62 72 63 73)
+
+        movq    mm3, MMWORD [wk(6)]     ; mm3=tmp12L
+        movq    mm0, MMWORD [wk(7)]     ; mm0=tmp12H
+        movq    mm4, MMWORD [wk(10)]    ; mm4=tmp1L
+        movq    mm2, MMWORD [wk(11)]    ; mm2=tmp1H
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(00 10 01 11)
+        movq    MMWORD [wk(1)], mm6     ; wk(1)=(02 12 03 13)
+        movq    MMWORD [wk(4)], mm7     ; wk(4)=(60 70 61 71)
+        movq    MMWORD [wk(5)], mm1     ; wk(5)=(62 72 63 73)
+
+        movq    mm5,mm3
+        movq    mm6,mm0
+        paddd   mm3,mm4                 ; mm3=data2L
+        paddd   mm0,mm2                 ; mm0=data2H
+        psubd   mm5,mm4                 ; mm5=data5L
+        psubd   mm6,mm2                 ; mm6=data5H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm7=[PD_DESCALE_P1]
+
+        paddd   mm3,mm7
+        paddd   mm0,mm7
+        psrad   mm3,DESCALE_P1
+        psrad   mm0,DESCALE_P1
+        paddd   mm5,mm7
+        paddd   mm6,mm7
+        psrad   mm5,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm3,mm0               ; mm3=data2=(20 21 22 23)
+        packssdw  mm5,mm6               ; mm5=data5=(50 51 52 53)
+
+        movq    mm1, MMWORD [wk(2)]     ; mm1=tmp13L
+        movq    mm4, MMWORD [wk(3)]     ; mm4=tmp13H
+        movq    mm2, MMWORD [wk(8)]     ; mm2=tmp0L
+        movq    mm7, MMWORD [wk(9)]     ; mm7=tmp0H
+
+        movq    mm0,mm1
+        movq    mm6,mm4
+        paddd   mm1,mm2                 ; mm1=data3L
+        paddd   mm4,mm7                 ; mm4=data3H
+        psubd   mm0,mm2                 ; mm0=data4L
+        psubd   mm6,mm7                 ; mm6=data4H
+
+        movq    mm2,[GOTOFF(ebx,PD_DESCALE_P1)] ; mm2=[PD_DESCALE_P1]
+
+        paddd   mm1,mm2
+        paddd   mm4,mm2
+        psrad   mm1,DESCALE_P1
+        psrad   mm4,DESCALE_P1
+        paddd   mm0,mm2
+        paddd   mm6,mm2
+        psrad   mm0,DESCALE_P1
+        psrad   mm6,DESCALE_P1
+
+        packssdw  mm1,mm4               ; mm1=data3=(30 31 32 33)
+        packssdw  mm0,mm6               ; mm0=data4=(40 41 42 43)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=(00 10 01 11)
+        movq    mm2, MMWORD [wk(1)]     ; mm2=(02 12 03 13)
+
+        movq      mm4,mm3               ; transpose coefficients(phase 1)
+        punpcklwd mm3,mm1               ; mm3=(20 30 21 31)
+        punpckhwd mm4,mm1               ; mm4=(22 32 23 33)
+        movq      mm6,mm0               ; transpose coefficients(phase 1)
+        punpcklwd mm0,mm5               ; mm0=(40 50 41 51)
+        punpckhwd mm6,mm5               ; mm6=(42 52 43 53)
+
+        movq      mm1,mm7               ; transpose coefficients(phase 2)
+        punpckldq mm7,mm3               ; mm7=(00 10 20 30)
+        punpckhdq mm1,mm3               ; mm1=(01 11 21 31)
+        movq      mm5,mm2               ; transpose coefficients(phase 2)
+        punpckldq mm2,mm4               ; mm2=(02 12 22 32)
+        punpckhdq mm5,mm4               ; mm5=(03 13 23 33)
+
+        movq    mm3, MMWORD [wk(4)]     ; mm3=(60 70 61 71)
+        movq    mm4, MMWORD [wk(5)]     ; mm4=(62 72 63 73)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm7
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm5
+
+        movq      mm7,mm0               ; transpose coefficients(phase 2)
+        punpckldq mm0,mm3               ; mm0=(40 50 60 70)
+        punpckhdq mm7,mm3               ; mm7=(41 51 61 71)
+        movq      mm1,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm4               ; mm6=(42 52 62 72)
+        punpckhdq mm1,mm4               ; mm1=(43 53 63 73)
+
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm7
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_JCOEF)], mm1
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_ISLOW_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.rowloop:
+
+        ; -- Even part
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movq      mm4,mm1               ; mm1=in2=z2
+        movq      mm5,mm1
+        punpcklwd mm4,mm3               ; mm3=in6=z3
+        punpckhwd mm5,mm3
+        movq      mm1,mm4
+        movq      mm3,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F130_F054)]        ; mm4=tmp3L
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F130_F054)]        ; mm5=tmp3H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F054_MF130)]       ; mm1=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F054_MF130)]       ; mm3=tmp2H
+
+        movq      mm6,mm0
+        paddw     mm0,mm2               ; mm0=in0+in4
+        psubw     mm6,mm2               ; mm6=in0-in4
+
+        pxor      mm7,mm7
+        pxor      mm2,mm2
+        punpcklwd mm7,mm0               ; mm7=tmp0L
+        punpckhwd mm2,mm0               ; mm2=tmp0H
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+        psrad     mm2,(16-CONST_BITS)   ; psrad mm2,16 & pslld mm2,CONST_BITS
+
+        movq    mm0,mm7
+        paddd   mm7,mm4                 ; mm7=tmp10L
+        psubd   mm0,mm4                 ; mm0=tmp13L
+        movq    mm4,mm2
+        paddd   mm2,mm5                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp13H
+
+        movq    MMWORD [wk(0)], mm7     ; wk(0)=tmp10L
+        movq    MMWORD [wk(1)], mm2     ; wk(1)=tmp10H
+        movq    MMWORD [wk(2)], mm0     ; wk(2)=tmp13L
+        movq    MMWORD [wk(3)], mm4     ; wk(3)=tmp13H
+
+        pxor      mm5,mm5
+        pxor      mm7,mm7
+        punpcklwd mm5,mm6               ; mm5=tmp1L
+        punpckhwd mm7,mm6               ; mm7=tmp1H
+        psrad     mm5,(16-CONST_BITS)   ; psrad mm5,16 & pslld mm5,CONST_BITS
+        psrad     mm7,(16-CONST_BITS)   ; psrad mm7,16 & pslld mm7,CONST_BITS
+
+        movq    mm2,mm5
+        paddd   mm5,mm1                 ; mm5=tmp11L
+        psubd   mm2,mm1                 ; mm2=tmp12L
+        movq    mm0,mm7
+        paddd   mm7,mm3                 ; mm7=tmp11H
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        movq    MMWORD [wk(4)], mm5     ; wk(4)=tmp11L
+        movq    MMWORD [wk(5)], mm7     ; wk(5)=tmp11H
+        movq    MMWORD [wk(6)], mm2     ; wk(6)=tmp12L
+        movq    MMWORD [wk(7)], mm0     ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movq    mm4, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm6, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq    mm5,mm6
+        movq    mm7,mm4
+        paddw   mm5,mm3                 ; mm5=z3
+        paddw   mm7,mm1                 ; mm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movq      mm2,mm5
+        movq      mm0,mm5
+        punpcklwd mm2,mm7
+        punpckhwd mm0,mm7
+        movq      mm5,mm2
+        movq      mm7,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF078_F117)]       ; mm2=z3L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF078_F117)]       ; mm0=z3H
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F117_F078)]        ; mm5=z4L
+        pmaddwd   mm7,[GOTOFF(ebx,PW_F117_F078)]        ; mm7=z4H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=z3L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movq      mm2,mm3
+        movq      mm0,mm3
+        punpcklwd mm2,mm4
+        punpckhwd mm0,mm4
+        movq      mm3,mm2
+        movq      mm4,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm2=tmp0L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF060_MF089)]      ; mm0=tmp0H
+        pmaddwd   mm3,[GOTOFF(ebx,PW_MF089_F060)]       ; mm3=tmp3L
+        pmaddwd   mm4,[GOTOFF(ebx,PW_MF089_F060)]       ; mm4=tmp3H
+
+        paddd   mm2, MMWORD [wk(10)]    ; mm2=tmp0L
+        paddd   mm0, MMWORD [wk(11)]    ; mm0=tmp0H
+        paddd   mm3,mm5                 ; mm3=tmp3L
+        paddd   mm4,mm7                 ; mm4=tmp3H
+
+        movq    MMWORD [wk(8)], mm2     ; wk(8)=tmp0L
+        movq    MMWORD [wk(9)], mm0     ; wk(9)=tmp0H
+
+        movq      mm2,mm1
+        movq      mm0,mm1
+        punpcklwd mm2,mm6
+        punpckhwd mm0,mm6
+        movq      mm1,mm2
+        movq      mm6,mm0
+        pmaddwd   mm2,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm2=tmp1L
+        pmaddwd   mm0,[GOTOFF(ebx,PW_MF050_MF256)]      ; mm0=tmp1H
+        pmaddwd   mm1,[GOTOFF(ebx,PW_MF256_F050)]       ; mm1=tmp2L
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF256_F050)]       ; mm6=tmp2H
+
+        paddd   mm2,mm5                 ; mm2=tmp1L
+        paddd   mm0,mm7                 ; mm0=tmp1H
+        paddd   mm1, MMWORD [wk(10)]    ; mm1=tmp2L
+        paddd   mm6, MMWORD [wk(11)]    ; mm6=tmp2H
+
+        movq    MMWORD [wk(10)], mm2    ; wk(10)=tmp1L
+        movq    MMWORD [wk(11)], mm0    ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movq    mm5, MMWORD [wk(0)]     ; mm5=tmp10L
+        movq    mm7, MMWORD [wk(1)]     ; mm7=tmp10H
+
+        movq    mm2,mm5
+        movq    mm0,mm7
+        paddd   mm5,mm3                 ; mm5=data0L
+        paddd   mm7,mm4                 ; mm7=data0H
+        psubd   mm2,mm3                 ; mm2=data7L
+        psubd   mm0,mm4                 ; mm0=data7H
+
+        movq    mm3,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm3=[PD_DESCALE_P2]
+
+        paddd   mm5,mm3
+        paddd   mm7,mm3
+        psrad   mm5,DESCALE_P2
+        psrad   mm7,DESCALE_P2
+        paddd   mm2,mm3
+        paddd   mm0,mm3
+        psrad   mm2,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+
+        packssdw  mm5,mm7               ; mm5=data0=(00 10 20 30)
+        packssdw  mm2,mm0               ; mm2=data7=(07 17 27 37)
+
+        movq    mm4, MMWORD [wk(4)]     ; mm4=tmp11L
+        movq    mm3, MMWORD [wk(5)]     ; mm3=tmp11H
+
+        movq    mm7,mm4
+        movq    mm0,mm3
+        paddd   mm4,mm1                 ; mm4=data1L
+        paddd   mm3,mm6                 ; mm3=data1H
+        psubd   mm7,mm1                 ; mm7=data6L
+        psubd   mm0,mm6                 ; mm0=data6H
+
+        movq    mm1,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm1=[PD_DESCALE_P2]
+
+        paddd   mm4,mm1
+        paddd   mm3,mm1
+        psrad   mm4,DESCALE_P2
+        psrad   mm3,DESCALE_P2
+        paddd   mm7,mm1
+        paddd   mm0,mm1
+        psrad   mm7,DESCALE_P2
+        psrad   mm0,DESCALE_P2
+
+        packssdw  mm4,mm3               ; mm4=data1=(01 11 21 31)
+        packssdw  mm7,mm0               ; mm7=data6=(06 16 26 36)
+
+        packsswb  mm5,mm7               ; mm5=(00 10 20 30 06 16 26 36)
+        packsswb  mm4,mm2               ; mm4=(01 11 21 31 07 17 27 37)
+
+        movq    mm6, MMWORD [wk(6)]     ; mm6=tmp12L
+        movq    mm1, MMWORD [wk(7)]     ; mm1=tmp12H
+        movq    mm3, MMWORD [wk(10)]    ; mm3=tmp1L
+        movq    mm0, MMWORD [wk(11)]    ; mm0=tmp1H
+
+        movq    MMWORD [wk(0)], mm5     ; wk(0)=(00 10 20 30 06 16 26 36)
+        movq    MMWORD [wk(1)], mm4     ; wk(1)=(01 11 21 31 07 17 27 37)
+
+        movq    mm7,mm6
+        movq    mm2,mm1
+        paddd   mm6,mm3                 ; mm6=data2L
+        paddd   mm1,mm0                 ; mm1=data2H
+        psubd   mm7,mm3                 ; mm7=data5L
+        psubd   mm2,mm0                 ; mm2=data5H
+
+        movq    mm5,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm5=[PD_DESCALE_P2]
+
+        paddd   mm6,mm5
+        paddd   mm1,mm5
+        psrad   mm6,DESCALE_P2
+        psrad   mm1,DESCALE_P2
+        paddd   mm7,mm5
+        paddd   mm2,mm5
+        psrad   mm7,DESCALE_P2
+        psrad   mm2,DESCALE_P2
+
+        packssdw  mm6,mm1               ; mm6=data2=(02 12 22 32)
+        packssdw  mm7,mm2               ; mm7=data5=(05 15 25 35)
+
+        movq    mm4, MMWORD [wk(2)]     ; mm4=tmp13L
+        movq    mm3, MMWORD [wk(3)]     ; mm3=tmp13H
+        movq    mm0, MMWORD [wk(8)]     ; mm0=tmp0L
+        movq    mm5, MMWORD [wk(9)]     ; mm5=tmp0H
+
+        movq    mm1,mm4
+        movq    mm2,mm3
+        paddd   mm4,mm0                 ; mm4=data3L
+        paddd   mm3,mm5                 ; mm3=data3H
+        psubd   mm1,mm0                 ; mm1=data4L
+        psubd   mm2,mm5                 ; mm2=data4H
+
+        movq    mm0,[GOTOFF(ebx,PD_DESCALE_P2)] ; mm0=[PD_DESCALE_P2]
+
+        paddd   mm4,mm0
+        paddd   mm3,mm0
+        psrad   mm4,DESCALE_P2
+        psrad   mm3,DESCALE_P2
+        paddd   mm1,mm0
+        paddd   mm2,mm0
+        psrad   mm1,DESCALE_P2
+        psrad   mm2,DESCALE_P2
+
+        movq      mm5,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm5=[PB_CENTERJSAMP]
+
+        packssdw  mm4,mm3               ; mm4=data3=(03 13 23 33)
+        packssdw  mm1,mm2               ; mm1=data4=(04 14 24 34)
+
+        movq      mm0, MMWORD [wk(0)]   ; mm0=(00 10 20 30 06 16 26 36)
+        movq      mm3, MMWORD [wk(1)]   ; mm3=(01 11 21 31 07 17 27 37)
+
+        packsswb  mm6,mm1               ; mm6=(02 12 22 32 04 14 24 34)
+        packsswb  mm4,mm7               ; mm4=(03 13 23 33 05 15 25 35)
+
+        paddb     mm0,mm5
+        paddb     mm3,mm5
+        paddb     mm6,mm5
+        paddb     mm4,mm5
+
+        movq      mm2,mm0               ; transpose coefficients(phase 1)
+        punpcklbw mm0,mm3               ; mm0=(00 01 10 11 20 21 30 31)
+        punpckhbw mm2,mm3               ; mm2=(06 07 16 17 26 27 36 37)
+        movq      mm1,mm6               ; transpose coefficients(phase 1)
+        punpcklbw mm6,mm4               ; mm6=(02 03 12 13 22 23 32 33)
+        punpckhbw mm1,mm4               ; mm1=(04 05 14 15 24 25 34 35)
+
+        movq      mm7,mm0               ; transpose coefficients(phase 2)
+        punpcklwd mm0,mm6               ; mm0=(00 01 02 03 10 11 12 13)
+        punpckhwd mm7,mm6               ; mm7=(20 21 22 23 30 31 32 33)
+        movq      mm5,mm1               ; transpose coefficients(phase 2)
+        punpcklwd mm1,mm2               ; mm1=(04 05 06 07 14 15 16 17)
+        punpckhwd mm5,mm2               ; mm5=(24 25 26 27 34 35 36 37)
+
+        movq      mm3,mm0               ; transpose coefficients(phase 3)
+        punpckldq mm0,mm1               ; mm0=(00 01 02 03 04 05 06 07)
+        punpckhdq mm3,mm1               ; mm3=(10 11 12 13 14 15 16 17)
+        movq      mm4,mm7               ; transpose coefficients(phase 3)
+        punpckldq mm7,mm5               ; mm7=(20 21 22 23 24 25 26 27)
+        punpckhdq mm4,mm5               ; mm4=(30 31 32 33 34 35 36 37)
+
+        pushpic ebx                     ; save GOT address
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm0
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm3
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     ebx, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    MMWORD [edx+eax*SIZEOF_JSAMPLE], mm7
+        movq    MMWORD [ebx+eax*SIZEOF_JSAMPLE], mm4
+
+        poppic  ebx                     ; restore GOT address
+
+        add     esi, byte 4*SIZEOF_JCOEF        ; wsptr
+        add     edi, byte 4*SIZEOF_JSAMPROW
+        dec     ecx                             ; ctr
+        jnz     near .rowloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,848 @@
+;
+; jidctint.asm - accurate integer IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_sse2) PRIVATE
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = jpeg_component_info *compptr
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          12
+
+        align   16
+        global  EXTN(jsimd_idct_islow_sse2) PRIVATE
+
+EXTN(jsimd_idct_islow_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm5, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm5,PASS1_BITS
+
+        movdqa    xmm4,xmm5             ; xmm5=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm5,xmm5             ; xmm5=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm4,xmm4             ; xmm4=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm7,xmm5,0x00          ; xmm7=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm6,xmm5,0x55          ; xmm6=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm1,xmm5,0xAA          ; xmm1=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm5,xmm5,0xFF          ; xmm5=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm0,xmm4,0x00          ; xmm0=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm3,xmm4,0x55          ; xmm3=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm2,xmm4,0xAA          ; xmm2=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm4,xmm4,0xFF          ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(8)], xmm6   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm5   ; wk(9)=col3
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm4,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm4,xmm3             ; xmm3=in6=z3
+        punpckhwd xmm5,xmm3
+        movdqa    xmm1,xmm4
+        movdqa    xmm3,xmm5
+        pmaddwd   xmm4,[rel PW_F130_F054]       ; xmm4=tmp3L
+        pmaddwd   xmm5,[rel PW_F130_F054]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=tmp2L
+        pmaddwd   xmm3,[rel PW_F054_MF130]      ; xmm3=tmp2H
+
+        movdqa    xmm6,xmm0
+        paddw     xmm0,xmm2             ; xmm0=in0+in4
+        psubw     xmm6,xmm2             ; xmm6=in0-in4
+
+        pxor      xmm7,xmm7
+        pxor      xmm2,xmm2
+        punpcklwd xmm7,xmm0             ; xmm7=tmp0L
+        punpckhwd xmm2,xmm0             ; xmm2=tmp0H
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+        psrad     xmm2,(16-CONST_BITS)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm4               ; xmm7=tmp10L
+        psubd   xmm0,xmm4               ; xmm0=tmp13L
+        movdqa  xmm4,xmm2
+        paddd   xmm2,xmm5               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm0   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm4   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm7,xmm7
+        punpcklwd xmm5,xmm6             ; xmm5=tmp1L
+        punpckhwd xmm7,xmm6             ; xmm7=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+        movdqa  xmm2,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm2,xmm1               ; xmm2=tmp12L
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm3               ; xmm7=tmp11H
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm7   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm0   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm7,xmm4
+        paddw   xmm5,xmm3               ; xmm5=z3
+        paddw   xmm7,xmm1               ; xmm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm2,xmm5
+        movdqa    xmm0,xmm5
+        punpcklwd xmm2,xmm7
+        punpckhwd xmm0,xmm7
+        movdqa    xmm5,xmm2
+        movdqa    xmm7,xmm0
+        pmaddwd   xmm2,[rel PW_MF078_F117]      ; xmm2=z3L
+        pmaddwd   xmm0,[rel PW_MF078_F117]      ; xmm0=z3H
+        pmaddwd   xmm5,[rel PW_F117_F078]       ; xmm5=z4L
+        pmaddwd   xmm7,[rel PW_F117_F078]       ; xmm7=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm2,xmm3
+        movdqa    xmm0,xmm3
+        punpcklwd xmm2,xmm4
+        punpckhwd xmm0,xmm4
+        movdqa    xmm3,xmm2
+        movdqa    xmm4,xmm0
+        pmaddwd   xmm2,[rel PW_MF060_MF089]     ; xmm2=tmp0L
+        pmaddwd   xmm0,[rel PW_MF060_MF089]     ; xmm0=tmp0H
+        pmaddwd   xmm3,[rel PW_MF089_F060]      ; xmm3=tmp3L
+        pmaddwd   xmm4,[rel PW_MF089_F060]      ; xmm4=tmp3H
+
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp0L
+        paddd   xmm0, XMMWORD [wk(11)]  ; xmm0=tmp0H
+        paddd   xmm3,xmm5               ; xmm3=tmp3L
+        paddd   xmm4,xmm7               ; xmm4=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm2   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm0   ; wk(9)=tmp0H
+
+        movdqa    xmm2,xmm1
+        movdqa    xmm0,xmm1
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm0,xmm6
+        movdqa    xmm1,xmm2
+        movdqa    xmm6,xmm0
+        pmaddwd   xmm2,[rel PW_MF050_MF256]     ; xmm2=tmp1L
+        pmaddwd   xmm0,[rel PW_MF050_MF256]     ; xmm0=tmp1H
+        pmaddwd   xmm1,[rel PW_MF256_F050]      ; xmm1=tmp2L
+        pmaddwd   xmm6,[rel PW_MF256_F050]      ; xmm6=tmp2H
+
+        paddd   xmm2,xmm5               ; xmm2=tmp1L
+        paddd   xmm0,xmm7               ; xmm0=tmp1H
+        paddd   xmm1, XMMWORD [wk(10)]  ; xmm1=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=tmp10H
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm0,xmm7
+        paddd   xmm5,xmm3               ; xmm5=data0L
+        paddd   xmm7,xmm4               ; xmm7=data0H
+        psubd   xmm2,xmm3               ; xmm2=data7L
+        psubd   xmm0,xmm4               ; xmm0=data7H
+
+        movdqa  xmm3,[rel PD_DESCALE_P1]        ; xmm3=[rel PD_DESCALE_P1]
+
+        paddd   xmm5,xmm3
+        paddd   xmm7,xmm3
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm7,DESCALE_P1
+        paddd   xmm2,xmm3
+        paddd   xmm0,xmm3
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm5,xmm7             ; xmm5=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm2,xmm0             ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+        movdqa  xmm4, XMMWORD [wk(4)]   ; xmm4=tmp11L
+        movdqa  xmm3, XMMWORD [wk(5)]   ; xmm3=tmp11H
+
+        movdqa  xmm7,xmm4
+        movdqa  xmm0,xmm3
+        paddd   xmm4,xmm1               ; xmm4=data1L
+        paddd   xmm3,xmm6               ; xmm3=data1H
+        psubd   xmm7,xmm1               ; xmm7=data6L
+        psubd   xmm0,xmm6               ; xmm0=data6H
+
+        movdqa  xmm1,[rel PD_DESCALE_P1]        ; xmm1=[rel PD_DESCALE_P1]
+
+        paddd   xmm4,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+        paddd   xmm7,xmm1
+        paddd   xmm0,xmm1
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm4,xmm3             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm7,xmm0             ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+        movdqa    xmm6,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm4             ; xmm5=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4             ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm2             ; xmm7=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm1,xmm2             ; xmm1=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(6)]   ; xmm3=tmp12L
+        movdqa  xmm0, XMMWORD [wk(7)]   ; xmm0=tmp12H
+        movdqa  xmm4, XMMWORD [wk(10)]  ; xmm4=tmp1L
+        movdqa  xmm2, XMMWORD [wk(11)]  ; xmm2=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(5)], xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm5,xmm3
+        movdqa  xmm6,xmm0
+        paddd   xmm3,xmm4               ; xmm3=data2L
+        paddd   xmm0,xmm2               ; xmm0=data2H
+        psubd   xmm5,xmm4               ; xmm5=data5L
+        psubd   xmm6,xmm2               ; xmm6=data5H
+
+        movdqa  xmm7,[rel PD_DESCALE_P1]        ; xmm7=[rel PD_DESCALE_P1]
+
+        paddd   xmm3,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm3,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+        paddd   xmm5,xmm7
+        paddd   xmm6,xmm7
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm3,xmm0             ; xmm3=data2=(20 21 22 23 24 25 26 27)
+        packssdw  xmm5,xmm6             ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+        movdqa  xmm1, XMMWORD [wk(2)]   ; xmm1=tmp13L
+        movdqa  xmm4, XMMWORD [wk(3)]   ; xmm4=tmp13H
+        movdqa  xmm2, XMMWORD [wk(8)]   ; xmm2=tmp0L
+        movdqa  xmm7, XMMWORD [wk(9)]   ; xmm7=tmp0H
+
+        movdqa  xmm0,xmm1
+        movdqa  xmm6,xmm4
+        paddd   xmm1,xmm2               ; xmm1=data3L
+        paddd   xmm4,xmm7               ; xmm4=data3H
+        psubd   xmm0,xmm2               ; xmm0=data4L
+        psubd   xmm6,xmm7               ; xmm6=data4H
+
+        movdqa  xmm2,[rel PD_DESCALE_P1]        ; xmm2=[rel PD_DESCALE_P1]
+
+        paddd   xmm1,xmm2
+        paddd   xmm4,xmm2
+        psrad   xmm1,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm0,xmm2
+        paddd   xmm6,xmm2
+        psrad   xmm0,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm1,xmm4             ; xmm1=data3=(30 31 32 33 34 35 36 37)
+        packssdw  xmm0,xmm6             ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(00 10 01 11 02 12 03 13)
+        movdqa  xmm2, XMMWORD [wk(1)]   ; xmm2=(04 14 05 15 06 16 07 17)
+
+        movdqa    xmm4,xmm3             ; transpose coefficients(phase 1)
+        punpcklwd xmm3,xmm1             ; xmm3=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm1             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm6,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm6,xmm5             ; xmm6=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm3             ; xmm7=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm1,xmm3             ; xmm1=(02 12 22 32 03 13 23 33)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm4             ; xmm2=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm5,xmm4             ; xmm5=(06 16 26 36 07 17 27 37)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=(60 70 61 71 62 72 63 73)
+        movdqa  xmm4, XMMWORD [wk(5)]   ; xmm4=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(7)], xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm2,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm2,xmm3             ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm4             ; xmm6=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm5,xmm4             ; xmm5=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm3,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm0            ; xmm7=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm3,xmm0            ; xmm3=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm4,xmm2            ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm0, XMMWORD [wk(6)]   ; xmm0=(04 14 24 34 05 15 25 35)
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(8)], xmm3   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm4   ; wk(9)=col3
+
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm3,xmm6            ; xmm3=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm4,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm5            ; xmm2=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm4,xmm5            ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm6,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm6,xmm2             ; xmm2=in6=z3
+        punpckhwd xmm5,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm2,xmm5
+        pmaddwd   xmm6,[rel PW_F130_F054]       ; xmm6=tmp3L
+        pmaddwd   xmm5,[rel PW_F130_F054]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[rel PW_F054_MF130]      ; xmm1=tmp2L
+        pmaddwd   xmm2,[rel PW_F054_MF130]      ; xmm2=tmp2H
+
+        movdqa    xmm3,xmm7
+        paddw     xmm7,xmm0             ; xmm7=in0+in4
+        psubw     xmm3,xmm0             ; xmm3=in0-in4
+
+        pxor      xmm4,xmm4
+        pxor      xmm0,xmm0
+        punpcklwd xmm4,xmm7             ; xmm4=tmp0L
+        punpckhwd xmm0,xmm7             ; xmm0=tmp0H
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+        psrad     xmm0,(16-CONST_BITS)  ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm6               ; xmm4=tmp10L
+        psubd   xmm7,xmm6               ; xmm7=tmp13L
+        movdqa  xmm6,xmm0
+        paddd   xmm0,xmm5               ; xmm0=tmp10H
+        psubd   xmm6,xmm5               ; xmm6=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm6   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm4,xmm4
+        punpcklwd xmm5,xmm3             ; xmm5=tmp1L
+        punpckhwd xmm4,xmm3             ; xmm4=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+        movdqa  xmm0,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm0,xmm1               ; xmm0=tmp12L
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm2               ; xmm4=tmp11H
+        psubd   xmm7,xmm2               ; xmm7=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm0   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm7   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm6, XMMWORD [wk(9)]   ; xmm6=col3
+        movdqa  xmm3, XMMWORD [wk(8)]   ; xmm3=col1
+        movdqa  xmm1, XMMWORD [wk(11)]  ; xmm1=col7
+        movdqa  xmm2, XMMWORD [wk(10)]  ; xmm2=col5
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm4,xmm3
+        paddw   xmm5,xmm1               ; xmm5=z3
+        paddw   xmm4,xmm2               ; xmm4=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm0,xmm5
+        movdqa    xmm7,xmm5
+        punpcklwd xmm0,xmm4
+        punpckhwd xmm7,xmm4
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm7
+        pmaddwd   xmm0,[rel PW_MF078_F117]      ; xmm0=z3L
+        pmaddwd   xmm7,[rel PW_MF078_F117]      ; xmm7=z3H
+        pmaddwd   xmm5,[rel PW_F117_F078]       ; xmm5=z4L
+        pmaddwd   xmm4,[rel PW_F117_F078]       ; xmm4=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm0,xmm1
+        movdqa    xmm7,xmm1
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm1,xmm0
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm0,[rel PW_MF060_MF089]     ; xmm0=tmp0L
+        pmaddwd   xmm7,[rel PW_MF060_MF089]     ; xmm7=tmp0H
+        pmaddwd   xmm1,[rel PW_MF089_F060]      ; xmm1=tmp3L
+        pmaddwd   xmm3,[rel PW_MF089_F060]      ; xmm3=tmp3H
+
+        paddd   xmm0, XMMWORD [wk(10)]  ; xmm0=tmp0L
+        paddd   xmm7, XMMWORD [wk(11)]  ; xmm7=tmp0H
+        paddd   xmm1,xmm5               ; xmm1=tmp3L
+        paddd   xmm3,xmm4               ; xmm3=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm0   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm7   ; wk(9)=tmp0H
+
+        movdqa    xmm0,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm0,xmm6
+        punpckhwd xmm7,xmm6
+        movdqa    xmm2,xmm0
+        movdqa    xmm6,xmm7
+        pmaddwd   xmm0,[rel PW_MF050_MF256]     ; xmm0=tmp1L
+        pmaddwd   xmm7,[rel PW_MF050_MF256]     ; xmm7=tmp1H
+        pmaddwd   xmm2,[rel PW_MF256_F050]      ; xmm2=tmp2L
+        pmaddwd   xmm6,[rel PW_MF256_F050]      ; xmm6=tmp2H
+
+        paddd   xmm0,xmm5               ; xmm0=tmp1L
+        paddd   xmm7,xmm4               ; xmm7=tmp1H
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm4, XMMWORD [wk(1)]   ; xmm4=tmp10H
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm7,xmm4
+        paddd   xmm5,xmm1               ; xmm5=data0L
+        paddd   xmm4,xmm3               ; xmm4=data0H
+        psubd   xmm0,xmm1               ; xmm0=data7L
+        psubd   xmm7,xmm3               ; xmm7=data7H
+
+        movdqa  xmm1,[rel PD_DESCALE_P2]        ; xmm1=[rel PD_DESCALE_P2]
+
+        paddd   xmm5,xmm1
+        paddd   xmm4,xmm1
+        psrad   xmm5,DESCALE_P2
+        psrad   xmm4,DESCALE_P2
+        paddd   xmm0,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm5,xmm4             ; xmm5=data0=(00 10 20 30 40 50 60 70)
+        packssdw  xmm0,xmm7             ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=tmp11L
+        movdqa  xmm1, XMMWORD [wk(5)]   ; xmm1=tmp11H
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm7,xmm1
+        paddd   xmm3,xmm2               ; xmm3=data1L
+        paddd   xmm1,xmm6               ; xmm1=data1H
+        psubd   xmm4,xmm2               ; xmm4=data6L
+        psubd   xmm7,xmm6               ; xmm7=data6H
+
+        movdqa  xmm2,[rel PD_DESCALE_P2]        ; xmm2=[rel PD_DESCALE_P2]
+
+        paddd   xmm3,xmm2
+        paddd   xmm1,xmm2
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm4,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm3,xmm1             ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        packssdw  xmm4,xmm7             ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+        packsswb  xmm5,xmm4             ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm0             ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm6, XMMWORD [wk(6)]   ; xmm6=tmp12L
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=tmp12H
+        movdqa  xmm1, XMMWORD [wk(10)]  ; xmm1=tmp1L
+        movdqa  xmm7, XMMWORD [wk(11)]  ; xmm7=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm4,xmm6
+        movdqa  xmm0,xmm2
+        paddd   xmm6,xmm1               ; xmm6=data2L
+        paddd   xmm2,xmm7               ; xmm2=data2H
+        psubd   xmm4,xmm1               ; xmm4=data5L
+        psubd   xmm0,xmm7               ; xmm0=data5H
+
+        movdqa  xmm5,[rel PD_DESCALE_P2]        ; xmm5=[rel PD_DESCALE_P2]
+
+        paddd   xmm6,xmm5
+        paddd   xmm2,xmm5
+        psrad   xmm6,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm4,xmm5
+        paddd   xmm0,xmm5
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        packssdw  xmm6,xmm2             ; xmm6=data2=(02 12 22 32 42 52 62 72)
+        packssdw  xmm4,xmm0             ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+        movdqa  xmm3, XMMWORD [wk(2)]   ; xmm3=tmp13L
+        movdqa  xmm1, XMMWORD [wk(3)]   ; xmm1=tmp13H
+        movdqa  xmm7, XMMWORD [wk(8)]   ; xmm7=tmp0L
+        movdqa  xmm5, XMMWORD [wk(9)]   ; xmm5=tmp0H
+
+        movdqa  xmm2,xmm3
+        movdqa  xmm0,xmm1
+        paddd   xmm3,xmm7               ; xmm3=data3L
+        paddd   xmm1,xmm5               ; xmm1=data3H
+        psubd   xmm2,xmm7               ; xmm2=data4L
+        psubd   xmm0,xmm5               ; xmm0=data4H
+
+        movdqa  xmm7,[rel PD_DESCALE_P2]        ; xmm7=[rel PD_DESCALE_P2]
+
+        paddd   xmm3,xmm7
+        paddd   xmm1,xmm7
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm2,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm2,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        movdqa    xmm5,[rel PB_CENTERJSAMP]     ; xmm5=[rel PB_CENTERJSAMP]
+
+        packssdw  xmm3,xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+        packssdw  xmm2,xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+        movdqa    xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        packsswb  xmm6,xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm3,xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm7,xmm5
+        paddb     xmm1,xmm5
+        paddb     xmm6,xmm5
+        paddb     xmm3,xmm5
+
+        movdqa    xmm0,xmm7     ; transpose coefficients(phase 1)
+        punpcklbw xmm7,xmm1     ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm1     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 1)
+        punpcklbw xmm6,xmm3     ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm2,xmm3     ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm7     ; transpose coefficients(phase 2)
+        punpcklwd xmm7,xmm6     ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm6     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm5,xmm2     ; transpose coefficients(phase 2)
+        punpcklwd xmm2,xmm0     ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm5,xmm0     ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm1,xmm7     ; transpose coefficients(phase 3)
+        punpckldq xmm7,xmm2     ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm1,xmm2     ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm3,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm5     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm3,xmm5     ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm6,xmm7,0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm1,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm2,xmm4,0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm5,xmm3,0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm7
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm1
+        mov     rdx, JSAMPROW [rdi+4*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+
+        mov     rdx, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm0
+        mov     rdx, JSAMPROW [rdi+5*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE], xmm2
+        movq    XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE], xmm5
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctint-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctint-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,859 @@
+;
+; jidctint.asm - accurate integer IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains a slow-but-accurate integer implementation of the
+; inverse DCT (Discrete Cosine Transform). The following code is based
+; directly on the IJG's original jidctint.c; see the jidctint.c for
+; more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1      (CONST_BITS-PASS1_BITS)
+%define DESCALE_P2      (CONST_BITS+PASS1_BITS+3)
+
+%if CONST_BITS == 13
+F_0_298 equ      2446           ; FIX(0.298631336)
+F_0_390 equ      3196           ; FIX(0.390180644)
+F_0_541 equ      4433           ; FIX(0.541196100)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_175 equ      9633           ; FIX(1.175875602)
+F_1_501 equ     12299           ; FIX(1.501321110)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_1_961 equ     16069           ; FIX(1.961570560)
+F_2_053 equ     16819           ; FIX(2.053119869)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_072 equ     25172           ; FIX(3.072711026)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_298 equ     DESCALE( 320652955,30-CONST_BITS)       ; FIX(0.298631336)
+F_0_390 equ     DESCALE( 418953276,30-CONST_BITS)       ; FIX(0.390180644)
+F_0_541 equ     DESCALE( 581104887,30-CONST_BITS)       ; FIX(0.541196100)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_175 equ     DESCALE(1262586813,30-CONST_BITS)       ; FIX(1.175875602)
+F_1_501 equ     DESCALE(1612031267,30-CONST_BITS)       ; FIX(1.501321110)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_1_961 equ     DESCALE(2106220350,30-CONST_BITS)       ; FIX(1.961570560)
+F_2_053 equ     DESCALE(2204520673,30-CONST_BITS)       ; FIX(2.053119869)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_072 equ     DESCALE(3299298341,30-CONST_BITS)       ; FIX(3.072711026)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_islow_sse2) PRIVATE
+
+EXTN(jconst_idct_islow_sse2):
+
+PW_F130_F054    times 4 dw  (F_0_541+F_0_765), F_0_541
+PW_F054_MF130   times 4 dw  F_0_541, (F_0_541-F_1_847)
+PW_MF078_F117   times 4 dw  (F_1_175-F_1_961), F_1_175
+PW_F117_F078    times 4 dw  F_1_175, (F_1_175-F_0_390)
+PW_MF060_MF089  times 4 dw  (F_0_298-F_0_899),-F_0_899
+PW_MF089_F060   times 4 dw -F_0_899, (F_1_501-F_0_899)
+PW_MF050_MF256  times 4 dw  (F_2_053-F_2_562),-F_2_562
+PW_MF256_F050   times 4 dw -F_2_562, (F_3_072-F_2_562)
+PD_DESCALE_P1   times 4 dd  1 << (DESCALE_P1-1)
+PD_DESCALE_P2   times 4 dd  1 << (DESCALE_P2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients.
+;
+; GLOBAL(void)
+; jsimd_idct_islow_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                        JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; jpeg_component_info *compptr
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          12
+
+        align   16
+        global  EXTN(jsimd_idct_islow_sse2) PRIVATE
+
+EXTN(jsimd_idct_islow_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_ISLOW_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     near .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm1,xmm0
+        packsswb xmm1,xmm1
+        packsswb xmm1,xmm1
+        movd    eax,xmm1
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm5, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm5,PASS1_BITS
+
+        movdqa    xmm4,xmm5             ; xmm5=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm5,xmm5             ; xmm5=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm4,xmm4             ; xmm4=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm7,xmm5,0x00          ; xmm7=col0=(00 00 00 00 00 00 00 00)
+        pshufd  xmm6,xmm5,0x55          ; xmm6=col1=(01 01 01 01 01 01 01 01)
+        pshufd  xmm1,xmm5,0xAA          ; xmm1=col2=(02 02 02 02 02 02 02 02)
+        pshufd  xmm5,xmm5,0xFF          ; xmm5=col3=(03 03 03 03 03 03 03 03)
+        pshufd  xmm0,xmm4,0x00          ; xmm0=col4=(04 04 04 04 04 04 04 04)
+        pshufd  xmm3,xmm4,0x55          ; xmm3=col5=(05 05 05 05 05 05 05 05)
+        pshufd  xmm2,xmm4,0xAA          ; xmm2=col6=(06 06 06 06 06 06 06 06)
+        pshufd  xmm4,xmm4,0xFF          ; xmm4=col7=(07 07 07 07 07 07 07 07)
+
+        movdqa  XMMWORD [wk(8)], xmm6   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm5   ; wk(9)=col3
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Even part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(4,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(4,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm4,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm4,xmm3             ; xmm3=in6=z3
+        punpckhwd xmm5,xmm3
+        movdqa    xmm1,xmm4
+        movdqa    xmm3,xmm5
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F130_F054)]       ; xmm4=tmp3L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F130_F054)]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=tmp2L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm3=tmp2H
+
+        movdqa    xmm6,xmm0
+        paddw     xmm0,xmm2             ; xmm0=in0+in4
+        psubw     xmm6,xmm2             ; xmm6=in0-in4
+
+        pxor      xmm7,xmm7
+        pxor      xmm2,xmm2
+        punpcklwd xmm7,xmm0             ; xmm7=tmp0L
+        punpckhwd xmm2,xmm0             ; xmm2=tmp0H
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+        psrad     xmm2,(16-CONST_BITS)  ; psrad xmm2,16 & pslld xmm2,CONST_BITS
+
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm4               ; xmm7=tmp10L
+        psubd   xmm0,xmm4               ; xmm0=tmp13L
+        movdqa  xmm4,xmm2
+        paddd   xmm2,xmm5               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm7   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm2   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm0   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm4   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm7,xmm7
+        punpcklwd xmm5,xmm6             ; xmm5=tmp1L
+        punpckhwd xmm7,xmm6             ; xmm7=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm7,(16-CONST_BITS)  ; psrad xmm7,16 & pslld xmm7,CONST_BITS
+
+        movdqa  xmm2,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm2,xmm1               ; xmm2=tmp12L
+        movdqa  xmm0,xmm7
+        paddd   xmm7,xmm3               ; xmm7=tmp11H
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm7   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm0   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm7,xmm4
+        paddw   xmm5,xmm3               ; xmm5=z3
+        paddw   xmm7,xmm1               ; xmm7=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm2,xmm5
+        movdqa    xmm0,xmm5
+        punpcklwd xmm2,xmm7
+        punpckhwd xmm0,xmm7
+        movdqa    xmm5,xmm2
+        movdqa    xmm7,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm2=z3L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm0=z3H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F117_F078)]       ; xmm5=z4L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F117_F078)]       ; xmm7=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm2,xmm3
+        movdqa    xmm0,xmm3
+        punpcklwd xmm2,xmm4
+        punpckhwd xmm0,xmm4
+        movdqa    xmm3,xmm2
+        movdqa    xmm4,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm2=tmp0L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm0=tmp0H
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm3=tmp3L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm4=tmp3H
+
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp0L
+        paddd   xmm0, XMMWORD [wk(11)]  ; xmm0=tmp0H
+        paddd   xmm3,xmm5               ; xmm3=tmp3L
+        paddd   xmm4,xmm7               ; xmm4=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm2   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm0   ; wk(9)=tmp0H
+
+        movdqa    xmm2,xmm1
+        movdqa    xmm0,xmm1
+        punpcklwd xmm2,xmm6
+        punpckhwd xmm0,xmm6
+        movdqa    xmm1,xmm2
+        movdqa    xmm6,xmm0
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm2=tmp1L
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm0=tmp1H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm1=tmp2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm6=tmp2H
+
+        paddd   xmm2,xmm5               ; xmm2=tmp1L
+        paddd   xmm0,xmm7               ; xmm0=tmp1H
+        paddd   xmm1, XMMWORD [wk(10)]  ; xmm1=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm2  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm0  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm7, XMMWORD [wk(1)]   ; xmm7=tmp10H
+
+        movdqa  xmm2,xmm5
+        movdqa  xmm0,xmm7
+        paddd   xmm5,xmm3               ; xmm5=data0L
+        paddd   xmm7,xmm4               ; xmm7=data0H
+        psubd   xmm2,xmm3               ; xmm2=data7L
+        psubd   xmm0,xmm4               ; xmm0=data7H
+
+        movdqa  xmm3,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm3=[PD_DESCALE_P1]
+
+        paddd   xmm5,xmm3
+        paddd   xmm7,xmm3
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm7,DESCALE_P1
+        paddd   xmm2,xmm3
+        paddd   xmm0,xmm3
+        psrad   xmm2,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm5,xmm7             ; xmm5=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm2,xmm0             ; xmm2=data7=(70 71 72 73 74 75 76 77)
+
+        movdqa  xmm4, XMMWORD [wk(4)]   ; xmm4=tmp11L
+        movdqa  xmm3, XMMWORD [wk(5)]   ; xmm3=tmp11H
+
+        movdqa  xmm7,xmm4
+        movdqa  xmm0,xmm3
+        paddd   xmm4,xmm1               ; xmm4=data1L
+        paddd   xmm3,xmm6               ; xmm3=data1H
+        psubd   xmm7,xmm1               ; xmm7=data6L
+        psubd   xmm0,xmm6               ; xmm0=data6H
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm1=[PD_DESCALE_P1]
+
+        paddd   xmm4,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm4,DESCALE_P1
+        psrad   xmm3,DESCALE_P1
+        paddd   xmm7,xmm1
+        paddd   xmm0,xmm1
+        psrad   xmm7,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+
+        packssdw  xmm4,xmm3             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm7,xmm0             ; xmm7=data6=(60 61 62 63 64 65 66 67)
+
+        movdqa    xmm6,xmm5             ; transpose coefficients(phase 1)
+        punpcklwd xmm5,xmm4             ; xmm5=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4             ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 1)
+        punpcklwd xmm7,xmm2             ; xmm7=(60 70 61 71 62 72 63 73)
+        punpckhwd xmm1,xmm2             ; xmm1=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(6)]   ; xmm3=tmp12L
+        movdqa  xmm0, XMMWORD [wk(7)]   ; xmm0=tmp12H
+        movdqa  xmm4, XMMWORD [wk(10)]  ; xmm4=tmp1L
+        movdqa  xmm2, XMMWORD [wk(11)]  ; xmm2=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 01 11 02 12 03 13)
+        movdqa  XMMWORD [wk(1)], xmm6   ; wk(1)=(04 14 05 15 06 16 07 17)
+        movdqa  XMMWORD [wk(4)], xmm7   ; wk(4)=(60 70 61 71 62 72 63 73)
+        movdqa  XMMWORD [wk(5)], xmm1   ; wk(5)=(64 74 65 75 66 76 67 77)
+
+        movdqa  xmm5,xmm3
+        movdqa  xmm6,xmm0
+        paddd   xmm3,xmm4               ; xmm3=data2L
+        paddd   xmm0,xmm2               ; xmm0=data2H
+        psubd   xmm5,xmm4               ; xmm5=data5L
+        psubd   xmm6,xmm2               ; xmm6=data5H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm7=[PD_DESCALE_P1]
+
+        paddd   xmm3,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm3,DESCALE_P1
+        psrad   xmm0,DESCALE_P1
+        paddd   xmm5,xmm7
+        paddd   xmm6,xmm7
+        psrad   xmm5,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm3,xmm0             ; xmm3=data2=(20 21 22 23 24 25 26 27)
+        packssdw  xmm5,xmm6             ; xmm5=data5=(50 51 52 53 54 55 56 57)
+
+        movdqa  xmm1, XMMWORD [wk(2)]   ; xmm1=tmp13L
+        movdqa  xmm4, XMMWORD [wk(3)]   ; xmm4=tmp13H
+        movdqa  xmm2, XMMWORD [wk(8)]   ; xmm2=tmp0L
+        movdqa  xmm7, XMMWORD [wk(9)]   ; xmm7=tmp0H
+
+        movdqa  xmm0,xmm1
+        movdqa  xmm6,xmm4
+        paddd   xmm1,xmm2               ; xmm1=data3L
+        paddd   xmm4,xmm7               ; xmm4=data3H
+        psubd   xmm0,xmm2               ; xmm0=data4L
+        psubd   xmm6,xmm7               ; xmm6=data4H
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P1)]        ; xmm2=[PD_DESCALE_P1]
+
+        paddd   xmm1,xmm2
+        paddd   xmm4,xmm2
+        psrad   xmm1,DESCALE_P1
+        psrad   xmm4,DESCALE_P1
+        paddd   xmm0,xmm2
+        paddd   xmm6,xmm2
+        psrad   xmm0,DESCALE_P1
+        psrad   xmm6,DESCALE_P1
+
+        packssdw  xmm1,xmm4             ; xmm1=data3=(30 31 32 33 34 35 36 37)
+        packssdw  xmm0,xmm6             ; xmm0=data4=(40 41 42 43 44 45 46 47)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=(00 10 01 11 02 12 03 13)
+        movdqa  xmm2, XMMWORD [wk(1)]   ; xmm2=(04 14 05 15 06 16 07 17)
+
+        movdqa    xmm4,xmm3             ; transpose coefficients(phase 1)
+        punpcklwd xmm3,xmm1             ; xmm3=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm4,xmm1             ; xmm4=(24 34 25 35 26 36 27 37)
+        movdqa    xmm6,xmm0             ; transpose coefficients(phase 1)
+        punpcklwd xmm0,xmm5             ; xmm0=(40 50 41 51 42 52 43 53)
+        punpckhwd xmm6,xmm5             ; xmm6=(44 54 45 55 46 56 47 57)
+
+        movdqa    xmm1,xmm7             ; transpose coefficients(phase 2)
+        punpckldq xmm7,xmm3             ; xmm7=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm1,xmm3             ; xmm1=(02 12 22 32 03 13 23 33)
+        movdqa    xmm5,xmm2             ; transpose coefficients(phase 2)
+        punpckldq xmm2,xmm4             ; xmm2=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm5,xmm4             ; xmm5=(06 16 26 36 07 17 27 37)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=(60 70 61 71 62 72 63 73)
+        movdqa  xmm4, XMMWORD [wk(5)]   ; xmm4=(64 74 65 75 66 76 67 77)
+
+        movdqa  XMMWORD [wk(6)], xmm2   ; wk(6)=(04 14 24 34 05 15 25 35)
+        movdqa  XMMWORD [wk(7)], xmm5   ; wk(7)=(06 16 26 36 07 17 27 37)
+
+        movdqa    xmm2,xmm0             ; transpose coefficients(phase 2)
+        punpckldq xmm0,xmm3             ; xmm0=(40 50 60 70 41 51 61 71)
+        punpckhdq xmm2,xmm3             ; xmm2=(42 52 62 72 43 53 63 73)
+        movdqa    xmm5,xmm6             ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm4             ; xmm6=(44 54 64 74 45 55 65 75)
+        punpckhdq xmm5,xmm4             ; xmm5=(46 56 66 76 47 57 67 77)
+
+        movdqa     xmm3,xmm7            ; transpose coefficients(phase 3)
+        punpcklqdq xmm7,xmm0            ; xmm7=col0=(00 10 20 30 40 50 60 70)
+        punpckhqdq xmm3,xmm0            ; xmm3=col1=(01 11 21 31 41 51 61 71)
+        movdqa     xmm4,xmm1            ; transpose coefficients(phase 3)
+        punpcklqdq xmm1,xmm2            ; xmm1=col2=(02 12 22 32 42 52 62 72)
+        punpckhqdq xmm4,xmm2            ; xmm4=col3=(03 13 23 33 43 53 63 73)
+
+        movdqa  xmm0, XMMWORD [wk(6)]   ; xmm0=(04 14 24 34 05 15 25 35)
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=(06 16 26 36 07 17 27 37)
+
+        movdqa  XMMWORD [wk(8)], xmm3   ; wk(8)=col1
+        movdqa  XMMWORD [wk(9)], xmm4   ; wk(9)=col3
+
+        movdqa     xmm3,xmm0            ; transpose coefficients(phase 3)
+        punpcklqdq xmm0,xmm6            ; xmm0=col4=(04 14 24 34 44 54 64 74)
+        punpckhqdq xmm3,xmm6            ; xmm3=col5=(05 15 25 35 45 55 65 75)
+        movdqa     xmm4,xmm2            ; transpose coefficients(phase 3)
+        punpcklqdq xmm2,xmm5            ; xmm2=col6=(06 16 26 36 46 56 66 76)
+        punpckhqdq xmm4,xmm5            ; xmm4=col7=(07 17 27 37 47 57 67 77)
+
+        movdqa  XMMWORD [wk(10)], xmm3  ; wk(10)=col5
+        movdqa  XMMWORD [wk(11)], xmm4  ; wk(11)=col7
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        ; xmm7=col0, xmm1=col2, xmm0=col4, xmm2=col6
+
+        ; (Original)
+        ; z1 = (z2 + z3) * 0.541196100;
+        ; tmp2 = z1 + z3 * -1.847759065;
+        ; tmp3 = z1 + z2 * 0.765366865;
+        ;
+        ; (This implementation)
+        ; tmp2 = z2 * 0.541196100 + z3 * (0.541196100 - 1.847759065);
+        ; tmp3 = z2 * (0.541196100 + 0.765366865) + z3 * 0.541196100;
+
+        movdqa    xmm6,xmm1             ; xmm1=in2=z2
+        movdqa    xmm5,xmm1
+        punpcklwd xmm6,xmm2             ; xmm2=in6=z3
+        punpckhwd xmm5,xmm2
+        movdqa    xmm1,xmm6
+        movdqa    xmm2,xmm5
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_F130_F054)]       ; xmm6=tmp3L
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F130_F054)]       ; xmm5=tmp3H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm1=tmp2L
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F054_MF130)]      ; xmm2=tmp2H
+
+        movdqa    xmm3,xmm7
+        paddw     xmm7,xmm0             ; xmm7=in0+in4
+        psubw     xmm3,xmm0             ; xmm3=in0-in4
+
+        pxor      xmm4,xmm4
+        pxor      xmm0,xmm0
+        punpcklwd xmm4,xmm7             ; xmm4=tmp0L
+        punpckhwd xmm0,xmm7             ; xmm0=tmp0H
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+        psrad     xmm0,(16-CONST_BITS)  ; psrad xmm0,16 & pslld xmm0,CONST_BITS
+
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm6               ; xmm4=tmp10L
+        psubd   xmm7,xmm6               ; xmm7=tmp13L
+        movdqa  xmm6,xmm0
+        paddd   xmm0,xmm5               ; xmm0=tmp10H
+        psubd   xmm6,xmm5               ; xmm6=tmp13H
+
+        movdqa  XMMWORD [wk(0)], xmm4   ; wk(0)=tmp10L
+        movdqa  XMMWORD [wk(1)], xmm0   ; wk(1)=tmp10H
+        movdqa  XMMWORD [wk(2)], xmm7   ; wk(2)=tmp13L
+        movdqa  XMMWORD [wk(3)], xmm6   ; wk(3)=tmp13H
+
+        pxor      xmm5,xmm5
+        pxor      xmm4,xmm4
+        punpcklwd xmm5,xmm3             ; xmm5=tmp1L
+        punpckhwd xmm4,xmm3             ; xmm4=tmp1H
+        psrad     xmm5,(16-CONST_BITS)  ; psrad xmm5,16 & pslld xmm5,CONST_BITS
+        psrad     xmm4,(16-CONST_BITS)  ; psrad xmm4,16 & pslld xmm4,CONST_BITS
+
+        movdqa  xmm0,xmm5
+        paddd   xmm5,xmm1               ; xmm5=tmp11L
+        psubd   xmm0,xmm1               ; xmm0=tmp12L
+        movdqa  xmm7,xmm4
+        paddd   xmm4,xmm2               ; xmm4=tmp11H
+        psubd   xmm7,xmm2               ; xmm7=tmp12H
+
+        movdqa  XMMWORD [wk(4)], xmm5   ; wk(4)=tmp11L
+        movdqa  XMMWORD [wk(5)], xmm4   ; wk(5)=tmp11H
+        movdqa  XMMWORD [wk(6)], xmm0   ; wk(6)=tmp12L
+        movdqa  XMMWORD [wk(7)], xmm7   ; wk(7)=tmp12H
+
+        ; -- Odd part
+
+        movdqa  xmm6, XMMWORD [wk(9)]   ; xmm6=col3
+        movdqa  xmm3, XMMWORD [wk(8)]   ; xmm3=col1
+        movdqa  xmm1, XMMWORD [wk(11)]  ; xmm1=col7
+        movdqa  xmm2, XMMWORD [wk(10)]  ; xmm2=col5
+
+        movdqa  xmm5,xmm6
+        movdqa  xmm4,xmm3
+        paddw   xmm5,xmm1               ; xmm5=z3
+        paddw   xmm4,xmm2               ; xmm4=z4
+
+        ; (Original)
+        ; z5 = (z3 + z4) * 1.175875602;
+        ; z3 = z3 * -1.961570560;  z4 = z4 * -0.390180644;
+        ; z3 += z5;  z4 += z5;
+        ;
+        ; (This implementation)
+        ; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
+        ; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
+
+        movdqa    xmm0,xmm5
+        movdqa    xmm7,xmm5
+        punpcklwd xmm0,xmm4
+        punpckhwd xmm7,xmm4
+        movdqa    xmm5,xmm0
+        movdqa    xmm4,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm0=z3L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF078_F117)]      ; xmm7=z3H
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F117_F078)]       ; xmm5=z4L
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F117_F078)]       ; xmm4=z4H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=z3L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=z3H
+
+        ; (Original)
+        ; z1 = tmp0 + tmp3;  z2 = tmp1 + tmp2;
+        ; tmp0 = tmp0 * 0.298631336;  tmp1 = tmp1 * 2.053119869;
+        ; tmp2 = tmp2 * 3.072711026;  tmp3 = tmp3 * 1.501321110;
+        ; z1 = z1 * -0.899976223;  z2 = z2 * -2.562915447;
+        ; tmp0 += z1 + z3;  tmp1 += z2 + z4;
+        ; tmp2 += z2 + z3;  tmp3 += z1 + z4;
+        ;
+        ; (This implementation)
+        ; tmp0 = tmp0 * (0.298631336 - 0.899976223) + tmp3 * -0.899976223;
+        ; tmp1 = tmp1 * (2.053119869 - 2.562915447) + tmp2 * -2.562915447;
+        ; tmp2 = tmp1 * -2.562915447 + tmp2 * (3.072711026 - 2.562915447);
+        ; tmp3 = tmp0 * -0.899976223 + tmp3 * (1.501321110 - 0.899976223);
+        ; tmp0 += z3;  tmp1 += z4;
+        ; tmp2 += z3;  tmp3 += z4;
+
+        movdqa    xmm0,xmm1
+        movdqa    xmm7,xmm1
+        punpcklwd xmm0,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm1,xmm0
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm0=tmp0L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF089)]     ; xmm7=tmp0H
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm1=tmp3L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_MF089_F060)]      ; xmm3=tmp3H
+
+        paddd   xmm0, XMMWORD [wk(10)]  ; xmm0=tmp0L
+        paddd   xmm7, XMMWORD [wk(11)]  ; xmm7=tmp0H
+        paddd   xmm1,xmm5               ; xmm1=tmp3L
+        paddd   xmm3,xmm4               ; xmm3=tmp3H
+
+        movdqa  XMMWORD [wk(8)], xmm0   ; wk(8)=tmp0L
+        movdqa  XMMWORD [wk(9)], xmm7   ; wk(9)=tmp0H
+
+        movdqa    xmm0,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm0,xmm6
+        punpckhwd xmm7,xmm6
+        movdqa    xmm2,xmm0
+        movdqa    xmm6,xmm7
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm0=tmp1L
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF050_MF256)]     ; xmm7=tmp1H
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm2=tmp2L
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF256_F050)]      ; xmm6=tmp2H
+
+        paddd   xmm0,xmm5               ; xmm0=tmp1L
+        paddd   xmm7,xmm4               ; xmm7=tmp1H
+        paddd   xmm2, XMMWORD [wk(10)]  ; xmm2=tmp2L
+        paddd   xmm6, XMMWORD [wk(11)]  ; xmm6=tmp2H
+
+        movdqa  XMMWORD [wk(10)], xmm0  ; wk(10)=tmp1L
+        movdqa  XMMWORD [wk(11)], xmm7  ; wk(11)=tmp1H
+
+        ; -- Final output stage
+
+        movdqa  xmm5, XMMWORD [wk(0)]   ; xmm5=tmp10L
+        movdqa  xmm4, XMMWORD [wk(1)]   ; xmm4=tmp10H
+
+        movdqa  xmm0,xmm5
+        movdqa  xmm7,xmm4
+        paddd   xmm5,xmm1               ; xmm5=data0L
+        paddd   xmm4,xmm3               ; xmm4=data0H
+        psubd   xmm0,xmm1               ; xmm0=data7L
+        psubd   xmm7,xmm3               ; xmm7=data7H
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm1=[PD_DESCALE_P2]
+
+        paddd   xmm5,xmm1
+        paddd   xmm4,xmm1
+        psrad   xmm5,DESCALE_P2
+        psrad   xmm4,DESCALE_P2
+        paddd   xmm0,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm0,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm5,xmm4             ; xmm5=data0=(00 10 20 30 40 50 60 70)
+        packssdw  xmm0,xmm7             ; xmm0=data7=(07 17 27 37 47 57 67 77)
+
+        movdqa  xmm3, XMMWORD [wk(4)]   ; xmm3=tmp11L
+        movdqa  xmm1, XMMWORD [wk(5)]   ; xmm1=tmp11H
+
+        movdqa  xmm4,xmm3
+        movdqa  xmm7,xmm1
+        paddd   xmm3,xmm2               ; xmm3=data1L
+        paddd   xmm1,xmm6               ; xmm1=data1H
+        psubd   xmm4,xmm2               ; xmm4=data6L
+        psubd   xmm7,xmm6               ; xmm7=data6H
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm2=[PD_DESCALE_P2]
+
+        paddd   xmm3,xmm2
+        paddd   xmm1,xmm2
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm4,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm7,DESCALE_P2
+
+        packssdw  xmm3,xmm1             ; xmm3=data1=(01 11 21 31 41 51 61 71)
+        packssdw  xmm4,xmm7             ; xmm4=data6=(06 16 26 36 46 56 66 76)
+
+        packsswb  xmm5,xmm4             ; xmm5=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        packsswb  xmm3,xmm0             ; xmm3=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm6, XMMWORD [wk(6)]   ; xmm6=tmp12L
+        movdqa  xmm2, XMMWORD [wk(7)]   ; xmm2=tmp12H
+        movdqa  xmm1, XMMWORD [wk(10)]  ; xmm1=tmp1L
+        movdqa  xmm7, XMMWORD [wk(11)]  ; xmm7=tmp1H
+
+        movdqa  XMMWORD [wk(0)], xmm5   ; wk(0)=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        movdqa  xmm4,xmm6
+        movdqa  xmm0,xmm2
+        paddd   xmm6,xmm1               ; xmm6=data2L
+        paddd   xmm2,xmm7               ; xmm2=data2H
+        psubd   xmm4,xmm1               ; xmm4=data5L
+        psubd   xmm0,xmm7               ; xmm0=data5H
+
+        movdqa  xmm5,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm5=[PD_DESCALE_P2]
+
+        paddd   xmm6,xmm5
+        paddd   xmm2,xmm5
+        psrad   xmm6,DESCALE_P2
+        psrad   xmm2,DESCALE_P2
+        paddd   xmm4,xmm5
+        paddd   xmm0,xmm5
+        psrad   xmm4,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        packssdw  xmm6,xmm2             ; xmm6=data2=(02 12 22 32 42 52 62 72)
+        packssdw  xmm4,xmm0             ; xmm4=data5=(05 15 25 35 45 55 65 75)
+
+        movdqa  xmm3, XMMWORD [wk(2)]   ; xmm3=tmp13L
+        movdqa  xmm1, XMMWORD [wk(3)]   ; xmm1=tmp13H
+        movdqa  xmm7, XMMWORD [wk(8)]   ; xmm7=tmp0L
+        movdqa  xmm5, XMMWORD [wk(9)]   ; xmm5=tmp0H
+
+        movdqa  xmm2,xmm3
+        movdqa  xmm0,xmm1
+        paddd   xmm3,xmm7               ; xmm3=data3L
+        paddd   xmm1,xmm5               ; xmm1=data3H
+        psubd   xmm2,xmm7               ; xmm2=data4L
+        psubd   xmm0,xmm5               ; xmm0=data4H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P2)]        ; xmm7=[PD_DESCALE_P2]
+
+        paddd   xmm3,xmm7
+        paddd   xmm1,xmm7
+        psrad   xmm3,DESCALE_P2
+        psrad   xmm1,DESCALE_P2
+        paddd   xmm2,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm2,DESCALE_P2
+        psrad   xmm0,DESCALE_P2
+
+        movdqa    xmm5,[GOTOFF(ebx,PB_CENTERJSAMP)]     ; xmm5=[PB_CENTERJSAMP]
+
+        packssdw  xmm3,xmm1             ; xmm3=data3=(03 13 23 33 43 53 63 73)
+        packssdw  xmm2,xmm0             ; xmm2=data4=(04 14 24 34 44 54 64 74)
+
+        movdqa    xmm7, XMMWORD [wk(0)] ; xmm7=(00 10 20 30 40 50 60 70 06 16 26 36 46 56 66 76)
+        movdqa    xmm1, XMMWORD [wk(1)] ; xmm1=(01 11 21 31 41 51 61 71 07 17 27 37 47 57 67 77)
+
+        packsswb  xmm6,xmm2             ; xmm6=(02 12 22 32 42 52 62 72 04 14 24 34 44 54 64 74)
+        packsswb  xmm3,xmm4             ; xmm3=(03 13 23 33 43 53 63 73 05 15 25 35 45 55 65 75)
+
+        paddb     xmm7,xmm5
+        paddb     xmm1,xmm5
+        paddb     xmm6,xmm5
+        paddb     xmm3,xmm5
+
+        movdqa    xmm0,xmm7     ; transpose coefficients(phase 1)
+        punpcklbw xmm7,xmm1     ; xmm7=(00 01 10 11 20 21 30 31 40 41 50 51 60 61 70 71)
+        punpckhbw xmm0,xmm1     ; xmm0=(06 07 16 17 26 27 36 37 46 47 56 57 66 67 76 77)
+        movdqa    xmm2,xmm6     ; transpose coefficients(phase 1)
+        punpcklbw xmm6,xmm3     ; xmm6=(02 03 12 13 22 23 32 33 42 43 52 53 62 63 72 73)
+        punpckhbw xmm2,xmm3     ; xmm2=(04 05 14 15 24 25 34 35 44 45 54 55 64 65 74 75)
+
+        movdqa    xmm4,xmm7     ; transpose coefficients(phase 2)
+        punpcklwd xmm7,xmm6     ; xmm7=(00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33)
+        punpckhwd xmm4,xmm6     ; xmm4=(40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73)
+        movdqa    xmm5,xmm2     ; transpose coefficients(phase 2)
+        punpcklwd xmm2,xmm0     ; xmm2=(04 05 06 07 14 15 16 17 24 25 26 27 34 35 36 37)
+        punpckhwd xmm5,xmm0     ; xmm5=(44 45 46 47 54 55 56 57 64 65 66 67 74 75 76 77)
+
+        movdqa    xmm1,xmm7     ; transpose coefficients(phase 3)
+        punpckldq xmm7,xmm2     ; xmm7=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
+        punpckhdq xmm1,xmm2     ; xmm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
+        movdqa    xmm3,xmm4     ; transpose coefficients(phase 3)
+        punpckldq xmm4,xmm5     ; xmm4=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
+        punpckhdq xmm3,xmm5     ; xmm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
+
+        pshufd  xmm6,xmm7,0x4E  ; xmm6=(10 11 12 13 14 15 16 17 00 01 02 03 04 05 06 07)
+        pshufd  xmm0,xmm1,0x4E  ; xmm0=(30 31 32 33 34 35 36 37 20 21 22 23 24 25 26 27)
+        pshufd  xmm2,xmm4,0x4E  ; xmm2=(50 51 52 53 54 55 56 57 40 41 42 43 44 45 46 47)
+        pshufd  xmm5,xmm3,0x4E  ; xmm5=(70 71 72 73 74 75 76 77 60 61 62 63 64 65 66 67)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm7
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm1
+        mov     edx, JSAMPROW [edi+4*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+6*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm6
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm0
+        mov     edx, JSAMPROW [edi+5*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+7*SIZEOF_JSAMPROW]
+        movq    XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE], xmm2
+        movq    XMM_MMWORD [esi+eax*SIZEOF_JSAMPLE], xmm5
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,706 @@
+;
+; jidctred.asm - reduced-size IDCT (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_mmx) PRIVATE
+
+EXTN(jconst_idct_red_mmx):
+
+PW_F184_MF076   times 2 dw  F_1_847,-F_0_765
+PW_F256_F089    times 2 dw  F_2_562, F_0_899
+PW_F106_MF217   times 2 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 2 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 2 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 2 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 2 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 2 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 2 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 2 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 2 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 8 db  CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_mmx (void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_MMWORD  ; mmword wk[WK_NUM]
+%define WK_NUM          2
+%define workspace       wk(0)-DCTSIZE2*SIZEOF_JCOEF
+                                        ; JCOEF workspace[DCTSIZE2]
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_mmx) PRIVATE
+
+EXTN(jsimd_idct_4x4_mmx):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_MMWORD)      ; align to 64 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [workspace]
+        pushpic ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input, store into work array.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+        lea     edi, [workspace]                        ; JCOEF *wsptr
+        mov     ecx, DCTSIZE/4                          ; ctr
+        alignx  16,7
+.columnloop:
+%ifndef NO_ZERO_COLUMN_TEST_4X4_MMX
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     mm1, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     mm0,mm1
+        packsswb mm0,mm0
+        movd    eax,mm0
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   mm0,PASS1_BITS
+
+        movq      mm2,mm0               ; mm0=in0=(00 01 02 03)
+        punpcklwd mm0,mm0               ; mm0=(00 00 01 01)
+        punpckhwd mm2,mm2               ; mm2=(02 02 03 03)
+
+        movq      mm1,mm0
+        punpckldq mm0,mm0               ; mm0=(00 00 00 00)
+        punpckhdq mm1,mm1               ; mm1=(01 01 01 01)
+        movq      mm3,mm2
+        punpckldq mm2,mm2               ; mm2=(02 02 02 02)
+        punpckhdq mm3,mm3               ; mm3=(03 03 03 03)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+        jmp     near .nextcolumn
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movq      mm4,mm0
+        movq      mm5,mm0
+        punpcklwd mm4,mm1
+        punpckhwd mm5,mm1
+        movq      mm0,mm4
+        movq      mm1,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F256_F089)]        ; mm4=(tmp2L)
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F256_F089)]        ; mm5=(tmp2H)
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F106_MF217)]       ; mm0=(tmp0L)
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F106_MF217)]       ; mm1=(tmp0H)
+
+        movq      mm6,mm2
+        movq      mm7,mm2
+        punpcklwd mm6,mm3
+        punpckhwd mm7,mm3
+        movq      mm2,mm6
+        movq      mm3,mm7
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm6=(tmp2L)
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm7=(tmp2H)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F145_MF021)]       ; mm2=(tmp0L)
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F145_MF021)]       ; mm3=(tmp0H)
+
+        paddd   mm6,mm4                 ; mm6=tmp2L
+        paddd   mm7,mm5                 ; mm7=tmp2H
+        paddd   mm2,mm0                 ; mm2=tmp0L
+        paddd   mm3,mm1                 ; mm3=tmp0H
+
+        movq    MMWORD [wk(0)], mm2     ; wk(0)=tmp0L
+        movq    MMWORD [wk(1)], mm3     ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movq    mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm4, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm0, MMWORD [MMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      mm1,mm1
+        pxor      mm2,mm2
+        punpcklwd mm1,mm4               ; mm1=tmp0L
+        punpckhwd mm2,mm4               ; mm2=tmp0H
+        psrad     mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+        psrad     mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+        movq      mm3,mm5               ; mm5=in2=z2
+        punpcklwd mm5,mm0               ; mm0=in6=z3
+        punpckhwd mm3,mm0
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F184_MF076)]       ; mm5=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F184_MF076)]       ; mm3=tmp2H
+
+        movq    mm4,mm1
+        movq    mm0,mm2
+        paddd   mm1,mm5                 ; mm1=tmp10L
+        paddd   mm2,mm3                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp12L
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        ; -- Final output stage
+
+        movq    mm5,mm1
+        movq    mm3,mm2
+        paddd   mm1,mm6                 ; mm1=data0L
+        paddd   mm2,mm7                 ; mm2=data0H
+        psubd   mm5,mm6                 ; mm5=data3L
+        psubd   mm3,mm7                 ; mm3=data3H
+
+        movq    mm6,[GOTOFF(ebx,PD_DESCALE_P1_4)]       ; mm6=[PD_DESCALE_P1_4]
+
+        paddd   mm1,mm6
+        paddd   mm2,mm6
+        psrad   mm1,DESCALE_P1_4
+        psrad   mm2,DESCALE_P1_4
+        paddd   mm5,mm6
+        paddd   mm3,mm6
+        psrad   mm5,DESCALE_P1_4
+        psrad   mm3,DESCALE_P1_4
+
+        packssdw  mm1,mm2               ; mm1=data0=(00 01 02 03)
+        packssdw  mm5,mm3               ; mm5=data3=(30 31 32 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp0L
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp0H
+
+        movq    mm2,mm4
+        movq    mm3,mm0
+        paddd   mm4,mm7                 ; mm4=data1L
+        paddd   mm0,mm6                 ; mm0=data1H
+        psubd   mm2,mm7                 ; mm2=data2L
+        psubd   mm3,mm6                 ; mm3=data2H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1_4)]       ; mm7=[PD_DESCALE_P1_4]
+
+        paddd   mm4,mm7
+        paddd   mm0,mm7
+        psrad   mm4,DESCALE_P1_4
+        psrad   mm0,DESCALE_P1_4
+        paddd   mm2,mm7
+        paddd   mm3,mm7
+        psrad   mm2,DESCALE_P1_4
+        psrad   mm3,DESCALE_P1_4
+
+        packssdw  mm4,mm0               ; mm4=data1=(10 11 12 13)
+        packssdw  mm2,mm3               ; mm2=data2=(20 21 22 23)
+
+        movq      mm6,mm1               ; transpose coefficients(phase 1)
+        punpcklwd mm1,mm4               ; mm1=(00 10 01 11)
+        punpckhwd mm6,mm4               ; mm6=(02 12 03 13)
+        movq      mm7,mm2               ; transpose coefficients(phase 1)
+        punpcklwd mm2,mm5               ; mm2=(20 30 21 31)
+        punpckhwd mm7,mm5               ; mm7=(22 32 23 33)
+
+        movq      mm0,mm1               ; transpose coefficients(phase 2)
+        punpckldq mm1,mm2               ; mm1=(00 10 20 30)
+        punpckhdq mm0,mm2               ; mm0=(01 11 21 31)
+        movq      mm3,mm6               ; transpose coefficients(phase 2)
+        punpckldq mm6,mm7               ; mm6=(02 12 22 32)
+        punpckhdq mm3,mm7               ; mm3=(03 13 23 33)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_JCOEF)], mm3
+
+.nextcolumn:
+        add     esi, byte 4*SIZEOF_JCOEF                ; coef_block
+        add     edx, byte 4*SIZEOF_ISLOW_MULT_TYPE      ; quantptr
+        add     edi, byte 4*DCTSIZE*SIZEOF_JCOEF        ; wsptr
+        dec     ecx                                     ; ctr
+        jnz     near .columnloop
+
+        ; ---- Pass 2: process rows from work array, store into output array.
+
+        mov     eax, [original_ebp]
+        lea     esi, [workspace]                        ; JCOEF *wsptr
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+
+        movq      mm4,mm0
+        movq      mm5,mm0
+        punpcklwd mm4,mm1
+        punpckhwd mm5,mm1
+        movq      mm0,mm4
+        movq      mm1,mm5
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F256_F089)]        ; mm4=(tmp2L)
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F256_F089)]        ; mm5=(tmp2H)
+        pmaddwd   mm0,[GOTOFF(ebx,PW_F106_MF217)]       ; mm0=(tmp0L)
+        pmaddwd   mm1,[GOTOFF(ebx,PW_F106_MF217)]       ; mm1=(tmp0H)
+
+        movq      mm6,mm2
+        movq      mm7,mm2
+        punpcklwd mm6,mm3
+        punpckhwd mm7,mm3
+        movq      mm2,mm6
+        movq      mm3,mm7
+        pmaddwd   mm6,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm6=(tmp2L)
+        pmaddwd   mm7,[GOTOFF(ebx,PW_MF060_MF050)]      ; mm7=(tmp2H)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F145_MF021)]       ; mm2=(tmp0L)
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F145_MF021)]       ; mm3=(tmp0H)
+
+        paddd   mm6,mm4                 ; mm6=tmp2L
+        paddd   mm7,mm5                 ; mm7=tmp2H
+        paddd   mm2,mm0                 ; mm2=tmp0L
+        paddd   mm3,mm1                 ; mm3=tmp0H
+
+        movq    MMWORD [wk(0)], mm2     ; wk(0)=tmp0L
+        movq    MMWORD [wk(1)], mm3     ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movq    mm4, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movq    mm0, MMWORD [MMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+
+        pxor      mm1,mm1
+        pxor      mm2,mm2
+        punpcklwd mm1,mm4               ; mm1=tmp0L
+        punpckhwd mm2,mm4               ; mm2=tmp0H
+        psrad     mm1,(16-CONST_BITS-1) ; psrad mm1,16 & pslld mm1,CONST_BITS+1
+        psrad     mm2,(16-CONST_BITS-1) ; psrad mm2,16 & pslld mm2,CONST_BITS+1
+
+        movq      mm3,mm5               ; mm5=in2=z2
+        punpcklwd mm5,mm0               ; mm0=in6=z3
+        punpckhwd mm3,mm0
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F184_MF076)]       ; mm5=tmp2L
+        pmaddwd   mm3,[GOTOFF(ebx,PW_F184_MF076)]       ; mm3=tmp2H
+
+        movq    mm4,mm1
+        movq    mm0,mm2
+        paddd   mm1,mm5                 ; mm1=tmp10L
+        paddd   mm2,mm3                 ; mm2=tmp10H
+        psubd   mm4,mm5                 ; mm4=tmp12L
+        psubd   mm0,mm3                 ; mm0=tmp12H
+
+        ; -- Final output stage
+
+        movq    mm5,mm1
+        movq    mm3,mm2
+        paddd   mm1,mm6                 ; mm1=data0L
+        paddd   mm2,mm7                 ; mm2=data0H
+        psubd   mm5,mm6                 ; mm5=data3L
+        psubd   mm3,mm7                 ; mm3=data3H
+
+        movq    mm6,[GOTOFF(ebx,PD_DESCALE_P2_4)]       ; mm6=[PD_DESCALE_P2_4]
+
+        paddd   mm1,mm6
+        paddd   mm2,mm6
+        psrad   mm1,DESCALE_P2_4
+        psrad   mm2,DESCALE_P2_4
+        paddd   mm5,mm6
+        paddd   mm3,mm6
+        psrad   mm5,DESCALE_P2_4
+        psrad   mm3,DESCALE_P2_4
+
+        packssdw  mm1,mm2               ; mm1=data0=(00 10 20 30)
+        packssdw  mm5,mm3               ; mm5=data3=(03 13 23 33)
+
+        movq    mm7, MMWORD [wk(0)]     ; mm7=tmp0L
+        movq    mm6, MMWORD [wk(1)]     ; mm6=tmp0H
+
+        movq    mm2,mm4
+        movq    mm3,mm0
+        paddd   mm4,mm7                 ; mm4=data1L
+        paddd   mm0,mm6                 ; mm0=data1H
+        psubd   mm2,mm7                 ; mm2=data2L
+        psubd   mm3,mm6                 ; mm3=data2H
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P2_4)]       ; mm7=[PD_DESCALE_P2_4]
+
+        paddd   mm4,mm7
+        paddd   mm0,mm7
+        psrad   mm4,DESCALE_P2_4
+        psrad   mm0,DESCALE_P2_4
+        paddd   mm2,mm7
+        paddd   mm3,mm7
+        psrad   mm2,DESCALE_P2_4
+        psrad   mm3,DESCALE_P2_4
+
+        packssdw  mm4,mm0               ; mm4=data1=(01 11 21 31)
+        packssdw  mm2,mm3               ; mm2=data2=(02 12 22 32)
+
+        movq      mm6,[GOTOFF(ebx,PB_CENTERJSAMP)]      ; mm6=[PB_CENTERJSAMP]
+
+        packsswb  mm1,mm2               ; mm1=(00 10 20 30 02 12 22 32)
+        packsswb  mm4,mm5               ; mm4=(01 11 21 31 03 13 23 33)
+        paddb     mm1,mm6
+        paddb     mm4,mm6
+
+        movq      mm7,mm1               ; transpose coefficients(phase 1)
+        punpcklbw mm1,mm4               ; mm1=(00 01 10 11 20 21 30 31)
+        punpckhbw mm7,mm4               ; mm7=(02 03 12 13 22 23 32 33)
+
+        movq      mm0,mm1               ; transpose coefficients(phase 2)
+        punpcklwd mm1,mm7               ; mm1=(00 01 02 03 10 11 12 13)
+        punpckhwd mm0,mm7               ; mm0=(20 21 22 23 30 31 32 33)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        movd    DWORD [edx+eax*SIZEOF_JSAMPLE], mm1
+        movd    DWORD [esi+eax*SIZEOF_JSAMPLE], mm0
+
+        psrlq   mm1,4*BYTE_BIT
+        psrlq   mm0,4*BYTE_BIT
+
+        mov     edx, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movd    DWORD [edx+eax*SIZEOF_JSAMPLE], mm1
+        movd    DWORD [esi+eax*SIZEOF_JSAMPLE], mm0
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_mmx (void *dct_table, JCOEFPTR coef_block,
+;                     JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_mmx) PRIVATE
+
+EXTN(jsimd_idct_2x2_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     edx, POINTER [dct_table(ebp)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(ebp)]         ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movq    mm0, MMWORD [MMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm0, MMWORD [MMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm2, MMWORD [MMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movq    mm3, MMWORD [MMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  mm2, MMWORD [MMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm3, MMWORD [MMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm0=(10 11 ** 13), mm1=(30 31 ** 33)
+        ; mm2=(50 51 ** 53), mm3=(70 71 ** 73)
+
+        pcmpeqd   mm7,mm7
+        pslld     mm7,WORD_BIT          ; mm7={0x0000 0xFFFF 0x0000 0xFFFF}
+
+        movq      mm4,mm0               ; mm4=(10 11 ** 13)
+        movq      mm5,mm2               ; mm5=(50 51 ** 53)
+        punpcklwd mm4,mm1               ; mm4=(10 30 11 31)
+        punpcklwd mm5,mm3               ; mm5=(50 70 51 71)
+        pmaddwd   mm4,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        psrld   mm0,WORD_BIT            ; mm0=(11 -- 13 --)
+        pand    mm1,mm7                 ; mm1=(-- 31 -- 33)
+        psrld   mm2,WORD_BIT            ; mm2=(51 -- 53 --)
+        pand    mm3,mm7                 ; mm3=(-- 71 -- 73)
+        por     mm0,mm1                 ; mm0=(11 31 13 33)
+        por     mm2,mm3                 ; mm2=(51 71 53 73)
+        pmaddwd mm0,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd mm2,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   mm4,mm5                 ; mm4=tmp0[col0 col1]
+
+        movq    mm6, MMWORD [MMBLOCK(1,1,esi,SIZEOF_JCOEF)]
+        movq    mm1, MMWORD [MMBLOCK(3,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm6, MMWORD [MMBLOCK(1,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm1, MMWORD [MMBLOCK(3,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movq    mm3, MMWORD [MMBLOCK(5,1,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(7,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm3, MMWORD [MMBLOCK(5,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(7,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm6=(** 15 ** 17), mm1=(** 35 ** 37)
+        ; mm3=(** 55 ** 57), mm5=(** 75 ** 77)
+
+        psrld   mm6,WORD_BIT            ; mm6=(15 -- 17 --)
+        pand    mm1,mm7                 ; mm1=(-- 35 -- 37)
+        psrld   mm3,WORD_BIT            ; mm3=(55 -- 57 --)
+        pand    mm5,mm7                 ; mm5=(-- 75 -- 77)
+        por     mm6,mm1                 ; mm6=(15 35 17 37)
+        por     mm3,mm5                 ; mm3=(55 75 57 77)
+        pmaddwd mm6,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd mm3,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   mm0,mm2                 ; mm0=tmp0[col1 col3]
+        paddd   mm6,mm3                 ; mm6=tmp0[col5 col7]
+
+        ; -- Even part
+
+        movq    mm1, MMWORD [MMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movq    mm5, MMWORD [MMBLOCK(0,1,esi,SIZEOF_JCOEF)]
+        pmullw  mm1, MMWORD [MMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  mm5, MMWORD [MMBLOCK(0,1,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; mm1=(00 01 ** 03), mm5=(** 05 ** 07)
+
+        movq    mm2,mm1                         ; mm2=(00 01 ** 03)
+        pslld   mm1,WORD_BIT                    ; mm1=(-- 00 -- **)
+        psrad   mm1,(WORD_BIT-CONST_BITS-2)     ; mm1=tmp10[col0 ****]
+
+        pand    mm2,mm7                         ; mm2=(-- 01 -- 03)
+        pand    mm5,mm7                         ; mm5=(-- 05 -- 07)
+        psrad   mm2,(WORD_BIT-CONST_BITS-2)     ; mm2=tmp10[col1 col3]
+        psrad   mm5,(WORD_BIT-CONST_BITS-2)     ; mm5=tmp10[col5 col7]
+
+        ; -- Final output stage
+
+        movq      mm3,mm1
+        paddd     mm1,mm4               ; mm1=data0[col0 ****]=(A0 **)
+        psubd     mm3,mm4               ; mm3=data1[col0 ****]=(B0 **)
+        punpckldq mm1,mm3               ; mm1=(A0 B0)
+
+        movq    mm7,[GOTOFF(ebx,PD_DESCALE_P1_2)]       ; mm7=[PD_DESCALE_P1_2]
+
+        movq    mm4,mm2
+        movq    mm3,mm5
+        paddd   mm2,mm0                 ; mm2=data0[col1 col3]=(A1 A3)
+        paddd   mm5,mm6                 ; mm5=data0[col5 col7]=(A5 A7)
+        psubd   mm4,mm0                 ; mm4=data1[col1 col3]=(B1 B3)
+        psubd   mm3,mm6                 ; mm3=data1[col5 col7]=(B5 B7)
+
+        paddd   mm1,mm7
+        psrad   mm1,DESCALE_P1_2
+
+        paddd   mm2,mm7
+        paddd   mm5,mm7
+        psrad   mm2,DESCALE_P1_2
+        psrad   mm5,DESCALE_P1_2
+        paddd   mm4,mm7
+        paddd   mm3,mm7
+        psrad   mm4,DESCALE_P1_2
+        psrad   mm3,DESCALE_P1_2
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     edi, JSAMPARRAY [output_buf(ebp)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(ebp)]
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  mm2,mm4               ; mm2=(A1 A3 B1 B3)
+        packssdw  mm5,mm3               ; mm5=(A5 A7 B5 B7)
+        pmaddwd   mm2,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   mm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd     mm2,mm5               ; mm2=tmp0[row0 row1]
+
+        ; -- Even part
+
+        pslld     mm1,(CONST_BITS+2)    ; mm1=tmp10[row0 row1]
+
+        ; -- Final output stage
+
+        movq      mm0,[GOTOFF(ebx,PD_DESCALE_P2_2)]     ; mm0=[PD_DESCALE_P2_2]
+
+        movq      mm6,mm1
+        paddd     mm1,mm2               ; mm1=data0[row0 row1]=(C0 C1)
+        psubd     mm6,mm2               ; mm6=data1[row0 row1]=(D0 D1)
+
+        paddd     mm1,mm0
+        paddd     mm6,mm0
+        psrad     mm1,DESCALE_P2_2
+        psrad     mm6,DESCALE_P2_2
+
+        movq      mm7,mm1               ; transpose coefficients
+        punpckldq mm1,mm6               ; mm1=(C0 D0)
+        punpckhdq mm7,mm6               ; mm7=(C1 D1)
+
+        packssdw  mm1,mm7               ; mm1=(C0 D0 C1 D1)
+        packsswb  mm1,mm1               ; mm1=(C0 D0 C1 D1 C0 D0 C1 D1)
+        paddb     mm1,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        movd    ecx,mm1
+        movd    ebx,mm1                 ; ebx=(C0 D0 C1 D1)
+        shr     ecx,2*BYTE_BIT          ; ecx=(C1 D1 -- --)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     WORD [edx+eax*SIZEOF_JSAMPLE], bx
+        mov     WORD [esi+eax*SIZEOF_JSAMPLE], cx
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,576 @@
+;
+; jidctred.asm - reduced-size IDCT (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_sse2) PRIVATE
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4 dw  F_1_847,-F_0_765
+PW_F256_F089    times 4 dw  F_2_562, F_0_899
+PW_F106_MF217   times 4 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 4 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 4 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 4 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 4 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 4 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 4 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 4 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 4 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+%define original_rbp    rbp+0
+%define wk(i)           rbp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_sse2) PRIVATE
+
+EXTN(jsimd_idct_4x4_sse2):
+        push    rbp
+        mov     rax,rsp                         ; rax = original rbp
+        sub     rsp, byte 4
+        and     rsp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [rsp],rax
+        mov     rbp,rsp                         ; rbp = aligned rbp
+        lea     rsp, [wk(0)]
+        collect_args
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        por     xmm0,xmm1
+        packsswb xmm0,xmm0
+        packsswb xmm0,xmm0
+        movd    eax,xmm0
+        test    rax,rax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm0,PASS1_BITS
+
+        movdqa    xmm3,xmm0     ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0     ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm3,xmm3     ; xmm3=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm1,xmm0,0x50  ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01)
+        pshufd  xmm0,xmm0,0xFA  ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03)
+        pshufd  xmm6,xmm3,0x50  ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05)
+        pshufd  xmm3,xmm3,0xFA  ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07)
+
+        jmp     near .column_end
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm5,xmm0
+        punpcklwd xmm4,xmm1
+        punpckhwd xmm5,xmm1
+        movdqa    xmm0,xmm4
+        movdqa    xmm1,xmm5
+        pmaddwd   xmm4,[rel PW_F256_F089]       ; xmm4=(tmp2L)
+        pmaddwd   xmm5,[rel PW_F256_F089]       ; xmm5=(tmp2H)
+        pmaddwd   xmm0,[rel PW_F106_MF217]      ; xmm0=(tmp0L)
+        pmaddwd   xmm1,[rel PW_F106_MF217]      ; xmm1=(tmp0H)
+
+        movdqa    xmm6,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm6,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm2,xmm6
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm6,[rel PW_MF060_MF050]     ; xmm6=(tmp2L)
+        pmaddwd   xmm7,[rel PW_MF060_MF050]     ; xmm7=(tmp2H)
+        pmaddwd   xmm2,[rel PW_F145_MF021]      ; xmm2=(tmp0L)
+        pmaddwd   xmm3,[rel PW_F145_MF021]      ; xmm3=(tmp0H)
+
+        paddd   xmm6,xmm4               ; xmm6=tmp2L
+        paddd   xmm7,xmm5               ; xmm7=tmp2H
+        paddd   xmm2,xmm0               ; xmm2=tmp0L
+        paddd   xmm3,xmm1               ; xmm3=tmp0H
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp0L
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm0, XMMWORD [XMMBLOCK(6,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(2,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(6,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      xmm1,xmm1
+        pxor      xmm2,xmm2
+        punpcklwd xmm1,xmm4             ; xmm1=tmp0L
+        punpckhwd xmm2,xmm4             ; xmm2=tmp0H
+        psrad     xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+        psrad     xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+        movdqa    xmm3,xmm5             ; xmm5=in2=z2
+        punpcklwd xmm5,xmm0             ; xmm0=in6=z3
+        punpckhwd xmm3,xmm0
+        pmaddwd   xmm5,[rel PW_F184_MF076]      ; xmm5=tmp2L
+        pmaddwd   xmm3,[rel PW_F184_MF076]      ; xmm3=tmp2H
+
+        movdqa  xmm4,xmm1
+        movdqa  xmm0,xmm2
+        paddd   xmm1,xmm5               ; xmm1=tmp10L
+        paddd   xmm2,xmm3               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp12L
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        ; -- Final output stage
+
+        movdqa  xmm5,xmm1
+        movdqa  xmm3,xmm2
+        paddd   xmm1,xmm6               ; xmm1=data0L
+        paddd   xmm2,xmm7               ; xmm2=data0H
+        psubd   xmm5,xmm6               ; xmm5=data3L
+        psubd   xmm3,xmm7               ; xmm3=data3H
+
+        movdqa  xmm6,[rel PD_DESCALE_P1_4]      ; xmm6=[rel PD_DESCALE_P1_4]
+
+        paddd   xmm1,xmm6
+        paddd   xmm2,xmm6
+        psrad   xmm1,DESCALE_P1_4
+        psrad   xmm2,DESCALE_P1_4
+        paddd   xmm5,xmm6
+        paddd   xmm3,xmm6
+        psrad   xmm5,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm1,xmm2             ; xmm1=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm5,xmm3             ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp0L
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp0H
+
+        movdqa  xmm2,xmm4
+        movdqa  xmm3,xmm0
+        paddd   xmm4,xmm7               ; xmm4=data1L
+        paddd   xmm0,xmm6               ; xmm0=data1H
+        psubd   xmm2,xmm7               ; xmm2=data2L
+        psubd   xmm3,xmm6               ; xmm3=data2H
+
+        movdqa  xmm7,[rel PD_DESCALE_P1_4]      ; xmm7=[rel PD_DESCALE_P1_4]
+
+        paddd   xmm4,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm4,DESCALE_P1_4
+        psrad   xmm0,DESCALE_P1_4
+        paddd   xmm2,xmm7
+        paddd   xmm3,xmm7
+        psrad   xmm2,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm4,xmm0             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm2,xmm3             ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+        movdqa    xmm6,xmm1     ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm4     ; xmm1=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4     ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm7,xmm2     ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm5     ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm7,xmm5     ; xmm7=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm2     ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm0,xmm2     ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33)
+        movdqa    xmm3,xmm6     ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7     ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm3,xmm7     ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     rax, [original_rbp]
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; -- Even part
+
+        pxor      xmm4,xmm4
+        punpcklwd xmm4,xmm1             ; xmm4=tmp0
+        psrad     xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+        ; -- Odd part
+
+        punpckhwd xmm1,xmm0
+        punpckhwd xmm6,xmm3
+        movdqa    xmm5,xmm1
+        movdqa    xmm2,xmm6
+        pmaddwd   xmm1,[rel PW_F256_F089]       ; xmm1=(tmp2)
+        pmaddwd   xmm6,[rel PW_MF060_MF050]     ; xmm6=(tmp2)
+        pmaddwd   xmm5,[rel PW_F106_MF217]      ; xmm5=(tmp0)
+        pmaddwd   xmm2,[rel PW_F145_MF021]      ; xmm2=(tmp0)
+
+        paddd     xmm6,xmm1             ; xmm6=tmp2
+        paddd     xmm2,xmm5             ; xmm2=tmp0
+
+        ; -- Even part
+
+        punpcklwd xmm0,xmm3
+        pmaddwd   xmm0,[rel PW_F184_MF076]      ; xmm0=tmp2
+
+        movdqa    xmm7,xmm4
+        paddd     xmm4,xmm0             ; xmm4=tmp10
+        psubd     xmm7,xmm0             ; xmm7=tmp12
+
+        ; -- Final output stage
+
+        movdqa  xmm1,[rel PD_DESCALE_P2_4]      ; xmm1=[rel PD_DESCALE_P2_4]
+
+        movdqa  xmm5,xmm4
+        movdqa  xmm3,xmm7
+        paddd   xmm4,xmm6               ; xmm4=data0=(00 10 20 30)
+        paddd   xmm7,xmm2               ; xmm7=data1=(01 11 21 31)
+        psubd   xmm5,xmm6               ; xmm5=data3=(03 13 23 33)
+        psubd   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+
+        paddd   xmm4,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm4,DESCALE_P2_4
+        psrad   xmm7,DESCALE_P2_4
+        paddd   xmm5,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm5,DESCALE_P2_4
+        psrad   xmm3,DESCALE_P2_4
+
+        packssdw  xmm4,xmm3             ; xmm4=(00 10 20 30 02 12 22 32)
+        packssdw  xmm7,xmm5             ; xmm7=(01 11 21 31 03 13 23 33)
+
+        movdqa    xmm0,xmm4             ; transpose coefficients(phase 1)
+        punpcklwd xmm4,xmm7             ; xmm4=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm0,xmm7             ; xmm0=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm6,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm6,xmm0             ; xmm6=(20 21 22 23 30 31 32 33)
+
+        packsswb  xmm4,xmm6             ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+        paddb     xmm4,[rel PB_CENTERJSAMP]
+
+        pshufd    xmm2,xmm4,0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+        pshufd    xmm1,xmm4,0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+        pshufd    xmm3,xmm4,0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm4
+        movd    XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm2
+        mov     rdx, JSAMPROW [rdi+2*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+3*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [rdx+rax*SIZEOF_JSAMPLE], xmm1
+        movd    XMM_DWORD [rsi+rax*SIZEOF_JSAMPLE], xmm3
+
+        uncollect_args
+        mov     rsp,rbp         ; rsp <- aligned rbp
+        pop     rsp             ; rsp <- original rbp
+        pop     rbp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+; r10 = void *dct_table
+; r11 = JCOEFPTR coef_block
+; r12 = JSAMPARRAY output_buf
+; r13 = JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_sse2) PRIVATE
+
+EXTN(jsimd_idct_2x2_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     rdx, r10                ; quantptr
+        mov     rsi, r11                ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,rsi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+        ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+        pcmpeqd   xmm7,xmm7
+        pslld     xmm7,WORD_BIT         ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+        movdqa    xmm4,xmm0             ; xmm4=(10 11 ** 13 ** 15 ** 17)
+        movdqa    xmm5,xmm2             ; xmm5=(50 51 ** 53 ** 55 ** 57)
+        punpcklwd xmm4,xmm1             ; xmm4=(10 30 11 31 ** ** 13 33)
+        punpcklwd xmm5,xmm3             ; xmm5=(50 70 51 71 ** ** 53 73)
+        pmaddwd   xmm4,[rel PW_F362_MF127]
+        pmaddwd   xmm5,[rel PW_F085_MF072]
+
+        psrld   xmm0,WORD_BIT           ; xmm0=(11 -- 13 -- 15 -- 17 --)
+        pand    xmm1,xmm7               ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+        psrld   xmm2,WORD_BIT           ; xmm2=(51 -- 53 -- 55 -- 57 --)
+        pand    xmm3,xmm7               ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+        por     xmm0,xmm1               ; xmm0=(11 31 13 33 15 35 17 37)
+        por     xmm2,xmm3               ; xmm2=(51 71 53 73 55 75 57 77)
+        pmaddwd xmm0,[rel PW_F362_MF127]
+        pmaddwd xmm2,[rel PW_F085_MF072]
+
+        paddd   xmm4,xmm5               ; xmm4=tmp0[col0 col1 **** col3]
+        paddd   xmm0,xmm2               ; xmm0=tmp0[col1 col3 col5 col7]
+
+        ; -- Even part
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_JCOEF)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+        movdqa  xmm1,xmm6               ; xmm1=(00 01 ** 03 ** 05 ** 07)
+        pslld   xmm6,WORD_BIT           ; xmm6=(-- 00 -- ** -- ** -- **)
+        pand    xmm1,xmm7               ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+        psrad   xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****]
+        psrad   xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7]
+
+        ; -- Final output stage
+
+        movdqa  xmm3,xmm6
+        movdqa  xmm5,xmm1
+        paddd   xmm6,xmm4       ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **)
+        paddd   xmm1,xmm0       ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7)
+        psubd   xmm3,xmm4       ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **)
+        psubd   xmm5,xmm0       ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7)
+
+        movdqa  xmm2,[rel PD_DESCALE_P1_2]      ; xmm2=[rel PD_DESCALE_P1_2]
+
+        punpckldq  xmm6,xmm3            ; xmm6=(A0 B0 ** **)
+
+        movdqa     xmm7,xmm1
+        punpcklqdq xmm1,xmm5            ; xmm1=(A1 A3 B1 B3)
+        punpckhqdq xmm7,xmm5            ; xmm7=(A5 A7 B5 B7)
+
+        paddd   xmm6,xmm2
+        psrad   xmm6,DESCALE_P1_2
+
+        paddd   xmm1,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm1,DESCALE_P1_2
+        psrad   xmm7,DESCALE_P1_2
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [rsi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     rdi, r12        ; (JSAMPROW *)
+        mov     eax, r13d
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  xmm1,xmm1             ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+        packssdw  xmm7,xmm7             ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+        pmaddwd   xmm1,[rel PW_F362_MF127]
+        pmaddwd   xmm7,[rel PW_F085_MF072]
+
+        paddd     xmm1,xmm7             ; xmm1=tmp0[row0 row1 row0 row1]
+
+        ; -- Even part
+
+        pslld     xmm6,(CONST_BITS+2)   ; xmm6=tmp10[row0 row1 **** ****]
+
+        ; -- Final output stage
+
+        movdqa    xmm4,xmm6
+        paddd     xmm6,xmm1     ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **)
+        psubd     xmm4,xmm1     ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **)
+
+        punpckldq xmm6,xmm4     ; xmm6=(C0 D0 C1 D1)
+
+        paddd     xmm6,[rel PD_DESCALE_P2_2]
+        psrad     xmm6,DESCALE_P2_2
+
+        packssdw  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+        packsswb  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+        paddb     xmm6,[rel PB_CENTERJSAMP]
+
+        pextrw  ebx,xmm6,0x00           ; ebx=(C0 D0 -- --)
+        pextrw  ecx,xmm6,0x01           ; ecx=(C1 D1 -- --)
+
+        mov     rdx, JSAMPROW [rdi+0*SIZEOF_JSAMPROW]
+        mov     rsi, JSAMPROW [rdi+1*SIZEOF_JSAMPROW]
+        mov     WORD [rdx+rax*SIZEOF_JSAMPLE], bx
+        mov     WORD [rsi+rax*SIZEOF_JSAMPLE], cx
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jidctred-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jidctred-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,594 @@
+;
+; jidctred.asm - reduced-size IDCT (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; This file contains inverse-DCT routines that produce reduced-size
+; output: either 4x4 or 2x2 pixels from an 8x8 DCT block.
+; The following code is based directly on the IJG's original jidctred.c;
+; see the jidctred.c for more details.
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+
+%define CONST_BITS      13
+%define PASS1_BITS      2
+
+%define DESCALE_P1_4    (CONST_BITS-PASS1_BITS+1)
+%define DESCALE_P2_4    (CONST_BITS+PASS1_BITS+3+1)
+%define DESCALE_P1_2    (CONST_BITS-PASS1_BITS+2)
+%define DESCALE_P2_2    (CONST_BITS+PASS1_BITS+3+2)
+
+%if CONST_BITS == 13
+F_0_211 equ      1730           ; FIX(0.211164243)
+F_0_509 equ      4176           ; FIX(0.509795579)
+F_0_601 equ      4926           ; FIX(0.601344887)
+F_0_720 equ      5906           ; FIX(0.720959822)
+F_0_765 equ      6270           ; FIX(0.765366865)
+F_0_850 equ      6967           ; FIX(0.850430095)
+F_0_899 equ      7373           ; FIX(0.899976223)
+F_1_061 equ      8697           ; FIX(1.061594337)
+F_1_272 equ     10426           ; FIX(1.272758580)
+F_1_451 equ     11893           ; FIX(1.451774981)
+F_1_847 equ     15137           ; FIX(1.847759065)
+F_2_172 equ     17799           ; FIX(2.172734803)
+F_2_562 equ     20995           ; FIX(2.562915447)
+F_3_624 equ     29692           ; FIX(3.624509785)
+%else
+; NASM cannot do compile-time arithmetic on floating-point constants.
+%define DESCALE(x,n)  (((x)+(1<<((n)-1)))>>(n))
+F_0_211 equ     DESCALE( 226735879,30-CONST_BITS)       ; FIX(0.211164243)
+F_0_509 equ     DESCALE( 547388834,30-CONST_BITS)       ; FIX(0.509795579)
+F_0_601 equ     DESCALE( 645689155,30-CONST_BITS)       ; FIX(0.601344887)
+F_0_720 equ     DESCALE( 774124714,30-CONST_BITS)       ; FIX(0.720959822)
+F_0_765 equ     DESCALE( 821806413,30-CONST_BITS)       ; FIX(0.765366865)
+F_0_850 equ     DESCALE( 913142361,30-CONST_BITS)       ; FIX(0.850430095)
+F_0_899 equ     DESCALE( 966342111,30-CONST_BITS)       ; FIX(0.899976223)
+F_1_061 equ     DESCALE(1139878239,30-CONST_BITS)       ; FIX(1.061594337)
+F_1_272 equ     DESCALE(1366614119,30-CONST_BITS)       ; FIX(1.272758580)
+F_1_451 equ     DESCALE(1558831516,30-CONST_BITS)       ; FIX(1.451774981)
+F_1_847 equ     DESCALE(1984016188,30-CONST_BITS)       ; FIX(1.847759065)
+F_2_172 equ     DESCALE(2332956230,30-CONST_BITS)       ; FIX(2.172734803)
+F_2_562 equ     DESCALE(2751909506,30-CONST_BITS)       ; FIX(2.562915447)
+F_3_624 equ     DESCALE(3891787747,30-CONST_BITS)       ; FIX(3.624509785)
+%endif
+
+; --------------------------------------------------------------------------
+        SECTION SEG_CONST
+
+        alignz  16
+        global  EXTN(jconst_idct_red_sse2) PRIVATE
+
+EXTN(jconst_idct_red_sse2):
+
+PW_F184_MF076   times 4 dw  F_1_847,-F_0_765
+PW_F256_F089    times 4 dw  F_2_562, F_0_899
+PW_F106_MF217   times 4 dw  F_1_061,-F_2_172
+PW_MF060_MF050  times 4 dw -F_0_601,-F_0_509
+PW_F145_MF021   times 4 dw  F_1_451,-F_0_211
+PW_F362_MF127   times 4 dw  F_3_624,-F_1_272
+PW_F085_MF072   times 4 dw  F_0_850,-F_0_720
+PD_DESCALE_P1_4 times 4 dd  1 << (DESCALE_P1_4-1)
+PD_DESCALE_P2_4 times 4 dd  1 << (DESCALE_P2_4-1)
+PD_DESCALE_P1_2 times 4 dd  1 << (DESCALE_P1_2-1)
+PD_DESCALE_P2_2 times 4 dd  1 << (DESCALE_P2_2-1)
+PB_CENTERJSAMP  times 16 db CENTERJSAMPLE
+
+        alignz  16
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 4x4 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_4x4_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+%define original_ebp    ebp+0
+%define wk(i)           ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
+%define WK_NUM          2
+
+        align   16
+        global  EXTN(jsimd_idct_4x4_sse2) PRIVATE
+
+EXTN(jsimd_idct_4x4_sse2):
+        push    ebp
+        mov     eax,esp                         ; eax = original ebp
+        sub     esp, byte 4
+        and     esp, byte (-SIZEOF_XMMWORD)     ; align to 128 bits
+        mov     [esp],eax
+        mov     ebp,esp                         ; ebp = aligned ebp
+        lea     esp, [wk(0)]
+        pushpic ebx
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+;       mov     eax, [original_ebp]
+        mov     edx, POINTER [dct_table(eax)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(eax)]         ; inptr
+
+%ifndef NO_ZERO_COLUMN_TEST_4X4_SSE2
+        mov     eax, DWORD [DWBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        or      eax, DWORD [DWBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        jnz     short .columnDCT
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        por     xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        por     xmm1, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        por     xmm0,xmm1
+        packsswb xmm0,xmm0
+        packsswb xmm0,xmm0
+        movd    eax,xmm0
+        test    eax,eax
+        jnz     short .columnDCT
+
+        ; -- AC terms all zero
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        psllw   xmm0,PASS1_BITS
+
+        movdqa    xmm3,xmm0     ; xmm0=in0=(00 01 02 03 04 05 06 07)
+        punpcklwd xmm0,xmm0     ; xmm0=(00 00 01 01 02 02 03 03)
+        punpckhwd xmm3,xmm3     ; xmm3=(04 04 05 05 06 06 07 07)
+
+        pshufd  xmm1,xmm0,0x50  ; xmm1=[col0 col1]=(00 00 00 00 01 01 01 01)
+        pshufd  xmm0,xmm0,0xFA  ; xmm0=[col2 col3]=(02 02 02 02 03 03 03 03)
+        pshufd  xmm6,xmm3,0x50  ; xmm6=[col4 col5]=(04 04 04 04 05 05 05 05)
+        pshufd  xmm3,xmm3,0xFA  ; xmm3=[col6 col7]=(06 06 06 06 07 07 07 07)
+
+        jmp     near .column_end
+        alignx  16,7
+%endif
+.columnDCT:
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        movdqa    xmm4,xmm0
+        movdqa    xmm5,xmm0
+        punpcklwd xmm4,xmm1
+        punpckhwd xmm5,xmm1
+        movdqa    xmm0,xmm4
+        movdqa    xmm1,xmm5
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F256_F089)]       ; xmm4=(tmp2L)
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F256_F089)]       ; xmm5=(tmp2H)
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm0=(tmp0L)
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm1=(tmp0H)
+
+        movdqa    xmm6,xmm2
+        movdqa    xmm7,xmm2
+        punpcklwd xmm6,xmm3
+        punpckhwd xmm7,xmm3
+        movdqa    xmm2,xmm6
+        movdqa    xmm3,xmm7
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm6=(tmp2L)
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm7=(tmp2H)
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm2=(tmp0L)
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm3=(tmp0H)
+
+        paddd   xmm6,xmm4               ; xmm6=tmp2L
+        paddd   xmm7,xmm5               ; xmm7=tmp2H
+        paddd   xmm2,xmm0               ; xmm2=tmp0L
+        paddd   xmm3,xmm1               ; xmm3=tmp0H
+
+        movdqa  XMMWORD [wk(0)], xmm2   ; wk(0)=tmp0L
+        movdqa  XMMWORD [wk(1)], xmm3   ; wk(1)=tmp0H
+
+        ; -- Even part
+
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm0, XMMWORD [XMMBLOCK(6,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm4, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm5, XMMWORD [XMMBLOCK(2,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(6,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        pxor      xmm1,xmm1
+        pxor      xmm2,xmm2
+        punpcklwd xmm1,xmm4             ; xmm1=tmp0L
+        punpckhwd xmm2,xmm4             ; xmm2=tmp0H
+        psrad     xmm1,(16-CONST_BITS-1) ; psrad xmm1,16 & pslld xmm1,CONST_BITS+1
+        psrad     xmm2,(16-CONST_BITS-1) ; psrad xmm2,16 & pslld xmm2,CONST_BITS+1
+
+        movdqa    xmm3,xmm5             ; xmm5=in2=z2
+        punpcklwd xmm5,xmm0             ; xmm0=in6=z3
+        punpckhwd xmm3,xmm0
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm5=tmp2L
+        pmaddwd   xmm3,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm3=tmp2H
+
+        movdqa  xmm4,xmm1
+        movdqa  xmm0,xmm2
+        paddd   xmm1,xmm5               ; xmm1=tmp10L
+        paddd   xmm2,xmm3               ; xmm2=tmp10H
+        psubd   xmm4,xmm5               ; xmm4=tmp12L
+        psubd   xmm0,xmm3               ; xmm0=tmp12H
+
+        ; -- Final output stage
+
+        movdqa  xmm5,xmm1
+        movdqa  xmm3,xmm2
+        paddd   xmm1,xmm6               ; xmm1=data0L
+        paddd   xmm2,xmm7               ; xmm2=data0H
+        psubd   xmm5,xmm6               ; xmm5=data3L
+        psubd   xmm3,xmm7               ; xmm3=data3H
+
+        movdqa  xmm6,[GOTOFF(ebx,PD_DESCALE_P1_4)]      ; xmm6=[PD_DESCALE_P1_4]
+
+        paddd   xmm1,xmm6
+        paddd   xmm2,xmm6
+        psrad   xmm1,DESCALE_P1_4
+        psrad   xmm2,DESCALE_P1_4
+        paddd   xmm5,xmm6
+        paddd   xmm3,xmm6
+        psrad   xmm5,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm1,xmm2             ; xmm1=data0=(00 01 02 03 04 05 06 07)
+        packssdw  xmm5,xmm3             ; xmm5=data3=(30 31 32 33 34 35 36 37)
+
+        movdqa  xmm7, XMMWORD [wk(0)]   ; xmm7=tmp0L
+        movdqa  xmm6, XMMWORD [wk(1)]   ; xmm6=tmp0H
+
+        movdqa  xmm2,xmm4
+        movdqa  xmm3,xmm0
+        paddd   xmm4,xmm7               ; xmm4=data1L
+        paddd   xmm0,xmm6               ; xmm0=data1H
+        psubd   xmm2,xmm7               ; xmm2=data2L
+        psubd   xmm3,xmm6               ; xmm3=data2H
+
+        movdqa  xmm7,[GOTOFF(ebx,PD_DESCALE_P1_4)]      ; xmm7=[PD_DESCALE_P1_4]
+
+        paddd   xmm4,xmm7
+        paddd   xmm0,xmm7
+        psrad   xmm4,DESCALE_P1_4
+        psrad   xmm0,DESCALE_P1_4
+        paddd   xmm2,xmm7
+        paddd   xmm3,xmm7
+        psrad   xmm2,DESCALE_P1_4
+        psrad   xmm3,DESCALE_P1_4
+
+        packssdw  xmm4,xmm0             ; xmm4=data1=(10 11 12 13 14 15 16 17)
+        packssdw  xmm2,xmm3             ; xmm2=data2=(20 21 22 23 24 25 26 27)
+
+        movdqa    xmm6,xmm1     ; transpose coefficients(phase 1)
+        punpcklwd xmm1,xmm4     ; xmm1=(00 10 01 11 02 12 03 13)
+        punpckhwd xmm6,xmm4     ; xmm6=(04 14 05 15 06 16 07 17)
+        movdqa    xmm7,xmm2     ; transpose coefficients(phase 1)
+        punpcklwd xmm2,xmm5     ; xmm2=(20 30 21 31 22 32 23 33)
+        punpckhwd xmm7,xmm5     ; xmm7=(24 34 25 35 26 36 27 37)
+
+        movdqa    xmm0,xmm1     ; transpose coefficients(phase 2)
+        punpckldq xmm1,xmm2     ; xmm1=[col0 col1]=(00 10 20 30 01 11 21 31)
+        punpckhdq xmm0,xmm2     ; xmm0=[col2 col3]=(02 12 22 32 03 13 23 33)
+        movdqa    xmm3,xmm6     ; transpose coefficients(phase 2)
+        punpckldq xmm6,xmm7     ; xmm6=[col4 col5]=(04 14 24 34 05 15 25 35)
+        punpckhdq xmm3,xmm7     ; xmm3=[col6 col7]=(06 16 26 36 07 17 27 37)
+.column_end:
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     eax, [original_ebp]
+        mov     edi, JSAMPARRAY [output_buf(eax)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(eax)]
+
+        ; -- Even part
+
+        pxor      xmm4,xmm4
+        punpcklwd xmm4,xmm1             ; xmm4=tmp0
+        psrad     xmm4,(16-CONST_BITS-1) ; psrad xmm4,16 & pslld xmm4,CONST_BITS+1
+
+        ; -- Odd part
+
+        punpckhwd xmm1,xmm0
+        punpckhwd xmm6,xmm3
+        movdqa    xmm5,xmm1
+        movdqa    xmm2,xmm6
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F256_F089)]       ; xmm1=(tmp2)
+        pmaddwd   xmm6,[GOTOFF(ebx,PW_MF060_MF050)]     ; xmm6=(tmp2)
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F106_MF217)]      ; xmm5=(tmp0)
+        pmaddwd   xmm2,[GOTOFF(ebx,PW_F145_MF021)]      ; xmm2=(tmp0)
+
+        paddd     xmm6,xmm1             ; xmm6=tmp2
+        paddd     xmm2,xmm5             ; xmm2=tmp0
+
+        ; -- Even part
+
+        punpcklwd xmm0,xmm3
+        pmaddwd   xmm0,[GOTOFF(ebx,PW_F184_MF076)]      ; xmm0=tmp2
+
+        movdqa    xmm7,xmm4
+        paddd     xmm4,xmm0             ; xmm4=tmp10
+        psubd     xmm7,xmm0             ; xmm7=tmp12
+
+        ; -- Final output stage
+
+        movdqa  xmm1,[GOTOFF(ebx,PD_DESCALE_P2_4)]      ; xmm1=[PD_DESCALE_P2_4]
+
+        movdqa  xmm5,xmm4
+        movdqa  xmm3,xmm7
+        paddd   xmm4,xmm6               ; xmm4=data0=(00 10 20 30)
+        paddd   xmm7,xmm2               ; xmm7=data1=(01 11 21 31)
+        psubd   xmm5,xmm6               ; xmm5=data3=(03 13 23 33)
+        psubd   xmm3,xmm2               ; xmm3=data2=(02 12 22 32)
+
+        paddd   xmm4,xmm1
+        paddd   xmm7,xmm1
+        psrad   xmm4,DESCALE_P2_4
+        psrad   xmm7,DESCALE_P2_4
+        paddd   xmm5,xmm1
+        paddd   xmm3,xmm1
+        psrad   xmm5,DESCALE_P2_4
+        psrad   xmm3,DESCALE_P2_4
+
+        packssdw  xmm4,xmm3             ; xmm4=(00 10 20 30 02 12 22 32)
+        packssdw  xmm7,xmm5             ; xmm7=(01 11 21 31 03 13 23 33)
+
+        movdqa    xmm0,xmm4             ; transpose coefficients(phase 1)
+        punpcklwd xmm4,xmm7             ; xmm4=(00 01 10 11 20 21 30 31)
+        punpckhwd xmm0,xmm7             ; xmm0=(02 03 12 13 22 23 32 33)
+
+        movdqa    xmm6,xmm4             ; transpose coefficients(phase 2)
+        punpckldq xmm4,xmm0             ; xmm4=(00 01 02 03 10 11 12 13)
+        punpckhdq xmm6,xmm0             ; xmm6=(20 21 22 23 30 31 32 33)
+
+        packsswb  xmm4,xmm6             ; xmm4=(00 01 02 03 10 11 12 13 20 ..)
+        paddb     xmm4,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        pshufd    xmm2,xmm4,0x39        ; xmm2=(10 11 12 13 20 21 22 23 30 ..)
+        pshufd    xmm1,xmm4,0x4E        ; xmm1=(20 21 22 23 30 31 32 33 00 ..)
+        pshufd    xmm3,xmm4,0x93        ; xmm3=(30 31 32 33 00 01 02 03 10 ..)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm4
+        movd    XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm2
+        mov     edx, JSAMPROW [edi+2*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+3*SIZEOF_JSAMPROW]
+        movd    XMM_DWORD [edx+eax*SIZEOF_JSAMPLE], xmm1
+        movd    XMM_DWORD [esi+eax*SIZEOF_JSAMPLE], xmm3
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+        poppic  ebx
+        mov     esp,ebp         ; esp <- aligned ebp
+        pop     esp             ; esp <- original ebp
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Perform dequantization and inverse DCT on one block of coefficients,
+; producing a reduced-size 2x2 output block.
+;
+; GLOBAL(void)
+; jsimd_idct_2x2_sse2 (void *dct_table, JCOEFPTR coef_block,
+;                      JSAMPARRAY output_buf, JDIMENSION output_col)
+;
+
+%define dct_table(b)    (b)+8           ; void *dct_table
+%define coef_block(b)   (b)+12          ; JCOEFPTR coef_block
+%define output_buf(b)   (b)+16          ; JSAMPARRAY output_buf
+%define output_col(b)   (b)+20          ; JDIMENSION output_col
+
+        align   16
+        global  EXTN(jsimd_idct_2x2_sse2) PRIVATE
+
+EXTN(jsimd_idct_2x2_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        get_GOT ebx             ; get GOT address
+
+        ; ---- Pass 1: process columns from input.
+
+        mov     edx, POINTER [dct_table(ebp)]           ; quantptr
+        mov     esi, JCOEFPTR [coef_block(ebp)]         ; inptr
+
+        ; | input:                  | result:        |
+        ; | 00 01 ** 03 ** 05 ** 07 |                |
+        ; | 10 11 ** 13 ** 15 ** 17 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 30 31 ** 33 ** 35 ** 37 | A0 A1 A3 A5 A7 |
+        ; | ** ** ** ** ** ** ** ** | B0 B1 B3 B5 B7 |
+        ; | 50 51 ** 53 ** 55 ** 57 |                |
+        ; | ** ** ** ** ** ** ** ** |                |
+        ; | 70 71 ** 73 ** 75 ** 77 |                |
+
+        ; -- Odd part
+
+        movdqa  xmm0, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm1, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm0, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm1, XMMWORD [XMMBLOCK(3,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        movdqa  xmm2, XMMWORD [XMMBLOCK(5,0,esi,SIZEOF_JCOEF)]
+        movdqa  xmm3, XMMWORD [XMMBLOCK(7,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm2, XMMWORD [XMMBLOCK(5,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+        pmullw  xmm3, XMMWORD [XMMBLOCK(7,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm0=(10 11 ** 13 ** 15 ** 17), xmm1=(30 31 ** 33 ** 35 ** 37)
+        ; xmm2=(50 51 ** 53 ** 55 ** 57), xmm3=(70 71 ** 73 ** 75 ** 77)
+
+        pcmpeqd   xmm7,xmm7
+        pslld     xmm7,WORD_BIT         ; xmm7={0x0000 0xFFFF 0x0000 0xFFFF ..}
+
+        movdqa    xmm4,xmm0             ; xmm4=(10 11 ** 13 ** 15 ** 17)
+        movdqa    xmm5,xmm2             ; xmm5=(50 51 ** 53 ** 55 ** 57)
+        punpcklwd xmm4,xmm1             ; xmm4=(10 30 11 31 ** ** 13 33)
+        punpcklwd xmm5,xmm3             ; xmm5=(50 70 51 71 ** ** 53 73)
+        pmaddwd   xmm4,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   xmm5,[GOTOFF(ebx,PW_F085_MF072)]
+
+        psrld   xmm0,WORD_BIT           ; xmm0=(11 -- 13 -- 15 -- 17 --)
+        pand    xmm1,xmm7               ; xmm1=(-- 31 -- 33 -- 35 -- 37)
+        psrld   xmm2,WORD_BIT           ; xmm2=(51 -- 53 -- 55 -- 57 --)
+        pand    xmm3,xmm7               ; xmm3=(-- 71 -- 73 -- 75 -- 77)
+        por     xmm0,xmm1               ; xmm0=(11 31 13 33 15 35 17 37)
+        por     xmm2,xmm3               ; xmm2=(51 71 53 73 55 75 57 77)
+        pmaddwd xmm0,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd xmm2,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd   xmm4,xmm5               ; xmm4=tmp0[col0 col1 **** col3]
+        paddd   xmm0,xmm2               ; xmm0=tmp0[col1 col3 col5 col7]
+
+        ; -- Even part
+
+        movdqa  xmm6, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_JCOEF)]
+        pmullw  xmm6, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_ISLOW_MULT_TYPE)]
+
+        ; xmm6=(00 01 ** 03 ** 05 ** 07)
+
+        movdqa  xmm1,xmm6               ; xmm1=(00 01 ** 03 ** 05 ** 07)
+        pslld   xmm6,WORD_BIT           ; xmm6=(-- 00 -- ** -- ** -- **)
+        pand    xmm1,xmm7               ; xmm1=(-- 01 -- 03 -- 05 -- 07)
+        psrad   xmm6,(WORD_BIT-CONST_BITS-2) ; xmm6=tmp10[col0 **** **** ****]
+        psrad   xmm1,(WORD_BIT-CONST_BITS-2) ; xmm1=tmp10[col1 col3 col5 col7]
+
+        ; -- Final output stage
+
+        movdqa  xmm3,xmm6
+        movdqa  xmm5,xmm1
+        paddd   xmm6,xmm4       ; xmm6=data0[col0 **** **** ****]=(A0 ** ** **)
+        paddd   xmm1,xmm0       ; xmm1=data0[col1 col3 col5 col7]=(A1 A3 A5 A7)
+        psubd   xmm3,xmm4       ; xmm3=data1[col0 **** **** ****]=(B0 ** ** **)
+        psubd   xmm5,xmm0       ; xmm5=data1[col1 col3 col5 col7]=(B1 B3 B5 B7)
+
+        movdqa  xmm2,[GOTOFF(ebx,PD_DESCALE_P1_2)]      ; xmm2=[PD_DESCALE_P1_2]
+
+        punpckldq  xmm6,xmm3            ; xmm6=(A0 B0 ** **)
+
+        movdqa     xmm7,xmm1
+        punpcklqdq xmm1,xmm5            ; xmm1=(A1 A3 B1 B3)
+        punpckhqdq xmm7,xmm5            ; xmm7=(A5 A7 B5 B7)
+
+        paddd   xmm6,xmm2
+        psrad   xmm6,DESCALE_P1_2
+
+        paddd   xmm1,xmm2
+        paddd   xmm7,xmm2
+        psrad   xmm1,DESCALE_P1_2
+        psrad   xmm7,DESCALE_P1_2
+
+        ; -- Prefetch the next coefficient block
+
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 0*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 1*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 2*32]
+        prefetchnta [esi + DCTSIZE2*SIZEOF_JCOEF + 3*32]
+
+        ; ---- Pass 2: process rows, store into output array.
+
+        mov     edi, JSAMPARRAY [output_buf(ebp)]       ; (JSAMPROW *)
+        mov     eax, JDIMENSION [output_col(ebp)]
+
+        ; | input:| result:|
+        ; | A0 B0 |        |
+        ; | A1 B1 | C0 C1  |
+        ; | A3 B3 | D0 D1  |
+        ; | A5 B5 |        |
+        ; | A7 B7 |        |
+
+        ; -- Odd part
+
+        packssdw  xmm1,xmm1             ; xmm1=(A1 A3 B1 B3 A1 A3 B1 B3)
+        packssdw  xmm7,xmm7             ; xmm7=(A5 A7 B5 B7 A5 A7 B5 B7)
+        pmaddwd   xmm1,[GOTOFF(ebx,PW_F362_MF127)]
+        pmaddwd   xmm7,[GOTOFF(ebx,PW_F085_MF072)]
+
+        paddd     xmm1,xmm7             ; xmm1=tmp0[row0 row1 row0 row1]
+
+        ; -- Even part
+
+        pslld     xmm6,(CONST_BITS+2)   ; xmm6=tmp10[row0 row1 **** ****]
+
+        ; -- Final output stage
+
+        movdqa    xmm4,xmm6
+        paddd     xmm6,xmm1     ; xmm6=data0[row0 row1 **** ****]=(C0 C1 ** **)
+        psubd     xmm4,xmm1     ; xmm4=data1[row0 row1 **** ****]=(D0 D1 ** **)
+
+        punpckldq xmm6,xmm4     ; xmm6=(C0 D0 C1 D1)
+
+        paddd     xmm6,[GOTOFF(ebx,PD_DESCALE_P2_2)]
+        psrad     xmm6,DESCALE_P2_2
+
+        packssdw  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1)
+        packsswb  xmm6,xmm6             ; xmm6=(C0 D0 C1 D1 C0 D0 C1 D1 ..)
+        paddb     xmm6,[GOTOFF(ebx,PB_CENTERJSAMP)]
+
+        pextrw  ebx,xmm6,0x00           ; ebx=(C0 D0 -- --)
+        pextrw  ecx,xmm6,0x01           ; ecx=(C1 D1 -- --)
+
+        mov     edx, JSAMPROW [edi+0*SIZEOF_JSAMPROW]
+        mov     esi, JSAMPROW [edi+1*SIZEOF_JSAMPROW]
+        mov     WORD [edx+eax*SIZEOF_JSAMPLE], bx
+        mov     WORD [esi+eax*SIZEOF_JSAMPLE], cx
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jpeg_nbits_table.inc glmark2-2021.02/src/libjpeg-turbo/simd/jpeg_nbits_table.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jpeg_nbits_table.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jpeg_nbits_table.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4097 @@
+jpeg_nbits_table db  \
+   0,  1,  2,  2,  3,  3,  3,  3,  4,  4,  4,  4,  4,  4,  4,  4,  \
+   5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  5,  \
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  \
+   6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  6,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  7,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  8,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+   9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  9,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16,  \
+  16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-3dn.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquant-3dn.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-3dn.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquant-3dn.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,233 @@
+;
+; jquant.asm - sample data conversion and quantization (3DNow! & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_3dnow (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                             FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_3dnow) PRIVATE
+
+EXTN(jsimd_convsamp_float_3dnow):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  mm7,mm7
+        psllw    mm7,7
+        packsswb mm7,mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   mm0,mm7                         ; mm0=(01234567)
+        psubb   mm1,mm7                         ; mm1=(89ABCDEF)
+
+        punpcklbw mm2,mm0                       ; mm2=(*0*1*2*3)
+        punpckhbw mm0,mm0                       ; mm0=(*4*5*6*7)
+        punpcklbw mm3,mm1                       ; mm3=(*8*9*A*B)
+        punpckhbw mm1,mm1                       ; mm1=(*C*D*E*F)
+
+        punpcklwd mm4,mm2                       ; mm4=(***0***1)
+        punpckhwd mm2,mm2                       ; mm2=(***2***3)
+        punpcklwd mm5,mm0                       ; mm5=(***4***5)
+        punpckhwd mm0,mm0                       ; mm0=(***6***7)
+
+        psrad   mm4,(DWORD_BIT-BYTE_BIT)        ; mm4=(01)
+        psrad   mm2,(DWORD_BIT-BYTE_BIT)        ; mm2=(23)
+        pi2fd   mm4,mm4
+        pi2fd   mm2,mm2
+        psrad   mm5,(DWORD_BIT-BYTE_BIT)        ; mm5=(45)
+        psrad   mm0,(DWORD_BIT-BYTE_BIT)        ; mm0=(67)
+        pi2fd   mm5,mm5
+        pi2fd   mm0,mm0
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm2
+        movq    MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
+        movq    MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
+
+        punpcklwd mm6,mm3                       ; mm6=(***8***9)
+        punpckhwd mm3,mm3                       ; mm3=(***A***B)
+        punpcklwd mm4,mm1                       ; mm4=(***C***D)
+        punpckhwd mm1,mm1                       ; mm1=(***E***F)
+
+        psrad   mm6,(DWORD_BIT-BYTE_BIT)        ; mm6=(89)
+        psrad   mm3,(DWORD_BIT-BYTE_BIT)        ; mm3=(AB)
+        pi2fd   mm6,mm6
+        pi2fd   mm3,mm3
+        psrad   mm4,(DWORD_BIT-BYTE_BIT)        ; mm4=(CD)
+        psrad   mm1,(DWORD_BIT-BYTE_BIT)        ; mm1=(EF)
+        pi2fd   mm4,mm4
+        pi2fd   mm1,mm1
+
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm6
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm3
+        movq    MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm4
+        movq    MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .convloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_3dnow (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                             FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_3dnow) PRIVATE
+
+EXTN(jsimd_quantize_float_3dnow):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov       eax, 0x4B400000       ; (float)0x00C00000 (rndint_magic)
+        movd      mm7,eax
+        punpckldq mm7,mm7               ; mm7={12582912.0F 12582912.0F}
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movq    mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm2, MMWORD [MMBLOCK(0,2,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(0,3,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm3, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)]
+
+        pfadd   mm0,mm7                 ; mm0=(00 ** 01 **)
+        pfadd   mm1,mm7                 ; mm1=(02 ** 03 **)
+        pfadd   mm2,mm7                 ; mm0=(04 ** 05 **)
+        pfadd   mm3,mm7                 ; mm1=(06 ** 07 **)
+
+        movq      mm4,mm0
+        punpcklwd mm0,mm1               ; mm0=(00 02 ** **)
+        punpckhwd mm4,mm1               ; mm4=(01 03 ** **)
+        movq      mm5,mm2
+        punpcklwd mm2,mm3               ; mm2=(04 06 ** **)
+        punpckhwd mm5,mm3               ; mm5=(05 07 ** **)
+
+        punpcklwd mm0,mm4               ; mm0=(00 01 02 03)
+        punpcklwd mm2,mm5               ; mm2=(04 05 06 07)
+
+        movq    mm6, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm1, MMWORD [MMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm6, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm1, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+        movq    mm3, MMWORD [MMBLOCK(1,2,esi,SIZEOF_FAST_FLOAT)]
+        movq    mm4, MMWORD [MMBLOCK(1,3,esi,SIZEOF_FAST_FLOAT)]
+        pfmul   mm3, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)]
+        pfmul   mm4, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)]
+
+        pfadd   mm6,mm7                 ; mm0=(10 ** 11 **)
+        pfadd   mm1,mm7                 ; mm4=(12 ** 13 **)
+        pfadd   mm3,mm7                 ; mm0=(14 ** 15 **)
+        pfadd   mm4,mm7                 ; mm4=(16 ** 17 **)
+
+        movq      mm5,mm6
+        punpcklwd mm6,mm1               ; mm6=(10 12 ** **)
+        punpckhwd mm5,mm1               ; mm5=(11 13 ** **)
+        movq      mm1,mm3
+        punpcklwd mm3,mm4               ; mm3=(14 16 ** **)
+        punpckhwd mm1,mm4               ; mm1=(15 17 ** **)
+
+        punpcklwd mm6,mm5               ; mm6=(10 11 12 13)
+        punpcklwd mm3,mm1               ; mm3=(14 15 16 17)
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm6
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     near .quantloop
+
+        femms           ; empty MMX/3DNow! state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquantf-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquantf-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquantf-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquantf-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,158 @@
+;
+; jquantf.asm - sample data conversion and quantization (64-bit SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                            FAST_FLOAT *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11 = JDIMENSION start_col
+; r12 = FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        pcmpeqw  xmm7,xmm7
+        psllw    xmm7,7
+        packsswb xmm7,xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov rsi, r10
+        mov     eax, r11d
+        mov rdi, r12
+        mov     rcx, DCTSIZE/2
+.convloop:
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]       ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]
+        movq    xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]
+
+        psubb   xmm0,xmm7                       ; xmm0=(01234567)
+        psubb   xmm1,xmm7                       ; xmm1=(89ABCDEF)
+
+        punpcklbw xmm0,xmm0                     ; xmm0=(*0*1*2*3*4*5*6*7)
+        punpcklbw xmm1,xmm1                     ; xmm1=(*8*9*A*B*C*D*E*F)
+
+        punpcklwd xmm2,xmm0                     ; xmm2=(***0***1***2***3)
+        punpckhwd xmm0,xmm0                     ; xmm0=(***4***5***6***7)
+        punpcklwd xmm3,xmm1                     ; xmm3=(***8***9***A***B)
+        punpckhwd xmm1,xmm1                     ; xmm1=(***C***D***E***F)
+
+        psrad     xmm2,(DWORD_BIT-BYTE_BIT)     ; xmm2=(0123)
+        psrad     xmm0,(DWORD_BIT-BYTE_BIT)     ; xmm0=(4567)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=(0123)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=(4567)
+        psrad     xmm3,(DWORD_BIT-BYTE_BIT)     ; xmm3=(89AB)
+        psrad     xmm1,(DWORD_BIT-BYTE_BIT)     ; xmm1=(CDEF)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=(89AB)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(0,1,rdi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(1,1,rdi,SIZEOF_FAST_FLOAT)], xmm1
+
+        add     rsi, byte 2*SIZEOF_JSAMPROW
+        add     rdi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     rcx
+        jnz     short .convloop
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                         FAST_FLOAT *workspace);
+;
+
+; r10 = JCOEFPTR coef_block
+; r11 = FAST_FLOAT *divisors
+; r12 = FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse2) PRIVATE
+
+EXTN(jsimd_quantize_float_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov rsi, r12
+        mov rdx, r11
+        mov rdi, r10
+        mov     rax, DCTSIZE2/16
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,rsi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,rdx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,rdx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,rsi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,rdx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,rdx,SIZEOF_FAST_FLOAT)]
+
+        cvtps2dq xmm0,xmm0
+        cvtps2dq xmm1,xmm1
+        cvtps2dq xmm2,xmm2
+        cvtps2dq xmm3,xmm3
+
+        packssdw xmm0,xmm1
+        packssdw xmm2,xmm3
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_JCOEF)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_JCOEF)], xmm2
+
+        add     rsi, byte 16*SIZEOF_FAST_FLOAT
+        add     rdx, byte 16*SIZEOF_FAST_FLOAT
+        add     rdi, byte 16*SIZEOF_JCOEF
+        dec     rax
+        jnz     short .quantloop
+
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquantf-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquantf-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquantf-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquantf-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,171 @@
+;
+; jquantf.asm - sample data conversion and quantization (SSE & SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                            FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  xmm7,xmm7
+        psllw    xmm7,7
+        packsswb xmm7,xmm7              ; xmm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   xmm0,xmm7                       ; xmm0=(01234567)
+        psubb   xmm1,xmm7                       ; xmm1=(89ABCDEF)
+
+        punpcklbw xmm0,xmm0                     ; xmm0=(*0*1*2*3*4*5*6*7)
+        punpcklbw xmm1,xmm1                     ; xmm1=(*8*9*A*B*C*D*E*F)
+
+        punpcklwd xmm2,xmm0                     ; xmm2=(***0***1***2***3)
+        punpckhwd xmm0,xmm0                     ; xmm0=(***4***5***6***7)
+        punpcklwd xmm3,xmm1                     ; xmm3=(***8***9***A***B)
+        punpckhwd xmm1,xmm1                     ; xmm1=(***C***D***E***F)
+
+        psrad     xmm2,(DWORD_BIT-BYTE_BIT)     ; xmm2=(0123)
+        psrad     xmm0,(DWORD_BIT-BYTE_BIT)     ; xmm0=(4567)
+        cvtdq2ps  xmm2,xmm2                     ; xmm2=(0123)
+        cvtdq2ps  xmm0,xmm0                     ; xmm0=(4567)
+        psrad     xmm3,(DWORD_BIT-BYTE_BIT)     ; xmm3=(89AB)
+        psrad     xmm1,(DWORD_BIT-BYTE_BIT)     ; xmm1=(CDEF)
+        cvtdq2ps  xmm3,xmm3                     ; xmm3=(89AB)
+        cvtdq2ps  xmm1,xmm1                     ; xmm1=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm3
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm1
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     short .convloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse2 (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                            FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse2) PRIVATE
+
+EXTN(jsimd_quantize_float_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        cvtps2dq xmm0,xmm0
+        cvtps2dq xmm1,xmm1
+        cvtps2dq xmm2,xmm2
+        cvtps2dq xmm3,xmm3
+
+        packssdw xmm0,xmm1
+        packssdw xmm2,xmm3
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_JCOEF)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_JCOEF)], xmm2
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     short .quantloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-altivec.c glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-altivec.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-altivec.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-altivec.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,252 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+/* INTEGER QUANTIZATION AND SAMPLE CONVERSION */
+
+#include "jsimd_altivec.h"
+
+
+/* NOTE: The address will either be aligned or offset by 8 bytes, so we can
+ * always get the data we want by using a single vector load (although we may
+ * have to permute the result.)
+ */
+#if __BIG_ENDIAN__
+
+#define LOAD_ROW(row) {  \
+  elemptr = sample_data[row] + start_col;  \
+  in##row = vec_ld(0, elemptr);  \
+  if ((size_t)elemptr & 15)  \
+    in##row = vec_perm(in##row, in##row, vec_lvsl(0, elemptr));  \
+}
+
+#else
+
+#define LOAD_ROW(row) {  \
+  elemptr = sample_data[row] + start_col;  \
+  in##row = vec_vsx_ld(0, elemptr);  \
+}
+
+#endif
+
+
+void
+jsimd_convsamp_altivec (JSAMPARRAY sample_data, JDIMENSION start_col,
+                        DCTELEM *workspace)
+{
+  JSAMPROW elemptr;
+
+  __vector unsigned char in0, in1, in2, in3, in4, in5, in6, in7;
+  __vector short out0, out1, out2, out3, out4, out5, out6, out7;
+
+  /* Constants */
+  __vector short pw_centerjsamp = { __8X(CENTERJSAMPLE) };
+  __vector unsigned char pb_zero = { __16X(0) };
+
+  LOAD_ROW(0);
+  LOAD_ROW(1);
+  LOAD_ROW(2);
+  LOAD_ROW(3);
+  LOAD_ROW(4);
+  LOAD_ROW(5);
+  LOAD_ROW(6);
+  LOAD_ROW(7);
+
+  out0 = (__vector short)VEC_UNPACKHU(in0);
+  out1 = (__vector short)VEC_UNPACKHU(in1);
+  out2 = (__vector short)VEC_UNPACKHU(in2);
+  out3 = (__vector short)VEC_UNPACKHU(in3);
+  out4 = (__vector short)VEC_UNPACKHU(in4);
+  out5 = (__vector short)VEC_UNPACKHU(in5);
+  out6 = (__vector short)VEC_UNPACKHU(in6);
+  out7 = (__vector short)VEC_UNPACKHU(in7);
+
+  out0 = vec_sub(out0, pw_centerjsamp);
+  out1 = vec_sub(out1, pw_centerjsamp);
+  out2 = vec_sub(out2, pw_centerjsamp);
+  out3 = vec_sub(out3, pw_centerjsamp);
+  out4 = vec_sub(out4, pw_centerjsamp);
+  out5 = vec_sub(out5, pw_centerjsamp);
+  out6 = vec_sub(out6, pw_centerjsamp);
+  out7 = vec_sub(out7, pw_centerjsamp);
+
+  vec_st(out0, 0, workspace);
+  vec_st(out1, 16, workspace);
+  vec_st(out2, 32, workspace);
+  vec_st(out3, 48, workspace);
+  vec_st(out4, 64, workspace);
+  vec_st(out5, 80, workspace);
+  vec_st(out6, 96, workspace);
+  vec_st(out7, 112, workspace);
+}
+
+
+#define WORD_BIT 16
+
+/* There is no AltiVec 16-bit unsigned multiply instruction, hence this.
+   We basically need an unsigned equivalent of vec_madds(). */
+
+#define MULTIPLY(vs0, vs1, out) {  \
+  tmpe = vec_mule((__vector unsigned short)vs0,  \
+                  (__vector unsigned short)vs1);  \
+  tmpo = vec_mulo((__vector unsigned short)vs0,  \
+                  (__vector unsigned short)vs1);  \
+  out = (__vector short)vec_perm((__vector unsigned short)tmpe,  \
+                                 (__vector unsigned short)tmpo,  \
+                                 shift_pack_index);  \
+}
+
+void
+jsimd_quantize_altivec (JCOEFPTR coef_block, DCTELEM *divisors,
+                        DCTELEM *workspace)
+{
+  __vector short row0, row1, row2, row3, row4, row5, row6, row7,
+    row0s, row1s, row2s, row3s, row4s, row5s, row6s, row7s,
+    corr0, corr1, corr2, corr3, corr4, corr5, corr6, corr7,
+    recip0, recip1, recip2, recip3, recip4, recip5, recip6, recip7,
+    scale0, scale1, scale2, scale3, scale4, scale5, scale6, scale7;
+  __vector unsigned int tmpe, tmpo;
+
+  /* Constants */
+  __vector unsigned short pw_word_bit_m1 = { __8X(WORD_BIT - 1) };
+#if __BIG_ENDIAN__
+  __vector unsigned char shift_pack_index =
+    {0,1,16,17,4,5,20,21,8,9,24,25,12,13,28,29};
+#else
+  __vector unsigned char shift_pack_index =
+    {2,3,18,19,6,7,22,23,10,11,26,27,14,15,30,31};
+#endif
+
+  row0 = vec_ld(0, workspace);
+  row1 = vec_ld(16, workspace);
+  row2 = vec_ld(32, workspace);
+  row3 = vec_ld(48, workspace);
+  row4 = vec_ld(64, workspace);
+  row5 = vec_ld(80, workspace);
+  row6 = vec_ld(96, workspace);
+  row7 = vec_ld(112, workspace);
+
+  /* Branch-less absolute value */
+  row0s = vec_sra(row0, pw_word_bit_m1);
+  row1s = vec_sra(row1, pw_word_bit_m1);
+  row2s = vec_sra(row2, pw_word_bit_m1);
+  row3s = vec_sra(row3, pw_word_bit_m1);
+  row4s = vec_sra(row4, pw_word_bit_m1);
+  row5s = vec_sra(row5, pw_word_bit_m1);
+  row6s = vec_sra(row6, pw_word_bit_m1);
+  row7s = vec_sra(row7, pw_word_bit_m1);
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  corr0 = vec_ld(DCTSIZE2 * 2, divisors);
+  corr1 = vec_ld(DCTSIZE2 * 2 + 16, divisors);
+  corr2 = vec_ld(DCTSIZE2 * 2 + 32, divisors);
+  corr3 = vec_ld(DCTSIZE2 * 2 + 48, divisors);
+  corr4 = vec_ld(DCTSIZE2 * 2 + 64, divisors);
+  corr5 = vec_ld(DCTSIZE2 * 2 + 80, divisors);
+  corr6 = vec_ld(DCTSIZE2 * 2 + 96, divisors);
+  corr7 = vec_ld(DCTSIZE2 * 2 + 112, divisors);
+
+  row0 = vec_add(row0, corr0);
+  row1 = vec_add(row1, corr1);
+  row2 = vec_add(row2, corr2);
+  row3 = vec_add(row3, corr3);
+  row4 = vec_add(row4, corr4);
+  row5 = vec_add(row5, corr5);
+  row6 = vec_add(row6, corr6);
+  row7 = vec_add(row7, corr7);
+
+  recip0 = vec_ld(0, divisors);
+  recip1 = vec_ld(16, divisors);
+  recip2 = vec_ld(32, divisors);
+  recip3 = vec_ld(48, divisors);
+  recip4 = vec_ld(64, divisors);
+  recip5 = vec_ld(80, divisors);
+  recip6 = vec_ld(96, divisors);
+  recip7 = vec_ld(112, divisors);
+
+  MULTIPLY(row0, recip0, row0);
+  MULTIPLY(row1, recip1, row1);
+  MULTIPLY(row2, recip2, row2);
+  MULTIPLY(row3, recip3, row3);
+  MULTIPLY(row4, recip4, row4);
+  MULTIPLY(row5, recip5, row5);
+  MULTIPLY(row6, recip6, row6);
+  MULTIPLY(row7, recip7, row7);
+
+  scale0 = vec_ld(DCTSIZE2 * 4, divisors);
+  scale1 = vec_ld(DCTSIZE2 * 4 + 16, divisors);
+  scale2 = vec_ld(DCTSIZE2 * 4 + 32, divisors);
+  scale3 = vec_ld(DCTSIZE2 * 4 + 48, divisors);
+  scale4 = vec_ld(DCTSIZE2 * 4 + 64, divisors);
+  scale5 = vec_ld(DCTSIZE2 * 4 + 80, divisors);
+  scale6 = vec_ld(DCTSIZE2 * 4 + 96, divisors);
+  scale7 = vec_ld(DCTSIZE2 * 4 + 112, divisors);
+
+  MULTIPLY(row0, scale0, row0);
+  MULTIPLY(row1, scale1, row1);
+  MULTIPLY(row2, scale2, row2);
+  MULTIPLY(row3, scale3, row3);
+  MULTIPLY(row4, scale4, row4);
+  MULTIPLY(row5, scale5, row5);
+  MULTIPLY(row6, scale6, row6);
+  MULTIPLY(row7, scale7, row7);
+
+  row0 = vec_xor(row0, row0s);
+  row1 = vec_xor(row1, row1s);
+  row2 = vec_xor(row2, row2s);
+  row3 = vec_xor(row3, row3s);
+  row4 = vec_xor(row4, row4s);
+  row5 = vec_xor(row5, row5s);
+  row6 = vec_xor(row6, row6s);
+  row7 = vec_xor(row7, row7s);
+  row0 = vec_sub(row0, row0s);
+  row1 = vec_sub(row1, row1s);
+  row2 = vec_sub(row2, row2s);
+  row3 = vec_sub(row3, row3s);
+  row4 = vec_sub(row4, row4s);
+  row5 = vec_sub(row5, row5s);
+  row6 = vec_sub(row6, row6s);
+  row7 = vec_sub(row7, row7s);
+
+  vec_st(row0, 0, coef_block);
+  vec_st(row1, 16, coef_block);
+  vec_st(row2, 32, coef_block);
+  vec_st(row3, 48, coef_block);
+  vec_st(row4, 64, coef_block);
+  vec_st(row5, 80, coef_block);
+  vec_st(row6, 96, coef_block);
+  vec_st(row7, 112, coef_block);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-sse2-64.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-sse2-64.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-sse2-64.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-sse2-64.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,187 @@
+;
+; jquanti.asm - sample data conversion and quantization (64-bit SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2009 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    64
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                      DCTELEM *workspace);
+;
+
+; r10 = JSAMPARRAY sample_data
+; r11 = JDIMENSION start_col
+; r12 = DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+        push    rbx
+
+        pxor    xmm6,xmm6               ; xmm6=(all 0's)
+        pcmpeqw xmm7,xmm7
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        mov rsi, r10
+        mov eax, r11d
+        mov rdi, r12
+        mov     rcx, DCTSIZE/4
+.convloop:
+        mov     rbx, JSAMPROW [rsi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov rdx, JSAMPROW [rsi+1*SIZEOF_JSAMPROW]       ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]       ; xmm0=(01234567)
+        movq    xmm1, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]       ; xmm1=(89ABCDEF)
+
+        mov     rbx, JSAMPROW [rsi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     rdx, JSAMPROW [rsi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm2, XMM_MMWORD [rbx+rax*SIZEOF_JSAMPLE]       ; xmm2=(GHIJKLMN)
+        movq    xmm3, XMM_MMWORD [rdx+rax*SIZEOF_JSAMPLE]       ; xmm3=(OPQRSTUV)
+
+        punpcklbw xmm0,xmm6             ; xmm0=(01234567)
+        punpcklbw xmm1,xmm6             ; xmm1=(89ABCDEF)
+        paddw     xmm0,xmm7
+        paddw     xmm1,xmm7
+        punpcklbw xmm2,xmm6             ; xmm2=(GHIJKLMN)
+        punpcklbw xmm3,xmm6             ; xmm3=(OPQRSTUV)
+        paddw     xmm2,xmm7
+        paddw     xmm3,xmm7
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3
+
+        add     rsi, byte 4*SIZEOF_JSAMPROW
+        add     rdi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     rcx
+        jnz     short .convloop
+
+        pop     rbx
+        uncollect_args
+        pop     rbp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM *divisors,
+;                      DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+
+; r10 = JCOEFPTR coef_block
+; r11 = DCTELEM *divisors
+; r12 = DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_sse2) PRIVATE
+
+EXTN(jsimd_quantize_sse2):
+        push    rbp
+        mov     rax,rsp
+        mov     rbp,rsp
+        collect_args
+
+        mov rsi, r12
+        mov rdx, r11
+        mov rdi, r10
+        mov     rax, DCTSIZE2/32
+.quantloop:
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(1,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(2,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(3,0,rsi,SIZEOF_DCTELEM)]
+        movdqa  xmm0,xmm4
+        movdqa  xmm1,xmm5
+        movdqa  xmm2,xmm6
+        movdqa  xmm3,xmm7
+        psraw   xmm4,(WORD_BIT-1)
+        psraw   xmm5,(WORD_BIT-1)
+        psraw   xmm6,(WORD_BIT-1)
+        psraw   xmm7,(WORD_BIT-1)
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4               ; if (xmm0 < 0) xmm0 = -xmm0;
+        psubw   xmm1,xmm5               ; if (xmm1 < 0) xmm1 = -xmm1;
+        psubw   xmm2,xmm6               ; if (xmm2 < 0) xmm2 = -xmm2;
+        psubw   xmm3,xmm7               ; if (xmm3 < 0) xmm3 = -xmm3;
+
+        paddw   xmm0, XMMWORD [CORRECTION(0,0,rdx)]  ; correction + roundfactor
+        paddw   xmm1, XMMWORD [CORRECTION(1,0,rdx)]
+        paddw   xmm2, XMMWORD [CORRECTION(2,0,rdx)]
+        paddw   xmm3, XMMWORD [CORRECTION(3,0,rdx)]
+        pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,rdx)]  ; reciprocal
+        pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,rdx)]
+        pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,rdx)]
+        pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,rdx)]
+        pmulhuw xmm0, XMMWORD [SCALE(0,0,rdx)]  ; scale
+        pmulhuw xmm1, XMMWORD [SCALE(1,0,rdx)]
+        pmulhuw xmm2, XMMWORD [SCALE(2,0,rdx)]
+        pmulhuw xmm3, XMMWORD [SCALE(3,0,rdx)]
+
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4
+        psubw   xmm1,xmm5
+        psubw   xmm2,xmm6
+        psubw   xmm3,xmm7
+        movdqa  XMMWORD [XMMBLOCK(0,0,rdi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,rdi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,rdi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,rdi,SIZEOF_DCTELEM)], xmm3
+
+        add     rsi, byte 32*SIZEOF_DCTELEM
+        add     rdx, byte 32*SIZEOF_DCTELEM
+        add     rdi, byte 32*SIZEOF_JCOEF
+        dec     rax
+        jnz     near .quantloop
+
+        uncollect_args
+        pop     rbp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-sse2.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-sse2.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquanti-sse2.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquanti-sse2.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,200 @@
+;
+; jquanti.asm - sample data conversion and quantization (SSE2)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_sse2 (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                      DCTELEM *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_sse2) PRIVATE
+
+EXTN(jsimd_convsamp_sse2):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pxor    xmm6,xmm6               ; xmm6=(all 0's)
+        pcmpeqw xmm7,xmm7
+        psllw   xmm7,7                  ; xmm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm0, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]       ; xmm0=(01234567)
+        movq    xmm1, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]       ; xmm1=(89ABCDEF)
+
+        mov     ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    xmm2, XMM_MMWORD [ebx+eax*SIZEOF_JSAMPLE]       ; xmm2=(GHIJKLMN)
+        movq    xmm3, XMM_MMWORD [edx+eax*SIZEOF_JSAMPLE]       ; xmm3=(OPQRSTUV)
+
+        punpcklbw xmm0,xmm6             ; xmm0=(01234567)
+        punpcklbw xmm1,xmm6             ; xmm1=(89ABCDEF)
+        paddw     xmm0,xmm7
+        paddw     xmm1,xmm7
+        punpcklbw xmm2,xmm6             ; xmm2=(GHIJKLMN)
+        punpcklbw xmm3,xmm6             ; xmm3=(OPQRSTUV)
+        paddw     xmm2,xmm7
+        paddw     xmm3,xmm7
+
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3
+
+        add     esi, byte 4*SIZEOF_JSAMPROW
+        add     edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     short .convloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_sse2 (JCOEFPTR coef_block, DCTELEM *divisors,
+;                      DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) XMMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) XMMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      XMMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; DCTELEM *divisors
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_sse2) PRIVATE
+
+EXTN(jsimd_quantize_sse2):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/32
+        alignx  16,7
+.quantloop:
+        movdqa  xmm4, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm5, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm6, XMMWORD [XMMBLOCK(2,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm7, XMMWORD [XMMBLOCK(3,0,esi,SIZEOF_DCTELEM)]
+        movdqa  xmm0,xmm4
+        movdqa  xmm1,xmm5
+        movdqa  xmm2,xmm6
+        movdqa  xmm3,xmm7
+        psraw   xmm4,(WORD_BIT-1)
+        psraw   xmm5,(WORD_BIT-1)
+        psraw   xmm6,(WORD_BIT-1)
+        psraw   xmm7,(WORD_BIT-1)
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4               ; if (xmm0 < 0) xmm0 = -xmm0;
+        psubw   xmm1,xmm5               ; if (xmm1 < 0) xmm1 = -xmm1;
+        psubw   xmm2,xmm6               ; if (xmm2 < 0) xmm2 = -xmm2;
+        psubw   xmm3,xmm7               ; if (xmm3 < 0) xmm3 = -xmm3;
+
+        paddw   xmm0, XMMWORD [CORRECTION(0,0,edx)]  ; correction + roundfactor
+        paddw   xmm1, XMMWORD [CORRECTION(1,0,edx)]
+        paddw   xmm2, XMMWORD [CORRECTION(2,0,edx)]
+        paddw   xmm3, XMMWORD [CORRECTION(3,0,edx)]
+        pmulhuw xmm0, XMMWORD [RECIPROCAL(0,0,edx)]  ; reciprocal
+        pmulhuw xmm1, XMMWORD [RECIPROCAL(1,0,edx)]
+        pmulhuw xmm2, XMMWORD [RECIPROCAL(2,0,edx)]
+        pmulhuw xmm3, XMMWORD [RECIPROCAL(3,0,edx)]
+        pmulhuw xmm0, XMMWORD [SCALE(0,0,edx)]  ; scale
+        pmulhuw xmm1, XMMWORD [SCALE(1,0,edx)]
+        pmulhuw xmm2, XMMWORD [SCALE(2,0,edx)]
+        pmulhuw xmm3, XMMWORD [SCALE(3,0,edx)]
+
+        pxor    xmm0,xmm4
+        pxor    xmm1,xmm5
+        pxor    xmm2,xmm6
+        pxor    xmm3,xmm7
+        psubw   xmm0,xmm4
+        psubw   xmm1,xmm5
+        psubw   xmm2,xmm6
+        psubw   xmm3,xmm7
+        movdqa  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_DCTELEM)], xmm0
+        movdqa  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_DCTELEM)], xmm1
+        movdqa  XMMWORD [XMMBLOCK(2,0,edi,SIZEOF_DCTELEM)], xmm2
+        movdqa  XMMWORD [XMMBLOCK(3,0,edi,SIZEOF_DCTELEM)], xmm3
+
+        add     esi, byte 32*SIZEOF_DCTELEM
+        add     edx, byte 32*SIZEOF_DCTELEM
+        add     edi, byte 32*SIZEOF_JCOEF
+        dec     eax
+        jnz     near .quantloop
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-mmx.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquant-mmx.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-mmx.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquant-mmx.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,274 @@
+;
+; jquant.asm - sample data conversion and quantization (MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_mmx (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                     DCTELEM *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_mmx) PRIVATE
+
+EXTN(jsimd_convsamp_mmx):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pxor    mm6,mm6                 ; mm6=(all 0's)
+        pcmpeqw mm7,mm7
+        psllw   mm7,7                   ; mm7={0xFF80 0xFF80 0xFF80 0xFF80}
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/4
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]    ; mm0=(01234567)
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]    ; mm1=(89ABCDEF)
+
+        mov     ebx, JSAMPROW [esi+2*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+3*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm2, MMWORD [ebx+eax*SIZEOF_JSAMPLE]    ; mm2=(GHIJKLMN)
+        movq    mm3, MMWORD [edx+eax*SIZEOF_JSAMPLE]    ; mm3=(OPQRSTUV)
+
+        movq      mm4,mm0
+        punpcklbw mm0,mm6               ; mm0=(0123)
+        punpckhbw mm4,mm6               ; mm4=(4567)
+        movq      mm5,mm1
+        punpcklbw mm1,mm6               ; mm1=(89AB)
+        punpckhbw mm5,mm6               ; mm5=(CDEF)
+
+        paddw   mm0,mm7
+        paddw   mm4,mm7
+        paddw   mm1,mm7
+        paddw   mm5,mm7
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm4
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_DCTELEM)], mm1
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_DCTELEM)], mm5
+
+        movq      mm0,mm2
+        punpcklbw mm2,mm6               ; mm2=(GHIJ)
+        punpckhbw mm0,mm6               ; mm0=(KLMN)
+        movq      mm4,mm3
+        punpcklbw mm3,mm6               ; mm3=(OPQR)
+        punpckhbw mm4,mm6               ; mm4=(STUV)
+
+        paddw   mm2,mm7
+        paddw   mm0,mm7
+        paddw   mm3,mm7
+        paddw   mm4,mm7
+
+        movq    MMWORD [MMBLOCK(2,0,edi,SIZEOF_DCTELEM)], mm2
+        movq    MMWORD [MMBLOCK(2,1,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(3,0,edi,SIZEOF_DCTELEM)], mm3
+        movq    MMWORD [MMBLOCK(3,1,edi,SIZEOF_DCTELEM)], mm4
+
+        add     esi, byte 4*SIZEOF_JSAMPROW
+        add     edi, byte 4*DCTSIZE*SIZEOF_DCTELEM
+        dec     ecx
+        jnz     short .convloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; This implementation is based on an algorithm described in
+;   "How to optimize for the Pentium family of microprocessors"
+;   (http://www.agner.org/assem/).
+;
+; GLOBAL(void)
+; jsimd_quantize_mmx (JCOEFPTR coef_block, DCTELEM *divisors,
+;                     DCTELEM *workspace);
+;
+
+%define RECIPROCAL(m,n,b) MMBLOCK(DCTSIZE*0+(m),(n),(b),SIZEOF_DCTELEM)
+%define CORRECTION(m,n,b) MMBLOCK(DCTSIZE*1+(m),(n),(b),SIZEOF_DCTELEM)
+%define SCALE(m,n,b)      MMBLOCK(DCTSIZE*2+(m),(n),(b),SIZEOF_DCTELEM)
+%define SHIFT(m,n,b)      MMBLOCK(DCTSIZE*3+(m),(n),(b),SIZEOF_DCTELEM)
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; DCTELEM *divisors
+%define workspace       ebp+16          ; DCTELEM *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_mmx) PRIVATE
+
+EXTN(jsimd_quantize_mmx):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     ah, 2
+        alignx  16,7
+.quantloop1:
+        mov     al, DCTSIZE2/8/2
+        alignx  16,7
+.quantloop2:
+        movq    mm2, MMWORD [MMBLOCK(0,0,esi,SIZEOF_DCTELEM)]
+        movq    mm3, MMWORD [MMBLOCK(0,1,esi,SIZEOF_DCTELEM)]
+
+        movq    mm0,mm2
+        movq    mm1,mm3
+
+        psraw   mm2,(WORD_BIT-1)  ; -1 if value < 0, 0 otherwise
+        psraw   mm3,(WORD_BIT-1)
+
+        pxor    mm0,mm2   ; val = -val
+        pxor    mm1,mm3
+        psubw   mm0,mm2
+        psubw   mm1,mm3
+
+        ;
+        ; MMX is an annoyingly crappy instruction set. It has two
+        ; misfeatures that are causing problems here:
+        ;
+        ; - All multiplications are signed.
+        ;
+        ; - The second operand for the shifts is not treated as packed.
+        ;
+        ;
+        ; We work around the first problem by implementing this algorithm:
+        ;
+        ; unsigned long unsigned_multiply(unsigned short x, unsigned short y)
+        ; {
+        ;   enum { SHORT_BIT = 16 };
+        ;   signed short sx = (signed short) x;
+        ;   signed short sy = (signed short) y;
+        ;   signed long sz;
+        ;
+        ;   sz = (long) sx * (long) sy;     /* signed multiply */
+        ;
+        ;   if (sx < 0) sz += (long) sy << SHORT_BIT;
+        ;   if (sy < 0) sz += (long) sx << SHORT_BIT;
+        ;
+        ;   return (unsigned long) sz;
+        ; }
+        ;
+        ; (note that a negative sx adds _sy_ and vice versa)
+        ;
+        ; For the second problem, we replace the shift by a multiplication.
+        ; Unfortunately that means we have to deal with the signed issue again.
+        ;
+
+        paddw   mm0, MMWORD [CORRECTION(0,0,edx)]   ; correction + roundfactor
+        paddw   mm1, MMWORD [CORRECTION(0,1,edx)]
+
+        movq    mm4,mm0   ; store current value for later
+        movq    mm5,mm1
+        pmulhw  mm0, MMWORD [RECIPROCAL(0,0,edx)]   ; reciprocal
+        pmulhw  mm1, MMWORD [RECIPROCAL(0,1,edx)]
+        paddw   mm0,mm4         ; reciprocal is always negative (MSB=1),
+        paddw   mm1,mm5   ; so we always need to add the initial value
+                        ; (input value is never negative as we
+                        ; inverted it at the start of this routine)
+
+        ; here it gets a bit tricky as both scale
+        ; and mm0/mm1 can be negative
+        movq    mm6, MMWORD [SCALE(0,0,edx)]    ; scale
+        movq    mm7, MMWORD [SCALE(0,1,edx)]
+        movq    mm4,mm0
+        movq    mm5,mm1
+        pmulhw  mm0,mm6
+        pmulhw  mm1,mm7
+
+        psraw   mm6,(WORD_BIT-1)    ; determine if scale is negative
+        psraw   mm7,(WORD_BIT-1)
+
+        pand    mm6,mm4             ; and add input if it is
+        pand    mm7,mm5
+        paddw   mm0,mm6
+        paddw   mm1,mm7
+
+        psraw   mm4,(WORD_BIT-1)    ; then check if negative input
+        psraw   mm5,(WORD_BIT-1)
+
+        pand    mm4, MMWORD [SCALE(0,0,edx)]    ; and add scale if it is
+        pand    mm5, MMWORD [SCALE(0,1,edx)]
+        paddw   mm0,mm4
+        paddw   mm1,mm5
+
+        pxor    mm0,mm2   ; val = -val
+        pxor    mm1,mm3
+        psubw   mm0,mm2
+        psubw   mm1,mm3
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_DCTELEM)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_DCTELEM)], mm1
+
+        add     esi, byte 8*SIZEOF_DCTELEM
+        add     edx, byte 8*SIZEOF_DCTELEM
+        add     edi, byte 8*SIZEOF_JCOEF
+        dec     al
+        jnz     near .quantloop2
+        dec     ah
+        jnz     near .quantloop1        ; to avoid branch misprediction
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-sse.asm glmark2-2021.02/src/libjpeg-turbo/simd/jquant-sse.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jquant-sse.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jquant-sse.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,211 @@
+;
+; jquant.asm - sample data conversion and quantization (SSE & MMX)
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+%include "jdct.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Load data into workspace, applying unsigned->signed conversion
+;
+; GLOBAL(void)
+; jsimd_convsamp_float_sse (JSAMPARRAY sample_data, JDIMENSION start_col,
+;                           FAST_FLOAT *workspace);
+;
+
+%define sample_data     ebp+8           ; JSAMPARRAY sample_data
+%define start_col       ebp+12          ; JDIMENSION start_col
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_convsamp_float_sse) PRIVATE
+
+EXTN(jsimd_convsamp_float_sse):
+        push    ebp
+        mov     ebp,esp
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        pcmpeqw  mm7,mm7
+        psllw    mm7,7
+        packsswb mm7,mm7                ; mm7 = PB_CENTERJSAMPLE (0x808080..)
+
+        mov     esi, JSAMPARRAY [sample_data]   ; (JSAMPROW *)
+        mov     eax, JDIMENSION [start_col]
+        mov     edi, POINTER [workspace]        ; (DCTELEM *)
+        mov     ecx, DCTSIZE/2
+        alignx  16,7
+.convloop:
+        mov     ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+        mov     edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW]   ; (JSAMPLE *)
+
+        movq    mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]
+        movq    mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]
+
+        psubb   mm0,mm7                         ; mm0=(01234567)
+        psubb   mm1,mm7                         ; mm1=(89ABCDEF)
+
+        punpcklbw mm2,mm0                       ; mm2=(*0*1*2*3)
+        punpckhbw mm0,mm0                       ; mm0=(*4*5*6*7)
+        punpcklbw mm3,mm1                       ; mm3=(*8*9*A*B)
+        punpckhbw mm1,mm1                       ; mm1=(*C*D*E*F)
+
+        punpcklwd mm4,mm2                       ; mm4=(***0***1)
+        punpckhwd mm2,mm2                       ; mm2=(***2***3)
+        punpcklwd mm5,mm0                       ; mm5=(***4***5)
+        punpckhwd mm0,mm0                       ; mm0=(***6***7)
+
+        psrad     mm4,(DWORD_BIT-BYTE_BIT)      ; mm4=(01)
+        psrad     mm2,(DWORD_BIT-BYTE_BIT)      ; mm2=(23)
+        cvtpi2ps  xmm0,mm4                      ; xmm0=(01**)
+        cvtpi2ps  xmm1,mm2                      ; xmm1=(23**)
+        psrad     mm5,(DWORD_BIT-BYTE_BIT)      ; mm5=(45)
+        psrad     mm0,(DWORD_BIT-BYTE_BIT)      ; mm0=(67)
+        cvtpi2ps  xmm2,mm5                      ; xmm2=(45**)
+        cvtpi2ps  xmm3,mm0                      ; xmm3=(67**)
+
+        punpcklwd mm6,mm3                       ; mm6=(***8***9)
+        punpckhwd mm3,mm3                       ; mm3=(***A***B)
+        punpcklwd mm4,mm1                       ; mm4=(***C***D)
+        punpckhwd mm1,mm1                       ; mm1=(***E***F)
+
+        psrad     mm6,(DWORD_BIT-BYTE_BIT)      ; mm6=(89)
+        psrad     mm3,(DWORD_BIT-BYTE_BIT)      ; mm3=(AB)
+        cvtpi2ps  xmm4,mm6                      ; xmm4=(89**)
+        cvtpi2ps  xmm5,mm3                      ; xmm5=(AB**)
+        psrad     mm4,(DWORD_BIT-BYTE_BIT)      ; mm4=(CD)
+        psrad     mm1,(DWORD_BIT-BYTE_BIT)      ; mm1=(EF)
+        cvtpi2ps  xmm6,mm4                      ; xmm6=(CD**)
+        cvtpi2ps  xmm7,mm1                      ; xmm7=(EF**)
+
+        movlhps   xmm0,xmm1                     ; xmm0=(0123)
+        movlhps   xmm2,xmm3                     ; xmm2=(4567)
+        movlhps   xmm4,xmm5                     ; xmm4=(89AB)
+        movlhps   xmm6,xmm7                     ; xmm6=(CDEF)
+
+        movaps  XMMWORD [XMMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], xmm0
+        movaps  XMMWORD [XMMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], xmm2
+        movaps  XMMWORD [XMMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], xmm4
+        movaps  XMMWORD [XMMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], xmm6
+
+        add     esi, byte 2*SIZEOF_JSAMPROW
+        add     edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
+        dec     ecx
+        jnz     near .convloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        pop     ebp
+        ret
+
+
+; --------------------------------------------------------------------------
+;
+; Quantize/descale the coefficients, and store into coef_block
+;
+; GLOBAL(void)
+; jsimd_quantize_float_sse (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+;                           FAST_FLOAT *workspace);
+;
+
+%define coef_block      ebp+8           ; JCOEFPTR coef_block
+%define divisors        ebp+12          ; FAST_FLOAT *divisors
+%define workspace       ebp+16          ; FAST_FLOAT *workspace
+
+        align   16
+        global  EXTN(jsimd_quantize_float_sse) PRIVATE
+
+EXTN(jsimd_quantize_float_sse):
+        push    ebp
+        mov     ebp,esp
+;       push    ebx             ; unused
+;       push    ecx             ; unused
+;       push    edx             ; need not be preserved
+        push    esi
+        push    edi
+
+        mov     esi, POINTER [workspace]
+        mov     edx, POINTER [divisors]
+        mov     edi, JCOEFPTR [coef_block]
+        mov     eax, DCTSIZE2/16
+        alignx  16,7
+.quantloop:
+        movaps  xmm0, XMMWORD [XMMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm1, XMMWORD [XMMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm0, XMMWORD [XMMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm1, XMMWORD [XMMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
+        movaps  xmm2, XMMWORD [XMMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
+        movaps  xmm3, XMMWORD [XMMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
+        mulps   xmm2, XMMWORD [XMMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
+        mulps   xmm3, XMMWORD [XMMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
+
+        movhlps  xmm4,xmm0
+        movhlps  xmm5,xmm1
+
+        cvtps2pi mm0,xmm0
+        cvtps2pi mm1,xmm1
+        cvtps2pi mm4,xmm4
+        cvtps2pi mm5,xmm5
+
+        movhlps  xmm6,xmm2
+        movhlps  xmm7,xmm3
+
+        cvtps2pi mm2,xmm2
+        cvtps2pi mm3,xmm3
+        cvtps2pi mm6,xmm6
+        cvtps2pi mm7,xmm7
+
+        packssdw mm0,mm4
+        packssdw mm1,mm5
+        packssdw mm2,mm6
+        packssdw mm3,mm7
+
+        movq    MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
+        movq    MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm1
+        movq    MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm2
+        movq    MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3
+
+        add     esi, byte 16*SIZEOF_FAST_FLOAT
+        add     edx, byte 16*SIZEOF_FAST_FLOAT
+        add     edi, byte 16*SIZEOF_JCOEF
+        dec     eax
+        jnz     short .quantloop
+
+        emms            ; empty MMX state
+
+        pop     edi
+        pop     esi
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; unused
+;       pop     ebx             ; unused
+        pop     ebp
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_altivec.h glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_altivec.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_altivec.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_altivec.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,99 @@
+/*
+ * AltiVec optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * All rights reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+#include <altivec.h>
+
+
+/* Common code */
+
+#define __4X(a) a, a, a, a
+#define __4X2(a, b) a, b, a, b, a, b, a, b
+#define __8X(a) __4X(a), __4X(a)
+#define __16X(a) __8X(a), __8X(a)
+
+#define TRANSPOSE(row, col)  \
+{  \
+  __vector short row04l, row04h, row15l, row15h,  \
+                 row26l, row26h, row37l, row37h;  \
+  __vector short col01e, col01o, col23e, col23o,  \
+                 col45e, col45o, col67e, col67o;  \
+  \
+                                       /* transpose coefficients (phase 1) */ \
+  row04l = vec_mergeh(row##0, row##4); /* row04l=(00 40 01 41 02 42 03 43) */ \
+  row04h = vec_mergel(row##0, row##4); /* row04h=(04 44 05 45 06 46 07 47) */ \
+  row15l = vec_mergeh(row##1, row##5); /* row15l=(10 50 11 51 12 52 13 53) */ \
+  row15h = vec_mergel(row##1, row##5); /* row15h=(14 54 15 55 16 56 17 57) */ \
+  row26l = vec_mergeh(row##2, row##6); /* row26l=(20 60 21 61 22 62 23 63) */ \
+  row26h = vec_mergel(row##2, row##6); /* row26h=(24 64 25 65 26 66 27 67) */ \
+  row37l = vec_mergeh(row##3, row##7); /* row37l=(30 70 31 71 32 72 33 73) */ \
+  row37h = vec_mergel(row##3, row##7); /* row37h=(34 74 35 75 36 76 37 77) */ \
+  \
+                                       /* transpose coefficients (phase 2) */ \
+  col01e = vec_mergeh(row04l, row26l); /* col01e=(00 20 40 60 01 21 41 61) */ \
+  col23e = vec_mergel(row04l, row26l); /* col23e=(02 22 42 62 03 23 43 63) */ \
+  col45e = vec_mergeh(row04h, row26h); /* col45e=(04 24 44 64 05 25 45 65) */ \
+  col67e = vec_mergel(row04h, row26h); /* col67e=(06 26 46 66 07 27 47 67) */ \
+  col01o = vec_mergeh(row15l, row37l); /* col01o=(10 30 50 70 11 31 51 71) */ \
+  col23o = vec_mergel(row15l, row37l); /* col23o=(12 32 52 72 13 33 53 73) */ \
+  col45o = vec_mergeh(row15h, row37h); /* col45o=(14 34 54 74 15 35 55 75) */ \
+  col67o = vec_mergel(row15h, row37h); /* col67o=(16 36 56 76 17 37 57 77) */ \
+  \
+                                       /* transpose coefficients (phase 3) */ \
+  col##0 = vec_mergeh(col01e, col01o); /* col0=(00 10 20 30 40 50 60 70) */   \
+  col##1 = vec_mergel(col01e, col01o); /* col1=(01 11 21 31 41 51 61 71) */   \
+  col##2 = vec_mergeh(col23e, col23o); /* col2=(02 12 22 32 42 52 62 72) */   \
+  col##3 = vec_mergel(col23e, col23o); /* col3=(03 13 23 33 43 53 63 73) */   \
+  col##4 = vec_mergeh(col45e, col45o); /* col4=(04 14 24 34 44 54 64 74) */   \
+  col##5 = vec_mergel(col45e, col45o); /* col5=(05 15 25 35 45 55 65 75) */   \
+  col##6 = vec_mergeh(col67e, col67o); /* col6=(06 16 26 36 46 56 66 76) */   \
+  col##7 = vec_mergel(col67e, col67o); /* col7=(07 17 27 37 47 57 67 77) */   \
+}
+
+#ifndef min
+#define min(a,b) ((a) < (b) ? (a) : (b))
+#endif
+
+
+/* Macros to abstract big/little endian bit twiddling */
+
+#if __BIG_ENDIAN__
+
+#define VEC_LD(a, b) vec_ld(a, b)
+#define VEC_ST(a, b, c) vec_st(a, b, c)
+#define VEC_UNPACKHU(a) vec_mergeh(pb_zero, a)
+#define VEC_UNPACKLU(a) vec_mergel(pb_zero, a)
+
+#else
+
+#define VEC_LD(a, b) vec_vsx_ld(a, b)
+#define VEC_ST(a, b, c) vec_vsx_st(a, b, c)
+#define VEC_UNPACKHU(a) vec_mergeh(a, pb_zero)
+#define VEC_UNPACKLU(a) vec_mergel(a, pb_zero)
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm64.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm64.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm64.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm64.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,802 @@
+/*
+ * jsimd_arm64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015-2016 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit ARM architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+#define JSIMD_FASTLD3 1
+#define JSIMD_FASTST3 2
+#define JSIMD_FASTTBL 4
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+static unsigned int simd_features = JSIMD_FASTLD3 | JSIMD_FASTST3 |
+                                    JSIMD_FASTTBL;
+
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+
+#define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
+
+LOCAL(int)
+check_cpuinfo (char *buffer, const char *field, char *value)
+{
+  char *p;
+  if (*value == 0)
+    return 0;
+  if (strncmp(buffer, field, strlen(field)) != 0)
+    return 0;
+  buffer += strlen(field);
+  while (isspace(*buffer))
+    buffer++;
+
+  /* Check if 'value' is present in the buffer as a separate word */
+  while ((p = strstr(buffer, value))) {
+    if (p > buffer && !isspace(*(p - 1))) {
+      buffer++;
+      continue;
+    }
+    p += strlen(value);
+    if (*p != 0 && !isspace(*p)) {
+      buffer++;
+      continue;
+    }
+    return 1;
+  }
+  return 0;
+}
+
+LOCAL(int)
+parse_proc_cpuinfo (int bufsize)
+{
+  char *buffer = (char *)malloc(bufsize);
+  FILE *fd;
+
+  if (!buffer)
+    return 0;
+
+  fd = fopen("/proc/cpuinfo", "r");
+  if (fd) {
+    while (fgets(buffer, bufsize, fd)) {
+      if (!strchr(buffer, '\n') && !feof(fd)) {
+        /* "impossible" happened - insufficient size of the buffer! */
+        fclose(fd);
+        free(buffer);
+        return 0;
+      }
+      if (check_cpuinfo(buffer, "CPU part", "0xd03") ||
+          check_cpuinfo(buffer, "CPU part", "0xd07"))
+        /* The Cortex-A53 has a slow tbl implementation.  We can gain a few
+           percent speedup by disabling the use of that instruction.  The
+           speedup on Cortex-A57 is more subtle but still measurable. */
+        simd_features &= ~JSIMD_FASTTBL;
+      else if (check_cpuinfo(buffer, "CPU part", "0x0a1"))
+        /* The SIMD version of Huffman encoding is slower than the C version on
+           Cavium ThunderX.  Also, ld3 and st3 are abyssmally slow on that
+           CPU. */
+        simd_huffman = simd_features = 0;
+    }
+    fclose(fd);
+  }
+  free(buffer);
+  return 1;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+
+/*
+ * ARMv8 architectures support NEON extensions by default.
+ * It is no longer optional as it was with ARMv7.
+ */
+
+
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  int bufsize = 1024; /* an initial guess for the line buffer size limit */
+#endif
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+  simd_support |= JSIMD_ARM_NEON;
+#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+  while (!parse_proc_cpuinfo(bufsize)) {
+    bufsize *= 2;
+    if (bufsize > SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT)
+      break;
+  }
+#endif
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENEON");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_ARM_NEON;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+  env = getenv("JSIMD_FASTLD3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTLD3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTLD3;
+  env = getenv("JSIMD_FASTST3");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_features |= JSIMD_FASTST3;
+  if ((env != NULL) && (strcmp(env, "0") == 0))
+    simd_features &= ~JSIMD_FASTST3;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extrgb_ycc_convert_neon;
+      else
+        neonfct=jsimd_extrgb_ycc_convert_neon_slowld3;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct=jsimd_extrgbx_ycc_convert_neon;
+      break;
+    case JCS_EXT_BGR:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extbgr_ycc_convert_neon;
+      else
+        neonfct=jsimd_extbgr_ycc_convert_neon_slowld3;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct=jsimd_extbgrx_ycc_convert_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct=jsimd_extxbgr_ycc_convert_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct=jsimd_extxrgb_ycc_convert_neon;
+      break;
+    default:
+      if (simd_features & JSIMD_FASTLD3)
+        neonfct=jsimd_extrgb_ycc_convert_neon;
+      else
+        neonfct=jsimd_extrgb_ycc_convert_neon_slowld3;
+      break;
+  }
+
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extrgb_convert_neon;
+      else
+        neonfct=jsimd_ycc_extrgb_convert_neon_slowst3;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      neonfct=jsimd_ycc_extrgbx_convert_neon;
+      break;
+    case JCS_EXT_BGR:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extbgr_convert_neon;
+      else
+        neonfct=jsimd_ycc_extbgr_convert_neon_slowst3;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      neonfct=jsimd_ycc_extbgrx_convert_neon;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      neonfct=jsimd_ycc_extxbgr_convert_neon;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      neonfct=jsimd_ycc_extxrgb_convert_neon;
+      break;
+    default:
+      if (simd_features & JSIMD_FASTST3)
+        neonfct=jsimd_ycc_extrgb_convert_neon;
+      else
+        neonfct=jsimd_ycc_extrgb_convert_neon_slowst3;
+      break;
+  }
+
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_neon(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_neon(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_neon(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  if (simd_features & JSIMD_FASTTBL)
+    return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                            dctbl, actbl);
+  else
+    return jsimd_huff_encode_one_block_neon_slowtbl(state, buffer, block,
+                                                    last_dc_val, dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm64_neon.S glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm64_neon.S
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm64_neon.S	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm64_neon.S	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,3453 @@
+/*
+ * ARMv8 NEON optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies).
+ * All rights reserved.
+ * Author: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2013-2014, Linaro Limited
+ * Author: Ragesh Radhakrishnan <ragesh.r@linaro.org>
+ * Copyright (C) 2014-2016, D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016, Matthieu Darbois.  All Rights Reserved.
+ * Copyright (C) 2016, Siarhei Siamashka.  All Rights Reserved.
+ *
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
+#endif
+
+.text
+
+
+#define RESPECT_STRICT_ALIGNMENT 1
+
+
+/*****************************************************************************/
+
+/* Supplementary macro for setting function attributes */
+.macro asm_function fname
+#ifdef __APPLE__
+    .globl _\fname
+_\fname:
+#else
+    .global \fname
+#ifdef __ELF__
+    .hidden \fname
+    .type \fname, %function
+#endif
+\fname:
+#endif
+.endm
+
+/* Transpose elements of single 128 bit registers */
+.macro transpose_single x0, x1, xi, xilen, literal
+    ins             \xi\xilen[0], \x0\xilen[0]
+    ins             \x1\xilen[0], \x0\xilen[1]
+    trn1            \x0\literal, \x0\literal, \x1\literal
+    trn2            \x1\literal, \xi\literal, \x1\literal
+.endm
+
+/* Transpose elements of 2 differnet registers */
+.macro transpose x0, x1, xi, xilen, literal
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\literal, \x0\literal, \x1\literal
+    trn2            \x1\literal, \xi\literal, \x1\literal
+.endm
+
+/* Transpose a block of 4x4 coefficients in four 64-bit registers */
+.macro transpose_4x4_32 x0, x0len, x1, x1len, x2, x2len, x3, x3len, xi, xilen
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\x0len, \x0\x0len, \x2\x2len
+    trn2            \x2\x2len, \xi\x0len, \x2\x2len
+    mov             \xi\xilen, \x1\xilen
+    trn1            \x1\x1len, \x1\x1len, \x3\x3len
+    trn2            \x3\x3len, \xi\x1len, \x3\x3len
+.endm
+
+.macro transpose_4x4_16 x0, x0len, x1, x1len, x2, x2len, x3, x3len, xi, xilen
+    mov             \xi\xilen, \x0\xilen
+    trn1            \x0\x0len, \x0\x0len, \x1\x1len
+    trn2            \x1\x2len, \xi\x0len, \x1\x2len
+    mov             \xi\xilen, \x2\xilen
+    trn1            \x2\x2len, \x2\x2len, \x3\x3len
+    trn2            \x3\x2len, \xi\x1len, \x3\x3len
+.endm
+
+.macro transpose_4x4 x0, x1, x2, x3, x5
+    transpose_4x4_16 \x0, .4h, \x1, .4h, \x2, .4h, \x3, .4h, \x5, .16b
+    transpose_4x4_32 \x0, .2s, \x1, .2s, \x2, .2s, \x3, .2s, \x5, .16b
+.endm
+
+.macro transpose_8x8 l0, l1, l2, l3, l4, l5, l6, l7, t0, t1, t2, t3
+    trn1            \t0\().8h, \l0\().8h, \l1\().8h
+    trn1            \t1\().8h, \l2\().8h, \l3\().8h
+    trn1            \t2\().8h, \l4\().8h, \l5\().8h
+    trn1            \t3\().8h, \l6\().8h, \l7\().8h
+    trn2            \l1\().8h, \l0\().8h, \l1\().8h
+    trn2            \l3\().8h, \l2\().8h, \l3\().8h
+    trn2            \l5\().8h, \l4\().8h, \l5\().8h
+    trn2            \l7\().8h, \l6\().8h, \l7\().8h
+
+    trn1            \l4\().4s, \t2\().4s, \t3\().4s
+    trn2            \t3\().4s, \t2\().4s, \t3\().4s
+    trn1            \t2\().4s, \t0\().4s, \t1\().4s
+    trn2            \l2\().4s, \t0\().4s, \t1\().4s
+    trn1            \t0\().4s, \l1\().4s, \l3\().4s
+    trn2            \l3\().4s, \l1\().4s, \l3\().4s
+    trn2            \t1\().4s, \l5\().4s, \l7\().4s
+    trn1            \l5\().4s, \l5\().4s, \l7\().4s
+
+    trn2            \l6\().2d, \l2\().2d, \t3\().2d
+    trn1            \l0\().2d, \t2\().2d, \l4\().2d
+    trn1            \l1\().2d, \t0\().2d, \l5\().2d
+    trn2            \l7\().2d, \l3\().2d, \t1\().2d
+    trn1            \l2\().2d, \l2\().2d, \t3\().2d
+    trn2            \l4\().2d, \t2\().2d, \l4\().2d
+    trn1            \l3\().2d, \l3\().2d, \t1\().2d
+    trn2            \l5\().2d, \t0\().2d, \l5\().2d
+.endm
+
+
+#define CENTERJSAMPLE 128
+
+/*****************************************************************************/
+
+/*
+ * Perform dequantization and inverse DCT on one block of coefficients.
+ *
+ * GLOBAL(void)
+ * jsimd_idct_islow_neon (void *dct_table, JCOEFPTR coef_block,
+ *                        JSAMPARRAY output_buf, JDIMENSION output_col)
+ */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+
+#define F_0_298  2446  /* FIX(0.298631336) */
+#define F_0_390  3196  /* FIX(0.390180644) */
+#define F_0_541  4433  /* FIX(0.541196100) */
+#define F_0_765  6270  /* FIX(0.765366865) */
+#define F_0_899  7373  /* FIX(0.899976223) */
+#define F_1_175  9633  /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_idct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+
+#define XFIX_P_0_298 v0.h[0]
+#define XFIX_N_0_390 v0.h[1]
+#define XFIX_P_0_541 v0.h[2]
+#define XFIX_P_0_765 v0.h[3]
+#define XFIX_N_0_899 v0.h[4]
+#define XFIX_P_1_175 v0.h[5]
+#define XFIX_P_1_501 v0.h[6]
+#define XFIX_N_1_847 v0.h[7]
+#define XFIX_N_1_961 v1.h[0]
+#define XFIX_P_2_053 v1.h[1]
+#define XFIX_N_2_562 v1.h[2]
+#define XFIX_P_3_072 v1.h[3]
+
+asm_function jsimd_idct_islow_neon
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x9
+    TMP4            .req x10
+    TMP5            .req x11
+    TMP6            .req x12
+    TMP7            .req x13
+    TMP8            .req x14
+
+    sub             sp, sp, #64
+    adr             x15, Ljsimd_idct_islow_neon_consts
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], #32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], #32
+    ld1             {v0.8h, v1.8h}, [x15]
+    ld1             {v2.8h, v3.8h, v4.8h, v5.8h}, [COEF_BLOCK], #64
+    ld1             {v18.8h, v19.8h, v20.8h, v21.8h}, [DCT_TABLE], #64
+    ld1             {v6.8h, v7.8h, v8.8h, v9.8h}, [COEF_BLOCK], #64
+    ld1             {v22.8h, v23.8h, v24.8h, v25.8h}, [DCT_TABLE], #64
+
+    cmeq            v16.8h, v3.8h, #0
+    cmeq            v26.8h, v4.8h, #0
+    cmeq            v27.8h, v5.8h, #0
+    cmeq            v28.8h, v6.8h, #0
+    cmeq            v29.8h, v7.8h, #0
+    cmeq            v30.8h, v8.8h, #0
+    cmeq            v31.8h, v9.8h, #0
+
+    and             v10.16b, v16.16b, v26.16b
+    and             v11.16b, v27.16b, v28.16b
+    and             v12.16b, v29.16b, v30.16b
+    and             v13.16b, v31.16b, v10.16b
+    and             v14.16b, v11.16b, v12.16b
+    mul             v2.8h, v2.8h, v18.8h
+    and             v15.16b, v13.16b, v14.16b
+    shl             v10.8h, v2.8h, #(PASS1_BITS)
+    sqxtn           v16.8b, v15.8h
+    mov             TMP1, v16.d[0]
+    sub             sp, sp, #64
+    mvn             TMP2, TMP1
+
+    cbnz            TMP2, 2f
+    /* case all AC coeffs are zeros */
+    dup             v2.2d, v10.d[0]
+    dup             v6.2d, v10.d[1]
+    mov             v3.16b, v2.16b
+    mov             v7.16b, v6.16b
+    mov             v4.16b, v2.16b
+    mov             v8.16b, v6.16b
+    mov             v5.16b, v2.16b
+    mov             v9.16b, v6.16b
+1:
+    /* for this transpose, we should organise data like this:
+     * 00, 01, 02, 03, 40, 41, 42, 43
+     * 10, 11, 12, 13, 50, 51, 52, 53
+     * 20, 21, 22, 23, 60, 61, 62, 63
+     * 30, 31, 32, 33, 70, 71, 72, 73
+     * 04, 05, 06, 07, 44, 45, 46, 47
+     * 14, 15, 16, 17, 54, 55, 56, 57
+     * 24, 25, 26, 27, 64, 65, 66, 67
+     * 34, 35, 36, 37, 74, 75, 76, 77
+     */
+    trn1            v28.8h, v2.8h, v3.8h
+    trn1            v29.8h, v4.8h, v5.8h
+    trn1            v30.8h, v6.8h, v7.8h
+    trn1            v31.8h, v8.8h, v9.8h
+    trn2            v16.8h, v2.8h, v3.8h
+    trn2            v17.8h, v4.8h, v5.8h
+    trn2            v18.8h, v6.8h, v7.8h
+    trn2            v19.8h, v8.8h, v9.8h
+    trn1            v2.4s, v28.4s, v29.4s
+    trn1            v6.4s, v30.4s, v31.4s
+    trn1            v3.4s, v16.4s, v17.4s
+    trn1            v7.4s, v18.4s, v19.4s
+    trn2            v4.4s, v28.4s, v29.4s
+    trn2            v8.4s, v30.4s, v31.4s
+    trn2            v5.4s, v16.4s, v17.4s
+    trn2            v9.4s, v18.4s, v19.4s
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    shrn            v2.4h, v18.4s, #16  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v9.4h, v20.4s, #16  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn            v3.4h, v22.4s, #16  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v8.4h, v24.4s, #16  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn            v4.4h, v26.4s, #16  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v7.4h, v28.4s, #16  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn            v5.4h, v14.4s, #16  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn            v6.4h, v16.4s, #16  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v2.8h, v19.4s, #16  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v9.8h, v21.4s, #16  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v3.8h, v23.4s, #16  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v8.8h, v25.4s, #16  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v4.8h, v27.4s, #16  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v7.8h, v29.4s, #16  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v5.8h, v15.4s, #16  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS+PASS1_BITS+3) */
+    shrn2           v6.8h, v17.4s, #16  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS+PASS1_BITS+3) */
+    movi            v0.16b, #(CENTERJSAMPLE)
+    /* Prepare pointers (dual-issue with NEON instructions) */
+      ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    sqrshrn         v28.8b, v2.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP3, TMP4, [OUTPUT_BUF], 16
+    sqrshrn         v29.8b, v3.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP1, TMP1, OUTPUT_COL
+    sqrshrn         v30.8b, v4.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP2, TMP2, OUTPUT_COL
+    sqrshrn         v31.8b, v5.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP3, TMP3, OUTPUT_COL
+    sqrshrn2        v28.16b, v6.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP4, TMP4, OUTPUT_COL
+    sqrshrn2        v29.16b, v7.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP5, TMP6, [OUTPUT_BUF], 16
+    sqrshrn2        v30.16b, v8.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      ldp             TMP7, TMP8, [OUTPUT_BUF], 16
+    sqrshrn2        v31.16b, v9.8h, #(CONST_BITS+PASS1_BITS+3-16)
+      add             TMP5, TMP5, OUTPUT_COL
+    add             v16.16b, v28.16b, v0.16b
+      add             TMP6, TMP6, OUTPUT_COL
+    add             v18.16b, v29.16b, v0.16b
+      add             TMP7, TMP7, OUTPUT_COL
+    add             v20.16b, v30.16b, v0.16b
+      add             TMP8, TMP8, OUTPUT_COL
+    add             v22.16b, v31.16b, v0.16b
+
+    /* Transpose the final 8-bit samples */
+    trn1            v28.16b, v16.16b, v18.16b
+    trn1            v30.16b, v20.16b, v22.16b
+    trn2            v29.16b, v16.16b, v18.16b
+    trn2            v31.16b, v20.16b, v22.16b
+
+    trn1            v16.8h, v28.8h, v30.8h
+    trn2            v18.8h, v28.8h, v30.8h
+    trn1            v20.8h, v29.8h, v31.8h
+    trn2            v22.8h, v29.8h, v31.8h
+
+    uzp1            v28.4s, v16.4s, v18.4s
+    uzp2            v30.4s, v16.4s, v18.4s
+    uzp1            v29.4s, v20.4s, v22.4s
+    uzp2            v31.4s, v20.4s, v22.4s
+
+    /* Store results to the output buffer */
+    st1             {v28.d}[0], [TMP1]
+    st1             {v29.d}[0], [TMP2]
+    st1             {v28.d}[1], [TMP3]
+    st1             {v29.d}[1], [TMP4]
+    st1             {v30.d}[0], [TMP5]
+    st1             {v31.d}[0], [TMP6]
+    st1             {v30.d}[1], [TMP7]
+    st1             {v31.d}[1], [TMP8]
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], #32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], #32
+    blr             x30
+
+.balign 16
+2:
+    mul             v3.8h, v3.8h, v19.8h
+    mul             v4.8h, v4.8h, v20.8h
+    mul             v5.8h, v5.8h, v21.8h
+    add             TMP4, xzr, TMP2, LSL #32
+    mul             v6.8h, v6.8h, v22.8h
+    mul             v7.8h, v7.8h, v23.8h
+    adds            TMP3, xzr, TMP2, LSR #32
+    mul             v8.8h, v8.8h, v24.8h
+    mul             v9.8h, v9.8h, v25.8h
+    b.ne            3f
+    /* Right AC coef is zero */
+    dup             v15.2d, v10.d[1]
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.4h, v4.4h, v8.4h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.4h, v2.4h, v6.4h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    sub             v26.4h, v2.4h, v6.4h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.4h, v9.4h, v5.4h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.4h, v7.4h, v3.4h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.4h, v9.4h, v3.4h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.4h, v7.4h, v5.4h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.4h, v22.4h, v24.4h  /* z5 = z3 + z4 */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    mov             v6.16b, v15.16b
+    mov             v7.16b, v15.16b
+    mov             v8.16b, v15.16b
+    mov             v9.16b, v15.16b
+    b               1b
+
+.balign 16
+3:
+    cbnz            TMP4, 4f
+    /* Left AC coef is zero */
+    dup             v14.2d, v10.d[0]
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    mov             v2.16b, v14.16b
+    mov             v3.16b, v14.16b
+    mov             v4.16b, v14.16b
+    mov             v5.16b, v14.16b
+    rshrn           v6.4h, v19.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+.balign 16
+4:
+    /* "No" AC coef is zero */
+    /* Even part: reverse the even part of the forward DCT. */
+    add             v18.8h, v4.8h, v8.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*2], quantptr[DCTSIZE*2]) + DEQUANTIZE(inptr[DCTSIZE*6], quantptr[DCTSIZE*6]) */
+    add             v22.8h, v2.8h, v6.8h           /* z2 + z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) + DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull2          v19.4s, v18.8h, XFIX_P_0_541   /* z1h z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sub             v26.8h, v2.8h, v6.8h           /* z2 - z3 = DEQUANTIZE(inptr[DCTSIZE*0], quantptr[DCTSIZE*0]) - DEQUANTIZE(inptr[DCTSIZE*4], quantptr[DCTSIZE*4]) */
+    smull           v18.4s, v18.4h, XFIX_P_0_541   /* z1l z1 = MULTIPLY(z2 + z3, FIX_0_541196100); */
+    sshll2          v23.4s, v22.8h, #(CONST_BITS)  /* tmp0h tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    mov             v21.16b, v19.16b               /* tmp3 = z1 */
+    mov             v20.16b, v18.16b               /* tmp3 = z1 */
+    smlal2          v19.4s, v8.8h, XFIX_N_1_847    /* tmp2h tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    smlal           v18.4s, v8.4h, XFIX_N_1_847    /* tmp2l tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); */
+    sshll2          v27.4s, v26.8h, #(CONST_BITS)  /* tmp1h tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    smlal2          v21.4s, v4.8h, XFIX_P_0_765    /* tmp3h tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    smlal           v20.4s, v4.4h, XFIX_P_0_765    /* tmp3l tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); */
+    sshll           v22.4s, v22.4h, #(CONST_BITS)  /* tmp0l tmp0 = LEFT_SHIFT(z2 + z3, CONST_BITS); */
+    sshll           v26.4s, v26.4h, #(CONST_BITS)  /* tmp1l tmp1 = LEFT_SHIFT(z2 - z3, CONST_BITS); */
+    add             v2.4s, v22.4s, v20.4s          /* tmp10l tmp10 = tmp0 + tmp3; */
+    sub             v6.4s, v22.4s, v20.4s          /* tmp13l tmp13 = tmp0 - tmp3; */
+    add             v8.4s, v26.4s, v18.4s          /* tmp11l tmp11 = tmp1 + tmp2; */
+    sub             v4.4s, v26.4s, v18.4s          /* tmp12l tmp12 = tmp1 - tmp2; */
+    add             v28.4s, v23.4s, v21.4s         /* tmp10h tmp10 = tmp0 + tmp3; */
+    sub             v31.4s, v23.4s, v21.4s         /* tmp13h tmp13 = tmp0 - tmp3; */
+    add             v29.4s, v27.4s, v19.4s         /* tmp11h tmp11 = tmp1 + tmp2; */
+    sub             v30.4s, v27.4s, v19.4s         /* tmp12h tmp12 = tmp1 - tmp2; */
+
+    /* Odd part per figure 8; the matrix is unitary and hence its
+     * transpose is its inverse.  i0..i3 are y7,y5,y3,y1 respectively.
+     */
+
+    add             v22.8h, v9.8h, v5.8h    /* z3 = tmp0 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v24.8h, v7.8h, v3.8h    /* z4 = tmp1 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v18.8h, v9.8h, v3.8h    /* z1 = tmp0 + tmp3 = DEQUANTIZE(inptr[DCTSIZE*7], quantptr[DCTSIZE*7]) + DEQUANTIZE(inptr[DCTSIZE*1], quantptr[DCTSIZE*1]) */
+    add             v20.8h, v7.8h, v5.8h    /* z2 = tmp1 + tmp2 = DEQUANTIZE(inptr[DCTSIZE*5], quantptr[DCTSIZE*5]) + DEQUANTIZE(inptr[DCTSIZE*3], quantptr[DCTSIZE*3]) */
+    add             v26.8h, v22.8h, v24.8h  /* z5 = z3 + z4 */
+
+    smull2          v11.4s, v9.8h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull2          v13.4s, v7.8h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull2          v15.4s, v5.8h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull2          v17.4s, v3.8h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull2          v27.4s, v26.8h, XFIX_P_1_175  /* z5h z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull2          v23.4s, v22.8h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull2          v25.4s, v24.8h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull2          v19.4s, v18.8h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull2          v21.4s, v20.8h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    smull           v10.4s, v9.4h, XFIX_P_0_298   /* tmp0 = MULTIPLY(tmp0, FIX_0_298631336) */
+    smull           v12.4s, v7.4h, XFIX_P_2_053   /* tmp1 = MULTIPLY(tmp1, FIX_2_053119869) */
+    smull           v14.4s, v5.4h, XFIX_P_3_072   /* tmp2 = MULTIPLY(tmp2, FIX_3_072711026) */
+    smull           v16.4s, v3.4h, XFIX_P_1_501   /* tmp3 = MULTIPLY(tmp3, FIX_1_501321110) */
+    smull           v26.4s, v26.4h, XFIX_P_1_175  /* z5l z5 = MULTIPLY(z3 + z4, FIX_1_175875602) */
+    smull           v22.4s, v22.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560) */
+    smull           v24.4s, v24.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644) */
+    smull           v18.4s, v18.4h, XFIX_N_0_899  /* z1 = MULTIPLY(z1, - FIX_0_899976223) */
+    smull           v20.4s, v20.4h, XFIX_N_2_562  /* z2 = MULTIPLY(z2, - FIX_2_562915447) */
+
+    add             v23.4s, v23.4s, v27.4s  /* z3 += z5 */
+    add             v22.4s, v22.4s, v26.4s  /* z3 += z5 */
+    add             v25.4s, v25.4s, v27.4s  /* z4 += z5 */
+    add             v24.4s, v24.4s, v26.4s  /* z4 += z5 */
+
+    add             v11.4s, v11.4s, v19.4s  /* tmp0 += z1 */
+    add             v10.4s, v10.4s, v18.4s  /* tmp0 += z1 */
+    add             v13.4s, v13.4s, v21.4s  /* tmp1 += z2 */
+    add             v12.4s, v12.4s, v20.4s  /* tmp1 += z2 */
+    add             v15.4s, v15.4s, v21.4s  /* tmp2 += z2 */
+    add             v14.4s, v14.4s, v20.4s  /* tmp2 += z2 */
+    add             v17.4s, v17.4s, v19.4s  /* tmp3 += z1 */
+    add             v16.4s, v16.4s, v18.4s  /* tmp3 += z1 */
+
+    add             v11.4s, v11.4s, v23.4s  /* tmp0 += z3 */
+    add             v10.4s, v10.4s, v22.4s  /* tmp0 += z3 */
+    add             v13.4s, v13.4s, v25.4s  /* tmp1 += z4 */
+    add             v12.4s, v12.4s, v24.4s  /* tmp1 += z4 */
+    add             v17.4s, v17.4s, v25.4s  /* tmp3 += z4 */
+    add             v16.4s, v16.4s, v24.4s  /* tmp3 += z4 */
+    add             v15.4s, v15.4s, v23.4s  /* tmp2 += z3 */
+    add             v14.4s, v14.4s, v22.4s  /* tmp2 += z3 */
+
+    /* Final output stage: inputs are tmp10..tmp13, tmp0..tmp3 */
+
+    add             v18.4s, v2.4s, v16.4s   /* tmp10 + tmp3 */
+    add             v19.4s, v28.4s, v17.4s  /* tmp10 + tmp3 */
+    sub             v20.4s, v2.4s, v16.4s   /* tmp10 - tmp3 */
+    sub             v21.4s, v28.4s, v17.4s  /* tmp10 - tmp3 */
+    add             v22.4s, v8.4s, v14.4s   /* tmp11 + tmp2 */
+    add             v23.4s, v29.4s, v15.4s  /* tmp11 + tmp2 */
+    sub             v24.4s, v8.4s, v14.4s   /* tmp11 - tmp2 */
+    sub             v25.4s, v29.4s, v15.4s  /* tmp11 - tmp2 */
+    add             v26.4s, v4.4s, v12.4s   /* tmp12 + tmp1 */
+    add             v27.4s, v30.4s, v13.4s  /* tmp12 + tmp1 */
+    sub             v28.4s, v4.4s, v12.4s   /* tmp12 - tmp1 */
+    sub             v29.4s, v30.4s, v13.4s  /* tmp12 - tmp1 */
+    add             v14.4s, v6.4s, v10.4s   /* tmp13 + tmp0 */
+    add             v15.4s, v31.4s, v11.4s  /* tmp13 + tmp0 */
+    sub             v16.4s, v6.4s, v10.4s   /* tmp13 - tmp0 */
+    sub             v17.4s, v31.4s, v11.4s  /* tmp13 - tmp0 */
+
+    rshrn           v2.4h, v18.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v3.4h, v22.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v4.4h, v26.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v5.4h, v14.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn           v6.4h, v19.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*0] = (int) DESCALE(tmp10 + tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn           v7.4h, v23.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*1] = (int) DESCALE(tmp11 + tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn           v8.4h, v27.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*2] = (int) DESCALE(tmp12 + tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn           v9.4h, v15.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*3] = (int) DESCALE(tmp13 + tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v2.8h, v16.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v3.8h, v28.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v4.8h, v24.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v5.8h, v20.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    rshrn2          v6.8h, v17.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*4] = (int) DESCALE(tmp13 - tmp0, CONST_BITS-PASS1_BITS) */
+    rshrn2          v7.8h, v29.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*5] = (int) DESCALE(tmp12 - tmp1, CONST_BITS-PASS1_BITS) */
+    rshrn2          v8.8h, v25.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*6] = (int) DESCALE(tmp11 - tmp2, CONST_BITS-PASS1_BITS) */
+    rshrn2          v9.8h, v21.4s, #(CONST_BITS-PASS1_BITS)  /* wsptr[DCTSIZE*7] = (int) DESCALE(tmp10 - tmp3, CONST_BITS-PASS1_BITS) */
+    b               1b
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+
+#undef CENTERJSAMPLE
+#undef CONST_BITS
+#undef PASS1_BITS
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_ifast_neon
+ *
+ * This function contains a fast, not so accurate integer implementation of
+ * the inverse DCT (Discrete Cosine Transform). It uses the same calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_ifast'
+ * function from jidctfst.c
+ *
+ * Normally 1-D AAN DCT needs 5 multiplications and 29 additions.
+ * But in ARM NEON case some extra additions are required because VQDMULH
+ * instruction can't handle the constants larger than 1. So the expressions
+ * like "x * 1.082392200" have to be converted to "x * 0.082392200 + x",
+ * which introduces an extra addition. Overall, there are 6 extra additions
+ * per 1-D IDCT pass, totalling to 5 VQDMULH and 35 VADD/VSUB instructions.
+ */
+
+#define XFIX_1_082392200 v0.h[0]
+#define XFIX_1_414213562 v0.h[1]
+#define XFIX_1_847759065 v0.h[2]
+#define XFIX_2_613125930 v0.h[3]
+
+.balign 16
+Ljsimd_idct_ifast_neon_consts:
+  .short (277 * 128 - 256 * 128)  /* XFIX_1_082392200 */
+  .short (362 * 128 - 256 * 128)  /* XFIX_1_414213562 */
+  .short (473 * 128 - 256 * 128)  /* XFIX_1_847759065 */
+  .short (669 * 128 - 512 * 128)  /* XFIX_2_613125930 */
+
+asm_function jsimd_idct_ifast_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x9
+    TMP4            .req x10
+    TMP5            .req x11
+    TMP6            .req x12
+    TMP7            .req x13
+    TMP8            .req x14
+
+    /* Load and dequantize coefficients into NEON registers
+     * with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17     ( v16.8h )
+     *   1 | d18     | d19     ( v17.8h )
+     *   2 | d20     | d21     ( v18.8h )
+     *   3 | d22     | d23     ( v19.8h )
+     *   4 | d24     | d25     ( v20.8h )
+     *   5 | d26     | d27     ( v21.8h )
+     *   6 | d28     | d29     ( v22.8h )
+     *   7 | d30     | d31     ( v23.8h )
+     */
+    /* Save NEON registers used in fast IDCT */
+    adr             TMP5, Ljsimd_idct_ifast_neon_consts
+    ld1             {v16.8h, v17.8h}, [COEF_BLOCK], 32
+    ld1             {v0.8h, v1.8h}, [DCT_TABLE], 32
+    ld1             {v18.8h, v19.8h}, [COEF_BLOCK], 32
+    mul             v16.8h, v16.8h, v0.8h
+    ld1             {v2.8h, v3.8h}, [DCT_TABLE], 32
+    mul             v17.8h, v17.8h, v1.8h
+    ld1             {v20.8h, v21.8h}, [COEF_BLOCK], 32
+    mul             v18.8h, v18.8h, v2.8h
+    ld1             {v0.8h, v1.8h}, [DCT_TABLE], 32
+    mul             v19.8h, v19.8h, v3.8h
+    ld1             {v22.8h, v23.8h}, [COEF_BLOCK], 32
+    mul             v20.8h, v20.8h, v0.8h
+    ld1             {v2.8h, v3.8h}, [DCT_TABLE], 32
+    mul             v22.8h, v22.8h, v2.8h
+    mul             v21.8h, v21.8h, v1.8h
+    ld1             {v0.4h}, [TMP5]        /* load constants */
+    mul             v23.8h, v23.8h, v3.8h
+
+    /* 1-D IDCT, pass 1 */
+    sub             v2.8h, v18.8h, v22.8h
+    add             v22.8h, v18.8h, v22.8h
+    sub             v1.8h, v19.8h, v21.8h
+    add             v21.8h, v19.8h, v21.8h
+    sub             v5.8h, v17.8h, v23.8h
+    add             v23.8h, v17.8h, v23.8h
+    sqdmulh         v4.8h, v2.8h, XFIX_1_414213562
+    sqdmulh         v6.8h, v1.8h, XFIX_2_613125930
+    add             v3.8h, v1.8h, v1.8h
+    sub             v1.8h, v5.8h, v1.8h
+    add             v18.8h, v2.8h, v4.8h
+    sqdmulh         v4.8h, v1.8h, XFIX_1_847759065
+    sub             v2.8h, v23.8h, v21.8h
+    add             v3.8h, v3.8h, v6.8h
+    sqdmulh         v6.8h, v2.8h, XFIX_1_414213562
+    add             v1.8h, v1.8h, v4.8h
+    sqdmulh         v4.8h, v5.8h, XFIX_1_082392200
+    sub             v18.8h, v18.8h, v22.8h
+    add             v2.8h, v2.8h, v6.8h
+    sub             v6.8h, v16.8h, v20.8h
+    add             v20.8h, v16.8h, v20.8h
+    add             v17.8h, v5.8h, v4.8h
+    add             v5.8h, v6.8h, v18.8h
+    sub             v18.8h, v6.8h, v18.8h
+    add             v6.8h, v23.8h, v21.8h
+    add             v16.8h, v20.8h, v22.8h
+    sub             v3.8h, v6.8h, v3.8h
+    sub             v20.8h, v20.8h, v22.8h
+    sub             v3.8h, v3.8h, v1.8h
+    sub             v1.8h, v17.8h, v1.8h
+    add             v2.8h, v3.8h, v2.8h
+    sub             v23.8h, v16.8h, v6.8h
+    add             v1.8h, v1.8h, v2.8h
+    add             v16.8h, v16.8h, v6.8h
+    add             v22.8h, v5.8h, v3.8h
+    sub             v17.8h, v5.8h, v3.8h
+    sub             v21.8h, v18.8h, v2.8h
+    add             v18.8h, v18.8h, v2.8h
+    sub             v19.8h, v20.8h, v1.8h
+    add             v20.8h, v20.8h, v1.8h
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v28, v29, v30, v31
+    /* 1-D IDCT, pass 2 */
+    sub             v2.8h, v18.8h, v22.8h
+    add             v22.8h, v18.8h, v22.8h
+    sub             v1.8h, v19.8h, v21.8h
+    add             v21.8h, v19.8h, v21.8h
+    sub             v5.8h, v17.8h, v23.8h
+    add             v23.8h, v17.8h, v23.8h
+    sqdmulh         v4.8h, v2.8h, XFIX_1_414213562
+    sqdmulh         v6.8h, v1.8h, XFIX_2_613125930
+    add             v3.8h, v1.8h, v1.8h
+    sub             v1.8h, v5.8h, v1.8h
+    add             v18.8h, v2.8h, v4.8h
+    sqdmulh         v4.8h, v1.8h, XFIX_1_847759065
+    sub             v2.8h, v23.8h, v21.8h
+    add             v3.8h, v3.8h, v6.8h
+    sqdmulh         v6.8h, v2.8h, XFIX_1_414213562
+    add             v1.8h, v1.8h, v4.8h
+    sqdmulh         v4.8h, v5.8h, XFIX_1_082392200
+    sub             v18.8h, v18.8h, v22.8h
+    add             v2.8h, v2.8h, v6.8h
+    sub             v6.8h, v16.8h, v20.8h
+    add             v20.8h, v16.8h, v20.8h
+    add             v17.8h, v5.8h, v4.8h
+    add             v5.8h, v6.8h, v18.8h
+    sub             v18.8h, v6.8h, v18.8h
+    add             v6.8h, v23.8h, v21.8h
+    add             v16.8h, v20.8h, v22.8h
+    sub             v3.8h, v6.8h, v3.8h
+    sub             v20.8h, v20.8h, v22.8h
+    sub             v3.8h, v3.8h, v1.8h
+    sub             v1.8h, v17.8h, v1.8h
+    add             v2.8h, v3.8h, v2.8h
+    sub             v23.8h, v16.8h, v6.8h
+    add             v1.8h, v1.8h, v2.8h
+    add             v16.8h, v16.8h, v6.8h
+    add             v22.8h, v5.8h, v3.8h
+    sub             v17.8h, v5.8h, v3.8h
+    sub             v21.8h, v18.8h, v2.8h
+    add             v18.8h, v18.8h, v2.8h
+    sub             v19.8h, v20.8h, v1.8h
+    add             v20.8h, v20.8h, v1.8h
+    /* Descale to 8-bit and range limit */
+    movi            v0.16b, #0x80
+      /* Prepare pointers (dual-issue with NEON instructions) */
+      ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    sqshrn          v28.8b, v16.8h, #5
+      ldp             TMP3, TMP4, [OUTPUT_BUF], 16
+    sqshrn          v29.8b, v17.8h, #5
+      add             TMP1, TMP1, OUTPUT_COL
+    sqshrn          v30.8b, v18.8h, #5
+      add             TMP2, TMP2, OUTPUT_COL
+    sqshrn          v31.8b, v19.8h, #5
+      add             TMP3, TMP3, OUTPUT_COL
+    sqshrn2         v28.16b, v20.8h, #5
+      add             TMP4, TMP4, OUTPUT_COL
+    sqshrn2         v29.16b, v21.8h, #5
+      ldp             TMP5, TMP6, [OUTPUT_BUF], 16
+    sqshrn2         v30.16b, v22.8h, #5
+      ldp             TMP7, TMP8, [OUTPUT_BUF], 16
+    sqshrn2         v31.16b, v23.8h, #5
+      add             TMP5, TMP5, OUTPUT_COL
+    add             v16.16b, v28.16b, v0.16b
+      add             TMP6, TMP6, OUTPUT_COL
+    add             v18.16b, v29.16b, v0.16b
+      add             TMP7, TMP7, OUTPUT_COL
+    add             v20.16b, v30.16b, v0.16b
+      add             TMP8, TMP8, OUTPUT_COL
+    add             v22.16b, v31.16b, v0.16b
+
+    /* Transpose the final 8-bit samples */
+    trn1            v28.16b, v16.16b, v18.16b
+    trn1            v30.16b, v20.16b, v22.16b
+    trn2            v29.16b, v16.16b, v18.16b
+    trn2            v31.16b, v20.16b, v22.16b
+
+    trn1            v16.8h, v28.8h, v30.8h
+    trn2            v18.8h, v28.8h, v30.8h
+    trn1            v20.8h, v29.8h, v31.8h
+    trn2            v22.8h, v29.8h, v31.8h
+
+    uzp1            v28.4s, v16.4s, v18.4s
+    uzp2            v30.4s, v16.4s, v18.4s
+    uzp1            v29.4s, v20.4s, v22.4s
+    uzp2            v31.4s, v20.4s, v22.4s
+
+    /* Store results to the output buffer */
+    st1             {v28.d}[0], [TMP1]
+    st1             {v29.d}[0], [TMP2]
+    st1             {v28.d}[1], [TMP3]
+    st1             {v29.d}[1], [TMP4]
+    st1             {v30.d}[0], [TMP5]
+    st1             {v31.d}[0], [TMP6]
+    st1             {v30.d}[1], [TMP7]
+    st1             {v31.d}[1], [TMP8]
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_4x4_neon
+ *
+ * This function contains inverse-DCT code for getting reduced-size
+ * 4x4 pixels output from an 8x8 DCT block. It uses the same  calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_4x4'
+ * function from jpeg-6b (jidctred.c).
+ *
+ * NOTE: jpeg-8 has an improved implementation of 4x4 inverse-DCT, which
+ *       requires much less arithmetic operations and hence should be faster.
+ *       The primary purpose of this particular NEON optimized function is
+ *       bit exact compatibility with jpeg-6b.
+ *
+ * TODO: a bit better instructions scheduling can be achieved by expanding
+ *       idct_helper/transpose_4x4 macros and reordering instructions,
+ *       but readability will suffer somewhat.
+ */
+
+#define CONST_BITS  13
+
+#define FIX_0_211164243 (1730)   /* FIX(0.211164243) */
+#define FIX_0_509795579 (4176)   /* FIX(0.509795579) */
+#define FIX_0_601344887 (4926)   /* FIX(0.601344887) */
+#define FIX_0_720959822 (5906)   /* FIX(0.720959822) */
+#define FIX_0_765366865 (6270)   /* FIX(0.765366865) */
+#define FIX_0_850430095 (6967)   /* FIX(0.850430095) */
+#define FIX_0_899976223 (7373)   /* FIX(0.899976223) */
+#define FIX_1_061594337 (8697)   /* FIX(1.061594337) */
+#define FIX_1_272758580 (10426)  /* FIX(1.272758580) */
+#define FIX_1_451774981 (11893)  /* FIX(1.451774981) */
+#define FIX_1_847759065 (15137)  /* FIX(1.847759065) */
+#define FIX_2_172734803 (17799)  /* FIX(2.172734803) */
+#define FIX_2_562915447 (20995)  /* FIX(2.562915447) */
+#define FIX_3_624509785 (29692)  /* FIX(3.624509785) */
+
+.balign 16
+Ljsimd_idct_4x4_neon_consts:
+  .short FIX_1_847759065      /* v0.h[0] */
+  .short -FIX_0_765366865     /* v0.h[1] */
+  .short -FIX_0_211164243     /* v0.h[2] */
+  .short FIX_1_451774981      /* v0.h[3] */
+  .short -FIX_2_172734803     /* d1[0] */
+  .short FIX_1_061594337      /* d1[1] */
+  .short -FIX_0_509795579     /* d1[2] */
+  .short -FIX_0_601344887     /* d1[3] */
+  .short FIX_0_899976223      /* v2.h[0] */
+  .short FIX_2_562915447      /* v2.h[1] */
+  .short 1 << (CONST_BITS+1)  /* v2.h[2] */
+  .short 0                    /* v2.h[3] */
+
+.macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29
+    smull           v28.4s, \x4, v2.h[2]
+    smlal           v28.4s, \x8, v0.h[0]
+    smlal           v28.4s, \x14, v0.h[1]
+
+    smull           v26.4s, \x16, v1.h[2]
+    smlal           v26.4s, \x12, v1.h[3]
+    smlal           v26.4s, \x10, v2.h[0]
+    smlal           v26.4s, \x6, v2.h[1]
+
+    smull           v30.4s, \x4, v2.h[2]
+    smlsl           v30.4s, \x8, v0.h[0]
+    smlsl           v30.4s, \x14, v0.h[1]
+
+    smull           v24.4s, \x16, v0.h[2]
+    smlal           v24.4s, \x12, v0.h[3]
+    smlal           v24.4s, \x10, v1.h[0]
+    smlal           v24.4s, \x6, v1.h[1]
+
+    add             v20.4s, v28.4s, v26.4s
+    sub             v28.4s, v28.4s, v26.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v28.4s, v28.4s, #\shift
+    xtn             \y26, v20.4s
+    xtn             \y29, v28.4s
+  .else
+    rshrn           \y26, v20.4s, #\shift
+    rshrn           \y29, v28.4s, #\shift
+  .endif
+
+    add             v20.4s, v30.4s, v24.4s
+    sub             v30.4s, v30.4s, v24.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v30.4s, v30.4s, #\shift
+    xtn             \y27, v20.4s
+    xtn             \y28, v30.4s
+  .else
+    rshrn           \y27, v20.4s, #\shift
+    rshrn           \y28, v30.4s, #\shift
+  .endif
+.endm
+
+asm_function jsimd_idct_4x4_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x1
+    TMP3            .req x2
+    TMP4            .req x15
+
+    /* Save all used NEON registers */
+    sub             sp, sp, 272
+    str             x15, [sp], 16
+    /* Load constants (v3.4h is just used for padding) */
+    adr             TMP4, Ljsimd_idct_4x4_neon_consts
+    st1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    ld1             {v0.4h, v1.4h, v2.4h, v3.4h}, [TMP4]
+
+    /* Load all COEF_BLOCK into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | v4.4h   | v5.4h
+     *   1 | v6.4h   | v7.4h
+     *   2 | v8.4h   | v9.4h
+     *   3 | v10.4h  | v11.4h
+     *   4 | -       | -
+     *   5 | v12.4h  | v13.4h
+     *   6 | v14.4h  | v15.4h
+     *   7 | v16.4h  | v17.4h
+     */
+    ld1             {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32
+    ld1             {v8.4h, v9.4h, v10.4h, v11.4h}, [COEF_BLOCK], 32
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v12.4h, v13.4h, v14.4h, v15.4h}, [COEF_BLOCK], 32
+    ld1             {v16.4h, v17.4h}, [COEF_BLOCK], 16
+    /* dequantize */
+    ld1             {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32
+    mul             v4.4h, v4.4h, v18.4h
+    mul             v5.4h, v5.4h, v19.4h
+    ins             v4.d[1], v5.d[0]              /* 128 bit q4 */
+    ld1             {v22.4h, v23.4h, v24.4h, v25.4h}, [DCT_TABLE], 32
+    mul             v6.4h, v6.4h, v20.4h
+    mul             v7.4h, v7.4h, v21.4h
+    ins             v6.d[1], v7.d[0]              /* 128 bit q6 */
+    mul             v8.4h, v8.4h, v22.4h
+    mul             v9.4h, v9.4h, v23.4h
+    ins             v8.d[1], v9.d[0]              /* 128 bit q8 */
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v26.4h, v27.4h, v28.4h, v29.4h}, [DCT_TABLE], 32
+    mul             v10.4h, v10.4h, v24.4h
+    mul             v11.4h, v11.4h, v25.4h
+    ins             v10.d[1], v11.d[0]            /* 128 bit q10 */
+    mul             v12.4h, v12.4h, v26.4h
+    mul             v13.4h, v13.4h, v27.4h
+    ins             v12.d[1], v13.d[0]            /* 128 bit q12 */
+    ld1             {v30.4h, v31.4h}, [DCT_TABLE], 16
+    mul             v14.4h, v14.4h, v28.4h
+    mul             v15.4h, v15.4h, v29.4h
+    ins             v14.d[1], v15.d[0]            /* 128 bit q14 */
+    mul             v16.4h, v16.4h, v30.4h
+    mul             v17.4h, v17.4h, v31.4h
+    ins             v16.d[1], v17.d[0]            /* 128 bit q16 */
+
+    /* Pass 1 */
+    idct_helper     v4.4h, v6.4h, v8.4h, v10.4h, v12.4h, v14.4h, v16.4h, 12, \
+                    v4.4h, v6.4h, v8.4h, v10.4h
+    transpose_4x4   v4, v6, v8, v10, v3
+    ins             v10.d[1], v11.d[0]
+    idct_helper     v5.4h, v7.4h, v9.4h, v11.4h, v13.4h, v15.4h, v17.4h, 12, \
+                    v5.4h, v7.4h, v9.4h, v11.4h
+    transpose_4x4   v5, v7, v9, v11, v3
+    ins             v10.d[1], v11.d[0]
+
+    /* Pass 2 */
+    idct_helper     v4.4h, v6.4h, v8.4h, v10.4h, v7.4h, v9.4h, v11.4h, 19, \
+                    v26.4h, v27.4h, v28.4h, v29.4h
+    transpose_4x4   v26, v27, v28, v29, v3
+
+    /* Range limit */
+    movi            v30.8h, #0x80
+    ins             v26.d[1], v27.d[0]
+    ins             v28.d[1], v29.d[0]
+    add             v26.8h, v26.8h, v30.8h
+    add             v28.8h, v28.8h, v30.8h
+    sqxtun          v26.8b, v26.8h
+    sqxtun          v27.8b, v28.8h
+
+    /* Store results to the output buffer */
+    ldp             TMP1, TMP2, [OUTPUT_BUF], 16
+    ldp             TMP3, TMP4, [OUTPUT_BUF]
+    add             TMP1, TMP1, OUTPUT_COL
+    add             TMP2, TMP2, OUTPUT_COL
+    add             TMP3, TMP3, OUTPUT_COL
+    add             TMP4, TMP4, OUTPUT_COL
+
+#if defined(__ARMEL__) && !RESPECT_STRICT_ALIGNMENT
+    /* We can use much less instructions on little endian systems if the
+     * OS kernel is not configured to trap unaligned memory accesses
+     */
+    st1             {v26.s}[0], [TMP1], 4
+    st1             {v27.s}[0], [TMP3], 4
+    st1             {v26.s}[1], [TMP2], 4
+    st1             {v27.s}[1], [TMP4], 4
+#else
+    st1             {v26.b}[0], [TMP1], 1
+    st1             {v27.b}[0], [TMP3], 1
+    st1             {v26.b}[1], [TMP1], 1
+    st1             {v27.b}[1], [TMP3], 1
+    st1             {v26.b}[2], [TMP1], 1
+    st1             {v27.b}[2], [TMP3], 1
+    st1             {v26.b}[3], [TMP1], 1
+    st1             {v27.b}[3], [TMP3], 1
+
+    st1             {v26.b}[4], [TMP2], 1
+    st1             {v27.b}[4], [TMP4], 1
+    st1             {v26.b}[5], [TMP2], 1
+    st1             {v27.b}[5], [TMP4], 1
+    st1             {v26.b}[6], [TMP2], 1
+    st1             {v27.b}[6], [TMP4], 1
+    st1             {v26.b}[7], [TMP2], 1
+    st1             {v27.b}[7], [TMP4], 1
+#endif
+
+    /* vpop            {v8.4h - v15.4h}    ;not available */
+    sub             sp, sp, #272
+    ldr             x15, [sp], 16
+    ld1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+
+.purgem idct_helper
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_idct_2x2_neon
+ *
+ * This function contains inverse-DCT code for getting reduced-size
+ * 2x2 pixels output from an 8x8 DCT block. It uses the same  calculations
+ * and produces exactly the same output as IJG's original 'jpeg_idct_2x2'
+ * function from jpeg-6b (jidctred.c).
+ *
+ * NOTE: jpeg-8 has an improved implementation of 2x2 inverse-DCT, which
+ *       requires much less arithmetic operations and hence should be faster.
+ *       The primary purpose of this particular NEON optimized function is
+ *       bit exact compatibility with jpeg-6b.
+ */
+
+.balign 8
+Ljsimd_idct_2x2_neon_consts:
+  .short -FIX_0_720959822  /* v14[0] */
+  .short FIX_0_850430095   /* v14[1] */
+  .short -FIX_1_272758580  /* v14[2] */
+  .short FIX_3_624509785   /* v14[3] */
+
+.macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27
+    sshll           v15.4s, \x4, #15
+    smull           v26.4s, \x6, v14.h[3]
+    smlal           v26.4s, \x10, v14.h[2]
+    smlal           v26.4s, \x12, v14.h[1]
+    smlal           v26.4s, \x16, v14.h[0]
+
+    add             v20.4s, v15.4s, v26.4s
+    sub             v15.4s, v15.4s, v26.4s
+
+  .if \shift > 16
+    srshr           v20.4s, v20.4s, #\shift
+    srshr           v15.4s, v15.4s, #\shift
+    xtn             \y26, v20.4s
+    xtn             \y27, v15.4s
+  .else
+    rshrn           \y26, v20.4s, #\shift
+    rshrn           \y27, v15.4s, #\shift
+  .endif
+.endm
+
+asm_function jsimd_idct_2x2_neon
+
+    DCT_TABLE       .req x0
+    COEF_BLOCK      .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_COL      .req x3
+    TMP1            .req x0
+    TMP2            .req x15
+
+    /* vpush           {v8.4h - v15.4h}            ; not available */
+    sub             sp, sp, 208
+    str             x15, [sp], 16
+
+    /* Load constants */
+    adr             TMP2, Ljsimd_idct_2x2_neon_consts
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v21.8b, v22.8b}, [sp], 16
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v30.8b, v31.8b}, [sp], 16
+    ld1             {v14.4h}, [TMP2]
+
+    /* Load all COEF_BLOCK into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | v4.4h   | v5.4h
+     *   1 | v6.4h   | v7.4h
+     *   2 | -       | -
+     *   3 | v10.4h  | v11.4h
+     *   4 | -       | -
+     *   5 | v12.4h  | v13.4h
+     *   6 | -       | -
+     *   7 | v16.4h  | v17.4h
+     */
+    ld1             {v4.4h, v5.4h, v6.4h, v7.4h}, [COEF_BLOCK], 32
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v10.4h, v11.4h}, [COEF_BLOCK], 16
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v12.4h, v13.4h}, [COEF_BLOCK], 16
+    add             COEF_BLOCK, COEF_BLOCK, #16
+    ld1             {v16.4h, v17.4h}, [COEF_BLOCK], 16
+    /* Dequantize */
+    ld1             {v18.4h, v19.4h, v20.4h, v21.4h}, [DCT_TABLE], 32
+    mul             v4.4h, v4.4h, v18.4h
+    mul             v5.4h, v5.4h, v19.4h
+    ins             v4.d[1], v5.d[0]
+    mul             v6.4h, v6.4h, v20.4h
+    mul             v7.4h, v7.4h, v21.4h
+    ins             v6.d[1], v7.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v24.4h, v25.4h}, [DCT_TABLE], 16
+    mul             v10.4h, v10.4h, v24.4h
+    mul             v11.4h, v11.4h, v25.4h
+    ins             v10.d[1], v11.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v26.4h, v27.4h}, [DCT_TABLE], 16
+    mul             v12.4h, v12.4h, v26.4h
+    mul             v13.4h, v13.4h, v27.4h
+    ins             v12.d[1], v13.d[0]
+    add             DCT_TABLE, DCT_TABLE, #16
+    ld1             {v30.4h, v31.4h}, [DCT_TABLE], 16
+    mul             v16.4h, v16.4h, v30.4h
+    mul             v17.4h, v17.4h, v31.4h
+    ins             v16.d[1], v17.d[0]
+
+    /* Pass 1 */
+#if 0
+    idct_helper     v4.4h, v6.4h, v10.4h, v12.4h, v16.4h, 13, v4.4h, v6.4h
+    transpose_4x4   v4.4h, v6.4h, v8.4h, v10.4h
+    idct_helper     v5.4h, v7.4h, v11.4h, v13.4h, v17.4h, 13, v5.4h, v7.4h
+    transpose_4x4   v5.4h, v7.4h, v9.4h, v11.4h
+#else
+    smull           v26.4s, v6.4h, v14.h[3]
+    smlal           v26.4s, v10.4h, v14.h[2]
+    smlal           v26.4s, v12.4h, v14.h[1]
+    smlal           v26.4s, v16.4h, v14.h[0]
+    smull           v24.4s, v7.4h, v14.h[3]
+    smlal           v24.4s, v11.4h, v14.h[2]
+    smlal           v24.4s, v13.4h, v14.h[1]
+    smlal           v24.4s, v17.4h, v14.h[0]
+    sshll           v15.4s, v4.4h, #15
+    sshll           v30.4s, v5.4h, #15
+    add             v20.4s, v15.4s, v26.4s
+    sub             v15.4s, v15.4s, v26.4s
+    rshrn           v4.4h, v20.4s, #13
+    rshrn           v6.4h, v15.4s, #13
+    add             v20.4s, v30.4s, v24.4s
+    sub             v15.4s, v30.4s, v24.4s
+    rshrn           v5.4h, v20.4s, #13
+    rshrn           v7.4h, v15.4s, #13
+    ins             v4.d[1], v5.d[0]
+    ins             v6.d[1], v7.d[0]
+    transpose       v4, v6, v3, .16b, .8h
+    transpose       v6, v10, v3, .16b, .4s
+    ins             v11.d[0], v10.d[1]
+    ins             v7.d[0], v6.d[1]
+#endif
+
+    /* Pass 2 */
+    idct_helper     v4.4h, v6.4h, v10.4h, v7.4h, v11.4h, 20, v26.4h, v27.4h
+
+    /* Range limit */
+    movi            v30.8h, #0x80
+    ins             v26.d[1], v27.d[0]
+    add             v26.8h, v26.8h, v30.8h
+    sqxtun          v30.8b, v26.8h
+    ins             v26.d[0], v30.d[0]
+    sqxtun          v27.8b, v26.8h
+
+    /* Store results to the output buffer */
+    ldp             TMP1, TMP2, [OUTPUT_BUF]
+    add             TMP1, TMP1, OUTPUT_COL
+    add             TMP2, TMP2, OUTPUT_COL
+
+    st1             {v26.b}[0], [TMP1], 1
+    st1             {v27.b}[4], [TMP1], 1
+    st1             {v26.b}[1], [TMP2], 1
+    st1             {v27.b}[5], [TMP2], 1
+
+    sub             sp, sp, #208
+    ldr             x15, [sp], 16
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v21.8b, v22.8b}, [sp], 16
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v30.8b, v31.8b}, [sp], 16
+    blr             x30
+
+    .unreq          DCT_TABLE
+    .unreq          COEF_BLOCK
+    .unreq          OUTPUT_BUF
+    .unreq          OUTPUT_COL
+    .unreq          TMP1
+    .unreq          TMP2
+
+.purgem idct_helper
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_ycc_extrgb_convert_neon
+ * jsimd_ycc_extbgr_convert_neon
+ * jsimd_ycc_extrgbx_convert_neon
+ * jsimd_ycc_extbgrx_convert_neon
+ * jsimd_ycc_extxbgr_convert_neon
+ * jsimd_ycc_extxrgb_convert_neon
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro do_load size
+  .if \size == 8
+    ld1             {v4.8b}, [U], 8
+    ld1             {v5.8b}, [V], 8
+    ld1             {v0.8b}, [Y], 8
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+  .elseif \size == 4
+    ld1             {v4.b}[0], [U], 1
+    ld1             {v4.b}[1], [U], 1
+    ld1             {v4.b}[2], [U], 1
+    ld1             {v4.b}[3], [U], 1
+    ld1             {v5.b}[0], [V], 1
+    ld1             {v5.b}[1], [V], 1
+    ld1             {v5.b}[2], [V], 1
+    ld1             {v5.b}[3], [V], 1
+    ld1             {v0.b}[0], [Y], 1
+    ld1             {v0.b}[1], [Y], 1
+    ld1             {v0.b}[2], [Y], 1
+    ld1             {v0.b}[3], [Y], 1
+  .elseif \size == 2
+    ld1             {v4.b}[4], [U], 1
+    ld1             {v4.b}[5], [U], 1
+    ld1             {v5.b}[4], [V], 1
+    ld1             {v5.b}[5], [V], 1
+    ld1             {v0.b}[4], [Y], 1
+    ld1             {v0.b}[5], [Y], 1
+  .elseif \size == 1
+    ld1             {v4.b}[6], [U], 1
+    ld1             {v5.b}[6], [V], 1
+    ld1             {v0.b}[6], [Y], 1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_store bpp, size, fast_st3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_st3 == 1
+        st3         {v10.8b, v11.8b, v12.8b}, [RGB], 24
+      .else
+        st1         {v10.b}[0], [RGB], #1
+        st1         {v11.b}[0], [RGB], #1
+        st1         {v12.b}[0], [RGB], #1
+
+        st1         {v10.b}[1], [RGB], #1
+        st1         {v11.b}[1], [RGB], #1
+        st1         {v12.b}[1], [RGB], #1
+
+        st1         {v10.b}[2], [RGB], #1
+        st1         {v11.b}[2], [RGB], #1
+        st1         {v12.b}[2], [RGB], #1
+
+        st1         {v10.b}[3], [RGB], #1
+        st1         {v11.b}[3], [RGB], #1
+        st1         {v12.b}[3], [RGB], #1
+
+        st1         {v10.b}[4], [RGB], #1
+        st1         {v11.b}[4], [RGB], #1
+        st1         {v12.b}[4], [RGB], #1
+
+        st1         {v10.b}[5], [RGB], #1
+        st1         {v11.b}[5], [RGB], #1
+        st1         {v12.b}[5], [RGB], #1
+
+        st1         {v10.b}[6], [RGB], #1
+        st1         {v11.b}[6], [RGB], #1
+        st1         {v12.b}[6], [RGB], #1
+
+        st1         {v10.b}[7], [RGB], #1
+        st1         {v11.b}[7], [RGB], #1
+        st1         {v12.b}[7], [RGB], #1
+      .endif
+    .elseif \size == 4
+      st3           {v10.b, v11.b, v12.b}[0], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[1], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[2], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[3], [RGB], 3
+    .elseif \size == 2
+      st3           {v10.b, v11.b, v12.b}[4], [RGB], 3
+      st3           {v10.b, v11.b, v12.b}[5], [RGB], 3
+    .elseif \size == 1
+      st3           {v10.b, v11.b, v12.b}[6], [RGB], 3
+    .else
+     .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      st4           {v10.8b, v11.8b, v12.8b, v13.8b}, [RGB], 32
+    .elseif \size == 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[0], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[1], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[2], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[3], [RGB], 4
+    .elseif \size == 2
+      st4           {v10.b, v11.b, v12.b, v13.b}[4], [RGB], 4
+      st4           {v10.b, v11.b, v12.b, v13.b}[5], [RGB], 4
+    .elseif \size == 1
+      st4           {v10.b, v11.b, v12.b, v13.b}[6], [RGB], 4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp==16
+    .if \size == 8
+      st1           {v25.8h}, [RGB], 16
+    .elseif \size == 4
+      st1           {v25.4h}, [RGB], 8
+    .elseif \size == 2
+      st1           {v25.h}[4], [RGB], 2
+      st1           {v25.h}[5], [RGB], 2
+    .elseif \size == 1
+      st1           {v25.h}[6], [RGB], 2
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, rsize, \
+                                           g_offs, gsize, b_offs, bsize, \
+                                           defsize, fast_st3
+
+/*
+ * 2-stage pipelined YCbCr->RGB conversion
+ */
+
+.macro do_yuv_to_rgb_stage1
+    uaddw           v6.8h, v2.8h, v4.8b     /* q3 = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+    smull           v28.4s, v6.4h, v1.h[3]  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h[3]  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb_stage2
+    rshrn           v20.4h, v20.4s, #15
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16
+    sqxtun          v1\g_offs\defsize, v20.8h
+    sqxtun          v1\r_offs\defsize, v24.8h
+    sqxtun          v1\b_offs\defsize, v28.8h
+  .else
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    sri             v25.8h, v21.8h, #5
+    sri             v25.8h, v29.8h, #11
+  .endif
+.endm
+
+.macro do_yuv_to_rgb_stage2_store_load_stage1 fast_st3
+    rshrn           v20.4h, v20.4s, #15
+    rshrn           v24.4h, v24.4s, #14
+    rshrn           v28.4h, v28.4s, #14
+    ld1             {v4.8b}, [U], 8
+    rshrn2          v20.8h, v22.4s, #15
+    rshrn2          v24.8h, v26.4s, #14
+    rshrn2          v28.8h, v30.4s, #14
+    ld1             {v5.8b}, [V], 8
+    uaddw           v20.8h, v20.8h, v0.8b
+    uaddw           v24.8h, v24.8h, v0.8b
+    uaddw           v28.8h, v28.8h, v0.8b
+  .if \bpp != 16  /**************** rgb24/rgb32 ******************************/
+    sqxtun          v1\g_offs\defsize, v20.8h
+    ld1             {v0.8b}, [Y], 8
+    sqxtun          v1\r_offs\defsize, v24.8h
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+    sqxtun          v1\b_offs\defsize, v28.8h
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+  .else  /**************************** rgb565 ********************************/
+    sqshlu          v21.8h, v20.8h, #8
+    sqshlu          v25.8h, v24.8h, #8
+    sqshlu          v29.8h, v28.8h, #8
+    uaddw           v6.8h, v2.8h, v4.8b     /* v6.16b = u - 128 */
+    uaddw           v8.8h, v2.8h, v5.8b     /* q2 = v - 128 */
+    ld1             {v0.8b}, [Y], 8
+    smull           v20.4s, v6.4h, v1.h[1]  /* multiply by -11277 */
+    smlal           v20.4s, v8.4h, v1.h[2]  /* multiply by -23401 */
+    smull2          v22.4s, v6.8h, v1.h[1]  /* multiply by -11277 */
+    smlal2          v22.4s, v8.8h, v1.h[2]  /* multiply by -23401 */
+    sri             v25.8h, v21.8h, #5
+    smull           v24.4s, v8.4h, v1.h[0]  /* multiply by 22971 */
+    smull2          v26.4s, v8.8h, v1.h[0]  /* multiply by 22971 */
+    prfm            pldl1keep, [U, #64]
+    prfm            pldl1keep, [V, #64]
+    prfm            pldl1keep, [Y, #64]
+    sri             v25.8h, v29.8h, #11
+  .endif
+    do_store        \bpp, 8, \fast_st3
+    smull           v28.4s, v6.4h, v1.h[3]  /* multiply by 29033 */
+    smull2          v30.4s, v6.8h, v1.h[3]  /* multiply by 29033 */
+.endm
+
+.macro do_yuv_to_rgb
+    do_yuv_to_rgb_stage1
+    do_yuv_to_rgb_stage2
+.endm
+
+/* Apple gas crashes on adrl, work around that by using adr.
+ * But this requires a copy of these constants for each function.
+ */
+
+.balign 16
+.if \fast_st3 == 1
+Ljsimd_ycc_\colorid\()_neon_consts:
+.else
+Ljsimd_ycc_\colorid\()_neon_slowst3_consts:
+.endif
+  .short 0,      0,     0,      0
+  .short 22971, -11277, -23401, 29033
+  .short -128,  -128,   -128,   -128
+  .short -128,  -128,   -128,   -128
+
+.if \fast_st3 == 1
+asm_function jsimd_ycc_\colorid\()_convert_neon
+.else
+asm_function jsimd_ycc_\colorid\()_convert_neon_slowst3
+.endif
+    OUTPUT_WIDTH    .req x0
+    INPUT_BUF       .req x1
+    INPUT_ROW       .req x2
+    OUTPUT_BUF      .req x3
+    NUM_ROWS        .req x4
+
+    INPUT_BUF0      .req x5
+    INPUT_BUF1      .req x6
+    INPUT_BUF2      .req x1
+
+    RGB             .req x7
+    Y               .req x8
+    U               .req x9
+    V               .req x10
+    N               .req x15
+
+    sub             sp, sp, 336
+    str             x15, [sp], 16
+
+    /* Load constants to d1, d2, d3 (v0.4h is just used for padding) */
+    .if \fast_st3 == 1
+      adr           x15, Ljsimd_ycc_\colorid\()_neon_consts
+    .else
+      adr           x15, Ljsimd_ycc_\colorid\()_neon_slowst3_consts
+    .endif
+
+    /* Save NEON registers */
+    st1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    st1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    st1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    st1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    st1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    st1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    ld1             {v0.4h, v1.4h}, [x15], 16
+    ld1             {v2.8h}, [x15]
+
+    /* Save ARM registers and handle input arguments */
+    /* push            {x4, x5, x6, x7, x8, x9, x10, x30} */
+    stp             x4, x5, [sp], 16
+    stp             x6, x7, [sp], 16
+    stp             x8, x9, [sp], 16
+    stp             x10, x30, [sp], 16
+    ldr             INPUT_BUF0, [INPUT_BUF]
+    ldr             INPUT_BUF1, [INPUT_BUF, #8]
+    ldr             INPUT_BUF2, [INPUT_BUF, #16]
+    .unreq          INPUT_BUF
+
+    /* Initially set v10, v11.4h, v12.8b, d13 to 0xFF */
+    movi            v10.16b, #255
+    movi            v13.16b, #255
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    lsl             x16, INPUT_ROW, #3
+    ldr             Y, [INPUT_BUF0, x16]
+    ldr             U, [INPUT_BUF1, x16]
+    mov             N, OUTPUT_WIDTH
+    ldr             V, [INPUT_BUF2, x16]
+    add             INPUT_ROW, INPUT_ROW, #1
+    ldr             RGB, [OUTPUT_BUF], #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         8
+    do_yuv_to_rgb_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_yuv_to_rgb_stage2_store_load_stage1 \fast_st3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_yuv_to_rgb_stage2
+    do_store        \bpp, 8, \fast_st3
+    tst             N, #7
+    b.eq            8f
+3:
+    tst             N, #4
+    b.eq            3f
+    do_load         4
+3:
+    tst             N, #2
+    b.eq            4f
+    do_load         2
+4:
+    tst             N, #1
+    b.eq            5f
+    do_load         1
+5:
+    do_yuv_to_rgb
+    tst             N, #4
+    b.eq            6f
+    do_store        \bpp, 4, \fast_st3
+6:
+    tst             N, #2
+    b.eq            7f
+    do_store        \bpp, 2, \fast_st3
+7:
+    tst             N, #1
+    b.eq            8f
+    do_store        \bpp, 1, \fast_st3
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    sub             sp, sp, #336
+    ldr             x15, [sp], 16
+    ld1             {v0.8b, v1.8b, v2.8b, v3.8b}, [sp], 32
+    ld1             {v4.8b, v5.8b, v6.8b, v7.8b}, [sp], 32
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    ld1             {v16.8b, v17.8b, v18.8b, v19.8b}, [sp], 32
+    ld1             {v20.8b, v21.8b, v22.8b, v23.8b}, [sp], 32
+    ld1             {v24.8b, v25.8b, v26.8b, v27.8b}, [sp], 32
+    ld1             {v28.8b, v29.8b, v30.8b, v31.8b}, [sp], 32
+    /* pop             {r4, r5, r6, r7, r8, r9, r10, pc} */
+    ldp             x4, x5, [sp], 16
+    ldp             x6, x7, [sp], 16
+    ldp             x8, x9, [sp], 16
+    ldp             x10, x30, [sp], 16
+    br              x30
+    .unreq          OUTPUT_WIDTH
+    .unreq          INPUT_ROW
+    .unreq          OUTPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          INPUT_BUF0
+    .unreq          INPUT_BUF1
+    .unreq          INPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_yuv_to_rgb
+.purgem do_yuv_to_rgb_stage1
+.purgem do_yuv_to_rgb_stage2
+.purgem do_yuv_to_rgb_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  rsize G  gsize B  bsize defsize fast_st3*/
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extrgbx, 32, 0, .4h,  1, .4h,  2, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, .4h,  1, .4h,  0, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, .4h,  2, .4h,  1, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, .4h,  2, .4h,  3, .4h,  .8b,    1
+generate_jsimd_ycc_rgb_convert_neon rgb565,  16, 0, .4h,  0, .4h,  0, .4h,  .8b,    1
+
+generate_jsimd_ycc_rgb_convert_neon extrgb,  24, 0, .4h,  1, .4h,  2, .4h,  .8b,    0
+generate_jsimd_ycc_rgb_convert_neon extbgr,  24, 2, .4h,  1, .4h,  0, .4h,  .8b,    0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_extrgb_ycc_convert_neon
+ * jsimd_extbgr_ycc_convert_neon
+ * jsimd_extrgbx_ycc_convert_neon
+ * jsimd_extbgrx_ycc_convert_neon
+ * jsimd_extxbgr_ycc_convert_neon
+ * jsimd_extxrgb_ycc_convert_neon
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro do_store size
+  .if \size == 8
+    st1             {v20.8b}, [Y], #8
+    st1             {v21.8b}, [U], #8
+    st1             {v22.8b}, [V], #8
+  .elseif \size == 4
+    st1             {v20.b}[0], [Y], #1
+    st1             {v20.b}[1], [Y], #1
+    st1             {v20.b}[2], [Y], #1
+    st1             {v20.b}[3], [Y], #1
+    st1             {v21.b}[0], [U], #1
+    st1             {v21.b}[1], [U], #1
+    st1             {v21.b}[2], [U], #1
+    st1             {v21.b}[3], [U], #1
+    st1             {v22.b}[0], [V], #1
+    st1             {v22.b}[1], [V], #1
+    st1             {v22.b}[2], [V], #1
+    st1             {v22.b}[3], [V], #1
+  .elseif \size == 2
+    st1             {v20.b}[4], [Y], #1
+    st1             {v20.b}[5], [Y], #1
+    st1             {v21.b}[4], [U], #1
+    st1             {v21.b}[5], [U], #1
+    st1             {v22.b}[4], [V], #1
+    st1             {v22.b}[5], [V], #1
+  .elseif \size == 1
+    st1             {v20.b}[6], [Y], #1
+    st1             {v21.b}[6], [U], #1
+    st1             {v22.b}[6], [V], #1
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size, fast_ld3
+  .if \bpp == 24
+    .if \size == 8
+      .if \fast_ld3 == 1
+        ld3         {v10.8b, v11.8b, v12.8b}, [RGB], #24
+      .else
+        ld1         {v10.b}[0], [RGB], #1
+        ld1         {v11.b}[0], [RGB], #1
+        ld1         {v12.b}[0], [RGB], #1
+
+        ld1         {v10.b}[1], [RGB], #1
+        ld1         {v11.b}[1], [RGB], #1
+        ld1         {v12.b}[1], [RGB], #1
+
+        ld1         {v10.b}[2], [RGB], #1
+        ld1         {v11.b}[2], [RGB], #1
+        ld1         {v12.b}[2], [RGB], #1
+
+        ld1         {v10.b}[3], [RGB], #1
+        ld1         {v11.b}[3], [RGB], #1
+        ld1         {v12.b}[3], [RGB], #1
+
+        ld1         {v10.b}[4], [RGB], #1
+        ld1         {v11.b}[4], [RGB], #1
+        ld1         {v12.b}[4], [RGB], #1
+
+        ld1         {v10.b}[5], [RGB], #1
+        ld1         {v11.b}[5], [RGB], #1
+        ld1         {v12.b}[5], [RGB], #1
+
+        ld1         {v10.b}[6], [RGB], #1
+        ld1         {v11.b}[6], [RGB], #1
+        ld1         {v12.b}[6], [RGB], #1
+
+        ld1         {v10.b}[7], [RGB], #1
+        ld1         {v11.b}[7], [RGB], #1
+        ld1         {v12.b}[7], [RGB], #1
+      .endif
+      prfm          pldl1keep, [RGB, #128]
+    .elseif \size == 4
+      ld3           {v10.b, v11.b, v12.b}[0], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[1], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[2], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[3], [RGB], #3
+    .elseif \size == 2
+      ld3           {v10.b, v11.b, v12.b}[4], [RGB], #3
+      ld3           {v10.b, v11.b, v12.b}[5], [RGB], #3
+    .elseif \size == 1
+      ld3           {v10.b, v11.b, v12.b}[6], [RGB], #3
+    .else
+      .error unsupported macroblock size
+    .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      ld4           {v10.8b, v11.8b, v12.8b, v13.8b}, [RGB], #32
+      prfm          pldl1keep, [RGB, #128]
+    .elseif \size == 4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[0], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[1], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[2], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[3], [RGB], #4
+    .elseif \size == 2
+      ld4           {v10.b, v11.b, v12.b, v13.b}[4], [RGB], #4
+      ld4           {v10.b, v11.b, v12.b, v13.b}[5], [RGB], #4
+    .elseif \size == 1
+      ld4           {v10.b, v11.b, v12.b, v13.b}[6], [RGB], #4
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
+.endm
+
+.macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, \
+                                           b_offs, fast_ld3
+
+/*
+ * 2-stage pipelined RGB->YCbCr conversion
+ */
+
+.macro do_rgb_to_yuv_stage1
+    ushll           v4.8h, v1\r_offs\().8b, #0  /* r = v4 */
+    ushll           v6.8h, v1\g_offs\().8b, #0  /* g = v6 */
+    ushll           v8.8h, v1\b_offs\().8b, #0  /* b = v8 */
+    rev64           v18.4s, v1.4s
+    rev64           v26.4s, v1.4s
+    rev64           v28.4s, v1.4s
+    rev64           v30.4s, v1.4s
+    umull           v14.4s, v4.4h, v0.h[0]
+    umull2          v16.4s, v4.8h, v0.h[0]
+    umlsl           v18.4s, v4.4h, v0.h[3]
+    umlsl2          v26.4s, v4.8h, v0.h[3]
+    umlal           v28.4s, v4.4h, v0.h[5]
+    umlal2          v30.4s, v4.8h, v0.h[5]
+    umlal           v14.4s, v6.4h, v0.h[1]
+    umlal2          v16.4s, v6.8h, v0.h[1]
+    umlsl           v18.4s, v6.4h, v0.h[4]
+    umlsl2          v26.4s, v6.8h, v0.h[4]
+    umlsl           v28.4s, v6.4h, v0.h[6]
+    umlsl2          v30.4s, v6.8h, v0.h[6]
+    umlal           v14.4s, v8.4h, v0.h[2]
+    umlal2          v16.4s, v8.8h, v0.h[2]
+    umlal           v18.4s, v8.4h, v0.h[5]
+    umlal2          v26.4s, v8.8h, v0.h[5]
+    umlsl           v28.4s, v8.4h, v0.h[7]
+    umlsl2          v30.4s, v8.8h, v0.h[7]
+.endm
+
+.macro do_rgb_to_yuv_stage2
+    rshrn           v20.4h, v14.4s, #16
+    shrn            v22.4h, v18.4s, #16
+    shrn            v24.4h, v28.4s, #16
+    rshrn2          v20.8h, v16.4s, #16
+    shrn2           v22.8h, v26.4s, #16
+    shrn2           v24.8h, v30.4s, #16
+    xtn             v20.8b, v20.8h       /* v20 = y */
+    xtn             v21.8b, v22.8h       /* v21 = u */
+    xtn             v22.8b, v24.8h       /* v22 = v */
+.endm
+
+.macro do_rgb_to_yuv
+    do_rgb_to_yuv_stage1
+    do_rgb_to_yuv_stage2
+.endm
+
+/* TODO: expand macros and interleave instructions if some in-order
+ *       ARM64 processor actually can dual-issue LOAD/STORE with ALU */
+.macro do_rgb_to_yuv_stage2_store_load_stage1 fast_ld3
+    do_rgb_to_yuv_stage2
+    do_load         \bpp, 8, \fast_ld3
+    st1             {v20.8b}, [Y], #8
+    st1             {v21.8b}, [U], #8
+    st1             {v22.8b}, [V], #8
+    do_rgb_to_yuv_stage1
+.endm
+
+.balign 16
+.if \fast_ld3 == 1
+Ljsimd_\colorid\()_ycc_neon_consts:
+.else
+Ljsimd_\colorid\()_ycc_neon_slowld3_consts:
+.endif
+  .short 19595, 38470, 7471, 11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128, 32767, 128
+  .short 32767, 128, 32767, 128
+
+.if \fast_ld3 == 1
+asm_function jsimd_\colorid\()_ycc_convert_neon
+.else
+asm_function jsimd_\colorid\()_ycc_convert_neon_slowld3
+.endif
+    OUTPUT_WIDTH    .req w0
+    INPUT_BUF       .req x1
+    OUTPUT_BUF      .req x2
+    OUTPUT_ROW      .req x3
+    NUM_ROWS        .req x4
+
+    OUTPUT_BUF0     .req x5
+    OUTPUT_BUF1     .req x6
+    OUTPUT_BUF2     .req x2  /* OUTPUT_BUF */
+
+    RGB             .req x7
+    Y               .req x9
+    U               .req x10
+    V               .req x11
+    N               .req w12
+
+    /* Load constants to d0, d1, d2, d3 */
+    .if \fast_ld3 == 1
+      adr           x13, Ljsimd_\colorid\()_ycc_neon_consts
+    .else
+      adr           x13, Ljsimd_\colorid\()_ycc_neon_slowld3_consts
+    .endif
+    ld1             {v0.8h, v1.8h}, [x13]
+
+    ldr             OUTPUT_BUF0, [OUTPUT_BUF]
+    ldr             OUTPUT_BUF1, [OUTPUT_BUF, #8]
+    ldr             OUTPUT_BUF2, [OUTPUT_BUF, #16]
+    .unreq          OUTPUT_BUF
+
+    /* Save NEON registers */
+    sub             sp, sp, #64
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    /* Outer loop over scanlines */
+    cmp             NUM_ROWS, #1
+    b.lt            9f
+0:
+    ldr             Y, [OUTPUT_BUF0, OUTPUT_ROW, lsl #3]
+    ldr             U, [OUTPUT_BUF1, OUTPUT_ROW, lsl #3]
+    mov             N, OUTPUT_WIDTH
+    ldr             V, [OUTPUT_BUF2, OUTPUT_ROW, lsl #3]
+    add             OUTPUT_ROW, OUTPUT_ROW, #1
+    ldr             RGB, [INPUT_BUF], #8
+
+    /* Inner loop over pixels */
+    subs            N, N, #8
+    b.lt            3f
+    do_load         \bpp, 8, \fast_ld3
+    do_rgb_to_yuv_stage1
+    subs            N, N, #8
+    b.lt            2f
+1:
+    do_rgb_to_yuv_stage2_store_load_stage1 \fast_ld3
+    subs            N, N, #8
+    b.ge            1b
+2:
+    do_rgb_to_yuv_stage2
+    do_store        8
+    tst             N, #7
+    b.eq            8f
+3:
+    tbz             N, #2, 3f
+    do_load         \bpp, 4, \fast_ld3
+3:
+    tbz             N, #1, 4f
+    do_load         \bpp, 2, \fast_ld3
+4:
+    tbz             N, #0, 5f
+    do_load         \bpp, 1, \fast_ld3
+5:
+    do_rgb_to_yuv
+    tbz             N, #2, 6f
+    do_store        4
+6:
+    tbz             N, #1, 7f
+    do_store        2
+7:
+    tbz             N, #0, 8f
+    do_store        1
+8:
+    subs            NUM_ROWS, NUM_ROWS, #1
+    b.gt            0b
+9:
+    /* Restore all registers and return */
+    sub             sp, sp, #64
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+    br              x30
+
+    .unreq          OUTPUT_WIDTH
+    .unreq          OUTPUT_ROW
+    .unreq          INPUT_BUF
+    .unreq          NUM_ROWS
+    .unreq          OUTPUT_BUF0
+    .unreq          OUTPUT_BUF1
+    .unreq          OUTPUT_BUF2
+    .unreq          RGB
+    .unreq          Y
+    .unreq          U
+    .unreq          V
+    .unreq          N
+
+.purgem do_rgb_to_yuv
+.purgem do_rgb_to_yuv_stage1
+.purgem do_rgb_to_yuv_stage2
+.purgem do_rgb_to_yuv_stage2_store_load_stage1
+
+.endm
+
+/*--------------------------------- id ----- bpp R  G  B  Fast LD3 */
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extrgbx, 32, 0, 1, 2, 1
+generate_jsimd_rgb_ycc_convert_neon extbgrx, 32, 2, 1, 0, 1
+generate_jsimd_rgb_ycc_convert_neon extxbgr, 32, 3, 2, 1, 1
+generate_jsimd_rgb_ycc_convert_neon extxrgb, 32, 1, 2, 3, 1
+
+generate_jsimd_rgb_ycc_convert_neon extrgb,  24, 0, 1, 2, 0
+generate_jsimd_rgb_ycc_convert_neon extbgr,  24, 2, 1, 0, 0
+
+.purgem do_load
+.purgem do_store
+
+
+/*****************************************************************************/
+
+/*
+ * Load data into workspace, applying unsigned->signed conversion
+ *
+ * TODO: can be combined with 'jsimd_fdct_ifast_neon' to get
+ *       rid of VST1.16 instructions
+ */
+
+asm_function jsimd_convsamp_neon
+    SAMPLE_DATA     .req x0
+    START_COL       .req x1
+    WORKSPACE       .req x2
+    TMP1            .req x9
+    TMP2            .req x10
+    TMP3            .req x11
+    TMP4            .req x12
+    TMP5            .req x13
+    TMP6            .req x14
+    TMP7            .req x15
+    TMP8            .req x4
+    TMPDUP          .req w3
+
+    mov             TMPDUP, #128
+    ldp             TMP1, TMP2, [SAMPLE_DATA], 16
+    ldp             TMP3, TMP4, [SAMPLE_DATA], 16
+    dup             v0.8b, TMPDUP
+    add             TMP1, TMP1, START_COL
+    add             TMP2, TMP2, START_COL
+    ldp             TMP5, TMP6, [SAMPLE_DATA], 16
+    add             TMP3, TMP3, START_COL
+    add             TMP4, TMP4, START_COL
+    ldp             TMP7, TMP8, [SAMPLE_DATA], 16
+    add             TMP5, TMP5, START_COL
+    add             TMP6, TMP6, START_COL
+    ld1             {v16.8b}, [TMP1]
+    add             TMP7, TMP7, START_COL
+    add             TMP8, TMP8, START_COL
+    ld1             {v17.8b}, [TMP2]
+    usubl           v16.8h, v16.8b, v0.8b
+    ld1             {v18.8b}, [TMP3]
+    usubl           v17.8h, v17.8b, v0.8b
+    ld1             {v19.8b}, [TMP4]
+    usubl           v18.8h, v18.8b, v0.8b
+    ld1             {v20.8b}, [TMP5]
+    usubl           v19.8h, v19.8b, v0.8b
+    ld1             {v21.8b}, [TMP6]
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [WORKSPACE], 64
+    usubl           v20.8h, v20.8b, v0.8b
+    ld1             {v22.8b}, [TMP7]
+    usubl           v21.8h, v21.8b, v0.8b
+    ld1             {v23.8b}, [TMP8]
+    usubl           v22.8h, v22.8b, v0.8b
+    usubl           v23.8h, v23.8b, v0.8b
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [WORKSPACE], 64
+
+    br              x30
+
+    .unreq          SAMPLE_DATA
+    .unreq          START_COL
+    .unreq          WORKSPACE
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMP4
+    .unreq          TMP5
+    .unreq          TMP6
+    .unreq          TMP7
+    .unreq          TMP8
+    .unreq          TMPDUP
+
+/*****************************************************************************/
+
+/*
+ * jsimd_fdct_islow_neon
+ *
+ * This file contains a slow-but-accurate integer implementation of the
+ * forward DCT (Discrete Cosine Transform). The following code is based
+ * directly on the IJG''s original jfdctint.c; see the jfdctint.c for
+ * more details.
+ *
+ * TODO: can be combined with 'jsimd_convsamp_neon' to get
+ *       rid of a bunch of VLD1.16 instructions
+ */
+
+#define CONST_BITS 13
+#define PASS1_BITS 2
+
+#define DESCALE_P1 (CONST_BITS-PASS1_BITS)
+#define DESCALE_P2 (CONST_BITS+PASS1_BITS)
+
+#define F_0_298  2446  /* FIX(0.298631336) */
+#define F_0_390  3196  /* FIX(0.390180644) */
+#define F_0_541  4433  /* FIX(0.541196100) */
+#define F_0_765  6270  /* FIX(0.765366865) */
+#define F_0_899  7373  /* FIX(0.899976223) */
+#define F_1_175  9633  /* FIX(1.175875602) */
+#define F_1_501 12299  /* FIX(1.501321110) */
+#define F_1_847 15137  /* FIX(1.847759065) */
+#define F_1_961 16069  /* FIX(1.961570560) */
+#define F_2_053 16819  /* FIX(2.053119869) */
+#define F_2_562 20995  /* FIX(2.562915447) */
+#define F_3_072 25172  /* FIX(3.072711026) */
+
+.balign 16
+Ljsimd_fdct_islow_neon_consts:
+  .short F_0_298
+  .short -F_0_390
+  .short F_0_541
+  .short F_0_765
+  .short - F_0_899
+  .short F_1_175
+  .short F_1_501
+  .short - F_1_847
+  .short - F_1_961
+  .short F_2_053
+  .short - F_2_562
+  .short F_3_072
+  .short 0          /* padding */
+  .short 0
+  .short 0
+  .short 0
+
+#undef F_0_298
+#undef F_0_390
+#undef F_0_541
+#undef F_0_765
+#undef F_0_899
+#undef F_1_175
+#undef F_1_501
+#undef F_1_847
+#undef F_1_961
+#undef F_2_053
+#undef F_2_562
+#undef F_3_072
+#define XFIX_P_0_298 v0.h[0]
+#define XFIX_N_0_390 v0.h[1]
+#define XFIX_P_0_541 v0.h[2]
+#define XFIX_P_0_765 v0.h[3]
+#define XFIX_N_0_899 v0.h[4]
+#define XFIX_P_1_175 v0.h[5]
+#define XFIX_P_1_501 v0.h[6]
+#define XFIX_N_1_847 v0.h[7]
+#define XFIX_N_1_961 v1.h[0]
+#define XFIX_P_2_053 v1.h[1]
+#define XFIX_N_2_562 v1.h[2]
+#define XFIX_P_3_072 v1.h[3]
+
+asm_function jsimd_fdct_islow_neon
+
+    DATA            .req x0
+    TMP             .req x9
+
+    /* Load constants */
+    adr             TMP, Ljsimd_fdct_islow_neon_consts
+    ld1             {v0.8h, v1.8h}, [TMP]
+
+    /* Save NEON registers */
+    sub             sp, sp, #64
+    st1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    st1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    /* Load all DATA into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17    | v16.8h
+     *   1 | d18     | d19    | v17.8h
+     *   2 | d20     | d21    | v18.8h
+     *   3 | d22     | d23    | v19.8h
+     *   4 | d24     | d25    | v20.8h
+     *   5 | d26     | d27    | v21.8h
+     *   6 | d28     | d29    | v22.8h
+     *   7 | d30     | d31    | v23.8h
+     */
+
+    ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+    sub             DATA, DATA, #64
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr[0] + dataptr[7]; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr[0] - dataptr[7]; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr[1] + dataptr[6]; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr[1] - dataptr[6]; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr[2] + dataptr[5]; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr[2] - dataptr[5]; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr[3] + dataptr[4]; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr[3] - dataptr[4]; */
+
+    /* even part */
+
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    shl             v16.8h, v16.8h, #PASS1_BITS  /* dataptr[0] = (DCTELEM) LEFT_SHIFT(tmp10 + tmp11, PASS1_BITS); */
+    shl             v20.8h, v20.8h, #PASS1_BITS  /* dataptr[4] = (DCTELEM) LEFT_SHIFT(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P1
+    rshrn           v22.4h, v22.4s, #DESCALE_P1
+    rshrn2          v18.8h, v24.4s, #DESCALE_P1  /* dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P1  /* dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+
+    add             v8.8h, v28.8h, v31.8h        /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h        /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h       /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h       /* z4 = tmp5 + tmp7; */
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, - FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, - FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s  /* z3 += z5 */
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s  /* z4 += z5 */
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P1
+    rshrn           v21.4h, v29.4s, #DESCALE_P1
+    rshrn           v19.4h, v30.4s, #DESCALE_P1
+    rshrn           v17.4h, v31.4s, #DESCALE_P1
+    rshrn2          v23.8h, v24.4s, #DESCALE_P1  /* dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P1  /* dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P1  /* dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P1  /* dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v31, v2, v3, v4
+
+    /* 1-D FDCT */
+    add             v24.8h, v16.8h, v23.8h  /* tmp0 = dataptr[0] + dataptr[7]; */
+    sub             v31.8h, v16.8h, v23.8h  /* tmp7 = dataptr[0] - dataptr[7]; */
+    add             v25.8h, v17.8h, v22.8h  /* tmp1 = dataptr[1] + dataptr[6]; */
+    sub             v30.8h, v17.8h, v22.8h  /* tmp6 = dataptr[1] - dataptr[6]; */
+    add             v26.8h, v18.8h, v21.8h  /* tmp2 = dataptr[2] + dataptr[5]; */
+    sub             v29.8h, v18.8h, v21.8h  /* tmp5 = dataptr[2] - dataptr[5]; */
+    add             v27.8h, v19.8h, v20.8h  /* tmp3 = dataptr[3] + dataptr[4]; */
+    sub             v28.8h, v19.8h, v20.8h  /* tmp4 = dataptr[3] - dataptr[4]; */
+
+    /* even part */
+    add             v8.8h, v24.8h, v27.8h   /* tmp10 = tmp0 + tmp3; */
+    sub             v9.8h, v24.8h, v27.8h   /* tmp13 = tmp0 - tmp3; */
+    add             v10.8h, v25.8h, v26.8h  /* tmp11 = tmp1 + tmp2; */
+    sub             v11.8h, v25.8h, v26.8h  /* tmp12 = tmp1 - tmp2; */
+
+    add             v16.8h, v8.8h, v10.8h  /* tmp10 + tmp11 */
+    sub             v20.8h, v8.8h, v10.8h  /* tmp10 - tmp11 */
+
+    add             v18.8h, v11.8h, v9.8h  /* tmp12 + tmp13 */
+
+    srshr           v16.8h, v16.8h, #PASS1_BITS  /* dataptr[0] = (DCTELEM) DESCALE(tmp10 + tmp11, PASS1_BITS); */
+    srshr           v20.8h, v20.8h, #PASS1_BITS  /* dataptr[4] = (DCTELEM) DESCALE(tmp10 - tmp11, PASS1_BITS); */
+
+    smull2          v24.4s, v18.8h, XFIX_P_0_541  /* z1 hi = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    smull           v18.4s, v18.4h, XFIX_P_0_541  /* z1 lo = MULTIPLY(tmp12 + tmp13, XFIX_P_0_541); */
+    mov             v22.16b, v18.16b
+    mov             v25.16b, v24.16b
+
+    smlal           v18.4s, v9.4h, XFIX_P_0_765   /* lo z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal2          v24.4s, v9.8h, XFIX_P_0_765   /* hi z1 + MULTIPLY(tmp13, XFIX_P_0_765) */
+    smlal           v22.4s, v11.4h, XFIX_N_1_847  /* lo z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+    smlal2          v25.4s, v11.8h, XFIX_N_1_847  /* hi z1 + MULTIPLY(tmp12, XFIX_N_1_847) */
+
+    rshrn           v18.4h, v18.4s, #DESCALE_P2
+    rshrn           v22.4h, v22.4s, #DESCALE_P2
+    rshrn2          v18.8h, v24.4s, #DESCALE_P2  /* dataptr[2] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp13, XFIX_P_0_765), CONST_BITS-PASS1_BITS); */
+    rshrn2          v22.8h, v25.4s, #DESCALE_P2  /* dataptr[6] = (DCTELEM) DESCALE(z1 + MULTIPLY(tmp12, XFIX_N_1_847), CONST_BITS-PASS1_BITS); */
+
+    /* Odd part */
+    add             v8.8h, v28.8h, v31.8h   /* z1 = tmp4 + tmp7; */
+    add             v9.8h, v29.8h, v30.8h   /* z2 = tmp5 + tmp6; */
+    add             v10.8h, v28.8h, v30.8h  /* z3 = tmp4 + tmp6; */
+    add             v11.8h, v29.8h, v31.8h  /* z4 = tmp5 + tmp7; */
+
+    smull           v4.4s, v10.4h, XFIX_P_1_175  /* z5 lo = z3 lo * XFIX_P_1_175 */
+    smull2          v5.4s, v10.8h, XFIX_P_1_175
+    smlal           v4.4s, v11.4h, XFIX_P_1_175  /* z5 = MULTIPLY(z3 + z4, FIX_1_175875602); */
+    smlal2          v5.4s, v11.8h, XFIX_P_1_175
+
+    smull2          v24.4s, v28.8h, XFIX_P_0_298
+    smull2          v25.4s, v29.8h, XFIX_P_2_053
+    smull2          v26.4s, v30.8h, XFIX_P_3_072
+    smull2          v27.4s, v31.8h, XFIX_P_1_501
+    smull           v28.4s, v28.4h, XFIX_P_0_298  /* tmp4 = MULTIPLY(tmp4, FIX_0_298631336); */
+    smull           v29.4s, v29.4h, XFIX_P_2_053  /* tmp5 = MULTIPLY(tmp5, FIX_2_053119869); */
+    smull           v30.4s, v30.4h, XFIX_P_3_072  /* tmp6 = MULTIPLY(tmp6, FIX_3_072711026); */
+    smull           v31.4s, v31.4h, XFIX_P_1_501  /* tmp7 = MULTIPLY(tmp7, FIX_1_501321110); */
+
+    smull2          v12.4s, v8.8h, XFIX_N_0_899
+    smull2          v13.4s, v9.8h, XFIX_N_2_562
+    smull2          v14.4s, v10.8h, XFIX_N_1_961
+    smull2          v15.4s, v11.8h, XFIX_N_0_390
+    smull           v8.4s, v8.4h, XFIX_N_0_899    /* z1 = MULTIPLY(z1, - FIX_0_899976223); */
+    smull           v9.4s, v9.4h, XFIX_N_2_562    /* z2 = MULTIPLY(z2, - FIX_2_562915447); */
+    smull           v10.4s, v10.4h, XFIX_N_1_961  /* z3 = MULTIPLY(z3, - FIX_1_961570560); */
+    smull           v11.4s, v11.4h, XFIX_N_0_390  /* z4 = MULTIPLY(z4, - FIX_0_390180644); */
+
+    add             v10.4s, v10.4s, v4.4s
+    add             v14.4s, v14.4s, v5.4s
+    add             v11.4s, v11.4s, v4.4s
+    add             v15.4s, v15.4s, v5.4s
+
+    add             v28.4s, v28.4s, v8.4s   /* tmp4 += z1 */
+    add             v24.4s, v24.4s, v12.4s
+    add             v29.4s, v29.4s, v9.4s   /* tmp5 += z2 */
+    add             v25.4s, v25.4s, v13.4s
+    add             v30.4s, v30.4s, v10.4s  /* tmp6 += z3 */
+    add             v26.4s, v26.4s, v14.4s
+    add             v31.4s, v31.4s, v11.4s  /* tmp7 += z4 */
+    add             v27.4s, v27.4s, v15.4s
+
+    add             v28.4s, v28.4s, v10.4s  /* tmp4 += z3 */
+    add             v24.4s, v24.4s, v14.4s
+    add             v29.4s, v29.4s, v11.4s  /* tmp5 += z4 */
+    add             v25.4s, v25.4s, v15.4s
+    add             v30.4s, v30.4s, v9.4s   /* tmp6 += z2 */
+    add             v26.4s, v26.4s, v13.4s
+    add             v31.4s, v31.4s, v8.4s   /* tmp7 += z1 */
+    add             v27.4s, v27.4s, v12.4s
+
+    rshrn           v23.4h, v28.4s, #DESCALE_P2
+    rshrn           v21.4h, v29.4s, #DESCALE_P2
+    rshrn           v19.4h, v30.4s, #DESCALE_P2
+    rshrn           v17.4h, v31.4s, #DESCALE_P2
+    rshrn2          v23.8h, v24.4s, #DESCALE_P2  /* dataptr[7] = (DCTELEM) DESCALE(tmp4 + z1 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v21.8h, v25.4s, #DESCALE_P2  /* dataptr[5] = (DCTELEM) DESCALE(tmp5 + z2 + z4, CONST_BITS-PASS1_BITS); */
+    rshrn2          v19.8h, v26.4s, #DESCALE_P2  /* dataptr[3] = (DCTELEM) DESCALE(tmp6 + z2 + z3, CONST_BITS-PASS1_BITS); */
+    rshrn2          v17.8h, v27.4s, #DESCALE_P2  /* dataptr[1] = (DCTELEM) DESCALE(tmp7 + z1 + z4, CONST_BITS-PASS1_BITS); */
+
+    /* store results */
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+
+    /* Restore NEON registers */
+    sub             sp, sp, #64
+    ld1             {v8.8b, v9.8b, v10.8b, v11.8b}, [sp], 32
+    ld1             {v12.8b, v13.8b, v14.8b, v15.8b}, [sp], 32
+
+    br              x30
+
+    .unreq          DATA
+    .unreq          TMP
+
+#undef XFIX_P_0_298
+#undef XFIX_N_0_390
+#undef XFIX_P_0_541
+#undef XFIX_P_0_765
+#undef XFIX_N_0_899
+#undef XFIX_P_1_175
+#undef XFIX_P_1_501
+#undef XFIX_N_1_847
+#undef XFIX_N_1_961
+#undef XFIX_P_2_053
+#undef XFIX_N_2_562
+#undef XFIX_P_3_072
+
+
+/*****************************************************************************/
+
+/*
+ * jsimd_fdct_ifast_neon
+ *
+ * This function contains a fast, not so accurate integer implementation of
+ * the forward DCT (Discrete Cosine Transform). It uses the same calculations
+ * and produces exactly the same output as IJG's original 'jpeg_fdct_ifast'
+ * function from jfdctfst.c
+ *
+ * TODO: can be combined with 'jsimd_convsamp_neon' to get
+ *       rid of a bunch of VLD1.16 instructions
+ */
+
+#undef XFIX_0_541196100
+#define XFIX_0_382683433 v0.h[0]
+#define XFIX_0_541196100 v0.h[1]
+#define XFIX_0_707106781 v0.h[2]
+#define XFIX_1_306562965 v0.h[3]
+
+.balign 16
+Ljsimd_fdct_ifast_neon_consts:
+  .short (98 * 128)               /* XFIX_0_382683433 */
+  .short (139 * 128)              /* XFIX_0_541196100 */
+  .short (181 * 128)              /* XFIX_0_707106781 */
+  .short (334 * 128 - 256 * 128)  /* XFIX_1_306562965 */
+
+asm_function jsimd_fdct_ifast_neon
+
+    DATA            .req x0
+    TMP             .req x9
+
+    /* Load constants */
+    adr             TMP, Ljsimd_fdct_ifast_neon_consts
+    ld1             {v0.4h}, [TMP]
+
+    /* Load all DATA into NEON registers with the following allocation:
+     *       0 1 2 3 | 4 5 6 7
+     *      ---------+--------
+     *   0 | d16     | d17    | v0.8h
+     *   1 | d18     | d19    | q9
+     *   2 | d20     | d21    | q10
+     *   3 | d22     | d23    | q11
+     *   4 | d24     | d25    | q12
+     *   5 | d26     | d27    | q13
+     *   6 | d28     | d29    | q14
+     *   7 | d30     | d31    | q15
+     */
+
+    ld1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    ld1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+    mov             TMP, #2
+    sub             DATA, DATA, #64
+1:
+    /* Transpose */
+    transpose_8x8   v16, v17, v18, v19, v20, v21, v22, v23, v1, v2, v3, v4
+    subs            TMP, TMP, #1
+    /* 1-D FDCT */
+    add             v4.8h, v19.8h, v20.8h
+    sub             v20.8h, v19.8h, v20.8h
+    sub             v28.8h, v18.8h, v21.8h
+    add             v18.8h, v18.8h, v21.8h
+    sub             v29.8h, v17.8h, v22.8h
+    add             v17.8h, v17.8h, v22.8h
+    sub             v21.8h, v16.8h, v23.8h
+    add             v16.8h, v16.8h, v23.8h
+    sub             v6.8h, v17.8h, v18.8h
+    sub             v7.8h, v16.8h, v4.8h
+    add             v5.8h, v17.8h, v18.8h
+    add             v6.8h, v6.8h, v7.8h
+    add             v4.8h, v16.8h, v4.8h
+    sqdmulh         v6.8h, v6.8h, XFIX_0_707106781
+    add             v19.8h, v20.8h, v28.8h
+    add             v16.8h, v4.8h, v5.8h
+    sub             v20.8h, v4.8h, v5.8h
+    add             v5.8h, v28.8h, v29.8h
+    add             v29.8h, v29.8h, v21.8h
+    sqdmulh         v5.8h, v5.8h, XFIX_0_707106781
+    sub             v28.8h, v19.8h, v29.8h
+    add             v18.8h, v7.8h, v6.8h
+    sqdmulh         v28.8h, v28.8h, XFIX_0_382683433
+    sub             v22.8h, v7.8h, v6.8h
+    sqdmulh         v19.8h, v19.8h, XFIX_0_541196100
+    sqdmulh         v7.8h, v29.8h, XFIX_1_306562965
+    add             v6.8h, v21.8h, v5.8h
+    sub             v5.8h, v21.8h, v5.8h
+    add             v29.8h, v29.8h, v28.8h
+    add             v19.8h, v19.8h, v28.8h
+    add             v29.8h, v29.8h, v7.8h
+    add             v21.8h, v5.8h, v19.8h
+    sub             v19.8h, v5.8h, v19.8h
+    add             v17.8h, v6.8h, v29.8h
+    sub             v23.8h, v6.8h, v29.8h
+
+    b.ne            1b
+
+    /* store results */
+    st1             {v16.8h, v17.8h, v18.8h, v19.8h}, [DATA], 64
+    st1             {v20.8h, v21.8h, v22.8h, v23.8h}, [DATA]
+
+    br              x30
+
+    .unreq          DATA
+    .unreq          TMP
+#undef XFIX_0_382683433
+#undef XFIX_0_541196100
+#undef XFIX_0_707106781
+#undef XFIX_1_306562965
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(void)
+ * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM *divisors,
+ *                      DCTELEM *workspace);
+ *
+ */
+asm_function jsimd_quantize_neon
+
+    COEF_BLOCK      .req x0
+    DIVISORS        .req x1
+    WORKSPACE       .req x2
+
+    RECIPROCAL      .req DIVISORS
+    CORRECTION      .req x9
+    SHIFT           .req x10
+    LOOP_COUNT      .req x11
+
+    mov             LOOP_COUNT, #2
+    add             CORRECTION, DIVISORS, #(64 * 2)
+    add             SHIFT, DIVISORS, #(64 * 6)
+1:
+    subs            LOOP_COUNT, LOOP_COUNT, #1
+    ld1             {v0.8h, v1.8h, v2.8h, v3.8h}, [WORKSPACE], 64
+    ld1             {v4.8h, v5.8h, v6.8h, v7.8h}, [CORRECTION], 64
+    abs             v20.8h, v0.8h
+    abs             v21.8h, v1.8h
+    abs             v22.8h, v2.8h
+    abs             v23.8h, v3.8h
+    ld1             {v28.8h, v29.8h, v30.8h, v31.8h}, [RECIPROCAL], 64
+    add             v20.8h, v20.8h, v4.8h  /* add correction */
+    add             v21.8h, v21.8h, v5.8h
+    add             v22.8h, v22.8h, v6.8h
+    add             v23.8h, v23.8h, v7.8h
+    umull           v4.4s, v20.4h, v28.4h  /* multiply by reciprocal */
+    umull2          v16.4s, v20.8h, v28.8h
+    umull           v5.4s, v21.4h, v29.4h
+    umull2          v17.4s, v21.8h, v29.8h
+    umull           v6.4s, v22.4h, v30.4h  /* multiply by reciprocal */
+    umull2          v18.4s, v22.8h, v30.8h
+    umull           v7.4s, v23.4h, v31.4h
+    umull2          v19.4s, v23.8h, v31.8h
+    ld1             {v24.8h, v25.8h, v26.8h, v27.8h}, [SHIFT], 64
+    shrn            v4.4h, v4.4s, #16
+    shrn            v5.4h, v5.4s, #16
+    shrn            v6.4h, v6.4s, #16
+    shrn            v7.4h, v7.4s, #16
+    shrn2           v4.8h, v16.4s, #16
+    shrn2           v5.8h, v17.4s, #16
+    shrn2           v6.8h, v18.4s, #16
+    shrn2           v7.8h, v19.4s, #16
+    neg             v24.8h, v24.8h
+    neg             v25.8h, v25.8h
+    neg             v26.8h, v26.8h
+    neg             v27.8h, v27.8h
+    sshr            v0.8h, v0.8h, #15  /* extract sign */
+    sshr            v1.8h, v1.8h, #15
+    sshr            v2.8h, v2.8h, #15
+    sshr            v3.8h, v3.8h, #15
+    ushl            v4.8h, v4.8h, v24.8h  /* shift */
+    ushl            v5.8h, v5.8h, v25.8h
+    ushl            v6.8h, v6.8h, v26.8h
+    ushl            v7.8h, v7.8h, v27.8h
+
+    eor             v4.16b, v4.16b, v0.16b  /* restore sign */
+    eor             v5.16b, v5.16b, v1.16b
+    eor             v6.16b, v6.16b, v2.16b
+    eor             v7.16b, v7.16b, v3.16b
+    sub             v4.8h, v4.8h, v0.8h
+    sub             v5.8h, v5.8h, v1.8h
+    sub             v6.8h, v6.8h, v2.8h
+    sub             v7.8h, v7.8h, v3.8h
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [COEF_BLOCK], 64
+
+    b.ne            1b
+
+    br              x30  /* return */
+
+    .unreq          COEF_BLOCK
+    .unreq          DIVISORS
+    .unreq          WORKSPACE
+    .unreq          RECIPROCAL
+    .unreq          CORRECTION
+    .unreq          SHIFT
+    .unreq          LOOP_COUNT
+
+
+/*****************************************************************************/
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 1:1 vertical,
+ * without smoothing.
+ *
+ * GLOBAL(void)
+ * jsimd_h2v1_downsample_neon (JDIMENSION image_width, int max_v_samp_factor,
+ *                             JDIMENSION v_samp_factor,
+ *                             JDIMENSION width_blocks, JSAMPARRAY input_data,
+ *                             JSAMPARRAY output_data);
+ */
+
+.balign 16
+Ljsimd_h2_downsample_neon_consts:
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F  /* diff 0 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0E  /* diff 1 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0D, 0x0D  /* diff 2 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0C, 0x0C, 0x0C  /* diff 3 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0B, 0x0B, 0x0B, 0x0B, 0x0B  /* diff 4 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A, 0x0A  /* diff 5 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09, 0x09  /* diff 6 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08, 0x08  /* diff 7 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
+        0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07, 0x07  /* diff 8 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x06, \
+        0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06, 0x06  /* diff 9 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x05, 0x05, \
+        0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05, 0x05  /* diff 10 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x04, 0x04, 0x04, 0x04, \
+        0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04, 0x04  /* diff 11 */
+  .byte 0x00, 0x01, 0x02, 0x03, 0x03, 0x03, 0x03, 0x03, \
+        0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03  /* diff 12 */
+  .byte 0x00, 0x01, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, \
+        0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02, 0x02  /* diff 13 */
+  .byte 0x00, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, \
+        0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01  /* diff 14 */
+  .byte 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, \
+        0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00  /* diff 15 */
+
+asm_function jsimd_h2v1_downsample_neon
+    IMAGE_WIDTH     .req x0
+    MAX_V_SAMP      .req x1
+    V_SAMP          .req x2
+    BLOCK_WIDTH     .req x3
+    INPUT_DATA      .req x4
+    OUTPUT_DATA     .req x5
+    OUTPTR          .req x9
+    INPTR           .req x10
+    TMP1            .req x11
+    TMP2            .req x12
+    TMP3            .req x13
+    TMPDUP          .req w15
+
+    mov             TMPDUP, #0x10000
+    lsl             TMP2, BLOCK_WIDTH, #4
+    sub             TMP2, TMP2, IMAGE_WIDTH
+    adr             TMP3, Ljsimd_h2_downsample_neon_consts
+    add             TMP3, TMP3, TMP2, lsl #4
+    dup             v16.4s, TMPDUP
+    ld1             {v18.16b}, [TMP3]
+
+1:  /* row loop */
+    ldr             INPTR, [INPUT_DATA], #8
+    ldr             OUTPTR, [OUTPUT_DATA], #8
+    subs            TMP1, BLOCK_WIDTH, #1
+    b.eq            3f
+2:  /* columns */
+    ld1             {v0.16b}, [INPTR], #16
+    mov             v4.16b, v16.16b
+    subs            TMP1, TMP1, #1
+    uadalp          v4.8h, v0.16b
+    shrn            v6.8b, v4.8h, #1
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            2b
+3:  /* last columns */
+    ld1             {v0.16b}, [INPTR]
+    mov             v4.16b, v16.16b
+    subs            V_SAMP, V_SAMP, #1
+    /* expand right */
+    tbl             v2.16b, {v0.16b}, v18.16b
+    uadalp          v4.8h, v2.16b
+    shrn            v6.8b, v4.8h, #1
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            1b
+
+    br              x30
+
+    .unreq          IMAGE_WIDTH
+    .unreq          MAX_V_SAMP
+    .unreq          V_SAMP
+    .unreq          BLOCK_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA
+    .unreq          OUTPTR
+    .unreq          INPTR
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMPDUP
+
+
+/*****************************************************************************/
+
+/*
+ * Downsample pixel values of a single component.
+ * This version handles the common case of 2:1 horizontal and 2:1 vertical,
+ * without smoothing.
+ *
+ * GLOBAL(void)
+ * jsimd_h2v2_downsample_neon (JDIMENSION image_width, int max_v_samp_factor,
+ *                             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+ *                             JSAMPARRAY input_data, JSAMPARRAY output_data);
+ */
+
+.balign 16
+asm_function jsimd_h2v2_downsample_neon
+    IMAGE_WIDTH     .req x0
+    MAX_V_SAMP      .req x1
+    V_SAMP          .req x2
+    BLOCK_WIDTH     .req x3
+    INPUT_DATA      .req x4
+    OUTPUT_DATA     .req x5
+    OUTPTR          .req x9
+    INPTR0          .req x10
+    INPTR1          .req x14
+    TMP1            .req x11
+    TMP2            .req x12
+    TMP3            .req x13
+    TMPDUP          .req w15
+
+    mov             TMPDUP, #1
+    lsl             TMP2, BLOCK_WIDTH, #4
+    lsl             TMPDUP, TMPDUP, #17
+    sub             TMP2, TMP2, IMAGE_WIDTH
+    adr             TMP3, Ljsimd_h2_downsample_neon_consts
+    orr             TMPDUP, TMPDUP, #1
+    add             TMP3, TMP3, TMP2, lsl #4
+    dup             v16.4s, TMPDUP
+    ld1             {v18.16b}, [TMP3]
+
+1:  /* row loop */
+    ldr             INPTR0, [INPUT_DATA], #8
+    ldr             OUTPTR, [OUTPUT_DATA], #8
+    ldr             INPTR1, [INPUT_DATA], #8
+    subs            TMP1, BLOCK_WIDTH, #1
+    b.eq            3f
+2:  /* columns */
+    ld1             {v0.16b}, [INPTR0], #16
+    ld1             {v1.16b}, [INPTR1], #16
+    mov             v4.16b, v16.16b
+    subs            TMP1, TMP1, #1
+    uadalp          v4.8h, v0.16b
+    uadalp          v4.8h, v1.16b
+    shrn            v6.8b, v4.8h, #2
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            2b
+3:  /* last columns */
+    ld1             {v0.16b}, [INPTR0], #16
+    ld1             {v1.16b}, [INPTR1], #16
+    mov             v4.16b, v16.16b
+    subs            V_SAMP, V_SAMP, #1
+    /* expand right */
+    tbl             v2.16b, {v0.16b}, v18.16b
+    tbl             v3.16b, {v1.16b}, v18.16b
+    uadalp          v4.8h, v2.16b
+    uadalp          v4.8h, v3.16b
+    shrn            v6.8b, v4.8h, #2
+    st1             {v6.8b}, [OUTPTR], #8
+    b.ne            1b
+
+    br              x30
+
+    .unreq          IMAGE_WIDTH
+    .unreq          MAX_V_SAMP
+    .unreq          V_SAMP
+    .unreq          BLOCK_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA
+    .unreq          OUTPTR
+    .unreq          INPTR0
+    .unreq          INPTR1
+    .unreq          TMP1
+    .unreq          TMP2
+    .unreq          TMP3
+    .unreq          TMPDUP
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(JOCTET*)
+ * jsimd_huff_encode_one_block (working_state *state, JOCTET *buffer,
+ *                              JCOEFPTR block, int last_dc_val,
+ *                              c_derived_tbl *dctbl, c_derived_tbl *actbl)
+ *
+ */
+
+    BUFFER          .req x1
+    PUT_BUFFER      .req x6
+    PUT_BITS        .req x7
+    PUT_BITSw       .req w7
+
+.macro emit_byte
+    sub             PUT_BITS, PUT_BITS, #0x8
+    lsr             x19, PUT_BUFFER, PUT_BITS
+    uxtb            w19, w19
+    strb            w19, [BUFFER, #1]!
+    cmp             w19, #0xff
+    b.ne            14f
+    strb            wzr, [BUFFER, #1]!
+14:
+.endm
+.macro put_bits CODE, SIZE
+    lsl             PUT_BUFFER, PUT_BUFFER, \SIZE
+    add             PUT_BITS, PUT_BITS, \SIZE
+    orr             PUT_BUFFER, PUT_BUFFER, \CODE
+.endm
+.macro checkbuf31
+    cmp             PUT_BITS, #0x20
+    b.lt            31f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+31:
+.endm
+.macro checkbuf47
+    cmp             PUT_BITS, #0x30
+    b.lt            47f
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+    emit_byte
+47:
+.endm
+
+.macro generate_jsimd_huff_encode_one_block fast_tbl
+
+.balign 16
+.if \fast_tbl == 1
+Ljsimd_huff_encode_one_block_neon_consts:
+.else
+Ljsimd_huff_encode_one_block_neon_slowtbl_consts:
+.endif
+    .byte 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, \
+          0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80
+.if \fast_tbl == 1
+    .byte    0,   1,   2,   3,  16,  17,  32,  33, \
+            18,  19,   4,   5,   6,   7,  20,  21  /* L0 => L3 : 4 lines OK */
+    .byte   34,  35,  48,  49, 255, 255,  50,  51, \
+            36,  37,  22,  23,   8,   9,  10,  11  /* L0 => L3 : 4 lines OK */
+    .byte    8,   9,  22,  23,  36,  37,  50,  51, \
+           255, 255, 255, 255, 255, 255,  52,  53  /* L1 => L4 : 4 lines OK */
+    .byte   54,  55,  40,  41,  26,  27,  12,  13, \
+            14,  15,  28,  29,  42,  43,  56,  57  /* L0 => L3 : 4 lines OK */
+    .byte    6,   7,  20,  21,  34,  35,  48,  49, \
+            50,  51,  36,  37,  22,  23,   8,   9  /* L4 => L7 : 4 lines OK */
+    .byte   42,  43,  28,  29,  14,  15,  30,  31, \
+            44,  45,  58,  59, 255, 255, 255, 255  /* L1 => L4 : 4 lines OK */
+    .byte  255, 255, 255, 255,  56,  57,  42,  43, \
+            28,  29,  14,  15,  30,  31,  44,  45  /* L3 => L6 : 4 lines OK */
+    .byte   26,  27,  40,  41,  42,  43,  28,  29, \
+            14,  15,  30,  31,  44,  45,  46,  47  /* L5 => L7 : 3 lines OK */
+    .byte  255, 255, 255, 255,   0,   1, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L4 : 1 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+             0,   1,  16,  17,   2,   3, 255, 255  /* L5 => L6 : 2 lines OK */
+    .byte  255, 255, 255, 255, 255, 255, 255, 255, \
+           255, 255, 255, 255,   8,   9,  22,  23  /* L5 => L6 : 2 lines OK */
+    .byte    4,   5,   6,   7, 255, 255, 255, 255, \
+           255, 255, 255, 255, 255, 255, 255, 255  /* L7 : 1 line OK */
+.endif
+
+.if \fast_tbl == 1
+asm_function jsimd_huff_encode_one_block_neon
+.else
+asm_function jsimd_huff_encode_one_block_neon_slowtbl
+.endif
+    sub             sp, sp, 272
+    sub             BUFFER, BUFFER, #0x1    /* BUFFER=buffer-- */
+    /* Save ARM registers */
+    stp             x19, x20, [sp], 16
+.if \fast_tbl == 1
+    adr             x15, Ljsimd_huff_encode_one_block_neon_consts
+.else
+    adr             x15, Ljsimd_huff_encode_one_block_neon_slowtbl_consts
+.endif
+    ldr             PUT_BUFFER, [x0, #0x10]
+    ldr             PUT_BITSw, [x0, #0x18]
+    ldrsh           w12, [x2]               /* load DC coeff in w12 */
+    /* prepare data */
+.if \fast_tbl == 1
+    ld1             {v23.16b}, [x15], #16
+    ld1             {v0.16b, v1.16b, v2.16b, v3.16b}, [x15], #64
+    ld1             {v4.16b, v5.16b, v6.16b, v7.16b}, [x15], #64
+    ld1             {v16.16b, v17.16b, v18.16b, v19.16b}, [x15], #64
+    ld1             {v24.16b, v25.16b, v26.16b, v27.16b}, [x2], #64
+    ld1             {v28.16b, v29.16b, v30.16b, v31.16b}, [x2], #64
+    sub             w12, w12, w3      /* last_dc_val, not used afterwards */
+    /* ZigZag 8x8 */
+    tbl             v0.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v0.16b
+    tbl             v1.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v1.16b
+    tbl             v2.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v2.16b
+    tbl             v3.16b, {v24.16b, v25.16b, v26.16b, v27.16b}, v3.16b
+    tbl             v4.16b, {v28.16b, v29.16b, v30.16b, v31.16b}, v4.16b
+    tbl             v5.16b, {v25.16b, v26.16b, v27.16b, v28.16b}, v5.16b
+    tbl             v6.16b, {v27.16b, v28.16b, v29.16b, v30.16b}, v6.16b
+    tbl             v7.16b, {v29.16b, v30.16b, v31.16b}, v7.16b
+    ins             v0.h[0], w12
+    tbx             v1.16b, {v28.16b}, v16.16b
+    tbx             v2.16b, {v29.16b, v30.16b}, v17.16b
+    tbx             v5.16b, {v29.16b, v30.16b}, v18.16b
+    tbx             v6.16b, {v31.16b}, v19.16b
+.else
+      add             x13, x2, #0x22
+      sub             w12, w12, w3    /* last_dc_val, not used afterwards */
+    ld1             {v23.16b}, [x15]
+      add             x14, x2, #0x18
+      add             x3, x2, #0x36
+    ins             v0.h[0], w12
+      add             x9, x2, #0x2
+    ld1             {v1.h}[0], [x13]
+      add             x15, x2, #0x30
+    ld1             {v2.h}[0], [x14]
+      add             x19, x2, #0x26
+    ld1             {v3.h}[0], [x3]
+      add             x20, x2, #0x28
+    ld1             {v0.h}[1], [x9]
+      add             x12, x2, #0x10
+    ld1             {v1.h}[1], [x15]
+      add             x13, x2, #0x40
+    ld1             {v2.h}[1], [x19]
+      add             x14, x2, #0x34
+    ld1             {v3.h}[1], [x20]
+      add             x3, x2, #0x1a
+    ld1             {v0.h}[2], [x12]
+      add             x9, x2, #0x20
+    ld1             {v1.h}[2], [x13]
+      add             x15, x2, #0x32
+    ld1             {v2.h}[2], [x14]
+      add             x19, x2, #0x42
+    ld1             {v3.h}[2], [x3]
+      add             x20, x2, #0xc
+    ld1             {v0.h}[3], [x9]
+      add             x12, x2, #0x12
+    ld1             {v1.h}[3], [x15]
+      add             x13, x2, #0x24
+    ld1             {v2.h}[3], [x19]
+      add             x14, x2, #0x50
+    ld1             {v3.h}[3], [x20]
+      add             x3, x2, #0xe
+    ld1             {v0.h}[4], [x12]
+      add             x9, x2, #0x4
+    ld1             {v1.h}[4], [x13]
+      add             x15, x2, #0x16
+    ld1             {v2.h}[4], [x14]
+      add             x19, x2, #0x60
+    ld1             {v3.h}[4], [x3]
+      add             x20, x2, #0x1c
+    ld1             {v0.h}[5], [x9]
+      add             x12, x2, #0x6
+    ld1             {v1.h}[5], [x15]
+      add             x13, x2, #0x8
+    ld1             {v2.h}[5], [x19]
+      add             x14, x2, #0x52
+    ld1             {v3.h}[5], [x20]
+      add             x3, x2, #0x2a
+    ld1             {v0.h}[6], [x12]
+      add             x9, x2, #0x14
+    ld1             {v1.h}[6], [x13]
+      add             x15, x2, #0xa
+    ld1             {v2.h}[6], [x14]
+      add             x19, x2, #0x44
+    ld1             {v3.h}[6], [x3]
+      add             x20, x2, #0x38
+    ld1             {v0.h}[7], [x9]
+      add             x12, x2, #0x46
+    ld1             {v1.h}[7], [x15]
+      add             x13, x2, #0x3a
+    ld1             {v2.h}[7], [x19]
+      add             x14, x2, #0x74
+    ld1             {v3.h}[7], [x20]
+      add             x3, x2, #0x6a
+    ld1             {v4.h}[0], [x12]
+      add             x9, x2, #0x54
+    ld1             {v5.h}[0], [x13]
+      add             x15, x2, #0x2c
+    ld1             {v6.h}[0], [x14]
+      add             x19, x2, #0x76
+    ld1             {v7.h}[0], [x3]
+      add             x20, x2, #0x78
+    ld1             {v4.h}[1], [x9]
+      add             x12, x2, #0x62
+    ld1             {v5.h}[1], [x15]
+      add             x13, x2, #0x1e
+    ld1             {v6.h}[1], [x19]
+      add             x14, x2, #0x68
+    ld1             {v7.h}[1], [x20]
+      add             x3, x2, #0x7a
+    ld1             {v4.h}[2], [x12]
+      add             x9, x2, #0x70
+    ld1             {v5.h}[2], [x13]
+      add             x15, x2, #0x2e
+    ld1             {v6.h}[2], [x14]
+      add             x19, x2, #0x5a
+    ld1             {v7.h}[2], [x3]
+      add             x20, x2, #0x6c
+    ld1             {v4.h}[3], [x9]
+      add             x12, x2, #0x72
+    ld1             {v5.h}[3], [x15]
+      add             x13, x2, #0x3c
+    ld1             {v6.h}[3], [x19]
+      add             x14, x2, #0x4c
+    ld1             {v7.h}[3], [x20]
+      add             x3, x2, #0x5e
+    ld1             {v4.h}[4], [x12]
+      add             x9, x2, #0x64
+    ld1             {v5.h}[4], [x13]
+      add             x15, x2, #0x4a
+    ld1             {v6.h}[4], [x14]
+      add             x19, x2, #0x3e
+    ld1             {v7.h}[4], [x3]
+      add             x20, x2, #0x6e
+    ld1             {v4.h}[5], [x9]
+      add             x12, x2, #0x56
+    ld1             {v5.h}[5], [x15]
+      add             x13, x2, #0x58
+    ld1             {v6.h}[5], [x19]
+      add             x14, x2, #0x4e
+    ld1             {v7.h}[5], [x20]
+      add             x3, x2, #0x7c
+    ld1             {v4.h}[6], [x12]
+      add             x9, x2, #0x48
+    ld1             {v5.h}[6], [x13]
+      add             x15, x2, #0x66
+    ld1             {v6.h}[6], [x14]
+      add             x19, x2, #0x5c
+    ld1             {v7.h}[6], [x3]
+      add             x20, x2, #0x7e
+    ld1             {v4.h}[7], [x9]
+    ld1             {v5.h}[7], [x15]
+    ld1             {v6.h}[7], [x19]
+    ld1             {v7.h}[7], [x20]
+.endif
+    cmlt            v24.8h, v0.8h, #0
+    cmlt            v25.8h, v1.8h, #0
+    cmlt            v26.8h, v2.8h, #0
+    cmlt            v27.8h, v3.8h, #0
+    cmlt            v28.8h, v4.8h, #0
+    cmlt            v29.8h, v5.8h, #0
+    cmlt            v30.8h, v6.8h, #0
+    cmlt            v31.8h, v7.8h, #0
+    abs             v0.8h, v0.8h
+    abs             v1.8h, v1.8h
+    abs             v2.8h, v2.8h
+    abs             v3.8h, v3.8h
+    abs             v4.8h, v4.8h
+    abs             v5.8h, v5.8h
+    abs             v6.8h, v6.8h
+    abs             v7.8h, v7.8h
+    eor             v24.16b, v24.16b, v0.16b
+    eor             v25.16b, v25.16b, v1.16b
+    eor             v26.16b, v26.16b, v2.16b
+    eor             v27.16b, v27.16b, v3.16b
+    eor             v28.16b, v28.16b, v4.16b
+    eor             v29.16b, v29.16b, v5.16b
+    eor             v30.16b, v30.16b, v6.16b
+    eor             v31.16b, v31.16b, v7.16b
+    cmeq            v16.8h, v0.8h, #0
+    cmeq            v17.8h, v1.8h, #0
+    cmeq            v18.8h, v2.8h, #0
+    cmeq            v19.8h, v3.8h, #0
+    cmeq            v20.8h, v4.8h, #0
+    cmeq            v21.8h, v5.8h, #0
+    cmeq            v22.8h, v6.8h, #0
+    xtn             v16.8b, v16.8h
+    xtn             v18.8b, v18.8h
+    xtn             v20.8b, v20.8h
+    xtn             v22.8b, v22.8h
+      umov            w14, v0.h[0]
+    xtn2            v16.16b, v17.8h
+      umov            w13, v24.h[0]
+    xtn2            v18.16b, v19.8h
+      clz             w14, w14
+    xtn2            v20.16b, v21.8h
+      lsl             w13, w13, w14
+    cmeq            v17.8h, v7.8h, #0
+      sub             w12, w14, #32
+    xtn2            v22.16b, v17.8h
+      lsr             w13, w13, w14
+    and             v16.16b, v16.16b, v23.16b
+      neg             w12, w12
+    and             v18.16b, v18.16b, v23.16b
+      add             x3, x4, #0x400           /* r1 = dctbl->ehufsi */
+    and             v20.16b, v20.16b, v23.16b
+      add             x15, sp, #0x80           /* x15 = t2 */
+    and             v22.16b, v22.16b, v23.16b
+      ldr             w10, [x4, x12, lsl #2]
+    addp            v16.16b, v16.16b, v18.16b
+      ldrb            w11, [x3, x12]
+    addp            v20.16b, v20.16b, v22.16b
+      checkbuf47
+    addp            v16.16b, v16.16b, v20.16b
+      put_bits        x10, x11
+    addp            v16.16b, v16.16b, v18.16b
+      checkbuf47
+    umov            x9,v16.D[0]
+      put_bits        x13, x12
+    cnt             v17.8b, v16.8b
+      mvn             x9, x9
+    addv            B18, v17.8b
+      add             x4, x5, #0x400   /* x4 = actbl->ehufsi */
+    umov            w12, v18.b[0]
+      lsr             x9, x9, #0x1     /* clear AC coeff */
+    ldr             w13, [x5, #0x3c0]  /* x13 = actbl->ehufco[0xf0] */
+    rbit            x9, x9             /* x9 = index0 */
+    ldrb            w14, [x4, #0xf0]   /* x14 = actbl->ehufsi[0xf0] */
+    cmp             w12, #(64-8)
+    mov             x11, sp
+    b.lt            4f
+    cbz             x9, 6f
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x11], #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x11], #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, [x11], #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, [x11], #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w20, [x15, #-126]
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    clz             w20, w20
+    ldrh            w3, [x15, #2]!
+    sub             w11, w20, #32
+    lsl             w3, w3, w20
+    neg             w11, w11
+    lsr             w3, w3, w20
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, [x5, x2, lsl #2]
+    ldrb            w10, [x4, x2]
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+    b               6f
+4:
+    movi            v21.8h, #0x0010
+    clz             v0.8h, v0.8h
+    clz             v1.8h, v1.8h
+    clz             v2.8h, v2.8h
+    clz             v3.8h, v3.8h
+    clz             v4.8h, v4.8h
+    clz             v5.8h, v5.8h
+    clz             v6.8h, v6.8h
+    clz             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    neg             v0.8h, v0.8h
+    neg             v1.8h, v1.8h
+    neg             v2.8h, v2.8h
+    neg             v3.8h, v3.8h
+    neg             v4.8h, v4.8h
+    neg             v5.8h, v5.8h
+    neg             v6.8h, v6.8h
+    neg             v7.8h, v7.8h
+    ushl            v24.8h, v24.8h, v0.8h
+    ushl            v25.8h, v25.8h, v1.8h
+    ushl            v26.8h, v26.8h, v2.8h
+    ushl            v27.8h, v27.8h, v3.8h
+    ushl            v28.8h, v28.8h, v4.8h
+    ushl            v29.8h, v29.8h, v5.8h
+    ushl            v30.8h, v30.8h, v6.8h
+    ushl            v31.8h, v31.8h, v7.8h
+    add             v0.8h, v21.8h, v0.8h
+    add             v1.8h, v21.8h, v1.8h
+    add             v2.8h, v21.8h, v2.8h
+    add             v3.8h, v21.8h, v3.8h
+    add             v4.8h, v21.8h, v4.8h
+    add             v5.8h, v21.8h, v5.8h
+    add             v6.8h, v21.8h, v6.8h
+    add             v7.8h, v21.8h, v7.8h
+    st1             {v0.8h, v1.8h, v2.8h, v3.8h}, [x11], #64
+    st1             {v4.8h, v5.8h, v6.8h, v7.8h}, [x11], #64
+    st1             {v24.8h, v25.8h, v26.8h, v27.8h}, [x11], #64
+    st1             {v28.8h, v29.8h, v30.8h, v31.8h}, [x11], #64
+1:
+    clz             x2, x9
+    add             x15, x15, x2, lsl #1
+    lsl             x9, x9, x2
+    ldrh            w11, [x15, #-126]
+2:
+    cmp             x2, #0x10
+    b.lt            3f
+    sub             x2, x2, #0x10
+    checkbuf47
+    put_bits        x13, x14
+    b               2b
+3:
+    ldrh            w3, [x15, #2]!
+    add             x2, x11, x2, lsl #4
+    lsl             x9, x9, #0x1
+    ldr             w12, [x5, x2, lsl #2]
+    ldrb            w10, [x4, x2]
+    checkbuf31
+    put_bits        x12, x10
+    put_bits        x3, x11
+    cbnz            x9, 1b
+6:
+    add             x13, sp, #0xfe
+    cmp             x15, x13
+    b.hs            1f
+    ldr             w12, [x5]
+    ldrb            w14, [x4]
+    checkbuf47
+    put_bits        x12, x14
+1:
+    sub             sp, sp, 16
+    str             PUT_BUFFER, [x0, #0x10]
+    str             PUT_BITSw, [x0, #0x18]
+    ldp             x19, x20, [sp], 16
+    add             x0, BUFFER, #0x1
+    add             sp, sp, 256
+    br              x30
+
+.endm
+
+generate_jsimd_huff_encode_one_block 1
+generate_jsimd_huff_encode_one_block 0
+
+    .unreq          BUFFER
+    .unreq          PUT_BUFFER
+    .unreq          PUT_BITS
+    .unreq          PUT_BITSw
+
+.purgem emit_byte
+.purgem put_bits
+.purgem checkbuf31
+.purgem checkbuf47
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm.c	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm.c	2021-04-25 01:47:22.000000000 +0800
@@ -2,17 +2,16 @@
  * jsimd_arm.c
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009-2011 D. R. Commander
- * 
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015-2016 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
  *
  * This file contains the interface between the "normal" portions
- * of the library and the SIMD implementations when running on
- * ARM architecture.
- *
- * Based on the stubs from 'jsimd_none.c'
+ * of the library and the SIMD implementations when running on a
+ * 32-bit ARM architecture.
  */
 
 #define JPEG_INTERNALS
@@ -28,8 +27,9 @@
 #include <ctype.h>
 
 static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
 
-#if defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
 
 #define SOMEWHAT_SANE_PROC_CPUINFO_SIZE_LIMIT (1024 * 1024)
 
@@ -100,7 +100,7 @@
 init_simd (void)
 {
   char *env = NULL;
-#if !defined(__ARM_NEON__) && defined(__linux__) || defined(ANDROID) || defined(__ANDROID__)
+#if !defined(__ARM_NEON__) && (defined(__linux__) || defined(ANDROID) || defined(__ANDROID__))
   int bufsize = 1024; /* an initial guess for the line buffer size limit */
 #endif
 
@@ -123,12 +123,15 @@
 #endif
 
   /* Force different settings through environment variables */
-  env = getenv("JSIMD_FORCE_ARM_NEON");
+  env = getenv("JSIMD_FORCENEON");
   if ((env != NULL) && (strcmp(env, "1") == 0))
     simd_support &= JSIMD_ARM_NEON;
-  env = getenv("JSIMD_FORCE_NO_SIMD");
+  env = getenv("JSIMD_FORCENONE");
   if ((env != NULL) && (strcmp(env, "1") == 0))
     simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
 }
 
 GLOBAL(int)
@@ -170,6 +173,24 @@
     return 0;
   if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
     return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
   if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
@@ -183,8 +204,7 @@
 {
   void (*neonfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
 
-  switch(cinfo->in_color_space)
-  {
+  switch(cinfo->in_color_space) {
     case JCS_EXT_RGB:
       neonfct=jsimd_extrgb_ycc_convert_neon;
       break;
@@ -212,9 +232,7 @@
       break;
   }
 
-  if (simd_support & JSIMD_ARM_NEON)
-    neonfct(cinfo->image_width, input_buf,
-        output_buf, output_row, num_rows);
+  neonfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
 }
 
 GLOBAL(void)
@@ -231,8 +249,7 @@
 {
   void (*neonfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
 
-  switch(cinfo->out_color_space)
-  {
+  switch(cinfo->out_color_space) {
     case JCS_EXT_RGB:
       neonfct=jsimd_ycc_extrgb_convert_neon;
       break;
@@ -255,14 +272,21 @@
     case JCS_EXT_ARGB:
       neonfct=jsimd_ycc_extxrgb_convert_neon;
       break;
-  default:
+    default:
       neonfct=jsimd_ycc_extrgb_convert_neon;
       break;
   }
 
-  if (simd_support & JSIMD_ARM_NEON)
-    neonfct(cinfo->output_width, input_buf,
-        input_row, output_buf, num_rows);
+  neonfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  jsimd_ycc_rgb565_convert_neon(cinfo->output_width, input_buf, input_row,
+                                output_buf, num_rows);
 }
 
 GLOBAL(int)
@@ -282,13 +306,13 @@
 }
 
 GLOBAL(void)
-jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
 }
 
 GLOBAL(void)
-jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
                        JSAMPARRAY input_data, JSAMPARRAY output_data)
 {
 }
@@ -311,17 +335,17 @@
 
 GLOBAL(void)
 jsimd_h2v2_upsample (j_decompress_ptr cinfo,
-                     jpeg_component_info * compptr, 
+                     jpeg_component_info *compptr,
                      JSAMPARRAY input_data,
-                     JSAMPARRAY * output_data_ptr)
+                     JSAMPARRAY *output_data_ptr)
 {
 }
 
 GLOBAL(void)
 jsimd_h2v1_upsample (j_decompress_ptr cinfo,
-                     jpeg_component_info * compptr, 
+                     jpeg_component_info *compptr,
                      JSAMPARRAY input_data,
-                     JSAMPARRAY * output_data_ptr)
+                     JSAMPARRAY *output_data_ptr)
 {
 }
 
@@ -338,23 +362,35 @@
 {
   init_simd();
 
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON)
+    return 1;
+
   return 0;
 }
 
 GLOBAL(void)
 jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
-                           jpeg_component_info * compptr, 
+                           jpeg_component_info *compptr,
                            JSAMPARRAY input_data,
-                           JSAMPARRAY * output_data_ptr)
+                           JSAMPARRAY *output_data_ptr)
 {
 }
 
 GLOBAL(void)
 jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
-                           jpeg_component_info * compptr, 
+                           jpeg_component_info *compptr,
                            JSAMPARRAY input_data,
-                           JSAMPARRAY * output_data_ptr)
+                           JSAMPARRAY *output_data_ptr)
 {
+  jsimd_h2v1_fancy_upsample_neon(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
 }
 
 GLOBAL(int)
@@ -420,15 +456,14 @@
 
 GLOBAL(void)
 jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
-                DCTELEM * workspace)
+                DCTELEM *workspace)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_convsamp_neon(sample_data, start_col, workspace);
+  jsimd_convsamp_neon(sample_data, start_col, workspace);
 }
 
 GLOBAL(void)
 jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
-                      FAST_FLOAT * workspace)
+                      FAST_FLOAT *workspace)
 {
 }
 
@@ -466,19 +501,18 @@
 }
 
 GLOBAL(void)
-jsimd_fdct_islow (DCTELEM * data)
+jsimd_fdct_islow (DCTELEM *data)
 {
 }
 
 GLOBAL(void)
-jsimd_fdct_ifast (DCTELEM * data)
+jsimd_fdct_ifast (DCTELEM *data)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_fdct_ifast_neon(data);
+  jsimd_fdct_ifast_neon(data);
 }
 
 GLOBAL(void)
-jsimd_fdct_float (FAST_FLOAT * data)
+jsimd_fdct_float (FAST_FLOAT *data)
 {
 }
 
@@ -510,16 +544,15 @@
 }
 
 GLOBAL(void)
-jsimd_quantize (JCOEFPTR coef_block, DCTELEM * divisors,
-                DCTELEM * workspace)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
 {
-  if (simd_support & JSIMD_ARM_NEON)
-    jsimd_quantize_neon(coef_block, divisors, workspace);
+  jsimd_quantize_neon(coef_block, divisors, workspace);
 }
 
 GLOBAL(void)
-jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT * divisors,
-                      FAST_FLOAT * workspace)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
 {
 }
 
@@ -540,7 +573,7 @@
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
@@ -563,28 +596,28 @@
   if (sizeof(ISLOW_MULT_TYPE) != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
 }
 
 GLOBAL(void)
-jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
                 JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_2x2_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info * compptr,
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
                 JCOEFPTR coef_block, JSAMPARRAY output_buf,
                 JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_4x4_neon(compptr->dct_table, coef_block, output_buf,
+                      output_col);
 }
 
 GLOBAL(int)
@@ -629,7 +662,7 @@
   if (IFAST_SCALE_BITS != 2)
     return 0;
 
-  if ((simd_support & JSIMD_ARM_NEON))
+  if (simd_support & JSIMD_ARM_NEON)
     return 1;
 
   return 0;
@@ -644,27 +677,51 @@
 }
 
 GLOBAL(void)
-jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_islow_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
-  if ((simd_support & JSIMD_ARM_NEON))
-    jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf, output_col);
+  jsimd_idct_ifast_neon(compptr->dct_table, coef_block, output_buf,
+                        output_col);
 }
 
 GLOBAL(void)
-jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info * compptr,
-                JCOEFPTR coef_block, JSAMPARRAY output_buf,
-                JDIMENSION output_col)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
 {
 }
 
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ARM_NEON && simd_huffman)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_neon(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm_neon.S glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm_neon.S
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_arm_neon.S	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_arm_neon.S	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,13 @@
 /*
- * ARM NEON optimizations for libjpeg-turbo
+ * ARMv7 NEON optimizations for libjpeg-turbo
  *
  * Copyright (C) 2009-2011 Nokia Corporation and/or its subsidiary(-ies).
  * All rights reserved.
  * Author: Siarhei Siamashka <siarhei.siamashka@nokia.com>
+ * Copyright (C) 2014 Siarhei Siamashka.  All Rights Reserved.
+ * Copyright (C) 2014 Linaro Limited.  All Rights Reserved.
+ * Copyright (C) 2015 D. R. Commander.  All Rights Reserved.
+ * Copyright (C) 2015-2016 Matthieu Darbois.  All Rights Reserved.
  *
  * This software is provided 'as-is', without any express or implied
  * warranty.  In no event will the authors be held liable for any damages
@@ -23,7 +27,7 @@
  */
 
 #if defined(__linux__) && defined(__ELF__)
-.section .note.GNU-stack,"",%progbits /* mark stack as non-executable */
+.section .note.GNU-stack, "", %progbits  /* mark stack as non-executable */
 #endif
 
 .text
@@ -31,20 +35,20 @@
 .arch armv7a
 .object_arch armv4
 .arm
+.syntax unified
 
 
 #define RESPECT_STRICT_ALIGNMENT 1
 
+
 /*****************************************************************************/
 
 /* Supplementary macro for setting function attributes */
 .macro asm_function fname
 #ifdef __APPLE__
-    .func _\fname
     .globl _\fname
 _\fname:
 #else
-    .func \fname
     .global \fname
 #ifdef __ELF__
     .hidden \fname
@@ -56,12 +60,13 @@
 
 /* Transpose a block of 4x4 coefficients in four 64-bit registers */
 .macro transpose_4x4 x0, x1, x2, x3
-    vtrn.16 \x0, \x1
-    vtrn.16 \x2, \x3
-    vtrn.32 \x0, \x2
-    vtrn.32 \x1, \x3
+    vtrn.16         \x0, \x1
+    vtrn.16         \x2, \x3
+    vtrn.32         \x0, \x2
+    vtrn.32         \x1, \x3
 .endm
 
+
 #define CENTERJSAMPLE 128
 
 /*****************************************************************************/
@@ -70,22 +75,22 @@
  * Perform dequantization and inverse DCT on one block of coefficients.
  *
  * GLOBAL(void)
- * jsimd_idct_islow_neon (void * dct_table, JCOEFPTR coef_block,
+ * jsimd_idct_islow_neon (void *dct_table, JCOEFPTR coef_block,
  *                        JSAMPARRAY output_buf, JDIMENSION output_col)
  */
 
-#define FIX_0_298631336  (2446)
-#define FIX_0_390180644  (3196)
-#define FIX_0_541196100  (4433)
-#define FIX_0_765366865  (6270)
-#define FIX_0_899976223  (7373)
-#define FIX_1_175875602  (9633)
-#define FIX_1_501321110  (12299)
-#define FIX_1_847759065  (15137)
-#define FIX_1_961570560  (16069)
-#define FIX_2_053119869  (16819)
-#define FIX_2_562915447  (20995)
-#define FIX_3_072711026  (25172)
+#define FIX_0_298631336 (2446)
+#define FIX_0_390180644 (3196)
+#define FIX_0_541196100 (4433)
+#define FIX_0_765366865 (6270)
+#define FIX_0_899976223 (7373)
+#define FIX_1_175875602 (9633)
+#define FIX_1_501321110 (12299)
+#define FIX_1_847759065 (15137)
+#define FIX_1_961570560 (16069)
+#define FIX_2_053119869 (16819)
+#define FIX_2_562915447 (20995)
+#define FIX_3_072711026 (25172)
 
 #define FIX_1_175875602_MINUS_1_961570560 (FIX_1_175875602 - FIX_1_961570560)
 #define FIX_1_175875602_MINUS_0_390180644 (FIX_1_175875602 - FIX_0_390180644)
@@ -103,8 +108,8 @@
 #define REF_1D_IDCT(xrow0, xrow1, xrow2, xrow3, xrow4, xrow5, xrow6, xrow7)   \
 {                                                                             \
     DCTELEM row0, row1, row2, row3, row4, row5, row6, row7;                   \
-    INT32   q1, q2, q3, q4, q5, q6, q7;                                       \
-    INT32   tmp11_plus_tmp2, tmp11_minus_tmp2;                                \
+    JLONG   q1, q2, q3, q4, q5, q6, q7;                                       \
+    JLONG   tmp11_plus_tmp2, tmp11_minus_tmp2;                                \
                                                                               \
     /* 1-D iDCT input data */                                                 \
     row0 = xrow0;                                                             \
@@ -125,7 +130,7 @@
     q2 = MULTIPLY(row2, FIX_0_541196100) +                                    \
          MULTIPLY(row6, FIX_0_541196100_MINUS_1_847759065);                   \
     q4 = q6;                                                                  \
-    q3 = ((INT32) row0 - (INT32) row4) << 13;                                 \
+    q3 = ((JLONG) row0 - (JLONG) row4) << 13;                                 \
     q6 += MULTIPLY(row5, -FIX_2_562915447) +                                  \
           MULTIPLY(row3, FIX_3_072711026_MINUS_2_562915447);                  \
     /* now we can use q1 (reloadable constants have been used up) */          \
@@ -152,7 +157,7 @@
     /* (tmp11 - tmp2) has been calculated (out_row6 before descale) */        \
     tmp11_minus_tmp2 = q1;                                                    \
                                                                               \
-    q1 = ((INT32) row0 + (INT32) row4) << 13;                                 \
+    q1 = ((JLONG) row0 + (JLONG) row4) << 13;                                 \
     q2 = q1 + q6;                                                             \
     q1 = q1 - q6;                                                             \
                                                                               \
@@ -167,34 +172,34 @@
     tmp13 = q1;                                                               \
 }
 
-#define XFIX_0_899976223                    d0[0]
-#define XFIX_0_541196100                    d0[1]
-#define XFIX_2_562915447                    d0[2]
-#define XFIX_0_298631336_MINUS_0_899976223  d0[3]
-#define XFIX_1_501321110_MINUS_0_899976223  d1[0]
-#define XFIX_2_053119869_MINUS_2_562915447  d1[1]
-#define XFIX_0_541196100_PLUS_0_765366865   d1[2]
-#define XFIX_1_175875602                    d1[3]
-#define XFIX_1_175875602_MINUS_0_390180644  d2[0]
-#define XFIX_0_541196100_MINUS_1_847759065  d2[1]
-#define XFIX_3_072711026_MINUS_2_562915447  d2[2]
-#define XFIX_1_175875602_MINUS_1_961570560  d2[3]
+#define XFIX_0_899976223                   d0[0]
+#define XFIX_0_541196100                   d0[1]
+#define XFIX_2_562915447                   d0[2]
+#define XFIX_0_298631336_MINUS_0_899976223 d0[3]
+#define XFIX_1_501321110_MINUS_0_899976223 d1[0]
+#define XFIX_2_053119869_MINUS_2_562915447 d1[1]
+#define XFIX_0_541196100_PLUS_0_765366865  d1[2]
+#define XFIX_1_175875602                   d1[3]
+#define XFIX_1_175875602_MINUS_0_390180644 d2[0]
+#define XFIX_0_541196100_MINUS_1_847759065 d2[1]
+#define XFIX_3_072711026_MINUS_2_562915447 d2[2]
+#define XFIX_1_175875602_MINUS_1_961570560 d2[3]
 
 .balign 16
 jsimd_idct_islow_neon_consts:
-    .short FIX_0_899976223                    /* d0[0] */
-    .short FIX_0_541196100                    /* d0[1] */
-    .short FIX_2_562915447                    /* d0[2] */
-    .short FIX_0_298631336_MINUS_0_899976223  /* d0[3] */
-    .short FIX_1_501321110_MINUS_0_899976223  /* d1[0] */
-    .short FIX_2_053119869_MINUS_2_562915447  /* d1[1] */
-    .short FIX_0_541196100_PLUS_0_765366865   /* d1[2] */
-    .short FIX_1_175875602                    /* d1[3] */
-    /* reloadable constants */
-    .short FIX_1_175875602_MINUS_0_390180644  /* d2[0] */
-    .short FIX_0_541196100_MINUS_1_847759065  /* d2[1] */
-    .short FIX_3_072711026_MINUS_2_562915447  /* d2[2] */
-    .short FIX_1_175875602_MINUS_1_961570560  /* d2[3] */
+  .short FIX_0_899976223                    /* d0[0] */
+  .short FIX_0_541196100                    /* d0[1] */
+  .short FIX_2_562915447                    /* d0[2] */
+  .short FIX_0_298631336_MINUS_0_899976223  /* d0[3] */
+  .short FIX_1_501321110_MINUS_0_899976223  /* d1[0] */
+  .short FIX_2_053119869_MINUS_2_562915447  /* d1[1] */
+  .short FIX_0_541196100_PLUS_0_765366865   /* d1[2] */
+  .short FIX_1_175875602                    /* d1[3] */
+  /* reloadable constants */
+  .short FIX_1_175875602_MINUS_0_390180644  /* d2[0] */
+  .short FIX_0_541196100_MINUS_1_847759065  /* d2[1] */
+  .short FIX_3_072711026_MINUS_2_562915447  /* d2[2] */
+  .short FIX_1_175875602_MINUS_1_961570560  /* d2[3] */
 
 asm_function jsimd_idct_islow_neon
 
@@ -253,140 +258,141 @@
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
     vmul.s16        q14, q14, q2
     vmul.s16        q13, q13, q1
-    vld1.16         {d0, d1, d2, d3}, [ip, :128] /* load constants */
+    vld1.16         {d0, d1, d2, d3}, [ip, :128]  /* load constants */
     add             ip, ip, #16
     vmul.s16        q15, q15, q3
-    vpush           {d8-d15} /* save NEON registers */
+    vpush           {d8-d15}                      /* save NEON registers */
     /* 1-D IDCT, pass 1, left 4x8 half */
-    vadd.s16        d4,    ROW7L, ROW3L
-    vadd.s16        d5,    ROW5L, ROW1L
-    vmull.s16       q6,    d4,    XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    d5,    XFIX_1_175875602
-    vmull.s16       q7,    d4,    XFIX_1_175875602
+    vadd.s16        d4, ROW7L, ROW3L
+    vadd.s16        d5, ROW5L, ROW1L
+    vmull.s16       q6, d4, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d5, XFIX_1_175875602
+    vmull.s16       q7, d4, XFIX_1_175875602
       /* Check for the zero coefficients in the right 4x8 half */
       push            {r4, r5}
-    vmlal.s16       q7,    d5,    XFIX_1_175875602_MINUS_0_390180644
-    vsubl.s16       q3,    ROW0L, ROW4L
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))]
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vmlal.s16       q2,    ROW6L, XFIX_0_541196100_MINUS_1_847759065
-      orr             r0,    r4,    r5
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5L, XFIX_2_562915447
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))]
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vshl.s32        q3,    q3,    #13
-      orr             r0,    r0,    r4
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-      orr             r0,    r0,    r5
-    vadd.s32        q1,    q3,    q2
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))]
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-      orr             r0,    r0,    r4
-    vmlsl.s16       q7,    ROW7L, XFIX_0_899976223
-      orr             r0,    r0,    r5
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vrshrn.s32      ROW1L, q1,    #11
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))]
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5L, XFIX_2_053119869_MINUS_2_562915447
-      orr             r0,    r0,    r4
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-      orr             r0,    r0,    r5
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))]
-    vmlal.s16       q6,    ROW6L, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-      orr             r0,    r0,    r4
-    vrshrn.s32      ROW6L, q1,    #11
-      orr             r0,    r0,    r5
-    vadd.s32        q1,    q3,    q5
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))]
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0L, ROW4L
-      orr             r0,    r0,    r4
-    vrshrn.s32      ROW2L, q1,    #11
-      orr             r0,    r0,    r5
-    vrshrn.s32      ROW5L, q3,    #11
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))]
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7L, XFIX_0_298631336_MINUS_0_899976223
-      orr             r0,    r0,    r4
-    vadd.s32        q2,    q5,    q6
-      orrs            r0,    r0,    r5
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-      ldrd            r4,    [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))]
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-      orr             r0,    r4,    r5
-    vsub.s32        q3,    q1,    q4
+    vmlal.s16       q7, d5, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW4L
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 1 * 8))]
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW6L, XFIX_0_541196100_MINUS_1_847759065
+      orr             r0, r4, r5
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5L, XFIX_2_562915447
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 2 * 8))]
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+      orr             r0, r0, r4
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q2
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 3 * 8))]
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+      orr             r0, r0, r4
+    vmlsl.s16       q7, ROW7L, XFIX_0_899976223
+      orr             r0, r0, r5
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1L, q1, #11
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 4 * 8))]
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5L, XFIX_2_053119869_MINUS_2_562915447
+      orr             r0, r0, r4
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+      orr             r0, r0, r5
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 5 * 8))]
+    vmlal.s16       q6, ROW6L, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+      orr             r0, r0, r4
+    vrshrn.s32      ROW6L, q1, #11
+      orr             r0, r0, r5
+    vadd.s32        q1, q3, q5
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 6 * 8))]
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW4L
+      orr             r0, r0, r4
+    vrshrn.s32      ROW2L, q1, #11
+      orr             r0, r0, r5
+    vrshrn.s32      ROW5L, q3, #11
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 7 * 8))]
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7L, XFIX_0_298631336_MINUS_0_899976223
+      orr             r0, r0, r4
+    vadd.s32        q2, q5, q6
+      orrs            r0, r0, r5
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+      ldrd            r4, [COEF_BLOCK, #(-96 + 2 * (4 + 0 * 8))]
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+      orr             r0, r4, r5
+    vsub.s32        q3, q1, q4
       pop             {r4, r5}
-    vrshrn.s32      ROW7L, q2,    #11
-    vrshrn.s32      ROW3L, q5,    #11
-    vrshrn.s32      ROW0L, q6,    #11
-    vrshrn.s32      ROW4L, q3,    #11
+    vrshrn.s32      ROW7L, q2, #11
+    vrshrn.s32      ROW3L, q5, #11
+    vrshrn.s32      ROW0L, q6, #11
+    vrshrn.s32      ROW4L, q3, #11
 
-      beq             3f /* Go to do some special handling for the sparse right 4x8 half */
+      beq             3f  /* Go to do some special handling for the sparse
+                             right 4x8 half */
 
     /* 1-D IDCT, pass 1, right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vadd.s16        d10,   ROW7R, ROW3R
-    vadd.s16        d8,    ROW5R, ROW1R
+    vld1.s16        {d2}, [ip, :64]  /* reload constants */
+    vadd.s16        d10, ROW7R, ROW3R
+    vadd.s16        d8, ROW5R, ROW1R
       /* Transpose left 4x8 half */
       vtrn.16         ROW6L, ROW7L
-    vmull.s16       q6,    d10,   XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    d8,    XFIX_1_175875602
+    vmull.s16       q6, d10, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, d8, XFIX_1_175875602
       vtrn.16         ROW2L, ROW3L
-    vmull.s16       q7,    d10,   XFIX_1_175875602
-    vmlal.s16       q7,    d8,    XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q7, d10, XFIX_1_175875602
+    vmlal.s16       q7, d8, XFIX_1_175875602_MINUS_0_390180644
       vtrn.16         ROW0L, ROW1L
-    vsubl.s16       q3,    ROW0R, ROW4R
-    vmull.s16       q2,    ROW2R, XFIX_0_541196100
-    vmlal.s16       q2,    ROW6R, XFIX_0_541196100_MINUS_1_847759065
+    vsubl.s16       q3, ROW0R, ROW4R
+    vmull.s16       q2, ROW2R, XFIX_0_541196100
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
       vtrn.16         ROW4L, ROW5L
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5R, XFIX_2_562915447
-    vmlal.s16       q6,    ROW3R, XFIX_3_072711026_MINUS_2_562915447
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW3R, XFIX_3_072711026_MINUS_2_562915447
       vtrn.32         ROW1L, ROW3L
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW1R, XFIX_0_899976223
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1R, XFIX_0_899976223
       vtrn.32         ROW4L, ROW6L
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
       vtrn.32         ROW0L, ROW2L
-    vmlsl.s16       q7,    ROW7R, XFIX_0_899976223
-    vmlal.s16       q7,    ROW1R, XFIX_1_501321110_MINUS_0_899976223
-    vrshrn.s32      ROW1R, q1,    #11
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW1R, XFIX_1_501321110_MINUS_0_899976223
+    vrshrn.s32      ROW1R, q1, #11
       vtrn.32         ROW5L, ROW7L
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5R, XFIX_2_053119869_MINUS_2_562915447
-    vmlsl.s16       q5,    ROW3R, XFIX_2_562915447
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2R, XFIX_0_541196100_PLUS_0_765366865
-    vmlal.s16       q6,    ROW6R, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-    vrshrn.s32      ROW6R, q1,    #11
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0R, ROW4R
-    vrshrn.s32      ROW2R, q1,    #11
-    vrshrn.s32      ROW5R, q3,    #11
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7R, XFIX_0_298631336_MINUS_0_899976223
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vrshrn.s32      ROW7R, q2,    #11
-    vrshrn.s32      ROW3R, q5,    #11
-    vrshrn.s32      ROW0R, q6,    #11
-    vrshrn.s32      ROW4R, q3,    #11
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW3R, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2R, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vrshrn.s32      ROW6R, q1, #11
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0R, ROW4R
+    vrshrn.s32      ROW2R, q1, #11
+    vrshrn.s32      ROW5R, q3, #11
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vrshrn.s32      ROW7R, q2, #11
+    vrshrn.s32      ROW3R, q5, #11
+    vrshrn.s32      ROW0R, q6, #11
+    vrshrn.s32      ROW4R, q3, #11
     /* Transpose right 4x8 half */
     vtrn.16         ROW6R, ROW7R
     vtrn.16         ROW2R, ROW3R
@@ -398,122 +404,122 @@
     vtrn.32         ROW5R, ROW7R
 
 1:  /* 1-D IDCT, pass 2 (normal variant), left 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW1R, XFIX_1_175875602 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW1L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW3R, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */
-    vmlal.s16       q6,    ROW3L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW3R, XFIX_1_175875602 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW3L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW1R, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */
-    vmlal.s16       q7,    ROW1L, XFIX_1_175875602_MINUS_0_390180644
-    vsubl.s16       q3,    ROW0L, ROW0R /* ROW4L <-> ROW0R */
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vmlal.s16       q2,    ROW2R, XFIX_0_541196100_MINUS_1_847759065 /* ROW6L <-> ROW2R */
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW1R, XFIX_2_562915447 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-    vmlsl.s16       q7,    ROW3R, XFIX_0_899976223 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vshrn.s32       ROW1L, q1,    #16
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW1R, XFIX_2_053119869_MINUS_2_562915447 /* ROW5L <-> ROW1R */
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-    vmlal.s16       q6,    ROW2R, XFIX_0_541196100 /* ROW6L <-> ROW2R */
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW2R, q1,    #16 /* ROW6L <-> ROW2R */
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW0L, ROW0R /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW2L, q1,    #16
-    vshrn.s32       ROW1R, q3,    #16 /* ROW5L <-> ROW1R */
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW3R, XFIX_0_298631336_MINUS_0_899976223 /* ROW7L <-> ROW3R */
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW3R, q2,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW3L, q5,    #16
-    vshrn.s32       ROW0L, q6,    #16
-    vshrn.s32       ROW0R, q3,    #16 /* ROW4L <-> ROW0R */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW1R, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3R, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3R, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1R, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vsubl.s16       q3, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vmlal.s16       q2, ROW2R, XFIX_0_541196100_MINUS_1_847759065  /* ROW6L <-> ROW2R */
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW1R, XFIX_2_562915447   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW3R, XFIX_0_899976223   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW1R, XFIX_2_053119869_MINUS_2_562915447  /* ROW5L <-> ROW1R */
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vmlal.s16       q6, ROW2R, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW0L, ROW0R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW3R, XFIX_0_298631336_MINUS_0_899976223  /* ROW7L <-> ROW3R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
     /* 1-D IDCT, pass 2, right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW5R, XFIX_1_175875602
-    vmlal.s16       q6,    ROW5L, XFIX_1_175875602 /* ROW5L <-> ROW1R */
-    vmlal.s16       q6,    ROW7R, XFIX_1_175875602_MINUS_1_961570560
-    vmlal.s16       q6,    ROW7L, XFIX_1_175875602_MINUS_1_961570560 /* ROW7L <-> ROW3R */
-    vmull.s16       q7,    ROW7R, XFIX_1_175875602
-    vmlal.s16       q7,    ROW7L, XFIX_1_175875602 /* ROW7L <-> ROW3R */
-    vmlal.s16       q7,    ROW5R, XFIX_1_175875602_MINUS_0_390180644
-    vmlal.s16       q7,    ROW5L, XFIX_1_175875602_MINUS_0_390180644 /* ROW5L <-> ROW1R */
-    vsubl.s16       q3,    ROW4L, ROW4R /* ROW4L <-> ROW0R */
-    vmull.s16       q2,    ROW6L, XFIX_0_541196100 /* ROW6L <-> ROW2R */
-    vmlal.s16       q2,    ROW6R, XFIX_0_541196100_MINUS_1_847759065
-    vmov            q4,    q6
-    vmlsl.s16       q6,    ROW5R, XFIX_2_562915447
-    vmlal.s16       q6,    ROW7L, XFIX_3_072711026_MINUS_2_562915447 /* ROW7L <-> ROW3R */
-    vshl.s32        q3,    q3,    #13
-    vmlsl.s16       q4,    ROW5L, XFIX_0_899976223 /* ROW5L <-> ROW1R */
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vadd.s32        q1,    q1,    q6
-    vmlsl.s16       q7,    ROW7R, XFIX_0_899976223
-    vmlal.s16       q7,    ROW5L, XFIX_1_501321110_MINUS_0_899976223 /* ROW5L <-> ROW1R */
-    vshrn.s32       ROW5L, q1,    #16 /* ROW5L <-> ROW1R */
-    vsub.s32        q1,    q1,    q6
-    vmlal.s16       q5,    ROW5R, XFIX_2_053119869_MINUS_2_562915447
-    vmlsl.s16       q5,    ROW7L, XFIX_2_562915447 /* ROW7L <-> ROW3R */
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW6L, XFIX_0_541196100_PLUS_0_765366865 /* ROW6L <-> ROW2R */
-    vmlal.s16       q6,    ROW6R, XFIX_0_541196100
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW6R, q1,    #16
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vaddl.s16       q5,    ROW4L, ROW4R /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW6L, q1,    #16 /* ROW6L <-> ROW2R */
-    vshrn.s32       ROW5R, q3,    #16
-    vshl.s32        q5,    q5,    #13
-    vmlal.s16       q4,    ROW7R, XFIX_0_298631336_MINUS_0_899976223
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW7R, q2,    #16
-    vshrn.s32       ROW7L, q5,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW4L, q6,    #16 /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW4R, q3,    #16
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW5R, XFIX_1_175875602
+    vmlal.s16       q6, ROW5L, XFIX_1_175875602   /* ROW5L <-> ROW1R */
+    vmlal.s16       q6, ROW7R, XFIX_1_175875602_MINUS_1_961570560
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560  /* ROW7L <-> ROW3R */
+    vmull.s16       q7, ROW7R, XFIX_1_175875602
+    vmlal.s16       q7, ROW7L, XFIX_1_175875602   /* ROW7L <-> ROW3R */
+    vmlal.s16       q7, ROW5R, XFIX_1_175875602_MINUS_0_390180644
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644  /* ROW5L <-> ROW1R */
+    vsubl.s16       q3, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vmull.s16       q2, ROW6L, XFIX_0_541196100   /* ROW6L <-> ROW2R */
+    vmlal.s16       q2, ROW6R, XFIX_0_541196100_MINUS_1_847759065
+    vmov            q4, q6
+    vmlsl.s16       q6, ROW5R, XFIX_2_562915447
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447  /* ROW7L <-> ROW3R */
+    vshl.s32        q3, q3, #13
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223   /* ROW5L <-> ROW1R */
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vadd.s32        q1, q1, q6
+    vmlsl.s16       q7, ROW7R, XFIX_0_899976223
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223  /* ROW5L <-> ROW1R */
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmlal.s16       q5, ROW5R, XFIX_2_053119869_MINUS_2_562915447
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447   /* ROW7L <-> ROW3R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865  /* ROW6L <-> ROW2R */
+    vmlal.s16       q6, ROW6R, XFIX_0_541196100
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vaddl.s16       q5, ROW4L, ROW4R              /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vshl.s32        q5, q5, #13
+    vmlal.s16       q4, ROW7R, XFIX_0_298631336_MINUS_0_899976223
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
 
 2:  /* Descale to 8-bit and range limit */
-    vqrshrn.s16     d16,   q8,    #2
-    vqrshrn.s16     d17,   q9,    #2
-    vqrshrn.s16     d18,   q10,   #2
-    vqrshrn.s16     d19,   q11,   #2
-    vpop            {d8-d15} /* restore NEON registers */
-    vqrshrn.s16     d20,   q12,   #2
+    vqrshrn.s16     d16, q8, #2
+    vqrshrn.s16     d17, q9, #2
+    vqrshrn.s16     d18, q10, #2
+    vqrshrn.s16     d19, q11, #2
+    vpop            {d8-d15}                      /* restore NEON registers */
+    vqrshrn.s16     d20, q12, #2
       /* Transpose the final 8-bit samples and do signed->unsigned conversion */
-      vtrn.16         q8,    q9
-    vqrshrn.s16     d21,   q13,   #2
-    vqrshrn.s16     d22,   q14,   #2
-      vmov.u8         q0,    #(CENTERJSAMPLE)
-    vqrshrn.s16     d23,   q15,   #2
-      vtrn.8          d16,   d17
-      vtrn.8          d18,   d19
-      vadd.u8         q8,    q8,    q0
-      vadd.u8         q9,    q9,    q0
-      vtrn.16         q10,   q11
+      vtrn.16         q8, q9
+    vqrshrn.s16     d21, q13, #2
+    vqrshrn.s16     d22, q14, #2
+      vmov.u8         q0, #(CENTERJSAMPLE)
+    vqrshrn.s16     d23, q15, #2
+      vtrn.8          d16, d17
+      vtrn.8          d18, d19
+      vadd.u8         q8, q8, q0
+      vadd.u8         q9, q9, q0
+      vtrn.16         q10, q11
         /* Store results to the output buffer */
         ldmia           OUTPUT_BUF!, {TMP1, TMP2}
         add             TMP1, TMP1, OUTPUT_COL
@@ -525,7 +531,7 @@
         add             TMP1, TMP1, OUTPUT_COL
         add             TMP2, TMP2, OUTPUT_COL
         vst1.8          {d18}, [TMP1]
-      vadd.u8         q10,   q10,   q0
+      vadd.u8         q10, q10, q0
         vst1.8          {d19}, [TMP2]
         ldmia           OUTPUT_BUF, {TMP1, TMP2, TMP3, TMP4}
         add             TMP1, TMP1, OUTPUT_COL
@@ -534,7 +540,7 @@
         add             TMP4, TMP4, OUTPUT_COL
       vtrn.8          d22, d23
         vst1.8          {d20}, [TMP1]
-      vadd.u8         q11,   q11,   q0
+      vadd.u8         q11, q11, q0
         vst1.8          {d21}, [TMP2]
         vst1.8          {d22}, [TMP3]
         vst1.8          {d23}, [TMP4]
@@ -547,14 +553,15 @@
     vtrn.16         ROW2L, ROW3L
     vtrn.16         ROW0L, ROW1L
     vtrn.16         ROW4L, ROW5L
-    vshl.s16        ROW0R, ROW0R, #2 /* PASS1_BITS */
+    vshl.s16        ROW0R, ROW0R, #2  /* PASS1_BITS */
     vtrn.32         ROW1L, ROW3L
     vtrn.32         ROW4L, ROW6L
     vtrn.32         ROW0L, ROW2L
     vtrn.32         ROW5L, ROW7L
 
     cmp             r0, #0
-    beq             4f /* Right 4x8 half has all zeros, go to 'sparse' second pass */
+    beq             4f  /* Right 4x8 half has all zeros, go to 'sparse' second
+                           pass */
 
     /* Only row 0 is non-zero for the right 4x8 half  */
     vdup.s16        ROW1R, ROW0R[1]
@@ -565,83 +572,83 @@
     vdup.s16        ROW6R, ROW0R[2]
     vdup.s16        ROW7R, ROW0R[3]
     vdup.s16        ROW0R, ROW0R[0]
-    b               1b /* Go to 'normal' second pass */
+    b               1b  /* Go to 'normal' second pass */
 
 4:  /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), left 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW1L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW3L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW3L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW1L, XFIX_1_175875602_MINUS_0_390180644
-    vmull.s16       q2,    ROW2L, XFIX_0_541196100
-    vshll.s16       q3,    ROW0L, #13
-    vmov            q4,    q6
-    vmlal.s16       q6,    ROW3L, XFIX_3_072711026_MINUS_2_562915447
-    vmlsl.s16       q4,    ROW1L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vmlal.s16       q7,    ROW1L, XFIX_1_501321110_MINUS_0_899976223
-    vadd.s32        q1,    q1,    q6
-    vadd.s32        q6,    q6,    q6
-    vmlsl.s16       q5,    ROW3L, XFIX_2_562915447
-    vshrn.s32       ROW1L, q1,    #16
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW2L, XFIX_0_541196100_PLUS_0_765366865
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW2R, q1,    #16 /* ROW6L <-> ROW2R */
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vshll.s16       q5,    ROW0L, #13
-    vshrn.s32       ROW2L, q1,    #16
-    vshrn.s32       ROW1R, q3,    #16 /* ROW5L <-> ROW1R */
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW3R, q2,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW3L, q5,    #16
-    vshrn.s32       ROW0L, q6,    #16
-    vshrn.s32       ROW0R, q3,    #16 /* ROW4L <-> ROW0R */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW1L, XFIX_1_175875602
+    vmlal.s16       q6, ROW3L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW3L, XFIX_1_175875602
+    vmlal.s16       q7, ROW1L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW2L, XFIX_0_541196100
+    vshll.s16       q3, ROW0L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW3L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW1L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW1L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW3L, XFIX_2_562915447
+    vshrn.s32       ROW1L, q1, #16
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW2L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW2R, q1, #16                /* ROW6L <-> ROW2R */
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW0L, #13
+    vshrn.s32       ROW2L, q1, #16
+    vshrn.s32       ROW1R, q3, #16                /* ROW5L <-> ROW1R */
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW3R, q2, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW3L, q5, #16
+    vshrn.s32       ROW0L, q6, #16
+    vshrn.s32       ROW0R, q3, #16                /* ROW4L <-> ROW0R */
     /* 1-D IDCT, pass 2 (sparse variant with zero rows 4-7), right 4x8 half */
-    vld1.s16        {d2},  [ip, :64]    /* reload constants */
-    vmull.s16       q6,    ROW5L, XFIX_1_175875602
-    vmlal.s16       q6,    ROW7L, XFIX_1_175875602_MINUS_1_961570560
-    vmull.s16       q7,    ROW7L, XFIX_1_175875602
-    vmlal.s16       q7,    ROW5L, XFIX_1_175875602_MINUS_0_390180644
-    vmull.s16       q2,    ROW6L, XFIX_0_541196100
-    vshll.s16       q3,    ROW4L, #13
-    vmov            q4,    q6
-    vmlal.s16       q6,    ROW7L, XFIX_3_072711026_MINUS_2_562915447
-    vmlsl.s16       q4,    ROW5L, XFIX_0_899976223
-    vadd.s32        q1,    q3,    q2
-    vmov            q5,    q7
-    vmlal.s16       q7,    ROW5L, XFIX_1_501321110_MINUS_0_899976223
-    vadd.s32        q1,    q1,    q6
-    vadd.s32        q6,    q6,    q6
-    vmlsl.s16       q5,    ROW7L, XFIX_2_562915447
-    vshrn.s32       ROW5L, q1,    #16 /* ROW5L <-> ROW1R */
-    vsub.s32        q1,    q1,    q6
-    vmull.s16       q6,    ROW6L, XFIX_0_541196100_PLUS_0_765366865
-    vsub.s32        q3,    q3,    q2
-    vshrn.s32       ROW6R, q1,    #16
-    vadd.s32        q1,    q3,    q5
-    vsub.s32        q3,    q3,    q5
-    vshll.s16       q5,    ROW4L, #13
-    vshrn.s32       ROW6L, q1,    #16 /* ROW6L <-> ROW2R */
-    vshrn.s32       ROW5R, q3,    #16
-    vadd.s32        q2,    q5,    q6
-    vsub.s32        q1,    q5,    q6
-    vadd.s32        q6,    q2,    q7
-    vsub.s32        q2,    q2,    q7
-    vadd.s32        q5,    q1,    q4
-    vsub.s32        q3,    q1,    q4
-    vshrn.s32       ROW7R, q2,    #16
-    vshrn.s32       ROW7L, q5,    #16 /* ROW7L <-> ROW3R */
-    vshrn.s32       ROW4L, q6,    #16 /* ROW4L <-> ROW0R */
-    vshrn.s32       ROW4R, q3,    #16
-    b               2b /* Go to epilogue */
+    vld1.s16        {d2}, [ip, :64]               /* reload constants */
+    vmull.s16       q6, ROW5L, XFIX_1_175875602
+    vmlal.s16       q6, ROW7L, XFIX_1_175875602_MINUS_1_961570560
+    vmull.s16       q7, ROW7L, XFIX_1_175875602
+    vmlal.s16       q7, ROW5L, XFIX_1_175875602_MINUS_0_390180644
+    vmull.s16       q2, ROW6L, XFIX_0_541196100
+    vshll.s16       q3, ROW4L, #13
+    vmov            q4, q6
+    vmlal.s16       q6, ROW7L, XFIX_3_072711026_MINUS_2_562915447
+    vmlsl.s16       q4, ROW5L, XFIX_0_899976223
+    vadd.s32        q1, q3, q2
+    vmov            q5, q7
+    vmlal.s16       q7, ROW5L, XFIX_1_501321110_MINUS_0_899976223
+    vadd.s32        q1, q1, q6
+    vadd.s32        q6, q6, q6
+    vmlsl.s16       q5, ROW7L, XFIX_2_562915447
+    vshrn.s32       ROW5L, q1, #16                /* ROW5L <-> ROW1R */
+    vsub.s32        q1, q1, q6
+    vmull.s16       q6, ROW6L, XFIX_0_541196100_PLUS_0_765366865
+    vsub.s32        q3, q3, q2
+    vshrn.s32       ROW6R, q1, #16
+    vadd.s32        q1, q3, q5
+    vsub.s32        q3, q3, q5
+    vshll.s16       q5, ROW4L, #13
+    vshrn.s32       ROW6L, q1, #16                /* ROW6L <-> ROW2R */
+    vshrn.s32       ROW5R, q3, #16
+    vadd.s32        q2, q5, q6
+    vsub.s32        q1, q5, q6
+    vadd.s32        q6, q2, q7
+    vsub.s32        q2, q2, q7
+    vadd.s32        q5, q1, q4
+    vsub.s32        q3, q1, q4
+    vshrn.s32       ROW7R, q2, #16
+    vshrn.s32       ROW7L, q5, #16                /* ROW7L <-> ROW3R */
+    vshrn.s32       ROW4L, q6, #16                /* ROW4L <-> ROW0R */
+    vshrn.s32       ROW4R, q3, #16
+    b               2b                            /* Go to epilogue */
 
     .unreq          DCT_TABLE
     .unreq          COEF_BLOCK
@@ -668,7 +675,7 @@
     .unreq          ROW6R
     .unreq          ROW7L
     .unreq          ROW7R
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -695,10 +702,10 @@
 
 .balign 16
 jsimd_idct_ifast_neon_consts:
-    .short (277 * 128 - 256 * 128) /* XFIX_1_082392200 */
-    .short (362 * 128 - 256 * 128) /* XFIX_1_414213562 */
-    .short (473 * 128 - 256 * 128) /* XFIX_1_847759065 */
-    .short (669 * 128 - 512 * 128) /* XFIX_2_613125930 */
+  .short (277 * 128 - 256 * 128)  /* XFIX_1_082392200 */
+  .short (362 * 128 - 256 * 128)  /* XFIX_1_414213562 */
+  .short (473 * 128 - 256 * 128)  /* XFIX_1_847759065 */
+  .short (669 * 128 - 512 * 128)  /* XFIX_2_613125930 */
 
 asm_function jsimd_idct_ifast_neon
 
@@ -728,9 +735,9 @@
     vld1.16         {d16, d17, d18, d19}, [COEF_BLOCK, :128]!
     vld1.16         {d0, d1, d2, d3}, [DCT_TABLE, :128]!
     vld1.16         {d20, d21, d22, d23}, [COEF_BLOCK, :128]!
-    vmul.s16        q8,  q8,  q0
+    vmul.s16        q8, q8, q0
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
-    vmul.s16        q9,  q9,  q1
+    vmul.s16        q9, q9, q1
     vld1.16         {d24, d25, d26, d27}, [COEF_BLOCK, :128]!
     vmul.s16        q10, q10, q2
     vld1.16         {d0, d1, d2, d3}, [DCT_TABLE, :128]!
@@ -740,124 +747,124 @@
     vld1.16         {d4, d5, d6, d7}, [DCT_TABLE, :128]!
     vmul.s16        q14, q14, q2
     vmul.s16        q13, q13, q1
-    vld1.16         {d0}, [ip, :64] /* load constants */
+    vld1.16         {d0}, [ip, :64]  /* load constants */
     vmul.s16        q15, q15, q3
-    vpush           {d8-d13}        /* save NEON registers */
+    vpush           {d8-d13}         /* save NEON registers */
     /* 1-D IDCT, pass 1 */
-    vsub.s16        q2,  q10, q14
+    vsub.s16        q2, q10, q14
     vadd.s16        q14, q10, q14
-    vsub.s16        q1,  q11, q13
+    vsub.s16        q1, q11, q13
     vadd.s16        q13, q11, q13
-    vsub.s16        q5,  q9,  q15
-    vadd.s16        q15, q9,  q15
-    vqdmulh.s16     q4,  q2,  XFIX_1_414213562
-    vqdmulh.s16     q6,  q1,  XFIX_2_613125930
-    vadd.s16        q3,  q1,  q1
-    vsub.s16        q1,  q5,  q1
-    vadd.s16        q10, q2,  q4
-    vqdmulh.s16     q4,  q1,  XFIX_1_847759065
-    vsub.s16        q2,  q15, q13
-    vadd.s16        q3,  q3,  q6
-    vqdmulh.s16     q6,  q2,  XFIX_1_414213562
-    vadd.s16        q1,  q1,  q4
-    vqdmulh.s16     q4,  q5,  XFIX_1_082392200
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
     vsub.s16        q10, q10, q14
-    vadd.s16        q2,  q2,  q6
-    vsub.s16        q6,  q8,  q12
-    vadd.s16        q12, q8,  q12
-    vadd.s16        q9,  q5,  q4
-    vadd.s16        q5,  q6,  q10
-    vsub.s16        q10, q6,  q10
-    vadd.s16        q6,  q15, q13
-    vadd.s16        q8,  q12, q14
-    vsub.s16        q3,  q6,  q3
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
     vsub.s16        q12, q12, q14
-    vsub.s16        q3,  q3,  q1
-    vsub.s16        q1,  q9,  q1
-    vadd.s16        q2,  q3,  q2
-    vsub.s16        q15, q8,  q6
-    vadd.s16        q1,  q1,  q2
-    vadd.s16        q8,  q8,  q6
-    vadd.s16        q14, q5,  q3
-    vsub.s16        q9,  q5,  q3
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
     vsub.s16        q13, q10, q2
     vadd.s16        q10, q10, q2
       /* Transpose */
-      vtrn.16         q8,  q9
+      vtrn.16         q8, q9
     vsub.s16        q11, q12, q1
       vtrn.16         q14, q15
     vadd.s16        q12, q12, q1
       vtrn.16         q10, q11
       vtrn.16         q12, q13
-      vtrn.32         q9,  q11
+      vtrn.32         q9, q11
       vtrn.32         q12, q14
-      vtrn.32         q8,  q10
+      vtrn.32         q8, q10
       vtrn.32         q13, q15
       vswp            d28, d21
       vswp            d26, d19
     /* 1-D IDCT, pass 2 */
-    vsub.s16        q2,  q10, q14
+    vsub.s16        q2, q10, q14
       vswp            d30, d23
     vadd.s16        q14, q10, q14
       vswp            d24, d17
-    vsub.s16        q1,  q11, q13
+    vsub.s16        q1, q11, q13
     vadd.s16        q13, q11, q13
-    vsub.s16        q5,  q9,  q15
-    vadd.s16        q15, q9,  q15
-    vqdmulh.s16     q4,  q2,  XFIX_1_414213562
-    vqdmulh.s16     q6,  q1,  XFIX_2_613125930
-    vadd.s16        q3,  q1,  q1
-    vsub.s16        q1,  q5,  q1
-    vadd.s16        q10, q2,  q4
-    vqdmulh.s16     q4,  q1,  XFIX_1_847759065
-    vsub.s16        q2,  q15, q13
-    vadd.s16        q3,  q3,  q6
-    vqdmulh.s16     q6,  q2,  XFIX_1_414213562
-    vadd.s16        q1,  q1,  q4
-    vqdmulh.s16     q4,  q5,  XFIX_1_082392200
+    vsub.s16        q5, q9, q15
+    vadd.s16        q15, q9, q15
+    vqdmulh.s16     q4, q2, XFIX_1_414213562
+    vqdmulh.s16     q6, q1, XFIX_2_613125930
+    vadd.s16        q3, q1, q1
+    vsub.s16        q1, q5, q1
+    vadd.s16        q10, q2, q4
+    vqdmulh.s16     q4, q1, XFIX_1_847759065
+    vsub.s16        q2, q15, q13
+    vadd.s16        q3, q3, q6
+    vqdmulh.s16     q6, q2, XFIX_1_414213562
+    vadd.s16        q1, q1, q4
+    vqdmulh.s16     q4, q5, XFIX_1_082392200
     vsub.s16        q10, q10, q14
-    vadd.s16        q2,  q2,  q6
-    vsub.s16        q6,  q8,  q12
-    vadd.s16        q12, q8,  q12
-    vadd.s16        q9,  q5,  q4
-    vadd.s16        q5,  q6,  q10
-    vsub.s16        q10, q6,  q10
-    vadd.s16        q6,  q15, q13
-    vadd.s16        q8,  q12, q14
-    vsub.s16        q3,  q6,  q3
+    vadd.s16        q2, q2, q6
+    vsub.s16        q6, q8, q12
+    vadd.s16        q12, q8, q12
+    vadd.s16        q9, q5, q4
+    vadd.s16        q5, q6, q10
+    vsub.s16        q10, q6, q10
+    vadd.s16        q6, q15, q13
+    vadd.s16        q8, q12, q14
+    vsub.s16        q3, q6, q3
     vsub.s16        q12, q12, q14
-    vsub.s16        q3,  q3,  q1
-    vsub.s16        q1,  q9,  q1
-    vadd.s16        q2,  q3,  q2
-    vsub.s16        q15, q8,  q6
-    vadd.s16        q1,  q1,  q2
-    vadd.s16        q8,  q8,  q6
-    vadd.s16        q14, q5,  q3
-    vsub.s16        q9,  q5,  q3
+    vsub.s16        q3, q3, q1
+    vsub.s16        q1, q9, q1
+    vadd.s16        q2, q3, q2
+    vsub.s16        q15, q8, q6
+    vadd.s16        q1, q1, q2
+    vadd.s16        q8, q8, q6
+    vadd.s16        q14, q5, q3
+    vsub.s16        q9, q5, q3
     vsub.s16        q13, q10, q2
-    vpop            {d8-d13}        /* restore NEON registers */
+    vpop            {d8-d13}      /* restore NEON registers */
     vadd.s16        q10, q10, q2
     vsub.s16        q11, q12, q1
     vadd.s16        q12, q12, q1
     /* Descale to 8-bit and range limit */
-    vmov.u8         q0,  #0x80
-    vqshrn.s16      d16, q8,  #5
-    vqshrn.s16      d17, q9,  #5
+    vmov.u8         q0, #0x80
+    vqshrn.s16      d16, q8, #5
+    vqshrn.s16      d17, q9, #5
     vqshrn.s16      d18, q10, #5
     vqshrn.s16      d19, q11, #5
     vqshrn.s16      d20, q12, #5
     vqshrn.s16      d21, q13, #5
     vqshrn.s16      d22, q14, #5
     vqshrn.s16      d23, q15, #5
-    vadd.u8         q8,  q8,  q0
-    vadd.u8         q9,  q9,  q0
+    vadd.u8         q8, q8, q0
+    vadd.u8         q9, q9, q0
     vadd.u8         q10, q10, q0
     vadd.u8         q11, q11, q0
     /* Transpose the final 8-bit samples */
-    vtrn.16         q8,  q9
+    vtrn.16         q8, q9
     vtrn.16         q10, q11
-    vtrn.32         q8,  q10
-    vtrn.32         q9,  q11
+    vtrn.32         q8, q10
+    vtrn.32         q9, q11
     vtrn.8          d16, d17
     vtrn.8          d18, d19
       /* Store results to the output buffer */
@@ -892,7 +899,7 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -916,81 +923,80 @@
 
 #define CONST_BITS  13
 
-#define FIX_0_211164243  (1730)  /* FIX(0.211164243) */
-#define FIX_0_509795579  (4176)  /* FIX(0.509795579) */
-#define FIX_0_601344887  (4926)  /* FIX(0.601344887) */
-#define FIX_0_720959822  (5906)  /* FIX(0.720959822) */
-#define FIX_0_765366865  (6270)  /* FIX(0.765366865) */
-#define FIX_0_850430095  (6967)  /* FIX(0.850430095) */
-#define FIX_0_899976223  (7373)  /* FIX(0.899976223) */
-#define FIX_1_061594337  (8697)  /* FIX(1.061594337) */
-#define FIX_1_272758580  (10426) /* FIX(1.272758580) */
-#define FIX_1_451774981  (11893) /* FIX(1.451774981) */
-#define FIX_1_847759065  (15137) /* FIX(1.847759065) */
-#define FIX_2_172734803  (17799) /* FIX(2.172734803) */
-#define FIX_2_562915447  (20995) /* FIX(2.562915447) */
-#define FIX_3_624509785  (29692) /* FIX(3.624509785) */
+#define FIX_0_211164243 (1730)   /* FIX(0.211164243) */
+#define FIX_0_509795579 (4176)   /* FIX(0.509795579) */
+#define FIX_0_601344887 (4926)   /* FIX(0.601344887) */
+#define FIX_0_720959822 (5906)   /* FIX(0.720959822) */
+#define FIX_0_765366865 (6270)   /* FIX(0.765366865) */
+#define FIX_0_850430095 (6967)   /* FIX(0.850430095) */
+#define FIX_0_899976223 (7373)   /* FIX(0.899976223) */
+#define FIX_1_061594337 (8697)   /* FIX(1.061594337) */
+#define FIX_1_272758580 (10426)  /* FIX(1.272758580) */
+#define FIX_1_451774981 (11893)  /* FIX(1.451774981) */
+#define FIX_1_847759065 (15137)  /* FIX(1.847759065) */
+#define FIX_2_172734803 (17799)  /* FIX(2.172734803) */
+#define FIX_2_562915447 (20995)  /* FIX(2.562915447) */
+#define FIX_3_624509785 (29692)  /* FIX(3.624509785) */
 
 .balign 16
 jsimd_idct_4x4_neon_consts:
-    .short     FIX_1_847759065     /* d0[0] */
-    .short     -FIX_0_765366865    /* d0[1] */
-    .short     -FIX_0_211164243    /* d0[2] */
-    .short     FIX_1_451774981     /* d0[3] */
-    .short     -FIX_2_172734803    /* d1[0] */
-    .short     FIX_1_061594337     /* d1[1] */
-    .short     -FIX_0_509795579    /* d1[2] */
-    .short     -FIX_0_601344887    /* d1[3] */
-    .short     FIX_0_899976223     /* d2[0] */
-    .short     FIX_2_562915447     /* d2[1] */
-    .short     1 << (CONST_BITS+1) /* d2[2] */
-    .short     0                   /* d2[3] */
+  .short FIX_1_847759065      /* d0[0] */
+  .short -FIX_0_765366865     /* d0[1] */
+  .short -FIX_0_211164243     /* d0[2] */
+  .short FIX_1_451774981      /* d0[3] */
+  .short -FIX_2_172734803     /* d1[0] */
+  .short FIX_1_061594337      /* d1[1] */
+  .short -FIX_0_509795579     /* d1[2] */
+  .short -FIX_0_601344887     /* d1[3] */
+  .short FIX_0_899976223      /* d2[0] */
+  .short FIX_2_562915447      /* d2[1] */
+  .short 1 << (CONST_BITS+1)  /* d2[2] */
+  .short 0                    /* d2[3] */
 
 .macro idct_helper x4, x6, x8, x10, x12, x14, x16, shift, y26, y27, y28, y29
-    vmull.s16       q14, \x4,  d2[2]
-    vmlal.s16       q14, \x8,  d0[0]
+    vmull.s16       q14, \x4, d2[2]
+    vmlal.s16       q14, \x8, d0[0]
     vmlal.s16       q14, \x14, d0[1]
 
     vmull.s16       q13, \x16, d1[2]
     vmlal.s16       q13, \x12, d1[3]
     vmlal.s16       q13, \x10, d2[0]
-    vmlal.s16       q13, \x6,  d2[1]
+    vmlal.s16       q13, \x6, d2[1]
 
-    vmull.s16       q15, \x4,  d2[2]
-    vmlsl.s16       q15, \x8,  d0[0]
+    vmull.s16       q15, \x4, d2[2]
+    vmlsl.s16       q15, \x8, d0[0]
     vmlsl.s16       q15, \x14, d0[1]
 
     vmull.s16       q12, \x16, d0[2]
     vmlal.s16       q12, \x12, d0[3]
     vmlal.s16       q12, \x10, d1[0]
-    vmlal.s16       q12, \x6,  d1[1]
+    vmlal.s16       q12, \x6, d1[1]
 
     vadd.s32        q10, q14, q13
     vsub.s32        q14, q14, q13
 
-.if \shift > 16
-    vrshr.s32       q10,  q10, #\shift
-    vrshr.s32       q14,  q14, #\shift
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q14, q14, #\shift
     vmovn.s32       \y26, q10
     vmovn.s32       \y29, q14
-.else
+  .else
     vrshrn.s32      \y26, q10, #\shift
     vrshrn.s32      \y29, q14, #\shift
-.endif
+  .endif
 
     vadd.s32        q10, q15, q12
     vsub.s32        q15, q15, q12
 
-.if \shift > 16
-    vrshr.s32       q10,  q10, #\shift
-    vrshr.s32       q15,  q15, #\shift
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q15, q15, #\shift
     vmovn.s32       \y27, q10
     vmovn.s32       \y28, q15
-.else
+  .else
     vrshrn.s32      \y27, q10, #\shift
     vrshrn.s32      \y28, q15, #\shift
-.endif
-
+  .endif
 .endm
 
 asm_function jsimd_idct_4x4_neon
@@ -1104,10 +1110,10 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
 
 .purgem idct_helper
 
+
 /*****************************************************************************/
 
 /*
@@ -1126,31 +1132,30 @@
 
 .balign 8
 jsimd_idct_2x2_neon_consts:
-    .short     -FIX_0_720959822    /* d0[0] */
-    .short     FIX_0_850430095     /* d0[1] */
-    .short     -FIX_1_272758580    /* d0[2] */
-    .short     FIX_3_624509785     /* d0[3] */
+  .short -FIX_0_720959822  /* d0[0] */
+  .short FIX_0_850430095   /* d0[1] */
+  .short -FIX_1_272758580  /* d0[2] */
+  .short FIX_3_624509785   /* d0[3] */
 
 .macro idct_helper x4, x6, x10, x12, x16, shift, y26, y27
-    vshll.s16  q14,  \x4,  #15
-    vmull.s16  q13,  \x6,  d0[3]
-    vmlal.s16  q13,  \x10, d0[2]
-    vmlal.s16  q13,  \x12, d0[1]
-    vmlal.s16  q13,  \x16, d0[0]
-
-    vadd.s32   q10,  q14,  q13
-    vsub.s32   q14,  q14,  q13
-
-.if \shift > 16
-    vrshr.s32  q10,  q10,  #\shift
-    vrshr.s32  q14,  q14,  #\shift
-    vmovn.s32  \y26, q10
-    vmovn.s32  \y27, q14
-.else
-    vrshrn.s32 \y26, q10,  #\shift
-    vrshrn.s32 \y27, q14,  #\shift
-.endif
+    vshll.s16       q14, \x4, #15
+    vmull.s16       q13, \x6, d0[3]
+    vmlal.s16       q13, \x10, d0[2]
+    vmlal.s16       q13, \x12, d0[1]
+    vmlal.s16       q13, \x16, d0[0]
 
+    vadd.s32        q10, q14, q13
+    vsub.s32        q14, q14, q13
+
+  .if \shift > 16
+    vrshr.s32       q10, q10, #\shift
+    vrshr.s32       q14, q14, #\shift
+    vmovn.s32       \y26, q10
+    vmovn.s32       \y27, q14
+  .else
+    vrshrn.s32      \y26, q10, #\shift
+    vrshrn.s32      \y27, q14, #\shift
+  .endif
 .endm
 
 asm_function jsimd_idct_2x2_neon
@@ -1204,30 +1209,30 @@
     /* Pass 1 */
 #if 0
     idct_helper     d4, d6, d10, d12, d16, 13, d4, d6
-    transpose_4x4   d4, d6, d8,  d10
+    transpose_4x4   d4, d6, d8, d10
     idct_helper     d5, d7, d11, d13, d17, 13, d5, d7
-    transpose_4x4   d5, d7, d9,  d11
+    transpose_4x4   d5, d7, d9, d11
 #else
-    vmull.s16       q13, d6,  d0[3]
+    vmull.s16       q13, d6, d0[3]
     vmlal.s16       q13, d10, d0[2]
     vmlal.s16       q13, d12, d0[1]
     vmlal.s16       q13, d16, d0[0]
-    vmull.s16       q12, d7,  d0[3]
+    vmull.s16       q12, d7, d0[3]
     vmlal.s16       q12, d11, d0[2]
     vmlal.s16       q12, d13, d0[1]
     vmlal.s16       q12, d17, d0[0]
-    vshll.s16       q14, d4,  #15
-    vshll.s16       q15, d5,  #15
+    vshll.s16       q14, d4, #15
+    vshll.s16       q15, d5, #15
     vadd.s32        q10, q14, q13
     vsub.s32        q14, q14, q13
-    vrshrn.s32      d4,  q10, #13
-    vrshrn.s32      d6,  q14, #13
+    vrshrn.s32      d4, q10, #13
+    vrshrn.s32      d6, q14, #13
     vadd.s32        q10, q15, q12
     vsub.s32        q14, q15, q12
-    vrshrn.s32      d5,  q10, #13
-    vrshrn.s32      d7,  q14, #13
-    vtrn.16         q2,  q3
-    vtrn.32         q3,  q5
+    vrshrn.s32      d5, q10, #13
+    vrshrn.s32      d7, q14, #13
+    vtrn.16         q2, q3
+    vtrn.32         q3, q5
 #endif
 
     /* Pass 2 */
@@ -1258,10 +1263,10 @@
     .unreq          OUTPUT_COL
     .unreq          TMP1
     .unreq          TMP2
-.endfunc
 
 .purgem idct_helper
 
+
 /*****************************************************************************/
 
 /*
@@ -1277,97 +1282,110 @@
 
 
 .macro do_load size
+  .if \size == 8
+    vld1.8          {d4}, [U, :64]!
+    vld1.8          {d5}, [V, :64]!
+    vld1.8          {d0}, [Y, :64]!
+    pld             [U, #64]
+    pld             [V, #64]
+    pld             [Y, #64]
+  .elseif \size == 4
+    vld1.8          {d4[0]}, [U]!
+    vld1.8          {d4[1]}, [U]!
+    vld1.8          {d4[2]}, [U]!
+    vld1.8          {d4[3]}, [U]!
+    vld1.8          {d5[0]}, [V]!
+    vld1.8          {d5[1]}, [V]!
+    vld1.8          {d5[2]}, [V]!
+    vld1.8          {d5[3]}, [V]!
+    vld1.8          {d0[0]}, [Y]!
+    vld1.8          {d0[1]}, [Y]!
+    vld1.8          {d0[2]}, [Y]!
+    vld1.8          {d0[3]}, [Y]!
+  .elseif \size == 2
+    vld1.8          {d4[4]}, [U]!
+    vld1.8          {d4[5]}, [U]!
+    vld1.8          {d5[4]}, [V]!
+    vld1.8          {d5[5]}, [V]!
+    vld1.8          {d0[4]}, [Y]!
+    vld1.8          {d0[5]}, [Y]!
+  .elseif \size == 1
+    vld1.8          {d4[6]}, [U]!
+    vld1.8          {d5[6]}, [V]!
+    vld1.8          {d0[6]}, [Y]!
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_store bpp, size
+  .if \bpp == 24
     .if \size == 8
-        vld1.8  {d4}, [U, :64]!
-        vld1.8  {d5}, [V, :64]!
-        vld1.8  {d0}, [Y, :64]!
-        pld     [U, #64]
-        pld     [V, #64]
-        pld     [Y, #64]
+      vst3.8        {d10, d11, d12}, [RGB]!
     .elseif \size == 4
-        vld1.8  {d4[0]}, [U]!
-        vld1.8  {d4[1]}, [U]!
-        vld1.8  {d4[2]}, [U]!
-        vld1.8  {d4[3]}, [U]!
-        vld1.8  {d5[0]}, [V]!
-        vld1.8  {d5[1]}, [V]!
-        vld1.8  {d5[2]}, [V]!
-        vld1.8  {d5[3]}, [V]!
-        vld1.8  {d0[0]}, [Y]!
-        vld1.8  {d0[1]}, [Y]!
-        vld1.8  {d0[2]}, [Y]!
-        vld1.8  {d0[3]}, [Y]!
+      vst3.8        {d10[0], d11[0], d12[0]}, [RGB]!
+      vst3.8        {d10[1], d11[1], d12[1]}, [RGB]!
+      vst3.8        {d10[2], d11[2], d12[2]}, [RGB]!
+      vst3.8        {d10[3], d11[3], d12[3]}, [RGB]!
     .elseif \size == 2
-        vld1.8  {d4[4]}, [U]!
-        vld1.8  {d4[5]}, [U]!
-        vld1.8  {d5[4]}, [V]!
-        vld1.8  {d5[5]}, [V]!
-        vld1.8  {d0[4]}, [Y]!
-        vld1.8  {d0[5]}, [Y]!
+      vst3.8        {d10[4], d11[4], d12[4]}, [RGB]!
+      vst3.8        {d10[5], d11[5], d12[5]}, [RGB]!
     .elseif \size == 1
-        vld1.8  {d4[6]}, [U]!
-        vld1.8  {d5[6]}, [V]!
-        vld1.8  {d0[6]}, [Y]!
+      vst3.8        {d10[6], d11[6], d12[6]}, [RGB]!
     .else
-        .error unsupported macroblock size
+      .error unsupported macroblock size
     .endif
-.endm
-
-.macro do_store bpp, size
-    .if \bpp == 24
-        .if \size == 8
-            vst3.8  {d10, d11, d12}, [RGB]!
-        .elseif \size == 4
-            vst3.8  {d10[0], d11[0], d12[0]}, [RGB]!
-            vst3.8  {d10[1], d11[1], d12[1]}, [RGB]!
-            vst3.8  {d10[2], d11[2], d12[2]}, [RGB]!
-            vst3.8  {d10[3], d11[3], d12[3]}, [RGB]!
-        .elseif \size == 2
-            vst3.8  {d10[4], d11[4], d12[4]}, [RGB]!
-            vst3.8  {d10[5], d11[5], d12[5]}, [RGB]!
-        .elseif \size == 1
-            vst3.8  {d10[6], d11[6], d12[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
-    .elseif \bpp == 32
-        .if \size == 8
-            vst4.8  {d10, d11, d12, d13}, [RGB]!
-        .elseif \size == 4
-            vst4.8  {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
-            vst4.8  {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
-            vst4.8  {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
-            vst4.8  {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
-        .elseif \size == 2
-            vst4.8  {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
-            vst4.8  {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
-        .elseif \size == 1
-            vst4.8  {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      vst4.8        {d10, d11, d12, d13}, [RGB]!
+    .elseif \size == 4
+      vst4.8        {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
+      vst4.8        {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
+      vst4.8        {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
+      vst4.8        {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
+    .elseif \size == 2
+      vst4.8        {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
+      vst4.8        {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
+    .elseif \size == 1
+      vst4.8        {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
     .else
-        .error unsupported bpp
+      .error unsupported macroblock size
     .endif
+  .elseif \bpp == 16
+    .if \size == 8
+      vst1.16       {q15}, [RGB]!
+    .elseif \size == 4
+      vst1.16       {d30}, [RGB]!
+    .elseif \size == 2
+      vst1.16       {d31[0]}, [RGB]!
+      vst1.16       {d31[1]}, [RGB]!
+    .elseif \size == 1
+      vst1.16       {d31[2]}, [RGB]!
+    .else
+      .error unsupported macroblock size
+    .endif
+  .else
+    .error unsupported bpp
+  .endif
 .endm
 
 .macro generate_jsimd_ycc_rgb_convert_neon colorid, bpp, r_offs, g_offs, b_offs
 
 /*
- * 2 stage pipelined YCbCr->RGB conversion
+ * 2-stage pipelined YCbCr->RGB conversion
  */
 
 .macro do_yuv_to_rgb_stage1
-    vaddw.u8        q3, q1, d4     /* q3 = u - 128 */
-    vaddw.u8        q4, q1, d5     /* q2 = v - 128 */
-    vmull.s16       q10, d6, d1[1] /* multiply by -11277 */
-    vmlal.s16       q10, d8, d1[2] /* multiply by -23401 */
-    vmull.s16       q11, d7, d1[1] /* multiply by -11277 */
-    vmlal.s16       q11, d9, d1[2] /* multiply by -23401 */
-    vmull.s16       q12, d8, d1[0] /* multiply by 22971 */
-    vmull.s16       q13, d9, d1[0] /* multiply by 22971 */
-    vmull.s16       q14, d6, d1[3] /* multiply by 29033 */
-    vmull.s16       q15, d7, d1[3] /* multiply by 29033 */
+    vaddw.u8        q3, q1, d4      /* q3 = u - 128 */
+    vaddw.u8        q4, q1, d5      /* q2 = v - 128 */
+    vmull.s16       q10, d6, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q10, d8, d1[2]  /* multiply by -23401 */
+    vmull.s16       q11, d7, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q11, d9, d1[2]  /* multiply by -23401 */
+    vmull.s16       q12, d8, d1[0]  /* multiply by 22971 */
+    vmull.s16       q13, d9, d1[0]  /* multiply by 22971 */
+    vmull.s16       q14, d6, d1[3]  /* multiply by 29033 */
+    vmull.s16       q15, d7, d1[3]  /* multiply by 29033 */
 .endm
 
 .macro do_yuv_to_rgb_stage2
@@ -1377,44 +1395,71 @@
     vrshrn.s32      d25, q13, #14
     vrshrn.s32      d28, q14, #14
     vrshrn.s32      d29, q15, #14
-    vaddw.u8        q10, q10, d0
+    vaddw.u8        q11, q10, d0
     vaddw.u8        q12, q12, d0
     vaddw.u8        q14, q14, d0
-    vqmovun.s16     d1\g_offs, q10
+  .if \bpp != 16
+    vqmovun.s16     d1\g_offs, q11
     vqmovun.s16     d1\r_offs, q12
     vqmovun.s16     d1\b_offs, q14
+  .else  /* rgb565 */
+    vqshlu.s16      q13, q11, #8
+    vqshlu.s16      q15, q12, #8
+    vqshlu.s16      q14, q14, #8
+    vsri.u16        q15, q13, #5
+    vsri.u16        q15, q14, #11
+  .endif
 .endm
 
 .macro do_yuv_to_rgb_stage2_store_load_stage1
+                                       /* "do_yuv_to_rgb_stage2" and "store" */
+                                       vrshrn.s32      d20, q10, #15
+    /* "load" and "do_yuv_to_rgb_stage1" */
+    pld             [U, #64]
+                                       vrshrn.s32      d21, q11, #15
+    pld             [V, #64]
+                                       vrshrn.s32      d24, q12, #14
+                                       vrshrn.s32      d25, q13, #14
     vld1.8          {d4}, [U, :64]!
-      vrshrn.s32      d20, q10, #15
-      vrshrn.s32      d21, q11, #15
-      vrshrn.s32      d24, q12, #14
-      vrshrn.s32      d25, q13, #14
-      vrshrn.s32      d28, q14, #14
+                                       vrshrn.s32      d28, q14, #14
     vld1.8          {d5}, [V, :64]!
-      vrshrn.s32      d29, q15, #14
-      vaddw.u8        q10, q10, d0
-      vaddw.u8        q12, q12, d0
-      vaddw.u8        q14, q14, d0
-      vqmovun.s16     d1\g_offs, q10
+                                       vrshrn.s32      d29, q15, #14
+    vaddw.u8        q3, q1, d4      /* q3 = u - 128 */
+    vaddw.u8        q4, q1, d5      /* q2 = v - 128 */
+                                       vaddw.u8        q11, q10, d0
+    vmull.s16       q10, d6, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q10, d8, d1[2]  /* multiply by -23401 */
+                                       vaddw.u8        q12, q12, d0
+                                       vaddw.u8        q14, q14, d0
+  .if \bpp != 16  /**************** rgb24/rgb32 ******************************/
+                                       vqmovun.s16     d1\g_offs, q11
+    pld             [Y, #64]
+                                       vqmovun.s16     d1\r_offs, q12
     vld1.8          {d0}, [Y, :64]!
-      vqmovun.s16     d1\r_offs, q12
-    pld             [U, #64]
-    pld             [V, #64]
+                                       vqmovun.s16     d1\b_offs, q14
+    vmull.s16       q11, d7, d1[1]  /* multiply by -11277 */
+    vmlal.s16       q11, d9, d1[2]  /* multiply by -23401 */
+                                       do_store        \bpp, 8
+    vmull.s16       q12, d8, d1[0]  /* multiply by 22971 */
+    vmull.s16       q13, d9, d1[0]  /* multiply by 22971 */
+    vmull.s16       q14, d6, d1[3]  /* multiply by 29033 */
+    vmull.s16       q15, d7, d1[3]  /* multiply by 29033 */
+  .else  /**************************** rgb565 ********************************/
+                                       vqshlu.s16      q13, q11, #8
     pld             [Y, #64]
-      vqmovun.s16     d1\b_offs, q14
-    vaddw.u8        q3, q1, d4     /* q3 = u - 128 */
-    vaddw.u8        q4, q1, d5     /* q2 = v - 128 */
-      do_store        \bpp, 8
-    vmull.s16       q10, d6, d1[1] /* multiply by -11277 */
-    vmlal.s16       q10, d8, d1[2] /* multiply by -23401 */
-    vmull.s16       q11, d7, d1[1] /* multiply by -11277 */
-    vmlal.s16       q11, d9, d1[2] /* multiply by -23401 */
-    vmull.s16       q12, d8, d1[0] /* multiply by 22971 */
-    vmull.s16       q13, d9, d1[0] /* multiply by 22971 */
-    vmull.s16       q14, d6, d1[3] /* multiply by 29033 */
-    vmull.s16       q15, d7, d1[3] /* multiply by 29033 */
+                                       vqshlu.s16      q15, q12, #8
+                                       vqshlu.s16      q14, q14, #8
+    vld1.8          {d0}, [Y, :64]!
+    vmull.s16       q11, d7, d1[1]
+    vmlal.s16       q11, d9, d1[2]
+                                       vsri.u16        q15, q13, #5
+    vmull.s16       q12, d8, d1[0]
+                                       vsri.u16        q15, q14, #11
+    vmull.s16       q13, d9, d1[0]
+    vmull.s16       q14, d6, d1[3]
+                                       do_store        \bpp, 8
+    vmull.s16       q15, d7, d1[3]
+  .endif
 .endm
 
 .macro do_yuv_to_rgb
@@ -1428,10 +1473,10 @@
 
 .balign 16
 jsimd_ycc_\colorid\()_neon_consts:
-    .short          0,      0,     0,      0
-    .short          22971, -11277, -23401, 29033
-    .short          -128,  -128,   -128,   -128
-    .short          -128,  -128,   -128,   -128
+  .short 0,      0,     0,      0
+  .short 22971, -11277, -23401, 29033
+  .short -128,  -128,   -128,   -128
+  .short -128,  -128,   -128,   -128
 
 asm_function jsimd_ycc_\colorid\()_convert_neon
     OUTPUT_WIDTH    .req r0
@@ -1541,7 +1586,6 @@
     .unreq          U
     .unreq          V
     .unreq          N
-.endfunc
 
 .purgem do_yuv_to_rgb
 .purgem do_yuv_to_rgb_stage1
@@ -1557,10 +1601,12 @@
 generate_jsimd_ycc_rgb_convert_neon extbgrx, 32, 2, 1, 0
 generate_jsimd_ycc_rgb_convert_neon extxbgr, 32, 3, 2, 1
 generate_jsimd_ycc_rgb_convert_neon extxrgb, 32, 1, 2, 3
+generate_jsimd_ycc_rgb_convert_neon rgb565,  16, 0, 0, 0
 
 .purgem do_load
 .purgem do_store
 
+
 /*****************************************************************************/
 
 /*
@@ -1575,123 +1621,123 @@
  */
 
 .macro do_store size
+  .if \size == 8
+    vst1.8          {d20}, [Y]!
+    vst1.8          {d21}, [U]!
+    vst1.8          {d22}, [V]!
+  .elseif \size == 4
+    vst1.8          {d20[0]}, [Y]!
+    vst1.8          {d20[1]}, [Y]!
+    vst1.8          {d20[2]}, [Y]!
+    vst1.8          {d20[3]}, [Y]!
+    vst1.8          {d21[0]}, [U]!
+    vst1.8          {d21[1]}, [U]!
+    vst1.8          {d21[2]}, [U]!
+    vst1.8          {d21[3]}, [U]!
+    vst1.8          {d22[0]}, [V]!
+    vst1.8          {d22[1]}, [V]!
+    vst1.8          {d22[2]}, [V]!
+    vst1.8          {d22[3]}, [V]!
+  .elseif \size == 2
+    vst1.8          {d20[4]}, [Y]!
+    vst1.8          {d20[5]}, [Y]!
+    vst1.8          {d21[4]}, [U]!
+    vst1.8          {d21[5]}, [U]!
+    vst1.8          {d22[4]}, [V]!
+    vst1.8          {d22[5]}, [V]!
+  .elseif \size == 1
+    vst1.8          {d20[6]}, [Y]!
+    vst1.8          {d21[6]}, [U]!
+    vst1.8          {d22[6]}, [V]!
+  .else
+    .error unsupported macroblock size
+  .endif
+.endm
+
+.macro do_load bpp, size
+  .if \bpp == 24
     .if \size == 8
-        vst1.8  {d20}, [Y]!
-        vst1.8  {d21}, [U]!
-        vst1.8  {d22}, [V]!
+      vld3.8        {d10, d11, d12}, [RGB]!
+      pld           [RGB, #128]
     .elseif \size == 4
-        vst1.8  {d20[0]}, [Y]!
-        vst1.8  {d20[1]}, [Y]!
-        vst1.8  {d20[2]}, [Y]!
-        vst1.8  {d20[3]}, [Y]!
-        vst1.8  {d21[0]}, [U]!
-        vst1.8  {d21[1]}, [U]!
-        vst1.8  {d21[2]}, [U]!
-        vst1.8  {d21[3]}, [U]!
-        vst1.8  {d22[0]}, [V]!
-        vst1.8  {d22[1]}, [V]!
-        vst1.8  {d22[2]}, [V]!
-        vst1.8  {d22[3]}, [V]!
+      vld3.8        {d10[0], d11[0], d12[0]}, [RGB]!
+      vld3.8        {d10[1], d11[1], d12[1]}, [RGB]!
+      vld3.8        {d10[2], d11[2], d12[2]}, [RGB]!
+      vld3.8        {d10[3], d11[3], d12[3]}, [RGB]!
     .elseif \size == 2
-        vst1.8  {d20[4]}, [Y]!
-        vst1.8  {d20[5]}, [Y]!
-        vst1.8  {d21[4]}, [U]!
-        vst1.8  {d21[5]}, [U]!
-        vst1.8  {d22[4]}, [V]!
-        vst1.8  {d22[5]}, [V]!
+      vld3.8        {d10[4], d11[4], d12[4]}, [RGB]!
+      vld3.8        {d10[5], d11[5], d12[5]}, [RGB]!
     .elseif \size == 1
-        vst1.8  {d20[6]}, [Y]!
-        vst1.8  {d21[6]}, [U]!
-        vst1.8  {d22[6]}, [V]!
+      vld3.8        {d10[6], d11[6], d12[6]}, [RGB]!
     .else
-        .error unsupported macroblock size
+      .error unsupported macroblock size
     .endif
-.endm
-
-.macro do_load bpp, size
-    .if \bpp == 24
-        .if \size == 8
-            vld3.8  {d10, d11, d12}, [RGB]!
-            pld     [RGB, #128]
-        .elseif \size == 4
-            vld3.8  {d10[0], d11[0], d12[0]}, [RGB]!
-            vld3.8  {d10[1], d11[1], d12[1]}, [RGB]!
-            vld3.8  {d10[2], d11[2], d12[2]}, [RGB]!
-            vld3.8  {d10[3], d11[3], d12[3]}, [RGB]!
-        .elseif \size == 2
-            vld3.8  {d10[4], d11[4], d12[4]}, [RGB]!
-            vld3.8  {d10[5], d11[5], d12[5]}, [RGB]!
-        .elseif \size == 1
-            vld3.8  {d10[6], d11[6], d12[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
-    .elseif \bpp == 32
-        .if \size == 8
-            vld4.8  {d10, d11, d12, d13}, [RGB]!
-            pld     [RGB, #128]
-        .elseif \size == 4
-            vld4.8  {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
-            vld4.8  {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
-            vld4.8  {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
-            vld4.8  {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
-        .elseif \size == 2
-            vld4.8  {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
-            vld4.8  {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
-        .elseif \size == 1
-            vld4.8  {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
-        .else
-            .error unsupported macroblock size
-        .endif
+  .elseif \bpp == 32
+    .if \size == 8
+      vld4.8        {d10, d11, d12, d13}, [RGB]!
+      pld           [RGB, #128]
+    .elseif \size == 4
+      vld4.8        {d10[0], d11[0], d12[0], d13[0]}, [RGB]!
+      vld4.8        {d10[1], d11[1], d12[1], d13[1]}, [RGB]!
+      vld4.8        {d10[2], d11[2], d12[2], d13[2]}, [RGB]!
+      vld4.8        {d10[3], d11[3], d12[3], d13[3]}, [RGB]!
+    .elseif \size == 2
+      vld4.8        {d10[4], d11[4], d12[4], d13[4]}, [RGB]!
+      vld4.8        {d10[5], d11[5], d12[5], d13[5]}, [RGB]!
+    .elseif \size == 1
+      vld4.8        {d10[6], d11[6], d12[6], d13[6]}, [RGB]!
     .else
-        .error unsupported bpp
+      .error unsupported macroblock size
     .endif
+  .else
+    .error unsupported bpp
+  .endif
 .endm
 
 .macro generate_jsimd_rgb_ycc_convert_neon colorid, bpp, r_offs, g_offs, b_offs
 
 /*
- * 2 stage pipelined RGB->YCbCr conversion
+ * 2-stage pipelined RGB->YCbCr conversion
  */
 
 .macro do_rgb_to_yuv_stage1
-    vmovl.u8    q2, d1\r_offs /* r = { d4, d5 } */
-    vmovl.u8    q3, d1\g_offs /* g = { d6, d7 } */
-    vmovl.u8    q4, d1\b_offs /* b = { d8, d9 } */
-    vmull.u16   q7, d4, d0[0]
-    vmlal.u16   q7, d6, d0[1]
-    vmlal.u16   q7, d8, d0[2]
-    vmull.u16   q8, d5, d0[0]
-    vmlal.u16   q8, d7, d0[1]
-    vmlal.u16   q8, d9, d0[2]
-    vrev64.32   q9,  q1
-    vrev64.32   q13, q1
-    vmlsl.u16   q9,  d4, d0[3]
-    vmlsl.u16   q9,  d6, d1[0]
-    vmlal.u16   q9,  d8, d1[1]
-    vmlsl.u16   q13, d5, d0[3]
-    vmlsl.u16   q13, d7, d1[0]
-    vmlal.u16   q13, d9, d1[1]
-    vrev64.32   q14, q1
-    vrev64.32   q15, q1
-    vmlal.u16   q14, d4, d1[1]
-    vmlsl.u16   q14, d6, d1[2]
-    vmlsl.u16   q14, d8, d1[3]
-    vmlal.u16   q15, d5, d1[1]
-    vmlsl.u16   q15, d7, d1[2]
-    vmlsl.u16   q15, d9, d1[3]
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d0[0]
+    vmlal.u16       q7, d6, d0[1]
+    vmlal.u16       q7, d8, d0[2]
+    vmull.u16       q8, d5, d0[0]
+    vmlal.u16       q8, d7, d0[1]
+    vmlal.u16       q8, d9, d0[2]
+    vrev64.32       q9, q1
+    vrev64.32       q13, q1
+    vmlsl.u16       q9, d4, d0[3]
+    vmlsl.u16       q9, d6, d1[0]
+    vmlal.u16       q9, d8, d1[1]
+    vmlsl.u16       q13, d5, d0[3]
+    vmlsl.u16       q13, d7, d1[0]
+    vmlal.u16       q13, d9, d1[1]
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d1[1]
+    vmlsl.u16       q14, d6, d1[2]
+    vmlsl.u16       q14, d8, d1[3]
+    vmlal.u16       q15, d5, d1[1]
+    vmlsl.u16       q15, d7, d1[2]
+    vmlsl.u16       q15, d9, d1[3]
 .endm
 
 .macro do_rgb_to_yuv_stage2
-    vrshrn.u32  d20, q7,  #16
-    vrshrn.u32  d21, q8,  #16
-    vshrn.u32   d22, q9,  #16
-    vshrn.u32   d23, q13, #16
-    vshrn.u32   d24, q14, #16
-    vshrn.u32   d25, q15, #16
-    vmovn.u16   d20, q10      /* d20 = y */
-    vmovn.u16   d21, q11      /* d21 = u */
-    vmovn.u16   d22, q12      /* d22 = v */
+    vrshrn.u32      d20, q7, #16
+    vrshrn.u32      d21, q8, #16
+    vshrn.u32       d22, q9, #16
+    vshrn.u32       d23, q13, #16
+    vshrn.u32       d24, q14, #16
+    vshrn.u32       d25, q15, #16
+    vmovn.u16       d20, q10       /* d20 = y */
+    vmovn.u16       d21, q11       /* d21 = u */
+    vmovn.u16       d22, q12       /* d22 = v */
 .endm
 
 .macro do_rgb_to_yuv
@@ -1700,52 +1746,52 @@
 .endm
 
 .macro do_rgb_to_yuv_stage2_store_load_stage1
-      vrshrn.u32  d20, q7,  #16
-      vrshrn.u32  d21, q8,  #16
-      vshrn.u32   d22, q9,  #16
-    vrev64.32   q9,  q1
-      vshrn.u32   d23, q13, #16
-    vrev64.32   q13, q1
-      vshrn.u32   d24, q14, #16
-      vshrn.u32   d25, q15, #16
-    do_load     \bpp, 8
-      vmovn.u16   d20, q10      /* d20 = y */
-    vmovl.u8    q2, d1\r_offs   /* r = { d4, d5 } */
-      vmovn.u16   d21, q11      /* d21 = u */
-    vmovl.u8    q3, d1\g_offs   /* g = { d6, d7 } */
-      vmovn.u16   d22, q12      /* d22 = v */
-    vmovl.u8    q4, d1\b_offs   /* b = { d8, d9 } */
-    vmull.u16   q7, d4, d0[0]
-    vmlal.u16   q7, d6, d0[1]
-    vmlal.u16   q7, d8, d0[2]
-      vst1.8      {d20}, [Y]!
-    vmull.u16   q8, d5, d0[0]
-    vmlal.u16   q8, d7, d0[1]
-    vmlal.u16   q8, d9, d0[2]
-    vmlsl.u16   q9,  d4, d0[3]
-    vmlsl.u16   q9,  d6, d1[0]
-    vmlal.u16   q9,  d8, d1[1]
-      vst1.8      {d21}, [U]!
-    vmlsl.u16   q13, d5, d0[3]
-    vmlsl.u16   q13, d7, d1[0]
-    vmlal.u16   q13, d9, d1[1]
-    vrev64.32   q14, q1
-    vrev64.32   q15, q1
-    vmlal.u16   q14, d4, d1[1]
-    vmlsl.u16   q14, d6, d1[2]
-    vmlsl.u16   q14, d8, d1[3]
-      vst1.8      {d22}, [V]!
-    vmlal.u16   q15, d5, d1[1]
-    vmlsl.u16   q15, d7, d1[2]
-    vmlsl.u16   q15, d9, d1[3]
+      vrshrn.u32      d20, q7, #16
+      vrshrn.u32      d21, q8, #16
+      vshrn.u32       d22, q9, #16
+    vrev64.32       q9, q1
+      vshrn.u32       d23, q13, #16
+    vrev64.32       q13, q1
+      vshrn.u32       d24, q14, #16
+      vshrn.u32       d25, q15, #16
+    do_load         \bpp, 8
+      vmovn.u16       d20, q10     /* d20 = y */
+    vmovl.u8        q2, d1\r_offs  /* r = { d4, d5 } */
+      vmovn.u16       d21, q11     /* d21 = u */
+    vmovl.u8        q3, d1\g_offs  /* g = { d6, d7 } */
+      vmovn.u16       d22, q12     /* d22 = v */
+    vmovl.u8        q4, d1\b_offs  /* b = { d8, d9 } */
+    vmull.u16       q7, d4, d0[0]
+    vmlal.u16       q7, d6, d0[1]
+    vmlal.u16       q7, d8, d0[2]
+      vst1.8          {d20}, [Y]!
+    vmull.u16       q8, d5, d0[0]
+    vmlal.u16       q8, d7, d0[1]
+    vmlal.u16       q8, d9, d0[2]
+    vmlsl.u16       q9, d4, d0[3]
+    vmlsl.u16       q9, d6, d1[0]
+    vmlal.u16       q9, d8, d1[1]
+      vst1.8          {d21}, [U]!
+    vmlsl.u16       q13, d5, d0[3]
+    vmlsl.u16       q13, d7, d1[0]
+    vmlal.u16       q13, d9, d1[1]
+    vrev64.32       q14, q1
+    vrev64.32       q15, q1
+    vmlal.u16       q14, d4, d1[1]
+    vmlsl.u16       q14, d6, d1[2]
+    vmlsl.u16       q14, d8, d1[3]
+      vst1.8          {d22}, [V]!
+    vmlal.u16       q15, d5, d1[1]
+    vmlsl.u16       q15, d7, d1[2]
+    vmlsl.u16       q15, d9, d1[3]
 .endm
 
 .balign 16
 jsimd_\colorid\()_ycc_neon_consts:
-    .short          19595, 38470, 7471,  11059
-    .short          21709, 32768, 27439, 5329
-    .short          32767, 128,   32767, 128
-    .short          32767, 128,   32767, 128
+  .short 19595, 38470, 7471,  11059
+  .short 21709, 32768, 27439, 5329
+  .short 32767, 128,   32767, 128
+  .short 32767, 128,   32767, 128
 
 asm_function jsimd_\colorid\()_ycc_convert_neon
     OUTPUT_WIDTH    .req r0
@@ -1851,7 +1897,6 @@
     .unreq          U
     .unreq          V
     .unreq          N
-.endfunc
 
 .purgem do_rgb_to_yuv
 .purgem do_rgb_to_yuv_stage1
@@ -1871,6 +1916,7 @@
 .purgem do_load
 .purgem do_store
 
+
 /*****************************************************************************/
 
 /*
@@ -1932,7 +1978,7 @@
     .unreq          TMP2
     .unreq          TMP3
     .unreq          TMP4
-.endfunc
+
 
 /*****************************************************************************/
 
@@ -1955,10 +2001,10 @@
 
 .balign 16
 jsimd_fdct_ifast_neon_consts:
-    .short (98 * 128)              /* XFIX_0_382683433 */
-    .short (139 * 128)             /* XFIX_0_541196100 */
-    .short (181 * 128)             /* XFIX_0_707106781 */
-    .short (334 * 128 - 256 * 128) /* XFIX_1_306562965 */
+  .short (98 * 128)               /* XFIX_0_382683433 */
+  .short (139 * 128)              /* XFIX_0_541196100 */
+  .short (181 * 128)              /* XFIX_0_707106781 */
+  .short (334 * 128 - 256 * 128)  /* XFIX_1_306562965 */
 
 asm_function jsimd_fdct_ifast_neon
 
@@ -1995,52 +2041,52 @@
     /* Transpose */
     vtrn.16         q12, q13
     vtrn.16         q10, q11
-    vtrn.16         q8,  q9
+    vtrn.16         q8, q9
     vtrn.16         q14, q15
-    vtrn.32         q9,  q11
+    vtrn.32         q9, q11
     vtrn.32         q13, q15
-    vtrn.32         q8,  q10
+    vtrn.32         q8, q10
     vtrn.32         q12, q14
     vswp            d30, d23
     vswp            d24, d17
     vswp            d26, d19
       /* 1-D FDCT */
-      vadd.s16        q2,  q11, q12
+      vadd.s16        q2, q11, q12
     vswp            d28, d21
       vsub.s16        q12, q11, q12
-      vsub.s16        q6,  q10, q13
+      vsub.s16        q6, q10, q13
       vadd.s16        q10, q10, q13
-      vsub.s16        q7,  q9,  q14
-      vadd.s16        q9,  q9,  q14
-      vsub.s16        q1,  q8,  q15
-      vadd.s16        q8,  q8,  q15
-      vsub.s16        q4,  q9,  q10
-      vsub.s16        q5,  q8,  q2
-      vadd.s16        q3,  q9,  q10
-      vadd.s16        q4,  q4,  q5
-      vadd.s16        q2,  q8,  q2
-      vqdmulh.s16     q4,  q4,  XFIX_0_707106781
+      vsub.s16        q7, q9, q14
+      vadd.s16        q9, q9, q14
+      vsub.s16        q1, q8, q15
+      vadd.s16        q8, q8, q15
+      vsub.s16        q4, q9, q10
+      vsub.s16        q5, q8, q2
+      vadd.s16        q3, q9, q10
+      vadd.s16        q4, q4, q5
+      vadd.s16        q2, q8, q2
+      vqdmulh.s16     q4, q4, XFIX_0_707106781
       vadd.s16        q11, q12, q6
-      vadd.s16        q8,  q2,  q3
-      vsub.s16        q12, q2,  q3
-      vadd.s16        q3,  q6,  q7
-      vadd.s16        q7,  q7,  q1
-      vqdmulh.s16     q3,  q3,  XFIX_0_707106781
-      vsub.s16        q6,  q11, q7
-      vadd.s16        q10, q5,  q4
-      vqdmulh.s16     q6,  q6,  XFIX_0_382683433
-      vsub.s16        q14, q5,  q4
+      vadd.s16        q8, q2, q3
+      vsub.s16        q12, q2, q3
+      vadd.s16        q3, q6, q7
+      vadd.s16        q7, q7, q1
+      vqdmulh.s16     q3, q3, XFIX_0_707106781
+      vsub.s16        q6, q11, q7
+      vadd.s16        q10, q5, q4
+      vqdmulh.s16     q6, q6, XFIX_0_382683433
+      vsub.s16        q14, q5, q4
       vqdmulh.s16     q11, q11, XFIX_0_541196100
-      vqdmulh.s16     q5,  q7,  XFIX_1_306562965
-      vadd.s16        q4,  q1,  q3
-      vsub.s16        q3,  q1,  q3
-      vadd.s16        q7,  q7,  q6
+      vqdmulh.s16     q5, q7, XFIX_1_306562965
+      vadd.s16        q4, q1, q3
+      vsub.s16        q3, q1, q3
+      vadd.s16        q7, q7, q6
       vadd.s16        q11, q11, q6
-      vadd.s16        q7,  q7,  q5
-      vadd.s16        q13, q3,  q11
-      vsub.s16        q11, q3,  q11
-      vadd.s16        q9,  q4,  q7
-      vsub.s16        q15, q4,  q7
+      vadd.s16        q7, q7, q5
+      vadd.s16        q13, q3, q11
+      vsub.s16        q11, q3, q11
+      vadd.s16        q9, q4, q7
+      vsub.s16        q15, q4, q7
     subs            TMP, TMP, #1
     bne             1b
 
@@ -2055,14 +2101,14 @@
 
     .unreq          DATA
     .unreq          TMP
-.endfunc
+
 
 /*****************************************************************************/
 
 /*
  * GLOBAL(void)
- * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM * divisors,
- *                      DCTELEM * workspace);
+ * jsimd_quantize_neon (JCOEFPTR coef_block, DCTELEM *divisors,
+ *                      DCTELEM *workspace);
  *
  * Note: the code uses 2 stage pipelining in order to improve instructions
  *       scheduling and eliminate stalls (this provides ~15% better
@@ -2089,22 +2135,22 @@
     vld1.16         {d20, d21, d22, d23}, [CORRECTION, :128]!
     vabs.s16        q13, q1
     vld1.16         {d16, d17, d18, d19}, [RECIPROCAL, :128]!
-    vadd.u16        q12, q12, q10 /* add correction */
+    vadd.u16        q12, q12, q10  /* add correction */
     vadd.u16        q13, q13, q11
-    vmull.u16       q10, d24, d16 /* multiply by reciprocal */
+    vmull.u16       q10, d24, d16  /* multiply by reciprocal */
     vmull.u16       q11, d25, d17
-    vmull.u16       q8,  d26, d18
-    vmull.u16       q9,  d27, d19
+    vmull.u16       q8, d26, d18
+    vmull.u16       q9, d27, d19
     vld1.16         {d24, d25, d26, d27}, [SHIFT, :128]!
     vshrn.u32       d20, q10, #16
     vshrn.u32       d21, q11, #16
-    vshrn.u32       d22, q8,  #16
-    vshrn.u32       d23, q9,  #16
+    vshrn.u32       d22, q8, #16
+    vshrn.u32       d23, q9, #16
     vneg.s16        q12, q12
     vneg.s16        q13, q13
-    vshr.s16        q2,  q0,  #15 /* extract sign */
-    vshr.s16        q3,  q1,  #15
-    vshl.u16        q14, q10, q12 /* shift */
+    vshr.s16        q2, q0, #15    /* extract sign */
+    vshr.s16        q3, q1, #15
+    vshl.u16        q14, q10, q12  /* shift */
     vshl.u16        q15, q11, q13
 
     push            {r4, r5}
@@ -2117,25 +2163,25 @@
     vabs.s16        q13, q1
       veor.u16        q15, q15, q3
     vld1.16         {d16, d17, d18, d19}, [RECIPROCAL, :128]!
-    vadd.u16        q12, q12, q10 /* add correction */
+    vadd.u16        q12, q12, q10  /* add correction */
     vadd.u16        q13, q13, q11
-    vmull.u16       q10, d24, d16 /* multiply by reciprocal */
+    vmull.u16       q10, d24, d16  /* multiply by reciprocal */
     vmull.u16       q11, d25, d17
-    vmull.u16       q8,  d26, d18
-    vmull.u16       q9,  d27, d19
+    vmull.u16       q8, d26, d18
+    vmull.u16       q9, d27, d19
       vsub.u16        q14, q14, q2
     vld1.16         {d24, d25, d26, d27}, [SHIFT, :128]!
       vsub.u16        q15, q15, q3
     vshrn.u32       d20, q10, #16
     vshrn.u32       d21, q11, #16
       vst1.16         {d28, d29, d30, d31}, [COEF_BLOCK, :128]!
-    vshrn.u32       d22, q8,  #16
-    vshrn.u32       d23, q9,  #16
+    vshrn.u32       d22, q8, #16
+    vshrn.u32       d23, q9, #16
     vneg.s16        q12, q12
     vneg.s16        q13, q13
-    vshr.s16        q2,  q0,  #15 /* extract sign */
-    vshr.s16        q3,  q1,  #15
-    vshl.u16        q14, q10, q12 /* shift */
+    vshr.s16        q2, q0, #15    /* extract sign */
+    vshr.s16        q3, q1, #15
+    vshl.u16        q14, q10, q12  /* shift */
     vshl.u16        q15, q11, q13
     subs            LOOP_COUNT, LOOP_COUNT, #1
     bne             1b
@@ -2147,7 +2193,7 @@
       vsub.u16        q15, q15, q3
       vst1.16         {d28, d29, d30, d31}, [COEF_BLOCK, :128]!
 
-    bx              lr /* return */
+    bx              lr  /* return */
 
     .unreq          COEF_BLOCK
     .unreq          DIVISORS
@@ -2156,4 +2202,677 @@
     .unreq          CORRECTION
     .unreq          SHIFT
     .unreq          LOOP_COUNT
-.endfunc
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(void)
+ * jsimd_h2v1_fancy_upsample_neon (int max_v_samp_factor,
+ *                                 JDIMENSION downsampled_width,
+ *                                 JSAMPARRAY input_data,
+ *                                 JSAMPARRAY *output_data_ptr);
+ *
+ * Note: the use of unaligned writes is the main remaining bottleneck in
+ *       this code, which can be potentially solved to get up to tens
+ *       of percents performance improvement on Cortex-A8/Cortex-A9.
+ */
+
+/*
+ * Upsample 16 source pixels to 32 destination pixels. The new 16 source
+ * pixels are loaded to q0. The previous 16 source pixels are in q1. The
+ * shifted-by-one source pixels are constructed in q2 by using q0 and q1.
+ * Register d28 is used for multiplication by 3. Register q15 is used
+ * for adding +1 bias.
+ */
+.macro upsample16 OUTPTR, INPTR
+    vld1.8          {q0}, [\INPTR]!
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+    vmov            q1, q0        /* backup source pixels to q1 */
+    vrshrn.u16      d6, q8, #2
+    vrshrn.u16      d7, q9, #2
+    vshrn.u16       d8, q10, #2
+    vshrn.u16       d9, q11, #2
+    vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+.endm
+
+/*
+ * Upsample 32 source pixels to 64 destination pixels. Compared to 'usample16'
+ * macro, the roles of q0 and q1 registers are reversed for even and odd
+ * groups of 16 pixels, that's why "vmov q1, q0" instructions are not needed.
+ * Also this unrolling allows to reorder loads and stores to compensate
+ * multiplication latency and reduce stalls.
+ */
+.macro upsample32 OUTPTR, INPTR
+    /* even 16 pixels group */
+    vld1.8          {q0}, [\INPTR]!
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+      /* odd 16 pixels group */
+      vld1.8          {q1}, [\INPTR]!
+    vrshrn.u16      d6, q8, #2
+    vrshrn.u16      d7, q9, #2
+    vshrn.u16       d8, q10, #2
+    vshrn.u16       d9, q11, #2
+      vmovl.u8        q8, d2
+      vext.8          q2, q0, q1, #15
+      vmovl.u8        q9, d3
+      vaddw.u8        q10, q15, d4
+      vaddw.u8        q11, q15, d5
+      vmlal.u8        q8, d4, d28
+      vmlal.u8        q9, d5, d28
+      vmlal.u8        q10, d2, d28
+      vmlal.u8        q11, d3, d28
+    vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+      vrshrn.u16      d6, q8, #2
+      vrshrn.u16      d7, q9, #2
+      vshrn.u16       d8, q10, #2
+      vshrn.u16       d9, q11, #2
+      vst2.8          {d6, d7, d8, d9}, [\OUTPTR]!
+.endm
+
+/*
+ * Upsample a row of WIDTH pixels from INPTR to OUTPTR.
+ */
+.macro upsample_row OUTPTR, INPTR, WIDTH, TMP1
+    /* special case for the first and last pixels */
+    sub             \WIDTH, \WIDTH, #1
+    add             \OUTPTR, \OUTPTR, #1
+    ldrb            \TMP1, [\INPTR, \WIDTH]
+    strb            \TMP1, [\OUTPTR, \WIDTH, asl #1]
+    ldrb            \TMP1, [\INPTR], #1
+    strb            \TMP1, [\OUTPTR, #-1]
+    vmov.8          d3[7], \TMP1
+
+    subs            \WIDTH, \WIDTH, #32
+    blt             5f
+0:  /* process 32 pixels per iteration */
+    upsample32      \OUTPTR, \INPTR
+    subs            \WIDTH, \WIDTH, #32
+    bge             0b
+5:
+    adds            \WIDTH, \WIDTH, #16
+    blt             1f
+0:  /* process 16 pixels if needed */
+    upsample16      \OUTPTR, \INPTR
+    subs            \WIDTH, \WIDTH, #16
+1:
+    adds            \WIDTH, \WIDTH, #16
+    beq             9f
+
+    /* load the remaining 1-15 pixels */
+    add             \INPTR, \INPTR, \WIDTH
+    tst             \WIDTH, #1
+    beq             2f
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #2
+    beq             2f
+    vext.8          d0, d0, d0, #6
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[1]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #4
+    beq             2f
+    vrev64.32       d0, d0
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[3]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[2]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[1]}, [\INPTR]
+    sub             \INPTR, \INPTR, #1
+    vld1.8          {d0[0]}, [\INPTR]
+2:
+    tst             \WIDTH, #8
+    beq             2f
+    vmov            d1, d0
+    sub             \INPTR, \INPTR, #8
+    vld1.8          {d0}, [\INPTR]
+2:  /* upsample the remaining pixels */
+    vmovl.u8        q8, d0
+    vext.8          q2, q1, q0, #15
+    vmovl.u8        q9, d1
+    vaddw.u8        q10, q15, d4
+    vaddw.u8        q11, q15, d5
+    vmlal.u8        q8, d4, d28
+    vmlal.u8        q9, d5, d28
+    vmlal.u8        q10, d0, d28
+    vmlal.u8        q11, d1, d28
+    vrshrn.u16      d10, q8, #2
+    vrshrn.u16      d12, q9, #2
+    vshrn.u16       d11, q10, #2
+    vshrn.u16       d13, q11, #2
+    vzip.8          d10, d11
+    vzip.8          d12, d13
+    /* store the remaining pixels */
+    tst             \WIDTH, #8
+    beq             2f
+    vst1.8          {d10, d11}, [\OUTPTR]!
+    vmov            q5, q6
+2:
+    tst             \WIDTH, #4
+    beq             2f
+    vst1.8          {d10}, [\OUTPTR]!
+    vmov            d10, d11
+2:
+    tst             \WIDTH, #2
+    beq             2f
+    vst1.8          {d10[0]}, [\OUTPTR]!
+    vst1.8          {d10[1]}, [\OUTPTR]!
+    vst1.8          {d10[2]}, [\OUTPTR]!
+    vst1.8          {d10[3]}, [\OUTPTR]!
+    vext.8          d10, d10, d10, #4
+2:
+    tst             \WIDTH, #1
+    beq             2f
+    vst1.8          {d10[0]}, [\OUTPTR]!
+    vst1.8          {d10[1]}, [\OUTPTR]!
+2:
+9:
+.endm
+
+asm_function jsimd_h2v1_fancy_upsample_neon
+
+    MAX_V_SAMP_FACTOR .req r0
+    DOWNSAMPLED_WIDTH .req r1
+    INPUT_DATA        .req r2
+    OUTPUT_DATA_PTR   .req r3
+    OUTPUT_DATA       .req OUTPUT_DATA_PTR
+
+    OUTPTR            .req r4
+    INPTR             .req r5
+    WIDTH             .req ip
+    TMP               .req lr
+
+    push            {r4, r5, r6, lr}
+    vpush           {d8-d15}
+
+    ldr             OUTPUT_DATA, [OUTPUT_DATA_PTR]
+    cmp             MAX_V_SAMP_FACTOR, #0
+    ble             99f
+
+    /* initialize constants */
+    vmov.u8         d28, #3
+    vmov.u16        q15, #1
+11:
+    ldr             INPTR, [INPUT_DATA], #4
+    ldr             OUTPTR, [OUTPUT_DATA], #4
+    mov             WIDTH, DOWNSAMPLED_WIDTH
+    upsample_row    OUTPTR, INPTR, WIDTH, TMP
+    subs            MAX_V_SAMP_FACTOR, MAX_V_SAMP_FACTOR, #1
+    bgt             11b
+
+99:
+    vpop            {d8-d15}
+    pop             {r4, r5, r6, pc}
+
+    .unreq          MAX_V_SAMP_FACTOR
+    .unreq          DOWNSAMPLED_WIDTH
+    .unreq          INPUT_DATA
+    .unreq          OUTPUT_DATA_PTR
+    .unreq          OUTPUT_DATA
+
+    .unreq          OUTPTR
+    .unreq          INPTR
+    .unreq          WIDTH
+    .unreq          TMP
+
+.purgem upsample16
+.purgem upsample32
+.purgem upsample_row
+
+
+/*****************************************************************************/
+
+/*
+ * GLOBAL(JOCTET*)
+ * jsimd_huff_encode_one_block (working_state *state, JOCTET *buffer,
+ *                              JCOEFPTR block, int last_dc_val,
+ *                              c_derived_tbl *dctbl, c_derived_tbl *actbl)
+ *
+ */
+
+.macro emit_byte BUFFER, PUT_BUFFER, PUT_BITS, ZERO, TMP
+    sub             \PUT_BITS, \PUT_BITS, #0x8
+    lsr             \TMP, \PUT_BUFFER, \PUT_BITS
+    uxtb            \TMP, \TMP
+    strb            \TMP, [\BUFFER, #1]!
+    cmp             \TMP, #0xff
+    /*it eq*/
+    strbeq          \ZERO, [\BUFFER, #1]!
+.endm
+
+.macro put_bits PUT_BUFFER, PUT_BITS, CODE, SIZE
+    /*lsl             \PUT_BUFFER, \PUT_BUFFER, \SIZE*/
+    add             \PUT_BITS, \SIZE
+    /*orr             \PUT_BUFFER, \PUT_BUFFER, \CODE*/
+    orr             \PUT_BUFFER, \CODE, \PUT_BUFFER, lsl \SIZE
+.endm
+
+.macro checkbuf15 BUFFER, PUT_BUFFER, PUT_BITS, ZERO, TMP
+  cmp               \PUT_BITS, #0x10
+  blt               15f
+    eor               \ZERO, \ZERO, \ZERO
+    emit_byte         \BUFFER, \PUT_BUFFER, \PUT_BITS, \ZERO, \TMP
+    emit_byte         \BUFFER, \PUT_BUFFER, \PUT_BITS, \ZERO, \TMP
+15:
+.endm
+
+.balign 16
+jsimd_huff_encode_one_block_neon_consts:
+  .byte 0x01
+  .byte 0x02
+  .byte 0x04
+  .byte 0x08
+  .byte 0x10
+  .byte 0x20
+  .byte 0x40
+  .byte 0x80
+
+asm_function jsimd_huff_encode_one_block_neon
+    push            {r4, r5, r6, r7, r8, r9, r10, r11, lr}
+    add             r7, sp, #0x1c
+    sub             r4, sp, #0x40
+    bfc             r4, #0, #5
+    mov             sp, r4           /* align sp on 32 bytes */
+    vst1.64         {d8, d9, d10, d11}, [r4, :128]!
+    vst1.64         {d12, d13, d14, d15}, [r4, :128]
+    sub             sp, #0x140       /* reserve 320 bytes */
+    str             r0, [sp, #0x18]  /* working state > sp + Ox18 */
+    add             r4, sp, #0x20    /* r4 = t1 */
+    ldr             lr, [r7, #0x8]   /* lr = dctbl */
+    sub             r10, r1, #0x1    /* r10=buffer-- */
+    ldrsh           r1, [r2]
+    mov             r9, #0x10
+    mov             r8, #0x1
+    adr             r5, jsimd_huff_encode_one_block_neon_consts
+    /* prepare data */
+    vld1.8          {d26}, [r5, :64]
+    veor            q8, q8, q8
+    veor            q9, q9, q9
+    vdup.16         q14, r9
+    vdup.16         q15, r8
+    veor            q10, q10, q10
+    veor            q11, q11, q11
+    sub             r1, r1, r3
+    add             r9, r2, #0x22
+    add             r8, r2, #0x18
+    add             r3, r2, #0x36
+    vmov.16         d0[0], r1
+    vld1.16         {d2[0]}, [r9, :16]
+    vld1.16         {d4[0]}, [r8, :16]
+    vld1.16         {d6[0]}, [r3, :16]
+    add             r1, r2, #0x2
+    add             r9, r2, #0x30
+    add             r8, r2, #0x26
+    add             r3, r2, #0x28
+    vld1.16         {d0[1]}, [r1, :16]
+    vld1.16         {d2[1]}, [r9, :16]
+    vld1.16         {d4[1]}, [r8, :16]
+    vld1.16         {d6[1]}, [r3, :16]
+    add             r1, r2, #0x10
+    add             r9, r2, #0x40
+    add             r8, r2, #0x34
+    add             r3, r2, #0x1a
+    vld1.16         {d0[2]}, [r1, :16]
+    vld1.16         {d2[2]}, [r9, :16]
+    vld1.16         {d4[2]}, [r8, :16]
+    vld1.16         {d6[2]}, [r3, :16]
+    add             r1, r2, #0x20
+    add             r9, r2, #0x32
+    add             r8, r2, #0x42
+    add             r3, r2, #0xc
+    vld1.16         {d0[3]}, [r1, :16]
+    vld1.16         {d2[3]}, [r9, :16]
+    vld1.16         {d4[3]}, [r8, :16]
+    vld1.16         {d6[3]}, [r3, :16]
+    add             r1, r2, #0x12
+    add             r9, r2, #0x24
+    add             r8, r2, #0x50
+    add             r3, r2, #0xe
+    vld1.16         {d1[0]}, [r1, :16]
+    vld1.16         {d3[0]}, [r9, :16]
+    vld1.16         {d5[0]}, [r8, :16]
+    vld1.16         {d7[0]}, [r3, :16]
+    add             r1, r2, #0x4
+    add             r9, r2, #0x16
+    add             r8, r2, #0x60
+    add             r3, r2, #0x1c
+    vld1.16         {d1[1]}, [r1, :16]
+    vld1.16         {d3[1]}, [r9, :16]
+    vld1.16         {d5[1]}, [r8, :16]
+    vld1.16         {d7[1]}, [r3, :16]
+    add             r1, r2, #0x6
+    add             r9, r2, #0x8
+    add             r8, r2, #0x52
+    add             r3, r2, #0x2a
+    vld1.16         {d1[2]}, [r1, :16]
+    vld1.16         {d3[2]}, [r9, :16]
+    vld1.16         {d5[2]}, [r8, :16]
+    vld1.16         {d7[2]}, [r3, :16]
+    add             r1, r2, #0x14
+    add             r9, r2, #0xa
+    add             r8, r2, #0x44
+    add             r3, r2, #0x38
+    vld1.16         {d1[3]}, [r1, :16]
+    vld1.16         {d3[3]}, [r9, :16]
+    vld1.16         {d5[3]}, [r8, :16]
+    vld1.16         {d7[3]}, [r3, :16]
+    vcgt.s16        q8, q8, q0
+    vcgt.s16        q9, q9, q1
+    vcgt.s16        q10, q10, q2
+    vcgt.s16        q11, q11, q3
+    vabs.s16        q0, q0
+    vabs.s16        q1, q1
+    vabs.s16        q2, q2
+    vabs.s16        q3, q3
+    veor            q8, q8, q0
+    veor            q9, q9, q1
+    veor            q10, q10, q2
+    veor            q11, q11, q3
+    add             r9, r4, #0x20
+    add             r8, r4, #0x80
+    add             r3, r4, #0xa0
+    vclz.i16        q0, q0
+    vclz.i16        q1, q1
+    vclz.i16        q2, q2
+    vclz.i16        q3, q3
+    vsub.i16        q0, q14, q0
+    vsub.i16        q1, q14, q1
+    vsub.i16        q2, q14, q2
+    vsub.i16        q3, q14, q3
+    vst1.16         {d0, d1, d2, d3}, [r4, :256]
+    vst1.16         {d4, d5, d6, d7}, [r9, :256]
+    vshl.s16        q0, q15, q0
+    vshl.s16        q1, q15, q1
+    vshl.s16        q2, q15, q2
+    vshl.s16        q3, q15, q3
+    vsub.i16        q0, q0, q15
+    vsub.i16        q1, q1, q15
+    vsub.i16        q2, q2, q15
+    vsub.i16        q3, q3, q15
+    vand            q8, q8, q0
+    vand            q9, q9, q1
+    vand            q10, q10, q2
+    vand            q11, q11, q3
+    vst1.16         {d16, d17, d18, d19}, [r8, :256]
+    vst1.16         {d20, d21, d22, d23}, [r3, :256]
+    add             r1, r2, #0x46
+    add             r9, r2, #0x3a
+    add             r8, r2, #0x74
+    add             r3, r2, #0x6a
+    vld1.16         {d8[0]}, [r1, :16]
+    vld1.16         {d10[0]}, [r9, :16]
+    vld1.16         {d12[0]}, [r8, :16]
+    vld1.16         {d14[0]}, [r3, :16]
+    veor            q8, q8, q8
+    veor            q9, q9, q9
+    veor            q10, q10, q10
+    veor            q11, q11, q11
+    add             r1, r2, #0x54
+    add             r9, r2, #0x2c
+    add             r8, r2, #0x76
+    add             r3, r2, #0x78
+    vld1.16         {d8[1]}, [r1, :16]
+    vld1.16         {d10[1]}, [r9, :16]
+    vld1.16         {d12[1]}, [r8, :16]
+    vld1.16         {d14[1]}, [r3, :16]
+    add             r1, r2, #0x62
+    add             r9, r2, #0x1e
+    add             r8, r2, #0x68
+    add             r3, r2, #0x7a
+    vld1.16         {d8[2]}, [r1, :16]
+    vld1.16         {d10[2]}, [r9, :16]
+    vld1.16         {d12[2]}, [r8, :16]
+    vld1.16         {d14[2]}, [r3, :16]
+    add             r1, r2, #0x70
+    add             r9, r2, #0x2e
+    add             r8, r2, #0x5a
+    add             r3, r2, #0x6c
+    vld1.16         {d8[3]}, [r1, :16]
+    vld1.16         {d10[3]}, [r9, :16]
+    vld1.16         {d12[3]}, [r8, :16]
+    vld1.16         {d14[3]}, [r3, :16]
+    add             r1, r2, #0x72
+    add             r9, r2, #0x3c
+    add             r8, r2, #0x4c
+    add             r3, r2, #0x5e
+    vld1.16         {d9[0]}, [r1, :16]
+    vld1.16         {d11[0]}, [r9, :16]
+    vld1.16         {d13[0]}, [r8, :16]
+    vld1.16         {d15[0]}, [r3, :16]
+    add             r1, r2, #0x64
+    add             r9, r2, #0x4a
+    add             r8, r2, #0x3e
+    add             r3, r2, #0x6e
+    vld1.16         {d9[1]}, [r1, :16]
+    vld1.16         {d11[1]}, [r9, :16]
+    vld1.16         {d13[1]}, [r8, :16]
+    vld1.16         {d15[1]}, [r3, :16]
+    add             r1, r2, #0x56
+    add             r9, r2, #0x58
+    add             r8, r2, #0x4e
+    add             r3, r2, #0x7c
+    vld1.16         {d9[2]}, [r1, :16]
+    vld1.16         {d11[2]}, [r9, :16]
+    vld1.16         {d13[2]}, [r8, :16]
+    vld1.16         {d15[2]}, [r3, :16]
+    add             r1, r2, #0x48
+    add             r9, r2, #0x66
+    add             r8, r2, #0x5c
+    add             r3, r2, #0x7e
+    vld1.16         {d9[3]}, [r1, :16]
+    vld1.16         {d11[3]}, [r9, :16]
+    vld1.16         {d13[3]}, [r8, :16]
+    vld1.16         {d15[3]}, [r3, :16]
+    vcgt.s16        q8, q8, q4
+    vcgt.s16        q9, q9, q5
+    vcgt.s16        q10, q10, q6
+    vcgt.s16        q11, q11, q7
+    vabs.s16        q4, q4
+    vabs.s16        q5, q5
+    vabs.s16        q6, q6
+    vabs.s16        q7, q7
+    veor            q8, q8, q4
+    veor            q9, q9, q5
+    veor            q10, q10, q6
+    veor            q11, q11, q7
+    add             r1, r4, #0x40
+    add             r9, r4, #0x60
+    add             r8, r4, #0xc0
+    add             r3, r4, #0xe0
+    vclz.i16        q4, q4
+    vclz.i16        q5, q5
+    vclz.i16        q6, q6
+    vclz.i16        q7, q7
+    vsub.i16        q4, q14, q4
+    vsub.i16        q5, q14, q5
+    vsub.i16        q6, q14, q6
+    vsub.i16        q7, q14, q7
+    vst1.16         {d8, d9, d10, d11}, [r1, :256]
+    vst1.16         {d12, d13, d14, d15}, [r9, :256]
+    vshl.s16        q4, q15, q4
+    vshl.s16        q5, q15, q5
+    vshl.s16        q6, q15, q6
+    vshl.s16        q7, q15, q7
+    vsub.i16        q4, q4, q15
+    vsub.i16        q5, q5, q15
+    vsub.i16        q6, q6, q15
+    vsub.i16        q7, q7, q15
+    vand            q8, q8, q4
+    vand            q9, q9, q5
+    vand            q10, q10, q6
+    vand            q11, q11, q7
+    vst1.16         {d16, d17, d18, d19}, [r8, :256]
+    vst1.16         {d20, d21, d22, d23}, [r3, :256]
+    ldr             r12, [r7, #0xc]       /* r12 = actbl */
+    add             r1, lr, #0x400        /* r1 = dctbl->ehufsi */
+    mov             r9, r12               /* r9 = actbl */
+    add             r6, r4, #0x80         /* r6 = t2 */
+    ldr             r11, [r0, #0x8]       /* r11 = put_buffer */
+    ldr             r4, [r0, #0xc]        /* r4  = put_bits */
+    ldrh            r2, [r6, #-128]       /* r2  = nbits */
+    ldrh            r3, [r6]              /* r3  = temp2 & (((JLONG) 1)<<nbits) - 1; */
+    ldr             r0, [lr, r2, lsl #2]
+    ldrb            r5, [r1, r2]
+    put_bits        r11, r4, r0, r5
+    checkbuf15      r10, r11, r4, r5, r0
+    put_bits        r11, r4, r3, r2
+    checkbuf15      r10, r11, r4, r5, r0
+    mov             lr, r6                /* lr = t2 */
+    add             r5, r9, #0x400        /* r5 = actbl->ehufsi */
+    ldrsb           r6, [r5, #0xf0]       /* r6 = actbl->ehufsi[0xf0] */
+    veor            q8, q8, q8
+    vceq.i16        q0, q0, q8
+    vceq.i16        q1, q1, q8
+    vceq.i16        q2, q2, q8
+    vceq.i16        q3, q3, q8
+    vceq.i16        q4, q4, q8
+    vceq.i16        q5, q5, q8
+    vceq.i16        q6, q6, q8
+    vceq.i16        q7, q7, q8
+    vmovn.i16       d0, q0
+    vmovn.i16       d2, q1
+    vmovn.i16       d4, q2
+    vmovn.i16       d6, q3
+    vmovn.i16       d8, q4
+    vmovn.i16       d10, q5
+    vmovn.i16       d12, q6
+    vmovn.i16       d14, q7
+    vand            d0, d0, d26
+    vand            d2, d2, d26
+    vand            d4, d4, d26
+    vand            d6, d6, d26
+    vand            d8, d8, d26
+    vand            d10, d10, d26
+    vand            d12, d12, d26
+    vand            d14, d14, d26
+    vpadd.i8        d0, d0, d2
+    vpadd.i8        d4, d4, d6
+    vpadd.i8        d8, d8, d10
+    vpadd.i8        d12, d12, d14
+    vpadd.i8        d0, d0, d4
+    vpadd.i8        d8, d8, d12
+    vpadd.i8        d0, d0, d8
+    vmov.32         r1, d0[1]
+    vmov.32         r8, d0[0]
+    mvn             r1, r1
+    mvn             r8, r8
+    lsrs            r1, r1, #0x1
+    rrx             r8, r8            /* shift in last r1 bit while shifting out DC bit */
+    rbit            r1, r1            /* r1 = index1 */
+    rbit            r8, r8            /* r8 = index0 */
+    ldr             r0, [r9, #0x3c0]  /* r0 = actbl->ehufco[0xf0] */
+    str             r1, [sp, #0x14]   /* index1 > sp + 0x14 */
+    cmp             r8, #0x0
+    beq             6f
+1:
+    clz             r2, r8
+    add             lr, lr, r2, lsl #1
+    lsl             r8, r8, r2
+    ldrh            r1, [lr, #-126]
+2:
+    cmp             r2, #0x10
+    blt             3f
+    sub             r2, r2, #0x10
+    put_bits        r11, r4, r0, r6
+    cmp             r4, #0x10
+    blt             2b
+    eor             r3, r3, r3
+    emit_byte       r10, r11, r4, r3, r12
+    emit_byte       r10, r11, r4, r3, r12
+    b               2b
+3:
+    add             r2, r1, r2, lsl #4
+    ldrh            r3, [lr, #2]!
+    ldr             r12, [r9, r2, lsl #2]
+    ldrb            r2, [r5, r2]
+    put_bits        r11, r4, r12, r2
+    checkbuf15      r10, r11, r4, r2, r12
+    put_bits        r11, r4, r3, r1
+    checkbuf15      r10, r11, r4, r2, r12
+    lsls            r8, r8, #0x1
+    bne             1b
+6:
+    add             r12, sp, #0x20   /* r12 = t1 */
+    ldr             r8, [sp, #0x14]  /* r8 = index1 */
+    adds            r12, #0xc0       /* r12 = t2 + (DCTSIZE2/2) */
+    cmp             r8, #0x0
+    beq             6f
+    clz             r2, r8
+    sub             r12, r12, lr
+    lsl             r8, r8, r2
+    add             r2, r2, r12, lsr #1
+    add             lr, lr, r2, lsl #1
+    b               7f
+1:
+    clz             r2, r8
+    add             lr, lr, r2, lsl #1
+    lsl             r8, r8, r2
+7:
+    ldrh            r1, [lr, #-126]
+2:
+    cmp             r2, #0x10
+    blt             3f
+    sub             r2, r2, #0x10
+    put_bits        r11, r4, r0, r6
+    cmp             r4, #0x10
+    blt             2b
+    eor             r3, r3, r3
+    emit_byte       r10, r11, r4, r3, r12
+    emit_byte       r10, r11, r4, r3, r12
+    b               2b
+3:
+    add             r2, r1, r2, lsl #4
+    ldrh            r3, [lr, #2]!
+    ldr             r12, [r9, r2, lsl #2]
+    ldrb            r2, [r5, r2]
+    put_bits        r11, r4, r12, r2
+    checkbuf15      r10, r11, r4, r2, r12
+    put_bits        r11, r4, r3, r1
+    checkbuf15      r10, r11, r4, r2, r12
+    lsls            r8, r8, #0x1
+    bne             1b
+6:
+    add             r0, sp, #0x20
+    add             r0, #0xfe
+    cmp             lr, r0
+    bhs             1f
+    ldr             r1, [r9]
+    ldrb            r0, [r5]
+    put_bits        r11, r4, r1, r0
+    checkbuf15      r10, r11, r4, r0, r1
+1:
+    ldr             r12, [sp, #0x18]
+    str             r11, [r12, #0x8]
+    str             r4, [r12, #0xc]
+    add             r0, r10, #0x1
+    add             r4, sp, #0x140
+    vld1.64         {d8, d9, d10, d11}, [r4, :128]!
+    vld1.64         {d12, d13, d14, d15}, [r4, :128]
+    sub             r4, r7, #0x1c
+    mov             sp, r4
+    pop             {r4, r5, r6, r7, r8, r9, r10, r11, pc}
+
+.purgem emit_byte
+.purgem put_bits
+.purgem checkbuf15
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcfg.inc glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcfg.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcfg.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcfg.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,92 @@
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+;
+; -- jpeglib.h
+;
+%define DCTSIZE 8
+%define DCTSIZE2 64
+;
+; -- jmorecfg.h
+;
+%define RGB_RED 0
+%define RGB_GREEN 1
+%define RGB_BLUE 2
+%define RGB_PIXELSIZE 3
+%define EXT_RGB_RED 0
+%define EXT_RGB_GREEN 1
+%define EXT_RGB_BLUE 2
+%define EXT_RGB_PIXELSIZE 3
+%define EXT_RGBX_RED 0
+%define EXT_RGBX_GREEN 1
+%define EXT_RGBX_BLUE 2
+%define EXT_RGBX_PIXELSIZE 4
+%define EXT_BGR_RED 2
+%define EXT_BGR_GREEN 1
+%define EXT_BGR_BLUE 0
+%define EXT_BGR_PIXELSIZE 3
+%define EXT_BGRX_RED 2
+%define EXT_BGRX_GREEN 1
+%define EXT_BGRX_BLUE 0
+%define EXT_BGRX_PIXELSIZE 4
+%define EXT_XBGR_RED 3
+%define EXT_XBGR_GREEN 2
+%define EXT_XBGR_BLUE 1
+%define EXT_XBGR_PIXELSIZE 4
+%define EXT_XRGB_RED 1
+%define EXT_XRGB_GREEN 2
+%define EXT_XRGB_BLUE 3
+%define EXT_XRGB_PIXELSIZE 4
+%define RGBX_FILLER_0XFF 1
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+%define JSAMPLE byte ; unsigned char
+%define SIZEOF_JSAMPLE SIZEOF_BYTE ; sizeof(JSAMPLE)
+%define CENTERJSAMPLE 128
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF word ; short
+%define SIZEOF_JCOEF SIZEOF_WORD ; sizeof(JCOEF)
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION dword ; unsigned int
+%define SIZEOF_JDIMENSION SIZEOF_DWORD ; sizeof(JDIMENSION)
+%define JSAMPROW POINTER ; JSAMPLE * (jpeglib.h)
+%define JSAMPARRAY POINTER ; JSAMPROW * (jpeglib.h)
+%define JSAMPIMAGE POINTER ; JSAMPARRAY * (jpeglib.h)
+%define JCOEFPTR POINTER ; JCOEF * (jpeglib.h)
+%define SIZEOF_JSAMPROW SIZEOF_POINTER ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY SIZEOF_POINTER ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE SIZEOF_POINTER ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR SIZEOF_POINTER ; sizeof(JCOEFPTR)
+;
+; -- jdct.h
+;
+; A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM word ; short
+%define SIZEOF_DCTELEM SIZEOF_WORD ; sizeof(DCTELEM)
+%define float FP32 ; float
+%define SIZEOF_FAST_FLOAT SIZEOF_FP32 ; sizeof(float)
+; To maximize parallelism, Type short is changed to short.
+;
+%define ISLOW_MULT_TYPE word ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE SIZEOF_WORD ; sizeof(ISLOW_MULT_TYPE)
+%define IFAST_MULT_TYPE word ; must be short
+%define SIZEOF_IFAST_MULT_TYPE SIZEOF_WORD ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS 2 ; fractional bits in scale factors
+%define FLOAT_MULT_TYPE FP32 ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE SIZEOF_FP32 ; sizeof(FLOAT_MULT_TYPE)
+;
+; -- jsimd.h
+;
+%define JSIMD_NONE 0x00
+%define JSIMD_MMX 0x01
+%define JSIMD_3DNOW 0x02
+%define JSIMD_SSE 0x04
+%define JSIMD_SSE2 0x08
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcfg.inc.h glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcfg.inc.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcfg.inc.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcfg.inc.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,130 @@
+// This file generates the include file for the assembly
+// implementations by abusing the C preprocessor.
+//
+// Note: Some things are manually defined as they need to
+// be mapped to NASM types.
+
+;
+; Automatically generated include file from jsimdcfg.inc.h
+;
+
+#define JPEG_INTERNALS
+
+#include "../jpeglib.h"
+#include "../jconfig.h"
+#include "../jmorecfg.h"
+#include "jsimd.h"
+
+;
+; -- jpeglib.h
+;
+
+%define _cpp_protection_DCTSIZE DCTSIZE
+%define _cpp_protection_DCTSIZE2 DCTSIZE2
+
+;
+; -- jmorecfg.h
+;
+
+%define _cpp_protection_RGB_RED RGB_RED
+%define _cpp_protection_RGB_GREEN RGB_GREEN
+%define _cpp_protection_RGB_BLUE RGB_BLUE
+%define _cpp_protection_RGB_PIXELSIZE RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGB_RED EXT_RGB_RED
+%define _cpp_protection_EXT_RGB_GREEN EXT_RGB_GREEN
+%define _cpp_protection_EXT_RGB_BLUE EXT_RGB_BLUE
+%define _cpp_protection_EXT_RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+
+%define _cpp_protection_EXT_RGBX_RED EXT_RGBX_RED
+%define _cpp_protection_EXT_RGBX_GREEN EXT_RGBX_GREEN
+%define _cpp_protection_EXT_RGBX_BLUE EXT_RGBX_BLUE
+%define _cpp_protection_EXT_RGBX_PIXELSIZE EXT_RGBX_PIXELSIZE
+
+%define _cpp_protection_EXT_BGR_RED EXT_BGR_RED
+%define _cpp_protection_EXT_BGR_GREEN EXT_BGR_GREEN
+%define _cpp_protection_EXT_BGR_BLUE EXT_BGR_BLUE
+%define _cpp_protection_EXT_BGR_PIXELSIZE EXT_BGR_PIXELSIZE
+
+%define _cpp_protection_EXT_BGRX_RED EXT_BGRX_RED
+%define _cpp_protection_EXT_BGRX_GREEN EXT_BGRX_GREEN
+%define _cpp_protection_EXT_BGRX_BLUE EXT_BGRX_BLUE
+%define _cpp_protection_EXT_BGRX_PIXELSIZE EXT_BGRX_PIXELSIZE
+
+%define _cpp_protection_EXT_XBGR_RED EXT_XBGR_RED
+%define _cpp_protection_EXT_XBGR_GREEN EXT_XBGR_GREEN
+%define _cpp_protection_EXT_XBGR_BLUE EXT_XBGR_BLUE
+%define _cpp_protection_EXT_XBGR_PIXELSIZE EXT_XBGR_PIXELSIZE
+
+%define _cpp_protection_EXT_XRGB_RED EXT_XRGB_RED
+%define _cpp_protection_EXT_XRGB_GREEN EXT_XRGB_GREEN
+%define _cpp_protection_EXT_XRGB_BLUE EXT_XRGB_BLUE
+%define _cpp_protection_EXT_XRGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+
+%define RGBX_FILLER_0XFF        1
+
+; Representation of a single sample (pixel element value).
+; On this SIMD implementation, this must be 'unsigned char'.
+;
+
+%define JSAMPLE                 byte          ; unsigned char
+%define SIZEOF_JSAMPLE          SIZEOF_BYTE   ; sizeof(JSAMPLE)
+
+%define _cpp_protection_CENTERJSAMPLE CENTERJSAMPLE
+
+; Representation of a DCT frequency coefficient.
+; On this SIMD implementation, this must be 'short'.
+;
+%define JCOEF                   word          ; short
+%define SIZEOF_JCOEF            SIZEOF_WORD   ; sizeof(JCOEF)
+
+; Datatype used for image dimensions.
+; On this SIMD implementation, this must be 'unsigned int'.
+;
+%define JDIMENSION              dword         ; unsigned int
+%define SIZEOF_JDIMENSION       SIZEOF_DWORD  ; sizeof(JDIMENSION)
+
+%define JSAMPROW                POINTER       ; JSAMPLE *     (jpeglib.h)
+%define JSAMPARRAY              POINTER       ; JSAMPROW *    (jpeglib.h)
+%define JSAMPIMAGE              POINTER       ; JSAMPARRAY *  (jpeglib.h)
+%define JCOEFPTR                POINTER       ; JCOEF *       (jpeglib.h)
+%define SIZEOF_JSAMPROW         SIZEOF_POINTER  ; sizeof(JSAMPROW)
+%define SIZEOF_JSAMPARRAY       SIZEOF_POINTER  ; sizeof(JSAMPARRAY)
+%define SIZEOF_JSAMPIMAGE       SIZEOF_POINTER  ; sizeof(JSAMPIMAGE)
+%define SIZEOF_JCOEFPTR         SIZEOF_POINTER  ; sizeof(JCOEFPTR)
+
+;
+; -- jdct.h
+;
+
+; A forward DCT routine is given a pointer to a work area of type DCTELEM[];
+; the DCT is to be performed in-place in that buffer.
+; To maximize parallelism, Type DCTELEM is changed to short (originally, int).
+;
+%define DCTELEM                 word          ; short
+%define SIZEOF_DCTELEM          SIZEOF_WORD   ; sizeof(DCTELEM)
+
+%define FAST_FLOAT              FP32            ; float
+%define SIZEOF_FAST_FLOAT       SIZEOF_FP32     ; sizeof(FAST_FLOAT)
+
+; To maximize parallelism, Type MULTIPLIER is changed to short.
+;
+%define ISLOW_MULT_TYPE         word          ; must be short
+%define SIZEOF_ISLOW_MULT_TYPE  SIZEOF_WORD   ; sizeof(ISLOW_MULT_TYPE)
+
+%define IFAST_MULT_TYPE         word          ; must be short
+%define SIZEOF_IFAST_MULT_TYPE  SIZEOF_WORD   ; sizeof(IFAST_MULT_TYPE)
+%define IFAST_SCALE_BITS        2             ; fractional bits in scale factors
+
+%define FLOAT_MULT_TYPE         FP32          ; must be float
+%define SIZEOF_FLOAT_MULT_TYPE  SIZEOF_FP32   ; sizeof(FLOAT_MULT_TYPE)
+
+;
+; -- jsimd.h
+;
+
+%define _cpp_protection_JSIMD_NONE JSIMD_NONE
+%define _cpp_protection_JSIMD_MMX JSIMD_MMX
+%define _cpp_protection_JSIMD_3DNOW JSIMD_3DNOW
+%define _cpp_protection_JSIMD_SSE JSIMD_SSE
+%define _cpp_protection_JSIMD_SSE2 JSIMD_SSE2
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcpu.asm glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcpu.asm
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdcpu.asm	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimdcpu.asm	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,105 @@
+;
+; jsimdcpu.asm - SIMD instruction support check
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+; For conditions of distribution and use, see copyright notice in jsimdext.inc
+;
+; This file should be assembled with NASM (Netwide Assembler),
+; can *not* be assembled with Microsoft's MASM or any compatible
+; assembler (including Borland's Turbo Assembler).
+; NASM is available from http://nasm.sourceforge.net/ or
+; http://sourceforge.net/project/showfiles.php?group_id=6208
+;
+; [TAB8]
+
+%include "jsimdext.inc"
+
+; --------------------------------------------------------------------------
+        SECTION SEG_TEXT
+        BITS    32
+;
+; Check if the CPU supports SIMD instructions
+;
+; GLOBAL(unsigned int)
+; jpeg_simd_cpu_support (void)
+;
+
+        align   16
+        global  EXTN(jpeg_simd_cpu_support) PRIVATE
+
+EXTN(jpeg_simd_cpu_support):
+        push    ebx
+;       push    ecx             ; need not be preserved
+;       push    edx             ; need not be preserved
+;       push    esi             ; unused
+        push    edi
+
+        xor     edi,edi                 ; simd support flag
+
+        pushfd
+        pop     eax
+        mov     edx,eax
+        xor     eax, 1<<21              ; flip ID bit in EFLAGS
+        push    eax
+        popfd
+        pushfd
+        pop     eax
+        xor     eax,edx
+        jz      short .return           ; CPUID is not supported
+
+        ; Check for MMX instruction support
+        xor     eax,eax
+        cpuid
+        test    eax,eax
+        jz      short .return
+
+        xor     eax,eax
+        inc     eax
+        cpuid
+        mov     eax,edx                 ; eax = Standard feature flags
+
+        test    eax, 1<<23              ; bit23:MMX
+        jz      short .no_mmx
+        or      edi, byte JSIMD_MMX
+.no_mmx:
+        test    eax, 1<<25              ; bit25:SSE
+        jz      short .no_sse
+        or      edi, byte JSIMD_SSE
+.no_sse:
+        test    eax, 1<<26              ; bit26:SSE2
+        jz      short .no_sse2
+        or      edi, byte JSIMD_SSE2
+.no_sse2:
+
+        ; Check for 3DNow! instruction support
+        mov     eax, 0x80000000
+        cpuid
+        cmp     eax, 0x80000000
+        jbe     short .return
+
+        mov     eax, 0x80000001
+        cpuid
+        mov     eax,edx                 ; eax = Extended feature flags
+
+        test    eax, 1<<31              ; bit31:3DNow!(vendor independent)
+        jz      short .no_3dnow
+        or      edi, byte JSIMD_3DNOW
+.no_3dnow:
+
+.return:
+        mov     eax,edi
+
+        pop     edi
+;       pop     esi             ; unused
+;       pop     edx             ; need not be preserved
+;       pop     ecx             ; need not be preserved
+        pop     ebx
+        ret
+
+; For some reason, the OS X linker does not honor the request to align the
+; segment unless we do this.
+        align   16
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdext.inc glmark2-2021.02/src/libjpeg-turbo/simd/jsimdext.inc
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimdext.inc	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimdext.inc	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,386 @@
+;
+; jsimdext.inc - common declarations
+;
+; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+; Copyright 2010 D. R. Commander
+;
+; Based on
+; x86 SIMD extension for IJG JPEG library - version 1.02
+;
+; Copyright (C) 1999-2006, MIYASAKA Masaru.
+;
+; This software is provided 'as-is', without any express or implied
+; warranty.  In no event will the authors be held liable for any damages
+; arising from the use of this software.
+;
+; Permission is granted to anyone to use this software for any purpose,
+; including commercial applications, and to alter it and redistribute it
+; freely, subject to the following restrictions:
+;
+; 1. The origin of this software must not be misrepresented; you must not
+;    claim that you wrote the original software. If you use this software
+;    in a product, an acknowledgment in the product documentation would be
+;    appreciated but is not required.
+; 2. Altered source versions must be plainly marked as such, and must not be
+;    misrepresented as being the original software.
+; 3. This notice may not be removed or altered from any source distribution.
+;
+; [TAB8]
+
+; ==========================================================================
+;  System-dependent configurations
+
+%ifdef WIN32    ; ----(nasm -fwin32 -DWIN32 ...)--------
+; * Microsoft Visual C++
+; * MinGW (Minimalist GNU for Windows)
+; * CygWin
+; * LCC-Win32
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
+%define SEG_TEXT    .text  align=16 public use32 class=CODE
+%define SEG_CONST   .rdata align=16 public use32 class=CONST
+%endif
+
+%elifdef WIN64  ; ----(nasm -fwin64 -DWIN64 ...)--------
+; * Microsoft Visual C++
+
+; -- segment definition --
+;
+%ifdef __YASM_VER__
+%define SEG_TEXT    .text  align=16
+%define SEG_CONST   .rdata align=16
+%else
+%define SEG_TEXT    .text  align=16 public use64 class=CODE
+%define SEG_CONST   .rdata align=16 public use64 class=CONST
+%endif
+%define EXTN(name)  name                        ; foo() -> foo
+
+%elifdef OBJ32  ; ----(nasm -fobj -DOBJ32 ...)----------
+; * Borland C++ (Win32)
+
+; -- segment definition --
+;
+%define SEG_TEXT    _text  align=16 public use32 class=CODE
+%define SEG_CONST   _data  align=16 public use32 class=DATA
+
+%elifdef ELF    ; ----(nasm -felf[64] -DELF ...)------------
+; * Linux
+; * *BSD family Unix using elf format
+; * Unix System V, including Solaris x86, UnixWare and SCO Unix
+
+; mark stack as non-executable
+section .note.GNU-stack noalloc noexec nowrite progbits
+
+; -- segment definition --
+;
+%ifdef __x86_64__
+%define SEG_TEXT    .text   progbits align=16
+%define SEG_CONST   .rodata progbits align=16
+%else
+%define SEG_TEXT    .text   progbits alloc exec   nowrite align=16
+%define SEG_CONST   .rodata progbits alloc noexec nowrite align=16
+%endif
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  _GLOBAL_OFFSET_TABLE_       ; ELF supports PIC
+%define EXTN(name)  name                        ; foo() -> foo
+
+%elifdef AOUT   ; ----(nasm -faoutb/aout -DAOUT ...)----
+; * Older Linux using a.out format  (nasm -f aout -DAOUT ...)
+; * *BSD family Unix using a.out format  (nasm -f aoutb -DAOUT ...)
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text
+%define SEG_CONST   .data
+
+; To make the code position-independent, append -DPIC to the commandline
+;
+%define GOT_SYMBOL  __GLOBAL_OFFSET_TABLE_      ; BSD-style a.out supports PIC
+
+%elifdef MACHO  ; ----(nasm -fmacho -DMACHO ...)--------
+; * NeXTstep/OpenStep/Rhapsody/Darwin/MacOS X (Mach-O format)
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text  ;align=16    ; nasm doesn't accept align=16. why?
+%define SEG_CONST   .rodata align=16
+
+; The generation of position-independent code (PIC) is the default on Darwin.
+;
+%define PIC
+%define GOT_SYMBOL  _MACHO_PIC_         ; Mach-O style code-relative addressing
+
+%else           ; ----(Other case)----------------------
+
+; -- segment definition --
+;
+%define SEG_TEXT    .text
+%define SEG_CONST   .data
+
+%endif  ; ----------------------------------------------
+
+; ==========================================================================
+
+; --------------------------------------------------------------------------
+;  Common types
+;
+%ifdef __x86_64__
+%define POINTER                 qword           ; general pointer type
+%define SIZEOF_POINTER          SIZEOF_QWORD    ; sizeof(POINTER)
+%define POINTER_BIT             QWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%else
+%define POINTER                 dword           ; general pointer type
+%define SIZEOF_POINTER          SIZEOF_DWORD    ; sizeof(POINTER)
+%define POINTER_BIT             DWORD_BIT       ; sizeof(POINTER)*BYTE_BIT
+%endif
+
+%define INT                     dword           ; signed integer type
+%define SIZEOF_INT              SIZEOF_DWORD    ; sizeof(INT)
+%define INT_BIT                 DWORD_BIT       ; sizeof(INT)*BYTE_BIT
+
+%define FP32                    dword           ; IEEE754 single
+%define SIZEOF_FP32             SIZEOF_DWORD    ; sizeof(FP32)
+%define FP32_BIT                DWORD_BIT       ; sizeof(FP32)*BYTE_BIT
+
+%define MMWORD                  qword           ; int64  (MMX register)
+%define SIZEOF_MMWORD           SIZEOF_QWORD    ; sizeof(MMWORD)
+%define MMWORD_BIT              QWORD_BIT       ; sizeof(MMWORD)*BYTE_BIT
+
+; NASM is buggy and doesn't properly handle operand sizes for SSE
+; instructions, so for now we have to define XMMWORD as blank.
+%define XMMWORD                                 ; int128 (SSE register)
+%define SIZEOF_XMMWORD          SIZEOF_OWORD    ; sizeof(XMMWORD)
+%define XMMWORD_BIT             OWORD_BIT       ; sizeof(XMMWORD)*BYTE_BIT
+
+; Similar hacks for when we load a dword or MMWORD into an xmm# register
+%define XMM_DWORD
+%define XMM_MMWORD
+
+%define SIZEOF_BYTE             1               ; sizeof(BYTE)
+%define SIZEOF_WORD             2               ; sizeof(WORD)
+%define SIZEOF_DWORD            4               ; sizeof(DWORD)
+%define SIZEOF_QWORD            8               ; sizeof(QWORD)
+%define SIZEOF_OWORD            16              ; sizeof(OWORD)
+
+%define BYTE_BIT                8               ; CHAR_BIT in C
+%define WORD_BIT                16              ; sizeof(WORD)*BYTE_BIT
+%define DWORD_BIT               32              ; sizeof(DWORD)*BYTE_BIT
+%define QWORD_BIT               64              ; sizeof(QWORD)*BYTE_BIT
+%define OWORD_BIT               128             ; sizeof(OWORD)*BYTE_BIT
+
+; --------------------------------------------------------------------------
+;  External Symbol Name
+;
+%ifndef EXTN
+%define EXTN(name)   _ %+ name          ; foo() -> _foo
+%endif
+
+; --------------------------------------------------------------------------
+;  Macros for position-independent code (PIC) support
+;
+%ifndef GOT_SYMBOL
+%undef PIC
+%endif
+
+%ifdef PIC ; -------------------------------------------
+
+%ifidn GOT_SYMBOL,_MACHO_PIC_ ; --------------------
+
+; At present, nasm doesn't seem to support PIC generation for Mach-O.
+; The PIC support code below is a little tricky.
+
+        SECTION SEG_CONST
+const_base:
+
+%define GOTOFF(got,sym) (got) + (sym) - const_base
+
+%imacro get_GOT 1
+        ; NOTE: this macro destroys ecx resister.
+        call    %%geteip
+        add     ecx, byte (%%ref - $)
+        jmp     short %%adjust
+%%geteip:
+        mov     ecx, POINTER [esp]
+        ret
+%%adjust:
+        push    ebp
+        xor     ebp,ebp         ; ebp = 0
+%ifidni %1,ebx  ; (%1 == ebx)
+        ; db 0x8D,0x9C + jmp near const_base =
+        ;   lea ebx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,9C,E9,(offset32)
+        db      0x8D,0x9C               ; 8D,9C
+        jmp     near const_base         ; E9,(const_base-%%ref)
+%%ref:
+%else  ; (%1 != ebx)
+        ; db 0x8D,0x8C + jmp near const_base =
+        ;   lea ecx, [ecx+ebp*8+(const_base-%%ref)] ; 8D,8C,E9,(offset32)
+        db      0x8D,0x8C               ; 8D,8C
+        jmp     near const_base         ; E9,(const_base-%%ref)
+%%ref:  mov     %1, ecx
+%endif ; (%1 == ebx)
+        pop     ebp
+%endmacro
+
+%else   ; GOT_SYMBOL != _MACHO_PIC_ ----------------
+
+%define GOTOFF(got,sym) (got) + (sym) wrt ..gotoff
+
+%imacro get_GOT 1
+        extern  GOT_SYMBOL
+        call    %%geteip
+        add     %1, GOT_SYMBOL + $$ - $ wrt ..gotpc
+        jmp     short %%done
+%%geteip:
+        mov     %1, POINTER [esp]
+        ret
+%%done:
+%endmacro
+
+%endif  ; GOT_SYMBOL == _MACHO_PIC_ ----------------
+
+%imacro pushpic 1.nolist
+        push    %1
+%endmacro
+%imacro poppic  1.nolist
+        pop     %1
+%endmacro
+%imacro movpic  2.nolist
+        mov     %1,%2
+%endmacro
+
+%else   ; !PIC -----------------------------------------
+
+%define GOTOFF(got,sym) (sym)
+
+%imacro get_GOT 1.nolist
+%endmacro
+%imacro pushpic 1.nolist
+%endmacro
+%imacro poppic  1.nolist
+%endmacro
+%imacro movpic  2.nolist
+%endmacro
+
+%endif  ;  PIC -----------------------------------------
+
+; --------------------------------------------------------------------------
+;  Align the next instruction on {2,4,8,16,..}-byte boundary.
+;  ".balign n,,m" in GNU as
+;
+%define MSKLE(x,y)  (~(((y) & 0xFFFF) - ((x) & 0xFFFF)) >> 16)
+%define FILLB(b,n)  (($$-(b)) & ((n)-1))
+
+%imacro alignx 1-2.nolist 0xFFFF
+%%bs:   times MSKLE(FILLB(%%bs,%1),%2) & MSKLE(16,FILLB($,%1)) & FILLB($,%1) \
+               db 0x90                               ; nop
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/9 \
+               db 0x8D,0x9C,0x23,0x00,0x00,0x00,0x00 ; lea ebx,[ebx+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/7 \
+               db 0x8D,0xAC,0x25,0x00,0x00,0x00,0x00 ; lea ebp,[ebp+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/6 \
+               db 0x8D,0xAD,0x00,0x00,0x00,0x00      ; lea ebp,[ebp+0x00000000]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/4 \
+               db 0x8D,0x6C,0x25,0x00                ; lea ebp,[ebp+0x00]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/3 \
+               db 0x8D,0x6D,0x00                     ; lea ebp,[ebp+0x00]
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/2 \
+               db 0x8B,0xED                          ; mov ebp,ebp
+        times MSKLE(FILLB(%%bs,%1),%2) & FILLB($,%1)/1 \
+               db 0x90                               ; nop
+%endmacro
+
+; Align the next data on {2,4,8,16,..}-byte boundary.
+;
+%imacro alignz 1.nolist
+        align %1, db 0          ; filling zeros
+%endmacro
+
+%ifdef __x86_64__
+
+%ifdef WIN64
+
+%imacro collect_args 0
+        push r12
+        push r13
+        push r14
+        push r15
+        mov r10, rcx
+        mov r11, rdx
+        mov r12, r8
+        mov r13, r9
+        mov r14, [rax+48]
+        mov r15, [rax+56]
+        push rsi
+        push rdi
+        sub     rsp, SIZEOF_XMMWORD
+        movaps  XMMWORD [rsp], xmm6
+        sub     rsp, SIZEOF_XMMWORD
+        movaps  XMMWORD [rsp], xmm7
+%endmacro
+
+%imacro uncollect_args 0
+        movaps  xmm7, XMMWORD [rsp]
+        add     rsp, SIZEOF_XMMWORD
+        movaps  xmm6, XMMWORD [rsp]
+        add     rsp, SIZEOF_XMMWORD
+        pop rdi
+        pop rsi
+        pop r15
+        pop r14
+        pop r13
+        pop r12
+%endmacro
+
+%else
+
+%imacro collect_args 0
+        push r10
+        push r11
+        push r12
+        push r13
+        push r14
+        push r15
+        mov r10, rdi
+        mov r11, rsi
+        mov r12, rdx
+        mov r13, rcx
+        mov r14, r8
+        mov r15, r9
+%endmacro
+
+%imacro uncollect_args 0
+        pop r15
+        pop r14
+        pop r13
+        pop r12
+        pop r11
+        pop r10
+%endmacro
+
+%endif
+
+%endif
+
+; Begin chromium edits
+%ifdef MACHO ; ----(nasm -fmacho -DMACHO ...)--------
+%define PRIVATE :private_extern
+%elifdef ELF ; ----(nasm -felf[64] -DELF ...)------------
+%define PRIVATE :hidden
+%else
+%define PRIVATE
+%endif
+; End chromium edits
+
+; --------------------------------------------------------------------------
+;  Defines picked up from the C headers
+;
+%include "jsimdcfg.inc"
+
+; --------------------------------------------------------------------------
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd.h glmark2-2021.02/src/libjpeg-turbo/simd/jsimd.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd.h	2021-04-25 01:47:22.000000000 +0800
@@ -2,8 +2,11 @@
  * simd/jsimd.h
  *
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2011 D. R. Commander
- * 
+ * Copyright (C) 2011, 2014-2016 D. R. Commander
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California
+ * Copyright (C) 2014 Linaro Limited
+ * Copyright (C) 2015-2016 Matthieu Darbois
+ *
  * Based on the x86 SIMD extension for IJG JPEG library,
  * Copyright (C) 1999-2006, MIYASAKA Masaru.
  * For conditions of distribution and use, see copyright notice in jsimdext.inc
@@ -18,649 +21,851 @@
 #define JSIMD_SSE        0x04
 #define JSIMD_SSE2       0x08
 #define JSIMD_ARM_NEON   0x10
-
-/* Short forms of external names for systems with brain-damaged linkers. */
-
-#ifdef NEED_SHORT_EXTERNAL_NAMES
-#define jpeg_simd_cpu_support                 jSiCpuSupport
-#define jsimd_rgb_ycc_convert_mmx             jSRGBYCCM
-#define jsimd_extrgb_ycc_convert_mmx          jSEXTRGBYCCM
-#define jsimd_extrgbx_ycc_convert_mmx         jSEXTRGBXYCCM
-#define jsimd_extbgr_ycc_convert_mmx          jSEXTBGRYCCM
-#define jsimd_extbgrx_ycc_convert_mmx         jSEXTBGRXYCCM
-#define jsimd_extxbgr_ycc_convert_mmx         jSEXTXBGRYCCM
-#define jsimd_extxrgb_ycc_convert_mmx         jSEXTXRGBYCCM
-#define jsimd_rgb_gray_convert_mmx            jSRGBGRYM
-#define jsimd_extrgb_gray_convert_mmx         jSEXTRGBGRYM
-#define jsimd_extrgbx_gray_convert_mmx        jSEXTRGBXGRYM
-#define jsimd_extbgr_gray_convert_mmx         jSEXTBGRGRYM
-#define jsimd_extbgrx_gray_convert_mmx        jSEXTBGRXGRYM
-#define jsimd_extxbgr_gray_convert_mmx        jSEXTXBGRGRYM
-#define jsimd_extxrgb_gray_convert_mmx        jSEXTXRGBGRYM
-#define jsimd_ycc_rgb_convert_mmx             jSYCCRGBM
-#define jsimd_ycc_extrgb_convert_mmx          jSYCCEXTRGBM
-#define jsimd_ycc_extrgbx_convert_mmx         jSYCCEXTRGBXM
-#define jsimd_ycc_extbgr_convert_mmx          jSYCCEXTBGRM
-#define jsimd_ycc_extbgrx_convert_mmx         jSYCCEXTBGRXM
-#define jsimd_ycc_extxbgr_convert_mmx         jSYCCEXTXBGRM
-#define jsimd_ycc_extxrgb_convert_mmx         jSYCCEXTXRGBM
-#define jconst_rgb_ycc_convert_sse2           jSCRGBYCCS2
-#define jsimd_rgb_ycc_convert_sse2            jSRGBYCCS2
-#define jsimd_extrgb_ycc_convert_sse2         jSEXTRGBYCCS2
-#define jsimd_extrgbx_ycc_convert_sse2        jSEXTRGBXYCCS2
-#define jsimd_extbgr_ycc_convert_sse2         jSEXTBGRYCCS2
-#define jsimd_extbgrx_ycc_convert_sse2        jSEXTBGRXYCCS2
-#define jsimd_extxbgr_ycc_convert_sse2        jSEXTXBGRYCCS2
-#define jsimd_extxrgb_ycc_convert_sse2        jSEXTXRGBYCCS2
-#define jconst_rgb_gray_convert_sse2          jSCRGBGRYS2
-#define jsimd_rgb_gray_convert_sse2           jSRGBGRYS2
-#define jsimd_extrgb_gray_convert_sse2        jSEXTRGBGRYS2
-#define jsimd_extrgbx_gray_convert_sse2       jSEXTRGBXGRYS2
-#define jsimd_extbgr_gray_convert_sse2        jSEXTBGRGRYS2
-#define jsimd_extbgrx_gray_convert_sse2       jSEXTBGRXGRYS2
-#define jsimd_extxbgr_gray_convert_sse2       jSEXTXBGRGRYS2
-#define jsimd_extxrgb_gray_convert_sse2       jSEXTXRGBGRYS2
-#define jconst_ycc_rgb_convert_sse2           jSCYCCRGBS2
-#define jsimd_ycc_rgb_convert_sse2            jSYCCRGBS2
-#define jsimd_ycc_extrgb_convert_sse2         jSYCCEXTRGBS2
-#define jsimd_ycc_extrgbx_convert_sse2        jSYCCEXTRGBXS2
-#define jsimd_ycc_extbgr_convert_sse2         jSYCCEXTBGRS2
-#define jsimd_ycc_extbgrx_convert_sse2        jSYCCEXTBGRXS2
-#define jsimd_ycc_extxbgr_convert_sse2        jSYCCEXTXBGRS2
-#define jsimd_ycc_extxrgb_convert_sse2        jSYCCEXTXRGBS2
-#define jsimd_h2v2_downsample_mmx             jSDnH2V2M
-#define jsimd_h2v1_downsample_mmx             jSDnH2V1M
-#define jsimd_h2v2_downsample_sse2            jSDnH2V2S2
-#define jsimd_h2v1_downsample_sse2            jSDnH2V1S2
-#define jsimd_h2v2_upsample_mmx               jSUpH2V2M
-#define jsimd_h2v1_upsample_mmx               jSUpH2V1M
-#define jsimd_h2v2_fancy_upsample_mmx         jSFUpH2V2M
-#define jsimd_h2v1_fancy_upsample_mmx         jSFUpH2V1M
-#define jsimd_h2v2_merged_upsample_mmx        jSMUpH2V2M
-#define jsimd_h2v2_extrgb_merged_upsample_mmx jSMUpH2V2EXTRGBM
-#define jsimd_h2v2_extrgbx_merged_upsample_mmx jSMUpH2V2EXTRGBXM
-#define jsimd_h2v2_extbgr_merged_upsample_mmx jSMUpH2V2EXTBGRM
-#define jsimd_h2v2_extbgrx_merged_upsample_mmx jSMUpH2V2EXTBGRXM
-#define jsimd_h2v2_extxbgr_merged_upsample_mmx jSMUpH2V2EXTXBGRM
-#define jsimd_h2v2_extxrgb_merged_upsample_mmx jSMUpH2V2EXTXRGBM
-#define jsimd_h2v1_merged_upsample_mmx        jSMUpH2V1M
-#define jsimd_h2v1_extrgb_merged_upsample_mmx jSMUpH2V1EXTRGBM
-#define jsimd_h2v1_extrgbx_merged_upsample_mmx jSMUpH2V1EXTRGBXM
-#define jsimd_h2v1_extbgr_merged_upsample_mmx jSMUpH2V1EXTBGRM
-#define jsimd_h2v1_extbgrx_merged_upsample_mmx jSMUpH2V1EXTBGRXM
-#define jsimd_h2v1_extxbgr_merged_upsample_mmx jSMUpH2V1EXTXBGRM
-#define jsimd_h2v1_extxrgb_merged_upsample_mmx jSMUpH2V1EXTXRGBM
-#define jsimd_h2v2_upsample_sse2              jSUpH2V2S2
-#define jsimd_h2v1_upsample_sse2              jSUpH2V1S2
-#define jconst_fancy_upsample_sse2            jSCFUpS2
-#define jsimd_h2v2_fancy_upsample_sse2        jSFUpH2V2S2
-#define jsimd_h2v1_fancy_upsample_sse2        jSFUpH2V1S2
-#define jconst_merged_upsample_sse2           jSCMUpS2
-#define jsimd_h2v2_merged_upsample_sse2       jSMUpH2V2S2
-#define jsimd_h2v2_extrgb_merged_upsample_sse2 jSMUpH2V2EXTRGBS2
-#define jsimd_h2v2_extrgbx_merged_upsample_sse2 jSMUpH2V2EXTRGBXS2
-#define jsimd_h2v2_extbgr_merged_upsample_sse2 jSMUpH2V2EXTBGRS2
-#define jsimd_h2v2_extbgrx_merged_upsample_sse2 jSMUpH2V2EXTBGRXS2
-#define jsimd_h2v2_extxbgr_merged_upsample_sse2 jSMUpH2V2EXTXBGRS2
-#define jsimd_h2v2_extxrgb_merged_upsample_sse2 jSMUpH2V2EXTXRGBS2
-#define jsimd_h2v1_merged_upsample_sse2       jSMUpH2V1S2
-#define jsimd_h2v1_extrgb_merged_upsample_sse2 jSMUpH2V1EXTRGBS2
-#define jsimd_h2v1_extrgbx_merged_upsample_sse2 jSMUpH2V1EXTRGBXS2
-#define jsimd_h2v1_extbgr_merged_upsample_sse2 jSMUpH2V1EXTBGRS2
-#define jsimd_h2v1_extbgrx_merged_upsample_sse2 jSMUpH2V1EXTBGRXS2
-#define jsimd_h2v1_extxbgr_merged_upsample_sse2 jSMUpH2V1EXTXBGRS2
-#define jsimd_h2v1_extxrgb_merged_upsample_sse2 jSMUpH2V1EXTXRGBS2
-#define jsimd_convsamp_mmx                    jSConvM
-#define jsimd_convsamp_sse2                   jSConvS2
-#define jsimd_convsamp_float_3dnow            jSConvF3D
-#define jsimd_convsamp_float_sse              jSConvFS
-#define jsimd_convsamp_float_sse2             jSConvFS2
-#define jsimd_fdct_islow_mmx                  jSFDMIS
-#define jsimd_fdct_ifast_mmx                  jSFDMIF
-#define jconst_fdct_islow_sse2                jSCFDS2IS
-#define jsimd_fdct_islow_sse2                 jSFDS2IS
-#define jconst_fdct_ifast_sse2                jSCFDS2IF
-#define jsimd_fdct_ifast_sse2                 jSFDS2IF
-#define jsimd_fdct_float_3dnow                jSFD3DF
-#define jconst_fdct_float_sse                 jSCFDSF
-#define jsimd_fdct_float_sse                  jSFDSF
-#define jsimd_quantize_mmx                    jSQuantM
-#define jsimd_quantize_sse2                   jSQuantS2
-#define jsimd_quantize_float_3dnow            jSQuantF3D
-#define jsimd_quantize_float_sse              jSQuantFS
-#define jsimd_quantize_float_sse2             jSQuantFS2
-#define jsimd_idct_2x2_mmx                    jSIDM22
-#define jsimd_idct_4x4_mmx                    jSIDM44
-#define jconst_idct_red_sse2                  jSCIDS2R
-#define jsimd_idct_2x2_sse2                   jSIDS222
-#define jsimd_idct_4x4_sse2                   jSIDS244
-#define jsimd_idct_islow_mmx                  jSIDMIS
-#define jsimd_idct_ifast_mmx                  jSIDMIF
-#define jconst_idct_islow_sse2                jSCIDS2IS
-#define jsimd_idct_islow_sse2                 jSIDS2IS
-#define jconst_idct_ifast_sse2                jSCIDS2IF
-#define jsimd_idct_ifast_sse2                 jSIDS2IF
-#define jsimd_idct_float_3dnow                jSID3DF
-#define jconst_fdct_float_sse                 jSCIDSF
-#define jsimd_idct_float_sse                  jSIDSF
-#define jconst_fdct_float_sse2                jSCIDS2F
-#define jsimd_idct_float_sse2                 jSIDS2F
-#endif /* NEED_SHORT_EXTERNAL_NAMES */
+#define JSIMD_MIPS_DSPR2 0x20
+#define JSIMD_ALTIVEC    0x40
 
 /* SIMD Ext: retrieve SIMD/CPU information */
-EXTERN(unsigned int) jpeg_simd_cpu_support JPP((void));
+EXTERN(unsigned int) jpeg_simd_cpu_support (void);
 
-/* SIMD Color Space Conversion */
+/* RGB & extended RGB --> YCC Colorspace Conversion */
 EXTERN(void) jsimd_rgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_ycc_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-
-EXTERN(void) jsimd_rgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgbx_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgr_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgrx_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxbgr_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxrgb_gray_convert_mmx
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-
-EXTERN(void) jsimd_ycc_rgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extrgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extrgbx_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extbgr_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extbgrx_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extxbgr_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-EXTERN(void) jsimd_ycc_extxrgb_convert_mmx
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
 extern const int jconst_rgb_ycc_convert_sse2[];
 EXTERN(void) jsimd_rgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_ycc_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_neon
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_extrgb_ycc_convert_neon_slowld3
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_neon_slowld3
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_ycc_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+/* RGB & extended RGB --> Grayscale Colorspace Conversion */
+EXTERN(void) jsimd_rgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mmx
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 
 extern const int jconst_rgb_gray_convert_sse2[];
 EXTERN(void) jsimd_rgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extrgbx_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgr_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extbgrx_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxbgr_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
 EXTERN(void) jsimd_extxrgb_gray_convert_sse2
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+EXTERN(void) jsimd_rgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extrgbx_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgr_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extbgrx_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxbgr_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+EXTERN(void) jsimd_extxrgb_gray_convert_altivec
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows);
+
+/* YCC --> RGB & extended RGB Colorspace Conversion */
+EXTERN(void) jsimd_ycc_rgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mmx
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 
 extern const int jconst_ycc_rgb_convert_sse2[];
 EXTERN(void) jsimd_ycc_rgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgbx_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgr_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgrx_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxbgr_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxrgb_convert_sse2
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
-
-EXTERN(void) jsimd_rgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extrgbx_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgr_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extbgrx_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxbgr_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
-EXTERN(void) jsimd_extxrgb_ycc_convert_neon
-        JPP((JDIMENSION img_width,
-             JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
-             JDIMENSION output_row, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 
 EXTERN(void) jsimd_ycc_rgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extrgbx_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgr_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extbgrx_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxbgr_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
 EXTERN(void) jsimd_ycc_extxrgb_convert_neon
-        JPP((JDIMENSION out_width,
-             JSAMPIMAGE input_buf, JDIMENSION input_row,
-             JSAMPARRAY output_buf, int num_rows));
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_rgb565_convert_neon
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_extrgb_convert_neon_slowst3
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_neon_slowst3
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_mips_dspr2
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+EXTERN(void) jsimd_ycc_rgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extrgbx_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgr_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extbgrx_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxbgr_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+EXTERN(void) jsimd_ycc_extxrgb_convert_altivec
+        (JDIMENSION out_width, JSAMPIMAGE input_buf, JDIMENSION input_row,
+         JSAMPARRAY output_buf, int num_rows);
+
+/* NULL Colorspace Conversion */
+EXTERN(void) jsimd_c_null_convert_mips_dspr2
+        (JDIMENSION img_width, JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+         JDIMENSION output_row, int num_rows, int num_components);
 
-/* SIMD Downsample */
-EXTERN(void) jsimd_h2v2_downsample_mmx
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+/* h2v1 Downsampling */
 EXTERN(void) jsimd_h2v1_downsample_mmx
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-EXTERN(void) jsimd_h2v2_downsample_sse2
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
 EXTERN(void) jsimd_h2v1_downsample_sse2
-        JPP((JDIMENSION image_width, int max_v_samp_factor,
-             JDIMENSION v_samp_factor, JDIMENSION width_blocks,
-             JSAMPARRAY input_data, JSAMPARRAY output_data));
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_neon
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_mips_dspr2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v1_downsample_altivec
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
 
-/* SIMD Upsample */
-EXTERN(void) jsimd_h2v2_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+/* h2v2 Downsampling */
+EXTERN(void) jsimd_h2v2_downsample_mmx
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_sse2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_neon
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_mips_dspr2
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+EXTERN(void) jsimd_h2v2_downsample_altivec
+        (JDIMENSION image_width, int max_v_samp_factor,
+         JDIMENSION v_samp_factor, JDIMENSION width_blocks,
+         JSAMPARRAY input_data, JSAMPARRAY output_data);
+
+/* h2v2 Smooth Downsampling */
+EXTERN(void) jsimd_h2v2_smooth_downsample_mips_dspr2
+        (JSAMPARRAY input_data, JSAMPARRAY output_data,
+         JDIMENSION v_samp_factor, int max_v_samp_factor,
+         int smoothing_factor, JDIMENSION width_blocks,
+         JDIMENSION image_width);
+
+
+/* Upsampling */
 EXTERN(void) jsimd_h2v1_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_mmx
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
 
-EXTERN(void) jsimd_h2v2_fancy_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+EXTERN(void) jsimd_h2v1_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_int_upsample_mips_dspr2
+        (UINT8 h_expand, UINT8 v_expand, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr, JDIMENSION output_width,
+         int max_v_samp_factor);
+
+EXTERN(void) jsimd_h2v1_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION output_width, JSAMPARRAY input_data,
+         JSAMPARRAY *output_data_ptr);
+
+/* Fancy Upsampling */
 EXTERN(void) jsimd_h2v1_fancy_upsample_mmx
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mmx
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
 
-EXTERN(void) jsimd_h2v2_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+extern const int jconst_fancy_upsample_sse2[];
+EXTERN(void) jsimd_h2v1_fancy_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_sse2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_neon
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_mips_dspr2
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+EXTERN(void) jsimd_h2v1_fancy_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+EXTERN(void) jsimd_h2v2_fancy_upsample_altivec
+        (int max_v_samp_factor, JDIMENSION downsampled_width,
+         JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr);
+
+/* Merged Upsampling */
 EXTERN(void) jsimd_h2v1_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mmx
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-
-EXTERN(void) jsimd_h2v2_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-EXTERN(void) jsimd_h2v1_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION output_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
-extern const int jconst_fancy_upsample_sse2[];
-EXTERN(void) jsimd_h2v2_fancy_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
-EXTERN(void) jsimd_h2v1_fancy_upsample_sse2
-        JPP((int max_v_samp_factor, JDIMENSION downsampled_width,
-             JSAMPARRAY input_data, JSAMPARRAY * output_data_ptr));
+EXTERN(void) jsimd_h2v2_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mmx
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
 extern const int jconst_merged_upsample_sse2[];
-EXTERN(void) jsimd_h2v2_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
-EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
 EXTERN(void) jsimd_h2v1_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_sse2
-        JPP((JDIMENSION output_width, JSAMPIMAGE input_buf,
-             JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf));
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_sse2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
 
-/* SIMD Sample Conversion */
-EXTERN(void) jsimd_convsamp_mmx JPP((JSAMPARRAY sample_data,
-                                     JDIMENSION start_col,
-                                     DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_sse2 JPP((JSAMPARRAY sample_data,
-                                      JDIMENSION start_col,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_neon JPP((JSAMPARRAY sample_data,
-                                      JDIMENSION start_col,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_convsamp_float_3dnow JPP((JSAMPARRAY sample_data,
-                                             JDIMENSION start_col,
-                                             FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_convsamp_float_sse JPP((JSAMPARRAY sample_data,
-                                           JDIMENSION start_col,
-                                           FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_convsamp_float_sse2 JPP((JSAMPARRAY sample_data,
-                                            JDIMENSION start_col,
-                                            FAST_FLOAT * workspace));
-
-/* SIMD Forward DCT */
-EXTERN(void) jsimd_fdct_islow_mmx JPP((DCTELEM * data));
-EXTERN(void) jsimd_fdct_ifast_mmx JPP((DCTELEM * data));
+EXTERN(void) jsimd_h2v1_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf, JSAMPLE* range);
+
+EXTERN(void) jsimd_h2v1_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extrgbx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extbgrx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v1_extxrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+EXTERN(void) jsimd_h2v2_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extrgbx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extbgrx_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxbgr_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+EXTERN(void) jsimd_h2v2_extxrgb_merged_upsample_altivec
+        (JDIMENSION output_width, JSAMPIMAGE input_buf,
+         JDIMENSION in_row_group_ctr, JSAMPARRAY output_buf);
+
+/* Sample Conversion */
+EXTERN(void) jsimd_convsamp_mmx
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_sse2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_neon
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_mips_dspr2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+EXTERN(void) jsimd_convsamp_altivec
+        (JSAMPARRAY sample_data, JDIMENSION start_col, DCTELEM *workspace);
+
+/* Floating Point Sample Conversion */
+EXTERN(void) jsimd_convsamp_float_3dnow
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_sse2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_convsamp_float_mips_dspr2
+        (JSAMPARRAY sample_data, JDIMENSION start_col, FAST_FLOAT *workspace);
+
+/* Slow Integer Forward DCT */
+EXTERN(void) jsimd_fdct_islow_mmx (DCTELEM *data);
 
-extern const int jconst_fdct_ifast_sse2[];
-EXTERN(void) jsimd_fdct_islow_sse2 JPP((DCTELEM * data));
 extern const int jconst_fdct_islow_sse2[];
-EXTERN(void) jsimd_fdct_ifast_sse2 JPP((DCTELEM * data));
+EXTERN(void) jsimd_fdct_islow_sse2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_neon (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_mips_dspr2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_islow_altivec (DCTELEM *data);
 
-EXTERN(void) jsimd_fdct_ifast_neon JPP((DCTELEM * data));
+/* Fast Integer Forward DCT */
+EXTERN(void) jsimd_fdct_ifast_mmx (DCTELEM *data);
 
-EXTERN(void) jsimd_fdct_float_3dnow JPP((FAST_FLOAT * data));
+extern const int jconst_fdct_ifast_sse2[];
+EXTERN(void) jsimd_fdct_ifast_sse2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_neon (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_mips_dspr2 (DCTELEM *data);
+
+EXTERN(void) jsimd_fdct_ifast_altivec (DCTELEM *data);
+
+/* Floating Point Forward DCT */
+EXTERN(void) jsimd_fdct_float_3dnow (FAST_FLOAT *data);
 
 extern const int jconst_fdct_float_sse[];
-EXTERN(void) jsimd_fdct_float_sse JPP((FAST_FLOAT * data));
+EXTERN(void) jsimd_fdct_float_sse (FAST_FLOAT *data);
+
+/* Quantization */
+EXTERN(void) jsimd_quantize_mmx
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_sse2
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_neon
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
 
-/* SIMD Quantization */
-EXTERN(void) jsimd_quantize_mmx JPP((JCOEFPTR coef_block,
-                                     DCTELEM * divisors,
-                                     DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_sse2 JPP((JCOEFPTR coef_block,
-                                      DCTELEM * divisors,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_neon JPP((JCOEFPTR coef_block,
-                                      DCTELEM * divisors,
-                                      DCTELEM * workspace));
-
-EXTERN(void) jsimd_quantize_float_3dnow JPP((JCOEFPTR coef_block,
-                                             FAST_FLOAT * divisors,
-                                             FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_quantize_float_sse JPP((JCOEFPTR coef_block,
-                                           FAST_FLOAT * divisors,
-                                           FAST_FLOAT * workspace));
-
-EXTERN(void) jsimd_quantize_float_sse2 JPP((JCOEFPTR coef_block,
-                                            FAST_FLOAT * divisors,
-                                            FAST_FLOAT * workspace));
-
-/* SIMD Reduced Inverse DCT */
-EXTERN(void) jsimd_idct_2x2_mmx JPP((void * dct_table,
-                                     JCOEFPTR coef_block,
-                                     JSAMPARRAY output_buf,
-                                     JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_mmx JPP((void * dct_table,
-                                     JCOEFPTR coef_block,
-                                     JSAMPARRAY output_buf,
-                                     JDIMENSION output_col));
+EXTERN(void) jsimd_quantize_mips_dspr2
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+EXTERN(void) jsimd_quantize_altivec
+        (JCOEFPTR coef_block, DCTELEM *divisors, DCTELEM *workspace);
+
+/* Floating Point Quantization */
+EXTERN(void) jsimd_quantize_float_3dnow
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_sse2
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+EXTERN(void) jsimd_quantize_float_mips_dspr2
+        (JCOEFPTR coef_block, FAST_FLOAT *divisors, FAST_FLOAT *workspace);
+
+/* Scaled Inverse DCT */
+EXTERN(void) jsimd_idct_2x2_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_red_sse2[];
-EXTERN(void) jsimd_idct_2x2_sse2 JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_sse2 JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_2x2_neon JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-EXTERN(void) jsimd_idct_4x4_neon JPP((void * dct_table,
-                                      JCOEFPTR coef_block,
-                                      JSAMPARRAY output_buf,
-                                      JDIMENSION output_col));
-
-/* SIMD Inverse DCT */
-EXTERN(void) jsimd_idct_islow_mmx JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast_mmx JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
+EXTERN(void) jsimd_idct_2x2_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_2x2_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_4x4_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col, int *workspace);
+EXTERN(void) jsimd_idct_6x6_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+EXTERN(void) jsimd_idct_12x12_pass1_mips_dspr2
+        (JCOEFPTR coef_block, void *dct_table, int *workspace);
+EXTERN(void) jsimd_idct_12x12_pass2_mips_dspr2
+        (int *workspace, int *output);
+
+/* Slow Integer Inverse DCT */
+EXTERN(void) jsimd_idct_islow_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_islow_sse2[];
-EXTERN(void) jsimd_idct_islow_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
+EXTERN(void) jsimd_idct_islow_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_islow_mips_dspr2
+        (void *dct_table, JCOEFPTR coef_block, int *output_buf,
+         JSAMPLE *output_col);
+
+EXTERN(void) jsimd_idct_islow_altivec
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Fast Integer Inverse DCT */
+EXTERN(void) jsimd_idct_ifast_mmx
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
 extern const int jconst_idct_ifast_sse2[];
-EXTERN(void) jsimd_idct_ifast_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_islow_neon JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-EXTERN(void) jsimd_idct_ifast_neon JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
-EXTERN(void) jsimd_idct_float_3dnow JPP((void * dct_table,
-                                         JCOEFPTR coef_block,
-                                         JSAMPARRAY output_buf,
-                                         JDIMENSION output_col));
+EXTERN(void) jsimd_idct_ifast_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_neon
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+EXTERN(void) jsimd_idct_ifast_cols_mips_dspr2
+        (JCOEF *inptr, IFAST_MULT_TYPE *quantptr, DCTELEM *wsptr,
+         const int *idct_coefs);
+EXTERN(void) jsimd_idct_ifast_rows_mips_dspr2
+        (DCTELEM *wsptr, JSAMPARRAY output_buf, JDIMENSION output_col,
+         const int *idct_coefs);
+
+EXTERN(void) jsimd_idct_ifast_altivec
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Floating Point Inverse DCT */
+EXTERN(void) jsimd_idct_float_3dnow
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_float_sse[];
-EXTERN(void) jsimd_idct_float_sse JPP((void * dct_table,
-                                       JCOEFPTR coef_block,
-                                       JSAMPARRAY output_buf,
-                                       JDIMENSION output_col));
+EXTERN(void) jsimd_idct_float_sse
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
 
 extern const int jconst_idct_float_sse2[];
-EXTERN(void) jsimd_idct_float_sse2 JPP((void * dct_table,
-                                        JCOEFPTR coef_block,
-                                        JSAMPARRAY output_buf,
-                                        JDIMENSION output_col));
-
+EXTERN(void) jsimd_idct_float_sse2
+        (void *dct_table, JCOEFPTR coef_block, JSAMPARRAY output_buf,
+         JDIMENSION output_col);
+
+/* Huffman coding */
+extern const int jconst_huff_encode_one_block[];
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_sse2
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_neon
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
+
+EXTERN(JOCTET*) jsimd_huff_encode_one_block_neon_slowtbl
+        (void *state, JOCTET *buffer, JCOEFPTR block, int last_dc_val,
+         c_derived_tbl *dctbl, c_derived_tbl *actbl);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_i386.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_i386.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_i386.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_i386.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1091 @@
+/*
+ * jsimd_i386.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2013-2014, 2016 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 32-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order) (((unsigned)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = jpeg_simd_cpu_support();
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCEMMX");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_MMX;
+  env = getenv("JSIMD_FORCE3DNOW");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_3DNOW|JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE|JSIMD_MMX;
+  env = getenv("JSIMD_FORCESSE2");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support &= JSIMD_SSE2;
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_ycc_convert_sse2;
+      mmxfct=jsimd_extrgb_ycc_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_ycc_convert_sse2;
+      mmxfct=jsimd_extrgbx_ycc_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_ycc_convert_sse2;
+      mmxfct=jsimd_extbgr_ycc_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_ycc_convert_sse2;
+      mmxfct=jsimd_extbgrx_ycc_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_ycc_convert_sse2;
+      mmxfct=jsimd_extxbgr_ycc_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_ycc_convert_sse2;
+      mmxfct=jsimd_extxrgb_ycc_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_rgb_ycc_convert_sse2;
+      mmxfct=jsimd_rgb_ycc_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+  void (*mmxfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      mmxfct=jsimd_extrgb_gray_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      mmxfct=jsimd_extrgbx_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      mmxfct=jsimd_extbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      mmxfct=jsimd_extbgrx_gray_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      mmxfct=jsimd_extxbgr_gray_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      mmxfct=jsimd_extxrgb_gray_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      mmxfct=jsimd_rgb_gray_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_ycc_extrgb_convert_sse2;
+      mmxfct=jsimd_ycc_extrgb_convert_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_ycc_extrgbx_convert_sse2;
+      mmxfct=jsimd_ycc_extrgbx_convert_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_ycc_extbgr_convert_sse2;
+      mmxfct=jsimd_ycc_extbgr_convert_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_ycc_extbgrx_convert_sse2;
+      mmxfct=jsimd_ycc_extbgrx_convert_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_ycc_extxbgr_convert_sse2;
+      mmxfct=jsimd_ycc_extxbgr_convert_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_ycc_extxrgb_convert_sse2;
+      mmxfct=jsimd_ycc_extxrgb_convert_mmx;
+      break;
+    default:
+      sse2fct=jsimd_ycc_rgb_convert_sse2;
+      mmxfct=jsimd_ycc_rgb_convert_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                               compptr->v_samp_factor,
+                               compptr->width_in_blocks, input_data,
+                               output_data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_downsample_mmx(cinfo->image_width, cinfo->max_v_samp_factor,
+                              compptr->v_samp_factor, compptr->width_in_blocks,
+                              input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                             input_data, output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_upsample_mmx(cinfo->max_v_samp_factor, cinfo->output_width,
+                            input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v2_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                   compptr->downsampled_width, input_data,
+                                   output_data_ptr);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_h2v1_fancy_upsample_mmx(cinfo->max_v_samp_factor,
+                                  compptr->downsampled_width, input_data,
+                                  output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extrgb_merged_upsample_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extrgbx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extbgrx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extxbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_extxrgb_merged_upsample_mmx;
+      break;
+    default:
+      sse2fct=jsimd_h2v2_merged_upsample_sse2;
+      mmxfct=jsimd_h2v2_merged_upsample_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+  void (*mmxfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extrgb_merged_upsample_mmx;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extrgbx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extbgrx_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extxbgr_merged_upsample_mmx;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_extxrgb_merged_upsample_mmx;
+      break;
+    default:
+      sse2fct=jsimd_h2v1_merged_upsample_sse2;
+      mmxfct=jsimd_h2v1_merged_upsample_mmx;
+      break;
+  }
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+  else if (simd_support & JSIMD_MMX)
+    mmxfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_sse2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_convsamp_mmx(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_convsamp_float_sse(sample_data, start_col, workspace);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_convsamp_float_3dnow(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    jsimd_fdct_islow_sse2(data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_fdct_islow_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    jsimd_fdct_ifast_sse2(data);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_fdct_ifast_mmx(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    jsimd_fdct_float_sse(data);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_fdct_float_3dnow(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+  if (simd_support & JSIMD_SSE)
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_sse2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_quantize_mmx(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_SSE2)
+    jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_SSE)
+    jsimd_quantize_float_sse(coef_block, divisors, workspace);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_quantize_float_3dnow(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_2x2_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_4x4_mmx(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+  if (simd_support & JSIMD_MMX)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    return 1;
+  if (simd_support & JSIMD_3DNOW)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_islow_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if (simd_support & JSIMD_MMX)
+    jsimd_idct_ifast_mmx(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                          output_col);
+  else if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_idct_float_sse))
+    jsimd_idct_float_sse(compptr->dct_table, coef_block, output_buf,
+                         output_col);
+  else if (simd_support & JSIMD_3DNOW)
+    jsimd_idct_float_3dnow(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1130 @@
+/*
+ * jsimd_mips.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014 D. R. Commander
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * MIPS architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+#include <stdio.h>
+#include <string.h>
+#include <ctype.h>
+
+static unsigned int simd_support = ~0;
+
+#if defined(__linux__)
+
+LOCAL(int)
+parse_proc_cpuinfo(const char* search_string)
+{
+  const char* file_name = "/proc/cpuinfo";
+  char cpuinfo_line[256];
+  FILE* f = NULL;
+  simd_support = 0;
+
+  if ((f = fopen(file_name, "r")) != NULL) {
+    while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) {
+      if (strstr(cpuinfo_line, search_string) != NULL) {
+        fclose(f);
+        simd_support |= JSIMD_MIPS_DSPR2;
+        return 1;
+      }
+    }
+    fclose(f);
+  }
+  /* Did not find string in the proc file, or not Linux ELF. */
+  return 0;
+}
+
+#endif
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = 0;
+
+#if defined(__MIPSEL__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2)
+  simd_support |= JSIMD_MIPS_DSPR2;
+#elif defined(__linux__)
+  /* We still have a chance to use MIPS DSPR2 regardless of globally used
+   * -mdspr2 options passed to gcc by performing runtime detection via
+   * /proc/cpuinfo parsing on linux */
+  if (!parse_proc_cpuinfo("MIPS 74K"))
+    return;
+#endif
+}
+
+static const int mips_idct_ifast_coefs[4] = {
+  0x45404540,           // FIX( 1.082392200 / 2) =  17734 = 0x4546
+  0x5A805A80,           // FIX( 1.414213562 / 2) =  23170 = 0x5A82
+  0x76407640,           // FIX( 1.847759065 / 2) =  30274 = 0x7642
+  0xAC60AC60            // FIX(-2.613125930 / 4) = -21407 = 0xAC61
+};
+
+/* The following struct is borrowed from jdsample.c */
+typedef void (*upsample1_ptr) (j_decompress_ptr cinfo,
+                               jpeg_component_info *compptr,
+                               JSAMPARRAY input_data,
+                               JSAMPARRAY *output_data_ptr);
+
+typedef struct {
+  struct jpeg_upsampler pub;
+  JSAMPARRAY color_buf[MAX_COMPONENTS];
+  upsample1_ptr methods[MAX_COMPONENTS];
+  int next_row_out;
+  JDIMENSION rows_to_go;
+  int rowgroup_height[MAX_COMPONENTS];
+  UINT8 h_expand[MAX_COMPONENTS];
+  UINT8 v_expand[MAX_COMPONENTS];
+} my_upsampler;
+
+typedef my_upsampler *my_upsample_ptr;
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_c_can_null_convert (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_extrgbx_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_extbgr_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_extbgrx_ycc_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_extxbgr_ycc_convert_mips_dspr2;
+
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_extxrgb_ycc_convert_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_extrgb_ycc_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_extrgbx_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_extbgr_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_extbgrx_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_extxbgr_gray_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_extxrgb_gray_convert_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_extrgb_gray_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->image_width, input_buf, output_buf, output_row,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_ycc_extrgbx_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_ycc_extbgr_convert_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_ycc_extbgrx_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_ycc_extxbgr_convert_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_ycc_extxrgb_convert_mips_dspr2;
+      break;
+  default:
+      mipsdspr2fct=jsimd_ycc_extrgb_convert_mips_dspr2;
+      break;
+  }
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    mipsdspr2fct(cinfo->output_width, input_buf, input_row, output_buf,
+                 num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(void)
+jsimd_c_null_convert (j_compress_ptr cinfo,
+                      JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                      JDIMENSION output_row, int num_rows)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_c_null_convert_mips_dspr2(cinfo->image_width, input_buf,
+                                    output_buf, output_row, num_rows,
+                                    cinfo->num_components);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_smooth_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if(DCTSIZE != 8)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_downsample_mips_dspr2(cinfo->image_width,
+                                     cinfo->max_v_samp_factor,
+                                     compptr->v_samp_factor,
+                                     compptr->width_in_blocks, input_data,
+                                     output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v2_smooth_downsample (j_compress_ptr cinfo,
+                              jpeg_component_info *compptr,
+                              JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_smooth_downsample_mips_dspr2(input_data, output_data,
+                                          compptr->v_samp_factor,
+                                          cinfo->max_v_samp_factor,
+                                          cinfo->smoothing_factor,
+                                          compptr->width_in_blocks,
+                                          cinfo->image_width);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_downsample_mips_dspr2(cinfo->image_width,
+                                     cinfo->max_v_samp_factor,
+                                     compptr->v_samp_factor,
+                                     compptr->width_in_blocks,
+                                     input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_int_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                   cinfo->output_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                   cinfo->output_width, input_data,
+                                   output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_int_upsample (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                    JSAMPARRAY input_data, JSAMPARRAY *output_data_ptr)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+
+  jsimd_int_upsample_mips_dspr2(upsample->h_expand[compptr->component_index],
+                                upsample->v_expand[compptr->component_index],
+                                input_data, output_data_ptr,
+                                cinfo->output_width,
+                                cinfo->max_v_samp_factor);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v2_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                         compptr->downsampled_width,
+                                         input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_h2v1_fancy_upsample_mips_dspr2(cinfo->max_v_samp_factor,
+                                         compptr->downsampled_width,
+                                         input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY,
+                       JSAMPLE *);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_h2v2_extbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_h2v2_extrgb_merged_upsample_mips_dspr2;
+      break;
+  }
+
+  mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+               cinfo->sample_range_limit);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*mipsdspr2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY,
+                       JSAMPLE *);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      mipsdspr2fct=jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGR:
+      mipsdspr2fct=jsimd_h2v1_extbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      mipsdspr2fct=jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      mipsdspr2fct=jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      mipsdspr2fct=jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2;
+      break;
+    default:
+      mipsdspr2fct=jsimd_h2v1_extrgb_merged_upsample_mips_dspr2;
+      break;
+  }
+
+  mipsdspr2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf,
+               cinfo->sample_range_limit);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_convsamp_mips_dspr2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  if ((simd_support & JSIMD_MIPS_DSPR2))
+    jsimd_convsamp_float_mips_dspr2(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_fdct_islow_mips_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_fdct_ifast_mips_dspr2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_quantize_mips_dspr2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_quantize_float_mips_dspr2(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_6x6 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_12x12 (void)
+{
+  init_simd();
+
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    jsimd_idct_2x2_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                              output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int workspace[DCTSIZE*4];  /* buffers data between passes */
+    jsimd_idct_4x4_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                              output_col, workspace);
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_6x6 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+           JCOEFPTR coef_block, JSAMPARRAY output_buf,
+           JDIMENSION output_col)
+{
+    if (simd_support & JSIMD_MIPS_DSPR2)
+      jsimd_idct_6x6_mips_dspr2(compptr->dct_table, coef_block, output_buf,
+                                output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_12x12 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block,
+                  JSAMPARRAY output_buf, JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int workspace[96];
+    int output[12] = {
+      (int)(output_buf[0] + output_col),
+      (int)(output_buf[1] + output_col),
+      (int)(output_buf[2] + output_col),
+      (int)(output_buf[3] + output_col),
+      (int)(output_buf[4] + output_col),
+      (int)(output_buf[5] + output_col),
+      (int)(output_buf[6] + output_col),
+      (int)(output_buf[7] + output_col),
+      (int)(output_buf[8] + output_col),
+      (int)(output_buf[9] + output_col),
+      (int)(output_buf[10] + output_col),
+      (int)(output_buf[11] + output_col),
+    };
+    jsimd_idct_12x12_pass1_mips_dspr2(coef_block, compptr->dct_table,
+                                      workspace);
+    jsimd_idct_12x12_pass2_mips_dspr2(workspace, output);
+  }
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if (simd_support & JSIMD_MIPS_DSPR2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    int output[8] = {
+      (int)(output_buf[0] + output_col),
+      (int)(output_buf[1] + output_col),
+      (int)(output_buf[2] + output_col),
+      (int)(output_buf[3] + output_col),
+      (int)(output_buf[4] + output_col),
+      (int)(output_buf[5] + output_col),
+      (int)(output_buf[6] + output_col),
+      (int)(output_buf[7] + output_col),
+    };
+
+    jsimd_idct_islow_mips_dspr2(coef_block, compptr->dct_table,
+                                output, IDCT_range_limit(cinfo));
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  if (simd_support & JSIMD_MIPS_DSPR2) {
+    JCOEFPTR inptr;
+    IFAST_MULT_TYPE *quantptr;
+    DCTELEM workspace[DCTSIZE2];  /* buffers data between passes */
+
+    /* Pass 1: process columns from input, store into work array. */
+
+    inptr = coef_block;
+    quantptr = (IFAST_MULT_TYPE *) compptr->dct_table;
+
+    jsimd_idct_ifast_cols_mips_dspr2(inptr, quantptr,
+                                     workspace, mips_idct_ifast_coefs);
+
+    /* Pass 2: process rows from work array, store into output array. */
+    /* Note that we must descale the results by a factor of 8 == 2**3, */
+    /* and also undo the PASS1_BITS scaling. */
+
+    jsimd_idct_ifast_rows_mips_dspr2(workspace, output_buf,
+                                     output_col, mips_idct_ifast_coefs);
+  }
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips_dspr2_asm.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,285 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013, MIPS Technologies, Inc., California.
+ * All rights reserved.
+ * Authors:  Teodora Novkovic (teodora.novkovic@imgtec.com)
+ *           Darko Laus       (darko.laus@imgtec.com)
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#define zero $0
+#define AT   $1
+#define v0   $2
+#define v1   $3
+#define a0   $4
+#define a1   $5
+#define a2   $6
+#define a3   $7
+#define t0   $8
+#define t1   $9
+#define t2   $10
+#define t3   $11
+#define t4   $12
+#define t5   $13
+#define t6   $14
+#define t7   $15
+#define s0   $16
+#define s1   $17
+#define s2   $18
+#define s3   $19
+#define s4   $20
+#define s5   $21
+#define s6   $22
+#define s7   $23
+#define t8   $24
+#define t9   $25
+#define k0   $26
+#define k1   $27
+#define gp   $28
+#define sp   $29
+#define fp   $30
+#define s8   $30
+#define ra   $31
+
+#define f0   $f0
+#define f1   $f1
+#define f2   $f2
+#define f3   $f3
+#define f4   $f4
+#define f5   $f5
+#define f6   $f6
+#define f7   $f7
+#define f8   $f8
+#define f9   $f9
+#define f10  $f10
+#define f11  $f11
+#define f12  $f12
+#define f13  $f13
+#define f14  $f14
+#define f15  $f15
+#define f16  $f16
+#define f17  $f17
+#define f18  $f18
+#define f19  $f19
+#define f20  $f20
+#define f21  $f21
+#define f22  $f22
+#define f23  $f23
+#define f24  $f24
+#define f25  $f25
+#define f26  $f26
+#define f27  $f27
+#define f28  $f28
+#define f29  $f29
+#define f30  $f30
+#define f31  $f31
+
+/*
+ * LEAF_MIPS32R2 - declare leaf routine for MIPS32r2
+ */
+#define LEAF_MIPS32R2(symbol)                           \
+                .globl  symbol;                         \
+                .align  2;                              \
+                .type   symbol, @function;              \
+                .ent    symbol, 0;                      \
+symbol:         .frame  sp, 0, ra;                      \
+                .set    push;                           \
+                .set    arch=mips32r2;                  \
+                .set    noreorder;                      \
+                .set    noat;
+
+/*
+ * LEAF_MIPS_DSPR2 - declare leaf routine for MIPS DSPr2
+ */
+#define LEAF_MIPS_DSPR2(symbol)                         \
+LEAF_MIPS32R2(symbol)                                   \
+                .set    dspr2;
+
+/*
+ * END - mark end of function
+ */
+#define END(function)                                   \
+                .set    pop;                            \
+                .end    function;                       \
+                .size   function,.-function
+
+/*
+ * Checks if stack offset is big enough for storing/restoring regs_num
+ * number of register to/from stack. Stack offset must be greater than
+ * or equal to the number of bytes needed for storing registers (regs_num*4).
+ * Since MIPS ABI allows usage of first 16 bytes of stack frame (this is
+ * preserved for input arguments of the functions, already stored in a0-a3),
+ * stack size can be further optimized by utilizing this space.
+ */
+.macro CHECK_STACK_OFFSET regs_num, stack_offset
+.if \stack_offset < \regs_num * 4 - 16
+.error "Stack offset too small."
+.endif
+.endm
+
+/*
+ * Saves set of registers on stack. Maximum number of registers that
+ * can be saved on stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * before registers are pushed in order to provide enough space on stack
+ * (offset must be multiple of 4, and must be big enough, as described by
+ * CHECK_STACK_OFFSET macro). This macro is intended to be used in
+ * combination with RESTORE_REGS_FROM_STACK macro. Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro SAVE_REGS_ON_STACK stack_offset = 0, r1, \
+                          r2  = 0, r3  = 0, r4  = 0, \
+                          r5  = 0, r6  = 0, r7  = 0, \
+                          r8  = 0, r9  = 0, r10 = 0, \
+                          r11 = 0, r12 = 0, r13 = 0, \
+                          r14 = 0
+    .if (\stack_offset < 0) || (\stack_offset - (\stack_offset / 4) * 4)
+    .error "Stack offset must be pozitive and multiple of 4."
+    .endif
+    .if \stack_offset != 0
+    addiu           sp, sp, -\stack_offset
+    .endif
+    sw              \r1, 0(sp)
+    .if \r2 != 0
+    sw              \r2, 4(sp)
+    .endif
+    .if \r3 != 0
+    sw              \r3, 8(sp)
+    .endif
+    .if \r4 != 0
+    sw              \r4, 12(sp)
+    .endif
+    .if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    sw              \r5, 16(sp)
+    .endif
+    .if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    sw              \r6, 20(sp)
+    .endif
+    .if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    sw              \r7, 24(sp)
+    .endif
+    .if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    sw              \r8, 28(sp)
+    .endif
+    .if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    sw              \r9, 32(sp)
+    .endif
+    .if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    sw              \r10, 36(sp)
+    .endif
+    .if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    sw              \r11, 40(sp)
+    .endif
+    .if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    sw              \r12, 44(sp)
+    .endif
+    .if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    sw              \r13, 48(sp)
+    .endif
+    .if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    sw              \r14, 52(sp)
+    .endif
+.endm
+
+/*
+ * Restores set of registers from stack. Maximum number of registers that
+ * can be restored from stack is limitted to 14 (a0-a3, v0-v1 and s0-s7).
+ * Stack offset is number of bytes that are added to stack pointer (sp)
+ * after registers are restored (offset must be multiple of 4, and must
+ * be big enough, as described by CHECK_STACK_OFFSET macro). This macro is
+ * intended to be used in combination with RESTORE_REGS_FROM_STACK macro.
+ * Example:
+ *  SAVE_REGS_ON_STACK      4, v0, v1, s0, s1
+ *  RESTORE_REGS_FROM_STACK 4, v0, v1, s0, s1
+ */
+.macro RESTORE_REGS_FROM_STACK stack_offset = 0, r1, \
+                               r2  = 0, r3  = 0, r4  = 0, \
+                               r5  = 0, r6  = 0, r7  = 0, \
+                               r8  = 0, r9  = 0, r10 = 0, \
+                               r11 = 0, r12 = 0, r13 = 0, \
+                               r14 = 0
+    .if (\stack_offset < 0) || (\stack_offset - (\stack_offset/4)*4)
+    .error "Stack offset must be pozitive and multiple of 4."
+    .endif
+    lw              \r1, 0(sp)
+    .if \r2 != 0
+    lw              \r2, 4(sp)
+    .endif
+    .if \r3 != 0
+    lw              \r3, 8(sp)
+    .endif
+    .if \r4 != 0
+    lw              \r4, 12(sp)
+    .endif
+    .if \r5 != 0
+    CHECK_STACK_OFFSET 5, \stack_offset
+    lw              \r5, 16(sp)
+    .endif
+    .if \r6 != 0
+    CHECK_STACK_OFFSET 6, \stack_offset
+    lw              \r6, 20(sp)
+    .endif
+    .if \r7 != 0
+    CHECK_STACK_OFFSET 7, \stack_offset
+    lw              \r7, 24(sp)
+    .endif
+    .if \r8 != 0
+    CHECK_STACK_OFFSET 8, \stack_offset
+    lw              \r8, 28(sp)
+    .endif
+    .if \r9 != 0
+    CHECK_STACK_OFFSET 9, \stack_offset
+    lw              \r9, 32(sp)
+    .endif
+    .if \r10 != 0
+    CHECK_STACK_OFFSET 10, \stack_offset
+    lw              \r10, 36(sp)
+    .endif
+    .if \r11 != 0
+    CHECK_STACK_OFFSET 11, \stack_offset
+    lw              \r11, 40(sp)
+    .endif
+    .if \r12 != 0
+    CHECK_STACK_OFFSET 12, \stack_offset
+    lw              \r12, 44(sp)
+    .endif
+    .if \r13 != 0
+    CHECK_STACK_OFFSET 13, \stack_offset
+    lw              \r13, 48(sp)
+    .endif
+    .if \r14 != 0
+    CHECK_STACK_OFFSET 14, \stack_offset
+    lw              \r14, 52(sp)
+    .endif
+    .if \stack_offset != 0
+    addiu           sp, sp, \stack_offset
+    .endif
+.endm
+
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_mips_dspr2.S	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,4487 @@
+/*
+ * MIPS DSPr2 optimizations for libjpeg-turbo
+ *
+ * Copyright (C) 2013-2014, MIPS Technologies, Inc., California.
+ * All rights reserved.
+ * Authors:  Teodora Novkovic (teodora.novkovic@imgtec.com)
+ *           Darko Laus       (darko.laus@imgtec.com)
+ * Copyright (C) 2015, D. R. Commander.  All Rights Reserved.
+ * This software is provided 'as-is', without any express or implied
+ * warranty.  In no event will the authors be held liable for any damages
+ * arising from the use of this software.
+ *
+ * Permission is granted to anyone to use this software for any purpose,
+ * including commercial applications, and to alter it and redistribute it
+ * freely, subject to the following restrictions:
+ *
+ * 1. The origin of this software must not be misrepresented; you must not
+ *    claim that you wrote the original software. If you use this software
+ *    in a product, an acknowledgment in the product documentation would be
+ *    appreciated but is not required.
+ * 2. Altered source versions must be plainly marked as such, and must not be
+ *    misrepresented as being the original software.
+ * 3. This notice may not be removed or altered from any source distribution.
+ */
+
+#include "jsimd_mips_dspr2_asm.h"
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_c_null_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ * 20(sp) - cinfo->num_components
+ *
+ * Null conversion for compression
+ */
+
+    SAVE_REGS_ON_STACK 8, s0, s1
+
+    lw        t9, 24(sp)   // t9 = num_rows
+    lw        s0, 28(sp)   // s0 = cinfo->num_components
+    andi      t0, a0, 3    // t0 = cinfo->image_width & 3
+    beqz      t0, 4f       // no residual
+     nop
+0:
+    addiu     t9, t9, -1
+    bltz      t9, 7f
+     li       t1, 0
+1:
+    sll       t3, t1, 2
+    lwx       t5, t3(a2)   // t5 = outptr = output_buf[ci]
+    lw        t2, 0(a1)    // t2 = inptr = *input_buf
+    sll       t4, a3, 2
+    lwx       t5, t4(t5)   // t5 = outptr = output_buf[ci][output_row]
+    addu      t2, t2, t1
+    addu      s1, t5, a0
+    addu      t6, t5, t0
+2:
+    lbu       t3, 0(t2)
+    addiu     t5, t5, 1
+    sb        t3, -1(t5)
+    bne       t6, t5, 2b
+     addu     t2, t2, s0
+3:
+    lbu       t3, 0(t2)
+    addu      t4, t2, s0
+    addu      t7, t4, s0
+    addu      t8, t7, s0
+    addu      t2, t8, s0
+    lbu       t4, 0(t4)
+    lbu       t7, 0(t7)
+    lbu       t8, 0(t8)
+    addiu     t5, t5, 4
+    sb        t3, -4(t5)
+    sb        t4, -3(t5)
+    sb        t7, -2(t5)
+    bne       s1, t5, 3b
+     sb       t8, -1(t5)
+    addiu     t1, t1, 1
+    bne       t1, s0, 1b
+     nop
+    addiu     a1, a1, 4
+    bgez      t9, 0b
+     addiu    a3, a3, 1
+    b         7f
+     nop
+4:
+    addiu     t9, t9, -1
+    bltz      t9, 7f
+     li       t1, 0
+5:
+    sll       t3, t1, 2
+    lwx       t5, t3(a2)   // t5 = outptr = output_buf[ci]
+    lw        t2, 0(a1)    // t2 = inptr = *input_buf
+    sll       t4, a3, 2
+    lwx       t5, t4(t5)   // t5 = outptr = output_buf[ci][output_row]
+    addu      t2, t2, t1
+    addu      s1, t5, a0
+    addu      t6, t5, t0
+6:
+    lbu       t3, 0(t2)
+    addu      t4, t2, s0
+    addu      t7, t4, s0
+    addu      t8, t7, s0
+    addu      t2, t8, s0
+    lbu       t4, 0(t4)
+    lbu       t7, 0(t7)
+    lbu       t8, 0(t8)
+    addiu     t5, t5, 4
+    sb        t3, -4(t5)
+    sb        t4, -3(t5)
+    sb        t7, -2(t5)
+    bne       s1, t5, 6b
+     sb       t8, -1(t5)
+    addiu     t1, t1, 1
+    bne       t1, s0, 5b
+     nop
+    addiu     a1, a1, 4
+    bgez      t9, 4b
+     addiu    a3, a3, 1
+7:
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j         ra
+     nop
+
+END(jsimd_c_null_convert_mips_dspr2)
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_ycc_convert_mips_dspr2
+ * jsimd_extbgr_ycc_convert_mips_dspr2
+ * jsimd_extrgbx_ycc_convert_mips_dspr2
+ * jsimd_extbgrx_ycc_convert_mips_dspr2
+ * jsimd_extxbgr_ycc_convert_mips_dspr2
+ * jsimd_extxrgb_ycc_convert_mips_dspr2
+ *
+ * Colorspace conversion RGB -> YCbCr
+ */
+
+.macro GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_YCC r,    \
+                     g,    \
+                     b,    \
+                     inptr
+    lbu     \r, \r_offs(\inptr)
+    lbu     \g, \g_offs(\inptr)
+    lbu     \b, \b_offs(\inptr)
+    addiu   \inptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_\colorid\()_ycc_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw      t7, 48(sp)        // t7 = num_rows
+    li      s0, 0x4c8b        // FIX(0.29900)
+    li      s1, 0x9646        // FIX(0.58700)
+    li      s2, 0x1d2f        // FIX(0.11400)
+    li      s3, 0xffffd4cd    // -FIX(0.16874)
+    li      s4, 0xffffab33    // -FIX(0.33126)
+    li      s5, 0x8000        // FIX(0.50000)
+    li      s6, 0xffff94d1    // -FIX(0.41869)
+    li      s7, 0xffffeb2f    // -FIX(0.08131)
+    li      t8, 0x807fff      // CBCR_OFFSET + ONE_HALF-1
+
+0:
+    addiu   t7, -1            // --num_rows
+    lw      t6, 0(a1)         // t6 = input_buf[0]
+    lw      t0, 0(a2)
+    lw      t1, 4(a2)
+    lw      t2, 8(a2)
+    sll     t3, a3, 2
+    lwx     t0, t3(t0)        // t0 = output_buf[0][output_row]
+    lwx     t1, t3(t1)        // t1 = output_buf[1][output_row]
+    lwx     t2, t3(t2)        // t2 = output_buf[2][output_row]
+
+    addu    t9, t2, a0        // t9 = end address
+    addiu   a3, 1
+
+1:
+    DO_RGB_TO_YCC t3, t4, t5, t6
+
+    mtlo    s5, $ac0
+    mtlo    t8, $ac1
+    mtlo    t8, $ac2
+    maddu   $ac0, s2, t5
+    maddu   $ac1, s5, t5
+    maddu   $ac2, s5, t3
+    maddu   $ac0, s0, t3
+    maddu   $ac1, s3, t3
+    maddu   $ac2, s6, t4
+    maddu   $ac0, s1, t4
+    maddu   $ac1, s4, t4
+    maddu   $ac2, s7, t5
+    extr.w  t3, $ac0, 16
+    extr.w  t4, $ac1, 16
+    extr.w  t5, $ac2, 16
+    sb      t3, 0(t0)
+    sb      t4, 0(t1)
+    sb      t5, 0(t2)
+    addiu   t0, 1
+    addiu   t2, 1
+    bne     t2, t9, 1b
+     addiu  t1, 1
+    bgtz    t7, 0b
+     addiu  a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_\colorid\()_ycc_convert_mips_dspr2)
+
+.purgem DO_RGB_TO_YCC
+
+.endm
+
+/*------------------------------------------id -- pix R  G  B */
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_YCC_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3
+
+/*****************************************************************************/
+/*
+ * jsimd_ycc_extrgb_convert_mips_dspr2
+ * jsimd_ycc_extbgr_convert_mips_dspr2
+ * jsimd_ycc_extrgbx_convert_mips_dspr2
+ * jsimd_ycc_extbgrx_convert_mips_dspr2
+ * jsimd_ycc_extxbgr_convert_mips_dspr2
+ * jsimd_ycc_extxrgb_convert_mips_dspr2
+ *
+ * Colorspace conversion YCbCr -> RGB
+ */
+
+.macro GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs, a_offs
+
+.macro STORE_YCC_TO_RGB  scratch0 \
+                         scratch1 \
+                         scratch2 \
+                         outptr
+    sb       \scratch0, \r_offs(\outptr)
+    sb       \scratch1, \g_offs(\outptr)
+    sb       \scratch2, \b_offs(\outptr)
+.if (\pixel_size == 4)
+    li       t0, 0xFF
+    sb       t0, \a_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_ycc_\colorid\()_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - input_row
+ * a3     - output_buf
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw         s1, 48(sp)
+    li         t3, 0x8000
+    li         t4, 0x166e9     // FIX(1.40200)
+    li         t5, 0x1c5a2     // FIX(1.77200)
+    li         t6, 0xffff492e  // -FIX(0.71414)
+    li         t7, 0xffffa7e6  // -FIX(0.34414)
+    repl.ph    t8, 128
+
+0:
+    lw         s0, 0(a3)
+    lw         t0, 0(a1)
+    lw         t1, 4(a1)
+    lw         t2, 8(a1)
+    sll        s5, a2, 2
+    addiu      s1, -1
+    lwx        s2, s5(t0)
+    lwx        s3, s5(t1)
+    lwx        s4, s5(t2)
+    addu       t9, s2, a0
+    addiu      a2, 1
+
+1:
+    lbu        s7, 0(s4)       // cr
+    lbu        s6, 0(s3)       // cb
+    lbu        s5, 0(s2)       // y
+    addiu      s2, 1
+    addiu      s4, 1
+    addiu      s7, -128
+    addiu      s6, -128
+    mul        t2, t7, s6
+    mul        t0, t6, s7      // Crgtab[cr]
+    sll        s7, 15
+    mulq_rs.w  t1, t4, s7      // Crrtab[cr]
+    sll        s6, 15
+    addu       t2, t3          // Cbgtab[cb]
+    addu       t2, t0
+
+    mulq_rs.w  t0, t5, s6      // Cbbtab[cb]
+    sra        t2, 16
+    addu       t1, s5
+    addu       t2, s5          // add y
+    ins        t2, t1, 16, 16
+    subu.ph    t2, t2, t8
+    addu       t0, s5
+    shll_s.ph  t2, t2, 8
+    subu       t0, 128
+    shra.ph    t2, t2, 8
+    shll_s.w   t0, t0, 24
+    addu.ph    t2, t2, t8      // clip & store
+    sra        t0, t0, 24
+    sra        t1, t2, 16
+    addiu      t0, 128
+
+    STORE_YCC_TO_RGB t1, t2, t0, s0
+
+    bne        s2, t9, 1b
+     addiu     s3, 1
+    bgtz       s1, 0b
+     addiu     a3, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_ycc_\colorid\()_convert_mips_dspr2)
+
+.purgem STORE_YCC_TO_RGB
+
+.endm
+
+/*------------------------------------------id -- pix R  G  B  A */
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0, 3
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1, 0
+GENERATE_JSIMD_YCC_RGB_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3, 0
+
+/*****************************************************************************/
+/*
+ * jsimd_extrgb_gray_convert_mips_dspr2
+ * jsimd_extbgr_gray_convert_mips_dspr2
+ * jsimd_extrgbx_gray_convert_mips_dspr2
+ * jsimd_extbgrx_gray_convert_mips_dspr2
+ * jsimd_extxbgr_gray_convert_mips_dspr2
+ * jsimd_extxrgb_gray_convert_mips_dspr2
+ *
+ * Colorspace conversion RGB -> GRAY
+ */
+
+.macro GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 colorid, pixel_size, r_offs, g_offs, b_offs
+
+.macro DO_RGB_TO_GRAY r,    \
+                      g,    \
+                      b,    \
+                      inptr
+    lbu     \r, \r_offs(\inptr)
+    lbu     \g, \g_offs(\inptr)
+    lbu     \b, \b_offs(\inptr)
+    addiu   \inptr, \pixel_size
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_\colorid\()_gray_convert_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - input_buf
+ * a2     - output_buf
+ * a3     - output_row
+ * 16(sp) - num_rows
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    li      s0, 0x4c8b             // s0 = FIX(0.29900)
+    li      s1, 0x9646             // s1 = FIX(0.58700)
+    li      s2, 0x1d2f             // s2 = FIX(0.11400)
+    li      s7, 0x8000             // s7 = FIX(0.50000)
+    lw      s6, 48(sp)
+    andi    t7, a0, 3
+
+0:
+    addiu   s6, -1                 // s6 = num_rows
+    lw      t0, 0(a1)
+    lw      t1, 0(a2)
+    sll     t3, a3, 2
+    lwx     t1, t3(t1)
+    addiu   a3, 1
+    addu    t9, t1, a0
+    subu    t8, t9, t7
+    beq     t1, t8, 2f
+     nop
+
+1:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    maddu   $ac0, s0, t3
+    mtlo    s7, $ac1
+    maddu   $ac1, s2, s5
+    maddu   $ac1, s1, s4
+    maddu   $ac1, s0, s3
+    extr.w  t6, $ac0, 16
+
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+    DO_RGB_TO_GRAY s3, s4, s5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    extr.w  t2, $ac1, 16
+    maddu   $ac0, s0, t3
+    mtlo    s7, $ac1
+    maddu   $ac1, s2, s5
+    maddu   $ac1, s1, s4
+    maddu   $ac1, s0, s3
+    extr.w  t5, $ac0, 16
+    sb      t6, 0(t1)
+    sb      t2, 1(t1)
+    extr.w  t3, $ac1, 16
+    addiu   t1, 4
+    sb      t5, -2(t1)
+    sb      t3, -1(t1)
+    bne     t1, t8, 1b
+     nop
+
+2:
+    beqz    t7, 4f
+     nop
+
+3:
+    DO_RGB_TO_GRAY t3, t4, t5, t0
+
+    mtlo    s7, $ac0
+    maddu   $ac0, s2, t5
+    maddu   $ac0, s1, t4
+    maddu   $ac0, s0, t3
+    extr.w  t6, $ac0, 16
+    sb      t6, 0(t1)
+    addiu   t1, 1
+    bne     t1, t9, 3b
+     nop
+
+4:
+    bgtz    s6, 0b
+     addiu  a1, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j ra
+     nop
+END(jsimd_\colorid\()_gray_convert_mips_dspr2)
+
+.purgem DO_RGB_TO_GRAY
+
+.endm
+
+/*------------------------------------------id --  pix R  G  B */
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgb,  3, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgr,  3, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extrgbx, 4, 0, 1, 2
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extbgrx, 4, 2, 1, 0
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxbgr, 4, 3, 2, 1
+GENERATE_JSIMD_RGB_GRAY_CONVERT_MIPS_DSPR2 extxrgb, 4, 1, 2, 3
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extrgb_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extrgbx_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extbgrx_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extxbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v2_extxrgb_merged_upsample_mips_dspr2
+ *
+ * Merged h2v2 upsample routines
+ */
+.macro GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 colorid,    \
+                                                pixel_size, \
+                                                r1_offs,    \
+                                                g1_offs,    \
+                                                b1_offs,    \
+                                                a1_offs,    \
+                                                r2_offs,    \
+                                                g2_offs,    \
+                                                b2_offs,    \
+                                                a2_offs
+
+.macro STORE_H2V2_2_PIXELS  scratch0 \
+                            scratch1 \
+                            scratch2 \
+                            scratch3 \
+                            scratch4 \
+                            scratch5 \
+                            outptr
+    sb       \scratch0, \r1_offs(\outptr)
+    sb       \scratch1, \g1_offs(\outptr)
+    sb       \scratch2, \b1_offs(\outptr)
+    sb       \scratch3, \r2_offs(\outptr)
+    sb       \scratch4, \g2_offs(\outptr)
+    sb       \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li       \scratch0, 0xFF
+    sb       \scratch0, \a1_offs(\outptr)
+    sb       \scratch0, \a2_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V2_1_PIXEL  scratch0 \
+                           scratch1 \
+                           scratch2 \
+                           outptr
+    sb    \scratch0, \r1_offs(\outptr)
+    sb    \scratch1, \g1_offs(\outptr)
+    sb    \scratch2, \b1_offs(\outptr)
+
+.if (\pixel_size == 8)
+    li    t0, 0xFF
+    sb    t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->output_width
+ * a1     - input_buf
+ * a2     - in_row_group_ctr
+ * a3     - output_buf
+ * 16(sp) - cinfo->sample_range_limit
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    lw           t9, 56(sp)        // cinfo->sample_range_limit
+    lw           v0, 0(a1)
+    lw           v1, 4(a1)
+    lw           t0, 8(a1)
+    sll          t1, a2, 3
+    addiu        t2, t1, 4
+    sll          t3, a2, 2
+    lw           t4, 0(a3)         // t4 = output_buf[0]
+    lwx          t1, t1(v0)        // t1 = input_buf[0][in_row_group_ctr*2]
+    lwx          t2, t2(v0)        // t2 = input_buf[0][in_row_group_ctr*2 + 1]
+    lwx          t5, t3(v1)        // t5 = input_buf[1][in_row_group_ctr]
+    lwx          t6, t3(t0)        // t6 = input_buf[2][in_row_group_ctr]
+    lw           t7, 4(a3)         // t7 = output_buf[1]
+    li           s1, 0xe6ea
+    addiu        t8, s1, 0x7fff    // t8 = 0x166e9 [FIX(1.40200)]
+    addiu        s0, t8, 0x5eb9    // s0 = 0x1c5a2 [FIX(1.77200)]
+    addiu        s1, zero, 0xa7e6  // s4 = 0xffffa7e6 [-FIX(0.34414)]
+    xori         s2, s1, 0xeec8    // s3 = 0xffff492e [-FIX(0.71414)]
+    srl          t3, a0, 1
+    blez         t3, 2f
+     addu        t0, t5, t3        // t0 = end address
+ 1:
+    lbu          t3, 0(t5)
+    lbu          s3, 0(t6)
+    addiu        t5, t5, 1
+    addiu        t3, t3, -128      // (cb - 128)
+    addiu        s3, s3, -128      // (cr - 128)
+    mult         $ac1, s1, t3
+    madd         $ac1, s2, s3
+    sll          s3, s3, 15
+    sll          t3, t3, 15
+    mulq_rs.w    s4, t8, s3        // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS
+    extr_r.w     s5, $ac1, 16
+    mulq_rs.w    s6, s0, t3        // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS
+    lbu          v0, 0(t1)
+    addiu        t6, t6, 1
+    addiu        t1, t1, 2
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          AT, 0(t3)
+    lbu          s7, 0(s3)
+    lbu          ra, 0(v1)
+    lbu          v0, -1(t1)
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+    lbu          v0, 0(t2)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t4
+
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          AT, 0(t3)
+    lbu          s7, 0(s3)
+    lbu          ra, 0(v1)
+    lbu          v0, 1(t2)
+    addiu        t2, t2, 2
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+
+    STORE_H2V2_2_PIXELS AT, s7, ra, t3, s3, v1, t7
+
+    bne          t0, t5, 1b
+     nop
+2:
+    andi         t0, a0, 1
+    beqz         t0, 4f
+     lbu          t3, 0(t5)
+    lbu          s3, 0(t6)
+    addiu        t3, t3, -128      // (cb - 128)
+    addiu        s3, s3, -128      // (cr - 128)
+    mult         $ac1, s1, t3
+    madd         $ac1, s2, s3
+    sll          s3, s3, 15
+    sll          t3, t3, 15
+    lbu          v0, 0(t1)
+    extr_r.w     s5, $ac1, 16
+    mulq_rs.w    s4, t8, s3        // s4 = (C1 * cr + ONE_HALF)>> SCALEBITS
+    mulq_rs.w    s6, s0, t3        // s6 = (C2 * cb + ONE_HALF)>> SCALEBITS
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+    lbu          v0, 0(t2)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t4
+
+    addu         t3, v0, s4        // y+cred
+    addu         s3, v0, s5        // y+cgreen
+    addu         v1, v0, s6        // y+cblue
+    addu         t3, t9, t3        // y+cred
+    addu         s3, t9, s3        // y+cgreen
+    addu         v1, t9, v1        // y+cblue
+    lbu          t3, 0(t3)
+    lbu          s3, 0(s3)
+    lbu          v1, 0(v1)
+
+    STORE_H2V2_1_PIXEL t3, s3, v1, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_\colorid\()_merged_upsample_mips_dspr2)
+
+.purgem STORE_H2V2_1_PIXEL
+.purgem STORE_H2V2_2_PIXELS
+.endm
+
+/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V2_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+/*****************************************************************************/
+/*
+ * jsimd_h2v1_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extrgb_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extrgbx_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extbgrx_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extxbgr_merged_upsample_mips_dspr2
+ * jsimd_h2v1_extxrgb_merged_upsample_mips_dspr2
+ *
+ * Merged h2v1 upsample routines
+ */
+
+.macro GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 colorid,    \
+                                                pixel_size, \
+                                                r1_offs,    \
+                                                g1_offs,    \
+                                                b1_offs,    \
+                                                a1_offs,    \
+                                                r2_offs,    \
+                                                g2_offs,    \
+                                                b2_offs,    \
+                                                a2_offs
+
+.macro STORE_H2V1_2_PIXELS  scratch0 \
+                            scratch1 \
+                            scratch2 \
+                            scratch3 \
+                            scratch4 \
+                            scratch5 \
+                            outptr
+    sb       \scratch0, \r1_offs(\outptr)
+    sb       \scratch1, \g1_offs(\outptr)
+    sb       \scratch2, \b1_offs(\outptr)
+    sb       \scratch3, \r2_offs(\outptr)
+    sb       \scratch4, \g2_offs(\outptr)
+    sb       \scratch5, \b2_offs(\outptr)
+.if (\pixel_size == 8)
+    li       t0, 0xFF
+    sb       t0, \a1_offs(\outptr)
+    sb       t0, \a2_offs(\outptr)
+.endif
+    addiu    \outptr, \pixel_size
+.endm
+
+.macro STORE_H2V1_1_PIXEL  scratch0 \
+                           scratch1 \
+                           scratch2 \
+                           outptr
+    sb    \scratch0, \r1_offs(\outptr)
+    sb    \scratch1, \g1_offs(\outptr)
+    sb    \scratch2, \b1_offs(\outptr)
+.if (\pixel_size == 8)
+    li    t0, 0xFF
+    sb    t0, \a1_offs(\outptr)
+.endif
+.endm
+
+LEAF_MIPS_DSPR2(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->output_width
+ * a1     - input_buf
+ * a2     - in_row_group_ctr
+ * a3     - output_buf
+ * 16(sp) - range_limit
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    li           t0, 0xe6ea
+    lw           t1, 0(a1)         // t1 = input_buf[0]
+    lw           t2, 4(a1)         // t2 = input_buf[1]
+    lw           t3, 8(a1)         // t3 = input_buf[2]
+    lw           t8, 56(sp)        // t8 = range_limit
+    addiu        s1, t0, 0x7fff    // s1 = 0x166e9 [FIX(1.40200)]
+    addiu        s2, s1, 0x5eb9    // s2 = 0x1c5a2 [FIX(1.77200)]
+    addiu        s0, t0, 0x9916    // s0 = 0x8000
+    addiu        s4, zero, 0xa7e6  // s4 = 0xffffa7e6 [-FIX(0.34414)]
+    xori         s3, s4, 0xeec8    // s3 = 0xffff492e [-FIX(0.71414)]
+    srl          t0, a0, 1
+    sll          t4, a2, 2
+    lwx          s5, t4(t1)        // s5 = inptr0
+    lwx          s6, t4(t2)        // s6 = inptr1
+    lwx          s7, t4(t3)        // s7 = inptr2
+    lw           t7, 0(a3)         // t7 = outptr
+    blez         t0, 2f
+     addu        t9, s6, t0        // t9 = end address
+1:
+    lbu          t2, 0(s6)         // t2 = cb
+    lbu          t0, 0(s7)         // t0 = cr
+    lbu          t1, 0(s5)         // t1 = y
+    addiu        t2, t2, -128      // t2 = cb - 128
+    addiu        t0, t0, -128      // t0 = cr - 128
+    mult         $ac1, s4, t2
+    madd         $ac1, s3, t0
+    sll          t0, t0, 15
+    sll          t2, t2, 15
+    mulq_rs.w    t0, s1, t0        // t0 = (C1*cr + ONE_HALF)>> SCALEBITS
+    extr_r.w     t5, $ac1, 16
+    mulq_rs.w    t6, s2, t2        // t6 = (C2*cb + ONE_HALF)>> SCALEBITS
+    addiu        s7, s7, 1
+    addiu        s6, s6, 1
+    addu         t2, t1, t0        // t2 = y + cred
+    addu         t3, t1, t5        // t3 = y + cgreen
+    addu         t4, t1, t6        // t4 = y + cblue
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t1, 1(s5)
+    lbu          v0, 0(t2)
+    lbu          v1, 0(t3)
+    lbu          ra, 0(t4)
+    addu         t2, t1, t0
+    addu         t3, t1, t5
+    addu         t4, t1, t6
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t2, 0(t2)
+    lbu          t3, 0(t3)
+    lbu          t4, 0(t4)
+
+    STORE_H2V1_2_PIXELS v0, v1, ra, t2, t3, t4, t7
+
+    bne          t9, s6, 1b
+     addiu       s5, s5, 2
+2:
+    andi         t0, a0, 1
+    beqz         t0, 4f
+     nop
+3:
+    lbu          t2, 0(s6)
+    lbu          t0, 0(s7)
+    lbu          t1, 0(s5)
+    addiu        t2, t2, -128      //(cb - 128)
+    addiu        t0, t0, -128      //(cr - 128)
+    mul          t3, s4, t2
+    mul          t4, s3, t0
+    sll          t0, t0, 15
+    sll          t2, t2, 15
+    mulq_rs.w    t0, s1, t0       // (C1*cr + ONE_HALF)>> SCALEBITS
+    mulq_rs.w    t6, s2, t2       // (C2*cb + ONE_HALF)>> SCALEBITS
+    addu         t3, t3, s0
+    addu         t3, t4, t3
+    sra          t5, t3, 16       // (C4*cb + ONE_HALF + C3*cr)>> SCALEBITS
+    addu         t2, t1, t0       // y + cred
+    addu         t3, t1, t5       // y + cgreen
+    addu         t4, t1, t6       // y + cblue
+    addu         t2, t8, t2
+    addu         t3, t8, t3
+    addu         t4, t8, t4
+    lbu          t2, 0(t2)
+    lbu          t3, 0(t3)
+    lbu          t4, 0(t4)
+
+    STORE_H2V1_1_PIXEL t2, t3, t4, t7
+4:
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, ra
+
+    j            ra
+     nop
+
+END(jsimd_h2v1_\colorid\()_merged_upsample_mips_dspr2)
+
+.purgem STORE_H2V1_1_PIXEL
+.purgem STORE_H2V1_2_PIXELS
+.endm
+
+/*-----------------------------------------id -- pix R1 G1 B1 A1 R2 G2 B2 A2 */
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgb,  6, 0, 1, 2, 6, 3, 4, 5, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgr,  6, 2, 1, 0, 3, 5, 4, 3, 6
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extrgbx, 8, 0, 1, 2, 3, 4, 5, 6, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extbgrx, 8, 2, 1, 0, 3, 6, 5, 4, 7
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxbgr, 8, 3, 2, 1, 0, 7, 6, 5, 4
+GENERATE_H2V1_MERGED_UPSAMPLE_MIPS_DSPR2 extxrgb, 8, 1, 2, 3, 0, 5, 6, 7, 4
+/*****************************************************************************/
+/*
+ * jsimd_h2v2_fancy_upsample_mips_dspr2
+ *
+ * Fancy processing for the common case of 2:1 horizontal and 2:1 vertical.
+ */
+LEAF_MIPS_DSPR2(jsimd_h2v2_fancy_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - downsampled_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    li             s4, 0
+    lw             s2, 0(a3)       // s2 = *output_data_ptr
+0:
+    li             t9, 2
+    lw             s1, -4(a2)      // s1 = inptr1
+
+1:
+    lw             s0, 0(a2)       // s0 = inptr0
+    lwx            s3, s4(s2)
+    addiu          s5, a1, -2      // s5 = downsampled_width - 2
+    srl            t4, s5, 1
+    sll            t4, t4, 1
+    lbu            t0, 0(s0)
+    lbu            t1, 1(s0)
+    lbu            t2, 0(s1)
+    lbu            t3, 1(s1)
+    addiu          s0, 2
+    addiu          s1, 2
+    addu           t8, s0, t4      // t8 = end address
+    andi           s5, s5, 1       // s5 = residual
+    sll            t4, t0, 1
+    sll            t6, t1, 1
+    addu           t0, t0, t4      // t0 = (*inptr0++) * 3
+    addu           t1, t1, t6      // t1 = (*inptr0++) * 3
+    addu           t7, t0, t2      // t7 = thiscolsum
+    addu           t6, t1, t3      // t5 = nextcolsum
+    sll            t0, t7, 2       // t0 = thiscolsum * 4
+    subu           t1, t0, t7      // t1 = thiscolsum * 3
+    shra_r.w       t0, t0, 4
+    addiu          t1, 7
+    addu           t1, t1, t6
+    srl            t1, t1, 4
+    sb             t0, 0(s3)
+    sb             t1, 1(s3)
+    beq            t8, s0, 22f     // skip to final iteration if width == 3
+     addiu          s3, 2
+2:
+    lh             t0, 0(s0)       // t0 = A3|A2
+    lh             t2, 0(s1)       // t2 = B3|B2
+    addiu          s0, 2
+    addiu          s1, 2
+    preceu.ph.qbr  t0, t0          // t0 = 0|A3|0|A2
+    preceu.ph.qbr  t2, t2          // t2 = 0|B3|0|B2
+    shll.ph        t1, t0, 1
+    sll            t3, t6, 1
+    addu.ph        t0, t1, t0      // t0 = A3*3|A2*3
+    addu           t3, t3, t6      // t3 = this * 3
+    addu.ph        t0, t0, t2      // t0 = next2|next1
+    addu           t1, t3, t7
+    andi           t7, t0, 0xFFFF  // t7 = next1
+    sll            t2, t7, 1
+    addu           t2, t7, t2      // t2 = next1*3
+    addu           t4, t2, t6
+    srl            t6, t0, 16      // t6 = next2
+    shra_r.w       t1, t1, 4       // t1 = (this*3 + last + 8) >> 4
+    addu           t0, t3, t7
+    addiu          t0, 7
+    srl            t0, t0, 4       // t0 = (this*3 + next1 + 7) >> 4
+    shra_r.w       t4, t4, 4       // t3 = (next1*3 + this + 8) >> 4
+    addu           t2, t2, t6
+    addiu          t2, 7
+    srl            t2, t2, 4       // t2 = (next1*3 + next2 + 7) >> 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    sb             t4, 2(s3)
+    sb             t2, 3(s3)
+    bne            t8, s0, 2b
+     addiu         s3, 4
+22:
+    beqz           s5, 4f
+     addu          t8, s0, s5
+3:
+    lbu            t0, 0(s0)
+    lbu            t2, 0(s1)
+    addiu          s0, 1
+    addiu          s1, 1
+    sll            t3, t6, 1
+    sll            t1, t0, 1
+    addu           t1, t0, t1      // t1 = inptr0 * 3
+    addu           t3, t3, t6      // t3 = thiscolsum * 3
+    addu           t5, t1, t2
+    addu           t1, t3, t7
+    shra_r.w       t1, t1, 4
+    addu           t0, t3, t5
+    addiu          t0, 7
+    srl            t0, t0, 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    addiu          s3, 2
+    move           t7, t6
+    bne            t8, s0, 3b
+     move          t6, t5
+4:
+    sll            t0, t6, 2       // t0 = thiscolsum * 4
+    subu           t1, t0, t6      // t1 = thiscolsum * 3
+    addu           t1, t1, t7
+    addiu          s4, 4
+    shra_r.w       t1, t1, 4
+    addiu          t0, 7
+    srl            t0, t0, 4
+    sb             t1, 0(s3)
+    sb             t0, 1(s3)
+    addiu          t9, -1
+    addiu          s3, 2
+    bnez           t9, 1b
+     lw            s1, 4(a2)
+    srl            t0, s4, 2
+    subu           t0, a0, t0
+    bgtz           t0, 0b
+     addiu         a2, 4
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j ra
+     nop
+END(jsimd_h2v2_fancy_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_fancy_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - downsampled_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    .set at
+
+    beqz           a0, 3f
+     sll           t0, a0, 2
+    lw             s1, 0(a3)
+    li             s3, 0x10001
+    addu           s0, s1, t0
+0:
+    addiu          t8, a1, -2
+    srl            t9, t8, 2
+    lw             t7, 0(a2)
+    lw             s2, 0(s1)
+    lbu            t0, 0(t7)
+    lbu            t1, 1(t7)   // t1 = inptr[1]
+    sll            t2, t0, 1
+    addu           t2, t2, t0  // t2 = invalue*3
+    addu           t2, t2, t1
+    shra_r.w       t2, t2, 2
+    sb             t0, 0(s2)
+    sb             t2, 1(s2)
+    beqz           t9, 11f
+     addiu         s2, 2
+1:
+    ulw            t0, 0(t7)   // t0 = |P3|P2|P1|P0|
+    ulw            t1, 1(t7)
+    ulh            t2, 4(t7)   // t2 = |0|0|P5|P4|
+    preceu.ph.qbl  t3, t0      // t3 = |0|P3|0|P2|
+    preceu.ph.qbr  t0, t0      // t0 = |0|P1|0|P0|
+    preceu.ph.qbr  t2, t2      // t2 = |0|P5|0|P4|
+    preceu.ph.qbl  t4, t1      // t4 = |0|P4|0|P3|
+    preceu.ph.qbr  t1, t1      // t1 = |0|P2|0|P1|
+    shll.ph        t5, t4, 1
+    shll.ph        t6, t1, 1
+    addu.ph        t5, t5, t4  // t5 = |P4*3|P3*3|
+    addu.ph        t6, t6, t1  // t6 = |P2*3|P1*3|
+    addu.ph        t4, t3, s3
+    addu.ph        t0, t0, s3
+    addu.ph        t4, t4, t5
+    addu.ph        t0, t0, t6
+    shrl.ph        t4, t4, 2   // t4 = |0|P3|0|P2|
+    shrl.ph        t0, t0, 2   // t0 = |0|P1|0|P0|
+    addu.ph        t2, t2, t5
+    addu.ph        t3, t3, t6
+    shra_r.ph      t2, t2, 2   // t2 = |0|P5|0|P4|
+    shra_r.ph      t3, t3, 2   // t3 = |0|P3|0|P2|
+    shll.ph        t2, t2, 8
+    shll.ph        t3, t3, 8
+    or             t2, t4, t2
+    or             t3, t3, t0
+    addiu          t9, -1
+    usw            t3, 0(s2)
+    usw            t2, 4(s2)
+    addiu          s2, 8
+    bgtz           t9, 1b
+     addiu         t7, 4
+11:
+    andi           t8, 3
+    beqz           t8, 22f
+     addiu         t7, 1
+
+2:
+    lbu            t0, 0(t7)
+    addiu          t7, 1
+    sll            t1, t0, 1
+    addu           t2, t0, t1  // t2 = invalue
+    lbu            t3, -2(t7)
+    lbu            t4, 0(t7)
+    addiu          t3, 1
+    addiu          t4, 2
+    addu           t3, t3, t2
+    addu           t4, t4, t2
+    srl            t3, 2
+    srl            t4, 2
+    sb             t3, 0(s2)
+    sb             t4, 1(s2)
+    addiu          t8, -1
+    bgtz           t8, 2b
+     addiu         s2, 2
+
+22:
+    lbu            t0, 0(t7)
+    lbu            t2, -1(t7)
+    sll            t1, t0, 1
+    addu           t1, t1, t0 // t1 = invalue * 3
+    addu           t1, t1, t2
+    addiu          t1, 1
+    srl            t1, t1, 2
+    sb             t1, 0(s2)
+    sb             t0, 1(s2)
+    addiu          s1, 4
+    bne            s1, s0, 0b
+     addiu         a2, 4
+3:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j              ra
+     nop
+END(jsimd_h2v1_fancy_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_downsample_mips_dspr2)
+/*
+ * a0     - cinfo->image_width
+ * a1     - cinfo->max_v_samp_factor
+ * a2     - compptr->v_samp_factor
+ * a3     - compptr->width_in_blocks
+ * 16(sp) - input_data
+ * 20(sp) - output_data
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4
+
+    beqz        a2, 7f
+     lw         s1, 44(sp)  // s1 = output_data
+    lw          s0, 40(sp)  // s0 = input_data
+    srl         s2, a0, 2
+    andi        t9, a0, 2
+    srl         t7, t9, 1
+    addu        s2, t7, s2
+    sll         t0, a3, 3   // t0 = width_in_blocks*DCT
+    srl         t7, t0, 1
+    subu        s2, t7, s2
+0:
+    andi        t6, a0, 1   // t6 = temp_index
+    addiu       t6, -1
+    lw          t4, 0(s1)   // t4 = outptr
+    lw          t5, 0(s0)   // t5 = inptr0
+    li          s3, 0       // s3 = bias
+    srl         t7, a0, 1   // t7 = image_width1
+    srl         s4, t7, 2
+    andi        t8, t7, 3
+1:
+    ulhu        t0, 0(t5)
+    ulhu        t1, 2(t5)
+    ulhu        t2, 4(t5)
+    ulhu        t3, 6(t5)
+    raddu.w.qb  t0, t0
+    raddu.w.qb  t1, t1
+    raddu.w.qb  t2, t2
+    raddu.w.qb  t3, t3
+    shra.ph     t0, t0, 1
+    shra_r.ph   t1, t1, 1
+    shra.ph     t2, t2, 1
+    shra_r.ph   t3, t3, 1
+    sb          t0, 0(t4)
+    sb          t1, 1(t4)
+    sb          t2, 2(t4)
+    sb          t3, 3(t4)
+    addiu       s4, -1
+    addiu       t4, 4
+    bgtz        s4, 1b
+     addiu      t5, 8
+    beqz        t8, 3f
+     addu       s4, t4, t8
+2:
+    ulhu        t0, 0(t5)
+    raddu.w.qb  t0, t0
+    addqh.w     t0, t0, s3
+    xori        s3, s3, 1
+    sb          t0, 0(t4)
+    addiu       t4, 1
+    bne         t4, s4, 2b
+     addiu      t5, 2
+3:
+    lbux        t1, t6(t5)
+    sll         t1, 1
+    addqh.w     t2, t1, s3  // t2 = pixval1
+    xori        s3, s3, 1
+    addqh.w     t3, t1, s3  // t3 = pixval2
+    blez        s2, 5f
+     append     t3, t2,  8
+    addu        t5, t4, s2  // t5 = loop_end2
+4:
+    ush         t3, 0(t4)
+    addiu       s2, -1
+    bgtz        s2, 4b
+     addiu      t4,  2
+5:
+    beqz        t9, 6f
+     nop
+    sb          t2, 0(t4)
+6:
+    addiu       s1, 4
+    addiu       a2, -1
+    bnez        a2, 0b
+     addiu      s0, 4
+7:
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4
+
+    j           ra
+    nop
+END(jsimd_h2v1_downsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_downsample_mips_dspr2)
+
+/*
+ * a0     - cinfo->image_width
+ * a1     - cinfo->max_v_samp_factor
+ * a2     - compptr->v_samp_factor
+ * a3     - compptr->width_in_blocks
+ * 16(sp) - input_data
+ * 20(sp) - output_data
+ */
+    .set at
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    beqz         a2, 8f
+     lw          s1, 52(sp)      // s1 = output_data
+    lw           s0, 48(sp)      // s0 = input_data
+
+    andi         t6, a0, 1       // t6 = temp_index
+    addiu        t6, -1
+    srl          t7, a0, 1       // t7 = image_width1
+    srl          s4, t7, 2
+    andi         t8, t7, 3
+    andi         t9, a0, 2
+    srl          s2, a0, 2
+    srl          t7, t9, 1
+    addu         s2, t7, s2
+    sll          t0, a3, 3       // s2 = width_in_blocks*DCT
+    srl          t7, t0, 1
+    subu         s2, t7, s2
+0:
+    lw           t4, 0(s1)       // t4 = outptr
+    lw           t5, 0(s0)       // t5 = inptr0
+    lw           s7, 4(s0)       // s7 = inptr1
+    li           s6, 1           // s6 = bias
+2:
+    ulw          t0, 0(t5)       // t0 = |P3|P2|P1|P0|
+    ulw          t1, 0(s7)       // t1 = |Q3|Q2|Q1|Q0|
+    ulw          t2, 4(t5)
+    ulw          t3, 4(s7)
+    precrq.ph.w  t7, t0, t1      // t2 = |P3|P2|Q3|Q2|
+    ins          t0, t1, 16, 16  // t0 = |Q1|Q0|P1|P0|
+    raddu.w.qb   t1, t7
+    raddu.w.qb   t0, t0
+    shra_r.w     t1, t1, 2
+    addiu        t0, 1
+    srl          t0, 2
+    precrq.ph.w  t7, t2, t3
+    ins          t2, t3, 16, 16
+    raddu.w.qb   t7, t7
+    raddu.w.qb   t2, t2
+    shra_r.w     t7, t7, 2
+    addiu        t2, 1
+    srl          t2, 2
+    sb           t0, 0(t4)
+    sb           t1, 1(t4)
+    sb           t2, 2(t4)
+    sb           t7, 3(t4)
+    addiu        t4, 4
+    addiu        t5, 8
+    addiu        s4, s4, -1
+    bgtz         s4, 2b
+     addiu       s7, 8
+    beqz         t8, 4f
+     addu        t8, t4, t8
+3:
+    ulhu         t0, 0(t5)
+    ulhu         t1, 0(s7)
+    ins          t0, t1, 16, 16
+    raddu.w.qb   t0, t0
+    addu         t0, t0, s6
+    srl          t0, 2
+    xori         s6, s6, 3
+    sb           t0, 0(t4)
+    addiu        t5, 2
+    addiu        t4, 1
+    bne          t8, t4, 3b
+     addiu       s7, 2
+4:
+    lbux         t1, t6(t5)
+    sll          t1, 1
+    lbux         t0, t6(s7)
+    sll          t0, 1
+    addu         t1, t1, t0
+    addu         t3, t1, s6
+    srl          t0, t3, 2       // t2 = pixval1
+    xori         s6, s6, 3
+    addu         t2, t1, s6
+    srl          t1, t2, 2       // t3 = pixval2
+    blez         s2, 6f
+     append      t1, t0, 8
+5:
+    ush          t1, 0(t4)
+    addiu        s2, -1
+    bgtz         s2, 5b
+     addiu       t4, 2
+6:
+    beqz         t9, 7f
+     nop
+    sb           t0, 0(t4)
+7:
+    addiu        s1, 4
+    addiu        a2, -1
+    bnez         a2, 0b
+     addiu       s0, 8
+8:
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j            ra
+     nop
+END(jsimd_h2v2_downsample_mips_dspr2)
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_smooth_downsample_mips_dspr2)
+/*
+ * a0     - input_data
+ * a1     - output_data
+ * a2     - compptr->v_samp_factor
+ * a3     - cinfo->max_v_samp_factor
+ * 16(sp) - cinfo->smoothing_factor
+ * 20(sp) - compptr->width_in_blocks
+ * 24(sp) - cinfo->image_width
+ */
+
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw          s7, 52(sp)      // compptr->width_in_blocks
+    lw          s0, 56(sp)      // cinfo->image_width
+    lw          s6, 48(sp)      // cinfo->smoothing_factor
+    sll         s7, 3           // output_cols = width_in_blocks * DCTSIZE
+    sll         v0, s7, 1
+    subu        v0, v0, s0
+    blez        v0, 2f
+    move        v1, zero
+    addiu       t0, a3, 2       // t0 = cinfo->max_v_samp_factor + 2
+0:
+    addiu       t1, a0, -4
+    sll         t2, v1, 2
+    lwx         t1, t2(t1)
+    move        t3, v0
+    addu        t1, t1, s0
+    lbu         t2, -1(t1)
+1:
+    addiu       t3, t3, -1
+    sb          t2, 0(t1)
+    bgtz        t3, 1b
+    addiu       t1, t1, 1
+    addiu       v1, v1, 1
+    bne         v1, t0, 0b
+    nop
+2:
+    li          v0, 80
+    mul         v0, s6, v0
+    li          v1, 16384
+    move        t4, zero
+    move        t5, zero
+    subu        t6, v1, v0      // t6 = 16384 - tmp_smoot_f * 80
+    sll         t7, s6, 4       // t7 = tmp_smoot_f * 16
+3:
+/* Special case for first column: pretend column -1 is same as column 0 */
+    sll         v0, t4, 2
+    lwx         t8, v0(a1)      //  outptr = output_data[outrow]
+    sll         v1, t5, 2
+    addiu       t9, v1, 4
+    addiu       s0, v1, -4
+    addiu       s1, v1, 8
+    lwx         s2, v1(a0)      // inptr0 = input_data[inrow]
+    lwx         t9, t9(a0)      // inptr1 = input_data[inrow+1]
+    lwx         s0, s0(a0)      // above_ptr = input_data[inrow-1]
+    lwx         s1, s1(a0)      // below_ptr = input_data[inrow+2]
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, 0(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, 0(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1,t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, 0(s0)
+    lbu         t0, 0(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1,s3, t7
+    extr_r.w    v0, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    addiu       s1, s1, 2
+    sb          v0, -1(t8)
+    addiu       s4, s7, -2
+    and         s4, s4, 3
+    addu        s5, s4, t8      //end adress
+4:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    addiu       t8, t8, 1
+    addiu       s2, s2, 2
+    addiu       t9, t9, 2
+    addiu       s0, s0, 2
+    sb          t2, -1(t8)
+    bne         s5, t8, 4b
+    addiu       s1, s1, 2
+    addiu       s5, s7, -2
+    subu        s5, s5, s4
+    addu        s5, s5, t8      //end adress
+5:
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 2(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 2(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 2(s0)
+    addu        t0, t0, v0
+    lbu         t3, 2(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    lh          v1, 2(t9)
+    addu        t0, t0, v0
+    lh          v0, 2(s2)
+    addu        s3, t0, s3
+    lh          t0, 2(s0)
+    lh          t1, 2(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 4(s2)
+    lbu         t0, 1(t9)
+    lbu         t1, 4(t9)
+    sb          t2, 0(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 1(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 4(s0)
+    addu        t0, t0, v0
+    lbu         v0, 1(s0)
+    addu        s3, t0, s3
+    lbu         t0, 1(s1)
+    lbu         t3, 4(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 4(t9)
+    addu        t0, t0, v0
+    lh          v0, 4(s2)
+    addu        s3, t0, s3
+    lh          t0, 4(s0)
+    lh          t1, 4(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t2, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 6(s2)
+    lbu         t0, 3(t9)
+    lbu         t1, 6(t9)
+    sb          t2, 1(t8)
+    raddu.w.qb  t3, v0
+    lbu         v0, 3(s2)
+    addu        t0, t0,t1
+    mult        $ac1, t3, t6
+    addu        v0, v0, v1
+    lbu         t2, 6(s0)
+    addu        t0, t0, v0
+    lbu         v0, 3(s0)
+    addu        s3, t0, s3
+    lbu         t0, 3(s1)
+    lbu         t3, 6(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    lh          v1, 6(t9)
+    addu        t0, t0, v0
+    lh          v0, 6(s2)
+    addu        s3, t0, s3
+    lh          t0, 6(s0)
+    lh          t1, 6(s1)
+    madd        $ac1, s3, t7
+    extr_r.w    t3, $ac1, 16
+    ins         t0, t1, 16, 16
+    ins         v0, v1, 16, 16
+    raddu.w.qb  s3, t0
+    lbu         v1, 8(s2)
+    lbu         t0, 5(t9)
+    lbu         t1, 8(t9)
+    sb          t3, 2(t8)
+    raddu.w.qb  t2, v0
+    lbu         v0, 5(s2)
+    addu        t0, t0, t1
+    mult        $ac1, t2, t6
+    addu        v0, v0, v1
+    lbu         t2, 8(s0)
+    addu        t0, t0, v0
+    lbu         v0, 5(s0)
+    addu        s3, t0, s3
+    lbu         t0, 5(s1)
+    lbu         t3, 8(s1)
+    addu        v0, v0, t2
+    sll         s3, s3, 1
+    addu        t0, t0, t3
+    addiu       t8, t8, 4
+    addu        t0, t0, v0
+    addiu       s2, s2, 8
+    addu        s3, t0, s3
+    addiu       t9, t9, 8
+    madd        $ac1, s3, t7
+    extr_r.w    t1, $ac1, 16
+    addiu       s0, s0, 8
+    addiu       s1, s1, 8
+    bne         s5, t8, 5b
+    sb          t1, -1(t8)
+/* Special case for last column */
+    lh          v0, 0(s2)
+    lh          v1, 0(t9)
+    lh          t0, 0(s0)
+    lh          t1, 0(s1)
+    ins         v0, v1, 16, 16
+    ins         t0, t1, 16, 16
+    raddu.w.qb  t2, v0
+    raddu.w.qb  s3, t0
+    lbu         v0, -1(s2)
+    lbu         v1, 1(s2)
+    lbu         t0, -1(t9)
+    lbu         t1, 1(t9)
+    addu        v0, v0, v1
+    mult        $ac1, t2, t6
+    addu        t0, t0, t1
+    lbu         t2, 1(s0)
+    addu        t0, t0, v0
+    lbu         t3, 1(s1)
+    addu        s3, t0, s3
+    lbu         v0, -1(s0)
+    lbu         t0, -1(s1)
+    sll         s3, s3, 1
+    addu        v0, v0, t2
+    addu        t0, t0, t3
+    addu        t0, t0, v0
+    addu        s3, t0, s3
+    madd        $ac1, s3, t7
+    extr_r.w    t0, $ac1, 16
+    addiu       t5, t5, 2
+    sb          t0, 0(t8)
+    addiu       t4, t4, 1
+    bne         t4, a2, 3b
+    addiu       t5, t5, 2
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j           ra
+     nop
+
+END(jsimd_h2v2_smooth_downsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_int_upsample_mips_dspr2)
+/*
+ * a0     - upsample->h_expand[compptr->component_index]
+ * a1     - upsample->v_expand[compptr->component_index]
+ * a2     - input_data
+ * a3     - output_data_ptr
+ * 16(sp) - cinfo->output_width
+ * 20(sp) - cinfo->max_v_samp_factor
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    lw      s0, 0(a3)    // s0 = output_data
+    lw      s1, 32(sp)   // s1 = cinfo->output_width
+    lw      s2, 36(sp)   // s2 = cinfo->max_v_samp_factor
+    li      t6, 0        // t6 = inrow
+    beqz    s2, 10f
+     li     s3, 0        // s3 = outrow
+0:
+    addu    t0, a2, t6
+    addu    t7, s0, s3
+    lw      t3, 0(t0)    // t3 = inptr
+    lw      t8, 0(t7)    // t8 = outptr
+    beqz    s1, 4f
+     addu   t5, t8, s1   // t5 = outend
+1:
+    lb      t2, 0(t3)    // t2 = invalue = *inptr++
+    addiu   t3, 1
+    beqz    a0, 3f
+     move   t0, a0       // t0 = h_expand
+2:
+    sb      t2, 0(t8)
+    addiu   t0, -1
+    bgtz    t0, 2b
+     addiu  t8, 1
+3:
+    bgt     t5, t8, 1b
+     nop
+4:
+    addiu   t9, a1, -1   // t9 = v_expand - 1
+    blez    t9, 9f
+     nop
+5:
+    lw      t3, 0(s0)
+    lw      t4, 4(s0)
+    subu    t0, s1, 0xF
+    blez    t0, 7f
+     addu   t5, t3, s1   // t5 = end address
+    andi    t7, s1, 0xF  // t7 = residual
+    subu    t8, t5, t7
+6:
+    ulw     t0, 0(t3)
+    ulw     t1, 4(t3)
+    ulw     t2, 8(t3)
+    usw     t0, 0(t4)
+    ulw     t0, 12(t3)
+    usw     t1, 4(t4)
+    usw     t2, 8(t4)
+    usw     t0, 12(t4)
+    addiu   t3, 16
+    bne     t3, t8, 6b
+     addiu  t4, 16
+    beqz    t7, 8f
+     nop
+7:
+    lbu     t0, 0(t3)
+    sb      t0, 0(t4)
+    addiu   t3, 1
+    bne     t3, t5, 7b
+     addiu  t4, 1
+8:
+    addiu   t9, -1
+    bgtz    t9, 5b
+     addiu  s0, 8
+9:
+    addu    s3, s3, a1
+    bne     s3, s2, 0b
+     addiu  t6, 1
+10:
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j       ra
+     nop
+END(jsimd_int_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v1_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - cinfo->output_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+    lw      t7, 0(a3)       // t7 = output_data
+    andi    t8, a1, 0xf     // t8 = residual
+    sll     t0, a0, 2
+    blez    a0, 4f
+     addu   t9, t7, t0      // t9 = output_data end address
+0:
+    lw      t5, 0(t7)       // t5 = outptr
+    lw      t6, 0(a2)       // t6 = inptr
+    addu    t3, t5, a1      // t3 = outptr + output_width (end address)
+    subu    t3, t8          // t3 = end address - residual
+    beq     t5, t3, 2f
+     move   t4, t8
+1:
+    ulw     t0, 0(t6)       // t0 = |P3|P2|P1|P0|
+    ulw     t2, 4(t6)       // t2 = |P7|P6|P5|P4|
+    srl     t1, t0, 16      // t1 = |X|X|P3|P2|
+    ins     t0, t0, 16, 16  // t0 = |P1|P0|P1|P0|
+    ins     t1, t1, 16, 16  // t1 = |P3|P2|P3|P2|
+    ins     t0, t0, 8, 16   // t0 = |P1|P1|P0|P0|
+    ins     t1, t1, 8, 16   // t1 = |P3|P3|P2|P2|
+    usw     t0, 0(t5)
+    usw     t1, 4(t5)
+    srl     t0, t2, 16      // t0 = |X|X|P7|P6|
+    ins     t2, t2, 16, 16  // t2 = |P5|P4|P5|P4|
+    ins     t0, t0, 16, 16  // t0 = |P7|P6|P7|P6|
+    ins     t2, t2, 8, 16   // t2 = |P5|P5|P4|P4|
+    ins     t0, t0, 8, 16   // t0 = |P7|P7|P6|P6|
+    usw     t2, 8(t5)
+    usw     t0, 12(t5)
+    addiu   t5, 16
+    bne     t5, t3, 1b
+     addiu  t6, 8
+    beqz    t8, 3f
+     move   t4, t8
+2:
+    lbu     t1, 0(t6)
+    sb      t1, 0(t5)
+    sb      t1, 1(t5)
+    addiu   t4, -2
+    addiu   t6, 1
+    bgtz    t4, 2b
+     addiu  t5, 2
+3:
+    addiu   t7, 4
+    bne     t9, t7, 0b
+     addiu  a2, 4
+4:
+    j       ra
+     nop
+END(jsimd_h2v1_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_h2v2_upsample_mips_dspr2)
+/*
+ * a0     - cinfo->max_v_samp_factor
+ * a1     - cinfo->output_width
+ * a2     - input_data
+ * a3     - output_data_ptr
+ */
+    lw      t7, 0(a3)
+    blez    a0, 7f
+     andi   t9, a1, 0xf     // t9 = residual
+0:
+    lw      t6, 0(a2)       // t6 = inptr
+    lw      t5, 0(t7)       // t5 = outptr
+    addu    t8, t5, a1      // t8 = outptr end address
+    subu    t8, t9          // t8 = end address - residual
+    beq     t5, t8, 2f
+     move   t4, t9
+1:
+    ulw     t0, 0(t6)
+    srl     t1, t0, 16
+    ins     t0, t0, 16, 16
+    ins     t0, t0, 8, 16
+    ins     t1, t1, 16, 16
+    ins     t1, t1, 8, 16
+    ulw     t2, 4(t6)
+    usw     t0, 0(t5)
+    usw     t1, 4(t5)
+    srl     t3, t2, 16
+    ins     t2, t2, 16, 16
+    ins     t2, t2, 8, 16
+    ins     t3, t3, 16, 16
+    ins     t3, t3, 8, 16
+    usw     t2, 8(t5)
+    usw     t3, 12(t5)
+    addiu   t5, 16
+    bne     t5, t8, 1b
+     addiu  t6, 8
+    beqz    t9, 3f
+     move   t4, t9
+2:
+    lbu     t0, 0(t6)
+    sb      t0, 0(t5)
+    sb      t0, 1(t5)
+    addiu   t4, -2
+    addiu   t6, 1
+    bgtz    t4, 2b
+     addiu  t5, 2
+3:
+    lw      t6, 0(t7)       // t6 = outptr[0]
+    lw      t5, 4(t7)       // t5 = outptr[1]
+    addu    t4, t6, a1      // t4 = new end address
+    beq     a1, t9, 5f
+     subu   t8, t4, t9
+4:
+    ulw     t0, 0(t6)
+    ulw     t1, 4(t6)
+    ulw     t2, 8(t6)
+    usw     t0, 0(t5)
+    ulw     t0, 12(t6)
+    usw     t1, 4(t5)
+    usw     t2, 8(t5)
+    usw     t0, 12(t5)
+    addiu   t6, 16
+    bne     t6, t8, 4b
+     addiu  t5, 16
+    beqz    t9, 6f
+     nop
+5:
+    lbu     t0, 0(t6)
+    sb      t0, 0(t5)
+    addiu   t6, 1
+    bne     t6, t4, 5b
+     addiu  t5, 1
+6:
+    addiu   t7, 8
+    addiu   a0, -2
+    bgtz    a0, 0b
+     addiu  a2, 4
+7:
+    j       ra
+     nop
+END(jsimd_h2v2_upsample_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_islow_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - compptr->dcttable
+ * a2     - output
+ * a3     - range_limit
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu     sp, sp, -256
+    move      v0, sp
+    addiu     v1, zero, 8      // v1 = DCTSIZE = 8
+1:
+    lh        s4, 32(a0)       // s4 = inptr[16]
+    lh        s5, 64(a0)       // s5 = inptr[32]
+    lh        s6, 96(a0)       // s6 = inptr[48]
+    lh        t1, 112(a0)      // t1 = inptr[56]
+    lh        t7, 16(a0)       // t7 = inptr[8]
+    lh        t5, 80(a0)       // t5 = inptr[40]
+    lh        t3, 48(a0)       // t3 = inptr[24]
+    or        s4, s4, t1
+    or        s4, s4, t3
+    or        s4, s4, t5
+    or        s4, s4, t7
+    or        s4, s4, s5
+    or        s4, s4, s6
+    bnez      s4, 2f
+     addiu    v1, v1, -1
+    lh        s5, 0(a1)        // quantptr[DCTSIZE*0]
+    lh        s6, 0(a0)        // inptr[DCTSIZE*0]
+    mul       s5, s5, s6       // DEQUANTIZE(inptr[0], quantptr[0])
+    sll       s5, s5, 2
+    sw        s5, 0(v0)
+    sw        s5, 32(v0)
+    sw        s5, 64(v0)
+    sw        s5, 96(v0)
+    sw        s5, 128(v0)
+    sw        s5, 160(v0)
+    sw        s5, 192(v0)
+    b         3f
+     sw       s5, 224(v0)
+2:
+    lh        t0, 112(a1)
+    lh        t2, 48(a1)
+    lh        t4, 80(a1)
+    lh        t6, 16(a1)
+    mul       t0, t0, t1       // DEQUANTIZE(inptr[DCTSIZE*7],quant[DCTSIZE*7])
+    mul       t1, t2, t3       // DEQUANTIZE(inptr[DCTSIZE*3],quant[DCTSIZE*3])
+    mul       t2, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*5],quant[DCTSIZE*5])
+    mul       t3, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*1],quant[DCTSIZE*1])
+    lh        t4, 32(a1)
+    lh        t5, 32(a0)
+    lh        t6, 96(a1)
+    lh        t7, 96(a0)
+    addu      s0, t0, t1       // z3 = tmp0 + tmp2
+    addu      s1, t1, t2       // z2 = tmp1 + tmp2
+    addu      s2, t2, t3       // z4 = tmp1 + tmp3
+    addu      s3, s0, s2       // z3 + z4
+    addiu     t9, zero, 9633   // FIX_1_175875602
+    mul       s3, s3, t9       // z5 = MULTIPLY(z3 + z4, FIX_1_175875602)
+    addu      t8, t0, t3       // z1 = tmp0 + tmp3
+    addiu     t9, zero, 2446   // FIX_0_298631336
+    mul       t0, t0, t9       // tmp0 = MULTIPLY(tmp0, FIX_0_298631336)
+    addiu     t9, zero, 16819  // FIX_2_053119869
+    mul       t2, t2, t9       // tmp1 = MULTIPLY(tmp1, FIX_2_053119869)
+    addiu     t9, zero, 25172  // FIX_3_072711026
+    mul       t1, t1, t9       // tmp2 = MULTIPLY(tmp2, FIX_3_072711026)
+    addiu     t9, zero, 12299  // FIX_1_501321110
+    mul       t3, t3, t9       // tmp3 = MULTIPLY(tmp3, FIX_1_501321110)
+    addiu     t9, zero, 16069  // FIX_1_961570560
+    mul       s0, s0, t9       // -z3 = MULTIPLY(z3, FIX_1_961570560)
+    addiu     t9, zero, 3196   // FIX_0_390180644
+    mul       s2, s2, t9       // -z4 = MULTIPLY(z4, FIX_0_390180644)
+    addiu     t9, zero, 7373   // FIX_0_899976223
+    mul       t8, t8, t9       // -z1 = MULTIPLY(z1, FIX_0_899976223)
+    addiu     t9, zero, 20995  // FIX_2_562915447
+    mul       s1, s1, t9       // -z2 = MULTIPLY(z2, FIX_2_562915447)
+    subu      s0, s3, s0       // z3 += z5
+    addu      t0, t0, s0       // tmp0 += z3
+    addu      t1, t1, s0       // tmp2 += z3
+    subu      s2, s3, s2       // z4 += z5
+    addu      t2, t2, s2       // tmp1 += z4
+    addu      t3, t3, s2       // tmp3 += z4
+    subu      t0, t0, t8       // tmp0 += z1
+    subu      t1, t1, s1       // tmp2 += z2
+    subu      t2, t2, s1       // tmp1 += z2
+    subu      t3, t3, t8       // tmp3 += z1
+    mul       s0, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*2],quant[DCTSIZE*2])
+    addiu     t9, zero, 6270   // FIX_0_765366865
+    mul       s1, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*6],quant[DCTSIZE*6])
+    lh        t4, 0(a1)
+    lh        t5, 0(a0)
+    lh        t6, 64(a1)
+    lh        t7, 64(a0)
+    mul       s2, t9, s0       // MULTIPLY(z2, FIX_0_765366865)
+    mul       t5, t4, t5       // DEQUANTIZE(inptr[DCTSIZE*0],quant[DCTSIZE*0])
+    mul       t6, t6, t7       // DEQUANTIZE(inptr[DCTSIZE*4],quant[DCTSIZE*4])
+    addiu     t9, zero, 4433   // FIX_0_541196100
+    addu      s3, s0, s1       // z2 + z3
+    mul       s3, s3, t9       // z1 = MULTIPLY(z2 + z3, FIX_0_541196100)
+    addiu     t9, zero, 15137  // FIX_1_847759065
+    mul       t8, s1, t9       // MULTIPLY(z3, FIX_1_847759065)
+    addu      t4, t5, t6
+    subu      t5, t5, t6
+    sll       t4, t4, 13       // tmp0 = (z2 + z3) << CONST_BITS
+    sll       t5, t5, 13       // tmp1 = (z2 - z3) << CONST_BITS
+    addu      t7, s3, s2       // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865)
+    subu      t6, s3, t8       // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065)
+    addu      s0, t4, t7
+    subu      s1, t4, t7
+    addu      s2, t5, t6
+    subu      s3, t5, t6
+    addu      t4, s0, t3
+    subu      s0, s0, t3
+    addu      t3, s2, t1
+    subu      s2, s2, t1
+    addu      t1, s3, t2
+    subu      s3, s3, t2
+    addu      t2, s1, t0
+    subu      s1, s1, t0
+    shra_r.w  t4, t4, 11
+    shra_r.w  t3, t3, 11
+    shra_r.w  t1, t1, 11
+    shra_r.w  t2, t2, 11
+    shra_r.w  s1, s1, 11
+    shra_r.w  s3, s3, 11
+    shra_r.w  s2, s2, 11
+    shra_r.w  s0, s0, 11
+    sw        t4, 0(v0)
+    sw        t3, 32(v0)
+    sw        t1, 64(v0)
+    sw        t2, 96(v0)
+    sw        s1, 128(v0)
+    sw        s3, 160(v0)
+    sw        s2, 192(v0)
+    sw        s0, 224(v0)
+3:
+    addiu     a1, a1, 2
+    addiu     a0, a0, 2
+    bgtz      v1, 1b
+     addiu    v0, v0, 4
+    move      v0, sp
+    addiu     v1, zero, 8
+4:
+    lw        t0, 8(v0)        // z2 = (JLONG) wsptr[2]
+    lw        t1, 24(v0)       // z3 = (JLONG) wsptr[6]
+    lw        t2, 0(v0)        // (JLONG) wsptr[0]
+    lw        t3, 16(v0)       // (JLONG) wsptr[4]
+    lw        s4, 4(v0)        // (JLONG) wsptr[1]
+    lw        s5, 12(v0)       // (JLONG) wsptr[3]
+    lw        s6, 20(v0)       // (JLONG) wsptr[5]
+    lw        s7, 28(v0)       // (JLONG) wsptr[7]
+    or        s4, s4, t0
+    or        s4, s4, t1
+    or        s4, s4, t3
+    or        s4, s4, s7
+    or        s4, s4, s5
+    or        s4, s4, s6
+    bnez      s4, 5f
+     addiu    v1, v1, -1
+    shra_r.w  s5, t2, 5
+    andi      s5, s5, 0x3ff
+    lbux      s5, s5(a3)
+    lw        s1, 0(a2)
+    replv.qb  s5, s5
+    usw       s5, 0(s1)
+    usw       s5, 4(s1)
+    b         6f
+     nop
+5:
+    addu      t4, t0, t1       // z2 + z3
+    addiu     t8, zero, 4433   // FIX_0_541196100
+    mul       t5, t4, t8       // z1 = MULTIPLY(z2 + z3, FIX_0_541196100)
+    addiu     t8, zero, 15137  // FIX_1_847759065
+    mul       t1, t1, t8       // MULTIPLY(z3, FIX_1_847759065)
+    addiu     t8, zero, 6270   // FIX_0_765366865
+    mul       t0, t0, t8       // MULTIPLY(z2, FIX_0_765366865)
+    addu      t4, t2, t3       // (JLONG) wsptr[0] + (JLONG) wsptr[4]
+    subu      t2, t2, t3       // (JLONG) wsptr[0] - (JLONG) wsptr[4]
+    sll       t4, t4, 13       // tmp0 = ((wsptr[0] + wsptr[4]) << CONST_BITS
+    sll       t2, t2, 13       // tmp1 = ((wsptr[0] - wsptr[4]) << CONST_BITS
+    subu      t1, t5, t1       // tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065)
+    subu      t3, t2, t1       // tmp12 = tmp1 - tmp2
+    addu      t2, t2, t1       // tmp11 = tmp1 + tmp2
+    addu      t5, t5, t0       // tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865)
+    subu      t1, t4, t5       // tmp13 = tmp0 - tmp3
+    addu      t0, t4, t5       // tmp10 = tmp0 + tmp3
+    lw        t4, 28(v0)       // tmp0 = (JLONG) wsptr[7]
+    lw        t6, 12(v0)       // tmp2 = (JLONG) wsptr[3]
+    lw        t5, 20(v0)       // tmp1 = (JLONG) wsptr[5]
+    lw        t7, 4(v0)        // tmp3 = (JLONG) wsptr[1]
+    addu      s0, t4, t6       // z3 = tmp0 + tmp2
+    addiu     t8, zero, 9633   // FIX_1_175875602
+    addu      s1, t5, t7       // z4 = tmp1 + tmp3
+    addu      s2, s0, s1       // z3 + z4
+    mul       s2, s2, t8       // z5 = MULTIPLY(z3 + z4, FIX_1_175875602)
+    addu      s3, t4, t7       // z1 = tmp0 + tmp3
+    addu      t9, t5, t6       // z2 = tmp1 + tmp2
+    addiu     t8, zero, 16069  // FIX_1_961570560
+    mul       s0, s0, t8       // -z3 = MULTIPLY(z3, FIX_1_961570560)
+    addiu     t8, zero, 3196   // FIX_0_390180644
+    mul       s1, s1, t8       // -z4 = MULTIPLY(z4, FIX_0_390180644)
+    addiu     t8, zero, 2446   // FIX_0_298631336
+    mul       t4, t4, t8       // tmp0 = MULTIPLY(tmp0, FIX_0_298631336)
+    addiu     t8, zero, 7373   // FIX_0_899976223
+    mul       s3, s3, t8       // -z1 = MULTIPLY(z1, FIX_0_899976223)
+    addiu     t8, zero, 16819  // FIX_2_053119869
+    mul       t5, t5, t8       // tmp1 = MULTIPLY(tmp1, FIX_2_053119869)
+    addiu     t8, zero, 20995  // FIX_2_562915447
+    mul       t9, t9, t8       // -z2 = MULTIPLY(z2, FIX_2_562915447)
+    addiu     t8, zero, 25172  // FIX_3_072711026
+    mul       t6, t6, t8       // tmp2 = MULTIPLY(tmp2, FIX_3_072711026)
+    addiu     t8, zero, 12299  // FIX_1_501321110
+    mul       t7, t7, t8       // tmp3 = MULTIPLY(tmp3, FIX_1_501321110)
+    subu      s0, s2, s0       // z3 += z5
+    subu      s1, s2, s1       // z4 += z5
+    addu      t4, t4, s0
+    subu      t4, t4, s3       // tmp0
+    addu      t5, t5, s1
+    subu      t5, t5, t9       // tmp1
+    addu      t6, t6, s0
+    subu      t6, t6, t9       // tmp2
+    addu      t7, t7, s1
+    subu      t7, t7, s3       // tmp3
+    addu      s0, t0, t7
+    subu      t0, t0, t7
+    addu      t7, t2, t6
+    subu      t2, t2, t6
+    addu      t6, t3, t5
+    subu      t3, t3, t5
+    addu      t5, t1, t4
+    subu      t1, t1, t4
+    shra_r.w  s0, s0, 18
+    shra_r.w  t7, t7, 18
+    shra_r.w  t6, t6, 18
+    shra_r.w  t5, t5, 18
+    shra_r.w  t1, t1, 18
+    shra_r.w  t3, t3, 18
+    shra_r.w  t2, t2, 18
+    shra_r.w  t0, t0, 18
+    andi      s0, s0, 0x3ff
+    andi      t7, t7, 0x3ff
+    andi      t6, t6, 0x3ff
+    andi      t5, t5, 0x3ff
+    andi      t1, t1, 0x3ff
+    andi      t3, t3, 0x3ff
+    andi      t2, t2, 0x3ff
+    andi      t0, t0, 0x3ff
+    lw        s1, 0(a2)
+    lbux      s0, s0(a3)
+    lbux      t7, t7(a3)
+    lbux      t6, t6(a3)
+    lbux      t5, t5(a3)
+    lbux      t1, t1(a3)
+    lbux      t3, t3(a3)
+    lbux      t2, t2(a3)
+    lbux      t0, t0(a3)
+    sb        s0, 0(s1)
+    sb        t7, 1(s1)
+    sb        t6, 2(s1)
+    sb        t5, 3(s1)
+    sb        t1, 4(s1)
+    sb        t3, 5(s1)
+    sb        t2, 6(s1)
+    sb        t0, 7(s1)
+6:
+    addiu     v0, v0, 32
+    bgtz      v1, 4b
+     addiu    a2, a2, 4
+    addiu     sp, sp, 256
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+
+END(jsimd_idct_islow_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_ifast_cols_mips_dspr2)
+/*
+ * a0     - inptr
+ * a1     - quantptr
+ * a2     - wsptr
+ * a3     - mips_idct_ifast_coefs
+ */
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu          t9, a0, 16            // end address
+    or             AT, a3, zero
+
+0:
+    lw             s0, 0(a1)             // quantptr[DCTSIZE*0]
+    lw             t0, 0(a0)             // inptr[DCTSIZE*0]
+    lw             t1, 16(a0)            // inptr[DCTSIZE*1]
+    muleq_s.w.phl  v0, t0, s0            // tmp0 ...
+    lw             t2, 32(a0)            // inptr[DCTSIZE*2]
+    lw             t3, 48(a0)            // inptr[DCTSIZE*3]
+    lw             t4, 64(a0)            // inptr[DCTSIZE*4]
+    lw             t5, 80(a0)            // inptr[DCTSIZE*5]
+    muleq_s.w.phr  t0, t0, s0            // ... tmp0 ...
+    lw             t6, 96(a0)            // inptr[DCTSIZE*6]
+    lw             t7, 112(a0)           // inptr[DCTSIZE*7]
+    or             s4, t1, t2
+    or             s5, t3, t4
+    bnez           s4, 1f
+     ins           t0, v0, 16, 16        // ... tmp0
+    bnez           s5, 1f
+     or            s6, t5, t6
+    or             s6, s6, t7
+    bnez           s6, 1f
+     sw            t0, 0(a2)             // wsptr[DCTSIZE*0]
+    sw             t0, 16(a2)            // wsptr[DCTSIZE*1]
+    sw             t0, 32(a2)            // wsptr[DCTSIZE*2]
+    sw             t0, 48(a2)            // wsptr[DCTSIZE*3]
+    sw             t0, 64(a2)            // wsptr[DCTSIZE*4]
+    sw             t0, 80(a2)            // wsptr[DCTSIZE*5]
+    sw             t0, 96(a2)            // wsptr[DCTSIZE*6]
+    sw             t0, 112(a2)           // wsptr[DCTSIZE*7]
+    addiu          a0, a0, 4
+    b              2f
+     addiu         a1, a1, 4
+
+1:
+    lw             s1, 32(a1)            // quantptr[DCTSIZE*2]
+    lw             s2, 64(a1)            // quantptr[DCTSIZE*4]
+    muleq_s.w.phl  v0, t2, s1            // tmp1 ...
+    muleq_s.w.phr  t2, t2, s1            // ... tmp1 ...
+    lw             s0, 16(a1)            // quantptr[DCTSIZE*1]
+    lw             s1, 48(a1)            // quantptr[DCTSIZE*3]
+    lw             s3, 96(a1)            // quantptr[DCTSIZE*6]
+    muleq_s.w.phl  v1, t4, s2            // tmp2 ...
+    muleq_s.w.phr  t4, t4, s2            // ... tmp2 ...
+    lw             s2, 80(a1)            // quantptr[DCTSIZE*5]
+    lw             t8, 4(AT)             // FIX(1.414213562)
+    ins            t2, v0, 16, 16        // ... tmp1
+    muleq_s.w.phl  v0, t6, s3            // tmp3 ...
+    muleq_s.w.phr  t6, t6, s3            // ... tmp3 ...
+    ins            t4, v1, 16, 16        // ... tmp2
+    addq.ph        s4, t0, t4            // tmp10
+    subq.ph        s5, t0, t4            // tmp11
+    ins            t6, v0, 16, 16        // ... tmp3
+    subq.ph        s6, t2, t6            // tmp12 ...
+    addq.ph        s7, t2, t6            // tmp13
+    mulq_s.ph      s6, s6, t8            // ... tmp12 ...
+    addq.ph        t0, s4, s7            // tmp0
+    subq.ph        t6, s4, s7            // tmp3
+    muleq_s.w.phl  v0, t1, s0            // tmp4 ...
+    muleq_s.w.phr  t1, t1, s0            // ... tmp4 ...
+    shll_s.ph      s6, s6, 1             // x2
+    lw             s3, 112(a1)           // quantptr[DCTSIZE*7]
+    subq.ph        s6, s6, s7            // ... tmp12
+    muleq_s.w.phl  v1, t7, s3            // tmp7 ...
+    muleq_s.w.phr  t7, t7, s3            // ... tmp7 ...
+    ins            t1, v0, 16, 16        // ... tmp4
+    addq.ph        t2, s5, s6            // tmp1
+    subq.ph        t4, s5, s6            // tmp2
+    muleq_s.w.phl  v0, t5, s2            // tmp6 ...
+    muleq_s.w.phr  t5, t5, s2            // ... tmp6 ...
+    ins            t7, v1, 16, 16        // ... tmp7
+    addq.ph        s5, t1, t7            // z11
+    subq.ph        s6, t1, t7            // z12
+    muleq_s.w.phl  v1, t3, s1            // tmp5 ...
+    muleq_s.w.phr  t3, t3, s1            // ... tmp5 ...
+    ins            t5, v0, 16, 16        // ... tmp6
+    ins            t3, v1, 16, 16        // ... tmp5
+    addq.ph        s7, t5, t3            // z13
+    subq.ph        v0, t5, t3            // z10
+    addq.ph        t7, s5, s7            // tmp7
+    subq.ph        s5, s5, s7            // tmp11 ...
+    addq.ph        v1, v0, s6            // z5 ...
+    mulq_s.ph      s5, s5, t8            // ... tmp11
+    lw             t8, 8(AT)             // FIX(1.847759065)
+    lw             s4, 0(AT)             // FIX(1.082392200)
+    addq.ph        s0, t0, t7
+    subq.ph        s1, t0, t7
+    mulq_s.ph      v1, v1, t8            // ... z5
+    shll_s.ph      s5, s5, 1             // x2
+    lw             t8, 12(AT)            // FIX(-2.613125930)
+    sw             s0, 0(a2)             // wsptr[DCTSIZE*0]
+    shll_s.ph      v0, v0, 1             // x4
+    mulq_s.ph      v0, v0, t8            // tmp12 ...
+    mulq_s.ph      s4, s6, s4            // tmp10 ...
+    shll_s.ph      v1, v1, 1             // x2
+    addiu          a0, a0, 4
+    addiu          a1, a1, 4
+    sw             s1, 112(a2)           // wsptr[DCTSIZE*7]
+    shll_s.ph      s6, v0, 1             // x4
+    shll_s.ph      s4, s4, 1             // x2
+    addq.ph        s6, s6, v1            // ... tmp12
+    subq.ph        t5, s6, t7            // tmp6
+    subq.ph        s4, s4, v1            // ... tmp10
+    subq.ph        t3, s5, t5            // tmp5
+    addq.ph        s2, t2, t5
+    addq.ph        t1, s4, t3            // tmp4
+    subq.ph        s3, t2, t5
+    sw             s2, 16(a2)            // wsptr[DCTSIZE*1]
+    sw             s3, 96(a2)            // wsptr[DCTSIZE*6]
+    addq.ph        v0, t4, t3
+    subq.ph        v1, t4, t3
+    sw             v0, 32(a2)            // wsptr[DCTSIZE*2]
+    sw             v1, 80(a2)            // wsptr[DCTSIZE*5]
+    addq.ph        v0, t6, t1
+    subq.ph        v1, t6, t1
+    sw             v0, 64(a2)            // wsptr[DCTSIZE*4]
+    sw             v1, 48(a2)            // wsptr[DCTSIZE*3]
+
+2:
+    bne            a0, t9, 0b
+     addiu         a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j              ra
+     nop
+
+END(jsimd_idct_ifast_cols_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_ifast_rows_mips_dspr2)
+/*
+ * a0     - wsptr
+ * a1     - output_buf
+ * a2     - output_col
+ * a3     - mips_idct_ifast_coefs
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    addiu         t9, a0, 128        // end address
+    lui           s8, 0x8080
+    ori           s8, s8, 0x8080
+
+0:
+    lw            AT, 36(sp)         // restore $a3 (mips_idct_ifast_coefs)
+    lw            t0, 0(a0)          // wsptr[DCTSIZE*0+0/1]  b a
+    lw            s0, 16(a0)         // wsptr[DCTSIZE*1+0/1]  B A
+    lw            t2, 4(a0)          // wsptr[DCTSIZE*0+2/3]  d c
+    lw            s2, 20(a0)         // wsptr[DCTSIZE*1+2/3]  D C
+    lw            t4, 8(a0)          // wsptr[DCTSIZE*0+4/5]  f e
+    lw            s4, 24(a0)         // wsptr[DCTSIZE*1+4/5]  F E
+    lw            t6, 12(a0)         // wsptr[DCTSIZE*0+6/7]  h g
+    lw            s6, 28(a0)         // wsptr[DCTSIZE*1+6/7]  H G
+    precrq.ph.w   t1, s0, t0         // B b
+    ins           t0, s0, 16, 16     // A a
+    bnez          t1, 1f
+     or           s0, t2, s2
+    bnez          s0, 1f
+     or           s0, t4, s4
+    bnez          s0, 1f
+     or           s0, t6, s6
+    bnez          s0, 1f
+     shll_s.ph    s0, t0, 2          // A a
+    lw            a3, 0(a1)
+    lw            AT, 4(a1)
+    precrq.ph.w   t0, s0, s0         // A A
+    ins           s0, s0, 16, 16     // a a
+    addu          a3, a3, a2
+    addu          AT, AT, a2
+    precrq.qb.ph  t0, t0, t0         // A A A A
+    precrq.qb.ph  s0, s0, s0         // a a a a
+    addu.qb       s0, s0, s8
+    addu.qb       t0, t0, s8
+    sw            s0, 0(a3)
+    sw            s0, 4(a3)
+    sw            t0, 0(AT)
+    sw            t0, 4(AT)
+    addiu         a0, a0, 32
+    bne           a0, t9, 0b
+     addiu        a1, a1, 8
+    b             2f
+     nop
+
+1:
+    precrq.ph.w   t3, s2, t2
+    ins           t2, s2, 16, 16
+    precrq.ph.w   t5, s4, t4
+    ins           t4, s4, 16, 16
+    precrq.ph.w   t7, s6, t6
+    ins           t6, s6, 16, 16
+    lw            t8, 4(AT)          // FIX(1.414213562)
+    addq.ph       s4, t0, t4         // tmp10
+    subq.ph       s5, t0, t4         // tmp11
+    subq.ph       s6, t2, t6         // tmp12 ...
+    addq.ph       s7, t2, t6         // tmp13
+    mulq_s.ph     s6, s6, t8         // ... tmp12 ...
+    addq.ph       t0, s4, s7         // tmp0
+    subq.ph       t6, s4, s7         // tmp3
+    shll_s.ph     s6, s6, 1          // x2
+    subq.ph       s6, s6, s7         // ... tmp12
+    addq.ph       t2, s5, s6         // tmp1
+    subq.ph       t4, s5, s6         // tmp2
+    addq.ph       s5, t1, t7         // z11
+    subq.ph       s6, t1, t7         // z12
+    addq.ph       s7, t5, t3         // z13
+    subq.ph       v0, t5, t3         // z10
+    addq.ph       t7, s5, s7         // tmp7
+    subq.ph       s5, s5, s7         // tmp11 ...
+    addq.ph       v1, v0, s6         // z5 ...
+    mulq_s.ph     s5, s5, t8         // ... tmp11
+    lw            t8, 8(AT)          // FIX(1.847759065)
+    lw            s4, 0(AT)          // FIX(1.082392200)
+    addq.ph       s0, t0, t7         // tmp0 + tmp7
+    subq.ph       s7, t0, t7         // tmp0 - tmp7
+    mulq_s.ph     v1, v1, t8         // ... z5
+    lw            a3, 0(a1)
+    lw            t8, 12(AT)         // FIX(-2.613125930)
+    shll_s.ph     s5, s5, 1          // x2
+    addu          a3, a3, a2
+    shll_s.ph     v0, v0, 1          // x4
+    mulq_s.ph     v0, v0, t8         // tmp12 ...
+    mulq_s.ph     s4, s6, s4         // tmp10 ...
+    shll_s.ph     v1, v1, 1          // x2
+    addiu         a0, a0, 32
+    addiu         a1, a1, 8
+    shll_s.ph     s6, v0, 1          // x4
+    shll_s.ph     s4, s4, 1          // x2
+    addq.ph       s6, s6, v1         // ... tmp12
+    shll_s.ph     s0, s0, 2
+    subq.ph       t5, s6, t7         // tmp6
+    subq.ph       s4, s4, v1         // ... tmp10
+    subq.ph       t3, s5, t5         // tmp5
+    shll_s.ph     s7, s7, 2
+    addq.ph       t1, s4, t3         // tmp4
+    addq.ph       s1, t2, t5         // tmp1 + tmp6
+    subq.ph       s6, t2, t5         // tmp1 - tmp6
+    addq.ph       s2, t4, t3         // tmp2 + tmp5
+    subq.ph       s5, t4, t3         // tmp2 - tmp5
+    addq.ph       s4, t6, t1         // tmp3 + tmp4
+    subq.ph       s3, t6, t1         // tmp3 - tmp4
+    shll_s.ph     s1, s1, 2
+    shll_s.ph     s2, s2, 2
+    shll_s.ph     s3, s3, 2
+    shll_s.ph     s4, s4, 2
+    shll_s.ph     s5, s5, 2
+    shll_s.ph     s6, s6, 2
+    precrq.ph.w   t0, s1, s0         // B A
+    ins           s0, s1, 16, 16     // b a
+    precrq.ph.w   t2, s3, s2         // D C
+    ins           s2, s3, 16, 16     // d c
+    precrq.ph.w   t4, s5, s4         // F E
+    ins           s4, s5, 16, 16     // f e
+    precrq.ph.w   t6, s7, s6         // H G
+    ins           s6, s7, 16, 16     // h g
+    precrq.qb.ph  t0, t2, t0         // D C B A
+    precrq.qb.ph  s0, s2, s0         // d c b a
+    precrq.qb.ph  t4, t6, t4         // H G F E
+    precrq.qb.ph  s4, s6, s4         // h g f e
+    addu.qb       s0, s0, s8
+    addu.qb       s4, s4, s8
+    sw            s0, 0(a3)          // outptr[0/1/2/3]       d c b a
+    sw            s4, 4(a3)          // outptr[4/5/6/7]       h g f e
+    lw            a3, -4(a1)
+    addu.qb       t0, t0, s8
+    addu          a3, a3, a2
+    addu.qb       t4, t4, s8
+    sw            t0, 0(a3)          // outptr[0/1/2/3]       D C B A
+    bne           a0, t9, 0b
+     sw           t4, 4(a3)          // outptr[4/5/6/7]       H G F E
+
+2:
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8, a3
+
+    j             ra
+     nop
+
+END(jsimd_idct_ifast_rows_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_fdct_islow_mips_dspr2)
+/*
+ * a0     - data
+ */
+
+    SAVE_REGS_ON_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    lui       t0, 6437
+    ori       t0, 2260
+    lui       t1, 9633
+    ori       t1, 11363
+    lui       t2, 0xd39e
+    ori       t2, 0xe6dc
+    lui       t3, 0xf72d
+    ori       t3, 9633
+    lui       t4, 2261
+    ori       t4, 9633
+    lui       t5, 0xd39e
+    ori       t5, 6437
+    lui       t6, 9633
+    ori       t6, 0xd39d
+    lui       t7, 0xe6dc
+    ori       t7, 2260
+    lui       t8, 4433
+    ori       t8, 10703
+    lui       t9, 0xd630
+    ori       t9, 4433
+    li        s8, 8
+    move      a1, a0
+1:
+    lw        s0, 0(a1)     // tmp0 = 1|0
+    lw        s1, 4(a1)     // tmp1 = 3|2
+    lw        s2, 8(a1)     // tmp2 = 5|4
+    lw        s3, 12(a1)    // tmp3 = 7|6
+    packrl.ph s1, s1, s1    // tmp1 = 2|3
+    packrl.ph s3, s3, s3    // tmp3 = 6|7
+    subq.ph   s7, s1, s2    // tmp7 = 2-5|3-4 = t5|t4
+    subq.ph   s5, s0, s3    // tmp5 = 1-6|0-7 = t6|t7
+    mult      $0, $0        // ac0  = 0
+    dpa.w.ph  $ac0, s7, t0  // ac0 += t5*  6437 + t4*  2260
+    dpa.w.ph  $ac0, s5, t1  // ac0 += t6*  9633 + t7* 11363
+    mult      $ac1, $0, $0  // ac1  = 0
+    dpa.w.ph  $ac1, s7, t2  // ac1 += t5*-11362 + t4* -6436
+    dpa.w.ph  $ac1, s5, t3  // ac1 += t6* -2259 + t7*  9633
+    mult      $ac2, $0, $0  // ac2  = 0
+    dpa.w.ph  $ac2, s7, t4  // ac2 += t5*  2261 + t4*  9633
+    dpa.w.ph  $ac2, s5, t5  // ac2 += t6*-11362 + t7*  6437
+    mult      $ac3, $0, $0  // ac3  = 0
+    dpa.w.ph  $ac3, s7, t6  // ac3 += t5*  9633 + t4*-11363
+    dpa.w.ph  $ac3, s5, t7  // ac3 += t6* -6436 + t7*  2260
+    addq.ph   s6, s1, s2    // tmp6 = 2+5|3+4 = t2|t3
+    addq.ph   s4, s0, s3    // tmp4 = 1+6|0+7 = t1|t0
+    extr_r.w  s0, $ac0, 11  // tmp0 = (ac0 + 1024) >> 11
+    extr_r.w  s1, $ac1, 11  // tmp1 = (ac1 + 1024) >> 11
+    extr_r.w  s2, $ac2, 11  // tmp2 = (ac2 + 1024) >> 11
+    extr_r.w  s3, $ac3, 11  // tmp3 = (ac3 + 1024) >> 11
+    addq.ph   s5, s4, s6    // tmp5 = t1+t2|t0+t3 = t11|t10
+    subq.ph   s7, s4, s6    // tmp7 = t1-t2|t0-t3 = t12|t13
+    sh        s0, 2(a1)
+    sh        s1, 6(a1)
+    sh        s2, 10(a1)
+    sh        s3, 14(a1)
+    mult      $0, $0        // ac0  = 0
+    dpa.w.ph  $ac0, s7, t8  // ac0 += t12*  4433 + t13* 10703
+    mult      $ac1, $0, $0  // ac1  = 0
+    dpa.w.ph  $ac1, s7, t9  // ac1 += t12*-10704 + t13*  4433
+    sra       s4, s5, 16    // tmp4 = t11
+    addiu     a1, a1, 16
+    addiu     s8, s8, -1
+    extr_r.w  s0, $ac0, 11  // tmp0 = (ac0 + 1024) >> 11
+    extr_r.w  s1, $ac1, 11  // tmp1 = (ac1 + 1024) >> 11
+    addu      s2, s5, s4    // tmp2 = t10 + t11
+    subu      s3, s5, s4    // tmp3 = t10 - t11
+    sll       s2, s2, 2     // tmp2 = (t10 + t11) << 2
+    sll       s3, s3, 2     // tmp3 = (t10 - t11) << 2
+    sh        s2, -16(a1)
+    sh        s3, -8(a1)
+    sh        s0, -12(a1)
+    bgtz      s8, 1b
+     sh       s1, -4(a1)
+    li        t0, 2260
+    li        t1, 11363
+    li        t2, 9633
+    li        t3, 6436
+    li        t4, 6437
+    li        t5, 2261
+    li        t6, 11362
+    li        t7, 2259
+    li        t8, 4433
+    li        t9, 10703
+    li        a1, 10704
+    li        s8, 8
+
+2:
+    lh        a2, 0(a0)     // 0
+    lh        a3, 16(a0)    // 8
+    lh        v0, 32(a0)    // 16
+    lh        v1, 48(a0)    // 24
+    lh        s4, 64(a0)    // 32
+    lh        s5, 80(a0)    // 40
+    lh        s6, 96(a0)    // 48
+    lh        s7, 112(a0)   // 56
+    addu      s2, v0, s5    // tmp2 = 16 + 40
+    subu      s5, v0, s5    // tmp5 = 16 - 40
+    addu      s3, v1, s4    // tmp3 = 24 + 32
+    subu      s4, v1, s4    // tmp4 = 24 - 32
+    addu      s0, a2, s7    // tmp0 =  0 + 56
+    subu      s7, a2, s7    // tmp7 =  0 - 56
+    addu      s1, a3, s6    // tmp1 =  8 + 48
+    subu      s6, a3, s6    // tmp6 =  8 - 48
+    addu      a2, s0, s3    // tmp10 = tmp0 + tmp3
+    subu      v1, s0, s3    // tmp13 = tmp0 - tmp3
+    addu      a3, s1, s2    // tmp11 = tmp1 + tmp2
+    subu      v0, s1, s2    // tmp12 = tmp1 - tmp2
+    mult      s7, t1        // ac0  = tmp7 * c1
+    madd      s4, t0        // ac0 += tmp4 * c0
+    madd      s5, t4        // ac0 += tmp5 * c4
+    madd      s6, t2        // ac0 += tmp6 * c2
+    mult      $ac1, s7, t2  // ac1  = tmp7 * c2
+    msub      $ac1, s4, t3  // ac1 -= tmp4 * c3
+    msub      $ac1, s5, t6  // ac1 -= tmp5 * c6
+    msub      $ac1, s6, t7  // ac1 -= tmp6 * c7
+    mult      $ac2, s7, t4  // ac2  = tmp7 * c4
+    madd      $ac2, s4, t2  // ac2 += tmp4 * c2
+    madd      $ac2, s5, t5  // ac2 += tmp5 * c5
+    msub      $ac2, s6, t6  // ac2 -= tmp6 * c6
+    mult      $ac3, s7, t0  // ac3  = tmp7 * c0
+    msub      $ac3, s4, t1  // ac3 -= tmp4 * c1
+    madd      $ac3, s5, t2  // ac3 += tmp5 * c2
+    msub      $ac3, s6, t3  // ac3 -= tmp6 * c3
+    extr_r.w  s0, $ac0, 15  // tmp0 = (ac0 + 16384) >> 15
+    extr_r.w  s1, $ac1, 15  // tmp1 = (ac1 + 16384) >> 15
+    extr_r.w  s2, $ac2, 15  // tmp2 = (ac2 + 16384) >> 15
+    extr_r.w  s3, $ac3, 15  // tmp3 = (ac3 + 16384) >> 15
+    addiu     s8, s8, -1
+    addu      s4, a2, a3    // tmp4 = tmp10 + tmp11
+    subu      s5, a2, a3    // tmp5 = tmp10 - tmp11
+    sh        s0, 16(a0)
+    sh        s1, 48(a0)
+    sh        s2, 80(a0)
+    sh        s3, 112(a0)
+    mult      v0, t8        // ac0  = tmp12 * c8
+    madd      v1, t9        // ac0 += tmp13 * c9
+    mult      $ac1, v1, t8  // ac1  = tmp13 * c8
+    msub      $ac1, v0, a1  // ac1 -= tmp12 * c10
+    addiu     a0, a0, 2
+    extr_r.w  s6, $ac0, 15  // tmp6 = (ac0 + 16384) >> 15
+    extr_r.w  s7, $ac1, 15  // tmp7 = (ac1 + 16384) >> 15
+    shra_r.w  s4, s4, 2     // tmp4 = (tmp4 + 2) >> 2
+    shra_r.w  s5, s5, 2     // tmp5 = (tmp5 + 2) >> 2
+    sh        s4, -2(a0)
+    sh        s5, 62(a0)
+    sh        s6, 30(a0)
+    bgtz      s8, 2b
+     sh       s7, 94(a0)
+
+    RESTORE_REGS_FROM_STACK 40, s0, s1, s2, s3, s4, s5, s6, s7, s8
+
+    jr       ra
+     nop
+
+END(jsimd_fdct_islow_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_fdct_ifast_mips_dspr2)
+/*
+ * a0     - data
+ */
+    .set at
+    SAVE_REGS_ON_STACK 8, s0, s1
+    li           a1, 0x014e014e  // FIX_1_306562965 (334 << 16)|(334 & 0xffff)
+    li           a2, 0x008b008b  // FIX_0_541196100 (139 << 16)|(139 & 0xffff)
+    li           a3, 0x00620062  // FIX_0_382683433 (98 << 16) |(98 & 0xffff)
+    li           s1, 0x00b500b5  // FIX_0_707106781 (181 << 16)|(181 & 0xffff)
+
+    move         v0, a0
+    addiu        v1, v0, 128     // end address
+
+0:
+    lw           t0, 0(v0)       // tmp0 = 1|0
+    lw           t1, 4(v0)       // tmp1 = 3|2
+    lw           t2, 8(v0)       // tmp2 = 5|4
+    lw           t3, 12(v0)      // tmp3 = 7|6
+    packrl.ph    t1, t1, t1      // tmp1 = 2|3
+    packrl.ph    t3, t3, t3      // tmp3 = 6|7
+    subq.ph      t7, t1, t2      // tmp7 = 2-5|3-4 = t5|t4
+    subq.ph      t5, t0, t3      // tmp5 = 1-6|0-7 = t6|t7
+    addq.ph      t6, t1, t2      // tmp6 = 2+5|3+4 = t2|t3
+    addq.ph      t4, t0, t3      // tmp4 = 1+6|0+7 = t1|t0
+    addq.ph      t8, t4, t6      // tmp5 = t1+t2|t0+t3 = t11|t10
+    subq.ph      t9, t4, t6      // tmp7 = t1-t2|t0-t3 = t12|t13
+    sra          t4, t8, 16      // tmp4 = t11
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t9, s1
+    mult         $ac1, $0, $0    // ac1  = 0
+    dpa.w.ph     $ac1, t7, a3    // ac1 += t4*98 + t5*98
+    dpsx.w.ph    $ac1, t5, a3    // ac1 += t6*98 + t7*98
+    mult         $ac2, $0, $0    // ac2  = 0
+    dpa.w.ph     $ac2, t7, a2    // ac2 += t4*139 + t5*139
+    mult         $ac3, $0, $0    // ac3  = 0
+    dpa.w.ph     $ac3, t5, a1    // ac3 += t6*334 + t7*334
+    precrq.ph.w  t0, t5, t7      // t0 = t5|t6
+    addq.ph      t2, t8, t4      // tmp2 = t10 + t11
+    subq.ph      t3, t8, t4      // tmp3 = t10 - t11
+    extr.w       t4, $ac0, 8
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t0, s1    // ac0 += t5*181 + t6*181
+    extr.w       t0, $ac1, 8     // t0 = z5
+    extr.w       t1, $ac2, 8     // t1 = MULTIPLY(tmp10, 139)
+    extr.w       t7, $ac3, 8     // t2 = MULTIPLY(tmp12, 334)
+    extr.w       t8, $ac0, 8     // t8 = z3 = MULTIPLY(tmp11, 181)
+    add          t6, t1, t0      // t6 = z2
+    add          t7, t7, t0      // t7 = z4
+    subq.ph      t0, t5, t8      // t0 = z13 = tmp7 - z3
+    addq.ph      t8, t5, t8      // t9 = z11 = tmp7 + z3
+    addq.ph      t1, t0, t6      // t1 = z13 + z2
+    subq.ph      t6, t0, t6      // t6 = z13 - z2
+    addq.ph      t0, t8, t7      // t0 = z11 + z4
+    subq.ph      t7, t8, t7      // t7 = z11 - z4
+    addq.ph      t5, t4, t9
+    subq.ph      t4, t9, t4
+    sh           t2, 0(v0)
+    sh           t5, 4(v0)
+    sh           t3, 8(v0)
+    sh           t4, 12(v0)
+    sh           t1, 10(v0)
+    sh           t6, 6(v0)
+    sh           t0, 2(v0)
+    sh           t7, 14(v0)
+    addiu        v0, 16
+    bne          v1, v0, 0b
+     nop
+    move         v0, a0
+    addiu        v1, v0, 16
+
+1:
+    lh           t0, 0(v0)       // 0
+    lh           t1, 16(v0)      // 8
+    lh           t2, 32(v0)      // 16
+    lh           t3, 48(v0)      // 24
+    lh           t4, 64(v0)      // 32
+    lh           t5, 80(v0)      // 40
+    lh           t6, 96(v0)      // 48
+    lh           t7, 112(v0)     // 56
+    add          t8, t0, t7      // t8 = tmp0
+    sub          t7, t0, t7      // t7 = tmp7
+    add          t0, t1, t6      // t0 = tmp1
+    sub          t1, t1, t6      // t1 = tmp6
+    add          t6, t2, t5      // t6 = tmp2
+    sub          t5, t2, t5      // t5 = tmp5
+    add          t2, t3, t4      // t2 = tmp3
+    sub          t3, t3, t4      // t3 = tmp4
+    add          t4, t8, t2      // t4 = tmp10 = tmp0 + tmp3
+    sub          t8, t8, t2      // t8 = tmp13 = tmp0 - tmp3
+    sub          s0, t0, t6      // s0 = tmp12 = tmp1 - tmp2
+    ins          t8, s0, 16, 16  // t8 = tmp12|tmp13
+    add          t2, t0, t6      // t2 = tmp11 = tmp1 + tmp2
+    mult         $0, $0          // ac0  = 0
+    dpa.w.ph     $ac0, t8, s1    // ac0 += t12*181 + t13*181
+    add          s0, t4, t2      // t8 = tmp10+tmp11
+    sub          t4, t4, t2      // t4 = tmp10-tmp11
+    sh           s0, 0(v0)
+    sh           t4, 64(v0)
+    extr.w       t2, $ac0, 8     // z1 = MULTIPLY(tmp12+tmp13,FIX_0_707106781)
+    addq.ph      t4, t8, t2      // t9 = tmp13 + z1
+    subq.ph      t8, t8, t2      // t2 = tmp13 - z1
+    sh           t4, 32(v0)
+    sh           t8, 96(v0)
+    add          t3, t3, t5      // t3 = tmp10 = tmp4 + tmp5
+    add          t0, t5, t1      // t0 = tmp11 = tmp5 + tmp6
+    add          t1, t1, t7      // t1 = tmp12 = tmp6 + tmp7
+    andi         t4, a1, 0xffff
+    mul          s0, t1, t4
+    sra          s0, s0, 8       // s0 = z4 = MULTIPLY(tmp12, FIX_1_306562965)
+    ins          t1, t3, 16, 16  // t1 = tmp10|tmp12
+    mult         $0, $0          // ac0  = 0
+    mulsa.w.ph   $ac0, t1, a3    // ac0 += t10*98 - t12*98
+    extr.w       t8, $ac0, 8     // z5 = MULTIPLY(tmp10-tmp12,FIX_0_382683433)
+    add          t2, t7, t8      // t2 = tmp7 + z5
+    sub          t7, t7, t8      // t7 = tmp7 - z5
+    andi         t4, a2, 0xffff
+    mul          t8, t3, t4
+    sra          t8, t8, 8       // t8 = z2 = MULTIPLY(tmp10, FIX_0_541196100)
+    andi         t4, s1, 0xffff
+    mul          t6, t0, t4
+    sra          t6, t6, 8       // t6 = z3 = MULTIPLY(tmp11, FIX_0_707106781)
+    add          t0, t6, t8      // t0 = z3 + z2
+    sub          t1, t6, t8      // t1 = z3 - z2
+    add          t3, t6, s0      // t3 = z3 + z4
+    sub          t4, t6, s0      // t4 = z3 - z4
+    sub          t5, t2, t1      // t5 = dataptr[5]
+    sub          t6, t7, t0      // t6 = dataptr[3]
+    add          t3, t2, t3      // t3 = dataptr[1]
+    add          t4, t7, t4      // t4 = dataptr[7]
+    sh           t5, 80(v0)
+    sh           t6, 48(v0)
+    sh           t3, 16(v0)
+    sh           t4, 112(v0)
+    addiu        v0, 2
+    bne          v0, v1, 1b
+     nop
+
+    RESTORE_REGS_FROM_STACK 8, s0, s1
+
+    j            ra
+     nop
+END(jsimd_fdct_ifast_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_quantize_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - divisors
+ * a2     - workspace
+ */
+
+    .set at
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2
+
+    addiu   v0, a2, 124  // v0 = workspace_end
+    lh      t0, 0(a2)
+    lh      t1, 0(a1)
+    lh      t2, 128(a1)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0, t3
+    lh      t4, 384(a1)
+    lh      t5, 130(a1)
+    lh      t6, 2(a2)
+    lh      t7, 2(a1)
+    lh      t8, 386(a1)
+
+1:
+    andi    t1, 0xffff
+    add     t9, t0, t2
+    andi    t9, 0xffff
+    mul     v1, t9, t1
+    sra     s0, t6, 15
+    sll     s0, s0, 1
+    addiu   s0, s0, 1
+    addiu   t9, t4, 16
+    srav    v1, v1, t9
+    mul     v1, v1, t3
+    mul     t6, t6, s0
+    andi    t7, 0xffff
+    addiu   a2, a2, 4
+    addiu   a1, a1, 4
+    add     s1, t6, t5
+    andi    s1, 0xffff
+    sh      v1, 0(a0)
+
+    mul     s2, s1, t7
+    addiu   s1, t8, 16
+    srav    s2, s2, s1
+    mul     s2,s2, s0
+    lh      t0, 0(a2)
+    lh      t1, 0(a1)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0, t3
+    lh      t2, 128(a1)
+    lh      t4, 384(a1)
+    lh      t5, 130(a1)
+    lh      t8, 386(a1)
+    lh      t6, 2(a2)
+    lh      t7, 2(a1)
+    sh      s2, 2(a0)
+    lh      t0, 0(a2)
+    sra     t3, t0, 15
+    sll     t3, t3, 1
+    addiu   t3, t3, 1
+    mul     t0, t0,t3
+    bne     a2, v0, 1b
+     addiu  a0, a0, 4
+
+    andi    t1, 0xffff
+    add     t9, t0, t2
+    andi    t9, 0xffff
+    mul     v1, t9, t1
+    sra     s0, t6, 15
+    sll     s0, s0, 1
+    addiu   s0, s0, 1
+    addiu   t9, t4, 16
+    srav    v1, v1, t9
+    mul     v1, v1, t3
+    mul     t6, t6, s0
+    andi    t7, 0xffff
+    sh      v1, 0(a0)
+    add     s1, t6, t5
+    andi    s1, 0xffff
+    mul     s2, s1, t7
+    addiu   s1, t8, 16
+    addiu   a2, a2, 4
+    addiu   a1, a1, 4
+    srav    s2, s2, s1
+    mul     s2, s2, s0
+    sh      s2, 2(a0)
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2
+
+    j       ra
+     nop
+
+END(jsimd_quantize_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_quantize_float_mips_dspr2)
+/*
+ * a0     - coef_block
+ * a1     - divisors
+ * a2     - workspace
+ */
+
+    .set at
+
+    li         t1, 0x46800100     //integer representation 16384.5
+    mtc1       t1, f0
+    li         t0, 63
+0:
+    lwc1       f2, 0(a2)
+    lwc1       f10, 0(a1)
+    lwc1       f4, 4(a2)
+    lwc1       f12, 4(a1)
+    lwc1       f6, 8(a2)
+    lwc1       f14, 8(a1)
+    lwc1       f8, 12(a2)
+    lwc1       f16, 12(a1)
+    madd.s     f2, f0, f2, f10
+    madd.s     f4, f0, f4, f12
+    madd.s     f6, f0, f6, f14
+    madd.s     f8, f0, f8, f16
+    lwc1       f10, 16(a1)
+    lwc1       f12, 20(a1)
+    trunc.w.s  f2, f2
+    trunc.w.s  f4, f4
+    trunc.w.s  f6, f6
+    trunc.w.s  f8, f8
+    lwc1       f14, 24(a1)
+    lwc1       f16, 28(a1)
+    mfc1       t1, f2
+    mfc1       t2, f4
+    mfc1       t3, f6
+    mfc1       t4, f8
+    lwc1       f2, 16(a2)
+    lwc1       f4, 20(a2)
+    lwc1       f6, 24(a2)
+    lwc1       f8, 28(a2)
+    madd.s     f2, f0, f2, f10
+    madd.s     f4, f0, f4, f12
+    madd.s     f6, f0, f6, f14
+    madd.s     f8, f0, f8, f16
+    addiu      t1, t1, -16384
+    addiu      t2, t2, -16384
+    addiu      t3, t3, -16384
+    addiu      t4, t4, -16384
+    trunc.w.s  f2, f2
+    trunc.w.s  f4, f4
+    trunc.w.s  f6, f6
+    trunc.w.s  f8, f8
+    sh         t1, 0(a0)
+    sh         t2, 2(a0)
+    sh         t3, 4(a0)
+    sh         t4, 6(a0)
+    mfc1       t1, f2
+    mfc1       t2, f4
+    mfc1       t3, f6
+    mfc1       t4, f8
+    addiu      t0, t0, -8
+    addiu      a2, a2, 32
+    addiu      a1, a1, 32
+    addiu      t1, t1, -16384
+    addiu      t2, t2, -16384
+    addiu      t3, t3, -16384
+    addiu      t4, t4, -16384
+    sh         t1, 8(a0)
+    sh         t2, 10(a0)
+    sh         t3, 12(a0)
+    sh         t4, 14(a0)
+    bgez       t0, 0b
+     addiu     a0, a0, 16
+
+    j          ra
+     nop
+
+END(jsimd_quantize_float_mips_dspr2)
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_2x2_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 24, s0, s1, s2, s3, s4, s5
+
+    addiu     sp, sp, -40
+    move      v0, sp
+    addiu     s2, zero, 29692
+    addiu     s3, zero, -10426
+    addiu     s4, zero, 6967
+    addiu     s5, zero, -5906
+    lh        t0, 0(a1)         // t0 = inptr[DCTSIZE*0]
+    lh        t5, 0(a0)         // t5 = quantptr[DCTSIZE*0]
+    lh        t1, 48(a1)        // t1 = inptr[DCTSIZE*3]
+    lh        t6, 48(a0)        // t6 = quantptr[DCTSIZE*3]
+    mul       t4, t5, t0
+    lh        t0, 16(a1)        // t0 = inptr[DCTSIZE*1]
+    lh        t5, 16(a0)        // t5 = quantptr[DCTSIZE*1]
+    mul       t6, t6, t1
+    mul       t5, t5, t0
+    lh        t2, 80(a1)        // t2 = inptr[DCTSIZE*5]
+    lh        t7, 80(a0)        // t7 = quantptr[DCTSIZE*5]
+    lh        t3, 112(a1)       // t3 = inptr[DCTSIZE*7]
+    lh        t8, 112(a0)       // t8 = quantptr[DCTSIZE*7]
+    mul       t7, t7, t2
+    mult      zero, zero
+    mul       t8, t8, t3
+    li        s0, 0x73FCD746    // s0 = (29692 << 16) | (-10426 & 0xffff)
+    li        s1, 0x1B37E8EE    // s1 = (6967 << 16) | (-5906 & 0xffff)
+    ins       t6, t5, 16, 16    // t6 = t5|t6
+    sll       t4, t4, 15
+    dpa.w.ph  $ac0, t6, s0
+    lh        t1, 2(a1)
+    lh        t6, 2(a0)
+    ins       t8, t7, 16, 16    // t8 = t7|t8
+    dpa.w.ph  $ac0, t8, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 18(a1)
+    lh        t6, 18(a0)
+    lh        t2, 50(a1)
+    lh        t7, 50(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 82(a1)
+    lh        t2, 82(a0)
+    lh        t3, 114(a1)
+    lh        t4, 114(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 0(v0)
+    sw        t8, 20(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 6(a1)
+    lh        t6, 6(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 22(a1)
+    lh        t6, 22(a0)
+    lh        t2, 54(a1)
+    lh        t7, 54(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 86(a1)
+    lh        t2, 86(a0)
+    lh        t3, 118(a1)
+    lh        t4, 118(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 4(v0)
+    sw        t8, 24(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 10(a1)
+    lh        t6, 10(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 26(a1)
+    lh        t6, 26(a0)
+    lh        t2, 58(a1)
+    lh        t7, 58(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 90(a1)
+    lh        t2, 90(a0)
+    lh        t3, 122(a1)
+    lh        t4, 122(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 8(v0)
+    sw        t8, 28(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    lh        t1, 14(a1)
+    lh        t6, 14(a0)
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    mul       t5, t6, t1
+    lh        t1, 30(a1)
+    lh        t6, 30(a0)
+    lh        t2, 62(a1)
+    lh        t7, 62(a0)
+    mul       t6, t6, t1
+    subu      t8, t4, t0
+    mul       t7, t7, t2
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    lh        t1, 94(a1)
+    lh        t2, 94(a0)
+    lh        t3, 126(a1)
+    lh        t4, 126(a0)
+    shra_r.w  t8, t8, 13
+    mul       t1, t1, t2
+    mul       t3, t3, t4
+    sw        t0, 12(v0)
+    sw        t8, 32(v0)
+    sll       t4, t5, 15
+    ins       t7, t6, 16, 16
+    mult      zero, zero
+    dpa.w.ph  $ac0, t7, s0
+    ins       t3, t1, 16, 16
+    dpa.w.ph  $ac0, t3, s1
+    mflo      t0, $ac0
+    lw        t9, 0(a2)
+    lw        t3, 0(v0)
+    lw        t7, 4(v0)
+    lw        t1, 8(v0)
+    addu      t9, t9, a3
+    sll       t3, t3, 15
+    subu      t8, t4, t0
+    addu      t0, t4, t0
+    shra_r.w  t0, t0, 13
+    shra_r.w  t8, t8, 13
+    sw        t0, 16(v0)
+    sw        t8, 36(v0)
+    lw        t5, 12(v0)
+    lw        t6, 16(v0)
+    mult      t7, s2
+    madd      t1, s3
+    madd      t5, s4
+    madd      t6, s5
+    lw        t5, 24(v0)
+    lw        t7, 28(v0)
+    mflo      t0, $ac0
+    lw        t8, 32(v0)
+    lw        t2, 36(v0)
+    mult      $ac1, t5, s2
+    madd      $ac1, t7, s3
+    madd      $ac1, t8, s4
+    madd      $ac1, t2, s5
+    addu      t1, t3, t0
+    subu      t6, t3, t0
+    shra_r.w  t1, t1, 20
+    shra_r.w  t6, t6, 20
+    mflo      t4, $ac1
+    shll_s.w  t1, t1, 24
+    shll_s.w  t6, t6, 24
+    sra       t1, t1, 24
+    sra       t6, t6, 24
+    addiu     t1, t1, 128
+    addiu     t6, t6, 128
+    lw        t0, 20(v0)
+    sb        t1, 0(t9)
+    sb        t6, 1(t9)
+    sll       t0, t0, 15
+    lw        t9, 4(a2)
+    addu      t1, t0, t4
+    subu      t6, t0, t4
+    addu      t9, t9, a3
+    shra_r.w  t1, t1, 20
+    shra_r.w  t6, t6, 20
+    shll_s.w  t1, t1, 24
+    shll_s.w  t6, t6, 24
+    sra       t1, t1, 24
+    sra       t6, t6, 24
+    addiu     t1, t1, 128
+    addiu     t6, t6, 128
+    sb        t1, 0(t9)
+    sb        t6, 1(t9)
+    addiu     sp, sp, 40
+
+    RESTORE_REGS_FROM_STACK 24, s0, s1, s2, s3, s4, s5
+
+    j         ra
+     nop
+
+END(jsimd_idct_2x2_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_4x4_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ * 16(sp) - workspace[DCTSIZE*4];  // buffers data between passes
+ */
+
+    .set at
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    lw        v1, 48(sp)
+    move      t0, a1
+    move      t1, v1
+    li        t9, 4
+    li        s0, 0x2e75f93e
+    li        s1, 0x21f9ba79
+    li        s2, 0xecc2efb0
+    li        s3, 0x52031ccd
+
+0:
+    lh        s6, 32(t0)        // inptr[DCTSIZE*2]
+    lh        t6, 32(a0)        // quantptr[DCTSIZE*2]
+    lh        s7, 96(t0)        // inptr[DCTSIZE*6]
+    lh        t7, 96(a0)        // quantptr[DCTSIZE*6]
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        s4, 0(t0)         // inptr[DCTSIZE*0]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s5, 0(a0)         // quantptr[0]
+    li        s6, 15137
+    li        s7, 6270
+    mul       t2, s4, s5        // tmp0 = (inptr[0] * quantptr[0])
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        t5, 112(t0)       // inptr[DCTSIZE*7]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s4, 112(a0)       // quantptr[DCTSIZE*7]
+    lh        v0, 80(t0)        // inptr[DCTSIZE*5]
+    lh        s5, 80(a0)        // quantptr[DCTSIZE*5]
+    lh        s6, 48(a0)        // quantptr[DCTSIZE*3]
+    sll       t2, t2, 14        // tmp0 <<= (CONST_BITS+1)
+    lh        s7, 16(a0)        // quantptr[DCTSIZE*1]
+    lh        t8, 16(t0)        // inptr[DCTSIZE*1]
+    subu      t6, t6, t7        // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6)
+    lh        t7, 48(t0)        // inptr[DCTSIZE*3]
+    mul       t5, s4, t5        // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7])
+    mul       v0, s5, v0        // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5])
+    mul       t7, s6, t7        // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3])
+    mul       t8, s7, t8        // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1])
+    addu      t3, t2, t6        // tmp10 = tmp0 + z2
+    subu      t4, t2, t6        // tmp10 = tmp0 - z2
+    mult      $ac0, zero, zero
+    mult      $ac1, zero, zero
+    ins       t5, v0, 16, 16
+    ins       t7, t8, 16, 16
+    addiu     t9, t9, -1
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    mflo      s4, $ac0
+    mflo      s5, $ac1
+    addiu     a0, a0, 2
+    addiu     t1, t1, 4
+    addiu     t0, t0, 2
+    addu      t6, t4, s4
+    subu      t5, t4, s4
+    addu      s6, t3, s5
+    subu      s7, t3, s5
+    shra_r.w  t6, t6, 12        // DESCALE(tmp12 + temp1, 12)
+    shra_r.w  t5, t5, 12        // DESCALE(tmp12 - temp1, 12)
+    shra_r.w  s6, s6, 12        // DESCALE(tmp10 + temp2, 12)
+    shra_r.w  s7, s7, 12        // DESCALE(tmp10 - temp2, 12)
+    sw        t6, 28(t1)
+    sw        t5, 60(t1)
+    sw        s6, -4(t1)
+    bgtz      t9, 0b
+     sw       s7, 92(t1)
+    // second loop three pass
+    li        t9, 3
+1:
+    lh        s6, 34(t0)        // inptr[DCTSIZE*2]
+    lh        t6, 34(a0)        // quantptr[DCTSIZE*2]
+    lh        s7, 98(t0)        // inptr[DCTSIZE*6]
+    lh        t7, 98(a0)        // quantptr[DCTSIZE*6]
+    mul       t6, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        s4, 2(t0)         // inptr[DCTSIZE*0]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s5, 2(a0)         // quantptr[DCTSIZE*0]
+    li        s6, 15137
+    li        s7, 6270
+    mul       t2, s4, s5        // tmp0 = (inptr[0] * quantptr[0])
+    mul       v0, s6, t6        // z2 = (inptr[DCTSIZE*2] * quantptr[DCTSIZE*2])
+    lh        t5, 114(t0)       // inptr[DCTSIZE*7]
+    mul       t7, s7, t7        // z3 = (inptr[DCTSIZE*6] * quantptr[DCTSIZE*6])
+    lh        s4, 114(a0)       // quantptr[DCTSIZE*7]
+    lh        s5, 82(a0)        // quantptr[DCTSIZE*5]
+    lh        t6, 82(t0)        // inptr[DCTSIZE*5]
+    sll       t2, t2, 14        // tmp0 <<= (CONST_BITS+1)
+    lh        s6, 50(a0)        // quantptr[DCTSIZE*3]
+    lh        t8, 18(t0)        // inptr[DCTSIZE*1]
+    subu      v0, v0, t7        // tmp2 = MULTIPLY(z2, t5) - MULTIPLY(z3, t6)
+    lh        t7, 50(t0)        // inptr[DCTSIZE*3]
+    lh        s7, 18(a0)        // quantptr[DCTSIZE*1]
+    mul       t5, s4, t5        // z1 = (inptr[DCTSIZE*7] * quantptr[DCTSIZE*7])
+    mul       t6, s5, t6        // z2 = (inptr[DCTSIZE*5] * quantptr[DCTSIZE*5])
+    mul       t7, s6, t7        // z3 = (inptr[DCTSIZE*3] * quantptr[DCTSIZE*3])
+    mul       t8, s7, t8        // z4 = (inptr[DCTSIZE*1] * quantptr[DCTSIZE*1])
+    addu      t3, t2, v0        // tmp10 = tmp0 + z2
+    subu      t4, t2, v0        // tmp10 = tmp0 - z2
+    mult      $ac0, zero, zero
+    mult      $ac1, zero, zero
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    mflo      t5, $ac0
+    mflo      t6, $ac1
+    addiu     t9, t9, -1
+    addiu     t0, t0, 2
+    addiu     a0, a0, 2
+    addiu     t1, t1, 4
+    addu      s5, t4, t5
+    subu      s4, t4, t5
+    addu      s6, t3, t6
+    subu      s7, t3, t6
+    shra_r.w  s5, s5, 12        // DESCALE(tmp12 + temp1, 12)
+    shra_r.w  s4, s4, 12        // DESCALE(tmp12 - temp1, 12)
+    shra_r.w  s6, s6, 12        // DESCALE(tmp10 + temp2, 12)
+    shra_r.w  s7, s7, 12        // DESCALE(tmp10 - temp2, 12)
+    sw        s5, 32(t1)
+    sw        s4, 64(t1)
+    sw        s6, 0(t1)
+    bgtz      t9, 1b
+     sw       s7, 96(t1)
+    move      t1, v1
+    li        s4, 15137
+    lw        s6, 8(t1)         // wsptr[2]
+    li        s5, 6270
+    lw        s7, 24(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 0(t1)         // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 28(t1)        // wsptr[7]
+    lh        t6, 20(t1)        // wsptr[5]
+    lh        t7, 12(t1)        // wsptr[3]
+    lh        t8, 4(t1)         // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 0(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    // 2
+    li        s4, 15137
+    lw        s6, 40(t1)        // wsptr[2]
+    li        s5, 6270
+    lw        s7, 56(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 32(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 60(t1)        // wsptr[7]
+    lh        t6, 52(t1)        // wsptr[5]
+    lh        t7, 44(t1)        // wsptr[3]
+    lh        t8, 36(t1)        // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, CONST_BITS-PASS1_BITS+1)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, CONST_BITS-PASS1_BITS+1)
+    sll       s4, t9, 2
+    lw        v0, 4(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    // 3
+    li        s4, 15137
+    lw        s6, 72(t1)        // wsptr[2]
+    li        s5, 6270
+    lw        s7, 88(t1)        // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 64(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], - FIX_0_765366865)
+    lh        t5, 92(t1)        // wsptr[7]
+    lh        t6, 84(t1)        // wsptr[5]
+    lh        t7, 76(t1)        // wsptr[3]
+    lh        t8, 68(t1)        // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 8(a2)         // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+    li        s4, 15137
+    lw        s6, 104(t1)       // wsptr[2]
+    li        s5, 6270
+    lw        s7, 120(t1)       // wsptr[6]
+    mul       s4, s4, s6        // MULTIPLY((JLONG) wsptr[2], FIX_1_847759065)
+    lw        t2, 96(t1)        // wsptr[0]
+    mul       s5, s5, s7        // MULTIPLY((JLONG) wsptr[6], -FIX_0_765366865)
+    lh        t5, 124(t1)       // wsptr[7]
+    lh        t6, 116(t1)       // wsptr[5]
+    lh        t7, 108(t1)       // wsptr[3]
+    lh        t8, 100(t1)       // wsptr[1]
+    ins       t5, t6, 16, 16
+    ins       t7, t8, 16, 16
+    mult      $ac0, zero, zero
+    dpa.w.ph  $ac0, t5, s0
+    dpa.w.ph  $ac0, t7, s1
+    mult      $ac1, zero, zero
+    dpa.w.ph  $ac1, t5, s2
+    dpa.w.ph  $ac1, t7, s3
+    sll       t2, t2, 14        // tmp0 = ((JLONG) wsptr[0]) << (CONST_BITS+1)
+    mflo      s6, $ac0
+    // MULTIPLY(wsptr[2], FIX_1_847759065 + MULTIPLY(wsptr[6], -FIX_0_765366865)
+    subu      s4, s4, s5
+    addu      t3, t2, s4        // tmp10 = tmp0 + z2;
+    mflo      s7, $ac1
+    subu      t4, t2, s4        // tmp10 = tmp0 - z2;
+    addu      t7, t4, s6
+    subu      t8, t4, s6
+    addu      t5, t3, s7
+    subu      t6, t3, s7
+    shra_r.w  t5, t5, 19        // DESCALE(tmp10 + temp2, 19)
+    shra_r.w  t6, t6, 19        // DESCALE(tmp10 - temp2, 19)
+    shra_r.w  t7, t7, 19        // DESCALE(tmp12 + temp1, 19)
+    shra_r.w  t8, t8, 19        // DESCALE(tmp12 - temp1, 19)
+    sll       s4, t9, 2
+    lw        v0, 12(a2)        // output_buf[ctr]
+    shll_s.w  t5, t5, 24
+    shll_s.w  t6, t6, 24
+    shll_s.w  t7, t7, 24
+    shll_s.w  t8, t8, 24
+    sra       t5, t5, 24
+    sra       t6, t6, 24
+    sra       t7, t7, 24
+    sra       t8, t8, 24
+    addu      v0, v0, a3        // outptr = output_buf[ctr] + output_col
+    addiu     t5, t5, 128
+    addiu     t6, t6, 128
+    addiu     t7, t7, 128
+    addiu     t8, t8, 128
+    sb        t5, 0(v0)
+    sb        t7, 1(v0)
+    sb        t8, 2(v0)
+    sb        t6, 3(v0)
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+END(jsimd_idct_4x4_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_6x6_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - output_buf
+ * a3     - output_col
+ */
+    .set at
+
+    SAVE_REGS_ON_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    addiu     sp, sp, -144
+    move      v0, sp
+    addiu     v1, v0, 24
+    addiu     t9, zero, 5793
+    addiu     s0, zero, 10033
+    addiu     s1, zero, 2998
+
+1:
+    lh        s2, 0(a0)   // q0 = quantptr[ 0]
+    lh        s3, 32(a0)  // q1 = quantptr[16]
+    lh        s4, 64(a0)  // q2 = quantptr[32]
+    lh        t2, 64(a1)  // tmp2 = inptr[32]
+    lh        t1, 32(a1)  // tmp1 = inptr[16]
+    lh        t0, 0(a1)   // tmp0 = inptr[ 0]
+    mul       t2, t2, s4  // tmp2 = tmp2 * q2
+    mul       t1, t1, s3  // tmp1 = tmp1 * q1
+    mul       t0, t0, s2  // tmp0 = tmp0 * q0
+    lh        t6, 16(a1)  // z1 = inptr[ 8]
+    lh        t8, 80(a1)  // z3 = inptr[40]
+    lh        t7, 48(a1)  // z2 = inptr[24]
+    lh        s2, 16(a0)  // q0 = quantptr[ 8]
+    lh        s4, 80(a0)  // q2 = quantptr[40]
+    lh        s3, 48(a0)  // q1 = quantptr[24]
+    mul       t2, t2, t9  // tmp2 = tmp2 * 5793
+    mul       t1, t1, s0  // tmp1 = tmp1 * 10033
+    sll       t0, t0, 13  // tmp0 = tmp0 << 13
+    mul       t6, t6, s2  // z1 = z1 * q0
+    mul       t8, t8, s4  // z3 = z3 * q2
+    mul       t7, t7, s3  // z2 = z2 * q1
+    addu      t3, t0, t2  // tmp10 = tmp0 + tmp2
+    sll       t2, t2, 1   // tmp2 = tmp2 << 2
+    subu      t4, t0, t2  // tmp11 = tmp0 - tmp2;
+    subu      t5, t3, t1  // tmp12 = tmp10 - tmp1
+    addu      t3, t3, t1  // tmp10 = tmp10 + tmp1
+    addu      t1, t6, t8  // tmp1 = z1 + z3
+    mul       t1, t1, s1  // tmp1 = tmp1 * 2998
+    shra_r.w  t4, t4, 11  // tmp11 = (tmp11 + 1024) >> 11
+    subu      t2, t6, t8  // tmp2 = z1 - z3
+    subu      t2, t2, t7  // tmp2 = tmp2 - z2
+    sll       t2, t2, 2   // tmp2 = tmp2 << 2
+    addu      t0, t6, t7  // tmp0 = z1 + z2
+    sll       t0, t0, 13  // tmp0 = tmp0 << 13
+    subu      s2, t8, t7  // q0 = z3 - z2
+    sll       s2, s2, 13  // q0 = q0 << 13
+    addu      t0, t0, t1  // tmp0 = tmp0 + tmp1
+    addu      t1, s2, t1  // tmp1 = q0 + tmp1
+    addu      s2, t4, t2  // q0 = tmp11 + tmp2
+    subu      s3, t4, t2  // q1 = tmp11 - tmp2
+    addu      t6, t3, t0  // z1 = tmp10 + tmp0
+    subu      t7, t3, t0  // z2 = tmp10 - tmp0
+    addu      t4, t5, t1  // tmp11 = tmp12 + tmp1
+    subu      t5, t5, t1  // tmp12 = tmp12 - tmp1
+    shra_r.w  t6, t6, 11  // z1 = (z1 + 1024) >> 11
+    shra_r.w  t7, t7, 11  // z2 = (z2 + 1024) >> 11
+    shra_r.w  t4, t4, 11  // tmp11 = (tmp11 + 1024) >> 11
+    shra_r.w  t5, t5, 11  // tmp12 = (tmp12 + 1024) >> 11
+    sw        s2, 24(v0)
+    sw        s3, 96(v0)
+    sw        t6, 0(v0)
+    sw        t7, 120(v0)
+    sw        t4, 48(v0)
+    sw        t5, 72(v0)
+    addiu     v0, v0, 4
+    addiu     a1, a1, 2
+    bne       v0, v1, 1b
+     addiu    a0, a0, 2
+
+    /* Pass 2: process 6 rows from work array, store into output array. */
+    move      v0, sp
+    addiu     v1, v0, 144
+
+2:
+    lw        t0, 0(v0)
+    lw        t2, 16(v0)
+    lw        s5, 0(a2)
+    addiu     t0, t0, 16
+    sll       t0, t0, 13
+    mul       t3, t2, t9
+    lw        t6, 4(v0)
+    lw        t8, 20(v0)
+    lw        t7, 12(v0)
+    addu      s5, s5, a3
+    addu      s6, t6, t8
+    mul       s6, s6, s1
+    addu      t1, t0, t3
+    subu      t4, t0, t3
+    subu      t4, t4, t3
+    lw        t3, 8(v0)
+    mul       t0, t3, s0
+    addu      s7, t6, t7
+    sll       s7, s7, 13
+    addu      s7, s6, s7
+    subu      t2, t8, t7
+    sll       t2, t2, 13
+    addu      t2, s6, t2
+    subu      s6, t6, t7
+    subu      s6, s6, t8
+    sll       s6, s6, 13
+    addu      t3, t1, t0
+    subu      t5, t1, t0
+    addu      t6, t3, s7
+    subu      t3, t3, s7
+    addu      t7, t4, s6
+    subu      t4, t4, s6
+    addu      t8, t5, t2
+    subu      t5, t5, t2
+    shll_s.w  t6, t6, 6
+    shll_s.w  t3, t3, 6
+    shll_s.w  t7, t7, 6
+    shll_s.w  t4, t4, 6
+    shll_s.w  t8, t8, 6
+    shll_s.w  t5, t5, 6
+    sra       t6, t6, 24
+    addiu     t6, t6, 128
+    sra       t3, t3, 24
+    addiu     t3, t3, 128
+    sb        t6, 0(s5)
+    sra       t7, t7, 24
+    addiu     t7, t7, 128
+    sb        t3, 5(s5)
+    sra       t4, t4, 24
+    addiu     t4, t4, 128
+    sb        t7, 1(s5)
+    sra       t8, t8, 24
+    addiu     t8, t8, 128
+    sb        t4, 4(s5)
+    addiu     v0, v0, 24
+    sra       t5, t5, 24
+    addiu     t5, t5, 128
+    sb        t8, 2(s5)
+    addiu     a2, a2,  4
+    bne       v0, v1, 2b
+     sb       t5, 3(s5)
+
+    addiu     sp, sp, 144
+
+    RESTORE_REGS_FROM_STACK 32, s0, s1, s2, s3, s4, s5, s6, s7
+
+    j         ra
+     nop
+
+END(jsimd_idct_6x6_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass1_mips_dspr2)
+/*
+ * a0     - compptr->dct_table
+ * a1     - coef_block
+ * a2     - workspace
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li         a3, 8
+
+1:
+    // odd part
+    lh         t0, 48(a1)
+    lh         t1, 48(a0)
+    lh         t2, 16(a1)
+    lh         t3, 16(a0)
+    lh         t4, 80(a1)
+    lh         t5, 80(a0)
+    lh         t6, 112(a1)
+    lh         t7, 112(a0)
+    mul        t0, t0, t1    // z2
+    mul        t1, t2, t3    // z1
+    mul        t2, t4, t5    // z3
+    mul        t3, t6, t7    // z4
+    li         t4, 10703     // FIX(1.306562965)
+    li         t5, 4433      // FIX_0_541196100
+    li         t6, 7053      // FIX(0.860918669)
+    mul        t4, t0,t4     // tmp11
+    mul        t5, t0,t5     // -tmp14
+    addu       t7, t1,t2     // tmp10
+    addu       t8, t7,t3     // tmp10 + z4
+    mul        t6, t6, t8    // tmp15
+    li         t8, 2139      // FIX(0.261052384)
+    mul        t8, t7, t8    // MULTIPLY(tmp10, FIX(0.261052384))
+    li         t7, 2295      // FIX(0.280143716)
+    mul        t7, t1, t7    // MULTIPLY(z1, FIX(0.280143716))
+    addu       t9, t2, t3    // z3 + z4
+    li         s0, 8565      // FIX(1.045510580)
+    mul        t9, t9, s0    // -tmp13
+    li         s0, 12112     // FIX(1.478575242)
+    mul        s0, t2, s0    // MULTIPLY(z3, FIX(1.478575242)
+    li         s1, 12998     // FIX(1.586706681)
+    mul        s1, t3, s1    // MULTIPLY(z4, FIX(1.586706681))
+    li         s2, 5540      // FIX(0.676326758)
+    mul        s2, t1, s2    // MULTIPLY(z1, FIX(0.676326758))
+    li         s3, 16244     // FIX(1.982889723)
+    mul        s3, t3, s3    // MULTIPLY(z4, FIX(1.982889723))
+    subu       t1, t1, t3    // z1-=z4
+    subu       t0, t0, t2    // z2-=z3
+    addu       t2, t0, t1    // z1+z2
+    li         t3, 4433      // FIX_0_541196100
+    mul        t2, t2, t3    // z3
+    li         t3, 6270      // FIX_0_765366865
+    mul        t1, t1, t3    // MULTIPLY(z1, FIX_0_765366865)
+    li         t3, 15137     // FIX_0_765366865
+    mul        t0, t0, t3    // MULTIPLY(z2, FIX_1_847759065)
+    addu       t8, t6, t8    // tmp12
+    addu       t3, t8, t4    // tmp12 + tmp11
+    addu       t3, t3, t7    // tmp10
+    subu       t8, t8, t9    // tmp12 + tmp13
+    addu       s0, t5, s0
+    subu       t8, t8, s0    // tmp12
+    subu       t9, t6, t9
+    subu       s1, s1, t4
+    addu       t9, t9, s1    // tmp13
+    subu       t6, t6, t5
+    subu       t6, t6, s2
+    subu       t6, t6, s3    // tmp15
+    // even part start
+    lh         t4, 64(a1)
+    lh         t5, 64(a0)
+    lh         t7, 32(a1)
+    lh         s0, 32(a0)
+    lh         s1, 0(a1)
+    lh         s2, 0(a0)
+    lh         s3, 96(a1)
+    lh         v0, 96(a0)
+    mul        t4, t4, t5    // DEQUANTIZE(inptr[DCTSIZE*4],quantptr[DCTSIZE*4])
+    mul        t5, t7, s0    // DEQUANTIZE(inptr[DCTSIZE*2],quantptr[DCTSIZE*2])
+    mul        t7, s1, s2    // DEQUANTIZE(inptr[DCTSIZE*0],quantptr[DCTSIZE*0])
+    mul        s0, s3, v0    // DEQUANTIZE(inptr[DCTSIZE*6],quantptr[DCTSIZE*6])
+    // odd part end
+    addu       t1, t2, t1    // tmp11
+    subu       t0, t2, t0    // tmp14
+    // update counter and pointers
+    addiu      a3, a3, -1
+    addiu      a0, a0, 2
+    addiu      a1, a1, 2
+    // even part rest
+    li         s1, 10033
+    li         s2, 11190
+    mul        t4, t4, s1    // z4
+    mul        s1, t5, s2    // z4
+    sll        t5, t5, 13    // z1
+    sll        t7, t7, 13
+    addiu      t7, t7, 1024  // z3
+    sll        s0, s0, 13    // z2
+    addu       s2, t7, t4    // tmp10
+    subu       t4, t7, t4    // tmp11
+    subu       s3, t5, s0    // tmp12
+    addu       t2, t7, s3    // tmp21
+    subu       s3, t7, s3    // tmp24
+    addu       t7, s1, s0    // tmp12
+    addu       v0, s2, t7    // tmp20
+    subu       s2, s2, t7    // tmp25
+    subu       s1, s1, t5    // z4 - z1
+    subu       s1, s1, s0    // tmp12
+    addu       s0, t4, s1    // tmp22
+    subu       t4, t4, s1    // tmp23
+    // final output stage
+    addu       t5, v0, t3
+    subu       v0, v0, t3
+    addu       t3, t2, t1
+    subu       t2, t2, t1
+    addu       t1, s0, t8
+    subu       s0, s0, t8
+    addu       t8, t4, t9
+    subu       t4, t4, t9
+    addu       t9, s3, t0
+    subu       s3, s3, t0
+    addu       t0, s2, t6
+    subu       s2, s2, t6
+    sra        t5, t5, 11
+    sra        t3, t3, 11
+    sra        t1, t1, 11
+    sra        t8, t8, 11
+    sra        t9, t9, 11
+    sra        t0, t0, 11
+    sra        s2, s2, 11
+    sra        s3, s3, 11
+    sra        t4, t4, 11
+    sra        s0, s0, 11
+    sra        t2, t2, 11
+    sra        v0, v0, 11
+    sw         t5, 0(a2)
+    sw         t3, 32(a2)
+    sw         t1, 64(a2)
+    sw         t8, 96(a2)
+    sw         t9, 128(a2)
+    sw         t0, 160(a2)
+    sw         s2, 192(a2)
+    sw         s3, 224(a2)
+    sw         t4, 256(a2)
+    sw         s0, 288(a2)
+    sw         t2, 320(a2)
+    sw         v0, 352(a2)
+    bgtz       a3, 1b
+     addiu     a2, a2, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    j          ra
+     nop
+
+END(jsimd_idct_12x12_pass1_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_idct_12x12_pass2_mips_dspr2)
+/*
+ * a0     - workspace
+ * a1     - output
+ */
+
+    SAVE_REGS_ON_STACK 16, s0, s1, s2, s3
+
+    li        a3, 12
+
+1:
+    // Odd part
+    lw        t0, 12(a0)
+    lw        t1, 4(a0)
+    lw        t2, 20(a0)
+    lw        t3, 28(a0)
+    li        t4, 10703     // FIX(1.306562965)
+    li        t5, 4433      // FIX_0_541196100
+    mul       t4, t0, t4    // tmp11
+    mul       t5, t0, t5    // -tmp14
+    addu      t6, t1, t2    // tmp10
+    li        t7, 2139      // FIX(0.261052384)
+    mul       t7, t6, t7    // MULTIPLY(tmp10, FIX(0.261052384))
+    addu      t6, t6, t3    // tmp10 + z4
+    li        t8, 7053      // FIX(0.860918669)
+    mul       t6, t6, t8    // tmp15
+    li        t8, 2295      // FIX(0.280143716)
+    mul       t8, t1, t8    // MULTIPLY(z1, FIX(0.280143716))
+    addu      t9, t2, t3    // z3 + z4
+    li        s0, 8565      // FIX(1.045510580)
+    mul       t9, t9, s0    // -tmp13
+    li        s0, 12112     // FIX(1.478575242)
+    mul       s0, t2, s0    // MULTIPLY(z3, FIX(1.478575242))
+    li        s1, 12998     // FIX(1.586706681)
+    mul       s1, t3, s1    // MULTIPLY(z4, FIX(1.586706681))
+    li        s2, 5540      // FIX(0.676326758)
+    mul       s2, t1, s2    // MULTIPLY(z1, FIX(0.676326758))
+    li        s3, 16244     // FIX(1.982889723)
+    mul       s3, t3, s3    // MULTIPLY(z4, FIX(1.982889723))
+    subu      t1, t1, t3    // z1 -= z4
+    subu      t0, t0, t2    // z2 -= z3
+    addu      t2, t1, t0    // z1 + z2
+    li        t3, 4433      // FIX_0_541196100
+    mul       t2, t2, t3    // z3
+    li        t3, 6270      // FIX_0_765366865
+    mul       t1, t1, t3    // MULTIPLY(z1, FIX_0_765366865)
+    li        t3, 15137     // FIX_1_847759065
+    mul       t0, t0, t3    // MULTIPLY(z2, FIX_1_847759065)
+    addu      t3, t6, t7    // tmp12
+    addu      t7, t3, t4
+    addu      t7, t7, t8    // tmp10
+    subu      t3, t3, t9
+    subu      t3, t3, t5
+    subu      t3, t3, s0    // tmp12
+    subu      t9, t6, t9
+    subu      t9, t9, t4
+    addu      t9, t9, s1    // tmp13
+    subu      t6, t6, t5
+    subu      t6, t6, s2
+    subu      t6, t6, s3    // tmp15
+    addu      t1, t2, t1    // tmp11
+    subu      t0, t2, t0    // tmp14
+    // even part
+    lw        t2, 16(a0)    // z4
+    lw        t4, 8(a0)     // z1
+    lw        t5, 0(a0)     // z3
+    lw        t8, 24(a0)    // z2
+    li        s0, 10033     // FIX(1.224744871)
+    li        s1, 11190     // FIX(1.366025404)
+    mul       t2, t2, s0    // z4
+    mul       s0, t4, s1    // z4
+    addiu     t5, t5, 0x10
+    sll       t5, t5, 13    // z3
+    sll       t4, t4, 13    // z1
+    sll       t8, t8, 13    // z2
+    subu      s1, t4, t8    // tmp12
+    addu      s2, t5, t2    // tmp10
+    subu      t2, t5, t2    // tmp11
+    addu      s3, t5, s1    // tmp21
+    subu      s1, t5, s1    // tmp24
+    addu      t5, s0, t8    // tmp12
+    addu      v0, s2, t5    // tmp20
+    subu      t5, s2, t5    // tmp25
+    subu      t4, s0, t4
+    subu      t4, t4, t8    // tmp12
+    addu      t8, t2, t4    // tmp22
+    subu      t2, t2, t4    // tmp23
+    // increment counter and pointers
+    addiu     a3, a3, -1
+    addiu     a0, a0, 32
+    // Final stage
+    addu      t4, v0, t7
+    subu      v0, v0, t7
+    addu      t7, s3, t1
+    subu      s3, s3, t1
+    addu      t1, t8, t3
+    subu      t8, t8, t3
+    addu      t3, t2, t9
+    subu      t2, t2, t9
+    addu      t9, s1, t0
+    subu      s1, s1, t0
+    addu      t0, t5, t6
+    subu      t5, t5, t6
+    sll       t4, t4, 4
+    sll       t7, t7, 4
+    sll       t1, t1, 4
+    sll       t3, t3, 4
+    sll       t9, t9, 4
+    sll       t0, t0, 4
+    sll       t5, t5, 4
+    sll       s1, s1, 4
+    sll       t2, t2, 4
+    sll       t8, t8, 4
+    sll       s3, s3, 4
+    sll       v0, v0, 4
+    shll_s.w  t4, t4, 2
+    shll_s.w  t7, t7, 2
+    shll_s.w  t1, t1, 2
+    shll_s.w  t3, t3, 2
+    shll_s.w  t9, t9, 2
+    shll_s.w  t0, t0, 2
+    shll_s.w  t5, t5, 2
+    shll_s.w  s1, s1, 2
+    shll_s.w  t2, t2, 2
+    shll_s.w  t8, t8, 2
+    shll_s.w  s3, s3, 2
+    shll_s.w  v0, v0, 2
+    srl       t4, t4, 24
+    srl       t7, t7, 24
+    srl       t1, t1, 24
+    srl       t3, t3, 24
+    srl       t9, t9, 24
+    srl       t0, t0, 24
+    srl       t5, t5, 24
+    srl       s1, s1, 24
+    srl       t2, t2, 24
+    srl       t8, t8, 24
+    srl       s3, s3, 24
+    srl       v0, v0, 24
+    lw        t6, 0(a1)
+    addiu     t4, t4, 0x80
+    addiu     t7, t7, 0x80
+    addiu     t1, t1, 0x80
+    addiu     t3, t3, 0x80
+    addiu     t9, t9, 0x80
+    addiu     t0, t0, 0x80
+    addiu     t5, t5, 0x80
+    addiu     s1, s1, 0x80
+    addiu     t2, t2, 0x80
+    addiu     t8, t8, 0x80
+    addiu     s3, s3, 0x80
+    addiu     v0, v0, 0x80
+    sb        t4, 0(t6)
+    sb        t7, 1(t6)
+    sb        t1, 2(t6)
+    sb        t3, 3(t6)
+    sb        t9, 4(t6)
+    sb        t0, 5(t6)
+    sb        t5, 6(t6)
+    sb        s1, 7(t6)
+    sb        t2, 8(t6)
+    sb        t8, 9(t6)
+    sb        s3, 10(t6)
+    sb        v0, 11(t6)
+    bgtz      a3, 1b
+     addiu    a1, a1, 4
+
+    RESTORE_REGS_FROM_STACK 16, s0, s1, s2, s3
+
+    jr        ra
+     nop
+
+END(jsimd_idct_12x12_pass2_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_convsamp_mips_dspr2)
+/*
+ * a0     - sample_data
+ * a1     - start_col
+ * a2     - workspace
+ */
+
+    lw             t0, 0(a0)
+    li             t7, 0xff80ff80
+    addu           t0, t0, a1
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    lw             t0, 4(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 0(a2)
+    usw            t4, 4(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 8(a2)
+    usw            t6, 12(a2)
+
+    lw             t0, 8(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 16(a2)
+    usw            t4, 20(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 24(a2)
+    usw            t6, 28(a2)
+
+    lw             t0, 12(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 32(a2)
+    usw            t4, 36(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 40(a2)
+    usw            t6, 44(a2)
+
+    lw             t0, 16(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 48(a2)
+    usw            t4, 52(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 56(a2)
+    usw            t6, 60(a2)
+
+    lw             t0, 20(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 64(a2)
+    usw            t4, 68(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 72(a2)
+    usw            t6, 76(a2)
+
+    lw             t0, 24(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 80(a2)
+    usw            t4, 84(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 88(a2)
+    usw            t6, 92(a2)
+
+    lw             t0, 28(a0)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu           t0, t0, a1
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    ulw            t1, 0(t0)
+    ulw            t2, 4(t0)
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 96(a2)
+    usw            t4, 100(a2)
+    preceu.ph.qbr  t3, t1
+    preceu.ph.qbl  t4, t1
+    usw            t5, 104(a2)
+    usw            t6, 108(a2)
+    preceu.ph.qbr  t5, t2
+    preceu.ph.qbl  t6, t2
+    addu.ph        t3, t3, t7
+    addu.ph        t4, t4, t7
+    addu.ph        t5, t5, t7
+    addu.ph        t6, t6, t7
+    usw            t3, 112(a2)
+    usw            t4, 116(a2)
+    usw            t5, 120(a2)
+    usw            t6, 124(a2)
+
+    j              ra
+     nop
+
+END(jsimd_convsamp_mips_dspr2)
+
+/*****************************************************************************/
+LEAF_MIPS_DSPR2(jsimd_convsamp_float_mips_dspr2)
+/*
+ * a0     - sample_data
+ * a1     - start_col
+ * a2     - workspace
+ */
+
+    .set at
+
+    lw       t0, 0(a0)
+    addu     t0, t0, a1
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 4(a0)
+    swc1     f2, 0(a2)
+    swc1     f4, 4(a2)
+    swc1     f6, 8(a2)
+    addu     t0, t0, a1
+    swc1     f8, 12(a2)
+    swc1     f10, 16(a2)
+    swc1     f12, 20(a2)
+    swc1     f14, 24(a2)
+    swc1     f16, 28(a2)
+    //elemr 1
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 8(a0)
+    swc1     f2, 32(a2)
+    swc1     f4, 36(a2)
+    swc1     f6, 40(a2)
+    addu     t0, t0, a1
+    swc1     f8, 44(a2)
+    swc1     f10, 48(a2)
+    swc1     f12, 52(a2)
+    swc1     f14, 56(a2)
+    swc1     f16, 60(a2)
+    //elemr 2
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 12(a0)
+    swc1     f2, 64(a2)
+    swc1     f4, 68(a2)
+    swc1     f6, 72(a2)
+    addu     t0, t0, a1
+    swc1     f8, 76(a2)
+    swc1     f10, 80(a2)
+    swc1     f12, 84(a2)
+    swc1     f14, 88(a2)
+    swc1     f16, 92(a2)
+    //elemr 3
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 16(a0)
+    swc1     f2, 96(a2)
+    swc1     f4, 100(a2)
+    swc1     f6, 104(a2)
+    addu     t0, t0, a1
+    swc1     f8, 108(a2)
+    swc1     f10, 112(a2)
+    swc1     f12, 116(a2)
+    swc1     f14, 120(a2)
+    swc1     f16, 124(a2)
+    //elemr 4
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 20(a0)
+    swc1     f2, 128(a2)
+    swc1     f4, 132(a2)
+    swc1     f6, 136(a2)
+    addu     t0, t0, a1
+    swc1     f8, 140(a2)
+    swc1     f10, 144(a2)
+    swc1     f12, 148(a2)
+    swc1     f14, 152(a2)
+    swc1     f16, 156(a2)
+    //elemr 5
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 24(a0)
+    swc1     f2, 160(a2)
+    swc1     f4, 164(a2)
+    swc1     f6, 168(a2)
+    addu     t0, t0, a1
+    swc1     f8, 172(a2)
+    swc1     f10, 176(a2)
+    swc1     f12, 180(a2)
+    swc1     f14, 184(a2)
+    swc1     f16, 188(a2)
+    //elemr 6
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    lw       t0, 28(a0)
+    swc1     f2, 192(a2)
+    swc1     f4, 196(a2)
+    swc1     f6, 200(a2)
+    addu     t0, t0, a1
+    swc1     f8, 204(a2)
+    swc1     f10, 208(a2)
+    swc1     f12, 212(a2)
+    swc1     f14, 216(a2)
+    swc1     f16, 220(a2)
+    //elemr 7
+    lbu      t1, 0(t0)
+    lbu      t2, 1(t0)
+    lbu      t3, 2(t0)
+    lbu      t4, 3(t0)
+    lbu      t5, 4(t0)
+    lbu      t6, 5(t0)
+    lbu      t7, 6(t0)
+    lbu      t8, 7(t0)
+    addiu    t1, t1, -128
+    addiu    t2, t2, -128
+    addiu    t3, t3, -128
+    addiu    t4, t4, -128
+    addiu    t5, t5, -128
+    addiu    t6, t6, -128
+    addiu    t7, t7, -128
+    addiu    t8, t8, -128
+    mtc1     t1, f2
+    mtc1     t2, f4
+    mtc1     t3, f6
+    mtc1     t4, f8
+    mtc1     t5, f10
+    mtc1     t6, f12
+    mtc1     t7, f14
+    mtc1     t8, f16
+    cvt.s.w  f2, f2
+    cvt.s.w  f4, f4
+    cvt.s.w  f6, f6
+    cvt.s.w  f8, f8
+    cvt.s.w  f10, f10
+    cvt.s.w  f12, f12
+    cvt.s.w  f14, f14
+    cvt.s.w  f16, f16
+    swc1     f2, 224(a2)
+    swc1     f4, 228(a2)
+    swc1     f6, 232(a2)
+    swc1     f8, 236(a2)
+    swc1     f10, 240(a2)
+    swc1     f12, 244(a2)
+    swc1     f14, 248(a2)
+    swc1     f16, 252(a2)
+
+    j        ra
+     nop
+
+END(jsimd_convsamp_float_mips_dspr2)
+
+/*****************************************************************************/
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_powerpc.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_powerpc.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_powerpc.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_powerpc.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,741 @@
+/*
+ * jsimd_powerpc.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014-2015 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * PowerPC architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+static unsigned int simd_support = ~0;
+
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = JSIMD_ALTIVEC;
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_extrgb_ycc_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_extrgbx_ycc_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_extbgr_ycc_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_extbgrx_ycc_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_extxbgr_ycc_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_extxrgb_ycc_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_rgb_ycc_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_extrgb_gray_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_extrgbx_gray_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_extbgr_gray_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_extbgrx_gray_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_extxbgr_gray_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_extxrgb_gray_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_rgb_gray_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_ycc_extrgb_convert_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_ycc_extrgbx_convert_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_ycc_extbgr_convert_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_ycc_extbgrx_convert_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_ycc_extxbgr_convert_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_ycc_extxrgb_convert_altivec;
+      break;
+    default:
+      altivecfct=jsimd_ycc_rgb_convert_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks,
+                                input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_altivec(cinfo->image_width, cinfo->max_v_samp_factor,
+                                compptr->v_samp_factor,
+                                compptr->width_in_blocks,
+                                input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_altivec(cinfo->max_v_samp_factor, cinfo->output_width,
+                              input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_altivec(cinfo->max_v_samp_factor,
+                                    compptr->downsampled_width, input_data,
+                                    output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_h2v2_extrgb_merged_upsample_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_h2v2_extrgbx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_h2v2_extbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_h2v2_extbgrx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_h2v2_extxbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_h2v2_extxrgb_merged_upsample_altivec;
+      break;
+    default:
+      altivecfct=jsimd_h2v2_merged_upsample_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*altivecfct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      altivecfct=jsimd_h2v1_extrgb_merged_upsample_altivec;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      altivecfct=jsimd_h2v1_extrgbx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGR:
+      altivecfct=jsimd_h2v1_extbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      altivecfct=jsimd_h2v1_extbgrx_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      altivecfct=jsimd_h2v1_extxbgr_merged_upsample_altivec;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      altivecfct=jsimd_h2v1_extxrgb_merged_upsample_altivec;
+      break;
+    default:
+      altivecfct=jsimd_h2v1_merged_upsample_altivec;
+      break;
+  }
+
+  altivecfct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_altivec(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_altivec(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_altivec(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_ALTIVEC)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_altivec(compptr->dct_table, coef_block, output_buf,
+                           output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return NULL;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_x86_64.c glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_x86_64.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/jsimd_x86_64.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/jsimd_x86_64.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,887 @@
+/*
+ * jsimd_x86_64.c
+ *
+ * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
+ * Copyright 2009-2011, 2014, 2016 D. R. Commander
+ * Copyright 2015 Matthieu Darbois
+ *
+ * Based on the x86 SIMD extension for IJG JPEG library,
+ * Copyright (C) 1999-2006, MIYASAKA Masaru.
+ * For conditions of distribution and use, see copyright notice in jsimdext.inc
+ *
+ * This file contains the interface between the "normal" portions
+ * of the library and the SIMD implementations when running on a
+ * 64-bit x86 architecture.
+ */
+
+#define JPEG_INTERNALS
+#include "../jinclude.h"
+#include "../jpeglib.h"
+#include "../jsimd.h"
+#include "../jdct.h"
+#include "../jsimddct.h"
+#include "jsimd.h"
+
+/*
+ * In the PIC cases, we have no guarantee that constants will keep
+ * their alignment. This macro allows us to verify it at runtime.
+ */
+#define IS_ALIGNED(ptr, order) (((size_t)ptr & ((1 << order) - 1)) == 0)
+
+#define IS_ALIGNED_SSE(ptr) (IS_ALIGNED(ptr, 4)) /* 16 byte alignment */
+
+static unsigned int simd_support = ~0;
+static unsigned int simd_huffman = 1;
+
+/*
+ * Check what SIMD accelerations are supported.
+ *
+ * FIXME: This code is racy under a multi-threaded environment.
+ */
+LOCAL(void)
+init_simd (void)
+{
+  char *env = NULL;
+
+  if (simd_support != ~0U)
+    return;
+
+  simd_support = JSIMD_SSE2 | JSIMD_SSE;
+
+  /* Force different settings through environment variables */
+  env = getenv("JSIMD_FORCENONE");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_support = 0;
+  env = getenv("JSIMD_NOHUFFENC");
+  if ((env != NULL) && (strcmp(env, "1") == 0))
+    simd_huffman = 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_ycc (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_ycc_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_rgb_gray (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_rgb_gray_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if ((RGB_PIXELSIZE != 3) && (RGB_PIXELSIZE != 4))
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_ycc_rgb_convert_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_ycc_rgb565 (void)
+{
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_rgb_ycc_convert (j_compress_ptr cinfo,
+                       JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                       JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_ycc_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_ycc_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_ycc_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_ycc_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_ycc_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_ycc_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_rgb_ycc_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_rgb_gray_convert (j_compress_ptr cinfo,
+                        JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                        JDIMENSION output_row, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPARRAY, JSAMPIMAGE, JDIMENSION, int);
+
+  switch(cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_extrgb_gray_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_extrgbx_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_extbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_extbgrx_gray_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_extxbgr_gray_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_extxrgb_gray_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_rgb_gray_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->image_width, input_buf, output_buf, output_row, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb_convert (j_decompress_ptr cinfo,
+                       JSAMPIMAGE input_buf, JDIMENSION input_row,
+                       JSAMPARRAY output_buf, int num_rows)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY, int);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_ycc_extrgb_convert_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_ycc_extrgbx_convert_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_ycc_extbgr_convert_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_ycc_extbgrx_convert_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_ycc_extxbgr_convert_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_ycc_extxrgb_convert_sse2;
+      break;
+    default:
+      sse2fct=jsimd_ycc_rgb_convert_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, input_row, output_buf, num_rows);
+}
+
+GLOBAL(void)
+jsimd_ycc_rgb565_convert (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_downsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v2_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(void)
+jsimd_h2v1_downsample (j_compress_ptr cinfo, jpeg_component_info *compptr,
+                       JSAMPARRAY input_data, JSAMPARRAY output_data)
+{
+  jsimd_h2v1_downsample_sse2(cinfo->image_width, cinfo->max_v_samp_factor,
+                             compptr->v_samp_factor, compptr->width_in_blocks,
+                             input_data, output_data);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_upsample (j_decompress_ptr cinfo,
+                     jpeg_component_info *compptr,
+                     JSAMPARRAY input_data,
+                     JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_upsample_sse2(cinfo->max_v_samp_factor, cinfo->output_width,
+                           input_data, output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_fancy_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_fancy_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v2_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(void)
+jsimd_h2v1_fancy_upsample (j_decompress_ptr cinfo,
+                           jpeg_component_info *compptr,
+                           JSAMPARRAY input_data,
+                           JSAMPARRAY *output_data_ptr)
+{
+  jsimd_h2v1_fancy_upsample_sse2(cinfo->max_v_samp_factor,
+                                 compptr->downsampled_width, input_data,
+                                 output_data_ptr);
+}
+
+GLOBAL(int)
+jsimd_can_h2v2_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_h2v1_merged_upsample (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) &&
+      IS_ALIGNED_SSE(jconst_merged_upsample_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_h2v2_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v2_extrgb_merged_upsample_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v2_extrgbx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v2_extbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v2_extbgrx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v2_extxbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v2_extxrgb_merged_upsample_sse2;
+      break;
+    default:
+      sse2fct=jsimd_h2v2_merged_upsample_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(void)
+jsimd_h2v1_merged_upsample (j_decompress_ptr cinfo,
+                            JSAMPIMAGE input_buf,
+                            JDIMENSION in_row_group_ctr,
+                            JSAMPARRAY output_buf)
+{
+  void (*sse2fct)(JDIMENSION, JSAMPIMAGE, JDIMENSION, JSAMPARRAY);
+
+  switch(cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      sse2fct=jsimd_h2v1_extrgb_merged_upsample_sse2;
+      break;
+    case JCS_EXT_RGBX:
+    case JCS_EXT_RGBA:
+      sse2fct=jsimd_h2v1_extrgbx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGR:
+      sse2fct=jsimd_h2v1_extbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_BGRX:
+    case JCS_EXT_BGRA:
+      sse2fct=jsimd_h2v1_extbgrx_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XBGR:
+    case JCS_EXT_ABGR:
+      sse2fct=jsimd_h2v1_extxbgr_merged_upsample_sse2;
+      break;
+    case JCS_EXT_XRGB:
+    case JCS_EXT_ARGB:
+      sse2fct=jsimd_h2v1_extxrgb_merged_upsample_sse2;
+      break;
+    default:
+      sse2fct=jsimd_h2v1_merged_upsample_sse2;
+      break;
+  }
+
+  sse2fct(cinfo->output_width, input_buf, in_row_group_ctr, output_buf);
+}
+
+GLOBAL(int)
+jsimd_can_convsamp (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_convsamp_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_convsamp (JSAMPARRAY sample_data, JDIMENSION start_col,
+                DCTELEM *workspace)
+{
+  jsimd_convsamp_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(void)
+jsimd_convsamp_float (JSAMPARRAY sample_data, JDIMENSION start_col,
+                      FAST_FLOAT *workspace)
+{
+  jsimd_convsamp_float_sse2(sample_data, start_col, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_fdct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_fdct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_fdct_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE) && IS_ALIGNED_SSE(jconst_fdct_float_sse))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_fdct_islow (DCTELEM *data)
+{
+  jsimd_fdct_islow_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_ifast (DCTELEM *data)
+{
+  jsimd_fdct_ifast_sse2(data);
+}
+
+GLOBAL(void)
+jsimd_fdct_float (FAST_FLOAT *data)
+{
+  jsimd_fdct_float_sse(data);
+}
+
+GLOBAL(int)
+jsimd_can_quantize (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(DCTELEM) != 2)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_quantize_float (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+
+  if (simd_support & JSIMD_SSE2)
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_quantize (JCOEFPTR coef_block, DCTELEM *divisors,
+                DCTELEM *workspace)
+{
+  jsimd_quantize_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(void)
+jsimd_quantize_float (JCOEFPTR coef_block, FAST_FLOAT *divisors,
+                      FAST_FLOAT *workspace)
+{
+  jsimd_quantize_float_sse2(coef_block, divisors, workspace);
+}
+
+GLOBAL(int)
+jsimd_can_idct_2x2 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_4x4 (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_red_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_2x2 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_2x2_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_4x4 (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                JDIMENSION output_col)
+{
+  jsimd_idct_4x4_sse2(compptr->dct_table, coef_block, output_buf, output_col);
+}
+
+GLOBAL(int)
+jsimd_can_idct_islow (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(ISLOW_MULT_TYPE) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_islow_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_ifast (void)
+{
+  init_simd();
+
+  /* The code is optimised for these values only */
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(IFAST_MULT_TYPE) != 2)
+    return 0;
+  if (IFAST_SCALE_BITS != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_ifast_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(int)
+jsimd_can_idct_float (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+  if (BITS_IN_JSAMPLE != 8)
+    return 0;
+  if (sizeof(JDIMENSION) != 4)
+    return 0;
+  if (sizeof(FAST_FLOAT) != 4)
+    return 0;
+  if (sizeof(FLOAT_MULT_TYPE) != 4)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && IS_ALIGNED_SSE(jconst_idct_float_sse2))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(void)
+jsimd_idct_islow (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_islow_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_ifast (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_ifast_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(void)
+jsimd_idct_float (j_decompress_ptr cinfo, jpeg_component_info *compptr,
+                  JCOEFPTR coef_block, JSAMPARRAY output_buf,
+                  JDIMENSION output_col)
+{
+  jsimd_idct_float_sse2(compptr->dct_table, coef_block, output_buf,
+                        output_col);
+}
+
+GLOBAL(int)
+jsimd_can_huff_encode_one_block (void)
+{
+  init_simd();
+
+  if (DCTSIZE != 8)
+    return 0;
+  if (sizeof(JCOEF) != 2)
+    return 0;
+
+  if ((simd_support & JSIMD_SSE2) && simd_huffman &&
+      IS_ALIGNED_SSE(jconst_huff_encode_one_block))
+    return 1;
+
+  return 0;
+}
+
+GLOBAL(JOCTET*)
+jsimd_huff_encode_one_block (void *state, JOCTET *buffer, JCOEFPTR block,
+                             int last_dc_val, c_derived_tbl *dctbl,
+                             c_derived_tbl *actbl)
+{
+  return jsimd_huff_encode_one_block_sse2(state, buffer, block, last_dc_val,
+                                          dctbl, actbl);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/Makefile.am glmark2-2021.02/src/libjpeg-turbo/simd/Makefile.am
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/Makefile.am	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/Makefile.am	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,97 @@
+noinst_LTLIBRARIES = libsimd.la
+
+BUILT_SOURCES = jsimdcfg.inc
+
+EXTRA_DIST = nasm_lt.sh CMakeLists.txt \
+	jccolext-mmx.asm   jcgryext-mmx.asm   jdcolext-mmx.asm   jdmrgext-mmx.asm \
+	jccolext-sse2.asm  jcgryext-sse2.asm  jdcolext-sse2.asm  jdmrgext-sse2.asm \
+	jccolext-sse2-64.asm  jcgryext-sse2-64.asm  jdcolext-sse2-64.asm \
+	jdmrgext-sse2-64.asm  jccolext-altivec.c    jcgryext-altivec.c \
+	jdcolext-altivec.c    jdmrgext-altivec.c
+
+if SIMD_X86_64
+
+libsimd_la_SOURCES = jsimd_x86_64.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jfdctflt-sse-64.asm \
+	jccolor-sse2-64.asm   jcgray-sse2-64.asm    jchuff-sse2-64.asm \
+	jcsample-sse2-64.asm  jdcolor-sse2-64.asm   jdmerge-sse2-64.asm \
+	jdsample-sse2-64.asm  jfdctfst-sse2-64.asm  jfdctint-sse2-64.asm \
+	jidctflt-sse2-64.asm  jidctfst-sse2-64.asm  jidctint-sse2-64.asm \
+	jidctred-sse2-64.asm  jquantf-sse2-64.asm   jquanti-sse2-64.asm
+
+jccolor-sse2-64.lo:  jccolext-sse2-64.asm
+jcgray-sse2-64.lo:   jcgryext-sse2-64.asm
+jdcolor-sse2-64.lo:  jdcolext-sse2-64.asm
+jdmerge-sse2-64.lo:  jdmrgext-sse2-64.asm
+
+endif
+
+if SIMD_I386
+
+libsimd_la_SOURCES = jsimd_i386.c jsimd.h jsimdcfg.inc.h jsimdext.inc \
+	jcolsamp.inc jdct.inc jpeg_nbits_table.inc jsimdcpu.asm \
+	jfdctflt-3dn.asm   jidctflt-3dn.asm   jquant-3dn.asm \
+	jccolor-mmx.asm    jcgray-mmx.asm     jcsample-mmx.asm \
+	jdcolor-mmx.asm    jdmerge-mmx.asm    jdsample-mmx.asm \
+	jfdctfst-mmx.asm   jfdctint-mmx.asm   jidctfst-mmx.asm \
+	jidctint-mmx.asm   jidctred-mmx.asm   jquant-mmx.asm \
+	jfdctflt-sse.asm   jidctflt-sse.asm   jquant-sse.asm \
+	jccolor-sse2.asm   jcgray-sse2.asm    jchuff-sse2.asm \
+	jcsample-sse2.asm  jdcolor-sse2.asm   jdmerge-sse2.asm \
+	jdsample-sse2.asm  jfdctfst-sse2.asm  jfdctint-sse2.asm \
+	jidctflt-sse2.asm  jidctfst-sse2.asm  jidctint-sse2.asm \
+	jidctred-sse2.asm  jquantf-sse2.asm   jquanti-sse2.asm
+
+jccolor-mmx.lo:   jccolext-mmx.asm
+jcgray.-mmx.lo:   jcgryext-mmx.asm
+jdcolor-mmx.lo:   jdcolext-mmx.asm
+jdmerge-mmx.lo:   jdmrgext-mmx.asm
+jccolor-sse2.lo:  jccolext-sse2.asm
+jcgray-sse2.lo:   jcgryext-sse2.asm
+jdcolor-sse2.lo:  jdcolext-sse2.asm
+jdmerge-sse2.lo:  jdmrgext-sse2.asm
+
+endif
+
+if SIMD_ARM
+
+libsimd_la_SOURCES = jsimd_arm.c jsimd_arm_neon.S
+
+endif
+
+if SIMD_ARM_64
+
+libsimd_la_SOURCES = jsimd_arm64.c jsimd_arm64_neon.S
+
+endif
+
+if SIMD_MIPS
+
+libsimd_la_SOURCES = jsimd_mips.c jsimd_mips_dspr2_asm.h jsimd_mips_dspr2.S
+
+endif
+
+if SIMD_POWERPC
+
+libsimd_la_SOURCES = jsimd_powerpc.c jsimd_altivec.h jcsample.h \
+	jccolor-altivec.c     jcgray-altivec.c      jcsample-altivec.c \
+	jdcolor-altivec.c     jdmerge-altivec.c     jdsample-altivec.c \
+	jfdctfst-altivec.c    jfdctint-altivec.c \
+	jidctfst-altivec.c    jidctint-altivec.c \
+	jquanti-altivec.c
+libsimd_la_CFLAGS = -maltivec
+
+jccolor-altivec.lo:  jccolext-altivec.c
+jcgray-altivec.lo:   jcgryext-altivec.c
+jdcolor-altivec.lo:  jdcolext-altivec.c
+jdmerge-altivec.lo:  jdmrgext-altivec.c
+
+endif
+
+AM_CPPFLAGS = -I$(top_srcdir)
+
+.asm.lo:
+	$(AM_V_GEN) $(LIBTOOL) $(AM_V_lt) --mode=compile --tag NASM $(srcdir)/nasm_lt.sh $(AM_V_lt) $(NASM) $(NAFLAGS) -I$(srcdir) -I. $< -o $@
+
+jsimdcfg.inc: $(srcdir)/jsimdcfg.inc.h ../jpeglib.h ../jconfig.h ../jmorecfg.h
+	$(AM_V_GEN) $(CPP) -I$(top_builddir) -I$(top_builddir)/simd $(srcdir)/jsimdcfg.inc.h | $(EGREP) "^[\;%]|^\ %" | sed 's%_cpp_protection_%%' | sed 's@% define@%define@g' > $@
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/nasm_lt.sh glmark2-2021.02/src/libjpeg-turbo/simd/nasm_lt.sh
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/simd/nasm_lt.sh	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/simd/nasm_lt.sh	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,60 @@
+#! /bin/sh
+command=""
+infile=""
+o_opt=no
+pic=no
+while [ $# -gt 0 ]; do
+    case "$1" in
+        --silent)
+            exec > /dev/null
+            ;;
+        -DPIC|-fPIC|-fpic|-Kpic|-KPIC)
+            if [ "$pic" != "yes" ] ; then
+                command="$command -DPIC"
+                pic=yes
+            fi
+            ;;
+        -f|-fbin|-faout|-faoutb|-fcoff|-felf|-felf64|-fas86| \
+        -fobj|-fwin32|-fwin64|-frdf|-fieee|-fmacho|-fmacho64)
+            # it's a file format specifier for nasm.
+            command="$command $1"
+            ;;
+        -f*)
+            # maybe a code-generation flag for gcc.
+            ;;
+        -[Ii]*)
+            incdir=`echo "$1" | sed 's/^-[Ii]//'`
+            if [ "x$incdir" = x -a "x$2" != x ] ; then
+                case "$2" in
+                    -*) ;;
+                    *) incdir="$2"; shift;;
+                esac
+            fi
+            if [ "x$incdir" != x ] ; then
+                # In the case of NASM, the trailing slash is necessary.
+                incdir=`echo "$incdir" | sed 's%/*$%/%'`
+                command="$command -I$incdir"
+            fi
+            ;;
+        -o*)
+            o_opt=yes
+            command="$command $1"
+            ;;
+        *.asm)
+            infile=$1
+            command="$command $1"
+            ;;
+        *)
+            command="$command $1"
+            ;;
+    esac
+    shift
+done
+if [ "$o_opt" != yes ] ; then
+    # By default, NASM creates an output file
+    # in the same directory as the input file.
+    outfile="-o `echo $infile | sed -e 's%^.*/%%' -e 's%\.[^.]*$%%'`.o"
+    command="$command $outfile"
+fi
+echo $command
+exec $command
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/structure.txt glmark2-2021.02/src/libjpeg-turbo/structure.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/structure.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/structure.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,906 @@
+IJG JPEG LIBRARY:  SYSTEM ARCHITECTURE
+
+This file was part of the Independent JPEG Group's software:
+Copyright (C) 1991-2012, Thomas G. Lane, Guido Vollbeding.
+It was modified by The libjpeg-turbo Project to include only information
+relevant to libjpeg-turbo.
+For conditions of distribution and use, see the accompanying README.ijg file.
+
+
+This file provides an overview of the architecture of the IJG JPEG software;
+that is, the functions of the various modules in the system and the interfaces
+between modules.  For more precise details about any data structure or calling
+convention, see the include files and comments in the source code.
+
+We assume that the reader is already somewhat familiar with the JPEG standard.
+The README.ijg file includes references for learning about JPEG.  The file
+libjpeg.txt describes the library from the viewpoint of an application
+programmer using the library; it's best to read that file before this one.
+Also, the file coderules.txt describes the coding style conventions we use.
+
+In this document, JPEG-specific terminology follows the JPEG standard:
+  A "component" means a color channel, e.g., Red or Luminance.
+  A "sample" is a single component value (i.e., one number in the image data).
+  A "coefficient" is a frequency coefficient (a DCT transform output number).
+  A "block" is an 8x8 group of samples or coefficients.
+  An "MCU" (minimum coded unit) is an interleaved set of blocks of size
+        determined by the sampling factors, or a single block in a
+        noninterleaved scan.
+We do not use the terms "pixel" and "sample" interchangeably.  When we say
+pixel, we mean an element of the full-size image, while a sample is an element
+of the downsampled image.  Thus the number of samples may vary across
+components while the number of pixels does not.  (This terminology is not used
+rigorously throughout the code, but it is used in places where confusion would
+otherwise result.)
+
+
+*** System features ***
+
+The IJG distribution contains two parts:
+  * A subroutine library for JPEG compression and decompression.
+  * cjpeg/djpeg, two sample applications that use the library to transform
+    JFIF JPEG files to and from several other image formats.
+cjpeg/djpeg are of no great intellectual complexity: they merely add a simple
+command-line user interface and I/O routines for several uncompressed image
+formats.  This document concentrates on the library itself.
+
+We desire the library to be capable of supporting all JPEG baseline, extended
+sequential, and progressive DCT processes.  Hierarchical processes are not
+supported.
+
+The library does not support the lossless (spatial) JPEG process.  Lossless
+JPEG shares little or no code with lossy JPEG, and would normally be used
+without the extensive pre- and post-processing provided by this library.
+We feel that lossless JPEG is better handled by a separate library.
+
+Within these limits, any set of compression parameters allowed by the JPEG
+spec should be readable for decompression.  (We can be more restrictive about
+what formats we can generate.)  Although the system design allows for all
+parameter values, some uncommon settings are not yet implemented and may
+never be; nonintegral sampling ratios are the prime example.  Furthermore,
+we treat 8-bit vs. 12-bit data precision as a compile-time switch, not a
+run-time option, because most machines can store 8-bit pixels much more
+compactly than 12-bit.
+
+By itself, the library handles only interchange JPEG datastreams --- in
+particular the widely used JFIF file format.  The library can be used by
+surrounding code to process interchange or abbreviated JPEG datastreams that
+are embedded in more complex file formats.  (For example, libtiff uses this
+library to implement JPEG compression within the TIFF file format.)
+
+The library includes a substantial amount of code that is not covered by the
+JPEG standard but is necessary for typical applications of JPEG.  These
+functions preprocess the image before JPEG compression or postprocess it after
+decompression.  They include colorspace conversion, downsampling/upsampling,
+and color quantization.  This code can be omitted if not needed.
+
+A wide range of quality vs. speed tradeoffs are possible in JPEG processing,
+and even more so in decompression postprocessing.  The decompression library
+provides multiple implementations that cover most of the useful tradeoffs,
+ranging from very-high-quality down to fast-preview operation.  On the
+compression side we have generally not provided low-quality choices, since
+compression is normally less time-critical.  It should be understood that the
+low-quality modes may not meet the JPEG standard's accuracy requirements;
+nonetheless, they are useful for viewers.
+
+
+*** System overview ***
+
+The compressor and decompressor are each divided into two main sections:
+the JPEG compressor or decompressor proper, and the preprocessing or
+postprocessing functions.  The interface between these two sections is the
+image data that the official JPEG spec regards as its input or output: this
+data is in the colorspace to be used for compression, and it is downsampled
+to the sampling factors to be used.  The preprocessing and postprocessing
+steps are responsible for converting a normal image representation to or from
+this form.  (Those few applications that want to deal with YCbCr downsampled
+data can skip the preprocessing or postprocessing step.)
+
+Looking more closely, the compressor library contains the following main
+elements:
+
+  Preprocessing:
+    * Color space conversion (e.g., RGB to YCbCr).
+    * Edge expansion and downsampling.  Optionally, this step can do simple
+      smoothing --- this is often helpful for low-quality source data.
+  JPEG proper:
+    * MCU assembly, DCT, quantization.
+    * Entropy coding (sequential or progressive, Huffman or arithmetic).
+
+In addition to these modules we need overall control, marker generation,
+and support code (memory management & error handling).  There is also a
+module responsible for physically writing the output data --- typically
+this is just an interface to fwrite(), but some applications may need to
+do something else with the data.
+
+The decompressor library contains the following main elements:
+
+  JPEG proper:
+    * Entropy decoding (sequential or progressive, Huffman or arithmetic).
+    * Dequantization, inverse DCT, MCU disassembly.
+  Postprocessing:
+    * Upsampling.  Optionally, this step may be able to do more general
+      rescaling of the image.
+    * Color space conversion (e.g., YCbCr to RGB).  This step may also
+      provide gamma adjustment [ currently it does not ].
+    * Optional color quantization (e.g., reduction to 256 colors).
+    * Optional color precision reduction (e.g., 24-bit to 15-bit color).
+      [This feature is not currently implemented.]
+
+We also need overall control, marker parsing, and a data source module.
+The support code (memory management & error handling) can be shared with
+the compression half of the library.
+
+There may be several implementations of each of these elements, particularly
+in the decompressor, where a wide range of speed/quality tradeoffs is very
+useful.  It must be understood that some of the best speedups involve
+merging adjacent steps in the pipeline.  For example, upsampling, color space
+conversion, and color quantization might all be done at once when using a
+low-quality ordered-dither technique.  The system architecture is designed to
+allow such merging where appropriate.
+
+
+Note: it is convenient to regard edge expansion (padding to block boundaries)
+as a preprocessing/postprocessing function, even though the JPEG spec includes
+it in compression/decompression.  We do this because downsampling/upsampling
+can be simplified a little if they work on padded data: it's not necessary to
+have special cases at the right and bottom edges.  Therefore the interface
+buffer is always an integral number of blocks wide and high, and we expect
+compression preprocessing to pad the source data properly.  Padding will occur
+only to the next block (8-sample) boundary.  In an interleaved-scan situation,
+additional dummy blocks may be used to fill out MCUs, but the MCU assembly and
+disassembly logic will create or discard these blocks internally.  (This is
+advantageous for speed reasons, since we avoid DCTing the dummy blocks.
+It also permits a small reduction in file size, because the compressor can
+choose dummy block contents so as to minimize their size in compressed form.
+Finally, it makes the interface buffer specification independent of whether
+the file is actually interleaved or not.)  Applications that wish to deal
+directly with the downsampled data must provide similar buffering and padding
+for odd-sized images.
+
+
+*** Poor man's object-oriented programming ***
+
+It should be clear by now that we have a lot of quasi-independent processing
+steps, many of which have several possible behaviors.  To avoid cluttering the
+code with lots of switch statements, we use a simple form of object-style
+programming to separate out the different possibilities.
+
+For example, two different color quantization algorithms could be implemented
+as two separate modules that present the same external interface; at runtime,
+the calling code will access the proper module indirectly through an "object".
+
+We can get the limited features we need while staying within portable C.
+The basic tool is a function pointer.  An "object" is just a struct
+containing one or more function pointer fields, each of which corresponds to
+a method name in real object-oriented languages.  During initialization we
+fill in the function pointers with references to whichever module we have
+determined we need to use in this run.  Then invocation of the module is done
+by indirecting through a function pointer; on most machines this is no more
+expensive than a switch statement, which would be the only other way of
+making the required run-time choice.  The really significant benefit, of
+course, is keeping the source code clean and well structured.
+
+We can also arrange to have private storage that varies between different
+implementations of the same kind of object.  We do this by making all the
+module-specific object structs be separately allocated entities, which will
+be accessed via pointers in the master compression or decompression struct.
+The "public" fields or methods for a given kind of object are specified by
+a commonly known struct.  But a module's initialization code can allocate
+a larger struct that contains the common struct as its first member, plus
+additional private fields.  With appropriate pointer casting, the module's
+internal functions can access these private fields.  (For a simple example,
+see jdatadst.c, which implements the external interface specified by struct
+jpeg_destination_mgr, but adds extra fields.)
+
+(Of course this would all be a lot easier if we were using C++, but we are
+not yet prepared to assume that everyone has a C++ compiler.)
+
+An important benefit of this scheme is that it is easy to provide multiple
+versions of any method, each tuned to a particular case.  While a lot of
+precalculation might be done to select an optimal implementation of a method,
+the cost per invocation is constant.  For example, the upsampling step might
+have a "generic" method, plus one or more "hardwired" methods for the most
+popular sampling factors; the hardwired methods would be faster because they'd
+use straight-line code instead of for-loops.  The cost to determine which
+method to use is paid only once, at startup, and the selection criteria are
+hidden from the callers of the method.
+
+This plan differs a little bit from usual object-oriented structures, in that
+only one instance of each object class will exist during execution.  The
+reason for having the class structure is that on different runs we may create
+different instances (choose to execute different modules).  You can think of
+the term "method" as denoting the common interface presented by a particular
+set of interchangeable functions, and "object" as denoting a group of related
+methods, or the total shared interface behavior of a group of modules.
+
+
+*** Overall control structure ***
+
+We previously mentioned the need for overall control logic in the compression
+and decompression libraries.  In IJG implementations prior to v5, overall
+control was mostly provided by "pipeline control" modules, which proved to be
+large, unwieldy, and hard to understand.  To improve the situation, the
+control logic has been subdivided into multiple modules.  The control modules
+consist of:
+
+1. Master control for module selection and initialization.  This has two
+responsibilities:
+
+   1A.  Startup initialization at the beginning of image processing.
+        The individual processing modules to be used in this run are selected
+        and given initialization calls.
+
+   1B.  Per-pass control.  This determines how many passes will be performed
+        and calls each active processing module to configure itself
+        appropriately at the beginning of each pass.  End-of-pass processing,
+        where necessary, is also invoked from the master control module.
+
+   Method selection is partially distributed, in that a particular processing
+   module may contain several possible implementations of a particular method,
+   which it will select among when given its initialization call.  The master
+   control code need only be concerned with decisions that affect more than
+   one module.
+
+2. Data buffering control.  A separate control module exists for each
+   inter-processing-step data buffer.  This module is responsible for
+   invoking the processing steps that write or read that data buffer.
+
+Each buffer controller sees the world as follows:
+
+input data => processing step A => buffer => processing step B => output data
+                      |              |               |
+              ------------------ controller ------------------
+
+The controller knows the dataflow requirements of steps A and B: how much data
+they want to accept in one chunk and how much they output in one chunk.  Its
+function is to manage its buffer and call A and B at the proper times.
+
+A data buffer control module may itself be viewed as a processing step by a
+higher-level control module; thus the control modules form a binary tree with
+elementary processing steps at the leaves of the tree.
+
+The control modules are objects.  A considerable amount of flexibility can
+be had by replacing implementations of a control module.  For example:
+* Merging of adjacent steps in the pipeline is done by replacing a control
+  module and its pair of processing-step modules with a single processing-
+  step module.  (Hence the possible merges are determined by the tree of
+  control modules.)
+* In some processing modes, a given interstep buffer need only be a "strip"
+  buffer large enough to accommodate the desired data chunk sizes.  In other
+  modes, a full-image buffer is needed and several passes are required.
+  The control module determines which kind of buffer is used and manipulates
+  virtual array buffers as needed.  One or both processing steps may be
+  unaware of the multi-pass behavior.
+
+In theory, we might be able to make all of the data buffer controllers
+interchangeable and provide just one set of implementations for all.  In
+practice, each one contains considerable special-case processing for its
+particular job.  The buffer controller concept should be regarded as an
+overall system structuring principle, not as a complete description of the
+task performed by any one controller.
+
+
+*** Compression object structure ***
+
+Here is a sketch of the logical structure of the JPEG compression library:
+
+                                                 |-- Colorspace conversion
+                  |-- Preprocessing controller --|
+                  |                              |-- Downsampling
+Main controller --|
+                  |                            |-- Forward DCT, quantize
+                  |-- Coefficient controller --|
+                                               |-- Entropy encoding
+
+This sketch also describes the flow of control (subroutine calls) during
+typical image data processing.  Each of the components shown in the diagram is
+an "object" which may have several different implementations available.  One
+or more source code files contain the actual implementation(s) of each object.
+
+The objects shown above are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the preprocessed input data.  This controller invokes preprocessing to
+  fill the subsampled-data buffer, and JPEG compression to empty it.  There is
+  usually no need for a full-image buffer here; a strip buffer is adequate.
+
+* Preprocessing controller: buffer controller for the downsampling input data
+  buffer, which lies between colorspace conversion and downsampling.  Note
+  that a unified conversion/downsampling module would probably replace this
+  controller entirely.
+
+* Colorspace conversion: converts application image data into the desired
+  JPEG color space; also changes the data from pixel-interleaved layout to
+  separate component planes.  Processes one pixel row at a time.
+
+* Downsampling: performs reduction of chroma components as required.
+  Optionally may perform pixel-level smoothing as well.  Processes a "row
+  group" at a time, where a row group is defined as Vmax pixel rows of each
+  component before downsampling, and Vk sample rows afterwards (remember Vk
+  differs across components).  Some downsampling or smoothing algorithms may
+  require context rows above and below the current row group; the
+  preprocessing controller is responsible for supplying these rows via proper
+  buffering.  The downsampler is responsible for edge expansion at the right
+  edge (i.e., extending each sample row to a multiple of 8 samples); but the
+  preprocessing controller is responsible for vertical edge expansion (i.e.,
+  duplicating the bottom sample row as needed to make a multiple of 8 rows).
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU assembly, including insertion of dummy DCT
+  blocks when needed at the right or bottom edge.  When performing
+  Huffman-code optimization or emitting a multiscan JPEG file, this
+  controller is responsible for buffering the full image.  The equivalent of
+  one fully interleaved MCU row of subsampled data is processed per call,
+  even when the JPEG file is noninterleaved.
+
+* Forward DCT and quantization: Perform DCT, quantize, and emit coefficients.
+  Works on one or more DCT blocks at a time.  (Note: the coefficients are now
+  emitted in normal array order, which the entropy encoder is expected to
+  convert to zigzag order as necessary.  Prior versions of the IJG code did
+  the conversion to zigzag order within the quantization step.)
+
+* Entropy encoding: Perform Huffman or arithmetic entropy coding and emit the
+  coded data to the data destination module.  Works on one MCU per call.
+  For progressive JPEG, the same DCT blocks are fed to the entropy coder
+  during each pass, and the coder must emit the appropriate subset of
+  coefficients.
+
+In addition to the above objects, the compression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.
+
+* Marker writing: generates JPEG markers (except for RSTn, which is emitted
+  by the entropy encoder when needed).
+
+* Data destination manager: writes the output JPEG datastream to its final
+  destination (e.g., a file).  The destination manager supplied with the
+  library knows how to write to a stdio stream or to a memory buffer;
+  for other behaviors, the surrounding application may provide its own
+  destination manager.
+
+* Memory manager: allocates and releases memory, controls virtual arrays
+  (with backing store management, where required).
+
+* Error handler: performs formatting and output of error and trace messages;
+  determines handling of nonfatal errors.  The surrounding application may
+  override some or all of this object's methods to change error handling.
+
+* Progress monitor: supports output of "percent-done" progress reports.
+  This object represents an optional callback to the surrounding application:
+  if wanted, it must be supplied by the application.
+
+The error handler, destination manager, and progress monitor objects are
+defined as separate objects in order to simplify application-specific
+customization of the JPEG library.  A surrounding application may override
+individual methods or supply its own all-new implementation of one of these
+objects.  The object interfaces for these objects are therefore treated as
+part of the application interface of the library, whereas the other objects
+are internal to the library.
+
+The error handler and memory manager are shared by JPEG compression and
+decompression; the progress monitor, if used, may be shared as well.
+
+
+*** Decompression object structure ***
+
+Here is a sketch of the logical structure of the JPEG decompression library:
+
+                                               |-- Entropy decoding
+                  |-- Coefficient controller --|
+                  |                            |-- Dequantize, Inverse DCT
+Main controller --|
+                  |                               |-- Upsampling
+                  |-- Postprocessing controller --|   |-- Colorspace conversion
+                                                  |-- Color quantization
+                                                  |-- Color precision reduction
+
+As before, this diagram also represents typical control flow.  The objects
+shown are:
+
+* Main controller: buffer controller for the subsampled-data buffer, which
+  holds the output of JPEG decompression proper.  This controller's primary
+  task is to feed the postprocessing procedure.  Some upsampling algorithms
+  may require context rows above and below the current row group; when this
+  is true, the main controller is responsible for managing its buffer so as
+  to make context rows available.  In the current design, the main buffer is
+  always a strip buffer; a full-image buffer is never required.
+
+* Coefficient controller: buffer controller for the DCT-coefficient data.
+  This controller handles MCU disassembly, including deletion of any dummy
+  DCT blocks at the right or bottom edge.  When reading a multiscan JPEG
+  file, this controller is responsible for buffering the full image.
+  (Buffering DCT coefficients, rather than samples, is necessary to support
+  progressive JPEG.)  The equivalent of one fully interleaved MCU row of
+  subsampled data is processed per call, even when the source JPEG file is
+  noninterleaved.
+
+* Entropy decoding: Read coded data from the data source module and perform
+  Huffman or arithmetic entropy decoding.  Works on one MCU per call.
+  For progressive JPEG decoding, the coefficient controller supplies the prior
+  coefficients of each MCU (initially all zeroes), which the entropy decoder
+  modifies in each scan.
+
+* Dequantization and inverse DCT: like it says.  Note that the coefficients
+  buffered by the coefficient controller have NOT been dequantized; we
+  merge dequantization and inverse DCT into a single step for speed reasons.
+  When scaled-down output is asked for, simplified DCT algorithms may be used
+  that emit fewer samples per DCT block, not the full 8x8.  Works on one DCT
+  block at a time.
+
+* Postprocessing controller: buffer controller for the color quantization
+  input buffer, when quantization is in use.  (Without quantization, this
+  controller just calls the upsampler.)  For two-pass quantization, this
+  controller is responsible for buffering the full-image data.
+
+* Upsampling: restores chroma components to full size.  (May support more
+  general output rescaling, too.  Note that if undersized DCT outputs have
+  been emitted by the DCT module, this module must adjust so that properly
+  sized outputs are created.)  Works on one row group at a time.  This module
+  also calls the color conversion module, so its top level is effectively a
+  buffer controller for the upsampling->color conversion buffer.  However, in
+  all but the highest-quality operating modes, upsampling and color
+  conversion are likely to be merged into a single step.
+
+* Colorspace conversion: convert from JPEG color space to output color space,
+  and change data layout from separate component planes to pixel-interleaved.
+  Works on one pixel row at a time.
+
+* Color quantization: reduce the data to colormapped form, using either an
+  externally specified colormap or an internally generated one.  This module
+  is not used for full-color output.  Works on one pixel row at a time; may
+  require two passes to generate a color map.  Note that the output will
+  always be a single component representing colormap indexes.  In the current
+  design, the output values are JSAMPLEs, so an 8-bit compilation cannot
+  quantize to more than 256 colors.  This is unlikely to be a problem in
+  practice.
+
+* Color reduction: this module handles color precision reduction, e.g.,
+  generating 15-bit color (5 bits/primary) from JPEG's 24-bit output.
+  Not quite clear yet how this should be handled... should we merge it with
+  colorspace conversion???
+
+Note that some high-speed operating modes might condense the entire
+postprocessing sequence to a single module (upsample, color convert, and
+quantize in one step).
+
+In addition to the above objects, the decompression library includes these
+objects:
+
+* Master control: determines the number of passes required, controls overall
+  and per-pass initialization of the other modules.  This is subdivided into
+  input and output control: jdinput.c controls only input-side processing,
+  while jdmaster.c handles overall initialization and output-side control.
+
+* Marker reading: decodes JPEG markers (except for RSTn).
+
+* Data source manager: supplies the input JPEG datastream.  The source
+  manager supplied with the library knows how to read from a stdio stream
+  or from a memory buffer;  for other behaviors, the surrounding application
+  may provide its own source manager.
+
+* Memory manager: same as for compression library.
+
+* Error handler: same as for compression library.
+
+* Progress monitor: same as for compression library.
+
+As with compression, the data source manager, error handler, and progress
+monitor are candidates for replacement by a surrounding application.
+
+
+*** Decompression input and output separation ***
+
+To support efficient incremental display of progressive JPEG files, the
+decompressor is divided into two sections that can run independently:
+
+1. Data input includes marker parsing, entropy decoding, and input into the
+   coefficient controller's DCT coefficient buffer.  Note that this
+   processing is relatively cheap and fast.
+
+2. Data output reads from the DCT coefficient buffer and performs the IDCT
+   and all postprocessing steps.
+
+For a progressive JPEG file, the data input processing is allowed to get
+arbitrarily far ahead of the data output processing.  (This occurs only
+if the application calls jpeg_consume_input(); otherwise input and output
+run in lockstep, since the input section is called only when the output
+section needs more data.)  In this way the application can avoid making
+extra display passes when data is arriving faster than the display pass
+can run.  Furthermore, it is possible to abort an output pass without
+losing anything, since the coefficient buffer is read-only as far as the
+output section is concerned.  See libjpeg.txt for more detail.
+
+A full-image coefficient array is only created if the JPEG file has multiple
+scans (or if the application specifies buffered-image mode anyway).  When
+reading a single-scan file, the coefficient controller normally creates only
+a one-MCU buffer, so input and output processing must run in lockstep in this
+case.  jpeg_consume_input() is effectively a no-op in this situation.
+
+The main impact of dividing the decompressor in this fashion is that we must
+be very careful with shared variables in the cinfo data structure.  Each
+variable that can change during the course of decompression must be
+classified as belonging to data input or data output, and each section must
+look only at its own variables.  For example, the data output section may not
+depend on any of the variables that describe the current scan in the JPEG
+file, because these may change as the data input section advances into a new
+scan.
+
+The progress monitor is (somewhat arbitrarily) defined to treat input of the
+file as one pass when buffered-image mode is not used, and to ignore data
+input work completely when buffered-image mode is used.  Note that the
+library has no reliable way to predict the number of passes when dealing
+with a progressive JPEG file, nor can it predict the number of output passes
+in buffered-image mode.  So the work estimate is inherently bogus anyway.
+
+No comparable division is currently made in the compression library, because
+there isn't any real need for it.
+
+
+*** Data formats ***
+
+Arrays of pixel sample values use the following data structure:
+
+    typedef something JSAMPLE;          a pixel component value, 0..MAXJSAMPLE
+    typedef JSAMPLE *JSAMPROW;          ptr to a row of samples
+    typedef JSAMPROW *JSAMPARRAY;       ptr to a list of rows
+    typedef JSAMPARRAY *JSAMPIMAGE;     ptr to a list of color-component arrays
+
+The basic element type JSAMPLE will typically be one of unsigned char,
+(signed) char, or short.  Short will be used if samples wider than 8 bits are
+to be supported (this is a compile-time option).  Otherwise, unsigned char is
+used if possible.  If the compiler only supports signed chars, then it is
+necessary to mask off the value when reading.  Thus, all reads of JSAMPLE
+values must be coded as "GETJSAMPLE(value)", where the macro will be defined
+as "((value) & 0xFF)" on signed-char machines and "((int) (value))" elsewhere.
+
+With these conventions, JSAMPLE values can be assumed to be >= 0.  This helps
+simplify correct rounding during downsampling, etc.  The JPEG standard's
+specification that sample values run from -128..127 is accommodated by
+subtracting 128 from the sample value in the DCT step.  Similarly, during
+decompression the output of the IDCT step will be immediately shifted back to
+0..255.  (NB: different values are required when 12-bit samples are in use.
+The code is written in terms of MAXJSAMPLE and CENTERJSAMPLE, which will be
+defined as 255 and 128 respectively in an 8-bit implementation, and as 4095
+and 2048 in a 12-bit implementation.)
+
+We use a pointer per row, rather than a two-dimensional JSAMPLE array.  This
+choice costs only a small amount of memory and has several benefits:
+* Code using the data structure doesn't need to know the allocated width of
+  the rows.  This simplifies edge expansion/compression, since we can work
+  in an array that's wider than the logical picture width.
+* Indexing doesn't require multiplication; this is a performance win on many
+  machines.
+* Arrays with more than 64K total elements can be supported even on machines
+  where malloc() cannot allocate chunks larger than 64K.
+* The rows forming a component array may be allocated at different times
+  without extra copying.  This trick allows some speedups in smoothing steps
+  that need access to the previous and next rows.
+
+Note that each color component is stored in a separate array; we don't use the
+traditional layout in which the components of a pixel are stored together.
+This simplifies coding of modules that work on each component independently,
+because they don't need to know how many components there are.  Furthermore,
+we can read or write each component to a temporary file independently, which
+is helpful when dealing with noninterleaved JPEG files.
+
+In general, a specific sample value is accessed by code such as
+        GETJSAMPLE(image[colorcomponent][row][col])
+where col is measured from the image left edge, but row is measured from the
+first sample row currently in memory.  Either of the first two indexings can
+be precomputed by copying the relevant pointer.
+
+
+Since most image-processing applications prefer to work on images in which
+the components of a pixel are stored together, the data passed to or from the
+surrounding application uses the traditional convention: a single pixel is
+represented by N consecutive JSAMPLE values, and an image row is an array of
+(# of color components)*(image width) JSAMPLEs.  One or more rows of data can
+be represented by a pointer of type JSAMPARRAY in this scheme.  This scheme is
+converted to component-wise storage inside the JPEG library.  (Applications
+that want to skip JPEG preprocessing or postprocessing will have to contend
+with component-wise storage.)
+
+
+Arrays of DCT-coefficient values use the following data structure:
+
+    typedef short JCOEF;                a 16-bit signed integer
+    typedef JCOEF JBLOCK[DCTSIZE2];     an 8x8 block of coefficients
+    typedef JBLOCK *JBLOCKROW;          ptr to one horizontal row of 8x8 blocks
+    typedef JBLOCKROW *JBLOCKARRAY;     ptr to a list of such rows
+    typedef JBLOCKARRAY *JBLOCKIMAGE;   ptr to a list of color component arrays
+
+The underlying type is at least a 16-bit signed integer; while "short" is big
+enough on all machines of interest, on some machines it is preferable to use
+"int" for speed reasons, despite the storage cost.  Coefficients are grouped
+into 8x8 blocks (but we always use #defines DCTSIZE and DCTSIZE2 rather than
+"8" and "64").
+
+The contents of a coefficient block may be in either "natural" or zigzagged
+order, and may be true values or divided by the quantization coefficients,
+depending on where the block is in the processing pipeline.  In the current
+library, coefficient blocks are kept in natural order everywhere; the entropy
+codecs zigzag or dezigzag the data as it is written or read.  The blocks
+contain quantized coefficients everywhere outside the DCT/IDCT subsystems.
+(This latter decision may need to be revisited to support variable
+quantization a la JPEG Part 3.)
+
+Notice that the allocation unit is now a row of 8x8 blocks, corresponding to
+eight rows of samples.  Otherwise the structure is much the same as for
+samples, and for the same reasons.
+
+
+*** Suspendable processing ***
+
+In some applications it is desirable to use the JPEG library as an
+incremental, memory-to-memory filter.  In this situation the data source or
+destination may be a limited-size buffer, and we can't rely on being able to
+empty or refill the buffer at arbitrary times.  Instead the application would
+like to have control return from the library at buffer overflow/underrun, and
+then resume compression or decompression at a later time.
+
+This scenario is supported for simple cases.  (For anything more complex, we
+recommend that the application "bite the bullet" and develop real multitasking
+capability.)  The libjpeg.txt file goes into more detail about the usage and
+limitations of this capability; here we address the implications for library
+structure.
+
+The essence of the problem is that the entropy codec (coder or decoder) must
+be prepared to stop at arbitrary times.  In turn, the controllers that call
+the entropy codec must be able to stop before having produced or consumed all
+the data that they normally would handle in one call.  That part is reasonably
+straightforward: we make the controller call interfaces include "progress
+counters" which indicate the number of data chunks successfully processed, and
+we require callers to test the counter rather than just assume all of the data
+was processed.
+
+Rather than trying to restart at an arbitrary point, the current Huffman
+codecs are designed to restart at the beginning of the current MCU after a
+suspension due to buffer overflow/underrun.  At the start of each call, the
+codec's internal state is loaded from permanent storage (in the JPEG object
+structures) into local variables.  On successful completion of the MCU, the
+permanent state is updated.  (This copying is not very expensive, and may even
+lead to *improved* performance if the local variables can be registerized.)
+If a suspension occurs, the codec simply returns without updating the state,
+thus effectively reverting to the start of the MCU.  Note that this implies
+leaving some data unprocessed in the source/destination buffer (ie, the
+compressed partial MCU).  The data source/destination module interfaces are
+specified so as to make this possible.  This also implies that the data buffer
+must be large enough to hold a worst-case compressed MCU; a couple thousand
+bytes should be enough.
+
+In a successive-approximation AC refinement scan, the progressive Huffman
+decoder has to be able to undo assignments of newly nonzero coefficients if it
+suspends before the MCU is complete, since decoding requires distinguishing
+previously-zero and previously-nonzero coefficients.  This is a bit tedious
+but probably won't have much effect on performance.  Other variants of Huffman
+decoding need not worry about this, since they will just store the same values
+again if forced to repeat the MCU.
+
+This approach would probably not work for an arithmetic codec, since its
+modifiable state is quite large and couldn't be copied cheaply.  Instead it
+would have to suspend and resume exactly at the point of the buffer end.
+
+The JPEG marker reader is designed to cope with suspension at an arbitrary
+point.  It does so by backing up to the start of the marker parameter segment,
+so the data buffer must be big enough to hold the largest marker of interest.
+Again, a couple KB should be adequate.  (A special "skip" convention is used
+to bypass COM and APPn markers, so these can be larger than the buffer size
+without causing problems; otherwise a 64K buffer would be needed in the worst
+case.)
+
+The JPEG marker writer currently does *not* cope with suspension.
+We feel that this is not necessary; it is much easier simply to require
+the application to ensure there is enough buffer space before starting.  (An
+empty 2K buffer is more than sufficient for the header markers; and ensuring
+there are a dozen or two bytes available before calling jpeg_finish_compress()
+will suffice for the trailer.)  This would not work for writing multi-scan
+JPEG files, but we simply do not intend to support that capability with
+suspension.
+
+
+*** Memory manager services ***
+
+The JPEG library's memory manager controls allocation and deallocation of
+memory, and it manages large "virtual" data arrays on machines where the
+operating system does not provide virtual memory.  Note that the same
+memory manager serves both compression and decompression operations.
+
+In all cases, allocated objects are tied to a particular compression or
+decompression master record, and they will be released when that master
+record is destroyed.
+
+The memory manager does not provide explicit deallocation of objects.
+Instead, objects are created in "pools" of free storage, and a whole pool
+can be freed at once.  This approach helps prevent storage-leak bugs, and
+it speeds up operations whenever malloc/free are slow (as they often are).
+The pools can be regarded as lifetime identifiers for objects.  Two
+pools/lifetimes are defined:
+  * JPOOL_PERMANENT     lasts until master record is destroyed
+  * JPOOL_IMAGE         lasts until done with image (JPEG datastream)
+Permanent lifetime is used for parameters and tables that should be carried
+across from one datastream to another; this includes all application-visible
+parameters.  Image lifetime is used for everything else.  (A third lifetime,
+JPOOL_PASS = one processing pass, was originally planned.  However it was
+dropped as not being worthwhile.  The actual usage patterns are such that the
+peak memory usage would be about the same anyway; and having per-pass storage
+substantially complicates the virtual memory allocation rules --- see below.)
+
+The memory manager deals with three kinds of object:
+1. "Small" objects.  Typically these require no more than 10K-20K total.
+2. "Large" objects.  These may require tens to hundreds of K depending on
+   image size.  Semantically they behave the same as small objects, but we
+   distinguish them because pool allocation heuristics may differ for large and
+   small objects (historically, large objects were also referenced by far
+   pointers on MS-DOS machines.)  Note that individual "large" objects cannot
+   exceed the size allowed by type size_t, which may be 64K or less on some
+   machines.
+3. "Virtual" objects.  These are large 2-D arrays of JSAMPLEs or JBLOCKs
+   (typically large enough for the entire image being processed).  The
+   memory manager provides stripwise access to these arrays.  On machines
+   without virtual memory, the rest of the array may be swapped out to a
+   temporary file.
+
+(Note: JSAMPARRAY and JBLOCKARRAY data structures are a combination of large
+objects for the data proper and small objects for the row pointers.  For
+convenience and speed, the memory manager provides single routines to create
+these structures.  Similarly, virtual arrays include a small control block
+and a JSAMPARRAY or JBLOCKARRAY working buffer, all created with one call.)
+
+In the present implementation, virtual arrays are only permitted to have image
+lifespan.  (Permanent lifespan would not be reasonable, and pass lifespan is
+not very useful since a virtual array's raison d'etre is to store data for
+multiple passes through the image.)  We also expect that only "small" objects
+will be given permanent lifespan, though this restriction is not required by
+the memory manager.
+
+In a non-virtual-memory machine, some performance benefit can be gained by
+making the in-memory buffers for virtual arrays be as large as possible.
+(For small images, the buffers might fit entirely in memory, so blind
+swapping would be very wasteful.)  The memory manager will adjust the height
+of the buffers to fit within a prespecified maximum memory usage.  In order
+to do this in a reasonably optimal fashion, the manager needs to allocate all
+of the virtual arrays at once.  Therefore, there isn't a one-step allocation
+routine for virtual arrays; instead, there is a "request" routine that simply
+allocates the control block, and a "realize" routine (called just once) that
+determines space allocation and creates all of the actual buffers.  The
+realize routine must allow for space occupied by non-virtual large objects.
+(We don't bother to factor in the space needed for small objects, on the
+grounds that it isn't worth the trouble.)
+
+To support all this, we establish the following protocol for doing business
+with the memory manager:
+  1. Modules must request virtual arrays (which may have only image lifespan)
+     during the initial setup phase, i.e., in their jinit_xxx routines.
+  2. All "large" objects (including JSAMPARRAYs and JBLOCKARRAYs) must also be
+     allocated during initial setup.
+  3. realize_virt_arrays will be called at the completion of initial setup.
+     The above conventions ensure that sufficient information is available
+     for it to choose a good size for virtual array buffers.
+Small objects of any lifespan may be allocated at any time.  We expect that
+the total space used for small objects will be small enough to be negligible
+in the realize_virt_arrays computation.
+
+In a virtual-memory machine, we simply pretend that the available space is
+infinite, thus causing realize_virt_arrays to decide that it can allocate all
+the virtual arrays as full-size in-memory buffers.  The overhead of the
+virtual-array access protocol is very small when no swapping occurs.
+
+A virtual array can be specified to be "pre-zeroed"; when this flag is set,
+never-yet-written sections of the array are set to zero before being made
+available to the caller.  If this flag is not set, never-written sections
+of the array contain garbage.  (This feature exists primarily because the
+equivalent logic would otherwise be needed in jdcoefct.c for progressive
+JPEG mode; we may as well make it available for possible other uses.)
+
+The first write pass on a virtual array is required to occur in top-to-bottom
+order; read passes, as well as any write passes after the first one, may
+access the array in any order.  This restriction exists partly to simplify
+the virtual array control logic, and partly because some file systems may not
+support seeking beyond the current end-of-file in a temporary file.  The main
+implication of this restriction is that rearrangement of rows (such as
+converting top-to-bottom data order to bottom-to-top) must be handled while
+reading data out of the virtual array, not while putting it in.
+
+
+*** Memory manager internal structure ***
+
+To isolate system dependencies as much as possible, we have broken the
+memory manager into two parts.  There is a reasonably system-independent
+"front end" (jmemmgr.c) and a "back end" that contains only the code
+likely to change across systems.  All of the memory management methods
+outlined above are implemented by the front end.  The back end provides
+the following routines for use by the front end (none of these routines
+are known to the rest of the JPEG code):
+
+jpeg_mem_init, jpeg_mem_term    system-dependent initialization/shutdown
+
+jpeg_get_small, jpeg_free_small interface to malloc and free library routines
+                                (or their equivalents)
+
+jpeg_get_large, jpeg_free_large historically was used to interface with
+                                FAR malloc/free on MS-DOS machines;  now the
+                                same as jpeg_get_small/jpeg_free_small
+
+jpeg_mem_available              estimate available memory
+
+jpeg_open_backing_store         create a backing-store object
+
+read_backing_store,             manipulate a backing-store object
+write_backing_store,
+close_backing_store
+
+On some systems there will be more than one type of backing-store object
+(specifically, in MS-DOS a backing store file might be an area of extended
+memory as well as a disk file).  jpeg_open_backing_store is responsible for
+choosing how to implement a given object.  The read/write/close routines
+are method pointers in the structure that describes a given object; this
+lets them be different for different object types.
+
+It may be necessary to ensure that backing store objects are explicitly
+released upon abnormal program termination.  For example, MS-DOS won't free
+extended memory by itself.  To support this, we will expect the main program
+or surrounding application to arrange to call self_destruct (typically via
+jpeg_destroy) upon abnormal termination.  This may require a SIGINT signal
+handler or equivalent.  We don't want to have the back end module install its
+own signal handler, because that would pre-empt the surrounding application's
+ability to control signal handling.
+
+The IJG distribution includes several memory manager back end implementations.
+Usually the same back end should be suitable for all applications on a given
+system, but it is possible for an application to supply its own back end at
+need.
+
+
+*** Implications of DNL marker ***
+
+Some JPEG files may use a DNL marker to postpone definition of the image
+height (this would be useful for a fax-like scanner's output, for instance).
+In these files the SOF marker claims the image height is 0, and you only
+find out the true image height at the end of the first scan.
+
+We could read these files as follows:
+1. Upon seeing zero image height, replace it by 65535 (the maximum allowed).
+2. When the DNL is found, update the image height in the global image
+   descriptor.
+This implies that control modules must avoid making copies of the image
+height, and must re-test for termination after each MCU row.  This would
+be easy enough to do.
+
+In cases where image-size data structures are allocated, this approach will
+result in very inefficient use of virtual memory or much-larger-than-necessary
+temporary files.  This seems acceptable for something that probably won't be a
+mainstream usage.  People might have to forgo use of memory-hogging options
+(such as two-pass color quantization or noninterleaved JPEG files) if they
+want efficient conversion of such files.  (One could improve efficiency by
+demanding a user-supplied upper bound for the height, less than 65536; in most
+cases it could be much less.)
+
+The standard also permits the SOF marker to overestimate the image height,
+with a DNL to give the true, smaller height at the end of the first scan.
+This would solve the space problems if the overestimate wasn't too great.
+However, it implies that you don't even know whether DNL will be used.
+
+This leads to a couple of very serious objections:
+1. Testing for a DNL marker must occur in the inner loop of the decompressor's
+   Huffman decoder; this implies a speed penalty whether the feature is used
+   or not.
+2. There is no way to hide the last-minute change in image height from an
+   application using the decoder.  Thus *every* application using the IJG
+   library would suffer a complexity penalty whether it cared about DNL or
+   not.
+We currently do not support DNL because of these problems.
+
+A different approach is to insist that DNL-using files be preprocessed by a
+separate program that reads ahead to the DNL, then goes back and fixes the SOF
+marker.  This is a much simpler solution and is probably far more efficient.
+Even if one wants piped input, buffering the first scan of the JPEG file needs
+a lot smaller temp file than is implied by the maximum-height method.  For
+this approach we'd simply treat DNL as a no-op in the decompressor (at most,
+check that it matches the SOF image height).
+
+We will not worry about making the compressor capable of outputting DNL.
+Something similar to the first scheme above could be applied if anyone ever
+wants to make that work.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbench.c glmark2-2021.02/src/libjpeg-turbo/tjbench.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbench.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjbench.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,990 @@
+/*
+ * Copyright (C)2009-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+#include <math.h>
+#include <errno.h>
+#include <cdjpeg.h>
+#include "./bmp.h"
+#include "./tjutil.h"
+#include "./turbojpeg.h"
+
+
+#define _throw(op, err) {  \
+	printf("ERROR in line %d while %s:\n%s\n", __LINE__, op, err);  \
+  retval=-1;  goto bailout;}
+#define _throwunix(m) _throw(m, strerror(errno))
+#define _throwtj(m) _throw(m, tjGetErrorStr())
+#define _throwbmp(m) _throw(m, bmpgeterr())
+
+int flags=TJFLAG_NOREALLOC, componly=0, decomponly=0, doyuv=0, quiet=0,
+	dotile=0, pf=TJPF_BGR, yuvpad=1, warmup=1, dowrite=1;
+char *ext="ppm";
+const char *pixFormatStr[TJ_NUMPF]=
+{
+	"RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "GRAY", "", "", "", "", "CMYK"
+};
+const char *subNameLong[TJ_NUMSAMP]=
+{
+	"4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *csName[TJ_NUMCS]=
+{
+	"RGB", "YCbCr", "GRAY", "CMYK", "YCCK"
+};
+const char *subName[TJ_NUMSAMP]={"444", "422", "420", "GRAY", "440", "411"};
+tjscalingfactor *scalingfactors=NULL, sf={1, 1};  int nsf=0;
+int xformop=TJXOP_NONE, xformopt=0;
+int (*customFilter)(short *, tjregion, tjregion, int, int, tjtransform *);
+double benchtime=5.0;
+
+
+char *formatName(int subsamp, int cs, char *buf)
+{
+	if(cs==TJCS_YCbCr) return (char *)subNameLong[subsamp];
+	else if(cs==TJCS_YCCK)
+	{
+		snprintf(buf, 80, "%s %s", csName[cs], subNameLong[subsamp]);
+		return buf;
+	}
+	else return (char *)csName[cs];
+}
+
+
+char *sigfig(double val, int figs, char *buf, int len)
+{
+	char format[80];
+	int digitsafterdecimal=figs-(int)ceil(log10(fabs(val)));
+	if(digitsafterdecimal<1) snprintf(format, 80, "%%.0f");
+	else snprintf(format, 80, "%%.%df", digitsafterdecimal);
+	snprintf(buf, len, format, val);
+	return buf;
+}
+
+
+/* Custom DCT filter which produces a negative of the image */
+int dummyDCTFilter(short *coeffs, tjregion arrayRegion, tjregion planeRegion,
+	int componentIndex, int transformIndex, tjtransform *transform)
+{
+	int i;
+	for(i=0; i<arrayRegion.w*arrayRegion.h; i++) coeffs[i]=-coeffs[i];
+	return 0;
+}
+
+
+/* Decompression test */
+int decomp(unsigned char *srcbuf, unsigned char **jpegbuf,
+	unsigned long *jpegsize, unsigned char *dstbuf, int w, int h,
+	int subsamp, int jpegqual, char *filename, int tilew, int tileh)
+{
+	char tempstr[1024], sizestr[20]="\0", qualstr[6]="\0", *ptr;
+	FILE *file=NULL;  tjhandle handle=NULL;
+	int row, col, iter=0, dstbufalloc=0, retval=0;
+	double elapsed, elapsedDecode;
+	int ps=tjPixelSize[pf];
+	int scaledw=TJSCALED(w, sf);
+	int scaledh=TJSCALED(h, sf);
+	int pitch=scaledw*ps;
+	int ntilesw=(w+tilew-1)/tilew, ntilesh=(h+tileh-1)/tileh;
+	unsigned char *dstptr, *dstptr2, *yuvbuf=NULL;
+
+	if(jpegqual>0)
+	{
+		snprintf(qualstr, 6, "_Q%d", jpegqual);
+		qualstr[5]=0;
+	}
+
+	if((handle=tjInitDecompress())==NULL)
+		_throwtj("executing tjInitDecompress()");
+
+	if(dstbuf==NULL)
+	{
+		if((dstbuf=(unsigned char *)malloc(pitch*scaledh))==NULL)
+			_throwunix("allocating destination buffer");
+		dstbufalloc=1;
+	}
+	/* Set the destination buffer to gray so we know whether the decompressor
+	   attempted to write to it */
+	memset(dstbuf, 127, pitch*scaledh);
+
+	if(doyuv)
+	{
+		int width=dotile? tilew:scaledw;
+		int height=dotile? tileh:scaledh;
+		int yuvsize=tjBufSizeYUV2(width, yuvpad, height, subsamp);
+		if((yuvbuf=(unsigned char *)malloc(yuvsize))==NULL)
+			_throwunix("allocating YUV buffer");
+		memset(yuvbuf, 127, yuvsize);
+	}
+
+	/* Benchmark */
+	iter=-warmup;
+	elapsed=elapsedDecode=0.;
+	while(1)
+	{
+		int tile=0;
+		double start=gettime();
+		for(row=0, dstptr=dstbuf; row<ntilesh; row++, dstptr+=pitch*tileh)
+		{
+			for(col=0, dstptr2=dstptr; col<ntilesw; col++, tile++, dstptr2+=ps*tilew)
+			{
+				int width=dotile? min(tilew, w-col*tilew):scaledw;
+				int height=dotile? min(tileh, h-row*tileh):scaledh;
+				if(doyuv)
+				{
+					double startDecode;
+					if(tjDecompressToYUV2(handle, jpegbuf[tile], jpegsize[tile], yuvbuf,
+						width, yuvpad, height, flags)==-1)
+						_throwtj("executing tjDecompressToYUV2()");
+					startDecode=gettime();
+					if(tjDecodeYUV(handle, yuvbuf, yuvpad, subsamp, dstptr2, width,
+						pitch, height, pf, flags)==-1)
+						_throwtj("executing tjDecodeYUV()");
+					if(iter>=0) elapsedDecode+=gettime()-startDecode;
+				}
+				else
+					if(tjDecompress2(handle, jpegbuf[tile], jpegsize[tile], dstptr2,
+						width, pitch, height, pf, flags)==-1)
+						_throwtj("executing tjDecompress2()");
+			}
+		}
+		iter++;
+		if(iter>=1)
+		{
+			elapsed+=gettime()-start;
+			if(elapsed>=benchtime) break;
+		}
+	}
+	if(doyuv) elapsed-=elapsedDecode;
+
+	if(tjDestroy(handle)==-1) _throwtj("executing tjDestroy()");
+	handle=NULL;
+
+	if(quiet)
+	{
+		printf("%-6s%s",
+			sigfig((double)(w*h)/1000000.*(double)iter/elapsed, 4, tempstr, 1024),
+			quiet==2? "\n":"  ");
+		if(doyuv)
+			printf("%s\n",
+				sigfig((double)(w*h)/1000000.*(double)iter/elapsedDecode, 4, tempstr,
+					1024));
+		else if(quiet!=2) printf("\n");
+	}
+	else
+	{
+		printf("%s --> Frame rate:         %f fps\n",
+			doyuv? "Decomp to YUV":"Decompress   ", (double)iter/elapsed);
+		printf("                  Throughput:         %f Megapixels/sec\n",
+			(double)(w*h)/1000000.*(double)iter/elapsed);
+		if(doyuv)
+		{
+			printf("YUV Decode    --> Frame rate:         %f fps\n",
+				(double)iter/elapsedDecode);
+			printf("                  Throughput:         %f Megapixels/sec\n",
+				(double)(w*h)/1000000.*(double)iter/elapsedDecode);
+		}
+	}
+
+	if (!dowrite) goto bailout;
+
+	if(sf.num!=1 || sf.denom!=1)
+		snprintf(sizestr, 20, "%d_%d", sf.num, sf.denom);
+	else if(tilew!=w || tileh!=h)
+		snprintf(sizestr, 20, "%dx%d", tilew, tileh);
+	else snprintf(sizestr, 20, "full");
+	if(decomponly)
+		snprintf(tempstr, 1024, "%s_%s.%s", filename, sizestr, ext);
+	else
+		snprintf(tempstr, 1024, "%s_%s%s_%s.%s", filename, subName[subsamp],
+			qualstr, sizestr, ext);
+
+	if(savebmp(tempstr, dstbuf, scaledw, scaledh, pf,
+		(flags&TJFLAG_BOTTOMUP)!=0)==-1)
+		_throwbmp("saving bitmap");
+	ptr=strrchr(tempstr, '.');
+	snprintf(ptr, 1024-(ptr-tempstr), "-err.%s", ext);
+	if(srcbuf && sf.num==1 && sf.denom==1)
+	{
+		if(!quiet) printf("Compression error written to %s.\n", tempstr);
+		if(subsamp==TJ_GRAYSCALE)
+		{
+			int index, index2;
+			for(row=0, index=0; row<h; row++, index+=pitch)
+			{
+				for(col=0, index2=index; col<w; col++, index2+=ps)
+				{
+					int rindex=index2+tjRedOffset[pf];
+					int gindex=index2+tjGreenOffset[pf];
+					int bindex=index2+tjBlueOffset[pf];
+					int y=(int)((double)srcbuf[rindex]*0.299
+						+ (double)srcbuf[gindex]*0.587
+						+ (double)srcbuf[bindex]*0.114 + 0.5);
+					if(y>255) y=255;  if(y<0) y=0;
+					dstbuf[rindex]=abs(dstbuf[rindex]-y);
+					dstbuf[gindex]=abs(dstbuf[gindex]-y);
+					dstbuf[bindex]=abs(dstbuf[bindex]-y);
+				}
+			}
+		}
+		else
+		{
+			for(row=0; row<h; row++)
+				for(col=0; col<w*ps; col++)
+					dstbuf[pitch*row+col]
+						=abs(dstbuf[pitch*row+col]-srcbuf[pitch*row+col]);
+		}
+		if(savebmp(tempstr, dstbuf, w, h, pf,
+			(flags&TJFLAG_BOTTOMUP)!=0)==-1)
+			_throwbmp("saving bitmap");
+	}
+
+	bailout:
+	if(file) fclose(file);
+	if(handle) tjDestroy(handle);
+	if(dstbuf && dstbufalloc) free(dstbuf);
+	if(yuvbuf) free(yuvbuf);
+	return retval;
+}
+
+
+int fullTest(unsigned char *srcbuf, int w, int h, int subsamp, int jpegqual,
+	char *filename)
+{
+	char tempstr[1024], tempstr2[80];
+	FILE *file=NULL;  tjhandle handle=NULL;
+	unsigned char **jpegbuf=NULL, *yuvbuf=NULL, *tmpbuf=NULL, *srcptr, *srcptr2;
+	double start, elapsed, elapsedEncode;
+	int totaljpegsize=0, row, col, i, tilew=w, tileh=h, retval=0;
+	int iter, yuvsize=0;
+	unsigned long *jpegsize=NULL;
+	int ps=tjPixelSize[pf];
+	int ntilesw=1, ntilesh=1, pitch=w*ps;
+	const char *pfStr=pixFormatStr[pf];
+
+	if((tmpbuf=(unsigned char *)malloc(pitch*h)) == NULL)
+		_throwunix("allocating temporary image buffer");
+
+	if(!quiet)
+		printf(">>>>>  %s (%s) <--> JPEG %s Q%d  <<<<<\n", pfStr,
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-up":"Top-down", subNameLong[subsamp],
+			jpegqual);
+
+	for(tilew=dotile? 8:w, tileh=dotile? 8:h; ; tilew*=2, tileh*=2)
+	{
+		if(tilew>w) tilew=w;  if(tileh>h) tileh=h;
+		ntilesw=(w+tilew-1)/tilew;  ntilesh=(h+tileh-1)/tileh;
+
+		if((jpegbuf=(unsigned char **)malloc(sizeof(unsigned char *)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG tile array");
+		memset(jpegbuf, 0, sizeof(unsigned char *)*ntilesw*ntilesh);
+		if((jpegsize=(unsigned long *)malloc(sizeof(unsigned long)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG size array");
+		memset(jpegsize, 0, sizeof(unsigned long)*ntilesw*ntilesh);
+
+		if((flags&TJFLAG_NOREALLOC)!=0)
+			for(i=0; i<ntilesw*ntilesh; i++)
+			{
+				if((jpegbuf[i]=(unsigned char *)tjAlloc(tjBufSize(tilew, tileh,
+					subsamp)))==NULL)
+					_throwunix("allocating JPEG tiles");
+			}
+
+		/* Compression test */
+		if(quiet==1)
+			printf("%-4s (%s)  %-5s    %-3d   ", pfStr,
+				(flags&TJFLAG_BOTTOMUP)? "BU":"TD", subNameLong[subsamp], jpegqual);
+		for(i=0; i<h; i++)
+			memcpy(&tmpbuf[pitch*i], &srcbuf[w*ps*i], w*ps);
+		if((handle=tjInitCompress())==NULL)
+			_throwtj("executing tjInitCompress()");
+
+		if(doyuv)
+		{
+			yuvsize=tjBufSizeYUV2(tilew, yuvpad, tileh, subsamp);
+			if((yuvbuf=(unsigned char *)malloc(yuvsize))==NULL)
+				_throwunix("allocating YUV buffer");
+			memset(yuvbuf, 127, yuvsize);
+		}
+
+		/* Benchmark */
+		iter=-warmup;
+		elapsed=elapsedEncode=0.;
+		while(1)
+		{
+			int tile=0;
+			totaljpegsize=0;
+			start=gettime();
+			for(row=0, srcptr=srcbuf; row<ntilesh; row++, srcptr+=pitch*tileh)
+			{
+				for(col=0, srcptr2=srcptr; col<ntilesw; col++, tile++,
+					srcptr2+=ps*tilew)
+				{
+					int width=min(tilew, w-col*tilew);
+					int height=min(tileh, h-row*tileh);
+					if(doyuv)
+					{
+						double startEncode=gettime();
+						if(tjEncodeYUV3(handle, srcptr2, width, pitch, height, pf, yuvbuf,
+							yuvpad, subsamp, flags)==-1)
+							_throwtj("executing tjEncodeYUV3()");
+						if(iter>=0) elapsedEncode+=gettime()-startEncode;
+						if(tjCompressFromYUV(handle, yuvbuf, width, yuvpad, height,
+							subsamp, &jpegbuf[tile], &jpegsize[tile], jpegqual, flags)==-1)
+							_throwtj("executing tjCompressFromYUV()");
+					}
+					else
+					{
+						if(tjCompress2(handle, srcptr2, width, pitch, height, pf,
+							&jpegbuf[tile], &jpegsize[tile], subsamp, jpegqual, flags)==-1)
+							_throwtj("executing tjCompress2()");
+					}
+					totaljpegsize+=jpegsize[tile];
+				}
+			}
+			iter++;
+			if(iter>=1)
+			{
+				elapsed+=gettime()-start;
+				if(elapsed>=benchtime) break;
+			}
+		}
+		if(doyuv) elapsed-=elapsedEncode;
+
+		if(tjDestroy(handle)==-1) _throwtj("executing tjDestroy()");
+		handle=NULL;
+
+		if(quiet==1) printf("%-5d  %-5d   ", tilew, tileh);
+		if(quiet)
+		{
+			if(doyuv)
+				printf("%-6s%s",
+					sigfig((double)(w*h)/1000000.*(double)iter/elapsedEncode, 4, tempstr,
+						1024), quiet==2? "\n":"  ");
+			printf("%-6s%s",
+				sigfig((double)(w*h)/1000000.*(double)iter/elapsed, 4,	tempstr, 1024),
+				quiet==2? "\n":"  ");
+			printf("%-6s%s",
+				sigfig((double)(w*h*ps)/(double)totaljpegsize, 4, tempstr2, 80),
+				quiet==2? "\n":"  ");
+		}
+		else
+		{
+			printf("\n%s size: %d x %d\n", dotile? "Tile":"Image", tilew,
+				tileh);
+			if(doyuv)
+			{
+				printf("Encode YUV    --> Frame rate:         %f fps\n",
+					(double)iter/elapsedEncode);
+				printf("                  Output image size:  %d bytes\n", yuvsize);
+				printf("                  Compression ratio:  %f:1\n",
+					(double)(w*h*ps)/(double)yuvsize);
+				printf("                  Throughput:         %f Megapixels/sec\n",
+					(double)(w*h)/1000000.*(double)iter/elapsedEncode);
+				printf("                  Output bit stream:  %f Megabits/sec\n",
+					(double)yuvsize*8./1000000.*(double)iter/elapsedEncode);
+			}
+			printf("%s --> Frame rate:         %f fps\n",
+				doyuv? "Comp from YUV":"Compress     ", (double)iter/elapsed);
+			printf("                  Output image size:  %d bytes\n",
+				totaljpegsize);
+			printf("                  Compression ratio:  %f:1\n",
+				(double)(w*h*ps)/(double)totaljpegsize);
+			printf("                  Throughput:         %f Megapixels/sec\n",
+				(double)(w*h)/1000000.*(double)iter/elapsed);
+			printf("                  Output bit stream:  %f Megabits/sec\n",
+				(double)totaljpegsize*8./1000000.*(double)iter/elapsed);
+		}
+		if(tilew==w && tileh==h && dowrite)
+		{
+			snprintf(tempstr, 1024, "%s_%s_Q%d.jpg", filename, subName[subsamp],
+				jpegqual);
+			if((file=fopen(tempstr, "wb"))==NULL)
+				_throwunix("opening reference image");
+			if(fwrite(jpegbuf[0], jpegsize[0], 1, file)!=1)
+				_throwunix("writing reference image");
+			fclose(file);  file=NULL;
+			if(!quiet) printf("Reference image written to %s\n", tempstr);
+		}
+
+		/* Decompression test */
+		if(!componly)
+		{
+			if(decomp(srcbuf, jpegbuf, jpegsize, tmpbuf, w, h, subsamp, jpegqual,
+				filename, tilew, tileh)==-1)
+				goto bailout;
+		}
+
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+		free(jpegsize);  jpegsize=NULL;
+		if(doyuv)
+		{
+			free(yuvbuf);  yuvbuf=NULL;
+		}
+
+		if(tilew==w && tileh==h) break;
+	}
+
+	bailout:
+	if(file) {fclose(file);  file=NULL;}
+	if(jpegbuf)
+	{
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+	}
+	if(yuvbuf) {free(yuvbuf);  yuvbuf=NULL;}
+	if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+	if(tmpbuf) {free(tmpbuf);  tmpbuf=NULL;}
+	if(handle) {tjDestroy(handle);  handle=NULL;}
+	return retval;
+}
+
+
+int decompTest(char *filename)
+{
+	FILE *file=NULL;  tjhandle handle=NULL;
+	unsigned char **jpegbuf=NULL, *srcbuf=NULL;
+	unsigned long *jpegsize=NULL, srcsize, totaljpegsize;
+	tjtransform *t=NULL;
+	int w=0, h=0, subsamp=-1, cs=-1, _w, _h, _tilew, _tileh,
+		_ntilesw, _ntilesh, _subsamp;
+	char *temp=NULL, tempstr[80], tempstr2[80];
+	int row, col, i, iter, tilew, tileh, ntilesw=1, ntilesh=1, retval=0;
+	double start, elapsed;
+	int ps=tjPixelSize[pf], tile;
+
+	if((file=fopen(filename, "rb"))==NULL)
+		_throwunix("opening file");
+	if(fseek(file, 0, SEEK_END)<0 || (srcsize=ftell(file))==(unsigned long)-1)
+		_throwunix("determining file size");
+	if((srcbuf=(unsigned char *)malloc(srcsize))==NULL)
+		_throwunix("allocating memory");
+	if(fseek(file, 0, SEEK_SET)<0)
+		_throwunix("setting file position");
+	if(fread(srcbuf, srcsize, 1, file)<1)
+		_throwunix("reading JPEG data");
+	fclose(file);  file=NULL;
+
+	temp=strrchr(filename, '.');
+	if(temp!=NULL) *temp='\0';
+
+	if((handle=tjInitTransform())==NULL)
+		_throwtj("executing tjInitTransform()");
+	if(tjDecompressHeader3(handle, srcbuf, srcsize, &w, &h, &subsamp, &cs)==-1)
+		_throwtj("executing tjDecompressHeader3()");
+	if(cs==TJCS_YCCK || cs==TJCS_CMYK)
+	{
+		pf=TJPF_CMYK;  ps=tjPixelSize[pf];
+	}
+
+	if(quiet==1)
+	{
+		printf("All performance values in Mpixels/sec\n\n");
+		printf("Bitmap     JPEG   JPEG     %s  %s   Xform   Comp    Decomp  ",
+			dotile? "Tile ":"Image", dotile? "Tile ":"Image");
+		if(doyuv) printf("Decode");
+		printf("\n");
+		printf("Format     CS     Subsamp  Width  Height  Perf    Ratio   Perf    ");
+		if(doyuv) printf("Perf");
+		printf("\n\n");
+	}
+	else if(!quiet)
+		printf(">>>>>  JPEG %s --> %s (%s)  <<<<<\n",
+			formatName(subsamp, cs, tempstr), pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-up":"Top-down");
+
+	for(tilew=dotile? 16:w, tileh=dotile? 16:h; ; tilew*=2, tileh*=2)
+	{
+		if(tilew>w) tilew=w;  if(tileh>h) tileh=h;
+		ntilesw=(w+tilew-1)/tilew;  ntilesh=(h+tileh-1)/tileh;
+
+		if((jpegbuf=(unsigned char **)malloc(sizeof(unsigned char *)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG tile array");
+		memset(jpegbuf, 0, sizeof(unsigned char *)*ntilesw*ntilesh);
+		if((jpegsize=(unsigned long *)malloc(sizeof(unsigned long)
+			*ntilesw*ntilesh))==NULL)
+			_throwunix("allocating JPEG size array");
+		memset(jpegsize, 0, sizeof(unsigned long)*ntilesw*ntilesh);
+
+		if((flags&TJFLAG_NOREALLOC)!=0 || !dotile)
+			for(i=0; i<ntilesw*ntilesh; i++)
+			{
+				if((jpegbuf[i]=(unsigned char *)tjAlloc(tjBufSize(tilew, tileh,
+					subsamp)))==NULL)
+					_throwunix("allocating JPEG tiles");
+			}
+
+		_w=w;  _h=h;  _tilew=tilew;  _tileh=tileh;
+		if(!quiet)
+		{
+			printf("\n%s size: %d x %d", dotile? "Tile":"Image", _tilew,
+				_tileh);
+			if(sf.num!=1 || sf.denom!=1)
+				printf(" --> %d x %d", TJSCALED(_w, sf), TJSCALED(_h, sf));
+			printf("\n");
+		}
+		else if(quiet==1)
+		{
+			printf("%-4s (%s)  %-5s  %-5s    ", pixFormatStr[pf],
+				(flags&TJFLAG_BOTTOMUP)? "BU":"TD", csName[cs], subNameLong[subsamp]);
+			printf("%-5d  %-5d   ", tilew, tileh);
+		}
+
+		_subsamp=subsamp;
+		if(dotile || xformop!=TJXOP_NONE || xformopt!=0 || customFilter)
+		{
+			if((t=(tjtransform *)malloc(sizeof(tjtransform)*ntilesw*ntilesh))
+				==NULL)
+				_throwunix("allocating image transform array");
+
+			if(xformop==TJXOP_TRANSPOSE || xformop==TJXOP_TRANSVERSE
+				|| xformop==TJXOP_ROT90 || xformop==TJXOP_ROT270)
+			{
+				_w=h;  _h=w;  _tilew=tileh;  _tileh=tilew;
+			}
+
+			if(xformopt&TJXOPT_GRAY) _subsamp=TJ_GRAYSCALE;
+			if(xformop==TJXOP_HFLIP || xformop==TJXOP_ROT180)
+				_w=_w-(_w%tjMCUWidth[_subsamp]);
+			if(xformop==TJXOP_VFLIP || xformop==TJXOP_ROT180)
+				_h=_h-(_h%tjMCUHeight[_subsamp]);
+			if(xformop==TJXOP_TRANSVERSE || xformop==TJXOP_ROT90)
+				_w=_w-(_w%tjMCUHeight[_subsamp]);
+			if(xformop==TJXOP_TRANSVERSE || xformop==TJXOP_ROT270)
+				_h=_h-(_h%tjMCUWidth[_subsamp]);
+			_ntilesw=(_w+_tilew-1)/_tilew;
+			_ntilesh=(_h+_tileh-1)/_tileh;
+
+			if(xformop==TJXOP_TRANSPOSE || xformop==TJXOP_TRANSVERSE
+				|| xformop==TJXOP_ROT90 || xformop==TJXOP_ROT270)
+			{
+				if(_subsamp==TJSAMP_422) _subsamp=TJSAMP_440;
+				else if(_subsamp==TJSAMP_440) _subsamp=TJSAMP_422;
+			}
+
+			for(row=0, tile=0; row<_ntilesh; row++)
+			{
+				for(col=0; col<_ntilesw; col++, tile++)
+				{
+					t[tile].r.w=min(_tilew, _w-col*_tilew);
+					t[tile].r.h=min(_tileh, _h-row*_tileh);
+					t[tile].r.x=col*_tilew;
+					t[tile].r.y=row*_tileh;
+					t[tile].op=xformop;
+					t[tile].options=xformopt|TJXOPT_TRIM;
+					t[tile].customFilter=customFilter;
+					if(t[tile].options&TJXOPT_NOOUTPUT && jpegbuf[tile])
+					{
+						tjFree(jpegbuf[tile]);  jpegbuf[tile]=NULL;
+					}
+				}
+			}
+
+			iter=-warmup;
+			elapsed=0.;
+			while(1)
+			{
+				start=gettime();
+				if(tjTransform(handle, srcbuf, srcsize, _ntilesw*_ntilesh, jpegbuf,
+					jpegsize, t, flags)==-1)
+					_throwtj("executing tjTransform()");
+				iter++;
+				if(iter>=1)
+				{
+					elapsed+=gettime()-start;
+					if(elapsed>=benchtime) break;
+				}
+			}
+
+			free(t);  t=NULL;
+
+			for(tile=0, totaljpegsize=0; tile<_ntilesw*_ntilesh; tile++)
+				totaljpegsize+=jpegsize[tile];
+
+			if(quiet)
+			{
+				printf("%-6s%s%-6s%s",
+					sigfig((double)(w*h)/1000000./elapsed, 4, tempstr, 80),
+					quiet==2? "\n":"  ",
+					sigfig((double)(w*h*ps)/(double)totaljpegsize, 4, tempstr2, 80),
+					quiet==2? "\n":"  ");
+			}
+			else if(!quiet)
+			{
+				printf("Transform     --> Frame rate:         %f fps\n", 1.0/elapsed);
+				printf("                  Output image size:  %lu bytes\n", totaljpegsize);
+				printf("                  Compression ratio:  %f:1\n",
+					(double)(w*h*ps)/(double)totaljpegsize);
+				printf("                  Throughput:         %f Megapixels/sec\n",
+					(double)(w*h)/1000000./elapsed);
+				printf("                  Output bit stream:  %f Megabits/sec\n",
+					(double)totaljpegsize*8./1000000./elapsed);
+			}
+		}
+		else
+		{
+			if(quiet==1) printf("N/A     N/A     ");
+			jpegsize[0]=srcsize;
+			memcpy(jpegbuf[0], srcbuf, srcsize);
+		}
+
+		if(w==tilew) _tilew=_w;
+		if(h==tileh) _tileh=_h;
+		if(!(xformopt&TJXOPT_NOOUTPUT))
+		{
+			if(decomp(NULL, jpegbuf, jpegsize, NULL, _w, _h, _subsamp, 0,
+				filename, _tilew, _tileh)==-1)
+				goto bailout;
+		}
+		else if(quiet==1) printf("N/A\n");
+
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+		if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+
+		if(tilew==w && tileh==h) break;
+	}
+
+	bailout:
+	if(file) {fclose(file);  file=NULL;}
+	if(jpegbuf)
+	{
+		for(i=0; i<ntilesw*ntilesh; i++)
+		{
+			if(jpegbuf[i]) tjFree(jpegbuf[i]);  jpegbuf[i]=NULL;
+		}
+		free(jpegbuf);  jpegbuf=NULL;
+	}
+	if(jpegsize) {free(jpegsize);  jpegsize=NULL;}
+	if(srcbuf) {free(srcbuf);  srcbuf=NULL;}
+	if(t) {free(t);  t=NULL;}
+	if(handle) {tjDestroy(handle);  handle=NULL;}
+	return retval;
+}
+
+
+void usage(char *progname)
+{
+	int i;
+	printf("USAGE: %s\n", progname);
+	printf("       <Inputfile (BMP|PPM)> <Quality> [options]\n\n");
+	printf("       %s\n", progname);
+	printf("       <Inputfile (JPG)> [options]\n\n");
+	printf("Options:\n\n");
+	printf("-alloc = Dynamically allocate JPEG image buffers\n");
+	printf("-bmp = Generate output images in Windows Bitmap format (default = PPM)\n");
+	printf("-bottomup = Test bottom-up compression/decompression\n");
+	printf("-tile = Test performance of the codec when the image is encoded as separate\n");
+	printf("     tiles of varying sizes.\n");
+	printf("-rgb, -bgr, -rgbx, -bgrx, -xbgr, -xrgb =\n");
+	printf("     Test the specified color conversion path in the codec (default = BGR)\n");
+	printf("-cmyk = Indirectly test YCCK JPEG compression/decompression (the source\n");
+	printf("     and destination bitmaps are still RGB.  The conversion is done\n");
+	printf("     internally prior to compression or after decompression.)\n");
+	printf("-fastupsample = Use the fastest chrominance upsampling algorithm available in\n");
+	printf("     the underlying codec\n");
+	printf("-fastdct = Use the fastest DCT/IDCT algorithms available in the underlying\n");
+	printf("     codec\n");
+	printf("-accuratedct = Use the most accurate DCT/IDCT algorithms available in the\n");
+	printf("     underlying codec\n");
+	printf("-subsamp <s> = When testing JPEG compression, this option specifies the level\n");
+	printf("     of chrominance subsampling to use (<s> = 444, 422, 440, 420, 411, or\n");
+	printf("     GRAY).  The default is to test Grayscale, 4:2:0, 4:2:2, and 4:4:4 in\n");
+	printf("     sequence.\n");
+	printf("-quiet = Output results in tabular rather than verbose format\n");
+	printf("-yuv = Test YUV encoding/decoding functions\n");
+	printf("-yuvpad <p> = If testing YUV encoding/decoding, this specifies the number of\n");
+	printf("     bytes to which each row of each plane in the intermediate YUV image is\n");
+	printf("     padded (default = 1)\n");
+	printf("-scale M/N = Scale down the width/height of the decompressed JPEG image by a\n");
+	printf("     factor of M/N (M/N = ");
+	for(i=0; i<nsf; i++)
+	{
+		printf("%d/%d", scalingfactors[i].num, scalingfactors[i].denom);
+		if(nsf==2 && i!=nsf-1) printf(" or ");
+		else if(nsf>2)
+		{
+			if(i!=nsf-1) printf(", ");
+			if(i==nsf-2) printf("or ");
+		}
+		if(i%8==0 && i!=0) printf("\n     ");
+	}
+	printf(")\n");
+	printf("-hflip, -vflip, -transpose, -transverse, -rot90, -rot180, -rot270 =\n");
+	printf("     Perform the corresponding lossless transform prior to\n");
+	printf("     decompression (these options are mutually exclusive)\n");
+	printf("-grayscale = Perform lossless grayscale conversion prior to decompression\n");
+	printf("     test (can be combined with the other transforms above)\n");
+	printf("-benchtime <t> = Run each benchmark for at least <t> seconds (default = 5.0)\n");
+	printf("-warmup <w> = Execute each benchmark <w> times to prime the cache before\n");
+	printf("     taking performance measurements (default = 1)\n");
+	printf("-componly = Stop after running compression tests.  Do not test decompression.\n");
+	printf("-nowrite = Do not write reference or output images (improves consistency of\n");
+	printf("     performance measurements.)\n\n");
+	printf("NOTE:  If the quality is specified as a range (e.g. 90-100), a separate\n");
+	printf("test will be performed for all quality values in the range.\n\n");
+	exit(1);
+}
+
+
+int main(int argc, char *argv[])
+{
+	unsigned char *srcbuf=NULL;  int w=0, h=0, i, j;
+	int minqual=-1, maxqual=-1;  char *temp;
+	int minarg=2, retval=0, subsamp=-1;
+
+	if((scalingfactors=tjGetScalingFactors(&nsf))==NULL || nsf==0)
+		_throwtj("executing tjGetScalingFactors()");
+
+	if(argc<minarg) usage(argv[0]);
+
+	temp=strrchr(argv[1], '.');
+	if(temp!=NULL)
+	{
+		if(!strcasecmp(temp, ".bmp")) ext="bmp";
+		if(!strcasecmp(temp, ".jpg") || !strcasecmp(temp, ".jpeg")) decomponly=1;
+	}
+
+	printf("\n");
+
+	if(!decomponly)
+	{
+		minarg=3;
+		if(argc<minarg) usage(argv[0]);
+		if((minqual=atoi(argv[2]))<1 || minqual>100)
+		{
+			puts("ERROR: Quality must be between 1 and 100.");
+			exit(1);
+		}
+		if((temp=strchr(argv[2], '-'))!=NULL && strlen(temp)>1
+			&& sscanf(&temp[1], "%d", &maxqual)==1 && maxqual>minqual && maxqual>=1
+			&& maxqual<=100) {}
+		else maxqual=minqual;
+	}
+
+	if(argc>minarg)
+	{
+		for(i=minarg; i<argc; i++)
+		{
+			if(!strcasecmp(argv[i], "-tile"))
+			{
+				dotile=1;  xformopt|=TJXOPT_CROP;
+			}
+			if(!strcasecmp(argv[i], "-fastupsample"))
+			{
+				printf("Using fast upsampling code\n\n");
+				flags|=TJFLAG_FASTUPSAMPLE;
+			}
+			if(!strcasecmp(argv[i], "-fastdct"))
+			{
+				printf("Using fastest DCT/IDCT algorithm\n\n");
+				flags|=TJFLAG_FASTDCT;
+			}
+			if(!strcasecmp(argv[i], "-accuratedct"))
+			{
+				printf("Using most accurate DCT/IDCT algorithm\n\n");
+				flags|=TJFLAG_ACCURATEDCT;
+			}
+			if(!strcasecmp(argv[i], "-rgb")) pf=TJPF_RGB;
+			if(!strcasecmp(argv[i], "-rgbx")) pf=TJPF_RGBX;
+			if(!strcasecmp(argv[i], "-bgr")) pf=TJPF_BGR;
+			if(!strcasecmp(argv[i], "-bgrx")) pf=TJPF_BGRX;
+			if(!strcasecmp(argv[i], "-xbgr")) pf=TJPF_XBGR;
+			if(!strcasecmp(argv[i], "-xrgb")) pf=TJPF_XRGB;
+			if(!strcasecmp(argv[i], "-cmyk")) pf=TJPF_CMYK;
+			if(!strcasecmp(argv[i], "-bottomup")) flags|=TJFLAG_BOTTOMUP;
+			if(!strcasecmp(argv[i], "-quiet")) quiet=1;
+			if(!strcasecmp(argv[i], "-qq")) quiet=2;
+			if(!strcasecmp(argv[i], "-scale") && i<argc-1)
+			{
+				int temp1=0, temp2=0, match=0;
+				if(sscanf(argv[++i], "%d/%d", &temp1, &temp2)==2)
+				{
+					for(j=0; j<nsf; j++)
+					{
+						if((double)temp1/(double)temp2
+							== (double)scalingfactors[j].num/(double)scalingfactors[j].denom)
+						{
+							sf=scalingfactors[j];
+							match=1;  break;
+						}
+					}
+					if(!match) usage(argv[0]);
+				}
+				else usage(argv[0]);
+			}
+			if(!strcasecmp(argv[i], "-hflip")) xformop=TJXOP_HFLIP;
+			if(!strcasecmp(argv[i], "-vflip")) xformop=TJXOP_VFLIP;
+			if(!strcasecmp(argv[i], "-transpose")) xformop=TJXOP_TRANSPOSE;
+			if(!strcasecmp(argv[i], "-transverse")) xformop=TJXOP_TRANSVERSE;
+			if(!strcasecmp(argv[i], "-rot90")) xformop=TJXOP_ROT90;
+			if(!strcasecmp(argv[i], "-rot180")) xformop=TJXOP_ROT180;
+			if(!strcasecmp(argv[i], "-rot270")) xformop=TJXOP_ROT270;
+			if(!strcasecmp(argv[i], "-grayscale")) xformopt|=TJXOPT_GRAY;
+			if(!strcasecmp(argv[i], "-custom")) customFilter=dummyDCTFilter;
+			if(!strcasecmp(argv[i], "-nooutput")) xformopt|=TJXOPT_NOOUTPUT;
+			if(!strcasecmp(argv[i], "-benchtime") && i<argc-1)
+			{
+				double temp=atof(argv[++i]);
+				if(temp>0.0) benchtime=temp;
+				else usage(argv[0]);
+			}
+			if(!strcasecmp(argv[i], "-warmup") && i<argc-1)
+			{
+				int temp=atoi(argv[++i]);
+				if(temp>=0)
+				{
+					warmup=temp;
+					printf("Warmup runs = %d\n\n", warmup);
+				}
+				else usage(argv[0]);
+			}
+			if(!strcmp(argv[i], "-?")) usage(argv[0]);
+			if(!strcasecmp(argv[i], "-alloc")) flags&=(~TJFLAG_NOREALLOC);
+			if(!strcasecmp(argv[i], "-bmp")) ext="bmp";
+			if(!strcasecmp(argv[i], "-yuv"))
+			{
+				printf("Testing YUV planar encoding/decoding\n\n");
+				doyuv=1;
+			}
+			if(!strcasecmp(argv[i], "-yuvpad") && i<argc-1)
+			{
+				int temp=atoi(argv[++i]);
+				if(temp>=1) yuvpad=temp;
+			}
+			if(!strcasecmp(argv[i], "-subsamp") && i<argc-1)
+			{
+				i++;
+				if(toupper(argv[i][0])=='G') subsamp=TJSAMP_GRAY;
+				else
+				{
+					int temp=atoi(argv[i]);
+					switch(temp)
+					{
+						case 444:  subsamp=TJSAMP_444;  break;
+						case 422:  subsamp=TJSAMP_422;  break;
+						case 440:  subsamp=TJSAMP_440;  break;
+						case 420:  subsamp=TJSAMP_420;  break;
+						case 411:  subsamp=TJSAMP_411;  break;
+					}
+				}
+			}
+			if(!strcasecmp(argv[i], "-componly")) componly=1;
+			if(!strcasecmp(argv[i], "-nowrite")) dowrite=0;
+		}
+	}
+
+	if((sf.num!=1 || sf.denom!=1) && dotile)
+	{
+		printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+		printf("work when scaled decompression is enabled.\n");
+		dotile=0;
+	}
+
+	if((flags&TJFLAG_NOREALLOC)==0 && dotile)
+	{
+		printf("Disabling tiled compression/decompression tests, because those tests do not\n");
+		printf("work when dynamic JPEG buffer allocation is enabled.\n\n");
+		dotile=0;
+	}
+
+	if(!decomponly)
+	{
+		if(loadbmp(argv[1], &srcbuf, &w, &h, pf, (flags&TJFLAG_BOTTOMUP)!=0)==-1)
+			_throwbmp("loading bitmap");
+		temp=strrchr(argv[1], '.');
+		if(temp!=NULL) *temp='\0';
+	}
+
+	if(quiet==1 && !decomponly)
+	{
+		printf("All performance values in Mpixels/sec\n\n");
+		printf("Bitmap     JPEG     JPEG  %s  %s   ",
+			dotile? "Tile ":"Image", dotile? "Tile ":"Image");
+		if(doyuv) printf("Encode  ");
+		printf("Comp    Comp    Decomp  ");
+		if(doyuv) printf("Decode");
+		printf("\n");
+		printf("Format     Subsamp  Qual  Width  Height  ");
+		if(doyuv) printf("Perf    ");
+		printf("Perf    Ratio   Perf    ");
+		if(doyuv) printf("Perf");
+		printf("\n\n");
+	}
+
+	if(decomponly)
+	{
+		decompTest(argv[1]);
+		printf("\n");
+		goto bailout;
+	}
+	if(subsamp>=0 && subsamp<TJ_NUMSAMP)
+	{
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, subsamp, i, argv[1]);
+		printf("\n");
+	}
+	else
+	{
+		if(pf!=TJPF_CMYK)
+		{
+			for(i=maxqual; i>=minqual; i--)
+				fullTest(srcbuf, w, h, TJSAMP_GRAY, i, argv[1]);
+			printf("\n");
+		}
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_420, i, argv[1]);
+		printf("\n");
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_422, i, argv[1]);
+		printf("\n");
+		for(i=maxqual; i>=minqual; i--)
+			fullTest(srcbuf, w, h, TJSAMP_444, i, argv[1]);
+		printf("\n");
+	}
+
+	bailout:
+	if(srcbuf) free(srcbuf);
+	return retval;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbenchtest.in glmark2-2021.02/src/libjpeg-turbo/tjbenchtest.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbenchtest.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjbenchtest.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,249 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+EXT=bmp
+IMAGES="vgl_5674_0098.${EXT} vgl_6434_0018a.${EXT} vgl_6548_0026a.${EXT} nightshot_iso_100.${EXT}"
+IMGDIR=@srcdir@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_output.XXXXXX`
+EXEDIR=.
+BMPARG=
+NSARG=
+YUVARG=
+ALLOC=0
+ALLOCARG=
+if [ "$EXT" = "bmp" ]; then BMPARG=-bmp; fi
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjbenchtest.log
+
+if [ $# -gt 0 ]; then
+	if [ "$1" = "-yuv" ]; then
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.${EXT}"
+	fi
+	if [ "$1" = "-alloc" ]; then
+		ALLOCARG=-alloc
+		ALLOC=1
+	fi
+fi
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .${EXT}`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.${EXT}
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth $BMPARG -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.${EXT} $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG $ALLOCARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG
+		for samp in GRAY 444; do
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		runme $EXEDIR/tjbench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG
+		for samp in 420 422; do
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+			runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${scale}_djpeg.${EXT}
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG $ALLOCARG
+			if [ $ALLOC = 1 ]; then
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_full.${EXT} $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_full.${EXT}
+			else
+				for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].${EXT} \
+					$OUTDIR/${basename}_${samp}_Q95_full.${EXT}; do
+					runme cmp $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.${EXT}
+					rm $i
+				done
+			fi
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG $BMPARG -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $EXEDIR/tjbench $OUTDIR/${basename}_${samp}_Q95.jpg $BMPARG -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG $ALLOCARG
+				runme cmp $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT} $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.${EXT}
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.${EXT}
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbenchtest.java.in glmark2-2021.02/src/libjpeg-turbo/tjbenchtest.java.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjbenchtest.java.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjbenchtest.java.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,207 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@srcdir@/testimages
+OUTDIR=`mktemp -d /tmp/__tjbenchtest_java_output.XXXXXX`
+EXEDIR=.
+JAVA="@JAVA@ -cp java/turbojpeg.jar -Djava.library.path=.libs"
+BMPARG=
+NSARG=
+YUVARG=
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjbenchtest-java.log
+
+if [ $# -gt 0 ]; then
+	if [ "$1" = "-yuv" ]; then
+		NSARG=-nosmooth
+		YUVARG=-yuv
+
+# NOTE: The combination of tjEncodeYUV*() and tjCompressFromYUV*() does not
+# always produce bitwise-identical results to tjCompress*() if subsampling is
+# enabled.  In both cases, if the image width or height are not evenly
+# divisible by the MCU width/height, then the bottom and/or right edge are
+# expanded.  However, the libjpeg code performs this expansion prior to
+# downsampling, and TurboJPEG performs it in tjCompressFromYUV*(), which is
+# after downsampling.  Thus, the two will agree only if the width/height along
+# each downsampled dimension is an odd number or is evenly divisible by the MCU
+# width/height.  This disagreement basically amounts to a round-off error, but
+# there is no easy way around it, so for now, we just test the only image that
+# works.  (NOTE: nightshot_iso_100 does not suffer from the above issue, but
+# it suffers from an unrelated problem whereby the combination of
+# tjDecompressToYUV*() and tjDecodeYUV*() do not produce bitwise-identical
+# results to tjDecompress*() if decompression scaling is enabled.  This latter
+# phenomenon is not yet fully understood but is also believed to be some sort
+# of round-off error.)
+		IMAGES="vgl_6548_0026a.bmp"
+	fi
+fi
+
+# Standard tests
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -grayscale -outfile $OUTDIR/${basename}_GRAY_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 -outfile $OUTDIR/${basename}_420_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 -outfile $OUTDIR/${basename}_422_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 -outfile $OUTDIR/${basename}_444_fast_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -grayscale -outfile $OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 -outfile $OUTDIR/${basename}_420_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 -outfile $OUTDIR/${basename}_422_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	runme $EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 -outfile $OUTDIR/${basename}_444_accurate_cjpeg.jpg $IMGDIR/${basename}.bmp
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_default_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_fast_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+	for samp in 420 422; do
+		runme $EXEDIR/djpeg -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct fast -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+		runme $EXEDIR/djpeg -dct int -nosmooth -bmp -outfile $OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg
+	done
+
+	# Compression
+	for dct in accurate fast; do
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -quiet -benchtime 0.01 -warmup 0 -${dct}dct $YUVARG
+		for samp in GRAY 420 422 444; do
+			runme cmp $OUTDIR/${basename}_${samp}_Q95.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	for dct in fast accurate default; do
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			dctarg=
+		fi
+
+		# Tiled compression & decompression
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG
+		for samp in GRAY 444; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		runme $JAVA TJBench $OUTDIR/$image 95 -rgb -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG
+		for samp in 420 422; do
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+
+		# Tiled decompression
+		for samp in GRAY 444; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 ${dctarg} $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample ${dctarg} $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp $i -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		runme $EXEDIR/jpegtran -flip horizontal -trim -outfile $OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -flip vertical -trim -outfile $OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transpose -trim -outfile $OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -transverse -trim -outfile $OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 90 -trim -outfile $OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 180 -trim -outfile $OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+		runme $EXEDIR/jpegtran -rotate 270 -trim -outfile $OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg $OUTDIR/${basename}_${samp}_Q95.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444; do
+			runme $EXEDIR/djpeg -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+		for samp in 420 422; do
+			runme $EXEDIR/djpeg -nosmooth -rgb -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -fastupsample $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -tile -quiet -benchtime 0.01 -warmup 0 -grayscale $YUVARG
+			for i in $OUTDIR/${basename}_${samp}_Q95_[0-9]*[0-9]x[0-9]*[0-9].bmp \
+				$OUTDIR/${basename}_${samp}_Q95_full.bmp; do
+				runme cmp -i 54:54 $i $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+				rm $i
+			done
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				runme $EXEDIR/djpeg -rgb -scale ${scalearg} $NSARG -bmp -outfile $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+				runme $JAVA TJBench $OUTDIR/${basename}_${samp}_Q95.jpg -$xform -scale ${scalearg} -quiet -benchtime 0.01 -warmup 0 $YUVARG
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_Q95_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjexampletest.in glmark2-2021.02/src/libjpeg-turbo/tjexampletest.in
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjexampletest.in	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjexampletest.in	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,150 @@
+#!/bin/bash
+
+set -u
+set -e
+trap onexit INT
+trap onexit TERM
+trap onexit EXIT
+
+onexit()
+{
+	if [ -d $OUTDIR ]; then
+		rm -rf $OUTDIR
+	fi
+}
+
+runme()
+{
+	echo \*\*\* $*
+	$*
+}
+
+IMAGES="vgl_5674_0098.bmp vgl_6434_0018a.bmp vgl_6548_0026a.bmp nightshot_iso_100.bmp"
+IMGDIR=@srcdir@/testimages
+OUTDIR=__tjexampletest_output
+EXEDIR=.
+JAVA="@JAVA@ -cp java/turbojpeg.jar -Djava.library.path=.libs"
+
+if [ -d $OUTDIR ]; then
+	rm -rf $OUTDIR
+fi
+mkdir -p $OUTDIR
+
+exec >$EXEDIR/tjexampletest.log
+
+for image in $IMAGES; do
+
+	cp $IMGDIR/$image $OUTDIR
+	basename=`basename $image .bmp`
+	$EXEDIR/cjpeg -quality 95 -dct fast -grayscale $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_GRAY_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 2x2 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_420_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 2x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_422_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct fast -sample 1x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_444_fast_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -grayscale $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_GRAY_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 2x2 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_420_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 2x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_422_accurate_cjpeg.jpg
+	$EXEDIR/cjpeg -quality 95 -dct int -sample 1x1 $IMGDIR/${basename}.bmp >$OUTDIR/${basename}_444_accurate_cjpeg.jpg
+	for samp in GRAY 420 422 444; do
+		$EXEDIR/djpeg -rgb -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_default_djpeg.bmp
+		$EXEDIR/djpeg -dct fast -rgb -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_fast_djpeg.bmp
+		$EXEDIR/djpeg -dct int -rgb -bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg >$OUTDIR/${basename}_${samp}_accurate_djpeg.bmp
+	done
+	for samp in 420 422; do
+		$EXEDIR/djpeg -nosmooth -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_default_nosmooth_djpeg.bmp
+		$EXEDIR/djpeg -dct fast -nosmooth -bmp $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_fast_nosmooth_djpeg.bmp
+		$EXEDIR/djpeg -dct int -nosmooth -bmp $OUTDIR/${basename}_${samp}_accurate_cjpeg.jpg >$OUTDIR/${basename}_${samp}_accurate_nosmooth_djpeg.bmp
+	done
+
+	# Compression
+	for dct in fast accurate; do
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/$image $OUTDIR/${basename}_${samp}_${dct}.jpg -q 95 -samp ${samp} -${dct}dct
+			runme cmp $OUTDIR/${basename}_${samp}_${dct}.jpg $OUTDIR/${basename}_${samp}_${dct}_cjpeg.jpg
+		done
+	done
+
+	# Decompression
+	for dct in fast accurate default; do
+		srcdct=${dct}
+		dctarg=-${dct}dct
+		if [ "${dct}" = "default" ]; then
+			srcdct=fast
+			dctarg=
+		fi
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}.bmp ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}.bmp $OUTDIR/${basename}_${samp}_${dct}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}.bmp
+		done
+		for samp in 420 422; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_${srcdct}.jpg $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp -fastupsample ${dctarg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp $OUTDIR/${basename}_${samp}_${dct}_nosmooth_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${dct}_nosmooth.bmp
+		done
+	done
+
+	# Scaled decompression
+	for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+		scalearg=`echo $scale | sed s@_@/@g`
+		for samp in GRAY 420 422 444; do
+			$EXEDIR/djpeg -rgb -bmp -scale ${scalearg} $OUTDIR/${basename}_${samp}_fast_cjpeg.jpg >$OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${scale}.bmp -scale ${scalearg}
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${scale}.bmp $OUTDIR/${basename}_${samp}_${scale}_djpeg.bmp
+			rm $OUTDIR/${basename}_${samp}_${scale}.bmp
+		done
+	done
+
+	# Transforms
+	for samp in GRAY 420 422 444; do
+		$EXEDIR/jpegtran -crop 70x60+16+16 -flip horizontal -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_hflip_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -flip vertical -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_vflip_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -transpose -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_transpose_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -transverse -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_transverse_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 90 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot90_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 180 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot180_jpegtran.jpg
+		$EXEDIR/jpegtran -crop 70x60+16+16 -rotate 270 -trim $OUTDIR/${basename}_${samp}_fast.jpg >$OUTDIR/${basename}_${samp}_rot270_jpegtran.jpg
+	done
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 420 422 444; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -crop 16,16,70x60
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg
+			$EXEDIR/djpeg -rgb -bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 16,16,70x60
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+		for samp in 420 422; do
+			$EXEDIR/djpeg -nosmooth -rgb -bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -crop 16,16,70x60 -fastupsample
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_${samp}_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Grayscale transform
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.jpg -$xform -grayscale -crop 16,16,70x60
+			runme cmp $OUTDIR/${basename}_${samp}_${xform}.jpg $OUTDIR/${basename}_GRAY_${xform}_jpegtran.jpg
+			runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}.bmp -$xform -grayscale -crop 16,16,70x60
+			runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}.bmp $OUTDIR/${basename}_GRAY_${xform}_jpegtran.bmp
+			rm $OUTDIR/${basename}_${samp}_${xform}.bmp
+		done
+	done
+
+	# Transforms with scaling
+	for xform in hflip vflip transpose transverse rot90 rot180 rot270; do
+		for samp in GRAY 444 422 420; do
+			for scale in 2_1 15_8 7_4 13_8 3_2 11_8 5_4 9_8 7_8 3_4 5_8 1_2 3_8 1_4 1_8; do
+				scalearg=`echo $scale | sed s@_@/@g`
+				$EXEDIR/djpeg -rgb -bmp -scale ${scalearg} $OUTDIR/${basename}_${samp}_${xform}_jpegtran.jpg >$OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				runme $JAVA TJExample $OUTDIR/${basename}_${samp}_fast.jpg $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp -$xform -scale ${scalearg} -crop 16,16,70x60
+				runme cmp -i 54:54 $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp $OUTDIR/${basename}_${samp}_${xform}_${scale}_jpegtran.bmp
+				rm $OUTDIR/${basename}_${samp}_${xform}_${scale}.bmp
+			done
+		done
+	done
+
+done
+
+echo SUCCESS!
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjunittest.c glmark2-2021.02/src/libjpeg-turbo/tjunittest.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjunittest.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjunittest.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,735 @@
+/*
+ * Copyright (C)2009-2014 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * This program tests the various code paths in the TurboJPEG C Wrapper
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include "./tjutil.h"
+#include "./turbojpeg.h"
+#ifdef _WIN32
+ #include <time.h>
+ #define random() rand()
+#endif
+
+
+void usage(char *progName)
+{
+	printf("\nUSAGE: %s [options]\n", progName);
+	printf("Options:\n");
+	printf("-yuv = test YUV encoding/decoding support\n");
+	printf("-noyuvpad = do not pad each line of each Y, U, and V plane to the nearest\n");
+	printf("            4-byte boundary\n");
+	printf("-alloc = test automatic buffer allocation\n");
+	exit(1);
+}
+
+
+#define _throwtj() {printf("TurboJPEG ERROR:\n%s\n", tjGetErrorStr());  \
+	bailout();}
+#define _tj(f) {if((f)==-1) _throwtj();}
+#define _throw(m) {printf("ERROR: %s\n", m);  bailout();}
+
+const char *subNameLong[TJ_NUMSAMP]=
+{
+	"4:4:4", "4:2:2", "4:2:0", "GRAY", "4:4:0", "4:1:1"
+};
+const char *subName[TJ_NUMSAMP]={"444", "422", "420", "GRAY", "440", "411"};
+
+const char *pixFormatStr[TJ_NUMPF]=
+{
+	"RGB", "BGR", "RGBX", "BGRX", "XBGR", "XRGB", "Grayscale",
+	"RGBA", "BGRA", "ABGR", "ARGB", "CMYK"
+};
+
+const int alphaOffset[TJ_NUMPF] = {-1, -1, -1, -1, -1, -1, -1, 3, 3, 0, 0, -1};
+
+const int _3byteFormats[]={TJPF_RGB, TJPF_BGR};
+const int _4byteFormats[]={TJPF_RGBX, TJPF_BGRX, TJPF_XBGR, TJPF_XRGB,
+	TJPF_CMYK};
+const int _onlyGray[]={TJPF_GRAY};
+const int _onlyRGB[]={TJPF_RGB};
+
+int doyuv=0, alloc=0, pad=4;
+
+int exitStatus=0;
+#define bailout() {exitStatus=-1;  goto bailout;}
+
+
+void initBuf(unsigned char *buf, int w, int h, int pf, int flags)
+{
+	int roffset=tjRedOffset[pf];
+	int goffset=tjGreenOffset[pf];
+	int boffset=tjBlueOffset[pf];
+	int ps=tjPixelSize[pf];
+	int index, row, col, halfway=16;
+
+	if(pf==TJPF_GRAY)
+	{
+		memset(buf, 0, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0) buf[index]=(row<halfway)? 255:0;
+				else buf[index]=(row<halfway)? 76:226;
+			}
+		}
+	}
+	else if(pf==TJPF_CMYK)
+	{
+		memset(buf, 255, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0)
+				{
+					if(row>=halfway) buf[index*ps+3]=0;
+				}
+				else
+				{
+					buf[index*ps+2]=0;
+					if(row<halfway) buf[index*ps+1]=0;
+				}
+			}
+		}
+	}
+	else
+	{
+		memset(buf, 0, w*h*ps);
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				if(((row/8)+(col/8))%2==0)
+				{
+					if(row<halfway)
+					{
+						buf[index*ps+roffset]=255;
+						buf[index*ps+goffset]=255;
+						buf[index*ps+boffset]=255;
+					}
+				}
+				else
+				{
+					buf[index*ps+roffset]=255;
+					if(row>=halfway) buf[index*ps+goffset]=255;
+				}
+			}
+		}
+	}
+}
+
+
+#define checkval(v, cv) { \
+	if(v<cv-1 || v>cv+1) { \
+		printf("\nComp. %s at %d,%d should be %d, not %d\n",  \
+			#v, row, col, cv, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+#define checkval0(v) { \
+	if(v>1) { \
+		printf("\nComp. %s at %d,%d should be 0, not %d\n", #v, row, col, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+#define checkval255(v) { \
+	if(v<254) { \
+		printf("\nComp. %s at %d,%d should be 255, not %d\n", #v, row, col, v); \
+		retval=0;  exitStatus=-1;  goto bailout; \
+	}}
+
+
+int checkBuf(unsigned char *buf, int w, int h, int pf, int subsamp,
+	tjscalingfactor sf, int flags)
+{
+	int roffset=tjRedOffset[pf];
+	int goffset=tjGreenOffset[pf];
+	int boffset=tjBlueOffset[pf];
+	int aoffset=alphaOffset[pf];
+	int ps=tjPixelSize[pf];
+	int index, row, col, retval=1;
+	int halfway=16*sf.num/sf.denom;
+	int blocksize=8*sf.num/sf.denom;
+
+	if(pf==TJPF_CMYK)
+	{
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				unsigned char c, m, y, k;
+				if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+				else index=row*w+col;
+				c=buf[index*ps];
+				m=buf[index*ps+1];
+				y=buf[index*ps+2];
+				k=buf[index*ps+3];
+				if(((row/blocksize)+(col/blocksize))%2==0)
+				{
+					checkval255(c);  checkval255(m);  checkval255(y);
+					if(row<halfway) checkval255(k)
+					else checkval0(k)
+				}
+				else
+				{
+					checkval255(c);  checkval0(y);  checkval255(k);
+					if(row<halfway) checkval0(m)
+					else checkval255(m)
+				}
+			}
+		}
+		return 1;
+	}
+
+	for(row=0; row<h; row++)
+	{
+		for(col=0; col<w; col++)
+		{
+			unsigned char r, g, b, a;
+			if(flags&TJFLAG_BOTTOMUP) index=(h-row-1)*w+col;
+			else index=row*w+col;
+			r=buf[index*ps+roffset];
+			g=buf[index*ps+goffset];
+			b=buf[index*ps+boffset];
+			a=aoffset>=0? buf[index*ps+aoffset]:0xFF;
+			if(((row/blocksize)+(col/blocksize))%2==0)
+			{
+				if(row<halfway)
+				{
+					checkval255(r);  checkval255(g);  checkval255(b);
+				}
+				else
+				{
+					checkval0(r);  checkval0(g);  checkval0(b);
+				}
+			}
+			else
+			{
+				if(subsamp==TJSAMP_GRAY)
+				{
+					if(row<halfway)
+					{
+						checkval(r, 76);  checkval(g, 76);  checkval(b, 76);
+					}
+					else
+					{
+						checkval(r, 226);  checkval(g, 226);  checkval(b, 226);
+					}
+				}
+				else
+				{
+					if(row<halfway)
+					{
+						checkval255(r);  checkval0(g);  checkval0(b);
+					}
+					else
+					{
+						checkval255(r);  checkval255(g);  checkval0(b);
+					}
+				}
+			}
+			checkval255(a);
+		}
+	}
+
+	bailout:
+	if(retval==0)
+	{
+		for(row=0; row<h; row++)
+		{
+			for(col=0; col<w; col++)
+			{
+				if(pf==TJPF_CMYK)
+					printf("%.3d/%.3d/%.3d/%.3d ", buf[(row*w+col)*ps],
+						buf[(row*w+col)*ps+1], buf[(row*w+col)*ps+2],
+						buf[(row*w+col)*ps+3]);
+				else
+					printf("%.3d/%.3d/%.3d ", buf[(row*w+col)*ps+roffset],
+						buf[(row*w+col)*ps+goffset], buf[(row*w+col)*ps+boffset]);
+			}
+			printf("\n");
+		}
+	}
+	return retval;
+}
+
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+
+int checkBufYUV(unsigned char *buf, int w, int h, int subsamp,
+	tjscalingfactor sf)
+{
+	int row, col;
+	int hsf=tjMCUWidth[subsamp]/8, vsf=tjMCUHeight[subsamp]/8;
+	int pw=PAD(w, hsf), ph=PAD(h, vsf);
+	int cw=pw/hsf, ch=ph/vsf;
+	int ypitch=PAD(pw, pad), uvpitch=PAD(cw, pad);
+	int retval=1;
+	int halfway=16*sf.num/sf.denom;
+	int blocksize=8*sf.num/sf.denom;
+
+	for(row=0; row<ph; row++)
+	{
+		for(col=0; col<pw; col++)
+		{
+			unsigned char y=buf[ypitch*row+col];
+			if(((row/blocksize)+(col/blocksize))%2==0)
+			{
+				if(row<halfway) checkval255(y)  else checkval0(y);
+			}
+			else
+			{
+				if(row<halfway) checkval(y, 76)  else checkval(y, 226);
+			}
+		}
+	}
+	if(subsamp!=TJSAMP_GRAY)
+	{
+		int halfway=16/vsf*sf.num/sf.denom;
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+			{
+				unsigned char u=buf[ypitch*ph + (uvpitch*row+col)],
+					v=buf[ypitch*ph + uvpitch*ch + (uvpitch*row+col)];
+				if(((row*vsf/blocksize)+(col*hsf/blocksize))%2==0)
+				{
+					checkval(u, 128);  checkval(v, 128);
+				}
+				else
+				{
+					if(row<halfway)
+					{
+						checkval(u, 85);  checkval255(v);
+					}
+					else
+					{
+						checkval0(u);  checkval(v, 149);
+					}
+				}
+			}
+		}
+	}
+
+	bailout:
+	if(retval==0)
+	{
+		for(row=0; row<ph; row++)
+		{
+			for(col=0; col<pw; col++)
+				printf("%.3d ", buf[ypitch*row+col]);
+			printf("\n");
+		}
+		printf("\n");
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+				printf("%.3d ", buf[ypitch*ph + (uvpitch*row+col)]);
+			printf("\n");
+		}
+		printf("\n");
+		for(row=0; row<ch; row++)
+		{
+			for(col=0; col<cw; col++)
+				printf("%.3d ", buf[ypitch*ph + uvpitch*ch + (uvpitch*row+col)]);
+			printf("\n");
+		}
+	}
+
+	return retval;
+}
+
+
+void writeJPEG(unsigned char *jpegBuf, unsigned long jpegSize, char *filename)
+{
+	FILE *file=fopen(filename, "wb");
+	if(!file || fwrite(jpegBuf, jpegSize, 1, file)!=1)
+	{
+		printf("ERROR: Could not write to %s.\n%s\n", filename, strerror(errno));
+		bailout();
+	}
+
+	bailout:
+	if(file) fclose(file);
+}
+
+
+void compTest(tjhandle handle, unsigned char **dstBuf,
+	unsigned long *dstSize, int w, int h, int pf, char *basename,
+	int subsamp, int jpegQual, int flags)
+{
+	char tempStr[1024];  unsigned char *srcBuf=NULL, *yuvBuf=NULL;
+	const char *pfStr=pixFormatStr[pf];
+	const char *buStrLong=(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ";
+	const char *buStr=(flags&TJFLAG_BOTTOMUP)? "BU":"TD";
+
+	if((srcBuf=(unsigned char *)malloc(w*h*tjPixelSize[pf]))==NULL)
+		_throw("Memory allocation failure");
+	initBuf(srcBuf, w, h, pf, flags);
+
+	if(*dstBuf && *dstSize>0) memset(*dstBuf, 0, *dstSize);
+
+
+	if(!alloc) flags|=TJFLAG_NOREALLOC;
+	if(doyuv)
+	{
+		unsigned long yuvSize=tjBufSizeYUV2(w, pad, h, subsamp);
+		tjscalingfactor sf={1, 1};
+		tjhandle handle2=tjInitCompress();
+		if(!handle2) _throwtj();
+
+		if((yuvBuf=(unsigned char *)malloc(yuvSize))==NULL)
+			_throw("Memory allocation failure");
+		memset(yuvBuf, 0, yuvSize);
+
+		printf("%s %s -> YUV %s ... ", pfStr, buStrLong, subNameLong[subsamp]);
+		_tj(tjEncodeYUV3(handle2, srcBuf, w, 0, h, pf, yuvBuf, pad, subsamp,
+			flags));
+		tjDestroy(handle2);
+		if(checkBufYUV(yuvBuf, w, h, subsamp, sf)) printf("Passed.\n");
+		else printf("FAILED!\n");
+
+		printf("YUV %s %s -> JPEG Q%d ... ", subNameLong[subsamp], buStrLong,
+			jpegQual);
+		_tj(tjCompressFromYUV(handle, yuvBuf, w, pad, h, subsamp, dstBuf,
+			dstSize, jpegQual, flags));
+	}
+	else
+	{
+		printf("%s %s -> %s Q%d ... ", pfStr, buStrLong, subNameLong[subsamp],
+			jpegQual);
+		_tj(tjCompress2(handle, srcBuf, w, 0, h, pf, dstBuf, dstSize, subsamp,
+			jpegQual, flags));
+	}
+
+	snprintf(tempStr, 1024, "%s_enc_%s_%s_%s_Q%d.jpg", basename, pfStr, buStr,
+		subName[subsamp], jpegQual);
+	writeJPEG(*dstBuf, *dstSize, tempStr);
+	printf("Done.\n  Result in %s\n", tempStr);
+
+	bailout:
+	if(yuvBuf) free(yuvBuf);
+	if(srcBuf) free(srcBuf);
+}
+
+
+void _decompTest(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, int w, int h, int pf, char *basename, int subsamp,
+	int flags, tjscalingfactor sf)
+{
+	unsigned char *dstBuf=NULL, *yuvBuf=NULL;
+	int _hdrw=0, _hdrh=0, _hdrsubsamp=-1;
+	int scaledWidth=TJSCALED(w, sf);
+	int scaledHeight=TJSCALED(h, sf);
+	unsigned long dstSize=0;
+
+	_tj(tjDecompressHeader2(handle, jpegBuf, jpegSize, &_hdrw, &_hdrh,
+		&_hdrsubsamp));
+	if(_hdrw!=w || _hdrh!=h || _hdrsubsamp!=subsamp)
+		_throw("Incorrect JPEG header");
+
+	dstSize=scaledWidth*scaledHeight*tjPixelSize[pf];
+	if((dstBuf=(unsigned char *)malloc(dstSize))==NULL)
+		_throw("Memory allocation failure");
+	memset(dstBuf, 0, dstSize);
+
+	if(doyuv)
+	{
+		unsigned long yuvSize=tjBufSizeYUV2(scaledWidth, pad, scaledHeight,
+			subsamp);
+		tjhandle handle2=tjInitDecompress();
+		if(!handle2) _throwtj();
+
+		if((yuvBuf=(unsigned char *)malloc(yuvSize))==NULL)
+			_throw("Memory allocation failure");
+		memset(yuvBuf, 0, yuvSize);
+
+		printf("JPEG -> YUV %s ", subNameLong[subsamp]);
+		if(sf.num!=1 || sf.denom!=1)
+			printf("%d/%d ... ", sf.num, sf.denom);
+		else printf("... ");
+		_tj(tjDecompressToYUV2(handle, jpegBuf, jpegSize, yuvBuf, scaledWidth,
+			pad, scaledHeight, flags));
+		if(checkBufYUV(yuvBuf, scaledWidth, scaledHeight, subsamp, sf))
+			printf("Passed.\n");
+		else printf("FAILED!\n");
+
+		printf("YUV %s -> %s %s ... ", subNameLong[subsamp], pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ");
+		_tj(tjDecodeYUV(handle2, yuvBuf, pad, subsamp, dstBuf, scaledWidth, 0,
+			scaledHeight, pf, flags));
+		tjDestroy(handle2);
+	}
+	else
+	{
+		printf("JPEG -> %s %s ", pixFormatStr[pf],
+			(flags&TJFLAG_BOTTOMUP)? "Bottom-Up":"Top-Down ");
+		if(sf.num!=1 || sf.denom!=1)
+			printf("%d/%d ... ", sf.num, sf.denom);
+		else printf("... ");
+		_tj(tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, scaledWidth, 0,
+			scaledHeight, pf, flags));
+	}
+
+	if(checkBuf(dstBuf, scaledWidth, scaledHeight, pf, subsamp, sf, flags))
+		printf("Passed.");
+	else printf("FAILED!");
+	printf("\n");
+
+	bailout:
+	if(yuvBuf) free(yuvBuf);
+	if(dstBuf) free(dstBuf);
+}
+
+
+void decompTest(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, int w, int h, int pf, char *basename, int subsamp,
+	int flags)
+{
+	int i, n=0;
+	tjscalingfactor *sf=tjGetScalingFactors(&n);
+	if(!sf || !n) _throwtj();
+
+	for(i=0; i<n; i++)
+	{
+		if(subsamp==TJSAMP_444 || subsamp==TJSAMP_GRAY ||
+			(subsamp==TJSAMP_411 && sf[i].num==1 &&
+				(sf[i].denom==2 || sf[i].denom==1)) ||
+			(subsamp!=TJSAMP_411 && sf[i].num==1 &&
+				(sf[i].denom==4 || sf[i].denom==2 || sf[i].denom==1)))
+			_decompTest(handle, jpegBuf, jpegSize, w, h, pf, basename, subsamp,
+				flags, sf[i]);
+	}
+
+	bailout:
+	return;
+}
+
+
+void doTest(int w, int h, const int *formats, int nformats, int subsamp,
+	char *basename)
+{
+	tjhandle chandle=NULL, dhandle=NULL;
+	unsigned char *dstBuf=NULL;
+	unsigned long size=0;  int pfi, pf, i;
+
+	if(!alloc)
+		size=tjBufSize(w, h, subsamp);
+	if(size!=0)
+		if((dstBuf=(unsigned char *)tjAlloc(size))==NULL)
+			_throw("Memory allocation failure.");
+
+	if((chandle=tjInitCompress())==NULL || (dhandle=tjInitDecompress())==NULL)
+		_throwtj();
+
+	for(pfi=0; pfi<nformats; pfi++)
+	{
+		for(i=0; i<2; i++)
+		{
+			int flags=0;
+			if(subsamp==TJSAMP_422 || subsamp==TJSAMP_420 || subsamp==TJSAMP_440 ||
+				subsamp==TJSAMP_411)
+				flags|=TJFLAG_FASTUPSAMPLE;
+			if(i==1) flags|=TJFLAG_BOTTOMUP;
+			pf=formats[pfi];
+			compTest(chandle, &dstBuf, &size, w, h, pf, basename, subsamp, 100,
+				flags);
+			decompTest(dhandle, dstBuf, size, w, h, pf, basename, subsamp,
+				flags);
+			if(pf>=TJPF_RGBX && pf<=TJPF_XRGB)
+			{
+				printf("\n");
+				decompTest(dhandle, dstBuf, size, w, h, pf+(TJPF_RGBA-TJPF_RGBX),
+					basename, subsamp, flags);
+			}
+			printf("\n");
+		}
+	}
+	printf("--------------------\n\n");
+
+	bailout:
+	if(chandle) tjDestroy(chandle);
+	if(dhandle) tjDestroy(dhandle);
+
+	if(dstBuf) tjFree(dstBuf);
+}
+
+
+void bufSizeTest(void)
+{
+	int w, h, i, subsamp;
+	unsigned char *srcBuf=NULL, *dstBuf=NULL;
+	tjhandle handle=NULL;
+	unsigned long dstSize=0;
+
+	if((handle=tjInitCompress())==NULL) _throwtj();
+
+	printf("Buffer size regression test\n");
+	for(subsamp=0; subsamp<TJ_NUMSAMP; subsamp++)
+	{
+		for(w=1; w<48; w++)
+		{
+			int maxh=(w==1)? 2048:48;
+			for(h=1; h<maxh; h++)
+			{
+				if(h%100==0) printf("%.4d x %.4d\b\b\b\b\b\b\b\b\b\b\b", w, h);
+				if((srcBuf=(unsigned char *)malloc(w*h*4))==NULL)
+					_throw("Memory allocation failure");
+				if(!alloc || doyuv)
+				{
+					if(doyuv) dstSize=tjBufSizeYUV2(w, pad, h, subsamp);
+					else dstSize=tjBufSize(w, h, subsamp);
+					if((dstBuf=(unsigned char *)tjAlloc(dstSize))==NULL)
+						_throw("Memory allocation failure");
+				}
+
+				for(i=0; i<w*h*4; i++)
+				{
+					if(random()<RAND_MAX/2) srcBuf[i]=0;
+					else srcBuf[i]=255;
+				}
+
+				if(doyuv)
+				{
+					_tj(tjEncodeYUV3(handle, srcBuf, w, 0, h, TJPF_BGRX, dstBuf, pad,
+						subsamp, 0));
+				}
+				else
+				{
+					_tj(tjCompress2(handle, srcBuf, w, 0, h, TJPF_BGRX, &dstBuf,
+						&dstSize, subsamp, 100, alloc? 0:TJFLAG_NOREALLOC));
+				}
+				free(srcBuf);  srcBuf=NULL;
+				if(!alloc || doyuv)
+				{
+					tjFree(dstBuf);  dstBuf=NULL;
+				}
+
+				if((srcBuf=(unsigned char *)malloc(h*w*4))==NULL)
+					_throw("Memory allocation failure");
+				if(!alloc || doyuv)
+				{
+					if(doyuv) dstSize=tjBufSizeYUV2(h, pad, w, subsamp);
+					else dstSize=tjBufSize(h, w, subsamp);
+					if((dstBuf=(unsigned char *)tjAlloc(dstSize))==NULL)
+						_throw("Memory allocation failure");
+				}
+
+				for(i=0; i<h*w*4; i++)
+				{
+					if(random()<RAND_MAX/2) srcBuf[i]=0;
+					else srcBuf[i]=255;
+				}
+
+				if(doyuv)
+				{
+					_tj(tjEncodeYUV3(handle, srcBuf, h, 0, w, TJPF_BGRX, dstBuf, pad,
+						subsamp, 0));
+				}
+				else
+				{
+					_tj(tjCompress2(handle, srcBuf, h, 0, w, TJPF_BGRX, &dstBuf,
+						&dstSize, subsamp, 100, alloc? 0:TJFLAG_NOREALLOC));
+				}
+				free(srcBuf);  srcBuf=NULL;
+				if(!alloc || doyuv)
+				{
+					tjFree(dstBuf);  dstBuf=NULL;
+				}
+			}
+		}
+	}
+	printf("Done.      \n");
+
+	bailout:
+	if(srcBuf) free(srcBuf);
+	if(dstBuf) tjFree(dstBuf);
+	if(handle) tjDestroy(handle);
+}
+
+
+int main(int argc, char *argv[])
+{
+	int i, num4bf=5;
+	#ifdef _WIN32
+	srand((unsigned int)time(NULL));
+	#endif
+	if(argc>1)
+	{
+		for(i=1; i<argc; i++)
+		{
+			if(!strcasecmp(argv[i], "-yuv")) doyuv=1;
+			if(!strcasecmp(argv[i], "-noyuvpad")) pad=1;
+			if(!strcasecmp(argv[i], "-alloc")) alloc=1;
+			if(!strncasecmp(argv[i], "-h", 2) || !strcasecmp(argv[i], "-?"))
+				usage(argv[0]);
+		}
+	}
+	if(alloc) printf("Testing automatic buffer allocation\n");
+	if(doyuv) num4bf=4;
+	doTest(35, 39, _3byteFormats, 2, TJSAMP_444, "test");
+	doTest(39, 41, _4byteFormats, num4bf, TJSAMP_444, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_422, "test");
+	doTest(35, 39, _4byteFormats, num4bf, TJSAMP_422, "test");
+	doTest(39, 41, _3byteFormats, 2, TJSAMP_420, "test");
+	doTest(41, 35, _4byteFormats, num4bf, TJSAMP_420, "test");
+	doTest(35, 39, _3byteFormats, 2, TJSAMP_440, "test");
+	doTest(39, 41, _4byteFormats, num4bf, TJSAMP_440, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_411, "test");
+	doTest(35, 39, _4byteFormats, num4bf, TJSAMP_411, "test");
+	doTest(39, 41, _onlyGray, 1, TJSAMP_GRAY, "test");
+	doTest(41, 35, _3byteFormats, 2, TJSAMP_GRAY, "test");
+	doTest(35, 39, _4byteFormats, 4, TJSAMP_GRAY, "test");
+	bufSizeTest();
+	if(doyuv)
+	{
+		printf("\n--------------------\n\n");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_444, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_422, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_420, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_440, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_411, "test_yuv0");
+		doTest(48, 48, _onlyRGB, 1, TJSAMP_GRAY, "test_yuv0");
+		doTest(48, 48, _onlyGray, 1, TJSAMP_GRAY, "test_yuv0");
+	}
+
+	return exitStatus;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjutil.c glmark2-2021.02/src/libjpeg-turbo/tjutil.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjutil.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjutil.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,66 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+
+#include <windows.h>
+
+static double getfreq(void)
+{
+	LARGE_INTEGER freq;
+	if(!QueryPerformanceFrequency(&freq)) return 0.0;
+	return (double)freq.QuadPart;
+}
+
+static double f=-1.0;
+
+double gettime(void)
+{
+	LARGE_INTEGER t;
+	if(f<0.0) f=getfreq();
+	if(f==0.0) return (double)GetTickCount()/1000.;
+	else
+	{
+		QueryPerformanceCounter(&t);
+		return (double)t.QuadPart/f;
+	}
+}
+
+#else
+
+#include <stdlib.h>
+#include <sys/time.h>
+
+double gettime(void)
+{
+	struct timeval tv;
+	if(gettimeofday(&tv, NULL)<0) return 0.0;
+	else return (double)tv.tv_sec+((double)tv.tv_usec/1000000.);
+}
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjutil.h glmark2-2021.02/src/libjpeg-turbo/tjutil.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/tjutil.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/tjutil.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,47 @@
+/*
+ * Copyright (C)2011 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifdef _WIN32
+	#ifndef __MINGW32__
+		#include <stdio.h>
+		#define snprintf(str, n, format, ...)  \
+			_snprintf_s(str, n, _TRUNCATE, format, __VA_ARGS__)
+	#endif
+	#define strcasecmp stricmp
+	#define strncasecmp strnicmp
+#endif
+
+#ifndef min
+ #define min(a,b) ((a)<(b)?(a):(b))
+#endif
+
+#ifndef max
+ #define max(a,b) ((a)>(b)?(a):(b))
+#endif
+
+extern double gettime(void);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/transupp.c glmark2-2021.02/src/libjpeg-turbo/transupp.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/transupp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/transupp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1621 @@
+/*
+ * transupp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2011, Thomas G. Lane, Guido Vollbeding.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2010, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains image transformation routines and other utility code
+ * used by the jpegtran sample application.  These are NOT part of the core
+ * JPEG library.  But we keep these routines separate from jpegtran.c to
+ * ease the task of maintaining jpegtran-like programs that have other user
+ * interfaces.
+ */
+
+/* Although this file really shouldn't have access to the library internals,
+ * it's helpful to let it call jround_up() and jcopy_block_row().
+ */
+#define JPEG_INTERNALS
+
+#include "jinclude.h"
+#include "jpeglib.h"
+#include "transupp.h"           /* My own external interface */
+#include "jpegcomp.h"
+#include <ctype.h>              /* to declare isdigit() */
+
+
+#if JPEG_LIB_VERSION >= 70
+#define dstinfo_min_DCT_h_scaled_size dstinfo->min_DCT_h_scaled_size
+#define dstinfo_min_DCT_v_scaled_size dstinfo->min_DCT_v_scaled_size
+#else
+#define dstinfo_min_DCT_h_scaled_size DCTSIZE
+#define dstinfo_min_DCT_v_scaled_size DCTSIZE
+#endif
+
+
+#if TRANSFORMS_SUPPORTED
+
+/*
+ * Lossless image transformation routines.  These routines work on DCT
+ * coefficient arrays and thus do not require any lossy decompression
+ * or recompression of the image.
+ * Thanks to Guido Vollbeding for the initial design and code of this feature,
+ * and to Ben Jackson for introducing the cropping feature.
+ *
+ * Horizontal flipping is done in-place, using a single top-to-bottom
+ * pass through the virtual source array.  It will thus be much the
+ * fastest option for images larger than main memory.
+ *
+ * The other routines require a set of destination virtual arrays, so they
+ * need twice as much memory as jpegtran normally does.  The destination
+ * arrays are always written in normal scan order (top to bottom) because
+ * the virtual array manager expects this.  The source arrays will be scanned
+ * in the corresponding order, which means multiple passes through the source
+ * arrays for most of the transforms.  That could result in much thrashing
+ * if the image is larger than main memory.
+ *
+ * If cropping or trimming is involved, the destination arrays may be smaller
+ * than the source arrays.  Note it is not possible to do horizontal flip
+ * in-place when a nonzero Y crop offset is specified, since we'd have to move
+ * data from one block row to another but the virtual array manager doesn't
+ * guarantee we can touch more than one row at a time.  So in that case,
+ * we have to use a separate destination array.
+ *
+ * Some notes about the operating environment of the individual transform
+ * routines:
+ * 1. Both the source and destination virtual arrays are allocated from the
+ *    source JPEG object, and therefore should be manipulated by calling the
+ *    source's memory manager.
+ * 2. The destination's component count should be used.  It may be smaller
+ *    than the source's when forcing to grayscale.
+ * 3. Likewise the destination's sampling factors should be used.  When
+ *    forcing to grayscale the destination's sampling factors will be all 1,
+ *    and we may as well take that as the effective iMCU size.
+ * 4. When "trim" is in effect, the destination's dimensions will be the
+ *    trimmed values but the source's will be untrimmed.
+ * 5. When "crop" is in effect, the destination's dimensions will be the
+ *    cropped values but the source's will be uncropped.  Each transform
+ *    routine is responsible for picking up source data starting at the
+ *    correct X and Y offset for the crop region.  (The X and Y offsets
+ *    passed to the transform routines are measured in iMCU blocks of the
+ *    destination.)
+ * 6. All the routines assume that the source and destination buffers are
+ *    padded out to a full iMCU boundary.  This is true, although for the
+ *    source buffer it is an undocumented property of jdcoefct.c.
+ */
+
+
+LOCAL(void)
+do_crop (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+         jvirt_barray_ptr *src_coef_arrays,
+         jvirt_barray_ptr *dst_coef_arrays)
+/* Crop.  This is only used when no rotate/flip is requested with the crop. */
+{
+  JDIMENSION dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  jpeg_component_info *compptr;
+
+  /* We simply have to copy the right amount of data (the destination's
+   * image size) starting at the given X and Y offsets in the source.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+         dst_blk_y + y_crop_blocks,
+         (JDIMENSION) compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        jcopy_block_row(src_buffer[offset_y] + x_crop_blocks,
+                        dst_buffer[offset_y],
+                        compptr->width_in_blocks);
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h_no_crop (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                   JDIMENSION x_crop_offset,
+                   jvirt_barray_ptr *src_coef_arrays)
+/* Horizontal flip; done in-place, so no separate dest array is required.
+ * NB: this only works when y_crop_offset is zero.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, blk_x, blk_y, x_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY buffer;
+  JCOEFPTR ptr1, ptr2;
+  JCOEF temp1, temp2;
+  jpeg_component_info *compptr;
+
+  /* Horizontal mirroring of DCT blocks is accomplished by swapping
+   * pairs of blocks in-place.  Within a DCT block, we perform horizontal
+   * mirroring by changing the signs of odd-numbered columns.
+   * Partial iMCUs at the right edge are left untouched.
+   */
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    for (blk_y = 0; blk_y < compptr->height_in_blocks;
+         blk_y += compptr->v_samp_factor) {
+      buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci], blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        /* Do the mirroring */
+        for (blk_x = 0; blk_x * 2 < comp_width; blk_x++) {
+          ptr1 = buffer[offset_y][blk_x];
+          ptr2 = buffer[offset_y][comp_width - blk_x - 1];
+          /* this unrolled loop doesn't need to know which row it's on... */
+          for (k = 0; k < DCTSIZE2; k += 2) {
+            temp1 = *ptr1;      /* swap even column */
+            temp2 = *ptr2;
+            *ptr1++ = temp2;
+            *ptr2++ = temp1;
+            temp1 = *ptr1;      /* swap odd column with sign change */
+            temp2 = *ptr2;
+            *ptr1++ = -temp2;
+            *ptr2++ = -temp1;
+          }
+        }
+        if (x_crop_blocks > 0) {
+          /* Now left-justify the portion of the data to be kept.
+           * We can't use a single jcopy_block_row() call because that routine
+           * depends on memcpy(), whose behavior is unspecified for overlapping
+           * source and destination areas.  Sigh.
+           */
+          for (blk_x = 0; blk_x < compptr->width_in_blocks; blk_x++) {
+            jcopy_block_row(buffer[offset_y] + blk_x + x_crop_blocks,
+                            buffer[offset_y] + blk_x,
+                            (JDIMENSION) 1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_h (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* Horizontal flip in general cropping case */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, k, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Here we must output into a separate array because we can't touch
+   * different rows of a single virtual array simultaneously.  Otherwise,
+   * this is essentially the same as the routine above.
+   */
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      src_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+         dst_blk_y + y_crop_blocks,
+         (JDIMENSION) compptr->v_samp_factor, FALSE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_buffer[offset_y];
+        src_row_ptr = src_buffer[offset_y];
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Do the mirrorable blocks */
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+            /* this unrolled loop doesn't need to know which row it's on... */
+            for (k = 0; k < DCTSIZE2; k += 2) {
+              *dst_ptr++ = *src_ptr++;   /* copy even column */
+              *dst_ptr++ = - *src_ptr++; /* copy odd column with sign change */
+            }
+          } else {
+            /* Copy last partial block(s) verbatim */
+            jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                            dst_row_ptr + dst_blk_x,
+                            (JDIMENSION) 1);
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_flip_v (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* Vertical flip */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* We output into a separate array because we can't touch different
+   * rows of the source virtual array simultaneously.  Otherwise, this
+   * is a pretty straightforward analog of horizontal flip.
+   * Within a DCT block, vertical mirroring is done by changing the signs
+   * of odd-numbered rows.
+   * Partial iMCUs at the bottom edge are copied verbatim.
+   */
+  MCU_rows = srcinfo->output_height /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION) compptr->v_samp_factor,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge blocks will be copied verbatim. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          dst_row_ptr = dst_buffer[offset_y];
+          src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+          src_row_ptr += x_crop_blocks;
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+               dst_blk_x++) {
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            src_ptr = src_row_ptr[dst_blk_x];
+            for (i = 0; i < DCTSIZE; i += 2) {
+              /* copy even row */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = *src_ptr++;
+              /* copy odd row with sign change */
+              for (j = 0; j < DCTSIZE; j++)
+                *dst_ptr++ = - *src_ptr++;
+            }
+          }
+        } else {
+          /* Just copy row verbatim. */
+          jcopy_block_row(src_buffer[offset_y] + x_crop_blocks,
+                          dst_buffer[offset_y],
+                          compptr->width_in_blocks);
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transpose (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+              JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+              jvirt_barray_ptr *src_coef_arrays,
+              jvirt_barray_ptr *dst_coef_arrays)
+/* Transpose source into destination */
+{
+  JDIMENSION dst_blk_x, dst_blk_y, x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Transposing pixels within a block just requires transposing the
+   * DCT coefficients.
+   * Partial iMCUs at the edges require no special treatment; we simply
+   * process all the available DCT blocks for every component.
+   */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION) compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            src_ptr = src_buffer[offset_x][dst_blk_y + offset_y + y_crop_blocks];
+            for (i = 0; i < DCTSIZE; i++)
+              for (j = 0; j < DCTSIZE; j++)
+                dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_90 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+           JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+           jvirt_barray_ptr *src_coef_arrays,
+           jvirt_barray_ptr *dst_coef_arrays)
+/* 90 degree rotation is equivalent to
+ *   1. Transposing the image;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, comp_width, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) right edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_cols = srcinfo->output_height /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION) compptr->h_samp_factor,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          } else {
+            /* Edge blocks are transposed but not mirrored. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                i++;
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_buffer[offset_x]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_270 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+            JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+            jvirt_barray_ptr *src_coef_arrays,
+            jvirt_barray_ptr *dst_coef_arrays)
+/* 270 degree rotation is equivalent to
+ *   1. Horizontal mirroring;
+ *   2. Transposing the image.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_rows, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  /* Because of the horizontal mirror step, we can't process partial iMCUs
+   * at the (output) bottom edge properly.  They just get transposed and
+   * not mirrored.
+   */
+  MCU_rows = srcinfo->output_width /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          src_buffer = (*srcinfo->mem->access_virt_barray)
+            ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+             dst_blk_x + x_crop_blocks,
+             (JDIMENSION) compptr->h_samp_factor, FALSE);
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              /* Block is within the mirrorable area. */
+              src_ptr = src_buffer[offset_x]
+                [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+              for (i = 0; i < DCTSIZE; i++) {
+                for (j = 0; j < DCTSIZE; j++) {
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  j++;
+                  dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                }
+              }
+            } else {
+              /* Edge blocks are transposed but not mirrored. */
+              src_ptr = src_buffer[offset_x]
+                [dst_blk_y + offset_y + y_crop_blocks];
+              for (i = 0; i < DCTSIZE; i++)
+                for (j = 0; j < DCTSIZE; j++)
+                  dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_rot_180 (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+            JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+            jvirt_barray_ptr *src_coef_arrays,
+            jvirt_barray_ptr *dst_coef_arrays)
+/* 180 degree rotation is equivalent to
+ *   1. Vertical mirroring;
+ *   2. Horizontal mirroring.
+ * These two steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JBLOCKROW src_row_ptr, dst_row_ptr;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_width /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_height /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      if (y_crop_blocks + dst_blk_y < comp_height) {
+        /* Row is within the vertically mirrorable area. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           comp_height - y_crop_blocks - dst_blk_y -
+           (JDIMENSION) compptr->v_samp_factor,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      } else {
+        /* Bottom-edge rows are only mirrored horizontally. */
+        src_buffer = (*srcinfo->mem->access_virt_barray)
+          ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+           dst_blk_y + y_crop_blocks,
+           (JDIMENSION) compptr->v_samp_factor, FALSE);
+      }
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        dst_row_ptr = dst_buffer[offset_y];
+        if (y_crop_blocks + dst_blk_y < comp_height) {
+          /* Row is within the mirrorable area. */
+          src_row_ptr = src_buffer[compptr->v_samp_factor - offset_y - 1];
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+            dst_ptr = dst_row_ptr[dst_blk_x];
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored both ways. */
+              src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+              for (i = 0; i < DCTSIZE; i += 2) {
+                /* For even row, negate every odd column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = *src_ptr++;
+                  *dst_ptr++ = - *src_ptr++;
+                }
+                /* For odd row, negate every even column. */
+                for (j = 0; j < DCTSIZE; j += 2) {
+                  *dst_ptr++ = - *src_ptr++;
+                  *dst_ptr++ = *src_ptr++;
+                }
+              }
+            } else {
+              /* Any remaining right-edge blocks are only mirrored vertically. */
+              src_ptr = src_row_ptr[x_crop_blocks + dst_blk_x];
+              for (i = 0; i < DCTSIZE; i += 2) {
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = *src_ptr++;
+                for (j = 0; j < DCTSIZE; j++)
+                  *dst_ptr++ = - *src_ptr++;
+              }
+            }
+          }
+        } else {
+          /* Remaining rows are just mirrored horizontally. */
+          src_row_ptr = src_buffer[offset_y];
+          for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks; dst_blk_x++) {
+            if (x_crop_blocks + dst_blk_x < comp_width) {
+              /* Process the blocks that can be mirrored. */
+              dst_ptr = dst_row_ptr[dst_blk_x];
+              src_ptr = src_row_ptr[comp_width - x_crop_blocks - dst_blk_x - 1];
+              for (i = 0; i < DCTSIZE2; i += 2) {
+                *dst_ptr++ = *src_ptr++;
+                *dst_ptr++ = - *src_ptr++;
+              }
+            } else {
+              /* Any remaining right-edge blocks are only copied. */
+              jcopy_block_row(src_row_ptr + dst_blk_x + x_crop_blocks,
+                              dst_row_ptr + dst_blk_x,
+                              (JDIMENSION) 1);
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+LOCAL(void)
+do_transverse (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+               JDIMENSION x_crop_offset, JDIMENSION y_crop_offset,
+               jvirt_barray_ptr *src_coef_arrays,
+               jvirt_barray_ptr *dst_coef_arrays)
+/* Transverse transpose is equivalent to
+ *   1. 180 degree rotation;
+ *   2. Transposition;
+ * or
+ *   1. Horizontal mirroring;
+ *   2. Transposition;
+ *   3. Horizontal mirroring.
+ * These steps are merged into a single processing routine.
+ */
+{
+  JDIMENSION MCU_cols, MCU_rows, comp_width, comp_height, dst_blk_x, dst_blk_y;
+  JDIMENSION x_crop_blocks, y_crop_blocks;
+  int ci, i, j, offset_x, offset_y;
+  JBLOCKARRAY src_buffer, dst_buffer;
+  JCOEFPTR src_ptr, dst_ptr;
+  jpeg_component_info *compptr;
+
+  MCU_cols = srcinfo->output_height /
+    (dstinfo->max_h_samp_factor * dstinfo_min_DCT_h_scaled_size);
+  MCU_rows = srcinfo->output_width /
+    (dstinfo->max_v_samp_factor * dstinfo_min_DCT_v_scaled_size);
+
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    comp_width = MCU_cols * compptr->h_samp_factor;
+    comp_height = MCU_rows * compptr->v_samp_factor;
+    x_crop_blocks = x_crop_offset * compptr->h_samp_factor;
+    y_crop_blocks = y_crop_offset * compptr->v_samp_factor;
+    for (dst_blk_y = 0; dst_blk_y < compptr->height_in_blocks;
+         dst_blk_y += compptr->v_samp_factor) {
+      dst_buffer = (*srcinfo->mem->access_virt_barray)
+        ((j_common_ptr) srcinfo, dst_coef_arrays[ci], dst_blk_y,
+         (JDIMENSION) compptr->v_samp_factor, TRUE);
+      for (offset_y = 0; offset_y < compptr->v_samp_factor; offset_y++) {
+        for (dst_blk_x = 0; dst_blk_x < compptr->width_in_blocks;
+             dst_blk_x += compptr->h_samp_factor) {
+          if (x_crop_blocks + dst_blk_x < comp_width) {
+            /* Block is within the mirrorable area. */
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               comp_width - x_crop_blocks - dst_blk_x -
+               (JDIMENSION) compptr->h_samp_factor,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          } else {
+            src_buffer = (*srcinfo->mem->access_virt_barray)
+              ((j_common_ptr) srcinfo, src_coef_arrays[ci],
+               dst_blk_x + x_crop_blocks,
+               (JDIMENSION) compptr->h_samp_factor, FALSE);
+          }
+          for (offset_x = 0; offset_x < compptr->h_samp_factor; offset_x++) {
+            dst_ptr = dst_buffer[offset_y][dst_blk_x + offset_x];
+            if (y_crop_blocks + dst_blk_y < comp_height) {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Block is within the mirrorable area. */
+                src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                  [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                  }
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  }
+                }
+              } else {
+                /* Right-edge blocks are mirrored in y only */
+                src_ptr = src_buffer[offset_x]
+                  [comp_height - y_crop_blocks - dst_blk_y - offset_y - 1];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++) {
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                    j++;
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                  }
+                }
+              }
+            } else {
+              if (x_crop_blocks + dst_blk_x < comp_width) {
+                /* Bottom-edge blocks are mirrored in x only */
+                src_ptr = src_buffer[compptr->h_samp_factor - offset_x - 1]
+                  [dst_blk_y + offset_y + y_crop_blocks];
+                for (i = 0; i < DCTSIZE; i++) {
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+                  i++;
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = -src_ptr[i*DCTSIZE+j];
+                }
+              } else {
+                /* At lower right corner, just transpose, no mirroring */
+                src_ptr = src_buffer[offset_x]
+                  [dst_blk_y + offset_y + y_crop_blocks];
+                for (i = 0; i < DCTSIZE; i++)
+                  for (j = 0; j < DCTSIZE; j++)
+                    dst_ptr[j*DCTSIZE+i] = src_ptr[i*DCTSIZE+j];
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
+
+/* Parse an unsigned integer: subroutine for jtransform_parse_crop_spec.
+ * Returns TRUE if valid integer found, FALSE if not.
+ * *strptr is advanced over the digit string, and *result is set to its value.
+ */
+
+LOCAL(boolean)
+jt_read_integer (const char **strptr, JDIMENSION *result)
+{
+  const char *ptr = *strptr;
+  JDIMENSION val = 0;
+
+  for (; isdigit(*ptr); ptr++) {
+    val = val * 10 + (JDIMENSION) (*ptr - '0');
+  }
+  *result = val;
+  if (ptr == *strptr)
+    return FALSE;               /* oops, no digits */
+  *strptr = ptr;
+  return TRUE;
+}
+
+
+/* Parse a crop specification (written in X11 geometry style).
+ * The routine returns TRUE if the spec string is valid, FALSE if not.
+ *
+ * The crop spec string should have the format
+ *      <width>[f]x<height>[f]{+-}<xoffset>{+-}<yoffset>
+ * where width, height, xoffset, and yoffset are unsigned integers.
+ * Each of the elements can be omitted to indicate a default value.
+ * (A weakness of this style is that it is not possible to omit xoffset
+ * while specifying yoffset, since they look alike.)
+ *
+ * This code is loosely based on XParseGeometry from the X11 distribution.
+ */
+
+GLOBAL(boolean)
+jtransform_parse_crop_spec (jpeg_transform_info *info, const char *spec)
+{
+  info->crop = FALSE;
+  info->crop_width_set = JCROP_UNSET;
+  info->crop_height_set = JCROP_UNSET;
+  info->crop_xoffset_set = JCROP_UNSET;
+  info->crop_yoffset_set = JCROP_UNSET;
+
+  if (isdigit(*spec)) {
+    /* fetch width */
+    if (! jt_read_integer(&spec, &info->crop_width))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_width_set = JCROP_FORCE;
+    } else
+      info->crop_width_set = JCROP_POS;
+  }
+  if (*spec == 'x' || *spec == 'X') {
+    /* fetch height */
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_height))
+      return FALSE;
+    if (*spec == 'f' || *spec == 'F') {
+      spec++;
+      info->crop_height_set = JCROP_FORCE;
+    } else
+      info->crop_height_set = JCROP_POS;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch xoffset */
+    info->crop_xoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_xoffset))
+      return FALSE;
+  }
+  if (*spec == '+' || *spec == '-') {
+    /* fetch yoffset */
+    info->crop_yoffset_set = (*spec == '-') ? JCROP_NEG : JCROP_POS;
+    spec++;
+    if (! jt_read_integer(&spec, &info->crop_yoffset))
+      return FALSE;
+  }
+  /* We had better have gotten to the end of the string. */
+  if (*spec != '\0')
+    return FALSE;
+  info->crop = TRUE;
+  return TRUE;
+}
+
+
+/* Trim off any partial iMCUs on the indicated destination edge */
+
+LOCAL(void)
+trim_right_edge (jpeg_transform_info *info, JDIMENSION full_width)
+{
+  JDIMENSION MCU_cols;
+
+  MCU_cols = info->output_width / info->iMCU_sample_width;
+  if (MCU_cols > 0 && info->x_crop_offset + MCU_cols ==
+      full_width / info->iMCU_sample_width)
+    info->output_width = MCU_cols * info->iMCU_sample_width;
+}
+
+LOCAL(void)
+trim_bottom_edge (jpeg_transform_info *info, JDIMENSION full_height)
+{
+  JDIMENSION MCU_rows;
+
+  MCU_rows = info->output_height / info->iMCU_sample_height;
+  if (MCU_rows > 0 && info->y_crop_offset + MCU_rows ==
+      full_height / info->iMCU_sample_height)
+    info->output_height = MCU_rows * info->iMCU_sample_height;
+}
+
+
+/* Request any required workspace.
+ *
+ * This routine figures out the size that the output image will be
+ * (which implies that all the transform parameters must be set before
+ * it is called).
+ *
+ * We allocate the workspace virtual arrays from the source decompression
+ * object, so that all the arrays (both the original data and the workspace)
+ * will be taken into account while making memory management decisions.
+ * Hence, this routine must be called after jpeg_read_header (which reads
+ * the image dimensions) and before jpeg_read_coefficients (which realizes
+ * the source's virtual arrays).
+ *
+ * This function returns FALSE right away if -perfect is given
+ * and transformation is not perfect.  Otherwise returns TRUE.
+ */
+
+GLOBAL(boolean)
+jtransform_request_workspace (j_decompress_ptr srcinfo,
+                              jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *coef_arrays;
+  boolean need_workspace, transpose_it;
+  jpeg_component_info *compptr;
+  JDIMENSION xoffset, yoffset;
+  JDIMENSION width_in_iMCUs, height_in_iMCUs;
+  JDIMENSION width_in_blocks, height_in_blocks;
+  int ci, h_samp_factor, v_samp_factor;
+
+  /* Determine number of components in output image */
+  if (info->force_grayscale &&
+      srcinfo->jpeg_color_space == JCS_YCbCr &&
+      srcinfo->num_components == 3)
+    /* We'll only process the first component */
+    info->num_components = 1;
+  else
+    /* Process all the components */
+    info->num_components = srcinfo->num_components;
+
+  /* Compute output image dimensions and related values. */
+#if JPEG_LIB_VERSION >= 80
+  jpeg_core_output_dimensions(srcinfo);
+#else
+  srcinfo->output_width = srcinfo->image_width;
+  srcinfo->output_height = srcinfo->image_height;
+#endif
+
+  /* Return right away if -perfect is given and transformation is not perfect.
+   */
+  if (info->perfect) {
+    if (info->num_components == 1) {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    } else {
+      if (!jtransform_perfect_transform(srcinfo->output_width,
+          srcinfo->output_height,
+          srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size,
+          srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size,
+          info->transform))
+        return FALSE;
+    }
+  }
+
+  /* If there is only one output component, force the iMCU size to be 1;
+   * else use the source iMCU size.  (This allows us to do the right thing
+   * when reducing color to grayscale, and also provides a handy way of
+   * cleaning up "funny" grayscale images whose sampling factors are not 1x1.)
+   */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+    info->output_width = srcinfo->output_height;
+    info->output_height = srcinfo->output_width;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_h_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+    }
+    break;
+  default:
+    info->output_width = srcinfo->output_width;
+    info->output_height = srcinfo->output_height;
+    if (info->num_components == 1) {
+      info->iMCU_sample_width = srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height = srcinfo->_min_DCT_v_scaled_size;
+    } else {
+      info->iMCU_sample_width =
+        srcinfo->max_h_samp_factor * srcinfo->_min_DCT_h_scaled_size;
+      info->iMCU_sample_height =
+        srcinfo->max_v_samp_factor * srcinfo->_min_DCT_v_scaled_size;
+    }
+    break;
+  }
+
+  /* If cropping has been requested, compute the crop area's position and
+   * dimensions, ensuring that its upper left corner falls at an iMCU boundary.
+   */
+  if (info->crop) {
+    /* Insert default values for unset crop parameters */
+    if (info->crop_xoffset_set == JCROP_UNSET)
+      info->crop_xoffset = 0;   /* default to +0 */
+    if (info->crop_yoffset_set == JCROP_UNSET)
+      info->crop_yoffset = 0;   /* default to +0 */
+    if (info->crop_xoffset >= info->output_width ||
+        info->crop_yoffset >= info->output_height)
+      ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+    if (info->crop_width_set == JCROP_UNSET)
+      info->crop_width = info->output_width - info->crop_xoffset;
+    if (info->crop_height_set == JCROP_UNSET)
+      info->crop_height = info->output_height - info->crop_yoffset;
+    /* Ensure parameters are valid */
+    if (info->crop_width <= 0 || info->crop_width > info->output_width ||
+        info->crop_height <= 0 || info->crop_height > info->output_height ||
+        info->crop_xoffset > info->output_width - info->crop_width ||
+        info->crop_yoffset > info->output_height - info->crop_height)
+      ERREXIT(srcinfo, JERR_BAD_CROP_SPEC);
+    /* Convert negative crop offsets into regular offsets */
+    if (info->crop_xoffset_set == JCROP_NEG)
+      xoffset = info->output_width - info->crop_width - info->crop_xoffset;
+    else
+      xoffset = info->crop_xoffset;
+    if (info->crop_yoffset_set == JCROP_NEG)
+      yoffset = info->output_height - info->crop_height - info->crop_yoffset;
+    else
+      yoffset = info->crop_yoffset;
+    /* Now adjust so that upper left corner falls at an iMCU boundary */
+    if (info->crop_width_set == JCROP_FORCE)
+      info->output_width = info->crop_width;
+    else
+      info->output_width =
+        info->crop_width + (xoffset % info->iMCU_sample_width);
+    if (info->crop_height_set == JCROP_FORCE)
+      info->output_height = info->crop_height;
+    else
+      info->output_height =
+        info->crop_height + (yoffset % info->iMCU_sample_height);
+    /* Save x/y offsets measured in iMCUs */
+    info->x_crop_offset = xoffset / info->iMCU_sample_width;
+    info->y_crop_offset = yoffset / info->iMCU_sample_height;
+  } else {
+    info->x_crop_offset = 0;
+    info->y_crop_offset = 0;
+  }
+
+  /* Figure out whether we need workspace arrays,
+   * and if so whether they are transposed relative to the source.
+   */
+  need_workspace = FALSE;
+  transpose_it = FALSE;
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->x_crop_offset != 0 || info->y_crop_offset != 0)
+      need_workspace = TRUE;
+    /* No workspace needed if neither cropping nor transforming */
+    break;
+  case JXFORM_FLIP_H:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_width);
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      need_workspace = TRUE;
+    /* do_flip_h_no_crop doesn't need a workspace array */
+    break;
+  case JXFORM_FLIP_V:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_TRANSPOSE:
+    /* transpose does NOT have to trim anything */
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_TRANSVERSE:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_height);
+      trim_bottom_edge(info, srcinfo->output_width);
+    }
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_90:
+    if (info->trim)
+      trim_right_edge(info, srcinfo->output_height);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  case JXFORM_ROT_180:
+    if (info->trim) {
+      trim_right_edge(info, srcinfo->output_width);
+      trim_bottom_edge(info, srcinfo->output_height);
+    }
+    /* Need workspace arrays having same dimensions as source image. */
+    need_workspace = TRUE;
+    break;
+  case JXFORM_ROT_270:
+    if (info->trim)
+      trim_bottom_edge(info, srcinfo->output_width);
+    /* Need workspace arrays having transposed dimensions. */
+    need_workspace = TRUE;
+    transpose_it = TRUE;
+    break;
+  }
+
+  /* Allocate workspace if needed.
+   * Note that we allocate arrays padded out to the next iMCU boundary,
+   * so that transform routines need not worry about missing edge blocks.
+   */
+  if (need_workspace) {
+    coef_arrays = (jvirt_barray_ptr *)
+      (*srcinfo->mem->alloc_small) ((j_common_ptr) srcinfo, JPOOL_IMAGE,
+                sizeof(jvirt_barray_ptr) * info->num_components);
+    width_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long) info->output_width,
+                    (long) info->iMCU_sample_width);
+    height_in_iMCUs = (JDIMENSION)
+      jdiv_round_up((long) info->output_height,
+                    (long) info->iMCU_sample_height);
+    for (ci = 0; ci < info->num_components; ci++) {
+      compptr = srcinfo->comp_info + ci;
+      if (info->num_components == 1) {
+        /* we're going to force samp factors to 1x1 in this case */
+        h_samp_factor = v_samp_factor = 1;
+      } else if (transpose_it) {
+        h_samp_factor = compptr->v_samp_factor;
+        v_samp_factor = compptr->h_samp_factor;
+      } else {
+        h_samp_factor = compptr->h_samp_factor;
+        v_samp_factor = compptr->v_samp_factor;
+      }
+      width_in_blocks = width_in_iMCUs * h_samp_factor;
+      height_in_blocks = height_in_iMCUs * v_samp_factor;
+      coef_arrays[ci] = (*srcinfo->mem->request_virt_barray)
+        ((j_common_ptr) srcinfo, JPOOL_IMAGE, FALSE,
+         width_in_blocks, height_in_blocks, (JDIMENSION) v_samp_factor);
+    }
+    info->workspace_coef_arrays = coef_arrays;
+  } else
+    info->workspace_coef_arrays = NULL;
+
+  return TRUE;
+}
+
+
+/* Transpose destination image parameters */
+
+LOCAL(void)
+transpose_critical_parameters (j_compress_ptr dstinfo)
+{
+  int tblno, i, j, ci, itemp;
+  jpeg_component_info *compptr;
+  JQUANT_TBL *qtblptr;
+  JDIMENSION jtemp;
+  UINT16 qtemp;
+
+  /* Transpose image dimensions */
+  jtemp = dstinfo->image_width;
+  dstinfo->image_width = dstinfo->image_height;
+  dstinfo->image_height = jtemp;
+#if JPEG_LIB_VERSION >= 70
+  itemp = dstinfo->min_DCT_h_scaled_size;
+  dstinfo->min_DCT_h_scaled_size = dstinfo->min_DCT_v_scaled_size;
+  dstinfo->min_DCT_v_scaled_size = itemp;
+#endif
+
+  /* Transpose sampling factors */
+  for (ci = 0; ci < dstinfo->num_components; ci++) {
+    compptr = dstinfo->comp_info + ci;
+    itemp = compptr->h_samp_factor;
+    compptr->h_samp_factor = compptr->v_samp_factor;
+    compptr->v_samp_factor = itemp;
+  }
+
+  /* Transpose quantization tables */
+  for (tblno = 0; tblno < NUM_QUANT_TBLS; tblno++) {
+    qtblptr = dstinfo->quant_tbl_ptrs[tblno];
+    if (qtblptr != NULL) {
+      for (i = 0; i < DCTSIZE; i++) {
+        for (j = 0; j < i; j++) {
+          qtemp = qtblptr->quantval[i*DCTSIZE+j];
+          qtblptr->quantval[i*DCTSIZE+j] = qtblptr->quantval[j*DCTSIZE+i];
+          qtblptr->quantval[j*DCTSIZE+i] = qtemp;
+        }
+      }
+    }
+  }
+}
+
+
+/* Adjust Exif image parameters.
+ *
+ * We try to adjust the Tags ExifImageWidth and ExifImageHeight if possible.
+ */
+
+#if JPEG_LIB_VERSION >= 70
+LOCAL(void)
+adjust_exif_parameters (JOCTET *data, unsigned int length,
+                        JDIMENSION new_width, JDIMENSION new_height)
+{
+  boolean is_motorola; /* Flag for byte order */
+  unsigned int number_of_tags, tagnum;
+  unsigned int firstoffset, offset;
+  JDIMENSION new_value;
+
+  if (length < 12) return; /* Length of an IFD entry */
+
+  /* Discover byte order */
+  if (GETJOCTET(data[0]) == 0x49 && GETJOCTET(data[1]) == 0x49)
+    is_motorola = FALSE;
+  else if (GETJOCTET(data[0]) == 0x4D && GETJOCTET(data[1]) == 0x4D)
+    is_motorola = TRUE;
+  else
+    return;
+
+  /* Check Tag Mark */
+  if (is_motorola) {
+    if (GETJOCTET(data[2]) != 0) return;
+    if (GETJOCTET(data[3]) != 0x2A) return;
+  } else {
+    if (GETJOCTET(data[3]) != 0) return;
+    if (GETJOCTET(data[2]) != 0x2A) return;
+  }
+
+  /* Get first IFD offset (offset to IFD0) */
+  if (is_motorola) {
+    if (GETJOCTET(data[4]) != 0) return;
+    if (GETJOCTET(data[5]) != 0) return;
+    firstoffset = GETJOCTET(data[6]);
+    firstoffset <<= 8;
+    firstoffset += GETJOCTET(data[7]);
+  } else {
+    if (GETJOCTET(data[7]) != 0) return;
+    if (GETJOCTET(data[6]) != 0) return;
+    firstoffset = GETJOCTET(data[5]);
+    firstoffset <<= 8;
+    firstoffset += GETJOCTET(data[4]);
+  }
+  if (firstoffset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this IFD */
+  if (is_motorola) {
+    number_of_tags = GETJOCTET(data[firstoffset]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[firstoffset+1]);
+  } else {
+    number_of_tags = GETJOCTET(data[firstoffset+1]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[firstoffset]);
+  }
+  if (number_of_tags == 0) return;
+  firstoffset += 2;
+
+  /* Search for ExifSubIFD offset Tag in IFD0 */
+  for (;;) {
+    if (firstoffset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = GETJOCTET(data[firstoffset]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[firstoffset+1]);
+    } else {
+      tagnum = GETJOCTET(data[firstoffset+1]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[firstoffset]);
+    }
+    if (tagnum == 0x8769) break; /* found ExifSubIFD offset Tag */
+    if (--number_of_tags == 0) return;
+    firstoffset += 12;
+  }
+
+  /* Get the ExifSubIFD offset */
+  if (is_motorola) {
+    if (GETJOCTET(data[firstoffset+8]) != 0) return;
+    if (GETJOCTET(data[firstoffset+9]) != 0) return;
+    offset = GETJOCTET(data[firstoffset+10]);
+    offset <<= 8;
+    offset += GETJOCTET(data[firstoffset+11]);
+  } else {
+    if (GETJOCTET(data[firstoffset+11]) != 0) return;
+    if (GETJOCTET(data[firstoffset+10]) != 0) return;
+    offset = GETJOCTET(data[firstoffset+9]);
+    offset <<= 8;
+    offset += GETJOCTET(data[firstoffset+8]);
+  }
+  if (offset > length - 2) return; /* check end of data segment */
+
+  /* Get the number of directory entries contained in this SubIFD */
+  if (is_motorola) {
+    number_of_tags = GETJOCTET(data[offset]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[offset+1]);
+  } else {
+    number_of_tags = GETJOCTET(data[offset+1]);
+    number_of_tags <<= 8;
+    number_of_tags += GETJOCTET(data[offset]);
+  }
+  if (number_of_tags < 2) return;
+  offset += 2;
+
+  /* Search for ExifImageWidth and ExifImageHeight Tags in this SubIFD */
+  do {
+    if (offset > length - 12) return; /* check end of data segment */
+    /* Get Tag number */
+    if (is_motorola) {
+      tagnum = GETJOCTET(data[offset]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[offset+1]);
+    } else {
+      tagnum = GETJOCTET(data[offset+1]);
+      tagnum <<= 8;
+      tagnum += GETJOCTET(data[offset]);
+    }
+    if (tagnum == 0xA002 || tagnum == 0xA003) {
+      if (tagnum == 0xA002)
+        new_value = new_width; /* ExifImageWidth Tag */
+      else
+        new_value = new_height; /* ExifImageHeight Tag */
+      if (is_motorola) {
+        data[offset+2] = 0; /* Format = unsigned long (4 octets) */
+        data[offset+3] = 4;
+        data[offset+4] = 0; /* Number Of Components = 1 */
+        data[offset+5] = 0;
+        data[offset+6] = 0;
+        data[offset+7] = 1;
+        data[offset+8] = 0;
+        data[offset+9] = 0;
+        data[offset+10] = (JOCTET)((new_value >> 8) & 0xFF);
+        data[offset+11] = (JOCTET)(new_value & 0xFF);
+      } else {
+        data[offset+2] = 4; /* Format = unsigned long (4 octets) */
+        data[offset+3] = 0;
+        data[offset+4] = 1; /* Number Of Components = 1 */
+        data[offset+5] = 0;
+        data[offset+6] = 0;
+        data[offset+7] = 0;
+        data[offset+8] = (JOCTET)(new_value & 0xFF);
+        data[offset+9] = (JOCTET)((new_value >> 8) & 0xFF);
+        data[offset+10] = 0;
+        data[offset+11] = 0;
+      }
+    }
+    offset += 12;
+  } while (--number_of_tags);
+}
+#endif
+
+
+/* Adjust output image parameters as needed.
+ *
+ * This must be called after jpeg_copy_critical_parameters()
+ * and before jpeg_write_coefficients().
+ *
+ * The return value is the set of virtual coefficient arrays to be written
+ * (either the ones allocated by jtransform_request_workspace, or the
+ * original source data arrays).  The caller will need to pass this value
+ * to jpeg_write_coefficients().
+ */
+
+GLOBAL(jvirt_barray_ptr *)
+jtransform_adjust_parameters (j_decompress_ptr srcinfo,
+                              j_compress_ptr dstinfo,
+                              jvirt_barray_ptr *src_coef_arrays,
+                              jpeg_transform_info *info)
+{
+  /* If force-to-grayscale is requested, adjust destination parameters */
+  if (info->force_grayscale) {
+    /* First, ensure we have YCbCr or grayscale data, and that the source's
+     * Y channel is full resolution.  (No reasonable person would make Y
+     * be less than full resolution, so actually coping with that case
+     * isn't worth extra code space.  But we check it to avoid crashing.)
+     */
+    if (((dstinfo->jpeg_color_space == JCS_YCbCr &&
+          dstinfo->num_components == 3) ||
+         (dstinfo->jpeg_color_space == JCS_GRAYSCALE &&
+          dstinfo->num_components == 1)) &&
+        srcinfo->comp_info[0].h_samp_factor == srcinfo->max_h_samp_factor &&
+        srcinfo->comp_info[0].v_samp_factor == srcinfo->max_v_samp_factor) {
+      /* We use jpeg_set_colorspace to make sure subsidiary settings get fixed
+       * properly.  Among other things, it sets the target h_samp_factor &
+       * v_samp_factor to 1, which typically won't match the source.
+       * We have to preserve the source's quantization table number, however.
+       */
+      int sv_quant_tbl_no = dstinfo->comp_info[0].quant_tbl_no;
+      jpeg_set_colorspace(dstinfo, JCS_GRAYSCALE);
+      dstinfo->comp_info[0].quant_tbl_no = sv_quant_tbl_no;
+    } else {
+      /* Sorry, can't do it */
+      ERREXIT(dstinfo, JERR_CONVERSION_NOTIMPL);
+    }
+  } else if (info->num_components == 1) {
+    /* For a single-component source, we force the destination sampling factors
+     * to 1x1, with or without force_grayscale.  This is useful because some
+     * decoders choke on grayscale images with other sampling factors.
+     */
+    dstinfo->comp_info[0].h_samp_factor = 1;
+    dstinfo->comp_info[0].v_samp_factor = 1;
+  }
+
+  /* Correct the destination's image dimensions as necessary
+   * for rotate/flip, resize, and crop operations.
+   */
+#if JPEG_LIB_VERSION >= 70
+  dstinfo->jpeg_width = info->output_width;
+  dstinfo->jpeg_height = info->output_height;
+#endif
+
+  /* Transpose destination image parameters */
+  switch (info->transform) {
+  case JXFORM_TRANSPOSE:
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_90:
+  case JXFORM_ROT_270:
+#if JPEG_LIB_VERSION < 70
+    dstinfo->image_width = info->output_height;
+    dstinfo->image_height = info->output_width;
+#endif
+    transpose_critical_parameters(dstinfo);
+    break;
+  default:
+#if JPEG_LIB_VERSION < 70
+    dstinfo->image_width = info->output_width;
+    dstinfo->image_height = info->output_height;
+#endif
+    break;
+  }
+
+  /* Adjust Exif properties */
+  if (srcinfo->marker_list != NULL &&
+      srcinfo->marker_list->marker == JPEG_APP0+1 &&
+      srcinfo->marker_list->data_length >= 6 &&
+      GETJOCTET(srcinfo->marker_list->data[0]) == 0x45 &&
+      GETJOCTET(srcinfo->marker_list->data[1]) == 0x78 &&
+      GETJOCTET(srcinfo->marker_list->data[2]) == 0x69 &&
+      GETJOCTET(srcinfo->marker_list->data[3]) == 0x66 &&
+      GETJOCTET(srcinfo->marker_list->data[4]) == 0 &&
+      GETJOCTET(srcinfo->marker_list->data[5]) == 0) {
+    /* Suppress output of JFIF marker */
+    dstinfo->write_JFIF_header = FALSE;
+#if JPEG_LIB_VERSION >= 70
+    /* Adjust Exif image parameters */
+    if (dstinfo->jpeg_width != srcinfo->image_width ||
+        dstinfo->jpeg_height != srcinfo->image_height)
+      /* Align data segment to start of TIFF structure for parsing */
+      adjust_exif_parameters(srcinfo->marker_list->data + 6,
+        srcinfo->marker_list->data_length - 6,
+        dstinfo->jpeg_width, dstinfo->jpeg_height);
+#endif
+  }
+
+  /* Return the appropriate output data set */
+  if (info->workspace_coef_arrays != NULL)
+    return info->workspace_coef_arrays;
+  return src_coef_arrays;
+}
+
+
+/* Execute the actual transformation, if any.
+ *
+ * This must be called *after* jpeg_write_coefficients, because it depends
+ * on jpeg_write_coefficients to have computed subsidiary values such as
+ * the per-component width and height fields in the destination object.
+ *
+ * Note that some transformations will modify the source data arrays!
+ */
+
+GLOBAL(void)
+jtransform_execute_transform (j_decompress_ptr srcinfo,
+                              j_compress_ptr dstinfo,
+                              jvirt_barray_ptr *src_coef_arrays,
+                              jpeg_transform_info *info)
+{
+  jvirt_barray_ptr *dst_coef_arrays = info->workspace_coef_arrays;
+
+  /* Note: conditions tested here should match those in switch statement
+   * in jtransform_request_workspace()
+   */
+  switch (info->transform) {
+  case JXFORM_NONE:
+    if (info->x_crop_offset != 0 || info->y_crop_offset != 0)
+      do_crop(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_FLIP_H:
+    if (info->y_crop_offset != 0 || info->slow_hflip)
+      do_flip_h(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                src_coef_arrays, dst_coef_arrays);
+    else
+      do_flip_h_no_crop(srcinfo, dstinfo, info->x_crop_offset,
+                        src_coef_arrays);
+    break;
+  case JXFORM_FLIP_V:
+    do_flip_v(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSPOSE:
+    do_transpose(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                 src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_TRANSVERSE:
+    do_transverse(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+                  src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_90:
+    do_rot_90(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+              src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_180:
+    do_rot_180(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  case JXFORM_ROT_270:
+    do_rot_270(srcinfo, dstinfo, info->x_crop_offset, info->y_crop_offset,
+               src_coef_arrays, dst_coef_arrays);
+    break;
+  }
+}
+
+/* jtransform_perfect_transform
+ *
+ * Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ *
+ * Inputs:
+ *   image_width, image_height: source image dimensions.
+ *   MCU_width, MCU_height: pixel dimensions of MCU.
+ *   transform: transformation identifier.
+ * Parameter sources from initialized jpeg_struct
+ * (after reading source header):
+ *   image_width = cinfo.image_width
+ *   image_height = cinfo.image_height
+ *   MCU_width = cinfo.max_h_samp_factor * cinfo.block_size
+ *   MCU_height = cinfo.max_v_samp_factor * cinfo.block_size
+ * Result:
+ *   TRUE = perfect transformation possible
+ *   FALSE = perfect transformation not possible
+ *           (may use custom action then)
+ */
+
+GLOBAL(boolean)
+jtransform_perfect_transform(JDIMENSION image_width, JDIMENSION image_height,
+                             int MCU_width, int MCU_height,
+                             JXFORM_CODE transform)
+{
+  boolean result = TRUE; /* initialize TRUE */
+
+  switch (transform) {
+  case JXFORM_FLIP_H:
+  case JXFORM_ROT_270:
+    if (image_width % (JDIMENSION) MCU_width)
+      result = FALSE;
+    break;
+  case JXFORM_FLIP_V:
+  case JXFORM_ROT_90:
+    if (image_height % (JDIMENSION) MCU_height)
+      result = FALSE;
+    break;
+  case JXFORM_TRANSVERSE:
+  case JXFORM_ROT_180:
+    if (image_width % (JDIMENSION) MCU_width)
+      result = FALSE;
+    if (image_height % (JDIMENSION) MCU_height)
+      result = FALSE;
+    break;
+  default:
+    break;
+  }
+
+  return result;
+}
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/* Setup decompression object to save desired markers in memory.
+ * This must be called before jpeg_read_header() to have the desired effect.
+ */
+
+GLOBAL(void)
+jcopy_markers_setup (j_decompress_ptr srcinfo, JCOPY_OPTION option)
+{
+#ifdef SAVE_MARKERS_SUPPORTED
+  int m;
+
+  /* Save comments except under NONE option */
+  if (option != JCOPYOPT_NONE) {
+    jpeg_save_markers(srcinfo, JPEG_COM, 0xFFFF);
+  }
+  /* Save all types of APPn markers iff ALL option */
+  if (option == JCOPYOPT_ALL) {
+    for (m = 0; m < 16; m++)
+      jpeg_save_markers(srcinfo, JPEG_APP0 + m, 0xFFFF);
+  }
+#endif /* SAVE_MARKERS_SUPPORTED */
+}
+
+/* Copy markers saved in the given source object to the destination object.
+ * This should be called just after jpeg_start_compress() or
+ * jpeg_write_coefficients().
+ * Note that those routines will have written the SOI, and also the
+ * JFIF APP0 or Adobe APP14 markers if selected.
+ */
+
+GLOBAL(void)
+jcopy_markers_execute (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+                       JCOPY_OPTION option)
+{
+  jpeg_saved_marker_ptr marker;
+
+  /* In the current implementation, we don't actually need to examine the
+   * option flag here; we just copy everything that got saved.
+   * But to avoid confusion, we do not output JFIF and Adobe APP14 markers
+   * if the encoder library already wrote one.
+   */
+  for (marker = srcinfo->marker_list; marker != NULL; marker = marker->next) {
+    if (dstinfo->write_JFIF_header &&
+        marker->marker == JPEG_APP0 &&
+        marker->data_length >= 5 &&
+        GETJOCTET(marker->data[0]) == 0x4A &&
+        GETJOCTET(marker->data[1]) == 0x46 &&
+        GETJOCTET(marker->data[2]) == 0x49 &&
+        GETJOCTET(marker->data[3]) == 0x46 &&
+        GETJOCTET(marker->data[4]) == 0)
+      continue;                 /* reject duplicate JFIF */
+    if (dstinfo->write_Adobe_marker &&
+        marker->marker == JPEG_APP0+14 &&
+        marker->data_length >= 5 &&
+        GETJOCTET(marker->data[0]) == 0x41 &&
+        GETJOCTET(marker->data[1]) == 0x64 &&
+        GETJOCTET(marker->data[2]) == 0x6F &&
+        GETJOCTET(marker->data[3]) == 0x62 &&
+        GETJOCTET(marker->data[4]) == 0x65)
+      continue;                 /* reject duplicate Adobe */
+    jpeg_write_marker(dstinfo, marker->marker,
+                      marker->data, marker->data_length);
+  }
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/transupp.h glmark2-2021.02/src/libjpeg-turbo/transupp.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/transupp.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/transupp.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,207 @@
+/*
+ * transupp.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1997-2011, Thomas G. Lane, Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code relevant
+ * to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains declarations for image transformation routines and
+ * other utility code used by the jpegtran sample application.  These are
+ * NOT part of the core JPEG library.  But we keep these routines separate
+ * from jpegtran.c to ease the task of maintaining jpegtran-like programs
+ * that have other user interfaces.
+ *
+ * NOTE: all the routines declared here have very specific requirements
+ * about when they are to be executed during the reading and writing of the
+ * source and destination files.  See the comments in transupp.c, or see
+ * jpegtran.c for an example of correct usage.
+ */
+
+/* If you happen not to want the image transform support, disable it here */
+#ifndef TRANSFORMS_SUPPORTED
+#define TRANSFORMS_SUPPORTED 1          /* 0 disables transform code */
+#endif
+
+/*
+ * Although rotating and flipping data expressed as DCT coefficients is not
+ * hard, there is an asymmetry in the JPEG format specification for images
+ * whose dimensions aren't multiples of the iMCU size.  The right and bottom
+ * image edges are padded out to the next iMCU boundary with junk data; but
+ * no padding is possible at the top and left edges.  If we were to flip
+ * the whole image including the pad data, then pad garbage would become
+ * visible at the top and/or left, and real pixels would disappear into the
+ * pad margins --- perhaps permanently, since encoders & decoders may not
+ * bother to preserve DCT blocks that appear to be completely outside the
+ * nominal image area.  So, we have to exclude any partial iMCUs from the
+ * basic transformation.
+ *
+ * Transpose is the only transformation that can handle partial iMCUs at the
+ * right and bottom edges completely cleanly.  flip_h can flip partial iMCUs
+ * at the bottom, but leaves any partial iMCUs at the right edge untouched.
+ * Similarly flip_v leaves any partial iMCUs at the bottom edge untouched.
+ * The other transforms are defined as combinations of these basic transforms
+ * and process edge blocks in a way that preserves the equivalence.
+ *
+ * The "trim" option causes untransformable partial iMCUs to be dropped;
+ * this is not strictly lossless, but it usually gives the best-looking
+ * result for odd-size images.  Note that when this option is active,
+ * the expected mathematical equivalences between the transforms may not hold.
+ * (For example, -rot 270 -trim trims only the bottom edge, but -rot 90 -trim
+ * followed by -rot 180 -trim trims both edges.)
+ *
+ * We also offer a lossless-crop option, which discards data outside a given
+ * image region but losslessly preserves what is inside.  Like the rotate and
+ * flip transforms, lossless crop is restricted by the JPEG format: the upper
+ * left corner of the selected region must fall on an iMCU boundary.  If this
+ * does not hold for the given crop parameters, we silently move the upper left
+ * corner up and/or left to make it so, simultaneously increasing the region
+ * dimensions to keep the lower right crop corner unchanged.  (Thus, the
+ * output image covers at least the requested region, but may cover more.)
+ * The adjustment of the region dimensions may be optionally disabled.
+ *
+ * We also provide a lossless-resize option, which is kind of a lossless-crop
+ * operation in the DCT coefficient block domain - it discards higher-order
+ * coefficients and losslessly preserves lower-order coefficients of a
+ * sub-block.
+ *
+ * Rotate/flip transform, resize, and crop can be requested together in a
+ * single invocation.  The crop is applied last --- that is, the crop region
+ * is specified in terms of the destination image after transform/resize.
+ *
+ * We also offer a "force to grayscale" option, which simply discards the
+ * chrominance channels of a YCbCr image.  This is lossless in the sense that
+ * the luminance channel is preserved exactly.  It's not the same kind of
+ * thing as the rotate/flip transformations, but it's convenient to handle it
+ * as part of this package, mainly because the transformation routines have to
+ * be aware of the option to know how many components to work on.
+ */
+
+
+/*
+ * Codes for supported types of image transformations.
+ */
+
+typedef enum {
+  JXFORM_NONE,            /* no transformation */
+  JXFORM_FLIP_H,          /* horizontal flip */
+  JXFORM_FLIP_V,          /* vertical flip */
+  JXFORM_TRANSPOSE,       /* transpose across UL-to-LR axis */
+  JXFORM_TRANSVERSE,      /* transpose across UR-to-LL axis */
+  JXFORM_ROT_90,          /* 90-degree clockwise rotation */
+  JXFORM_ROT_180,         /* 180-degree rotation */
+  JXFORM_ROT_270          /* 270-degree clockwise (or 90 ccw) */
+} JXFORM_CODE;
+
+/*
+ * Codes for crop parameters, which can individually be unspecified,
+ * positive or negative for xoffset or yoffset,
+ * positive or forced for width or height.
+ */
+
+typedef enum {
+  JCROP_UNSET,
+  JCROP_POS,
+  JCROP_NEG,
+  JCROP_FORCE
+} JCROP_CODE;
+
+/*
+ * Transform parameters struct.
+ * NB: application must not change any elements of this struct after
+ * calling jtransform_request_workspace.
+ */
+
+typedef struct {
+  /* Options: set by caller */
+  JXFORM_CODE transform;        /* image transform operator */
+  boolean perfect;              /* if TRUE, fail if partial MCUs are requested */
+  boolean trim;                 /* if TRUE, trim partial MCUs as needed */
+  boolean force_grayscale;      /* if TRUE, convert color image to grayscale */
+  boolean crop;                 /* if TRUE, crop source image */
+  boolean slow_hflip;  /* For best performance, the JXFORM_FLIP_H transform
+                          normally modifies the source coefficients in place.
+                          Setting this to TRUE will instead use a slower,
+                          double-buffered algorithm, which leaves the source
+                          coefficients in tact (necessary if other transformed
+                          images must be generated from the same set of
+                          coefficients. */
+
+  /* Crop parameters: application need not set these unless crop is TRUE.
+   * These can be filled in by jtransform_parse_crop_spec().
+   */
+  JDIMENSION crop_width;        /* Width of selected region */
+  JCROP_CODE crop_width_set;    /* (forced disables adjustment) */
+  JDIMENSION crop_height;       /* Height of selected region */
+  JCROP_CODE crop_height_set;   /* (forced disables adjustment) */
+  JDIMENSION crop_xoffset;      /* X offset of selected region */
+  JCROP_CODE crop_xoffset_set;  /* (negative measures from right edge) */
+  JDIMENSION crop_yoffset;      /* Y offset of selected region */
+  JCROP_CODE crop_yoffset_set;  /* (negative measures from bottom edge) */
+
+  /* Internal workspace: caller should not touch these */
+  int num_components;           /* # of components in workspace */
+  jvirt_barray_ptr *workspace_coef_arrays; /* workspace for transformations */
+  JDIMENSION output_width;      /* cropped destination dimensions */
+  JDIMENSION output_height;
+  JDIMENSION x_crop_offset;     /* destination crop offsets measured in iMCUs */
+  JDIMENSION y_crop_offset;
+  int iMCU_sample_width;        /* destination iMCU size */
+  int iMCU_sample_height;
+} jpeg_transform_info;
+
+
+#if TRANSFORMS_SUPPORTED
+
+/* Parse a crop specification (written in X11 geometry style) */
+EXTERN(boolean) jtransform_parse_crop_spec
+        (jpeg_transform_info *info, const char *spec);
+/* Request any required workspace */
+EXTERN(boolean) jtransform_request_workspace
+        (j_decompress_ptr srcinfo, jpeg_transform_info *info);
+/* Adjust output image parameters */
+EXTERN(jvirt_barray_ptr *) jtransform_adjust_parameters
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         jvirt_barray_ptr *src_coef_arrays, jpeg_transform_info *info);
+/* Execute the actual transformation, if any */
+EXTERN(void) jtransform_execute_transform
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         jvirt_barray_ptr *src_coef_arrays, jpeg_transform_info *info);
+/* Determine whether lossless transformation is perfectly
+ * possible for a specified image and transformation.
+ */
+EXTERN(boolean) jtransform_perfect_transform
+        (JDIMENSION image_width, JDIMENSION image_height, int MCU_width,
+         int MCU_height, JXFORM_CODE transform);
+
+/* jtransform_execute_transform used to be called
+ * jtransform_execute_transformation, but some compilers complain about
+ * routine names that long.  This macro is here to avoid breaking any
+ * old source code that uses the original name...
+ */
+#define jtransform_execute_transformation       jtransform_execute_transform
+
+#endif /* TRANSFORMS_SUPPORTED */
+
+
+/*
+ * Support for copying optional markers from source to destination file.
+ */
+
+typedef enum {
+  JCOPYOPT_NONE,          /* copy no optional markers */
+  JCOPYOPT_COMMENTS,      /* copy only comment (COM) markers */
+  JCOPYOPT_ALL            /* copy all optional markers */
+} JCOPY_OPTION;
+
+#define JCOPYOPT_DEFAULT  JCOPYOPT_COMMENTS     /* recommended default */
+
+/* Setup decompression object to save desired markers in memory */
+EXTERN(void) jcopy_markers_setup
+        (j_decompress_ptr srcinfo, JCOPY_OPTION option);
+/* Copy markers saved in the given source object to the destination object */
+EXTERN(void) jcopy_markers_execute
+        (j_decompress_ptr srcinfo, j_compress_ptr dstinfo,
+         JCOPY_OPTION option);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg.c glmark2-2021.02/src/libjpeg-turbo/turbojpeg.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/turbojpeg.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,2121 @@
+/*
+ * Copyright (C)2009-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/* TurboJPEG/LJT:  this implements the TurboJPEG API using libjpeg or
+   libjpeg-turbo */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <jinclude.h>
+#define JPEG_INTERNALS
+#include <jpeglib.h>
+#include <jerror.h>
+#include <setjmp.h>
+#include "./turbojpeg.h"
+#include "./tjutil.h"
+#include "transupp.h"
+#include "./jpegcomp.h"
+
+extern void jpeg_mem_dest_tj(j_compress_ptr, unsigned char **,
+	unsigned long *, boolean);
+extern void jpeg_mem_src_tj(j_decompress_ptr, const unsigned char *,
+	unsigned long);
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+#define isPow2(x) (((x)&(x-1))==0)
+
+
+/* Error handling (based on example in example.c) */
+
+static char errStr[JMSG_LENGTH_MAX]="No error";
+
+struct my_error_mgr
+{
+	struct jpeg_error_mgr pub;
+	jmp_buf setjmp_buffer;
+	void (*emit_message)(j_common_ptr, int);
+	boolean warning;
+};
+typedef struct my_error_mgr *my_error_ptr;
+
+static void my_error_exit(j_common_ptr cinfo)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	(*cinfo->err->output_message)(cinfo);
+	longjmp(myerr->setjmp_buffer, 1);
+}
+
+/* Based on output_message() in jerror.c */
+
+static void my_output_message(j_common_ptr cinfo)
+{
+	(*cinfo->err->format_message)(cinfo, errStr);
+}
+
+static void my_emit_message(j_common_ptr cinfo, int msg_level)
+{
+	my_error_ptr myerr=(my_error_ptr)cinfo->err;
+	myerr->emit_message(cinfo, msg_level);
+	if(msg_level<0) myerr->warning=TRUE;
+}
+
+
+/* Global structures, macros, etc. */
+
+enum {COMPRESS=1, DECOMPRESS=2};
+
+typedef struct _tjinstance
+{
+	struct jpeg_compress_struct cinfo;
+	struct jpeg_decompress_struct dinfo;
+	struct my_error_mgr jerr;
+	int init, headerRead;
+} tjinstance;
+
+static const int pixelsize[TJ_NUMSAMP]={3, 3, 3, 1, 3, 3};
+
+static const JXFORM_CODE xformtypes[TJ_NUMXOP]=
+{
+	JXFORM_NONE, JXFORM_FLIP_H, JXFORM_FLIP_V, JXFORM_TRANSPOSE,
+	JXFORM_TRANSVERSE, JXFORM_ROT_90, JXFORM_ROT_180, JXFORM_ROT_270
+};
+
+#define NUMSF 16
+static const tjscalingfactor sf[NUMSF]={
+	{2, 1},
+	{15, 8},
+	{7, 4},
+	{13, 8},
+	{3, 2},
+	{11, 8},
+	{5, 4},
+	{9, 8},
+	{1, 1},
+	{7, 8},
+	{3, 4},
+	{5, 8},
+	{1, 2},
+	{3, 8},
+	{1, 4},
+	{1, 8}
+};
+
+#define _throw(m) {snprintf(errStr, JMSG_LENGTH_MAX, "%s", m);  \
+	retval=-1;  goto bailout;}
+#define getinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_compress_ptr cinfo=NULL;  j_decompress_ptr dinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	cinfo=&this->cinfo;  dinfo=&this->dinfo;  \
+	this->jerr.warning=FALSE;
+#define getcinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_compress_ptr cinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	cinfo=&this->cinfo;  \
+	this->jerr.warning=FALSE;
+#define getdinstance(handle) tjinstance *this=(tjinstance *)handle;  \
+	j_decompress_ptr dinfo=NULL;  \
+	if(!this) {snprintf(errStr, JMSG_LENGTH_MAX, "Invalid handle");  \
+		return -1;}  \
+	dinfo=&this->dinfo;  \
+	this->jerr.warning=FALSE;
+
+static int getPixelFormat(int pixelSize, int flags)
+{
+	if(pixelSize==1) return TJPF_GRAY;
+	if(pixelSize==3)
+	{
+		if(flags&TJ_BGR) return TJPF_BGR;
+		else return TJPF_RGB;
+	}
+	if(pixelSize==4)
+	{
+		if(flags&TJ_ALPHAFIRST)
+		{
+			if(flags&TJ_BGR) return TJPF_XBGR;
+			else return TJPF_XRGB;
+		}
+		else
+		{
+			if(flags&TJ_BGR) return TJPF_BGRX;
+			else return TJPF_RGBX;
+		}
+	}
+	return -1;
+}
+
+static int setCompDefaults(struct jpeg_compress_struct *cinfo,
+	int pixelFormat, int subsamp, int jpegQual, int flags)
+{
+	int retval=0;
+	char *env=NULL;
+
+	switch(pixelFormat)
+	{
+		case TJPF_GRAY:
+			cinfo->in_color_space=JCS_GRAYSCALE;  break;
+		#if JCS_EXTENSIONS==1
+		case TJPF_RGB:
+			cinfo->in_color_space=JCS_EXT_RGB;  break;
+		case TJPF_BGR:
+			cinfo->in_color_space=JCS_EXT_BGR;  break;
+		case TJPF_RGBX:
+		case TJPF_RGBA:
+			cinfo->in_color_space=JCS_EXT_RGBX;  break;
+		case TJPF_BGRX:
+		case TJPF_BGRA:
+			cinfo->in_color_space=JCS_EXT_BGRX;  break;
+		case TJPF_XRGB:
+		case TJPF_ARGB:
+			cinfo->in_color_space=JCS_EXT_XRGB;  break;
+		case TJPF_XBGR:
+		case TJPF_ABGR:
+			cinfo->in_color_space=JCS_EXT_XBGR;  break;
+		#else
+		case TJPF_RGB:
+		case TJPF_BGR:
+		case TJPF_RGBX:
+		case TJPF_BGRX:
+		case TJPF_XRGB:
+		case TJPF_XBGR:
+		case TJPF_RGBA:
+		case TJPF_BGRA:
+		case TJPF_ARGB:
+		case TJPF_ABGR:
+			cinfo->in_color_space=JCS_RGB;  pixelFormat=TJPF_RGB;
+			break;
+		#endif
+		case TJPF_CMYK:
+			cinfo->in_color_space=JCS_CMYK;  break;
+	}
+
+	cinfo->input_components=tjPixelSize[pixelFormat];
+	jpeg_set_defaults(cinfo);
+
+#ifndef NO_GETENV
+	if((env=getenv("TJ_OPTIMIZE"))!=NULL && strlen(env)>0 && !strcmp(env, "1"))
+		cinfo->optimize_coding=TRUE;
+	if((env=getenv("TJ_ARITHMETIC"))!=NULL && strlen(env)>0	&& !strcmp(env, "1"))
+		cinfo->arith_code=TRUE;
+	if((env=getenv("TJ_RESTART"))!=NULL && strlen(env)>0)
+	{
+		int temp=-1;  char tempc=0;
+		if(sscanf(env, "%d%c", &temp, &tempc)>=1 && temp>=0 && temp<=65535)
+		{
+			if(toupper(tempc)=='B')
+			{
+				cinfo->restart_interval=temp;
+				cinfo->restart_in_rows=0;
+			}
+			else
+				cinfo->restart_in_rows=temp;
+		}
+	}
+#endif
+
+	if(jpegQual>=0)
+	{
+		jpeg_set_quality(cinfo, jpegQual, TRUE);
+		if(jpegQual>=96 || flags&TJFLAG_ACCURATEDCT) cinfo->dct_method=JDCT_ISLOW;
+		else cinfo->dct_method=JDCT_FASTEST;
+	}
+	if(subsamp==TJSAMP_GRAY)
+		jpeg_set_colorspace(cinfo, JCS_GRAYSCALE);
+	else if(pixelFormat==TJPF_CMYK)
+		jpeg_set_colorspace(cinfo, JCS_YCCK);
+	else jpeg_set_colorspace(cinfo, JCS_YCbCr);
+
+#ifndef NO_GETENV
+	if((env=getenv("TJ_PROGRESSIVE"))!=NULL && strlen(env)>0
+		&& !strcmp(env, "1"))
+		jpeg_simple_progression(cinfo);
+#endif
+
+	cinfo->comp_info[0].h_samp_factor=tjMCUWidth[subsamp]/8;
+	cinfo->comp_info[1].h_samp_factor=1;
+	cinfo->comp_info[2].h_samp_factor=1;
+	if(cinfo->num_components>3)
+		cinfo->comp_info[3].h_samp_factor=tjMCUWidth[subsamp]/8;
+	cinfo->comp_info[0].v_samp_factor=tjMCUHeight[subsamp]/8;
+	cinfo->comp_info[1].v_samp_factor=1;
+	cinfo->comp_info[2].v_samp_factor=1;
+	if(cinfo->num_components>3)
+		cinfo->comp_info[3].v_samp_factor=tjMCUHeight[subsamp]/8;
+
+	return retval;
+}
+
+static int setDecompDefaults(struct jpeg_decompress_struct *dinfo,
+	int pixelFormat, int flags)
+{
+	int retval=0;
+
+	switch(pixelFormat)
+	{
+		case TJPF_GRAY:
+			dinfo->out_color_space=JCS_GRAYSCALE;  break;
+		#if JCS_EXTENSIONS==1
+		case TJPF_RGB:
+			dinfo->out_color_space=JCS_EXT_RGB;  break;
+		case TJPF_BGR:
+			dinfo->out_color_space=JCS_EXT_BGR;  break;
+		case TJPF_RGBX:
+			dinfo->out_color_space=JCS_EXT_RGBX;  break;
+		case TJPF_BGRX:
+			dinfo->out_color_space=JCS_EXT_BGRX;  break;
+		case TJPF_XRGB:
+			dinfo->out_color_space=JCS_EXT_XRGB;  break;
+		case TJPF_XBGR:
+			dinfo->out_color_space=JCS_EXT_XBGR;  break;
+		#if JCS_ALPHA_EXTENSIONS==1
+		case TJPF_RGBA:
+			dinfo->out_color_space=JCS_EXT_RGBA;  break;
+		case TJPF_BGRA:
+			dinfo->out_color_space=JCS_EXT_BGRA;  break;
+		case TJPF_ARGB:
+			dinfo->out_color_space=JCS_EXT_ARGB;  break;
+		case TJPF_ABGR:
+			dinfo->out_color_space=JCS_EXT_ABGR;  break;
+		#endif
+		#else
+		case TJPF_RGB:
+		case TJPF_BGR:
+		case TJPF_RGBX:
+		case TJPF_BGRX:
+		case TJPF_XRGB:
+		case TJPF_XBGR:
+		case TJPF_RGBA:
+		case TJPF_BGRA:
+		case TJPF_ARGB:
+		case TJPF_ABGR:
+			dinfo->out_color_space=JCS_RGB;  break;
+		#endif
+		case TJPF_CMYK:
+			dinfo->out_color_space=JCS_CMYK;  break;
+		default:
+			_throw("Unsupported pixel format");
+	}
+
+	if(flags&TJFLAG_FASTDCT) dinfo->dct_method=JDCT_FASTEST;
+
+	bailout:
+	return retval;
+}
+
+
+static int getSubsamp(j_decompress_ptr dinfo)
+{
+	int retval=-1, i, k;
+
+	/* The sampling factors actually have no meaning with grayscale JPEG files,
+	   and in fact it's possible to generate grayscale JPEGs with sampling
+	   factors > 1 (even though those sampling factors are ignored by the
+	   decompressor.)  Thus, we need to treat grayscale as a special case. */
+	if(dinfo->num_components==1 && dinfo->jpeg_color_space==JCS_GRAYSCALE)
+		return TJSAMP_GRAY;
+
+	for(i=0; i<NUMSUBOPT; i++)
+	{
+		if(dinfo->num_components==pixelsize[i]
+			|| ((dinfo->jpeg_color_space==JCS_YCCK
+				|| dinfo->jpeg_color_space==JCS_CMYK)
+					&& pixelsize[i]==3 && dinfo->num_components==4))
+		{
+			if(dinfo->comp_info[0].h_samp_factor==tjMCUWidth[i]/8
+				&& dinfo->comp_info[0].v_samp_factor==tjMCUHeight[i]/8)
+			{
+				int match=0;
+				for(k=1; k<dinfo->num_components; k++)
+				{
+					int href=1, vref=1;
+					if(dinfo->jpeg_color_space==JCS_YCCK && k==3)
+					{
+						href=tjMCUWidth[i]/8;  vref=tjMCUHeight[i]/8;
+					}
+					if(dinfo->comp_info[k].h_samp_factor==href
+						&& dinfo->comp_info[k].v_samp_factor==vref)
+						match++;
+				}
+				if(match==dinfo->num_components-1)
+				{
+					retval=i;  break;
+				}
+			}
+		}
+	}
+	return retval;
+}
+
+
+#ifndef JCS_EXTENSIONS
+
+/* Conversion functions to emulate the colorspace extensions.  This allows the
+   TurboJPEG wrapper to be used with libjpeg */
+
+#define TORGB(PS, ROFFSET, GOFFSET, BOFFSET) {  \
+	int rowPad=pitch-width*PS;  \
+	while(height--)  \
+	{  \
+		unsigned char *endOfRow=src+width*PS;  \
+		while(src<endOfRow)  \
+		{  \
+			dst[RGB_RED]=src[ROFFSET];  \
+			dst[RGB_GREEN]=src[GOFFSET];  \
+			dst[RGB_BLUE]=src[BOFFSET];  \
+			dst+=RGB_PIXELSIZE;  src+=PS;  \
+		}  \
+		src+=rowPad;  \
+	}  \
+}
+
+static unsigned char *toRGB(unsigned char *src, int width, int pitch,
+	int height, int pixelFormat, unsigned char *dst)
+{
+	unsigned char *retval=src;
+	switch(pixelFormat)
+	{
+		case TJPF_RGB:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=3
+			retval=dst;  TORGB(3, 0, 1, 2);
+			#endif
+			break;
+		case TJPF_BGR:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=3
+			retval=dst;  TORGB(3, 2, 1, 0);
+			#endif
+			break;
+		case TJPF_RGBX:
+		case TJPF_RGBA:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 0, 1, 2);
+			#endif
+			break;
+		case TJPF_BGRX:
+		case TJPF_BGRA:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 2, 1, 0);
+			#endif
+			break;
+		case TJPF_XRGB:
+		case TJPF_ARGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 1, 2, 3);
+			#endif
+			break;
+		case TJPF_XBGR:
+		case TJPF_ABGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			retval=dst;  TORGB(4, 3, 2, 1);
+			#endif
+			break;
+	}
+	return retval;
+}
+
+#define FROMRGB(PS, ROFFSET, GOFFSET, BOFFSET, SETALPHA) {  \
+	int rowPad=pitch-width*PS;  \
+	while(height--)  \
+	{  \
+		unsigned char *endOfRow=dst+width*PS;  \
+		while(dst<endOfRow)  \
+		{  \
+			dst[ROFFSET]=src[RGB_RED];  \
+			dst[GOFFSET]=src[RGB_GREEN];  \
+			dst[BOFFSET]=src[RGB_BLUE];  \
+			SETALPHA  \
+			dst+=PS;  src+=RGB_PIXELSIZE;  \
+		}  \
+		dst+=rowPad;  \
+	}  \
+}
+
+static void fromRGB(unsigned char *src, unsigned char *dst, int width,
+	int pitch, int height, int pixelFormat)
+{
+	switch(pixelFormat)
+	{
+		case TJPF_RGB:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=3
+			FROMRGB(3, 0, 1, 2,);
+			#endif
+			break;
+		case TJPF_BGR:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=3
+			FROMRGB(3, 2, 1, 0,);
+			#endif
+			break;
+		case TJPF_RGBX:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 0, 1, 2,);
+			#endif
+			break;
+		case TJPF_RGBA:
+			#if RGB_RED!=0 || RGB_GREEN!=1 || RGB_BLUE!=2 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 0, 1, 2, dst[3]=0xFF;);
+			#endif
+			break;
+		case TJPF_BGRX:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 2, 1, 0,);
+			#endif
+			break;
+		case TJPF_BGRA:
+			#if RGB_RED!=2 || RGB_GREEN!=1 || RGB_BLUE!=0 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 2, 1, 0, dst[3]=0xFF;);  return;
+			#endif
+			break;
+		case TJPF_XRGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 1, 2, 3,);  return;
+			#endif
+			break;
+		case TJPF_ARGB:
+			#if RGB_RED!=1 || RGB_GREEN!=2 || RGB_BLUE!=3 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 1, 2, 3, dst[0]=0xFF;);  return;
+			#endif
+			break;
+		case TJPF_XBGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 3, 2, 1,);  return;
+			#endif
+			break;
+		case TJPF_ABGR:
+			#if RGB_RED!=3 || RGB_GREEN!=2 || RGB_BLUE!=1 || RGB_PIXELSIZE!=4
+			FROMRGB(4, 3, 2, 1, dst[0]=0xFF;);  return;
+			#endif
+			break;
+	}
+}
+
+#endif
+
+
+/* General API functions */
+
+DLLEXPORT char* DLLCALL tjGetErrorStr(void)
+{
+	return errStr;
+}
+
+
+DLLEXPORT int DLLCALL tjDestroy(tjhandle handle)
+{
+	getinstance(handle);
+	if(setjmp(this->jerr.setjmp_buffer)) return -1;
+	if(this->init&COMPRESS) jpeg_destroy_compress(cinfo);
+	if(this->init&DECOMPRESS) jpeg_destroy_decompress(dinfo);
+	free(this);
+	return 0;
+}
+
+
+/* These are exposed mainly because Windows can't malloc() and free() across
+   DLL boundaries except when the CRT DLL is used, and we don't use the CRT DLL
+   with turbojpeg.dll for compatibility reasons.  However, these functions
+   can potentially be used for other purposes by different implementations. */
+
+DLLEXPORT void DLLCALL tjFree(unsigned char *buf)
+{
+	if(buf) free(buf);
+}
+
+
+DLLEXPORT unsigned char *DLLCALL tjAlloc(int bytes)
+{
+	return (unsigned char *)malloc(bytes);
+}
+
+
+/* Compressor  */
+
+static tjhandle _tjInitCompress(tjinstance *this)
+{
+	static unsigned char buffer[1];
+	unsigned char *buf=buffer;  unsigned long size=1;
+
+	/* This is also straight out of example.c */
+	this->cinfo.err=jpeg_std_error(&this->jerr.pub);
+	this->jerr.pub.error_exit=my_error_exit;
+	this->jerr.pub.output_message=my_output_message;
+	this->jerr.emit_message=this->jerr.pub.emit_message;
+	this->jerr.pub.emit_message=my_emit_message;
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		if(this) free(this);  return NULL;
+	}
+
+	jpeg_create_compress(&this->cinfo);
+	/* Make an initial call so it will create the destination manager */
+	jpeg_mem_dest_tj(&this->cinfo, &buf, &size, 0);
+
+	this->init|=COMPRESS;
+	return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle DLLCALL tjInitCompress(void)
+{
+	tjinstance *this=NULL;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitCompress(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	return _tjInitCompress(this);
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjBufSize(int width, int height,
+	int jpegSubsamp)
+{
+	unsigned long retval=0;  int mcuw, mcuh, chromasf;
+	if(width<1 || height<1 || jpegSubsamp<0 || jpegSubsamp>=NUMSUBOPT)
+		_throw("tjBufSize(): Invalid argument");
+
+	/* This allows for rare corner cases in which a JPEG image can actually be
+	   larger than the uncompressed input (we wouldn't mention it if it hadn't
+	   happened before.) */
+	mcuw=tjMCUWidth[jpegSubsamp];
+	mcuh=tjMCUHeight[jpegSubsamp];
+	chromasf=jpegSubsamp==TJSAMP_GRAY? 0: 4*64/(mcuw*mcuh);
+	retval=PAD(width, mcuw) * PAD(height, mcuh) * (2 + chromasf) + 2048;
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZE(int width, int height)
+{
+	unsigned long retval=0;
+	if(width<1 || height<1)
+		_throw("TJBUFSIZE(): Invalid argument");
+
+	/* This allows for rare corner cases in which a JPEG image can actually be
+	   larger than the uncompressed input (we wouldn't mention it if it hadn't
+	   happened before.) */
+	retval=PAD(width, 16) * PAD(height, 16) * 6 + 2048;
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV2(int width, int pad, int height,
+	int subsamp)
+{
+	int retval=0, nc, i;
+
+	if(subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjBufSizeYUV2(): Invalid argument");
+
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	for(i=0; i<nc; i++)
+	{
+		int pw=tjPlaneWidth(i, width, subsamp);
+		int stride=PAD(pw, pad);
+		int ph=tjPlaneHeight(i, height, subsamp);
+		if(pw<0 || ph<0) return -1;
+		else retval+=stride*ph;
+	}
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV(int width, int height,
+	int subsamp)
+{
+	return tjBufSizeYUV2(width, 4, height, subsamp);
+}
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZEYUV(int width, int height,
+	int subsamp)
+{
+	return tjBufSizeYUV(width, height, subsamp);
+}
+
+
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp)
+{
+	int pw, nc, retval=0;
+
+	if(width<1 || subsamp<0 || subsamp>=TJ_NUMSAMP)
+		_throw("tjPlaneWidth(): Invalid argument");
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	if(componentID<0 || componentID>=nc)
+		_throw("tjPlaneWidth(): Invalid argument");
+
+	pw=PAD(width, tjMCUWidth[subsamp]/8);
+	if(componentID==0)
+		retval=pw;
+	else
+		retval=pw*8/tjMCUWidth[subsamp];
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp)
+{
+	int ph, nc, retval=0;
+
+	if(height<1 || subsamp<0 || subsamp>=TJ_NUMSAMP)
+		_throw("tjPlaneHeight(): Invalid argument");
+	nc=(subsamp==TJSAMP_GRAY? 1:3);
+	if(componentID<0 || componentID>=nc)
+		_throw("tjPlaneHeight(): Invalid argument");
+
+	ph=PAD(height, tjMCUHeight[subsamp]/8);
+	if(componentID==0)
+		retval=ph;
+	else
+		retval=ph*8/tjMCUHeight[subsamp];
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT unsigned long DLLCALL tjPlaneSizeYUV(int componentID, int width,
+	int stride, int height, int subsamp)
+{
+	unsigned long retval=0;
+	int pw, ph;
+
+	if(width<1 || height<1 || subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjPlaneSizeYUV(): Invalid argument");
+
+	pw=tjPlaneWidth(componentID, width, subsamp);
+	ph=tjPlaneHeight(componentID, height, subsamp);
+	if(pw<0 || ph<0) return -1;
+
+	if(stride==0) stride=pw;
+	else stride=abs(stride);
+
+	retval=stride*(ph-1)+pw;
+
+	bailout:
+	return retval;
+}
+
+
+DLLEXPORT int DLLCALL tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelFormat, unsigned char **jpegBuf,
+	unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags)
+{
+	int i, retval=0, alloc=1;  JSAMPROW *row_pointer=NULL;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	#endif
+
+	getcinstance(handle)
+	if((this->init&COMPRESS)==0)
+		_throw("tjCompress2(): Instance has not been initialized for compression");
+
+	if(srcBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF || jpegBuf==NULL || jpegSize==NULL
+		|| jpegSubsamp<0 || jpegSubsamp>=NUMSUBOPT || jpegQual<0 || jpegQual>100)
+		_throw("tjCompress2(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK)
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*RGB_PIXELSIZE);
+		if(!rgbBuf) _throw("tjCompress2(): Memory allocation failure");
+		srcBuf=toRGB(srcBuf, width, pitch, height, pixelFormat, rgbBuf);
+		pitch=width*RGB_PIXELSIZE;
+	}
+	#endif
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(flags&TJFLAG_NOREALLOC)
+	{
+		alloc=0;  *jpegSize=tjBufSize(width, height, jpegSubsamp);
+	}
+	jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+	if(setCompDefaults(cinfo, pixelFormat, jpegSubsamp, jpegQual, flags)==-1)
+		return -1;
+
+	jpeg_start_compress(cinfo, TRUE);
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*height))==NULL)
+		_throw("tjCompress2(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=(JSAMPROW)&srcBuf[(height-i-1)*pitch];
+		else row_pointer[i]=(JSAMPROW)&srcBuf[i*pitch];
+	}
+	while(cinfo->next_scanline<cinfo->image_height)
+	{
+		jpeg_write_scanlines(cinfo, &row_pointer[cinfo->next_scanline],
+			cinfo->image_height-cinfo->next_scanline);
+	}
+	jpeg_finish_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjCompress(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelSize, unsigned char *jpegBuf,
+	unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags)
+{
+	int retval=0;  unsigned long size;
+	if(flags&TJ_YUV)
+	{
+		size=tjBufSizeYUV(width, height, jpegSubsamp);
+		retval=tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+			getPixelFormat(pixelSize, flags), jpegBuf, jpegSubsamp, flags);
+	}
+	else
+	{
+		retval=tjCompress2(handle, srcBuf, width, pitch, height,
+			getPixelFormat(pixelSize, flags), &jpegBuf, &size, jpegSubsamp, jpegQual,
+			flags|TJFLAG_NOREALLOC);
+	}
+	*jpegSize=size;
+	return retval;
+}
+
+
+DLLEXPORT int DLLCALL tjEncodeYUVPlanes(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pitch, int height,
+	int pixelFormat, unsigned char **dstPlanes, int *strides, int subsamp,
+	int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	JSAMPLE *_tmpbuf[MAX_COMPONENTS], *_tmpbuf2[MAX_COMPONENTS];
+	JSAMPROW *tmpbuf[MAX_COMPONENTS], *tmpbuf2[MAX_COMPONENTS];
+	JSAMPROW *outbuf[MAX_COMPONENTS];
+	int row, pw0, ph0, pw[MAX_COMPONENTS], ph[MAX_COMPONENTS];
+	JSAMPLE *ptr;
+	jpeg_component_info *compptr;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	#endif
+
+	getcinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  _tmpbuf[i]=NULL;
+		tmpbuf2[i]=NULL;  _tmpbuf2[i]=NULL;  outbuf[i]=NULL;
+	}
+
+	if((this->init&COMPRESS)==0)
+		_throw("tjEncodeYUVPlanes(): Instance has not been initialized for compression");
+
+	if(srcBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF || !dstPlanes || !dstPlanes[0] || subsamp<0
+		|| subsamp>=NUMSUBOPT)
+		_throw("tjEncodeYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!dstPlanes[1] || !dstPlanes[2]))
+		_throw("tjEncodeYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pixelFormat==TJPF_CMYK)
+		_throw("tjEncodeYUVPlanes(): Cannot generate YUV images from CMYK pixels");
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK)
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*RGB_PIXELSIZE);
+		if(!rgbBuf) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		srcBuf=toRGB(srcBuf, width, pitch, height, pixelFormat, rgbBuf);
+		pitch=width*RGB_PIXELSIZE;
+	}
+	#endif
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setCompDefaults(cinfo, pixelFormat, subsamp, -1, flags)==-1) return -1;
+
+	/* Execute only the parts of jpeg_start_compress() that we need.  If we
+	   were to call the whole jpeg_start_compress() function, then it would try
+	   to write the file headers, which could overflow the output buffer if the
+	   YUV image were very small. */
+	if(cinfo->global_state!=CSTATE_START)
+		_throw("tjEncodeYUVPlanes(): libjpeg API is in the wrong state");
+	(*cinfo->err->reset_error_mgr)((j_common_ptr)cinfo);
+	jinit_c_master_control(cinfo, FALSE);
+	jinit_color_converter(cinfo);
+	jinit_downsampler(cinfo);
+	(*cinfo->cconvert->start_pass)(cinfo);
+
+	pw0=PAD(width, cinfo->max_h_samp_factor);
+	ph0=PAD(height, cinfo->max_v_samp_factor);
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph0))==NULL)
+		_throw("tjEncodeYUVPlanes(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=(JSAMPROW)&srcBuf[(height-i-1)*pitch];
+		else row_pointer[i]=(JSAMPROW)&srcBuf[i*pitch];
+	}
+	if(height<ph0)
+		for(i=height; i<ph0; i++) row_pointer[i]=row_pointer[height-1];
+
+	for(i=0; i<cinfo->num_components; i++)
+	{
+		compptr=&cinfo->comp_info[i];
+		_tmpbuf[i]=(JSAMPLE *)malloc(
+			PAD((compptr->width_in_blocks*cinfo->max_h_samp_factor*DCTSIZE)
+				/compptr->h_samp_factor, 16) * cinfo->max_v_samp_factor + 16);
+		if(!_tmpbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*cinfo->max_v_samp_factor);
+		if(!tmpbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<cinfo->max_v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf[i], 16);
+			tmpbuf[i][row]=&_tmpbuf_aligned[
+				PAD((compptr->width_in_blocks*cinfo->max_h_samp_factor*DCTSIZE)
+					/compptr->h_samp_factor, 16) * row];
+		}
+		_tmpbuf2[i]=(JSAMPLE *)malloc(PAD(compptr->width_in_blocks*DCTSIZE, 16)
+			* compptr->v_samp_factor + 16);
+		if(!_tmpbuf2[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		tmpbuf2[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*compptr->v_samp_factor);
+		if(!tmpbuf2[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<compptr->v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf2_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf2[i], 16);
+			tmpbuf2[i][row]=&_tmpbuf2_aligned[
+				PAD(compptr->width_in_blocks*DCTSIZE, 16) * row];
+		}
+		pw[i]=pw0*compptr->h_samp_factor/cinfo->max_h_samp_factor;
+		ph[i]=ph0*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+		outbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]);
+		if(!outbuf[i]) _throw("tjEncodeYUVPlanes(): Memory allocation failure");
+		ptr=dstPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			outbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+
+	for(row=0; row<ph0; row+=cinfo->max_v_samp_factor)
+	{
+		(*cinfo->cconvert->color_convert)(cinfo, &row_pointer[row], tmpbuf, 0,
+			cinfo->max_v_samp_factor);
+		(cinfo->downsample->downsample)(cinfo, tmpbuf, 0, tmpbuf2, 0);
+		for(i=0, compptr=cinfo->comp_info; i<cinfo->num_components; i++, compptr++)
+			jcopy_sample_rows(tmpbuf2[i], 0, outbuf[i],
+				row*compptr->v_samp_factor/cinfo->max_v_samp_factor,
+				compptr->v_samp_factor, pw[i]);
+	}
+	cinfo->next_scanline+=height;
+	jpeg_abort_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]!=NULL) free(tmpbuf[i]);
+		if(_tmpbuf[i]!=NULL) free(_tmpbuf[i]);
+		if(tmpbuf2[i]!=NULL) free(tmpbuf2[i]);
+		if(_tmpbuf2[i]!=NULL) free(_tmpbuf2[i]);
+		if(outbuf[i]!=NULL) free(outbuf[i]);
+	}
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV3(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pitch, int height,
+	int pixelFormat, unsigned char *dstBuf, int pad, int subsamp, int flags)
+{
+	unsigned char *dstPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(width<=0 || height<=0 || dstBuf==NULL || pad<0 || !isPow2(pad)
+		|| subsamp<0 || subsamp>=NUMSUBOPT)
+		_throw("tjEncodeYUV3(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	dstPlanes[0]=dstBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		dstPlanes[1]=dstPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		dstPlanes[1]=dstPlanes[0]+strides[0]*ph0;
+		dstPlanes[2]=dstPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjEncodeYUVPlanes(handle, srcBuf, width, pitch, height, pixelFormat,
+		dstPlanes, strides, subsamp, flags);
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV2(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelFormat, unsigned char *dstBuf,
+	int subsamp, int flags)
+{
+	return tjEncodeYUV3(handle, srcBuf, width, pitch, height, pixelFormat,
+		dstBuf, 4, subsamp, flags);
+}
+
+DLLEXPORT int DLLCALL tjEncodeYUV(tjhandle handle, unsigned char *srcBuf,
+	int width, int pitch, int height, int pixelSize, unsigned char *dstBuf,
+	int subsamp, int flags)
+{
+	return tjEncodeYUV2(handle, srcBuf, width, pitch, height,
+		getPixelFormat(pixelSize, flags), dstBuf, subsamp, flags);
+}
+
+
+DLLEXPORT int DLLCALL tjCompressFromYUVPlanes(tjhandle handle,
+	const unsigned char **srcPlanes, int width, const int *strides, int height,
+	int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual,
+	int flags)
+{
+	int i, row, retval=0, alloc=1;  JSAMPROW *inbuf[MAX_COMPONENTS];
+	int pw[MAX_COMPONENTS], ph[MAX_COMPONENTS], iw[MAX_COMPONENTS],
+		tmpbufsize=0, usetmpbuf=0, th[MAX_COMPONENTS];
+	JSAMPLE *_tmpbuf=NULL, *ptr;  JSAMPROW *tmpbuf[MAX_COMPONENTS];
+
+	getcinstance(handle)
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  inbuf[i]=NULL;
+	}
+
+	if((this->init&COMPRESS)==0)
+		_throw("tjCompressFromYUVPlanes(): Instance has not been initialized for compression");
+
+	if(!srcPlanes || !srcPlanes[0] || width<=0 || height<=0 || subsamp<0
+		|| subsamp>=NUMSUBOPT || jpegBuf==NULL || jpegSize==NULL || jpegQual<0
+		|| jpegQual>100)
+		_throw("tjCompressFromYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!srcPlanes[1] || !srcPlanes[2]))
+		_throw("tjCompressFromYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	cinfo->image_width=width;
+	cinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(flags&TJFLAG_NOREALLOC)
+	{
+		alloc=0;  *jpegSize=tjBufSize(width, height, subsamp);
+	}
+	jpeg_mem_dest_tj(cinfo, jpegBuf, jpegSize, alloc);
+	if(setCompDefaults(cinfo, TJPF_RGB, subsamp, jpegQual, flags)==-1)
+		return -1;
+	cinfo->raw_data_in=TRUE;
+
+	jpeg_start_compress(cinfo, TRUE);
+	for(i=0; i<cinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&cinfo->comp_info[i];
+		int ih;
+		iw[i]=compptr->width_in_blocks*DCTSIZE;
+		ih=compptr->height_in_blocks*DCTSIZE;
+		pw[i]=PAD(cinfo->image_width, cinfo->max_h_samp_factor)
+			*compptr->h_samp_factor/cinfo->max_h_samp_factor;
+		ph[i]=PAD(cinfo->image_height, cinfo->max_v_samp_factor)
+			*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+		if(iw[i]!=pw[i] || ih!=ph[i]) usetmpbuf=1;
+		th[i]=compptr->v_samp_factor*DCTSIZE;
+		tmpbufsize+=iw[i]*th[i];
+		if((inbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]))==NULL)
+			_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+		ptr=(JSAMPLE *)srcPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			inbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+	if(usetmpbuf)
+	{
+		if((_tmpbuf=(JSAMPLE *)malloc(sizeof(JSAMPLE)*tmpbufsize))==NULL)
+			_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+		ptr=_tmpbuf;
+		for(i=0; i<cinfo->num_components; i++)
+		{
+			if((tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*th[i]))==NULL)
+				_throw("tjCompressFromYUVPlanes(): Memory allocation failure");
+			for(row=0; row<th[i]; row++)
+			{
+				tmpbuf[i][row]=ptr;
+				ptr+=iw[i];
+			}
+		}
+	}
+
+	for(row=0; row<(int)cinfo->image_height;
+		row+=cinfo->max_v_samp_factor*DCTSIZE)
+	{
+		JSAMPARRAY yuvptr[MAX_COMPONENTS];
+		int crow[MAX_COMPONENTS];
+		for(i=0; i<cinfo->num_components; i++)
+		{
+			jpeg_component_info *compptr=&cinfo->comp_info[i];
+			crow[i]=row*compptr->v_samp_factor/cinfo->max_v_samp_factor;
+			if(usetmpbuf)
+			{
+				int j, k;
+				for(j=0; j<min(th[i], ph[i]-crow[i]); j++)
+				{
+					memcpy(tmpbuf[i][j], inbuf[i][crow[i]+j], pw[i]);
+					/* Duplicate last sample in row to fill out MCU */
+					for(k=pw[i]; k<iw[i]; k++) tmpbuf[i][j][k]=tmpbuf[i][j][pw[i]-1];
+				}
+				/* Duplicate last row to fill out MCU */
+				for(j=ph[i]-crow[i]; j<th[i]; j++)
+					memcpy(tmpbuf[i][j], tmpbuf[i][ph[i]-crow[i]-1], iw[i]);
+				yuvptr[i]=tmpbuf[i];
+			}
+			else
+				yuvptr[i]=&inbuf[i][crow[i]];
+		}
+		jpeg_write_raw_data(cinfo, yuvptr, cinfo->max_v_samp_factor*DCTSIZE);
+	}
+	jpeg_finish_compress(cinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]) free(tmpbuf[i]);
+		if(inbuf[i]) free(inbuf[i]);
+	}
+	if(_tmpbuf) free(_tmpbuf);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjCompressFromYUV(tjhandle handle,
+	const unsigned char *srcBuf, int width, int pad, int height, int subsamp,
+	unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags)
+{
+	const unsigned char *srcPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(srcBuf==NULL || width<=0 || pad<1 || height<=0 || subsamp<0
+		|| subsamp>=NUMSUBOPT)
+		_throw("tjCompressFromYUV(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	srcPlanes[0]=srcBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		srcPlanes[1]=srcPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		srcPlanes[1]=srcPlanes[0]+strides[0]*ph0;
+		srcPlanes[2]=srcPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjCompressFromYUVPlanes(handle, srcPlanes, width, strides, height,
+		subsamp, jpegBuf, jpegSize, jpegQual, flags);
+
+	bailout:
+	return retval;
+}
+
+
+/* Decompressor */
+
+static tjhandle _tjInitDecompress(tjinstance *this)
+{
+	static unsigned char buffer[1];
+
+	/* This is also straight out of example.c */
+	this->dinfo.err=jpeg_std_error(&this->jerr.pub);
+	this->jerr.pub.error_exit=my_error_exit;
+	this->jerr.pub.output_message=my_output_message;
+	this->jerr.emit_message=this->jerr.pub.emit_message;
+	this->jerr.pub.emit_message=my_emit_message;
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		if(this) free(this);  return NULL;
+	}
+
+	jpeg_create_decompress(&this->dinfo);
+	/* Make an initial call so it will create the source manager */
+	jpeg_mem_src_tj(&this->dinfo, buffer, 1);
+
+	this->init|=DECOMPRESS;
+	return (tjhandle)this;
+}
+
+DLLEXPORT tjhandle DLLCALL tjInitDecompress(void)
+{
+	tjinstance *this;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitDecompress(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	return _tjInitDecompress(this);
+}
+
+
+DLLEXPORT int DLLCALL tjDecompressHeader3(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, int *width,
+	int *height, int *jpegSubsamp, int *jpegColorspace)
+{
+	int retval=0;
+
+	getdinstance(handle);
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompressHeader3(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || width==NULL || height==NULL
+		|| jpegSubsamp==NULL || jpegColorspace==NULL)
+		_throw("tjDecompressHeader3(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		return -1;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+
+	*width=dinfo->image_width;
+	*height=dinfo->image_height;
+	*jpegSubsamp=getSubsamp(dinfo);
+	switch(dinfo->jpeg_color_space)
+	{
+		case JCS_GRAYSCALE:  *jpegColorspace=TJCS_GRAY;  break;
+		case JCS_RGB:        *jpegColorspace=TJCS_RGB;  break;
+		case JCS_YCbCr:      *jpegColorspace=TJCS_YCbCr;  break;
+		case JCS_CMYK:       *jpegColorspace=TJCS_CMYK;  break;
+		case JCS_YCCK:       *jpegColorspace=TJCS_YCCK;  break;
+		default:             *jpegColorspace=-1;  break;
+	}
+
+	jpeg_abort_decompress(dinfo);
+
+	if(*jpegSubsamp<0)
+		_throw("tjDecompressHeader3(): Could not determine subsampling type for JPEG image");
+	if(*jpegColorspace<0)
+		_throw("tjDecompressHeader3(): Could not determine colorspace of JPEG image");
+	if(*width<1 || *height<1)
+		_throw("tjDecompressHeader3(): Invalid data returned in header");
+
+	bailout:
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressHeader2(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height,
+	int *jpegSubsamp)
+{
+	int jpegColorspace;
+	return tjDecompressHeader3(handle, jpegBuf, jpegSize, width, height,
+		jpegSubsamp, &jpegColorspace);
+}
+
+DLLEXPORT int DLLCALL tjDecompressHeader(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height)
+{
+	int jpegSubsamp;
+	return tjDecompressHeader2(handle, jpegBuf, jpegSize, width, height,
+		&jpegSubsamp);
+}
+
+
+DLLEXPORT tjscalingfactor* DLLCALL tjGetScalingFactors(int *numscalingfactors)
+{
+	if(numscalingfactors==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjGetScalingFactors(): Invalid argument");
+		return NULL;
+	}
+
+	*numscalingfactors=NUMSF;
+	return (tjscalingfactor *)sf;
+}
+
+
+DLLEXPORT int DLLCALL tjDecompress2(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int width, int pitch, int height, int pixelFormat, int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	int jpegwidth, jpegheight, scaledw, scaledh;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	unsigned char *_dstBuf=NULL;  int _pitch=0;
+	#endif
+
+	getdinstance(handle);
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompress2(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || dstBuf==NULL || width<0 || pitch<0
+		|| height<0 || pixelFormat<0 || pixelFormat>=TJ_NUMPF)
+		_throw("tjDecompress2(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+	if(setDecompDefaults(dinfo, pixelFormat, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+
+	if(flags&TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling=FALSE;
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompress2(): Could not scale down to desired image dimensions");
+	width=scaledw;  height=scaledh;
+	dinfo->scale_num=sf[i].num;
+	dinfo->scale_denom=sf[i].denom;
+
+	jpeg_start_decompress(dinfo);
+	if(pitch==0) pitch=dinfo->output_width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK &&
+		(RGB_RED!=tjRedOffset[pixelFormat] ||
+			RGB_GREEN!=tjGreenOffset[pixelFormat] ||
+			RGB_BLUE!=tjBlueOffset[pixelFormat] ||
+			RGB_PIXELSIZE!=tjPixelSize[pixelFormat]))
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*3);
+		if(!rgbBuf) _throw("tjDecompress2(): Memory allocation failure");
+		_pitch=pitch;  pitch=width*3;
+		_dstBuf=dstBuf;  dstBuf=rgbBuf;
+	}
+	#endif
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)
+		*dinfo->output_height))==NULL)
+		_throw("tjDecompress2(): Memory allocation failure");
+	for(i=0; i<(int)dinfo->output_height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP)
+			row_pointer[i]=&dstBuf[(dinfo->output_height-i-1)*pitch];
+		else row_pointer[i]=&dstBuf[i*pitch];
+	}
+	while(dinfo->output_scanline<dinfo->output_height)
+	{
+		jpeg_read_scanlines(dinfo, &row_pointer[dinfo->output_scanline],
+			dinfo->output_height-dinfo->output_scanline);
+	}
+	jpeg_finish_decompress(dinfo);
+
+	#ifndef JCS_EXTENSIONS
+	fromRGB(rgbBuf, _dstBuf, width, _pitch, height, pixelFormat);
+	#endif
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompress(tjhandle handle, unsigned char *jpegBuf,
+	unsigned long jpegSize, unsigned char *dstBuf, int width, int pitch,
+	int height, int pixelSize, int flags)
+{
+	if(flags&TJ_YUV)
+		return tjDecompressToYUV(handle, jpegBuf, jpegSize, dstBuf, flags);
+	else
+		return tjDecompress2(handle, jpegBuf, jpegSize, dstBuf, width, pitch,
+			height, getPixelFormat(pixelSize, flags), flags);
+}
+
+
+static int setDecodeDefaults(struct jpeg_decompress_struct *dinfo,
+	int pixelFormat, int subsamp, int flags)
+{
+	int i;
+
+	dinfo->scale_num=dinfo->scale_denom=1;
+
+	if(subsamp==TJSAMP_GRAY)
+	{
+		dinfo->num_components=dinfo->comps_in_scan=1;
+		dinfo->jpeg_color_space=JCS_GRAYSCALE;
+	}
+	else
+	{
+		dinfo->num_components=dinfo->comps_in_scan=3;
+		dinfo->jpeg_color_space=JCS_YCbCr;
+	}
+
+	dinfo->comp_info=(jpeg_component_info *)
+		(*dinfo->mem->alloc_small)((j_common_ptr)dinfo, JPOOL_IMAGE,
+			dinfo->num_components*sizeof(jpeg_component_info));
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&dinfo->comp_info[i];
+		compptr->h_samp_factor=(i==0)? tjMCUWidth[subsamp]/8:1;
+		compptr->v_samp_factor=(i==0)? tjMCUHeight[subsamp]/8:1;
+		compptr->component_index=i;
+		compptr->component_id=i+1;
+		compptr->quant_tbl_no=compptr->dc_tbl_no=compptr->ac_tbl_no=
+			(i==0)? 0:1;
+		dinfo->cur_comp_info[i]=compptr;
+	}
+	dinfo->data_precision=8;
+	for(i=0; i<2; i++)
+	{
+		if(dinfo->quant_tbl_ptrs[i]==NULL)
+			dinfo->quant_tbl_ptrs[i]=jpeg_alloc_quant_table((j_common_ptr)dinfo);
+	}
+
+	return 0;
+}
+
+
+int my_read_markers(j_decompress_ptr dinfo)
+{
+	return JPEG_REACHED_SOS;
+}
+
+void my_reset_marker_reader(j_decompress_ptr dinfo)
+{
+}
+
+DLLEXPORT int DLLCALL tjDecodeYUVPlanes(tjhandle handle,
+	const unsigned char **srcPlanes, const int *strides, int subsamp,
+	unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat,
+	int flags)
+{
+	int i, retval=0;  JSAMPROW *row_pointer=NULL;
+	JSAMPLE *_tmpbuf[MAX_COMPONENTS];
+	JSAMPROW *tmpbuf[MAX_COMPONENTS], *inbuf[MAX_COMPONENTS];
+	int row, pw0, ph0, pw[MAX_COMPONENTS], ph[MAX_COMPONENTS];
+	JSAMPLE *ptr;
+	jpeg_component_info *compptr;
+	#ifndef JCS_EXTENSIONS
+	unsigned char *rgbBuf=NULL;
+	unsigned char *_dstBuf=NULL;  int _pitch=0;
+	#endif
+	int (*old_read_markers)(j_decompress_ptr);
+	void (*old_reset_marker_reader)(j_decompress_ptr);
+
+	getdinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  _tmpbuf[i]=NULL;  inbuf[i]=NULL;
+	}
+
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecodeYUVPlanes(): Instance has not been initialized for decompression");
+
+	if(!srcPlanes || !srcPlanes[0] || subsamp<0 || subsamp>=NUMSUBOPT
+		|| dstBuf==NULL || width<=0 || pitch<0 || height<=0 || pixelFormat<0
+		|| pixelFormat>=TJ_NUMPF)
+		_throw("tjDecodeYUVPlanes(): Invalid argument");
+	if(subsamp!=TJSAMP_GRAY && (!srcPlanes[1] || !srcPlanes[2]))
+		_throw("tjDecodeYUVPlanes(): Invalid argument");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(pixelFormat==TJPF_CMYK)
+		_throw("tjDecodeYUVPlanes(): Cannot decode YUV images into CMYK pixels.");
+
+	if(pitch==0) pitch=width*tjPixelSize[pixelFormat];
+	dinfo->image_width=width;
+	dinfo->image_height=height;
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setDecodeDefaults(dinfo, pixelFormat, subsamp, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+	old_read_markers=dinfo->marker->read_markers;
+	dinfo->marker->read_markers=my_read_markers;
+	old_reset_marker_reader=dinfo->marker->reset_marker_reader;
+	dinfo->marker->reset_marker_reader=my_reset_marker_reader;
+	jpeg_read_header(dinfo, TRUE);
+	dinfo->marker->read_markers=old_read_markers;
+	dinfo->marker->reset_marker_reader=old_reset_marker_reader;
+
+	if(setDecompDefaults(dinfo, pixelFormat, flags)==-1)
+	{
+		retval=-1;  goto bailout;
+	}
+	dinfo->do_fancy_upsampling=FALSE;
+	dinfo->Se=DCTSIZE2-1;
+	jinit_master_decompress(dinfo);
+	(*dinfo->upsample->start_pass)(dinfo);
+
+	pw0=PAD(width, dinfo->max_h_samp_factor);
+	ph0=PAD(height, dinfo->max_v_samp_factor);
+
+	if(pitch==0) pitch=dinfo->output_width*tjPixelSize[pixelFormat];
+
+	#ifndef JCS_EXTENSIONS
+	if(pixelFormat!=TJPF_GRAY && pixelFormat!=TJPF_CMYK &&
+		(RGB_RED!=tjRedOffset[pixelFormat] ||
+			RGB_GREEN!=tjGreenOffset[pixelFormat] ||
+			RGB_BLUE!=tjBlueOffset[pixelFormat] ||
+			RGB_PIXELSIZE!=tjPixelSize[pixelFormat]))
+	{
+		rgbBuf=(unsigned char *)malloc(width*height*3);
+		if(!rgbBuf) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		_pitch=pitch;  pitch=width*3;
+		_dstBuf=dstBuf;  dstBuf=rgbBuf;
+	}
+	#endif
+
+	if((row_pointer=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph0))==NULL)
+		_throw("tjDecodeYUVPlanes(): Memory allocation failure");
+	for(i=0; i<height; i++)
+	{
+		if(flags&TJFLAG_BOTTOMUP) row_pointer[i]=&dstBuf[(height-i-1)*pitch];
+		else row_pointer[i]=&dstBuf[i*pitch];
+	}
+	if(height<ph0)
+		for(i=height; i<ph0; i++) row_pointer[i]=row_pointer[height-1];
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		compptr=&dinfo->comp_info[i];
+		_tmpbuf[i]=(JSAMPLE *)malloc(PAD(compptr->width_in_blocks*DCTSIZE, 16)
+			* compptr->v_samp_factor + 16);
+		if(!_tmpbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*compptr->v_samp_factor);
+		if(!tmpbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		for(row=0; row<compptr->v_samp_factor; row++)
+		{
+			unsigned char *_tmpbuf_aligned=
+				(unsigned char *)PAD((size_t)_tmpbuf[i], 16);
+			tmpbuf[i][row]=&_tmpbuf_aligned[
+				PAD(compptr->width_in_blocks*DCTSIZE, 16) * row];
+		}
+		pw[i]=pw0*compptr->h_samp_factor/dinfo->max_h_samp_factor;
+		ph[i]=ph0*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+		inbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]);
+		if(!inbuf[i]) _throw("tjDecodeYUVPlanes(): Memory allocation failure");
+		ptr=(JSAMPLE *)srcPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			inbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+
+	for(row=0; row<ph0; row+=dinfo->max_v_samp_factor)
+	{
+		JDIMENSION inrow=0, outrow=0;
+		for(i=0, compptr=dinfo->comp_info; i<dinfo->num_components; i++, compptr++)
+			jcopy_sample_rows(inbuf[i],
+				row*compptr->v_samp_factor/dinfo->max_v_samp_factor, tmpbuf[i], 0,
+				compptr->v_samp_factor, pw[i]);
+		(dinfo->upsample->upsample)(dinfo, tmpbuf, &inrow,
+			dinfo->max_v_samp_factor, &row_pointer[row], &outrow,
+			dinfo->max_v_samp_factor);
+	}
+	jpeg_abort_decompress(dinfo);
+
+	#ifndef JCS_EXTENSIONS
+	fromRGB(rgbBuf, _dstBuf, width, _pitch, height, pixelFormat);
+	#endif
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	#ifndef JCS_EXTENSIONS
+	if(rgbBuf) free(rgbBuf);
+	#endif
+	if(row_pointer) free(row_pointer);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]!=NULL) free(tmpbuf[i]);
+		if(_tmpbuf[i]!=NULL) free(_tmpbuf[i]);
+		if(inbuf[i]!=NULL) free(inbuf[i]);
+	}
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+	int pad, int subsamp, unsigned char *dstBuf, int width, int pitch,
+	int height, int pixelFormat, int flags)
+{
+	const unsigned char *srcPlanes[3];
+	int pw0, ph0, strides[3], retval=-1;
+
+	if(srcBuf==NULL || pad<0 || !isPow2(pad) || subsamp<0 || subsamp>=NUMSUBOPT
+		|| width<=0 || height<=0)
+		_throw("tjDecodeYUV(): Invalid argument");
+
+	pw0=tjPlaneWidth(0, width, subsamp);
+	ph0=tjPlaneHeight(0, height, subsamp);
+	srcPlanes[0]=srcBuf;
+	strides[0]=PAD(pw0, pad);
+	if(subsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		srcPlanes[1]=srcPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, subsamp);
+		int ph1=tjPlaneHeight(1, height, subsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		srcPlanes[1]=srcPlanes[0]+strides[0]*ph0;
+		srcPlanes[2]=srcPlanes[1]+strides[1]*ph1;
+	}
+
+	return tjDecodeYUVPlanes(handle, srcPlanes, strides, subsamp, dstBuf, width,
+		pitch, height, pixelFormat, flags);
+
+	bailout:
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUVPlanes(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize,
+	unsigned char **dstPlanes, int width, int *strides, int height, int flags)
+{
+	int i, sfi, row, retval=0;  JSAMPROW *outbuf[MAX_COMPONENTS];
+	int jpegwidth, jpegheight, jpegSubsamp, scaledw, scaledh;
+	int pw[MAX_COMPONENTS], ph[MAX_COMPONENTS], iw[MAX_COMPONENTS],
+		tmpbufsize=0, usetmpbuf=0, th[MAX_COMPONENTS];
+	JSAMPLE *_tmpbuf=NULL, *ptr;  JSAMPROW *tmpbuf[MAX_COMPONENTS];
+	int dctsize;
+
+	getdinstance(handle);
+
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		tmpbuf[i]=NULL;  outbuf[i]=NULL;
+	}
+
+	if((this->init&DECOMPRESS)==0)
+		_throw("tjDecompressToYUVPlanes(): Instance has not been initialized for decompression");
+
+	if(jpegBuf==NULL || jpegSize<=0 || !dstPlanes || !dstPlanes[0] || width<0
+		|| height<0)
+		_throw("tjDecompressToYUVPlanes(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	if(!this->headerRead)
+	{
+		jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+		jpeg_read_header(dinfo, TRUE);
+	}
+	this->headerRead=0;
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjDecompressToYUVPlanes(): Could not determine subsampling type for JPEG image");
+
+	if(jpegSubsamp!=TJSAMP_GRAY && (!dstPlanes[1] || !dstPlanes[2]))
+		_throw("tjDecompressToYUVPlanes(): Invalid argument");
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompressToYUVPlanes(): Could not scale down to desired image dimensions");
+	if(dinfo->num_components>3)
+		_throw("tjDecompressToYUVPlanes(): JPEG image must have 3 or fewer components");
+
+	width=scaledw;  height=scaledh;
+	dinfo->scale_num=sf[i].num;
+	dinfo->scale_denom=sf[i].denom;
+	sfi=i;
+	jpeg_calc_output_dimensions(dinfo);
+
+	dctsize=DCTSIZE*sf[sfi].num/sf[sfi].denom;
+
+	for(i=0; i<dinfo->num_components; i++)
+	{
+		jpeg_component_info *compptr=&dinfo->comp_info[i];
+		int ih;
+		iw[i]=compptr->width_in_blocks*dctsize;
+		ih=compptr->height_in_blocks*dctsize;
+		pw[i]=PAD(dinfo->output_width, dinfo->max_h_samp_factor)
+			*compptr->h_samp_factor/dinfo->max_h_samp_factor;
+		ph[i]=PAD(dinfo->output_height, dinfo->max_v_samp_factor)
+			*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+		if(iw[i]!=pw[i] || ih!=ph[i]) usetmpbuf=1;
+		th[i]=compptr->v_samp_factor*dctsize;
+		tmpbufsize+=iw[i]*th[i];
+		if((outbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*ph[i]))==NULL)
+			_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+		ptr=dstPlanes[i];
+		for(row=0; row<ph[i]; row++)
+		{
+			outbuf[i][row]=ptr;
+			ptr+=(strides && strides[i]!=0)? strides[i]:pw[i];
+		}
+	}
+	if(usetmpbuf)
+	{
+		if((_tmpbuf=(JSAMPLE *)malloc(sizeof(JSAMPLE)*tmpbufsize))==NULL)
+			_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+		ptr=_tmpbuf;
+		for(i=0; i<dinfo->num_components; i++)
+		{
+			if((tmpbuf[i]=(JSAMPROW *)malloc(sizeof(JSAMPROW)*th[i]))==NULL)
+				_throw("tjDecompressToYUVPlanes(): Memory allocation failure");
+			for(row=0; row<th[i]; row++)
+			{
+				tmpbuf[i][row]=ptr;
+				ptr+=iw[i];
+			}
+		}
+	}
+
+	if(flags&TJFLAG_FASTUPSAMPLE) dinfo->do_fancy_upsampling=FALSE;
+	if(flags&TJFLAG_FASTDCT) dinfo->dct_method=JDCT_FASTEST;
+	dinfo->raw_data_out=TRUE;
+
+	jpeg_start_decompress(dinfo);
+	for(row=0; row<(int)dinfo->output_height;
+		row+=dinfo->max_v_samp_factor*dinfo->_min_DCT_scaled_size)
+	{
+		JSAMPARRAY yuvptr[MAX_COMPONENTS];
+		int crow[MAX_COMPONENTS];
+		for(i=0; i<dinfo->num_components; i++)
+		{
+			jpeg_component_info *compptr=&dinfo->comp_info[i];
+			if(jpegSubsamp==TJ_420)
+			{
+				/* When 4:2:0 subsampling is used with IDCT scaling, libjpeg will try
+				   to be clever and use the IDCT to perform upsampling on the U and V
+				   planes.  For instance, if the output image is to be scaled by 1/2
+				   relative to the JPEG image, then the scaling factor and upsampling
+				   effectively cancel each other, so a normal 8x8 IDCT can be used.
+				   However, this is not desirable when using the decompress-to-YUV
+				   functionality in TurboJPEG, since we want to output the U and V
+				   planes in their subsampled form.  Thus, we have to override some
+				   internal libjpeg parameters to force it to use the "scaled" IDCT
+				   functions on the U and V planes. */
+				compptr->_DCT_scaled_size=dctsize;
+				compptr->MCU_sample_width=tjMCUWidth[jpegSubsamp]*
+					sf[sfi].num/sf[sfi].denom*
+					compptr->v_samp_factor/dinfo->max_v_samp_factor;
+				dinfo->idct->inverse_DCT[i] = dinfo->idct->inverse_DCT[0];
+			}
+			crow[i]=row*compptr->v_samp_factor/dinfo->max_v_samp_factor;
+			if(usetmpbuf) yuvptr[i]=tmpbuf[i];
+			else yuvptr[i]=&outbuf[i][crow[i]];
+		}
+		jpeg_read_raw_data(dinfo, yuvptr,
+			dinfo->max_v_samp_factor*dinfo->_min_DCT_scaled_size);
+		if(usetmpbuf)
+		{
+			int j;
+			for(i=0; i<dinfo->num_components; i++)
+			{
+				for(j=0; j<min(th[i], ph[i]-crow[i]); j++)
+				{
+					memcpy(outbuf[i][crow[i]+j], tmpbuf[i][j], pw[i]);
+				}
+			}
+		}
+	}
+	jpeg_finish_decompress(dinfo);
+
+	bailout:
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	for(i=0; i<MAX_COMPONENTS; i++)
+	{
+		if(tmpbuf[i]) free(tmpbuf[i]);
+		if(outbuf[i]) free(outbuf[i]);
+	}
+	if(_tmpbuf) free(_tmpbuf);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUV2(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int width, int pad, int height, int flags)
+{
+	unsigned char *dstPlanes[3];
+	int pw0, ph0, strides[3], retval=-1, jpegSubsamp=-1;
+	int i, jpegwidth, jpegheight, scaledw, scaledh;
+
+	getdinstance(handle);
+
+	if(jpegBuf==NULL || jpegSize<=0 || dstBuf==NULL || width<0 || pad<1
+		|| !isPow2(pad) || height<0)
+		_throw("tjDecompressToYUV2(): Invalid argument");
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+	jpeg_read_header(dinfo, TRUE);
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjDecompressToYUV2(): Could not determine subsampling type for JPEG image");
+
+	jpegwidth=dinfo->image_width;  jpegheight=dinfo->image_height;
+	if(width==0) width=jpegwidth;
+	if(height==0) height=jpegheight;
+
+	for(i=0; i<NUMSF; i++)
+	{
+		scaledw=TJSCALED(jpegwidth, sf[i]);
+		scaledh=TJSCALED(jpegheight, sf[i]);
+		if(scaledw<=width && scaledh<=height)
+			break;
+	}
+	if(i>=NUMSF)
+		_throw("tjDecompressToYUV2(): Could not scale down to desired image dimensions");
+
+	pw0=tjPlaneWidth(0, width, jpegSubsamp);
+	ph0=tjPlaneHeight(0, height, jpegSubsamp);
+	dstPlanes[0]=dstBuf;
+	strides[0]=PAD(pw0, pad);
+	if(jpegSubsamp==TJSAMP_GRAY)
+	{
+		strides[1]=strides[2]=0;
+		dstPlanes[1]=dstPlanes[2]=NULL;
+	}
+	else
+	{
+		int pw1=tjPlaneWidth(1, width, jpegSubsamp);
+		int ph1=tjPlaneHeight(1, height, jpegSubsamp);
+		strides[1]=strides[2]=PAD(pw1, pad);
+		dstPlanes[1]=dstPlanes[0]+strides[0]*ph0;
+		dstPlanes[2]=dstPlanes[1]+strides[1]*ph1;
+	}
+
+	this->headerRead=1;
+	return tjDecompressToYUVPlanes(handle, jpegBuf, jpegSize, dstPlanes, width,
+		strides, height, flags);
+
+	bailout:
+	return retval;
+
+}
+
+DLLEXPORT int DLLCALL tjDecompressToYUV(tjhandle handle,
+	unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+	int flags)
+{
+	return tjDecompressToYUV2(handle, jpegBuf, jpegSize, dstBuf, 0, 4, 0, flags);
+}
+
+
+/* Transformer */
+
+DLLEXPORT tjhandle DLLCALL tjInitTransform(void)
+{
+	tjinstance *this=NULL;  tjhandle handle=NULL;
+	if((this=(tjinstance *)malloc(sizeof(tjinstance)))==NULL)
+	{
+		snprintf(errStr, JMSG_LENGTH_MAX,
+			"tjInitTransform(): Memory allocation failure");
+		return NULL;
+	}
+	MEMZERO(this, sizeof(tjinstance));
+	handle=_tjInitCompress(this);
+	if(!handle) return NULL;
+	handle=_tjInitDecompress(this);
+	return handle;
+}
+
+
+DLLEXPORT int DLLCALL tjTransform(tjhandle handle,
+	const unsigned char *jpegBuf, unsigned long jpegSize, int n,
+	unsigned char **dstBufs, unsigned long *dstSizes, tjtransform *t, int flags)
+{
+	jpeg_transform_info *xinfo=NULL;
+	jvirt_barray_ptr *srccoefs, *dstcoefs;
+	int retval=0, i, jpegSubsamp;
+
+	getinstance(handle);
+	if((this->init&COMPRESS)==0 || (this->init&DECOMPRESS)==0)
+		_throw("tjTransform(): Instance has not been initialized for transformation");
+
+	if(jpegBuf==NULL || jpegSize<=0 || n<1 || dstBufs==NULL || dstSizes==NULL
+		|| t==NULL || flags<0)
+		_throw("tjTransform(): Invalid argument");
+
+	if(flags&TJFLAG_FORCEMMX) putenv("JSIMD_FORCEMMX=1");
+	else if(flags&TJFLAG_FORCESSE) putenv("JSIMD_FORCESSE=1");
+	else if(flags&TJFLAG_FORCESSE2) putenv("JSIMD_FORCESSE2=1");
+
+	if(setjmp(this->jerr.setjmp_buffer))
+	{
+		/* If we get here, the JPEG code has signaled an error. */
+		retval=-1;
+		goto bailout;
+	}
+
+	jpeg_mem_src_tj(dinfo, jpegBuf, jpegSize);
+
+	if((xinfo=(jpeg_transform_info *)malloc(sizeof(jpeg_transform_info)*n))
+		==NULL)
+		_throw("tjTransform(): Memory allocation failure");
+	MEMZERO(xinfo, sizeof(jpeg_transform_info)*n);
+
+	for(i=0; i<n; i++)
+	{
+		xinfo[i].transform=xformtypes[t[i].op];
+		xinfo[i].perfect=(t[i].options&TJXOPT_PERFECT)? 1:0;
+		xinfo[i].trim=(t[i].options&TJXOPT_TRIM)? 1:0;
+		xinfo[i].force_grayscale=(t[i].options&TJXOPT_GRAY)? 1:0;
+		xinfo[i].crop=(t[i].options&TJXOPT_CROP)? 1:0;
+		if(n!=1 && t[i].op==TJXOP_HFLIP) xinfo[i].slow_hflip=1;
+		else xinfo[i].slow_hflip=0;
+
+		if(xinfo[i].crop)
+		{
+			xinfo[i].crop_xoffset=t[i].r.x;  xinfo[i].crop_xoffset_set=JCROP_POS;
+			xinfo[i].crop_yoffset=t[i].r.y;  xinfo[i].crop_yoffset_set=JCROP_POS;
+			if(t[i].r.w!=0)
+			{
+				xinfo[i].crop_width=t[i].r.w;  xinfo[i].crop_width_set=JCROP_POS;
+			}
+			else xinfo[i].crop_width=JCROP_UNSET;
+			if(t[i].r.h!=0)
+			{
+				xinfo[i].crop_height=t[i].r.h;  xinfo[i].crop_height_set=JCROP_POS;
+			}
+			else xinfo[i].crop_height=JCROP_UNSET;
+		}
+	}
+
+	jcopy_markers_setup(dinfo, JCOPYOPT_ALL);
+	jpeg_read_header(dinfo, TRUE);
+	jpegSubsamp=getSubsamp(dinfo);
+	if(jpegSubsamp<0)
+		_throw("tjTransform(): Could not determine subsampling type for JPEG image");
+
+	for(i=0; i<n; i++)
+	{
+		if(!jtransform_request_workspace(dinfo, &xinfo[i]))
+			_throw("tjTransform(): Transform is not perfect");
+
+		if(xinfo[i].crop)
+		{
+			if((t[i].r.x%xinfo[i].iMCU_sample_width)!=0
+				|| (t[i].r.y%xinfo[i].iMCU_sample_height)!=0)
+			{
+				snprintf(errStr, JMSG_LENGTH_MAX,
+					"To crop this JPEG image, x must be a multiple of %d\n"
+					"and y must be a multiple of %d.\n",
+					xinfo[i].iMCU_sample_width, xinfo[i].iMCU_sample_height);
+				retval=-1;  goto bailout;
+			}
+		}
+	}
+
+	srccoefs=jpeg_read_coefficients(dinfo);
+
+	for(i=0; i<n; i++)
+	{
+		int w, h, alloc=1;
+		if(!xinfo[i].crop)
+		{
+			w=dinfo->image_width;  h=dinfo->image_height;
+		}
+		else
+		{
+			w=xinfo[i].crop_width;  h=xinfo[i].crop_height;
+		}
+		if(flags&TJFLAG_NOREALLOC)
+		{
+			alloc=0;  dstSizes[i]=tjBufSize(w, h, jpegSubsamp);
+		}
+		if(!(t[i].options&TJXOPT_NOOUTPUT))
+			jpeg_mem_dest_tj(cinfo, &dstBufs[i], &dstSizes[i], alloc);
+		jpeg_copy_critical_parameters(dinfo, cinfo);
+		dstcoefs=jtransform_adjust_parameters(dinfo, cinfo, srccoefs,
+			&xinfo[i]);
+		if(!(t[i].options&TJXOPT_NOOUTPUT))
+		{
+			jpeg_write_coefficients(cinfo, dstcoefs);
+			jcopy_markers_execute(dinfo, cinfo, JCOPYOPT_ALL);
+		}
+		else jinit_c_master_control(cinfo, TRUE);
+		jtransform_execute_transformation(dinfo, cinfo, srccoefs,
+			&xinfo[i]);
+		if(t[i].customFilter)
+		{
+			int ci, y;  JDIMENSION by;
+			for(ci=0; ci<cinfo->num_components; ci++)
+			{
+				jpeg_component_info *compptr=&cinfo->comp_info[ci];
+				tjregion arrayRegion={0, 0, compptr->width_in_blocks*DCTSIZE,
+					DCTSIZE};
+				tjregion planeRegion={0, 0, compptr->width_in_blocks*DCTSIZE,
+					compptr->height_in_blocks*DCTSIZE};
+				for(by=0; by<compptr->height_in_blocks; by+=compptr->v_samp_factor)
+				{
+					JBLOCKARRAY barray=(dinfo->mem->access_virt_barray)
+						((j_common_ptr)dinfo, dstcoefs[ci], by, compptr->v_samp_factor,
+						TRUE);
+					for(y=0; y<compptr->v_samp_factor; y++)
+					{
+						if(t[i].customFilter(barray[y][0], arrayRegion, planeRegion,
+							ci, i, &t[i])==-1)
+							_throw("tjTransform(): Error in custom filter");
+						arrayRegion.y+=DCTSIZE;
+					}
+				}
+			}
+		}
+		if(!(t[i].options&TJXOPT_NOOUTPUT)) jpeg_finish_compress(cinfo);
+	}
+
+	jpeg_finish_decompress(dinfo);
+
+	bailout:
+	if(cinfo->global_state>CSTATE_START) jpeg_abort_compress(cinfo);
+	if(dinfo->global_state>DSTATE_START) jpeg_abort_decompress(dinfo);
+	if(xinfo) free(xinfo);
+	if(this->jerr.warning) retval=-1;
+	return retval;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg.h glmark2-2021.02/src/libjpeg-turbo/turbojpeg.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/turbojpeg.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1542 @@
+/*
+ * Copyright (C)2009-2015 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __TURBOJPEG_H__
+#define __TURBOJPEG_H__
+
+#if defined(_WIN32) && defined(DLLDEFINE)
+#define DLLEXPORT __declspec(dllexport)
+#else
+#define DLLEXPORT
+#endif
+#define DLLCALL
+
+
+/**
+ * @addtogroup TurboJPEG
+ * TurboJPEG API.  This API provides an interface for generating, decoding, and
+ * transforming planar YUV and JPEG images in memory.
+ *
+ * @anchor YUVnotes
+ * YUV Image Format Notes
+ * ----------------------
+ * Technically, the JPEG format uses the YCbCr colorspace (which is technically
+ * not a colorspace but a color transform), but per the convention of the
+ * digital video community, the TurboJPEG API uses "YUV" to refer to an image
+ * format consisting of Y, Cb, and Cr image planes.
+ *
+ * Each plane is simply a 2D array of bytes, each byte representing the value
+ * of one of the components (Y, Cb, or Cr) at a particular location in the
+ * image.  The width and height of each plane are determined by the image
+ * width, height, and level of chrominance subsampling.   The luminance plane
+ * width is the image width padded to the nearest multiple of the horizontal
+ * subsampling factor (2 in the case of 4:2:0 and 4:2:2, 4 in the case of
+ * 4:1:1, 1 in the case of 4:4:4 or grayscale.)  Similarly, the luminance plane
+ * height is the image height padded to the nearest multiple of the vertical
+ * subsampling factor (2 in the case of 4:2:0 or 4:4:0, 1 in the case of 4:4:4
+ * or grayscale.)  This is irrespective of any additional padding that may be
+ * specified as an argument to the various YUV functions.  The chrominance
+ * plane width is equal to the luminance plane width divided by the horizontal
+ * subsampling factor, and the chrominance plane height is equal to the
+ * luminance plane height divided by the vertical subsampling factor.
+ *
+ * For example, if the source image is 35 x 35 pixels and 4:2:2 subsampling is
+ * used, then the luminance plane would be 36 x 35 bytes, and each of the
+ * chrominance planes would be 18 x 35 bytes.  If you specify a line padding of
+ * 4 bytes on top of this, then the luminance plane would be 36 x 35 bytes, and
+ * each of the chrominance planes would be 20 x 35 bytes.
+ *
+ * @{
+ */
+
+
+/**
+ * The number of chrominance subsampling options
+ */
+#define TJ_NUMSAMP 6
+
+/**
+ * Chrominance subsampling options.
+ * When pixels are converted from RGB to YCbCr (see #TJCS_YCbCr) or from CMYK
+ * to YCCK (see #TJCS_YCCK) as part of the JPEG compression process, some of
+ * the Cb and Cr (chrominance) components can be discarded or averaged together
+ * to produce a smaller image with little perceptible loss of image clarity
+ * (the human eye is more sensitive to small changes in brightness than to
+ * small changes in color.)  This is called "chrominance subsampling".
+ */
+enum TJSAMP
+{
+  /**
+   * 4:4:4 chrominance subsampling (no chrominance subsampling).  The JPEG or
+   * YUV image will contain one chrominance component for every pixel in the
+   * source image.
+   */
+  TJSAMP_444=0,
+  /**
+   * 4:2:2 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x1 block of pixels in the source image.
+   */
+  TJSAMP_422,
+  /**
+   * 4:2:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 2x2 block of pixels in the source image.
+   */
+  TJSAMP_420,
+  /**
+   * Grayscale.  The JPEG or YUV image will contain no chrominance components.
+   */
+  TJSAMP_GRAY,
+  /**
+   * 4:4:0 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 1x2 block of pixels in the source image.
+   *
+   * @note 4:4:0 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_440,
+  /**
+   * 4:1:1 chrominance subsampling.  The JPEG or YUV image will contain one
+   * chrominance component for every 4x1 block of pixels in the source image.
+   * JPEG images compressed with 4:1:1 subsampling will be almost exactly the
+   * same size as those compressed with 4:2:0 subsampling, and in the
+   * aggregate, both subsampling methods produce approximately the same
+   * perceptual quality.  However, 4:1:1 is better able to reproduce sharp
+   * horizontal features.
+   *
+   * @note 4:1:1 subsampling is not fully accelerated in libjpeg-turbo.
+   */
+  TJSAMP_411
+};
+
+/**
+ * MCU block width (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUWidth[TJ_NUMSAMP]  = {8, 16, 16, 8, 8, 32};
+
+/**
+ * MCU block height (in pixels) for a given level of chrominance subsampling.
+ * MCU block sizes:
+ * - 8x8 for no subsampling or grayscale
+ * - 16x8 for 4:2:2
+ * - 8x16 for 4:4:0
+ * - 16x16 for 4:2:0
+ * - 32x8 for 4:1:1
+ */
+static const int tjMCUHeight[TJ_NUMSAMP] = {8, 8, 16, 8, 16, 8};
+
+
+/**
+ * The number of pixel formats
+ */
+#define TJ_NUMPF 12
+
+/**
+ * Pixel formats
+ */
+enum TJPF
+{
+  /**
+   * RGB pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_RGB=0,
+  /**
+   * BGR pixel format.  The red, green, and blue components in the image are
+   * stored in 3-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.
+   */
+  TJPF_BGR,
+  /**
+   * RGBX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_RGBX,
+  /**
+   * BGRX pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from lowest to highest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_BGRX,
+  /**
+   * XBGR pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order R, G, B from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XBGR,
+  /**
+   * XRGB pixel format.  The red, green, and blue components in the image are
+   * stored in 4-byte pixels in the order B, G, R from highest to lowest byte
+   * address within each pixel.  The X component is ignored when compressing
+   * and undefined when decompressing.
+   */
+  TJPF_XRGB,
+  /**
+   * Grayscale pixel format.  Each 1-byte pixel represents a luminance
+   * (brightness) level from 0 to 255.
+   */
+  TJPF_GRAY,
+  /**
+   * RGBA pixel format.  This is the same as @ref TJPF_RGBX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_RGBA,
+  /**
+   * BGRA pixel format.  This is the same as @ref TJPF_BGRX, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_BGRA,
+  /**
+   * ABGR pixel format.  This is the same as @ref TJPF_XBGR, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ABGR,
+  /**
+   * ARGB pixel format.  This is the same as @ref TJPF_XRGB, except that when
+   * decompressing, the X component is guaranteed to be 0xFF, which can be
+   * interpreted as an opaque alpha channel.
+   */
+  TJPF_ARGB,
+  /**
+   * CMYK pixel format.  Unlike RGB, which is an additive color model used
+   * primarily for display, CMYK (Cyan/Magenta/Yellow/Key) is a subtractive
+   * color model used primarily for printing.  In the CMYK color model, the
+   * value of each color component typically corresponds to an amount of cyan,
+   * magenta, yellow, or black ink that is applied to a white background.  In
+   * order to convert between CMYK and RGB, it is necessary to use a color
+   * management system (CMS.)  A CMS will attempt to map colors within the
+   * printer's gamut to perceptually similar colors in the display's gamut and
+   * vice versa, but the mapping is typically not 1:1 or reversible, nor can it
+   * be defined with a simple formula.  Thus, such a conversion is out of scope
+   * for a codec library.  However, the TurboJPEG API allows for compressing
+   * CMYK pixels into a YCCK JPEG image (see #TJCS_YCCK) and decompressing YCCK
+   * JPEG images into CMYK pixels.
+   */
+  TJPF_CMYK
+};
+
+
+/**
+ * Red offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the red component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel[]</tt>,
+ * then the red component will be <tt>pixel[tjRedOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjRedOffset[TJ_NUMPF] = {0, 2, 0, 2, 3, 1, 0, 0, 2, 3, 1, -1};
+/**
+ * Green offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the green component is offset from the start of the pixel.
+ * For instance, if a pixel of format TJ_BGRX is stored in
+ * <tt>char pixel[]</tt>, then the green component will be
+ * <tt>pixel[tjGreenOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjGreenOffset[TJ_NUMPF] = {1, 1, 1, 1, 2, 2, 0, 1, 1, 2, 2, -1};
+/**
+ * Blue offset (in bytes) for a given pixel format.  This specifies the number
+ * of bytes that the Blue component is offset from the start of the pixel.  For
+ * instance, if a pixel of format TJ_BGRX is stored in <tt>char pixel[]</tt>,
+ * then the blue component will be <tt>pixel[tjBlueOffset[TJ_BGRX]]</tt>.
+ */
+static const int tjBlueOffset[TJ_NUMPF] = {2, 0, 2, 0, 1, 3, 0, 2, 0, 1, 3, -1};
+
+/**
+ * Pixel size (in bytes) for a given pixel format.
+ */
+static const int tjPixelSize[TJ_NUMPF] = {3, 3, 4, 4, 4, 4, 1, 4, 4, 4, 4, 4};
+
+
+/**
+ * The number of JPEG colorspaces
+ */
+#define TJ_NUMCS 5
+
+/**
+ * JPEG colorspaces
+ */
+enum TJCS
+{
+  /**
+   * RGB colorspace.  When compressing the JPEG image, the R, G, and B
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  RGB JPEG images can be
+   * decompressed to any of the extended RGB pixel formats or grayscale, but
+   * they cannot be decompressed to YUV images.
+   */
+  TJCS_RGB=0,
+  /**
+   * YCbCr colorspace.  YCbCr is not an absolute colorspace but rather a
+   * mathematical transformation of RGB designed solely for storage and
+   * transmission.  YCbCr images must be converted to RGB before they can
+   * actually be displayed.  In the YCbCr colorspace, the Y (luminance)
+   * component represents the black & white portion of the original image, and
+   * the Cb and Cr (chrominance) components represent the color portion of the
+   * original image.  Originally, the analog equivalent of this transformation
+   * allowed the same signal to drive both black & white and color televisions,
+   * but JPEG images use YCbCr primarily because it allows the color data to be
+   * optionally subsampled for the purposes of reducing bandwidth or disk
+   * space.  YCbCr is the most common JPEG colorspace, and YCbCr JPEG images
+   * can be compressed from and decompressed to any of the extended RGB pixel
+   * formats or grayscale, or they can be decompressed to YUV planar images.
+   */
+  TJCS_YCbCr,
+  /**
+   * Grayscale colorspace.  The JPEG image retains only the luminance data (Y
+   * component), and any color data from the source image is discarded.
+   * Grayscale JPEG images can be compressed from and decompressed to any of
+   * the extended RGB pixel formats or grayscale, or they can be decompressed
+   * to YUV planar images.
+   */
+  TJCS_GRAY,
+  /**
+   * CMYK colorspace.  When compressing the JPEG image, the C, M, Y, and K
+   * components in the source image are reordered into image planes, but no
+   * colorspace conversion or subsampling is performed.  CMYK JPEG images can
+   * only be decompressed to CMYK pixels.
+   */
+  TJCS_CMYK,
+  /**
+   * YCCK colorspace.  YCCK (AKA "YCbCrK") is not an absolute colorspace but
+   * rather a mathematical transformation of CMYK designed solely for storage
+   * and transmission.  It is to CMYK as YCbCr is to RGB.  CMYK pixels can be
+   * reversibly transformed into YCCK, and as with YCbCr, the chrominance
+   * components in the YCCK pixels can be subsampled without incurring major
+   * perceptual loss.  YCCK JPEG images can only be compressed from and
+   * decompressed to CMYK pixels.
+   */
+  TJCS_YCCK
+};
+
+
+/**
+ * The uncompressed source/destination image is stored in bottom-up (Windows,
+ * OpenGL) order, not top-down (X11) order.
+ */
+#define TJFLAG_BOTTOMUP        2
+/**
+ * When decompressing an image that was compressed using chrominance
+ * subsampling, use the fastest chrominance upsampling algorithm available in
+ * the underlying codec.  The default is to use smooth upsampling, which
+ * creates a smooth transition between neighboring chrominance components in
+ * order to reduce upsampling artifacts in the decompressed image.
+ */
+#define TJFLAG_FASTUPSAMPLE  256
+/**
+ * Disable buffer (re)allocation.  If passed to #tjCompress2() or
+ * #tjTransform(), this flag will cause those functions to generate an error if
+ * the JPEG image buffer is invalid or too small rather than attempting to
+ * allocate or reallocate that buffer.  This reproduces the behavior of earlier
+ * versions of TurboJPEG.
+ */
+#define TJFLAG_NOREALLOC     1024
+/**
+ * Use the fastest DCT/IDCT algorithm available in the underlying codec.  The
+ * default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg[-turbo] uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_FASTDCT       2048
+/**
+ * Use the most accurate DCT/IDCT algorithm available in the underlying codec.
+ * The default if this flag is not specified is implementation-specific.  For
+ * example, the implementation of TurboJPEG for libjpeg[-turbo] uses the fast
+ * algorithm by default when compressing, because this has been shown to have
+ * only a very slight effect on accuracy, but it uses the accurate algorithm
+ * when decompressing, because this has been shown to have a larger effect.
+ */
+#define TJFLAG_ACCURATEDCT   4096
+
+
+/**
+ * The number of transform operations
+ */
+#define TJ_NUMXOP 8
+
+/**
+ * Transform operations for #tjTransform()
+ */
+enum TJXOP
+{
+  /**
+   * Do not transform the position of the image pixels
+   */
+  TJXOP_NONE=0,
+  /**
+   * Flip (mirror) image horizontally.  This transform is imperfect if there
+   * are any partial MCU blocks on the right edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_HFLIP,
+  /**
+   * Flip (mirror) image vertically.  This transform is imperfect if there are
+   * any partial MCU blocks on the bottom edge (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_VFLIP,
+  /**
+   * Transpose image (flip/mirror along upper left to lower right axis.)  This
+   * transform is always perfect.
+   */
+  TJXOP_TRANSPOSE,
+  /**
+   * Transverse transpose image (flip/mirror along upper right to lower left
+   * axis.)  This transform is imperfect if there are any partial MCU blocks in
+   * the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_TRANSVERSE,
+  /**
+   * Rotate image clockwise by 90 degrees.  This transform is imperfect if
+   * there are any partial MCU blocks on the bottom edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT90,
+  /**
+   * Rotate image 180 degrees.  This transform is imperfect if there are any
+   * partial MCU blocks in the image (see #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT180,
+  /**
+   * Rotate image counter-clockwise by 90 degrees.  This transform is imperfect
+   * if there are any partial MCU blocks on the right edge (see
+   * #TJXOPT_PERFECT.)
+   */
+  TJXOP_ROT270
+};
+
+
+/**
+ * This option will cause #tjTransform() to return an error if the transform is
+ * not perfect.  Lossless transforms operate on MCU blocks, whose size depends
+ * on the level of chrominance subsampling used (see #tjMCUWidth
+ * and #tjMCUHeight.)  If the image's width or height is not evenly divisible
+ * by the MCU block size, then there will be partial MCU blocks on the right
+ * and/or bottom edges.  It is not possible to move these partial MCU blocks to
+ * the top or left of the image, so any transform that would require that is
+ * "imperfect."  If this option is not specified, then any partial MCU blocks
+ * that cannot be transformed will be left in place, which will create
+ * odd-looking strips on the right or bottom edge of the image.
+ */
+#define TJXOPT_PERFECT  1
+/**
+ * This option will cause #tjTransform() to discard any partial MCU blocks that
+ * cannot be transformed.
+ */
+#define TJXOPT_TRIM     2
+/**
+ * This option will enable lossless cropping.  See #tjTransform() for more
+ * information.
+ */
+#define TJXOPT_CROP     4
+/**
+ * This option will discard the color data in the input image and produce
+ * a grayscale output image.
+ */
+#define TJXOPT_GRAY     8
+/**
+ * This option will prevent #tjTransform() from outputting a JPEG image for
+ * this particular transform (this can be used in conjunction with a custom
+ * filter to capture the transformed DCT coefficients without transcoding
+ * them.)
+ */
+#define TJXOPT_NOOUTPUT 16
+
+
+/**
+ * Scaling factor
+ */
+typedef struct
+{
+  /**
+   * Numerator
+   */
+  int num;
+  /**
+   * Denominator
+   */
+  int denom;
+} tjscalingfactor;
+
+/**
+ * Cropping region
+ */
+typedef struct
+{
+  /**
+   * The left boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block width (see #tjMCUWidth.)
+   */
+  int x;
+  /**
+   * The upper boundary of the cropping region.  This must be evenly divisible
+   * by the MCU block height (see #tjMCUHeight.)
+   */
+  int y;
+  /**
+   * The width of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the width of the source JPEG image - x.
+   */
+  int w;
+  /**
+   * The height of the cropping region. Setting this to 0 is the equivalent of
+   * setting it to the height of the source JPEG image - y.
+   */
+  int h;
+} tjregion;
+
+/**
+ * Lossless transform
+ */
+typedef struct tjtransform
+{
+  /**
+   * Cropping region
+   */
+  tjregion r;
+  /**
+   * One of the @ref TJXOP "transform operations"
+   */
+  int op;
+  /**
+   * The bitwise OR of one of more of the @ref TJXOPT_CROP "transform options"
+   */
+  int options;
+  /**
+   * Arbitrary data that can be accessed within the body of the callback
+   * function
+   */
+  void *data;
+  /**
+   * A callback function that can be used to modify the DCT coefficients
+   * after they are losslessly transformed but before they are transcoded to a
+   * new JPEG image.  This allows for custom filters or other transformations
+   * to be applied in the frequency domain.
+   *
+   * @param coeffs pointer to an array of transformed DCT coefficients.  (NOTE:
+   * this pointer is not guaranteed to be valid once the callback returns, so
+   * applications wishing to hand off the DCT coefficients to another function
+   * or library should make a copy of them within the body of the callback.)
+   *
+   * @param arrayRegion #tjregion structure containing the width and height of
+   * the array pointed to by <tt>coeffs</tt> as well as its offset relative to
+   * the component plane.  TurboJPEG implementations may choose to split each
+   * component plane into multiple DCT coefficient arrays and call the callback
+   * function once for each array.
+   *
+   * @param planeRegion #tjregion structure containing the width and height of
+   * the component plane to which <tt>coeffs</tt> belongs
+   *
+   * @param componentID ID number of the component plane to which
+   * <tt>coeffs</tt> belongs (Y, Cb, and Cr have, respectively, ID's of 0, 1,
+   * and 2 in typical JPEG images.)
+   *
+   * @param transformID ID number of the transformed image to which
+   * <tt>coeffs</tt> belongs.  This is the same as the index of the transform
+   * in the <tt>transforms</tt> array that was passed to #tjTransform().
+   *
+   * @param transform a pointer to a #tjtransform structure that specifies the
+   * parameters and/or cropping region for this transform
+   *
+   * @return 0 if the callback was successful, or -1 if an error occurred.
+   */
+  int (*customFilter)(short *coeffs, tjregion arrayRegion,
+    tjregion planeRegion, int componentIndex, int transformIndex,
+    struct tjtransform *transform);
+} tjtransform;
+
+/**
+ * TurboJPEG instance handle
+ */
+typedef void* tjhandle;
+
+
+/**
+ * Pad the given width to the nearest 32-bit boundary
+ */
+#define TJPAD(width) (((width)+3)&(~3))
+
+/**
+ * Compute the scaled value of <tt>dimension</tt> using the given scaling
+ * factor.  This macro performs the integer equivalent of <tt>ceil(dimension *
+ * scalingFactor)</tt>.
+ */
+#define TJSCALED(dimension, scalingFactor) ((dimension * scalingFactor.num \
+  + scalingFactor.denom - 1) / scalingFactor.denom)
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+/**
+ * Create a TurboJPEG compressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT tjhandle DLLCALL tjInitCompress(void);
+
+
+/**
+ * Compress an RGB, grayscale, or CMYK image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB, grayscale, or
+ * CMYK pixels to be compressed
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer
+ * to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompress2(tjhandle handle, const unsigned char *srcBuf,
+  int width, int pitch, int height, int pixelFormat, unsigned char **jpegBuf,
+  unsigned long *jpegSize, int jpegSubsamp, int jpegQual, int flags);
+
+
+/**
+ * Compress a YUV planar image into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * compressed.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param pad the line padding used in the source image.  For instance, if each
+ * line in each plane of the YUV image is padded to the nearest multiple of 4
+ * bytes, then <tt>pad</tt> should be set to 4.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompressFromYUV(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pad, int height, int subsamp,
+  unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual, int flags);
+
+
+/**
+ * Compress a set of Y, U (Cb), and V (Cr) image planes into a JPEG image.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if compressing a grayscale image) that contain a YUV
+ * image to be compressed.  These planes can be contiguous or non-contiguous in
+ * memory.  The size of each plane should match the value returned by
+ * #tjPlaneSizeYUV() for the given image width, height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width width (in pixels) of the source image.  If the width is not an
+ * even multiple of the MCU block width (see #tjMCUWidth), then an intermediate
+ * buffer copy will be performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to create a JPEG image from a subregion of a larger
+ * YUV planar image.
+ *
+ * @param height height (in pixels) of the source image.  If the height is not
+ * an even multiple of the MCU block height (see #tjMCUHeight), then an
+ * intermediate buffer copy will be performed within TurboJPEG.
+ *
+ * @param subsamp the level of chrominance subsampling used in the source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegBuf address of a pointer to an image buffer that will receive the
+ * JPEG image.  TurboJPEG has the ability to reallocate the JPEG buffer to
+ * accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>*jpegBuf</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize().  This should ensure that the buffer never has to be
+ * re-allocated (setting #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>*jpegSize</tt> should be set to the size of your
+ * pre-allocated buffer.  In any case, unless you have set #TJFLAG_NOREALLOC,
+ * you should always check <tt>*jpegBuf</tt> upon return from this function, as
+ * it may have changed.
+ *
+ * @param jpegSize pointer to an unsigned long variable that holds the size of
+ * the JPEG image buffer.  If <tt>*jpegBuf</tt> points to a pre-allocated
+ * buffer, then <tt>*jpegSize</tt> should be set to the size of the buffer.
+ * Upon return, <tt>*jpegSize</tt> will contain the size of the JPEG image (in
+ * bytes.)  If <tt>*jpegBuf</tt> points to a JPEG image buffer that is being
+ * reused from a previous call to one of the JPEG compression functions, then
+ * <tt>*jpegSize</tt> is ignored.
+ *
+ * @param jpegQual the image quality of the generated JPEG image (1 = worst,
+ * 100 = best)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjCompressFromYUVPlanes(tjhandle handle,
+  const unsigned char **srcPlanes, int width, const int *strides, int height,
+  int subsamp, unsigned char **jpegBuf, unsigned long *jpegSize, int jpegQual,
+  int flags);
+
+
+/**
+ * The maximum size of the buffer (in bytes) required to hold a JPEG image with
+ * the given parameters.  The number of bytes returned by this function is
+ * larger than the size of the uncompressed source image.  The reason for this
+ * is that the JPEG format uses 16-bit coefficients, and it is thus possible
+ * for a very high-quality JPEG image with very high-frequency content to
+ * expand rather than compress when converted to the JPEG format.  Such images
+ * represent a very rare corner case, but since there is no way to predict the
+ * size of a JPEG image prior to compression, the corner case has to be
+ * handled.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param jpegSubsamp the level of chrominance subsampling to be used when
+ * generating the JPEG image (see @ref TJSAMP
+ * "Chrominance subsampling options".)
+ *
+ * @return the maximum size of the buffer (in bytes) required to hold the
+ * image, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjBufSize(int width, int height,
+  int jpegSubsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV planar image with
+ * the given parameters.
+ *
+ * @param width width (in pixels) of the image
+ *
+ * @param pad the width of each line in each plane of the image is padded to
+ * the nearest multiple of this number of bytes (must be a power of 2.)
+ *
+ * @param height height (in pixels) of the image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the image, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV2(int width, int pad, int height,
+  int subsamp);
+
+
+/**
+ * The size of the buffer (in bytes) required to hold a YUV image plane with
+ * the given parameters.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image.  NOTE: this is the width of
+ * the whole image, not the plane width.
+ *
+ * @param stride bytes per line in the image plane.  Setting this to 0 is the
+ * equivalent of setting it to the plane width.
+ *
+ * @param height height (in pixels) of the YUV image.  NOTE: this is the height
+ * of the whole image, not the plane height.
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the size of the buffer (in bytes) required to hold the YUV image
+ * plane, or -1 if the arguments are out of bounds.
+ */
+DLLEXPORT unsigned long DLLCALL tjPlaneSizeYUV(int componentID, int width,
+  int stride, int height, int subsamp);
+
+
+/**
+ * The plane width of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane width.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param width width (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane width of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneWidth(int componentID, int width, int subsamp);
+
+
+/**
+ * The plane height of a YUV image plane with the given parameters.  Refer to
+ * @ref YUVnotes "YUV Image Format Notes" for a description of plane height.
+ *
+ * @param componentID ID number of the image plane (0 = Y, 1 = U/Cb, 2 = V/Cr)
+ *
+ * @param height height (in pixels) of the YUV image
+ *
+ * @param subsamp level of chrominance subsampling in the image (see
+ * @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @return the plane height of a YUV image plane with the given parameters, or
+ * -1 if the arguments are out of bounds.
+ */
+DLLEXPORT int tjPlaneHeight(int componentID, int height, int subsamp);
+
+
+/**
+ * Encode an RGB or grayscale image into a YUV planar image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG compression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of chrominance subsampling.
+ * The Y, U (Cb), and V (Cr) image planes will be stored sequentially in the
+ * buffer (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjEncodeYUV3(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pitch, int height,
+  int pixelFormat, unsigned char *dstBuf, int pad, int subsamp, int flags);
+
+
+/**
+ * Encode an RGB or grayscale image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * compression process.
+ *
+ * @param handle a handle to a TurboJPEG compressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing RGB or grayscale pixels
+ * to be encoded
+ *
+ * @param width width (in pixels) of the source image
+ *
+ * @param pitch bytes per line in the source image.  Normally, this should be
+ * <tt>width * #tjPixelSize[pixelFormat]</tt> if the image is unpadded, or
+ * <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line of the image
+ * is padded to the nearest 32-bit boundary, as is the case for Windows
+ * bitmaps.  You can also be clever and use this parameter to skip lines, etc.
+ * Setting this parameter to 0 is the equivalent of setting it to
+ * <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source image
+ *
+ * @param pixelFormat pixel format of the source image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if generating a grayscale image) that will receive the
+ * encoded image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the image width, height, strides, and level of chrominance subsampling.
+ * Refer to @ref YUVnotes "YUV Image Format Notes" for more details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to add an arbitrary amount of line
+ * padding to each plane or to encode an RGB or grayscale image into a
+ * subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling to be used when
+ * generating the YUV image (see @ref TJSAMP
+ * "Chrominance subsampling options".)  To generate images suitable for X
+ * Video, <tt>subsamp</tt> should be set to @ref TJSAMP_420.  This produces an
+ * image compatible with the I420 (AKA "YUV420P") format.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjEncodeYUVPlanes(tjhandle handle,
+  const unsigned char *srcBuf, int width, int pitch, int height,
+  int pixelFormat, unsigned char **dstPlanes, int *strides, int subsamp,
+  int flags);
+
+
+/**
+ * Create a TurboJPEG decompressor instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT tjhandle DLLCALL tjInitDecompress(void);
+
+
+/**
+ * Retrieve information about a JPEG image without decompressing it.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing a JPEG image
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param width pointer to an integer variable that will receive the width (in
+ * pixels) of the JPEG image
+ *
+ * @param height pointer to an integer variable that will receive the height
+ * (in pixels) of the JPEG image
+ *
+ * @param jpegSubsamp pointer to an integer variable that will receive the
+ * level of chrominance subsampling used when the JPEG image was compressed
+ * (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param jpegColorspace pointer to an integer variable that will receive one
+ * of the JPEG colorspace constants, indicating the colorspace of the JPEG
+ * image (see @ref TJCS "JPEG colorspaces".)
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+*/
+DLLEXPORT int DLLCALL tjDecompressHeader3(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, int *width,
+  int *height, int *jpegSubsamp, int *jpegColorspace);
+
+
+/**
+ * Returns a list of fractional scaling factors that the JPEG decompressor in
+ * this implementation of TurboJPEG supports.
+ *
+ * @param numscalingfactors pointer to an integer variable that will receive
+ * the number of elements in the list
+ *
+ * @return a pointer to a list of fractional scaling factors, or NULL if an
+ * error is encountered (see #tjGetErrorStr().)
+*/
+DLLEXPORT tjscalingfactor* DLLCALL tjGetScalingFactors(int *numscalingfactors);
+
+
+/**
+ * Decompress a JPEG image to an RGB, grayscale, or CMYK image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decompressed
+ * image.  This buffer should normally be <tt>pitch * scaledHeight</tt> bytes
+ * in size, where <tt>scaledHeight</tt> can be determined by calling
+ * #TJSCALED() with the JPEG image height and one of the scaling factors
+ * returned by #tjGetScalingFactors().  The <tt>dstBuf</tt> pointer may also be
+ * used to decompress into a specific region of a larger buffer.
+ *
+ * @param width desired width (in pixels) of the destination image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this is
+ * <tt>scaledWidth * #tjPixelSize[pixelFormat]</tt> if the decompressed image
+ * is unpadded, else <tt>#TJPAD(scaledWidth * #tjPixelSize[pixelFormat])</tt>
+ * if each line of the decompressed image is padded to the nearest 32-bit
+ * boundary, as is the case for Windows bitmaps.  (NOTE: <tt>scaledWidth</tt>
+ * can be determined by calling #TJSCALED() with the JPEG image width and one
+ * of the scaling factors returned by #tjGetScalingFactors().)  You can also be
+ * clever and use the pitch parameter to skip lines, etc.  Setting this
+ * parameter to 0 is the equivalent of setting it to
+ * <tt>scaledWidth * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height desired height (in pixels) of the destination image.  If this
+ * is different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref
+ * TJPF "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompress2(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pitch, int height, int pixelFormat, int flags);
+
+
+/**
+ * Decompress a JPEG image to a YUV planar image.  This function performs JPEG
+ * decompression but leaves out the color conversion step, so a planar YUV
+ * image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the YUV image.
+ * Use #tjBufSizeYUV2() to determine the appropriate size for this buffer based
+ * on the image width, height, padding, and level of subsampling.  The Y,
+ * U (Cb), and V (Cr) image planes will be stored sequentially in the buffer
+ * (refer to @ref YUVnotes "YUV Image Format Notes".)
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param pad the width of each line in each plane of the YUV image will be
+ * padded to the nearest multiple of this number of bytes (must be a power of
+ * 2.)  To generate images suitable for X Video, <tt>pad</tt> should be set to
+ * 4.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompressToYUV2(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pad, int height, int flags);
+
+
+/**
+ * Decompress a JPEG image into separate Y, U (Cb), and V (Cr) image
+ * planes.  This function performs JPEG decompression but leaves out the color
+ * conversion step, so a planar YUV image is generated instead of an RGB image.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG image to decompress
+ *
+ * @param jpegSize size of the JPEG image (in bytes)
+ *
+ * @param dstPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decompressing a grayscale image) that will receive
+ * the YUV image.  These planes can be contiguous or non-contiguous in memory.
+ * Use #tjPlaneSizeYUV() to determine the appropriate size for each plane based
+ * on the scaled image width, scaled image height, strides, and level of
+ * chrominance subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes"
+ * for more details.
+ *
+ * @param width desired width (in pixels) of the YUV image.  If this is
+ * different than the width of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired width.  If <tt>width</tt> is
+ * set to 0, then only the height will be considered when determining the
+ * scaled image size.  If the scaled width is not an even multiple of the MCU
+ * block width (see #tjMCUWidth), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the output image.  Setting the stride for
+ * any plane to 0 is the same as setting it to the scaled plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective scaled plane
+ * widths.  You can adjust the strides in order to add an arbitrary amount of
+ * line padding to each plane or to decompress the JPEG image into a subregion
+ * of a larger YUV planar image.
+ *
+ * @param height desired height (in pixels) of the YUV image.  If this is
+ * different than the height of the JPEG image being decompressed, then
+ * TurboJPEG will use scaling in the JPEG decompressor to generate the largest
+ * possible image that will fit within the desired height.  If <tt>height</tt>
+ * is set to 0, then only the width will be considered when determining the
+ * scaled image size.  If the scaled height is not an even multiple of the MCU
+ * block height (see #tjMCUHeight), then an intermediate buffer copy will be
+ * performed within TurboJPEG.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecompressToYUVPlanes(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize,
+  unsigned char **dstPlanes, int width, int *strides, int height, int flags);
+
+
+/**
+ * Decode a YUV planar image into an RGB or grayscale image.  This function
+ * uses the accelerated color conversion routines in the underlying
+ * codec but does not execute any of the other steps in the JPEG decompression
+ * process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcBuf pointer to an image buffer containing a YUV planar image to be
+ * decoded.  The size of this buffer should match the value returned by
+ * #tjBufSizeYUV2() for the given image width, height, padding, and level of
+ * chrominance subsampling.  The Y, U (Cb), and V (Cr) image planes should be
+ * stored sequentially in the source buffer (refer to @ref YUVnotes
+ * "YUV Image Format Notes".)
+ *
+ * @param pad Use this parameter to specify that the width of each line in each
+ * plane of the YUV source image is padded to the nearest multiple of this
+ * number of bytes (must be a power of 2.)
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSize[pixelFormat]</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecodeYUV(tjhandle handle, const unsigned char *srcBuf,
+  int pad, int subsamp, unsigned char *dstBuf, int width, int pitch,
+  int height, int pixelFormat, int flags);
+
+
+/**
+ * Decode a set of Y, U (Cb), and V (Cr) image planes into an RGB or grayscale
+ * image.  This function uses the accelerated color conversion routines in the
+ * underlying codec but does not execute any of the other steps in the JPEG
+ * decompression process.
+ *
+ * @param handle a handle to a TurboJPEG decompressor or transformer instance
+ *
+ * @param srcPlanes an array of pointers to Y, U (Cb), and V (Cr) image planes
+ * (or just a Y plane, if decoding a grayscale image) that contain a YUV image
+ * to be decoded.  These planes can be contiguous or non-contiguous in memory.
+ * The size of each plane should match the value returned by #tjPlaneSizeYUV()
+ * for the given image width, height, strides, and level of chrominance
+ * subsampling.  Refer to @ref YUVnotes "YUV Image Format Notes" for more
+ * details.
+ *
+ * @param strides an array of integers, each specifying the number of bytes per
+ * line in the corresponding plane of the YUV source image.  Setting the stride
+ * for any plane to 0 is the same as setting it to the plane width (see
+ * @ref YUVnotes "YUV Image Format Notes".)  If <tt>strides</tt> is NULL, then
+ * the strides for all planes will be set to their respective plane widths.
+ * You can adjust the strides in order to specify an arbitrary amount of line
+ * padding in each plane or to decode a subregion of a larger YUV planar image.
+ *
+ * @param subsamp the level of chrominance subsampling used in the YUV source
+ * image (see @ref TJSAMP "Chrominance subsampling options".)
+ *
+ * @param dstBuf pointer to an image buffer that will receive the decoded
+ * image.  This buffer should normally be <tt>pitch * height</tt> bytes in
+ * size, but the <tt>dstBuf</tt> pointer can also be used to decode into a
+ * specific region of a larger buffer.
+ *
+ * @param width width (in pixels) of the source and destination images
+ *
+ * @param pitch bytes per line in the destination image.  Normally, this should
+ * be <tt>width * #tjPixelSize[pixelFormat]</tt> if the destination image is
+ * unpadded, or <tt>#TJPAD(width * #tjPixelSize[pixelFormat])</tt> if each line
+ * of the destination image should be padded to the nearest 32-bit boundary, as
+ * is the case for Windows bitmaps.  You can also be clever and use the pitch
+ * parameter to skip lines, etc.  Setting this parameter to 0 is the equivalent
+ * of setting it to <tt>width * #tjPixelSize[pixelFormat]</tt>.
+ *
+ * @param height height (in pixels) of the source and destination images
+ *
+ * @param pixelFormat pixel format of the destination image (see @ref TJPF
+ * "Pixel formats".)
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDecodeYUVPlanes(tjhandle handle,
+  const unsigned char **srcPlanes, const int *strides, int subsamp,
+  unsigned char *dstBuf, int width, int pitch, int height, int pixelFormat,
+  int flags);
+
+
+/**
+ * Create a new TurboJPEG transformer instance.
+ *
+ * @return a handle to the newly-created instance, or NULL if an error
+ * occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT tjhandle DLLCALL tjInitTransform(void);
+
+
+/**
+ * Losslessly transform a JPEG image into another JPEG image.  Lossless
+ * transforms work by moving the raw DCT coefficients from one JPEG image
+ * structure to another without altering the values of the coefficients.  While
+ * this is typically faster than decompressing the image, transforming it, and
+ * re-compressing it, lossless transforms are not free.  Each lossless
+ * transform requires reading and performing Huffman decoding on all of the
+ * coefficients in the source image, regardless of the size of the destination
+ * image.  Thus, this function provides a means of generating multiple
+ * transformed images from the same source or  applying multiple
+ * transformations simultaneously, in order to eliminate the need to read the
+ * source coefficients multiple times.
+ *
+ * @param handle a handle to a TurboJPEG transformer instance
+ *
+ * @param jpegBuf pointer to a buffer containing the JPEG source image to
+ * transform
+ *
+ * @param jpegSize size of the JPEG source image (in bytes)
+ *
+ * @param n the number of transformed JPEG images to generate
+ *
+ * @param dstBufs pointer to an array of n image buffers.  <tt>dstBufs[i]</tt>
+ * will receive a JPEG image that has been transformed using the parameters in
+ * <tt>transforms[i]</tt>.  TurboJPEG has the ability to reallocate the JPEG
+ * buffer to accommodate the size of the JPEG image.  Thus, you can choose to:
+ * -# pre-allocate the JPEG buffer with an arbitrary size using #tjAlloc() and
+ * let TurboJPEG grow the buffer as needed,
+ * -# set <tt>dstBufs[i]</tt> to NULL to tell TurboJPEG to allocate the buffer
+ * for you, or
+ * -# pre-allocate the buffer to a "worst case" size determined by calling
+ * #tjBufSize() with the transformed or cropped width and height.  This should
+ * ensure that the buffer never has to be re-allocated (setting
+ * #TJFLAG_NOREALLOC guarantees this.)
+ * .
+ * If you choose option 1, <tt>dstSizes[i]</tt> should be set to the size of
+ * your pre-allocated buffer.  In any case, unless you have set
+ * #TJFLAG_NOREALLOC, you should always check <tt>dstBufs[i]</tt> upon return
+ * from this function, as it may have changed.
+ *
+ * @param dstSizes pointer to an array of n unsigned long variables that will
+ * receive the actual sizes (in bytes) of each transformed JPEG image.  If
+ * <tt>dstBufs[i]</tt> points to a pre-allocated buffer, then
+ * <tt>dstSizes[i]</tt> should be set to the size of the buffer.  Upon return,
+ * <tt>dstSizes[i]</tt> will contain the size of the JPEG image (in bytes.)
+ *
+ * @param transforms pointer to an array of n #tjtransform structures, each of
+ * which specifies the transform parameters and/or cropping region for the
+ * corresponding transformed output image.
+ *
+ * @param flags the bitwise OR of one or more of the @ref TJFLAG_BOTTOMUP
+ * "flags"
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjTransform(tjhandle handle,
+  const unsigned char *jpegBuf, unsigned long jpegSize, int n,
+  unsigned char **dstBufs, unsigned long *dstSizes, tjtransform *transforms,
+  int flags);
+
+
+/**
+ * Destroy a TurboJPEG compressor, decompressor, or transformer instance.
+ *
+ * @param handle a handle to a TurboJPEG compressor, decompressor or
+ * transformer instance
+ *
+ * @return 0 if successful, or -1 if an error occurred (see #tjGetErrorStr().)
+ */
+DLLEXPORT int DLLCALL tjDestroy(tjhandle handle);
+
+
+/**
+ * Allocate an image buffer for use with TurboJPEG.  You should always use
+ * this function to allocate the JPEG destination buffer(s) for #tjCompress2()
+ * and #tjTransform() unless you are disabling automatic buffer
+ * (re)allocation (by setting #TJFLAG_NOREALLOC.)
+ *
+ * @param bytes the number of bytes to allocate
+ *
+ * @return a pointer to a newly-allocated buffer with the specified number of
+ * bytes.
+ *
+ * @sa tjFree()
+ */
+DLLEXPORT unsigned char* DLLCALL tjAlloc(int bytes);
+
+
+/**
+ * Free an image buffer previously allocated by TurboJPEG.  You should always
+ * use this function to free JPEG destination buffer(s) that were automatically
+ * (re)allocated by #tjCompress2() or #tjTransform() or that were manually
+ * allocated using #tjAlloc().
+ *
+ * @param buffer address of the buffer to free
+ *
+ * @sa tjAlloc()
+ */
+DLLEXPORT void DLLCALL tjFree(unsigned char *buffer);
+
+
+/**
+ * Returns a descriptive error message explaining why the last command failed.
+ *
+ * @return a descriptive error message explaining why the last command failed.
+ */
+DLLEXPORT char* DLLCALL tjGetErrorStr(void);
+
+
+/* Deprecated functions and macros */
+#define TJFLAG_FORCEMMX        8
+#define TJFLAG_FORCESSE       16
+#define TJFLAG_FORCESSE2      32
+#define TJFLAG_FORCESSE3     128
+
+
+/* Backward compatibility functions and macros (nothing to see here) */
+#define NUMSUBOPT TJ_NUMSAMP
+#define TJ_444 TJSAMP_444
+#define TJ_422 TJSAMP_422
+#define TJ_420 TJSAMP_420
+#define TJ_411 TJSAMP_420
+#define TJ_GRAYSCALE TJSAMP_GRAY
+
+#define TJ_BGR 1
+#define TJ_BOTTOMUP TJFLAG_BOTTOMUP
+#define TJ_FORCEMMX TJFLAG_FORCEMMX
+#define TJ_FORCESSE TJFLAG_FORCESSE
+#define TJ_FORCESSE2 TJFLAG_FORCESSE2
+#define TJ_ALPHAFIRST 64
+#define TJ_FORCESSE3 TJFLAG_FORCESSE3
+#define TJ_FASTUPSAMPLE TJFLAG_FASTUPSAMPLE
+#define TJ_YUV 512
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZE(int width, int height);
+
+DLLEXPORT unsigned long DLLCALL TJBUFSIZEYUV(int width, int height,
+  int jpegSubsamp);
+
+DLLEXPORT unsigned long DLLCALL tjBufSizeYUV(int width, int height,
+  int subsamp);
+
+DLLEXPORT int DLLCALL tjCompress(tjhandle handle, unsigned char *srcBuf,
+  int width, int pitch, int height, int pixelSize, unsigned char *dstBuf,
+  unsigned long *compressedSize, int jpegSubsamp, int jpegQual, int flags);
+
+DLLEXPORT int DLLCALL tjEncodeYUV(tjhandle handle,
+  unsigned char *srcBuf, int width, int pitch, int height, int pixelSize,
+  unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int DLLCALL tjEncodeYUV2(tjhandle handle,
+  unsigned char *srcBuf, int width, int pitch, int height, int pixelFormat,
+  unsigned char *dstBuf, int subsamp, int flags);
+
+DLLEXPORT int DLLCALL tjDecompressHeader(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height);
+
+DLLEXPORT int DLLCALL tjDecompressHeader2(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, int *width, int *height,
+  int *jpegSubsamp);
+
+DLLEXPORT int DLLCALL tjDecompress(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int width, int pitch, int height, int pixelSize, int flags);
+
+DLLEXPORT int DLLCALL tjDecompressToYUV(tjhandle handle,
+  unsigned char *jpegBuf, unsigned long jpegSize, unsigned char *dstBuf,
+  int flags);
+
+
+/**
+ * @}
+ */
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-jni.c glmark2-2021.02/src/libjpeg-turbo/turbojpeg-jni.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-jni.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/turbojpeg-jni.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1166 @@
+/*
+ * Copyright (C)2011-2016 D. R. Commander.  All Rights Reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * - Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ * - Neither the name of the libjpeg-turbo Project nor the names of its
+ *   contributors may be used to endorse or promote products derived from this
+ *   software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS",
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDERS OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdlib.h>
+#include <string.h>
+#include "turbojpeg.h"
+#ifdef WIN32
+#include "tjutil.h"
+#endif
+#include <jni.h>
+#include "java/org_libjpegturbo_turbojpeg_TJCompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJDecompressor.h"
+#include "java/org_libjpegturbo_turbojpeg_TJ.h"
+
+#define PAD(v, p) ((v+(p)-1)&(~((p)-1)))
+
+#define _throw(msg, exceptionClass) {  \
+	jclass _exccls=(*env)->FindClass(env, exceptionClass);  \
+	if(!_exccls || (*env)->ExceptionCheck(env)) goto bailout;  \
+	(*env)->ThrowNew(env, _exccls, msg);  \
+	goto bailout;  \
+}
+
+#define _throwtj() _throw(tjGetErrorStr(), "org/libjpegturbo/turbojpeg/TJException")
+
+#define _throwarg(msg) _throw(msg, "java/lang/IllegalArgumentException")
+
+#define _throwmem() _throw("Memory allocation failure", "java/lang/OutOfMemoryError");
+
+#define bailif0(f) {if(!(f) || (*env)->ExceptionCheck(env)) {  \
+	goto bailout;  \
+}}
+
+#define gethandle()  \
+	jclass _cls=(*env)->GetObjectClass(env, obj);  \
+	jfieldID _fid;  \
+	if(!_cls || (*env)->ExceptionCheck(env)) goto bailout;  \
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "handle", "J"));  \
+	handle=(tjhandle)(size_t)(*env)->GetLongField(env, obj, _fid);  \
+
+#ifdef _WIN32
+#define setenv(envvar, value, dummy) _putenv_s(envvar, value)
+#endif
+
+#define prop2env(property, envvar)  \
+{  \
+	if((jName=(*env)->NewStringUTF(env, property))!=NULL  \
+		&& (jValue=(*env)->CallStaticObjectMethod(env, cls, mid, jName))!=NULL)  \
+	{  \
+		if((value=(*env)->GetStringUTFChars(env, jValue, 0))!=NULL)  \
+		{  \
+			setenv(envvar, value, 1);  \
+			(*env)->ReleaseStringUTFChars(env, jValue, value);  \
+		}  \
+	}  \
+}
+
+int ProcessSystemProperties(JNIEnv *env)
+{
+	jclass cls;  jmethodID mid;
+	jstring jName, jValue;
+	const char *value;
+
+	bailif0(cls=(*env)->FindClass(env, "java/lang/System"));
+	bailif0(mid=(*env)->GetStaticMethodID(env, cls, "getProperty",
+		"(Ljava/lang/String;)Ljava/lang/String;"));
+
+	prop2env("turbojpeg.optimize", "TJ_OPTIMIZE");
+	prop2env("turbojpeg.arithmetic", "TJ_ARITHMETIC");
+	prop2env("turbojpeg.restart", "TJ_RESTART");
+	prop2env("turbojpeg.progressive", "TJ_PROGRESSIVE");
+	return 0;
+
+	bailout:
+	return -1;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSize() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSize
+	(JNIEnv *env, jclass cls, jint width, jint height, jint jpegSubsamp)
+{
+	jint retval=(jint)tjBufSize(width, height, jpegSubsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII
+	(JNIEnv *env, jclass cls, jint width, jint pad, jint height, jint subsamp)
+{
+	jint retval=(jint)tjBufSizeYUV2(width, pad, height, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.2.x: TJ::bufSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III
+	(JNIEnv *env, jclass cls, jint width, jint height, jint subsamp)
+{
+	return Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII(env, cls, width,
+		4, height, subsamp);
+}
+
+/* TurboJPEG 1.4.x: TJ::planeSizeYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII
+	(JNIEnv *env, jclass cls, jint componentID, jint width, jint stride,
+		jint height, jint subsamp)
+{
+	jint retval=(jint)tjPlaneSizeYUV(componentID, width, stride, height,
+		subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeWidth() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III
+	(JNIEnv *env, jclass cls, jint componentID, jint width, jint subsamp)
+{
+	jint retval=(jint)tjPlaneWidth(componentID, width, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.4.x: TJ::planeHeight() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III
+	(JNIEnv *env, jclass cls, jint componentID, jint height, jint subsamp)
+{
+	jint retval=(jint)tjPlaneHeight(componentID, height, subsamp);
+	if(retval==-1) _throwarg(tjGetErrorStr());
+
+	bailout:
+	return retval;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitCompress())==NULL)
+		_throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+static jint TJCompressor_compress
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jbyteArray dst,
+		jint jpegSubsamp, jint jpegQual, jint flags)
+{
+	tjhandle handle=0;
+	unsigned long jpegSize=0;
+	jsize arraySize=0, actualPitch;
+	unsigned char *srcBuf=NULL, *jpegBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0)
+		_throwarg("Invalid argument in compress()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+	jpegSize=tjBufSize(width, height, jpegSubsamp);
+	if((*env)->GetArrayLength(env, dst)<(jsize)jpegSize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(ProcessSystemProperties(env)<0) goto bailout;
+
+	if(tjCompress2(handle, &srcBuf[y*actualPitch + x*tjPixelSize[pf]], width,
+		pitch, height, pf, &jpegBuf, &jpegSize, jpegSubsamp, jpegQual,
+		flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, jpegBuf, 0);
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	return (jint)jpegSize;
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+		jint pitch, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+		jint jpegQual, jint flags)
+{
+	return TJCompressor_compress(env, obj, src, 1, x, y, width, pitch, height,
+		pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() byte source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+		jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+		jint flags)
+{
+	return TJCompressor_compress(env, obj, src, 1, 0, 0, width, pitch, height,
+		pf, dst, jpegSubsamp, jpegQual, flags);
+}
+
+/* TurboJPEG 1.3.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII
+	(JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+		jint stride, jint height, jint pf, jbyteArray dst, jint jpegSubsamp,
+		jint jpegQual, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in compress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+	return TJCompressor_compress(env, obj, src, sizeof(jint), x, y, width,
+		stride*sizeof(jint), height, pf, dst, jpegSubsamp, jpegQual, flags);
+
+	bailout:
+	return 0;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::compress() int source */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII
+	(JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+		jint height, jint pf, jbyteArray dst, jint jpegSubsamp, jint jpegQual,
+		jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in compress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when compressing from an integer buffer.");
+
+	return TJCompressor_compress(env, obj, src, sizeof(jint), 0, 0, width,
+		stride*sizeof(jint), height, pf, dst, jpegSubsamp, jpegQual, flags);
+
+	bailout:
+	return 0;
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::compressFromYUV() */
+JNIEXPORT jint JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jint width, jintArray jSrcStrides, jint height, jint subsamp,
+		jbyteArray dst, jint jpegQual, jint flags)
+{
+	tjhandle handle=0;
+	unsigned long jpegSize=0;
+	jbyteArray jSrcPlanes[3]={NULL, NULL, NULL};
+	const unsigned char *srcPlanes[3];
+	unsigned char *jpegBuf=NULL;
+	int *srcOffsets=NULL, *srcStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(subsamp<0 || subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in compressFromYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, srcobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	jpegSize=tjBufSize(width, height, subsamp);
+	if((*env)->GetArrayLength(env, dst)<(jsize)jpegSize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcOffsets=(*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+	bailif0(srcStrides=(*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, srcStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(srcOffsets[i]<0)
+			_throwarg("Invalid argument in compressFromYUV()");
+		if(srcStrides[i]<0 && srcOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jSrcPlanes[i]=(*env)->GetObjectArrayElement(env, srcobjs, i));
+		if((*env)->GetArrayLength(env, jSrcPlanes[i])<srcOffsets[i]+planeSize)
+			_throwarg("Source plane is not large enough");
+
+		bailif0(srcPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jSrcPlanes[i],
+			0));
+		srcPlanes[i]=&srcPlanes[i][srcOffsets[i]];
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(ProcessSystemProperties(env)<0) goto bailout;
+
+	if(tjCompressFromYUVPlanes(handle, srcPlanes, width, srcStrides, height,
+		subsamp, &jpegBuf, &jpegSize, jpegQual, flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, jpegBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(srcPlanes[i] && jSrcPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jSrcPlanes[i],
+				(unsigned char *)srcPlanes[i], 0);
+	}
+	if(srcStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcStrides, srcStrides, 0);
+	if(srcOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcOffsets, srcOffsets, 0);
+	return (jint)jpegSize;
+}
+
+static void TJCompressor_encodeYUV
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	jbyteArray jDstPlanes[3]={NULL, NULL, NULL};
+	unsigned char *srcBuf=NULL, *dstPlanes[3];
+	int *dstOffsets=NULL, *dstStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0 || subsamp<0 || subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF
+		|| org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, dstobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jDstOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jDstStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+
+	bailif0(dstOffsets=(*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+	bailif0(dstStrides=(*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, dstStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(dstOffsets[i]<0)
+			_throwarg("Invalid argument in encodeYUV()");
+		if(dstStrides[i]<0 && dstOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jDstPlanes[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((*env)->GetArrayLength(env, jDstPlanes[i])<dstOffsets[i]+planeSize)
+			_throwarg("Destination plane is not large enough");
+
+		bailif0(dstPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jDstPlanes[i],
+			0));
+		dstPlanes[i]=&dstPlanes[i][dstOffsets[i]];
+	}
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjEncodeYUVPlanes(handle, &srcBuf[y*actualPitch + x*tjPixelSize[pf]],
+		width, pitch, height, pf, dstPlanes, dstStrides, subsamp, flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(dstPlanes[i] && jDstPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jDstPlanes[i], dstPlanes[i],
+				0);
+	}
+	if(dstStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstStrides, dstStrides, 0);
+	if(dstOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstOffsets, dstOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III
+	(JNIEnv *env, jobject obj, jbyteArray src, jint x, jint y, jint width,
+		jint pitch, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	TJCompressor_encodeYUV(env, obj, src, 1, x, y, width, pitch, height, pf,
+		dstobjs, jDstOffsets, jDstStrides, subsamp, flags);
+}
+
+/* TurboJPEG 1.4.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III
+	(JNIEnv *env, jobject obj, jintArray src, jint x, jint y, jint width,
+		jint stride, jint height, jint pf, jobjectArray dstobjs,
+		jintArray jDstOffsets, jintArray jDstStrides, jint subsamp, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+	TJCompressor_encodeYUV(env, obj, src, sizeof(jint), x, y, width,
+		stride*sizeof(jint), height, pf, dstobjs, jDstOffsets, jDstStrides,
+		subsamp, flags);
+
+	bailout:
+	return;
+}
+
+JNIEXPORT void JNICALL TJCompressor_encodeYUV_12
+	(JNIEnv *env, jobject obj, jarray src, jint srcElementSize, jint width,
+		jint pitch, jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0;
+	unsigned char *srcBuf=NULL, *dstBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || width<1 || height<1
+		|| pitch<0)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	arraySize=(pitch==0)? width*tjPixelSize[pf]*height:pitch*height;
+	if((*env)->GetArrayLength(env, src)*srcElementSize<arraySize)
+		_throwarg("Source buffer is not large enough");
+	if((*env)->GetArrayLength(env, dst)
+		<(jsize)tjBufSizeYUV(width, height, subsamp))
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjEncodeYUV2(handle, srcBuf, width, pitch, height, pf, dstBuf, subsamp,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(srcBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, srcBuf, 0);
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() byte source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint width, jint pitch,
+		jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	TJCompressor_encodeYUV_12(env, obj, src, 1, width, pitch, height, pf, dst,
+		subsamp, flags);
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::encodeYUV() int source */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII
+	(JNIEnv *env, jobject obj, jintArray src, jint width, jint stride,
+		jint height, jint pf, jbyteArray dst, jint subsamp, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in encodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when encoding from an integer buffer.");
+
+	TJCompressor_encodeYUV_12(env, obj, src, sizeof(jint), width,
+		stride*sizeof(jint), height, pf, dst, subsamp, flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJCompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy
+	(JNIEnv *env, jobject obj)
+{
+	tjhandle handle=0;
+
+	gethandle();
+
+	if(tjDestroy(handle)==-1) _throwtj();
+	(*env)->SetLongField(env, obj, _fid, 0);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitDecompress())==NULL) _throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::getScalingFactors() */
+JNIEXPORT jobjectArray JNICALL Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors
+	(JNIEnv *env, jclass cls)
+{
+	jclass sfcls=NULL;  jfieldID fid=0;
+	tjscalingfactor *sf=NULL;  int n=0, i;
+	jobject sfobj=NULL;
+	jobjectArray sfjava=NULL;
+
+	if((sf=tjGetScalingFactors(&n))==NULL || n==0)
+		_throwarg(tjGetErrorStr());
+
+	bailif0(sfcls=(*env)->FindClass(env, "org/libjpegturbo/turbojpeg/TJScalingFactor"));
+	bailif0(sfjava=(jobjectArray)(*env)->NewObjectArray(env, n, sfcls, 0));
+
+	for(i=0; i<n; i++)
+	{
+		bailif0(sfobj=(*env)->AllocObject(env, sfcls));
+		bailif0(fid=(*env)->GetFieldID(env, sfcls, "num", "I"));
+		(*env)->SetIntField(env, sfobj, fid, sf[i].num);
+		bailif0(fid=(*env)->GetFieldID(env, sfcls, "denom", "I"));
+		(*env)->SetIntField(env, sfobj, fid, sf[i].denom);
+		(*env)->SetObjectArrayElement(env, sfjava, i, sfobj);
+	}
+
+	bailout:
+	return sfjava;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressHeader() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize)
+{
+	tjhandle handle=0;
+	unsigned char *jpegBuf=NULL;
+	int width=0, height=0, jpegSubsamp=-1, jpegColorspace=-1;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjDecompressHeader3(handle, jpegBuf, (unsigned long)jpegSize,
+		&width, &height, &jpegSubsamp, &jpegColorspace)==-1)
+		_throwtj();
+
+	(*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);  jpegBuf=NULL;
+
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	(*env)->SetIntField(env, obj, _fid, jpegSubsamp);
+	if((_fid=(*env)->GetFieldID(env, _cls, "jpegColorspace", "I"))==0)
+		(*env)->ExceptionClear(env);
+	else
+		(*env)->SetIntField(env, obj, _fid, jpegColorspace);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	(*env)->SetIntField(env, obj, _fid, width);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	(*env)->SetIntField(env, obj, _fid, height);
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+static void TJDecompressor_decompress
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jarray dst,
+		jint dstElementSize, jint x, jint y, jint width, jint pitch, jint height,
+		jint pf, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	unsigned char *jpegBuf=NULL, *dstBuf=NULL;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, dst)*dstElementSize<arraySize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecompress2(handle, jpegBuf, (unsigned long)jpegSize,
+		&dstBuf[y*actualPitch + x*tjPixelSize[pf]], width, pitch, height, pf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, x, y, width,
+		pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, 1, 0, 0, width,
+		pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.3.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+		jint x, jint y, jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), x, y,
+		width, stride*sizeof(jint), height, pf, flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompress() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jintArray dst,
+		jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decompress()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decompressing to an integer buffer.");
+
+	TJDecompressor_decompress(env, obj, src, jpegSize, dst, sizeof(jint), 0, 0,
+		width, stride*sizeof(jint), height, pf, flags);
+
+	bailout:
+	return;
+
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize,
+		jobjectArray dstobjs, jintArray jDstOffsets, jint desiredWidth,
+		jintArray jDstStrides, jint desiredHeight, jint flags)
+{
+	tjhandle handle=0;
+	jbyteArray jDstPlanes[3]={NULL, NULL, NULL};
+	unsigned char *jpegBuf=NULL, *dstPlanes[3];
+	int *dstOffsets=NULL, *dstStrides=NULL;
+	int jpegSubsamp=-1, jpegWidth=0, jpegHeight=0;
+	int nc=0, i, width, height, scaledWidth, scaledHeight, nsf=0;
+	tjscalingfactor *sf;
+
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+
+	nc=(jpegSubsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3);
+
+	width=desiredWidth;  height=desiredHeight;
+	if(width==0) width=jpegWidth;
+	if(height==0) height=jpegHeight;
+	sf=tjGetScalingFactors(&nsf);
+	if(!sf || nsf<1)
+		_throwarg(tjGetErrorStr());
+	for(i=0; i<nsf; i++)
+	{
+		scaledWidth=TJSCALED(jpegWidth, sf[i]);
+		scaledHeight=TJSCALED(jpegHeight, sf[i]);
+		if(scaledWidth<=width && scaledHeight<=height)
+			break;
+	}
+	if(i>=nsf)
+		_throwarg("Could not scale down to desired image dimensions");
+
+	bailif0(dstOffsets=(*env)->GetPrimitiveArrayCritical(env, jDstOffsets, 0));
+	bailif0(dstStrides=(*env)->GetPrimitiveArrayCritical(env, jDstStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, scaledWidth, dstStrides[i], scaledHeight,
+			jpegSubsamp);
+		int pw=tjPlaneWidth(i, scaledWidth, jpegSubsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(dstOffsets[i]<0)
+			_throwarg("Invalid argument in decompressToYUV()");
+		if(dstStrides[i]<0 && dstOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jDstPlanes[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((*env)->GetArrayLength(env, jDstPlanes[i])<dstOffsets[i]+planeSize)
+			_throwarg("Destination plane is not large enough");
+
+		bailif0(dstPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jDstPlanes[i],
+			0));
+		dstPlanes[i]=&dstPlanes[i][dstOffsets[i]];
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+
+	if(tjDecompressToYUVPlanes(handle, jpegBuf, (unsigned long)jpegSize,
+		dstPlanes, desiredWidth, dstStrides, desiredHeight, flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(dstPlanes[i] && jDstPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jDstPlanes[i], dstPlanes[i],
+				0);
+	}
+	if(dstStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstStrides, dstStrides, 0);
+	if(dstOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jDstOffsets, dstOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::decompressToYUV() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI
+	(JNIEnv *env, jobject obj, jbyteArray src, jint jpegSize, jbyteArray dst,
+		jint flags)
+{
+	tjhandle handle=0;
+	unsigned char *jpegBuf=NULL, *dstBuf=NULL;
+	int jpegSubsamp=-1, jpegWidth=0, jpegHeight=0;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, src)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+	if((*env)->GetArrayLength(env, dst)
+		<(jsize)tjBufSizeYUV(jpegWidth, jpegHeight, jpegSubsamp))
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, src, 0));
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecompressToYUV(handle, jpegBuf, (unsigned long)jpegSize, dstBuf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, src, jpegBuf, 0);
+	return;
+}
+
+static void TJDecompressor_decodeYUV
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jarray dst, jint dstElementSize,
+		jint x, jint y, jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	tjhandle handle=0;
+	jsize arraySize=0, actualPitch;
+	jbyteArray jSrcPlanes[3]={NULL, NULL, NULL};
+	const unsigned char *srcPlanes[3];
+	unsigned char *dstBuf=NULL;
+	int *srcOffsets=NULL, *srcStrides=NULL;
+	int nc=(subsamp==org_libjpegturbo_turbojpeg_TJ_SAMP_GRAY? 1:3), i;
+
+	gethandle();
+
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF || subsamp<0
+		|| subsamp>=org_libjpegturbo_turbojpeg_TJ_NUMSAMP)
+		_throwarg("Invalid argument in decodeYUV()");
+	if(org_libjpegturbo_turbojpeg_TJ_NUMPF!=TJ_NUMPF
+		|| org_libjpegturbo_turbojpeg_TJ_NUMSAMP!=TJ_NUMSAMP)
+		_throwarg("Mismatch between Java and C API");
+
+	if((*env)->GetArrayLength(env, srcobjs)<nc)
+		_throwarg("Planes array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcOffsets)<nc)
+		_throwarg("Offsets array is too small for the subsampling type");
+	if((*env)->GetArrayLength(env, jSrcStrides)<nc)
+		_throwarg("Strides array is too small for the subsampling type");
+
+	actualPitch=(pitch==0)? width*tjPixelSize[pf]:pitch;
+	arraySize=(y+height-1)*actualPitch + (x+width)*tjPixelSize[pf];
+	if((*env)->GetArrayLength(env, dst)*dstElementSize<arraySize)
+		_throwarg("Destination buffer is not large enough");
+
+	bailif0(srcOffsets=(*env)->GetPrimitiveArrayCritical(env, jSrcOffsets, 0));
+	bailif0(srcStrides=(*env)->GetPrimitiveArrayCritical(env, jSrcStrides, 0));
+	for(i=0; i<nc; i++)
+	{
+		int planeSize=tjPlaneSizeYUV(i, width, srcStrides[i], height, subsamp);
+		int pw=tjPlaneWidth(i, width, subsamp);
+
+		if(planeSize<0 || pw<0)
+			_throwarg(tjGetErrorStr());
+
+		if(srcOffsets[i]<0)
+			_throwarg("Invalid argument in decodeYUV()");
+		if(srcStrides[i]<0 && srcOffsets[i]-planeSize+pw<0)
+			_throwarg("Negative plane stride would cause memory to be accessed below plane boundary");
+
+		bailif0(jSrcPlanes[i]=(*env)->GetObjectArrayElement(env, srcobjs, i));
+		if((*env)->GetArrayLength(env, jSrcPlanes[i])<srcOffsets[i]+planeSize)
+			_throwarg("Source plane is not large enough");
+
+		bailif0(srcPlanes[i]=(*env)->GetPrimitiveArrayCritical(env, jSrcPlanes[i],
+			0));
+		srcPlanes[i]=&srcPlanes[i][srcOffsets[i]];
+	}
+	bailif0(dstBuf=(*env)->GetPrimitiveArrayCritical(env, dst, 0));
+
+	if(tjDecodeYUVPlanes(handle, srcPlanes, srcStrides, subsamp,
+		&dstBuf[y*actualPitch + x*tjPixelSize[pf]], width, pitch, height, pf,
+		flags)==-1)
+		_throwtj();
+
+	bailout:
+	if(dstBuf) (*env)->ReleasePrimitiveArrayCritical(env, dst, dstBuf, 0);
+	for(i=0; i<nc; i++)
+	{
+		if(srcPlanes[i] && jSrcPlanes[i])
+			(*env)->ReleasePrimitiveArrayCritical(env, jSrcPlanes[i],
+				(unsigned char *)srcPlanes[i], 0);
+	}
+	if(srcStrides)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcStrides, srcStrides, 0);
+	if(srcOffsets)
+		(*env)->ReleasePrimitiveArrayCritical(env, jSrcOffsets, srcOffsets, 0);
+	return;
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() byte destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jbyteArray dst, jint x, jint y,
+		jint width, jint pitch, jint height, jint pf, jint flags)
+{
+	TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+		subsamp, dst, 1, x, y, width, pitch, height, pf, flags);
+}
+
+/* TurboJPEG 1.4.x: TJDecompressor::decodeYUV() int destination */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII
+	(JNIEnv *env, jobject obj, jobjectArray srcobjs, jintArray jSrcOffsets,
+		jintArray jSrcStrides, jint subsamp, jintArray dst, jint x, jint y,
+		jint width, jint stride, jint height, jint pf, jint flags)
+{
+	if(pf<0 || pf>=org_libjpegturbo_turbojpeg_TJ_NUMPF)
+		_throwarg("Invalid argument in decodeYUV()");
+	if(tjPixelSize[pf]!=sizeof(jint))
+		_throwarg("Pixel format must be 32-bit when decoding to an integer buffer.");
+
+	TJDecompressor_decodeYUV(env, obj, srcobjs, jSrcOffsets, jSrcStrides,
+		subsamp, dst, sizeof(jint), x, y, width, stride*sizeof(jint), height, pf,
+		flags);
+
+	bailout:
+	return;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::init() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_init
+	(JNIEnv *env, jobject obj)
+{
+	jclass cls;
+	jfieldID fid;
+	tjhandle handle;
+
+	if((handle=tjInitTransform())==NULL) _throwtj();
+
+	bailif0(cls=(*env)->GetObjectClass(env, obj));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "handle", "J"));
+	(*env)->SetLongField(env, obj, fid, (size_t)handle);
+
+	bailout:
+	return;
+}
+
+typedef struct _JNICustomFilterParams
+{
+	JNIEnv *env;
+	jobject tobj;
+	jobject cfobj;
+} JNICustomFilterParams;
+
+static int JNICustomFilter(short *coeffs, tjregion arrayRegion,
+	tjregion planeRegion, int componentIndex, int transformIndex,
+	tjtransform *transform)
+{
+	JNICustomFilterParams *params=(JNICustomFilterParams *)transform->data;
+	JNIEnv *env=params->env;
+	jobject tobj=params->tobj, cfobj=params->cfobj;
+	jobject arrayRegionObj, planeRegionObj, bufobj, borobj;
+	jclass cls;  jmethodID mid;  jfieldID fid;
+
+	bailif0(bufobj=(*env)->NewDirectByteBuffer(env, coeffs,
+		sizeof(short)*arrayRegion.w*arrayRegion.h));
+	bailif0(cls=(*env)->FindClass(env, "java/nio/ByteOrder"));
+	bailif0(mid=(*env)->GetStaticMethodID(env, cls, "nativeOrder",
+		"()Ljava/nio/ByteOrder;"));
+	bailif0(borobj=(*env)->CallStaticObjectMethod(env, cls, mid));
+	bailif0(cls=(*env)->GetObjectClass(env, bufobj));
+	bailif0(mid=(*env)->GetMethodID(env, cls, "order",
+		"(Ljava/nio/ByteOrder;)Ljava/nio/ByteBuffer;"));
+	(*env)->CallObjectMethod(env, bufobj, mid, borobj);
+	bailif0(mid=(*env)->GetMethodID(env, cls, "asShortBuffer",
+		"()Ljava/nio/ShortBuffer;"));
+	bailif0(bufobj=(*env)->CallObjectMethod(env, bufobj, mid));
+
+	bailif0(cls=(*env)->FindClass(env, "java/awt/Rectangle"));
+	bailif0(arrayRegionObj=(*env)->AllocObject(env, cls));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "x", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.x);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "y", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.y);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "width", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.w);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "height", "I"));
+	(*env)->SetIntField(env, arrayRegionObj, fid, arrayRegion.h);
+
+	bailif0(planeRegionObj=(*env)->AllocObject(env, cls));
+	bailif0(fid=(*env)->GetFieldID(env, cls, "x", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.x);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "y", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.y);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "width", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.w);
+	bailif0(fid=(*env)->GetFieldID(env, cls, "height", "I"));
+	(*env)->SetIntField(env, planeRegionObj, fid, planeRegion.h);
+
+	bailif0(cls=(*env)->GetObjectClass(env, cfobj));
+	bailif0(mid=(*env)->GetMethodID(env, cls, "customFilter",
+		"(Ljava/nio/ShortBuffer;Ljava/awt/Rectangle;Ljava/awt/Rectangle;IILorg/libjpegturbo/turbojpeg/TJTransform;)V"));
+	(*env)->CallVoidMethod(env, cfobj, mid, bufobj, arrayRegionObj,
+		planeRegionObj, componentIndex, transformIndex, tobj);
+
+	return 0;
+
+	bailout:
+	return -1;
+}
+
+/* TurboJPEG 1.2.x: TJTransformer::transform() */
+JNIEXPORT jintArray JNICALL Java_org_libjpegturbo_turbojpeg_TJTransformer_transform
+	(JNIEnv *env, jobject obj, jbyteArray jsrcBuf, jint jpegSize,
+		jobjectArray dstobjs, jobjectArray tobjs, jint flags)
+{
+	tjhandle handle=0;  int i;
+	unsigned char *jpegBuf=NULL, **dstBufs=NULL;  jsize n=0;
+	unsigned long *dstSizes=NULL;  tjtransform *t=NULL;
+	jbyteArray *jdstBufs=NULL;
+	int jpegWidth=0, jpegHeight=0, jpegSubsamp;
+	jintArray jdstSizes=0;  jint *dstSizesi=NULL;
+	JNICustomFilterParams *params=NULL;
+
+	gethandle();
+
+	if((*env)->GetArrayLength(env, jsrcBuf)<jpegSize)
+		_throwarg("Source buffer is not large enough");
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegWidth", "I"));
+	jpegWidth=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegHeight", "I"));
+	jpegHeight=(int)(*env)->GetIntField(env, obj, _fid);
+	bailif0(_fid=(*env)->GetFieldID(env, _cls, "jpegSubsamp", "I"));
+	jpegSubsamp=(int)(*env)->GetIntField(env, obj, _fid);
+
+	n=(*env)->GetArrayLength(env, dstobjs);
+	if(n!=(*env)->GetArrayLength(env, tobjs))
+		_throwarg("Mismatch between size of transforms array and destination buffers array");
+
+	if((dstBufs=(unsigned char **)malloc(sizeof(unsigned char *)*n))==NULL)
+		_throwmem();
+	if((jdstBufs=(jbyteArray *)malloc(sizeof(jbyteArray)*n))==NULL)
+		_throwmem();
+	if((dstSizes=(unsigned long *)malloc(sizeof(unsigned long)*n))==NULL)
+		_throwmem();
+	if((t=(tjtransform *)malloc(sizeof(tjtransform)*n))==NULL)
+		_throwmem();
+	if((params=(JNICustomFilterParams *)malloc(sizeof(JNICustomFilterParams)*n))
+		==NULL)
+		_throwmem();
+	for(i=0; i<n; i++)
+	{
+		dstBufs[i]=NULL;  jdstBufs[i]=NULL;  dstSizes[i]=0;
+		memset(&t[i], 0, sizeof(tjtransform));
+		memset(&params[i], 0, sizeof(JNICustomFilterParams));
+	}
+
+	for(i=0; i<n; i++)
+	{
+		jobject tobj, cfobj;
+
+		bailif0(tobj=(*env)->GetObjectArrayElement(env, tobjs, i));
+		bailif0(_cls=(*env)->GetObjectClass(env, tobj));
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "op", "I"));
+		t[i].op=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "options", "I"));
+		t[i].options=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "x", "I"));
+		t[i].r.x=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "y", "I"));
+		t[i].r.y=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "width", "I"));
+		t[i].r.w=(*env)->GetIntField(env, tobj, _fid);
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "height", "I"));
+		t[i].r.h=(*env)->GetIntField(env, tobj, _fid);
+
+		bailif0(_fid=(*env)->GetFieldID(env, _cls, "cf",
+			"Lorg/libjpegturbo/turbojpeg/TJCustomFilter;"));
+		cfobj=(*env)->GetObjectField(env, tobj, _fid);
+		if(cfobj)
+		{
+			params[i].env=env;
+			params[i].tobj=tobj;
+			params[i].cfobj=cfobj;
+			t[i].customFilter=JNICustomFilter;
+			t[i].data=(void *)&params[i];
+		}
+	}
+
+	for(i=0; i<n; i++)
+	{
+		int w=jpegWidth, h=jpegHeight;
+		if(t[i].r.w!=0) w=t[i].r.w;
+		if(t[i].r.h!=0) h=t[i].r.h;
+		bailif0(jdstBufs[i]=(*env)->GetObjectArrayElement(env, dstobjs, i));
+		if((unsigned long)(*env)->GetArrayLength(env, jdstBufs[i])
+			<tjBufSize(w, h, jpegSubsamp))
+			_throwarg("Destination buffer is not large enough");
+	}
+	bailif0(jpegBuf=(*env)->GetPrimitiveArrayCritical(env, jsrcBuf, 0));
+	for(i=0; i<n; i++)
+		bailif0(dstBufs[i]=(*env)->GetPrimitiveArrayCritical(env, jdstBufs[i], 0));
+
+	if(tjTransform(handle, jpegBuf, jpegSize, n, dstBufs, dstSizes, t,
+		flags|TJFLAG_NOREALLOC)==-1)
+		_throwtj();
+
+	for(i=0; i<n; i++)
+	{
+		(*env)->ReleasePrimitiveArrayCritical(env, jdstBufs[i], dstBufs[i], 0);
+		dstBufs[i]=NULL;
+	}
+	(*env)->ReleasePrimitiveArrayCritical(env, jsrcBuf, jpegBuf, 0);
+	jpegBuf=NULL;
+
+	jdstSizes=(*env)->NewIntArray(env, n);
+	bailif0(dstSizesi=(*env)->GetIntArrayElements(env, jdstSizes, 0));
+	for(i=0; i<n; i++) dstSizesi[i]=(int)dstSizes[i];
+
+	bailout:
+	if(dstSizesi) (*env)->ReleaseIntArrayElements(env, jdstSizes, dstSizesi, 0);
+	if(dstBufs)
+	{
+		for(i=0; i<n; i++)
+		{
+			if(dstBufs[i] && jdstBufs && jdstBufs[i])
+				(*env)->ReleasePrimitiveArrayCritical(env, jdstBufs[i], dstBufs[i], 0);
+		}
+		free(dstBufs);
+	}
+	if(jpegBuf) (*env)->ReleasePrimitiveArrayCritical(env, jsrcBuf, jpegBuf, 0);
+	if(jdstBufs) free(jdstBufs);
+	if(dstSizes) free(dstSizes);
+	if(t) free(t);
+	return jdstSizes;
+}
+
+/* TurboJPEG 1.2.x: TJDecompressor::destroy() */
+JNIEXPORT void JNICALL Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy
+	(JNIEnv *env, jobject obj)
+{
+	Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy(env, obj);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-mapfile glmark2-2021.02/src/libjpeg-turbo/turbojpeg-mapfile
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-mapfile	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/turbojpeg-mapfile	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+TURBOJPEG_1.0
+{
+	global:
+		tjInitCompress;
+		tjCompress;
+		TJBUFSIZE;
+		tjInitDecompress;
+		tjDecompressHeader;
+		tjDecompress;
+		tjDestroy;
+		tjGetErrorStr;
+	local:
+		*;
+};
+
+TURBOJPEG_1.1
+{
+	global:
+		TJBUFSIZEYUV;
+		tjDecompressHeader2;
+		tjDecompressToYUV;
+		tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+	global:
+		tjAlloc;
+		tjBufSize;
+		tjBufSizeYUV;
+		tjCompress2;
+		tjDecompress2;
+		tjEncodeYUV2;
+		tjFree;
+		tjGetScalingFactors;
+		tjInitTransform;
+		tjTransform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.4
+{
+	global:
+		tjBufSizeYUV2;
+		tjCompressFromYUV;
+		tjCompressFromYUVPlanes;
+		tjDecodeYUV;
+		tjDecodeYUVPlanes;
+		tjDecompressHeader3;
+		tjDecompressToYUV2;
+		tjDecompressToYUVPlanes;
+		tjEncodeYUV3;
+		tjEncodeYUVPlanes;
+		tjPlaneHeight;
+		tjPlaneSizeYUV;
+		tjPlaneWidth;
+} TURBOJPEG_1.2;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-mapfile.jni glmark2-2021.02/src/libjpeg-turbo/turbojpeg-mapfile.jni
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/turbojpeg-mapfile.jni	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/turbojpeg-mapfile.jni	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,92 @@
+TURBOJPEG_1.0
+{
+	global:
+		tjInitCompress;
+		tjCompress;
+		TJBUFSIZE;
+		tjInitDecompress;
+		tjDecompressHeader;
+		tjDecompress;
+		tjDestroy;
+		tjGetErrorStr;
+	local:
+		*;
+};
+
+TURBOJPEG_1.1
+{
+	global:
+		TJBUFSIZEYUV;
+		tjDecompressHeader2;
+		tjDecompressToYUV;
+		tjEncodeYUV;
+} TURBOJPEG_1.0;
+
+TURBOJPEG_1.2
+{
+	global:
+		tjAlloc;
+		tjBufSize;
+		tjBufSizeYUV;
+		tjCompress2;
+		tjDecompress2;
+		tjEncodeYUV2;
+		tjFree;
+		tjGetScalingFactors;
+		tjInitTransform;
+		tjTransform;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSize;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__III;
+		Java_org_libjpegturbo_turbojpeg_TJ_getScalingFactors;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_init;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIII_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIII_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_destroy;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_init;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressHeader;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3BI;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_destroy;
+		Java_org_libjpegturbo_turbojpeg_TJTransformer_init;
+		Java_org_libjpegturbo_turbojpeg_TJTransformer_transform;
+} TURBOJPEG_1.1;
+
+TURBOJPEG_1.3
+{
+	global:
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3BIIIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compress___3IIIIIII_3BIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3BIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompress___3BI_3IIIIIIII;
+} TURBOJPEG_1.2;
+
+TURBOJPEG_1.4
+{
+	global:
+		tjBufSizeYUV2;
+		tjCompressFromYUV;
+		tjCompressFromYUVPlanes;
+		tjDecodeYUV;
+		tjDecodeYUVPlanes;
+		tjDecompressHeader3;
+		tjDecompressToYUV2;
+		tjDecompressToYUVPlanes;
+		tjEncodeYUV3;
+		tjEncodeYUVPlanes;
+		tjPlaneHeight;
+		tjPlaneSizeYUV;
+		tjPlaneWidth;
+		Java_org_libjpegturbo_turbojpeg_TJ_bufSizeYUV__IIII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_compressFromYUV___3_3B_3II_3III_3BII;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3BIIIIII_3_3B_3I_3III;
+		Java_org_libjpegturbo_turbojpeg_TJCompressor_encodeYUV___3IIIIIII_3_3B_3I_3III;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decompressToYUV___3BI_3_3B_3II_3III;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3BIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJDecompressor_decodeYUV___3_3B_3I_3II_3IIIIIIII;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeHeight__III;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeSizeYUV__IIIII;
+		Java_org_libjpegturbo_turbojpeg_TJ_planeWidth__III;
+} TURBOJPEG_1.3;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/usage.txt glmark2-2021.02/src/libjpeg-turbo/usage.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/usage.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/usage.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,660 @@
+NOTE:  This file was modified by The libjpeg-turbo Project to include only
+information relevant to libjpeg-turbo and to wordsmith certain sections.
+
+USAGE instructions for the Independent JPEG Group's JPEG software
+=================================================================
+
+This file describes usage of the JPEG conversion programs cjpeg and djpeg,
+as well as the utility programs jpegtran, rdjpgcom and wrjpgcom.  (See
+the other documentation files if you wish to use the JPEG library within
+your own programs.)
+
+If you are on a Unix machine you may prefer to read the Unix-style manual
+pages in files cjpeg.1, djpeg.1, jpegtran.1, rdjpgcom.1, wrjpgcom.1.
+
+
+INTRODUCTION
+
+These programs implement JPEG image encoding, decoding, and transcoding.
+JPEG (pronounced "jay-peg") is a standardized compression method for
+full-color and grayscale images.
+
+
+GENERAL USAGE
+
+We provide two programs, cjpeg to compress an image file into JPEG format,
+and djpeg to decompress a JPEG file back into a conventional image format.
+
+On Unix-like systems, you say:
+        cjpeg [switches] [imagefile] >jpegfile
+or
+        djpeg [switches] [jpegfile]  >imagefile
+The programs read the specified input file, or standard input if none is
+named.  They always write to standard output (with trace/error messages to
+standard error).  These conventions are handy for piping images between
+programs.
+
+On most non-Unix systems, you say:
+        cjpeg [switches] imagefile jpegfile
+or
+        djpeg [switches] jpegfile  imagefile
+i.e., both the input and output files are named on the command line.  This
+style is a little more foolproof, and it loses no functionality if you don't
+have pipes.  (You can get this style on Unix too, if you prefer, by defining
+TWO_FILE_COMMANDLINE when you compile the programs; see install.txt.)
+
+You can also say:
+        cjpeg [switches] -outfile jpegfile  imagefile
+or
+        djpeg [switches] -outfile imagefile  jpegfile
+This syntax works on all systems, so it is useful for scripts.
+
+The currently supported image file formats are: PPM (PBMPLUS color format),
+PGM (PBMPLUS grayscale format), BMP, Targa, and RLE (Utah Raster Toolkit
+format).  (RLE is supported only if the URT library is available, which it
+isn't on most non-Unix systems.)  cjpeg recognizes the input image format
+automatically, with the exception of some Targa files.  You have to tell djpeg
+which format to generate.
+
+JPEG files are in the defacto standard JFIF file format.  There are other,
+less widely used JPEG-based file formats, but we don't support them.
+
+All switch names may be abbreviated; for example, -grayscale may be written
+-gray or -gr.  Most of the "basic" switches can be abbreviated to as little as
+one letter.  Upper and lower case are equivalent (-BMP is the same as -bmp).
+British spellings are also accepted (e.g., -greyscale), though for brevity
+these are not mentioned below.
+
+
+CJPEG DETAILS
+
+The basic command line switches for cjpeg are:
+
+        -quality N[,...]  Scale quantization tables to adjust image quality.
+                          Quality is 0 (worst) to 100 (best); default is 75.
+                          (See below for more info.)
+
+        -grayscale      Create monochrome JPEG file from color input.
+                        Be sure to use this switch when compressing a grayscale
+                        BMP file, because cjpeg isn't bright enough to notice
+                        whether a BMP file uses only shades of gray.  By
+                        saying -grayscale, you'll get a smaller JPEG file that
+                        takes less time to process.
+
+        -rgb            Create RGB JPEG file.
+                        Using this switch suppresses the conversion from RGB
+                        colorspace input to the default YCbCr JPEG colorspace.
+
+        -optimize       Perform optimization of entropy encoding parameters.
+                        Without this, default encoding parameters are used.
+                        -optimize usually makes the JPEG file a little smaller,
+                        but cjpeg runs somewhat slower and needs much more
+                        memory.  Image quality and speed of decompression are
+                        unaffected by -optimize.
+
+        -progressive    Create progressive JPEG file (see below).
+
+        -targa          Input file is Targa format.  Targa files that contain
+                        an "identification" field will not be automatically
+                        recognized by cjpeg; for such files you must specify
+                        -targa to make cjpeg treat the input as Targa format.
+                        For most Targa files, you won't need this switch.
+
+The -quality switch lets you trade off compressed file size against quality of
+the reconstructed image: the higher the quality setting, the larger the JPEG
+file, and the closer the output image will be to the original input.  Normally
+you want to use the lowest quality setting (smallest file) that decompresses
+into something visually indistinguishable from the original image.  For this
+purpose the quality setting should generally be between 50 and 95 (the default
+is 75) for photographic images.  If you see defects at -quality 75, then go up
+5 or 10 counts at a time until you are happy with the output image.  (The
+optimal setting will vary from one image to another.)
+
+-quality 100 will generate a quantization table of all 1's, minimizing loss
+in the quantization step (but there is still information loss in subsampling,
+as well as roundoff error.)  For most images, specifying a quality value above
+about 95 will increase the size of the compressed file dramatically, and while
+the quality gain from these higher quality values is measurable (using metrics
+such as PSNR or SSIM), it is rarely perceivable by human vision.
+
+In the other direction, quality values below 50 will produce very small files
+of low image quality.  Settings around 5 to 10 might be useful in preparing an
+index of a large image library, for example.  Try -quality 2 (or so) for some
+amusing Cubist effects.  (Note: quality values below about 25 generate 2-byte
+quantization tables, which are considered optional in the JPEG standard.
+cjpeg emits a warning message when you give such a quality value, because some
+other JPEG programs may be unable to decode the resulting file.  Use -baseline
+if you need to ensure compatibility at low quality values.)
+
+The -quality option has been extended in this version of cjpeg to support
+separate quality settings for luminance and chrominance (or, in general,
+separate settings for every quantization table slot.)  The principle is the
+same as chrominance subsampling:  since the human eye is more sensitive to
+spatial changes in brightness than spatial changes in color, the chrominance
+components can be quantized more than the luminance components without
+incurring any visible image quality loss.  However, unlike subsampling, this
+feature reduces data in the frequency domain instead of the spatial domain,
+which allows for more fine-grained control.  This option is useful in
+quality-sensitive applications, for which the artifacts generated by
+subsampling may be unacceptable.
+
+The -quality option accepts a comma-separated list of parameters, which
+respectively refer to the quality levels that should be assigned to the
+quantization table slots.  If there are more q-table slots than parameters,
+then the last parameter is replicated.  Thus, if only one quality parameter is
+given, this is used for both luminance and chrominance (slots 0 and 1,
+respectively), preserving the legacy behavior of cjpeg v6b and prior.  More (or
+customized) quantization tables can be set with the -qtables option and
+assigned to components with the -qslots option (see the "wizard" switches
+below.)
+
+JPEG  files  generated  with separate luminance and chrominance quality are
+fully compliant with standard JPEG decoders.
+
+CAUTION: For this setting to be useful, be sure to pass an argument of
+-sample 1x1 to cjpeg to disable chrominance subsampling.  Otherwise, the
+default subsampling level (2x2, AKA "4:2:0") will be used.
+
+The -progressive switch creates a "progressive JPEG" file.  In this type of
+JPEG file, the data is stored in multiple scans of increasing quality.  If the
+file is being transmitted over a slow communications link, the decoder can use
+the first scan to display a low-quality image very quickly, and can then
+improve the display with each subsequent scan.  The final image is exactly
+equivalent to a standard JPEG file of the same quality setting, and the total
+file size is about the same --- often a little smaller.
+
+Switches for advanced users:
+
+        -arithmetic     Use arithmetic coding.  CAUTION: arithmetic coded JPEG
+                        is not yet widely implemented, so many decoders will
+                        be unable to view an arithmetic coded JPEG file at
+                        all.
+
+        -dct int        Use integer DCT method (default).
+        -dct fast       Use fast integer DCT (less accurate).
+                        In libjpeg-turbo, the fast method is generally about
+                        5-15% faster than the int method when using the
+                        x86/x86-64 SIMD extensions (results may vary with other
+                        SIMD implementations, or when using libjpeg-turbo
+                        without SIMD extensions.)  For quality levels of 90 and
+                        below, there should be little or no perceptible
+                        difference between the two algorithms.  For quality
+                        levels above 90, however, the difference between
+                        the fast and the int methods becomes more pronounced.
+                        With quality=97, for instance, the fast method incurs
+                        generally about a 1-3 dB loss (in PSNR) relative to
+                        the int method, but this can be larger for some images.
+                        Do not use the fast method with quality levels above
+                        97.  The algorithm often degenerates at quality=98 and
+                        above and can actually produce a more lossy image than
+                        if lower quality levels had been used.  Also, in
+                        libjpeg-turbo, the fast method is not fully accerated
+                        for quality levels above 97, so it will be slower than
+                        the int method.
+        -dct float      Use floating-point DCT method.
+                        The float method is mainly a legacy feature.  It does
+                        not produce significantly more accurate results than
+                        the int method, and it is much slower.  The float
+                        method may also give different results on different
+                        machines due to varying roundoff behavior, whereas the
+                        integer methods should give the same results on all
+                        machines.
+
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+                        -restart 0 (the default) means no restart markers.
+
+        -smooth N       Smooth the input image to eliminate dithering noise.
+                        N, ranging from 1 to 100, indicates the strength of
+                        smoothing.  0 (the default) means no smoothing.
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, temporary files will be used.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+The -restart option inserts extra markers that allow a JPEG decoder to
+resynchronize after a transmission error.  Without restart markers, any damage
+to a compressed file will usually ruin the image from the point of the error
+to the end of the image; with restart markers, the damage is usually confined
+to the portion of the image up to the next restart marker.  Of course, the
+restart markers occupy extra space.  We recommend -restart 1 for images that
+will be transmitted across unreliable networks such as Usenet.
+
+The -smooth option filters the input to eliminate fine-scale noise.  This is
+often useful when converting dithered images to JPEG: a moderate smoothing
+factor of 10 to 50 gets rid of dithering patterns in the input file, resulting
+in a smaller JPEG file and a better-looking image.  Too large a smoothing
+factor will visibly blur the image, however.
+
+Switches for wizards:
+
+        -baseline       Force baseline-compatible quantization tables to be
+                        generated.  This clamps quantization values to 8 bits
+                        even at low quality settings.  (This switch is poorly
+                        named, since it does not ensure that the output is
+                        actually baseline JPEG.  For example, you can use
+                        -baseline and -progressive together.)
+
+        -qtables file   Use the quantization tables given in the specified
+                        text file.
+
+        -qslots N[,...] Select which quantization table to use for each color
+                        component.
+
+        -sample HxV[,...]  Set JPEG sampling factors for each color component.
+
+        -scans file     Use the scan script given in the specified text file.
+
+The "wizard" switches are intended for experimentation with JPEG.  If you
+don't know what you are doing, DON'T USE THEM.  These switches are documented
+further in the file wizard.txt.
+
+
+DJPEG DETAILS
+
+The basic command line switches for djpeg are:
+
+        -colors N       Reduce image to at most N colors.  This reduces the
+        or -quantize N  number of colors used in the output image, so that it
+                        can be displayed on a colormapped display or stored in
+                        a colormapped file format.  For example, if you have
+                        an 8-bit display, you'd need to reduce to 256 or fewer
+                        colors.  (-colors is the recommended name, -quantize
+                        is provided only for backwards compatibility.)
+
+        -fast           Select recommended processing options for fast, low
+                        quality output.  (The default options are chosen for
+                        highest quality output.)  Currently, this is equivalent
+                        to "-dct fast -nosmooth -onepass -dither ordered".
+
+        -grayscale      Force grayscale output even if JPEG file is color.
+                        Useful for viewing on monochrome displays; also,
+                        djpeg runs noticeably faster in this mode.
+
+        -rgb            Force RGB output even if JPEG file is grayscale.
+
+        -scale M/N      Scale the output image by a factor M/N.  Currently
+                        the scale factor must be M/8, where M is an integer
+                        between 1 and 16 inclusive, or any reduced fraction
+                        thereof (such as 1/2, 3/4, etc.  Scaling is handy if
+                        the image is larger than your screen; also, djpeg runs
+                        much faster when scaling down the output.
+
+        -bmp            Select BMP output format (Windows flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -gif            Select GIF output format.  Since GIF does not support
+                        more than 256 colors, -colors 256 is assumed (unless
+                        you specify a smaller number of colors).  If you
+                        specify -fast, the default number of colors is 216.
+
+        -os2            Select BMP output format (OS/2 1.x flavor).  8-bit
+                        colormapped format is emitted if -colors or -grayscale
+                        is specified, or if the JPEG file is grayscale;
+                        otherwise, 24-bit full-color format is emitted.
+
+        -pnm            Select PBMPLUS (PPM/PGM) output format (this is the
+                        default format).  PGM is emitted if the JPEG file is
+                        grayscale or if -grayscale is specified; otherwise
+                        PPM is emitted.
+
+        -rle            Select RLE output format.  (Requires URT library.)
+
+        -targa          Select Targa output format.  Grayscale format is
+                        emitted if the JPEG file is grayscale or if
+                        -grayscale is specified; otherwise, colormapped format
+                        is emitted if -colors is specified; otherwise, 24-bit
+                        full-color format is emitted.
+
+Switches for advanced users:
+
+        -dct int        Use integer DCT method (default).
+        -dct fast       Use fast integer DCT (less accurate).
+                        In libjpeg-turbo, the fast method is generally about
+                        5-15% faster than the int method when using the
+                        x86/x86-64 SIMD extensions (results may vary with other
+                        SIMD implementations, or when using libjpeg-turbo
+                        without SIMD extensions.)  If the JPEG image was
+                        compressed using a quality level of 85 or below, then
+                        there should be little or no perceptible difference
+                        between the two algorithms.  When decompressing images
+                        that were compressed using quality levels above 85,
+                        however, the difference between the fast and int
+                        methods becomes more pronounced.  With images
+                        compressed using quality=97, for instance, the fast
+                        method incurs generally about a 4-6 dB loss (in PSNR)
+                        relative to the int method, but this can be larger for
+                        some images.  If you can avoid it, do not use the fast
+                        method when decompressing images that were compressed
+                        using quality levels above 97.  The algorithm often
+                        degenerates for such images and can actually produce
+                        a more lossy output image than if the JPEG image had
+                        been compressed using lower quality levels.
+        -dct float      Use floating-point DCT method.
+                        The float method is mainly a legacy feature.  It does
+                        not produce significantly more accurate results than
+                        the int method, and it is much slower.  The float
+                        method may also give different results on different
+                        machines due to varying roundoff behavior, whereas the
+                        integer methods should give the same results on all
+                        machines.
+
+        -dither fs      Use Floyd-Steinberg dithering in color quantization.
+        -dither ordered Use ordered dithering in color quantization.
+        -dither none    Do not use dithering in color quantization.
+                        By default, Floyd-Steinberg dithering is applied when
+                        quantizing colors; this is slow but usually produces
+                        the best results.  Ordered dither is a compromise
+                        between speed and quality; no dithering is fast but
+                        usually looks awful.  Note that these switches have
+                        no effect unless color quantization is being done.
+                        Ordered dither is only available in -onepass mode.
+
+        -map FILE       Quantize to the colors used in the specified image
+                        file.  This is useful for producing multiple files
+                        with identical color maps, or for forcing a predefined
+                        set of colors to be used.  The FILE must be a GIF
+                        or PPM file.  This option overrides -colors and
+                        -onepass.
+
+        -nosmooth       Use a faster, lower-quality upsampling routine.
+
+        -onepass        Use one-pass instead of two-pass color quantization.
+                        The one-pass method is faster and needs less memory,
+                        but it produces a lower-quality image.  -onepass is
+                        ignored unless you also say -colors N.  Also,
+                        the one-pass method is always used for grayscale
+                        output (the two-pass method is no improvement then).
+
+        -maxmemory N    Set limit for amount of memory to use in processing
+                        large images.  Value is in thousands of bytes, or
+                        millions of bytes if "M" is attached to the number.
+                        For example, -max 4m selects 4000000 bytes.  If more
+                        space is needed, temporary files will be used.
+
+        -verbose        Enable debug printout.  More -v's give more printout.
+        or  -debug      Also, version information is printed at startup.
+
+
+HINTS FOR CJPEG
+
+Color GIF files are not the ideal input for JPEG; JPEG is really intended for
+compressing full-color (24-bit) images.  In particular, don't try to convert
+cartoons, line drawings, and other images that have only a few distinct
+colors.  GIF works great on these, JPEG does not.  If you want to convert a
+GIF to JPEG, you should experiment with cjpeg's -quality and -smooth options
+to get a satisfactory conversion.  -smooth 10 or so is often helpful.
+
+Avoid running an image through a series of JPEG compression/decompression
+cycles.  Image quality loss will accumulate; after ten or so cycles the image
+may be noticeably worse than it was after one cycle.  It's best to use a
+lossless format while manipulating an image, then convert to JPEG format when
+you are ready to file the image away.
+
+The -optimize option to cjpeg is worth using when you are making a "final"
+version for posting or archiving.  It's also a win when you are using low
+quality settings to make very small JPEG files; the percentage improvement
+is often a lot more than it is on larger files.  (At present, -optimize
+mode is always selected when generating progressive JPEG files.)
+
+Support for GIF input files was removed in cjpeg v6b due to concerns over
+the Unisys LZW patent.  Although this patent expired in 2006, cjpeg still
+lacks GIF support, for these historical reasons.  (Conversion of GIF files to
+JPEG is usually a bad idea anyway.)
+
+
+HINTS FOR DJPEG
+
+To get a quick preview of an image, use the -grayscale and/or -scale switches.
+"-grayscale -scale 1/8" is the fastest case.
+
+Several options are available that trade off image quality to gain speed.
+"-fast" turns on the recommended settings.
+
+"-dct fast" and/or "-nosmooth" gain speed at a small sacrifice in quality.
+When producing a color-quantized image, "-onepass -dither ordered" is fast but
+much lower quality than the default behavior.  "-dither none" may give
+acceptable results in two-pass mode, but is seldom tolerable in one-pass mode.
+
+Two-pass color quantization requires a good deal of memory; on MS-DOS machines
+it may run out of memory even with -maxmemory 0.  In that case you can still
+decompress, with some loss of image quality, by specifying -onepass for
+one-pass quantization.
+
+To avoid the Unisys LZW patent (now expired), djpeg produces uncompressed GIF
+files.  These are larger than they should be, but are readable by standard GIF
+decoders.
+
+
+HINTS FOR BOTH PROGRAMS
+
+If more space is needed than will fit in the available main memory (as
+determined by -maxmemory), temporary files will be used.  (MS-DOS versions
+will try to get extended or expanded memory first.)  The temporary files are
+often rather large: in typical cases they occupy three bytes per pixel, for
+example 3*800*600 = 1.44Mb for an 800x600 image.  If you don't have enough
+free disk space, leave out -progressive and -optimize (for cjpeg) or specify
+-onepass (for djpeg).
+
+On MS-DOS, the temporary files are created in the directory named by the TMP
+or TEMP environment variable, or in the current directory if neither of those
+exist.  Amiga implementations put the temp files in the directory named by
+JPEGTMP:, so be sure to assign JPEGTMP: to a disk partition with adequate free
+space.
+
+The default memory usage limit (-maxmemory) is set when the software is
+compiled.  If you get an "insufficient memory" error, try specifying a smaller
+-maxmemory value, even -maxmemory 0 to use the absolute minimum space.  You
+may want to recompile with a smaller default value if this happens often.
+
+On machines that have "environment" variables, you can define the environment
+variable JPEGMEM to set the default memory limit.  The value is specified as
+described for the -maxmemory switch.  JPEGMEM overrides the default value
+specified when the program was compiled, and itself is overridden by an
+explicit -maxmemory switch.
+
+On MS-DOS machines, -maxmemory is the amount of main (conventional) memory to
+use.  (Extended or expanded memory is also used if available.)  Most
+DOS-specific versions of this software do their own memory space estimation
+and do not need you to specify -maxmemory.
+
+
+JPEGTRAN
+
+jpegtran performs various useful transformations of JPEG files.
+It can translate the coded representation from one variant of JPEG to another,
+for example from baseline JPEG to progressive JPEG or vice versa.  It can also
+perform some rearrangements of the image data, for example turning an image
+from landscape to portrait format by rotation.  For EXIF files and JPEG files
+containing Exif data, you may prefer to use exiftran instead.
+
+jpegtran works by rearranging the compressed data (DCT coefficients), without
+ever fully decoding the image.  Therefore, its transformations are lossless:
+there is no image degradation at all, which would not be true if you used
+djpeg followed by cjpeg to accomplish the same conversion.  But by the same
+token, jpegtran cannot perform lossy operations such as changing the image
+quality.  However, while the image data is losslessly transformed, metadata
+can be removed.  See the -copy option for specifics.
+
+jpegtran uses a command line syntax similar to cjpeg or djpeg.
+On Unix-like systems, you say:
+        jpegtran [switches] [inputfile] >outputfile
+On most non-Unix systems, you say:
+        jpegtran [switches] inputfile outputfile
+where both the input and output files are JPEG files.
+
+To specify the coded JPEG representation used in the output file,
+jpegtran accepts a subset of the switches recognized by cjpeg:
+        -optimize       Perform optimization of entropy encoding parameters.
+        -progressive    Create progressive JPEG file.
+        -arithmetic     Use arithmetic coding.
+        -restart N      Emit a JPEG restart marker every N MCU rows, or every
+                        N MCU blocks if "B" is attached to the number.
+        -scans file     Use the scan script given in the specified text file.
+See the previous discussion of cjpeg for more details about these switches.
+If you specify none of these switches, you get a plain baseline-JPEG output
+file.  The quality setting and so forth are determined by the input file.
+
+The image can be losslessly transformed by giving one of these switches:
+        -flip horizontal        Mirror image horizontally (left-right).
+        -flip vertical          Mirror image vertically (top-bottom).
+        -rotate 90              Rotate image 90 degrees clockwise.
+        -rotate 180             Rotate image 180 degrees.
+        -rotate 270             Rotate image 270 degrees clockwise (or 90 ccw).
+        -transpose              Transpose image (across UL-to-LR axis).
+        -transverse             Transverse transpose (across UR-to-LL axis).
+
+The transpose transformation has no restrictions regarding image dimensions.
+The other transformations operate rather oddly if the image dimensions are not
+a multiple of the iMCU size (usually 8 or 16 pixels), because they can only
+transform complete blocks of DCT coefficient data in the desired way.
+
+jpegtran's default behavior when transforming an odd-size image is designed
+to preserve exact reversibility and mathematical consistency of the
+transformation set.  As stated, transpose is able to flip the entire image
+area.  Horizontal mirroring leaves any partial iMCU column at the right edge
+untouched, but is able to flip all rows of the image.  Similarly, vertical
+mirroring leaves any partial iMCU row at the bottom edge untouched, but is
+able to flip all columns.  The other transforms can be built up as sequences
+of transpose and flip operations; for consistency, their actions on edge
+pixels are defined to be the same as the end result of the corresponding
+transpose-and-flip sequence.
+
+For practical use, you may prefer to discard any untransformable edge pixels
+rather than having a strange-looking strip along the right and/or bottom edges
+of a transformed image.  To do this, add the -trim switch:
+        -trim           Drop non-transformable edge blocks.
+Obviously, a transformation with -trim is not reversible, so strictly speaking
+jpegtran with this switch is not lossless.  Also, the expected mathematical
+equivalences between the transformations no longer hold.  For example,
+"-rot 270 -trim" trims only the bottom edge, but "-rot 90 -trim" followed by
+"-rot 180 -trim" trims both edges.
+
+If you are only interested in perfect transformations, add the -perfect switch:
+        -perfect        Fail with an error if the transformation is not
+                        perfect.
+For example, you may want to do
+  jpegtran -rot 90 -perfect foo.jpg || djpeg foo.jpg | pnmflip -r90 | cjpeg
+to do a perfect rotation, if available, or an approximated one if not.
+
+This version of jpegtran also offers a lossless crop option, which discards
+data outside of a given image region but losslessly preserves what is inside.
+Like the rotate and flip transforms, lossless crop is restricted by the current
+JPEG format; the upper left corner of the selected region must fall on an iMCU
+boundary.  If it doesn't, then it is silently moved up and/or left to the
+nearest iMCU boundary (the lower right corner is unchanged.)  Thus, the output
+image covers at least the requested region, but it may cover more.  The
+adjustment of the region dimensions may be optionally disabled by attaching an
+'f' character ("force") to the width or height number.
+
+The image can be losslessly cropped by giving the switch:
+        -crop WxH+X+Y   Crop to a rectangular region of width W and height H,
+                        starting at point X,Y.
+
+Other not-strictly-lossless transformation switches are:
+
+        -grayscale      Force grayscale output.
+This option discards the chrominance channels if the input image is YCbCr
+(ie, a standard color JPEG), resulting in a grayscale JPEG file.  The
+luminance channel is preserved exactly, so this is a better method of reducing
+to grayscale than decompression, conversion, and recompression.  This switch
+is particularly handy for fixing a monochrome picture that was mistakenly
+encoded as a color JPEG.  (In such a case, the space savings from getting rid
+of the near-empty chroma channels won't be large; but the decoding time for
+a grayscale JPEG is substantially less than that for a color JPEG.)
+
+jpegtran also recognizes these switches that control what to do with "extra"
+markers, such as comment blocks:
+        -copy none      Copy no extra markers from source file.  This setting
+                        suppresses all comments and other metadata in the
+                        source file.
+        -copy comments  Copy only comment markers.  This setting copies
+                        comments from the source file but discards any other
+                        metadata.
+        -copy all       Copy all extra markers.  This setting preserves
+                        miscellaneous markers found in the source file, such
+                        as JFIF thumbnails, Exif data, and Photoshop settings.
+                        In some files, these extra markers can be sizable.
+                        Note that this option will copy thumbnails as-is;
+                        they will not be transformed.
+The default behavior is -copy comments.  (Note: in IJG releases v6 and v6a,
+jpegtran always did the equivalent of -copy none.)
+
+Additional switches recognized by jpegtran are:
+        -outfile filename
+        -maxmemory N
+        -verbose
+        -debug
+These work the same as in cjpeg or djpeg.
+
+
+THE COMMENT UTILITIES
+
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+
+We provide two utility programs to display COM block contents and add COM
+blocks to a JPEG file.
+
+rdjpgcom searches a JPEG file and prints the contents of any COM blocks on
+standard output.  The command line syntax is
+        rdjpgcom [-raw] [-verbose] [inputfilename]
+The switch "-raw" (or just "-r") causes rdjpgcom to output non-printable
+characters in JPEG comments.  These characters are normally escaped for
+security reasons.
+The switch "-verbose" (or just "-v") causes rdjpgcom to also display the JPEG
+image dimensions.  If you omit the input file name from the command line,
+the JPEG file is read from standard input.  (This may not work on some
+operating systems, if binary data can't be read from stdin.)
+
+wrjpgcom adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks, but you
+can delete the old COM blocks if you wish.  wrjpgcom produces a new JPEG
+file; it does not modify the input file.  DO NOT try to overwrite the input
+file by directing wrjpgcom's output back into it; on most systems this will
+just destroy your file.
+
+The command line syntax for wrjpgcom is similar to cjpeg's.  On Unix-like
+systems, it is
+        wrjpgcom [switches] [inputfilename]
+The output file is written to standard output.  The input file comes from
+the named file, or from standard input if no input file is named.
+
+On most non-Unix systems, the syntax is
+        wrjpgcom [switches] inputfilename outputfilename
+where both input and output file names must be given explicitly.
+
+wrjpgcom understands three switches:
+        -replace                 Delete any existing COM blocks from the file.
+        -comment "Comment text"  Supply new COM text on command line.
+        -cfile name              Read text for new COM block from named file.
+(Switch names can be abbreviated.)  If you have only one line of comment text
+to add, you can provide it on the command line with -comment.  The comment
+text must be surrounded with quotes so that it is treated as a single
+argument.  Longer comments can be read from a text file.
+
+If you give neither -comment nor -cfile, then wrjpgcom will read the comment
+text from standard input.  (In this case an input image file name MUST be
+supplied, so that the source JPEG file comes from somewhere else.)  You can
+enter multiple lines, up to 64KB worth.  Type an end-of-file indicator
+(usually control-D or control-Z) to terminate the comment text entry.
+
+wrjpgcom will not add a COM block if the provided comment string is empty.
+Therefore -replace -comment "" can be used to delete all COM blocks from a
+file.
+
+These utility programs do not depend on the IJG JPEG library.  In
+particular, the source code for rdjpgcom is intended as an illustration of
+the minimum amount of code required to parse a JPEG file header correctly.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wizard.txt glmark2-2021.02/src/libjpeg-turbo/wizard.txt
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wizard.txt	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wizard.txt	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,211 @@
+Advanced usage instructions for the Independent JPEG Group's JPEG software
+==========================================================================
+
+This file describes cjpeg's "switches for wizards".
+
+The "wizard" switches are intended for experimentation with JPEG by persons
+who are reasonably knowledgeable about the JPEG standard.  If you don't know
+what you are doing, DON'T USE THESE SWITCHES.  You'll likely produce files
+with worse image quality and/or poorer compression than you'd get from the
+default settings.  Furthermore, these switches must be used with caution
+when making files intended for general use, because not all JPEG decoders
+will support unusual JPEG parameter settings.
+
+
+Quantization Table Adjustment
+-----------------------------
+
+Ordinarily, cjpeg starts with a default set of tables (the same ones given
+as examples in the JPEG standard) and scales them up or down according to
+the -quality setting.  The details of the scaling algorithm can be found in
+jcparam.c.  At very low quality settings, some quantization table entries
+can get scaled up to values exceeding 255.  Although 2-byte quantization
+values are supported by the IJG software, this feature is not in baseline
+JPEG and is not supported by all implementations.  If you need to ensure
+wide compatibility of low-quality files, you can constrain the scaled
+quantization values to no more than 255 by giving the -baseline switch.
+Note that use of -baseline will result in poorer quality for the same file
+size, since more bits than necessary are expended on higher AC coefficients.
+
+You can substitute a different set of quantization values by using the
+-qtables switch:
+
+        -qtables file   Use the quantization tables given in the named file.
+
+The specified file should be a text file containing decimal quantization
+values.  The file should contain one to four tables, each of 64 elements.
+The tables are implicitly numbered 0,1,etc. in order of appearance.  Table
+entries appear in normal array order (NOT in the zigzag order in which they
+will be stored in the JPEG file).
+
+Quantization table files are free format, in that arbitrary whitespace can
+appear between numbers.  Also, comments can be included: a comment starts
+with '#' and extends to the end of the line.  Here is an example file that
+duplicates the default quantization tables:
+
+        # Quantization tables given in JPEG spec, section K.1
+
+        # This is table 0 (the luminance table):
+          16  11  10  16  24  40  51  61
+          12  12  14  19  26  58  60  55
+          14  13  16  24  40  57  69  56
+          14  17  22  29  51  87  80  62
+          18  22  37  56  68 109 103  77
+          24  35  55  64  81 104 113  92
+          49  64  78  87 103 121 120 101
+          72  92  95  98 112 100 103  99
+
+        # This is table 1 (the chrominance table):
+          17  18  24  47  99  99  99  99
+          18  21  26  66  99  99  99  99
+          24  26  56  99  99  99  99  99
+          47  66  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+          99  99  99  99  99  99  99  99
+
+If the -qtables switch is used without -quality, then the specified tables
+are used exactly as-is.  If both -qtables and -quality are used, then the
+tables taken from the file are scaled in the same fashion that the default
+tables would be scaled for that quality setting.  If -baseline appears, then
+the quantization values are constrained to the range 1-255.
+
+By default, cjpeg will use quantization table 0 for luminance components and
+table 1 for chrominance components.  To override this choice, use the -qslots
+switch:
+
+        -qslots N[,...]         Select which quantization table to use for
+                                each color component.
+
+The -qslots switch specifies a quantization table number for each color
+component, in the order in which the components appear in the JPEG SOF marker.
+For example, to create a separate table for each of Y,Cb,Cr, you could
+provide a -qtables file that defines three quantization tables and say
+"-qslots 0,1,2".  If -qslots gives fewer table numbers than there are color
+components, then the last table number is repeated as necessary.
+
+
+Sampling Factor Adjustment
+--------------------------
+
+By default, cjpeg uses 2:1 horizontal and vertical downsampling when
+compressing YCbCr data, and no downsampling for all other color spaces.
+You can override this default with the -sample switch:
+
+        -sample HxV[,...]       Set JPEG sampling factors for each color
+                                component.
+
+The -sample switch specifies the JPEG sampling factors for each color
+component, in the order in which they appear in the JPEG SOF marker.
+If you specify fewer HxV pairs than there are components, the remaining
+components are set to 1x1 sampling.  For example, the default YCbCr setting
+is equivalent to "-sample 2x2,1x1,1x1", which can be abbreviated to
+"-sample 2x2".
+
+There are still some JPEG decoders in existence that support only 2x1
+sampling (also called 4:2:2 sampling).  Compatibility with such decoders can
+be achieved by specifying "-sample 2x1".  This is not recommended unless
+really necessary, since it increases file size and encoding/decoding time
+with very little quality gain.
+
+
+Multiple Scan / Progression Control
+-----------------------------------
+
+By default, cjpeg emits a single-scan sequential JPEG file.  The
+-progressive switch generates a progressive JPEG file using a default series
+of progression parameters.  You can create multiple-scan sequential JPEG
+files or progressive JPEG files with custom progression parameters by using
+the -scans switch:
+
+        -scans file     Use the scan sequence given in the named file.
+
+The specified file should be a text file containing a "scan script".
+The script specifies the contents and ordering of the scans to be emitted.
+Each entry in the script defines one scan.  A scan definition specifies
+the components to be included in the scan, and for progressive JPEG it also
+specifies the progression parameters Ss,Se,Ah,Al for the scan.  Scan
+definitions are separated by semicolons (';').  A semicolon after the last
+scan definition is optional.
+
+Each scan definition contains one to four component indexes, optionally
+followed by a colon (':') and the four progressive-JPEG parameters.  The
+component indexes denote which color component(s) are to be transmitted in
+the scan.  Components are numbered in the order in which they appear in the
+JPEG SOF marker, with the first component being numbered 0.  (Note that these
+indexes are not the "component ID" codes assigned to the components, just
+positional indexes.)
+
+The progression parameters for each scan are:
+        Ss      Zigzag index of first coefficient included in scan
+        Se      Zigzag index of last coefficient included in scan
+        Ah      Zero for first scan of a coefficient, else Al of prior scan
+        Al      Successive approximation low bit position for scan
+If the progression parameters are omitted, the values 0,63,0,0 are used,
+producing a sequential JPEG file.  cjpeg automatically determines whether
+the script represents a progressive or sequential file, by observing whether
+Ss and Se values other than 0 and 63 appear.  (The -progressive switch is
+not needed to specify this; in fact, it is ignored when -scans appears.)
+The scan script must meet the JPEG restrictions on progression sequences.
+(cjpeg checks that the spec's requirements are obeyed.)
+
+Scan script files are free format, in that arbitrary whitespace can appear
+between numbers and around punctuation.  Also, comments can be included: a
+comment starts with '#' and extends to the end of the line.  For additional
+legibility, commas or dashes can be placed between values.  (Actually, any
+single punctuation character other than ':' or ';' can be inserted.)  For
+example, the following two scan definitions are equivalent:
+        0 1 2: 0 63 0 0;
+        0,1,2 : 0-63, 0,0 ;
+
+Here is an example of a scan script that generates a partially interleaved
+sequential JPEG file:
+
+        0;                      # Y only in first scan
+        1 2;                    # Cb and Cr in second scan
+
+Here is an example of a progressive scan script using only spectral selection
+(no successive approximation):
+
+        # Interleaved DC scan for Y,Cb,Cr:
+        0,1,2: 0-0,   0, 0 ;
+        # AC scans:
+        0:     1-2,   0, 0 ;    # First two Y AC coefficients
+        0:     3-5,   0, 0 ;    # Three more
+        1:     1-63,  0, 0 ;    # All AC coefficients for Cb
+        2:     1-63,  0, 0 ;    # All AC coefficients for Cr
+        0:     6-9,   0, 0 ;    # More Y coefficients
+        0:     10-63, 0, 0 ;    # Remaining Y coefficients
+
+Here is an example of a successive-approximation script.  This is equivalent
+to the default script used by "cjpeg -progressive" for YCbCr images:
+
+        # Initial DC scan for Y,Cb,Cr (lowest bit not sent)
+        0,1,2: 0-0,   0, 1 ;
+        # First AC scan: send first 5 Y AC coefficients, minus 2 lowest bits:
+        0:     1-5,   0, 2 ;
+        # Send all Cr,Cb AC coefficients, minus lowest bit:
+        # (chroma data is usually too small to be worth subdividing further;
+        #  but note we send Cr first since eye is least sensitive to Cb)
+        2:     1-63,  0, 1 ;
+        1:     1-63,  0, 1 ;
+        # Send remaining Y AC coefficients, minus 2 lowest bits:
+        0:     6-63,  0, 2 ;
+        # Send next-to-lowest bit of all Y AC coefficients:
+        0:     1-63,  2, 1 ;
+        # At this point we've sent all but the lowest bit of all coefficients.
+        # Send lowest bit of DC coefficients
+        0,1,2: 0-0,   1, 0 ;
+        # Send lowest bit of AC coefficients
+        2:     1-63,  1, 0 ;
+        1:     1-63,  1, 0 ;
+        # Y AC lowest bit scan is last; it's usually the largest scan
+        0:     1-63,  1, 0 ;
+
+It may be worth pointing out that this script is tuned for quality settings
+of around 50 to 75.  For lower quality settings, you'd probably want to use
+a script with fewer stages of successive approximation (otherwise the
+initial scans will be really bad).  For higher quality settings, you might
+want to use more stages of successive approximation (so that the initial
+scans are not too large).
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrbmp.c glmark2-2021.02/src/libjpeg-turbo/wrbmp.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrbmp.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrbmp.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,493 @@
+/*
+ * wrbmp.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2013, Linaro Limited.
+ * Copyright (C) 2014-2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Microsoft "BMP"
+ * format (MS Windows 3.x and OS/2 1.x flavors).
+ * Either 8-bit colormapped or 24-bit full-color format can be written.
+ * No compression is supported.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * This code contributed by James Arthur Boucher.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "jconfigint.h"
+
+#ifdef BMP_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+/*
+ * Since BMP stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * BMP file during finish_output.  The virtual array contains one JSAMPLE per
+ * pixel if the output is grayscale or colormapped, three if it is full color.
+ */
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  boolean is_os2;               /* saves the OS2 format request flag */
+
+  jvirt_sarray_ptr whole_image; /* needed to reverse row order */
+  JDIMENSION data_width;        /* JSAMPLEs per row */
+  JDIMENSION row_width;         /* physical width of one row in the BMP file */
+  int pad_bytes;                /* number of padding bytes needed per row */
+  JDIMENSION cur_output_row;    /* next row# to write to virtual array */
+} bmp_dest_struct;
+
+typedef bmp_dest_struct *bmp_dest_ptr;
+
+
+/* Forward declarations */
+LOCAL(void) write_colormap
+        (j_decompress_ptr cinfo, bmp_dest_ptr dest, int map_colors,
+         int map_entry_size);
+
+
+static INLINE boolean is_big_endian(void)
+{
+  int test_value = 1;
+  if(*(char *)&test_value != 1)
+    return TRUE;
+  return FALSE;
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+/* This version is for writing 24-bit pixels */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data.  Note destination values must be in BGR order
+   * (even though Microsoft's own documents say the opposite).
+   */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+
+  if(cinfo->out_color_space == JCS_RGB565) {
+    boolean big_endian = is_big_endian();
+    unsigned short *inptr2 = (unsigned short *)inptr;
+    for (col = cinfo->output_width; col > 0; col--) {
+      if (big_endian) {
+        outptr[0] = (*inptr2 >> 5) & 0xF8;
+        outptr[1] = ((*inptr2 << 5) & 0xE0) | ((*inptr2 >> 11) & 0x1C);
+        outptr[2] = *inptr2 & 0xF8;
+      } else {
+        outptr[0] = (*inptr2 << 3) & 0xF8;
+        outptr[1] = (*inptr2 >> 3) & 0xFC;
+        outptr[2] = (*inptr2 >> 8) & 0xF8;
+      }
+      outptr += 3;
+      inptr2++;
+    }
+  } else {
+    for (col = cinfo->output_width; col > 0; col--) {
+      outptr[2] = *inptr++;       /* can omit GETJSAMPLE() safely */
+      outptr[1] = *inptr++;
+      outptr[0] = *inptr++;
+      outptr += 3;
+    }
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* This version is for grayscale OR quantized color output */
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  int pad;
+
+  /* Access next row in virtual array */
+  image_ptr = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->whole_image,
+     dest->cur_output_row, (JDIMENSION) 1, TRUE);
+  dest->cur_output_row++;
+
+  /* Transfer data. */
+  inptr = dest->pub.buffer[0];
+  outptr = image_ptr[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = *inptr++;       /* can omit GETJSAMPLE() safely */
+  }
+
+  /* Zero out the pad bytes. */
+  pad = dest->pad_bytes;
+  while (--pad >= 0)
+    *outptr++ = 0;
+}
+
+
+/*
+ * Startup: normally writes the file header.
+ * In this module we may as well postpone everything until finish_output.
+ */
+
+METHODDEF(void)
+start_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* no work here */
+}
+
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the BMP file.
+ *
+ * First, routines to write the Windows and OS/2 variants of the file header.
+ */
+
+LOCAL(void)
+write_bmp_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write a Windows-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpinfoheader[40];
+#define PUT_2B(array,offset,value)  \
+        (array[offset] = (char) ((value) & 0xFF), \
+         array[offset+1] = (char) (((value) >> 8) & 0xFF))
+#define PUT_4B(array,offset,value)  \
+        (array[offset] = (char) ((value) & 0xFF), \
+         array[offset+1] = (char) (((value) >> 8) & 0xFF), \
+         array[offset+2] = (char) (((value) >> 16) & 0xFF), \
+         array[offset+3] = (char) (((value) >> 24) & 0xFF))
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 40 + cmap_entries * 4; /* Header and colormap */
+  bfSize = headersize + (long) dest->row_width * (long) cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpinfoheader, sizeof(bmpinfoheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPINFOHEADER) */
+  PUT_2B(bmpinfoheader, 0, 40); /* biSize */
+  PUT_4B(bmpinfoheader, 4, cinfo->output_width); /* biWidth */
+  PUT_4B(bmpinfoheader, 8, cinfo->output_height); /* biHeight */
+  PUT_2B(bmpinfoheader, 12, 1); /* biPlanes - must be 1 */
+  PUT_2B(bmpinfoheader, 14, bits_per_pixel); /* biBitCount */
+  /* we leave biCompression = 0, for none */
+  /* we leave biSizeImage = 0; this is correct for uncompressed data */
+  if (cinfo->density_unit == 2) { /* if have density in dots/cm, then */
+    PUT_4B(bmpinfoheader, 24, (long) (cinfo->X_density*100)); /* XPels/M */
+    PUT_4B(bmpinfoheader, 28, (long) (cinfo->Y_density*100)); /* XPels/M */
+  }
+  PUT_2B(bmpinfoheader, 32, cmap_entries); /* biClrUsed */
+  /* we leave biClrImportant = 0 */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpinfoheader, 40) != (size_t) 40)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 4);
+}
+
+
+LOCAL(void)
+write_os2_header (j_decompress_ptr cinfo, bmp_dest_ptr dest)
+/* Write an OS2-style BMP file header, including colormap if needed */
+{
+  char bmpfileheader[14];
+  char bmpcoreheader[12];
+  long headersize, bfSize;
+  int bits_per_pixel, cmap_entries;
+
+  /* Compute colormap size and total file size */
+  if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* Colormapped RGB */
+      bits_per_pixel = 8;
+      cmap_entries = 256;
+    } else {
+      /* Unquantized, full color RGB */
+      bits_per_pixel = 24;
+      cmap_entries = 0;
+    }
+  } else if (cinfo->out_color_space == JCS_RGB565) {
+    bits_per_pixel = 24;
+    cmap_entries   = 0;
+  } else {
+    /* Grayscale output.  We need to fake a 256-entry colormap. */
+    bits_per_pixel = 8;
+    cmap_entries = 256;
+  }
+  /* File size */
+  headersize = 14 + 12 + cmap_entries * 3; /* Header and colormap */
+  bfSize = headersize + (long) dest->row_width * (long) cinfo->output_height;
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(bmpfileheader, sizeof(bmpfileheader));
+  MEMZERO(bmpcoreheader, sizeof(bmpcoreheader));
+
+  /* Fill the file header */
+  bmpfileheader[0] = 0x42;      /* first 2 bytes are ASCII 'B', 'M' */
+  bmpfileheader[1] = 0x4D;
+  PUT_4B(bmpfileheader, 2, bfSize); /* bfSize */
+  /* we leave bfReserved1 & bfReserved2 = 0 */
+  PUT_4B(bmpfileheader, 10, headersize); /* bfOffBits */
+
+  /* Fill the info header (Microsoft calls this a BITMAPCOREHEADER) */
+  PUT_2B(bmpcoreheader, 0, 12); /* bcSize */
+  PUT_2B(bmpcoreheader, 4, cinfo->output_width); /* bcWidth */
+  PUT_2B(bmpcoreheader, 6, cinfo->output_height); /* bcHeight */
+  PUT_2B(bmpcoreheader, 8, 1);  /* bcPlanes - must be 1 */
+  PUT_2B(bmpcoreheader, 10, bits_per_pixel); /* bcBitCount */
+
+  if (JFWRITE(dest->pub.output_file, bmpfileheader, 14) != (size_t) 14)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+  if (JFWRITE(dest->pub.output_file, bmpcoreheader, 12) != (size_t) 12)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+
+  if (cmap_entries > 0)
+    write_colormap(cinfo, dest, cmap_entries, 3);
+}
+
+
+/*
+ * Write the colormap.
+ * Windows uses BGR0 map entries; OS/2 uses BGR entries.
+ */
+
+LOCAL(void)
+write_colormap (j_decompress_ptr cinfo, bmp_dest_ptr dest,
+                int map_colors, int map_entry_size)
+{
+  JSAMPARRAY colormap = cinfo->colormap;
+  int num_colors = cinfo->actual_number_of_colors;
+  FILE *outfile = dest->pub.output_file;
+  int i;
+
+  if (colormap != NULL) {
+    if (cinfo->out_color_components == 3) {
+      /* Normal case with RGB colormap */
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(colormap[2][i]), outfile);
+        putc(GETJSAMPLE(colormap[1][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    } else {
+      /* Grayscale colormap (only happens with grayscale quantization) */
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        putc(GETJSAMPLE(colormap[0][i]), outfile);
+        if (map_entry_size == 4)
+          putc(0, outfile);
+      }
+    }
+  } else {
+    /* If no colormap, must be grayscale data.  Generate a linear "map". */
+    for (i = 0; i < 256; i++) {
+      putc(i, outfile);
+      putc(i, outfile);
+      putc(i, outfile);
+      if (map_entry_size == 4)
+        putc(0, outfile);
+    }
+  }
+  /* Pad colormap with zeros to ensure specified number of colormap entries */
+  if (i > map_colors)
+    ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, i);
+  for (; i < map_colors; i++) {
+    putc(0, outfile);
+    putc(0, outfile);
+    putc(0, outfile);
+    if (map_entry_size == 4)
+      putc(0, outfile);
+  }
+}
+
+
+METHODDEF(void)
+finish_output_bmp (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  bmp_dest_ptr dest = (bmp_dest_ptr) dinfo;
+  register FILE *outfile = dest->pub.output_file;
+  JSAMPARRAY image_ptr;
+  register JSAMPROW data_ptr;
+  JDIMENSION row;
+  register JDIMENSION col;
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+
+  /* Write the header and colormap */
+  if (dest->is_os2)
+    write_os2_header(cinfo, dest);
+  else
+    write_bmp_header(cinfo, dest);
+
+  /* Write the file body from our virtual array */
+  for (row = cinfo->output_height; row > 0; row--) {
+    if (progress != NULL) {
+      progress->pub.pass_counter = (long) (cinfo->output_height - row);
+      progress->pub.pass_limit = (long) cinfo->output_height;
+      (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+    }
+    image_ptr = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->whole_image, row-1, (JDIMENSION) 1, FALSE);
+    data_ptr = image_ptr[0];
+    for (col = dest->row_width; col > 0; col--) {
+      putc(GETJSAMPLE(*data_ptr), outfile);
+      data_ptr++;
+    }
+  }
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+
+  /* Make sure we wrote the output file OK */
+  fflush(outfile);
+  if (ferror(outfile))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for BMP format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_bmp (j_decompress_ptr cinfo, boolean is_os2)
+{
+  bmp_dest_ptr dest;
+  JDIMENSION row_width;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (bmp_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(bmp_dest_struct));
+  dest->pub.start_output = start_output_bmp;
+  dest->pub.finish_output = finish_output_bmp;
+  dest->is_os2 = is_os2;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_gray_rows;
+    else
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else if(cinfo->out_color_space == JCS_RGB565 ) {
+      dest->pub.put_pixel_rows = put_pixel_rows;
+  } else {
+    ERREXIT(cinfo, JERR_BMP_COLORSPACE);
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Determine width of rows in the BMP file (padded to 4-byte boundary). */
+  if (cinfo->out_color_space == JCS_RGB565) {
+    row_width = cinfo->output_width * 2;
+    dest->row_width = dest->data_width = cinfo->output_width * 3;
+  } else {
+    row_width = cinfo->output_width * cinfo->output_components;
+    dest->row_width = dest->data_width = row_width;
+  }
+  while ((dest->row_width & 3) != 0) dest->row_width++;
+  dest->pad_bytes = (int) (dest->row_width - dest->data_width);
+  if (cinfo->out_color_space == JCS_RGB565) {
+    while ((row_width & 3) != 0) row_width++;
+  } else {
+    row_width = dest->row_width;
+  }
+
+
+  /* Allocate space for inversion array, prepare for write pass */
+  dest->whole_image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     dest->row_width, cinfo->output_height, (JDIMENSION) 1);
+  dest->cur_output_row = 0;
+  if (cinfo->progress != NULL) {
+    cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+    progress->total_extra_passes++; /* count file input as separate pass */
+  }
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, row_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* BMP_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrgif.c glmark2-2021.02/src/libjpeg-turbo/wrgif.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrgif.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrgif.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,402 @@
+/*
+ * wrgif.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2015, D. R. Commander.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in GIF format.
+ *
+ **************************************************************************
+ * NOTE: to avoid entanglements with Unisys' patent on LZW compression,   *
+ * this code has been modified to output "uncompressed GIF" files.        *
+ * There is no trace of the LZW algorithm in this file.                   *
+ **************************************************************************
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+/*
+ * This code is loosely based on ppmtogif from the PBMPLUS distribution
+ * of Feb. 1991.  That file contains the following copyright notice:
+ *    Based on GIFENCODE by David Rowley <mgardi@watdscu.waterloo.edu>.
+ *    Lempel-Ziv compression based on "compress" by Spencer W. Thomas et al.
+ *    Copyright (C) 1989 by Jef Poskanzer.
+ *    Permission to use, copy, modify, and distribute this software and its
+ *    documentation for any purpose and without fee is hereby granted, provided
+ *    that the above copyright notice appear in all copies and that both that
+ *    copyright notice and this permission notice appear in supporting
+ *    documentation.  This software is provided "as is" without express or
+ *    implied warranty.
+ *
+ * We are also required to state that
+ *    "The Graphics Interchange Format(c) is the Copyright property of
+ *    CompuServe Incorporated. GIF(sm) is a Service Mark property of
+ *    CompuServe Incorporated."
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef GIF_SUPPORTED
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  j_decompress_ptr cinfo;       /* back link saves passing separate parm */
+
+  /* State for packing variable-width codes into a bitstream */
+  int n_bits;                   /* current number of bits/code */
+  int maxcode;                  /* maximum code, given n_bits */
+  long cur_accum;               /* holds bits not yet output */
+  int cur_bits;                 /* # of bits in cur_accum */
+
+  /* State for GIF code assignment */
+  int ClearCode;                /* clear code (doesn't change) */
+  int EOFCode;                  /* EOF code (ditto) */
+  int code_counter;             /* counts output symbols */
+
+  /* GIF data packet construction buffer */
+  int bytesinpkt;               /* # of bytes in current packet */
+  char packetbuf[256];          /* workspace for accumulating packet */
+
+} gif_dest_struct;
+
+typedef gif_dest_struct *gif_dest_ptr;
+
+/* Largest value that will fit in N bits */
+#define MAXCODE(n_bits) ((1 << (n_bits)) - 1)
+
+
+/*
+ * Routines to package finished data bytes into GIF data blocks.
+ * A data block consists of a count byte (1..255) and that many data bytes.
+ */
+
+LOCAL(void)
+flush_packet (gif_dest_ptr dinfo)
+/* flush any accumulated data */
+{
+  if (dinfo->bytesinpkt > 0) {  /* never write zero-length packet */
+    dinfo->packetbuf[0] = (char) dinfo->bytesinpkt++;
+    if (JFWRITE(dinfo->pub.output_file, dinfo->packetbuf, dinfo->bytesinpkt)
+        != (size_t) dinfo->bytesinpkt)
+      ERREXIT(dinfo->cinfo, JERR_FILE_WRITE);
+    dinfo->bytesinpkt = 0;
+  }
+}
+
+
+/* Add a character to current packet; flush to disk if necessary */
+#define CHAR_OUT(dinfo,c)  \
+        { (dinfo)->packetbuf[++(dinfo)->bytesinpkt] = (char) (c);  \
+            if ((dinfo)->bytesinpkt >= 255)  \
+              flush_packet(dinfo);  \
+        }
+
+
+/* Routine to convert variable-width codes into a byte stream */
+
+LOCAL(void)
+output (gif_dest_ptr dinfo, int code)
+/* Emit a code of n_bits bits */
+/* Uses cur_accum and cur_bits to reblock into 8-bit bytes */
+{
+  dinfo->cur_accum |= ((long) code) << dinfo->cur_bits;
+  dinfo->cur_bits += dinfo->n_bits;
+
+  while (dinfo->cur_bits >= 8) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+    dinfo->cur_accum >>= 8;
+    dinfo->cur_bits -= 8;
+  }
+}
+
+
+/* The pseudo-compression algorithm.
+ *
+ * In this module we simply output each pixel value as a separate symbol;
+ * thus, no compression occurs.  In fact, there is expansion of one bit per
+ * pixel, because we use a symbol width one bit wider than the pixel width.
+ *
+ * GIF ordinarily uses variable-width symbols, and the decoder will expect
+ * to ratchet up the symbol width after a fixed number of symbols.
+ * To simplify the logic and keep the expansion penalty down, we emit a
+ * GIF Clear code to reset the decoder just before the width would ratchet up.
+ * Thus, all the symbols in the output file will have the same bit width.
+ * Note that emitting the Clear codes at the right times is a mere matter of
+ * counting output symbols and is in no way dependent on the LZW patent.
+ *
+ * With a small basic pixel width (low color count), Clear codes will be
+ * needed very frequently, causing the file to expand even more.  So this
+ * simplistic approach wouldn't work too well on bilevel images, for example.
+ * But for output of JPEG conversions the pixel width will usually be 8 bits
+ * (129 to 256 colors), so the overhead added by Clear symbols is only about
+ * one symbol in every 256.
+ */
+
+LOCAL(void)
+compress_init (gif_dest_ptr dinfo, int i_bits)
+/* Initialize pseudo-compressor */
+{
+  /* init all the state variables */
+  dinfo->n_bits = i_bits;
+  dinfo->maxcode = MAXCODE(dinfo->n_bits);
+  dinfo->ClearCode = (1 << (i_bits - 1));
+  dinfo->EOFCode = dinfo->ClearCode + 1;
+  dinfo->code_counter = dinfo->ClearCode + 2;
+  /* init output buffering vars */
+  dinfo->bytesinpkt = 0;
+  dinfo->cur_accum = 0;
+  dinfo->cur_bits = 0;
+  /* GIF specifies an initial Clear code */
+  output(dinfo, dinfo->ClearCode);
+}
+
+
+LOCAL(void)
+compress_pixel (gif_dest_ptr dinfo, int c)
+/* Accept and "compress" one pixel value.
+ * The given value must be less than n_bits wide.
+ */
+{
+  /* Output the given pixel value as a symbol. */
+  output(dinfo, c);
+  /* Issue Clear codes often enough to keep the reader from ratcheting up
+   * its symbol size.
+   */
+  if (dinfo->code_counter < dinfo->maxcode) {
+    dinfo->code_counter++;
+  } else {
+    output(dinfo, dinfo->ClearCode);
+    dinfo->code_counter = dinfo->ClearCode + 2; /* reset the counter */
+  }
+}
+
+
+LOCAL(void)
+compress_term (gif_dest_ptr dinfo)
+/* Clean up at end */
+{
+  /* Send an EOF code */
+  output(dinfo, dinfo->EOFCode);
+  /* Flush the bit-packing buffer */
+  if (dinfo->cur_bits > 0) {
+    CHAR_OUT(dinfo, dinfo->cur_accum & 0xFF);
+  }
+  /* Flush the packet buffer */
+  flush_packet(dinfo);
+}
+
+
+/* GIF header construction */
+
+
+LOCAL(void)
+put_word (gif_dest_ptr dinfo, unsigned int w)
+/* Emit a 16-bit word, LSB first */
+{
+  putc(w & 0xFF, dinfo->pub.output_file);
+  putc((w >> 8) & 0xFF, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+put_3bytes (gif_dest_ptr dinfo, int val)
+/* Emit 3 copies of same byte value --- handy subr for colormap construction */
+{
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+  putc(val, dinfo->pub.output_file);
+}
+
+
+LOCAL(void)
+emit_header (gif_dest_ptr dinfo, int num_colors, JSAMPARRAY colormap)
+/* Output the GIF file header, including color map */
+/* If colormap==NULL, synthesize a grayscale colormap */
+{
+  int BitsPerPixel, ColorMapSize, InitCodeSize, FlagByte;
+  int cshift = dinfo->cinfo->data_precision - 8;
+  int i;
+
+  if (num_colors > 256)
+    ERREXIT1(dinfo->cinfo, JERR_TOO_MANY_COLORS, num_colors);
+  /* Compute bits/pixel and related values */
+  BitsPerPixel = 1;
+  while (num_colors > (1 << BitsPerPixel))
+    BitsPerPixel++;
+  ColorMapSize = 1 << BitsPerPixel;
+  if (BitsPerPixel <= 1)
+    InitCodeSize = 2;
+  else
+    InitCodeSize = BitsPerPixel;
+  /*
+   * Write the GIF header.
+   * Note that we generate a plain GIF87 header for maximum compatibility.
+   */
+  putc('G', dinfo->pub.output_file);
+  putc('I', dinfo->pub.output_file);
+  putc('F', dinfo->pub.output_file);
+  putc('8', dinfo->pub.output_file);
+  putc('7', dinfo->pub.output_file);
+  putc('a', dinfo->pub.output_file);
+  /* Write the Logical Screen Descriptor */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  FlagByte = 0x80;              /* Yes, there is a global color table */
+  FlagByte |= (BitsPerPixel-1) << 4; /* color resolution */
+  FlagByte |= (BitsPerPixel-1); /* size of global color table */
+  putc(FlagByte, dinfo->pub.output_file);
+  putc(0, dinfo->pub.output_file); /* Background color index */
+  putc(0, dinfo->pub.output_file); /* Reserved (aspect ratio in GIF89) */
+  /* Write the Global Color Map */
+  /* If the color map is more than 8 bits precision, */
+  /* we reduce it to 8 bits by shifting */
+  for (i=0; i < ColorMapSize; i++) {
+    if (i < num_colors) {
+      if (colormap != NULL) {
+        if (dinfo->cinfo->out_color_space == JCS_RGB) {
+          /* Normal case: RGB color map */
+          putc(GETJSAMPLE(colormap[0][i]) >> cshift, dinfo->pub.output_file);
+          putc(GETJSAMPLE(colormap[1][i]) >> cshift, dinfo->pub.output_file);
+          putc(GETJSAMPLE(colormap[2][i]) >> cshift, dinfo->pub.output_file);
+        } else {
+          /* Grayscale "color map": possible if quantizing grayscale image */
+          put_3bytes(dinfo, GETJSAMPLE(colormap[0][i]) >> cshift);
+        }
+      } else {
+        /* Create a grayscale map of num_colors values, range 0..255 */
+        put_3bytes(dinfo, (i * 255 + (num_colors-1)/2) / (num_colors-1));
+      }
+    } else {
+      /* fill out the map to a power of 2 */
+      put_3bytes(dinfo, 0);
+    }
+  }
+  /* Write image separator and Image Descriptor */
+  putc(',', dinfo->pub.output_file); /* separator */
+  put_word(dinfo, 0);           /* left/top offset */
+  put_word(dinfo, 0);
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_width); /* image size */
+  put_word(dinfo, (unsigned int) dinfo->cinfo->output_height);
+  /* flag byte: not interlaced, no local color map */
+  putc(0x00, dinfo->pub.output_file);
+  /* Write Initial Code Size byte */
+  putc(InitCodeSize, dinfo->pub.output_file);
+
+  /* Initialize for "compression" of image data */
+  compress_init(dinfo, InitCodeSize+1);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  if (cinfo->quantize_colors)
+    emit_header(dest, cinfo->actual_number_of_colors, cinfo->colormap);
+  else
+    emit_header(dest, 256, (JSAMPARRAY) NULL);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  for (col = cinfo->output_width; col > 0; col--) {
+    compress_pixel(dest, GETJSAMPLE(*ptr++));
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_gif (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  gif_dest_ptr dest = (gif_dest_ptr) dinfo;
+
+  /* Flush "compression" mechanism */
+  compress_term(dest);
+  /* Write a zero-length data block to end the series */
+  putc(0, dest->pub.output_file);
+  /* Write the GIF terminator mark */
+  putc(';', dest->pub.output_file);
+  /* Make sure we wrote the output file OK */
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for GIF format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_gif (j_decompress_ptr cinfo)
+{
+  gif_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (gif_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(gif_dest_struct));
+  dest->cinfo = cinfo;          /* make back link for subroutines */
+  dest->pub.start_output = start_output_gif;
+  dest->pub.put_pixel_rows = put_pixel_rows;
+  dest->pub.finish_output = finish_output_gif;
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_GIF_COLORSPACE);
+
+  /* Force quantization if color or if > 8 bits input */
+  if (cinfo->out_color_space != JCS_GRAYSCALE || cinfo->data_precision > 8) {
+    /* Force quantization to at most 256 colors */
+    cinfo->quantize_colors = TRUE;
+    if (cinfo->desired_number_of_colors > 256)
+      cinfo->desired_number_of_colors = 256;
+  }
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  if (cinfo->output_components != 1) /* safety check: just one component? */
+    ERREXIT(cinfo, JERR_GIF_BUG);
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, cinfo->output_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* GIF_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrjpgcom.1 glmark2-2021.02/src/libjpeg-turbo/wrjpgcom.1
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrjpgcom.1	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrjpgcom.1	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,103 @@
+.TH WRJPGCOM 1 "15 June 1995"
+.SH NAME
+wrjpgcom \- insert text comments into a JPEG file
+.SH SYNOPSIS
+.B wrjpgcom
+[
+.B \-replace
+]
+[
+.BI \-comment " text"
+]
+[
+.BI \-cfile " name"
+]
+[
+.I filename
+]
+.LP
+.SH DESCRIPTION
+.LP
+.B wrjpgcom
+reads the named JPEG/JFIF file, or the standard input if no file is named,
+and generates a new JPEG/JFIF file on standard output.  A comment block is
+added to the file.
+.PP
+The JPEG standard allows "comment" (COM) blocks to occur within a JPEG file.
+Although the standard doesn't actually define what COM blocks are for, they
+are widely used to hold user-supplied text strings.  This lets you add
+annotations, titles, index terms, etc to your JPEG files, and later retrieve
+them as text.  COM blocks do not interfere with the image stored in the JPEG
+file.  The maximum size of a COM block is 64K, but you can have as many of
+them as you like in one JPEG file.
+.PP
+.B wrjpgcom
+adds a COM block, containing text you provide, to a JPEG file.
+Ordinarily, the COM block is added after any existing COM blocks; but you
+can delete the old COM blocks if you wish.
+.SH OPTIONS
+Switch names may be abbreviated, and are not case sensitive.
+.TP
+.B \-replace
+Delete any existing COM blocks from the file.
+.TP
+.BI \-comment " text"
+Supply text for new COM block on command line.
+.TP
+.BI \-cfile " name"
+Read text for new COM block from named file.
+.PP
+If you have only one line of comment text to add, you can provide it on the
+command line with
+.BR \-comment .
+The comment text must be surrounded with quotes so that it is treated as a
+single argument.  Longer comments can be read from a text file.
+.PP
+If you give neither
+.B \-comment
+nor
+.BR \-cfile ,
+then
+.B wrjpgcom
+will read the comment text from standard input.  (In this case an input image
+file name MUST be supplied, so that the source JPEG file comes from somewhere
+else.)  You can enter multiple lines, up to 64KB worth.  Type an end-of-file
+indicator (usually control-D) to terminate the comment text entry.
+.PP
+.B wrjpgcom
+will not add a COM block if the provided comment string is empty.  Therefore
+\fB\-replace \-comment ""\fR can be used to delete all COM blocks from a file.
+.SH EXAMPLES
+.LP
+Add a short comment to in.jpg, producing out.jpg:
+.IP
+.B wrjpgcom \-c
+\fI"View of my back yard" in.jpg
+.B >
+.I out.jpg
+.PP
+Attach a long comment previously stored in comment.txt:
+.IP
+.B wrjpgcom
+.I in.jpg
+.B <
+.I comment.txt
+.B >
+.I out.jpg
+.PP
+or equivalently
+.IP
+.B wrjpgcom
+.B -cfile
+.I comment.txt
+.B <
+.I in.jpg
+.B >
+.I out.jpg
+.SH SEE ALSO
+.BR cjpeg (1),
+.BR djpeg (1),
+.BR jpegtran (1),
+.BR rdjpgcom (1)
+.SH AUTHOR
+Independent JPEG Group
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrjpgcom.c glmark2-2021.02/src/libjpeg-turbo/wrjpgcom.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrjpgcom.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrjpgcom.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,592 @@
+/*
+ * wrjpgcom.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994-1997, Thomas G. Lane.
+ * libjpeg-turbo Modifications:
+ * Copyright (C) 2014, D. R. Commander
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains a very simple stand-alone application that inserts
+ * user-supplied text as a COM (comment) marker in a JFIF file.
+ * This may be useful as an example of the minimum logic needed to parse
+ * JPEG markers.
+ */
+
+#define JPEG_CJPEG_DJPEG        /* to get the command-line config symbols */
+#include "jinclude.h"           /* get auto-config symbols, <stdio.h> */
+
+#ifndef HAVE_STDLIB_H           /* <stdlib.h> should declare malloc() */
+extern void *malloc ();
+#endif
+#include <ctype.h>              /* to declare isupper(), tolower() */
+#ifdef USE_SETMODE
+#include <fcntl.h>              /* to declare setmode()'s parameter macros */
+/* If you have setmode() but not <io.h>, just delete this line: */
+#include <io.h>                 /* to declare setmode() */
+#endif
+
+#ifdef USE_CCOMMAND             /* command-line reader for Macintosh */
+#ifdef __MWERKS__
+#include <SIOUX.h>              /* Metrowerks needs this */
+#include <console.h>            /* ... and this */
+#endif
+#ifdef THINK_C
+#include <console.h>            /* Think declares it here */
+#endif
+#endif
+
+#ifdef DONT_USE_B_MODE          /* define mode parameters for fopen() */
+#define READ_BINARY     "r"
+#define WRITE_BINARY    "w"
+#else
+#define READ_BINARY     "rb"
+#define WRITE_BINARY    "wb"
+#endif
+
+#ifndef EXIT_FAILURE            /* define exit() codes if not provided */
+#define EXIT_FAILURE  1
+#endif
+#ifndef EXIT_SUCCESS
+#define EXIT_SUCCESS  0
+#endif
+
+/* Reduce this value if your malloc() can't allocate blocks up to 64K.
+ * On DOS, compiling in large model is usually a better solution.
+ */
+
+#ifndef MAX_COM_LENGTH
+#define MAX_COM_LENGTH 65000L   /* must be <= 65533 in any case */
+#endif
+
+
+/*
+ * These macros are used to read the input file and write the output file.
+ * To reuse this code in another application, you might need to change these.
+ */
+
+static FILE *infile;            /* input JPEG file */
+
+/* Return next input byte, or EOF if no more */
+#define NEXTBYTE()  getc(infile)
+
+static FILE *outfile;           /* output JPEG file */
+
+/* Emit an output byte */
+#define PUTBYTE(x)  putc((x), outfile)
+
+
+/* Error exit handler */
+#define ERREXIT(msg)  (fprintf(stderr, "%s\n", msg), exit(EXIT_FAILURE))
+
+
+/* Read one byte, testing for EOF */
+static int
+read_1_byte (void)
+{
+  int c;
+
+  c = NEXTBYTE();
+  if (c == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return c;
+}
+
+/* Read 2 bytes, convert to unsigned int */
+/* All 2-byte quantities in JPEG markers are MSB first */
+static unsigned int
+read_2_bytes (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  if (c1 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  c2 = NEXTBYTE();
+  if (c2 == EOF)
+    ERREXIT("Premature EOF in JPEG file");
+  return (((unsigned int) c1) << 8) + ((unsigned int) c2);
+}
+
+
+/* Routines to write data to output file */
+
+static void
+write_1_byte (int c)
+{
+  PUTBYTE(c);
+}
+
+static void
+write_2_bytes (unsigned int val)
+{
+  PUTBYTE((val >> 8) & 0xFF);
+  PUTBYTE(val & 0xFF);
+}
+
+static void
+write_marker (int marker)
+{
+  PUTBYTE(0xFF);
+  PUTBYTE(marker);
+}
+
+static void
+copy_rest_of_file (void)
+{
+  int c;
+
+  while ((c = NEXTBYTE()) != EOF)
+    PUTBYTE(c);
+}
+
+
+/*
+ * JPEG markers consist of one or more 0xFF bytes, followed by a marker
+ * code byte (which is not an FF).  Here are the marker codes of interest
+ * in this program.  (See jdmarker.c for a more complete list.)
+ */
+
+#define M_SOF0  0xC0            /* Start Of Frame N */
+#define M_SOF1  0xC1            /* N indicates which compression process */
+#define M_SOF2  0xC2            /* Only SOF0-SOF2 are now in common use */
+#define M_SOF3  0xC3
+#define M_SOF5  0xC5            /* NB: codes C4 and CC are NOT SOF markers */
+#define M_SOF6  0xC6
+#define M_SOF7  0xC7
+#define M_SOF9  0xC9
+#define M_SOF10 0xCA
+#define M_SOF11 0xCB
+#define M_SOF13 0xCD
+#define M_SOF14 0xCE
+#define M_SOF15 0xCF
+#define M_SOI   0xD8            /* Start Of Image (beginning of datastream) */
+#define M_EOI   0xD9            /* End Of Image (end of datastream) */
+#define M_SOS   0xDA            /* Start Of Scan (begins compressed data) */
+#define M_COM   0xFE            /* COMment */
+
+
+/*
+ * Find the next JPEG marker and return its marker code.
+ * We expect at least one FF byte, possibly more if the compressor used FFs
+ * to pad the file.  (Padding FFs will NOT be replicated in the output file.)
+ * There could also be non-FF garbage between markers.  The treatment of such
+ * garbage is unspecified; we choose to skip over it but emit a warning msg.
+ * NB: this routine must not be used after seeing SOS marker, since it will
+ * not deal correctly with FF/00 sequences in the compressed image data...
+ */
+
+static int
+next_marker (void)
+{
+  int c;
+  int discarded_bytes = 0;
+
+  /* Find 0xFF byte; count and skip any non-FFs. */
+  c = read_1_byte();
+  while (c != 0xFF) {
+    discarded_bytes++;
+    c = read_1_byte();
+  }
+  /* Get marker code byte, swallowing any duplicate FF bytes.  Extra FFs
+   * are legal as pad bytes, so don't count them in discarded_bytes.
+   */
+  do {
+    c = read_1_byte();
+  } while (c == 0xFF);
+
+  if (discarded_bytes != 0) {
+    fprintf(stderr, "Warning: garbage data found in JPEG file\n");
+  }
+
+  return c;
+}
+
+
+/*
+ * Read the initial marker, which should be SOI.
+ * For a JFIF file, the first two bytes of the file should be literally
+ * 0xFF M_SOI.  To be more general, we could use next_marker, but if the
+ * input file weren't actually JPEG at all, next_marker might read the whole
+ * file and then return a misleading error message...
+ */
+
+static int
+first_marker (void)
+{
+  int c1, c2;
+
+  c1 = NEXTBYTE();
+  c2 = NEXTBYTE();
+  if (c1 != 0xFF || c2 != M_SOI)
+    ERREXIT("Not a JPEG file");
+  return c2;
+}
+
+
+/*
+ * Most types of marker are followed by a variable-length parameter segment.
+ * This routine skips over the parameters for any marker we don't otherwise
+ * want to process.
+ * Note that we MUST skip the parameter segment explicitly in order not to
+ * be fooled by 0xFF bytes that might appear within the parameter segment;
+ * such bytes do NOT introduce new markers.
+ */
+
+static void
+copy_variable (void)
+/* Copy an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  write_2_bytes(length);
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    write_1_byte(read_1_byte());
+    length--;
+  }
+}
+
+static void
+skip_variable (void)
+/* Skip over an unknown or uninteresting variable-length marker */
+{
+  unsigned int length;
+
+  /* Get the marker parameter length count */
+  length = read_2_bytes();
+  /* Length includes itself, so must be at least 2 */
+  if (length < 2)
+    ERREXIT("Erroneous JPEG marker length");
+  length -= 2;
+  /* Skip over the remaining bytes */
+  while (length > 0) {
+    (void) read_1_byte();
+    length--;
+  }
+}
+
+
+/*
+ * Parse the marker stream until SOFn or EOI is seen;
+ * copy data to output, but discard COM markers unless keep_COM is true.
+ */
+
+static int
+scan_JPEG_header (int keep_COM)
+{
+  int marker;
+
+  /* Expect SOI at start of file */
+  if (first_marker() != M_SOI)
+    ERREXIT("Expected SOI marker first");
+  write_marker(M_SOI);
+
+  /* Scan miscellaneous markers until we reach SOFn. */
+  for (;;) {
+    marker = next_marker();
+    switch (marker) {
+      /* Note that marker codes 0xC4, 0xC8, 0xCC are not, and must not be,
+       * treated as SOFn.  C4 in particular is actually DHT.
+       */
+    case M_SOF0:                /* Baseline */
+    case M_SOF1:                /* Extended sequential, Huffman */
+    case M_SOF2:                /* Progressive, Huffman */
+    case M_SOF3:                /* Lossless, Huffman */
+    case M_SOF5:                /* Differential sequential, Huffman */
+    case M_SOF6:                /* Differential progressive, Huffman */
+    case M_SOF7:                /* Differential lossless, Huffman */
+    case M_SOF9:                /* Extended sequential, arithmetic */
+    case M_SOF10:               /* Progressive, arithmetic */
+    case M_SOF11:               /* Lossless, arithmetic */
+    case M_SOF13:               /* Differential sequential, arithmetic */
+    case M_SOF14:               /* Differential progressive, arithmetic */
+    case M_SOF15:               /* Differential lossless, arithmetic */
+      return marker;
+
+    case M_SOS:                 /* should not see compressed data before SOF */
+      ERREXIT("SOS without prior SOFn");
+      break;
+
+    case M_EOI:                 /* in case it's a tables-only JPEG stream */
+      return marker;
+
+    case M_COM:                 /* Existing COM: conditionally discard */
+      if (keep_COM) {
+        write_marker(marker);
+        copy_variable();
+      } else {
+        skip_variable();
+      }
+      break;
+
+    default:                    /* Anything else just gets copied */
+      write_marker(marker);
+      copy_variable();          /* we assume it has a parameter count... */
+      break;
+    }
+  } /* end loop */
+}
+
+
+/* Command line parsing code */
+
+static const char *progname;    /* program name for error messages */
+
+
+static void
+usage (void)
+/* complain about bad command line */
+{
+  fprintf(stderr, "wrjpgcom inserts a textual comment in a JPEG file.\n");
+  fprintf(stderr, "You can add to or replace any existing comment(s).\n");
+
+  fprintf(stderr, "Usage: %s [switches] ", progname);
+#ifdef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "inputfile outputfile\n");
+#else
+  fprintf(stderr, "[inputfile]\n");
+#endif
+
+  fprintf(stderr, "Switches (names may be abbreviated):\n");
+  fprintf(stderr, "  -replace         Delete any existing comments\n");
+  fprintf(stderr, "  -comment \"text\"  Insert comment with given text\n");
+  fprintf(stderr, "  -cfile name      Read comment from named file\n");
+  fprintf(stderr, "Notice that you must put quotes around the comment text\n");
+  fprintf(stderr, "when you use -comment.\n");
+  fprintf(stderr, "If you do not give either -comment or -cfile on the command line,\n");
+  fprintf(stderr, "then the comment text is read from standard input.\n");
+  fprintf(stderr, "It can be multiple lines, up to %u characters total.\n",
+          (unsigned int) MAX_COM_LENGTH);
+#ifndef TWO_FILE_COMMANDLINE
+  fprintf(stderr, "You must specify an input JPEG file name when supplying\n");
+  fprintf(stderr, "comment text from standard input.\n");
+#endif
+
+  exit(EXIT_FAILURE);
+}
+
+
+static int
+keymatch (char *arg, const char *keyword, int minchars)
+/* Case-insensitive matching of (possibly abbreviated) keyword switches. */
+/* keyword is the constant keyword (must be lower case already), */
+/* minchars is length of minimum legal abbreviation. */
+{
+  register int ca, ck;
+  register int nmatched = 0;
+
+  while ((ca = *arg++) != '\0') {
+    if ((ck = *keyword++) == '\0')
+      return 0;                 /* arg longer than keyword, no good */
+    if (isupper(ca))            /* force arg to lcase (assume ck is already) */
+      ca = tolower(ca);
+    if (ca != ck)
+      return 0;                 /* no good */
+    nmatched++;                 /* count matched characters */
+  }
+  /* reached end of argument; fail if it's too short for unique abbrev */
+  if (nmatched < minchars)
+    return 0;
+  return 1;                     /* A-OK */
+}
+
+
+/*
+ * The main program.
+ */
+
+int
+main (int argc, char **argv)
+{
+  int argn;
+  char *arg;
+  int keep_COM = 1;
+  char *comment_arg = NULL;
+  FILE *comment_file = NULL;
+  unsigned int comment_length = 0;
+  int marker;
+
+  /* On Mac, fetch a command line. */
+#ifdef USE_CCOMMAND
+  argc = ccommand(&argv);
+#endif
+
+  progname = argv[0];
+  if (progname == NULL || progname[0] == 0)
+    progname = "wrjpgcom";      /* in case C library doesn't provide it */
+
+  /* Parse switches, if any */
+  for (argn = 1; argn < argc; argn++) {
+    arg = argv[argn];
+    if (arg[0] != '-')
+      break;                    /* not switch, must be file name */
+    arg++;                      /* advance over '-' */
+    if (keymatch(arg, "replace", 1)) {
+      keep_COM = 0;
+    } else if (keymatch(arg, "cfile", 2)) {
+      if (++argn >= argc) usage();
+      if ((comment_file = fopen(argv[argn], "r")) == NULL) {
+        fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+        exit(EXIT_FAILURE);
+      }
+    } else if (keymatch(arg, "comment", 1)) {
+      if (++argn >= argc) usage();
+      comment_arg = argv[argn];
+      /* If the comment text starts with '"', then we are probably running
+       * under MS-DOG and must parse out the quoted string ourselves.  Sigh.
+       */
+      if (comment_arg[0] == '"') {
+        comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+        if (comment_arg == NULL)
+          ERREXIT("Insufficient memory");
+        if (strlen(argv[argn]) + 2 >= (size_t) MAX_COM_LENGTH) {
+          fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                  (unsigned int) MAX_COM_LENGTH);
+          exit(EXIT_FAILURE);
+        }
+        strcpy(comment_arg, argv[argn]+1);
+        for (;;) {
+          comment_length = (unsigned int) strlen(comment_arg);
+          if (comment_length > 0 && comment_arg[comment_length-1] == '"') {
+            comment_arg[comment_length-1] = '\0'; /* zap terminating quote */
+            break;
+          }
+          if (++argn >= argc)
+            ERREXIT("Missing ending quote mark");
+          if (strlen(comment_arg) + strlen(argv[argn]) + 2 >=
+              (size_t) MAX_COM_LENGTH) {
+            fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                    (unsigned int) MAX_COM_LENGTH);
+            exit(EXIT_FAILURE);
+          }
+          strcat(comment_arg, " ");
+          strcat(comment_arg, argv[argn]);
+        }
+      } else if (strlen(argv[argn]) >= (size_t) MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int) MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_length = (unsigned int) strlen(comment_arg);
+    } else
+      usage();
+  }
+
+  /* Cannot use both -comment and -cfile. */
+  if (comment_arg != NULL && comment_file != NULL)
+    usage();
+  /* If there is neither -comment nor -cfile, we will read the comment text
+   * from stdin; in this case there MUST be an input JPEG file name.
+   */
+  if (comment_arg == NULL && comment_file == NULL && argn >= argc)
+    usage();
+
+  /* Open the input file. */
+  if (argn < argc) {
+    if ((infile = fopen(argv[argn], READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open %s\n", progname, argv[argn]);
+      exit(EXIT_FAILURE);
+    }
+  } else {
+    /* default input file is stdin */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+    setmode(fileno(stdin), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+    if ((infile = fdopen(fileno(stdin), READ_BINARY)) == NULL) {
+      fprintf(stderr, "%s: can't open stdin\n", progname);
+      exit(EXIT_FAILURE);
+    }
+#else
+    infile = stdin;
+#endif
+  }
+
+  /* Open the output file. */
+#ifdef TWO_FILE_COMMANDLINE
+  /* Must have explicit output file name */
+  if (argn != argc-2) {
+    fprintf(stderr, "%s: must name one input and one output file\n",
+            progname);
+    usage();
+  }
+  if ((outfile = fopen(argv[argn+1], WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open %s\n", progname, argv[argn+1]);
+    exit(EXIT_FAILURE);
+  }
+#else
+  /* Unix style: expect zero or one file name */
+  if (argn < argc-1) {
+    fprintf(stderr, "%s: only one input file\n", progname);
+    usage();
+  }
+  /* default output file is stdout */
+#ifdef USE_SETMODE              /* need to hack file mode? */
+  setmode(fileno(stdout), O_BINARY);
+#endif
+#ifdef USE_FDOPEN               /* need to re-open in binary mode? */
+  if ((outfile = fdopen(fileno(stdout), WRITE_BINARY)) == NULL) {
+    fprintf(stderr, "%s: can't open stdout\n", progname);
+    exit(EXIT_FAILURE);
+  }
+#else
+  outfile = stdout;
+#endif
+#endif /* TWO_FILE_COMMANDLINE */
+
+  /* Collect comment text from comment_file or stdin, if necessary */
+  if (comment_arg == NULL) {
+    FILE *src_file;
+    int c;
+
+    comment_arg = (char *) malloc((size_t) MAX_COM_LENGTH);
+    if (comment_arg == NULL)
+      ERREXIT("Insufficient memory");
+    comment_length = 0;
+    src_file = (comment_file != NULL ? comment_file : stdin);
+    while ((c = getc(src_file)) != EOF) {
+      if (comment_length >= (unsigned int) MAX_COM_LENGTH) {
+        fprintf(stderr, "Comment text may not exceed %u bytes\n",
+                (unsigned int) MAX_COM_LENGTH);
+        exit(EXIT_FAILURE);
+      }
+      comment_arg[comment_length++] = (char) c;
+    }
+    if (comment_file != NULL)
+      fclose(comment_file);
+  }
+
+  /* Copy JPEG headers until SOFn marker;
+   * we will insert the new comment marker just before SOFn.
+   * This (a) causes the new comment to appear after, rather than before,
+   * existing comments; and (b) ensures that comments come after any JFIF
+   * or JFXX markers, as required by the JFIF specification.
+   */
+  marker = scan_JPEG_header(keep_COM);
+  /* Insert the new COM marker, but only if nonempty text has been supplied */
+  if (comment_length > 0) {
+    write_marker(M_COM);
+    write_2_bytes(comment_length + 2);
+    while (comment_length > 0) {
+      write_1_byte(*comment_arg++);
+      comment_length--;
+    }
+  }
+  /* Duplicate the remainder of the source file.
+   * Note that any COM markers occuring after SOF will not be touched.
+   */
+  write_marker(marker);
+  copy_rest_of_file();
+
+  /* All done. */
+  exit(EXIT_SUCCESS);
+  return 0;                     /* suppress no-return-value warnings */
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrppm.c glmark2-2021.02/src/libjpeg-turbo/wrppm.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrppm.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrppm.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,252 @@
+/*
+ * wrppm.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * Modified 2009 by Guido Vollbeding.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in PPM/PGM format.
+ * The extended 2-byte-per-sample raw PPM/PGM formats are supported.
+ * The PBMPLUS library is NOT required to compile this software
+ * (but it is highly useful as a set of PPM image manipulation programs).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+#include "wrppm.h"
+
+#ifdef PPM_SUPPORTED
+
+
+/*
+ * For 12-bit JPEG data, we either downscale the values to 8 bits
+ * (to write standard byte-per-sample PPM/PGM files), or output
+ * nonstandard word-per-sample PPM/PGM files.  Downscaling is done
+ * if PPM_NORAWWORD is defined (this can be done in the Makefile
+ * or in jconfig.h).
+ * (When the core library supports data precision reduction, a cleaner
+ * implementation will be to ask for that instead.)
+ */
+
+#if BITS_IN_JSAMPLE == 8
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) (v)
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+#ifdef PPM_NORAWWORD
+#define PUTPPMSAMPLE(ptr,v)  *ptr++ = (char) ((v) >> (BITS_IN_JSAMPLE-8))
+#define BYTESPERSAMPLE 1
+#define PPM_MAXVAL 255
+#else
+/* The word-per-sample format always puts the MSB first. */
+#define PUTPPMSAMPLE(ptr,v)                     \
+        { register int val_ = v;                \
+          *ptr++ = (char) ((val_ >> 8) & 0xFF); \
+          *ptr++ = (char) (val_ & 0xFF);        \
+        }
+#define BYTESPERSAMPLE 2
+#define PPM_MAXVAL ((1<<BITS_IN_JSAMPLE)-1)
+#endif
+#endif
+
+
+/*
+ * When JSAMPLE is the same size as char, we can just fwrite() the
+ * decompressed data to the PPM or PGM file.
+ */
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ *
+ * put_pixel_rows handles the "normal" 8-bit case where the decompressor
+ * output buffer is physically the same as the fwrite buffer.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * This code is used when we have to copy the data and apply a pixel
+ * format translation.  Typically this only happens in 12-bit mode.
+ */
+
+METHODDEF(void)
+copy_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                 JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = dest->samples_per_row; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(*ptr++));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some pixel data when color quantization is in effect.
+ * We have to demap the color index values to straight data.
+ */
+
+METHODDEF(void)
+put_demapped_rgb (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                  JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register int pixval;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JSAMPROW color_map1 = cinfo->colormap[1];
+  register JSAMPROW color_map2 = cinfo->colormap[2];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    pixval = GETJSAMPLE(*ptr++);
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map0[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map1[pixval]));
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map2[pixval]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+  register char *bufferptr;
+  register JSAMPROW ptr;
+  register JSAMPROW color_map = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  ptr = dest->pub.buffer[0];
+  bufferptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    PUTPPMSAMPLE(bufferptr, GETJSAMPLE(color_map[GETJSAMPLE(*ptr++)]));
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  ppm_dest_ptr dest = (ppm_dest_ptr) dinfo;
+
+  /* Emit file header */
+  switch (cinfo->out_color_space) {
+  case JCS_GRAYSCALE:
+    /* emit header for raw PGM format */
+    fprintf(dest->pub.output_file, "P5\n%ld %ld\n%d\n",
+            (long) cinfo->output_width, (long) cinfo->output_height,
+            PPM_MAXVAL);
+    break;
+  case JCS_RGB:
+    /* emit header for raw PPM format */
+    fprintf(dest->pub.output_file, "P6\n%ld %ld\n%d\n",
+            (long) cinfo->output_width, (long) cinfo->output_height,
+            PPM_MAXVAL);
+    break;
+  default:
+    ERREXIT(cinfo, JERR_PPM_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_ppm (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for PPM format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_ppm (j_decompress_ptr cinfo)
+{
+  ppm_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (ppm_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(ppm_dest_struct));
+  dest->pub.start_output = start_output_ppm;
+  dest->pub.finish_output = finish_output_ppm;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create physical I/O buffer */
+  dest->samples_per_row = cinfo->output_width * cinfo->out_color_components;
+  dest->buffer_width = dest->samples_per_row * (BYTESPERSAMPLE * sizeof(char));
+  dest->iobuffer = (char *) (*cinfo->mem->alloc_small)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width);
+
+  if (cinfo->quantize_colors || BITS_IN_JSAMPLE != 8 ||
+      sizeof(JSAMPLE) != sizeof(char)) {
+    /* When quantizing, we need an output buffer for colormap indexes
+     * that's separate from the physical I/O buffer.  We also need a
+     * separate buffer if pixel format translation must take place.
+     */
+    dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE,
+       cinfo->output_width * cinfo->output_components, (JDIMENSION) 1);
+    dest->pub.buffer_height = 1;
+    if (! cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = copy_pixel_rows;
+    else if (cinfo->out_color_space == JCS_GRAYSCALE)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_demapped_rgb;
+  } else {
+    /* We will fwrite() directly from decompressor output buffer. */
+    /* Synthesize a JSAMPARRAY pointer structure */
+    dest->pixrow = (JSAMPROW) dest->iobuffer;
+    dest->pub.buffer = & dest->pixrow;
+    dest->pub.buffer_height = 1;
+    dest->pub.put_pixel_rows = put_pixel_rows;
+  }
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* PPM_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrppm.h glmark2-2021.02/src/libjpeg-turbo/wrppm.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrppm.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrppm.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,26 @@
+/*
+ * wrppm.h
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1994, Thomas G. Lane.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ */
+
+#ifdef PPM_SUPPORTED
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  /* Usually these two pointers point to the same place: */
+  char *iobuffer;               /* fwrite's I/O buffer */
+  JSAMPROW pixrow;              /* decompressor output buffer */
+  size_t buffer_width;          /* width of I/O buffer */
+  JDIMENSION samples_per_row;   /* JSAMPLEs per output row */
+} ppm_dest_struct;
+
+typedef ppm_dest_struct *ppm_dest_ptr;
+
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrrle.c glmark2-2021.02/src/libjpeg-turbo/wrrle.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrrle.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrrle.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,307 @@
+/*
+ * wrrle.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in RLE format.
+ * The Utah Raster Toolkit library is required (version 3.1 or later).
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Mike Lijewski,
+ * with updates from Robert Hutchinson.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef RLE_SUPPORTED
+
+/* rle.h is provided by the Utah Raster Toolkit. */
+
+#include <rle.h>
+
+/*
+ * We assume that JSAMPLE has the same representation as rle_pixel,
+ * to wit, "unsigned char".  Hence we can't cope with 12- or 16-bit samples.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/*
+ * Since RLE stores scanlines bottom-to-top, we have to invert the image
+ * from JPEG's top-to-bottom order.  To do this, we save the outgoing data
+ * in a virtual array during put_pixel_row calls, then actually emit the
+ * RLE file during finish_output.
+ */
+
+
+/*
+ * For now, if we emit an RLE color map then it is always 256 entries long,
+ * though not all of the entries need be used.
+ */
+
+#define CMAPBITS        8
+#define CMAPLENGTH      (1<<(CMAPBITS))
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  jvirt_sarray_ptr image;       /* virtual array to store the output image */
+  rle_map *colormap;            /* RLE-style color map, or NULL if none */
+  rle_pixel **rle_row;          /* To pass rows to rle_putrow() */
+
+} rle_dest_struct;
+
+typedef rle_dest_struct *rle_dest_ptr;
+
+/* Forward declarations */
+METHODDEF(void) rle_put_pixel_rows
+        (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+         JDIMENSION rows_supplied);
+
+
+/*
+ * Write the file header.
+ *
+ * In this module it's easier to wait till finish_output to write anything.
+ */
+
+METHODDEF(void)
+start_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  size_t cmapsize;
+  int i, ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /*
+   * Make sure the image can be stored in RLE format.
+   *
+   * - RLE stores image dimensions as *signed* 16 bit integers.  JPEG
+   *   uses unsigned, so we have to check the width.
+   *
+   * - Colorspace is expected to be grayscale or RGB.
+   *
+   * - The number of channels (components) is expected to be 1 (grayscale/
+   *   pseudocolor) or 3 (truecolor/directcolor).
+   *   (could be 2 or 4 if using an alpha channel, but we aren't)
+   */
+
+  if (cinfo->output_width > 32767 || cinfo->output_height > 32767)
+    ERREXIT2(cinfo, JERR_RLE_DIMENSIONS, cinfo->output_width,
+             cinfo->output_height);
+
+  if (cinfo->out_color_space != JCS_GRAYSCALE &&
+      cinfo->out_color_space != JCS_RGB)
+    ERREXIT(cinfo, JERR_RLE_COLORSPACE);
+
+  if (cinfo->output_components != 1 && cinfo->output_components != 3)
+    ERREXIT1(cinfo, JERR_RLE_TOOMANYCHANNELS, cinfo->num_components);
+
+  /* Convert colormap, if any, to RLE format. */
+
+  dest->colormap = NULL;
+
+  if (cinfo->quantize_colors) {
+    /* Allocate storage for RLE-style cmap, zero any extra entries */
+    cmapsize = cinfo->out_color_components * CMAPLENGTH * sizeof(rle_map);
+    dest->colormap = (rle_map *) (*cinfo->mem->alloc_small)
+      ((j_common_ptr) cinfo, JPOOL_IMAGE, cmapsize);
+    MEMZERO(dest->colormap, cmapsize);
+
+    /* Save away data in RLE format --- note 8-bit left shift! */
+    /* Shifting would need adjustment for JSAMPLEs wider than 8 bits. */
+    for (ci = 0; ci < cinfo->out_color_components; ci++) {
+      for (i = 0; i < cinfo->actual_number_of_colors; i++) {
+        dest->colormap[ci * CMAPLENGTH + i] =
+          GETJSAMPLE(cinfo->colormap[ci][i]) << 8;
+      }
+    }
+  }
+
+  /* Set the output buffer to the first row */
+  dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+    ((j_common_ptr) cinfo, dest->image, (JDIMENSION) 0, (JDIMENSION) 1, TRUE);
+  dest->pub.buffer_height = 1;
+
+  dest->pub.put_pixel_rows = rle_put_pixel_rows;
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->total_extra_passes++;  /* count file writing as separate pass */
+  }
+#endif
+}
+
+
+/*
+ * Write some pixel data.
+ *
+ * This routine just saves the data away in a virtual array.
+ */
+
+METHODDEF(void)
+rle_put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                    JDIMENSION rows_supplied)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+
+  if (cinfo->output_scanline < cinfo->output_height) {
+    dest->pub.buffer = (*cinfo->mem->access_virt_sarray)
+      ((j_common_ptr) cinfo, dest->image,
+       cinfo->output_scanline, (JDIMENSION) 1, TRUE);
+  }
+}
+
+/*
+ * Finish up at the end of the file.
+ *
+ * Here is where we really output the RLE file.
+ */
+
+METHODDEF(void)
+finish_output_rle (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  rle_dest_ptr dest = (rle_dest_ptr) dinfo;
+  rle_hdr header;               /* Output file information */
+  rle_pixel **rle_row, *red, *green, *blue;
+  JSAMPROW output_row;
+  char cmapcomment[80];
+  int row, col;
+  int ci;
+#ifdef PROGRESS_REPORT
+  cd_progress_ptr progress = (cd_progress_ptr) cinfo->progress;
+#endif
+
+  /* Initialize the header info */
+  header = *rle_hdr_init(NULL);
+  header.rle_file = dest->pub.output_file;
+  header.xmin     = 0;
+  header.xmax     = cinfo->output_width  - 1;
+  header.ymin     = 0;
+  header.ymax     = cinfo->output_height - 1;
+  header.alpha    = 0;
+  header.ncolors  = cinfo->output_components;
+  for (ci = 0; ci < cinfo->output_components; ci++) {
+    RLE_SET_BIT(header, ci);
+  }
+  if (cinfo->quantize_colors) {
+    header.ncmap   = cinfo->out_color_components;
+    header.cmaplen = CMAPBITS;
+    header.cmap    = dest->colormap;
+    /* Add a comment to the output image with the true colormap length. */
+    sprintf(cmapcomment, "color_map_length=%d", cinfo->actual_number_of_colors);
+    rle_putcom(cmapcomment, &header);
+  }
+
+  /* Emit the RLE header and color map (if any) */
+  rle_put_setup(&header);
+
+  /* Now output the RLE data from our virtual array.
+   * We assume here that rle_pixel is represented the same as JSAMPLE.
+   */
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL) {
+    progress->pub.pass_limit = cinfo->output_height;
+    progress->pub.pass_counter = 0;
+    (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+  }
+#endif
+
+  if (cinfo->output_components == 1) {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) (*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+         (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  } else {
+    for (row = cinfo->output_height-1; row >= 0; row--) {
+      rle_row = (rle_pixel **) dest->rle_row;
+      output_row = *(*cinfo->mem->access_virt_sarray)
+        ((j_common_ptr) cinfo, dest->image,
+         (JDIMENSION) row, (JDIMENSION) 1, FALSE);
+      red = rle_row[0];
+      green = rle_row[1];
+      blue = rle_row[2];
+      for (col = cinfo->output_width; col > 0; col--) {
+        *red++ = GETJSAMPLE(*output_row++);
+        *green++ = GETJSAMPLE(*output_row++);
+        *blue++ = GETJSAMPLE(*output_row++);
+      }
+      rle_putrow(rle_row, (int) cinfo->output_width, &header);
+#ifdef PROGRESS_REPORT
+      if (progress != NULL) {
+        progress->pub.pass_counter++;
+        (*progress->pub.progress_monitor) ((j_common_ptr) cinfo);
+      }
+#endif
+    }
+  }
+
+#ifdef PROGRESS_REPORT
+  if (progress != NULL)
+    progress->completed_extra_passes++;
+#endif
+
+  /* Emit file trailer */
+  rle_puteof(&header);
+  fflush(dest->pub.output_file);
+  if (ferror(dest->pub.output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for RLE format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_rle (j_decompress_ptr cinfo)
+{
+  rle_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (rle_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(rle_dest_struct));
+  dest->pub.start_output = start_output_rle;
+  dest->pub.finish_output = finish_output_rle;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Allocate a work array for output to the RLE library. */
+  dest->rle_row = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE,
+     cinfo->output_width, (JDIMENSION) cinfo->output_components);
+
+  /* Allocate a virtual array to hold the image. */
+  dest->image = (*cinfo->mem->request_virt_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, FALSE,
+     (JDIMENSION) (cinfo->output_width * cinfo->output_components),
+     cinfo->output_height, (JDIMENSION) 1);
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* RLE_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrtarga.c glmark2-2021.02/src/libjpeg-turbo/wrtarga.c
--- glmark2-2014.03+git20150611.fa71af2d/src/libjpeg-turbo/wrtarga.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/libjpeg-turbo/wrtarga.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,247 @@
+/*
+ * wrtarga.c
+ *
+ * This file was part of the Independent JPEG Group's software:
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * It was modified by The libjpeg-turbo Project to include only code and
+ * information relevant to libjpeg-turbo.
+ * For conditions of distribution and use, see the accompanying README.ijg
+ * file.
+ *
+ * This file contains routines to write output images in Targa format.
+ *
+ * These routines may need modification for non-Unix environments or
+ * specialized applications.  As they stand, they assume output to
+ * an ordinary stdio stream.
+ *
+ * Based on code contributed by Lee Daniel Crocker.
+ */
+
+#include "cdjpeg.h"             /* Common decls for cjpeg/djpeg applications */
+
+#ifdef TARGA_SUPPORTED
+
+
+/*
+ * To support 12-bit JPEG data, we'd have to scale output down to 8 bits.
+ * This is not yet implemented.
+ */
+
+#if BITS_IN_JSAMPLE != 8
+  Sorry, this code only copes with 8-bit JSAMPLEs. /* deliberate syntax err */
+#endif
+
+
+/* Private version of data destination object */
+
+typedef struct {
+  struct djpeg_dest_struct pub; /* public fields */
+
+  char *iobuffer;               /* physical I/O buffer */
+  JDIMENSION buffer_width;      /* width of one row */
+} tga_dest_struct;
+
+typedef tga_dest_struct *tga_dest_ptr;
+
+
+LOCAL(void)
+write_header (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo, int num_colors)
+/* Create and write a Targa header */
+{
+  char targaheader[18];
+
+  /* Set unused fields of header to 0 */
+  MEMZERO(targaheader, sizeof(targaheader));
+
+  if (num_colors > 0) {
+    targaheader[1] = 1;         /* color map type 1 */
+    targaheader[5] = (char) (num_colors & 0xFF);
+    targaheader[6] = (char) (num_colors >> 8);
+    targaheader[7] = 24;        /* 24 bits per cmap entry */
+  }
+
+  targaheader[12] = (char) (cinfo->output_width & 0xFF);
+  targaheader[13] = (char) (cinfo->output_width >> 8);
+  targaheader[14] = (char) (cinfo->output_height & 0xFF);
+  targaheader[15] = (char) (cinfo->output_height >> 8);
+  targaheader[17] = 0x20;       /* Top-down, non-interlaced */
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    targaheader[2] = 3;         /* image type = uncompressed grayscale */
+    targaheader[16] = 8;        /* bits per pixel */
+  } else {                      /* must be RGB */
+    if (num_colors > 0) {
+      targaheader[2] = 1;       /* image type = colormapped RGB */
+      targaheader[16] = 8;
+    } else {
+      targaheader[2] = 2;       /* image type = uncompressed RGB */
+      targaheader[16] = 24;
+    }
+  }
+
+  if (JFWRITE(dinfo->output_file, targaheader, 18) != (size_t) 18)
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * Write some pixel data.
+ * In this module rows_supplied will always be 1.
+ */
+
+METHODDEF(void)
+put_pixel_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                JDIMENSION rows_supplied)
+/* used for unquantized full-color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    outptr[0] = (char) GETJSAMPLE(inptr[2]); /* RGB to BGR order */
+    outptr[1] = (char) GETJSAMPLE(inptr[1]);
+    outptr[2] = (char) GETJSAMPLE(inptr[0]);
+    inptr += 3, outptr += 3;
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+METHODDEF(void)
+put_gray_rows (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+               JDIMENSION rows_supplied)
+/* used for grayscale OR quantized color output */
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(*inptr++);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Write some demapped pixel data when color quantization is in effect.
+ * For Targa, this is only applied to grayscale data.
+ */
+
+METHODDEF(void)
+put_demapped_gray (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo,
+                   JDIMENSION rows_supplied)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  register JSAMPROW inptr;
+  register char *outptr;
+  register JSAMPROW color_map0 = cinfo->colormap[0];
+  register JDIMENSION col;
+
+  inptr = dest->pub.buffer[0];
+  outptr = dest->iobuffer;
+  for (col = cinfo->output_width; col > 0; col--) {
+    *outptr++ = (char) GETJSAMPLE(color_map0[GETJSAMPLE(*inptr++)]);
+  }
+  (void) JFWRITE(dest->pub.output_file, dest->iobuffer, dest->buffer_width);
+}
+
+
+/*
+ * Startup: write the file header.
+ */
+
+METHODDEF(void)
+start_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  tga_dest_ptr dest = (tga_dest_ptr) dinfo;
+  int num_colors, i;
+  FILE *outfile;
+
+  if (cinfo->out_color_space == JCS_GRAYSCALE) {
+    /* Targa doesn't have a mapped grayscale format, so we will */
+    /* demap quantized gray output.  Never emit a colormap. */
+    write_header(cinfo, dinfo, 0);
+    if (cinfo->quantize_colors)
+      dest->pub.put_pixel_rows = put_demapped_gray;
+    else
+      dest->pub.put_pixel_rows = put_gray_rows;
+  } else if (cinfo->out_color_space == JCS_RGB) {
+    if (cinfo->quantize_colors) {
+      /* We only support 8-bit colormap indexes, so only 256 colors */
+      num_colors = cinfo->actual_number_of_colors;
+      if (num_colors > 256)
+        ERREXIT1(cinfo, JERR_TOO_MANY_COLORS, num_colors);
+      write_header(cinfo, dinfo, num_colors);
+      /* Write the colormap.  Note Targa uses BGR byte order */
+      outfile = dest->pub.output_file;
+      for (i = 0; i < num_colors; i++) {
+        putc(GETJSAMPLE(cinfo->colormap[2][i]), outfile);
+        putc(GETJSAMPLE(cinfo->colormap[1][i]), outfile);
+        putc(GETJSAMPLE(cinfo->colormap[0][i]), outfile);
+      }
+      dest->pub.put_pixel_rows = put_gray_rows;
+    } else {
+      write_header(cinfo, dinfo, 0);
+      dest->pub.put_pixel_rows = put_pixel_rows;
+    }
+  } else {
+    ERREXIT(cinfo, JERR_TGA_COLORSPACE);
+  }
+}
+
+
+/*
+ * Finish up at the end of the file.
+ */
+
+METHODDEF(void)
+finish_output_tga (j_decompress_ptr cinfo, djpeg_dest_ptr dinfo)
+{
+  /* Make sure we wrote the output file OK */
+  fflush(dinfo->output_file);
+  if (ferror(dinfo->output_file))
+    ERREXIT(cinfo, JERR_FILE_WRITE);
+}
+
+
+/*
+ * The module selection routine for Targa format output.
+ */
+
+GLOBAL(djpeg_dest_ptr)
+jinit_write_targa (j_decompress_ptr cinfo)
+{
+  tga_dest_ptr dest;
+
+  /* Create module interface object, fill in method pointers */
+  dest = (tga_dest_ptr)
+      (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                  sizeof(tga_dest_struct));
+  dest->pub.start_output = start_output_tga;
+  dest->pub.finish_output = finish_output_tga;
+
+  /* Calculate output image dimensions so we can allocate space */
+  jpeg_calc_output_dimensions(cinfo);
+
+  /* Create I/O buffer. */
+  dest->buffer_width = cinfo->output_width * cinfo->output_components;
+  dest->iobuffer = (char *)
+    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
+                                (size_t) (dest->buffer_width * sizeof(char)));
+
+  /* Create decompressor output buffer. */
+  dest->pub.buffer = (*cinfo->mem->alloc_sarray)
+    ((j_common_ptr) cinfo, JPOOL_IMAGE, dest->buffer_width, (JDIMENSION) 1);
+  dest->pub.buffer_height = 1;
+
+  return (djpeg_dest_ptr) dest;
+}
+
+#endif /* TARGA_SUPPORTED */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/log.cc glmark2-2021.02/src/libmatrix/log.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/log.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/log.cc	2021-04-25 01:47:22.000000000 +0800
@@ -10,7 +10,6 @@
 //     Alexandros Frantzis <alexandros.frantzis@linaro.org>
 //     Jesse Barker <jesse.barker@linaro.org>
 //
-#include <unistd.h>
 #include <cstdio>
 #include <cstdarg>
 #include <string>
@@ -22,6 +21,13 @@
 #include <android/log.h>
 #endif
 
+#ifdef _WIN32
+// On windows 'isatty' is found in <io.h>
+#include <io.h>
+#else
+#include <unistd.h>
+#endif
+
 using std::string;
 
 const string Log::continuation_prefix("\x10");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/mat.h glmark2-2021.02/src/libmatrix/mat.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/mat.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/mat.h	2021-04-25 01:47:22.000000000 +0800
@@ -109,9 +109,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat2& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
@@ -411,9 +408,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat3& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
@@ -793,9 +787,6 @@
     // NOTE: If this is non-invertible, we will
     //       throw to avoid undefined behavior.
     tmat4& inverse()
-#ifdef USE_EXCEPTIONS
-        throw(std::runtime_error)
-#endif // USE_EXCEPTIONS
     {
         T d(determinant());
         if (d == static_cast<T>(0))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/program.cc glmark2-2021.02/src/libmatrix/program.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/program.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/program.cc	2021-04-25 01:47:22.000000000 +0800
@@ -190,7 +190,7 @@
     }
 
     shader.attach(handle_);
-    shaders_.push_back(shader);
+    shaders_.push_back(std::move(shader));
     return;
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/program.h glmark2-2021.02/src/libmatrix/program.h
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/program.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/program.h	2021-04-25 01:47:22.000000000 +0800
@@ -35,6 +35,13 @@
         message_(shader.message_),
         ready_(shader.ready_),
         valid_(shader.valid_) {}
+    Shader(Shader&& shader) noexcept:
+        handle_(std::exchange(shader.handle_, 0)),
+        type_(std::exchange(shader.type_, 0)),
+        source_(std::move(shader.source_)),
+        message_(std::move(shader.message_)),
+        ready_(std::exchange(shader.ready_, false)),
+        valid_(std::exchange(shader.valid_, false)) {}
     Shader(unsigned int type, const std::string& source);
     ~Shader();
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/shader-source.cc glmark2-2021.02/src/libmatrix/shader-source.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/shader-source.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/shader-source.cc	2021-04-25 01:47:22.000000000 +0800
@@ -18,6 +18,36 @@
 #include "vec.h"
 #include "util.h"
 
+namespace
+{
+
+bool is_valid_precision_value(ShaderSource::PrecisionValue precision_value)
+{
+    switch(precision_value) {
+        case ShaderSource::PrecisionValueLow:
+        case ShaderSource::PrecisionValueMedium:
+        case ShaderSource::PrecisionValueHigh:
+        case ShaderSource::PrecisionValueDefault:
+            return true;
+        default:
+            return false;
+    }
+}
+
+bool is_valid_shader_type(ShaderSource::ShaderType shader_type)
+{
+    switch(shader_type) {
+        case ShaderSource::ShaderTypeVertex:
+        case ShaderSource::ShaderTypeFragment:
+        case ShaderSource::ShaderTypeUnknown:
+            return true;
+        default:
+            return false;
+    }
+}
+
+}
+
 /**
  * Holds default precision values for all shader types
  * (even the unknown type, which is hardwired to default precision values)
@@ -34,7 +64,7 @@
 bool
 ShaderSource::load_file(const std::string& filename, std::string& str)
 {
-    std::auto_ptr<std::istream> is_ptr(Util::get_resource(filename));
+    std::unique_ptr<std::istream> is_ptr(Util::get_resource(filename));
     std::istream& inputFile(*is_ptr);
 
     if (!inputFile)
@@ -421,7 +451,9 @@
             ss << "#endif" << std::endl;
         }
     }
-    else if (val >= 0 && val < ShaderSource::PrecisionValueDefault) {
+    else if (is_valid_precision_value(val) &&
+             val != ShaderSource::PrecisionValueDefault)
+    {
         ss << "precision " << precision_map[val] << " ";
         ss << type_str << ";" << std::endl;
     }
@@ -512,7 +544,7 @@
 ShaderSource::default_precision(const ShaderSource::Precision& precision,
                                 ShaderSource::ShaderType type)
 {
-    if (type < 0 || type > ShaderSource::ShaderTypeUnknown)
+    if (!is_valid_shader_type(type))
         type = ShaderSource::ShaderTypeUnknown;
 
     if (type == ShaderSource::ShaderTypeUnknown) {
@@ -537,7 +569,7 @@
 const ShaderSource::Precision&
 ShaderSource::default_precision(ShaderSource::ShaderType type)
 {
-    if (type < 0 || type > ShaderSource::ShaderTypeUnknown)
+    if (!is_valid_shader_type(type))
         type = ShaderSource::ShaderTypeUnknown;
 
     return default_precision_[type];
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/util.cc glmark2-2021.02/src/libmatrix/util.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/libmatrix/util.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/libmatrix/util.cc	2021-04-25 01:47:22.000000000 +0800
@@ -220,10 +220,10 @@
 uint64_t
 Util::get_timestamp_us()
 {
-    struct timeval tv;
-    gettimeofday(&tv, NULL);
-    uint64_t now = static_cast<uint64_t>(tv.tv_sec) * 1000000 +
-                   static_cast<double>(tv.tv_usec);
+    struct timespec ts;
+    clock_gettime(CLOCK_MONOTONIC, &ts);
+    uint64_t now = static_cast<uint64_t>(ts.tv_sec) * 1000000 +
+                   static_cast<double>(ts.tv_nsec)/1000.0;
     return now;
 }
 
@@ -244,7 +244,7 @@
 std::istream *
 Util::get_resource(const std::string &path)
 {
-    std::ifstream *ifs = new std::ifstream(path.c_str());
+    std::ifstream *ifs = new std::ifstream(path.c_str(), std::ios::binary);
 
     return static_cast<std::istream *>(ifs);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/main.cpp glmark2-2021.02/src/main.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/main.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/main.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -42,12 +42,18 @@
 #include "native-state-mir.h"
 #elif GLMARK2_USE_WAYLAND
 #include "native-state-wayland.h"
+#elif GLMARK2_USE_DISPMANX
+#include "native-state-dispmanx.h"
+#elif GLMARK2_USE_WIN32
+#include "native-state-win32.h"
 #endif
 
 #if GLMARK2_USE_EGL
 #include "gl-state-egl.h"
 #elif GLMARK2_USE_GLX
 #include "gl-state-glx.h"
+#elif GLMARK2_USE_WGL
+#include "gl-state-wgl.h"
 #endif
 
 using std::vector;
@@ -138,7 +144,6 @@
 int
 main(int argc, char *argv[])
 {
-
     if (!Options::parse_args(argc, argv))
         return 1;
 
@@ -168,12 +173,18 @@
     NativeStateMir native_state;
 #elif GLMARK2_USE_WAYLAND
     NativeStateWayland native_state;
+#elif GLMARK2_USE_DISPMANX
+    NativeStateDispmanx native_state;
+#elif GLMARK2_USE_WIN32
+    NativeStateWin32 native_state;
 #endif
 
 #if GLMARK2_USE_EGL
     GLStateEGL gl_state;
 #elif GLMARK2_USE_GLX
     GLStateGLX gl_state;
+#elif GLMARK2_USE_WGL
+    GLStateWGL gl_state;
 #endif
 
     CanvasGeneric canvas(native_state, gl_state, Options::size.first, Options::size.second);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/main-loop.cpp glmark2-2021.02/src/main-loop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/main-loop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/main-loop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -80,9 +80,9 @@
 
         /* If we have found a valid scene, set it up */
         if (bench_iter_ != benchmarks_.end()) {
+            before_scene_setup();
             if (!Options::reuse_context)
                 canvas_.reset();
-            before_scene_setup();
             scene_ = &(*bench_iter_)->setup_scene();
             if (!scene_->running()) {
                 if (!scene_->supported(false))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/mesh.cpp glmark2-2021.02/src/mesh.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/mesh.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/mesh.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -396,8 +396,9 @@
         /* Update the current range from the vertex data */
         float *dest(array + nfloats * ri->first);
         for (size_t n = ri->first; n <= ri->second; n++) {
-            for (size_t i = 0; i < nfloats; i++)
-                *dest++ = vertices_[n][offset + i];
+            float *src(vertices_[n].data() + offset);
+            std::copy(src, src + nfloats, dest);
+            dest += nfloats;
         }
 
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/meson.build glmark2-2021.02/src/meson.build
--- glmark2-2014.03+git20150611.fa71af2d/src/meson.build	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/meson.build	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,263 @@
+common_sources = [
+    'benchmark-collection.cpp',
+    'benchmark.cpp',
+    'canvas-generic.cpp',
+    'gl-headers.cpp',
+    'gl-visual-config.cpp',
+    'image-reader.cpp',
+    'libmatrix/log.cc',
+    'libmatrix/mat.cc',
+    'libmatrix/program.cc',
+    'libmatrix/shader-source.cc',
+    'libmatrix/util.cc',
+    'main-loop.cpp',
+    'mesh.cpp',
+    'model.cpp',
+    'options.cpp',
+    'scene-buffer.cpp',
+    'scene-build.cpp',
+    'scene-bump.cpp',
+    'scene-clear.cpp',
+    'scene-conditionals.cpp',
+    'scene.cpp',
+    'scene-default-options.cpp',
+    'scene-desktop.cpp',
+    'scene-effect-2d.cpp',
+    'scene-function.cpp',
+    'scene-grid.cpp',
+    'scene-ideas/a.cc',
+    'scene-ideas.cpp',
+    'scene-ideas/d.cc',
+    'scene-ideas/e.cc',
+    'scene-ideas/i.cc',
+    'scene-ideas/lamp.cc',
+    'scene-ideas/logo.cc',
+    'scene-ideas/m.cc',
+    'scene-ideas/n.cc',
+    'scene-ideas/o.cc',
+    'scene-ideas/s.cc',
+    'scene-ideas/splines.cc',
+    'scene-ideas/table.cc',
+    'scene-ideas/t.cc',
+    'scene-jellyfish.cpp',
+    'scene-loop.cpp',
+    'scene-pulsar.cpp',
+    'scene-refract.cpp',
+    'scene-shading.cpp',
+    'scene-shadow.cpp',
+    'scene-terrain/base-renderer.cpp',
+    'scene-terrain/blur-renderer.cpp',
+    'scene-terrain/copy-renderer.cpp',
+    'scene-terrain.cpp',
+    'scene-terrain/luminance-renderer.cpp',
+    'scene-terrain/normal-from-height-renderer.cpp',
+    'scene-terrain/overlay-renderer.cpp',
+    'scene-terrain/renderer-chain.cpp',
+    'scene-terrain/simplex-noise-renderer.cpp',
+    'scene-terrain/terrain-renderer.cpp',
+    'scene-terrain/texture-renderer.cpp',
+    'scene-texture.cpp',
+    'shared-library.cpp',
+    'text-renderer.cpp',
+    'texture.cpp'
+]
+
+libmatrix_headers_dep = declare_dependency(
+    include_directories: include_directories('libmatrix'),
+    compile_args: ['-DUSE_EXCEPTIONS'],
+)
+
+common_deps = [
+    m_dep,
+    dl_dep,
+    libjpeg_dep,
+    libpng_dep,
+    libmatrix_headers_dep,
+]
+    
+if need_drm
+    native_drm_lib = static_library(
+        'native-drm',
+        'native-state-drm.cpp',
+        dependencies: [libdrm_dep, libudev_dep, gbm_dep, libmatrix_headers_dep],
+        )
+    native_drm_dep = declare_dependency(
+        link_with: native_drm_lib,
+        dependencies: libdrm_dep,
+        compile_args: ['-DGLMARK2_USE_DRM', '-D__GBM__'],
+        )
+else
+    native_drm_dep = declare_dependency()
+endif
+
+if need_wayland
+    wayland_scanner = find_program(wayland_scanner_dep.get_pkgconfig_variable('wayland_scanner'))
+    wayland_protocols_dir = wayland_protocols_dep.get_pkgconfig_variable('pkgdatadir')
+
+    xdg_shell_xml_path = wayland_protocols_dir + '/stable/xdg-shell/xdg-shell.xml'
+    xdg_shell_client_header = custom_target(
+        'xdg-shell client-header',
+        command: [ wayland_scanner, 'client-header', '@INPUT@', '@OUTPUT@' ],
+        input: xdg_shell_xml_path,
+        output: 'xdg-shell-client-protocol.h',
+        )
+    xdg_shell_private_code = custom_target(
+        'xdg-shell private-code',
+        command: [ wayland_scanner, 'private-code', '@INPUT@', '@OUTPUT@' ],
+        input: xdg_shell_xml_path,
+        output: 'xdg-shell-protocol.c',
+        )
+
+    native_wayland_lib = static_library(
+        'native-wayland',
+        'native-state-wayland.cpp',
+        xdg_shell_client_header,
+        xdg_shell_private_code,
+        dependencies: [libmatrix_headers_dep, wayland_client_dep, wayland_cursor_dep, wayland_egl_dep],
+        )
+
+    native_wayland_dep = declare_dependency(
+        link_with: native_wayland_lib,
+        dependencies: [wayland_client_dep, wayland_cursor_dep, wayland_egl_dep],
+        compile_args: ['-DGLMARK2_USE_WAYLAND', '-DWL_EGL_PLATFORM'],
+        )
+else
+    native_wayland_dep = declare_dependency()
+endif
+
+if need_x11
+    native_x11_lib = static_library(
+        'native-x11',
+        'native-state-x11.cpp',
+        dependencies: [x11_dep, libmatrix_headers_dep],
+        )
+
+    native_x11_dep = declare_dependency(
+        link_with: native_x11_lib,
+        compile_args: ['-DGLMARK2_USE_X11'],
+        )
+else
+    native_x11_dep = declare_dependency()
+endif
+
+if need_gl
+    gl_gl_lib = static_library(
+        'gl-gl',
+        'glad/src/gl.c',
+        include_directories: include_directories('glad/include'),
+        )
+    gl_gl_dep = declare_dependency(
+        link_with: gl_gl_lib,
+        include_directories: include_directories('glad/include'),
+        compile_args: ['-DGLMARK2_USE_GL'],
+        )
+
+    lib_common_gl = static_library(
+        'common-gl',
+        common_sources,
+        dependencies: [gl_gl_dep, common_deps],
+        include_directories: [
+            include_directories('scene-ideas'),
+            include_directories('scene-terrain'),
+            ]
+        )
+    common_gl_dep = declare_dependency(
+        link_with: lib_common_gl,
+        include_directories: include_directories('libmatrix'),
+        )
+else
+    gl_gl_dep = declare_dependency()
+    common_gl_dep = declare_dependency()
+endif
+
+if need_glesv2
+    gl_glesv2_lib = static_library(
+        'gl-glesv2',
+        'glad/src/gles2.c',
+        include_directories: include_directories('glad/include'),
+        )
+    gl_glesv2_dep = declare_dependency(
+        link_with: gl_glesv2_lib,
+        include_directories: include_directories('glad/include'),
+        compile_args: ['-DGLMARK2_USE_GLESv2'],
+        )
+
+    lib_common_glesv2 = static_library(
+        'common-glesv2',
+        common_sources,
+        dependencies: [gl_glesv2_dep, common_deps],
+        include_directories: [
+            include_directories('scene-ideas'),
+            include_directories('scene-terrain'),
+            ]
+        )
+    common_glesv2_dep = declare_dependency(
+        link_with: lib_common_glesv2,
+        include_directories: include_directories('libmatrix'),
+        )
+else
+    gl_glesv2_dep = declare_dependency()
+    common_glesv2_dep = declare_dependency()
+endif
+
+if need_egl
+    # gl-state-egl builds depend on both the GL type and the native type. To
+    # avoid creating all the required combinations explicitly, make this a
+    # source dependency which will be built properly as part of the final
+    # glmark2 executable.
+    wsi_egl_dep = declare_dependency(
+        sources: ['gl-state-egl.cpp', 'glad/src/egl.c'],
+        include_directories: [
+            include_directories('glad/include'),
+            include_directories('libmatrix'),
+            ],
+        compile_args: ['-DGLMARK2_USE_EGL'],
+        )
+else
+    wsi_egl_dep = declare_dependency()
+endif
+
+if need_glx
+    # Although there is a single valid flavor for glx (x11-gl) and we could
+    # build gl-state-glx here, make this a source dependency for convenience
+    # and parity with gl-state-egl.
+    wsi_glx_dep = declare_dependency(
+        sources: ['gl-state-glx.cpp', 'glad/src/glx.c'],
+        include_directories: [
+            include_directories('glad/include'),
+            include_directories('libmatrix'),
+            ],
+        compile_args: ['-DGLMARK2_USE_GLX'],
+        )
+else
+    wsi_glx_dep = declare_dependency()
+endif
+
+# Flavor info : [ binary, native, gl, wsi ]
+flavor_info = {
+  'drm-gl' : ['glmark2-drm', native_drm_dep, gl_gl_dep, wsi_egl_dep],
+  'drm-glesv2' : ['glmark2-es2-drm', native_drm_dep, gl_glesv2_dep, wsi_egl_dep],
+  'wayland-gl' : ['glmark2-wayland', native_wayland_dep, gl_gl_dep, wsi_egl_dep],
+  'wayland-glesv2' : ['glmark2-es2-wayland', native_wayland_dep, gl_glesv2_dep, wsi_egl_dep],
+  'x11-gl' : ['glmark2', native_x11_dep, gl_gl_dep, wsi_glx_dep],
+  'x11-glesv2' : ['glmark2-es2', native_x11_dep, gl_glesv2_dep, wsi_egl_dep],
+}
+
+foreach flavor : flavors
+    info = flavor_info[flavor]
+    binary = info[0]
+    native_dep = info[1]
+    gl_dep = info[2]
+    wsi_dep = info[3]
+    if gl_dep == gl_gl_dep
+        common_dep = common_gl_dep
+    elif gl_dep == gl_glesv2_dep
+        common_dep = common_glesv2_dep
+    endif
+    executable(
+        binary,
+        ['main.cpp', 'canvas-generic.cpp'],
+        dependencies: [common_dep, native_dep, gl_dep, wsi_dep],
+        install: true,
+    )
+endforeach
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/model.cpp glmark2-2021.02/src/model.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/model.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/model.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -361,7 +361,7 @@
 
     Log::debug("Loading model from 3ds file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& input_file(*input_file_ptr);
 
     if (!input_file) {
@@ -385,7 +385,7 @@
             return false;
         }
 
-        //Read the lenght of the chunk
+        //Read the length of the chunk
         read_or_fail(input_file, &chunk_length, 4);
 
         switch (chunk_id)
@@ -393,7 +393,7 @@
             //----------------- MAIN3DS -----------------
             // Description: Main chunk, contains all the other chunks
             // Chunk ID: 4d4d
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x4d4d:
                 break;
@@ -401,7 +401,7 @@
             //----------------- EDIT3DS -----------------
             // Description: 3D Editor chunk, objects layout info
             // Chunk ID: 3d3d (hex)
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x3d3d:
                 break;
@@ -409,7 +409,7 @@
             //--------------- EDIT_OBJECT ---------------
             // Description: Object block, info for each object
             // Chunk ID: 4000 (hex)
-            // Chunk Lenght: len(object name) + sub chunks
+            // Chunk Length: len(object name) + sub chunks
             //-------------------------------------------
             case 0x4000:
                 {
@@ -429,7 +429,7 @@
             //--------------- OBJ_TRIMESH ---------------
             // Description: Triangular mesh, contains chunks for 3d mesh info
             // Chunk ID: 4100 (hex)
-            // Chunk Lenght: 0 + sub chunks
+            // Chunk Length: 0 + sub chunks
             //-------------------------------------------
             case 0x4100:
                 break;
@@ -437,7 +437,7 @@
             //--------------- TRI_VERTEXL ---------------
             // Description: Vertices list
             // Chunk ID: 4110 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of vertices)
+            // Chunk Length: 1 x unsigned short (number of vertices)
             //             + 3 x float (vertex coordinates) x (number of vertices)
             //             + sub chunks
             //-------------------------------------------
@@ -461,7 +461,7 @@
             //--------------- TRI_FACEL1 ----------------
             // Description: Polygons (faces) list
             // Chunk ID: 4120 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of polygons)
+            // Chunk Length: 1 x unsigned short (number of polygons)
             //             + 3 x unsigned short (polygon points) x (number of polygons)
             //             + sub chunks
             //-------------------------------------------
@@ -484,7 +484,7 @@
             //------------- TRI_MAPPINGCOORS ------------
             // Description: Vertices list
             // Chunk ID: 4140 (hex)
-            // Chunk Lenght: 1 x unsigned short (number of mapping points)
+            // Chunk Length: 1 x unsigned short (number of mapping points)
             //             + 2 x float (mapping coordinates) x (number of mapping points)
             //             + sub chunks
             //-------------------------------------------
@@ -505,7 +505,7 @@
 
             //----------- Skip unknow chunks ------------
             //We need to skip all the chunks that currently we don't use
-            //We use the chunk lenght information to set the file pointer
+            //We use the chunk length information to set the file pointer
             //to the same level next chunk
             //-------------------------------------------
             default:
@@ -681,7 +681,7 @@
 {
     Log::debug("Loading model from obj file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& inputFile(*input_file_ptr);
     if (!inputFile)
     {
@@ -814,7 +814,7 @@
         return ModelPrivate::modelMap;
     }
     vector<string> pathVec;
-    string dataDir(GLMARK_DATA_PATH"/models");
+    string dataDir(Options::data_path + "/models");
     Util::list_files(dataDir, pathVec);
 #ifdef GLMARK_EXTRAS_PATH
     string extrasDir(GLMARK_EXTRAS_PATH"/models");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-dispmanx.cpp glmark2-2021.02/src/native-state-dispmanx.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-dispmanx.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/native-state-dispmanx.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,132 @@
+/*
+ * Copyright © 2013 Rafal Mielniczuk
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Rafal Mielniczuk <rafal.mielniczuk2@gmail.com>
+ */
+
+#include "native-state-dispmanx.h"
+#include "log.h"
+
+#include <cstring>
+#include <csignal>
+
+NativeStateDispmanx::NativeStateDispmanx()
+{
+    memset(&properties_, 0, sizeof(properties_));
+    memset(&egl_dispmanx_window, 0, sizeof(egl_dispmanx_window));
+}
+
+NativeStateDispmanx::~NativeStateDispmanx()
+{
+}
+
+bool
+NativeStateDispmanx::init_display()
+{
+    bcm_host_init();
+
+    return true;
+}
+
+void*
+NativeStateDispmanx::display()
+{
+    return EGL_DEFAULT_DISPLAY;
+}
+
+bool
+NativeStateDispmanx::create_window(WindowProperties const& properties)
+{
+    int dispmanx_output = 0; /* LCD */
+
+    if (!properties.fullscreen) {
+        Log::error("Error: Dispmanx only supports full screen rendering.\n");
+	return false;
+    }
+
+    unsigned screen_width, screen_height;
+    if (graphics_get_display_size(dispmanx_output,
+				  &screen_width, &screen_height) < 0) {
+        Log::error("Error: Couldn't get screen width/height.\n");
+	return false;
+    }
+
+    properties_.fullscreen = properties.fullscreen;
+    properties_.visual_id = properties.visual_id;
+    properties_.width = screen_width;
+    properties_.height = screen_height;
+
+    dispmanx_display = vc_dispmanx_display_open(dispmanx_output);
+
+    VC_RECT_T dst_rect;
+    VC_RECT_T src_rect;
+    dst_rect.x = 0;
+    dst_rect.y = 0;
+    dst_rect.width = screen_width;
+    dst_rect.height = screen_height;
+
+    src_rect.x = 0;
+    src_rect.y = 0;
+    src_rect.width = screen_width << 16;
+    src_rect.height = screen_height << 16;
+
+    dispmanx_update = vc_dispmanx_update_start(0);
+
+    dispmanx_element = vc_dispmanx_element_add(dispmanx_update,
+					       dispmanx_display,
+					       0 /*layer*/,
+					       &dst_rect,
+					       0 /*src*/,
+					       &src_rect,
+					       DISPMANX_PROTECTION_NONE,
+					       0 /*alpha*/,
+					       0 /*clamp*/,
+					       DISPMANX_NO_ROTATE);
+
+    egl_dispmanx_window.element = dispmanx_element;
+    egl_dispmanx_window.width = dst_rect.width;
+    egl_dispmanx_window.height = dst_rect.height;
+    vc_dispmanx_update_submit_sync(dispmanx_update);
+
+    return true;
+}
+
+void*
+NativeStateDispmanx::window(WindowProperties &properties)
+{
+    properties = properties_;
+    return &egl_dispmanx_window;
+}
+
+void
+NativeStateDispmanx::visible(bool /*v*/)
+{
+}
+
+bool
+NativeStateDispmanx::should_quit()
+{
+    return false;
+}
+
+void
+NativeStateDispmanx::flip()
+{
+}
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-dispmanx.h glmark2-2021.02/src/native-state-dispmanx.h
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-dispmanx.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/native-state-dispmanx.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+/*
+ * Copyright © 2013 Rafal Mielniczuk
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Rafal Mielniczuk <rafal.mielniczuk2@gmail.com>
+ */
+
+#ifndef GLMARK2_NATIVE_STATE_DISPMANX_H_
+#define GLMARK2_NATIVE_STATE_DISPMANX_H_
+
+#include <vector>
+#include "bcm_host.h"
+#include "GLES/gl.h"
+#include "EGL/egl.h"
+#include "EGL/eglext.h"
+
+#include "native-state.h"
+
+class NativeStateDispmanx : public NativeState
+{
+public:
+    NativeStateDispmanx();
+    ~NativeStateDispmanx();
+
+    bool init_display();
+    void* display();
+    bool create_window(WindowProperties const& properties);
+    void* window(WindowProperties& properties);
+    void visible(bool v);
+    bool should_quit();
+    void flip();
+
+private:
+    DISPMANX_DISPLAY_HANDLE_T dispmanx_display;
+    DISPMANX_UPDATE_HANDLE_T dispmanx_update;
+    DISPMANX_ELEMENT_HANDLE_T dispmanx_element;
+    EGL_DISPMANX_WINDOW_T egl_dispmanx_window;
+    WindowProperties properties_;
+};
+
+#endif /* GLMARK2_NATIVE_STATE_DISPMANX_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-drm.cpp glmark2-2021.02/src/native-state-drm.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-drm.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-drm.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -25,6 +25,11 @@
 #include "native-state-drm.h"
 #include "log.h"
 
+#include <fcntl.h>
+#include <libudev.h>
+#include <cstring>
+#include <string>
+
 /******************
  * Public methods *
  ******************/
@@ -106,7 +111,8 @@
     FD_ZERO(&fds);
     FD_SET(fd_, &fds);
     drmEventContext evCtx;
-    evCtx.version = DRM_EVENT_CONTEXT_VERSION;
+    memset(&evCtx, 0, sizeof(evCtx));
+    evCtx.version = 2;
     evCtx.page_flip_handler = page_flip_handler;
 
     while (waiting) {
@@ -134,6 +140,269 @@
  * Private methods *
  *******************/
 
+/* Simple helpers */
+
+inline static bool valid_fd(int fd)
+{
+    return fd >= 0;
+}
+
+inline static bool valid_drm_node_path(std::string const& provided_node_path)
+{
+    return !provided_node_path.empty();
+}
+
+inline static bool invalid_drm_node_path(std::string const& provided_node_path)
+{
+    return !(valid_drm_node_path(provided_node_path));
+}
+
+/* Udev methods */
+// Udev detection functions
+#define UDEV_TEST_FUNC_SIGNATURE(udev_identifier, device_identifier, syspath_identifier) \
+    struct udev * __restrict const udev_identifier, \
+    struct udev_device * __restrict const device_identifier, \
+    char const * __restrict syspath_identifier
+
+/* Omitting the parameter names is kind of ugly but is the only way
+ * to force G++ to forget about the unused parameters.
+ * Having big warnings during the compilation isn't very nice.
+ *
+ * These functions will be used as function pointers and should have
+ * the same signature to avoid weird stack related issues.
+ *
+ * Another way to deal with that issue will be to mark unused parameters
+ * with __attribute__((unused))
+ */
+static bool udev_drm_test_virtual(
+    UDEV_TEST_FUNC_SIGNATURE(,,tested_node_syspath))
+{
+    return strstr(tested_node_syspath, "virtual") != NULL;
+}
+
+static bool udev_drm_test_not_virtual(
+    UDEV_TEST_FUNC_SIGNATURE(udev, current_device, tested_node_syspath))
+{
+    return !udev_drm_test_virtual(udev,
+                                  current_device,
+                                  tested_node_syspath);
+}
+
+static bool
+udev_drm_test_primary_gpu(UDEV_TEST_FUNC_SIGNATURE(, current_device,))
+{
+    bool is_main_gpu = false;
+
+    auto const drm_node_parent = udev_device_get_parent(current_device);
+
+    /* While tempting, using udev_device_unref will generate issues
+     * when unreferencing the child in udev_get_node_that_pass_in_enum
+     *
+     * udev_device_unref WILL unreference the parent, so avoid doing
+     * that here.
+     *
+     * ( See udev sources : src/libudev/libudev-device.c )
+     */
+    if (drm_node_parent != NULL) {
+        is_main_gpu =
+            (udev_device_get_sysattr_value(drm_node_parent, "boot_vga")
+             != NULL);
+    }
+
+    return is_main_gpu;
+}
+
+/* Test if the drm-device is actually modeset capable.
+ * Render-only devices cannot drive an actual display,
+ * so the GETRESOURCES ioctl will fail in that case.
+ */
+static bool udev_drm_test_modeset(std::string const& dev_path)
+{
+    struct drm_mode_card_res res {};
+    int fd, ret;
+
+    fd = open(dev_path.c_str(), O_RDWR);
+    if (!valid_fd(fd))
+        return false;
+
+    ret = drmIoctl(fd, DRM_IOCTL_MODE_GETRESOURCES, &res);
+    drmClose(fd);
+
+    return !ret;
+}
+
+static std::string
+udev_get_node_that_pass_in_enum(
+    struct udev * __restrict const udev,
+    struct udev_enumerate * __restrict const dev_enum,
+    bool (* check_function)(UDEV_TEST_FUNC_SIGNATURE(,,)))
+{
+    std::string result;
+
+    auto current_element = udev_enumerate_get_list_entry(dev_enum);
+
+    while (current_element && result.empty()) {
+        char const * __restrict current_element_sys_path =
+            udev_list_entry_get_name(current_element);
+
+        if (current_element_sys_path) {
+            struct udev_device * current_device =
+                udev_device_new_from_syspath(udev,
+                                             current_element_sys_path);
+            auto check_passed = check_function(
+                udev, current_device, current_element_sys_path);
+
+            if (check_passed) {
+                const char * device_node_path =
+                    udev_device_get_devnode(current_device);
+
+                if (device_node_path &&
+                    udev_drm_test_modeset(device_node_path)) {
+                        result = device_node_path;
+                }
+
+            }
+
+            udev_device_unref(current_device);
+        }
+
+        current_element = udev_list_entry_get_next(current_element);
+    }
+
+    return result;
+}
+
+/* Inspired by KWin detection mechanism */
+/* And yet KWin got it wrong too, it seems.
+ * 1 - Looking for the primary GPU by checking the flag 'boot_vga'
+ *     won't get you far with some embedded chipsets, like Rockchip.
+ * 2 - Looking for a GPU plugged in PCI will fail on various embedded
+ *     devices !
+ * 3 - Looking for a render node is not guaranteed to work on some
+ *     poorly maintained DRM drivers, which plague some embedded
+ *     devices too...
+ *
+ * So, we won't play too smart here.
+ * - We first check for a primary GPU plugged in PCI with the 'boot_vga'
+ *   attribute, to take care of Desktop users using multiple GPU.
+ * - Then, we just check for a DRM node that is not virtual
+ * - At least, we use the first virtual node we get, if we didn't find
+ *   anything yet.
+ * This should take care of almost every potential use case.
+ *
+ * The remaining ones will be dealt with an additional option to
+ * specify the DRM dev node manually.
+ */
+static std::string udev_main_gpu_drm_node_path()
+{
+    Log::debug("Using Udev to detect the right DRM node to use\n");
+    auto udev = udev_new();
+    auto dev_enumeration = udev_enumerate_new(udev);
+
+    udev_enumerate_add_match_subsystem(dev_enumeration, "drm");
+    udev_enumerate_add_match_sysname(dev_enumeration, "card[0-9]*");
+    udev_enumerate_scan_devices(dev_enumeration);
+
+    Log::debug("Looking for the main GPU DRM node...\n");
+    std::string node_path = udev_get_node_that_pass_in_enum(
+        udev, dev_enumeration, udev_drm_test_primary_gpu);
+
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found!\n");
+        Log::debug("Looking for a concrete GPU DRM node...\n");
+        node_path = udev_get_node_that_pass_in_enum(
+            udev, dev_enumeration, udev_drm_test_not_virtual);
+    }
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found!?\n");
+        Log::debug("Looking for a virtual GPU DRM node...\n");
+        node_path = udev_get_node_that_pass_in_enum(
+            udev, dev_enumeration, udev_drm_test_virtual);
+    }
+    if (invalid_drm_node_path(node_path)) {
+        Log::debug("Not found.\n");
+        Log::debug("Cannot find a single DRM node using UDEV...\n");
+    }
+
+    if (valid_drm_node_path(node_path)) {
+        Log::debug("Success!\n");
+    }
+
+    udev_enumerate_unref(dev_enumeration);
+    udev_unref(udev);
+
+    return node_path;
+}
+
+static int open_using_udev_scan()
+{
+    auto dev_path = udev_main_gpu_drm_node_path();
+
+    int fd = -1;
+    if (valid_drm_node_path(dev_path)) {
+        Log::debug("Trying to use the DRM node %s\n", dev_path.c_str());
+        fd = open(dev_path.c_str(), O_RDWR);
+    }
+    else {
+        Log::error("Can't determine the main graphic card "
+                   "DRM device node\n");
+    }
+
+    if (!valid_fd(fd)) {
+        // %m is GLIBC specific... Maybe use strerror here...
+        Log::error("Tried to use '%s' but failed.\nReason : %m\n",
+                   dev_path.c_str());
+    }
+    else
+        Log::debug("Success!\n");
+
+    return fd;
+}
+
+/* -- End of Udev helpers -- */
+
+/*
+ * This method is there to take care of cases that would not be handled
+ * by open_using_udev_scan. This should not happen.
+ *
+ * If your driver defines a /dev/dri/cardX node and open_using_udev_scan
+ * were not able to detect it, you should probably file an issue.
+ *
+ */
+static int open_using_module_checking()
+{
+    static const char* drm_modules[] = {
+        "i915",
+        "imx-drm",
+        "nouveau",
+        "radeon",
+        "vmgfx",
+        "omapdrm",
+        "exynos",
+        "pl111",
+        "vc4",
+        "msm",
+        "meson",
+        "rockchip",
+        "sun4i-drm",
+        "stm",
+    };
+
+    int fd = -1;
+    unsigned int num_modules(sizeof(drm_modules)/sizeof(drm_modules[0]));
+    for (unsigned int m = 0; m < num_modules; m++) {
+        fd = drmOpen(drm_modules[m], 0);
+        if (fd < 0) {
+            Log::debug("Failed to open DRM module '%s'\n", drm_modules[m]);
+            continue;
+        }
+        Log::debug("Opened DRM module '%s'\n", drm_modules[m]);
+        break;
+    }
+
+    return fd;
+}
+
 void
 NativeStateDRM::fb_destroy_callback(gbm_bo* bo, void* data)
 {
@@ -197,34 +466,23 @@
 bool
 NativeStateDRM::init()
 {
-    // TODO: Replace this with something that explicitly probes for the loaded
-    // driver (udev?).
-    static const char* drm_modules[] = {
-        "i915",
-        "nouveau",
-        "radeon",
-        "vmgfx",
-        "omapdrm",
-        "exynos"
-    };
+    // TODO: The user should be able to define *exactly* which device
+    //       node to open and the program should try to open only
+    //       this node, in order to take care of unknown use cases.
+    int fd = open_using_udev_scan();
 
-    unsigned int num_modules(sizeof(drm_modules)/sizeof(drm_modules[0]));
-    for (unsigned int m = 0; m < num_modules; m++) {
-        fd_ = drmOpen(drm_modules[m], 0);
-        if (fd_ < 0) {
-            Log::debug("Failed to open DRM module '%s'\n", drm_modules[m]);
-            continue;
-        }
-        Log::debug("Opened DRM module '%s'\n", drm_modules[m]);
-        break;
+    if (!valid_fd(fd)) {
+        fd = open_using_module_checking();
     }
 
-    if (fd_ < 0) {
+    if (!valid_fd(fd)) {
         Log::error("Failed to find a suitable DRM device\n");
         return false;
     }
 
-    resources_ = drmModeGetResources(fd_);
+    fd_ = fd;
+
+    resources_ = drmModeGetResources(fd);
     if (!resources_) {
         Log::error("drmModeGetResources failed\n");
         return false;
@@ -232,7 +490,7 @@
 
     // Find a connected connector
     for (int c = 0; c < resources_->count_connectors; c++) {
-        connector_ = drmModeGetConnector(fd_, resources_->connectors[c]);
+        connector_ = drmModeGetConnector(fd, resources_->connectors[c]);
         if (DRM_MODE_CONNECTED == connector_->connection) {
             break;
         }
@@ -263,14 +521,40 @@
 
     // Find a suitable encoder
     for (int e = 0; e < resources_->count_encoders; e++) {
+        bool found = false;
         encoder_ = drmModeGetEncoder(fd_, resources_->encoders[e]);
-        if (encoder_ && encoder_->encoder_id == connector_->encoder_id) {
-            break;
+        for (int ce = 0; ce < connector_->count_encoders; ce++) {
+            if (encoder_ && encoder_->encoder_id == connector_->encoders[ce]) {
+                found = true;
+                break;
+            }
         }
+        if (found)
+            break;
         drmModeFreeEncoder(encoder_);
         encoder_ = 0;
     }
 
+    // If encoder is not connected to the connector try to find
+    // a suitable one
+    if (!encoder_) {
+        for (int e = 0; e < connector_->count_encoders; e++) {
+            encoder_ = drmModeGetEncoder(fd, connector_->encoders[e]);
+            for (int c = 0; c < resources_->count_crtcs; c++) {
+                if (encoder_->possible_crtcs & (1 << c)) {
+                    encoder_->crtc_id = resources_->crtcs[c];
+                    break;
+                }
+            }
+            if (encoder_->crtc_id) {
+                break;
+            }
+
+            drmModeFreeEncoder(encoder_);
+            encoder_ = 0;
+        }
+    }
+
     if (!encoder_) {
         Log::error("Failed to find a suitable encoder\n");
         return false;
@@ -282,8 +566,13 @@
 
     crtc_ = drmModeGetCrtc(fd_, encoder_->crtc_id);
     if (!crtc_) {
-        Log::error("Failed to get current CRTC\n");
-        return false;
+        // if there is no current CRTC, make sure to attach a suitable one
+        for (int c = 0; c < resources_->count_crtcs; c++) {
+            if (encoder_->possible_crtcs & (1 << c)) {
+                encoder_->crtc_id = resources_->crtcs[c];
+                break;
+            }
+        }
     }
 
     signal(SIGINT, &NativeStateDRM::quit_handler);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-drm.h glmark2-2021.02/src/native-state-drm.h
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-drm.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-drm.h	2021-04-25 01:47:22.000000000 +0800
@@ -41,6 +41,7 @@
         resources_(0),
         connector_(0),
         encoder_(0),
+        crtc_(0),
         mode_(0),
         dev_(0),
         surface_(0),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state.h glmark2-2021.02/src/native-state.h
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,12 +22,14 @@
 #ifndef GLMARK2_NATIVE_STATE_H_
 #define GLMARK2_NATIVE_STATE_H_
 
+#include <stdint.h>
+
 class NativeState
 {
 public:
     struct WindowProperties
     {
-        WindowProperties(int w, int h, bool f, int v)
+        WindowProperties(int w, int h, bool f, intptr_t v)
             : width(w), height(h), fullscreen(f), visual_id(v) {}
         WindowProperties()
             : width(0), height(0), fullscreen(false), visual_id(0) {}
@@ -35,7 +37,7 @@
         int width;
         int height;
         bool fullscreen;
-        int visual_id;
+        intptr_t visual_id;
     };
 
     virtual ~NativeState() {}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-mir.cpp glmark2-2021.02/src/native-state-mir.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-mir.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-mir.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -157,7 +157,7 @@
 bool
 NativeStateMir::create_window(WindowProperties const& properties)
 {
-    static const char *win_name("glmark2 "GLMARK_VERSION);
+    static const char *win_name("glmark2 " GLMARK_VERSION);
 
     if (!mir_connection_is_valid(mir_connection_)) {
         Log::error("Cannot create a Mir surface without a valid connection "
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-wayland.cpp glmark2-2021.02/src/native-state-wayland.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-wayland.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-wayland.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -21,19 +21,29 @@
  */
 
 #include "native-state-wayland.h"
+#include "log.h"
 
+#include <linux/input.h>
 #include <cstring>
 #include <csignal>
+#include <unistd.h>
 
 const struct wl_registry_listener NativeStateWayland::registry_listener_ = {
     NativeStateWayland::registry_handle_global,
     NativeStateWayland::registry_handle_global_remove
 };
 
-const struct wl_shell_surface_listener NativeStateWayland::shell_surface_listener_ = {
-    NativeStateWayland::shell_surface_handle_ping,
-    NativeStateWayland::shell_surface_handle_configure,
-    NativeStateWayland::shell_surface_handle_popup_done
+const struct xdg_toplevel_listener NativeStateWayland::xdg_toplevel_listener_ = {
+    NativeStateWayland::xdg_toplevel_handle_configure,
+    NativeStateWayland::xdg_toplevel_handle_close
+};
+
+const struct xdg_surface_listener NativeStateWayland::xdg_surface_listener_ = {
+    NativeStateWayland::xdg_surface_handle_configure
+};
+
+const struct xdg_wm_base_listener NativeStateWayland::xdg_wm_base_listener_ = {
+    NativeStateWayland::xdg_wm_base_handle_ping
 };
 
 const struct wl_output_listener NativeStateWayland::output_listener_ = {
@@ -43,29 +53,58 @@
     NativeStateWayland::output_handle_scale
 };
 
+const struct wl_seat_listener NativeStateWayland::seat_listener_ = {
+    NativeStateWayland::seat_handle_capabilities,
+};
+
+const struct wl_pointer_listener NativeStateWayland::pointer_listener_ = {
+    NativeStateWayland::pointer_handle_enter,
+    NativeStateWayland::pointer_handle_leave,
+    NativeStateWayland::pointer_handle_motion,
+    NativeStateWayland::pointer_handle_button,
+    NativeStateWayland::pointer_handle_axis,
+};
+
+const struct wl_keyboard_listener NativeStateWayland::keyboard_listener_ = {
+    NativeStateWayland::keyboard_handle_keymap,
+    NativeStateWayland::keyboard_handle_enter,
+    NativeStateWayland::keyboard_handle_leave,
+    NativeStateWayland::keyboard_handle_key,
+    NativeStateWayland::keyboard_handle_modifiers,
+};
+
 volatile bool NativeStateWayland::should_quit_ = false;
 
-NativeStateWayland::NativeStateWayland() : display_(0), window_(0)
+NativeStateWayland::NativeStateWayland() : cursor_(0), display_(0), window_(0)
 {
 }
 
 NativeStateWayland::~NativeStateWayland()
 {
     if (window_) {
-        if (window_->shell_surface)
-            wl_shell_surface_destroy(window_->shell_surface);
-        if (window_->opaque_reqion)
-            wl_region_destroy(window_->opaque_reqion);
-        if (window_->surface)
-            wl_surface_destroy(window_->surface);
+        if (window_->xdg_toplevel)
+            xdg_toplevel_destroy(window_->xdg_toplevel);
+        if (window_->xdg_surface)
+            xdg_surface_destroy(window_->xdg_surface);
         if (window_->native)
             wl_egl_window_destroy(window_->native);
+        if (window_->surface)
+            wl_surface_destroy(window_->surface);
         delete window_;
     }
 
+    if (cursor_) {
+        if (cursor_->cursor_surface)
+            wl_surface_destroy(cursor_->cursor_surface);
+        if (cursor_->cursor_theme)
+            wl_cursor_theme_destroy(cursor_->cursor_theme);
+
+        delete cursor_;
+    }
+
     if (display_) {
-        if (display_->shell)
-            wl_shell_destroy(display_->shell);
+        if (display_->xdg_wm_base)
+            xdg_wm_base_destroy(display_->xdg_wm_base);
 
         for (OutputsVector::iterator it = display_->outputs.begin();
              it != display_->outputs.end(); ++it) {
@@ -88,30 +127,66 @@
 void
 NativeStateWayland::registry_handle_global(void *data, struct wl_registry *registry,
                                            uint32_t id, const char *interface,
-                                           uint32_t /*version*/)
+                                           uint32_t version)
 {
     NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
     if (strcmp(interface, "wl_compositor") == 0) {
         that->display_->compositor =
                 static_cast<struct wl_compositor *>(
                     wl_registry_bind(registry,
-                                     id, &wl_compositor_interface, 1));
-    } else if (strcmp(interface, "wl_shell") == 0) {
-        that->display_->shell =
-                static_cast<struct wl_shell *>(
+                                     id, &wl_compositor_interface, std::min(version, 4U)));
+    } else if (strcmp(interface, "xdg_wm_base") == 0) {
+        that->display_->xdg_wm_base =
+                static_cast<struct xdg_wm_base *>(
                     wl_registry_bind(registry,
-                                     id, &wl_shell_interface, 1));
+                                     id, &xdg_wm_base_interface, std::min(version, 2U)));
+        xdg_wm_base_add_listener(that->display_->xdg_wm_base, &xdg_wm_base_listener_, that);
     } else if (strcmp(interface, "wl_output") == 0) {
-        struct my_output *my_output = new struct my_output;
+        struct my_output *my_output = new struct my_output();
         memset(my_output, 0, sizeof(*my_output));
+        my_output->scale = 1;
         my_output->output =
                 static_cast<struct wl_output *>(
                     wl_registry_bind(registry,
-                                     id, &wl_output_interface, 2));
+                                     id, &wl_output_interface, std::min(version, 2U)));
         that->display_->outputs.push_back(my_output);
 
         wl_output_add_listener(my_output->output, &output_listener_, my_output);
         wl_display_roundtrip(that->display_->display);
+    } else if (strcmp(interface, "wl_seat") == 0) {
+        that->display_->seat =
+            static_cast<struct wl_seat *>(
+            wl_registry_bind(registry, id, &wl_seat_interface, 1));
+        wl_seat_add_listener(that->display_->seat, &seat_listener_, that);
+    } else if (strcmp(interface, "wl_shm") == 0) {
+        that->display_->shm =
+            static_cast<struct wl_shm *>(
+                wl_registry_bind(registry, id, &wl_shm_interface, 1));
+
+        struct my_cursor *my_cursor = new struct my_cursor();
+        my_cursor->cursor_surface =
+             wl_compositor_create_surface(that->display_->compositor);
+        my_cursor->cursor_theme = wl_cursor_theme_load(NULL, 32, that->display_->shm);
+
+        if (!my_cursor->cursor_theme) {
+            Log::error("unable to load default theme\n");
+            wl_surface_destroy(my_cursor->cursor_surface);
+            delete my_cursor;
+            return;
+        }
+
+        my_cursor->default_cursor =
+            wl_cursor_theme_get_cursor(my_cursor->cursor_theme, "left_ptr");
+
+        if (!my_cursor->default_cursor) {
+            wl_surface_destroy(my_cursor->cursor_surface);
+            // assume above cursor_theme was set
+            wl_cursor_theme_destroy(my_cursor->cursor_theme);
+            delete my_cursor;
+            return;
+        }
+
+        that->cursor_ = my_cursor;
     }
 }
 
@@ -152,32 +227,84 @@
 }
 
 void
-NativeStateWayland::output_handle_scale(void * /*data*/, struct wl_output * /*wl_output*/,
-                                        int32_t /*factor*/)
+NativeStateWayland::output_handle_scale(void *data, struct wl_output * /*wl_output*/,
+                                        int32_t factor)
 {
+    struct my_output *my_output = static_cast<struct my_output *>(data);
+    my_output->scale = factor;
 }
 
 void
-NativeStateWayland::shell_surface_handle_ping(void * /*data*/, struct wl_shell_surface *shell_surface,
-                                              uint32_t serial)
+NativeStateWayland::xdg_wm_base_handle_ping(void * /*data*/, struct xdg_wm_base *xdg_wm_base,
+                                            uint32_t serial)
 {
-    wl_shell_surface_pong(shell_surface, serial);
+    xdg_wm_base_pong(xdg_wm_base, serial);
 }
 
 void
-NativeStateWayland::shell_surface_handle_popup_done(void * /*data*/,
-                                                    struct wl_shell_surface * /*shell_surface*/)
+NativeStateWayland::xdg_toplevel_handle_configure(void *data, struct xdg_toplevel * /*xdg_toplevel*/,
+                                                  int32_t width, int32_t height,
+                                                  struct wl_array *states)
 {
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+    uint32_t *state;
+    bool want_fullscreen = false;
+    uint32_t scale = 1;
+
+    that->window_->waiting_for_configure = false;
+
+    if (!that->display_->outputs.empty()) scale = that->display_->outputs.at(0)->scale;
+
+    for (state = (uint32_t *) states->data;
+         state < (uint32_t *) ((char *) states->data + states->size);
+         state++) {
+        if (*state == XDG_TOPLEVEL_STATE_FULLSCREEN)
+            want_fullscreen = true;
+    }
+
+    if (want_fullscreen) {
+        width *= scale;
+        height *= scale;
+    }
+
+    if (want_fullscreen == that->window_->properties.fullscreen &&
+        (width == 0 || width == that->window_->properties.width) &&
+        (height == 0 || height == that->window_->properties.fullscreen)) {
+        return;
+    }
+
+    that->window_->properties.fullscreen = want_fullscreen;
+    that->window_->properties.width = width;
+    that->window_->properties.height = height;
+
+    if (!that->window_->native)
+        return;
+
+    wl_egl_window_resize(that->window_->native, width, height, 0, 0);
+
+    struct wl_region *opaque_reqion = wl_compositor_create_region(that->display_->compositor);
+    wl_region_add(opaque_reqion, 0, 0,
+                  that->window_->properties.width,
+                  that->window_->properties.height);
+    wl_surface_set_opaque_region(that->window_->surface, opaque_reqion);
+    wl_region_destroy(opaque_reqion);
 }
 
 void
-NativeStateWayland::shell_surface_handle_configure(void *data, struct wl_shell_surface * /*shell_surface*/,
-                                                   uint32_t /*edges*/, int32_t width, int32_t height)
+NativeStateWayland::xdg_toplevel_handle_close(void *data,
+                                              struct xdg_toplevel * /*xdg_toplevel */)
 {
     NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
-    that->window_->properties.width = width;
-    that->window_->properties.height = height;
-    wl_egl_window_resize(that->window_->native, width, height, 0, 0);
+
+    that->should_quit_ = true;
+}
+
+void
+NativeStateWayland::xdg_surface_handle_configure(void * /*data*/,
+                                                 struct xdg_surface *xdg_surface,
+                                                 uint32_t serial)
+{
+    xdg_surface_ack_configure(xdg_surface, serial);
 }
 
 bool
@@ -191,7 +318,7 @@
     sigaction(SIGINT, &sa, NULL);
     sigaction(SIGTERM, &sa, NULL);
 
-    display_ = new struct my_display;
+    display_ = new struct my_display();
 
     if (!display_) {
         return false;
@@ -223,42 +350,51 @@
 {
     struct my_output *output = 0;
     if (!display_->outputs.empty()) output = display_->outputs.at(0);
-    window_ = new struct my_window;
+    window_ = new struct my_window();
     window_->properties = properties;
-    window_->surface = wl_compositor_create_surface(display_->compositor);
-    if (window_->properties.fullscreen && output) {
-        window_->native = wl_egl_window_create(window_->surface,
-                                               output->width, output->height);
-        window_->properties.width = output->width;
-        window_->properties.height = output->height;
-    } else {
-        window_->native = wl_egl_window_create(window_->surface,
-                                               properties.width, properties.height);
-    }
 
-    window_->opaque_reqion = wl_compositor_create_region(display_->compositor);
-    wl_region_add(window_->opaque_reqion, 0, 0,
-                  window_->properties.width,
-                  window_->properties.height);
-    wl_surface_set_opaque_region(window_->surface, window_->opaque_reqion);
-
-    window_->shell_surface = wl_shell_get_shell_surface(display_->shell,
-                                                        window_->surface);
+    window_->surface = wl_compositor_create_surface(display_->compositor);
+    window_->xdg_surface = xdg_wm_base_get_xdg_surface(display_->xdg_wm_base,
+                                                       window_->surface);
+    xdg_surface_add_listener(window_->xdg_surface, &xdg_surface_listener_, this);
+    window_->xdg_toplevel = xdg_surface_get_toplevel(window_->xdg_surface);
+    xdg_toplevel_add_listener(window_->xdg_toplevel, &xdg_toplevel_listener_, this);
+
+    xdg_toplevel_set_app_id(window_->xdg_toplevel, "com.github.glmark2.glmark2");
+    xdg_toplevel_set_title(window_->xdg_toplevel, "glmark2");
+    if (window_->properties.fullscreen && output)
+        xdg_toplevel_set_fullscreen(window_->xdg_toplevel, output->output);
+    wl_surface_commit(window_->surface);
+    window_->waiting_for_configure = true;
+
+    /* xdg-shell requires us to wait for the compositor to tell us what its
+     * desired size for us is */
+    while (window_->waiting_for_configure)
+        wl_display_roundtrip(display_->display);
+
+    if (output && window_->properties.fullscreen) {
+        if (window_->properties.width <= 0 ||
+             window_->properties.height <= 0) {
+            window_->properties.width = output->width * output->scale;
+            window_->properties.height = output->height * output->scale;
+        }
 
-    if (window_->shell_surface) {
-        wl_shell_surface_add_listener(window_->shell_surface,
-                                      &shell_surface_listener_, this);
+        if (wl_proxy_get_version((struct wl_proxy *) output) >=
+            WL_OUTPUT_SCALE_SINCE_VERSION) {
+            wl_surface_set_buffer_scale(window_->surface, output->scale);
+        }
     }
 
-    wl_shell_surface_set_title(window_->shell_surface, "glmark2");
+    window_->native = wl_egl_window_create(window_->surface,
+                                           window_->properties.width,
+                                           window_->properties.height);
 
-    if (window_->properties.fullscreen) {
-        wl_shell_surface_set_fullscreen(window_->shell_surface,
-                                        WL_SHELL_SURFACE_FULLSCREEN_METHOD_DRIVER,
-                                        output->refresh, output->output);
-    } else {
-        wl_shell_surface_set_toplevel(window_->shell_surface);
-    }
+    struct wl_region *opaque_reqion = wl_compositor_create_region(display_->compositor);
+    wl_region_add(opaque_reqion, 0, 0,
+                  window_->properties.width,
+                  window_->properties.height);
+    wl_surface_set_opaque_region(window_->surface, opaque_reqion);
+    wl_region_destroy(opaque_reqion);
 
     return true;
 }
@@ -298,3 +434,135 @@
     should_quit_ = true;
 }
 
+void
+NativeStateWayland::seat_handle_capabilities(void *data, struct wl_seat *seat, uint32_t caps)
+{
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    if ((caps & WL_SEAT_CAPABILITY_POINTER) && !that->display_->pointer) {
+        that->display_->pointer = wl_seat_get_pointer(seat);
+        wl_pointer_add_listener(that->display_->pointer, &pointer_listener_, that);
+    } else if (!(caps & WL_SEAT_CAPABILITY_POINTER) && that->display_->pointer) {
+        wl_pointer_destroy(that->display_->pointer);
+        that->display_->pointer = NULL;
+    }
+
+    if ((caps & WL_SEAT_CAPABILITY_KEYBOARD) && !that->display_->keyboard) {
+        that->display_->keyboard = wl_seat_get_keyboard(seat);
+        wl_keyboard_add_listener(that->display_->keyboard, &keyboard_listener_, that);
+    } else if (!(caps & WL_SEAT_CAPABILITY_KEYBOARD) && that->display_->keyboard) {
+        wl_keyboard_destroy(that->display_->keyboard);
+        that->display_->keyboard = NULL;
+    }
+}
+
+
+void
+NativeStateWayland::pointer_handle_enter(void *data, struct wl_pointer *pointer,
+                                         uint32_t serial, struct wl_surface *surface,
+                                         wl_fixed_t sx, wl_fixed_t sy)
+{
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    struct wl_buffer *buffer;
+    struct my_cursor *my_cursor = that->cursor_;
+    struct wl_cursor *cursor;
+    struct wl_cursor_image *image;
+
+    if (!my_cursor)
+        return;
+
+    cursor = my_cursor->default_cursor;
+
+    if (that->window_->properties.fullscreen) {
+        wl_pointer_set_cursor(pointer, serial, NULL, 0, 0);
+    } else {
+        if (cursor) {
+            image = my_cursor->default_cursor->images[0];
+            buffer = wl_cursor_image_get_buffer(image);
+            if (!buffer)
+                return;
+
+            wl_pointer_set_cursor(pointer, serial,
+                                  my_cursor->cursor_surface,
+                                  image->hotspot_x,
+                                  image->hotspot_y);
+            wl_surface_attach(my_cursor->cursor_surface, buffer, 0, 0);
+            wl_surface_damage(my_cursor->cursor_surface, 0, 0,
+                              image->width, image->height);
+            wl_surface_commit(my_cursor->cursor_surface);
+        }
+    }
+}
+
+void
+NativeStateWayland::pointer_handle_leave(void *data, struct wl_pointer *pointer,
+                     uint32_t serial, struct wl_surface *surface)
+{
+}
+
+void
+NativeStateWayland::pointer_handle_motion(void *data, struct wl_pointer *pointer,
+                      uint32_t time, wl_fixed_t sx, wl_fixed_t sy)
+{
+}
+
+void
+NativeStateWayland::pointer_handle_button(void *data, struct wl_pointer *wl_pointer,
+                                          uint32_t serial, uint32_t time, uint32_t button,
+                                          uint32_t state)
+{
+    /* just show that window_ can be used */
+    NativeStateWayland *that = static_cast<NativeStateWayland *>(data);
+
+    if (!that->window_->xdg_toplevel)
+        return;
+
+    if (state == WL_POINTER_BUTTON_STATE_PRESSED && button == BTN_LEFT)
+        xdg_toplevel_move(that->window_->xdg_toplevel,
+                          that->display_->seat, serial);
+}
+
+void
+NativeStateWayland::pointer_handle_axis(void *data, struct wl_pointer *wl_pointer,
+                                        uint32_t time, uint32_t axis, wl_fixed_t value)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_keymap(void *data, struct wl_keyboard *keyboard,
+                                           uint32_t format, int fd, uint32_t size)
+{
+    close(fd);
+}
+
+void
+NativeStateWayland::keyboard_handle_enter(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, struct wl_surface *surface,
+                                          struct wl_array *keys)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_leave(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, struct wl_surface *surface)
+{
+}
+
+void
+NativeStateWayland::keyboard_handle_key(void *data, struct wl_keyboard *keyboard,
+                                        uint32_t serial, uint32_t time, uint32_t key,
+                                        uint32_t state)
+{
+    if (state == WL_KEYBOARD_KEY_STATE_PRESSED &&
+        (key == KEY_ESC || key == KEY_Q))
+        should_quit_ = true;
+}
+
+void
+NativeStateWayland::keyboard_handle_modifiers(void *data, struct wl_keyboard *keyboard,
+                                              uint32_t serial, uint32_t mods_depressed,
+                                              uint32_t mods_latched, uint32_t mods_locked,
+                                              uint32_t group)
+{
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-wayland.h glmark2-2021.02/src/native-state-wayland.h
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-wayland.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-wayland.h	2021-04-25 01:47:47.000000000 +0800
@@ -26,6 +26,8 @@
 #include <vector>
 #include <wayland-client.h>
 #include <wayland-egl.h>
+#include <wayland-cursor.h>
+#include "xdg-shell-client-protocol.h"
 
 #include "native-state.h"
 
@@ -47,8 +49,13 @@
     static void quit_handler(int signum);
 
     static const struct wl_registry_listener registry_listener_;
-    static const struct wl_shell_surface_listener shell_surface_listener_;
+    static const struct xdg_wm_base_listener xdg_wm_base_listener_;
+    static const struct xdg_surface_listener xdg_surface_listener_;
+    static const struct xdg_toplevel_listener xdg_toplevel_listener_;
     static const struct wl_output_listener output_listener_;
+    static const struct wl_seat_listener seat_listener_;
+    static const struct wl_pointer_listener pointer_listener_;
+    static const struct wl_keyboard_listener keyboard_listener_;
 
     static void
     registry_handle_global(void *data, struct wl_registry *registry,
@@ -76,39 +83,87 @@
     output_handle_scale(void *data, struct wl_output *wl_output, int32_t factor);
 
     static void
-    shell_surface_handle_ping(void *data, struct wl_shell_surface *shell_surface,
-                              uint32_t serial);
+    xdg_wm_base_handle_ping(void *data, struct xdg_wm_base *xdg_wm_base,
+                            uint32_t serial);
     static void
-    shell_surface_handle_configure(void *data,
-                                   struct wl_shell_surface *shell_surface,
-                                   uint32_t edges,
-                                   int32_t width, int32_t height);
-    static void
-    shell_surface_handle_popup_done(void *data,
-                                    struct wl_shell_surface *shell_surface);
-
+    xdg_toplevel_handle_configure(void *data,
+                                  struct xdg_toplevel *xdg_toplevel,
+                                  int32_t width, int32_t height,
+                                  struct wl_array *states);
+    static void
+    xdg_toplevel_handle_close(void *data, struct xdg_toplevel *xdg_toplevel);
+    static void
+    xdg_surface_handle_configure(void *data, struct xdg_surface *xdg_surface,
+                                 uint32_t serial);
+
+    static void seat_handle_capabilities(void *data,
+                                         struct wl_seat *seat, uint32_t caps);
+
+    static void pointer_handle_enter(void *data, struct wl_pointer *pointer,
+                                     uint32_t serial, struct wl_surface *surface,
+                                     wl_fixed_t sx, wl_fixed_t sy);
+
+    static void pointer_handle_leave(void *data, struct wl_pointer *pointer,
+                                     uint32_t serial, struct wl_surface *surface);
+
+    static void pointer_handle_motion(void *data, struct wl_pointer *pointer,
+                                      uint32_t time, wl_fixed_t sx, wl_fixed_t sy);
+
+    static void pointer_handle_button(void *data, struct wl_pointer *wl_pointer,
+                                      uint32_t serial, uint32_t time,
+                                      uint32_t button, uint32_t state);
+
+    static void pointer_handle_axis(void *data, struct wl_pointer *wl_pointer,
+                                    uint32_t time, uint32_t axis, wl_fixed_t value);
+
+    static void keyboard_handle_keymap(void *data, struct wl_keyboard *keyboard,
+                                       uint32_t format, int fd, uint32_t size);
+    static void keyboard_handle_enter(void *data, struct wl_keyboard *keyboard,
+                                      uint32_t serial, struct wl_surface *surface,
+                                      struct wl_array *keys);
+    static void keyboard_handle_leave(void *data, struct wl_keyboard *keyboard,
+                                      uint32_t serial, struct wl_surface *surface);
+    static void keyboard_handle_key(void *data, struct wl_keyboard *keyboard,
+                                    uint32_t serial, uint32_t time, uint32_t key,
+                                    uint32_t state);
+    static void keyboard_handle_modifiers(void *data, struct wl_keyboard *keyboard,
+                                          uint32_t serial, uint32_t mods_depressed,
+                                          uint32_t mods_latched, uint32_t mods_locked,
+                                          uint32_t group);
     struct my_output {
         wl_output *output;
         int32_t width, height;
         int32_t refresh;
+        int32_t scale;
     };
 
     typedef std::vector<struct my_output *> OutputsVector;
 
+    struct my_cursor {
+        wl_cursor_theme *cursor_theme;
+        wl_cursor *default_cursor;
+        wl_surface *cursor_surface;
+    } *cursor_;
+
     struct my_display {
         wl_display *display;
         wl_registry *registry;
         wl_compositor *compositor;
-        wl_shell *shell;
+        wl_shm *shm;
+        wl_seat *seat;
+        wl_pointer *pointer;
+        wl_keyboard *keyboard;
+        struct xdg_wm_base *xdg_wm_base;
         OutputsVector outputs;
     } *display_;
 
     struct my_window {
         WindowProperties properties;
+        bool waiting_for_configure;
         struct wl_surface *surface;
-        struct wl_region *opaque_reqion;
         struct wl_egl_window *native;
-        struct wl_shell_surface *shell_surface;
+        struct xdg_surface *xdg_surface;
+        struct xdg_toplevel *xdg_toplevel;
     } *window_;
 
     static volatile bool should_quit_;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-win32.cpp glmark2-2021.02/src/native-state-win32.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-win32.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/native-state-win32.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,248 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#include "native-state-win32.h"
+
+static const char *win_name("glmark2 " GLMARK_VERSION);
+static const char *child_win_name("glmark2 " GLMARK_VERSION " child");
+
+/******************
+ * Public methods *
+ ******************/
+
+NativeStateWin32::NativeStateWin32() : parent_window_(0), child_window_(0), native_display_(0), should_quit_(false) {}
+
+NativeStateWin32::~NativeStateWin32() { cleanup(); }
+
+bool
+NativeStateWin32::init_display()
+{
+    cleanup();
+
+    static const LONG initial_width = 128;
+    static const LONG initial_height = 128;
+
+    // Work around compile error from not defining "UNICODE" while Chromium does
+    const LPSTR idcArrow = MAKEINTRESOURCEA(32512);
+
+    WNDCLASSEXA parentWindowClass = {};
+    parentWindowClass.cbSize = sizeof(WNDCLASSEXA);
+    parentWindowClass.style = 0;
+    parentWindowClass.lpfnWndProc = wnd_proc;
+    parentWindowClass.cbClsExtra = 0;
+    parentWindowClass.cbWndExtra = 0;
+    parentWindowClass.hInstance = GetModuleHandle(nullptr);
+    parentWindowClass.hIcon = nullptr;
+    parentWindowClass.hCursor = LoadCursorA(nullptr, idcArrow);
+    parentWindowClass.hbrBackground = 0;
+    parentWindowClass.lpszMenuName = nullptr;
+    parentWindowClass.lpszClassName = win_name;
+    if (!RegisterClassExA(&parentWindowClass))
+    {
+        return false;
+    }
+
+    WNDCLASSEXA childWindowClass = {};
+    childWindowClass.cbSize = sizeof(WNDCLASSEXA);
+    childWindowClass.style = CS_OWNDC;
+    childWindowClass.lpfnWndProc = wnd_proc;
+    childWindowClass.cbClsExtra = 0;
+    childWindowClass.cbWndExtra = 0;
+    childWindowClass.hInstance = GetModuleHandle(nullptr);
+    childWindowClass.hIcon = nullptr;
+    childWindowClass.hCursor = LoadCursorA(nullptr, idcArrow);
+    childWindowClass.hbrBackground = 0;
+    childWindowClass.lpszMenuName = nullptr;
+    childWindowClass.lpszClassName = child_win_name;
+    if (!RegisterClassExA(&childWindowClass))
+    {
+        return false;
+    }
+
+    DWORD parentStyle = WS_CAPTION | WS_THICKFRAME | WS_MINIMIZEBOX | WS_MAXIMIZEBOX | WS_SYSMENU;
+    DWORD parentExtendedStyle = WS_EX_APPWINDOW;
+
+    RECT sizeRect = { 0, 0, initial_width, initial_height };
+    AdjustWindowRectEx(&sizeRect, parentStyle, FALSE, parentExtendedStyle);
+
+    parent_window_ = CreateWindowExA(parentExtendedStyle, win_name, win_name,
+        parentStyle, CW_USEDEFAULT, CW_USEDEFAULT,
+        sizeRect.right - sizeRect.left, sizeRect.bottom - sizeRect.top,
+        nullptr, nullptr, GetModuleHandle(nullptr), this);
+
+    child_window_ = CreateWindowExA(0, child_win_name, win_name, WS_CHILD, 0, 0,
+        static_cast<int>(initial_width), static_cast<int>(initial_height),
+        parent_window_, nullptr, GetModuleHandle(nullptr), this);
+
+    native_display_ = GetDC(child_window_);
+    if (!native_display_)
+    {
+        cleanup();
+    }
+
+    return native_display_;
+}
+
+void*
+NativeStateWin32::display()
+{
+    return native_display_;
+}
+
+bool
+NativeStateWin32::create_window(WindowProperties const& properties)
+{
+    properties_ = properties;
+
+    RECT windowRect;
+    if (!GetWindowRect(parent_window_, &windowRect))
+    {
+        return false;
+    }
+
+    RECT clientRect;
+    if (!GetClientRect(parent_window_, &clientRect))
+    {
+        return false;
+    }
+
+    LONG diffX = (windowRect.right - windowRect.left) - clientRect.right;
+    LONG diffY = (windowRect.bottom - windowRect.top) - clientRect.bottom;
+    if (!MoveWindow(parent_window_, windowRect.left, windowRect.top, properties.width + diffX, properties.height + diffY,
+        TRUE))
+    {
+        return false;
+    }
+
+    if (!MoveWindow(child_window_, 0, 0, properties.width, properties.height, FALSE))
+    {
+        return false;
+    }
+
+    return true;
+}
+
+void*
+NativeStateWin32::window(WindowProperties& properties)
+{
+    properties = properties_;
+    return child_window_;
+}
+
+void
+NativeStateWin32::visible(bool visible)
+{
+    int flag = (visible ? SW_SHOW : SW_HIDE);
+
+    if (parent_window_)
+    {
+        ShowWindow(parent_window_, flag);
+    }
+
+    if (child_window_)
+    {
+        ShowWindow(child_window_, flag);
+    }
+}
+
+bool
+NativeStateWin32::should_quit()
+{
+    pump_messages();
+    return should_quit_;
+}
+
+void
+NativeStateWin32::flip()
+{
+    pump_messages();
+}
+
+/*******************
+ * Private methods *
+ *******************/
+
+void
+NativeStateWin32::cleanup()
+{
+    if (native_display_)
+    {
+        ReleaseDC(child_window_, native_display_);
+        native_display_ = 0;
+    }
+
+    if (child_window_)
+    {
+        DestroyWindow(child_window_);
+        child_window_ = 0;
+    }
+
+    if (parent_window_)
+    {
+        DestroyWindow(parent_window_);
+        parent_window_ = 0;
+    }
+
+    UnregisterClassA(win_name, nullptr);
+    UnregisterClassA(child_win_name, nullptr);
+}
+
+void
+NativeStateWin32::pump_messages()
+{
+    MSG msg;
+    while (PeekMessage(&msg, nullptr, 0, 0, PM_REMOVE))
+    {
+        TranslateMessage(&msg);
+        DispatchMessage(&msg);
+    }
+}
+
+LRESULT CALLBACK
+NativeStateWin32::wnd_proc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam)
+{
+    switch (message)
+    {
+        case WM_NCCREATE:
+        {
+            LPCREATESTRUCT pCreateStruct = reinterpret_cast<LPCREATESTRUCT>(lParam);
+            SetWindowLongPtr(hWnd, GWLP_USERDATA,
+                reinterpret_cast<LONG_PTR>(pCreateStruct->lpCreateParams));
+            return DefWindowProcA(hWnd, message, wParam, lParam);
+        }
+    }
+
+    NativeStateWin32 *window = reinterpret_cast<NativeStateWin32 *>(GetWindowLongPtr(hWnd, GWLP_USERDATA));
+    if (window)
+    {
+        switch (message)
+        {
+            case WM_CLOSE:
+            case WM_DESTROY:
+                window->should_quit_ = true;
+                break;
+            default:
+                break;
+        }
+    }
+
+    return DefWindowProcA(hWnd, message, wParam, lParam);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-win32.h glmark2-2021.02/src/native-state-win32.h
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-win32.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/native-state-win32.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,56 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#ifndef GLMARK2_NATIVE_STATE_WIN32_H_
+#define GLMARK2_NATIVE_STATE_WIN32_H_
+
+#include "native-state.h"
+
+#include <windows.h>
+
+class NativeStateWin32 : public NativeState
+{
+public:
+    NativeStateWin32();
+    ~NativeStateWin32();
+
+    bool init_display();
+    void* display();
+    bool create_window(WindowProperties const& properties);
+    void* window(WindowProperties& properties);
+    void visible(bool v);
+    bool should_quit();
+    void flip();
+
+private:
+    void cleanup();
+    void pump_messages();
+
+    static LRESULT CALLBACK wnd_proc(HWND hWnd, UINT message, WPARAM wParam, LPARAM lParam);
+
+    HWND parent_window_;
+    HWND child_window_;
+    HDC native_display_;
+    WindowProperties properties_;
+    bool should_quit_;
+};
+
+#endif /* GLMARK2_NATIVE_STATE_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/native-state-x11.cpp glmark2-2021.02/src/native-state-x11.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/native-state-x11.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/native-state-x11.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -60,7 +60,7 @@
 bool
 NativeStateX11::create_window(WindowProperties const& properties)
 {
-    static const char *win_name("glmark2 "GLMARK_VERSION);
+    static const char *win_name("glmark2 " GLMARK_VERSION);
 
     if (!xdpy_) {
         Log::error("Error: X11 Display has not been initialized!\n");
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/options.cpp glmark2-2021.02/src/options.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/options.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/options.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -31,6 +31,7 @@
 std::vector<std::string> Options::benchmarks;
 std::vector<std::string> Options::benchmark_files;
 bool Options::validate = false;
+std::string Options::data_path = std::string(GLMARK_DATA_PATH);
 Options::FrameEnd Options::frame_end = Options::FrameEndDefault;
 std::pair<int,int> Options::size(800, 600);
 bool Options::list_scenes = false;
@@ -48,6 +49,7 @@
     {"benchmark", 1, 0, 0},
     {"benchmark-file", 1, 0, 0},
     {"validate", 0, 0, 0},
+    {"data-path", 1, 0, 0},
     {"frame-end", 1, 0, 0},
     {"off-screen", 0, 0, 0},
     {"visual-config", 1, 0, 0},
@@ -116,13 +118,15 @@
     printf("A benchmark for Open GL (ES) 2.0\n"
            "\n"
            "Options:\n"
-           "  -b, --benchmark BENCH  A benchmark to run: 'scene(:opt1=val1)*'\n"
+           "  -b, --benchmark BENCH  A benchmark or options to run: '(scene)?(:opt1=val1)*'\n"
            "                         (the option can be used multiple times)\n"
            "  -f, --benchmark-file F Load benchmarks to run from a file containing a\n"
            "                         list of benchmark descriptions (one per line)\n"
            "                         (the option can be used multiple times)\n"
            "      --validate         Run a quick output validation test instead of \n"
            "                         running the benchmarks\n"
+           "      --data-path PATH   Path to glmark2 models, shaders and textures\n"
+           "                         Default: " GLMARK_DATA_PATH "\n"
            "      --frame-end METHOD How to end a frame [default,none,swap,finish,readpixels]\n"
            "      --off-screen       Render to an off-screen surface\n"
            "      --visual-config C  The visual configuration to use for the rendering\n"
@@ -171,6 +175,8 @@
             Options::benchmark_files.push_back(optarg);
         else if (!strcmp(optname, "validate"))
             Options::validate = true;
+        else if (!strcmp(optname, "data-path"))
+            Options::data_path = std::string(optarg);
         else if (!strcmp(optname, "frame-end"))
             Options::frame_end = frame_end_from_str(optarg);
         else if (!strcmp(optname, "off-screen"))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/options.h glmark2-2021.02/src/options.h
--- glmark2-2014.03+git20150611.fa71af2d/src/options.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/options.h	2021-04-25 01:47:22.000000000 +0800
@@ -43,6 +43,7 @@
     static std::vector<std::string> benchmarks;
     static std::vector<std::string> benchmark_files;
     static bool validate;
+    static std::string data_path;
     static FrameEnd frame_end;
     static std::pair<int,int> size;
     static bool list_scenes;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-buffer.cpp glmark2-2021.02/src/scene-buffer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-buffer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-buffer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "util.h"
@@ -244,9 +245,9 @@
     {
         /* Set up shaders */
         static const std::string vtx_shader_filename(
-                GLMARK_DATA_PATH"/shaders/buffer-wireframe.vert");
+                Options::data_path + "/shaders/buffer-wireframe.vert");
         static const std::string frg_shader_filename(
-                GLMARK_DATA_PATH"/shaders/buffer-wireframe.frag");
+                Options::data_path + "/shaders/buffer-wireframe.frag");
 
         ShaderSource vtx_source(vtx_shader_filename);
         ShaderSource frg_source(frg_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-build.cpp glmark2-2021.02/src/scene-build.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-build.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-build.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -25,6 +25,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "model.h"
@@ -89,8 +90,8 @@
         return false;
 
     /* Set up shaders */
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/light-basic.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     static const LibMatrix::vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-bump.cpp glmark2-2021.02/src/scene-bump.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-bump.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-bump.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,6 +23,7 @@
 #include "scene.h"
 #include "log.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "shader-source.h"
 #include "model.h"
@@ -61,8 +62,8 @@
 bool
 SceneBump::setup_model_plain(const std::string &type)
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-poly.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-poly.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-poly.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-poly.frag");
     static const std::string low_poly_filename("asteroid-low");
     static const std::string high_poly_filename("asteroid-high");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
@@ -115,8 +116,8 @@
 bool
 SceneBump::setup_model_normals()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-normals.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-normals.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
@@ -170,8 +171,8 @@
 bool
 SceneBump::setup_model_normals_tangent()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals-tangent.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-normals-tangent.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-normals-tangent.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-normals-tangent.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
@@ -228,8 +229,8 @@
 bool
 SceneBump::setup_model_height()
 {
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/bump-height.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/bump-height.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/bump-height.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/bump-height.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     Model model;
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-clear.cpp glmark2-2021.02/src/scene-clear.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-clear.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-clear.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2013 Linaro Limited
+// Copyright © 2013 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-collection.h glmark2-2021.02/src/scene-collection.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-collection.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-collection.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2013 Linaro Limited
+// Copyright © 2013 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-conditionals.cpp glmark2-2021.02/src/scene-conditionals.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-conditionals.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-conditionals.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,6 +21,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -29,7 +30,7 @@
 
 #include <cmath>
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/conditionals");
+static const std::string shader_file_base("/shaders/conditionals");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -56,14 +57,14 @@
 static std::string
 get_vertex_shader_source(int steps, bool conditionals)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
 
     for (int i = 0; i < steps; i++) {
         if (conditionals)
-            source_main.append_file(step_conditional_file);
+            source_main.append_file(Options::data_path + step_conditional_file);
         else
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
@@ -74,14 +75,14 @@
 static std::string
 get_fragment_shader_source(int steps, bool conditionals)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
 
     for (int i = 0; i < steps; i++) {
         if (conditionals)
-            source_main.append_file(step_conditional_file);
+            source_main.append_file(Options::data_path + step_conditional_file);
         else
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene.cpp glmark2-2021.02/src/scene.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -43,10 +43,12 @@
 Scene::Scene(Canvas &pCanvas, const string &name) :
     canvas_(pCanvas), name_(name),
     startTime_(0), lastUpdateTime_(0), currentFrame_(0),
-    running_(0), duration_(0)
+    running_(0), duration_(0), nframes_(0)
 {
     options_["duration"] = Scene::Option("duration", "10.0",
                                          "The duration of each benchmark in seconds");
+    options_["nframes"] = Scene::Option("nframes", "",
+                                         "The number of frames to render");
     options_["vertex-precision"] = Scene::Option("vertex-precision",
                                                  "default,default,default,default",
                                                  "The precision values for the vertex shader (\"int,float,sampler2d,samplercube\")");
@@ -96,6 +98,7 @@
 Scene::setup()
 {
     duration_ = Util::fromString<double>(options_["duration"].value);
+    nframes_ = Util::fromString<unsigned>(options_["nframes"].value);
 
     ShaderSource::default_precision(
             ShaderSource::Precision(options_["vertex-precision"].value),
@@ -132,6 +135,9 @@
 
     if (elapsed_time >= duration_)
         running_ = false;
+
+    if (nframes_ > 0 && currentFrame_ >= nframes_)
+        running_ = false;
 }
 
 void
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-desktop.cpp glmark2-2021.02/src/scene-desktop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-desktop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-desktop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -20,11 +20,13 @@
  *  Alexandros Frantzis (glmark2)
  *  Jesse Barker (glmark2)
  */
+#include <algorithm>
 #include <cmath>
 #include <cstdlib>
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -43,8 +45,8 @@
 create_blur_shaders(ShaderSource& vtx_source, ShaderSource& frg_source,
                     unsigned int radius, float sigma, BlurDirection direction)
 {
-    vtx_source.append_file(GLMARK_DATA_PATH"/shaders/desktop.vert");
-    frg_source.append_file(GLMARK_DATA_PATH"/shaders/desktop-blur.frag");
+    vtx_source.append_file(Options::data_path + "/shaders/desktop.vert");
+    frg_source.append_file(Options::data_path + "/shaders/desktop-blur.frag");
 
     /* Don't let the gaussian curve become too narrow */
     if (sigma < 1.0)
@@ -122,6 +124,9 @@
         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
         glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
 
+        GLint prev_fbo;
+        glGetIntegerv(GL_FRAMEBUFFER_BINDING, &prev_fbo);
+
         /* Create a FBO */
         glGenFramebuffers(1, &fbo_);
         glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
@@ -144,12 +149,12 @@
             }
         }
 
-        glBindFramebuffer(GL_FRAMEBUFFER, 0);
+        glBindFramebuffer(GL_FRAMEBUFFER, prev_fbo);
 
         /* Load the shader program when this class if first used */
         if (RenderObject::use_count == 0) {
-            ShaderSource vtx_source(GLMARK_DATA_PATH"/shaders/desktop.vert");
-            ShaderSource frg_source(GLMARK_DATA_PATH"/shaders/desktop.frag");
+            ShaderSource vtx_source(Options::data_path + "/shaders/desktop.vert");
+            ShaderSource frg_source(Options::data_path + "/shaders/desktop.frag");
             Scene::load_shaders_from_strings(main_program, vtx_source.str(),
                                              frg_source.str());
         }
@@ -196,6 +201,8 @@
         /* Recreate the backing texture with correct size */
         if (size_.x() != size.x() || size_.y() != size.y()) {
             size_ = size;
+            if (fbo_ == 0)
+                return;
             /* If we're resizing the texture, we need to tell the framebuffer*/
             glBindTexture(GL_TEXTURE_2D, texture_);
             glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, size_.x(), size_.y(), 0,
@@ -377,9 +384,11 @@
  */
 class RenderScreen : public RenderObject
 {
+    Canvas &canvas_;
 public:
-    RenderScreen(Canvas &canvas) { fbo_ = canvas.fbo(); }
-    virtual void init() {}
+    RenderScreen(Canvas &canvas) : canvas_(canvas) {}
+    virtual void init() { fbo_ = canvas_.fbo(); }
+    virtual void size(const LibMatrix::vec2& size) { size_ = size; clear(); }
     virtual void release() {}
 };
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-effect-2d.cpp glmark2-2021.02/src/scene-effect-2d.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-effect-2d.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-effect-2d.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,6 +26,7 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -96,7 +97,7 @@
 create_convolution_fragment_shader(Canvas &canvas, std::vector<float> &array,
                                    unsigned int width, unsigned int height)
 {
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/effect-2d-convolution.frag");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/effect-2d-convolution.frag");
     ShaderSource source(frg_shader_filename);
 
     if (width * height != array.size()) {
@@ -303,7 +304,7 @@
 
     Texture::find_textures();
 
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/effect-2d.vert");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/effect-2d.vert");
 
     std::vector<float> kernel;
     unsigned int kernel_width = 0;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-function.cpp glmark2-2021.02/src/scene-function.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-function.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-function.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,13 +23,14 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
 #include "shader-source.h"
 #include "util.h"
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/function");
+static const std::string shader_file_base("/shaders/function");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -61,7 +62,7 @@
 static std::string
 get_vertex_shader_source(int steps, bool function, std::string &complexity)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
     std::string step_file;
 
@@ -72,13 +73,13 @@
 
     for (int i = 0; i < steps; i++) {
         if (function)
-            source_main.append_file(call_file);
+            source_main.append_file(Options::data_path + call_file);
         else
-            source_main.append_file(step_file);
+            source_main.append_file(Options::data_path + step_file);
     }
 
     if (function)
-        source.replace_with_file("$PROCESS$", step_file);
+        source.replace_with_file("$PROCESS$", Options::data_path + step_file);
     else
         source.replace("$PROCESS$", "");
 
@@ -90,7 +91,7 @@
 static std::string
 get_fragment_shader_source(int steps, bool function, std::string &complexity)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
     std::string step_file;
 
@@ -101,13 +102,13 @@
 
     for (int i = 0; i < steps; i++) {
         if (function)
-            source_main.append_file(call_file);
+            source_main.append_file(Options::data_path + call_file);
         else
-            source_main.append_file(step_file);
+            source_main.append_file(Options::data_path + step_file);
     }
 
     if (function)
-        source.replace_with_file("$PROCESS$", step_file);
+        source.replace_with_file("$PROCESS$", Options::data_path + step_file);
     else
         source.replace("$PROCESS$", "");
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene.h glmark2-2021.02/src/scene.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene.h	2021-04-25 01:47:22.000000000 +0800
@@ -234,6 +234,7 @@
     unsigned currentFrame_;
     bool running_;
     double duration_;      // Duration of run in seconds
+    unsigned nframes_;
 };
 
 /*
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/a.cc glmark2-2021.02/src/scene-ideas/a.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/a.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/a.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'a'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/characters.h glmark2-2021.02/src/scene-ideas/characters.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/characters.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/characters.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/d.cc glmark2-2021.02/src/scene-ideas/d.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/d.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/d.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'd'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/e.cc glmark2-2021.02/src/scene-ideas/e.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/e.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/e.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'e'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/i.cc glmark2-2021.02/src/scene-ideas/i.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/i.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/i.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'i'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/lamp.cc glmark2-2021.02/src/scene-ideas/lamp.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/lamp.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/lamp.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the lamp
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -23,6 +23,7 @@
  *  Jesse Barker
  */
 #include "lamp.h"
+#include "options.h"
 #include "shader-source.h"
 #include "log.h"
 #include "scene.h"
@@ -170,8 +171,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // The program for handling lighting...
-    string lit_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-lit.vert");
-    string lit_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-lit.frag");
+    string lit_vtx_filename(Options::data_path + "/shaders/ideas-lamp-lit.vert");
+    string lit_frg_filename(Options::data_path + "/shaders/ideas-lamp-lit.frag");
     ShaderSource lit_vtx_source(lit_vtx_filename);
     ShaderSource lit_frg_source(lit_frg_filename);
     if (!Scene::load_shaders_from_strings(litProgram_, lit_vtx_source.str(),
@@ -182,8 +183,8 @@
     }
 
     // The simple program with no lighting...
-    string unlit_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-unlit.vert");
-    string unlit_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-lamp-unlit.frag");
+    string unlit_vtx_filename(Options::data_path + "/shaders/ideas-lamp-unlit.vert");
+    string unlit_frg_filename(Options::data_path + "/shaders/ideas-lamp-unlit.frag");
     ShaderSource unlit_vtx_source(unlit_vtx_filename);
     ShaderSource unlit_frg_source(unlit_frg_filename);
     if (!Scene::load_shaders_from_strings(unlitProgram_, unlit_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/lamp.h glmark2-2021.02/src/scene-ideas/lamp.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/lamp.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/lamp.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/logo.cc glmark2-2021.02/src/scene-ideas/logo.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/logo.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/logo.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the old Silicon Graphics logo
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -23,6 +23,7 @@
  *  Jesse Barker
  */
 #include "logo.h"
+#include "options.h"
 #include "scene.h"
 #include "shader-source.h"
 #include "log.h"
@@ -423,8 +424,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // The program for handling the main object with lighting...
-    string logo_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo.vert");
-    string logo_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo.frag");
+    string logo_vtx_filename(Options::data_path + "/shaders/ideas-logo.vert");
+    string logo_frg_filename(Options::data_path + "/shaders/ideas-logo.frag");
     ShaderSource logo_vtx_source(logo_vtx_filename);
     ShaderSource logo_frg_source(logo_frg_filename);
     if (!Scene::load_shaders_from_strings(normalProgram_, logo_vtx_source.str(),
@@ -437,8 +438,8 @@
     normalNormalIndex_ = normalProgram_[normalAttribName_].location();
 
     // The program for handling the flat object...
-    string logo_flat_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-flat.vert");
-    string logo_flat_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-flat.frag");
+    string logo_flat_vtx_filename(Options::data_path + "/shaders/ideas-logo-flat.vert");
+    string logo_flat_frg_filename(Options::data_path + "/shaders/ideas-logo-flat.frag");
     ShaderSource logo_flat_vtx_source(logo_flat_vtx_filename);
     ShaderSource logo_flat_frg_source(logo_flat_frg_filename);
     if (!Scene::load_shaders_from_strings(flatProgram_, logo_flat_vtx_source.str(),
@@ -450,8 +451,8 @@
     flatVertexIndex_ = flatProgram_[vertexAttribName_].location();
 
     // The program for handling the shadow object with texturing...
-    string logo_shadow_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-shadow.vert");
-    string logo_shadow_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-logo-shadow.frag");
+    string logo_shadow_vtx_filename(Options::data_path + "/shaders/ideas-logo-shadow.vert");
+    string logo_shadow_frg_filename(Options::data_path + "/shaders/ideas-logo-shadow.frag");
     ShaderSource logo_shadow_vtx_source(logo_shadow_vtx_filename);
     ShaderSource logo_shadow_frg_source(logo_shadow_frg_filename);
     if (!Scene::load_shaders_from_strings(shadowProgram_, logo_shadow_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/logo.h glmark2-2021.02/src/scene-ideas/logo.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/logo.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/logo.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/m.cc glmark2-2021.02/src/scene-ideas/m.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/m.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/m.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'm'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/n.cc glmark2-2021.02/src/scene-ideas/n.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/n.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/n.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'o'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/o.cc glmark2-2021.02/src/scene-ideas/o.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/o.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/o.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 'o'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/s.cc glmark2-2021.02/src/scene-ideas/s.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/s.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/s.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 's'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/splines.cc glmark2-2021.02/src/scene-ideas/splines.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/splines.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/splines.cc	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/splines.h glmark2-2021.02/src/scene-ideas/splines.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/splines.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/splines.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/table.cc glmark2-2021.02/src/scene-ideas/table.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/table.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/table.cc	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -20,6 +20,7 @@
  * Authors:
  *  Jesse Barker
  */
+#include "options.h"
 #include "table.h"
 #include "scene.h"
 #include "shader-source.h"
@@ -101,8 +102,8 @@
 
     // Initialize shader sources from input files and create programs from them
     // Program to render the table with lighting and a time-based fade...
-    string table_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-table.vert");
-    string table_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-table.frag");
+    string table_vtx_filename(Options::data_path + "/shaders/ideas-table.vert");
+    string table_frg_filename(Options::data_path + "/shaders/ideas-table.frag");
     ShaderSource table_vtx_source(table_vtx_filename);
     ShaderSource table_frg_source(table_frg_filename);
     if (!Scene::load_shaders_from_strings(tableProgram_, table_vtx_source.str(),
@@ -111,11 +112,11 @@
         Log::error("No valid program for table rendering.\n");
         return;
     }
-    textVertexIndex_ = tableProgram_[vertexAttribName_].location();
+    tableVertexIndex_ = tableProgram_[vertexAttribName_].location();
 
     // Program to render the paper with lighting and a time-based fade...
-    string paper_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-paper.vert");
-    string paper_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-paper.frag");
+    string paper_vtx_filename(Options::data_path + "/shaders/ideas-paper.vert");
+    string paper_frg_filename(Options::data_path + "/shaders/ideas-paper.frag");
     ShaderSource paper_vtx_source(paper_vtx_filename);
     ShaderSource paper_frg_source(paper_frg_filename);
     if (!Scene::load_shaders_from_strings(paperProgram_, paper_vtx_source.str(),
@@ -127,8 +128,8 @@
     paperVertexIndex_ = paperProgram_[vertexAttribName_].location();
 
     // Program to handle the text (time-based color fade)...
-    string text_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-text.vert");
-    string text_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-text.frag");
+    string text_vtx_filename(Options::data_path + "/shaders/ideas-text.vert");
+    string text_frg_filename(Options::data_path + "/shaders/ideas-text.frag");
     ShaderSource text_vtx_source(text_vtx_filename);
     ShaderSource text_frg_source(text_frg_filename);
     if (!Scene::load_shaders_from_strings(textProgram_, text_vtx_source.str(),
@@ -140,8 +141,8 @@
     textVertexIndex_ = textProgram_[vertexAttribName_].location();
 
     // Program for the drawUnder functionality (just paint it black)...
-    string under_table_vtx_filename(GLMARK_DATA_PATH"/shaders/ideas-under-table.vert");
-    string under_table_frg_filename(GLMARK_DATA_PATH"/shaders/ideas-under-table.frag");
+    string under_table_vtx_filename(Options::data_path + "/shaders/ideas-under-table.vert");
+    string under_table_frg_filename(Options::data_path + "/shaders/ideas-under-table.frag");
     ShaderSource under_table_vtx_source(under_table_vtx_filename);
     ShaderSource under_table_frg_source(under_table_frg_filename);
     if (!Scene::load_shaders_from_strings(underProgram_, under_table_vtx_source.str(),
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/table.h glmark2-2021.02/src/scene-ideas/table.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/table.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/table.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/t.cc glmark2-2021.02/src/scene-ideas/t.cc
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas/t.cc	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas/t.cc	2021-04-25 01:47:22.000000000 +0800
@@ -2,7 +2,7 @@
  * Vertex position data describing the letter 't'
  *
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas.cpp glmark2-2021.02/src/scene-ideas.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-ideas.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-ideas.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 /*
  * (c) Copyright 1993, Silicon Graphics, Inc.
- * Copyright © 2012 Linaro Limited
+ * Copyright © 2012 Linaro Limited
  *
  * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
  *
@@ -48,7 +48,7 @@
         timeOffset_(START_TIME_)
     {
         startTime_.tv_sec = 0;
-        startTime_.tv_usec = 0; 
+        startTime_.tv_nsec = 0; 
     }
     ~SceneIdeasPrivate()
     {
@@ -69,7 +69,7 @@
     float currentSpeed_;
     float currentTime_;
     float timeOffset_;
-    struct timeval startTime_;
+    struct timespec startTime_;
     static const float CYCLE_TIME_;
     static const float TIME_;
     static const float START_TIME_;
@@ -163,17 +163,17 @@
 SceneIdeasPrivate::reset_time()
 {
     timeOffset_ = START_TIME_;
-    gettimeofday(&startTime_, NULL);
+    clock_gettime(CLOCK_MONOTONIC, &startTime_);
 }
 
 void
 SceneIdeasPrivate::update_time()
 {
     // Compute new time
-    struct timeval current = {0, 0};
-    gettimeofday(&current, NULL);
+    struct timespec current = {0, 0};
+    clock_gettime(CLOCK_MONOTONIC, &current);
     float timediff = (current.tv_sec - startTime_.tv_sec) + 
-        static_cast<double>(current.tv_usec - startTime_.tv_usec) / 1000000.0;
+        static_cast<double>(current.tv_nsec - startTime_.tv_nsec) / 1000000000.0;
     float sceneTime = timediff * currentSpeed_ + timeOffset_;
 
     // Keep the current time in [START_TIME_..CYCLE_TIME_)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-jellyfish.cpp glmark2-2021.02/src/scene-jellyfish.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-jellyfish.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-jellyfish.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -24,6 +24,7 @@
 #include <fstream>
 #include <memory>
 #include <iomanip>
+#include "options.h"
 #include "scene.h"
 #include "scene-jellyfish.h"
 #include "log.h"
@@ -115,8 +116,8 @@
 GradientRenderer::init()
 {
     // Program set up
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/gradient.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/gradient.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/gradient.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/gradient.frag");
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
     if (!Scene::load_shaders_from_strings(program_, vtx_source.str(),
@@ -272,7 +273,7 @@
 {
     Log::debug("Loading model from file '%s'\n", filename.c_str());
 
-    const std::auto_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
+    const std::unique_ptr<std::istream> input_file_ptr(Util::get_resource(filename));
     std::istream& inputFile(*input_file_ptr);
     if (!inputFile)
     {
@@ -375,7 +376,7 @@
 bool
 JellyfishPrivate::initialize()
 {
-    static const string modelFilename(GLMARK_DATA_PATH"/models/jellyfish.jobj");
+    static const string modelFilename(Options::data_path + "/models/jellyfish.jobj");
     if (!load_obj(modelFilename))
     {
         return false;
@@ -410,8 +411,8 @@
     // Set up program first so we can store attribute and uniform locations
     // away for the 
     using std::string;
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/jellyfish.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/jellyfish.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/jellyfish.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/jellyfish.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-jellyfish.h glmark2-2021.02/src/scene-jellyfish.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-jellyfish.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-jellyfish.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-loop.cpp glmark2-2021.02/src/scene-loop.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-loop.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-loop.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,13 +23,14 @@
 
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
 #include "shader-source.h"
 #include "util.h"
 
-static const std::string shader_file_base(GLMARK_DATA_PATH"/shaders/loop");
+static const std::string shader_file_base("/shaders/loop");
 
 static const std::string vtx_file(shader_file_base + ".vert");
 static const std::string frg_file(shader_file_base + ".frag");
@@ -62,11 +63,11 @@
 static std::string
 get_fragment_shader_source(int steps, bool loop, bool uniform)
 {
-    ShaderSource source(frg_file);
+    ShaderSource source(Options::data_path + frg_file);
     ShaderSource source_main;
 
     if (loop) {
-        source_main.append_file(step_loop_file);
+        source_main.append_file(Options::data_path + step_loop_file);
         if (uniform) {
             source_main.replace("$NLOOPS$", "FragmentLoops");
         }
@@ -76,7 +77,7 @@
     }
     else {
         for (int i = 0; i < steps; i++)
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
@@ -87,11 +88,11 @@
 static std::string
 get_vertex_shader_source(int steps, bool loop, bool uniform)
 {
-    ShaderSource source(vtx_file);
+    ShaderSource source(Options::data_path + vtx_file);
     ShaderSource source_main;
 
     if (loop) {
-        source_main.append_file(step_loop_file);
+        source_main.append_file(Options::data_path + step_loop_file);
         if (uniform) {
             source_main.replace("$NLOOPS$", "VertexLoops");
         }
@@ -101,7 +102,7 @@
     }
     else {
         for (int i = 0; i < steps; i++)
-            source_main.append_file(step_simple_file);
+            source_main.append_file(Options::data_path + step_simple_file);
     }
 
     source.replace("$MAIN$", source_main.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-pulsar.cpp glmark2-2021.02/src/scene-pulsar.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-pulsar.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-pulsar.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -26,6 +26,7 @@
 #include <stdlib.h>
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -112,19 +113,19 @@
     std::string frg_shader_filename;
     static const vec4 lightPosition(-20.0f, 20.0f,-20.0f, 1.0f);
     if (options_["light"].value == "true") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/pulsar-light.vert";
+        vtx_shader_filename = Options::data_path + "/shaders/pulsar-light.vert";
     } else {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/pulsar.vert";
+        vtx_shader_filename = Options::data_path + "/shaders/pulsar.vert";
     }
 
     if (options_["texture"].value == "true") {
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic-tex.frag";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic-tex.frag";
         Texture::find_textures();
         if (!Texture::load("crate-base", &texture_, GL_NEAREST, GL_NEAREST, 0))
             return false;
 
     } else {
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.frag";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic.frag";
     }
 
     ShaderSource vtx_source(vtx_shader_filename);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-refract.cpp glmark2-2021.02/src/scene-refract.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-refract.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-refract.cpp	2021-04-25 01:47:47.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -21,6 +21,7 @@
 //
 #include "scene-refract.h"
 #include "model.h"
+#include "options.h"
 #include "texture.h"
 #include "util.h"
 #include "log.h"
@@ -176,10 +177,10 @@
 //
 
 bool
-DistanceRenderTarget::setup(unsigned int width, unsigned int height)
+DistanceRenderTarget::setup(unsigned int canvas_fbo, unsigned int width, unsigned int height)
 {
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/depth.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/depth.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/depth.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/depth.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
@@ -192,6 +193,7 @@
     canvas_height_ = height;
     width_ = canvas_width_ * 2;
     height_ = canvas_height_ * 2;
+    canvas_fbo_ = canvas_fbo;
 
     // If the texture will be too large for the implemnetation, we need to
     // clamp the dimensions but maintain the aspect ratio.
@@ -235,7 +237,7 @@
         Log::error("DistanceRenderTarget::setup: glCheckFramebufferStatus failed (0x%x)\n", status);
         return false;
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
 
     return true;
 }
@@ -272,7 +274,7 @@
 
 void DistanceRenderTarget::disable()
 {
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
     glViewport(0, 0, canvas_width_, canvas_height_);
     glCullFace(GL_BACK);
 }
@@ -281,8 +283,8 @@
 RefractPrivate::setup(map<string, Scene::Option>& options)
 {
     // Program object setup
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-refract.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-refract.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/light-refract.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/light-refract.frag");
     static const vec4 lightColor(0.4, 0.4, 0.4, 1.0);
 
     ShaderSource vtx_source(vtx_shader_filename);
@@ -382,7 +384,7 @@
                                       0.0, 0.0, 0.0,
                                       0.0, 1.0, 0.0);
 
-    if (!depthTarget_.setup(canvas_.width(), canvas_.height())) {
+    if (!depthTarget_.setup(canvas_.fbo(), canvas_.width(), canvas_.height())) {
         Log::error("Failed to set up the render target for the depth pass\n");
         return false;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-refract.h glmark2-2021.02/src/scene-refract.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-refract.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-refract.h	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -46,6 +46,7 @@
     unsigned int height_;
     unsigned int tex_[2];
     unsigned int fbo_;
+    unsigned int canvas_fbo_;
 public:
     DistanceRenderTarget() :
         canvas_width_(0),
@@ -57,7 +58,7 @@
         tex_[DEPTH] = tex_[COLOR] = 0;
     }
     ~DistanceRenderTarget() {}
-    bool setup(unsigned int width, unsigned int height);
+    bool setup(unsigned int canvas_fbo, unsigned int width, unsigned int height);
     void teardown();
     void enable(const LibMatrix::mat4& mvp);
     void disable();
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-shading.cpp glmark2-2021.02/src/scene-shading.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-shading.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-shading.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,6 +24,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -154,24 +155,24 @@
     ShaderSource vtx_source;
     ShaderSource frg_source;
     if (shading == "gouraud") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-basic.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-basic.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-basic.frag";
         frg_source.append_file(frg_shader_filename);
         vtx_source.append_file(vtx_shader_filename);
         vtx_source.add_const("LightSourcePosition", lightPosition);
         vtx_source.add_const("MaterialDiffuse", materialDiffuse);
     }
     else if (shading == "blinn-phong-inf") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-advanced.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-advanced.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-advanced.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-advanced.frag";
         frg_source.append_file(frg_shader_filename);
         frg_source.add_const("LightSourcePosition", lightPosition);
         frg_source.add_const("LightSourceHalfVector", halfVector);
         vtx_source.append_file(vtx_shader_filename);
     }
     else if (shading == "phong") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-phong.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-phong.frag";
         unsigned int num_lights = Util::fromString<unsigned int>(options_["num-lights"].value);
         string fragsource = get_fragment_shader_source(frg_shader_filename, num_lights);
         frg_source.append(fragsource);
@@ -179,8 +180,8 @@
         vtx_source.append_file(vtx_shader_filename);
     }
     else if (shading == "cel") {
-        vtx_shader_filename = GLMARK_DATA_PATH"/shaders/light-phong.vert";
-        frg_shader_filename = GLMARK_DATA_PATH"/shaders/light-cel.frag";
+        vtx_shader_filename = Options::data_path + "/shaders/light-phong.vert";
+        frg_shader_filename = Options::data_path + "/shaders/light-cel.frag";
         vtx_source.append_file(vtx_shader_filename);
         frg_source.append_file(frg_shader_filename);
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-shadow.cpp glmark2-2021.02/src/scene-shadow.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-shadow.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-shadow.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -1,5 +1,5 @@
 //
-// Copyright © 2012 Linaro Limited
+// Copyright © 2012 Linaro Limited
 //
 // This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
 //
@@ -21,6 +21,7 @@
 //
 #include "scene.h"
 #include "model.h"
+#include "options.h"
 #include "util.h"
 #include "log.h"
 #include "shader-source.h"
@@ -53,6 +54,7 @@
     unsigned int height_;
     unsigned int tex_;
     unsigned int fbo_;
+    unsigned int canvas_fbo_;
 public:
     DepthRenderTarget() :
         canvas_width_(0),
@@ -62,7 +64,7 @@
         tex_(0),
         fbo_(0) {}
     ~DepthRenderTarget() {}
-    bool setup(unsigned int width, unsigned int height);
+    bool setup(unsigned int canvas_fbo, unsigned int width, unsigned int height);
     void teardown();
     void enable(const mat4& mvp);
     void disable();
@@ -71,10 +73,10 @@
 };
 
 bool
-DepthRenderTarget::setup(unsigned int width, unsigned int height)
+DepthRenderTarget::setup(unsigned int canvas_fbo, unsigned int width, unsigned int height)
 {
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/depth.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/depth.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/depth.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/depth.frag");
 
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
@@ -85,6 +87,7 @@
 
     canvas_width_ = width;
     canvas_height_ = height;
+    canvas_fbo_ = canvas_fbo;
     width_ = canvas_width_ * 2;
     height_ = canvas_height_ * 2;
 
@@ -120,7 +123,7 @@
         Log::error("DepthRenderTarget::setup: glCheckFramebufferStatus failed (0x%x)\n", status);
         return false;
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
 
     return true;
 }
@@ -155,7 +158,7 @@
 
 void DepthRenderTarget::disable()
 {
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_fbo_);
     glViewport(0, 0, canvas_width_, canvas_height_);
     glColorMask(GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
 }
@@ -201,8 +204,8 @@
 
     // Program set up
     static const vec4 materialDiffuse(0.3f, 0.3f, 0.3f, 1.0f);
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/shadow.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/shadow.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/shadow.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/shadow.frag");
     ShaderSource vtx_source(vtx_shader_filename);
     ShaderSource frg_source(frg_shader_filename);
 
@@ -314,8 +317,8 @@
 ShadowPrivate::setup(map<string, Scene::Option>& options)
 {
     // Program object setup
-    static const string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.frag");
+    static const string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const string frg_shader_filename(Options::data_path + "/shaders/light-basic.frag");
     static const vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
     ShaderSource vtx_source(vtx_shader_filename);
@@ -372,7 +375,7 @@
     float aspect(static_cast<float>(canvas_.width())/static_cast<float>(canvas_.height()));
     projection_.perspective(fovy, aspect, 2.0, 50.0);
 
-    if (!depthTarget_.setup(canvas_.width(), canvas_.height())) {
+    if (!depthTarget_.setup(canvas_.fbo(), canvas_.width(), canvas_.height())) {
         Log::error("Failed to set up the render target for the depth pass\n");
         return false;
     }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/base-renderer.cpp glmark2-2021.02/src/scene-terrain/base-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/base-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/base-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,26 +21,33 @@
  */
 #include "renderer.h"
 
-BaseRenderer::BaseRenderer(const LibMatrix::vec2 &size) :
+BaseRenderer::BaseRenderer() :
     texture_(0), input_texture_(0), fbo_(0), depth_renderbuffer_(0),
-    min_filter_(GL_LINEAR), mag_filter_(GL_LINEAR),
+    owns_fbo_(false), min_filter_(GL_LINEAR), mag_filter_(GL_LINEAR),
     wrap_s_(GL_CLAMP_TO_EDGE), wrap_t_(GL_CLAMP_TO_EDGE)
 {
-    setup(size, true, true);
 }
 
 BaseRenderer::~BaseRenderer()
 {
     glDeleteTextures(1, &texture_);
     glDeleteRenderbuffers(1, &depth_renderbuffer_);
-    glDeleteFramebuffers(1, &fbo_);
+    if (owns_fbo_)
+        glDeleteFramebuffers(1, &fbo_);
+}
+
+void
+BaseRenderer::setup_onscreen(Canvas& canvas)
+{
+    size_ = LibMatrix::vec2(canvas.width(), canvas.height());
+    recreate(&canvas, false);
 }
 
 void
-BaseRenderer::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+BaseRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     size_ = size;
-    recreate(onscreen, has_depth);
+    recreate(nullptr, has_depth);
 }
 
 void
@@ -59,7 +66,7 @@
 {
     glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
     glViewport(0, 0, size_.x(), size_.y());
-    if (!fbo_ || depth_renderbuffer_) {
+    if (!owns_fbo_ || depth_renderbuffer_) {
         glEnable(GL_DEPTH_TEST);
         glDepthMask(GL_TRUE);
     }
@@ -79,21 +86,26 @@
 }
 
 void
-BaseRenderer::recreate(bool onscreen, bool has_depth)
+BaseRenderer::recreate(Canvas* canvas, bool has_depth)
 {
     if (texture_) {
         glDeleteTextures(1, &texture_);
         texture_ = 0;
     }
-    if (fbo_) {
+    if (owns_fbo_ && fbo_) {
         glDeleteRenderbuffers(1, &depth_renderbuffer_);
         depth_renderbuffer_ = 0;
         glDeleteFramebuffers(1, &fbo_);
         fbo_ = 0;
     }
-    if (!onscreen) {
+
+    if (canvas) {
+        fbo_ = canvas->fbo();
+        owns_fbo_ = false;
+    } else {
         create_texture();
         create_fbo(has_depth);
+        owns_fbo_ = true;
     }
 }
 
@@ -131,6 +143,9 @@
                 size_.x(), size_.y());
     }
 
+    GLint prev_fbo;
+    glGetIntegerv(GL_FRAMEBUFFER_BINDING, &prev_fbo);
+
     /* Create the FBO and attach the texture and the renderebuffer */
     glGenFramebuffers(1, &fbo_);
     glBindFramebuffer(GL_FRAMEBUFFER, fbo_);
@@ -140,5 +155,6 @@
         glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT,
                 GL_RENDERBUFFER, depth_renderbuffer_);
     }
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+
+    glBindFramebuffer(GL_FRAMEBUFFER, prev_fbo);
 }
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/blur-renderer.cpp glmark2-2021.02/src/scene-terrain/blur-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/blur-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/blur-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,18 +19,21 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
+#include <cstdlib>
+
 void
 create_blur_shaders(ShaderSource& vtx_source, ShaderSource& frg_source,
                     unsigned int radius, float sigma, BlurRenderer::BlurDirection direction,
                     float tilt_shift);
 
-BlurRenderer::BlurRenderer(const LibMatrix::vec2 &size, int radius, float sigma,
+BlurRenderer::BlurRenderer(int radius, float sigma,
                            BlurDirection dir, const LibMatrix::vec2 &step, float tilt_shift) :
-        TextureRenderer(size, *blur_program(true, radius, sigma, dir, step, tilt_shift))
+        TextureRenderer(*blur_program(true, radius, sigma, dir, step, tilt_shift))
 {
     blur_program_ = blur_program(false, radius, sigma, dir, step, tilt_shift);
 }
@@ -74,15 +77,15 @@
                     unsigned int radius, float sigma, BlurRenderer::BlurDirection direction,
                     float tilt_shift)
 {
-    vtx_source.append_file(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-    frg_source.append_file(GLMARK_DATA_PATH"/shaders/terrain-blur.frag");
+    vtx_source.append_file(Options::data_path + "/shaders/terrain-texture.vert");
+    frg_source.append_file(Options::data_path + "/shaders/terrain-blur.frag");
 
     /* Don't let the gaussian curve become too narrow */
     if (sigma < 1.0)
         sigma = 1.0;
 
     unsigned int side = 2 * radius + 1;
-    float values[radius];
+    std::vector<float> values(radius + 1);
     float sum = 0.0;
 
     for (unsigned int i = 0; i < radius + 1; i++) {
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/copy-renderer.cpp glmark2-2021.02/src/scene-terrain/copy-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/copy-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/copy-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,12 +19,13 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-CopyRenderer::CopyRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *copy_program(true))
+CopyRenderer::CopyRenderer() :
+    TextureRenderer(*copy_program(true))
 {
     copy_program_ = copy_program(false);
 }
@@ -38,8 +39,8 @@
 
     if (!copy_program) {
         copy_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-overlay.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-overlay.frag");
 
         Scene::load_shaders_from_strings(*copy_program, vtx_shader.str(), frg_shader.str());
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/luminance-renderer.cpp glmark2-2021.02/src/scene-terrain/luminance-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/luminance-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/luminance-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,12 +19,13 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-LuminanceRenderer::LuminanceRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *luminance_program(true))
+LuminanceRenderer::LuminanceRenderer() :
+    TextureRenderer(*luminance_program(true))
 {
     luminance_program_ = luminance_program(false);
 }
@@ -38,8 +39,8 @@
 
     if (!luminance_program) {
         luminance_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-luminance.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-luminance.frag");
 
         Scene::load_shaders_from_strings(*luminance_program,
                                          vtx_shader.str(), frg_shader.str());
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/normal-from-height-renderer.cpp glmark2-2021.02/src/scene-terrain/normal-from-height-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/normal-from-height-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/normal-from-height-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,19 +19,40 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
-NormalFromHeightRenderer::NormalFromHeightRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *normal_from_height_program(size, true))
+NormalFromHeightRenderer::NormalFromHeightRenderer() :
+    TextureRenderer(*normal_from_height_program(true))
 {
-    normal_from_height_program_ = normal_from_height_program(size, false);
+    normal_from_height_program_ = normal_from_height_program(false);
+}
+
+void
+NormalFromHeightRenderer::setup_onscreen(Canvas& canvas)
+{
+    TextureRenderer::setup_onscreen(canvas);
+
+    normal_from_height_program_->start();
+    (*normal_from_height_program_)["resolution"] =
+        LibMatrix::vec2(canvas.width(), canvas.height());
+    normal_from_height_program_->stop();
+}
+
+void
+NormalFromHeightRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
+{
+    TextureRenderer::setup_offscreen(size, has_depth);
+
+    normal_from_height_program_->start();
+    (*normal_from_height_program_)["resolution"] = size;
+    normal_from_height_program_->stop();
 }
 
 Program *
-NormalFromHeightRenderer::normal_from_height_program(const LibMatrix::vec2 &size,
-                                                     bool create_new)
+NormalFromHeightRenderer::normal_from_height_program(bool create_new)
 {
     static Program *normal_from_height_program(0);
     if (create_new)
@@ -39,15 +60,14 @@
 
     if (!normal_from_height_program) {
         normal_from_height_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-normalmap.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-normalmap.frag");
 
         Scene::load_shaders_from_strings(*normal_from_height_program,
                                          vtx_shader.str(), frg_shader.str());
 
         normal_from_height_program->start();
         (*normal_from_height_program)["heightMap"] = 0;
-        (*normal_from_height_program)["resolution"] = size;
         (*normal_from_height_program)["height"] = 0.05f;
         (*normal_from_height_program)["uvOffset"] = LibMatrix::vec2(0.0f, 0.0f);
         (*normal_from_height_program)["uvScale"] = LibMatrix::vec2(1.0f, 1.0f);
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/overlay-renderer.cpp glmark2-2021.02/src/scene-terrain/overlay-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/overlay-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/overlay-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,6 +19,7 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
@@ -33,10 +34,14 @@
 
 
 void
-OverlayRenderer::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+OverlayRenderer::setup_onscreen(Canvas &canvas)
+{
+}
+
+void
+OverlayRenderer::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     static_cast<void>(size);
-    static_cast<void>(onscreen);
     static_cast<void>(has_depth);
 }
 
@@ -107,8 +112,8 @@
 void
 OverlayRenderer::create_program()
 {
-    ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-    ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-overlay.frag");
+    ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+    ShaderSource frg_shader(Options::data_path + "/shaders/terrain-overlay.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_shader.str(), frg_shader.str()))
         return;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/renderer-chain.cpp glmark2-2021.02/src/scene-terrain/renderer-chain.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/renderer-chain.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/renderer-chain.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -22,10 +22,15 @@
 #include "renderer.h"
 
 void
-RendererChain::setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth)
+RendererChain::setup_onscreen(Canvas &canvas)
+{
+    static_cast<void>(canvas);
+}
+
+void
+RendererChain::setup_offscreen(const LibMatrix::vec2 &size, bool has_depth)
 {
     static_cast<void>(size);
-    static_cast<void>(onscreen);
     static_cast<void>(has_depth);
 }
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/renderer.h glmark2-2021.02/src/scene-terrain/renderer.h
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/renderer.h	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/renderer.h	2021-04-25 01:47:22.000000000 +0800
@@ -22,6 +22,7 @@
 #include <stdint.h>
 #include <vector>
 
+#include "canvas.h"
 #include "mesh.h"
 #include "vec.h"
 #include "program.h"
@@ -36,7 +37,8 @@
     /** 
      * Sets up the renderer's target.
      */
-    virtual void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth) = 0;
+    virtual void setup_onscreen(Canvas& canvas) = 0;
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth) = 0;
     /** 
      * Sets up the renderer's target texture (if any).
      */
@@ -81,7 +83,8 @@
     virtual ~RendererChain() {}
 
     /* IRenderer methods */
-    void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    void setup_onscreen(Canvas& canvas);
+    void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     void setup_texture(GLint min_filter, GLint mag_filter,
                        GLint wrap_s, GLint wrap_t);
     void input_texture(GLuint t);
@@ -108,11 +111,12 @@
 class BaseRenderer : public IRenderer
 {
 public:
-    BaseRenderer(const LibMatrix::vec2 &size);
+    BaseRenderer();
     virtual ~BaseRenderer();
 
     /* IRenderer methods */
-    virtual void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     virtual void setup_texture(GLint min_filter, GLint mag_filter,
                                GLint wrap_s, GLint wrap_t);
     virtual void input_texture(GLuint t) { input_texture_ = t; }
@@ -123,7 +127,7 @@
     virtual void render() = 0;
 
 protected:
-    void recreate(bool onscreen, bool has_depth);
+    void recreate(Canvas* canvas, bool has_depth);
     void create_texture();
     void update_texture_parameters();
     void create_fbo(bool has_depth);
@@ -133,6 +137,7 @@
     GLuint input_texture_;
     GLuint fbo_;
     GLuint depth_renderbuffer_;
+    bool owns_fbo_;
     GLint min_filter_;
     GLint mag_filter_;
     GLint wrap_s_;
@@ -146,7 +151,7 @@
 class TextureRenderer : public BaseRenderer
 {
 public:
-    TextureRenderer(const LibMatrix::vec2 &size, Program &program);
+    TextureRenderer(Program &program);
     virtual ~TextureRenderer() { }
 
     /* IRenderer/BaseRenderer methods */
@@ -170,7 +175,7 @@
 class CopyRenderer : public TextureRenderer
 {
 public:
-    CopyRenderer(const LibMatrix::vec2 &size);
+    CopyRenderer();
 
     virtual ~CopyRenderer() { delete copy_program_; }
 
@@ -186,7 +191,7 @@
 class SimplexNoiseRenderer : public TextureRenderer
 {
 public:
-    SimplexNoiseRenderer(const LibMatrix::vec2 &size);
+    SimplexNoiseRenderer();
     virtual ~SimplexNoiseRenderer() { delete noise_program_; }
 
     LibMatrix::vec2 uv_scale() { return uv_scale_; }
@@ -204,12 +209,14 @@
 class NormalFromHeightRenderer : public TextureRenderer
 {
 public:
-    NormalFromHeightRenderer(const LibMatrix::vec2 &size);
+    NormalFromHeightRenderer();
     virtual ~NormalFromHeightRenderer() { delete normal_from_height_program_; }
 
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
+
 private:
-    static Program *normal_from_height_program(const LibMatrix::vec2 &size, 
-                                               bool create_new);
+    static Program *normal_from_height_program(bool create_new);
 
     Program *normal_from_height_program_;
 };
@@ -220,7 +227,7 @@
 class LuminanceRenderer : public TextureRenderer
 {
 public:
-    LuminanceRenderer(const LibMatrix::vec2 &size);
+    LuminanceRenderer();
     virtual ~LuminanceRenderer() { delete luminance_program_; }
 
 private:
@@ -242,8 +249,8 @@
         BlurDirectionBoth
     };
 
-    BlurRenderer(const LibMatrix::vec2 &size, int radius, float sigma,
-                 BlurDirection dir, const LibMatrix::vec2 &step, float tilt_shift);
+    BlurRenderer(int radius, float sigma, BlurDirection dir,
+                 const LibMatrix::vec2 &step, float tilt_shift);
     virtual ~BlurRenderer() { delete blur_program_; }
 
 private:
@@ -265,7 +272,8 @@
     virtual ~OverlayRenderer() { }
 
     /* IRenderable Methods */
-    void setup(const LibMatrix::vec2 &size, bool onscreen, bool has_depth);
+    virtual void setup_onscreen(Canvas& canvas);
+    virtual void setup_offscreen(const LibMatrix::vec2 &size, bool has_depth);
     void setup_texture(GLint min_filter, GLint mag_filter,
                        GLint wrap_s, GLint wrap_t);
     void input_texture(GLuint t) { input_texture_ = t; }
@@ -293,7 +301,7 @@
 class TerrainRenderer : public BaseRenderer
 {
 public:
-    TerrainRenderer(const LibMatrix::vec2 &size, const LibMatrix::vec2 &repeat_overlay);
+    TerrainRenderer(const LibMatrix::vec2 &repeat_overlay);
     virtual ~TerrainRenderer();
 
     /* IRenderable Methods */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/simplex-noise-renderer.cpp glmark2-2021.02/src/scene-terrain/simplex-noise-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/simplex-noise-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/simplex-noise-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,14 +19,15 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "shader-source.h"
 
 LibMatrix::vec2 SimplexNoiseRenderer::uv_scale_(1.5f, 1.5);
 
-SimplexNoiseRenderer::SimplexNoiseRenderer(const LibMatrix::vec2 &size) :
-    TextureRenderer(size, *noise_program(true))
+SimplexNoiseRenderer::SimplexNoiseRenderer() :
+    TextureRenderer(*noise_program(true))
 {
     noise_program_ = noise_program(false);
 }
@@ -40,8 +41,8 @@
 
     if (!noise_program) {
         noise_program = new Program();
-        ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain-texture.vert");
-        ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain-noise.frag");
+        ShaderSource vtx_shader(Options::data_path + "/shaders/terrain-texture.vert");
+        ShaderSource frg_shader(Options::data_path + "/shaders/terrain-noise.frag");
 
         Scene::load_shaders_from_strings(*noise_program, vtx_shader.str(), frg_shader.str());
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/terrain-renderer.cpp glmark2-2021.02/src/scene-terrain/terrain-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/terrain-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/terrain-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -19,14 +19,14 @@
  * Authors:
  *  Alexandros Frantzis
  */
+#include "options.h"
 #include "scene.h"
 #include "renderer.h"
 #include "texture.h"
 #include "shader-source.h"
 
-TerrainRenderer::TerrainRenderer(const LibMatrix::vec2 &size,
-                                 const LibMatrix::vec2 &repeat_overlay) :
-    BaseRenderer(size), height_map_tex_(0), normal_map_tex_(0),
+TerrainRenderer::TerrainRenderer(const LibMatrix::vec2 &repeat_overlay) :
+    BaseRenderer(), height_map_tex_(0), normal_map_tex_(0),
     specular_map_tex_(0), repeat_overlay_(repeat_overlay)
 {
     create_mesh();
@@ -85,8 +85,8 @@
 void
 TerrainRenderer::init_program()
 {
-    ShaderSource vtx_shader(GLMARK_DATA_PATH"/shaders/terrain.vert");
-    ShaderSource frg_shader(GLMARK_DATA_PATH"/shaders/terrain.frag");
+    ShaderSource vtx_shader(Options::data_path + "/shaders/terrain.vert");
+    ShaderSource frg_shader(Options::data_path + "/shaders/terrain.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_shader.str(), frg_shader.str()))
         return;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/texture-renderer.cpp glmark2-2021.02/src/scene-terrain/texture-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain/texture-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain/texture-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,8 +21,8 @@
  */
 #include "renderer.h"
 
-TextureRenderer::TextureRenderer(const LibMatrix::vec2 &size, Program &program) :
-    BaseRenderer(size), program_(program)
+TextureRenderer::TextureRenderer(Program &program) :
+    BaseRenderer(), program_(program)
 {
     /* Create the mesh (quad) used for rendering */
     create_mesh();
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain.cpp glmark2-2021.02/src/scene-terrain.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-terrain.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-terrain.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -64,63 +64,67 @@
         const vec2 bloom_res(256.0f, 256.0f);
         const vec2 grass_res(512.0f, 512.0f);
 
-        height_map_renderer = new SimplexNoiseRenderer(map_res);
-        height_map_renderer->setup(height_map_renderer->size(), false, false);
+        height_map_renderer = new SimplexNoiseRenderer();
+        height_map_renderer->setup_offscreen(map_res, false);
 
-        normal_map_renderer = new NormalFromHeightRenderer(map_res);
-        normal_map_renderer->setup(normal_map_renderer->size(), false, false);
+        normal_map_renderer = new NormalFromHeightRenderer();
+        normal_map_renderer->setup_offscreen(map_res, false);
 
-        specular_map_renderer = new LuminanceRenderer(grass_res);
+        specular_map_renderer = new LuminanceRenderer();
+        specular_map_renderer->setup_offscreen(grass_res, false);
         specular_map_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                              GL_REPEAT, GL_REPEAT);
-        specular_map_renderer->setup(specular_map_renderer->size(), false, false);
 
-        terrain_renderer = new TerrainRenderer(screen_res, repeat_overlay);
-        terrain_renderer->setup(terrain_renderer->size(),
-                                !use_bloom && !use_tilt_shift,
-                                true);
+        terrain_renderer = new TerrainRenderer(repeat_overlay);
+        if (!use_bloom && !use_tilt_shift)
+            terrain_renderer->setup_onscreen(canvas);
+        else
+            terrain_renderer->setup_offscreen(screen_res, true);
         terrain_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                         GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
 
         /* Bloom */
         if (use_bloom) {
-            bloom_h_renderer = new BlurRenderer(bloom_res, 2, 4.0,
+            bloom_h_renderer = new BlurRenderer(2, 4.0,
                                                 BlurRenderer::BlurDirectionHorizontal,
                                                 vec2(1.0, 1.0) / screen_res,
                                                 0.0);
+            bloom_h_renderer->setup_offscreen(bloom_res, false);
             bloom_h_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                             GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            bloom_h_renderer->setup(bloom_h_renderer->size(), false, false);
 
-            bloom_v_renderer = new BlurRenderer(bloom_res, 2, 4.0,
+            bloom_v_renderer = new BlurRenderer(2, 4.0,
                                                 BlurRenderer::BlurDirectionVertical,
                                                 vec2(1.0, 1.0) / bloom_res,
                                                 0.0);
+            bloom_v_renderer->setup_offscreen(bloom_res, false);
             bloom_v_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                             GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            bloom_v_renderer->setup(bloom_v_renderer->size(), false, false);
             overlay_renderer = new OverlayRenderer(*terrain_renderer, 0.6);
         }
 
         /* Tilt-shift */
         if (use_tilt_shift) {
-            tilt_h_renderer = new BlurRenderer(screen_res, 4, 2.7,
+            tilt_h_renderer = new BlurRenderer(4, 2.7,
                                                BlurRenderer::BlurDirectionHorizontal,
                                                vec2(1.0, 1.0) / screen_res,
                                                0.5);
+            tilt_h_renderer->setup_offscreen(screen_res, false);
             tilt_h_renderer->setup_texture(GL_LINEAR_MIPMAP_LINEAR, GL_LINEAR,
                                            GL_CLAMP_TO_EDGE, GL_CLAMP_TO_EDGE);
-            tilt_h_renderer->setup(tilt_h_renderer->size(), false, false);
 
-            tilt_v_renderer = new BlurRenderer(screen_res, 4, 2.7,
+            tilt_v_renderer = new BlurRenderer(4, 2.7,
                                                BlurRenderer::BlurDirectionVertical,
                                                vec2(1.0, 1.0) / screen_res,
                                                0.5);
+            tilt_v_renderer->setup_onscreen(canvas);
         }
 
         /* Copy renderer */
-        if (use_bloom && !use_tilt_shift)
-            copy_renderer = new CopyRenderer(screen_res);
+        if (use_bloom && !use_tilt_shift) {
+            copy_renderer = new CopyRenderer();
+            copy_renderer->setup_onscreen(canvas);
+        }
 
         /* Height normal chain */
         height_normal_chain = new RendererChain();
@@ -310,7 +314,7 @@
     /* Create the specular map */
     priv_->specular_map_renderer->render();
 
-    glBindFramebuffer(GL_FRAMEBUFFER, 0);
+    glBindFramebuffer(GL_FRAMEBUFFER, canvas_.fbo());
     glViewport(0, 0, canvas_.width(), canvas_.height());
 
     currentFrame_ = 0;
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/scene-texture.cpp glmark2-2021.02/src/scene-texture.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/scene-texture.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/scene-texture.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -24,6 +24,7 @@
  */
 #include "scene.h"
 #include "mat.h"
+#include "options.h"
 #include "stack.h"
 #include "vec.h"
 #include "log.h"
@@ -109,10 +110,10 @@
     if (!Scene::setup())
         return false;
 
-    static const std::string vtx_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic.vert");
-    static const std::string vtx_shader_texgen_filename(GLMARK_DATA_PATH"/shaders/light-basic-texgen.vert");
-    static const std::string frg_shader_filename(GLMARK_DATA_PATH"/shaders/light-basic-tex.frag");
-    static const std::string frg_shader_bilinear_filename(GLMARK_DATA_PATH"/shaders/light-basic-tex-bilinear.frag");
+    static const std::string vtx_shader_filename(Options::data_path + "/shaders/light-basic.vert");
+    static const std::string vtx_shader_texgen_filename(Options::data_path + "/shaders/light-basic-texgen.vert");
+    static const std::string frg_shader_filename(Options::data_path + "/shaders/light-basic-tex.frag");
+    static const std::string frg_shader_bilinear_filename(Options::data_path + "/shaders/light-basic-tex-bilinear.frag");
     static const LibMatrix::vec4 lightPosition(20.0f, 20.0f, 10.0f, 1.0f);
     static const LibMatrix::vec4 materialDiffuse(1.0f, 1.0f, 1.0f, 1.0f);
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/shared-library.cpp glmark2-2021.02/src/shared-library.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/shared-library.cpp	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/shared-library.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,90 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#include "shared-library.h"
+
+#include <stdio.h>
+
+#if defined(WIN32)
+#include <windows.h>
+#else
+#include <dlfcn.h>
+#endif
+
+/******************
+ * Public methods *
+ ******************/
+SharedLibrary::SharedLibrary() : handle_(nullptr) {}
+
+SharedLibrary::~SharedLibrary()
+{
+    close();
+}
+
+bool SharedLibrary::open(const char *libName)
+{
+#if defined(WIN32)
+    handle_ = LoadLibraryA(libName);
+#else
+    handle_ = dlopen(libName, RTLD_NOW | RTLD_NODELETE);
+#endif
+    return (handle_ != nullptr);
+}
+
+bool SharedLibrary::open_from_alternatives(std::initializer_list<const char*> alt_names)
+{
+    for (const char *name : alt_names) {
+        if (open(name))
+            return true;
+    }
+
+    return false;
+}
+
+void SharedLibrary::close()
+{
+    if (handle_)
+    {
+#if defined(WIN32)
+        FreeLibrary(reinterpret_cast<HMODULE>(handle_));
+#else
+        dlclose(handle_);
+#endif
+    }
+}
+
+void *SharedLibrary::handle() const
+{
+    return handle_;
+}
+
+void *SharedLibrary::load(const char *symbol) const
+{
+#if defined(WIN32)
+    return reinterpret_cast<void*>(GetProcAddress(reinterpret_cast<HMODULE>(handle_), symbol));
+#else
+    return dlsym(handle_, symbol);
+#endif
+}
+
+/*******************
+ * Private methods *
+ *******************/
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/shared-library.h glmark2-2021.02/src/shared-library.h
--- glmark2-2014.03+git20150611.fa71af2d/src/shared-library.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/shared-library.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,44 @@
+/*
+ * Copyright © 2019 Jamie Madill
+ *
+ * This file is part of the glmark2 OpenGL (ES) 2.0 benchmark.
+ *
+ * glmark2 is free software: you can redistribute it and/or modify it under the
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation, either version 3 of the License, or (at your option) any later
+ * version.
+ *
+ * glmark2 is distributed in the hope that it will be useful, but WITHOUT ANY
+ * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * glmark2.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Authors:
+ *  Jamie Madill
+ */
+#ifndef GLMARK2_SHARED_LIBRARY_H_
+#define GLMARK2_SHARED_LIBRARY_H_
+
+#include <initializer_list>
+
+class SharedLibrary
+{
+public:
+    SharedLibrary();
+    ~SharedLibrary();
+
+    bool open(const char *name);
+    bool open_from_alternatives(std::initializer_list<const char*> alt_names);
+    void close();
+
+    void *handle() const;
+    void *load(const char *symbol) const;
+
+private:
+    void *handle_;
+};
+
+#endif /* GLMARK2_SHARED_LIBRARY_H_ */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/text-renderer.cpp glmark2-2021.02/src/text-renderer.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/text-renderer.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/text-renderer.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -21,6 +21,7 @@
  */
 #include "text-renderer.h"
 #include "gl-headers.h"
+#include "options.h"
 #include "scene.h"
 #include "shader-source.h"
 #include "vec.h"
@@ -49,8 +50,8 @@
     size(0.03);
 
     glGenBuffers(2, vbo_);
-    ShaderSource vtx_source(GLMARK_DATA_PATH"/shaders/text-renderer.vert");
-    ShaderSource frg_source(GLMARK_DATA_PATH"/shaders/text-renderer.frag");
+    ShaderSource vtx_source(Options::data_path + "/shaders/text-renderer.vert");
+    ShaderSource frg_source(Options::data_path + "/shaders/text-renderer.frag");
 
     if (!Scene::load_shaders_from_strings(program_, vtx_source.str(),
                                           frg_source.str()))
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/texture.cpp glmark2-2021.02/src/texture.cpp
--- glmark2-2014.03+git20150611.fa71af2d/src/texture.cpp	2015-06-11 22:39:35.000000000 +0800
+++ glmark2-2021.02/src/texture.cpp	2021-04-25 01:47:22.000000000 +0800
@@ -23,9 +23,11 @@
  */
 #include "texture.h"
 #include "log.h"
+#include "options.h"
 #include "util.h"
 #include "image-reader.h"
 
+#include <algorithm>
 #include <cstdarg>
 #include <vector>
 
@@ -149,7 +151,7 @@
         return TexturePrivate::textureMap;
     }
     vector<string> pathVec;
-    string dataDir(GLMARK_DATA_PATH"/textures");
+    string dataDir(Options::data_path + "/textures");
     Util::list_files(dataDir, pathVec);
     // Now that we have a list of all of the image files available to us,
     // let's go through and pull out the names and what format they're in
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/wscript_build glmark2-2021.02/src/wscript_build
--- glmark2-2014.03+git20150611.fa71af2d/src/wscript_build	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/src/wscript_build	2021-04-25 01:47:22.000000000 +0800
@@ -1,70 +1,251 @@
+# Detect a Windows build.
+is_win = 'WIN32' in ' '.join(bld.env.keys())
+
+# Detect MSVC
+is_msvc = bld.env.CXX_NAME == 'msvc'
+
 all_sources = bld.path.ant_glob('*.cpp scene-ideas/*.cc scene-terrain/*.cpp')
 common_sources = [f for f in all_sources if f.name.find('canvas-') == -1 and
                                             f.name.find('android') == -1 and
                                             f.name.find('native-state-') == -1 and
-                                            f.name.find('gl-state-') == -1]
-common_uselibs = ['libpng']
+                                            f.name.find('gl-state-') == -1 and
+                                            f.name.find('main.cpp') == -1]
+
 common_defines = ['USE_EXCEPTIONS']
 
 libmatrix_sources = [f for f in bld.path.ant_glob('libmatrix/*.cc')
                      if not f.name.endswith('test.cc')]
+
+common_flavor_sources = ['main.cpp', 'canvas-generic.cpp']
+
+libpng_local_sources = [f for f in bld.path.ant_glob('libpng/*.c')]
+zlib_local_sources = [f for f in bld.path.ant_glob('zlib/*.c')]
+
+# Some c sources in libjpeg-turbo are included like headers. So we cannot use globs.
+libjpeg_local_sources = [
+  'jcapimin', 'jcapistd', 'jccoefct', 'jccolor', 'jcdctmgr', 'jchuff', 'jcinit',
+  'jcmainct', 'jcmarker', 'jcmaster', 'jcomapi', 'jcparam', 'jcphuff', 'jcprepct',
+  'jcsample', 'jdapimin', 'jdapistd', 'jdatadst', 'jdatasrc', 'jdcoefct', 'jdcolor',
+  'jddctmgr', 'jdhuff', 'jdinput', 'jdmainct', 'jdmarker', 'jdmaster', 'jdmerge',
+  'jdphuff', 'jdpostct', 'jdsample', 'jerror', 'jfdctflt', 'jfdctfst', 'jfdctint',
+  'jidctflt', 'jidctfst', 'jidctint', 'jidctred', 'jmemmgr', 'jmemnobs', 'jquant1',
+  'jquant2', 'jsimd_none', 'jutils',
+]
+
+libjpeg_turbo_local_sources = ['libjpeg-turbo/%s.c' % f for f in libjpeg_local_sources]
+
+platform_includes = []
+platform_libs = []
+platform_uselibs = []
+
+if is_win:
+  platform_uselibs += ['libpng-local', 'zlib-local', 'libjpeg-turbo-local']
+  platform_libs = ['user32']
+
+  # On windows include some overrides for posix headers.
+  if is_msvc:
+    platform_includes += ['include']
+else:
+  platform_uselibs += ['libpng']
+  platform_libs = ['m', 'jpeg', 'dl']
+
+if 'WAYLAND_SCANNER_wayland_scanner' in bld.env.keys():
+    def wayland_scanner_cmd(arg, src):
+        return '%s %s < %s > ${TGT}' % (bld.env['WAYLAND_SCANNER_wayland_scanner'], arg, src)
+
+    def wayland_proto_src_path(proto, ver):
+        wp_dir = bld.env['WAYLAND_PROTOCOLS_pkgdatadir']
+        if ver == 'stable':
+            return '%s/stable/%s/%s.xml' % (wp_dir, proto, proto)
+        else:
+            return '%s/unstable/%s/%s-unstable-%s.xml' % (wp_dir, proto, proto, ver)
+
+    def wayland_client_protocol(proto, ver, dst_base):
+        src_path = wayland_proto_src_path(proto, ver)
+        out = '%s-client-protocol.h' % dst_base
+        return bld(rule = wayland_scanner_cmd('client-header', src_path),
+                   target = out,
+                   name = out,
+                   ext_out = ['.h'])
+
+    def wayland_protocol_code(proto, ver, dst_base):
+        src_path = wayland_proto_src_path(proto, ver)
+        out = '%s-protocol.c' % dst_base
+        return bld(rule = wayland_scanner_cmd('private-code', src_path),
+                   target = out,
+                   name = out)
+
+    wayland_client_protocol('xdg-shell', 'stable', 'xdg-shell')
+    wayland_protocol_code('xdg-shell', 'stable', 'xdg-shell')
+
 flavor_sources = {
-  'x11-gl' : ['canvas-generic.cpp', 'native-state-x11.cpp', 'gl-state-glx.cpp'],
-  'x11-glesv2' : ['canvas-generic.cpp', 'native-state-x11.cpp', 'gl-state-egl.cpp'],
-  'drm-gl' : ['canvas-generic.cpp', 'native-state-drm.cpp', 'gl-state-egl.cpp'],
-  'drm-glesv2' : ['canvas-generic.cpp', 'native-state-drm.cpp', 'gl-state-egl.cpp'],
-  'mir-gl' : ['canvas-generic.cpp', 'native-state-mir.cpp', 'gl-state-egl.cpp'],
-  'mir-glesv2' : ['canvas-generic.cpp', 'native-state-mir.cpp', 'gl-state-egl.cpp'],
-  'wayland-gl' : ['canvas-generic.cpp', 'native-state-wayland.cpp', 'gl-state-egl.cpp'],
-  'wayland-glesv2' : ['canvas-generic.cpp', 'native-state-wayland.cpp', 'gl-state-egl.cpp']
+  'dispmanx-glesv2' : common_flavor_sources + ['native-state-dispmanx.cpp', 'gl-state-egl.cpp'],
+  'drm-gl' : common_flavor_sources + ['native-state-drm.cpp', 'gl-state-egl.cpp'],
+  'drm-glesv2' : common_flavor_sources + ['native-state-drm.cpp', 'gl-state-egl.cpp'],
+  'mir-gl' : common_flavor_sources + ['native-state-mir.cpp', 'gl-state-egl.cpp'],
+  'mir-glesv2' : common_flavor_sources + ['native-state-mir.cpp', 'gl-state-egl.cpp'],
+  'wayland-gl' : common_flavor_sources + ['native-state-wayland.cpp', 'gl-state-egl.cpp'],
+  'wayland-glesv2' : common_flavor_sources + ['native-state-wayland.cpp', 'gl-state-egl.cpp'],
+  'win32-gl': common_flavor_sources + ['native-state-win32.cpp', 'gl-state-wgl.cpp'],
+  'win32-glesv2': common_flavor_sources + ['native-state-win32.cpp', 'gl-state-egl.cpp'],
+  'x11-gl' : common_flavor_sources + ['native-state-x11.cpp', 'gl-state-glx.cpp'],
+  'x11-glesv2' : common_flavor_sources + ['native-state-x11.cpp', 'gl-state-egl.cpp'],
 }
 flavor_uselibs = {
-  'x11-gl' : ['x11', 'gl', 'matrix-gl'],
-  'x11-glesv2' : ['x11', 'egl', 'glesv2', 'matrix-glesv2'],
-  'drm-gl' : ['drm', 'gbm', 'egl', 'gl', 'matrix-gl'],
-  'drm-glesv2' : ['drm', 'gbm', 'egl', 'glesv2', 'matrix-glesv2'],
-  'mir-gl' : ['mirclient', 'egl', 'gl', 'matrix-gl'],
-  'mir-glesv2' : ['mirclient', 'egl', 'glesv2', 'matrix-glesv2'],
-  'wayland-gl' : ['wayland-client', 'wayland-egl', 'egl', 'gl', 'matrix-gl'],
-  'wayland-glesv2' : ['wayland-client', 'wayland-egl', 'egl', 'glesv2', 'matrix-glesv2']
+  'dispmanx-glesv2' : ['glad-egl-dispmanx', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2',  'dispmanx'],
+  'drm-gl' : ['drm', 'gbm', 'udev', 'glad-egl-drm', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'drm-glesv2' : ['drm', 'gbm', 'udev', 'glad-egl-drm', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'mir-gl' : ['mirclient', 'glad-egl-mir', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'mir-glesv2' : ['mirclient', 'glad-egl-mir', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'wayland-gl' : ['wayland-client', 'wayland-egl', 'wayland-cursor', 'glad-egl-wayland', 'glad-gl', 'matrix-gl', 'common-gl'],
+  'wayland-glesv2' : ['wayland-client', 'wayland-egl', 'wayland-cursor', 'glad-egl-wayland', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'win32-gl': ['glad-gl', 'glad-wgl', 'matrix-gl', 'common-gl'],
+  'win32-glesv2': ['glad-egl-win32', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
+  'x11-gl' : ['x11', 'glad-gl', 'glad-glx', 'matrix-gl', 'common-gl'],
+  'x11-glesv2' : ['x11', 'glad-egl-x11', 'glad-glesv2', 'matrix-glesv2', 'common-glesv2'],
 }
+
 flavor_defines = {
-  'x11-gl' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GL', 'GLMARK2_USE_GLX'],
-  'x11-glesv2' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
-  'drm-gl' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL', '__GBM__'],
-  'drm-glesv2' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL', '__GBM__'],
+  'dispmanx-glesv2' : ['GLMARK2_USE_DISPMANX', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'drm-gl' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
+  'drm-glesv2' : ['GLMARK2_USE_DRM', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
   'mir-gl' : ['GLMARK2_USE_MIR', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
   'mir-glesv2' : ['GLMARK2_USE_MIR', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
   'wayland-gl' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GL', 'GLMARK2_USE_EGL'],
-  'wayland-glesv2' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL']
+  'wayland-glesv2' : ['GLMARK2_USE_WAYLAND', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'win32-gl': ['GLMARK2_USE_WIN32', 'GLMARK2_USE_WGL', 'GLMARK2_USE_GL'],
+  'win32-glesv2': ['GLMARK2_USE_WIN32', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+  'x11-gl' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GL', 'GLMARK2_USE_GLX'],
+  'x11-glesv2' : ['GLMARK2_USE_X11', 'GLMARK2_USE_GLESv2', 'GLMARK2_USE_EGL'],
+}
+flavor_libs = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : [],
+  'wayland-glesv2' : [],
+  'win32-gl': ['opengl32', 'gdi32'],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+flavor_depends_on = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : ['xdg-shell-client-protocol.h', 'xdg-shell-protocol.c'],
+  'wayland-glesv2' : ['xdg-shell-client-protocol.h', 'xdg-shell-protocol.c'],
+  'win32-gl': [],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+flavor_sources_gen = {
+  'dispmanx-glesv2' : [],
+  'drm-gl' : [],
+  'drm-glesv2' : [],
+  'mir-gl' : [],
+  'mir-glesv2' : [],
+  'wayland-gl' : [bld.path.find_or_declare('xdg-shell-protocol.c')],
+  'wayland-glesv2' : [bld.path.find_or_declare('xdg-shell-protocol.c')],
+  'win32-gl': [],
+  'win32-glesv2' : [],
+  'x11-gl' : [],
+  'x11-glesv2' : [],
+}
+egl_platform_defines = {
+  'dispmanx' : ['MESA_EGL_NO_X11_HEADERS'],
+  'drm' : ['__GBM__'],
+  'mir' : ['MESA_EGL_NO_X11_HEADERS'],
+  'wayland' : ['WL_EGL_PLATFORM'],
+  'win32' : [],
+  'x11' : [],
 }
 
-includes = ['.', 'scene-ideas', 'scene-terrain']
+includes = ['.', 'scene-ideas', 'scene-terrain'] + platform_includes
 
 all_uselibs = set()
 
 for name in bld.env.keys():
     if name.startswith('FLAVOR_') and bld.env[name]:
         flavor = name.replace('FLAVOR_', '').lower().replace('_', '-')
+        egl_platform = flavor.split('-')[0]
         target = bld.env[name]
-        bld(
+        node = bld(
             features     = ['cxx', 'cprogram'],
-            source       = common_sources + flavor_sources[flavor],
+            source       = flavor_sources[flavor],
             target       = target,
-            use          = common_uselibs + flavor_uselibs[flavor],
-            lib          = ['m', 'jpeg', 'dl'],
-            includes     = includes,
-            defines      = common_defines + flavor_defines[flavor]
+            use          = platform_uselibs + flavor_uselibs[flavor],
+            lib          = platform_libs + flavor_libs[flavor],
+            includes     = ['.'] + platform_includes,
+            defines      = common_defines + flavor_defines[flavor] +
+                           egl_platform_defines[egl_platform],
+            depends_on   = flavor_depends_on[flavor]
             )
-        all_uselibs |= set(flavor_uselibs[flavor])
+        if flavor_sources_gen[flavor]:
+            node.source.extend(flavor_sources_gen[flavor])
+        all_uselibs |= set(flavor_uselibs[flavor] + platform_uselibs)
+
+# Build glad-egl for all used EGL platforms
+for egl_target in (v for v in all_uselibs if v.startswith('glad-egl')):
+    egl_platform = egl_target.split('-')[2]
+    bld(
+        features = ['c'],
+        source   = ['glad/src/egl.c'],
+        target   = egl_target,
+        includes = ['glad/include'],
+        export_includes = 'glad/include',
+        defines = egl_platform_defines[egl_platform]
+        )
+
+if 'glad-glx' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/glx.c'],
+        target   = 'glad-glx',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-gl' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/gl.c'],
+        target   = 'glad-gl',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-wgl' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/wgl.c'],
+        target   = 'glad-wgl',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
+
+if 'glad-glesv2' in all_uselibs:
+    bld(
+        features = ['c'],
+        source   = ['glad/src/gles2.c'],
+        target   = 'glad-glesv2',
+        includes = ['glad/include'],
+        export_includes = 'glad/include'
+        )
         
 if 'matrix-gl' in all_uselibs:
     bld(
         features = ['cxx', 'cxxstlib'],
         source   = libmatrix_sources,
         target   = 'matrix-gl',
+        use      = ['glad-gl'],
         lib      = ['m'],
-        includes = ['.'],
+        includes = ['.'] + platform_includes,
         export_includes = 'libmatrix',
         defines  = ['GLMARK2_USE_GL', 'USE_EXCEPTIONS']
         )
@@ -74,8 +255,56 @@
         features = ['cxx', 'cxxstlib'],
         source   = libmatrix_sources,
         target   = 'matrix-glesv2',
+        use      = ['glad-glesv2'],
         lib      = ['m'],
-        includes = ['.'],
+        includes = ['.'] + platform_includes,
         export_includes = 'libmatrix',
         defines  = ['GLMARK2_USE_GLESv2', 'USE_EXCEPTIONS']
         )
+
+if 'libpng-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = libpng_local_sources,
+        target    = 'libpng-local',
+        use       = ['zlib-local'],
+        export_includes = 'libpng'
+        )
+
+if 'zlib-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = zlib_local_sources,
+        target    = 'zlib-local',
+        export_includes = 'zlib'
+        )
+
+if 'libjpeg-turbo-local' in all_uselibs:
+    bld(
+        features  = ['c'],
+        source    = libjpeg_turbo_local_sources,
+        target    = 'libjpeg-turbo-local',
+        export_includes = 'libjpeg-turbo'
+        )
+
+if 'common-gl' in all_uselibs:
+    bld(
+        features = ['cxx', 'cxxstlib'],
+        source   = common_sources,
+        target   = 'common-gl',
+        use      = platform_uselibs + ['glad-gl', 'matrix-gl'],
+        lib      = ['m', 'jpeg', 'dl'],
+        includes = includes,
+        defines  = ['GLMARK2_USE_GL', 'USE_EXCEPTIONS']
+        )
+
+if 'common-glesv2' in all_uselibs:
+    bld(
+        features = ['cxx', 'cxxstlib'],
+        source   = common_sources,
+        target   = 'common-glesv2',
+        use      = platform_uselibs + ['glad-glesv2', 'matrix-glesv2'],
+        lib      = ['m', 'jpeg', 'dl'],
+        includes = includes,
+        defines  = ['GLMARK2_USE_GLESv2', 'USE_EXCEPTIONS']
+        )
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/adler32.c glmark2-2021.02/src/zlib/adler32.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/adler32.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/adler32.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,186 @@
+/* adler32.c -- compute the Adler-32 checksum of a data stream
+ * Copyright (C) 1995-2011, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#include "zutil.h"
+
+local uLong adler32_combine_ OF((uLong adler1, uLong adler2, z_off64_t len2));
+
+#define BASE 65521U     /* largest prime smaller than 65536 */
+#define NMAX 5552
+/* NMAX is the largest n such that 255n(n+1)/2 + (n+1)(BASE-1) <= 2^32-1 */
+
+#define DO1(buf,i)  {adler += (buf)[i]; sum2 += adler;}
+#define DO2(buf,i)  DO1(buf,i); DO1(buf,i+1);
+#define DO4(buf,i)  DO2(buf,i); DO2(buf,i+2);
+#define DO8(buf,i)  DO4(buf,i); DO4(buf,i+4);
+#define DO16(buf)   DO8(buf,0); DO8(buf,8);
+
+/* use NO_DIVIDE if your processor does not do division in hardware --
+   try it both ways to see which is faster */
+#ifdef NO_DIVIDE
+/* note that this assumes BASE is 65521, where 65536 % 65521 == 15
+   (thank you to John Reiser for pointing this out) */
+#  define CHOP(a) \
+    do { \
+        unsigned long tmp = a >> 16; \
+        a &= 0xffffUL; \
+        a += (tmp << 4) - tmp; \
+    } while (0)
+#  define MOD28(a) \
+    do { \
+        CHOP(a); \
+        if (a >= BASE) a -= BASE; \
+    } while (0)
+#  define MOD(a) \
+    do { \
+        CHOP(a); \
+        MOD28(a); \
+    } while (0)
+#  define MOD63(a) \
+    do { /* this assumes a is not negative */ \
+        z_off64_t tmp = a >> 32; \
+        a &= 0xffffffffL; \
+        a += (tmp << 8) - (tmp << 5) + tmp; \
+        tmp = a >> 16; \
+        a &= 0xffffL; \
+        a += (tmp << 4) - tmp; \
+        tmp = a >> 16; \
+        a &= 0xffffL; \
+        a += (tmp << 4) - tmp; \
+        if (a >= BASE) a -= BASE; \
+    } while (0)
+#else
+#  define MOD(a) a %= BASE
+#  define MOD28(a) a %= BASE
+#  define MOD63(a) a %= BASE
+#endif
+
+/* ========================================================================= */
+uLong ZEXPORT adler32_z(adler, buf, len)
+    uLong adler;
+    const Bytef *buf;
+    z_size_t len;
+{
+    unsigned long sum2;
+    unsigned n;
+
+    /* split Adler-32 into component sums */
+    sum2 = (adler >> 16) & 0xffff;
+    adler &= 0xffff;
+
+    /* in case user likes doing a byte at a time, keep it fast */
+    if (len == 1) {
+        adler += buf[0];
+        if (adler >= BASE)
+            adler -= BASE;
+        sum2 += adler;
+        if (sum2 >= BASE)
+            sum2 -= BASE;
+        return adler | (sum2 << 16);
+    }
+
+    /* initial Adler-32 value (deferred check for len == 1 speed) */
+    if (buf == Z_NULL)
+        return 1L;
+
+    /* in case short lengths are provided, keep it somewhat fast */
+    if (len < 16) {
+        while (len--) {
+            adler += *buf++;
+            sum2 += adler;
+        }
+        if (adler >= BASE)
+            adler -= BASE;
+        MOD28(sum2);            /* only added so many BASE's */
+        return adler | (sum2 << 16);
+    }
+
+    /* do length NMAX blocks -- requires just one modulo operation */
+    while (len >= NMAX) {
+        len -= NMAX;
+        n = NMAX / 16;          /* NMAX is divisible by 16 */
+        do {
+            DO16(buf);          /* 16 sums unrolled */
+            buf += 16;
+        } while (--n);
+        MOD(adler);
+        MOD(sum2);
+    }
+
+    /* do remaining bytes (less than NMAX, still just one modulo) */
+    if (len) {                  /* avoid modulos if none remaining */
+        while (len >= 16) {
+            len -= 16;
+            DO16(buf);
+            buf += 16;
+        }
+        while (len--) {
+            adler += *buf++;
+            sum2 += adler;
+        }
+        MOD(adler);
+        MOD(sum2);
+    }
+
+    /* return recombined sums */
+    return adler | (sum2 << 16);
+}
+
+/* ========================================================================= */
+uLong ZEXPORT adler32(adler, buf, len)
+    uLong adler;
+    const Bytef *buf;
+    uInt len;
+{
+    return adler32_z(adler, buf, len);
+}
+
+/* ========================================================================= */
+local uLong adler32_combine_(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off64_t len2;
+{
+    unsigned long sum1;
+    unsigned long sum2;
+    unsigned rem;
+
+    /* for negative len, return invalid adler32 as a clue for debugging */
+    if (len2 < 0)
+        return 0xffffffffUL;
+
+    /* the derivation of this formula is left as an exercise for the reader */
+    MOD63(len2);                /* assumes len2 >= 0 */
+    rem = (unsigned)len2;
+    sum1 = adler1 & 0xffff;
+    sum2 = rem * sum1;
+    MOD(sum2);
+    sum1 += (adler2 & 0xffff) + BASE - 1;
+    sum2 += ((adler1 >> 16) & 0xffff) + ((adler2 >> 16) & 0xffff) + BASE - rem;
+    if (sum1 >= BASE) sum1 -= BASE;
+    if (sum1 >= BASE) sum1 -= BASE;
+    if (sum2 >= ((unsigned long)BASE << 1)) sum2 -= ((unsigned long)BASE << 1);
+    if (sum2 >= BASE) sum2 -= BASE;
+    return sum1 | (sum2 << 16);
+}
+
+/* ========================================================================= */
+uLong ZEXPORT adler32_combine(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off_t len2;
+{
+    return adler32_combine_(adler1, adler2, len2);
+}
+
+uLong ZEXPORT adler32_combine64(adler1, adler2, len2)
+    uLong adler1;
+    uLong adler2;
+    z_off64_t len2;
+{
+    return adler32_combine_(adler1, adler2, len2);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/compress.c glmark2-2021.02/src/zlib/compress.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/compress.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/compress.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,86 @@
+/* compress.c -- compress a memory buffer
+ * Copyright (C) 1995-2005, 2014, 2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#define ZLIB_INTERNAL
+#include "zlib.h"
+
+/* ===========================================================================
+     Compresses the source buffer into the destination buffer. The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer. Upon entry, destLen is the total size of the
+   destination buffer, which must be at least 0.1% larger than sourceLen plus
+   12 bytes. Upon exit, destLen is the actual size of the compressed buffer.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+int ZEXPORT compress2 (dest, destLen, source, sourceLen, level)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+    int level;
+{
+    z_stream stream;
+    int err;
+    const uInt max = (uInt)-1;
+    uLong left;
+
+    left = *destLen;
+    *destLen = 0;
+
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+    stream.opaque = (voidpf)0;
+
+    err = deflateInit(&stream, level);
+    if (err != Z_OK) return err;
+
+    stream.next_out = dest;
+    stream.avail_out = 0;
+    stream.next_in = (z_const Bytef *)source;
+    stream.avail_in = 0;
+
+    do {
+        if (stream.avail_out == 0) {
+            stream.avail_out = left > (uLong)max ? max : (uInt)left;
+            left -= stream.avail_out;
+        }
+        if (stream.avail_in == 0) {
+            stream.avail_in = sourceLen > (uLong)max ? max : (uInt)sourceLen;
+            sourceLen -= stream.avail_in;
+        }
+        err = deflate(&stream, sourceLen ? Z_NO_FLUSH : Z_FINISH);
+    } while (err == Z_OK);
+
+    *destLen = stream.total_out;
+    deflateEnd(&stream);
+    return err == Z_STREAM_END ? Z_OK : err;
+}
+
+/* ===========================================================================
+ */
+int ZEXPORT compress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    return compress2(dest, destLen, source, sourceLen, Z_DEFAULT_COMPRESSION);
+}
+
+/* ===========================================================================
+     If the default memLevel or windowBits for deflateInit() is changed, then
+   this function needs to be updated.
+ */
+uLong ZEXPORT compressBound (sourceLen)
+    uLong sourceLen;
+{
+    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
+           (sourceLen >> 25) + 13;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/crc32.c glmark2-2021.02/src/zlib/crc32.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/crc32.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/crc32.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,442 @@
+/* crc32.c -- compute the CRC-32 of a data stream
+ * Copyright (C) 1995-2006, 2010, 2011, 2012, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ *
+ * Thanks to Rodney Brown <rbrown64@csc.com.au> for his contribution of faster
+ * CRC methods: exclusive-oring 32 bits of data at a time, and pre-computing
+ * tables for updating the shift register in one step with three exclusive-ors
+ * instead of four steps with four exclusive-ors.  This results in about a
+ * factor of two increase in speed on a Power PC G4 (PPC7455) using gcc -O3.
+ */
+
+/* @(#) $Id$ */
+
+/*
+  Note on the use of DYNAMIC_CRC_TABLE: there is no mutex or semaphore
+  protection on the static variables used to control the first-use generation
+  of the crc tables.  Therefore, if you #define DYNAMIC_CRC_TABLE, you should
+  first call get_crc_table() to initialize the tables before allowing more than
+  one thread to use crc32().
+
+  DYNAMIC_CRC_TABLE and MAKECRCH can be #defined to write out crc32.h.
+ */
+
+#ifdef MAKECRCH
+#  include <stdio.h>
+#  ifndef DYNAMIC_CRC_TABLE
+#    define DYNAMIC_CRC_TABLE
+#  endif /* !DYNAMIC_CRC_TABLE */
+#endif /* MAKECRCH */
+
+#include "zutil.h"      /* for STDC and FAR definitions */
+
+/* Definitions for doing the crc four data bytes at a time. */
+#if !defined(NOBYFOUR) && defined(Z_U4)
+#  define BYFOUR
+#endif
+#ifdef BYFOUR
+   local unsigned long crc32_little OF((unsigned long,
+                        const unsigned char FAR *, z_size_t));
+   local unsigned long crc32_big OF((unsigned long,
+                        const unsigned char FAR *, z_size_t));
+#  define TBLS 8
+#else
+#  define TBLS 1
+#endif /* BYFOUR */
+
+/* Local functions for crc concatenation */
+local unsigned long gf2_matrix_times OF((unsigned long *mat,
+                                         unsigned long vec));
+local void gf2_matrix_square OF((unsigned long *square, unsigned long *mat));
+local uLong crc32_combine_ OF((uLong crc1, uLong crc2, z_off64_t len2));
+
+
+#ifdef DYNAMIC_CRC_TABLE
+
+local volatile int crc_table_empty = 1;
+local z_crc_t FAR crc_table[TBLS][256];
+local void make_crc_table OF((void));
+#ifdef MAKECRCH
+   local void write_table OF((FILE *, const z_crc_t FAR *));
+#endif /* MAKECRCH */
+/*
+  Generate tables for a byte-wise 32-bit CRC calculation on the polynomial:
+  x^32+x^26+x^23+x^22+x^16+x^12+x^11+x^10+x^8+x^7+x^5+x^4+x^2+x+1.
+
+  Polynomials over GF(2) are represented in binary, one bit per coefficient,
+  with the lowest powers in the most significant bit.  Then adding polynomials
+  is just exclusive-or, and multiplying a polynomial by x is a right shift by
+  one.  If we call the above polynomial p, and represent a byte as the
+  polynomial q, also with the lowest power in the most significant bit (so the
+  byte 0xb1 is the polynomial x^7+x^3+x+1), then the CRC is (q*x^32) mod p,
+  where a mod b means the remainder after dividing a by b.
+
+  This calculation is done using the shift-register method of multiplying and
+  taking the remainder.  The register is initialized to zero, and for each
+  incoming bit, x^32 is added mod p to the register if the bit is a one (where
+  x^32 mod p is p+x^32 = x^26+...+1), and the register is multiplied mod p by
+  x (which is shifting right by one and adding x^32 mod p if the bit shifted
+  out is a one).  We start with the highest power (least significant bit) of
+  q and repeat for all eight bits of q.
+
+  The first table is simply the CRC of all possible eight bit values.  This is
+  all the information needed to generate CRCs on data a byte at a time for all
+  combinations of CRC register values and incoming bytes.  The remaining tables
+  allow for word-at-a-time CRC calculation for both big-endian and little-
+  endian machines, where a word is four bytes.
+*/
+local void make_crc_table()
+{
+    z_crc_t c;
+    int n, k;
+    z_crc_t poly;                       /* polynomial exclusive-or pattern */
+    /* terms of polynomial defining this crc (except x^32): */
+    static volatile int first = 1;      /* flag to limit concurrent making */
+    static const unsigned char p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
+
+    /* See if another task is already doing this (not thread-safe, but better
+       than nothing -- significantly reduces duration of vulnerability in
+       case the advice about DYNAMIC_CRC_TABLE is ignored) */
+    if (first) {
+        first = 0;
+
+        /* make exclusive-or pattern from polynomial (0xedb88320UL) */
+        poly = 0;
+        for (n = 0; n < (int)(sizeof(p)/sizeof(unsigned char)); n++)
+            poly |= (z_crc_t)1 << (31 - p[n]);
+
+        /* generate a crc for every 8-bit value */
+        for (n = 0; n < 256; n++) {
+            c = (z_crc_t)n;
+            for (k = 0; k < 8; k++)
+                c = c & 1 ? poly ^ (c >> 1) : c >> 1;
+            crc_table[0][n] = c;
+        }
+
+#ifdef BYFOUR
+        /* generate crc for each value followed by one, two, and three zeros,
+           and then the byte reversal of those as well as the first table */
+        for (n = 0; n < 256; n++) {
+            c = crc_table[0][n];
+            crc_table[4][n] = ZSWAP32(c);
+            for (k = 1; k < 4; k++) {
+                c = crc_table[0][c & 0xff] ^ (c >> 8);
+                crc_table[k][n] = c;
+                crc_table[k + 4][n] = ZSWAP32(c);
+            }
+        }
+#endif /* BYFOUR */
+
+        crc_table_empty = 0;
+    }
+    else {      /* not first */
+        /* wait for the other guy to finish (not efficient, but rare) */
+        while (crc_table_empty)
+            ;
+    }
+
+#ifdef MAKECRCH
+    /* write out CRC tables to crc32.h */
+    {
+        FILE *out;
+
+        out = fopen("crc32.h", "w");
+        if (out == NULL) return;
+        fprintf(out, "/* crc32.h -- tables for rapid CRC calculation\n");
+        fprintf(out, " * Generated automatically by crc32.c\n */\n\n");
+        fprintf(out, "local const z_crc_t FAR ");
+        fprintf(out, "crc_table[TBLS][256] =\n{\n  {\n");
+        write_table(out, crc_table[0]);
+#  ifdef BYFOUR
+        fprintf(out, "#ifdef BYFOUR\n");
+        for (k = 1; k < 8; k++) {
+            fprintf(out, "  },\n  {\n");
+            write_table(out, crc_table[k]);
+        }
+        fprintf(out, "#endif\n");
+#  endif /* BYFOUR */
+        fprintf(out, "  }\n};\n");
+        fclose(out);
+    }
+#endif /* MAKECRCH */
+}
+
+#ifdef MAKECRCH
+local void write_table(out, table)
+    FILE *out;
+    const z_crc_t FAR *table;
+{
+    int n;
+
+    for (n = 0; n < 256; n++)
+        fprintf(out, "%s0x%08lxUL%s", n % 5 ? "" : "    ",
+                (unsigned long)(table[n]),
+                n == 255 ? "\n" : (n % 5 == 4 ? ",\n" : ", "));
+}
+#endif /* MAKECRCH */
+
+#else /* !DYNAMIC_CRC_TABLE */
+/* ========================================================================
+ * Tables of CRC-32s of all single-byte values, made by make_crc_table().
+ */
+#include "crc32.h"
+#endif /* DYNAMIC_CRC_TABLE */
+
+/* =========================================================================
+ * This function can be used by asm versions of crc32()
+ */
+const z_crc_t FAR * ZEXPORT get_crc_table()
+{
+#ifdef DYNAMIC_CRC_TABLE
+    if (crc_table_empty)
+        make_crc_table();
+#endif /* DYNAMIC_CRC_TABLE */
+    return (const z_crc_t FAR *)crc_table;
+}
+
+/* ========================================================================= */
+#define DO1 crc = crc_table[0][((int)crc ^ (*buf++)) & 0xff] ^ (crc >> 8)
+#define DO8 DO1; DO1; DO1; DO1; DO1; DO1; DO1; DO1
+
+/* ========================================================================= */
+unsigned long ZEXPORT crc32_z(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    if (buf == Z_NULL) return 0UL;
+
+#ifdef DYNAMIC_CRC_TABLE
+    if (crc_table_empty)
+        make_crc_table();
+#endif /* DYNAMIC_CRC_TABLE */
+
+#ifdef BYFOUR
+    if (sizeof(void *) == sizeof(ptrdiff_t)) {
+        z_crc_t endian;
+
+        endian = 1;
+        if (*((unsigned char *)(&endian)))
+            return crc32_little(crc, buf, len);
+        else
+            return crc32_big(crc, buf, len);
+    }
+#endif /* BYFOUR */
+    crc = crc ^ 0xffffffffUL;
+    while (len >= 8) {
+        DO8;
+        len -= 8;
+    }
+    if (len) do {
+        DO1;
+    } while (--len);
+    return crc ^ 0xffffffffUL;
+}
+
+/* ========================================================================= */
+unsigned long ZEXPORT crc32(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    uInt len;
+{
+    return crc32_z(crc, buf, len);
+}
+
+#ifdef BYFOUR
+
+/*
+   This BYFOUR code accesses the passed unsigned char * buffer with a 32-bit
+   integer pointer type. This violates the strict aliasing rule, where a
+   compiler can assume, for optimization purposes, that two pointers to
+   fundamentally different types won't ever point to the same memory. This can
+   manifest as a problem only if one of the pointers is written to. This code
+   only reads from those pointers. So long as this code remains isolated in
+   this compilation unit, there won't be a problem. For this reason, this code
+   should not be copied and pasted into a compilation unit in which other code
+   writes to the buffer that is passed to these routines.
+ */
+
+/* ========================================================================= */
+#define DOLIT4 c ^= *buf4++; \
+        c = crc_table[3][c & 0xff] ^ crc_table[2][(c >> 8) & 0xff] ^ \
+            crc_table[1][(c >> 16) & 0xff] ^ crc_table[0][c >> 24]
+#define DOLIT32 DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4; DOLIT4
+
+/* ========================================================================= */
+local unsigned long crc32_little(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    register z_crc_t c;
+    register const z_crc_t FAR *buf4;
+
+    c = (z_crc_t)crc;
+    c = ~c;
+    while (len && ((ptrdiff_t)buf & 3)) {
+        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
+        len--;
+    }
+
+    buf4 = (const z_crc_t FAR *)(const void FAR *)buf;
+    while (len >= 32) {
+        DOLIT32;
+        len -= 32;
+    }
+    while (len >= 4) {
+        DOLIT4;
+        len -= 4;
+    }
+    buf = (const unsigned char FAR *)buf4;
+
+    if (len) do {
+        c = crc_table[0][(c ^ *buf++) & 0xff] ^ (c >> 8);
+    } while (--len);
+    c = ~c;
+    return (unsigned long)c;
+}
+
+/* ========================================================================= */
+#define DOBIG4 c ^= *buf4++; \
+        c = crc_table[4][c & 0xff] ^ crc_table[5][(c >> 8) & 0xff] ^ \
+            crc_table[6][(c >> 16) & 0xff] ^ crc_table[7][c >> 24]
+#define DOBIG32 DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4; DOBIG4
+
+/* ========================================================================= */
+local unsigned long crc32_big(crc, buf, len)
+    unsigned long crc;
+    const unsigned char FAR *buf;
+    z_size_t len;
+{
+    register z_crc_t c;
+    register const z_crc_t FAR *buf4;
+
+    c = ZSWAP32((z_crc_t)crc);
+    c = ~c;
+    while (len && ((ptrdiff_t)buf & 3)) {
+        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
+        len--;
+    }
+
+    buf4 = (const z_crc_t FAR *)(const void FAR *)buf;
+    while (len >= 32) {
+        DOBIG32;
+        len -= 32;
+    }
+    while (len >= 4) {
+        DOBIG4;
+        len -= 4;
+    }
+    buf = (const unsigned char FAR *)buf4;
+
+    if (len) do {
+        c = crc_table[4][(c >> 24) ^ *buf++] ^ (c << 8);
+    } while (--len);
+    c = ~c;
+    return (unsigned long)(ZSWAP32(c));
+}
+
+#endif /* BYFOUR */
+
+#define GF2_DIM 32      /* dimension of GF(2) vectors (length of CRC) */
+
+/* ========================================================================= */
+local unsigned long gf2_matrix_times(mat, vec)
+    unsigned long *mat;
+    unsigned long vec;
+{
+    unsigned long sum;
+
+    sum = 0;
+    while (vec) {
+        if (vec & 1)
+            sum ^= *mat;
+        vec >>= 1;
+        mat++;
+    }
+    return sum;
+}
+
+/* ========================================================================= */
+local void gf2_matrix_square(square, mat)
+    unsigned long *square;
+    unsigned long *mat;
+{
+    int n;
+
+    for (n = 0; n < GF2_DIM; n++)
+        square[n] = gf2_matrix_times(mat, mat[n]);
+}
+
+/* ========================================================================= */
+local uLong crc32_combine_(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off64_t len2;
+{
+    int n;
+    unsigned long row;
+    unsigned long even[GF2_DIM];    /* even-power-of-two zeros operator */
+    unsigned long odd[GF2_DIM];     /* odd-power-of-two zeros operator */
+
+    /* degenerate case (also disallow negative lengths) */
+    if (len2 <= 0)
+        return crc1;
+
+    /* put operator for one zero bit in odd */
+    odd[0] = 0xedb88320UL;          /* CRC-32 polynomial */
+    row = 1;
+    for (n = 1; n < GF2_DIM; n++) {
+        odd[n] = row;
+        row <<= 1;
+    }
+
+    /* put operator for two zero bits in even */
+    gf2_matrix_square(even, odd);
+
+    /* put operator for four zero bits in odd */
+    gf2_matrix_square(odd, even);
+
+    /* apply len2 zeros to crc1 (first square will put the operator for one
+       zero byte, eight zero bits, in even) */
+    do {
+        /* apply zeros operator for this bit of len2 */
+        gf2_matrix_square(even, odd);
+        if (len2 & 1)
+            crc1 = gf2_matrix_times(even, crc1);
+        len2 >>= 1;
+
+        /* if no more bits set, then done */
+        if (len2 == 0)
+            break;
+
+        /* another iteration of the loop with odd and even swapped */
+        gf2_matrix_square(odd, even);
+        if (len2 & 1)
+            crc1 = gf2_matrix_times(odd, crc1);
+        len2 >>= 1;
+
+        /* if no more bits set, then done */
+    } while (len2 != 0);
+
+    /* return combined crc */
+    crc1 ^= crc2;
+    return crc1;
+}
+
+/* ========================================================================= */
+uLong ZEXPORT crc32_combine(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off_t len2;
+{
+    return crc32_combine_(crc1, crc2, len2);
+}
+
+uLong ZEXPORT crc32_combine64(crc1, crc2, len2)
+    uLong crc1;
+    uLong crc2;
+    z_off64_t len2;
+{
+    return crc32_combine_(crc1, crc2, len2);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/crc32.h glmark2-2021.02/src/zlib/crc32.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/crc32.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/crc32.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,441 @@
+/* crc32.h -- tables for rapid CRC calculation
+ * Generated automatically by crc32.c
+ */
+
+local const z_crc_t FAR crc_table[TBLS][256] =
+{
+  {
+    0x00000000UL, 0x77073096UL, 0xee0e612cUL, 0x990951baUL, 0x076dc419UL,
+    0x706af48fUL, 0xe963a535UL, 0x9e6495a3UL, 0x0edb8832UL, 0x79dcb8a4UL,
+    0xe0d5e91eUL, 0x97d2d988UL, 0x09b64c2bUL, 0x7eb17cbdUL, 0xe7b82d07UL,
+    0x90bf1d91UL, 0x1db71064UL, 0x6ab020f2UL, 0xf3b97148UL, 0x84be41deUL,
+    0x1adad47dUL, 0x6ddde4ebUL, 0xf4d4b551UL, 0x83d385c7UL, 0x136c9856UL,
+    0x646ba8c0UL, 0xfd62f97aUL, 0x8a65c9ecUL, 0x14015c4fUL, 0x63066cd9UL,
+    0xfa0f3d63UL, 0x8d080df5UL, 0x3b6e20c8UL, 0x4c69105eUL, 0xd56041e4UL,
+    0xa2677172UL, 0x3c03e4d1UL, 0x4b04d447UL, 0xd20d85fdUL, 0xa50ab56bUL,
+    0x35b5a8faUL, 0x42b2986cUL, 0xdbbbc9d6UL, 0xacbcf940UL, 0x32d86ce3UL,
+    0x45df5c75UL, 0xdcd60dcfUL, 0xabd13d59UL, 0x26d930acUL, 0x51de003aUL,
+    0xc8d75180UL, 0xbfd06116UL, 0x21b4f4b5UL, 0x56b3c423UL, 0xcfba9599UL,
+    0xb8bda50fUL, 0x2802b89eUL, 0x5f058808UL, 0xc60cd9b2UL, 0xb10be924UL,
+    0x2f6f7c87UL, 0x58684c11UL, 0xc1611dabUL, 0xb6662d3dUL, 0x76dc4190UL,
+    0x01db7106UL, 0x98d220bcUL, 0xefd5102aUL, 0x71b18589UL, 0x06b6b51fUL,
+    0x9fbfe4a5UL, 0xe8b8d433UL, 0x7807c9a2UL, 0x0f00f934UL, 0x9609a88eUL,
+    0xe10e9818UL, 0x7f6a0dbbUL, 0x086d3d2dUL, 0x91646c97UL, 0xe6635c01UL,
+    0x6b6b51f4UL, 0x1c6c6162UL, 0x856530d8UL, 0xf262004eUL, 0x6c0695edUL,
+    0x1b01a57bUL, 0x8208f4c1UL, 0xf50fc457UL, 0x65b0d9c6UL, 0x12b7e950UL,
+    0x8bbeb8eaUL, 0xfcb9887cUL, 0x62dd1ddfUL, 0x15da2d49UL, 0x8cd37cf3UL,
+    0xfbd44c65UL, 0x4db26158UL, 0x3ab551ceUL, 0xa3bc0074UL, 0xd4bb30e2UL,
+    0x4adfa541UL, 0x3dd895d7UL, 0xa4d1c46dUL, 0xd3d6f4fbUL, 0x4369e96aUL,
+    0x346ed9fcUL, 0xad678846UL, 0xda60b8d0UL, 0x44042d73UL, 0x33031de5UL,
+    0xaa0a4c5fUL, 0xdd0d7cc9UL, 0x5005713cUL, 0x270241aaUL, 0xbe0b1010UL,
+    0xc90c2086UL, 0x5768b525UL, 0x206f85b3UL, 0xb966d409UL, 0xce61e49fUL,
+    0x5edef90eUL, 0x29d9c998UL, 0xb0d09822UL, 0xc7d7a8b4UL, 0x59b33d17UL,
+    0x2eb40d81UL, 0xb7bd5c3bUL, 0xc0ba6cadUL, 0xedb88320UL, 0x9abfb3b6UL,
+    0x03b6e20cUL, 0x74b1d29aUL, 0xead54739UL, 0x9dd277afUL, 0x04db2615UL,
+    0x73dc1683UL, 0xe3630b12UL, 0x94643b84UL, 0x0d6d6a3eUL, 0x7a6a5aa8UL,
+    0xe40ecf0bUL, 0x9309ff9dUL, 0x0a00ae27UL, 0x7d079eb1UL, 0xf00f9344UL,
+    0x8708a3d2UL, 0x1e01f268UL, 0x6906c2feUL, 0xf762575dUL, 0x806567cbUL,
+    0x196c3671UL, 0x6e6b06e7UL, 0xfed41b76UL, 0x89d32be0UL, 0x10da7a5aUL,
+    0x67dd4accUL, 0xf9b9df6fUL, 0x8ebeeff9UL, 0x17b7be43UL, 0x60b08ed5UL,
+    0xd6d6a3e8UL, 0xa1d1937eUL, 0x38d8c2c4UL, 0x4fdff252UL, 0xd1bb67f1UL,
+    0xa6bc5767UL, 0x3fb506ddUL, 0x48b2364bUL, 0xd80d2bdaUL, 0xaf0a1b4cUL,
+    0x36034af6UL, 0x41047a60UL, 0xdf60efc3UL, 0xa867df55UL, 0x316e8eefUL,
+    0x4669be79UL, 0xcb61b38cUL, 0xbc66831aUL, 0x256fd2a0UL, 0x5268e236UL,
+    0xcc0c7795UL, 0xbb0b4703UL, 0x220216b9UL, 0x5505262fUL, 0xc5ba3bbeUL,
+    0xb2bd0b28UL, 0x2bb45a92UL, 0x5cb36a04UL, 0xc2d7ffa7UL, 0xb5d0cf31UL,
+    0x2cd99e8bUL, 0x5bdeae1dUL, 0x9b64c2b0UL, 0xec63f226UL, 0x756aa39cUL,
+    0x026d930aUL, 0x9c0906a9UL, 0xeb0e363fUL, 0x72076785UL, 0x05005713UL,
+    0x95bf4a82UL, 0xe2b87a14UL, 0x7bb12baeUL, 0x0cb61b38UL, 0x92d28e9bUL,
+    0xe5d5be0dUL, 0x7cdcefb7UL, 0x0bdbdf21UL, 0x86d3d2d4UL, 0xf1d4e242UL,
+    0x68ddb3f8UL, 0x1fda836eUL, 0x81be16cdUL, 0xf6b9265bUL, 0x6fb077e1UL,
+    0x18b74777UL, 0x88085ae6UL, 0xff0f6a70UL, 0x66063bcaUL, 0x11010b5cUL,
+    0x8f659effUL, 0xf862ae69UL, 0x616bffd3UL, 0x166ccf45UL, 0xa00ae278UL,
+    0xd70dd2eeUL, 0x4e048354UL, 0x3903b3c2UL, 0xa7672661UL, 0xd06016f7UL,
+    0x4969474dUL, 0x3e6e77dbUL, 0xaed16a4aUL, 0xd9d65adcUL, 0x40df0b66UL,
+    0x37d83bf0UL, 0xa9bcae53UL, 0xdebb9ec5UL, 0x47b2cf7fUL, 0x30b5ffe9UL,
+    0xbdbdf21cUL, 0xcabac28aUL, 0x53b39330UL, 0x24b4a3a6UL, 0xbad03605UL,
+    0xcdd70693UL, 0x54de5729UL, 0x23d967bfUL, 0xb3667a2eUL, 0xc4614ab8UL,
+    0x5d681b02UL, 0x2a6f2b94UL, 0xb40bbe37UL, 0xc30c8ea1UL, 0x5a05df1bUL,
+    0x2d02ef8dUL
+#ifdef BYFOUR
+  },
+  {
+    0x00000000UL, 0x191b3141UL, 0x32366282UL, 0x2b2d53c3UL, 0x646cc504UL,
+    0x7d77f445UL, 0x565aa786UL, 0x4f4196c7UL, 0xc8d98a08UL, 0xd1c2bb49UL,
+    0xfaefe88aUL, 0xe3f4d9cbUL, 0xacb54f0cUL, 0xb5ae7e4dUL, 0x9e832d8eUL,
+    0x87981ccfUL, 0x4ac21251UL, 0x53d92310UL, 0x78f470d3UL, 0x61ef4192UL,
+    0x2eaed755UL, 0x37b5e614UL, 0x1c98b5d7UL, 0x05838496UL, 0x821b9859UL,
+    0x9b00a918UL, 0xb02dfadbUL, 0xa936cb9aUL, 0xe6775d5dUL, 0xff6c6c1cUL,
+    0xd4413fdfUL, 0xcd5a0e9eUL, 0x958424a2UL, 0x8c9f15e3UL, 0xa7b24620UL,
+    0xbea97761UL, 0xf1e8e1a6UL, 0xe8f3d0e7UL, 0xc3de8324UL, 0xdac5b265UL,
+    0x5d5daeaaUL, 0x44469febUL, 0x6f6bcc28UL, 0x7670fd69UL, 0x39316baeUL,
+    0x202a5aefUL, 0x0b07092cUL, 0x121c386dUL, 0xdf4636f3UL, 0xc65d07b2UL,
+    0xed705471UL, 0xf46b6530UL, 0xbb2af3f7UL, 0xa231c2b6UL, 0x891c9175UL,
+    0x9007a034UL, 0x179fbcfbUL, 0x0e848dbaUL, 0x25a9de79UL, 0x3cb2ef38UL,
+    0x73f379ffUL, 0x6ae848beUL, 0x41c51b7dUL, 0x58de2a3cUL, 0xf0794f05UL,
+    0xe9627e44UL, 0xc24f2d87UL, 0xdb541cc6UL, 0x94158a01UL, 0x8d0ebb40UL,
+    0xa623e883UL, 0xbf38d9c2UL, 0x38a0c50dUL, 0x21bbf44cUL, 0x0a96a78fUL,
+    0x138d96ceUL, 0x5ccc0009UL, 0x45d73148UL, 0x6efa628bUL, 0x77e153caUL,
+    0xbabb5d54UL, 0xa3a06c15UL, 0x888d3fd6UL, 0x91960e97UL, 0xded79850UL,
+    0xc7cca911UL, 0xece1fad2UL, 0xf5facb93UL, 0x7262d75cUL, 0x6b79e61dUL,
+    0x4054b5deUL, 0x594f849fUL, 0x160e1258UL, 0x0f152319UL, 0x243870daUL,
+    0x3d23419bUL, 0x65fd6ba7UL, 0x7ce65ae6UL, 0x57cb0925UL, 0x4ed03864UL,
+    0x0191aea3UL, 0x188a9fe2UL, 0x33a7cc21UL, 0x2abcfd60UL, 0xad24e1afUL,
+    0xb43fd0eeUL, 0x9f12832dUL, 0x8609b26cUL, 0xc94824abUL, 0xd05315eaUL,
+    0xfb7e4629UL, 0xe2657768UL, 0x2f3f79f6UL, 0x362448b7UL, 0x1d091b74UL,
+    0x04122a35UL, 0x4b53bcf2UL, 0x52488db3UL, 0x7965de70UL, 0x607eef31UL,
+    0xe7e6f3feUL, 0xfefdc2bfUL, 0xd5d0917cUL, 0xcccba03dUL, 0x838a36faUL,
+    0x9a9107bbUL, 0xb1bc5478UL, 0xa8a76539UL, 0x3b83984bUL, 0x2298a90aUL,
+    0x09b5fac9UL, 0x10aecb88UL, 0x5fef5d4fUL, 0x46f46c0eUL, 0x6dd93fcdUL,
+    0x74c20e8cUL, 0xf35a1243UL, 0xea412302UL, 0xc16c70c1UL, 0xd8774180UL,
+    0x9736d747UL, 0x8e2de606UL, 0xa500b5c5UL, 0xbc1b8484UL, 0x71418a1aUL,
+    0x685abb5bUL, 0x4377e898UL, 0x5a6cd9d9UL, 0x152d4f1eUL, 0x0c367e5fUL,
+    0x271b2d9cUL, 0x3e001cddUL, 0xb9980012UL, 0xa0833153UL, 0x8bae6290UL,
+    0x92b553d1UL, 0xddf4c516UL, 0xc4eff457UL, 0xefc2a794UL, 0xf6d996d5UL,
+    0xae07bce9UL, 0xb71c8da8UL, 0x9c31de6bUL, 0x852aef2aUL, 0xca6b79edUL,
+    0xd37048acUL, 0xf85d1b6fUL, 0xe1462a2eUL, 0x66de36e1UL, 0x7fc507a0UL,
+    0x54e85463UL, 0x4df36522UL, 0x02b2f3e5UL, 0x1ba9c2a4UL, 0x30849167UL,
+    0x299fa026UL, 0xe4c5aeb8UL, 0xfdde9ff9UL, 0xd6f3cc3aUL, 0xcfe8fd7bUL,
+    0x80a96bbcUL, 0x99b25afdUL, 0xb29f093eUL, 0xab84387fUL, 0x2c1c24b0UL,
+    0x350715f1UL, 0x1e2a4632UL, 0x07317773UL, 0x4870e1b4UL, 0x516bd0f5UL,
+    0x7a468336UL, 0x635db277UL, 0xcbfad74eUL, 0xd2e1e60fUL, 0xf9ccb5ccUL,
+    0xe0d7848dUL, 0xaf96124aUL, 0xb68d230bUL, 0x9da070c8UL, 0x84bb4189UL,
+    0x03235d46UL, 0x1a386c07UL, 0x31153fc4UL, 0x280e0e85UL, 0x674f9842UL,
+    0x7e54a903UL, 0x5579fac0UL, 0x4c62cb81UL, 0x8138c51fUL, 0x9823f45eUL,
+    0xb30ea79dUL, 0xaa1596dcUL, 0xe554001bUL, 0xfc4f315aUL, 0xd7626299UL,
+    0xce7953d8UL, 0x49e14f17UL, 0x50fa7e56UL, 0x7bd72d95UL, 0x62cc1cd4UL,
+    0x2d8d8a13UL, 0x3496bb52UL, 0x1fbbe891UL, 0x06a0d9d0UL, 0x5e7ef3ecUL,
+    0x4765c2adUL, 0x6c48916eUL, 0x7553a02fUL, 0x3a1236e8UL, 0x230907a9UL,
+    0x0824546aUL, 0x113f652bUL, 0x96a779e4UL, 0x8fbc48a5UL, 0xa4911b66UL,
+    0xbd8a2a27UL, 0xf2cbbce0UL, 0xebd08da1UL, 0xc0fdde62UL, 0xd9e6ef23UL,
+    0x14bce1bdUL, 0x0da7d0fcUL, 0x268a833fUL, 0x3f91b27eUL, 0x70d024b9UL,
+    0x69cb15f8UL, 0x42e6463bUL, 0x5bfd777aUL, 0xdc656bb5UL, 0xc57e5af4UL,
+    0xee530937UL, 0xf7483876UL, 0xb809aeb1UL, 0xa1129ff0UL, 0x8a3fcc33UL,
+    0x9324fd72UL
+  },
+  {
+    0x00000000UL, 0x01c26a37UL, 0x0384d46eUL, 0x0246be59UL, 0x0709a8dcUL,
+    0x06cbc2ebUL, 0x048d7cb2UL, 0x054f1685UL, 0x0e1351b8UL, 0x0fd13b8fUL,
+    0x0d9785d6UL, 0x0c55efe1UL, 0x091af964UL, 0x08d89353UL, 0x0a9e2d0aUL,
+    0x0b5c473dUL, 0x1c26a370UL, 0x1de4c947UL, 0x1fa2771eUL, 0x1e601d29UL,
+    0x1b2f0bacUL, 0x1aed619bUL, 0x18abdfc2UL, 0x1969b5f5UL, 0x1235f2c8UL,
+    0x13f798ffUL, 0x11b126a6UL, 0x10734c91UL, 0x153c5a14UL, 0x14fe3023UL,
+    0x16b88e7aUL, 0x177ae44dUL, 0x384d46e0UL, 0x398f2cd7UL, 0x3bc9928eUL,
+    0x3a0bf8b9UL, 0x3f44ee3cUL, 0x3e86840bUL, 0x3cc03a52UL, 0x3d025065UL,
+    0x365e1758UL, 0x379c7d6fUL, 0x35dac336UL, 0x3418a901UL, 0x3157bf84UL,
+    0x3095d5b3UL, 0x32d36beaUL, 0x331101ddUL, 0x246be590UL, 0x25a98fa7UL,
+    0x27ef31feUL, 0x262d5bc9UL, 0x23624d4cUL, 0x22a0277bUL, 0x20e69922UL,
+    0x2124f315UL, 0x2a78b428UL, 0x2bbade1fUL, 0x29fc6046UL, 0x283e0a71UL,
+    0x2d711cf4UL, 0x2cb376c3UL, 0x2ef5c89aUL, 0x2f37a2adUL, 0x709a8dc0UL,
+    0x7158e7f7UL, 0x731e59aeUL, 0x72dc3399UL, 0x7793251cUL, 0x76514f2bUL,
+    0x7417f172UL, 0x75d59b45UL, 0x7e89dc78UL, 0x7f4bb64fUL, 0x7d0d0816UL,
+    0x7ccf6221UL, 0x798074a4UL, 0x78421e93UL, 0x7a04a0caUL, 0x7bc6cafdUL,
+    0x6cbc2eb0UL, 0x6d7e4487UL, 0x6f38fadeUL, 0x6efa90e9UL, 0x6bb5866cUL,
+    0x6a77ec5bUL, 0x68315202UL, 0x69f33835UL, 0x62af7f08UL, 0x636d153fUL,
+    0x612bab66UL, 0x60e9c151UL, 0x65a6d7d4UL, 0x6464bde3UL, 0x662203baUL,
+    0x67e0698dUL, 0x48d7cb20UL, 0x4915a117UL, 0x4b531f4eUL, 0x4a917579UL,
+    0x4fde63fcUL, 0x4e1c09cbUL, 0x4c5ab792UL, 0x4d98dda5UL, 0x46c49a98UL,
+    0x4706f0afUL, 0x45404ef6UL, 0x448224c1UL, 0x41cd3244UL, 0x400f5873UL,
+    0x4249e62aUL, 0x438b8c1dUL, 0x54f16850UL, 0x55330267UL, 0x5775bc3eUL,
+    0x56b7d609UL, 0x53f8c08cUL, 0x523aaabbUL, 0x507c14e2UL, 0x51be7ed5UL,
+    0x5ae239e8UL, 0x5b2053dfUL, 0x5966ed86UL, 0x58a487b1UL, 0x5deb9134UL,
+    0x5c29fb03UL, 0x5e6f455aUL, 0x5fad2f6dUL, 0xe1351b80UL, 0xe0f771b7UL,
+    0xe2b1cfeeUL, 0xe373a5d9UL, 0xe63cb35cUL, 0xe7fed96bUL, 0xe5b86732UL,
+    0xe47a0d05UL, 0xef264a38UL, 0xeee4200fUL, 0xeca29e56UL, 0xed60f461UL,
+    0xe82fe2e4UL, 0xe9ed88d3UL, 0xebab368aUL, 0xea695cbdUL, 0xfd13b8f0UL,
+    0xfcd1d2c7UL, 0xfe976c9eUL, 0xff5506a9UL, 0xfa1a102cUL, 0xfbd87a1bUL,
+    0xf99ec442UL, 0xf85cae75UL, 0xf300e948UL, 0xf2c2837fUL, 0xf0843d26UL,
+    0xf1465711UL, 0xf4094194UL, 0xf5cb2ba3UL, 0xf78d95faUL, 0xf64fffcdUL,
+    0xd9785d60UL, 0xd8ba3757UL, 0xdafc890eUL, 0xdb3ee339UL, 0xde71f5bcUL,
+    0xdfb39f8bUL, 0xddf521d2UL, 0xdc374be5UL, 0xd76b0cd8UL, 0xd6a966efUL,
+    0xd4efd8b6UL, 0xd52db281UL, 0xd062a404UL, 0xd1a0ce33UL, 0xd3e6706aUL,
+    0xd2241a5dUL, 0xc55efe10UL, 0xc49c9427UL, 0xc6da2a7eUL, 0xc7184049UL,
+    0xc25756ccUL, 0xc3953cfbUL, 0xc1d382a2UL, 0xc011e895UL, 0xcb4dafa8UL,
+    0xca8fc59fUL, 0xc8c97bc6UL, 0xc90b11f1UL, 0xcc440774UL, 0xcd866d43UL,
+    0xcfc0d31aUL, 0xce02b92dUL, 0x91af9640UL, 0x906dfc77UL, 0x922b422eUL,
+    0x93e92819UL, 0x96a63e9cUL, 0x976454abUL, 0x9522eaf2UL, 0x94e080c5UL,
+    0x9fbcc7f8UL, 0x9e7eadcfUL, 0x9c381396UL, 0x9dfa79a1UL, 0x98b56f24UL,
+    0x99770513UL, 0x9b31bb4aUL, 0x9af3d17dUL, 0x8d893530UL, 0x8c4b5f07UL,
+    0x8e0de15eUL, 0x8fcf8b69UL, 0x8a809decUL, 0x8b42f7dbUL, 0x89044982UL,
+    0x88c623b5UL, 0x839a6488UL, 0x82580ebfUL, 0x801eb0e6UL, 0x81dcdad1UL,
+    0x8493cc54UL, 0x8551a663UL, 0x8717183aUL, 0x86d5720dUL, 0xa9e2d0a0UL,
+    0xa820ba97UL, 0xaa6604ceUL, 0xaba46ef9UL, 0xaeeb787cUL, 0xaf29124bUL,
+    0xad6fac12UL, 0xacadc625UL, 0xa7f18118UL, 0xa633eb2fUL, 0xa4755576UL,
+    0xa5b73f41UL, 0xa0f829c4UL, 0xa13a43f3UL, 0xa37cfdaaUL, 0xa2be979dUL,
+    0xb5c473d0UL, 0xb40619e7UL, 0xb640a7beUL, 0xb782cd89UL, 0xb2cddb0cUL,
+    0xb30fb13bUL, 0xb1490f62UL, 0xb08b6555UL, 0xbbd72268UL, 0xba15485fUL,
+    0xb853f606UL, 0xb9919c31UL, 0xbcde8ab4UL, 0xbd1ce083UL, 0xbf5a5edaUL,
+    0xbe9834edUL
+  },
+  {
+    0x00000000UL, 0xb8bc6765UL, 0xaa09c88bUL, 0x12b5afeeUL, 0x8f629757UL,
+    0x37def032UL, 0x256b5fdcUL, 0x9dd738b9UL, 0xc5b428efUL, 0x7d084f8aUL,
+    0x6fbde064UL, 0xd7018701UL, 0x4ad6bfb8UL, 0xf26ad8ddUL, 0xe0df7733UL,
+    0x58631056UL, 0x5019579fUL, 0xe8a530faUL, 0xfa109f14UL, 0x42acf871UL,
+    0xdf7bc0c8UL, 0x67c7a7adUL, 0x75720843UL, 0xcdce6f26UL, 0x95ad7f70UL,
+    0x2d111815UL, 0x3fa4b7fbUL, 0x8718d09eUL, 0x1acfe827UL, 0xa2738f42UL,
+    0xb0c620acUL, 0x087a47c9UL, 0xa032af3eUL, 0x188ec85bUL, 0x0a3b67b5UL,
+    0xb28700d0UL, 0x2f503869UL, 0x97ec5f0cUL, 0x8559f0e2UL, 0x3de59787UL,
+    0x658687d1UL, 0xdd3ae0b4UL, 0xcf8f4f5aUL, 0x7733283fUL, 0xeae41086UL,
+    0x525877e3UL, 0x40edd80dUL, 0xf851bf68UL, 0xf02bf8a1UL, 0x48979fc4UL,
+    0x5a22302aUL, 0xe29e574fUL, 0x7f496ff6UL, 0xc7f50893UL, 0xd540a77dUL,
+    0x6dfcc018UL, 0x359fd04eUL, 0x8d23b72bUL, 0x9f9618c5UL, 0x272a7fa0UL,
+    0xbafd4719UL, 0x0241207cUL, 0x10f48f92UL, 0xa848e8f7UL, 0x9b14583dUL,
+    0x23a83f58UL, 0x311d90b6UL, 0x89a1f7d3UL, 0x1476cf6aUL, 0xaccaa80fUL,
+    0xbe7f07e1UL, 0x06c36084UL, 0x5ea070d2UL, 0xe61c17b7UL, 0xf4a9b859UL,
+    0x4c15df3cUL, 0xd1c2e785UL, 0x697e80e0UL, 0x7bcb2f0eUL, 0xc377486bUL,
+    0xcb0d0fa2UL, 0x73b168c7UL, 0x6104c729UL, 0xd9b8a04cUL, 0x446f98f5UL,
+    0xfcd3ff90UL, 0xee66507eUL, 0x56da371bUL, 0x0eb9274dUL, 0xb6054028UL,
+    0xa4b0efc6UL, 0x1c0c88a3UL, 0x81dbb01aUL, 0x3967d77fUL, 0x2bd27891UL,
+    0x936e1ff4UL, 0x3b26f703UL, 0x839a9066UL, 0x912f3f88UL, 0x299358edUL,
+    0xb4446054UL, 0x0cf80731UL, 0x1e4da8dfUL, 0xa6f1cfbaUL, 0xfe92dfecUL,
+    0x462eb889UL, 0x549b1767UL, 0xec277002UL, 0x71f048bbUL, 0xc94c2fdeUL,
+    0xdbf98030UL, 0x6345e755UL, 0x6b3fa09cUL, 0xd383c7f9UL, 0xc1366817UL,
+    0x798a0f72UL, 0xe45d37cbUL, 0x5ce150aeUL, 0x4e54ff40UL, 0xf6e89825UL,
+    0xae8b8873UL, 0x1637ef16UL, 0x048240f8UL, 0xbc3e279dUL, 0x21e91f24UL,
+    0x99557841UL, 0x8be0d7afUL, 0x335cb0caUL, 0xed59b63bUL, 0x55e5d15eUL,
+    0x47507eb0UL, 0xffec19d5UL, 0x623b216cUL, 0xda874609UL, 0xc832e9e7UL,
+    0x708e8e82UL, 0x28ed9ed4UL, 0x9051f9b1UL, 0x82e4565fUL, 0x3a58313aUL,
+    0xa78f0983UL, 0x1f336ee6UL, 0x0d86c108UL, 0xb53aa66dUL, 0xbd40e1a4UL,
+    0x05fc86c1UL, 0x1749292fUL, 0xaff54e4aUL, 0x322276f3UL, 0x8a9e1196UL,
+    0x982bbe78UL, 0x2097d91dUL, 0x78f4c94bUL, 0xc048ae2eUL, 0xd2fd01c0UL,
+    0x6a4166a5UL, 0xf7965e1cUL, 0x4f2a3979UL, 0x5d9f9697UL, 0xe523f1f2UL,
+    0x4d6b1905UL, 0xf5d77e60UL, 0xe762d18eUL, 0x5fdeb6ebUL, 0xc2098e52UL,
+    0x7ab5e937UL, 0x680046d9UL, 0xd0bc21bcUL, 0x88df31eaUL, 0x3063568fUL,
+    0x22d6f961UL, 0x9a6a9e04UL, 0x07bda6bdUL, 0xbf01c1d8UL, 0xadb46e36UL,
+    0x15080953UL, 0x1d724e9aUL, 0xa5ce29ffUL, 0xb77b8611UL, 0x0fc7e174UL,
+    0x9210d9cdUL, 0x2aacbea8UL, 0x38191146UL, 0x80a57623UL, 0xd8c66675UL,
+    0x607a0110UL, 0x72cfaefeUL, 0xca73c99bUL, 0x57a4f122UL, 0xef189647UL,
+    0xfdad39a9UL, 0x45115eccUL, 0x764dee06UL, 0xcef18963UL, 0xdc44268dUL,
+    0x64f841e8UL, 0xf92f7951UL, 0x41931e34UL, 0x5326b1daUL, 0xeb9ad6bfUL,
+    0xb3f9c6e9UL, 0x0b45a18cUL, 0x19f00e62UL, 0xa14c6907UL, 0x3c9b51beUL,
+    0x842736dbUL, 0x96929935UL, 0x2e2efe50UL, 0x2654b999UL, 0x9ee8defcUL,
+    0x8c5d7112UL, 0x34e11677UL, 0xa9362eceUL, 0x118a49abUL, 0x033fe645UL,
+    0xbb838120UL, 0xe3e09176UL, 0x5b5cf613UL, 0x49e959fdUL, 0xf1553e98UL,
+    0x6c820621UL, 0xd43e6144UL, 0xc68bceaaUL, 0x7e37a9cfUL, 0xd67f4138UL,
+    0x6ec3265dUL, 0x7c7689b3UL, 0xc4caeed6UL, 0x591dd66fUL, 0xe1a1b10aUL,
+    0xf3141ee4UL, 0x4ba87981UL, 0x13cb69d7UL, 0xab770eb2UL, 0xb9c2a15cUL,
+    0x017ec639UL, 0x9ca9fe80UL, 0x241599e5UL, 0x36a0360bUL, 0x8e1c516eUL,
+    0x866616a7UL, 0x3eda71c2UL, 0x2c6fde2cUL, 0x94d3b949UL, 0x090481f0UL,
+    0xb1b8e695UL, 0xa30d497bUL, 0x1bb12e1eUL, 0x43d23e48UL, 0xfb6e592dUL,
+    0xe9dbf6c3UL, 0x516791a6UL, 0xccb0a91fUL, 0x740cce7aUL, 0x66b96194UL,
+    0xde0506f1UL
+  },
+  {
+    0x00000000UL, 0x96300777UL, 0x2c610eeeUL, 0xba510999UL, 0x19c46d07UL,
+    0x8ff46a70UL, 0x35a563e9UL, 0xa395649eUL, 0x3288db0eUL, 0xa4b8dc79UL,
+    0x1ee9d5e0UL, 0x88d9d297UL, 0x2b4cb609UL, 0xbd7cb17eUL, 0x072db8e7UL,
+    0x911dbf90UL, 0x6410b71dUL, 0xf220b06aUL, 0x4871b9f3UL, 0xde41be84UL,
+    0x7dd4da1aUL, 0xebe4dd6dUL, 0x51b5d4f4UL, 0xc785d383UL, 0x56986c13UL,
+    0xc0a86b64UL, 0x7af962fdUL, 0xecc9658aUL, 0x4f5c0114UL, 0xd96c0663UL,
+    0x633d0ffaUL, 0xf50d088dUL, 0xc8206e3bUL, 0x5e10694cUL, 0xe44160d5UL,
+    0x727167a2UL, 0xd1e4033cUL, 0x47d4044bUL, 0xfd850dd2UL, 0x6bb50aa5UL,
+    0xfaa8b535UL, 0x6c98b242UL, 0xd6c9bbdbUL, 0x40f9bcacUL, 0xe36cd832UL,
+    0x755cdf45UL, 0xcf0dd6dcUL, 0x593dd1abUL, 0xac30d926UL, 0x3a00de51UL,
+    0x8051d7c8UL, 0x1661d0bfUL, 0xb5f4b421UL, 0x23c4b356UL, 0x9995bacfUL,
+    0x0fa5bdb8UL, 0x9eb80228UL, 0x0888055fUL, 0xb2d90cc6UL, 0x24e90bb1UL,
+    0x877c6f2fUL, 0x114c6858UL, 0xab1d61c1UL, 0x3d2d66b6UL, 0x9041dc76UL,
+    0x0671db01UL, 0xbc20d298UL, 0x2a10d5efUL, 0x8985b171UL, 0x1fb5b606UL,
+    0xa5e4bf9fUL, 0x33d4b8e8UL, 0xa2c90778UL, 0x34f9000fUL, 0x8ea80996UL,
+    0x18980ee1UL, 0xbb0d6a7fUL, 0x2d3d6d08UL, 0x976c6491UL, 0x015c63e6UL,
+    0xf4516b6bUL, 0x62616c1cUL, 0xd8306585UL, 0x4e0062f2UL, 0xed95066cUL,
+    0x7ba5011bUL, 0xc1f40882UL, 0x57c40ff5UL, 0xc6d9b065UL, 0x50e9b712UL,
+    0xeab8be8bUL, 0x7c88b9fcUL, 0xdf1ddd62UL, 0x492dda15UL, 0xf37cd38cUL,
+    0x654cd4fbUL, 0x5861b24dUL, 0xce51b53aUL, 0x7400bca3UL, 0xe230bbd4UL,
+    0x41a5df4aUL, 0xd795d83dUL, 0x6dc4d1a4UL, 0xfbf4d6d3UL, 0x6ae96943UL,
+    0xfcd96e34UL, 0x468867adUL, 0xd0b860daUL, 0x732d0444UL, 0xe51d0333UL,
+    0x5f4c0aaaUL, 0xc97c0dddUL, 0x3c710550UL, 0xaa410227UL, 0x10100bbeUL,
+    0x86200cc9UL, 0x25b56857UL, 0xb3856f20UL, 0x09d466b9UL, 0x9fe461ceUL,
+    0x0ef9de5eUL, 0x98c9d929UL, 0x2298d0b0UL, 0xb4a8d7c7UL, 0x173db359UL,
+    0x810db42eUL, 0x3b5cbdb7UL, 0xad6cbac0UL, 0x2083b8edUL, 0xb6b3bf9aUL,
+    0x0ce2b603UL, 0x9ad2b174UL, 0x3947d5eaUL, 0xaf77d29dUL, 0x1526db04UL,
+    0x8316dc73UL, 0x120b63e3UL, 0x843b6494UL, 0x3e6a6d0dUL, 0xa85a6a7aUL,
+    0x0bcf0ee4UL, 0x9dff0993UL, 0x27ae000aUL, 0xb19e077dUL, 0x44930ff0UL,
+    0xd2a30887UL, 0x68f2011eUL, 0xfec20669UL, 0x5d5762f7UL, 0xcb676580UL,
+    0x71366c19UL, 0xe7066b6eUL, 0x761bd4feUL, 0xe02bd389UL, 0x5a7ada10UL,
+    0xcc4add67UL, 0x6fdfb9f9UL, 0xf9efbe8eUL, 0x43beb717UL, 0xd58eb060UL,
+    0xe8a3d6d6UL, 0x7e93d1a1UL, 0xc4c2d838UL, 0x52f2df4fUL, 0xf167bbd1UL,
+    0x6757bca6UL, 0xdd06b53fUL, 0x4b36b248UL, 0xda2b0dd8UL, 0x4c1b0aafUL,
+    0xf64a0336UL, 0x607a0441UL, 0xc3ef60dfUL, 0x55df67a8UL, 0xef8e6e31UL,
+    0x79be6946UL, 0x8cb361cbUL, 0x1a8366bcUL, 0xa0d26f25UL, 0x36e26852UL,
+    0x95770cccUL, 0x03470bbbUL, 0xb9160222UL, 0x2f260555UL, 0xbe3bbac5UL,
+    0x280bbdb2UL, 0x925ab42bUL, 0x046ab35cUL, 0xa7ffd7c2UL, 0x31cfd0b5UL,
+    0x8b9ed92cUL, 0x1daede5bUL, 0xb0c2649bUL, 0x26f263ecUL, 0x9ca36a75UL,
+    0x0a936d02UL, 0xa906099cUL, 0x3f360eebUL, 0x85670772UL, 0x13570005UL,
+    0x824abf95UL, 0x147ab8e2UL, 0xae2bb17bUL, 0x381bb60cUL, 0x9b8ed292UL,
+    0x0dbed5e5UL, 0xb7efdc7cUL, 0x21dfdb0bUL, 0xd4d2d386UL, 0x42e2d4f1UL,
+    0xf8b3dd68UL, 0x6e83da1fUL, 0xcd16be81UL, 0x5b26b9f6UL, 0xe177b06fUL,
+    0x7747b718UL, 0xe65a0888UL, 0x706a0fffUL, 0xca3b0666UL, 0x5c0b0111UL,
+    0xff9e658fUL, 0x69ae62f8UL, 0xd3ff6b61UL, 0x45cf6c16UL, 0x78e20aa0UL,
+    0xeed20dd7UL, 0x5483044eUL, 0xc2b30339UL, 0x612667a7UL, 0xf71660d0UL,
+    0x4d476949UL, 0xdb776e3eUL, 0x4a6ad1aeUL, 0xdc5ad6d9UL, 0x660bdf40UL,
+    0xf03bd837UL, 0x53aebca9UL, 0xc59ebbdeUL, 0x7fcfb247UL, 0xe9ffb530UL,
+    0x1cf2bdbdUL, 0x8ac2bacaUL, 0x3093b353UL, 0xa6a3b424UL, 0x0536d0baUL,
+    0x9306d7cdUL, 0x2957de54UL, 0xbf67d923UL, 0x2e7a66b3UL, 0xb84a61c4UL,
+    0x021b685dUL, 0x942b6f2aUL, 0x37be0bb4UL, 0xa18e0cc3UL, 0x1bdf055aUL,
+    0x8def022dUL
+  },
+  {
+    0x00000000UL, 0x41311b19UL, 0x82623632UL, 0xc3532d2bUL, 0x04c56c64UL,
+    0x45f4777dUL, 0x86a75a56UL, 0xc796414fUL, 0x088ad9c8UL, 0x49bbc2d1UL,
+    0x8ae8effaUL, 0xcbd9f4e3UL, 0x0c4fb5acUL, 0x4d7eaeb5UL, 0x8e2d839eUL,
+    0xcf1c9887UL, 0x5112c24aUL, 0x1023d953UL, 0xd370f478UL, 0x9241ef61UL,
+    0x55d7ae2eUL, 0x14e6b537UL, 0xd7b5981cUL, 0x96848305UL, 0x59981b82UL,
+    0x18a9009bUL, 0xdbfa2db0UL, 0x9acb36a9UL, 0x5d5d77e6UL, 0x1c6c6cffUL,
+    0xdf3f41d4UL, 0x9e0e5acdUL, 0xa2248495UL, 0xe3159f8cUL, 0x2046b2a7UL,
+    0x6177a9beUL, 0xa6e1e8f1UL, 0xe7d0f3e8UL, 0x2483dec3UL, 0x65b2c5daUL,
+    0xaaae5d5dUL, 0xeb9f4644UL, 0x28cc6b6fUL, 0x69fd7076UL, 0xae6b3139UL,
+    0xef5a2a20UL, 0x2c09070bUL, 0x6d381c12UL, 0xf33646dfUL, 0xb2075dc6UL,
+    0x715470edUL, 0x30656bf4UL, 0xf7f32abbUL, 0xb6c231a2UL, 0x75911c89UL,
+    0x34a00790UL, 0xfbbc9f17UL, 0xba8d840eUL, 0x79dea925UL, 0x38efb23cUL,
+    0xff79f373UL, 0xbe48e86aUL, 0x7d1bc541UL, 0x3c2ade58UL, 0x054f79f0UL,
+    0x447e62e9UL, 0x872d4fc2UL, 0xc61c54dbUL, 0x018a1594UL, 0x40bb0e8dUL,
+    0x83e823a6UL, 0xc2d938bfUL, 0x0dc5a038UL, 0x4cf4bb21UL, 0x8fa7960aUL,
+    0xce968d13UL, 0x0900cc5cUL, 0x4831d745UL, 0x8b62fa6eUL, 0xca53e177UL,
+    0x545dbbbaUL, 0x156ca0a3UL, 0xd63f8d88UL, 0x970e9691UL, 0x5098d7deUL,
+    0x11a9ccc7UL, 0xd2fae1ecUL, 0x93cbfaf5UL, 0x5cd76272UL, 0x1de6796bUL,
+    0xdeb55440UL, 0x9f844f59UL, 0x58120e16UL, 0x1923150fUL, 0xda703824UL,
+    0x9b41233dUL, 0xa76bfd65UL, 0xe65ae67cUL, 0x2509cb57UL, 0x6438d04eUL,
+    0xa3ae9101UL, 0xe29f8a18UL, 0x21cca733UL, 0x60fdbc2aUL, 0xafe124adUL,
+    0xeed03fb4UL, 0x2d83129fUL, 0x6cb20986UL, 0xab2448c9UL, 0xea1553d0UL,
+    0x29467efbUL, 0x687765e2UL, 0xf6793f2fUL, 0xb7482436UL, 0x741b091dUL,
+    0x352a1204UL, 0xf2bc534bUL, 0xb38d4852UL, 0x70de6579UL, 0x31ef7e60UL,
+    0xfef3e6e7UL, 0xbfc2fdfeUL, 0x7c91d0d5UL, 0x3da0cbccUL, 0xfa368a83UL,
+    0xbb07919aUL, 0x7854bcb1UL, 0x3965a7a8UL, 0x4b98833bUL, 0x0aa99822UL,
+    0xc9fab509UL, 0x88cbae10UL, 0x4f5def5fUL, 0x0e6cf446UL, 0xcd3fd96dUL,
+    0x8c0ec274UL, 0x43125af3UL, 0x022341eaUL, 0xc1706cc1UL, 0x804177d8UL,
+    0x47d73697UL, 0x06e62d8eUL, 0xc5b500a5UL, 0x84841bbcUL, 0x1a8a4171UL,
+    0x5bbb5a68UL, 0x98e87743UL, 0xd9d96c5aUL, 0x1e4f2d15UL, 0x5f7e360cUL,
+    0x9c2d1b27UL, 0xdd1c003eUL, 0x120098b9UL, 0x533183a0UL, 0x9062ae8bUL,
+    0xd153b592UL, 0x16c5f4ddUL, 0x57f4efc4UL, 0x94a7c2efUL, 0xd596d9f6UL,
+    0xe9bc07aeUL, 0xa88d1cb7UL, 0x6bde319cUL, 0x2aef2a85UL, 0xed796bcaUL,
+    0xac4870d3UL, 0x6f1b5df8UL, 0x2e2a46e1UL, 0xe136de66UL, 0xa007c57fUL,
+    0x6354e854UL, 0x2265f34dUL, 0xe5f3b202UL, 0xa4c2a91bUL, 0x67918430UL,
+    0x26a09f29UL, 0xb8aec5e4UL, 0xf99fdefdUL, 0x3accf3d6UL, 0x7bfde8cfUL,
+    0xbc6ba980UL, 0xfd5ab299UL, 0x3e099fb2UL, 0x7f3884abUL, 0xb0241c2cUL,
+    0xf1150735UL, 0x32462a1eUL, 0x73773107UL, 0xb4e17048UL, 0xf5d06b51UL,
+    0x3683467aUL, 0x77b25d63UL, 0x4ed7facbUL, 0x0fe6e1d2UL, 0xccb5ccf9UL,
+    0x8d84d7e0UL, 0x4a1296afUL, 0x0b238db6UL, 0xc870a09dUL, 0x8941bb84UL,
+    0x465d2303UL, 0x076c381aUL, 0xc43f1531UL, 0x850e0e28UL, 0x42984f67UL,
+    0x03a9547eUL, 0xc0fa7955UL, 0x81cb624cUL, 0x1fc53881UL, 0x5ef42398UL,
+    0x9da70eb3UL, 0xdc9615aaUL, 0x1b0054e5UL, 0x5a314ffcUL, 0x996262d7UL,
+    0xd85379ceUL, 0x174fe149UL, 0x567efa50UL, 0x952dd77bUL, 0xd41ccc62UL,
+    0x138a8d2dUL, 0x52bb9634UL, 0x91e8bb1fUL, 0xd0d9a006UL, 0xecf37e5eUL,
+    0xadc26547UL, 0x6e91486cUL, 0x2fa05375UL, 0xe836123aUL, 0xa9070923UL,
+    0x6a542408UL, 0x2b653f11UL, 0xe479a796UL, 0xa548bc8fUL, 0x661b91a4UL,
+    0x272a8abdUL, 0xe0bccbf2UL, 0xa18dd0ebUL, 0x62defdc0UL, 0x23efe6d9UL,
+    0xbde1bc14UL, 0xfcd0a70dUL, 0x3f838a26UL, 0x7eb2913fUL, 0xb924d070UL,
+    0xf815cb69UL, 0x3b46e642UL, 0x7a77fd5bUL, 0xb56b65dcUL, 0xf45a7ec5UL,
+    0x370953eeUL, 0x763848f7UL, 0xb1ae09b8UL, 0xf09f12a1UL, 0x33cc3f8aUL,
+    0x72fd2493UL
+  },
+  {
+    0x00000000UL, 0x376ac201UL, 0x6ed48403UL, 0x59be4602UL, 0xdca80907UL,
+    0xebc2cb06UL, 0xb27c8d04UL, 0x85164f05UL, 0xb851130eUL, 0x8f3bd10fUL,
+    0xd685970dUL, 0xe1ef550cUL, 0x64f91a09UL, 0x5393d808UL, 0x0a2d9e0aUL,
+    0x3d475c0bUL, 0x70a3261cUL, 0x47c9e41dUL, 0x1e77a21fUL, 0x291d601eUL,
+    0xac0b2f1bUL, 0x9b61ed1aUL, 0xc2dfab18UL, 0xf5b56919UL, 0xc8f23512UL,
+    0xff98f713UL, 0xa626b111UL, 0x914c7310UL, 0x145a3c15UL, 0x2330fe14UL,
+    0x7a8eb816UL, 0x4de47a17UL, 0xe0464d38UL, 0xd72c8f39UL, 0x8e92c93bUL,
+    0xb9f80b3aUL, 0x3cee443fUL, 0x0b84863eUL, 0x523ac03cUL, 0x6550023dUL,
+    0x58175e36UL, 0x6f7d9c37UL, 0x36c3da35UL, 0x01a91834UL, 0x84bf5731UL,
+    0xb3d59530UL, 0xea6bd332UL, 0xdd011133UL, 0x90e56b24UL, 0xa78fa925UL,
+    0xfe31ef27UL, 0xc95b2d26UL, 0x4c4d6223UL, 0x7b27a022UL, 0x2299e620UL,
+    0x15f32421UL, 0x28b4782aUL, 0x1fdeba2bUL, 0x4660fc29UL, 0x710a3e28UL,
+    0xf41c712dUL, 0xc376b32cUL, 0x9ac8f52eUL, 0xada2372fUL, 0xc08d9a70UL,
+    0xf7e75871UL, 0xae591e73UL, 0x9933dc72UL, 0x1c259377UL, 0x2b4f5176UL,
+    0x72f11774UL, 0x459bd575UL, 0x78dc897eUL, 0x4fb64b7fUL, 0x16080d7dUL,
+    0x2162cf7cUL, 0xa4748079UL, 0x931e4278UL, 0xcaa0047aUL, 0xfdcac67bUL,
+    0xb02ebc6cUL, 0x87447e6dUL, 0xdefa386fUL, 0xe990fa6eUL, 0x6c86b56bUL,
+    0x5bec776aUL, 0x02523168UL, 0x3538f369UL, 0x087faf62UL, 0x3f156d63UL,
+    0x66ab2b61UL, 0x51c1e960UL, 0xd4d7a665UL, 0xe3bd6464UL, 0xba032266UL,
+    0x8d69e067UL, 0x20cbd748UL, 0x17a11549UL, 0x4e1f534bUL, 0x7975914aUL,
+    0xfc63de4fUL, 0xcb091c4eUL, 0x92b75a4cUL, 0xa5dd984dUL, 0x989ac446UL,
+    0xaff00647UL, 0xf64e4045UL, 0xc1248244UL, 0x4432cd41UL, 0x73580f40UL,
+    0x2ae64942UL, 0x1d8c8b43UL, 0x5068f154UL, 0x67023355UL, 0x3ebc7557UL,
+    0x09d6b756UL, 0x8cc0f853UL, 0xbbaa3a52UL, 0xe2147c50UL, 0xd57ebe51UL,
+    0xe839e25aUL, 0xdf53205bUL, 0x86ed6659UL, 0xb187a458UL, 0x3491eb5dUL,
+    0x03fb295cUL, 0x5a456f5eUL, 0x6d2fad5fUL, 0x801b35e1UL, 0xb771f7e0UL,
+    0xeecfb1e2UL, 0xd9a573e3UL, 0x5cb33ce6UL, 0x6bd9fee7UL, 0x3267b8e5UL,
+    0x050d7ae4UL, 0x384a26efUL, 0x0f20e4eeUL, 0x569ea2ecUL, 0x61f460edUL,
+    0xe4e22fe8UL, 0xd388ede9UL, 0x8a36abebUL, 0xbd5c69eaUL, 0xf0b813fdUL,
+    0xc7d2d1fcUL, 0x9e6c97feUL, 0xa90655ffUL, 0x2c101afaUL, 0x1b7ad8fbUL,
+    0x42c49ef9UL, 0x75ae5cf8UL, 0x48e900f3UL, 0x7f83c2f2UL, 0x263d84f0UL,
+    0x115746f1UL, 0x944109f4UL, 0xa32bcbf5UL, 0xfa958df7UL, 0xcdff4ff6UL,
+    0x605d78d9UL, 0x5737bad8UL, 0x0e89fcdaUL, 0x39e33edbUL, 0xbcf571deUL,
+    0x8b9fb3dfUL, 0xd221f5ddUL, 0xe54b37dcUL, 0xd80c6bd7UL, 0xef66a9d6UL,
+    0xb6d8efd4UL, 0x81b22dd5UL, 0x04a462d0UL, 0x33cea0d1UL, 0x6a70e6d3UL,
+    0x5d1a24d2UL, 0x10fe5ec5UL, 0x27949cc4UL, 0x7e2adac6UL, 0x494018c7UL,
+    0xcc5657c2UL, 0xfb3c95c3UL, 0xa282d3c1UL, 0x95e811c0UL, 0xa8af4dcbUL,
+    0x9fc58fcaUL, 0xc67bc9c8UL, 0xf1110bc9UL, 0x740744ccUL, 0x436d86cdUL,
+    0x1ad3c0cfUL, 0x2db902ceUL, 0x4096af91UL, 0x77fc6d90UL, 0x2e422b92UL,
+    0x1928e993UL, 0x9c3ea696UL, 0xab546497UL, 0xf2ea2295UL, 0xc580e094UL,
+    0xf8c7bc9fUL, 0xcfad7e9eUL, 0x9613389cUL, 0xa179fa9dUL, 0x246fb598UL,
+    0x13057799UL, 0x4abb319bUL, 0x7dd1f39aUL, 0x3035898dUL, 0x075f4b8cUL,
+    0x5ee10d8eUL, 0x698bcf8fUL, 0xec9d808aUL, 0xdbf7428bUL, 0x82490489UL,
+    0xb523c688UL, 0x88649a83UL, 0xbf0e5882UL, 0xe6b01e80UL, 0xd1dadc81UL,
+    0x54cc9384UL, 0x63a65185UL, 0x3a181787UL, 0x0d72d586UL, 0xa0d0e2a9UL,
+    0x97ba20a8UL, 0xce0466aaUL, 0xf96ea4abUL, 0x7c78ebaeUL, 0x4b1229afUL,
+    0x12ac6fadUL, 0x25c6adacUL, 0x1881f1a7UL, 0x2feb33a6UL, 0x765575a4UL,
+    0x413fb7a5UL, 0xc429f8a0UL, 0xf3433aa1UL, 0xaafd7ca3UL, 0x9d97bea2UL,
+    0xd073c4b5UL, 0xe71906b4UL, 0xbea740b6UL, 0x89cd82b7UL, 0x0cdbcdb2UL,
+    0x3bb10fb3UL, 0x620f49b1UL, 0x55658bb0UL, 0x6822d7bbUL, 0x5f4815baUL,
+    0x06f653b8UL, 0x319c91b9UL, 0xb48adebcUL, 0x83e01cbdUL, 0xda5e5abfUL,
+    0xed3498beUL
+  },
+  {
+    0x00000000UL, 0x6567bcb8UL, 0x8bc809aaUL, 0xeeafb512UL, 0x5797628fUL,
+    0x32f0de37UL, 0xdc5f6b25UL, 0xb938d79dUL, 0xef28b4c5UL, 0x8a4f087dUL,
+    0x64e0bd6fUL, 0x018701d7UL, 0xb8bfd64aUL, 0xddd86af2UL, 0x3377dfe0UL,
+    0x56106358UL, 0x9f571950UL, 0xfa30a5e8UL, 0x149f10faUL, 0x71f8ac42UL,
+    0xc8c07bdfUL, 0xada7c767UL, 0x43087275UL, 0x266fcecdUL, 0x707fad95UL,
+    0x1518112dUL, 0xfbb7a43fUL, 0x9ed01887UL, 0x27e8cf1aUL, 0x428f73a2UL,
+    0xac20c6b0UL, 0xc9477a08UL, 0x3eaf32a0UL, 0x5bc88e18UL, 0xb5673b0aUL,
+    0xd00087b2UL, 0x6938502fUL, 0x0c5fec97UL, 0xe2f05985UL, 0x8797e53dUL,
+    0xd1878665UL, 0xb4e03addUL, 0x5a4f8fcfUL, 0x3f283377UL, 0x8610e4eaUL,
+    0xe3775852UL, 0x0dd8ed40UL, 0x68bf51f8UL, 0xa1f82bf0UL, 0xc49f9748UL,
+    0x2a30225aUL, 0x4f579ee2UL, 0xf66f497fUL, 0x9308f5c7UL, 0x7da740d5UL,
+    0x18c0fc6dUL, 0x4ed09f35UL, 0x2bb7238dUL, 0xc518969fUL, 0xa07f2a27UL,
+    0x1947fdbaUL, 0x7c204102UL, 0x928ff410UL, 0xf7e848a8UL, 0x3d58149bUL,
+    0x583fa823UL, 0xb6901d31UL, 0xd3f7a189UL, 0x6acf7614UL, 0x0fa8caacUL,
+    0xe1077fbeUL, 0x8460c306UL, 0xd270a05eUL, 0xb7171ce6UL, 0x59b8a9f4UL,
+    0x3cdf154cUL, 0x85e7c2d1UL, 0xe0807e69UL, 0x0e2fcb7bUL, 0x6b4877c3UL,
+    0xa20f0dcbUL, 0xc768b173UL, 0x29c70461UL, 0x4ca0b8d9UL, 0xf5986f44UL,
+    0x90ffd3fcUL, 0x7e5066eeUL, 0x1b37da56UL, 0x4d27b90eUL, 0x284005b6UL,
+    0xc6efb0a4UL, 0xa3880c1cUL, 0x1ab0db81UL, 0x7fd76739UL, 0x9178d22bUL,
+    0xf41f6e93UL, 0x03f7263bUL, 0x66909a83UL, 0x883f2f91UL, 0xed589329UL,
+    0x546044b4UL, 0x3107f80cUL, 0xdfa84d1eUL, 0xbacff1a6UL, 0xecdf92feUL,
+    0x89b82e46UL, 0x67179b54UL, 0x027027ecUL, 0xbb48f071UL, 0xde2f4cc9UL,
+    0x3080f9dbUL, 0x55e74563UL, 0x9ca03f6bUL, 0xf9c783d3UL, 0x176836c1UL,
+    0x720f8a79UL, 0xcb375de4UL, 0xae50e15cUL, 0x40ff544eUL, 0x2598e8f6UL,
+    0x73888baeUL, 0x16ef3716UL, 0xf8408204UL, 0x9d273ebcUL, 0x241fe921UL,
+    0x41785599UL, 0xafd7e08bUL, 0xcab05c33UL, 0x3bb659edUL, 0x5ed1e555UL,
+    0xb07e5047UL, 0xd519ecffUL, 0x6c213b62UL, 0x094687daUL, 0xe7e932c8UL,
+    0x828e8e70UL, 0xd49eed28UL, 0xb1f95190UL, 0x5f56e482UL, 0x3a31583aUL,
+    0x83098fa7UL, 0xe66e331fUL, 0x08c1860dUL, 0x6da63ab5UL, 0xa4e140bdUL,
+    0xc186fc05UL, 0x2f294917UL, 0x4a4ef5afUL, 0xf3762232UL, 0x96119e8aUL,
+    0x78be2b98UL, 0x1dd99720UL, 0x4bc9f478UL, 0x2eae48c0UL, 0xc001fdd2UL,
+    0xa566416aUL, 0x1c5e96f7UL, 0x79392a4fUL, 0x97969f5dUL, 0xf2f123e5UL,
+    0x05196b4dUL, 0x607ed7f5UL, 0x8ed162e7UL, 0xebb6de5fUL, 0x528e09c2UL,
+    0x37e9b57aUL, 0xd9460068UL, 0xbc21bcd0UL, 0xea31df88UL, 0x8f566330UL,
+    0x61f9d622UL, 0x049e6a9aUL, 0xbda6bd07UL, 0xd8c101bfUL, 0x366eb4adUL,
+    0x53090815UL, 0x9a4e721dUL, 0xff29cea5UL, 0x11867bb7UL, 0x74e1c70fUL,
+    0xcdd91092UL, 0xa8beac2aUL, 0x46111938UL, 0x2376a580UL, 0x7566c6d8UL,
+    0x10017a60UL, 0xfeaecf72UL, 0x9bc973caUL, 0x22f1a457UL, 0x479618efUL,
+    0xa939adfdUL, 0xcc5e1145UL, 0x06ee4d76UL, 0x6389f1ceUL, 0x8d2644dcUL,
+    0xe841f864UL, 0x51792ff9UL, 0x341e9341UL, 0xdab12653UL, 0xbfd69aebUL,
+    0xe9c6f9b3UL, 0x8ca1450bUL, 0x620ef019UL, 0x07694ca1UL, 0xbe519b3cUL,
+    0xdb362784UL, 0x35999296UL, 0x50fe2e2eUL, 0x99b95426UL, 0xfcdee89eUL,
+    0x12715d8cUL, 0x7716e134UL, 0xce2e36a9UL, 0xab498a11UL, 0x45e63f03UL,
+    0x208183bbUL, 0x7691e0e3UL, 0x13f65c5bUL, 0xfd59e949UL, 0x983e55f1UL,
+    0x2106826cUL, 0x44613ed4UL, 0xaace8bc6UL, 0xcfa9377eUL, 0x38417fd6UL,
+    0x5d26c36eUL, 0xb389767cUL, 0xd6eecac4UL, 0x6fd61d59UL, 0x0ab1a1e1UL,
+    0xe41e14f3UL, 0x8179a84bUL, 0xd769cb13UL, 0xb20e77abUL, 0x5ca1c2b9UL,
+    0x39c67e01UL, 0x80fea99cUL, 0xe5991524UL, 0x0b36a036UL, 0x6e511c8eUL,
+    0xa7166686UL, 0xc271da3eUL, 0x2cde6f2cUL, 0x49b9d394UL, 0xf0810409UL,
+    0x95e6b8b1UL, 0x7b490da3UL, 0x1e2eb11bUL, 0x483ed243UL, 0x2d596efbUL,
+    0xc3f6dbe9UL, 0xa6916751UL, 0x1fa9b0ccUL, 0x7ace0c74UL, 0x9461b966UL,
+    0xf10605deUL
+#endif
+  }
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/deflate.c glmark2-2021.02/src/zlib/deflate.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/deflate.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/deflate.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,2163 @@
+/* deflate.c -- compress data using the deflation algorithm
+ * Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process depends on being able to identify portions
+ *      of the input text which are identical to earlier input (within a
+ *      sliding window trailing behind the input currently being processed).
+ *
+ *      The most straightforward technique turns out to be the fastest for
+ *      most input files: try all possible matches and select the longest.
+ *      The key feature of this algorithm is that insertions into the string
+ *      dictionary are very simple and thus fast, and deletions are avoided
+ *      completely. Insertions are performed at each input character, whereas
+ *      string matches are performed only when the previous match ends. So it
+ *      is preferable to spend more time in matches to allow very fast string
+ *      insertions and avoid deletions. The matching algorithm for small
+ *      strings is inspired from that of Rabin & Karp. A brute force approach
+ *      is used to find longer strings when a small match has been found.
+ *      A similar algorithm is used in comic (by Jan-Mark Wams) and freeze
+ *      (by Leonid Broukhis).
+ *         A previous version of this file used a more sophisticated algorithm
+ *      (by Fiala and Greene) which is guaranteed to run in linear amortized
+ *      time, but has a larger average cost, uses more memory and is patented.
+ *      However the F&G algorithm may be faster for some highly redundant
+ *      files if the parameter max_chain_length (described below) is too large.
+ *
+ *  ACKNOWLEDGEMENTS
+ *
+ *      The idea of lazy evaluation of matches is due to Jan-Mark Wams, and
+ *      I found it in 'freeze' written by Leonid Broukhis.
+ *      Thanks to many people for bug reports and testing.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"DEFLATE Compressed Data Format Specification".
+ *      Available in http://tools.ietf.org/html/rfc1951
+ *
+ *      A description of the Rabin and Karp algorithm is given in the book
+ *         "Algorithms" by R. Sedgewick, Addison-Wesley, p252.
+ *
+ *      Fiala,E.R., and Greene,D.H.
+ *         Data Compression with Finite Windows, Comm.ACM, 32,4 (1989) 490-595
+ *
+ */
+
+/* @(#) $Id$ */
+
+#include "deflate.h"
+
+const char deflate_copyright[] =
+   " deflate 1.2.11 Copyright 1995-2017 Jean-loup Gailly and Mark Adler ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+
+/* ===========================================================================
+ *  Function prototypes.
+ */
+typedef enum {
+    need_more,      /* block not completed, need more input or more output */
+    block_done,     /* block flush performed */
+    finish_started, /* finish started, need only more output at next deflate */
+    finish_done     /* finish done, accept no more input or output */
+} block_state;
+
+typedef block_state (*compress_func) OF((deflate_state *s, int flush));
+/* Compression function. Returns the block state after the call. */
+
+local int deflateStateCheck      OF((z_streamp strm));
+local void slide_hash     OF((deflate_state *s));
+local void fill_window    OF((deflate_state *s));
+local block_state deflate_stored OF((deflate_state *s, int flush));
+local block_state deflate_fast   OF((deflate_state *s, int flush));
+#ifndef FASTEST
+local block_state deflate_slow   OF((deflate_state *s, int flush));
+#endif
+local block_state deflate_rle    OF((deflate_state *s, int flush));
+local block_state deflate_huff   OF((deflate_state *s, int flush));
+local void lm_init        OF((deflate_state *s));
+local void putShortMSB    OF((deflate_state *s, uInt b));
+local void flush_pending  OF((z_streamp strm));
+local unsigned read_buf   OF((z_streamp strm, Bytef *buf, unsigned size));
+#ifdef ASMV
+#  pragma message("Assembler code may have bugs -- use at your own risk")
+      void match_init OF((void)); /* asm code initialization */
+      uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#else
+local uInt longest_match  OF((deflate_state *s, IPos cur_match));
+#endif
+
+#ifdef ZLIB_DEBUG
+local  void check_match OF((deflate_state *s, IPos start, IPos match,
+                            int length));
+#endif
+
+/* ===========================================================================
+ * Local data
+ */
+
+#define NIL 0
+/* Tail of hash chains */
+
+#ifndef TOO_FAR
+#  define TOO_FAR 4096
+#endif
+/* Matches of length 3 are discarded if their distance exceeds TOO_FAR */
+
+/* Values for max_lazy_match, good_match and max_chain_length, depending on
+ * the desired pack level (0..9). The values given below have been tuned to
+ * exclude worst case performance for pathological files. Better values may be
+ * found for specific files.
+ */
+typedef struct config_s {
+   ush good_length; /* reduce lazy search above this match length */
+   ush max_lazy;    /* do not perform lazy search above this match length */
+   ush nice_length; /* quit search above this match length */
+   ush max_chain;
+   compress_func func;
+} config;
+
+#ifdef FASTEST
+local const config configuration_table[2] = {
+/*      good lazy nice chain */
+/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
+/* 1 */ {4,    4,  8,    4, deflate_fast}}; /* max speed, no lazy matches */
+#else
+local const config configuration_table[10] = {
+/*      good lazy nice chain */
+/* 0 */ {0,    0,  0,    0, deflate_stored},  /* store only */
+/* 1 */ {4,    4,  8,    4, deflate_fast}, /* max speed, no lazy matches */
+/* 2 */ {4,    5, 16,    8, deflate_fast},
+/* 3 */ {4,    6, 32,   32, deflate_fast},
+
+/* 4 */ {4,    4, 16,   16, deflate_slow},  /* lazy matches */
+/* 5 */ {8,   16, 32,   32, deflate_slow},
+/* 6 */ {8,   16, 128, 128, deflate_slow},
+/* 7 */ {8,   32, 128, 256, deflate_slow},
+/* 8 */ {32, 128, 258, 1024, deflate_slow},
+/* 9 */ {32, 258, 258, 4096, deflate_slow}}; /* max compression */
+#endif
+
+/* Note: the deflate() code requires max_lazy >= MIN_MATCH and max_chain >= 4
+ * For deflate_fast() (levels <= 3) good is ignored and lazy has a different
+ * meaning.
+ */
+
+/* rank Z_BLOCK between Z_NO_FLUSH and Z_PARTIAL_FLUSH */
+#define RANK(f) (((f) * 2) - ((f) > 4 ? 9 : 0))
+
+/* ===========================================================================
+ * Update a hash value with the given input byte
+ * IN  assertion: all calls to UPDATE_HASH are made with consecutive input
+ *    characters, so that a running hash key can be computed from the previous
+ *    key instead of complete recalculation each time.
+ */
+#define UPDATE_HASH(s,h,c) (h = (((h)<<s->hash_shift) ^ (c)) & s->hash_mask)
+
+
+/* ===========================================================================
+ * Insert string str in the dictionary and set match_head to the previous head
+ * of the hash chain (the most recent string with same hash key). Return
+ * the previous length of the hash chain.
+ * If this file is compiled with -DFASTEST, the compression level is forced
+ * to 1, and no hash chains are maintained.
+ * IN  assertion: all calls to INSERT_STRING are made with consecutive input
+ *    characters and the first MIN_MATCH bytes of str are valid (except for
+ *    the last MIN_MATCH-1 bytes of the input file).
+ */
+#ifdef FASTEST
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    match_head = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#else
+#define INSERT_STRING(s, str, match_head) \
+   (UPDATE_HASH(s, s->ins_h, s->window[(str) + (MIN_MATCH-1)]), \
+    match_head = s->prev[(str) & s->w_mask] = s->head[s->ins_h], \
+    s->head[s->ins_h] = (Pos)(str))
+#endif
+
+/* ===========================================================================
+ * Initialize the hash table (avoiding 64K overflow for 16 bit systems).
+ * prev[] will be initialized on the fly.
+ */
+#define CLEAR_HASH(s) \
+    s->head[s->hash_size-1] = NIL; \
+    zmemzero((Bytef *)s->head, (unsigned)(s->hash_size-1)*sizeof(*s->head));
+
+/* ===========================================================================
+ * Slide the hash table when sliding the window down (could be avoided with 32
+ * bit values at the expense of memory usage). We slide even when level == 0 to
+ * keep the hash table consistent if we switch back to level > 0 later.
+ */
+local void slide_hash(s)
+    deflate_state *s;
+{
+    unsigned n, m;
+    Posf *p;
+    uInt wsize = s->w_size;
+
+    n = s->hash_size;
+    p = &s->head[n];
+    do {
+        m = *--p;
+        *p = (Pos)(m >= wsize ? m - wsize : NIL);
+    } while (--n);
+    n = wsize;
+#ifndef FASTEST
+    p = &s->prev[n];
+    do {
+        m = *--p;
+        *p = (Pos)(m >= wsize ? m - wsize : NIL);
+        /* If n is not on any hash chain, prev[n] is garbage but
+         * its value will never be used.
+         */
+    } while (--n);
+#endif
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateInit_(strm, level, version, stream_size)
+    z_streamp strm;
+    int level;
+    const char *version;
+    int stream_size;
+{
+    return deflateInit2_(strm, level, Z_DEFLATED, MAX_WBITS, DEF_MEM_LEVEL,
+                         Z_DEFAULT_STRATEGY, version, stream_size);
+    /* To do: ignore strm->next_in if we use it as window */
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateInit2_(strm, level, method, windowBits, memLevel, strategy,
+                  version, stream_size)
+    z_streamp strm;
+    int  level;
+    int  method;
+    int  windowBits;
+    int  memLevel;
+    int  strategy;
+    const char *version;
+    int stream_size;
+{
+    deflate_state *s;
+    int wrap = 1;
+    static const char my_version[] = ZLIB_VERSION;
+
+    ushf *overlay;
+    /* We overlay pending_buf and d_buf+l_buf. This works since the average
+     * output size for (length,distance) codes is <= 24 bits.
+     */
+
+    if (version == Z_NULL || version[0] != my_version[0] ||
+        stream_size != sizeof(z_stream)) {
+        return Z_VERSION_ERROR;
+    }
+    if (strm == Z_NULL) return Z_STREAM_ERROR;
+
+    strm->msg = Z_NULL;
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zfree = zcfree;
+#endif
+
+#ifdef FASTEST
+    if (level != 0) level = 1;
+#else
+    if (level == Z_DEFAULT_COMPRESSION) level = 6;
+#endif
+
+    if (windowBits < 0) { /* suppress zlib wrapper */
+        wrap = 0;
+        windowBits = -windowBits;
+    }
+#ifdef GZIP
+    else if (windowBits > 15) {
+        wrap = 2;       /* write gzip wrapper instead */
+        windowBits -= 16;
+    }
+#endif
+    if (memLevel < 1 || memLevel > MAX_MEM_LEVEL || method != Z_DEFLATED ||
+        windowBits < 8 || windowBits > 15 || level < 0 || level > 9 ||
+        strategy < 0 || strategy > Z_FIXED || (windowBits == 8 && wrap != 1)) {
+        return Z_STREAM_ERROR;
+    }
+    if (windowBits == 8) windowBits = 9;  /* until 256-byte window bug fixed */
+    s = (deflate_state *) ZALLOC(strm, 1, sizeof(deflate_state));
+    if (s == Z_NULL) return Z_MEM_ERROR;
+    strm->state = (struct internal_state FAR *)s;
+    s->strm = strm;
+    s->status = INIT_STATE;     /* to pass state test in deflateReset() */
+
+    s->wrap = wrap;
+    s->gzhead = Z_NULL;
+    s->w_bits = (uInt)windowBits;
+    s->w_size = 1 << s->w_bits;
+    s->w_mask = s->w_size - 1;
+
+    s->hash_bits = (uInt)memLevel + 7;
+    s->hash_size = 1 << s->hash_bits;
+    s->hash_mask = s->hash_size - 1;
+    s->hash_shift =  ((s->hash_bits+MIN_MATCH-1)/MIN_MATCH);
+
+    s->window = (Bytef *) ZALLOC(strm, s->w_size, 2*sizeof(Byte));
+    s->prev   = (Posf *)  ZALLOC(strm, s->w_size, sizeof(Pos));
+    s->head   = (Posf *)  ZALLOC(strm, s->hash_size, sizeof(Pos));
+
+    s->high_water = 0;      /* nothing written to s->window yet */
+
+    s->lit_bufsize = 1 << (memLevel + 6); /* 16K elements by default */
+
+    overlay = (ushf *) ZALLOC(strm, s->lit_bufsize, sizeof(ush)+2);
+    s->pending_buf = (uchf *) overlay;
+    s->pending_buf_size = (ulg)s->lit_bufsize * (sizeof(ush)+2L);
+
+    if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
+        s->pending_buf == Z_NULL) {
+        s->status = FINISH_STATE;
+        strm->msg = ERR_MSG(Z_MEM_ERROR);
+        deflateEnd (strm);
+        return Z_MEM_ERROR;
+    }
+    s->d_buf = overlay + s->lit_bufsize/sizeof(ush);
+    s->l_buf = s->pending_buf + (1+sizeof(ush))*s->lit_bufsize;
+
+    s->level = level;
+    s->strategy = strategy;
+    s->method = (Byte)method;
+
+    return deflateReset(strm);
+}
+
+/* =========================================================================
+ * Check for a valid deflate stream state. Return 0 if ok, 1 if not.
+ */
+local int deflateStateCheck (strm)
+    z_streamp strm;
+{
+    deflate_state *s;
+    if (strm == Z_NULL ||
+        strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0)
+        return 1;
+    s = strm->state;
+    if (s == Z_NULL || s->strm != strm || (s->status != INIT_STATE &&
+#ifdef GZIP
+                                           s->status != GZIP_STATE &&
+#endif
+                                           s->status != EXTRA_STATE &&
+                                           s->status != NAME_STATE &&
+                                           s->status != COMMENT_STATE &&
+                                           s->status != HCRC_STATE &&
+                                           s->status != BUSY_STATE &&
+                                           s->status != FINISH_STATE))
+        return 1;
+    return 0;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateSetDictionary (strm, dictionary, dictLength)
+    z_streamp strm;
+    const Bytef *dictionary;
+    uInt  dictLength;
+{
+    deflate_state *s;
+    uInt str, n;
+    int wrap;
+    unsigned avail;
+    z_const unsigned char *next;
+
+    if (deflateStateCheck(strm) || dictionary == Z_NULL)
+        return Z_STREAM_ERROR;
+    s = strm->state;
+    wrap = s->wrap;
+    if (wrap == 2 || (wrap == 1 && s->status != INIT_STATE) || s->lookahead)
+        return Z_STREAM_ERROR;
+
+    /* when using zlib wrappers, compute Adler-32 for provided dictionary */
+    if (wrap == 1)
+        strm->adler = adler32(strm->adler, dictionary, dictLength);
+    s->wrap = 0;                    /* avoid computing Adler-32 in read_buf */
+
+    /* if dictionary would fill window, just replace the history */
+    if (dictLength >= s->w_size) {
+        if (wrap == 0) {            /* already empty otherwise */
+            CLEAR_HASH(s);
+            s->strstart = 0;
+            s->block_start = 0L;
+            s->insert = 0;
+        }
+        dictionary += dictLength - s->w_size;  /* use the tail */
+        dictLength = s->w_size;
+    }
+
+    /* insert dictionary into window and hash */
+    avail = strm->avail_in;
+    next = strm->next_in;
+    strm->avail_in = dictLength;
+    strm->next_in = (z_const Bytef *)dictionary;
+    fill_window(s);
+    while (s->lookahead >= MIN_MATCH) {
+        str = s->strstart;
+        n = s->lookahead - (MIN_MATCH-1);
+        do {
+            UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+            s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+            s->head[s->ins_h] = (Pos)str;
+            str++;
+        } while (--n);
+        s->strstart = str;
+        s->lookahead = MIN_MATCH-1;
+        fill_window(s);
+    }
+    s->strstart += s->lookahead;
+    s->block_start = (long)s->strstart;
+    s->insert = s->lookahead;
+    s->lookahead = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    strm->next_in = next;
+    strm->avail_in = avail;
+    s->wrap = wrap;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateGetDictionary (strm, dictionary, dictLength)
+    z_streamp strm;
+    Bytef *dictionary;
+    uInt  *dictLength;
+{
+    deflate_state *s;
+    uInt len;
+
+    if (deflateStateCheck(strm))
+        return Z_STREAM_ERROR;
+    s = strm->state;
+    len = s->strstart + s->lookahead;
+    if (len > s->w_size)
+        len = s->w_size;
+    if (dictionary != Z_NULL && len)
+        zmemcpy(dictionary, s->window + s->strstart + s->lookahead - len, len);
+    if (dictLength != Z_NULL)
+        *dictLength = len;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateResetKeep (strm)
+    z_streamp strm;
+{
+    deflate_state *s;
+
+    if (deflateStateCheck(strm)) {
+        return Z_STREAM_ERROR;
+    }
+
+    strm->total_in = strm->total_out = 0;
+    strm->msg = Z_NULL; /* use zfree if we ever allocate msg dynamically */
+    strm->data_type = Z_UNKNOWN;
+
+    s = (deflate_state *)strm->state;
+    s->pending = 0;
+    s->pending_out = s->pending_buf;
+
+    if (s->wrap < 0) {
+        s->wrap = -s->wrap; /* was made negative by deflate(..., Z_FINISH); */
+    }
+    s->status =
+#ifdef GZIP
+        s->wrap == 2 ? GZIP_STATE :
+#endif
+        s->wrap ? INIT_STATE : BUSY_STATE;
+    strm->adler =
+#ifdef GZIP
+        s->wrap == 2 ? crc32(0L, Z_NULL, 0) :
+#endif
+        adler32(0L, Z_NULL, 0);
+    s->last_flush = Z_NO_FLUSH;
+
+    _tr_init(s);
+
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateReset (strm)
+    z_streamp strm;
+{
+    int ret;
+
+    ret = deflateResetKeep(strm);
+    if (ret == Z_OK)
+        lm_init(strm->state);
+    return ret;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateSetHeader (strm, head)
+    z_streamp strm;
+    gz_headerp head;
+{
+    if (deflateStateCheck(strm) || strm->state->wrap != 2)
+        return Z_STREAM_ERROR;
+    strm->state->gzhead = head;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflatePending (strm, pending, bits)
+    unsigned *pending;
+    int *bits;
+    z_streamp strm;
+{
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    if (pending != Z_NULL)
+        *pending = strm->state->pending;
+    if (bits != Z_NULL)
+        *bits = strm->state->bi_valid;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflatePrime (strm, bits, value)
+    z_streamp strm;
+    int bits;
+    int value;
+{
+    deflate_state *s;
+    int put;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+    if ((Bytef *)(s->d_buf) < s->pending_out + ((Buf_size + 7) >> 3))
+        return Z_BUF_ERROR;
+    do {
+        put = Buf_size - s->bi_valid;
+        if (put > bits)
+            put = bits;
+        s->bi_buf |= (ush)((value & ((1 << put) - 1)) << s->bi_valid);
+        s->bi_valid += put;
+        _tr_flush_bits(s);
+        value >>= put;
+        bits -= put;
+    } while (bits);
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateParams(strm, level, strategy)
+    z_streamp strm;
+    int level;
+    int strategy;
+{
+    deflate_state *s;
+    compress_func func;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+
+#ifdef FASTEST
+    if (level != 0) level = 1;
+#else
+    if (level == Z_DEFAULT_COMPRESSION) level = 6;
+#endif
+    if (level < 0 || level > 9 || strategy < 0 || strategy > Z_FIXED) {
+        return Z_STREAM_ERROR;
+    }
+    func = configuration_table[s->level].func;
+
+    if ((strategy != s->strategy || func != configuration_table[level].func) &&
+        s->high_water) {
+        /* Flush the last buffer: */
+        int err = deflate(strm, Z_BLOCK);
+        if (err == Z_STREAM_ERROR)
+            return err;
+        if (strm->avail_out == 0)
+            return Z_BUF_ERROR;
+    }
+    if (s->level != level) {
+        if (s->level == 0 && s->matches != 0) {
+            if (s->matches == 1)
+                slide_hash(s);
+            else
+                CLEAR_HASH(s);
+            s->matches = 0;
+        }
+        s->level = level;
+        s->max_lazy_match   = configuration_table[level].max_lazy;
+        s->good_match       = configuration_table[level].good_length;
+        s->nice_match       = configuration_table[level].nice_length;
+        s->max_chain_length = configuration_table[level].max_chain;
+    }
+    s->strategy = strategy;
+    return Z_OK;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateTune(strm, good_length, max_lazy, nice_length, max_chain)
+    z_streamp strm;
+    int good_length;
+    int max_lazy;
+    int nice_length;
+    int max_chain;
+{
+    deflate_state *s;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+    s = strm->state;
+    s->good_match = (uInt)good_length;
+    s->max_lazy_match = (uInt)max_lazy;
+    s->nice_match = nice_length;
+    s->max_chain_length = (uInt)max_chain;
+    return Z_OK;
+}
+
+/* =========================================================================
+ * For the default windowBits of 15 and memLevel of 8, this function returns
+ * a close to exact, as well as small, upper bound on the compressed size.
+ * They are coded as constants here for a reason--if the #define's are
+ * changed, then this function needs to be changed as well.  The return
+ * value for 15 and 8 only works for those exact settings.
+ *
+ * For any setting other than those defaults for windowBits and memLevel,
+ * the value returned is a conservative worst case for the maximum expansion
+ * resulting from using fixed blocks instead of stored blocks, which deflate
+ * can emit on compressed data for some combinations of the parameters.
+ *
+ * This function could be more sophisticated to provide closer upper bounds for
+ * every combination of windowBits and memLevel.  But even the conservative
+ * upper bound of about 14% expansion does not seem onerous for output buffer
+ * allocation.
+ */
+uLong ZEXPORT deflateBound(strm, sourceLen)
+    z_streamp strm;
+    uLong sourceLen;
+{
+    deflate_state *s;
+    uLong complen, wraplen;
+
+    /* conservative upper bound for compressed data */
+    complen = sourceLen +
+              ((sourceLen + 7) >> 3) + ((sourceLen + 63) >> 6) + 5;
+
+    /* if can't get parameters, return conservative bound plus zlib wrapper */
+    if (deflateStateCheck(strm))
+        return complen + 6;
+
+    /* compute wrapper length */
+    s = strm->state;
+    switch (s->wrap) {
+    case 0:                                 /* raw deflate */
+        wraplen = 0;
+        break;
+    case 1:                                 /* zlib wrapper */
+        wraplen = 6 + (s->strstart ? 4 : 0);
+        break;
+#ifdef GZIP
+    case 2:                                 /* gzip wrapper */
+        wraplen = 18;
+        if (s->gzhead != Z_NULL) {          /* user-supplied gzip header */
+            Bytef *str;
+            if (s->gzhead->extra != Z_NULL)
+                wraplen += 2 + s->gzhead->extra_len;
+            str = s->gzhead->name;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            str = s->gzhead->comment;
+            if (str != Z_NULL)
+                do {
+                    wraplen++;
+                } while (*str++);
+            if (s->gzhead->hcrc)
+                wraplen += 2;
+        }
+        break;
+#endif
+    default:                                /* for compiler happiness */
+        wraplen = 6;
+    }
+
+    /* if not default parameters, return conservative bound */
+    if (s->w_bits != 15 || s->hash_bits != 8 + 7)
+        return complen + wraplen;
+
+    /* default settings: return tight bound for that case */
+    return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
+           (sourceLen >> 25) + 13 - 6 + wraplen;
+}
+
+/* =========================================================================
+ * Put a short in the pending buffer. The 16-bit value is put in MSB order.
+ * IN assertion: the stream state is correct and there is enough room in
+ * pending_buf.
+ */
+local void putShortMSB (s, b)
+    deflate_state *s;
+    uInt b;
+{
+    put_byte(s, (Byte)(b >> 8));
+    put_byte(s, (Byte)(b & 0xff));
+}
+
+/* =========================================================================
+ * Flush as much pending output as possible. All deflate() output, except for
+ * some deflate_stored() output, goes through this function so some
+ * applications may wish to modify it to avoid allocating a large
+ * strm->next_out buffer and copying into it. (See also read_buf()).
+ */
+local void flush_pending(strm)
+    z_streamp strm;
+{
+    unsigned len;
+    deflate_state *s = strm->state;
+
+    _tr_flush_bits(s);
+    len = s->pending;
+    if (len > strm->avail_out) len = strm->avail_out;
+    if (len == 0) return;
+
+    zmemcpy(strm->next_out, s->pending_out, len);
+    strm->next_out  += len;
+    s->pending_out  += len;
+    strm->total_out += len;
+    strm->avail_out -= len;
+    s->pending      -= len;
+    if (s->pending == 0) {
+        s->pending_out = s->pending_buf;
+    }
+}
+
+/* ===========================================================================
+ * Update the header CRC with the bytes s->pending_buf[beg..s->pending - 1].
+ */
+#define HCRC_UPDATE(beg) \
+    do { \
+        if (s->gzhead->hcrc && s->pending > (beg)) \
+            strm->adler = crc32(strm->adler, s->pending_buf + (beg), \
+                                s->pending - (beg)); \
+    } while (0)
+
+/* ========================================================================= */
+int ZEXPORT deflate (strm, flush)
+    z_streamp strm;
+    int flush;
+{
+    int old_flush; /* value of flush param for previous deflate call */
+    deflate_state *s;
+
+    if (deflateStateCheck(strm) || flush > Z_BLOCK || flush < 0) {
+        return Z_STREAM_ERROR;
+    }
+    s = strm->state;
+
+    if (strm->next_out == Z_NULL ||
+        (strm->avail_in != 0 && strm->next_in == Z_NULL) ||
+        (s->status == FINISH_STATE && flush != Z_FINISH)) {
+        ERR_RETURN(strm, Z_STREAM_ERROR);
+    }
+    if (strm->avail_out == 0) ERR_RETURN(strm, Z_BUF_ERROR);
+
+    old_flush = s->last_flush;
+    s->last_flush = flush;
+
+    /* Flush as much pending output as possible */
+    if (s->pending != 0) {
+        flush_pending(strm);
+        if (strm->avail_out == 0) {
+            /* Since avail_out is 0, deflate will be called again with
+             * more output space, but possibly with both pending and
+             * avail_in equal to zero. There won't be anything to do,
+             * but this is not an error situation so make sure we
+             * return OK instead of BUF_ERROR at next call of deflate:
+             */
+            s->last_flush = -1;
+            return Z_OK;
+        }
+
+    /* Make sure there is something to do and avoid duplicate consecutive
+     * flushes. For repeated and useless calls with Z_FINISH, we keep
+     * returning Z_STREAM_END instead of Z_BUF_ERROR.
+     */
+    } else if (strm->avail_in == 0 && RANK(flush) <= RANK(old_flush) &&
+               flush != Z_FINISH) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* User must not provide more input after the first FINISH: */
+    if (s->status == FINISH_STATE && strm->avail_in != 0) {
+        ERR_RETURN(strm, Z_BUF_ERROR);
+    }
+
+    /* Write the header */
+    if (s->status == INIT_STATE) {
+        /* zlib header */
+        uInt header = (Z_DEFLATED + ((s->w_bits-8)<<4)) << 8;
+        uInt level_flags;
+
+        if (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2)
+            level_flags = 0;
+        else if (s->level < 6)
+            level_flags = 1;
+        else if (s->level == 6)
+            level_flags = 2;
+        else
+            level_flags = 3;
+        header |= (level_flags << 6);
+        if (s->strstart != 0) header |= PRESET_DICT;
+        header += 31 - (header % 31);
+
+        putShortMSB(s, header);
+
+        /* Save the adler32 of the preset dictionary: */
+        if (s->strstart != 0) {
+            putShortMSB(s, (uInt)(strm->adler >> 16));
+            putShortMSB(s, (uInt)(strm->adler & 0xffff));
+        }
+        strm->adler = adler32(0L, Z_NULL, 0);
+        s->status = BUSY_STATE;
+
+        /* Compression must start with an empty pending buffer */
+        flush_pending(strm);
+        if (s->pending != 0) {
+            s->last_flush = -1;
+            return Z_OK;
+        }
+    }
+#ifdef GZIP
+    if (s->status == GZIP_STATE) {
+        /* gzip header */
+        strm->adler = crc32(0L, Z_NULL, 0);
+        put_byte(s, 31);
+        put_byte(s, 139);
+        put_byte(s, 8);
+        if (s->gzhead == Z_NULL) {
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, 0);
+            put_byte(s, s->level == 9 ? 2 :
+                     (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
+                      4 : 0));
+            put_byte(s, OS_CODE);
+            s->status = BUSY_STATE;
+
+            /* Compression must start with an empty pending buffer */
+            flush_pending(strm);
+            if (s->pending != 0) {
+                s->last_flush = -1;
+                return Z_OK;
+            }
+        }
+        else {
+            put_byte(s, (s->gzhead->text ? 1 : 0) +
+                     (s->gzhead->hcrc ? 2 : 0) +
+                     (s->gzhead->extra == Z_NULL ? 0 : 4) +
+                     (s->gzhead->name == Z_NULL ? 0 : 8) +
+                     (s->gzhead->comment == Z_NULL ? 0 : 16)
+                     );
+            put_byte(s, (Byte)(s->gzhead->time & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 8) & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 16) & 0xff));
+            put_byte(s, (Byte)((s->gzhead->time >> 24) & 0xff));
+            put_byte(s, s->level == 9 ? 2 :
+                     (s->strategy >= Z_HUFFMAN_ONLY || s->level < 2 ?
+                      4 : 0));
+            put_byte(s, s->gzhead->os & 0xff);
+            if (s->gzhead->extra != Z_NULL) {
+                put_byte(s, s->gzhead->extra_len & 0xff);
+                put_byte(s, (s->gzhead->extra_len >> 8) & 0xff);
+            }
+            if (s->gzhead->hcrc)
+                strm->adler = crc32(strm->adler, s->pending_buf,
+                                    s->pending);
+            s->gzindex = 0;
+            s->status = EXTRA_STATE;
+        }
+    }
+    if (s->status == EXTRA_STATE) {
+        if (s->gzhead->extra != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            uInt left = (s->gzhead->extra_len & 0xffff) - s->gzindex;
+            while (s->pending + left > s->pending_buf_size) {
+                uInt copy = s->pending_buf_size - s->pending;
+                zmemcpy(s->pending_buf + s->pending,
+                        s->gzhead->extra + s->gzindex, copy);
+                s->pending = s->pending_buf_size;
+                HCRC_UPDATE(beg);
+                s->gzindex += copy;
+                flush_pending(strm);
+                if (s->pending != 0) {
+                    s->last_flush = -1;
+                    return Z_OK;
+                }
+                beg = 0;
+                left -= copy;
+            }
+            zmemcpy(s->pending_buf + s->pending,
+                    s->gzhead->extra + s->gzindex, left);
+            s->pending += left;
+            HCRC_UPDATE(beg);
+            s->gzindex = 0;
+        }
+        s->status = NAME_STATE;
+    }
+    if (s->status == NAME_STATE) {
+        if (s->gzhead->name != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            int val;
+            do {
+                if (s->pending == s->pending_buf_size) {
+                    HCRC_UPDATE(beg);
+                    flush_pending(strm);
+                    if (s->pending != 0) {
+                        s->last_flush = -1;
+                        return Z_OK;
+                    }
+                    beg = 0;
+                }
+                val = s->gzhead->name[s->gzindex++];
+                put_byte(s, val);
+            } while (val != 0);
+            HCRC_UPDATE(beg);
+            s->gzindex = 0;
+        }
+        s->status = COMMENT_STATE;
+    }
+    if (s->status == COMMENT_STATE) {
+        if (s->gzhead->comment != Z_NULL) {
+            ulg beg = s->pending;   /* start of bytes to update crc */
+            int val;
+            do {
+                if (s->pending == s->pending_buf_size) {
+                    HCRC_UPDATE(beg);
+                    flush_pending(strm);
+                    if (s->pending != 0) {
+                        s->last_flush = -1;
+                        return Z_OK;
+                    }
+                    beg = 0;
+                }
+                val = s->gzhead->comment[s->gzindex++];
+                put_byte(s, val);
+            } while (val != 0);
+            HCRC_UPDATE(beg);
+        }
+        s->status = HCRC_STATE;
+    }
+    if (s->status == HCRC_STATE) {
+        if (s->gzhead->hcrc) {
+            if (s->pending + 2 > s->pending_buf_size) {
+                flush_pending(strm);
+                if (s->pending != 0) {
+                    s->last_flush = -1;
+                    return Z_OK;
+                }
+            }
+            put_byte(s, (Byte)(strm->adler & 0xff));
+            put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
+            strm->adler = crc32(0L, Z_NULL, 0);
+        }
+        s->status = BUSY_STATE;
+
+        /* Compression must start with an empty pending buffer */
+        flush_pending(strm);
+        if (s->pending != 0) {
+            s->last_flush = -1;
+            return Z_OK;
+        }
+    }
+#endif
+
+    /* Start a new block or continue the current one.
+     */
+    if (strm->avail_in != 0 || s->lookahead != 0 ||
+        (flush != Z_NO_FLUSH && s->status != FINISH_STATE)) {
+        block_state bstate;
+
+        bstate = s->level == 0 ? deflate_stored(s, flush) :
+                 s->strategy == Z_HUFFMAN_ONLY ? deflate_huff(s, flush) :
+                 s->strategy == Z_RLE ? deflate_rle(s, flush) :
+                 (*(configuration_table[s->level].func))(s, flush);
+
+        if (bstate == finish_started || bstate == finish_done) {
+            s->status = FINISH_STATE;
+        }
+        if (bstate == need_more || bstate == finish_started) {
+            if (strm->avail_out == 0) {
+                s->last_flush = -1; /* avoid BUF_ERROR next call, see above */
+            }
+            return Z_OK;
+            /* If flush != Z_NO_FLUSH && avail_out == 0, the next call
+             * of deflate should use the same flush parameter to make sure
+             * that the flush is complete. So we don't have to output an
+             * empty block here, this will be done at next call. This also
+             * ensures that for a very small output buffer, we emit at most
+             * one empty block.
+             */
+        }
+        if (bstate == block_done) {
+            if (flush == Z_PARTIAL_FLUSH) {
+                _tr_align(s);
+            } else if (flush != Z_BLOCK) { /* FULL_FLUSH or SYNC_FLUSH */
+                _tr_stored_block(s, (char*)0, 0L, 0);
+                /* For a full flush, this empty block will be recognized
+                 * as a special marker by inflate_sync().
+                 */
+                if (flush == Z_FULL_FLUSH) {
+                    CLEAR_HASH(s);             /* forget history */
+                    if (s->lookahead == 0) {
+                        s->strstart = 0;
+                        s->block_start = 0L;
+                        s->insert = 0;
+                    }
+                }
+            }
+            flush_pending(strm);
+            if (strm->avail_out == 0) {
+              s->last_flush = -1; /* avoid BUF_ERROR at next call, see above */
+              return Z_OK;
+            }
+        }
+    }
+
+    if (flush != Z_FINISH) return Z_OK;
+    if (s->wrap <= 0) return Z_STREAM_END;
+
+    /* Write the trailer */
+#ifdef GZIP
+    if (s->wrap == 2) {
+        put_byte(s, (Byte)(strm->adler & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 8) & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 16) & 0xff));
+        put_byte(s, (Byte)((strm->adler >> 24) & 0xff));
+        put_byte(s, (Byte)(strm->total_in & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 8) & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 16) & 0xff));
+        put_byte(s, (Byte)((strm->total_in >> 24) & 0xff));
+    }
+    else
+#endif
+    {
+        putShortMSB(s, (uInt)(strm->adler >> 16));
+        putShortMSB(s, (uInt)(strm->adler & 0xffff));
+    }
+    flush_pending(strm);
+    /* If avail_out is zero, the application will call deflate again
+     * to flush the rest.
+     */
+    if (s->wrap > 0) s->wrap = -s->wrap; /* write the trailer only once! */
+    return s->pending != 0 ? Z_OK : Z_STREAM_END;
+}
+
+/* ========================================================================= */
+int ZEXPORT deflateEnd (strm)
+    z_streamp strm;
+{
+    int status;
+
+    if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
+
+    status = strm->state->status;
+
+    /* Deallocate in reverse order of allocations: */
+    TRY_FREE(strm, strm->state->pending_buf);
+    TRY_FREE(strm, strm->state->head);
+    TRY_FREE(strm, strm->state->prev);
+    TRY_FREE(strm, strm->state->window);
+
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+
+    return status == BUSY_STATE ? Z_DATA_ERROR : Z_OK;
+}
+
+/* =========================================================================
+ * Copy the source state to the destination state.
+ * To simplify the source, this is not supported for 16-bit MSDOS (which
+ * doesn't have enough memory anyway to duplicate compression states).
+ */
+int ZEXPORT deflateCopy (dest, source)
+    z_streamp dest;
+    z_streamp source;
+{
+#ifdef MAXSEG_64K
+    return Z_STREAM_ERROR;
+#else
+    deflate_state *ds;
+    deflate_state *ss;
+    ushf *overlay;
+
+
+    if (deflateStateCheck(source) || dest == Z_NULL) {
+        return Z_STREAM_ERROR;
+    }
+
+    ss = source->state;
+
+    zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
+
+    ds = (deflate_state *) ZALLOC(dest, 1, sizeof(deflate_state));
+    if (ds == Z_NULL) return Z_MEM_ERROR;
+    dest->state = (struct internal_state FAR *) ds;
+    zmemcpy((voidpf)ds, (voidpf)ss, sizeof(deflate_state));
+    ds->strm = dest;
+
+    ds->window = (Bytef *) ZALLOC(dest, ds->w_size, 2*sizeof(Byte));
+    ds->prev   = (Posf *)  ZALLOC(dest, ds->w_size, sizeof(Pos));
+    ds->head   = (Posf *)  ZALLOC(dest, ds->hash_size, sizeof(Pos));
+    overlay = (ushf *) ZALLOC(dest, ds->lit_bufsize, sizeof(ush)+2);
+    ds->pending_buf = (uchf *) overlay;
+
+    if (ds->window == Z_NULL || ds->prev == Z_NULL || ds->head == Z_NULL ||
+        ds->pending_buf == Z_NULL) {
+        deflateEnd (dest);
+        return Z_MEM_ERROR;
+    }
+    /* following zmemcpy do not work for 16-bit MSDOS */
+    zmemcpy(ds->window, ss->window, ds->w_size * 2 * sizeof(Byte));
+    zmemcpy((voidpf)ds->prev, (voidpf)ss->prev, ds->w_size * sizeof(Pos));
+    zmemcpy((voidpf)ds->head, (voidpf)ss->head, ds->hash_size * sizeof(Pos));
+    zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
+
+    ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
+    ds->d_buf = overlay + ds->lit_bufsize/sizeof(ush);
+    ds->l_buf = ds->pending_buf + (1+sizeof(ush))*ds->lit_bufsize;
+
+    ds->l_desc.dyn_tree = ds->dyn_ltree;
+    ds->d_desc.dyn_tree = ds->dyn_dtree;
+    ds->bl_desc.dyn_tree = ds->bl_tree;
+
+    return Z_OK;
+#endif /* MAXSEG_64K */
+}
+
+/* ===========================================================================
+ * Read a new buffer from the current input stream, update the adler32
+ * and total number of bytes read.  All deflate() input goes through
+ * this function so some applications may wish to modify it to avoid
+ * allocating a large strm->next_in buffer and copying from it.
+ * (See also flush_pending()).
+ */
+local unsigned read_buf(strm, buf, size)
+    z_streamp strm;
+    Bytef *buf;
+    unsigned size;
+{
+    unsigned len = strm->avail_in;
+
+    if (len > size) len = size;
+    if (len == 0) return 0;
+
+    strm->avail_in  -= len;
+
+    zmemcpy(buf, strm->next_in, len);
+    if (strm->state->wrap == 1) {
+        strm->adler = adler32(strm->adler, buf, len);
+    }
+#ifdef GZIP
+    else if (strm->state->wrap == 2) {
+        strm->adler = crc32(strm->adler, buf, len);
+    }
+#endif
+    strm->next_in  += len;
+    strm->total_in += len;
+
+    return len;
+}
+
+/* ===========================================================================
+ * Initialize the "longest match" routines for a new zlib stream
+ */
+local void lm_init (s)
+    deflate_state *s;
+{
+    s->window_size = (ulg)2L*s->w_size;
+
+    CLEAR_HASH(s);
+
+    /* Set the default configuration parameters:
+     */
+    s->max_lazy_match   = configuration_table[s->level].max_lazy;
+    s->good_match       = configuration_table[s->level].good_length;
+    s->nice_match       = configuration_table[s->level].nice_length;
+    s->max_chain_length = configuration_table[s->level].max_chain;
+
+    s->strstart = 0;
+    s->block_start = 0L;
+    s->lookahead = 0;
+    s->insert = 0;
+    s->match_length = s->prev_length = MIN_MATCH-1;
+    s->match_available = 0;
+    s->ins_h = 0;
+#ifndef FASTEST
+#ifdef ASMV
+    match_init(); /* initialize the asm code */
+#endif
+#endif
+}
+
+#ifndef FASTEST
+/* ===========================================================================
+ * Set match_start to the longest match starting at the given string and
+ * return its length. Matches shorter or equal to prev_length are discarded,
+ * in which case the result is equal to prev_length and match_start is
+ * garbage.
+ * IN assertions: cur_match is the head of the hash chain for the current
+ *   string (strstart) and its distance is <= MAX_DIST, and prev_length >= 1
+ * OUT assertion: the match length is not greater than s->lookahead.
+ */
+#ifndef ASMV
+/* For 80x86 and 680x0, an optimized version will be provided in match.asm or
+ * match.S. The code will be functionally equivalent.
+ */
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    unsigned chain_length = s->max_chain_length;/* max hash chain length */
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                      /* matched string */
+    register int len;                           /* length of current match */
+    int best_len = (int)s->prev_length;         /* best match length so far */
+    int nice_match = s->nice_match;             /* stop if match long enough */
+    IPos limit = s->strstart > (IPos)MAX_DIST(s) ?
+        s->strstart - (IPos)MAX_DIST(s) : NIL;
+    /* Stop when cur_match becomes <= limit. To simplify the code,
+     * we prevent matches with the string of window index 0.
+     */
+    Posf *prev = s->prev;
+    uInt wmask = s->w_mask;
+
+#ifdef UNALIGNED_OK
+    /* Compare two bytes at a time. Note: this is not always beneficial.
+     * Try with and without -DUNALIGNED_OK to check.
+     */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH - 1;
+    register ush scan_start = *(ushf*)scan;
+    register ush scan_end   = *(ushf*)(scan+best_len-1);
+#else
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+    register Byte scan_end1  = scan[best_len-1];
+    register Byte scan_end   = scan[best_len];
+#endif
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    /* Do not waste too much time if we already have a good match: */
+    if (s->prev_length >= s->good_match) {
+        chain_length >>= 2;
+    }
+    /* Do not look for matches beyond the end of the input. This is necessary
+     * to make deflate deterministic.
+     */
+    if ((uInt)nice_match > s->lookahead) nice_match = (int)s->lookahead;
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    do {
+        Assert(cur_match < s->strstart, "no future");
+        match = s->window + cur_match;
+
+        /* Skip to next match if the match length cannot increase
+         * or if the match length is less than 2.  Note that the checks below
+         * for insufficient lookahead only occur occasionally for performance
+         * reasons.  Therefore uninitialized memory will be accessed, and
+         * conditional jumps will be made that depend on those values.
+         * However the length of the match is limited to the lookahead, so
+         * the output of deflate is not affected by the uninitialized values.
+         */
+#if (defined(UNALIGNED_OK) && MAX_MATCH == 258)
+        /* This code assumes sizeof(unsigned short) == 2. Do not use
+         * UNALIGNED_OK if your compiler uses a different size.
+         */
+        if (*(ushf*)(match+best_len-1) != scan_end ||
+            *(ushf*)match != scan_start) continue;
+
+        /* It is not necessary to compare scan[2] and match[2] since they are
+         * always equal when the other bytes match, given that the hash keys
+         * are equal and that HASH_BITS >= 8. Compare 2 bytes at a time at
+         * strstart+3, +5, ... up to strstart+257. We check for insufficient
+         * lookahead only every 4th comparison; the 128th check will be made
+         * at strstart+257. If MAX_MATCH-2 is not a multiple of 8, it is
+         * necessary to put more guard bytes at the end of the window, or
+         * to check more often for insufficient lookahead.
+         */
+        Assert(scan[2] == match[2], "scan[2]?");
+        scan++, match++;
+        do {
+        } while (*(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 *(ushf*)(scan+=2) == *(ushf*)(match+=2) &&
+                 scan < strend);
+        /* The funny "do {}" generates better code on most compilers */
+
+        /* Here, scan <= window+strstart+257 */
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+        if (*scan == *match) scan++;
+
+        len = (MAX_MATCH - 1) - (int)(strend-scan);
+        scan = strend - (MAX_MATCH-1);
+
+#else /* UNALIGNED_OK */
+
+        if (match[best_len]   != scan_end  ||
+            match[best_len-1] != scan_end1 ||
+            *match            != *scan     ||
+            *++match          != scan[1])      continue;
+
+        /* The check at best_len-1 can be removed because it will be made
+         * again later. (This heuristic is not always a win.)
+         * It is not necessary to compare scan[2] and match[2] since they
+         * are always equal when the other bytes match, given that
+         * the hash keys are equal and that HASH_BITS >= 8.
+         */
+        scan += 2, match++;
+        Assert(*scan == *match, "match[2]?");
+
+        /* We check for insufficient lookahead only every 8th comparison;
+         * the 256th check will be made at strstart+258.
+         */
+        do {
+        } while (*++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 *++scan == *++match && *++scan == *++match &&
+                 scan < strend);
+
+        Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+        len = MAX_MATCH - (int)(strend - scan);
+        scan = strend - MAX_MATCH;
+
+#endif /* UNALIGNED_OK */
+
+        if (len > best_len) {
+            s->match_start = cur_match;
+            best_len = len;
+            if (len >= nice_match) break;
+#ifdef UNALIGNED_OK
+            scan_end = *(ushf*)(scan+best_len-1);
+#else
+            scan_end1  = scan[best_len-1];
+            scan_end   = scan[best_len];
+#endif
+        }
+    } while ((cur_match = prev[cur_match & wmask]) > limit
+             && --chain_length != 0);
+
+    if ((uInt)best_len <= s->lookahead) return (uInt)best_len;
+    return s->lookahead;
+}
+#endif /* ASMV */
+
+#else /* FASTEST */
+
+/* ---------------------------------------------------------------------------
+ * Optimized version for FASTEST only
+ */
+local uInt longest_match(s, cur_match)
+    deflate_state *s;
+    IPos cur_match;                             /* current match */
+{
+    register Bytef *scan = s->window + s->strstart; /* current string */
+    register Bytef *match;                       /* matched string */
+    register int len;                           /* length of current match */
+    register Bytef *strend = s->window + s->strstart + MAX_MATCH;
+
+    /* The code is optimized for HASH_BITS >= 8 and MAX_MATCH-2 multiple of 16.
+     * It is easy to get rid of this optimization if necessary.
+     */
+    Assert(s->hash_bits >= 8 && MAX_MATCH == 258, "Code too clever");
+
+    Assert((ulg)s->strstart <= s->window_size-MIN_LOOKAHEAD, "need lookahead");
+
+    Assert(cur_match < s->strstart, "no future");
+
+    match = s->window + cur_match;
+
+    /* Return failure if the match length is less than 2:
+     */
+    if (match[0] != scan[0] || match[1] != scan[1]) return MIN_MATCH-1;
+
+    /* The check at best_len-1 can be removed because it will be made
+     * again later. (This heuristic is not always a win.)
+     * It is not necessary to compare scan[2] and match[2] since they
+     * are always equal when the other bytes match, given that
+     * the hash keys are equal and that HASH_BITS >= 8.
+     */
+    scan += 2, match += 2;
+    Assert(*scan == *match, "match[2]?");
+
+    /* We check for insufficient lookahead only every 8th comparison;
+     * the 256th check will be made at strstart+258.
+     */
+    do {
+    } while (*++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             *++scan == *++match && *++scan == *++match &&
+             scan < strend);
+
+    Assert(scan <= s->window+(unsigned)(s->window_size-1), "wild scan");
+
+    len = MAX_MATCH - (int)(strend - scan);
+
+    if (len < MIN_MATCH) return MIN_MATCH - 1;
+
+    s->match_start = cur_match;
+    return (uInt)len <= s->lookahead ? (uInt)len : s->lookahead;
+}
+
+#endif /* FASTEST */
+
+#ifdef ZLIB_DEBUG
+
+#define EQUAL 0
+/* result of memcmp for equal strings */
+
+/* ===========================================================================
+ * Check that the match at match_start is indeed a match.
+ */
+local void check_match(s, start, match, length)
+    deflate_state *s;
+    IPos start, match;
+    int length;
+{
+    /* check that the match is indeed a match */
+    if (zmemcmp(s->window + match,
+                s->window + start, length) != EQUAL) {
+        fprintf(stderr, " start %u, match %u, length %d\n",
+                start, match, length);
+        do {
+            fprintf(stderr, "%c%c", s->window[match++], s->window[start++]);
+        } while (--length != 0);
+        z_error("invalid match");
+    }
+    if (z_verbose > 1) {
+        fprintf(stderr,"\\[%d,%d]", start-match, length);
+        do { putc(s->window[start++], stderr); } while (--length != 0);
+    }
+}
+#else
+#  define check_match(s, start, match, length)
+#endif /* ZLIB_DEBUG */
+
+/* ===========================================================================
+ * Fill the window when the lookahead becomes insufficient.
+ * Updates strstart and lookahead.
+ *
+ * IN assertion: lookahead < MIN_LOOKAHEAD
+ * OUT assertions: strstart <= window_size-MIN_LOOKAHEAD
+ *    At least one byte has been read, or avail_in == 0; reads are
+ *    performed for at least two bytes (required for the zip translate_eol
+ *    option -- not supported here).
+ */
+local void fill_window(s)
+    deflate_state *s;
+{
+    unsigned n;
+    unsigned more;    /* Amount of free space at the end of the window. */
+    uInt wsize = s->w_size;
+
+    Assert(s->lookahead < MIN_LOOKAHEAD, "already enough lookahead");
+
+    do {
+        more = (unsigned)(s->window_size -(ulg)s->lookahead -(ulg)s->strstart);
+
+        /* Deal with !@#$% 64K limit: */
+        if (sizeof(int) <= 2) {
+            if (more == 0 && s->strstart == 0 && s->lookahead == 0) {
+                more = wsize;
+
+            } else if (more == (unsigned)(-1)) {
+                /* Very unlikely, but possible on 16 bit machine if
+                 * strstart == 0 && lookahead == 1 (input done a byte at time)
+                 */
+                more--;
+            }
+        }
+
+        /* If the window is almost full and there is insufficient lookahead,
+         * move the upper half to the lower one to make room in the upper half.
+         */
+        if (s->strstart >= wsize+MAX_DIST(s)) {
+
+            zmemcpy(s->window, s->window+wsize, (unsigned)wsize - more);
+            s->match_start -= wsize;
+            s->strstart    -= wsize; /* we now have strstart >= MAX_DIST */
+            s->block_start -= (long) wsize;
+            slide_hash(s);
+            more += wsize;
+        }
+        if (s->strm->avail_in == 0) break;
+
+        /* If there was no sliding:
+         *    strstart <= WSIZE+MAX_DIST-1 && lookahead <= MIN_LOOKAHEAD - 1 &&
+         *    more == window_size - lookahead - strstart
+         * => more >= window_size - (MIN_LOOKAHEAD-1 + WSIZE + MAX_DIST-1)
+         * => more >= window_size - 2*WSIZE + 2
+         * In the BIG_MEM or MMAP case (not yet supported),
+         *   window_size == input_size + MIN_LOOKAHEAD  &&
+         *   strstart + s->lookahead <= input_size => more >= MIN_LOOKAHEAD.
+         * Otherwise, window_size == 2*WSIZE so more >= 2.
+         * If there was sliding, more >= WSIZE. So in all cases, more >= 2.
+         */
+        Assert(more >= 2, "more < 2");
+
+        n = read_buf(s->strm, s->window + s->strstart + s->lookahead, more);
+        s->lookahead += n;
+
+        /* Initialize the hash value now that we have some input: */
+        if (s->lookahead + s->insert >= MIN_MATCH) {
+            uInt str = s->strstart - s->insert;
+            s->ins_h = s->window[str];
+            UPDATE_HASH(s, s->ins_h, s->window[str + 1]);
+#if MIN_MATCH != 3
+            Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+            while (s->insert) {
+                UPDATE_HASH(s, s->ins_h, s->window[str + MIN_MATCH-1]);
+#ifndef FASTEST
+                s->prev[str & s->w_mask] = s->head[s->ins_h];
+#endif
+                s->head[s->ins_h] = (Pos)str;
+                str++;
+                s->insert--;
+                if (s->lookahead + s->insert < MIN_MATCH)
+                    break;
+            }
+        }
+        /* If the whole input has less than MIN_MATCH bytes, ins_h is garbage,
+         * but this is not important since only literal bytes will be emitted.
+         */
+
+    } while (s->lookahead < MIN_LOOKAHEAD && s->strm->avail_in != 0);
+
+    /* If the WIN_INIT bytes after the end of the current data have never been
+     * written, then zero those bytes in order to avoid memory check reports of
+     * the use of uninitialized (or uninitialised as Julian writes) bytes by
+     * the longest match routines.  Update the high water mark for the next
+     * time through here.  WIN_INIT is set to MAX_MATCH since the longest match
+     * routines allow scanning to strstart + MAX_MATCH, ignoring lookahead.
+     */
+    if (s->high_water < s->window_size) {
+        ulg curr = s->strstart + (ulg)(s->lookahead);
+        ulg init;
+
+        if (s->high_water < curr) {
+            /* Previous high water mark below current data -- zero WIN_INIT
+             * bytes or up to end of window, whichever is less.
+             */
+            init = s->window_size - curr;
+            if (init > WIN_INIT)
+                init = WIN_INIT;
+            zmemzero(s->window + curr, (unsigned)init);
+            s->high_water = curr + init;
+        }
+        else if (s->high_water < (ulg)curr + WIN_INIT) {
+            /* High water mark at or above current data, but below current data
+             * plus WIN_INIT -- zero out to current data plus WIN_INIT, or up
+             * to end of window, whichever is less.
+             */
+            init = (ulg)curr + WIN_INIT - s->high_water;
+            if (init > s->window_size - s->high_water)
+                init = s->window_size - s->high_water;
+            zmemzero(s->window + s->high_water, (unsigned)init);
+            s->high_water += init;
+        }
+    }
+
+    Assert((ulg)s->strstart <= s->window_size - MIN_LOOKAHEAD,
+           "not enough room for search");
+}
+
+/* ===========================================================================
+ * Flush the current block, with given end-of-file flag.
+ * IN assertion: strstart is set to the end of the current match.
+ */
+#define FLUSH_BLOCK_ONLY(s, last) { \
+   _tr_flush_block(s, (s->block_start >= 0L ? \
+                   (charf *)&s->window[(unsigned)s->block_start] : \
+                   (charf *)Z_NULL), \
+                (ulg)((long)s->strstart - s->block_start), \
+                (last)); \
+   s->block_start = s->strstart; \
+   flush_pending(s->strm); \
+   Tracev((stderr,"[FLUSH]")); \
+}
+
+/* Same but force premature exit if necessary. */
+#define FLUSH_BLOCK(s, last) { \
+   FLUSH_BLOCK_ONLY(s, last); \
+   if (s->strm->avail_out == 0) return (last) ? finish_started : need_more; \
+}
+
+/* Maximum stored block length in deflate format (not including header). */
+#define MAX_STORED 65535
+
+/* Minimum of a and b. */
+#define MIN(a, b) ((a) > (b) ? (b) : (a))
+
+/* ===========================================================================
+ * Copy without compression as much as possible from the input stream, return
+ * the current block state.
+ *
+ * In case deflateParams() is used to later switch to a non-zero compression
+ * level, s->matches (otherwise unused when storing) keeps track of the number
+ * of hash table slides to perform. If s->matches is 1, then one hash table
+ * slide will be done when switching. If s->matches is 2, the maximum value
+ * allowed here, then the hash table will be cleared, since two or more slides
+ * is the same as a clear.
+ *
+ * deflate_stored() is written to minimize the number of times an input byte is
+ * copied. It is most efficient with large input and output buffers, which
+ * maximizes the opportunites to have a single copy from next_in to next_out.
+ */
+local block_state deflate_stored(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    /* Smallest worthy block size when not flushing or finishing. By default
+     * this is 32K. This can be as small as 507 bytes for memLevel == 1. For
+     * large input and output buffers, the stored block size will be larger.
+     */
+    unsigned min_block = MIN(s->pending_buf_size - 5, s->w_size);
+
+    /* Copy as many min_block or larger stored blocks directly to next_out as
+     * possible. If flushing, copy the remaining available input to next_out as
+     * stored blocks, if there is enough space.
+     */
+    unsigned len, left, have, last = 0;
+    unsigned used = s->strm->avail_in;
+    do {
+        /* Set len to the maximum size block that we can copy directly with the
+         * available input data and output space. Set left to how much of that
+         * would be copied from what's left in the window.
+         */
+        len = MAX_STORED;       /* maximum deflate stored block length */
+        have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
+        if (s->strm->avail_out < have)          /* need room for header */
+            break;
+            /* maximum stored block length that will fit in avail_out: */
+        have = s->strm->avail_out - have;
+        left = s->strstart - s->block_start;    /* bytes left in window */
+        if (len > (ulg)left + s->strm->avail_in)
+            len = left + s->strm->avail_in;     /* limit len to the input */
+        if (len > have)
+            len = have;                         /* limit len to the output */
+
+        /* If the stored block would be less than min_block in length, or if
+         * unable to copy all of the available input when flushing, then try
+         * copying to the window and the pending buffer instead. Also don't
+         * write an empty block when flushing -- deflate() does that.
+         */
+        if (len < min_block && ((len == 0 && flush != Z_FINISH) ||
+                                flush == Z_NO_FLUSH ||
+                                len != left + s->strm->avail_in))
+            break;
+
+        /* Make a dummy stored block in pending to get the header bytes,
+         * including any pending bits. This also updates the debugging counts.
+         */
+        last = flush == Z_FINISH && len == left + s->strm->avail_in ? 1 : 0;
+        _tr_stored_block(s, (char *)0, 0L, last);
+
+        /* Replace the lengths in the dummy stored block with len. */
+        s->pending_buf[s->pending - 4] = len;
+        s->pending_buf[s->pending - 3] = len >> 8;
+        s->pending_buf[s->pending - 2] = ~len;
+        s->pending_buf[s->pending - 1] = ~len >> 8;
+
+        /* Write the stored block header bytes. */
+        flush_pending(s->strm);
+
+#ifdef ZLIB_DEBUG
+        /* Update debugging counts for the data about to be copied. */
+        s->compressed_len += len << 3;
+        s->bits_sent += len << 3;
+#endif
+
+        /* Copy uncompressed bytes from the window to next_out. */
+        if (left) {
+            if (left > len)
+                left = len;
+            zmemcpy(s->strm->next_out, s->window + s->block_start, left);
+            s->strm->next_out += left;
+            s->strm->avail_out -= left;
+            s->strm->total_out += left;
+            s->block_start += left;
+            len -= left;
+        }
+
+        /* Copy uncompressed bytes directly from next_in to next_out, updating
+         * the check value.
+         */
+        if (len) {
+            read_buf(s->strm, s->strm->next_out, len);
+            s->strm->next_out += len;
+            s->strm->avail_out -= len;
+            s->strm->total_out += len;
+        }
+    } while (last == 0);
+
+    /* Update the sliding window with the last s->w_size bytes of the copied
+     * data, or append all of the copied data to the existing window if less
+     * than s->w_size bytes were copied. Also update the number of bytes to
+     * insert in the hash tables, in the event that deflateParams() switches to
+     * a non-zero compression level.
+     */
+    used -= s->strm->avail_in;      /* number of input bytes directly copied */
+    if (used) {
+        /* If any input was used, then no unused input remains in the window,
+         * therefore s->block_start == s->strstart.
+         */
+        if (used >= s->w_size) {    /* supplant the previous history */
+            s->matches = 2;         /* clear hash */
+            zmemcpy(s->window, s->strm->next_in - s->w_size, s->w_size);
+            s->strstart = s->w_size;
+        }
+        else {
+            if (s->window_size - s->strstart <= used) {
+                /* Slide the window down. */
+                s->strstart -= s->w_size;
+                zmemcpy(s->window, s->window + s->w_size, s->strstart);
+                if (s->matches < 2)
+                    s->matches++;   /* add a pending slide_hash() */
+            }
+            zmemcpy(s->window + s->strstart, s->strm->next_in - used, used);
+            s->strstart += used;
+        }
+        s->block_start = s->strstart;
+        s->insert += MIN(used, s->w_size - s->insert);
+    }
+    if (s->high_water < s->strstart)
+        s->high_water = s->strstart;
+
+    /* If the last block was written to next_out, then done. */
+    if (last)
+        return finish_done;
+
+    /* If flushing and all input has been consumed, then done. */
+    if (flush != Z_NO_FLUSH && flush != Z_FINISH &&
+        s->strm->avail_in == 0 && (long)s->strstart == s->block_start)
+        return block_done;
+
+    /* Fill the window with any remaining input. */
+    have = s->window_size - s->strstart - 1;
+    if (s->strm->avail_in > have && s->block_start >= (long)s->w_size) {
+        /* Slide the window down. */
+        s->block_start -= s->w_size;
+        s->strstart -= s->w_size;
+        zmemcpy(s->window, s->window + s->w_size, s->strstart);
+        if (s->matches < 2)
+            s->matches++;           /* add a pending slide_hash() */
+        have += s->w_size;          /* more space now */
+    }
+    if (have > s->strm->avail_in)
+        have = s->strm->avail_in;
+    if (have) {
+        read_buf(s->strm, s->window + s->strstart, have);
+        s->strstart += have;
+    }
+    if (s->high_water < s->strstart)
+        s->high_water = s->strstart;
+
+    /* There was not enough avail_out to write a complete worthy or flushed
+     * stored block to next_out. Write a stored block to pending instead, if we
+     * have enough input for a worthy block, or if flushing and there is enough
+     * room for the remaining input as a stored block in the pending buffer.
+     */
+    have = (s->bi_valid + 42) >> 3;         /* number of header bytes */
+        /* maximum stored block length that will fit in pending: */
+    have = MIN(s->pending_buf_size - have, MAX_STORED);
+    min_block = MIN(have, s->w_size);
+    left = s->strstart - s->block_start;
+    if (left >= min_block ||
+        ((left || flush == Z_FINISH) && flush != Z_NO_FLUSH &&
+         s->strm->avail_in == 0 && left <= have)) {
+        len = MIN(left, have);
+        last = flush == Z_FINISH && s->strm->avail_in == 0 &&
+               len == left ? 1 : 0;
+        _tr_stored_block(s, (charf *)s->window + s->block_start, len, last);
+        s->block_start += len;
+        flush_pending(s->strm);
+    }
+
+    /* We've done all we can with the available input and output. */
+    return last ? finish_started : need_more;
+}
+
+/* ===========================================================================
+ * Compress as much as possible from the input stream, return the current
+ * block state.
+ * This function does not perform lazy evaluation of matches and inserts
+ * new strings in the dictionary only for unmatched strings or for short
+ * matches. It is used only for the fast compression options.
+ */
+local block_state deflate_fast(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head;       /* head of the hash chain */
+    int bflush;           /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        hash_head = NIL;
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         * At this point we have always match_length < MIN_MATCH
+         */
+        if (hash_head != NIL && s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
+        }
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->match_start, s->match_length);
+
+            _tr_tally_dist(s, s->strstart - s->match_start,
+                           s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+
+            /* Insert new strings in the hash table only if the match length
+             * is not too large. This saves time but degrades compression.
+             */
+#ifndef FASTEST
+            if (s->match_length <= s->max_insert_length &&
+                s->lookahead >= MIN_MATCH) {
+                s->match_length--; /* string at strstart already in table */
+                do {
+                    s->strstart++;
+                    INSERT_STRING(s, s->strstart, hash_head);
+                    /* strstart never exceeds WSIZE-MAX_MATCH, so there are
+                     * always MIN_MATCH bytes ahead.
+                     */
+                } while (--s->match_length != 0);
+                s->strstart++;
+            } else
+#endif
+            {
+                s->strstart += s->match_length;
+                s->match_length = 0;
+                s->ins_h = s->window[s->strstart];
+                UPDATE_HASH(s, s->ins_h, s->window[s->strstart+1]);
+#if MIN_MATCH != 3
+                Call UPDATE_HASH() MIN_MATCH-3 more times
+#endif
+                /* If lookahead < MIN_MATCH, ins_h is garbage, but it does not
+                 * matter since it will be recomputed at next deflate call.
+                 */
+            }
+        } else {
+            /* No match, output a literal byte */
+            Tracevv((stderr,"%c", s->window[s->strstart]));
+            _tr_tally_lit (s, s->window[s->strstart], bflush);
+            s->lookahead--;
+            s->strstart++;
+        }
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+
+#ifndef FASTEST
+/* ===========================================================================
+ * Same as above, but achieves better compression. We use a lazy
+ * evaluation for matches: a match is finally adopted only if there is
+ * no better match at the next window position.
+ */
+local block_state deflate_slow(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    IPos hash_head;          /* head of hash chain */
+    int bflush;              /* set if current block must be flushed */
+
+    /* Process the input block. */
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the next match, plus MIN_MATCH bytes to insert the
+         * string following the next match.
+         */
+        if (s->lookahead < MIN_LOOKAHEAD) {
+            fill_window(s);
+            if (s->lookahead < MIN_LOOKAHEAD && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* Insert the string window[strstart .. strstart+2] in the
+         * dictionary, and set hash_head to the head of the hash chain:
+         */
+        hash_head = NIL;
+        if (s->lookahead >= MIN_MATCH) {
+            INSERT_STRING(s, s->strstart, hash_head);
+        }
+
+        /* Find the longest match, discarding those <= prev_length.
+         */
+        s->prev_length = s->match_length, s->prev_match = s->match_start;
+        s->match_length = MIN_MATCH-1;
+
+        if (hash_head != NIL && s->prev_length < s->max_lazy_match &&
+            s->strstart - hash_head <= MAX_DIST(s)) {
+            /* To simplify the code, we prevent matches with the string
+             * of window index 0 (in particular we have to avoid a match
+             * of the string with itself at the start of the input file).
+             */
+            s->match_length = longest_match (s, hash_head);
+            /* longest_match() sets match_start */
+
+            if (s->match_length <= 5 && (s->strategy == Z_FILTERED
+#if TOO_FAR <= 32767
+                || (s->match_length == MIN_MATCH &&
+                    s->strstart - s->match_start > TOO_FAR)
+#endif
+                )) {
+
+                /* If prev_match is also MIN_MATCH, match_start is garbage
+                 * but we will ignore the current match anyway.
+                 */
+                s->match_length = MIN_MATCH-1;
+            }
+        }
+        /* If there was a match at the previous step and the current
+         * match is not better, output the previous match:
+         */
+        if (s->prev_length >= MIN_MATCH && s->match_length <= s->prev_length) {
+            uInt max_insert = s->strstart + s->lookahead - MIN_MATCH;
+            /* Do not insert strings in hash table beyond this. */
+
+            check_match(s, s->strstart-1, s->prev_match, s->prev_length);
+
+            _tr_tally_dist(s, s->strstart -1 - s->prev_match,
+                           s->prev_length - MIN_MATCH, bflush);
+
+            /* Insert in hash table all strings up to the end of the match.
+             * strstart-1 and strstart are already inserted. If there is not
+             * enough lookahead, the last two strings are not inserted in
+             * the hash table.
+             */
+            s->lookahead -= s->prev_length-1;
+            s->prev_length -= 2;
+            do {
+                if (++s->strstart <= max_insert) {
+                    INSERT_STRING(s, s->strstart, hash_head);
+                }
+            } while (--s->prev_length != 0);
+            s->match_available = 0;
+            s->match_length = MIN_MATCH-1;
+            s->strstart++;
+
+            if (bflush) FLUSH_BLOCK(s, 0);
+
+        } else if (s->match_available) {
+            /* If there was no match at the previous position, output a
+             * single literal. If there was a match but the current match
+             * is longer, truncate the previous match to a single literal.
+             */
+            Tracevv((stderr,"%c", s->window[s->strstart-1]));
+            _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+            if (bflush) {
+                FLUSH_BLOCK_ONLY(s, 0);
+            }
+            s->strstart++;
+            s->lookahead--;
+            if (s->strm->avail_out == 0) return need_more;
+        } else {
+            /* There is no previous match to compare with, wait for
+             * the next step to decide.
+             */
+            s->match_available = 1;
+            s->strstart++;
+            s->lookahead--;
+        }
+    }
+    Assert (flush != Z_NO_FLUSH, "no flush?");
+    if (s->match_available) {
+        Tracevv((stderr,"%c", s->window[s->strstart-1]));
+        _tr_tally_lit(s, s->window[s->strstart-1], bflush);
+        s->match_available = 0;
+    }
+    s->insert = s->strstart < MIN_MATCH-1 ? s->strstart : MIN_MATCH-1;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+#endif /* FASTEST */
+
+/* ===========================================================================
+ * For Z_RLE, simply look for runs of bytes, generate matches only of distance
+ * one.  Do not maintain a hash table.  (It will be regenerated if this run of
+ * deflate switches away from Z_RLE.)
+ */
+local block_state deflate_rle(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    int bflush;             /* set if current block must be flushed */
+    uInt prev;              /* byte at distance one to match */
+    Bytef *scan, *strend;   /* scan goes up to strend for length of run */
+
+    for (;;) {
+        /* Make sure that we always have enough lookahead, except
+         * at the end of the input file. We need MAX_MATCH bytes
+         * for the longest run, plus one for the unrolled loop.
+         */
+        if (s->lookahead <= MAX_MATCH) {
+            fill_window(s);
+            if (s->lookahead <= MAX_MATCH && flush == Z_NO_FLUSH) {
+                return need_more;
+            }
+            if (s->lookahead == 0) break; /* flush the current block */
+        }
+
+        /* See how many times the previous byte repeats */
+        s->match_length = 0;
+        if (s->lookahead >= MIN_MATCH && s->strstart > 0) {
+            scan = s->window + s->strstart - 1;
+            prev = *scan;
+            if (prev == *++scan && prev == *++scan && prev == *++scan) {
+                strend = s->window + s->strstart + MAX_MATCH;
+                do {
+                } while (prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         prev == *++scan && prev == *++scan &&
+                         scan < strend);
+                s->match_length = MAX_MATCH - (uInt)(strend - scan);
+                if (s->match_length > s->lookahead)
+                    s->match_length = s->lookahead;
+            }
+            Assert(scan <= s->window+(uInt)(s->window_size-1), "wild scan");
+        }
+
+        /* Emit match if have run of MIN_MATCH or longer, else emit literal */
+        if (s->match_length >= MIN_MATCH) {
+            check_match(s, s->strstart, s->strstart - 1, s->match_length);
+
+            _tr_tally_dist(s, 1, s->match_length - MIN_MATCH, bflush);
+
+            s->lookahead -= s->match_length;
+            s->strstart += s->match_length;
+            s->match_length = 0;
+        } else {
+            /* No match, output a literal byte */
+            Tracevv((stderr,"%c", s->window[s->strstart]));
+            _tr_tally_lit (s, s->window[s->strstart], bflush);
+            s->lookahead--;
+            s->strstart++;
+        }
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
+
+/* ===========================================================================
+ * For Z_HUFFMAN_ONLY, do not look for matches.  Do not maintain a hash table.
+ * (It will be regenerated if this run of deflate switches away from Huffman.)
+ */
+local block_state deflate_huff(s, flush)
+    deflate_state *s;
+    int flush;
+{
+    int bflush;             /* set if current block must be flushed */
+
+    for (;;) {
+        /* Make sure that we have a literal to write. */
+        if (s->lookahead == 0) {
+            fill_window(s);
+            if (s->lookahead == 0) {
+                if (flush == Z_NO_FLUSH)
+                    return need_more;
+                break;      /* flush the current block */
+            }
+        }
+
+        /* Output a literal byte */
+        s->match_length = 0;
+        Tracevv((stderr,"%c", s->window[s->strstart]));
+        _tr_tally_lit (s, s->window[s->strstart], bflush);
+        s->lookahead--;
+        s->strstart++;
+        if (bflush) FLUSH_BLOCK(s, 0);
+    }
+    s->insert = 0;
+    if (flush == Z_FINISH) {
+        FLUSH_BLOCK(s, 1);
+        return finish_done;
+    }
+    if (s->last_lit)
+        FLUSH_BLOCK(s, 0);
+    return block_done;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/deflate.h glmark2-2021.02/src/zlib/deflate.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/deflate.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/deflate.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,349 @@
+/* deflate.h -- internal compression state
+ * Copyright (C) 1995-2016 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id$ */
+
+#ifndef DEFLATE_H
+#define DEFLATE_H
+
+#include "zutil.h"
+
+/* define NO_GZIP when compiling if you want to disable gzip header and
+   trailer creation by deflate().  NO_GZIP would be used to avoid linking in
+   the crc code when it is not needed.  For shared libraries, gzip encoding
+   should be left enabled. */
+#ifndef NO_GZIP
+#  define GZIP
+#endif
+
+/* ===========================================================================
+ * Internal compression state.
+ */
+
+#define LENGTH_CODES 29
+/* number of length codes, not counting the special END_BLOCK code */
+
+#define LITERALS  256
+/* number of literal bytes 0..255 */
+
+#define L_CODES (LITERALS+1+LENGTH_CODES)
+/* number of Literal or Length codes, including the END_BLOCK code */
+
+#define D_CODES   30
+/* number of distance codes */
+
+#define BL_CODES  19
+/* number of codes used to transfer the bit lengths */
+
+#define HEAP_SIZE (2*L_CODES+1)
+/* maximum heap size */
+
+#define MAX_BITS 15
+/* All codes must not exceed MAX_BITS bits */
+
+#define Buf_size 16
+/* size of bit buffer in bi_buf */
+
+#define INIT_STATE    42    /* zlib header -> BUSY_STATE */
+#ifdef GZIP
+#  define GZIP_STATE  57    /* gzip header -> BUSY_STATE | EXTRA_STATE */
+#endif
+#define EXTRA_STATE   69    /* gzip extra block -> NAME_STATE */
+#define NAME_STATE    73    /* gzip file name -> COMMENT_STATE */
+#define COMMENT_STATE 91    /* gzip comment -> HCRC_STATE */
+#define HCRC_STATE   103    /* gzip header CRC -> BUSY_STATE */
+#define BUSY_STATE   113    /* deflate -> FINISH_STATE */
+#define FINISH_STATE 666    /* stream complete */
+/* Stream status */
+
+
+/* Data structure describing a single value and its code string. */
+typedef struct ct_data_s {
+    union {
+        ush  freq;       /* frequency count */
+        ush  code;       /* bit string */
+    } fc;
+    union {
+        ush  dad;        /* father node in Huffman tree */
+        ush  len;        /* length of bit string */
+    } dl;
+} FAR ct_data;
+
+#define Freq fc.freq
+#define Code fc.code
+#define Dad  dl.dad
+#define Len  dl.len
+
+typedef struct static_tree_desc_s  static_tree_desc;
+
+typedef struct tree_desc_s {
+    ct_data *dyn_tree;           /* the dynamic tree */
+    int     max_code;            /* largest code with non zero frequency */
+    const static_tree_desc *stat_desc;  /* the corresponding static tree */
+} FAR tree_desc;
+
+typedef ush Pos;
+typedef Pos FAR Posf;
+typedef unsigned IPos;
+
+/* A Pos is an index in the character window. We use short instead of int to
+ * save space in the various tables. IPos is used only for parameter passing.
+ */
+
+typedef struct internal_state {
+    z_streamp strm;      /* pointer back to this zlib stream */
+    int   status;        /* as the name implies */
+    Bytef *pending_buf;  /* output still pending */
+    ulg   pending_buf_size; /* size of pending_buf */
+    Bytef *pending_out;  /* next pending byte to output to the stream */
+    ulg   pending;       /* nb of bytes in the pending buffer */
+    int   wrap;          /* bit 0 true for zlib, bit 1 true for gzip */
+    gz_headerp  gzhead;  /* gzip header information to write */
+    ulg   gzindex;       /* where in extra, name, or comment */
+    Byte  method;        /* can only be DEFLATED */
+    int   last_flush;    /* value of flush param for previous deflate call */
+
+                /* used by deflate.c: */
+
+    uInt  w_size;        /* LZ77 window size (32K by default) */
+    uInt  w_bits;        /* log2(w_size)  (8..16) */
+    uInt  w_mask;        /* w_size - 1 */
+
+    Bytef *window;
+    /* Sliding window. Input bytes are read into the second half of the window,
+     * and move to the first half later to keep a dictionary of at least wSize
+     * bytes. With this organization, matches are limited to a distance of
+     * wSize-MAX_MATCH bytes, but this ensures that IO is always
+     * performed with a length multiple of the block size. Also, it limits
+     * the window size to 64K, which is quite useful on MSDOS.
+     * To do: use the user input buffer as sliding window.
+     */
+
+    ulg window_size;
+    /* Actual size of window: 2*wSize, except when the user input buffer
+     * is directly used as sliding window.
+     */
+
+    Posf *prev;
+    /* Link to older string with same hash index. To limit the size of this
+     * array to 64K, this link is maintained only for the last 32K strings.
+     * An index in this array is thus a window index modulo 32K.
+     */
+
+    Posf *head; /* Heads of the hash chains or NIL. */
+
+    uInt  ins_h;          /* hash index of string to be inserted */
+    uInt  hash_size;      /* number of elements in hash table */
+    uInt  hash_bits;      /* log2(hash_size) */
+    uInt  hash_mask;      /* hash_size-1 */
+
+    uInt  hash_shift;
+    /* Number of bits by which ins_h must be shifted at each input
+     * step. It must be such that after MIN_MATCH steps, the oldest
+     * byte no longer takes part in the hash key, that is:
+     *   hash_shift * MIN_MATCH >= hash_bits
+     */
+
+    long block_start;
+    /* Window position at the beginning of the current output block. Gets
+     * negative when the window is moved backwards.
+     */
+
+    uInt match_length;           /* length of best match */
+    IPos prev_match;             /* previous match */
+    int match_available;         /* set if previous match exists */
+    uInt strstart;               /* start of string to insert */
+    uInt match_start;            /* start of matching string */
+    uInt lookahead;              /* number of valid bytes ahead in window */
+
+    uInt prev_length;
+    /* Length of the best match at previous step. Matches not greater than this
+     * are discarded. This is used in the lazy match evaluation.
+     */
+
+    uInt max_chain_length;
+    /* To speed up deflation, hash chains are never searched beyond this
+     * length.  A higher limit improves compression ratio but degrades the
+     * speed.
+     */
+
+    uInt max_lazy_match;
+    /* Attempt to find a better match only when the current match is strictly
+     * smaller than this value. This mechanism is used only for compression
+     * levels >= 4.
+     */
+#   define max_insert_length  max_lazy_match
+    /* Insert new strings in the hash table only if the match length is not
+     * greater than this length. This saves time but degrades compression.
+     * max_insert_length is used only for compression levels <= 3.
+     */
+
+    int level;    /* compression level (1..9) */
+    int strategy; /* favor or force Huffman coding*/
+
+    uInt good_match;
+    /* Use a faster search when the previous match is longer than this */
+
+    int nice_match; /* Stop searching when current match exceeds this */
+
+                /* used by trees.c: */
+    /* Didn't use ct_data typedef below to suppress compiler warning */
+    struct ct_data_s dyn_ltree[HEAP_SIZE];   /* literal and length tree */
+    struct ct_data_s dyn_dtree[2*D_CODES+1]; /* distance tree */
+    struct ct_data_s bl_tree[2*BL_CODES+1];  /* Huffman tree for bit lengths */
+
+    struct tree_desc_s l_desc;               /* desc. for literal tree */
+    struct tree_desc_s d_desc;               /* desc. for distance tree */
+    struct tree_desc_s bl_desc;              /* desc. for bit length tree */
+
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    int heap[2*L_CODES+1];      /* heap used to build the Huffman trees */
+    int heap_len;               /* number of elements in the heap */
+    int heap_max;               /* element of largest frequency */
+    /* The sons of heap[n] are heap[2*n] and heap[2*n+1]. heap[0] is not used.
+     * The same heap array is used to build all trees.
+     */
+
+    uch depth[2*L_CODES+1];
+    /* Depth of each subtree used as tie breaker for trees of equal frequency
+     */
+
+    uchf *l_buf;          /* buffer for literals or lengths */
+
+    uInt  lit_bufsize;
+    /* Size of match buffer for literals/lengths.  There are 4 reasons for
+     * limiting lit_bufsize to 64K:
+     *   - frequencies can be kept in 16 bit counters
+     *   - if compression is not successful for the first block, all input
+     *     data is still in the window so we can still emit a stored block even
+     *     when input comes from standard input.  (This can also be done for
+     *     all blocks if lit_bufsize is not greater than 32K.)
+     *   - if compression is not successful for a file smaller than 64K, we can
+     *     even emit a stored file instead of a stored block (saving 5 bytes).
+     *     This is applicable only for zip (not gzip or zlib).
+     *   - creating new Huffman trees less frequently may not provide fast
+     *     adaptation to changes in the input data statistics. (Take for
+     *     example a binary file with poorly compressible code followed by
+     *     a highly compressible string table.) Smaller buffer sizes give
+     *     fast adaptation but have of course the overhead of transmitting
+     *     trees more frequently.
+     *   - I can't count above 4
+     */
+
+    uInt last_lit;      /* running index in l_buf */
+
+    ushf *d_buf;
+    /* Buffer for distances. To simplify the code, d_buf and l_buf have
+     * the same number of elements. To use different lengths, an extra flag
+     * array would be necessary.
+     */
+
+    ulg opt_len;        /* bit length of current block with optimal trees */
+    ulg static_len;     /* bit length of current block with static trees */
+    uInt matches;       /* number of string matches in current block */
+    uInt insert;        /* bytes at end of window left to insert */
+
+#ifdef ZLIB_DEBUG
+    ulg compressed_len; /* total bit length of compressed file mod 2^32 */
+    ulg bits_sent;      /* bit length of compressed data sent mod 2^32 */
+#endif
+
+    ush bi_buf;
+    /* Output buffer. bits are inserted starting at the bottom (least
+     * significant bits).
+     */
+    int bi_valid;
+    /* Number of valid bits in bi_buf.  All bits above the last valid bit
+     * are always zero.
+     */
+
+    ulg high_water;
+    /* High water mark offset in window for initialized bytes -- bytes above
+     * this are set to zero in order to avoid memory check warnings when
+     * longest match routines access bytes past the input.  This is then
+     * updated to the new high water mark.
+     */
+
+} FAR deflate_state;
+
+/* Output a byte on the stream.
+ * IN assertion: there is enough room in pending_buf.
+ */
+#define put_byte(s, c) {s->pending_buf[s->pending++] = (Bytef)(c);}
+
+
+#define MIN_LOOKAHEAD (MAX_MATCH+MIN_MATCH+1)
+/* Minimum amount of lookahead, except at the end of the input file.
+ * See deflate.c for comments about the MIN_MATCH+1.
+ */
+
+#define MAX_DIST(s)  ((s)->w_size-MIN_LOOKAHEAD)
+/* In order to simplify the code, particularly on 16 bit machines, match
+ * distances are limited to MAX_DIST instead of WSIZE.
+ */
+
+#define WIN_INIT MAX_MATCH
+/* Number of bytes after end of data in window to initialize in order to avoid
+   memory checker errors from longest match routines */
+
+        /* in trees.c */
+void ZLIB_INTERNAL _tr_init OF((deflate_state *s));
+int ZLIB_INTERNAL _tr_tally OF((deflate_state *s, unsigned dist, unsigned lc));
+void ZLIB_INTERNAL _tr_flush_block OF((deflate_state *s, charf *buf,
+                        ulg stored_len, int last));
+void ZLIB_INTERNAL _tr_flush_bits OF((deflate_state *s));
+void ZLIB_INTERNAL _tr_align OF((deflate_state *s));
+void ZLIB_INTERNAL _tr_stored_block OF((deflate_state *s, charf *buf,
+                        ulg stored_len, int last));
+
+#define d_code(dist) \
+   ((dist) < 256 ? _dist_code[dist] : _dist_code[256+((dist)>>7)])
+/* Mapping from a distance to a distance code. dist is the distance - 1 and
+ * must not have side effects. _dist_code[256] and _dist_code[257] are never
+ * used.
+ */
+
+#ifndef ZLIB_DEBUG
+/* Inline versions of _tr_tally for speed: */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+  extern uch ZLIB_INTERNAL _length_code[];
+  extern uch ZLIB_INTERNAL _dist_code[];
+#else
+  extern const uch ZLIB_INTERNAL _length_code[];
+  extern const uch ZLIB_INTERNAL _dist_code[];
+#endif
+
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->d_buf[s->last_lit] = 0; \
+    s->l_buf[s->last_lit++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (uch)(length); \
+    ush dist = (ush)(distance); \
+    s->d_buf[s->last_lit] = dist; \
+    s->l_buf[s->last_lit++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->last_lit == s->lit_bufsize-1); \
+  }
+#else
+# define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
+# define _tr_tally_dist(s, distance, length, flush) \
+              flush = _tr_tally(s, distance, length)
+#endif
+
+#endif /* DEFLATE_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzclose.c glmark2-2021.02/src/zlib/gzclose.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzclose.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/gzclose.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,25 @@
+/* gzclose.c -- zlib gzclose() function
+ * Copyright (C) 2004, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* gzclose() is in a separate file so that it is linked in only if it is used.
+   That way the other gzclose functions can be used instead to avoid linking in
+   unneeded compression or decompression routines. */
+int ZEXPORT gzclose(file)
+    gzFile file;
+{
+#ifndef NO_GZCOMPRESS
+    gz_statep state;
+
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    return state->mode == GZ_READ ? gzclose_r(file) : gzclose_w(file);
+#else
+    return gzclose_r(file);
+#endif
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzguts.h glmark2-2021.02/src/zlib/gzguts.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzguts.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/gzguts.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,218 @@
+/* gzguts.h -- zlib internal header definitions for gz* operations
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#ifdef _LARGEFILE64_SOURCE
+#  ifndef _LARGEFILE_SOURCE
+#    define _LARGEFILE_SOURCE 1
+#  endif
+#  ifdef _FILE_OFFSET_BITS
+#    undef _FILE_OFFSET_BITS
+#  endif
+#endif
+
+#ifdef HAVE_HIDDEN
+#  define ZLIB_INTERNAL __attribute__((visibility ("hidden")))
+#else
+#  define ZLIB_INTERNAL
+#endif
+
+#include <stdio.h>
+#include "zlib.h"
+#ifdef STDC
+#  include <string.h>
+#  include <stdlib.h>
+#  include <limits.h>
+#endif
+
+#ifndef _POSIX_SOURCE
+#  define _POSIX_SOURCE
+#endif
+#include <fcntl.h>
+
+#ifdef _WIN32
+#  include <stddef.h>
+#endif
+
+#if defined(__TURBOC__) || defined(_MSC_VER) || defined(_WIN32)
+#  include <io.h>
+#endif
+
+#if defined(_WIN32) || defined(__CYGWIN__)
+#  define WIDECHAR
+#endif
+
+#ifdef WINAPI_FAMILY
+#  define open _open
+#  define read _read
+#  define write _write
+#  define close _close
+#endif
+
+#ifdef NO_DEFLATE       /* for compatibility with old definition */
+#  define NO_GZCOMPRESS
+#endif
+
+#if defined(STDC99) || (defined(__TURBOC__) && __TURBOC__ >= 0x550)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#if defined(__CYGWIN__)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#if defined(MSDOS) && defined(__BORLANDC__) && (BORLANDC > 0x410)
+#  ifndef HAVE_VSNPRINTF
+#    define HAVE_VSNPRINTF
+#  endif
+#endif
+
+#ifndef HAVE_VSNPRINTF
+#  ifdef MSDOS
+/* vsnprintf may exist on some MS-DOS compilers (DJGPP?),
+   but for now we just assume it doesn't. */
+#    define NO_vsnprintf
+#  endif
+#  ifdef __TURBOC__
+#    define NO_vsnprintf
+#  endif
+#  ifdef WIN32
+/* In Win32, vsnprintf is available as the "non-ANSI" _vsnprintf. */
+#    if !defined(vsnprintf) && !defined(NO_vsnprintf)
+#      if !defined(_MSC_VER) || ( defined(_MSC_VER) && _MSC_VER < 1500 )
+#         define vsnprintf _vsnprintf
+#      endif
+#    endif
+#  endif
+#  ifdef __SASC
+#    define NO_vsnprintf
+#  endif
+#  ifdef VMS
+#    define NO_vsnprintf
+#  endif
+#  ifdef __OS400__
+#    define NO_vsnprintf
+#  endif
+#  ifdef __MVS__
+#    define NO_vsnprintf
+#  endif
+#endif
+
+/* unlike snprintf (which is required in C99), _snprintf does not guarantee
+   null termination of the result -- however this is only used in gzlib.c where
+   the result is assured to fit in the space provided */
+#if defined(_MSC_VER) && _MSC_VER < 1900
+#  define snprintf _snprintf
+#endif
+
+#ifndef local
+#  define local static
+#endif
+/* since "static" is used to mean two completely different things in C, we
+   define "local" for the non-static meaning of "static", for readability
+   (compile with -Dlocal if your debugger can't find static symbols) */
+
+/* gz* functions always use library allocation functions */
+#ifndef STDC
+  extern voidp  malloc OF((uInt size));
+  extern void   free   OF((voidpf ptr));
+#endif
+
+/* get errno and strerror definition */
+#if defined UNDER_CE
+#  include <windows.h>
+#  define zstrerror() gz_strwinerror((DWORD)GetLastError())
+#else
+#  ifndef NO_STRERROR
+#    include <errno.h>
+#    define zstrerror() strerror(errno)
+#  else
+#    define zstrerror() "stdio error (consult errno)"
+#  endif
+#endif
+
+/* provide prototypes for these when building zlib without LFS */
+#if !defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0
+    ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+    ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int));
+    ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile));
+    ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile));
+#endif
+
+/* default memLevel */
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+
+/* default i/o buffer size -- double this for output when reading (this and
+   twice this must be able to fit in an unsigned type) */
+#define GZBUFSIZE 8192
+
+/* gzip modes, also provide a little integrity check on the passed structure */
+#define GZ_NONE 0
+#define GZ_READ 7247
+#define GZ_WRITE 31153
+#define GZ_APPEND 1     /* mode set to GZ_WRITE after the file is opened */
+
+/* values for gz_state how */
+#define LOOK 0      /* look for a gzip header */
+#define COPY 1      /* copy input directly */
+#define GZIP 2      /* decompress a gzip stream */
+
+/* internal gzip file state data structure */
+typedef struct {
+        /* exposed contents for gzgetc() macro */
+    struct gzFile_s x;      /* "x" for exposed */
+                            /* x.have: number of bytes available at x.next */
+                            /* x.next: next output data to deliver or write */
+                            /* x.pos: current position in uncompressed data */
+        /* used for both reading and writing */
+    int mode;               /* see gzip modes above */
+    int fd;                 /* file descriptor */
+    char *path;             /* path or fd for error messages */
+    unsigned size;          /* buffer size, zero if not allocated yet */
+    unsigned want;          /* requested buffer size, default is GZBUFSIZE */
+    unsigned char *in;      /* input buffer (double-sized when writing) */
+    unsigned char *out;     /* output buffer (double-sized when reading) */
+    int direct;             /* 0 if processing gzip, 1 if transparent */
+        /* just for reading */
+    int how;                /* 0: get header, 1: copy, 2: decompress */
+    z_off64_t start;        /* where the gzip data started, for rewinding */
+    int eof;                /* true if end of input file reached */
+    int past;               /* true if read requested past end */
+        /* just for writing */
+    int level;              /* compression level */
+    int strategy;           /* compression strategy */
+        /* seek request */
+    z_off64_t skip;         /* amount to skip (already rewound if backwards) */
+    int seek;               /* true if seek request pending */
+        /* error information */
+    int err;                /* error code */
+    char *msg;              /* error message */
+        /* zlib inflate or deflate stream */
+    z_stream strm;          /* stream structure in-place (not a pointer) */
+} gz_state;
+typedef gz_state FAR *gz_statep;
+
+/* shared functions */
+void ZLIB_INTERNAL gz_error OF((gz_statep, int, const char *));
+#if defined UNDER_CE
+char ZLIB_INTERNAL *gz_strwinerror OF((DWORD error));
+#endif
+
+/* GT_OFF(x), where x is an unsigned value, is true if x > maximum z_off64_t
+   value -- needed when comparing unsigned to z_off64_t, which is signed
+   (possible z_off64_t types off_t, off64_t, and long are all signed) */
+#ifdef INT_MAX
+#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > INT_MAX)
+#else
+unsigned ZLIB_INTERNAL gz_intmax OF((void));
+#  define GT_OFF(x) (sizeof(int) == sizeof(z_off64_t) && (x) > gz_intmax())
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzlib.c glmark2-2021.02/src/zlib/gzlib.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzlib.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/gzlib.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,637 @@
+/* gzlib.c -- zlib functions common to reading and writing gzip files
+ * Copyright (C) 2004-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+#if defined(_WIN32) && !defined(__BORLANDC__) && !defined(__MINGW32__)
+#  define LSEEK _lseeki64
+#else
+#if defined(_LARGEFILE64_SOURCE) && _LFS64_LARGEFILE-0
+#  define LSEEK lseek64
+#else
+#  define LSEEK lseek
+#endif
+#endif
+
+/* Local functions */
+local void gz_reset OF((gz_statep));
+local gzFile gz_open OF((const void *, int, const char *));
+
+#if defined UNDER_CE
+
+/* Map the Windows error number in ERROR to a locale-dependent error message
+   string and return a pointer to it.  Typically, the values for ERROR come
+   from GetLastError.
+
+   The string pointed to shall not be modified by the application, but may be
+   overwritten by a subsequent call to gz_strwinerror
+
+   The gz_strwinerror function does not change the current setting of
+   GetLastError. */
+char ZLIB_INTERNAL *gz_strwinerror (error)
+     DWORD error;
+{
+    static char buf[1024];
+
+    wchar_t *msgbuf;
+    DWORD lasterr = GetLastError();
+    DWORD chars = FormatMessage(FORMAT_MESSAGE_FROM_SYSTEM
+        | FORMAT_MESSAGE_ALLOCATE_BUFFER,
+        NULL,
+        error,
+        0, /* Default language */
+        (LPVOID)&msgbuf,
+        0,
+        NULL);
+    if (chars != 0) {
+        /* If there is an \r\n appended, zap it.  */
+        if (chars >= 2
+            && msgbuf[chars - 2] == '\r' && msgbuf[chars - 1] == '\n') {
+            chars -= 2;
+            msgbuf[chars] = 0;
+        }
+
+        if (chars > sizeof (buf) - 1) {
+            chars = sizeof (buf) - 1;
+            msgbuf[chars] = 0;
+        }
+
+        wcstombs(buf, msgbuf, chars + 1);
+        LocalFree(msgbuf);
+    }
+    else {
+        sprintf(buf, "unknown win32 error (%ld)", error);
+    }
+
+    SetLastError(lasterr);
+    return buf;
+}
+
+#endif /* UNDER_CE */
+
+/* Reset gzip file state */
+local void gz_reset(state)
+    gz_statep state;
+{
+    state->x.have = 0;              /* no output data available */
+    if (state->mode == GZ_READ) {   /* for reading ... */
+        state->eof = 0;             /* not at end of file */
+        state->past = 0;            /* have not read past end yet */
+        state->how = LOOK;          /* look for gzip header */
+    }
+    state->seek = 0;                /* no seek request pending */
+    gz_error(state, Z_OK, NULL);    /* clear error */
+    state->x.pos = 0;               /* no uncompressed data yet */
+    state->strm.avail_in = 0;       /* no input data yet */
+}
+
+/* Open a gzip file either by name or file descriptor. */
+local gzFile gz_open(path, fd, mode)
+    const void *path;
+    int fd;
+    const char *mode;
+{
+    gz_statep state;
+    z_size_t len;
+    int oflag;
+#ifdef O_CLOEXEC
+    int cloexec = 0;
+#endif
+#ifdef O_EXCL
+    int exclusive = 0;
+#endif
+
+    /* check input */
+    if (path == NULL)
+        return NULL;
+
+    /* allocate gzFile structure to return */
+    state = (gz_statep)malloc(sizeof(gz_state));
+    if (state == NULL)
+        return NULL;
+    state->size = 0;            /* no buffers allocated yet */
+    state->want = GZBUFSIZE;    /* requested buffer size */
+    state->msg = NULL;          /* no error message yet */
+
+    /* interpret mode */
+    state->mode = GZ_NONE;
+    state->level = Z_DEFAULT_COMPRESSION;
+    state->strategy = Z_DEFAULT_STRATEGY;
+    state->direct = 0;
+    while (*mode) {
+        if (*mode >= '0' && *mode <= '9')
+            state->level = *mode - '0';
+        else
+            switch (*mode) {
+            case 'r':
+                state->mode = GZ_READ;
+                break;
+#ifndef NO_GZCOMPRESS
+            case 'w':
+                state->mode = GZ_WRITE;
+                break;
+            case 'a':
+                state->mode = GZ_APPEND;
+                break;
+#endif
+            case '+':       /* can't read and write at the same time */
+                free(state);
+                return NULL;
+            case 'b':       /* ignore -- will request binary anyway */
+                break;
+#ifdef O_CLOEXEC
+            case 'e':
+                cloexec = 1;
+                break;
+#endif
+#ifdef O_EXCL
+            case 'x':
+                exclusive = 1;
+                break;
+#endif
+            case 'f':
+                state->strategy = Z_FILTERED;
+                break;
+            case 'h':
+                state->strategy = Z_HUFFMAN_ONLY;
+                break;
+            case 'R':
+                state->strategy = Z_RLE;
+                break;
+            case 'F':
+                state->strategy = Z_FIXED;
+                break;
+            case 'T':
+                state->direct = 1;
+                break;
+            default:        /* could consider as an error, but just ignore */
+                ;
+            }
+        mode++;
+    }
+
+    /* must provide an "r", "w", or "a" */
+    if (state->mode == GZ_NONE) {
+        free(state);
+        return NULL;
+    }
+
+    /* can't force transparent read */
+    if (state->mode == GZ_READ) {
+        if (state->direct) {
+            free(state);
+            return NULL;
+        }
+        state->direct = 1;      /* for empty file */
+    }
+
+    /* save the path name for error messages */
+#ifdef WIDECHAR
+    if (fd == -2) {
+        len = wcstombs(NULL, path, 0);
+        if (len == (z_size_t)-1)
+            len = 0;
+    }
+    else
+#endif
+        len = strlen((const char *)path);
+    state->path = (char *)malloc(len + 1);
+    if (state->path == NULL) {
+        free(state);
+        return NULL;
+    }
+#ifdef WIDECHAR
+    if (fd == -2)
+        if (len)
+            wcstombs(state->path, path, len + 1);
+        else
+            *(state->path) = 0;
+    else
+#endif
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+        (void)snprintf(state->path, len + 1, "%s", (const char *)path);
+#else
+        strcpy(state->path, path);
+#endif
+
+    /* compute the flags for open() */
+    oflag =
+#ifdef O_LARGEFILE
+        O_LARGEFILE |
+#endif
+#ifdef O_BINARY
+        O_BINARY |
+#endif
+#ifdef O_CLOEXEC
+        (cloexec ? O_CLOEXEC : 0) |
+#endif
+        (state->mode == GZ_READ ?
+         O_RDONLY :
+         (O_WRONLY | O_CREAT |
+#ifdef O_EXCL
+          (exclusive ? O_EXCL : 0) |
+#endif
+          (state->mode == GZ_WRITE ?
+           O_TRUNC :
+           O_APPEND)));
+
+    /* open the file with the appropriate flags (or just use fd) */
+    state->fd = fd > -1 ? fd : (
+#ifdef WIDECHAR
+        fd == -2 ? _wopen(path, oflag, 0666) :
+#endif
+        open((const char *)path, oflag, 0666));
+    if (state->fd == -1) {
+        free(state->path);
+        free(state);
+        return NULL;
+    }
+    if (state->mode == GZ_APPEND) {
+        LSEEK(state->fd, 0, SEEK_END);  /* so gzoffset() is correct */
+        state->mode = GZ_WRITE;         /* simplify later checks */
+    }
+
+    /* save the current position for rewinding (only if reading) */
+    if (state->mode == GZ_READ) {
+        state->start = LSEEK(state->fd, 0, SEEK_CUR);
+        if (state->start == -1) state->start = 0;
+    }
+
+    /* initialize stream */
+    gz_reset(state);
+
+    /* return stream */
+    return (gzFile)state;
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzopen(path, mode)
+    const char *path;
+    const char *mode;
+{
+    return gz_open(path, -1, mode);
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzopen64(path, mode)
+    const char *path;
+    const char *mode;
+{
+    return gz_open(path, -1, mode);
+}
+
+/* -- see zlib.h -- */
+gzFile ZEXPORT gzdopen(fd, mode)
+    int fd;
+    const char *mode;
+{
+    char *path;         /* identifier for error messages */
+    gzFile gz;
+
+    if (fd == -1 || (path = (char *)malloc(7 + 3 * sizeof(int))) == NULL)
+        return NULL;
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    (void)snprintf(path, 7 + 3 * sizeof(int), "<fd:%d>", fd);
+#else
+    sprintf(path, "<fd:%d>", fd);   /* for debugging */
+#endif
+    gz = gz_open(path, fd, mode);
+    free(path);
+    return gz;
+}
+
+/* -- see zlib.h -- */
+#ifdef WIDECHAR
+gzFile ZEXPORT gzopen_w(path, mode)
+    const wchar_t *path;
+    const char *mode;
+{
+    return gz_open(path, -2, mode);
+}
+#endif
+
+/* -- see zlib.h -- */
+int ZEXPORT gzbuffer(file, size)
+    gzFile file;
+    unsigned size;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* make sure we haven't already allocated memory */
+    if (state->size != 0)
+        return -1;
+
+    /* check and set requested size */
+    if ((size << 1) < size)
+        return -1;              /* need to be able to double it */
+    if (size < 2)
+        size = 2;               /* need two bytes to check magic header */
+    state->want = size;
+    return 0;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzrewind(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* back up and start over */
+    if (LSEEK(state->fd, state->start, SEEK_SET) == -1)
+        return -1;
+    gz_reset(state);
+    return 0;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gzseek64(file, offset, whence)
+    gzFile file;
+    z_off64_t offset;
+    int whence;
+{
+    unsigned n;
+    z_off64_t ret;
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* check that there's no error */
+    if (state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+
+    /* can only seek from start or relative to current position */
+    if (whence != SEEK_SET && whence != SEEK_CUR)
+        return -1;
+
+    /* normalize offset to a SEEK_CUR specification */
+    if (whence == SEEK_SET)
+        offset -= state->x.pos;
+    else if (state->seek)
+        offset += state->skip;
+    state->seek = 0;
+
+    /* if within raw area while reading, just go there */
+    if (state->mode == GZ_READ && state->how == COPY &&
+            state->x.pos + offset >= 0) {
+        ret = LSEEK(state->fd, offset - state->x.have, SEEK_CUR);
+        if (ret == -1)
+            return -1;
+        state->x.have = 0;
+        state->eof = 0;
+        state->past = 0;
+        state->seek = 0;
+        gz_error(state, Z_OK, NULL);
+        state->strm.avail_in = 0;
+        state->x.pos += offset;
+        return state->x.pos;
+    }
+
+    /* calculate skip amount, rewinding if needed for back seek when reading */
+    if (offset < 0) {
+        if (state->mode != GZ_READ)         /* writing -- can't go backwards */
+            return -1;
+        offset += state->x.pos;
+        if (offset < 0)                     /* before start of file! */
+            return -1;
+        if (gzrewind(file) == -1)           /* rewind, then skip to offset */
+            return -1;
+    }
+
+    /* if reading, skip what's in output buffer (one less gzgetc() check) */
+    if (state->mode == GZ_READ) {
+        n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > offset ?
+            (unsigned)offset : state->x.have;
+        state->x.have -= n;
+        state->x.next += n;
+        state->x.pos += n;
+        offset -= n;
+    }
+
+    /* request skip (if not zero) */
+    if (offset) {
+        state->seek = 1;
+        state->skip = offset;
+    }
+    return state->x.pos + offset;
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gzseek(file, offset, whence)
+    gzFile file;
+    z_off_t offset;
+    int whence;
+{
+    z_off64_t ret;
+
+    ret = gzseek64(file, (z_off64_t)offset, whence);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gztell64(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* return position */
+    return state->x.pos + (state->seek ? state->skip : 0);
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gztell(file)
+    gzFile file;
+{
+    z_off64_t ret;
+
+    ret = gztell64(file);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+z_off64_t ZEXPORT gzoffset64(file)
+    gzFile file;
+{
+    z_off64_t offset;
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return -1;
+
+    /* compute and return effective offset in file */
+    offset = LSEEK(state->fd, 0, SEEK_CUR);
+    if (offset == -1)
+        return -1;
+    if (state->mode == GZ_READ)             /* reading */
+        offset -= state->strm.avail_in;     /* don't count buffered input */
+    return offset;
+}
+
+/* -- see zlib.h -- */
+z_off_t ZEXPORT gzoffset(file)
+    gzFile file;
+{
+    z_off64_t ret;
+
+    ret = gzoffset64(file);
+    return ret == (z_off_t)ret ? (z_off_t)ret : -1;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzeof(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return 0;
+
+    /* return end-of-file state */
+    return state->mode == GZ_READ ? state->past : 0;
+}
+
+/* -- see zlib.h -- */
+const char * ZEXPORT gzerror(file, errnum)
+    gzFile file;
+    int *errnum;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return NULL;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return NULL;
+
+    /* return error information */
+    if (errnum != NULL)
+        *errnum = state->err;
+    return state->err == Z_MEM_ERROR ? "out of memory" :
+                                       (state->msg == NULL ? "" : state->msg);
+}
+
+/* -- see zlib.h -- */
+void ZEXPORT gzclearerr(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure and check integrity */
+    if (file == NULL)
+        return;
+    state = (gz_statep)file;
+    if (state->mode != GZ_READ && state->mode != GZ_WRITE)
+        return;
+
+    /* clear error and end-of-file */
+    if (state->mode == GZ_READ) {
+        state->eof = 0;
+        state->past = 0;
+    }
+    gz_error(state, Z_OK, NULL);
+}
+
+/* Create an error message in allocated memory and set state->err and
+   state->msg accordingly.  Free any previous error message already there.  Do
+   not try to free or allocate space if the error is Z_MEM_ERROR (out of
+   memory).  Simply save the error message as a static string.  If there is an
+   allocation failure constructing the error message, then convert the error to
+   out of memory. */
+void ZLIB_INTERNAL gz_error(state, err, msg)
+    gz_statep state;
+    int err;
+    const char *msg;
+{
+    /* free previously allocated message and clear */
+    if (state->msg != NULL) {
+        if (state->err != Z_MEM_ERROR)
+            free(state->msg);
+        state->msg = NULL;
+    }
+
+    /* if fatal, set state->x.have to 0 so that the gzgetc() macro fails */
+    if (err != Z_OK && err != Z_BUF_ERROR)
+        state->x.have = 0;
+
+    /* set error code, and if no message, then done */
+    state->err = err;
+    if (msg == NULL)
+        return;
+
+    /* for an out of memory error, return literal string when requested */
+    if (err == Z_MEM_ERROR)
+        return;
+
+    /* construct error message with path */
+    if ((state->msg = (char *)malloc(strlen(state->path) + strlen(msg) + 3)) ==
+            NULL) {
+        state->err = Z_MEM_ERROR;
+        return;
+    }
+#if !defined(NO_snprintf) && !defined(NO_vsnprintf)
+    (void)snprintf(state->msg, strlen(state->path) + strlen(msg) + 3,
+                   "%s%s%s", state->path, ": ", msg);
+#else
+    strcpy(state->msg, state->path);
+    strcat(state->msg, ": ");
+    strcat(state->msg, msg);
+#endif
+}
+
+#ifndef INT_MAX
+/* portably return maximum value for an int (when limits.h presumed not
+   available) -- we need to do this to cover cases where 2's complement not
+   used, since C standard permits 1's complement and sign-bit representations,
+   otherwise we could just use ((unsigned)-1) >> 1 */
+unsigned ZLIB_INTERNAL gz_intmax()
+{
+    unsigned p, q;
+
+    p = 1;
+    do {
+        q = p;
+        p <<= 1;
+        p++;
+    } while (p > q);
+    return q >> 1;
+}
+#endif
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzread.c glmark2-2021.02/src/zlib/gzread.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzread.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/gzread.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,654 @@
+/* gzread.c -- zlib functions for reading gzip files
+ * Copyright (C) 2004, 2005, 2010, 2011, 2012, 2013, 2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* Local functions */
+local int gz_load OF((gz_statep, unsigned char *, unsigned, unsigned *));
+local int gz_avail OF((gz_statep));
+local int gz_look OF((gz_statep));
+local int gz_decomp OF((gz_statep));
+local int gz_fetch OF((gz_statep));
+local int gz_skip OF((gz_statep, z_off64_t));
+local z_size_t gz_read OF((gz_statep, voidp, z_size_t));
+
+/* Use read() to load a buffer -- return -1 on error, otherwise 0.  Read from
+   state->fd, and update state->eof, state->err, and state->msg as appropriate.
+   This function needs to loop on read(), since read() is not guaranteed to
+   read the number of bytes requested, depending on the type of descriptor. */
+local int gz_load(state, buf, len, have)
+    gz_statep state;
+    unsigned char *buf;
+    unsigned len;
+    unsigned *have;
+{
+    int ret;
+    unsigned get, max = ((unsigned)-1 >> 2) + 1;
+
+    *have = 0;
+    do {
+        get = len - *have;
+        if (get > max)
+            get = max;
+        ret = read(state->fd, buf + *have, get);
+        if (ret <= 0)
+            break;
+        *have += (unsigned)ret;
+    } while (*have < len);
+    if (ret < 0) {
+        gz_error(state, Z_ERRNO, zstrerror());
+        return -1;
+    }
+    if (ret == 0)
+        state->eof = 1;
+    return 0;
+}
+
+/* Load up input buffer and set eof flag if last data loaded -- return -1 on
+   error, 0 otherwise.  Note that the eof flag is set when the end of the input
+   file is reached, even though there may be unused data in the buffer.  Once
+   that data has been used, no more attempts will be made to read the file.
+   If strm->avail_in != 0, then the current data is moved to the beginning of
+   the input buffer, and then the remainder of the buffer is loaded with the
+   available data from the input file. */
+local int gz_avail(state)
+    gz_statep state;
+{
+    unsigned got;
+    z_streamp strm = &(state->strm);
+
+    if (state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+    if (state->eof == 0) {
+        if (strm->avail_in) {       /* copy what's there to the start */
+            unsigned char *p = state->in;
+            unsigned const char *q = strm->next_in;
+            unsigned n = strm->avail_in;
+            do {
+                *p++ = *q++;
+            } while (--n);
+        }
+        if (gz_load(state, state->in + strm->avail_in,
+                    state->size - strm->avail_in, &got) == -1)
+            return -1;
+        strm->avail_in += got;
+        strm->next_in = state->in;
+    }
+    return 0;
+}
+
+/* Look for gzip header, set up for inflate or copy.  state->x.have must be 0.
+   If this is the first time in, allocate required memory.  state->how will be
+   left unchanged if there is no more input data available, will be set to COPY
+   if there is no gzip header and direct copying will be performed, or it will
+   be set to GZIP for decompression.  If direct copying, then leftover input
+   data from the input buffer will be copied to the output buffer.  In that
+   case, all further file reads will be directly to either the output buffer or
+   a user buffer.  If decompressing, the inflate state will be initialized.
+   gz_look() will return 0 on success or -1 on failure. */
+local int gz_look(state)
+    gz_statep state;
+{
+    z_streamp strm = &(state->strm);
+
+    /* allocate read buffers and inflate memory */
+    if (state->size == 0) {
+        /* allocate buffers */
+        state->in = (unsigned char *)malloc(state->want);
+        state->out = (unsigned char *)malloc(state->want << 1);
+        if (state->in == NULL || state->out == NULL) {
+            free(state->out);
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        state->size = state->want;
+
+        /* allocate inflate memory */
+        state->strm.zalloc = Z_NULL;
+        state->strm.zfree = Z_NULL;
+        state->strm.opaque = Z_NULL;
+        state->strm.avail_in = 0;
+        state->strm.next_in = Z_NULL;
+        if (inflateInit2(&(state->strm), 15 + 16) != Z_OK) {    /* gunzip */
+            free(state->out);
+            free(state->in);
+            state->size = 0;
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+    }
+
+    /* get at least the magic bytes in the input buffer */
+    if (strm->avail_in < 2) {
+        if (gz_avail(state) == -1)
+            return -1;
+        if (strm->avail_in == 0)
+            return 0;
+    }
+
+    /* look for gzip magic bytes -- if there, do gzip decoding (note: there is
+       a logical dilemma here when considering the case of a partially written
+       gzip file, to wit, if a single 31 byte is written, then we cannot tell
+       whether this is a single-byte file, or just a partially written gzip
+       file -- for here we assume that if a gzip file is being written, then
+       the header will be written in a single operation, so that reading a
+       single byte is sufficient indication that it is not a gzip file) */
+    if (strm->avail_in > 1 &&
+            strm->next_in[0] == 31 && strm->next_in[1] == 139) {
+        inflateReset(strm);
+        state->how = GZIP;
+        state->direct = 0;
+        return 0;
+    }
+
+    /* no gzip header -- if we were decoding gzip before, then this is trailing
+       garbage.  Ignore the trailing garbage and finish. */
+    if (state->direct == 0) {
+        strm->avail_in = 0;
+        state->eof = 1;
+        state->x.have = 0;
+        return 0;
+    }
+
+    /* doing raw i/o, copy any leftover input to output -- this assumes that
+       the output buffer is larger than the input buffer, which also assures
+       space for gzungetc() */
+    state->x.next = state->out;
+    if (strm->avail_in) {
+        memcpy(state->x.next, strm->next_in, strm->avail_in);
+        state->x.have = strm->avail_in;
+        strm->avail_in = 0;
+    }
+    state->how = COPY;
+    state->direct = 1;
+    return 0;
+}
+
+/* Decompress from input to the provided next_out and avail_out in the state.
+   On return, state->x.have and state->x.next point to the just decompressed
+   data.  If the gzip stream completes, state->how is reset to LOOK to look for
+   the next gzip stream or raw data, once state->x.have is depleted.  Returns 0
+   on success, -1 on failure. */
+local int gz_decomp(state)
+    gz_statep state;
+{
+    int ret = Z_OK;
+    unsigned had;
+    z_streamp strm = &(state->strm);
+
+    /* fill output buffer up to end of deflate stream */
+    had = strm->avail_out;
+    do {
+        /* get more input for inflate() */
+        if (strm->avail_in == 0 && gz_avail(state) == -1)
+            return -1;
+        if (strm->avail_in == 0) {
+            gz_error(state, Z_BUF_ERROR, "unexpected end of file");
+            break;
+        }
+
+        /* decompress and handle errors */
+        ret = inflate(strm, Z_NO_FLUSH);
+        if (ret == Z_STREAM_ERROR || ret == Z_NEED_DICT) {
+            gz_error(state, Z_STREAM_ERROR,
+                     "internal error: inflate stream corrupt");
+            return -1;
+        }
+        if (ret == Z_MEM_ERROR) {
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        if (ret == Z_DATA_ERROR) {              /* deflate stream invalid */
+            gz_error(state, Z_DATA_ERROR,
+                     strm->msg == NULL ? "compressed data error" : strm->msg);
+            return -1;
+        }
+    } while (strm->avail_out && ret != Z_STREAM_END);
+
+    /* update available output */
+    state->x.have = had - strm->avail_out;
+    state->x.next = strm->next_out - state->x.have;
+
+    /* if the gzip stream completed successfully, look for another */
+    if (ret == Z_STREAM_END)
+        state->how = LOOK;
+
+    /* good decompression */
+    return 0;
+}
+
+/* Fetch data and put it in the output buffer.  Assumes state->x.have is 0.
+   Data is either copied from the input file or decompressed from the input
+   file depending on state->how.  If state->how is LOOK, then a gzip header is
+   looked for to determine whether to copy or decompress.  Returns -1 on error,
+   otherwise 0.  gz_fetch() will leave state->how as COPY or GZIP unless the
+   end of the input file has been reached and all data has been processed.  */
+local int gz_fetch(state)
+    gz_statep state;
+{
+    z_streamp strm = &(state->strm);
+
+    do {
+        switch(state->how) {
+        case LOOK:      /* -> LOOK, COPY (only if never GZIP), or GZIP */
+            if (gz_look(state) == -1)
+                return -1;
+            if (state->how == LOOK)
+                return 0;
+            break;
+        case COPY:      /* -> COPY */
+            if (gz_load(state, state->out, state->size << 1, &(state->x.have))
+                    == -1)
+                return -1;
+            state->x.next = state->out;
+            return 0;
+        case GZIP:      /* -> GZIP or LOOK (if end of gzip stream) */
+            strm->avail_out = state->size << 1;
+            strm->next_out = state->out;
+            if (gz_decomp(state) == -1)
+                return -1;
+        }
+    } while (state->x.have == 0 && (!state->eof || strm->avail_in));
+    return 0;
+}
+
+/* Skip len uncompressed bytes of output.  Return -1 on error, 0 on success. */
+local int gz_skip(state, len)
+    gz_statep state;
+    z_off64_t len;
+{
+    unsigned n;
+
+    /* skip over len bytes or reach end-of-file, whichever comes first */
+    while (len)
+        /* skip over whatever is in output buffer */
+        if (state->x.have) {
+            n = GT_OFF(state->x.have) || (z_off64_t)state->x.have > len ?
+                (unsigned)len : state->x.have;
+            state->x.have -= n;
+            state->x.next += n;
+            state->x.pos += n;
+            len -= n;
+        }
+
+        /* output buffer empty -- return if we're at the end of the input */
+        else if (state->eof && state->strm.avail_in == 0)
+            break;
+
+        /* need more data to skip -- load up output buffer */
+        else {
+            /* get more output, looking for header if required */
+            if (gz_fetch(state) == -1)
+                return -1;
+        }
+    return 0;
+}
+
+/* Read len bytes into buf from file, or less than len up to the end of the
+   input.  Return the number of bytes read.  If zero is returned, either the
+   end of file was reached, or there was an error.  state->err must be
+   consulted in that case to determine which. */
+local z_size_t gz_read(state, buf, len)
+    gz_statep state;
+    voidp buf;
+    z_size_t len;
+{
+    z_size_t got;
+    unsigned n;
+
+    /* if len is zero, avoid unnecessary operations */
+    if (len == 0)
+        return 0;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return 0;
+    }
+
+    /* get len bytes to buf, or less than len if at the end */
+    got = 0;
+    do {
+        /* set n to the maximum amount of len that fits in an unsigned int */
+        n = -1;
+        if (n > len)
+            n = len;
+
+        /* first just try copying data from the output buffer */
+        if (state->x.have) {
+            if (state->x.have < n)
+                n = state->x.have;
+            memcpy(buf, state->x.next, n);
+            state->x.next += n;
+            state->x.have -= n;
+        }
+
+        /* output buffer empty -- return if we're at the end of the input */
+        else if (state->eof && state->strm.avail_in == 0) {
+            state->past = 1;        /* tried to read past end */
+            break;
+        }
+
+        /* need output data -- for small len or new stream load up our output
+           buffer */
+        else if (state->how == LOOK || n < (state->size << 1)) {
+            /* get more output, looking for header if required */
+            if (gz_fetch(state) == -1)
+                return 0;
+            continue;       /* no progress yet -- go back to copy above */
+            /* the copy above assures that we will leave with space in the
+               output buffer, allowing at least one gzungetc() to succeed */
+        }
+
+        /* large len -- read directly into user buffer */
+        else if (state->how == COPY) {      /* read directly */
+            if (gz_load(state, (unsigned char *)buf, n, &n) == -1)
+                return 0;
+        }
+
+        /* large len -- decompress directly into user buffer */
+        else {  /* state->how == GZIP */
+            state->strm.avail_out = n;
+            state->strm.next_out = (unsigned char *)buf;
+            if (gz_decomp(state) == -1)
+                return 0;
+            n = state->x.have;
+            state->x.have = 0;
+        }
+
+        /* update progress */
+        len -= n;
+        buf = (char *)buf + n;
+        got += n;
+        state->x.pos += n;
+    } while (len);
+
+    /* return number of bytes read into user buffer */
+    return got;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzread(file, buf, len)
+    gzFile file;
+    voidp buf;
+    unsigned len;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* since an int is returned, make sure len fits in one, otherwise return
+       with an error (this avoids a flaw in the interface) */
+    if ((int)len < 0) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in an int");
+        return -1;
+    }
+
+    /* read len or fewer bytes to buf */
+    len = gz_read(state, buf, len);
+
+    /* check for an error */
+    if (len == 0 && state->err != Z_OK && state->err != Z_BUF_ERROR)
+        return -1;
+
+    /* return the number of bytes read (this is assured to fit in an int) */
+    return (int)len;
+}
+
+/* -- see zlib.h -- */
+z_size_t ZEXPORT gzfread(buf, size, nitems, file)
+    voidp buf;
+    z_size_t size;
+    z_size_t nitems;
+    gzFile file;
+{
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+            (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return 0;
+
+    /* compute bytes to read -- error on overflow */
+    len = nitems * size;
+    if (size && len / size != nitems) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
+        return 0;
+    }
+
+    /* read len or fewer bytes to buf, return the number of full items read */
+    return len ? gz_read(state, buf, len) / size : 0;
+}
+
+/* -- see zlib.h -- */
+#ifdef Z_PREFIX_SET
+#  undef z_gzgetc
+#else
+#  undef gzgetc
+#endif
+int ZEXPORT gzgetc(file)
+    gzFile file;
+{
+    int ret;
+    unsigned char buf[1];
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* try output buffer (no need to check for skip request) */
+    if (state->x.have) {
+        state->x.have--;
+        state->x.pos++;
+        return *(state->x.next)++;
+    }
+
+    /* nothing there -- try gz_read() */
+    ret = gz_read(state, buf, 1);
+    return ret < 1 ? -1 : buf[0];
+}
+
+int ZEXPORT gzgetc_(file)
+gzFile file;
+{
+    return gzgetc(file);
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzungetc(c, file)
+    int c;
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return -1;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return -1;
+    }
+
+    /* can't push EOF */
+    if (c < 0)
+        return -1;
+
+    /* if output buffer empty, put byte at end (allows more pushing) */
+    if (state->x.have == 0) {
+        state->x.have = 1;
+        state->x.next = state->out + (state->size << 1) - 1;
+        state->x.next[0] = (unsigned char)c;
+        state->x.pos--;
+        state->past = 0;
+        return c;
+    }
+
+    /* if no room, give up (must have already done a gzungetc()) */
+    if (state->x.have == (state->size << 1)) {
+        gz_error(state, Z_DATA_ERROR, "out of room to push characters");
+        return -1;
+    }
+
+    /* slide output data if needed and insert byte before existing data */
+    if (state->x.next == state->out) {
+        unsigned char *src = state->out + state->x.have;
+        unsigned char *dest = state->out + (state->size << 1);
+        while (src > state->out)
+            *--dest = *--src;
+        state->x.next = dest;
+    }
+    state->x.have++;
+    state->x.next--;
+    state->x.next[0] = (unsigned char)c;
+    state->x.pos--;
+    state->past = 0;
+    return c;
+}
+
+/* -- see zlib.h -- */
+char * ZEXPORT gzgets(file, buf, len)
+    gzFile file;
+    char *buf;
+    int len;
+{
+    unsigned left, n;
+    char *str;
+    unsigned char *eol;
+    gz_statep state;
+
+    /* check parameters and get internal structure */
+    if (file == NULL || buf == NULL || len < 1)
+        return NULL;
+    state = (gz_statep)file;
+
+    /* check that we're reading and that there's no (serious) error */
+    if (state->mode != GZ_READ ||
+        (state->err != Z_OK && state->err != Z_BUF_ERROR))
+        return NULL;
+
+    /* process a skip request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_skip(state, state->skip) == -1)
+            return NULL;
+    }
+
+    /* copy output bytes up to new line or len - 1, whichever comes first --
+       append a terminating zero to the string (we don't check for a zero in
+       the contents, let the user worry about that) */
+    str = buf;
+    left = (unsigned)len - 1;
+    if (left) do {
+        /* assure that something is in the output buffer */
+        if (state->x.have == 0 && gz_fetch(state) == -1)
+            return NULL;                /* error */
+        if (state->x.have == 0) {       /* end of file */
+            state->past = 1;            /* read past end */
+            break;                      /* return what we have */
+        }
+
+        /* look for end-of-line in current output buffer */
+        n = state->x.have > left ? left : state->x.have;
+        eol = (unsigned char *)memchr(state->x.next, '\n', n);
+        if (eol != NULL)
+            n = (unsigned)(eol - state->x.next) + 1;
+
+        /* copy through end-of-line, or remainder if not found */
+        memcpy(buf, state->x.next, n);
+        state->x.have -= n;
+        state->x.next += n;
+        state->x.pos += n;
+        left -= n;
+        buf += n;
+    } while (left && eol == NULL);
+
+    /* return terminated string, or if nothing, end of file */
+    if (buf == str)
+        return NULL;
+    buf[0] = 0;
+    return str;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzdirect(file)
+    gzFile file;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* if the state is not known, but we can find out, then do so (this is
+       mainly for right after a gzopen() or gzdopen()) */
+    if (state->mode == GZ_READ && state->how == LOOK && state->x.have == 0)
+        (void)gz_look(state);
+
+    /* return 1 if transparent, 0 if processing a gzip stream */
+    return state->direct;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzclose_r(file)
+    gzFile file;
+{
+    int ret, err;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're reading */
+    if (state->mode != GZ_READ)
+        return Z_STREAM_ERROR;
+
+    /* free memory and close file */
+    if (state->size) {
+        inflateEnd(&(state->strm));
+        free(state->out);
+        free(state->in);
+    }
+    err = state->err == Z_BUF_ERROR ? Z_BUF_ERROR : Z_OK;
+    gz_error(state, Z_OK, NULL);
+    free(state->path);
+    ret = close(state->fd);
+    free(state);
+    return ret ? Z_ERRNO : err;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzwrite.c glmark2-2021.02/src/zlib/gzwrite.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/gzwrite.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/gzwrite.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,665 @@
+/* gzwrite.c -- zlib functions for writing gzip files
+ * Copyright (C) 2004-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "gzguts.h"
+
+/* Local functions */
+local int gz_init OF((gz_statep));
+local int gz_comp OF((gz_statep, int));
+local int gz_zero OF((gz_statep, z_off64_t));
+local z_size_t gz_write OF((gz_statep, voidpc, z_size_t));
+
+/* Initialize state for writing a gzip file.  Mark initialization by setting
+   state->size to non-zero.  Return -1 on a memory allocation failure, or 0 on
+   success. */
+local int gz_init(state)
+    gz_statep state;
+{
+    int ret;
+    z_streamp strm = &(state->strm);
+
+    /* allocate input buffer (double size for gzprintf) */
+    state->in = (unsigned char *)malloc(state->want << 1);
+    if (state->in == NULL) {
+        gz_error(state, Z_MEM_ERROR, "out of memory");
+        return -1;
+    }
+
+    /* only need output buffer and deflate state if compressing */
+    if (!state->direct) {
+        /* allocate output buffer */
+        state->out = (unsigned char *)malloc(state->want);
+        if (state->out == NULL) {
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+
+        /* allocate deflate memory, set up for gzip compression */
+        strm->zalloc = Z_NULL;
+        strm->zfree = Z_NULL;
+        strm->opaque = Z_NULL;
+        ret = deflateInit2(strm, state->level, Z_DEFLATED,
+                           MAX_WBITS + 16, DEF_MEM_LEVEL, state->strategy);
+        if (ret != Z_OK) {
+            free(state->out);
+            free(state->in);
+            gz_error(state, Z_MEM_ERROR, "out of memory");
+            return -1;
+        }
+        strm->next_in = NULL;
+    }
+
+    /* mark state as initialized */
+    state->size = state->want;
+
+    /* initialize write buffer if compressing */
+    if (!state->direct) {
+        strm->avail_out = state->size;
+        strm->next_out = state->out;
+        state->x.next = strm->next_out;
+    }
+    return 0;
+}
+
+/* Compress whatever is at avail_in and next_in and write to the output file.
+   Return -1 if there is an error writing to the output file or if gz_init()
+   fails to allocate memory, otherwise 0.  flush is assumed to be a valid
+   deflate() flush value.  If flush is Z_FINISH, then the deflate() state is
+   reset to start a new gzip stream.  If gz->direct is true, then simply write
+   to the output file without compressing, and ignore flush. */
+local int gz_comp(state, flush)
+    gz_statep state;
+    int flush;
+{
+    int ret, writ;
+    unsigned have, put, max = ((unsigned)-1 >> 2) + 1;
+    z_streamp strm = &(state->strm);
+
+    /* allocate memory if this is the first time through */
+    if (state->size == 0 && gz_init(state) == -1)
+        return -1;
+
+    /* write directly if requested */
+    if (state->direct) {
+        while (strm->avail_in) {
+            put = strm->avail_in > max ? max : strm->avail_in;
+            writ = write(state->fd, strm->next_in, put);
+            if (writ < 0) {
+                gz_error(state, Z_ERRNO, zstrerror());
+                return -1;
+            }
+            strm->avail_in -= (unsigned)writ;
+            strm->next_in += writ;
+        }
+        return 0;
+    }
+
+    /* run deflate() on provided input until it produces no more output */
+    ret = Z_OK;
+    do {
+        /* write out current buffer contents if full, or if flushing, but if
+           doing Z_FINISH then don't write until we get to Z_STREAM_END */
+        if (strm->avail_out == 0 || (flush != Z_NO_FLUSH &&
+            (flush != Z_FINISH || ret == Z_STREAM_END))) {
+            while (strm->next_out > state->x.next) {
+                put = strm->next_out - state->x.next > (int)max ? max :
+                      (unsigned)(strm->next_out - state->x.next);
+                writ = write(state->fd, state->x.next, put);
+                if (writ < 0) {
+                    gz_error(state, Z_ERRNO, zstrerror());
+                    return -1;
+                }
+                state->x.next += writ;
+            }
+            if (strm->avail_out == 0) {
+                strm->avail_out = state->size;
+                strm->next_out = state->out;
+                state->x.next = state->out;
+            }
+        }
+
+        /* compress */
+        have = strm->avail_out;
+        ret = deflate(strm, flush);
+        if (ret == Z_STREAM_ERROR) {
+            gz_error(state, Z_STREAM_ERROR,
+                      "internal error: deflate stream corrupt");
+            return -1;
+        }
+        have -= strm->avail_out;
+    } while (have);
+
+    /* if that completed a deflate stream, allow another to start */
+    if (flush == Z_FINISH)
+        deflateReset(strm);
+
+    /* all done, no errors */
+    return 0;
+}
+
+/* Compress len zeros to output.  Return -1 on a write error or memory
+   allocation failure by gz_comp(), or 0 on success. */
+local int gz_zero(state, len)
+    gz_statep state;
+    z_off64_t len;
+{
+    int first;
+    unsigned n;
+    z_streamp strm = &(state->strm);
+
+    /* consume whatever's left in the input buffer */
+    if (strm->avail_in && gz_comp(state, Z_NO_FLUSH) == -1)
+        return -1;
+
+    /* compress len zeros (len guaranteed > 0) */
+    first = 1;
+    while (len) {
+        n = GT_OFF(state->size) || (z_off64_t)state->size > len ?
+            (unsigned)len : state->size;
+        if (first) {
+            memset(state->in, 0, n);
+            first = 0;
+        }
+        strm->avail_in = n;
+        strm->next_in = state->in;
+        state->x.pos += n;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return -1;
+        len -= n;
+    }
+    return 0;
+}
+
+/* Write len bytes from buf to file.  Return the number of bytes written.  If
+   the returned value is less than len, then there was an error. */
+local z_size_t gz_write(state, buf, len)
+    gz_statep state;
+    voidpc buf;
+    z_size_t len;
+{
+    z_size_t put = len;
+
+    /* if len is zero, avoid unnecessary operations */
+    if (len == 0)
+        return 0;
+
+    /* allocate memory if this is the first time through */
+    if (state->size == 0 && gz_init(state) == -1)
+        return 0;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return 0;
+    }
+
+    /* for small len, copy to input buffer, otherwise compress directly */
+    if (len < state->size) {
+        /* copy to input buffer, compress when full */
+        do {
+            unsigned have, copy;
+
+            if (state->strm.avail_in == 0)
+                state->strm.next_in = state->in;
+            have = (unsigned)((state->strm.next_in + state->strm.avail_in) -
+                              state->in);
+            copy = state->size - have;
+            if (copy > len)
+                copy = len;
+            memcpy(state->in + have, buf, copy);
+            state->strm.avail_in += copy;
+            state->x.pos += copy;
+            buf = (const char *)buf + copy;
+            len -= copy;
+            if (len && gz_comp(state, Z_NO_FLUSH) == -1)
+                return 0;
+        } while (len);
+    }
+    else {
+        /* consume whatever's left in the input buffer */
+        if (state->strm.avail_in && gz_comp(state, Z_NO_FLUSH) == -1)
+            return 0;
+
+        /* directly compress user buffer to file */
+        state->strm.next_in = (z_const Bytef *)buf;
+        do {
+            unsigned n = (unsigned)-1;
+            if (n > len)
+                n = len;
+            state->strm.avail_in = n;
+            state->x.pos += n;
+            if (gz_comp(state, Z_NO_FLUSH) == -1)
+                return 0;
+            len -= n;
+        } while (len);
+    }
+
+    /* input was all buffered or compressed */
+    return put;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzwrite(file, buf, len)
+    gzFile file;
+    voidpc buf;
+    unsigned len;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return 0;
+
+    /* since an int is returned, make sure len fits in one, otherwise return
+       with an error (this avoids a flaw in the interface) */
+    if ((int)len < 0) {
+        gz_error(state, Z_DATA_ERROR, "requested length does not fit in int");
+        return 0;
+    }
+
+    /* write len bytes from buf (the return value will fit in an int) */
+    return (int)gz_write(state, buf, len);
+}
+
+/* -- see zlib.h -- */
+z_size_t ZEXPORT gzfwrite(buf, size, nitems, file)
+    voidpc buf;
+    z_size_t size;
+    z_size_t nitems;
+    gzFile file;
+{
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return 0;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return 0;
+
+    /* compute bytes to read -- error on overflow */
+    len = nitems * size;
+    if (size && len / size != nitems) {
+        gz_error(state, Z_STREAM_ERROR, "request does not fit in a size_t");
+        return 0;
+    }
+
+    /* write len bytes to buf, return the number of full items written */
+    return len ? gz_write(state, buf, len) / size : 0;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzputc(file, c)
+    gzFile file;
+    int c;
+{
+    unsigned have;
+    unsigned char buf[1];
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return -1;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return -1;
+    }
+
+    /* try writing to input buffer for speed (state->size == 0 if buffer not
+       initialized) */
+    if (state->size) {
+        if (strm->avail_in == 0)
+            strm->next_in = state->in;
+        have = (unsigned)((strm->next_in + strm->avail_in) - state->in);
+        if (have < state->size) {
+            state->in[have] = (unsigned char)c;
+            strm->avail_in++;
+            state->x.pos++;
+            return c & 0xff;
+        }
+    }
+
+    /* no room in buffer or not initialized, use gz_write() */
+    buf[0] = (unsigned char)c;
+    if (gz_write(state, buf, 1) != 1)
+        return -1;
+    return c & 0xff;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzputs(file, str)
+    gzFile file;
+    const char *str;
+{
+    int ret;
+    z_size_t len;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return -1;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return -1;
+
+    /* write string */
+    len = strlen(str);
+    ret = gz_write(state, str, len);
+    return ret == 0 && len != 0 ? -1 : ret;
+}
+
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#include <stdarg.h>
+
+/* -- see zlib.h -- */
+int ZEXPORTVA gzvprintf(gzFile file, const char *format, va_list va)
+{
+    int len;
+    unsigned left;
+    char *next;
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* make sure we have some buffer space */
+    if (state->size == 0 && gz_init(state) == -1)
+        return state->err;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* do the printf() into the input buffer, put length in len -- the input
+       buffer is double-sized just for this function, so there is guaranteed to
+       be state->size bytes available after the current contents */
+    if (strm->avail_in == 0)
+        strm->next_in = state->in;
+    next = (char *)(state->in + (strm->next_in - state->in) + strm->avail_in);
+    next[state->size - 1] = 0;
+#ifdef NO_vsnprintf
+#  ifdef HAS_vsprintf_void
+    (void)vsprintf(next, format, va);
+    for (len = 0; len < state->size; len++)
+        if (next[len] == 0) break;
+#  else
+    len = vsprintf(next, format, va);
+#  endif
+#else
+#  ifdef HAS_vsnprintf_void
+    (void)vsnprintf(next, state->size, format, va);
+    len = strlen(next);
+#  else
+    len = vsnprintf(next, state->size, format, va);
+#  endif
+#endif
+
+    /* check that printf() results fit in buffer */
+    if (len == 0 || (unsigned)len >= state->size || next[state->size - 1] != 0)
+        return 0;
+
+    /* update buffer and position, compress first half if past that */
+    strm->avail_in += (unsigned)len;
+    state->x.pos += len;
+    if (strm->avail_in >= state->size) {
+        left = strm->avail_in - state->size;
+        strm->avail_in = state->size;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return state->err;
+        memcpy(state->in, state->in + state->size, left);
+        strm->next_in = state->in;
+        strm->avail_in = left;
+    }
+    return len;
+}
+
+int ZEXPORTVA gzprintf(gzFile file, const char *format, ...)
+{
+    va_list va;
+    int ret;
+
+    va_start(va, format);
+    ret = gzvprintf(file, format, va);
+    va_end(va);
+    return ret;
+}
+
+#else /* !STDC && !Z_HAVE_STDARG_H */
+
+/* -- see zlib.h -- */
+int ZEXPORTVA gzprintf (file, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+                       a11, a12, a13, a14, a15, a16, a17, a18, a19, a20)
+    gzFile file;
+    const char *format;
+    int a1, a2, a3, a4, a5, a6, a7, a8, a9, a10,
+        a11, a12, a13, a14, a15, a16, a17, a18, a19, a20;
+{
+    unsigned len, left;
+    char *next;
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that can really pass pointer in ints */
+    if (sizeof(int) != sizeof(void *))
+        return Z_STREAM_ERROR;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* make sure we have some buffer space */
+    if (state->size == 0 && gz_init(state) == -1)
+        return state->error;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->error;
+    }
+
+    /* do the printf() into the input buffer, put length in len -- the input
+       buffer is double-sized just for this function, so there is guaranteed to
+       be state->size bytes available after the current contents */
+    if (strm->avail_in == 0)
+        strm->next_in = state->in;
+    next = (char *)(strm->next_in + strm->avail_in);
+    next[state->size - 1] = 0;
+#ifdef NO_snprintf
+#  ifdef HAS_sprintf_void
+    sprintf(next, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11, a12,
+            a13, a14, a15, a16, a17, a18, a19, a20);
+    for (len = 0; len < size; len++)
+        if (next[len] == 0)
+            break;
+#  else
+    len = sprintf(next, format, a1, a2, a3, a4, a5, a6, a7, a8, a9, a10, a11,
+                  a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#  endif
+#else
+#  ifdef HAS_snprintf_void
+    snprintf(next, state->size, format, a1, a2, a3, a4, a5, a6, a7, a8, a9,
+             a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+    len = strlen(next);
+#  else
+    len = snprintf(next, state->size, format, a1, a2, a3, a4, a5, a6, a7, a8,
+                   a9, a10, a11, a12, a13, a14, a15, a16, a17, a18, a19, a20);
+#  endif
+#endif
+
+    /* check that printf() results fit in buffer */
+    if (len == 0 || len >= state->size || next[state->size - 1] != 0)
+        return 0;
+
+    /* update buffer and position, compress first half if past that */
+    strm->avail_in += len;
+    state->x.pos += len;
+    if (strm->avail_in >= state->size) {
+        left = strm->avail_in - state->size;
+        strm->avail_in = state->size;
+        if (gz_comp(state, Z_NO_FLUSH) == -1)
+            return state->err;
+        memcpy(state->in, state->in + state->size, left);
+        strm->next_in = state->in;
+        strm->avail_in = left;
+    }
+    return (int)len;
+}
+
+#endif
+
+/* -- see zlib.h -- */
+int ZEXPORT gzflush(file, flush)
+    gzFile file;
+    int flush;
+{
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* check flush parameter */
+    if (flush < 0 || flush > Z_FINISH)
+        return Z_STREAM_ERROR;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* compress remaining data with requested flush */
+    (void)gz_comp(state, flush);
+    return state->err;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzsetparams(file, level, strategy)
+    gzFile file;
+    int level;
+    int strategy;
+{
+    gz_statep state;
+    z_streamp strm;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+    strm = &(state->strm);
+
+    /* check that we're writing and that there's no error */
+    if (state->mode != GZ_WRITE || state->err != Z_OK)
+        return Z_STREAM_ERROR;
+
+    /* if no change is requested, then do nothing */
+    if (level == state->level && strategy == state->strategy)
+        return Z_OK;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            return state->err;
+    }
+
+    /* change compression parameters for subsequent input */
+    if (state->size) {
+        /* flush previous input with previous parameters before changing */
+        if (strm->avail_in && gz_comp(state, Z_BLOCK) == -1)
+            return state->err;
+        deflateParams(strm, level, strategy);
+    }
+    state->level = level;
+    state->strategy = strategy;
+    return Z_OK;
+}
+
+/* -- see zlib.h -- */
+int ZEXPORT gzclose_w(file)
+    gzFile file;
+{
+    int ret = Z_OK;
+    gz_statep state;
+
+    /* get internal structure */
+    if (file == NULL)
+        return Z_STREAM_ERROR;
+    state = (gz_statep)file;
+
+    /* check that we're writing */
+    if (state->mode != GZ_WRITE)
+        return Z_STREAM_ERROR;
+
+    /* check for seek request */
+    if (state->seek) {
+        state->seek = 0;
+        if (gz_zero(state, state->skip) == -1)
+            ret = state->err;
+    }
+
+    /* flush, free memory, and close file */
+    if (gz_comp(state, Z_FINISH) == -1)
+        ret = state->err;
+    if (state->size) {
+        if (!state->direct) {
+            (void)deflateEnd(&(state->strm));
+            free(state->out);
+        }
+        free(state->in);
+    }
+    gz_error(state, Z_OK, NULL);
+    free(state->path);
+    if (close(state->fd) == -1)
+        ret = Z_ERRNO;
+    free(state);
+    return ret;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/infback.c glmark2-2021.02/src/zlib/infback.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/infback.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/infback.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,640 @@
+/* infback.c -- inflate using a call-back interface
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+   This code is largely copied from inflate.c.  Normally either infback.o or
+   inflate.o would be linked into an application--not both.  The interface
+   with inffast.c is retained so that optimized assembler-coded versions of
+   inflate_fast() can be used with either inflate.c or infback.c.
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+/* function prototypes */
+local void fixedtables OF((struct inflate_state FAR *state));
+
+/*
+   strm provides memory allocation functions in zalloc and zfree, or
+   Z_NULL to use the library memory allocation functions.
+
+   windowBits is in the range 8..15, and window is a user-supplied
+   window and output buffer that is 2**windowBits bytes.
+ */
+int ZEXPORT inflateBackInit_(strm, windowBits, window, version, stream_size)
+z_streamp strm;
+int windowBits;
+unsigned char FAR *window;
+const char *version;
+int stream_size;
+{
+    struct inflate_state FAR *state;
+
+    if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
+        stream_size != (int)(sizeof(z_stream)))
+        return Z_VERSION_ERROR;
+    if (strm == Z_NULL || window == Z_NULL ||
+        windowBits < 8 || windowBits > 15)
+        return Z_STREAM_ERROR;
+    strm->msg = Z_NULL;                 /* in case we return an error */
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+    strm->zfree = zcfree;
+#endif
+    state = (struct inflate_state FAR *)ZALLOC(strm, 1,
+                                               sizeof(struct inflate_state));
+    if (state == Z_NULL) return Z_MEM_ERROR;
+    Tracev((stderr, "inflate: allocated\n"));
+    strm->state = (struct internal_state FAR *)state;
+    state->dmax = 32768U;
+    state->wbits = (uInt)windowBits;
+    state->wsize = 1U << windowBits;
+    state->window = window;
+    state->wnext = 0;
+    state->whave = 0;
+    return Z_OK;
+}
+
+/*
+   Return state with length and distance decoding tables and index sizes set to
+   fixed code decoding.  Normally this returns fixed tables from inffixed.h.
+   If BUILDFIXED is defined, then instead this routine builds the tables the
+   first time it's called, and returns those tables the first time and
+   thereafter.  This reduces the size of the code by about 2K bytes, in
+   exchange for a little execution time.  However, BUILDFIXED should not be
+   used for threaded applications, since the rewriting of the tables and virgin
+   may not be thread-safe.
+ */
+local void fixedtables(state)
+struct inflate_state FAR *state;
+{
+#ifdef BUILDFIXED
+    static int virgin = 1;
+    static code *lenfix, *distfix;
+    static code fixed[544];
+
+    /* build fixed huffman tables if first call (may not be thread safe) */
+    if (virgin) {
+        unsigned sym, bits;
+        static code *next;
+
+        /* literal/length table */
+        sym = 0;
+        while (sym < 144) state->lens[sym++] = 8;
+        while (sym < 256) state->lens[sym++] = 9;
+        while (sym < 280) state->lens[sym++] = 7;
+        while (sym < 288) state->lens[sym++] = 8;
+        next = fixed;
+        lenfix = next;
+        bits = 9;
+        inflate_table(LENS, state->lens, 288, &(next), &(bits), state->work);
+
+        /* distance table */
+        sym = 0;
+        while (sym < 32) state->lens[sym++] = 5;
+        distfix = next;
+        bits = 5;
+        inflate_table(DISTS, state->lens, 32, &(next), &(bits), state->work);
+
+        /* do this just once */
+        virgin = 0;
+    }
+#else /* !BUILDFIXED */
+#   include "inffixed.h"
+#endif /* BUILDFIXED */
+    state->lencode = lenfix;
+    state->lenbits = 9;
+    state->distcode = distfix;
+    state->distbits = 5;
+}
+
+/* Macros for inflateBack(): */
+
+/* Load returned state from inflate_fast() */
+#define LOAD() \
+    do { \
+        put = strm->next_out; \
+        left = strm->avail_out; \
+        next = strm->next_in; \
+        have = strm->avail_in; \
+        hold = state->hold; \
+        bits = state->bits; \
+    } while (0)
+
+/* Set state from registers for inflate_fast() */
+#define RESTORE() \
+    do { \
+        strm->next_out = put; \
+        strm->avail_out = left; \
+        strm->next_in = next; \
+        strm->avail_in = have; \
+        state->hold = hold; \
+        state->bits = bits; \
+    } while (0)
+
+/* Clear the input bit accumulator */
+#define INITBITS() \
+    do { \
+        hold = 0; \
+        bits = 0; \
+    } while (0)
+
+/* Assure that some input is available.  If input is requested, but denied,
+   then return a Z_BUF_ERROR from inflateBack(). */
+#define PULL() \
+    do { \
+        if (have == 0) { \
+            have = in(in_desc, &next); \
+            if (have == 0) { \
+                next = Z_NULL; \
+                ret = Z_BUF_ERROR; \
+                goto inf_leave; \
+            } \
+        } \
+    } while (0)
+
+/* Get a byte of input into the bit accumulator, or return from inflateBack()
+   with an error if there is no input available. */
+#define PULLBYTE() \
+    do { \
+        PULL(); \
+        have--; \
+        hold += (unsigned long)(*next++) << bits; \
+        bits += 8; \
+    } while (0)
+
+/* Assure that there are at least n bits in the bit accumulator.  If there is
+   not enough available input to do that, then return from inflateBack() with
+   an error. */
+#define NEEDBITS(n) \
+    do { \
+        while (bits < (unsigned)(n)) \
+            PULLBYTE(); \
+    } while (0)
+
+/* Return the low n bits of the bit accumulator (n < 16) */
+#define BITS(n) \
+    ((unsigned)hold & ((1U << (n)) - 1))
+
+/* Remove n bits from the bit accumulator */
+#define DROPBITS(n) \
+    do { \
+        hold >>= (n); \
+        bits -= (unsigned)(n); \
+    } while (0)
+
+/* Remove zero to seven bits as needed to go to a byte boundary */
+#define BYTEBITS() \
+    do { \
+        hold >>= bits & 7; \
+        bits -= bits & 7; \
+    } while (0)
+
+/* Assure that some output space is available, by writing out the window
+   if it's full.  If the write fails, return from inflateBack() with a
+   Z_BUF_ERROR. */
+#define ROOM() \
+    do { \
+        if (left == 0) { \
+            put = state->window; \
+            left = state->wsize; \
+            state->whave = left; \
+            if (out(out_desc, put, left)) { \
+                ret = Z_BUF_ERROR; \
+                goto inf_leave; \
+            } \
+        } \
+    } while (0)
+
+/*
+   strm provides the memory allocation functions and window buffer on input,
+   and provides information on the unused input on return.  For Z_DATA_ERROR
+   returns, strm will also provide an error message.
+
+   in() and out() are the call-back input and output functions.  When
+   inflateBack() needs more input, it calls in().  When inflateBack() has
+   filled the window with output, or when it completes with data in the
+   window, it calls out() to write out the data.  The application must not
+   change the provided input until in() is called again or inflateBack()
+   returns.  The application must not change the window/output buffer until
+   inflateBack() returns.
+
+   in() and out() are called with a descriptor parameter provided in the
+   inflateBack() call.  This parameter can be a structure that provides the
+   information required to do the read or write, as well as accumulated
+   information on the input and output such as totals and check values.
+
+   in() should return zero on failure.  out() should return non-zero on
+   failure.  If either in() or out() fails, than inflateBack() returns a
+   Z_BUF_ERROR.  strm->next_in can be checked for Z_NULL to see whether it
+   was in() or out() that caused in the error.  Otherwise,  inflateBack()
+   returns Z_STREAM_END on success, Z_DATA_ERROR for an deflate format
+   error, or Z_MEM_ERROR if it could not allocate memory for the state.
+   inflateBack() can also return Z_STREAM_ERROR if the input parameters
+   are not correct, i.e. strm is Z_NULL or the state was not initialized.
+ */
+int ZEXPORT inflateBack(strm, in, in_desc, out, out_desc)
+z_streamp strm;
+in_func in;
+void FAR *in_desc;
+out_func out;
+void FAR *out_desc;
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *next;    /* next input */
+    unsigned char FAR *put;     /* next output */
+    unsigned have, left;        /* available input and output */
+    unsigned long hold;         /* bit buffer */
+    unsigned bits;              /* bits in bit buffer */
+    unsigned copy;              /* number of stored or match bytes to copy */
+    unsigned char FAR *from;    /* where to copy match bytes from */
+    code here;                  /* current decoding table entry */
+    code last;                  /* parent table entry */
+    unsigned len;               /* length to copy for repeats, bits to drop */
+    int ret;                    /* return code */
+    static const unsigned short order[19] = /* permutation of code lengths */
+        {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+    /* Check that the strm exists and that the state was initialized */
+    if (strm == Z_NULL || strm->state == Z_NULL)
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* Reset the state */
+    strm->msg = Z_NULL;
+    state->mode = TYPE;
+    state->last = 0;
+    state->whave = 0;
+    next = strm->next_in;
+    have = next != Z_NULL ? strm->avail_in : 0;
+    hold = 0;
+    bits = 0;
+    put = state->window;
+    left = state->wsize;
+
+    /* Inflate until end of block marked as last */
+    for (;;)
+        switch (state->mode) {
+        case TYPE:
+            /* determine and dispatch block type */
+            if (state->last) {
+                BYTEBITS();
+                state->mode = DONE;
+                break;
+            }
+            NEEDBITS(3);
+            state->last = BITS(1);
+            DROPBITS(1);
+            switch (BITS(2)) {
+            case 0:                             /* stored block */
+                Tracev((stderr, "inflate:     stored block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = STORED;
+                break;
+            case 1:                             /* fixed block */
+                fixedtables(state);
+                Tracev((stderr, "inflate:     fixed codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = LEN;              /* decode codes */
+                break;
+            case 2:                             /* dynamic block */
+                Tracev((stderr, "inflate:     dynamic codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = TABLE;
+                break;
+            case 3:
+                strm->msg = (char *)"invalid block type";
+                state->mode = BAD;
+            }
+            DROPBITS(2);
+            break;
+
+        case STORED:
+            /* get and verify stored block length */
+            BYTEBITS();                         /* go to byte boundary */
+            NEEDBITS(32);
+            if ((hold & 0xffff) != ((hold >> 16) ^ 0xffff)) {
+                strm->msg = (char *)"invalid stored block lengths";
+                state->mode = BAD;
+                break;
+            }
+            state->length = (unsigned)hold & 0xffff;
+            Tracev((stderr, "inflate:       stored length %u\n",
+                    state->length));
+            INITBITS();
+
+            /* copy stored block from input to output */
+            while (state->length != 0) {
+                copy = state->length;
+                PULL();
+                ROOM();
+                if (copy > have) copy = have;
+                if (copy > left) copy = left;
+                zmemcpy(put, next, copy);
+                have -= copy;
+                next += copy;
+                left -= copy;
+                put += copy;
+                state->length -= copy;
+            }
+            Tracev((stderr, "inflate:       stored end\n"));
+            state->mode = TYPE;
+            break;
+
+        case TABLE:
+            /* get dynamic table entries descriptor */
+            NEEDBITS(14);
+            state->nlen = BITS(5) + 257;
+            DROPBITS(5);
+            state->ndist = BITS(5) + 1;
+            DROPBITS(5);
+            state->ncode = BITS(4) + 4;
+            DROPBITS(4);
+#ifndef PKZIP_BUG_WORKAROUND
+            if (state->nlen > 286 || state->ndist > 30) {
+                strm->msg = (char *)"too many length or distance symbols";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracev((stderr, "inflate:       table sizes ok\n"));
+
+            /* get code length code lengths (not a typo) */
+            state->have = 0;
+            while (state->have < state->ncode) {
+                NEEDBITS(3);
+                state->lens[order[state->have++]] = (unsigned short)BITS(3);
+                DROPBITS(3);
+            }
+            while (state->have < 19)
+                state->lens[order[state->have++]] = 0;
+            state->next = state->codes;
+            state->lencode = (code const FAR *)(state->next);
+            state->lenbits = 7;
+            ret = inflate_table(CODES, state->lens, 19, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid code lengths set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       code lengths ok\n"));
+
+            /* get length and distance code code lengths */
+            state->have = 0;
+            while (state->have < state->nlen + state->ndist) {
+                for (;;) {
+                    here = state->lencode[BITS(state->lenbits)];
+                    if ((unsigned)(here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                if (here.val < 16) {
+                    DROPBITS(here.bits);
+                    state->lens[state->have++] = here.val;
+                }
+                else {
+                    if (here.val == 16) {
+                        NEEDBITS(here.bits + 2);
+                        DROPBITS(here.bits);
+                        if (state->have == 0) {
+                            strm->msg = (char *)"invalid bit length repeat";
+                            state->mode = BAD;
+                            break;
+                        }
+                        len = (unsigned)(state->lens[state->have - 1]);
+                        copy = 3 + BITS(2);
+                        DROPBITS(2);
+                    }
+                    else if (here.val == 17) {
+                        NEEDBITS(here.bits + 3);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 3 + BITS(3);
+                        DROPBITS(3);
+                    }
+                    else {
+                        NEEDBITS(here.bits + 7);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 11 + BITS(7);
+                        DROPBITS(7);
+                    }
+                    if (state->have + copy > state->nlen + state->ndist) {
+                        strm->msg = (char *)"invalid bit length repeat";
+                        state->mode = BAD;
+                        break;
+                    }
+                    while (copy--)
+                        state->lens[state->have++] = (unsigned short)len;
+                }
+            }
+
+            /* handle error breaks in while */
+            if (state->mode == BAD) break;
+
+            /* check for end-of-block code (better have one) */
+            if (state->lens[256] == 0) {
+                strm->msg = (char *)"invalid code -- missing end-of-block";
+                state->mode = BAD;
+                break;
+            }
+
+            /* build code tables -- note: do not change the lenbits or distbits
+               values here (9 and 6) without reading the comments in inftrees.h
+               concerning the ENOUGH constants, which depend on those values */
+            state->next = state->codes;
+            state->lencode = (code const FAR *)(state->next);
+            state->lenbits = 9;
+            ret = inflate_table(LENS, state->lens, state->nlen, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid literal/lengths set";
+                state->mode = BAD;
+                break;
+            }
+            state->distcode = (code const FAR *)(state->next);
+            state->distbits = 6;
+            ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist,
+                            &(state->next), &(state->distbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid distances set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       codes ok\n"));
+            state->mode = LEN;
+
+        case LEN:
+            /* use inflate_fast() if we have enough input and output */
+            if (have >= 6 && left >= 258) {
+                RESTORE();
+                if (state->whave < state->wsize)
+                    state->whave = state->wsize - left;
+                inflate_fast(strm, state->wsize);
+                LOAD();
+                break;
+            }
+
+            /* get a literal, length, or end-of-block code */
+            for (;;) {
+                here = state->lencode[BITS(state->lenbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if (here.op && (here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->lencode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+            }
+            DROPBITS(here.bits);
+            state->length = (unsigned)here.val;
+
+            /* process literal */
+            if (here.op == 0) {
+                Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                        "inflate:         literal '%c'\n" :
+                        "inflate:         literal 0x%02x\n", here.val));
+                ROOM();
+                *put++ = (unsigned char)(state->length);
+                left--;
+                state->mode = LEN;
+                break;
+            }
+
+            /* process end of block */
+            if (here.op & 32) {
+                Tracevv((stderr, "inflate:         end of block\n"));
+                state->mode = TYPE;
+                break;
+            }
+
+            /* invalid code */
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid literal/length code";
+                state->mode = BAD;
+                break;
+            }
+
+            /* length code -- get extra bits, if any */
+            state->extra = (unsigned)(here.op) & 15;
+            if (state->extra != 0) {
+                NEEDBITS(state->extra);
+                state->length += BITS(state->extra);
+                DROPBITS(state->extra);
+            }
+            Tracevv((stderr, "inflate:         length %u\n", state->length));
+
+            /* get distance code */
+            for (;;) {
+                here = state->distcode[BITS(state->distbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if ((here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->distcode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+            }
+            DROPBITS(here.bits);
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+            state->offset = (unsigned)here.val;
+
+            /* get distance extra bits, if any */
+            state->extra = (unsigned)(here.op) & 15;
+            if (state->extra != 0) {
+                NEEDBITS(state->extra);
+                state->offset += BITS(state->extra);
+                DROPBITS(state->extra);
+            }
+            if (state->offset > state->wsize - (state->whave < state->wsize ?
+                                                left : 0)) {
+                strm->msg = (char *)"invalid distance too far back";
+                state->mode = BAD;
+                break;
+            }
+            Tracevv((stderr, "inflate:         distance %u\n", state->offset));
+
+            /* copy match from window to output */
+            do {
+                ROOM();
+                copy = state->wsize - state->offset;
+                if (copy < left) {
+                    from = put + copy;
+                    copy = left - copy;
+                }
+                else {
+                    from = put - state->offset;
+                    copy = left;
+                }
+                if (copy > state->length) copy = state->length;
+                state->length -= copy;
+                left -= copy;
+                do {
+                    *put++ = *from++;
+                } while (--copy);
+            } while (state->length != 0);
+            break;
+
+        case DONE:
+            /* inflate stream terminated properly -- write leftover output */
+            ret = Z_STREAM_END;
+            if (left < state->wsize) {
+                if (out(out_desc, state->window, state->wsize - left))
+                    ret = Z_BUF_ERROR;
+            }
+            goto inf_leave;
+
+        case BAD:
+            ret = Z_DATA_ERROR;
+            goto inf_leave;
+
+        default:                /* can't happen, but makes compilers happy */
+            ret = Z_STREAM_ERROR;
+            goto inf_leave;
+        }
+
+    /* Return unused input */
+  inf_leave:
+    strm->next_in = next;
+    strm->avail_in = have;
+    return ret;
+}
+
+int ZEXPORT inflateBackEnd(strm)
+z_streamp strm;
+{
+    if (strm == Z_NULL || strm->state == Z_NULL || strm->zfree == (free_func)0)
+        return Z_STREAM_ERROR;
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+    Tracev((stderr, "inflate: end\n"));
+    return Z_OK;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffast.c glmark2-2021.02/src/zlib/inffast.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffast.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inffast.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,323 @@
+/* inffast.c -- fast decoding
+ * Copyright (C) 1995-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+#ifdef ASMINF
+#  pragma message("Assembler code may have bugs -- use at your own risk")
+#else
+
+/*
+   Decode literal, length, and distance codes and write out the resulting
+   literal and match bytes until either not enough input or output is
+   available, an end-of-block is encountered, or a data error is encountered.
+   When large enough input and output buffers are supplied to inflate(), for
+   example, a 16K input buffer and a 64K output buffer, more than 95% of the
+   inflate execution time is spent in this routine.
+
+   Entry assumptions:
+
+        state->mode == LEN
+        strm->avail_in >= 6
+        strm->avail_out >= 258
+        start >= strm->avail_out
+        state->bits < 8
+
+   On return, state->mode is one of:
+
+        LEN -- ran out of enough output space or enough available input
+        TYPE -- reached end of block code, inflate() to interpret next block
+        BAD -- error in block data
+
+   Notes:
+
+    - The maximum input bits used by a length/distance pair is 15 bits for the
+      length code, 5 bits for the length extra, 15 bits for the distance code,
+      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
+      Therefore if strm->avail_in >= 6, then there is enough input to avoid
+      checking for available input while decoding.
+
+    - The maximum bytes that a single length/distance pair can output is 258
+      bytes, which is the maximum length that can be coded.  inflate_fast()
+      requires strm->avail_out >= 258 for each loop to avoid checking for
+      output space.
+ */
+void ZLIB_INTERNAL inflate_fast(strm, start)
+z_streamp strm;
+unsigned start;         /* inflate()'s starting value for strm->avail_out */
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *in;      /* local strm->next_in */
+    z_const unsigned char FAR *last;    /* have enough input while in < last */
+    unsigned char FAR *out;     /* local strm->next_out */
+    unsigned char FAR *beg;     /* inflate()'s initial strm->next_out */
+    unsigned char FAR *end;     /* while out < end, enough space available */
+#ifdef INFLATE_STRICT
+    unsigned dmax;              /* maximum distance from zlib header */
+#endif
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned wnext;             /* window write index */
+    unsigned char FAR *window;  /* allocated sliding window, if wsize != 0 */
+    unsigned long hold;         /* local strm->hold */
+    unsigned bits;              /* local strm->bits */
+    code const FAR *lcode;      /* local strm->lencode */
+    code const FAR *dcode;      /* local strm->distcode */
+    unsigned lmask;             /* mask for first level of length codes */
+    unsigned dmask;             /* mask for first level of distance codes */
+    code here;                  /* retrieved table entry */
+    unsigned op;                /* code bits, operation, extra bits, or */
+                                /*  window position, window bytes to copy */
+    unsigned len;               /* match length, unused bytes */
+    unsigned dist;              /* match distance */
+    unsigned char FAR *from;    /* where to copy match from */
+
+    /* copy state to local variables */
+    state = (struct inflate_state FAR *)strm->state;
+    in = strm->next_in;
+    last = in + (strm->avail_in - 5);
+    out = strm->next_out;
+    beg = out - (start - strm->avail_out);
+    end = out + (strm->avail_out - 257);
+#ifdef INFLATE_STRICT
+    dmax = state->dmax;
+#endif
+    wsize = state->wsize;
+    whave = state->whave;
+    wnext = state->wnext;
+    window = state->window;
+    hold = state->hold;
+    bits = state->bits;
+    lcode = state->lencode;
+    dcode = state->distcode;
+    lmask = (1U << state->lenbits) - 1;
+    dmask = (1U << state->distbits) - 1;
+
+    /* decode literals and length/distances until end-of-block or not enough
+       input data or output space */
+    do {
+        if (bits < 15) {
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+            hold += (unsigned long)(*in++) << bits;
+            bits += 8;
+        }
+        here = lcode[hold & lmask];
+      dolen:
+        op = (unsigned)(here.bits);
+        hold >>= op;
+        bits -= op;
+        op = (unsigned)(here.op);
+        if (op == 0) {                          /* literal */
+            Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                    "inflate:         literal '%c'\n" :
+                    "inflate:         literal 0x%02x\n", here.val));
+            *out++ = (unsigned char)(here.val);
+        }
+        else if (op & 16) {                     /* length base */
+            len = (unsigned)(here.val);
+            op &= 15;                           /* number of extra bits */
+            if (op) {
+                if (bits < op) {
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                }
+                len += (unsigned)hold & ((1U << op) - 1);
+                hold >>= op;
+                bits -= op;
+            }
+            Tracevv((stderr, "inflate:         length %u\n", len));
+            if (bits < 15) {
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+                hold += (unsigned long)(*in++) << bits;
+                bits += 8;
+            }
+            here = dcode[hold & dmask];
+          dodist:
+            op = (unsigned)(here.bits);
+            hold >>= op;
+            bits -= op;
+            op = (unsigned)(here.op);
+            if (op & 16) {                      /* distance base */
+                dist = (unsigned)(here.val);
+                op &= 15;                       /* number of extra bits */
+                if (bits < op) {
+                    hold += (unsigned long)(*in++) << bits;
+                    bits += 8;
+                    if (bits < op) {
+                        hold += (unsigned long)(*in++) << bits;
+                        bits += 8;
+                    }
+                }
+                dist += (unsigned)hold & ((1U << op) - 1);
+#ifdef INFLATE_STRICT
+                if (dist > dmax) {
+                    strm->msg = (char *)"invalid distance too far back";
+                    state->mode = BAD;
+                    break;
+                }
+#endif
+                hold >>= op;
+                bits -= op;
+                Tracevv((stderr, "inflate:         distance %u\n", dist));
+                op = (unsigned)(out - beg);     /* max distance in output */
+                if (dist > op) {                /* see if copy from window */
+                    op = dist - op;             /* distance back in window */
+                    if (op > whave) {
+                        if (state->sane) {
+                            strm->msg =
+                                (char *)"invalid distance too far back";
+                            state->mode = BAD;
+                            break;
+                        }
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                        if (len <= op - whave) {
+                            do {
+                                *out++ = 0;
+                            } while (--len);
+                            continue;
+                        }
+                        len -= op - whave;
+                        do {
+                            *out++ = 0;
+                        } while (--op > whave);
+                        if (op == 0) {
+                            from = out - dist;
+                            do {
+                                *out++ = *from++;
+                            } while (--len);
+                            continue;
+                        }
+#endif
+                    }
+                    from = window;
+                    if (wnext == 0) {           /* very common case */
+                        from += wsize - op;
+                        if (op < len) {         /* some from window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = out - dist;  /* rest from output */
+                        }
+                    }
+                    else if (wnext < op) {      /* wrap around window */
+                        from += wsize + wnext - op;
+                        op -= wnext;
+                        if (op < len) {         /* some from end of window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = window;
+                            if (wnext < len) {  /* some from start of window */
+                                op = wnext;
+                                len -= op;
+                                do {
+                                    *out++ = *from++;
+                                } while (--op);
+                                from = out - dist;      /* rest from output */
+                            }
+                        }
+                    }
+                    else {                      /* contiguous in window */
+                        from += wnext - op;
+                        if (op < len) {         /* some from window */
+                            len -= op;
+                            do {
+                                *out++ = *from++;
+                            } while (--op);
+                            from = out - dist;  /* rest from output */
+                        }
+                    }
+                    while (len > 2) {
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        len -= 3;
+                    }
+                    if (len) {
+                        *out++ = *from++;
+                        if (len > 1)
+                            *out++ = *from++;
+                    }
+                }
+                else {
+                    from = out - dist;          /* copy direct from output */
+                    do {                        /* minimum length is three */
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        *out++ = *from++;
+                        len -= 3;
+                    } while (len > 2);
+                    if (len) {
+                        *out++ = *from++;
+                        if (len > 1)
+                            *out++ = *from++;
+                    }
+                }
+            }
+            else if ((op & 64) == 0) {          /* 2nd level distance code */
+                here = dcode[here.val + (hold & ((1U << op) - 1))];
+                goto dodist;
+            }
+            else {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+        }
+        else if ((op & 64) == 0) {              /* 2nd level length code */
+            here = lcode[here.val + (hold & ((1U << op) - 1))];
+            goto dolen;
+        }
+        else if (op & 32) {                     /* end-of-block */
+            Tracevv((stderr, "inflate:         end of block\n"));
+            state->mode = TYPE;
+            break;
+        }
+        else {
+            strm->msg = (char *)"invalid literal/length code";
+            state->mode = BAD;
+            break;
+        }
+    } while (in < last && out < end);
+
+    /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
+    len = bits >> 3;
+    in -= len;
+    bits -= len << 3;
+    hold &= (1U << bits) - 1;
+
+    /* update state and return */
+    strm->next_in = in;
+    strm->next_out = out;
+    strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last));
+    strm->avail_out = (unsigned)(out < end ?
+                                 257 + (end - out) : 257 - (out - end));
+    state->hold = hold;
+    state->bits = bits;
+    return;
+}
+
+/*
+   inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe):
+   - Using bit fields for code structure
+   - Different op definition to avoid & for extra bits (do & for table bits)
+   - Three separate decoding do-loops for direct, window, and wnext == 0
+   - Special case for distance > 1 copies to do overlapped load and store copy
+   - Explicit branch predictions (based on measured branch probabilities)
+   - Deferring match copy and interspersed it with decoding subsequent codes
+   - Swapping literal/length else
+   - Swapping window/direct else
+   - Larger unrolled copy loops (three is about right)
+   - Moving len -= 3 statement into middle of loop
+ */
+
+#endif /* !ASMINF */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffast.h glmark2-2021.02/src/zlib/inffast.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffast.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inffast.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+/* inffast.h -- header to use inffast.c
+ * Copyright (C) 1995-2003, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+void ZLIB_INTERNAL inflate_fast OF((z_streamp strm, unsigned start));
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffixed.h glmark2-2021.02/src/zlib/inffixed.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inffixed.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inffixed.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,94 @@
+    /* inffixed.h -- table for decoding fixed codes
+     * Generated automatically by makefixed().
+     */
+
+    /* WARNING: this file should *not* be used by applications.
+       It is part of the implementation of this library and is
+       subject to change. Applications should only use zlib.h.
+     */
+
+    static const code lenfix[512] = {
+        {96,7,0},{0,8,80},{0,8,16},{20,8,115},{18,7,31},{0,8,112},{0,8,48},
+        {0,9,192},{16,7,10},{0,8,96},{0,8,32},{0,9,160},{0,8,0},{0,8,128},
+        {0,8,64},{0,9,224},{16,7,6},{0,8,88},{0,8,24},{0,9,144},{19,7,59},
+        {0,8,120},{0,8,56},{0,9,208},{17,7,17},{0,8,104},{0,8,40},{0,9,176},
+        {0,8,8},{0,8,136},{0,8,72},{0,9,240},{16,7,4},{0,8,84},{0,8,20},
+        {21,8,227},{19,7,43},{0,8,116},{0,8,52},{0,9,200},{17,7,13},{0,8,100},
+        {0,8,36},{0,9,168},{0,8,4},{0,8,132},{0,8,68},{0,9,232},{16,7,8},
+        {0,8,92},{0,8,28},{0,9,152},{20,7,83},{0,8,124},{0,8,60},{0,9,216},
+        {18,7,23},{0,8,108},{0,8,44},{0,9,184},{0,8,12},{0,8,140},{0,8,76},
+        {0,9,248},{16,7,3},{0,8,82},{0,8,18},{21,8,163},{19,7,35},{0,8,114},
+        {0,8,50},{0,9,196},{17,7,11},{0,8,98},{0,8,34},{0,9,164},{0,8,2},
+        {0,8,130},{0,8,66},{0,9,228},{16,7,7},{0,8,90},{0,8,26},{0,9,148},
+        {20,7,67},{0,8,122},{0,8,58},{0,9,212},{18,7,19},{0,8,106},{0,8,42},
+        {0,9,180},{0,8,10},{0,8,138},{0,8,74},{0,9,244},{16,7,5},{0,8,86},
+        {0,8,22},{64,8,0},{19,7,51},{0,8,118},{0,8,54},{0,9,204},{17,7,15},
+        {0,8,102},{0,8,38},{0,9,172},{0,8,6},{0,8,134},{0,8,70},{0,9,236},
+        {16,7,9},{0,8,94},{0,8,30},{0,9,156},{20,7,99},{0,8,126},{0,8,62},
+        {0,9,220},{18,7,27},{0,8,110},{0,8,46},{0,9,188},{0,8,14},{0,8,142},
+        {0,8,78},{0,9,252},{96,7,0},{0,8,81},{0,8,17},{21,8,131},{18,7,31},
+        {0,8,113},{0,8,49},{0,9,194},{16,7,10},{0,8,97},{0,8,33},{0,9,162},
+        {0,8,1},{0,8,129},{0,8,65},{0,9,226},{16,7,6},{0,8,89},{0,8,25},
+        {0,9,146},{19,7,59},{0,8,121},{0,8,57},{0,9,210},{17,7,17},{0,8,105},
+        {0,8,41},{0,9,178},{0,8,9},{0,8,137},{0,8,73},{0,9,242},{16,7,4},
+        {0,8,85},{0,8,21},{16,8,258},{19,7,43},{0,8,117},{0,8,53},{0,9,202},
+        {17,7,13},{0,8,101},{0,8,37},{0,9,170},{0,8,5},{0,8,133},{0,8,69},
+        {0,9,234},{16,7,8},{0,8,93},{0,8,29},{0,9,154},{20,7,83},{0,8,125},
+        {0,8,61},{0,9,218},{18,7,23},{0,8,109},{0,8,45},{0,9,186},{0,8,13},
+        {0,8,141},{0,8,77},{0,9,250},{16,7,3},{0,8,83},{0,8,19},{21,8,195},
+        {19,7,35},{0,8,115},{0,8,51},{0,9,198},{17,7,11},{0,8,99},{0,8,35},
+        {0,9,166},{0,8,3},{0,8,131},{0,8,67},{0,9,230},{16,7,7},{0,8,91},
+        {0,8,27},{0,9,150},{20,7,67},{0,8,123},{0,8,59},{0,9,214},{18,7,19},
+        {0,8,107},{0,8,43},{0,9,182},{0,8,11},{0,8,139},{0,8,75},{0,9,246},
+        {16,7,5},{0,8,87},{0,8,23},{64,8,0},{19,7,51},{0,8,119},{0,8,55},
+        {0,9,206},{17,7,15},{0,8,103},{0,8,39},{0,9,174},{0,8,7},{0,8,135},
+        {0,8,71},{0,9,238},{16,7,9},{0,8,95},{0,8,31},{0,9,158},{20,7,99},
+        {0,8,127},{0,8,63},{0,9,222},{18,7,27},{0,8,111},{0,8,47},{0,9,190},
+        {0,8,15},{0,8,143},{0,8,79},{0,9,254},{96,7,0},{0,8,80},{0,8,16},
+        {20,8,115},{18,7,31},{0,8,112},{0,8,48},{0,9,193},{16,7,10},{0,8,96},
+        {0,8,32},{0,9,161},{0,8,0},{0,8,128},{0,8,64},{0,9,225},{16,7,6},
+        {0,8,88},{0,8,24},{0,9,145},{19,7,59},{0,8,120},{0,8,56},{0,9,209},
+        {17,7,17},{0,8,104},{0,8,40},{0,9,177},{0,8,8},{0,8,136},{0,8,72},
+        {0,9,241},{16,7,4},{0,8,84},{0,8,20},{21,8,227},{19,7,43},{0,8,116},
+        {0,8,52},{0,9,201},{17,7,13},{0,8,100},{0,8,36},{0,9,169},{0,8,4},
+        {0,8,132},{0,8,68},{0,9,233},{16,7,8},{0,8,92},{0,8,28},{0,9,153},
+        {20,7,83},{0,8,124},{0,8,60},{0,9,217},{18,7,23},{0,8,108},{0,8,44},
+        {0,9,185},{0,8,12},{0,8,140},{0,8,76},{0,9,249},{16,7,3},{0,8,82},
+        {0,8,18},{21,8,163},{19,7,35},{0,8,114},{0,8,50},{0,9,197},{17,7,11},
+        {0,8,98},{0,8,34},{0,9,165},{0,8,2},{0,8,130},{0,8,66},{0,9,229},
+        {16,7,7},{0,8,90},{0,8,26},{0,9,149},{20,7,67},{0,8,122},{0,8,58},
+        {0,9,213},{18,7,19},{0,8,106},{0,8,42},{0,9,181},{0,8,10},{0,8,138},
+        {0,8,74},{0,9,245},{16,7,5},{0,8,86},{0,8,22},{64,8,0},{19,7,51},
+        {0,8,118},{0,8,54},{0,9,205},{17,7,15},{0,8,102},{0,8,38},{0,9,173},
+        {0,8,6},{0,8,134},{0,8,70},{0,9,237},{16,7,9},{0,8,94},{0,8,30},
+        {0,9,157},{20,7,99},{0,8,126},{0,8,62},{0,9,221},{18,7,27},{0,8,110},
+        {0,8,46},{0,9,189},{0,8,14},{0,8,142},{0,8,78},{0,9,253},{96,7,0},
+        {0,8,81},{0,8,17},{21,8,131},{18,7,31},{0,8,113},{0,8,49},{0,9,195},
+        {16,7,10},{0,8,97},{0,8,33},{0,9,163},{0,8,1},{0,8,129},{0,8,65},
+        {0,9,227},{16,7,6},{0,8,89},{0,8,25},{0,9,147},{19,7,59},{0,8,121},
+        {0,8,57},{0,9,211},{17,7,17},{0,8,105},{0,8,41},{0,9,179},{0,8,9},
+        {0,8,137},{0,8,73},{0,9,243},{16,7,4},{0,8,85},{0,8,21},{16,8,258},
+        {19,7,43},{0,8,117},{0,8,53},{0,9,203},{17,7,13},{0,8,101},{0,8,37},
+        {0,9,171},{0,8,5},{0,8,133},{0,8,69},{0,9,235},{16,7,8},{0,8,93},
+        {0,8,29},{0,9,155},{20,7,83},{0,8,125},{0,8,61},{0,9,219},{18,7,23},
+        {0,8,109},{0,8,45},{0,9,187},{0,8,13},{0,8,141},{0,8,77},{0,9,251},
+        {16,7,3},{0,8,83},{0,8,19},{21,8,195},{19,7,35},{0,8,115},{0,8,51},
+        {0,9,199},{17,7,11},{0,8,99},{0,8,35},{0,9,167},{0,8,3},{0,8,131},
+        {0,8,67},{0,9,231},{16,7,7},{0,8,91},{0,8,27},{0,9,151},{20,7,67},
+        {0,8,123},{0,8,59},{0,9,215},{18,7,19},{0,8,107},{0,8,43},{0,9,183},
+        {0,8,11},{0,8,139},{0,8,75},{0,9,247},{16,7,5},{0,8,87},{0,8,23},
+        {64,8,0},{19,7,51},{0,8,119},{0,8,55},{0,9,207},{17,7,15},{0,8,103},
+        {0,8,39},{0,9,175},{0,8,7},{0,8,135},{0,8,71},{0,9,239},{16,7,9},
+        {0,8,95},{0,8,31},{0,9,159},{20,7,99},{0,8,127},{0,8,63},{0,9,223},
+        {18,7,27},{0,8,111},{0,8,47},{0,9,191},{0,8,15},{0,8,143},{0,8,79},
+        {0,9,255}
+    };
+
+    static const code distfix[32] = {
+        {16,5,1},{23,5,257},{19,5,17},{27,5,4097},{17,5,5},{25,5,1025},
+        {21,5,65},{29,5,16385},{16,5,3},{24,5,513},{20,5,33},{28,5,8193},
+        {18,5,9},{26,5,2049},{22,5,129},{64,5,0},{16,5,2},{23,5,385},
+        {19,5,25},{27,5,6145},{17,5,7},{25,5,1537},{21,5,97},{29,5,24577},
+        {16,5,4},{24,5,769},{20,5,49},{28,5,12289},{18,5,13},{26,5,3073},
+        {22,5,193},{64,5,0}
+    };
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inflate.c glmark2-2021.02/src/zlib/inflate.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inflate.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inflate.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1561 @@
+/* inflate.c -- zlib decompression
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ * Change history:
+ *
+ * 1.2.beta0    24 Nov 2002
+ * - First version -- complete rewrite of inflate to simplify code, avoid
+ *   creation of window when not needed, minimize use of window when it is
+ *   needed, make inffast.c even faster, implement gzip decoding, and to
+ *   improve code readability and style over the previous zlib inflate code
+ *
+ * 1.2.beta1    25 Nov 2002
+ * - Use pointers for available input and output checking in inffast.c
+ * - Remove input and output counters in inffast.c
+ * - Change inffast.c entry and loop from avail_in >= 7 to >= 6
+ * - Remove unnecessary second byte pull from length extra in inffast.c
+ * - Unroll direct copy to three copies per loop in inffast.c
+ *
+ * 1.2.beta2    4 Dec 2002
+ * - Change external routine names to reduce potential conflicts
+ * - Correct filename to inffixed.h for fixed tables in inflate.c
+ * - Make hbuf[] unsigned char to match parameter type in inflate.c
+ * - Change strm->next_out[-state->offset] to *(strm->next_out - state->offset)
+ *   to avoid negation problem on Alphas (64 bit) in inflate.c
+ *
+ * 1.2.beta3    22 Dec 2002
+ * - Add comments on state->bits assertion in inffast.c
+ * - Add comments on op field in inftrees.h
+ * - Fix bug in reuse of allocated window after inflateReset()
+ * - Remove bit fields--back to byte structure for speed
+ * - Remove distance extra == 0 check in inflate_fast()--only helps for lengths
+ * - Change post-increments to pre-increments in inflate_fast(), PPC biased?
+ * - Add compile time option, POSTINC, to use post-increments instead (Intel?)
+ * - Make MATCH copy in inflate() much faster for when inflate_fast() not used
+ * - Use local copies of stream next and avail values, as well as local bit
+ *   buffer and bit count in inflate()--for speed when inflate_fast() not used
+ *
+ * 1.2.beta4    1 Jan 2003
+ * - Split ptr - 257 statements in inflate_table() to avoid compiler warnings
+ * - Move a comment on output buffer sizes from inffast.c to inflate.c
+ * - Add comments in inffast.c to introduce the inflate_fast() routine
+ * - Rearrange window copies in inflate_fast() for speed and simplification
+ * - Unroll last copy for window match in inflate_fast()
+ * - Use local copies of window variables in inflate_fast() for speed
+ * - Pull out common wnext == 0 case for speed in inflate_fast()
+ * - Make op and len in inflate_fast() unsigned for consistency
+ * - Add FAR to lcode and dcode declarations in inflate_fast()
+ * - Simplified bad distance check in inflate_fast()
+ * - Added inflateBackInit(), inflateBack(), and inflateBackEnd() in new
+ *   source file infback.c to provide a call-back interface to inflate for
+ *   programs like gzip and unzip -- uses window as output buffer to avoid
+ *   window copying
+ *
+ * 1.2.beta5    1 Jan 2003
+ * - Improved inflateBack() interface to allow the caller to provide initial
+ *   input in strm.
+ * - Fixed stored blocks bug in inflateBack()
+ *
+ * 1.2.beta6    4 Jan 2003
+ * - Added comments in inffast.c on effectiveness of POSTINC
+ * - Typecasting all around to reduce compiler warnings
+ * - Changed loops from while (1) or do {} while (1) to for (;;), again to
+ *   make compilers happy
+ * - Changed type of window in inflateBackInit() to unsigned char *
+ *
+ * 1.2.beta7    27 Jan 2003
+ * - Changed many types to unsigned or unsigned short to avoid warnings
+ * - Added inflateCopy() function
+ *
+ * 1.2.0        9 Mar 2003
+ * - Changed inflateBack() interface to provide separate opaque descriptors
+ *   for the in() and out() functions
+ * - Changed inflateBack() argument and in_func typedef to swap the length
+ *   and buffer address return values for the input function
+ * - Check next_in and next_out for Z_NULL on entry to inflate()
+ *
+ * The history for versions after 1.2.0 are in ChangeLog in zlib distribution.
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+#include "inflate.h"
+#include "inffast.h"
+
+#ifdef MAKEFIXED
+#  ifndef BUILDFIXED
+#    define BUILDFIXED
+#  endif
+#endif
+
+/* function prototypes */
+local int inflateStateCheck OF((z_streamp strm));
+local void fixedtables OF((struct inflate_state FAR *state));
+local int updatewindow OF((z_streamp strm, const unsigned char FAR *end,
+                           unsigned copy));
+#ifdef BUILDFIXED
+   void makefixed OF((void));
+#endif
+local unsigned syncsearch OF((unsigned FAR *have, const unsigned char FAR *buf,
+                              unsigned len));
+
+local int inflateStateCheck(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (strm == Z_NULL ||
+        strm->zalloc == (alloc_func)0 || strm->zfree == (free_func)0)
+        return 1;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state == Z_NULL || state->strm != strm ||
+        state->mode < HEAD || state->mode > SYNC)
+        return 1;
+    return 0;
+}
+
+int ZEXPORT inflateResetKeep(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    strm->total_in = strm->total_out = state->total = 0;
+    strm->msg = Z_NULL;
+    if (state->wrap)        /* to support ill-conceived Java test suite */
+        strm->adler = state->wrap & 1;
+    state->mode = HEAD;
+    state->last = 0;
+    state->havedict = 0;
+    state->dmax = 32768U;
+    state->head = Z_NULL;
+    state->hold = 0;
+    state->bits = 0;
+    state->lencode = state->distcode = state->next = state->codes;
+    state->sane = 1;
+    state->back = -1;
+    Tracev((stderr, "inflate: reset\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateReset(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    state->wsize = 0;
+    state->whave = 0;
+    state->wnext = 0;
+    return inflateResetKeep(strm);
+}
+
+int ZEXPORT inflateReset2(strm, windowBits)
+z_streamp strm;
+int windowBits;
+{
+    int wrap;
+    struct inflate_state FAR *state;
+
+    /* get the state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* extract wrap request from windowBits parameter */
+    if (windowBits < 0) {
+        wrap = 0;
+        windowBits = -windowBits;
+    }
+    else {
+        wrap = (windowBits >> 4) + 5;
+#ifdef GUNZIP
+        if (windowBits < 48)
+            windowBits &= 15;
+#endif
+    }
+
+    /* set number of window bits, free window if different */
+    if (windowBits && (windowBits < 8 || windowBits > 15))
+        return Z_STREAM_ERROR;
+    if (state->window != Z_NULL && state->wbits != (unsigned)windowBits) {
+        ZFREE(strm, state->window);
+        state->window = Z_NULL;
+    }
+
+    /* update state and reset the rest of it */
+    state->wrap = wrap;
+    state->wbits = (unsigned)windowBits;
+    return inflateReset(strm);
+}
+
+int ZEXPORT inflateInit2_(strm, windowBits, version, stream_size)
+z_streamp strm;
+int windowBits;
+const char *version;
+int stream_size;
+{
+    int ret;
+    struct inflate_state FAR *state;
+
+    if (version == Z_NULL || version[0] != ZLIB_VERSION[0] ||
+        stream_size != (int)(sizeof(z_stream)))
+        return Z_VERSION_ERROR;
+    if (strm == Z_NULL) return Z_STREAM_ERROR;
+    strm->msg = Z_NULL;                 /* in case we return an error */
+    if (strm->zalloc == (alloc_func)0) {
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zalloc = zcalloc;
+        strm->opaque = (voidpf)0;
+#endif
+    }
+    if (strm->zfree == (free_func)0)
+#ifdef Z_SOLO
+        return Z_STREAM_ERROR;
+#else
+        strm->zfree = zcfree;
+#endif
+    state = (struct inflate_state FAR *)
+            ZALLOC(strm, 1, sizeof(struct inflate_state));
+    if (state == Z_NULL) return Z_MEM_ERROR;
+    Tracev((stderr, "inflate: allocated\n"));
+    strm->state = (struct internal_state FAR *)state;
+    state->strm = strm;
+    state->window = Z_NULL;
+    state->mode = HEAD;     /* to pass state test in inflateReset2() */
+    ret = inflateReset2(strm, windowBits);
+    if (ret != Z_OK) {
+        ZFREE(strm, state);
+        strm->state = Z_NULL;
+    }
+    return ret;
+}
+
+int ZEXPORT inflateInit_(strm, version, stream_size)
+z_streamp strm;
+const char *version;
+int stream_size;
+{
+    return inflateInit2_(strm, DEF_WBITS, version, stream_size);
+}
+
+int ZEXPORT inflatePrime(strm, bits, value)
+z_streamp strm;
+int bits;
+int value;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (bits < 0) {
+        state->hold = 0;
+        state->bits = 0;
+        return Z_OK;
+    }
+    if (bits > 16 || state->bits + (uInt)bits > 32) return Z_STREAM_ERROR;
+    value &= (1L << bits) - 1;
+    state->hold += (unsigned)value << state->bits;
+    state->bits += (uInt)bits;
+    return Z_OK;
+}
+
+/*
+   Return state with length and distance decoding tables and index sizes set to
+   fixed code decoding.  Normally this returns fixed tables from inffixed.h.
+   If BUILDFIXED is defined, then instead this routine builds the tables the
+   first time it's called, and returns those tables the first time and
+   thereafter.  This reduces the size of the code by about 2K bytes, in
+   exchange for a little execution time.  However, BUILDFIXED should not be
+   used for threaded applications, since the rewriting of the tables and virgin
+   may not be thread-safe.
+ */
+local void fixedtables(state)
+struct inflate_state FAR *state;
+{
+#ifdef BUILDFIXED
+    static int virgin = 1;
+    static code *lenfix, *distfix;
+    static code fixed[544];
+
+    /* build fixed huffman tables if first call (may not be thread safe) */
+    if (virgin) {
+        unsigned sym, bits;
+        static code *next;
+
+        /* literal/length table */
+        sym = 0;
+        while (sym < 144) state->lens[sym++] = 8;
+        while (sym < 256) state->lens[sym++] = 9;
+        while (sym < 280) state->lens[sym++] = 7;
+        while (sym < 288) state->lens[sym++] = 8;
+        next = fixed;
+        lenfix = next;
+        bits = 9;
+        inflate_table(LENS, state->lens, 288, &(next), &(bits), state->work);
+
+        /* distance table */
+        sym = 0;
+        while (sym < 32) state->lens[sym++] = 5;
+        distfix = next;
+        bits = 5;
+        inflate_table(DISTS, state->lens, 32, &(next), &(bits), state->work);
+
+        /* do this just once */
+        virgin = 0;
+    }
+#else /* !BUILDFIXED */
+#   include "inffixed.h"
+#endif /* BUILDFIXED */
+    state->lencode = lenfix;
+    state->lenbits = 9;
+    state->distcode = distfix;
+    state->distbits = 5;
+}
+
+#ifdef MAKEFIXED
+#include <stdio.h>
+
+/*
+   Write out the inffixed.h that is #include'd above.  Defining MAKEFIXED also
+   defines BUILDFIXED, so the tables are built on the fly.  makefixed() writes
+   those tables to stdout, which would be piped to inffixed.h.  A small program
+   can simply call makefixed to do this:
+
+    void makefixed(void);
+
+    int main(void)
+    {
+        makefixed();
+        return 0;
+    }
+
+   Then that can be linked with zlib built with MAKEFIXED defined and run:
+
+    a.out > inffixed.h
+ */
+void makefixed()
+{
+    unsigned low, size;
+    struct inflate_state state;
+
+    fixedtables(&state);
+    puts("    /* inffixed.h -- table for decoding fixed codes");
+    puts("     * Generated automatically by makefixed().");
+    puts("     */");
+    puts("");
+    puts("    /* WARNING: this file should *not* be used by applications.");
+    puts("       It is part of the implementation of this library and is");
+    puts("       subject to change. Applications should only use zlib.h.");
+    puts("     */");
+    puts("");
+    size = 1U << 9;
+    printf("    static const code lenfix[%u] = {", size);
+    low = 0;
+    for (;;) {
+        if ((low % 7) == 0) printf("\n        ");
+        printf("{%u,%u,%d}", (low & 127) == 99 ? 64 : state.lencode[low].op,
+               state.lencode[low].bits, state.lencode[low].val);
+        if (++low == size) break;
+        putchar(',');
+    }
+    puts("\n    };");
+    size = 1U << 5;
+    printf("\n    static const code distfix[%u] = {", size);
+    low = 0;
+    for (;;) {
+        if ((low % 6) == 0) printf("\n        ");
+        printf("{%u,%u,%d}", state.distcode[low].op, state.distcode[low].bits,
+               state.distcode[low].val);
+        if (++low == size) break;
+        putchar(',');
+    }
+    puts("\n    };");
+}
+#endif /* MAKEFIXED */
+
+/*
+   Update the window with the last wsize (normally 32K) bytes written before
+   returning.  If window does not exist yet, create it.  This is only called
+   when a window is already in use, or when output has been written during this
+   inflate call, but the end of the deflate stream has not been reached yet.
+   It is also called to create a window for dictionary data when a dictionary
+   is loaded.
+
+   Providing output buffers larger than 32K to inflate() should provide a speed
+   advantage, since only the last 32K of output is copied to the sliding window
+   upon return from inflate(), and since all distances after the first 32K of
+   output will fall in the output data, making match copies simpler and faster.
+   The advantage may be dependent on the size of the processor's data caches.
+ */
+local int updatewindow(strm, end, copy)
+z_streamp strm;
+const Bytef *end;
+unsigned copy;
+{
+    struct inflate_state FAR *state;
+    unsigned dist;
+
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* if it hasn't been done already, allocate space for the window */
+    if (state->window == Z_NULL) {
+        state->window = (unsigned char FAR *)
+                        ZALLOC(strm, 1U << state->wbits,
+                               sizeof(unsigned char));
+        if (state->window == Z_NULL) return 1;
+    }
+
+    /* if window not in use yet, initialize */
+    if (state->wsize == 0) {
+        state->wsize = 1U << state->wbits;
+        state->wnext = 0;
+        state->whave = 0;
+    }
+
+    /* copy state->wsize or less output bytes into the circular window */
+    if (copy >= state->wsize) {
+        zmemcpy(state->window, end - state->wsize, state->wsize);
+        state->wnext = 0;
+        state->whave = state->wsize;
+    }
+    else {
+        dist = state->wsize - state->wnext;
+        if (dist > copy) dist = copy;
+        zmemcpy(state->window + state->wnext, end - copy, dist);
+        copy -= dist;
+        if (copy) {
+            zmemcpy(state->window, end - copy, copy);
+            state->wnext = copy;
+            state->whave = state->wsize;
+        }
+        else {
+            state->wnext += dist;
+            if (state->wnext == state->wsize) state->wnext = 0;
+            if (state->whave < state->wsize) state->whave += dist;
+        }
+    }
+    return 0;
+}
+
+/* Macros for inflate(): */
+
+/* check function to use adler32() for zlib or crc32() for gzip */
+#ifdef GUNZIP
+#  define UPDATE(check, buf, len) \
+    (state->flags ? crc32(check, buf, len) : adler32(check, buf, len))
+#else
+#  define UPDATE(check, buf, len) adler32(check, buf, len)
+#endif
+
+/* check macros for header crc */
+#ifdef GUNZIP
+#  define CRC2(check, word) \
+    do { \
+        hbuf[0] = (unsigned char)(word); \
+        hbuf[1] = (unsigned char)((word) >> 8); \
+        check = crc32(check, hbuf, 2); \
+    } while (0)
+
+#  define CRC4(check, word) \
+    do { \
+        hbuf[0] = (unsigned char)(word); \
+        hbuf[1] = (unsigned char)((word) >> 8); \
+        hbuf[2] = (unsigned char)((word) >> 16); \
+        hbuf[3] = (unsigned char)((word) >> 24); \
+        check = crc32(check, hbuf, 4); \
+    } while (0)
+#endif
+
+/* Load registers with state in inflate() for speed */
+#define LOAD() \
+    do { \
+        put = strm->next_out; \
+        left = strm->avail_out; \
+        next = strm->next_in; \
+        have = strm->avail_in; \
+        hold = state->hold; \
+        bits = state->bits; \
+    } while (0)
+
+/* Restore state from registers in inflate() */
+#define RESTORE() \
+    do { \
+        strm->next_out = put; \
+        strm->avail_out = left; \
+        strm->next_in = next; \
+        strm->avail_in = have; \
+        state->hold = hold; \
+        state->bits = bits; \
+    } while (0)
+
+/* Clear the input bit accumulator */
+#define INITBITS() \
+    do { \
+        hold = 0; \
+        bits = 0; \
+    } while (0)
+
+/* Get a byte of input into the bit accumulator, or return from inflate()
+   if there is no input available. */
+#define PULLBYTE() \
+    do { \
+        if (have == 0) goto inf_leave; \
+        have--; \
+        hold += (unsigned long)(*next++) << bits; \
+        bits += 8; \
+    } while (0)
+
+/* Assure that there are at least n bits in the bit accumulator.  If there is
+   not enough available input to do that, then return from inflate(). */
+#define NEEDBITS(n) \
+    do { \
+        while (bits < (unsigned)(n)) \
+            PULLBYTE(); \
+    } while (0)
+
+/* Return the low n bits of the bit accumulator (n < 16) */
+#define BITS(n) \
+    ((unsigned)hold & ((1U << (n)) - 1))
+
+/* Remove n bits from the bit accumulator */
+#define DROPBITS(n) \
+    do { \
+        hold >>= (n); \
+        bits -= (unsigned)(n); \
+    } while (0)
+
+/* Remove zero to seven bits as needed to go to a byte boundary */
+#define BYTEBITS() \
+    do { \
+        hold >>= bits & 7; \
+        bits -= bits & 7; \
+    } while (0)
+
+/*
+   inflate() uses a state machine to process as much input data and generate as
+   much output data as possible before returning.  The state machine is
+   structured roughly as follows:
+
+    for (;;) switch (state) {
+    ...
+    case STATEn:
+        if (not enough input data or output space to make progress)
+            return;
+        ... make progress ...
+        state = STATEm;
+        break;
+    ...
+    }
+
+   so when inflate() is called again, the same case is attempted again, and
+   if the appropriate resources are provided, the machine proceeds to the
+   next state.  The NEEDBITS() macro is usually the way the state evaluates
+   whether it can proceed or should return.  NEEDBITS() does the return if
+   the requested bits are not available.  The typical use of the BITS macros
+   is:
+
+        NEEDBITS(n);
+        ... do something with BITS(n) ...
+        DROPBITS(n);
+
+   where NEEDBITS(n) either returns from inflate() if there isn't enough
+   input left to load n bits into the accumulator, or it continues.  BITS(n)
+   gives the low n bits in the accumulator.  When done, DROPBITS(n) drops
+   the low n bits off the accumulator.  INITBITS() clears the accumulator
+   and sets the number of available bits to zero.  BYTEBITS() discards just
+   enough bits to put the accumulator on a byte boundary.  After BYTEBITS()
+   and a NEEDBITS(8), then BITS(8) would return the next byte in the stream.
+
+   NEEDBITS(n) uses PULLBYTE() to get an available byte of input, or to return
+   if there is no input available.  The decoding of variable length codes uses
+   PULLBYTE() directly in order to pull just enough bytes to decode the next
+   code, and no more.
+
+   Some states loop until they get enough input, making sure that enough
+   state information is maintained to continue the loop where it left off
+   if NEEDBITS() returns in the loop.  For example, want, need, and keep
+   would all have to actually be part of the saved state in case NEEDBITS()
+   returns:
+
+    case STATEw:
+        while (want < need) {
+            NEEDBITS(n);
+            keep[want++] = BITS(n);
+            DROPBITS(n);
+        }
+        state = STATEx;
+    case STATEx:
+
+   As shown above, if the next state is also the next case, then the break
+   is omitted.
+
+   A state may also return if there is not enough output space available to
+   complete that state.  Those states are copying stored data, writing a
+   literal byte, and copying a matching string.
+
+   When returning, a "goto inf_leave" is used to update the total counters,
+   update the check value, and determine whether any progress has been made
+   during that inflate() call in order to return the proper return code.
+   Progress is defined as a change in either strm->avail_in or strm->avail_out.
+   When there is a window, goto inf_leave will update the window with the last
+   output written.  If a goto inf_leave occurs in the middle of decompression
+   and there is no window currently, goto inf_leave will create one and copy
+   output to the window for the next call of inflate().
+
+   In this implementation, the flush parameter of inflate() only affects the
+   return code (per zlib.h).  inflate() always writes as much as possible to
+   strm->next_out, given the space available and the provided input--the effect
+   documented in zlib.h of Z_SYNC_FLUSH.  Furthermore, inflate() always defers
+   the allocation of and copying into a sliding window until necessary, which
+   provides the effect documented in zlib.h for Z_FINISH when the entire input
+   stream available.  So the only thing the flush parameter actually does is:
+   when flush is set to Z_FINISH, inflate() cannot return Z_OK.  Instead it
+   will return Z_BUF_ERROR if it has not reached the end of the stream.
+ */
+
+int ZEXPORT inflate(strm, flush)
+z_streamp strm;
+int flush;
+{
+    struct inflate_state FAR *state;
+    z_const unsigned char FAR *next;    /* next input */
+    unsigned char FAR *put;     /* next output */
+    unsigned have, left;        /* available input and output */
+    unsigned long hold;         /* bit buffer */
+    unsigned bits;              /* bits in bit buffer */
+    unsigned in, out;           /* save starting available input and output */
+    unsigned copy;              /* number of stored or match bytes to copy */
+    unsigned char FAR *from;    /* where to copy match bytes from */
+    code here;                  /* current decoding table entry */
+    code last;                  /* parent table entry */
+    unsigned len;               /* length to copy for repeats, bits to drop */
+    int ret;                    /* return code */
+#ifdef GUNZIP
+    unsigned char hbuf[4];      /* buffer for gzip header crc calculation */
+#endif
+    static const unsigned short order[19] = /* permutation of code lengths */
+        {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
+
+    if (inflateStateCheck(strm) || strm->next_out == Z_NULL ||
+        (strm->next_in == Z_NULL && strm->avail_in != 0))
+        return Z_STREAM_ERROR;
+
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->mode == TYPE) state->mode = TYPEDO;      /* skip check */
+    LOAD();
+    in = have;
+    out = left;
+    ret = Z_OK;
+    for (;;)
+        switch (state->mode) {
+        case HEAD:
+            if (state->wrap == 0) {
+                state->mode = TYPEDO;
+                break;
+            }
+            NEEDBITS(16);
+#ifdef GUNZIP
+            if ((state->wrap & 2) && hold == 0x8b1f) {  /* gzip header */
+                if (state->wbits == 0)
+                    state->wbits = 15;
+                state->check = crc32(0L, Z_NULL, 0);
+                CRC2(state->check, hold);
+                INITBITS();
+                state->mode = FLAGS;
+                break;
+            }
+            state->flags = 0;           /* expect zlib header */
+            if (state->head != Z_NULL)
+                state->head->done = -1;
+            if (!(state->wrap & 1) ||   /* check if zlib header allowed */
+#else
+            if (
+#endif
+                ((BITS(8) << 8) + (hold >> 8)) % 31) {
+                strm->msg = (char *)"incorrect header check";
+                state->mode = BAD;
+                break;
+            }
+            if (BITS(4) != Z_DEFLATED) {
+                strm->msg = (char *)"unknown compression method";
+                state->mode = BAD;
+                break;
+            }
+            DROPBITS(4);
+            len = BITS(4) + 8;
+            if (state->wbits == 0)
+                state->wbits = len;
+            if (len > 15 || len > state->wbits) {
+                strm->msg = (char *)"invalid window size";
+                state->mode = BAD;
+                break;
+            }
+            state->dmax = 1U << len;
+            Tracev((stderr, "inflate:   zlib header ok\n"));
+            strm->adler = state->check = adler32(0L, Z_NULL, 0);
+            state->mode = hold & 0x200 ? DICTID : TYPE;
+            INITBITS();
+            break;
+#ifdef GUNZIP
+        case FLAGS:
+            NEEDBITS(16);
+            state->flags = (int)(hold);
+            if ((state->flags & 0xff) != Z_DEFLATED) {
+                strm->msg = (char *)"unknown compression method";
+                state->mode = BAD;
+                break;
+            }
+            if (state->flags & 0xe000) {
+                strm->msg = (char *)"unknown header flags set";
+                state->mode = BAD;
+                break;
+            }
+            if (state->head != Z_NULL)
+                state->head->text = (int)((hold >> 8) & 1);
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC2(state->check, hold);
+            INITBITS();
+            state->mode = TIME;
+        case TIME:
+            NEEDBITS(32);
+            if (state->head != Z_NULL)
+                state->head->time = hold;
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC4(state->check, hold);
+            INITBITS();
+            state->mode = OS;
+        case OS:
+            NEEDBITS(16);
+            if (state->head != Z_NULL) {
+                state->head->xflags = (int)(hold & 0xff);
+                state->head->os = (int)(hold >> 8);
+            }
+            if ((state->flags & 0x0200) && (state->wrap & 4))
+                CRC2(state->check, hold);
+            INITBITS();
+            state->mode = EXLEN;
+        case EXLEN:
+            if (state->flags & 0x0400) {
+                NEEDBITS(16);
+                state->length = (unsigned)(hold);
+                if (state->head != Z_NULL)
+                    state->head->extra_len = (unsigned)hold;
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    CRC2(state->check, hold);
+                INITBITS();
+            }
+            else if (state->head != Z_NULL)
+                state->head->extra = Z_NULL;
+            state->mode = EXTRA;
+        case EXTRA:
+            if (state->flags & 0x0400) {
+                copy = state->length;
+                if (copy > have) copy = have;
+                if (copy) {
+                    if (state->head != Z_NULL &&
+                        state->head->extra != Z_NULL) {
+                        len = state->head->extra_len - state->length;
+                        zmemcpy(state->head->extra + len, next,
+                                len + copy > state->head->extra_max ?
+                                state->head->extra_max - len : copy);
+                    }
+                    if ((state->flags & 0x0200) && (state->wrap & 4))
+                        state->check = crc32(state->check, next, copy);
+                    have -= copy;
+                    next += copy;
+                    state->length -= copy;
+                }
+                if (state->length) goto inf_leave;
+            }
+            state->length = 0;
+            state->mode = NAME;
+        case NAME:
+            if (state->flags & 0x0800) {
+                if (have == 0) goto inf_leave;
+                copy = 0;
+                do {
+                    len = (unsigned)(next[copy++]);
+                    if (state->head != Z_NULL &&
+                            state->head->name != Z_NULL &&
+                            state->length < state->head->name_max)
+                        state->head->name[state->length++] = (Bytef)len;
+                } while (len && copy < have);
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    state->check = crc32(state->check, next, copy);
+                have -= copy;
+                next += copy;
+                if (len) goto inf_leave;
+            }
+            else if (state->head != Z_NULL)
+                state->head->name = Z_NULL;
+            state->length = 0;
+            state->mode = COMMENT;
+        case COMMENT:
+            if (state->flags & 0x1000) {
+                if (have == 0) goto inf_leave;
+                copy = 0;
+                do {
+                    len = (unsigned)(next[copy++]);
+                    if (state->head != Z_NULL &&
+                            state->head->comment != Z_NULL &&
+                            state->length < state->head->comm_max)
+                        state->head->comment[state->length++] = (Bytef)len;
+                } while (len && copy < have);
+                if ((state->flags & 0x0200) && (state->wrap & 4))
+                    state->check = crc32(state->check, next, copy);
+                have -= copy;
+                next += copy;
+                if (len) goto inf_leave;
+            }
+            else if (state->head != Z_NULL)
+                state->head->comment = Z_NULL;
+            state->mode = HCRC;
+        case HCRC:
+            if (state->flags & 0x0200) {
+                NEEDBITS(16);
+                if ((state->wrap & 4) && hold != (state->check & 0xffff)) {
+                    strm->msg = (char *)"header crc mismatch";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+            }
+            if (state->head != Z_NULL) {
+                state->head->hcrc = (int)((state->flags >> 9) & 1);
+                state->head->done = 1;
+            }
+            strm->adler = state->check = crc32(0L, Z_NULL, 0);
+            state->mode = TYPE;
+            break;
+#endif
+        case DICTID:
+            NEEDBITS(32);
+            strm->adler = state->check = ZSWAP32(hold);
+            INITBITS();
+            state->mode = DICT;
+        case DICT:
+            if (state->havedict == 0) {
+                RESTORE();
+                return Z_NEED_DICT;
+            }
+            strm->adler = state->check = adler32(0L, Z_NULL, 0);
+            state->mode = TYPE;
+        case TYPE:
+            if (flush == Z_BLOCK || flush == Z_TREES) goto inf_leave;
+        case TYPEDO:
+            if (state->last) {
+                BYTEBITS();
+                state->mode = CHECK;
+                break;
+            }
+            NEEDBITS(3);
+            state->last = BITS(1);
+            DROPBITS(1);
+            switch (BITS(2)) {
+            case 0:                             /* stored block */
+                Tracev((stderr, "inflate:     stored block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = STORED;
+                break;
+            case 1:                             /* fixed block */
+                fixedtables(state);
+                Tracev((stderr, "inflate:     fixed codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = LEN_;             /* decode codes */
+                if (flush == Z_TREES) {
+                    DROPBITS(2);
+                    goto inf_leave;
+                }
+                break;
+            case 2:                             /* dynamic block */
+                Tracev((stderr, "inflate:     dynamic codes block%s\n",
+                        state->last ? " (last)" : ""));
+                state->mode = TABLE;
+                break;
+            case 3:
+                strm->msg = (char *)"invalid block type";
+                state->mode = BAD;
+            }
+            DROPBITS(2);
+            break;
+        case STORED:
+            BYTEBITS();                         /* go to byte boundary */
+            NEEDBITS(32);
+            if ((hold & 0xffff) != ((hold >> 16) ^ 0xffff)) {
+                strm->msg = (char *)"invalid stored block lengths";
+                state->mode = BAD;
+                break;
+            }
+            state->length = (unsigned)hold & 0xffff;
+            Tracev((stderr, "inflate:       stored length %u\n",
+                    state->length));
+            INITBITS();
+            state->mode = COPY_;
+            if (flush == Z_TREES) goto inf_leave;
+        case COPY_:
+            state->mode = COPY;
+        case COPY:
+            copy = state->length;
+            if (copy) {
+                if (copy > have) copy = have;
+                if (copy > left) copy = left;
+                if (copy == 0) goto inf_leave;
+                zmemcpy(put, next, copy);
+                have -= copy;
+                next += copy;
+                left -= copy;
+                put += copy;
+                state->length -= copy;
+                break;
+            }
+            Tracev((stderr, "inflate:       stored end\n"));
+            state->mode = TYPE;
+            break;
+        case TABLE:
+            NEEDBITS(14);
+            state->nlen = BITS(5) + 257;
+            DROPBITS(5);
+            state->ndist = BITS(5) + 1;
+            DROPBITS(5);
+            state->ncode = BITS(4) + 4;
+            DROPBITS(4);
+#ifndef PKZIP_BUG_WORKAROUND
+            if (state->nlen > 286 || state->ndist > 30) {
+                strm->msg = (char *)"too many length or distance symbols";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracev((stderr, "inflate:       table sizes ok\n"));
+            state->have = 0;
+            state->mode = LENLENS;
+        case LENLENS:
+            while (state->have < state->ncode) {
+                NEEDBITS(3);
+                state->lens[order[state->have++]] = (unsigned short)BITS(3);
+                DROPBITS(3);
+            }
+            while (state->have < 19)
+                state->lens[order[state->have++]] = 0;
+            state->next = state->codes;
+            state->lencode = (const code FAR *)(state->next);
+            state->lenbits = 7;
+            ret = inflate_table(CODES, state->lens, 19, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid code lengths set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       code lengths ok\n"));
+            state->have = 0;
+            state->mode = CODELENS;
+        case CODELENS:
+            while (state->have < state->nlen + state->ndist) {
+                for (;;) {
+                    here = state->lencode[BITS(state->lenbits)];
+                    if ((unsigned)(here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                if (here.val < 16) {
+                    DROPBITS(here.bits);
+                    state->lens[state->have++] = here.val;
+                }
+                else {
+                    if (here.val == 16) {
+                        NEEDBITS(here.bits + 2);
+                        DROPBITS(here.bits);
+                        if (state->have == 0) {
+                            strm->msg = (char *)"invalid bit length repeat";
+                            state->mode = BAD;
+                            break;
+                        }
+                        len = state->lens[state->have - 1];
+                        copy = 3 + BITS(2);
+                        DROPBITS(2);
+                    }
+                    else if (here.val == 17) {
+                        NEEDBITS(here.bits + 3);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 3 + BITS(3);
+                        DROPBITS(3);
+                    }
+                    else {
+                        NEEDBITS(here.bits + 7);
+                        DROPBITS(here.bits);
+                        len = 0;
+                        copy = 11 + BITS(7);
+                        DROPBITS(7);
+                    }
+                    if (state->have + copy > state->nlen + state->ndist) {
+                        strm->msg = (char *)"invalid bit length repeat";
+                        state->mode = BAD;
+                        break;
+                    }
+                    while (copy--)
+                        state->lens[state->have++] = (unsigned short)len;
+                }
+            }
+
+            /* handle error breaks in while */
+            if (state->mode == BAD) break;
+
+            /* check for end-of-block code (better have one) */
+            if (state->lens[256] == 0) {
+                strm->msg = (char *)"invalid code -- missing end-of-block";
+                state->mode = BAD;
+                break;
+            }
+
+            /* build code tables -- note: do not change the lenbits or distbits
+               values here (9 and 6) without reading the comments in inftrees.h
+               concerning the ENOUGH constants, which depend on those values */
+            state->next = state->codes;
+            state->lencode = (const code FAR *)(state->next);
+            state->lenbits = 9;
+            ret = inflate_table(LENS, state->lens, state->nlen, &(state->next),
+                                &(state->lenbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid literal/lengths set";
+                state->mode = BAD;
+                break;
+            }
+            state->distcode = (const code FAR *)(state->next);
+            state->distbits = 6;
+            ret = inflate_table(DISTS, state->lens + state->nlen, state->ndist,
+                            &(state->next), &(state->distbits), state->work);
+            if (ret) {
+                strm->msg = (char *)"invalid distances set";
+                state->mode = BAD;
+                break;
+            }
+            Tracev((stderr, "inflate:       codes ok\n"));
+            state->mode = LEN_;
+            if (flush == Z_TREES) goto inf_leave;
+        case LEN_:
+            state->mode = LEN;
+        case LEN:
+            if (have >= 6 && left >= 258) {
+                RESTORE();
+                inflate_fast(strm, out);
+                LOAD();
+                if (state->mode == TYPE)
+                    state->back = -1;
+                break;
+            }
+            state->back = 0;
+            for (;;) {
+                here = state->lencode[BITS(state->lenbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if (here.op && (here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->lencode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+                state->back += last.bits;
+            }
+            DROPBITS(here.bits);
+            state->back += here.bits;
+            state->length = (unsigned)here.val;
+            if ((int)(here.op) == 0) {
+                Tracevv((stderr, here.val >= 0x20 && here.val < 0x7f ?
+                        "inflate:         literal '%c'\n" :
+                        "inflate:         literal 0x%02x\n", here.val));
+                state->mode = LIT;
+                break;
+            }
+            if (here.op & 32) {
+                Tracevv((stderr, "inflate:         end of block\n"));
+                state->back = -1;
+                state->mode = TYPE;
+                break;
+            }
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid literal/length code";
+                state->mode = BAD;
+                break;
+            }
+            state->extra = (unsigned)(here.op) & 15;
+            state->mode = LENEXT;
+        case LENEXT:
+            if (state->extra) {
+                NEEDBITS(state->extra);
+                state->length += BITS(state->extra);
+                DROPBITS(state->extra);
+                state->back += state->extra;
+            }
+            Tracevv((stderr, "inflate:         length %u\n", state->length));
+            state->was = state->length;
+            state->mode = DIST;
+        case DIST:
+            for (;;) {
+                here = state->distcode[BITS(state->distbits)];
+                if ((unsigned)(here.bits) <= bits) break;
+                PULLBYTE();
+            }
+            if ((here.op & 0xf0) == 0) {
+                last = here;
+                for (;;) {
+                    here = state->distcode[last.val +
+                            (BITS(last.bits + last.op) >> last.bits)];
+                    if ((unsigned)(last.bits + here.bits) <= bits) break;
+                    PULLBYTE();
+                }
+                DROPBITS(last.bits);
+                state->back += last.bits;
+            }
+            DROPBITS(here.bits);
+            state->back += here.bits;
+            if (here.op & 64) {
+                strm->msg = (char *)"invalid distance code";
+                state->mode = BAD;
+                break;
+            }
+            state->offset = (unsigned)here.val;
+            state->extra = (unsigned)(here.op) & 15;
+            state->mode = DISTEXT;
+        case DISTEXT:
+            if (state->extra) {
+                NEEDBITS(state->extra);
+                state->offset += BITS(state->extra);
+                DROPBITS(state->extra);
+                state->back += state->extra;
+            }
+#ifdef INFLATE_STRICT
+            if (state->offset > state->dmax) {
+                strm->msg = (char *)"invalid distance too far back";
+                state->mode = BAD;
+                break;
+            }
+#endif
+            Tracevv((stderr, "inflate:         distance %u\n", state->offset));
+            state->mode = MATCH;
+        case MATCH:
+            if (left == 0) goto inf_leave;
+            copy = out - left;
+            if (state->offset > copy) {         /* copy from window */
+                copy = state->offset - copy;
+                if (copy > state->whave) {
+                    if (state->sane) {
+                        strm->msg = (char *)"invalid distance too far back";
+                        state->mode = BAD;
+                        break;
+                    }
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+                    Trace((stderr, "inflate.c too far\n"));
+                    copy -= state->whave;
+                    if (copy > state->length) copy = state->length;
+                    if (copy > left) copy = left;
+                    left -= copy;
+                    state->length -= copy;
+                    do {
+                        *put++ = 0;
+                    } while (--copy);
+                    if (state->length == 0) state->mode = LEN;
+                    break;
+#endif
+                }
+                if (copy > state->wnext) {
+                    copy -= state->wnext;
+                    from = state->window + (state->wsize - copy);
+                }
+                else
+                    from = state->window + (state->wnext - copy);
+                if (copy > state->length) copy = state->length;
+            }
+            else {                              /* copy from output */
+                from = put - state->offset;
+                copy = state->length;
+            }
+            if (copy > left) copy = left;
+            left -= copy;
+            state->length -= copy;
+            do {
+                *put++ = *from++;
+            } while (--copy);
+            if (state->length == 0) state->mode = LEN;
+            break;
+        case LIT:
+            if (left == 0) goto inf_leave;
+            *put++ = (unsigned char)(state->length);
+            left--;
+            state->mode = LEN;
+            break;
+        case CHECK:
+            if (state->wrap) {
+                NEEDBITS(32);
+                out -= left;
+                strm->total_out += out;
+                state->total += out;
+                if ((state->wrap & 4) && out)
+                    strm->adler = state->check =
+                        UPDATE(state->check, put - out, out);
+                out = left;
+                if ((state->wrap & 4) && (
+#ifdef GUNZIP
+                     state->flags ? hold :
+#endif
+                     ZSWAP32(hold)) != state->check) {
+                    strm->msg = (char *)"incorrect data check";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+                Tracev((stderr, "inflate:   check matches trailer\n"));
+            }
+#ifdef GUNZIP
+            state->mode = LENGTH;
+        case LENGTH:
+            if (state->wrap && state->flags) {
+                NEEDBITS(32);
+                if (hold != (state->total & 0xffffffffUL)) {
+                    strm->msg = (char *)"incorrect length check";
+                    state->mode = BAD;
+                    break;
+                }
+                INITBITS();
+                Tracev((stderr, "inflate:   length matches trailer\n"));
+            }
+#endif
+            state->mode = DONE;
+        case DONE:
+            ret = Z_STREAM_END;
+            goto inf_leave;
+        case BAD:
+            ret = Z_DATA_ERROR;
+            goto inf_leave;
+        case MEM:
+            return Z_MEM_ERROR;
+        case SYNC:
+        default:
+            return Z_STREAM_ERROR;
+        }
+
+    /*
+       Return from inflate(), updating the total counts and the check value.
+       If there was no progress during the inflate() call, return a buffer
+       error.  Call updatewindow() to create and/or update the window state.
+       Note: a memory error from inflate() is non-recoverable.
+     */
+  inf_leave:
+    RESTORE();
+    if (state->wsize || (out != strm->avail_out && state->mode < BAD &&
+            (state->mode < CHECK || flush != Z_FINISH)))
+        if (updatewindow(strm, strm->next_out, out - strm->avail_out)) {
+            state->mode = MEM;
+            return Z_MEM_ERROR;
+        }
+    in -= strm->avail_in;
+    out -= strm->avail_out;
+    strm->total_in += in;
+    strm->total_out += out;
+    state->total += out;
+    if ((state->wrap & 4) && out)
+        strm->adler = state->check =
+            UPDATE(state->check, strm->next_out - out, out);
+    strm->data_type = (int)state->bits + (state->last ? 64 : 0) +
+                      (state->mode == TYPE ? 128 : 0) +
+                      (state->mode == LEN_ || state->mode == COPY_ ? 256 : 0);
+    if (((in == 0 && out == 0) || flush == Z_FINISH) && ret == Z_OK)
+        ret = Z_BUF_ERROR;
+    return ret;
+}
+
+int ZEXPORT inflateEnd(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (inflateStateCheck(strm))
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->window != Z_NULL) ZFREE(strm, state->window);
+    ZFREE(strm, strm->state);
+    strm->state = Z_NULL;
+    Tracev((stderr, "inflate: end\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateGetDictionary(strm, dictionary, dictLength)
+z_streamp strm;
+Bytef *dictionary;
+uInt *dictLength;
+{
+    struct inflate_state FAR *state;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+
+    /* copy dictionary */
+    if (state->whave && dictionary != Z_NULL) {
+        zmemcpy(dictionary, state->window + state->wnext,
+                state->whave - state->wnext);
+        zmemcpy(dictionary + state->whave - state->wnext,
+                state->window, state->wnext);
+    }
+    if (dictLength != Z_NULL)
+        *dictLength = state->whave;
+    return Z_OK;
+}
+
+int ZEXPORT inflateSetDictionary(strm, dictionary, dictLength)
+z_streamp strm;
+const Bytef *dictionary;
+uInt dictLength;
+{
+    struct inflate_state FAR *state;
+    unsigned long dictid;
+    int ret;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (state->wrap != 0 && state->mode != DICT)
+        return Z_STREAM_ERROR;
+
+    /* check for correct dictionary identifier */
+    if (state->mode == DICT) {
+        dictid = adler32(0L, Z_NULL, 0);
+        dictid = adler32(dictid, dictionary, dictLength);
+        if (dictid != state->check)
+            return Z_DATA_ERROR;
+    }
+
+    /* copy dictionary to window using updatewindow(), which will amend the
+       existing dictionary if appropriate */
+    ret = updatewindow(strm, dictionary + dictLength, dictLength);
+    if (ret) {
+        state->mode = MEM;
+        return Z_MEM_ERROR;
+    }
+    state->havedict = 1;
+    Tracev((stderr, "inflate:   dictionary set\n"));
+    return Z_OK;
+}
+
+int ZEXPORT inflateGetHeader(strm, head)
+z_streamp strm;
+gz_headerp head;
+{
+    struct inflate_state FAR *state;
+
+    /* check state */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if ((state->wrap & 2) == 0) return Z_STREAM_ERROR;
+
+    /* save header structure */
+    state->head = head;
+    head->done = 0;
+    return Z_OK;
+}
+
+/*
+   Search buf[0..len-1] for the pattern: 0, 0, 0xff, 0xff.  Return when found
+   or when out of input.  When called, *have is the number of pattern bytes
+   found in order so far, in 0..3.  On return *have is updated to the new
+   state.  If on return *have equals four, then the pattern was found and the
+   return value is how many bytes were read including the last byte of the
+   pattern.  If *have is less than four, then the pattern has not been found
+   yet and the return value is len.  In the latter case, syncsearch() can be
+   called again with more data and the *have state.  *have is initialized to
+   zero for the first call.
+ */
+local unsigned syncsearch(have, buf, len)
+unsigned FAR *have;
+const unsigned char FAR *buf;
+unsigned len;
+{
+    unsigned got;
+    unsigned next;
+
+    got = *have;
+    next = 0;
+    while (next < len && got < 4) {
+        if ((int)(buf[next]) == (got < 2 ? 0 : 0xff))
+            got++;
+        else if (buf[next])
+            got = 0;
+        else
+            got = 4 - got;
+        next++;
+    }
+    *have = got;
+    return next;
+}
+
+int ZEXPORT inflateSync(strm)
+z_streamp strm;
+{
+    unsigned len;               /* number of bytes to look at or looked at */
+    unsigned long in, out;      /* temporary to save total_in and total_out */
+    unsigned char buf[4];       /* to restore bit buffer to byte string */
+    struct inflate_state FAR *state;
+
+    /* check parameters */
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (strm->avail_in == 0 && state->bits < 8) return Z_BUF_ERROR;
+
+    /* if first time, start search in bit buffer */
+    if (state->mode != SYNC) {
+        state->mode = SYNC;
+        state->hold <<= state->bits & 7;
+        state->bits -= state->bits & 7;
+        len = 0;
+        while (state->bits >= 8) {
+            buf[len++] = (unsigned char)(state->hold);
+            state->hold >>= 8;
+            state->bits -= 8;
+        }
+        state->have = 0;
+        syncsearch(&(state->have), buf, len);
+    }
+
+    /* search available input */
+    len = syncsearch(&(state->have), strm->next_in, strm->avail_in);
+    strm->avail_in -= len;
+    strm->next_in += len;
+    strm->total_in += len;
+
+    /* return no joy or set up to restart inflate() on a new block */
+    if (state->have != 4) return Z_DATA_ERROR;
+    in = strm->total_in;  out = strm->total_out;
+    inflateReset(strm);
+    strm->total_in = in;  strm->total_out = out;
+    state->mode = TYPE;
+    return Z_OK;
+}
+
+/*
+   Returns true if inflate is currently at the end of a block generated by
+   Z_SYNC_FLUSH or Z_FULL_FLUSH. This function is used by one PPP
+   implementation to provide an additional safety check. PPP uses
+   Z_SYNC_FLUSH but removes the length bytes of the resulting empty stored
+   block. When decompressing, PPP checks that at the end of input packet,
+   inflate is waiting for these length bytes.
+ */
+int ZEXPORT inflateSyncPoint(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    return state->mode == STORED && state->bits == 0;
+}
+
+int ZEXPORT inflateCopy(dest, source)
+z_streamp dest;
+z_streamp source;
+{
+    struct inflate_state FAR *state;
+    struct inflate_state FAR *copy;
+    unsigned char FAR *window;
+    unsigned wsize;
+
+    /* check input */
+    if (inflateStateCheck(source) || dest == Z_NULL)
+        return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)source->state;
+
+    /* allocate space */
+    copy = (struct inflate_state FAR *)
+           ZALLOC(source, 1, sizeof(struct inflate_state));
+    if (copy == Z_NULL) return Z_MEM_ERROR;
+    window = Z_NULL;
+    if (state->window != Z_NULL) {
+        window = (unsigned char FAR *)
+                 ZALLOC(source, 1U << state->wbits, sizeof(unsigned char));
+        if (window == Z_NULL) {
+            ZFREE(source, copy);
+            return Z_MEM_ERROR;
+        }
+    }
+
+    /* copy state */
+    zmemcpy((voidpf)dest, (voidpf)source, sizeof(z_stream));
+    zmemcpy((voidpf)copy, (voidpf)state, sizeof(struct inflate_state));
+    copy->strm = dest;
+    if (state->lencode >= state->codes &&
+        state->lencode <= state->codes + ENOUGH - 1) {
+        copy->lencode = copy->codes + (state->lencode - state->codes);
+        copy->distcode = copy->codes + (state->distcode - state->codes);
+    }
+    copy->next = copy->codes + (state->next - state->codes);
+    if (window != Z_NULL) {
+        wsize = 1U << state->wbits;
+        zmemcpy(window, state->window, wsize);
+    }
+    copy->window = window;
+    dest->state = (struct internal_state FAR *)copy;
+    return Z_OK;
+}
+
+int ZEXPORT inflateUndermine(strm, subvert)
+z_streamp strm;
+int subvert;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+#ifdef INFLATE_ALLOW_INVALID_DISTANCE_TOOFAR_ARRR
+    state->sane = !subvert;
+    return Z_OK;
+#else
+    (void)subvert;
+    state->sane = 1;
+    return Z_DATA_ERROR;
+#endif
+}
+
+int ZEXPORT inflateValidate(strm, check)
+z_streamp strm;
+int check;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm)) return Z_STREAM_ERROR;
+    state = (struct inflate_state FAR *)strm->state;
+    if (check)
+        state->wrap |= 4;
+    else
+        state->wrap &= ~4;
+    return Z_OK;
+}
+
+long ZEXPORT inflateMark(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+
+    if (inflateStateCheck(strm))
+        return -(1L << 16);
+    state = (struct inflate_state FAR *)strm->state;
+    return (long)(((unsigned long)((long)state->back)) << 16) +
+        (state->mode == COPY ? state->length :
+            (state->mode == MATCH ? state->was - state->length : 0));
+}
+
+unsigned long ZEXPORT inflateCodesUsed(strm)
+z_streamp strm;
+{
+    struct inflate_state FAR *state;
+    if (inflateStateCheck(strm)) return (unsigned long)-1;
+    state = (struct inflate_state FAR *)strm->state;
+    return (unsigned long)(state->next - state->codes);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inflate.h glmark2-2021.02/src/zlib/inflate.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inflate.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inflate.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,125 @@
+/* inflate.h -- internal inflate state definition
+ * Copyright (C) 1995-2016 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* define NO_GZIP when compiling if you want to disable gzip header and
+   trailer decoding by inflate().  NO_GZIP would be used to avoid linking in
+   the crc code when it is not needed.  For shared libraries, gzip decoding
+   should be left enabled. */
+#ifndef NO_GZIP
+#  define GUNZIP
+#endif
+
+/* Possible inflate modes between inflate() calls */
+typedef enum {
+    HEAD = 16180,   /* i: waiting for magic header */
+    FLAGS,      /* i: waiting for method and flags (gzip) */
+    TIME,       /* i: waiting for modification time (gzip) */
+    OS,         /* i: waiting for extra flags and operating system (gzip) */
+    EXLEN,      /* i: waiting for extra length (gzip) */
+    EXTRA,      /* i: waiting for extra bytes (gzip) */
+    NAME,       /* i: waiting for end of file name (gzip) */
+    COMMENT,    /* i: waiting for end of comment (gzip) */
+    HCRC,       /* i: waiting for header crc (gzip) */
+    DICTID,     /* i: waiting for dictionary check value */
+    DICT,       /* waiting for inflateSetDictionary() call */
+        TYPE,       /* i: waiting for type bits, including last-flag bit */
+        TYPEDO,     /* i: same, but skip check to exit inflate on new block */
+        STORED,     /* i: waiting for stored size (length and complement) */
+        COPY_,      /* i/o: same as COPY below, but only first time in */
+        COPY,       /* i/o: waiting for input or output to copy stored block */
+        TABLE,      /* i: waiting for dynamic block table lengths */
+        LENLENS,    /* i: waiting for code length code lengths */
+        CODELENS,   /* i: waiting for length/lit and distance code lengths */
+            LEN_,       /* i: same as LEN below, but only first time in */
+            LEN,        /* i: waiting for length/lit/eob code */
+            LENEXT,     /* i: waiting for length extra bits */
+            DIST,       /* i: waiting for distance code */
+            DISTEXT,    /* i: waiting for distance extra bits */
+            MATCH,      /* o: waiting for output space to copy string */
+            LIT,        /* o: waiting for output space to write literal */
+    CHECK,      /* i: waiting for 32-bit check value */
+    LENGTH,     /* i: waiting for 32-bit length (gzip) */
+    DONE,       /* finished check, done -- remain here until reset */
+    BAD,        /* got a data error -- remain here until reset */
+    MEM,        /* got an inflate() memory error -- remain here until reset */
+    SYNC        /* looking for synchronization bytes to restart inflate() */
+} inflate_mode;
+
+/*
+    State transitions between above modes -
+
+    (most modes can go to BAD or MEM on error -- not shown for clarity)
+
+    Process header:
+        HEAD -> (gzip) or (zlib) or (raw)
+        (gzip) -> FLAGS -> TIME -> OS -> EXLEN -> EXTRA -> NAME -> COMMENT ->
+                  HCRC -> TYPE
+        (zlib) -> DICTID or TYPE
+        DICTID -> DICT -> TYPE
+        (raw) -> TYPEDO
+    Read deflate blocks:
+            TYPE -> TYPEDO -> STORED or TABLE or LEN_ or CHECK
+            STORED -> COPY_ -> COPY -> TYPE
+            TABLE -> LENLENS -> CODELENS -> LEN_
+            LEN_ -> LEN
+    Read deflate codes in fixed or dynamic block:
+                LEN -> LENEXT or LIT or TYPE
+                LENEXT -> DIST -> DISTEXT -> MATCH -> LEN
+                LIT -> LEN
+    Process trailer:
+        CHECK -> LENGTH -> DONE
+ */
+
+/* State maintained between inflate() calls -- approximately 7K bytes, not
+   including the allocated sliding window, which is up to 32K bytes. */
+struct inflate_state {
+    z_streamp strm;             /* pointer back to this zlib stream */
+    inflate_mode mode;          /* current inflate mode */
+    int last;                   /* true if processing last block */
+    int wrap;                   /* bit 0 true for zlib, bit 1 true for gzip,
+                                   bit 2 true to validate check value */
+    int havedict;               /* true if dictionary provided */
+    int flags;                  /* gzip header method and flags (0 if zlib) */
+    unsigned dmax;              /* zlib header max distance (INFLATE_STRICT) */
+    unsigned long check;        /* protected copy of check value */
+    unsigned long total;        /* protected copy of output count */
+    gz_headerp head;            /* where to save gzip header information */
+        /* sliding window */
+    unsigned wbits;             /* log base 2 of requested window size */
+    unsigned wsize;             /* window size or zero if not using window */
+    unsigned whave;             /* valid bytes in the window */
+    unsigned wnext;             /* window write index */
+    unsigned char FAR *window;  /* allocated sliding window, if needed */
+        /* bit accumulator */
+    unsigned long hold;         /* input bit accumulator */
+    unsigned bits;              /* number of bits in "in" */
+        /* for string and stored block copying */
+    unsigned length;            /* literal or length of data to copy */
+    unsigned offset;            /* distance back to copy string from */
+        /* for table and code decoding */
+    unsigned extra;             /* extra bits needed */
+        /* fixed and dynamic code tables */
+    code const FAR *lencode;    /* starting table for length/literal codes */
+    code const FAR *distcode;   /* starting table for distance codes */
+    unsigned lenbits;           /* index bits for lencode */
+    unsigned distbits;          /* index bits for distcode */
+        /* dynamic table building */
+    unsigned ncode;             /* number of code length code lengths */
+    unsigned nlen;              /* number of length code lengths */
+    unsigned ndist;             /* number of distance code lengths */
+    unsigned have;              /* number of code lengths in lens[] */
+    code FAR *next;             /* next available space in codes[] */
+    unsigned short lens[320];   /* temporary storage for code lengths */
+    unsigned short work[288];   /* work area for code table building */
+    code codes[ENOUGH];         /* space for code tables */
+    int sane;                   /* if false, allow invalid distance too far */
+    int back;                   /* bits back of last unprocessed length/lit */
+    unsigned was;               /* initial length of match */
+};
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inftrees.c glmark2-2021.02/src/zlib/inftrees.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inftrees.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inftrees.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,304 @@
+/* inftrees.c -- generate Huffman trees for efficient decoding
+ * Copyright (C) 1995-2017 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+#include "zutil.h"
+#include "inftrees.h"
+
+#define MAXBITS 15
+
+const char inflate_copyright[] =
+   " inflate 1.2.11 Copyright 1995-2017 Mark Adler ";
+/*
+  If you use the zlib library in a product, an acknowledgment is welcome
+  in the documentation of your product. If for some reason you cannot
+  include such an acknowledgment, I would appreciate that you keep this
+  copyright string in the executable of your product.
+ */
+
+/*
+   Build a set of tables to decode the provided canonical Huffman code.
+   The code lengths are lens[0..codes-1].  The result starts at *table,
+   whose indices are 0..2^bits-1.  work is a writable array of at least
+   lens shorts, which is used as a work area.  type is the type of code
+   to be generated, CODES, LENS, or DISTS.  On return, zero is success,
+   -1 is an invalid code, and +1 means that ENOUGH isn't enough.  table
+   on return points to the next available entry's address.  bits is the
+   requested root table index bits, and on return it is the actual root
+   table index bits.  It will differ if the request is greater than the
+   longest code or if it is less than the shortest code.
+ */
+int ZLIB_INTERNAL inflate_table(type, lens, codes, table, bits, work)
+codetype type;
+unsigned short FAR *lens;
+unsigned codes;
+code FAR * FAR *table;
+unsigned FAR *bits;
+unsigned short FAR *work;
+{
+    unsigned len;               /* a code's length in bits */
+    unsigned sym;               /* index of code symbols */
+    unsigned min, max;          /* minimum and maximum code lengths */
+    unsigned root;              /* number of index bits for root table */
+    unsigned curr;              /* number of index bits for current table */
+    unsigned drop;              /* code bits to drop for sub-table */
+    int left;                   /* number of prefix codes available */
+    unsigned used;              /* code entries in table used */
+    unsigned huff;              /* Huffman code */
+    unsigned incr;              /* for incrementing code, index */
+    unsigned fill;              /* index for replicating entries */
+    unsigned low;               /* low bits for current root entry */
+    unsigned mask;              /* mask for low root bits */
+    code here;                  /* table entry for duplication */
+    code FAR *next;             /* next available space in table */
+    const unsigned short FAR *base;     /* base value table to use */
+    const unsigned short FAR *extra;    /* extra bits table to use */
+    unsigned match;             /* use base and extra for symbol >= match */
+    unsigned short count[MAXBITS+1];    /* number of codes of each length */
+    unsigned short offs[MAXBITS+1];     /* offsets in table for each length */
+    static const unsigned short lbase[31] = { /* Length codes 257..285 base */
+        3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
+        35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
+    static const unsigned short lext[31] = { /* Length codes 257..285 extra */
+        16, 16, 16, 16, 16, 16, 16, 16, 17, 17, 17, 17, 18, 18, 18, 18,
+        19, 19, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 16, 77, 202};
+    static const unsigned short dbase[32] = { /* Distance codes 0..29 base */
+        1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
+        257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
+        8193, 12289, 16385, 24577, 0, 0};
+    static const unsigned short dext[32] = { /* Distance codes 0..29 extra */
+        16, 16, 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22,
+        23, 23, 24, 24, 25, 25, 26, 26, 27, 27,
+        28, 28, 29, 29, 64, 64};
+
+    /*
+       Process a set of code lengths to create a canonical Huffman code.  The
+       code lengths are lens[0..codes-1].  Each length corresponds to the
+       symbols 0..codes-1.  The Huffman code is generated by first sorting the
+       symbols by length from short to long, and retaining the symbol order
+       for codes with equal lengths.  Then the code starts with all zero bits
+       for the first code of the shortest length, and the codes are integer
+       increments for the same length, and zeros are appended as the length
+       increases.  For the deflate format, these bits are stored backwards
+       from their more natural integer increment ordering, and so when the
+       decoding tables are built in the large loop below, the integer codes
+       are incremented backwards.
+
+       This routine assumes, but does not check, that all of the entries in
+       lens[] are in the range 0..MAXBITS.  The caller must assure this.
+       1..MAXBITS is interpreted as that code length.  zero means that that
+       symbol does not occur in this code.
+
+       The codes are sorted by computing a count of codes for each length,
+       creating from that a table of starting indices for each length in the
+       sorted table, and then entering the symbols in order in the sorted
+       table.  The sorted table is work[], with that space being provided by
+       the caller.
+
+       The length counts are used for other purposes as well, i.e. finding
+       the minimum and maximum length codes, determining if there are any
+       codes at all, checking for a valid set of lengths, and looking ahead
+       at length counts to determine sub-table sizes when building the
+       decoding tables.
+     */
+
+    /* accumulate lengths for codes (assumes lens[] all in 0..MAXBITS) */
+    for (len = 0; len <= MAXBITS; len++)
+        count[len] = 0;
+    for (sym = 0; sym < codes; sym++)
+        count[lens[sym]]++;
+
+    /* bound code lengths, force root to be within code lengths */
+    root = *bits;
+    for (max = MAXBITS; max >= 1; max--)
+        if (count[max] != 0) break;
+    if (root > max) root = max;
+    if (max == 0) {                     /* no symbols to code at all */
+        here.op = (unsigned char)64;    /* invalid code marker */
+        here.bits = (unsigned char)1;
+        here.val = (unsigned short)0;
+        *(*table)++ = here;             /* make a table to force an error */
+        *(*table)++ = here;
+        *bits = 1;
+        return 0;     /* no symbols, but wait for decoding to report error */
+    }
+    for (min = 1; min < max; min++)
+        if (count[min] != 0) break;
+    if (root < min) root = min;
+
+    /* check for an over-subscribed or incomplete set of lengths */
+    left = 1;
+    for (len = 1; len <= MAXBITS; len++) {
+        left <<= 1;
+        left -= count[len];
+        if (left < 0) return -1;        /* over-subscribed */
+    }
+    if (left > 0 && (type == CODES || max != 1))
+        return -1;                      /* incomplete set */
+
+    /* generate offsets into symbol table for each length for sorting */
+    offs[1] = 0;
+    for (len = 1; len < MAXBITS; len++)
+        offs[len + 1] = offs[len] + count[len];
+
+    /* sort symbols by length, by symbol order within each length */
+    for (sym = 0; sym < codes; sym++)
+        if (lens[sym] != 0) work[offs[lens[sym]]++] = (unsigned short)sym;
+
+    /*
+       Create and fill in decoding tables.  In this loop, the table being
+       filled is at next and has curr index bits.  The code being used is huff
+       with length len.  That code is converted to an index by dropping drop
+       bits off of the bottom.  For codes where len is less than drop + curr,
+       those top drop + curr - len bits are incremented through all values to
+       fill the table with replicated entries.
+
+       root is the number of index bits for the root table.  When len exceeds
+       root, sub-tables are created pointed to by the root entry with an index
+       of the low root bits of huff.  This is saved in low to check for when a
+       new sub-table should be started.  drop is zero when the root table is
+       being filled, and drop is root when sub-tables are being filled.
+
+       When a new sub-table is needed, it is necessary to look ahead in the
+       code lengths to determine what size sub-table is needed.  The length
+       counts are used for this, and so count[] is decremented as codes are
+       entered in the tables.
+
+       used keeps track of how many table entries have been allocated from the
+       provided *table space.  It is checked for LENS and DIST tables against
+       the constants ENOUGH_LENS and ENOUGH_DISTS to guard against changes in
+       the initial root table size constants.  See the comments in inftrees.h
+       for more information.
+
+       sym increments through all symbols, and the loop terminates when
+       all codes of length max, i.e. all codes, have been processed.  This
+       routine permits incomplete codes, so another loop after this one fills
+       in the rest of the decoding tables with invalid code markers.
+     */
+
+    /* set up for code type */
+    switch (type) {
+    case CODES:
+        base = extra = work;    /* dummy value--not used */
+        match = 20;
+        break;
+    case LENS:
+        base = lbase;
+        extra = lext;
+        match = 257;
+        break;
+    default:    /* DISTS */
+        base = dbase;
+        extra = dext;
+        match = 0;
+    }
+
+    /* initialize state for loop */
+    huff = 0;                   /* starting code */
+    sym = 0;                    /* starting code symbol */
+    len = min;                  /* starting code length */
+    next = *table;              /* current table to fill in */
+    curr = root;                /* current table index bits */
+    drop = 0;                   /* current bits to drop from code for index */
+    low = (unsigned)(-1);       /* trigger new sub-table when len > root */
+    used = 1U << root;          /* use root table entries */
+    mask = used - 1;            /* mask for comparing low */
+
+    /* check available table space */
+    if ((type == LENS && used > ENOUGH_LENS) ||
+        (type == DISTS && used > ENOUGH_DISTS))
+        return 1;
+
+    /* process all codes and make table entries */
+    for (;;) {
+        /* create table entry */
+        here.bits = (unsigned char)(len - drop);
+        if (work[sym] + 1U < match) {
+            here.op = (unsigned char)0;
+            here.val = work[sym];
+        }
+        else if (work[sym] >= match) {
+            here.op = (unsigned char)(extra[work[sym] - match]);
+            here.val = base[work[sym] - match];
+        }
+        else {
+            here.op = (unsigned char)(32 + 64);         /* end of block */
+            here.val = 0;
+        }
+
+        /* replicate for those indices with low len bits equal to huff */
+        incr = 1U << (len - drop);
+        fill = 1U << curr;
+        min = fill;                 /* save offset to next table */
+        do {
+            fill -= incr;
+            next[(huff >> drop) + fill] = here;
+        } while (fill != 0);
+
+        /* backwards increment the len-bit code huff */
+        incr = 1U << (len - 1);
+        while (huff & incr)
+            incr >>= 1;
+        if (incr != 0) {
+            huff &= incr - 1;
+            huff += incr;
+        }
+        else
+            huff = 0;
+
+        /* go to next symbol, update count, len */
+        sym++;
+        if (--(count[len]) == 0) {
+            if (len == max) break;
+            len = lens[work[sym]];
+        }
+
+        /* create new sub-table if needed */
+        if (len > root && (huff & mask) != low) {
+            /* if first time, transition to sub-tables */
+            if (drop == 0)
+                drop = root;
+
+            /* increment past last table */
+            next += min;            /* here min is 1 << curr */
+
+            /* determine length of next table */
+            curr = len - drop;
+            left = (int)(1 << curr);
+            while (curr + drop < max) {
+                left -= count[curr + drop];
+                if (left <= 0) break;
+                curr++;
+                left <<= 1;
+            }
+
+            /* check for enough space */
+            used += 1U << curr;
+            if ((type == LENS && used > ENOUGH_LENS) ||
+                (type == DISTS && used > ENOUGH_DISTS))
+                return 1;
+
+            /* point entry in root table to sub-table */
+            low = huff & mask;
+            (*table)[low].op = (unsigned char)curr;
+            (*table)[low].bits = (unsigned char)root;
+            (*table)[low].val = (unsigned short)(next - *table);
+        }
+    }
+
+    /* fill in remaining table entry if code is incomplete (guaranteed to have
+       at most one remaining entry, since if the code is incomplete, the
+       maximum code length that was allowed to get this far is one bit) */
+    if (huff != 0) {
+        here.op = (unsigned char)64;            /* invalid code marker */
+        here.bits = (unsigned char)(len - drop);
+        here.val = (unsigned short)0;
+        next[huff] = here;
+    }
+
+    /* set return parameters */
+    *table += used;
+    *bits = root;
+    return 0;
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/inftrees.h glmark2-2021.02/src/zlib/inftrees.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/inftrees.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/inftrees.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,62 @@
+/* inftrees.h -- header to use inftrees.c
+ * Copyright (C) 1995-2005, 2010 Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* Structure for decoding tables.  Each entry provides either the
+   information needed to do the operation requested by the code that
+   indexed that table entry, or it provides a pointer to another
+   table that indexes more bits of the code.  op indicates whether
+   the entry is a pointer to another table, a literal, a length or
+   distance, an end-of-block, or an invalid code.  For a table
+   pointer, the low four bits of op is the number of index bits of
+   that table.  For a length or distance, the low four bits of op
+   is the number of extra bits to get after the code.  bits is
+   the number of bits in this code or part of the code to drop off
+   of the bit buffer.  val is the actual byte to output in the case
+   of a literal, the base length or distance, or the offset from
+   the current table to the next table.  Each entry is four bytes. */
+typedef struct {
+    unsigned char op;           /* operation, extra bits, table bits */
+    unsigned char bits;         /* bits in this part of the code */
+    unsigned short val;         /* offset in table or code value */
+} code;
+
+/* op values as set by inflate_table():
+    00000000 - literal
+    0000tttt - table link, tttt != 0 is the number of table index bits
+    0001eeee - length or distance, eeee is the number of extra bits
+    01100000 - end of block
+    01000000 - invalid code
+ */
+
+/* Maximum size of the dynamic table.  The maximum number of code structures is
+   1444, which is the sum of 852 for literal/length codes and 592 for distance
+   codes.  These values were found by exhaustive searches using the program
+   examples/enough.c found in the zlib distribtution.  The arguments to that
+   program are the number of symbols, the initial root table size, and the
+   maximum bit length of a code.  "enough 286 9 15" for literal/length codes
+   returns returns 852, and "enough 30 6 15" for distance codes returns 592.
+   The initial root table size (9 or 6) is found in the fifth argument of the
+   inflate_table() calls in inflate.c and infback.c.  If the root table size is
+   changed, then these maximum sizes would be need to be recalculated and
+   updated. */
+#define ENOUGH_LENS 852
+#define ENOUGH_DISTS 592
+#define ENOUGH (ENOUGH_LENS+ENOUGH_DISTS)
+
+/* Type of code to build for inflate_table() */
+typedef enum {
+    CODES,
+    LENS,
+    DISTS
+} codetype;
+
+int ZLIB_INTERNAL inflate_table OF((codetype type, unsigned short FAR *lens,
+                             unsigned codes, code FAR * FAR *table,
+                             unsigned FAR *bits, unsigned short FAR *work));
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/README glmark2-2021.02/src/zlib/README
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/README	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/README	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,115 @@
+ZLIB DATA COMPRESSION LIBRARY
+
+zlib 1.2.11 is a general purpose data compression library.  All the code is
+thread safe.  The data format used by the zlib library is described by RFCs
+(Request for Comments) 1950 to 1952 in the files
+http://tools.ietf.org/html/rfc1950 (zlib format), rfc1951 (deflate format) and
+rfc1952 (gzip format).
+
+All functions of the compression library are documented in the file zlib.h
+(volunteer to write man pages welcome, contact zlib@gzip.org).  A usage example
+of the library is given in the file test/example.c which also tests that
+the library is working correctly.  Another example is given in the file
+test/minigzip.c.  The compression library itself is composed of all source
+files in the root directory.
+
+To compile all files and run the test program, follow the instructions given at
+the top of Makefile.in.  In short "./configure; make test", and if that goes
+well, "make install" should work for most flavors of Unix.  For Windows, use
+one of the special makefiles in win32/ or contrib/vstudio/ .  For VMS, use
+make_vms.com.
+
+Questions about zlib should be sent to <zlib@gzip.org>, or to Gilles Vollant
+<info@winimage.com> for the Windows DLL version.  The zlib home page is
+http://zlib.net/ .  Before reporting a problem, please check this site to
+verify that you have the latest version of zlib; otherwise get the latest
+version and check whether the problem still exists or not.
+
+PLEASE read the zlib FAQ http://zlib.net/zlib_faq.html before asking for help.
+
+Mark Nelson <markn@ieee.org> wrote an article about zlib for the Jan.  1997
+issue of Dr.  Dobb's Journal; a copy of the article is available at
+http://marknelson.us/1997/01/01/zlib-engine/ .
+
+The changes made in version 1.2.11 are documented in the file ChangeLog.
+
+Unsupported third party contributions are provided in directory contrib/ .
+
+zlib is available in Java using the java.util.zip package, documented at
+http://java.sun.com/developer/technicalArticles/Programming/compression/ .
+
+A Perl interface to zlib written by Paul Marquess <pmqs@cpan.org> is available
+at CPAN (Comprehensive Perl Archive Network) sites, including
+http://search.cpan.org/~pmqs/IO-Compress-Zlib/ .
+
+A Python interface to zlib written by A.M. Kuchling <amk@amk.ca> is
+available in Python 1.5 and later versions, see
+http://docs.python.org/library/zlib.html .
+
+zlib is built into tcl: http://wiki.tcl.tk/4610 .
+
+An experimental package to read and write files in .zip format, written on top
+of zlib by Gilles Vollant <info@winimage.com>, is available in the
+contrib/minizip directory of zlib.
+
+
+Notes for some targets:
+
+- For Windows DLL versions, please see win32/DLL_FAQ.txt
+
+- For 64-bit Irix, deflate.c must be compiled without any optimization. With
+  -O, one libpng test fails. The test works in 32 bit mode (with the -n32
+  compiler flag). The compiler bug has been reported to SGI.
+
+- zlib doesn't work with gcc 2.6.3 on a DEC 3000/300LX under OSF/1 2.1 it works
+  when compiled with cc.
+
+- On Digital Unix 4.0D (formely OSF/1) on AlphaServer, the cc option -std1 is
+  necessary to get gzprintf working correctly. This is done by configure.
+
+- zlib doesn't work on HP-UX 9.05 with some versions of /bin/cc. It works with
+  other compilers. Use "make test" to check your compiler.
+
+- gzdopen is not supported on RISCOS or BEOS.
+
+- For PalmOs, see http://palmzlib.sourceforge.net/
+
+
+Acknowledgments:
+
+  The deflate format used by zlib was defined by Phil Katz.  The deflate and
+  zlib specifications were written by L.  Peter Deutsch.  Thanks to all the
+  people who reported problems and suggested various improvements in zlib; they
+  are too numerous to cite here.
+
+Copyright notice:
+
+ (C) 1995-2017 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+If you use the zlib library in a product, we would appreciate *not* receiving
+lengthy legal documents to sign.  The sources are provided for free but without
+warranty of any kind.  The library has been entirely written by Jean-loup
+Gailly and Mark Adler; it does not include third-party code.
+
+If you redistribute modified sources, we would appreciate that you include in
+the file ChangeLog history information documenting your changes.  Please read
+the FAQ for more information on the distribution of modified source versions.
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/trees.c glmark2-2021.02/src/zlib/trees.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/trees.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/trees.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1203 @@
+/* trees.c -- output deflated data using Huffman coding
+ * Copyright (C) 1995-2017 Jean-loup Gailly
+ * detect_data_type() function provided freely by Cosmin Truta, 2006
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/*
+ *  ALGORITHM
+ *
+ *      The "deflation" process uses several Huffman trees. The more
+ *      common source values are represented by shorter bit sequences.
+ *
+ *      Each code tree is stored in a compressed form which is itself
+ * a Huffman encoding of the lengths of all the code strings (in
+ * ascending order by source values).  The actual code strings are
+ * reconstructed from the lengths in the inflate process, as described
+ * in the deflate specification.
+ *
+ *  REFERENCES
+ *
+ *      Deutsch, L.P.,"'Deflate' Compressed Data Format Specification".
+ *      Available in ftp.uu.net:/pub/archiving/zip/doc/deflate-1.1.doc
+ *
+ *      Storer, James A.
+ *          Data Compression:  Methods and Theory, pp. 49-50.
+ *          Computer Science Press, 1988.  ISBN 0-7167-8156-5.
+ *
+ *      Sedgewick, R.
+ *          Algorithms, p290.
+ *          Addison-Wesley, 1983. ISBN 0-201-06672-6.
+ */
+
+/* @(#) $Id$ */
+
+/* #define GEN_TREES_H */
+
+#include "deflate.h"
+
+#ifdef ZLIB_DEBUG
+#  include <ctype.h>
+#endif
+
+/* ===========================================================================
+ * Constants
+ */
+
+#define MAX_BL_BITS 7
+/* Bit length codes must not exceed MAX_BL_BITS bits */
+
+#define END_BLOCK 256
+/* end of block literal code */
+
+#define REP_3_6      16
+/* repeat previous bit length 3-6 times (2 bits of repeat count) */
+
+#define REPZ_3_10    17
+/* repeat a zero length 3-10 times  (3 bits of repeat count) */
+
+#define REPZ_11_138  18
+/* repeat a zero length 11-138 times  (7 bits of repeat count) */
+
+local const int extra_lbits[LENGTH_CODES] /* extra bits for each length code */
+   = {0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0};
+
+local const int extra_dbits[D_CODES] /* extra bits for each distance code */
+   = {0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13};
+
+local const int extra_blbits[BL_CODES]/* extra bits for each bit length code */
+   = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,2,3,7};
+
+local const uch bl_order[BL_CODES]
+   = {16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15};
+/* The lengths of the bit length codes are sent in order of decreasing
+ * probability, to avoid transmitting the lengths for unused bit length codes.
+ */
+
+/* ===========================================================================
+ * Local data. These are initialized only once.
+ */
+
+#define DIST_CODE_LEN  512 /* see definition of array dist_code below */
+
+#if defined(GEN_TREES_H) || !defined(STDC)
+/* non ANSI compilers may not accept trees.h */
+
+local ct_data static_ltree[L_CODES+2];
+/* The static literal tree. Since the bit lengths are imposed, there is no
+ * need for the L_CODES extra codes used during heap construction. However
+ * The codes 286 and 287 are needed to build a canonical tree (see _tr_init
+ * below).
+ */
+
+local ct_data static_dtree[D_CODES];
+/* The static distance tree. (Actually a trivial tree since all codes use
+ * 5 bits.)
+ */
+
+uch _dist_code[DIST_CODE_LEN];
+/* Distance codes. The first 256 values correspond to the distances
+ * 3 .. 258, the last 256 values correspond to the top 8 bits of
+ * the 15 bit distances.
+ */
+
+uch _length_code[MAX_MATCH-MIN_MATCH+1];
+/* length code for each normalized match length (0 == MIN_MATCH) */
+
+local int base_length[LENGTH_CODES];
+/* First normalized length for each code (0 = MIN_MATCH) */
+
+local int base_dist[D_CODES];
+/* First normalized distance for each code (0 = distance of 1) */
+
+#else
+#  include "trees.h"
+#endif /* GEN_TREES_H */
+
+struct static_tree_desc_s {
+    const ct_data *static_tree;  /* static tree or NULL */
+    const intf *extra_bits;      /* extra bits for each code or NULL */
+    int     extra_base;          /* base index for extra_bits */
+    int     elems;               /* max number of elements in the tree */
+    int     max_length;          /* max bit length for the codes */
+};
+
+local const static_tree_desc  static_l_desc =
+{static_ltree, extra_lbits, LITERALS+1, L_CODES, MAX_BITS};
+
+local const static_tree_desc  static_d_desc =
+{static_dtree, extra_dbits, 0,          D_CODES, MAX_BITS};
+
+local const static_tree_desc  static_bl_desc =
+{(const ct_data *)0, extra_blbits, 0,   BL_CODES, MAX_BL_BITS};
+
+/* ===========================================================================
+ * Local (static) routines in this file.
+ */
+
+local void tr_static_init OF((void));
+local void init_block     OF((deflate_state *s));
+local void pqdownheap     OF((deflate_state *s, ct_data *tree, int k));
+local void gen_bitlen     OF((deflate_state *s, tree_desc *desc));
+local void gen_codes      OF((ct_data *tree, int max_code, ushf *bl_count));
+local void build_tree     OF((deflate_state *s, tree_desc *desc));
+local void scan_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local void send_tree      OF((deflate_state *s, ct_data *tree, int max_code));
+local int  build_bl_tree  OF((deflate_state *s));
+local void send_all_trees OF((deflate_state *s, int lcodes, int dcodes,
+                              int blcodes));
+local void compress_block OF((deflate_state *s, const ct_data *ltree,
+                              const ct_data *dtree));
+local int  detect_data_type OF((deflate_state *s));
+local unsigned bi_reverse OF((unsigned value, int length));
+local void bi_windup      OF((deflate_state *s));
+local void bi_flush       OF((deflate_state *s));
+
+#ifdef GEN_TREES_H
+local void gen_trees_header OF((void));
+#endif
+
+#ifndef ZLIB_DEBUG
+#  define send_code(s, c, tree) send_bits(s, tree[c].Code, tree[c].Len)
+   /* Send a code of the given tree. c and tree must not have side effects */
+
+#else /* !ZLIB_DEBUG */
+#  define send_code(s, c, tree) \
+     { if (z_verbose>2) fprintf(stderr,"\ncd %3d ",(c)); \
+       send_bits(s, tree[c].Code, tree[c].Len); }
+#endif
+
+/* ===========================================================================
+ * Output a short LSB first on the stream.
+ * IN assertion: there is enough room in pendingBuf.
+ */
+#define put_short(s, w) { \
+    put_byte(s, (uch)((w) & 0xff)); \
+    put_byte(s, (uch)((ush)(w) >> 8)); \
+}
+
+/* ===========================================================================
+ * Send a value on a given number of bits.
+ * IN assertion: length <= 16 and value fits in length bits.
+ */
+#ifdef ZLIB_DEBUG
+local void send_bits      OF((deflate_state *s, int value, int length));
+
+local void send_bits(s, value, length)
+    deflate_state *s;
+    int value;  /* value to send */
+    int length; /* number of bits */
+{
+    Tracevv((stderr," l %2d v %4x ", length, value));
+    Assert(length > 0 && length <= 15, "invalid length");
+    s->bits_sent += (ulg)length;
+
+    /* If not enough room in bi_buf, use (valid) bits from bi_buf and
+     * (16 - bi_valid) bits from value, leaving (width - (16-bi_valid))
+     * unused bits in value.
+     */
+    if (s->bi_valid > (int)Buf_size - length) {
+        s->bi_buf |= (ush)value << s->bi_valid;
+        put_short(s, s->bi_buf);
+        s->bi_buf = (ush)value >> (Buf_size - s->bi_valid);
+        s->bi_valid += length - Buf_size;
+    } else {
+        s->bi_buf |= (ush)value << s->bi_valid;
+        s->bi_valid += length;
+    }
+}
+#else /* !ZLIB_DEBUG */
+
+#define send_bits(s, value, length) \
+{ int len = length;\
+  if (s->bi_valid > (int)Buf_size - len) {\
+    int val = (int)value;\
+    s->bi_buf |= (ush)val << s->bi_valid;\
+    put_short(s, s->bi_buf);\
+    s->bi_buf = (ush)val >> (Buf_size - s->bi_valid);\
+    s->bi_valid += len - Buf_size;\
+  } else {\
+    s->bi_buf |= (ush)(value) << s->bi_valid;\
+    s->bi_valid += len;\
+  }\
+}
+#endif /* ZLIB_DEBUG */
+
+
+/* the arguments must not have side effects */
+
+/* ===========================================================================
+ * Initialize the various 'constant' tables.
+ */
+local void tr_static_init()
+{
+#if defined(GEN_TREES_H) || !defined(STDC)
+    static int static_init_done = 0;
+    int n;        /* iterates over tree elements */
+    int bits;     /* bit counter */
+    int length;   /* length value */
+    int code;     /* code value */
+    int dist;     /* distance index */
+    ush bl_count[MAX_BITS+1];
+    /* number of codes at each bit length for an optimal tree */
+
+    if (static_init_done) return;
+
+    /* For some embedded targets, global variables are not initialized: */
+#ifdef NO_INIT_GLOBAL_POINTERS
+    static_l_desc.static_tree = static_ltree;
+    static_l_desc.extra_bits = extra_lbits;
+    static_d_desc.static_tree = static_dtree;
+    static_d_desc.extra_bits = extra_dbits;
+    static_bl_desc.extra_bits = extra_blbits;
+#endif
+
+    /* Initialize the mapping length (0..255) -> length code (0..28) */
+    length = 0;
+    for (code = 0; code < LENGTH_CODES-1; code++) {
+        base_length[code] = length;
+        for (n = 0; n < (1<<extra_lbits[code]); n++) {
+            _length_code[length++] = (uch)code;
+        }
+    }
+    Assert (length == 256, "tr_static_init: length != 256");
+    /* Note that the length 255 (match length 258) can be represented
+     * in two different ways: code 284 + 5 bits or code 285, so we
+     * overwrite length_code[255] to use the best encoding:
+     */
+    _length_code[length-1] = (uch)code;
+
+    /* Initialize the mapping dist (0..32K) -> dist code (0..29) */
+    dist = 0;
+    for (code = 0 ; code < 16; code++) {
+        base_dist[code] = dist;
+        for (n = 0; n < (1<<extra_dbits[code]); n++) {
+            _dist_code[dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: dist != 256");
+    dist >>= 7; /* from now on, all distances are divided by 128 */
+    for ( ; code < D_CODES; code++) {
+        base_dist[code] = dist << 7;
+        for (n = 0; n < (1<<(extra_dbits[code]-7)); n++) {
+            _dist_code[256 + dist++] = (uch)code;
+        }
+    }
+    Assert (dist == 256, "tr_static_init: 256+dist != 512");
+
+    /* Construct the codes of the static literal tree */
+    for (bits = 0; bits <= MAX_BITS; bits++) bl_count[bits] = 0;
+    n = 0;
+    while (n <= 143) static_ltree[n++].Len = 8, bl_count[8]++;
+    while (n <= 255) static_ltree[n++].Len = 9, bl_count[9]++;
+    while (n <= 279) static_ltree[n++].Len = 7, bl_count[7]++;
+    while (n <= 287) static_ltree[n++].Len = 8, bl_count[8]++;
+    /* Codes 286 and 287 do not exist, but we must include them in the
+     * tree construction to get a canonical Huffman tree (longest code
+     * all ones)
+     */
+    gen_codes((ct_data *)static_ltree, L_CODES+1, bl_count);
+
+    /* The static distance tree is trivial: */
+    for (n = 0; n < D_CODES; n++) {
+        static_dtree[n].Len = 5;
+        static_dtree[n].Code = bi_reverse((unsigned)n, 5);
+    }
+    static_init_done = 1;
+
+#  ifdef GEN_TREES_H
+    gen_trees_header();
+#  endif
+#endif /* defined(GEN_TREES_H) || !defined(STDC) */
+}
+
+/* ===========================================================================
+ * Genererate the file trees.h describing the static trees.
+ */
+#ifdef GEN_TREES_H
+#  ifndef ZLIB_DEBUG
+#    include <stdio.h>
+#  endif
+
+#  define SEPARATOR(i, last, width) \
+      ((i) == (last)? "\n};\n\n" :    \
+       ((i) % (width) == (width)-1 ? ",\n" : ", "))
+
+void gen_trees_header()
+{
+    FILE *header = fopen("trees.h", "w");
+    int i;
+
+    Assert (header != NULL, "Can't open trees.h");
+    fprintf(header,
+            "/* header created automatically with -DGEN_TREES_H */\n\n");
+
+    fprintf(header, "local const ct_data static_ltree[L_CODES+2] = {\n");
+    for (i = 0; i < L_CODES+2; i++) {
+        fprintf(header, "{{%3u},{%3u}}%s", static_ltree[i].Code,
+                static_ltree[i].Len, SEPARATOR(i, L_CODES+1, 5));
+    }
+
+    fprintf(header, "local const ct_data static_dtree[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+        fprintf(header, "{{%2u},{%2u}}%s", static_dtree[i].Code,
+                static_dtree[i].Len, SEPARATOR(i, D_CODES-1, 5));
+    }
+
+    fprintf(header, "const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = {\n");
+    for (i = 0; i < DIST_CODE_LEN; i++) {
+        fprintf(header, "%2u%s", _dist_code[i],
+                SEPARATOR(i, DIST_CODE_LEN-1, 20));
+    }
+
+    fprintf(header,
+        "const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= {\n");
+    for (i = 0; i < MAX_MATCH-MIN_MATCH+1; i++) {
+        fprintf(header, "%2u%s", _length_code[i],
+                SEPARATOR(i, MAX_MATCH-MIN_MATCH, 20));
+    }
+
+    fprintf(header, "local const int base_length[LENGTH_CODES] = {\n");
+    for (i = 0; i < LENGTH_CODES; i++) {
+        fprintf(header, "%1u%s", base_length[i],
+                SEPARATOR(i, LENGTH_CODES-1, 20));
+    }
+
+    fprintf(header, "local const int base_dist[D_CODES] = {\n");
+    for (i = 0; i < D_CODES; i++) {
+        fprintf(header, "%5u%s", base_dist[i],
+                SEPARATOR(i, D_CODES-1, 10));
+    }
+
+    fclose(header);
+}
+#endif /* GEN_TREES_H */
+
+/* ===========================================================================
+ * Initialize the tree data structures for a new zlib stream.
+ */
+void ZLIB_INTERNAL _tr_init(s)
+    deflate_state *s;
+{
+    tr_static_init();
+
+    s->l_desc.dyn_tree = s->dyn_ltree;
+    s->l_desc.stat_desc = &static_l_desc;
+
+    s->d_desc.dyn_tree = s->dyn_dtree;
+    s->d_desc.stat_desc = &static_d_desc;
+
+    s->bl_desc.dyn_tree = s->bl_tree;
+    s->bl_desc.stat_desc = &static_bl_desc;
+
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+#ifdef ZLIB_DEBUG
+    s->compressed_len = 0L;
+    s->bits_sent = 0L;
+#endif
+
+    /* Initialize the first block of the first file: */
+    init_block(s);
+}
+
+/* ===========================================================================
+ * Initialize a new block.
+ */
+local void init_block(s)
+    deflate_state *s;
+{
+    int n; /* iterates over tree elements */
+
+    /* Initialize the trees. */
+    for (n = 0; n < L_CODES;  n++) s->dyn_ltree[n].Freq = 0;
+    for (n = 0; n < D_CODES;  n++) s->dyn_dtree[n].Freq = 0;
+    for (n = 0; n < BL_CODES; n++) s->bl_tree[n].Freq = 0;
+
+    s->dyn_ltree[END_BLOCK].Freq = 1;
+    s->opt_len = s->static_len = 0L;
+    s->last_lit = s->matches = 0;
+}
+
+#define SMALLEST 1
+/* Index within the heap array of least frequent node in the Huffman tree */
+
+
+/* ===========================================================================
+ * Remove the smallest element from the heap and recreate the heap with
+ * one less element. Updates heap and heap_len.
+ */
+#define pqremove(s, tree, top) \
+{\
+    top = s->heap[SMALLEST]; \
+    s->heap[SMALLEST] = s->heap[s->heap_len--]; \
+    pqdownheap(s, tree, SMALLEST); \
+}
+
+/* ===========================================================================
+ * Compares to subtrees, using the tree depth as tie breaker when
+ * the subtrees have equal frequency. This minimizes the worst case length.
+ */
+#define smaller(tree, n, m, depth) \
+   (tree[n].Freq < tree[m].Freq || \
+   (tree[n].Freq == tree[m].Freq && depth[n] <= depth[m]))
+
+/* ===========================================================================
+ * Restore the heap property by moving down the tree starting at node k,
+ * exchanging a node with the smallest of its two sons if necessary, stopping
+ * when the heap property is re-established (each father smaller than its
+ * two sons).
+ */
+local void pqdownheap(s, tree, k)
+    deflate_state *s;
+    ct_data *tree;  /* the tree to restore */
+    int k;               /* node to move down */
+{
+    int v = s->heap[k];
+    int j = k << 1;  /* left son of k */
+    while (j <= s->heap_len) {
+        /* Set j to the smallest of the two sons: */
+        if (j < s->heap_len &&
+            smaller(tree, s->heap[j+1], s->heap[j], s->depth)) {
+            j++;
+        }
+        /* Exit if v is smaller than both sons */
+        if (smaller(tree, v, s->heap[j], s->depth)) break;
+
+        /* Exchange v with the smallest son */
+        s->heap[k] = s->heap[j];  k = j;
+
+        /* And continue down the tree, setting j to the left son of k */
+        j <<= 1;
+    }
+    s->heap[k] = v;
+}
+
+/* ===========================================================================
+ * Compute the optimal bit lengths for a tree and update the total bit length
+ * for the current block.
+ * IN assertion: the fields freq and dad are set, heap[heap_max] and
+ *    above are the tree nodes sorted by increasing frequency.
+ * OUT assertions: the field len is set to the optimal bit length, the
+ *     array bl_count contains the frequencies for each bit length.
+ *     The length opt_len is updated; static_len is also updated if stree is
+ *     not null.
+ */
+local void gen_bitlen(s, desc)
+    deflate_state *s;
+    tree_desc *desc;    /* the tree descriptor */
+{
+    ct_data *tree        = desc->dyn_tree;
+    int max_code         = desc->max_code;
+    const ct_data *stree = desc->stat_desc->static_tree;
+    const intf *extra    = desc->stat_desc->extra_bits;
+    int base             = desc->stat_desc->extra_base;
+    int max_length       = desc->stat_desc->max_length;
+    int h;              /* heap index */
+    int n, m;           /* iterate over the tree elements */
+    int bits;           /* bit length */
+    int xbits;          /* extra bits */
+    ush f;              /* frequency */
+    int overflow = 0;   /* number of elements with bit length too large */
+
+    for (bits = 0; bits <= MAX_BITS; bits++) s->bl_count[bits] = 0;
+
+    /* In a first pass, compute the optimal bit lengths (which may
+     * overflow in the case of the bit length tree).
+     */
+    tree[s->heap[s->heap_max]].Len = 0; /* root of the heap */
+
+    for (h = s->heap_max+1; h < HEAP_SIZE; h++) {
+        n = s->heap[h];
+        bits = tree[tree[n].Dad].Len + 1;
+        if (bits > max_length) bits = max_length, overflow++;
+        tree[n].Len = (ush)bits;
+        /* We overwrite tree[n].Dad which is no longer needed */
+
+        if (n > max_code) continue; /* not a leaf node */
+
+        s->bl_count[bits]++;
+        xbits = 0;
+        if (n >= base) xbits = extra[n-base];
+        f = tree[n].Freq;
+        s->opt_len += (ulg)f * (unsigned)(bits + xbits);
+        if (stree) s->static_len += (ulg)f * (unsigned)(stree[n].Len + xbits);
+    }
+    if (overflow == 0) return;
+
+    Tracev((stderr,"\nbit length overflow\n"));
+    /* This happens for example on obj2 and pic of the Calgary corpus */
+
+    /* Find the first bit length which could increase: */
+    do {
+        bits = max_length-1;
+        while (s->bl_count[bits] == 0) bits--;
+        s->bl_count[bits]--;      /* move one leaf down the tree */
+        s->bl_count[bits+1] += 2; /* move one overflow item as its brother */
+        s->bl_count[max_length]--;
+        /* The brother of the overflow item also moves one step up,
+         * but this does not affect bl_count[max_length]
+         */
+        overflow -= 2;
+    } while (overflow > 0);
+
+    /* Now recompute all bit lengths, scanning in increasing frequency.
+     * h is still equal to HEAP_SIZE. (It is simpler to reconstruct all
+     * lengths instead of fixing only the wrong ones. This idea is taken
+     * from 'ar' written by Haruhiko Okumura.)
+     */
+    for (bits = max_length; bits != 0; bits--) {
+        n = s->bl_count[bits];
+        while (n != 0) {
+            m = s->heap[--h];
+            if (m > max_code) continue;
+            if ((unsigned) tree[m].Len != (unsigned) bits) {
+                Tracev((stderr,"code %d bits %d->%d\n", m, tree[m].Len, bits));
+                s->opt_len += ((ulg)bits - tree[m].Len) * tree[m].Freq;
+                tree[m].Len = (ush)bits;
+            }
+            n--;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Generate the codes for a given tree and bit counts (which need not be
+ * optimal).
+ * IN assertion: the array bl_count contains the bit length statistics for
+ * the given tree and the field len is set for all tree elements.
+ * OUT assertion: the field code is set for all tree elements of non
+ *     zero code length.
+ */
+local void gen_codes (tree, max_code, bl_count)
+    ct_data *tree;             /* the tree to decorate */
+    int max_code;              /* largest code with non zero frequency */
+    ushf *bl_count;            /* number of codes at each bit length */
+{
+    ush next_code[MAX_BITS+1]; /* next code value for each bit length */
+    unsigned code = 0;         /* running code value */
+    int bits;                  /* bit index */
+    int n;                     /* code index */
+
+    /* The distribution counts are first used to generate the code values
+     * without bit reversal.
+     */
+    for (bits = 1; bits <= MAX_BITS; bits++) {
+        code = (code + bl_count[bits-1]) << 1;
+        next_code[bits] = (ush)code;
+    }
+    /* Check that the bit counts in bl_count are consistent. The last code
+     * must be all ones.
+     */
+    Assert (code + bl_count[MAX_BITS]-1 == (1<<MAX_BITS)-1,
+            "inconsistent bit counts");
+    Tracev((stderr,"\ngen_codes: max_code %d ", max_code));
+
+    for (n = 0;  n <= max_code; n++) {
+        int len = tree[n].Len;
+        if (len == 0) continue;
+        /* Now reverse the bits */
+        tree[n].Code = (ush)bi_reverse(next_code[len]++, len);
+
+        Tracecv(tree != static_ltree, (stderr,"\nn %3d %c l %2d c %4x (%x) ",
+             n, (isgraph(n) ? n : ' '), len, tree[n].Code, next_code[len]-1));
+    }
+}
+
+/* ===========================================================================
+ * Construct one Huffman tree and assigns the code bit strings and lengths.
+ * Update the total bit length for the current block.
+ * IN assertion: the field freq is set for all tree elements.
+ * OUT assertions: the fields len and code are set to the optimal bit length
+ *     and corresponding code. The length opt_len is updated; static_len is
+ *     also updated if stree is not null. The field max_code is set.
+ */
+local void build_tree(s, desc)
+    deflate_state *s;
+    tree_desc *desc; /* the tree descriptor */
+{
+    ct_data *tree         = desc->dyn_tree;
+    const ct_data *stree  = desc->stat_desc->static_tree;
+    int elems             = desc->stat_desc->elems;
+    int n, m;          /* iterate over heap elements */
+    int max_code = -1; /* largest code with non zero frequency */
+    int node;          /* new node being created */
+
+    /* Construct the initial heap, with least frequent element in
+     * heap[SMALLEST]. The sons of heap[n] are heap[2*n] and heap[2*n+1].
+     * heap[0] is not used.
+     */
+    s->heap_len = 0, s->heap_max = HEAP_SIZE;
+
+    for (n = 0; n < elems; n++) {
+        if (tree[n].Freq != 0) {
+            s->heap[++(s->heap_len)] = max_code = n;
+            s->depth[n] = 0;
+        } else {
+            tree[n].Len = 0;
+        }
+    }
+
+    /* The pkzip format requires that at least one distance code exists,
+     * and that at least one bit should be sent even if there is only one
+     * possible code. So to avoid special checks later on we force at least
+     * two codes of non zero frequency.
+     */
+    while (s->heap_len < 2) {
+        node = s->heap[++(s->heap_len)] = (max_code < 2 ? ++max_code : 0);
+        tree[node].Freq = 1;
+        s->depth[node] = 0;
+        s->opt_len--; if (stree) s->static_len -= stree[node].Len;
+        /* node is 0 or 1 so it does not have extra bits */
+    }
+    desc->max_code = max_code;
+
+    /* The elements heap[heap_len/2+1 .. heap_len] are leaves of the tree,
+     * establish sub-heaps of increasing lengths:
+     */
+    for (n = s->heap_len/2; n >= 1; n--) pqdownheap(s, tree, n);
+
+    /* Construct the Huffman tree by repeatedly combining the least two
+     * frequent nodes.
+     */
+    node = elems;              /* next internal node of the tree */
+    do {
+        pqremove(s, tree, n);  /* n = node of least frequency */
+        m = s->heap[SMALLEST]; /* m = node of next least frequency */
+
+        s->heap[--(s->heap_max)] = n; /* keep the nodes sorted by frequency */
+        s->heap[--(s->heap_max)] = m;
+
+        /* Create a new node father of n and m */
+        tree[node].Freq = tree[n].Freq + tree[m].Freq;
+        s->depth[node] = (uch)((s->depth[n] >= s->depth[m] ?
+                                s->depth[n] : s->depth[m]) + 1);
+        tree[n].Dad = tree[m].Dad = (ush)node;
+#ifdef DUMP_BL_TREE
+        if (tree == s->bl_tree) {
+            fprintf(stderr,"\nnode %d(%d), sons %d(%d) %d(%d)",
+                    node, tree[node].Freq, n, tree[n].Freq, m, tree[m].Freq);
+        }
+#endif
+        /* and insert the new node in the heap */
+        s->heap[SMALLEST] = node++;
+        pqdownheap(s, tree, SMALLEST);
+
+    } while (s->heap_len >= 2);
+
+    s->heap[--(s->heap_max)] = s->heap[SMALLEST];
+
+    /* At this point, the fields freq and dad are set. We can now
+     * generate the bit lengths.
+     */
+    gen_bitlen(s, (tree_desc *)desc);
+
+    /* The field len is now set, we can generate the bit codes */
+    gen_codes ((ct_data *)tree, max_code, s->bl_count);
+}
+
+/* ===========================================================================
+ * Scan a literal or distance tree to determine the frequencies of the codes
+ * in the bit length tree.
+ */
+local void scan_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree;   /* the tree to be scanned */
+    int max_code;    /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    if (nextlen == 0) max_count = 138, min_count = 3;
+    tree[max_code+1].Len = (ush)0xffff; /* guard */
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            s->bl_tree[curlen].Freq += count;
+        } else if (curlen != 0) {
+            if (curlen != prevlen) s->bl_tree[curlen].Freq++;
+            s->bl_tree[REP_3_6].Freq++;
+        } else if (count <= 10) {
+            s->bl_tree[REPZ_3_10].Freq++;
+        } else {
+            s->bl_tree[REPZ_11_138].Freq++;
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Send a literal or distance tree in compressed form, using the codes in
+ * bl_tree.
+ */
+local void send_tree (s, tree, max_code)
+    deflate_state *s;
+    ct_data *tree; /* the tree to be scanned */
+    int max_code;       /* and its largest code of non zero frequency */
+{
+    int n;                     /* iterates over all tree elements */
+    int prevlen = -1;          /* last emitted length */
+    int curlen;                /* length of current code */
+    int nextlen = tree[0].Len; /* length of next code */
+    int count = 0;             /* repeat count of the current code */
+    int max_count = 7;         /* max repeat count */
+    int min_count = 4;         /* min repeat count */
+
+    /* tree[max_code+1].Len = -1; */  /* guard already set */
+    if (nextlen == 0) max_count = 138, min_count = 3;
+
+    for (n = 0; n <= max_code; n++) {
+        curlen = nextlen; nextlen = tree[n+1].Len;
+        if (++count < max_count && curlen == nextlen) {
+            continue;
+        } else if (count < min_count) {
+            do { send_code(s, curlen, s->bl_tree); } while (--count != 0);
+
+        } else if (curlen != 0) {
+            if (curlen != prevlen) {
+                send_code(s, curlen, s->bl_tree); count--;
+            }
+            Assert(count >= 3 && count <= 6, " 3_6?");
+            send_code(s, REP_3_6, s->bl_tree); send_bits(s, count-3, 2);
+
+        } else if (count <= 10) {
+            send_code(s, REPZ_3_10, s->bl_tree); send_bits(s, count-3, 3);
+
+        } else {
+            send_code(s, REPZ_11_138, s->bl_tree); send_bits(s, count-11, 7);
+        }
+        count = 0; prevlen = curlen;
+        if (nextlen == 0) {
+            max_count = 138, min_count = 3;
+        } else if (curlen == nextlen) {
+            max_count = 6, min_count = 3;
+        } else {
+            max_count = 7, min_count = 4;
+        }
+    }
+}
+
+/* ===========================================================================
+ * Construct the Huffman tree for the bit lengths and return the index in
+ * bl_order of the last bit length code to send.
+ */
+local int build_bl_tree(s)
+    deflate_state *s;
+{
+    int max_blindex;  /* index of last bit length code of non zero freq */
+
+    /* Determine the bit length frequencies for literal and distance trees */
+    scan_tree(s, (ct_data *)s->dyn_ltree, s->l_desc.max_code);
+    scan_tree(s, (ct_data *)s->dyn_dtree, s->d_desc.max_code);
+
+    /* Build the bit length tree: */
+    build_tree(s, (tree_desc *)(&(s->bl_desc)));
+    /* opt_len now includes the length of the tree representations, except
+     * the lengths of the bit lengths codes and the 5+5+4 bits for the counts.
+     */
+
+    /* Determine the number of bit length codes to send. The pkzip format
+     * requires that at least 4 bit length codes be sent. (appnote.txt says
+     * 3 but the actual value used is 4.)
+     */
+    for (max_blindex = BL_CODES-1; max_blindex >= 3; max_blindex--) {
+        if (s->bl_tree[bl_order[max_blindex]].Len != 0) break;
+    }
+    /* Update opt_len to include the bit length tree and counts */
+    s->opt_len += 3*((ulg)max_blindex+1) + 5+5+4;
+    Tracev((stderr, "\ndyn trees: dyn %ld, stat %ld",
+            s->opt_len, s->static_len));
+
+    return max_blindex;
+}
+
+/* ===========================================================================
+ * Send the header for a block using dynamic Huffman trees: the counts, the
+ * lengths of the bit length codes, the literal tree and the distance tree.
+ * IN assertion: lcodes >= 257, dcodes >= 1, blcodes >= 4.
+ */
+local void send_all_trees(s, lcodes, dcodes, blcodes)
+    deflate_state *s;
+    int lcodes, dcodes, blcodes; /* number of codes for each tree */
+{
+    int rank;                    /* index in bl_order */
+
+    Assert (lcodes >= 257 && dcodes >= 1 && blcodes >= 4, "not enough codes");
+    Assert (lcodes <= L_CODES && dcodes <= D_CODES && blcodes <= BL_CODES,
+            "too many codes");
+    Tracev((stderr, "\nbl counts: "));
+    send_bits(s, lcodes-257, 5); /* not +255 as stated in appnote.txt */
+    send_bits(s, dcodes-1,   5);
+    send_bits(s, blcodes-4,  4); /* not -3 as stated in appnote.txt */
+    for (rank = 0; rank < blcodes; rank++) {
+        Tracev((stderr, "\nbl code %2d ", bl_order[rank]));
+        send_bits(s, s->bl_tree[bl_order[rank]].Len, 3);
+    }
+    Tracev((stderr, "\nbl tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_ltree, lcodes-1); /* literal tree */
+    Tracev((stderr, "\nlit tree: sent %ld", s->bits_sent));
+
+    send_tree(s, (ct_data *)s->dyn_dtree, dcodes-1); /* distance tree */
+    Tracev((stderr, "\ndist tree: sent %ld", s->bits_sent));
+}
+
+/* ===========================================================================
+ * Send a stored block
+ */
+void ZLIB_INTERNAL _tr_stored_block(s, buf, stored_len, last)
+    deflate_state *s;
+    charf *buf;       /* input block */
+    ulg stored_len;   /* length of input block */
+    int last;         /* one if this is the last block for a file */
+{
+    send_bits(s, (STORED_BLOCK<<1)+last, 3);    /* send block type */
+    bi_windup(s);        /* align on byte boundary */
+    put_short(s, (ush)stored_len);
+    put_short(s, (ush)~stored_len);
+    zmemcpy(s->pending_buf + s->pending, (Bytef *)buf, stored_len);
+    s->pending += stored_len;
+#ifdef ZLIB_DEBUG
+    s->compressed_len = (s->compressed_len + 3 + 7) & (ulg)~7L;
+    s->compressed_len += (stored_len + 4) << 3;
+    s->bits_sent += 2*16;
+    s->bits_sent += stored_len<<3;
+#endif
+}
+
+/* ===========================================================================
+ * Flush the bits in the bit buffer to pending output (leaves at most 7 bits)
+ */
+void ZLIB_INTERNAL _tr_flush_bits(s)
+    deflate_state *s;
+{
+    bi_flush(s);
+}
+
+/* ===========================================================================
+ * Send one empty static block to give enough lookahead for inflate.
+ * This takes 10 bits, of which 7 may remain in the bit buffer.
+ */
+void ZLIB_INTERNAL _tr_align(s)
+    deflate_state *s;
+{
+    send_bits(s, STATIC_TREES<<1, 3);
+    send_code(s, END_BLOCK, static_ltree);
+#ifdef ZLIB_DEBUG
+    s->compressed_len += 10L; /* 3 for block type, 7 for EOB */
+#endif
+    bi_flush(s);
+}
+
+/* ===========================================================================
+ * Determine the best encoding for the current block: dynamic trees, static
+ * trees or store, and write out the encoded block.
+ */
+void ZLIB_INTERNAL _tr_flush_block(s, buf, stored_len, last)
+    deflate_state *s;
+    charf *buf;       /* input block, or NULL if too old */
+    ulg stored_len;   /* length of input block */
+    int last;         /* one if this is the last block for a file */
+{
+    ulg opt_lenb, static_lenb; /* opt_len and static_len in bytes */
+    int max_blindex = 0;  /* index of last bit length code of non zero freq */
+
+    /* Build the Huffman trees unless a stored block is forced */
+    if (s->level > 0) {
+
+        /* Check if the file is binary or text */
+        if (s->strm->data_type == Z_UNKNOWN)
+            s->strm->data_type = detect_data_type(s);
+
+        /* Construct the literal and distance trees */
+        build_tree(s, (tree_desc *)(&(s->l_desc)));
+        Tracev((stderr, "\nlit data: dyn %ld, stat %ld", s->opt_len,
+                s->static_len));
+
+        build_tree(s, (tree_desc *)(&(s->d_desc)));
+        Tracev((stderr, "\ndist data: dyn %ld, stat %ld", s->opt_len,
+                s->static_len));
+        /* At this point, opt_len and static_len are the total bit lengths of
+         * the compressed block data, excluding the tree representations.
+         */
+
+        /* Build the bit length tree for the above two trees, and get the index
+         * in bl_order of the last bit length code to send.
+         */
+        max_blindex = build_bl_tree(s);
+
+        /* Determine the best encoding. Compute the block lengths in bytes. */
+        opt_lenb = (s->opt_len+3+7)>>3;
+        static_lenb = (s->static_len+3+7)>>3;
+
+        Tracev((stderr, "\nopt %lu(%lu) stat %lu(%lu) stored %lu lit %u ",
+                opt_lenb, s->opt_len, static_lenb, s->static_len, stored_len,
+                s->last_lit));
+
+        if (static_lenb <= opt_lenb) opt_lenb = static_lenb;
+
+    } else {
+        Assert(buf != (char*)0, "lost buf");
+        opt_lenb = static_lenb = stored_len + 5; /* force a stored block */
+    }
+
+#ifdef FORCE_STORED
+    if (buf != (char*)0) { /* force stored block */
+#else
+    if (stored_len+4 <= opt_lenb && buf != (char*)0) {
+                       /* 4: two words for the lengths */
+#endif
+        /* The test buf != NULL is only necessary if LIT_BUFSIZE > WSIZE.
+         * Otherwise we can't have processed more than WSIZE input bytes since
+         * the last block flush, because compression would have been
+         * successful. If LIT_BUFSIZE <= WSIZE, it is never too late to
+         * transform a block into a stored block.
+         */
+        _tr_stored_block(s, buf, stored_len, last);
+
+#ifdef FORCE_STATIC
+    } else if (static_lenb >= 0) { /* force static trees */
+#else
+    } else if (s->strategy == Z_FIXED || static_lenb == opt_lenb) {
+#endif
+        send_bits(s, (STATIC_TREES<<1)+last, 3);
+        compress_block(s, (const ct_data *)static_ltree,
+                       (const ct_data *)static_dtree);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 3 + s->static_len;
+#endif
+    } else {
+        send_bits(s, (DYN_TREES<<1)+last, 3);
+        send_all_trees(s, s->l_desc.max_code+1, s->d_desc.max_code+1,
+                       max_blindex+1);
+        compress_block(s, (const ct_data *)s->dyn_ltree,
+                       (const ct_data *)s->dyn_dtree);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 3 + s->opt_len;
+#endif
+    }
+    Assert (s->compressed_len == s->bits_sent, "bad compressed size");
+    /* The above check is made mod 2^32, for files larger than 512 MB
+     * and uLong implemented on 32 bits.
+     */
+    init_block(s);
+
+    if (last) {
+        bi_windup(s);
+#ifdef ZLIB_DEBUG
+        s->compressed_len += 7;  /* align on byte boundary */
+#endif
+    }
+    Tracev((stderr,"\ncomprlen %lu(%lu) ", s->compressed_len>>3,
+           s->compressed_len-7*last));
+}
+
+/* ===========================================================================
+ * Save the match info and tally the frequency counts. Return true if
+ * the current block must be flushed.
+ */
+int ZLIB_INTERNAL _tr_tally (s, dist, lc)
+    deflate_state *s;
+    unsigned dist;  /* distance of matched string */
+    unsigned lc;    /* match length-MIN_MATCH or unmatched char (if dist==0) */
+{
+    s->d_buf[s->last_lit] = (ush)dist;
+    s->l_buf[s->last_lit++] = (uch)lc;
+    if (dist == 0) {
+        /* lc is the unmatched char */
+        s->dyn_ltree[lc].Freq++;
+    } else {
+        s->matches++;
+        /* Here, lc is the match length - MIN_MATCH */
+        dist--;             /* dist = match distance - 1 */
+        Assert((ush)dist < (ush)MAX_DIST(s) &&
+               (ush)lc <= (ush)(MAX_MATCH-MIN_MATCH) &&
+               (ush)d_code(dist) < (ush)D_CODES,  "_tr_tally: bad match");
+
+        s->dyn_ltree[_length_code[lc]+LITERALS+1].Freq++;
+        s->dyn_dtree[d_code(dist)].Freq++;
+    }
+
+#ifdef TRUNCATE_BLOCK
+    /* Try to guess if it is profitable to stop the current block here */
+    if ((s->last_lit & 0x1fff) == 0 && s->level > 2) {
+        /* Compute an upper bound for the compressed length */
+        ulg out_length = (ulg)s->last_lit*8L;
+        ulg in_length = (ulg)((long)s->strstart - s->block_start);
+        int dcode;
+        for (dcode = 0; dcode < D_CODES; dcode++) {
+            out_length += (ulg)s->dyn_dtree[dcode].Freq *
+                (5L+extra_dbits[dcode]);
+        }
+        out_length >>= 3;
+        Tracev((stderr,"\nlast_lit %u, in %ld, out ~%ld(%ld%%) ",
+               s->last_lit, in_length, out_length,
+               100L - out_length*100L/in_length));
+        if (s->matches < s->last_lit/2 && out_length < in_length/2) return 1;
+    }
+#endif
+    return (s->last_lit == s->lit_bufsize-1);
+    /* We avoid equality with lit_bufsize because of wraparound at 64K
+     * on 16 bit machines and because stored blocks are restricted to
+     * 64K-1 bytes.
+     */
+}
+
+/* ===========================================================================
+ * Send the block data compressed using the given Huffman trees
+ */
+local void compress_block(s, ltree, dtree)
+    deflate_state *s;
+    const ct_data *ltree; /* literal tree */
+    const ct_data *dtree; /* distance tree */
+{
+    unsigned dist;      /* distance of matched string */
+    int lc;             /* match length or unmatched char (if dist == 0) */
+    unsigned lx = 0;    /* running index in l_buf */
+    unsigned code;      /* the code to send */
+    int extra;          /* number of extra bits to send */
+
+    if (s->last_lit != 0) do {
+        dist = s->d_buf[lx];
+        lc = s->l_buf[lx++];
+        if (dist == 0) {
+            send_code(s, lc, ltree); /* send a literal byte */
+            Tracecv(isgraph(lc), (stderr," '%c' ", lc));
+        } else {
+            /* Here, lc is the match length - MIN_MATCH */
+            code = _length_code[lc];
+            send_code(s, code+LITERALS+1, ltree); /* send the length code */
+            extra = extra_lbits[code];
+            if (extra != 0) {
+                lc -= base_length[code];
+                send_bits(s, lc, extra);       /* send the extra length bits */
+            }
+            dist--; /* dist is now the match distance - 1 */
+            code = d_code(dist);
+            Assert (code < D_CODES, "bad d_code");
+
+            send_code(s, code, dtree);       /* send the distance code */
+            extra = extra_dbits[code];
+            if (extra != 0) {
+                dist -= (unsigned)base_dist[code];
+                send_bits(s, dist, extra);   /* send the extra distance bits */
+            }
+        } /* literal or match pair ? */
+
+        /* Check that the overlay between pending_buf and d_buf+l_buf is ok: */
+        Assert((uInt)(s->pending) < s->lit_bufsize + 2*lx,
+               "pendingBuf overflow");
+
+    } while (lx < s->last_lit);
+
+    send_code(s, END_BLOCK, ltree);
+}
+
+/* ===========================================================================
+ * Check if the data type is TEXT or BINARY, using the following algorithm:
+ * - TEXT if the two conditions below are satisfied:
+ *    a) There are no non-portable control characters belonging to the
+ *       "black list" (0..6, 14..25, 28..31).
+ *    b) There is at least one printable character belonging to the
+ *       "white list" (9 {TAB}, 10 {LF}, 13 {CR}, 32..255).
+ * - BINARY otherwise.
+ * - The following partially-portable control characters form a
+ *   "gray list" that is ignored in this detection algorithm:
+ *   (7 {BEL}, 8 {BS}, 11 {VT}, 12 {FF}, 26 {SUB}, 27 {ESC}).
+ * IN assertion: the fields Freq of dyn_ltree are set.
+ */
+local int detect_data_type(s)
+    deflate_state *s;
+{
+    /* black_mask is the bit mask of black-listed bytes
+     * set bits 0..6, 14..25, and 28..31
+     * 0xf3ffc07f = binary 11110011111111111100000001111111
+     */
+    unsigned long black_mask = 0xf3ffc07fUL;
+    int n;
+
+    /* Check for non-textual ("black-listed") bytes. */
+    for (n = 0; n <= 31; n++, black_mask >>= 1)
+        if ((black_mask & 1) && (s->dyn_ltree[n].Freq != 0))
+            return Z_BINARY;
+
+    /* Check for textual ("white-listed") bytes. */
+    if (s->dyn_ltree[9].Freq != 0 || s->dyn_ltree[10].Freq != 0
+            || s->dyn_ltree[13].Freq != 0)
+        return Z_TEXT;
+    for (n = 32; n < LITERALS; n++)
+        if (s->dyn_ltree[n].Freq != 0)
+            return Z_TEXT;
+
+    /* There are no "black-listed" or "white-listed" bytes:
+     * this stream either is empty or has tolerated ("gray-listed") bytes only.
+     */
+    return Z_BINARY;
+}
+
+/* ===========================================================================
+ * Reverse the first len bits of a code, using straightforward code (a faster
+ * method would use a table)
+ * IN assertion: 1 <= len <= 15
+ */
+local unsigned bi_reverse(code, len)
+    unsigned code; /* the value to invert */
+    int len;       /* its bit length */
+{
+    register unsigned res = 0;
+    do {
+        res |= code & 1;
+        code >>= 1, res <<= 1;
+    } while (--len > 0);
+    return res >> 1;
+}
+
+/* ===========================================================================
+ * Flush the bit buffer, keeping at most 7 bits in it.
+ */
+local void bi_flush(s)
+    deflate_state *s;
+{
+    if (s->bi_valid == 16) {
+        put_short(s, s->bi_buf);
+        s->bi_buf = 0;
+        s->bi_valid = 0;
+    } else if (s->bi_valid >= 8) {
+        put_byte(s, (Byte)s->bi_buf);
+        s->bi_buf >>= 8;
+        s->bi_valid -= 8;
+    }
+}
+
+/* ===========================================================================
+ * Flush the bit buffer and align the output on a byte boundary
+ */
+local void bi_windup(s)
+    deflate_state *s;
+{
+    if (s->bi_valid > 8) {
+        put_short(s, s->bi_buf);
+    } else if (s->bi_valid > 0) {
+        put_byte(s, (Byte)s->bi_buf);
+    }
+    s->bi_buf = 0;
+    s->bi_valid = 0;
+#ifdef ZLIB_DEBUG
+    s->bits_sent = (s->bits_sent+7) & ~7;
+#endif
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/trees.h glmark2-2021.02/src/zlib/trees.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/trees.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/trees.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,128 @@
+/* header created automatically with -DGEN_TREES_H */
+
+local const ct_data static_ltree[L_CODES+2] = {
+{{ 12},{  8}}, {{140},{  8}}, {{ 76},{  8}}, {{204},{  8}}, {{ 44},{  8}},
+{{172},{  8}}, {{108},{  8}}, {{236},{  8}}, {{ 28},{  8}}, {{156},{  8}},
+{{ 92},{  8}}, {{220},{  8}}, {{ 60},{  8}}, {{188},{  8}}, {{124},{  8}},
+{{252},{  8}}, {{  2},{  8}}, {{130},{  8}}, {{ 66},{  8}}, {{194},{  8}},
+{{ 34},{  8}}, {{162},{  8}}, {{ 98},{  8}}, {{226},{  8}}, {{ 18},{  8}},
+{{146},{  8}}, {{ 82},{  8}}, {{210},{  8}}, {{ 50},{  8}}, {{178},{  8}},
+{{114},{  8}}, {{242},{  8}}, {{ 10},{  8}}, {{138},{  8}}, {{ 74},{  8}},
+{{202},{  8}}, {{ 42},{  8}}, {{170},{  8}}, {{106},{  8}}, {{234},{  8}},
+{{ 26},{  8}}, {{154},{  8}}, {{ 90},{  8}}, {{218},{  8}}, {{ 58},{  8}},
+{{186},{  8}}, {{122},{  8}}, {{250},{  8}}, {{  6},{  8}}, {{134},{  8}},
+{{ 70},{  8}}, {{198},{  8}}, {{ 38},{  8}}, {{166},{  8}}, {{102},{  8}},
+{{230},{  8}}, {{ 22},{  8}}, {{150},{  8}}, {{ 86},{  8}}, {{214},{  8}},
+{{ 54},{  8}}, {{182},{  8}}, {{118},{  8}}, {{246},{  8}}, {{ 14},{  8}},
+{{142},{  8}}, {{ 78},{  8}}, {{206},{  8}}, {{ 46},{  8}}, {{174},{  8}},
+{{110},{  8}}, {{238},{  8}}, {{ 30},{  8}}, {{158},{  8}}, {{ 94},{  8}},
+{{222},{  8}}, {{ 62},{  8}}, {{190},{  8}}, {{126},{  8}}, {{254},{  8}},
+{{  1},{  8}}, {{129},{  8}}, {{ 65},{  8}}, {{193},{  8}}, {{ 33},{  8}},
+{{161},{  8}}, {{ 97},{  8}}, {{225},{  8}}, {{ 17},{  8}}, {{145},{  8}},
+{{ 81},{  8}}, {{209},{  8}}, {{ 49},{  8}}, {{177},{  8}}, {{113},{  8}},
+{{241},{  8}}, {{  9},{  8}}, {{137},{  8}}, {{ 73},{  8}}, {{201},{  8}},
+{{ 41},{  8}}, {{169},{  8}}, {{105},{  8}}, {{233},{  8}}, {{ 25},{  8}},
+{{153},{  8}}, {{ 89},{  8}}, {{217},{  8}}, {{ 57},{  8}}, {{185},{  8}},
+{{121},{  8}}, {{249},{  8}}, {{  5},{  8}}, {{133},{  8}}, {{ 69},{  8}},
+{{197},{  8}}, {{ 37},{  8}}, {{165},{  8}}, {{101},{  8}}, {{229},{  8}},
+{{ 21},{  8}}, {{149},{  8}}, {{ 85},{  8}}, {{213},{  8}}, {{ 53},{  8}},
+{{181},{  8}}, {{117},{  8}}, {{245},{  8}}, {{ 13},{  8}}, {{141},{  8}},
+{{ 77},{  8}}, {{205},{  8}}, {{ 45},{  8}}, {{173},{  8}}, {{109},{  8}},
+{{237},{  8}}, {{ 29},{  8}}, {{157},{  8}}, {{ 93},{  8}}, {{221},{  8}},
+{{ 61},{  8}}, {{189},{  8}}, {{125},{  8}}, {{253},{  8}}, {{ 19},{  9}},
+{{275},{  9}}, {{147},{  9}}, {{403},{  9}}, {{ 83},{  9}}, {{339},{  9}},
+{{211},{  9}}, {{467},{  9}}, {{ 51},{  9}}, {{307},{  9}}, {{179},{  9}},
+{{435},{  9}}, {{115},{  9}}, {{371},{  9}}, {{243},{  9}}, {{499},{  9}},
+{{ 11},{  9}}, {{267},{  9}}, {{139},{  9}}, {{395},{  9}}, {{ 75},{  9}},
+{{331},{  9}}, {{203},{  9}}, {{459},{  9}}, {{ 43},{  9}}, {{299},{  9}},
+{{171},{  9}}, {{427},{  9}}, {{107},{  9}}, {{363},{  9}}, {{235},{  9}},
+{{491},{  9}}, {{ 27},{  9}}, {{283},{  9}}, {{155},{  9}}, {{411},{  9}},
+{{ 91},{  9}}, {{347},{  9}}, {{219},{  9}}, {{475},{  9}}, {{ 59},{  9}},
+{{315},{  9}}, {{187},{  9}}, {{443},{  9}}, {{123},{  9}}, {{379},{  9}},
+{{251},{  9}}, {{507},{  9}}, {{  7},{  9}}, {{263},{  9}}, {{135},{  9}},
+{{391},{  9}}, {{ 71},{  9}}, {{327},{  9}}, {{199},{  9}}, {{455},{  9}},
+{{ 39},{  9}}, {{295},{  9}}, {{167},{  9}}, {{423},{  9}}, {{103},{  9}},
+{{359},{  9}}, {{231},{  9}}, {{487},{  9}}, {{ 23},{  9}}, {{279},{  9}},
+{{151},{  9}}, {{407},{  9}}, {{ 87},{  9}}, {{343},{  9}}, {{215},{  9}},
+{{471},{  9}}, {{ 55},{  9}}, {{311},{  9}}, {{183},{  9}}, {{439},{  9}},
+{{119},{  9}}, {{375},{  9}}, {{247},{  9}}, {{503},{  9}}, {{ 15},{  9}},
+{{271},{  9}}, {{143},{  9}}, {{399},{  9}}, {{ 79},{  9}}, {{335},{  9}},
+{{207},{  9}}, {{463},{  9}}, {{ 47},{  9}}, {{303},{  9}}, {{175},{  9}},
+{{431},{  9}}, {{111},{  9}}, {{367},{  9}}, {{239},{  9}}, {{495},{  9}},
+{{ 31},{  9}}, {{287},{  9}}, {{159},{  9}}, {{415},{  9}}, {{ 95},{  9}},
+{{351},{  9}}, {{223},{  9}}, {{479},{  9}}, {{ 63},{  9}}, {{319},{  9}},
+{{191},{  9}}, {{447},{  9}}, {{127},{  9}}, {{383},{  9}}, {{255},{  9}},
+{{511},{  9}}, {{  0},{  7}}, {{ 64},{  7}}, {{ 32},{  7}}, {{ 96},{  7}},
+{{ 16},{  7}}, {{ 80},{  7}}, {{ 48},{  7}}, {{112},{  7}}, {{  8},{  7}},
+{{ 72},{  7}}, {{ 40},{  7}}, {{104},{  7}}, {{ 24},{  7}}, {{ 88},{  7}},
+{{ 56},{  7}}, {{120},{  7}}, {{  4},{  7}}, {{ 68},{  7}}, {{ 36},{  7}},
+{{100},{  7}}, {{ 20},{  7}}, {{ 84},{  7}}, {{ 52},{  7}}, {{116},{  7}},
+{{  3},{  8}}, {{131},{  8}}, {{ 67},{  8}}, {{195},{  8}}, {{ 35},{  8}},
+{{163},{  8}}, {{ 99},{  8}}, {{227},{  8}}
+};
+
+local const ct_data static_dtree[D_CODES] = {
+{{ 0},{ 5}}, {{16},{ 5}}, {{ 8},{ 5}}, {{24},{ 5}}, {{ 4},{ 5}},
+{{20},{ 5}}, {{12},{ 5}}, {{28},{ 5}}, {{ 2},{ 5}}, {{18},{ 5}},
+{{10},{ 5}}, {{26},{ 5}}, {{ 6},{ 5}}, {{22},{ 5}}, {{14},{ 5}},
+{{30},{ 5}}, {{ 1},{ 5}}, {{17},{ 5}}, {{ 9},{ 5}}, {{25},{ 5}},
+{{ 5},{ 5}}, {{21},{ 5}}, {{13},{ 5}}, {{29},{ 5}}, {{ 3},{ 5}},
+{{19},{ 5}}, {{11},{ 5}}, {{27},{ 5}}, {{ 7},{ 5}}, {{23},{ 5}}
+};
+
+const uch ZLIB_INTERNAL _dist_code[DIST_CODE_LEN] = {
+ 0,  1,  2,  3,  4,  4,  5,  5,  6,  6,  6,  6,  7,  7,  7,  7,  8,  8,  8,  8,
+ 8,  8,  8,  8,  9,  9,  9,  9,  9,  9,  9,  9, 10, 10, 10, 10, 10, 10, 10, 10,
+10, 10, 10, 10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
+11, 11, 11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
+12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13,
+13, 13, 13, 13, 13, 13, 13, 13, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14,
+14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 14, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,
+15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15, 15,  0,  0, 16, 17,
+18, 18, 19, 19, 20, 20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28, 28,
+28, 28, 28, 28, 28, 28, 28, 28, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29,
+29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29, 29
+};
+
+const uch ZLIB_INTERNAL _length_code[MAX_MATCH-MIN_MATCH+1]= {
+ 0,  1,  2,  3,  4,  5,  6,  7,  8,  8,  9,  9, 10, 10, 11, 11, 12, 12, 12, 12,
+13, 13, 13, 13, 14, 14, 14, 14, 15, 15, 15, 15, 16, 16, 16, 16, 16, 16, 16, 16,
+17, 17, 17, 17, 17, 17, 17, 17, 18, 18, 18, 18, 18, 18, 18, 18, 19, 19, 19, 19,
+19, 19, 19, 19, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20, 20,
+21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 21, 22, 22, 22, 22,
+22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23,
+23, 23, 23, 23, 23, 23, 23, 23, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25,
+25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 25, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26, 26,
+26, 26, 26, 26, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27,
+27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 27, 28
+};
+
+local const int base_length[LENGTH_CODES] = {
+0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 20, 24, 28, 32, 40, 48, 56,
+64, 80, 96, 112, 128, 160, 192, 224, 0
+};
+
+local const int base_dist[D_CODES] = {
+    0,     1,     2,     3,     4,     6,     8,    12,    16,    24,
+   32,    48,    64,    96,   128,   192,   256,   384,   512,   768,
+ 1024,  1536,  2048,  3072,  4096,  6144,  8192, 12288, 16384, 24576
+};
+
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/uncompr.c glmark2-2021.02/src/zlib/uncompr.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/uncompr.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/uncompr.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,93 @@
+/* uncompr.c -- decompress a memory buffer
+ * Copyright (C) 1995-2003, 2010, 2014, 2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#define ZLIB_INTERNAL
+#include "zlib.h"
+
+/* ===========================================================================
+     Decompresses the source buffer into the destination buffer.  *sourceLen is
+   the byte length of the source buffer. Upon entry, *destLen is the total size
+   of the destination buffer, which must be large enough to hold the entire
+   uncompressed data. (The size of the uncompressed data must have been saved
+   previously by the compressor and transmitted to the decompressor by some
+   mechanism outside the scope of this compression library.) Upon exit,
+   *destLen is the size of the decompressed data and *sourceLen is the number
+   of source bytes consumed. Upon return, source + *sourceLen points to the
+   first unused input byte.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer, or
+   Z_DATA_ERROR if the input data was corrupted, including if the input data is
+   an incomplete zlib stream.
+*/
+int ZEXPORT uncompress2 (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong *sourceLen;
+{
+    z_stream stream;
+    int err;
+    const uInt max = (uInt)-1;
+    uLong len, left;
+    Byte buf[1];    /* for detection of incomplete stream when *destLen == 0 */
+
+    len = *sourceLen;
+    if (*destLen) {
+        left = *destLen;
+        *destLen = 0;
+    }
+    else {
+        left = 1;
+        dest = buf;
+    }
+
+    stream.next_in = (z_const Bytef *)source;
+    stream.avail_in = 0;
+    stream.zalloc = (alloc_func)0;
+    stream.zfree = (free_func)0;
+    stream.opaque = (voidpf)0;
+
+    err = inflateInit(&stream);
+    if (err != Z_OK) return err;
+
+    stream.next_out = dest;
+    stream.avail_out = 0;
+
+    do {
+        if (stream.avail_out == 0) {
+            stream.avail_out = left > (uLong)max ? max : (uInt)left;
+            left -= stream.avail_out;
+        }
+        if (stream.avail_in == 0) {
+            stream.avail_in = len > (uLong)max ? max : (uInt)len;
+            len -= stream.avail_in;
+        }
+        err = inflate(&stream, Z_NO_FLUSH);
+    } while (err == Z_OK);
+
+    *sourceLen -= len + stream.avail_in;
+    if (dest != buf)
+        *destLen = stream.total_out;
+    else if (stream.total_out && err == Z_BUF_ERROR)
+        left = 1;
+
+    inflateEnd(&stream);
+    return err == Z_STREAM_END ? Z_OK :
+           err == Z_NEED_DICT ? Z_DATA_ERROR  :
+           err == Z_BUF_ERROR && left + stream.avail_out ? Z_DATA_ERROR :
+           err;
+}
+
+int ZEXPORT uncompress (dest, destLen, source, sourceLen)
+    Bytef *dest;
+    uLongf *destLen;
+    const Bytef *source;
+    uLong sourceLen;
+{
+    return uncompress2(dest, destLen, source, &sourceLen);
+}
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/zconf.h glmark2-2021.02/src/zlib/zconf.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/zconf.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/zconf.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,534 @@
+/* zconf.h -- configuration of the zlib compression library
+ * Copyright (C) 1995-2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#ifndef ZCONF_H
+#define ZCONF_H
+
+/*
+ * If you *really* need a unique prefix for all types and library functions,
+ * compile with -DZ_PREFIX. The "standard" zlib should be compiled without it.
+ * Even better than compiling with -DZ_PREFIX would be to use configure to set
+ * this permanently in zconf.h using "./configure --zprefix".
+ */
+#ifdef Z_PREFIX     /* may be set to #if 1 by ./configure */
+#  define Z_PREFIX_SET
+
+/* all linked symbols and init macros */
+#  define _dist_code            z__dist_code
+#  define _length_code          z__length_code
+#  define _tr_align             z__tr_align
+#  define _tr_flush_bits        z__tr_flush_bits
+#  define _tr_flush_block       z__tr_flush_block
+#  define _tr_init              z__tr_init
+#  define _tr_stored_block      z__tr_stored_block
+#  define _tr_tally             z__tr_tally
+#  define adler32               z_adler32
+#  define adler32_combine       z_adler32_combine
+#  define adler32_combine64     z_adler32_combine64
+#  define adler32_z             z_adler32_z
+#  ifndef Z_SOLO
+#    define compress              z_compress
+#    define compress2             z_compress2
+#    define compressBound         z_compressBound
+#  endif
+#  define crc32                 z_crc32
+#  define crc32_combine         z_crc32_combine
+#  define crc32_combine64       z_crc32_combine64
+#  define crc32_z               z_crc32_z
+#  define deflate               z_deflate
+#  define deflateBound          z_deflateBound
+#  define deflateCopy           z_deflateCopy
+#  define deflateEnd            z_deflateEnd
+#  define deflateGetDictionary  z_deflateGetDictionary
+#  define deflateInit           z_deflateInit
+#  define deflateInit2          z_deflateInit2
+#  define deflateInit2_         z_deflateInit2_
+#  define deflateInit_          z_deflateInit_
+#  define deflateParams         z_deflateParams
+#  define deflatePending        z_deflatePending
+#  define deflatePrime          z_deflatePrime
+#  define deflateReset          z_deflateReset
+#  define deflateResetKeep      z_deflateResetKeep
+#  define deflateSetDictionary  z_deflateSetDictionary
+#  define deflateSetHeader      z_deflateSetHeader
+#  define deflateTune           z_deflateTune
+#  define deflate_copyright     z_deflate_copyright
+#  define get_crc_table         z_get_crc_table
+#  ifndef Z_SOLO
+#    define gz_error              z_gz_error
+#    define gz_intmax             z_gz_intmax
+#    define gz_strwinerror        z_gz_strwinerror
+#    define gzbuffer              z_gzbuffer
+#    define gzclearerr            z_gzclearerr
+#    define gzclose               z_gzclose
+#    define gzclose_r             z_gzclose_r
+#    define gzclose_w             z_gzclose_w
+#    define gzdirect              z_gzdirect
+#    define gzdopen               z_gzdopen
+#    define gzeof                 z_gzeof
+#    define gzerror               z_gzerror
+#    define gzflush               z_gzflush
+#    define gzfread               z_gzfread
+#    define gzfwrite              z_gzfwrite
+#    define gzgetc                z_gzgetc
+#    define gzgetc_               z_gzgetc_
+#    define gzgets                z_gzgets
+#    define gzoffset              z_gzoffset
+#    define gzoffset64            z_gzoffset64
+#    define gzopen                z_gzopen
+#    define gzopen64              z_gzopen64
+#    ifdef _WIN32
+#      define gzopen_w              z_gzopen_w
+#    endif
+#    define gzprintf              z_gzprintf
+#    define gzputc                z_gzputc
+#    define gzputs                z_gzputs
+#    define gzread                z_gzread
+#    define gzrewind              z_gzrewind
+#    define gzseek                z_gzseek
+#    define gzseek64              z_gzseek64
+#    define gzsetparams           z_gzsetparams
+#    define gztell                z_gztell
+#    define gztell64              z_gztell64
+#    define gzungetc              z_gzungetc
+#    define gzvprintf             z_gzvprintf
+#    define gzwrite               z_gzwrite
+#  endif
+#  define inflate               z_inflate
+#  define inflateBack           z_inflateBack
+#  define inflateBackEnd        z_inflateBackEnd
+#  define inflateBackInit       z_inflateBackInit
+#  define inflateBackInit_      z_inflateBackInit_
+#  define inflateCodesUsed      z_inflateCodesUsed
+#  define inflateCopy           z_inflateCopy
+#  define inflateEnd            z_inflateEnd
+#  define inflateGetDictionary  z_inflateGetDictionary
+#  define inflateGetHeader      z_inflateGetHeader
+#  define inflateInit           z_inflateInit
+#  define inflateInit2          z_inflateInit2
+#  define inflateInit2_         z_inflateInit2_
+#  define inflateInit_          z_inflateInit_
+#  define inflateMark           z_inflateMark
+#  define inflatePrime          z_inflatePrime
+#  define inflateReset          z_inflateReset
+#  define inflateReset2         z_inflateReset2
+#  define inflateResetKeep      z_inflateResetKeep
+#  define inflateSetDictionary  z_inflateSetDictionary
+#  define inflateSync           z_inflateSync
+#  define inflateSyncPoint      z_inflateSyncPoint
+#  define inflateUndermine      z_inflateUndermine
+#  define inflateValidate       z_inflateValidate
+#  define inflate_copyright     z_inflate_copyright
+#  define inflate_fast          z_inflate_fast
+#  define inflate_table         z_inflate_table
+#  ifndef Z_SOLO
+#    define uncompress            z_uncompress
+#    define uncompress2           z_uncompress2
+#  endif
+#  define zError                z_zError
+#  ifndef Z_SOLO
+#    define zcalloc               z_zcalloc
+#    define zcfree                z_zcfree
+#  endif
+#  define zlibCompileFlags      z_zlibCompileFlags
+#  define zlibVersion           z_zlibVersion
+
+/* all zlib typedefs in zlib.h and zconf.h */
+#  define Byte                  z_Byte
+#  define Bytef                 z_Bytef
+#  define alloc_func            z_alloc_func
+#  define charf                 z_charf
+#  define free_func             z_free_func
+#  ifndef Z_SOLO
+#    define gzFile                z_gzFile
+#  endif
+#  define gz_header             z_gz_header
+#  define gz_headerp            z_gz_headerp
+#  define in_func               z_in_func
+#  define intf                  z_intf
+#  define out_func              z_out_func
+#  define uInt                  z_uInt
+#  define uIntf                 z_uIntf
+#  define uLong                 z_uLong
+#  define uLongf                z_uLongf
+#  define voidp                 z_voidp
+#  define voidpc                z_voidpc
+#  define voidpf                z_voidpf
+
+/* all zlib structs in zlib.h and zconf.h */
+#  define gz_header_s           z_gz_header_s
+#  define internal_state        z_internal_state
+
+#endif
+
+#if defined(__MSDOS__) && !defined(MSDOS)
+#  define MSDOS
+#endif
+#if (defined(OS_2) || defined(__OS2__)) && !defined(OS2)
+#  define OS2
+#endif
+#if defined(_WINDOWS) && !defined(WINDOWS)
+#  define WINDOWS
+#endif
+#if defined(_WIN32) || defined(_WIN32_WCE) || defined(__WIN32__)
+#  ifndef WIN32
+#    define WIN32
+#  endif
+#endif
+#if (defined(MSDOS) || defined(OS2) || defined(WINDOWS)) && !defined(WIN32)
+#  if !defined(__GNUC__) && !defined(__FLAT__) && !defined(__386__)
+#    ifndef SYS16BIT
+#      define SYS16BIT
+#    endif
+#  endif
+#endif
+
+/*
+ * Compile with -DMAXSEG_64K if the alloc function cannot allocate more
+ * than 64k bytes at a time (needed on systems with 16-bit int).
+ */
+#ifdef SYS16BIT
+#  define MAXSEG_64K
+#endif
+#ifdef MSDOS
+#  define UNALIGNED_OK
+#endif
+
+#ifdef __STDC_VERSION__
+#  ifndef STDC
+#    define STDC
+#  endif
+#  if __STDC_VERSION__ >= 199901L
+#    ifndef STDC99
+#      define STDC99
+#    endif
+#  endif
+#endif
+#if !defined(STDC) && (defined(__STDC__) || defined(__cplusplus))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(__GNUC__) || defined(__BORLANDC__))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(MSDOS) || defined(WINDOWS) || defined(WIN32))
+#  define STDC
+#endif
+#if !defined(STDC) && (defined(OS2) || defined(__HOS_AIX__))
+#  define STDC
+#endif
+
+#if defined(__OS400__) && !defined(STDC)    /* iSeries (formerly AS/400). */
+#  define STDC
+#endif
+
+#ifndef STDC
+#  ifndef const /* cannot use !defined(STDC) && !defined(const) on Mac */
+#    define const       /* note: need a more gentle solution here */
+#  endif
+#endif
+
+#if defined(ZLIB_CONST) && !defined(z_const)
+#  define z_const const
+#else
+#  define z_const
+#endif
+
+#ifdef Z_SOLO
+   typedef unsigned long z_size_t;
+#else
+#  define z_longlong long long
+#  if defined(NO_SIZE_T)
+     typedef unsigned NO_SIZE_T z_size_t;
+#  elif defined(STDC)
+#    include <stddef.h>
+     typedef size_t z_size_t;
+#  else
+     typedef unsigned long z_size_t;
+#  endif
+#  undef z_longlong
+#endif
+
+/* Maximum value for memLevel in deflateInit2 */
+#ifndef MAX_MEM_LEVEL
+#  ifdef MAXSEG_64K
+#    define MAX_MEM_LEVEL 8
+#  else
+#    define MAX_MEM_LEVEL 9
+#  endif
+#endif
+
+/* Maximum value for windowBits in deflateInit2 and inflateInit2.
+ * WARNING: reducing MAX_WBITS makes minigzip unable to extract .gz files
+ * created by gzip. (Files created by minigzip can still be extracted by
+ * gzip.)
+ */
+#ifndef MAX_WBITS
+#  define MAX_WBITS   15 /* 32K LZ77 window */
+#endif
+
+/* The memory requirements for deflate are (in bytes):
+            (1 << (windowBits+2)) +  (1 << (memLevel+9))
+ that is: 128K for windowBits=15  +  128K for memLevel = 8  (default values)
+ plus a few kilobytes for small objects. For example, if you want to reduce
+ the default memory requirements from 256K to 128K, compile with
+     make CFLAGS="-O -DMAX_WBITS=14 -DMAX_MEM_LEVEL=7"
+ Of course this will generally degrade compression (there's no free lunch).
+
+   The memory requirements for inflate are (in bytes) 1 << windowBits
+ that is, 32K for windowBits=15 (default value) plus about 7 kilobytes
+ for small objects.
+*/
+
+                        /* Type declarations */
+
+#ifndef OF /* function prototypes */
+#  ifdef STDC
+#    define OF(args)  args
+#  else
+#    define OF(args)  ()
+#  endif
+#endif
+
+#ifndef Z_ARG /* function prototypes for stdarg */
+#  if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#    define Z_ARG(args)  args
+#  else
+#    define Z_ARG(args)  ()
+#  endif
+#endif
+
+/* The following definitions for FAR are needed only for MSDOS mixed
+ * model programming (small or medium model with some far allocations).
+ * This was tested only with MSC; for other MSDOS compilers you may have
+ * to define NO_MEMCPY in zutil.h.  If you don't need the mixed model,
+ * just define FAR to be empty.
+ */
+#ifdef SYS16BIT
+#  if defined(M_I86SM) || defined(M_I86MM)
+     /* MSC small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef _MSC_VER
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#  if (defined(__SMALL__) || defined(__MEDIUM__))
+     /* Turbo C small or medium model */
+#    define SMALL_MEDIUM
+#    ifdef __BORLANDC__
+#      define FAR _far
+#    else
+#      define FAR far
+#    endif
+#  endif
+#endif
+
+#if defined(WINDOWS) || defined(WIN32)
+   /* If building or using zlib as a DLL, define ZLIB_DLL.
+    * This is not mandatory, but it offers a little performance increase.
+    */
+#  ifdef ZLIB_DLL
+#    if defined(WIN32) && (!defined(__BORLANDC__) || (__BORLANDC__ >= 0x500))
+#      ifdef ZLIB_INTERNAL
+#        define ZEXTERN extern __declspec(dllexport)
+#      else
+#        define ZEXTERN extern __declspec(dllimport)
+#      endif
+#    endif
+#  endif  /* ZLIB_DLL */
+   /* If building or using zlib with the WINAPI/WINAPIV calling convention,
+    * define ZLIB_WINAPI.
+    * Caution: the standard ZLIB1.DLL is NOT compiled using ZLIB_WINAPI.
+    */
+#  ifdef ZLIB_WINAPI
+#    ifdef FAR
+#      undef FAR
+#    endif
+#    include <windows.h>
+     /* No need for _export, use ZLIB.DEF instead. */
+     /* For complete Windows compatibility, use WINAPI, not __stdcall. */
+#    define ZEXPORT WINAPI
+#    ifdef WIN32
+#      define ZEXPORTVA WINAPIV
+#    else
+#      define ZEXPORTVA FAR CDECL
+#    endif
+#  endif
+#endif
+
+#if defined (__BEOS__)
+#  ifdef ZLIB_DLL
+#    ifdef ZLIB_INTERNAL
+#      define ZEXPORT   __declspec(dllexport)
+#      define ZEXPORTVA __declspec(dllexport)
+#    else
+#      define ZEXPORT   __declspec(dllimport)
+#      define ZEXPORTVA __declspec(dllimport)
+#    endif
+#  endif
+#endif
+
+#ifndef ZEXTERN
+#  define ZEXTERN extern
+#endif
+#ifndef ZEXPORT
+#  define ZEXPORT
+#endif
+#ifndef ZEXPORTVA
+#  define ZEXPORTVA
+#endif
+
+#ifndef FAR
+#  define FAR
+#endif
+
+#if !defined(__MACTYPES__)
+typedef unsigned char  Byte;  /* 8 bits */
+#endif
+typedef unsigned int   uInt;  /* 16 bits or more */
+typedef unsigned long  uLong; /* 32 bits or more */
+
+#ifdef SMALL_MEDIUM
+   /* Borland C/C++ and some old MSC versions ignore FAR inside typedef */
+#  define Bytef Byte FAR
+#else
+   typedef Byte  FAR Bytef;
+#endif
+typedef char  FAR charf;
+typedef int   FAR intf;
+typedef uInt  FAR uIntf;
+typedef uLong FAR uLongf;
+
+#ifdef STDC
+   typedef void const *voidpc;
+   typedef void FAR   *voidpf;
+   typedef void       *voidp;
+#else
+   typedef Byte const *voidpc;
+   typedef Byte FAR   *voidpf;
+   typedef Byte       *voidp;
+#endif
+
+#if !defined(Z_U4) && !defined(Z_SOLO) && defined(STDC)
+#  include <limits.h>
+#  if (UINT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned
+#  elif (ULONG_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned long
+#  elif (USHRT_MAX == 0xffffffffUL)
+#    define Z_U4 unsigned short
+#  endif
+#endif
+
+#ifdef Z_U4
+   typedef Z_U4 z_crc_t;
+#else
+   typedef unsigned long z_crc_t;
+#endif
+
+#ifdef HAVE_UNISTD_H    /* may be set to #if 1 by ./configure */
+#  define Z_HAVE_UNISTD_H
+#endif
+
+#ifdef HAVE_STDARG_H    /* may be set to #if 1 by ./configure */
+#  define Z_HAVE_STDARG_H
+#endif
+
+#ifdef STDC
+#  ifndef Z_SOLO
+#    include <sys/types.h>      /* for off_t */
+#  endif
+#endif
+
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+#    include <stdarg.h>         /* for va_list */
+#  endif
+#endif
+
+#ifdef _WIN32
+#  ifndef Z_SOLO
+#    include <stddef.h>         /* for wchar_t */
+#  endif
+#endif
+
+/* a little trick to accommodate both "#define _LARGEFILE64_SOURCE" and
+ * "#define _LARGEFILE64_SOURCE 1" as requesting 64-bit operations, (even
+ * though the former does not conform to the LFS document), but considering
+ * both "#undef _LARGEFILE64_SOURCE" and "#define _LARGEFILE64_SOURCE 0" as
+ * equivalently requesting no 64-bit operations
+ */
+#if defined(_LARGEFILE64_SOURCE) && -_LARGEFILE64_SOURCE - -1 == 1
+#  undef _LARGEFILE64_SOURCE
+#endif
+
+#if defined(__WATCOMC__) && !defined(Z_HAVE_UNISTD_H)
+#  define Z_HAVE_UNISTD_H
+#endif
+#ifndef Z_SOLO
+#  if defined(Z_HAVE_UNISTD_H) || defined(_LARGEFILE64_SOURCE)
+#    include <unistd.h>         /* for SEEK_*, off_t, and _LFS64_LARGEFILE */
+#    ifdef VMS
+#      include <unixio.h>       /* for off_t */
+#    endif
+#    ifndef z_off_t
+#      define z_off_t off_t
+#    endif
+#  endif
+#endif
+
+#if defined(_LFS64_LARGEFILE) && _LFS64_LARGEFILE-0
+#  define Z_LFS64
+#endif
+
+#if defined(_LARGEFILE64_SOURCE) && defined(Z_LFS64)
+#  define Z_LARGE64
+#endif
+
+#if defined(_FILE_OFFSET_BITS) && _FILE_OFFSET_BITS-0 == 64 && defined(Z_LFS64)
+#  define Z_WANT64
+#endif
+
+#if !defined(SEEK_SET) && !defined(Z_SOLO)
+#  define SEEK_SET        0       /* Seek from beginning of file.  */
+#  define SEEK_CUR        1       /* Seek from current position.  */
+#  define SEEK_END        2       /* Set file pointer to EOF plus "offset" */
+#endif
+
+#ifndef z_off_t
+#  define z_off_t long
+#endif
+
+#if !defined(_WIN32) && defined(Z_LARGE64)
+#  define z_off64_t off64_t
+#else
+#  if defined(_WIN32) && !defined(__GNUC__) && !defined(Z_SOLO)
+#    define z_off64_t __int64
+#  else
+#    define z_off64_t z_off_t
+#  endif
+#endif
+
+/* MVS linker does not support external names larger than 8 bytes */
+#if defined(__MVS__)
+  #pragma map(deflateInit_,"DEIN")
+  #pragma map(deflateInit2_,"DEIN2")
+  #pragma map(deflateEnd,"DEEND")
+  #pragma map(deflateBound,"DEBND")
+  #pragma map(inflateInit_,"ININ")
+  #pragma map(inflateInit2_,"ININ2")
+  #pragma map(inflateEnd,"INEND")
+  #pragma map(inflateSync,"INSY")
+  #pragma map(inflateSetDictionary,"INSEDI")
+  #pragma map(compressBound,"CMBND")
+  #pragma map(inflate_table,"INTABL")
+  #pragma map(inflate_fast,"INFA")
+  #pragma map(inflate_copyright,"INCOPY")
+#endif
+
+#endif /* ZCONF_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/zlib.h glmark2-2021.02/src/zlib/zlib.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/zlib.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/zlib.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,1912 @@
+/* zlib.h -- interface of the 'zlib' general purpose compression library
+  version 1.2.11, January 15th, 2017
+
+  Copyright (C) 1995-2017 Jean-loup Gailly and Mark Adler
+
+  This software is provided 'as-is', without any express or implied
+  warranty.  In no event will the authors be held liable for any damages
+  arising from the use of this software.
+
+  Permission is granted to anyone to use this software for any purpose,
+  including commercial applications, and to alter it and redistribute it
+  freely, subject to the following restrictions:
+
+  1. The origin of this software must not be misrepresented; you must not
+     claim that you wrote the original software. If you use this software
+     in a product, an acknowledgment in the product documentation would be
+     appreciated but is not required.
+  2. Altered source versions must be plainly marked as such, and must not be
+     misrepresented as being the original software.
+  3. This notice may not be removed or altered from any source distribution.
+
+  Jean-loup Gailly        Mark Adler
+  jloup@gzip.org          madler@alumni.caltech.edu
+
+
+  The data format used by the zlib library is described by RFCs (Request for
+  Comments) 1950 to 1952 in the files http://tools.ietf.org/html/rfc1950
+  (zlib format), rfc1951 (deflate format) and rfc1952 (gzip format).
+*/
+
+#ifndef ZLIB_H
+#define ZLIB_H
+
+#include "zconf.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define ZLIB_VERSION "1.2.11"
+#define ZLIB_VERNUM 0x12b0
+#define ZLIB_VER_MAJOR 1
+#define ZLIB_VER_MINOR 2
+#define ZLIB_VER_REVISION 11
+#define ZLIB_VER_SUBREVISION 0
+
+/*
+    The 'zlib' compression library provides in-memory compression and
+  decompression functions, including integrity checks of the uncompressed data.
+  This version of the library supports only one compression method (deflation)
+  but other algorithms will be added later and will have the same stream
+  interface.
+
+    Compression can be done in a single step if the buffers are large enough,
+  or can be done by repeated calls of the compression function.  In the latter
+  case, the application must provide more input and/or consume the output
+  (providing more output space) before each call.
+
+    The compressed data format used by default by the in-memory functions is
+  the zlib format, which is a zlib wrapper documented in RFC 1950, wrapped
+  around a deflate stream, which is itself documented in RFC 1951.
+
+    The library also supports reading and writing files in gzip (.gz) format
+  with an interface similar to that of stdio using the functions that start
+  with "gz".  The gzip format is different from the zlib format.  gzip is a
+  gzip wrapper, documented in RFC 1952, wrapped around a deflate stream.
+
+    This library can optionally read and write gzip and raw deflate streams in
+  memory as well.
+
+    The zlib format was designed to be compact and fast for use in memory
+  and on communications channels.  The gzip format was designed for single-
+  file compression on file systems, has a larger header than zlib to maintain
+  directory information, and uses a different, slower check method than zlib.
+
+    The library does not install any signal handler.  The decoder checks
+  the consistency of the compressed data, so the library should never crash
+  even in the case of corrupted input.
+*/
+
+typedef voidpf (*alloc_func) OF((voidpf opaque, uInt items, uInt size));
+typedef void   (*free_func)  OF((voidpf opaque, voidpf address));
+
+struct internal_state;
+
+typedef struct z_stream_s {
+    z_const Bytef *next_in;     /* next input byte */
+    uInt     avail_in;  /* number of bytes available at next_in */
+    uLong    total_in;  /* total number of input bytes read so far */
+
+    Bytef    *next_out; /* next output byte will go here */
+    uInt     avail_out; /* remaining free space at next_out */
+    uLong    total_out; /* total number of bytes output so far */
+
+    z_const char *msg;  /* last error message, NULL if no error */
+    struct internal_state FAR *state; /* not visible by applications */
+
+    alloc_func zalloc;  /* used to allocate the internal state */
+    free_func  zfree;   /* used to free the internal state */
+    voidpf     opaque;  /* private data object passed to zalloc and zfree */
+
+    int     data_type;  /* best guess about the data type: binary or text
+                           for deflate, or the decoding state for inflate */
+    uLong   adler;      /* Adler-32 or CRC-32 value of the uncompressed data */
+    uLong   reserved;   /* reserved for future use */
+} z_stream;
+
+typedef z_stream FAR *z_streamp;
+
+/*
+     gzip header information passed to and from zlib routines.  See RFC 1952
+  for more details on the meanings of these fields.
+*/
+typedef struct gz_header_s {
+    int     text;       /* true if compressed data believed to be text */
+    uLong   time;       /* modification time */
+    int     xflags;     /* extra flags (not used when writing a gzip file) */
+    int     os;         /* operating system */
+    Bytef   *extra;     /* pointer to extra field or Z_NULL if none */
+    uInt    extra_len;  /* extra field length (valid if extra != Z_NULL) */
+    uInt    extra_max;  /* space at extra (only when reading header) */
+    Bytef   *name;      /* pointer to zero-terminated file name or Z_NULL */
+    uInt    name_max;   /* space at name (only when reading header) */
+    Bytef   *comment;   /* pointer to zero-terminated comment or Z_NULL */
+    uInt    comm_max;   /* space at comment (only when reading header) */
+    int     hcrc;       /* true if there was or will be a header crc */
+    int     done;       /* true when done reading gzip header (not used
+                           when writing a gzip file) */
+} gz_header;
+
+typedef gz_header FAR *gz_headerp;
+
+/*
+     The application must update next_in and avail_in when avail_in has dropped
+   to zero.  It must update next_out and avail_out when avail_out has dropped
+   to zero.  The application must initialize zalloc, zfree and opaque before
+   calling the init function.  All other fields are set by the compression
+   library and must not be updated by the application.
+
+     The opaque value provided by the application will be passed as the first
+   parameter for calls of zalloc and zfree.  This can be useful for custom
+   memory management.  The compression library attaches no meaning to the
+   opaque value.
+
+     zalloc must return Z_NULL if there is not enough memory for the object.
+   If zlib is used in a multi-threaded application, zalloc and zfree must be
+   thread safe.  In that case, zlib is thread-safe.  When zalloc and zfree are
+   Z_NULL on entry to the initialization function, they are set to internal
+   routines that use the standard library functions malloc() and free().
+
+     On 16-bit systems, the functions zalloc and zfree must be able to allocate
+   exactly 65536 bytes, but will not be required to allocate more than this if
+   the symbol MAXSEG_64K is defined (see zconf.h).  WARNING: On MSDOS, pointers
+   returned by zalloc for objects of exactly 65536 bytes *must* have their
+   offset normalized to zero.  The default allocation function provided by this
+   library ensures this (see zutil.c).  To reduce memory requirements and avoid
+   any allocation of 64K objects, at the expense of compression ratio, compile
+   the library with -DMAX_WBITS=14 (see zconf.h).
+
+     The fields total_in and total_out can be used for statistics or progress
+   reports.  After compression, total_in holds the total size of the
+   uncompressed data and may be saved for use by the decompressor (particularly
+   if the decompressor wants to decompress everything in a single step).
+*/
+
+                        /* constants */
+
+#define Z_NO_FLUSH      0
+#define Z_PARTIAL_FLUSH 1
+#define Z_SYNC_FLUSH    2
+#define Z_FULL_FLUSH    3
+#define Z_FINISH        4
+#define Z_BLOCK         5
+#define Z_TREES         6
+/* Allowed flush values; see deflate() and inflate() below for details */
+
+#define Z_OK            0
+#define Z_STREAM_END    1
+#define Z_NEED_DICT     2
+#define Z_ERRNO        (-1)
+#define Z_STREAM_ERROR (-2)
+#define Z_DATA_ERROR   (-3)
+#define Z_MEM_ERROR    (-4)
+#define Z_BUF_ERROR    (-5)
+#define Z_VERSION_ERROR (-6)
+/* Return codes for the compression/decompression functions. Negative values
+ * are errors, positive values are used for special but normal events.
+ */
+
+#define Z_NO_COMPRESSION         0
+#define Z_BEST_SPEED             1
+#define Z_BEST_COMPRESSION       9
+#define Z_DEFAULT_COMPRESSION  (-1)
+/* compression levels */
+
+#define Z_FILTERED            1
+#define Z_HUFFMAN_ONLY        2
+#define Z_RLE                 3
+#define Z_FIXED               4
+#define Z_DEFAULT_STRATEGY    0
+/* compression strategy; see deflateInit2() below for details */
+
+#define Z_BINARY   0
+#define Z_TEXT     1
+#define Z_ASCII    Z_TEXT   /* for compatibility with 1.2.2 and earlier */
+#define Z_UNKNOWN  2
+/* Possible values of the data_type field for deflate() */
+
+#define Z_DEFLATED   8
+/* The deflate compression method (the only one supported in this version) */
+
+#define Z_NULL  0  /* for initializing zalloc, zfree, opaque */
+
+#define zlib_version zlibVersion()
+/* for compatibility with versions < 1.0.2 */
+
+
+                        /* basic functions */
+
+ZEXTERN const char * ZEXPORT zlibVersion OF((void));
+/* The application can compare zlibVersion and ZLIB_VERSION for consistency.
+   If the first character differs, the library code actually used is not
+   compatible with the zlib.h header file used by the application.  This check
+   is automatically made by deflateInit and inflateInit.
+ */
+
+/*
+ZEXTERN int ZEXPORT deflateInit OF((z_streamp strm, int level));
+
+     Initializes the internal stream state for compression.  The fields
+   zalloc, zfree and opaque must be initialized before by the caller.  If
+   zalloc and zfree are set to Z_NULL, deflateInit updates them to use default
+   allocation functions.
+
+     The compression level must be Z_DEFAULT_COMPRESSION, or between 0 and 9:
+   1 gives best speed, 9 gives best compression, 0 gives no compression at all
+   (the input data is simply copied a block at a time).  Z_DEFAULT_COMPRESSION
+   requests a default compromise between speed and compression (currently
+   equivalent to level 6).
+
+     deflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if level is not a valid compression level, or
+   Z_VERSION_ERROR if the zlib library version (zlib_version) is incompatible
+   with the version assumed by the caller (ZLIB_VERSION).  msg is set to null
+   if there is no error message.  deflateInit does not perform any compression:
+   this will be done by deflate().
+*/
+
+
+ZEXTERN int ZEXPORT deflate OF((z_streamp strm, int flush));
+/*
+    deflate compresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full.  It may introduce
+  some output latency (reading input without producing any output) except when
+  forced to flush.
+
+    The detailed semantics are as follows.  deflate performs one or both of the
+  following actions:
+
+  - Compress more input starting at next_in and update next_in and avail_in
+    accordingly.  If not all input can be processed (because there is not
+    enough room in the output buffer), next_in and avail_in are updated and
+    processing will resume at this point for the next call of deflate().
+
+  - Generate more output starting at next_out and update next_out and avail_out
+    accordingly.  This action is forced if the parameter flush is non zero.
+    Forcing flush frequently degrades the compression ratio, so this parameter
+    should be set only when necessary.  Some output may be provided even if
+    flush is zero.
+
+    Before the call of deflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming more
+  output, and updating avail_in or avail_out accordingly; avail_out should
+  never be zero before the call.  The application can consume the compressed
+  output when it wants, for example when the output buffer is full (avail_out
+  == 0), or after each call of deflate().  If deflate returns Z_OK and with
+  zero avail_out, it must be called again after making room in the output
+  buffer because there might be more output pending. See deflatePending(),
+  which can be used if desired to determine whether or not there is more ouput
+  in that case.
+
+    Normally the parameter flush is set to Z_NO_FLUSH, which allows deflate to
+  decide how much data to accumulate before producing output, in order to
+  maximize compression.
+
+    If the parameter flush is set to Z_SYNC_FLUSH, all pending output is
+  flushed to the output buffer and the output is aligned on a byte boundary, so
+  that the decompressor can get all input data available so far.  (In
+  particular avail_in is zero after the call if enough output space has been
+  provided before the call.) Flushing may degrade compression for some
+  compression algorithms and so it should be used only when necessary.  This
+  completes the current deflate block and follows it with an empty stored block
+  that is three bits plus filler bits to the next byte, followed by four bytes
+  (00 00 ff ff).
+
+    If flush is set to Z_PARTIAL_FLUSH, all pending output is flushed to the
+  output buffer, but the output is not aligned to a byte boundary.  All of the
+  input data so far will be available to the decompressor, as for Z_SYNC_FLUSH.
+  This completes the current deflate block and follows it with an empty fixed
+  codes block that is 10 bits long.  This assures that enough bytes are output
+  in order for the decompressor to finish the block before the empty fixed
+  codes block.
+
+    If flush is set to Z_BLOCK, a deflate block is completed and emitted, as
+  for Z_SYNC_FLUSH, but the output is not aligned on a byte boundary, and up to
+  seven bits of the current block are held to be written as the next byte after
+  the next deflate block is completed.  In this case, the decompressor may not
+  be provided enough bits at this point in order to complete decompression of
+  the data provided so far to the compressor.  It may need to wait for the next
+  block to be emitted.  This is for advanced applications that need to control
+  the emission of deflate blocks.
+
+    If flush is set to Z_FULL_FLUSH, all output is flushed as with
+  Z_SYNC_FLUSH, and the compression state is reset so that decompression can
+  restart from this point if previous compressed data has been damaged or if
+  random access is desired.  Using Z_FULL_FLUSH too often can seriously degrade
+  compression.
+
+    If deflate returns with avail_out == 0, this function must be called again
+  with the same value of the flush parameter and more output space (updated
+  avail_out), until the flush is complete (deflate returns with non-zero
+  avail_out).  In the case of a Z_FULL_FLUSH or Z_SYNC_FLUSH, make sure that
+  avail_out is greater than six to avoid repeated flush markers due to
+  avail_out == 0 on return.
+
+    If the parameter flush is set to Z_FINISH, pending input is processed,
+  pending output is flushed and deflate returns with Z_STREAM_END if there was
+  enough output space.  If deflate returns with Z_OK or Z_BUF_ERROR, this
+  function must be called again with Z_FINISH and more output space (updated
+  avail_out) but no more input data, until it returns with Z_STREAM_END or an
+  error.  After deflate has returned Z_STREAM_END, the only possible operations
+  on the stream are deflateReset or deflateEnd.
+
+    Z_FINISH can be used in the first deflate call after deflateInit if all the
+  compression is to be done in a single step.  In order to complete in one
+  call, avail_out must be at least the value returned by deflateBound (see
+  below).  Then deflate is guaranteed to return Z_STREAM_END.  If not enough
+  output space is provided, deflate will not return Z_STREAM_END, and it must
+  be called again as described above.
+
+    deflate() sets strm->adler to the Adler-32 checksum of all input read
+  so far (that is, total_in bytes).  If a gzip stream is being generated, then
+  strm->adler will be the CRC-32 checksum of the input read so far.  (See
+  deflateInit2 below.)
+
+    deflate() may update strm->data_type if it can make a good guess about
+  the input data type (Z_BINARY or Z_TEXT).  If in doubt, the data is
+  considered binary.  This field is only for information purposes and does not
+  affect the compression algorithm in any manner.
+
+    deflate() returns Z_OK if some progress has been made (more input
+  processed or more output produced), Z_STREAM_END if all input has been
+  consumed and all output has been produced (only when flush is set to
+  Z_FINISH), Z_STREAM_ERROR if the stream state was inconsistent (for example
+  if next_in or next_out was Z_NULL or the state was inadvertently written over
+  by the application), or Z_BUF_ERROR if no progress is possible (for example
+  avail_in or avail_out was zero).  Note that Z_BUF_ERROR is not fatal, and
+  deflate() can be called again with more input and more output space to
+  continue compressing.
+*/
+
+
+ZEXTERN int ZEXPORT deflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any pending
+   output.
+
+     deflateEnd returns Z_OK if success, Z_STREAM_ERROR if the
+   stream state was inconsistent, Z_DATA_ERROR if the stream was freed
+   prematurely (some input or output was discarded).  In the error case, msg
+   may be set but then points to a static string (which must not be
+   deallocated).
+*/
+
+
+/*
+ZEXTERN int ZEXPORT inflateInit OF((z_streamp strm));
+
+     Initializes the internal stream state for decompression.  The fields
+   next_in, avail_in, zalloc, zfree and opaque must be initialized before by
+   the caller.  In the current version of inflate, the provided input is not
+   read or consumed.  The allocation of a sliding window will be deferred to
+   the first call of inflate (if the decompression does not complete on the
+   first call).  If zalloc and zfree are set to Z_NULL, inflateInit updates
+   them to use default allocation functions.
+
+     inflateInit returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_VERSION_ERROR if the zlib library version is incompatible with the
+   version assumed by the caller, or Z_STREAM_ERROR if the parameters are
+   invalid, such as a null pointer to the structure.  msg is set to null if
+   there is no error message.  inflateInit does not perform any decompression.
+   Actual decompression will be done by inflate().  So next_in, and avail_in,
+   next_out, and avail_out are unused and unchanged.  The current
+   implementation of inflateInit() does not process any header information --
+   that is deferred until inflate() is called.
+*/
+
+
+ZEXTERN int ZEXPORT inflate OF((z_streamp strm, int flush));
+/*
+    inflate decompresses as much data as possible, and stops when the input
+  buffer becomes empty or the output buffer becomes full.  It may introduce
+  some output latency (reading input without producing any output) except when
+  forced to flush.
+
+  The detailed semantics are as follows.  inflate performs one or both of the
+  following actions:
+
+  - Decompress more input starting at next_in and update next_in and avail_in
+    accordingly.  If not all input can be processed (because there is not
+    enough room in the output buffer), then next_in and avail_in are updated
+    accordingly, and processing will resume at this point for the next call of
+    inflate().
+
+  - Generate more output starting at next_out and update next_out and avail_out
+    accordingly.  inflate() provides as much output as possible, until there is
+    no more input data or no more space in the output buffer (see below about
+    the flush parameter).
+
+    Before the call of inflate(), the application should ensure that at least
+  one of the actions is possible, by providing more input and/or consuming more
+  output, and updating the next_* and avail_* values accordingly.  If the
+  caller of inflate() does not provide both available input and available
+  output space, it is possible that there will be no progress made.  The
+  application can consume the uncompressed output when it wants, for example
+  when the output buffer is full (avail_out == 0), or after each call of
+  inflate().  If inflate returns Z_OK and with zero avail_out, it must be
+  called again after making room in the output buffer because there might be
+  more output pending.
+
+    The flush parameter of inflate() can be Z_NO_FLUSH, Z_SYNC_FLUSH, Z_FINISH,
+  Z_BLOCK, or Z_TREES.  Z_SYNC_FLUSH requests that inflate() flush as much
+  output as possible to the output buffer.  Z_BLOCK requests that inflate()
+  stop if and when it gets to the next deflate block boundary.  When decoding
+  the zlib or gzip format, this will cause inflate() to return immediately
+  after the header and before the first block.  When doing a raw inflate,
+  inflate() will go ahead and process the first block, and will return when it
+  gets to the end of that block, or when it runs out of data.
+
+    The Z_BLOCK option assists in appending to or combining deflate streams.
+  To assist in this, on return inflate() always sets strm->data_type to the
+  number of unused bits in the last byte taken from strm->next_in, plus 64 if
+  inflate() is currently decoding the last block in the deflate stream, plus
+  128 if inflate() returned immediately after decoding an end-of-block code or
+  decoding the complete header up to just before the first byte of the deflate
+  stream.  The end-of-block will not be indicated until all of the uncompressed
+  data from that block has been written to strm->next_out.  The number of
+  unused bits may in general be greater than seven, except when bit 7 of
+  data_type is set, in which case the number of unused bits will be less than
+  eight.  data_type is set as noted here every time inflate() returns for all
+  flush options, and so can be used to determine the amount of currently
+  consumed input in bits.
+
+    The Z_TREES option behaves as Z_BLOCK does, but it also returns when the
+  end of each deflate block header is reached, before any actual data in that
+  block is decoded.  This allows the caller to determine the length of the
+  deflate block header for later use in random access within a deflate block.
+  256 is added to the value of strm->data_type when inflate() returns
+  immediately after reaching the end of the deflate block header.
+
+    inflate() should normally be called until it returns Z_STREAM_END or an
+  error.  However if all decompression is to be performed in a single step (a
+  single call of inflate), the parameter flush should be set to Z_FINISH.  In
+  this case all pending input is processed and all pending output is flushed;
+  avail_out must be large enough to hold all of the uncompressed data for the
+  operation to complete.  (The size of the uncompressed data may have been
+  saved by the compressor for this purpose.)  The use of Z_FINISH is not
+  required to perform an inflation in one step.  However it may be used to
+  inform inflate that a faster approach can be used for the single inflate()
+  call.  Z_FINISH also informs inflate to not maintain a sliding window if the
+  stream completes, which reduces inflate's memory footprint.  If the stream
+  does not complete, either because not all of the stream is provided or not
+  enough output space is provided, then a sliding window will be allocated and
+  inflate() can be called again to continue the operation as if Z_NO_FLUSH had
+  been used.
+
+     In this implementation, inflate() always flushes as much output as
+  possible to the output buffer, and always uses the faster approach on the
+  first call.  So the effects of the flush parameter in this implementation are
+  on the return value of inflate() as noted below, when inflate() returns early
+  when Z_BLOCK or Z_TREES is used, and when inflate() avoids the allocation of
+  memory for a sliding window when Z_FINISH is used.
+
+     If a preset dictionary is needed after this call (see inflateSetDictionary
+  below), inflate sets strm->adler to the Adler-32 checksum of the dictionary
+  chosen by the compressor and returns Z_NEED_DICT; otherwise it sets
+  strm->adler to the Adler-32 checksum of all output produced so far (that is,
+  total_out bytes) and returns Z_OK, Z_STREAM_END or an error code as described
+  below.  At the end of the stream, inflate() checks that its computed Adler-32
+  checksum is equal to that saved by the compressor and returns Z_STREAM_END
+  only if the checksum is correct.
+
+    inflate() can decompress and check either zlib-wrapped or gzip-wrapped
+  deflate data.  The header type is detected automatically, if requested when
+  initializing with inflateInit2().  Any information contained in the gzip
+  header is not retained unless inflateGetHeader() is used.  When processing
+  gzip-wrapped deflate data, strm->adler32 is set to the CRC-32 of the output
+  produced so far.  The CRC-32 is checked against the gzip trailer, as is the
+  uncompressed length, modulo 2^32.
+
+    inflate() returns Z_OK if some progress has been made (more input processed
+  or more output produced), Z_STREAM_END if the end of the compressed data has
+  been reached and all uncompressed output has been produced, Z_NEED_DICT if a
+  preset dictionary is needed at this point, Z_DATA_ERROR if the input data was
+  corrupted (input stream not conforming to the zlib format or incorrect check
+  value, in which case strm->msg points to a string with a more specific
+  error), Z_STREAM_ERROR if the stream structure was inconsistent (for example
+  next_in or next_out was Z_NULL, or the state was inadvertently written over
+  by the application), Z_MEM_ERROR if there was not enough memory, Z_BUF_ERROR
+  if no progress was possible or if there was not enough room in the output
+  buffer when Z_FINISH is used.  Note that Z_BUF_ERROR is not fatal, and
+  inflate() can be called again with more input and more output space to
+  continue decompressing.  If Z_DATA_ERROR is returned, the application may
+  then call inflateSync() to look for a good compression block if a partial
+  recovery of the data is to be attempted.
+*/
+
+
+ZEXTERN int ZEXPORT inflateEnd OF((z_streamp strm));
+/*
+     All dynamically allocated data structures for this stream are freed.
+   This function discards any unprocessed input and does not flush any pending
+   output.
+
+     inflateEnd returns Z_OK if success, or Z_STREAM_ERROR if the stream state
+   was inconsistent.
+*/
+
+
+                        /* Advanced functions */
+
+/*
+    The following functions are needed only in some special applications.
+*/
+
+/*
+ZEXTERN int ZEXPORT deflateInit2 OF((z_streamp strm,
+                                     int  level,
+                                     int  method,
+                                     int  windowBits,
+                                     int  memLevel,
+                                     int  strategy));
+
+     This is another version of deflateInit with more compression options.  The
+   fields next_in, zalloc, zfree and opaque must be initialized before by the
+   caller.
+
+     The method parameter is the compression method.  It must be Z_DEFLATED in
+   this version of the library.
+
+     The windowBits parameter is the base two logarithm of the window size
+   (the size of the history buffer).  It should be in the range 8..15 for this
+   version of the library.  Larger values of this parameter result in better
+   compression at the expense of memory usage.  The default value is 15 if
+   deflateInit is used instead.
+
+     For the current implementation of deflate(), a windowBits value of 8 (a
+   window size of 256 bytes) is not supported.  As a result, a request for 8
+   will result in 9 (a 512-byte window).  In that case, providing 8 to
+   inflateInit2() will result in an error when the zlib header with 9 is
+   checked against the initialization of inflate().  The remedy is to not use 8
+   with deflateInit2() with this initialization, or at least in that case use 9
+   with inflateInit2().
+
+     windowBits can also be -8..-15 for raw deflate.  In this case, -windowBits
+   determines the window size.  deflate() will then generate raw deflate data
+   with no zlib header or trailer, and will not compute a check value.
+
+     windowBits can also be greater than 15 for optional gzip encoding.  Add
+   16 to windowBits to write a simple gzip header and trailer around the
+   compressed data instead of a zlib wrapper.  The gzip header will have no
+   file name, no extra data, no comment, no modification time (set to zero), no
+   header crc, and the operating system will be set to the appropriate value,
+   if the operating system was determined at compile time.  If a gzip stream is
+   being written, strm->adler is a CRC-32 instead of an Adler-32.
+
+     For raw deflate or gzip encoding, a request for a 256-byte window is
+   rejected as invalid, since only the zlib header provides a means of
+   transmitting the window size to the decompressor.
+
+     The memLevel parameter specifies how much memory should be allocated
+   for the internal compression state.  memLevel=1 uses minimum memory but is
+   slow and reduces compression ratio; memLevel=9 uses maximum memory for
+   optimal speed.  The default value is 8.  See zconf.h for total memory usage
+   as a function of windowBits and memLevel.
+
+     The strategy parameter is used to tune the compression algorithm.  Use the
+   value Z_DEFAULT_STRATEGY for normal data, Z_FILTERED for data produced by a
+   filter (or predictor), Z_HUFFMAN_ONLY to force Huffman encoding only (no
+   string match), or Z_RLE to limit match distances to one (run-length
+   encoding).  Filtered data consists mostly of small values with a somewhat
+   random distribution.  In this case, the compression algorithm is tuned to
+   compress them better.  The effect of Z_FILTERED is to force more Huffman
+   coding and less string matching; it is somewhat intermediate between
+   Z_DEFAULT_STRATEGY and Z_HUFFMAN_ONLY.  Z_RLE is designed to be almost as
+   fast as Z_HUFFMAN_ONLY, but give better compression for PNG image data.  The
+   strategy parameter only affects the compression ratio but not the
+   correctness of the compressed output even if it is not set appropriately.
+   Z_FIXED prevents the use of dynamic Huffman codes, allowing for a simpler
+   decoder for special applications.
+
+     deflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_STREAM_ERROR if any parameter is invalid (such as an invalid
+   method), or Z_VERSION_ERROR if the zlib library version (zlib_version) is
+   incompatible with the version assumed by the caller (ZLIB_VERSION).  msg is
+   set to null if there is no error message.  deflateInit2 does not perform any
+   compression: this will be done by deflate().
+*/
+
+ZEXTERN int ZEXPORT deflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the compression dictionary from the given byte sequence
+   without producing any compressed output.  When using the zlib format, this
+   function must be called immediately after deflateInit, deflateInit2 or
+   deflateReset, and before any call of deflate.  When doing raw deflate, this
+   function must be called either before any call of deflate, or immediately
+   after the completion of a deflate block, i.e. after all input has been
+   consumed and all output has been delivered when using any of the flush
+   options Z_BLOCK, Z_PARTIAL_FLUSH, Z_SYNC_FLUSH, or Z_FULL_FLUSH.  The
+   compressor and decompressor must use exactly the same dictionary (see
+   inflateSetDictionary).
+
+     The dictionary should consist of strings (byte sequences) that are likely
+   to be encountered later in the data to be compressed, with the most commonly
+   used strings preferably put towards the end of the dictionary.  Using a
+   dictionary is most useful when the data to be compressed is short and can be
+   predicted with good accuracy; the data can then be compressed better than
+   with the default empty dictionary.
+
+     Depending on the size of the compression data structures selected by
+   deflateInit or deflateInit2, a part of the dictionary may in effect be
+   discarded, for example if the dictionary is larger than the window size
+   provided in deflateInit or deflateInit2.  Thus the strings most likely to be
+   useful should be put at the end of the dictionary, not at the front.  In
+   addition, the current implementation of deflate will use at most the window
+   size minus 262 bytes of the provided dictionary.
+
+     Upon return of this function, strm->adler is set to the Adler-32 value
+   of the dictionary; the decompressor may later use this value to determine
+   which dictionary has been used by the compressor.  (The Adler-32 value
+   applies to the whole dictionary even if only a subset of the dictionary is
+   actually used by the compressor.) If a raw deflate was requested, then the
+   Adler-32 value is not computed and strm->adler is not set.
+
+     deflateSetDictionary returns Z_OK if success, or Z_STREAM_ERROR if a
+   parameter is invalid (e.g.  dictionary being Z_NULL) or the stream state is
+   inconsistent (for example if deflate has already been called for this stream
+   or if not at a block boundary for raw deflate).  deflateSetDictionary does
+   not perform any compression: this will be done by deflate().
+*/
+
+ZEXTERN int ZEXPORT deflateGetDictionary OF((z_streamp strm,
+                                             Bytef *dictionary,
+                                             uInt  *dictLength));
+/*
+     Returns the sliding dictionary being maintained by deflate.  dictLength is
+   set to the number of bytes in the dictionary, and that many bytes are copied
+   to dictionary.  dictionary must have enough space, where 32768 bytes is
+   always enough.  If deflateGetDictionary() is called with dictionary equal to
+   Z_NULL, then only the dictionary length is returned, and nothing is copied.
+   Similary, if dictLength is Z_NULL, then it is not set.
+
+     deflateGetDictionary() may return a length less than the window size, even
+   when more than the window size in input has been provided. It may return up
+   to 258 bytes less in that case, due to how zlib's implementation of deflate
+   manages the sliding window and lookahead for matches, where matches can be
+   up to 258 bytes long. If the application needs the last window-size bytes of
+   input, then that would need to be saved by the application outside of zlib.
+
+     deflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the
+   stream state is inconsistent.
+*/
+
+ZEXTERN int ZEXPORT deflateCopy OF((z_streamp dest,
+                                    z_streamp source));
+/*
+     Sets the destination stream as a complete copy of the source stream.
+
+     This function can be useful when several compression strategies will be
+   tried, for example when there are several ways of pre-processing the input
+   data with a filter.  The streams that will be discarded should then be freed
+   by calling deflateEnd.  Note that deflateCopy duplicates the internal
+   compression state which can be quite large, so this strategy is slow and can
+   consume lots of memory.
+
+     deflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if the source stream state was inconsistent
+   (such as zalloc being Z_NULL).  msg is left unchanged in both source and
+   destination.
+*/
+
+ZEXTERN int ZEXPORT deflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to deflateEnd followed by deflateInit, but
+   does not free and reallocate the internal compression state.  The stream
+   will leave the compression level and any other attributes that may have been
+   set unchanged.
+
+     deflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL).
+*/
+
+ZEXTERN int ZEXPORT deflateParams OF((z_streamp strm,
+                                      int level,
+                                      int strategy));
+/*
+     Dynamically update the compression level and compression strategy.  The
+   interpretation of level and strategy is as in deflateInit2().  This can be
+   used to switch between compression and straight copy of the input data, or
+   to switch to a different kind of input data requiring a different strategy.
+   If the compression approach (which is a function of the level) or the
+   strategy is changed, and if any input has been consumed in a previous
+   deflate() call, then the input available so far is compressed with the old
+   level and strategy using deflate(strm, Z_BLOCK).  There are three approaches
+   for the compression levels 0, 1..3, and 4..9 respectively.  The new level
+   and strategy will take effect at the next call of deflate().
+
+     If a deflate(strm, Z_BLOCK) is performed by deflateParams(), and it does
+   not have enough output space to complete, then the parameter change will not
+   take effect.  In this case, deflateParams() can be called again with the
+   same parameters and more output space to try again.
+
+     In order to assure a change in the parameters on the first try, the
+   deflate stream should be flushed using deflate() with Z_BLOCK or other flush
+   request until strm.avail_out is not zero, before calling deflateParams().
+   Then no more input data should be provided before the deflateParams() call.
+   If this is done, the old level and strategy will be applied to the data
+   compressed before deflateParams(), and the new level and strategy will be
+   applied to the the data compressed after deflateParams().
+
+     deflateParams returns Z_OK on success, Z_STREAM_ERROR if the source stream
+   state was inconsistent or if a parameter was invalid, or Z_BUF_ERROR if
+   there was not enough output space to complete the compression of the
+   available input data before a change in the strategy or approach.  Note that
+   in the case of a Z_BUF_ERROR, the parameters are not changed.  A return
+   value of Z_BUF_ERROR is not fatal, in which case deflateParams() can be
+   retried with more output space.
+*/
+
+ZEXTERN int ZEXPORT deflateTune OF((z_streamp strm,
+                                    int good_length,
+                                    int max_lazy,
+                                    int nice_length,
+                                    int max_chain));
+/*
+     Fine tune deflate's internal compression parameters.  This should only be
+   used by someone who understands the algorithm used by zlib's deflate for
+   searching for the best matching string, and even then only by the most
+   fanatic optimizer trying to squeeze out the last compressed bit for their
+   specific input data.  Read the deflate.c source code for the meaning of the
+   max_lazy, good_length, nice_length, and max_chain parameters.
+
+     deflateTune() can be called after deflateInit() or deflateInit2(), and
+   returns Z_OK on success, or Z_STREAM_ERROR for an invalid deflate stream.
+ */
+
+ZEXTERN uLong ZEXPORT deflateBound OF((z_streamp strm,
+                                       uLong sourceLen));
+/*
+     deflateBound() returns an upper bound on the compressed size after
+   deflation of sourceLen bytes.  It must be called after deflateInit() or
+   deflateInit2(), and after deflateSetHeader(), if used.  This would be used
+   to allocate an output buffer for deflation in a single pass, and so would be
+   called before deflate().  If that first deflate() call is provided the
+   sourceLen input bytes, an output buffer allocated to the size returned by
+   deflateBound(), and the flush value Z_FINISH, then deflate() is guaranteed
+   to return Z_STREAM_END.  Note that it is possible for the compressed size to
+   be larger than the value returned by deflateBound() if flush options other
+   than Z_FINISH or Z_NO_FLUSH are used.
+*/
+
+ZEXTERN int ZEXPORT deflatePending OF((z_streamp strm,
+                                       unsigned *pending,
+                                       int *bits));
+/*
+     deflatePending() returns the number of bytes and bits of output that have
+   been generated, but not yet provided in the available output.  The bytes not
+   provided would be due to the available output space having being consumed.
+   The number of bits of output not provided are between 0 and 7, where they
+   await more bits to join them in order to fill out a full byte.  If pending
+   or bits are Z_NULL, then those values are not set.
+
+     deflatePending returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+ */
+
+ZEXTERN int ZEXPORT deflatePrime OF((z_streamp strm,
+                                     int bits,
+                                     int value));
+/*
+     deflatePrime() inserts bits in the deflate output stream.  The intent
+   is that this function is used to start off the deflate output with the bits
+   leftover from a previous deflate stream when appending to it.  As such, this
+   function can only be used for raw deflate, and must be used before the first
+   deflate() call after a deflateInit2() or deflateReset().  bits must be less
+   than or equal to 16, and that many of the least significant bits of value
+   will be inserted in the output.
+
+     deflatePrime returns Z_OK if success, Z_BUF_ERROR if there was not enough
+   room in the internal buffer to insert the bits, or Z_STREAM_ERROR if the
+   source stream state was inconsistent.
+*/
+
+ZEXTERN int ZEXPORT deflateSetHeader OF((z_streamp strm,
+                                         gz_headerp head));
+/*
+     deflateSetHeader() provides gzip header information for when a gzip
+   stream is requested by deflateInit2().  deflateSetHeader() may be called
+   after deflateInit2() or deflateReset() and before the first call of
+   deflate().  The text, time, os, extra field, name, and comment information
+   in the provided gz_header structure are written to the gzip header (xflag is
+   ignored -- the extra flags are set according to the compression level).  The
+   caller must assure that, if not Z_NULL, name and comment are terminated with
+   a zero byte, and that if extra is not Z_NULL, that extra_len bytes are
+   available there.  If hcrc is true, a gzip header crc is included.  Note that
+   the current versions of the command-line version of gzip (up through version
+   1.3.x) do not support header crc's, and will report that it is a "multi-part
+   gzip file" and give up.
+
+     If deflateSetHeader is not used, the default gzip header has text false,
+   the time set to zero, and os set to 255, with no extra, name, or comment
+   fields.  The gzip header is returned to the default state by deflateReset().
+
+     deflateSetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+/*
+ZEXTERN int ZEXPORT inflateInit2 OF((z_streamp strm,
+                                     int  windowBits));
+
+     This is another version of inflateInit with an extra parameter.  The
+   fields next_in, avail_in, zalloc, zfree and opaque must be initialized
+   before by the caller.
+
+     The windowBits parameter is the base two logarithm of the maximum window
+   size (the size of the history buffer).  It should be in the range 8..15 for
+   this version of the library.  The default value is 15 if inflateInit is used
+   instead.  windowBits must be greater than or equal to the windowBits value
+   provided to deflateInit2() while compressing, or it must be equal to 15 if
+   deflateInit2() was not used.  If a compressed stream with a larger window
+   size is given as input, inflate() will return with the error code
+   Z_DATA_ERROR instead of trying to allocate a larger window.
+
+     windowBits can also be zero to request that inflate use the window size in
+   the zlib header of the compressed stream.
+
+     windowBits can also be -8..-15 for raw inflate.  In this case, -windowBits
+   determines the window size.  inflate() will then process raw deflate data,
+   not looking for a zlib or gzip header, not generating a check value, and not
+   looking for any check values for comparison at the end of the stream.  This
+   is for use with other formats that use the deflate compressed data format
+   such as zip.  Those formats provide their own check values.  If a custom
+   format is developed using the raw deflate format for compressed data, it is
+   recommended that a check value such as an Adler-32 or a CRC-32 be applied to
+   the uncompressed data as is done in the zlib, gzip, and zip formats.  For
+   most applications, the zlib format should be used as is.  Note that comments
+   above on the use in deflateInit2() applies to the magnitude of windowBits.
+
+     windowBits can also be greater than 15 for optional gzip decoding.  Add
+   32 to windowBits to enable zlib and gzip decoding with automatic header
+   detection, or add 16 to decode only the gzip format (the zlib format will
+   return a Z_DATA_ERROR).  If a gzip stream is being decoded, strm->adler is a
+   CRC-32 instead of an Adler-32.  Unlike the gunzip utility and gzread() (see
+   below), inflate() will not automatically decode concatenated gzip streams.
+   inflate() will return Z_STREAM_END at the end of the gzip stream.  The state
+   would need to be reset to continue decoding a subsequent gzip stream.
+
+     inflateInit2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_VERSION_ERROR if the zlib library version is incompatible with the
+   version assumed by the caller, or Z_STREAM_ERROR if the parameters are
+   invalid, such as a null pointer to the structure.  msg is set to null if
+   there is no error message.  inflateInit2 does not perform any decompression
+   apart from possibly reading the zlib header if present: actual decompression
+   will be done by inflate().  (So next_in and avail_in may be modified, but
+   next_out and avail_out are unused and unchanged.) The current implementation
+   of inflateInit2() does not process any header information -- that is
+   deferred until inflate() is called.
+*/
+
+ZEXTERN int ZEXPORT inflateSetDictionary OF((z_streamp strm,
+                                             const Bytef *dictionary,
+                                             uInt  dictLength));
+/*
+     Initializes the decompression dictionary from the given uncompressed byte
+   sequence.  This function must be called immediately after a call of inflate,
+   if that call returned Z_NEED_DICT.  The dictionary chosen by the compressor
+   can be determined from the Adler-32 value returned by that call of inflate.
+   The compressor and decompressor must use exactly the same dictionary (see
+   deflateSetDictionary).  For raw inflate, this function can be called at any
+   time to set the dictionary.  If the provided dictionary is smaller than the
+   window and there is already data in the window, then the provided dictionary
+   will amend what's there.  The application must insure that the dictionary
+   that was used for compression is provided.
+
+     inflateSetDictionary returns Z_OK if success, Z_STREAM_ERROR if a
+   parameter is invalid (e.g.  dictionary being Z_NULL) or the stream state is
+   inconsistent, Z_DATA_ERROR if the given dictionary doesn't match the
+   expected one (incorrect Adler-32 value).  inflateSetDictionary does not
+   perform any decompression: this will be done by subsequent calls of
+   inflate().
+*/
+
+ZEXTERN int ZEXPORT inflateGetDictionary OF((z_streamp strm,
+                                             Bytef *dictionary,
+                                             uInt  *dictLength));
+/*
+     Returns the sliding dictionary being maintained by inflate.  dictLength is
+   set to the number of bytes in the dictionary, and that many bytes are copied
+   to dictionary.  dictionary must have enough space, where 32768 bytes is
+   always enough.  If inflateGetDictionary() is called with dictionary equal to
+   Z_NULL, then only the dictionary length is returned, and nothing is copied.
+   Similary, if dictLength is Z_NULL, then it is not set.
+
+     inflateGetDictionary returns Z_OK on success, or Z_STREAM_ERROR if the
+   stream state is inconsistent.
+*/
+
+ZEXTERN int ZEXPORT inflateSync OF((z_streamp strm));
+/*
+     Skips invalid compressed data until a possible full flush point (see above
+   for the description of deflate with Z_FULL_FLUSH) can be found, or until all
+   available input is skipped.  No output is provided.
+
+     inflateSync searches for a 00 00 FF FF pattern in the compressed data.
+   All full flush points have this pattern, but not all occurrences of this
+   pattern are full flush points.
+
+     inflateSync returns Z_OK if a possible full flush point has been found,
+   Z_BUF_ERROR if no more input was provided, Z_DATA_ERROR if no flush point
+   has been found, or Z_STREAM_ERROR if the stream structure was inconsistent.
+   In the success case, the application may save the current current value of
+   total_in which indicates where valid compressed data was found.  In the
+   error case, the application may repeatedly call inflateSync, providing more
+   input each time, until success or end of the input data.
+*/
+
+ZEXTERN int ZEXPORT inflateCopy OF((z_streamp dest,
+                                    z_streamp source));
+/*
+     Sets the destination stream as a complete copy of the source stream.
+
+     This function can be useful when randomly accessing a large stream.  The
+   first pass through the stream can periodically record the inflate state,
+   allowing restarting inflate at those points when randomly accessing the
+   stream.
+
+     inflateCopy returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_STREAM_ERROR if the source stream state was inconsistent
+   (such as zalloc being Z_NULL).  msg is left unchanged in both source and
+   destination.
+*/
+
+ZEXTERN int ZEXPORT inflateReset OF((z_streamp strm));
+/*
+     This function is equivalent to inflateEnd followed by inflateInit,
+   but does not free and reallocate the internal decompression state.  The
+   stream will keep attributes that may have been set by inflateInit2.
+
+     inflateReset returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL).
+*/
+
+ZEXTERN int ZEXPORT inflateReset2 OF((z_streamp strm,
+                                      int windowBits));
+/*
+     This function is the same as inflateReset, but it also permits changing
+   the wrap and window size requests.  The windowBits parameter is interpreted
+   the same as it is for inflateInit2.  If the window size is changed, then the
+   memory allocated for the window is freed, and the window will be reallocated
+   by inflate() if needed.
+
+     inflateReset2 returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent (such as zalloc or state being Z_NULL), or if
+   the windowBits parameter is invalid.
+*/
+
+ZEXTERN int ZEXPORT inflatePrime OF((z_streamp strm,
+                                     int bits,
+                                     int value));
+/*
+     This function inserts bits in the inflate input stream.  The intent is
+   that this function is used to start inflating at a bit position in the
+   middle of a byte.  The provided bits will be used before any bytes are used
+   from next_in.  This function should only be used with raw inflate, and
+   should be used before the first inflate() call after inflateInit2() or
+   inflateReset().  bits must be less than or equal to 16, and that many of the
+   least significant bits of value will be inserted in the input.
+
+     If bits is negative, then the input stream bit buffer is emptied.  Then
+   inflatePrime() can be called again to put bits in the buffer.  This is used
+   to clear out bits leftover after feeding inflate a block description prior
+   to feeding inflate codes.
+
+     inflatePrime returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+ZEXTERN long ZEXPORT inflateMark OF((z_streamp strm));
+/*
+     This function returns two values, one in the lower 16 bits of the return
+   value, and the other in the remaining upper bits, obtained by shifting the
+   return value down 16 bits.  If the upper value is -1 and the lower value is
+   zero, then inflate() is currently decoding information outside of a block.
+   If the upper value is -1 and the lower value is non-zero, then inflate is in
+   the middle of a stored block, with the lower value equaling the number of
+   bytes from the input remaining to copy.  If the upper value is not -1, then
+   it is the number of bits back from the current bit position in the input of
+   the code (literal or length/distance pair) currently being processed.  In
+   that case the lower value is the number of bytes already emitted for that
+   code.
+
+     A code is being processed if inflate is waiting for more input to complete
+   decoding of the code, or if it has completed decoding but is waiting for
+   more output space to write the literal or match data.
+
+     inflateMark() is used to mark locations in the input data for random
+   access, which may be at bit positions, and to note those cases where the
+   output of a code may span boundaries of random access blocks.  The current
+   location in the input stream can be determined from avail_in and data_type
+   as noted in the description for the Z_BLOCK flush parameter for inflate.
+
+     inflateMark returns the value noted above, or -65536 if the provided
+   source stream state was inconsistent.
+*/
+
+ZEXTERN int ZEXPORT inflateGetHeader OF((z_streamp strm,
+                                         gz_headerp head));
+/*
+     inflateGetHeader() requests that gzip header information be stored in the
+   provided gz_header structure.  inflateGetHeader() may be called after
+   inflateInit2() or inflateReset(), and before the first call of inflate().
+   As inflate() processes the gzip stream, head->done is zero until the header
+   is completed, at which time head->done is set to one.  If a zlib stream is
+   being decoded, then head->done is set to -1 to indicate that there will be
+   no gzip header information forthcoming.  Note that Z_BLOCK or Z_TREES can be
+   used to force inflate() to return immediately after header processing is
+   complete and before any actual data is decompressed.
+
+     The text, time, xflags, and os fields are filled in with the gzip header
+   contents.  hcrc is set to true if there is a header CRC.  (The header CRC
+   was valid if done is set to one.) If extra is not Z_NULL, then extra_max
+   contains the maximum number of bytes to write to extra.  Once done is true,
+   extra_len contains the actual extra field length, and extra contains the
+   extra field, or that field truncated if extra_max is less than extra_len.
+   If name is not Z_NULL, then up to name_max characters are written there,
+   terminated with a zero unless the length is greater than name_max.  If
+   comment is not Z_NULL, then up to comm_max characters are written there,
+   terminated with a zero unless the length is greater than comm_max.  When any
+   of extra, name, or comment are not Z_NULL and the respective field is not
+   present in the header, then that field is set to Z_NULL to signal its
+   absence.  This allows the use of deflateSetHeader() with the returned
+   structure to duplicate the header.  However if those fields are set to
+   allocated memory, then the application will need to save those pointers
+   elsewhere so that they can be eventually freed.
+
+     If inflateGetHeader is not used, then the header information is simply
+   discarded.  The header is always checked for validity, including the header
+   CRC if present.  inflateReset() will reset the process to discard the header
+   information.  The application would need to call inflateGetHeader() again to
+   retrieve the header from the next gzip stream.
+
+     inflateGetHeader returns Z_OK if success, or Z_STREAM_ERROR if the source
+   stream state was inconsistent.
+*/
+
+/*
+ZEXTERN int ZEXPORT inflateBackInit OF((z_streamp strm, int windowBits,
+                                        unsigned char FAR *window));
+
+     Initialize the internal stream state for decompression using inflateBack()
+   calls.  The fields zalloc, zfree and opaque in strm must be initialized
+   before the call.  If zalloc and zfree are Z_NULL, then the default library-
+   derived memory allocation routines are used.  windowBits is the base two
+   logarithm of the window size, in the range 8..15.  window is a caller
+   supplied buffer of that size.  Except for special applications where it is
+   assured that deflate was used with small window sizes, windowBits must be 15
+   and a 32K byte window must be supplied to be able to decompress general
+   deflate streams.
+
+     See inflateBack() for the usage of these routines.
+
+     inflateBackInit will return Z_OK on success, Z_STREAM_ERROR if any of
+   the parameters are invalid, Z_MEM_ERROR if the internal state could not be
+   allocated, or Z_VERSION_ERROR if the version of the library does not match
+   the version of the header file.
+*/
+
+typedef unsigned (*in_func) OF((void FAR *,
+                                z_const unsigned char FAR * FAR *));
+typedef int (*out_func) OF((void FAR *, unsigned char FAR *, unsigned));
+
+ZEXTERN int ZEXPORT inflateBack OF((z_streamp strm,
+                                    in_func in, void FAR *in_desc,
+                                    out_func out, void FAR *out_desc));
+/*
+     inflateBack() does a raw inflate with a single call using a call-back
+   interface for input and output.  This is potentially more efficient than
+   inflate() for file i/o applications, in that it avoids copying between the
+   output and the sliding window by simply making the window itself the output
+   buffer.  inflate() can be faster on modern CPUs when used with large
+   buffers.  inflateBack() trusts the application to not change the output
+   buffer passed by the output function, at least until inflateBack() returns.
+
+     inflateBackInit() must be called first to allocate the internal state
+   and to initialize the state with the user-provided window buffer.
+   inflateBack() may then be used multiple times to inflate a complete, raw
+   deflate stream with each call.  inflateBackEnd() is then called to free the
+   allocated state.
+
+     A raw deflate stream is one with no zlib or gzip header or trailer.
+   This routine would normally be used in a utility that reads zip or gzip
+   files and writes out uncompressed files.  The utility would decode the
+   header and process the trailer on its own, hence this routine expects only
+   the raw deflate stream to decompress.  This is different from the default
+   behavior of inflate(), which expects a zlib header and trailer around the
+   deflate stream.
+
+     inflateBack() uses two subroutines supplied by the caller that are then
+   called by inflateBack() for input and output.  inflateBack() calls those
+   routines until it reads a complete deflate stream and writes out all of the
+   uncompressed data, or until it encounters an error.  The function's
+   parameters and return types are defined above in the in_func and out_func
+   typedefs.  inflateBack() will call in(in_desc, &buf) which should return the
+   number of bytes of provided input, and a pointer to that input in buf.  If
+   there is no input available, in() must return zero -- buf is ignored in that
+   case -- and inflateBack() will return a buffer error.  inflateBack() will
+   call out(out_desc, buf, len) to write the uncompressed data buf[0..len-1].
+   out() should return zero on success, or non-zero on failure.  If out()
+   returns non-zero, inflateBack() will return with an error.  Neither in() nor
+   out() are permitted to change the contents of the window provided to
+   inflateBackInit(), which is also the buffer that out() uses to write from.
+   The length written by out() will be at most the window size.  Any non-zero
+   amount of input may be provided by in().
+
+     For convenience, inflateBack() can be provided input on the first call by
+   setting strm->next_in and strm->avail_in.  If that input is exhausted, then
+   in() will be called.  Therefore strm->next_in must be initialized before
+   calling inflateBack().  If strm->next_in is Z_NULL, then in() will be called
+   immediately for input.  If strm->next_in is not Z_NULL, then strm->avail_in
+   must also be initialized, and then if strm->avail_in is not zero, input will
+   initially be taken from strm->next_in[0 ..  strm->avail_in - 1].
+
+     The in_desc and out_desc parameters of inflateBack() is passed as the
+   first parameter of in() and out() respectively when they are called.  These
+   descriptors can be optionally used to pass any information that the caller-
+   supplied in() and out() functions need to do their job.
+
+     On return, inflateBack() will set strm->next_in and strm->avail_in to
+   pass back any unused input that was provided by the last in() call.  The
+   return values of inflateBack() can be Z_STREAM_END on success, Z_BUF_ERROR
+   if in() or out() returned an error, Z_DATA_ERROR if there was a format error
+   in the deflate stream (in which case strm->msg is set to indicate the nature
+   of the error), or Z_STREAM_ERROR if the stream was not properly initialized.
+   In the case of Z_BUF_ERROR, an input or output error can be distinguished
+   using strm->next_in which will be Z_NULL only if in() returned an error.  If
+   strm->next_in is not Z_NULL, then the Z_BUF_ERROR was due to out() returning
+   non-zero.  (in() will always be called before out(), so strm->next_in is
+   assured to be defined if out() returns non-zero.)  Note that inflateBack()
+   cannot return Z_OK.
+*/
+
+ZEXTERN int ZEXPORT inflateBackEnd OF((z_streamp strm));
+/*
+     All memory allocated by inflateBackInit() is freed.
+
+     inflateBackEnd() returns Z_OK on success, or Z_STREAM_ERROR if the stream
+   state was inconsistent.
+*/
+
+ZEXTERN uLong ZEXPORT zlibCompileFlags OF((void));
+/* Return flags indicating compile-time options.
+
+    Type sizes, two bits each, 00 = 16 bits, 01 = 32, 10 = 64, 11 = other:
+     1.0: size of uInt
+     3.2: size of uLong
+     5.4: size of voidpf (pointer)
+     7.6: size of z_off_t
+
+    Compiler, assembler, and debug options:
+     8: ZLIB_DEBUG
+     9: ASMV or ASMINF -- use ASM code
+     10: ZLIB_WINAPI -- exported functions use the WINAPI calling convention
+     11: 0 (reserved)
+
+    One-time table building (smaller code, but not thread-safe if true):
+     12: BUILDFIXED -- build static block decoding tables when needed
+     13: DYNAMIC_CRC_TABLE -- build CRC calculation tables when needed
+     14,15: 0 (reserved)
+
+    Library content (indicates missing functionality):
+     16: NO_GZCOMPRESS -- gz* functions cannot compress (to avoid linking
+                          deflate code when not needed)
+     17: NO_GZIP -- deflate can't write gzip streams, and inflate can't detect
+                    and decode gzip streams (to avoid linking crc code)
+     18-19: 0 (reserved)
+
+    Operation variations (changes in library functionality):
+     20: PKZIP_BUG_WORKAROUND -- slightly more permissive inflate
+     21: FASTEST -- deflate algorithm with only one, lowest compression level
+     22,23: 0 (reserved)
+
+    The sprintf variant used by gzprintf (zero is best):
+     24: 0 = vs*, 1 = s* -- 1 means limited to 20 arguments after the format
+     25: 0 = *nprintf, 1 = *printf -- 1 means gzprintf() not secure!
+     26: 0 = returns value, 1 = void -- 1 means inferred string length returned
+
+    Remainder:
+     27-31: 0 (reserved)
+ */
+
+#ifndef Z_SOLO
+
+                        /* utility functions */
+
+/*
+     The following utility functions are implemented on top of the basic
+   stream-oriented functions.  To simplify the interface, some default options
+   are assumed (compression level and memory usage, standard memory allocation
+   functions).  The source code of these utility functions can be modified if
+   you need special options.
+*/
+
+ZEXTERN int ZEXPORT compress OF((Bytef *dest,   uLongf *destLen,
+                                 const Bytef *source, uLong sourceLen));
+/*
+     Compresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer.  Upon entry, destLen is the total size
+   of the destination buffer, which must be at least the value returned by
+   compressBound(sourceLen).  Upon exit, destLen is the actual size of the
+   compressed data.  compress() is equivalent to compress2() with a level
+   parameter of Z_DEFAULT_COMPRESSION.
+
+     compress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer.
+*/
+
+ZEXTERN int ZEXPORT compress2 OF((Bytef *dest,   uLongf *destLen,
+                                  const Bytef *source, uLong sourceLen,
+                                  int level));
+/*
+     Compresses the source buffer into the destination buffer.  The level
+   parameter has the same meaning as in deflateInit.  sourceLen is the byte
+   length of the source buffer.  Upon entry, destLen is the total size of the
+   destination buffer, which must be at least the value returned by
+   compressBound(sourceLen).  Upon exit, destLen is the actual size of the
+   compressed data.
+
+     compress2 returns Z_OK if success, Z_MEM_ERROR if there was not enough
+   memory, Z_BUF_ERROR if there was not enough room in the output buffer,
+   Z_STREAM_ERROR if the level parameter is invalid.
+*/
+
+ZEXTERN uLong ZEXPORT compressBound OF((uLong sourceLen));
+/*
+     compressBound() returns an upper bound on the compressed size after
+   compress() or compress2() on sourceLen bytes.  It would be used before a
+   compress() or compress2() call to allocate the destination buffer.
+*/
+
+ZEXTERN int ZEXPORT uncompress OF((Bytef *dest,   uLongf *destLen,
+                                   const Bytef *source, uLong sourceLen));
+/*
+     Decompresses the source buffer into the destination buffer.  sourceLen is
+   the byte length of the source buffer.  Upon entry, destLen is the total size
+   of the destination buffer, which must be large enough to hold the entire
+   uncompressed data.  (The size of the uncompressed data must have been saved
+   previously by the compressor and transmitted to the decompressor by some
+   mechanism outside the scope of this compression library.) Upon exit, destLen
+   is the actual size of the uncompressed data.
+
+     uncompress returns Z_OK if success, Z_MEM_ERROR if there was not
+   enough memory, Z_BUF_ERROR if there was not enough room in the output
+   buffer, or Z_DATA_ERROR if the input data was corrupted or incomplete.  In
+   the case where there is not enough room, uncompress() will fill the output
+   buffer with the uncompressed data up to that point.
+*/
+
+ZEXTERN int ZEXPORT uncompress2 OF((Bytef *dest,   uLongf *destLen,
+                                    const Bytef *source, uLong *sourceLen));
+/*
+     Same as uncompress, except that sourceLen is a pointer, where the
+   length of the source is *sourceLen.  On return, *sourceLen is the number of
+   source bytes consumed.
+*/
+
+                        /* gzip file access functions */
+
+/*
+     This library supports reading and writing files in gzip (.gz) format with
+   an interface similar to that of stdio, using the functions that start with
+   "gz".  The gzip format is different from the zlib format.  gzip is a gzip
+   wrapper, documented in RFC 1952, wrapped around a deflate stream.
+*/
+
+typedef struct gzFile_s *gzFile;    /* semi-opaque gzip file descriptor */
+
+/*
+ZEXTERN gzFile ZEXPORT gzopen OF((const char *path, const char *mode));
+
+     Opens a gzip (.gz) file for reading or writing.  The mode parameter is as
+   in fopen ("rb" or "wb") but can also include a compression level ("wb9") or
+   a strategy: 'f' for filtered data as in "wb6f", 'h' for Huffman-only
+   compression as in "wb1h", 'R' for run-length encoding as in "wb1R", or 'F'
+   for fixed code compression as in "wb9F".  (See the description of
+   deflateInit2 for more information about the strategy parameter.)  'T' will
+   request transparent writing or appending with no compression and not using
+   the gzip format.
+
+     "a" can be used instead of "w" to request that the gzip stream that will
+   be written be appended to the file.  "+" will result in an error, since
+   reading and writing to the same gzip file is not supported.  The addition of
+   "x" when writing will create the file exclusively, which fails if the file
+   already exists.  On systems that support it, the addition of "e" when
+   reading or writing will set the flag to close the file on an execve() call.
+
+     These functions, as well as gzip, will read and decode a sequence of gzip
+   streams in a file.  The append function of gzopen() can be used to create
+   such a file.  (Also see gzflush() for another way to do this.)  When
+   appending, gzopen does not test whether the file begins with a gzip stream,
+   nor does it look for the end of the gzip streams to begin appending.  gzopen
+   will simply append a gzip stream to the existing file.
+
+     gzopen can be used to read a file which is not in gzip format; in this
+   case gzread will directly read from the file without decompression.  When
+   reading, this will be detected automatically by looking for the magic two-
+   byte gzip header.
+
+     gzopen returns NULL if the file could not be opened, if there was
+   insufficient memory to allocate the gzFile state, or if an invalid mode was
+   specified (an 'r', 'w', or 'a' was not provided, or '+' was provided).
+   errno can be checked to determine if the reason gzopen failed was that the
+   file could not be opened.
+*/
+
+ZEXTERN gzFile ZEXPORT gzdopen OF((int fd, const char *mode));
+/*
+     gzdopen associates a gzFile with the file descriptor fd.  File descriptors
+   are obtained from calls like open, dup, creat, pipe or fileno (if the file
+   has been previously opened with fopen).  The mode parameter is as in gzopen.
+
+     The next call of gzclose on the returned gzFile will also close the file
+   descriptor fd, just like fclose(fdopen(fd, mode)) closes the file descriptor
+   fd.  If you want to keep fd open, use fd = dup(fd_keep); gz = gzdopen(fd,
+   mode);.  The duplicated descriptor should be saved to avoid a leak, since
+   gzdopen does not close fd if it fails.  If you are using fileno() to get the
+   file descriptor from a FILE *, then you will have to use dup() to avoid
+   double-close()ing the file descriptor.  Both gzclose() and fclose() will
+   close the associated file descriptor, so they need to have different file
+   descriptors.
+
+     gzdopen returns NULL if there was insufficient memory to allocate the
+   gzFile state, if an invalid mode was specified (an 'r', 'w', or 'a' was not
+   provided, or '+' was provided), or if fd is -1.  The file descriptor is not
+   used until the next gz* read, write, seek, or close operation, so gzdopen
+   will not detect if fd is invalid (unless fd is -1).
+*/
+
+ZEXTERN int ZEXPORT gzbuffer OF((gzFile file, unsigned size));
+/*
+     Set the internal buffer size used by this library's functions.  The
+   default buffer size is 8192 bytes.  This function must be called after
+   gzopen() or gzdopen(), and before any other calls that read or write the
+   file.  The buffer memory allocation is always deferred to the first read or
+   write.  Three times that size in buffer space is allocated.  A larger buffer
+   size of, for example, 64K or 128K bytes will noticeably increase the speed
+   of decompression (reading).
+
+     The new buffer size also affects the maximum length for gzprintf().
+
+     gzbuffer() returns 0 on success, or -1 on failure, such as being called
+   too late.
+*/
+
+ZEXTERN int ZEXPORT gzsetparams OF((gzFile file, int level, int strategy));
+/*
+     Dynamically update the compression level or strategy.  See the description
+   of deflateInit2 for the meaning of these parameters.  Previously provided
+   data is flushed before the parameter change.
+
+     gzsetparams returns Z_OK if success, Z_STREAM_ERROR if the file was not
+   opened for writing, Z_ERRNO if there is an error writing the flushed data,
+   or Z_MEM_ERROR if there is a memory allocation error.
+*/
+
+ZEXTERN int ZEXPORT gzread OF((gzFile file, voidp buf, unsigned len));
+/*
+     Reads the given number of uncompressed bytes from the compressed file.  If
+   the input file is not in gzip format, gzread copies the given number of
+   bytes into the buffer directly from the file.
+
+     After reaching the end of a gzip stream in the input, gzread will continue
+   to read, looking for another gzip stream.  Any number of gzip streams may be
+   concatenated in the input file, and will all be decompressed by gzread().
+   If something other than a gzip stream is encountered after a gzip stream,
+   that remaining trailing garbage is ignored (and no error is returned).
+
+     gzread can be used to read a gzip file that is being concurrently written.
+   Upon reaching the end of the input, gzread will return with the available
+   data.  If the error code returned by gzerror is Z_OK or Z_BUF_ERROR, then
+   gzclearerr can be used to clear the end of file indicator in order to permit
+   gzread to be tried again.  Z_OK indicates that a gzip stream was completed
+   on the last gzread.  Z_BUF_ERROR indicates that the input file ended in the
+   middle of a gzip stream.  Note that gzread does not return -1 in the event
+   of an incomplete gzip stream.  This error is deferred until gzclose(), which
+   will return Z_BUF_ERROR if the last gzread ended in the middle of a gzip
+   stream.  Alternatively, gzerror can be used before gzclose to detect this
+   case.
+
+     gzread returns the number of uncompressed bytes actually read, less than
+   len for end of file, or -1 for error.  If len is too large to fit in an int,
+   then nothing is read, -1 is returned, and the error state is set to
+   Z_STREAM_ERROR.
+*/
+
+ZEXTERN z_size_t ZEXPORT gzfread OF((voidp buf, z_size_t size, z_size_t nitems,
+                                     gzFile file));
+/*
+     Read up to nitems items of size size from file to buf, otherwise operating
+   as gzread() does.  This duplicates the interface of stdio's fread(), with
+   size_t request and return types.  If the library defines size_t, then
+   z_size_t is identical to size_t.  If not, then z_size_t is an unsigned
+   integer type that can contain a pointer.
+
+     gzfread() returns the number of full items read of size size, or zero if
+   the end of the file was reached and a full item could not be read, or if
+   there was an error.  gzerror() must be consulted if zero is returned in
+   order to determine if there was an error.  If the multiplication of size and
+   nitems overflows, i.e. the product does not fit in a z_size_t, then nothing
+   is read, zero is returned, and the error state is set to Z_STREAM_ERROR.
+
+     In the event that the end of file is reached and only a partial item is
+   available at the end, i.e. the remaining uncompressed data length is not a
+   multiple of size, then the final partial item is nevetheless read into buf
+   and the end-of-file flag is set.  The length of the partial item read is not
+   provided, but could be inferred from the result of gztell().  This behavior
+   is the same as the behavior of fread() implementations in common libraries,
+   but it prevents the direct use of gzfread() to read a concurrently written
+   file, reseting and retrying on end-of-file, when size is not 1.
+*/
+
+ZEXTERN int ZEXPORT gzwrite OF((gzFile file,
+                                voidpc buf, unsigned len));
+/*
+     Writes the given number of uncompressed bytes into the compressed file.
+   gzwrite returns the number of uncompressed bytes written or 0 in case of
+   error.
+*/
+
+ZEXTERN z_size_t ZEXPORT gzfwrite OF((voidpc buf, z_size_t size,
+                                      z_size_t nitems, gzFile file));
+/*
+     gzfwrite() writes nitems items of size size from buf to file, duplicating
+   the interface of stdio's fwrite(), with size_t request and return types.  If
+   the library defines size_t, then z_size_t is identical to size_t.  If not,
+   then z_size_t is an unsigned integer type that can contain a pointer.
+
+     gzfwrite() returns the number of full items written of size size, or zero
+   if there was an error.  If the multiplication of size and nitems overflows,
+   i.e. the product does not fit in a z_size_t, then nothing is written, zero
+   is returned, and the error state is set to Z_STREAM_ERROR.
+*/
+
+ZEXTERN int ZEXPORTVA gzprintf Z_ARG((gzFile file, const char *format, ...));
+/*
+     Converts, formats, and writes the arguments to the compressed file under
+   control of the format string, as in fprintf.  gzprintf returns the number of
+   uncompressed bytes actually written, or a negative zlib error code in case
+   of error.  The number of uncompressed bytes written is limited to 8191, or
+   one less than the buffer size given to gzbuffer().  The caller should assure
+   that this limit is not exceeded.  If it is exceeded, then gzprintf() will
+   return an error (0) with nothing written.  In this case, there may also be a
+   buffer overflow with unpredictable consequences, which is possible only if
+   zlib was compiled with the insecure functions sprintf() or vsprintf()
+   because the secure snprintf() or vsnprintf() functions were not available.
+   This can be determined using zlibCompileFlags().
+*/
+
+ZEXTERN int ZEXPORT gzputs OF((gzFile file, const char *s));
+/*
+     Writes the given null-terminated string to the compressed file, excluding
+   the terminating null character.
+
+     gzputs returns the number of characters written, or -1 in case of error.
+*/
+
+ZEXTERN char * ZEXPORT gzgets OF((gzFile file, char *buf, int len));
+/*
+     Reads bytes from the compressed file until len-1 characters are read, or a
+   newline character is read and transferred to buf, or an end-of-file
+   condition is encountered.  If any characters are read or if len == 1, the
+   string is terminated with a null character.  If no characters are read due
+   to an end-of-file or len < 1, then the buffer is left untouched.
+
+     gzgets returns buf which is a null-terminated string, or it returns NULL
+   for end-of-file or in case of error.  If there was an error, the contents at
+   buf are indeterminate.
+*/
+
+ZEXTERN int ZEXPORT gzputc OF((gzFile file, int c));
+/*
+     Writes c, converted to an unsigned char, into the compressed file.  gzputc
+   returns the value that was written, or -1 in case of error.
+*/
+
+ZEXTERN int ZEXPORT gzgetc OF((gzFile file));
+/*
+     Reads one byte from the compressed file.  gzgetc returns this byte or -1
+   in case of end of file or error.  This is implemented as a macro for speed.
+   As such, it does not do all of the checking the other functions do.  I.e.
+   it does not check to see if file is NULL, nor whether the structure file
+   points to has been clobbered or not.
+*/
+
+ZEXTERN int ZEXPORT gzungetc OF((int c, gzFile file));
+/*
+     Push one character back onto the stream to be read as the first character
+   on the next read.  At least one character of push-back is allowed.
+   gzungetc() returns the character pushed, or -1 on failure.  gzungetc() will
+   fail if c is -1, and may fail if a character has been pushed but not read
+   yet.  If gzungetc is used immediately after gzopen or gzdopen, at least the
+   output buffer size of pushed characters is allowed.  (See gzbuffer above.)
+   The pushed character will be discarded if the stream is repositioned with
+   gzseek() or gzrewind().
+*/
+
+ZEXTERN int ZEXPORT gzflush OF((gzFile file, int flush));
+/*
+     Flushes all pending output into the compressed file.  The parameter flush
+   is as in the deflate() function.  The return value is the zlib error number
+   (see function gzerror below).  gzflush is only permitted when writing.
+
+     If the flush parameter is Z_FINISH, the remaining data is written and the
+   gzip stream is completed in the output.  If gzwrite() is called again, a new
+   gzip stream will be started in the output.  gzread() is able to read such
+   concatenated gzip streams.
+
+     gzflush should be called only when strictly necessary because it will
+   degrade compression if called too often.
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile file,
+                                   z_off_t offset, int whence));
+
+     Sets the starting position for the next gzread or gzwrite on the given
+   compressed file.  The offset represents a number of bytes in the
+   uncompressed data stream.  The whence parameter is defined as in lseek(2);
+   the value SEEK_END is not supported.
+
+     If the file is opened for reading, this function is emulated but can be
+   extremely slow.  If the file is opened for writing, only forward seeks are
+   supported; gzseek then compresses a sequence of zeroes up to the new
+   starting position.
+
+     gzseek returns the resulting offset location as measured in bytes from
+   the beginning of the uncompressed stream, or -1 in case of error, in
+   particular if the file is opened for writing and the new starting position
+   would be before the current position.
+*/
+
+ZEXTERN int ZEXPORT    gzrewind OF((gzFile file));
+/*
+     Rewinds the given file. This function is supported only for reading.
+
+     gzrewind(file) is equivalent to (int)gzseek(file, 0L, SEEK_SET)
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT    gztell OF((gzFile file));
+
+     Returns the starting position for the next gzread or gzwrite on the given
+   compressed file.  This position represents a number of bytes in the
+   uncompressed data stream, and is zero when starting, even if appending or
+   reading a gzip stream from the middle of a file using gzdopen().
+
+     gztell(file) is equivalent to gzseek(file, 0L, SEEK_CUR)
+*/
+
+/*
+ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile file));
+
+     Returns the current offset in the file being read or written.  This offset
+   includes the count of bytes that precede the gzip stream, for example when
+   appending or when using gzdopen() for reading.  When reading, the offset
+   does not include as yet unused buffered input.  This information can be used
+   for a progress indicator.  On error, gzoffset() returns -1.
+*/
+
+ZEXTERN int ZEXPORT gzeof OF((gzFile file));
+/*
+     Returns true (1) if the end-of-file indicator has been set while reading,
+   false (0) otherwise.  Note that the end-of-file indicator is set only if the
+   read tried to go past the end of the input, but came up short.  Therefore,
+   just like feof(), gzeof() may return false even if there is no more data to
+   read, in the event that the last read request was for the exact number of
+   bytes remaining in the input file.  This will happen if the input file size
+   is an exact multiple of the buffer size.
+
+     If gzeof() returns true, then the read functions will return no more data,
+   unless the end-of-file indicator is reset by gzclearerr() and the input file
+   has grown since the previous end of file was detected.
+*/
+
+ZEXTERN int ZEXPORT gzdirect OF((gzFile file));
+/*
+     Returns true (1) if file is being copied directly while reading, or false
+   (0) if file is a gzip stream being decompressed.
+
+     If the input file is empty, gzdirect() will return true, since the input
+   does not contain a gzip stream.
+
+     If gzdirect() is used immediately after gzopen() or gzdopen() it will
+   cause buffers to be allocated to allow reading the file to determine if it
+   is a gzip file.  Therefore if gzbuffer() is used, it should be called before
+   gzdirect().
+
+     When writing, gzdirect() returns true (1) if transparent writing was
+   requested ("wT" for the gzopen() mode), or false (0) otherwise.  (Note:
+   gzdirect() is not needed when writing.  Transparent writing must be
+   explicitly requested, so the application already knows the answer.  When
+   linking statically, using gzdirect() will include all of the zlib code for
+   gzip file reading and decompression, which may not be desired.)
+*/
+
+ZEXTERN int ZEXPORT    gzclose OF((gzFile file));
+/*
+     Flushes all pending output if necessary, closes the compressed file and
+   deallocates the (de)compression state.  Note that once file is closed, you
+   cannot call gzerror with file, since its structures have been deallocated.
+   gzclose must not be called more than once on the same file, just as free
+   must not be called more than once on the same allocation.
+
+     gzclose will return Z_STREAM_ERROR if file is not valid, Z_ERRNO on a
+   file operation error, Z_MEM_ERROR if out of memory, Z_BUF_ERROR if the
+   last read ended in the middle of a gzip stream, or Z_OK on success.
+*/
+
+ZEXTERN int ZEXPORT gzclose_r OF((gzFile file));
+ZEXTERN int ZEXPORT gzclose_w OF((gzFile file));
+/*
+     Same as gzclose(), but gzclose_r() is only for use when reading, and
+   gzclose_w() is only for use when writing or appending.  The advantage to
+   using these instead of gzclose() is that they avoid linking in zlib
+   compression or decompression code that is not used when only reading or only
+   writing respectively.  If gzclose() is used, then both compression and
+   decompression code will be included the application when linking to a static
+   zlib library.
+*/
+
+ZEXTERN const char * ZEXPORT gzerror OF((gzFile file, int *errnum));
+/*
+     Returns the error message for the last error which occurred on the given
+   compressed file.  errnum is set to zlib error number.  If an error occurred
+   in the file system and not in the compression library, errnum is set to
+   Z_ERRNO and the application may consult errno to get the exact error code.
+
+     The application must not modify the returned string.  Future calls to
+   this function may invalidate the previously returned string.  If file is
+   closed, then the string previously returned by gzerror will no longer be
+   available.
+
+     gzerror() should be used to distinguish errors from end-of-file for those
+   functions above that do not distinguish those cases in their return values.
+*/
+
+ZEXTERN void ZEXPORT gzclearerr OF((gzFile file));
+/*
+     Clears the error and end-of-file flags for file.  This is analogous to the
+   clearerr() function in stdio.  This is useful for continuing to read a gzip
+   file that is being written concurrently.
+*/
+
+#endif /* !Z_SOLO */
+
+                        /* checksum functions */
+
+/*
+     These functions are not related to compression but are exported
+   anyway because they might be useful in applications using the compression
+   library.
+*/
+
+ZEXTERN uLong ZEXPORT adler32 OF((uLong adler, const Bytef *buf, uInt len));
+/*
+     Update a running Adler-32 checksum with the bytes buf[0..len-1] and
+   return the updated checksum.  If buf is Z_NULL, this function returns the
+   required initial value for the checksum.
+
+     An Adler-32 checksum is almost as reliable as a CRC-32 but can be computed
+   much faster.
+
+   Usage example:
+
+     uLong adler = adler32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       adler = adler32(adler, buffer, length);
+     }
+     if (adler != original_adler) error();
+*/
+
+ZEXTERN uLong ZEXPORT adler32_z OF((uLong adler, const Bytef *buf,
+                                    z_size_t len));
+/*
+     Same as adler32(), but with a size_t length.
+*/
+
+/*
+ZEXTERN uLong ZEXPORT adler32_combine OF((uLong adler1, uLong adler2,
+                                          z_off_t len2));
+
+     Combine two Adler-32 checksums into one.  For two sequences of bytes, seq1
+   and seq2 with lengths len1 and len2, Adler-32 checksums were calculated for
+   each, adler1 and adler2.  adler32_combine() returns the Adler-32 checksum of
+   seq1 and seq2 concatenated, requiring only adler1, adler2, and len2.  Note
+   that the z_off_t type (like off_t) is a signed integer.  If len2 is
+   negative, the result has no meaning or utility.
+*/
+
+ZEXTERN uLong ZEXPORT crc32   OF((uLong crc, const Bytef *buf, uInt len));
+/*
+     Update a running CRC-32 with the bytes buf[0..len-1] and return the
+   updated CRC-32.  If buf is Z_NULL, this function returns the required
+   initial value for the crc.  Pre- and post-conditioning (one's complement) is
+   performed within this function so it shouldn't be done by the application.
+
+   Usage example:
+
+     uLong crc = crc32(0L, Z_NULL, 0);
+
+     while (read_buffer(buffer, length) != EOF) {
+       crc = crc32(crc, buffer, length);
+     }
+     if (crc != original_crc) error();
+*/
+
+ZEXTERN uLong ZEXPORT crc32_z OF((uLong adler, const Bytef *buf,
+                                  z_size_t len));
+/*
+     Same as crc32(), but with a size_t length.
+*/
+
+/*
+ZEXTERN uLong ZEXPORT crc32_combine OF((uLong crc1, uLong crc2, z_off_t len2));
+
+     Combine two CRC-32 check values into one.  For two sequences of bytes,
+   seq1 and seq2 with lengths len1 and len2, CRC-32 check values were
+   calculated for each, crc1 and crc2.  crc32_combine() returns the CRC-32
+   check value of seq1 and seq2 concatenated, requiring only crc1, crc2, and
+   len2.
+*/
+
+
+                        /* various hacks, don't look :) */
+
+/* deflateInit and inflateInit are macros to allow checking the zlib version
+ * and the compiler's view of z_stream:
+ */
+ZEXTERN int ZEXPORT deflateInit_ OF((z_streamp strm, int level,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT inflateInit_ OF((z_streamp strm,
+                                     const char *version, int stream_size));
+ZEXTERN int ZEXPORT deflateInit2_ OF((z_streamp strm, int  level, int  method,
+                                      int windowBits, int memLevel,
+                                      int strategy, const char *version,
+                                      int stream_size));
+ZEXTERN int ZEXPORT inflateInit2_ OF((z_streamp strm, int  windowBits,
+                                      const char *version, int stream_size));
+ZEXTERN int ZEXPORT inflateBackInit_ OF((z_streamp strm, int windowBits,
+                                         unsigned char FAR *window,
+                                         const char *version,
+                                         int stream_size));
+#ifdef Z_PREFIX_SET
+#  define z_deflateInit(strm, level) \
+          deflateInit_((strm), (level), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_inflateInit(strm) \
+          inflateInit_((strm), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_deflateInit2(strm, level, method, windowBits, memLevel, strategy) \
+          deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\
+                        (strategy), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define z_inflateInit2(strm, windowBits) \
+          inflateInit2_((strm), (windowBits), ZLIB_VERSION, \
+                        (int)sizeof(z_stream))
+#  define z_inflateBackInit(strm, windowBits, window) \
+          inflateBackInit_((strm), (windowBits), (window), \
+                           ZLIB_VERSION, (int)sizeof(z_stream))
+#else
+#  define deflateInit(strm, level) \
+          deflateInit_((strm), (level), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define inflateInit(strm) \
+          inflateInit_((strm), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define deflateInit2(strm, level, method, windowBits, memLevel, strategy) \
+          deflateInit2_((strm),(level),(method),(windowBits),(memLevel),\
+                        (strategy), ZLIB_VERSION, (int)sizeof(z_stream))
+#  define inflateInit2(strm, windowBits) \
+          inflateInit2_((strm), (windowBits), ZLIB_VERSION, \
+                        (int)sizeof(z_stream))
+#  define inflateBackInit(strm, windowBits, window) \
+          inflateBackInit_((strm), (windowBits), (window), \
+                           ZLIB_VERSION, (int)sizeof(z_stream))
+#endif
+
+#ifndef Z_SOLO
+
+/* gzgetc() macro and its supporting function and exposed data structure.  Note
+ * that the real internal state is much larger than the exposed structure.
+ * This abbreviated structure exposes just enough for the gzgetc() macro.  The
+ * user should not mess with these exposed elements, since their names or
+ * behavior could change in the future, perhaps even capriciously.  They can
+ * only be used by the gzgetc() macro.  You have been warned.
+ */
+struct gzFile_s {
+    unsigned have;
+    unsigned char *next;
+    z_off64_t pos;
+};
+ZEXTERN int ZEXPORT gzgetc_ OF((gzFile file));  /* backward compatibility */
+#ifdef Z_PREFIX_SET
+#  undef z_gzgetc
+#  define z_gzgetc(g) \
+          ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : (gzgetc)(g))
+#else
+#  define gzgetc(g) \
+          ((g)->have ? ((g)->have--, (g)->pos++, *((g)->next)++) : (gzgetc)(g))
+#endif
+
+/* provide 64-bit offset functions if _LARGEFILE64_SOURCE defined, and/or
+ * change the regular functions to 64 bits if _FILE_OFFSET_BITS is 64 (if
+ * both are true, the application gets the *64 functions, and the regular
+ * functions are changed to 64 bits) -- in case these are set on systems
+ * without large file support, _LFS64_LARGEFILE must also be true
+ */
+#ifdef Z_LARGE64
+   ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+   ZEXTERN z_off64_t ZEXPORT gzseek64 OF((gzFile, z_off64_t, int));
+   ZEXTERN z_off64_t ZEXPORT gztell64 OF((gzFile));
+   ZEXTERN z_off64_t ZEXPORT gzoffset64 OF((gzFile));
+   ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off64_t));
+   ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off64_t));
+#endif
+
+#if !defined(ZLIB_INTERNAL) && defined(Z_WANT64)
+#  ifdef Z_PREFIX_SET
+#    define z_gzopen z_gzopen64
+#    define z_gzseek z_gzseek64
+#    define z_gztell z_gztell64
+#    define z_gzoffset z_gzoffset64
+#    define z_adler32_combine z_adler32_combine64
+#    define z_crc32_combine z_crc32_combine64
+#  else
+#    define gzopen gzopen64
+#    define gzseek gzseek64
+#    define gztell gztell64
+#    define gzoffset gzoffset64
+#    define adler32_combine adler32_combine64
+#    define crc32_combine crc32_combine64
+#  endif
+#  ifndef Z_LARGE64
+     ZEXTERN gzFile ZEXPORT gzopen64 OF((const char *, const char *));
+     ZEXTERN z_off_t ZEXPORT gzseek64 OF((gzFile, z_off_t, int));
+     ZEXTERN z_off_t ZEXPORT gztell64 OF((gzFile));
+     ZEXTERN z_off_t ZEXPORT gzoffset64 OF((gzFile));
+     ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t));
+     ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t));
+#  endif
+#else
+   ZEXTERN gzFile ZEXPORT gzopen OF((const char *, const char *));
+   ZEXTERN z_off_t ZEXPORT gzseek OF((gzFile, z_off_t, int));
+   ZEXTERN z_off_t ZEXPORT gztell OF((gzFile));
+   ZEXTERN z_off_t ZEXPORT gzoffset OF((gzFile));
+   ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t));
+   ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t));
+#endif
+
+#else /* Z_SOLO */
+
+   ZEXTERN uLong ZEXPORT adler32_combine OF((uLong, uLong, z_off_t));
+   ZEXTERN uLong ZEXPORT crc32_combine OF((uLong, uLong, z_off_t));
+
+#endif /* !Z_SOLO */
+
+/* undocumented functions */
+ZEXTERN const char   * ZEXPORT zError           OF((int));
+ZEXTERN int            ZEXPORT inflateSyncPoint OF((z_streamp));
+ZEXTERN const z_crc_t FAR * ZEXPORT get_crc_table    OF((void));
+ZEXTERN int            ZEXPORT inflateUndermine OF((z_streamp, int));
+ZEXTERN int            ZEXPORT inflateValidate OF((z_streamp, int));
+ZEXTERN unsigned long  ZEXPORT inflateCodesUsed OF ((z_streamp));
+ZEXTERN int            ZEXPORT inflateResetKeep OF((z_streamp));
+ZEXTERN int            ZEXPORT deflateResetKeep OF((z_streamp));
+#if (defined(_WIN32) || defined(__CYGWIN__)) && !defined(Z_SOLO)
+ZEXTERN gzFile         ZEXPORT gzopen_w OF((const wchar_t *path,
+                                            const char *mode));
+#endif
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifndef Z_SOLO
+ZEXTERN int            ZEXPORTVA gzvprintf Z_ARG((gzFile file,
+                                                  const char *format,
+                                                  va_list va));
+#  endif
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* ZLIB_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/zutil.c glmark2-2021.02/src/zlib/zutil.c
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/zutil.c	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/zutil.c	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,325 @@
+/* zutil.c -- target dependent utility functions for the compression library
+ * Copyright (C) 1995-2017 Jean-loup Gailly
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* @(#) $Id$ */
+
+#include "zutil.h"
+#ifndef Z_SOLO
+#  include "gzguts.h"
+#endif
+
+z_const char * const z_errmsg[10] = {
+    (z_const char *)"need dictionary",     /* Z_NEED_DICT       2  */
+    (z_const char *)"stream end",          /* Z_STREAM_END      1  */
+    (z_const char *)"",                    /* Z_OK              0  */
+    (z_const char *)"file error",          /* Z_ERRNO         (-1) */
+    (z_const char *)"stream error",        /* Z_STREAM_ERROR  (-2) */
+    (z_const char *)"data error",          /* Z_DATA_ERROR    (-3) */
+    (z_const char *)"insufficient memory", /* Z_MEM_ERROR     (-4) */
+    (z_const char *)"buffer error",        /* Z_BUF_ERROR     (-5) */
+    (z_const char *)"incompatible version",/* Z_VERSION_ERROR (-6) */
+    (z_const char *)""
+};
+
+
+const char * ZEXPORT zlibVersion()
+{
+    return ZLIB_VERSION;
+}
+
+uLong ZEXPORT zlibCompileFlags()
+{
+    uLong flags;
+
+    flags = 0;
+    switch ((int)(sizeof(uInt))) {
+    case 2:     break;
+    case 4:     flags += 1;     break;
+    case 8:     flags += 2;     break;
+    default:    flags += 3;
+    }
+    switch ((int)(sizeof(uLong))) {
+    case 2:     break;
+    case 4:     flags += 1 << 2;        break;
+    case 8:     flags += 2 << 2;        break;
+    default:    flags += 3 << 2;
+    }
+    switch ((int)(sizeof(voidpf))) {
+    case 2:     break;
+    case 4:     flags += 1 << 4;        break;
+    case 8:     flags += 2 << 4;        break;
+    default:    flags += 3 << 4;
+    }
+    switch ((int)(sizeof(z_off_t))) {
+    case 2:     break;
+    case 4:     flags += 1 << 6;        break;
+    case 8:     flags += 2 << 6;        break;
+    default:    flags += 3 << 6;
+    }
+#ifdef ZLIB_DEBUG
+    flags += 1 << 8;
+#endif
+#if defined(ASMV) || defined(ASMINF)
+    flags += 1 << 9;
+#endif
+#ifdef ZLIB_WINAPI
+    flags += 1 << 10;
+#endif
+#ifdef BUILDFIXED
+    flags += 1 << 12;
+#endif
+#ifdef DYNAMIC_CRC_TABLE
+    flags += 1 << 13;
+#endif
+#ifdef NO_GZCOMPRESS
+    flags += 1L << 16;
+#endif
+#ifdef NO_GZIP
+    flags += 1L << 17;
+#endif
+#ifdef PKZIP_BUG_WORKAROUND
+    flags += 1L << 20;
+#endif
+#ifdef FASTEST
+    flags += 1L << 21;
+#endif
+#if defined(STDC) || defined(Z_HAVE_STDARG_H)
+#  ifdef NO_vsnprintf
+    flags += 1L << 25;
+#    ifdef HAS_vsprintf_void
+    flags += 1L << 26;
+#    endif
+#  else
+#    ifdef HAS_vsnprintf_void
+    flags += 1L << 26;
+#    endif
+#  endif
+#else
+    flags += 1L << 24;
+#  ifdef NO_snprintf
+    flags += 1L << 25;
+#    ifdef HAS_sprintf_void
+    flags += 1L << 26;
+#    endif
+#  else
+#    ifdef HAS_snprintf_void
+    flags += 1L << 26;
+#    endif
+#  endif
+#endif
+    return flags;
+}
+
+#ifdef ZLIB_DEBUG
+#include <stdlib.h>
+#  ifndef verbose
+#    define verbose 0
+#  endif
+int ZLIB_INTERNAL z_verbose = verbose;
+
+void ZLIB_INTERNAL z_error (m)
+    char *m;
+{
+    fprintf(stderr, "%s\n", m);
+    exit(1);
+}
+#endif
+
+/* exported to allow conversion of error code to string for compress() and
+ * uncompress()
+ */
+const char * ZEXPORT zError(err)
+    int err;
+{
+    return ERR_MSG(err);
+}
+
+#if defined(_WIN32_WCE)
+    /* The Microsoft C Run-Time Library for Windows CE doesn't have
+     * errno.  We define it as a global variable to simplify porting.
+     * Its value is always 0 and should not be used.
+     */
+    int errno = 0;
+#endif
+
+#ifndef HAVE_MEMCPY
+
+void ZLIB_INTERNAL zmemcpy(dest, source, len)
+    Bytef* dest;
+    const Bytef* source;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = *source++; /* ??? to be unrolled */
+    } while (--len != 0);
+}
+
+int ZLIB_INTERNAL zmemcmp(s1, s2, len)
+    const Bytef* s1;
+    const Bytef* s2;
+    uInt  len;
+{
+    uInt j;
+
+    for (j = 0; j < len; j++) {
+        if (s1[j] != s2[j]) return 2*(s1[j] > s2[j])-1;
+    }
+    return 0;
+}
+
+void ZLIB_INTERNAL zmemzero(dest, len)
+    Bytef* dest;
+    uInt  len;
+{
+    if (len == 0) return;
+    do {
+        *dest++ = 0;  /* ??? to be unrolled */
+    } while (--len != 0);
+}
+#endif
+
+#ifndef Z_SOLO
+
+#ifdef SYS16BIT
+
+#ifdef __TURBOC__
+/* Turbo C in 16-bit mode */
+
+#  define MY_ZCALLOC
+
+/* Turbo C malloc() does not allow dynamic allocation of 64K bytes
+ * and farmalloc(64K) returns a pointer with an offset of 8, so we
+ * must fix the pointer. Warning: the pointer must be put back to its
+ * original form in order to free it, use zcfree().
+ */
+
+#define MAX_PTR 10
+/* 10*64K = 640K */
+
+local int next_ptr = 0;
+
+typedef struct ptr_table_s {
+    voidpf org_ptr;
+    voidpf new_ptr;
+} ptr_table;
+
+local ptr_table table[MAX_PTR];
+/* This table is used to remember the original form of pointers
+ * to large buffers (64K). Such pointers are normalized with a zero offset.
+ * Since MSDOS is not a preemptive multitasking OS, this table is not
+ * protected from concurrent access. This hack doesn't work anyway on
+ * a protected system like OS/2. Use Microsoft C instead.
+ */
+
+voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, unsigned items, unsigned size)
+{
+    voidpf buf;
+    ulg bsize = (ulg)items*size;
+
+    (void)opaque;
+
+    /* If we allocate less than 65520 bytes, we assume that farmalloc
+     * will return a usable pointer which doesn't have to be normalized.
+     */
+    if (bsize < 65520L) {
+        buf = farmalloc(bsize);
+        if (*(ush*)&buf != 0) return buf;
+    } else {
+        buf = farmalloc(bsize + 16L);
+    }
+    if (buf == NULL || next_ptr >= MAX_PTR) return NULL;
+    table[next_ptr].org_ptr = buf;
+
+    /* Normalize the pointer to seg:0 */
+    *((ush*)&buf+1) += ((ush)((uch*)buf-0) + 15) >> 4;
+    *(ush*)&buf = 0;
+    table[next_ptr++].new_ptr = buf;
+    return buf;
+}
+
+void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr)
+{
+    int n;
+
+    (void)opaque;
+
+    if (*(ush*)&ptr != 0) { /* object < 64K */
+        farfree(ptr);
+        return;
+    }
+    /* Find the original pointer */
+    for (n = 0; n < next_ptr; n++) {
+        if (ptr != table[n].new_ptr) continue;
+
+        farfree(table[n].org_ptr);
+        while (++n < next_ptr) {
+            table[n-1] = table[n];
+        }
+        next_ptr--;
+        return;
+    }
+    Assert(0, "zcfree: ptr not found");
+}
+
+#endif /* __TURBOC__ */
+
+
+#ifdef M_I86
+/* Microsoft C in 16-bit mode */
+
+#  define MY_ZCALLOC
+
+#if (!defined(_MSC_VER) || (_MSC_VER <= 600))
+#  define _halloc  halloc
+#  define _hfree   hfree
+#endif
+
+voidpf ZLIB_INTERNAL zcalloc (voidpf opaque, uInt items, uInt size)
+{
+    (void)opaque;
+    return _halloc((long)items, size);
+}
+
+void ZLIB_INTERNAL zcfree (voidpf opaque, voidpf ptr)
+{
+    (void)opaque;
+    _hfree(ptr);
+}
+
+#endif /* M_I86 */
+
+#endif /* SYS16BIT */
+
+
+#ifndef MY_ZCALLOC /* Any system without a special alloc function */
+
+#ifndef STDC
+extern voidp  malloc OF((uInt size));
+extern voidp  calloc OF((uInt items, uInt size));
+extern void   free   OF((voidpf ptr));
+#endif
+
+voidpf ZLIB_INTERNAL zcalloc (opaque, items, size)
+    voidpf opaque;
+    unsigned items;
+    unsigned size;
+{
+    (void)opaque;
+    return sizeof(uInt) > 2 ? (voidpf)malloc(items * size) :
+                              (voidpf)calloc(items, size);
+}
+
+void ZLIB_INTERNAL zcfree (opaque, ptr)
+    voidpf opaque;
+    voidpf ptr;
+{
+    (void)opaque;
+    free(ptr);
+}
+
+#endif /* MY_ZCALLOC */
+
+#endif /* !Z_SOLO */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/src/zlib/zutil.h glmark2-2021.02/src/zlib/zutil.h
--- glmark2-2014.03+git20150611.fa71af2d/src/zlib/zutil.h	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/src/zlib/zutil.h	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,271 @@
+/* zutil.h -- internal interface and configuration of the compression library
+ * Copyright (C) 1995-2016 Jean-loup Gailly, Mark Adler
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* @(#) $Id$ */
+
+#ifndef ZUTIL_H
+#define ZUTIL_H
+
+#ifdef HAVE_HIDDEN
+#  define ZLIB_INTERNAL __attribute__((visibility ("hidden")))
+#else
+#  define ZLIB_INTERNAL
+#endif
+
+#include "zlib.h"
+
+#if defined(STDC) && !defined(Z_SOLO)
+#  if !(defined(_WIN32_WCE) && defined(_MSC_VER))
+#    include <stddef.h>
+#  endif
+#  include <string.h>
+#  include <stdlib.h>
+#endif
+
+#ifdef Z_SOLO
+   typedef long ptrdiff_t;  /* guess -- will be caught if guess is wrong */
+#endif
+
+#ifndef local
+#  define local static
+#endif
+/* since "static" is used to mean two completely different things in C, we
+   define "local" for the non-static meaning of "static", for readability
+   (compile with -Dlocal if your debugger can't find static symbols) */
+
+typedef unsigned char  uch;
+typedef uch FAR uchf;
+typedef unsigned short ush;
+typedef ush FAR ushf;
+typedef unsigned long  ulg;
+
+extern z_const char * const z_errmsg[10]; /* indexed by 2-zlib_error */
+/* (size given to avoid silly warnings with Visual C++) */
+
+#define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)]
+
+#define ERR_RETURN(strm,err) \
+  return (strm->msg = ERR_MSG(err), (err))
+/* To be used only when the state is known to be valid */
+
+        /* common constants */
+
+#ifndef DEF_WBITS
+#  define DEF_WBITS MAX_WBITS
+#endif
+/* default windowBits for decompression. MAX_WBITS is for compression only */
+
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+/* default memLevel */
+
+#define STORED_BLOCK 0
+#define STATIC_TREES 1
+#define DYN_TREES    2
+/* The three kinds of block type */
+
+#define MIN_MATCH  3
+#define MAX_MATCH  258
+/* The minimum and maximum match lengths */
+
+#define PRESET_DICT 0x20 /* preset dictionary flag in zlib header */
+
+        /* target dependencies */
+
+#if defined(MSDOS) || (defined(WINDOWS) && !defined(WIN32))
+#  define OS_CODE  0x00
+#  ifndef Z_SOLO
+#    if defined(__TURBOC__) || defined(__BORLANDC__)
+#      if (__STDC__ == 1) && (defined(__LARGE__) || defined(__COMPACT__))
+         /* Allow compilation with ANSI keywords only enabled */
+         void _Cdecl farfree( void *block );
+         void *_Cdecl farmalloc( unsigned long nbytes );
+#      else
+#        include <alloc.h>
+#      endif
+#    else /* MSC or DJGPP */
+#      include <malloc.h>
+#    endif
+#  endif
+#endif
+
+#ifdef AMIGA
+#  define OS_CODE  1
+#endif
+
+#if defined(VAXC) || defined(VMS)
+#  define OS_CODE  2
+#  define F_OPEN(name, mode) \
+     fopen((name), (mode), "mbc=60", "ctx=stm", "rfm=fix", "mrs=512")
+#endif
+
+#ifdef __370__
+#  if __TARGET_LIB__ < 0x20000000
+#    define OS_CODE 4
+#  elif __TARGET_LIB__ < 0x40000000
+#    define OS_CODE 11
+#  else
+#    define OS_CODE 8
+#  endif
+#endif
+
+#if defined(ATARI) || defined(atarist)
+#  define OS_CODE  5
+#endif
+
+#ifdef OS2
+#  define OS_CODE  6
+#  if defined(M_I86) && !defined(Z_SOLO)
+#    include <malloc.h>
+#  endif
+#endif
+
+#if defined(MACOS) || defined(TARGET_OS_MAC)
+#  define OS_CODE  7
+#  ifndef Z_SOLO
+#    if defined(__MWERKS__) && __dest_os != __be_os && __dest_os != __win32_os
+#      include <unix.h> /* for fdopen */
+#    else
+#      ifndef fdopen
+#        define fdopen(fd,mode) NULL /* No fdopen() */
+#      endif
+#    endif
+#  endif
+#endif
+
+#ifdef __acorn
+#  define OS_CODE 13
+#endif
+
+#if defined(WIN32) && !defined(__CYGWIN__)
+#  define OS_CODE  10
+#endif
+
+#ifdef _BEOS_
+#  define OS_CODE  16
+#endif
+
+#ifdef __TOS_OS400__
+#  define OS_CODE 18
+#endif
+
+#ifdef __APPLE__
+#  define OS_CODE 19
+#endif
+
+#if defined(_BEOS_) || defined(RISCOS)
+#  define fdopen(fd,mode) NULL /* No fdopen() */
+#endif
+
+#if (defined(_MSC_VER) && (_MSC_VER > 600)) && !defined __INTERIX
+#  if defined(_WIN32_WCE)
+#    define fdopen(fd,mode) NULL /* No fdopen() */
+#    ifndef _PTRDIFF_T_DEFINED
+       typedef int ptrdiff_t;
+#      define _PTRDIFF_T_DEFINED
+#    endif
+#  else
+#    define fdopen(fd,type)  _fdopen(fd,type)
+#  endif
+#endif
+
+#if defined(__BORLANDC__) && !defined(MSDOS)
+  #pragma warn -8004
+  #pragma warn -8008
+  #pragma warn -8066
+#endif
+
+/* provide prototypes for these when building zlib without LFS */
+#if !defined(_WIN32) && \
+    (!defined(_LARGEFILE64_SOURCE) || _LFS64_LARGEFILE-0 == 0)
+    ZEXTERN uLong ZEXPORT adler32_combine64 OF((uLong, uLong, z_off_t));
+    ZEXTERN uLong ZEXPORT crc32_combine64 OF((uLong, uLong, z_off_t));
+#endif
+
+        /* common defaults */
+
+#ifndef OS_CODE
+#  define OS_CODE  3     /* assume Unix */
+#endif
+
+#ifndef F_OPEN
+#  define F_OPEN(name, mode) fopen((name), (mode))
+#endif
+
+         /* functions */
+
+#if defined(pyr) || defined(Z_SOLO)
+#  define NO_MEMCPY
+#endif
+#if defined(SMALL_MEDIUM) && !defined(_MSC_VER) && !defined(__SC__)
+ /* Use our own functions for small and medium model with MSC <= 5.0.
+  * You may have to use the same strategy for Borland C (untested).
+  * The __SC__ check is for Symantec.
+  */
+#  define NO_MEMCPY
+#endif
+#if defined(STDC) && !defined(HAVE_MEMCPY) && !defined(NO_MEMCPY)
+#  define HAVE_MEMCPY
+#endif
+#ifdef HAVE_MEMCPY
+#  ifdef SMALL_MEDIUM /* MSDOS small or medium model */
+#    define zmemcpy _fmemcpy
+#    define zmemcmp _fmemcmp
+#    define zmemzero(dest, len) _fmemset(dest, 0, len)
+#  else
+#    define zmemcpy memcpy
+#    define zmemcmp memcmp
+#    define zmemzero(dest, len) memset(dest, 0, len)
+#  endif
+#else
+   void ZLIB_INTERNAL zmemcpy OF((Bytef* dest, const Bytef* source, uInt len));
+   int ZLIB_INTERNAL zmemcmp OF((const Bytef* s1, const Bytef* s2, uInt len));
+   void ZLIB_INTERNAL zmemzero OF((Bytef* dest, uInt len));
+#endif
+
+/* Diagnostic functions */
+#ifdef ZLIB_DEBUG
+#  include <stdio.h>
+   extern int ZLIB_INTERNAL z_verbose;
+   extern void ZLIB_INTERNAL z_error OF((char *m));
+#  define Assert(cond,msg) {if(!(cond)) z_error(msg);}
+#  define Trace(x) {if (z_verbose>=0) fprintf x ;}
+#  define Tracev(x) {if (z_verbose>0) fprintf x ;}
+#  define Tracevv(x) {if (z_verbose>1) fprintf x ;}
+#  define Tracec(c,x) {if (z_verbose>0 && (c)) fprintf x ;}
+#  define Tracecv(c,x) {if (z_verbose>1 && (c)) fprintf x ;}
+#else
+#  define Assert(cond,msg)
+#  define Trace(x)
+#  define Tracev(x)
+#  define Tracevv(x)
+#  define Tracec(c,x)
+#  define Tracecv(c,x)
+#endif
+
+#ifndef Z_SOLO
+   voidpf ZLIB_INTERNAL zcalloc OF((voidpf opaque, unsigned items,
+                                    unsigned size));
+   void ZLIB_INTERNAL zcfree  OF((voidpf opaque, voidpf ptr));
+#endif
+
+#define ZALLOC(strm, items, size) \
+           (*((strm)->zalloc))((strm)->opaque, (items), (size))
+#define ZFREE(strm, addr)  (*((strm)->zfree))((strm)->opaque, (voidpf)(addr))
+#define TRY_FREE(s, p) {if (p) ZFREE(s, p);}
+
+/* Reverse the bytes in a 32-bit value */
+#define ZSWAP32(q) ((((q) >> 24) & 0xff) + (((q) >> 8) & 0xff00) + \
+                    (((q) & 0xff00) << 8) + (((q) & 0xff) << 24))
+
+#endif /* ZUTIL_H */
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waf glmark2-2021.02/waf
--- glmark2-2014.03+git20150611.fa71af2d/waf	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waf	2021-04-25 01:47:22.000000000 +0800
@@ -1,7 +1,6 @@
-#!/usr/bin/env python
-# encoding: ISO8859-1
-# Thomas Nagy, 2005-2011
-
+#!/usr/bin/env python3
+# Thomas Nagy, 2005-2016
+#
 """
 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions
@@ -30,13 +29,15 @@
 POSSIBILITY OF SUCH DAMAGE.
 """
 
-import os, sys
+import os, sys, inspect
 
-VERSION="1.6.11"
-REVISION="30618c54883417962c38f5d395f83584"
+VERSION="1.9.3"
+REVISION="826456a3e8d1d27787c240c6ba69d280"
+GIT="b022313c197614245fae732c0dd1d8167b9ee5dd"
 INSTALL=''
-C1='#*'
+C1='#)'
 C2='#%'
+C3='#$'
 cwd = os.getcwd()
 join = os.path.join
 
@@ -53,8 +54,8 @@
 	print(('\033[91mError: %s\033[0m' % m))
 	sys.exit(1)
 
-def unpack_wafdir(dir):
-	f = open(sys.argv[0],'rb')
+def unpack_wafdir(dir, src):
+	f = open(src,'rb')
 	c = 'corrupt archive (%d)'
 	while 1:
 		line = f.readline()
@@ -65,22 +66,22 @@
 			if f.readline() != b('#<==\n'): err(c % 2)
 			break
 	if not txt: err(c % 3)
-	txt = txt[1:-1].replace(b(C1), b('\n')).replace(b(C2), b('\r'))
+	txt = txt[1:-1].replace(b(C1), b('\n')).replace(b(C2), b('\r')).replace(b(C3), b('\x00'))
 
 	import shutil, tarfile
 	try: shutil.rmtree(dir)
 	except OSError: pass
 	try:
-		for x in ['Tools', 'extras']:
+		for x in ('Tools', 'extras'):
 			os.makedirs(join(dir, 'waflib', x))
 	except OSError:
-		err("Cannot unpack waf lib into %s\nMove waf into a writeable directory" % dir)
+		err("Cannot unpack waf lib into %s\nMove waf in a writable directory" % dir)
 
 	os.chdir(dir)
 	tmp = 't.bz2'
 	t = open(tmp,'wb')
-	t.write(txt)
-	t.close()
+	try: t.write(txt)
+	finally: t.close()
 
 	try:
 		t = tarfile.open(tmp)
@@ -95,10 +96,12 @@
 			except OSError: pass
 			err("Waf cannot be unpacked, check that bzip2 support is present")
 
-	for x in t: t.extract(x)
-	t.close()
+	try:
+		for x in t: t.extract(x)
+	finally:
+		t.close()
 
-	for x in ['Tools', 'extras']:
+	for x in ('Tools', 'extras'):
 		os.chmod(join('waflib',x), 493)
 
 	if sys.hexversion<0x300000f:
@@ -106,7 +109,7 @@
 		import fixpy2
 		fixpy2.fixdir(dir)
 
-	os.unlink(tmp)
+	os.remove(tmp)
 	os.chdir(cwd)
 
 	try: dir = unicode(dir, 'mbcs')
@@ -125,8 +128,8 @@
 		pass
 
 def find_lib():
-	name = sys.argv[0]
-	base = os.path.dirname(os.path.abspath(name))
+	src = os.path.abspath(inspect.getfile(inspect.getmodule(err)))
+	base, name = os.path.split(src)
 
 	#devs use $WAFDIR
 	w=test(os.environ.get('WAFDIR', ''))
@@ -139,7 +142,7 @@
 		err('waf-light requires waflib -> export WAFDIR=/folder')
 
 	dirname = '%s-%s-%s' % (WAF, VERSION, REVISION)
-	for i in [INSTALL,'/usr','/usr/local','/opt']:
+	for i in (INSTALL,'/usr','/usr/local','/opt'):
 		w = test(i + '/lib/' + dirname)
 		if w: return w
 
@@ -149,14 +152,14 @@
 	if w: return w
 
 	#unpack
-	unpack_wafdir(dir)
+	unpack_wafdir(dir, src)
 	return dir
 
 wafdir = find_lib()
 sys.path.insert(0, wafdir)
 
 if __name__ == '__main__':
-	import waflib.extras.compat15
+
 	from waflib import Scripting
 	Scripting.waf_entry_point(cwd, VERSION, wafdir)
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/ansiterm.py glmark2-2021.02/waflib/ansiterm.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/ansiterm.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/ansiterm.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,52 +1,84 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys,os
+import os,re,sys
+from waflib import Utils
+wlock=Utils.threading.Lock()
 try:
-	if not(sys.stderr.isatty()and sys.stdout.isatty()):
-		raise ValueError('not a tty')
-	from ctypes import*
+	from ctypes import Structure,windll,c_short,c_ushort,c_ulong,c_int,byref,c_wchar,POINTER,c_long
+except ImportError:
+	class AnsiTerm(object):
+		def __init__(self,stream):
+			self.stream=stream
+			try:
+				self.errors=self.stream.errors
+			except AttributeError:
+				pass
+			self.encoding=self.stream.encoding
+		def write(self,txt):
+			try:
+				wlock.acquire()
+				self.stream.write(txt)
+				self.stream.flush()
+			finally:
+				wlock.release()
+		def fileno(self):
+			return self.stream.fileno()
+		def flush(self):
+			self.stream.flush()
+		def isatty(self):
+			return self.stream.isatty()
+else:
 	class COORD(Structure):
 		_fields_=[("X",c_short),("Y",c_short)]
 	class SMALL_RECT(Structure):
 		_fields_=[("Left",c_short),("Top",c_short),("Right",c_short),("Bottom",c_short)]
 	class CONSOLE_SCREEN_BUFFER_INFO(Structure):
-		_fields_=[("Size",COORD),("CursorPosition",COORD),("Attributes",c_short),("Window",SMALL_RECT),("MaximumWindowSize",COORD)]
+		_fields_=[("Size",COORD),("CursorPosition",COORD),("Attributes",c_ushort),("Window",SMALL_RECT),("MaximumWindowSize",COORD)]
 	class CONSOLE_CURSOR_INFO(Structure):
 		_fields_=[('dwSize',c_ulong),('bVisible',c_int)]
-	sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
-	csinfo=CONSOLE_CURSOR_INFO()
-	hconsole=windll.kernel32.GetStdHandle(-11)
-	windll.kernel32.GetConsoleScreenBufferInfo(hconsole,byref(sbinfo))
-	if sbinfo.Size.X<9 or sbinfo.Size.Y<9:raise ValueError('small console')
-	windll.kernel32.GetConsoleCursorInfo(hconsole,byref(csinfo))
-except Exception:
-	pass
-else:
-	import re,threading
-	is_vista=getattr(sys,"getwindowsversion",None)and sys.getwindowsversion()[0]>=6
 	try:
 		_type=unicode
-	except:
+	except NameError:
 		_type=str
 	to_int=lambda number,default:number and int(number)or default
-	wlock=threading.Lock()
 	STD_OUTPUT_HANDLE=-11
 	STD_ERROR_HANDLE=-12
+	windll.kernel32.GetStdHandle.argtypes=[c_ulong]
+	windll.kernel32.GetStdHandle.restype=c_ulong
+	windll.kernel32.GetConsoleScreenBufferInfo.argtypes=[c_ulong,POINTER(CONSOLE_SCREEN_BUFFER_INFO)]
+	windll.kernel32.GetConsoleScreenBufferInfo.restype=c_long
+	windll.kernel32.SetConsoleTextAttribute.argtypes=[c_ulong,c_ushort]
+	windll.kernel32.SetConsoleTextAttribute.restype=c_long
+	windll.kernel32.FillConsoleOutputCharacterW.argtypes=[c_ulong,c_wchar,c_ulong,POINTER(COORD),POINTER(c_ulong)]
+	windll.kernel32.FillConsoleOutputCharacterW.restype=c_long
+	windll.kernel32.FillConsoleOutputAttribute.argtypes=[c_ulong,c_ushort,c_ulong,POINTER(COORD),POINTER(c_ulong)]
+	windll.kernel32.FillConsoleOutputAttribute.restype=c_long
+	windll.kernel32.SetConsoleCursorPosition.argtypes=[c_ulong,POINTER(COORD)]
+	windll.kernel32.SetConsoleCursorPosition.restype=c_long
+	windll.kernel32.SetConsoleCursorInfo.argtypes=[c_ulong,POINTER(CONSOLE_CURSOR_INFO)]
+	windll.kernel32.SetConsoleCursorInfo.restype=c_long
 	class AnsiTerm(object):
-		def __init__(self):
-			self.encoding=sys.stdout.encoding
-			self.hconsole=windll.kernel32.GetStdHandle(STD_OUTPUT_HANDLE)
+		def __init__(self,s):
+			self.stream=s
+			try:
+				self.errors=s.errors
+			except AttributeError:
+				pass
+			self.encoding=s.encoding
 			self.cursor_history=[]
-			self.orig_sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
-			self.orig_csinfo=CONSOLE_CURSOR_INFO()
-			windll.kernel32.GetConsoleScreenBufferInfo(self.hconsole,byref(self.orig_sbinfo))
-			windll.kernel32.GetConsoleCursorInfo(hconsole,byref(self.orig_csinfo))
+			handle=(s.fileno()==2)and STD_ERROR_HANDLE or STD_OUTPUT_HANDLE
+			self.hconsole=windll.kernel32.GetStdHandle(handle)
+			self._sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
+			self._csinfo=CONSOLE_CURSOR_INFO()
+			windll.kernel32.GetConsoleCursorInfo(self.hconsole,byref(self._csinfo))
+			self._orig_sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
+			r=windll.kernel32.GetConsoleScreenBufferInfo(self.hconsole,byref(self._orig_sbinfo))
+			self._isatty=r==1
 		def screen_buffer_info(self):
-			sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
-			windll.kernel32.GetConsoleScreenBufferInfo(self.hconsole,byref(sbinfo))
-			return sbinfo
+			windll.kernel32.GetConsoleScreenBufferInfo(self.hconsole,byref(self._sbinfo))
+			return self._sbinfo
 		def clear_line(self,param):
 			mode=param and int(param)or 0
 			sbinfo=self.screen_buffer_info()
@@ -59,8 +91,8 @@
 			else:
 				line_start=sbinfo.CursorPosition
 				line_length=sbinfo.Size.X-sbinfo.CursorPosition.X
-			chars_written=c_int()
-			windll.kernel32.FillConsoleOutputCharacterA(self.hconsole,c_wchar(' '),line_length,line_start,byref(chars_written))
+			chars_written=c_ulong()
+			windll.kernel32.FillConsoleOutputCharacterW(self.hconsole,c_wchar(' '),line_length,line_start,byref(chars_written))
 			windll.kernel32.FillConsoleOutputAttribute(self.hconsole,sbinfo.Attributes,line_length,line_start,byref(chars_written))
 		def clear_screen(self,param):
 			mode=to_int(param,0)
@@ -75,8 +107,8 @@
 			else:
 				clear_start=sbinfo.CursorPosition
 				clear_length=((sbinfo.Size.X-sbinfo.CursorPosition.X)+sbinfo.Size.X*(sbinfo.Size.Y-sbinfo.CursorPosition.Y))
-			chars_written=c_int()
-			windll.kernel32.FillConsoleOutputCharacterA(self.hconsole,c_wchar(' '),clear_length,clear_start,byref(chars_written))
+			chars_written=c_ulong()
+			windll.kernel32.FillConsoleOutputCharacterW(self.hconsole,c_wchar(' '),clear_length,clear_start,byref(chars_written))
 			windll.kernel32.FillConsoleOutputAttribute(self.hconsole,sbinfo.Attributes,clear_length,clear_start,byref(chars_written))
 		def push_cursor(self,param):
 			sbinfo=self.screen_buffer_info()
@@ -119,20 +151,16 @@
 			return((c&1)<<2)|(c&2)|((c&4)>>2)
 		def set_color(self,param):
 			cols=param.split(';')
-			sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
-			windll.kernel32.GetConsoleScreenBufferInfo(self.hconsole,byref(sbinfo))
+			sbinfo=self.screen_buffer_info()
 			attr=sbinfo.Attributes
 			for c in cols:
-				if is_vista:
-					c=int(c)
-				else:
-					c=to_int(c,0)
-				if c in range(30,38):
+				c=to_int(c,0)
+				if 29<c<38:
 					attr=(attr&0xfff0)|self.rgb2bgr(c-30)
-				elif c in range(40,48):
+				elif 39<c<48:
 					attr=(attr&0xff0f)|(self.rgb2bgr(c-40)<<4)
 				elif c==0:
-					attr=self.orig_sbinfo.Attributes
+					attr=self._orig_sbinfo.Attributes
 				elif c==1:
 					attr|=0x08
 				elif c==4:
@@ -141,37 +169,70 @@
 					attr=(attr&0xff88)|((attr&0x70)>>4)|((attr&0x07)<<4)
 			windll.kernel32.SetConsoleTextAttribute(self.hconsole,attr)
 		def show_cursor(self,param):
-			csinfo.bVisible=1
-			windll.kernel32.SetConsoleCursorInfo(self.hconsole,byref(csinfo))
+			self._csinfo.bVisible=1
+			windll.kernel32.SetConsoleCursorInfo(self.hconsole,byref(self._csinfo))
 		def hide_cursor(self,param):
-			csinfo.bVisible=0
-			windll.kernel32.SetConsoleCursorInfo(self.hconsole,byref(csinfo))
+			self._csinfo.bVisible=0
+			windll.kernel32.SetConsoleCursorInfo(self.hconsole,byref(self._csinfo))
 		ansi_command_table={'A':move_up,'B':move_down,'C':move_right,'D':move_left,'E':next_line,'F':prev_line,'G':set_column,'H':set_cursor,'f':set_cursor,'J':clear_screen,'K':clear_line,'h':show_cursor,'l':hide_cursor,'m':set_color,'s':push_cursor,'u':pop_cursor,}
 		ansi_tokens=re.compile('(?:\x1b\[([0-9?;]*)([a-zA-Z])|([^\x1b]+))')
 		def write(self,text):
 			try:
 				wlock.acquire()
-				for param,cmd,txt in self.ansi_tokens.findall(text):
-					if cmd:
-						cmd_func=self.ansi_command_table.get(cmd)
-						if cmd_func:
-							cmd_func(self,param)
-					else:
-						self.writeconsole(txt)
+				if self._isatty:
+					for param,cmd,txt in self.ansi_tokens.findall(text):
+						if cmd:
+							cmd_func=self.ansi_command_table.get(cmd)
+							if cmd_func:
+								cmd_func(self,param)
+						else:
+							self.writeconsole(txt)
+				else:
+					self.stream.write(text)
 			finally:
 				wlock.release()
 		def writeconsole(self,txt):
-			chars_written=c_int()
+			chars_written=c_ulong()
 			writeconsole=windll.kernel32.WriteConsoleA
 			if isinstance(txt,_type):
 				writeconsole=windll.kernel32.WriteConsoleW
-			TINY_STEP=3000
-			for x in range(0,len(txt),TINY_STEP):
-				tiny=txt[x:x+TINY_STEP]
-				writeconsole(self.hconsole,tiny,len(tiny),byref(chars_written),None)
+			done=0
+			todo=len(txt)
+			chunk=32<<10
+			while todo!=0:
+				doing=min(chunk,todo)
+				buf=txt[done:done+doing]
+				r=writeconsole(self.hconsole,buf,doing,byref(chars_written),None)
+				if r==0:
+					chunk>>=1
+					continue
+				done+=doing
+				todo-=doing
+		def fileno(self):
+			return self.stream.fileno()
 		def flush(self):
 			pass
 		def isatty(self):
-			return True
-	sys.stderr=sys.stdout=AnsiTerm()
-	os.environ['TERM']='vt100'
+			return self._isatty
+	if sys.stdout.isatty()or sys.stderr.isatty():
+		handle=sys.stdout.isatty()and STD_OUTPUT_HANDLE or STD_ERROR_HANDLE
+		console=windll.kernel32.GetStdHandle(handle)
+		sbinfo=CONSOLE_SCREEN_BUFFER_INFO()
+		def get_term_cols():
+			windll.kernel32.GetConsoleScreenBufferInfo(console,byref(sbinfo))
+			return sbinfo.Size.X-1
+try:
+	import struct,fcntl,termios
+except ImportError:
+	pass
+else:
+	if(sys.stdout.isatty()or sys.stderr.isatty())and os.environ.get('TERM','')not in('dumb','emacs'):
+		FD=sys.stdout.isatty()and sys.stdout.fileno()or sys.stderr.fileno()
+		def fun():
+			return struct.unpack("HHHH",fcntl.ioctl(FD,termios.TIOCGWINSZ,struct.pack("HHHH",0,0,0,0)))[1]
+		try:
+			fun()
+		except Exception as e:
+			pass
+		else:
+			get_term_cols=fun
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Build.py glmark2-2021.02/waflib/Build.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Build.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Build.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,21 +1,24 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,errno,re,shutil
-try:import cPickle
-except:import pickle as cPickle
-from waflib import Runner,TaskGen,Utils,ConfigSet,Task,Logs,Options,Context,Errors
-import waflib.Node
+import os,sys,errno,re,shutil,stat
+try:
+	import cPickle
+except ImportError:
+	import pickle as cPickle
+from waflib import Node,Runner,TaskGen,Utils,ConfigSet,Task,Logs,Options,Context,Errors
 CACHE_DIR='c4che'
 CACHE_SUFFIX='_cache.py'
 INSTALL=1337
 UNINSTALL=-1337
-SAVED_ATTRS='root node_deps raw_deps task_sigs'.split()
+SAVED_ATTRS='root node_sigs task_sigs imp_sigs raw_deps node_deps'.split()
 CFG_FILES='cfg_files'
 POST_AT_ONCE=0
 POST_LAZY=1
-POST_BOTH=2
+PROTOCOL=-1
+if sys.platform=='cli':
+	PROTOCOL=0
 class BuildContext(Context.Context):
 	'''executes the build'''
 	cmd='build'
@@ -24,29 +27,31 @@
 		super(BuildContext,self).__init__(**kw)
 		self.is_install=0
 		self.top_dir=kw.get('top_dir',Context.top_dir)
-		self.run_dir=kw.get('run_dir',Context.run_dir)
-		self.post_mode=POST_AT_ONCE
 		self.out_dir=kw.get('out_dir',Context.out_dir)
-		self.cache_dir=kw.get('cache_dir',None)
+		self.run_dir=kw.get('run_dir',Context.run_dir)
+		self.launch_dir=Context.launch_dir
+		self.post_mode=POST_LAZY
+		self.cache_dir=kw.get('cache_dir')
 		if not self.cache_dir:
-			self.cache_dir=self.out_dir+os.sep+CACHE_DIR
+			self.cache_dir=os.path.join(self.out_dir,CACHE_DIR)
 		self.all_envs={}
+		self.node_sigs={}
 		self.task_sigs={}
+		self.imp_sigs={}
 		self.node_deps={}
 		self.raw_deps={}
-		self.cache_dir_contents={}
 		self.task_gen_cache_names={}
-		self.launch_dir=Context.launch_dir
 		self.jobs=Options.options.jobs
 		self.targets=Options.options.targets
 		self.keep=Options.options.keep
-		self.cache_global=Options.cache_global
-		self.nocache=Options.options.nocache
 		self.progress_bar=Options.options.progress_bar
 		self.deps_man=Utils.defaultdict(list)
 		self.current_group=0
 		self.groups=[]
 		self.group_names={}
+		for v in SAVED_ATTRS:
+			if not hasattr(self,v):
+				setattr(self,v,{})
 	def get_variant_dir(self):
 		if not self.variant:
 			return self.out_dir
@@ -56,16 +61,16 @@
 		kw['bld']=self
 		ret=TaskGen.task_gen(*k,**kw)
 		self.task_gen_cache_names={}
-		self.add_to_group(ret,group=kw.get('group',None))
+		self.add_to_group(ret,group=kw.get('group'))
 		return ret
+	def rule(self,*k,**kw):
+		def f(rule):
+			ret=self(*k,**kw)
+			ret.rule=rule
+			return ret
+		return f
 	def __copy__(self):
-		raise Errors.WafError('build contexts are not supposed to be copied')
-	def install_files(self,*k,**kw):
-		pass
-	def install_as(self,*k,**kw):
-		pass
-	def symlink_as(self,*k,**kw):
-		pass
+		raise Errors.WafError('build contexts cannot be copied')
 	def load_envs(self):
 		node=self.root.find_node(self.cache_dir)
 		if not node:
@@ -79,12 +84,8 @@
 			self.all_envs[name]=env
 			for f in env[CFG_FILES]:
 				newnode=self.root.find_resource(f)
-				try:
-					h=Utils.h_file(newnode.abspath())
-				except(IOError,AttributeError):
-					Logs.error('cannot find %r'%f)
-					h=Utils.SIG_NIL
-				newnode.sig=h
+				if not newnode or not newnode.exists():
+					raise Errors.WafError('Missing configuration file %r, reconfigure the project!'%f)
 	def init_dirs(self):
 		if not(os.path.isabs(self.top_dir)and os.path.isabs(self.out_dir)):
 			raise Errors.WafError('The project was not configured: run "waf configure" first!')
@@ -97,56 +98,52 @@
 			self.load_envs()
 		self.execute_build()
 	def execute_build(self):
-		Logs.info("Waf: Entering directory `%s'"%self.variant_dir)
+		Logs.info("Waf: Entering directory `%s'",self.variant_dir)
 		self.recurse([self.run_dir])
 		self.pre_build()
 		self.timer=Utils.Timer()
-		if self.progress_bar:
-			sys.stderr.write(Logs.colors.cursor_off)
 		try:
 			self.compile()
 		finally:
-			if self.progress_bar==1:
-				c=len(self.returned_tasks)or 1
-				self.to_log(self.progress_line(c,c,Logs.colors.BLUE,Logs.colors.NORMAL))
-				print('')
-				sys.stdout.flush()
-				sys.stderr.write(Logs.colors.cursor_on)
-			Logs.info("Waf: Leaving directory `%s'"%self.variant_dir)
+			if self.progress_bar==1 and sys.stderr.isatty():
+				c=self.producer.processed or 1
+				m=self.progress_line(c,c,Logs.colors.BLUE,Logs.colors.NORMAL)
+				Logs.info(m,extra={'stream':sys.stderr,'c1':Logs.colors.cursor_off,'c2':Logs.colors.cursor_on})
+			Logs.info("Waf: Leaving directory `%s'",self.variant_dir)
+		try:
+			self.producer.bld=None
+			del self.producer
+		except AttributeError:
+			pass
 		self.post_build()
 	def restore(self):
 		try:
 			env=ConfigSet.ConfigSet(os.path.join(self.cache_dir,'build.config.py'))
-		except(IOError,OSError):
+		except EnvironmentError:
 			pass
 		else:
-			if env['version']<Context.HEXVERSION:
+			if env.version<Context.HEXVERSION:
 				raise Errors.WafError('Version mismatch! reconfigure the project')
-			for t in env['tools']:
+			for t in env.tools:
 				self.setup(**t)
-		f=None
+		dbfn=os.path.join(self.variant_dir,Context.DBFILE)
 		try:
-			dbfn=os.path.join(self.variant_dir,Context.DBFILE)
+			data=Utils.readf(dbfn,'rb')
+		except(EnvironmentError,EOFError):
+			Logs.debug('build: Could not load the build cache %s (missing)',dbfn)
+		else:
 			try:
-				f=open(dbfn,'rb')
-			except(IOError,EOFError):
-				Logs.debug('build: could not load the build cache %s (missing)'%dbfn)
-			else:
+				Node.pickle_lock.acquire()
+				Node.Nod3=self.node_class
 				try:
-					waflib.Node.pickle_lock.acquire()
-					waflib.Node.Nod3=self.node_class
-					try:
-						data=cPickle.load(f)
-					except Exception ,e:
-						Logs.debug('build: could not pickle the build cache %s: %r'%(dbfn,e))
-					else:
-						for x in SAVED_ATTRS:
-							setattr(self,x,data[x])
-				finally:
-					waflib.Node.pickle_lock.release()
-		finally:
-			if f:
-				f.close()
+					data=cPickle.loads(data)
+				except Exception as e:
+					Logs.debug('build: Could not pickle the build cache %s: %r',dbfn,e)
+				else:
+					for x in SAVED_ATTRS:
+						setattr(self,x,data[x])
+			finally:
+				Node.pickle_lock.release()
 		self.init_dirs()
 	def store(self):
 		data={}
@@ -154,20 +151,15 @@
 			data[x]=getattr(self,x)
 		db=os.path.join(self.variant_dir,Context.DBFILE)
 		try:
-			waflib.Node.pickle_lock.acquire()
-			waflib.Node.Nod3=self.node_class
-			f=None
-			try:
-				f=open(db+'.tmp','wb')
-				cPickle.dump(data,f)
-			finally:
-				if f:
-					f.close()
+			Node.pickle_lock.acquire()
+			Node.Nod3=self.node_class
+			x=cPickle.dumps(data,PROTOCOL)
 		finally:
-			waflib.Node.pickle_lock.release()
+			Node.pickle_lock.release()
+		Utils.writef(db+'.tmp',x,m='wb')
 		try:
 			st=os.stat(db)
-			os.unlink(db)
+			os.remove(db)
 			if not Utils.is_win32:
 				os.chown(db+'.tmp',st.st_uid,st.st_gid)
 		except(AttributeError,OSError):
@@ -177,7 +169,6 @@
 		Logs.debug('build: compile()')
 		self.producer=Runner.Parallel(self,self.jobs)
 		self.producer.biter=self.get_build_iterator()
-		self.returned_tasks=[]
 		try:
 			self.producer.start()
 		except KeyboardInterrupt:
@@ -203,13 +194,20 @@
 		self.all_envs[self.variant]=val
 	env=property(get_env,set_env)
 	def add_manual_dependency(self,path,value):
-		if isinstance(path,waflib.Node.Node):
+		if not path:
+			raise ValueError('Invalid input path %r'%path)
+		if isinstance(path,Node.Node):
 			node=path
 		elif os.path.isabs(path):
 			node=self.root.find_resource(path)
 		else:
 			node=self.path.find_resource(path)
-		self.deps_man[id(node)].append(value)
+		if not node:
+			raise ValueError('Could not find the path %r'%path)
+		if isinstance(value,list):
+			self.deps_man[node].extend(value)
+		else:
+			self.deps_man[node].append(value)
 	def launch_node(self):
 		try:
 			return self.p_ln
@@ -232,9 +230,8 @@
 			except KeyError:
 				pass
 		lst=[env[a]for a in vars_lst]
-		ret=Utils.h_list(lst)
+		cache[idx]=ret=Utils.h_list(lst)
 		Logs.debug('envhash: %s %r',Utils.to_hex(ret),lst)
-		cache[idx]=ret
 		return ret
 	def get_tgen_by_name(self,name):
 		cache=self.task_gen_cache_names
@@ -249,20 +246,21 @@
 			return cache[name]
 		except KeyError:
 			raise Errors.WafError('Could not find a task generator for the name %r'%name)
-	def progress_line(self,state,total,col1,col2):
+	def progress_line(self,idx,total,col1,col2):
+		if not sys.stderr.isatty():
+			return''
 		n=len(str(total))
 		Utils.rot_idx+=1
 		ind=Utils.rot_chr[Utils.rot_idx%4]
-		pc=(100.*state)/total
-		eta=str(self.timer)
-		fs="[%%%dd/%%%dd][%%s%%2d%%%%%%s][%s]["%(n,n,ind)
-		left=fs%(state,total,col1,pc,col2)
-		right='][%s%s%s]'%(col1,eta,col2)
+		pc=(100.*idx)/total
+		fs="[%%%dd/%%d][%%s%%2d%%%%%%s][%s]["%(n,ind)
+		left=fs%(idx,total,col1,pc,col2)
+		right='][%s%s%s]'%(col1,self.timer,col2)
 		cols=Logs.get_term_cols()-len(left)-len(right)+2*len(col1)+2*len(col2)
 		if cols<7:cols=7
-		ratio=((cols*state)//total)-1
+		ratio=((cols*idx)//total)-1
 		bar=('='*ratio+'>').ljust(cols)
-		msg=Utils.indicator%(left,bar,right)
+		msg=Logs.indicator%(left,bar,right)
 		return msg
 	def declare_chain(self,*k,**kw):
 		return TaskGen.declare_chain(*k,**kw)
@@ -303,14 +301,14 @@
 		return''
 	def get_group_idx(self,tg):
 		se=id(tg)
-		for i in range(len(self.groups)):
-			for t in self.groups[i]:
+		for i,tmp in enumerate(self.groups):
+			for t in tmp:
 				if id(t)==se:
 					return i
 		return None
 	def add_group(self,name=None,move=True):
 		if name and name in self.group_names:
-			Logs.error('add_group: name %s already present'%name)
+			raise Errors.WafError('add_group: name %s already present',name)
 		g=[]
 		self.group_names[name]=g
 		self.groups.append(g)
@@ -319,9 +317,10 @@
 	def set_group(self,idx):
 		if isinstance(idx,str):
 			g=self.group_names[idx]
-			for i in range(len(self.groups)):
-				if id(g)==id(self.groups[i]):
+			for i,tmp in enumerate(self.groups):
+				if id(g)==id(tmp):
 					self.current_group=i
+					break
 		else:
 			self.current_group=idx
 	def total(self):
@@ -338,8 +337,6 @@
 		min_grp=0
 		for name in self.targets.split(','):
 			tg=self.get_tgen_by_name(name)
-			if not tg:
-				raise Errors.WafError('target %r does not exist'%name)
 			m=self.get_group_idx(tg)
 			if m>min_grp:
 				min_grp=m
@@ -347,6 +344,11 @@
 			elif m==min_grp:
 				to_post.append(tg)
 		return(min_grp,to_post)
+	def get_all_task_gen(self):
+		lst=[]
+		for g in self.groups:
+			lst.extend(g)
+		return lst
 	def post_group(self):
 		if self.targets=='*':
 			for tg in self.groups[self.cur]:
@@ -370,6 +372,12 @@
 					tg.post()
 		else:
 			ln=self.launch_node()
+			if ln.is_child_of(self.bldnode):
+				Logs.warn('Building from the build directory, forcing --targets=*')
+				ln=self.srcnode
+			elif not ln.is_child_of(self.srcnode):
+				Logs.warn('CWD %s is not under %s, forcing --targets=* (run distclean?)',ln.abspath(),self.srcnode.abspath())
+				ln=self.srcnode
 			for tg in self.groups[self.cur]:
 				try:
 					f=tg.post
@@ -381,10 +389,10 @@
 	def get_tasks_group(self,idx):
 		tasks=[]
 		for tg in self.groups[idx]:
-			if isinstance(tg,Task.TaskBase):
-				tasks.append(tg)
-			else:
+			try:
 				tasks.extend(tg.tasks)
+			except AttributeError:
+				tasks.append(tg)
 		return tasks
 	def get_build_iterator(self):
 		self.cur=0
@@ -409,74 +417,149 @@
 			yield tasks
 		while 1:
 			yield[]
+	def install_files(self,dest,files,**kw):
+		assert(dest)
+		tg=self(features='install_task',install_to=dest,install_from=files,**kw)
+		tg.dest=tg.install_to
+		tg.type='install_files'
+		if not kw.get('postpone',True):
+			tg.post()
+		return tg
+	def install_as(self,dest,srcfile,**kw):
+		assert(dest)
+		tg=self(features='install_task',install_to=dest,install_from=srcfile,**kw)
+		tg.dest=tg.install_to
+		tg.type='install_as'
+		if not kw.get('postpone',True):
+			tg.post()
+		return tg
+	def symlink_as(self,dest,src,**kw):
+		assert(dest)
+		tg=self(features='install_task',install_to=dest,install_from=src,**kw)
+		tg.dest=tg.install_to
+		tg.type='symlink_as'
+		tg.link=src
+		if not kw.get('postpone',True):
+			tg.post()
+		return tg
+@TaskGen.feature('install_task')
+@TaskGen.before_method('process_rule','process_source')
+def process_install_task(self):
+	self.add_install_task(**self.__dict__)
+@TaskGen.taskgen_method
+def add_install_task(self,**kw):
+	if not self.bld.is_install:
+		return
+	if not kw['install_to']:
+		return
+	if kw['type']=='symlink_as'and Utils.is_win32:
+		if kw.get('win32_install'):
+			kw['type']='install_as'
+		else:
+			return
+	tsk=self.install_task=self.create_task('inst')
+	tsk.chmod=kw.get('chmod',Utils.O644)
+	tsk.link=kw.get('link','')or kw.get('install_from','')
+	tsk.relative_trick=kw.get('relative_trick',False)
+	tsk.type=kw['type']
+	tsk.install_to=tsk.dest=kw['install_to']
+	tsk.install_from=kw['install_from']
+	tsk.relative_base=kw.get('cwd')or kw.get('relative_base',self.path)
+	tsk.init_files()
+	if not kw.get('postpone',True):
+		tsk.run_now()
+	return tsk
+@TaskGen.taskgen_method
+def add_install_files(self,**kw):
+	kw['type']='install_files'
+	return self.add_install_task(**kw)
+@TaskGen.taskgen_method
+def add_install_as(self,**kw):
+	kw['type']='install_as'
+	return self.add_install_task(**kw)
+@TaskGen.taskgen_method
+def add_symlink_as(self,**kw):
+	kw['type']='symlink_as'
+	return self.add_install_task(**kw)
 class inst(Task.Task):
-	color='CYAN'
-	def post(self):
-		buf=[]
-		for x in self.source:
-			if isinstance(x,waflib.Node.Node):
-				y=x
-			else:
-				y=self.path.find_resource(x)
-				if not y:
-					if Logs.verbose:
-						Logs.warn('Could not find %s immediately (may cause broken builds)'%x)
-					idx=self.generator.bld.get_group_idx(self)
-					for tg in self.generator.bld.groups[idx]:
-						if not isinstance(tg,inst)and id(tg)!=id(self):
-							tg.post()
-						y=self.path.find_resource(x)
-						if y:
-							break
-					else:
-						raise Errors.WafError('could not find %r in %r'%(x,self.path))
-			buf.append(y)
-		self.inputs=buf
+	def __str__(self):
+		return''
+	def uid(self):
+		lst=self.inputs+self.outputs+[self.link,self.generator.path.abspath()]
+		return Utils.h_list(lst)
+	def init_files(self):
+		if self.type=='symlink_as':
+			inputs=[]
+		else:
+			inputs=self.generator.to_nodes(self.install_from)
+			if self.type=='install_as':
+				assert len(inputs)==1
+		self.set_inputs(inputs)
+		dest=self.get_install_path()
+		outputs=[]
+		if self.type=='symlink_as':
+			if self.relative_trick:
+				self.link=os.path.relpath(self.link,os.path.dirname(dest))
+			outputs.append(self.generator.bld.root.make_node(dest))
+		elif self.type=='install_as':
+			outputs.append(self.generator.bld.root.make_node(dest))
+		else:
+			for y in inputs:
+				if self.relative_trick:
+					destfile=os.path.join(dest,y.path_from(self.relative_base))
+				else:
+					destfile=os.path.join(dest,y.name)
+				outputs.append(self.generator.bld.root.make_node(destfile))
+		self.set_outputs(outputs)
 	def runnable_status(self):
 		ret=super(inst,self).runnable_status()
-		if ret==Task.SKIP_ME:
+		if ret==Task.SKIP_ME and self.generator.bld.is_install:
 			return Task.RUN_ME
 		return ret
-	def __str__(self):
-		return''
-	def run(self):
-		return self.generator.exec_task()
+	def post_run(self):
+		pass
 	def get_install_path(self,destdir=True):
-		dest=Utils.subst_vars(self.dest,self.env)
-		dest=dest.replace('/',os.sep)
+		if isinstance(self.install_to,Node.Node):
+			dest=self.install_to.abspath()
+		else:
+			dest=Utils.subst_vars(self.install_to,self.env)
 		if destdir and Options.options.destdir:
 			dest=os.path.join(Options.options.destdir,os.path.splitdrive(dest)[1].lstrip(os.sep))
 		return dest
-	def exec_install_files(self):
-		destpath=self.get_install_path()
-		if not destpath:
-			raise Errors.WafError('unknown installation path %r'%self.generator)
-		for x,y in zip(self.source,self.inputs):
-			if self.relative_trick:
-				destfile=os.path.join(destpath,y.path_from(self.path))
-				Utils.check_dir(os.path.dirname(destfile))
-			else:
-				destfile=os.path.join(destpath,y.name)
-			self.generator.bld.do_install(y.abspath(),destfile,self.chmod)
-	def exec_install_as(self):
-		destfile=self.get_install_path()
-		self.generator.bld.do_install(self.inputs[0].abspath(),destfile,self.chmod)
-	def exec_symlink_as(self):
-		destfile=self.get_install_path()
-		self.generator.bld.do_link(self.link,destfile)
-class InstallContext(BuildContext):
-	'''installs the targets on the system'''
-	cmd='install'
-	def __init__(self,**kw):
-		super(InstallContext,self).__init__(**kw)
-		self.uninstall=[]
-		self.is_install=INSTALL
-	def do_install(self,src,tgt,chmod=Utils.O644):
-		d,_=os.path.split(tgt)
-		if not d:
-			raise Errors.WafError('Invalid installation given %r->%r'%(src,tgt))
-		Utils.check_dir(d)
-		srclbl=src.replace(self.srcnode.abspath()+os.sep,'')
+	def copy_fun(self,src,tgt):
+		if Utils.is_win32 and len(tgt)>259 and not tgt.startswith('\\\\?\\'):
+			tgt='\\\\?\\'+tgt
+		shutil.copy2(src,tgt)
+		os.chmod(tgt,self.chmod)
+	def rm_empty_dirs(self,tgt):
+		while tgt:
+			tgt=os.path.dirname(tgt)
+			try:
+				os.rmdir(tgt)
+			except OSError:
+				break
+	def run(self):
+		is_install=self.generator.bld.is_install
+		if not is_install:
+			return
+		for x in self.outputs:
+			if is_install==INSTALL:
+				x.parent.mkdir()
+		if self.type=='symlink_as':
+			fun=is_install==INSTALL and self.do_link or self.do_unlink
+			fun(self.link,self.outputs[0].abspath())
+		else:
+			fun=is_install==INSTALL and self.do_install or self.do_uninstall
+			launch_node=self.generator.bld.launch_node()
+			for x,y in zip(self.inputs,self.outputs):
+				fun(x.abspath(),y.abspath(),x.path_from(launch_node))
+	def run_now(self):
+		status=self.runnable_status()
+		if status not in(Task.RUN_ME,Task.SKIP_ME):
+			raise Errors.TaskNotReady('Could not process %r: status %r'%(self,status))
+		self.run()
+		self.hasrun=Task.SUCCESS
+	def do_install(self,src,tgt,lbl,**kw):
 		if not Options.options.force:
 			try:
 				st1=os.stat(tgt)
@@ -485,124 +568,72 @@
 				pass
 			else:
 				if st1.st_mtime+2>=st2.st_mtime and st1.st_size==st2.st_size:
-					if not self.progress_bar:
-						Logs.info('- install %s (from %s)'%(tgt,srclbl))
+					if not self.generator.bld.progress_bar:
+						Logs.info('- install %s (from %s)',tgt,lbl)
 					return False
-		if not self.progress_bar:
-			Logs.info('+ install %s (from %s)'%(tgt,srclbl))
+		if not self.generator.bld.progress_bar:
+			Logs.info('+ install %s (from %s)',tgt,lbl)
+		try:
+			os.chmod(tgt,Utils.O644|stat.S_IMODE(os.stat(tgt).st_mode))
+		except EnvironmentError:
+			pass
 		try:
 			os.remove(tgt)
 		except OSError:
 			pass
 		try:
-			shutil.copy2(src,tgt)
-			os.chmod(tgt,chmod)
-		except IOError:
+			self.copy_fun(src,tgt)
+		except EnvironmentError as e:
+			if not os.path.exists(src):
+				Logs.error('File %r does not exist',src)
+			elif not os.path.isfile(src):
+				Logs.error('Input %r is not a file',src)
+			raise Errors.WafError('Could not install the file %r'%tgt,e)
+	def do_link(self,src,tgt,**kw):
+		if os.path.islink(tgt)and os.readlink(tgt)==src:
+			if not self.generator.bld.progress_bar:
+				Logs.info('- symlink %s (to %s)',tgt,src)
+		else:
 			try:
-				os.stat(src)
-			except(OSError,IOError):
-				Logs.error('File %r does not exist'%src)
-			raise Errors.WafError('Could not install the file %r'%tgt)
-	def do_link(self,src,tgt):
-		d,_=os.path.split(tgt)
-		Utils.check_dir(d)
-		link=False
-		if not os.path.islink(tgt):
-			link=True
-		elif os.readlink(tgt)!=src:
-			link=True
-		if link:
-			try:os.remove(tgt)
-			except OSError:pass
-			if not self.progress_bar:
-				Logs.info('+ symlink %s (to %s)'%(tgt,src))
+				os.remove(tgt)
+			except OSError:
+				pass
+			if not self.generator.bld.progress_bar:
+				Logs.info('+ symlink %s (to %s)',tgt,src)
 			os.symlink(src,tgt)
-		else:
-			if not self.progress_bar:
-				Logs.info('- symlink %s (to %s)'%(tgt,src))
-	def run_task_now(self,tsk,postpone):
-		tsk.post()
-		if not postpone:
-			if tsk.runnable_status()==Task.ASK_LATER:
-				raise self.WafError('cannot post the task %r'%tsk)
-			tsk.run()
-	def install_files(self,dest,files,env=None,chmod=Utils.O644,relative_trick=False,cwd=None,add=True,postpone=True):
-		tsk=inst(env=env or self.env)
-		tsk.bld=self
-		tsk.path=cwd or self.path
-		tsk.chmod=chmod
-		if isinstance(files,waflib.Node.Node):
-			tsk.source=[files]
-		else:
-			tsk.source=Utils.to_list(files)
-		tsk.dest=dest
-		tsk.exec_task=tsk.exec_install_files
-		tsk.relative_trick=relative_trick
-		if add:self.add_to_group(tsk)
-		self.run_task_now(tsk,postpone)
-		return tsk
-	def install_as(self,dest,srcfile,env=None,chmod=Utils.O644,cwd=None,add=True,postpone=True):
-		tsk=inst(env=env or self.env)
-		tsk.bld=self
-		tsk.path=cwd or self.path
-		tsk.chmod=chmod
-		tsk.source=[srcfile]
-		tsk.dest=dest
-		tsk.exec_task=tsk.exec_install_as
-		if add:self.add_to_group(tsk)
-		self.run_task_now(tsk,postpone)
-		return tsk
-	def symlink_as(self,dest,src,env=None,cwd=None,add=True,postpone=True):
-		if Utils.is_win32:
-			return
-		tsk=inst(env=env or self.env)
-		tsk.bld=self
-		tsk.dest=dest
-		tsk.path=cwd or self.path
-		tsk.source=[]
-		tsk.link=src
-		tsk.exec_task=tsk.exec_symlink_as
-		if add:self.add_to_group(tsk)
-		self.run_task_now(tsk,postpone)
-		return tsk
-class UninstallContext(InstallContext):
-	'''removes the targets installed'''
-	cmd='uninstall'
-	def __init__(self,**kw):
-		super(UninstallContext,self).__init__(**kw)
-		self.is_install=UNINSTALL
-	def do_install(self,src,tgt,chmod=Utils.O644):
-		if not self.progress_bar:
-			Logs.info('- remove %s'%tgt)
-		self.uninstall.append(tgt)
+	def do_uninstall(self,src,tgt,lbl,**kw):
+		if not self.generator.bld.progress_bar:
+			Logs.info('- remove %s',tgt)
 		try:
 			os.remove(tgt)
-		except OSError ,e:
+		except OSError as e:
 			if e.errno!=errno.ENOENT:
 				if not getattr(self,'uninstall_error',None):
 					self.uninstall_error=True
 					Logs.warn('build: some files could not be uninstalled (retry with -vv to list them)')
 				if Logs.verbose>1:
-					Logs.warn('could not remove %s (error code %r)'%(e.filename,e.errno))
-		while tgt:
-			tgt=os.path.dirname(tgt)
-			try:
-				os.rmdir(tgt)
-			except OSError:
-				break
-	def do_link(self,src,tgt):
+					Logs.warn('Could not remove %s (error code %r)',e.filename,e.errno)
+		self.rm_empty_dirs(tgt)
+	def do_unlink(self,src,tgt,**kw):
 		try:
-			if not self.progress_bar:
-				Logs.info('- unlink %s'%tgt)
+			if not self.generator.bld.progress_bar:
+				Logs.info('- remove %s',tgt)
 			os.remove(tgt)
 		except OSError:
 			pass
-		while tgt:
-			tgt=os.path.dirname(tgt)
-			try:
-				os.rmdir(tgt)
-			except OSError:
-				break
+		self.rm_empty_dirs(tgt)
+class InstallContext(BuildContext):
+	'''installs the targets on the system'''
+	cmd='install'
+	def __init__(self,**kw):
+		super(InstallContext,self).__init__(**kw)
+		self.is_install=INSTALL
+class UninstallContext(InstallContext):
+	'''removes the targets installed'''
+	cmd='uninstall'
+	def __init__(self,**kw):
+		super(UninstallContext,self).__init__(**kw)
+		self.is_install=UNINSTALL
 	def execute(self):
 		try:
 			def runnable_status(self):
@@ -627,13 +658,17 @@
 	def clean(self):
 		Logs.debug('build: clean called')
 		if self.bldnode!=self.srcnode:
-			lst=[self.root.find_or_declare(f)for f in self.env[CFG_FILES]]
-			for n in self.bldnode.ant_glob('**/*',excl='lock* *conf_check_*/** config.log c4che/*',quiet=True):
+			lst=[]
+			for env in self.all_envs.values():
+				lst.extend(self.root.find_or_declare(f)for f in env[CFG_FILES])
+			for n in self.bldnode.ant_glob('**/*',excl='.lock* *conf_check_*/** config.log c4che/*',quiet=True):
 				if n in lst:
 					continue
 				n.delete()
 		self.root.children={}
-		for v in'node_deps task_sigs raw_deps'.split():
+		for v in SAVED_ATTRS:
+			if v=='root':
+				continue
 			setattr(self,v,{})
 class ListContext(BuildContext):
 	'''lists the targets to execute'''
@@ -655,11 +690,9 @@
 					f()
 		try:
 			self.get_tgen_by_name('')
-		except:
+		except Errors.WafError:
 			pass
-		lst=list(self.task_gen_cache_names.keys())
-		lst.sort()
-		for k in lst:
+		for k in sorted(self.task_gen_cache_names.keys()):
 			Logs.pprint('GREEN',k)
 class StepContext(BuildContext):
 	'''executes tasks in a step-by-step fashion, for debugging'''
@@ -672,8 +705,13 @@
 			Logs.warn('Add a pattern for the debug build, for example "waf step --files=main.c,app"')
 			BuildContext.compile(self)
 			return
+		targets=[]
+		if self.targets and self.targets!='*':
+			targets=self.targets.split(',')
 		for g in self.groups:
 			for tg in g:
+				if targets and tg.name not in targets:
+					continue
 				try:
 					f=tg.post
 				except AttributeError:
@@ -699,7 +737,7 @@
 								break
 						if do_exec:
 							ret=tsk.run()
-							Logs.info('%s -> exit %r'%(str(tsk),ret))
+							Logs.info('%s -> exit %r',tsk,ret)
 	def get_matcher(self,pat):
 		inn=True
 		out=True
@@ -727,5 +765,10 @@
 			else:
 				return pattern.match(node.abspath())
 		return match
-BuildContext.store=Utils.nogc(BuildContext.store)
-BuildContext.restore=Utils.nogc(BuildContext.restore)
+class EnvContext(BuildContext):
+	fun=cmd=None
+	def execute(self):
+		self.restore()
+		if not self.all_envs:
+			self.load_envs()
+		self.recurse([self.run_dir])
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/ConfigSet.py glmark2-2021.02/waflib/ConfigSet.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/ConfigSet.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/ConfigSet.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,7 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
 import copy,re,os
 from waflib import Logs,Utils
 re_imp=re.compile('^(#)*?([^#=]*?)\ =\ (.*?)$',re.M)
@@ -26,12 +24,14 @@
 		keys=list(keys)
 		keys.sort()
 		return keys
+	def __iter__(self):
+		return iter(self.keys())
 	def __str__(self):
 		return"\n".join(["%r %r"%(x,self.__getitem__(x))for x in self.keys()])
 	def __getitem__(self,key):
 		try:
 			while 1:
-				x=self.table.get(key,None)
+				x=self.table.get(key)
 				if not x is None:
 					return x
 				self=self.parent
@@ -71,6 +71,7 @@
 			for x in keys:
 				tbl[x]=copy.deepcopy(tbl[x])
 			self.table=tbl
+		return self
 	def get_flat(self,key):
 		s=self[key]
 		if isinstance(s,str):return s
@@ -79,21 +80,24 @@
 		try:
 			value=self.table[key]
 		except KeyError:
-			try:value=self.parent[key]
-			except AttributeError:value=[]
-			if isinstance(value,list):
-				value=value[:]
+			try:
+				value=self.parent[key]
+			except AttributeError:
+				value=[]
 			else:
-				value=[value]
+				if isinstance(value,list):
+					value=value[:]
+				else:
+					value=[value]
+			self.table[key]=value
 		else:
 			if not isinstance(value,list):
-				value=[value]
-		self.table[key]=value
+				self.table[key]=value=[value]
 		return value
 	def append_value(self,var,val):
-		current_value=self._get_list_value_for_modification(var)
 		if isinstance(val,str):
 			val=[val]
+		current_value=self._get_list_value_for_modification(var)
 		current_value.extend(val)
 	def prepend_value(self,var,val):
 		if isinstance(val,str):
@@ -122,30 +126,34 @@
 			os.makedirs(os.path.split(filename)[0])
 		except OSError:
 			pass
-		f=None
+		buf=[]
+		merged_table=self.get_merged_dict()
+		keys=list(merged_table.keys())
+		keys.sort()
 		try:
-			f=open(filename,'w')
-			merged_table=self.get_merged_dict()
-			keys=list(merged_table.keys())
-			keys.sort()
-			for k in keys:
-				if k!='undo_stack':
-					f.write('%s = %r\n'%(k,merged_table[k]))
-		finally:
-			if f:
-				f.close()
+			fun=ascii
+		except NameError:
+			fun=repr
+		for k in keys:
+			if k!='undo_stack':
+				buf.append('%s = %s\n'%(k,fun(merged_table[k])))
+		Utils.writef(filename,''.join(buf))
 	def load(self,filename):
 		tbl=self.table
-		code=Utils.readf(filename)
+		code=Utils.readf(filename,m='rU')
 		for m in re_imp.finditer(code):
 			g=m.group
 			tbl[g(2)]=eval(g(3))
-		Logs.debug('env: %s'%str(self.table))
+		Logs.debug('env: %s',self.table)
 	def update(self,d):
-		for k,v in d.items():
-			self[k]=v
+		self.table.update(d)
 	def stash(self):
-		self.undo_stack=self.undo_stack+[self.table]
-		self.table=self.table.copy()
+		orig=self.table
+		tbl=self.table=self.table.copy()
+		for x in tbl.keys():
+			tbl[x]=copy.deepcopy(tbl[x])
+		self.undo_stack=self.undo_stack+[orig]
+	def commit(self):
+		self.undo_stack.pop(-1)
 	def revert(self):
 		self.table=self.undo_stack.pop(-1)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Configure.py glmark2-2021.02/waflib/Configure.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Configure.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Configure.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,54 +1,15 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,shlex,sys,time
+import os,shlex,sys,time,re,shutil
 from waflib import ConfigSet,Utils,Options,Logs,Context,Build,Errors
-try:
-	from urllib import request
-except:
-	from urllib import urlopen
-else:
-	urlopen=request.urlopen
-BREAK='break'
-CONTINUE='continue'
 WAF_CONFIG_LOG='config.log'
 autoconfig=False
 conf_template='''# project %(app)s configured on %(now)s by
 # waf %(wafver)s (abi %(abi)s, python %(pyver)x on %(systype)s)
 # using %(args)s
 #'''
-def download_check(node):
-	pass
-def download_tool(tool,force=False,ctx=None):
-	for x in Utils.to_list(Context.remote_repo):
-		for sub in Utils.to_list(Context.remote_locs):
-			url='/'.join((x,sub,tool+'.py'))
-			try:
-				web=urlopen(url)
-				try:
-					if web.getcode()!=200:
-						continue
-				except AttributeError:
-					pass
-			except Exception:
-				continue
-			else:
-				tmp=ctx.root.make_node(os.sep.join((Context.waf_dir,'waflib','extras',tool+'.py')))
-				tmp.write(web.read())
-				Logs.warn('Downloaded %s from %s'%(tool,url))
-				download_check(tmp)
-				try:
-					module=Context.load_tool(tool)
-				except:
-					Logs.warn('The tool %s from %s is unusable'%(tool,url))
-					try:
-						tmp.delete()
-					except:
-						pass
-					continue
-				return module
-	raise Errors.WafError('Could not load the Waf tool')
 class ConfigurationContext(Context.Context):
 	'''configures the project'''
 	cmd='configure'
@@ -96,10 +57,11 @@
 			out=getattr(Context.g_module,Context.OUT,None)
 		if not out:
 			out=Options.lockfile.replace('.lock-waf_%s_'%sys.platform,'').replace('.lock-waf','')
+		out=os.path.realpath(out)
 		self.bldnode=(os.path.isabs(out)and self.root or self.path).make_node(out)
 		self.bldnode.mkdir()
 		if not os.path.isdir(self.bldnode.abspath()):
-			conf.fatal('could not create the build directory %s'%self.bldnode.abspath())
+			conf.fatal('Could not create the build directory %s'%self.bldnode.abspath())
 	def execute(self):
 		self.init_dirs()
 		self.cachedir=self.bldnode.make_node(Build.CACHE_DIR)
@@ -111,17 +73,12 @@
 			ver=getattr(Context.g_module,'VERSION','')
 			if ver:
 				app="%s (%s)"%(app,ver)
-		now=time.ctime()
-		pyver=sys.hexversion
-		systype=sys.platform
-		args=" ".join(sys.argv)
-		wafver=Context.WAFVERSION
-		abi=Context.ABI
-		self.to_log(conf_template%vars())
+		params={'now':time.ctime(),'pyver':sys.hexversion,'systype':sys.platform,'args':" ".join(sys.argv),'wafver':Context.WAFVERSION,'abi':Context.ABI,'app':app}
+		self.to_log(conf_template%params)
 		self.msg('Setting top to',self.srcnode.abspath())
 		self.msg('Setting out to',self.bldnode.abspath())
 		if id(self.srcnode)==id(self.bldnode):
-			Logs.warn('Setting top == out (remember to use "update_outputs")')
+			Logs.warn('Setting top == out')
 		elif id(self.path)!=id(self.srcnode):
 			if self.srcnode.is_child_of(self.path):
 				Logs.warn('Are you certain that you do not want to set top="." ?')
@@ -130,27 +87,37 @@
 		Context.top_dir=self.srcnode.abspath()
 		Context.out_dir=self.bldnode.abspath()
 		env=ConfigSet.ConfigSet()
-		env['argv']=sys.argv
-		env['options']=Options.options.__dict__
+		env.argv=sys.argv
+		env.options=Options.options.__dict__
+		env.config_cmd=self.cmd
 		env.run_dir=Context.run_dir
 		env.top_dir=Context.top_dir
 		env.out_dir=Context.out_dir
-		env['hash']=self.hash
-		env['files']=self.files
-		env['environ']=dict(self.environ)
-		if not self.env.NO_LOCK_IN_RUN:
-			env.store(Context.run_dir+os.sep+Options.lockfile)
-		if not self.env.NO_LOCK_IN_TOP:
-			env.store(Context.top_dir+os.sep+Options.lockfile)
-		if not self.env.NO_LOCK_IN_OUT:
-			env.store(Context.out_dir+os.sep+Options.lockfile)
+		env.hash=self.hash
+		env.files=self.files
+		env.environ=dict(self.environ)
+		if not self.env.NO_LOCK_IN_RUN and not getattr(Options.options,'no_lock_in_run'):
+			env.store(os.path.join(Context.run_dir,Options.lockfile))
+		if not self.env.NO_LOCK_IN_TOP and not getattr(Options.options,'no_lock_in_top'):
+			env.store(os.path.join(Context.top_dir,Options.lockfile))
+		if not self.env.NO_LOCK_IN_OUT and not getattr(Options.options,'no_lock_in_out'):
+			env.store(os.path.join(Context.out_dir,Options.lockfile))
 	def prepare_env(self,env):
 		if not env.PREFIX:
-			env.PREFIX=os.path.abspath(os.path.expanduser(Options.options.prefix))
+			if Options.options.prefix or Utils.is_win32:
+				env.PREFIX=Utils.sane_path(Options.options.prefix)
+			else:
+				env.PREFIX=''
 		if not env.BINDIR:
-			env.BINDIR=Utils.subst_vars('${PREFIX}/bin',env)
+			if Options.options.bindir:
+				env.BINDIR=Utils.sane_path(Options.options.bindir)
+			else:
+				env.BINDIR=Utils.subst_vars('${PREFIX}/bin',env)
 		if not env.LIBDIR:
-			env.LIBDIR=Utils.subst_vars('${PREFIX}/lib',env)
+			if Options.options.libdir:
+				env.LIBDIR=Utils.sane_path(Options.options.libdir)
+			else:
+				env.LIBDIR=Utils.subst_vars('${PREFIX}/lib%s'%Utils.lib64(),env)
 	def store(self):
 		n=self.cachedir.make_node('build.config.py')
 		n.write('version = 0x%x\ntools = %r\n'%(Context.HEXVERSION,self.tools))
@@ -159,26 +126,22 @@
 		for key in self.all_envs:
 			tmpenv=self.all_envs[key]
 			tmpenv.store(os.path.join(self.cachedir.abspath(),key+Build.CACHE_SUFFIX))
-	def load(self,input,tooldir=None,funs=None,download=True):
+	def load(self,input,tooldir=None,funs=None,with_sys_path=True,cache=False):
 		tools=Utils.to_list(input)
 		if tooldir:tooldir=Utils.to_list(tooldir)
 		for tool in tools:
-			mag=(tool,id(self.env),funs)
-			if mag in self.tool_cache:
-				self.to_log('(tool %s is already loaded, skipping)'%tool)
-				continue
-			self.tool_cache.append(mag)
+			if cache:
+				mag=(tool,id(self.env),tooldir,funs)
+				if mag in self.tool_cache:
+					self.to_log('(tool %s is already loaded, skipping)'%tool)
+					continue
+				self.tool_cache.append(mag)
 			module=None
 			try:
-				module=Context.load_tool(tool,tooldir)
-			except ImportError ,e:
-				if Options.options.download:
-					module=download_tool(tool,ctx=self)
-					if not module:
-						self.fatal('Could not load the Waf tool %r or download a suitable replacement from the repository (sys.path %r)\n%s'%(tool,sys.path,e))
-				else:
-					self.fatal('Could not load the Waf tool %r from %r (try the --download option?):\n%s'%(tool,sys.path,e))
-			except Exception ,e:
+				module=Context.load_tool(tool,tooldir,ctx=self,with_sys_path=with_sys_path)
+			except ImportError as e:
+				self.fatal('Could not load the Waf tool %r from %r\n%s'%(tool,sys.path,e))
+			except Exception as e:
 				self.to_log('imp %r (%r & %r)'%(tool,tooldir,funs))
 				self.to_log(Utils.ex_stack())
 				raise
@@ -192,25 +155,15 @@
 			self.tools.append({'tool':tool,'tooldir':tooldir,'funs':funs})
 	def post_recurse(self,node):
 		super(ConfigurationContext,self).post_recurse(node)
-		self.hash=hash((self.hash,node.read('rb')))
+		self.hash=Utils.h_list((self.hash,node.read('rb')))
 		self.files.append(node.abspath())
 	def eval_rules(self,rules):
 		self.rules=Utils.to_list(rules)
 		for x in self.rules:
 			f=getattr(self,x)
-			if not f:self.fatal("No such method '%s'."%x)
-			try:
-				f()
-			except Exception ,e:
-				ret=self.err_handler(x,e)
-				if ret==BREAK:
-					break
-				elif ret==CONTINUE:
-					continue
-				else:
-					raise
-	def err_handler(self,fun,error):
-		pass
+			if not f:
+				self.fatal('No such configuration function %r'%x)
+			f()
 def conf(f):
 	def fun(*k,**kw):
 		mandatory=True
@@ -219,97 +172,196 @@
 			del kw['mandatory']
 		try:
 			return f(*k,**kw)
-		except Errors.ConfigurationError ,e:
+		except Errors.ConfigurationError:
 			if mandatory:
-				raise e
+				raise
+	fun.__name__=f.__name__
 	setattr(ConfigurationContext,f.__name__,fun)
 	setattr(Build.BuildContext,f.__name__,fun)
 	return f
-def add_os_flags(self,var,dest=None):
-	try:self.env.append_value(dest or var,shlex.split(self.environ[var]))
-	except KeyError:pass
+@conf
+def add_os_flags(self,var,dest=None,dup=False):
+	try:
+		flags=shlex.split(self.environ[var])
+	except KeyError:
+		return
+	if dup or''.join(flags)not in''.join(Utils.to_list(self.env[dest or var])):
+		self.env.append_value(dest or var,flags)
+@conf
 def cmd_to_list(self,cmd):
-	if isinstance(cmd,str)and cmd.find(' '):
-		try:
-			os.stat(cmd)
-		except OSError:
+	if isinstance(cmd,str):
+		if os.path.isfile(cmd):
+			return[cmd]
+		if os.sep=='/':
 			return shlex.split(cmd)
 		else:
-			return[cmd]
+			try:
+				return shlex.split(cmd,posix=False)
+			except TypeError:
+				return shlex.split(cmd)
 	return cmd
-def check_waf_version(self,mini='1.6.0',maxi='1.7.0'):
-	self.start_msg('Checking for waf version in %s-%s'%(str(mini),str(maxi)))
+@conf
+def check_waf_version(self,mini='1.8.99',maxi='2.0.0',**kw):
+	self.start_msg('Checking for waf version in %s-%s'%(str(mini),str(maxi)),**kw)
 	ver=Context.HEXVERSION
 	if Utils.num2ver(mini)>ver:
 		self.fatal('waf version should be at least %r (%r found)'%(Utils.num2ver(mini),ver))
 	if Utils.num2ver(maxi)<ver:
 		self.fatal('waf version should be at most %r (%r found)'%(Utils.num2ver(maxi),ver))
-	self.end_msg('ok')
+	self.end_msg('ok',**kw)
+@conf
 def find_file(self,filename,path_list=[]):
 	for n in Utils.to_list(filename):
 		for d in Utils.to_list(path_list):
-			p=os.path.join(d,n)
+			p=os.path.expanduser(os.path.join(d,n))
 			if os.path.exists(p):
 				return p
 	self.fatal('Could not find %r'%filename)
+@conf
 def find_program(self,filename,**kw):
 	exts=kw.get('exts',Utils.is_win32 and'.exe,.com,.bat,.cmd'or',.sh,.pl,.py')
-	environ=kw.get('environ',os.environ)
+	environ=kw.get('environ',getattr(self,'environ',os.environ))
 	ret=''
 	filename=Utils.to_list(filename)
+	msg=kw.get('msg',', '.join(filename))
 	var=kw.get('var','')
 	if not var:
-		var=filename[0].upper()
-	if self.env[var]:
-		ret=self.env[var]
-	elif var in environ:
-		ret=environ[var]
+		var=re.sub(r'[-.]','_',filename[0].upper())
 	path_list=kw.get('path_list','')
-	if not ret:
-		if path_list:
-			path_list=Utils.to_list(path_list)
+	if path_list:
+		path_list=Utils.to_list(path_list)
+	else:
+		path_list=environ.get('PATH','').split(os.pathsep)
+	if kw.get('value'):
+		ret=self.cmd_to_list(kw['value'])
+	elif var in environ:
+		ret=self.cmd_to_list(environ[var])
+	elif self.env[var]:
+		ret=self.cmd_to_list(self.env[var])
+	else:
+		if not ret:
+			ret=self.find_binary(filename,exts.split(','),path_list)
+		if not ret and Utils.winreg:
+			ret=Utils.get_registry_app_path(Utils.winreg.HKEY_CURRENT_USER,filename)
+		if not ret and Utils.winreg:
+			ret=Utils.get_registry_app_path(Utils.winreg.HKEY_LOCAL_MACHINE,filename)
+		ret=self.cmd_to_list(ret)
+	if ret:
+		if len(ret)==1:
+			retmsg=ret[0]
 		else:
-			path_list=environ.get('PATH','').split(os.pathsep)
-		if not isinstance(filename,list):
-			filename=[filename]
-		for a in exts.split(','):
-			if ret:
-				break
-			for b in filename:
-				if ret:
-					break
-				for c in path_list:
-					if ret:
-						break
-					x=os.path.expanduser(os.path.join(c,b+a))
-					if os.path.isfile(x):
-						ret=x
-	if not ret and Utils.winreg:
-		ret=Utils.get_registry_app_path(Utils.winreg.HKEY_CURRENT_USER,filename)
-	if not ret and Utils.winreg:
-		ret=Utils.get_registry_app_path(Utils.winreg.HKEY_LOCAL_MACHINE,filename)
-	self.msg('Checking for program '+','.join(filename),ret or False)
-	self.to_log('find program=%r paths=%r var=%r -> %r'%(filename,path_list,var,ret))
+			retmsg=ret
+	else:
+		retmsg=False
+	self.msg('Checking for program %r'%msg,retmsg,**kw)
+	if not kw.get('quiet'):
+		self.to_log('find program=%r paths=%r var=%r -> %r'%(filename,path_list,var,ret))
 	if not ret:
-		self.fatal(kw.get('errmsg','')or'Could not find the program %s'%','.join(filename))
-	if var:
+		self.fatal(kw.get('errmsg','')or'Could not find the program %r'%filename)
+	interpreter=kw.get('interpreter')
+	if interpreter is None:
+		if not Utils.check_exe(ret[0],env=environ):
+			self.fatal('Program %r is not executable'%ret)
 		self.env[var]=ret
+	else:
+		self.env[var]=self.env[interpreter]+ret
 	return ret
-def find_perl_program(self,filename,path_list=[],var=None,environ=None,exts=''):
+@conf
+def find_binary(self,filenames,exts,paths):
+	for f in filenames:
+		for ext in exts:
+			exe_name=f+ext
+			if os.path.isabs(exe_name):
+				if os.path.isfile(exe_name):
+					return exe_name
+			else:
+				for path in paths:
+					x=os.path.expanduser(os.path.join(path,exe_name))
+					if os.path.isfile(x):
+						return x
+	return None
+@conf
+def run_build(self,*k,**kw):
+	lst=[str(v)for(p,v)in kw.items()if p!='env']
+	h=Utils.h_list(lst)
+	dir=self.bldnode.abspath()+os.sep+(not Utils.is_win32 and'.'or'')+'conf_check_'+Utils.to_hex(h)
+	try:
+		os.makedirs(dir)
+	except OSError:
+		pass
+	try:
+		os.stat(dir)
+	except OSError:
+		self.fatal('cannot use the configuration test folder %r'%dir)
+	cachemode=getattr(Options.options,'confcache',None)
+	if cachemode==1:
+		try:
+			proj=ConfigSet.ConfigSet(os.path.join(dir,'cache_run_build'))
+		except EnvironmentError:
+			pass
+		else:
+			ret=proj['cache_run_build']
+			if isinstance(ret,str)and ret.startswith('Test does not build'):
+				self.fatal(ret)
+			return ret
+	bdir=os.path.join(dir,'testbuild')
+	if not os.path.exists(bdir):
+		os.makedirs(bdir)
+	self.test_bld=bld=Context.create_context('build',top_dir=dir,out_dir=bdir)
+	bld.init_dirs()
+	bld.progress_bar=0
+	bld.targets='*'
+	bld.logger=self.logger
+	bld.all_envs.update(self.all_envs)
+	bld.env=kw['env']
+	bld.kw=kw
+	bld.conf=self
+	kw['build_fun'](bld)
+	ret=-1
 	try:
-		app=self.find_program(filename,path_list=path_list,var=var,environ=environ,exts=exts)
-	except:
-		self.find_program('perl',var='PERL')
-		app=self.find_file(filename,os.environ['PATH'].split(os.pathsep))
-		if not app:
+		try:
+			bld.compile()
+		except Errors.WafError:
+			ret='Test does not build: %s'%Utils.ex_stack()
+			self.fatal(ret)
+		else:
+			ret=getattr(bld,'retval',0)
+	finally:
+		if cachemode==1:
+			proj=ConfigSet.ConfigSet()
+			proj['cache_run_build']=ret
+			proj.store(os.path.join(dir,'cache_run_build'))
+		else:
+			shutil.rmtree(dir)
+	return ret
+@conf
+def ret_msg(self,msg,args):
+	if isinstance(msg,str):
+		return msg
+	return msg(args)
+@conf
+def test(self,*k,**kw):
+	if not'env'in kw:
+		kw['env']=self.env.derive()
+	if kw.get('validate'):
+		kw['validate'](kw)
+	self.start_msg(kw['msg'],**kw)
+	ret=None
+	try:
+		ret=self.run_build(*k,**kw)
+	except self.errors.ConfigurationError:
+		self.end_msg(kw['errmsg'],'YELLOW',**kw)
+		if Logs.verbose>1:
 			raise
-		if var:
-			self.env[var]=Utils.to_list(self.env['PERL'])+[app]
-	self.msg('Checking for %r'%filename,app)
-
-conf(add_os_flags)
-conf(cmd_to_list)
-conf(check_waf_version)
-conf(find_file)
-conf(find_program)
-conf(find_perl_program)
\ No newline at end of file
+		else:
+			self.fatal('The configuration failed')
+	else:
+		kw['success']=ret
+	if kw.get('post_check'):
+		ret=kw['post_check'](kw)
+	if ret:
+		self.end_msg(kw['errmsg'],'YELLOW',**kw)
+		self.fatal('The configuration failed %r'%ret)
+	else:
+		self.end_msg(self.ret_msg(kw['okmsg'],kw),**kw)
+	return ret
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Context.py glmark2-2021.02/waflib/Context.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Context.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Context.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,15 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,imp,sys
+import os,re,imp,sys
 from waflib import Utils,Errors,Logs
 import waflib.Node
-HEXVERSION=0x1060b00
-WAFVERSION="1.6.11"
-WAFREVISION="a7e69d6b81b04729804754c4d5214da063779a65"
-ABI=98
-DBFILE='.wafpickle-%d'%ABI
+HEXVERSION=0x1090300
+WAFVERSION="1.9.3"
+WAFREVISION="d31398c4a3e8b8f055821d3db6cf6d2ef1b09a40"
+ABI=99
+DBFILE='.wafpickle-%s-%d-%d'%(sys.platform,sys.hexversion,ABI)
 APPNAME='APPNAME'
 VERSION='VERSION'
 TOP='top'
@@ -20,9 +20,6 @@
 top_dir=''
 out_dir=''
 waf_dir=''
-local_repo=''
-remote_repo='http://waf.googlecode.com/git/'
-remote_locs=['waflib/extras','waflib/Tools']
 g_module=None
 STDOUT=1
 STDERR=-1
@@ -40,7 +37,7 @@
 	def __init__(cls,name,bases,dict):
 		super(store_context,cls).__init__(name,bases,dict)
 		name=cls.__name__
-		if name=='ctx'or name=='Context':
+		if name in('ctx','Context'):
 			return
 		try:
 			cls.cmd
@@ -60,11 +57,8 @@
 		except KeyError:
 			global run_dir
 			rd=run_dir
-		class node_class(waflib.Node.Node):
-			pass
-		self.node_class=node_class
-		self.node_class.__module__="waflib.Node"
-		self.node_class.__name__="Nod3"
+		self.node_class=type('Nod3',(waflib.Node.Node,),{})
+		self.node_class.__module__='waflib.Node'
 		self.node_class.ctx=self
 		self.root=self.node_class('',None)
 		self.cur_script=None
@@ -72,13 +66,20 @@
 		self.stack_path=[]
 		self.exec_dict={'ctx':self,'conf':self,'bld':self,'opt':self}
 		self.logger=None
-	def __hash__(self):
-		return id(self)
+	def finalize(self):
+		try:
+			logger=self.logger
+		except AttributeError:
+			pass
+		else:
+			Logs.free_logger(logger)
+			delattr(self,'logger')
 	def load(self,tool_list,*k,**kw):
 		tools=Utils.to_list(tool_list)
 		path=Utils.to_list(kw.get('tooldir',''))
+		with_sys_path=kw.get('with_sys_path',True)
 		for t in tools:
-			module=load_tool(t,path)
+			module=load_tool(t,path,with_sys_path=with_sys_path)
 			fun=getattr(module,kw.get('name',self.fun),None)
 			if fun:
 				fun(self)
@@ -93,10 +94,10 @@
 		self.cur_script=self.stack_path.pop()
 		if self.cur_script:
 			self.path=self.cur_script.parent
-	def recurse(self,dirs,name=None,mandatory=True,once=True):
+	def recurse(self,dirs,name=None,mandatory=True,once=True,encoding=None):
 		try:
 			cache=self.recurse_cache
-		except:
+		except AttributeError:
 			cache=self.recurse_cache={}
 		for d in Utils.to_list(dirs):
 			if not os.path.isabs(d):
@@ -108,7 +109,7 @@
 				cache[node]=True
 				self.pre_recurse(node)
 				try:
-					function_code=node.read('rU')
+					function_code=node.read('rU',encoding)
 					exec(compile(function_code,node.abspath(),'exec'),self.exec_dict)
 				finally:
 					self.post_recurse(node)
@@ -119,7 +120,7 @@
 					cache[tup]=True
 					self.pre_recurse(node)
 					try:
-						wscript_module=load_module(node.abspath())
+						wscript_module=load_module(node.abspath(),encoding=encoding)
 						user_function=getattr(wscript_module,(name or self.fun),None)
 						if not user_function:
 							if not mandatory:
@@ -131,32 +132,60 @@
 				elif not node:
 					if not mandatory:
 						continue
+					try:
+						os.listdir(d)
+					except OSError:
+						raise Errors.WafError('Cannot read the folder %r'%d)
 					raise Errors.WafError('No wscript file in directory %s'%d)
 	def exec_command(self,cmd,**kw):
 		subprocess=Utils.subprocess
 		kw['shell']=isinstance(cmd,str)
-		Logs.debug('runner: %r'%cmd)
-		Logs.debug('runner_env: kw=%s'%kw)
+		Logs.debug('runner: %r',cmd)
+		Logs.debug('runner_env: kw=%s',kw)
+		if self.logger:
+			self.logger.info(cmd)
+		if'stdout'not in kw:
+			kw['stdout']=subprocess.PIPE
+		if'stderr'not in kw:
+			kw['stderr']=subprocess.PIPE
+		if Logs.verbose and not kw['shell']and not Utils.check_exe(cmd[0]):
+			raise Errors.WafError('Program %s not found!'%cmd[0])
+		wargs={}
+		if'timeout'in kw:
+			if kw['timeout']is not None:
+				wargs['timeout']=kw['timeout']
+			del kw['timeout']
+		if'input'in kw:
+			if kw['input']:
+				wargs['input']=kw['input']
+				kw['stdin']=subprocess.PIPE
+			del kw['input']
+		if'cwd'in kw:
+			if not isinstance(kw['cwd'],str):
+				kw['cwd']=kw['cwd'].abspath()
 		try:
+			ret,out,err=Utils.run_process(cmd,kw,wargs)
+		except Exception as e:
+			raise Errors.WafError('Execution failure: %s'%str(e),ex=e)
+		if out:
+			if not isinstance(out,str):
+				out=out.decode(sys.stdout.encoding or'iso8859-1')
 			if self.logger:
-				self.logger.info(cmd)
-				kw['stdout']=kw['stderr']=subprocess.PIPE
-				p=subprocess.Popen(cmd,**kw)
-				(out,err)=p.communicate()
-				if out:
-					self.logger.debug('out: %s'%out.decode(sys.stdout.encoding or'iso8859-1'))
-				if err:
-					self.logger.error('err: %s'%err.decode(sys.stdout.encoding or'iso8859-1'))
-				return p.returncode
+				self.logger.debug('out: %s',out)
 			else:
-				p=subprocess.Popen(cmd,**kw)
-				return p.wait()
-		except OSError:
-			return-1
+				Logs.info(out,extra={'stream':sys.stdout,'c1':''})
+		if err:
+			if not isinstance(err,str):
+				err=err.decode(sys.stdout.encoding or'iso8859-1')
+			if self.logger:
+				self.logger.error('err: %s'%err)
+			else:
+				Logs.info(err,extra={'stream':sys.stderr,'c1':''})
+		return ret
 	def cmd_and_log(self,cmd,**kw):
 		subprocess=Utils.subprocess
 		kw['shell']=isinstance(cmd,str)
-		Logs.debug('runner: %r'%cmd)
+		Logs.debug('runner: %r',cmd)
 		if'quiet'in kw:
 			quiet=kw['quiet']
 			del kw['quiet']
@@ -167,13 +196,27 @@
 			del kw['output']
 		else:
 			to_ret=STDOUT
+		if Logs.verbose and not kw['shell']and not Utils.check_exe(cmd[0]):
+			raise Errors.WafError('Program %r not found!'%cmd[0])
 		kw['stdout']=kw['stderr']=subprocess.PIPE
 		if quiet is None:
 			self.to_log(cmd)
+		wargs={}
+		if'timeout'in kw:
+			if kw['timeout']is not None:
+				wargs['timeout']=kw['timeout']
+			del kw['timeout']
+		if'input'in kw:
+			if kw['input']:
+				wargs['input']=kw['input']
+				kw['stdin']=subprocess.PIPE
+			del kw['input']
+		if'cwd'in kw:
+			if not isinstance(kw['cwd'],str):
+				kw['cwd']=kw['cwd'].abspath()
 		try:
-			p=subprocess.Popen(cmd,**kw)
-			(out,err)=p.communicate()
-		except Exception ,e:
+			ret,out,err=Utils.run_process(cmd,kw,wargs)
+		except Exception as e:
 			raise Errors.WafError('Execution failure: %s'%str(e),ex=e)
 		if not isinstance(out,str):
 			out=out.decode(sys.stdout.encoding or'iso8859-1')
@@ -183,9 +226,9 @@
 			self.to_log('out: %s'%out)
 		if err and quiet!=STDERR and quiet!=BOTH:
 			self.to_log('err: %s'%err)
-		if p.returncode:
-			e=Errors.WafError('Command %r returned %r'%(cmd,p.returncode))
-			e.returncode=p.returncode
+		if ret:
+			e=Errors.WafError('Command %r returned %r'%(cmd,ret))
+			e.returncode=ret
 			e.stderr=err
 			e.stdout=out
 			raise e
@@ -199,7 +242,7 @@
 			self.logger.info('from %s: %s'%(self.path.abspath(),msg))
 		try:
 			msg='%s\n(complete log in %s)'%(msg,self.logger.handlers[0].baseFilename)
-		except:
+		except AttributeError:
 			pass
 		raise self.errors.ConfigurationError(msg,ex=ex)
 	def to_log(self,msg):
@@ -210,17 +253,29 @@
 		else:
 			sys.stderr.write(str(msg))
 			sys.stderr.flush()
-	def msg(self,msg,result,color=None):
-		self.start_msg(msg)
+	def msg(self,*k,**kw):
+		try:
+			msg=kw['msg']
+		except KeyError:
+			msg=k[0]
+		self.start_msg(msg,**kw)
+		try:
+			result=kw['result']
+		except KeyError:
+			result=k[1]
+		color=kw.get('color')
 		if not isinstance(color,str):
 			color=result and'GREEN'or'YELLOW'
-		self.end_msg(result,color)
-	def start_msg(self,msg):
+		self.end_msg(result,color,**kw)
+	def start_msg(self,*k,**kw):
+		if kw.get('quiet'):
+			return
+		msg=kw.get('msg')or k[0]
 		try:
 			if self.in_msg:
 				self.in_msg+=1
 				return
-		except:
+		except AttributeError:
 			self.in_msg=0
 		self.in_msg+=1
 		try:
@@ -230,10 +285,13 @@
 		for x in(self.line_just*'-',msg):
 			self.to_log(x)
 		Logs.pprint('NORMAL',"%s :"%msg.ljust(self.line_just),sep='')
-	def end_msg(self,result,color=None):
+	def end_msg(self,*k,**kw):
+		if kw.get('quiet'):
+			return
 		self.in_msg-=1
 		if self.in_msg:
 			return
+		result=kw.get('result')or k[0]
 		defcolor='GREEN'
 		if result==True:
 			msg='ok'
@@ -243,57 +301,92 @@
 		else:
 			msg=str(result)
 		self.to_log(msg)
-		Logs.pprint(color or defcolor,msg)
+		try:
+			color=kw['color']
+		except KeyError:
+			if len(k)>1 and k[1]in Logs.colors_lst:
+				color=k[1]
+			else:
+				color=defcolor
+		Logs.pprint(color,msg)
 	def load_special_tools(self,var,ban=[]):
 		global waf_dir
-		lst=self.root.find_node(waf_dir).find_node('waflib/extras').ant_glob(var)
-		for x in lst:
-			if not x.name in ban:
-				load_tool(x.name.replace('.py',''))
+		if os.path.isdir(waf_dir):
+			lst=self.root.find_node(waf_dir).find_node('waflib/extras').ant_glob(var)
+			for x in lst:
+				if not x.name in ban:
+					load_tool(x.name.replace('.py',''))
+		else:
+			from zipfile import PyZipFile
+			waflibs=PyZipFile(waf_dir)
+			lst=waflibs.namelist()
+			for x in lst:
+				if not re.match('waflib/extras/%s'%var.replace('*','.*'),var):
+					continue
+				f=os.path.basename(x)
+				doban=False
+				for b in ban:
+					r=b.replace('*','.*')
+					if re.match(r,f):
+						doban=True
+				if not doban:
+					f=f.replace('.py','')
+					load_tool(f)
 cache_modules={}
-def load_module(path):
+def load_module(path,encoding=None):
 	try:
 		return cache_modules[path]
 	except KeyError:
 		pass
 	module=imp.new_module(WSCRIPT_FILE)
 	try:
-		code=Utils.readf(path,m='rU')
-	except(IOError,OSError):
+		code=Utils.readf(path,m='rU',encoding=encoding)
+	except EnvironmentError:
 		raise Errors.WafError('Could not read the file %r'%path)
 	module_dir=os.path.dirname(path)
 	sys.path.insert(0,module_dir)
-	exec(compile(code,path,'exec'),module.__dict__)
-	sys.path.remove(module_dir)
+	try:
+		exec(compile(code,path,'exec'),module.__dict__)
+	finally:
+		sys.path.remove(module_dir)
 	cache_modules[path]=module
 	return module
-def load_tool(tool,tooldir=None):
-	tool=tool.replace('++','xx')
-	tool=tool.replace('java','javaw')
-	tool=tool.replace('compiler_cc','compiler_c')
-	if tooldir:
-		assert isinstance(tooldir,list)
-		sys.path=tooldir+sys.path
-		try:
-			__import__(tool)
+def load_tool(tool,tooldir=None,ctx=None,with_sys_path=True):
+	if tool=='java':
+		tool='javaw'
+	else:
+		tool=tool.replace('++','xx')
+	if not with_sys_path:
+		back_path=sys.path
+		sys.path=[]
+	try:
+		if tooldir:
+			assert isinstance(tooldir,list)
+			sys.path=tooldir+sys.path
+			try:
+				__import__(tool)
+			finally:
+				for d in tooldir:
+					sys.path.remove(d)
 			ret=sys.modules[tool]
 			Context.tools[tool]=ret
 			return ret
-		finally:
-			for d in tooldir:
-				sys.path.remove(d)
-	else:
-		global waf_dir
-		try:
-			os.stat(os.path.join(waf_dir,'waflib','extras',tool+'.py'))
-			d='waflib.extras.%s'%tool
-		except:
+		else:
+			if not with_sys_path:sys.path.insert(0,waf_dir)
 			try:
-				os.stat(os.path.join(waf_dir,'waflib','Tools',tool+'.py'))
-				d='waflib.Tools.%s'%tool
-			except:
-				d=tool
-		__import__(d)
-		ret=sys.modules[d]
-		Context.tools[tool]=ret
-		return ret
+				for x in('waflib.Tools.%s','waflib.extras.%s','waflib.%s','%s'):
+					try:
+						__import__(x%tool)
+						break
+					except ImportError:
+						x=None
+				else:
+					__import__(tool)
+			finally:
+				if not with_sys_path:sys.path.remove(waf_dir)
+			ret=sys.modules[x%tool]
+			Context.tools[tool]=ret
+			return ret
+	finally:
+		if not with_sys_path:
+			sys.path+=back_path
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Errors.py glmark2-2021.02/waflib/Errors.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Errors.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Errors.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import traceback,sys
 class WafError(Exception):
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/extras/compat15.py glmark2-2021.02/waflib/extras/compat15.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/extras/compat15.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/extras/compat15.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,8 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import sys
 from waflib import ConfigSet,Logs,Options,Scripting,Task,Build,Configure,Node,Runner,TaskGen,Utils,Errors,Context
 sys.modules['Environment']=ConfigSet
 ConfigSet.Environment=ConfigSet.ConfigSet
@@ -18,18 +16,50 @@
 sys.modules['Runner']=Runner
 sys.modules['TaskGen']=TaskGen
 sys.modules['Utils']=Utils
+sys.modules['Constants']=Context
+Context.SRCDIR=''
+Context.BLDDIR=''
 from waflib.Tools import c_preproc
 sys.modules['preproc']=c_preproc
 from waflib.Tools import c_config
 sys.modules['config_c']=c_config
 ConfigSet.ConfigSet.copy=ConfigSet.ConfigSet.derive
 ConfigSet.ConfigSet.set_variant=Utils.nada
+Utils.pproc=Utils.subprocess
 Build.BuildContext.add_subdirs=Build.BuildContext.recurse
 Build.BuildContext.new_task_gen=Build.BuildContext.__call__
 Build.BuildContext.is_install=0
 Node.Node.relpath_gen=Node.Node.path_from
+Utils.pproc=Utils.subprocess
+Utils.get_term_cols=Logs.get_term_cols
+def cmd_output(cmd,**kw):
+	silent=False
+	if'silent'in kw:
+		silent=kw['silent']
+		del(kw['silent'])
+	if'e'in kw:
+		tmp=kw['e']
+		del(kw['e'])
+		kw['env']=tmp
+	kw['shell']=isinstance(cmd,str)
+	kw['stdout']=Utils.subprocess.PIPE
+	if silent:
+		kw['stderr']=Utils.subprocess.PIPE
+	try:
+		p=Utils.subprocess.Popen(cmd,**kw)
+		output=p.communicate()[0]
+	except OSError as e:
+		raise ValueError(str(e))
+	if p.returncode:
+		if not silent:
+			msg="command execution failed: %s -> %r"%(cmd,str(output))
+			raise ValueError(msg)
+		output=''
+	return output
+Utils.cmd_output=cmd_output
 def name_to_obj(self,s,env=None):
-	Logs.warn('compat: change "name_to_obj(name, env)" by "get_tgen_by_name(name)"')
+	if Logs.verbose:
+		Logs.warn('compat: change "name_to_obj(name, env)" by "get_tgen_by_name(name)"')
 	return self.get_tgen_by_name(s)
 Build.BuildContext.name_to_obj=name_to_obj
 def env_of_name(self,name):
@@ -51,13 +81,15 @@
 		self.prepare_env(env)
 		self.all_envs[name]=env
 	else:
-		if fromenv:Logs.warn("The environment %s may have been configured already"%name)
+		if fromenv:
+			Logs.warn('The environment %s may have been configured already',name)
 	return env
 Configure.ConfigurationContext.retrieve=retrieve
 Configure.ConfigurationContext.sub_config=Configure.ConfigurationContext.recurse
 Configure.ConfigurationContext.check_tool=Configure.ConfigurationContext.load
 Configure.conftest=Configure.conf
 Configure.ConfigurationError=Errors.ConfigurationError
+Utils.WafError=Errors.WafError
 Options.OptionsContext.sub_options=Options.OptionsContext.recurse
 Options.OptionsContext.tool_options=Context.Context.load
 Options.Handler=Options.OptionsContext
@@ -68,6 +100,8 @@
 		key=key[1:]
 	self.table[key]=value
 ConfigSet.ConfigSet.__setitem__=setitem
+@TaskGen.feature('d')
+@TaskGen.before('apply_incpaths')
 def old_importpaths(self):
 	if getattr(self,'importpaths',[]):
 		self.includes=self.importpaths
@@ -76,40 +110,61 @@
 def load_tool(*k,**kw):
 	ret=eld(*k,**kw)
 	if'set_options'in ret.__dict__:
-		Logs.warn('compat: rename "set_options" to options')
+		if Logs.verbose:
+			Logs.warn('compat: rename "set_options" to options')
 		ret.options=ret.set_options
 	if'detect'in ret.__dict__:
-		Logs.warn('compat: rename "detect" to "configure"')
+		if Logs.verbose:
+			Logs.warn('compat: rename "detect" to "configure"')
 		ret.configure=ret.detect
 	return ret
 Context.load_tool=load_tool
+def get_curdir(self):
+	return self.path.abspath()
+Context.Context.curdir=property(get_curdir,Utils.nada)
+def get_srcdir(self):
+	return self.srcnode.abspath()
+Configure.ConfigurationContext.srcdir=property(get_srcdir,Utils.nada)
+def get_blddir(self):
+	return self.bldnode.abspath()
+Configure.ConfigurationContext.blddir=property(get_blddir,Utils.nada)
+Configure.ConfigurationContext.check_message_1=Configure.ConfigurationContext.start_msg
+Configure.ConfigurationContext.check_message_2=Configure.ConfigurationContext.end_msg
 rev=Context.load_module
-def load_module(path):
-	ret=rev(path)
+def load_module(path,encoding=None):
+	ret=rev(path,encoding)
 	if'set_options'in ret.__dict__:
-		Logs.warn('compat: rename "set_options" to "options" (%r)'%path)
+		if Logs.verbose:
+			Logs.warn('compat: rename "set_options" to "options" (%r)',path)
 		ret.options=ret.set_options
 	if'srcdir'in ret.__dict__:
-		Logs.warn('compat: rename "srcdir" to "top" (%r)'%path)
+		if Logs.verbose:
+			Logs.warn('compat: rename "srcdir" to "top" (%r)',path)
 		ret.top=ret.srcdir
 	if'blddir'in ret.__dict__:
-		Logs.warn('compat: rename "blddir" to "out" (%r)'%path)
+		if Logs.verbose:
+			Logs.warn('compat: rename "blddir" to "out" (%r)',path)
 		ret.out=ret.blddir
+	Utils.g_module=Context.g_module
+	Options.launch_dir=Context.launch_dir
 	return ret
 Context.load_module=load_module
 old_post=TaskGen.task_gen.post
 def post(self):
 	self.features=self.to_list(self.features)
 	if'cc'in self.features:
-		Logs.warn('compat: the feature cc does not exist anymore (use "c")')
+		if Logs.verbose:
+			Logs.warn('compat: the feature cc does not exist anymore (use "c")')
 		self.features.remove('cc')
 		self.features.append('c')
 	if'cstaticlib'in self.features:
-		Logs.warn('compat: the feature cstaticlib does not exist anymore (use "cstlib" or "cxxstlib")')
+		if Logs.verbose:
+			Logs.warn('compat: the feature cstaticlib does not exist anymore (use "cstlib" or "cxxstlib")')
 		self.features.remove('cstaticlib')
 		self.features.append(('cxx'in self.features)and'cxxstlib'or'cstlib')
 	if getattr(self,'ccflags',None):
-		Logs.warn('compat: "ccflags" was renamed to "cflags"')
+		if Logs.verbose:
+			Logs.warn('compat: "ccflags" was renamed to "cflags"')
 		self.cflags=self.ccflags
 	return old_post(self)
 TaskGen.task_gen.post=post
@@ -117,6 +172,9 @@
 	Logs.warn('wrong version (waf_version was removed in waf 1.6)')
 Utils.waf_version=waf_version
 import os
+@TaskGen.feature('c','cxx','d')
+@TaskGen.before('apply_incpaths','propagate_uselib_vars')
+@TaskGen.after('apply_link','process_source')
 def apply_uselib_local(self):
 	env=self.env
 	from waflib.Tools.ccroot import stlink_task
@@ -125,9 +183,11 @@
 	names=self.to_list(getattr(self,'uselib_local',[]))
 	get=self.bld.get_tgen_by_name
 	seen=set([])
+	seen_uselib=set([])
 	tmp=Utils.deque(names)
 	if tmp:
-		Logs.warn('compat: "uselib_local" is deprecated, replace by "use"')
+		if Logs.verbose:
+			Logs.warn('compat: "uselib_local" is deprecated, replace by "use"')
 	while tmp:
 		lib_name=tmp.popleft()
 		if lib_name in seen:
@@ -154,11 +214,15 @@
 			if not tmp_path in env['LIBPATH']:
 				env.prepend_value('LIBPATH',[tmp_path])
 		for v in self.to_list(getattr(y,'uselib',[])):
-			if not env['STLIB_'+v]:
-				if not v in self.uselib:
-					self.uselib.insert(0,v)
+			if v not in seen_uselib:
+				seen_uselib.add(v)
+				if not env['STLIB_'+v]:
+					if not v in self.uselib:
+						self.uselib.insert(0,v)
 		if getattr(y,'export_includes',None):
 			self.includes.extend(y.to_incnodes(y.export_includes))
+@TaskGen.feature('cprogram','cxxprogram','cstlib','cxxstlib','cshlib','cxxshlib','dprogram','dstlib','dshlib')
+@TaskGen.after('apply_link')
 def apply_objdeps(self):
 	names=getattr(self,'add_objects',[])
 	if not names:
@@ -185,39 +249,53 @@
 		seen.append(x)
 		for t in getattr(y,'compiled_tasks',[]):
 			self.link_task.inputs.extend(t.outputs)
+@TaskGen.after('apply_link')
 def process_obj_files(self):
 	if not hasattr(self,'obj_files'):
 		return
 	for x in self.obj_files:
 		node=self.path.find_resource(x)
 		self.link_task.inputs.append(node)
+@TaskGen.taskgen_method
 def add_obj_file(self,file):
 	if not hasattr(self,'obj_files'):self.obj_files=[]
 	if not'process_obj_files'in self.meths:self.meths.append('process_obj_files')
 	self.obj_files.append(file)
 old_define=Configure.ConfigurationContext.__dict__['define']
-def define(self,key,val,quote=True):
-	old_define(self,key,val,quote)
+@Configure.conf
+def define(self,key,val,quote=True,comment=''):
+	old_define(self,key,val,quote,comment)
 	if key.startswith('HAVE_'):
 		self.env[key]=1
 old_undefine=Configure.ConfigurationContext.__dict__['undefine']
-def undefine(self,key):
-	old_undefine(self,key)
+@Configure.conf
+def undefine(self,key,comment=''):
+	old_undefine(self,key,comment)
 	if key.startswith('HAVE_'):
 		self.env[key]=0
 def set_incdirs(self,val):
 	Logs.warn('compat: change "export_incdirs" by "export_includes"')
 	self.export_includes=val
 TaskGen.task_gen.export_incdirs=property(None,set_incdirs)
-
-TaskGen.feature('d')(old_importpaths)
-TaskGen.before('apply_incpaths')(old_importpaths)
-TaskGen.feature('c','cxx','d')(apply_uselib_local)
-TaskGen.before('apply_incpaths','propagate_uselib_vars')(apply_uselib_local)
-TaskGen.after('apply_link','process_source')(apply_uselib_local)
-TaskGen.feature('cprogram','cxxprogram','cstlib','cxxstlib','cshlib','cxxshlib','dprogram','dstlib','dshlib')(apply_objdeps)
-TaskGen.after('apply_link')(apply_objdeps)
-TaskGen.after('apply_link')(process_obj_files)
-TaskGen.taskgen_method(add_obj_file)
-Configure.conf(define)
-Configure.conf(undefine)
\ No newline at end of file
+def install_dir(self,path):
+	if not path:
+		return[]
+	destpath=Utils.subst_vars(path,self.env)
+	if self.is_install>0:
+		Logs.info('* creating %s',destpath)
+		Utils.check_dir(destpath)
+	elif self.is_install<0:
+		Logs.info('* removing %s',destpath)
+		try:
+			os.remove(destpath)
+		except OSError:
+			pass
+Build.BuildContext.install_dir=install_dir
+repl={'apply_core':'process_source','apply_lib_vars':'process_source','apply_obj_vars':'propagate_uselib_vars','exec_rule':'process_rule'}
+def after(*k):
+	k=[repl.get(key,key)for key in k]
+	return TaskGen.after_method(*k)
+def before(*k):
+	k=[repl.get(key,key)for key in k]
+	return TaskGen.before_method(*k)
+TaskGen.before=before
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/extras/__init__.py glmark2-2021.02/waflib/extras/__init__.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/extras/__init__.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/extras/__init__.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,4 +1,4 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/fixpy2.py glmark2-2021.02/waflib/fixpy2.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/fixpy2.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/fixpy2.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os
 all_modifs={}
@@ -21,12 +21,16 @@
 		return
 	filename=os.path.join(dir,name)
 	f=open(filename,'r')
-	txt=f.read()
-	f.close()
+	try:
+		txt=f.read()
+	finally:
+		f.close()
 	txt=fun(txt)
 	f=open(filename,'w')
-	f.write(txt)
-	f.close()
+	try:
+		f.write(txt)
+	finally:
+		f.close()
 def subst(*k):
 	def do_subst(fun):
 		global all_modifs
@@ -37,14 +41,14 @@
 				all_modifs[x]=[fun]
 		return fun
 	return do_subst
+@subst('*')
 def r1(code):
-	code=code.replace(',e:',',e:')
-	code=code.replace("",'')
-	code=code.replace('','')
-	return code
+	code=code.replace('as e:',',e:')
+	code=code.replace(".decode(sys.stdout.encoding or 'iso8859-1')",'')
+	return code.replace('.encode()','')
+@subst('Runner.py')
 def r4(code):
-	code=code.replace('next(self.biter)','self.biter.next()')
-	return code
-
-subst('*')(r1)
-subst('Runner.py')(r4)
\ No newline at end of file
+	return code.replace('next(self.biter)','self.biter.next()')
+@subst('Context.py')
+def r5(code):
+	return code.replace("('Execution failure: %s'%str(e),ex=e)","('Execution failure: %s'%str(e),ex=e),None,sys.exc_info()[2]")
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/__init__.py glmark2-2021.02/waflib/__init__.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/__init__.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/__init__.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,4 +1,4 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Logs.py glmark2-2021.02/waflib/Logs.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Logs.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Logs.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,54 +1,53 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os,re,traceback,sys
-_nocolor=os.environ.get('NOCOLOR','no')not in('no','0','false')
-try:
-	if not _nocolor:
-		import waflib.ansiterm
-except:
-	pass
+from waflib import Utils,ansiterm
+if not os.environ.get('NOSYNC',False):
+	if sys.stdout.isatty()and id(sys.stdout)==id(sys.__stdout__):
+		sys.stdout=ansiterm.AnsiTerm(sys.stdout)
+	if sys.stderr.isatty()and id(sys.stderr)==id(sys.__stderr__):
+		sys.stderr=ansiterm.AnsiTerm(sys.stderr)
 import logging
-LOG_FORMAT="%(asctime)s %(c1)s%(zone)s%(c2)s %(message)s"
-HOUR_FORMAT="%H:%M:%S"
-zones=''
+LOG_FORMAT=os.environ.get('WAF_LOG_FORMAT','%(asctime)s %(c1)s%(zone)s%(c2)s %(message)s')
+HOUR_FORMAT=os.environ.get('WAF_HOUR_FORMAT','%H:%M:%S')
+zones=[]
 verbose=0
-colors_lst={'USE':True,'BOLD':'\x1b[01;1m','RED':'\x1b[01;31m','GREEN':'\x1b[32m','YELLOW':'\x1b[33m','PINK':'\x1b[35m','BLUE':'\x1b[01;34m','CYAN':'\x1b[36m','NORMAL':'\x1b[0m','cursor_on':'\x1b[?25h','cursor_off':'\x1b[?25l',}
-got_tty=not os.environ.get('TERM','dumb')in['dumb','emacs']
-if got_tty:
-	try:
-		got_tty=sys.stderr.isatty()
-	except AttributeError:
-		got_tty=False
-if(not got_tty and os.environ.get('TERM','dumb')!='msys')or _nocolor:
-	colors_lst['USE']=False
-def get_term_cols():
-	return 80
+colors_lst={'USE':True,'BOLD':'\x1b[01;1m','RED':'\x1b[01;31m','GREEN':'\x1b[32m','YELLOW':'\x1b[33m','PINK':'\x1b[35m','BLUE':'\x1b[01;34m','CYAN':'\x1b[36m','GREY':'\x1b[37m','NORMAL':'\x1b[0m','cursor_on':'\x1b[?25h','cursor_off':'\x1b[?25l',}
+indicator='\r\x1b[K%s%s%s'
 try:
-	import struct,fcntl,termios
-except ImportError:
-	pass
-else:
-	if got_tty:
-		def get_term_cols_real():
-			dummy_lines,cols=struct.unpack("HHHH",fcntl.ioctl(sys.stderr.fileno(),termios.TIOCGWINSZ,struct.pack("HHHH",0,0,0,0)))[:2]
-			return cols
-		try:
-			get_term_cols_real()
-		except:
-			pass
+	unicode
+except NameError:
+	unicode=None
+def enable_colors(use):
+	if use==1:
+		if not(sys.stderr.isatty()or sys.stdout.isatty()):
+			use=0
+		if Utils.is_win32 and os.name!='java':
+			term=os.environ.get('TERM','')
 		else:
-			get_term_cols=get_term_cols_real
+			term=os.environ.get('TERM','dumb')
+		if term in('dumb','emacs'):
+			use=0
+	if use>=1:
+		os.environ['TERM']='vt100'
+	colors_lst['USE']=use
+try:
+	get_term_cols=ansiterm.get_term_cols
+except AttributeError:
+	def get_term_cols():
+		return 80
 get_term_cols.__doc__="""
-	Get the console width in characters.
+	Returns the console width in characters.
 
 	:return: the number of characters per line
 	:rtype: int
 	"""
 def get_color(cl):
-	if not colors_lst['USE']:return''
-	return colors_lst.get(cl,'')
+	if colors_lst['USE']:
+		return colors_lst.get(cl,'')
+	return''
 class color_dict(object):
 	def __getattr__(self,a):
 		return get_color(a)
@@ -57,19 +56,12 @@
 colors=color_dict()
 re_log=re.compile(r'(\w+): (.*)',re.M)
 class log_filter(logging.Filter):
-	def __init__(self,name=None):
-		pass
+	def __init__(self,name=''):
+		logging.Filter.__init__(self,name)
 	def filter(self,rec):
-		rec.c1=colors.PINK
-		rec.c2=colors.NORMAL
+		global verbose
 		rec.zone=rec.module
 		if rec.levelno>=logging.INFO:
-			if rec.levelno>=logging.ERROR:
-				rec.c1=colors.RED
-			elif rec.levelno>=logging.WARNING:
-				rec.c1=colors.YELLOW
-			else:
-				rec.c1=colors.GREEN
 			return True
 		m=re_log.match(rec.msg)
 		if m:
@@ -80,26 +72,82 @@
 		elif not verbose>2:
 			return False
 		return True
+class log_handler(logging.StreamHandler):
+	def emit(self,record):
+		try:
+			try:
+				self.stream=record.stream
+			except AttributeError:
+				if record.levelno>=logging.WARNING:
+					record.stream=self.stream=sys.stderr
+				else:
+					record.stream=self.stream=sys.stdout
+			self.emit_override(record)
+			self.flush()
+		except(KeyboardInterrupt,SystemExit):
+			raise
+		except:
+			self.handleError(record)
+	def emit_override(self,record,**kw):
+		self.terminator=getattr(record,'terminator','\n')
+		stream=self.stream
+		if unicode:
+			msg=self.formatter.format(record)
+			fs='%s'+self.terminator
+			try:
+				if(isinstance(msg,unicode)and getattr(stream,'encoding',None)):
+					fs=fs.decode(stream.encoding)
+					try:
+						stream.write(fs%msg)
+					except UnicodeEncodeError:
+						stream.write((fs%msg).encode(stream.encoding))
+				else:
+					stream.write(fs%msg)
+			except UnicodeError:
+				stream.write((fs%msg).encode('utf-8'))
+		else:
+			logging.StreamHandler.emit(self,record)
 class formatter(logging.Formatter):
 	def __init__(self):
 		logging.Formatter.__init__(self,LOG_FORMAT,HOUR_FORMAT)
 	def format(self,rec):
-		if rec.levelno>=logging.WARNING or rec.levelno==logging.INFO:
-			try:
-				msg=rec.msg.decode('utf-8')
-			except:
-				msg=rec.msg
-			return'%s%s%s'%(rec.c1,msg,rec.c2)
+		try:
+			msg=rec.msg.decode('utf-8')
+		except Exception:
+			msg=rec.msg
+		use=colors_lst['USE']
+		if(use==1 and rec.stream.isatty())or use==2:
+			c1=getattr(rec,'c1',None)
+			if c1 is None:
+				c1=''
+				if rec.levelno>=logging.ERROR:
+					c1=colors.RED
+				elif rec.levelno>=logging.WARNING:
+					c1=colors.YELLOW
+				elif rec.levelno>=logging.INFO:
+					c1=colors.GREEN
+			c2=getattr(rec,'c2',colors.NORMAL)
+			msg='%s%s%s'%(c1,msg,c2)
+		else:
+			msg=re.sub(r'\r(?!\n)|\x1B\[(K|.*?(m|h|l))','',msg)
+		if rec.levelno>=logging.INFO:
+			if rec.args:
+				return msg%rec.args
+			return msg
+		rec.msg=msg
+		rec.c1=colors.PINK
+		rec.c2=colors.NORMAL
 		return logging.Formatter.format(self,rec)
 log=None
 def debug(*k,**kw):
+	global verbose
 	if verbose:
 		k=list(k)
 		k[0]=k[0].replace('\n',' ')
 		global log
 		log.debug(*k,**kw)
 def error(*k,**kw):
-	global log
+	global log,verbose
 	log.error(*k,**kw)
 	if verbose>2:
 		st=traceback.extract_stack()
@@ -107,10 +155,10 @@
 			st=st[:-1]
 			buf=[]
 			for filename,lineno,name,line in st:
-				buf.append('  File "%s", line %d, in %s'%(filename,lineno,name))
+				buf.append('  File %r, line %d, in %s'%(filename,lineno,name))
 				if line:
 					buf.append('	%s'%line.strip())
-			if buf:log.error("\n".join(buf))
+			if buf:log.error('\n'.join(buf))
 def warn(*k,**kw):
 	global log
 	log.warn(*k,**kw)
@@ -122,7 +170,7 @@
 	log=logging.getLogger('waflib')
 	log.handlers=[]
 	log.filters=[]
-	hdlr=logging.StreamHandler()
+	hdlr=log_handler()
 	hdlr.setFormatter(formatter())
 	log.addHandler(hdlr)
 	log.addFilter(log_filter())
@@ -135,7 +183,7 @@
 	logger.addHandler(hdlr)
 	logger.setLevel(logging.DEBUG)
 	return logger
-def make_mem_logger(name,to_log,size=10000):
+def make_mem_logger(name,to_log,size=8192):
 	from logging.handlers import MemoryHandler
 	logger=logging.getLogger(name)
 	hdlr=MemoryHandler(size,target=to_log)
@@ -145,5 +193,13 @@
 	logger.memhandler=hdlr
 	logger.setLevel(logging.DEBUG)
 	return logger
-def pprint(col,str,label='',sep='\n'):
-	sys.stderr.write("%s%s%s %s%s"%(colors(col),str,colors.NORMAL,label,sep))
+def free_logger(logger):
+	try:
+		for x in logger.handlers:
+			x.close()
+			logger.removeHandler(x)
+	except Exception:
+		pass
+def pprint(col,msg,label='',sep='\n'):
+	global info
+	info('%s%s%s %s',colors(col),msg,colors.NORMAL,label,extra={'terminator':sep})
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Node.py glmark2-2021.02/waflib/Node.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Node.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Node.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,7 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
 import os,re,sys,shutil
 from waflib import Utils,Errors
 exclude_regs='''
@@ -35,28 +33,11 @@
 **/{arch}
 **/_darcs
 **/_darcs/**
+**/.intlcache
 **/.DS_Store'''
-def split_path(path):
-	return path.split('/')
-def split_path_cygwin(path):
-	if path.startswith('//'):
-		ret=path.split('/')[2:]
-		ret[0]='/'+ret[0]
-		return ret
-	return path.split('/')
-re_sp=re.compile('[/\\\\]')
-def split_path_win32(path):
-	if path.startswith('\\\\'):
-		ret=re.split(re_sp,path)[2:]
-		ret[0]='\\'+ret[0]
-		return ret
-	return re.split(re_sp,path)
-if sys.platform=='cygwin':
-	split_path=split_path_cygwin
-elif Utils.is_win32:
-	split_path=split_path_win32
 class Node(object):
-	__slots__=('name','sig','children','parent','cache_abspath','cache_isdir')
+	dict_class=dict
+	__slots__=('name','parent','children','cache_abspath','cache_isdir')
 	def __init__(self,name,parent):
 		self.name=name
 		self.parent=parent
@@ -68,45 +49,71 @@
 		self.name=data[0]
 		self.parent=data[1]
 		if data[2]is not None:
-			self.children=data[2]
-		if data[3]is not None:
-			self.sig=data[3]
+			self.children=self.dict_class(data[2])
 	def __getstate__(self):
-		return(self.name,self.parent,getattr(self,'children',None),getattr(self,'sig',None))
+		return(self.name,self.parent,getattr(self,'children',None))
 	def __str__(self):
-		return self.name
+		return self.abspath()
 	def __repr__(self):
 		return self.abspath()
-	def __hash__(self):
-		return id(self)
-	def __eq__(self,node):
-		return id(self)==id(node)
 	def __copy__(self):
 		raise Errors.WafError('nodes are not supposed to be copied')
-	def read(self,flags='r'):
-		return Utils.readf(self.abspath(),flags)
-	def write(self,data,flags='w'):
-		f=None
-		try:
-			f=open(self.abspath(),flags)
-			f.write(data)
-		finally:
-			if f:
-				f.close()
+	def read(self,flags='r',encoding='ISO8859-1'):
+		return Utils.readf(self.abspath(),flags,encoding)
+	def write(self,data,flags='w',encoding='ISO8859-1'):
+		Utils.writef(self.abspath(),data,flags,encoding)
+	def read_json(self,convert=True,encoding='utf-8'):
+		import json
+		object_pairs_hook=None
+		if convert and sys.hexversion<0x3000000:
+			try:
+				_type=unicode
+			except NameError:
+				_type=str
+			def convert(value):
+				if isinstance(value,list):
+					return[convert(element)for element in value]
+				elif isinstance(value,_type):
+					return str(value)
+				else:
+					return value
+			def object_pairs(pairs):
+				return dict((str(pair[0]),convert(pair[1]))for pair in pairs)
+			object_pairs_hook=object_pairs
+		return json.loads(self.read(encoding=encoding),object_pairs_hook=object_pairs_hook)
+	def write_json(self,data,pretty=True):
+		import json
+		indent=2
+		separators=(',',': ')
+		sort_keys=pretty
+		newline=os.linesep
+		if not pretty:
+			indent=None
+			separators=(',',':')
+			newline=''
+		output=json.dumps(data,indent=indent,separators=separators,sort_keys=sort_keys)+newline
+		self.write(output,encoding='utf-8')
+	def exists(self):
+		return os.path.exists(self.abspath())
+	def isdir(self):
+		return os.path.isdir(self.abspath())
 	def chmod(self,val):
 		os.chmod(self.abspath(),val)
-	def delete(self):
-		try:
-			if getattr(self,'children',None):
-				shutil.rmtree(self.abspath())
-			else:
-				os.unlink(self.abspath())
-		except:
-			pass
+	def delete(self,evict=True):
 		try:
-			delattr(self,'children')
-		except:
-			pass
+			try:
+				if os.path.isdir(self.abspath()):
+					shutil.rmtree(self.abspath())
+				else:
+					os.remove(self.abspath())
+			except OSError as e:
+				if os.path.exists(self.abspath()):
+					raise e
+		finally:
+			if evict:
+				self.evict()
+	def evict(self):
+		del self.parent.children[self.name]
 	def suffix(self):
 		k=max(0,self.name.rfind('.'))
 		return self.name[k:]
@@ -122,86 +129,80 @@
 		lst.sort()
 		return lst
 	def mkdir(self):
-		if getattr(self,'cache_isdir',None):
+		if self.isdir():
 			return
 		try:
 			self.parent.mkdir()
-		except:
+		except OSError:
 			pass
 		if self.name:
 			try:
 				os.makedirs(self.abspath())
 			except OSError:
 				pass
-			if not os.path.isdir(self.abspath()):
-				raise Errors.WafError('Could not create the directory %s'%self.abspath())
+			if not self.isdir():
+				raise Errors.WafError('Could not create the directory %r'%self)
 			try:
 				self.children
-			except:
-				self.children={}
-		self.cache_isdir=True
+			except AttributeError:
+				self.children=self.dict_class()
 	def find_node(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
 		cur=self
 		for x in lst:
 			if x=='..':
 				cur=cur.parent or cur
 				continue
 			try:
-				if x in cur.children:
-					cur=cur.children[x]
+				ch=cur.children
+			except AttributeError:
+				cur.children=self.dict_class()
+			else:
+				try:
+					cur=ch[x]
 					continue
-			except:
-				cur.children={}
+				except KeyError:
+					pass
 			cur=self.__class__(x,cur)
-			try:
-				os.stat(cur.abspath())
-			except:
-				del cur.parent.children[x]
+			if not cur.exists():
+				cur.evict()
 				return None
-		ret=cur
-		try:
-			os.stat(ret.abspath())
-		except:
-			del ret.parent.children[ret.name]
+		if not cur.exists():
+			cur.evict()
 			return None
-		try:
-			while not getattr(cur.parent,'cache_isdir',None):
-				cur=cur.parent
-				cur.cache_isdir=True
-		except AttributeError:
-			pass
-		return ret
+		return cur
 	def make_node(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
 		cur=self
 		for x in lst:
 			if x=='..':
 				cur=cur.parent or cur
 				continue
-			if getattr(cur,'children',{}):
-				if x in cur.children:
-					cur=cur.children[x]
-					continue
+			try:
+				cur=cur.children[x]
+			except AttributeError:
+				cur.children=self.dict_class()
+			except KeyError:
+				pass
 			else:
-				cur.children={}
+				continue
 			cur=self.__class__(x,cur)
 		return cur
-	def search(self,lst):
+	def search_node(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
 		cur=self
-		try:
-			for x in lst:
-				if x=='..':
-					cur=cur.parent or cur
-				else:
+		for x in lst:
+			if x=='..':
+				cur=cur.parent or cur
+			else:
+				try:
 					cur=cur.children[x]
-			return cur
-		except:
-			pass
+				except(AttributeError,KeyError):
+					return None
+		return cur
 	def path_from(self,node):
 		c1=self
 		c2=node
@@ -217,59 +218,70 @@
 			up+=1
 			c2=c2.parent
 			c2h-=1
-		while id(c1)!=id(c2):
+		while not c1 is c2:
 			lst.append(c1.name)
 			up+=1
 			c1=c1.parent
 			c2=c2.parent
-		for i in range(up):
-			lst.append('..')
+		if c1.parent:
+			for i in range(up):
+				lst.append('..')
+		else:
+			if lst and not Utils.is_win32:
+				lst.append('')
 		lst.reverse()
 		return os.sep.join(lst)or'.'
 	def abspath(self):
 		try:
 			return self.cache_abspath
-		except:
+		except AttributeError:
 			pass
-		if os.sep=='/':
-			if not self.parent:
-				val=os.sep
-			elif not self.parent.name:
-				val=os.sep+self.name
-			else:
-				val=self.parent.abspath()+os.sep+self.name
+		if not self.parent:
+			val=os.sep
+		elif not self.parent.name:
+			val=os.sep+self.name
 		else:
+			val=self.parent.abspath()+os.sep+self.name
+		self.cache_abspath=val
+		return val
+	if Utils.is_win32:
+		def abspath(self):
+			try:
+				return self.cache_abspath
+			except AttributeError:
+				pass
 			if not self.parent:
 				val=''
 			elif not self.parent.name:
 				val=self.name+os.sep
 			else:
 				val=self.parent.abspath().rstrip(os.sep)+os.sep+self.name
-		self.cache_abspath=val
-		return val
+			self.cache_abspath=val
+			return val
 	def is_child_of(self,node):
 		p=self
 		diff=self.height()-node.height()
 		while diff>0:
 			diff-=1
 			p=p.parent
-		return id(p)==id(node)
+		return p is node
 	def ant_iter(self,accept=None,maxdepth=25,pats=[],dir=False,src=True,remove=True):
 		dircont=self.listdir()
 		dircont.sort()
 		try:
 			lst=set(self.children.keys())
+		except AttributeError:
+			self.children=self.dict_class()
+		else:
 			if remove:
 				for x in lst-set(dircont):
-					del self.children[x]
-		except:
-			self.children={}
+					self.children[x].evict()
 		for name in dircont:
 			npats=accept(name,pats)
 			if npats and npats[0]:
 				accepted=[]in npats[0]
 				node=self.make_node([name])
-				isdir=os.path.isdir(node.abspath())
+				isdir=node.isdir()
 				if accepted:
 					if isdir:
 						if dir:
@@ -277,17 +289,17 @@
 					else:
 						if src:
 							yield node
-				if getattr(node,'cache_isdir',None)or isdir:
+				if isdir:
 					node.cache_isdir=True
 					if maxdepth:
 						for k in node.ant_iter(accept=accept,maxdepth=maxdepth-1,pats=npats,dir=dir,src=src,remove=remove):
 							yield k
-		raise StopIteration
 	def ant_glob(self,*k,**kw):
 		src=kw.get('src',True)
 		dir=kw.get('dir',False)
 		excl=kw.get('excl',exclude_regs)
 		incl=k and k[0]or kw.get('incl','**')
+		reflags=kw.get('ignorecase',0)and re.I
 		def to_pat(s):
 			lst=Utils.to_list(s)
 			ret=[]
@@ -304,9 +316,9 @@
 						k=k.replace('.','[.]').replace('*','.*').replace('?','.').replace('+','\\+')
 						k='^%s$'%k
 						try:
-							accu.append(re.compile(k))
-						except Exception ,e:
-							raise Errors.WafError("Invalid pattern: %s"%k,e)
+							accu.append(re.compile(k,flags=reflags))
+						except Exception as e:
+							raise Errors.WafError('Invalid pattern: %s'%k,e)
 				ret.append(accu)
 			return ret
 		def filtre(name,nn):
@@ -330,77 +342,52 @@
 			if[]in nrej:
 				nacc=[]
 			return[nacc,nrej]
-		ret=[x for x in self.ant_iter(accept=accept,pats=[to_pat(incl),to_pat(excl)],maxdepth=25,dir=dir,src=src,remove=kw.get('remove',True))]
+		ret=[x for x in self.ant_iter(accept=accept,pats=[to_pat(incl),to_pat(excl)],maxdepth=kw.get('maxdepth',25),dir=dir,src=src,remove=kw.get('remove',True))]
 		if kw.get('flat',False):
 			return' '.join([x.path_from(self)for x in ret])
 		return ret
-	def find_nodes(self,find_dirs=True,find_files=True,match_fun=lambda x:True):
-		x="""
-		Recursively finds nodes::
-
-			def configure(cnf):
-				cnf.find_nodes()
-
-		:param find_dirs: whether to return directories
-		:param find_files: whether to return files
-		:param match_fun: matching function, taking a node as parameter
-		:rtype generator
-		:return: a generator that iterates over all the requested files
-		"""
-		files=self.listdir()
-		for f in files:
-			node=self.make_node([f])
-			if os.path.isdir(node.abspath()):
-				if find_dirs and match_fun(node):
-					yield node
-				gen=node.find_nodes(find_dirs,find_files,match_fun)
-				for g in gen:
-					yield g
-			else:
-				if find_files and match_fun(node):
-					yield node
 	def is_src(self):
 		cur=self
-		x=id(self.ctx.srcnode)
-		y=id(self.ctx.bldnode)
+		x=self.ctx.srcnode
+		y=self.ctx.bldnode
 		while cur.parent:
-			if id(cur)==y:
+			if cur is y:
 				return False
-			if id(cur)==x:
+			if cur is x:
 				return True
 			cur=cur.parent
 		return False
 	def is_bld(self):
 		cur=self
-		y=id(self.ctx.bldnode)
+		y=self.ctx.bldnode
 		while cur.parent:
-			if id(cur)==y:
+			if cur is y:
 				return True
 			cur=cur.parent
 		return False
 	def get_src(self):
 		cur=self
-		x=id(self.ctx.srcnode)
-		y=id(self.ctx.bldnode)
+		x=self.ctx.srcnode
+		y=self.ctx.bldnode
 		lst=[]
 		while cur.parent:
-			if id(cur)==y:
+			if cur is y:
 				lst.reverse()
-				return self.ctx.srcnode.make_node(lst)
-			if id(cur)==x:
+				return x.make_node(lst)
+			if cur is x:
 				return self
 			lst.append(cur.name)
 			cur=cur.parent
 		return self
 	def get_bld(self):
 		cur=self
-		x=id(self.ctx.srcnode)
-		y=id(self.ctx.bldnode)
+		x=self.ctx.srcnode
+		y=self.ctx.bldnode
 		lst=[]
 		while cur.parent:
-			if id(cur)==y:
+			if cur is y:
 				return self
-			if id(cur)==x:
+			if cur is x:
 				lst.reverse()
 				return self.ctx.bldnode.make_node(lst)
 			lst.append(cur.name)
@@ -411,51 +398,33 @@
 		return self.ctx.bldnode.make_node(['__root__']+lst)
 	def find_resource(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
-		node=self.get_bld().search(lst)
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
+		node=self.get_bld().search_node(lst)
 		if not node:
-			self=self.get_src()
-			node=self.find_node(lst)
-		try:
-			pat=node.abspath()
-			if os.path.isdir(pat):
-				return None
-		except:
-			pass
+			node=self.get_src().find_node(lst)
+		if node and node.isdir():
+			return None
 		return node
 	def find_or_declare(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
-		node=self.get_bld().search(lst)
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
+		node=self.get_bld().search_node(lst)
 		if node:
 			if not os.path.isfile(node.abspath()):
-				node.sig=None
-				try:
-					node.parent.mkdir()
-				except:
-					pass
+				node.parent.mkdir()
 			return node
 		self=self.get_src()
 		node=self.find_node(lst)
 		if node:
-			if not os.path.isfile(node.abspath()):
-				node.sig=None
-				try:
-					node.parent.mkdir()
-				except:
-					pass
 			return node
 		node=self.get_bld().make_node(lst)
 		node.parent.mkdir()
 		return node
 	def find_dir(self,lst):
 		if isinstance(lst,str):
-			lst=[x for x in split_path(lst)if x and x!='.']
+			lst=[x for x in Utils.split_path(lst)if x and x!='.']
 		node=self.find_node(lst)
-		try:
-			if not os.path.isdir(node.abspath()):
-				return None
-		except(OSError,AttributeError):
+		if node and not node.isdir():
 			return None
 		return node
 	def change_ext(self,ext,ext_in=None):
@@ -469,38 +438,49 @@
 		else:
 			name=name[:-len(ext_in)]+ext
 		return self.parent.find_or_declare([name])
-	def nice_path(self,env=None):
-		return self.path_from(self.ctx.launch_node())
 	def bldpath(self):
 		return self.path_from(self.ctx.bldnode)
 	def srcpath(self):
 		return self.path_from(self.ctx.srcnode)
 	def relpath(self):
 		cur=self
-		x=id(self.ctx.bldnode)
+		x=self.ctx.bldnode
 		while cur.parent:
-			if id(cur)==x:
+			if cur is x:
 				return self.bldpath()
 			cur=cur.parent
 		return self.srcpath()
 	def bld_dir(self):
 		return self.parent.bldpath()
-	def bld_base(self):
-		s=os.path.splitext(self.name)[0]
-		return self.bld_dir()+os.sep+s
+	def h_file(self):
+		return Utils.h_file(self.abspath())
 	def get_bld_sig(self):
 		try:
-			ret=self.ctx.hash_cache[id(self)]
-		except KeyError:
-			pass
+			cache=self.ctx.cache_sig
 		except AttributeError:
-			self.ctx.hash_cache={}
-		else:
-			return ret
-		if not self.is_bld()or self.ctx.bldnode is self.ctx.srcnode:
-			self.sig=Utils.h_file(self.abspath())
-		self.ctx.hash_cache[id(self)]=ret=self.sig
+			cache=self.ctx.cache_sig={}
+		try:
+			ret=cache[self]
+		except KeyError:
+			p=self.abspath()
+			try:
+				ret=cache[self]=self.h_file()
+			except EnvironmentError:
+				if self.isdir():
+					st=os.stat(p)
+					ret=cache[self]=Utils.h_list([p,st.st_ino,st.st_mode])
+					return ret
+				raise
 		return ret
+	def get_sig(self):
+		return self.h_file()
+	def set_sig(self,val):
+		try:
+			del self.get_bld_sig.__cache__[(self,)]
+		except(AttributeError,KeyError):
+			pass
+	sig=property(get_sig,set_sig)
+	cache_sig=property(get_sig,set_sig)
 pickle_lock=Utils.threading.Lock()
 class Nod3(Node):
 	pass
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Options.py glmark2-2021.02/waflib/Options.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Options.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Options.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,63 +1,30 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os,tempfile,optparse,sys,re
-from waflib import Logs,Utils,Context
-cmds='distclean configure build install clean uninstall check dist distcheck'.split()
+from waflib import Logs,Utils,Context,Errors
 options={}
 commands=[]
+envvars=[]
 lockfile=os.environ.get('WAFLOCK','.lock-waf_%s_build'%sys.platform)
-try:cache_global=os.path.abspath(os.environ['WAFCACHE'])
-except KeyError:cache_global=''
-platform=Utils.unversioned_sys_platform()
 class opt_parser(optparse.OptionParser):
 	def __init__(self,ctx):
 		optparse.OptionParser.__init__(self,conflict_handler="resolve",version='waf %s (%s)'%(Context.WAFVERSION,Context.WAFREVISION))
 		self.formatter.width=Logs.get_term_cols()
-		p=self.add_option
 		self.ctx=ctx
-		jobs=ctx.jobs()
-		p('-j','--jobs',dest='jobs',default=jobs,type='int',help='amount of parallel jobs (%r)'%jobs)
-		p('-k','--keep',dest='keep',default=0,action='count',help='keep running happily even if errors are found')
-		p('-v','--verbose',dest='verbose',default=0,action='count',help='verbosity level -v -vv or -vvv [default: 0]')
-		p('--nocache',dest='nocache',default=False,action='store_true',help='ignore the WAFCACHE (if set)')
-		p('--zones',dest='zones',default='',action='store',help='debugging zones (task_gen, deps, tasks, etc)')
-		gr=optparse.OptionGroup(self,'configure options')
-		self.add_option_group(gr)
-		gr.add_option('-o','--out',action='store',default='',help='build dir for the project',dest='out')
-		gr.add_option('-t','--top',action='store',default='',help='src dir for the project',dest='top')
-		default_prefix=os.environ.get('PREFIX')
-		if not default_prefix:
-			if platform=='win32':
-				d=tempfile.gettempdir()
-				default_prefix=d[0].upper()+d[1:]
-			else:
-				default_prefix='/usr/local/'
-		gr.add_option('--prefix',dest='prefix',default=default_prefix,help='installation prefix [default: %r]'%default_prefix)
-		gr.add_option('--download',dest='download',default=False,action='store_true',help='try to download the tools if missing')
-		gr=optparse.OptionGroup(self,'build and install options')
-		self.add_option_group(gr)
-		gr.add_option('-p','--progress',dest='progress_bar',default=0,action='count',help='-p: progress bar; -pp: ide output')
-		gr.add_option('--targets',dest='targets',default='',action='store',help='task generators, e.g. "target1,target2"')
-		gr=optparse.OptionGroup(self,'step options')
-		self.add_option_group(gr)
-		gr.add_option('--files',dest='files',default='',action='store',help='files to process, by regexp, e.g. "*/main.c,*/test/main.o"')
-		default_destdir=os.environ.get('DESTDIR','')
-		gr=optparse.OptionGroup(self,'install/uninstall options')
-		self.add_option_group(gr)
-		gr.add_option('--destdir',help='installation root [default: %r]'%default_destdir,default=default_destdir,dest='destdir')
-		gr.add_option('-f','--force',dest='force',default=False,action='store_true',help='force file installation')
+	def print_usage(self,file=None):
+		return self.print_help(file)
 	def get_usage(self):
 		cmds_str={}
 		for cls in Context.classes:
-			if not cls.cmd or cls.cmd=='options':
+			if not cls.cmd or cls.cmd=='options'or cls.cmd.startswith('_'):
 				continue
 			s=cls.__doc__ or''
 			cmds_str[cls.cmd]=s
 		if Context.g_module:
 			for(k,v)in Context.g_module.__dict__.items():
-				if k in['options','init','shutdown']:
+				if k in('options','init','shutdown'):
 					continue
 				if type(v)is type(Context.create_context):
 					if v.__doc__ and not k.startswith('_'):
@@ -80,6 +47,45 @@
 		super(OptionsContext,self).__init__(**kw)
 		self.parser=opt_parser(self)
 		self.option_groups={}
+		jobs=self.jobs()
+		p=self.add_option
+		color=os.environ.get('NOCOLOR','')and'no'or'auto'
+		p('-c','--color',dest='colors',default=color,action='store',help='whether to use colors (yes/no/auto) [default: auto]',choices=('yes','no','auto'))
+		p('-j','--jobs',dest='jobs',default=jobs,type='int',help='amount of parallel jobs (%r)'%jobs)
+		p('-k','--keep',dest='keep',default=0,action='count',help='continue despite errors (-kk to try harder)')
+		p('-v','--verbose',dest='verbose',default=0,action='count',help='verbosity level -v -vv or -vvv [default: 0]')
+		p('--zones',dest='zones',default='',action='store',help='debugging zones (task_gen, deps, tasks, etc)')
+		p('--profile',dest='profile',default='',action='store_true',help=optparse.SUPPRESS_HELP)
+		gr=self.add_option_group('Configuration options')
+		self.option_groups['configure options']=gr
+		gr.add_option('-o','--out',action='store',default='',help='build dir for the project',dest='out')
+		gr.add_option('-t','--top',action='store',default='',help='src dir for the project',dest='top')
+		gr.add_option('--no-lock-in-run',action='store_true',default='',help=optparse.SUPPRESS_HELP,dest='no_lock_in_run')
+		gr.add_option('--no-lock-in-out',action='store_true',default='',help=optparse.SUPPRESS_HELP,dest='no_lock_in_out')
+		gr.add_option('--no-lock-in-top',action='store_true',default='',help=optparse.SUPPRESS_HELP,dest='no_lock_in_top')
+		default_prefix=getattr(Context.g_module,'default_prefix',os.environ.get('PREFIX'))
+		if not default_prefix:
+			if Utils.unversioned_sys_platform()=='win32':
+				d=tempfile.gettempdir()
+				default_prefix=d[0].upper()+d[1:]
+			else:
+				default_prefix='/usr/local/'
+		gr.add_option('--prefix',dest='prefix',default=default_prefix,help='installation prefix [default: %r]'%default_prefix)
+		gr.add_option('--bindir',dest='bindir',help='bindir')
+		gr.add_option('--libdir',dest='libdir',help='libdir')
+		gr=self.add_option_group('Build and installation options')
+		self.option_groups['build and install options']=gr
+		gr.add_option('-p','--progress',dest='progress_bar',default=0,action='count',help='-p: progress bar; -pp: ide output')
+		gr.add_option('--targets',dest='targets',default='',action='store',help='task generators, e.g. "target1,target2"')
+		gr=self.add_option_group('Step options')
+		self.option_groups['step options']=gr
+		gr.add_option('--files',dest='files',default='',action='store',help='files to process, by regexp, e.g. "*/main.c,*/test/main.o"')
+		default_destdir=os.environ.get('DESTDIR','')
+		gr=self.add_option_group('Installation and uninstallation options')
+		self.option_groups['install/uninstall options']=gr
+		gr.add_option('--destdir',help='installation root [default: %r]'%default_destdir,default=default_destdir,dest='destdir')
+		gr.add_option('-f','--force',dest='force',default=False,action='store_true',help='force file installation')
+		gr.add_option('--distcheck-args',metavar='ARGS',help='arguments to pass to distcheck',default=None,action='store')
 	def jobs(self):
 		count=int(os.environ.get('JOBS',0))
 		if count<1:
@@ -94,7 +100,7 @@
 				if not count and os.name not in('nt','java'):
 					try:
 						tmp=self.cmd_and_log(['sysctl','-n','hw.ncpu'],quiet=0)
-					except Exception:
+					except Errors.WafError:
 						pass
 					else:
 						if re.match('^[0-9]+$',tmp):
@@ -105,11 +111,11 @@
 			count=1024
 		return count
 	def add_option(self,*k,**kw):
-		self.parser.add_option(*k,**kw)
+		return self.parser.add_option(*k,**kw)
 	def add_option_group(self,*k,**kw):
 		try:
 			gr=self.option_groups[k[0]]
-		except:
+		except KeyError:
 			gr=self.parser.add_option_group(*k,**kw)
 		self.option_groups[k[0]]=gr
 		return gr
@@ -122,13 +128,20 @@
 					return group
 			return None
 	def parse_args(self,_args=None):
-		global options,commands
+		global options,commands,envvars
 		(options,leftover_args)=self.parser.parse_args(args=_args)
-		commands=leftover_args
+		for arg in leftover_args:
+			if'='in arg:
+				envvars.append(arg)
+			else:
+				commands.append(arg)
 		if options.destdir:
-			options.destdir=os.path.abspath(os.path.expanduser(options.destdir))
+			options.destdir=Utils.sane_path(options.destdir)
 		if options.verbose>=1:
 			self.load('errcheck')
+		colors={'yes':2,'auto':1,'no':0}[options.colors]
+		Logs.enable_colors(colors)
 	def execute(self):
 		super(OptionsContext,self).execute()
 		self.parse_args()
+		Utils.alloc_process_pool(options.jobs)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/processor.py glmark2-2021.02/waflib/processor.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/processor.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/processor.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,40 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+import sys,traceback,base64
+try:
+	import cPickle
+except ImportError:
+	import pickle as cPickle
+try:
+	import subprocess32 as subprocess
+except ImportError:
+	import subprocess
+def run():
+	txt=sys.stdin.readline().strip()
+	if not txt:
+		sys.exit(1)
+	[cmd,kwargs,cargs]=cPickle.loads(base64.b64decode(txt))
+	cargs=cargs or{}
+	ret=1
+	out,err,ex,trace=(None,None,None,None)
+	try:
+		proc=subprocess.Popen(cmd,**kwargs)
+		out,err=proc.communicate(**cargs)
+		ret=proc.returncode
+	except(OSError,ValueError,Exception)as e:
+		exc_type,exc_value,tb=sys.exc_info()
+		exc_lines=traceback.format_exception(exc_type,exc_value,tb)
+		trace=str(cmd)+'\n'+''.join(exc_lines)
+		ex=e.__class__.__name__
+	tmp=[ret,out,err,ex,trace]
+	obj=base64.b64encode(cPickle.dumps(tmp))
+	sys.stdout.write(obj.decode())
+	sys.stdout.write('\n')
+	sys.stdout.flush()
+while 1:
+	try:
+		run()
+	except KeyboardInterrupt:
+		break
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Runner.py glmark2-2021.02/waflib/Runner.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Runner.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Runner.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,57 +1,57 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import random,atexit
+import random
 try:
 	from queue import Queue
-except:
+except ImportError:
 	from Queue import Queue
 from waflib import Utils,Task,Errors,Logs
-GAP=10
-class TaskConsumer(Utils.threading.Thread):
-	def __init__(self):
+GAP=20
+class Consumer(Utils.threading.Thread):
+	__slots__=('task','spawner')
+	def __init__(self,spawner,task):
 		Utils.threading.Thread.__init__(self)
-		self.ready=Queue()
+		self.task=task
+		self.spawner=spawner
+		self.setDaemon(1)
+		self.start()
+	def run(self):
+		try:
+			if not self.spawner.master.stop:
+				self.task.process()
+		finally:
+			self.spawner.sem.release()
+			self.spawner.master.out.put(self.task)
+			self.task=None
+			self.spawner=None
+class Spawner(Utils.threading.Thread):
+	def __init__(self,master):
+		Utils.threading.Thread.__init__(self)
+		self.master=master
+		self.sem=Utils.threading.Semaphore(master.numjobs)
 		self.setDaemon(1)
 		self.start()
 	def run(self):
 		try:
 			self.loop()
-		except:
+		except Exception:
 			pass
 	def loop(self):
+		master=self.master
 		while 1:
-			tsk=self.ready.get()
-			if not isinstance(tsk,Task.TaskBase):
-				tsk(self)
-			else:
-				tsk.process()
-pool=Queue()
-def get_pool():
-	try:
-		return pool.get(False)
-	except:
-		return TaskConsumer()
-def put_pool(x):
-	pool.put(x)
-def _free_resources():
-	global pool
-	lst=[]
-	while pool.qsize():
-		lst.append(pool.get())
-	for x in lst:
-		x.ready.put(None)
-	for x in lst:
-		x.join()
-	pool=None
-atexit.register(_free_resources)
+			task=master.ready.get()
+			self.sem.acquire()
+			task.log_display(task.generator.bld)
+			Consumer(self,task)
 class Parallel(object):
 	def __init__(self,bld,j=2):
 		self.numjobs=j
 		self.bld=bld
-		self.outstanding=[]
-		self.frozen=[]
+		self.outstanding=Utils.deque()
+		self.frozen=Utils.deque()
+		self.ready=Queue(0)
 		self.out=Queue(0)
 		self.count=0
 		self.processed=1
@@ -59,13 +59,14 @@
 		self.error=[]
 		self.biter=None
 		self.dirty=False
+		self.spawner=Spawner(self)
 	def get_next_task(self):
 		if not self.outstanding:
 			return None
-		return self.outstanding.pop(0)
+		return self.outstanding.popleft()
 	def postpone(self,tsk):
 		if random.randint(0,1):
-			self.frozen.insert(0,tsk)
+			self.frozen.appendleft(tsk)
 		else:
 			self.frozen.append(tsk)
 	def refill_task_list(self):
@@ -77,7 +78,7 @@
 			elif self.frozen:
 				try:
 					cond=self.deadlock==self.processed
-				except:
+				except AttributeError:
 					pass
 				else:
 					if cond:
@@ -92,15 +93,15 @@
 						raise Errors.WafError('Deadlock detected: %s%s'%(msg,''.join(lst)))
 				self.deadlock=self.processed
 			if self.frozen:
-				self.outstanding+=self.frozen
-				self.frozen=[]
+				self.outstanding.extend(self.frozen)
+				self.frozen.clear()
 			elif not self.count:
-				self.outstanding.extend(self.biter.next())
+				self.outstanding.extend(next(self.biter))
 				self.total=self.bld.total()
 				break
 	def add_more_tasks(self,tsk):
 		if getattr(tsk,'more_tasks',None):
-			self.outstanding+=tsk.more_tasks
+			self.outstanding.extend(tsk.more_tasks)
 			self.total+=len(tsk.more_tasks)
 	def get_out(self):
 		tsk=self.out.get()
@@ -109,40 +110,38 @@
 		self.count-=1
 		self.dirty=True
 		return tsk
+	def add_task(self,tsk):
+		self.ready.put(tsk)
+	def skip(self,tsk):
+		tsk.hasrun=Task.SKIPPED
 	def error_handler(self,tsk):
+		if hasattr(tsk,'scan')and hasattr(tsk,'uid'):
+			try:
+				del self.bld.imp_sigs[tsk.uid()]
+			except KeyError:
+				pass
 		if not self.bld.keep:
 			self.stop=True
 		self.error.append(tsk)
-	def add_task(self,tsk):
+	def task_status(self,tsk):
 		try:
-			self.pool
-		except AttributeError:
-			self.init_task_pool()
-		self.ready.put(tsk)
-	def init_task_pool(self):
-		pool=self.pool=[get_pool()for i in range(self.numjobs)]
-		self.ready=Queue(0)
-		def setq(consumer):
-			consumer.ready=self.ready
-		for x in pool:
-			x.ready.put(setq)
-		return pool
-	def free_task_pool(self):
-		def setq(consumer):
-			consumer.ready=Queue(0)
-			self.out.put(self)
-		try:
-			pool=self.pool
-		except:
-			pass
-		else:
-			for x in pool:
-				self.ready.put(setq)
-			for x in pool:
-				self.get_out()
-			for x in pool:
-				put_pool(x)
-			self.pool=[]
+			return tsk.runnable_status()
+		except Exception:
+			self.processed+=1
+			tsk.err_msg=Utils.ex_stack()
+			if not self.stop and self.bld.keep:
+				self.skip(tsk)
+				if self.bld.keep==1:
+					if Logs.verbose>1 or not self.error:
+						self.error.append(tsk)
+					self.stop=True
+				else:
+					if Logs.verbose>1:
+						self.error.append(tsk)
+				return Task.EXCEPTION
+			tsk.hasrun=Task.EXCEPTION
+			self.error_handler(tsk)
+			return Task.EXCEPTION
 	def start(self):
 		self.total=self.bld.total()
 		while not self.stop:
@@ -158,40 +157,25 @@
 				continue
 			if self.stop:
 				break
-			try:
-				st=tsk.runnable_status()
-			except Exception:
+			st=self.task_status(tsk)
+			if st==Task.RUN_ME:
+				self.count+=1
 				self.processed+=1
-				tsk.err_msg=Utils.ex_stack()
-				if not self.stop and self.bld.keep:
-					tsk.hasrun=Task.SKIPPED
-					if self.bld.keep==1:
-						if Logs.verbose>1 or not self.error:
-							self.error.append(tsk)
-						self.stop=True
-					else:
-						if Logs.verbose>1:
-							self.error.append(tsk)
-					continue
-				tsk.hasrun=Task.EXCEPTION
-				self.error_handler(tsk)
-				continue
+				if self.numjobs==1:
+					tsk.log_display(tsk.generator.bld)
+					try:
+						tsk.process()
+					finally:
+						self.out.put(tsk)
+				else:
+					self.add_task(tsk)
 			if st==Task.ASK_LATER:
 				self.postpone(tsk)
 			elif st==Task.SKIP_ME:
 				self.processed+=1
-				tsk.hasrun=Task.SKIPPED
+				self.skip(tsk)
 				self.add_more_tasks(tsk)
-			else:
-				tsk.position=(self.processed,self.total)
-				self.count+=1
-				tsk.master=self
-				self.processed+=1
-				if self.numjobs==1:
-					tsk.process()
-				else:
-					self.add_task(tsk)
 		while self.error and self.count:
 			self.get_out()
+		self.ready.put(None)
 		assert(self.count==0 or self.stop)
-		self.free_task_pool()
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Scripting.py glmark2-2021.02/waflib/Scripting.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Scripting.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Scripting.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,8 +1,8 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,shutil,traceback,errno,sys,stat
+import os,shlex,shutil,traceback,errno,sys,stat
 from waflib import Utils,Configure,Logs,Options,ConfigSet,Context,Errors,Build,Node
 build_dir_override=None
 no_climb_commands=['configure']
@@ -10,7 +10,7 @@
 def waf_entry_point(current_directory,version,wafdir):
 	Logs.init_log()
 	if Context.WAFVERSION!=version:
-		Logs.error('Waf script %r and library %r do not match (directory %r)'%(version,Context.WAFVERSION,wafdir))
+		Logs.error('Waf script %r and library %r do not match (directory %r)',version,Context.WAFVERSION,wafdir)
 		sys.exit(1)
 	if'--version'in sys.argv:
 		Context.run_dir=current_directory
@@ -18,26 +18,43 @@
 		ctx.curdir=current_directory
 		ctx.parse_args()
 		sys.exit(0)
+	if len(sys.argv)>1:
+		potential_wscript=os.path.join(current_directory,sys.argv[1])
+		if os.path.basename(potential_wscript)=='wscript'and os.path.isfile(potential_wscript):
+			current_directory=os.path.normpath(os.path.dirname(potential_wscript))
+			sys.argv.pop(1)
 	Context.waf_dir=wafdir
 	Context.launch_dir=current_directory
-	no_climb=os.environ.get('NOCLIMB',None)
+	no_climb=os.environ.get('NOCLIMB')
 	if not no_climb:
 		for k in no_climb_commands:
-			if k in sys.argv:
-				no_climb=True
-				break
+			for y in sys.argv:
+				if y.startswith(k):
+					no_climb=True
+					break
+	for i,x in enumerate(sys.argv):
+		if x.startswith('--top='):
+			Context.run_dir=Context.top_dir=Utils.sane_path(x[6:])
+			sys.argv[i]='--top='+Context.run_dir
+		if x.startswith('--out='):
+			Context.out_dir=Utils.sane_path(x[6:])
+			sys.argv[i]='--out='+Context.out_dir
 	cur=current_directory
-	while cur:
-		lst=os.listdir(cur)
+	while cur and not Context.top_dir:
+		try:
+			lst=os.listdir(cur)
+		except OSError:
+			lst=[]
+			Logs.error('Directory %r is unreadable!',cur)
 		if Options.lockfile in lst:
 			env=ConfigSet.ConfigSet()
 			try:
 				env.load(os.path.join(cur,Options.lockfile))
 				ino=os.stat(cur)[stat.ST_INO]
-			except Exception:
+			except EnvironmentError:
 				pass
 			else:
-				for x in[env.run_dir,env.top_dir,env.out_dir]:
+				for x in(env.run_dir,env.top_dir,env.out_dir):
 					if Utils.is_win32:
 						if cur==x:
 							load=True
@@ -45,14 +62,14 @@
 					else:
 						try:
 							ino2=os.stat(x)[stat.ST_INO]
-						except:
+						except OSError:
 							pass
 						else:
 							if ino==ino2:
 								load=True
 								break
 				else:
-					Logs.warn('invalid lock file in %s'%cur)
+					Logs.warn('invalid lock file in %s',cur)
 					load=False
 				if load:
 					Context.run_dir=env.run_dir
@@ -76,36 +93,44 @@
 			ctx.curdir=current_directory
 			ctx.parse_args()
 			sys.exit(0)
-		Logs.error('Waf: Run from a directory containing a file named %r'%Context.WSCRIPT_FILE)
+		Logs.error('Waf: Run from a directory containing a file named %r',Context.WSCRIPT_FILE)
 		sys.exit(1)
 	try:
 		os.chdir(Context.run_dir)
 	except OSError:
-		Logs.error('Waf: The folder %r is unreadable'%Context.run_dir)
+		Logs.error('Waf: The folder %r is unreadable',Context.run_dir)
 		sys.exit(1)
 	try:
-		set_main_module(Context.run_dir+os.sep+Context.WSCRIPT_FILE)
-	except Errors.WafError ,e:
+		set_main_module(os.path.normpath(os.path.join(Context.run_dir,Context.WSCRIPT_FILE)))
+	except Errors.WafError as e:
 		Logs.pprint('RED',e.verbose_msg)
 		Logs.error(str(e))
 		sys.exit(1)
-	except Exception ,e:
-		Logs.error('Waf: The wscript in %r is unreadable'%Context.run_dir,e)
-		traceback.print_exc(file=sys.stdout)
-		sys.exit(2)
-	try:
-		run_commands()
-	except Errors.WafError ,e:
-		if Logs.verbose>1:
-			Logs.pprint('RED',e.verbose_msg)
-		Logs.error(e.msg)
-		sys.exit(1)
-	except Exception ,e:
+	except Exception as e:
+		Logs.error('Waf: The wscript in %r is unreadable',Context.run_dir)
 		traceback.print_exc(file=sys.stdout)
 		sys.exit(2)
-	except KeyboardInterrupt:
-		Logs.pprint('RED','Interrupted')
-		sys.exit(68)
+	if'--profile'in sys.argv:
+		import cProfile,pstats
+		cProfile.runctx('from waflib import Scripting; Scripting.run_commands()',{},{},'profi.txt')
+		p=pstats.Stats('profi.txt')
+		p.sort_stats('time').print_stats(75)
+	else:
+		try:
+			run_commands()
+		except Errors.WafError as e:
+			if Logs.verbose>1:
+				Logs.pprint('RED',e.verbose_msg)
+			Logs.error(e.msg)
+			sys.exit(1)
+		except SystemExit:
+			raise
+		except Exception as e:
+			traceback.print_exc(file=sys.stdout)
+			sys.exit(2)
+		except KeyboardInterrupt:
+			Logs.pprint('RED','Interrupted')
+			sys.exit(68)
 def set_main_module(file_path):
 	Context.g_module=Context.load_module(file_path)
 	Context.g_module.root_path=file_path
@@ -113,7 +138,7 @@
 		name=obj.__name__
 		if not name in Context.g_module.__dict__:
 			setattr(Context.g_module,name,obj)
-	for k in[update,dist,distclean,distcheck,update]:
+	for k in(dist,distclean,distcheck):
 		set_def(k)
 	if not'init'in Context.g_module.__dict__:
 		Context.g_module.init=Utils.nada
@@ -123,11 +148,13 @@
 		Context.g_module.options=Utils.nada
 def parse_options():
 	Context.create_context('options').execute()
+	for var in Options.envvars:
+		(name,value)=var.split('=',1)
+		os.environ[name.strip()]=value
 	if not Options.commands:
 		Options.commands=[default_cmd]
 	Options.commands=[x for x in Options.commands if x!='options']
 	Logs.verbose=Options.options.verbose
-	Logs.init_log()
 	if Options.options.zones:
 		Logs.zones=Options.options.zones.split(',')
 		if not Logs.verbose:
@@ -138,43 +165,39 @@
 		Logs.zones=['*']
 def run_command(cmd_name):
 	ctx=Context.create_context(cmd_name)
+	ctx.log_timer=Utils.Timer()
 	ctx.options=Options.options
 	ctx.cmd=cmd_name
-	ctx.execute()
+	try:
+		ctx.execute()
+	finally:
+		ctx.finalize()
 	return ctx
 def run_commands():
 	parse_options()
 	run_command('init')
 	while Options.commands:
 		cmd_name=Options.commands.pop(0)
-		timer=Utils.Timer()
-		run_command(cmd_name)
-		if not Options.options.progress_bar:
-			elapsed=' (%s)'%str(timer)
-			Logs.info('%r finished successfully%s'%(cmd_name,elapsed))
+		ctx=run_command(cmd_name)
+		Logs.info('%r finished successfully (%s)',cmd_name,ctx.log_timer)
 	run_command('shutdown')
-def _can_distclean(name):
-	for k in'.o .moc .exe'.split():
-		if name.endswith(k):
-			return True
-	return False
 def distclean_dir(dirname):
 	for(root,dirs,files)in os.walk(dirname):
 		for f in files:
-			if _can_distclean(f):
-				fname=root+os.sep+f
+			if f.endswith(('.o','.moc','.exe')):
+				fname=os.path.join(root,f)
 				try:
-					os.unlink(fname)
-				except:
-					Logs.warn('could not remove %r'%fname)
-	for x in[Context.DBFILE,'config.log']:
+					os.remove(fname)
+				except OSError:
+					Logs.warn('Could not remove %r',fname)
+	for x in(Context.DBFILE,'config.log'):
 		try:
-			os.unlink(x)
-		except:
+			os.remove(x)
+		except OSError:
 			pass
 	try:
 		shutil.rmtree('c4che')
-	except:
+	except OSError:
 		pass
 def distclean(ctx):
 	'''removes the build directory'''
@@ -183,28 +206,30 @@
 		if f==Options.lockfile:
 			try:
 				proj=ConfigSet.ConfigSet(f)
-			except:
-				Logs.warn('could not read %r'%f)
+			except IOError:
+				Logs.warn('Could not read %r',f)
 				continue
 			if proj['out_dir']!=proj['top_dir']:
 				try:
 					shutil.rmtree(proj['out_dir'])
-				except IOError:
-					pass
-				except OSError ,e:
+				except EnvironmentError as e:
 					if e.errno!=errno.ENOENT:
-						Logs.warn('project %r cannot be removed'%proj[Context.OUT])
+						Logs.warn('Could not remove %r',proj['out_dir'])
 			else:
 				distclean_dir(proj['out_dir'])
 			for k in(proj['out_dir'],proj['top_dir'],proj['run_dir']):
+				p=os.path.join(k,Options.lockfile)
 				try:
-					os.remove(os.path.join(k,Options.lockfile))
-				except OSError ,e:
+					os.remove(p)
+				except OSError as e:
 					if e.errno!=errno.ENOENT:
-						Logs.warn('file %r cannot be removed'%f)
-		if f.startswith('.waf')and not Options.commands:
-			shutil.rmtree(f,ignore_errors=True)
+						Logs.warn('Could not remove %r',p)
+		if not Options.commands:
+			for x in'.waf-1. waf-1. .waf3-1. waf3-1.'.split():
+				if f.startswith(x):
+					shutil.rmtree(f,ignore_errors=True)
 class Dist(Context.Context):
+	'''creates an archive containing the project source code'''
 	cmd='dist'
 	fun='dist'
 	algo='tar.bz2'
@@ -217,12 +242,12 @@
 		arch_name=self.get_arch_name()
 		try:
 			self.base_path
-		except:
+		except AttributeError:
 			self.base_path=self.path
 		node=self.base_path.make_node(arch_name)
 		try:
 			node.delete()
-		except:
+		except OSError:
 			pass
 		files=self.get_files()
 		if self.algo.startswith('tar.'):
@@ -238,16 +263,14 @@
 				zip.write(x.abspath(),archive_name,zipfile.ZIP_DEFLATED)
 			zip.close()
 		else:
-			self.fatal('Valid algo types are tar.bz2, tar.gz or zip')
+			self.fatal('Valid algo types are tar.bz2, tar.gz, tar.xz or zip')
 		try:
-			from hashlib import sha1 as sha
+			from hashlib import sha1
 		except ImportError:
-			from sha import sha
-		try:
-			digest=" (sha=%r)"%sha(node.read()).hexdigest()
-		except:
 			digest=''
-		Logs.info('New archive created: %s%s'%(self.arch_name,digest))
+		else:
+			digest=' (sha=%r)'%sha1(node.read(flags='rb')).hexdigest()
+		Logs.info('New archive created: %s%s',self.arch_name,digest)
 	def get_tar_path(self,node):
 		return node.abspath()
 	def add_tar_file(self,x,tar):
@@ -257,28 +280,29 @@
 		tinfo.gid=0
 		tinfo.uname='root'
 		tinfo.gname='root'
-		fu=None
-		try:
+		if os.path.isfile(p):
 			fu=open(p,'rb')
-			tar.addfile(tinfo,fileobj=fu)
-		finally:
-			if fu:
+			try:
+				tar.addfile(tinfo,fileobj=fu)
+			finally:
 				fu.close()
+		else:
+			tar.addfile(tinfo)
 	def get_tar_prefix(self):
 		try:
 			return self.tar_prefix
-		except:
+		except AttributeError:
 			return self.get_base_name()
 	def get_arch_name(self):
 		try:
 			self.arch_name
-		except:
+		except AttributeError:
 			self.arch_name=self.get_base_name()+'.'+self.ext_algo.get(self.algo,self.algo)
 		return self.arch_name
 	def get_base_name(self):
 		try:
 			self.base_name
-		except:
+		except AttributeError:
 			appname=getattr(Context.g_module,Context.APPNAME,'noname')
 			version=getattr(Context.g_module,Context.VERSION,'1.0')
 			self.base_name=appname+'-'+version
@@ -286,16 +310,17 @@
 	def get_excl(self):
 		try:
 			return self.excl
-		except:
-			self.excl=Node.exclude_regs+' **/waf-1.6.* **/.waf-1.6* **/waf3-1.6.* **/.waf3-1.6* **/*~ **/*.rej **/*.orig **/*.pyc **/*.pyo **/*.bak **/*.swp **/.lock-w*'
-			nd=self.root.find_node(Context.out_dir)
-			if nd:
-				self.excl+=' '+nd.path_from(self.base_path)
+		except AttributeError:
+			self.excl=Node.exclude_regs+' **/waf-1.8.* **/.waf-1.8* **/waf3-1.8.* **/.waf3-1.8* **/*~ **/*.rej **/*.orig **/*.pyc **/*.pyo **/*.bak **/*.swp **/.lock-w*'
+			if Context.out_dir:
+				nd=self.root.find_node(Context.out_dir)
+				if nd:
+					self.excl+=' '+nd.path_from(self.base_path)
 			return self.excl
 	def get_files(self):
 		try:
 			files=self.files
-		except:
+		except AttributeError:
 			files=self.base_path.ant_glob('**/*',excl=self.get_excl())
 		return files
 def dist(ctx):
@@ -310,35 +335,27 @@
 		self.check()
 	def check(self):
 		import tempfile,tarfile
-		t=None
 		try:
 			t=tarfile.open(self.get_arch_name())
 			for x in t:
 				t.extract(x)
 		finally:
-			if t:
-				t.close()
+			t.close()
+		cfg=[]
+		if Options.options.distcheck_args:
+			cfg=shlex.split(Options.options.distcheck_args)
+		else:
+			cfg=[x for x in sys.argv if x.startswith('-')]
 		instdir=tempfile.mkdtemp('.inst',self.get_base_name())
-		ret=Utils.subprocess.Popen([sys.argv[0],'configure','install','uninstall','--destdir='+instdir],cwd=self.get_base_name()).wait()
+		ret=Utils.subprocess.Popen([sys.executable,sys.argv[0],'configure','install','uninstall','--destdir='+instdir]+cfg,cwd=self.get_base_name()).wait()
 		if ret:
-			raise Errors.WafError('distcheck failed with code %i'%ret)
+			raise Errors.WafError('distcheck failed with code %r'%ret)
 		if os.path.exists(instdir):
 			raise Errors.WafError('distcheck succeeded, but files were left in %s'%instdir)
 		shutil.rmtree(self.get_base_name())
 def distcheck(ctx):
 	'''checks if the project compiles (tarball from 'dist')'''
 	pass
-def update(ctx):
-	'''updates the plugins from the *waflib/extras* directory'''
-	lst=Options.options.files.split(',')
-	if not lst:
-		lst=[x for x in Utils.listdir(Context.waf_dir+'/waflib/extras')if x.endswith('.py')]
-	for x in lst:
-		tool=x.replace('.py','')
-		try:
-			Configure.download_tool(tool,force=True,ctx=ctx)
-		except Errors.WafError:
-			Logs.error('Could not find the tool %s in the remote repository'%x)
 def autoconfigure(execute_method):
 	def execute(self):
 		if not Configure.autoconfig:
@@ -347,7 +364,7 @@
 		do_config=False
 		try:
 			env.load(os.path.join(Context.top_dir,Options.lockfile))
-		except Exception:
+		except EnvironmentError:
 			Logs.warn('Configuring the project')
 			do_config=True
 		else:
@@ -355,13 +372,27 @@
 				do_config=True
 			else:
 				h=0
-				for f in env['files']:
-					h=hash((h,Utils.readf(f,'rb')))
-				do_config=h!=env.hash
+				for f in env.files:
+					try:
+						h=Utils.h_list((h,Utils.readf(f,'rb')))
+					except EnvironmentError:
+						do_config=True
+						break
+				else:
+					do_config=h!=env.hash
 		if do_config:
-			Options.commands.insert(0,self.cmd)
-			Options.commands.insert(0,'configure')
-			return
-		return execute_method(self)
+			cmd=env.config_cmd or'configure'
+			if Configure.autoconfig=='clobber':
+				tmp=Options.options.__dict__
+				Options.options.__dict__=env.options
+				try:
+					run_command(cmd)
+				finally:
+					Options.options.__dict__=tmp
+			else:
+				run_command(cmd)
+			run_command(self.cmd)
+		else:
+			return execute_method(self)
 	return execute
 Build.BuildContext.execute=autoconfigure(Build.BuildContext.execute)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/TaskGen.py glmark2-2021.02/waflib/TaskGen.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/TaskGen.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/TaskGen.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,21 +1,18 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
 import copy,re,os
-from waflib import Task,Utils,Logs,Errors,ConfigSet
+from waflib import Task,Utils,Logs,Errors,ConfigSet,Node
 feats=Utils.defaultdict(set)
+HEADER_EXTS=['.h','.hpp','.hxx','.hh']
 class task_gen(object):
-	mappings={}
+	mappings=Utils.ordered_iter_dict()
 	prec=Utils.defaultdict(list)
 	def __init__(self,*k,**kw):
 		self.source=''
 		self.target=''
 		self.meths=[]
-		self.prec=Utils.defaultdict(list)
-		self.mappings={}
 		self.features=[]
 		self.tasks=[]
 		if not'bld'in kw:
@@ -27,20 +24,22 @@
 			self.env=self.bld.env.derive()
 			self.path=self.bld.path
 			try:
-				self.idx=self.bld.idx[id(self.path)]=self.bld.idx.get(id(self.path),0)+1
+				self.idx=self.bld.idx[self.path]=self.bld.idx.get(self.path,0)+1
 			except AttributeError:
 				self.bld.idx={}
-				self.idx=self.bld.idx[id(self.path)]=1
+				self.idx=self.bld.idx[self.path]=1
 		for key,val in kw.items():
 			setattr(self,key,val)
 	def __str__(self):
 		return"<task_gen %r declared in %s>"%(self.name,self.path.abspath())
 	def __repr__(self):
 		lst=[]
-		for x in self.__dict__.keys():
-			if x not in['env','bld','compiled_tasks','tasks']:
+		for x in self.__dict__:
+			if x not in('env','bld','compiled_tasks','tasks'):
 				lst.append("%s=%s"%(x,repr(getattr(self,x))))
 		return"bld(%s) in %s"%(", ".join(lst),self.path.abspath())
+	def get_cwd(self):
+		return self.bld.bldnode
 	def get_name(self):
 		try:
 			return self._name
@@ -55,8 +54,10 @@
 		self._name=name
 	name=property(get_name,set_name)
 	def to_list(self,val):
-		if isinstance(val,str):return val.split()
-		else:return val
+		if isinstance(val,str):
+			return val.split()
+		else:
+			return val
 	def post(self):
 		if getattr(self,'posted',None):
 			return False
@@ -67,10 +68,10 @@
 			st=feats[x]
 			if not st:
 				if not x in Task.classes:
-					Logs.warn('feature %r does not exist - bind at least one method to it'%x)
+					Logs.warn('feature %r does not exist - bind at least one method to it',x)
 			keys.update(list(st))
 		prec={}
-		prec_tbl=self.prec or task_gen.prec
+		prec_tbl=self.prec
 		for x in prec_tbl:
 			if x in keys:
 				prec[x]=prec_tbl[x]
@@ -80,6 +81,7 @@
 				if a in x:break
 			else:
 				tmp.append(a)
+		tmp.sort()
 		out=[]
 		while tmp:
 			e=tmp.pop()
@@ -97,42 +99,45 @@
 					else:
 						tmp.append(x)
 		if prec:
-			raise Errors.WafError('Cycle detected in the method execution %r'%prec)
+			txt='\n'.join(['- %s after %s'%(k,repr(v))for k,v in prec.items()])
+			raise Errors.WafError('Cycle detected in the method execution\n%s'%txt)
 		out.reverse()
 		self.meths=out
-		Logs.debug('task_gen: posting %s %d'%(self,id(self)))
+		Logs.debug('task_gen: posting %s %d',self,id(self))
 		for x in out:
 			try:
 				v=getattr(self,x)
 			except AttributeError:
 				raise Errors.WafError('%r is not a valid task generator method'%x)
-			Logs.debug('task_gen: -> %s (%d)'%(x,id(self)))
+			Logs.debug('task_gen: -> %s (%d)',x,id(self))
 			v()
-		Logs.debug('task_gen: posted %s'%self.name)
+		Logs.debug('task_gen: posted %s',self.name)
 		return True
 	def get_hook(self,node):
 		name=node.name
 		for k in self.mappings:
-			if name.endswith(k):
-				return self.mappings[k]
-		for k in task_gen.mappings:
-			if name.endswith(k):
-				return task_gen.mappings[k]
-		raise Errors.WafError("File %r has no mapping in %r (did you forget to load a waf tool?)"%(node,task_gen.mappings.keys()))
-	def create_task(self,name,src=None,tgt=None):
+			try:
+				if name.endswith(k):
+					return self.mappings[k]
+			except TypeError:
+				if k.match(name):
+					return self.mappings[k]
+		raise Errors.WafError("File %r has no mapping in %r (have you forgotten to load a waf tool?)"%(node,self.mappings.keys()))
+	def create_task(self,name,src=None,tgt=None,**kw):
 		task=Task.classes[name](env=self.env.derive(),generator=self)
 		if src:
 			task.set_inputs(src)
 		if tgt:
 			task.set_outputs(tgt)
+		task.__dict__.update(kw)
 		self.tasks.append(task)
 		return task
 	def clone(self,env):
 		newobj=self.bld()
 		for x in self.__dict__:
-			if x in['env','bld']:
+			if x in('env','bld'):
 				continue
-			elif x in['path','features']:
+			elif x in('path','features'):
 				setattr(newobj,x,getattr(self,x))
 			else:
 				setattr(newobj,x,copy.copy(getattr(self,x)))
@@ -149,12 +154,11 @@
 		name=rule
 	cls=Task.task_factory(name,rule,color=color,ext_in=ext_in,ext_out=ext_out,before=before,after=after,scan=scan,shell=shell)
 	def x_file(self,node):
-		ext=decider and decider(self,node)or cls.ext_out
 		if ext_in:
 			_ext_in=ext_in[0]
 		tsk=self.create_task(name,node)
 		cnt=0
-		keys=self.mappings.keys()+self.__class__.mappings.keys()
+		ext=decider(self,node)if decider else cls.ext_out
 		for x in ext:
 			k=node.change_ext(x,ext_in=_ext_in)
 			tsk.outputs.append(k)
@@ -162,13 +166,13 @@
 				if cnt<int(reentrant):
 					self.source.append(k)
 			else:
-				for y in keys:
+				for y in self.mappings:
 					if k.name.endswith(y):
 						self.source.append(k)
 						break
 			cnt+=1
 		if install_path:
-			self.bld.install_files(install_path,tsk.outputs)
+			self.install_task=self.add_install_files(install_to=install_path,install_from=tsk.outputs)
 		return tsk
 	for x in cls.ext_in:
 		task_gen.mappings[x]=x_file
@@ -208,11 +212,12 @@
 			task_gen.mappings[x]=func
 		return func
 	return deco
+@taskgen_method
 def to_nodes(self,lst,path=None):
 	tmp=[]
 	path=path or self.path
 	find=path.find_resource
-	if isinstance(lst,self.path.__class__):
+	if isinstance(lst,Node.Node):
 		lst=[lst]
 	for x in Utils.to_list(lst):
 		if isinstance(x,str):
@@ -223,15 +228,57 @@
 			raise Errors.WafError("source not found: %r in %r"%(x,self))
 		tmp.append(node)
 	return tmp
+@feature('*')
 def process_source(self):
 	self.source=self.to_nodes(getattr(self,'source',[]))
 	for node in self.source:
 		self.get_hook(node)(self,node)
+@feature('*')
+@before_method('process_source')
 def process_rule(self):
 	if not getattr(self,'rule',None):
 		return
-	name=str(getattr(self,'name',None)or self.target or self.rule)
-	cls=Task.task_factory(name,self.rule,getattr(self,'vars',[]),shell=getattr(self,'shell',True),color=getattr(self,'color','BLUE'))
+	name=str(getattr(self,'name',None)or self.target or getattr(self.rule,'__name__',self.rule))
+	try:
+		cache=self.bld.cache_rule_attr
+	except AttributeError:
+		cache=self.bld.cache_rule_attr={}
+	cls=None
+	if getattr(self,'cache_rule','True'):
+		try:
+			cls=cache[(name,self.rule)]
+		except KeyError:
+			pass
+	if not cls:
+		rule=self.rule
+		if hasattr(self,'chmod'):
+			def chmod_fun(tsk):
+				for x in tsk.outputs:
+					os.chmod(x.abspath(),self.chmod)
+			rule=(self.rule,chmod_fun)
+		cls=Task.task_factory(name,rule,getattr(self,'vars',[]),shell=getattr(self,'shell',True),color=getattr(self,'color','BLUE'),scan=getattr(self,'scan',None))
+		if getattr(self,'scan',None):
+			cls.scan=self.scan
+		elif getattr(self,'deps',None):
+			def scan(self):
+				nodes=[]
+				for x in self.generator.to_list(getattr(self.generator,'deps',None)):
+					node=self.generator.path.find_resource(x)
+					if not node:
+						self.generator.bld.fatal('Could not find %r (was it declared?)'%x)
+					nodes.append(node)
+				return[nodes,[]]
+			cls.scan=scan
+		if getattr(self,'always',None):
+			cls.always_run=True
+		for x in('after','before','ext_in','ext_out'):
+			setattr(cls,x,getattr(self,x,[]))
+		if getattr(self,'cache_rule','True'):
+			cache[(name,self.rule)]=cls
+		if getattr(self,'cls_str',None):
+			setattr(cls,'__str__',self.cls_str)
+		if getattr(self,'cls_keyword',None):
+			setattr(cls,'keyword',self.cls_keyword)
 	tsk=self.create_task(name)
 	if getattr(self,'target',None):
 		if isinstance(self.target,str):
@@ -245,30 +292,13 @@
 				x.parent.mkdir()
 				tsk.outputs.append(x)
 		if getattr(self,'install_path',None):
-			self.bld.install_files(self.install_path,tsk.outputs)
+			self.install_task=self.add_install_files(install_to=self.install_path,install_from=tsk.outputs,chmod=getattr(self,'chmod',Utils.O644))
 	if getattr(self,'source',None):
 		tsk.inputs=self.to_nodes(self.source)
 		self.source=[]
-	if getattr(self,'scan',None):
-		cls.scan=self.scan
-	elif getattr(self,'deps',None):
-		def scan(self):
-			nodes=[]
-			for x in self.generator.to_list(self.generator.deps):
-				node=self.generator.path.find_resource(x)
-				if not node:
-					self.generator.bld.fatal('Could not find %r (was it declared?)'%x)
-				nodes.append(node)
-			return[nodes,[]]
-		cls.scan=scan
 	if getattr(self,'cwd',None):
 		tsk.cwd=self.cwd
-	if getattr(self,'update_outputs',None)or getattr(self,'on_results',None):
-		Task.update_outputs(cls)
-	if getattr(self,'always',None):
-		Task.always_run(cls)
-	for x in['after','before','ext_in','ext_out']:
-		setattr(cls,x,getattr(self,x,[]))
+@feature('seq')
 def sequence_order(self):
 	if self.meths and self.meths[-1]!='sequence_order':
 		self.meths.append('sequence_order')
@@ -283,8 +313,28 @@
 	self.bld.prev=self
 re_m4=re.compile('@(\w+)@',re.M)
 class subst_pc(Task.Task):
+	def force_permissions(self):
+		if getattr(self.generator,'chmod',None):
+			for x in self.outputs:
+				os.chmod(x.abspath(),self.generator.chmod)
 	def run(self):
-		code=self.inputs[0].read()
+		if getattr(self.generator,'is_copy',None):
+			for i,x in enumerate(self.outputs):
+				x.write(self.inputs[i].read('rb'),'wb')
+			self.force_permissions()
+			return None
+		if getattr(self.generator,'fun',None):
+			ret=self.generator.fun(self)
+			if not ret:
+				self.force_permissions()
+			return ret
+		code=self.inputs[0].read(encoding=getattr(self.generator,'encoding','ISO8859-1'))
+		if getattr(self.generator,'subst_fun',None):
+			code=self.generator.subst_fun(self,code)
+			if code is not None:
+				self.outputs[0].write(code,encoding=getattr(self.generator,'encoding','ISO8859-1'))
+			self.force_permissions()
+			return None
 		code=code.replace('%','%%')
 		lst=[]
 		def repl(match):
@@ -293,61 +343,90 @@
 				lst.append(g(1))
 				return"%%(%s)s"%g(1)
 			return''
-		code=re_m4.sub(repl,code)
+		global re_m4
+		code=getattr(self.generator,'re_m4',re_m4).sub(repl,code)
 		try:
 			d=self.generator.dct
 		except AttributeError:
 			d={}
 			for x in lst:
-				tmp=getattr(self.generator,x,'')or self.env.get_flat(x)or self.env.get_flat(x.upper())
-				d[x]=str(tmp)
-		self.outputs[0].write(code%d)
-		self.generator.bld.raw_deps[self.uid()]=self.dep_vars=lst
+				tmp=getattr(self.generator,x,'')or self.env[x]or self.env[x.upper()]
+				try:
+					tmp=''.join(tmp)
+				except TypeError:
+					tmp=str(tmp)
+				d[x]=tmp
+		code=code%d
+		self.outputs[0].write(code,encoding=getattr(self.generator,'encoding','ISO8859-1'))
+		self.generator.bld.raw_deps[self.uid()]=lst
 		try:delattr(self,'cache_sig')
 		except AttributeError:pass
-		if getattr(self.generator,'chmod',None):
-			os.chmod(self.outputs[0].abspath(),self.generator.chmod)
+		self.force_permissions()
 	def sig_vars(self):
 		bld=self.generator.bld
 		env=self.env
 		upd=self.m.update
+		if getattr(self.generator,'fun',None):
+			upd(Utils.h_fun(self.generator.fun).encode())
+		if getattr(self.generator,'subst_fun',None):
+			upd(Utils.h_fun(self.generator.subst_fun).encode())
 		vars=self.generator.bld.raw_deps.get(self.uid(),[])
 		act_sig=bld.hash_env_vars(env,vars)
 		upd(act_sig)
 		lst=[getattr(self.generator,x,'')for x in vars]
 		upd(Utils.h_list(lst))
 		return self.m.digest()
+@extension('.pc.in')
 def add_pcfile(self,node):
 	tsk=self.create_task('subst_pc',node,node.change_ext('.pc','.pc.in'))
-	self.bld.install_files(getattr(self,'install_path','${LIBDIR}/pkgconfig/'),tsk.outputs)
+	self.install_task=self.add_install_files(install_to=getattr(self,'install_path','${LIBDIR}/pkgconfig/'),install_from=tsk.outputs)
 class subst(subst_pc):
 	pass
+@feature('subst')
+@before_method('process_source','process_rule')
 def process_subst(self):
-	src=self.to_nodes(getattr(self,'source',[]))
-	tgt=getattr(self,'target',[])
-	if isinstance(tgt,self.path.__class__):
+	src=Utils.to_list(getattr(self,'source',[]))
+	if isinstance(src,Node.Node):
+		src=[src]
+	tgt=Utils.to_list(getattr(self,'target',[]))
+	if isinstance(tgt,Node.Node):
 		tgt=[tgt]
-	tgt=[isinstance(x,self.path.__class__)and x or self.path.find_or_declare(x)for x in Utils.to_list(tgt)]
 	if len(src)!=len(tgt):
-		raise Errors.WafError('invalid source or target for %r'%self)
+		raise Errors.WafError('invalid number of source/target for %r'%self)
 	for x,y in zip(src,tgt):
-		if not(x and y):
-			raise Errors.WafError('invalid source or target for %r'%self)
-		tsk=self.create_task('subst',x,y)
-		for a in('after','before','ext_in','ext_out'):
-			val=getattr(self,a,None)
+		if not x or not y:
+			raise Errors.WafError('null source or target for %r'%self)
+		a,b=None,None
+		if isinstance(x,str)and isinstance(y,str)and x==y:
+			a=self.path.find_node(x)
+			b=self.path.get_bld().make_node(y)
+			if not os.path.isfile(b.abspath()):
+				b.parent.mkdir()
+		else:
+			if isinstance(x,str):
+				a=self.path.find_resource(x)
+			elif isinstance(x,Node.Node):
+				a=x
+			if isinstance(y,str):
+				b=self.path.find_or_declare(y)
+			elif isinstance(y,Node.Node):
+				b=y
+		if not a:
+			raise Errors.WafError('could not find %r for %r'%(x,self))
+		has_constraints=False
+		tsk=self.create_task('subst',a,b)
+		for k in('after','before','ext_in','ext_out'):
+			val=getattr(self,k,None)
 			if val:
-				setattr(tsk,a,val)
-	inst_to=getattr(self,'install_path',None)
-	if inst_to:
-		self.bld.install_files(inst_to,tgt,chmod=getattr(self,'chmod',Utils.O644))
+				has_constraints=True
+				setattr(tsk,k,val)
+		if not has_constraints:
+			global HEADER_EXTS
+			for xt in HEADER_EXTS:
+				if b.name.endswith(xt):
+					tsk.before=[k for k in('c','cxx')if k in Task.classes]
+					break
+		inst_to=getattr(self,'install_path',None)
+		if inst_to:
+			self.install_task=self.add_install_files(install_to=inst_to,install_from=b,chmod=getattr(self,'chmod',Utils.O644))
 	self.source=[]
-
-taskgen_method(to_nodes)
-feature('*')(process_source)
-feature('*')(process_rule)
-before_method('process_source')(process_rule)
-feature('seq')(sequence_order)
-extension('.pc.in')(add_pcfile)
-feature('subst')(process_subst)
-before_method('process_source','process_rule')(process_subst)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Task.py glmark2-2021.02/waflib/Task.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Task.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Task.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,8 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import os,shutil,re,tempfile
+import os,re,sys,tempfile
 from waflib import Utils,Logs,Errors
 NOT_RUN=0
 MISSING=1
@@ -20,64 +18,49 @@
 	env = tsk.env
 	gen = tsk.generator
 	bld = gen.bld
-	wd = getattr(tsk, 'cwd', None)
+	cwdx = tsk.get_cwd()
 	p = env.get_flat
 	tsk.last_cmd = cmd = \'\'\' %s \'\'\' % s
-	return tsk.exec_command(cmd, cwd=wd, env=env.env or None)
+	return tsk.exec_command(cmd, cwd=cwdx, env=env.env or None)
 '''
 COMPILE_TEMPLATE_NOSHELL='''
 def f(tsk):
 	env = tsk.env
 	gen = tsk.generator
 	bld = gen.bld
-	wd = getattr(tsk, 'cwd', None)
+	cwdx = tsk.get_cwd()
 	def to_list(xx):
 		if isinstance(xx, str): return [xx]
 		return xx
-	tsk.last_cmd = lst = []
+	def merge(lst1, lst2):
+		if lst1 and lst2:
+			return lst1[:-1] + [lst1[-1] + lst2[0]] + lst2[1:]
+		return lst1 + lst2
+	lst = []
 	%s
-	lst = [x for x in lst if x]
-	return tsk.exec_command(lst, cwd=wd, env=env.env or None)
+	if '' in lst:
+		lst = [x for x in lst if x]
+	tsk.last_cmd = lst
+	return tsk.exec_command(lst, cwd=cwdx, env=env.env or None)
 '''
-def cache_outputs(cls):
-	m1=cls.run
-	def run(self):
-		bld=self.generator.bld
-		if bld.cache_global and not bld.nocache:
-			if self.can_retrieve_cache():
-				return 0
-		return m1(self)
-	cls.run=run
-	m2=cls.post_run
-	def post_run(self):
-		bld=self.generator.bld
-		ret=m2(self)
-		if bld.cache_global and not bld.nocache:
-			self.put_files_cache()
-		return ret
-	cls.post_run=post_run
-	return cls
 classes={}
 class store_task_type(type):
 	def __init__(cls,name,bases,dict):
 		super(store_task_type,cls).__init__(name,bases,dict)
 		name=cls.__name__
-		if name.endswith('_task'):
-			name=name.replace('_task','')
 		if name!='evil'and name!='TaskBase':
 			global classes
 			if getattr(cls,'run_str',None):
 				(f,dvars)=compile_fun(cls.run_str,cls.shell)
-				cls.hcode=cls.run_str
+				cls.hcode=Utils.h_cmd(cls.run_str)
+				cls.orig_run_str=cls.run_str
 				cls.run_str=None
 				cls.run=f
 				cls.vars=list(set(cls.vars+dvars))
 				cls.vars.sort()
 			elif getattr(cls,'run',None)and not'hcode'in cls.__dict__:
-				cls.hcode=Utils.h_fun(cls.run)
-			if not getattr(cls,'nocache',None):
-				cls=cache_outputs(cls)
-			classes[name]=cls
+				cls.hcode=Utils.h_cmd(cls.run)
+			getattr(cls,'register',classes)[name]=cls
 evil=store_task_type('evil',(object,),{})
 class TaskBase(evil):
 	color='GREEN'
@@ -86,6 +69,8 @@
 	before=[]
 	after=[]
 	hcode=''
+	keep_last_cmd=False
+	__slots__=('hasrun','generator')
 	def __init__(self,*k,**kw):
 		self.hasrun=NOT_RUN
 		try:
@@ -96,38 +81,65 @@
 		return'\n\t{task %r: %s %s}'%(self.__class__.__name__,id(self),str(getattr(self,'fun','')))
 	def __str__(self):
 		if hasattr(self,'fun'):
-			return'executing: %s\n'%self.fun.__name__
-		return self.__class__.__name__+'\n'
-	def __hash__(self):
-		return id(self)
-	def exec_command(self,cmd,**kw):
+			return self.fun.__name__
+		return self.__class__.__name__
+	def keyword(self):
+		if hasattr(self,'fun'):
+			return'Function'
+		return'Processing'
+	def get_cwd(self):
 		bld=self.generator.bld
-		try:
-			if not kw.get('cwd',None):
-				kw['cwd']=bld.cwd
-		except AttributeError:
-			bld.cwd=kw['cwd']=bld.variant_dir
-		return bld.exec_command(cmd,**kw)
+		ret=getattr(self,'cwd',None)or getattr(self.generator.bld,'cwd',bld.bldnode)
+		if isinstance(ret,str):
+			self.generator.bld.fatal('Working folders given to tasks must be Node objects')
+		return ret
+	def quote_flag(self,x):
+		old=x
+		if'\\'in x:
+			x=x.replace('\\','\\\\')
+		if'"'in x:
+			x=x.replace('"','\\"')
+		if old!=x or' 'in x or'\t'in x or"'"in x:
+			x='"%s"'%x
+		return x
+	def split_argfile(self,cmd):
+		return([cmd[0]],[self.quote_flag(x)for x in cmd[1:]])
+	def exec_command(self,cmd,**kw):
+		if not'cwd'in kw:
+			kw['cwd']=self.get_cwd()
+		if self.env.PATH:
+			env=kw['env']=dict(kw.get('env')or self.env.env or os.environ)
+			env['PATH']=self.env.PATH if isinstance(self.env.PATH,str)else os.pathsep.join(self.env.PATH)
+		if not isinstance(cmd,str)and(len(repr(cmd))>=8192 if Utils.is_win32 else len(cmd)>200000):
+			cmd,args=self.split_argfile(cmd)
+			try:
+				(fd,tmp)=tempfile.mkstemp()
+				os.write(fd,'\r\n'.join(args).encode())
+				os.close(fd)
+				return self.generator.bld.exec_command(cmd+['@'+tmp],**kw)
+			finally:
+				try:
+					os.remove(tmp)
+				except OSError:
+					pass
+		else:
+			return self.generator.bld.exec_command(cmd,**kw)
 	def runnable_status(self):
 		return RUN_ME
+	def uid(self):
+		return Utils.SIG_NIL
 	def process(self):
-		m=self.master
-		if m.stop:
-			m.out.put(self)
-			return
+		m=self.generator.bld.producer
 		try:
 			del self.generator.bld.task_sigs[self.uid()]
-		except:
+		except KeyError:
 			pass
 		try:
-			self.generator.bld.returned_tasks.append(self)
-			self.log_display(self.generator.bld)
 			ret=self.run()
 		except Exception:
 			self.err_msg=Utils.ex_stack()
 			self.hasrun=EXCEPTION
 			m.error_handler(self)
-			m.out.put(self)
 			return
 		if ret:
 			self.err_code=ret
@@ -144,7 +156,6 @@
 				self.hasrun=SUCCESS
 		if self.hasrun!=SUCCESS:
 			m.error_handler(self)
-		m.out.put(self)
 	def run(self):
 		if hasattr(self,'fun'):
 			return self.fun(self)
@@ -152,11 +163,24 @@
 	def post_run(self):
 		pass
 	def log_display(self,bld):
-		bld.to_log(self.display())
+		if self.generator.bld.progress_bar==3:
+			return
+		s=self.display()
+		if s:
+			if bld.logger:
+				logger=bld.logger
+			else:
+				logger=Logs
+			if self.generator.bld.progress_bar==1:
+				c1=Logs.colors.cursor_off
+				c2=Logs.colors.cursor_on
+				logger.info(s,extra={'stream':sys.stderr,'terminator':'','c1':c1,'c2':c2})
+			else:
+				logger.info(s,extra={'terminator':'','c1':'','c2':''})
 	def display(self):
 		col1=Logs.colors(self.color)
 		col2=Logs.colors.NORMAL
-		master=self.master
+		master=self.generator.bld.producer
 		def cur():
 			tmp=-1
 			if hasattr(master,'ready'):
@@ -180,17 +204,15 @@
 			return None
 		total=master.total
 		n=len(str(total))
-		fs='[%%%dd/%%%dd] %%s%%s%%s'%(n,n)
-		return fs%(cur(),total,col1,s,col2)
-	def attr(self,att,default=None):
-		ret=getattr(self,att,self)
-		if ret is self:return getattr(self.__class__,att,default)
-		return ret
+		fs='[%%%dd/%%%dd] %%s%%s%%s%%s\n'%(n,n)
+		kw=self.keyword()
+		if kw:
+			kw+=' '
+		return fs%(cur(),total,kw,col1,s,col2)
 	def hash_constraints(self):
 		cls=self.__class__
 		tup=(str(cls.before),str(cls.after),str(cls.ext_in),str(cls.ext_out),cls.__name__,cls.hcode)
-		h=hash(tup)
-		return h
+		return hash(tup)
 	def format_error(self):
 		msg=getattr(self,'last_cmd','')
 		name=getattr(self.generator,'name','')
@@ -209,6 +231,8 @@
 			return'invalid status for task in %r: %r'%(name,self.hasrun)
 	def colon(self,var1,var2):
 		tmp=self.env[var1]
+		if not tmp:
+			return[]
 		if isinstance(var2,str):
 			it=self.env[var2]
 		else:
@@ -216,8 +240,6 @@
 		if isinstance(tmp,str):
 			return[tmp%x for x in it]
 		else:
-			if Logs.verbose and not tmp and it:
-				Logs.warn('Missing env variable %r for task %r (generator %r)'%(var1,self,self.generator))
 			lst=[]
 			for y in it:
 				lst.extend(tmp)
@@ -225,6 +247,7 @@
 			return lst
 class Task(TaskBase):
 	vars=[]
+	always_run=False
 	shell=False
 	def __init__(self,*k,**kw):
 		TaskBase.__init__(self,*k,**kw)
@@ -234,21 +257,47 @@
 		self.dep_nodes=[]
 		self.run_after=set([])
 	def __str__(self):
-		env=self.env
-		src_str=' '.join([a.nice_path(env)for a in self.inputs])
-		tgt_str=' '.join([a.nice_path(env)for a in self.outputs])
+		name=self.__class__.__name__
+		if self.outputs:
+			if name.endswith(('lib','program'))or not self.inputs:
+				node=self.outputs[0]
+				return node.path_from(node.ctx.launch_node())
+		if not(self.inputs or self.outputs):
+			return self.__class__.__name__
+		if len(self.inputs)==1:
+			node=self.inputs[0]
+			return node.path_from(node.ctx.launch_node())
+		src_str=' '.join([a.path_from(a.ctx.launch_node())for a in self.inputs])
+		tgt_str=' '.join([a.path_from(a.ctx.launch_node())for a in self.outputs])
 		if self.outputs:sep=' -> '
 		else:sep=''
-		return'%s: %s%s%s\n'%(self.__class__.__name__.replace('_task',''),src_str,sep,tgt_str)
+		return'%s: %s%s%s'%(self.__class__.__name__,src_str,sep,tgt_str)
+	def keyword(self):
+		name=self.__class__.__name__
+		if name.endswith(('lib','program')):
+			return'Linking'
+		if len(self.inputs)==1 and len(self.outputs)==1:
+			return'Compiling'
+		if not self.inputs:
+			if self.outputs:
+				return'Creating'
+			else:
+				return'Running'
+		return'Processing'
 	def __repr__(self):
-		return"".join(['\n\t{task %r: '%id(self),self.__class__.__name__," ",",".join([x.name for x in self.inputs])," -> ",",".join([x.name for x in self.outputs]),'}'])
+		try:
+			ins=",".join([x.name for x in self.inputs])
+			outs=",".join([x.name for x in self.outputs])
+		except AttributeError:
+			ins=",".join([str(x)for x in self.inputs])
+			outs=",".join([str(x)for x in self.outputs])
+		return"".join(['\n\t{task %r: '%id(self),self.__class__.__name__," ",ins," -> ",outs,'}'])
 	def uid(self):
 		try:
 			return self.uid_
 		except AttributeError:
-			m=Utils.md5()
+			m=Utils.md5(self.__class__.__name__)
 			up=m.update
-			up(self.__class__.__name__)
 			for x in self.inputs+self.outputs:
 				up(x.abspath())
 			self.uid_=m.digest()
@@ -263,10 +312,11 @@
 		assert isinstance(task,TaskBase)
 		self.run_after.add(task)
 	def signature(self):
-		try:return self.cache_sig
-		except AttributeError:pass
-		self.m=Utils.md5()
-		self.m.update(self.hcode)
+		try:
+			return self.cache_sig
+		except AttributeError:
+			pass
+		self.m=Utils.md5(self.hcode)
 		self.sig_explicit_deps()
 		self.sig_vars()
 		if self.scan:
@@ -280,124 +330,108 @@
 		for t in self.run_after:
 			if not t.hasrun:
 				return ASK_LATER
-		bld=self.generator.bld
 		try:
 			new_sig=self.signature()
 		except Errors.TaskNotReady:
 			return ASK_LATER
+		bld=self.generator.bld
 		key=self.uid()
 		try:
 			prev_sig=bld.task_sigs[key]
 		except KeyError:
-			Logs.debug("task: task %r must run as it was never run before or the task code changed"%self)
+			Logs.debug('task: task %r must run: it was never run before or the task code changed',self)
 			return RUN_ME
-		for node in self.outputs:
-			try:
-				if node.sig!=new_sig:
-					return RUN_ME
-			except AttributeError:
-				Logs.debug("task: task %r must run as the output nodes do not exist"%self)
-				return RUN_ME
 		if new_sig!=prev_sig:
+			Logs.debug('task: task %r must run: the task signature changed',self)
 			return RUN_ME
-		return SKIP_ME
+		for node in self.outputs:
+			sig=bld.node_sigs.get(node)
+			if not sig:
+				Logs.debug('task: task %r must run: an output node has no signature',self)
+				return RUN_ME
+			if sig!=key:
+				Logs.debug('task: task %r must run: an output node was produced by another task',self)
+				return RUN_ME
+			if not node.exists():
+				Logs.debug('task: task %r must run: an output node does not exist',self)
+				return RUN_ME
+		return(self.always_run and RUN_ME)or SKIP_ME
 	def post_run(self):
 		bld=self.generator.bld
-		sig=self.signature()
 		for node in self.outputs:
-			try:
-				os.stat(node.abspath())
-			except OSError:
+			if not node.exists():
 				self.hasrun=MISSING
 				self.err_msg='-> missing file: %r'%node.abspath()
 				raise Errors.WafError(self.err_msg)
-			node.sig=sig
-		bld.task_sigs[self.uid()]=self.cache_sig
+			bld.node_sigs[node]=self.uid()
+		bld.task_sigs[self.uid()]=self.signature()
+		if not self.keep_last_cmd:
+			try:
+				del self.last_cmd
+			except AttributeError:
+				pass
 	def sig_explicit_deps(self):
 		bld=self.generator.bld
 		upd=self.m.update
 		for x in self.inputs+self.dep_nodes:
-			try:
-				upd(x.get_bld_sig())
-			except(AttributeError,TypeError):
-				raise Errors.WafError('Missing node signature for %r (required by %r)'%(x,self))
+			upd(x.get_bld_sig())
 		if bld.deps_man:
 			additional_deps=bld.deps_man
 			for x in self.inputs+self.outputs:
 				try:
-					d=additional_deps[id(x)]
+					d=additional_deps[x]
 				except KeyError:
 					continue
 				for v in d:
 					if isinstance(v,bld.root.__class__):
-						try:
-							v=v.get_bld_sig()
-						except AttributeError:
-							raise Errors.WafError('Missing node signature for %r (required by %r)'%(v,self))
+						v=v.get_bld_sig()
 					elif hasattr(v,'__call__'):
 						v=v()
 					upd(v)
-		return self.m.digest()
 	def sig_vars(self):
-		bld=self.generator.bld
-		env=self.env
-		upd=self.m.update
-		act_sig=bld.hash_env_vars(env,self.__class__.vars)
-		upd(act_sig)
-		dep_vars=getattr(self,'dep_vars',None)
-		if dep_vars:
-			upd(bld.hash_env_vars(env,dep_vars))
-		return self.m.digest()
+		sig=self.generator.bld.hash_env_vars(self.env,self.__class__.vars)
+		self.m.update(sig)
 	scan=None
 	def sig_implicit_deps(self):
 		bld=self.generator.bld
 		key=self.uid()
-		prev=bld.task_sigs.get((key,'imp'),[])
+		prev=bld.imp_sigs.get(key,[])
 		if prev:
 			try:
 				if prev==self.compute_sig_implicit_deps():
 					return prev
-			except:
+			except Errors.TaskNotReady:
+				raise
+			except EnvironmentError:
 				for x in bld.node_deps.get(self.uid(),[]):
-					if x.is_child_of(bld.srcnode):
+					if not x.is_bld()and not x.exists():
 						try:
-							os.stat(x.abspath())
-						except:
-							try:
-								del x.parent.children[x.name]
-							except:
-								pass
-			del bld.task_sigs[(key,'imp')]
+							del x.parent.children[x.name]
+						except KeyError:
+							pass
+			del bld.imp_sigs[key]
 			raise Errors.TaskRescan('rescan')
-		(nodes,names)=self.scan()
+		(bld.node_deps[key],bld.raw_deps[key])=self.scan()
 		if Logs.verbose:
-			Logs.debug('deps: scanner for %s returned %s %s'%(str(self),str(nodes),str(names)))
-		bld.node_deps[key]=nodes
-		bld.raw_deps[key]=names
-		self.are_implicit_nodes_ready()
+			Logs.debug('deps: scanner for %s: %r; unresolved: %r',self,bld.node_deps[key],bld.raw_deps[key])
 		try:
-			bld.task_sigs[(key,'imp')]=sig=self.compute_sig_implicit_deps()
-		except:
-			if Logs.verbose:
-				for k in bld.node_deps.get(self.uid(),[]):
-					try:
-						k.get_bld_sig()
-					except:
-						Logs.warn('Missing signature for node %r (may cause rebuilds)'%k)
-		else:
-			return sig
+			bld.imp_sigs[key]=self.compute_sig_implicit_deps()
+		except EnvironmentError:
+			for k in bld.node_deps.get(self.uid(),[]):
+				if not k.exists():
+					Logs.warn('Dependency %r for %r is missing: check the task declaration and the build order!',k,self)
+			raise
 	def compute_sig_implicit_deps(self):
 		upd=self.m.update
-		bld=self.generator.bld
 		self.are_implicit_nodes_ready()
-		for k in bld.node_deps.get(self.uid(),[]):
+		for k in self.generator.bld.node_deps.get(self.uid(),[]):
 			upd(k.get_bld_sig())
 		return self.m.digest()
 	def are_implicit_nodes_ready(self):
 		bld=self.generator.bld
 		try:
 			cache=bld.dct_implicit_nodes
-		except:
+		except AttributeError:
 			bld.dct_implicit_nodes=cache={}
 		try:
 			dct=cache[bld.cur]
@@ -415,71 +449,19 @@
 			for tsk in self.run_after:
 				if not tsk.hasrun:
 					raise Errors.TaskNotReady('not ready')
-	def can_retrieve_cache(self):
-		if not getattr(self,'outputs',None):
-			return None
-		sig=self.signature()
-		ssig=Utils.to_hex(self.uid())+Utils.to_hex(sig)
-		dname=os.path.join(self.generator.bld.cache_global,ssig)
-		try:
-			t1=os.stat(dname).st_mtime
-		except OSError:
-			return None
-		for node in self.outputs:
-			orig=os.path.join(dname,node.name)
-			try:
-				shutil.copy2(orig,node.abspath())
-				os.utime(orig,None)
-			except(OSError,IOError):
-				Logs.debug('task: failed retrieving file')
-				return None
-		try:
-			t2=os.stat(dname).st_mtime
-		except OSError:
-			return None
-		if t1!=t2:
-			return None
-		for node in self.outputs:
-			node.sig=sig
-			if self.generator.bld.progress_bar<1:
-				self.generator.bld.to_log('restoring from cache %r\n'%node.abspath())
-		self.cached=True
-		return True
-	def put_files_cache(self):
-		if getattr(self,'cached',None):
-			return None
-		if not getattr(self,'outputs',None):
-			return None
-		sig=self.signature()
-		ssig=Utils.to_hex(self.uid())+Utils.to_hex(sig)
-		dname=os.path.join(self.generator.bld.cache_global,ssig)
-		tmpdir=tempfile.mkdtemp(prefix=self.generator.bld.cache_global+os.sep+'waf')
-		try:
-			shutil.rmtree(dname)
-		except:
-			pass
+if sys.hexversion>0x3000000:
+	def uid(self):
 		try:
-			for node in self.outputs:
-				dest=os.path.join(tmpdir,node.name)
-				shutil.copy2(node.abspath(),dest)
-		except(OSError,IOError):
-			try:
-				shutil.rmtree(tmpdir)
-			except:
-				pass
-		else:
-			try:
-				os.rename(tmpdir,dname)
-			except OSError:
-				try:
-					shutil.rmtree(tmpdir)
-				except:
-					pass
-			else:
-				try:
-					os.chmod(dname,Utils.O755)
-				except:
-					pass
+			return self.uid_
+		except AttributeError:
+			m=Utils.md5(self.__class__.__name__.encode('iso8859-1','xmlcharrefreplace'))
+			up=m.update
+			for x in self.inputs+self.outputs:
+				up(x.abspath().encode('iso8859-1','xmlcharrefreplace'))
+			self.uid_=m.digest()
+			return self.uid_
+	uid.__doc__=Task.uid.__doc__
+	Task.uid=uid
 def is_before(t1,t2):
 	to_list=Utils.to_list
 	for k in to_list(t2.ext_in):
@@ -521,109 +503,178 @@
 				b=i
 			else:
 				continue
+			aval=set(cstr_groups[keys[a]])
 			for x in cstr_groups[keys[b]]:
-				x.run_after.update(cstr_groups[keys[a]])
+				x.run_after.update(aval)
 def funex(c):
 	dc={}
 	exec(c,dc)
 	return dc['f']
-reg_act=re.compile(r"(?P<backslash>\\)|(?P<dollar>\$\$)|(?P<subst>\$\{(?P<var>\w+)(?P<code>.*?)\})",re.M)
+re_cond=re.compile('(?P<var>\w+)|(?P<or>\|)|(?P<and>&)')
+re_novar=re.compile(r'^(SRC|TGT)\W+.*?$')
+reg_act=re.compile(r'(?P<backslash>\\)|(?P<dollar>\$\$)|(?P<subst>\$\{(?P<var>\w+)(?P<code>.*?)\})',re.M)
 def compile_fun_shell(line):
 	extr=[]
 	def repl(match):
 		g=match.group
-		if g('dollar'):return"$"
-		elif g('backslash'):return'\\\\'
-		elif g('subst'):extr.append((g('var'),g('code')));return"%s"
+		if g('dollar'):
+			return"$"
+		elif g('backslash'):
+			return'\\\\'
+		elif g('subst'):
+			extr.append((g('var'),g('code')))
+			return"%s"
 		return None
 	line=reg_act.sub(repl,line)or line
+	def replc(m):
+		if m.group('and'):
+			return' and '
+		elif m.group('or'):
+			return' or '
+		else:
+			x=m.group('var')
+			if x not in dvars:
+				dvars.append(x)
+			return'env[%r]'%x
 	parm=[]
 	dvars=[]
 	app=parm.append
 	for(var,meth)in extr:
 		if var=='SRC':
 			if meth:app('tsk.inputs%s'%meth)
-			else:app('" ".join([a.path_from(bld.bldnode) for a in tsk.inputs])')
+			else:app('" ".join([a.path_from(cwdx) for a in tsk.inputs])')
 		elif var=='TGT':
 			if meth:app('tsk.outputs%s'%meth)
-			else:app('" ".join([a.path_from(bld.bldnode) for a in tsk.outputs])')
+			else:app('" ".join([a.path_from(cwdx) for a in tsk.outputs])')
 		elif meth:
 			if meth.startswith(':'):
+				if var not in dvars:
+					dvars.append(var)
 				m=meth[1:]
 				if m=='SRC':
-					m='[a.path_from(bld.bldnode) for a in tsk.inputs]'
+					m='[a.path_from(cwdx) for a in tsk.inputs]'
 				elif m=='TGT':
-					m='[a.path_from(bld.bldnode) for a in tsk.outputs]'
+					m='[a.path_from(cwdx) for a in tsk.outputs]'
+				elif re_novar.match(m):
+					m='[tsk.inputs%s]'%m[3:]
+				elif re_novar.match(m):
+					m='[tsk.outputs%s]'%m[3:]
 				elif m[:3]not in('tsk','gen','bld'):
-					dvars.extend([var,meth[1:]])
+					dvars.append(meth[1:])
 					m='%r'%m
 				app('" ".join(tsk.colon(%r, %s))'%(var,m))
+			elif meth.startswith('?'):
+				expr=re_cond.sub(replc,meth[1:])
+				app('p(%r) if (%s) else ""'%(var,expr))
 			else:
 				app('%s%s'%(var,meth))
 		else:
-			if not var in dvars:dvars.append(var)
+			if var not in dvars:
+				dvars.append(var)
 			app("p('%s')"%var)
 	if parm:parm="%% (%s) "%(',\n\t\t'.join(parm))
 	else:parm=''
 	c=COMPILE_TEMPLATE_SHELL%(line,parm)
-	Logs.debug('action: %s'%c)
+	Logs.debug('action: %s',c.strip().splitlines())
 	return(funex(c),dvars)
+reg_act_noshell=re.compile(r"(?P<space>\s+)|(?P<subst>\$\{(?P<var>\w+)(?P<code>.*?)\})|(?P<text>\S+)",re.M)
 def compile_fun_noshell(line):
-	extr=[]
-	def repl(match):
-		g=match.group
-		if g('dollar'):return"$"
-		elif g('subst'):extr.append((g('var'),g('code')));return"<<|@|>>"
-		return None
-	line2=reg_act.sub(repl,line)
-	params=line2.split('<<|@|>>')
-	assert(extr)
 	buf=[]
 	dvars=[]
+	merge=False
 	app=buf.append
-	for x in range(len(extr)):
-		params[x]=params[x].strip()
-		if params[x]:
-			app("lst.extend(%r)"%params[x].split())
-		(var,meth)=extr[x]
-		if var=='SRC':
-			if meth:app('lst.append(tsk.inputs%s)'%meth)
-			else:app("lst.extend([a.path_from(bld.bldnode) for a in tsk.inputs])")
-		elif var=='TGT':
-			if meth:app('lst.append(tsk.outputs%s)'%meth)
-			else:app("lst.extend([a.path_from(bld.bldnode) for a in tsk.outputs])")
-		elif meth:
-			if meth.startswith(':'):
-				m=meth[1:]
-				if m=='SRC':
-					m='[a.path_from(bld.bldnode) for a in tsk.inputs]'
-				elif m=='TGT':
-					m='[a.path_from(bld.bldnode) for a in tsk.outputs]'
-				elif m[:3]not in('tsk','gen','bld'):
-					dvars.extend([var,m])
-					m='%r'%m
-				app('lst.extend(tsk.colon(%r, %s))'%(var,m))
-			else:
-				app('lst.extend(gen.to_list(%s%s))'%(var,meth))
+	def replc(m):
+		if m.group('and'):
+			return' and '
+		elif m.group('or'):
+			return' or '
 		else:
-			app('lst.extend(to_list(env[%r]))'%var)
-			if not var in dvars:dvars.append(var)
-	if extr:
-		if params[-1]:
-			app("lst.extend(%r)"%params[-1].split())
+			x=m.group('var')
+			if x not in dvars:
+				dvars.append(x)
+			return'env[%r]'%x
+	for m in reg_act_noshell.finditer(line):
+		if m.group('space'):
+			merge=False
+			continue
+		elif m.group('text'):
+			app('[%r]'%m.group('text'))
+		elif m.group('subst'):
+			var=m.group('var')
+			code=m.group('code')
+			if var=='SRC':
+				if code:
+					app('[tsk.inputs%s]'%code)
+				else:
+					app('[a.path_from(cwdx) for a in tsk.inputs]')
+			elif var=='TGT':
+				if code:
+					app('[tsk.outputs%s]'%code)
+				else:
+					app('[a.path_from(cwdx) for a in tsk.outputs]')
+			elif code:
+				if code.startswith(':'):
+					if not var in dvars:
+						dvars.append(var)
+					m=code[1:]
+					if m=='SRC':
+						m='[a.path_from(cwdx) for a in tsk.inputs]'
+					elif m=='TGT':
+						m='[a.path_from(cwdx) for a in tsk.outputs]'
+					elif re_novar.match(m):
+						m='[tsk.inputs%s]'%m[3:]
+					elif re_novar.match(m):
+						m='[tsk.outputs%s]'%m[3:]
+					elif m[:3]not in('tsk','gen','bld'):
+						dvars.append(m)
+						m='%r'%m
+					app('tsk.colon(%r, %s)'%(var,m))
+				elif code.startswith('?'):
+					expr=re_cond.sub(replc,code[1:])
+					app('to_list(env[%r] if (%s) else [])'%(var,expr))
+				else:
+					app('gen.to_list(%s%s)'%(var,code))
+			else:
+				app('to_list(env[%r])'%var)
+				if not var in dvars:
+					dvars.append(var)
+		if merge:
+			tmp='merge(%s, %s)'%(buf[-2],buf[-1])
+			del buf[-1]
+			buf[-1]=tmp
+		merge=True
+	buf=['lst.extend(%s)'%x for x in buf]
 	fun=COMPILE_TEMPLATE_NOSHELL%"\n\t".join(buf)
-	Logs.debug('action: %s'%fun)
+	Logs.debug('action: %s',fun.strip().splitlines())
 	return(funex(fun),dvars)
 def compile_fun(line,shell=False):
-	if line.find('<')>0 or line.find('>')>0 or line.find('&&')>0:
-		shell=True
+	if isinstance(line,str):
+		if line.find('<')>0 or line.find('>')>0 or line.find('&&')>0:
+			shell=True
+	else:
+		dvars_lst=[]
+		funs_lst=[]
+		for x in line:
+			if isinstance(x,str):
+				fun,dvars=compile_fun(x,shell)
+				dvars_lst+=dvars
+				funs_lst.append(fun)
+			else:
+				funs_lst.append(x)
+		def composed_fun(task):
+			for x in funs_lst:
+				ret=x(task)
+				if ret:
+					return ret
+			return None
+		return composed_fun,dvars
 	if shell:
 		return compile_fun_shell(line)
 	else:
 		return compile_fun_noshell(line)
 def task_factory(name,func=None,vars=None,color='GREEN',ext_in=[],ext_out=[],before=[],after=[],shell=False,scan=None):
 	params={'vars':vars or[],'color':color,'name':name,'ext_in':Utils.to_list(ext_in),'ext_out':Utils.to_list(ext_out),'before':Utils.to_list(before),'after':Utils.to_list(after),'shell':shell,'scan':scan,}
-	if isinstance(func,str):
+	if isinstance(func,str)or isinstance(func,tuple):
 		params['run_str']=func
 	else:
 		params['run']=func
@@ -632,41 +683,8 @@
 	classes[name]=cls
 	return cls
 def always_run(cls):
-	old=cls.runnable_status
-	def always(self):
-		ret=old(self)
-		if ret==SKIP_ME:
-			ret=RUN_ME
-		return ret
-	cls.runnable_status=always
+	Logs.warn('This decorator is deprecated, set always_run on the task class instead!')
+	cls.always_run=True
 	return cls
 def update_outputs(cls):
-	old_post_run=cls.post_run
-	def post_run(self):
-		old_post_run(self)
-		for node in self.outputs:
-			node.sig=Utils.h_file(node.abspath())
-			self.generator.bld.task_sigs[node.abspath()]=self.uid()
-	cls.post_run=post_run
-	old_runnable_status=cls.runnable_status
-	def runnable_status(self):
-		status=old_runnable_status(self)
-		if status!=RUN_ME:
-			return status
-		try:
-			bld=self.generator.bld
-			prev_sig=bld.task_sigs[self.uid()]
-			if prev_sig==self.signature():
-				for x in self.outputs:
-					if not x.sig or bld.task_sigs[x.abspath()]!=self.uid():
-						return RUN_ME
-				return SKIP_ME
-		except KeyError:
-			pass
-		except IndexError:
-			pass
-		except AttributeError:
-			pass
-		return RUN_ME
-	cls.runnable_status=runnable_status
 	return cls
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ar.py glmark2-2021.02/waflib/Tools/ar.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ar.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/ar.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,13 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib.Configure import conf
+@conf
 def find_ar(conf):
 	conf.load('ar')
 def configure(conf):
 	conf.find_program('ar',var='AR')
-	conf.env.ARFLAGS='rcs'
-
-conf(find_ar)
\ No newline at end of file
+	conf.add_os_flags('ARFLAGS')
+	if not conf.env.ARFLAGS:
+		conf.env.ARFLAGS=['rcs']
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/asm.py glmark2-2021.02/waflib/Tools/asm.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/asm.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/asm.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,25 +1,23 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
-from waflib import Task,Utils
-import waflib.Task
+from waflib import Task
 from waflib.Tools.ccroot import link_task,stlink_task
-from waflib.TaskGen import extension,feature
+from waflib.TaskGen import extension
 class asm(Task.Task):
 	color='BLUE'
-	run_str='${AS} ${ASFLAGS} ${CPPPATH_ST:INCPATHS} ${AS_SRC_F}${SRC} ${AS_TGT_F}${TGT}'
+	run_str='${AS} ${ASFLAGS} ${ASMPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${AS_SRC_F}${SRC} ${AS_TGT_F}${TGT}'
+@extension('.s','.S','.asm','.ASM','.spp','.SPP')
 def asm_hook(self,node):
 	return self.create_compiled_task('asm',node)
 class asmprogram(link_task):
 	run_str='${ASLINK} ${ASLINKFLAGS} ${ASLNK_TGT_F}${TGT} ${ASLNK_SRC_F}${SRC}'
 	ext_out=['.bin']
 	inst_to='${BINDIR}'
-	chmod=Utils.O755
 class asmshlib(asmprogram):
 	inst_to='${LIBDIR}'
 class asmstlib(stlink_task):
 	pass
-
-extension('.s','.S','.asm','.ASM','.spp','.SPP')(asm_hook)
\ No newline at end of file
+def configure(conf):
+	conf.env.ASMPATH_ST='-I%s'
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/bison.py glmark2-2021.02/waflib/Tools/bison.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/bison.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/bison.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib import Task
 from waflib.TaskGen import extension
@@ -8,8 +8,9 @@
 	color='BLUE'
 	run_str='${BISON} ${BISONFLAGS} ${SRC[0].abspath()} -o ${TGT[0].name}'
 	ext_out=['.h']
+@extension('.y','.yc','.yy')
 def big_bison(self,node):
-	has_h='-d'in self.env['BISONFLAGS']
+	has_h='-d'in self.env.BISONFLAGS
 	outs=[]
 	if node.name.endswith('.yc'):
 		outs.append(node.change_ext('.tab.cc'))
@@ -20,10 +21,8 @@
 		if has_h:
 			outs.append(node.change_ext('.tab.h'))
 	tsk=self.create_task('bison',node,outs)
-	tsk.cwd=node.parent.get_bld().abspath()
+	tsk.cwd=node.parent.get_bld()
 	self.source.append(outs[0])
 def configure(conf):
 	conf.find_program('bison',var='BISON')
 	conf.env.BISONFLAGS=['-d']
-
-extension('.y','.yc','.yy')(big_bison)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_aliases.py glmark2-2021.02/waflib/Tools/c_aliases.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_aliases.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c_aliases.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,56 +1,60 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,re
-from waflib import Utils,Build
+from waflib import Utils,Errors
 from waflib.Configure import conf
 def get_extensions(lst):
 	ret=[]
 	for x in Utils.to_list(lst):
-		try:
-			if not isinstance(x,str):
-				x=x.name
-			ret.append(x[x.rfind('.')+1:])
-		except:
-			pass
+		if not isinstance(x,str):
+			x=x.name
+		ret.append(x[x.rfind('.')+1:])
 	return ret
 def sniff_features(**kw):
 	exts=get_extensions(kw['source'])
-	type=kw['_type']
+	typ=kw['typ']
 	feats=[]
-	if'cxx'in exts or'cpp'in exts or'c++'in exts or'cc'in exts or'C'in exts:
-		feats.append('cxx')
-	if'c'in exts or'vala'in exts:
+	for x in'cxx cpp c++ cc C'.split():
+		if x in exts:
+			feats.append('cxx')
+			break
+	if'c'in exts or'vala'in exts or'gs'in exts:
 		feats.append('c')
+	for x in'f f90 F F90 for FOR'.split():
+		if x in exts:
+			feats.append('fc')
+			break
 	if'd'in exts:
 		feats.append('d')
 	if'java'in exts:
 		feats.append('java')
-	if'java'in exts:
 		return'java'
-	if type in['program','shlib','stlib']:
+	if typ in('program','shlib','stlib'):
+		will_link=False
 		for x in feats:
-			if x in['cxx','d','c']:
-				feats.append(x+type)
+			if x in('cxx','d','fc','c'):
+				feats.append(x+typ)
+				will_link=True
+		if not will_link and not kw.get('features',[]):
+			raise Errors.WafError('Cannot link from %r, try passing eg: features="c cprogram"?'%kw)
 	return feats
-def set_features(kw,_type):
-	kw['_type']=_type
+def set_features(kw,typ):
+	kw['typ']=typ
 	kw['features']=Utils.to_list(kw.get('features',[]))+Utils.to_list(sniff_features(**kw))
+@conf
 def program(bld,*k,**kw):
 	set_features(kw,'program')
 	return bld(*k,**kw)
+@conf
 def shlib(bld,*k,**kw):
 	set_features(kw,'shlib')
 	return bld(*k,**kw)
+@conf
 def stlib(bld,*k,**kw):
 	set_features(kw,'stlib')
 	return bld(*k,**kw)
+@conf
 def objects(bld,*k,**kw):
 	set_features(kw,'objects')
 	return bld(*k,**kw)
-
-conf(program)
-conf(shlib)
-conf(stlib)
-conf(objects)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_config.py glmark2-2021.02/waflib/Tools/c_config.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_config.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c_config.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,108 +1,135 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import os,imp,sys,re,shlex,shutil
-from waflib import Build,Utils,Configure,Task,Options,Logs,TaskGen,Errors,ConfigSet,Runner
-from waflib.TaskGen import before_method,after_method,feature
+import os,re,shlex
+from waflib import Build,Utils,Task,Options,Logs,Errors,Runner
+from waflib.TaskGen import after_method,feature
 from waflib.Configure import conf
 WAF_CONFIG_H='config.h'
 DEFKEYS='define_key'
 INCKEYS='include_key'
 cfg_ver={'atleast-version':'>=','exact-version':'==','max-version':'<=',}
 SNIP_FUNCTION='''
-	int main() {
-	void *p;
-	p=(void*)(%s);
-	return 0;
+int main(int argc, char **argv) {
+	void (*p)();
+	(void)argc; (void)argv;
+	p=(void(*)())(%s);
+	return !p;
 }
 '''
 SNIP_TYPE='''
-int main() {
+int main(int argc, char **argv) {
+	(void)argc; (void)argv;
 	if ((%(type_name)s *) 0) return 0;
 	if (sizeof (%(type_name)s)) return 0;
-}
-'''
-SNIP_CLASS='''
-int main() {
-	if (
+	return 1;
 }
 '''
 SNIP_EMPTY_PROGRAM='''
-int main() {
+int main(int argc, char **argv) {
+	(void)argc; (void)argv;
 	return 0;
 }
 '''
 SNIP_FIELD='''
-int main() {
+int main(int argc, char **argv) {
 	char *off;
+	(void)argc; (void)argv;
 	off = (char*) &((%(type_name)s*)0)->%(field_name)s;
 	return (size_t) off < sizeof(%(type_name)s);
 }
 '''
-MACRO_TO_DESTOS={'__linux__':'linux','__GNU__':'gnu','__FreeBSD__':'freebsd','__NetBSD__':'netbsd','__OpenBSD__':'openbsd','__sun':'sunos','__hpux':'hpux','__sgi':'irix','_AIX':'aix','__CYGWIN__':'cygwin','__MSYS__':'msys','_UWIN':'uwin','_WIN64':'win32','_WIN32':'win32','__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__':'darwin','__ENVIRONMENT_IPHONE_OS_VERSION_MIN_REQUIRED__':'darwin','__QNX__':'qnx','__native_client__':'nacl'}
-MACRO_TO_DEST_CPU={'__x86_64__':'x86_64','__i386__':'x86','__ia64__':'ia','__mips__':'mips','__sparc__':'sparc','__alpha__':'alpha','__arm__':'arm','__hppa__':'hppa','__powerpc__':'powerpc',}
-def parse_flags(self,line,uselib,env=None,force_static=False):
+MACRO_TO_DESTOS={'__linux__':'linux','__GNU__':'gnu','__FreeBSD__':'freebsd','__NetBSD__':'netbsd','__OpenBSD__':'openbsd','__sun':'sunos','__hpux':'hpux','__sgi':'irix','_AIX':'aix','__CYGWIN__':'cygwin','__MSYS__':'cygwin','_UWIN':'uwin','_WIN64':'win32','_WIN32':'win32','__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__':'darwin','__ENVIRONMENT_IPHONE_OS_VERSION_MIN_REQUIRED__':'darwin','__QNX__':'qnx','__native_client__':'nacl'}
+MACRO_TO_DEST_CPU={'__x86_64__':'x86_64','__amd64__':'x86_64','__i386__':'x86','__ia64__':'ia','__mips__':'mips','__sparc__':'sparc','__alpha__':'alpha','__aarch64__':'aarch64','__thumb__':'thumb','__arm__':'arm','__hppa__':'hppa','__powerpc__':'powerpc','__ppc__':'powerpc','__convex__':'convex','__m68k__':'m68k','__s390x__':'s390x','__s390__':'s390','__sh__':'sh','__xtensa__':'xtensa',}
+@conf
+def parse_flags(self,line,uselib_store,env=None,force_static=False,posix=None):
 	assert(isinstance(line,str))
 	env=env or self.env
-	app=env.append_value
-	appu=env.append_unique
-	lex=shlex.shlex(line,posix=False)
+	if posix is None:
+		posix=True
+		if'\\'in line:
+			posix=('\\ 'in line)or('\\\\'in line)
+	lex=shlex.shlex(line,posix=posix)
 	lex.whitespace_split=True
 	lex.commenters=''
 	lst=list(lex)
+	uselib=uselib_store
+	def app(var,val):
+		env.append_value('%s_%s'%(var,uselib),val)
+	def appu(var,val):
+		env.append_unique('%s_%s'%(var,uselib),val)
+	static=False
 	while lst:
 		x=lst.pop(0)
 		st=x[:2]
 		ot=x[2:]
 		if st=='-I'or st=='/I':
-			if not ot:ot=lst.pop(0)
-			appu('INCLUDES_'+uselib,[ot])
-		elif st=='-include':
+			if not ot:
+				ot=lst.pop(0)
+			appu('INCLUDES',ot)
+		elif st=='-i':
 			tmp=[x,lst.pop(0)]
 			app('CFLAGS',tmp)
 			app('CXXFLAGS',tmp)
-		elif st=='-D'or(self.env.CXX_NAME=='msvc'and st=='/D'):
-			if not ot:ot=lst.pop(0)
-			app('DEFINES_'+uselib,[ot])
+		elif st=='-D'or(env.CXX_NAME=='msvc'and st=='/D'):
+			if not ot:
+				ot=lst.pop(0)
+			app('DEFINES',ot)
 		elif st=='-l':
-			if not ot:ot=lst.pop(0)
-			prefix=force_static and'STLIB_'or'LIB_'
-			appu(prefix+uselib,[ot])
+			if not ot:
+				ot=lst.pop(0)
+			prefix='STLIB'if(force_static or static)else'LIB'
+			app(prefix,ot)
 		elif st=='-L':
-			if not ot:ot=lst.pop(0)
-			appu('LIBPATH_'+uselib,[ot])
-		elif x=='-pthread'or x.startswith('+')or x.startswith('-std'):
-			app('CFLAGS_'+uselib,[x])
-			app('CXXFLAGS_'+uselib,[x])
-			app('LINKFLAGS_'+uselib,[x])
+			if not ot:
+				ot=lst.pop(0)
+			prefix='STLIBPATH'if(force_static or static)else'LIBPATH'
+			appu(prefix,ot)
+		elif x.startswith('/LIBPATH:'):
+			prefix='STLIBPATH'if(force_static or static)else'LIBPATH'
+			appu(prefix,x.replace('/LIBPATH:',''))
+		elif x.startswith('-std='):
+			prefix='CXXFLAGS'if'++'in x else'CFLAGS'
+			app(prefix,x)
+		elif x=='-pthread'or x.startswith('+'):
+			app('CFLAGS',x)
+			app('CXXFLAGS',x)
+			app('LINKFLAGS',x)
 		elif x=='-framework':
-			appu('FRAMEWORK_'+uselib,[lst.pop(0)])
+			appu('FRAMEWORK',lst.pop(0))
 		elif x.startswith('-F'):
-			appu('FRAMEWORKPATH_'+uselib,[x[2:]])
+			appu('FRAMEWORKPATH',x[2:])
+		elif x=='-Wl,-rpath'or x=='-Wl,-R':
+			app('RPATH',lst.pop(0).lstrip('-Wl,'))
+		elif x.startswith('-Wl,-R,'):
+			app('RPATH',x[7:])
+		elif x.startswith('-Wl,-R'):
+			app('RPATH',x[6:])
+		elif x.startswith('-Wl,-rpath,'):
+			app('RPATH',x[11:])
+		elif x=='-Wl,-Bstatic'or x=='-Bstatic':
+			static=True
+		elif x=='-Wl,-Bdynamic'or x=='-Bdynamic':
+			static=False
 		elif x.startswith('-Wl'):
-			app('LINKFLAGS_'+uselib,[x])
-		elif x.startswith('-m')or x.startswith('-f')or x.startswith('-dynamic'):
-			app('CFLAGS_'+uselib,[x])
-			app('CXXFLAGS_'+uselib,[x])
+			app('LINKFLAGS',x)
+		elif x.startswith(('-m','-f','-dynamic')):
+			app('CFLAGS',x)
+			app('CXXFLAGS',x)
 		elif x.startswith('-bundle'):
-			app('LINKFLAGS_'+uselib,[x])
-		elif x.startswith('-undefined'):
+			app('LINKFLAGS',x)
+		elif x.startswith(('-undefined','-Xlinker')):
 			arg=lst.pop(0)
-			app('LINKFLAGS_'+uselib,[x,arg])
-		elif x.startswith('-arch')or x.startswith('-isysroot'):
+			app('LINKFLAGS',[x,arg])
+		elif x.startswith(('-arch','-isysroot')):
 			tmp=[x,lst.pop(0)]
-			app('CFLAGS_'+uselib,tmp)
-			app('CXXFLAGS_'+uselib,tmp)
-			app('LINKFLAGS_'+uselib,tmp)
-		elif x.endswith('.a')or x.endswith('.so')or x.endswith('.dylib'):
-			appu('LINKFLAGS_'+uselib,[x])
-def ret_msg(self,f,kw):
-	if isinstance(f,str):
-		return f
-	return f(kw)
+			app('CFLAGS',tmp)
+			app('CXXFLAGS',tmp)
+			app('LINKFLAGS',tmp)
+		elif x.endswith(('.a','.so','.dylib','.lib')):
+			appu('LINKFLAGS',x)
+@conf
 def validate_cfg(self,kw):
 	if not'path'in kw:
 		if not self.env.PKGCONFIG:
@@ -119,67 +146,93 @@
 	if'modversion'in kw:
 		if not'msg'in kw:
 			kw['msg']='Checking for %r version'%kw['modversion']
+		if not'uselib_store'in kw:
+			kw['uselib_store']=kw['modversion']
+		if not'define_name'in kw:
+			kw['define_name']='%s_VERSION'%Utils.quote_define_name(kw['uselib_store'])
 		return
-	for x in cfg_ver.keys():
+	if not'package'in kw:
+		raise ValueError('a package name is required')
+	if not'uselib_store'in kw:
+		kw['uselib_store']=kw['package'].upper()
+	if not'define_name'in kw:
+		kw['define_name']=self.have_define(kw['uselib_store'])
+	if not'msg'in kw:
+		kw['msg']='Checking for %r'%(kw['package']or kw['path'])
+	for x in cfg_ver:
 		y=x.replace('-','_')
 		if y in kw:
-			if not'package'in kw:
-				raise ValueError('%s requires a package'%x)
+			package=kw['package']
+			if Logs.verbose:
+				Logs.warn('Passing %r to conf.check_cfg() is obsolete, pass parameters directly, eg:',y)
+				Logs.warn(" conf.check_cfg(package='%s', args=['--libs', '--cflags', '%s >= 1.6'])",package,package)
 			if not'msg'in kw:
-				kw['msg']='Checking for %r %s %s'%(kw['package'],cfg_ver[x],kw[y])
-			return
-	if not'msg'in kw:
-		kw['msg']='Checking for %r'%(kw['package']or kw['path'])
+				kw['msg']='Checking for %r %s %s'%(package,cfg_ver[x],kw[y])
+			break
+@conf
 def exec_cfg(self,kw):
+	path=Utils.to_list(kw['path'])
+	env=self.env.env or None
+	if kw.get('pkg_config_path'):
+		if not env:
+			env=dict(self.environ)
+		env['PKG_CONFIG_PATH']=kw['pkg_config_path']
+	def define_it():
+		define_name=kw['define_name']
+		if kw.get('global_define',1):
+			self.define(define_name,1,False)
+		else:
+			self.env.append_unique('DEFINES_%s'%kw['uselib_store'],"%s=1"%define_name)
+		if kw.get('add_have_to_env',1):
+			self.env[define_name]=1
 	if'atleast_pkgconfig_version'in kw:
-		cmd=[kw['path'],'--atleast-pkgconfig-version=%s'%kw['atleast_pkgconfig_version']]
-		self.cmd_and_log(cmd)
+		cmd=path+['--atleast-pkgconfig-version=%s'%kw['atleast_pkgconfig_version']]
+		self.cmd_and_log(cmd,env=env)
 		if not'okmsg'in kw:
 			kw['okmsg']='yes'
 		return
 	for x in cfg_ver:
 		y=x.replace('-','_')
 		if y in kw:
-			self.cmd_and_log([kw['path'],'--%s=%s'%(x,kw[y]),kw['package']])
+			self.cmd_and_log(path+['--%s=%s'%(x,kw[y]),kw['package']],env=env)
 			if not'okmsg'in kw:
 				kw['okmsg']='yes'
-			self.define(self.have_define(kw.get('uselib_store',kw['package'])),1,0)
+			define_it()
 			break
 	if'modversion'in kw:
-		version=self.cmd_and_log([kw['path'],'--modversion',kw['modversion']]).strip()
-		self.define('%s_VERSION'%Utils.quote_define_name(kw.get('uselib_store',kw['modversion'])),version)
+		version=self.cmd_and_log(path+['--modversion',kw['modversion']],env=env).strip()
+		self.define(kw['define_name'],version)
 		return version
-	lst=[kw['path']]
-	defi=kw.get('define_variable',None)
+	lst=[]+path
+	defi=kw.get('define_variable')
 	if not defi:
 		defi=self.env.PKG_CONFIG_DEFINES or{}
 	for key,val in defi.items():
 		lst.append('--define-variable=%s=%s'%(key,val))
-	if kw['package']:
-		lst.extend(Utils.to_list(kw['package']))
+	static=kw.get('force_static',False)
+	if'args'in kw:
+		args=Utils.to_list(kw['args'])
+		if'--static'in args or'--static-libs'in args:
+			static=True
+		lst+=args
+	lst.extend(Utils.to_list(kw['package']))
 	if'variables'in kw:
-		env=kw.get('env',self.env)
-		uselib=kw.get('uselib_store',kw['package'].upper())
+		v_env=kw.get('env',self.env)
 		vars=Utils.to_list(kw['variables'])
 		for v in vars:
-			val=self.cmd_and_log(lst+['--variable='+v]).strip()
-			var='%s_%s'%(uselib,v)
-			env[var]=val
+			val=self.cmd_and_log(lst+['--variable='+v],env=env).strip()
+			var='%s_%s'%(kw['uselib_store'],v)
+			v_env[var]=val
 		if not'okmsg'in kw:
 			kw['okmsg']='yes'
 		return
-	static=False
-	if'args'in kw:
-		args=Utils.to_list(kw['args'])
-		if'--static'in args or'--static-libs'in args:
-			static=True
-		lst+=args
-	ret=self.cmd_and_log(lst)
+	ret=self.cmd_and_log(lst,env=env)
 	if not'okmsg'in kw:
 		kw['okmsg']='yes'
-	self.define(self.have_define(kw.get('uselib_store',kw['package'])),1,0)
-	self.parse_flags(ret,kw.get('uselib_store',kw['package'].upper()),kw.get('env',self.env),force_static=static)
+	define_it()
+	self.parse_flags(ret,kw['uselib_store'],kw.get('env',self.env),force_static=static,posix=kw.get('posix'))
 	return ret
+@conf
 def check_cfg(self,*k,**kw):
 	if k:
 		lst=k[0].split()
@@ -187,34 +240,48 @@
 		kw['args']=' '.join(lst[1:])
 	self.validate_cfg(kw)
 	if'msg'in kw:
-		self.start_msg(kw['msg'])
+		self.start_msg(kw['msg'],**kw)
 	ret=None
 	try:
 		ret=self.exec_cfg(kw)
-	except self.errors.WafError ,e:
+	except self.errors.WafError:
 		if'errmsg'in kw:
-			self.end_msg(kw['errmsg'],'YELLOW')
+			self.end_msg(kw['errmsg'],'YELLOW',**kw)
 		if Logs.verbose>1:
 			raise
 		else:
 			self.fatal('The configuration failed')
 	else:
+		if not ret:
+			ret=True
 		kw['success']=ret
 		if'okmsg'in kw:
-			self.end_msg(self.ret_msg(kw['okmsg'],kw))
+			self.end_msg(self.ret_msg(kw['okmsg'],kw),**kw)
 	return ret
+def build_fun(bld):
+	if bld.kw['compile_filename']:
+		node=bld.srcnode.make_node(bld.kw['compile_filename'])
+		node.write(bld.kw['code'])
+	o=bld(features=bld.kw['features'],source=bld.kw['compile_filename'],target='testprog')
+	for k,v in bld.kw.items():
+		setattr(o,k,v)
+	if not bld.kw.get('quiet'):
+		bld.conf.to_log("==>\n%s\n<=="%bld.kw['code'])
+@conf
 def validate_c(self,kw):
+	if not'build_fun'in kw:
+		kw['build_fun']=build_fun
 	if not'env'in kw:
 		kw['env']=self.env.derive()
 	env=kw['env']
 	if not'compiler'in kw and not'features'in kw:
 		kw['compiler']='c'
-		if env['CXX_NAME']and Task.classes.get('cxx',None):
+		if env.CXX_NAME and Task.classes.get('cxx'):
 			kw['compiler']='cxx'
-			if not self.env['CXX']:
+			if not self.env.CXX:
 				self.fatal('a c++ compiler is required')
 		else:
-			if not self.env['CC']:
+			if not self.env.CC:
 				self.fatal('a c compiler is required')
 	if not'compile_mode'in kw:
 		kw['compile_mode']='c'
@@ -223,7 +290,10 @@
 	if not'type'in kw:
 		kw['type']='cprogram'
 	if not'features'in kw:
-		kw['features']=[kw['compile_mode'],kw['type']]
+		if not'header_name'in kw or kw.get('link_header_test',True):
+			kw['features']=[kw['compile_mode'],kw['type']]
+		else:
+			kw['features']=[kw['compile_mode']]
 	else:
 		kw['features']=Utils.to_list(kw['features'])
 	if not'compile_filename'in kw:
@@ -241,7 +311,7 @@
 			if not'header_name'in kw:
 				kw['header_name']=[]
 			fwk='%s/%s.h'%(fwkname,fwkname)
-			if kw.get('remove_dot_h',None):
+			if kw.get('remove_dot_h'):
 				fwk=fwk[:-2]
 			kw['header_name']=Utils.to_list(kw['header_name'])+[fwk]
 		kw['msg']='Checking for framework %s'%fwkname
@@ -276,7 +346,7 @@
 		if not'msg'in kw:
 			kw['msg']='Checking for header %s'%kw['header_name']
 		l=Utils.to_list(kw['header_name'])
-		assert len(l)>0,'list of headers in header_name is empty'
+		assert len(l),'list of headers in header_name is empty'
 		kw['code']=to_header(kw)+SNIP_EMPTY_PROGRAM
 		if not'uselib_store'in kw:
 			kw['uselib_store']=l[0].upper()
@@ -298,7 +368,7 @@
 			kw['msg']='Checking for code snippet'
 		if not'errmsg'in kw:
 			kw['errmsg']='no'
-	for(flagsname,flagstype)in[('cxxflags','compiler'),('cflags','compiler'),('linkflags','linker')]:
+	for(flagsname,flagstype)in(('cxxflags','compiler'),('cflags','compiler'),('linkflags','linker')):
 		if flagsname in kw:
 			if not'msg'in kw:
 				kw['msg']='Checking for %s flags %s'%(flagstype,kw[flagsname])
@@ -308,6 +378,7 @@
 		kw['execute']=False
 	if kw['execute']:
 		kw['features'].append('test_exec')
+		kw['chmod']=Utils.O755
 	if not'errmsg'in kw:
 		kw['errmsg']='not found'
 	if not'okmsg'in kw:
@@ -316,10 +387,15 @@
 		kw['code']=SNIP_EMPTY_PROGRAM
 	if self.env[INCKEYS]:
 		kw['code']='\n'.join(['#include <%s>'%x for x in self.env[INCKEYS]])+'\n'+kw['code']
+	if kw.get('merge_config_header',False)or env.merge_config_header:
+		kw['code']='%s\n\n%s'%(self.get_config_header(),kw['code'])
+		env.DEFINES=[]
 	if not kw.get('success'):kw['success']=None
 	if'define_name'in kw:
 		self.undefine(kw['define_name'])
-	assert'msg'in kw,'invalid parameters, read http://freehackers.org/~tnagy/wafbook/single.html#config_helpers_c'
+	if not'msg'in kw:
+		self.fatal('missing "msg" in conf.check(...)')
+@conf
 def post_check(self,*k,**kw):
 	is_success=0
 	if kw['execute']:
@@ -330,15 +406,31 @@
 				is_success=(kw['success']==0)
 	else:
 		is_success=(kw['success']==0)
-	if'define_name'in kw:
-		if'header_name'in kw or'function_name'in kw or'type_name'in kw or'fragment'in kw:
-			nm=kw['define_name']
-			if kw['execute']and kw.get('define_ret',None)and isinstance(is_success,str):
-				self.define(kw['define_name'],is_success,quote=kw.get('quote',1))
+	if kw.get('define_name'):
+		comment=kw.get('comment','')
+		define_name=kw['define_name']
+		if kw['execute']and kw.get('define_ret')and isinstance(is_success,str):
+			if kw.get('global_define',1):
+				self.define(define_name,is_success,quote=kw.get('quote',1),comment=comment)
 			else:
-				self.define_cond(kw['define_name'],is_success)
+				if kw.get('quote',1):
+					succ='"%s"'%is_success
+				else:
+					succ=int(is_success)
+				val='%s=%s'%(define_name,succ)
+				var='DEFINES_%s'%kw['uselib_store']
+				self.env.append_value(var,val)
 		else:
-			self.define_cond(kw['define_name'],is_success)
+			if kw.get('global_define',1):
+				self.define_cond(define_name,is_success,comment=comment)
+			else:
+				var='DEFINES_%s'%kw['uselib_store']
+				self.env.append_value(var,'%s=%s'%(define_name,int(is_success)))
+		if kw.get('add_have_to_env',1):
+			if kw.get('uselib_store'):
+				self.env[self.have_define(kw['uselib_store'])]=1
+			else:
+				self.env[define_name]=int(is_success)
 	if'header_name'in kw:
 		if kw.get('auto_add_header_name',False):
 			self.env.append_value(INCKEYS,Utils.to_list(kw['header_name']))
@@ -349,33 +441,31 @@
 			if x in ccroot.USELIB_VARS:
 				_vars|=ccroot.USELIB_VARS[x]
 		for k in _vars:
-			lk=k.lower()
-			if k=='INCLUDES':lk='includes'
-			if k=='DEFINES':lk='defines'
-			if lk in kw:
-				val=kw[lk]
-				if isinstance(val,str):
-					val=val.rstrip(os.path.sep)
-				self.env.append_unique(k+'_'+kw['uselib_store'],val)
+			x=k.lower()
+			if x in kw:
+				self.env.append_value(k+'_'+kw['uselib_store'],kw[x])
 	return is_success
+@conf
 def check(self,*k,**kw):
 	self.validate_c(kw)
-	self.start_msg(kw['msg'])
+	self.start_msg(kw['msg'],**kw)
 	ret=None
 	try:
-		ret=self.run_c_code(*k,**kw)
-	except self.errors.ConfigurationError ,e:
-		self.end_msg(kw['errmsg'],'YELLOW')
+		ret=self.run_build(*k,**kw)
+	except self.errors.ConfigurationError:
+		self.end_msg(kw['errmsg'],'YELLOW',**kw)
 		if Logs.verbose>1:
 			raise
 		else:
 			self.fatal('The configuration failed')
 	else:
 		kw['success']=ret
-		self.end_msg(self.ret_msg(kw['okmsg'],kw))
 	ret=self.post_check(*k,**kw)
 	if not ret:
+		self.end_msg(kw['errmsg'],'YELLOW',**kw)
 		self.fatal('The configuration failed %r'%ret)
+	else:
+		self.end_msg(self.ret_msg(kw['okmsg'],kw),**kw)
 	return ret
 class test_exec(Task.Task):
 	color='PINK'
@@ -394,80 +484,44 @@
 				self.generator.bld.retval=self.generator.bld.cmd_and_log([self.inputs[0].abspath()],env=env)
 			else:
 				self.generator.bld.retval=self.generator.bld.exec_command([self.inputs[0].abspath()],env=env)
+@feature('test_exec')
+@after_method('apply_link')
 def test_exec_fun(self):
 	self.create_task('test_exec',self.link_task.outputs[0])
-CACHE_RESULTS=1
-COMPILE_ERRORS=2
-def run_c_code(self,*k,**kw):
-	lst=[str(v)for(p,v)in kw.items()if p!='env']
-	h=Utils.h_list(lst)
-	dir=self.bldnode.abspath()+os.sep+(not Utils.is_win32 and'.'or'')+'conf_check_'+Utils.to_hex(h)
-	try:
-		os.makedirs(dir)
-	except:
-		pass
-	try:
-		os.stat(dir)
-	except:
-		self.fatal('cannot use the configuration test folder %r'%dir)
-	cachemode=getattr(Options.options,'confcache',None)
-	if cachemode==CACHE_RESULTS:
-		try:
-			proj=ConfigSet.ConfigSet(os.path.join(dir,'cache_run_c_code'))
-			ret=proj['cache_run_c_code']
-		except:
-			pass
-		else:
-			if isinstance(ret,str)and ret.startswith('Test does not build'):
-				self.fatal(ret)
-			return ret
-	bdir=os.path.join(dir,'testbuild')
-	if not os.path.exists(bdir):
-		os.makedirs(bdir)
-	self.test_bld=bld=Build.BuildContext(top_dir=dir,out_dir=bdir)
-	bld.init_dirs()
-	bld.progress_bar=0
-	bld.targets='*'
-	if kw['compile_filename']:
-		node=bld.srcnode.make_node(kw['compile_filename'])
-		node.write(kw['code'])
-	bld.logger=self.logger
-	bld.all_envs.update(self.all_envs)
-	bld.env=kw['env']
-	o=bld(features=kw['features'],source=kw['compile_filename'],target='testprog')
-	for k,v in kw.items():
-		setattr(o,k,v)
-	self.to_log("==>\n%s\n<=="%kw['code'])
-	bld.targets='*'
-	ret=-1
-	try:
-		try:
-			bld.compile()
-		except Errors.WafError:
-			ret='Test does not build: %s'%Utils.ex_stack()
-			self.fatal(ret)
-		else:
-			ret=getattr(bld,'retval',0)
-	finally:
-		proj=ConfigSet.ConfigSet()
-		proj['cache_run_c_code']=ret
-		proj.store(os.path.join(dir,'cache_run_c_code'))
-	return ret
+@conf
 def check_cxx(self,*k,**kw):
 	kw['compiler']='cxx'
 	return self.check(*k,**kw)
+@conf
 def check_cc(self,*k,**kw):
 	kw['compiler']='c'
 	return self.check(*k,**kw)
-def define(self,key,val,quote=True):
-	assert key and isinstance(key,str)
+@conf
+def set_define_comment(self,key,comment):
+	coms=self.env.DEFINE_COMMENTS
+	if not coms:
+		coms=self.env.DEFINE_COMMENTS={}
+	coms[key]=comment or''
+@conf
+def get_define_comment(self,key):
+	coms=self.env.DEFINE_COMMENTS or{}
+	return coms.get(key,'')
+@conf
+def define(self,key,val,quote=True,comment=''):
+	assert isinstance(key,str)
+	if not key:
+		return
+	if val is True:
+		val=1
+	elif val in(False,None):
+		val=0
 	if isinstance(val,int)or isinstance(val,float):
 		s='%s=%s'
 	else:
 		s=quote and'%s="%s"'or'%s=%s'
 	app=s%(key,str(val))
 	ban=key+'='
-	lst=self.env['DEFINES']
+	lst=self.env.DEFINES
 	for x in lst:
 		if x.startswith(ban):
 			lst[lst.index(x)]=app
@@ -475,102 +529,129 @@
 	else:
 		self.env.append_value('DEFINES',app)
 	self.env.append_unique(DEFKEYS,key)
-def undefine(self,key):
-	assert key and isinstance(key,str)
+	self.set_define_comment(key,comment)
+@conf
+def undefine(self,key,comment=''):
+	assert isinstance(key,str)
+	if not key:
+		return
 	ban=key+'='
-	lst=[x for x in self.env['DEFINES']if not x.startswith(ban)]
-	self.env['DEFINES']=lst
+	lst=[x for x in self.env.DEFINES if not x.startswith(ban)]
+	self.env.DEFINES=lst
 	self.env.append_unique(DEFKEYS,key)
-def define_cond(self,key,val):
-	assert key and isinstance(key,str)
+	self.set_define_comment(key,comment)
+@conf
+def define_cond(self,key,val,comment=''):
+	assert isinstance(key,str)
+	if not key:
+		return
 	if val:
-		self.define(key,1)
+		self.define(key,1,comment=comment)
 	else:
-		self.undefine(key)
+		self.undefine(key,comment=comment)
+@conf
 def is_defined(self,key):
 	assert key and isinstance(key,str)
 	ban=key+'='
-	for x in self.env['DEFINES']:
+	for x in self.env.DEFINES:
 		if x.startswith(ban):
 			return True
 	return False
+@conf
 def get_define(self,key):
 	assert key and isinstance(key,str)
 	ban=key+'='
-	for x in self.env['DEFINES']:
+	for x in self.env.DEFINES:
 		if x.startswith(ban):
 			return x[len(ban):]
 	return None
+@conf
 def have_define(self,key):
-	return self.__dict__.get('HAVE_PAT','HAVE_%s')%Utils.quote_define_name(key)
-def write_config_header(self,configfile='',guard='',top=False,env=None,defines=True,headers=False,remove=True):
+	return(self.env.HAVE_PAT or'HAVE_%s')%Utils.quote_define_name(key)
+@conf
+def write_config_header(self,configfile='',guard='',top=False,defines=True,headers=False,remove=True,define_prefix=''):
 	if not configfile:configfile=WAF_CONFIG_H
-	waf_guard=guard or'_%s_WAF'%Utils.quote_define_name(configfile)
+	waf_guard=guard or'W_%s_WAF'%Utils.quote_define_name(configfile)
 	node=top and self.bldnode or self.path.get_bld()
 	node=node.make_node(configfile)
 	node.parent.mkdir()
 	lst=['/* WARNING! All changes made to this file will be lost! */\n']
 	lst.append('#ifndef %s\n#define %s\n'%(waf_guard,waf_guard))
-	lst.append(self.get_config_header(defines,headers))
+	lst.append(self.get_config_header(defines,headers,define_prefix=define_prefix))
 	lst.append('\n#endif /* %s */\n'%waf_guard)
 	node.write('\n'.join(lst))
-	env=env or self.env
-	env.append_unique(Build.CFG_FILES,[node.abspath()])
+	self.env.append_unique(Build.CFG_FILES,[node.abspath()])
 	if remove:
 		for key in self.env[DEFKEYS]:
 			self.undefine(key)
 		self.env[DEFKEYS]=[]
-def get_config_header(self,defines=True,headers=False):
+@conf
+def get_config_header(self,defines=True,headers=False,define_prefix=''):
 	lst=[]
+	if self.env.WAF_CONFIG_H_PRELUDE:
+		lst.append(self.env.WAF_CONFIG_H_PRELUDE)
 	if headers:
 		for x in self.env[INCKEYS]:
 			lst.append('#include <%s>'%x)
 	if defines:
-		for x in self.env[DEFKEYS]:
-			if self.is_defined(x):
-				val=self.get_define(x)
-				lst.append('#define %s %s'%(x,val))
-			else:
-				lst.append('/* #undef %s */'%x)
+		tbl={}
+		for k in self.env.DEFINES:
+			a,_,b=k.partition('=')
+			tbl[a]=b
+		for k in self.env[DEFKEYS]:
+			caption=self.get_define_comment(k)
+			if caption:
+				caption=' /* %s */'%caption
+			try:
+				txt='#define %s%s %s%s'%(define_prefix,k,tbl[k],caption)
+			except KeyError:
+				txt='/* #undef %s%s */%s'%(define_prefix,k,caption)
+			lst.append(txt)
 	return"\n".join(lst)
+@conf
 def cc_add_flags(conf):
-	conf.add_os_flags('CPPFLAGS','CFLAGS')
-	conf.add_os_flags('CFLAGS')
+	conf.add_os_flags('CPPFLAGS',dup=False)
+	conf.add_os_flags('CFLAGS',dup=False)
+@conf
 def cxx_add_flags(conf):
-	conf.add_os_flags('CPPFLAGS','CXXFLAGS')
-	conf.add_os_flags('CXXFLAGS')
+	conf.add_os_flags('CPPFLAGS',dup=False)
+	conf.add_os_flags('CXXFLAGS',dup=False)
+@conf
 def link_add_flags(conf):
-	conf.add_os_flags('LINKFLAGS')
-	conf.add_os_flags('LDFLAGS','LINKFLAGS')
+	conf.add_os_flags('LINKFLAGS',dup=False)
+	conf.add_os_flags('LDFLAGS',dup=False)
+@conf
 def cc_load_tools(conf):
 	if not conf.env.DEST_OS:
 		conf.env.DEST_OS=Utils.unversioned_sys_platform()
 	conf.load('c')
+@conf
 def cxx_load_tools(conf):
 	if not conf.env.DEST_OS:
 		conf.env.DEST_OS=Utils.unversioned_sys_platform()
 	conf.load('cxx')
-def get_cc_version(conf,cc,gcc=False,icc=False):
+@conf
+def get_cc_version(conf,cc,gcc=False,icc=False,clang=False):
 	cmd=cc+['-dM','-E','-']
 	env=conf.env.env or None
 	try:
-		p=Utils.subprocess.Popen(cmd,stdin=Utils.subprocess.PIPE,stdout=Utils.subprocess.PIPE,stderr=Utils.subprocess.PIPE,env=env)
-		p.stdin.write('\n')
-		out=p.communicate()[0]
-	except:
+		out,err=conf.cmd_and_log(cmd,output=0,input='\n'.encode(),env=env)
+	except Exception:
 		conf.fatal('Could not determine the compiler version %r'%cmd)
-	if not isinstance(out,str):
-		out=out.decode(sys.stdout.encoding)
 	if gcc:
 		if out.find('__INTEL_COMPILER')>=0:
 			conf.fatal('The intel compiler pretends to be gcc')
-		if out.find('__GNUC__')<0:
+		if out.find('__GNUC__')<0 and out.find('__clang__')<0:
 			conf.fatal('Could not determine the compiler type')
 	if icc and out.find('__INTEL_COMPILER')<0:
 		conf.fatal('Not icc/icpc')
+	if clang and out.find('__clang__')<0:
+		conf.fatal('Not clang/clang++')
+	if not clang and out.find('__clang__')>=0:
+		conf.fatal('Could not find gcc/g++ (only Clang), if renamed try eg: CC=gcc48 CXX=g++48 waf configure')
 	k={}
-	if icc or gcc:
-		out=out.split('\n')
+	if icc or gcc or clang:
+		out=out.splitlines()
 		for line in out:
 			lst=shlex.split(line)
 			if len(lst)>2:
@@ -579,8 +660,6 @@
 				k[key]=val
 		def isD(var):
 			return var in k
-		def isT(var):
-			return var in k and k[var]!='0'
 		if not conf.env.DEST_OS:
 			conf.env.DEST_OS=''
 		for i in MACRO_TO_DESTOS:
@@ -594,9 +673,9 @@
 				conf.env.DEST_OS='generic'
 		if isD('__ELF__'):
 			conf.env.DEST_BINFMT='elf'
-		elif isD('__WINNT__')or isD('__CYGWIN__'):
+		elif isD('__WINNT__')or isD('__CYGWIN__')or isD('_WIN32'):
 			conf.env.DEST_BINFMT='pe'
-			conf.env.LIBDIR=conf.env['PREFIX']+'/bin'
+			conf.env.LIBDIR=conf.env.BINDIR
 		elif isD('__APPLE__'):
 			conf.env.DEST_BINFMT='mac-o'
 		if not conf.env.DEST_BINFMT:
@@ -608,31 +687,59 @@
 		Logs.debug('ccroot: dest platform: '+' '.join([conf.env[x]or'?'for x in('DEST_OS','DEST_BINFMT','DEST_CPU')]))
 		if icc:
 			ver=k['__INTEL_COMPILER']
-			conf.env['CC_VERSION']=(ver[:-2],ver[-2],ver[-1])
+			conf.env.CC_VERSION=(ver[:-2],ver[-2],ver[-1])
 		else:
-			conf.env['CC_VERSION']=(k['__GNUC__'],k['__GNUC_MINOR__'],k['__GNUC_PATCHLEVEL__'])
+			if isD('__clang__')and isD('__clang_major__'):
+				conf.env.CC_VERSION=(k['__clang_major__'],k['__clang_minor__'],k['__clang_patchlevel__'])
+			else:
+				conf.env.CC_VERSION=(k['__GNUC__'],k['__GNUC_MINOR__'],k.get('__GNUC_PATCHLEVEL__','0'))
 	return k
+@conf
 def get_xlc_version(conf,cc):
-	version_re=re.compile(r"IBM XL C/C\+\+.*, V(?P<major>\d*)\.(?P<minor>\d*)",re.I).search
 	cmd=cc+['-qversion']
 	try:
 		out,err=conf.cmd_and_log(cmd,output=0)
 	except Errors.WafError:
 		conf.fatal('Could not find xlc %r'%cmd)
-	if out:match=version_re(out)
-	else:match=version_re(err)
-	if not match:
+	for v in(r"IBM XL C/C\+\+.* V(?P<major>\d*)\.(?P<minor>\d*)",):
+		version_re=re.compile(v,re.I).search
+		match=version_re(out or err)
+		if match:
+			k=match.groupdict()
+			conf.env.CC_VERSION=(k['major'],k['minor'])
+			break
+	else:
 		conf.fatal('Could not determine the XLC version.')
-	k=match.groupdict()
-	conf.env['CC_VERSION']=(k['major'],k['minor'])
+@conf
+def get_suncc_version(conf,cc):
+	cmd=cc+['-V']
+	try:
+		out,err=conf.cmd_and_log(cmd,output=0)
+	except Errors.WafError as e:
+		if not(hasattr(e,'returncode')and hasattr(e,'stdout')and hasattr(e,'stderr')):
+			conf.fatal('Could not find suncc %r'%cmd)
+		out=e.stdout
+		err=e.stderr
+	version=(out or err)
+	version=version.splitlines()[0]
+	version_re=re.compile(r'cc: (studio.*?|\s+)?(sun\s+(c\+\+|c)|(WorkShop\s+Compilers))?\s+(?P<major>\d*)\.(?P<minor>\d*)',re.I).search
+	match=version_re(version)
+	if match:
+		k=match.groupdict()
+		conf.env.CC_VERSION=(k['major'],k['minor'])
+	else:
+		conf.fatal('Could not determine the suncc version.')
+@conf
 def add_as_needed(self):
 	if self.env.DEST_BINFMT=='elf'and'gcc'in(self.env.CXX_NAME,self.env.CC_NAME):
-		self.env.append_unique('LINKFLAGS','--as-needed')
+		self.env.append_unique('LINKFLAGS','-Wl,--as-needed')
 class cfgtask(Task.TaskBase):
 	def display(self):
 		return''
 	def runnable_status(self):
 		return Task.RUN_ME
+	def uid(self):
+		return Utils.SIG_NIL
 	def run(self):
 		conf=self.conf
 		bld=Build.BuildContext(top_dir=conf.srcnode.abspath(),out_dir=conf.bldnode.abspath())
@@ -640,26 +747,40 @@
 		bld.init_dirs()
 		bld.in_msg=1
 		bld.logger=self.logger
+		bld.multicheck_task=self
+		args=self.args
 		try:
-			bld.check(**self.args)
-		except:
+			if'func'in args:
+				bld.test(build_fun=args['func'],msg=args.get('msg',''),okmsg=args.get('okmsg',''),errmsg=args.get('errmsg',''),)
+			else:
+				args['multicheck_mandatory']=args.get('mandatory',True)
+				args['mandatory']=True
+				try:
+					bld.check(**args)
+				finally:
+					args['mandatory']=args['multicheck_mandatory']
+		except Exception:
 			return 1
+@conf
 def multicheck(self,*k,**kw):
-	self.start_msg(kw.get('msg','Executing %d configuration tests'%len(k)))
+	self.start_msg(kw.get('msg','Executing %d configuration tests'%len(k)),**kw)
+	for var in('DEFINES',DEFKEYS):
+		self.env.append_value(var,[])
+	self.env.DEFINE_COMMENTS=self.env.DEFINE_COMMENTS or{}
 	class par(object):
 		def __init__(self):
 			self.keep=False
-			self.cache_global=Options.cache_global
-			self.nocache=Options.options.nocache
-			self.returned_tasks=[]
+			self.task_sigs={}
+			self.progress_bar=0
 		def total(self):
 			return len(tasks)
 		def to_log(self,*k,**kw):
 			return
 	bld=par()
+	bld.keep=kw.get('run_all_tests',True)
 	tasks=[]
 	for dct in k:
-		x=cfgtask(bld=bld)
+		x=Task.classes['cfgtask'](bld=bld)
 		tasks.append(x)
 		x.args=dct
 		x.bld=bld
@@ -670,44 +791,35 @@
 		yield tasks
 		while 1:
 			yield[]
-	p=Runner.Parallel(bld,Options.options.jobs)
+	bld.producer=p=Runner.Parallel(bld,Options.options.jobs)
 	p.biter=it()
 	p.start()
 	for x in tasks:
 		x.logger.memhandler.flush()
+	if p.error:
+		for x in p.error:
+			if getattr(x,'err_msg',None):
+				self.to_log(x.err_msg)
+				self.end_msg('fail',color='RED')
+				raise Errors.WafError('There is an error in the library, read config.log for more information')
+	failure_count=0
+	for x in tasks:
+		if x.hasrun not in(Task.SUCCESS,Task.NOT_RUN):
+			failure_count+=1
+	if failure_count:
+		self.end_msg(kw.get('errmsg','%s test failed'%failure_count),color='YELLOW',**kw)
+	else:
+		self.end_msg('all ok',**kw)
+	for x in tasks:
+		if'msg'in x.args:
+			self.start_msg(x.args['msg'])
+			if x.hasrun==Task.NOT_RUN:
+				self.end_msg('test cancelled','YELLOW')
+			elif x.hasrun!=Task.SUCCESS:
+				self.end_msg(x.args.get('errmsg','no'),'YELLOW')
+			else:
+				self.end_msg(x.args.get('okmsg','yes'),'GREEN')
 	for x in tasks:
 		if x.hasrun!=Task.SUCCESS:
-			self.end_msg(kw.get('errmsg','no'),color='YELLOW')
-			self.fatal(kw.get('fatalmsg',None)or'One of the tests has failed, see the config.log for more information')
-	self.end_msg('ok')
-
-conf(parse_flags)
-conf(ret_msg)
-conf(validate_cfg)
-conf(exec_cfg)
-conf(check_cfg)
-conf(validate_c)
-conf(post_check)
-conf(check)
-feature('test_exec')(test_exec_fun)
-after_method('apply_link')(test_exec_fun)
-conf(run_c_code)
-conf(check_cxx)
-conf(check_cc)
-conf(define)
-conf(undefine)
-conf(define_cond)
-conf(is_defined)
-conf(get_define)
-conf(have_define)
-conf(write_config_header)
-conf(get_config_header)
-conf(cc_add_flags)
-conf(cxx_add_flags)
-conf(link_add_flags)
-conf(cc_load_tools)
-conf(cxx_load_tools)
-conf(get_cc_version)
-conf(get_xlc_version)
-conf(add_as_needed)
-conf(multicheck)
\ No newline at end of file
+			if x.args.get('mandatory',True):
+				self.fatal(kw.get('fatalmsg')or'One of the tests has failed, read config.log for more information')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ccroot.py glmark2-2021.02/waflib/Tools/ccroot.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ccroot.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/ccroot.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,28 +1,26 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import os,sys,re
-from waflib import TaskGen,Task,Utils,Logs,Build,Options,Node,Errors
-from waflib.Logs import error,debug,warn
+import os,re
+from waflib import Task,Utils,Node,Errors,Logs
 from waflib.TaskGen import after_method,before_method,feature,taskgen_method,extension
 from waflib.Tools import c_aliases,c_preproc,c_config,c_osx,c_tests
 from waflib.Configure import conf
+SYSTEM_LIB_PATHS=['/usr/lib64','/usr/lib','/usr/local/lib64','/usr/local/lib']
 USELIB_VARS=Utils.defaultdict(set)
 USELIB_VARS['c']=set(['INCLUDES','FRAMEWORKPATH','DEFINES','CPPFLAGS','CCDEPS','CFLAGS','ARCH'])
 USELIB_VARS['cxx']=set(['INCLUDES','FRAMEWORKPATH','DEFINES','CPPFLAGS','CXXDEPS','CXXFLAGS','ARCH'])
 USELIB_VARS['d']=set(['INCLUDES','DFLAGS'])
-USELIB_VARS['cprogram']=USELIB_VARS['cxxprogram']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS','FRAMEWORK','FRAMEWORKPATH','ARCH'])
-USELIB_VARS['cshlib']=USELIB_VARS['cxxshlib']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS','FRAMEWORK','FRAMEWORKPATH','ARCH'])
+USELIB_VARS['includes']=set(['INCLUDES','FRAMEWORKPATH','ARCH'])
+USELIB_VARS['cprogram']=USELIB_VARS['cxxprogram']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS','FRAMEWORK','FRAMEWORKPATH','ARCH','LDFLAGS'])
+USELIB_VARS['cshlib']=USELIB_VARS['cxxshlib']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS','FRAMEWORK','FRAMEWORKPATH','ARCH','LDFLAGS'])
 USELIB_VARS['cstlib']=USELIB_VARS['cxxstlib']=set(['ARFLAGS','LINKDEPS'])
 USELIB_VARS['dprogram']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS'])
 USELIB_VARS['dshlib']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS'])
 USELIB_VARS['dstlib']=set(['ARFLAGS','LINKDEPS'])
-USELIB_VARS['go']=set(['GOCFLAGS'])
-USELIB_VARS['goprogram']=set(['GOLFLAGS'])
 USELIB_VARS['asm']=set(['ASFLAGS'])
+@taskgen_method
 def create_compiled_task(self,name,node):
 	out='%s.%d.o'%(node.name,self.idx)
 	task=self.create_task(name,node,node.parent.find_or_declare(out))
@@ -31,6 +29,7 @@
 	except AttributeError:
 		self.compiled_tasks=[task]
 	return task
+@taskgen_method
 def to_incnodes(self,inlst):
 	lst=[]
 	seen=set([])
@@ -55,28 +54,71 @@
 				lst.append(p)
 				lst.append(v)
 	return lst
+@feature('c','cxx','d','asm','fc','includes')
+@after_method('propagate_uselib_vars','process_source')
 def apply_incpaths(self):
-	lst=self.to_incnodes(self.to_list(getattr(self,'includes',[]))+self.env['INCLUDES'])
+	lst=self.to_incnodes(self.to_list(getattr(self,'includes',[]))+self.env.INCLUDES)
 	self.includes_nodes=lst
-	self.env['INCPATHS']=[x.abspath()for x in lst]
+	cwd=self.get_cwd()
+	self.env.INCPATHS=[x.path_from(cwd)for x in lst]
 class link_task(Task.Task):
 	color='YELLOW'
 	inst_to=None
-	chmod=Utils.O644
+	chmod=Utils.O755
 	def add_target(self,target):
 		if isinstance(target,str):
+			base=self.generator.path
+			if target.startswith('#'):
+				target=target[1:]
+				base=self.generator.bld.bldnode
 			pattern=self.env[self.__class__.__name__+'_PATTERN']
 			if not pattern:
 				pattern='%s'
 			folder,name=os.path.split(target)
-			if self.__class__.__name__.find('shlib')>0:
-				if self.env.DEST_BINFMT=='pe'and getattr(self.generator,'vnum',None):
-					name=name+'-'+self.generator.vnum.split('.')[0]
-			tmp=folder+os.sep+pattern%name
-			target=self.generator.path.find_or_declare(tmp)
+			if self.__class__.__name__.find('shlib')>0 and getattr(self.generator,'vnum',None):
+				nums=self.generator.vnum.split('.')
+				if self.env.DEST_BINFMT=='pe':
+					name=name+'-'+nums[0]
+				elif self.env.DEST_OS=='openbsd':
+					pattern='%s.%s'%(pattern,nums[0])
+					if len(nums)>=2:
+						pattern+='.%s'%nums[1]
+			if folder:
+				tmp=folder+os.sep+pattern%name
+			else:
+				tmp=pattern%name
+			target=base.find_or_declare(tmp)
 		self.set_outputs(target)
+	def exec_command(self,*k,**kw):
+		ret=super(link_task,self).exec_command(*k,**kw)
+		if not ret and self.env.DO_MANIFEST:
+			ret=self.exec_mf()
+		return ret
+	def exec_mf(self):
+		if not self.env.MT:
+			return 0
+		manifest=None
+		for out_node in self.outputs:
+			if out_node.name.endswith('.manifest'):
+				manifest=out_node.abspath()
+				break
+		else:
+			return 0
+		mode=''
+		for x in Utils.to_list(self.generator.features):
+			if x in('cprogram','cxxprogram','fcprogram','fcprogram_test'):
+				mode=1
+			elif x in('cshlib','cxxshlib','fcshlib'):
+				mode=2
+		Logs.debug('msvc: embedding manifest in mode %r',mode)
+		lst=[]+self.env.MT
+		lst.extend(Utils.to_list(self.env.MTFLAGS))
+		lst.extend(['-manifest',manifest])
+		lst.append('-outputresource:%s;%s'%(self.outputs[0].abspath(),mode))
+		return super(link_task,self).exec_command(lst)
 class stlink_task(link_task):
 	run_str='${AR} ${ARFLAGS} ${AR_TGT_F}${TGT} ${AR_SRC_F}${SRC}'
+	chmod=Utils.O644
 def rm_tgt(cls):
 	old=cls.run
 	def wrap(self):
@@ -85,6 +127,8 @@
 		return old(self)
 	setattr(cls,'run',wrap)
 rm_tgt(stlink_task)
+@feature('c','cxx','d','fc','asm')
+@after_method('process_source')
 def apply_link(self):
 	for x in self.features:
 		if x=='cprogram'and'cxx'in self.features:
@@ -105,7 +149,8 @@
 	except AttributeError:
 		inst_to=self.link_task.__class__.inst_to
 	if inst_to:
-		self.install_task=self.bld.install_files(inst_to,self.link_task.outputs[:],env=self.env,chmod=self.link_task.chmod)
+		self.install_task=self.add_install_files(install_to=inst_to,install_from=self.link_task.outputs[:],chmod=self.link_task.chmod,task=self.link_task)
+@taskgen_method
 def use_rec(self,name,**kw):
 	if name in self.tmp_use_not or name in self.tmp_use_seen:
 		return
@@ -125,21 +170,26 @@
 		y.tmp_use_var=''
 	else:
 		objects=False
-		if not isinstance(y.link_task,stlink_task):
+		if not isinstance(link_task,stlink_task):
 			stlib=False
 			y.tmp_use_var='LIB'
 		else:
 			y.tmp_use_var='STLIB'
 	p=self.tmp_use_prec
 	for x in self.to_list(getattr(y,'use',[])):
+		if self.env["STLIB_"+x]:
+			continue
 		try:
 			p[x].append(name)
-		except:
+		except KeyError:
 			p[x]=[name]
 		self.use_rec(x,objects=objects,stlib=stlib)
+@feature('c','cxx','d','use','fc')
+@before_method('apply_incpaths','propagate_uselib_vars')
+@after_method('apply_link','process_source')
 def process_use(self):
 	use_not=self.tmp_use_not=set([])
-	use_seen=self.tmp_use_seen=[]
+	self.tmp_use_seen=[]
 	use_prec=self.tmp_use_prec={}
 	self.uselib=self.to_list(getattr(self,'uselib',[]))
 	self.includes=self.to_list(getattr(self,'includes',[]))
@@ -149,7 +199,7 @@
 	for x in use_not:
 		if x in use_prec:
 			del use_prec[x]
-	out=[]
+	out=self.tmp_use_sorted=[]
 	tmp=[]
 	for x in self.tmp_use_seen:
 		for k in use_prec.values():
@@ -180,26 +230,32 @@
 		y=self.bld.get_tgen_by_name(x)
 		var=y.tmp_use_var
 		if var and link_task:
-			if var=='LIB'or y.tmp_use_stlib:
+			if var=='LIB'or y.tmp_use_stlib or x in names:
 				self.env.append_value(var,[y.target[y.target.rfind(os.sep)+1:]])
 				self.link_task.dep_nodes.extend(y.link_task.outputs)
-				tmp_path=y.link_task.outputs[0].parent.path_from(self.bld.bldnode)
-				self.env.append_value(var+'PATH',[tmp_path])
+				tmp_path=y.link_task.outputs[0].parent.path_from(self.get_cwd())
+				self.env.append_unique(var+'PATH',[tmp_path])
 		else:
 			if y.tmp_use_objects:
 				self.add_objects_from_tgen(y)
 		if getattr(y,'export_includes',None):
 			self.includes.extend(y.to_incnodes(y.export_includes))
+		if getattr(y,'export_defines',None):
+			self.env.append_value('DEFINES',self.to_list(y.export_defines))
 	for x in names:
 		try:
 			y=self.bld.get_tgen_by_name(x)
-		except:
+		except Errors.WafError:
 			if not self.env['STLIB_'+x]and not x in self.uselib:
 				self.uselib.append(x)
 		else:
-			for k in self.to_list(getattr(y,'uselib',[])):
+			for k in self.to_list(getattr(y,'use',[])):
 				if not self.env['STLIB_'+k]and not k in self.uselib:
 					self.uselib.append(k)
+@taskgen_method
+def accept_node_to_link(self,node):
+	return not node.name.endswith('.pdb')
+@taskgen_method
 def add_objects_from_tgen(self,tg):
 	try:
 		link_task=self.link_task
@@ -208,27 +264,33 @@
 	else:
 		for tsk in getattr(tg,'compiled_tasks',[]):
 			for x in tsk.outputs:
-				if x.name.endswith('.o')or x.name.endswith('.obj'):
+				if self.accept_node_to_link(x):
 					link_task.inputs.append(x)
+@taskgen_method
 def get_uselib_vars(self):
 	_vars=set([])
 	for x in self.features:
 		if x in USELIB_VARS:
 			_vars|=USELIB_VARS[x]
 	return _vars
+@feature('c','cxx','d','fc','javac','cs','uselib','asm')
+@after_method('process_use')
 def propagate_uselib_vars(self):
 	_vars=self.get_uselib_vars()
 	env=self.env
-	for x in _vars:
-		y=x.lower()
-		env.append_unique(x,self.to_list(getattr(self,y,[])))
-	for x in self.features:
-		for var in _vars:
-			compvar='%s_%s'%(var,x)
-			env.append_value(var,env[compvar])
-	for x in self.to_list(getattr(self,'uselib',[])):
-		for v in _vars:
-			env.append_value(v,env[v+'_'+x])
+	app=env.append_value
+	feature_uselib=self.features+self.to_list(getattr(self,'uselib',[]))
+	for var in _vars:
+		y=var.lower()
+		val=getattr(self,y,[])
+		if val:
+			app(var,self.to_list(val))
+		for x in feature_uselib:
+			val=env['%s_%s'%(var,x)]
+			if val:
+				app(var,val)
+@feature('cshlib','cxxshlib','fcshlib')
+@after_method('apply_link')
 def apply_implib(self):
 	if not self.env.DEST_BINFMT=='pe':
 		return
@@ -237,58 +299,92 @@
 		name=self.target.name
 	else:
 		name=os.path.split(self.target)[1]
-	implib=self.env['implib_PATTERN']%name
+	implib=self.env.implib_PATTERN%name
 	implib=dll.parent.find_or_declare(implib)
-	self.env.append_value('LINKFLAGS',self.env['IMPLIB_ST']%implib.bldpath())
+	self.env.append_value('LINKFLAGS',self.env.IMPLIB_ST%implib.bldpath())
 	self.link_task.outputs.append(implib)
 	if getattr(self,'defs',None)and self.env.DEST_BINFMT=='pe':
 		node=self.path.find_resource(self.defs)
 		if not node:
 			raise Errors.WafError('invalid def file %r'%self.defs)
 		if'msvc'in(self.env.CC_NAME,self.env.CXX_NAME):
-			self.env.append_value('LINKFLAGS','/def:%s'%node.path_from(self.bld.bldnode))
+			self.env.append_value('LINKFLAGS','/def:%s'%node.path_from(self.get_cwd()))
 			self.link_task.dep_nodes.append(node)
 		else:
 			self.link_task.inputs.append(node)
-	try:
-		inst_to=self.install_path
-	except AttributeError:
-		inst_to=self.link_task.__class__.inst_to
-	if not inst_to:
-		return
-	self.implib_install_task=self.bld.install_as('${PREFIX}/lib/%s'%implib.name,implib,self.env)
+	if getattr(self,'install_task',None):
+		try:
+			inst_to=self.install_path_implib
+		except AttributeError:
+			try:
+				inst_to=self.install_path
+			except AttributeError:
+				inst_to='${IMPLIBDIR}'
+				self.install_task.install_to='${BINDIR}'
+				if not self.env.IMPLIBDIR:
+					self.env.IMPLIBDIR=self.env.LIBDIR
+		self.implib_install_task=self.add_install_files(install_to=inst_to,install_from=implib,chmod=self.link_task.chmod,task=self.link_task)
+re_vnum=re.compile('^([1-9]\\d*|0)([.]([1-9]\\d*|0)){0,2}?$')
+@feature('cshlib','cxxshlib','dshlib','fcshlib','vnum')
+@after_method('apply_link','propagate_uselib_vars')
 def apply_vnum(self):
 	if not getattr(self,'vnum','')or os.name!='posix'or self.env.DEST_BINFMT not in('elf','mac-o'):
 		return
 	link=self.link_task
+	if not re_vnum.match(self.vnum):
+		raise Errors.WafError('Invalid vnum %r for target %r'%(self.vnum,getattr(self,'name',self)))
 	nums=self.vnum.split('.')
 	node=link.outputs[0]
+	cnum=getattr(self,'cnum',str(nums[0]))
+	cnums=cnum.split('.')
+	if len(cnums)>len(nums)or nums[0:len(cnums)]!=cnums:
+		raise Errors.WafError('invalid compatibility version %s'%cnum)
 	libname=node.name
 	if libname.endswith('.dylib'):
 		name3=libname.replace('.dylib','.%s.dylib'%self.vnum)
-		name2=libname.replace('.dylib','.%s.dylib'%nums[0])
+		name2=libname.replace('.dylib','.%s.dylib'%cnum)
 	else:
 		name3=libname+'.'+self.vnum
-		name2=libname+'.'+nums[0]
+		name2=libname+'.'+cnum
 	if self.env.SONAME_ST:
 		v=self.env.SONAME_ST%name2
 		self.env.append_value('LINKFLAGS',v.split())
-	tsk=self.create_task('vnum',node,[node.parent.find_or_declare(name2),node.parent.find_or_declare(name3)])
-	if getattr(self.bld,'is_install',None):
+	if self.env.DEST_OS!='openbsd':
+		outs=[node.parent.make_node(name3)]
+		if name2!=name3:
+			outs.append(node.parent.make_node(name2))
+		self.create_task('vnum',node,outs)
+	if getattr(self,'install_task',None):
 		self.install_task.hasrun=Task.SKIP_ME
-		bld=self.bld
-		path=self.install_task.dest
-		t1=bld.install_as(path+os.sep+name3,node,env=self.env,chmod=self.link_task.chmod)
-		t2=bld.symlink_as(path+os.sep+name2,name3)
-		t3=bld.symlink_as(path+os.sep+libname,name3)
-		self.vnum_install_task=(t1,t2,t3)
-	if'-dynamiclib'in self.env['LINKFLAGS']and getattr(self,'install_task',None):
-		path=os.path.join(self.install_task.get_install_path(),self.link_task.outputs[0].name)
-		self.env.append_value('LINKFLAGS',['-install_name',path])
+		path=self.install_task.install_to
+		if self.env.DEST_OS=='openbsd':
+			libname=self.link_task.outputs[0].name
+			t1=self.add_install_as(install_to='%s/%s'%(path,libname),install_from=node,chmod=self.link_task.chmod)
+			self.vnum_install_task=(t1,)
+		else:
+			t1=self.add_install_as(install_to=path+os.sep+name3,install_from=node,chmod=self.link_task.chmod)
+			t3=self.add_symlink_as(install_to=path+os.sep+libname,install_from=name3)
+			if name2!=name3:
+				t2=self.add_symlink_as(install_to=path+os.sep+name2,install_from=name3)
+				self.vnum_install_task=(t1,t2,t3)
+			else:
+				self.vnum_install_task=(t1,t3)
+	if'-dynamiclib'in self.env.LINKFLAGS:
+		try:
+			inst_to=self.install_path
+		except AttributeError:
+			inst_to=self.link_task.__class__.inst_to
+		if inst_to:
+			p=Utils.subst_vars(inst_to,self.env)
+			path=os.path.join(p,name2)
+			self.env.append_value('LINKFLAGS',['-install_name',path])
+			self.env.append_value('LINKFLAGS','-Wl,-compatibility_version,%s'%cnum)
+			self.env.append_value('LINKFLAGS','-Wl,-current_version,%s'%self.vnum)
 class vnum(Task.Task):
 	color='CYAN'
-	quient=True
 	ext_in=['.bin']
+	def keyword(self):
+		return'Symlinking'
 	def run(self):
 		for x in self.outputs:
 			path=x.abspath()
@@ -305,26 +401,25 @@
 		for t in self.run_after:
 			if not t.hasrun:
 				return Task.ASK_LATER
-		for x in self.outputs:
-			x.sig=Utils.h_file(x.abspath())
 		return Task.SKIP_ME
 class fake_stlib(stlink_task):
 	def runnable_status(self):
 		for t in self.run_after:
 			if not t.hasrun:
 				return Task.ASK_LATER
-		for x in self.outputs:
-			x.sig=Utils.h_file(x.abspath())
 		return Task.SKIP_ME
-def read_shlib(self,name,paths=[]):
-	return self(name=name,features='fake_lib',lib_paths=paths,lib_type='shlib')
-def read_stlib(self,name,paths=[]):
-	return self(name=name,features='fake_lib',lib_paths=paths,lib_type='stlib')
-lib_patterns={'shlib':['lib%s.so','%s.so','lib%s.dll','%s.dll'],'stlib':['lib%s.a','%s.a','lib%s.dll','%s.dll','lib%s.lib','%s.lib'],}
+@conf
+def read_shlib(self,name,paths=[],export_includes=[],export_defines=[]):
+	return self(name=name,features='fake_lib',lib_paths=paths,lib_type='shlib',export_includes=export_includes,export_defines=export_defines)
+@conf
+def read_stlib(self,name,paths=[],export_includes=[],export_defines=[]):
+	return self(name=name,features='fake_lib',lib_paths=paths,lib_type='stlib',export_includes=export_includes,export_defines=export_defines)
+lib_patterns={'shlib':['lib%s.so','%s.so','lib%s.dylib','lib%s.dll','%s.dll'],'stlib':['lib%s.a','%s.a','lib%s.dll','%s.dll','lib%s.lib','%s.lib'],}
+@feature('fake_lib')
 def process_lib(self):
 	node=None
 	names=[x%self.name for x in lib_patterns[self.lib_type]]
-	for x in self.lib_paths+[self.path,'/usr/lib64','/usr/lib','/usr/local/lib64','/usr/local/lib']:
+	for x in self.lib_paths+[self.path]+SYSTEM_LIB_PATHS:
 		if not isinstance(x,Node.Node):
 			x=self.bld.root.find_node(x)or self.path.find_node(x)
 			if not x:
@@ -332,7 +427,10 @@
 		for y in names:
 			node=x.find_node(y)
 			if node:
-				node.sig=Utils.h_file(node.abspath())
+				try:
+					Utils.h_file(node.abspath())
+				except EnvironmentError:
+					raise ValueError('Could not read %r'%y)
 				break
 		else:
 			continue
@@ -344,32 +442,35 @@
 class fake_o(Task.Task):
 	def runnable_status(self):
 		return Task.SKIP_ME
+@extension('.o','.obj')
 def add_those_o_files(self,node):
 	tsk=self.create_task('fake_o',[],node)
 	try:
 		self.compiled_tasks.append(tsk)
 	except AttributeError:
 		self.compiled_tasks=[tsk]
-
-taskgen_method(create_compiled_task)
-taskgen_method(to_incnodes)
-feature('c','cxx','d','go','asm','fc','includes')(apply_incpaths)
-after_method('propagate_uselib_vars','process_source')(apply_incpaths)
-feature('c','cxx','d','go','fc','asm')(apply_link)
-after_method('process_source')(apply_link)
-taskgen_method(use_rec)
-feature('c','cxx','d','use','fc')(process_use)
-before_method('apply_incpaths','propagate_uselib_vars')(process_use)
-after_method('apply_link','process_source')(process_use)
-taskgen_method(add_objects_from_tgen)
-taskgen_method(get_uselib_vars)
-feature('c','cxx','d','fc','javac','cs','uselib')(propagate_uselib_vars)
-after_method('process_use')(propagate_uselib_vars)
-feature('cshlib','cxxshlib','fcshlib')(apply_implib)
-after_method('apply_link')(apply_implib)
-feature('cshlib','cxxshlib','dshlib','fcshlib','vnum')(apply_vnum)
-after_method('apply_link')(apply_vnum)
-conf(read_shlib)
-conf(read_stlib)
-feature('fake_lib')(process_lib)
-extension('.o','.obj')(add_those_o_files)
\ No newline at end of file
+@feature('fake_obj')
+@before_method('process_source')
+def process_objs(self):
+	for node in self.to_nodes(self.source):
+		self.add_those_o_files(node)
+	self.source=[]
+@conf
+def read_object(self,obj):
+	if not isinstance(obj,self.path.__class__):
+		obj=self.path.find_resource(obj)
+	return self(features='fake_obj',source=obj,name=obj.name)
+@feature('cxxprogram','cprogram')
+@after_method('apply_link','process_use')
+def set_full_paths_hpux(self):
+	if self.env.DEST_OS!='hp-ux':
+		return
+	base=self.bld.bldnode.abspath()
+	for var in['LIBPATH','STLIBPATH']:
+		lst=[]
+		for x in self.env[var]:
+			if x.startswith('/'):
+				lst.append(x)
+			else:
+				lst.append(os.path.normpath(os.path.join(base,x)))
+		self.env[var]=lst
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/clang.py glmark2-2021.02/waflib/Tools/clang.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/clang.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/clang.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,20 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+from waflib.Tools import ccroot,ar,gcc
+from waflib.Configure import conf
+@conf
+def find_clang(conf):
+	cc=conf.find_program('clang',var='CC')
+	conf.get_cc_version(cc,clang=True)
+	conf.env.CC_NAME='clang'
+def configure(conf):
+	conf.find_clang()
+	conf.find_program(['llvm-ar','ar'],var='AR')
+	conf.find_ar()
+	conf.gcc_common_flags()
+	conf.gcc_modifier_platform()
+	conf.cc_load_tools()
+	conf.cc_add_flags()
+	conf.link_add_flags()
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/clangxx.py glmark2-2021.02/waflib/Tools/clangxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/clangxx.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/clangxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,20 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+from waflib.Tools import ccroot,ar,gxx
+from waflib.Configure import conf
+@conf
+def find_clangxx(conf):
+	cxx=conf.find_program('clang++',var='CXX')
+	conf.get_cc_version(cxx,clang=True)
+	conf.env.CXX_NAME='clang'
+def configure(conf):
+	conf.find_clangxx()
+	conf.find_program(['llvm-ar','ar'],var='AR')
+	conf.find_ar()
+	conf.gxx_common_flags()
+	conf.gxx_modifier_platform()
+	conf.cxx_load_tools()
+	conf.cxx_add_flags()
+	conf.link_add_flags()
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_c.py glmark2-2021.02/waflib/Tools/compiler_c.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_c.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/compiler_c.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,39 +1,44 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,imp,types
+import re
 from waflib.Tools import ccroot
-from waflib import Utils,Configure
+from waflib import Utils
 from waflib.Logs import debug
-c_compiler={'win32':['msvc','gcc'],'cygwin':['gcc'],'darwin':['gcc'],'aix':['xlc','gcc'],'linux':['gcc','icc'],'sunos':['suncc','gcc'],'irix':['gcc','irixcc'],'hpux':['gcc'],'gnu':['gcc'],'java':['gcc','msvc','icc'],'default':['gcc'],}
+c_compiler={'win32':['msvc','gcc','clang'],'cygwin':['gcc'],'darwin':['clang','gcc'],'aix':['xlc','gcc','clang'],'linux':['gcc','clang','icc'],'sunos':['suncc','gcc'],'irix':['gcc','irixcc'],'hpux':['gcc'],'osf1V':['gcc'],'gnu':['gcc','clang'],'java':['gcc','msvc','clang','icc'],'default':['clang','gcc'],}
+def default_compilers():
+	build_platform=Utils.unversioned_sys_platform()
+	possible_compiler_list=c_compiler.get(build_platform,c_compiler['default'])
+	return' '.join(possible_compiler_list)
 def configure(conf):
-	try:test_for_compiler=conf.options.check_c_compiler
-	except AttributeError:conf.fatal("Add options(opt): opt.load('compiler_c')")
-	for compiler in test_for_compiler.split():
+	try:
+		test_for_compiler=conf.options.check_c_compiler or default_compilers()
+	except AttributeError:
+		conf.fatal("Add options(opt): opt.load('compiler_c')")
+	for compiler in re.split('[ ,]+',test_for_compiler):
 		conf.env.stash()
-		conf.start_msg('Checking for %r (c compiler)'%compiler)
+		conf.start_msg('Checking for %r (C compiler)'%compiler)
 		try:
 			conf.load(compiler)
-		except conf.errors.ConfigurationError ,e:
+		except conf.errors.ConfigurationError as e:
 			conf.env.revert()
 			conf.end_msg(False)
-			debug('compiler_c: %r'%e)
+			debug('compiler_c: %r',e)
 		else:
-			if conf.env['CC']:
+			if conf.env.CC:
 				conf.end_msg(conf.env.get_flat('CC'))
-				conf.env['COMPILER_CC']=compiler
+				conf.env.COMPILER_CC=compiler
+				conf.env.commit()
 				break
+			conf.env.revert()
 			conf.end_msg(False)
 	else:
-		conf.fatal('could not configure a c compiler!')
+		conf.fatal('could not configure a C compiler!')
 def options(opt):
+	test_for_compiler=default_compilers()
 	opt.load_special_tools('c_*.py',ban=['c_dumbpreproc.py'])
-	global c_compiler
-	build_platform=Utils.unversioned_sys_platform()
-	possible_compiler_list=c_compiler[build_platform in c_compiler and build_platform or'default']
-	test_for_compiler=' '.join(possible_compiler_list)
-	cc_compiler_opts=opt.add_option_group("C Compiler Options")
-	cc_compiler_opts.add_option('--check-c-compiler',default="%s"%test_for_compiler,help='On this platform (%s) the following C-Compiler will be checked by default: "%s"'%(build_platform,test_for_compiler),dest="check_c_compiler")
+	cc_compiler_opts=opt.add_option_group('Configuration options')
+	cc_compiler_opts.add_option('--check-c-compiler',default=None,help='list of C compilers to try [%s]'%test_for_compiler,dest="check_c_compiler")
 	for x in test_for_compiler.split():
 		opt.load('%s'%x)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_cxx.py glmark2-2021.02/waflib/Tools/compiler_cxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_cxx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/compiler_cxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,39 +1,44 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,imp,types
+import re
 from waflib.Tools import ccroot
-from waflib import Utils,Configure
+from waflib import Utils
 from waflib.Logs import debug
-cxx_compiler={'win32':['msvc','g++'],'cygwin':['g++'],'darwin':['g++'],'aix':['xlc++','g++'],'linux':['g++','icpc'],'sunos':['sunc++','g++'],'irix':['g++'],'hpux':['g++'],'gnu':['g++'],'java':['g++','msvc','icpc'],'default':['g++']}
+cxx_compiler={'win32':['msvc','g++','clang++'],'cygwin':['g++'],'darwin':['clang++','g++'],'aix':['xlc++','g++','clang++'],'linux':['g++','clang++','icpc'],'sunos':['sunc++','g++'],'irix':['g++'],'hpux':['g++'],'osf1V':['g++'],'gnu':['g++','clang++'],'java':['g++','msvc','clang++','icpc'],'default':['clang++','g++']}
+def default_compilers():
+	build_platform=Utils.unversioned_sys_platform()
+	possible_compiler_list=cxx_compiler.get(build_platform,cxx_compiler['default'])
+	return' '.join(possible_compiler_list)
 def configure(conf):
-	try:test_for_compiler=conf.options.check_cxx_compiler
-	except AttributeError:conf.fatal("Add options(opt): opt.load('compiler_cxx')")
-	for compiler in test_for_compiler.split():
+	try:
+		test_for_compiler=conf.options.check_cxx_compiler or default_compilers()
+	except AttributeError:
+		conf.fatal("Add options(opt): opt.load('compiler_cxx')")
+	for compiler in re.split('[ ,]+',test_for_compiler):
 		conf.env.stash()
-		conf.start_msg('Checking for %r (c++ compiler)'%compiler)
+		conf.start_msg('Checking for %r (C++ compiler)'%compiler)
 		try:
 			conf.load(compiler)
-		except conf.errors.ConfigurationError ,e:
+		except conf.errors.ConfigurationError as e:
 			conf.env.revert()
 			conf.end_msg(False)
-			debug('compiler_cxx: %r'%e)
+			debug('compiler_cxx: %r',e)
 		else:
-			if conf.env['CXX']:
+			if conf.env.CXX:
 				conf.end_msg(conf.env.get_flat('CXX'))
-				conf.env['COMPILER_CXX']=compiler
+				conf.env.COMPILER_CXX=compiler
+				conf.env.commit()
 				break
+			conf.env.revert()
 			conf.end_msg(False)
 	else:
-		conf.fatal('could not configure a c++ compiler!')
+		conf.fatal('could not configure a C++ compiler!')
 def options(opt):
+	test_for_compiler=default_compilers()
 	opt.load_special_tools('cxx_*.py')
-	global cxx_compiler
-	build_platform=Utils.unversioned_sys_platform()
-	possible_compiler_list=cxx_compiler[build_platform in cxx_compiler and build_platform or'default']
-	test_for_compiler=' '.join(possible_compiler_list)
-	cxx_compiler_opts=opt.add_option_group('C++ Compiler Options')
-	cxx_compiler_opts.add_option('--check-cxx-compiler',default="%s"%test_for_compiler,help='On this platform (%s) the following C++ Compiler will be checked by default: "%s"'%(build_platform,test_for_compiler),dest="check_cxx_compiler")
+	cxx_compiler_opts=opt.add_option_group('Configuration options')
+	cxx_compiler_opts.add_option('--check-cxx-compiler',default=None,help='list of C++ compilers to try [%s]'%test_for_compiler,dest="check_cxx_compiler")
 	for x in test_for_compiler.split():
 		opt.load('%s'%x)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_d.py glmark2-2021.02/waflib/Tools/compiler_d.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_d.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/compiler_d.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,30 +1,41 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,imp,types
-from waflib import Utils,Configure,Options,Logs
+import re
+from waflib import Utils,Logs
+d_compiler={'default':['gdc','dmd','ldc2']}
+def default_compilers():
+	build_platform=Utils.unversioned_sys_platform()
+	possible_compiler_list=d_compiler.get(build_platform,d_compiler['default'])
+	return' '.join(possible_compiler_list)
 def configure(conf):
-	for compiler in conf.options.dcheck.split(','):
+	try:
+		test_for_compiler=conf.options.check_d_compiler or default_compilers()
+	except AttributeError:
+		conf.fatal("Add options(opt): opt.load('compiler_d')")
+	for compiler in re.split('[ ,]+',test_for_compiler):
 		conf.env.stash()
-		conf.start_msg('Checking for %r (d compiler)'%compiler)
+		conf.start_msg('Checking for %r (D compiler)'%compiler)
 		try:
 			conf.load(compiler)
-		except conf.errors.ConfigurationError ,e:
+		except conf.errors.ConfigurationError as e:
 			conf.env.revert()
 			conf.end_msg(False)
-			Logs.debug('compiler_cxx: %r'%e)
+			Logs.debug('compiler_d: %r',e)
 		else:
 			if conf.env.D:
 				conf.end_msg(conf.env.get_flat('D'))
-				conf.env['COMPILER_D']=compiler
-				conf.env.D_COMPILER=conf.env.D
+				conf.env.COMPILER_D=compiler
+				conf.env.commit()
 				break
+			conf.env.revert()
 			conf.end_msg(False)
 	else:
-		conf.fatal('no suitable d compiler was found')
+		conf.fatal('could not configure a D compiler!')
 def options(opt):
-	d_compiler_opts=opt.add_option_group('D Compiler Options')
-	d_compiler_opts.add_option('--check-d-compiler',default='gdc,dmd',action='store',help='check for the compiler [Default:gdc,dmd]',dest='dcheck')
-	for d_compiler in['gdc','dmd']:
-		opt.load('%s'%d_compiler)
+	test_for_compiler=default_compilers()
+	d_compiler_opts=opt.add_option_group('Configuration options')
+	d_compiler_opts.add_option('--check-d-compiler',default=None,help='list of D compilers to try [%s]'%test_for_compiler,dest='check_d_compiler')
+	for x in test_for_compiler.split():
+		opt.load('%s'%x)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_fc.py glmark2-2021.02/waflib/Tools/compiler_fc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/compiler_fc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/compiler_fc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,43 +1,43 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,imp,types
-from waflib import Utils,Configure,Options,Logs,Errors
+import re
+from waflib import Utils,Logs
 from waflib.Tools import fc
 fc_compiler={'win32':['gfortran','ifort'],'darwin':['gfortran','g95','ifort'],'linux':['gfortran','g95','ifort'],'java':['gfortran','g95','ifort'],'default':['gfortran'],'aix':['gfortran']}
-def __list_possible_compiler(platform):
-	try:
-		return fc_compiler[platform]
-	except KeyError:
-		return fc_compiler["default"]
+def default_compilers():
+	build_platform=Utils.unversioned_sys_platform()
+	possible_compiler_list=fc_compiler.get(build_platform,fc_compiler['default'])
+	return' '.join(possible_compiler_list)
 def configure(conf):
-	try:test_for_compiler=conf.options.check_fc
-	except AttributeError:conf.fatal("Add options(opt): opt.load('compiler_fc')")
-	for compiler in test_for_compiler.split():
+	try:
+		test_for_compiler=conf.options.check_fortran_compiler or default_compilers()
+	except AttributeError:
+		conf.fatal("Add options(opt): opt.load('compiler_fc')")
+	for compiler in re.split('[ ,]+',test_for_compiler):
 		conf.env.stash()
-		conf.start_msg('Checking for %r (fortran compiler)'%compiler)
+		conf.start_msg('Checking for %r (Fortran compiler)'%compiler)
 		try:
 			conf.load(compiler)
-		except conf.errors.ConfigurationError ,e:
+		except conf.errors.ConfigurationError as e:
 			conf.env.revert()
 			conf.end_msg(False)
-			Logs.debug('compiler_fortran: %r'%e)
+			Logs.debug('compiler_fortran: %r',e)
 		else:
-			if conf.env['FC']:
+			if conf.env.FC:
 				conf.end_msg(conf.env.get_flat('FC'))
 				conf.env.COMPILER_FORTRAN=compiler
+				conf.env.commit()
 				break
+			conf.env.revert()
 			conf.end_msg(False)
 	else:
-		conf.fatal('could not configure a fortran compiler!')
+		conf.fatal('could not configure a Fortran compiler!')
 def options(opt):
+	test_for_compiler=default_compilers()
 	opt.load_special_tools('fc_*.py')
-	build_platform=Utils.unversioned_sys_platform()
-	detected_platform=Options.platform
-	possible_compiler_list=__list_possible_compiler(detected_platform)
-	test_for_compiler=' '.join(possible_compiler_list)
-	fortran_compiler_opts=opt.add_option_group("Fortran Compiler Options")
-	fortran_compiler_opts.add_option('--check-fortran-compiler',default="%s"%test_for_compiler,help='On this platform (%s) the following Fortran Compiler will be checked by default: "%s"'%(detected_platform,test_for_compiler),dest="check_fc")
-	for compiler in test_for_compiler.split():
-		opt.load('%s'%compiler)
+	fortran_compiler_opts=opt.add_option_group('Configuration options')
+	fortran_compiler_opts.add_option('--check-fortran-compiler',default=None,help='list of Fortran compiler to try [%s]'%test_for_compiler,dest="check_fortran_compiler")
+	for x in test_for_compiler.split():
+		opt.load('%s'%x)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_osx.py glmark2-2021.02/waflib/Tools/c_osx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_osx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c_osx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,9 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,shutil,sys,platform
-from waflib import TaskGen,Task,Build,Options,Utils,Errors
+import os,shutil,platform
+from waflib import Task,Utils
 from waflib.TaskGen import taskgen_method,feature,after_method,before_method
 app_info='''
 <?xml version="1.0" encoding="UTF-8"?>
@@ -19,18 +19,19 @@
 	<key>NOTE</key>
 	<string>THIS IS A GENERATED FILE, DO NOT MODIFY</string>
 	<key>CFBundleExecutable</key>
-	<string>%s</string>
+	<string>{app_name}</string>
 </dict>
 </plist>
 '''
+@feature('c','cxx')
 def set_macosx_deployment_target(self):
-	if self.env['MACOSX_DEPLOYMENT_TARGET']:
-		os.environ['MACOSX_DEPLOYMENT_TARGET']=self.env['MACOSX_DEPLOYMENT_TARGET']
+	if self.env.MACOSX_DEPLOYMENT_TARGET:
+		os.environ['MACOSX_DEPLOYMENT_TARGET']=self.env.MACOSX_DEPLOYMENT_TARGET
 	elif'MACOSX_DEPLOYMENT_TARGET'not in os.environ:
 		if Utils.unversioned_sys_platform()=='darwin':
 			os.environ['MACOSX_DEPLOYMENT_TARGET']='.'.join(platform.mac_ver()[0].split('.')[:2])
+@taskgen_method
 def create_bundle_dirs(self,name,out):
-	bld=self.bld
 	dir=out.parent.find_or_declare(name)
 	dir.mkdir()
 	macos=dir.find_or_declare(['Contents','MacOS'])
@@ -44,40 +45,44 @@
 	else:
 		name=name+'.app'
 	return name
+@feature('cprogram','cxxprogram')
+@after_method('apply_link')
 def create_task_macapp(self):
-	if self.env['MACAPP']or getattr(self,'mac_app',False):
+	if self.env.MACAPP or getattr(self,'mac_app',False):
 		out=self.link_task.outputs[0]
 		name=bundle_name_for_output(out)
 		dir=self.create_bundle_dirs(name,out)
 		n1=dir.find_or_declare(['Contents','MacOS',out.name])
 		self.apptask=self.create_task('macapp',self.link_task.outputs,n1)
 		inst_to=getattr(self,'install_path','/Applications')+'/%s/Contents/MacOS/'%name
-		self.bld.install_files(inst_to,n1,chmod=Utils.O755)
-		if getattr(self,'mac_resources',None):
+		self.add_install_files(install_to=inst_to,install_from=n1,chmod=Utils.O755)
+		if getattr(self,'mac_files',None):
+			mac_files_root=getattr(self,'mac_files_root',None)
+			if isinstance(mac_files_root,str):
+				mac_files_root=self.path.find_node(mac_files_root)
+				if not mac_files_root:
+					self.bld.fatal('Invalid mac_files_root %r'%self.mac_files_root)
 			res_dir=n1.parent.parent.make_node('Resources')
 			inst_to=getattr(self,'install_path','/Applications')+'/%s/Resources'%name
-			for x in self.to_list(self.mac_resources):
-				node=self.path.find_node(x)
-				if not node:
-					raise Errors.WafError('Missing mac_resource %r in %r'%(x,self))
-				parent=node.parent
-				if os.path.isdir(node.abspath()):
-					nodes=node.ant_glob('**')
-				else:
-					nodes=[node]
-				for node in nodes:
-					rel=node.path_from(parent)
-					tsk=self.create_task('macapp',node,res_dir.make_node(rel))
-					self.bld.install_as(inst_to+'/%s'%rel,node)
+			for node in self.to_nodes(self.mac_files):
+				relpath=node.path_from(mac_files_root or node.parent)
+				self.create_task('macapp',node,res_dir.make_node(relpath))
+				self.add_install_as(install_to=os.path.join(inst_to,relpath),install_source=node)
 		if getattr(self.bld,'is_install',None):
 			self.install_task.hasrun=Task.SKIP_ME
+@feature('cprogram','cxxprogram')
+@after_method('apply_link')
 def create_task_macplist(self):
-	if self.env['MACAPP']or getattr(self,'mac_app',False):
+	if self.env.MACAPP or getattr(self,'mac_app',False):
 		out=self.link_task.outputs[0]
 		name=bundle_name_for_output(out)
 		dir=self.create_bundle_dirs(name,out)
 		n1=dir.find_or_declare(['Contents','Info.plist'])
 		self.plisttask=plisttask=self.create_task('macplist',[],n1)
+		plisttask.context={'app_name':self.link_task.outputs[0].name,'env':self.env}
+		plist_ctx=getattr(self,'plist_context',None)
+		if(plist_ctx):
+			plisttask.context.update(plist_ctx)
 		if getattr(self,'mac_plist',False):
 			node=self.path.find_resource(self.mac_plist)
 			if node:
@@ -85,13 +90,15 @@
 			else:
 				plisttask.code=self.mac_plist
 		else:
-			plisttask.code=app_info%self.link_task.outputs[0].name
+			plisttask.code=app_info
 		inst_to=getattr(self,'install_path','/Applications')+'/%s/Contents/'%name
-		self.bld.install_files(inst_to,n1)
+		self.add_install_files(install_to=inst_to,install_from=n1)
+@feature('cshlib','cxxshlib')
+@before_method('apply_link','propagate_uselib_vars')
 def apply_bundle(self):
-	if self.env['MACBUNDLE']or getattr(self,'mac_bundle',False):
-		self.env['LINKFLAGS_cshlib']=self.env['LINKFLAGS_cxxshlib']=[]
-		self.env['cshlib_PATTERN']=self.env['cxxshlib_PATTERN']=self.env['macbundle_PATTERN']
+	if self.env.MACBUNDLE or getattr(self,'mac_bundle',False):
+		self.env.LINKFLAGS_cshlib=self.env.LINKFLAGS_cxxshlib=[]
+		self.env.cshlib_PATTERN=self.env.cxxshlib_PATTERN=self.env.macbundle_PATTERN
 		use=self.use=self.to_list(getattr(self,'use',[]))
 		if not'MACBUNDLE'in use:
 			use.append('MACBUNDLE')
@@ -109,13 +116,6 @@
 			txt=self.code
 		else:
 			txt=self.inputs[0].read()
+		context=getattr(self,'context',{})
+		txt=txt.format(**context)
 		self.outputs[0].write(txt)
-
-feature('c','cxx')(set_macosx_deployment_target)
-taskgen_method(create_bundle_dirs)
-feature('cprogram','cxxprogram')(create_task_macapp)
-after_method('apply_link')(create_task_macapp)
-feature('cprogram','cxxprogram')(create_task_macplist)
-after_method('apply_link')(create_task_macplist)
-feature('cshlib','cxxshlib')(apply_bundle)
-before_method('apply_link','propagate_uselib_vars')(apply_bundle)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_preproc.py glmark2-2021.02/waflib/Tools/c_preproc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_preproc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c_preproc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,9 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import re,sys,os,string,traceback
-from waflib import Logs,Build,Utils,Errors
-from waflib.Logs import debug,error
+import re,string,traceback
+from waflib import Logs,Utils,Errors
 class PreprocError(Errors.WafError):
 	pass
 POPFILE='-'
@@ -17,13 +14,13 @@
 	standard_includes=[]
 use_trigraphs=0
 strict_quotes=0
-g_optrans={'not':'!','and':'&&','bitand':'&','and_eq':'&=','or':'||','bitor':'|','or_eq':'|=','xor':'^','xor_eq':'^=','compl':'~',}
-re_lines=re.compile('^[ \t]*(#|%:)[ \t]*(ifdef|ifndef|if|else|elif|endif|include|import|define|undef|pragma)[ \t]*(.*)\r*$',re.IGNORECASE|re.MULTILINE)
+g_optrans={'not':'!','not_eq':'!','and':'&&','and_eq':'&=','or':'||','or_eq':'|=','xor':'^','xor_eq':'^=','bitand':'&','bitor':'|','compl':'~',}
+re_lines=re.compile('^[ \t]*(?:#|%:)[ \t]*(ifdef|ifndef|if|else|elif|endif|include|import|define|undef|pragma)[ \t]*(.*)\r*$',re.IGNORECASE|re.MULTILINE)
 re_mac=re.compile("^[a-zA-Z_]\w*")
 re_fun=re.compile('^[a-zA-Z_][a-zA-Z0-9_]*[(]')
 re_pragma_once=re.compile('^\s*once\s*',re.IGNORECASE)
 re_nl=re.compile('\\\\\r*\n',re.MULTILINE)
-re_cpp=re.compile(r"""(/\*[^*]*\*+([^/*][^*]*\*+)*/)|//[^\n]*|("(\\.|[^"\\])*"|'(\\.|[^'\\])*'|.[^/"'\\]*)""",re.MULTILINE)
+re_cpp=re.compile(r'//.*?$|/\*.*?\*/|\'(?:\\.|[^\\\'])*\'|"(?:\\.|[^\\"])*"',re.DOTALL|re.MULTILINE)
 trig_def=[('??'+a,b)for a,b in zip("=-/!'()<>",r'#~\|^[]{}')]
 chr_esc={'0':0,'a':7,'b':8,'t':9,'n':10,'f':11,'v':12,'r':13,'\\':92,"'":39}
 NUM='i'
@@ -39,21 +36,13 @@
 undefined='u'
 skipped='s'
 def repl(m):
-	s=m.group(1)
-	if s:
+	s=m.group()
+	if s[0]=='/':
 		return' '
-	return m.group(3)or''
-def filter_comments(filename):
-	code=Utils.readf(filename)
-	if use_trigraphs:
-		for(a,b)in trig_def:code=code.split(a).join(b)
-	code=re_nl.sub('',code)
-	code=re_cpp.sub(repl,code)
-	return[(m.group(2),m.group(3))for m in re.finditer(re_lines,code)]
+	return s
 prec={}
 ops=['* / %','+ -','<< >>','< <= >= >','== !=','& | ^','&& ||',',']
-for x in range(len(ops)):
-	syms=ops[x]
+for x,syms in enumerate(ops):
 	for u in syms.split():
 		prec[u]=x
 def trimquotes(s):
@@ -73,23 +62,23 @@
 	elif d=='*':c=a*b
 	elif d=='/':c=a/b
 	elif d=='^':c=a^b
-	elif d=='|':c=a|b
-	elif d=='||':c=int(a or b)
-	elif d=='&':c=a&b
-	elif d=='&&':c=int(a and b)
 	elif d=='==':c=int(a==b)
-	elif d=='!=':c=int(a!=b)
+	elif d=='|'or d=='bitor':c=a|b
+	elif d=='||'or d=='or':c=int(a or b)
+	elif d=='&'or d=='bitand':c=a&b
+	elif d=='&&'or d=='and':c=int(a and b)
+	elif d=='!='or d=='not_eq':c=int(a!=b)
+	elif d=='^'or d=='xor':c=int(a^b)
 	elif d=='<=':c=int(a<=b)
 	elif d=='<':c=int(a<b)
 	elif d=='>':c=int(a>b)
 	elif d=='>=':c=int(a>=b)
-	elif d=='^':c=int(a^b)
 	elif d=='<<':c=a<<b
 	elif d=='>>':c=a>>b
 	else:c=0
 	return c
 def get_num(lst):
-	if not lst:raise PreprocError("empty list for get_num")
+	if not lst:raise PreprocError('empty list for get_num')
 	(p,v)=lst[0]
 	if p==OP:
 		if v=='(':
@@ -106,7 +95,7 @@
 						count_par+=1
 				i+=1
 			else:
-				raise PreprocError("rparen expected %r"%lst)
+				raise PreprocError('rparen expected %r'%lst)
 			(num,_)=get_term(lst[1:i])
 			return(num,lst[i+1:])
 		elif v=='+':
@@ -118,27 +107,24 @@
 			num,lst=get_num(lst[1:])
 			return(int(not int(num)),lst)
 		elif v=='~':
+			num,lst=get_num(lst[1:])
 			return(~int(num),lst)
 		else:
-			raise PreprocError("Invalid op token %r for get_num"%lst)
+			raise PreprocError('Invalid op token %r for get_num'%lst)
 	elif p==NUM:
 		return v,lst[1:]
 	elif p==IDENT:
 		return 0,lst[1:]
 	else:
-		raise PreprocError("Invalid token %r for get_num"%lst)
+		raise PreprocError('Invalid token %r for get_num'%lst)
 def get_term(lst):
-	if not lst:raise PreprocError("empty list for get_term")
+	if not lst:raise PreprocError('empty list for get_term')
 	num,lst=get_num(lst)
 	if not lst:
 		return(num,[])
 	(p,v)=lst[0]
 	if p==OP:
-		if v=='&&'and not num:
-			return(num,[])
-		elif v=='||'and num:
-			return(num,[])
-		elif v==',':
+		if v==',':
 			return get_term(lst[1:])
 		elif v=='?':
 			count_par=0
@@ -155,7 +141,7 @@
 							break
 				i+=1
 			else:
-				raise PreprocError("rparen expected %r"%lst)
+				raise PreprocError('rparen expected %r'%lst)
 			if int(num):
 				return get_term(lst[1:i])
 			else:
@@ -167,7 +153,7 @@
 				return get_term([(NUM,num2)]+lst)
 			p2,v2=lst[0]
 			if p2!=OP:
-				raise PreprocError("op expected %r"%lst)
+				raise PreprocError('op expected %r'%lst)
 			if prec[v2]>=prec[v]:
 				num2=reduce_nums(num,num2,v)
 				return get_term([(NUM,num2)]+lst)
@@ -175,7 +161,7 @@
 				num3,lst=get_num(lst[1:])
 				num3=reduce_nums(num2,num3,v2)
 				return get_term([(NUM,num),(p,v),(NUM,num3)]+lst)
-	raise PreprocError("cannot reduce %r"%lst)
+	raise PreprocError('cannot reduce %r'%lst)
 def reduce_eval(lst):
 	num,lst=get_term(lst)
 	return(NUM,num)
@@ -215,7 +201,7 @@
 					else:
 						lst[i]=(NUM,0)
 				else:
-					raise PreprocError("Invalid define expression %r"%lst)
+					raise PreprocError('Invalid define expression %r'%lst)
 		elif p==IDENT and v in defs:
 			if isinstance(defs[v],str):
 				a,b=extract_macro(defs[v])
@@ -224,17 +210,19 @@
 			to_add=macro_def[1]
 			if isinstance(macro_def[0],list):
 				del lst[i]
-				for x in range(len(to_add)):
-					lst.insert(i,to_add[x])
+				accu=to_add[:]
+				reduce_tokens(accu,defs,ban+[v])
+				for tmp in accu:
+					lst.insert(i,tmp)
 					i+=1
 			else:
 				args=[]
 				del lst[i]
 				if i>=len(lst):
-					raise PreprocError("expected '(' after %r (got nothing)"%v)
+					raise PreprocError('expected ( after %r (got nothing)'%v)
 				(p2,v2)=lst[i]
 				if p2!=OP or v2!='(':
-					raise PreprocError("expected '(' after %r"%v)
+					raise PreprocError('expected ( after %r'%v)
 				del lst[i]
 				one_param=[]
 				count_paren=0
@@ -249,7 +237,7 @@
 							if one_param:args.append(one_param)
 							break
 						elif v2==',':
-							if not one_param:raise PreprocError("empty param in funcall %s"%p)
+							if not one_param:raise PreprocError('empty param in funcall %r'%v)
 							args.append(one_param)
 							one_param=[]
 						else:
@@ -317,7 +305,7 @@
 		i+=1
 def eval_macro(lst,defs):
 	reduce_tokens(lst,defs,[])
-	if not lst:raise PreprocError("missing tokens to evaluate")
+	if not lst:raise PreprocError('missing tokens to evaluate')
 	(p,v)=reduce_eval(lst)
 	return int(v)!=0
 def extract_macro(txt):
@@ -325,7 +313,7 @@
 	if re_fun.search(txt):
 		p,name=t[0]
 		p,v=t[1]
-		if p!=OP:raise PreprocError("expected open parenthesis")
+		if p!=OP:raise PreprocError('expected (')
 		i=1
 		pindex=0
 		params={}
@@ -341,50 +329,55 @@
 				elif p==OP and v==')':
 					break
 				else:
-					raise PreprocError("unexpected token (3)")
+					raise PreprocError('unexpected token (3)')
 			elif prev==IDENT:
 				if p==OP and v==',':
 					prev=v
 				elif p==OP and v==')':
 					break
 				else:
-					raise PreprocError("comma or ... expected")
+					raise PreprocError('comma or ... expected')
 			elif prev==',':
 				if p==IDENT:
 					params[v]=pindex
 					pindex+=1
 					prev=p
 				elif p==OP and v=='...':
-					raise PreprocError("not implemented (1)")
+					raise PreprocError('not implemented (1)')
 				else:
-					raise PreprocError("comma or ... expected (2)")
+					raise PreprocError('comma or ... expected (2)')
 			elif prev=='...':
-				raise PreprocError("not implemented (2)")
+				raise PreprocError('not implemented (2)')
 			else:
-				raise PreprocError("unexpected else")
+				raise PreprocError('unexpected else')
 		return(name,[params,t[i+1:]])
 	else:
 		(p,v)=t[0]
-		return(v,[[],t[1:]])
-re_include=re.compile('^\s*(<(?P<a>.*)>|"(?P<b>.*)")')
+		if len(t)>1:
+			return(v,[[],t[1:]])
+		else:
+			return(v,[[],[('T','')]])
+re_include=re.compile('^\s*(<(?:.*)>|"(?:.*)")')
 def extract_include(txt,defs):
 	m=re_include.search(txt)
 	if m:
-		if m.group('a'):return'<',m.group('a')
-		if m.group('b'):return'"',m.group('b')
+		txt=m.group(1)
+		return txt[0],txt[1:-1]
 	toks=tokenize(txt)
 	reduce_tokens(toks,defs,['waf_include'])
 	if not toks:
-		raise PreprocError("could not parse include %s"%txt)
+		raise PreprocError('could not parse include %r'%txt)
 	if len(toks)==1:
 		if toks[0][0]==STR:
 			return'"',toks[0][1]
 	else:
 		if toks[0][1]=='<'and toks[-1][1]=='>':
-			return stringize(toks).lstrip('<').rstrip('>')
-	raise PreprocError("could not parse include %s."%txt)
+			ret='<',stringize(toks).lstrip('<').rstrip('>')
+			return ret
+	raise PreprocError('could not parse include %r'%txt)
 def parse_char(txt):
-	if not txt:raise PreprocError("attempted to parse a null char")
+	if not txt:
+		raise PreprocError('attempted to parse a null char')
 	if txt[0]!='\\':
 		return ord(txt)
 	c=txt[1]
@@ -398,8 +391,10 @@
 				return(1+i,int(txt[1:1+i],8))
 	else:
 		try:return chr_esc[c]
-		except KeyError:raise PreprocError("could not parse char literal '%s'"%txt)
+		except KeyError:raise PreprocError('could not parse char literal %r'%txt)
 def tokenize(s):
+	return tokenize_private(s)[:]
+def tokenize_private(s):
 	ret=[]
 	for match in re_clexer.finditer(s):
 		m=match.group
@@ -407,7 +402,9 @@
 			v=m(name)
 			if v:
 				if name==IDENT:
-					try:v=g_optrans[v];name=OP
+					try:
+						g_optrans[v]
+						name=OP
 					except KeyError:
 						if v.lower()=="true":
 							v=1
@@ -431,8 +428,18 @@
 				ret.append((name,v))
 				break
 	return ret
-def define_name(line):
-	return re_mac.match(line).group(0)
+def format_defines(lst):
+	ret=[]
+	for y in lst:
+		if y:
+			pos=y.find('=')
+			if pos==-1:
+				ret.append(y)
+			elif pos>0:
+				ret.append('%s %s'%(y[:pos],y[pos+1:]))
+			else:
+				raise ValueError('Invalid define expression %r'%y)
+	return ret
 class c_parser(object):
 	def __init__(self,nodepaths=None,defines=None):
 		self.lines=[]
@@ -450,81 +457,84 @@
 		self.ban_includes=set([])
 	def cached_find_resource(self,node,filename):
 		try:
-			nd=node.ctx.cache_nd
-		except:
-			nd=node.ctx.cache_nd={}
-		tup=(node,filename)
+			cache=node.ctx.preproc_cache_node
+		except AttributeError:
+			cache=node.ctx.preproc_cache_node=Utils.lru_cache(1000)
+		key=(node,filename)
 		try:
-			return nd[tup]
+			return cache[key]
 		except KeyError:
 			ret=node.find_resource(filename)
 			if ret:
 				if getattr(ret,'children',None):
 					ret=None
 				elif ret.is_child_of(node.ctx.bldnode):
-					tmp=node.ctx.srcnode.search(ret.path_from(node.ctx.bldnode))
+					tmp=node.ctx.srcnode.search_node(ret.path_from(node.ctx.bldnode))
 					if tmp and getattr(tmp,'children',None):
 						ret=None
-			nd[tup]=ret
+			cache[key]=ret
 			return ret
 	def tryfind(self,filename):
+		if filename.endswith('.moc'):
+			self.names.append(filename)
+			return None
 		self.curfile=filename
 		found=self.cached_find_resource(self.currentnode_stack[-1],filename)
 		for n in self.nodepaths:
 			if found:
 				break
 			found=self.cached_find_resource(n,filename)
-		if found:
+		if found and not found in self.ban_includes:
 			self.nodes.append(found)
-			if filename[-4:]!='.moc':
-				self.addlines(found)
+			self.addlines(found)
 		else:
 			if not filename in self.names:
 				self.names.append(filename)
 		return found
+	def filter_comments(self,node):
+		code=node.read()
+		if use_trigraphs:
+			for(a,b)in trig_def:code=code.split(a).join(b)
+		code=re_nl.sub('',code)
+		code=re_cpp.sub(repl,code)
+		return re_lines.findall(code)
+	def parse_lines(self,node):
+		try:
+			cache=node.ctx.preproc_cache_lines
+		except AttributeError:
+			cache=node.ctx.preproc_cache_lines=Utils.lru_cache(1000)
+		try:
+			return cache[node]
+		except KeyError:
+			cache[node]=lines=self.filter_comments(node)
+			lines.append((POPFILE,''))
+			lines.reverse()
+			return lines
 	def addlines(self,node):
 		self.currentnode_stack.append(node.parent)
-		filepath=node.abspath()
 		self.count_files+=1
 		if self.count_files>recursion_limit:
-			raise PreprocError("recursion limit exceeded")
-		pc=self.parse_cache
-		debug('preproc: reading file %r',filepath)
-		try:
-			lns=pc[filepath]
-		except KeyError:
-			pass
-		else:
-			self.lines.extend(lns)
-			return
+			raise PreprocError('recursion limit exceeded')
+		if Logs.verbose:
+			Logs.debug('preproc: reading file %r',node)
 		try:
-			lines=filter_comments(filepath)
-			lines.append((POPFILE,''))
-			lines.reverse()
-			pc[filepath]=lines
-			self.lines.extend(lines)
-		except IOError:
-			raise PreprocError("could not read the file %s"%filepath)
+			lines=self.parse_lines(node)
+		except EnvironmentError:
+			raise PreprocError('could not read the file %r'%node)
 		except Exception:
 			if Logs.verbose>0:
-				error("parsing %s failed"%filepath)
+				Logs.error('parsing %r failed',node)
 				traceback.print_exc()
+		else:
+			self.lines.extend(lines)
 	def start(self,node,env):
-		debug('preproc: scanning %s (in %s)',node.name,node.parent.name)
-		bld=node.ctx
-		try:
-			self.parse_cache=bld.parse_cache
-		except AttributeError:
-			bld.parse_cache={}
-			self.parse_cache=bld.parse_cache
+		Logs.debug('preproc: scanning %s (in %s)',node.name,node.parent.name)
+		self.current_file=node
 		self.addlines(node)
-		if env['DEFINES']:
-			try:
-				lst=['%s %s'%(x[0],trimquotes('='.join(x[1:])))for x in[y.split('=')for y in env['DEFINES']]]
-				lst.reverse()
-				self.lines.extend([('define',x)for x in lst])
-			except AttributeError:
-				pass
+		if env.DEFINES:
+			lst=format_defines(env.DEFINES)
+			lst.reverse()
+			self.lines.extend([('define',x)for x in lst])
 		while self.lines:
 			(token,line)=self.lines.pop()
 			if token==POPFILE:
@@ -533,7 +543,7 @@
 				continue
 			try:
 				ve=Logs.verbose
-				if ve:debug('preproc: line is %s - %s state is %s',token,line,self.state)
+				if ve:Logs.debug('preproc: line is %s - %s state is %s',token,line,self.state)
 				state=self.state
 				if token[:2]=='if':
 					state.append(undefined)
@@ -548,20 +558,19 @@
 					else:state[-1]=ignored
 				elif token=='ifdef':
 					m=re_mac.match(line)
-					if m and m.group(0)in self.defs:state[-1]=accepted
+					if m and m.group()in self.defs:state[-1]=accepted
 					else:state[-1]=ignored
 				elif token=='ifndef':
 					m=re_mac.match(line)
-					if m and m.group(0)in self.defs:state[-1]=ignored
+					if m and m.group()in self.defs:state[-1]=ignored
 					else:state[-1]=accepted
 				elif token=='include'or token=='import':
 					(kind,inc)=extract_include(line,self.defs)
-					if inc in self.ban_includes:
-						continue
-					if token=='import':self.ban_includes.add(inc)
-					if ve:debug('preproc: include found %s    (%s) ',inc,kind)
+					if ve:Logs.debug('preproc: include found %s    (%s) ',inc,kind)
 					if kind=='"'or not strict_quotes:
-						self.tryfind(inc)
+						self.current_file=self.tryfind(inc)
+						if token=='import':
+							self.ban_includes.add(self.current_file)
 				elif token=='elif':
 					if state[-1]==accepted:
 						state[-1]=skipped
@@ -573,19 +582,21 @@
 					elif state[-1]==ignored:state[-1]=accepted
 				elif token=='define':
 					try:
-						self.defs[define_name(line)]=line
-					except:
-						raise PreprocError("Invalid define line %s"%line)
+						self.defs[self.define_name(line)]=line
+					except AttributeError:
+						raise PreprocError('Invalid define line %r'%line)
 				elif token=='undef':
 					m=re_mac.match(line)
-					if m and m.group(0)in self.defs:
-						self.defs.__delitem__(m.group(0))
+					if m and m.group()in self.defs:
+						self.defs.__delitem__(m.group())
 				elif token=='pragma':
 					if re_pragma_once.match(line.lower()):
-						self.ban_includes.add(self.curfile)
-			except Exception ,e:
+						self.ban_includes.add(self.current_file)
+			except Exception as e:
 				if Logs.verbose:
-					debug('preproc: line parsing failed (%s): %s %s',e,line,Utils.ex_stack())
+					Logs.debug('preproc: line parsing failed (%s): %s %s',e,line,Utils.ex_stack())
+	def define_name(self,line):
+		return re_mac.match(line).group()
 def scan(task):
 	global go_absolute
 	try:
@@ -593,14 +604,9 @@
 	except AttributeError:
 		raise Errors.WafError('%r is missing a feature such as "c", "cxx" or "includes": '%task.generator)
 	if go_absolute:
-		nodepaths=incn+standard_includes
+		nodepaths=incn+[task.generator.bld.root.find_dir(x)for x in standard_includes]
 	else:
 		nodepaths=[x for x in incn if x.is_child_of(x.ctx.srcnode)or x.is_child_of(x.ctx.bldnode)]
 	tmp=c_parser(nodepaths)
 	tmp.start(task.inputs[0],task.env)
-	if Logs.verbose:
-		debug('deps: deps for %r: %r; unresolved %r'%(task.inputs,tmp.nodes,tmp.names))
 	return(tmp.nodes,tmp.names)
-
-Utils.run_once(tokenize)
-Utils.run_once(define_name)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c.py glmark2-2021.02/waflib/Tools/c.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,27 +1,26 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-from waflib import TaskGen,Task,Utils
+from waflib import TaskGen,Task
 from waflib.Tools import c_preproc
 from waflib.Tools.ccroot import link_task,stlink_task
+@TaskGen.extension('.c')
 def c_hook(self,node):
+	if not self.env.CC and self.env.CXX:
+		return self.create_compiled_task('cxx',node)
 	return self.create_compiled_task('c',node)
 class c(Task.Task):
-	run_str='${CC} ${ARCH_ST:ARCH} ${CFLAGS} ${CPPFLAGS} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${CPPPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${CC_SRC_F}${SRC} ${CC_TGT_F}${TGT}'
+	run_str='${CC} ${ARCH_ST:ARCH} ${CFLAGS} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${CPPPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${CC_SRC_F}${SRC} ${CC_TGT_F}${TGT[0].abspath()} ${CPPFLAGS}'
 	vars=['CCDEPS']
 	ext_in=['.h']
 	scan=c_preproc.scan
-Task.classes['cc']=cc=c
 class cprogram(link_task):
-	run_str='${LINK_CC} ${LINKFLAGS} ${CCLNK_SRC_F}${SRC} ${CCLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${FRAMEWORK_ST:FRAMEWORK} ${ARCH_ST:ARCH} ${STLIB_MARKER} ${STLIBPATH_ST:STLIBPATH} ${STLIB_ST:STLIB} ${SHLIB_MARKER} ${LIBPATH_ST:LIBPATH} ${LIB_ST:LIB}'
+	run_str='${LINK_CC} ${LINKFLAGS} ${CCLNK_SRC_F}${SRC} ${CCLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${FRAMEWORK_ST:FRAMEWORK} ${ARCH_ST:ARCH} ${STLIB_MARKER} ${STLIBPATH_ST:STLIBPATH} ${STLIB_ST:STLIB} ${SHLIB_MARKER} ${LIBPATH_ST:LIBPATH} ${LIB_ST:LIB} ${LDFLAGS}'
 	ext_out=['.bin']
 	vars=['LINKDEPS']
 	inst_to='${BINDIR}'
-	chmod=Utils.O755
 class cshlib(cprogram):
 	inst_to='${LIBDIR}'
 class cstlib(stlink_task):
 	pass
-
-TaskGen.extension('.c')(c_hook)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/cs.py glmark2-2021.02/waflib/Tools/cs.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/cs.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/cs.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,15 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-from waflib import Utils,Task,Options,Logs,Errors
+from waflib import Utils,Task,Options,Errors
 from waflib.TaskGen import before_method,after_method,feature
 from waflib.Tools import ccroot
 from waflib.Configure import conf
 ccroot.USELIB_VARS['cs']=set(['CSFLAGS','ASSEMBLIES','RESOURCES'])
 ccroot.lib_patterns['csshlib']=['%s']
+@feature('cs')
+@before_method('process_source')
 def apply_cs(self):
 	cs_nodes=[]
 	no_nodes=[]
@@ -19,14 +19,17 @@
 		else:
 			no_nodes.append(x)
 	self.source=no_nodes
-	bintype=getattr(self,'type',self.gen.endswith('.dll')and'library'or'exe')
+	bintype=getattr(self,'bintype',self.gen.endswith('.dll')and'library'or'exe')
 	self.cs_task=tsk=self.create_task('mcs',cs_nodes,self.path.find_or_declare(self.gen))
 	tsk.env.CSTYPE='/target:%s'%bintype
 	tsk.env.OUT='/out:%s'%tsk.outputs[0].abspath()
+	self.env.append_value('CSFLAGS','/platform:%s'%getattr(self,'platform','anycpu'))
 	inst_to=getattr(self,'install_path',bintype=='exe'and'${BINDIR}'or'${LIBDIR}')
 	if inst_to:
 		mod=getattr(self,'chmod',bintype=='exe'and Utils.O755 or Utils.O644)
-		self.install_task=self.bld.install_files(inst_to,self.cs_task.outputs[:],env=self.env,chmod=mod)
+		self.install_task=self.add_install_files(install_to=inst_to,install_from=self.cs_task.outputs[:],chmod=mod)
+@feature('cs')
+@after_method('apply_cs')
 def use_cs(self):
 	names=self.to_list(getattr(self,'use',[]))
 	get=self.bld.get_tgen_by_name
@@ -34,7 +37,7 @@
 		try:
 			y=get(x)
 		except Errors.WafError:
-			self.cs_task.env.append_value('CSFLAGS','/reference:%s'%x)
+			self.env.append_value('CSFLAGS','/reference:%s'%x)
 			continue
 		y.post()
 		tsk=getattr(y,'cs_task',None)or getattr(y,'link_task',None)
@@ -42,7 +45,9 @@
 			self.bld.fatal('cs task has no link task for use %r'%self)
 		self.cs_task.dep_nodes.extend(tsk.outputs)
 		self.cs_task.set_run_after(tsk)
-		self.cs_task.env.append_value('CSFLAGS','/reference:%s'%tsk.outputs[0].abspath())
+		self.env.append_value('CSFLAGS','/reference:%s'%tsk.outputs[0].abspath())
+@feature('cs')
+@after_method('apply_cs','use_cs')
 def debug_cs(self):
 	csdebug=getattr(self,'csdebug',self.env.CSDEBUG)
 	if not csdebug:
@@ -63,10 +68,14 @@
 		val=['/debug+','/debug:full']
 	else:
 		val=['/debug-']
-	self.cs_task.env.append_value('CSFLAGS',val)
+	self.env.append_value('CSFLAGS',val)
 class mcs(Task.Task):
 	color='YELLOW'
 	run_str='${MCS} ${CSTYPE} ${CSFLAGS} ${ASS_ST:ASSEMBLIES} ${RES_ST:RESOURCES} ${OUT} ${SRC}'
+	def exec_command(self,cmd,**kw):
+		if'/noconfig'in cmd:
+			raise ValueError('/noconfig is not allowed when using response files, check your flags!')
+		return super(self.__class__,self).exec_command(cmd,**kw)
 def configure(conf):
 	csc=getattr(Options.options,'cscbinary',None)
 	if csc:
@@ -83,16 +92,7 @@
 	color='YELLOW'
 	inst_to=None
 	def runnable_status(self):
-		for x in self.outputs:
-			x.sig=Utils.h_file(x.abspath())
 		return Task.SKIP_ME
+@conf
 def read_csshlib(self,name,paths=[]):
 	return self(name=name,features='fake_lib',lib_paths=paths,lib_type='csshlib')
-
-feature('cs')(apply_cs)
-before_method('process_source')(apply_cs)
-feature('cs')(use_cs)
-after_method('apply_cs')(use_cs)
-feature('cs')(debug_cs)
-after_method('apply_cs','use_cs')(debug_cs)
-conf(read_csshlib)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_tests.py glmark2-2021.02/waflib/Tools/c_tests.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/c_tests.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/c_tests.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,10 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib import Task
 from waflib.Configure import conf
 from waflib.TaskGen import feature,before_method,after_method
-import sys
 LIB_CODE='''
 #ifdef _MSC_VER
 #define testEXPORT __declspec(dllexport)
@@ -21,8 +20,13 @@
 #define testEXPORT
 #endif
 testEXPORT int lib_func(void);
-int main(void) {return !(lib_func() == 9);}
+int main(int argc, char **argv) {
+	(void)argc; (void)argv;
+	return !(lib_func() == 9);
+}
 '''
+@feature('link_lib_test')
+@before_method('process_source')
 def link_lib_test_fun(self):
 	def write_test_file(task):
 		task.outputs[0].write(task.generator.code)
@@ -37,12 +41,13 @@
 	bld(rule=write_test_file,target='main.'+mode,code=MAIN_CODE)
 	bld(features='%sshlib'%m,source='test.'+mode,target='test')
 	bld(features='%sprogram %s'%(m,ex),source='main.'+mode,target='app',use='test',rpath=rpath)
+@conf
 def check_library(self,mode=None,test_exec=True):
 	if not mode:
 		mode='c'
 		if self.env.CXX:
 			mode='cxx'
-	self.check(compile_filename=[],features='link_lib_test',msg='Checking for libraries',mode=mode,test_exec=test_exec,)
+	self.check(compile_filename=[],features='link_lib_test',msg='Checking for libraries',mode=mode,test_exec=test_exec)
 INLINE_CODE='''
 typedef int foo_t;
 static %s foo_t static_foo () {return 0; }
@@ -51,6 +56,7 @@
 }
 '''
 INLINE_VALUES=['inline','__inline__','__inline']
+@conf
 def check_inline(self,**kw):
 	self.start_msg('Checking for inline')
 	if not'define_name'in kw:
@@ -72,7 +78,13 @@
 				self.define('inline',x,quote=False)
 			return x
 	self.fatal('could not use inline functions')
-LARGE_FRAGMENT='#include <unistd.h>\nint main() { return !(sizeof(off_t) >= 8); }\n'
+LARGE_FRAGMENT='''#include <unistd.h>
+int main(int argc, char **argv) {
+	(void)argc; (void)argv;
+	return !(sizeof(off_t) >= 8);
+}
+'''
+@conf
 def check_large_file(self,**kw):
 	if not'define_name'in kw:
 		kw['define_name']='HAVE_LARGEFILE'
@@ -127,20 +139,14 @@
 			self.generator.tmp.append('big')
 		else:
 			return-1
+@feature('grep_for_endianness')
+@after_method('process_source')
 def grep_for_endianness_fun(self):
 	self.create_task('grep_for_endianness',self.compiled_tasks[0].outputs[0])
+@conf
 def check_endianness(self):
 	tmp=[]
 	def check_msg(self):
 		return tmp[0]
-	self.check(fragment=ENDIAN_FRAGMENT,features='c grep_for_endianness',msg="Checking for endianness",define='ENDIANNESS',tmp=tmp,okmsg=check_msg)
+	self.check(fragment=ENDIAN_FRAGMENT,features='c grep_for_endianness',msg='Checking for endianness',define='ENDIANNESS',tmp=tmp,okmsg=check_msg)
 	return tmp[0]
-
-feature('link_lib_test')(link_lib_test_fun)
-before_method('process_source')(link_lib_test_fun)
-conf(check_library)
-conf(check_inline)
-conf(check_large_file)
-feature('grep_for_endianness')(grep_for_endianness_fun)
-after_method('process_source')(grep_for_endianness_fun)
-conf(check_endianness)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/cxx.py glmark2-2021.02/waflib/Tools/cxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/cxx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/cxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,26 +1,25 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-from waflib import TaskGen,Task,Utils
+from waflib import TaskGen,Task
 from waflib.Tools import c_preproc
 from waflib.Tools.ccroot import link_task,stlink_task
+@TaskGen.extension('.cpp','.cc','.cxx','.C','.c++')
 def cxx_hook(self,node):
 	return self.create_compiled_task('cxx',node)
-TaskGen.extension('.cpp','.cc','.cxx','.C','.c++')(cxx_hook)
 if not'.c'in TaskGen.task_gen.mappings:
 	TaskGen.task_gen.mappings['.c']=TaskGen.task_gen.mappings['.cpp']
 class cxx(Task.Task):
-	run_str='${CXX} ${ARCH_ST:ARCH} ${CXXFLAGS} ${CPPFLAGS} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${CPPPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${CXX_SRC_F}${SRC} ${CXX_TGT_F}${TGT}'
+	run_str='${CXX} ${ARCH_ST:ARCH} ${CXXFLAGS} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${CPPPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${CXX_SRC_F}${SRC} ${CXX_TGT_F}${TGT[0].abspath()} ${CPPFLAGS}'
 	vars=['CXXDEPS']
 	ext_in=['.h']
 	scan=c_preproc.scan
 class cxxprogram(link_task):
-	run_str='${LINK_CXX} ${LINKFLAGS} ${CXXLNK_SRC_F}${SRC} ${CXXLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${FRAMEWORK_ST:FRAMEWORK} ${ARCH_ST:ARCH} ${STLIB_MARKER} ${STLIBPATH_ST:STLIBPATH} ${STLIB_ST:STLIB} ${SHLIB_MARKER} ${LIBPATH_ST:LIBPATH} ${LIB_ST:LIB}'
+	run_str='${LINK_CXX} ${LINKFLAGS} ${CXXLNK_SRC_F}${SRC} ${CXXLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FRAMEWORKPATH_ST:FRAMEWORKPATH} ${FRAMEWORK_ST:FRAMEWORK} ${ARCH_ST:ARCH} ${STLIB_MARKER} ${STLIBPATH_ST:STLIBPATH} ${STLIB_ST:STLIB} ${SHLIB_MARKER} ${LIBPATH_ST:LIBPATH} ${LIB_ST:LIB} ${LDFLAGS}'
 	vars=['LINKDEPS']
 	ext_out=['.bin']
 	inst_to='${BINDIR}'
-	chmod=Utils.O755
 class cxxshlib(cxxprogram):
 	inst_to='${LIBDIR}'
 class cxxstlib(stlink_task):
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/dbus.py glmark2-2021.02/waflib/Tools/dbus.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/dbus.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/dbus.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,17 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib import Task,Errors
 from waflib.TaskGen import taskgen_method,before_method
+@taskgen_method
 def add_dbus_file(self,filename,prefix,mode):
 	if not hasattr(self,'dbus_lst'):
 		self.dbus_lst=[]
 	if not'process_dbus'in self.meths:
 		self.meths.append('process_dbus')
 	self.dbus_lst.append([filename,prefix,mode])
+@before_method('apply_core')
 def process_dbus(self):
 	for filename,prefix,mode in getattr(self,'dbus_lst',[]):
 		node=self.path.find_resource(filename)
@@ -24,7 +26,4 @@
 	run_str='${DBUS_BINDING_TOOL} --prefix=${DBUS_BINDING_TOOL_PREFIX} --mode=${DBUS_BINDING_TOOL_MODE} --output=${TGT} ${SRC}'
 	shell=True
 def configure(conf):
-	dbus_binding_tool=conf.find_program('dbus-binding-tool',var='DBUS_BINDING_TOOL')
-
-taskgen_method(add_dbus_file)
-before_method('apply_core')(process_dbus)
\ No newline at end of file
+	conf.find_program('dbus-binding-tool',var='DBUS_BINDING_TOOL')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d_config.py glmark2-2021.02/waflib/Tools/d_config.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d_config.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/d_config.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,21 +1,27 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib import Utils
 from waflib.Configure import conf
+@conf
 def d_platform_flags(self):
 	v=self.env
 	if not v.DEST_OS:
 		v.DEST_OS=Utils.unversioned_sys_platform()
-	if Utils.destos_to_binfmt(self.env.DEST_OS)=='pe':
-		v['dprogram_PATTERN']='%s.exe'
-		v['dshlib_PATTERN']='lib%s.dll'
-		v['dstlib_PATTERN']='lib%s.a'
+	binfmt=Utils.destos_to_binfmt(self.env.DEST_OS)
+	if binfmt=='pe':
+		v.dprogram_PATTERN='%s.exe'
+		v.dshlib_PATTERN='lib%s.dll'
+		v.dstlib_PATTERN='lib%s.a'
+	elif binfmt=='mac-o':
+		v.dprogram_PATTERN='%s'
+		v.dshlib_PATTERN='lib%s.dylib'
+		v.dstlib_PATTERN='lib%s.a'
 	else:
-		v['dprogram_PATTERN']='%s'
-		v['dshlib_PATTERN']='lib%s.so'
-		v['dstlib_PATTERN']='lib%s.a'
+		v.dprogram_PATTERN='%s'
+		v.dshlib_PATTERN='lib%s.so'
+		v.dstlib_PATTERN='lib%s.a'
 DLIB='''
 version(D_Version2) {
 	import std.stdio;
@@ -39,9 +45,8 @@
 	}
 }
 '''
-def check_dlibrary(self):
-	ret=self.check_cc(features='d dprogram',fragment=DLIB,compile_filename='test.d',execute=True,define_ret=True)
-	self.env.DLIBRARY=ret.strip()
-
-conf(d_platform_flags)
-conf(check_dlibrary)
\ No newline at end of file
+@conf
+def check_dlibrary(self,execute=True):
+	ret=self.check_cc(features='d dprogram',fragment=DLIB,compile_filename='test.d',execute=execute,define_ret=True)
+	if execute:
+		self.env.DLIBRARY=ret.strip()
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/dmd.py glmark2-2021.02/waflib/Tools/dmd.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/dmd.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/dmd.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,47 +1,51 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import sys
 from waflib.Tools import ar,d
 from waflib.Configure import conf
+@conf
 def find_dmd(conf):
-	conf.find_program(['dmd','ldc'],var='D')
+	conf.find_program(['dmd','dmd2','ldc'],var='D')
+	out=conf.cmd_and_log(conf.env.D+['--help'])
+	if out.find("D Compiler v")==-1:
+		out=conf.cmd_and_log(conf.env.D+['-version'])
+		if out.find("based on DMD v1.")==-1:
+			conf.fatal("detected compiler is not dmd/ldc")
+@conf
 def common_flags_ldc(conf):
 	v=conf.env
-	v['DFLAGS']=['-d-version=Posix']
-	v['LINKFLAGS']=[]
-	v['DFLAGS_dshlib']=['-relocation-model=pic']
+	v.DFLAGS=['-d-version=Posix']
+	v.LINKFLAGS=[]
+	v.DFLAGS_dshlib=['-relocation-model=pic']
+@conf
 def common_flags_dmd(conf):
 	v=conf.env
-	v['D_SRC_F']=['-c']
-	v['D_TGT_F']='-of%s'
-	v['D_LINKER']=v['D']
-	v['DLNK_SRC_F']=''
-	v['DLNK_TGT_F']='-of%s'
-	v['DINC_ST']='-I%s'
-	v['DSHLIB_MARKER']=v['DSTLIB_MARKER']=''
-	v['DSTLIB_ST']=v['DSHLIB_ST']='-L-l%s'
-	v['DSTLIBPATH_ST']=v['DLIBPATH_ST']='-L-L%s'
-	v['LINKFLAGS_dprogram']=['-quiet']
-	v['DFLAGS_dshlib']=['-fPIC']
-	v['LINKFLAGS_dshlib']=['-L-shared']
-	v['DHEADER_ext']='.di'
+	v.D_SRC_F=['-c']
+	v.D_TGT_F='-of%s'
+	v.D_LINKER=v.D
+	v.DLNK_SRC_F=''
+	v.DLNK_TGT_F='-of%s'
+	v.DINC_ST='-I%s'
+	v.DSHLIB_MARKER=v.DSTLIB_MARKER=''
+	v.DSTLIB_ST=v.DSHLIB_ST='-L-l%s'
+	v.DSTLIBPATH_ST=v.DLIBPATH_ST='-L-L%s'
+	v.LINKFLAGS_dprogram=['-quiet']
+	v.DFLAGS_dshlib=['-fPIC']
+	v.LINKFLAGS_dshlib=['-L-shared']
+	v.DHEADER_ext='.di'
 	v.DFLAGS_d_with_header=['-H','-Hf']
-	v['D_HDR_F']='%s'
+	v.D_HDR_F='%s'
 def configure(conf):
 	conf.find_dmd()
 	if sys.platform=='win32':
-		out=conf.cmd_and_log([conf.env.D,'--help'])
-		if out.find("D Compiler v2.")>-1:
-			conf.fatal('dmd2 on Windows is not supported, use gdc or ldc instead')
+		out=conf.cmd_and_log(conf.env.D+['--help'])
+		if out.find('D Compiler v2.')>-1:
+			conf.fatal('dmd2 on Windows is not supported, use gdc or ldc2 instead')
 	conf.load('ar')
 	conf.load('d')
 	conf.common_flags_dmd()
 	conf.d_platform_flags()
 	if str(conf.env.D).find('ldc')>-1:
 		conf.common_flags_ldc()
-
-conf(find_dmd)
-conf(common_flags_ldc)
-conf(common_flags_dmd)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d.py glmark2-2021.02/waflib/Tools/d.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/d.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,6 +1,6 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib import Utils,Task,Errors
 from waflib.TaskGen import taskgen_method,feature,extension
@@ -18,11 +18,11 @@
 class dprogram(link_task):
 	run_str='${D_LINKER} ${LINKFLAGS} ${DLNK_SRC_F}${SRC} ${DLNK_TGT_F:TGT} ${RPATH_ST:RPATH} ${DSTLIB_MARKER} ${DSTLIBPATH_ST:STLIBPATH} ${DSTLIB_ST:STLIB} ${DSHLIB_MARKER} ${DLIBPATH_ST:LIBPATH} ${DSHLIB_ST:LIB}'
 	inst_to='${BINDIR}'
-	chmod=Utils.O755
 class dshlib(dprogram):
 	inst_to='${LIBDIR}'
 class dstlib(stlink_task):
 	pass
+@extension('.d','.di','.D')
 def d_hook(self,node):
 	ext=Utils.destos_to_binfmt(self.env.DEST_OS)=='pe'and'obj'or'o'
 	out='%s.%d.%s'%(node.name,self.idx,ext)
@@ -35,22 +35,20 @@
 		return task
 	if getattr(self,'generate_headers',None):
 		tsk=create_compiled_task(self,'d_with_header',node)
-		tsk.outputs.append(node.change_ext(self.env['DHEADER_ext']))
+		tsk.outputs.append(node.change_ext(self.env.DHEADER_ext))
 	else:
 		tsk=create_compiled_task(self,'d',node)
 	return tsk
-def generate_header(self,filename,install_path=None):
+@taskgen_method
+def generate_header(self,filename):
 	try:
-		self.header_lst.append([filename,install_path])
+		self.header_lst.append([filename,self.install_path])
 	except AttributeError:
-		self.header_lst=[[filename,install_path]]
+		self.header_lst=[[filename,self.install_path]]
+@feature('d')
 def process_header(self):
 	for i in getattr(self,'header_lst',[]):
 		node=self.path.find_resource(i[0])
 		if not node:
 			raise Errors.WafError('file %r not found on d obj'%i[0])
 		self.create_task('d_header',node,node.change_ext('.di'))
-
-extension('.d','.di','.D')(d_hook)
-taskgen_method(generate_header)
-feature('d')(process_header)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d_scan.py glmark2-2021.02/waflib/Tools/d_scan.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/d_scan.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/d_scan.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,9 +1,9 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import re
-from waflib import Utils,Logs
+from waflib import Utils
 def filter_comments(filename):
 	txt=Utils.readf(filename)
 	i=0
@@ -128,6 +128,4 @@
 	gruik.start(node)
 	nodes=gruik.nodes
 	names=gruik.names
-	if Logs.verbose:
-		Logs.debug('deps: deps for %s: %r; unresolved %r'%(str(node),nodes,names))
 	return(nodes,names)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/errcheck.py glmark2-2021.02/waflib/Tools/errcheck.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/errcheck.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/errcheck.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,24 +1,25 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-typos={'feature':'features','sources':'source','targets':'target','include':'includes','export_include':'export_includes','define':'defines','importpath':'includes','installpath':'install_path',}
+typos={'feature':'features','sources':'source','targets':'target','include':'includes','export_include':'export_includes','define':'defines','importpath':'includes','installpath':'install_path','iscopy':'is_copy',}
 meths_typos=['__call__','program','shlib','stlib','objects']
+import sys
 from waflib import Logs,Build,Node,Task,TaskGen,ConfigSet,Errors,Utils
-import waflib.Tools.ccroot
+from waflib.Tools import ccroot
 def check_same_targets(self):
 	mp=Utils.defaultdict(list)
 	uids={}
 	def check_task(tsk):
 		if not isinstance(tsk,Task.Task):
 			return
+		if hasattr(tsk,'no_errcheck_out'):
+			return
 		for node in tsk.outputs:
 			mp[node].append(tsk)
 		try:
 			uids[tsk.uid()].append(tsk)
-		except:
+		except KeyError:
 			uids[tsk.uid()]=[tsk]
 	for g in self.groups:
 		for tg in g:
@@ -31,19 +32,20 @@
 	for(k,v)in mp.items():
 		if len(v)>1:
 			dupe=True
-			msg='* Node %r is created by more than once%s. The task generators are:'%(k,Logs.verbose==1 and" (full message on 'waf -v -v')"or"")
+			msg='* Node %r is created more than once%s. The task generators are:'%(k,Logs.verbose==1 and" (full message on 'waf -v -v')"or"")
 			Logs.error(msg)
 			for x in v:
 				if Logs.verbose>1:
-					Logs.error('  %d. %r'%(1+v.index(x),x.generator))
+					Logs.error('  %d. %r',1+v.index(x),x.generator)
 				else:
-					Logs.error('  %d. %r in %r'%(1+v.index(x),x.generator.name,getattr(x.generator,'path',None)))
+					Logs.error('  %d. %r in %r',1+v.index(x),x.generator.name,getattr(x.generator,'path',None))
+			Logs.error('If you think that this is an error, set no_errcheck_out on the task instance')
 	if not dupe:
 		for(k,v)in uids.items():
 			if len(v)>1:
-				Logs.error('* Several tasks use the same identifier. Please check the information on\n   http://waf.googlecode.com/git/docs/apidocs/Task.html#waflib.Task.Task.uid')
+				Logs.error('* Several tasks use the same identifier. Please check the information on\n   https://waf.io/apidocs/Task.html?highlight=uid#waflib.Task.Task.uid')
 				for tsk in v:
-					Logs.error('  - object %r (%r) defined in %r'%(tsk.__class__.__name__,tsk,tsk.generator))
+					Logs.error('  - object %r (%r) defined in %r',tsk.__class__.__name__,tsk,tsk.generator)
 def check_invalid_constraints(self):
 	feat=set([])
 	for x in list(TaskGen.feats.values()):
@@ -56,22 +58,25 @@
 		ext.add(x.__name__)
 	invalid=ext&feat
 	if invalid:
-		Logs.error('The methods %r have invalid annotations:  @extension <-> @feature/@before_method/@after_method'%list(invalid))
+		Logs.error('The methods %r have invalid annotations:  @extension <-> @feature/@before_method/@after_method',list(invalid))
 	for cls in list(Task.classes.values()):
+		if sys.hexversion>0x3000000 and issubclass(cls,Task.Task)and isinstance(cls.hcode,str):
+			raise Errors.WafError('Class %r has hcode value %r of type <str>, expecting <bytes> (use Utils.h_cmd() ?)'%(cls,cls.hcode))
 		for x in('before','after'):
 			for y in Utils.to_list(getattr(cls,x,[])):
-				if not Task.classes.get(y,None):
-					Logs.error('Erroneous order constraint %r=%r on task class %r'%(x,y,cls.__name__))
+				if not Task.classes.get(y):
+					Logs.error('Erroneous order constraint %r=%r on task class %r',x,y,cls.__name__)
 		if getattr(cls,'rule',None):
-			Logs.error('Erroneous attribute "rule" on task class %r (rename to "run_str")'%cls.__name__)
+			Logs.error('Erroneous attribute "rule" on task class %r (rename to "run_str")',cls.__name__)
 def replace(m):
 	oldcall=getattr(Build.BuildContext,m)
 	def call(self,*k,**kw):
 		ret=oldcall(self,*k,**kw)
 		for x in typos:
 			if x in kw:
-				err=True
-				Logs.error('Fix the typo %r -> %r on %r'%(x,typos[x],ret))
+				if x=='iscopy'and'subst'in getattr(self,'features',''):
+					continue
+				Logs.error('Fix the typo %r -> %r on %r',x,typos[x],ret)
 		return ret
 	setattr(Build.BuildContext,m,call)
 def enhance_lib():
@@ -82,11 +87,11 @@
 			lst=Utils.to_list(k[0])
 			for pat in lst:
 				if'..'in pat.split('/'):
-					Logs.error("In ant_glob pattern %r: '..' means 'two dots', not 'parent directory'"%k[0])
+					Logs.error("In ant_glob pattern %r: '..' means 'two dots', not 'parent directory'",k[0])
 		if kw.get('remove',True):
 			try:
 				if self.is_child_of(self.ctx.bldnode)and not kw.get('quiet',False):
-					Logs.error('Using ant_glob on the build folder (%r) is dangerous (quiet=True to disable this warning)'%self)
+					Logs.error('Using ant_glob on the build folder (%r) is dangerous (quiet=True to disable this warning)',self)
 			except AttributeError:
 				pass
 		return self.old_ant_glob(*k,**kw)
@@ -96,7 +101,7 @@
 	def is_before(t1,t2):
 		ret=old(t1,t2)
 		if ret and old(t2,t1):
-			Logs.error('Contradictory order constraints in classes %r %r'%(t1,t2))
+			Logs.error('Contradictory order constraints in classes %r %r',t1,t2)
 		return ret
 	Task.is_before=is_before
 	def check_err_features(self):
@@ -105,18 +110,18 @@
 			Logs.error('feature shlib -> cshlib, dshlib or cxxshlib')
 		for x in('c','cxx','d','fc'):
 			if not x in lst and lst and lst[0]in[x+y for y in('program','shlib','stlib')]:
-				Logs.error('%r features is probably missing %r'%(self,x))
+				Logs.error('%r features is probably missing %r',self,x)
 	TaskGen.feature('*')(check_err_features)
 	def check_err_order(self):
-		if not hasattr(self,'rule'):
+		if not hasattr(self,'rule')and not'subst'in Utils.to_list(self.features):
 			for x in('before','after','ext_in','ext_out'):
 				if hasattr(self,x):
-					Logs.warn('Erroneous order constraint %r on non-rule based task generator %r'%(x,self))
+					Logs.warn('Erroneous order constraint %r on non-rule based task generator %r',x,self)
 		else:
 			for x in('before','after'):
 				for y in self.to_list(getattr(self,x,[])):
 					if not Task.classes.get(y,None):
-						Logs.error('Erroneous order constraint %s=%r on %r'%(x,y,self))
+						Logs.error('Erroneous order constraint %s=%r on %r (no such class)',x,y,self)
 	TaskGen.feature('*')(check_err_order)
 	def check_compile(self):
 		check_invalid_constraints(self)
@@ -145,7 +150,7 @@
 		self.orig_use_rec(name,**kw)
 	TaskGen.task_gen.orig_use_rec=TaskGen.task_gen.use_rec
 	TaskGen.task_gen.use_rec=use_rec
-	def getattri(self,name,default=None):
+	def _getattr(self,name,default=None):
 		if name=='append'or name=='add':
 			raise Errors.WafError('env.append and env.add do not exist: use env.append_value/env.append_unique')
 		elif name=='prepend':
@@ -154,8 +159,6 @@
 			return object.__getattr__(self,name,default)
 		else:
 			return self[name]
-	ConfigSet.ConfigSet.__getattr__=getattri
+	ConfigSet.ConfigSet.__getattr__=_getattr
 def options(opt):
 	enhance_lib()
-def configure(conf):
-	pass
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc_config.py glmark2-2021.02/waflib/Tools/fc_config.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc_config.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/fc_config.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,35 +1,44 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import re,shutil,os,sys,string,shlex
+import re,os,sys,shlex
 from waflib.Configure import conf
-from waflib.TaskGen import feature,after_method,before_method
-from waflib import Build,Utils
+from waflib.TaskGen import feature,before_method
 FC_FRAGMENT='        program main\n        end     program main\n'
 FC_FRAGMENT2='        PROGRAM MAIN\n        END\n'
+@conf
 def fc_flags(conf):
 	v=conf.env
-	v['FC_SRC_F']=[]
-	v['FC_TGT_F']=['-c','-o']
-	v['FCINCPATH_ST']='-I%s'
-	v['FCDEFINES_ST']='-D%s'
-	if not v['LINK_FC']:v['LINK_FC']=v['FC']
-	v['FCLNK_SRC_F']=[]
-	v['FCLNK_TGT_F']=['-o']
-	v['FCFLAGS_fcshlib']=['-fpic']
-	v['LINKFLAGS_fcshlib']=['-shared']
-	v['fcshlib_PATTERN']='lib%s.so'
-	v['fcstlib_PATTERN']='lib%s.a'
-	v['FCLIB_ST']='-l%s'
-	v['FCLIBPATH_ST']='-L%s'
-	v['FCSTLIB_ST']='-l%s'
-	v['FCSTLIBPATH_ST']='-L%s'
-	v['FCSTLIB_MARKER']='-Wl,-Bstatic'
-	v['FCSHLIB_MARKER']='-Wl,-Bdynamic'
-	v['SONAME_ST']='-Wl,-h,%s'
+	v.FC_SRC_F=[]
+	v.FC_TGT_F=['-c','-o']
+	v.FCINCPATH_ST='-I%s'
+	v.FCDEFINES_ST='-D%s'
+	if not v.LINK_FC:
+		v.LINK_FC=v.FC
+	v.FCLNK_SRC_F=[]
+	v.FCLNK_TGT_F=['-o']
+	v.FCFLAGS_fcshlib=['-fpic']
+	v.LINKFLAGS_fcshlib=['-shared']
+	v.fcshlib_PATTERN='lib%s.so'
+	v.fcstlib_PATTERN='lib%s.a'
+	v.FCLIB_ST='-l%s'
+	v.FCLIBPATH_ST='-L%s'
+	v.FCSTLIB_ST='-l%s'
+	v.FCSTLIBPATH_ST='-L%s'
+	v.FCSTLIB_MARKER='-Wl,-Bstatic'
+	v.FCSHLIB_MARKER='-Wl,-Bdynamic'
+	v.SONAME_ST='-Wl,-h,%s'
+@conf
+def fc_add_flags(conf):
+	conf.add_os_flags('FCPPFLAGS',dup=False)
+	conf.add_os_flags('FCFLAGS',dup=False)
+	conf.add_os_flags('LINKFLAGS',dup=False)
+	conf.add_os_flags('LDFLAGS',dup=False)
+@conf
 def check_fortran(self,*k,**kw):
 	self.check_cc(fragment=FC_FRAGMENT,compile_filename='test.f',features='fc fcprogram',msg='Compiling a simple fortran app')
+@conf
 def check_fc(self,*k,**kw):
 	kw['compiler']='fc'
 	if not'compile_mode'in kw:
@@ -41,32 +50,35 @@
 	if not'code'in kw:
 		kw['code']=FC_FRAGMENT
 	return self.check(*k,**kw)
+@conf
 def fortran_modifier_darwin(conf):
 	v=conf.env
-	v['FCFLAGS_fcshlib']=['-fPIC','-compatibility_version','1','-current_version','1']
-	v['LINKFLAGS_fcshlib']=['-dynamiclib']
-	v['fcshlib_PATTERN']='lib%s.dylib'
-	v['FRAMEWORKPATH_ST']='-F%s'
-	v['FRAMEWORK_ST']='-framework %s'
-	v['LINKFLAGS_fcstlib']=[]
-	v['FCSHLIB_MARKER']=''
-	v['FCSTLIB_MARKER']=''
-	v['SONAME_ST']=''
+	v.FCFLAGS_fcshlib=['-fPIC']
+	v.LINKFLAGS_fcshlib=['-dynamiclib']
+	v.fcshlib_PATTERN='lib%s.dylib'
+	v.FRAMEWORKPATH_ST='-F%s'
+	v.FRAMEWORK_ST='-framework %s'
+	v.LINKFLAGS_fcstlib=[]
+	v.FCSHLIB_MARKER=''
+	v.FCSTLIB_MARKER=''
+	v.SONAME_ST=''
+@conf
 def fortran_modifier_win32(conf):
 	v=conf.env
-	v['fcprogram_PATTERN']=v['fcprogram_test_PATTERN']='%s.exe'
-	v['fcshlib_PATTERN']='%s.dll'
-	v['implib_PATTERN']='lib%s.dll.a'
-	v['IMPLIB_ST']='-Wl,--out-implib,%s'
-	v['FCFLAGS_fcshlib']=[]
-	v.append_value('FCFLAGS_fcshlib',['-DDLL_EXPORT'])
+	v.fcprogram_PATTERN=v.fcprogram_test_PATTERN='%s.exe'
+	v.fcshlib_PATTERN='%s.dll'
+	v.implib_PATTERN='lib%s.dll.a'
+	v.IMPLIB_ST='-Wl,--out-implib,%s'
+	v.FCFLAGS_fcshlib=[]
 	v.append_value('LINKFLAGS',['-Wl,--enable-auto-import'])
+@conf
 def fortran_modifier_cygwin(conf):
 	fortran_modifier_win32(conf)
 	v=conf.env
-	v['fcshlib_PATTERN']='cyg%s.dll'
+	v.fcshlib_PATTERN='cyg%s.dll'
 	v.append_value('LINKFLAGS_fcshlib',['-Wl,--enable-auto-image-base'])
-	v['FCFLAGS_fcshlib']=[]
+	v.FCFLAGS_fcshlib=[]
+@conf
 def check_fortran_dummy_main(self,*k,**kw):
 	if not self.env.CC:
 		self.fatal('A c compiler is required for check_fortran_dummy_main')
@@ -93,6 +105,7 @@
 GCC_DRIVER_LINE=re.compile('^Driving:')
 POSIX_STATIC_EXT=re.compile('\S+\.a')
 POSIX_LIB_FLAGS=re.compile('-l\S+')
+@conf
 def is_link_verbose(self,txt):
 	assert isinstance(txt,str)
 	for line in txt.splitlines():
@@ -100,9 +113,10 @@
 			if POSIX_STATIC_EXT.search(line)or POSIX_LIB_FLAGS.search(line):
 				return True
 	return False
+@conf
 def check_fortran_verbose_flag(self,*k,**kw):
 	self.start_msg('fortran link verbose flag')
-	for x in['-v','--verbose','-verbose','-V']:
+	for x in('-v','--verbose','-verbose','-V'):
 		try:
 			self.check_cc(features='fc fcprogram_test',fragment=FC_FRAGMENT2,compile_filename='test.f',linkflags=[x],mandatory=True)
 		except self.errors.ConfigurationError:
@@ -135,41 +149,42 @@
 	return final_flags
 SPACE_OPTS=re.compile('^-[LRuYz]$')
 NOSPACE_OPTS=re.compile('^-[RL]')
+def _parse_flink_token(lexer,token,tmp_flags):
+	if _match_ignore(token):
+		pass
+	elif token.startswith('-lkernel32')and sys.platform=='cygwin':
+		tmp_flags.append(token)
+	elif SPACE_OPTS.match(token):
+		t=lexer.get_token()
+		if t.startswith('P,'):
+			t=t[2:]
+		for opt in t.split(os.pathsep):
+			tmp_flags.append('-L%s'%opt)
+	elif NOSPACE_OPTS.match(token):
+		tmp_flags.append(token)
+	elif POSIX_LIB_FLAGS.match(token):
+		tmp_flags.append(token)
+	else:
+		pass
+	t=lexer.get_token()
+	return t
 def _parse_flink_line(line,final_flags):
 	lexer=shlex.shlex(line,posix=True)
 	lexer.whitespace_split=True
 	t=lexer.get_token()
 	tmp_flags=[]
 	while t:
-		def parse(token):
-			if _match_ignore(token):
-				pass
-			elif token.startswith('-lkernel32')and sys.platform=='cygwin':
-				tmp_flags.append(token)
-			elif SPACE_OPTS.match(token):
-				t=lexer.get_token()
-				if t.startswith('P,'):
-					t=t[2:]
-				for opt in t.split(os.pathsep):
-					tmp_flags.append('-L%s'%opt)
-			elif NOSPACE_OPTS.match(token):
-				tmp_flags.append(token)
-			elif POSIX_LIB_FLAGS.match(token):
-				tmp_flags.append(token)
-			else:
-				pass
-			t=lexer.get_token()
-			return t
-		t=parse(t)
+		t=_parse_flink_token(lexer,t,tmp_flags)
 	final_flags.extend(tmp_flags)
 	return final_flags
+@conf
 def check_fortran_clib(self,autoadd=True,*k,**kw):
 	if not self.env.FC_VERBOSE_FLAG:
 		self.fatal('env.FC_VERBOSE_FLAG is not set: execute check_fortran_verbose_flag?')
 	self.start_msg('Getting fortran runtime link flags')
 	try:
 		self.check_cc(fragment=FC_FRAGMENT2,compile_filename='test.f',features='fc fcprogram_test',linkflags=[self.env.FC_VERBOSE_FLAG])
-	except:
+	except Exception:
 		self.end_msg(False)
 		if kw.get('mandatory',True):
 			conf.fatal('Could not find the c library flags')
@@ -181,24 +196,24 @@
 		return flags
 	return[]
 def getoutput(conf,cmd,stdin=False):
-	if stdin:
-		stdin=Utils.subprocess.PIPE
+	from waflib import Errors
+	if conf.env.env:
+		env=conf.env.env
 	else:
-		stdin=None
-	env=conf.env.env or None
+		env=dict(os.environ)
+		env['LANG']='C'
+	input=stdin and'\n'.encode()or None
 	try:
-		p=Utils.subprocess.Popen(cmd,stdin=stdin,stdout=Utils.subprocess.PIPE,stderr=Utils.subprocess.PIPE,env=env)
-		if stdin:
-			p.stdin.write('\n')
-		stdout,stderr=p.communicate()
-	except:
+		out,err=conf.cmd_and_log(cmd,env=env,output=0,input=input)
+	except Errors.WafError as e:
+		if not(hasattr(e,'stderr')and hasattr(e,'stdout')):
+			raise e
+		else:
+			out=e.stdout
+			err=e.stderr
+	except Exception:
 		conf.fatal('could not determine the compiler version %r'%cmd)
-	else:
-		if not isinstance(stdout,str):
-			stdout=stdout.decode(sys.stdout.encoding)
-		if not isinstance(stderr,str):
-			stderr=stderr.decode(sys.stdout.encoding)
-		return stdout,stderr
+	return(out,err)
 ROUTINES_CODE="""\
       subroutine foobar()
       return
@@ -216,6 +231,8 @@
   return 0;
 }
 """
+@feature('link_main_routines_func')
+@before_method('process_source')
 def link_main_routines_tg_method(self):
 	def write_test_file(task):
 		task.outputs[0].write(task.generator.code)
@@ -225,12 +242,13 @@
 	bld(features='fc fcstlib',source='test.f',target='test')
 	bld(features='c fcprogram',source='main.c',target='app',use='test')
 def mangling_schemes():
-	for u in['_','']:
-		for du in['','_']:
-			for c in["lower","upper"]:
+	for u in('_',''):
+		for du in('','_'):
+			for c in("lower","upper"):
 				yield(u,du,c)
 def mangle_name(u,du,c,name):
 	return getattr(name,c)()+u+(name.find('_')!=-1 and du or'')
+@conf
 def check_fortran_mangling(self,*k,**kw):
 	if not self.env.CC:
 		self.fatal('A c compiler is required for link_main_routines')
@@ -241,7 +259,7 @@
 	self.start_msg('Getting fortran mangling scheme')
 	for(u,du,c)in mangling_schemes():
 		try:
-			self.check_cc(compile_filename=[],features='link_main_routines_func',msg='nomsg',errmsg='nomsg',mandatory=True,dummy_func_nounder=mangle_name(u,du,c,"foobar"),dummy_func_under=mangle_name(u,du,c,"foo_bar"),main_func_name=self.env.FC_MAIN)
+			self.check_cc(compile_filename=[],features='link_main_routines_func',msg='nomsg',errmsg='nomsg',dummy_func_nounder=mangle_name(u,du,c,'foobar'),dummy_func_under=mangle_name(u,du,c,'foo_bar'),main_func_name=self.env.FC_MAIN)
 		except self.errors.ConfigurationError:
 			pass
 		else:
@@ -252,10 +270,13 @@
 		self.end_msg(False)
 		self.fatal('mangler not found')
 	return(u,du,c)
+@feature('pyext')
+@before_method('propagate_uselib_vars','apply_link')
 def set_lib_pat(self):
-	self.env['fcshlib_PATTERN']=self.env['pyext_PATTERN']
+	self.env.fcshlib_PATTERN=self.env.pyext_PATTERN
+@conf
 def detect_openmp(self):
-	for x in['-fopenmp','-openmp','-mp','-xopenmp','-omp','-qsmp=omp']:
+	for x in('-fopenmp','-openmp','-mp','-xopenmp','-omp','-qsmp=omp'):
 		try:
 			self.check_fc(msg='Checking for OpenMP flag %s'%x,fragment='program main\n  call omp_get_num_threads()\nend program main',fcflags=x,linkflags=x,uselib_store='OPENMP')
 		except self.errors.ConfigurationError:
@@ -264,20 +285,3 @@
 			break
 	else:
 		self.fatal('Could not find OpenMP')
-
-conf(fc_flags)
-conf(check_fortran)
-conf(check_fc)
-conf(fortran_modifier_darwin)
-conf(fortran_modifier_win32)
-conf(fortran_modifier_cygwin)
-conf(check_fortran_dummy_main)
-conf(is_link_verbose)
-conf(check_fortran_verbose_flag)
-conf(check_fortran_clib)
-feature('link_main_routines_func')(link_main_routines_tg_method)
-before_method('process_source')(link_main_routines_tg_method)
-conf(check_fortran_mangling)
-feature('pyext')(set_lib_pat)
-before_method('propagate_uselib_vars','apply_link')(set_lib_pat)
-conf(detect_openmp)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc.py glmark2-2021.02/waflib/Tools/fc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/fc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,22 +1,19 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import re
-from waflib import Utils,Task,TaskGen,Logs
+from waflib import Utils,Task
 from waflib.Tools import ccroot,fc_config,fc_scan
-from waflib.TaskGen import feature,before_method,after_method,extension
+from waflib.TaskGen import extension
 from waflib.Configure import conf
-ccroot.USELIB_VARS['fc']=set(['FCFLAGS','DEFINES','INCLUDES'])
+ccroot.USELIB_VARS['fc']=set(['FCFLAGS','DEFINES','INCLUDES','FCPPFLAGS'])
 ccroot.USELIB_VARS['fcprogram_test']=ccroot.USELIB_VARS['fcprogram']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS'])
 ccroot.USELIB_VARS['fcshlib']=set(['LIB','STLIB','LIBPATH','STLIBPATH','LINKFLAGS','RPATH','LINKDEPS'])
 ccroot.USELIB_VARS['fcstlib']=set(['ARFLAGS','LINKDEPS'])
-def dummy(self):
-	pass
+@extension('.f','.f90','.F','.F90','.for','.FOR')
 def fc_hook(self,node):
 	return self.create_compiled_task('fc',node)
+@conf
 def modfile(conf,name):
 	return{'lower':name.lower()+'.mod','lower.MOD':name.upper()+'.MOD','UPPER.mod':name.upper()+'.mod','UPPER':name.upper()+'.MOD'}[conf.env.FC_MOD_CAPITALIZATION or'lower']
 def get_fortran_tasks(tsk):
@@ -25,14 +22,12 @@
 	return[x for x in tasks if isinstance(x,fc)and not getattr(x,'nomod',None)and not getattr(x,'mod_fortran_done',None)]
 class fc(Task.Task):
 	color='GREEN'
-	run_str='${FC} ${FCFLAGS} ${FCINCPATH_ST:INCPATHS} ${FCDEFINES_ST:DEFINES} ${_FCMODOUTFLAGS} ${FC_TGT_F}${TGT[0].abspath()} ${FC_SRC_F}${SRC[0].abspath()}'
+	run_str='${FC} ${FCFLAGS} ${FCINCPATH_ST:INCPATHS} ${FCDEFINES_ST:DEFINES} ${_FCMODOUTFLAGS} ${FC_TGT_F}${TGT[0].abspath()} ${FC_SRC_F}${SRC[0].abspath()} ${FCPPFLAGS}'
 	vars=["FORTRANMODPATHFLAG"]
 	def scan(self):
 		tmp=fc_scan.fortran_parser(self.generator.includes_nodes)
 		tmp.task=self
 		tmp.start(self.inputs[0])
-		if Logs.verbose:
-			Logs.debug('deps: deps for %r: %r; unresolved %r'%(self.inputs,tmp.nodes,tmp.names))
 		return(tmp.nodes,tmp.names)
 	def runnable_status(self):
 		if getattr(self,'mod_fortran_done',None):
@@ -74,10 +69,7 @@
 				for t in outs[k]:
 					tmp.extend(t.outputs)
 				a.dep_nodes.extend(tmp)
-				try:
-					a.dep_nodes.sort(key=lambda x:x.abspath())
-				except:
-					a.dep_nodes.sort(lambda x,y:cmp(x.abspath(),y.abspath()))
+				a.dep_nodes.sort(key=lambda x:x.abspath())
 		for tsk in lst:
 			try:
 				delattr(tsk,'cache_sig')
@@ -86,14 +78,13 @@
 		return super(fc,self).runnable_status()
 class fcprogram(ccroot.link_task):
 	color='YELLOW'
-	run_str='${FC} ${LINKFLAGS} ${FCLNK_SRC_F}${SRC} ${FCLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FCSTLIB_MARKER} ${FCSTLIBPATH_ST:STLIBPATH} ${FCSTLIB_ST:STLIB} ${FCSHLIB_MARKER} ${FCLIBPATH_ST:LIBPATH} ${FCLIB_ST:LIB}'
+	run_str='${FC} ${LINKFLAGS} ${FCLNK_SRC_F}${SRC} ${FCLNK_TGT_F}${TGT[0].abspath()} ${RPATH_ST:RPATH} ${FCSTLIB_MARKER} ${FCSTLIBPATH_ST:STLIBPATH} ${FCSTLIB_ST:STLIB} ${FCSHLIB_MARKER} ${FCLIBPATH_ST:LIBPATH} ${FCLIB_ST:LIB} ${LDFLAGS}'
 	inst_to='${BINDIR}'
-	chmod=Utils.O755
 class fcshlib(fcprogram):
 	inst_to='${LIBDIR}'
+class fcstlib(ccroot.stlink_task):
+	pass
 class fcprogram_test(fcprogram):
-	def can_retrieve_cache(self):
-		return False
 	def runnable_status(self):
 		ret=super(fcprogram_test,self).runnable_status()
 		if ret==Task.SKIP_ME:
@@ -103,21 +94,15 @@
 		bld=self.generator.bld
 		kw['shell']=isinstance(cmd,str)
 		kw['stdout']=kw['stderr']=Utils.subprocess.PIPE
-		kw['cwd']=bld.variant_dir
+		kw['cwd']=self.get_cwd()
 		bld.out=bld.err=''
 		bld.to_log('command: %s\n'%cmd)
 		kw['output']=0
 		try:
 			(bld.out,bld.err)=bld.cmd_and_log(cmd,**kw)
-		except Exception ,e:
+		except Exception:
 			return-1
 		if bld.out:
-			bld.to_log("out: %s\n"%bld.out)
+			bld.to_log('out: %s\n'%bld.out)
 		if bld.err:
-			bld.to_log("err: %s\n"%bld.err)
-class fcstlib(ccroot.stlink_task):
-	pass
-
-feature('fcprogram','fcshlib','fcstlib','fcprogram_test')(dummy)
-extension('.f','.f90','.F','.F90','.for','.FOR')(fc_hook)
-conf(modfile)
\ No newline at end of file
+			bld.to_log('err: %s\n'%bld.err)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc_scan.py glmark2-2021.02/waflib/Tools/fc_scan.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/fc_scan.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/fc_scan.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,9 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import re
-from waflib import Utils,Task,TaskGen,Logs
-from waflib.TaskGen import feature,before_method,after_method,extension
-from waflib.Configure import conf
-INC_REGEX="""(?:^|['">]\s*;)\s*INCLUDE\s+(?:\w+_)?[<"'](.+?)(?=["'>])"""
+INC_REGEX="""(?:^|['">]\s*;)\s*(?:|#\s*)INCLUDE\s+(?:\w+_)?[<"'](.+?)(?=["'>])"""
 USE_REGEX="""(?:^|;)\s*USE(?:\s+|(?:(?:\s*,\s*(?:NON_)?INTRINSIC)?\s*::))\s*(\w+)"""
 MOD_REGEX="""(?:^|;)\s*MODULE(?!\s*PROCEDURE)(?:\s+|(?:(?:\s*,\s*(?:NON_)?INTRINSIC)?\s*::))\s*(\w+)"""
 re_inc=re.compile(INC_REGEX,re.I)
@@ -40,7 +37,6 @@
 			nd=self.waiting.pop(0)
 			self.iter(nd)
 	def iter(self,node):
-		path=node.abspath()
 		incs,uses,mods=self.find_deps(node)
 		for x in incs:
 			if x in self.seen:
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/flex.py glmark2-2021.02/waflib/Tools/flex.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/flex.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/flex.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,8 +1,10 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import waflib.TaskGen
+import os,re
+from waflib import Task,TaskGen
+from waflib.Tools import ccroot
 def decide_ext(self,node):
 	if'cxx'in self.features:
 		return['.lex.cc']
@@ -15,13 +17,21 @@
 		if isinstance(xx,str):return[xx]
 		return xx
 	tsk.last_cmd=lst=[]
-	lst.extend(to_list(env['FLEX']))
-	lst.extend(to_list(env['FLEXFLAGS']))
-	lst.extend([a.path_from(bld.bldnode)for a in tsk.inputs])
+	lst.extend(to_list(env.FLEX))
+	lst.extend(to_list(env.FLEXFLAGS))
+	inputs=[a.path_from(tsk.get_cwd())for a in tsk.inputs]
+	if env.FLEX_MSYS:
+		inputs=[x.replace(os.sep,'/')for x in inputs]
+	lst.extend(inputs)
 	lst=[x for x in lst if x]
 	txt=bld.cmd_and_log(lst,cwd=wd,env=env.env or None,quiet=0)
-	tsk.outputs[0].write(txt)
-waflib.TaskGen.declare_chain(name='flex',rule=flexfun,ext_in='.l',decider=decide_ext,)
+	tsk.outputs[0].write(txt.replace('\r\n','\n').replace('\r','\n'))
+TaskGen.declare_chain(name='flex',rule=flexfun,ext_in='.l',decider=decide_ext,)
+Task.classes['flex'].vars=['FLEXFLAGS','FLEX']
+ccroot.USELIB_VARS['c'].add('FLEXFLAGS')
+ccroot.USELIB_VARS['cxx'].add('FLEXFLAGS')
 def configure(conf):
 	conf.find_program('flex',var='FLEX')
 	conf.env.FLEXFLAGS=['-t']
+	if re.search(r"\\msys\\[0-9.]+\\bin\\flex.exe$",conf.env.FLEX[0]):
+		conf.env.FLEX_MSYS=True
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/g95.py glmark2-2021.02/waflib/Tools/g95.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/g95.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/g95.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,32 +1,38 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import re
 from waflib import Utils
-from waflib.Tools import fc,fc_config,fc_scan
+from waflib.Tools import fc,fc_config,fc_scan,ar
 from waflib.Configure import conf
+@conf
 def find_g95(conf):
 	fc=conf.find_program('g95',var='FC')
-	fc=conf.cmd_to_list(fc)
 	conf.get_g95_version(fc)
 	conf.env.FC_NAME='G95'
+@conf
 def g95_flags(conf):
 	v=conf.env
-	v['FCFLAGS_fcshlib']=['-fPIC']
-	v['FORTRANMODFLAG']=['-fmod=','']
-	v['FCFLAGS_DEBUG']=['-Werror']
+	v.FCFLAGS_fcshlib=['-fPIC']
+	v.FORTRANMODFLAG=['-fmod=','']
+	v.FCFLAGS_DEBUG=['-Werror']
+@conf
 def g95_modifier_win32(conf):
 	fc_config.fortran_modifier_win32(conf)
+@conf
 def g95_modifier_cygwin(conf):
 	fc_config.fortran_modifier_cygwin(conf)
+@conf
 def g95_modifier_darwin(conf):
 	fc_config.fortran_modifier_darwin(conf)
+@conf
 def g95_modifier_platform(conf):
-	dest_os=conf.env['DEST_OS']or Utils.unversioned_sys_platform()
+	dest_os=conf.env.DEST_OS or Utils.unversioned_sys_platform()
 	g95_modifier_func=getattr(conf,'g95_modifier_'+dest_os,None)
 	if g95_modifier_func:
 		g95_modifier_func()
+@conf
 def get_g95_version(conf,fc):
 	version_re=re.compile(r"g95\s*(?P<major>\d*)\.(?P<minor>\d*)").search
 	cmd=fc+['--version']
@@ -38,18 +44,11 @@
 	if not match:
 		conf.fatal('cannot determine g95 version')
 	k=match.groupdict()
-	conf.env['FC_VERSION']=(k['major'],k['minor'])
+	conf.env.FC_VERSION=(k['major'],k['minor'])
 def configure(conf):
 	conf.find_g95()
 	conf.find_ar()
 	conf.fc_flags()
+	conf.fc_add_flags()
 	conf.g95_flags()
 	conf.g95_modifier_platform()
-
-conf(find_g95)
-conf(g95_flags)
-conf(g95_modifier_win32)
-conf(g95_modifier_cygwin)
-conf(g95_modifier_darwin)
-conf(g95_modifier_platform)
-conf(get_g95_version)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gas.py glmark2-2021.02/waflib/Tools/gas.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gas.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gas.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,12 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import waflib.Tools.asm
 from waflib.Tools import ar
 def configure(conf):
-	conf.find_program(['gas','as','gcc'],var='AS')
-	conf.env.AS_TGT_F=['-o']
+	conf.find_program(['gas','gcc'],var='AS')
+	conf.env.AS_TGT_F=['-c','-o']
 	conf.env.ASLNK_TGT_F=['-o']
 	conf.find_ar()
+	conf.load('asm')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gcc.py glmark2-2021.02/waflib/Tools/gcc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gcc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gcc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,80 +1,94 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
-from waflib import Configure,Options,Utils
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_gcc(conf):
 	cc=conf.find_program(['gcc','cc'],var='CC')
-	cc=conf.cmd_to_list(cc)
 	conf.get_cc_version(cc,gcc=True)
 	conf.env.CC_NAME='gcc'
-	conf.env.CC=cc
+@conf
 def gcc_common_flags(conf):
 	v=conf.env
-	v['CC_SRC_F']=[]
-	v['CC_TGT_F']=['-c','-o']
-	if not v['LINK_CC']:v['LINK_CC']=v['CC']
-	v['CCLNK_SRC_F']=[]
-	v['CCLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['RPATH_ST']='-Wl,-rpath,%s'
-	v['SONAME_ST']='-Wl,-h,%s'
-	v['SHLIB_MARKER']='-Wl,-Bdynamic'
-	v['STLIB_MARKER']='-Wl,-Bstatic'
-	v['cprogram_PATTERN']='%s'
-	v['CFLAGS_cshlib']=['-fPIC']
-	v['LINKFLAGS_cshlib']=['-shared']
-	v['cshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cstlib']=['-Wl,-Bstatic']
-	v['cstlib_PATTERN']='lib%s.a'
-	v['LINKFLAGS_MACBUNDLE']=['-bundle','-undefined','dynamic_lookup']
-	v['CFLAGS_MACBUNDLE']=['-fPIC']
-	v['macbundle_PATTERN']='%s.bundle'
+	v.CC_SRC_F=[]
+	v.CC_TGT_F=['-c','-o']
+	if not v.LINK_CC:
+		v.LINK_CC=v.CC
+	v.CCLNK_SRC_F=[]
+	v.CCLNK_TGT_F=['-o']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.RPATH_ST='-Wl,-rpath,%s'
+	v.SONAME_ST='-Wl,-h,%s'
+	v.SHLIB_MARKER='-Wl,-Bdynamic'
+	v.STLIB_MARKER='-Wl,-Bstatic'
+	v.cprogram_PATTERN='%s'
+	v.CFLAGS_cshlib=['-fPIC']
+	v.LINKFLAGS_cshlib=['-shared']
+	v.cshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cstlib=['-Wl,-Bstatic']
+	v.cstlib_PATTERN='lib%s.a'
+	v.LINKFLAGS_MACBUNDLE=['-bundle','-undefined','dynamic_lookup']
+	v.CFLAGS_MACBUNDLE=['-fPIC']
+	v.macbundle_PATTERN='%s.bundle'
+@conf
 def gcc_modifier_win32(conf):
 	v=conf.env
-	v['cprogram_PATTERN']='%s.exe'
-	v['cshlib_PATTERN']='%s.dll'
-	v['implib_PATTERN']='lib%s.dll.a'
-	v['IMPLIB_ST']='-Wl,--out-implib,%s'
-	v['CFLAGS_cshlib']=[]
-	v.append_value('CFLAGS_cshlib',['-DDLL_EXPORT'])
+	v.cprogram_PATTERN='%s.exe'
+	v.cshlib_PATTERN='%s.dll'
+	v.implib_PATTERN='lib%s.dll.a'
+	v.IMPLIB_ST='-Wl,--out-implib,%s'
+	v.CFLAGS_cshlib=[]
 	v.append_value('LINKFLAGS',['-Wl,--enable-auto-import'])
+@conf
 def gcc_modifier_cygwin(conf):
 	gcc_modifier_win32(conf)
 	v=conf.env
-	v['cshlib_PATTERN']='cyg%s.dll'
+	v.cshlib_PATTERN='cyg%s.dll'
 	v.append_value('LINKFLAGS_cshlib',['-Wl,--enable-auto-image-base'])
-	v['CFLAGS_cshlib']=[]
+	v.CFLAGS_cshlib=[]
+@conf
 def gcc_modifier_darwin(conf):
 	v=conf.env
-	v['CFLAGS_cshlib']=['-fPIC','-compatibility_version','1','-current_version','1']
-	v['LINKFLAGS_cshlib']=['-dynamiclib']
-	v['cshlib_PATTERN']='lib%s.dylib'
-	v['FRAMEWORKPATH_ST']='-F%s'
-	v['FRAMEWORK_ST']=['-framework']
-	v['ARCH_ST']=['-arch']
-	v['LINKFLAGS_cstlib']=[]
-	v['SHLIB_MARKER']=[]
-	v['STLIB_MARKER']=[]
-	v['SONAME_ST']=[]
+	v.CFLAGS_cshlib=['-fPIC']
+	v.LINKFLAGS_cshlib=['-dynamiclib']
+	v.cshlib_PATTERN='lib%s.dylib'
+	v.FRAMEWORKPATH_ST='-F%s'
+	v.FRAMEWORK_ST=['-framework']
+	v.ARCH_ST=['-arch']
+	v.LINKFLAGS_cstlib=[]
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.SONAME_ST=[]
+@conf
 def gcc_modifier_aix(conf):
 	v=conf.env
-	v['LINKFLAGS_cprogram']=['-Wl,-brtl']
-	v['LINKFLAGS_cshlib']=['-shared','-Wl,-brtl,-bexpfull']
-	v['SHLIB_MARKER']=[]
+	v.LINKFLAGS_cprogram=['-Wl,-brtl']
+	v.LINKFLAGS_cshlib=['-shared','-Wl,-brtl,-bexpfull']
+	v.SHLIB_MARKER=[]
+@conf
 def gcc_modifier_hpux(conf):
 	v=conf.env
-	v['SHLIB_MARKER']=[]
-	v['CFLAGS_cshlib']=['-fPIC','-DPIC']
-	v['cshlib_PATTERN']='lib%s.sl'
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.CFLAGS_cshlib=['-fPIC','-DPIC']
+	v.cshlib_PATTERN='lib%s.sl'
+@conf
+def gcc_modifier_openbsd(conf):
+	conf.env.SONAME_ST=[]
+@conf
+def gcc_modifier_osf1V(conf):
+	v=conf.env
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.SONAME_ST=[]
+@conf
 def gcc_modifier_platform(conf):
 	gcc_modifier_func=getattr(conf,'gcc_modifier_'+conf.env.DEST_OS,None)
 	if gcc_modifier_func:
@@ -87,12 +101,3 @@
 	conf.cc_load_tools()
 	conf.cc_add_flags()
 	conf.link_add_flags()
-
-conf(find_gcc)
-conf(gcc_common_flags)
-conf(gcc_modifier_win32)
-conf(gcc_modifier_cygwin)
-conf(gcc_modifier_darwin)
-conf(gcc_modifier_aix)
-conf(gcc_modifier_hpux)
-conf(gcc_modifier_platform)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gdc.py glmark2-2021.02/waflib/Tools/gdc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gdc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gdc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,34 +1,35 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
 from waflib.Tools import ar,d
 from waflib.Configure import conf
+@conf
 def find_gdc(conf):
 	conf.find_program('gdc',var='D')
+	out=conf.cmd_and_log(conf.env.D+['--version'])
+	if out.find("gdc")==-1:
+		conf.fatal("detected compiler is not gdc")
+@conf
 def common_flags_gdc(conf):
 	v=conf.env
-	v['DFLAGS']=[]
-	v['D_SRC_F']=['-c']
-	v['D_TGT_F']='-o%s'
-	v['D_LINKER']=v['D']
-	v['DLNK_SRC_F']=''
-	v['DLNK_TGT_F']='-o%s'
-	v['DINC_ST']='-I%s'
-	v['DSHLIB_MARKER']=v['DSTLIB_MARKER']=''
-	v['DSTLIB_ST']=v['DSHLIB_ST']='-l%s'
-	v['DSTLIBPATH_ST']=v['DLIBPATH_ST']='-L%s'
-	v['LINKFLAGS_dshlib']=['-shared']
-	v['DHEADER_ext']='.di'
+	v.DFLAGS=[]
+	v.D_SRC_F=['-c']
+	v.D_TGT_F='-o%s'
+	v.D_LINKER=v.D
+	v.DLNK_SRC_F=''
+	v.DLNK_TGT_F='-o%s'
+	v.DINC_ST='-I%s'
+	v.DSHLIB_MARKER=v.DSTLIB_MARKER=''
+	v.DSTLIB_ST=v.DSHLIB_ST='-l%s'
+	v.DSTLIBPATH_ST=v.DLIBPATH_ST='-L%s'
+	v.LINKFLAGS_dshlib=['-shared']
+	v.DHEADER_ext='.di'
 	v.DFLAGS_d_with_header='-fintfc'
-	v['D_HDR_F']='-fintfc-file=%s'
+	v.D_HDR_F='-fintfc-file=%s'
 def configure(conf):
 	conf.find_gdc()
 	conf.load('ar')
 	conf.load('d')
 	conf.common_flags_gdc()
 	conf.d_platform_flags()
-
-conf(find_gdc)
-conf(common_flags_gdc)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gfortran.py glmark2-2021.02/waflib/Tools/gfortran.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gfortran.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gfortran.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,32 +1,38 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import re
 from waflib import Utils
-from waflib.Tools import fc,fc_config,fc_scan
+from waflib.Tools import fc,fc_config,fc_scan,ar
 from waflib.Configure import conf
+@conf
 def find_gfortran(conf):
 	fc=conf.find_program(['gfortran','g77'],var='FC')
-	fc=conf.cmd_to_list(fc)
 	conf.get_gfortran_version(fc)
 	conf.env.FC_NAME='GFORTRAN'
+@conf
 def gfortran_flags(conf):
 	v=conf.env
-	v['FCFLAGS_fcshlib']=['-fPIC']
-	v['FORTRANMODFLAG']=['-J','']
-	v['FCFLAGS_DEBUG']=['-Werror']
+	v.FCFLAGS_fcshlib=['-fPIC']
+	v.FORTRANMODFLAG=['-J','']
+	v.FCFLAGS_DEBUG=['-Werror']
+@conf
 def gfortran_modifier_win32(conf):
 	fc_config.fortran_modifier_win32(conf)
+@conf
 def gfortran_modifier_cygwin(conf):
 	fc_config.fortran_modifier_cygwin(conf)
+@conf
 def gfortran_modifier_darwin(conf):
 	fc_config.fortran_modifier_darwin(conf)
+@conf
 def gfortran_modifier_platform(conf):
-	dest_os=conf.env['DEST_OS']or Utils.unversioned_sys_platform()
+	dest_os=conf.env.DEST_OS or Utils.unversioned_sys_platform()
 	gfortran_modifier_func=getattr(conf,'gfortran_modifier_'+dest_os,None)
 	if gfortran_modifier_func:
 		gfortran_modifier_func()
+@conf
 def get_gfortran_version(conf,fc):
 	version_re=re.compile(r"GNU\s*Fortran",re.I).search
 	cmd=fc+['--version']
@@ -40,7 +46,7 @@
 	if out.find('__GNUC__')<0:
 		conf.fatal('Could not determine the compiler type')
 	k={}
-	out=out.split('\n')
+	out=out.splitlines()
 	import shlex
 	for line in out:
 		lst=shlex.split(line)
@@ -52,18 +58,11 @@
 		return var in k
 	def isT(var):
 		return var in k and k[var]!='0'
-	conf.env['FC_VERSION']=(k['__GNUC__'],k['__GNUC_MINOR__'],k['__GNUC_PATCHLEVEL__'])
+	conf.env.FC_VERSION=(k['__GNUC__'],k['__GNUC_MINOR__'],k['__GNUC_PATCHLEVEL__'])
 def configure(conf):
 	conf.find_gfortran()
 	conf.find_ar()
 	conf.fc_flags()
+	conf.fc_add_flags()
 	conf.gfortran_flags()
 	conf.gfortran_modifier_platform()
-
-conf(find_gfortran)
-conf(gfortran_flags)
-conf(gfortran_modifier_win32)
-conf(gfortran_modifier_cygwin)
-conf(gfortran_modifier_darwin)
-conf(gfortran_modifier_platform)
-conf(get_gfortran_version)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/glib2.py glmark2-2021.02/waflib/Tools/glib2.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/glib2.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/glib2.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,18 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os
-from waflib import Task,Utils,Options,Errors,Logs
-from waflib.TaskGen import taskgen_method,before_method,after_method,feature
+from waflib import Context,Task,Utils,Options,Errors,Logs
+from waflib.TaskGen import taskgen_method,before_method,feature,extension
+from waflib.Configure import conf
+@taskgen_method
 def add_marshal_file(self,filename,prefix):
 	if not hasattr(self,'marshal_list'):
 		self.marshal_list=[]
 	self.meths.append('process_marshal')
 	self.marshal_list.append((filename,prefix))
+@before_method('process_source')
 def process_marshal(self):
 	for f,prefix in getattr(self,'marshal_list',[]):
 		node=self.path.find_resource(f)
@@ -22,8 +25,11 @@
 	self.source=self.to_nodes(getattr(self,'source',[]))
 	self.source.append(c_node)
 class glib_genmarshal(Task.Task):
+	vars=['GLIB_GENMARSHAL_PREFIX','GLIB_GENMARSHAL']
+	color='BLUE'
+	ext_out=['.h']
 	def run(self):
-		bld=self.inputs[0].__class__.ctx
+		bld=self.generator.bld
 		get=self.env.get_flat
 		cmd1="%s %s --prefix=%s --header > %s"%(get('GLIB_GENMARSHAL'),self.inputs[0].srcpath(),get('GLIB_GENMARSHAL_PREFIX'),self.outputs[0].abspath())
 		ret=bld.exec_command(cmd1)
@@ -32,19 +38,19 @@
 		self.outputs[1].write(c)
 		cmd2="%s %s --prefix=%s --body >> %s"%(get('GLIB_GENMARSHAL'),self.inputs[0].srcpath(),get('GLIB_GENMARSHAL_PREFIX'),self.outputs[1].abspath())
 		return bld.exec_command(cmd2)
-	vars=['GLIB_GENMARSHAL_PREFIX','GLIB_GENMARSHAL']
-	color='BLUE'
-	ext_out=['.h']
+@taskgen_method
 def add_enums_from_template(self,source='',target='',template='',comments=''):
 	if not hasattr(self,'enums_list'):
 		self.enums_list=[]
 	self.meths.append('process_enums')
 	self.enums_list.append({'source':source,'target':target,'template':template,'file-head':'','file-prod':'','file-tail':'','enum-prod':'','value-head':'','value-prod':'','value-tail':'','comments':comments})
+@taskgen_method
 def add_enums(self,source='',target='',file_head='',file_prod='',file_tail='',enum_prod='',value_head='',value_prod='',value_tail='',comments=''):
 	if not hasattr(self,'enums_list'):
 		self.enums_list=[]
 	self.meths.append('process_enums')
 	self.enums_list.append({'source':source,'template':'','target':target,'file-head':file_head,'file-prod':file_prod,'file-tail':file_tail,'enum-prod':enum_prod,'value-head':value_head,'value-prod':value_prod,'value-tail':value_tail,'comments':comments})
+@before_method('process_source')
 def process_enums(self):
 	for enum in getattr(self,'enums_list',[]):
 		task=self.create_task('glib_mkenums')
@@ -55,13 +61,13 @@
 			raise Errors.WafError('missing source '+str(enum))
 		source_list=[self.path.find_resource(k)for k in source_list]
 		inputs+=source_list
-		env['GLIB_MKENUMS_SOURCE']=[k.abspath()for k in source_list]
+		env.GLIB_MKENUMS_SOURCE=[k.abspath()for k in source_list]
 		if not enum['target']:
 			raise Errors.WafError('missing target '+str(enum))
 		tgt_node=self.path.find_or_declare(enum['target'])
 		if tgt_node.name.endswith('.c'):
 			self.source.append(tgt_node)
-		env['GLIB_MKENUMS_TARGET']=tgt_node.abspath()
+		env.GLIB_MKENUMS_TARGET=tgt_node.abspath()
 		options=[]
 		if enum['template']:
 			template_node=self.path.find_resource(enum['template'])
@@ -71,104 +77,154 @@
 		for param,option in params.items():
 			if enum[param]:
 				options.append('%s %r'%(option,enum[param]))
-		env['GLIB_MKENUMS_OPTIONS']=' '.join(options)
+		env.GLIB_MKENUMS_OPTIONS=' '.join(options)
 		task.set_inputs(inputs)
 		task.set_outputs(tgt_node)
 class glib_mkenums(Task.Task):
 	run_str='${GLIB_MKENUMS} ${GLIB_MKENUMS_OPTIONS} ${GLIB_MKENUMS_SOURCE} > ${GLIB_MKENUMS_TARGET}'
 	color='PINK'
 	ext_out=['.h']
+@taskgen_method
 def add_settings_schemas(self,filename_list):
 	if not hasattr(self,'settings_schema_files'):
 		self.settings_schema_files=[]
 	if not isinstance(filename_list,list):
 		filename_list=[filename_list]
 	self.settings_schema_files.extend(filename_list)
+@taskgen_method
 def add_settings_enums(self,namespace,filename_list):
 	if hasattr(self,'settings_enum_namespace'):
-		raise Errors.WafError("Tried to add gsettings enums to '%s' more than once"%self.name)
+		raise Errors.WafError("Tried to add gsettings enums to %r more than once"%self.name)
 	self.settings_enum_namespace=namespace
 	if type(filename_list)!='list':
 		filename_list=[filename_list]
 	self.settings_enum_files=filename_list
-def r_change_ext(self,ext):
-	name=self.name
-	k=name.rfind('.')
-	if k>=0:
-		name=name[:k]+ext
-	else:
-		name=name+ext
-	return self.parent.find_or_declare([name])
+@feature('glib2')
 def process_settings(self):
 	enums_tgt_node=[]
 	install_files=[]
 	settings_schema_files=getattr(self,'settings_schema_files',[])
-	if settings_schema_files and not self.env['GLIB_COMPILE_SCHEMAS']:
+	if settings_schema_files and not self.env.GLIB_COMPILE_SCHEMAS:
 		raise Errors.WafError("Unable to process GSettings schemas - glib-compile-schemas was not found during configure")
 	if hasattr(self,'settings_enum_files'):
 		enums_task=self.create_task('glib_mkenums')
 		source_list=self.settings_enum_files
 		source_list=[self.path.find_resource(k)for k in source_list]
 		enums_task.set_inputs(source_list)
-		enums_task.env['GLIB_MKENUMS_SOURCE']=[k.abspath()for k in source_list]
+		enums_task.env.GLIB_MKENUMS_SOURCE=[k.abspath()for k in source_list]
 		target=self.settings_enum_namespace+'.enums.xml'
 		tgt_node=self.path.find_or_declare(target)
 		enums_task.set_outputs(tgt_node)
-		enums_task.env['GLIB_MKENUMS_TARGET']=tgt_node.abspath()
+		enums_task.env.GLIB_MKENUMS_TARGET=tgt_node.abspath()
 		enums_tgt_node=[tgt_node]
 		install_files.append(tgt_node)
 		options='--comments "<!-- @comment@ -->" --fhead "<schemalist>" --vhead "  <@type@ id=\\"%s.@EnumName@\\">" --vprod "    <value nick=\\"@valuenick@\\" value=\\"@valuenum@\\"/>" --vtail "  </@type@>" --ftail "</schemalist>" '%(self.settings_enum_namespace)
-		enums_task.env['GLIB_MKENUMS_OPTIONS']=options
+		enums_task.env.GLIB_MKENUMS_OPTIONS=options
 	for schema in settings_schema_files:
 		schema_task=self.create_task('glib_validate_schema')
 		schema_node=self.path.find_resource(schema)
 		if not schema_node:
-			raise Errors.WafError("Cannot find the schema file '%s'"%schema)
+			raise Errors.WafError("Cannot find the schema file %r"%schema)
 		install_files.append(schema_node)
 		source_list=enums_tgt_node+[schema_node]
 		schema_task.set_inputs(source_list)
-		schema_task.env['GLIB_COMPILE_SCHEMAS_OPTIONS']=[("--schema-file="+k.abspath())for k in source_list]
-		target_node=r_change_ext(schema_node,'.xml.valid')
+		schema_task.env.GLIB_COMPILE_SCHEMAS_OPTIONS=[("--schema-file="+k.abspath())for k in source_list]
+		target_node=schema_node.change_ext('.xml.valid')
 		schema_task.set_outputs(target_node)
-		schema_task.env['GLIB_VALIDATE_SCHEMA_OUTPUT']=target_node.abspath()
+		schema_task.env.GLIB_VALIDATE_SCHEMA_OUTPUT=target_node.abspath()
 	def compile_schemas_callback(bld):
 		if not bld.is_install:return
 		Logs.pprint('YELLOW','Updating GSettings schema cache')
 		command=Utils.subst_vars("${GLIB_COMPILE_SCHEMAS} ${GSETTINGSSCHEMADIR}",bld.env)
-		ret=self.bld.exec_command(command)
+		self.bld.exec_command(command)
 	if self.bld.is_install:
-		if not self.env['GSETTINGSSCHEMADIR']:
+		if not self.env.GSETTINGSSCHEMADIR:
 			raise Errors.WafError('GSETTINGSSCHEMADIR not defined (should have been set up automatically during configure)')
 		if install_files:
-			self.bld.install_files(self.env['GSETTINGSSCHEMADIR'],install_files)
+			self.add_install_files(install_to=self.env.GSETTINGSSCHEMADIR,install_from=install_files)
 			if not hasattr(self.bld,'_compile_schemas_registered'):
 				self.bld.add_post_fun(compile_schemas_callback)
 				self.bld._compile_schemas_registered=True
 class glib_validate_schema(Task.Task):
 	run_str='rm -f ${GLIB_VALIDATE_SCHEMA_OUTPUT} && ${GLIB_COMPILE_SCHEMAS} --dry-run ${GLIB_COMPILE_SCHEMAS_OPTIONS} && touch ${GLIB_VALIDATE_SCHEMA_OUTPUT}'
 	color='PINK'
-def configure(conf):
+@extension('.gresource.xml')
+def process_gresource_source(self,node):
+	if not self.env.GLIB_COMPILE_RESOURCES:
+		raise Errors.WafError("Unable to process GResource file - glib-compile-resources was not found during configure")
+	if'gresource'in self.features:
+		return
+	h_node=node.change_ext('_xml.h')
+	c_node=node.change_ext('_xml.c')
+	self.create_task('glib_gresource_source',node,[h_node,c_node])
+	self.source.append(c_node)
+@feature('gresource')
+def process_gresource_bundle(self):
+	for i in self.to_list(self.source):
+		node=self.path.find_resource(i)
+		task=self.create_task('glib_gresource_bundle',node,node.change_ext(''))
+		inst_to=getattr(self,'install_path',None)
+		if inst_to:
+			self.add_install_files(install_to=inst_to,install_from=task.outputs)
+class glib_gresource_base(Task.Task):
+	color='BLUE'
+	base_cmd='${GLIB_COMPILE_RESOURCES} --sourcedir=${SRC[0].parent.srcpath()} --sourcedir=${SRC[0].bld_dir()}'
+	def scan(self):
+		bld=self.generator.bld
+		kw={}
+		kw['cwd']=self.get_cwd()
+		kw['quiet']=Context.BOTH
+		cmd=Utils.subst_vars('${GLIB_COMPILE_RESOURCES} --sourcedir=%s --sourcedir=%s --generate-dependencies %s'%(self.inputs[0].parent.srcpath(),self.inputs[0].bld_dir(),self.inputs[0].bldpath()),self.env)
+		output=bld.cmd_and_log(cmd,**kw)
+		nodes=[]
+		names=[]
+		for dep in output.splitlines():
+			if dep:
+				node=bld.bldnode.find_node(dep)
+				if node:
+					nodes.append(node)
+				else:
+					names.append(dep)
+		return(nodes,names)
+class glib_gresource_source(glib_gresource_base):
+	vars=['GLIB_COMPILE_RESOURCES']
+	fun_h=Task.compile_fun_shell(glib_gresource_base.base_cmd+' --target=${TGT[0].abspath()} --generate-header ${SRC}')
+	fun_c=Task.compile_fun_shell(glib_gresource_base.base_cmd+' --target=${TGT[1].abspath()} --generate-source ${SRC}')
+	ext_out=['.h']
+	def run(self):
+		return self.fun_h[0](self)or self.fun_c[0](self)
+class glib_gresource_bundle(glib_gresource_base):
+	run_str=glib_gresource_base.base_cmd+' --target=${TGT} ${SRC}'
+	shell=True
+@conf
+def find_glib_genmarshal(conf):
 	conf.find_program('glib-genmarshal',var='GLIB_GENMARSHAL')
-	conf.find_perl_program('glib-mkenums',var='GLIB_MKENUMS')
-	conf.find_program('glib-compile-schemas',var='GLIB_COMPILE_SCHEMAS',mandatory=False)
+@conf
+def find_glib_mkenums(conf):
+	if not conf.env.PERL:
+		conf.find_program('perl',var='PERL')
+	conf.find_program('glib-mkenums',interpreter='PERL',var='GLIB_MKENUMS')
+@conf
+def find_glib_compile_schemas(conf):
+	conf.find_program('glib-compile-schemas',var='GLIB_COMPILE_SCHEMAS')
 	def getstr(varname):
 		return getattr(Options.options,varname,getattr(conf.env,varname,''))
 	gsettingsschemadir=getstr('GSETTINGSSCHEMADIR')
 	if not gsettingsschemadir:
 		datadir=getstr('DATADIR')
 		if not datadir:
-			prefix=conf.env['PREFIX']
+			prefix=conf.env.PREFIX
 			datadir=os.path.join(prefix,'share')
 		gsettingsschemadir=os.path.join(datadir,'glib-2.0','schemas')
-	conf.env['GSETTINGSSCHEMADIR']=gsettingsschemadir
+	conf.env.GSETTINGSSCHEMADIR=gsettingsschemadir
+@conf
+def find_glib_compile_resources(conf):
+	conf.find_program('glib-compile-resources',var='GLIB_COMPILE_RESOURCES')
+def configure(conf):
+	conf.find_glib_genmarshal()
+	conf.find_glib_mkenums()
+	conf.find_glib_compile_schemas(mandatory=False)
+	conf.find_glib_compile_resources(mandatory=False)
 def options(opt):
-	opt.add_option('--gsettingsschemadir',help='GSettings schema location [Default: ${datadir}/glib-2.0/schemas]',default='',dest='GSETTINGSSCHEMADIR')
-
-taskgen_method(add_marshal_file)
-before_method('process_source')(process_marshal)
-taskgen_method(add_enums_from_template)
-taskgen_method(add_enums)
-before_method('process_source')(process_enums)
-taskgen_method(add_settings_schemas)
-taskgen_method(add_settings_enums)
-feature('glib2')(process_settings)
\ No newline at end of file
+	gr=opt.add_option_group('Installation directories')
+	gr.add_option('--gsettingsschemadir',help='GSettings schema location [DATADIR/glib-2.0/schemas]',default='',dest='GSETTINGSSCHEMADIR')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gnu_dirs.py glmark2-2021.02/waflib/Tools/gnu_dirs.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gnu_dirs.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gnu_dirs.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,37 +1,38 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os
+import os,re
 from waflib import Utils,Options,Context
-_options=[x.split(', ')for x in'''
-bindir, user executables, ${EXEC_PREFIX}/bin
-sbindir, system admin executables, ${EXEC_PREFIX}/sbin
-libexecdir, program executables, ${EXEC_PREFIX}/libexec
-sysconfdir, read-only single-machine data, ${PREFIX}/etc
-sharedstatedir, modifiable architecture-independent data, ${PREFIX}/com
-localstatedir, modifiable single-machine data, ${PREFIX}/var
-libdir, object code libraries, ${EXEC_PREFIX}/lib
-includedir, C header files, ${PREFIX}/include
-oldincludedir, C header files for non-gcc, /usr/include
-datarootdir, read-only arch.-independent data root, ${PREFIX}/share
-datadir, read-only architecture-independent data, ${DATAROOTDIR}
-infodir, info documentation, ${DATAROOTDIR}/info
+gnuopts='''
+bindir, user commands, ${EXEC_PREFIX}/bin
+sbindir, system binaries, ${EXEC_PREFIX}/sbin
+libexecdir, program-specific binaries, ${EXEC_PREFIX}/libexec
+sysconfdir, host-specific configuration, ${PREFIX}/etc
+sharedstatedir, architecture-independent variable data, ${PREFIX}/com
+localstatedir, variable data, ${PREFIX}/var
+libdir, object code libraries, ${EXEC_PREFIX}/lib%s
+includedir, header files, ${PREFIX}/include
+oldincludedir, header files for non-GCC compilers, /usr/include
+datarootdir, architecture-independent data root, ${PREFIX}/share
+datadir, architecture-independent data, ${DATAROOTDIR}
+infodir, GNU "info" documentation, ${DATAROOTDIR}/info
 localedir, locale-dependent data, ${DATAROOTDIR}/locale
-mandir, man documentation, ${DATAROOTDIR}/man
+mandir, manual pages, ${DATAROOTDIR}/man
 docdir, documentation root, ${DATAROOTDIR}/doc/${PACKAGE}
-htmldir, html documentation, ${DOCDIR}
-dvidir, dvi documentation, ${DOCDIR}
-pdfdir, pdf documentation, ${DOCDIR}
-psdir, ps documentation, ${DOCDIR}
-'''.split('\n')if x]
+htmldir, HTML documentation, ${DOCDIR}
+dvidir, DVI documentation, ${DOCDIR}
+pdfdir, PDF documentation, ${DOCDIR}
+psdir, PostScript documentation, ${DOCDIR}
+'''%Utils.lib64()
+_options=[x.split(', ')for x in gnuopts.splitlines()if x]
 def configure(conf):
 	def get_param(varname,default):
 		return getattr(Options.options,varname,'')or default
 	env=conf.env
-	conf.env.LIBDIR=conf.env.BINDIR=[]
-	env['EXEC_PREFIX']=get_param('EXEC_PREFIX',env['PREFIX'])
-	env['PACKAGE']=getattr(Context.g_module,'APPNAME',None)or env['PACKAGE']
+	env.LIBDIR=env.BINDIR=[]
+	env.EXEC_PREFIX=get_param('EXEC_PREFIX',env.PREFIX)
+	env.PACKAGE=getattr(Context.g_module,'APPNAME',None)or env.PACKAGE
 	complete=False
 	iter=0
 	while not complete and iter<len(_options)+1:
@@ -45,10 +46,10 @@
 				except TypeError:
 					complete=False
 	if not complete:
-		lst=[name for name,_,_ in _options if not env[name.upper()]]
+		lst=[x for x,_,_ in _options if not env[x.upper()]]
 		raise conf.errors.WafError('Variable substitution failure %r'%lst)
 def options(opt):
-	inst_dir=opt.add_option_group('Installation directories','By default, "waf install" will put the files in\
+	inst_dir=opt.add_option_group('Installation prefix','By default, "waf install" will put the files in\
  "/usr/local/bin", "/usr/local/lib" etc. An installation prefix other\
  than "/usr/local" can be given using "--prefix", for example "--prefix=$HOME"')
 	for k in('--prefix','--destdir'):
@@ -56,10 +57,10 @@
 		if option:
 			opt.parser.remove_option(k)
 			inst_dir.add_option(option)
-	inst_dir.add_option('--exec-prefix',help='installation prefix [Default: ${PREFIX}]',default='',dest='EXEC_PREFIX')
-	dirs_options=opt.add_option_group('Pre-defined installation directories','')
+	inst_dir.add_option('--exec-prefix',help='installation prefix for binaries [PREFIX]',default='',dest='EXEC_PREFIX')
+	dirs_options=opt.add_option_group('Installation directories')
 	for name,help,default in _options:
 		option_name='--'+name
 		str_default=default
-		str_help='%s [Default: %s]'%(help,str_default)
+		str_help='%s [%s]'%(help,re.sub(r'\$\{([^}]+)\}',r'\1',str_default))
 		dirs_options.add_option(option_name,help=str_help,default='',dest=name.upper())
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gxx.py glmark2-2021.02/waflib/Tools/gxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/gxx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/gxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,80 +1,94 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
-from waflib import Configure,Options,Utils
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_gxx(conf):
 	cxx=conf.find_program(['g++','c++'],var='CXX')
-	cxx=conf.cmd_to_list(cxx)
 	conf.get_cc_version(cxx,gcc=True)
 	conf.env.CXX_NAME='gcc'
-	conf.env.CXX=cxx
+@conf
 def gxx_common_flags(conf):
 	v=conf.env
-	v['CXX_SRC_F']=[]
-	v['CXX_TGT_F']=['-c','-o']
-	if not v['LINK_CXX']:v['LINK_CXX']=v['CXX']
-	v['CXXLNK_SRC_F']=[]
-	v['CXXLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['RPATH_ST']='-Wl,-rpath,%s'
-	v['SONAME_ST']='-Wl,-h,%s'
-	v['SHLIB_MARKER']='-Wl,-Bdynamic'
-	v['STLIB_MARKER']='-Wl,-Bstatic'
-	v['cxxprogram_PATTERN']='%s'
-	v['CXXFLAGS_cxxshlib']=['-fPIC']
-	v['LINKFLAGS_cxxshlib']=['-shared']
-	v['cxxshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cxxstlib']=['-Wl,-Bstatic']
-	v['cxxstlib_PATTERN']='lib%s.a'
-	v['LINKFLAGS_MACBUNDLE']=['-bundle','-undefined','dynamic_lookup']
-	v['CXXFLAGS_MACBUNDLE']=['-fPIC']
-	v['macbundle_PATTERN']='%s.bundle'
+	v.CXX_SRC_F=[]
+	v.CXX_TGT_F=['-c','-o']
+	if not v.LINK_CXX:
+		v.LINK_CXX=v.CXX
+	v.CXXLNK_SRC_F=[]
+	v.CXXLNK_TGT_F=['-o']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.RPATH_ST='-Wl,-rpath,%s'
+	v.SONAME_ST='-Wl,-h,%s'
+	v.SHLIB_MARKER='-Wl,-Bdynamic'
+	v.STLIB_MARKER='-Wl,-Bstatic'
+	v.cxxprogram_PATTERN='%s'
+	v.CXXFLAGS_cxxshlib=['-fPIC']
+	v.LINKFLAGS_cxxshlib=['-shared']
+	v.cxxshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cxxstlib=['-Wl,-Bstatic']
+	v.cxxstlib_PATTERN='lib%s.a'
+	v.LINKFLAGS_MACBUNDLE=['-bundle','-undefined','dynamic_lookup']
+	v.CXXFLAGS_MACBUNDLE=['-fPIC']
+	v.macbundle_PATTERN='%s.bundle'
+@conf
 def gxx_modifier_win32(conf):
 	v=conf.env
-	v['cxxprogram_PATTERN']='%s.exe'
-	v['cxxshlib_PATTERN']='%s.dll'
-	v['implib_PATTERN']='lib%s.dll.a'
-	v['IMPLIB_ST']='-Wl,--out-implib,%s'
-	v['CXXFLAGS_cxxshlib']=[]
-	v.append_value('CXXFLAGS_cxxshlib',['-DDLL_EXPORT'])
+	v.cxxprogram_PATTERN='%s.exe'
+	v.cxxshlib_PATTERN='%s.dll'
+	v.implib_PATTERN='lib%s.dll.a'
+	v.IMPLIB_ST='-Wl,--out-implib,%s'
+	v.CXXFLAGS_cxxshlib=[]
 	v.append_value('LINKFLAGS',['-Wl,--enable-auto-import'])
+@conf
 def gxx_modifier_cygwin(conf):
 	gxx_modifier_win32(conf)
 	v=conf.env
-	v['cxxshlib_PATTERN']='cyg%s.dll'
+	v.cxxshlib_PATTERN='cyg%s.dll'
 	v.append_value('LINKFLAGS_cxxshlib',['-Wl,--enable-auto-image-base'])
-	v['CXXFLAGS_cxxshlib']=[]
+	v.CXXFLAGS_cxxshlib=[]
+@conf
 def gxx_modifier_darwin(conf):
 	v=conf.env
-	v['CXXFLAGS_cxxshlib']=['-fPIC','-compatibility_version','1','-current_version','1']
-	v['LINKFLAGS_cxxshlib']=['-dynamiclib']
-	v['cxxshlib_PATTERN']='lib%s.dylib'
-	v['FRAMEWORKPATH_ST']='-F%s'
-	v['FRAMEWORK_ST']=['-framework']
-	v['ARCH_ST']=['-arch']
-	v['LINKFLAGS_cxxstlib']=[]
-	v['SHLIB_MARKER']=[]
-	v['STLIB_MARKER']=[]
-	v['SONAME_ST']=[]
+	v.CXXFLAGS_cxxshlib=['-fPIC']
+	v.LINKFLAGS_cxxshlib=['-dynamiclib']
+	v.cxxshlib_PATTERN='lib%s.dylib'
+	v.FRAMEWORKPATH_ST='-F%s'
+	v.FRAMEWORK_ST=['-framework']
+	v.ARCH_ST=['-arch']
+	v.LINKFLAGS_cxxstlib=[]
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.SONAME_ST=[]
+@conf
 def gxx_modifier_aix(conf):
 	v=conf.env
-	v['LINKFLAGS_cxxprogram']=['-Wl,-brtl']
-	v['LINKFLAGS_cxxshlib']=['-shared','-Wl,-brtl,-bexpfull']
-	v['SHLIB_MARKER']=[]
+	v.LINKFLAGS_cxxprogram=['-Wl,-brtl']
+	v.LINKFLAGS_cxxshlib=['-shared','-Wl,-brtl,-bexpfull']
+	v.SHLIB_MARKER=[]
+@conf
 def gxx_modifier_hpux(conf):
 	v=conf.env
-	v['SHLIB_MARKER']=[]
-	v['CFLAGS_cxxshlib']=['-fPIC','-DPIC']
-	v['cxxshlib_PATTERN']='lib%s.sl'
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.CFLAGS_cxxshlib=['-fPIC','-DPIC']
+	v.cxxshlib_PATTERN='lib%s.sl'
+@conf
+def gxx_modifier_openbsd(conf):
+	conf.env.SONAME_ST=[]
+@conf
+def gcc_modifier_osf1V(conf):
+	v=conf.env
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.SONAME_ST=[]
+@conf
 def gxx_modifier_platform(conf):
 	gxx_modifier_func=getattr(conf,'gxx_modifier_'+conf.env.DEST_OS,None)
 	if gxx_modifier_func:
@@ -87,12 +101,3 @@
 	conf.cxx_load_tools()
 	conf.cxx_add_flags()
 	conf.link_add_flags()
-
-conf(find_gxx)
-conf(gxx_common_flags)
-conf(gxx_modifier_win32)
-conf(gxx_modifier_cygwin)
-conf(gxx_modifier_darwin)
-conf(gxx_modifier_aix)
-conf(gxx_modifier_hpux)
-conf(gxx_modifier_platform)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/icc.py glmark2-2021.02/waflib/Tools/icc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/icc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/icc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,24 +1,15 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
+import sys
 from waflib.Tools import ccroot,ar,gcc
 from waflib.Configure import conf
+@conf
 def find_icc(conf):
-	if sys.platform=='cygwin':
-		conf.fatal('The Intel compiler does not work on Cygwin')
-	v=conf.env
-	cc=None
-	if v['CC']:cc=v['CC']
-	elif'CC'in conf.environ:cc=conf.environ['CC']
-	if not cc:cc=conf.find_program('icc',var='CC')
-	if not cc:cc=conf.find_program('ICL',var='CC')
-	if not cc:conf.fatal('Intel C Compiler (icc) was not found')
-	cc=conf.cmd_to_list(cc)
+	cc=conf.find_program(['icc','ICL'],var='CC')
 	conf.get_cc_version(cc,icc=True)
-	v['CC']=cc
-	v['CC_NAME']='icc'
+	conf.env.CC_NAME='icc'
 def configure(conf):
 	conf.find_icc()
 	conf.find_ar()
@@ -27,5 +18,3 @@
 	conf.cc_load_tools()
 	conf.cc_add_flags()
 	conf.link_add_flags()
-
-conf(find_icc)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/icpc.py glmark2-2021.02/waflib/Tools/icpc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/icpc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/icpc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,23 +1,15 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
+import sys
 from waflib.Tools import ccroot,ar,gxx
 from waflib.Configure import conf
+@conf
 def find_icpc(conf):
-	if sys.platform=='cygwin':
-		conf.fatal('The Intel compiler does not work on Cygwin')
-	v=conf.env
-	cxx=None
-	if v['CXX']:cxx=v['CXX']
-	elif'CXX'in conf.environ:cxx=conf.environ['CXX']
-	if not cxx:cxx=conf.find_program('icpc',var='CXX')
-	if not cxx:conf.fatal('Intel C++ Compiler (icpc) was not found')
-	cxx=conf.cmd_to_list(cxx)
+	cxx=conf.find_program('icpc',var='CXX')
 	conf.get_cc_version(cxx,icc=True)
-	v['CXX']=cxx
-	v['CXX_NAME']='icc'
+	conf.env.CXX_NAME='icc'
 def configure(conf):
 	conf.find_icpc()
 	conf.find_ar()
@@ -26,5 +18,3 @@
 	conf.cxx_load_tools()
 	conf.cxx_add_flags()
 	conf.link_add_flags()
-
-conf(find_icpc)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ifort.py glmark2-2021.02/waflib/Tools/ifort.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ifort.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/ifort.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,49 +1,301 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import re
-from waflib import Utils
-from waflib.Tools import fc,fc_config,fc_scan
+import os,re
+from waflib import Utils,Logs,Errors
+from waflib.Tools import fc,fc_config,fc_scan,ar,ccroot
 from waflib.Configure import conf
+from waflib.TaskGen import after_method,feature
+@conf
 def find_ifort(conf):
 	fc=conf.find_program('ifort',var='FC')
-	fc=conf.cmd_to_list(fc)
 	conf.get_ifort_version(fc)
 	conf.env.FC_NAME='IFORT'
-def ifort_modifier_cygwin(conf):
-	raise NotImplementedError("Ifort on cygwin not yet implemented")
-def ifort_modifier_win32(conf):
-	fc_config.fortran_modifier_win32(conf)
+@conf
+def ifort_modifier_win32(self):
+	v=self.env
+	v.IFORT_WIN32=True
+	v.FCSTLIB_MARKER=''
+	v.FCSHLIB_MARKER=''
+	v.FCLIB_ST=v.FCSTLIB_ST='%s.lib'
+	v.FCLIBPATH_ST=v.STLIBPATH_ST='/LIBPATH:%s'
+	v.FCINCPATH_ST='/I%s'
+	v.FCDEFINES_ST='/D%s'
+	v.fcprogram_PATTERN=v.fcprogram_test_PATTERN='%s.exe'
+	v.fcshlib_PATTERN='%s.dll'
+	v.fcstlib_PATTERN=v.implib_PATTERN='%s.lib'
+	v.FCLNK_TGT_F='/out:'
+	v.FC_TGT_F=['/c','/o','']
+	v.FCFLAGS_fcshlib=''
+	v.LINKFLAGS_fcshlib='/DLL'
+	v.AR_TGT_F='/out:'
+	v.IMPLIB_ST='/IMPLIB:%s'
+	v.append_value('LINKFLAGS','/subsystem:console')
+	if v.IFORT_MANIFEST:
+		v.append_value('LINKFLAGS',['/MANIFEST'])
+@conf
 def ifort_modifier_darwin(conf):
 	fc_config.fortran_modifier_darwin(conf)
+@conf
 def ifort_modifier_platform(conf):
-	dest_os=conf.env['DEST_OS']or Utils.unversioned_sys_platform()
+	dest_os=conf.env.DEST_OS or Utils.unversioned_sys_platform()
 	ifort_modifier_func=getattr(conf,'ifort_modifier_'+dest_os,None)
 	if ifort_modifier_func:
 		ifort_modifier_func()
+@conf
 def get_ifort_version(conf,fc):
-	version_re=re.compile(r"ifort\s*\(IFORT\)\s*(?P<major>\d*)\.(?P<minor>\d*)",re.I).search
-	cmd=fc+['--version']
-	out,err=fc_config.getoutput(conf,cmd,stdin=False)
-	if out:
-		match=version_re(out)
+	version_re=re.compile(r"\bIntel\b.*\bVersion\s*(?P<major>\d*)\.(?P<minor>\d*)",re.I).search
+	if Utils.is_win32:
+		cmd=fc
 	else:
-		match=version_re(err)
+		cmd=fc+['-logo']
+	out,err=fc_config.getoutput(conf,cmd,stdin=False)
+	match=version_re(out)or version_re(err)
 	if not match:
 		conf.fatal('cannot determine ifort version.')
 	k=match.groupdict()
-	conf.env['FC_VERSION']=(k['major'],k['minor'])
+	conf.env.FC_VERSION=(k['major'],k['minor'])
 def configure(conf):
-	conf.find_ifort()
-	conf.find_program('xiar',var='AR')
-	conf.env.ARFLAGS='rcs'
-	conf.fc_flags()
-	conf.ifort_modifier_platform()
-
-conf(find_ifort)
-conf(ifort_modifier_cygwin)
-conf(ifort_modifier_win32)
-conf(ifort_modifier_darwin)
-conf(ifort_modifier_platform)
-conf(get_ifort_version)
\ No newline at end of file
+	if Utils.is_win32:
+		compiler,version,path,includes,libdirs,arch=conf.detect_ifort(True)
+		v=conf.env
+		v.DEST_CPU=arch
+		v.PATH=path
+		v.INCLUDES=includes
+		v.LIBPATH=libdirs
+		v.MSVC_COMPILER=compiler
+		try:
+			v.MSVC_VERSION=float(version)
+		except Exception:
+			raise
+			v.MSVC_VERSION=float(version[:-3])
+		conf.find_ifort_win32()
+		conf.ifort_modifier_win32()
+	else:
+		conf.find_ifort()
+		conf.find_program('xiar',var='AR')
+		conf.find_ar()
+		conf.fc_flags()
+		conf.fc_add_flags()
+		conf.ifort_modifier_platform()
+all_ifort_platforms=[('intel64','amd64'),('em64t','amd64'),('ia32','x86'),('Itanium','ia64')]
+@conf
+def gather_ifort_versions(conf,versions):
+	version_pattern=re.compile('^...?.?\....?.?')
+	try:
+		all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Intel\\Compilers\\Fortran')
+	except WindowsError:
+		try:
+			all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Intel\\Compilers\\Fortran')
+		except WindowsError:
+			return
+	index=0
+	while 1:
+		try:
+			version=Utils.winreg.EnumKey(all_versions,index)
+		except WindowsError:
+			break
+		index+=1
+		if not version_pattern.match(version):
+			continue
+		targets={}
+		for target,arch in all_ifort_platforms:
+			if target=='intel64':targetDir='EM64T_NATIVE'
+			else:targetDir=target
+			try:
+				Utils.winreg.OpenKey(all_versions,version+'\\'+targetDir)
+				icl_version=Utils.winreg.OpenKey(all_versions,version)
+				path,type=Utils.winreg.QueryValueEx(icl_version,'ProductDir')
+			except WindowsError:
+				pass
+			else:
+				batch_file=os.path.join(path,'bin','iclvars.bat')
+				if os.path.isfile(batch_file):
+					targets[target]=target_compiler(conf,'intel',arch,version,target,batch_file)
+		for target,arch in all_ifort_platforms:
+			try:
+				icl_version=Utils.winreg.OpenKey(all_versions,version+'\\'+target)
+				path,type=Utils.winreg.QueryValueEx(icl_version,'ProductDir')
+			except WindowsError:
+				continue
+			else:
+				batch_file=os.path.join(path,'bin','iclvars.bat')
+				if os.path.isfile(batch_file):
+					targets[target]=target_compiler(conf,'intel',arch,version,target,batch_file)
+		major=version[0:2]
+		versions['intel '+major]=targets
+@conf
+def setup_ifort(conf,versiondict):
+	platforms=Utils.to_list(conf.env.MSVC_TARGETS)or[i for i,j in all_ifort_platforms]
+	desired_versions=conf.env.MSVC_VERSIONS or list(reversed(list(versiondict.keys())))
+	for version in desired_versions:
+		try:
+			targets=versiondict[version]
+		except KeyError:
+			continue
+		for arch in platforms:
+			try:
+				cfg=targets[arch]
+			except KeyError:
+				continue
+			cfg.evaluate()
+			if cfg.is_valid:
+				compiler,revision=version.rsplit(' ',1)
+				return compiler,revision,cfg.bindirs,cfg.incdirs,cfg.libdirs,cfg.cpu
+	conf.fatal('ifort: Impossible to find a valid architecture for building %r - %r'%(desired_versions,list(versiondict.keys())))
+@conf
+def get_ifort_version_win32(conf,compiler,version,target,vcvars):
+	try:
+		conf.msvc_cnt+=1
+	except AttributeError:
+		conf.msvc_cnt=1
+	batfile=conf.bldnode.make_node('waf-print-msvc-%d.bat'%conf.msvc_cnt)
+	batfile.write("""@echo off
+set INCLUDE=
+set LIB=
+call "%s" %s
+echo PATH=%%PATH%%
+echo INCLUDE=%%INCLUDE%%
+echo LIB=%%LIB%%;%%LIBPATH%%
+"""%(vcvars,target))
+	sout=conf.cmd_and_log(['cmd.exe','/E:on','/V:on','/C',batfile.abspath()])
+	batfile.delete()
+	lines=sout.splitlines()
+	if not lines[0]:
+		lines.pop(0)
+	MSVC_PATH=MSVC_INCDIR=MSVC_LIBDIR=None
+	for line in lines:
+		if line.startswith('PATH='):
+			path=line[5:]
+			MSVC_PATH=path.split(';')
+		elif line.startswith('INCLUDE='):
+			MSVC_INCDIR=[i for i in line[8:].split(';')if i]
+		elif line.startswith('LIB='):
+			MSVC_LIBDIR=[i for i in line[4:].split(';')if i]
+	if None in(MSVC_PATH,MSVC_INCDIR,MSVC_LIBDIR):
+		conf.fatal('ifort: Could not find a valid architecture for building (get_ifort_version_win32)')
+	env=dict(os.environ)
+	env.update(PATH=path)
+	compiler_name,linker_name,lib_name=_get_prog_names(conf,compiler)
+	fc=conf.find_program(compiler_name,path_list=MSVC_PATH)
+	if'CL'in env:
+		del(env['CL'])
+	try:
+		conf.cmd_and_log(fc+['/help'],env=env)
+	except UnicodeError:
+		st=Utils.ex_stack()
+		if conf.logger:
+			conf.logger.error(st)
+		conf.fatal('ifort: Unicode error - check the code page?')
+	except Exception as e:
+		Logs.debug('ifort: get_ifort_version: %r %r %r -> failure %s',compiler,version,target,str(e))
+		conf.fatal('ifort: cannot run the compiler in get_ifort_version (run with -v to display errors)')
+	else:
+		Logs.debug('ifort: get_ifort_version: %r %r %r -> OK',compiler,version,target)
+	finally:
+		conf.env[compiler_name]=''
+	return(MSVC_PATH,MSVC_INCDIR,MSVC_LIBDIR)
+class target_compiler(object):
+	def __init__(self,ctx,compiler,cpu,version,bat_target,bat,callback=None):
+		self.conf=ctx
+		self.name=None
+		self.is_valid=False
+		self.is_done=False
+		self.compiler=compiler
+		self.cpu=cpu
+		self.version=version
+		self.bat_target=bat_target
+		self.bat=bat
+		self.callback=callback
+	def evaluate(self):
+		if self.is_done:
+			return
+		self.is_done=True
+		try:
+			vs=self.conf.get_msvc_version(self.compiler,self.version,self.bat_target,self.bat)
+		except Errors.ConfigurationError:
+			self.is_valid=False
+			return
+		if self.callback:
+			vs=self.callback(self,vs)
+		self.is_valid=True
+		(self.bindirs,self.incdirs,self.libdirs)=vs
+	def __str__(self):
+		return str((self.bindirs,self.incdirs,self.libdirs))
+	def __repr__(self):
+		return repr((self.bindirs,self.incdirs,self.libdirs))
+@conf
+def detect_ifort(self):
+	return self.setup_ifort(self.get_ifort_versions(False))
+@conf
+def get_ifort_versions(self,eval_and_save=True):
+	dct={}
+	self.gather_ifort_versions(dct)
+	return dct
+def _get_prog_names(self,compiler):
+	if compiler=='intel':
+		compiler_name='ifort'
+		linker_name='XILINK'
+		lib_name='XILIB'
+	else:
+		compiler_name='CL'
+		linker_name='LINK'
+		lib_name='LIB'
+	return compiler_name,linker_name,lib_name
+@conf
+def find_ifort_win32(conf):
+	v=conf.env
+	path=v.PATH
+	compiler=v.MSVC_COMPILER
+	version=v.MSVC_VERSION
+	compiler_name,linker_name,lib_name=_get_prog_names(conf,compiler)
+	v.IFORT_MANIFEST=(compiler=='intel'and version>=11)
+	fc=conf.find_program(compiler_name,var='FC',path_list=path)
+	env=dict(conf.environ)
+	if path:env.update(PATH=';'.join(path))
+	if not conf.cmd_and_log(fc+['/nologo','/help'],env=env):
+		conf.fatal('not intel fortran compiler could not be identified')
+	v.FC_NAME='IFORT'
+	if not v.LINK_FC:
+		conf.find_program(linker_name,var='LINK_FC',path_list=path,mandatory=True)
+	if not v.AR:
+		conf.find_program(lib_name,path_list=path,var='AR',mandatory=True)
+		v.ARFLAGS=['/nologo']
+	if v.IFORT_MANIFEST:
+		conf.find_program('MT',path_list=path,var='MT')
+		v.MTFLAGS=['/nologo']
+	try:
+		conf.load('winres')
+	except Errors.WafError:
+		Logs.warn('Resource compiler not found. Compiling resource file is disabled')
+@after_method('apply_link')
+@feature('fc')
+def apply_flags_ifort(self):
+	if not self.env.IFORT_WIN32 or not getattr(self,'link_task',None):
+		return
+	is_static=isinstance(self.link_task,ccroot.stlink_task)
+	subsystem=getattr(self,'subsystem','')
+	if subsystem:
+		subsystem='/subsystem:%s'%subsystem
+		flags=is_static and'ARFLAGS'or'LINKFLAGS'
+		self.env.append_value(flags,subsystem)
+	if not is_static:
+		for f in self.env.LINKFLAGS:
+			d=f.lower()
+			if d[1:]=='debug':
+				pdbnode=self.link_task.outputs[0].change_ext('.pdb')
+				self.link_task.outputs.append(pdbnode)
+				if getattr(self,'install_task',None):
+					self.pdb_install_task=self.add_install_files(install_to=self.install_task.install_to,install_from=pdbnode)
+				break
+@feature('fcprogram','fcshlib','fcprogram_test')
+@after_method('apply_link')
+def apply_manifest_ifort(self):
+	if self.env.IFORT_WIN32 and getattr(self,'link_task',None):
+		self.link_task.env.FC=self.env.LINK_FC
+	if self.env.IFORT_WIN32 and self.env.IFORT_MANIFEST and getattr(self,'link_task',None):
+		out_node=self.link_task.outputs[0]
+		man_node=out_node.parent.find_or_declare(out_node.name+'.manifest')
+		self.link_task.outputs.append(man_node)
+		self.env.DO_MANIFEST=True
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/__init__.py glmark2-2021.02/waflib/Tools/__init__.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/__init__.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/__init__.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,4 +1,4 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/intltool.py glmark2-2021.02/waflib/Tools/intltool.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/intltool.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/intltool.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,39 +1,58 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os,re
-from waflib import Configure,TaskGen,Task,Utils,Runner,Options,Build,Logs
+from waflib import Context,Task,Utils,Logs
 import waflib.Tools.ccroot
-from waflib.TaskGen import feature,before_method
+from waflib.TaskGen import feature,before_method,taskgen_method
 from waflib.Logs import error
+from waflib.Configure import conf
+_style_flags={'ba':'-b','desktop':'-d','keys':'-k','quoted':'--quoted-style','quotedxml':'--quotedxml-style','rfc822deb':'-r','schemas':'-s','xml':'-x',}
+@taskgen_method
+def ensure_localedir(self):
+	if not self.env.LOCALEDIR:
+		if self.env.DATAROOTDIR:
+			self.env.LOCALEDIR=os.path.join(self.env.DATAROOTDIR,'locale')
+		else:
+			self.env.LOCALEDIR=os.path.join(self.env.PREFIX,'share','locale')
+@before_method('process_source')
+@feature('intltool_in')
 def apply_intltool_in_f(self):
 	try:self.meths.remove('process_source')
 	except ValueError:pass
-	if not self.env.LOCALEDIR:
-		self.env.LOCALEDIR=self.env.PREFIX+'/share/locale'
+	self.ensure_localedir()
+	podir=getattr(self,'podir','.')
+	podirnode=self.path.find_dir(podir)
+	if not podirnode:
+		error("could not find the podir %r"%podir)
+		return
+	cache=getattr(self,'intlcache','.intlcache')
+	self.env.INTLCACHE=[os.path.join(str(self.path.get_bld()),podir,cache)]
+	self.env.INTLPODIR=podirnode.bldpath()
+	self.env.append_value('INTLFLAGS',getattr(self,'flags',self.env.INTLFLAGS_DEFAULT))
+	if'-c'in self.env.INTLFLAGS:
+		self.bld.fatal('Redundant -c flag in intltool task %r'%self)
+	style=getattr(self,'style',None)
+	if style:
+		try:
+			style_flag=_style_flags[style]
+		except KeyError:
+			self.bld.fatal('intltool_in style "%s" is not valid'%style)
+		self.env.append_unique('INTLFLAGS',[style_flag])
 	for i in self.to_list(self.source):
 		node=self.path.find_resource(i)
-		podir=getattr(self,'podir','po')
-		podirnode=self.path.find_dir(podir)
-		if not podirnode:
-			error("could not find the podir %r"%podir)
-			continue
-		cache=getattr(self,'intlcache','.intlcache')
-		self.env['INTLCACHE']=os.path.join(self.path.bldpath(),podir,cache)
-		self.env['INTLPODIR']=podirnode.bldpath()
-		self.env['INTLFLAGS']=getattr(self,'flags',['-q','-u','-c'])
 		task=self.create_task('intltool',node,node.change_ext(''))
-		inst=getattr(self,'install_path','${LOCALEDIR}')
+		inst=getattr(self,'install_path',None)
 		if inst:
-			self.bld.install_files(inst,task.outputs)
+			self.add_install_files(install_to=inst,install_from=task.outputs)
+@feature('intltool_po')
 def apply_intltool_po(self):
 	try:self.meths.remove('process_source')
 	except ValueError:pass
-	if not self.env.LOCALEDIR:
-		self.env.LOCALEDIR=self.env.PREFIX+'/share/locale'
-	appname=getattr(self,'appname','set_your_app_name')
-	podir=getattr(self,'podir','')
+	self.ensure_localedir()
+	appname=getattr(self,'appname',getattr(Context.g_module,Context.APPNAME,'set_your_app_name'))
+	podir=getattr(self,'podir','.')
 	inst=getattr(self,'install_path','${LOCALEDIR}')
 	linguas=self.path.find_node(os.path.join(podir,'LINGUAS'))
 	if linguas:
@@ -52,27 +71,27 @@
 					filename=task.outputs[0].name
 					(langname,ext)=os.path.splitext(filename)
 					inst_file=inst+os.sep+langname+os.sep+'LC_MESSAGES'+os.sep+appname+'.mo'
-					self.bld.install_as(inst_file,task.outputs[0],chmod=getattr(self,'chmod',Utils.O644),env=task.env)
+					self.add_install_as(install_to=inst_file,install_from=task.outputs[0],chmod=getattr(self,'chmod',Utils.O644))
 	else:
 		Logs.pprint('RED',"Error no LINGUAS file found in po directory")
 class po(Task.Task):
 	run_str='${MSGFMT} -o ${TGT} ${SRC}'
 	color='BLUE'
 class intltool(Task.Task):
-	run_str='${INTLTOOL} ${INTLFLAGS} ${INTLCACHE} ${INTLPODIR} ${SRC} ${TGT}'
+	run_str='${INTLTOOL} ${INTLFLAGS} ${INTLCACHE_ST:INTLCACHE} ${INTLPODIR} ${SRC} ${TGT}'
 	color='BLUE'
-def configure(conf):
+@conf
+def find_msgfmt(conf):
 	conf.find_program('msgfmt',var='MSGFMT')
-	conf.find_perl_program('intltool-merge',var='INTLTOOL')
-	prefix=conf.env.PREFIX
-	datadir=conf.env.DATADIR
-	if not datadir:
-		datadir=os.path.join(prefix,'share')
-	conf.define('LOCALEDIR',os.path.join(datadir,'locale').replace('\\','\\\\'))
-	conf.define('DATADIR',datadir.replace('\\','\\\\'))
+@conf
+def find_intltool_merge(conf):
+	if not conf.env.PERL:
+		conf.find_program('perl',var='PERL')
+	conf.env.INTLCACHE_ST='--cache=%s'
+	conf.env.INTLFLAGS_DEFAULT=['-q','-u']
+	conf.find_program('intltool-merge',interpreter='PERL',var='INTLTOOL')
+def configure(conf):
+	conf.find_msgfmt()
+	conf.find_intltool_merge()
 	if conf.env.CC or conf.env.CXX:
 		conf.check(header_name='locale.h')
-
-before_method('process_source')(apply_intltool_in_f)
-feature('intltool_in')(apply_intltool_in_f)
-feature('intltool_po')(apply_intltool_po)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/irixcc.py glmark2-2021.02/waflib/Tools/irixcc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/irixcc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/irixcc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,41 +1,45 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os
-from waflib import Utils
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_irixcc(conf):
 	v=conf.env
 	cc=None
-	if v['CC']:cc=v['CC']
-	elif'CC'in conf.environ:cc=conf.environ['CC']
-	if not cc:cc=conf.find_program('cc',var='CC')
-	if not cc:conf.fatal('irixcc was not found')
-	cc=conf.cmd_to_list(cc)
+	if v.CC:
+		cc=v.CC
+	elif'CC'in conf.environ:
+		cc=conf.environ['CC']
+	if not cc:
+		cc=conf.find_program('cc',var='CC')
+	if not cc:
+		conf.fatal('irixcc was not found')
 	try:
 		conf.cmd_and_log(cc+['-version'])
-	except:
+	except Exception:
 		conf.fatal('%r -version could not be executed'%cc)
-	v['CC']=cc
-	v['CC_NAME']='irix'
+	v.CC=cc
+	v.CC_NAME='irix'
+@conf
 def irixcc_common_flags(conf):
 	v=conf.env
-	v['CC_SRC_F']=''
-	v['CC_TGT_F']=['-c','-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	if not v['LINK_CC']:v['LINK_CC']=v['CC']
-	v['CCLNK_SRC_F']=''
-	v['CCLNK_TGT_F']=['-o']
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['cprogram_PATTERN']='%s'
-	v['cshlib_PATTERN']='lib%s.so'
-	v['cstlib_PATTERN']='lib%s.a'
+	v.CC_SRC_F=''
+	v.CC_TGT_F=['-c','-o']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	if not v.LINK_CC:
+		v.LINK_CC=v.CC
+	v.CCLNK_SRC_F=''
+	v.CCLNK_TGT_F=['-o']
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.cprogram_PATTERN='%s'
+	v.cshlib_PATTERN='lib%s.so'
+	v.cstlib_PATTERN='lib%s.a'
 def configure(conf):
 	conf.find_irixcc()
 	conf.find_cpp()
@@ -44,6 +48,3 @@
 	conf.cc_load_tools()
 	conf.cc_add_flags()
 	conf.link_add_flags()
-
-conf(find_irixcc)
-conf(irixcc_common_flags)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/javaw.py glmark2-2021.02/waflib/Tools/javaw.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/javaw.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/javaw.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,12 +1,10 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-import os,re,tempfile,shutil
+import os,shutil
+from waflib import Task,Utils,Errors,Node
 from waflib.Configure import conf
-from waflib import TaskGen,Task,Utils,Options,Build,Errors,Node,Logs
 from waflib.TaskGen import feature,before_method,after_method
 from waflib.Tools import ccroot
 ccroot.USELIB_VARS['javac']=set(['CLASSPATH','JAVACFLAGS'])
@@ -31,9 +29,10 @@
 	}
 }
 '''
+@feature('javac')
+@before_method('process_source')
 def apply_java(self):
 	Utils.def_attrs(self,jarname='',classpath='',sourcepath='.',srcdir='.',jar_mf_attributes={},jar_mf_classpath=[])
-	nodes_lst=[]
 	outdir=getattr(self,'outdir',None)
 	if outdir:
 		if not isinstance(outdir,Node.Node):
@@ -42,7 +41,7 @@
 		outdir=self.path.get_bld()
 	outdir.mkdir()
 	self.outdir=outdir
-	self.env['OUTDIR']=outdir.abspath()
+	self.env.OUTDIR=outdir.abspath()
 	self.javac_task=tsk=self.create_task('javac')
 	tmp=[]
 	srcdir=getattr(self,'srcdir','')
@@ -66,6 +65,8 @@
 		names=[x.srcpath()for x in tsk.srcdir]
 	if names:
 		tsk.env.append_value('JAVACFLAGS',['-sourcepath',names])
+@feature('javac')
+@after_method('apply_java')
 def use_javac_files(self):
 	lst=[]
 	self.uselib=self.to_list(getattr(self,'uselib',[]))
@@ -74,18 +75,26 @@
 	for x in names:
 		try:
 			y=get(x)
-		except:
+		except Errors.WafError:
 			self.uselib.append(x)
 		else:
 			y.post()
-			lst.append(y.jar_task.outputs[0].abspath())
-			self.javac_task.set_run_after(y.jar_task)
-	if lst:
-		self.env.append_value('CLASSPATH',lst)
+			if hasattr(y,'jar_task'):
+				lst.append(y.jar_task.outputs[0].abspath())
+				self.javac_task.set_run_after(y.jar_task)
+			else:
+				for tsk in y.tasks:
+					self.javac_task.set_run_after(tsk)
+	self.env.append_value('CLASSPATH',lst)
+@feature('javac')
+@after_method('apply_java','propagate_uselib_vars','use_javac_files')
 def set_classpath(self):
 	self.env.append_value('CLASSPATH',getattr(self,'classpath',[]))
 	for x in self.tasks:
 		x.env.CLASSPATH=os.pathsep.join(self.env.CLASSPATH)+os.pathsep
+@feature('jar')
+@after_method('apply_java','use_javac_files')
+@before_method('process_source')
 def jar_files(self):
 	destfile=getattr(self,'destfile','test.jar')
 	jaropts=getattr(self,'jaropts',[])
@@ -101,7 +110,10 @@
 	self.jar_task=tsk=self.create_task('jar_create')
 	if manifest:
 		jarcreate=getattr(self,'jarcreate','cfm')
-		node=self.path.find_node(manifest)
+		if not isinstance(manifest,Node.Node):
+			node=self.path.find_or_declare(manifest)
+		else:
+			node=manifest
 		tsk.dep_nodes.append(node)
 		jaropts.insert(0,node.abspath())
 	else:
@@ -115,24 +127,35 @@
 	jaropts.append('-C')
 	jaropts.append(basedir.bldpath())
 	jaropts.append('.')
-	tsk.env['JAROPTS']=jaropts
-	tsk.env['JARCREATE']=jarcreate
+	tsk.env.JAROPTS=jaropts
+	tsk.env.JARCREATE=jarcreate
 	if getattr(self,'javac_task',None):
 		tsk.set_run_after(self.javac_task)
+@feature('jar')
+@after_method('jar_files')
 def use_jar_files(self):
-	lst=[]
 	self.uselib=self.to_list(getattr(self,'uselib',[]))
 	names=self.to_list(getattr(self,'use',[]))
 	get=self.bld.get_tgen_by_name
 	for x in names:
 		try:
 			y=get(x)
-		except:
+		except Errors.WafError:
 			self.uselib.append(x)
 		else:
 			y.post()
 			self.jar_task.run_after.update(y.tasks)
-class jar_create(Task.Task):
+class JTask(Task.Task):
+	def split_argfile(self,cmd):
+		inline=[cmd[0]]
+		infile=[]
+		for x in cmd[1:]:
+			if x.startswith('-J'):
+				inline.append(x)
+			else:
+				infile.append(self.quote_flag(x))
+		return(inline,infile)
+class jar_create(JTask):
 	color='GREEN'
 	run_str='${JAR} ${JARCREATE} ${TGT} ${JAROPTS}'
 	def runnable_status(self):
@@ -143,13 +166,18 @@
 			global JAR_RE
 			try:
 				self.inputs=[x for x in self.basedir.ant_glob(JAR_RE,remove=False)if id(x)!=id(self.outputs[0])]
-			except:
+			except Exception:
 				raise Errors.WafError('Could not find the basedir %r for %r'%(self.basedir,self))
 		return super(jar_create,self).runnable_status()
-class javac(Task.Task):
+class javac(JTask):
 	color='BLUE'
-	nocache=True
+	run_str='${JAVAC} -classpath ${CLASSPATH} -d ${OUTDIR} ${JAVACFLAGS} ${SRC}'
 	vars=['CLASSPATH','JAVACFLAGS','JAVAC','OUTDIR']
+	def uid(self):
+		lst=[self.__class__.__name__,self.generator.outdir.abspath()]
+		for x in self.srcdir:
+			lst.append(x.abspath())
+		return Utils.h_list(lst)
 	def runnable_status(self):
 		for t in self.run_after:
 			if not t.hasrun:
@@ -160,87 +188,92 @@
 			for x in self.srcdir:
 				self.inputs.extend(x.ant_glob(SOURCE_RE,remove=False))
 		return super(javac,self).runnable_status()
+	def post_run(self):
+		for node in self.generator.outdir.ant_glob('**/*.class'):
+			self.generator.bld.node_sigs[node]=self.uid()
+		self.generator.bld.task_sigs[self.uid()]=self.cache_sig
+@feature('javadoc')
+@after_method('process_rule')
+def create_javadoc(self):
+	tsk=self.create_task('javadoc')
+	tsk.classpath=getattr(self,'classpath',[])
+	self.javadoc_package=Utils.to_list(self.javadoc_package)
+	if not isinstance(self.javadoc_output,Node.Node):
+		self.javadoc_output=self.bld.path.find_or_declare(self.javadoc_output)
+class javadoc(Task.Task):
+	color='BLUE'
+	def __str__(self):
+		return'%s: %s -> %s\n'%(self.__class__.__name__,self.generator.srcdir,self.generator.javadoc_output)
 	def run(self):
 		env=self.env
-		gen=self.generator
-		bld=gen.bld
-		wd=bld.bldnode.abspath()
-		def to_list(xx):
-			if isinstance(xx,str):return[xx]
-			return xx
-		cmd=[]
-		cmd.extend(to_list(env['JAVAC']))
-		cmd.extend(['-classpath'])
-		cmd.extend(to_list(env['CLASSPATH']))
-		cmd.extend(['-d'])
-		cmd.extend(to_list(env['OUTDIR']))
-		cmd.extend(to_list(env['JAVACFLAGS']))
-		files=[a.path_from(bld.bldnode)for a in self.inputs]
-		tmp=None
-		try:
-			if len(str(files))+len(str(cmd))>8192:
-				(fd,tmp)=tempfile.mkstemp(dir=bld.bldnode.abspath())
-				try:
-					os.write(fd,'\n'.join(files))
-				finally:
-					if tmp:
-						os.close(fd)
-				if Logs.verbose:
-					Logs.debug('runner: %r'%(cmd+files))
-				cmd.append('@'+tmp)
-			else:
-				cmd+=files
-			ret=self.exec_command(cmd,cwd=wd,env=env.env or None)
-		finally:
-			if tmp:
-				os.unlink(tmp)
-		return ret
+		bld=self.generator.bld
+		wd=bld.bldnode
+		srcpath=self.generator.path.abspath()+os.sep+self.generator.srcdir
+		srcpath+=os.pathsep
+		srcpath+=self.generator.path.get_bld().abspath()+os.sep+self.generator.srcdir
+		classpath=env.CLASSPATH
+		classpath+=os.pathsep
+		classpath+=os.pathsep.join(self.classpath)
+		classpath="".join(classpath)
+		self.last_cmd=lst=[]
+		lst.extend(Utils.to_list(env.JAVADOC))
+		lst.extend(['-d',self.generator.javadoc_output.abspath()])
+		lst.extend(['-sourcepath',srcpath])
+		lst.extend(['-classpath',classpath])
+		lst.extend(['-subpackages'])
+		lst.extend(self.generator.javadoc_package)
+		lst=[x for x in lst if x]
+		self.generator.bld.cmd_and_log(lst,cwd=wd,env=env.env or None,quiet=0)
 	def post_run(self):
-		for n in self.generator.outdir.ant_glob('**/*.class'):
-			n.sig=Utils.h_file(n.abspath())
+		nodes=self.generator.javadoc_output.ant_glob('**')
+		for node in nodes:
+			self.generator.bld.node_sigs[node]=self.uid()
 		self.generator.bld.task_sigs[self.uid()]=self.cache_sig
 def configure(self):
 	java_path=self.environ['PATH'].split(os.pathsep)
 	v=self.env
 	if'JAVA_HOME'in self.environ:
 		java_path=[os.path.join(self.environ['JAVA_HOME'],'bin')]+java_path
-		self.env['JAVA_HOME']=[self.environ['JAVA_HOME']]
-	for x in'javac java jar'.split():
+		self.env.JAVA_HOME=[self.environ['JAVA_HOME']]
+	for x in'javac java jar javadoc'.split():
 		self.find_program(x,var=x.upper(),path_list=java_path)
-		self.env[x.upper()]=self.cmd_to_list(self.env[x.upper()])
 	if'CLASSPATH'in self.environ:
-		v['CLASSPATH']=self.environ['CLASSPATH']
-	if not v['JAR']:self.fatal('jar is required for making java packages')
-	if not v['JAVAC']:self.fatal('javac is required for compiling java classes')
-	v['JARCREATE']='cf'
-	v['JAVACFLAGS']=[]
+		v.CLASSPATH=self.environ['CLASSPATH']
+	if not v.JAR:
+		self.fatal('jar is required for making java packages')
+	if not v.JAVAC:
+		self.fatal('javac is required for compiling java classes')
+	v.JARCREATE='cf'
+	v.JAVACFLAGS=[]
+@conf
 def check_java_class(self,classname,with_classpath=None):
 	javatestdir='.waf-javatest'
 	classpath=javatestdir
-	if self.env['CLASSPATH']:
-		classpath+=os.pathsep+self.env['CLASSPATH']
+	if self.env.CLASSPATH:
+		classpath+=os.pathsep+self.env.CLASSPATH
 	if isinstance(with_classpath,str):
 		classpath+=os.pathsep+with_classpath
 	shutil.rmtree(javatestdir,True)
 	os.mkdir(javatestdir)
-	java_file=open(os.path.join(javatestdir,'Test.java'),'w')
-	java_file.write(class_check_source)
-	java_file.close()
-	self.exec_command(self.env['JAVAC']+[os.path.join(javatestdir,'Test.java')],shell=False)
-	cmd=self.env['JAVA']+['-cp',classpath,'Test',classname]
+	Utils.writef(os.path.join(javatestdir,'Test.java'),class_check_source)
+	self.exec_command(self.env.JAVAC+[os.path.join(javatestdir,'Test.java')],shell=False)
+	cmd=self.env.JAVA+['-cp',classpath,'Test',classname]
 	self.to_log("%s\n"%str(cmd))
 	found=self.exec_command(cmd,shell=False)
 	self.msg('Checking for java class %s'%classname,not found)
 	shutil.rmtree(javatestdir,True)
 	return found
+@conf
 def check_jni_headers(conf):
 	if not conf.env.CC_NAME and not conf.env.CXX_NAME:
 		conf.fatal('load a compiler first (gcc, g++, ..)')
 	if not conf.env.JAVA_HOME:
 		conf.fatal('set JAVA_HOME in the system environment')
-	javaHome=conf.env['JAVA_HOME'][0]
+	javaHome=conf.env.JAVA_HOME[0]
 	dir=conf.root.find_dir(conf.env.JAVA_HOME[0]+'/include')
 	if dir is None:
+		dir=conf.root.find_dir(conf.env.JAVA_HOME[0]+'/../Headers')
+	if dir is None:
 		conf.fatal('JAVA_HOME does not seem to be set properly')
 	f=dir.ant_glob('**/(jni|jni_md).h')
 	incDirs=[x.parent.abspath()for x in f]
@@ -250,26 +283,14 @@
 	f=dir.ant_glob('**/*jvm.(lib)')
 	if f:
 		libDirs=[[x,y.parent.abspath()]for x in libDirs for y in f]
+	if conf.env.DEST_OS=='freebsd':
+		conf.env.append_unique('LINKFLAGS_JAVA','-pthread')
 	for d in libDirs:
 		try:
 			conf.check(header_name='jni.h',define_name='HAVE_JNI_H',lib='jvm',libpath=d,includes=incDirs,uselib_store='JAVA',uselib='JAVA')
-		except:
+		except Exception:
 			pass
 		else:
 			break
 	else:
 		conf.fatal('could not find lib jvm in %r (see config.log)'%libDirs)
-
-feature('javac')(apply_java)
-before_method('process_source')(apply_java)
-feature('javac')(use_javac_files)
-after_method('apply_java')(use_javac_files)
-feature('javac')(set_classpath)
-after_method('apply_java','propagate_uselib_vars','use_javac_files')(set_classpath)
-feature('jar')(jar_files)
-after_method('apply_java','use_javac_files')(jar_files)
-before_method('process_source')(jar_files)
-feature('jar')(use_jar_files)
-after_method('jar_files')(use_jar_files)
-conf(check_java_class)
-conf(check_jni_headers)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/kde4.py glmark2-2021.02/waflib/Tools/kde4.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/kde4.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/kde4.py	1970-01-01 08:00:00.000000000 +0800
@@ -1,49 +0,0 @@
-#! /usr/bin/env python
-# encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
-
-import os,sys,re
-from waflib import Options,TaskGen,Task,Utils
-from waflib.TaskGen import feature,after_method
-def apply_msgfmt(self):
-	for lang in self.to_list(self.langs):
-		node=self.path.find_resource(lang+'.po')
-		task=self.create_task('msgfmt',node,node.change_ext('.mo'))
-		langname=lang.split('/')
-		langname=langname[-1]
-		inst=getattr(self,'install_path','${KDE4_LOCALE_INSTALL_DIR}')
-		self.bld.install_as(inst+os.sep+langname+os.sep+'LC_MESSAGES'+os.sep+getattr(self,'appname','set_your_appname')+'.mo',task.outputs[0],chmod=getattr(self,'chmod',Utils.O644))
-class msgfmt(Task.Task):
-	color='BLUE'
-	run_str='${MSGFMT} ${SRC} -o ${TGT}'
-def configure(self):
-	kdeconfig=self.find_program('kde4-config')
-	prefix=self.cmd_and_log('%s --prefix'%kdeconfig).strip()
-	fname='%s/share/apps/cmake/modules/KDELibsDependencies.cmake'%prefix
-	try:os.stat(fname)
-	except OSError:
-		fname='%s/share/kde4/apps/cmake/modules/KDELibsDependencies.cmake'%prefix
-		try:os.stat(fname)
-		except OSError:self.fatal('could not open %s'%fname)
-	try:
-		txt=Utils.readf(fname)
-	except(OSError,IOError):
-		self.fatal('could not read %s'%fname)
-	txt=txt.replace('\\\n','\n')
-	fu=re.compile('#(.*)\n')
-	txt=fu.sub('',txt)
-	setregexp=re.compile('([sS][eE][tT]\s*\()\s*([^\s]+)\s+\"([^"]+)\"\)')
-	found=setregexp.findall(txt)
-	for(_,key,val)in found:
-		self.env[key]=val
-	self.env['LIB_KDECORE']=['kdecore']
-	self.env['LIB_KDEUI']=['kdeui']
-	self.env['LIB_KIO']=['kio']
-	self.env['LIB_KHTML']=['khtml']
-	self.env['LIB_KPARTS']=['kparts']
-	self.env['LIBPATH_KDECORE']=[self.env['KDE4_LIB_INSTALL_DIR']]
-	self.env['INCLUDES_KDECORE']=[self.env['KDE4_INCLUDE_INSTALL_DIR']]
-	self.env.append_value('INCLUDES_KDECORE',[self.env['KDE4_INCLUDE_INSTALL_DIR']+os.sep+'KDE'])
-	self.find_program('msgfmt',var='MSGFMT')
-
-feature('msgfmt')(apply_msgfmt)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ldc2.py glmark2-2021.02/waflib/Tools/ldc2.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ldc2.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/ldc2.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,36 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+from waflib.Tools import ar,d
+from waflib.Configure import conf
+@conf
+def find_ldc2(conf):
+	conf.find_program(['ldc2'],var='D')
+	out=conf.cmd_and_log(conf.env.D+['-version'])
+	if out.find("based on DMD v2.")==-1:
+		conf.fatal("detected compiler is not ldc2")
+@conf
+def common_flags_ldc2(conf):
+	v=conf.env
+	v.D_SRC_F=['-c']
+	v.D_TGT_F='-of%s'
+	v.D_LINKER=v.D
+	v.DLNK_SRC_F=''
+	v.DLNK_TGT_F='-of%s'
+	v.DINC_ST='-I%s'
+	v.DSHLIB_MARKER=v.DSTLIB_MARKER=''
+	v.DSTLIB_ST=v.DSHLIB_ST='-L-l%s'
+	v.DSTLIBPATH_ST=v.DLIBPATH_ST='-L-L%s'
+	v.LINKFLAGS_dshlib=['-L-shared']
+	v.DHEADER_ext='.di'
+	v.DFLAGS_d_with_header=['-H','-Hf']
+	v.D_HDR_F='%s'
+	v.LINKFLAGS=[]
+	v.DFLAGS_dshlib=['-relocation-model=pic']
+def configure(conf):
+	conf.find_ldc2()
+	conf.load('ar')
+	conf.load('d')
+	conf.common_flags_ldc2()
+	conf.d_platform_flags()
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/lua.py glmark2-2021.02/waflib/Tools/lua.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/lua.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/lua.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,19 +1,18 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib.TaskGen import extension
-from waflib import Task,Utils
+from waflib import Task
+@extension('.lua')
 def add_lua(self,node):
 	tsk=self.create_task('luac',node,node.change_ext('.luac'))
 	inst_to=getattr(self,'install_path',self.env.LUADIR and'${LUADIR}'or None)
 	if inst_to:
-		self.bld.install_files(inst_to,tsk.outputs)
+		self.add_install_files(install_to=inst_to,install_from=tsk.outputs)
 	return tsk
 class luac(Task.Task):
 	run_str='${LUAC} -s -o ${TGT} ${SRC}'
 	color='PINK'
 def configure(conf):
 	conf.find_program('luac',var='LUAC')
-
-extension('.lua')(add_lua)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/md5_tstamp.py glmark2-2021.02/waflib/Tools/md5_tstamp.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/md5_tstamp.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/md5_tstamp.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,25 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+import os,stat
+from waflib import Utils,Build,Node
+STRONGEST=True
+Build.SAVED_ATTRS.append('hashes_md5_tstamp')
+def h_file(self):
+	filename=self.abspath()
+	st=os.stat(filename)
+	cache=self.ctx.hashes_md5_tstamp
+	if filename in cache and cache[filename][0]==st.st_mtime:
+		return cache[filename][1]
+	global STRONGEST
+	if STRONGEST:
+		ret=Utils.h_file(filename)
+	else:
+		if stat.S_ISDIR(st[stat.ST_MODE]):
+			raise IOError('Not a file')
+		ret=Utils.md5(str((st.st_mtime,st.st_size)).encode()).digest()
+	cache[filename]=(st.st_mtime,ret)
+	return ret
+h_file.__doc__=Node.Node.h_file.__doc__
+Node.Node.h_file=h_file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/msvc.py glmark2-2021.02/waflib/Tools/msvc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/msvc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/msvc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,20 +1,12 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,re,tempfile
-try:
-	import _winreg
-except:
-	try:
-		import winreg as _winreg
-	except:
-		_winreg=None
-from waflib import Utils,TaskGen,Runner,Configure,Task,Options
-from waflib.Logs import debug,info,warn,error
-from waflib.TaskGen import after_method,before_method,feature
+import os,sys,re
+from waflib import Utils,Logs,Options,Errors
+from waflib.TaskGen import after_method,feature
 from waflib.Configure import conf
-from waflib.Tools import ccroot,c,cxx,ar,winres
+from waflib.Tools import ccroot,c,cxx,ar
 g_msvc_systemlibs='''
 aclui activeds ad1 adptif adsiid advapi32 asycfilt authz bhsupp bits bufferoverflowu cabinet
 cap certadm certidl ciuuid clusapi comctl32 comdlg32 comsupp comsuppd comsuppw comsuppwd comsvcs
@@ -35,52 +27,69 @@
 version vfw32 wbemuuid  webpost wiaguid wininet winmm winscard winspool winstrm
 wintrust wldap32 wmiutils wow32 ws2_32 wsnmp32 wsock32 wst wtsapi32 xaswitch xolehlp
 '''.split()
-all_msvc_platforms=[('x64','amd64'),('x86','x86'),('ia64','ia64'),('x86_amd64','amd64'),('x86_ia64','ia64')]
+all_msvc_platforms=[('x64','amd64'),('x86','x86'),('ia64','ia64'),('x86_amd64','amd64'),('x86_ia64','ia64'),('x86_arm','arm'),('amd64_x86','x86'),('amd64_arm','arm')]
 all_wince_platforms=[('armv4','arm'),('armv4i','arm'),('mipsii','mips'),('mipsii_fp','mips'),('mipsiv','mips'),('mipsiv_fp','mips'),('sh4','sh'),('x86','cex86')]
 all_icl_platforms=[('intel64','amd64'),('em64t','amd64'),('ia32','x86'),('Itanium','ia64')]
 def options(opt):
 	opt.add_option('--msvc_version',type='string',help='msvc version, eg: "msvc 10.0,msvc 9.0"',default='')
 	opt.add_option('--msvc_targets',type='string',help='msvc targets, eg: "x64,arm"',default='')
-def setup_msvc(conf,versions):
+	opt.add_option('--no-msvc-lazy',action='store_false',help='lazily check msvc target environments',default=True,dest='msvc_lazy')
+@conf
+def setup_msvc(conf,versiondict):
 	platforms=getattr(Options.options,'msvc_targets','').split(',')
 	if platforms==['']:
-		platforms=Utils.to_list(conf.env['MSVC_TARGETS'])or[i for i,j in all_msvc_platforms+all_icl_platforms+all_wince_platforms]
+		platforms=Utils.to_list(conf.env.MSVC_TARGETS)or[i for i,j in all_msvc_platforms+all_icl_platforms+all_wince_platforms]
 	desired_versions=getattr(Options.options,'msvc_version','').split(',')
 	if desired_versions==['']:
-		desired_versions=conf.env['MSVC_VERSIONS']or[v for v,_ in versions][::-1]
-	versiondict=dict(versions)
+		desired_versions=conf.env.MSVC_VERSIONS or list(reversed(list(versiondict.keys())))
+	lazy_detect=getattr(Options.options,'msvc_lazy',True)
+	if conf.env.MSVC_LAZY_AUTODETECT is False:
+		lazy_detect=False
+	if not lazy_detect:
+		for val in versiondict.values():
+			for arch in list(val.keys()):
+				cfg=val[arch]
+				cfg.evaluate()
+				if not cfg.is_valid:
+					del val[arch]
+		conf.env.MSVC_INSTALLED_VERSIONS=versiondict
 	for version in desired_versions:
 		try:
-			targets=dict(versiondict[version])
-			for target in platforms:
-				try:
-					arch,(p1,p2,p3)=targets[target]
-					compiler,revision=version.rsplit(' ',1)
-					return compiler,revision,p1,p2,p3
-				except KeyError:continue
-		except KeyError:continue
-	conf.fatal('msvc: Impossible to find a valid architecture for building (in setup_msvc)')
+			targets=versiondict[version]
+		except KeyError:
+			continue
+		for arch in platforms:
+			try:
+				cfg=targets[arch]
+			except KeyError:
+				continue
+			cfg.evaluate()
+			if cfg.is_valid:
+				compiler,revision=version.rsplit(' ',1)
+				return compiler,revision,cfg.bindirs,cfg.incdirs,cfg.libdirs,cfg.cpu
+	conf.fatal('msvc: Impossible to find a valid architecture for building %r - %r'%(desired_versions,list(versiondict.keys())))
+@conf
 def get_msvc_version(conf,compiler,version,target,vcvars):
-	debug('msvc: get_msvc_version: %r %r %r',compiler,version,target)
-	batfile=conf.bldnode.make_node('waf-print-msvc.bat')
+	Logs.debug('msvc: get_msvc_version: %r %r %r',compiler,version,target)
+	try:
+		conf.msvc_cnt+=1
+	except AttributeError:
+		conf.msvc_cnt=1
+	batfile=conf.bldnode.make_node('waf-print-msvc-%d.bat'%conf.msvc_cnt)
 	batfile.write("""@echo off
 set INCLUDE=
 set LIB=
 call "%s" %s
 echo PATH=%%PATH%%
 echo INCLUDE=%%INCLUDE%%
-echo LIB=%%LIB%%
+echo LIB=%%LIB%%;%%LIBPATH%%
 """%(vcvars,target))
-	sout=conf.cmd_and_log(['cmd','/E:on','/V:on','/C',batfile.abspath()])
+	sout=conf.cmd_and_log(['cmd.exe','/E:on','/V:on','/C',batfile.abspath()])
 	lines=sout.splitlines()
-	if not lines[0]:lines=lines[1:]
-	for x in('Setting environment','Setting SDK environment','Intel(R) C++ Compiler','Intel Parallel Studio'):
-		if lines[0].find(x)!=-1:
-			break
-	else:
-		debug('msvc: get_msvc_version: %r %r %r -> not found',compiler,version,target)
-		conf.fatal('msvc: Impossible to find a valid architecture for building (in get_msvc_version)')
-	for line in lines[1:]:
+	if not lines[0]:
+		lines.pop(0)
+	MSVC_PATH=MSVC_INCDIR=MSVC_LIBDIR=None
+	for line in lines:
 		if line.startswith('PATH='):
 			path=line[5:]
 			MSVC_PATH=path.split(';')
@@ -88,177 +97,206 @@
 			MSVC_INCDIR=[i for i in line[8:].split(';')if i]
 		elif line.startswith('LIB='):
 			MSVC_LIBDIR=[i for i in line[4:].split(';')if i]
-	env={}
-	env.update(os.environ)
+	if None in(MSVC_PATH,MSVC_INCDIR,MSVC_LIBDIR):
+		conf.fatal('msvc: Could not find a valid architecture for building (get_msvc_version_3)')
+	env=dict(os.environ)
 	env.update(PATH=path)
 	compiler_name,linker_name,lib_name=_get_prog_names(conf,compiler)
 	cxx=conf.find_program(compiler_name,path_list=MSVC_PATH)
-	cxx=conf.cmd_to_list(cxx)
 	if'CL'in env:
 		del(env['CL'])
 	try:
-		try:
-			conf.cmd_and_log(cxx+['/help'],env=env)
-		except Exception ,e:
-			debug('msvc: get_msvc_version: %r %r %r -> failure'%(compiler,version,target))
-			debug(str(e))
-			conf.fatal('msvc: cannot run the compiler (in get_msvc_version)')
-		else:
-			debug('msvc: get_msvc_version: %r %r %r -> OK',compiler,version,target)
+		conf.cmd_and_log(cxx+['/help'],env=env)
+	except UnicodeError:
+		st=Utils.ex_stack()
+		if conf.logger:
+			conf.logger.error(st)
+		conf.fatal('msvc: Unicode error - check the code page?')
+	except Exception as e:
+		Logs.debug('msvc: get_msvc_version: %r %r %r -> failure %s',compiler,version,target,str(e))
+		conf.fatal('msvc: cannot run the compiler in get_msvc_version (run with -v to display errors)')
+	else:
+		Logs.debug('msvc: get_msvc_version: %r %r %r -> OK',compiler,version,target)
 	finally:
 		conf.env[compiler_name]=''
 	return(MSVC_PATH,MSVC_INCDIR,MSVC_LIBDIR)
+@conf
 def gather_wsdk_versions(conf,versions):
 	version_pattern=re.compile('^v..?.?\...?.?')
 	try:
-		all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Microsoft\\Microsoft SDKs\\Windows')
+		all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Microsoft\\Microsoft SDKs\\Windows')
 	except WindowsError:
 		try:
-			all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Microsoft\\Microsoft SDKs\\Windows')
+			all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Microsoft\\Microsoft SDKs\\Windows')
 		except WindowsError:
 			return
 	index=0
 	while 1:
 		try:
-			version=_winreg.EnumKey(all_versions,index)
+			version=Utils.winreg.EnumKey(all_versions,index)
 		except WindowsError:
 			break
-		index=index+1
+		index+=1
 		if not version_pattern.match(version):
 			continue
 		try:
-			msvc_version=_winreg.OpenKey(all_versions,version)
-			path,type=_winreg.QueryValueEx(msvc_version,'InstallationFolder')
+			msvc_version=Utils.winreg.OpenKey(all_versions,version)
+			path,type=Utils.winreg.QueryValueEx(msvc_version,'InstallationFolder')
 		except WindowsError:
 			continue
-		if os.path.isfile(os.path.join(path,'bin','SetEnv.cmd')):
-			targets=[]
+		if path and os.path.isfile(os.path.join(path,'bin','SetEnv.cmd')):
+			targets={}
 			for target,arch in all_msvc_platforms:
-				try:
-					targets.append((target,(arch,conf.get_msvc_version('wsdk',version,'/'+target,os.path.join(path,'bin','SetEnv.cmd')))))
-				except conf.errors.ConfigurationError:
-					pass
-			versions.append(('wsdk '+version[1:],targets))
+				targets[target]=target_compiler(conf,'wsdk',arch,version,'/'+target,os.path.join(path,'bin','SetEnv.cmd'))
+			versions['wsdk '+version[1:]]=targets
 def gather_wince_supported_platforms():
 	supported_wince_platforms=[]
 	try:
-		ce_sdk=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Microsoft\\Windows CE Tools\\SDKs')
+		ce_sdk=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Microsoft\\Windows CE Tools\\SDKs')
 	except WindowsError:
 		try:
-			ce_sdk=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Microsoft\\Windows CE Tools\\SDKs')
+			ce_sdk=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Microsoft\\Windows CE Tools\\SDKs')
 		except WindowsError:
 			ce_sdk=''
 	if not ce_sdk:
 		return supported_wince_platforms
-	ce_index=0
+	index=0
 	while 1:
 		try:
-			sdk_device=_winreg.EnumKey(ce_sdk,ce_index)
+			sdk_device=Utils.winreg.EnumKey(ce_sdk,index)
+			sdk=Utils.winreg.OpenKey(ce_sdk,sdk_device)
 		except WindowsError:
 			break
-		ce_index=ce_index+1
-		sdk=_winreg.OpenKey(ce_sdk,sdk_device)
+		index+=1
 		try:
-			path,type=_winreg.QueryValueEx(sdk,'SDKRootDir')
+			path,type=Utils.winreg.QueryValueEx(sdk,'SDKRootDir')
 		except WindowsError:
 			try:
-				path,type=_winreg.QueryValueEx(sdk,'SDKInformation')
-				path,xml=os.path.split(path)
+				path,type=Utils.winreg.QueryValueEx(sdk,'SDKInformation')
 			except WindowsError:
 				continue
+			path,xml=os.path.split(path)
 		path=str(path)
 		path,device=os.path.split(path)
 		if not device:
 			path,device=os.path.split(path)
+		platforms=[]
 		for arch,compiler in all_wince_platforms:
-			platforms=[]
 			if os.path.isdir(os.path.join(path,device,'Lib',arch)):
 				platforms.append((arch,compiler,os.path.join(path,device,'Include',arch),os.path.join(path,device,'Lib',arch)))
-			if platforms:
-				supported_wince_platforms.append((device,platforms))
+		if platforms:
+			supported_wince_platforms.append((device,platforms))
 	return supported_wince_platforms
 def gather_msvc_detected_versions():
 	version_pattern=re.compile('^(\d\d?\.\d\d?)(Exp)?$')
 	detected_versions=[]
-	for vcver,vcvar in[('VCExpress','Exp'),('VisualStudio','')]:
+	for vcver,vcvar in(('VCExpress','Exp'),('VisualStudio','')):
+		prefix='SOFTWARE\\Wow6432node\\Microsoft\\'+vcver
 		try:
-			prefix='SOFTWARE\\Wow6432node\\Microsoft\\'+vcver
-			all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,prefix)
+			all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,prefix)
 		except WindowsError:
+			prefix='SOFTWARE\\Microsoft\\'+vcver
 			try:
-				prefix='SOFTWARE\\Microsoft\\'+vcver
-				all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,prefix)
+				all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,prefix)
 			except WindowsError:
 				continue
 		index=0
 		while 1:
 			try:
-				version=_winreg.EnumKey(all_versions,index)
+				version=Utils.winreg.EnumKey(all_versions,index)
 			except WindowsError:
 				break
-			index=index+1
+			index+=1
 			match=version_pattern.match(version)
-			if not match:
-				continue
-			else:
+			if match:
 				versionnumber=float(match.group(1))
-			detected_versions.append((versionnumber,version+vcvar,prefix+"\\"+version))
+			else:
+				continue
+			detected_versions.append((versionnumber,version+vcvar,prefix+'\\'+version))
 	def fun(tup):
 		return tup[0]
-	try:
-		detected_versions.sort(key=fun)
-	except:
-		detected_versions.sort(lambda x,y:cmp(x[0],y[0]))
+	detected_versions.sort(key=fun)
 	return detected_versions
+class target_compiler(object):
+	def __init__(self,ctx,compiler,cpu,version,bat_target,bat,callback=None):
+		self.conf=ctx
+		self.name=None
+		self.is_valid=False
+		self.is_done=False
+		self.compiler=compiler
+		self.cpu=cpu
+		self.version=version
+		self.bat_target=bat_target
+		self.bat=bat
+		self.callback=callback
+	def evaluate(self):
+		if self.is_done:
+			return
+		self.is_done=True
+		try:
+			vs=self.conf.get_msvc_version(self.compiler,self.version,self.bat_target,self.bat)
+		except Errors.ConfigurationError:
+			self.is_valid=False
+			return
+		if self.callback:
+			vs=self.callback(self,vs)
+		self.is_valid=True
+		(self.bindirs,self.incdirs,self.libdirs)=vs
+	def __str__(self):
+		return str((self.bindirs,self.incdirs,self.libdirs))
+	def __repr__(self):
+		return repr((self.bindirs,self.incdirs,self.libdirs))
+@conf
 def gather_msvc_targets(conf,versions,version,vc_path):
-	targets=[]
+	targets={}
 	if os.path.isfile(os.path.join(vc_path,'vcvarsall.bat')):
 		for target,realtarget in all_msvc_platforms[::-1]:
-			try:
-				targets.append((target,(realtarget,conf.get_msvc_version('msvc',version,target,os.path.join(vc_path,'vcvarsall.bat')))))
-			except conf.errors.ConfigurationError:
-				pass
+			targets[target]=target_compiler(conf,'msvc',realtarget,version,target,os.path.join(vc_path,'vcvarsall.bat'))
 	elif os.path.isfile(os.path.join(vc_path,'Common7','Tools','vsvars32.bat')):
-		try:
-			targets.append(('x86',('x86',conf.get_msvc_version('msvc',version,'x86',os.path.join(vc_path,'Common7','Tools','vsvars32.bat')))))
-		except conf.errors.ConfigurationError:
-			pass
+		targets['x86']=target_compiler(conf,'msvc','x86',version,'x86',os.path.join(vc_path,'Common7','Tools','vsvars32.bat'))
 	elif os.path.isfile(os.path.join(vc_path,'Bin','vcvars32.bat')):
-		try:
-			targets.append(('x86',('x86',conf.get_msvc_version('msvc',version,'',os.path.join(vc_path,'Bin','vcvars32.bat')))))
-		except conf.errors.ConfigurationError:
-			pass
-	versions.append(('msvc '+version,targets))
+		targets['x86']=target_compiler(conf,'msvc','x86',version,'',os.path.join(vc_path,'Bin','vcvars32.bat'))
+	if targets:
+		versions['msvc '+version]=targets
+@conf
 def gather_wince_targets(conf,versions,version,vc_path,vsvars,supported_platforms):
 	for device,platforms in supported_platforms:
-		cetargets=[]
+		targets={}
 		for platform,compiler,include,lib in platforms:
 			winCEpath=os.path.join(vc_path,'ce')
 			if not os.path.isdir(winCEpath):
 				continue
-			try:
-				common_bindirs,_1,_2=conf.get_msvc_version('msvc',version,'x86',vsvars)
-			except conf.errors.ConfigurationError:
-				continue
 			if os.path.isdir(os.path.join(winCEpath,'lib',platform)):
-				bindirs=[os.path.join(winCEpath,'bin',compiler),os.path.join(winCEpath,'bin','x86_'+compiler)]+common_bindirs
+				bindirs=[os.path.join(winCEpath,'bin',compiler),os.path.join(winCEpath,'bin','x86_'+compiler)]
 				incdirs=[os.path.join(winCEpath,'include'),os.path.join(winCEpath,'atlmfc','include'),include]
 				libdirs=[os.path.join(winCEpath,'lib',platform),os.path.join(winCEpath,'atlmfc','lib',platform),lib]
-				cetargets.append((platform,(platform,(bindirs,incdirs,libdirs))))
-		if cetargets:
-			versions.append((device+' '+version,cetargets))
+				def combine_common(obj,compiler_env):
+					(common_bindirs,_1,_2)=compiler_env
+					return(bindirs+common_bindirs,incdirs,libdirs)
+				targets[platform]=target_compiler(conf,'msvc',platform,version,'x86',vsvars,combine_common)
+		if targets:
+			versions[device+' '+version]=targets
+@conf
+def gather_winphone_targets(conf,versions,version,vc_path,vsvars):
+	targets={}
+	for target,realtarget in all_msvc_platforms[::-1]:
+		targets[target]=target_compiler(conf,'winphone',realtarget,version,target,vsvars)
+	if targets:
+		versions['winphone '+version]=targets
+@conf
 def gather_msvc_versions(conf,versions):
 	vc_paths=[]
 	for(v,version,reg)in gather_msvc_detected_versions():
 		try:
 			try:
-				msvc_version=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,reg+"\\Setup\\VC")
+				msvc_version=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,reg+"\\Setup\\VC")
 			except WindowsError:
-				msvc_version=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,reg+"\\Setup\\Microsoft Visual C++")
-			path,type=_winreg.QueryValueEx(msvc_version,'ProductDir')
-			vc_paths.append((version,os.path.abspath(str(path))))
+				msvc_version=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,reg+"\\Setup\\Microsoft Visual C++")
+			path,type=Utils.winreg.QueryValueEx(msvc_version,'ProductDir')
 		except WindowsError:
 			continue
+		else:
+			vc_paths.append((version,os.path.abspath(str(path))))
 	wince_supported_platforms=gather_wince_supported_platforms()
 	for version,vc_path in vc_paths:
 		vs_path=os.path.dirname(vc_path)
@@ -267,72 +305,127 @@
 			conf.gather_wince_targets(versions,version,vc_path,vsvars,wince_supported_platforms)
 	for version,vc_path in vc_paths:
 		vs_path=os.path.dirname(vc_path)
+		vsvars=os.path.join(vs_path,'VC','WPSDK','WP80','vcvarsphoneall.bat')
+		if os.path.isfile(vsvars):
+			conf.gather_winphone_targets(versions,'8.0',vc_path,vsvars)
+			break
+	for version,vc_path in vc_paths:
+		vs_path=os.path.dirname(vc_path)
 		conf.gather_msvc_targets(versions,version,vc_path)
+@conf
 def gather_icl_versions(conf,versions):
 	version_pattern=re.compile('^...?.?\....?.?')
 	try:
-		all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Intel\\Compilers\\C++')
+		all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Intel\\Compilers\\C++')
 	except WindowsError:
 		try:
-			all_versions=_winreg.OpenKey(_winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Intel\\Compilers\\C++')
+			all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Intel\\Compilers\\C++')
 		except WindowsError:
 			return
 	index=0
 	while 1:
 		try:
-			version=_winreg.EnumKey(all_versions,index)
+			version=Utils.winreg.EnumKey(all_versions,index)
 		except WindowsError:
 			break
-		index=index+1
+		index+=1
 		if not version_pattern.match(version):
 			continue
-		targets=[]
+		targets={}
 		for target,arch in all_icl_platforms:
+			if target=='intel64':targetDir='EM64T_NATIVE'
+			else:targetDir=target
 			try:
-				if target=='intel64':targetDir='EM64T_NATIVE'
-				else:targetDir=target
-				_winreg.OpenKey(all_versions,version+'\\'+targetDir)
-				icl_version=_winreg.OpenKey(all_versions,version)
-				path,type=_winreg.QueryValueEx(icl_version,'ProductDir')
-				if os.path.isfile(os.path.join(path,'bin','iclvars.bat')):
-					try:
-						targets.append((target,(arch,conf.get_msvc_version('intel',version,target,os.path.join(path,'bin','iclvars.bat')))))
-					except conf.errors.ConfigurationError:
-						pass
+				Utils.winreg.OpenKey(all_versions,version+'\\'+targetDir)
+				icl_version=Utils.winreg.OpenKey(all_versions,version)
+				path,type=Utils.winreg.QueryValueEx(icl_version,'ProductDir')
 			except WindowsError:
 				pass
+			else:
+				batch_file=os.path.join(path,'bin','iclvars.bat')
+				if os.path.isfile(batch_file):
+					targets[target]=target_compiler(conf,'intel',arch,version,target,batch_file)
 		for target,arch in all_icl_platforms:
 			try:
-				icl_version=_winreg.OpenKey(all_versions,version+'\\'+target)
-				path,type=_winreg.QueryValueEx(icl_version,'ProductDir')
-				if os.path.isfile(os.path.join(path,'bin','iclvars.bat')):
-					try:
-						targets.append((target,(arch,conf.get_msvc_version('intel',version,target,os.path.join(path,'bin','iclvars.bat')))))
-					except conf.errors.ConfigurationError:
-						pass
+				icl_version=Utils.winreg.OpenKey(all_versions,version+'\\'+target)
+				path,type=Utils.winreg.QueryValueEx(icl_version,'ProductDir')
 			except WindowsError:
 				continue
+			else:
+				batch_file=os.path.join(path,'bin','iclvars.bat')
+				if os.path.isfile(batch_file):
+					targets[target]=target_compiler(conf,'intel',arch,version,target,batch_file)
 		major=version[0:2]
-		versions.append(('intel '+major,targets))
-def get_msvc_versions(conf):
-	if not conf.env['MSVC_INSTALLED_VERSIONS']:
-		lst=[]
-		conf.gather_icl_versions(lst)
-		conf.gather_wsdk_versions(lst)
-		conf.gather_msvc_versions(lst)
-		conf.env['MSVC_INSTALLED_VERSIONS']=lst
-	return conf.env['MSVC_INSTALLED_VERSIONS']
-def print_all_msvc_detected(conf):
-	for version,targets in conf.env['MSVC_INSTALLED_VERSIONS']:
-		info(version)
-		for target,l in targets:
-			info("\t"+target)
-def detect_msvc(conf):
-	versions=get_msvc_versions(conf)
-	return setup_msvc(conf,versions)
+		versions['intel '+major]=targets
+@conf
+def gather_intel_composer_versions(conf,versions):
+	version_pattern=re.compile('^...?.?\...?.?.?')
+	try:
+		all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Wow6432node\\Intel\\Suites')
+	except WindowsError:
+		try:
+			all_versions=Utils.winreg.OpenKey(Utils.winreg.HKEY_LOCAL_MACHINE,'SOFTWARE\\Intel\\Suites')
+		except WindowsError:
+			return
+	index=0
+	while 1:
+		try:
+			version=Utils.winreg.EnumKey(all_versions,index)
+		except WindowsError:
+			break
+		index+=1
+		if not version_pattern.match(version):
+			continue
+		targets={}
+		for target,arch in all_icl_platforms:
+			if target=='intel64':targetDir='EM64T_NATIVE'
+			else:targetDir=target
+			try:
+				try:
+					defaults=Utils.winreg.OpenKey(all_versions,version+'\\Defaults\\C++\\'+targetDir)
+				except WindowsError:
+					if targetDir=='EM64T_NATIVE':
+						defaults=Utils.winreg.OpenKey(all_versions,version+'\\Defaults\\C++\\EM64T')
+					else:
+						raise
+				uid,type=Utils.winreg.QueryValueEx(defaults,'SubKey')
+				Utils.winreg.OpenKey(all_versions,version+'\\'+uid+'\\C++\\'+targetDir)
+				icl_version=Utils.winreg.OpenKey(all_versions,version+'\\'+uid+'\\C++')
+				path,type=Utils.winreg.QueryValueEx(icl_version,'ProductDir')
+			except WindowsError:
+				pass
+			else:
+				batch_file=os.path.join(path,'bin','iclvars.bat')
+				if os.path.isfile(batch_file):
+					targets[target]=target_compiler(conf,'intel',arch,version,target,batch_file)
+				compilervars_warning_attr='_compilervars_warning_key'
+				if version[0:2]=='13'and getattr(conf,compilervars_warning_attr,True):
+					setattr(conf,compilervars_warning_attr,False)
+					patch_url='http://software.intel.com/en-us/forums/topic/328487'
+					compilervars_arch=os.path.join(path,'bin','compilervars_arch.bat')
+					for vscomntool in('VS110COMNTOOLS','VS100COMNTOOLS'):
+						if vscomntool in os.environ:
+							vs_express_path=os.environ[vscomntool]+r'..\IDE\VSWinExpress.exe'
+							dev_env_path=os.environ[vscomntool]+r'..\IDE\devenv.exe'
+							if(r'if exist "%VS110COMNTOOLS%..\IDE\VSWinExpress.exe"'in Utils.readf(compilervars_arch)and not os.path.exists(vs_express_path)and not os.path.exists(dev_env_path)):
+								Logs.warn(('The Intel compilervar_arch.bat only checks for one Visual Studio SKU ''(VSWinExpress.exe) but it does not seem to be installed at %r. ''The intel command line set up will fail to configure unless the file %r''is patched. See: %s')%(vs_express_path,compilervars_arch,patch_url))
+		major=version[0:2]
+		versions['intel '+major]=targets
+@conf
+def detect_msvc(self):
+	return self.setup_msvc(self.get_msvc_versions())
+@conf
+def get_msvc_versions(self):
+	dct={}
+	self.gather_icl_versions(dct)
+	self.gather_intel_composer_versions(dct)
+	self.gather_wsdk_versions(dct)
+	self.gather_msvc_versions(dct)
+	return dct
+@conf
 def find_lt_names_msvc(self,libname,is_static=False):
 	lt_names=['lib%s.la'%libname,'%s.la'%libname,]
-	for path in self.env['LIBPATH']:
+	for path in self.env.LIBPATH:
 		for la in lt_names:
 			laf=os.path.join(path,la)
 			dll=None
@@ -357,6 +450,7 @@
 				else:
 					raise self.errors.WafError('invalid libtool object file: %s'%laf)
 	return(None,None,None)
+@conf
 def libname_msvc(self,libname,is_static=False):
 	lib=libname.lower()
 	lib=re.sub('\.lib$','',lib)
@@ -370,9 +464,9 @@
 		if lt_static==True:
 			return os.path.join(lt_path,lt_libname)
 	if lt_path!=None:
-		_libpaths=[lt_path]+self.env['LIBPATH']
+		_libpaths=[lt_path]+self.env.LIBPATH
 	else:
-		_libpaths=self.env['LIBPATH']
+		_libpaths=self.env.LIBPATH
 	static_libs=['lib%ss.lib'%lib,'lib%s.lib'%lib,'%ss.lib'%lib,'%s.lib'%lib,]
 	dynamic_libs=['lib%s.dll.lib'%lib,'lib%s.dll.a'%lib,'%s.dll.lib'%lib,'%s.dll.a'%lib,'lib%s_d.lib'%lib,'%s_d.lib'%lib,'%s.lib'%lib,]
 	libnames=static_libs
@@ -381,10 +475,11 @@
 	for path in _libpaths:
 		for libn in libnames:
 			if os.path.exists(os.path.join(path,libn)):
-				debug('msvc: lib found: %s'%os.path.join(path,libn))
+				Logs.debug('msvc: lib found: %s',os.path.join(path,libn))
 				return re.sub('\.lib$','',libn)
-	self.fatal("The library %r could not be found"%libname)
+	self.fatal('The library %r could not be found'%libname)
 	return re.sub('\.lib$','',libname)
+@conf
 def check_lib_msvc(self,libname,is_static=False,uselib_store=None):
 	libn=self.libname_msvc(libname,is_static)
 	if not uselib_store:
@@ -393,11 +488,12 @@
 		self.env['STLIB_'+uselib_store]=[libn]
 	else:
 		self.env['LIB_'+uselib_store]=[libn]
+@conf
 def check_libs_msvc(self,libnames,is_static=False):
 	for libname in Utils.to_list(libnames):
 		self.check_lib_msvc(libname,is_static)
 def configure(conf):
-	conf.autodetect()
+	conf.autodetect(True)
 	conf.find_msvc()
 	conf.msvc_common_flags()
 	conf.cc_load_tools()
@@ -406,22 +502,26 @@
 	conf.cxx_add_flags()
 	conf.link_add_flags()
 	conf.visual_studio_add_flags()
+@conf
 def no_autodetect(conf):
 	conf.env.NO_MSVC_DETECT=1
 	configure(conf)
-def autodetect(conf):
+@conf
+def autodetect(conf,arch=False):
 	v=conf.env
 	if v.NO_MSVC_DETECT:
 		return
-	compiler,version,path,includes,libdirs=conf.detect_msvc()
-	v['PATH']=path
-	v['INCLUDES']=includes
-	v['LIBPATH']=libdirs
-	v['MSVC_COMPILER']=compiler
+	compiler,version,path,includes,libdirs,cpu=conf.detect_msvc()
+	if arch:
+		v.DEST_CPU=cpu
+	v.PATH=path
+	v.INCLUDES=includes
+	v.LIBPATH=libdirs
+	v.MSVC_COMPILER=compiler
 	try:
-		v['MSVC_VERSION']=float(version)
-	except:
-		v['MSVC_VERSION']=float(version[:-3])
+		v.MSVC_VERSION=float(version)
+	except TypeError:
+		v.MSVC_VERSION=float(version[:-3])
 def _get_prog_names(conf,compiler):
 	if compiler=='intel':
 		compiler_name='ICL'
@@ -432,90 +532,89 @@
 		linker_name='LINK'
 		lib_name='LIB'
 	return compiler_name,linker_name,lib_name
+@conf
 def find_msvc(conf):
 	if sys.platform=='cygwin':
 		conf.fatal('MSVC module does not work under cygwin Python!')
 	v=conf.env
-	path=v['PATH']
-	compiler=v['MSVC_COMPILER']
-	version=v['MSVC_VERSION']
+	path=v.PATH
+	compiler=v.MSVC_COMPILER
+	version=v.MSVC_VERSION
 	compiler_name,linker_name,lib_name=_get_prog_names(conf,compiler)
 	v.MSVC_MANIFEST=(compiler=='msvc'and version>=8)or(compiler=='wsdk'and version>=6)or(compiler=='intel'and version>=11)
-	cxx=None
-	if v['CXX']:cxx=v['CXX']
-	elif'CXX'in conf.environ:cxx=conf.environ['CXX']
 	cxx=conf.find_program(compiler_name,var='CXX',path_list=path)
-	cxx=conf.cmd_to_list(cxx)
 	env=dict(conf.environ)
 	if path:env.update(PATH=';'.join(path))
 	if not conf.cmd_and_log(cxx+['/nologo','/help'],env=env):
 		conf.fatal('the msvc compiler could not be identified')
-	v['CC']=v['CXX']=cxx
-	v['CC_NAME']=v['CXX_NAME']='msvc'
-	if not v['LINK_CXX']:
-		link=conf.find_program(linker_name,path_list=path)
-		if link:v['LINK_CXX']=link
-		else:conf.fatal('%s was not found (linker)'%linker_name)
-		v['LINK']=link
-	if not v['LINK_CC']:
-		v['LINK_CC']=v['LINK_CXX']
-	if not v['AR']:
+	v.CC=v.CXX=cxx
+	v.CC_NAME=v.CXX_NAME='msvc'
+	if not v.LINK_CXX:
+		v.LINK_CXX=conf.find_program(linker_name,path_list=path,errmsg='%s was not found (linker)'%linker_name)
+	if not v.LINK_CC:
+		v.LINK_CC=v.LINK_CXX
+	if not v.AR:
 		stliblink=conf.find_program(lib_name,path_list=path,var='AR')
-		if not stliblink:return
-		v['ARFLAGS']=['/NOLOGO']
+		if not stliblink:
+			return
+		v.ARFLAGS=['/nologo']
 	if v.MSVC_MANIFEST:
-		mt=conf.find_program('MT',path_list=path,var='MT')
-		v['MTFLAGS']=['/NOLOGO']
-	conf.load('winres')
-	if not conf.env['WINRC']:
-		warn('Resource compiler not found. Compiling resource file is disabled')
+		conf.find_program('MT',path_list=path,var='MT')
+		v.MTFLAGS=['/nologo']
+	try:
+		conf.load('winres')
+	except Errors.ConfigurationError:
+		Logs.warn('Resource compiler not found. Compiling resource file is disabled')
+@conf
 def visual_studio_add_flags(self):
 	v=self.env
-	try:v.prepend_value('INCLUDES',self.environ['INCLUDE'].split(';'))
-	except:pass
-	try:v.prepend_value('LIBPATH',self.environ['LIB'].split(';'))
-	except:pass
+	if self.environ.get('INCLUDE'):
+		v.prepend_value('INCLUDES',[x for x in self.environ['INCLUDE'].split(';')if x])
+	if self.environ.get('LIB'):
+		v.prepend_value('LIBPATH',[x for x in self.environ['LIB'].split(';')if x])
+@conf
 def msvc_common_flags(conf):
 	v=conf.env
-	v['DEST_BINFMT']='pe'
+	v.DEST_BINFMT='pe'
 	v.append_value('CFLAGS',['/nologo'])
 	v.append_value('CXXFLAGS',['/nologo'])
-	v['DEFINES_ST']='/D%s'
-	v['CC_SRC_F']=''
-	v['CC_TGT_F']=['/c','/Fo']
-	if v['MSVC_VERSION']>=8:
-		v['CC_TGT_F']=['/FC']+v['CC_TGT_F']
-	v['CXX_SRC_F']=''
-	v['CXX_TGT_F']=['/c','/Fo']
-	if v['MSVC_VERSION']>=8:
-		v['CXX_TGT_F']=['/FC']+v['CXX_TGT_F']
-	v['CPPPATH_ST']='/I%s'
-	v['AR_TGT_F']=v['CCLNK_TGT_F']=v['CXXLNK_TGT_F']='/OUT:'
-	v['CFLAGS_CONSOLE']=v['CXXFLAGS_CONSOLE']=['/SUBSYSTEM:CONSOLE']
-	v['CFLAGS_NATIVE']=v['CXXFLAGS_NATIVE']=['/SUBSYSTEM:NATIVE']
-	v['CFLAGS_POSIX']=v['CXXFLAGS_POSIX']=['/SUBSYSTEM:POSIX']
-	v['CFLAGS_WINDOWS']=v['CXXFLAGS_WINDOWS']=['/SUBSYSTEM:WINDOWS']
-	v['CFLAGS_WINDOWSCE']=v['CXXFLAGS_WINDOWSCE']=['/SUBSYSTEM:WINDOWSCE']
-	v['CFLAGS_CRT_MULTITHREADED']=v['CXXFLAGS_CRT_MULTITHREADED']=['/MT']
-	v['CFLAGS_CRT_MULTITHREADED_DLL']=v['CXXFLAGS_CRT_MULTITHREADED_DLL']=['/MD']
-	v['CFLAGS_CRT_MULTITHREADED_DBG']=v['CXXFLAGS_CRT_MULTITHREADED_DBG']=['/MTd']
-	v['CFLAGS_CRT_MULTITHREADED_DLL_DBG']=v['CXXFLAGS_CRT_MULTITHREADED_DLL_DBG']=['/MDd']
-	v['LIB_ST']='%s.lib'
-	v['LIBPATH_ST']='/LIBPATH:%s'
-	v['STLIB_ST']='lib%s.lib'
-	v['STLIBPATH_ST']='/LIBPATH:%s'
-	v.append_value('LINKFLAGS',['/NOLOGO'])
-	if v['MSVC_MANIFEST']:
+	v.append_value('LINKFLAGS',['/nologo'])
+	v.DEFINES_ST='/D%s'
+	v.CC_SRC_F=''
+	v.CC_TGT_F=['/c','/Fo']
+	v.CXX_SRC_F=''
+	v.CXX_TGT_F=['/c','/Fo']
+	if(v.MSVC_COMPILER=='msvc'and v.MSVC_VERSION>=8)or(v.MSVC_COMPILER=='wsdk'and v.MSVC_VERSION>=6):
+		v.CC_TGT_F=['/FC']+v.CC_TGT_F
+		v.CXX_TGT_F=['/FC']+v.CXX_TGT_F
+	v.CPPPATH_ST='/I%s'
+	v.AR_TGT_F=v.CCLNK_TGT_F=v.CXXLNK_TGT_F='/OUT:'
+	v.CFLAGS_CONSOLE=v.CXXFLAGS_CONSOLE=['/SUBSYSTEM:CONSOLE']
+	v.CFLAGS_NATIVE=v.CXXFLAGS_NATIVE=['/SUBSYSTEM:NATIVE']
+	v.CFLAGS_POSIX=v.CXXFLAGS_POSIX=['/SUBSYSTEM:POSIX']
+	v.CFLAGS_WINDOWS=v.CXXFLAGS_WINDOWS=['/SUBSYSTEM:WINDOWS']
+	v.CFLAGS_WINDOWSCE=v.CXXFLAGS_WINDOWSCE=['/SUBSYSTEM:WINDOWSCE']
+	v.CFLAGS_CRT_MULTITHREADED=v.CXXFLAGS_CRT_MULTITHREADED=['/MT']
+	v.CFLAGS_CRT_MULTITHREADED_DLL=v.CXXFLAGS_CRT_MULTITHREADED_DLL=['/MD']
+	v.CFLAGS_CRT_MULTITHREADED_DBG=v.CXXFLAGS_CRT_MULTITHREADED_DBG=['/MTd']
+	v.CFLAGS_CRT_MULTITHREADED_DLL_DBG=v.CXXFLAGS_CRT_MULTITHREADED_DLL_DBG=['/MDd']
+	v.LIB_ST='%s.lib'
+	v.LIBPATH_ST='/LIBPATH:%s'
+	v.STLIB_ST='%s.lib'
+	v.STLIBPATH_ST='/LIBPATH:%s'
+	if v.MSVC_MANIFEST:
 		v.append_value('LINKFLAGS',['/MANIFEST'])
-	v['CFLAGS_cshlib']=[]
-	v['CXXFLAGS_cxxshlib']=[]
-	v['LINKFLAGS_cshlib']=v['LINKFLAGS_cxxshlib']=['/DLL']
-	v['cshlib_PATTERN']=v['cxxshlib_PATTERN']='%s.dll'
-	v['implib_PATTERN']='%s.lib'
-	v['IMPLIB_ST']='/IMPLIB:%s'
-	v['LINKFLAGS_cstlib']=[]
-	v['cstlib_PATTERN']=v['cxxstlib_PATTERN']='lib%s.lib'
-	v['cprogram_PATTERN']=v['cxxprogram_PATTERN']='%s.exe'
+	v.CFLAGS_cshlib=[]
+	v.CXXFLAGS_cxxshlib=[]
+	v.LINKFLAGS_cshlib=v.LINKFLAGS_cxxshlib=['/DLL']
+	v.cshlib_PATTERN=v.cxxshlib_PATTERN='%s.dll'
+	v.implib_PATTERN='%s.lib'
+	v.IMPLIB_ST='/IMPLIB:%s'
+	v.LINKFLAGS_cstlib=[]
+	v.cstlib_PATTERN=v.cxxstlib_PATTERN='%s.lib'
+	v.cprogram_PATTERN=v.cxxprogram_PATTERN='%s.exe'
+@after_method('apply_link')
+@feature('c','cxx')
 def apply_flags_msvc(self):
 	if self.env.CC_NAME!='msvc'or not getattr(self,'link_task',None):
 		return
@@ -531,124 +630,31 @@
 			if d[1:]=='debug':
 				pdbnode=self.link_task.outputs[0].change_ext('.pdb')
 				self.link_task.outputs.append(pdbnode)
-				try:
-					self.install_task.source.append(pdbnode)
-				except AttributeError:
-					pass
+				if getattr(self,'install_task',None):
+					self.pdb_install_task=self.add_install_files(install_to=self.install_task.install_to,install_from=pdbnode)
 				break
+@feature('cprogram','cshlib','cxxprogram','cxxshlib')
+@after_method('apply_link')
 def apply_manifest(self):
 	if self.env.CC_NAME=='msvc'and self.env.MSVC_MANIFEST and getattr(self,'link_task',None):
 		out_node=self.link_task.outputs[0]
 		man_node=out_node.parent.find_or_declare(out_node.name+'.manifest')
 		self.link_task.outputs.append(man_node)
-		self.link_task.do_manifest=True
-def exec_mf(self):
-	env=self.env
-	mtool=env['MT']
-	if not mtool:
-		return 0
-	self.do_manifest=False
-	outfile=self.outputs[0].abspath()
-	manifest=None
-	for out_node in self.outputs:
-		if out_node.name.endswith('.manifest'):
-			manifest=out_node.abspath()
-			break
-	if manifest is None:
-		return 0
-	mode=''
-	if'cprogram'in self.generator.features or'cxxprogram'in self.generator.features:
-		mode='1'
-	elif'cshlib'in self.generator.features or'cxxshlib'in self.generator.features:
-		mode='2'
-	debug('msvc: embedding manifest in mode %r'%mode)
-	lst=[]
-	lst.append(env['MT'])
-	lst.extend(Utils.to_list(env['MTFLAGS']))
-	lst.extend(['-manifest',manifest])
-	lst.append('-outputresource:%s;%s'%(outfile,mode))
-	lst=[lst]
-	return self.exec_command(*lst)
-def quote_response_command(self,flag):
-	if flag.find(' ')>-1:
-		for x in('/LIBPATH:','/IMPLIB:','/OUT:','/I'):
-			if flag.startswith(x):
-				flag='%s"%s"'%(x,flag[len(x):])
-				break
-		else:
-			flag='"%s"'%flag
-	return flag
-def exec_response_command(self,cmd,**kw):
-	try:
-		tmp=None
-		if sys.platform.startswith('win')and isinstance(cmd,list)and len(' '.join(cmd))>=8192:
-			program=cmd[0]
-			cmd=[self.quote_response_command(x)for x in cmd]
-			(fd,tmp)=tempfile.mkstemp()
-			os.write(fd,'\r\n'.join(i.replace('\\','\\\\')for i in cmd[1:]))
-			os.close(fd)
-			cmd=[program,'@'+tmp]
-		ret=self.generator.bld.exec_command(cmd,**kw)
-	finally:
-		if tmp:
-			try:
-				os.remove(tmp)
-			except:
-				pass
-	return ret
-def exec_command_msvc(self,*k,**kw):
-	if self.env['CC_NAME']=='msvc':
-		if isinstance(k[0],list):
-			lst=[]
-			carry=''
-			for a in k[0]:
-				if a=='/Fo'or a=='/doc'or a[-1]==':':
-					carry=a
-				else:
-					lst.append(carry+a)
-					carry=''
-			k=[lst]
-		if self.env['PATH']:
-			env=dict(os.environ)
-			env.update(PATH=';'.join(self.env['PATH']))
-			kw['env']=env
-	bld=self.generator.bld
-	try:
-		if not kw.get('cwd',None):
-			kw['cwd']=bld.cwd
-	except AttributeError:
-		bld.cwd=kw['cwd']=bld.variant_dir
-	ret=self.exec_response_command(k[0],**kw)
-	if not ret and getattr(self,'do_manifest',None):
-		ret=self.exec_mf()
-	return ret
-for k in'c cxx cprogram cxxprogram cshlib cxxshlib cstlib cxxstlib'.split():
-	cls=Task.classes.get(k,None)
-	if cls:
-		cls.exec_command=exec_command_msvc
-		cls.exec_response_command=exec_response_command
-		cls.quote_response_command=quote_response_command
-		cls.exec_mf=exec_mf
-
-conf(get_msvc_version)
-conf(gather_wsdk_versions)
-conf(gather_msvc_targets)
-conf(gather_wince_targets)
-conf(gather_msvc_versions)
-conf(gather_icl_versions)
-conf(get_msvc_versions)
-conf(print_all_msvc_detected)
-conf(detect_msvc)
-conf(find_lt_names_msvc)
-conf(libname_msvc)
-conf(check_lib_msvc)
-conf(check_libs_msvc)
-conf(no_autodetect)
-conf(autodetect)
-conf(find_msvc)
-conf(visual_studio_add_flags)
-conf(msvc_common_flags)
-after_method('apply_link')(apply_flags_msvc)
-feature('c','cxx')(apply_flags_msvc)
-feature('cprogram','cshlib','cxxprogram','cxxshlib')(apply_manifest)
-after_method('apply_link')(apply_manifest)
\ No newline at end of file
+		self.env.DO_MANIFEST=True
+def make_winapp(self,family):
+	append=self.env.append_unique
+	append('DEFINES','WINAPI_FAMILY=%s'%family)
+	append('CXXFLAGS',['/ZW','/TP'])
+	for lib_path in self.env.LIBPATH:
+		append('CXXFLAGS','/AI%s'%lib_path)
+@feature('winphoneapp')
+@after_method('process_use')
+@after_method('propagate_uselib_vars')
+def make_winphone_app(self):
+	make_winapp(self,'WINAPI_FAMILY_PHONE_APP')
+	conf.env.append_unique('LINKFLAGS',['/NODEFAULTLIB:ole32.lib','PhoneAppModelHost.lib'])
+@feature('winapp')
+@after_method('process_use')
+@after_method('propagate_uselib_vars')
+def make_windows_app(self):
+	make_winapp(self,'WINAPI_FAMILY_DESKTOP_APP')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/nasm.py glmark2-2021.02/waflib/Tools/nasm.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/nasm.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/nasm.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,14 +1,16 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
+import os
 import waflib.Tools.asm
 from waflib.TaskGen import feature
+@feature('asm')
 def apply_nasm_vars(self):
 	self.env.append_value('ASFLAGS',self.to_list(getattr(self,'nasm_flags',[])))
 def configure(conf):
-	nasm=conf.find_program(['nasm','yasm'],var='AS')
+	conf.find_program(['nasm','yasm'],var='AS')
 	conf.env.AS_TGT_F=['-o']
 	conf.env.ASLNK_TGT_F=['-o']
-
-feature('asm')(apply_nasm_vars)
\ No newline at end of file
+	conf.load('asm')
+	conf.env.ASMPATH_ST='-I%s'+os.sep
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/nobuild.py glmark2-2021.02/waflib/Tools/nobuild.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/nobuild.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/nobuild.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,11 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+from waflib import Task
+def build(bld):
+	def run(self):
+		for x in self.outputs:
+			x.write('')
+	for(name,cls)in Task.classes.items():
+		cls.run=run
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/perl.py glmark2-2021.02/waflib/Tools/perl.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/perl.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/perl.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,15 +1,18 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os
 from waflib import Task,Options,Utils
 from waflib.Configure import conf
 from waflib.TaskGen import extension,feature,before_method
+@before_method('apply_incpaths','apply_link','propagate_uselib_vars')
+@feature('perlext')
 def init_perlext(self):
 	self.uselib=self.to_list(getattr(self,'uselib',[]))
 	if not'PERLEXT'in self.uselib:self.uselib.append('PERLEXT')
-	self.env['cshlib_PATTERN']=self.env['cxxshlib_PATTERN']=self.env['perlext_PATTERN']
+	self.env.cshlib_PATTERN=self.env.cxxshlib_PATTERN=self.env.perlext_PATTERN
+@extension('.xs')
 def xsubpp_file(self,node):
 	outnode=node.change_ext('.c')
 	self.create_task('xsubpp',node,outnode)
@@ -18,6 +21,7 @@
 	run_str='${PERL} ${XSUBPP} -noprototypes -typemap ${EXTUTILS_TYPEMAP} ${SRC} > ${TGT}'
 	color='BLUE'
 	ext_out=['.h']
+@conf
 def check_perl_version(self,minver=None):
 	res=True
 	if minver:
@@ -25,14 +29,8 @@
 	else:
 		cver=''
 	self.start_msg('Checking for minimum perl version %s'%cver)
-	perl=getattr(Options.options,'perlbinary',None)
-	if not perl:
-		perl=self.find_program('perl',var='PERL')
-	if not perl:
-		self.end_msg("Perl not found",color="YELLOW")
-		return False
-	self.env['PERL']=perl
-	version=self.cmd_and_log([perl,"-e",'printf \"%vd\", $^V'])
+	perl=self.find_program('perl',var='PERL',value=getattr(Options.options,'perlbinary',None))
+	version=self.cmd_and_log(perl+["-e",'printf \"%vd\", $^V'])
 	if not version:
 		res=False
 		version="Unknown"
@@ -40,42 +38,47 @@
 		ver=tuple(map(int,version.split(".")))
 		if ver<minver:
 			res=False
-	self.end_msg(version,color=res and"GREEN"or"YELLOW")
+	self.end_msg(version,color=res and'GREEN'or'YELLOW')
 	return res
+@conf
 def check_perl_module(self,module):
-	cmd=[self.env['PERL'],'-e','use %s'%module]
+	cmd=self.env.PERL+['-e','use %s'%module]
 	self.start_msg('perl module %s'%module)
 	try:
 		r=self.cmd_and_log(cmd)
-	except:
+	except Exception:
 		self.end_msg(False)
 		return None
 	self.end_msg(r or True)
 	return r
+@conf
 def check_perl_ext_devel(self):
 	env=self.env
 	perl=env.PERL
 	if not perl:
 		self.fatal('find perl first')
-	def read_out(cmd):
-		return Utils.to_list(self.cmd_and_log(perl+cmd))
-	env['LINKFLAGS_PERLEXT']=read_out(" -MConfig -e'print $Config{lddlflags}'")
-	env['INCLUDES_PERLEXT']=read_out(" -MConfig -e'print \"$Config{archlib}/CORE\"'")
-	env['CFLAGS_PERLEXT']=read_out(" -MConfig -e'print \"$Config{ccflags} $Config{cccdlflags}\"'")
-	env['XSUBPP']=read_out(" -MConfig -e'print \"$Config{privlib}/ExtUtils/xsubpp$Config{exe_ext}\"'")
-	env['EXTUTILS_TYPEMAP']=read_out(" -MConfig -e'print \"$Config{privlib}/ExtUtils/typemap\"'")
+	def cmd_perl_config(s):
+		return perl+['-MConfig','-e','print \"%s\"'%s]
+	def cfg_str(cfg):
+		return self.cmd_and_log(cmd_perl_config(cfg))
+	def cfg_lst(cfg):
+		return Utils.to_list(cfg_str(cfg))
+	def find_xsubpp():
+		for var in('privlib','vendorlib'):
+			xsubpp=cfg_lst('$Config{%s}/ExtUtils/xsubpp$Config{exe_ext}'%var)
+			if xsubpp and os.path.isfile(xsubpp[0]):
+				return xsubpp
+		return self.find_program('xsubpp')
+	env.LINKFLAGS_PERLEXT=cfg_lst('$Config{lddlflags}')
+	env.INCLUDES_PERLEXT=cfg_lst('$Config{archlib}/CORE')
+	env.CFLAGS_PERLEXT=cfg_lst('$Config{ccflags} $Config{cccdlflags}')
+	env.EXTUTILS_TYPEMAP=cfg_lst('$Config{privlib}/ExtUtils/typemap')
+	env.XSUBPP=find_xsubpp()
 	if not getattr(Options.options,'perlarchdir',None):
-		env['ARCHDIR_PERL']=self.cmd_and_log(perl+" -MConfig -e'print $Config{sitearch}'")
+		env.ARCHDIR_PERL=cfg_str('$Config{sitearch}')
 	else:
-		env['ARCHDIR_PERL']=getattr(Options.options,'perlarchdir')
-	env['perlext_PATTERN']='%s.'+self.cmd_and_log(perl+" -MConfig -e'print $Config{dlext}'")
+		env.ARCHDIR_PERL=getattr(Options.options,'perlarchdir')
+	env.perlext_PATTERN='%s.'+cfg_str('$Config{dlext}')
 def options(opt):
 	opt.add_option('--with-perl-binary',type='string',dest='perlbinary',help='Specify alternate perl binary',default=None)
 	opt.add_option('--with-perl-archdir',type='string',dest='perlarchdir',help='Specify directory where to install arch specific files',default=None)
-
-before_method('apply_incpaths','apply_link','propagate_uselib_vars')(init_perlext)
-feature('perlext')(init_perlext)
-extension('.xs')(xsubpp_file)
-conf(check_perl_version)
-conf(check_perl_module)
-conf(check_perl_ext_devel)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/python.py glmark2-2021.02/waflib/Tools/python.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/python.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/python.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,10 +1,9 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os,sys
-from waflib import Utils,Options,Errors
-from waflib.Logs import debug,warn,info,error
+from waflib import Utils,Options,Errors,Logs,Task,Node
 from waflib.TaskGen import extension,before_method,after_method,feature
 from waflib.Configure import conf
 FRAG='''
@@ -17,8 +16,9 @@
 #ifdef __cplusplus
 }
 #endif
-int main()
+int main(int argc, char **argv)
 {
+   (void)argc; (void)argv;
    Py_Initialize();
    Py_Finalize();
    return 0;
@@ -26,79 +26,91 @@
 '''
 INST='''
 import sys, py_compile
-py_compile.compile(sys.argv[1], sys.argv[2], sys.argv[3])
+py_compile.compile(sys.argv[1], sys.argv[2], sys.argv[3], True)
 '''
 DISTUTILS_IMP=['from distutils.sysconfig import get_config_var, get_python_lib']
-def process_py(self,node):
-	try:
-		if not self.bld.is_install:
-			return
-	except:
-		return
-	try:
-		if not self.install_path:
-			return
-	except AttributeError:
-		self.install_path='${PYTHONDIR}'
-	def inst_py(ctx):
-		install_from=getattr(self,'install_from',None)
-		if install_from:
-			install_from=self.path.find_dir(install_from)
-		install_pyfile(self,node,install_from)
-	self.bld.add_post_fun(inst_py)
-def install_pyfile(self,node,install_from=None):
-	from_node=install_from or node.parent
-	tsk=self.bld.install_as(self.install_path+'/'+node.path_from(from_node),node,postpone=False)
-	path=tsk.get_install_path()
-	if self.bld.is_install<0:
-		info("+ removing byte compiled python files")
-		for x in'co':
-			try:
-				os.remove(path+x)
-			except OSError:
-				pass
-	if self.bld.is_install>0:
-		try:
-			st1=os.stat(path)
-		except:
-			error('The python file is missing, this should not happen')
-		for x in['c','o']:
-			do_inst=self.env['PY'+x.upper()]
-			try:
-				st2=os.stat(path+x)
-			except OSError:
-				pass
-			else:
-				if st1.st_mtime<=st2.st_mtime:
-					do_inst=False
-			if do_inst:
-				lst=(x=='o')and[self.env['PYFLAGS_OPT']]or[]
-				(a,b,c)=(path,path+x,tsk.get_install_path(destdir=False)+x)
-				argv=self.env['PYTHON']+lst+['-c',INST,a,b,c]
-				info('+ byte compiling %r'%(path+x))
-				env=self.env.env or None
-				ret=Utils.subprocess.Popen(argv,env=env).wait()
-				if ret:
-					raise Errors.WafError('py%s compilation failed %r'%(x,path))
+@before_method('process_source')
+@feature('py')
 def feature_py(self):
-	pass
+	self.install_path=getattr(self,'install_path','${PYTHONDIR}')
+	install_from=getattr(self,'install_from',None)
+	if install_from and not isinstance(install_from,Node.Node):
+		install_from=self.path.find_dir(install_from)
+	self.install_from=install_from
+	ver=self.env.PYTHON_VERSION
+	if not ver:
+		self.bld.fatal('Installing python files requires PYTHON_VERSION, try conf.check_python_version')
+	if int(ver.replace('.',''))>31:
+		self.install_32=True
+@extension('.py')
+def process_py(self,node):
+	assert(getattr(self,'install_path')),'add features="py"'
+	if self.install_path:
+		if self.install_from:
+			self.add_install_files(install_to=self.install_path,install_from=node,cwd=self.install_from,relative_trick=True)
+		else:
+			self.add_install_files(install_to=self.install_path,install_from=node,relative_trick=True)
+	lst=[]
+	if self.env.PYC:
+		lst.append('pyc')
+	if self.env.PYO:
+		lst.append('pyo')
+	if self.install_path:
+		if self.install_from:
+			pyd=Utils.subst_vars("%s/%s"%(self.install_path,node.path_from(self.install_from)),self.env)
+		else:
+			pyd=Utils.subst_vars("%s/%s"%(self.install_path,node.path_from(self.path)),self.env)
+	else:
+		pyd=node.abspath()
+	for ext in lst:
+		if self.env.PYTAG:
+			name=node.name[:-3]
+			pyobj=node.parent.get_bld().make_node('__pycache__').make_node("%s.%s.%s"%(name,self.env.PYTAG,ext))
+			pyobj.parent.mkdir()
+		else:
+			pyobj=node.change_ext(".%s"%ext)
+		tsk=self.create_task(ext,node,pyobj)
+		tsk.pyd=pyd
+		if self.install_path:
+			self.add_install_files(install_to=os.path.dirname(pyd),install_from=pyobj,cwd=node.parent.get_bld(),relative_trick=True)
+class pyc(Task.Task):
+	color='PINK'
+	def run(self):
+		cmd=[Utils.subst_vars('${PYTHON}',self.env),'-c',INST,self.inputs[0].abspath(),self.outputs[0].abspath(),self.pyd]
+		ret=self.generator.bld.exec_command(cmd)
+		return ret
+class pyo(Task.Task):
+	color='PINK'
+	def run(self):
+		cmd=[Utils.subst_vars('${PYTHON}',self.env),Utils.subst_vars('${PYFLAGS_OPT}',self.env),'-c',INST,self.inputs[0].abspath(),self.outputs[0].abspath(),self.pyd]
+		ret=self.generator.bld.exec_command(cmd)
+		return ret
+@feature('pyext')
+@before_method('propagate_uselib_vars','apply_link')
+@after_method('apply_bundle')
 def init_pyext(self):
+	self.uselib=self.to_list(getattr(self,'uselib',[]))
+	if not'PYEXT'in self.uselib:
+		self.uselib.append('PYEXT')
+	self.env.cshlib_PATTERN=self.env.cxxshlib_PATTERN=self.env.macbundle_PATTERN=self.env.pyext_PATTERN
+	self.env.fcshlib_PATTERN=self.env.dshlib_PATTERN=self.env.pyext_PATTERN
 	try:
 		if not self.install_path:
 			return
 	except AttributeError:
 		self.install_path='${PYTHONARCHDIR}'
-	self.uselib=self.to_list(getattr(self,'uselib',[]))
-	if not'PYEXT'in self.uselib:
-		self.uselib.append('PYEXT')
-	self.env['cshlib_PATTERN']=self.env['cxxshlib_PATTERN']=self.env['macbundle_PATTERN']=self.env['pyext_PATTERN']
+@feature('pyext')
+@before_method('apply_link','apply_bundle')
 def set_bundle(self):
 	if Utils.unversioned_sys_platform()=='darwin':
 		self.mac_bundle=True
+@before_method('propagate_uselib_vars')
+@feature('pyembed')
 def init_pyembed(self):
 	self.uselib=self.to_list(getattr(self,'uselib',[]))
 	if not'PYEMBED'in self.uselib:
 		self.uselib.append('PYEMBED')
+@conf
 def get_python_variables(self,variables,imports=None):
 	if not imports:
 		try:
@@ -118,48 +130,119 @@
 		out=self.cmd_and_log(self.env.PYTHON+['-c','\n'.join(program)],env=os_env)
 	except Errors.WafError:
 		self.fatal('The distutils module is unusable: install "python-devel"?')
+	self.to_log(out)
 	return_values=[]
-	for s in out.split('\n'):
+	for s in out.splitlines():
 		s=s.strip()
 		if not s:
 			continue
 		if s=='None':
 			return_values.append(None)
-		elif s[0]=="'"and s[-1]=="'":
-			return_values.append(s[1:-1])
+		elif(s[0]=="'"and s[-1]=="'")or(s[0]=='"'and s[-1]=='"'):
+			return_values.append(eval(s))
 		elif s[0].isdigit():
 			return_values.append(int(s))
 		else:break
 	return return_values
-def check_python_headers(conf):
-	if not conf.env['CC_NAME']and not conf.env['CXX_NAME']:
+@conf
+def test_pyembed(self,mode,msg='Testing pyembed configuration'):
+	self.check(header_name='Python.h',define_name='HAVE_PYEMBED',msg=msg,fragment=FRAG,errmsg='Could not build a python embedded interpreter',features='%s %sprogram pyembed'%(mode,mode))
+@conf
+def test_pyext(self,mode,msg='Testing pyext configuration'):
+	self.check(header_name='Python.h',define_name='HAVE_PYEXT',msg=msg,fragment=FRAG,errmsg='Could not build python extensions',features='%s %sshlib pyext'%(mode,mode))
+@conf
+def python_cross_compile(self,features='pyembed pyext'):
+	features=Utils.to_list(features)
+	if not('PYTHON_LDFLAGS'in self.environ or'PYTHON_PYEXT_LDFLAGS'in self.environ or'PYTHON_PYEMBED_LDFLAGS'in self.environ):
+		return False
+	for x in'PYTHON_VERSION PYTAG pyext_PATTERN'.split():
+		if not x in self.environ:
+			self.fatal('Please set %s in the os environment'%x)
+		else:
+			self.env[x]=self.environ[x]
+	xx=self.env.CXX_NAME and'cxx'or'c'
+	if'pyext'in features:
+		flags=self.environ.get('PYTHON_PYEXT_LDFLAGS',self.environ.get('PYTHON_LDFLAGS'))
+		if flags is None:
+			self.fatal('No flags provided through PYTHON_PYEXT_LDFLAGS as required')
+		else:
+			self.parse_flags(flags,'PYEXT')
+		self.test_pyext(xx)
+	if'pyembed'in features:
+		flags=self.environ.get('PYTHON_PYEMBED_LDFLAGS',self.environ.get('PYTHON_LDFLAGS'))
+		if flags is None:
+			self.fatal('No flags provided through PYTHON_PYEMBED_LDFLAGS as required')
+		else:
+			self.parse_flags(flags,'PYEMBED')
+		self.test_pyembed(xx)
+	return True
+@conf
+def check_python_headers(conf,features='pyembed pyext'):
+	features=Utils.to_list(features)
+	assert('pyembed'in features)or('pyext'in features),"check_python_headers features must include 'pyembed' and/or 'pyext'"
+	env=conf.env
+	if not env.CC_NAME and not env.CXX_NAME:
 		conf.fatal('load a compiler first (gcc, g++, ..)')
-	if not conf.env['PYTHON_VERSION']:
+	if conf.python_cross_compile(features):
+		return
+	if not env.PYTHON_VERSION:
 		conf.check_python_version()
-	env=conf.env
-	pybin=conf.env.PYTHON
+	pybin=env.PYTHON
 	if not pybin:
-		conf.fatal('could not find the python executable')
-	v='prefix SO LDFLAGS LIBDIR LIBPL INCLUDEPY Py_ENABLE_SHARED MACOSX_DEPLOYMENT_TARGET LDSHARED CFLAGS'.split()
+		conf.fatal('Could not find the python executable')
+	v='prefix SO LDFLAGS LIBDIR LIBPL INCLUDEPY Py_ENABLE_SHARED MACOSX_DEPLOYMENT_TARGET LDSHARED CFLAGS LDVERSION'.split()
 	try:
 		lst=conf.get_python_variables(["get_config_var('%s') or ''"%x for x in v])
 	except RuntimeError:
 		conf.fatal("Python development headers not found (-v for details).")
 	vals=['%s = %r'%(x,y)for(x,y)in zip(v,lst)]
-	conf.to_log("Configuration returned from %r:\n%r\n"%(pybin,'\n'.join(vals)))
+	conf.to_log("Configuration returned from %r:\n%s\n"%(pybin,'\n'.join(vals)))
 	dct=dict(zip(v,lst))
 	x='MACOSX_DEPLOYMENT_TARGET'
 	if dct[x]:
-		conf.env[x]=conf.environ[x]=dct[x]
-	env['pyext_PATTERN']='%s'+dct['SO']
+		env[x]=conf.environ[x]=dct[x]
+	env.pyext_PATTERN='%s'+dct['SO']
+	num='.'.join(env.PYTHON_VERSION.split('.')[:2])
+	conf.find_program([''.join(pybin)+'-config','python%s-config'%num,'python-config-%s'%num,'python%sm-config'%num],var='PYTHON_CONFIG',msg="python-config",mandatory=False)
+	if env.PYTHON_CONFIG:
+		all_flags=[['--cflags','--libs','--ldflags']]
+		if sys.hexversion<0x2070000:
+			all_flags=[[k]for k in all_flags[0]]
+		xx=env.CXX_NAME and'cxx'or'c'
+		if'pyembed'in features:
+			for flags in all_flags:
+				conf.check_cfg(msg='Asking python-config for pyembed %r flags'%' '.join(flags),path=env.PYTHON_CONFIG,package='',uselib_store='PYEMBED',args=flags)
+			try:
+				conf.test_pyembed(xx)
+			except conf.errors.ConfigurationError:
+				if dct['Py_ENABLE_SHARED']and dct['LIBDIR']:
+					env.append_unique('LIBPATH_PYEMBED',[dct['LIBDIR']])
+					conf.test_pyembed(xx)
+				else:
+					raise
+		if'pyext'in features:
+			for flags in all_flags:
+				conf.check_cfg(msg='Asking python-config for pyext %r flags'%' '.join(flags),path=env.PYTHON_CONFIG,package='',uselib_store='PYEXT',args=flags)
+			try:
+				conf.test_pyext(xx)
+			except conf.errors.ConfigurationError:
+				if dct['Py_ENABLE_SHARED']and dct['LIBDIR']:
+					env.append_unique('LIBPATH_PYEXT',[dct['LIBDIR']])
+					conf.test_pyext(xx)
+				else:
+					raise
+		conf.define('HAVE_PYTHON_H',1)
+		return
 	all_flags=dct['LDFLAGS']+' '+dct['CFLAGS']
 	conf.parse_flags(all_flags,'PYEMBED')
 	all_flags=dct['LDFLAGS']+' '+dct['LDSHARED']+' '+dct['CFLAGS']
 	conf.parse_flags(all_flags,'PYEXT')
 	result=None
-	for name in('python'+env['PYTHON_VERSION'],'python'+env['PYTHON_VERSION'].replace('.','')):
-		if not result and env['LIBPATH_PYEMBED']:
-			path=env['LIBPATH_PYEMBED']
+	if not dct["LDVERSION"]:
+		dct["LDVERSION"]=env.PYTHON_VERSION
+	for name in('python'+dct['LDVERSION'],'python'+env.PYTHON_VERSION+'m','python'+env.PYTHON_VERSION.replace('.','')):
+		if not result and env.LIBPATH_PYEMBED:
+			path=env.LIBPATH_PYEMBED
 			conf.to_log("\n\n# Trying default LIBPATH_PYEMBED: %r\n"%path)
 			result=conf.check(lib=name,uselib='PYEMBED',libpath=path,mandatory=False,msg='Checking for library %s in LIBPATH_PYEMBED'%name)
 		if not result and dct['LIBDIR']:
@@ -177,33 +260,20 @@
 		if result:
 			break
 	if result:
-		env['LIBPATH_PYEMBED']=path
+		env.LIBPATH_PYEMBED=path
 		env.append_value('LIB_PYEMBED',[name])
 	else:
 		conf.to_log("\n\n### LIB NOT FOUND\n")
-	if(Utils.is_win32 or sys.platform.startswith('os2')or dct['Py_ENABLE_SHARED']):
-		env['LIBPATH_PYEXT']=env['LIBPATH_PYEMBED']
-		env['LIB_PYEXT']=env['LIB_PYEMBED']
-	num='.'.join(env['PYTHON_VERSION'].split('.')[:2])
-	conf.find_program(['python%s-config'%num,'python-config-%s'%num,'python%sm-config'%num],var='PYTHON_CONFIG',mandatory=False)
-	includes=[]
-	if conf.env.PYTHON_CONFIG:
-		for incstr in conf.cmd_and_log([conf.env.PYTHON_CONFIG,'--includes']).strip().split():
-			if(incstr.startswith('-I')or incstr.startswith('/I')):
-				incstr=incstr[2:]
-			if incstr not in includes:
-				includes.append(incstr)
-		conf.to_log("Include path for Python extensions (found via python-config --includes): %r\n"%(includes,))
-		env['INCLUDES_PYEXT']=includes
-		env['INCLUDES_PYEMBED']=includes
-	else:
-		conf.to_log("Include path for Python extensions ""(found via distutils module): %r\n"%(dct['INCLUDEPY'],))
-		env['INCLUDES_PYEXT']=[dct['INCLUDEPY']]
-		env['INCLUDES_PYEMBED']=[dct['INCLUDEPY']]
-	if env['CC_NAME']=='gcc':
+	if Utils.is_win32 or dct['Py_ENABLE_SHARED']:
+		env.LIBPATH_PYEXT=env.LIBPATH_PYEMBED
+		env.LIB_PYEXT=env.LIB_PYEMBED
+	conf.to_log("Include path for Python extensions (found via distutils module): %r\n"%(dct['INCLUDEPY'],))
+	env.INCLUDES_PYEXT=[dct['INCLUDEPY']]
+	env.INCLUDES_PYEMBED=[dct['INCLUDEPY']]
+	if env.CC_NAME=='gcc':
 		env.append_value('CFLAGS_PYEMBED',['-fno-strict-aliasing'])
 		env.append_value('CFLAGS_PYEXT',['-fno-strict-aliasing'])
-	if env['CXX_NAME']=='gcc':
+	if env.CXX_NAME=='gcc':
 		env.append_value('CXXFLAGS_PYEMBED',['-fno-strict-aliasing'])
 		env.append_value('CXXFLAGS_PYEXT',['-fno-strict-aliasing'])
 	if env.CC_NAME=="msvc":
@@ -213,72 +283,74 @@
 		env.append_value('CFLAGS_PYEXT',dist_compiler.compile_options)
 		env.append_value('CXXFLAGS_PYEXT',dist_compiler.compile_options)
 		env.append_value('LINKFLAGS_PYEXT',dist_compiler.ldflags_shared)
-	try:
-		conf.check(header_name='Python.h',define_name='HAVE_PYTHON_H',uselib='PYEMBED',fragment=FRAG,errmsg='Could not find the python development headers')
-	except conf.errors.ConfigurationError:
-		conf.check_cfg(path=conf.env.PYTHON_CONFIG,package='',uselib_store='PYEMBED',args=['--cflags','--libs'])
-		conf.check(header_name='Python.h',define_name='HAVE_PYTHON_H',msg='Getting the python flags from python-config',uselib='PYEMBED',fragment=FRAG,errmsg='Could not find the python development headers elsewhere')
+	conf.check(header_name='Python.h',define_name='HAVE_PYTHON_H',uselib='PYEMBED',fragment=FRAG,errmsg='Distutils not installed? Broken python installation? Get python-config now!')
+@conf
 def check_python_version(conf,minver=None):
 	assert minver is None or isinstance(minver,tuple)
-	pybin=conf.env['PYTHON']
+	pybin=conf.env.PYTHON
 	if not pybin:
 		conf.fatal('could not find the python executable')
 	cmd=pybin+['-c','import sys\nfor x in sys.version_info: print(str(x))']
-	debug('python: Running python command %r'%cmd)
+	Logs.debug('python: Running python command %r',cmd)
 	lines=conf.cmd_and_log(cmd).split()
-	assert len(lines)==5,"found %i lines, expected 5: %r"%(len(lines),lines)
+	assert len(lines)==5,"found %r lines, expected 5: %r"%(len(lines),lines)
 	pyver_tuple=(int(lines[0]),int(lines[1]),int(lines[2]),lines[3],int(lines[4]))
 	result=(minver is None)or(pyver_tuple>=minver)
 	if result:
 		pyver='.'.join([str(x)for x in pyver_tuple[:2]])
-		conf.env['PYTHON_VERSION']=pyver
-		if'PYTHONDIR'in conf.environ:
+		conf.env.PYTHON_VERSION=pyver
+		if'PYTHONDIR'in conf.env:
+			pydir=conf.env.PYTHONDIR
+		elif'PYTHONDIR'in conf.environ:
 			pydir=conf.environ['PYTHONDIR']
 		else:
 			if Utils.is_win32:
-				(python_LIBDEST,pydir)=conf.get_python_variables(["get_config_var('LIBDEST') or ''","get_python_lib(standard_lib=0, prefix=%r) or ''"%conf.env['PREFIX']])
+				(python_LIBDEST,pydir)=conf.get_python_variables(["get_config_var('LIBDEST') or ''","get_python_lib(standard_lib=0) or ''"])
 			else:
 				python_LIBDEST=None
-				(pydir,)=conf.get_python_variables(["get_python_lib(standard_lib=0, prefix=%r) or ''"%conf.env['PREFIX']])
+				(pydir,)=conf.get_python_variables(["get_python_lib(standard_lib=0, prefix=%r) or ''"%conf.env.PREFIX])
 			if python_LIBDEST is None:
-				if conf.env['LIBDIR']:
-					python_LIBDEST=os.path.join(conf.env['LIBDIR'],"python"+pyver)
+				if conf.env.LIBDIR:
+					python_LIBDEST=os.path.join(conf.env.LIBDIR,'python'+pyver)
 				else:
-					python_LIBDEST=os.path.join(conf.env['PREFIX'],"lib","python"+pyver)
-		if'PYTHONARCHDIR'in conf.environ:
+					python_LIBDEST=os.path.join(conf.env.PREFIX,'lib','python'+pyver)
+		if'PYTHONARCHDIR'in conf.env:
+			pyarchdir=conf.env.PYTHONARCHDIR
+		elif'PYTHONARCHDIR'in conf.environ:
 			pyarchdir=conf.environ['PYTHONARCHDIR']
 		else:
-			(pyarchdir,)=conf.get_python_variables(["get_python_lib(plat_specific=1, standard_lib=0, prefix=%r) or ''"%conf.env['PREFIX']])
+			(pyarchdir,)=conf.get_python_variables(["get_python_lib(plat_specific=1, standard_lib=0, prefix=%r) or ''"%conf.env.PREFIX])
 			if not pyarchdir:
 				pyarchdir=pydir
 		if hasattr(conf,'define'):
 			conf.define('PYTHONDIR',pydir)
 			conf.define('PYTHONARCHDIR',pyarchdir)
-		conf.env['PYTHONDIR']=pydir
-		conf.env['PYTHONARCHDIR']=pyarchdir
+		conf.env.PYTHONDIR=pydir
+		conf.env.PYTHONARCHDIR=pyarchdir
 	pyver_full='.'.join(map(str,pyver_tuple[:3]))
 	if minver is None:
 		conf.msg('Checking for python version',pyver_full)
 	else:
 		minver_str='.'.join(map(str,minver))
-		conf.msg('Checking for python version',pyver_tuple,">= %s"%(minver_str,)and'GREEN'or'YELLOW')
+		conf.msg('Checking for python version >= %s'%(minver_str,),pyver_full,color=result and'GREEN'or'YELLOW')
 	if not result:
 		conf.fatal('The python version is too old, expecting %r'%(minver,))
 PYTHON_MODULE_TEMPLATE='''
 import %s as current_module
 version = getattr(current_module, '__version__', None)
 if version is not None:
-    print(str(version))
+	print(str(version))
 else:
-    print('unknown version')
+	print('unknown version')
 '''
+@conf
 def check_python_module(conf,module_name,condition=''):
-	msg='Python module %s'%module_name
+	msg="Checking for python module %r"%module_name
 	if condition:
 		msg='%s (%s)'%(msg,condition)
 	conf.start_msg(msg)
 	try:
-		ret=conf.cmd_and_log(conf.env['PYTHON']+['-c',PYTHON_MODULE_TEMPLATE%module_name])
+		ret=conf.cmd_and_log(conf.env.PYTHON+['-c',PYTHON_MODULE_TEMPLATE%module_name])
 	except Exception:
 		conf.end_msg(False)
 		conf.fatal('Could not find the python module %r'%module_name)
@@ -303,34 +375,24 @@
 		else:
 			conf.end_msg(ret)
 def configure(conf):
-	try:
-		conf.find_program('python',var='PYTHON')
-	except conf.errors.ConfigurationError:
-		warn("could not find a python executable, setting to sys.executable '%s'"%sys.executable)
-		conf.env.PYTHON=sys.executable
-	if conf.env.PYTHON!=sys.executable:
-		warn("python executable '%s' different from sys.executable '%s'"%(conf.env.PYTHON,sys.executable))
-	conf.env.PYTHON=conf.cmd_to_list(conf.env.PYTHON)
 	v=conf.env
-	v['PYCMD']='"import sys, py_compile;py_compile.compile(sys.argv[1], sys.argv[2])"'
-	v['PYFLAGS']=''
-	v['PYFLAGS_OPT']='-O'
-	v['PYC']=getattr(Options.options,'pyc',1)
-	v['PYO']=getattr(Options.options,'pyo',1)
+	if Options.options.pythondir:
+		v.PYTHONDIR=Options.options.pythondir
+	if Options.options.pythonarchdir:
+		v.PYTHONARCHDIR=Options.options.pythonarchdir
+	conf.find_program('python',var='PYTHON',value=Options.options.pythondir or sys.executable)
+	v.PYFLAGS=''
+	v.PYFLAGS_OPT='-O'
+	v.PYC=getattr(Options.options,'pyc',1)
+	v.PYO=getattr(Options.options,'pyo',1)
+	try:
+		v.PYTAG=conf.cmd_and_log(conf.env.PYTHON+['-c',"import imp;print(imp.get_tag())"]).strip()
+	except Errors.WafError:
+		pass
 def options(opt):
-	opt.add_option('--nopyc',action='store_false',default=1,help='Do not install bytecode compiled .pyc files (configuration) [Default:install]',dest='pyc')
-	opt.add_option('--nopyo',action='store_false',default=1,help='Do not install optimised compiled .pyo files (configuration) [Default:install]',dest='pyo')
-
-extension('.py')(process_py)
-feature('py')(feature_py)
-feature('pyext')(init_pyext)
-before_method('propagate_uselib_vars','apply_link')(init_pyext)
-after_method('apply_bundle')(init_pyext)
-feature('pyext')(set_bundle)
-before_method('apply_link','apply_bundle')(set_bundle)
-before_method('propagate_uselib_vars')(init_pyembed)
-feature('pyembed')(init_pyembed)
-conf(get_python_variables)
-conf(check_python_headers)
-conf(check_python_version)
-conf(check_python_module)
\ No newline at end of file
+	pyopt=opt.add_option_group("Python Options")
+	pyopt.add_option('--nopyc',dest='pyc',action='store_false',default=1,help='Do not install bytecode compiled .pyc files (configuration) [Default:install]')
+	pyopt.add_option('--nopyo',dest='pyo',action='store_false',default=1,help='Do not install optimised compiled .pyo files (configuration) [Default:install]')
+	pyopt.add_option('--python',dest="python",help='python binary to be used [Default: %s]'%sys.executable)
+	pyopt.add_option('--pythondir',dest='pythondir',help='Installation path for python modules (py, platform-independent .py and .pyc files)')
+	pyopt.add_option('--pythonarchdir',dest='pythonarchdir',help='Installation path for python extension (pyext, platform-dependent .so or .dylib files)')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/qt4.py glmark2-2021.02/waflib/Tools/qt4.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/qt4.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/qt4.py	1970-01-01 08:00:00.000000000 +0800
@@ -1,434 +0,0 @@
-#! /usr/bin/env python
-# encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
-
-import sys
-if sys.hexversion < 0x020400f0: from sets import Set as set
-try:
-	from xml.sax import make_parser
-	from xml.sax.handler import ContentHandler
-except ImportError:
-	has_xml=False
-	ContentHandler=object
-else:
-	has_xml=True
-import os,sys
-from waflib.Tools import c_preproc,cxx
-from waflib import Task,Utils,Options,Errors
-from waflib.TaskGen import feature,after_method,extension
-from waflib.Configure import conf
-from waflib import Logs
-MOC_H=['.h','.hpp','.hxx','.hh']
-EXT_RCC=['.qrc']
-EXT_UI=['.ui']
-EXT_QT4=['.cpp','.cc','.cxx','.C']
-QT4_LIBS="QtCore QtGui QtUiTools QtNetwork QtOpenGL QtSql QtSvg QtTest QtXml QtXmlPatterns QtWebKit Qt3Support QtHelp QtScript QtDeclarative"
-class qxx(cxx.cxx):
-	def __init__(self,*k,**kw):
-		Task.Task.__init__(self,*k,**kw)
-		self.moc_done=0
-	def scan(self):
-		(nodes,names)=c_preproc.scan(self)
-		for x in nodes:
-			if x.name.endswith('.moc'):
-				nodes.remove(x)
-				names.append(x.path_from(self.inputs[0].parent.get_bld()))
-		return(nodes,names)
-	def runnable_status(self):
-		if self.moc_done:
-			return Task.Task.runnable_status(self)
-		else:
-			for t in self.run_after:
-				if not t.hasrun:
-					return Task.ASK_LATER
-			self.add_moc_tasks()
-			return Task.Task.runnable_status(self)
-	def add_moc_tasks(self):
-		node=self.inputs[0]
-		bld=self.generator.bld
-		try:
-			self.signature()
-		except KeyError:
-			pass
-		else:
-			delattr(self,'cache_sig')
-		moctasks=[]
-		mocfiles=[]
-		try:
-			tmp_lst=bld.raw_deps[self.uid()]
-			bld.raw_deps[self.uid()]=[]
-		except KeyError:
-			tmp_lst=[]
-		for d in tmp_lst:
-			if not d.endswith('.moc'):
-				continue
-			if d in mocfiles:
-				Logs.error("paranoia owns")
-				continue
-			mocfiles.append(d)
-			h_node=None
-			try:ext=Options.options.qt_header_ext.split()
-			except AttributeError:pass
-			if not ext:ext=MOC_H
-			base2=d[:-4]
-			for x in[node.parent]+self.generator.includes_nodes:
-				for e in ext:
-					h_node=x.find_node(base2+e)
-					if h_node:
-						break
-				if h_node:
-					m_node=h_node.change_ext('.moc')
-					break
-			else:
-				for k in EXT_QT4:
-					if base2.endswith(k):
-						for x in[node.parent]+self.generator.includes_nodes:
-							h_node=x.find_node(base2)
-							if h_node:
-								break
-					if h_node:
-						m_node=h_node.change_ext(k+'.moc')
-						break
-			if not h_node:
-				raise Errors.WafError('no header found for %r which is a moc file'%d)
-			bld.node_deps[(self.inputs[0].parent.abspath(),m_node.name)]=h_node
-			task=Task.classes['moc'](env=self.env,generator=self.generator)
-			task.set_inputs(h_node)
-			task.set_outputs(m_node)
-			gen=bld.producer
-			gen.outstanding.insert(0,task)
-			gen.total+=1
-			moctasks.append(task)
-		tmp_lst=bld.raw_deps[self.uid()]=mocfiles
-		lst=bld.node_deps.get(self.uid(),())
-		for d in lst:
-			name=d.name
-			if name.endswith('.moc'):
-				task=Task.classes['moc'](env=self.env,generator=self.generator)
-				task.set_inputs(bld.node_deps[(self.inputs[0].parent.abspath(),name)])
-				task.set_outputs(d)
-				gen=bld.producer
-				gen.outstanding.insert(0,task)
-				gen.total+=1
-				moctasks.append(task)
-		self.run_after.update(set(moctasks))
-		self.moc_done=1
-	run=Task.classes['cxx'].__dict__['run']
-class trans_update(Task.Task):
-	run_str='${QT_LUPDATE} ${SRC} -ts ${TGT}'
-	color='BLUE'
-Task.update_outputs(trans_update)
-class XMLHandler(ContentHandler):
-	def __init__(self):
-		self.buf=[]
-		self.files=[]
-	def startElement(self,name,attrs):
-		if name=='file':
-			self.buf=[]
-	def endElement(self,name):
-		if name=='file':
-			self.files.append(str(''.join(self.buf)))
-	def characters(self,cars):
-		self.buf.append(cars)
-def create_rcc_task(self,node):
-	rcnode=node.change_ext('_rc.cpp')
-	rcctask=self.create_task('rcc',node,rcnode)
-	cpptask=self.create_task('cxx',rcnode,rcnode.change_ext('.o'))
-	try:
-		self.compiled_tasks.append(cpptask)
-	except AttributeError:
-		self.compiled_tasks=[cpptask]
-	return cpptask
-def create_uic_task(self,node):
-	uictask=self.create_task('ui4',node)
-	uictask.outputs=[self.path.find_or_declare(self.env['ui_PATTERN']%node.name[:-3])]
-def add_lang(self,node):
-	self.lang=self.to_list(getattr(self,'lang',[]))+[node]
-def apply_qt4(self):
-	if getattr(self,'lang',None):
-		qmtasks=[]
-		for x in self.to_list(self.lang):
-			if isinstance(x,str):
-				x=self.path.find_resource(x+'.ts')
-			qmtasks.append(self.create_task('ts2qm',x,x.change_ext('.qm')))
-		if getattr(self,'update',None)and Options.options.trans_qt4:
-			cxxnodes=[a.inputs[0]for a in self.compiled_tasks]+[a.inputs[0]for a in self.tasks if getattr(a,'inputs',None)and a.inputs[0].name.endswith('.ui')]
-			for x in qmtasks:
-				self.create_task('trans_update',cxxnodes,x.inputs)
-		if getattr(self,'langname',None):
-			qmnodes=[x.outputs[0]for x in qmtasks]
-			rcnode=self.langname
-			if isinstance(rcnode,str):
-				rcnode=self.path.find_or_declare(rcnode+'.qrc')
-			t=self.create_task('qm2rcc',qmnodes,rcnode)
-			k=create_rcc_task(self,t.outputs[0])
-			self.link_task.inputs.append(k.outputs[0])
-	lst=[]
-	for flag in self.to_list(self.env['CXXFLAGS']):
-		if len(flag)<2:continue
-		f=flag[0:2]
-		if f in['-D','-I','/D','/I']:
-			if(f[0]=='/'):
-				lst.append('-'+flag[1:])
-			else:
-				lst.append(flag)
-	self.env['MOC_FLAGS']=lst
-def cxx_hook(self,node):
-	return self.create_compiled_task('qxx',node)
-class rcc(Task.Task):
-	color='BLUE'
-	run_str='${QT_RCC} -name ${SRC[0].name} ${SRC[0].abspath()} ${RCC_ST} -o ${TGT}'
-	ext_out=['.h']
-	def scan(self):
-		node=self.inputs[0]
-		if not has_xml:
-			Logs.error('no xml support was found, the rcc dependencies will be incomplete!')
-			return([],[])
-		parser=make_parser()
-		curHandler=XMLHandler()
-		parser.setContentHandler(curHandler)
-		fi=open(self.inputs[0].abspath())
-		parser.parse(fi)
-		fi.close()
-		nodes=[]
-		names=[]
-		root=self.inputs[0].parent
-		for x in curHandler.files:
-			nd=root.find_resource(x)
-			if nd:nodes.append(nd)
-			else:names.append(x)
-		return(nodes,names)
-class moc(Task.Task):
-	color='BLUE'
-	run_str='${QT_MOC} ${MOC_FLAGS} ${MOCCPPPATH_ST:INCPATHS} ${MOCDEFINES_ST:DEFINES} ${SRC} ${MOC_ST} ${TGT}'
-class ui4(Task.Task):
-	color='BLUE'
-	run_str='${QT_UIC} ${SRC} -o ${TGT}'
-	ext_out=['.h']
-class ts2qm(Task.Task):
-	color='BLUE'
-	run_str='${QT_LRELEASE} ${QT_LRELEASE_FLAGS} ${SRC} -qm ${TGT}'
-class qm2rcc(Task.Task):
-	color='BLUE'
-	after='ts2qm'
-	def run(self):
-		txt='\n'.join(['<file>%s</file>'%k.path_from(self.outputs[0].parent)for k in self.inputs])
-		code='<!DOCTYPE RCC><RCC version="1.0">\n<qresource>\n%s\n</qresource>\n</RCC>'%txt
-		self.outputs[0].write(code)
-def configure(self):
-	self.find_qt4_binaries()
-	self.set_qt4_libs_to_check()
-	self.find_qt4_libraries()
-	self.add_qt4_rpath()
-	self.simplify_qt4_libs()
-def find_qt4_binaries(self):
-	env=self.env
-	opt=Options.options
-	qtdir=getattr(opt,'qtdir','')
-	qtbin=getattr(opt,'qtbin','')
-	paths=[]
-	if qtdir:
-		qtbin=os.path.join(qtdir,'bin')
-	if not qtdir:
-		qtdir=self.environ.get('QT4_ROOT','')
-		qtbin=os.path.join(qtdir,'bin')
-	if qtbin:
-		paths=[qtbin]
-	if not qtdir:
-		paths=os.environ.get('PATH','').split(os.pathsep)
-		paths.append('/usr/share/qt4/bin/')
-		try:
-			lst=Utils.listdir('/usr/local/Trolltech/')
-		except OSError:
-			pass
-		else:
-			if lst:
-				lst.sort()
-				lst.reverse()
-				qtdir='/usr/local/Trolltech/%s/'%lst[0]
-				qtbin=os.path.join(qtdir,'bin')
-				paths.append(qtbin)
-	cand=None
-	prev_ver=['4','0','0']
-	for qmk in['qmake-qt4','qmake4','qmake']:
-		try:
-			qmake=self.find_program(qmk,path_list=paths)
-		except self.errors.ConfigurationError:
-			pass
-		else:
-			try:
-				version=self.cmd_and_log([qmake,'-query','QT_VERSION']).strip()
-			except self.errors.ConfigurationError:
-				pass
-			else:
-				if version:
-					new_ver=version.split('.')
-					if new_ver>prev_ver:
-						cand=qmake
-						prev_ver=new_ver
-	if cand:
-		self.env.QMAKE=cand
-	else:
-		self.fatal('Could not find qmake for qt4')
-	qtbin=self.cmd_and_log([self.env.QMAKE,'-query','QT_INSTALL_BINS']).strip()+os.sep
-	def find_bin(lst,var):
-		for f in lst:
-			try:
-				ret=self.find_program(f,path_list=paths)
-			except self.errors.ConfigurationError:
-				pass
-			else:
-				env[var]=ret
-				break
-	find_bin(['uic-qt3','uic3'],'QT_UIC3')
-	find_bin(['uic-qt4','uic'],'QT_UIC')
-	if not env['QT_UIC']:
-		self.fatal('cannot find the uic compiler for qt4')
-	try:
-		uicver=self.cmd_and_log(env['QT_UIC']+" -version 2>&1").strip()
-	except self.errors.ConfigurationError:
-		self.fatal('this uic compiler is for qt3, add uic for qt4 to your path')
-	uicver=uicver.replace('Qt User Interface Compiler ','').replace('User Interface Compiler for Qt','')
-	self.msg('Checking for uic version','%s'%uicver)
-	if uicver.find(' 3.')!=-1:
-		self.fatal('this uic compiler is for qt3, add uic for qt4 to your path')
-	find_bin(['moc-qt4','moc'],'QT_MOC')
-	find_bin(['rcc'],'QT_RCC')
-	find_bin(['lrelease-qt4','lrelease'],'QT_LRELEASE')
-	find_bin(['lupdate-qt4','lupdate'],'QT_LUPDATE')
-	env['UIC3_ST']='%s -o %s'
-	env['UIC_ST']='%s -o %s'
-	env['MOC_ST']='-o'
-	env['ui_PATTERN']='ui_%s.h'
-	env['QT_LRELEASE_FLAGS']=['-silent']
-	env.MOCCPPPATH_ST='-I%s'
-	env.MOCDEFINES_ST='-D%s'
-def find_qt4_libraries(self):
-	qtlibs=getattr(Options.options,'qtlibs','')
-	if not qtlibs:
-		try:
-			qtlibs=self.cmd_and_log([self.env.QMAKE,'-query','QT_INSTALL_LIBS']).strip()
-		except Errors.WafError:
-			qtdir=self.cmd_and_log([self.env.QMAKE,'-query','QT_INSTALL_PREFIX']).strip()+os.sep
-			qtlibs=os.path.join(qtdir,'lib')
-	self.msg('Found the Qt4 libraries in',qtlibs)
-	qtincludes=self.cmd_and_log([self.env.QMAKE,'-query','QT_INSTALL_HEADERS']).strip()
-	env=self.env
-	if not'PKG_CONFIG_PATH'in os.environ:
-		os.environ['PKG_CONFIG_PATH']='%s:%s/pkgconfig:/usr/lib/qt4/lib/pkgconfig:/opt/qt4/lib/pkgconfig:/usr/lib/qt4/lib:/opt/qt4/lib'%(qtlibs,qtlibs)
-	try:
-		self.check_cfg(atleast_pkgconfig_version='0.1')
-	except self.errors.ConfigurationError:
-		for i in self.qt4_vars:
-			uselib=i.upper()
-			if Utils.unversioned_sys_platform()=="darwin":
-				frameworkName=i+".framework"
-				qtDynamicLib=os.path.join(qtlibs,frameworkName,i)
-				if os.path.exists(qtDynamicLib):
-					env.append_unique('FRAMEWORK_'+uselib,i)
-					self.msg('Checking for %s'%i,qtDynamicLib,'GREEN')
-				else:
-					self.msg('Checking for %s'%i,False,'YELLOW')
-				env.append_unique('INCLUDES_'+uselib,os.path.join(qtlibs,frameworkName,'Headers'))
-			elif sys.platform!="win32":
-				qtDynamicLib=os.path.join(qtlibs,"lib"+i+".so")
-				qtStaticLib=os.path.join(qtlibs,"lib"+i+".a")
-				if os.path.exists(qtDynamicLib):
-					env.append_unique('LIB_'+uselib,i)
-					self.msg('Checking for %s'%i,qtDynamicLib,'GREEN')
-				elif os.path.exists(qtStaticLib):
-					env.append_unique('LIB_'+uselib,i)
-					self.msg('Checking for %s'%i,qtStaticLib,'GREEN')
-				else:
-					self.msg('Checking for %s'%i,False,'YELLOW')
-				env.append_unique('LIBPATH_'+uselib,qtlibs)
-				env.append_unique('INCLUDES_'+uselib,qtincludes)
-				env.append_unique('INCLUDES_'+uselib,os.path.join(qtincludes,i))
-			else:
-				for k in("lib%s.a","lib%s4.a","%s.lib","%s4.lib"):
-					lib=os.path.join(qtlibs,k%i)
-					if os.path.exists(lib):
-						env.append_unique('LIB_'+uselib,i+k[k.find("%s")+2:k.find('.')])
-						self.msg('Checking for %s'%i,lib,'GREEN')
-						break
-				else:
-					self.msg('Checking for %s'%i,False,'YELLOW')
-				env.append_unique('LIBPATH_'+uselib,qtlibs)
-				env.append_unique('INCLUDES_'+uselib,qtincludes)
-				env.append_unique('INCLUDES_'+uselib,os.path.join(qtincludes,i))
-				uselib=i.upper()+"_debug"
-				for k in("lib%sd.a","lib%sd4.a","%sd.lib","%sd4.lib"):
-					lib=os.path.join(qtlibs,k%i)
-					if os.path.exists(lib):
-						env.append_unique('LIB_'+uselib,i+k[k.find("%s")+2:k.find('.')])
-						self.msg('Checking for %s'%i,lib,'GREEN')
-						break
-				else:
-					self.msg('Checking for %s'%i,False,'YELLOW')
-				env.append_unique('LIBPATH_'+uselib,qtlibs)
-				env.append_unique('INCLUDES_'+uselib,qtincludes)
-				env.append_unique('INCLUDES_'+uselib,os.path.join(qtincludes,i))
-	else:
-		for i in self.qt4_vars_debug+self.qt4_vars:
-			self.check_cfg(package=i,args='--cflags --libs',mandatory=False)
-def simplify_qt4_libs(self):
-	env=self.env
-	def process_lib(vars_,coreval):
-		for d in vars_:
-			var=d.upper()
-			if var=='QTCORE':
-				continue
-			value=env['LIBPATH_'+var]
-			if value:
-				core=env[coreval]
-				accu=[]
-				for lib in value:
-					if lib in core:
-						continue
-					accu.append(lib)
-				env['LIBPATH_'+var]=accu
-	process_lib(self.qt4_vars,'LIBPATH_QTCORE')
-	process_lib(self.qt4_vars_debug,'LIBPATH_QTCORE_DEBUG')
-def add_qt4_rpath(self):
-	env=self.env
-	if Options.options.want_rpath:
-		def process_rpath(vars_,coreval):
-			for d in vars_:
-				var=d.upper()
-				value=env['LIBPATH_'+var]
-				if value:
-					core=env[coreval]
-					accu=[]
-					for lib in value:
-						if var!='QTCORE':
-							if lib in core:
-								continue
-						accu.append('-Wl,--rpath='+lib)
-					env['RPATH_'+var]=accu
-		process_rpath(self.qt4_vars,'LIBPATH_QTCORE')
-		process_rpath(self.qt4_vars_debug,'LIBPATH_QTCORE_DEBUG')
-def set_qt4_libs_to_check(self):
-	if not hasattr(self,'qt4_vars'):
-		self.qt4_vars=QT4_LIBS
-	self.qt4_vars=Utils.to_list(self.qt4_vars)
-	if not hasattr(self,'qt4_vars_debug'):
-		self.qt4_vars_debug=[a+'_debug'for a in self.qt4_vars]
-	self.qt4_vars_debug=Utils.to_list(self.qt4_vars_debug)
-def options(opt):
-	opt.add_option('--want-rpath',action='store_true',default=False,dest='want_rpath',help='enable the rpath for qt libraries')
-	opt.add_option('--header-ext',type='string',default='',help='header extension for moc files',dest='qt_header_ext')
-	for i in'qtdir qtbin qtlibs'.split():
-		opt.add_option('--'+i,type='string',default='',dest=i)
-	opt.add_option('--translate',action="store_true",help="collect translation strings",dest="trans_qt4",default=False)
-
-extension(*EXT_RCC)(create_rcc_task)
-extension(*EXT_UI)(create_uic_task)
-extension('.ts')(add_lang)
-feature('qt4')(apply_qt4)
-after_method('apply_link')(apply_qt4)
-extension(*EXT_QT4)(cxx_hook)
-conf(find_qt4_binaries)
-conf(find_qt4_libraries)
-conf(simplify_qt4_libs)
-conf(add_qt4_rpath)
-conf(set_qt4_libs_to_check)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/qt5.py glmark2-2021.02/waflib/Tools/qt5.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/qt5.py	1970-01-01 08:00:00.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/qt5.py	2021-04-25 01:47:22.000000000 +0800
@@ -0,0 +1,506 @@
+#! /usr/bin/env python
+# encoding: utf-8
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
+
+try:
+	from xml.sax import make_parser
+	from xml.sax.handler import ContentHandler
+except ImportError:
+	has_xml=False
+	ContentHandler=object
+else:
+	has_xml=True
+import os,sys
+from waflib.Tools import cxx
+from waflib import Task,Utils,Options,Errors,Context
+from waflib.TaskGen import feature,after_method,extension,before_method
+from waflib.Configure import conf
+from waflib import Logs
+MOC_H=['.h','.hpp','.hxx','.hh']
+EXT_RCC=['.qrc']
+EXT_UI=['.ui']
+EXT_QT5=['.cpp','.cc','.cxx','.C']
+QT5_LIBS='''
+qtmain
+Qt5Bluetooth
+Qt5CLucene
+Qt5Concurrent
+Qt5Core
+Qt5DBus
+Qt5Declarative
+Qt5DesignerComponents
+Qt5Designer
+Qt5Gui
+Qt5Help
+Qt5MultimediaQuick_p
+Qt5Multimedia
+Qt5MultimediaWidgets
+Qt5Network
+Qt5Nfc
+Qt5OpenGL
+Qt5Positioning
+Qt5PrintSupport
+Qt5Qml
+Qt5QuickParticles
+Qt5Quick
+Qt5QuickTest
+Qt5Script
+Qt5ScriptTools
+Qt5Sensors
+Qt5SerialPort
+Qt5Sql
+Qt5Svg
+Qt5Test
+Qt5WebKit
+Qt5WebKitWidgets
+Qt5Widgets
+Qt5WinExtras
+Qt5X11Extras
+Qt5XmlPatterns
+Qt5Xml'''
+class qxx(Task.classes['cxx']):
+	def __init__(self,*k,**kw):
+		Task.Task.__init__(self,*k,**kw)
+		self.moc_done=0
+	def runnable_status(self):
+		if self.moc_done:
+			return Task.Task.runnable_status(self)
+		else:
+			for t in self.run_after:
+				if not t.hasrun:
+					return Task.ASK_LATER
+			self.add_moc_tasks()
+			return Task.Task.runnable_status(self)
+	def create_moc_task(self,h_node,m_node):
+		try:
+			moc_cache=self.generator.bld.moc_cache
+		except AttributeError:
+			moc_cache=self.generator.bld.moc_cache={}
+		try:
+			return moc_cache[h_node]
+		except KeyError:
+			tsk=moc_cache[h_node]=Task.classes['moc'](env=self.env,generator=self.generator)
+			tsk.set_inputs(h_node)
+			tsk.set_outputs(m_node)
+			tsk.env.append_unique('MOC_FLAGS','-i')
+			if self.generator:
+				self.generator.tasks.append(tsk)
+			gen=self.generator.bld.producer
+			gen.outstanding.appendleft(tsk)
+			gen.total+=1
+			return tsk
+		else:
+			delattr(self,'cache_sig')
+	def add_moc_tasks(self):
+		node=self.inputs[0]
+		bld=self.generator.bld
+		try:
+			self.signature()
+		except KeyError:
+			pass
+		else:
+			delattr(self,'cache_sig')
+		include_nodes=[node.parent]+self.generator.includes_nodes
+		moctasks=[]
+		mocfiles=set([])
+		for d in bld.raw_deps.get(self.uid(),[]):
+			if not d.endswith('.moc'):
+				continue
+			if d in mocfiles:
+				continue
+			mocfiles.add(d)
+			h_node=None
+			base2=d[:-4]
+			prefix=node.name[:node.name.rfind('.')]
+			if base2==prefix:
+				h_node=node
+			else:
+				for x in include_nodes:
+					for e in MOC_H:
+						h_node=x.find_node(base2+e)
+						if h_node:
+							break
+					else:
+						continue
+					break
+			if h_node:
+				m_node=h_node.change_ext('.moc')
+			else:
+				raise Errors.WafError('No source found for %r which is a moc file'%d)
+			task=self.create_moc_task(h_node,m_node)
+			moctasks.append(task)
+		self.run_after.update(set(moctasks))
+		self.moc_done=1
+class trans_update(Task.Task):
+	run_str='${QT_LUPDATE} ${SRC} -ts ${TGT}'
+	color='BLUE'
+class XMLHandler(ContentHandler):
+	def __init__(self):
+		self.buf=[]
+		self.files=[]
+	def startElement(self,name,attrs):
+		if name=='file':
+			self.buf=[]
+	def endElement(self,name):
+		if name=='file':
+			self.files.append(str(''.join(self.buf)))
+	def characters(self,cars):
+		self.buf.append(cars)
+@extension(*EXT_RCC)
+def create_rcc_task(self,node):
+	rcnode=node.change_ext('_rc.cpp')
+	self.create_task('rcc',node,rcnode)
+	cpptask=self.create_task('cxx',rcnode,rcnode.change_ext('.o'))
+	try:
+		self.compiled_tasks.append(cpptask)
+	except AttributeError:
+		self.compiled_tasks=[cpptask]
+	return cpptask
+@extension(*EXT_UI)
+def create_uic_task(self,node):
+	uictask=self.create_task('ui5',node)
+	uictask.outputs=[node.parent.find_or_declare(self.env.ui_PATTERN%node.name[:-3])]
+@extension('.ts')
+def add_lang(self,node):
+	self.lang=self.to_list(getattr(self,'lang',[]))+[node]
+@feature('qt5')
+@before_method('process_source')
+def process_mocs(self):
+	lst=self.to_nodes(getattr(self,'moc',[]))
+	self.source=self.to_list(getattr(self,'source',[]))
+	for x in lst:
+		prefix=x.name[:x.name.rfind('.')]
+		moc_target='moc_%s.%d.cpp'%(prefix,self.idx)
+		moc_node=x.parent.find_or_declare(moc_target)
+		self.source.append(moc_node)
+		self.create_task('moc',x,moc_node)
+@feature('qt5')
+@after_method('apply_link')
+def apply_qt5(self):
+	if getattr(self,'lang',None):
+		qmtasks=[]
+		for x in self.to_list(self.lang):
+			if isinstance(x,str):
+				x=self.path.find_resource(x+'.ts')
+			qmtasks.append(self.create_task('ts2qm',x,x.change_ext('.qm')))
+		if getattr(self,'update',None)and Options.options.trans_qt5:
+			cxxnodes=[a.inputs[0]for a in self.compiled_tasks]+[a.inputs[0]for a in self.tasks if getattr(a,'inputs',None)and a.inputs[0].name.endswith('.ui')]
+			for x in qmtasks:
+				self.create_task('trans_update',cxxnodes,x.inputs)
+		if getattr(self,'langname',None):
+			qmnodes=[x.outputs[0]for x in qmtasks]
+			rcnode=self.langname
+			if isinstance(rcnode,str):
+				rcnode=self.path.find_or_declare(rcnode+'.qrc')
+			t=self.create_task('qm2rcc',qmnodes,rcnode)
+			k=create_rcc_task(self,t.outputs[0])
+			self.link_task.inputs.append(k.outputs[0])
+	lst=[]
+	for flag in self.to_list(self.env.CXXFLAGS):
+		if len(flag)<2:continue
+		f=flag[0:2]
+		if f in('-D','-I','/D','/I'):
+			if(f[0]=='/'):
+				lst.append('-'+flag[1:])
+			else:
+				lst.append(flag)
+	self.env.append_value('MOC_FLAGS',lst)
+@extension(*EXT_QT5)
+def cxx_hook(self,node):
+	return self.create_compiled_task('qxx',node)
+class rcc(Task.Task):
+	color='BLUE'
+	run_str='${QT_RCC} -name ${tsk.rcname()} ${SRC[0].abspath()} ${RCC_ST} -o ${TGT}'
+	ext_out=['.h']
+	def rcname(self):
+		return os.path.splitext(self.inputs[0].name)[0]
+	def scan(self):
+		if not has_xml:
+			Logs.error('No xml.sax support was found, rcc dependencies will be incomplete!')
+			return([],[])
+		parser=make_parser()
+		curHandler=XMLHandler()
+		parser.setContentHandler(curHandler)
+		fi=open(self.inputs[0].abspath(),'r')
+		try:
+			parser.parse(fi)
+		finally:
+			fi.close()
+		nodes=[]
+		names=[]
+		root=self.inputs[0].parent
+		for x in curHandler.files:
+			nd=root.find_resource(x)
+			if nd:nodes.append(nd)
+			else:names.append(x)
+		return(nodes,names)
+class moc(Task.Task):
+	color='BLUE'
+	run_str='${QT_MOC} ${MOC_FLAGS} ${MOCCPPPATH_ST:INCPATHS} ${MOCDEFINES_ST:DEFINES} ${SRC} ${MOC_ST} ${TGT}'
+class ui5(Task.Task):
+	color='BLUE'
+	run_str='${QT_UIC} ${SRC} -o ${TGT}'
+	ext_out=['.h']
+class ts2qm(Task.Task):
+	color='BLUE'
+	run_str='${QT_LRELEASE} ${QT_LRELEASE_FLAGS} ${SRC} -qm ${TGT}'
+class qm2rcc(Task.Task):
+	color='BLUE'
+	after='ts2qm'
+	def run(self):
+		txt='\n'.join(['<file>%s</file>'%k.path_from(self.outputs[0].parent)for k in self.inputs])
+		code='<!DOCTYPE RCC><RCC version="1.0">\n<qresource>\n%s\n</qresource>\n</RCC>'%txt
+		self.outputs[0].write(code)
+def configure(self):
+	self.find_qt5_binaries()
+	self.set_qt5_libs_to_check()
+	self.set_qt5_defines()
+	self.find_qt5_libraries()
+	self.add_qt5_rpath()
+	self.simplify_qt5_libs()
+	if not has_xml:
+		Logs.error('No xml.sax support was found, rcc dependencies will be incomplete!')
+	frag='#include <QApplication>\nint main(int argc, char **argv) {return 0;}\n'
+	uses='QT5CORE QT5WIDGETS QT5GUI'
+	for flag in[[],'-fPIE','-fPIC','-std=c++11',['-std=c++11','-fPIE'],['-std=c++11','-fPIC']]:
+		msg='See if Qt files compile '
+		if flag:
+			msg+='with %s'%flag
+		try:
+			self.check(features='qt5 cxx',use=uses,uselib_store='qt5',cxxflags=flag,fragment=frag,msg=msg)
+		except self.errors.ConfigurationError:
+			pass
+		else:
+			break
+	else:
+		self.fatal('Could not build a simple Qt application')
+	from waflib import Utils
+	if Utils.unversioned_sys_platform()=='freebsd':
+		frag='#include <QApplication>\nint main(int argc, char **argv) { QApplication app(argc, argv); return NULL != (void*) (&app);}\n'
+		try:
+			self.check(features='qt5 cxx cxxprogram',use=uses,fragment=frag,msg='Can we link Qt programs on FreeBSD directly?')
+		except self.errors.ConfigurationError:
+			self.check(features='qt5 cxx cxxprogram',use=uses,uselib_store='qt5',libpath='/usr/local/lib',fragment=frag,msg='Is /usr/local/lib required?')
+@conf
+def find_qt5_binaries(self):
+	env=self.env
+	opt=Options.options
+	qtdir=getattr(opt,'qtdir','')
+	qtbin=getattr(opt,'qtbin','')
+	paths=[]
+	if qtdir:
+		qtbin=os.path.join(qtdir,'bin')
+	if not qtdir:
+		qtdir=self.environ.get('QT5_ROOT','')
+		qtbin=self.environ.get('QT5_BIN')or os.path.join(qtdir,'bin')
+	if qtbin:
+		paths=[qtbin]
+	if not qtdir:
+		paths=self.environ.get('PATH','').split(os.pathsep)
+		paths.extend(['/usr/share/qt5/bin','/usr/local/lib/qt5/bin'])
+		try:
+			lst=Utils.listdir('/usr/local/Trolltech/')
+		except OSError:
+			pass
+		else:
+			if lst:
+				lst.sort()
+				lst.reverse()
+				qtdir='/usr/local/Trolltech/%s/'%lst[0]
+				qtbin=os.path.join(qtdir,'bin')
+				paths.append(qtbin)
+	cand=None
+	prev_ver=['5','0','0']
+	for qmk in('qmake-qt5','qmake5','qmake'):
+		try:
+			qmake=self.find_program(qmk,path_list=paths)
+		except self.errors.ConfigurationError:
+			pass
+		else:
+			try:
+				version=self.cmd_and_log(qmake+['-query','QT_VERSION']).strip()
+			except self.errors.WafError:
+				pass
+			else:
+				if version:
+					new_ver=version.split('.')
+					if new_ver>prev_ver:
+						cand=qmake
+						prev_ver=new_ver
+	if not cand:
+		try:
+			self.find_program('qtchooser')
+		except self.errors.ConfigurationError:
+			pass
+		else:
+			cmd=self.env.QTCHOOSER+['-qt=5','-run-tool=qmake']
+			try:
+				version=self.cmd_and_log(cmd+['-query','QT_VERSION'])
+			except self.errors.WafError:
+				pass
+			else:
+				cand=cmd
+	if cand:
+		self.env.QMAKE=cand
+	else:
+		self.fatal('Could not find qmake for qt5')
+	self.env.QT_HOST_BINS=qtbin=self.cmd_and_log(self.env.QMAKE+['-query','QT_HOST_BINS']).strip()
+	paths.insert(0,qtbin)
+	def find_bin(lst,var):
+		if var in env:
+			return
+		for f in lst:
+			try:
+				ret=self.find_program(f,path_list=paths)
+			except self.errors.ConfigurationError:
+				pass
+			else:
+				env[var]=ret
+				break
+	find_bin(['uic-qt5','uic'],'QT_UIC')
+	if not env.QT_UIC:
+		self.fatal('cannot find the uic compiler for qt5')
+	self.start_msg('Checking for uic version')
+	uicver=self.cmd_and_log(env.QT_UIC+['-version'],output=Context.BOTH)
+	uicver=''.join(uicver).strip()
+	uicver=uicver.replace('Qt User Interface Compiler ','').replace('User Interface Compiler for Qt','')
+	self.end_msg(uicver)
+	if uicver.find(' 3.')!=-1 or uicver.find(' 4.')!=-1:
+		self.fatal('this uic compiler is for qt3 or qt5, add uic for qt5 to your path')
+	find_bin(['moc-qt5','moc'],'QT_MOC')
+	find_bin(['rcc-qt5','rcc'],'QT_RCC')
+	find_bin(['lrelease-qt5','lrelease'],'QT_LRELEASE')
+	find_bin(['lupdate-qt5','lupdate'],'QT_LUPDATE')
+	env.UIC_ST='%s -o %s'
+	env.MOC_ST='-o'
+	env.ui_PATTERN='ui_%s.h'
+	env.QT_LRELEASE_FLAGS=['-silent']
+	env.MOCCPPPATH_ST='-I%s'
+	env.MOCDEFINES_ST='-D%s'
+@conf
+def find_single_qt5_lib(self,name,uselib,qtlibs,qtincludes,force_static):
+	env=self.env
+	if force_static:
+		exts=('.a','.lib')
+		prefix='STLIB'
+	else:
+		exts=('.so','.lib')
+		prefix='LIB'
+	def lib_names():
+		for x in exts:
+			for k in('','5')if Utils.is_win32 else['']:
+				for p in('lib',''):
+					yield(p,name,k,x)
+	for tup in lib_names():
+		k=''.join(tup)
+		path=os.path.join(qtlibs,k)
+		if os.path.exists(path):
+			if env.DEST_OS=='win32':
+				libval=''.join(tup[:-1])
+			else:
+				libval=name
+			env.append_unique(prefix+'_'+uselib,libval)
+			env.append_unique('%sPATH_%s'%(prefix,uselib),qtlibs)
+			env.append_unique('INCLUDES_'+uselib,qtincludes)
+			env.append_unique('INCLUDES_'+uselib,os.path.join(qtincludes,name.replace('Qt5','Qt')))
+			return k
+	return False
+@conf
+def find_qt5_libraries(self):
+	env=self.env
+	qtlibs=getattr(Options.options,'qtlibs',None)or self.environ.get('QT5_LIBDIR')
+	if not qtlibs:
+		try:
+			qtlibs=self.cmd_and_log(env.QMAKE+['-query','QT_INSTALL_LIBS']).strip()
+		except Errors.WafError:
+			qtdir=self.cmd_and_log(env.QMAKE+['-query','QT_INSTALL_PREFIX']).strip()
+			qtlibs=os.path.join(qtdir,'lib')
+	self.msg('Found the Qt5 libraries in',qtlibs)
+	qtincludes=self.environ.get('QT5_INCLUDES')or self.cmd_and_log(env.QMAKE+['-query','QT_INSTALL_HEADERS']).strip()
+	force_static=self.environ.get('QT5_FORCE_STATIC')
+	try:
+		if self.environ.get('QT5_XCOMPILE'):
+			self.fatal('QT5_XCOMPILE Disables pkg-config detection')
+		self.check_cfg(atleast_pkgconfig_version='0.1')
+	except self.errors.ConfigurationError:
+		for i in self.qt5_vars:
+			uselib=i.upper()
+			if Utils.unversioned_sys_platform()=='darwin':
+				frameworkName=i+'.framework'
+				qtDynamicLib=os.path.join(qtlibs,frameworkName,i)
+				if os.path.exists(qtDynamicLib):
+					env.append_unique('FRAMEWORK_'+uselib,i)
+					self.msg('Checking for %s'%i,qtDynamicLib,'GREEN')
+				else:
+					self.msg('Checking for %s'%i,False,'YELLOW')
+				env.append_unique('INCLUDES_'+uselib,os.path.join(qtlibs,frameworkName,'Headers'))
+			else:
+				for j in('','d'):
+					k='_DEBUG'if j=='d'else''
+					ret=self.find_single_qt5_lib(i+j,uselib+k,qtlibs,qtincludes,force_static)
+					if not force_static and not ret:
+						ret=self.find_single_qt5_lib(i+j,uselib+k,qtlibs,qtincludes,True)
+					self.msg('Checking for %s'%(i+j),ret,'GREEN'if ret else'YELLOW')
+	else:
+		path='%s:%s:%s/pkgconfig:/usr/lib/qt5/lib/pkgconfig:/opt/qt5/lib/pkgconfig:/usr/lib/qt5/lib:/opt/qt5/lib'%(self.environ.get('PKG_CONFIG_PATH',''),qtlibs,qtlibs)
+		for i in self.qt5_vars_debug+self.qt5_vars:
+			self.check_cfg(package=i,args='--cflags --libs',mandatory=False,force_static=force_static,pkg_config_path=path)
+@conf
+def simplify_qt5_libs(self):
+	env=self.env
+	def process_lib(vars_,coreval):
+		for d in vars_:
+			var=d.upper()
+			if var=='QTCORE':
+				continue
+			value=env['LIBPATH_'+var]
+			if value:
+				core=env[coreval]
+				accu=[]
+				for lib in value:
+					if lib in core:
+						continue
+					accu.append(lib)
+				env['LIBPATH_'+var]=accu
+	process_lib(self.qt5_vars,'LIBPATH_QTCORE')
+	process_lib(self.qt5_vars_debug,'LIBPATH_QTCORE_DEBUG')
+@conf
+def add_qt5_rpath(self):
+	env=self.env
+	if getattr(Options.options,'want_rpath',False):
+		def process_rpath(vars_,coreval):
+			for d in vars_:
+				var=d.upper()
+				value=env['LIBPATH_'+var]
+				if value:
+					core=env[coreval]
+					accu=[]
+					for lib in value:
+						if var!='QTCORE':
+							if lib in core:
+								continue
+						accu.append('-Wl,--rpath='+lib)
+					env['RPATH_'+var]=accu
+		process_rpath(self.qt5_vars,'LIBPATH_QTCORE')
+		process_rpath(self.qt5_vars_debug,'LIBPATH_QTCORE_DEBUG')
+@conf
+def set_qt5_libs_to_check(self):
+	if not hasattr(self,'qt5_vars'):
+		self.qt5_vars=QT5_LIBS
+	self.qt5_vars=Utils.to_list(self.qt5_vars)
+	if not hasattr(self,'qt5_vars_debug'):
+		self.qt5_vars_debug=[a+'_DEBUG'for a in self.qt5_vars]
+	self.qt5_vars_debug=Utils.to_list(self.qt5_vars_debug)
+@conf
+def set_qt5_defines(self):
+	if sys.platform!='win32':
+		return
+	for x in self.qt5_vars:
+		y=x.replace('Qt5','Qt')[2:].upper()
+		self.env.append_unique('DEFINES_%s'%x.upper(),'QT_%s_LIB'%y)
+		self.env.append_unique('DEFINES_%s_DEBUG'%x.upper(),'QT_%s_LIB'%y)
+def options(opt):
+	opt.add_option('--want-rpath',action='store_true',default=False,dest='want_rpath',help='enable the rpath for qt libraries')
+	for i in'qtdir qtbin qtlibs'.split():
+		opt.add_option('--'+i,type='string',default='',dest=i)
+	opt.add_option('--translate',action='store_true',help='collect translation strings',dest='trans_qt5',default=False)
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ruby.py glmark2-2021.02/waflib/Tools/ruby.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/ruby.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/ruby.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,11 +1,13 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os
-from waflib import Task,Options,Utils
-from waflib.TaskGen import before_method,feature,after_method,Task,extension
+from waflib import Options,Utils,Task
+from waflib.TaskGen import before_method,feature,extension
 from waflib.Configure import conf
+@feature('rubyext')
+@before_method('apply_incpaths','process_source','apply_bundle','apply_link')
 def init_rubyext(self):
 	self.install_path='${ARCHDIR_RUBY}'
 	self.uselib=self.to_list(getattr(self,'uselib',''))
@@ -13,31 +15,29 @@
 		self.uselib.append('RUBY')
 	if not'RUBYEXT'in self.uselib:
 		self.uselib.append('RUBYEXT')
+@feature('rubyext')
+@before_method('apply_link','propagate_uselib_vars')
 def apply_ruby_so_name(self):
-	self.env['cshlib_PATTERN']=self.env['cxxshlib_PATTERN']=self.env['rubyext_PATTERN']
+	self.env.cshlib_PATTERN=self.env.cxxshlib_PATTERN=self.env.rubyext_PATTERN
+@conf
 def check_ruby_version(self,minver=()):
-	if Options.options.rubybinary:
-		self.env.RUBY=Options.options.rubybinary
-	else:
-		self.find_program('ruby',var='RUBY')
-	ruby=self.env.RUBY
+	ruby=self.find_program('ruby',var='RUBY',value=Options.options.rubybinary)
 	try:
-		version=self.cmd_and_log([ruby,'-e','puts defined?(VERSION) ? VERSION : RUBY_VERSION']).strip()
-	except:
+		version=self.cmd_and_log(ruby+['-e','puts defined?(VERSION) ? VERSION : RUBY_VERSION']).strip()
+	except Exception:
 		self.fatal('could not determine ruby version')
 	self.env.RUBY_VERSION=version
 	try:
 		ver=tuple(map(int,version.split(".")))
-	except:
+	except Exception:
 		self.fatal('unsupported ruby version %r'%version)
 	cver=''
 	if minver:
+		cver='> '+'.'.join(str(x)for x in minver)
 		if ver<minver:
 			self.fatal('ruby is too old %r'%ver)
-		cver='.'.join([str(x)for x in minver])
-	else:
-		cver=ver
-	self.msg('Checking for ruby version %s'%str(minver or''),cver)
+	self.msg('Checking for ruby version %s'%cver,version)
+@conf
 def check_ruby_ext_devel(self):
 	if not self.env.RUBY:
 		self.fatal('ruby detection is required first')
@@ -45,17 +45,17 @@
 		self.fatal('load a c/c++ compiler first')
 	version=tuple(map(int,self.env.RUBY_VERSION.split(".")))
 	def read_out(cmd):
-		return Utils.to_list(self.cmd_and_log([self.env.RUBY,'-rrbconfig','-e',cmd]))
+		return Utils.to_list(self.cmd_and_log(self.env.RUBY+['-rrbconfig','-e',cmd]))
 	def read_config(key):
-		return read_out('puts Config::CONFIG[%r]'%key)
-	ruby=self.env['RUBY']
-	archdir=read_config('archdir')
-	cpppath=archdir
+		return read_out('puts RbConfig::CONFIG[%r]'%key)
+	cpppath=archdir=read_config('archdir')
 	if version>=(1,9,0):
 		ruby_hdrdir=read_config('rubyhdrdir')
 		cpppath+=ruby_hdrdir
+		if version>=(2,0,0):
+			cpppath+=read_config('rubyarchhdrdir')
 		cpppath+=[os.path.join(ruby_hdrdir[0],read_config('arch')[0])]
-	self.check(header_name='ruby.h',includes=cpppath,errmsg='could not find ruby header file')
+	self.check(header_name='ruby.h',includes=cpppath,errmsg='could not find ruby header file',link_header_test=False)
 	self.env.LIBPATH_RUBYEXT=read_config('libdir')
 	self.env.LIBPATH_RUBYEXT+=archdir
 	self.env.INCLUDES_RUBYEXT=cpppath
@@ -77,28 +77,21 @@
 		self.env.LIBDIR_RUBY=Options.options.rubylibdir
 	else:
 		self.env.LIBDIR_RUBY=read_config('sitelibdir')[0]
+@conf
 def check_ruby_module(self,module_name):
 	self.start_msg('Ruby module %s'%module_name)
 	try:
-		self.cmd_and_log([self.env['RUBY'],'-e','require \'%s\';puts 1'%module_name])
-	except:
+		self.cmd_and_log(self.env.RUBY+['-e','require \'%s\';puts 1'%module_name])
+	except Exception:
 		self.end_msg(False)
 		self.fatal('Could not find the ruby module %r'%module_name)
 	self.end_msg(True)
+@extension('.rb')
 def process(self,node):
-	tsk=self.create_task('run_ruby',node)
+	return self.create_task('run_ruby',node)
 class run_ruby(Task.Task):
 	run_str='${RUBY} ${RBFLAGS} -I ${SRC[0].parent.abspath()} ${SRC}'
 def options(opt):
 	opt.add_option('--with-ruby-archdir',type='string',dest='rubyarchdir',help='Specify directory where to install arch specific files')
 	opt.add_option('--with-ruby-libdir',type='string',dest='rubylibdir',help='Specify alternate ruby library path')
 	opt.add_option('--with-ruby-binary',type='string',dest='rubybinary',help='Specify alternate ruby binary')
-
-feature('rubyext')(init_rubyext)
-before_method('apply_incpaths','apply_lib_vars','apply_bundle','apply_link')(init_rubyext)
-feature('rubyext')(apply_ruby_so_name)
-before_method('apply_link','propagate_uselib')(apply_ruby_so_name)
-conf(check_ruby_version)
-conf(check_ruby_ext_devel)
-conf(check_ruby_module)
-extension('.rb')(process)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/suncc.py glmark2-2021.02/waflib/Tools/suncc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/suncc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/suncc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,47 +1,43 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os
-from waflib import Utils
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_scc(conf):
 	v=conf.env
-	cc=None
-	if v['CC']:cc=v['CC']
-	elif'CC'in conf.environ:cc=conf.environ['CC']
-	if not cc:cc=conf.find_program('cc',var='CC')
-	if not cc:conf.fatal('Could not find a Sun C compiler')
-	cc=conf.cmd_to_list(cc)
+	cc=conf.find_program('cc',var='CC')
 	try:
 		conf.cmd_and_log(cc+['-flags'])
-	except:
+	except Exception:
 		conf.fatal('%r is not a Sun compiler'%cc)
-	v['CC']=cc
-	v['CC_NAME']='sun'
+	v.CC_NAME='sun'
+	conf.get_suncc_version(cc)
+@conf
 def scc_common_flags(conf):
 	v=conf.env
-	v['CC_SRC_F']=[]
-	v['CC_TGT_F']=['-c','-o']
-	if not v['LINK_CC']:v['LINK_CC']=v['CC']
-	v['CCLNK_SRC_F']=''
-	v['CCLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['SONAME_ST']='-Wl,-h,%s'
-	v['SHLIB_MARKER']='-Bdynamic'
-	v['STLIB_MARKER']='-Bstatic'
-	v['cprogram_PATTERN']='%s'
-	v['CFLAGS_cshlib']=['-Kpic','-DPIC']
-	v['LINKFLAGS_cshlib']=['-G']
-	v['cshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cstlib']=['-Bstatic']
-	v['cstlib_PATTERN']='lib%s.a'
+	v.CC_SRC_F=[]
+	v.CC_TGT_F=['-c','-o','']
+	if not v.LINK_CC:
+		v.LINK_CC=v.CC
+	v.CCLNK_SRC_F=''
+	v.CCLNK_TGT_F=['-o','']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.SONAME_ST='-Wl,-h,%s'
+	v.SHLIB_MARKER='-Bdynamic'
+	v.STLIB_MARKER='-Bstatic'
+	v.cprogram_PATTERN='%s'
+	v.CFLAGS_cshlib=['-xcode=pic32','-DPIC']
+	v.LINKFLAGS_cshlib=['-G']
+	v.cshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cstlib=['-Bstatic']
+	v.cstlib_PATTERN='lib%s.a'
 def configure(conf):
 	conf.find_scc()
 	conf.find_ar()
@@ -49,6 +45,3 @@
 	conf.cc_load_tools()
 	conf.cc_add_flags()
 	conf.link_add_flags()
-
-conf(find_scc)
-conf(scc_common_flags)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/suncxx.py glmark2-2021.02/waflib/Tools/suncxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/suncxx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/suncxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,48 +1,43 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os
-from waflib import Utils
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_sxx(conf):
 	v=conf.env
-	cc=None
-	if v['CXX']:cc=v['CXX']
-	elif'CXX'in conf.environ:cc=conf.environ['CXX']
-	if not cc:cc=conf.find_program('CC',var='CXX')
-	if not cc:cc=conf.find_program('c++',var='CXX')
-	if not cc:conf.fatal('Could not find a Sun C++ compiler')
-	cc=conf.cmd_to_list(cc)
+	cc=conf.find_program(['CC','c++'],var='CXX')
 	try:
 		conf.cmd_and_log(cc+['-flags'])
-	except:
+	except Exception:
 		conf.fatal('%r is not a Sun compiler'%cc)
-	v['CXX']=cc
-	v['CXX_NAME']='sun'
+	v.CXX_NAME='sun'
+	conf.get_suncc_version(cc)
+@conf
 def sxx_common_flags(conf):
 	v=conf.env
-	v['CXX_SRC_F']=[]
-	v['CXX_TGT_F']=['-c','-o']
-	if not v['LINK_CXX']:v['LINK_CXX']=v['CXX']
-	v['CXXLNK_SRC_F']=[]
-	v['CXXLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['SONAME_ST']='-Wl,-h,%s'
-	v['SHLIB_MARKER']='-Bdynamic'
-	v['STLIB_MARKER']='-Bstatic'
-	v['cxxprogram_PATTERN']='%s'
-	v['CXXFLAGS_cxxshlib']=['-Kpic','-DPIC']
-	v['LINKFLAGS_cxxshlib']=['-G']
-	v['cxxshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cxxstlib']=['-Bstatic']
-	v['cxxstlib_PATTERN']='lib%s.a'
+	v.CXX_SRC_F=[]
+	v.CXX_TGT_F=['-c','-o','']
+	if not v.LINK_CXX:
+		v.LINK_CXX=v.CXX
+	v.CXXLNK_SRC_F=[]
+	v.CXXLNK_TGT_F=['-o','']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.SONAME_ST='-Wl,-h,%s'
+	v.SHLIB_MARKER='-Bdynamic'
+	v.STLIB_MARKER='-Bstatic'
+	v.cxxprogram_PATTERN='%s'
+	v.CXXFLAGS_cxxshlib=['-xcode=pic32','-DPIC']
+	v.LINKFLAGS_cxxshlib=['-G']
+	v.cxxshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cxxstlib=['-Bstatic']
+	v.cxxstlib_PATTERN='lib%s.a'
 def configure(conf):
 	conf.find_sxx()
 	conf.find_ar()
@@ -50,6 +45,3 @@
 	conf.cxx_load_tools()
 	conf.cxx_add_flags()
 	conf.link_add_flags()
-
-conf(find_sxx)
-conf(sxx_common_flags)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/tex.py glmark2-2021.02/waflib/Tools/tex.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/tex.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/tex.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,33 +1,33 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 import os,re
-from waflib import Utils,Task,Errors
+from waflib import Utils,Task,Errors,Logs,Node
 from waflib.TaskGen import feature,before_method
-from waflib.Logs import error,warn,debug
 re_bibunit=re.compile(r'\\(?P<type>putbib)\[(?P<file>[^\[\]]*)\]',re.M)
 def bibunitscan(self):
 	node=self.inputs[0]
 	nodes=[]
 	if not node:return nodes
-	code=Utils.readf(node.abspath())
+	code=node.read()
 	for match in re_bibunit.finditer(code):
 		path=match.group('file')
 		if path:
-			for k in['','.bib']:
-				debug('tex: trying %s%s'%(path,k))
+			for k in('','.bib'):
+				Logs.debug('tex: trying %s%s',path,k)
 				fi=node.parent.find_resource(path+k)
 				if fi:
 					nodes.append(fi)
 			else:
-				debug('tex: could not find %s'%path)
-	debug("tex: found the following bibunit files: %s"%nodes)
+				Logs.debug('tex: could not find %s',path)
+	Logs.debug('tex: found the following bibunit files: %s',nodes)
 	return nodes
-exts_deps_tex=['','.ltx','.tex','.bib','.pdf','.png','.eps','.ps']
+exts_deps_tex=['','.ltx','.tex','.bib','.pdf','.png','.eps','.ps','.sty']
 exts_tex=['.ltx','.tex']
-re_tex=re.compile(r'\\(?P<type>include|bibliography|putbib|includegraphics|input|import|bringin|lstinputlisting)(\[[^\[\]]*\])?{(?P<file>[^{}]*)}',re.M)
+re_tex=re.compile(r'\\(?P<type>usepackage|RequirePackage|include|bibliography([^\[\]{}]*)|putbib|includegraphics|input|import|bringin|lstinputlisting)(\[[^\[\]]*\])?{(?P<file>[^{}]*)}',re.M)
 g_bibtex_re=re.compile('bibdata',re.M)
+g_glossaries_re=re.compile('\\@newglossary',re.M)
 class tex(Task.Task):
 	bibtex_fun,_=Task.compile_fun('${BIBTEX} ${BIBTEXFLAGS} ${SRCFILE}',shell=False)
 	bibtex_fun.__doc__="""
@@ -37,6 +37,13 @@
 	makeindex_fun.__doc__="""
 	Execute the program **makeindex**
 	"""
+	makeglossaries_fun,_=Task.compile_fun('${MAKEGLOSSARIES} ${SRCFILE}',shell=False)
+	makeglossaries_fun.__doc__="""
+	Execute the program **makeglossaries**
+	"""
+	def exec_command(self,cmd,**kw):
+		kw['stdout']=kw['stderr']=None
+		return super(tex,self).exec_command(cmd,**kw)
 	def scan_aux(self,node):
 		nodes=[node]
 		re_aux=re.compile(r'\\@input{(?P<file>[^{}]*)}',re.M)
@@ -46,7 +53,7 @@
 				path=match.group('file')
 				found=node.parent.find_or_declare(path)
 				if found and found not in nodes:
-					debug('tex: found aux node '+found.abspath())
+					Logs.debug('tex: found aux node %r',found)
 					nodes.append(found)
 					parse_node(found)
 		parse_node(node)
@@ -64,114 +71,157 @@
 			code=node.read()
 			global re_tex
 			for match in re_tex.finditer(code):
+				multibib=match.group('type')
+				if multibib and multibib.startswith('bibliography'):
+					multibib=multibib[len('bibliography'):]
+					if multibib.startswith('style'):
+						continue
+				else:
+					multibib=None
 				for path in match.group('file').split(','):
 					if path:
 						add_name=True
 						found=None
 						for k in exts_deps_tex:
-							debug('tex: trying %s%s'%(path,k))
-							found=node.parent.find_resource(path+k)
-							if found and not found in self.outputs:
+							for up in self.texinputs_nodes:
+								Logs.debug('tex: trying %s%s',path,k)
+								found=up.find_resource(path+k)
+								if found:
+									break
+							for tsk in self.generator.tasks:
+								if not found or found in tsk.outputs:
+									break
+							else:
 								nodes.append(found)
 								add_name=False
 								for ext in exts_tex:
 									if found.name.endswith(ext):
 										parse_node(found)
 										break
+							if found and multibib and found.name.endswith('.bib'):
+								try:
+									self.multibibs.append(found)
+								except AttributeError:
+									self.multibibs=[found]
 						if add_name:
 							names.append(path)
 		parse_node(node)
 		for x in nodes:
 			x.parent.get_bld().mkdir()
-		debug("tex: found the following : %s and names %s"%(nodes,names))
+		Logs.debug("tex: found the following : %s and names %s",nodes,names)
 		return(nodes,names)
 	def check_status(self,msg,retcode):
 		if retcode!=0:
-			raise Errors.WafError("%r command exit status %r"%(msg,retcode))
+			raise Errors.WafError('%r command exit status %r'%(msg,retcode))
 	def bibfile(self):
-		need_bibtex=False
-		try:
-			for aux_node in self.aux_nodes:
+		for aux_node in self.aux_nodes:
+			try:
 				ct=aux_node.read()
-				if g_bibtex_re.findall(ct):
-					need_bibtex=True
-					break
-		except(OSError,IOError):
-			error('error bibtex scan')
-		else:
-			if need_bibtex:
-				warn('calling bibtex')
+			except EnvironmentError:
+				Logs.error('Error reading %s: %r',aux_node.abspath())
+				continue
+			if g_bibtex_re.findall(ct):
+				Logs.info('calling bibtex')
 				self.env.env={}
 				self.env.env.update(os.environ)
-				self.env.env.update({'BIBINPUTS':self.TEXINPUTS,'BSTINPUTS':self.TEXINPUTS})
-				self.env.SRCFILE=self.aux_nodes[0].name[:-4]
+				self.env.env.update({'BIBINPUTS':self.texinputs(),'BSTINPUTS':self.texinputs()})
+				self.env.SRCFILE=aux_node.name[:-4]
 				self.check_status('error when calling bibtex',self.bibtex_fun())
+		for node in getattr(self,'multibibs',[]):
+			self.env.env={}
+			self.env.env.update(os.environ)
+			self.env.env.update({'BIBINPUTS':self.texinputs(),'BSTINPUTS':self.texinputs()})
+			self.env.SRCFILE=node.name[:-4]
+			self.check_status('error when calling bibtex',self.bibtex_fun())
 	def bibunits(self):
 		try:
 			bibunits=bibunitscan(self)
-		except FSError:
-			error('error bibunitscan')
+		except OSError:
+			Logs.error('error bibunitscan')
 		else:
 			if bibunits:
-				fn=['bu'+str(i)for i in xrange(1,len(bibunits)+1)]
+				fn=['bu'+str(i)for i in range(1,len(bibunits)+1)]
 				if fn:
-					warn('calling bibtex on bibunits')
+					Logs.info('calling bibtex on bibunits')
 				for f in fn:
-					self.env.env={'BIBINPUTS':self.TEXINPUTS,'BSTINPUTS':self.TEXINPUTS}
+					self.env.env={'BIBINPUTS':self.texinputs(),'BSTINPUTS':self.texinputs()}
 					self.env.SRCFILE=f
 					self.check_status('error when calling bibtex',self.bibtex_fun())
 	def makeindex(self):
+		self.idx_node=self.inputs[0].change_ext('.idx')
 		try:
 			idx_path=self.idx_node.abspath()
 			os.stat(idx_path)
 		except OSError:
-			warn('index file %s absent, not calling makeindex'%idx_path)
+			Logs.info('index file %s absent, not calling makeindex',idx_path)
 		else:
-			warn('calling makeindex')
+			Logs.info('calling makeindex')
 			self.env.SRCFILE=self.idx_node.name
 			self.env.env={}
 			self.check_status('error when calling makeindex %s'%idx_path,self.makeindex_fun())
+	def bibtopic(self):
+		p=self.inputs[0].parent.get_bld()
+		if os.path.exists(os.path.join(p.abspath(),'btaux.aux')):
+			self.aux_nodes+=p.ant_glob('*[0-9].aux')
+	def makeglossaries(self):
+		src_file=self.inputs[0].abspath()
+		base_file=os.path.basename(src_file)
+		base,_=os.path.splitext(base_file)
+		for aux_node in self.aux_nodes:
+			try:
+				ct=aux_node.read()
+			except EnvironmentError:
+				Logs.error('Error reading %s: %r',aux_node.abspath())
+				continue
+			if g_glossaries_re.findall(ct):
+				if not self.env.MAKEGLOSSARIES:
+					raise Errors.WafError("The program 'makeglossaries' is missing!")
+				Logs.warn('calling makeglossaries')
+				self.env.SRCFILE=base
+				self.check_status('error when calling makeglossaries %s'%base,self.makeglossaries_fun())
+				return
+	def texinputs(self):
+		return os.pathsep.join([k.abspath()for k in self.texinputs_nodes])+os.pathsep
 	def run(self):
 		env=self.env
-		if not env['PROMPT_LATEX']:
+		if not env.PROMPT_LATEX:
 			env.append_value('LATEXFLAGS','-interaction=batchmode')
 			env.append_value('PDFLATEXFLAGS','-interaction=batchmode')
 			env.append_value('XELATEXFLAGS','-interaction=batchmode')
-		fun=self.texfun
-		node=self.inputs[0]
-		srcfile=node.abspath()
-		texinputs=self.env.TEXINPUTS or''
-		self.TEXINPUTS=node.parent.get_bld().abspath()+os.pathsep+node.parent.get_src().abspath()+os.pathsep+texinputs+os.pathsep
-		self.aux_node=node.change_ext('.aux')
-		self.cwd=self.inputs[0].parent.get_bld().abspath()
-		warn('first pass on %s'%self.__class__.__name__)
-		self.env.env={}
-		self.env.env.update(os.environ)
-		self.env.env.update({'TEXINPUTS':self.TEXINPUTS})
-		self.env.SRCFILE=srcfile
-		self.check_status('error when calling latex',fun())
-		self.aux_nodes=self.scan_aux(node.change_ext('.aux'))
-		self.idx_node=node.change_ext('.idx')
+		self.cwd=self.inputs[0].parent.get_bld()
+		Logs.info('first pass on %s',self.__class__.__name__)
+		cur_hash=self.hash_aux_nodes()
+		self.call_latex()
+		self.hash_aux_nodes()
+		self.bibtopic()
 		self.bibfile()
 		self.bibunits()
 		self.makeindex()
-		hash=''
+		self.makeglossaries()
 		for i in range(10):
-			prev_hash=hash
-			try:
-				hashes=[Utils.h_file(x.abspath())for x in self.aux_nodes]
-				hash=Utils.h_list(hashes)
-			except(OSError,IOError):
-				error('could not read aux.h')
-				pass
-			if hash and hash==prev_hash:
+			prev_hash=cur_hash
+			cur_hash=self.hash_aux_nodes()
+			if not cur_hash:
+				Logs.error('No aux.h to process')
+			if cur_hash and cur_hash==prev_hash:
 				break
-			warn('calling %s'%self.__class__.__name__)
-			self.env.env={}
-			self.env.env.update(os.environ)
-			self.env.env.update({'TEXINPUTS':self.TEXINPUTS})
-			self.env.SRCFILE=srcfile
-			self.check_status('error when calling %s'%self.__class__.__name__,fun())
+			Logs.info('calling %s',self.__class__.__name__)
+			self.call_latex()
+	def hash_aux_nodes(self):
+		try:
+			self.aux_nodes
+		except AttributeError:
+			try:
+				self.aux_nodes=self.scan_aux(self.inputs[0].change_ext('.aux'))
+			except IOError:
+				return None
+		return Utils.h_list([Utils.h_file(x.abspath())for x in self.aux_nodes])
+	def call_latex(self):
+		self.env.env={}
+		self.env.env.update(os.environ)
+		self.env.env.update({'TEXINPUTS':self.texinputs()})
+		self.env.SRCFILE=self.inputs[0].abspath()
+		self.check_status('error when calling latex',self.texfun())
 class latex(tex):
 	texfun,vars=Task.compile_fun('${LATEX} ${LATEXFLAGS} ${SRCFILE}',shell=False)
 class pdflatex(tex):
@@ -190,18 +240,25 @@
 	run_str='${PDF2PS} ${PDF2PSFLAGS} ${SRC} ${TGT}'
 	color='BLUE'
 	after=['latex','pdflatex','xelatex']
+@feature('tex')
+@before_method('process_source')
 def apply_tex(self):
-	if not getattr(self,'type',None)in['latex','pdflatex','xelatex']:
+	if not getattr(self,'type',None)in('latex','pdflatex','xelatex'):
 		self.type='pdflatex'
-	tree=self.bld
 	outs=Utils.to_list(getattr(self,'outs',[]))
-	self.env['PROMPT_LATEX']=getattr(self,'prompt',1)
+	self.env.PROMPT_LATEX=getattr(self,'prompt',1)
 	deps_lst=[]
 	if getattr(self,'deps',None):
 		deps=self.to_list(self.deps)
-		for filename in deps:
-			n=self.path.find_resource(filename)
-			if not n in deps_lst:deps_lst.append(n)
+		for dep in deps:
+			if isinstance(dep,str):
+				n=self.path.find_resource(dep)
+				if not n:
+					self.bld.fatal('Could not find %r for %r'%(dep,self))
+				if not n in deps_lst:
+					deps_lst.append(n)
+			elif isinstance(dep,Node.Node):
+				deps_lst.append(dep)
 	for node in self.to_nodes(self.source):
 		if self.type=='latex':
 			task=self.create_task('latex',node,node.change_ext('.dvi'))
@@ -211,32 +268,44 @@
 			task=self.create_task('xelatex',node,node.change_ext('.pdf'))
 		task.env=self.env
 		if deps_lst:
-			try:
-				lst=tree.node_deps[task.uid()]
-				for n in deps_lst:
-					if not n in lst:
-						lst.append(n)
-			except KeyError:
-				tree.node_deps[task.uid()]=deps_lst
+			for n in deps_lst:
+				if not n in task.dep_nodes:
+					task.dep_nodes.append(n)
+		if hasattr(self,'texinputs_nodes'):
+			task.texinputs_nodes=self.texinputs_nodes
+		else:
+			task.texinputs_nodes=[node.parent,node.parent.get_bld(),self.path,self.path.get_bld()]
+			lst=os.environ.get('TEXINPUTS','')
+			if self.env.TEXINPUTS:
+				lst+=os.pathsep+self.env.TEXINPUTS
+			if lst:
+				lst=lst.split(os.pathsep)
+			for x in lst:
+				if x:
+					if os.path.isabs(x):
+						p=self.bld.root.find_node(x)
+						if p:
+							task.texinputs_nodes.append(p)
+						else:
+							Logs.error('Invalid TEXINPUTS folder %s',x)
+					else:
+						Logs.error('Cannot resolve relative paths in TEXINPUTS %s',x)
 		if self.type=='latex':
 			if'ps'in outs:
 				tsk=self.create_task('dvips',task.outputs,node.change_ext('.ps'))
-				tsk.env.env={'TEXINPUTS':node.parent.abspath()+os.pathsep+self.path.abspath()+os.pathsep+self.path.get_bld().abspath()}
+				tsk.env.env=dict(os.environ)
 			if'pdf'in outs:
 				tsk=self.create_task('dvipdf',task.outputs,node.change_ext('.pdf'))
-				tsk.env.env={'TEXINPUTS':node.parent.abspath()+os.pathsep+self.path.abspath()+os.pathsep+self.path.get_bld().abspath()}
+				tsk.env.env=dict(os.environ)
 		elif self.type=='pdflatex':
 			if'ps'in outs:
 				self.create_task('pdf2ps',task.outputs,node.change_ext('.ps'))
 	self.source=[]
 def configure(self):
 	v=self.env
-	for p in'tex latex pdflatex xelatex bibtex dvips dvipdf ps2pdf makeindex pdf2ps'.split():
+	for p in'tex latex pdflatex xelatex bibtex dvips dvipdf ps2pdf makeindex pdf2ps makeglossaries'.split():
 		try:
 			self.find_program(p,var=p.upper())
 		except self.errors.ConfigurationError:
 			pass
-	v['DVIPSFLAGS']='-Ppdf'
-
-feature('tex')(apply_tex)
-before_method('process_source')(apply_tex)
\ No newline at end of file
+	v.DVIPSFLAGS='-Ppdf'
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/vala.py glmark2-2021.02/waflib/Tools/vala.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/vala.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/vala.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,202 +1,203 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os.path,shutil,re
-from waflib import Context,Task,Utils,Logs,Options,Errors
-from waflib.TaskGen import extension
+import re
+from waflib import Context,Task,Utils,Logs,Options,Errors,Node
+from waflib.TaskGen import extension,taskgen_method
 from waflib.Configure import conf
 class valac(Task.Task):
 	vars=["VALAC","VALAC_VERSION","VALAFLAGS"]
 	ext_out=['.h']
 	def run(self):
-		env=self.env
-		cmd=[env['VALAC'],'-C','--quiet']
-		cmd.extend(Utils.to_list(env['VALAFLAGS']))
-		if self.threading:
-			cmd.append('--thread')
-		if self.profile:
-			cmd.append('--profile=%s'%self.profile)
-		if self.target_glib:
-			cmd.append('--target-glib=%s'%self.target_glib)
-		if self.is_lib:
-			cmd.append('--library='+self.target)
-			for x in self.outputs:
-				if x.name.endswith('.h'):
-					cmd.append('--header='+x.name)
-			if self.gir:
-				cmd.append('--gir=%s.gir'%self.gir)
-		for vapi_dir in self.vapi_dirs:
-			cmd.append('--vapidir=%s'%vapi_dir)
-		for package in self.packages:
-			cmd.append('--pkg=%s'%package)
-		for package in self.packages_private:
-			cmd.append('--pkg=%s'%package)
-		for define in self.vala_defines:
-			cmd.append('--define=%s'%define)
-		cmd.extend([a.abspath()for a in self.inputs])
-		ret=self.exec_command(cmd,cwd=self.outputs[0].parent.abspath())
+		cmd=self.env.VALAC+self.env.VALAFLAGS
+		resources=getattr(self,'vala_exclude',[])
+		cmd.extend([a.abspath()for a in self.inputs if a not in resources])
+		ret=self.exec_command(cmd,cwd=self.vala_dir_node.abspath())
 		if ret:
 			return ret
-		for x in self.outputs:
-			if id(x.parent)!=id(self.outputs[0].parent):
-				shutil.move(self.outputs[0].parent.abspath()+os.sep+x.name,x.abspath())
-		if self.packages and getattr(self,'deps_node',None):
-			self.deps_node.write('\n'.join(self.packages))
+		if self.generator.dump_deps_node:
+			self.generator.dump_deps_node.write('\n'.join(self.generator.packages))
 		return ret
-def vala_file(self,node):
-	valatask=getattr(self,"valatask",None)
-	if not valatask:
-		def _get_api_version():
-			api_version='1.0'
-			if hasattr(Context.g_module,'API_VERSION'):
-				version=Context.g_module.API_VERSION.split(".")
-				if version[0]=="0":
-					api_version="0."+version[1]
-				else:
-					api_version=version[0]+".0"
-			return api_version
-		valatask=self.create_task('valac')
-		self.valatask=valatask
-		self.includes=Utils.to_list(getattr(self,'includes',[]))
-		self.uselib=self.to_list(getattr(self,'uselib',[]))
-		valatask.packages=[]
-		valatask.packages_private=Utils.to_list(getattr(self,'packages_private',[]))
-		valatask.vapi_dirs=[]
-		valatask.target=self.target
-		valatask.threading=False
-		valatask.install_path=getattr(self,'install_path','')
-		valatask.profile=getattr(self,'profile','gobject')
-		valatask.vala_defines=getattr(self,'vala_defines',[])
-		valatask.target_glib=None
-		valatask.gir=getattr(self,'gir',None)
-		valatask.gir_path=getattr(self,'gir_path','${DATAROOTDIR}/gir-1.0')
-		valatask.vapi_path=getattr(self,'vapi_path','${DATAROOTDIR}/vala/vapi')
-		valatask.pkg_name=getattr(self,'pkg_name',self.env['PACKAGE'])
-		valatask.header_path=getattr(self,'header_path','${INCLUDEDIR}/%s-%s'%(valatask.pkg_name,_get_api_version()))
-		valatask.install_binding=getattr(self,'install_binding',True)
-		valatask.is_lib=False
-		if not'cprogram'in self.features:
-			valatask.is_lib=True
-		packages=Utils.to_list(getattr(self,'packages',[]))
-		vapi_dirs=Utils.to_list(getattr(self,'vapi_dirs',[]))
-		includes=[]
-		if hasattr(self,'use'):
-			local_packages=Utils.to_list(self.use)[:]
-			seen=[]
-			while len(local_packages)>0:
-				package=local_packages.pop()
-				if package in seen:
-					continue
-				seen.append(package)
-				try:
-					package_obj=self.bld.get_tgen_by_name(package)
-				except Errors.WafError:
-					continue
-				package_name=package_obj.target
-				package_node=package_obj.path
-				package_dir=package_node.path_from(self.path)
-				for task in package_obj.tasks:
-					for output in task.outputs:
-						if output.name==package_name+".vapi":
-							valatask.set_run_after(task)
-							if package_name not in packages:
-								packages.append(package_name)
-							if package_dir not in vapi_dirs:
-								vapi_dirs.append(package_dir)
-							if package_dir not in includes:
-								includes.append(package_dir)
-				if hasattr(package_obj,'use'):
-					lst=self.to_list(package_obj.use)
-					lst.reverse()
-					local_packages=[pkg for pkg in lst if pkg not in seen]+local_packages
-		valatask.packages=packages
-		for vapi_dir in vapi_dirs:
-			try:
-				valatask.vapi_dirs.append(self.path.find_dir(vapi_dir).abspath())
-				valatask.vapi_dirs.append(self.path.find_dir(vapi_dir).get_bld().abspath())
-			except AttributeError:
-				Logs.warn("Unable to locate Vala API directory: '%s'"%vapi_dir)
-		self.includes.append(self.bld.srcnode.abspath())
-		self.includes.append(self.bld.bldnode.abspath())
-		for include in includes:
+@taskgen_method
+def init_vala_task(self):
+	self.profile=getattr(self,'profile','gobject')
+	self.packages=packages=Utils.to_list(getattr(self,'packages',[]))
+	self.use=Utils.to_list(getattr(self,'use',[]))
+	if packages and not self.use:
+		self.use=packages[:]
+	if self.profile=='gobject':
+		if not'GOBJECT'in self.use:
+			self.use.append('GOBJECT')
+	def addflags(flags):
+		self.env.append_value('VALAFLAGS',flags)
+	if self.profile:
+		addflags('--profile=%s'%self.profile)
+	valatask=self.valatask
+	if hasattr(self,'vala_dir'):
+		if isinstance(self.vala_dir,str):
+			valatask.vala_dir_node=self.path.get_bld().make_node(self.vala_dir)
 			try:
-				self.includes.append(self.path.find_dir(include).abspath())
-				self.includes.append(self.path.find_dir(include).get_bld().abspath())
-			except AttributeError:
-				Logs.warn("Unable to locate include directory: '%s'"%include)
-		if valatask.profile=='gobject':
-			if hasattr(self,'target_glib'):
-				Logs.warn('target_glib on vala tasks is not supported --vala-target-glib=MAJOR.MINOR from the vala tool options')
-			if getattr(Options.options,'vala_target_glib',None):
-				valatask.target_glib=Options.options.vala_target_glib
-			if not'GOBJECT'in self.uselib:
-				self.uselib.append('GOBJECT')
-		if hasattr(self,'threading'):
-			if valatask.profile=='gobject':
-				valatask.threading=self.threading
-				if not'GTHREAD'in self.uselib:
-					self.uselib.append('GTHREAD')
+				valatask.vala_dir_node.mkdir()
+			except OSError:
+				raise self.bld.fatal('Cannot create the vala dir %r'%valatask.vala_dir_node)
+		else:
+			valatask.vala_dir_node=self.vala_dir
+	else:
+		valatask.vala_dir_node=self.path.get_bld()
+	addflags('--directory=%s'%valatask.vala_dir_node.abspath())
+	if hasattr(self,'thread'):
+		if self.profile=='gobject':
+			if not'GTHREAD'in self.use:
+				self.use.append('GTHREAD')
+		else:
+			Logs.warn('Profile %s means no threading support',self.profile)
+			self.thread=False
+		if self.thread:
+			addflags('--thread')
+	self.is_lib='cprogram'not in self.features
+	if self.is_lib:
+		addflags('--library=%s'%self.target)
+		h_node=valatask.vala_dir_node.find_or_declare('%s.h'%self.target)
+		valatask.outputs.append(h_node)
+		addflags('--header=%s'%h_node.name)
+		valatask.outputs.append(valatask.vala_dir_node.find_or_declare('%s.vapi'%self.target))
+		if getattr(self,'gir',None):
+			gir_node=valatask.vala_dir_node.find_or_declare('%s.gir'%self.gir)
+			addflags('--gir=%s'%gir_node.name)
+			valatask.outputs.append(gir_node)
+	self.vala_target_glib=getattr(self,'vala_target_glib',getattr(Options.options,'vala_target_glib',None))
+	if self.vala_target_glib:
+		addflags('--target-glib=%s'%self.vala_target_glib)
+	addflags(['--define=%s'%x for x in Utils.to_list(getattr(self,'vala_defines',[]))])
+	packages_private=Utils.to_list(getattr(self,'packages_private',[]))
+	addflags(['--pkg=%s'%x for x in packages_private])
+	def _get_api_version():
+		api_version='1.0'
+		if hasattr(Context.g_module,'API_VERSION'):
+			version=Context.g_module.API_VERSION.split(".")
+			if version[0]=="0":
+				api_version="0."+version[1]
 			else:
-				Logs.warn("Profile %s does not have threading support"%valatask.profile)
-		if valatask.is_lib:
-			valatask.outputs.append(self.path.find_or_declare('%s.h'%self.target))
-			valatask.outputs.append(self.path.find_or_declare('%s.vapi'%self.target))
-			if valatask.gir:
-				valatask.outputs.append(self.path.find_or_declare('%s.gir'%self.gir))
-			if valatask.packages:
-				d=self.path.find_or_declare('%s.deps'%self.target)
-				valatask.outputs.append(d)
-				valatask.deps_node=d
-	valatask.inputs.append(node)
-	c_node=node.change_ext('.c')
-	valatask.outputs.append(c_node)
-	self.source.append(c_node)
-	if valatask.is_lib and valatask.install_binding:
+				api_version=version[0]+".0"
+		return api_version
+	self.includes=Utils.to_list(getattr(self,'includes',[]))
+	valatask.install_path=getattr(self,'install_path','')
+	valatask.vapi_path=getattr(self,'vapi_path','${DATAROOTDIR}/vala/vapi')
+	valatask.pkg_name=getattr(self,'pkg_name',self.env.PACKAGE)
+	valatask.header_path=getattr(self,'header_path','${INCLUDEDIR}/%s-%s'%(valatask.pkg_name,_get_api_version()))
+	valatask.install_binding=getattr(self,'install_binding',True)
+	self.vapi_dirs=vapi_dirs=Utils.to_list(getattr(self,'vapi_dirs',[]))
+	if hasattr(self,'use'):
+		local_packages=Utils.to_list(self.use)[:]
+		seen=[]
+		while len(local_packages)>0:
+			package=local_packages.pop()
+			if package in seen:
+				continue
+			seen.append(package)
+			try:
+				package_obj=self.bld.get_tgen_by_name(package)
+			except Errors.WafError:
+				continue
+			package_name=package_obj.target
+			for task in package_obj.tasks:
+				for output in task.outputs:
+					if output.name==package_name+".vapi":
+						valatask.set_run_after(task)
+						if package_name not in packages:
+							packages.append(package_name)
+						if output.parent not in vapi_dirs:
+							vapi_dirs.append(output.parent)
+						if output.parent not in self.includes:
+							self.includes.append(output.parent)
+			if hasattr(package_obj,'use'):
+				lst=self.to_list(package_obj.use)
+				lst.reverse()
+				local_packages=[pkg for pkg in lst if pkg not in seen]+local_packages
+	addflags(['--pkg=%s'%p for p in packages])
+	for vapi_dir in vapi_dirs:
+		if isinstance(vapi_dir,Node.Node):
+			v_node=vapi_dir
+		else:
+			v_node=self.path.find_dir(vapi_dir)
+		if not v_node:
+			Logs.warn('Unable to locate Vala API directory: %r',vapi_dir)
+		else:
+			addflags('--vapidir=%s'%v_node.abspath())
+	self.dump_deps_node=None
+	if self.is_lib and self.packages:
+		self.dump_deps_node=valatask.vala_dir_node.find_or_declare('%s.deps'%self.target)
+		valatask.outputs.append(self.dump_deps_node)
+	if self.is_lib and valatask.install_binding:
 		headers_list=[o for o in valatask.outputs if o.suffix()==".h"]
 		try:
 			self.install_vheader.source=headers_list
 		except AttributeError:
-			self.install_vheader=self.bld.install_files(valatask.header_path,headers_list,self.env)
+			self.install_vheader=self.add_install_files(install_to=valatask.header_path,install_from=headers_list)
 		vapi_list=[o for o in valatask.outputs if(o.suffix()in(".vapi",".deps"))]
 		try:
 			self.install_vapi.source=vapi_list
 		except AttributeError:
-			self.install_vapi=self.bld.install_files(valatask.vapi_path,vapi_list,self.env)
-		gir_list=[o for o in valatask.outputs if o.suffix()==".gir"]
+			self.install_vapi=self.add_install_files(install_to=valatask.vapi_path,install_from=vapi_list)
+		gir_list=[o for o in valatask.outputs if o.suffix()=='.gir']
 		try:
 			self.install_gir.source=gir_list
 		except AttributeError:
-			self.install_gir=self.bld.install_files(valatask.gir_path,gir_list,self.env)
-valac=Task.update_outputs(valac)
+			self.install_gir=self.add_install_files(install_to=getattr(self,'gir_path','${DATAROOTDIR}/gir-1.0'),install_from=gir_list)
+	if hasattr(self,'vala_resources'):
+		nodes=self.to_nodes(self.vala_resources)
+		valatask.vala_exclude=getattr(valatask,'vala_exclude',[])+nodes
+		valatask.inputs.extend(nodes)
+		for x in nodes:
+			addflags(['--gresources',x.abspath()])
+@extension('.vala','.gs')
+def vala_file(self,node):
+	try:
+		valatask=self.valatask
+	except AttributeError:
+		valatask=self.valatask=self.create_task('valac')
+		self.init_vala_task()
+	valatask.inputs.append(node)
+	name=node.name[:node.name.rfind('.')]+'.c'
+	c_node=valatask.vala_dir_node.find_or_declare(name)
+	valatask.outputs.append(c_node)
+	self.source.append(c_node)
+@conf
 def find_valac(self,valac_name,min_version):
 	valac=self.find_program(valac_name,var='VALAC')
 	try:
-		output=self.cmd_and_log(valac+' --version')
+		output=self.cmd_and_log(valac+['--version'])
 	except Exception:
 		valac_version=None
 	else:
-		ver=re.search(r'\d+.\d+.\d+',output).group(0).split('.')
+		ver=re.search(r'\d+.\d+.\d+',output).group().split('.')
 		valac_version=tuple([int(x)for x in ver])
 	self.msg('Checking for %s version >= %r'%(valac_name,min_version),valac_version,valac_version and valac_version>=min_version)
 	if valac and valac_version<min_version:
 		self.fatal("%s version %r is too old, need >= %r"%(valac_name,valac_version,min_version))
-	self.env['VALAC_VERSION']=valac_version
+	self.env.VALAC_VERSION=valac_version
 	return valac
+@conf
 def check_vala(self,min_version=(0,8,0),branch=None):
+	if self.env.VALA_MINVER:
+		min_version=self.env.VALA_MINVER
+	if self.env.VALA_MINVER_BRANCH:
+		branch=self.env.VALA_MINVER_BRANCH
 	if not branch:
 		branch=min_version[:2]
 	try:
 		find_valac(self,'valac-%d.%d'%(branch[0],branch[1]),min_version)
 	except self.errors.ConfigurationError:
 		find_valac(self,'valac',min_version)
+@conf
 def check_vala_deps(self):
-	if not self.env['HAVE_GOBJECT']:
+	if not self.env.HAVE_GOBJECT:
 		pkg_args={'package':'gobject-2.0','uselib_store':'GOBJECT','args':'--cflags --libs'}
 		if getattr(Options.options,'vala_target_glib',None):
 			pkg_args['atleast_version']=Options.options.vala_target_glib
 		self.check_cfg(**pkg_args)
-	if not self.env['HAVE_GTHREAD']:
+	if not self.env.HAVE_GTHREAD:
 		pkg_args={'package':'gthread-2.0','uselib_store':'GTHREAD','args':'--cflags --libs'}
 		if getattr(Options.options,'vala_target_glib',None):
 			pkg_args['atleast_version']=Options.options.vala_target_glib
@@ -205,12 +206,9 @@
 	self.load('gnu_dirs')
 	self.check_vala_deps()
 	self.check_vala()
+	self.add_os_flags('VALAFLAGS')
+	self.env.append_unique('VALAFLAGS',['-C'])
 def options(opt):
 	opt.load('gnu_dirs')
 	valaopts=opt.add_option_group('Vala Compiler Options')
 	valaopts.add_option('--vala-target-glib',default=None,dest='vala_target_glib',metavar='MAJOR.MINOR',help='Target version of glib for Vala GObject code generation')
-
-extension('.vala','.gs')(vala_file)
-conf(find_valac)
-conf(check_vala)
-conf(check_vala_deps)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/waf_unit_test.py glmark2-2021.02/waflib/Tools/waf_unit_test.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/waf_unit_test.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/waf_unit_test.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,63 +1,99 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys
-from waflib.TaskGen import feature,after_method
+import os
+from waflib.TaskGen import feature,after_method,taskgen_method
 from waflib import Utils,Task,Logs,Options
+from waflib.Tools import ccroot
 testlock=Utils.threading.Lock()
+@feature('test')
+@after_method('apply_link','process_use')
 def make_test(self):
-	if getattr(self,'link_task',None):
-		self.create_task('utest',self.link_task.outputs)
+	if not getattr(self,'link_task',None):
+		return
+	tsk=self.create_task('utest',self.link_task.outputs)
+	if getattr(self,'ut_str',None):
+		self.ut_run,lst=Task.compile_fun(self.ut_str,shell=getattr(self,'ut_shell',False))
+		tsk.vars=lst+tsk.vars
+	if getattr(self,'ut_cwd',None):
+		if isinstance(self.ut_cwd,str):
+			if os.path.isabs(self.ut_cwd):
+				self.ut_cwd=self.bld.root.make_node(self.ut_cwd)
+			else:
+				self.ut_cwd=self.path.make_node(self.ut_cwd)
+	else:
+		self.ut_cwd=tsk.inputs[0].parent
+	if not hasattr(self,'ut_paths'):
+		paths=[]
+		for x in self.tmp_use_sorted:
+			try:
+				y=self.bld.get_tgen_by_name(x).link_task
+			except AttributeError:
+				pass
+			else:
+				if not isinstance(y,ccroot.stlink_task):
+					paths.append(y.outputs[0].parent.abspath())
+		self.ut_paths=os.pathsep.join(paths)+os.pathsep
+	if not hasattr(self,'ut_env'):
+		self.ut_env=dct=dict(os.environ)
+		def add_path(var):
+			dct[var]=self.ut_paths+dct.get(var,'')
+		if Utils.is_win32:
+			add_path('PATH')
+		elif Utils.unversioned_sys_platform()=='darwin':
+			add_path('DYLD_LIBRARY_PATH')
+			add_path('LD_LIBRARY_PATH')
+		else:
+			add_path('LD_LIBRARY_PATH')
+@taskgen_method
+def add_test_results(self,tup):
+	Logs.debug("ut: %r",tup)
+	self.utest_result=tup
+	try:
+		self.bld.utest_results.append(tup)
+	except AttributeError:
+		self.bld.utest_results=[tup]
 class utest(Task.Task):
 	color='PINK'
 	after=['vnum','inst']
 	vars=[]
 	def runnable_status(self):
+		if getattr(Options.options,'no_tests',False):
+			return Task.SKIP_ME
 		ret=super(utest,self).runnable_status()
 		if ret==Task.SKIP_ME:
 			if getattr(Options.options,'all_tests',False):
 				return Task.RUN_ME
 		return ret
+	def get_test_env(self):
+		return self.generator.ut_env
+	def post_run(self):
+		super(utest,self).post_run()
+		if getattr(Options.options,'clear_failed_tests',False)and self.waf_unit_test_results[1]:
+			self.generator.bld.task_sigs[self.uid()]=None
 	def run(self):
-		filename=self.inputs[0].abspath()
-		self.ut_exec=getattr(self,'ut_exec',[filename])
+		if hasattr(self.generator,'ut_run'):
+			return self.generator.ut_run(self)
+		self.ut_exec=getattr(self.generator,'ut_exec',[self.inputs[0].abspath()])
 		if getattr(self.generator,'ut_fun',None):
 			self.generator.ut_fun(self)
-		try:
-			fu=getattr(self.generator.bld,'all_test_paths')
-		except AttributeError:
-			fu=os.environ.copy()
-			self.generator.bld.all_test_paths=fu
-			lst=[]
-			for g in self.generator.bld.groups:
-				for tg in g:
-					if getattr(tg,'link_task',None):
-						lst.append(tg.link_task.outputs[0].parent.abspath())
-			def add_path(dct,path,var):
-				dct[var]=os.pathsep.join(Utils.to_list(path)+[os.environ.get(var,'')])
-			if Utils.is_win32:
-				add_path(fu,lst,'PATH')
-			elif Utils.unversioned_sys_platform()=='darwin':
-				add_path(fu,lst,'DYLD_LIBRARY_PATH')
-				add_path(fu,lst,'LD_LIBRARY_PATH')
-			else:
-				add_path(fu,lst,'LD_LIBRARY_PATH')
-		cwd=getattr(self.generator,'ut_cwd','')or self.inputs[0].parent.abspath()
-		proc=Utils.subprocess.Popen(self.ut_exec,cwd=cwd,env=fu,stderr=Utils.subprocess.PIPE,stdout=Utils.subprocess.PIPE)
+		testcmd=getattr(self.generator,'ut_cmd',False)or getattr(Options.options,'testcmd',False)
+		if testcmd:
+			self.ut_exec=(testcmd%' '.join(self.ut_exec)).split(' ')
+		return self.exec_command(self.ut_exec)
+	def exec_command(self,cmd,**kw):
+		Logs.debug('runner: %r',cmd)
+		proc=Utils.subprocess.Popen(cmd,cwd=self.get_cwd().abspath(),env=self.get_test_env(),stderr=Utils.subprocess.PIPE,stdout=Utils.subprocess.PIPE)
 		(stdout,stderr)=proc.communicate()
-		tup=(filename,proc.returncode,stdout,stderr)
-		self.generator.utest_result=tup
+		self.waf_unit_test_results=tup=(self.inputs[0].abspath(),proc.returncode,stdout,stderr)
 		testlock.acquire()
 		try:
-			bld=self.generator.bld
-			Logs.debug("ut: %r",tup)
-			try:
-				bld.utest_results.append(tup)
-			except AttributeError:
-				bld.utest_results=[tup]
+			return self.generator.add_test_results(tup)
 		finally:
 			testlock.release()
+	def get_cwd(self):
+		return self.generator.ut_cwd
 def summary(bld):
 	lst=getattr(bld,'utest_results',[])
 	if lst:
@@ -72,8 +108,18 @@
 		for(f,code,out,err)in lst:
 			if code:
 				Logs.pprint('CYAN','    %s'%f)
+def set_exit_code(bld):
+	lst=getattr(bld,'utest_results',[])
+	for(f,code,out,err)in lst:
+		if code:
+			msg=[]
+			if out:
+				msg.append('stdout:%s%s'%(os.linesep,out.decode('utf-8')))
+			if err:
+				msg.append('stderr:%s%s'%(os.linesep,err.decode('utf-8')))
+			bld.fatal(os.linesep.join(msg))
 def options(opt):
+	opt.add_option('--notests',action='store_true',default=False,help='Exec no unit tests',dest='no_tests')
 	opt.add_option('--alltests',action='store_true',default=False,help='Exec all unit tests',dest='all_tests')
-
-feature('test')(make_test)
-after_method('apply_link')(make_test)
\ No newline at end of file
+	opt.add_option('--clear-failed',action='store_true',default=False,help='Force failed unit tests to run again next time',dest='clear_failed_tests')
+	opt.add_option('--testcmd',action='store',default=False,help='Run the unit tests using the test-cmd string'' example "--test-cmd="valgrind --error-exitcode=1'' %s" to run under valgrind',dest='testcmd')
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/winres.py glmark2-2021.02/waflib/Tools/winres.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/winres.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/winres.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,34 +1,53 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
+import re
 from waflib import Task
 from waflib.TaskGen import extension
+from waflib.Tools import c_preproc
+@extension('.rc')
 def rc_file(self,node):
 	obj_ext='.rc.o'
-	if self.env['WINRC_TGT_F']=='/fo':
+	if self.env.WINRC_TGT_F=='/fo':
 		obj_ext='.res'
 	rctask=self.create_task('winrc',node,node.change_ext(obj_ext))
 	try:
 		self.compiled_tasks.append(rctask)
 	except AttributeError:
 		self.compiled_tasks=[rctask]
+re_lines=re.compile('(?:^[ \t]*(#|%:)[ \t]*(ifdef|ifndef|if|else|elif|endif|include|import|define|undef|pragma)[ \t]*(.*?)\s*$)|''(?:^\w+[ \t]*(ICON|BITMAP|CURSOR|HTML|FONT|MESSAGETABLE|TYPELIB|REGISTRY|D3DFX)[ \t]*(.*?)\s*$)',re.IGNORECASE|re.MULTILINE)
+class rc_parser(c_preproc.c_parser):
+	def filter_comments(self,node):
+		code=node.read()
+		if c_preproc.use_trigraphs:
+			for(a,b)in c_preproc.trig_def:code=code.split(a).join(b)
+		code=c_preproc.re_nl.sub('',code)
+		code=c_preproc.re_cpp.sub(c_preproc.repl,code)
+		ret=[]
+		for m in re.finditer(re_lines,code):
+			if m.group(2):
+				ret.append((m.group(2),m.group(3)))
+			else:
+				ret.append(('include',m.group(5)))
+		return ret
 class winrc(Task.Task):
 	run_str='${WINRC} ${WINRCFLAGS} ${CPPPATH_ST:INCPATHS} ${DEFINES_ST:DEFINES} ${WINRC_TGT_F} ${TGT} ${WINRC_SRC_F} ${SRC}'
 	color='BLUE'
+	def scan(self):
+		tmp=rc_parser(self.generator.includes_nodes)
+		tmp.start(self.inputs[0],self.env)
+		return(tmp.nodes,tmp.names)
 def configure(conf):
 	v=conf.env
-	v['WINRC_TGT_F']='-o'
-	v['WINRC_SRC_F']='-i'
-	if not conf.env.WINRC:
+	v.WINRC_TGT_F='-o'
+	v.WINRC_SRC_F='-i'
+	if not v.WINRC:
 		if v.CC_NAME=='msvc':
-			conf.find_program('RC',var='WINRC',path_list=v['PATH'])
-			v['WINRC_TGT_F']='/fo'
-			v['WINRC_SRC_F']=''
+			conf.find_program('RC',var='WINRC',path_list=v.PATH)
+			v.WINRC_TGT_F='/fo'
+			v.WINRC_SRC_F=''
 		else:
-			conf.find_program('windres',var='WINRC',path_list=v['PATH'])
-	if not conf.env.WINRC:
+			conf.find_program('windres',var='WINRC',path_list=v.PATH)
+	if not v.WINRC:
 		conf.fatal('winrc was not found!')
-	v['WINRCFLAGS']=[]
-
-extension('.rc')(rc_file)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/xlc.py glmark2-2021.02/waflib/Tools/xlc.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/xlc.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/xlc.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,39 +1,40 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_xlc(conf):
 	cc=conf.find_program(['xlc_r','xlc'],var='CC')
-	cc=conf.cmd_to_list(cc)
 	conf.get_xlc_version(cc)
 	conf.env.CC_NAME='xlc'
-	conf.env.CC=cc
+@conf
 def xlc_common_flags(conf):
 	v=conf.env
-	v['CC_SRC_F']=[]
-	v['CC_TGT_F']=['-c','-o']
-	if not v['LINK_CC']:v['LINK_CC']=v['CC']
-	v['CCLNK_SRC_F']=[]
-	v['CCLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['RPATH_ST']='-Wl,-rpath,%s'
-	v['SONAME_ST']=[]
-	v['SHLIB_MARKER']=[]
-	v['STLIB_MARKER']=[]
-	v['LINKFLAGS_cprogram']=['-Wl,-brtl']
-	v['cprogram_PATTERN']='%s'
-	v['CFLAGS_cshlib']=['-fPIC']
-	v['LINKFLAGS_cshlib']=['-G','-Wl,-brtl,-bexpfull']
-	v['cshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cstlib']=[]
-	v['cstlib_PATTERN']='lib%s.a'
+	v.CC_SRC_F=[]
+	v.CC_TGT_F=['-c','-o']
+	if not v.LINK_CC:
+		v.LINK_CC=v.CC
+	v.CCLNK_SRC_F=[]
+	v.CCLNK_TGT_F=['-o']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.RPATH_ST='-Wl,-rpath,%s'
+	v.SONAME_ST=[]
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.LINKFLAGS_cprogram=['-Wl,-brtl']
+	v.cprogram_PATTERN='%s'
+	v.CFLAGS_cshlib=['-fPIC']
+	v.LINKFLAGS_cshlib=['-G','-Wl,-brtl,-bexpfull']
+	v.cshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cstlib=[]
+	v.cstlib_PATTERN='lib%s.a'
 def configure(conf):
 	conf.find_xlc()
 	conf.find_ar()
@@ -41,6 +42,3 @@
 	conf.cc_load_tools()
 	conf.cc_add_flags()
 	conf.link_add_flags()
-
-conf(find_xlc)
-conf(xlc_common_flags)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/xlcxx.py glmark2-2021.02/waflib/Tools/xlcxx.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Tools/xlcxx.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Tools/xlcxx.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,39 +1,40 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
 from waflib.Tools import ccroot,ar
 from waflib.Configure import conf
+@conf
 def find_xlcxx(conf):
 	cxx=conf.find_program(['xlc++_r','xlc++'],var='CXX')
-	cxx=conf.cmd_to_list(cxx)
 	conf.get_xlc_version(cxx)
 	conf.env.CXX_NAME='xlc++'
-	conf.env.CXX=cxx
+@conf
 def xlcxx_common_flags(conf):
 	v=conf.env
-	v['CXX_SRC_F']=[]
-	v['CXX_TGT_F']=['-c','-o']
-	if not v['LINK_CXX']:v['LINK_CXX']=v['CXX']
-	v['CXXLNK_SRC_F']=[]
-	v['CXXLNK_TGT_F']=['-o']
-	v['CPPPATH_ST']='-I%s'
-	v['DEFINES_ST']='-D%s'
-	v['LIB_ST']='-l%s'
-	v['LIBPATH_ST']='-L%s'
-	v['STLIB_ST']='-l%s'
-	v['STLIBPATH_ST']='-L%s'
-	v['RPATH_ST']='-Wl,-rpath,%s'
-	v['SONAME_ST']=[]
-	v['SHLIB_MARKER']=[]
-	v['STLIB_MARKER']=[]
-	v['LINKFLAGS_cxxprogram']=['-Wl,-brtl']
-	v['cxxprogram_PATTERN']='%s'
-	v['CXXFLAGS_cxxshlib']=['-fPIC']
-	v['LINKFLAGS_cxxshlib']=['-G','-Wl,-brtl,-bexpfull']
-	v['cxxshlib_PATTERN']='lib%s.so'
-	v['LINKFLAGS_cxxstlib']=[]
-	v['cxxstlib_PATTERN']='lib%s.a'
+	v.CXX_SRC_F=[]
+	v.CXX_TGT_F=['-c','-o']
+	if not v.LINK_CXX:
+		v.LINK_CXX=v.CXX
+	v.CXXLNK_SRC_F=[]
+	v.CXXLNK_TGT_F=['-o']
+	v.CPPPATH_ST='-I%s'
+	v.DEFINES_ST='-D%s'
+	v.LIB_ST='-l%s'
+	v.LIBPATH_ST='-L%s'
+	v.STLIB_ST='-l%s'
+	v.STLIBPATH_ST='-L%s'
+	v.RPATH_ST='-Wl,-rpath,%s'
+	v.SONAME_ST=[]
+	v.SHLIB_MARKER=[]
+	v.STLIB_MARKER=[]
+	v.LINKFLAGS_cxxprogram=['-Wl,-brtl']
+	v.cxxprogram_PATTERN='%s'
+	v.CXXFLAGS_cxxshlib=['-fPIC']
+	v.LINKFLAGS_cxxshlib=['-G','-Wl,-brtl,-bexpfull']
+	v.cxxshlib_PATTERN='lib%s.so'
+	v.LINKFLAGS_cxxstlib=[]
+	v.cxxstlib_PATTERN='lib%s.a'
 def configure(conf):
 	conf.find_xlcxx()
 	conf.find_ar()
@@ -41,6 +42,3 @@
 	conf.cxx_load_tools()
 	conf.cxx_add_flags()
 	conf.link_add_flags()
-
-conf(find_xlcxx)
-conf(xlcxx_common_flags)
\ No newline at end of file
diff -Nru glmark2-2014.03+git20150611.fa71af2d/waflib/Utils.py glmark2-2021.02/waflib/Utils.py
--- glmark2-2014.03+git20150611.fa71af2d/waflib/Utils.py	2015-06-11 22:39:36.000000000 +0800
+++ glmark2-2021.02/waflib/Utils.py	2021-04-25 01:47:22.000000000 +0800
@@ -1,43 +1,40 @@
 #! /usr/bin/env python
 # encoding: utf-8
-# WARNING! Do not edit! http://waf.googlecode.com/git/docs/wafbook/single.html#_obtaining_the_waf_file
+# WARNING! Do not edit! https://waf.io/book/index.html#_obtaining_the_waf_file
 
-import os,sys,errno,traceback,inspect,re,shutil,datetime,gc
+import os,sys,errno,traceback,inspect,re,datetime,platform,base64
 try:
-	import subprocess
-except:
-	try:
-		import waflib.extras.subprocess as subprocess
-	except:
-		print("The subprocess module is missing (python2.3?):\n try calling 'waf update --files=subprocess'\n or add a copy of subprocess.py to the python libraries")
-try:
-	from collections import deque
+	import cPickle
 except ImportError:
-	class deque(list):
-		def popleft(self):
-			return self.pop(0)
+	import pickle as cPickle
+if os.name=='posix'and sys.version_info[0]<3:
+	try:
+		import subprocess32 as subprocess
+	except ImportError:
+		import subprocess
+else:
+	import subprocess
+from collections import deque,defaultdict
 try:
 	import _winreg as winreg
-except:
+except ImportError:
 	try:
 		import winreg
-	except:
+	except ImportError:
 		winreg=None
 from waflib import Errors
 try:
-	from collections import UserDict
-except:
-	from UserDict import UserDict
-try:
 	from hashlib import md5
-except:
+except ImportError:
 	try:
 		from md5 import md5
-	except:
+	except ImportError:
 		pass
 try:
 	import threading
-except:
+except ImportError:
+	if not'JOBS'in os.environ:
+		os.environ['JOBS']='1'
 	class threading(object):
 		pass
 	class Lock(object):
@@ -46,59 +43,182 @@
 		def release(self):
 			pass
 	threading.Lock=threading.Thread=Lock
-else:
-	run_old=threading.Thread.run
-	def run(*args,**kwargs):
-		try:
-			run_old(*args,**kwargs)
-		except(KeyboardInterrupt,SystemExit):
-			raise
-		except:
-			sys.excepthook(*sys.exc_info())
-	threading.Thread.run=run
-SIG_NIL='iluvcuteoverload'
+SIG_NIL='SIG_NIL_SIG_NIL_'.encode()
 O644=420
 O755=493
 rot_chr=['\\','|','/','-']
 rot_idx=0
-try:
-	from collections import defaultdict
-except ImportError:
-	class defaultdict(dict):
-		def __init__(self,default_factory):
-			super(defaultdict,self).__init__()
-			self.default_factory=default_factory
-		def __getitem__(self,key):
-			try:
-				return super(defaultdict,self).__getitem__(key)
-			except KeyError:
-				value=self.default_factory()
-				self[key]=value
-				return value
-is_win32=sys.platform in('win32','cli')
-indicator='\x1b[K%s%s%s\r'
-if is_win32 and'NOCOLOR'in os.environ:
-	indicator='%s%s%s\r'
-def readf(fname,m='r'):
+class ordered_iter_dict(dict):
+	def __init__(self,*k,**kw):
+		self.lst=deque()
+		dict.__init__(self,*k,**kw)
+	def clear(self):
+		dict.clear(self)
+		self.lst=deque()
+	def __setitem__(self,key,value):
+		if key in dict.keys(self):
+			self.lst.remove(key)
+		dict.__setitem__(self,key,value)
+		self.lst.append(key)
+	def __delitem__(self,key):
+		dict.__delitem__(self,key)
+		try:
+			self.lst.remove(key)
+		except ValueError:
+			pass
+	def __iter__(self):
+		return reversed(self.lst)
+	def keys(self):
+		return reversed(self.lst)
+class lru_node(object):
+	__slots__=('next','prev','key','val')
+	def __init__(self):
+		self.next=self
+		self.prev=self
+		self.key=None
+		self.val=None
+class lru_cache(object):
+	__slots__=('maxlen','table','head')
+	def __init__(self,maxlen=100):
+		self.maxlen=maxlen
+		self.table={}
+		self.head=lru_node()
+		for x in range(maxlen-1):
+			node=lru_node()
+			node.prev=self.head.prev
+			node.next=self.head
+			node.prev.next=node
+			node.next.prev=node
+	def __getitem__(self,key):
+		node=self.table[key]
+		if node is self.head:
+			return node.val
+		node.prev.next=node.next
+		node.next.prev=node.prev
+		node.next=self.head.next
+		node.prev=self.head
+		node.next.prev=node
+		node.prev.next=node
+		self.head=node
+		return node.val
+	def __setitem__(self,key,val):
+		node=self.head=self.head.next
+		try:
+			del self.table[node.key]
+		except KeyError:
+			pass
+		node.key=key
+		node.val=val
+		self.table[key]=node
+is_win32=os.sep=='\\'or sys.platform=='win32'
+def readf(fname,m='r',encoding='ISO8859-1'):
+	if sys.hexversion>0x3000000 and not'b'in m:
+		m+='b'
+		f=open(fname,m)
+		try:
+			txt=f.read()
+		finally:
+			f.close()
+		if encoding:
+			txt=txt.decode(encoding)
+		else:
+			txt=txt.decode()
+	else:
+		f=open(fname,m)
+		try:
+			txt=f.read()
+		finally:
+			f.close()
+	return txt
+def writef(fname,data,m='w',encoding='ISO8859-1'):
+	if sys.hexversion>0x3000000 and not'b'in m:
+		data=data.encode(encoding)
+		m+='b'
 	f=open(fname,m)
 	try:
-		txt=f.read()
+		f.write(data)
+	finally:
+		f.close()
+def h_file(fname):
+	f=open(fname,'rb')
+	m=md5()
+	try:
+		while fname:
+			fname=f.read(200000)
+			m.update(fname)
 	finally:
 		f.close()
+	return m.digest()
+def readf_win32(f,m='r',encoding='ISO8859-1'):
+	flags=os.O_NOINHERIT|os.O_RDONLY
+	if'b'in m:
+		flags|=os.O_BINARY
+	if'+'in m:
+		flags|=os.O_RDWR
+	try:
+		fd=os.open(f,flags)
+	except OSError:
+		raise IOError('Cannot read from %r'%f)
+	if sys.hexversion>0x3000000 and not'b'in m:
+		m+='b'
+		f=os.fdopen(fd,m)
+		try:
+			txt=f.read()
+		finally:
+			f.close()
+		if encoding:
+			txt=txt.decode(encoding)
+		else:
+			txt=txt.decode()
+	else:
+		f=os.fdopen(fd,m)
+		try:
+			txt=f.read()
+		finally:
+			f.close()
 	return txt
-def h_file(filename):
-	f=open(filename,'rb')
+def writef_win32(f,data,m='w',encoding='ISO8859-1'):
+	if sys.hexversion>0x3000000 and not'b'in m:
+		data=data.encode(encoding)
+		m+='b'
+	flags=os.O_CREAT|os.O_TRUNC|os.O_WRONLY|os.O_NOINHERIT
+	if'b'in m:
+		flags|=os.O_BINARY
+	if'+'in m:
+		flags|=os.O_RDWR
+	try:
+		fd=os.open(f,flags)
+	except OSError:
+		raise OSError('Cannot write to %r'%f)
+	f=os.fdopen(fd,m)
+	try:
+		f.write(data)
+	finally:
+		f.close()
+def h_file_win32(fname):
+	try:
+		fd=os.open(fname,os.O_BINARY|os.O_RDONLY|os.O_NOINHERIT)
+	except OSError:
+		raise OSError('Cannot read from %r'%fname)
+	f=os.fdopen(fd,'rb')
 	m=md5()
 	try:
-		while filename:
-			filename=f.read(100000)
-			m.update(filename)
+		while fname:
+			fname=f.read(200000)
+			m.update(fname)
 	finally:
 		f.close()
 	return m.digest()
+readf_unix=readf
+writef_unix=writef
+h_file_unix=h_file
+if hasattr(os,'O_NOINHERIT')and sys.hexversion<0x3040000:
+	readf=readf_win32
+	writef=writef_win32
+	h_file=h_file_win32
 try:
 	x=''.encode('hex')
-except:
+except LookupError:
 	import binascii
 	def to_hex(s):
 		ret=binascii.hexlify(s)
@@ -114,27 +234,27 @@
 :param s: string to convert
 :type s: string
 """
+def listdir_win32(s):
+	if not s:
+		try:
+			import ctypes
+		except ImportError:
+			return[x+':\\'for x in'ABCDEFGHIJKLMNOPQRSTUVWXYZ']
+		else:
+			dlen=4
+			maxdrives=26
+			buf=ctypes.create_string_buffer(maxdrives*dlen)
+			ndrives=ctypes.windll.kernel32.GetLogicalDriveStringsA(maxdrives*dlen,ctypes.byref(buf))
+			return[str(buf.raw[4*i:4*i+2].decode('ascii'))for i in range(int(ndrives/dlen))]
+	if len(s)==2 and s[1]==":":
+		s+=os.sep
+	if not os.path.isdir(s):
+		e=OSError('%s is not a directory'%s)
+		e.errno=errno.ENOENT
+		raise e
+	return os.listdir(s)
 listdir=os.listdir
 if is_win32:
-	def listdir_win32(s):
-		if not s:
-			try:
-				import ctypes
-			except:
-				return[x+':\\'for x in list('ABCDEFGHIJKLMNOPQRSTUVWXYZ')]
-			else:
-				dlen=4
-				maxdrives=26
-				buf=ctypes.create_string_buffer(maxdrives*dlen)
-				ndrives=ctypes.windll.kernel32.GetLogicalDriveStringsA(maxdrives,ctypes.byref(buf))
-				return[buf.raw[4*i:4*i+3].decode('ascii')for i in range(int(ndrives/dlen))]
-		if len(s)==2 and s[1]==":":
-			s+=os.sep
-		if not os.path.isdir(s):
-			e=OSError()
-			e.errno=errno.ENOENT
-			raise e
-		return os.listdir(s)
 	listdir=listdir_win32
 def num2ver(ver):
 	if isinstance(ver,str):
@@ -147,26 +267,13 @@
 		return ret
 	return ver
 def ex_stack():
-	exc_type,exc_value,tb=sys.exc_info()
-	exc_lines=traceback.format_exception(exc_type,exc_value,tb)
-	return''.join(exc_lines)
-def to_list(sth):
-	if isinstance(sth,str):
-		return sth.split()
+	return traceback.format_exc()
+def to_list(val):
+	if isinstance(val,str):
+		return val.split()
 	else:
-		return sth
-re_nl=re.compile('\r*\n',re.M)
-def str_to_dict(txt):
-	tbl={}
-	lines=re_nl.split(txt)
-	for x in lines:
-		x=x.strip()
-		if not x or x.startswith('#')or x.find('=')<0:
-			continue
-		tmp=x.split('=')
-		tbl[tmp[0].strip()]='='.join(tmp[1:]).strip()
-	return tbl
-def split_path(path):
+		return val
+def split_path_unix(path):
 	return path.split('/')
 def split_path_cygwin(path):
 	if path.startswith('//'):
@@ -174,56 +281,95 @@
 		ret[0]='/'+ret[0]
 		return ret
 	return path.split('/')
-re_sp=re.compile('[/\\\\]')
+re_sp=re.compile('[/\\\\]+')
 def split_path_win32(path):
 	if path.startswith('\\\\'):
-		ret=re.split(re_sp,path)[2:]
+		ret=re_sp.split(path)[2:]
 		ret[0]='\\'+ret[0]
 		return ret
-	return re.split(re_sp,path)
+	return re_sp.split(path)
+msysroot=None
+def split_path_msys(path):
+	if path.startswith(('/','\\'))and not path.startswith(('\\','\\\\')):
+		global msysroot
+		if not msysroot:
+			msysroot=subprocess.check_output(['cygpath','-w','/']).decode(sys.stdout.encoding or'iso8859-1')
+			msysroot=msysroot.strip()
+		path=os.path.normpath(msysroot+os.sep+path)
+	return split_path_win32(path)
 if sys.platform=='cygwin':
 	split_path=split_path_cygwin
 elif is_win32:
-	split_path=split_path_win32
+	if os.environ.get('MSYSTEM'):
+		split_path=split_path_msys
+	else:
+		split_path=split_path_win32
+else:
+	split_path=split_path_unix
 split_path.__doc__="""
-Split a path by / or \\. This function is not like os.path.split
+Splits a path by / or \\; do not confuse this function with with ``os.path.split``
 
 :type  path: string
 :param path: path to split
-:return:     list of strings
+:return:     list of string
 """
 def check_dir(path):
 	if not os.path.isdir(path):
 		try:
 			os.makedirs(path)
-		except OSError ,e:
+		except OSError as e:
 			if not os.path.isdir(path):
 				raise Errors.WafError('Cannot create the folder %r'%path,ex=e)
+def check_exe(name,env=None):
+	if not name:
+		raise ValueError('Cannot execute an empty string!')
+	def is_exe(fpath):
+		return os.path.isfile(fpath)and os.access(fpath,os.X_OK)
+	fpath,fname=os.path.split(name)
+	if fpath and is_exe(name):
+		return os.path.abspath(name)
+	else:
+		env=env or os.environ
+		for path in env['PATH'].split(os.pathsep):
+			path=path.strip('"')
+			exe_file=os.path.join(path,name)
+			if is_exe(exe_file):
+				return os.path.abspath(exe_file)
+	return None
 def def_attrs(cls,**kw):
 	for k,v in kw.items():
 		if not hasattr(cls,k):
 			setattr(cls,k,v)
 def quote_define_name(s):
-	fu=re.compile("[^a-zA-Z0-9]").sub("_",s)
+	fu=re.sub('[^a-zA-Z0-9]','_',s)
+	fu=re.sub('_+','_',fu)
 	fu=fu.upper()
 	return fu
 def h_list(lst):
-	m=md5()
-	m.update(str(lst))
-	return m.digest()
+	return md5(repr(lst).encode()).digest()
 def h_fun(fun):
 	try:
 		return fun.code
 	except AttributeError:
 		try:
 			h=inspect.getsource(fun)
-		except IOError:
-			h="nocode"
+		except EnvironmentError:
+			h='nocode'
 		try:
 			fun.code=h
 		except AttributeError:
 			pass
 		return h
+def h_cmd(ins):
+	if isinstance(ins,str):
+		ret=ins
+	elif isinstance(ins,list)or isinstance(ins,tuple):
+		ret=str([h_cmd(x)for x in ins])
+	else:
+		ret=str(h_fun(ins))
+	if sys.hexversion>0x3000000:
+		ret=ret.encode('iso8859-1','xmlcharrefreplace')
+	return ret
 reg_subst=re.compile(r"(\\\\)|(\$\$)|\$\{([^}]+)\}")
 def subst_vars(expr,params):
 	def repl_var(m):
@@ -244,7 +390,7 @@
 	return'elf'
 def unversioned_sys_platform():
 	s=sys.platform
-	if s=='java':
+	if s.startswith('java'):
 		from java.lang import System
 		s=System.getProperty('os.name')
 		if s=='Mac OS X':
@@ -254,13 +400,16 @@
 		elif s=='OS/2':
 			return'os2'
 		elif s=='HP-UX':
-			return'hpux'
+			return'hp-ux'
 		elif s in('SunOS','Solaris'):
 			return'sunos'
 		else:s=s.lower()
 	if s=='powerpc':
 		return'darwin'
-	if s=='win32'or s.endswith('os2')and s!='sunos2':return s
+	if s=='win32'or s=='os2':
+		return s
+	if s=='cli'and os.name=='nt':
+		return'win32'
 	return re.split('\d+$',s)[0]
 def nada(*k,**kw):
 	pass
@@ -269,10 +418,10 @@
 		self.start_time=datetime.datetime.utcnow()
 	def __str__(self):
 		delta=datetime.datetime.utcnow()-self.start_time
-		days=int(delta.days)
-		hours=delta.seconds//3600
-		minutes=(delta.seconds-hours*3600)//60
-		seconds=delta.seconds-hours*3600-minutes*60+float(delta.microseconds)/1000/1000
+		days=delta.days
+		hours,rem=divmod(delta.seconds,3600)
+		minutes,seconds=divmod(rem,60)
+		seconds+=delta.microseconds*1e-6
 		result=''
 		if days:
 			result+='%dd'%days
@@ -281,18 +430,6 @@
 		if days or hours or minutes:
 			result+='%dm'%minutes
 		return'%s%.3fs'%(result,seconds)
-if is_win32:
-	old=shutil.copy2
-	def copy2(src,dst):
-		old(src,dst)
-		shutil.copystat(src,dst)
-	setattr(shutil,'copy2',copy2)
-if os.name=='java':
-	try:
-		gc.disable()
-		gc.enable()
-	except NotImplementedError:
-		gc.disable=gc.enable
 def read_la_file(path):
 	sp=re.compile(r'^([^=]+)=\'(.*)\'$')
 	dc={}
@@ -303,26 +440,17 @@
 		except ValueError:
 			pass
 	return dc
-def nogc(fun):
-	def f(*k,**kw):
-		try:
-			gc.disable()
-			ret=fun(*k,**kw)
-		finally:
-			gc.enable()
-		return ret
-	f.__doc__=fun.__doc__
-	return f
 def run_once(fun):
 	cache={}
-	def wrap(k):
+	def wrap(*k):
 		try:
 			return cache[k]
 		except KeyError:
-			ret=fun(k)
+			ret=fun(*k)
 			cache[k]=ret
 			return ret
 	wrap.__cache__=cache
+	wrap.__name__=fun.__name__
 	return wrap
 def get_registry_app_path(key,filename):
 	if not winreg:
@@ -334,3 +462,74 @@
 	else:
 		if os.path.isfile(result):
 			return result
+def lib64():
+	if os.sep=='/':
+		if platform.architecture()[0]=='64bit':
+			if os.path.exists('/usr/lib64')and not os.path.exists('/usr/lib32'):
+				return'64'
+	return''
+def sane_path(p):
+	return os.path.abspath(os.path.expanduser(p))
+process_pool=[]
+def get_process():
+	try:
+		return process_pool.pop()
+	except IndexError:
+		filepath=os.path.dirname(os.path.abspath(__file__))+os.sep+'processor.py'
+		cmd=[sys.executable,'-c',readf(filepath)]
+		return subprocess.Popen(cmd,stdout=subprocess.PIPE,stdin=subprocess.PIPE,bufsize=0)
+def run_prefork_process(cmd,kwargs,cargs):
+	if not'env'in kwargs:
+		kwargs['env']=dict(os.environ)
+	try:
+		obj=base64.b64encode(cPickle.dumps([cmd,kwargs,cargs]))
+	except TypeError:
+		return run_regular_process(cmd,kwargs,cargs)
+	proc=get_process()
+	if not proc:
+		return run_regular_process(cmd,kwargs,cargs)
+	proc.stdin.write(obj)
+	proc.stdin.write('\n'.encode())
+	proc.stdin.flush()
+	obj=proc.stdout.readline()
+	if not obj:
+		raise OSError('Preforked sub-process %r died'%proc.pid)
+	process_pool.append(proc)
+	ret,out,err,ex,trace=cPickle.loads(base64.b64decode(obj))
+	if ex:
+		if ex=='OSError':
+			raise OSError(trace)
+		elif ex=='ValueError':
+			raise ValueError(trace)
+		else:
+			raise Exception(trace)
+	return ret,out,err
+def run_regular_process(cmd,kwargs,cargs={}):
+	proc=subprocess.Popen(cmd,**kwargs)
+	if kwargs.get('stdout')or kwargs.get('stderr'):
+		out,err=proc.communicate(**cargs)
+		status=proc.returncode
+	else:
+		out,err=(None,None)
+		status=proc.wait(**cargs)
+	return status,out,err
+def run_process(cmd,kwargs,cargs={}):
+	if kwargs.get('stdout')and kwargs.get('stderr'):
+		return run_prefork_process(cmd,kwargs,cargs)
+	else:
+		return run_regular_process(cmd,kwargs,cargs)
+def alloc_process_pool(n,force=False):
+	global run_process,get_process,alloc_process_pool
+	if not force:
+		n=max(n-len(process_pool),0)
+	try:
+		lst=[get_process()for x in range(n)]
+	except OSError:
+		run_process=run_regular_process
+		get_process=alloc_process_pool=nada
+	else:
+		for x in lst:
+			process_pool.append(x)
+if sys.platform=='cli'or not sys.executable:
+	run_process=run_regular_process
+	get_process=alloc_process_pool=nada
diff -Nru glmark2-2014.03+git20150611.fa71af2d/wscript glmark2-2021.02/wscript
--- glmark2-2014.03+git20150611.fa71af2d/wscript	2015-06-11 22:41:04.000000000 +0800
+++ glmark2-2021.02/wscript	2021-04-25 01:47:22.000000000 +0800
@@ -1,27 +1,32 @@
-import commands
-import subprocess
-import os
-import Options
-import Scripting
+import os, platform
 from waflib import Context
 
 out = 'build'
 top = '.'
 
-VERSION = '2014.03+git20150611.fa71af2d'
+VERSION = '2021.02'
 APPNAME = 'glmark2'
 
 FLAVORS = {
-    'x11-gl' : 'glmark2',
-    'x11-glesv2' : 'glmark2-es2',
+    'dispmanx-glesv2' : 'glmark2-es2-dispmanx',
     'drm-gl' : 'glmark2-drm',
     'drm-glesv2' : 'glmark2-es2-drm',
     'mir-gl' : 'glmark2-mir',
     'mir-glesv2' : 'glmark2-es2-mir',
     'wayland-gl' : 'glmark2-wayland',
-    'wayland-glesv2' : 'glmark2-es2-wayland'
+    'wayland-glesv2' : 'glmark2-es2-wayland',
+    'win32-gl': 'glmark2-win32',
+    'win32-glesv2': 'glmark2-es2',
+    'x11-gl' : 'glmark2',
+    'x11-glesv2' : 'glmark2-es2',
 }
-FLAVORS_STR = ", ".join(FLAVORS.keys())
+FLAVORS_STR = ", ".join(sorted(list(FLAVORS) + ['all-linux', 'all-win32']))
+
+def linux_flavors():
+    return [f for f in FLAVORS.keys() if not f.startswith('win32')]
+
+def win32_flavors():
+    return [f for f in FLAVORS.keys() if f.startswith('win32')]
 
 def option_list_cb(option, opt, value, parser):
     value = value.split(',')
@@ -36,68 +41,146 @@
     return False
 
 def options(opt):
-    opt.tool_options('gnu_dirs')
-    opt.tool_options('compiler_cc')
-    opt.tool_options('compiler_cxx')
+    opt.load('gnu_dirs')
+    opt.load('compiler_c')
+    opt.load('compiler_cxx')
 
     opt.add_option('--with-flavors', type = 'string', action='callback',
                    callback=option_list_cb,
                    dest = 'flavors',
-                   help = "a list of flavors to build (%s, all)" % FLAVORS_STR)
+                   help = "a list of flavors to build (%s, all-linux (except dispmanx-glesv2), all-win32)" % FLAVORS_STR)
     opt.parser.set_default('flavors', [])
 
+    opt.parser.add_option('--version-suffix', type='string', action='store', dest='versionsuffix',
+                          default='', help='add a suffix to the version number')
+
     opt.add_option('--no-debug', action='store_false', dest = 'debug',
                    default = True, help='disable compiler debug information')
     opt.add_option('--no-opt', action='store_false', dest = 'opt',
                    default = True, help='disable compiler optimizations')
-    opt.add_option('--no-werror', action='store_false', dest = 'werror',
-                   default = True, help='disable treating compiler warnings as errors')
     opt.add_option('--data-path', action='store', dest = 'data_path',
                    help='path to main data (also see --data(root)dir)')
     opt.add_option('--extras-path', action='store', dest = 'extras_path',
                    help='path to additional data (models, shaders, textures)')
 
+def get_data_path(ctx):
+    if ctx.options.data_path is not None:
+        return ctx.options.data_path
+    else:
+        return os.path.join(ctx.env.DATADIR, 'glmark2')
+
 def configure(ctx):
     # Special 'all' flavor
-    if 'all' in Options.options.flavors:
-        Options.options.flavors = list(set(Options.options.flavors) | set(FLAVORS.keys()))
-        Options.options.flavors.remove('all')
+    if 'all-linux' in ctx.options.flavors:
+        ctx.options.flavors = list(set(ctx.options.flavors) | set(linux_flavors()))
+        ctx.options.flavors.remove('all-linux')
+        # dispmanx is a special case, we don't want to include it in all
+        ctx.options.flavors.remove('dispmanx-glesv2')
+
+    if 'all-win32' in ctx.options.flavors:
+        ctx.options.flavors = list(set(ctx.options.flavors) | set(win32_flavors()))
+        ctx.options.flavors.remove('all-win32')
+
+    is_win = any(True for f in win32_flavors() if f in ctx.options.flavors)
+    is_linux = any(True for f in linux_flavors() if f in ctx.options.flavors)
+
+    if is_win and is_linux:
+        ctx.fatal('Simultaneous Windows and Linux builds are not supported')
 
     # Ensure the flavors are valid
-    for flavor in Options.options.flavors:
+    for flavor in ctx.options.flavors:
        if flavor not in FLAVORS:
             ctx.fatal('Unknown flavor: %s. Supported flavors are %s' % (flavor, FLAVORS_STR))
 
-    if not Options.options.flavors:
-        ctx.fatal('You need to select at least one flavor. Supported flavors are %s' % FLAVORS_STR)
+    if not ctx.options.flavors:
+        ctx.fatal('You need to select at least one flavor with --with-flavors=flavor1[,flavor2]...\n' +
+                  'Supported flavors are %s' % FLAVORS_STR)
 
     for flavor in FLAVORS:
-        if flavor in Options.options.flavors:
+        if flavor in ctx.options.flavors:
             ctx.env["FLAVOR_%s" % flavor.upper().replace('-','_')] = FLAVORS[flavor]
 
-    ctx.check_tool('gnu_dirs')
-    ctx.check_tool('compiler_cc')
-    ctx.check_tool('compiler_cxx')
+    ctx.load('gnu_dirs')
+    ctx.load('compiler_c')
+    ctx.load('compiler_cxx')
+
+    if is_win:
+        configure_win32(ctx)
+    else:
+        configure_linux(ctx)
+
+    ctx.env.append_unique('DEFINES', 'GLMARK_VERSION="%s"' % (VERSION + ctx.options.versionsuffix))
+    ctx.env.GLMARK2_VERSION = (VERSION + ctx.options.versionsuffix)
+
+    ctx.msg("Prefix", ctx.env.PREFIX, color = 'PINK')
+    ctx.msg("Data path", get_data_path(ctx), color = 'PINK')
+    ctx.msg("Including extras", "Yes" if ctx.env.HAVE_EXTRAS else "No",
+            color = 'PINK');
+    if ctx.env.HAVE_EXTRAS:
+        ctx.msg("Extras path", ctx.options.extras_path, color = 'PINK')
+    ctx.msg("Building flavors", ctx.options.flavors)
+
+
+def configure_win32(ctx):
+    # Check required headers
+    req_headers = ['stdlib.h', 'string.h', 'stdint.h', 'stdio.h', 'windows.h']
+    for header in req_headers:
+        ctx.check_cc(header_name = header, auto_add_header_name = True, mandatory = True)
+
+    req_libs = [('user32', 'user32'), ('opengl32', 'opengl32'), ('gdi32', 'gdi32')]
+    for (lib, uselib) in req_libs:
+        ctx.check_cc(lib = lib, uselib_store = uselib)
+
+    # Prepend CXX flags so that they can be overriden by the
+    # CXXFLAGS environment variable
+    if ctx.options.opt:
+        ctx.env.prepend_value('CXXFLAGS', '-O2')
+    if ctx.env.CXX_NAME != 'msvc':
+        if ctx.options.debug:
+            ctx.env.prepend_value('CXXFLAGS', '-g')
+        ctx.env.prepend_value('CXXFLAGS', '-std=c++14 -Wall -Wextra -Wnon-virtual-dtor'.split(' '))
+    else:
+        ctx.env.prepend_value('CXXFLAGS', '/EHsc /wd4312'.split(' '))
+
+    ctx.env.HAVE_EXTRAS = False
+    if ctx.options.extras_path is not None:
+        ctx.env.HAVE_EXTRAS = True
+        ctx.env.append_unique('GLMARK_EXTRAS_PATH', ctx.options.extras_path)
+        ctx.env.append_unique('DEFINES', 'GLMARK_EXTRAS_PATH="%s"' % ctx.options.extras_path)
+
+    data_path = get_data_path(ctx)
+
+    # Necessary for M_PI
+    ctx.env.append_unique('DEFINES', '_USE_MATH_DEFINES')
+    ctx.env.append_unique('DEFINES', 'WIN32')
+    # String contants have issues with Windows slashes
+    ctx.env.append_unique('DEFINES', 'GLMARK_DATA_PATH="%s"' % data_path.replace('\\', '/'))
+    ctx.env.append_unique('GLMARK_DATA_PATH', data_path)
 
+
+def configure_linux(ctx):
     # Check required headers
-    req_headers = ['stdlib.h', 'string.h', 'unistd.h', 'stdint.h', 'stdio.h', 'jpeglib.h']
+    req_headers = ['stdlib.h', 'string.h', 'stdint.h', 'stdio.h', 'dlfcn.h',
+                   'unistd.h', 'jpeglib.h', 'math.h', 'string.h']
     for header in req_headers:
-        ctx.check_cxx(header_name = header, auto_add_header_name = True, mandatory = True)
+        ctx.check_cc(header_name = header, auto_add_header_name = True, mandatory = True)
 
     # Check for required libs
     req_libs = [('m', 'm'), ('jpeg', 'jpeg')]
     for (lib, uselib) in req_libs:
-        ctx.check_cxx(lib = lib, uselib_store = uselib)
+        ctx.check_cc(lib = lib, uselib_store = uselib)
 
-    # Check required functions
-    req_funcs = [('memset', 'string.h', []) ,('sqrt', 'math.h', ['m'])]
-    for func, header, uselib in req_funcs:
-        ctx.check_cxx(function_name = func, header_name = header,
-                      uselib = uselib, mandatory = True)
+    # Check for required functions. Note that headers from req_headers
+    # are already included for this check.
+    ctx.check_cc(function_name = 'memset', mandatory = True)
+    # Special test code for sqrt, to make clang/libc++ happy.
+    ctx.check_cc(fragment = 'int main(void) { double (*p)(double); p = sqrt; }',
+                 msg = 'Checking for function sqrt',
+                 uselib = ['m'], mandatory = True)
 
     # Check for a supported version of libpng
-    supp_png_pkgs = (('libpng12', '1.2'), ('libpng15', '1.5'), ('libpng16', '1.6'),)
     have_png = False
+    supp_png_pkgs = (('libpng12', '1.2'), ('libpng15', '1.5'), ('libpng16', '1.6'),)
     for (pkg, atleast) in supp_png_pkgs:
         try:
             pkg_ver = ctx.check_cfg(package=pkg, uselib_store='libpng', atleast_version=atleast,
@@ -111,16 +194,28 @@
     if not have_png:
         ctx.fatal('You need to install a supported version of libpng: ' + str(supp_png_pkgs))
 
+    dispmanx = 'dispmanx-glesv2' in ctx.options.flavors
+    if dispmanx:
+        # dispmanx uses custom Broadcom libraries that don't follow standard
+        # Linux packaging.  Just force the library setup here.
+        if len(ctx.options.flavors) != 1:
+            ctx.fatal("dispmanx can't be built with any other flavor")
+
+        ctx.env.append_value('CXXFLAGS', '-I/opt/vc/include')
+
+        ctx.check_cxx(lib = 'brcmGLESv2', uselib_store = 'glesv2', libpath='/opt/vc/lib')
+        ctx.check_cxx(lib = ['brcmEGL', 'brcmGLESv2'], uselib_store = 'egl', libpath='/opt/vc/lib')
+        ctx.check_cxx(lib = ['bcm_host', 'vcos', 'vchiq_arm'], uselib_store = 'dispmanx', libpath='/opt/vc/lib')
+
     # Check optional packages
-    opt_pkgs = [('x11', 'x11', None, list_contains(Options.options.flavors, 'x11')),
-                ('gl', 'gl', None, list_contains(Options.options.flavors, 'gl$')),
-                ('egl', 'egl', None, list_contains(Options.options.flavors, 'glesv2$')),
-                ('glesv2', 'glesv2', None, list_contains(Options.options.flavors, 'glesv2$')),
-                ('libdrm','drm', None, list_contains(Options.options.flavors, 'drm')),
-                ('gbm','gbm', None, list_contains(Options.options.flavors, 'drm')),
-                ('mirclient','mirclient', '0.13', list_contains(Options.options.flavors, 'mir')),
-                ('wayland-client','wayland-client', None, list_contains(Options.options.flavors, 'wayland')),
-                ('wayland-egl','wayland-egl', None, list_contains(Options.options.flavors, 'wayland'))]
+    opt_pkgs = [('x11', 'x11', None, list_contains(ctx.options.flavors, 'x11')),
+                ('libdrm','drm', None, list_contains(ctx.options.flavors, 'drm')),
+                ('gbm','gbm', None, list_contains(ctx.options.flavors, 'drm')),
+                ('libudev', 'udev', None, list_contains(ctx.options.flavors, 'drm')),
+                ('mirclient','mirclient', '0.13', list_contains(ctx.options.flavors, 'mir')),
+                ('wayland-client','wayland-client', None, list_contains(ctx.options.flavors, 'wayland')),
+                ('wayland-cursor','wayland-cursor', None, list_contains(ctx.options.flavors, 'wayland')),
+                ('wayland-egl','wayland-egl', None, list_contains(ctx.options.flavors, 'wayland'))]
     for (pkg, uselib, atleast, mandatory) in opt_pkgs:
         if atleast is None:
             ctx.check_cfg(package = pkg, uselib_store = uselib,
@@ -129,40 +224,31 @@
             ctx.check_cfg(package = pkg, uselib_store = uselib, atleast_version=atleast,
                           args = '--cflags --libs', mandatory = mandatory)
 
+    if list_contains(ctx.options.flavors, 'wayland'):
+        # wayland-protocols >= 1.12 required for xdg-shell stable
+        ctx.check_cfg(package = 'wayland-protocols', atleast_version = '1.12',
+                      variables = ['pkgdatadir'], uselib_store = 'WAYLAND_PROTOCOLS')
+        ctx.check_cfg(package = 'wayland-scanner', variables = ['wayland_scanner'],
+                      uselib_store = 'WAYLAND_SCANNER')
 
     # Prepend CXX flags so that they can be overriden by the
     # CXXFLAGS environment variable
-    if Options.options.opt:
+    if ctx.options.opt:
         ctx.env.prepend_value('CXXFLAGS', '-O2')
-    if Options.options.debug:
+    if ctx.options.debug:
         ctx.env.prepend_value('CXXFLAGS', '-g')
-    if Options.options.werror:
-        ctx.env.prepend_value('CXXFLAGS', '-Werror')
-    ctx.env.prepend_value('CXXFLAGS', '-Wall -Wextra -Wnon-virtual-dtor'.split(' '))
+    ctx.env.prepend_value('CXXFLAGS', '-std=c++14 -Wall -Wextra -Wnon-virtual-dtor'.split(' '))
 
     ctx.env.HAVE_EXTRAS = False
-    if Options.options.extras_path is not None:
+    if ctx.options.extras_path is not None:
         ctx.env.HAVE_EXTRAS = True
-        ctx.env.append_unique('GLMARK_EXTRAS_PATH', Options.options.extras_path)
-        ctx.env.append_unique('DEFINES', 'GLMARK_EXTRAS_PATH="%s"' % Options.options.extras_path)
+        ctx.env.append_unique('GLMARK_EXTRAS_PATH', ctx.options.extras_path)
+        ctx.env.append_unique('DEFINES', 'GLMARK_EXTRAS_PATH="%s"' % ctx.options.extras_path)
 
-    if Options.options.data_path is not None:
-        data_path = Options.options.data_path 
-    else:
-        data_path = os.path.join(ctx.env.DATADIR, 'glmark2')
+    data_path = get_data_path(ctx)
 
-    ctx.env.append_unique('GLMARK_DATA_PATH', data_path)
     ctx.env.append_unique('DEFINES', 'GLMARK_DATA_PATH="%s"' % data_path)
-    ctx.env.append_unique('DEFINES', 'GLMARK_VERSION="%s"' % VERSION)
-    ctx.env.GLMARK2_VERSION = VERSION
-
-    ctx.msg("Prefix", ctx.env.PREFIX, color = 'PINK')
-    ctx.msg("Data path", data_path, color = 'PINK')
-    ctx.msg("Including extras", "Yes" if ctx.env.HAVE_EXTRAS else "No",
-            color = 'PINK');
-    if ctx.env.HAVE_EXTRAS:
-        ctx.msg("Extras path", Options.options.extras_path, color = 'PINK')
-    ctx.msg("Building flavors", Options.options.flavors)
+    ctx.env.append_unique('GLMARK_DATA_PATH', data_path)
 
 def build(ctx):
     ctx.recurse('src')